(Barresi, Antonello Fissore, Davide Pisano, Robe

DEVELOPMENTS IN
HEAT TRANSFER
Edited by
Marco Aurélio dos Santos Bernardes
Developments in Heat Transfer
Edited by Marco Aurélio dos Santos Bernardes
Published by InTech
Janeza Trdine 9, 51000 Rijeka, Croatia
Copyright © 2011 InTech

All chapters are Open Access articles distributed under the Creative Commons
Non Commercial Share Alike Attribution 3.0 license, which permits to copy,
distribute, transmit, and adapt the work in any medium, so long as the original
work is properly cited. After this work has been published by InTech, authors
have the right to republish it, in whole or part, in any publication of which they
are the author, and to make other personal use of the work. Any republication,
referencing or personal use of the work must explicitly identify the original source.
Statements and opinions expressed in the chapters are these of the individual contributors
and not necessarily those of the editors or publisher. No responsibility is accepted
for the accuracy of information contained in the published articles. The publisher
assumes no responsibility for any damage or injury to persons or property arising out
of the use of any materials, instructions, methods or ideas contained in the book.
Publishing Process Manager Ivana Lorković

Technical Editor Teodora Smiljanic
Cover Designer Jan Hyrat
Image Copyright slava_vn, 2010. Used under license from Shutterstock.com
First published August, 2011

Printed in Croatia
A free online edition of this book is available at www.intechopen.com

Additional hard copies can be obtained from [email protected]
Developments in Heat Transfer, Edited by Marco Aurélio dos Santos Bernardes

p. cm.
ISBN 978-953-307-569-3
Contents
Preface XI
Chapter 1 Thermal Effects in Optical Fibres 1

Paulo André, Ana Rocha, Fátima Domingues and Margarida Facão
Chapter 2 Heat Transfer for NDE: Landmine Detection 21

Fernando Pardo, Paula López and Diego Cabello
Chapter 3 The Heat Transfer Enhancement

Analysis and Experimental Investigation of
Non-Uniform Cross-Section Channel SEMOS Heat Pipe 47
Shang Fu-Min, Liu Jian-Hong and Liu Deng-Ying
Chapter 4 Magneto Hydro-Dynamics and

Heat Transfer in Liquid Metal Flows 55
J. S. Rao and Hari Sankar
Chapter 5 Thermal Anomaly and Strength of Atotsugawa Fault, Central

Japan, Inferred from Fission-Track Thermochronology 81
Ryuji Yamada and Kazuo Mizoguchi
Chapter 6 Heat Transfer in Freeze-Drying Apparatus 91

Roberto Pisano, Davide Fissore and Antonello A. Barresi
Chapter 7 Radiant Floor Heating System 115

Byung-Cheon Ahn
Chapter 8 Variable Property Effects in

Momentum and Heat Transfer 135
Yan Jin and Heinz Herwig
Chapter 9 Bioheat Transfer 153

Alireza Zolfaghari and Mehdi Maerefat
Chapter 10 The Manufacture of Microencapsulated Thermal

Energy Storage Compounds Suitable for Smart Textile 171
Salaün Fabien
VI Contents
Chapter 11 Heat Transfer and Thermal Air Management

in the Electronics and Process Industries 199
Harvey M. Thompson
Chapter 12 Unsteady Mixed Convection Flow in the Stagnation

Region of a Heated Vertical Plate Embedded in a Variable
Porosity Medium with Thermal Dispersion Effects 217
S. M. Alharbi and I. A. Hassanien
Chapter 13 Heat Generation and Transfer on Biological

Tissues Due to High-Intensity Laser Irradiation 227
Denise M. Zezell, Patricia A. Ana,
Thiago M. Pereira, Paulo R. Correa and Walter Velloso Jr.
Chapter 14 Entransy Dissipation Theory

and Its Application in Heat Transfer 247
Mingtian Xu
Chapter 15 Inverse Space Marching Method for

Determining Temperature and Stress
Distributions in Pressure Components 273
Jan Taler, Bohdan Weglowski,
Tomasz Sobota, Magdalena Jaremkiewicz and Dawid Taler
Chapter 16 Experimental Prediction of

Heat Transfer Correlations in Heat Exchangers 293
Tomasz Sobota
Chapter 17 High Temperature Thermal Devices for

Nuclear Process Heat Transfer Applications 309
Piyush Sabharwall and Eung Soo Kim
Chapter 18 Flow Properties and Heat Transfer of

Drag-Reducing Surfactant Solutions 331
Takashi Saeki
Chapter 19 Entransy - a Novel Theory in

Heat Transfer Analysis and Optimization 349
Qun Chen, Xin-Gang Liang and Zeng-Yuan Guo
Chapter 20 Transient Heat Transfer and Energy

Transport in Packed Bed Thermal Storage Systems 373
Pei Wen Li, Jon Van Lew, Wafaa Karaki,
Cho Lik Chan, Jake Stephens and James. E. O’Brien
Chapter 21 Role of Heat Transfer on Process

Characteristics During Electrical Discharge Machining 417
Ahsan Ali Khan
Contents VII
Chapter 22 Thermal Treatment of Granulated

Particles by Induction Thermal Plasma 437
M. Mofazzal Hossain, Takayuki Watanabe
Chapter 23 Method for Measurement of Single-Injector

Heat Transfer Characteristics and Its Application
in Studying Gas-Gas Injector Combustion Chamber 455
Guo-biao Cai, Xiao-wei Wang and Tao Chen
Chapter 24 Heat Transfer Related to

Gas Hydrate Formation/Dissociation 477
Bei Liu, Weixin Pang, Baozi Peng, Changyu Sun and Guangjin Chen
Chapter 25 Progress Works of High and

Super High Temperature Heat Pipes 503
Wei Qu
Chapter 26 Design of the Heat Conduction Structure

Based on the Topology Optimization 523
Yongcun Zhang, Shutian Liu and Heting Qiao
Chapter 27 Thermal Modelling for

Laser Treatment of Port Wine Stains 537
Dong Li, Ya-Ling He and Guo-Xiang Wang
Chapter 28 Study of the Heat Transfer Effect

in Moxibustion Practice 557
Chinlong Huang and Tony W. H. Sheu
Chapter 29 Heat and Mass Transfer in Jet Type Mold Cooling Pipe 573
Hideo Kawahara
Chapter 30 Thermal State and

Human Comfort in Underground Mining 589
Vidal F. Navarro Torres and Raghu N. Singh
Chapter 31 Heat Transfer in the Environment: Development and

Use of Fiber-Optic Distributed Temperature Sensing 611
Francisco Suárez, Mark B. Hausner,
Jeff Dozier, John S. Selker and Scott W. Tyler
Chapter 32 Prandtl Number Effect on

Heat Transfer Degradation in MHD Turbulent
Shear Flows by Means of High-Resolution DNS 637
Yoshinobu Yamamoto and Tomoaki Kunugi
Chapter 33 Effective Method of

Microcapsules Production for Smart Fabrics 649
Luz Sánchez-Silva, Paula Sánchez and Juan F. Rodríguez
VIII Contents
Chapter 34 Heat Conduction in Nonlinear Media 667

Michael M. Tilleman
Preface
Recent 40 years have witnessed considerable advances in experimental, theoretical and

calculation approaches as well as in industrial utilizations. In this way, this book aims
to bring together current progresses in some special heat transfer fields contributing to
the dissemination of high-quality research efforts. Therefore, issues comprising heat
transfer fundamental concepts and modes (specifically conduction, convection and
radiation), bioheat, entransy theory development, micro heat transfer, high
temperature applications, turbulent shear flows, mass transfer, heat pipes, design
optimization, medical therapies, fiber-optics, heat transfer of in surfactant solutions,
landmine detection, heat exchangers, radiant floor, packed bed thermal storage
systems, inverse space marching method, heat transfer in short slot ducts, freezing an
drying mechanisms, variable property effects in heat transfer, heat transfer in
electronics and process industries, fission-track thermochronology, combustion, heat
transfer in liquid metal flows, human comfort in underground mining, heat transfer
on electrical discharge machining, mixing convection are included in this book aiming.
The experimental and theoretical investigations, assessment and enhancement
techniques illustrated here aspire to be useful for many researchers, scientists,
engineers and graduate students.
Marco Aurélio dos Santos Bernardes

CRP Henri Tudor, CRTE
Esch-sur-Alzette
Luxembourg
1
Thermal Effects in Optical Fibres

Paulo André, Ana Rocha, Fátima Domingues and Margarida Facão
Instituto de Telecomunicações and Departamento de Física, Universidade de Aveiro
Portugal
1. Introduction
Optical fibres are essential components in the modern telecommunication scenario. From
the first works dealing with the optimization of optical fibres transmission characteristics to
accommodate long distance data transmission, realized by Charles Kao (Nobel Prize of
Physics in 2009), until the actual optical fibre communication networks, a long way was
paved.
The developments introduced in the optical communication systems have been focused in 3
main objectives: increase of the propagation distance, increase of the transmission capacity
(bitrate) and reduction of the deployment and operation costs. The achievement of these
objectives was only possible due to several technological breakthroughs, such as the
development of optical amplifiers and the introduction of wavelength multiplexing
techniques. However, the consequence of those developments was the increase of the total
optical power propagating along the fibres.
Moreover, in the last years, the evolution of the optical networks has been toward the
objective of deploying the fibre link end directly to the subscribers home (FTTH – fibre to
the home).
Thus, the conjugation of high power propagation and tight bending, resulting from the
actual FTTH infrastructures, is responsible for fibre lifetime reduction, mainly caused by the
local increase of the coating temperature. This effect can lead to the rupture of the fibre or to
the fibre fuse effect ignition with the consequent destruction of the optical fibre along
kilometres.
In this work, we analyze the thermal effects occurring in optical fibres, such as the coating
heating due to high power propagation in bent fibres and the fibre fuse effect. We describe
the actual state of the art of these phenomena and our contribution to the subject, which
consists on both experimental and numerical simulation results.
2. Literature review
The fibre fuse effect, named due to the similarity with a burning fuse, was first observed in
1987 (Kashyap, 1987; Kashyap et al., 1988). At that time, the effect was observed on a single
mode silica fibre illuminated by an optical signal with an average power density higher than
5MW/cm2. Like a burning fuse, after the optical fibre fuse ignition, the fuse zone propagates
towards the light source while a visible white light is emitted. After the fuse zone
propagation, the fibre core shows a string of voids, being permanently damaged. The
phenomenon was always associated with a thermal effect and although there are not yet
2 Developments in Heat Transfer
very accurate experimental data for the actual temperature achieved in the fibre core, it is
believe that the peak temperature is up above the silica vaporization point, around 3300 K.
Some authors also refer that the white light emission characteristic of this effect may indicate
temperatures that would allow plasma like fuse zone (104) (Hand et al., 1988b; Dianov et al.,
2006; Shuto, 2010).
The first explanation for the effect related it to a thermal self-focusing mechanism (Kashyap
et al., 1988). Afterwards, the fuse zone was identified as a soliton-like thermal shock-wave
which would occur by strong thermal dependent absorption due to the creation of
Ge-related defects in Ge doped core fibre. To sustain this hypothesis, the a Ge doped fibre
was heated up to 1000ºC in the absence of any propagating optical signal and the same kind
of periodic damaged pattern was produced (Driscoll et al., 1991), however this result has not
been reproduced by any others research groups.
At the time of these first observations the fuse effect did not represent a practical problem,
since the total power injected in the network optical fibres was well below the power
densities used in the experiments. However, the rise of optical communications demand and
the consequent increase of the injected power have promoted the fuse effect to one of the
fundamental issues which should be considered while developing and maintaining optical
networks. Hence, for several years the phenomenon was referred as the origin of the optical
fibres damage, but only in the presence of high powers. It was only a few years ago that the
scientific community turned to this effect in order to explain it better but also to design
devices able to detect and halt this catastrophic effect.
Nowadays, the most accepted explanation for the fuse effect describes it as an absorption
enhanced temperature rise that propagates toward the light source by thermal conduction
and driven by the optical power itself. The first numerical simulation of the fuse
propagation used an explicit finite-difference method where it was assumed that the
electrical conductivity and consequently the absorption of the core increase rapidly above a
given temperature, Tc. Using this thermally induced optical absorption, Tc of 1100 ºC and an
optical power of 1 W, the core temperature was shown to reach 100000 ºC (Shuto et al.,
2003), which is well above the temperature of the fuse zone measured by (Dianov et al.,
2006).
Also, the trigger to ignite this effect was studied. The trigger is a high loss local point in the
fibre network, usually in damaged or dirty connectors or in tight fibre bends that, combined
with high power signals, generate a heating point (Andre et al., 2010b; Seo et al., 2003;
Martins et al., 2009; Andre et al., 2010a). The specific mechanism associated with the fuse
effect generation in optical connectors was also studied and correlated with the absorption
of the dust particles in the connector end face (Shuto et al., 2004c).
Another important issue is the power density threshold to initiate and maintain the fibre
fuse propagation. The investigation so far indicates that the power density threshold is ~1-5
MW/cm2, depending on the type of fibre and on the signal wavelength (Davis et al., 1997;
Seo et al., 2003). Note that the first experiments using microstructured fibres have shown
that the optical power density threshold value to ignite the phenomena is 10 times higher in
these fibres than in traditional step index silica fibres (Dianov et al., 2004b).
The increase of absorption that is believed to take place during fuse propagation was related
with Ge’ defects, as mentioned above, but also with Si E’ defects in the Germanium doped
silica core optical fibres. These defects are induced at high temperatures, like the
temperatures present in the fibre drawing process (Hanafusa et al., 1985). The E’ defects are
Thermal Effects in Optical Fibres 3
associated with oxygen vacancies ≡Ge−Si≡ and are stable at temperatures above 870 K. The
conjugation of this temperature dependent absorption mechanism with the absorption of
the SiO specimen, produced by the thermal decomposition reaction of the Silica glass at high
temperatures, occurring for temperatures above 3000 ºC, was considered by Shuto et al to
numerically simulate the fuse effect ignition and propagation. He reported estimate for the
Silica absorption coefficient was 107 m-1 at 6000 K for a wavelength of 1064 nm (Shuto, 2010).
Dianov et al has experimentally demonstrated that the radiation spectrum for the optical
discharge, propagating through the silica fibre, is close to that of the blackbody with plasma
temperature values of 10 4 K. The observed optical discharge velocities were up to 10 m/s
on step index single mode fibre (Dianov et al., 2006) and 30 m/s for Erbium doped
fibre(Davis et al., 1997).
Atkins et al propose a model for the bubble and voids tracks based on the Rayleigh
instability due to the capillary effects in the molten silica that surrounds the vaporised fibre
core(Atkins et al., 2003). The void formation and other dynamics of the fibre fuse
propagation were exhaustively studied, leading to models for the voids and bubbles shape
(Todoroki, 2005b; Todoroki, 2005c; Yakovlenko, 2006a), and profile models for the optical
discharge (Todoroki, 2005a). Todoroki has also shown that is possible to have optical
discharge without the formation of voids, along short distances, being this responsible by
the irregular patterns on the voids trail (Todoroki, 2005d).
Other authors have also observed and studied the fibre fuse effect in special fibres like hole
assisted fibre (Hanzawa et al., 2010), high numerical aperture fibres (Wang et al., 2008),
polarization maintaining fibres(Lee et al., 2006) or in dispersion shift and non zero
dispersion shift fibres (Rocha et al., 2010; Andre et al., 2010a).
Recently, more accurate simulation models for fuse propagation have been proposed
(Yakovlenko, 2006b), or even alternative models based on ordinary differential equations
that represent time saving in the numerical integration (Facao et al., 2011).
The concern with the effects for the network structure caused by the triggering of the fuse
effect imposes the development of devices with the capacity to stop the fuse zone
propagation. An early solution proposed in 1989 was the use of single mode tapers (Hand et
al., 1989). The decrease of the fibre cladding led to expansion of the optical discharge plasma
and to decrease of the power density, this results in the termination of the fuse propagation
(Dianov et al., 2004a). Others proposed solutions to detect the fuse effect that are based in the
analysis of the electric spectrum of the back reflected optical signal (Abedin et al., 2009), or
in the fast temperature increase in the fibre outer surface (Rocha et al., 2011).
The deployment of FTTH networks imposes a new challenge, the dissemination of the
optical fibre infrastructure in the access networks, where the fibre installation conditions are
not always the more adequate. In these conditions, the deployed fibre is subject to tight
bending, which impose an additional attenuation for the network power budget. The
additional attenuation of waveguides subject to tight bending is a well know phenomenon,
studied in 1976 by Marcuse(Marcuse, 1976). Marcuse associated the additional losses in
bent waveguides with the optical signal radiated to the cladding region, this model was
later improved by other authors (Harris et al., 1986; Valiente et al., 1989; Schermer et al.,
2007).
Besides the new attenuation limits imposed by the bending, other constrain was observed.
For high propagation power signals, the optical modes irradiated to the cladding, are
absorbed in the primary coating, resulting in a temperature increase. This local heating
point can induce the fuse effect already referred above or can reduce the fibre lifetime
(Percival et al., 2000; Glaesemann et al., 2006; Sikora et al., 2007).
The first works on this subject have shown that the temperature achieved by the fibre
coating was linearly correlated to the propagated optical power (Logunov et al., 2003).
However, this linear model fails to describe the coating temperature for high power
propagation (> 1 W), and recently an improved model that considers a nonlinear absorption
coefficient for the coating was proposed (Andre et al., 2010b).
This topic has attracted the focus of the scientific community and many new achievements
have been reported in the last years technical conferences. Namely, the correlation of
temperature and fibre time failure (Davis et al., 2005), the definition of the safety bending
limits (Andre et al., 2009; Rocha et al., 2009a). Recently, this topic was also studied in the
new bend insensitive fibres (G.657), showing that the maximum power that can be injected
safely in these fibres without coating risk is > 3 W (Bigot-Astruc et al., 2008).
3. Fibre fuse effect

As described in the previous section, the fibre fuse effect is a phenomenon that can occur in
optical fibres in the presence of high optical powers and that may lead to the destruction of
the optical fibre, along several kilometres, and also reach the optical emitter equipment,
resulting in a permanent damage of the network active components.
However, the presence of high optical powers is not enough to ignite the fibre fuse but a
trigger consisting of a initial heating point is also required. During the fuse effect ignition,
this initial heating point causes a strong light absorption, due to the thermal induced
absorption increase, which in turn leads to a catastrophic temperature increase, up to values
that are high enough to vaporize the optical fibre core. This fuse zone propagates towards
the light source melting and vaporizing the fibre core while a visible white light is emitted,
as schematically illustrated in Fig 1. The propagation of the fuse zone only stops if the input
power is reduced below the threshold value or even shut down. After the fuse zone
propagation, the fibre core shows a string of voids, being permanently damaged.
Fig. 1. Schematic representation of the fuse effect ignition and propagation in an optical fibre
3.1 Experimental characterization of the fuse effect

As referred above, the fibre fuse effect is initiated in a local heating point, whenever the
optical signal have powers above a certain threshold value. Fig. 2 presents a controllable
experimental setup for the fibre fuse ignition.
Dummy fiber Test fiber
High Power
Laser
Fuse effect
ignition
Fig. 2. Experimental setup implemented to study the fibre fuse effect (Rocha et al., 2011)
This setup consists in a short length of fibre (~3m) connected to a high power laser. The
other end of the fibre is placed in contact with a metallic foil in order to produce a local
heating and promote the fuse effect ignition. In order to protect the optical source, an optical
isolator and a dummy fibre with 20 m are used between the test fibre and the laser.
Fig. 3 shows three frames from a movie, displaying the fuse propagation. In this movie, the
white light emitted (optical discharge) from the fuse zone is clearly seen. The fuse discharge
propagates at constant velocity towards the light source.
Fig. 3. Sequence of frames of the fibre fuse propagation in a SMF fibre, the time difference
between pictures is 0.1s
As mentioned above, if the optical power is reduced below a threshold value, the fuse
propagation stops and the optical discharge extinguishes. For standard single mode fibre
(SMF-28, manufactured by Corning) and a laser signal with a wavelength of 1480 nm, the
optical discharge extinguishes for an optical power of 1.39 W.
The fibre fuse propagation velocity increases with the optical power density, and could
reach values high as 10 m/s (Dianov et al., 2006; Rocha et al., 2011). Fig. 4 presents the
experimental velocities for the fuse effect propagation, ignited with a laser signal at 1480 nm
in a SMF-28 fibre. These experimental results indicate that, for this limited range of optical
power values, the fibre fuse propagation velocity is linearly dependent on the optical power
launched into the fibre, however, if we consider higher optical power values, the velocity
will be no longer a linear function function of the optical power (Dianov et al., 2006; Facao et
al., 2011).
0.6
0.5
Optical discharge velocity (ms )
-1
0.4
0.3
0.2
0.1
0.0
1 1.39 2 3 4
Optical power (W)
Fig. 4. Fuse discharge velocity as function of the injected optical power. The arrow
represents the power threshold and the line correspond to the data linear fit
(slope=0.110±0.002 m s-1 W-1, intercept= 0.190±0.004 m s-1, correlation coefficient > 0.993)
12.5
10.0
Temperature increase (K)
7.5
5.0
2.5
0.0
20 30 40 50 60 70
Time (s)
Fig. 5. Fibre surface temperature increase as a function of time, recorded by the FBG sensor
during the optical discharge for an optical power of 3 W
The propagation velocity of the fuse zone was measured using a setup based on FBG (Fibre
Bragg Grating) temperature sensors that measure the fibre outer interface temperature
(Andre et al., 2010a; Rocha et al., 2011). Two fibres Bragg gratings were placed in contact
with the optical fibre outer interface, in two positions separated by 2 m.
The optical discharge leads to a temperature increase in the outer fibre surface, which is
monitored by the FBG sensors. The time difference between the temperature peaks,
recorded at each FBG, is then used to obtain the velocity of the optical discharge. Fig. 5
displays the temperature increase in the fibre surface measured by one FBG. This graph
presents an abrupt temperature increase, followed by an exponential decrease. The
temperature peak corresponds to the optical discharge passing through the FBG location.
Although, the fiber core is believed to achieve temperatures around 104 K during the optical
discharge, the fiber surface temperature increases just a few degrees above the
environmental temperature, as results of the heat transfer mechanisms (conduction,
radiation and convection) that dissipate the thermal energy along the optical fiber and to the
surrounding environment.
After the optical discharge propagation, the fibre presents a chain of voids in the core region
that can be observed with an optical microscope. Fig 6 displays the optical microscopic
images of the SMF fibre, obtained after the optical discharge propagation.
Fig. 6. Microscopic images of the optical fibre after the optical discharge propagation for
optical powers of 2.5 W (right) and 4.0 W (left) (pictures obtained using an optical
magnification of ×50)
These pictures were taken after the removal of the fibre coating. In these pictures, the
damage caused by the fuse is clearly visible, the voids are created in the melted/vaporized
core region with a periodic spatial distribution. The size and the spatial interval of the voids
vary with the input power and the type of fibre (Andre et al., 2010a). Fig 7 shows the relation
between the void period and the optical signal power. For this limited range of optical
powers, the void period is linearly dependent on the optical power level.
3.2 Theoretical model

Even though many underlining phenomena that sustain the fuse effect are still not
understood, the general explanation says that the initial high temperature zone, that ignite
the effect, increases strongly the light absorption that, in turn, is responsible for the increase
of the fibre temperature around 104 K (Dianov et al., 2006) well above the silica vaporization
temperature. The localized high temperature zone spreads to neighbouring regions, due to
heat conduction, and propagates into the laser direction, where the optical power signal is
present to drive the spike up of the temperature. The process repeats causing the
propagation of the optical discharge.
15
14
Void Interval (μm)
13
12
11
10
1 2 3
P (W)
Fig. 7. Void period as function of the injected optical power at 1480 nm in a SMF-28 fibre.
Points are experimental data and the line corresponds to the data linear fit (slope=1.38±0.06
µm W-1, intercept=10.1±0.2 µm, correlation coefficient > 0.944)
To summarise, we assume that the main process taking place in the fibre during the fuse
effect is a positive feedback heating process induced by temperature enhanced light
absorption.
In the recent years, there has been substantial interest in the development of theoretical
models for the fibre fuse phenomenon. Several hypotheses have been put forward to explain
the strong absorption, but as we mentioned previously a lot of mechanisms are still to be
understood, especially because it has been hard to measure the optical absorption at such
high temperatures or even to chemically analyse the contents of the voids and their
surrounding on a fuse damaged fibre. Nevertheless, most of these works propose a
propagation model based on a heat conduction equation with a heat source term that
corresponds to the optical signal absorption which itself is enhanced by the temperature
rise. This equation is coupled to an ordinary differential equation (ODE) for the spatial
evolution of the optical signal power (Shuto et al., 2003; Shuto et al., 2004a; Facao et al., 2011).
Hence, let us model the fuse effect by a one-space-dimensional heat conduction equation
coupled to an equation for the optical power evolution along the fibre length, namely:
∂T ∂ 2T α P 2σ sε 4
ρcp =k 2 + − (T − Tr4 ),
∂t ∂z π R2 R (1)
dP
= −α P ,
dz
where T(t,z) is the fibre temperature, P is the optical power, t is time and z is the
longitudinal coordinate along the fibre distance, ρ, Cp and k are the density, the specific heat,
and the thermal conductivity of the fibre, respectively and R is the optical signal mode field
radius.
The second term of the heat conduction equation is the heat source, caused by light
absorption, where α(T) is the local absorption coefficient and the last term represents the
loss by radiation which is written in terms of the Stefan-Boltzmann constant, σs, the surface
emissivity, ε, and the environment temperature, Tr.
The increase of the optical signal absorption coefficient, α, with temperature plays the most
important role in the generation of the fibre fuse. It was reported that the absorption
coefficient is temperature dependent and rapidly increases above a critical temperature
(1000 ºC), moreover it achieves a very large value for temperatures above 2000ºC (Hand et
al., 1988a; Hand et al., 1988b; Shuto et al., 2004b). In 1988, Hand and Russell suggested that the
absorption increase is closely related with Ge defects that are supposedly created in the core
of the fiber once the temperature rises. In their model the absorption dependence with
temperature is described by an Arrhenius law (Hand et al., 1988a). Shuto et al (Shuto et al.,
2004b) have also proposed that the formation of Ge related defects could increase the
number of free electrons and the subsequent electrical conductivity of the fibre core then
enhances the absorption. In their opinion, this mechanism would explain the absorption
values reported by Kashyap (Kashyap et al., 1997). This latter model also results in an
Arrhenius law. But since the fuse effect also occurs in fibres without germanium, other
models of absorption increase that do not rely on the presence of germanium should be put
forward. One of them, also proposed by Shuto et al (Shuto et al., 2004b) relates the absorption
increase with the thermochemically formation of SiO at higher temperatures, this model
would allow absorption values as large as 104 m-1 for 2293 K. They also propose that, for
lower temperatures, the Ge-defects should be the main absorption mechanism but for higher
temperatures the presence of SiO should be more relevant. Moreover, an Arrhenius law was
also proposed to model this absorption of SiO. Therefore, even if more than one process
promotes the temperature rise in fibers, whether they are doped with germanium or not, the
experimental data that have been collected up to now seem to manifest that all of them are
thermally induced, so they are the kind of processes that are frequently modeled by an
Arrhenius law.
For the reasons presented above and since here we are mainly concerned with common
fibres used in telecommunications, with Ge doped cores, we use the following Arrhenius
law, proposed by Hand and Russell, to model the temperature dependent optical absorption
coefficient (Hand et al., 1988a).
⎛ Ef ⎞
α = α0 exp ⎜ − ⎟ (2)
⎝ bT ⎠
k
where Ef = 2.5 eV (Shuto et al., 2004a) is the formation energy of the Ge defects, kB is
Boltzmann's constant and α0 is a constant dependent on the light wavelength and on the
optical fibre type.
As already stated, the fibre fuse phenomenon is initiated only in a local heating point,
thus in our model we assume an initial hot zone with the temperature above the critical
value Tc.
3.2.1 Simulation of the fibre fuse effect for a single mode fibre (SMF)
The system of equations (1) and (2) was integrated using a numerical routine from the NAG
toolbox, d03pp, that integrates nonlinear parabolic differential equations with automatic
adaptive spatial remeshing.
In this calculation it was assumed an optical signal source at 1480 nm, a mode field radius of
5.025 µm, that is the mode field radius for the optical fibre used in the experimental
characterization (SMF-28) at this wavelength and the same thermal proprieties of the silica
glass, since the Ge concentration is very low. In this study the variation of the thermal
proprieties and energy loss during changes of phase or even chemical reactions that should
occur for such high temperatures were neglected, therefore, we have used the constant
thermal coefficients of the silica glass listed in table 1. The equations were solved for a 3 cm
long fiber with a initial hot zone of 0.2 mm centered within the fiber, and considering
Tc=2900K and an environment temperature of 300 K.
ρ 2200 kg m-1
k 2 Wm-1K-1
Cp 1430 JKg-1K-1
Table 1. Thermal coefficients of silica glass (Sergeev et al., 1982; Facao et al., 2011)
Since the absorption dependence with temperature is not exactly known and, particularly,
the parameter of the Arrhenius law here proposed were not yet determined for this
wavelength neither for such high temperatures, the parameter α0 was adjusted with the
model described in (Facao et al., 2011) which gave α 0 = 4.56 × 106 m-1.
18000
16000
14000
12000
Temperature (K)
10000
8000
6000
4000
2000
0
0 1 2 3 4
Time (ms)
Fig. 8. Fibre core temperature at a fixed point as function of time for an optical power of 3 W
Fig. 8 displays the temperature peak that corresponds to the hot zone passing through a
fixed point in the fiber. It shows an abrupt change in temperature followed by an
exponential decrease. This temperature pulse is similar to the one obtained in the
experimental characterization but achieves a much higher temperature. Recall that, in the
experimental characterization, the measured temperature was done in the outer surface of
the fiber and the temperature obtained numerically is estimated in the fiber core. Then, the
temperature pulses moves in the negative-z direction (direction of the light source) with
constant velocity (see Fig 9 (a)).
18000 3.0
16000
2.5
14000
Optical Power (W)

2.0
Temperature (K)
12000
10000 propagation direction propagation direction

1.5
8000
1.0
6000
4000 0.5
2000
0.0
0
12 13 14 15 16 17 18 12 13 14 15 16 17 18
Position (mm) Position (mm)
(a) (b)
Fig. 9. Temperature (a) and power (b) distribution profiles along the propagation axis at
several temporal moments spaced by 1 ms, for an optical power of 3W (the profile timing
increases from the right to the left)
0.8
0.7
Optical discharge velocity (ms )
-1
0.6
0.5
0.4
0.3
0.2
1 2 3 4 5 6 7
Optical Power (W)
Fig. 10. Experimental (points) and numerical values (line) for the optical discharge propagation
velocity as function of the injected optical power
Fig. 9 shows the travelling pulses of the optical discharge and the travelling fronts of optical
power in several temporal moments. In the fuse zone all the incident optical signal is
absorbed, scattered or reflected and the transmitted optical power decreases abruptly down
to a null value.
The velocities obtained in this calculation are compared with the experimental values in Fig.
10, showing a good agreement with a relative error smaller than 6%.
As mentioned in section 3.2, for a limited range of relatively low optical power values, the
fiber fuse propagation velocity is linearly dependent on the optical power, yet Fig 10 shows
that the relation of the velocity with power density is not linear if a wider range is
considered.
4. Optical fiber coating temperature increase

The increase of the temperature in an optical fibre is one of the responsible causes for its
degradation. This increase in the temperature can be due to the high optical powers
propagating in regions with small bending diameters. In these regions, part of the optical
signal transfers from the core of the fibre to the surrounding cladding and coating (Bigot-
Astruc et al., 2006; Giraldi et al., 2009; Rocha et al., 2009a; Rocha et al., 2009b). This effect can
lead to the degradation of the protection layers of the fibre, with the consequent rupture, or
even to the fibre fuse effect ignition.
Some recent works have studied the optical fiber resistance to high optical power (Logunov
et al., 2003). The maximum attained temperature value has been studied, based on a thermal
model for low power propagation signals (<1W) (Percival et al., 2000).
Here, we report our study about coating temperatures of an optical fiber, when subjected to
low bending and high power optical signals. The coating temperature and the optical power
loss were measured for different bend diameters. For that purpose, we have implemented
an experimental setup consisting in a circular loop whose radius is controllable, as sketched
and imaged in Fig. 11. The setup was designed in order to assure that the fiber suffer a bend
of 360º without any change of radius. The bend diameters under study comprised values
between 2.95 mm and 20.14 mm.
a) b)
Fig. 11. a) Diagram of the experimental setup, and b) photograph of the circular loop with a
diameter of 12.50mm
The temperature in the bent region was measured with an infrared thermal camera
(ThermaCAM™ Flir i40). The optical signal source was a Raman laser (IPG - RLR-10-1480),
emitting at the wavelength of 1480 nm, with a maximum optical power of 2W. To determine
the total attenuation in the curved section of the fiber, the optical signal output was
analyzed with an optical power meter (EXFO FPM-600).
The most common type of fiber used in optical networks is the SMF28.G652.D, thus this
fiber was the one studied on the work presented here. The fiber, produced by Corning, has
an outer diameter of 125 µm, a core diameter of 10 µm and a primary acrylate coating with
an external diameter of 250 µm. The environmental temperature at which the tests took
place was 23ºC.
Examples of thermal images captured during the experiments are presented in Fig. 12.
These two images were taken for two different bend diameters, after one minute of exposure
to high optical powers (1.75 W). It is perceptible that the temperature in the curved section is
rising considerably with the decrease of the bend diameter.
a) b)
Fig. 12. Thermal image of the fiber bending section for an injected optical power of 1.75W
and bending diameters of a) 4.86 mm and b) 9.95 mm
Through the analysis of the thermal images, it is evident that, after an initial increase, the
temperature stabilizes with time. In Fig. 13, the fiber temperature values obtained for several
bending diameters are represented as a function of time. The input optical power was 0.5W
for this test.
Considering the above results, all the subsequent measures were made in the stationary
regime, i.e., 60 s after the optical power has been turned on. The critical bend diameter is 20
mm, since for higher diameters the rise of the injected optical power as no significant impact
in the temperature increase (Andre et al., 2009).
Fig. 14 shows the maximum temperature increase values as function of the injected optical
power for different bending diameters.
The observed behavior confirms the relation previously described, between the maximum
temperature value obtained and the fiber bend diameter (Andre et al., 2010b; Andre et al.,
2010c).
After the exposure to high optical powers, the physical condition of the fiber bent section
was observed using an optical microscope (Olympus SZH-ILLD). Fig. 15 displays two
obtained images, showing a fiber that has been submitted to an optical power of 1.5W, and a
bend diameter of 2.9 mm during 60 s.
175
150
125 20.1 mm
Temperature (°C)
12.5 mm
10.0 mm
100
7.1 mm
4.9 mm
3.0 mm
75
50
25
0 30 60 90 120
Time (s)
Fig. 13. Temperature values in the bent region along time
160
140
Temperature increase (ºC)
120
100
80 20.1 mm
15.0 mm
60 12.5 mm
10.0 mm
40 7.1 mm
5.0 mm
20
0
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6
Optical Power (W )
Fig. 14. Maximum temperature increase achieved as function of the optical power injected
for several values of bending diameters
Fig. 15. Image of the fiber bend section, after being submitted to an optical power of 1.5W.
The bending diameter was 2.9mm and the exposure time was 60 s (magnification of X5)
There, the damage induced by the optical power on the fiber coating in the bent section is
well visible. There is degradation of the acrylate and its physical detachment from the silica.
4.1 Modelling the fibre temperature

The damage in the fiber coating is inflicted by the optical power loss in the bent section ,
which induces the temperature increase of the coating, resulting in an oxidation of the
acrylate layer. The optical power irradiated from the optical fiber core is given by:
⎛ ⎛ α fiber L ⎞ ⎞
ΔP = Pin ⎜ 1 − 10 ⎜⎜ − ⎟⎟ (3)
⎜
⎝ ⎝ 10 ⎟⎠ ⎟⎠
where Pin is the optical power injected in the fiber, αfiber is the power loss associated to the
bend, expressed in dB/m and L is the length of bent fiber section.
The relation between the temperature increase and the power loss in the bent region is
presented in Fig. 16. This graph shows that, independently of the bending diameter, the
maximum temperature changes nonlinearly with the optical power loss, showing some
stabilization for optical power losses around 0.3 W (Andre et al., 2010b). The differences
observed in the maximum temperature for the several bending diameter can be explained
by the fact that some of the energy that is lost from the core guided mode can be guided in
the cladding (Andre et al., 2010b).
Assuming a thermal model that considers that the optical signal energy transferred to non
guided modes is absorbed by the acrylate layer and then converted into heat, it is possible to
estimate the temperature in the bent region. Nevertheless, due to the absorption saturation
of the coating observed for higher optical powers not all the energy is absorbed (Andre et
al., 2010b). In the stationary regime, the heat source (Pin) given by the radiation absorbed in
the acrylate is converted in stored heat in the fiber (Qstored) and in heat loss to the exterior
(Qout), which can be modeled considering a heat transfer coefficient (h) (Andre et al., 2010b).
dT (4)
Qstored = ρ c p Al
dt
Qout = π dlh(T − T0 ) (5)
where d and l are the diameter and length of the fiber bent section, respectively. T0 is the
environment temperature, A is the optical fiber area and ρ and cp are the density and
specific heat of the fiber (considering an average value for the silica and acrylate),
respectively (Andre et al., 2010b)
160
140
120
100
80
60
20.1 mm
40 15.0 mm
12.0 mm
20 10.0 mm
4.9 mm
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Optical power loss (W)
Fig. 16. Temperature as function of the optical power loss for different bending diameters
for a propagation signal at 1480 nm
As stated before, in the presence of high optical powers, the absorption in the acrylate layer
exhibits a non-linear behavior and thus the power absorbed by the acrylate saturates in a
maximum value given by:
Pin = Pmax ( 1 − exp ( − β ΔP ) ) (6)
So, in equilibrium, the maximum temperature rise can be expressed as (Andre et al., 2010b):
Pmax ( 1 − exp ( − β ΔP ) )
ΔT = (7)
π dlh
Where ∆T is the maximum temperature increase relatively to the environment equilibrium
temperature, Pmax is the maximum optical power absorbed by the coating and β is the
activation constant (Andre et al., 2010b). The experimental results referred above for
maximum temperature increase versus injected optical power for different bending
diameters were fitted to the theoretical curve given by equation (7). The results are
presented in Fig. 17. The parameters for this theoretical fit were Pmax/h and β and the
reduced chi-square value obtained was of 0.4542.
These results show that is possible to correlate the bend diameter, the injected optical power
and the temperature increase in the optical fiber coating. This model can be useful for the
design of future systems, being an approach to limit the bending diameter to values that
depend on the injected power, in order to maintain the operational conditions below a
safety limit.
160
140
120
100
80 20.14 mm
15.04 mm
60 12.50 mm
9.95 mm
40 4.86 mm
Theoretical Fit
20
0
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6
Injected Power (W)
Fig. 17. Optical fiber maximum temperature increase as function of injected optical power
for different values of bending diameter. Dots are experimental results and the lines are the
fitting results, with Pmax/h=0.00245±0.00010 K⋅m2 and β=5.03±0.68 W-1
5. Conclusions
In this work, we have analyzed the thermal effects occurring in an optical fiber. We have
shown that the maximum power required to extinguish the optical discharge propagation is
1.39 W. It was also shown that the critical bending diameter for SMF fibers with high power
signal is 20 mm.
These results describe the main thermal affects occurring in an optical fibre used in
telecommunications.
6. References
Abedin, K. S. and T. Morioka (2009). Remote Detection of Fiber Fuse Propagating in Optical
Fibers, Proceedings of Conference on Optical Fiber Communication, collocated National
Fiber Optic Engineers Conference Vol. 1-5, pp. OThD5, San Diego, CA, USA.
Andre, P. S., M. Facao, A. M. Rocha, P. Antunes and A. Martins (2010a). Evaluation of the
fuse effect propagation in networks infrastructures with different types of fibers,
Proceedings of Conference on Optical Fiber Communication, collocated National Fiber
Optic Engineers Conference pp. JWA10, San Diego, CA, USA.
Andre, P. S., A. M. Rocha, F. Domingues and A. Martins (2010b). Improved thermal model
for optical fibre coating owing to small bending diameter and high power signals,
Electronics Letters, Vol. 46, No. 10, pp. 695-696, ISSN: 0013-5194.
Andre, P. S., A. M. Rocha, F. Domingues and A. Martins (2010c). Thermal model for optical
fiber coating under tight bending diameters, Proceedings of 15th OptoeElectronics and
Communications Conference pp. 318-319, Sapporo, Japan.
Andre, P. S., A. M. Rocha, B. Neto, A. Martins, M. Facao, J. L. Pinto, A. L. J. Teixeira, R. N.
Nogueira, M. J. Lima, G. Incerti, D. Forin and G. T. Beleffi (2009). Optical Fiber
Bending Limits for Optical Fiber Infraestructures, Proceedings of IEEE Africon, pp.
882-884, Nairobi, Kenya.
Atkins, R. M., P. G. Simpkins and A. D. Yablon (2003). Track of a fiber fuse: a Rayleigh
instability in optical waveguides, Optics Letters, Vol. 28, No. 12, pp. 974-976, ISSN:
0146-9592.
Bigot-Astruc, M., L. A. de Montmorillon and P. Sillard (2008). High-Power Resistance of
Bend-Optimized Single-Mode Fibers, Proceedings of Conference on Optical Fiber
communication / National Fiber Optic Engineers Conference, pp. JWA2, San Diego, CA,
USA.
Bigot-Astruc, M., P. Sillard, S. Gauchard, P. Le Roux and E. Brandon (2006). Analysis of
coating temperature increase in fibers under high power and tight bending,
Proceedings of Optical Fiber Communication Conference/National Fiber Optic Engineers
Conference, pp. OFK4, Anaheim, CA, USA.
Davis, D. D., S. C. Mettler and D. J. DiGiovanni (1997). A comparative evaluation of fiber
fuse models, Proceedings of SPIE, Vol. 2966, pp. 592-606.
Davis, I. M., G. S. Glaesemann, S. Ten and M. J. Winningham (2005). Optical fibres resilient
to failure in bending under high power, Proceedings of 31st European Conference on
Optical Communication, Vol. 3, pp. We3.4.5, ISBN 0537-9989.
Dianov, E. M., I. A. Bufetov and A. A. Frolov (2004a). Destruction of silica fiber cladding by
the fuse effect, Optics Letters, Vol. 29, No. 16, pp. 1852-1854, ISSN: 0146-9592.
Dianov, E. M., I. A. Bufetov, A. A. Frolov, Y. K. Chamorovsky, G. A. Ivanov and I. L.
Vorobjev (2004b). Fiber fuse effect in microstructured fibers, Ieee Photonics
Technology Letters, Vol. 16, No. 1, pp. 180-181, ISSN: 1041-1135.
Dianov, E. M., V. E. Fortov, I. A. Bufetov, V. P. Efremov, A. E. Rakitin, M. A. Melkumov, M.
I. Kulish and A. A. Frolov (2006). High-speed photography, spectra, and
temperature of optical discharge in silica-based fibers, Ieee Photonics Technology
Letters, Vol. 18, No. 5-8, pp. 752-754, ISSN: 1041-1135.
Driscoll, T. J., J. M. Calo and N. M. Lawandy (1991). Explaining the Optical Fuse, Optics
Letters, Vol. 16, No. 13, pp. 1046-1048, ISSN: 0146-9592.
Facao, M., A. M. Rocha and P. S. Andre (2011). Traveling Solutions of the Fuse Effect in
Optical Fibers, Journal of Lightwave Technology, Vol. 29, No. 1, pp. 109-114, ISSN:
0733-8724.
Giraldi, M. T. M. R., A. M. Rocha, B. Neto, C. Correia, M. E. V. Segatto, M. J. Pontes, A. P. L.
Barbero, J. C. W. Costa, M. A. G. Martinez, O. Frazao, J. M. Baptista, H. M. Salgado,
M. B. Marques, A. L. J. Teixeira and P. S. Andre (2009). Brillouin Effects in
Distributed Raman Amplifiers under Saturated Conditions, Proceedings of Ieee Mtt-S
International Microwave and Optoelectronics Conference pp. 831-835
Glaesemann, G. S., M. J. Winningham, D. A. Clark, J. Coon, S. E. DeMartino, S. L. Logunov and
C. K. Chien (2006). Mechanical failure of bent optical fiber subjected to high power,
Journal of the American Ceramic Society, Vol. 89, No. 1, pp. 50-56, ISSN: 0002-7820.
Hanafusa, H., Y. Hibino and F. Yamamoto (1985). Formation Mechanism of Drawing-

Induced E' Centers in Silica Optical Fibers, Journal of Applied Physics, Vol. 58, No. 3,
pp. 1356-1361, ISSN: 0021-8979.
Hand, D. P. and T. A. Birks (1989). Single-Mode Tapers as Fiber Fuse Damage Circuit-
Breakers, Electronics Letters, Vol. 25, No. 1, pp. 33-34, ISSN: 0013-5194.
Hand, D. P. and P. S. Russell (1988a). Solitary thermal shock waves and optical damage in
optical fibers: the fiber fuse, Optics Letters, Vol. 13, No. 9, pp. 767-769, ISSN: 0146-9592
Hand, D. P. and P. S. J. Russell (1988b). Soliton-like thermal shock-waves in optical fibres:
origin of periodic damage tracks, Proceedings of Fourteenth European Conference on
Optical Communication, pp. 111-114 vol.111.
Hanzawa, N., K. Kurokawa, K. Tsujikawa, K. Takenaga, S. Tanigawa, S. Matsuo and S.
Tomita (2010). Observation of a propagation mode of a fiber fuse with a long-
period damage track in hole-assisted fiber, Optics Letters, Vol. 35, No. 12, pp. 2004-
2006, ISSN: 0146-9592.
Harris, A. J. and P. F. Castle (1986). Bend Loss Measurements on High Numerical Aperture
Single-Mode Fibers as a Function of Wavelength and Bend Radius, Journal of
Lightwave Technology, Vol. 4, No. 1, pp. 34-40, ISSN: 0733-8724.
Kashyap, R. (1987). Self-propelled self-focusing damage in optical fibers, Proceedings of X
International Conference on Lasers, Lake Tahoe.
Kashyap, R. and K. J. Blow (1988). Observation of Catastrophic Self-Propelled Self-Focusing
in Optical Fibers, Electronics Letters, Vol. 24, No. 1, pp. 47-49, ISSN: 0013-5194.
Kashyap, R., A. H. Sayles and G. F. Cornwell (1997). Heat-flow modeling and visualization
of catastrophic self-propagating damage in single-mode optical fibers at low
powers, Proceedings of SPIE, Vol. 2966, pp. 586-591.
Lee, M. M., J. M. Roth, T. G. Ulmer and C. V. Cryan (2006). The fiber fuse phenomenon in
polarization-maintaining fibers at 1.55 μm, Proceedings of Lasers and Electro-Optics
and Quantum Electronics and Laser Science Conference, pp. JWB66, Long Beach, CA,
USA.
Logunov, S. L. and M. E. DeRosa (2003). Effect of coating heating by high power in optical
fibres at small bend diameters, Electronics Letters, Vol. 39, No. 12, pp. 897-898, ISSN:
0013-5194.
Marcuse, D. (1976). Curvature Loss Formula for Optical Fibers, Journal of the Optical Society of
America, Vol. 66, No. 3, pp. 216-220, ISSN: 0030-3941.
Martins, A., A. M. Rocha and P. S. Andre (2009). High power effects on fiber optic
connectors, Proceedings of International Microwave and Optoelectronics Conference
(IMOC), ISBN 1679-4389, Belem, Brazil.
Percival, R. M., E. S. R. Sikora and R. Wyatt (2000). Catastrophic damage and accelerated
ageing in bent fibres caused by high optical powers, Electronics Letters, Vol. 36, No.
5, pp. 414-416, ISSN: 0013-5194.
Rocha, A. M., P. Antunes, F. Domingues, M. Facao and P. S. Andre (2011) Detection of Fiber
Fuse Effect Using FBG Sensors. IEEE Sensors Journal, DOI:
10.1109/JSEN.2010.2094183
Rocha, A. M., M. Facão and P. S. André (2010). Study of fiber fuse effect on different types of
single mode optical fibers, Proceedings of Network and Optical Communications, pp.
75-79, Faro, Portugal.
Rocha, A. M., A. Martins, M. Facao and P. S. Andre (2009a). Effect of bending in SMF fibers
under high power, Proceedings of 11th International Conference on Transparent Optical
Networks, pp. Th.B2.1, Azores, Portugal.
Rocha, A. M., B. Neto, A. Martins, G. Incerti, D. Forin, G. Bellefi, M. Facao, J. L. Pinto, A. L. J.

Teixeira, R. N. Nogueira, M. J. Lima and P. S. Andre (2009b). The effect of high
power propagation in bended fibers, Proceedings of 10th International Conference
onTelecommunications. , pp. 303-304.
Schermer, R. T. and J. H. Cole (2007). Improved bend loss formula verified for optical fiber
by simulation and experiment, Ieee Journal of Quantum Electronics, Vol. 43, No. 9-10,
pp. 899-909, ISSN: 0018-9197.
Seo, K., N. Nishimura, M. Shiino, R. Yuguchi and H. Sasaki (2003). Evaluation of High-
power Endurance in Optical Fiber Links, Furukawa Review, Vol. 24, pp. 17-22,
Sergeev, O. A., A. G. Shashkov and A. S. Umanskii (1982). Thermophysical properties of
quartz glass, Journal of Engineering Physics and Thermophysics, Vol. 43, No. 6, pp.
1375-1383, ISSN: 1062-0125.
Shuto, Y. (2010). Evaluation of High-Temperature Absorption Coefficients of Ionized Gas
Plasmas in Optical Fibers, Ieee Photonics Technology Letters, Vol. 22, No. 3, pp. 134-
136, ISSN: 1041-1135.
Shuto, Y., S. Yanagi, S. Asakawa, M. Kobayashi and R. Nagase (2004a). Fiber fuse
phenomenon in step-index single-mode optical fibers, Ieee Journal of Quantum
Electronics, Vol. 40, No. 8, pp. 1113-1121, ISSN: 0018-9197.
Shuto, Y., S. Yanagi, S. Asakawa, M. Kobayashi and R. Nagase (2004b). Evaluation of high-
temperature absorption coefficients of optical fibers, Ieee Photonics Technology
Letters, Vol. 16, No. 4, pp. 1008-1010, ISSN: 1041-1135.
Shuto, Y., S. Yanagi, S. Asakawa, M. Kobayashi and R. Nagase (2004c). Fiber fuse generation
in single-mode fiber-optic connectors, Ieee Photonics Technology Letters, Vol. 16, No.
1, pp. 174-176, ISSN: 1041-1135.
Shuto, Y., S. Yanagi, S. Asakawa and R. Nagase (2003). Generation mechanism on fiber fuse
phenomenon in single-mode optical fibers, Electronics and Communications in Japan
Part Ii-Electronics, Vol. 86, No. 11, pp. 11-20, ISSN: 8756-663X.
Sikora, E. S. R., D. J. McCartney and J. V. Wright (2007). Impact of coating ageing on
susceptibility to high-power damage at fibre bends, Electronics Letters, Vol. 43, No.
4, pp. 208-210, ISSN: 0013-5194.
Todoroki, S. (2005a). In-situ observation of fiber-fuse propagation, Japanese Journal of Applied
Physics Part 1-Regular Papers Short Notes & Review Papers, Vol. 44, No. 6A, pp. 4022-
4024, ISSN: 0021-4922.
Todoroki, S. (2005b). Origin of periodic void formation during fiber fuse, Optics Express, Vol.
13, No. 17, pp. 6381-6389, ISSN: 1094-4087.
Todoroki, S. (2005c). Animation of fiber fuse damage, demonstrating periodic void
formation, Optics Letters, Vol. 30, No. 19, pp. 2551-2553, ISSN: 0146-9592.
Todoroki, S. (2005d). Transient propagation mode of fiber fuse leaving no voids, Optics
Express, Vol. 13, No. 23, pp. 9248-9256, ISSN: 1094-4087.
Valiente, I. and C. Vassallo (1989). New Formalism for Bending Losses in Coated Single-Mode
Optical Fibers, Electronics Letters, Vol. 25, No. 22, pp. 1544-1545, ISSN: 0013-5194.
Wang, J., S. Gray, D. Walton and L. Zantano (2008). Fiber Fuse in High power optical fiber,
Proceedings of SPIE, Vol. 7134, pp. 2E-1-9.
Yakovlenko, S. I. (2006a). Mechanism for the void formation in the bright spot of a fiber fuse,
Laser Physics, Vol. 16, No. 3, pp. 474-476, ISSN: 1054-660X.
Yakovlenko, S. I. (2006b). Physical processes upon the optical discharge propagation in
optical fiber, Laser Physics, Vol. 16, No. 9, pp. 1273-1290, ISSN: 1054-660X.
0
2
Heat Transfer for NDE: Landmine Detection

Fernando Pardo, Paula López and Diego Cabello
Universidad de Santiago de Compostela
Spain
1. Introduction
Although land mine problems existed in many regions, Bosnia (1995) and Afghanistan
(2001) gave the land mine issue a particular sense of urgency. Intended for warfare, these
mines remain buried after the end of the conflict. These mines are triggered by civilians
causing around 15,000-20,000 victims per year in 90 countries, ICBL (2006). The U.S. State
Department estimates that there are around 40-50 million of buried mines that need to be
cleared. According to Horowitz (1996) 100,000 mines are found and destroyed per year; thus
450 years will be necessary to clean all mines. However, each year, 1.9 million of new mines
are buried. In addition, the presence of mines also causes economic decline being one of the
major limitations to agricultural work on these regions, Cameron & Lawson (1998). Thus, it
is necessary to develop new techniques which allow to detect mines quickly and with high
precision. The Ottawa treaty, Ottawa (1997), banning the production and use of AP mines
was signed by 158 countries in 2007 however the most important AP manufacturers, China,
Russia, India and EE.UU, have not yet signed it.
Nowadays more than 350 types of mines exist, Vines & Thompson (1999); but they can be
broadly divided into two main categories:
• Antipersonnel (AP) mines.
• Antitank (AT) mines.
AT mines are relative big and heavy (2-5 Kg) and are usually laid on the ground forming
regular patterns and shallowly buried. AT mines have enough explosive to destroy a tank or
a truck, as well as to kill people in or around the vehicle; they also require more pressure to
be detonated than AP mines. On the contrary AP mines contain less explosive and are lighter
than AT mines. AP mines can be buried anywhere, they may lie on the surface or be shallowly
buried. Sometimes they are placed in a regular pattern to protect AT mines, however in most
cases they are placed randomly. Moreover, as AP mines are light and small, wind or rain can
easily move them making their location, even with the original pattern, more difficult. AP
mines are designed to damage foot soldiers avoiding their penetration into an specific area.
These mines can kill or disable their victims and are activated by pressure, tripwire or remote
detonation. These characteristics make AT mine detection and clearance easier than AP mine
detection.
Detection and clearance of buried mines is a big problem with lots of humanitarian,
environmental and economic implications. Current techniques for non-destructive evaluation
22
2 Developments inWill-be-set-by-IN-TECH
Heat Transfer
of soils fail to address the detection of small plastic landmines, which are the most difficult to
detect. There is no universal technique capable of detecting buried landmines in all situations.
The most widespread techniques in mine detection are metal detector and magnetometers.
Magnetometers are used to detect ferromagnetic objects and they measure the disturbance of
the earth’s natural electromagnetic field. Many modern mines have almost no metal parts
except for the small striker pin. Although metal detectors can be tuned to be sensitive enough
to detect these small items (current detectors can track a tenth of a gram of metal at a depth of
10 cm), such sensitivity detects more metal debris and increases considerably the rate of false
alarms. Increasing the sensitivity of metal detectors, therefore, does not solve the problem of
non-metal mines satisfactorily. Taking in mind this limitation new techniques are appearing
to detect the plastic landmines. Several techniques have been proposed for mine detection,
such as acoustic techniques, Sabatier & Xiang (2001); X-rays techniques, Lockwood et al.
(1997); biosensors, Larsson & Abrahamsson (1993); ion mobility spectrometers, Jankowski
et al. (1992); nuclear quadrupole resonance, Englelbeen (1998); neutron analysis, Bach et al.
(1996); ground penetrating radar (GPR) and infrared thermography (IRT), López et al. (2009);
Thanh et al. (2009; 2007; 2008). Each technique has its advantages and disadvantages, for a
more detailed description of theses techniques see Furuta & Ishikawa (2009); Gros & Bruschini
(1998); López (2003); Robledoa et al. (2009); Siegel (2002).
Infrared thermography is an attractive technique for some mine detection tasks because
it can be used from a considerable standoff distance, it provides information on several
mine properties, and it can rapidly survey large areas. Its ability to detect mines has been
recognized since the 1950s, Maksymonko & Le (1999). IRT sensors respond to electromagnetic
radiation in a sensor-specific wavelength range. The source of the received signal may be
either natural (i.e., thermal emission from the target or scattering of sunlight) or artificial
(e.g., an infrared illuminator), which leads to both passive and active sensor concepts. Each
material shows a characteristic thermal response to a given stimulus, also known as the thermal
signature. Thus, the cooling or heating process affects buried objects and the surrounding soil
in a different way. This difference is due to the fact that the mines are a better insulators
than the soil. Thus, in the case of passive thermography, the soil layer over the mine tends to
accumulate thermal energy during the day because the mine blocks the transport of thermal
energy. As a result of this process the soil over the mine tends to be warmer than the
surrounding soil. In the evening, the soil over the mine gives up thermal energy faster than the
surrounding soil and results cooler. Around midday the soil and the mine reach the thermal
equilibrium, which makes it impossible to perform the detection in this temporal gap. The
main limitation of this technique is the fact that temperature differences strongly depend on
the atmospheric conditions. On the other hand, active thermography uses artificial energy to
heat the soil under study, avoiding the dependence on the atmospheric conditions. In spite of
their long history, there is little compelling performance data available for infrared detection
of antipersonnel mines.
The use of IRT as non-destructive evaluation (NDE) for landmine detection consists of
subjecting the area under inspection to a source of natural or artificial heating/cooling process
and studying the soil’s response by means of the analysis of its thermal evolution given by a
temporal sequence of infrared images. In this sense the study of the basic phenomenology
lead to the development of mathematical models of the soil, England et al. (1992); England
(1990); Kahle (1977); Liou & England (1998); Pregowski et al. (2000). The idea underneath is to
Heat Transfer
Heat Transfer for NDE:
for NDE: Landmine Landmine Detection
Detection 233
characterize, and therefore predict, the thermal behavior of the unperturbed soil under given
conditions, i.e., its thermal signature. The presence of buried mines or other objects will alter
this signature, which if manifested as a thermal contrast on the surface. From the analysis
of this thermal contrast and, in particular of its dynamic evolution, it is possible to extract
relevant information about the nature of the objects. A deep analysis of the evolution of the
thermal contrast over a diurnal cycle can be found in Khanafer & Vafai (2002), where the
conditions for detection are studied, and sunrise and sunset are established as the periods
of the day in which the presence of buried mines induces a greater thermal contrast on the
surface (around 4-6°C).
An efficient way of extracting information from this data regarding the presence of landmines
using a 3D thermal model of the soil based on the solution of the heat equation was presented
in López et al. (2009; 2004). The process is divided in two steps. On the first one, the
forward problem, the soil is subjected to a heating process and a comparison between
temperatures measured at the soil surface (through IR imaging) and those obtained by
simulation using the model under the assumption of homogeneous soil and mine absence
is made. The differences between measured and simulated data put into evidence the
presence of unexpected objects on the soil. The second step is an inverse engineering
problem where the thermal model must be run for multiple soil configurations representing
different types of possible targets (mine, stone, ...) and depths of burial. The nearest
configuration to the measured data give us the estimated nature and location of the targets.
This approach and, particularly, the inverse engineering process, makes an intensive use of
the 3D thermal model that needs to be solved iteratively involving complex, coupled sets of
partial differential equations. The extensive computing power required makes impractical
its software implementation in personal computers . An alternative solution is the hardware
implementation of the Finite Difference (FD) representation of the thermal model. In fact,
different hardware implementations of FD solutions in the electromagnetics domain can be
found in the literature, Durbano et al. (2004); Placidi et al. (2002); Schneider et al. (2002). In
previous works, Pardo et al. (2009; 2010), we have presented an FPGA implementation of
a FD Heat Equation solver which speeds the computation up by a factor of 10 compared
to the purely software solution. The bottleneck of such an implementation is the access to
memory and the amount of available memory, which dramatically reduces the performance
of the system. However, this solution is hardware dependent an its cost is quite high. In
recent years, Graphic Processing Units (GPUs) have been proved to be a valuable hardware
platform to solve problems with a high degree of parallelism, Hwu et al. (2008). The total
speedup of the system in a typical simulation setup by a factor 40 compared to a Core2Duo
2.8 GHz implementation in C++, using a NVIDIA GTS250 GPU, but even higher speedups
could be achieved with more advanced GPUs.
The chapter is outlined as follows. In Section 2 the thermal model and the detection algorithm
are introduced. Section 3 addresses the architecture of the hardware implementation and its
GPU projection details. In Section 4 the main results are shown and, finally, the conclusions
are summarized.
2. Infrared thermography for NDE

Infrared thermography sensors, which respond to electromagnetic radiation in a
sensor-specific wavelength range, constitute an attractive NDE technique because they can
24
Heat Transfer

Fig. 1. Different heat transfer processes taking place in the soil volume and at the air-soil
interface.
be used from a considerable standoff distance to rapidly survey large areas while providing
information on several properties. Different materials show a characteristic response to
a given thermal stimulus over time, a property that can be successfully used for the
identification of buried targets. We will present a general procedure for the inspection of
soils based on the solution of the heat equation applied, but not limited, to the detection of
plastic antipersonnel mines.
2.1 Thermal model of the soil

The main processes considered are summarized in Fig. 1. We consider a soil volume, Ω,
subjected to a known thermal stimulus through the soil-air interface, Γ, where both the
soil and the targets are modeled as isotropic objects. We also assume that the thermal
diffusivity is constant and that the temporal variation of the moisture content and the mass
transference during the time of analysis are negligible, which are fair assumptions as long as
the duration of the experiment does not exceed a couple of hours and the depth of inspection
is limited to 10-15cm. A typical example is antipersonnel mine detection, given that mines
are usually either laid on the surface or only shallowly buried. Moreover, empirical evidence
demonstrates that it is possible to reduce the interval of interest to roughly one hour around
sunrise or sunset as it is at these times when their different thermal evolution is more clearly
manifested. Under these assumptions the overall process is described by the time-dependent
single-phase 3D heat equation:
∂T (−
→r , t) →
− div α(−
r ) grad T (−
→ k
r , t) = 0 , α = (1)
∂t ρc p
where −→r = ( x, y, z) with r ∈ Ω, ρ [kg/m3 ] is the density, c p [J/kg K] is the specific heat,
k[W/m K] is the thermal conductivity, α [m2 /s] is the thermal diffusivity and T [K] is the
distribution of temperatures inside the soil. To solve this equation the initial and boundary
Heat Transfer
Detection 255
conditions are required:
∂T (−
→r , t)
k = qnet (t) f or τ × Γ (2)
∂n
∂T (−
→r , t)
= 0 f or τ × ∂Ω\Γ (3)
∂n
T ( x, y, z → ∞, t) = T∞ (4)
T ( r , t = t ) = T (−
−
→
0 0
→r ) in Ω (5)
where τ is the time interval of analysis, qnet is the net heat that flows through the soil-air
interface Γ, n is the normal to the surface under consideration ∂Ω. Eq. (2) gives the boundary
condition at the air-soil interface; Eq. (3) shows the boundary conditions applied to the sides
of the volume not accessible for measurements, imposing a vanishing heat flux across them.
Eq. (4) is the deep-ground condition and it establishes that the temperature at a large enough
depth remains constant. Finally, Eq. (5) gives the initial conditions for the system. The net
heat flux at the soil-air interface, qnet , in Eq. (2) can be written as:
qnet (t) = qsun (t) + qrad (t) + qconv (t) (6)
where qsun is the short-wave radiation emitted by the sun and absorbed by the soil; qconv
represents the convection term at the soil-air interface; and qrad is the heat flux exchange due
to radiation. The first term of this expression, qsun , can be easily measured with the help of
low-cost equipment. The second term is given by qrad (t) = qsky (t) − qsoil (t). The term qsky (t)
is the longwave radiation from the atmosphere given by Stefan’s law. Finally, the third term
in Eq. (6) can be approximated by, qconv (t) = h( Tair (t) − Tsoil (t)), where h [W/m2 K] is the
convective heat transfer coefficient that is known to depend strongly on the wind speed.
2.2 Algebraic equations

Eq. (1) can be solved applying a FD approach, using either implicit or explicit methods. Using
the explicit method a set of equations that can be solved sequentially is obtained, its main
advantages being the simplicity of the formulation and its direct hardware translation. This
method, however, is only conditionally stable, Bejan (1993). On the other hand, the implicit
schemes are unconditionally stable but require the solution of the whole set of equations
simultaneously. The ADI method, Wang & Chen (2002), can overcome the drawbacks of
these two methods using a mixed approach, where only a subset of equations must be solved
simultaneously while retaining the unconditional stability property. Thus, high temporal
discretization steps can be used, effectively reducing the computing time when, for instance,
a system evolving toward a steady state is studied. However, in our case, the boundary
conditions change during the simulation time making necessary the use of small temporal
discretization steps and, therefore, the benefit of using an ADI method is lost. Due to this fact
and to the simplicity of the equations obtained, which allows to easily perform a hardware
implementation, we have chosen an explicit scheme to solve the heat equation. Under this
approach the derivatives can be approximated as:
26
Heat Transfer

Spatial discretization
x nx
y
z Δ Δ
ny

Δ
nz

(a) (b)
Fig. 2. (a) Spatial discretization scheme of the continuous soil into a set of discrete nodes. (b)
Considered neighbors to perform the updating process.
∂2 T (r, t) Ti+1,j,l − 2Ti,j,l + Ti−1,j,l

≈ (7)
∂x2 (Δx )2
∂T (r, t) T m +1 − T m
≈ (8)
∂t Δt
where Δx and Δt are, respectively, the spatial and temporal discretization steps; i, j, l are the
discretization indexes in the x, y and z directions respectively and m is the time discretization
index. The spatial discretization scheme of the continuous soil into a mesh of discrete nodes
can be seen in Fig. 2(a), where n x , ny and nz represent the number of nodes in the x, y and z
directions respectively (i = 0, 1, ...n x − 1, j = 0, 1, ...ny − 1 and l = 0, 1, ...nz − 1). Applying
this discretization scheme to Eq. (1), we obtain the following equations for a surface (Ti,j,0 )
and an internal node (Ti,j,l , with l > 0), assuming, without loss of generality, that the spatial
discretization steps are the same in all directions, that is, Δx = Δy = Δz,
m +1
Ti,j,0 = Ti,j,0
m
+ F0 [ ∑ ( Tneighbor
m
− Ti,j,0
m
) + 2( Ti,j,1
m
− Ti,j,0
m
)]
neighbors sur f ace (9)
sun + (2F0 H + 8F0 RTair )( Tair − Ti,j,0 ),
+2αsun F0 Sqm 3 m
∀i, j
m +1
Ti,j,l = Ti,j,l
m
+ F0 ∑ ( Tneighbor
m
− Ti,j,l
m
), ∀i, j, l > 0 (10)
neighbors
where F0 = F0 (i, j, l ), R, S and H are defined as,
αΔt σ Δx
F0 = R= Δx S= H = hS. (11)
(Δx )2 k k
As seen, every new temperature, T m+1 , is calculated as a function of the temperatures in the
previous time step, T m , and only the nearest neighbors are considered in order to facilitate the
hardware implementation, see Fig. 2(b). In the case of the surface layer, only the neighbors in
Heat Transfer
Detection 277
the same layer (z = 0) and in the layer beneath (z = 1) are considered. Using the explicit FD
method Eq. (10) can be written as,
m +1
Ti,j,l = Ti,j,l
m
(1 − 6F0 ) + F0 ( Tim+1,j,l + Tim−1,j,l + Ti,j
m
+1,l + Ti,j−1,l + Ti,j,l +1 + Ti,j,l −1 )
m m m
(12)
Thus, for an internal node the instability is avoided if the term (1 − 6F0 ) of Eq. (12) is non
negative, (Bejan, 1993; Incropera & DeWitt, 2004), which is translated into a small enough Δt
such that,
αΔt 1
F0 = ≤ (13)
(Δx ) 2 6
As can be noted, the stability criteria is a trade-off between the spatial and temporal
discretization steps. The small size of the mines constraints the value of Δx, which must be
small enough to have several nodes representing the mines. Thus, a small value of Δx must be
chosen which also implies to choose a small value of Δt. Meanwhile for a surface node Eq. (9)
the stability criteria can be written as:
m +1 3
Ti,j,0 = Ti,j,0
m
(1 − 6F0 − 2F0 H − 8F0 RTair )+
F0 ( Tim+1,j,0 + Tim−1,j,0 + Ti,j
m
+1,0 + Ti,j−1,0 + 2 Ti,j,1 )+
m m (14)
3
(2F0 H + 8F0 RTair ) Tair + 2αsun F0 Sqm
sun
Resulting in the following stability condition for a surface node:
1
F0 ≤ . (15)
6 + 2H + 8RTair
3
As shown, the maximum Δt for a given Δx is limited by the condition on F0 .

In summary, applying the FD method, the thermal model of the soil is discretized into a
set of nodes whose thermal behavior is given by algebraic equations involving additions,
subtractions and multiplications, making it suitable for a hardware implementation.
2.3 IRT-based soil inspection for the detection of buried objects

Many physical systems are characterized by the solution of a differential equation or system
of equations subject to known boundary conditions. This is the called forward problem. The
representation of a non-linear forward operator can take the form of a functional equation
involving a map F, which represents the connection between the model and the data:
y = F [ p] (16)
where F is a non-linear operator between Hilberts space Y and P, y is the measured data and p
the original distribution of parameters that gives rise to y under the application of operator F.
An inverse problem of this will be the reconstruction of the original distribution of parameters
based on the measurements of the resulting data.
Solving an inverse problem implies approximating the best solution p† = F −1 [y] of Eq. (16).
In general, y ∈ Y is never known exactly but up to an error of δ = 0. Therefore, we assume
28
Heat Transfer
that we know δ > 0 and yδ ∈ Y with y − yδ ≤ δ. Thus, yδ is the noisy data and δ is the
noise level. Under these situations is, generally, impossible to compute numerically a solution
of the problem unless making use of the regularization techniques, which makes it possible
to restore stability and existence/uniqueness of the solution and develop efficient numerical
algorithms, Kirsch (1996).
An example of such a technique is the non-linear Landweber iteration method which is
defined by the following recursive procedure, Engl et al. (1996):
pδk = pδk−1 + F [ pδk−1 ] T (yδ − F [ pδk−1 ]) k ∈ (17)
As initial guess pk=0 = p0 is set, incorporating a priori knowledge of an exact solution p† .

For non-linear problems, additional conditions about the stopping rule have to be imposed
to guarantee convergence rates, Engl et al. (1996). The inequality known as the discrepancy
principle will be used to define the stopping index k(δ, yδ ) with
yδ − F [ pk(δ,yδ ) ] ≤ μδ < yδ − F [ pδk ] 0 ≤ k ≤ k(δ, yδ ) (18)
if the parameter μ is chosen subject to the constraint
1+η
μ>2 (19)
1 − 2η
being η < 1/2. The proof that this stopping rule regularizes the Landweber iteration method
can be found in Engl et al. (1996).
Once the convergence criteria and the stopping rule have been established, the iterative
procedure is applied to update the estimate of the solution in practical situations. It can be
seen that the update term in the Landweber iterative method, see Eq. (17), is the negative
gradient at p = pδk−1 of the functional
J = F [ p] − yδ (20)
that is the misfit between the model and the measured data. Solving the inverse problem is
hence equivalent to minimize this functional.
The solution of the forward problem, F, is then the solution of the system of Eqs. (1)-(4) given
the boundary conditions in the surface of the soil, Γ, and a distribution of soil parameters
p within the volume Ω. According to Eq. (1), p corresponds to the value of the thermal
diffusivity at every point within the soil volume, α = α( x, y, z). Given the 3D nature of the
model, the application of the operator F to the set of parameters, p, produces a 3D solution
where the distribution of temperatures both on the surface and within the soil volume is
calculated. For the sake of clarity we denote as F [ p] the distribution of the temperatures on
the surface of the soil, which is compared with the temperature distribution acquired by the
IR camera, our noisy data yd . The reconstruction of the internal composition of the soil can be
viewed as an inverse problem of this,that can be formulated as: given the boundary conditions of
the system and a distribution of measurements y on Γ, derive the soil parameter distribution, p, within
Ω, that is, p = F −1 [y].
Now we will introduce the procedure for the non-destructive inspection of soils for the
identification and classification of mines and mine-like targets on infrared images based on the
Heat Transfer
Detection 299
solution of the heat equation and the use of inverse problems techniques, López (2003); López
et al. (2009; 2004). The process starts with the acquisition of a sequence of infrared images
of the surface of the soil under known heating and atmospheric conditions. As explained
before, sunrise and sunset are the preferred times for detection. We will also assume that
a pre-processing stage is run on a conventional PC in order to align the images and map
grayscale colors to temperature values on the surface. Next, the soil inspection procedure
itself starts. First, we run a detection procedure, as will be explained in the following section,
to obtain the mask of potential targets. Then, a quasi-inverse process operator is used to
identify the presence of antipersonnel mines among the potential targets. For those targets
that failed to be classified as mines (and are therefore labeled as unknown), a full inverse
procedure to extract their thermal diffusivity will be run in order to gain information about
their nature. The overall detection process is summarized in Fig. 3, where the processes
that require the use of the 3D thermal model are indicated with an ellipse. The detection,
quasi-inverse and full-inverse procedures are based on the solution of the heat equation for
different soil configurations. As explained, this is a very time consuming task that makes the
whole algorithm inefficient for real on-field applications.
2.3.1 Target detection

The use of IR cameras taking images of the soil under inspection gives us the exact distribution
of temperatures on the surface. On the other hand, the thermal model described previously
and extensively validated with experimental data permits us to predict the thermal signature
of the soil under given conditions, López (2003); López et al. (2004). The detection of the
presence of potential targets on the soil is then made by comparing the measured IR images
with the expected thermal behavior of the soil given by the solution of the forward problem
under the assumption of absence of mines on the field, mathematically,
α( x, y, z) = αsoil , ∀ x, y, z. (21)
For this set of soil parameters, p, the application of the functional in Eq. (20) determines
the surface positions ( x, y) where the behavior is different from that expected under the
assumption of mine absence, therefore revealing the presence of unexpected objects on the
soil. These positions will be classified as potential targets, whereas the rest of the pixels
(those that follow the expected pure-soil behavior) will be automatically classified as soil. This
process is not trivial. The most straightforward approach, the thresholded detection, has the
drawback of setting the threshold, which will vary not only for different image sequences, but
it is also likely to depend on the particular frame of the sequence, and on the characteristics of
the measured data such as lighting conditions and the nature and duration of the heating. For
this reason, the use of a reconfigurable structure, capable of adapting to varying experimental
conditions was proposed on López (2003); López et al. (2004). In this work they demonstrated
that it is possible to reduce the time frame of analysis to roughly one hour around sunrise
as it is at this time when the maximum thermal contrast at the surface is expected. This
phenomena can be better appreciated in Fig. 4, where a sequence of IR images of a mine field
taken between 07:40 am and 08:40 am is shown. Taking into account the short time interval
we can consider that the properties of the soil remain unaltered and that there is no mass
transference process during the simulation. The output of the detection stage is a black and
white image with the mask of the potential targets.
30
Heat Transfer
Fig. 3. Structure of the approach used to detect buried landmines using infrared
thermography.
(a) 7:40 am (b) 08:00 am (c) 08:30 am (d) 09:00 am
Fig. 4. Measured IR images of a minefield at sunrise.
2.3.2 Quasi-inverse operator for the classification of the detected targets

In the previous section we dealt with the identification of the ( x, y) position of the potential
targets on the soil. In this section we will propose an operator for their classification
into either mine or soil categories; any target that fails to fit into these categories will be
Heat Transfer
Detection 31
11
classified as unknown (a procedure for the retrieval of further information about the nature
of the unknown targets will be explained in the next section). For the mine category, the
depth of burial will be also estimated. In general, this reconstruction is not possible unless
additional information on the solution is incorporated in the model by means of the so-called
regularization techniques Engl et al. (1996); Kirsch (1996). It is, however, possible to solve
the inverse problem without the explicit use of a regularization strategy under proper
initialization conditions and the use of iteration methods.
The iterative procedure is based on evaluating Eq. (20), which expresses the deviation between
the observed IR data, yδ , and the one given by the solution of the forward problem using
known parameter distributions, F [ p]. Therefore the heat equation needs to be solved for each
of these distributions during the time of analysis (usually one hour around sunrise). In the
case of mine targets, we will assume that their thermal evolution is driven by the thermal
properties of the explosive used, which is commonly TNT composition B-3 or, less frequently,
Tetryl. Our initial guess will be to assume that, (i) all the targets detected in the detection step
are mines, that is,
αtarget = αmine , (22)
and (ii) the possible depths of burial constitute a discrete set z ∈ Z̃ being,
Z̃ = {k Δz, ∀k = 0, 1, ..., d}, (23)
with Δz the discretization step and d Δz the a depth of burial at which is satisfied the
deep-ground condition, see Eq. (4). The situation k = 0 corresponds to surface-laid mines.
These two assumptions imply a reduction of the search space, therefore the quasi-inverse
nature of the classification effort that will either confirm or reject them. Let,
• {yδs }, s = 1, ..., S, be the acquired IR image sequence, being S the total number of frames.
• F [ p]s,k , the modeled temperature distribution on the soil surface at time s. F [ p]s,k is
estimated by considering that all the detected targets are landmines buried at the depth
given by index k in Eq. (23).
Note that, in the following, we will concentrate only on those areas of the image that were
marked as possible targets in the detection phase. The classification map for the detected
targets is obtained through the definition of a classification operator which includes the
following computations:
1. For each time instant s = 1, ..., S and burial depth k = 0, ..., d, an error map, Js,k = F [ p]s,k −
yδs , is estimated by evaluating Eq. (20) for each pixel position ( x, y);
2. For each time instant s, a global error map (Js ) and a global classification map (Υs ) are
estimated iteratively by comparing the error maps Js,k , k = 0, ..., d, as follows:
• Initialization step: For each pixel ( x, y), we set Js ( x, y) = ε (where ε is a predefined
threshold error value); and Υs ( x, y) = soil.
• Iterative update step: For each depth of burial, k, with k = 0, ..., d, Js ( x, y) =
min( Js,k ( x, y), Js ( x, y)) and Υs ( x, y) = argmink ( Js,k ( x, y), Js ( x, y)) (the category for
which the error is smaller, i.e. the depth of burial). If Js ( x, y) > ε then Υs ( x, y) is
set to Unknown.
3. Once Js and Υs have been obtained, we combine all these partial maps (Js , resp. Υs ) into
single ones (J, resp. Υ) in the following way:
32
Heat Transfer
• Υ: Pixels classified as mines at any processing step are kept in the final classification
map. For the others, we keep the category that appears more times.
• J( x, y) = maxs ( Js ( x, y)). This is a very conservative approach aiming at reducing the
number of false negatives (failure to detect a buried mine) even at the cost of increasing
the false alarm rate of the system.
• To find a trade-off between the accuracy of the classification and the number of false
alarms, we define a cutoff error, emax . If the entry on the error map, J for a pixel exceeds
emax , the pixel will be automatically assigned to the category of Unknown. emax is
estimated empirically, however it could be estimated taking into account the pixels
classified as non-mine based their temperature variance using bootstrap techniques,
Zoubir & Iskander (2004).
2.3.3 Full-inverse procedure for the classification of non-mine targets

In this case, no assumption about the nature of the targets found in the detection phase is
made, although the set of possible depths at which the targets can be placed is still bounded
by Eq. (23). Under these assumptions, Eq. (22) does not hold and αtarget is unknown and could
take any value depending on the nature of the object. For this reason, it is necessary in this
case to use a systematic approach for the minimization of the functional J, which implies the
calculation of the gradient ∂J/∂p.
Let us consider the existence of a buried target in a 3D soil volume, Ω, with an unknown
α = α(r ), r = ( x, y, z) ∈ Ω. The thermal experiment is the following: at time t = t0 , the solid
is subject to a prescribed flux, qnet (r , t), on its surface Γ, being Γ the portion of the surface
∂Ω accessible for measurements. We then measure the temperature response θ (r , t) at the
boundary r ∈ Γ, during the time interval [t0 , tf ]. We rewrite our 3D forward problem in
Eq. (1) as,
∂θ
−div{α(r ) gradθ } + ∂t = 0, r∈Ω (24a)
θ (r, t = t0 ) = θ0 , r∈Ω (24b)
∂
∂n θ (r , t ) = qnet (r , t) r ∈ Γ, t ∈ [t0 , tf ]. (24c)
We look at the reconstruction of α(r ) from the knowledge of the surface response of
temperature, yδ = θ (r , t), to prescribed flux applied on the boundary qnet (r , t). We call data
the pair (θ (r , t), qnet (r , t)). As mentioned before, this is an ill-posed problem. It is intuitive
that the data parameters (r , t) belong to a 3D subset, because r ∈ Γ and t ∈ [t0 , tf ]. This is
sufficient enough for the reconstruction of the function α(r ), defined in a 3D volume. Let us
now introduce the model problem as an initial guess p, such that p(r ) = α(r ) (known data on
the boundary), with the following governing equations and boundary conditions,
∂u
−div{ p(r ) gradu} + ∂t =0 r∈Ω (25a)
u(r, t = t0 ) = u0 , r∈Ω (25b)
∂
∂n u (r, t ) = qnet (r , t ) r ∈ Γ, t ∈ [t0 , tf ]. (25c)
The solution of Eq. (25) is a well-posed problem, as opposed to Eq. (24), and will be denoted
by u(r, t; p). Our aim will be to control p in such a way that the difference between the model
and the observed data tends to zero. This goal is quantified by an objective function J to be
Heat Transfer
Detection 33
13
minimized. The functional to be minimized is the L2 norm of the misfit between the model
and the observation given by,
tf
1
J (u( p)) ≡ u(r , t; p) − θ (r , t)2 dS dt. (26)
2 t0 Γ
This is a classic optimization problem which implies the calculation of the gradient of the
functional J. To this aim we will make use of the variational method. If we introduce the
notation,

< u, v >Ω = Ω u(r ) v(r ) dΩ

< qnet , v >Γ = Γ qnet (r ) v(r ) dS

a p < u, v >= Ω p gradu gradv dΩ,
then the model problem, Eq. (25), is equivalent to the variational problem,
tf tf tf
∂u
< , v >Ω dt + ( a p < u, v >) dt − < qnet , v >Γ dt = 0, ∀v (27)
t0 ∂t t0 t0
Eq. (27) can be considered as the constraints in the minimization problem, see Eq. (26).
Therefore, we can introduce the Lagrange multiplier λ(r, t) and define the Lagrangian L as,
tf
∂u
L(u, p, λ) ≡ J (u) + {< , λ >Ω + a p < u, λ > − < qnet , λ >Γ } dt. (28)
t0 ∂t
Note that L = J if u is the solution of the model problem, Eq. (25), since Eq.(27) holds for
any λ. Thus, the minimum of J under the constraints in Eq. (27) is the stationary point of
the Lagrangian L. Conversely, if δL = 0 for arbitrary δλ, u and p being held fixed, it follows
necessarily that Eq. (27) holds. We consider,
∂L ∂L
δL = δu + δp, (29)
∂u ∂p
where,
∂L
tf tf ∂u

∂u δu ≡ t0 ( u − θ, ∂u )Γ dt + t0 {( δ ∂t , λ )Ω + a p ( δu, λ )} dt (30a)
∂L
tf
∂p δp ≡ t0 Ω δp gradu gradλ dΩ δμ. (30b)
We can restrict the choice of λ such that

∂L
δu = 0. (31)
∂u
This condition can be written as,
tf tf
∂λ
< u − θ, δu >Γ dt + {− < δu, > + a p < δu, λ >} dt+ < δu, λ >Ω |ttf0 = 0, (32)
t0 t0 ∂t Ω
34
Heat Transfer
where δu( x, 0) = 0. The last term of (32) vanishes if we impose,
λ(r, t ≥ tf ) = 0. (33)
By doing so we obtain the equation for the adjoint field λ,

∂λ
−div{ p gradλ} − ∂t = 0, r∈Ω (34a)
λ(r, t ≥ tf ) = 0, r∈Ω (34b)
∂
∂n λ (r , t ) = θ − u, r ∈ Γ. (34c)
This is the so called back diffusion equation for the adjoint field, and it is also a well posed
problem. With this choice of the adjoint field λ(r, t), the variation ∂J becomes
tf
∂J = ∂p gradu gradλ dΩ dt. (35)
t0 Ω
It results from Eq. (35) that the derivative of J in the p(r ) direction is known explicitly by
solving two problems, the direct problem for the field u and the adjoint problem for the field
λ. That is, the Z-integral,
tf
∂J
≡Z= gradu gradλ dt. (36)
∂p t0
Solving Eq. (25) and Eq. (34), both of them well-posed forward problems, and using Eq. (36),
the expression of the update of Eq. (17)) can be calculated in a straightforward manner. With
respect to the number of iterations of the Landweber method, the selection of the stopping
criteria of the algorithm must be made according to the discrepancy principle in Eq. (18)). The
bigger the η, the lower the number of iterations is, and the higher the error is. The selection
of η for a particular application must then be a trade-off between computational time and
accuracy of the solution.
2.4 Estimation of the computational cost

The algorithm described above is based on iterative procedures involving multiple solutions
of the heat equation for different soil configurations. This constitutes a time consuming
process not feasible for its use on the field as the computational complexity of the FD method,
if N = n x · ny · nz is the total number of grid nodes, is O(N · IT), where IT is the number
of iterations. As an example, we consider the analysis of a piece of soil (αsoil = 6.4 · 10−7
m/s2 ) of moderate dimensions of 1m×1m with a shallowly buried mine (αmine = 2.64 · 10−7
m/s2 ). Even if the depth resolution of IRT is barely 10-15 cm, the depth of analysis must be
set to at least 40-50 cm in order to apply the boundary condition in Eq. (4). Using a uniform
spatial discretization of Δx = Δy = Δz = 0.8 cm and assuming a temporal discretization step
of Δt = 6.25 s (F0 = 0.06), for a typical example the simulation of the behavior of the soil
during one hour using C++ (optimized for speed using O2 flag from Microsoft Visual C++
compiler) on a Intel Core2Duo 2.8GHz takes 30 seconds if single precision arithmetic is used to
represent the temperatures. Taking into account that the proposed inverse procedure requires
the solution of the model for multiple soil configurations, the total computing time assuming
that only 100 iterations are needed (a soft approach) will add up to 50 minutes. As this
jeopardizes its use for field experiments we have developed a hardware implementation of a
Heat Transfer
Detection 35
15

Fig. 5. GPU internal structure and memory hierarchy.
heat equation solver. In Pardo et al. (2009; 2010) we presented an FPGA-based implementation
of such a solver. However, the main drawback of an FPGA implementation is the requirement
of the system in terms of memory. The FPGA has a little amount of distributed memory
and the FPGA’s logic blocks can also be configured to behave like memory, however this is
an inefficient way of FPGA using. Some vendors offer cards where external memory and
FPGA are integrated on the same board, allowing to use the FPGA to deal with processing
issues. However, these are expensive solutions. GPUs offers a structure which perfectly
fits with the proposed problem and they have the advantage of being cheaper than FPGAs.
GPUs are present in all computers and therefore we avoid the necessity of having a dedicated
and expensive hardware to deal with our problem. Moreover, the GPU implementation is
hardware independent, in the sense that it can be used on GPUs from NVIDIA with none or
little changes, depending on GPU’s computing capabilities.
3. GPU thermal model implementation

The system that solves the thermal model using the explicit FD method was implemented
using CUDA language, NVIDIA (2010), and projected in a GPU from NVIDIA. The computing
structure of GPUs makes them a suitable candidate to implement algorithms requiring high
computing power. First we will introduce GPU characteristics and some basics abouts its
programming mode. Then, we will present the proposed GPU implementation that simulates
the thermal behavior of the soil and that speeds the computations up compared to a personnel
computer.
3.1 GPU structure

GPUs are made up of several multiprocessors that can perform parallel processing data, which
makes them suitable for processing in systems where it can be split up in independent portions
and processed independently. The structure of such a GPU can be seen in Fig. 5. The GPU is
made up of several multiprocessors, labeled as MP1 ... MPN in Fig. 5. Moreover, inside each
multiprocessor there are several cores, labeled as C1 ... CM in Fig. 5. One important issue of
GPU programming concerns to the use of the different memories available in the GPU, see
Fig. 5. The Global Memory is available to all multiprocessors and cores, whereas the Shared
Memory inside each multiprocessor is only available to the corresponding multiprocessor’s
cores. Additionally, each core has its own and private memory space. One key aspect of
a GPU-based system is the memory data organization and access, as they can impose a
36
Heat Transfer

Fig. 6. Structure of threads hierarchy in a GPU (reprinted from NVIDIA (2010)).
bottleneck in the system performance. The global memory has an access latency two orders
of magnitude higher than the access to the shared memory. Thus, it is important to minimize
the use of global memory and maximize, as far as possible, the use of shared memory because
this will increase the performance of the system.
Once the structure of the GPU has been briefly described we will introduce the basic aspects of
GPU programming required to understand the structure of the proposed system . Functions
in CUDA are called kernels and each kernel can be executed in parallel by several threads1 , as
contrary to ordinary C/C++ functions that can only be executed by one processor. A kernel
is not executed as a single thread, but it is executed as a block of threads, each of them
processing the same function on different data, following a single-program multiple data
(SPMD) computing model. Each thread inside the block has a 1D, 2D or 3D identifier (ID),
depending on the applications, which distinguishes the concrete thread, to compute elements
from a vector, matrix or volume of data. All the threads of a block are executed on the same
multiprocessor and therefore they must fit within the available resources. This sets a limit
on the maximum threads per block, which is limited to 512 in current GPUs. To avoid this
limitation a kernel can be executed in several blocks of threads, which are organized as 1D or
2D groups of threads. The only requirement concerning the block of threads is that they must
1 The thread is the basic element of processing

Heat Transfer
Detection 37
17
CUDA cores 128

CUDA Multiprocessors 8
Graphics Clock 738 MHz
Processor Clock 1836 MHz
Global Memory 512 MB
Memory Clock 1100 MHz
Memory Bandwidth 70.4 GB/s
Table 1. GTS 250 NVIDIA GPU main characteristics.

Fig. 7. Temperatures updating scheme on the GPU.
be independent from each other. Fig. 6 shows threads’ hierarchy and its organization in the
GPU.
3.2 GPU implementation of heat equation solver

GPU’s structure fits perfectly our problem, where the full data can be split up in independent
blocks that can be process the data in parallel. Each multiprocessor can work with a portion
of grid’s nodes increasing the performance of the system. The GPU used in this work was a
GTS-250 from NVIDIA (cost around 250e- 300 $), whose characteristics are summarized in
Table 1.
As was pointed, one of the main important aspects in an efficient CUDA-based system is the
correct management of the memory to reduce the access to the global memory. To this aim the
full grid of points, see Fig. 2(a), was divided into volume slices of size size_blockx × size_blocky,
where the nodes’ temperature of each slice is computed in a block of threads, see Fig. 7. Each
thread of the block is responsible for updating the temperature of the nodes with the same
(x,y) coordinates within the considered piece of soil. The threads advance as a wavefront,
updating the nodes’ temperature starting from the superficial layers to the inside of the soil,
see Fig 7. It can be noted that there are overlapping areas between different blocks of threads,
labeled as boundary nodes and indicated in grey in Fig. 7, which must be taking into account to
compute only once the new temperature value.
Concerning the memory usage, the initial temperatures are stored in the global memory, and
they have been transferred from the HOST memory to GPU global memory prior to the
computation of the new temperatures. The temperatures are duplicated in the memory, as
during one iteration we need to use one location to read temperatures and the other to write
38
Heat Transfer

Fig. 8. Data memory transferences during the updating process.
the updated values and in the following iteration the roles are interchanged. The remainder
constant values needed in the computations, such as F0 and values related to the boundary
conditions, see Eq. (14), are also stored in the global memory. The access to the global memory
should be minimized to increase the speed of the computations, because the global memory
has a high latency access. Thus, we use, during the updating process, the shared memory of
the multiprocessors to accelerate the access to the data. The memory operations are shown in
Fig. 8 where we can see the data transferences between the different memories of the GPU.
In Fig. 8 we will consider the temperature updating process from nodes in Layer K. In STEP
1 the temperatures of Layer K-1 nodes are stored in both the multiprocessor’s shared memory
and in the cores’ local memory. Moreover, nodes’ temperatures from Layer k are read from the
global memory and stored in the local cores’ memory . During STEP 2 the nodes’ temperatures
from Layer K replace those from Layer K-1 in the shared memory, at the same time, the nodes’
temperatures from Layer K+1 are read from global memory and stored in cores’ local memory.
In STEP 3 all data required to perform Layer K nodes’ temperature updating is available on
the local memory and shared memory. The same temperature of a Layer K is required to
update the temperature of several nodes (the node itself and its north, south, west and east
neighbors). If all nodes had to access global memory to read these values the process would
be slowed, however once they are read from cores’ local memory they are transferred to the
shared memory, where they are available to all threads of the block, thus reducing the time
access to the data. Once a thread has updated the temperature of a node, it uploads to the
main memory the updated value and it continues computing the following temperature node
updating. There is a synchronization process when a thread finishes one node’s temperature
updating because we must ensure that prior to continue with a node of the following layer all
threads have finished the temperatures updating of the current layer.
4. Results
In this section we will introduce the results of the complete system. We divide this section
into two main topics. On the one hand the results of the detection algorithm are shown for
a scenario from the TNO Physics and Electronics Laboratory, Jong et al. (1999). On the other
hand, we will show the performance of the GPU implementation and how it improves the
usability of the detection system reducing the processing time.
4.1 Landmine detection algorithm

Next, we will show the result of the previously described detection algorithm to images
acquired in a real test field. The scenario considered corresponds to the sand lane of the
Heat Transfer
Detection 39
19

Fig. 9. Ground truth of the test field used during the experiment corresponding to a sand
lane with different types of surrogated mines and non-mine targets.
Symbol Category Total

• Surface mine 14
◦ Mine at 1 cm 9
Mine at 6 cm 9
⊕ Mine at 10cm 1
Mine at 15cm 1
Undefined test object 5
Shell marker 4
Table 2. Symbols used to represent the different categories of targets present in the test field
in Fig. 9.
test facilities of the TNO Physics and Electronics Laboratory. For the experimental setup
considered, the measured thermal diffusivity of the soil was αsand = 6 × 10−7 m2 /s. With
respect to the test mine targets present, they are surrogated mines and most of them have been
built at TNO-FEL. In all test mines the same substitute for the explosive has been used, RTV,
having the same relevant properties as the real explosive and particularly αRTV = 1.13 × 10−7
m2 /s. Fig. 9 shows a sample image of the sand lane acquired with the IR sensor and the
position of the different targets considered. Table 2 summarizes the symbols used for
the different categories of targets present. The total number of targets is 43, 34 of which
correspond to landmines. The remaining nine targets are five undefined test objects and four
shells used as markers. We will concentrate on the results of the quasi-inverse and full-inverse
procedures for the classification of mines and non-mine targets respectively.
40
Heat Transfer
Location Detected and classified Total

Surface 12 12
Buried at 1 cm 6 6
Buried at 6 cm 5 6
Buried at 10 cm 1 1
Buried at 15 cm 0 1
Table 3. Summary of the mines correctly detected and classified after the application of the
quasi-inverse operator.
4.1.1 Results of the quasi-inverse operator for the classification of mine targets
The quasi-inverse operator classifies the detected targets as Mine or Unknown. Moreover,
for the Mine class, sub-categories corresponding to their depth of burial are produced. For
the experimental setup in Fig. 9, the results of the application of the quasi-inverse operator
are summarized in Table 3, showing the distribution of mine targets correctly detected and
classified according to their depth of burial with emax = 2.6. As can be seen, all the mine
targets on the surface or at 1cm depth were correctly detected and classified, and so were five
out of six of the mines buried at 6 cm. The results for depths of 10 and 15 cm are not conclusive
since only one of each is present, but a degraded performance of the quasi-inverse operator
with depth is to be expected.
The performance of the classification operator is evaluated by making use of two properties,
sensitivity and specificity, Hanley & McNeil (1982),
TP TN
Sensitivity = Specificity = (37)
TP + FN TN + FP
being TP the number of true positives, TN the number of true negatives, FP the number of
false positives and FN the number of false negatives. The performance depends strongly on
the election of emax , being a trade-off between sensitivity and specificity, i.e. , between the
number of mine targets correctly classified and the number of false alarms. In humanitarian
operations, the stress is put on the correct location of mines, while reducing the number
of false alarms, although highly desirable, is a secondary goal. With respect to the global
performance, it is clearly a function of the particular value of emax . For emax = 2.6 we find
that 24 mines were correctly detected and classified over a total of 26.At the same time, the
number of false positives is 13 compared to the 27 after the application of the detection stage
alone.
4.1.2 Results of the full-inverse approach for the classification of non-mine targets
Now, we will illustrate the process of estimating the thermal parameters of non-mine targets
making use of the full inverse process previously described. To this aim, we will consider the
test object V82 present on the minefield (see Fig. 9). This target was classified as unknown by
the quasi-inverse operator and we now aim to infer what type of object it is by estimating its
thermal properties. If we estimate the measurement error to be δ = 0.3◦ C and setting η = 0.2,
we have,
1+η
μ>2 = 4. (38)
1 − 2η
Heat Transfer
Detection 41
21
7 −5
x 10
1.8
6 1.6
1.4
5
1.2
α (m2/s)
Error
1
4
0.8
3 0.6
0.4
2
0.2
1 0
0 50 100 150 200 250 0 50 100 150 200 250
Number of Iterations Number of iterations
(a) (b)
Fig. 10. Test target V82: (a) Evolution of the error in the estimation of α; (b) Evolution of the
value of α during the inverse problem procedure.
For μ = 4.1, the discrepancy principle determines the stopping rule as,
yδ − F [ pδk(δ,yδ ) ] ≤ 1.23 < yδ − F [ pδk ], 0 ≤ k ≤ k(δ, yδ ) (39)
The evolution of the error for the Landweber iteration method is shown in figure 10(a). The
stopping criteria in (39) corresponds to a number of iterations of the algorithm Nit = 230.
Figure 10(b) shows the evolution of the estimation of the α parameter in this case. The final
result obtained for the non-mine target V82 is,
αV82 = 170 × 10−7 m2 /s.
This value is two orders of magnitude bigger than that of the sand (αsand ≈ 6 × 10−7 m2 /s)
which coincides with typical values of the thermal diffusivity of metallic solids.
4.2 GPU heat equation solver

We will now introduce the results obtained with the GPU implementation in terms of system
throughput. As was pointed in the introduction a NVIDIA GTS 250 GPU, a low-cost GPU,
was used to perform the comparison between a purely CPU implementation ( Core2Duo 2.8
GHz implementation in C++) of the heat equation solver and a GPU implementation. One
of the first issues is to think about the blocks threads’ distribution and partitioning. The full
volume of nodes which form the grid of points must be divided into blocks of threads, each
of which is responsible for the nodes’ temperature updating. The idea can be seen in Fig. 7,
where the volume has been divided into blocks of threads of size size_blockx×size_blocky which
are sent to the MP of th GPU, in this case a 1D array of blocks is shown for the shake of
clarity. Fig. 11(a) shows the performance of the GPU for various block sizes, where we have
chosen size_blockx = size_blocky, and a volume of 800×800×50 nodes. The results from these
simulations can be seen in Table 4, where the speedup is compared to a purely CPU simulation
of the full volume. It can be noted that the throughput of the system raises up as the size of
the block is increased. This is due to the fact that when small blocks are used there are a lot
42
Heat Transfer

(a) Throughput of GPU heat equation solver for different blocks’ size.

(b) Throughput of GPU and CPU heat equation solvers
implementations for different volume of simulated points. In
the GPU implementation the block grids’ size was set to 16×16.
Fig. 11. Performance results comparing GPU and CPU throughputs for different setups.
of such small blocks spread over all MP, and therefore there will be a long cue of pending
blocks to be processed. On the contrary, if the size of the blocks is increased we will have less
blocks and the cue of pending blocks to be processed by the MP will be reduced. There is a
limit, imposed by GPU’s structure, given by the maximum number of threads that a MP can
process (512). It can be seen from Fig. 11(a) and Table 4 that the throughput of the system is
increased one order of magnitude when we go from 2×2 to 16×16 blocks of threads. Thus
in the following simulations we will used this block’s size for GPU simulations. Fig. 11(b)
shows the GPU and CPU throughput for different nodes volumes, the data can be seen in
Table 5. It can be noted how the performance of the GPU grows up one order of magnitude
when the size of the volume is increased. This is due to the fact that for small volume of nodes
not all GPU’s resources are being used, whereas for big enough size volumes the inherent
parallelism of the GPU increases the throughput of the system. It is obvious that for very
big volumes the throughput will be low because there will be a cue of pending blocks to be
processed which degrades the throughput of the system (note the reduction of the throughput
for the 1024×1024×50 volume).
Heat Transfer
Detection 43
23
block_dimx × block_dimy GPU throughput (Mpoints/s) GPU time (s) Speedup

2×2 120.0 25.6 3.1
4×4 336.5 9.1 8.7
8×8 634.3 4.8 16.7
16×16 1400.0 2 40
Table 4. GPU throughput for a given volume of points and varying block thread’s
dimensions.
Volume GPU throughput (Mpoints/s) CPU throughput (Mpoint/s) Speedup

32×32×50 332 39.3 8.4
64×64×50 1289.33 39.3 32.6
256×256×50 1476.67 39.3 37.0
512×512×50 1496.02 39.3 39.3
1024×1024×50 1402.8 39.3 36.7
Table 5. Throughput and CPU and GPU solvers of the heat equation solver for different size
volumes.
5. Conclusions
In this chapter, two inverse procedures for the inspection of soils by non-invasive means with
application in antipersonnel mines detection have been presented. The first quasi-inverse
procedure aims at the detection of surface-laid and shallowly buried mines, giving an
estimation of their depth of burial that will be of critical importance during the removal stage.
In the second approach, a full inverse procedure for the identification of the thermal properties
of other objects present on the soil was presented. Both procedures need the recursive solution
of the heat equation problem for different soil configurations, which constitutes a very time
consuming task on a conventional computer. The efficient solution of the aforementioned
procedures is successfully solved using a heat equation solver accelerator based on the use of
GPUs, obtaining speed-up factors over 40. The speedup obtained with the proposed system
with respect to nowadays computers, together with its low-cost and portability justifies the
implementation as it permits its use on the field during demining operations.
6. References
Bach, P., Toumeur, P. L., Poumarkde, B. & Bretteand, M. (1996). Neutron activation and
analysis, EUREL International Conference Detection of Abandoned Landmines, Vol. 431,
pp. 58–61.
Bejan, A. (1993). Heat Transfer, John Wiley & Sons, Inc.
Cameron, M. & Lawson, R. (1998). To Walk Without Fear: The Global Movement to Ban Landmines,
Toronto: Oxford University Press.
Durbano, J., Ortiz, F., Humphrey, J. R., Curt, P. & Prather, D. (2004). Fpga-based acceleration
of the 3d finite-difference time-domain method, Proceedings of the 12th annual IEEE
symposium on Field-Programmable Custom Computing Machines, pp. 156–163.
44
Heat Transfer
Engl, H. W., Hanke, M. & Neubauer, A. (1996). Regularization of Inverse problems, Kluwer
Academic Publishers.
England, A., Galantowiz, J. & Schretter, M. (1992). The radiobrightness thermal inertia
measure of soil moisture, IEEE Transactions on Geoscience and Remote Sensing
30(1): 132–139.
England, A. W. (1990). Radiobrigthness of diurnally heated, freezing soil, IEEE Transactions on
Geoscience and Remote Sensing 28(4): 464–476.
Englelbeen, A. (1998). Nuclear quadrupole resonance mine detection, CLAWAR’98,
pp. 249–253.
Furuta, K. & Ishikawa, J. (eds) (2009). Anti-personnel landmine detection for humanitarian
demining: the current situation and future direction for Japanese research and development,
Springer-Verlag.
Gros, B. & Bruschini, C. (1998). A survey on sensor technology for landmine detection, Journal
of Humanitarian Demining pp. 172–187.
Hanley, J. & McNeil, B. (1982). The meaning and use of the area under a receiver operating
characteristic (ROC) curve, Radiology 143: 29–36.
Horowitz, P. (1996). New technological approaches to humanitarian demining, Technical Report
JSR-96-115, JASON MITRE.
Hwu, W. W., Keutzer, K. & Mattson, T. (2008). The concurrency challenge, IEEE Design & Test
of Computers 25(4): 312 – 320.
ICBL (2006). Landmine Monitor Report 2006, International campaign to can landmines (ICBL).
Incropera, F. & DeWitt, D. (2004). Introduction to Heat Transfer, 4th edn, John Wiley & Sons.
Jankowski, P., Mercado, A. & Hallowell, S. (1992). FAA explosive vapor/particle detection
technology, Applications of Signal and Image Processing in Explosives Detection Systems,
Vol. 1824, pp. 13–27.
Jong, W., Lensen, H. & Janssen, H. (1999). Sophisticated test facilities to detect land mines,
Detection and Remediation Technologies for Mines and Minelike Targets IV, Vol. 3710 of
Proceedings of the SPIE, pp. 1409–1418.
Kahle, A. B. (1977). A simple thermal model of the earth’s surface for geologic mapping by
remote sensing, Journal of Geophysical Research 82: 1673–1680.
Khanafer, K. & Vafai, K. (2002). Thermal analysis of buried land mines over a diurnal cycle,
IEEE Transactions on Geoscience and Remote Sensing 40(2): 461–473.
Kirsch, A. (1996). An introduction to the Mathematical Theory of Inverse problems, Vol. 120 of
Applied mathematical sciences, Springer-Verlag, New York.
Larsson, C. & Abrahamsson, S. (1993). Radar, multispectral and biosensor techniques for mine
detection, Symposium on Anti-Personnel Mines, pp. 179–202.
Liou, Y. & England, A. (1998). A land surface process/radiobrightness model with couple heat
and moisture transport for freezing soils, IEEE Transactions on Geoscience and Remote
Sensing 36(2): 669–677.
Lockwood, G., Shope, S., Bishop, L., Selph, M. & Jojola, J. (1997). Mine detection using
backscatered x-ray imaging of antitank and antipersonnel mines, Detection and
Remediation Technologies for Mines and Minelike Targets II 3079: 408–417.
López, P. (2003). Detection of Landmines from Measured Infrared Images using Thermal Modeling of
the Soil, PhD thesis, Universidad de Santiago de Compostela.
Heat Transfer
Detection 45
25
López, P., Pardo, F., Sahli, H. & Cabello, D. (2009). Non-destructive soil inspection using an
efficient 3d softwareâĂŞhardware heat equation solver, Inverse Problems in Science and
Engineering 6(17): 755–775.
López, P., van Kempen, L., Sahli, H. & Cabello, D. (2004). Improved thermal analysis of buried
landmines, IEEE Transactions Geoscience and Remote Sensing 42(9): 1955–1964.
Maksymonko, G. B. & Le, N. (1999). Performance comparison of standoff minefield detection
algorithms using thermal IR image data, Vol. 3710, SPIE, pp. 852–863.
URL: http://link.aip.org/link/?PSI/3710/852/1
NVIDIA (2010). NVIDIA CUDA C Programming Guide 3.1, NVIDIA Corporation Technical
Staff.
Ottawa (1997). Convention on the prohibition of the use, stockpiling, production and transfer
of anti-personnel mines and on their destruction.
Pardo, F., López, P., Cabello, D. & Balsi, M. (2009). Effcient software-hardware 3d heat
equation solver with applications on the non-destructive evaluation of minefields,
Computers & Geoscience 35: 2239–2249.
Pardo, F., López, P., Cabello, D. & Balsi, M. (2010). Fpga computation of the 3d heat equation,
Computational Geoscience 14: 649–664.
Placidi, P., Verducci, L., Matrella, G., Roselli, L. & Ciampiolini, P. (2002). A custom VLSI
architecture for the solution of FDTD equations, IEICE Transactions on Electronics
E85-C: 572–577.
Pregowski, P., Walczack, W. & Lamorski, K. (2000). Buried mine and soil temperature
prediction by numerical model, Proceedings of the SPIE, Detection and Remediation
Technologies for Mines and Minelike Targets V, Vol. 4038, pp. 1392–1403.
Robledoa, L., Carrascoa, M. & Merya, D. (2009). A survey of land mine detection technology,
International Journal of Remote Sensing 30(9): 2399–2410.
Sabatier, J. & Xiang, N. (2001). An investigation of a system that uses acoustic seismic coupling
to detect buried anti-tank mines, IEEE Transactions on Geoscience and Remote Sensing
39(6): 1146–1154.
Schneider, R., Turner, L. & Okoniewski, M. (2002). Application of FPGA technology to
accelerate the Finite-Difference Time-Domain (FD-TD) method, Proceedings of the 10th
ACM/SIGDA International Symposium on Field-Programmable Gate Arrays, pp. 97–105.
Siegel, R. (2002). Land mine detection, IEEE Instrumentation & Measurement Magazine
pp. 22–28.
Thanh, N., Hao, D. N. & Sahli, H. (2009). Augmented Vision Perception in Infrared Algorithms
and Applied Systems, Advances in Pattern Recognitiob, SpringerLink, chapter Infrared
Thermography for Land Mine Detection, p. 471.
Thanh, N., Sahli, H. & Hao, D. (2007). Finite-Difference methods and validity of a thermal
model for landmine detection with soil propierty estimation, IEEE Transactions on
Geoscience and Remote Sensing (4): 656–674.
Thanh, N., Sahli, H. & Hao, D. (2008). Infrared thermography for buried landmine
detection: inverse problem settin, IEEE Transactions on Geoscience and Remote Sensing
(12): 3987–4004.
Vines, A. & Thompson, H. (1999). Thompson, beyond the landmine ban: Eradicating a lethal
legacy, Technical report, Research Institute for the Study of Conflict and Terrorism.
46
Heat Transfer
Wang, T. & Chen, C. (2002). 3-D thermal-ADI: A linear-time chip level transient thermal
simulator, IEEE Transactions on Computer-Aided Design of Integrated Circuits and
Systems 21(12): 1434–1445.
Zoubir, A. M. & Iskander, R. (2004). Bootstrap Techniques for Signal Processing, Cambridge
University Press.
3
The Heat Transfer Enhancement Analysis and

Experimental Investigation of Non-Uniform
Cross-Section Channel SEMOS Heat Pipe
Shang Fu-Min1, Liu Jian-Hong1 and Liu Deng-Ying2
1Changchun Institute of Technology, Changchun 130012, Jilin,
2Institute of Engineering Thermo physics, Chinese Academy of Sciences, Beijing 100080,
China
1. Introduction
Along with the development of modern high-technology, the heat generation of many
equipments that were heated in unit area becomes more and more, usually it is over
100W/cm2, even 103~104W/cm2, such as super power integrate circuit, high temperature
battery for electricity generation, nuclear reactor, high power mini-machinery, mini-sized
steam turbine, superconductive electricity generation and distribution devices and airship.
Due to the trend of high rate heat generation, a new cooling technology is needed to solve
the problem between design flexibility, needed space and heat-dissipating performance.
A new period of heat pipe began with Pulsation Heat Pipe, invented by Japanese Doctor H.
Akachi in 1994, later is called Self-Exciting Mode Oscillating-Flow Heat Pipe[1, 2] (SEMOS
Heat Pipe). Although it is just known less than sixteen years, the SEMOS Heat Pipe is
drawing worldwide attention because of its immeasurable application potency. A lot of
research about the SEMOS Heat Pipe is doing in Japan, U.S.A, Germany, Russia, Ukraine
and China[3-5].
Notwithstanding all the research done is to understand the working principles of SEMOS
Heat Pipe because it is still at its initial stage, this new high effective heat transfer device is
catching attention on heat transfer due to phase transition all over the world. Thus the
contradiction between cooling device downsizing and heat generation increasing is still the
major technical problem to improve and bring the heat-dissipating performance to a higher
lever for all of the researchers along with the development of new high-technology.
2. Heat transfer characteristics of SEMOS Heat Pipe

According to the working principles of SEMOS Heat Pipe, there are two basic methods to
improve the heat transfer performance, one is to enhance the thermal conduction between
the interface of the pipe and the working gas-liquid medium in the pipe, the other method is
to increase the surge frequency and circulating power. Actually the thermal conduction of
the SEMOS Heat Pipe between the pipe surface and the working liquid in the pipe is the
non-steady phase transition heat transfer between the biphasic fluid and the interface of
pipe. The first method to enhance the thermal conductive performance is to improve the
liquid evaporation, the clotted phases transforming frequency, and intensity of phase
transforming. Another method is to increase the heat transfer rate between the working
liquid and the working surface. What needed to do to enhance the surge frequency and the
reliable circulating power is to increase the difference in temperatures between the hot and
cold liquid via enhancing the pulsing process inside the pipe. Obviously the two enhancing
methods discussed above could supplement each other.
On the principle of field coordination heat transfer enhancement[6, 7], which was put forward
by Guo Zengyuan academician, the heat transfer rate increases as field coordination
coefficient between velocity vector field and temperature grads field increases. Although the
convection heat transfer theory of single phase has been demonstrated, there still exists the
problem about the heat transfer during the phases transforming between two phases,
especially in a limited heat-dissipating space. Thus it needs further study on that if the
coordination theory could be used universally.
This experiment to enhance the heat transfer rate of SEMOS Heat Pipe is to validate the
application of field cooperation theory on the heat transfer field with phases transforming.
The SEMOS Heat Pipe with non-uniform cross-section is used for this experiment so as to
improve surging frequency and circulating power.
Water out
Water in
Surveying
point
Fig. 1. Experimental parts of variable cross-section SEMOS Heat Pipe

The Heat Transfer Enhancement Analysis and Experimental
Investigation of Non-Uniform Cross-Section Channel SEMOS Heat Pipe 49
3. Experiment set up and method

The object for this experiment is SEMOS Heat Pipes with closed loop, one is a SEMOS Heat
Pipe with a uniform cross-section of 3mm inner diameter, the other heat pipe based on the
uniform one with elliptic non-uniform cross-section is consisted of vertically intervened
heating section and insulating section of the pipes. As shown in figure 1 the material of the
heat pipe is brass, the working fluid in the pipe is distilled water with high purity, the
amount of working is ϕ =42%, the obliquity of the heat pipe is θ =550, the pressure in the
heat pipe is p =1.8×10-3Pa.
Figure 2 shows the experiment set up for the thermal performance measurement consisting
of the main test apparatus, a laser supply and its cooler and its power supply system, a data
acquisition system combined with personal computer to show the data collected. As shown
in figure 2, the laser supply is consisted of eight passages Quantum Well Laser Diode
Arrays, the maximum output power of every single channel is 50W, the heating electrical
current range is 5~40A, the wave length of the laser is 94nm. The heat input comes from
continual laser heating while every single could work individually or together as heater,
and 20 K-type thermocouples were mounted on the surface at diameter of 1mm totaling 20
in number. Accordingly, the data acquisition frequency is 1/s based on the 20 data
collecting channels and the data acquisition precision is at ms.
8 7 6
1—Unit of refrigeration cycle, 2—Power supply, 3—Laser supply, 4—Experimental table of SEMOS
Heat Pipe, 5—Water tank, 6—Flow meter, 7—Data acquisition system, 8—Personal computer
Fig. 2. Experimental system of SEMOS Heat Pipe heat transfer enhancement
4. Experiment result and discussions

4.1 Compare and contrast heat pipes uniform with non-uniform cross-section
Judging from the experiments above mentioned to the heat pipes with uniform cross-section
and non-uniform cross-section, figure 3 shows the transfer efficiency under different heating
power (heat electrical current I). The transfer efficiency is given by the following equation.
Po = Gc p (T2 − T1 ) (1)
Where, G denotes the amount of the cooling water, kg/s; cp is the specific heat of water
under constant pressure, J/(kg·K). T2 and T1 denote the output and input temperature of
cooling water, K.
As can be seen from the figure, the transfer rate of the heat pipe with non-uniform cross-
section is lower than that of the heat pipe with uniform cross-section when the heating
electrical current is relatively low, while the rate of heat pipe with non-uniform cross-section
would exceed that of the heat pipe with uniform cross-section as the heating electrical
current of laser supply increases. It is suggested from the trend of the graph of transfer rate
Po~I that the more heat input the more of the difference of transfer rate between the heat
pipe with non-uniform cross-section and the heat pipe with uniform cross-section becomes.
The transfer rate of heat pipe with non-uniform cross-section is 13.6% higher than that of the
heat pipe with uniform cross-section at the maximum heat input in this experiment, the
heating electrical current of every channel is 23A that is about 25.5W.
110
uniform cross-section
100
non-uniform cross-section
90
80
70
60
P0/W
50
40
30
20
10
0
8 10 12 14 16 18 20 22 24
I/A
Fig. 3. Compare transferred power uniform with non-uniform cross-section heat pipes
4.2 Compare uniform and non-uniform cross-section with pure conductor equivalent
thermal conductivity
Figure 4 shows the relationship between heat input and equivalent weight heat transfer
coefficient under three different conditions, SEMOS Heat Pipes with non-uniform cross-
section and uniform cross-section and the heat pipe without working fluid that means pure
conductor. According to the different output and the difference in temperature of
corresponding hot and cold end of the heat pipe, the equation of heat transfer coefficient is
shown as below, while
Pol (2)
λd =
( Th − Tc F)
Where, Po denotes transfer rate, W. l is the distance between the hot and cold end, m. Th
and Tc Mean temperature of hot end and cold end, K. F Total heat transfer area, m2, total
circulate cross section area for heat transfer, total cross section area for pure conductor.
11000
10000
9000 pure conductor
8000
7000
λ d /W(m.K)-1
6000
5000
4000
3000
2000
1000
0
8 10 12 14 16 18 20 22 24
I/A
Fig. 4. Compare equivalent thermal conductivity in varying cases
It can be seen from the figure that the equivalent weight heat transfer coefficients under all
the three conditions increase as the heat input increases, while the coefficient of the non-
uniform heat pipe increased the most and that of the pure conductor heat pipe the lest.
When the heat input is low, the heat transfer coefficient of uniform cross-section heat pipe is
higher than that of the non-uniform one. As the heat input increases, the heat transfer
coefficient of the non-uniform cross-section heat pipe becomes higher that that of the
uniform one and the difference between them becomes bigger and bigger, so it is suggested
that under certain condition the heat transfer rate could be improved by changing the form
of the cross-section of the SEMOS Heat Pipe. The figure also shows that the function on heat
transferring of heat pipes especially SEMOS Heat Pipes is better than the function on heat
conducting of brass pipe. Under that heat input range, the equivalent weight heat transfer
coefficient of heat pipe is as 4~15 times as that of the pure conductor one, and this difference
becomes bigger as the heat input increases. At maximum heat input (23A for every single
channel) in this experiment, the equivalent weight heat transfer rate of non-uniform cross-
section heat pipe is 19% higher than that of uniform cross-section one, 14 times higher than
that of the pure conductor one.
4.3 Heat transfer performance of the heat pipe with non-uniform cross-section
Theoretically, there are two reasons to answer that why non-uniform heat pipe consisted of
the heating section and insulating section could improve heat transfer rate. One is that the
portrait eddy formed for the Non-uniform Cross-Section of cross and adds parts of the
velocity on the pipe surface vertically, that means the temperature grads field and velocity
vector field co-operate together, then the heat transfer rate could be improved. Meanwhile
the resistance on the working fluid increases because of the non-uniformity of the cross-
section, certain amount of the heat input is needed for the SEMOS Heat Pipe with non-
uniform cross-section to improve heat transfer performance in word and deed. Because the
advantage of the SEMOS Heat Pipe with non-uniform cross-section on improving heat
transferring could be demonstrated easily when the portrait eddy becomes stronger and
circulating power becomes declining, the SEMOS Heat Pipe with non-uniform cross-section
is more suitable for the high density working fluid.
On the other side, as shown in figure 5 it could be found by monitoring the temperature of
the cold and hot end surface that the hot end surface temperature surging amplitude of the
non-uniform heat pipe is lower than that of the uniform one, while the surging frequency is
higher. In SEMOS Heat Pipe with non-uniform cross-section the circulating power increases
because of the alternation of liquid evaporation and gas clot. The accelerated motion inside
the pipe because of unsteady expanding and contracting process along with the vertical
velocity on the interface could finally improve the heat transfer performance.
110
105
100
T/ (K-273)
95
90
85
0 100 200 300 400 500
Time/s
Fig. 5. Compare wave with hot end surface temperature

5. Conclusions
1. The precondition to improve heat transfer performance with non-uniform SEMOS Heat
Pipe is that heat input must be high enough to overcome the resistance brought by the
non-uniform cross-section.
2. The heat transfer rate of SEMOS Heat Pipes with non-uniform cross-section and
uniform cross-section is much higher than that of the pure conductor under same
experimental condition. The SEMOS Heat Pipe with non-uniform cross-section is more
suitable for high density heat current because it especially shows its advantage when
the heat input is high.
3. The surging frequency of SEMOS heat pipe with non-uniform cross-section becomes
higher than that of the SEMOS Heat Pipe with uniform cross-section as the heat input
increase.
4. It is confirmed in this experiment that field coordination heat transfer enhancement
theory suit for heat transfer due to phase transition of the SEMOS Heat Pipe, which is
an exemplification that this law could be used in the heat transfer domain of
heterogeneous fluid. Which is the base of the future research and development for new
SEMOS Heat Pipe.
6. Acknowledgement
This work was Supported by the Key Project of Chinese Ministry of Education (No: 210050).
7. Nomenclature
P heating power, W
G rate of mass flow, kg·s-1
T temperature, K
T mean temperature, K
I heating electrical current, A
F Total heat transfer area, m2
c specific heat at constant pressure, J·kg-1·K-1
l distance between the hot and cold end, m
λ heat transfer coefficient, W·m-1·K-1
Subscripts
p pressure
o out
d equivalent
h hot end
c cold end
8. References
[1] H. Akachi. Looped Capillary Tube Heat Pipe. Proceedings of 71th General Meeting
Conference of JSME, 1994, 3(9): 1-10
[2] H. Akachi, F. Polasek, P. Stulc. Pulsating Heat Pipe. Proceedings of the 5th International
Heat Pipe Symposium, Australia, 1996: 17-20
[3] Qu Wei, Ma Tongze. Experimental Investigation on Flow and Heat Transfer of Pulsating
Heat Pipe. Journal of Engineering Thermo physics, 2002, 23(5): 596-598
[4] Cui Xiaoyu, Weng Jianhua, M. Groll. Experimental Investigation of Heat Performance for
The Copper/Water Pulsating Heat Pipe. Journal of Engineering Thermo physics,
2003, 24(5): 864-866
[5] Yasushi KATO, Takao NAGASAKI and Yutaka ITO. Study on Looped Heat Pipe with
Non-uniform Cross section. Proceedings of the 40th Heat Transfer Conference.
Japan, 2003, 5: 313-314
[6] Guo Zengyuan. Physical Mechanism and Control of Heat Convection: Speed and
Temperature Field Synergy. Chinese Science Bulletin, 2000, 45(19): 2118-2122
[7] Meng Jian, Chen Zejing, Li Zhixin and Guo Zengyuan. Field Coordination Analysis and
Convection Heat Transfer Enhancement in Duct. Journal of Engineering Thermo
physics, 2003, 24(4): 652-654
4
Magneto Hydro-Dynamics and

Heat Transfer in Liquid Metal Flows
J. S. Rao and Hari Sankar
Altair Engineering India Pvt. Ltd., Bengaluru, 560103
India
1. Introduction
Liquid metals are considered to be the most promising coolants for high temperature
applications like nuclear fusion reactors because of the inherent high thermal diffusivity,
thermal conductivity and hence excellent heat transfer characteristics. The coolant used in
nuclear reactor should have high heat extraction rate. The high melting point and boiling
point which eliminates the possibility of local boiling makes liquid metals more attractive to
high temperature applications. The thermal entrance length of liquid metals are relatively
high leading to flow never reaching fully developed condition which is always
advantageous for heat transfer applications as the Nusselt number value is higher in a
developing flow than a fully developed flow. The molecular properties of liquid metals are
such that the thermal diffusion is faster than momentum diffusion having Prandtl number
<< 1. The thermal boundary layer for liquid metal flow is not only confined to the near wall
region but also extends to the turbulent core region which makes the turbulent structures
important in transfer of heat. As turbulence plays an important role in transfer of heat from
the viscous sub-layer to the core flow, it is necessary to maintain high turbulence to achieve
high convective heat transfer rates. It may also be noted that surrounding magnetic fields
reduce the turbulence and flow becomes more streamlined.
Lithium is the lightest of all metals and has the highest specific heat per unit mass. Lithium
is characterized by large thermal conductivity and thermal diffusivity, low viscosity, low
vapor pressure as shown in table 1. Lithium is the most promising coolant for
thermonuclear power installations. Tritium is a component of fusionable fuel. Tritium does
not occur in nature in large amounts and is unstable with its half life 12 years. Tritium can
be obtained from lithium with nuclear reactions in fission nuclear reactors or blankets of
fusion reactors. Thus, lithium provides raw nuclear fuel to implement fusion reaction.
Density, Dynamic Viscosity, Specific Heat, Thermal Conductivity,

kg/m3 kg/m-s KJ/kg-K W/m-K
497 0.0004 4167 50
Table 1. Physical properties of liquid lithium at 650 °C, Davidson (1968)
The plasma generated in the nuclear fusion reactors is confined using an intense magnetic
field created by a series of toroidal and poloidal magnets generating magnetic field intensity
values of the order 10 Tesla, Kirillov (1994) and is very high compared to the magnetic field
intensity of the earth which is of the order 10-4 Tesla, Roberts (1967). The electrical
conductivity of liquid metals is of the order 106 to 107 1/Ω-m, Davidson (1968). This
introduces the term magneto-hydrodynamics relevant to design of liquid metal flows
especially in the application as coolant and breeding material in Tritium breeders of nuclear
fusion reactors.
Magneto-hydro-dynamics involves study of magnetic fields and fluid flows and their
interaction. Alfven (1942) was the first to introduce the term Magneto-hydrodynamics,
though Hartmann and Lazarus (1937) earlier performed both theoretical and experimental
studies in Magneto-hydrodynamic flows in ducts. Moffatt (1967) showed analytically that
the turbulent velocity fluctuation in a flow is suppressed by application of a uniform
magnetic field. This chapter is dedicated to the study of how magnetic field affects the
convective heat transfer characteristics in liquid metal flows by affecting the following
mechanisms.
1. Mean velocity distribution in the flow
2. Velocity fluctuations in time and space
2. The governing equations of magneto-hydrodynamic flows

The hydrodynamic governing equations are the mass (continuity), momentum and energy
equations. They are
Continuity equation
∂ρ (1)
+ ∇ ⋅ ρV = 0
∂t
where ρ is the density of the fluid and V is the velocity represented by V = ui + vj + zk. The
first term on the left is the accumulation/unsteady term which can be neglected for steady
state condition and second term on the left is mass flux term.
Momentum equation
∂V 1
+ (V ⋅ ∇ )V = − ∇p + ν∇ 2V + S (2)
∂t ρ
where ν is the kinematic viscosity which is a molecular property of the fluid, p is the
pressure. The first term on the left is the unsteady term and the second term on the left is the
convection term. The first term on the right is the pressure force term; the second term on
the right is the diffusion term which is due to the viscous effects. The third term on the right
is the source term which includes other forces such as gravitational forces, electromagnetic
forces etc.
Energy equation
⎡ ∂T ⎤
ρC p ⎢ + (V ⋅ ∇ ) T ⎥ = k∇ 2T + q''' (3)
⎣ ∂t ⎦
where k is the thermal conductivity which is a molecular property of the fluid, T is the
temperature and Cp is the specific heat of the fluid. The first term on the left is the unsteady
term and the second term on the left is the convection term. The first term on the right is the
diffusion term. The third term on the right is the source term which includes volumetric
heat sources.
Magneto Hydro-Dynamics and Heat Transfer in Liquid Metal Flows 57
The equations given above are generic equations and what differentiates the equation when
used for laminar and turbulent flow regimes is the replacement of molecular viscosity, ν and
thermal conductivity, k by effective viscosity, ν eff and effective thermal conductivity, keff in
the momentum and energy equation when the flow is turbulent.
ν eff = ν m + ν τ (4)
keff = km + kτ (5)
where ν τ is the turbulent viscosity which is the increment in momentum transfer due to
diffusion created by turbulence and kτ is the turbulent thermal conductivity which is the
increment in heat transfer due to diffusion created by turbulence.
The effect of magnetic field will be included by addition of Lorentz force term j × B and force
due to electric charge ρcE as a source term in the momentum equation and inclusion of Joule
2
dissipation term j as a volumetric heat generation term in the energy equation. The effect
σ
of magnetic field on turbulence can be included by altering the turbulent viscosity and
turbulent thermal conductivity in the momentum and energy equation respectively.
The modified momentum equation including the magnetic effects is given by,
∂V 1 1 1
+ (V ⋅ ∇ )V = − ∇p + ν∇ 2V + ( j × B) + ( ρ cE) + S (6)
∂t ρ ρ ρ
The modified energy equation including the Joule dissipation is given by,
⎛ ∂T ⎞ j2
ρC p ⎜ + (V ⋅ ∇ )T ⎟ = k∇ 2T + + q ''' (7)
⎝ ∂t ⎠ σ
The inclusion of new variables j, E and B in the hydrodynamic equations requires closure
equations given using Maxwell’s equations
∇ ⋅ D = ρc (8)
∂B
= −∇ × E (9)
∂t
∇⋅B = 0 (10)
∂D
∇×H = j + (11)
∂t
and Ohms law
j = σ ( E + V × B) (12)
where D is the electric displacement.

The above equations (1), (6)-(12) can be used for a range of problems, but the equations are
complex and may not be fully modeled to capture the physics of liquid metal flows and heat
transfer.
The Magneto-hydro-dynamic approximation

In applications where the flow velocity is low as compared to the speed of light, the force
due to electric charge ρcE can be neglected when U << c based on
ρ cE U2
≈ (13)
j×B c2
where c is the speed of light. This approximation will eliminate the inclusion of the
term ∂D ∂t from equation (11) and equation (8) can be ignored when U << c.
Using B = μ * H and combining (11) and (12)
1
( ∇ × B) = ( E + V × B) (14)
σμ *
1
Considering ∇ × of equation (14) assuming η * = is constant, we have
σμ *
{ }
η * ∇ ( ∇ ⋅ B) − ∇ 2 B = ∇ × E + ∇ × (V × B) (15)
Using (9) and (10), (15) becomes
∂B
= ∇ × (V × B ) + η * ∇ 2 B (16)
∂t
Equation (16) is called the magnetic induction equation; the first term on the right is the
induction term and the second term the diffusive term. The induction term describes the
interaction of the field with flow. If Rem << 1 then the diffusion term is more significant than
the induction term and hence the induction equation, for Rem << 1 becomes
∂B
= η * ∇2B (17)
∂t
1
j= ( ∇ × B) (18)
μ*
A summary of equations used for liquid metal flows involving heat transfer in presence of
magnetic field is given below namely, mass, momentum, energy and magnetic induction.
∂ρ
+ ∇ ⋅ ρV = 0 (19)
∂t
∂V 1 1⎡ 1 ⎛ B2 ⎞ ⎤
+ (V ⋅ ∇ )V = − ∇p + ν∇ 2V + ⎢ B ⋅ ∇B − ∇ ⎜⎜ ⎟⎟ ⎥ + S
(20)
∂t ρ ρ ⎣⎢ μ * ⎝ 2 μ * ⎠ ⎦⎥
⎛ 1 ⎞
⎜ (∇ × B) ⎟
⎛ ∂T ⎞ μ *
ρC p ⎜ + (V ⋅ ∇ )T ⎟ = k∇ 2T + ⎝ ⎠ + q''' (21)
⎝ ∂t ⎠ σ
∂B
= η * ∇2B (22)
∂t
The total number of unknowns is 9 including three scalars of velocity, three scalars of
magnetic field, pressure, density and temperature and the number of equations is 8
including continuity, 3 momentum equations, energy and 3 induction equations. We need a
closure equation for the set of equations. If the flow is incompressible then we can treat
density as constant and hence removing the need of an extra equation. If the flow is
compressible then we can use an equation for pressure as a function of temperature and
density i.e., p = f (T , ρ ) for example like ideal gas law for gases.
3. Non-dimensional numbers
Reynolds Number: The non-dimensional number that gives the ratio of the inertial forces to
the viscous forces is Reynolds number defined by
Inertial Forces ρU0 L

Re = =
Viscous Forces μ
where ρ is the density of the fluid (kg/m3), L is the characteristic length (m), U0 is the
velocity (m/s) and μ is the dynamic viscosity of the fluid (kg/m-s). For liquid metal flows in
fusion reactors, the typical values encountered is of the order 104 to 105. The transition from
laminar to turbulent flows occurs at Re ≈ 2300.
Hartmann Number: The non-dimensional number that gives the ratio of the electromagnetic
forces to the viscous forces is Hartmann number defined by
1
⎛ Electromagnetic Forces ⎞ 2 σ
Ha = ⎜ ⎟ = B0 L νρ
⎝ Viscous Forces ⎠
where, B0 is the magnetic field intensity (Tesla), σ is the electrical conductivity of the fluid
(1/Ωm) and ν is the kinematic viscosity of the fluid (m2/s2). For liquid metal flows in fusion
reactors, the typical values encountered for Ha are of the order 2×104 and the effect of
viscosity and turbulence on the flow is negligible in those situations.
Stuart Number: The non-dimensional number that gives the ratio of the electromagnetic
forces to the inertia forces is Stuart number (also called Interaction Parameter N).
Ha 2 Electromagnetic Forces σ B02 L

N ≡ St = = =
Re Intertial Forces ρU0
Stuart number can be used to characterize the effect of magnetic field on turbulence as it
represents the ratio of large-eddy turnover time τ to the Joule time τm. Stuart number shows
the ability of the magnetic field to change the isotropic turbulent structures to anisotropic
two-dimensional turbulence. When N is small, the anisotropy induced by Joule dissipation
is negligible and when it is larger than a critical value Nc, Lorentz force will drive turbulence
to two-dimensional situation. The significance of St is found in recent days as the transition
of turbulent flow will happen in the range 101 to 102 as reported by Uda et al. (2001) and a
sudden increase of turbulence is noted to happen at Stuart number equal to 10. Another
interpretation of Stuart number can be that when N is small, the electromagnetic forces will
be weak and will not be able to affect the flow.
Magnetic Reynolds Number: The non-dimensional number that gives the ratio of the
momentum diffusivity to the electromagnetic diffusivity defined by
Momentum Advection U 0 L
Re m = =
Magnetic Diffusion 1
μ *σ
where, μ* is the fluid magnetic permeability (h/m). It characterizes the effect of induced
magnetic field on the resultant magnetic field due to the imposed and induced magnetic
fields. At Re m << 1 the imposed magnetic field will guide the flow and will be independent
of the induced magnetic field as it will be very small compared to the imposed magnetic
field. The Lorentz force will be a linear function of velocity when Re m << 1 and quasi-static
approximation can be applied. At Re m >> 1 the dependence between the magnetic field and
velocity is non-linear as the imposed magnetic field is a function of velocity.
Another interpretation of magnetic Reynolds number is based on the ratio of characteristic
time scale of diffusion of magnetic field to the time scale of turbulence defined by
Time scale of diffusion of magnetic field vL

Re m = =
Time scale of turbulence 1
μ *σ
where v is the rms fluctuating velocity represented as
Rij
v= where Rij = ui u j
3
and ui is the fluctuating velocity. At Re m << 1 the distortion of magnetic field lines by the
fluid turbulence is small and the induced magnetic fluctuations with respect to the mean
magnetic field is also small.
Batchelor number (Magnetic Prandtl Number): The non-dimensional number that gives the
ratio of the momentum diffusivity to the electromagnetic diffusivity is Batchelor number
defined by
(Viscous diffusion length)2 ν

Bt ≡ Prm = = = μ * σν
(Elecro-magnetic diffusion length)2 νm
It characterizes the ratio of thickness of hydrodynamic boundary layer to magneto-

hydrodynamic layers. For liquid metal flows typical values encountered are of the order of
10-5 to 10-7. When Prm is small, the effect of an external magnetic field is known to be
stabilizing and weakly dependent on Prm in a variety of flow configurations.
4. Effect of MHD on mean velocity distribution in duct flows

Many analytical studies have been conducted in fully developed duct flows starting from
1930 onwards with some significant contributions by Hartmann (1937), Shercliff (1953) and
Hunt (1965). Most of the analytical studies are performed on rectangular cross-section
because of the ease of assigning the direction of the magnetic field and also due to the ease
of non-dimensionalizing the MHD equations over the channel height and width. An axial
magnetic field does not have much significance in redistribution of mean velocities in duct
flow whereas a transverse magnetic field significantly influences the mean velocity
distribution. This is because the value of the Lorentz force acting on the body is not
significant for a flow subjected to an axial magnetic field Kirillov (1994) as shown in figure 1,
because the net Lorentz force is zero (the cross product of two vectors in the same direction
is zero). Studies have shown that a transverse magnetic field affects the mean velocity
distribution near the walls based on the distribution of induced currents in the duct cross-
section depending on the electrical conductivity of the duct walls. The direction of the
Lorentz force will act against the flow direction as per right hand thumb rule.
Classical analytical studies with transverse magnetic fields showed that for Ha >> 1 there
exist exact solutions for the governing equations or solutions in the form of infinite series as
shown in Hartmann (1937), Shercliff (1953) and Hunt (1965). It was found that there exists a
core region in the flow where the pressure forces are balanced by the Lorentz force and
velocity is uniform in the region. Special boundary layers exist depending on the orientation
of the wall with respect to the magnetic field.
Hartmann Layer: Hartmann layer refers to the thin boundary layers formed near the walls
perpendicular to the magnetic field and hence has a normal magnetic field. The thickness of
the Hartmann layer is of the order O (Ha-1) and the viscous forces balance with the magnetic
forces in the Hartmann layer. Hartmann layers have a major influence on the core region
when the electrical conductivity of the Hartmann walls is low, Fink and Beaty (1999).
Shercliff/Shear/Side Layers: The boundary layers formed on the walls parallel to the magnetic
field (side walls) can have different characteristics based on the electrical conductivity of the
walls. The layers can either stay near the walls or exist some distance away from the wall in
the form of shear layers. The thickness of the boundary layers formed on the side wall is of
the order O (Ha-½) and is higher than that of Hartmann layers. The boundary layers formed
near the side walls influence the core region significantly only if the Hartmann walls are
electrically conducting and the side walls are insulating, Fink and Beaty (1999).
Case1: Axial Magnetic field

V
Lorentz
Force = 0
B
Case2: Transverse Magnetic field
V
Lorentz
Z Y Force
X B
Fig. 1. Schematic showing the direction of Lorentz Force

The schematic of the cross-section of a rectangular duct used for the analytical studies is
shown in figure 2. The direction of flow is along the z coordinate and the external magnetic
field B0 is applied in the y direction. The walls BB are the Hartmann walls as they are
perpendicular to the magnetic field and walls AA are the side walls as they are parallel to
the magnetic field. The dimensions of the duct in x and y directions are 2a and 2b
respectively. The dimensions of the duct are non-dimensionalized as η = y/a and ξ = x/b.
B0
y,η
x, ξ
2a
A A
2b
Fig. 2. Schematic of rectangular duct with magnetic field applied in y direction
4.1 Mean velocity distribution parallel to magnetic field (Hartmann layers)

The significance of Hartmann layer on the flow depends on the electrical conductivity of the
walls. If the electrical conductivity of the Hartmann walls is low (insulating), then the
currents induced in the core regions close in the Hartmann layers and hence making it
important in deciding the resistance offered by magnetic filed on the flow. If the Hartmann
walls are electrically conducting, then the electric current lines close in the walls and hence
the significance of the Hartmann layer on the flow reduces. The mean velocity distribution
perpendicular to the Hartmann wall is given using the relation given by Hartmann (1937).
The thickness of Hartmann layer is proportional to Ha-1. The normalized axial velocity
perpendicular to the Hartmann wall is solved analytically by Hartmann (1937) for a duct
having insulating Hartmann walls.
U Ha ⎡ cosh ( Haη ) ⎤
= ⎢1 − ⎥ (23)
U c Ha-tanh ( Ha ) ⎣⎢ cosh ( Ha ) ⎦⎥
If Ha grows, the velocity profile becomes more and more flattened as shown in figure 3. This
effect is known as the Hartmann effect. The thin layer near the wall where the flow velocity
changes from zero to Uc is called the Hartmann layer. The Hartmann effect is caused by the
Lorentz force, which accelerates the fluid in the Hartmann layers and slows it down in the
bulk. The case of Hartmann walls having high electrical conductivity is given by Hunt and
Stewartson (1965) and is similar to that of the solution shown in figure 3. This states that the
Hartmann layer velocity profile is independent of the electrical conductivity of the
Hartmann walls but its control on the core flow will be high if the electrical conductivity of
the Hartmann walls is high because the electric field will be stronger.
1.05
0.90
0.75
U/Uc 0.60
Ha - 5
0.45
Ha - 10
0.30
Ha - 20
0.15
0.00
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
η
y*
Fig. 3. Normalized axial velocity profiles in the Hartmann layer (centre of duct)
4.2 Mean velocity distribution perpendicular to magnetic field (side layers)

The mean velocity distribution depends on the distribution of electric current lines in the
duct which depends on the electrical conductivity of the duct walls. The side layer thickness
varies as a function of Ha-1/2. Different cases based on the electrical conductivity of walls AA
and BB, denoted by dA and dB respectively is given in sections 4.2.1 to 4.2.4.
4.2.1 Case 1: Rectangular duct with dB = 0, dA = 0

The mean velocity distribution perpendicular to the side wall for a rectangular duct with all
walls electrically insulating is given using the relation Shercliff (1953) and then developed
by Muller and Buhler (2001) as given by equation (24).
∞
U≈ ∑ ui ( y )cos(λi x ) (24)
i = 1,3,5
ki ⎪⎧ f i ( y ) ⎪⎫
ui ( y ) = ⎨1 − ⎬
λi2 ⎩⎪ f i ( 1 ) ⎭⎪
f i ( y ) = α i2 cosh ( pi1y ) − α i1 cosh ( pi2 y )
iπ
λi =
2d
2 sin ( λi d )
ki ≡
λi d
γ ≡ Ha 2 + 4λi2
1
pi 1,2 = (Ha ∓ γ )
2
α i 1,2 = sinh ( pi 1,2 )

The normalized axial velocity with respect to the normalized distance ξ from the side walls
is plotted for various Hartmann numbers as shown in figure 4.
1.20
1.00
0.80
U/Uc 0.60
Ha - 5
0.40 Ha - 10
Ha - 20
0.20
0.00
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
ξx*
Fig. 4. Normalized axial velocity profiles in the side wall boundary layer
4.2.2 Case 2: Rectangular duct with dB = 0, dA = ∞

The mean velocity distribution perpendicular to the side wall for a rectangular duct with
Hartmann walls electrically insulating and electrically conducting side walls is given by
Hunt and Stewartson (1965). The axial velocity normalized with respect to the core velocity
1
plotted against the Ha 2 (b − ξ ) is shown in figure 5. It can be seen that the velocity profile is
similar to case 1 i.e. Rectangular duct with dB=0, dA=0. This states that when the Hartmann
walls are electrically insulating then the velocity profile near the side walls is independent of
the electrical conductivity of the side walls.
1.20
1.00
0.80
U/Uc
0.60
0.40
0.20
0.00
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
1
Ha 2 (b − ξ )
4.2.3 Case 3: Rectangular duct with dB = ∞, dA = 0

The mean velocity distribution perpendicular to the side wall for a rectangular duct with all
walls electrically conducting is given by Hunt (1965). The mean velocity distribution
perpendicular to the side wall is given using the relation Hunt (1965) as Ha → ∞
∞ 2( −1) j cosα jη ⎡ ⎧ ⎫⎤
U≈∑
Ha 2α j
⎢1 − exp −λjξ
⎢⎣
( ) ⎪⎨cos ( λjξ ) − Ha
α
⎪
sin ( λjξ ) ⎬⎥
⎥
(25)
j=0 ⎪⎩ j ⎭⎪⎦
⎛ 1⎞
α j ≈ ⎜ j + ⎟π
⎝ 2 ⎠
1
⎛1 ⎞2
λj ≈ ⎜ α jHa ⎟
⎝ 2 ⎠
The axial velocity normalized with respect to the normalized distance from the side walls is
plotted for various Hartmann numbers as shown in figure 6. The maximum velocity in the
side layers is of the order O(Ha)Vc and the maximum velocity on the side wall jet tends to
0.25HaVc as Ha → ∞ , while the minimum velocity becomes locally negative for Ha > 89 and
tends to - 0.011 HaVc The high velocity gradients near the side walls generate instabilities at
high Hartmann numbers. The high velocity jets formed near the side walls results in
velocity deficit in the core region and in the Hartmann layers.
24.00 Ha - 5
Ha - 10
19.00 Ha - 20
Ha - 100
U/Uc 14.00
9.00
4.00
-1.00
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
ξ
y*
4.2.4 Case 4: Rectangular duct with dB = ∞, dA = ∞

The mean velocity distribution perpendicular to the side wall is given using the relation
given by Hunt (1965). The maximum velocities in the boundary layer are higher than the
core velocity but are of the same order unlike the case with d A = 0
∞ 2( −1) j cosα jη
U≈∑
Ha 2α j ⎣⎢ (
⎡ 1 − exp −λ ξ
j ){cos ( λjξ ) − sin ( λjξ )}⎤⎦⎥ (26)
j=0
The axial velocity normalized with respect to the normalized distance from the side walls is
plotted for various Hartmann numbers as shown in figure 7. The velocity near the side wall
increases with increase in magnetic field, but the maximum value of velocity jets approaches
1.25Vc when Ha → ∞ .
1.25
1.00
0.75
U/Uc
Ha - 5
0.50
Ha - 10
0.25 Ha - 20
0.00
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
ξ
y*
Case:1 (dB = 0, dA = 0) Case:2 (dB = 0, dA = ∞)
B0 B0
U/Uc U/Uc
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Case:3 (dB = ∞, dA = 0) Case:4 (dB = ∞, dA = ∞)
B0
B0
U/Uc U/Uc
0 1 2 3 4 5 6 7 0 0.25 0.5 0.75 1.0 1.25

Fig. 8. Axial velocity normalized with respect to the core velocity plotted for Ha = 20
4.3 Comparison on mean axial velocity profiles

The analytical solutions of the mean velocity distribution presented in the previous sections
are supported using numerical simulations performed for laminar ( Re ≈ 10) steady state
flow of an electrically conducting fluid in a square duct using FLUENT (version 6.3.26), see
figure 8. The velocity profile near the Hartmann walls is independent of the electrical
conductivity of the walls as stated in section 4.1 and the velocity gradients and thickness
seems to be similar. The velocity profiles near the side walls shows different characteristics
depending on the electrical conductivity of the walls. The velocity profiles for case 1 in
section 4.2.1 and case 2 in section 4.2.2 are similar as the induced field has low strength due
to the high resistance offered by the Hartmann walls.
In the cases where dB = 0 the net electrical current in the circuit will be low and current lines
will not be strong enough to create a redistribution in the mean velocity field as shown in
figure 8 and the imbalance between the Lorentz forces and pressure will not be significantly
high.
Case:1 (dB = 0, dA = 0) Case:2 (dB = 0, dA = ∞)

η B0 B0
B B
O(Ha-1)
Current Lines
A ξ A
O(Ha-1/2)
Case:3 (dB = ∞, dA = 0) B0 Case:4 (dB = ∞, dA = ∞)

B0
B B
A A
Fig. 9. Schematic of the current lines in the rectangular duct

The distribution of induced current lines in the cross section for different cases is given in
figure 9. The formation of shear layers in the case 3 given in section 4.2.3 where dB = ∞ and
1
dA = 0 is due to the imbalance in the Lorentz forces and pressure forces. If dAHa 2 << 1 , then
the current lines return to walls BB through boundary layers on wall AA and hence the
Lorentz force term drops near the boundary layer as compared to other regions making the
velocity near the side walls to shoot up as shown in figure 9.
1
If dAHa 2 >> 1 then the current lines returns to wall BB through walls AA which makes the
induced magnetic field and hence the Lorentz Force (j × B) term uniform in the duct cross-
section which happens in case 4 of section 4.2.4 shown in figure 8 and figure 9. Higher the
value of electrical conductivity of side walls, lower the value of the maximum velocity of the
jet for case 3 and case 4.
5. Effect of MHD on turbulence suppression

An external magnetic field suppresses the turbulent fluctuations in liquid metals (Low
magnetic Prandtl number fluid) independent of the direction of magnetic field with respect
to the flow and independent of the mean velocity distribution in the flow. The flow regimes
in magneto-hydrodynamic flows of liquid metals, i.e. Rm << 1, see Fink and Beaty (1989) is
shown in figure 10. It can be seen that the flow regime above line ‘A’ is the laminar regime.
1.E+05
10 5
N
1.E+04
10 4
Laminar
Electromagnetically
C Laminarized
1.E+03
103
A
1.E+02
102
Transition
1.E+01
10 1 B
Turbulent
1.E+00
10 0
1.E+01
10 1
1.E+02
102
1.E+03
103
1.E+04
10 4
1.E+05
10 5
Ha
Fig. 10. Turbulent flow regimes in low magnetic Reynolds number flows, Fink and Beaty
(1989)
The flow regime below line ’B’ is turbulent regime with low values of Stuart number N, in
which the influence of the magnetic field on the flow is less. The shaded region between line
‘A’ and line ‘B’ is the transition region in which the turbulence levels increase and will be
greater than the flow without magnetic field as stated in Uda et al. (2001 and 2002). The line
‘C’ in figure 10 represents the critical Reynolds number for laminar to turbulent transition
equal to 2000. The region on the left side of line ‘C’ is the classical laminar flow regime. The
region on the right side of line ‘C’ and above line ‘A’ is the region of electromagnetically
laminarized flow in which the turbulent flow is converted to so-called Stokes flow in which
the inertial term is negligible as mentioned in Happel (1981).
Axial magnetic field suppresses turbulent fluctuations and also increases the critical
Reynolds number for transition from laminar to turbulent regime as stated in Kirillov (1994).
A transverse magnetic field also has the effect of turbulence reduction. Moffat (1967). The
turbulent structures get aligned with the magnetic field lines. As shown in figure 11, when
the magnetic field increases, the MHD effect suppresses the turbulence significantly and the
large vortex structure breaks up into several small vortex structures. At the same time, the
vortex structure was dampened and stretches along the flow direction (perpendicular to the
magnetic field).
Without Magnetic Field With Magnetic Field
Fig. 11. Change of isotropic turbulent structures to anisotropic structures, Luo et al. (2003)
6. Effects of mean velocity distribution and turbulence suppression on

convective heat transfer
The convective heat transfer characteristic is affected by magnetic field due to two main
mechanisms. Evtushenko (1995)
1. Mean velocity distribution in the flow
2. Velocity fluctuations in time and space
The first mechanism, mean velocity distribution has to take into account two important
factors: (1a) the increase of velocity and velocity gradients near side walls parallel to the
magnetic field and (1b) Velocity (flow rate) redistribution in the magnetic field direction.
The second mechanism, effect of MHD on turbulent fluctuations and hence heat transfer has
to take into account important factors like (2a) Damping of turbulent fluctuations, (2b)
transition to two-dimensional structure and (2c) the instabilities created in the high velocity
jets near the side wall.
6.1 Effect of axial magnetic field on convective heat transfer

In case of flow of liquid metals in heated channels under the influence of a uniform axial
magnetic field shows a decrease of convective heat transfer at low and moderate Hartmann
numbers whereas the convective heat transfer and hence Nu increases at higher Hartmann
numbers as shown in Miyazaki (1988). It was stated in section 4 that an axial magnetic field
does not affect the mean velocity distribution so the modification of convective heat transfer
is due the variation in the turbulent fluctuations in time and space.
Reynolds number Hartmann Number Nu/Nu B=0

(2.5 − 5) ⋅ 10 3 360 0.83
4
(1 − 2) ⋅ 10 700 0.50
4
(3 − 4) ⋅ 10 1400 0.30
1 ⋅ 10 4 3600 2.75
Table 2. Variation of Nusselt Number with Axial Magnetic Field, Miyazaki (1988)
The values of Nu for various value of Hartmann numbers is shown in table 2. It can be seen
that the values of Nu decreases from its value, Nu/Nu B=0 = 1, at Ha = 0. The decrease in Nu
value for small and moderate Ha is more when the Reynolds number is high because of the
higher turbulence content in the flow. At lower values of Reynolds numbers, the flow will
be inherently laminar and therefore the reduction in Nu due to suppression of turbulent
fluctuations will be low.
At high values of Hartmann numbers, the Nusselt number was found to increase, violating
the earlier theories and studies, see Miyazaki (1988). Miyazaki attributes the increase in the
Nusselt number is due to the increase in turbulence levels in the flow as the effect of
buoyancy can be ruled out because the flow direction upwards.
6.2 Effect of transverse magnetic field on convective heat transfer

The studies in the field of effect of magnetic field on convective heat transfer in ducts
subjected to transverse magnetic field can be classified into two cases
1. Absence of high velocity jets near the side walls
2. Presence of high velocity jets near the side walls
6.2.1 Case 1: Absence of high velocity jets near the side walls
In case of ducts having Hartmann walls with zero conductivity, high velocity jets will not be
formed near the side walls. Gardener and Lykoudis (1971b) performed experiments with
flow of Mercury in horizontal electrically insulated pipe subjected to transverse magnetic
field. It was found that the velocity profile near the Hartmann wall becomes flat with
increase in magnetic field as discussed in section 4.1 and the velocity profile near side walls
becomes round as discussed in sections 4.2.1 and 4.2.2. The mean velocity distribution is not
much different with the increase in magnetic field, so the modification of turbulence
phenomenon by the magnetic field will affect the convective heat transfer predominantly for
this case. The Nusselt number distribution near the Hartmann and side walls for a range of
Reynolds numbers and Hartmann numbers is shown in figure 12. The decrease of Nusselt
number with increase in magnetic field is lesser at lower Reynolds number because the
turbulence content in the flow at low Reynolds number will be lesser.
40
40 No
No magnetic field
magnetic field
Hartmann wall 375
Hartmann – Ha - Ha 375
30
30
Side
Side wall
– Ha -375
Ha 375
Nuu
20
20
N
10
10
10
10 4
10000 105
100000 106
1000000
Re
Fig. 12. Nusselt number with magnetic field intensity, Gardener and Lykoudis (1971b)
The reduction in Nusselt number with increase in magnetic field is because of the reduction
in turbulence quantified using turbulence kinetic energy as shown in figure 13. It was found
that the turbulent kinetic energy decreases both near the Hartmann and side walls with
increase in magnetic field where r/R = 0 represents the centre of the duct and r/R = 1
represents the walls. The damping force within the Hartmann layer is much higher than at
the side region due to the high local electric current density. The turbulence in core is
suppressed initially and then the turbulence in the Hartmann layer followed by the
turbulence near the side wall.
Kinetic Energy
Energy
3.5
35000
3.0
30000 No magnetic field
2.5
25000 Hartmann wall - Ha 47
Turbulent Kinetic
2 2 2 2
/s/s
2.0
20000 Side wall - Ha 47
x 104mm
15000
1.5
10000
Turbulent
1.0
5000
0.5
0
0.0
0.00 0.20 0.40 0.60 0.80 1.00
r/R
Fig. 13. Turbulent kinetic energy vs. r/R for Re = 50,000, Gardener and Lykoudis (1971a)
A correlation for Nusselt number values is created from various experimental results by Ji
and Gardener (1997) and is given using the following relation as a function of Peclet number
Pe and Hartmann number Ha
0.00782Pe0.811
Nu = 7 + (27)
( 1 + 0.0004Ha 1.5
f ( Pe ) )
(
f ( Pe ) = 0.3 + 4.75 × 10 −5 Pe − 2.10 × 10 −9 Pe 2 )
6.2.2 Case 2: Presence of high velocity jets near the side walls
In case of ducts having Hartmann walls with finite conductivity, high velocity jets will be
formed near the side walls. The side layers with high velocity jets (M shaped profile, figure 8
case 3) carry high mass flux is the prime reason for increase of heat transfer near the side
walls. Miyazaki et al. (1986) performed experiments to determine the heat transfer
characteristics for liquid metal Lithium flow in annular duct with electrically conducting
walls under the influence of transverse magnetic fields. The Nusselt number plotted with
magnetic field is shown in figure 14. It can be seen that the Nusselt number increases near
the side walls and decreases near the Hartmann walls. A singular rise of Nusselt number
can be seen near both the walls.
8.5 Side Walls

8.0 Hartmann Walls
7.5
7.0
Nu
6.5
6.0
5.5
5.0
0.00 0.20 0.40 0.60 0.80 1.00
Magnetic FIeld, Tesla

Fig. 14. Nusselt number plotted with magnetic field intensity, Miyazaki (1986)
This effect of heat transfer enhancement near the side walls is caused by the generation and
development of large scale velocity fluctuations in the near wall area. The reduction in
Nusselt number near the Hartmann walls is created due to the turbulence reduction as
shown in figure 15.
0.8 Side Walls

RMS of Temperature
Hartmann Walls
Fluctuation (0C)
0.6
0.4
0.2
0.0
0.00 0.20 0.40 0.60 0.80 1.00
Magnetic FIeld, Tesla
Fig. 15. RMS of temperature fluctuation with magnetic field intensity, Miyazaki (1986)
7. Application of numerical codes

A difficulty in experimental study of the flow of liquid metals arises as the visualization is
not possible because of the opaque nature of liquid metals. Application of closed form
analytical solutions is limited to simple cases where the equations are not very complex.
This makes the application of numerical simulations useful for the study of liquid metal
magneto-hydro-dynamic flows. An example of application of a numerical code to explain
the mechanisms affecting heat transfer for flow subjected to transverse magnetic field is
explained using a series of simulations given in Rao and Sankar (2010), see figure 16.
Fluid Elements
B T5 Solid Elements
30
T4
T7
Heater Pin
T8
300
1.1
x 1.65
y
15.8 Height of the first

Flow cell = 1μm
7.6
Direction
(a) (b)
Fig. 16. (a) Schematic of model (b) Details of the computational mesh, Rao and Sankar (2010)
A numerical study is conducted in an annular duct formed by a SS316 circular tube with
electrically conducting walls and a coaxial heater pin, with liquid Lithium as the working
fluid for magnetic field ranging from 0 – 1 Tesla The Hartmann and Stuart number of the
study ranges from 0 – 700 and 0 – 50 respectively. The Reynolds number of the study is 104.
It was shown that the convective heat transfer and hence the Nusselt number decreases near
the walls perpendicular to the magnetic field due to reduction in turbulent fluctuations with
increase of magnetic field. It was observed that the Nusselt number value increases near the
walls parallel to the magnetic field as the mean velocity increases near the walls. A singular
rise was observed near both the walls near Stuart number ~ 10 which is due to the increase
of turbulence levels in the process of changing from turbulent to electromagnetically
laminarized flow, see figure 17.
When a very low Reynolds number ~ 300 is used, the reduction in Nusselt number near the
Hartmann walls is less as shown in figure 18. This shows that the reduction in Nusselt
number near the Hartmann walls for the high Reynolds number study is due to the
reduction in turbulent fluctuations. The Nusselt number was found to increase near the side
walls as the mean velocity increases near the walls. When an insulating duct is used the
Nusselt number near the parallel walls did not increase for the case with insulating walls as
in the case with conducting walls showing the contribution of the ‘M’ shaped velocity
profile in the Nusselt number increase near the parallel walls. The Nusselt number near the
perpendicular walls was found to decrease at a higher rate in case of insulating walls than
that of the study with conducting walls as shown in figure 19.
1.41.4 Hartmann
Hartmann wall wall
Side
Side wall
wall
1.21.2
Nu/NuB=0
1.01.0
0.80.8
0.60.6
0.0
0.0
0.2
0.2
0.4
0.4
0.6
0.6
0.8
0.8 1.0
1.0 Tesla
0.0 1.7 7.5 16.9 30.1 47.0 St
Fig. 17. High Reynolds number with conducting walls, Rao and Sankar (2010)
Hartmann
Hartmann wall
wall
1.61.6
Side
Side wall
wall
1.41.4
Nu/NuB=0
1.21.2
Nu/NuB=0
1.01.0
0.80.8
0.0 0.2 0.4 0.6 0.8
0.0 0.2 0.4 0.5 0.8 Tesla
0.00 1.78 7.52 16.93 30.10 St
Fig. 18. Low Reynolds number with conducting walls, Rao and Sankar (2010)
Side wall
Side wall
1.05
1.05 Hartmann wall
Hartmann wall
1.00
1.00
0.95
0.95
Nu/NuB=0
0.90
0.90
0.85
0.85
0.80
0.80
0.75
0.75
0.70
0.70
0.00
0.00 0.2
0.20 0.4
0.40 0.6
0.60 0.8
0.80 1.0 Tesla
1.00
0.00 1.78 7.52 16.93 30.10 47.04 St
Fig. 19. High Reynolds number with insulating walls, Rao and Sankar (2010)
8. Application of liquid metal MHD studies in nuclear fusion reactors

International Thermo-nuclear Experimental Reactor is an international organization formed
in 1985 comprising of researchers from US, EU, China, Japan, India, Korea and Russia
working towards development of a test reactor (TOKOMAK) which is expected to be
developed by 2020. The test reactor will be installed in France where the head office of ITER
is situated. Salient details of the reactor to be developed are shown in figure 20. The reactor
height will be close to 100 ft and would weigh around 38000 tons. The cryostat is the
external chamber around the TOKOMK which maintains high vacuum inside it to reduce
the heat load from atmosphere through conduction and convection. The fusion of Deuterium
and Tritium happens inside the plasma chamber. The magnets are used to confine the
plasma created inside the plasma chamber using a magnetic field of 4-8 Tesla.
Cryostat
Plasma
Chamber
Central
Solenoid
Toroidal
Magnets
Fig. 20. Details of the TOKOMAK

Tritium breeding modules are used in fusion reactors to produce Tritium by reacting
Lithium with neutrons a byproduct of the nuclear fusion reaction. The two basic breeder
concepts developed by ITER are liquid breeder and solid breeders. The advantages of liquid
breeder over solid breeder are the high Tritium breeding ratio and the Lead-Lithium
eutectic can also act as a coolant inside the breeding module which is subjected to high heat
from plasma and the heat generated in itself due to bombardment of neutrons. The major
disadvantages of liquid breeders over solid breeders is the pressure drop in the form of
Lorentz force and the reduction in convective heat transfer characteristics of the liquid metal
when it is flowing in the presence of intense magnetic field produced by the cryogenic
super-conducting magnets.
Wong et al. (2008) has mentioned about the various liquid metal breeders being developed
around the world details of which is given in the table 3. All the liquid breeder design uses
Lithium as the breeding material though most of them use a eutectic of Lead and Lithium
because of the lower electrical conductivity and the neutron multiplication ability of Lead.
Country Name of TBM Liquid Metal Used

US DCLL – Dual Coolant Lead Lithium PbLi
EU HCLL – Helium Cooled Lithium Lead PbLi
Korea HCML – Helium Cooled Molten Lithium Li
India LLCB - Lead-Lithium Cooled Ceramic Breeder PbLi
China DFLL – Dual Functional Lithium Lead PbLi
Table 3. Details of the liquid TBM developed in the various countries, Wong et al. (2008)
Pb-Li Inlet
First
Wall
PbLi
Poloidal View Exploded View

Fig. 21. Details of LLCB- TBM, Wong et al. (2008)
Indian Lead lithium Cooled Ceramic Breeder (LLCB) – The design description of LLCB is given
in Rao et al. (2008). The details of the exploded and cut section views of the LLCB – TBM is
shown in figure 21. The two coolants used in LLCB are Helium and a eutectic of Lead-
Lithium, Pb-Li. The two coolants are of different molecular properties as Pb-Li has very low
Prandtl Number of the order 10-2 and Helium gas has Prandtl number of ~0.65. The thermal
diffusivity of the two fluids were different as the main temperature difference for Helium in
straight ducts were concentrated at the viscous sub layer where as the temperature
difference for Pb-Li was also present in the mean core region.
The material of construction of the cooling channels is Ferritic-Martensitic Steel (FMS)
having electrical conductivity of the order 106 1/Ώ-m, so the pressure drop associated with
the flow was very high. Hence a coating of Alumina (Al2O3), which has very low electrical
conductivity (~10-8 1/ Ώ-m) is used on the wet surfaces of the cooling channels This makes
the configuration similar to the rectangular channel of Shercliff’s case with all walls
insulating i.e. dA = 0 and dB = 0 and hence as mentioned in 4.2.1, the velocity profiles will not
have a high velocity jet near the side walls. So the effect of turbulence modification is more
significant on the heat transfer characteristics as mentioned in section 6.2.1. The flow will be
electromagnetically laminarized and the heat transfer capacity of the Pb-Li deteriorates at
high Hartmann numbers.
9. Nomenclature
2a Distance between Hartmann walls
2b Distance between side walls
B0 Magnetic field
c Speed of light
Cp Specific heat
dA Electrical conductivity of wall AA
dB Electrical conductivity of wall BB
D Displacement current
E Electric field
H Magnetic field strength
Ha Hartmann Number
j Electric charge
k Thermal conductivity
keff Effective thermal conductivity
kΤ Turbulent thermal conductivity
L Characteristic length
N Interaction parameter
Nu Nusselt number
p Pressure
Pr Prandtl number
Prm Magnetic Prandtl number
q ''' Volumetric heat generation
S Source term
Re Reynolds number
Rem Magnetic Reynolds number

t Time
T Temperature
U Axial velocity
Uc Centre line velocity
U0 Mean velocity
σ Electrical conductivity of fluid

η Non-dimensionalized distance in y direction
ξ Non-dimensionalized distance in x direction
ν Kinematic viscosity of fluid
ν eff Effective viscosity
ντ Turbulent viscosity
ρ Density of fluid
μ Dynamic viscosity of fluid
μ* Magnetic permeability
ρc Electric charge density
10. Acknowledgement
We would like to acknowledge Altair Engineering India Pvt. Ltd., for providing an
opportunity to do the associated work
11. References
Alfven, H. (1942). Existence of electromagnetic-hydrodynamic waves. Nature, Vol.150,
(1942), pp.405-406.
Davidson, H. W. (1968). Compilation of thermo-physical properties of liquid Lithium.. NASA
Technical Note, Washington. D. C., 1968.
Evtushenko, I. A.; Hua, T. Q.; Kirillov. I. R.; Reed, C. B. & Sidorenkov, S. S. (1995). The effect
of a magnetic field on heat transfer in a slotted channel. Journal of Fusion Engineering
and Design, Vol. 27, (1995), pp. 587-592.
Fink, D. & Beaty, H. W. (October 1999). Standard handbook for electrical engineers (14th Edition).
McGraw Hill, ISBN 0070220050.
Gardener, R. A. & Lykoudis, P. S. (1971a). Magneto-fluid-mechanic pipe flow in a transverse
magnetic field Part 1 Isothermal flow. Journal of Fluid Mechanics, Vol.47, (1971), pp
737-764.
Gardener, R. A. & Lykoudis, P. S. (1971b). Magneto-fluid-mechanic pipe flow in a transverse
magnetic field Part 1 Heat Transfer. Journal of Fluid Mechanics, Vol.48, (1971), pp.
129-141.
Happel, J. & Brenner, H. (1981). Low Reynolds Number Hydrodynamics, Springer. ISBN
9001371159.
Hartmann, J. (1937) Theory of the laminar flow of electrically conductive liquid in a

homogeneous magnetic field, Hg-Dynamics, Kgl. Danske Videnskab. Selskab. Mat.-
fus. Medd., Vol.15, No.6, (1937)
Hartmann, J. & Lazarus. P. (1937). Experimental investigation of flow of Mercury in a
homogeneous magnetic field, Kgl. Damske Videnskabernes Selskab, Math-,Fys. Med,
Vol.14, No. 7, (1937).
Hunt, J. C. R. (1965). Magnetohydrodynamic flow in rectangular ducts. Journal of fluid
mechanics, Vol. 21, No. 4, (1965), pp. 577-590.
Hunt, J. C. R. & Stewartson, K. (1965). Magnetohydrodynamic flow in rectangular ducts. II.
Journal of fluid mechanics, Vol. 23, No.3, (1965), pp. 563-581.
Ji, H. C. & Gardener, R. A. (1997). Numerical analysis of turbulent pipe flow in a transverse
magnetic field. International Journal of Heat and Mass Transfer, Vol.40, No.8, (1997),
pp. 1839-1851.
Kirillov, I. R.; Reed, C. B.; Barleon, L. & Miyazaki, K. (1994). Present understanding of MHD
and heat transfer phenomenon for liquid metal blankets, Proceedings of 3rd
International Symposium of Fusion Nuclear Technology, Los Angeles, 1994.
Lielpeteris, J & Moreau, R. (1989). Liquid metal magnetohydrodynamics, Kluwer Academic
Publishers Group, ISBN 079230344X, Dordrecht, Boston.
Luo, X.; Ying, A. & Abodu, M. (2003). Experimental and computational simulation of free jet
characteristics under transverse field gradients. Journal of Fusion Science and
Technology, Vol 44, (July 2003), pp. 85-93.
Miyazaki, K.; Inoue, h.; Kimoto, T. ; Yamashita, S.; Inoue, S. & Yamaoka, N. (1986).
Heat transfer and temperature fluctuation of lithium flowing under transverse
magnetic field. Journal of Nuclear Science and Technology, Vol.23, (1986), pp. 582-
593.
Miyazaki, K.; Yokomizo, K.; Nakano, M.; Horiba, T. & Inoue, S. et al. (1988). Heat Transfer
and Pressure Drop of Lithium Flow under Longitudinal Strong Magnetic Field,
Proceedings of LIMET’88, Avignon, 1988.
Moffatt, H. K. (1967). On the suppression of turbulence by a uniform magnetic field. Journal
of Fluid Mechanics, Vol. 28, (1967), pp. 571–592.
Muller, U. & Buhler, H. (2001), Magneto-fluid-dynamics in Channels and Containers (1st
Edition), Springer, ISBN 978-3-540-41253-3.
Rao, J. S. et al.(2008). Design description document for the dual coolant Pb 17Li (DCLL) test blanket
module, Report to the ITER test blanket working group (TBWG), (2008), Institute of
Plasma Research, India.
Rao, J. S. & Sankar, H. (2011). Numerical Simulation of MHD Effects on Convective Heat
Transfer Characteristics of Flow of Liquid Metal in Annular Tube. Journal of Fusion
Engineering and Design, Vol.86, No.2-3, (March 2011), pp. 183-191.
Roberts, P. H. (1967). An Introduction to Magnetohydrodynamics, Longmans Green and Co Ltd,
1967, ISBN 978-0-582-44728-8.
Shercliff, J. A. (1953). Steady motion of conducting fluids in pipes under transverse magnetic
fields, Proceedings of Cambridge Philosophical Society, pp. 136-144, 1953.
Uda, N.; Miyazawa, A. ; Inoue, S.; Yamaoka, n.; Horiike, H. & Miyazaki, k. (2001). Forced
convection heat transfer and temperature fluctuations of lithium under
transverse magnetic field. Journal of Nuclear Science and Technology, Vol. 38, (2001),
pp. 936-943.
Uda, N.; Miyazawa, A.; Yamaoka, H. N.; Horiike, H. & Miyazaki, K. (2002). Heat transfer
enhancement in lithium annular flow under transverse magnetic field. Energy
Conversion and Management, Vol.43, (2002), pp. 441-447.
Wong, C. P. C. ; Salavy, J. F.; Kim, Y.; Kirillov, I.; Kumar, E. R.; Morley, n. B.; Tanaka, S. &
Wu, Y. C. (2008). Overview of liquid metal TBM concepts and programs. Journal of
Fusion Engineering and Design, Vol.83, (2008), pp. 850-857.
5
Thermal Anomaly and Strength of

Atotsugawa Fault, Central Japan,
Inferred from Fission-Track Thermochronology
Ryuji Yamada1 and Kazuo Mizoguchi2
1National Research Institute for Earth Science and Disaster Prevention
2Central Research Institute of Electric Power Industry
Japan
1. Introduction
Frictional slip induces temperature rise in a fault zone. Abundant frictional heat during an
earthquake sometimes produces melt of rocks (i.e. pseudotachylyte; e.g., Sibson, 1975). The
amount of heat production along faults provides the essential information to investigate
frictional strength of the faults that characterizes the earthquake generation processes.
Lachenbruch and Sass (1980) first estimated the coefficient of friction of the San Andreas
Fault to be 0.1-0.2 from the measurement of surface heat flow along the fault. Kano et al.
(2006) found a temperature rise of ~ 0.06 °C measured in a borehole drilled across the
Chelungpu fault six years after the 1999 Chi-Chi, Taiwan earthquake associated with this
fault. They found that very low coefficient of friction of 0.04-0.08 can explain the heat
anomaly along the Chelungpu fault. The above observations along the natural faults have
suggested a very low friction level compared with that of 0.6-0.8 evaluated in laboratory
rock friction experiments (Byerlee, 1978).
Fission-track (FT) thermochronology is an effective method to detect heat anomaly caused
by past faulting (e.g., Scholz et al., 1979; Camacho et al., 2001; Murakami et al., 2002;
Murakami and Tagami, 2004; Yamada et al., 2007a). In order to constrain the frictional
properties of faults, d’Alessio et al. (2003) measured apatite FT ages and lengths for samples
adjacent to and within the San Gabriel fault zone that is thought to be an abandoned major
trace of the San Andreas Fault system active from 13 to 4 Ma. They found no evidence of a
localized thermal anomaly in FT data even in samples within just 2 cm of the
ultracataclasite, and concluded that either there has never been an earthquake with > 4 m of
slip at this locality, or the average apparent coefficient of friction is < 0.4 based on the
modelling of heat generation and transport.
In this paper, we estimate the frictional strength of the Atotsugawa fault, central Japan,
using the method similar to that used by d’Alessio et al. (2003). In the Atotsugawa fault,
Yamada et al. (2009) performed FT thermochronologic analysis at an outcrop without visible
pseudotachylyte layers, and revealed a thermal anomaly at a several cm thick gouge whose
apatite age is significantly younger than those of other samples in the vicinity. Assuming
that the thermal anomaly is cause by frictional heating during a single earthquake, the
frictional coefficient and the ancient depth of gouge samples are evaluated by the thermal
modelling to satisfy the constraints given by the FT thermochronological data with respect
to the geometry and alignment of the gouges in the outcrop.
2. Fission-track thermochronology in Atotsugawa fault zone

The Atotsugawa fault is a right-lateral strike-slip one with a strike of N60°E and almost
vertical dip, located in the Hida metamorphic belt, central Japan (Fig. 1). From the trenching
surveys, a number of historical large earthquakes were detected along the Atotsugawa fault;
the most recent one is the 1858 Hietsu earthquake. The estimate of the magnitude ranges 7.0
(Usami, 1987), 7.3 (Matsu'ura et al., 2006) and 7.9 (Doke and Takeuchi, 2009). Geographical
Survey Institute, Japan (GSI; 1997) reported a creeping slip with a rate of 1.5 mm/yr in the
central section of the fault.
Fig. 1. Distribution of active fault around the Atotsugawa fault. An open circle (centre) and an
open triangle (bottom left) symbols indicate the locations of an outcrop and a reference site
of R2 in Fig. 2, respectively
In the creeping section, Yamada et al. (2009) performed FT thermochronologic analysis by
measuring ages of apatite and zircon grains separated from gouges and fractured rocks at
six fracture zones within a 15 m-wide fault zone without visible pseudotachylyte layers (Fig.
2a). This fault zone consists of six fault gouges (name, thickness; gouge-1, 10 cm; gouge-2, 8-
20 cm; gouge-3, 8-25 cm; gouge-4, 10 cm; gouge-5, 10-30 cm; gouge-6, 20 cm). FT ages of
zircon (c. 120-150 Ma) and apatite (c. 44-60 Ma) for samples except "gouge-1" agree well with
emplacement ages for the Funatsu granitic rocks that intruded the Hida Belt (Matsuda et al.,
1998). The discordance in zircon and apatite FT ages is interpreted to reflect the rock cooling
due to the regional uplift and associated erosion. A thermal anomaly was identified at the
gouge sample of "gouge-1" that showed an exceptionally young apatite age (32.1 ± 3.2 Ma,
1σ) with a unimodal FT length distribution, although its zircon age (121 ± 6 Ma, 1σ) was
well concordant with other samples (Fig. 2b; after Yamada et al., 2009). The creeping slip
Thermal Anomaly and Strength of Atotsugawa Fault,
Central Japan, Inferred from Fission-Track Thermochronology 83
observed in the central section of the Atotsugawa fault (GSI, 1997) could be a possible
source for this heat anomaly. Such a low slip rate of 1.5 mm/yr, however, causes much
smaller increase in temperature (< 20 °C; d’Alessio et al., 2003) in the fault zone. This
disagreement can therefore be attributed to the secondary heating induced by frictional slip
during an associated earthquake, and the young apatite age possibly gives a younger limit
of the initiation of the activity in the Hida Belt.
Fig. 2. (a) Sketch of the outcrop of the Atotsugawa fault zone and (b) fission-track age variation
in apatite (lower) and zircon (upper) across the outcrop (after Yamada et al., 2009). Open
circle, solid circle and square symbols indicate data of fault gouge, fractured rock and host
Hida metamorphic rock, respectively. Dashed lines indicate locations of the fault gouge
zones. Two reference samples of R1 and R2 (an open triangle in Fig. 1) were also collected
where no fractures were observed. Length distribution of apatite FTs for sample ATG1G is
also shown. Shaded bands behind the plot indicate the apatite and zircon FT age distributions
of the granitic rocks that intrude into the Hida Belt (Matsuda et al. 1998). Error bars show 2σ
uncertainty in age
3. Thermal modelling associated with frictional heating and estimation of

frictional strength
In order to estimate the frictional strength of the Atotsugawa fault based on the FT
thermochronological data, we modelled the temporal change in the temperature in and out
of the "gouge-1" where an exceptionally young apatite age was found (Yamada et al., 2009).
The FT data and the geometry of the occurrence of gouges in the outcrop indicate that the
apatite FT age in the 10 cm thick “gouge-1” zone was thermally reset but that in the
fractured rock 10 cm apart from “gouge-1” was not. Therefore, the model space for the thermal
modelling is composed of a central slip zone of 10 cm thickness and the surrounding rock
zone of 10 m thickness with a homogeneous temperature distribution at a certain depth in
the initial state (Fig. 3). It is assumed that a single fault slip occurs at a constant rate and all
of the frictional work converts into heat. One-dimensional heat transfer model is used to
describe the heat diffusion into the surrounding zone at a direction normal to the slip zone.
The effect of thermal diffusion by fluid flow is not considered because hydraulic properties
of the Atotsugawa fault zone at depth have not yet been investigated.
The equation of thermal diffusion with frictional heat source term for the slip zone is given
by
∂ T ⎛ V ⋅ μ ⋅σn ⎞ ∂ 2T
ρr ⋅ C p =⎜ ⎟+κ (1)
∂ t ⎝ Wc ⎠ ∂ x2
where T is temperature, t is the lapse time after a slip occurs, V is the slip rate of the fault, µ
is the coefficient of friction, σn is the effective normal stress on the fault, Cp is the heat
capacity of rock, Wc is the width of a slip zone, k is the thermal conductivity of rock, ρr is the
density of rock, and x is the distance normal to the fault from the centre of the slip zone.
The effective normal stress σn is equivalent to the effective overburden pressure given by
(ρr - ρw)·H·g, where ρw is the density of water, H is depth and g is gravity. For the surrounding
zone, Equation (1) without the heat source term (i.e., the first term in the right hand side) is
used.
Fig. 3. Fault model for thermal calculation associated with frictional heating. Distance is
measured from the centre of the fault
Table 1. Parameters used for thermal calculations

Constants of typical physical properties of rocks are used in the thermal modelling as shown
in Table 1 (c.f., Schön, 1996). Wc is set as 10 cm that is equivalent to the thickness of the
“gouge-1”. The initial temperature at a certain depth H is obtained from a geothermal
gradient of 30 °C/km (typical value for the upper crust in Japan; e.g., Tanaka et al., 1999)
multiplied by H plus a surface temperature of 20 °C, and the temperature distribution over
the fault zone is assumed to be uniform. The boundary condition of the temperature at the
edge of the model space is fixed at the initial value. Total slip of 5 m long is given for this
fault system because the estimate of the magnitude of the associated earthquake ranges
from 7.0 to 7.9 (Usami, 1987; Matsu'ura et al., 2006; Doke and Takeuchi, 2009) that
corresponds to the total slip of the order of 1-10 m, based on the empirical relationship
between the fault displacement and the magnitude (Matsuda, 1975). V is set at 1 m/s (e.g.,
Heaton, 1990) and therefore the slip duration is 5 sec. Considering the closure temperature
of apatite FT (100 ± 20 °C; e.g., Wagner and Van den Haute 1992), H should be shallower
than 3 km which corresponds to the environment temperature of 110 °C, and therefore
restricted to the three cases of 1, 2 and 3 km for the modelling assuming a geothermal
gradient of 30°C/km (e.g., Tanaka et al, 1999).
Fig. 4. Temperatures at the centre of the fault (T0, solid line) and at 10 cm apart from the
centre (T10, dashed line) are plotted as a function of time since a slip occurs
(a)
(b)
Fig. 5. Maximum temperature at the centre of the fault (T0; a) and at the location 10 cm apart
from the centre (T10; b) during an earthquake are plotted as a function of friction coefficient of
0.1 ~ 0.9 in cases of depth from 1 to 3 km. Square, triangle and circle symbols denote the data
at 1, 2 and 3 km, respectively. Shaded areas in (a) and (b) indicate ranges of T0 (upper) and
T10 (lower) inferred from apatite and zircon FT thermochronological analyses, respectively
Calculation results of the temporal change in temperature at the two locations of x = 0 cm
(D0) and x = 10 cm (D10) for the combination of µ (0.6) and H (3 km) parameters are shown as
representative cases in Fig. 4. These locations are chosen to approximate the positions of the
"gouge-1" and a surrounding rock sample, respectively. For any combinations of µ and H
parameters, the time during which the temperature in a specific location is maintained at its
maximum is almost invariant. The temperatures at D0 and D10 are preserved at the
maximums of T0 and T10 (named Tmax0 and Tmax10) for the order of ~10^2 sec and ~10^4 sec,
as indicated by shaded bands in Fig. 4, respectively. Fig. 5 shows the variations of Tmax0 and
Tmax10 for the combinations of µ (0.1-0.9) and H (1-3 km) parameters.
Whether FTs in apatite and zircon are annealed or not depends on the temperature and
duration of heating. Assuming that the frictional heat caused by an associated single palaeo-
earthquake event was responsible for the thermochronologic difference between the “gouge-
1” and other samples, calculation results above indicate that the effective heating duration
for the samples at D0 and D10 at Tmax0 and Tmax10 are estimated as the order of ~10^2 sec and
~10^4 sec, respectively. Note that the effective heating time is significantly longer than the
slip duration, and that FTs in minerals in the distance to the frictional centre are not
necessary annealed instantly due to the frictional slip event. The estimates of heating
durations and the kinetic relation of time-dependent FT annealing temperature of apatite
and zircon (Laslett and Galbraith, 1996; Yamada et al., 2007b) give the following constraints
on the Tmax0 and Tmax10 at the secondary heating event (Yamada et al., 2009). For the D0
sample, the fact that apatite FT age was totally reset although zircon FT age was not
indicates that Tmax0 is in the range of 400°C to 750°C (for the heating duration of ~10^2 sec).
For the Dmax10 sample, the fact that both apatite and zircon FT ages were not reset indicates
that Tmax10 does not exceed 250 °C (for ~10^4 sec). These constraints on Tmax0 and Tmax10 are
satisfied with the limited cases of µ > 0.6 for H = 2 km, and 0.4 < µ < 0.7 for H = 3 km, shown
as shaded areas in Fig. 5.
4. Discussion
The effect of the pore water is not taken into account in the modelling above. If the pore
water exists at the “gouge-1” zone, the temperature will be less than that in the dry
condition calculated above, because the pore pressure decreases the stress applied on the
fault and the frictional heat is diffused by fluid flow. Therefore, the estimate of the frictional
strength in the dry condition can be regarded as a lower bound. The increase in the total
amount of slip with the same slip velocity will considerably raise the temperature in the
fault zone. If the total amount of slip is doubled compared with the case for the above
calculation (= 5 m), the estimated increase in T0 and T10 is almost doubled and the estimate
of the coefficient of friction is reduced to almost half. Even in this case, however, the
estimated coefficient of friction is still large (µ > 0.3 for H = 2 km; 0.2 < µ < 0.4 for H = 3 km)
compared with that of 0.1-0.2 (Lachenbruch and Sass, 1980) and 0.04-0.08 (Kano et al., 2006).
Our estimates are obtained by assuming that the thermal anomaly found in “gouge-1” zone
is attributed to the frictional heating during a single earthquake associated at ~32 Ma
(Yamada et al., 2009). Although a number of earthquakes should have occurred thereafter
along the Atotsugawa Fault that remains active to date, an amount of heat generated by
each of these quakes might be insufficient to reduce FT length in apatite because the
partially annealed tracks are not observed for the "gouge-1" sample. The coefficient of
friction for the earthquakes occurred after 32 Ma are, therefore, inferred as < 0.6 for H = 2
km, and < 0.4 for H = 3 km. In addition, the effect of accumulated residual heat generated by
a number of earthquakes should be taken into account if the next heat generation may occur
before the temperature in the fault zone is reduced to the ambient temperature due to the
thermal diffusion in rocks. This effect should, however, be negligible considering the
recurrence interval of general active faults in Japan, ranging from 1000 to 10000 years
(Research Group for Active Faults of Japan, 1991) that would be sufficiently long for the
thermal diffusion.
Our modelled estimates of the coefficient of friction are approximately consistent with that
obtained in laboratory friction experiments on rocks (Byerlee, 1978). As for the Atotsugawa
fault, Mizoguchi et al. (2007) obtained the similar frictional strength of 0.5-0.6 by laboratory
friction experiments using fault gouge samples taken from the Atotsugawa borehole core
samples at a depth of 326 m, located near to the FT samples of Yamada et al. (2009). This
coincidence of frictional strength between the nature and laboratory has rarely reported in
the past. In previous studies, frictional strengths of natural faults are estimated much lower
than those in the laboratory (Lachenbruch and Sass, 1980; Kano et al., 2006; d’Alessio et al.,
2003). At the outcrop of the Atotsugawa fault in Yamada et al. (2009), however, the other
five gouges were not heated enough by frictional slip to reset their ages. We suggest that the
frictional strengths of the fault during earthquakes when the other gouges were activated
were less than that for the “gouge-1”. The variety of frictional strength of fault with every
event might reflect the complicated earthquake generation processes.
5. Conclusions
The fission-track analysis on fault-related rocks collected from a outcrop of the Atotsugawa
fault without visible pseudotachylyte layers revealed that the apatite FT age of the a gouge
sample is exceptionally younger than those of the surrounding other rocks (fault gouge,
fractured rocks and host rocks) in the vicinity, although the zircon FT age is well concordant
with other samples. To explain the thermal anomaly identified at this gouge sample, we
modelled the temporal change in the temperature in and out of the gouge after an associated
earthquake generated the frictional heat. The calculation results indicate that the effective
heating time is significantly longer than the slip duration, and that FTs in minerals in the
distance to the frictional centre are not necessary annealed instantly due to the frictional slip
event. The estimate of frictional strength for the “gouge-1” is larger than 0.6 for H = 2 km,
and between 0.4 and 0.7 for H = 3 km, which is similar to that obtained in laboratory friction
experiments using the gouge samples taken from the Atotsugawa Fault.
6. Acknowledgment
We are grateful to Drs. E. Fukuyama, H. Negishi and N. Hasebe for useful discussion and
help in preparing the manuscript.
7. References
Byerlee, J. D. (1978). Friction of rocks, Pure Appl. Geophys., Vol. 116, pp. 615-626
Camacho, A., McDougall, I., Armstrong, R., Braun, J. (2001). Evidence for shear heating,
Musgrave block, central Australia, J. Struct. Geol. Vol. 23, pp. 1007-1013
d’Alessio, M. A., Blythe, A. E., Bürgmann, R. (2003). No frictional heat along the San Gabriel fault,
California: Evidence from fission-track thermochronology, Geology Vol. 31, pp. 541-544
Doke, R., Takeuchi, A. (2009). The latest event at the eastern part of the Atotsugawa fault, inferred
from the outcrops at Sako, Hida City, central Japan (in Japanese with English abstract),
The Quaternary Res., Vol. 48, pp. 11-17
Geographical Survey Institute, Japan (1997). Crustal deformation in the Chubu and the Hokuriku
regions, Japan, Rep. Coord. Comm. Earthquake Predict. Vol. 57, pp. 520-524
Heaton, T. H. (1990). Evidence for and Implications of self-healing pulses of slip in earthquake
rupture, Phys. Earth Planet. Inter. Vol. 64, pp. 1-20
Kano, Y., Mori, J., Fujio, R., Ito, H., Yanagidani, T., Nakao, S., Ma, K. -F. (2006). Heat signature
on the Chelungpu fault associated with the 1999 Chi-Chi, Taiwan earthquake, Geophys.
Res. Lett. Vol. 33, doi:10.1029/2006GL026733
Lachenbruch, A. H., Sass J. H. (1980), Heat flow and energetics of the San Andreas fault zone, J.
Geophys. Res. Vol. 85, pp. 6185-6222
Laslett, G. M., Galbraith, R. (1996). Statistical modelling of thermal annealing of fission tracks in
apatite, Geochim. Cosmochim. Acta Vol. 60, pp. 5117-5131
Matsuda, T. (1975). Magnitude and recurrence interval of earthquakes from a fault (in Japanese
with English abstract), Zishin, J. Seisol. Soc. Japan, Vol. 28, pp. 269-283
Matsuda, T., Goto, A., Kano, T. (1998). Fission-track thermochronology of Jurassic granitic rocks
located in central part of Hida belt (in Japanese), Abstract of 105th Annual Meeting of
Geol. Soc. Jpn., Nagano, Japan
Matsu'ura, R., Nakamura, M., Karakama, I. (2006). Reexamination of hypocenters and
magnitudes for historical earthquakes Part 8 (in Japanese with English abstract),
Programme and Abstracts, The Seismol. Soc. Japan, 2006, Fall Meeting, B024
Mizoguchi, K., Fukuyama, E., Kitamura, K., Takahashi, M., Masuda, K., Omura, K. (2007).
Depth dependent strength of the fault gouge at the Atotsugawa fault, central Japan: A
possible mechanism for its creeping motion, Phys. Earth Planet. Inter. Vol. 161, pp. 115-
125
Murakami, M., Yamada, R., Tagami. T. (2002). Detection of frictional heating of fault motion by
zircon fission track thermochronology, Geochim. Cosmochim. Acta Vol. 66, pp. A537
Murakami, M., Tagami, T. (2004). Dating pseudotachylyte of the Nojima fault using the zircon
fission-track method, Geophys. Res. Lett. Vol. 31, doi:10.1029/2004GL020211
Ongirad, H., Yasue, K., Takeuchi, A., Nasu, T., Takami, A. (2001). A newly found fault outcrop
at the central part of the Atotsugawa fault, central Japan (in Japanese with English
abstract), Active Fault Res. Vol. 20, pp. 46-51.
Research Group for Active Faults of Japan (1991). Active faults in Japan: Sheet maps and
inventories (in Japanese with English abstract), (revised ed.) 437 pp., Univ. Tokyo,
Tokyo, ISBN 4130607006, Tokyo, Japan
Scholz, C.H., Beavan, J., Hanks, T. C. (1979). Frictional metamorphism, argon depletion, and tectonic
stress on the Alpine fault, New Zealand, J. Geophys. Res. Vol. 84, pp. 6770-6782.
Schön, J. H. (1996). Physical properties of rocks, 18: fundamentals and principles of
petrophysics, In: Handbook of Geophysical Exploration, Pergamon, ISBN 0-08-044346-
X, New York, USA
Sibson, R. H. (1975). Generation of pseudotachylyte by ancient seismic faulting, Geophys. J. R.
Astron. Soc. Vol. 43, pp. 775-794
Tanaka, A., Yano, Y., Sasada, M., Okubo, Y., Umeda, K., Nakatsuka, N., Akita, F. (1999).
Compilation of thermal gradient data in Japan on the basis of temperatures in boreholes (in
Japanese with English abstract), Bull. Geol. Surv. Jpn. Vol. 50, pp. 457-487
Usami, T. (1987). Materials for Comprehensive List of Destructive Earthquakes in Japan (New
Edition) (in Japanese), 434 pp., Univ. Tokyo, ISBN 413060712X, Tokyo, Japan
Wagner, G.A., Van den haute, P. (1992). Fission Track-Dating, Kluwer Academic Publishers,
285 pp. ISBN-10: 079231624X, Dordrecht, Netherlands
Yamada, R., Matsuda, T., & Omura, K. (2007a), Apatite and zircon fission-track dating from the
Hirabayashi-NIED borehole, Nojima Fault, Japan: evidence for anomalous heating in
fracture zones, Tectonophysics, Vol. 443, pp. 153-160
Yamada, R., Murakami, M., Tagami, T. (2007b). Statistical modelling of annealing kinetics of
fission tracks in zircon; Reassessment of laboratory experiments, Chemical Geology, Vol.
236, pp. 75-91
Yamada, R., Ongirad, H., Matsuda, T., Omura, K., Takeuchi, A., Iwano, H. (2009). Fission-
track analysis of the Atotsugawa Fault (Hida Metamorphic Belt, central Japan):
fault-related thermal anomaly and activation history, In: Thermochronological
Methods: From Paleotemperature Constraints to Landscape Evolution Models, Lisker, F.,
Ventura, B., Glasmacher, U. A. (Eds.), 331-337, The Geol. Soc., London, Special
Publications, 324, ISBN 978-1-86239-285-4, London, UK
6
Heat Transfer in Freeze-Drying Apparatus

Roberto Pisano, Davide Fissore and Antonello A. Barresi
Dipartimento di Scienza dei Materiali e Ingegneria Chimica, Politecnico di Torino
Italy
1. Introduction
Freeze-drying is a process used to remove water (or another solvent) from a frozen product,
thus increasing its shelf-life. It is extensively used in pharmaceuticals manufacturing, to
recover the active pharmaceutical ingredient (and the excipients) from an aqueous solution,
as well as in some food processes, because of the low operating temperatures that allow
preserving product quality. Moreover, the freeze-dried product has a high surface area and
can be easily re-hydrated.
In this chapter we focus on pharmaceuticals manufacturing, where the solution containing
the product is generally processed in vials, placed over the shelves in a drying chamber.
However, it is worth stating that, in industrial practice, other loading configurations can be
used to carry out the process, thus this study will be extended also to the case where vials
are loaded on trays, or the solution is directly poured in trays. The process consists of three
consecutive steps, namely:
1. Freezing: product temperature is lowered below the freezing point and, thus, most of
the solvent freezes, forming ice crystals. Part of the solvent can remain bounded to the
product, and must be desorbed. Also the product often forms an amorphous glass
which can retain a high amount of water.
2. Primary drying: in this step the pressure in the drying chamber is lowered, thus causing
ice sublimation. This phase is usually carried out at low temperature (ranging, in most
cases, from -40°C to -10°C) and, as sublimation requires energy, heat is transferred to
the product through the shelf, by acting on the temperature of the fluid flowing in the
coil inserted in the shelf.
3. Secondary drying: when the sublimation of the ice has been completed, shelf
temperature is raised (e.g. to 20-40°C) and chamber pressure is further decreased to
allow the desorption of the water bounded to the product, thus getting the target
moisture in the product.
The freeze-dryer comprises the drying chamber and a condenser where the water vapour is
sublimated on some cold surfaces in order to decrease the volumetric flow-rate arriving to
the vacuum pump. The pressure in the chamber can be modified either by acting on a valve
placed on pump discharge, or using the so called “controlled leakage”, i.e. manipulating the
flow rate of nitrogen (or another gas) introduced in the chamber. The drying chamber can be
isolated from the condenser by means of a valve that is usually placed in the duct
connecting the chamber to the condenser (Mellor, 1978; Jennings, 1999; Oetjen & Haseley,
2004; Franks, 2007).
Although it is generally considered a “soft” drying process, because of the low operating
temperatures, the heat transfer to the product has to be carefully controlled in order to avoid
product overheating. If fact, product temperature has to be maintained below a limit value
to avoid the occurrence of undesired events. In case of products that crystallize during
freezing, the limit temperature corresponds to the eutectic point: the goal is to avoid the
formation of a liquid phase and the successive boiling due to the low pressure. In case the
product remains amorphous during freezing, the maximum allowed product temperature is
close to the glass transition temperature in order to avoid the collapse of the dried cake: this
value can be very low, and is also dependent on the residual moisture. The occurrence of
collapse can increase the residual water content in the final product and the reconstitution
time, beside decreasing the activity of the pharmaceutical principle; moreover, a collapsed
product is often rejected because of the unattractive physical appearance (Pikal & Shah,
1990; Wang, 2000; Rambhatla et al., 2005; Sadikoglu et al., 2006). Beside product
temperature, also the residual amount of ice in the product has to be carefully controlled
during primary drying, thus identifying the ending point of this phase. In fact, if shelf
temperature is increased too early to the value required by the last phase of the cycle,
product temperature may exceed the maximum allowed value, thus causing melting or
collapse. Also in this case heat transfer to the product plays a key role as, at steady-state, the
following equation holds:
J q = ΔH s J w (1)
where Jq is the heat flux to the product, Jw is the solvent flux from the product to the
chamber, and ΔHs is the heat of sublimation.
In this chapter, thus, we will focus on heat transfer in vial, as well as bulk, freeze-drying as it
is one of the key factor affecting product dynamics and temperature. In particular, the
results previously presented by Pikal (2000) for vial freeze-drying are here extended.
The techniques allowing to calculate the heat transfer parameters, or to estimate their values
by means of experiments, will be briefly reviewed. We will point out that the heat flux
between the heating shelf and the container is the result of several mechanisms that depend
on dryer and container geometry, as well as on pressure and temperature of the
surrounding gas. Moreover the heat flux to the batch of vials is far from being uniform in a
freeze-dryer: the implications on recipe design and scale-up will be finally addressed.
2. Theoretical calculation of heat flux in vial freeze-drying

The heat flux to the product is proportional to the difference between the heating fluid
temperature (Tfluid) and the product temperature at the vial bottom (TB):
J q = K v (Tfluid − TB ) (2)
where Kv is the heat transfer coefficient. It has to be highlighted that in eq. (2) it has been
assumed that product temperature is uniform in the radial direction, as it has been
demonstrated by means of mathematical simulation and experimental investigations (Pikal,
1985; Sheehan & Liapis, 1998).
Various equations can be found in the literature to calculate the heat exchange coefficient Kv
in case the vial is placed directly over the shelf, i.e. if no tray is used. These relationships are
still valid in case the solution is directly poured in a tray.
Heat Transfer in Freeze-Drying Apparatus 93
Firstly, the heat transfer coefficient between the heating shelf and the bottom of the vial (Kv′)
is calculated as the sum of three terms:
K v' = Kc + K r + K g (3)
corresponding to the various heat transfer mechanisms between the fluid and the vial
bottom, namely the direct conduction from the shelf to the glass at the points of contact (Kc),
the radiation (Kr), and the conduction through the gas (Kg). In this study, we assume that all
these contributions can be referred to the same heat transfer area.
According to Smoluchowski theory, as outlined by Dushman & Lafferty (1962) and reported
by Pikal et al. (1984), Kg is a function of the average distance between the bottom of the vial
and the shelf (ℓ) and of chamber pressure:
αΛ0 Pc (4)
Kg =
⎛ αΛ ⎞
1 + A ⎜ 0 ⎟ Pc
⎝ λ0 ⎠
The parameter α is a function of the energy accommodation coefficient and of the absolute
temperature of the gas:
ac 273.2 (5)
α= ⋅
2 − ac T
When other gases, beside water vapour, are present in the drying chamber (e.g. nitrogen
entering the chamber when controlled leakage is used to regulate the pressure), eqs. (4) and
(5) have to be modified to take into account gas composition, as total pressure is due to the
sum of the contribution of water and nitrogen partial pressures (Brülls & Rasmuson, 2002).
It has to be remarked that the bottom of the vial is not flat, and the curvature depends on the
type of vial: thus, the value of Kg is a function of the type of vial considered.
Beside conduction in the gas, there are two radiative heat fluxes towards the product, one
from the shelf upon which the vials rest, and the other from the top. Each flux is
proportional to the difference in the fourth powers of the absolute temperatures of the two
surfaces, and to the effective emissivity for the heat exchange, which depends on the relative
areas of the two surfaces, their emissivities, and a geometrical view factor. According to
Pikal et al. (1984) the radiative heat transfer coefficient can be written as:
K r = 4 ( es + ev ) κ T 3 (6)
While the values of Λ0, λ0, and es can be found in the literature, the values of the parameters
Kc, ac, ℓ, and ev (in case radiation from upper shelf plays an important role), have to be
determined by regression analysis of experimental data, even if some values can be found in
the literature for some types of vials (Pikal et al., 1984).
It follows that Kv′ depends essentially on the operating conditions at which drying is carried
out: chamber pressure, shelf temperature and, in turn, temperature of chamber gas.
However, according to literature (Hottot et al., 2005) and to the results shown in the
following, the dependence of Kv′ on chamber gas temperature is not relevant. On the
contrary, the heat transfer between the shelf and the container significantly varies with
chamber pressure, and this dependence can be well described by a nonlinear equation with
the following structure:
C 2 ⋅ Pc
K v' = C 1 + (7)
1 + C 3 ⋅ Pc
where:
⎧
⎪
⎪C 1 = Kc + 4κ T ( eS + ev )
3
⎪
⎪ ac 273.2
⎨C 2 = Λ0 (8)
⎪ 2 − ac T
⎪ ⎛
⎪C 3 = A ⎜ Λ 0 ac 273.2 ⎞
⎜ λ 2−a ⎟
⎪⎩ ⎝ 0 c T ⎟⎠
At this point, K v' can be calculated, whichever are the processing conditions, provided that
ac, Kc and ℓ (and thus C1, C2 and C3) are known.
If vials are loaded on a tray, the shelf-vial system can be described as a set of resistors in
series, whose total resistance is the sum of their individual resistances. It follows that the
resistance for the mentioned system can be expressed as the sum of the resistance due to
tray thickness, as well as to the shelf-tray and tray-vial heat transfers that can be both
described by equations similar to eq. (7). Provided that the heat transfer between shelf and
tray ( K v' ,1 ), and from tray to vials ( K v' ,2 ), can be described by the same equations, we can
define K v' as an effective heat transfer coefficient that accounts for both contributions:
1 1 1 (9)
= +
K 'v K 'v ,1 K 'v ,2
It is worth noticing that, as a consequence of this assumption, the value of the global
coefficients of eq. (7) (i.e. C1, C2 and C3), as determined by a regression analysis of
experimental observations, are effective values that take into account both the heat transfer
between shelf and tray and tray and vials.
The coefficient K v* (from the shelf surface to product bottom) can be calculated by taking
into account the conduction in the glass at the bottom of the vial, as well as through the tray
(if it is used):
−1
Jq ⎛ 1 s g stray ⎞
K v* = =⎜ ' + + ⎟ (10)
Tshelf − TB ⎜⎝ K v λg λtray ⎟⎠
In addition, if the considered driving force for the heat transfer is the temperature difference
between the heating fluid and the product, we have also to include the resistance to heat
transfer through shelf (1/ks). Therefore, eq. (10) has to be substituted by:
−1
⎛ 1 s g stray 1 ⎞
Kv = ⎜ ' + + + ⎟ (11)
⎜ K v λg λtray ks ⎟
⎝ ⎠
Finally, it has to be remarked that the product in the vial can receive heat from the walls of
the container, as well as from chamber walls, due to radiation. Both contributions do not
appear explicitly in eq. (2), but the coefficient Kv can be regarded as an effective heat transfer
coefficient, weighing up these additional heat fluxes (Velardi & Barresi, 2008). However, as
it will be discussed in the following sections, it is worth remarking that the contribution of
radiative heat to the total energy balance of the vial can vary with its position over the shelf.
This phenomenon, together with the presence of fluid temperature gradients along the shelf,
is responsible for lot unevenness.
3. Experimental methods for Kv measurement

Various methods have been proposed in the past to determine the value of the coefficient Kv
(or that of K 'v ). They are briefly reviewed in the following.
A very simple test that allows determining the coefficient Kv consists of preparing a batch of
vials filled with the product, a placebo, or simply water and, then, to carry out the primary
drying for a time interval (∆t) and to measure the weight loss (∆m) and the temperature of
the ice at the vial bottom (TB). The coefficient Kv is given by:
Δm ⋅ ΔH s
Kv = (12)
Δt ⋅ (Tfluid − TB ) ⋅ Av
As the heat transfer coefficient is a function of chamber pressure, at least three

measurements at three different values of Pc are required to calculate the global coefficients
of eq. (7).
During the drying, the temperature of the heating fluid can vary along the shelf as the heat
is transferred from the technical fluid to the product. However, it must be said that
generally the inlet-outlet fluid temperature difference observed under full-load conditions is
small (typically lower than 1 K) and of the same order of magnitude of temperature sensor
uncertainty. A similar behaviour has been observed in all the experiments whose results are
reported in this chapter. It follows that it is not necessary to take into account explicitly the
fluid temperature gradient along the shelf in the calculation of Kv, as its effect can be
included in the uncertainty of the parameter.
On the other hands, the value of the coefficient Kv is not the same for all the vials of the lot,
as a consequence of the different contributions of the various heat transfer mechanisms: as
an example, the vials at the edges of the shelf receive also radiant heat from the chamber
walls, while the vials located in the central part of the batch are heated only from the heating
fluid (Pikal, 2000; Rambhatla & Pikal, 2003; Gan et al., 2005a, 2005b). As a consequence, a
different evolution of product temperature can be observed in vials located in various
positions of the batch. Therefore, to estimate accurately Kv for each class of vials, product
temperature has to be monitored, and thus thermocouples must be inserted both in vials
placed in the core and at the edge of the lot.
According to their position over the shelf, the vials of a lot can be classified in various
groups depending on the different heat transfer mechanisms involved. In particular, in case
vials are confined by a metal or plastic band, we can generally identify five zones (see Figure
1), which are characterized by different heat transfer mechanisms, as reported in Table 1.
A further refinement might be introduced to distinguish edge vials that are exposed to
different walls of the drying chamber. This is particularly true for laboratory scale
equipment, where the dryer door is not made of steel (like the other chamber walls), but of
Plexiglas. Since the two materials have a significantly different value of emissivity (0.36 vs.
0.95), vials close to the door may show a higher value of Kv with respect to the rest of the lot
(Rambhatla & Pikal, 2003). On the contrary, in a manufacturing plant, the dryer door is
usually made of steel and, thus, edge vials are less uneven. Nevertheless, it must be said that
in a large scale equipment only a limited fraction of vials are radiated, while in a laboratory
dryer (where the fraction of edge vials is much greater) the contribution of the radiative heat
can be strongly reduced by a proper shielding. These differences, however, must be taken
into account during scale-up of a recipe from laboratory to industrial scale dryers.
metal or plastic band
B D E E E E E
C D E E E E E
B D E E E E E
C D D D D D D
A B B B B B B
Fig. 1. Scheme of a lot, wherein vials have been classified depending on their position and
on the additional heat transfer mechanisms involved
Position Additional mechanisms to vial heat transfer

Vial
over the radiation from contact with the contact with
group
shelf chamber walls tray band “hot” vials
A corner yes yes yes
B peripheral yes yes yes
C peripheral yes no yes
D core no no yes
E core no no no
Table 1. Characteristics of the various vial families with respect to the position on the shelf
and the mechanisms of heat transfer involved
It must be remarked that the gravimetric measurement is the only method that gives a
detailed picture of single vial behaviour and of intervial variability, thus allowing to identify
the various vial families. A heat flux transducer can also be used for monitoring the
dynamics of the single vials composing the lot (Chen et al., 2008), but its use implies that the
monitored vials are loaded on the sensor surface altering de facto the heating conditions.
Equation (12) can be still used when the sublimation flux (i.e. Δm Δt ⋅ Av ) is obtained by
other means, for example using the Tunable Diode Laser Absorption Spectroscopy (TDLAS)
or the pressure rise test technique.
The Doppler-shifted near-infrared absorption spectroscopy is used for measuring the water
vapour concentration and gas flow velocity in the duct (that connects drying and condenser
chamber). At this point, these variables can be used, known the cross-sectional area of the
duct, to estimate the vapour flow rate (Kessler et al., 2006; Gieseler et al., 2007). It must be
remarked that TDLAS does not measure directly the mass flow, but only the gas flow
velocity in a limited number of points; therefore, the reliability of the vapour flow rate
estimation depends on how well the velocity profile is known. This is a critical issue, since
the velocity profile continuously develops in the duct, tightly depends on processing
conditions, can be strongly modified by the presence of Clean In Place systems, and even at
the duct exit is usually far from the parabolic profile corresponding to fully developed flow.
Other drawbacks of this technique are the high investment costs, the difficult to retrofit
existing units, and the difficulties in its calibration.
To measure the vapour flow rate (and hence the heat transfer coefficient), an alternative
solution, that can be easily implemented also in existing devices, is the pressure rise test
technique: the valve in the duct connecting the drying chamber to the condenser is closed
for a short time interval, and, thus, pressure in the chamber increases due to vapour
accumulation. Once the pressure rise curve has been acquired, the vapour flow rate (Jw) can
be calculated by simply evaluating its slope at the beginning of the test (Fissore et al., 2011a):
Vc M w dpw ,c (13)
Jw =
Asub RTc dt t = t0
This measure of Jw is reliable as it is based on the direct measurement of the effect of the
sublimation flow, the pressure increase, described by a simple physical law. Unlike TDLAS,
the pressure rise technique can be used in both laboratory and industrial scale dryers,
provided that drying and ice condenser chamber are separated, without any modifications
of existing units. Figure 2 shows an example of pressure rise curves measured (at various
time during the primary drying phase) in a laboratory and in an industrial scale dryer.
These results were obtained during the drying of a 5% by weight mannitol solution. As
stated before, the vapour flow rate is related to the slope of the pressure profile, which, as
expected, decreases as the drying proceeds.
Fig. 2. Examples of pressure rise curves measured at various times during the primary
drying stage in (left-side graph) a laboratory and (right-side graph) an industrial dryer
To calculate Kv from the vapour flow rate, the product temperature at the container bottom
has to be measured, for example, inserting thermocouples in some vials. For this purpose,
the use of wireless technology is promising (Corbellini et al., 2010; Schneid & Gieseler, 2008)
as it is compatible with the restrictions of industrial equipment that typically uses automatic
loading systems.
The pressure rise curve can be also described by a mathematical model, whose parameters
(e.g. product temperature at the beginning of the test, and Kv) are retrieved by looking for
the best fit between measured and calculated values of chamber pressure (Milton et al., 1997;
Chouvenc et al., 2004; Tang et al., 2006; Velardi et al., 2008). The optimal selection of
sampling frequency and duration of the test has been discussed by Fissore et al. (2011a),
who also proposed a modification of the Dynamic Parameters Estimation algorithm (Velardi
et al., 2008) to cope with the problem of ill-conditioning. It must be said that while the
estimation of Jw from the slope of the pressure rise curve can be also carried out with pure
ice, the previous algorithms (which take advantage of a model) require that a product,
which gives a mass flow resistance, is used. In addition, unlike previous techniques, the
pressure rise, coupled with one of the above cited algorithms, does not require an
independent measurement of product temperature.
It is worth noticing that the gravimetric measurement is the only method that allows to
determine the distribution of the values of Kv in the lot, while both the TDLAS sensor and
the pressure rise test based methods allow to determine only a mean value of Kv for the
batch. Nevertheless, it must be said that the fraction of edge-vials, and hence its contribution
to the mean value of Kv, is becoming less and less important as the size of the dryer becomes
larger.
In addition, it must be remarked that the above mentioned methods, differently from the
gravimetric test, allow to get an estimation of Kv vs. Pc, even in only one run (Kuu et al.,
2009; Pisano et al., 2010) if chamber pressure is properly varied during the test (e.g. as a
multi-step constant function), without requiring an excessive effort from the users.
Furthermore, these methods allow getting an estimation of Kv whichever is the scale of the
equipment. In particular, this eases the problem of Kv measurement for industrial scale
apparatus (as well as of process transfer) where the gravimetric procedure cannot be used
since the intervention of the user is limited. Therefore, following on from what stated above,
the use of global methods is strongly suggested for large scale apparatus.
4. Heat transfer in bulk and vials freeze-drying

In the following the influence of processing conditions, vial and tray characteristics and
equipment size will be discussed in detail, presenting a selected set of original experiments
planned by the authors to highlight the previous aspects.
In particular, experiments were carried out using both tubing vials (whose characteristics
are reported in Table 2) and trays made of stainless steel or polystyrene. In case of freeze-
drying in vials, it was also investigated the effect of various loading configurations (e.g.
comparing the value of Kv for vials processed directly over the heating shelf or in trays), as
well as of different pieces of equipment, not only on the mean value of Kv, but also on
intervial variability. With this respect, the gravimetric measurement is the only method that
allows to get a complete description of the heat transfer in the lot. However, results obtained
by the pressure rise test technique will be also presented, confirming the capacity of this
method to get a quick and reliable estimation of the parameter of interest, not only in a
laboratory equipment (LyoBeta 25TM by Telstar, Terrassa, Spain), but also in an industrial
scale dryer where gravimetric experiments cannot be easily carried out. Previous results had
already shown this, evidencing at the same time that inaccuracies of the PRT model can
affect its reliability (Tang et al., 2006).
Properties vials V1 vials V2 vials V3 vials V4

External diameter, mm 14.25 16.20 24.14 14.12
Thickness of side wall, mm 1.00 0.80 1.10 0.90
Thickness of vial bottom wall, mm 0.70 1.41 1.41 0.60
Estimated thickness of bottom vial gap (ℓ), mm 0.20 0.32 0.38 0.34
Table 2. Geometrical characteristics of the various glass vials investigated in this study
4.1 Influence of processing conditions on Kv
Figure 3 (left-side graph) displays the value of Kv vs. Pc measured by gravimetric way for a
lot of vials (type V1), which was directly loaded on the heating shelf and surrounded by a
metal band. Vials placed in the central part of the shelf, i.e. those far from the edges where
the effect of radiative heat coming from chamber walls can be relevant, are considered.
According to Brülls & Rasmuson (2002) and Hottot et al. (2005), Kv increases with Pc,
because of a higher heat conduction through the gas trapped between the tray and glass vial
bottom.
In Figure 3 (left-side graph), it can be also observed a fairly good agreement between the
value of Kv vs. Pc as measured by the pressure rise test technique (coupled with a modified
version of Dynamic Parameters Estimation algorithm, DPE+) and by gravimetric way. The
values of Kv vs. Pc predicted by the mathematical model described in section §2 is also
displayed. The model parameters ac, Kc and ℓ and, thus, the global coefficients C1, C2 and C3
(that describe the pressure dependence of Kv for vials V1) were obtained by a non-linear
regression of experimental data. The results so obtained are summarized in Table 3.
Fig. 3. Effect of chamber pressure on the value of Kv for the vial V1 (left-side graph) and a
stainless steel tray (right-side graph) processed in the laboratory freeze-dryer. The values
calculated using the theoretical model, and the parameters of Table 3, are also displayed
(solid line) together with the value estimated by pressure rise test technique ()
In addition, it must be noticed that there is a large variance of Kv, which cannot be explained
by the only observational error in the experiment, but is related to the heterogeneity of the
lot. This is confirmed by the contour plot of Figure 4, where it can be observed that, because
of side radiation and heat conduction though the metal frame, vials at the edge of the batch
have a higher value of Kv than those placed in the core. In particular, as already observed by
Pisano at al. (2008), these additional contributions increase only the sublimation rate of vials
placed just next to the additional heat sources, while their effect is reduced after the first
row. However, this issue will be better discussed in the following section.
container subset C1 C2 C3 σC
1
vials V1 all the lot 10.9 1.4 0.04 4.9

vials V1 B 21.9 1.4 0.04 6.3
vials V1 C 13.6 1.4 0.04 1.5
vials V1 D 9.7 1.4 0.04 0.5
vials V1 E 7.8 1.4 0.04 0.1
metal tray – 1.1 0.7 0.01 –
Table 3. Parameters required to calculate the value of K v' vs. Pc for a metal tray (made of
stainless steel) and the various groups of vials V1
Let’s now focus on the temperature influence (of the gas trapped in the gap at the container
bottom) on the value of Kv. In this study, as the gas temperature reasonably varies from the
product temperature at the interface and Tfluid, it can be approximated to their mean. Figure
5 compares the heat transfer coefficient as measured at two different temperatures of the
heating fluid, and thus of the chamber gas temperature, while the pressure inside the drying
chamber is kept constant. It can be observed that a significant variation in the fluid
temperature produces a modification of Kv that is of the same order of magnitude of the
parameter uncertainty. This confirms that the dependence on gas temperature of Kv is not
significant or, however, is very limited compared with the role of chamber pressure in the
heat transfer.
Fig. 4. Contour plot of the heat transfer coefficient for vials V1 (loaded directly over the
heating shelf) as measured over a lot. The primary drying phase was run at Tfluid=263 K and
Pc=5 Pa. Results refer to a half of the batch
Fig. 5. Effect of Tfluid, and in turn of chamber gas temperature, on the value of Kv for the vial
V2 processed in the laboratory freeze-dryer. Both measurements were carried out at Pc=10
Pa, while the fluid temperature was set to: ( ) Tfluid=275 K; ( ) Tfluid=253 K
Figure 3 (right-side graph) displays the value of Kv vs. Pc in case the process is carried out in
trays. It can be observed that, similarly to what has been already shown for vials V1, the heat
transfer coefficient is strongly influenced by the total chamber pressure. However, the value
of Kv for trays is significantly lower than that of vials. This is probably due a bad contact
between the tray bottom and the heating plate, as the resistance to heat transfer of the
stainless steel tray is negligible. In addition, as already observed by Bruttini et al. (1991) for
bulk freeze-drying of a solution of cloxacillin monosodium salt, it is confirmed that, for
freeze-drying in trays, the heat is mostly transferred by the conduction through the thin
layer of gas that separates shelf and tray surfaces.
In case of bulk freeze-drying, wherein the solution is poured in a metal tray, the heat
transfer between the heating fluid and the product can be still calculated as the sum of the
three terms of Equation (3). Therefore, the same mathematical model used for vials freeze-
drying is here used to determine (via non-linear regression of experimental data) the
optimal combination of ac, Kc and ℓ that describes the pressure dependence of Kv for trays.
The results so obtained are reported in Table 3.
4.2 Non uniformity of the lot

In the previous section it has been observed that the value of Kv significantly varies with the
position of the vial in the lot, see Figure 4. Therefore, following on from what stated in
section §3 (see Figure 1), the lot of vials has been divided in five zones, which are
characterized by different heat transfer mechanisms. The criteria chosen to identify the
various zones are summarized in Table 1.
As previously discussed, a further refinement might be introduced to distinguish vials
radiated by different walls of the chamber, but it has not been considered here: the
correctness of this approach is confirmed by the more detailed statistical analysis that is
presented in the following.
Figure 6 compares the value of Kv vs. Pc for the various groups of vials identified in Table 1.
Vials A were not considered in the analysis as only four samples can be monitored in each
test and, hence, the results are not statistically relevant. As it can be expected, the heat
transfer coefficient of edge-vials (i.e. vials B and C) is much higher than that of vials placed
in the core of the lot (i.e. vials D and E). In fact, according to Table 1, these vials receive an
additional contribution of heat due to radiant energy coming from chamber walls and tray
band, as well as, in case of vials B, the contribution due to conduction along the tray band.
Fig. 6. Values of Kv vs. Pc for the four groups of vials (type V1) identified in Table 1: (a) vials
B; (b) vials C; (c) vials D and (d) E
Now let’s focus on vials E, which are placed in the central part of the lot and, thus, are far
from the edges where the contribution of the radiant heat from chamber walls can be
relevant. Following on from what shown in section 2, the three parameters ac, Kc and ℓ of
vials V1, type E, can be obtained by a non-linear regression analysis of experimental data
(Figure 6, graph d) that describes the pressure dependence of Kv. To obtain the model
parameters of the other three groups of vials, we assume that ac and ℓ, and thus C2 and C3,
are known, and equal to those determined for vials E. This assumption is reasonable, since
these parameters depend on the container and not on the surrounding environment. On the
contrary, Kc, and thus C1, is optimized to get the best fitting between model predictions and
experimental observations. Table 3 summarizes the value of global parameters C1, C2 and C3
for the four groups of vials (V1 type).
Concerning the uncertainty on the parameter Kv, it must be said that ac and ℓ, thus C2 and C3
that define the pressure dependence of Kv, can be estimated, even if with some difficulties.
On the contrary, it is almost impossible to estimate C1 as it is affected by the contact between
the shelf surface and the bottom of the container, as well as the view factors for the radiative
heat. Furthermore, the effect on the final value of Kv due to the uncertainty on C2 and C3 (e.g.
deriving from a different geometry of the container bottom) is less important than that due
to the variability associated to transport phenomena involved in C1. For all these reasons,
we assume that the only responsible for the uncertainty on Kv is the parameter C1.
Nevertheless, it is worth noticing that the uncertainty, or variability, on ac and ℓ (and thus on
C2 and C3) is implicitly included in the uncertainty on the parameter C1, even if it is not
directly expressed. At this point, once the value of C2 and C3 is known, eq. (7) can be used
for retrieving the coefficient C1 of each vial of the lot (under investigation) from the value of
Kv that has been previously measured for vials V1.
Figure 7 displays the distribution of C1 (of vials V1) for the entire lot (graph a), as well as for
the various families of vials of Table 1 (graphs b–e). It can be observed that the various
groups of vials are characterized by a significantly different mean value of C1 and, as it can be
expected, edge-vials have a much higher C1 than that of vials placed in the core of the batch.
Fig. 7. Distribution of C1 of vials V1 for: (a) the entire lot; (b) vials B; (c) vials C; (d) vials D
and (e) vials E. The solid line is the Gaussian distribution calculated according to the mean
value and standard deviation reported in Table 3
In Figure 7 (graph a), the width of the C1 distribution for the entire lot is so high that cannot
be explained by the only observational error in the experiments but, as already discussed
above, is due to the lot unevenness. Thus, the set of data has been divided in the four
subsets previously identified and the distribution curve has been recalculated for each
group of vials (see graphs b–e). The correctness of this approach is confirmed by the fact that
the resulting curves can be well described by Gaussian distributions (that are typically used
for describing the observational error in an experiment) and their width is significantly
reduced. It can be noticed that the variance of C1 for vials B is still relevant (even if much
smaller than that of the entire lot), but it is now difficult to discern the contribution due to
the intrinsic variability of the object of measurement from the uncertainty of the method.
4.3 Role of the vials geometry and material

Following on from what stated in section §2, the shape of the container bottom significantly
affects the value of Kv, as well as its dependence on chamber pressure. To better investigate
this aspect, the heat transfer coefficient of two different types of vials (i.e. V1 and V2), which
have a similar geometry (in terms of external diameter and side-wall thickness), but a
different shape of the bottom, are compared.
The values of Kv vs. Pc for the two types of containers, and in case they are directly loaded
on the heating shelf and processed in the same equipment, are displayed in Figure 8. To
discern the contributions due to the characteristics of the vial from those of the environment,
the analysis is limited to only those vials that are placed in the centre of the lot. It can be
noticed that the two containers have a similar value of C1 even if vials V2 have a thicker vial
bottom wall.
Fig. 8. Comparison between the value of Kv vs. Pc for vials subset E (processed in the
laboratory freeze-dryer) in case of two different container geometries: (z) V1 and ({) V2
type. The values calculated using the theoretical model are also displayed: (solid line) vials
V1 and (dotted line) V2
In particular, as the containers have a similar geometry and the contact between the
container bottom and the shelf plate tightly depends on the characteristics of the vial
surface, this implies that, in this case, the heat conduction (at the points of contact between
the container bottom and the shelf surface) and the radiative heat coming from the heating
plate give a similar contribution to the energy balance.
On the contrary, it can be observed a significantly different pressure dependence of Kv.
This is due to a different contribution of the heat conduction through the gas trapped at
the container bottom to the total energy balance of the vial. According to manufacturer
specifications, the maximum value of the gap thickness is greater in vials
V1 (ℓmax=4.0x10-4 m) and this seems to be in disagreement with experimental results of
Figure 8. Nevertheless, it must be said that the heat conduction through the gas is a function
of the volume of the gas gap, which depends not only on ℓmax, but also on the shape of the
vial bottom. In particular, the visual inspection of the two systems confirms that vials V1,
despite a higher value of ℓmax, are characterized by a flatter gap at the bottom. As a
consequence, the heat transfer resistance due to the gas gap is lower in vials V1, which show
a marked pressure dependence of Kv. As further confirmation of the previous evidences, the
effective value of ℓ, as obtained by a non-linear regression of experimental data, of vials
V1 (ℓ =2.0x10-4 m) is smaller than that of vials V2 (ℓ =3.1x10-4 m).
A final comment concerns the effect of container treatments (e.g. a silane coating of the
internal vial walls) on the value of Kv. Figure 9 compares the overall heat transfer coefficient
of vials V2 in case they underwent a silane coating treatment or not. It can be noticed that
pre-treated vials show a lower value of Kv and this reduction is of the same order of
magnitude for all the four group of vials, thus the coating does not alter the contribution of
the radiative heat to the total energy balance of the vial, but probably only the conduction
through the vial side-wall.
Fig. 9. Heat transfer coefficient of the four groups of vials (V2 type) in case they underwent a
silane coating treatment (void bars) or not (filled bar). The lot of vials is loaded on the heating
shelf and enclosed by an external metal band. The measurement was carried out at Pc=10 Pa
4.4 Role of the configuration used for containers loading

In this section, we compare the value of Kv, as well as the heterogeneity of the lot, in case
various configurations are used for loading the product into the drying chamber; in
particular, the following configurations have been investigated:
• Vials are directly loaded on the heating shelf and enclosed by a metal frame;
• Vials are directly loaded on the heating shelf, but not surrounded by a frame;
• Vials are loaded on a metal tray.
Figure 10 compares the heat transfer coefficient of vials in case they are enclosed by a metal
frame or not (the comparison is carried out for vials V2). In particular, the use of a metal
frame can have a different effect on the heat transfer coefficient of the various groups of
vials. For example, the mean value of Kv for vials E is not significantly modified by the
presence of the surrounding band, while edge-vials (both vials B and C) show a much
higher value of Kv in case no metal frame is used. In fact, this frame acts as thermal shield
since it reduces the radiative heat coming from chamber walls, which are warmer than the
metal frame. The metal band also minimizes the unevenness of edge vials, as they are not
anymore exposed to surfaces with different characteristics and temperature. Nevertheless, it
must be said that edge-vials continue to receive an additional heat flow (due to the
conduction between the vial side-wall and the metal band at the points of contact), but this
has a minor effect with respect to chamber walls radiation. All these effects tightly depend
upon the type of frame used, as well as the geometry of the drying chamber; hence, it is
fundamental to take into account such phenomena during process transfer from one unit to
another one, mainly if the loading configuration is modified.
Fig. 10. Heat transfer coefficient (at Pc=10 Pa) of the four groups of vials (V2 type, silane
treated) in case of two different loading configurations: (void bars) the lot of vials is loaded
on the heating shelf; (filled bar) the lot of vials is loaded on the heating shelf and enclosed
by an external metal band
Figure 11, instead, compares the value of Kv in case vials are loaded on a metal tray or
directly loaded on the heating shelf; the comparison was carried out for vials V3. It can be
noticed that if on one side the use of a tray introduces an additional resistance to the heat
transfer, on the other side it allows to get a more uniform distribution of the heat over the lot
(in Figure 11, the error bar is much smaller in case vials are loaded on a tray), as well as a
strong reduction of the value of Kv for edge-vials due to the presence of the tray band. This
strongly reduces the difference of Kv between vials D and E, as those vials that are located
on the second row of the lot are now in contact with vials that are less warm and, thus,
receive a smaller amount of heat from edge-vials.
Figure 12 (a) compares the value of Kv for vials loaded on the heating plate or in trays made
of different materials, i.e. stainless steel or polystyrene. Even in this case we can observe a
strong reduction of Kv when vials are loaded on a tray. In particular, the value of Kv of
central vials (both group D and E) is not modified by the type of tray used; in fact, the
observed variations are of the same order of magnitude of the error bar. On the contrary, Kv
of edge-vials significantly changes with the type of tray used. This can be explained by a
different heat transfer between the tray band and the container side-walls, which, for
example, can be due to a different emissivity of the tray material.
4.5 Role of the equipment size

Figure 12 (b) compares the heat transfer coefficient of vials processed in laboratory and
manufacturing equipment. Results refer to vials loaded on a plastic tray and in case the drying
is carried out at Pc=10 Pa. It can be observed that the value of Kv of vials located in the central
part of the lot is not significantly modified, whereas edge-vials processed in the laboratory
freeze-dryer have a higher value of Kv than that observed in the large scale freeze-dryer.
This is due to a different contribution of the radiative heat (that, for example, can be caused
by a different geometry of the drying chamber) and has to be taken into account when we
have to transfer a process from one unit to another one, as well as during scale-up
operations, as the fraction of edge-vials into the lot varies with the scale of the equipment.
Fig. 11. Heat transfer coefficient (at Pc=10 Pa) of the four groups of vials (V3 type) in case of
two different loading configurations: (void bars) the lot of vials is loaded on a metal tray;
(filled bar) the lot of vials is loaded on the heating shelf and enclosed by an external metal
band
(a)
(b)
Fig. 12. Comparison between experimental values of Kv for the various groups of vials
(V4 type). (a) Influence of various loading configurations: vials loaded on a stainless steel
tray ( ), a polystyrene tray ( ) and directly on the heating shelf ( ); (b) values measured in
two pieces of equipment of different size: ( ) laboratory and ( ) industrial scale dryer
For the two pieces of equipment investigated in this study, central vials have almost the
same value of Kv, but, in general, it can happen to observe significant differences. In fact,
even if the value of ℓ is independent of the equipment used, the parameters ac and Kc can be
affected by the characteristics of the shelf surface. Thus, once the value of ℓ has been
determined for the laboratory freeze-dryer (using the procedure described in section §2),
and assuming that ac does not change, we can calculate the effective value of Kc for the four
groups of vials in case they are processed in the industrial freeze-dryer. It follows that the
global coefficients C2 and C3, which describe the pressure dependence of Kv, are the same for
both pieces of equipment, while the parameter C1 of peripheral vials is slightly different (as
it depends on the vial-tray and shelf-tray contact, and on radiation effect). The knowledge of
the pressure dependence of Kv in the two pieces of equipment is fundamental to make
decisions during the phase of scale-up of a recipe obtained in a laboratory scale equipment
to an industrial scale freeze-dryer.
5. Conclusion
Following on from what stated in the introduction, the heat transfer to the product has to be
carefully controlled in order to avoid product overheating. This result can be obtained in-
line, using a control system (Pisano et al., 2010, 2011), or building off-line the design space,
i.e. identifying the set of operating conditions (heating fluid temperature and chamber
pressure) that allows preserving product quality. Recently, the use of mathematical
modelling, coupled with few experiments to determine model parameters (e.g. the value of
Kv vs. Pc), has been proposed to get the design space quickly, taking also into account the
effect of parameters uncertainty and/or variability (Giordano et al., 2011).
Due to batch non-uniformity, vials have to be divided in various subsets (as shown
above), which are characterized by the same value of Kv, and the design space (that is
strongly dependent on the value of Kv) has to be built for each group (Fissore et al., 2011b).
Of course, since the design space of edge-vials is more contracted than that of vials placed
in the core, processing conditions has to be chosen according to the design space of those
vials that can be more easily damaged by product overheating (i.e. vials B). The recipe so
designed allows preserving product quality of the entire lot, but the resulting drying time
might be quite long. Therefore, as edge-vials are a small fraction of the lot, it might be
worth referring to central vials to reduce the duration of the process. Of course, this is
advantageous only if the energy saving deriving from a shorter drying time compensate
the loss due to product waste (that derives from the fact that edge vials did not meet
product quality criteria).
It must be remarked that whichever is the tool used to control the process, the resulting
recipe is strongly dependent on the value of Kv, which can vary with operating conditions,
type of container, equipment and configuration used for loading the product.
The experimental investigation carried out in this study has evidenced a strong dependence
of the heat transfer coefficient on processing conditions used for the drying. In particular, it
has been proved that while the gas temperature has a minor effect on the value of Kv, the
role of chamber pressure cannot be neglected. To be used by the model at the basis of the
various tools recently developed for the recipe design, the pressure dependence of Kv has
been described by a mathematical model, whose parameters have been obtained, for
various types of container (both glass vials and trays), by regression analysis of
experimental data.
In addition, the heat transfer of glass vials has been measured in case of various loading
configurations. It has been shown that, as expected, the introduction of a tray significantly
reduces the value of the heat transfer coefficient of the entire lot and, in particular, a more
marked reduction was observed for edge-vials. On the contrary, in case vials are directly
loaded on the shelf, but surrounded by a metal frame, the value of Kv of central vials is not
modified, while that of edge vials is strongly reduced. This is due to the contribution of the
tray band, which acts as thermal shield for the radiative heat coming from chamber walls.
Therefore, it must be remarked that, during the phase of process development, the user
has to take into account that the pressure dependence of Kv does not depend only on the
type of vials, but also on the configuration used for loading the product into the drying
chamber.
In general, the gravimetric procedure gives the best accuracy and robustness, even if it is
more time demanding with respect to other global methods available. However, the use of the
pressure rise test technique is strongly suggested in case of industrial apparatus, where the
gravimetric procedure is not practicable as the intervention of the user (to place temperature
sensors over the lot of vials) is limited. Therefore, it has been shown that the pressure rise
test technique (and in particular the latest developments like the DPE+ algorithm) can be
effectively used for measuring the value of Kv, whichever is the scale of the equipment,
without requiring an excessive effort from the users. In addition, an estimation of the mean
value of Kv is more than enough for an effective description of the heat transfer of the lot, as
the effect of batch non-uniformity in a manufacturing process is less marked. A further
advantage of the pressure rise test technique is that, with respect to other global methods
like TDLAS, it requires no modifications of the equipment and its hardware.
A final comment concerns the problem of scale-up, or process transfer, of a recipe from one
unit to another one. It has been proved that the heat transfer coefficient of a specific
container can varies significantly (mainly for edge vials) with the type of equipment used,
even if the same loading configuration is used. Therefore, if this difference is relevant, the
recipe, which is usually developed in laboratory and has to be transferred on manufacturing
equipment, should be adapted to take into account the different heat transfer of the
containers.
6. Acknowledgment
Development of PRT methods for industrial apparatus has been continuously supported by
Telstar S.A. (Terrassa, Spain), whose contribution of data obtained in large scale apparatus
and financial support for this chapter is gratefully acknowledged. The authors would like to
acknowledge Giovanni Accardo, Salvatore Genco e Daniele Sorce for their valuable support
in the experimental investigation.
7. Nomenclature
ac energy accommodation coefficient
Av cross sectional area of the vial, m2
Asub total sublimation area, m2
C1 parameter expressing the dependence of K 'v from radiation and the contact
between vial bottom and tray surface, J s-1m-2K-1
C2 parameter expressing the pressure dependence of K 'v , J s-1m-2K-1Pa-1
C3 parameter expressing the pressure dependence of K 'v , Pa-1
es emissivity for radiation heat exchange from the shelf to the bottom of the vial
ev emissivity for radiation heat exchange from the shelf to the top of the vial
ΔHs heat of sublimation, J kg-1
Jq heat flux to the product, J s-1m-2
Jw solvent flux, kg s-1m-2
ks heat transfer coefficient between the technical fluid and the shelf, J s-1m-2K-1
Kc heat transfer coefficient due to direct conduction from the shelf to the glass at the
points of contact , J s-1m-2K-1
Kg heat transfer coefficient due to conduction in the gas between the shelf and the vial
bottom , J s-1m-2K-1
Kr heat transfer coefficient between the shelf and the vial due to radiation, J s-1m-2K-1
Kv overall heat transfer coefficient between the heating fluid and the product at the
bottom of the vial, J s-1m-2K-1
K v' overall heat transfer coefficient between the heating shelf and the vial bottom (or
between shelf and tray, and tray and vials), J s-1m-2K-1
K v∗ overall heat transfer coefficient between the heating shelf and the product at the
bottom of the vial, J s-1m-2K-1
ℓ constant effective distance between the bottom of the vial and the shelf, m
m mass, kg
Mw molar mass of water, kg kmol-1
pw,c partial pressure of water in the drying chamber, Pa
Pc chamber pressure, Pa
R ideal gas constant, J kmol-1K-1
sg thickness of the glass at the bottom of the vial, m
stray thickness of the tray bottom, m
t time, s
T temperature, K
TB temperature of the product at the vial bottom, K
Tc temperature of the chamber gas, K
Tfluid temperature of the heating fluid, K
Tshelf temperature of the heating shelf, K
Vc volume of the drying chamber, m3
Greeks
α parameter used to calculate Kg
κ Stefan-Boltzman constant, J s-1m-2K-4
Λ0 free molecular heat conductivity at 0°C, J s-1m-1K-1
λ0 heat conductivity of the water vapour at ambient pressure, J s-1m-1K-1
λg heat conductivity of the glass, J s-1m-1K-1
λtray heat conductivity of the tray, J s-1m-1K-1
σC 1
standard deviation of the parameter C1, J s-1m-2K-1
8. References
Brülls, M., & Rasmuson, A. (2002). Heat transfer in vial lyophilization. International Journal of
Pharmaceutics, Vol. 246, pp. 1-16, ISSN 0378-5173.
Bruttini, R., Rovero, G., & Baldi, G. (1991). Experimentation and modelling of pharmaceutical
lyophilization using a pilot plant. The Chemical Engineering Journal, Vol. 45, pp. B67-
77, ISSN 1385-8947.
Chen, R., Slater, N. K. H., Gatlin, L. A., Kramer, T., & Shalaev, E. Y. (2008). Comparative
rates of freeze-drying for lactose and sucrose solutions as measured by
photographic recording, product temperature and heat flux transducer.

Pharmaceutical Development and Technology, Vol. 13, pp. 367-374, ISSN 1083-7450.
Chouvenc, P., Vessot, S., Andrieu, J., & Vacus P. (2004). Optimization of the freeze-drying
cycle: a new model for pressure rise analysis. Drying Technology, Vol. 22, pp. 1577-
1601, ISSN 1532-2300.
Corbellini, S., Parvis, M., & Vallan, A (2010). In-process temperature mapping system for
industrial freeze dryers. IEEE Transactions on Instrumentation and Measurement, Vol.
59, pp. 1134-1140, ISSN 0018-9456.
Dushman, S., & Lafferty, J. M. (1962). Scientific foundations of vacuum technique, Wiley, ISBN
978-047-1228-03-5, New York, USA.
Fissore, D., Pisano, R., & Barresi, A. A. (2011a). On the methods based on the Pressure Rise
Test for monitoring a freeze-drying process. Drying Technology, Vol. 29, pp. 73-90,
ISSN 1532-2300.
Fissore, D., Pisano, R., & Barresi, A. A. (2011b). Advanced approach to build the design
space for the primary drying of a pharmaceutical freeze-drying process. Submitted
to Journal of Pharmaceutical Sciences, ISSN 0022-3549.
Franks, F. (2007). Freeze-drying of pharmaceuticals and biopharmaceuticals, Royal Society of
Chemistry, ISBN 978-085-4042-68-5, Cambridge, UK.
Gan, K. H., Bruttini, R., Crosser, O. K., & Liapis, A. A. (2005a). Freeze-drying of
pharmaceuticals in vials on trays: effects of drying chamber wall temperature and
tray side on lyophilization performance. International Journal of Heat and Mass
Transfer, Vol. 48, pp. 1675-1687, ISSN 0017-9310.
Gan, K. H., Crosser, O. K., Liapis, A. I., & Bruttini, R. (2005b). Lyophilisation in vials on
trays: effects of tray side. Drying Technology, Vol. 23, pp. 341-363, ISSN 1532-2300.
Gieseler, H., Kessler, W. J., Finson, M., Davis, S. J., Mulhall, P. A., Bons, V., Debo, D. J., &
Pikal, M. J. (2007). Evaluation of Tunable Diode Laser Absorption Spectroscopy for
in-process water vapor mass flux measurement during freeze drying. Journal of
Pharmaceutical Sciences, Vol. 96, pp. 1776-1793, ISSN 0022-3549.
Giordano, A., Barresi, A. A., & Fissore, D. (2011). On the use of mathematical models to
build the design space for the primary drying phase of a pharmaceutical
lyophilization process. Journal of Pharmaceutical Sciences, Vol. 100, pp. 311-324, ISSN
0022-3549.
Hottot, A., Vessot, S., & Andrieu, J. (2005). Determination of mass and heat transfer
parameters during freeze-drying cycles of pharmaceutical products. PDA Journal of
Pharmaceutical Science and Technology, Vol. 59, pp. 138-53, ISSN 1079-7440.
Jennings, T. A. (1999) Lyophilization: introduction and basic principles, CRC Press, ISBN 978-
157-4910-81-0, Boca Raton, USA.
Kessler, W. J., Davis, S. J., Mulhall, P. A., & Finson, M. L. (2006). System for monitoring a
drying process. United States Patent No. 0208191 A1.
Kuu, W. Y., Nail, S. L., & Sacha, G. (2009). Rapid determination of vial heat transfer
parameters using tunable diode laser absorption spectroscopy (TDLAS) in response
to step-changes in pressure set-point during freeze-drying. Journal of Pharmaceutical
Sciences, Vol. 98, pp. 1136-1154, ISSN 0022-3549.
Mellor, J. D. (1978). Fundamentals of freeze-drying, Academic Press, ISBN 978-012-4900-50-9,
London, UK.
Milton, N., Pikal, M. J., Roy, M. L., & Nail, S. L. (1997). Evaluation of manometric
temperature measurement as a method of monitoring product temperature during
lyophilisation. PDA Journal of Pharmaceutical Science and Technology, Vol. 5, pp. 7-16,
ISSN 1079-7440.
Oetjen, G. W., & Haseley, P. (2004). Freeze-Drying, Wiely-VHC, ISBN 978-352-7306-20-6,
Weinheim, Germany.
Pikal, M. J. (1985). Use of laboratory data in freeze-drying process design: heat and mass
transfer coefficients and the computer simulation of freeze-drying. Journal of
Parenteral Science and Technology, Vol. 39, pp. 115-139, ISSN 0279-7976.
Pikal, M. J. (2000). Heat and mass transfer in low pressure gases: applications to freeze-
drying. In: Transport processes in pharmaceutical systems, Amidon, G. L., Lee, P. I., &
Topp, E. M., pp. 611-686, Marcel Dekker, ISBN 0-8247-66105, New York, USA.
Pikal, M. J., & Shah, S. (1990). The collapse temperature in freeze drying: dependence on
measurement methodology and rate of water removal from the glassy phase,
International Journal of Pharmaceutics, Vol. 62, pp. 165–186, ISSN 0378-5173.
Pikal, M. J., Roy, M. L., & Shah, S. (1984). Mass and heat transfer in vial freeze-drying of
pharmaceuticals: role of the vial. Journal of Pharmaceutical Sciences, Vol. 73, pp. 1224-
1237, ISSN 0022-3549.
Pisano, R., Fissore, D., & Barresi, A. A. (2011). Freeze-drying cycle optimization using Model
Predictive Control techniques. Industrial & Engineering Chemistry Research, Vol. 50,
pp. 7363-7379, ISSN 0888-5885.
Pisano, R., Fissore, D., Velardi, S. A., & Barresi, A. A. (2010). In-line optimization and control
of an industrial freeze-drying process for pharmaceuticals. Journal of Pharmaceutical
Sciences, Vol. 99, pp. 4691-4709, ISSN 0022-3549.
Pisano, R., Rasetto, V., Petitti, M., Barresi, A. A., & Vallan, A. (2008). Modelling and
experimental investigation of radiation effects in a freeze-drying process,
Proceedings of EMMC– 5th Chemical Engineering Conference for Collaborative Research in
Eastern Mediterranean Countries, pp. 394-398, Cetraro (CS), Italy, May 24-29, 2008.
Rambhatla, S., Obert, J. P., Luthra, S., Bhugra, C., & Pikal, M. J. (2005). Cake shrinkage
during freeze drying: a combined experimental and theoretical study,
Pharmaceutical Development & Technology, Vol. 1, pp. 33–40, ISSN 0265-2048.
Rambhatla, S., & Pikal, M. J. (2003). Heat and mass transfer scale-up issues during freeze-
drying, I: atypical radiation and edge vial effect. AAPS PharmSciTech, Vol. 4, Article
No. 14, ISSN: 1530-9932.
Sadikoglu, H., Ozdemir, M., & Seker, M. (2006). Freeze-drying of pharmaceutical products:
research and development needs. Drying Technology, Vol. 24, pp. 849-861, ISSN
0737-3937.
Schneid, S. & Gieseler, H. (2008). Evaluation of a new wireless temperature remote
interrogation system (TEMPRIS) to measure product temperature during freeze-
drying. AAPS PharmSciTech, Vol. 9, pp. 729-739, ISSN 1530-9932.
Sheehan, P., & Liapis, A. I. (1998). Modeling of the primary and secondary drying stages of
the freeze-drying of pharmaceutical product in vials: numerical results obtained
from the solution of a dynamic and spatially multi-dimensional lyophilisation
model for different operational policies. Biotechnology & Bioengineering, Vol. 60, pp.
712-728, ISSN 1097-0290.
Tang, X. C., Nail, S. L., & Pikal, M. J. (2006). Evaluation of manometric temperature
measurement (MTM), a process analytical technology tool in freeze-drying, part III:
heat and mass transfer measurement. AAPS PharmSciTech, Vol. 7, Article No. 97,
ISSN 1530-9932.
Velardi, S. A., & Barresi, A. A. (2008). Development of simplified models for the freeze-
drying process and investigation of the optimal operating conditions. Chemical
Engineering Research and Design, Vol. 86, pp. 9-22, ISSN 0263-8762.
Velardi, S. A., Rasetto, V., & Barresi A. A. (2008). Dynamic Parameters Estimation Method:
advanced Manometric Temperature Measurement approach for freeze-drying
monitoring of pharmaceutical solutions. Industrial Engineering Chemistry Research,
Vol. 47, pp. 8445-8457, ISSN 0888-5885.
Wang, W. (2000). Lyophilization and development of solid protein pharmaceuticals,
International Journal of Pharmaceutics, Vol. 203, pp. 1–60, ISSN 0378-5173.
7
Radiant Floor Heating System

Byung-Cheon Ahn
Department of Building Equipment System Engineering, Kyungwon University
Korea
1. Introduction
The radiant floor heating system controls indoor air temperature by heat transfer from
heated surface to indoor air, after the heat has been applied into a floor structure mass by
using hot water heating coil buried under the floor. In this case, hot water is provided by a
boiler, then, conveyed to indoor floor heating coil through pipe network.
The radiant floor heating system can operate transmitting power quietly and efficiently with
no noise at low costs of the initial investment and with low maintenance. However, since
hot water heating coil is buried under the floor, the system has a defect that it has large
thermal inertia by heavy heat capacitance of the floor structure mass. In addition, response
characteristics with long time delay will be caused due to certain amount of time needed to
heat up the structure mass(Ahn, 2010). Thus, saving energy and maintaining comfortable
indoor thermal environment would be possible only if a proper control method is applied
into the system, considering its thermal inertia.
This chapter introduces system features and mathematical background of radiant floor
heating system. Especially, it covers theoretical background of analysis on heat transfer
characteristics in pipes and indoor heat flow characteristics to help understand dynamic
characteristics of energy in the system. In addition, explanation is given on types and
characteristics of automatic thermostatic valves in the system that supplies hot water with
on-off or proportional control, and more information is demonstrated on heat flow
characteristics and heating performance of the radiant floor heating system in applying
various kinds of control systems to comfort indoor heat and save energy.
2. Heat transfer in pipes

In case of radiant floor heating system, hot water from the boiler will be streamed into
households through pipes, and these pipes can be distinguished into two types; outdoor
exposed pipe covered with heat insulator, and pipe buried under the floor structure mass.
Thus, separate mathematical analyzing method is suggested to explain two types of pipes.
Firstly, fig. 1 depicts pipe covered with heat insulator. In this case, the pipe has exposed
outdoor structure and constant outdoor temperature. Assuming that there is no
superheating or subcooling of the fluid that changes phase, and its pressure does not
change, the LMTD(Log Mean Temperature Difference) applies and in combination with a
heat balance(Stoecker, 1980) gives
(1)
/
where,
q Heat transfer
U Heat transfer coefficient
ρ Water density
Cp Specific heat
V Flow velocity
Ti, To Inlet and outlet temperature
Tao Ambient temperature
W Flow rate(πR12·V)
A Outside surface area of pipe(2πR1L)
Fig. 1. Insulated pipe

Therefore, To, outlet temperature of the pipe, can be indicated from the formula (2)
· 1 exp (2)
TL, average temperature of hot water considering the length of pipe, can be found as
1
· 1 1 / (3)
Where E=UA/ρWCp, and equivalent heat transfer coefficient, U, between fluid flow and
outdoor is
1
(4)
Where,
Kst, Thermal conductivity of pipe
Kin, Thermal conductivity of insulator
R1 and R2 Inside and outside diameter of pipe
R3 Outside diameter of insulator
The heat transfer coefficient of hot water inside the pipe, (hi), is
Radiant Floor Heating System 117
. .
k
h 0.023 · Re Pr · 2300 (5)
D
4.36 · 2300 (6)
Where Re is Reynolds number(V·Di/υ) and Pr is Prantl number (υ/α).

Also, heat transfer coefficient of exterior of the pipe, (ho), is
·
(7)
Where kf is thermal conductivity of air.

Nusselt number (NuD) can be solved differently regarding horizontal pipe and vertical pipe
(Holman, 1981).
In case of horizontal pipe,
0.53 · 10 · 10 (8)
0.13 · 10 · 10 (9)
And in case of vertical pipe,
0.59 · 10 · 10 (10)
0.1 · 10 · 10 (11)
Where Prf is Prantl number for air and GrD is Grashof number.
Considering outdoor temperature (Tao), and temperature difference (ΔT) between outdoor
and pipe’s external surface, Grashof number can be expressed as below.
· · ∆ · · ∆ ·
(12)
·
ΔT value is needed in order to figure out heat transfer coefficient, ho, while U value must be
solved to find ΔT. Thus, accurate value, ΔT, can be measured through repeated calculation,
assuming ΔT as a proper number.
Concerning that structure of hot water heating coil pipe is buried under the floor in radiant
floor heating system in general, heat transfer phenomena from hot water pipe to floor and
ceiling surface must be reviewed. Fig. 2 is a diagram of pipe buried under the household
floor. Considering thermal behavior from hot water through pipe gives the following.
Very small volume (A·dx) of the amount of heat in the hot water (Δq) can be formulated:
∆ · · · · (13)
where,
ρ Water density
A Cross sectional area of the pipe
Cp Specific heat of hot water
Tx Temperature of hot water

Hot water in a very small volume has a heat transfer loss after a very short time as follows.
∆ · · · · (14)
Fig. 2. Pipe buried in semi-infinite medium having isothermal surface

This value is a sum of the heat amounts emitting to the room floor (dΔqb) and to the ceiling
surface of the room below (dΔqc).
This occurs because an amount of heat from the heated water is transferred to the floor and
ceiling surface below.
∆ ∆ ∆ (15)
Where
d∆q U · A · T T · dx · dt
d∆q U · A · T T · dx · dt
1
U R R /R R B/R
K K
1
U R R /R R B /R
K K
where,
U1 Heat transfer coefficient from pipe surface to the floor surface
U2 Heat transfer coefficient from pipe surface to the ceiling surface of the bottom layer
hi Heat transfer coefficient of pipe inner surface
T1 Floor surface temperature
T2 Ceiling surface temperature of the bottom layer
Kp Thermal conductivity of pipe
Kb Equivalent thermal conductivity from pipe surface to the floor surface
Kb' Equivalent thermal conductivity from pipe surface to the ceiling surface of the
bottom layer
B Distance from the middle of pipe to the floor surface
B' Distance from the middle of pipe to the ceiling surface of the bottom layer
Ap Girth of pipe
R1 Inside radius of pipe
R2 Outside radius of pipe
After substituting equation (15) for equation (14), categorized according to hot water
temperature, assuming that T1 and T2 are steady for a very short time (dt), and integrating
for pipe length, the outlet temperature of hot water for length L can be expressed as the
following equation (16) for hot water inlet temperature
· ·
(16)
· · ·
where,
To Hot water temperature of pipe outlet
Ti Hot water temperature of pipe inlet
L The length of pipe
v Mean flow velocity(dx/dt)
If we replace To with Tx and L with x, then integrate this for an entire length, the mean
temperature of hot water could be achieved as in the following equation (17)
1 1
(17)
U U ·A
Where D
·A·CP ·
Fig. 3 depicts pipe network buried in a house of apartment building. Hot water is supplied
from supply header through 5 distinguished pipes separately. Average temperature of each
room can be found using formula (17)
3. Indoor heat transfer

Figure 4 shows heat amounts and temperatures of each part of the room; floor, ceiling, wall
and window. There are 3 routes for heat transfer; conduction in the floor, ceiling and wall,
convection with indoor air and radiant heat transfer between the heated floor and ceiling
surface, and a non-heated wall in the house.
The amount of thermal conduction (q1) from the heated water in the pipe to the floor
surface, and the convective amount (q2) from the floor to the indoor air can be shown as
equation (18), (19).
·
/ / (18)
.
2.41 ·
(19)
.
Where Ta is the temperature of indoor air and De is the equivalent diameter of floor slab.
Fig. 3. Pipe network buried in a house of apartment building
Fig. 4. Schematic diagram for heat flow for the heat transfer analysis in the room
The amount of the thermal convection to wall, window, and door surrounding indoor
air(ASHRAE, 2004)
.
1.87 ·
.
(20)
Where H is indoor wall height.

The amount of thermal conduction (q11) from hot water pipe buried under the floor of
upper level can be shown as equation (21).
·
′
(21)
′
In addition, the amount of radiant heat transfer (qrk) in each surface is calculated with the
Gebhart’s enclosure analysis method(Segel, 1981) as in the following equation (22).
· · · · · · ·
(22)
· · · ··· · · 1,2,···,
where,
qrk Radiant heat transfer for kth surface
Ak kth surface area
εk kth surface emissivity
σ boltzmann constant
Tk kth surface temperature
ρ1-n Reflectivity of surface
Fj-k Coefficient of form between the inside and outside surface
Fig. 5. Schematic of room for radiation heat transfer analysis

Temperatures of indoor air and each part of the room can be determined by an analysis of
these 3 heat transfers: conduction, convection and radiation. In this study, each temperature
is measured using the electrical resistance-capacitance circuit method(Sepsy, 1972) as shown
in Fig. 6. It is based on assumptions that heat capacity for each wall is concentrated to one
point in the wall, and that the temperature from the point to the wall surface is steady.
Equivalent heat resistances of either side from each central node in the wall are the same by
selecting the point. Heat loss to outdoor air is considered by setting the new central
node(Chang, 1996), which is the point at which both equivalent heat resistances of the
existing node and the surface are the same energy equation of each point is as follows
(23)
where,
Cp Capacitance of each part
Tp Temperature of each part
qin, qout Heat transfer by convection or conduction
Fig. 6. An equivalent R-C circuit for unsteady energy analysis
4. Automatic thermostatic valves

In case of radiant floor heating system, automatic temperature control valve is used in order
to consume energy effectively and maintain pleasant indoor temperature. This valve has
similar function as those of gate or glove valve, but it can be separated into electric powered
type and non-electric powered type in terms of the source of power that moves valve disk.
Electric powered type uses external force such as electricity, while non-electric powered
type uses only internal driving element such as shape memory alloy and spring. In general,
electric powered type operates by motors is composed of room temperature controller and
automatic thermostatic valve, and non-electric powered type only consists of automatic
thermostatic valve.
Furthermore, Automatic thermostatic valve itself can be differed into on-off type and
proportional control type in terms of control method. Fig. 7 shows control method of
automatic thermostatic valve expressed with flux supply method from temperature change.
Fig. 7. Characteristics of Flow rates vs temperature in control methods of automatic

thermostatic valves
Fig. 8 contains types of heating automatic thermostatic valve, and table 1 is about the
characteristics of each type of valves.
Fig. 8. Various classifications for automatic thermostatic valves

First, in case of electric powered automatic thermostatic valve, control part senses indoor air
temperature of each room and transmits signal to driving part (thermostatic valve) to
control the amount of heating hot water flowing through the pipe. Temperature sensor
called thermister is mostly used to sense temperature from indoor temperature controller,
and it is thermally sensitive semiconductor resistance thermometer. Thermister is widely
used as temperature sensor for household electric appliances and industrial machineries,
since its temperature usage range is -50~500℃, which can be applied to every possible range
that ordinary temperature control is needed, and it is not only cheap and tiny but highly
sensitive. Driving part of electric powered type automatic thermostatic valve can be
separated into 3 different methods using ball valve, cone valve, and solenoid. Solenoid valve
has 2 seconds of on-off response time, while ball valve has 10 seconds and cone valve has
several ten seconds to minutes
Non-electric powered type automatic thermostatic valve using shape memory alloy actively
controls on-off state of valve by sensing shape memory alloy element; closes valve
proportional to temperature due to the increase of returning water temperature, and opens
valve by returning spring due to decrease of returning water temperature.
As a merit, power supply is not necessary and response time is faster than thermal
expansion. Also, structure only consisting of thermal static valve is very simple and
endurance is superior, because Ti-Ni shape alloy spring is used as an operational element.
However, it has a demerit that it has to passively decide flux amount that fits to hot water
temperature amount considering consumer’s thermal surroundings after construction. Also,
it is significant to choose appropriate controlling components for types and characteristics of
installing heating system.
Indirect power type Direct type

Motor control Solenoid Thermal Shape memory Capillary tube
type control expansional alloy type Type
type control type
- Temperature - Temperature - Temperature - water - Water
Control sensing type sensing type sensing type temperature temperature
method - On-off type - On-off type - On-off type sensing type sensing type
- Proportional - Proportional
control type control type
Operating Electric motor Electromagnet Liquid thermal Shape memory Liquid thermal
source expansion alloy spring expansion
bellows
Response Within several Within several Within several Within several Within several
time ten seconds seconds minutes seconds ten minutes
- Fast response - Fastest - Electric power - Electric power

Merit - Easy response isnecessary is unnecessary
installation - Simple - Good endurance - Proportional
Structure - Proportional Type
- Low cost type
-Electric power - High Relatively slow Manual valve -Liquid quality
Demerit is necessary pressure loss response Setting variation
- High cost -Difficult
installation
Table 1. Characteristics of classified automatic thermostatic valves
5. Heating control system performance analysis

If indoor temperature were controlled by applying automatic thermostatic valve for radiant
floor heating system, response characteristics of long time delay on control response would
occur due to large heat capacitance of floor structure mass. In order to examine response
characteristics of radiant floor heating system, simulation was performed and was
compared and verified with experiment results by using mathematical analysis model of a
radiant floor heating system explained in previous 2nd, 3rd paragraphs.
Fig. 9 shows the results of the experiment and simulation (Ahn, 2010).
We measured temperature changes for 5 hours natural cooling after supplying hot-water for
3 hours. For the floor temperature, two temperatures (one at the nearest part to the pipe and
the other between the pipes), were measured and compared with temperatures from the
simulation data. The chiller to maintain temperature of artificial chamber in the test house
was set up to maintain an outdoor air temperature of 8°C. The reason for cooling is to secure
the constant temperature around the room for the indoor heating test.
Data obtained from the entering supply and outdoor temperatures into the simulation for
operation, were contrasted to the experimental data.
70
Supply water(experiment)
Return water(experiment)
Return water(simulation)
60 Floor surface1(experiment)
Supply water Floor surface(simulation)
Return water Floor surface2(experiment)
50 Indoor air(experiment)
Indoor air(simulation)
Temperature ( C)
Inner wall surface(experiment)

Inner wall surface(simulation)
o
40 Outdoor air(experiment)
Floor surface
30
20
Inner wall surface
Indoor air
Outdoor air
10
0
0 1 2 3 4 5 6 7 8
Time (Hour)
Fig. 9. Comparison of the simulation and experimental data (supply water temperature:
50°C, flow rate: 0.05L/s, outdoor temperature: 8°C)
Results showed that the simulation data agreed well with experimental data. Response
characteristics will be examined through using simulation model of radiant floor heating
system and applying various kinds of automatic thermostatic valves.
First of all, In terms of proportional control and On-Off control, which are types of
automatic thermostatic valves, each is classified according to sensing methods; water
temperature sensing and air temperature sensing. They are performed by testing simulation
and defining their features.
Each control of each flow rate and temperature has its own peculiarities as outdoor air
changes from the daily lowest temperature -5°C to the highest temperature 5°C over
24hours. Table 2 summarizes control characteristics for each of the four studied cases. For
the 4 controls: Case 1 does not adjust the flow rate, Case 2 controls the flow rate in
proportion to the difference between room temperature and setting point. Case 3 adjusts the
flow rate in proportion to the difference between returned-water temperature and setting
point, and Case 4 controls the On-off for the supply water by the differential gap according
to the difference between the room temperature and setting point.
Classification Description
Case1 No control method

Case2 Proportional valve control with air temperature feedback
Case3 Proportional valve control with water temperature feedback
Case4 On-off valve control with air temperature feedback
Table 2. Classification of control methods

Fig. 10 summarizes the results of changes in temperatures of return water, the floor and
indoor air over 24hours. The outdoor air is vibratory from -5°C to 5°C, the return water
increases to 43.3°C from the set point, and the floor and the indoor air rapidly increases to
20°C for 3hours, and then steadily to 32°C and then decreases to 27.5°C. At this time, the
mean temperatures of return water and indoor air are 42.3°C and 24.9°C, respectively.
Fig. 11 shows the temperature responses and a flow rate for 24hours as a result of the
proportional control for the indoor air temperature (Case 2) designated from 22.3°C to
23.3°C to maintain 22.8°C, the mean indoor temperature. The maximum flow rate helps
adjust flow and maintain an indoor air temperature of 23°C before reaching the lower limit
of 22.3°C. Controlling flow can offset the change of outdoor air temperature so that indoor
air temperature can be maintained.
With the exception of the first stage, the temperature of the return water shows a gentle
slope, increasing, and that of floor surface is 27.9°C continuously. In this case, each mean
temperature of return water and indoor air is 31.8°C, and 22.8°C. This air-temperature
proportional control maintains indoor temperature through light control regardless of
changes in outdoor air temperature.
Fig. 12 shows temperature responses and a flow rate for 24hours as a result of the
proportional control for return water temperature (Case 3) designated from 30.5°C to 34.5°C
to maintain 22.8°C, the mean indoor temperature. The indoor air maintains a temperature of
33.1°C after initially increasing to 24.1°C. The flow rate steadily decreases but its drop
dwindles except for the first stage. In this case, the mean temperatures of return water and
indoor air are 33°C, and 22.8°C, respectively.
Water proportional control does not cope with changes in indoor air temperature, but
maintains return water temperature, and if the indoor air temperature were at the best
condition, it would be difficult to find an upper and lower limit that can maintain it.
70 0.06
Flowrate
60
Supply water 0.05
50
Temperature ( C)
0.04
40
Flowrate(L/s)
o
Return water
Floor surface
30
0.03
20 Indoor air
0.02
10
0 Outdoor air 0.01
-10
0.00
0 3 6 9 12 15 18 21 24
Time (Hour)
Fig. 10. Various temperature responses for outdoor temperature change with no control
(Case 1)
70 0.06
Flowrate
60
Supply water 0.05
50
Temperature ( C)
Return water 0.04

40
Flowrate(L/s)
o
Floor surface
30
0.03
20
Indoor air
0.02
10
0 Outdoor air 0.01
-10
0.00
0 3 6 9 12 15 18 21 24
Time (Hour)
Fig. 11. Various temperature responses with proportional valve control with air temperature
feedback (Case 2)
70 0.06
60
Supply water
0.05
50
Temperature ( C)
Return water 0.04

40
Flowrate(L/s)
o
Floor surface
30
0.03
20
Flowrate Indoor air
0.02
10
0 Outdoor air 0.01
-10
0.00
0 3 6 9 12 15 18 21 24
Time (Hour)
Fig. 12. Various temperature responses with proportional valve control with return water
temperature feedback (Case 3)
70 0.06
Flowrate
60
Supply water 0.05
50
Temperature ( C)
0.04
40
Flowrate(L/s)
o
Return water
Floor surface
30
0.03
20
Indoor air
0.02
10
0 Outdoor air 0.01
-10
0.00
0 3 6 9 12 15 18 21 24
Time (Hour)
Fig. 13. Various temperature responses with on-off valve control with air temperature
feedback (Case 4)
Fig. 13 shows changes to temperature and flow rate for 24hours in flow-rate supply on–off
at a range of indoor air temperatures (Case 4), 22.3°C to 23.5°C.
The operation time is long due to the set point of indoor temperature of 20°C, and then
shortens due to intervals of 1°C of On-Off. The process is more time consuming when the
outdoor temperature is lower and the relatively non-operational time shortens. The indoor
air temperature vibrates from 22.5°C to 23.5°C according to the control, and the return water
temperature almost reduces to the floor temperature when the flow supply is stopped. At
this time, the mean temperatures of return water and indoor air are 34.4°C and 22.8°C,
respectively.
For the On-Off control, time differences are created between the hot water supply and non-
supply as the outdoor air temperature changes, and the indoor air temperature affects the
control. But the control does not maintain the indoor air and creates a fluctuation in
temperature. It is therefore important to set the proper point and use a differential gap.
Fig. 14 shows the mean temperature of the room air and total amount of heat supply
according to controls. No-Control (Case 1) is 2°C and 14,000Kcal higher than other controls
(Case 2~4). All controls, Cases 2~4, have a similar mean temperature but the water
temperature-sensing control (Case 3) uses about 8% more heat than the air temperature
sensing control (Case 2, Case 4).
The mean temperature of room air Total Energy Consumption
35 6.0E+04
5.0E+04
Energy consumption(Kcal/Day)
30
4.0E+04
Temperature(℃)
25 3.0E+04
2.0E+04
20
1.0E+04
15 0.0E+00
Case 1 Case 2 Case 3 Case 4
Fig. 14. Mean temperature of indoor air and total energy consumptions with different
control methods (supply water temperature of 50°C)
Control variables of radiant floor heating system can usually be considered as supply hot
water temperature and set value of indoor temperature. These control variables must be set
to a proper value to improve indoor thermal environment and save energy, and variable
values reflecting amount of outdoor temperature can be considered.
24
23
Temperature ( C)
o
22
21
20
Outdoor air temp. = -10oC / Supply water temp. = 45oC
Outdoor air temp. = -10oC / Supply water temp. = 52.5oC
Outdoor air temp. = -10oC / Supply water temp. = 60oC
19
0 3 6 9 12 15 18 21 24
Time (Hour)
Fig. 15. Indoor air temperature responses for various supply water temperatures
24
23
Temperature ( C)
o
22
21
20 o o
Outdoor air temp. = 10 C / Supply water temp. = 45 C
o o
Outdoor air temp. = 0 C / Supply water temp. = 52.5 C
o o
Outdoor air temp. = -10 C / Supply water temp. = 60 C
19
0 3 6 9 12 15 18 21 24
Time (Hour)
Fig. 16. Indoor air temperature responses for various supply water and outdoor air
temperatures
First of all, Fig. 15 shows characteristics of indoor air temperature alternation when supply
hot water temperature is 45℃, 52.5℃, 60℃ in case of setting outdoor temperature condition
as -10℃. The control method here is the result of applying on-off control method (the on-off
for the supply water by the differential gap according to the difference between the room
temperature and setting point), using indoor air temperature sensor.
The diagram indicates that the time to reach the setting point 22℃ is radically increasing
when using relatively low temperature supply water in case of decreasing outdoor
temperature. Analyzing from the data that more than 6 hours is needed for supply hot
water temperature at 45℃ to reach 22℃, considerable amount of time is consumed in order
to improve indoor thermal environment to a comfortable range in case when outdoor
temperature is low and low temperature supply water is provided. Thus, relatively warmer
hot water than cooler hot water is appropriate for heating in case of low outdoor
temperature (Ahn, 2005).
Fig. 16 shows dynamic characteristics of indoor air temperature change when supply water
temperature is 60℃, 52.5℃, and 45℃ at outdoor temperature condition of -10℃, 0℃, and
10℃. Overshoot of each 3 separate cases depicted similar amount on indoor air temperature
change, and the rise times to reach the set indoor air temperature 22℃ were close to 3 hours
without difference. By correctly determining supply hot water temperature in regard to
amount of outdoor temperature, transient response characteristics for indoor environment
can be improved.
Fig. 17 depicts characteristics of energy consumption and indoor air temperature average
considering temperature alternation of outdoor and supply hot water.
Energy consumption and indoor air temperature average entirely rises as temperature of
supply hot water or outdoor increases on the graph. The energy consumption gap was
moderate in terms of supply hot water temperature change, while it was relatively large
for outdoor temperature change. Therefore, changing supply hot water temperature in
terms of outdoor temperature variation will minimize overheating and benefit energy
savings.
Furthermore, if high temperature hot water were provided to make indoor air temperature
reach its set temperature, supplying time of hot water would be lessened due to rapid
increase of indoor air temperature despite of large consumption of heat amount depending
on temperature difference. On the other hand, when low temperature hot water is provided,
heat amount consumption will decrease, but supplying time will be delayed due to slow
increase rate of indoor air temperature so that amount difference of energy consumption
will be small.
Fig. 18 depicts result measured from the experiment about flow rate change characteristics
and temperature change of radiant floor heating system. This is to demonstrate general
thermal change features of radiant floor heating system from each boiler-installed house-
hold that is run by individual heating method.
Fig. 18 indicates temperatures of supply water, returning water, floor surface, indoor air,
outdoor air, and flow rate. Temperature of supply water from supply header was at
maximum 50℃ due to heat loss from pipe, while it is being operated as on-off type control
by boiler system.
Therefore, heat loss must be considered when determining set value of the supply water
temperature. In addition, while indoor air temperature is being controlled up to 23±0.5℃ by
o
Energy consumption for outdoor air temp. (-5 C)
o
Energy consumption for outdoor air temp. (0 C)
Energy consumption for outdoor air temp. (5oC)
Mean indoor temperature for outdoor air temp. (-5oC)
o
Mean indoor temperature for outdoor air temp. (0 C)
o
Mean indoor temperature for outdoor air temp. (5 C)
10000 23.5
8000
Energy consumption(kcal)
23.0
Temperature( C)
6000
o
22.5
4000
22.0
2000
0 21.5
45 55 65
Supply water temperature(℃)

Fig. 17. Energy consumption and mean indoor air temperature responses for various supply
water and outdoor air temperatures
12
Supply water Return water
50
40
Temperature ( C)
8
Flowrate(LPM)
30
o
20
Floor surface Indoor air Outdoor air Flowrate

10
4
-10
0
0 3 6 9 12 15 18 21 24
Time (Hour)
Fig. 18. Experimental results for the response characteristics of temperature and flow rate
on-off type, initial operation time period to reach set indoor temperature is relatively slower
than that of after second period, and temperature alternation of indoor air and floor surface
seems severely slow even after hot water is being supplied. Thus, thermal environment
would be improved if indoor air temperature increasing period were reduced by initially
supplying comparably hotter water temperature, and cooling it down afterwards.
6. Conclusions
Radiant floor heating system is mainly applied to residential buildings. Recent trend shows
increasing rate of countries applying the system to their residential buildings after
considering various profits of the system.
However, the radiant floor heating system may contain long time delay as inappropriate
response characteristics on indoor temperature control due to significant influence of
thermal inertia when performing indoor air temperature control.
In order to improve thermal environment and save energy by effectively controlling the
indoor air temperature, temperature value of supply hot water from boiler and indoor air
set temperature can be altered by forecasting the value of outdoor temperature and
characteristics variation.
Therefore, application of more effective system design and advanced control technique for
radiant floor heating system like HVAC system(Ahn, 2001) is required through consistent
technology research and development.
7. References
Ahn, B.C. and Mitchell, J.W. (2001). Optimal Control Development for Chilled Water Plants
using a Quadratic Representation, Energy and Buildings 33, pp.371-378, Elsevier,
ISSN 0378-7788
Ahn. B.C. and Lee, T.W. (2005). Strategy of Energy Saving and Thermal Environmental
Improvement for Intermittent Heating System in Apartment Buildings, Journal of
the SAREK, Vol.17, pp.88-93, ISSN 1229-6422
Ahn, B.C. and Song, J.Y. (2010). Control Characteristics and Heating Performance Analysis
of Automatic Thermostatic Valves for Radiant Slab Heating System in Residential
Apartment, Energy 35, pp.1615-1624, Elsevier, ISSN 0360-5442
ASHRAE Handbook, HVAC Systems and Equipment. (2004). Panael Heating and Cooling,
Chapter 6, 6.1-6.22. ISBN 1-931862-48-6, Atlanta, USA
Chang, H.W. and Ahn, B.C. (1996). The Energy Analysis and Control Characteristics of a Ho
Water Heating System for Apartment Houses, Journal of the SAREK, Vol.8, pp.76-87,
ISSN 1229-6422
Holman, H.J. (1981). Heat Transfer, 5th Ed., McGraw-Hill. , ISBN 0-07-029618-9, MI, USA
Segel Robert and Howell, J.R. (1981). Thermal Radiation Heat Transfer, McGraw-Hill. , ISBN
0-07-057316-6, New York, USA
Sepsy, C.F. (1972). A Thermal Analysis of the Building and the Heating and Cooling
Systems Selected for the Field Validation Test, ASHRAE Sym. Bulletine No. 72, pp.5-
9.
Stoecker, W.F. (1980). Design of Thermal Systems, McGraw-Hill, ISBN 0-07-061618-3, Tokyo,
Japan
8
Variable Property Effects in

Momentum and Heat Transfer
Yan Jin and Heinz Herwig
Institute for Thermo-Fluid Dynamics, Hamburg University of Technology (TUHH)
Germany
1. Introduction
“Variable property effects” is a theoretical construct since a real fluid always is subject to
variable properties when changes in temperature or pressure occur. Their influence
compared to a corresponding situation but with artificially constant properties may be small
and thus neglected in a first approximation. Those artificial “constant property results” may
then be corrected with respect to the initially neglected effects due to the variability of the
fluid properties. This concept assumes small variable property effects and therefore is not
applicable when the flow itself is basically generated by a variable property (like natural
convection, generated by density variations) or strongly affected by it (like strongly
compressible flow, determined by density variations).
Therefore we define “variable property effects” as those (small) artificial effects that would
be present if fluid properties could change from constant to variable. They have to be added
to a solution which is gained under the assumption of constant properties in order to
account for the fact that real fluids always have properties which are temperature and
pressure dependent. The variable property correction of a constant property solution can be
accomplished in different ways. Basically there are three methods which are widely used in
this context. They are
(1) Property ratio method
Results in terms of the nondimensional friction factor f and Nusselt number Nu are
gained by multiplying the constant property results f cp and Nucp with a property ratio
correction factor, i.e.
ma na
⎛a ⎞ ⎛a ⎞
f = f cp ⎜ 1 ⎟ , Nu = Nucp ⎜ 1 ⎟ (1)
⎝ a2 ⎠ ⎝ a2 ⎠
Here a1 , a2 are properties ( ρ , μ , k , c p ) at two different temperatures and ma , na are
empirical exponents. There may be more than one factor, depending on which properties
are assumed to contribute to the variable property effects. Studies with this approach are Li
et al. (2007) and Mahmood et al. (2003), for example.
(2) Reference temperature method
The constant property results in terms of f and Nu are evaluated at a certain temperature
for the properties that appear in f and Nu . This so-called reference temperature,
TR = T1 + j (T2 − T1 ) (2)
between two characteristic temperatures T1 and T2 of the problem must be chosen such
that f and Nu determined under the assumption of constant properties (and with T1 or
T2 as reference temperature) give the results for variable properties. For that purpose the
factor j in equ (2) must be determined properly. Basically, j is an empirical parameter like
the exponents in equ (1) are in the first method. Studies using this method are Jayari et al.
(1999) and Debrestian & Anderson (1994), for example.
(3) Asymptotic method
Since the effects of variable properties are assumed to be small the variable property
solution can be taken as a perturbation of the constant property solution. In a systematic
approach the problem is treated as a regular perturbation problem with a (small)
perturbation parameter ε linked to the transfer rate. The constant property solution is that
for ε = 0 and variable property effects are described as 1st, 2nd, … nth order effects of an
asymptotic expansion with respect to the perturbation parameter ε . This approach in an
early paper has been suggested by Carey and Mollendorf (1980) and afterwards has been
adopted by the second author for a wide variety of problems, see for example Herwig and
Wickern (1986), Herwig and Bauhaus (1986), Herwig et al. (1989) and Herwig and Schäfer
(1992).
The advantage of this method over the first two methods to account for variable properties
is its systematic generalization of the correction terms that appear within this approach.
These terms hold for all Newtonian fluids and all (small) transfer rates that are the reason
for variable property effects. Furthermore, with this approach the empirical parameters in
the first two method ( mα , nα and j ) can be embedded in the expansion approach and thus
be interpreted on a sound physical basis, as is demonstrated in Herwig and Wickern (1986),
for example.
For a long time all three methods have been applied to certain standard situations like pipe,
channel or boundary layer flows. Only recently it was shown how the asymptotic method
can also be applied to complex problems, such as turbulent mixed convection in a room that
is ventilated and heated by several heating elements, see Bünger and Herwig (2009), Jin and
Herwig (2010) and Jin and Herwig (2011).
This extension to complex convective heat transfer problems will be presented and illustrated
by some examples in the following.
2. The asymptotic approach: State of the art

Variable properties means that all fluid properties involved in a problem depend on
temperature and pressure. Whenever temperature and pressure variations occur in a problem
these variabilities affect the solution. Assuming the effects to be small, a perturbation approach
is straight forward. Since pressure effects almost always are negligibly small, we only account
for the temperature impact. The starting point for the temperature dependence is the Taylor
series expansion of all properties with a* representing ρ * (density), μ * (viscosity), k *
(thermal conductivity), and c p* (specific heat capacity), which is with aR* = a * TR* ( )
*
a :=
a
*
aR
1
2
1
n!
(
= 1 + ε K a1θ + ε 2 K a 2θ 2 + " + ε nK anθ n + O ε n + 1 ) (3)
Variable Property Effects in Momentum and Heat Transfer 137
ΔTR* (4)
ε :=
TR*
T * − TR* (5)
θ :=
ΔTR*
⎡ ∂a* T * ⎤
K a1 := ⎢ * * ⎥ (6)
⎣⎢ ∂T a ⎥⎦ R
⎡ ∂ 2 a* T *2 ⎤
Ka2 := ⎢ *2 * ⎥ (7)
⎢⎣ ∂T a ⎥⎦ R
⎡ ∂ n a* T *n ⎤
K an := ⎢ * n * ⎥ (8)
⎣⎢ ∂T a ⎦⎥ R
Here, ε is a small quantity which basically represents the strength of heat transfer ( ε = 0 :
isothermal situation) and formally serves as perturbation parameter. The dimensionless
temperature is θ which then is an O ( 1 ) quantity. The fluid behavior is covered by K aj
with j = 1, 2," n , which are O ( 1 ) quantities, representing the partial derivatives of a * with
respect to temperature.
The same expansion can be made with respect to the pressure p * . Since, however, for most
((
cases the corresponding Kˆ an values ∂ n a* ∂p * n )( ))
p * n a* and/or pressure variations are
very small, we restrict ourselves to the temperature effects.
In order to get solutions in terms of expansions which correspond to the Taylor-series
expansions of the properties, all dependent variables V * are expanded as
V*
V :=
VR* ⎣ ⎦ ( )
= 1 + ε ⎡K ρ 1Vρ 1 + K μ 1Vμ 1 + K k 1Vk 1 + Kc p 1Vc p 1 ⎤ + O ε 2 (9)
Here, V * represents the velocity components u* , v * , w * and the temperature T * which

appear in the basic equations of a problem. For many problems these are the Navier-Stokes
and the thermal energy equations, here shown for a steady flow ∂ ∂t = 0 , i.e:
( )
G
∇⋅ ρ u = 0 (10)
( ρ uG ⋅ ∇ ) uG = Re1 R
⎛
⎝
( G G 2
3
)
G ⎞
⎠
G
∇ ⋅ ⎜ μ ∇u + ∇ ⋅ uT − ∇ ⋅ uI ⎟ − ∇p + ρ g (11)
G
ρc p ( u ⋅ ∇ )T =
1
Re RPrR
( )
G
∇ ⋅ k ∇T + p∇ ⋅ u (12)
In the thermal energy equation (12) viscous dissipation is neglected since it contributes very
little to the energy balance. Equs (10)-(12) are nondimensional equations in which all fluid
properties a = a* aR* are marked by a box (note that a = 1 for constant properties).
The nondimensional groups are
ρ R* uR* L* μ R* c pR
*
Re R = , PrR = (13)
μ R* kR*
It is an important point that here they are formed with the reference fluid properties (index
R). When turbulent flows are treated by the RANS-approach, equs (10) – (12) must be time-
averaged and subject to turbulence modeling. Then, with a k − ε model, for example, two
more differential equations have to be solved in which fluid properties appear as well.
In the asymptotic method to account for variable property effects, all properties a (i.e. ρ ,
G
μ , k , c p ) and all variables υ (i.e. u , T ) are replaced by their expansions (3), (9) in the set
of equs (10) – (12). Then two different ways to proceed are possible which will be explained
in the next two subsections.
2.1 Higher order equations method (HOE)

In this method the equations are subject to the expansion process, i.e. all terms of O ( 1 ) ,
ε K ρ 1 , ε K μ 1 , …, ε 2 K ρ 2 , ε 2 K ρ 1K ρ 1 , … are collected separately and form the zero order, first
order, second order, … sets of equations. In the complete approach with all four properties
involved there is one zero order set of equations (corresponding to constant properties), four
first order sets and fourteen sets of second order. Only the zero order set of equations is a
nonlinear one, all subsequent equation sets have linear equations. All equations have to be
solved sequentially starting with the zero order equations for constant properties.
The final result in terms of the friction factor f and the Nusselt number Nu gains an
asymptotic form as well. For example, the Nusselt number for the thermal boundary
condition Tw = const , reads, with Nucp for the constant property case:
⎡ Tρ′ 1 Tμ′ 1 T′ Tc′p 1 ⎤

Nu
Nucp
= 1 + ε ⎢K ρ 1
⎢ T0′
+ Kμ 1
T0′
+ K k 1 k 1 + Kc p 1
T0′ T0′ ⎥
( )
⎥ +O ε2 (14)
⎣ ⎦
Here, Ta′1 are the temperature wall gradients since q w* in the Nusselt number is
(
q w* = − k *T *′ ) w
.
In the second order 14 additional terms in equ (14) appear so that altogether 19 sets of
equations have to be solved when second order accuracy is required. Then, however, the
result (14) is quite general, since it holds for all Newtonian fluids (different values for K ρ 1 ,
K μ 1 , … in (14)) and all small heating rates (different values for ε in equ (14)).
This method is straight forward for all laminar flows and has been applied in Herwig and
Wickern (1986) and Herwig and Bauhaus (1986), for example.
When turbulent flows are considered, the higher order equation method needs higher order
turbulence model equations. That may be a problem which does not occur in the alternative
method described next.
2.2 Higher order coefficients method (HOC)

Instead of solving higher order equations from which the influence of the single properties
can be deduced in the final results, like in equ (14) for example, the final results are taken in
its asymptotic form as the starting point. Again, shown for the Nusselt number, it reads up
to the second order:
Nu
= 1 + ε ⎡K ρ 1 Aρ + K μ 1 Aμ + K k 1 Ak + Kc p 1 Ac p ⎤ + ε 2 ⎡⎣K ρ 2 Aρ 2 + K μ 2 Aμ 2
Nucp ⎣ ⎦
+ K k 2 Ak 2 + Kc p 2 Ac p 2 + K ρ2 1 Aρρ + K μ2 1 Aμμ + K k21 Akk + +Kc2p 1 Ac p c p

(15)
( ) (
+ K ρ 1 K μ 1 Aρμ + K k 1 Aρ k + Kc p 1 Aρ c p + K μ 1 K k 1 Aμ k + Kc p 1 Aμ c p )
( )
+ K k 1Kc p 1 Akc p ⎤ + O ε 3
⎦
with the A-values Aρ , Aμ … as the crucial part of the result. Up to the second order there
are 4 + 14 = 18 different A-values.
With the nondimensional property a according to equ (3) is rewritten as
a = 1 + ha 1θ +
1
2
1
ha 2θ 2 + " + hanθ n + O ε n + 1
n!
( ) (16)
introducing
⎡ ∂ j a * ΔT * j ⎤
haj := ε j K aj = ⎢ * j * ⎥ , j = 1, 2," n. (17)
⎢⎣ ∂T a ⎥⎦ R
one immediately recognizes the A-values to be
1 ⎡ ∂Nu ⎤
Aa := ⎢ ⎥ (18)
Nucp ⎣ ∂ha1 ⎦ h = 0
1 ⎡ ∂ 2 Nu ⎤
Aaa := ⎢ ⎥ (19)
2Nucp ⎣⎢ ∂ha21 ⎦⎥ h = 0
1 ⎡ ∂Nu ⎤
Aa 2 := ⎢ ⎥ (20)
Nucp ⎣⎢ ∂ha 2 ⎦⎥ h = 0
1 ⎡ ∂ 2 Nu ⎤
Aab := ⎢ ⎥ (21)
Nucp ⎣⎢ ∂ha 1∂hb 1 ⎦⎥ h = 0
{ }
where a , b ∈ ρ , μ , k , c p , a ≠ b .
According to equs (18) – (21) the 18 A-values in (15) can be determined as first and second
derivatives of Nu with respect to the h-values in (16). These h-values quite generally can be
interpreted as combinations of ε and fluid properties ( K ρ 1 ," ). Since the method holds for
all Newtonian fluids it also holds for arbitrarily chosen fictitious fluids for which only one
property is temperature dependent (in some arbitrary way) with all other properties being
constant. Such fluids serve as “dummy fluids” from which the general A-values can be
determined.
This is done by solving the full equations (10) – (12) or their turbulent version for these
dummy fluids with certain values for ha 1 , hb 1 and ha 2 . Then the A values can be
determined by the following numerical approximations of (18) – (21):
1 ⎧⎪ Nu ( ha 1 = s ) − Nu ( ha1 = −s ) ⎫⎪
Aa = lim ⎨ ⎬ (22)
Nucp s →0 ⎩⎪ 2s ⎭⎪0
1 ⎧⎪ Nu ( ha 1 = s ) − 2Nucp + Nu ( ha 1 = −s ) ⎪⎫
Aaa = lim ⎨ ⎬ (23)
2Nucp s → 0 ⎩⎪ s2 ⎭⎪ 0
1 ⎧⎪ Nu ( ha 2 = s ) − Nu ( ha 2 = −s ) ⎫⎪
Aa 2 = lim ⎨ ⎬ (24)
Nucp s →0 ⎩⎪ 2s ⎭⎪0
1 1
Aab = lim
Nucp s →0 4s 2
{Nu ( ha1 = hb1 = s ) + Nu ( ha1 = hb1 = −s )
(25)
−Nu ( ha1 = −hb 1 = s ) − Nu ( −ha1 = hb 1 = s )} .
0
Here the limiting process lim … means that s must be small enough to avoid a higher order
s →0
influence, but not too small since otherwise truncation errors appear. The index 0 at the
right curly brackets indicates that all other variables hc 1 , hc 2 are set to zero. Each of the
eight terms
• Nu ( ha 1 = s )0 ,Nu ( ha 1 = −s )0
• Nu ( ha 2 = s )0 ,Nu ( ha 2 = −s )0
Nu ( ha 1 = hb 1 = s )0 ,Nu ( ha 1 = hb 1 = −s )0 ,
•
Nu ( ha 1 = − hb 1 = s )0 ,Nu ( ha 1 = − hb 1 = −s )0
corresponds to exactly one CFD-simulation. Hence, the calculation of (22), (23), and (24)
needs two CFD-simulations for a fixed small value of s , while the determination of Aab
according to (25) needs four CFD-simulations. With one further CFD-calculation the
constant property solution Nucp is determined.
In order to determine the complete set of all 18 A-values (second order accuracy of the
result) thus 40 variable property solutions and the one for constant properties are needed.
Since no higher order equations are determined, all kinds of turbulence modeling can be
incorporated.
Details of the approach and an application of the HOC-method to a complex geometry and
flow can be found in Bünger and Herwig (2009) and Jin and Herwig (2010), respectively. In
Fig. 1, as an example taken from Jin and Herwig (2010), the influence of variable properties
on the Nußelt number at two heating elements in a complex geometry is shown, determined
with the HOC-method.
There can be, however, a further improvement of the HOC-method which considerably
reduces the amount of numerical solutions that are needed for the determination of all A-
values. This will be described next.
3. A more efficient (asymptotic) approach

For a more efficient method, we first rewrite the basic equations (10)-(12). Instead of
referring the properties ρ * , μ * , k * and c p* to their values at a reference temperature TR* we
incorporate them in the nondimensional groups Re and Pr (which then are no longer Re R
and PrR since they now are local, variable quantities).
(a) (b)
A B C D E A B C D E
1.00 1.00
0.99
0.99
Nu/Nucp
Nu/Nucp
0.98
0.98
0.97
0.96 0.97
0 1 2 0 1 2
s s
(c) (d)
Fig. 1. Complex benchmark geometry, example from (Jin & Herwig, 2010); (a) 3D room for
experiments, (b) 2D approximation for simulations, 1 , 2 : heating elements ( q w* = 300 W m 2 ,
Re = 1.64 × 10 6 ) s * : coordinate around the heating elements and along the lower, upper, and
side walls A, B, C, D: corners of the heating elements (c) Heating element 1 (d) Heating
element 2 CFD: numerical solution, based on equs (10)-(12) HOC: asymptotic result, eq. (15)
Since ρ * , μ * , and k * appear within a spatial derivative in (10)-(12), additional terms
appear according to the product rule of differentiation. Instead of (10)-(12) one gets:
G G
∇ ⋅ ( u ) + Pρ ∇T ⋅ u = 0 (26)
∇p
( u ⋅ ∇ ) u = ⎜⎛
G G
⎝
1
Re ⎠⎝
(
⎞⎛ G G 2
3
G ⎞
⎠
)
∇ ⋅ + Pμ ∇T ⋅ ⎟⎜ ∇u + ∇uT − ∇ ⋅ uI ⎟ −
1 + ( ρ − 1)
+g (27)
⎛G Pk ⎞ 1
⎜u − ∇T ⎟ ⋅ ∇T = ∇ ⋅ ( ∇T ) (28)
⎝ PrRe ⎠ PrRe
with the parameters Pρ , Pμ , Pk as
ΔT * ∂a *
Pa = ; a* = ρ * , μ * , k * (29)
a* ∂T *
∇p
The only property that is left explicitly is ρ in the term − in equ (27), all other
1 + ( ρ − 1)
properties are “hidden” in Re , Pr and Pa . Asymptotically the parameters Pa are of order
O ( ε ) , since
Pa = ε K a1 + O ( ε ) (30)
( )
accounting for the fact that a * in K a1 is a * TR* but a * in Pa is a * T * . ( )
Based on equs (26)-(28) there are two simplified approaches to account for variable
properties which are motivated by the fact that the dimensionless groups account for the
variability of the properties leaving additional effects in the asymptotically small terms with
Pa and ( ρ − 1 ) .
3.1 High Reynolds number HOC-method (HR-HOC)

For high Reynolds numbers the energy equation (28) can be approximated by setting Pk = 0 ,
since Pk RePr → 0 for Re → ∞ . This reduces the number of A-values in the final result for
the Nusselt number. Instead of equ (15) one now has for the Nusselt number with
Kσ 1 = K k 1 − Kc p 1 , Kσ 2 = K k 2 − Kc p 2 − 2 K k 1Kc p 1 + 2 Kc2p 1 (31)
Nu
= 1 + ε ⎣⎡K ρ 1 Aρ + K μ 1 Aμ + Kσ 1 Aσ ⎦⎤ + ε 2 ⎣⎡K ρ 2 Aρ 2 + K μ 2 Aμ 2 + Kσ 2 Aσ 2
Nucp (32)
2 2 2
+ K ρ Aρρ + K μ Aμμ + Kσ Aσσ + K ρμ Aρμ + K ρσ Aρσ + K μσ Aμσ ⎤⎦ + O ε ( )3
The new haj -values (c.f. equ (17)) for the determination of the A-values are
(
hσ 1 = ε Kσ 1 = ε K k 1 − Kc p 1 ) (33)
(
hσ 2 = ε 2 Kσ 2 = ε 2 K k 2 − Kc p 2 − 2K k 1Kc p 1 + 2 Kc2p 1 ) (34)
Table 1 shows the reduction in terms and numerical solutions necessary for the determination
of the A-values in equ (32), compared to those for equ (15).
A-values Numerical solutions

Method
1. order 2. order 1. order 2. order
HOC 4 14 8 32
HR-HOC 3 9 6 18
Table 1. Number of A-values and numerical solutions
The complex benchmark geometry in Fig. 1 is now treated with the simplified HR-HOC
approach. Fig. 2 shows that the quality of the results is as good as in Fig. 1 for the full HOC-
method.
A B C D E A B C D E
1.00 1.00
0.99
0.99
Nu/Nucp
Nu/Nucp
0.98
0.98
0.97
0.96 0.97 0
0 1 2 1 2
s s
(a) (b)
Fig. 2. Complex benchmark geometry, s. Fig. 1, ( q w* = 300 W m 2 , Re = 1.64 × 106 ) (a) heating
element 1 (b) heating element 2
3.2 Locally constant property HOC-method (LCP-HOC)

Without a real asymptotic justification all variable property effects outside the nondimensional
groups are now neglected by setting all Pa = 0 and ρ = 1. Equations (26)-(28) then are the
equations for constant properties, but with locally variable properties in Re and Pr.
Solutions of these equations can be used as an easy to get answer with respect to the
question how strong the influence of variable properties is and if it makes sense to apply the
more sophisticated methods.
In this LCP-HOC approach Re and RePr in the reduced set of equations (26)-(28) are treated
like variable properties, being called α and β, i.e.
ρ *uR* L* (35)
α = Re =
μ*
ρ *c p* uR* L*
β = RePr = (36)
k*
Instead of equ (15) one now has for the Nusselt number with
Kα 1 = K ρ 1 − K μ 1
(37)
Kα 2 = K ρ 2 − K μ 2 − 2 K ρ 1K μ 1 + 2 K μ2 1
K β 1 = K ρ 1 + Kc p 1 − K k 1
(38)
K β 2 = Kc p 2 + K ρ 2 − K k 2 + 2 Kc p 1K ρ 1 − 2K k 1Kc p 1 − 2 K ρ 1K k 1 + 2 K k21
the form:
Nu
= 1 + ε ⎡⎣Kα 1 Aα + K β 1 Aβ ⎤⎦
Nucp (39)
+ε ⎡⎣Kα 2 Aα 2 + K β 2 Aβ 2 + Kα2 1 Aαα + K β2 1 Aββ + Kα 1K β 1 Aαβ ⎤⎦ + O ε 3
2
( )
The new haj -values (c.f. equ (17)) for the determination of the A-values are
hα 1 = ε Kα 1 ; hα 2 = ε 2 Kα 2 (40)
hβ 1 = ε K β 1 ; hβ 2 = ε 2 K β 2 (41)
Table 2 shows the reduction in terms and numerical solutions necessary for the determination
of the A-values in equ (39) compared to those for equ (15).
A-values Numerical solutions

Method
1. order 2. order 1. order 2. order
HOC 4 14 8 32
LCP-HOC 2 5 4 8
Table 2. Number of A-values and numerical solutions
1.00 A B C D E
1.00
0.99
0.99
Nu/Nucp
Nu/Nu cp
0.98
0.98
0.97
0.96 0.97
0.95
0 1 2 0.96
0 1 2
s s
(a) (b)
Fig. 3. Complex benchmark geometry, s. Fig. 1, ( q w* 2
= 300 W m , Re = 1.64 × 106 ) (a) heating
element 1 (b) heating element 2
Again, the complex benchmark geometry in Fig. 1 is chosen as a test case, now for the LCP-
HOC approach. Fig. 3 shows that the results are of the same order as with the more
comprehensive methods and thus can serve as a simple first approximation. So far, there is
no explanation for the obviously constant offset in the results.
4. Alternative determination HOC-method (AD-HOC)

So far the RANS-approach has been used to cope with turbulent flows. Computational times
are moderate and the relatively large number of numerical solutions necessary for the
various HOC-methods is available in acceptable time.
This is no longer true, however, when LES or DNS solutions are aimed at. Then, finding 32
solutions for the second order HOC method (c.f. table 1 or 2) is not a reasonable procedure.
In table 3 CPU-times for RANS, LES and DNS solutions of a simple benchmark problem are
compared to illustrate the discrepancy in typical CPU-times for the three methods. The
benchmark problem is that of a differentially heated cavity with adiabatic top and bottom
walls shown in Fig. 4. In x1-direction there is a periodic boundary condition L1 apart
corresponding to an overall 2D geometry. As a result for constant properties (Boussinesq-
approximation) the Nusselt number and shear stress distribution is shown for three
different Rayleigh numbers at the hot wall in Fig. 5. Whereas DNS results from (Trias et al,
2010a; Trias et al, 2010b; Trias et al, 2007) and our own LES results are close to each other,
RANS results show larger deviations. Some numerical details are given in table 3.
Fig. 4. Differentially heated cavity of aspect ratio L3/L1 = 4, L2/L1 = 1
NUMERICAL CPU-TIME in hours NUMERICAL DETAILS

METHOD Ra = 2×109 Ra = 4×109 Ra = 1010
RANS 1 1 1 k - ω SST model;
(own results) mesh resolution: 100×218;
steady simulation.
Computer: Linux cluster (4×2.66GHz)
LES 330 330 330 k -equation eddy-viscosity model;
(own results) mesh resolution: 64×100×218;
time step: 2.4×10-2
Computer: Linux cluster (4×2.66GHz)
DNS 6200 26000 Mesh resolution
(Trias et al, (estimated) (estimated) Ra = 2×109 : 64×144×318
2010a; Ra = 1010 : 128×190×462
Trias et al, time step
2010b; Ra = 2×109 : 1.27×10-3
Trias et al, Ra = 1010 : 6.25×10-4
2007) Computer: Nec-Sx5 vectorial parallel
machine.
CPU-TIME: sum of computational time of all processors in hours
Table 3. CPU-time and numerical details for different numerical methods applied to the
benchmark problem (heated cavity), see Fig. 4
250 0.5
(a)
200 DNS 0.4
LES
150 RANS 0.3
Ra du/dy
Nu
100 DNS
-1/4
0.2
LES
RANS
50 0.1
0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
X3 X3
0.5
300
(b) LES
RANS 0.4
200
Ra du/dy 0.3
Nu
LES
-1/4
0.2
100 RANS
0.1
0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
X3 X3
0.5
(c) 400
DNS 0.4
LES
300 RANS 0.3
du/dy
Nu
200 DNS
-1/4
0.2 LES
Ra
RANS
100
0.1
0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
X3 X3
Fig. 5. Constant property results of the flow and heat transfer in a differentially heated
cavity, see Fig. 4. (a) Ra = 2 × 109 (b) Ra = 4 × 109 (c) Ra = 1 × 1010
If now variable property effects are of interest, the A-coefficients should be determined with
a RANS-approach. It is assumed that the variable and the constant property RANS results
show basically the same errors (compared to the “exact” DNS results) and thus have
systematic errors that cancel when the difference is taken in the procedure to determine the
A-values.
DNS and LES constant property results can thus be corrected with respect to variable
property effects by “alternatively determined” (i.e. AD) A-values. This procedure is called
AD-HOC-method, but is not really an ad-hoc-method!
(a) 250 0.5
200 LES, constant properties 0.4

LES, variable properties
150 LES, AD-HOC-method 0.3
Ra du/dy
Nu
-1/4
100 0.2
LES, constant properties
50 0.1 LES, variable properties
LES, AD-HOC-method
0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
X3 X3
(b) 300 0.5
250 LES, constant properties 0.4

200 LES, AD-HOC-method 0.3
du/dy
Nu
150
-1/4
0.2 LES, constant properties

Ra
100
50 0.1 LES, AD-HOC-method
0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
X3 X3
(c) 0.5
400
0.4
300 LES, variable properties
0.3
du/dy
LES, AD-HOC-method
Nu
200
-1/4
0.2
Ra

100 LES, variable properties
0.1
LES, AD-HOC-method
0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
X3 X3
Fig. 6. Variable property results of the flow and heat transfer in a differentially heated cavity,
see Fig. 4. (a) Ra = 2 × 109 (b) Ra = 4 × 109 (c) Ra = 1 × 1010 Fluid: air with properties (42)-(45)
In Fig. 6 these AD-HOC-results that account for variable properties are compared to LES
results that account for variable properties from the beginning. They are results for air with
variable ρ * , μ * , k * and c p* asymptotically accounted for to the first order and for the LES
calculations taken as
( (
ρ * = 1.188 ⋅ 1 − γ * T * − TR* )) (42)
⎛ T * − TR* ⎛ T * − TR* ⎞ ⎞
2
* ⎜ −5
μ = 1.18185 × 10 ⋅ 1 + 0.775 − 0.176 ⎜⎜ ⎟ ⎟ (43)
⎜ TR* ⎝ TR* ⎟⎠ ⎟
⎝ ⎠
⎛ T * − TR* ⎛ T * − TR* ⎞ ⎞
2
k * = 0.025721 ⋅ ⎜ 1 + 0.891 − 0.1285 ⎜⎜ ⎟ ⎟ (44)

⎜ TR* ⎝ TR* ⎟⎠ ⎟
⎝ ⎠
⎛ T * − TR* ⎛ T * − TR* ⎞ ⎞
2
c p* = 1014 ⋅ ⎜ 1 + 0.068 − 0.038 ⎜

⎜ T* ⎟ ⎟ ⎟ ⎟ (45)
⎜ TR* ⎝ ⎠ ⎠
⎝ R
where the reference temperature TR* = 293K and the expansion coefficient γ * = 0.003413 .
The density variations are accounted for by the Boussinesq approximation like in (Trias et
al, 2010a; Trias et al, 2010b; Trias et al, 2007). Fig. 6 shows that with the influence of variable
properties determined by the AD-HOC-method there is a very good coincidence with LES
results that account for variable property effects from the beginning. Thus the AD-HOC-
method is an attractive approach when LES and DNS results have to be corrected with
respect to the influence of variable properties.
DNS
RANS
Fig. 7. Constant property results of the temperature field in a differentially heated cavity, see
Fig. 4, Ra = 2 × 108
This is also true with respect to the field variables of these cases. For example, Fig. 7 shows
that there is an appreciable difference when the temperature field is calculated by DNS
compared to the RANS results.
However, as shown in Fig. 8, the iso-temperature lines for variable properties, calculated by
DNS are well represented by iso-lines from the AD-HOC method, i.e. those lines from
constant property DNS results corrected by A-values from RANS solutions for variable
properties.
DNS, constant properties

DNS, variable properties
DNS, AD-HOC method
Fig. 8. Variable property results of the temperature field in a differentially heated cavity, see
Fig. 4, Ra = 2 × 108
Distribution of the first order A-values Aγ , Aμ , Ak and Ac p , computed by RANS, are
shown in Fig. 9. The variable properties behave differently in the core region, where a quasi-
laminar flow prevails and in the large vortex region near the bottom and top walls. Also,
signs within one region are different. For example Aγ and Aμ are negative in the core
region, whereas Ak and Ac p are positive in the same region. Altogether there is a non-
negligible effect of variable properties on the temperature distribution.
(a) (b)
Aγ Aμ
(c) (d)
Ak Ac p
Fig. 9. A-values of temperature computed by RANS in a differentially heated cavity, see

Fig. 4, Ra = 2 × 108 (a) Aγ ; (b) Aμ ; (c) Ak ; (d) Ac p
5. Conclusions
Various methods to account for variable property effects in complex geometries and highly
sophisticated numerical methods have been proposed. Due to the Taylor series expansions
of all properties, which are the starting point for all methods, the influence of variable
properties can be accounted for in a general manner, i.e. for all (small) heating rates and for
all Newtonian fluids. For a special problem with A-values determined once, the effect of
variable properties can be found in the final result by fixing ε (the heat transfer rate) and all
K-values (from the fluid of interest).
This way of treating variable property effects is much closer to the physics than empirical
methods like the property ratio and the reference temperature methods are.
6. Nomenclature
Aa , Aaa , Aab , Aa 2 { }
A-values, a , b ∈ ρ , γ , μ , k , c p , a ≠ b
cp specific heat capacity
f friction factor
G
g gravity vector
haj h-values, ε j K aj
j empirical parameter
K a , K a 2 , K an K-values, a = ρ , γ , μ , k , c p
k heat conductivity
L* characteristic length
ma , na empirical exponents
Nu Nußelt number
p pressure
Pr Prandtl number
Pa nondimensional properties about variable fluid property
q heat flux
Re Reynolds number
T temperature
G
u velocity vector
V variables
Greek symbols
α , β ,σ , δ variables composed of fluid properties
γ expansion coefficient
ΔT * temperature difference
ε nondimensional temperature difference
μ dynamic viscosity
ρ density
subscripts
cp constant properties
R reference state
* dimensional
7. Acknowledgement
This study was supported by the DFG (Deutsche Forschungsgemeinschaft).
8. References
Bünger F. & Herwig H. (2009). An extended similarity theory applied to heated flows in
complex geometries. ZAMP, Vol. 60, (2009), pp. 1095-1111.
Carey V. P. & Mollendorf J. C. (1980). Variable viscosity effects in several natural convection
flows. Int. J. Heat Mass Transfer, Vol. 23, (1980), pp. 95-109
Debrestian D. J. & Anderson J. D. (1994). Reference Temperature Method and Reynolds
Analogy for Chemically Reacting Non-equilibrium Flowfields. J. of Thermophysics
and Heat Transfer, Vol. 8, (1994), pp. 190-192
Herwig H. & Wickern G. (1986). The Effect of Variable Properties on Laminar Boundary
Layer Flows. Wärme- und Stoffübertragung, Vol. 20, (1986), pp. 47-57
Herwig H. & Bauhaus F. J. (1986). A Regular Perturbation Theory for Variable Properties
Applied to Compressible Boundary Layers. Proceedings of 8th International Heat
Transfer Conference, San Fransisco, Vol. 3, 1095-1101, 1986
Herwig H., Voigt M. & Bauhaus F. J. (1989). The Effect of Variable Properties on Momentum
and Heat Transfer in a Tube with constant Wall Temperature. Int. J. Heat Mass
Transfer, Vol. 32, (1989), pp. 1907-1915
Herwig H. & Schäfer P. (1992). Influence of variable properties on the stability of two-
dimensional boundary layers. J. Fluid Mechanics, Vol. 243, (1992), pp. 1-14.
Jayari S.; Dinesh K. K., & Pillai K. L. (1999). Thermophoresis in natural convection with
variable properties. Heat and Mass Transfer, Vol. 35, (1999), pp. 469-475
Jin Y. & Herwig H. (2010). Application of the Similarity Theory Including Variable Property
Effects to a Complex Benchmark Problem, ZAMP, Vol. 61, (2010), pp. 509-528
Jin Y. & Herwig H. (2010). Efficient method to account for variable property effects in
numerical momentum and heat transfer solutions, Int. J. Heat Mass Transfer, (2011),
in press
Li Z.; Huai X.; Tao Y. & Chen H. (2007). Effects of thermal property variations on the liquid
flow and heat transfer in micro-channel heat sinks. Applied Thermal Engineering, Vol.
27 (2007), pp. 2803–2814
Mahmood G. I.; Ligrani P. M. & Chen K. (2003). Variable property and Temperature Ratio
Effects on Nusselt Numbers in a Rectangular Channel with 45 Deg Angle Rib
Turbulators. J. Heat Transfer, Vol. 125 (2003), pp. 769-778
Trias F. X.; Soria M.; Oliva A. & Pérez-Segarra C. D. (2007). Direct numerical simulations of
two- and three-dimensional turbulent natural convection flows in a differentially
heated cavity of aspect ratio 4, J. Fluid Mech., Vol. 586, (2007), pp. 259–293
Trias F. X.; Gorobets A.; Soria M. & Oliva A. (2010a). Direct numerical simulation of a
differentially heated cavity of aspect ratio 4 with Rayleigh numbers up to 1011 –
Part I: Numerical methods and time-averaged flow, International Journal of Heat and
Mass Transfer, Vol. 53, (2010), pp. 665–673
Trias F. X.; Gorobets A.; Soria M. & Oliva A. (2010b). Direct numerical simulation of a
differentially heated cavity of aspect ratio 4 with Rayleigh numbers up to 1011 –
Part II: Numerical methods and time-averaged flow, International Journal of Heat and
Mass Transfer, Vol. 53, (2010), pp. 674–683
9
Bioheat Transfer
Alireza Zolfaghari1 and Mehdi Maerefat2
1Department of Mechanical Engineering, Birjand University, Birjand,
2Department of Mechanical Engineering, Tarbiat Modares University, Tehran,
Iran
1. Introduction
Heat transfer in living tissues is a complicated process because it involves a combination of
thermal conduction in tissues, convection and perfusion of blood, and metabolic heat
production. Over the years, several mathematical models have been developed to describe
heat transfer within living biological tissues. These models have been widely used in the
analysis of hyperthermia in cancer treatment, laser surgery, cryosurgery, cryopreservation,
thermal comfort, and many other applications. The most widely used bioheat model was
introduced by Pennes in 1948. Pennes proposed a new simplified bioheat model to describe
the effect of blood perfusion and metabolic heat generation on heat transfer within a living
tissue. Since the landmark paper by Pennes (1948), his model has been widely used by many
researchers for the analysis of bioheat transfer phenomena. And, also a large number of
bioheat transfer models have been proposed to overcome the shortcomings of Pennes’
equation. These models include the continuum models which consider the thermal impact
of all blood vessels as a global parameter and the vascular models which consider the
thermal impact of each vessel individually.
Although, several bioheat models have been developed in the recent years, the
thermoregulatory control mechanisms of the human body such as shivering, regulatory
sweating, vasodilation, and vasoconstriction have not been considered in these models. On
the other hand, these mechanisms may significantly influence the thermal conditions of the
human body. This causes a serious limitation in using the bioheat models for evaluating the
human body thermal response. In order to remove this limitation, Zolfaghari and Maerefat
(2010) developed a new Simplified Thermoregulatory Bioheat (STB) model based on the
combination of the well-known Pennes’ equation and Gagge’s thermal comfort model.
The present chapter aims at giving a concise introduction to bioheat transfer and the
mathematical models for evaluating the heat transfer within biological tissues. This chapter
is divided into six sections. The first section presents an introduction to the concept and
history of bioheat transfer. The structure of living tissues with blood perfusion is described
in section 2. Next, third section focuses on the mathematical modelling of heat transfer in
living tissues. In the mentioned section, a brief description of some of the most important
bioheat models (i.e. Pennes (1948) model, Wulff (1974) model, Klinger (1974) model, Chen
and Holmes (1980) model and so on) is presented. Afterwards, section 4 explains the
complexity of evaluating heat transfer within the tissues that thermally controlled by
thermoregulatory mechanisms such as shivering, regulatory sweating, vasodilation, and
vasoconstriction. Then, the Simplified Thermoregulatory Bioheat (STB) model is introduced

for evaluating heat transfer within the segments of the human body. Finally, section 5
outlines the main conclusions and recommendations of the research. Moreover, the selected
references are listed in the last section.
2. Structure of blood perfused tissues

Before we discuss the bioheat models, let us have a brief look at the structure of blood
perfused tissues. The biological tissues include the layers of skin, fat, muscle and bone.
Moreover, the skin is composed of two stratified layers: epidermis and dermis. Fig. 1 shows
a schematic geometry of the tissue structure. Furthermore, the thermophysical properties of
the human body tissue are provided in Table 1 (Lv & Liu, 2007; Sharma, 2010).
Fig. 1. Schematic geometry of the tissue structure (figure not to scale)
Specific Blood perfusion Thermal

Thickness Density
heat rate conductivity
l (m) ρ (kg/m3) C (J/kg K) Wbl (m3/s m3) k (W/m K)
Epidermis 80×10-6 1200 3589 0 0.24
Dermis 0.002 1200 3300 0.00125 0.45
Fat ≈ 0.010 937 3258 0.00125 0.21
Muscle ≈ 0.020 1000 4000 0.00125 0.5
Bone ≈ 0.008 1920 1440 - 0.44
Blood - 937 3889 - 0.64
Table 1. Geometrical information and thermal properties of the human body tissue (Lv & Liu,
2007; Sharma, 2010)
Blood circulation is a key mechanism for regulating the body temperature. The circulatory
system of the human body comprises of two sets of blood vessels (arteries and veins) which
carry blood from the heart and back. Blood leaves the heart through the aorta, which is the
largest artery (diameter ≈ 5000 μm). Vessels supplying blood to muscles are known as main
supply arteries and veins (SAV, 300-1000 μm diameter). They branch into primary arteries,
(P, 100-300 μm diameter) which feed the secondary arteries (s, 50-100 μm diameter). These
Bioheat Transfer 155
vessels deliver blood to the arterioles (20-40 μm diameter) which supply blood to the
smallest vessels known as capillaries (c, 5-15μm diameter). Blood is returned to the heart
through a system of vessels known as veins. Fig. 2 shows a schematic diagram of a typical
vascular structure (Jiji, 2009).
Fig. 2. Schematic diagram of the vascular system (Jiji, 2009)
Fig. 3. Schematic of temperature equilibration between the blood and the tissue (Datta, 2002)
Blood leaves the heart at the arterial temperature Tart. It remains essentially at this
temperature until it reaches the main arteries where equilibration with surrounding tissue
begins to take place. Equilibration becomes complete prior to reaching the arterioles and
capillaries. Beyond this point, blood temperature follows the solid tissue temperature (Tti)
through its spatial and time variations until blood reaches the terminal veins. At this point
the blood temperature ceases to equilibrate with the tissue, and remains virtually constant,
except as it mixes with other blood of different temperatures at venous confluences. Finally,
the cooler blood from peripheral regions and warmer blood from internal organs mix within
the vena cava and the right atrium and ventricle. Following thermal exchange in the
pulmonary circulation and remixing in the left heart, the blood attains the same temperature
it had at the start of the circuit (Datta, 2002). Fig. 3 shows a schematic of temperature
equilibration between the blood and the solid tissue.
3. Mathematical models of bioheat transfer

3.1 Pennes model
Over the years, the effects of blood flow on heat transfer in living tissue have been studied
by many researchers and a large number of bioheat transfer models have been developed on
the basis of two main approaches: the continuum approach and the discrete vessel
(vascular) approach. In the continuum approach, the thermal impact of all blood vessels
models with a single global parameter; and the vascular approach models the impact of
each vessel individually (Raaymakers et al., 2009). The most widely used continuum model
of perfused tissue was introduced in 1948 by Harry Pennes. The Pennes (1948) model was
initially developed for predicting heat transfer in the human forearm. Due to the simplicity
of the Pennes bioheat model, it was implemented in various biological research works such
as for therapeutic hyperthermia for the treatment of cancer (Minkowycz et al., 2009).
Pennes bioheat model is based on four simplifying assumptions (Jiji, 2009):
1. All pre-arteriole and post-venule heat transfer between blood and tissue is neglected.
2. The flow of blood in the small capillaries is assumed to be isotropic. This neglects the
effect of blood flow directionality.
3. Larger blood vessels in the vicinity of capillary beds play no role in the energy
exchange between tissue and capillary blood. Thus the Pennes model does not consider
the local vascular geometry.
4. Blood is assumed to reach the arterioles supplying the capillary beds at the body core
temperature. It instantaneously exchanges energy and equilibrates with the local tissue
temperature.
Based on these assumptions, Pennes (1948) modeled blood effect as an isotropic heat source
or sink which is proportional to blood flow rate and the difference between the body core
temperature and local tissue temperature. Therefore, Pennes (1948) proposed a model to
describe the effects of metabolism and blood perfusion on the energy balance within tissue.
These two effects were incorporated into the standard thermal diffusion equation, which is
written in its simplified form as:
∂Tti
ρ tiC ti = ∇.kti ∇Tti + ρ blC bl Wbl (Tart − Tti ) + qm (1)
∂t
where ρti, Cti, Tti and kti are, respectively, the density, specific heat, temperature and thermal
conductivity of tissue. Also, Tart is the temperature of arterial blood, qm is the metabolic heat
generation and ρbl, Cbl and Wbl are, respectively, the density, specific heat and perfusion rate
of blood. It should be noted that metabolic heat generation is assumed to be homogeneously
distributed throughout the tissue. Also, it is assumed that the blood perfusion effect is
homogeneous and isotropic and that thermal equilibration occurs in the microcirculatory
capillary bed. In this scenario, blood enters capillaries at the temperature of arterial blood,
Tart, where heat exchange occurs to bring the temperature to that of the surrounding tissue,
Tti. There is assumed to be no energy transfer either before or after the blood passes through
the capillaries, so that the temperature at which it enters the venous circulation is that of the
local tissue (Kreith, 2000).
Pennes (1948) performed a series of experimental studies to validate his model. Validations
have shown that the results of Pennes bioheat model are in a reasonable agreement with the
experimental data. Although Pennes bioheat model is often adequate for roughly describing
the effect of blood flow on the tissue temperature, there exist some serious shortcomings in
his model due to its inherent simplicity. The shortcomings of Pennes bioheat model come
from the basic assumptions that are introduced in this model. These shortcomings can be
listed as follows (Jiji, 2009):
1. Thermal equilibration does not occur in the capillaries, as Pennes assumed. Instead it
takes place in pre-arteriole and post-venule vessels having diameters ranging from 70-
500 μm.
2. Directionality of blood perfusion is an important factor in the interchange of energy
between vessels and tissue. The Pennes equation does not account for this effect.
3. Pennes equation does not consider the local vascular geometry. Thus significant
features of the circulatory system are not accounted for. This includes energy exchange
with large vessels, countercurrent heat transfer between artery-vein pairs and vessel
branching and diminution.
4. The arterial temperature varies continuously from the deep body temperature of the
aorta to the secondary arteries supplying the arterioles, and similarly for the venous
return. Thus, contrary to Pennes’ assumption, pre-arteriole blood temperature is not
equal to body core temperature and vein return temperature is not equal to the local
tissue temperature. Both approximations overestimate the effect of blood perfusion on
local tissue temperature.
To overcome these shortcomings, a considerable number of modifications have been
proposed by various researchers. Wulff (1974) and Klinger (1974) considered the local blood
mass flux to account the blood flow direction, while Chen and Holmes (1980) examined the
effect of thermal equilibration length on the blood temperature and added the dispersion
and microcirculatory perfusion terms to the Klinger equation (Vafai, 2011). In the following
sections, a brief review of the modified bioheat models will be given.
3.2 Wulff continuum model

Due to the simplicity of the Pennes model, many authors have looked into the validity of the
assumptions used to develop the Pennes bioheat equation. Wulff (1974) was one of the first
researchers that directly criticized the fundamental assumptions of the Pennes bioheat
equation and provided an alternate analysis (Cho, 1992). Wulff (1974) assumed that the heat
transfer between flowing blood and tissue should be modeled to be proportional to the
temperature difference between these two media rather than between the two bloodstream
temperatures (i.e., the temperature of the blood entering and leaving the tissue). Thus, the
energy flux at any point in the tissue should be expressed by (Minkowycz et al., 2009)
q = − kti∇Tti + ρ bl hbl vh (2)
where vh is the local mean blood velocity. Moreover, hbl is the specific enthalpy of the blood
and it is given by
Tbl P
hbl = ∫ C bl (Tbl* )dTbl* + + ΔH f (1 − φ ) (3)
To ρ bl
where P is the system pressure, ΔHf is the enthalpy of formation of the metabolic reaction,
and φ is the extent of reaction. Also, To and Tbl are the reference and blood temperatures,
respectively. Thus, the energy balance equation can be written as
∂Tti
ρ tiC ti = −∇.q (4)
∂t
Therefore,
∂Tti ⎡ ⎛ Tbl P ⎞⎤
ρ tiC ti = −∇. ⎢ − kti ∇Tti + ρ bl vh ⎜ ∫ C bl (Tbl* )dTbl* + + ΔH f (1 − φ ) ⎟ ⎥ (5)
∂t ⎢⎣ ⎝ To ρ bl ⎠ ⎥⎦
Neglecting the mechanical work term (P/ρbl ), setting the divergence of ρ bl vh to zero, and
assuming constant physical properties, Eq. (5) can be simplified as follows (Minkowycz et
al., 2009):
∂Tti
ρ tiC ti = kti∇ 2Tti − ρ blC bl vh ∇Tbl + ρ bl vh ΔH f ∇φ (6)
∂t
Since blood is effectively microcirculating within the tissue, it will likely be in thermal
equilibrium with the surrounding tissue. As such, Wulff (1974) assumed that Tbl is
equivalent to the tissue temperature Tti. In this condition, the metabolic reaction term
( ρ bl vh ΔH f ∇φ ) is equivalent to qm. Therefore, the final form of the bioheat equation that was
derived by Wulff (1974) is
∂T
ρ tiC ti = kti ∇ 2Tti − ρ blC bl vh .∇Tti + qm (7)
∂t
It should be noted that the main challenge in solving this bioheat equation is in the
evaluation of the local blood mass flux ρ bl vh (Minkowycz et al., 2009).
3.3 Klinger continuum model

In 1974, Klinger presented an analytical bioheat model that was conceptually similar to
Wulff bioheat model. Klinger (1974) argued that in utilizing the Pennes model, the effects of
nonunidirectional blood flow were being neglected and thus significant errors were being
introduced into the computed results. In order to correct this lack of directionality in the
formulation, Klinger (1974) proposed that the convection field inside the tissue should be
modeled based upon the in vivo vascular anatomy (Cho, 1992). Taking into account the
spatial and temporal variations of the velocity and heat source, and assuming constant
physical properties of tissue and incompressible blood flow, the Klinger bioheat equation
was expressed as:
∂T
ρ tiC ti = kti ∇ 2Tti − ρ blC bl v.∇Tti + qm (8)
∂t
This equation is similar to that derived by Wulff (1974), except it is written for the more
general case of a spatially and temporally nonuniform velocity field (v) and heat source (qm).
3.4 Chen-Holmes (CH) continuum model

Among the continuum bioheat models, the Chen-Holmes model (Chen & Holmes, 1980) is
the most developed (Kreith, 2000). Chen and Holmes (1980) showed that the major heat
transfer processes occur in the 50 to 500μm diameter vessels. Consequently, they proposed
that larger vessels be modeled separately from smaller vessels and tissue. Therefore, in the
Chen-Holmes bioheat model, the total tissue control volume is subdivided to the solid-
tissue subvolume (Vs) and blood subvolume (Vb) as shown in Fig. 4. By using this concept,
Chen and Holmes (1980) proposed a new modified relationship for calculating the blood
perfusion term (qbl) in their bioheat model:
qbl = ρ blC bl Wbl* (Tart

*
− Tti ) − ρ blC bl v.∇Tti + ∇.kp∇Tti (9)
Fig. 4. Schematic representation of tissue control volume as used in Chen-Holmes model

(Kreith, 2000)
In Eq. (9), the first term on the right hand side appears similar to Pennes’ perfusion term
except the perfusion rate ( Wbl* ) and the arterial temperature ( Tart
*
) are specific to the volume
*
being considered. It should be noted that Tart is essentially the temperature of blood
upstream of the arterioles and it is not equal to the body core temperature. The second term
in Eq. (9) accounts for energy convected due to equilibrated blood. Directionality of blood
flow is described by the vector v, which is the volumetric flow rate per unit area. The third
term in Eq. (9) describes conduction mechanisms associated with small temperature
fluctuations in equilibrated blood. The symbol kp denotes “perfusion conductivity”. It is a
function of blood flow velocity, vessel inclination angle relative to local temperature
gradient, vessel radius and number density.
Using a simplified volume-averaging technique, the Chen-Holmes bioheat equation can be
written as follows:
∂Tti
ρ ti,effC ti,eff
∂t
{
= ∇.kti,eff ∇Tti − ρ blC bl Wbl* (Tart
*
}
− Tti ) − ρ blC bl v.∇Tti + ∇.kp∇Tti + qm (10)
where
ρ ti,eff = (1 − ε bl )ρ ti + ε bl ρ bl (11)
C ti,eff = (1 − ε bl )C ti + ε blC bl (12)
kti,eff = (1 − ε bl )kti + ε bl kbl (13)
where ε bl is the porosity of the tissue where blood flows and Tti* is the local mean tissue
temperature expressed as (Minkowycz et al., 2009)
(1 − ε bl )ρ tiC tiTti + ε bl ρ blC bl kbl

Tti* = (14)
ρ ti,effC ti,eff
Since ε bl <<1, it follows that kti,eff is independent of blood flow and equal to the conductivity
of the solid tissue (kti).
Although the Chen-Holmes model represents a significant improvement over Pennes’
equation, it is not easy to implement since it requires detailed knowledge of the vascular
network and blood perfusion. Furthermore, the model does not explicitly address the effect
of closely spaced countercurrent artery-vein pairs (Kreith, 2000).
3.5 Weinbaum, Jiji and Lemons (WJL) bioheat model

The modeling of countercurrent vascular system, which was not explicitly addressed by the
Chen-Holmes model, developed separately from that of the continuum models. In 1984,
Weinbaum, Jiji and Lemons presented a new vascular bioheat model by considering the
countercurrent blood flow. This model was obtained based on a hypothesis that small
arteries and veins are parallel and the flow direction is countercurrent, resulting in
counterbalanced heating and cooling effects (Fig. 5). It should be noted that this assumption
is mainly applicable within the intermediate tissue of the skin (Minkowycz et al., 2009). In
an anatomic study performed on rabbit limbs, Weinbaum et al. (1984) identified three
vascular layers (deep, intermediate, and cutaneous) in the outer 1cm tissue layer. For the
countercurrent structure of the deep tissue layer, they proposed a system of three coupled
equations:
dTart
ρ blC bl π r 2 v. = −qart (15)
ds
dTv
ρ blC bl π r 2 v. = −q v (16)
ds
∂Tti ⎧ d(Tart − Tv ) ⎫
ρ tiC ti = ∇.kti ∇Tti + ⎨n g ρ blC bl (Tart − Tv ) − ρ blC bl nπ r 2 v. ⎬ + qm (17)
∂t ⎩ ds ⎭
where qart is the heat loss from the artery by conduction through its wall, qv is the heat gain
by conduction per unit length through the vein wall into the vein, Tart and Tv are the bulk
mean temperatures inside the blood vessel, r is the vessel radius, v is the mean velocity in
either the artery or vein, n is the number of arteries or veins, and g is the perfusion bleed-off
velocity per unit vessel surface area (s). The first two equations describe the heat transfer of
the thermally significant artery and vein, respectively. The third equation refers to the tissue
surrounding the artery–vein pair. In Eq. (17), the middle two right-hand-side terms
represent the capillary bleed-off energy exchange, and the net heat exchange between the
tissue and artery–vein pair, respectively. The capillary bleed-off term is similar to Pennes’
perfusion term except the bleed-off mass flow (g) is used. Their analysis showed that the
major heat transfer is due to the imperfect countercurrent heat exchange between artery–
vein pairs. They quantified the effect of perfusion bleed-off associated with this vascular
structure, and showed that Pennes’ perfusion formulation is negligible due to the
temperature differential (Kreith, 2000).
Fig. 5. Schematic of artery and vein pair in peripheral skin layer (Kreith, 2000)
Assumptions of the Weinbaum-Jiji-Lemons model include the following (Kreith, 2000):
1. Neglecting the lymphatic fluid loss, so that the mass flow rate in the artery is equal to
that of the vein.
2. Spatially uniform bleed-off perfusion.
3. Heat transfer in the plane normal to the artery–vein pair is greater than that along the
vessels (in order to apply the approximation of superposition of a line sink and source
in a pure conduction field).
4. A linear relationship for the temperature along the radial direction in the plane normal
to the artery and vein.
5. The artery–vein border temperature equals the mean of the artery and vein
temperature.
6. The blood exiting the bleed-off capillaries and entering the veins is at the venous blood
temperature.
The last assumption has drawn criticism based on studies that indicate the temperature to
be closer to tissue. Limitations of this model include the difficulty of implementation, and
that the artery and vein diameters must be identical (Kreith, 2000).
3.6 Simplified Weinbaum-Jiji (WJ) model

Since both Tart and Tv are unknowns in Equation (17), the tissue temperature Tti cannot be
determined. Therefore, Weinbaum and Jiji (1985) derived a simplified single equation to
study the influence of blood flow on the tissue temperature distribution. In order to
eliminate the artery and vein temperatures from their previous formulation (Weinbaum et
al., 1984), two major assumptions were used:
1. Tissue temperature Tti is approximated by the average of the local artery and vein
temperatures. That is
Tart + Tv (18)
Tti ≈
2
2. Heat from a paired artery is mostly conducted to the corresponding vein:
qart ≈ q v ≈ σ Δ kti (Tart − Tv ) (19)
where σ Δ is a geometrical shape factor and it is associated with the resistance to heat
transfer between two parallel vessels embedded in an infinite medium (Jiji, 2009). For the
case of vessels at uniform surface temperatures with center to center spacing l, the shape
factor is given by (Chato, 1980)
π (20)
σΔ =
cosh −1 ( l / 2r )
By using the mentioned assumptions and substituting the Eqs. (18) and (19) in Eqs. (15), (16)
and (17), Weinbaum and Jiji (1985) proposed a simplified equation for evaluating the tissue
temperature distribution:
∂Tti (21)
ρ tiC ti = ∇.keff ∇Tti + qm
∂t
where keff is the effective conductivity, defined as
keff = kti ⎡⎣1 + PeiV (ξ ) ⎦⎤ (22)
where ξ is a dimensionless distance and it defines as x/L and L is the tissue layer thickness.
Also, V (ξ ) is dimensionless vascular geometry function and it can be calculated if the
vascular data are available. Furthermore, Pei is the inlet Peclet number; which is defined as
(Jiji, 2009)
2 ρ blC bl ri vi (23)
Pei =
kbl
where ri and vi are the vessel radius and the blood velocity at the inlet to the tissue layer at
x=0.
The main limitations of the Weinbaum-Jiji bioheat equation are associated with the
importance of the countercurrent heat exchange. It was derived to describe heat transfer in
peripheral tissue only, where its fundamental assumptions are most applicable. In tissue
area containing a big blood vessel (>200 μm in diameter), the assumption that most of the
heat leaving the artery is recaptured by its countercurrent vein could be violated; thus, it is
not an accurate model to predict the temperature field. Furthermore, unlike the Pennes
bioheat equation, which requires only the value of local blood perfusion rate, the
Weinbaum-Jiji bioheat model requires many detailed anatomical and vascular data such as
the vessel number density, size, and artery-vein spacing for each vessel generation, as well
as the blood perfusion rate. These anatomic data are normally not available for most blood
vessels in the thermally significant range (Kutz, 2009).
4. Bioheat transfer in physiologically controlled tissues

4.1 Thermoregulatory control mechanisms of the human body
Thermoregulation is the ability of an organism to keep its body temperature within certain
boundaries, even when the surrounding temperature is very different. The hypothalamus
regulates the body temperature by the thermoregulatory mechanisms such as vasomotion,
shivering and regulatory sweating. It receives inputs from central and peripheral
temperature receptors situated in the ‘core’ and in the outer ‘shell’. Temperature-sensitive
receptors in the core are found in the hypothalamus, spinal cord, abdominal viscera and the
great veins. They respond to temperatures between 30°C and 42°C. Peripheral receptors are
located in the skin and they contain two types of thermoreceptors: warm receptors and cold
receptors. The hypothalamic response to a thermal stimulus depends on the integration of
both central and peripheral stimuli (Campbell, 2008). The intensity of these stimuli depends
on the difference between the temperature of each compartment of the body (core or skin)
and its related neutral temperature. In 1988, Doherty and Arens named the mentioned
temperature difference as the thermal signals. The thermoregulatory mechanisms of the
body are controlled by these thermal signals. The cold and warm signals of human body for
skin and core compartments are defined as follows (Doherty & Arens, 1988)
WSIGcr = Max{0, Tcr − Tcr,n } (24)
CSIGcr = Max{0, Tcr,n − Tcr } (25)
WSIGsk = Max{0, Tsk − Tsk,n } (26)
CSIGsk = Max{0, Tsk,n − Tsk } (27)
where CSIG and WSIG, respectively, represent cold and warm signals of the human body,
Tsk,n is neutral skin temperature (≈ 33.7ºC), and Tcr,n is neutral core temperature (≈ 36.8ºC).
The thermal neutrality state of the human body occurs when the body is able to maintain its
thermal equilibrium with the environment with minimal regulatory effort (Yigit, 1999).
The thermoregulatory mechanisms of the human body are related to the aforementioned
thermal signals of the body. One of these thermoregulatory mechanisms is vasomotion.
Vasomotion of blood vessels (vasoconstriction and vasodilation) is caused by cold/warm
thermal conditions and it changes the rate of blood flow ( m bl ) and also the fraction of body
mass concentrated in skin compartment (α). These parameters can be calculated as follows
(Kaynakli & Kilic, 2005)
α = 0.0418 + 0.745 /(3600m

bl + 0.585) (28)
where
6.3 + 200 WSIGcr (29)

bl =
m
3600(1 + 0.5CSIGsk )
The other thermoregulatory mechanism of the human body is shivering under cold
sensation. Shivering is an increase of heat production during cold exposure due to increased
contractile activity of skeletal muscles (Wan & Fan, 2008). Shivering and muscle tension may
generate additional metabolic heat. Total metabolic heat production of body includes the
metabolic rate due to activity (Mact) and the shivering metabolic rate (Mshiv). Therefore
M = Mact + Mshiv (30)
and
Mshiv = 19.4CSIGskCSIGcr (31)
Another thermoregulatory mechanism of the body is regulatory sweating. Sweating causes

the latent heat loss from the skin. The rate of the sweat production per unit of skin area can
be estimated by the following equation (Kaynakli & Kilic, 2005)
rsw = 4.7 × 10 −5 WSIGb exp( WSIGsk / 10.7)

m (32)
where
WSIGb = Max{0, Tb − Tb,n } (33)
and
Tb = α Tsk + (1 − α )Tcr (34)
Tb,n = α Tsk,n + (1 − α )Tcr,n (35)
where WSIGb is warm signal of body, Tb is body temperature (ºC), and Tb,n is the neutral
temperature of body (ºC).
The regulatory sweating leads to an increase in the skin wettedness. The total skin
wettedness is composed of wettedness due to diffusion through the skin (wdif) and
regulatory sweating (wrsw). Therefore
wskin = wdif + wrsw (36)
where
wdif = 0.06(1 − wrsw ) (37)
rsw hfg
m
wrsw = (38)
qevap,max
where hfg is the heat of vaporization of water and qevap,max is the maximum evaporative
potential and can be estimated by the following equation (Kaynakli & Kilic, 2005)
Psk(s) − Pa
qevap,max = (39)
Re,t
where Psk(s) is water vapor pressure in the saturated air at the skin temperature (kPa) and Re,t
is the total evaporative resistance between the body and the environment (m2kPa/W).
4.2 Simplified thermoregulatory bioheat (STB) model

The human body thermal response may be significantly affected by thermoregulatory
mechanisms of the human body such as shivering, regulatory sweating and vasomotion.
But, these thermoregulatory mechanisms have not been considered in the well-known
Pennes model and also in the other modified bioheat models. In addition, although the body
core temperature could be changed depending on personal/environmental conditions, it is
commonly assumed as a constant value in Pennes model. Therefore, it seems that the well-
known Pennes bioheat model must be modified for using in human thermal response
applications. In 2010, Zolfaghari and Maerefat (2010) developed a new simplified
thermoregulatory bioheat model (STB model) on the basis of two main objectives: the first is
to supplement the thermoregulatory mechanisms to Pennes bioheat model, and the second
is to consider the body core temperature as a variant parameter depending on
personal/environmental conditions. In order to reach the mentioned objectives, Zolfaghari
and Maerefat (2010) developed their bioheat model by combining Pennes’ equation and
Gagge’s two-node model. By using this concept, they presented an energy balance equation
for core compartment of the human body as follows
dTcr (K + C bl m
bl )(Tcr − Tsk )
(1 − α )ρ bC b = rm qm − eff (40)
dt Ab
where ρb is specific heat of the body (kg/m3), Cb is specific heat of body (J/kgºC), rm is
remaining metabolic coefficient, Qm is the volumetric metabolic heat generation (W/m3), Keff
is the effective conductance between core and skin compartments, Cbl is specific heat of
blood (J/kgºC), and A b is characteristic length of the body (m) and it is defined as follows
Vb
Ab = (41)
AD
where Vb is the volume of the human body (m3) and AD is the nude body surface area (m2).
AD is described by the well-known DuBois formula (DuBois and DuBois, 1916)
AD = 0.202m0.425l0.725 (42)
where m and l are body mass (kg) and height (m). Also, remaining metabolic coefficient (rm)
is defined as follows
rm = 1 − η − 0.0014(34 − Ta ) − 0.0173(5.87 − Pa ) (43)
where η is external mechanical efficiency (Fanger, 1970) and
W
η= (44)
M
It should be noted that the external mechanical efficiency (η) is insignificant in many human
thermal response applications. Also, the respiratory heat losses are negligible compared to
the total metabolic rate. Thus, the value of rm can be approximately estimated as unity.
Discretizing Eq. (40) gives
Δt ⎡ (K eff + C bl m old ⎤
bl )(Tcrold − Tsk )
Tcrnew = Tcrold + ⎢rm qm − ⎥ (45)
(1 − α )ρ bC b ⎢⎣ Ab ⎥⎦
Eq. (45) is used as a thermal boundary condition for the human body core in the STB model.
By implementing this approach, the core temperature is not treated as a constant value and
it varies depending on personal/environmental conditions. Therefore, the main governing
equation of the STB model is
∂Tti ∂ 2Tti
ρ tiC ti = kti + ρ blC bl Wbl (Tart − Tti ) + qm (46)
∂t ∂x 2
with time-dependent boundary conditions for skin surface and body core
⎧ ∂Tti 4 4
⎪− kti ∂x = h(Tti − Ta ) + σε ((Tti + 273) − (Ta + 273) )
⎪ at skin surface
⎪⎪ + (3.054 + 16.7 hwskin )(0.256Tti − 3.37 − Pa ),
⎨ (47)
⎪
⎪ new old Δt ⎡ (K eff + C bl m old ⎤
bl )(Tcrold − Tsk )
⎪Tcr = Tcr + ⎢rm qm − ⎥ , at body core
⎩⎪ (1 − α ) ρ C ⎢
b b ⎣ A b ⎦⎥
where Ta is the surrounding air temperature, Pa is the water vapor pressure in the air, h is
convective heat transfer coefficient, ε is the skin emissivity, and σ is Stefan-Boltzmann
constant ( 5.67 × 10 -8 W/m 2 K 4 ).
It should be noted that some physiological parameters in Eqs. (46) and (47) such as qm, wskin,
α and m bl are influenced by thermoregulatory mechanisms. Also, the metabolic heat
production is related to the physical activity of the human body and it can be increased by
shivering against cold. Hence,
qm = qm,act + qm,shiv (48)
and
19.4CSIGskCSIGcr
qm,shiv = (49)
Ab
Also, skin wettedness (wskin) can be calculated from Eqs. (36) to (39). In addition, α and m
bl
are, respectively, estimated by Eq. (28) and Eq. (29).
Fig. 6 illustrates calculation steps of the STB model (Zolfaghari & Maerefat, 2010). At the
beginning, personal and environmental variables are input. Then, the initial temperature
distribution in tissue is computed by solving the steady-state bioheat equation under the
initial conditions. It should be noted that the steady-state bioheat equation can be obtained
by eliminating time dependent derivations in the well-known Pennes bioheat equation.
Afterwards, at each time step, skin and core control signals are calculated and the
thermoregulatory parameters are subsequently computed. Then, the temperature
distribution in tissue can be obtained by solving the bioheat equation which expressed by
Eq. (46) and its related boundary conditions in Eq. (47). Because of the non-linearity of the
mentioned equations, they must be solved numerically. Zolafaghari and Maerefat (2010)
used the implicit finite difference scheme to find out the temperature distribution of the
human body.
Input Parameters
(Tf, Pa, qm,act, Re,t, η, Tcr,n,Tsk,n, …)
Initializing
Initial temperature distribution

in tissue
t = t + Δt
Tsk= T(0,t), Tcr= T(L,t), Ta=T(L,t)
Control signals
(CSIGcr, CSIGsk, WSIGcr, WSIGsk)
Thermoregulatory Parameters
bl , α, Tb, m
(m rsw , wskin, qm,shiv)
Solve the bio-heat equation (Eq. (46)

with its related B.C. in Eq. (47))
yes
t < tfinal
no
Output the desired values
Fig. 6. Flow chart of the STB model calculations

The STB model has been validated against the published experimental and analytical
results, where a good agreement has been found. Zolfaghari and Maerefat (2010) showed
that the thermal conditions of the human body may be significantly affected by the
thermoregulatory mechanisms. Therefore, neglecting the control signals and
thermoregulatory mechanisms of the human body can cause a significant error in evaluating
the body thermal conditions. Zolfaghari and Maerefat (2010) compared the results of STB
model with the results of Pennes bioheat model. This comparison was performed against
Stolwijk and Hardy (1966) measured data for a step change in ambient temperature from
30ºC/40%RH to 48ºC/30%RH for an exposure period of 2 hours followed by 1 hour of
environment at 30ºC/40%RH. Fig. 7 shows the measured skin temperature data of Stolwijk
and Hardy (1966) and the simulation results of Zolfaghari and Maerefat (2010) for STB and
Pennes bioheat models. It can be clearly seen that the Pennes bioheat model is not able to
accurately estimate the skin temperature under hot environmental conditions. As shown in
Fig. 7, the Pennes bioheat model overestimates the value of the skin temperature more than
3.5ºC under the mentioned extremely hot conditions. This inaccuracy may be caused by
neglecting the thermoregulatory mechanisms such as regulatory sweating and vasomotion
in Pennes bioheat model. However, as can be seen in Fig. 7, the results of the STB model are
in a good agreement with the experimental results.
42
40
Skin Temperature (ºC)
38
36
34
32 Experiments (Stolwijk and Hardy, 1966)

Pennes' Model
30
STB Model
28
0 30 60 90 120 150 180 210
Time (min)
Fig. 7. Comparison of measured (Stolwijk & Hardy, 1966) and simulated skin temperature
(Pennes’ model and the STB model) during temperature step change
Despite the simplicity of the STB model, it is able to accurately predict the temperature in
the cutaneous layer under a wide range of personal/environmental conditions. Also,
Zolfaghari and Maerefat (2010) showed that, because of the simplicity and reasonable
accuracy of the STB model, it can be widely used in predicting the thermal response of the
human body under both transient and steady-state environments. Therefore, the STB model
can be utilized for evaluating thermal comfort of the human body in variant
personal/environmental conditions. Furthermore, the model validations show that the
model results are sufficiently reliable under extremely hot/cold conditions (Zolfaghari &
Maerefat, 2010).
5. Conclusion
In this chapter, a brief review of bioheat transfer models (e.g. Pennes’ bioheat equation,
Wulff Continuum Model, Klinger Continuum Model, Chen-Holmes model, Weinbaum-Jiji-
Lemons vascular model and simplified Weinbaum-Jiji vascular Model) has been presented.
Also, a new simplified thermoregulatory bioheat (STB) model (Zolfaghari & Maerefat, 2010)
has been briefly introduced in the present chapter. The STB model has been developed by
combining the well-known Pennes’ equation with Gagge’s two-node model for evaluating
the temperature in the cutaneous layer under a wide range of personal/environmental
conditions. This model considers the effects of thermoregulatory mechanisms of the human
body by defining the thermal control signals of the body. The STB model has been validated
against the published experimental and analytical results, where a good agreement has been
found (Zolfaghari & Maerefat, 2010). Therefore, the results of the STB model are sufficiently
reliable for estimating the cutaneous themperature under both transient and steady-state
thermal conditions.
6. References
Chato, J.C. (1980). Heat Transfer to Blood Vessels, ASME Journal of Biomechanical Engineering,
Vol. 102, pp. 110-118, ISSN 0148-0731
Chen, M.M. & Holmes, K. R. (1980). Microvascular Contributions in Tissue Heat Transfer,
Annals of the New York Academy of Sciences, Vol. 335, pp. 137–150, ISSN 0077-8923
Cho, Y.I. (1992). Bioengineering Heat Transfer, In: Advances in Heat Transfer, J.P. Hartnett &
T.F. Irvine, (Ed.), Academic Press, Inc., ISBN 978-0-12-020022-8, San Diego, USA
Campbell, I. (2008). Body Temperature and its Regulation. Anaesthesia & Intensive Care
Medicine, Vol. 9, No. 6, pp. 259-263, ISSN 1472-0299
Datta, A.K. (2002). Biological and Bioenvironmental Heat and Mass Transfer, Marcel Dekker,
Inc., ISBN 978-0-8247-0775-3, New York, USA
Doherty, T. & Arens, E.A. (1988). Evaluation of the Physiological Bases of Thermal Comfort
Models. ASHRAE Transaction, Vol. 94, pp. 1371-1385, ISSN 0001-2505
DuBois D. & DuBois E.F. (1916) A Formula to Estimate Approximate Surface Area, if Height
and Weight are Known. Archives of Internal Medicine, Vol. 17, pp. 863-871, ISSN
0003-9926
Fanger P.O. (1970). Thermal Comfort Analysis and Applications in Environmental Engineering,
McGraw-Hill, ISBN 0-07-019915-9, New York, USA
Jiji, L.M. (2009). Heat Conduction, Third Edition, Springer, ISBN 978-3-642-01266-2, Berlin,
Germany
Kaynakli O. & Kilic M. (2005). Investigation of indoor thermal comfort under transient
conditions. Building and Environment, Vol. 40, No. 2, pp. 165-174, ISSN 0360-1323
Klinger, H.G. (1974). Heat transfer in perfused biological tissue. I. General theory. Bulletin of
Mathematical Biology, Vol. 36, pp. 403-415, ISSN 1522-9602
Kreith, F. (2000). The CRC Handbook of Thermal Engineering, CRC Press, ISBN 978-0-8493-
9581-9, Boca Raton, USA
Kutz, M. (2009). Biomedical Engineering and Design Handbook, Second Edition, McGraw-Hill,
ISBN 978-0-07-170472-4, New York, USA
Lv Y.G. & Liu J. (2007). Effect of transient temperature on thermoreceptor response and
thermal sensation. Building and Environment, Vol. 42, pp. 656-64, ISSN 0360-1323
Minkowycz, W.J., Sparrow, E.M. & Abraham, J.P. (2009). Advances in Numerical Heat Transfer:
Volume 3, CRC Press, ISBN 978-1-4200-9521-0, Boca Raton, USA
Pennes, H.H. (1948). Analysis of Tissue and Arterial Blood Temperatures in the Resting
Forearm, Journal of Applied Physiology, Vol. 1, pp. 93-122, ISSN 1522-1601
Raaymakers, B.W., Kotte, A.N.T.J. & Lagendijk, J.J.W. (2009). Discrete Vasculature (DIVA)
Model Simulating the Thermal Impact of Individual Blood Vessels for In Vivo Heat
Transfer, In: Advances in Numerical Heat Transfer: Volume 3, Minkowycz, W.J.,
Sparrow, E.M. & Abraham, J.P., pp. 121-148, CRC Press, ISBN 978-1-4200-9521-0,
Boca Raton, USA
Sharma, K.R. (2010). Transport Phenomena in Biomedical Engineering, McGraw-Hill, ISBN 978-
0-07-166398-4, New York, USA
Stolwijk J.A.J. & Hardy J.D. (1966). Temperature regulation in man — A theoretical study.
Pflügers Archiv European Journal of Physiology, Vol. 291, No. 2, pp. 129-62, ISSN 0031-
6768
Vafai, K. (2011). Porous Media: Applications in Biological Systems and Biotechnology, CRC Press,
ISBN 978-1-4200-6541-1, Boca Raton, USA
Wan, X. & Fan, J. (2008). A Transient Thermal Model of the Human Body-Clothing-
Environment System. Journal of Thermal Biology, Vol. 33, pp. 87-97, ISSN 0306-4565
Weinbaum, S., Jiji, L.M. & Lemons, D.E. (1984). Theory and experiment for the effect of
vascular microstructure on surface tissue heat transfer. Part I. Anatomical
foundation and model conceptualization. ASME Journal of Biomechanical
Engineering, Vol. 106, pp. 321-330, ISSN 0148-0731
Weinbaum, S. & Jiji, L.M. (1985). A new simplified bioheat equation for the effect of blood
flow on local average tissue temperature. ASME Journal of Biomechanical
Engineering, Vol. 107, pp. 131-139, ISSN 0148-0731
Wulff, W. (1974). The Energy Conservation Equation for Living Tissues. IEEE Transactions-
Biomedical Engineering, vol. 21, pp. 494-495, ISSN 0018-9294
Yigit, A. (1999). Combining Thermal Comfort Models. ASHRAE Transactions, Vol. 105, pp.
149-158, ISSN 0001-2505
Zolfaghari, A. & Maerefat, M. (2010). A New Simplified Thermoregulatory Bioheat Model
for Evaluating Thermal Response of the Human Body to Transient Environment.
Building and Environment, Vol. 45, No. 10, pp. 2068-2076, ISSN 0360-1323
10
The Manufacture of Microencapsulated

Thermal Energy Storage Compounds
Suitable for Smart Textile
Salaün Fabien1,2
1Univ Lille Nord de France,
2ENSAIT, GEMTEX;
France
1. Introduction
Smart textiles are able to sense electrical, thermal, chemical, magnetic, or other stimuli from
the environment and adapt or respond to them, using functionalities integrated into the
textile structure. As an important consideration in active wear, clothing comfort is closely
related to microclimate temperature and humidity between clothing and skin.
Since the end of the 80’s, functional textiles have been developed to enhance textile
performances according to the consumers’ demand and to include a large range of
properties with a higher added value. One of the possible ways to manufacture functional
or intelligent textile products is the incorporation of microcapsules or the use of
microencapsulation processes for textile finishing. Thermal storage by latent heat was early
recognised as an attractive alternative to sensible heat storage to improve the thermal
performance of clothing during the modifications of environmental temperature conditions.
Early efforts in the development of latent heat storage used organic phase change materials
(PCMs) for this purpose. PCMs are entrapped in a microcapsule of a few micrometers in
diameter to protect them and to prevent their leakage during its liquid phase. In the two
past decades, microencapsulated Phase Change Materials have drawn an increasing interest
to provide enhanced thermal functionalities in a wide variety of applications. When the
encapsulated PCMs is heated above its phase change temperature, it absorbs heat as it goes
from a solid state to a liquid state or during a solid to solid transition. It can be applied to
clothes technology, building insulation, energy storage as well as to coolant liquids. On a
more general basis, it can be used to design a broad variety of thermal transient regimes.
Recent research has investigated the incorporation of organic PCMs directly into the fabric
fibers (Zhang et al. (2006)) or coated on the substrate surface (Choi et al. (2004), Shin et al.
(2005)), creating functional and effective textile elements which can significantly affect
thermal insulation. Moreover, thermal comfort sensation is closely related to microclimate
temperature and humidity. Thus, Fan & Cheng (2005) denoted that a lower moisture
absorption rate is beneficial to thermal comfort. Therefore, the thermal functional
performance of a thermoregulated fabric is not only influenced by the latent heat of PCMs
but also by the design of the textile structure.
2. Classification of heat storage materials to melting temperature range and

textile application
Among the various heat storage technologies available, i.e. sensible heat based on increasing
the temperature without changing the phase of the material, latent heat based on the
transition of a material according to the temperature and thermo-chemical heat (or heat of
reaction) based on the thermophysics of the reactions; thermal heat storage in the form of
latent heat of phase change seems to be particularly attractive in textile fields. A variety of
PCMs are well-known for their thermal characteristics relating to their phase change stage.
These compounds possess the ability to absorb and store large amounts of latent heat during
the heating process and release this energy during the cooling process. Thus, materials being
converted from solid to liquid, from liquid to solid or solid 1 to solid 2 states are suitable to
be used in the manufacture of thermoregulated textiles. The selection of a PCMs formulation
depends typically on the required phase change temperature depending on end use. Indeed,
PCMs should react to changes in temperature of both the body and the outer layer of the
garment when they are incorporated in the textile substrate. Thus, for textile applications,
PCMs with a phase change within the ambient temperature and comfort range of humans
are suitable, i.e. in a temperature range from 15°C to 35°C.
Among the various ways to store energy, the most attractive form is latent heat storage in
phase change material, because of the advantages of high storage capacity in a small volume
and charging/discharging heat from the system at a nearly constant temperature (Abhat,
1983). Thermal storage by latent heat was recognised early as an attractive alternative to
sensible heat storage to improve the thermal performance of clothing during the changes of
environmental temperature conditions. The latent heat associated with a first-order phase
transition provides a mechanism for the thermal energy storage. The phase changes
comprise predominantly solid-liquid transitions for thermal storage applications in textiles.
The use of PCMs is linked to its latent heat of fusion for thermal storage. The latent heat of
fusion of a material is substantially greater than its sensible heat capacity. Stated differently,
the amount of energy that a material absorbs upon melting or releases upon freezing is
much greater than the amount of energy which it absorbs with a weak variation of
temperature. Upon melting or freezing, a PCM absorbs and releases substantially more
energy than a sensible heat storage material which is heated or cooled to the same
temperature range. The storage capacity of PCMs (Q) equals the phase change and the
sensible heat stored at the phase change temperature (1). Thus, the latent storage is always
increased by a significant extent by their sensible storage capacity. Furthermore, during the
complete phase change process, the temperature of the PCMs as well as the surrounding
media remains nearly constant (Feldman et al., 1986).
TTr Tend
Q = m. ∫ C p (T).dT + m.ΔH + m. ∫ C p (T).dT (1)
Tinitial TTr
For a textile application, the following PCMs properties should be required, i.e. a high value
of the heat of fusion and specific heat per unit volume and weight; a melting point in the
application range (between 15°C to 35°C); a high thermal conductivity; a chemical stability
and non-corrosiveness; PCMs should not be hazardous, non-flammable or poisonous; PCMs
should have a reproducible crystallisation without decomposition; PCMs should present a
small supercooling degree and high rate of crystal growth; they should have a small volume
Thermal Energy Storage Compounds Suitable for Smart Textile 173
variation during the phase change process; and PCMs should be sufficiently abundant at a
low cost.
The most common PCMs, with a phase change temperature suitable for textile application,
can be divided into two groups, i.e. organic compounds such as paraffins or linear alkyl
hydrocarbon and non paraffinic materials (hydrocarbon alcohol, hydrocarbon acid,
polyethylene or polytetramethylene glycol, aliphatic polyester…), and inorganic compounds
such as hydrated inorganic salts, eutectics or polyhydric alcohol-water solution (Zhang,
2001).
Fig. 1. Temperature ranges and corresponding melting enthalpy of suitable PCMs for textile
applications
2.1 Organic phase change materials

Organic solid–liquid PCMs include paraffin, alkyl esters and acids, polyethylene glycol and
its derivatives.
2.1.1 Polyhydric alcohols

Polyhydric alcohol, or plastic crystal, are solid-solid phase change materials, and even if
they are not suitable for textile application since their phase change temperature is higher
than the upper end-use limit, they present small volume change, lower undercooling, no
phase separation and no leakage. Amongst them, pentaerythritol (C(CH2OH)4), trimethylol
ethane ((CH3)C(CH2OH)3), neopentyl glycol ((CH3)2C(CH2OH)2), their NH2-substituted
compounds and 2-amino-2-methyl-1,3-propanediol ((CH2CH2OH)2CCH3NH2), 2-amino-2-
hydroxymethyl-1,3-propanediol ((CH2CH2OH)3CNH2), and their binary mixtures have an
endothermic or exothermic effect under their melting point and are cited as potential
candidate for thermal energy storage. Furthermore, the phase change temperatures and heat
of binary mixture of polyhydric alcohol are lower than of the raw material. Thus, Wang et
al. have found that the phase change of neopentylglycol/pentaerythritol and trihydroxy
methylaminomethane were between 30 and 40°C (Wang et al., 2000).
2.1.2 Polyethylene glycols

In the field of organic PCMs, polyethylene glycols (PEGs) (-O-CH2-CH2-)n and their
derivatives, such as poly(tetra methylene glycol) (PTMG), have been widely used as solid-
liquid thermal energy storage materials since they present high transition enthalpy
depending on their molecular weight, congruent melting behaviour, wide range of
transition temperature to obtain suitable melting point, in contrast to inorganic compounds.
Furthermore, their melting points and enthalpies depend on their molecular weight. Thus,
PEGs with a molecular weight comprised between 600 and 1500 g.mol-1 have melting points
in the temperature range between 17°C to 35°C and a melting enthalpy between 120 to
140 J.g-1. The use of PTMG as phase change material in textile field is limited due to a
supercooling phenomena even if the PTMG 3000 show a melting point about 33°C for a
melting enthalpy about 150 J.g-1.
Furthermore, great efforts were made to prepare new PEG solid-solid PCMs with another
compound to keep the material in solid state during the phase change; and thus prevent its
leakage. Nevertheless, these compounds lose their thermal properties after several heating-
cooling cycles.
2.1.3 Linear chain hydrocarbons

Linear hydrocarbons, CnH2n+2, with a melting temperature between 16°C to 38°C, are the
most common PCMs used for textiles. Thus, n-alkanes were preferred due to their high
latent heat, and they are chemically inert, non-toxic, non-corrosive and non-hygroscopic.
Thus, n-eicosane (Sarier & Onder, 2007a), n-octodecane (Pause, 2001; Shim et al., 2001; Choi
et al., 2004; Li & Zhu, 2004; Chung & Cho, 2004; Zhang et al., 2005b, 2006; Sarier & Onder;
2007a, 2007b) and n-hexadecane (Shim et al., 2001; Sarier & Onder, 2007a, 2007b) are the
most PCMs chosen to be applied. In these studies, pure compound was used as PCM
material and therefore the phase change occurs in a narrow temperature change at the
melting temperature of the n-alkane. The use of binary mixture of n-alkanes allows to adjust
the melting point (or phase change temperature) by modifying the composition, that also
leads to a decrease in the overall latent heat of fusion (Kumano et al., 2005). Furthermore,
some of the binary mixture, i.e. n-hexadecane/ n-octadecane (Métivaud, 1999), n-
octadecane/ n-hexadecane (Métivaud, 1999), or n-tetradecane/ n-hexadecane (Bo et al.,
1999) have a solid/liquid or a solid/solid transition which takes place in a relatively narrow
temperature range. The use of the binary mixture of n-hexadecane/ n-eicosane allows to
widen this temperature change and can be suitable for textiles application (Salaün et al.,
2008a, 2010a). Nevertheless, this widening is accompanied by a decrease of the latent heat
which can be caused by the formation of a concentration distribution in the liquid film due
to small scale segregation (Kumano et al., 2005). The binary mixture of n-hexadecane/ n-
eicosane was also studied by Sarier et al., they concluded that to enhance the thermal
capacity of fabrics, it should be better to use a combination of microcapsules containing
different types of PCMs rather than those including a mixture of them (Sarier & Onder,
2007a). They have observed that the mixture was satisfactory to provide a buffering effect
against temperature changes and to regulate the temperature at the target of desired value.
2.1.4 Fatty acids

Fatty acids are friendly PCMs and show several advantages such as a high phase change
enthalpy, congruent melting and freezing behaviour and good chemical and thermal
stability. Even if capric, lauric, myristic, palmitic or stearic acids have good thermal and heat
transfer characteristics, their melting points between to 32°C to 61°C are quite high for
textile energy storage (Feldman et al., 1989; Sari, 2003). Nevertheless, the eutectics of capric-
lauric, capric-palmitic, lauric-myristic, lauric-palmitic, and lauric-stearic binary mixture,
with melting points of 18.0, 21.8, 32.7, 34.2 and 34°C, and heats of melting of 120, 171, 150,
167 and 160 J.g-1, respectively seem to be suitable PCMs for textile applications (Sari, 2005;
Sari & Karaipekli, 2008).
Zeng et al. have analyzed the thermal properties of a binary mixture of palmitic acid and 1-
tetradecanol. The eutectic composition mixture (point at 29°C and melting enthalpy of 143
J.g-1) could be viewed as a new phase change material with large thermal energy storage
capacity (Zeng et al., 2009).
Although, the thermal properties of fatty acid ester were few studied, these compounds
show a solid–liquid phase transition within a narrow temperature range, which should be
promising for thermal energy storage (Li & Ding, 2007a, 2007b; Alkan et al., 2008; Sari et al.,
2009a). Thus, for isopropyl stearate, butyl steareate, and vinyl stearate and, the thermal
characteristics determined by Feldman et al. are as follow : melting point : 14, 19, and 27°C ,
respectively and the melting enthalpy range from 120 to 142 J.g-1 (Feldman et al., 1986;
Feldman et al., 1995).
2.2 Hydrated inorganic salts

Hydrated inorganic salts, e.g. salt with ‘n’ water molecules, are attractive materials with
their relatively high latent heat per unit weight (from 150 to 300 J.g-1), high thermal
conductivity (~0.5 W.M-1.K-1), non flammability property and moderate costs. Nevertheless,
their use is mainly limited by their incongruent melting, supercooling behaviour when there
are incorporated in the textile substrate, which can occur after several phase change cycle. If
LiNO3.3H2O is stable during at least 25 cycles, phase segregation of CaCl2.6H2O or
Na2SO4.10H2O should be observed after only few cycles. Thus, it was suggested to
overcome these problems to use an extra water amount to limit the formation of heavy
anhydrous salt, which results in the reduction of the storage density (Salaün et al., 2010b;
Biswas, 1977). The use of thickening to overcome phase segregation, such as betonite, or
nucleating agent to limit the supercooling phenomena, have also be suggested, but in all
cases the thermal properties of the inorganic PCMs are reduced. Furthermore, when
inorganic PCMs are encapsulated, the process as well as the morphology of the resulting
capsules affects the thermal properties of the materials (Salaün et al., 2010b).
2.3 Conclusion
Many substances have been studied as potential PCMs with regard to their typical range of
melting and melting enthalpy, within the human comfort range of 18°C to 35°C. This kind of
materials absorbs and emits heat while maintaining a nearly constant temperature, and thus
they are very effective since they can store 5 to 14 times more heat per unit volume than
sensible storage materials. Each type of PCMs presents some advantages or disadvantages.
On the one hand, organic PCMs have excellent stability concerning the thermal cycling, but
they are flammable and their melting enthalpy and storage density are relatively low
compared to salt hydrates. On the other hand, the main problem with inorganic PCMs are
their corrosiveness, their cycling stability and the existence of a subcooling phenomena
which be reduced by adding into the materials some nucleating agents.
3. Microencapsulation
For a textile use, PCMs should be surrounded by a protective shell to avoid any liquid
migration during the phase change step and therefore to preserve PCMs in the polymer
shell as long as possible through the heating/cooling cycles. Besides, the encapsulation step
allows to increase the heat transfer area, to decrease the reactivity of the core materials, to
limit the interference with other materials parameters, to enhance the low heat conductivity
and the manipulation of PCMs becomes easier. Microencapsulated PCMs have attracted
more and more attention since the end of the 1980s (Kaska & Chen, 1985; Bryant & Colvin,
1988). It was progressively established as an interesting alternative to the traditional
materials where the imprisoned air was the principal insulating element. The choice of the
polymer for the membrane synthesis of the microcapsules is linked to the considered
application and the required material processes. In the textile field, the polymers used
should have good thermomechanical properties to resist to the thermal and mechanical
requests during of the implementation processes, and during the daily use by the wearer.
The use of microencapsulated phase change material in textile requires a high thermal
stability of these particles that is influenced by the shell and the core materials. A review on
the various techniques available for the microencapsulation of Phase Change Materials is
presented. The methods can be classified as being either physical or chemical based.
3.1 Principle and generalities

Microencapsulation refers to the formation of polymeric particles with a range of diameters
from nanometers to millimeters, and which exhibits several kinds of morphologies, in which
the mains are so called microcapsules and microspheres (Benita, 1996). Microspheres consist
of polymeric network structure in which an active substance is enclosed, whereas in
microcapsules or core/shell structures exhibit a reservoir structure, i.e. the core substance is
surrounded by a polymeric layer. Even if, microencapsulation principle was firstly
introduced by Green and Schleicher in the 1950s to create microcapsules for carbonless
copying paper (Green & Schleicher, 1956 a & b, as cited in Arshady, 1990), textile
applications were mainly developed in labs since the early 1990’s and actually we can see
much more applications in the textile industry for developing new properties and added
values (Nelson, 2002).
Although, there are more than 200 microencapsulation methods described in scientific
literature and patent, most of them include three basic steps, i.e. enclosure of core
component, formation of the microparticles, and hardening of it. These methods are
generally divided or classified in three main groups, which are based on the mechanisms of
microparticle formation, such as mechanical, chemical and physico-chemical processes. The
choice of a method rather than another is based on the cost of processing, the use of organic
solvents for health and environment considerations. Nevertheless, the microencapsulation
processes used in textile industry are generally focused on chemical and physico-chemical
methods, because mechanical processes lead to the formation of particles with a mean
diameter higher than 100 µm. Even if microparticles with mean size distribution in the range
of 20 to 40 µm should be useful for textile application (Nelson, 2001), very small
microparticles (mean diameter ranging from 1 to 10 µm) are preferably incorporated within
textile fibers (Cox, 1998; Bryant, 1999) but according to Colvin (Colvin, 2000) particles with a
mean diameter less than 1 µm are suitable for fiber fabrication; and larger microcapsules
(mean diameter up to 100 µm) should be used into foams or coated formulation (Colvin &
Bryant, 1996; Pushaw, 1997).
Therefore, the most suitable methods include interfacial, in-situ and suspension
polymerization methods for chemical processes and simple or complex phase coacervation
for physico-chemical processes, and spray drying for the mechanical processes (Borreguero
et al., 2011; Hawlader et al., 2000 & 2003; Teixeira et al., 2004). In all cases, the
microencapsulation process includes two main steps, e.g. emulsification step which
determines the size and the size distribution of the microcapsules; and the formation of the
capsules. The emulsification step may be influenced at once by physical parameters such as
apparatus configuration, stirring rate and volume ratio of the two phases, and by
physicochemical properties such as interfacial tension, viscosities, densities and the chemical
compositions of the two phases. The formation of microcapsules is greatly affected by the
surfactant, which influences not only the mean diameter but also the stability of the
dispersion. The surfactants used in the system have two roles, one to reduce the interfacial
tension between oil and aqueous phases allowing formation of smaller microcapsules and
one to prevent coalescence by its adsorption on the oil-water interface and therefore by
forming a layer around the oil droplets. The second step, or the synthesis of a core/shell
particle or other possible morphologies, is mainly governed by the kinetic factors, i.e. the
ability of the monomers, pre-polymer or polymer to react or to cross-link, and
thermodynamic factors, i.e. the minimum total free energy exchange in the system. The
spreading coefficient should be used to predict the engulfment of PCMs within polymer in a
continuous immiscible phase (Torza & Mason, 1970; Sundberg & Sundberg, 1993, Loxley &
Vincent, 1998; Sánchez et al., 2007; Salaün et al., 2008a).
3.2 State of the art in microencapsulated PCM’s

The microencapsulation of phase change materials has been widely investigated since
the late 1970s to provide significantly enhanced thermal management for various
application fields. Thus, preliminary work on microencapsulation of phase change materials
was begun at General Electric under the sponsorship of the U.S Department of Energy. This
was the first such attempt and was directed toward the development of a two-component
heat transfer fluid for use in solar systems for residential applications (Colvin & Mulligan,
1986).
The microencapsulation of PCMs provides some benefit effects on the heat storage, since
the shape of the storage material is arbitrary, temperature gradient is more favourable
and PCMs lifetime increases. Furthermore, the product is always in a dried state since the
phase change occurs within the coating material. The microcapsules’ embedment in heat
storage is relatively simple, and the various PCMs formulation with different melting
temperatures should be used to obtain the desired temperature in which phase change
occurs. The use of microcapsules increase the specific exchange area, which is provided for
effective heat transfer; and thus large quantities of thermal energy can be stored and
released at a relatively constant temperature without significant volume change since in the
particles, the size remains constant and sufficient space or free volume exist. Finally, the
encapsulation step allows to reduce or to avoid problems as such supercooling and phase
separation (Hawlader et al., 2003). Therefore, to maintain the reversibility of the PCMs
functionality, they are encapsulated within the impermeable microcapsules walls for the
whole product life. Furthermore, the microcapsules need to be resistant to mechanical and
thermal stresses.
The PCMs, having a phase change temperature between -10 and 80°C, can be used for
encapsulation (Jahns). Furthermore, according Lane, over 200 phase change materials, with
a melting point from 10 to 90°C, are suitable to be encapsulated (Lane, 1980). Thus, among
all the potential PCMs used to be microencapsulated in a polyester resin, the best results
were obtained by using CaCl2.6H2O. From this study, various encapsulation strategies have
been explored to entrap inorganic compounds, since the organic PCMs tested migrated into
the polymer encapsulant and the thermal cycling stability was poor due to leakage problem
after 100 thermal cycle test runs, link to the thickness of the polymeric shell (Roy &
Sengupta, 1991) and to the mean size (Colvin & Mulligan, 1986). Thus, up to the 1990s, few
reports have been published on the microencapsulation of paraffin. Even if a primary use of
the mPCMs products is for the textile industry, mPCMs are also finding wide spread
applications in others areas such as electronic for cooling electrical component, as building
materials to increase the energy efficiency of building, or as storage solution for packaging
and agriculture to thermally protect products. Thus, since the middle of the 1990s, there has
been a renewed interest in using mPCMs to enhance performance of fluid convective heat
transfer (Yamagishi et al., 1999) or gas fluidised bed (Brown et al., 1998), textile (Bryant &
Colvin, 1988 & 1996; Cox, 1998; Pause, 1994 & 1995), building material and also solar energy
system (Hawlader et al., 2003).
However, at the current state of microencapsulation technology, most of commercial textile
application of PCM choice is restricted to paraffins or organic compounds (Mehling &
Cabeza, 2008); and these compounds are available as slurry or fluid dispersion and dried
powder. Nevertheless, Salaün et al. (2008b & 2010b) have tried to encapsulate hydrated salt.
3.3 Microencapsulation methods

One of the most typical methods which may be used to microencapsulate a PCM is to
disperse droplets of this molten material in an aqueous solution and to form walls around
the droplets. A literature survey on mPCMS indicates that the most suitable methods
include interfacial, in-situ and suspension polymerization methods for chemical processes
and simple or complex phase coacervation for physico-chemical processes and spray-drying
for the mechanical processes.
The choice of the shell materials has been widely investigated by several researchers (Table 1)
to enhance the structural anti-permeability or to bring up a solid protection and to improve
the thermal stability of these particles. The selection of the microencapsulation process is
mainly governed by the choice of the chemical nature of shell materials and by the desired
mean diameter. Thus, the in situ polymerization is suitable for the melamine derivatives, the
interfacial polymerizations were used to obtain polyurea and polyurethane walls, whereas
PMMA and styrenic shells were obtained from emulsion polymerization and suspension-
like polymerization reactions, respectively. The use of hydrosoluble polymers imply to
select a coacervation or a spray-drying method. Furthermore, since a few years, a new
process based on a sol-gel or core templating was used to obtain silica shell. Besides,
microencapsulation of PCMs is essentially restricted to organic phase change materials,
indeed the process technology does not allow to obtain a polymeric shell sufficiently
tightened to prevent the small water molecules diffusion during the phase changes of the
salt hydrate.
Encapsulation Microcapsule Melting Latent heat of
storage
Microcapsule shell Core material Method Reference
ratio diameter temperature melting
wt.% µm °C J.g-1
Polyurea n-Hexadecane 28.0 2.5 15.5 66.1 Interfacial polymerization Zou et al., 2004
Polyurea Butyl stearate 66.0 20–35 28.6 76.3 Interfacial polymerization Liang et al., 2009
Polyurea n-Octadecane 26.0-46.4 0.1-1 29.5-30.8 53.6-110.0 interfacial polymerization Cho et al., 2002
Polyurea n-Octadecane 70.0-75.0 7-16 26.8-28.2 153.2-188.9 interfacial polymerization Zhang & Wang, 2009a
Polyurea n-Eicosane 75.0 2-6 35.7 55.5 Interfacial polymerization Lan et al., 2004
Polyurethane n-Octadecane 40.0-70.0 5-10 ~29.5-30.5 110.0-117.5 Interfacial polymerization Su et al., 2007
Melamine–formaldehyde n-Hexadecane 67.2 2 16.4 160.0 In situ polymerization Salaün et al., 2009
Melamine–formaldehyde n-Octadecane 59.0 2.2 40.6 144.0 In situ polymerization Li et al., 2007
Melamine–formaldehyde n-Octadecane 70.0 9.2 30.5 169.0 In situ polymerization Zhang et al., 2004b
Melamine–formaldehyde n-Octadecane 40.0 1.3 --- 166.0 In situ polymerization Fan et al., 2005
Melamine–formaldehyde n-Octadecane 74.0 2-10 27.0 167.0 In situ polymerization Yamagishi et al., 1999
Melamine-formaldehyde n-Octadecane --- 3-5 28.6 153.7 In situ polymerization Chung & Cho, 2004
Thermal Energy Storage Compounds Suitable for Smart Textile
Melamine-formaldehyde n-Octadecane --- 1-1.5 28.0 141.0 In situ polymerization Choi et al., 2004
Melamine–formaldehyde Lauryl alcohol 75.0 9.2 --- --- In situ polymerization Su et al., 2005
n-Hexadecane/
Melamine–formaldehyde n-Eicosane/ 77.0 1-2 15.1 173.1 In situ polymerization Salaün et al., 2010c
TOES (48/48/4)
Urea–melamine–
n-Octadecane 71.4-78.8 0.2-5.6 33.23-35.15 172.7-190.6 in situ polymerization Zhang et al., 2004a
formaldehyde
Urea–melamine–
n-Octadecane 72.0 0.3-6.4 36.5 167.0 In situ polymerization Zhang et al., 2005a
formaldehyde
Urea–melamine–
n-Nonadecane 69.0 --- 30.3 233.0 In situ polymerization Zhang et al., 2005a
formaldehyde
Urea–melamine–
n-Eicosane 71.0 --- 45.3 172.0 In situ polymerization Zhang et al., 2005a
formaldehyde
Table 1. Shell composition and some properties of microcapsules suitable for thermal heat
179
180
Encapsulation Microcapsule Melting Latent heat of

Microcapsule shell Core material Method Reference
ratio diameter temperature melting
Melamine–formaldehyde/ n-Hexadecane / In situ
76.4 5-10 16.1 174.6 Salaün et al., 2008a
PVA-MDI Na2HPO4.12H2O interfacial polymerization
Melamine–formaldehyde/ n-Eicosane/ In situ
71.7 5-10 35.1 176.0 Salaün et al., 2008c
PVA-MDI Na2HPO4, 12H2O interfacial polymerization
thermal heat storage

Melamine-formaldehyde-
n-Octadecane 75.0 12.3 26.9 146.5 in situ polymerization Zhang & Wang, 2009b
resorcinol
Polymethylmethacrylate n-Docosane 28.0 0.16 41.0 54.6 Emulsion polymerization Alkan et al., 2009
Polymethylmethacrylate n-Eicosane 35.0 0.7 35.2 84.2 Emulsion polymerization Alkan et al., 2011
Polymethylmethacrylate n-Octacosane 43.0 0.25 50.6 86.4 Emulsion polymerization Sari et al., 2009b
Polymethylmethacrylate n-Heptadecane 38.0 0.26 18.2 81.5 Emulsion polymerization Sari et al., 2010
Polymethylmethacrylate Paraffin 61.2 0.5-2 24.0-33.0 101.0 Emulsion polymerization Ma et al., 2010
Styrene-methyl
Suspension-like
methacrylate PRS® paraffin wax 43.1 380 41.8 87.5 Sánchez-Silva et al., 2010
polymerisation reactions
copolymer
Suspension-like
Polystyrene PRS® paraffin wax 20.5 38 35.2 41.6 Sánchez et al., 2007
Suspension-like
Polystyrene n-Nonadecane 49.3 10.5 --- 119.8 Sánchez et al., 2007
Suspension-like
Polystyrene n-Octadecane 53.5 0.100-0.123 --- 124.4 Fang et al., 2008
cross-linked PVA n-Eicosane 70.0-85.0 50-250 43.4 218.7 --- Roy & Sengupta, 1991
Gelatin paraffin wax --- 500 --- 56.0-58.0 Simple Coacervation Uddin et al., 2002
Gelatin-arabic gum paraffin wax 33.3-66.6 50-100 --- 28.4-85.9 Complex Coacervation Hawlader et al., 2000
Gelatin-arabic gum paraffin wax 33.3-66.6 ~1 193.3-239.8 Complex Coacervation Hawlader et al., 2003
Gelatin-arabic gum PCM RT 27 49.0 9 25.2 79.0 Complex Coacervation Bayes-Garcia et al., 2010
Agar-Agar/arabic gum PCM RT 27 48.0 12 26.4 78.0 Complex Coacervation Bayes-Garcia et al., 2010
Table 1. (continues) Shell composition and some properties of microcapsules suitable for
3.3.1 Phase coacervation

Phase coacervation is one of the oldest and most widely used technique of
microencapsulation, and can be divided into two groups, i.e. simple coacervation which
implies the use of one colloidal solute such as gelatin (Uddin et al., 2002), and complex
coacervation, in which the aqueous polymeric solution is prepared from the interaction of
two oppositely charged colloids, such as gelatine-arabic gum (Hawlader et al., 2003; Onder
et al., 2008) or silk fibroin chitosan (Deveci & Bassal, 2009). Coacervation can be defined as
the separation of a macromolecular solution into two immiscible liquid phases, i.e. a dense
coacervate phase and a dilute equilibrium phase. The general outline of this method consists
in three consecutive steps carried out under agitation (Figure 2).
Fig. 2. Schematic representation of the various stages of the microencapsulation process by

the coacervation method
Firstly, the core PCMs material is dispersed in an aqueous solution containing the coating
polymer. Secondly, the deposition of the coating material is promoted by a reduction in the
total free interfacial energy of the system, brought by the decrease of the coating material
surface area during coalescence of the liquid polymer droplets. And thirdly, rigidizing of the
coating material by thermal, cross linking or desolvation techniques is done to obtain solid
particle.
The use of simple coavervation method was conducted by Uddin et al. (2002) to prepare
microencpasulated paraffin. The aim of this study was to investigate the thermal
perfomance of their microcapsules when they are subjected to cyclic operation. Thus, they
reproted that the microencapsulated paraffin has shown an energy storage and release
capacity about 56-58 J.g-1 during its phase change and this even after 1000 thermal cyles.
The complex coacervation method implies the use of two polyelectrolytes of opposite
charges, i.e. gelatine/arabic gum (or acacia), agar-agar/Arabic gum or chitosan/silk fibroin.
Thus, for the couple gelatine/gum Arabic, at pH<4.7, the gelatine is cationic, and gum
Arabic is anionic. For example, as for the simple coacervation phase, the core material is
initially dispersed in the gelatine solution at 40-60°C. To this dispersion, a solution of gum
Arabic is added, and the pH is adjusted to ~4.0. This causes a liquid complex coacervate of
gelatine-arabic gum and water to form, which surrounds the core to form primary capsules
shell. The system is then cooled to room temperature to form the shell. The shell is then
cross-linked by addition of glutaraldehyde or formaldehyde to react with amino groups on
the chains. After adjusting the pH to 9-11 with NaOH solution, the solution is cooled down
to 5-10°C during 2 to 4 hours.
Thus, Onder et al. (2008) have succeeded in the encapsulation of three paraffin waxes,
namely n-hexadecane, n-octadecane and n-nonadecane by complex coacervation based on
gelatin/arabic gum interactions before their incoporation into woven fabrics by coating.
They also found that the absorbed levels of coacervates were proportionnal with those of
PCMs, and tended to increase when the dispersed PCM contents in emulsions were higher.
Thus, the coacervates corresponding to 80 percent of n-hexadecane and n-octadecane
addition, provided significant enthalpy, i.e. 144.7 and 165.8 J.g-1, respectively.
3.3.2 Interfacial polycondensation

The microencapsulation of organic PCMs by interfacial polycondensation or polymerization
involves the dispersion of the liposoluble phase in an aqueous continuous phase to induce the
precipitation of the polymeric materials at the droplet interface. Each phase contains a
dissolved specific monomer suitable to react with the other present in the other phase. The
dispersed phase is as a good solvent for the monomers but acts as a non-solvent for the
produced polymer. Therefore during polymerization, the system is composed of three
mutually immiscible phases. Once, the various monomers were added to the system the
reactions occur at the interface resulting in the formation of insoluble oligomers in the
droplet, with tend to precipitate at the interface to form primary shell around the droplets.
Further crosslinking reactions lead to the formation of growing shell by diffusion of
monomers. When multifunctional monomers are used, a three-dimensional cross-linked
system is obtained. Nevertheless, the thickness of the shell increases slowly due to the
restricted diffusion of the starting monomers inducing layer morphology changes. The
porosity of the capsules increases with the precipitation of the polymer at the interface and
the solvent molecule diffusion through the wall. Furthermore, the main drawback of this
method is the use of organic solvent, e.g. cyclohexane, to solubilize both core material and
reactive liposoluble monomer. Based on this methods, polyurea, polyurethane, polyamide
and polyester microcapsule shells should be prepared. For polyurethane or polyurea
networks prepared from isocyanate and diols or diamine, the reaction of isocyanate groups
with water molecules was promoted during the shell formation process. Furthermore, Cho et
al. (2002) have observed that the formation of a polyurethane shell should be also provided
from the reaction of the isocyanate with the hydroxyl groupments of the surfactant.
Su et al. (2007) have successfully synthetized a series of PU-shell microPCMs containing n-
octadecane, their FTIR results have confirmed that the polyurethane shell was obtained
form the reaction provided between TDI and DETA with SMA as a dispersant. The surface
of the resulted microcapsules was observed as mostly smooth, and the shape was regularly
spherical with a diameter of about 5-10 μm.
For the most reference cited in Table 1, toluene 2,4 diisocyanate was used as liposoluble
monomer in cyclohexane or acetone, and three amino monomer were tested, i.e. ethylene
diamine, tetraethylene diamine and Jeffamine T403 to obtain a polyurea shell. It was also
found that the efficiency of n-alkane encapsulation increased as the core content decreased.
Nevertheless, it is possible to reach narrow size distribution and high loading content, up to
70 wt.%.
3.3.3 In situ polymerization

Until now, melamine-formaldehyde and urea-formaldehyde resins were usually used or
selected as microcapsule wall materials for the PCMs protection and also to improve the
thermal stability of these particles. The use of this prepolymer offers some advantages to
succeed in the synthesis of microcapsules due to their high reactivity, and therefore a short
reaction time and a controlled amount of polymer to form the particles shell with a high
loading content.
Fig. 3. Schematic representation of the various stages of the microencapsulation process by

in situ polymerization
The in situ polymerization microencapsulation process, described in Fig. 3, consisted in the

following steps: preparation of an oil in water emulsion from the dispersion of a PCMs
melted solution in an aqueous continuous phase containing surfactant under high stirring
rate; addition of melamine-formaldehyde pre-polymer for shell formation; induction of
polycondensation by rising temperature to 60-80°C and/or by adjusting the pH of the
emulsion to 4; when the desired size distribution was obtained, the stirring speed was
decreased and polymerization occurred during 1 to 3 hours according to the temperature
choice with continuous agitation; then the solution was cooled to room temperature, and
neutralized, free formaldehyde was removed by adding ammonia. The suspension was then
filtered, and the microcapsules were washed twice with methanol and distilled water.
This microencapsulation methods can start either from monomers, (i.e. melamine and
formaldehyde, urea and formaldehyde or urea, melamine and formaldehyde), or from
commercial prepolymers as such partially methylated trimethylomelamine, hexamethoxy-
methylolmelamine…(Šumiga et al., 2011). In all of these cases, the ability of amino resins for
self condensation around the core material droplet is linked to its surface activity and is an
enrichment of resin molecules within the interface. The reaction scheme for the synthesis
was given in Fig. 4.
Fig. 4. Formation of cross-linked amino resin shell

The formation of the microcapsule shell described in the liitterature shows that it occurs in
three consecutive stages (Salaün et al., 2009b). In the first step, the melamine-formaldehyde
pre-polymers obtain H+ changing into active pre-polymers under aciditic condition, which
results in an increase of the etherification degree due to a reduction in hydroxyl group
concentration. Then, the water solubility of the pre-polymer decreases and the material
separates from the continuous phase as a liquid solution relatively concentrated in active
pre-polymer. During the second step, the amino resins wet the PCMs droplets to create
primary shell around them. The concentration of resin in the boundary layer is enhanced by
the hydrophilic/hydrophobic interactions of the partial methylolated melamine, and
therefore by an enrichment of resin molecules within the interface. Thus, the resin
condensation proceeds much faster in the boundary layer than in the volume phase,
allowing the formation of tougher capsule walls. From this step, it is possible to recover
particle under powder form, nevertheless the capsule walls are not yet thermo-mechanically
stable and not completely harden, and the particles should collapse during drying. In the
last step the reaction rate of the polycondensation at the interface is increased with the result
of microcapsule shell formation. During the polycondensation process, the particle surface
becomes more and more hydrophobic, and according to the pH level, the shell is crosslinked
either by ether linkage or methylene as shown in Fig. 4. Furthermore, the surface morphology
depends mainly on the pH value, thus, a low pH value promotes the formation of
methylene bridges affecting the amino resin solubility and therefore the obtention of a
rough surface. The formation of higher oligomers of melamine-formaldehyde polymer with
the formation of methylene bridges can be described as the reaction of bound CH2OH group
with H.
The use of an in situ method to prepare PCMs microcapsules results in a high loading
content (from 60 to 80 wt.%) and a narrow size distribution (Table 1.); nevertheless the
weight ratio of the core/shell materials affected also the particle size, its distribution and the
surface morphology of the microcapsules (Salaün et al., 2009b). These particles have better
thermal and mechanical stability properties than those synthesized from other polymers
(Song et al., 2007). To improve their toughness, resorcinol can be added to the system (Zang
& Wang, 2009b). However, the presence of free-formaldehyde after forming the shell may
cause environmental and health problem; and it is relatively difficult to eliminate all
formaldehyde which can be continuously released from the resultant microcapsules. To
solve this problem, Li et al. (2007) have proposed to incorporate formaldehyde once and
urea or melamine for three times; on the other hand, Šumiga et al. (2011) have observed that
the use of ammonia during the post-treatment process allowed to decrease formaldehyde
content level in the suspension.
3.3.4 Other microencapsulation methods

For several years, research on microencapsulation process has been focussed on new
processes, either to obtain environmentally friendly products or to develop an easier,
cheaper and robuster method, such as suspension-like polymerization (Sánchez et al., 2007).
Thus, nanocapsules should be otained from an emulsion copolymerization of methyl
methacrylate and glycidyl methacrylate (Alay et al., 2010) or from emulsion polymerization
of methyl methacrylate (Sari et al., 2010). Even if, the entrapped PCMs have shown good
thermal reliability and chemical stability after 5000 melting and freezing cycling, the PCMs
loading content was found relatively poor (Table 1.). Sánchez et al., (2007) have suggested a
new microencapsulation method based on a free radical polymerization of styrene to entrap
PCMs, where the obtained particles contain almost 50 wt.% of PCMs. Furthermore, to
improve the encapsulation efficiency and the storage capacity of the microcapsules, they
also suggested to copolymerize styrene with methyl methacrylate (Sánchez et al., 2010).
3.4 New development in microencapsulation process to enhance thermal properties

One of the most limitations of these processes implying polymeric shell is the low heat
conductivity and thermal stability of the obtained microcapsules. Although various
methods have been explored for increasing the thermal conductivity of PCMs (Mills et al.,
2006), few of them should be transferred to the microencapsulation method. Thus, a
literature survey shows that there mainly are two ways to enhance it, either by using metal
filler, carbon nanofiller or fiber in the PCMs formulation or synthesize new shell with
enhance thermal properties.
On the one hand, Song et al., (2007) have improved the thermal stability of their mPCMs by
adding 3 wt.% of silver nanoparticles in the core formulation. Besides, it was observed by
Borreguero et al. (2011) that their microcapsules obtained by spray drying and containing
Rubitherm®RT27 with 2 wt.% of carbon nanofibers have higher thermal conductivity,
which promote a faster heat absorption while maintaining their thermal energy storage
capacity.
One the other hand, Salaün et al. (2008c) have observed that the addition of a reasonable
fraction of polymeric nanoparticles containing hydrated salt with low thermal conductivity
in the PCMs microcapsules results in a decrease by a factor of three of the effective thermal
conductivity. They concluded that this outcome was a significant improvement of the
thermal barrier in the steady state. In addition, organic-inorganic hybrids have attracted
more and more attention in order to combine distinct properties of these two kinds of
materials, and specially since the thermal conductivity of inorganic materials is higher than
that of organic materials. Thus, Zhang et al. (2010) have chosen a silica shell to synthesize
microcapsules with a size distribution between 7 to 16 µm. They observed that the thermal
conductivity of their particles increases up to 0.6212 W.m-1.K-1 for a n-Octadecane loading
content of 52.2 wt.%, which enhanced their heat transfer performance.
4. Textiles containing microencapsulated thermal energy storage compounds

4.1 Historical background
Since 1983 at Triangle Research and Development Corporation (TRDC), Colvin and co-
workers have begun investigations to demonstrate the possibility to microencapsulate
paraffinic PCM. After having demonstrated the possibility to manufacture textiles
containing mPCMS for the Air Force (Bryant & Colvin, 1988), TRDC participated, from 1987,
to develop the encapsulation of paraffin for NASA for use in astronaut gloves during
extravehicular activities. The patent issued to Triangle R&D in 1988 on the incorporation of
mPCMs within a fiber concerned polyester, nylon, acrylics and modacrylics fibers. Then,
Bryant & Colvin demonstrated the feasibility of incorporating PCMs, i.e. n-eicosane, 2,2-
dimethyl-1,3-propanediol and 2-hydroxymethyl-2-methyl-1,3-propanediol , within textiles
fibers, patented in 1994 (Bryant & Colvin, 1994). In 1991, Gateway Technologies (Outlast
Technologies from 1997) acquired the exclusive patent rights for incorporating phase-
change technology in commercial fibers and fabrics from TRDC. Since the 90s, Outlast
Technologies have developed and patented several processes to apply microcapsules into
textiles structures, to enhance wearer comfort and to provide temperature control for
consumers in bedding, medical supplies, sportwear, protective clothing…Furthermore, the
microencapsulation of phase change materials and the manufacture of smart textile
containing these microcapsules has drawn an increasing interest of researcher from
universities in the two past decades, as shown by the amount of published studies on the
manufacture and the thermal comfort characterization of these materials (Bendrowska,
2006).
4.2 Performance in clothing

When the mPCMs are applied into textiles substrate, they can absorb heat energy, from the
body or from the environment, to go from a solid state to a liquid state, and then create a
temporary cooling effect in the clothing layer, until the completed melting of the core
material. The reversibility of this effect can be obtained when the smart textile is worn in an
environement, where the temperature is below the crystallization point of the PCMs
formulation. Thus, the temperature of the smart fabrics falls down below the crystallization
point, and when the liquid PCMs change back to a solid state, the heat energy is released to
the surrounding medium, which leads to the wearer a temporary warming effect. From
these two effects, a thermo-regulating effect should be reach to keep the temperature of the
surrounding medium nearly constant. And therefore, this kind of textiles can be reacted as
an active thermal barrier effect to regulate the heat flux through the structure and to adapt
the heat flux to thermal needs according to the activity level or to the ambient temperature
changes (Pause, 2001).
Since the thermal comfort sensation refers to the state of mind that expresses satisfaction
with the thermal environment, this feeling is influenced by various parameters, in which the
garment design plays an important role. Thus, the use of mPCMs in clothing should act in
addition to the passive thermal insulation effect of the garment system. To design specific
clothing layer, the mPCMs quantity should be adjusted in regards to the activity level and
the duration of the garment worn, bearing in mind to maintain sufficient breathability,
flexibility and mechanical stability of the structure. Thus, in order to obtain the desired
effect, the manufacturer should pay attention to select the appropriate PCMs formulation, to
determine the sufficient quantity of the PCMs, to choose the appropriate fibrous susbtrate
and to design the product (Bendrowska, 2006). On the other hand, the magnitude and
duration of thermal effect are dependent upon design textile factors.
4.3 Manufacture of thermoregulated textile structures

The step of encapsulation allows to manufacture textile containing microcapsules by various
ways to fix the microcapsules within the fiber structure permanently, to embed them into a
binder or to mix them into foam. To produce these textiles, mPCMs should respect these
criteria: uniformity of particle size; a core to shell ratio with a PCMs content as high as
possible; stability to mechanical action and good thermal and chemical properties.
4.3.1 Lamination
The method to manufacture foam containing mPCMs (between 20 to 60 wt.%) was first
described by Colvin & Bryant in 1996. This foam is very useful in the heat-insulating
application, since it may be used as a lining for gloves, shoes, or outerwear, when it has
been integrated to the textiles complex by glue, fusion or lamination. The advantages of
this technology are the use of a greater amount of mPCMs, and also the possibility to
incorporate various PCMs formulation to cover a broad range of regulation temperature.
Furthermore, the passive insulation should be increased with the formation of honeycomb
structure which can entrap considerable amount of air (Sarier & Onder, 2007b).
Nevertheless, the interactions between the chemical structure of the microcapsules and
the polymeric material used to form the foam may lead to a defoaming process at high
loading content as observed by You et al., (2008). Thus, to increase the loading content (up
to 25.2 wt.%) and therefore the enthalpy heat (28 J.g-1), they proposed to synthetize
styrene-divinylbenzene (DVB) co-polymer as shell material rather than melamine-
formaldehyde (You et al., 2010).
4.3.2 Coating
For a coating process, the wetted mPCMs are dispersed throughout a polymeric binder, a
surfactant, a dispersant, an antifoam agent and a thickener, as preconized by Zuckerman
et al. (1997). All the common coating processes, such as knife over roll (Pause, 1995),
screen printing (Lottenbach & Sutter, 2002; Choi et al., 2004), pad-dry-cure (Shin et al.,
2005), knife over air, and gravure and dip coating (Bryant & Colvin, 1996; Zuckerman et
al., 2001) should be used. Besides, even if the method for manufacturing coating
composition was widely described in the patent literature, nevertheless few papers
published in the literature give account of the formulation of coating, finishing of fabrics
and therefore the evaluation of their characteristics more specially thermal and durability
properties. Furthermore, the use of polymeric binder has some drawnbacks, since the
incorporated amount should be enough to obtain permanent linkage which can alter the
fabric properties such as drape, air permeability, breathability, thermal resistance,
softness and tensile strength can be affected adversely as the percentage of binder add-on
increases. Salaün et al. (2009a) have proposed a thermodynamical method based on the
comparison of the surface energy component to select the most suitable polymeric binder
for melamine-formaldehyde microcapsules. They also determined that the polyurethane
binder plays a main role during the thirty seconds of a cold to warm transition allowing
to delay the temperature increase. Furthermore, the thermoregulating response depends
on the surface deposited weight and the mass ratio binder to microcapsules. Thus, an
interesting cooling effect is found for 20 g.m-² of binder and from 40 g.m-² of mPCMs. And
a mass ratio binder to mPCMs taken between 1:2 and 1:4 is suitable to manufacture
thermoregulating textile (Salaün et al., 2010c). Choi et al. (2004) have coated a polyester
fabric by screen printing and knife over role, and they have denoted that the first process
allowed to obtain less stiff and hard fabric, and higher air permeability and lower
moisture absorption than the other one. Chung et al. (2004) have succeeded in the
manufacture of vapour-permeable, water repellent and thermoregulated fabrics, in
coating melamine-formaldehyde microcapsules mixed in to polyurethane binder. The
amount of heat was kept constant (~14 J.g-1) after 30 launderings, and even if the water
repellence property was the same with and without mPCMs, they have also observed a
decrease for the water resistance with the presence of mPCMs. On the other hand,
Sánchez et al. (2010) have manufactured a cotton coating fabric (with an enthalpy heat
about 7.6 J.g-1) with 35% of polystyrene microcapsules related to the coating formulation.
They observed a thermoregulatory effect of 8.8, 6.3, 5.6 and 2.5°C after 6, 12, 44 and 75 s.
Furthermore, their results after rubbing and ironing tests were higher than those of Kim
& Cho (2002).
4.3.3 Spinning
The incorporation of mPCMs within fibers was first patented in 1988 (Bryant & Colvin,
1988). This kind of manufacture presents some advantages as a permanent incoporation and
no need to modify subsequent processing and after treatment. Thus, the process to obtain
artificial or synthetic fibers containing mPCMs can be achieved either by wet spinning for
e.g. polyacrylonitrile (Lennox, 1998), polyacrylonitrile-vinylidene chloride (Zhang et al.,
2005b), acrylic fiber (Cox, 2001), or by melt spinning (Gao et al., 2009) for polypropylene
(Leskovšek et al., 2004) or polyethylene fiber (Zhang et al., 2005b). Although, the touch,
drape, softness and color were not modified by the spinning process, the thermal heat
capacity of the obtained fiber was limited to a low microPCMs loading content. Up to now,
the main envisaged way to incorporate microcapsules directly into fibers was solvent
spinning rather than melt-spinning because of their tendency to aggregate themselves under
dried powder form, and also sometimes because of the high temperatures involved in melt-
spinning which could degrade the microcapsules. Thus, since 1997, polyacrylonitrile fibers
containing 7 wt.% of microcapsules were put onto market, evenf if their storage capacity is
relatively low (about 8-12 J.g-1). Since, Leskovšek et al. (2004) have shown the possibility of
introducing microcapsules into polypropylene fibers in the presence of a lubricant. In their
study, the added amount of microcapsules was relatively small (1%, 2% and 5%) and the PP
was in the form of a single monofilament yarn with a large diameter. They denoted that the
microcapsules did not impair the mechanical properties of the fibres. More recently, in 2005,
Zhang and co-workers have succeeded in increasing the loading mPCMs content up to 24%
in PP fibers, and they denoted that these fibers exhibited acceptable mechanical properties
as the content in mPCMs does not exceed 20 wt.%, for an enthalpy about 11 J.g-1. Wet-spun
thermoregulated polyacrylonitrile-vinylidene chloride fibers containing 30 wt.% of mPCMs
were also obtained by Zhang et al.(2006). Gao et al. (2009) have shown the possibility to
incorporate the mPCMs during the polymerization step of Acrylonitrile–methyl acrylate
copolymer before the melt-spinning process. Thus, they obtained thermoregulated fibers
containing 20 and 25 wt.% of mPCMs for a melting enthalpy about 21 to 25 J.g-1.
5. Conclusion
One of the advantages of microencapsulated PCMs is their use for the fabrication of any
kind of active smart textiles, meaning which sense and react to the conditions or simulus
and thus this technology is attractive to improve clothing comfort and this for a wide and
diversified market. MPCMs were firstly developed for use in space suits and glove to
protect astronauts from extreme temperature fluctuation during activities in space, specialy
as a glove liner. Actually, mPCMs are widely used to improve the thermal performance of
active wear garment (outwear, sportwear, underwear), shoes linning, accessories, bedding
products (comforters, pillows, or mattress covers), and in various activity fields such as
building, medical, to enhance the thermo-physical comfort of surgical clothing such as
gownd, caps and gloves; or as a product which support the effort to keep the patient warm
during an operation by providing insulation tailored to the body’s temperature; in
automotive sector to stabilise the interior temperature or in the seat to reduce unpleasant
temperature changes and to keep the skin temperatures more constant. In fact, more than
150 textile companies or brand use mPCMs licensed by Outlast technology.
On the other hand, a new kind of microcapsules, interacting with the absorbed water to
promote a cooling effect, has been recently developed to enhance mass transfer. This
technology is based on the thermo-chemical heat storage, which present numerous
advantages, i.e. high energy density, high discharge power due to high reaction enthalpy,
wide reaction temperature range, high heat and mass transfer rates, fast reaction kinetic, low
material prices, non toxic material. Thus, the combination of a porous shell with moisture-
sensitive compound as xylitol would be useful for a material design of new functional
microparticles for thermal and moisture management (Salaün et al., 2011).
6. Acknowledgment
I would like to thank the French Institute of Textiles and Clothing (IFTH, 2 rue de la
Recherche, 59650 Villeneuve d’Ascq, France) and Damartex (2, avenue de la Fosse-aux-
Chêne, 59100 Roubaix, France) for funding these researches.
7. References
Abhat, A. (1983). Low temperature latent heat thermal energy storage: heat storage
materials. Solar energy, Vol.30, No.4, (1983), pp. 313-332, ISSN 0038-092X
Alay, S., Göde, F. & Alkan, C. (2010). Preparation and Caharcterization of
Poly(methylmethacrylate-co-glycidyl methacrylate)/n-hexadecane Nanocapsules
as a fiber additive for Thermal energy Storage. Fibres and Polymers, Vol.11, No.8,
(December 2010), pp.1089-1093, ISSN 1229-9197
Alkan, C., Kaya, K. & Sari, A. (2008). Preparation and thermal properties of ethylene glycole
distearate as a novel phase change material for energy storage. Material Letters,
Vol.62, No.6-7, (March 2008), pp. 1122-1125, ISSN 0167-577X
Alkan, C., Sari, A., Karaipekli, A. & Uzun, O. (2009). Preparation, characterization, and
thermal properties of microencapsulated phase change material for thermal energy
storage. Solar Energy Materials and Solar Cells, Vol.93, No.1, (January 2009), pp. 143-
147, ISSN 0927-0248
Alkan, C., Sari, A. & Karaipekli, A. (2011). Preparation, thermal properties and thermal
reliability of microencapsulated n-eicosane as novel phase change material for
thermal energy storage. Energy Conversion and Management, Vol.52, No.1, (January
2011), pp. 687-692, ISSN 0196-8904
Arshady, R. (1990). Microspheres and Microcapsules, a Survey of Manufacturing
Techniques Part II: Coacervation. Polymer engineering and Science, Vol.30, No.15,
(Mid-August 1990), pp. 905-914, ISSN 0032-3888
Bayes-Garcia, L., Ventola, L., Cordobilla, R., Benages, R., Calvet, T.& Cuevas-Diarte,
M.A. (2010). Phase Change Materials (PCM) microcapsules with different shell
compositions: Preparation, characterization and thermal stability. Solar Energy
Materials and Solar Cells, Vol.94, No.7, (July 2010), pp. 1235-1240, ISSN 0927-
0248
Bendrowska, W. (2006). Intelligent textiles with PCMs, In: Intelligent textiles and clothing,
Mattila, pp.34-62, Woodhead Publishing in Textiles, ISBN 978-1-84569-005-2,
Cambridge, England
Benita, S. (1996). Microencapsulation: methods and industrial applications. Marcel Dekker Inc.,
ISBN 978-0-8247-2317-0, New York
Biswas, D.R. (1977). Thermal energy storage using sodium sulphate decahydrate and water.
Solar Energy, Vol.10, No.1, (January 1977), pp.99-100, ISSN 0038-092X
Bo, H., Gustafsson, E.M. & Setterwall F. (1999). Tetradecane and hexadecane binary
mixtures as phase change materials (PCMs) for cool storage in district cooling
systems. Energy, Vol.24, No. 12, (January 1999), pp. 1015-1028, ISSN 0360-5442
Borreguero, A.M., Valverde, J.L., Rodriguez, J.F., Barber, A.H., Cubillo, J.J. & Carmona, M.
(2011). Synthesis and characterization of microcapsules containing
Rubitherm®RT27 obtained by spray drying. Chemical Engineering Journal, Vol.166,
No.1, (January 2011), pp. 384-390, ISSN 1385-8947
Brown, R.C., Rasberry, J.D. & Overmann, S.P. (1998). Microencapsulated phase-change
materials as heat transfer media in gas-fluidized beds. Powder Technology, Vol.98,
No.3, (August 1998), pp. 217-222, ISSN 0032-5910
Bryant, Y.G. & Colvin, D.P. (1988). Fibre and Reversible Enhanced Thermal Storage
Properties and Fabric Made There From. US Patent 4,756,958, available from
http://patft.uspto.gov/
Bryant, Y.G. & Colvin, D.P. (1994). Fabric with reversible enhanced thermal properties. US
Patent 5,366,801, available from http://patft.uspto.gov/
Bryant, Y.G. & Colvin, D.P. (1996). Moldable foam insole with reversible enhanced thermal
storage properties. US Patent 5,499,460, available from http://patft.uspto.gov/
Bryant, Y.G. (1999). Melt spun fibres containing microencapsulated phase change material.
IN: Advances in Heat and Mass Transfer in Biotechnology; HTD- Vol.363/BED-VOl.44,
pp.225-234, ISBN 0791816435
Choi, K., Cho, G., Kim, P. & Cho, C. (2004). Thermal Storage/Release and Mechanical
Properties of Phase Change Materials on Polyester Fabrics. Textile Research Journal,
Vol.74, No.4, (April 2004), pp.292-296, ISSN 0040-5175
Cho, J.S., Kwon, A. & Cho, C.G. (2002). Microencapsulation of octadecane as a phase-change
material by interfacial polymerization in an emulsion system. Colloid Polymer
Science, Vol. 280, No.3, (March 2002), pp. 260-266, ISSN 0303-402X
Chung, H. & Cho, G. (2004). Thermal Properties and Physiological Responses of Vapor-
Permeable Water-Repellent Fabrics Treated with Microcapsule-containing PCMs.
Textile Research Journal, Vol.74, No.7, (July 2004), pp.571-575, ISSN 0040-5175
Colvin, D.P. (2000). Encapsulated phase change materials, The 2nd International Conference on
Safety & Protective Fabrics, , Winston-Salem, North Carolina, 28-30 April 2000
Colvin, D.P. & Bryant, Y.G. (1996). Thermally enhanced foam insulation. US Patent 5, 637,
389, available from http://patft.uspto.gov/
Colvin, D.P. & Bryant, Y.G. (1998). Protective clothing containing encapsulated phase
change materials. ASME: Advances in Heat and Mass Transfer, HTD-vol. 362/BED-
vol. 40, pp. 123-32, ISSN 0272-5673
Colvin, D.P. & Mulligan, J.C. (1986). Spacecraft Heat Rejection Methods : Active and Passive
Heat Transfer for Electronic Systems – Phase 1. Final Report for Period September 1985
– July 1986, AFWAL-TR-86-3074.
Cox, R. (1998). Synopsis of The new thermal regulation fiber Outlast. Chemical Fibers
International, Vol.48, No.6, (December 1998), pp. 475-479, ISSN 0340-3343
Cox, R. (2001). Repositionning acrylic fibres for the new millenium. Chemical Fibers
International, Vol.51, No.2, (May 2001), pp.118-120, ISSN 0340-3343
Deveci, S.S. & Basal, G. (2009). Preparation of PCM Microcapsules by Complex Coacervation
of Silk Fibroin and Chitosan. Colloid Polymer Science, Vol. 287, No.12, (December
2009), pp.1455-1467, ISSN 0303-402X
Fan, J. & Cheng, X.Y. (2005). Heat and Moisture Transfer with Sorption and Phase Change
Through Clothing Assemblies, Part II: Theoretical Modeling, Simulation, and
Comparison with Experimental Results. Textile Research Journal, Vol. 75, No.3,
(March 2005), pp.187-196, ISSN 0040-5175
Fan, Y.F., Zhang, X.X., Wu, S.Z. & Wang, X.C. (2005). Thermal stability and permeability of
microencapsulated n-octadecane and cyclohexane. Thermochimica Acta, Vol.429,
No.1, (May 2005), pp.25-29, ISSN 0040-6031
Fang, Y., Kuang, S., Gao, X. & Zhang, Z. (2008). Preparation and characterization of novel
nanoencapsulated phase change materials. Energy Conversion and Management,
Vol.49, No.12, (December 2008), pp. 3704-3707, ISSN 0196-8904
Feldman, D., Shapiro, M.M. & Banu, D. (1986). Organic phase change materials for thermal
energy storage. Solar Energy Materials, Vol.13, No.1, (January 1986), pp. 1-10, ISSN
0165-1633
Feldman, D., Shapiro, M.M., Banu, D. & Fuks, C.J. (1989). Fatty acids and their mixtures as
phase-change materials for thermal energy storage. Solar Energy Materials, Vol.18,
No.3-4, (March 1989), pp. 201-216, ISSN 0165-1633
Feldman, D., Banu, D. & Hawes, D. (1995). Low chain esters of stearic acid as phase change
materials for thermal energy storage in buildings. Solar Energy Materials and Solar
Cells, Vol.36, No.3, (March 1995), pp. 311-322, ISSN 0927-0248
Gao, X.Y., Han, N., Zhang, X.X. & Yu, W.Y. (2009). Melt-processable acrylonitrile–methyl
acrylate copolymers and melt-spun fibers containing MicroPCMs. Journal of
Materials Science, Vol.44, No.21, (November 2009), pp. 5877-5884, ISSN 1573-4803
Hawlader, M.N.A., Uddin, M.S. & Zhu, H.J. (2000). Preparation and evaluation of a novel
solar storage material: Microencapsulated paraffin. International Journal of
Sustainable Energy, Vol.20, No. 4, (September 2000), pp. 227-238, ISSN 0142-5919
Hawlader, M.N.A., Uddin M.S., & Khin, M.M. (2003). Microencapsulated PCM thermal-
energy storage system. Applied Energy, Vol.74, No.1-2, (January-February 2003),
pp. 195-202, ISSN 0306-2619
Jahns, E. (n.d.). Microencapsulated phase change material, available from:
http://www.fskab.com/annex10/WS_pres/Jahns.pdf
Jin, Y., Lee, W., Musina, Z. & Ding, Y. (2010). A one-step method for producing
microencapsulated phase change materials. Particuology, Vol.8, No.6, (December
2010), pp. 588-590, ISSN 1674-2001
Kaska, K.E. & Chen, M.M. (1985). Improvement of the Performance of Solar Energy of Waste
Heat Utilization Systems using Phase Change Slurry as Enhanced Heat Transfer
Storage Fluid. Journal of Solar Energy Engineering, Vol.107, No.3, (August 1985), pp.
229-236, ISSN 0199-6231
Kim, J. & Cho, G. (2002).Thermal storage/release, durability, and temperature sensing
properties of thermostatic fabrics treated with octadecane-containing
microcapsules. Textile Research Journal, Vol.72, No.12, (December 2002), pp. 1093-
1098, ISSN 0040-5175
Kumano, H., Saito, A., Okawa, S., Takeda, K. & Okuda, A. (2005). Study of direct contact
melting with hydrocarbon mixtures as the PCM. International Journal of Heat and
Mass Transfer, Vol.48, No.15, (July 2005), pp. 3212-3220, ISSN 0017-9310
Lan, X., Tan, Z., Zou, G., Sun L. & Zhang T. (2004). Microencapsulation of n-eicosane as
energy storage material. Chinese Journal of Chemistry, Vol.22, No.5, (May 2004), pp.
411–414, ISSN 1614-7065
Lane, G.A. (1980). Low temperature heat storage with phase change materials. International
Journal of Ambient Energy, Vol.1, No.3, (July 1980), pp. 155–168, ISSN 0143-0750
Lennox, K.P. (1998). Outlast Technologies adapts Space-age technology to keep us
comfortable, Technical Textile International, Vol.7, No.7, (July-August 1998), pp. 25-
26, ISSN 0964-5993
Leskovšek, M., Jedrinovic, G. & Stankovic-Elseni, U. (2004). Properties of propylene fibres
with incorporated microcapsules, Acta Chimica Slovenica, Vol.51, No.4, (December
2004), pp.699-715, ISSN 1318-0207
Li, W.D. & Ding, E.Y. (2007a). Preparation and characterization of a novel solid–liquid PCM:
Butanediol di-stearate. Material Letters, Vol.61, No.7, (March 2007), pp. 1526–1528,
ISSN 0167-577X
Li, W.D. & Ding, E.Y. (2007b). Preparation and characterization of a series of diol di-
stearates as phase change heat storage materials. Material Letters, Vol. 61, No.21,
(August 2007), pp. 4325–4328, ISSN 0167-577X
Li, W., Zhang, X.X., Wang, X.C. & Niu, J.J. (2007). Preparation and characterization of
microencapsulated phase change material with low remnant formaldehyde content.
Materials Chemistry and Physics, Vol.106, No.2-3, (December 2007), pp.437-442, ISSN
0254-0584
Li, Y. & Zhu, Q. (2004). A Model of Heat and Moisture Transfer in Porous Textiles with
Phase Change Materials. Textile Research Journal, Vol.74, No.5, (May 2004), pp. 447-
457, ISSN 0040-5175
Liang, C., Lingling, X., Hongbo, S. & Zhibin, Z. (2009). Microencapsulation of butyl stearate
as a phase change material by interfacial polycondensation in a polyurea system.
Energy Conversion and Management, Vol.50, No.3, (March 2009), pp. 723-729, ISSN
0196-8904
Lottenbach, R. and Sutter, S. (2002). Method for producing temperature-regulating surfaces
with phase change material. WO Patent 02095314
Loxley, A. & Vincent, B. (1998). Preparation of Poly(methylmethacrylate) Microcapsules
with Liquid Cores. Journal of Colloid and Interface Science, Vol.208, No.1, (December
1998), pp. 49-62, ISSN 0021-9797
Ma, S., Song, G., Li, W., Fan, P. & Tang, G. (2010). UV irradiation-initiated MMA
polymerization to prepare microcapsules containing phase change paraffin. Solar
Energy Materials and Solar Cells, Vol.94, No.10, (October 2010), pp. 1643-1647, ISSN
0927-0248
Mehling, H. & Cabeza, L.F. (2008). Heat and cold storage with PCM – An up date to introduction
into basics and applications, Springer, ISBN 978-3-540-68556-2
Métivaud, V. (1999). Systemes multicomposants d'alcanes normaux dans la gamme
C14H30–C25H52: aliances structurales et stabilite des echantilons mixtes.
Applications pour la protection thermique d'instalations de telecomunications et de

circuits optoelectroniques, European thesis, Université Bordeaux I, France.
Mills, A., Farid, M., Selman, J.R. & Al-Hallaj, S. (2006). Thermal conductivity enhancement
of phase change materials using a graphite matrix, Applied Thermal Engineering,
Vol.26, No.14-15, (October 2006), pp. 1652-1661, ISSN 1359-4311
Nelson, G. (2001). Microencapsulation in textile finishing. Review of Progress in Coloration
and Related Topics, Vol.31, No.1, (June 2001), pp. 57-64, ISSN 1478-4408
Nelson, G. (2002). Application of microencapsulation in textiles. International Journal of
Pharmaceutics, Vol.242, No.1-2, (August 2002), pp.55-62, ISSN 0378-5173
Onder, E., Nihal, S. & Cimen E. (2008). Encapsulation of phase change material by complex
coacervation to improve thermal performances of woven fabrics. Thermochimica
Acta, Vol.467, No.1-2, (January 2008), pp. 63-72, ISSN 0040-6031
Pause, B.H. (1994). Investigation of the heat insulation of protective textiles with
microencapsulated PCM. Techtextil-Symposium, 245, pp. 1-9
Pause, B.H. (1995). Development of heat and cold membrae structures with phase change
material. Journal of Coated Fabrics, Vol.25, No.7, (July 1995), pp.59-68, ISSN 0093-
4658
Pause, B.H. (2001). Interactive thermal insulating system having a layer treated with a
coating of energy absorbing phase change material adjacent a layer of fibers
containing energy absorbing phase change material. US Patent 6,217, 993, available
from http://patft.uspto.gov/
Pushaw, R.J. (1997). Coated skived foam and fabric article containing energy absorbing
phase change material. US Patent 5, 677, 048, available from
http://patft.uspto.gov/
Roy, S.K. & Sengupta, S. (1991). An evaluation of phase change microcapsles for use in
enhanced heat transfer fluids. International Communications in Heat and Mass
Transfer, Vol.18, No.4, (July-August 1991), pp. 495-507, ISSN 0735-1933
Salaün, F., Devaux, E., Bourbigot & S., Rumeau, P. (2008a). Preparation of multinuclear
microparticles using a polymerization in emulsion process. Journal of Applied
Polymer Science, Vol.107, No.4, (February 2008), pp. 2444-2452, ISSN 0021-8995
Salaün, F., Devaux, E., Bourbigot, S. & Rumeau, P. (2008b). Development of a precipitation
method intended for the entrapment of hydrated salt. Carbohydrate Polymers, Vol.73,
No.2, (July 2008), pp. 231-240, ISSN 0144-8617
Salaün, F., Devaux, E., Bourbigot, S., Rumeau, P., Chapuis, P.O, Saha, S.K., & Volz, S.
(2008c). Polymer nanoparticles to decrease thermal conductivity of phase change
materials. Thermochimica Acta, Vol.477, No.1-2, (October 2008), pp. 25-31, ISSN 0040-
6031
Salaün, F., Devaux, E., Bourbigot, S. & Rumeau, P. (2009a). Application of Contact Angle
Measurement to the Manufacture of Textiles containing Microcapsules. Textile
Research Journal, Vol.79, No.13, (September 2009), pp.1202-1212, ISSN 0040-5175
Salaün, F., Devaux, E., Bourbigot, S. & Rumeau, P. (2009b). Influence of process parameters
on microcapsules loaded with n-hexadecane prepared by in situ polymerization,
Chemical Engineering Journal, Vol.155, No. 1-2, (December 2009), pp. 457-465, ISSN
1385-8947
Salaün, F., Devaux, E., Bourbigot, S. & Rumeau, P. (2010a). Development of Phase
Change Materials in Clothing Part I: Formulation of Microencapsulated Phase
Change. Textile Research Journal, Vol.80, No.3, (February 2010), pp.195-205, ISSN
0040-5175
Salaün, F., Devaux, E., Bourbigot, S. & Rumeau, P. (2010b). Influence of the solvent on the
microencapsulation of an hydrated salt. Carbohydrate Polymers, Vol.79, No.4, (March
2010), pp. 964-974, ISSN 0144-8617
Salaün, F., Devaux, E., Bourbigot, S. & Rumeau, P. (2010c). Thermoregulating response of
cotton fabric containing microencapsulated phase change materials. Thermochimica
Acta, Vol.506, No.1-2, (July 2010), pp. 82-93, ISSN 0040-6031
Salaün, F., Bedek, G., Devaux, E. & Dupont, D. (2011). Influence of the washings on the
thermal properties of polyurea-urethane microcapsules containing xylitol to
provide a cooling effect. Materials Letters, Vol.65, No.2, (January 2011), pp.381-384,
ISSN 0167-577X
Sánchez, L., Sánchez, P., de Lucas, A., Carmona, M. & Rodriguez J.F. (2007).
Microencapsulation of PCMs with a polystyrene shell. Colloid polymer science,
Vol.285, No. 12, (July 2007), pp.1377-1385, ISSN 0303-402X
Sánchez-Silva, L., Rodriguez, J.F., Romero, A., Borreguero, A.M., Carmona, M. & Sánchez P.
(2010). Microencapsulation of PCMs with a styrene-methyl methacrylate copolymer
shell by suspension-like polymerisation. Chemical Engineering Journal, Vol.157, No.1,
(February 2010), pp. 216-222, ISSN 1385-8947
Sánchez, P., Sanchez-Fernandez, M.V., Romero, A., Rodriguez, J.F. & Sanchez-Silva, L.
(2010) Development of thermo-regulating textiles using paraffin wax
microcapsules. Thermochimica Acta, Vol.498, No.1-2, (January 2010), pp. 16-21, ISSN
0040-6031
Sari, A. (2003). Thermal reliability test of some fatty acids as PCMs used for solar thermal
latent heat storage applications. Energy Conversion and Management, Vol.44, No.14,
(August 2003), pp. 2277-2287, ISSN 0196-8904
Sari, A. (2005). Eutectic mixtures of some fatty acids for low temperature solar heating
applications: Thermal properties and thermal reliability. Applied Thermal
Engineering, Vol.25, No. 14-15, (October 2005), pp. 2100-2107, ISSN 1359-4311
Sari, A. & Karaipekli, A. (2008). Preparation and thermal properties of capric acid/palmitic
acid eutectic mixture as a phase change energy storage material. Materials Letters,
Vol.62, No.6-7, (15 March 2008), pp. 903-906, ISSN 0167-577X
Sari, A., Biçer, A. & & Karaipekli, A. (2009a). Synthesis, characterization, thermal properties
of a series of stearic acid esters as novel solid–liquid phase change materials,
Materials Letters, Vol.63, No. 13-14, (May 2009), pp. 1213-1216, ISSN 0167-577X
Sari, A., Alkan, C., Karaipekli, A. & Uzun, O. (2009b). Microencapsulated n-octacosane as
phase change material for thermal energy storage. Solar Energy, Vol.83, No.10,
(October 2009), pp. 1757-1763, ISSN 0038-092X
Sari, A., Alkan, C. & Karaipekli, A. (2010). Preparation, characterization and thermal
properties of PMMA/n-heptadecane microcapsules as novel solid-liquid
microPCM for thermal energy storage. Applied Energy, Vol.87, No.5, (May 2010), pp.
1529-1534, ISSN 0306-2619
Sarier, N. & Onder, E. (2007a). The manufacture of microencapsulated phase change

materials suitable for the design of thermally enhanced fabrics. Thermochimica Acta,
Vol.452, No.2, (January 2007), pp. 149-160, ISSN 0040-6031
Sarier, N. & Onder, E. (2007b). Thermal characteristics of polyurethane foams incorporated
with phase change materials. Thermochimica Acta, Vol.454, No.2, (March 2007), pp.
90-98, ISSN 0040-6031
Shim, H., McCullough, E.A. & Jones, B.W. (2001). Using Phase Change Materials in
Clothing. Textile Research Journal, Vol.71, No.6, (June 2001), pp.495-502, ISSN 0040-
5175
Shin, Y., Yoo, D. & Son, K. (2005). Development of thermoregulating textile materials with
microencapsulated phase change materials (PCM). II. Preparation and application
of PCM microcapsules. Journal of Applied Polymer Science, Vol.96, No.6, (June 2005),
pp. 2005-2010, ISSN0021-8995.
Song, Q., Li, Y., Xing, J., Hu, J.Y. & Marcus, Y. (2007). Thermal stability of composite phase
change material microcapsules incorporated with silver nano-particles. Polymer,
Vol.48, No.11, (May 2007), pp. 3317-3323, ISSN 0032-3861
Su, J.F., Ren, L. & Wang, L.X. (2005). Preparation and mechanical properties of thermal
energy storage microcapsules, Colloid and Polymer Science, Vol.284, No.2,
(November 2005), pp. 224–228, ISSN 0303-402X
Su, J.F., Wang, L.X. & Ren, L. (2007) Synthesis of polyurethane microPCMs containing n-
octadecane by interfacial polycondensation: Influence of styrene-maleic anhydride
as a surfactant. Colloids and Surfaces A: Physicochemical and Engineering Aspects,
Vol.299, No.1-3, (May 2007), pp. 268-275, ISSN 0927-7757
Šumiga, B., Knez, E., Vrtačnik, M., Savec, V.F., Starešinič, M. & Boh, B. (2011). Production of
Melamine-Formaldehyde PCM Microcapsules with Ammonia Scavenger used for
Residual Formaldehyde Reduction. Acta Chimica Slovenica, Vol.58, No.1, (March
2011), pp.14-25, ISSN 1318-0207
Sundberg, E.J. & Sundberg, D. C. (1993). Morphology development for three-component
emulsion polymers: Theory and experiments. Journal of Applied Science, Vol.47,
No.7, (February 1993), pp.1277-1294, ISSN 0021-8995
Teixeira, M.I., Andrade, L.R., Farina, M. & Rocha-Leao, M.H.M. (2004). Characterization of
short chain fatty acid microcapsules produced by spray drying. Materials Science
and Engineering C: Biomimetic and Supramolecular Systems, Vol.24, No.5, (November
2004), pp. 653–65, ISSN 0928-4931
Torza, S. & Mason, S.G. (1970). Three-phase interactions in shear and electrical fields. Journal
of colloid and Interface Science, Vol. 33, No.1 (May 1970), pp. 67-83, ISSN 0021-9797
Uddin, M.S., Zhu, H.J. & Hawlader, M.N.A. (2002). Effects of cyclic operation on the
characteristics of a microencapsulated PCM storage material. International Journal of
Solar Energy, Vol.22, No.3-4, (September-December 2002), pp.105-114, ISSN 0142-
5919
Wang, X., Lu, E., Lin, W., Liu, T., Shi, Z., Tang, R. & Wang, G. (2000). Heat storage
performance of the binary systems neopentyl glycol/pentaerythritol and neopentyl
glycol/trihydroxy methyl-aminomethane as solid-solid phase change materials.
Energy Conversion and Management, Vol.41, No.2, (January 2000), pp. 129-134, ISSN
0196-8904
Yamagishi, Y., Takeuchi, H., Pyatenko, A.T. & Kayukawa, N. (1999). Characteristics of
microencapsulated PCM slurry as a heat-tranfer fluid. AIChE Journal, Vol.45, No.4,
(April 1999), pp. 696-707, ISSN 0001-1541
You, M., Zhang, X.X., Li, W. & Wang, X.C. (2008) Effects of MicroPCMs on the fabrication of
MicroPCMs/polyurethane composite foams. Thermochimica Acta, Vol.472,No.1-2,
(June 2008), pp. 20-24, ISSN 0040-6031
You, M., Zhang, X.X., Wang, X.C., Li, W. & Wen, W. (2010). Effects of type and contents of
microencapsuled n-alkanes on properties of soft polyurethane foams.
Thermochimica Acta, Vol. 500, No.1-2, (March 2010), pp. 69-75, ISSN 0040-6031
Zeng, J., Cao, Z., Yang, D., Xu, F., Sun, L., Zhang, L. & Zhang, X. (2009). Phase diagram of
palmitic acid-tetradecanol mixtures obtained by DSC experiments. Journal of
Thermal Analysis and Calorimetry, Vol.95 , No.2, ( February 2009), pp. 501-505, ISSN
1388-5150
Zhang, X. (2001). Heat-storage and thermoregulated textiles and clothing, IN: Smart fibres,
fabrics and clothing, Tao, pp. 34-58, Woodhead Publishing Ltd., ISBN 1 85573 546 6,
Cambridge.
Zhang, H. & Wang, X. (2009a). Synthesis and properties of microencapsulated n-octadecane
with polyurea shells containing different soft segments for heat energy storage and
thermal regulation. Solar Energy Materials and Solar Cells, Vol.93, No.8, (August
2009), pp. 1366-1376, ISSN 0927-0248
Zhang, H. & Wang, X. (2009b). Fabrication and performances of microencapsulated phase
change materials based on n-octadecane core and resorcinol-modified melamine-
formaldehyde shell. Colloids and Surfaces A: Physicochemical and Engineering Aspects,
Vol.332, No.2-3, (January 2009), pp. 129-138, ISSN 0927-7757
Zhang, H., Wang, X. & Wu, D. (2010). Silica encapsulation of n-octadecane via sol-gel
process: A novel microencapsulated phase-change material with enhanced thermal
conductivity and performance. Journal of Colloid and Interface Science, Vol.343, No.1,
(March 2010), pp. 246-255, ISSN 0021-9797
Zhang, X.X., Tao, X.M., Yick, K.L. & Wang, X.C. (2004a). Structure and thermal stability of
microencapsulated phase-change materials. Colloid and polymer science, Vol. 282,
No.4, (February 2004), pp. 330-336, ISSN 0303-402X
Zhang, X.X., Fan, Y.F., Tao, X.M. & Yick, K.L. (2004b). Fabrication and properties of
microcapsules and nanocapsules containing n-octadecane. Materials Chemistry and
Physics, Vol.88, No.2-3, (December 2004), pp. 300-307, ISSN 0254-0584
Zhang, X.X., Fan, Y.F., Tao X.M. & Yick, K.L. (2005a). Crystallization and prevention of
supercooling of microencapsulated n-alkanes. Journal of Colloid and Interface Science,
Zhang, X.X., Wang, X.C., Tao, X.M. & Yick, K.L. (2005b). Energy storage
polymer/MicroPCMs blended chips and thermo-regulated fibers. Journal of
Materials Science, Vol.40, No.14, (July 2005), pp. 3729-3734, ISSN 0022-2461
Zhang, X.X., Wang, X.C., Tao, X.M. & Yick, K.L. (2006). Structures and Properties of Wet
Spun Thermo-regulated Polyacrylonitrile-Vinylidene Chloride Fibers. Textile
Research Journal, Vol.76, No.5, (May 2006), pp. 351-359, ISSN 0040-5175
Zou, G.L., Tan, Z.C., Lan, X.Z., Sun, L.X. & Zhang, T. (2004). Preparation and
characterization of microencapsulated hexadecane used for thermal energy storage.
Chinese Chemical Letters, Vol.15, No.6, (2004), pp. 729–732, ISSN 1001-8417
Zuckerman, J.L., Pushaw, R.J., Perry, B.T. & Wyner, D.M. (1997). Fabric containing anergy
absorbing phase change material and method of manufacturing same. US Patent 5,
514,362, available from http://patft.uspto.gov/
Zuckerman, J.L., Pushaw, R.J., Perry, B.T. & Wyner, D.M. (2001). Fabric coating composition
containing energy absorbing phase change material. US Patent 6,207,738, available
from http://patft.uspto.gov/
11
Heat Transfer and

Thermal Air Management in the
Electronics and Process Industries
Harvey M. Thompson
Institute of Engineering Thermofluids, Surfaces & Interfaces (iETSI)
School of Mechanical Engineering, University of Leeds, Leeds,
United Kingdom
1. Introduction
Rising energy costs and important legislative drivers are making the achievement of
efficient heat transfer and thermal management of crucial importance in energy intensive
industries. In the electronics industry, inexorable increases in microprocessor performance
due to the use of multiple cores on a single chip are creating an enormous challenge for the
cooling infrastructure, since almost all of the electrical energy consumed by the chip
package is released as heat (Anandan & Ramalingam, 2008). This is particularly relevant to
the rapidly increasing number of large scale data centres, see Figure 1, which form the
backbone of the digital society on which the world’s population is becoming increasingly
reliant. The power consumption of data centres is rising sharply, having doubled in the last
five years and is likely to double again in the next five years to over 100 billion KWh
(Scofield & Weaver, 2008). These enormous energy requirements are presenting
governments and industry with a serious energy supply problem (Shehabi et al., 2011) and
the importance of data centres’ energy efficiency has now been recognised at the
international level with the formation of several industry consortia such as the Green Grid,
the Uptime Institute and the Data Centre Alliance to promote energy efficiency and best
practices in the data centre industry.
Since most enterprise data centres run significant quantities of redundant power and
cooling systems to produce higher levels of resiliency, this had led to significant power
consumption inefficiencies. The latter are exacerbated by the inefficiencies in the
Information Technology (IT) hardware and cooling requirements, each accounting for
roughly 40% of the total energy usage. This results in each KWh of energy for data
processing requiring a further KWh for cooling (Almoli et al., 2011). In a typical data centre,
electrical energy is drawn from the main grid to power an uninterruptible power supply
(UPS) which then powers the IT equipment, supply power to offices and to power the
cooling infrastructure: computer room air conditioning (CRAC) units, building chilled water
pumps and water refrigeration plant. The IT load inefficiencies can be improved by server
virtualisation and improved semi-conductor technologies, while the chiller plant is
generally the biggest energy cooling component and increasing the set point temperature of
the chilled water leaving the chiller evaporator offers significant potential reductions in the
overall cooling plant energy consumption.
Fig. 1. A large scale data centre with several rows of server racks
A key strategy for efficient thermal air management in a data centre, as recommended in the
EU Code of Conduct on Data Centre Energy Efficiency, is to separate hot and cold air via a
layout of alternating hot and cold aisles (Rasmussen, 2006; Niemann, 2008). These are shown
schematically in Figure 2. In the cold aisle containment strategy, cold air is supplied from the
CRAC units through floor tiles or diffusers into cold aisles and the racks are arranged so
that all server fronts/intakes face cold aisles. This counteracts the problem that arises if all
rows are arranged with intakes facing the same way, when equipment malfunction is
inevitable due to server overheating (Cho et al., 2009). In the hot aisle containment strategy, it
is the hot air that is contained and this approach can have advantages in terms of obviating
the need for raised floor tiles and providing hotter air to the CRAC units, increasing their
overall efficiency of performance. The importance of good air flow management in data
centres has led to increasing use of Computational Fluid Dynamics (CFD) (Versteeg &
Malalasekera, 1995) to design data centre operations to ensure the thermal environment
within data centres conforms to narrow, acceptable bands. Care must, however, be taken to
ensure that CFD predictions are properly validated and the limitations of its key
assumptions (for example on the coupling between the small-scale server air flows and the
larger scale data centre air flows) are understood (Almoli et al., 2011). Once validated, CFD
models can be very useful for data centre air flow management in enabling a large number
of design scenarios to be investigated and optimal server rack configurations to be identified
much more quickly than would be possible experimentally.
Heat Transfer and Thermal Air Management in the Electronics and Process Industries 201
Cold Aisle Containment
H H H
R R
C A A C
R H C C C H R
A K K A
C C
plenum C C
H = hot air
C = cold air Hot Aisle Containment
C
C C
R R
C H A C
R C A C R
C C
A K A
K
C C
plenum H H
Fig. 2. Cold and hot aisle containment strategies

Relying on air as the primary heat transfer medium in data centres is becoming increasingly
problematical due to inexorable increases in power densities in IT equipment. The reduced
effectiveness of using air to cool servers is promoting much greater interest in a range of
promising alternative technologies based on direct liquid loop cooling, such as dielectric
liquid immersion and on-chip spray and jet impingement cooling (Garimella, 2000). This is
because the higher heat capacities and associated heat transfer coefficients of liquids mean
that they are much more efficient at transferring the waste heat, but with the disadvantages
of requiring liquid loops as close as possible to the heat source. Some of the most promising
liquid cooling technologies in electronics are discussed briefly in section 2.
Energy consumption in the process industries is also currently an area where a significant
amount of research is being conducted. Due to the enormous range of heat transfer
technologies deployed in the process industries, this chapter focuses on one important heat
transfer component of several industrial applications, namely the use of convective heat
transfer from impinging air jets within industrial ovens (Martin, 1977; Sarkar & Singh, 2004).
These are used in applications ranging from the tempering of glass, drying of paper, textiles
and precision coated products, to the cooling of metal sheets, turbine blades and, indeed,
electronic components, as well as several examples in the food processing and baking
industries. Forced-convection ovens in the coating, converting and baking industries
typically use arrays of hot air impingement jets to transfer heat into products in order to, in
the former cases, vaporise their solvent components, and in the latter cases to bake
important food products such as bread, see Figure 3.
Hot air Coated film bread
(a) (b)
Fig. 3. Schematic diagram of forced-convection ovens using hot air impinging jets in the (a)
coating and converting and (b) bread baking industry
In the coating and converting industries, drying capacity is often the key limitation on
production speed and as a result high speed air jets (typically between 10 and 100m/s) are
used to increase the heat transfer coefficients and hence heat transfer into the coatings
(Martin, 1977). This in turn leads to greater problems with surface disturbances when drying
air emerges from an array of nozzles arranged perpendicular to the machine direction and
disturbs the surface of the wet coating. A recent study into surface disturbances in
multilayer coated products has shown how carefully redistributing solvent so as to increase
the viscosity of the upper layers can significantly improve robustness to drying-air induced
disturbances, leading to important commercial benefits in terms of reduced drying load and
increased drying rates (Ikin & Thompson, 2007).
Baking is also a complex process of simultaneous heat, water and water vapour transport
within the product where the heat is supplied by a variety of indirect-fired and direct-fired
forced convection ovens (Zareifard et al., 2006). Indirect ovens rely on radiation from heated
elements within an oven, whereas forced convection ovens are now increasing in popularity
since they can offer greater levels of thermal efficiency (Khatir et al., 2011). In the bread
baking industry, the primary concern is the effect of heat transfer on the final product
quality, which is influenced by the rate and amount of heat application, the temperature
uniformity and humidity levels in the baking chamber and the overall baking time (Zhou &
Therdthai, 2007). Temperature distribution is particularly important since it affects the
enzymatic reaction, volume expansion, gelatinization, protein denaturation, non-enzymatic
browning reaction and water migration. The timing and application is also very important
since supplying too high a temperature can cause early crust formation and a shrunken loaf.
Forced convection ovens in the baking industry transfer heat to the product by convection
from the surrounding air, radiation from the oven walls to the product surfaces and
conduction from its containers. The relative importance of convection and radiation is
determined by the baking temperatures and the speeds of the impinging jets; for low air
speeds (~1m/s) radiation is the predominant mode while convection is much more
important for higher air speeds (Boulet et al., 2010). Most previous studies in the bread
baking industry have tended to focus on regimes with relatively low air speeds, where
radiative heat transfer is most influential (Kocer et al., 2007), although high air speeds are
now receiving greater attention in the literature.
For many years the design and control of baking ovens relied on empirical models,
correlating overall performance with simple global parameters such as chamber volume, the
temperature of the heating elements and inlet conditions (Carvalho & Nogueira, 1997).
However, the increasing need to reduce energy consumption during baking has led to far
greater use of sophisticated mathematical models in order to optimise baking conditions.
These include models of the internal temperature and moisture conditions inside the
dough/bread (Zheleva & Kambourova, 2005) and several analyses based on Computational
Fluid Dynamics, which predict the velocity and temperature distributions within baking
chambers. Recent studies by Zhou & Therdthai (2007) and Norton & Sun (2007) have shown
how a baking oven’s energy consumption can be reduced by manipulating airflow patterns
so as to increase the volume of airflow while reducing the energy supplied. CFD models can
also provide valuable insight into key baking issues that influence product quality, such as
temperature uniformity, that are difficult to measure experimentally.
This chapter presents a brief review of some of the key thermal management challenges in
the electronic and process industries that are being addressed by current research projects
both at the University of Leeds and at other institutions. In the electronics industry, the
focus is on the rapidly burgeoning data centres industry, where efficient thermal air
management is crucial. The current role, capabilities and limitations of CFD modelling in
this sector are discussed, as are the promising future liquid cooling technologies that will be
increasingly needed as the limits of air cooling methods are reached. In the process
industries, the particular focus is on the challenges of improving the energy efficiency of
forced convection ovens used throughout the coating, converting and bread baking
industries. The key role of CFD modelling in improving oven design and operation is
discussed, together with a brief overview of the future experimental and computational
research needed to embed computational design methods into industrial practice.
2. Heat transfer and thermal airflow management in data centres

2.1 Air cooling management
As discussed above, currently most data centre cooling is achieved using cold air supplied
by CRAC units into data centres through raised floor tiles that then passes through the
server racks, cools the electronic equipment and emerges from the back of the servers as a
hot air stream, see Figure 2. Maintaining temperature and humidity design conditions is
critical to the good operation of data centres and generally temperature conditions at the
inlet to the racks should be maintained between 20-30oC and 40-55% relative humidity in
order to prevent equipment malfunction (Cho et al., 2009). Recent figures from the ASHRAE
trends in rack heat load shows typical server heat fluxes of 27KW for a 19 inch rack
(Shrivastava et al., 2009) and these will be even larger today.
The European Commission has created an EU Code of Conduct in response to increasing
energy consumption in data centres and the need to reduce the related environmental,
economic and energy supply security impacts. The Code of Conduct aims to achieve this by
improving understanding of energy demand within a data centre, raising awareness and
recommending energy efficiency best practice and targets. The Code of Conduct makes
several important recommendations for air flow management in data centres in order to
improve overall energy efficiency. A key recommendation is that the hot/cold aisle layout
should be implemented which aims to minimise the amount of bypass air, which returns to
CRAC units without performing cooling, and the amount of mixing of cold and hot air
which leads to higher air intake temperatures into servers. As shown in Figure 2 the
hot/cold aisle concept aligns equipment airflow to create aisles between racks that are fed
cold air from which the electronic equipment draws intake air in conjunction with hot aisles
to which all equipment exhausts hot air.
Although the cold air containment strategy is probably the most common today, the
alternative approach, termed hot-aisle containment, is also increasing in popularity (Niemann,
2008). In this approach the hot air from the servers is contained and is cooled before being
recirculated back into the room. Key advantages of this approach that have been proposed
include:
• it does not impact on surrounding data centre infrastructure and obviates the need for
raised floor tiles
• it enables return air to be returned to CRAC units at higher temperatures, enabling the
chillers to operate more efficiently and increase the proportion of the year during which
free cooling technologies (where no compressor is required) can be utilised.
• reduced humidification and de-humidification costs, saving energy and water.
There are currently conflicting opinions about which containment strategy is the best in
practice, however maximising the use of free cooling is another key recommendation of the
EU Code of Conduct. Other key thermal air management recommendations of the Code of
Conduct include:
• the use of blanking plates where there is no electronic equipment in order to prevent
cold air passing through gaps in the rack;
• installing aperture brushes to cover all air leakage opportunities provided by floor
openings at the base of racks and gaps in their sides;
• use of overhead cabling to prevent obstructions in air flow paths that increase the fan
power needed to circulate air throughout the data centre.
In addition to encouraging imaginative use of the waste heat produced in data centres, such
as using the low grade heat for buildings and swimming pools, the ability to control the
thermal air environment in data centres more accurately enables the chilled water set point
temperature to be increased, maximising the use of free cooling and reducing compressor
energy consumption significantly.
2.2 CFD modelling of thermal air flows in data centres

Computational Fluid Dynamics (CFD) is now frequently used to design the layout of servers
within data centres. Thermal air flows in data centres are complex, recirculating air flows
characterised by multiple length scales, modes of heat transfer and flow regimes. Length
scales range from processor length scales (order of mm) to rack length scales (of the order of
metres) up to data centre length scales (order of several metres). A typical Reynolds number,
Re, based on a typical air inlet velocity from supply vents of 1 m/s and a rack length scale of
2m leads to an estimated Re≈105 indicating the turbulent flow regime (Almoli et al., 2011).
However, as discussed by Choi et al. (2008), for the flow through servers racks the Reynolds
numbers are typically much smaller and may even lie within the challenging laminar-
turbulent transition regime which requires different flow models from those that can be
used at the data centre length scales for fully developed turbulent conditions. At present
there is no effective multi-scale CFD model that integrates the thermal circuit modelling of
microprocessors and data centre scale thermal flow modelling and which is capable of
adapting to dynamic conditions within data centres.
However, most previous CFD studies of data centre airflows have simply assumed the flow
outside the racks is fully turbulent and have used Reynolds Averaged Navier Stokes
(RANS) flow models, see e.g. Cho et al. (2009), while modelling the racks in a compact
manner without explicit representations of internal components. These are based on the
following governing continuity and momentum equations, written in RANS format as
∇.U = 0 (1)
∂U
∂t
1
( 1
+ ∇. (UU ) = ∇. σ − ρ U ' U ' + S
ρ ρ ) (2)
(
where σ = − P I + μ ∇ U + (∇ U )T ) is the Newtonian stress tensor, µ is the air viscosity, ρ its
density, U and U’are the average and turbulent fluctuation velocity vectors respectively, P
is the pressure and I the unit tensor. The vector S represents the additional momentum
sources, which are discussed below, and the − ρ U ' U ' term is the Reynolds stress tensor that
requires additional model equations.
Most CFD models of data centre airflows use the standard k-ε model (Cho et al., 2009)
where the turbulence is modelled in terms of the turbulent kinetic energy (k) and turbulent
dissipation (ε) . The two additional transport equations for the k-ε model are:
∂k 1 ⎛μ ⎞ 2μ
+∇. ( kU ) = ∇. ⎜ t ∇k ⎟ + t SijSij − ε (3)
∂t ρ ⎝ ρk ⎠ ρ
∂ε 1 ⎛μ ⎞ C ε 2 μt ε2
+∇. ( ε U ) = ∇. ⎜ t ∇ε ⎟ + 1ε Sij Sij − C 2ε (4)
∂t ρ ⎝ ρε ⎠ ρk k
k2
with the turbulent viscosity defined via μt = ρ C μ , the Sij terms are the deformation tensor
ε
and the ρk, ρε, C1ε, C2ε and Cµ are five empirical constants (Boulet et al., 2010). The energy
equation is also solved and takes the form
∂T ⎛⎛ ν ν ⎞ ⎞ 1
+ ∇. (T U ) =∇. ⎜ ⎜ + T ⎟ ∇T ⎟ + S (5)
∂t ⎜ Pr Pr ⎟ ρC Q
⎝⎝ T ⎠ ⎠ p
where T and ν are the temperature and dynamic viscosity respectively and Pr is the Prandtl
number defined by
ν k
Pr = where α = , (6)
α ρ Cp
k is the thermal conductivity and Cp is the air’s specific heat capacity. The subscript T
indicates the turbulent flow and SQ is the source term of the energy equation, namely the
heat generated by the processors.
Several commercial CFD codes have now been used to solve air flows in data centres,
ranging from general purpose codes such as Ansys Fluent 12 (Almoli et al., 2011), to a
number of codes specifically developed for the rapidly growing data centre industry; the
latter include CFD software packages such as Flovent, Six Sigma and TileFlow which are
designed for maximum ease of use. However, it is important to recognize that CFD is still
largely unverified for data centre airflows (Shrivastava et al., 2009), and that a hierarchy of
models is required for the data centre air flows and air flows through the racks. All CFD
models of data centre air flows should ideally only be used after careful validation against
experimental data.
The recent study by Almoli et al. (2011) noted that previous CFD studies of data centre air
flows have provided very little explanation of the way the flow through server racks are
modelled. This makes it very difficult to carry out meaningful comparisons with previous
CFD studies. They proposed that an efficient coupling between the data centre air flows and
air flow through the racks could be achieved by treating the racks as porous media. Their
permeabilities can be estimated experimentally by measuring pressure drops across the rack
for a range of flow rates and the rate of heat generation by the IT equipment can be
estimated from manufacturer’s specifications. They used this approach to develop the first
CFD model for data centre cooling scenarios where a liquid loop heat exchanger is attached
at the rear of server racks (back doors) which can avoid the need to separate the cold and
hot air streams in traditional hot/cold aisle arrangements and can also significantly reduce
the load on the CRAC units. This study also investigated the effectiveness of additional fans
in the back door heat exchangers.
2.3 Alternative liquid cooling techniques

Relying solely on air as the primary heat transfer medium in data centres is becoming
increasingly problematical due to inexorable increases in power densities from IT
equipment. Since liquids have much higher heat capacities and heat transfer coefficients
than gases, liquid cooling can potentially be much more effective than gas cooling for high
power electronic components. However, until relatively recently problems with liquid
cooling systems due to leakage corrosion, extra weight and condensation have limited their
use to high power density situations where air cooling is simply not viable. As discussed in
the recent review by Anandan & Ramalingham (2008), a range of alternative liquid cooling
technologies are now beginning to be taken up within industry. A selection of some of the
most promising approaches is outlined briefly below.
2.3.1 Dielectric liquid immersion cooling

Here, electronic components are immersed in a dielectric fluid as shown schematically in
Figure 4. This involves the boiling of the working fluid on a heated surface and is highly
effective since the phase change from liquid to vapour increases the heat flux from the
heated surface significantly and the high thermal conductivity of the liquid increases the
accompanying convection. The main limitation of using these methods is the lack of suitable
dielectric fluids, which are usually refrigerant-type fluids whose effectiveness can be limited
by problems associated with the long term corrosion of computer components.
2.3.2 Spray cooling

In spray cooling, a cooling agent in form of jet of liquid droplets, is injected through nozzles
onto the electronic module. The spray is formed by a pressure drop across the nozzle,
impinges on the surface and forms a thin liquid film. The heat from the electronic module is
dissipated by evaporating the cooling agent. The resultant hot liquid and vapour is recycled
through a spray drain, as indicated in Figure 5.
Liquid reservoir
Pressure relief
valve
Saftey valve
vapour
Dielectric
fluid
Fig. 4. Direct liquid immersion cooling
Vapour escape
to atmosphere
Multi-chip module
Spray
Coolant in
Spray drain
Fig. 5. Schematic diagram of the spray cooling approach

Spray cooling is a very promising cooling method for high heat flux applications (Mudawar,
2001). It has specific advantages since spraying the heat source directly eliminates the
thermal resistance of the bonding layer in electronic equipment and offers attractive ratios of
power supplied for cooling to rate of heat removal. An important limitation to the wider
adoption of spray cooling is that these must be non-conducting, dielectric liquids. Water is
often used when a thin protective, coated layer is applied to electronic equipment to reduce
the risk of short circuits due to water’s low dielectric strength. Relatively few alternative
liquids have demonstrated their suitability for spray cooling applications (Chow et al.,
1997).
2.3.3 Indirect liquid cooling

As the name suggests, in indirect cooling the liquid cooling agent does not have direct
contact with the electronic module and instead a thermal pathway is formed between the
module and the cooling agent, as shown in Figure 6. The thermal pathway is often a cold
plate with high thermal conductivity and since there is no contact between the module and
cooling agent, the latter can be any suitable liquid. The high thermal conductivity and
environmental-friendliness of water make it the most common cooling agent, however
foam-filled cold plates are increasingly being used for high heat flux cooling applications
(Apollonov, 1999, 2000).
Filter
Air to water
heat exchanger
Pump
Water
reservoir
Cold plate
Electronic
module
Fig. 6. Indirect cooling of electronic modules
2.3.4 Liquid jet impingement cooling

Many applications in industry require localised cooling and use impinging liquid jets to
achieve this objective. Important examples include the cooling of metal sheets, turbine
blades and high power density electronic components. In electronic cooling, cold liquid jets
are typically directed towards a surface from which heat needs to be removed. Figure 7
shows schematic diagrams of common approaches to electronic cooling using impinging
liquid jets (Anandan & Ramalingham, 2008). These can be classified into free-surface,
submerged and confined submerged jets (Wolf et al., 1993).
Nozzle Gas
Liquid Nozzle
Gas
Gas
Liquid Liquid
(a) (b) (c)

Fig. 7. Jet impingement configurations: (a) free-surface jet, (b) submerged jet, (c) confined
submerged jet
Impinging liquid jets typically have large heat transfer coefficients immediately below the
point of impact and rapidly decay away from the point of impact. This variation in heat
transfer coefficient needs to be borne in mind when impinging liquid jet cooling is being
considered for specific applications (Sarkar & Singh, 2004).
3. Impinging jet heat transfer in the process industries

In contrast to the application of impinging jets of cold liquid to cool products, hot air
impingement jets are widely used in process industries to transfer energy into a variety of
products. They are used, for example, to dry coated products, paper and textiles (Ikin &
Thompson, 2007), or to bake a wide variety of food products (Norton & Sun, 2006), by
directing hot air jets towards a target product in order to transfer energy into it. The effect of
parameters such as jet velocity, jet diameter, nozzle to chip spacing, nozzle geometry,
turbulence level and fluid properties on the effective heat transfer coefficients have been
reviewed in detail by several authors, see for example Martin (1977), Webb & Ma (1995),
Lienhard (1995) and Garimella (2000).
These have revealed that the flow patterns from impinging air jets have 3 characteristic
regions, as shown in Figure 8 below, namely the free-jet, impingement/stagnation flow
and wall-jet regions (Olsson & Trägårdh, 2007). The free-jet region has also been
categorised into 3 sub-regions: the potential core, developing flow and developed flow
ones. In practice, there is a wide variation in the heat transfer coefficient, which decays
from its maximum value in the stagnation point region, and jets with 6 ≤ H/D ≤ 8 are
found to be ideal from a heat transfer perspective because they ensure that the potential
core is fully decayed but without excessive energy dissipation associated with very long
jets. The actual optimal value of H/D does, however, depend on the transition effect and
the induction of turbulence in the jet wake. Lyttle & Webb (1994), for example, showed that
increased turbulence with small plate separations leads to significantly increased local heat
transfer.
nozzle
Potential core region Developing flow

Mixing region region
Developed/free flow region
H Radial flow region
Stagnation point Impingement surface
Fig. 8. Flow structure in an impinging air jet

Martin (1997) provided an early, very comprehensive survey of heat and mass transfer from
impinging jet flow. He found that heat transfer conditions can generally be described in
dimensionless form, where the Nusselt number is a function of the flow Reynolds and
Prandl numbers and of the geometrical variables. In these expressions, the appropriate
characteristic length is the hydraulic diameter of the nozzles, S, and the Reynolds number is
formed by mean velocity at the nozzle exit, calculated from the total mass flow rate. For
single round and slot nozzles, the mean heat transfer coefficient may be described in
dimensionless form by an average Nusselt number of the form
Nu = F(Re,Pr,r/D,H/D) (7)
where r is the distance from the stagnation point. For a single round nozzle, this was found
to take the particularly simple form
Nu = 0.5Re 2/3 Pr0.42 (8)
In a wide variety of forced convection ovens in the process industries the hot air jets emerge
from an array of nozzles arranged perpendicular to the machine direction and transfer heat
into the target products by convection from the hot air jets, radiation from the oven walls
and conduction from its containers, see Figure 3 above.
3.1 The coating and converting industries

In the coating and converting industries, forced convection ovens typically use hot air
impingement jets in order to supply the energy needed to vaporise the solvent components
and hence dry the coated products. In such systems, drying capacity is often the key limitation
on production speed and as a result high speed air jets (typically between 10m/s and 100 m/s)
are used to increase the heat transfer coefficients and hence heat transfer into the coated
products. However, the aggressive drying due to high air speeds can lead to practical quality
problems due to surface non-uniformities in coating systems. One such problem is due to the
too rapid depletion of solvents near the surface of the coating which leads to skin formation
which can prevent subsequent solvent transport out of the surface of the coated film. This
problem can usually be alleviated by using less aggressive drying in the front zones of the
drying ovens while the viscosity of the coating increases due to solvent depletion.
Another important problem is particularly prevalent during the manufacture of high
quality, multi-layer coatings, where drying air induced disturbances to the free surface of
coated films can destroy product quality. A recent study into coated product robustness to
drying-air induced disturbances has shown that an effective strategy to overcome this
problem is to redistribute solvent from the upper layers to the lower layers so that the
uppermost layers are more viscous and hence resistant to drying-air induced disturbances
(Ikin & Thompson, 2007). For products where this redistribution of solvent is not possible,
for example in the wide variety of single-layer coating systems, alternative hot air jet drying
methods may be preferable. One such method is shown in Figure 9, the so-called air
floatation drying approach (Noakes et al., 2002) where the product is dried by the hot air
issuing from air floatation nozzles, arranged above and below the coated web. In this
approach the main difficulty is to arrange the nozzles so that they produce a stable,
sinusoidal web profile, however this can provide effective drying for high quality, highly
sensitive industrial coatings.
Pressure
Upper plenum
Web
Lower plenum Coanda plate Web
(a) (b)
Fig. 9. Industrial air floatation drying: (a) stable web configuration and (b) typical pressure
profile under each floatation nozzle
3.2 The bread baking industry

Baking is a complex process of simultaneous heat, water and water vapour transport within
the product, where heat is supplied by a variety of indirect-fired and direct-fired forced
convection ovens (Zareifard et al., 2006). Indirect ovens rely on radiation from heated
elements within the oven, whereas forced convection ovens heat the product by convection
from hot air impinging jets, radiation from the oven walls, and conduction from its
containers. During baking it is very important to control the temperature distribution and
uniformity throughout the oven since this dominates product quality due to its effect on the
enzymatic reaction and water migration. The timing of the application of temperature is also
very important since supplying too high a temperature can cause early crust formation and
shrunken bread loaves.
For many years the design and control of baking ovens relied on empirical models,
correlating overall performance with simple global parameters such as chamber volume, the
temperature of heating elements and inlet conditions (Carvalho & Nogueira, 1997).
However, since roughly half of the energy use in a bakery is consumed in the baking oven
(Thumann & Mehta, 2008), the need to reduce energy consumption in the baking industry
has led to a particular focus on developing better scientific understanding and control of this
important aspect of baking processes. This in turn has led to far greater use of mathematical
modelling to optimise baking predictions by predicting, for example, crust thickness as a
function of operating conditions or the internal dough/bread, temperature and moisture
conditions during baking (Zheleva & Kambourova, 2005). Until recently, previous scientific
studies in the bread baking industry have tended to focus on regimes with relatively low air
speeds (<1m/s), where radiative heat transfer is most influential (Kocer et al., 2007).
However, forced convection ovens with higher air speeds now appear to be gaining in
popularity since they can offer greater levels of thermal efficiency (Khatir et al., 2011).
3.2.1 CFD modelling of thermal air flows in bread baking ovens

CFD modelling is now being increasingly applied to a wide range of different food
processes in order to improve product quality and reduce operating costs (Norton & Sun,
2006). Several CFD models of the thermal airflows in forced convection baking ovens have
appeared recently (Zhou & Therdthai, 2007) which predict the air velocity and temperature
distributions within baking ovens. However, since the thermal airflows in baking ovens are
highly complex, recirculating flows the choice of an appropriate turbulence model and
proper experimental validation of its predictions are essential. Most previous CFD studies of
forced convection baking ovens have used Reynolds Averaged Navier-Stokes (RANS)
turbulence closure equations in baking applications, including the standard k-ε model
(Norton and Sun, 2006, 2007) and the realizable k-ε model for flow in complex geometries
(Boulet et al., 2010). Far fewer CFD studies on ovens with high speed impinging air jets have
appeared to date, however reasonable agreement between CFD predictions of temperature
distribution and experiments has been reported recently (Khatir et al., 2011). These CFD
studies have shown that reduced overall energy consumption can be achieved by control
strategies based on heat flux into the bread rather than simply controlling the air
temperatures. For example it is possible to reduce energy consumption for particular baking
temperature profiles by increasing the volume of airflow whilst reducing the heat supplied
to the oven. CFD models have also been used to study how operating conditions can be
adjusted to achieve the optimum temperature profile since increasing temperature
uniformity in baking ovens is known to lead to better quality baked products. These are
now being integrated with other mathematical models of quality attributes to estimate
weight loss from, and the crust colour of, the bread.
One important issue that has not been addressed satisfactorily so far in CFD studies is the
choice of the most appropriate length-scale for thermal air flows in forced convection ovens.
This is usually based on a typical oven length scale, leading to Reynolds numbers well
within the turbulent flow regime. However, in light of the rapid flow transition shown in
Figure 8, a more appropriate length scale may be that based on a typical nozzle width (of the
order of mm) which can reduce the effective Reynolds number significantly and may even
take the flow into the complex laminar-turbulent regime. At the very least the assumption of
fully turbulent flow to justify the use of k-ε models needs to be verified experimentally by a
comprehensive programme of measurements of the airflow velocities, temperatures and
heat fluxes under practical operating conditions. At present no such comprehensive
experimental database exists.
Finally, note that if the maximum practical benefit for the baking industry is to be derived
from CFD models, these will need to be incorporated into formal design optimization
frameworks that enable physically meaningful objective functions to be minimised. Recent
steps in this direction have been taken (Hadiyanto et al., 2008, 2009; Purlis, 2011), however
much more progress still needs to be made in embedding these approaches into industrial
oven design and operational practice.
4. Conclusions
The achievement of efficient heat transfer and thermal air management is becoming
increasingly important in all energy intensive industries. The electronics industry now faces
major challenges to provide the cooling required by the rapidly expanding data centres that
form the backbone of the digital society and which produce enormous quantities of waste
heat that must be managed efficiently. The seriousness of this problem and the importance
of improving the energy efficiency of data centres has been recognised by the formation of
several industry-led consortia and governmental initiatives. At present the majority of data
centre cooling is achieved through recycling cold air through server racks and CFD is now
an integral means of improving thermal air flow management in data centres. Despite
having achieved several successes there are still important areas of weakness of current CFD
methods and a need for greater transparency in terms of describing the all-important
boundary conditions and greater access to validation case study data.
The energy consumption of process industries is also receiving greater attention in the
scientific literature. This chapter has focussed on one important aspect of this enormous
subject, namely convective heat transfer from impinging air jets in forced convection ovens
used in the coating, converting and bread baking industries. Advances in CFD methods are
now being exploited within these industries and have shown, for example, how the required
heat flux into products can be achieved more efficiently by optimising the air flow velocity
and temperature conditions. Outstanding issues for the CFD modelling of these systems
include the variation of turbulence levels throughout the oven, the validity of popular
turbulence models used to model them and, once again, the need for a comprehensive
database of experimental data for validation purposes. In order for industry to derive the
maximum benefit from the improving capabilities of CFD modelling, CFD models will need
to be incorporated into formal design optimization frameworks that are capable of
minimising physically meaningful objective functions.
5. Acknowledgements
The author would like to thank several colleagues at the University of Leeds for their
contribution and support. Thanks are particularly due to Dr Nikil Kapur, Dr Jon Summers,
Dr Malcolm Lawes, Professor Phil Gaskell and Professor Vassili Toropov, to industrial
collaborators: Airedale International Air Conditioning Ltd, Spooner Industries Ltd and
Warburton’s Ltd and the UK’s Engineering and Physical Sciences Research Council (EPSRC)
for financial support in these research areas.
6. References
Almoli, A; Thompson, A., Kapur, N., Summers, J.L, Thompson, H.M. & Hannah, G. (2011).
Computational Fluid Dynamic Investigation of Liquid Rack Cooling in Data
Centres, to appear in Applied Energy, available from:
doi:10.1016/j.apenergy.2011.02.003
Anandan, S.S. & Ramalingam, V. (2008). Thermal management of electronics: a review of
literature, Thermal Science, vol. 12, no. 2, (April 2008), pp 5-26, ISSN 0354-9836.
Ansys Inc. Ansys Fluent 12, 18th April 2011, Available from:
http://www.ansys.com/Products/Simulation+Technology/Fluid+Dynamics/AN
SYS+FLUENT/
Apollonov, V.V. (1999). Intensification of heat transfer in high-power laser diode bars
bymeans of a porous metal heat-sink, Optics Express, vol. 4, pp 27-32, (June 1999),
ISSN 1094-4087.
Apollonov, V.V. (2000). Highly efficient heat exchangers for laser diode arrays, Proceedings
of SPIE, vol. 3889, pp 71-81, ISBN 1094-4087, Osaka, Japan.
Boulet, M; Marcos, B., Dostie, M & Moresoli, C. (2010). CFD modeling of heat transfer and a
flow field in a bakery pilot oven, J. Food Engineering, vol. 97, no. 3, (June 2010), pp
393-402, ISSN 0260-8774.
Carvalho, M. & Nogueira, M. (1997). Improvement of energy efficiency in glass-melting

furnace, cement kilns and baking ovens, Applied Thermal Engng., vol. 17, nos. 8-10,
(Aug-Oct 1997), pp 921-933, ISSN 1359-4311.
Cho, J; Lim, T. & Sean Kim, B. (2009). Measurements and predictions of the air distribution
systems in high compute density (Internet) data centers, Energy & Buildings, vol. 41,
(October 2009), pp 1107-1115, ISSN 0378-7788.
Choi, J; Kim, Y., Sivasubramaruam, A., Srebric, J., Wang, Q. & Lee. J. (2008). A CFD-based
tool for studying temperature in rack-mounted servers, IEEE Transactions on
Computers, vol. 57, no. 8, (August 2008), pp 1129-1142, ISSN 0018-9340.
Chow, L.S., Sehemby, M.S. & Pais, M.R. (1997). High-heat-flux spray cooling, Annual Review
of Heat Transfer, vol. 8, (January 1997), pp 291-318, ISSN 0970-9991.
Data Centre Alliance. 18th April 2011, Available from: http://www.datacentrealliance.org
European Commission, EU Code of Conduct on Data Centres Energy Efficiency, version 2.0,
18th April 2011. Available from: http://ec.europa.eu/dgs/jrc/index.cfm?id=1410&
obj_id=6260&dt_code=NWS&lang=en 0.
Future Facilities, Six Sigma, 19th April 2011, Available from:
http://www.futurefacilities.com/software/room/6SigmaRoom.php
S.V. Garimella (2000). Heat transfer and flow fields in confined jet impingement, Ann. Rev.
Heat Transfer, vol. 10, (January 2000), pp 413-449, ISSN 1049-0787.
Green Grid, 18th April 2011. Available from: http://www.thegreengrid.org
Hadiyanto, H; Esveld, D.C., Boom, R.M., Van Straten, G. & Van Boxtel, A.J.B. (2008). Control
vector parametrization with sensitivity based refinement applied to baking
optimization, Food and Bioproducts Processing, vol. 86, no. 2, (June 2008), pp 130-141,
ISSN 0960-3085.
Hadiyanto, H; Boom, R.M., Van Straten, G., Van Boxel, A.J.B. & Esveld, D.C. (2009). Multi-
objective optimization to improve the product range of baking systems, Journal of
Food Process Engineering, vol. 32, no. 5, (October 2009), pp 709-729, ISSN 0145-8876.
Ikin, J.B. & Thompson, H.M. (2007). Drying-air induced disturbances in multi-layer coating
systems, Chemical Engineering Science, vol. 62, no. 23, (December 2007), pp 6631-
6640, ISSN 0009-2509.
Innovative Research Inc., Tileflow, April 19th 2011, Available from:
http://inres.com/products/tileflow/overview.html
Khatir, Z; Paton, J., Thompson, H.M., Kapur, N., Toropov, V.V., Lawes, M. & Kirk, D. (2011).
Computational Fluid Dynamics (CFD) Investigation of air flow and temperature
distribution in a small scale bread baking oven, to appear in Applied Energy,
Available from: doi:10.1016/j.apenergy.2011.02.002.
Kocer, D; Nitin, N. & Karwe, M. (2007). Applications of CFD in Jet Impingement Oven, in:
Da-Wen Sun (Ed.), Computational Fluid Dynamics in Food Processing, CRC Press, pp.
469–485, ISBN 978-0-8493-9286-3, retrieved from
http://www.crcnetbase.com/isbn/9781420009217
Lienhard, J.H., (1995). Liquid jet impingement, Annual Review of Heat Transfer, vol. 6,
(January 1995), pp 199-270, ISSN 1049-0787.
Lyttle, D. & Webb, B.W. (1994). Air impingement heat transfer at low nozzle spacings, Int. J.
Heat Mass Transfer, vol. 37, no. 12, (August 1994), pp 1687-1697, ISSN 0017-9310.
Martin, H. (1977). Heat and mass transfer between impinging gas jets and solid surfaces,
Advances in Heat Transfer, vol. 13, (March 1977), pp 1-60, ISBN 0120200139.
Mentor Graphics, Flovent, 18th April 2011, Available from:

http://www.mentor.com/products/mechanical/products/flovent
Mudawar, I. (2001). Assessment of high-heat flux thermal management schemes, Trans.
Components and Packaging Technologies, vol. 24, no. 2, (April 2001), pp 122-141, ISSN
1521-3331.
Niemann, J; (2008). APC White Paper #135, Hot Aisle vs. Cold Aisle Containment, 19th April
2011, Available from: www.apcmedia.com/salestools/DBOY-7EDLE8_R0_EN.pdf
Noakes, C.J.; Thompson, H.M., Gaskell, P.H., Lowe, S. & Osborn, M. (2002). Heat transfer
characteristics of air flotation dryers, Paper Technology, vol. 43, no. 10, (October
2002), pp 46-50, ISSN 0958-6024.
Norton, T. & Sun, D-W. (2006). Computational fluid dynamics (CFD) - an effective and
efficient design and analysis tool for the food industry: A review. Trends in Food
Science & Technology, vol. 17, no. 11, (November 2006), pp 600-620, ISSN 0924-2244.
Norton, T. & Sun, D-W. (2007). An overview of CFD Applications in the Food Industry, in:
Da-Wen Sun (Ed.), Computational Fluid Dynamics in Food Processing, CRC Press, pp.
1–41, ISBN 0849392861, retrieved from
Olsson, E.M. & Trägårdh, C. (2007). CFD Modeling of Jet Impingement during Heating and
Cooling of Foods, in: Da-Wen Sun (Ed.), Computational Fluid Dynamics in Food
Processing, CRC Press, pp. 487–503, ISBN 0849392861, retrieved from
Purlis, E. (2011). Bread baking: technological considerations based on process modeling and
simulation, Journal of Food Engineering, vol. 103, no. 1, (March 2011), pp 92-102, ISSN
0260-8774.
Rasmussen, N. (2006). Implementing Energy Efficient Data Centers, APC White Paper#114, 19th
April 2011, Available from:
www.apcmedia.com/salestools/NRAN-6LXSHX_R1_EN.pdf
Sarkar, A. & Singh, R.P. (2004). Air impingement technology for food processing:
visualization studies, Lebens. Wiss. U. Tech., vol. 37, no. 8, (December 2004), pp 873-
879, ISSN 0023-6438.
Scofield, C.M. & Weaver, T.S. (2008). Data center cooling using wet-bulb economizers,
ASHRAE Journal, vol. 50, (August 2008), pp 52-58, ISSN 00012491.
Shehabi, A.; Masanet, E., Price, H., Horvath, A. & Nazaroff, W.W. (2011). Data center design
and location: consequences for electricity use and greenhouse-gas emissions,
Building and Environment, vol. 46, no. 5, (May 2011), pp 990-998, ISSN 0360-1323.
Shrivastava, S.K.; Iyengar, M., Sammakia, G., Schmidt, R. & VanGilder, J.W. (2009).
Experimental-numerical comparison for a high-density data center: hot spot heat
fluxes in excess of 500W/ft2, IEEE Transactions on Components and Packaging
Technologies, vol. 32, no. 1, (January 2009), pp 166-172, ISSN 1521-3331.
Thumann, A & Mehta, D.P. (2008). Handbook of energy engineering, Fairmont Press, ISBN
0881735817, Boca Raton, USA.
Uptime Institute, 19th April 2011, Available from: (http://www.uptimeinstitute.org)
Versteeg, H.K. & Malalasekera, W. (1995). An introduction to computational fluid dynamics: the
finite volume method, Longman, ISBN 0582 21884 5, London, UK.
Webb, B.W. & Ma, C.F. (1995). Single phase liquid jet impingement heat transfer, Advances in
Heat Transfer, vol. 26, (March 1995), pp 105-217, ISSN 0065-2717.
Wolf, D;, Incropera, F.P. & Viskanta, R. (1993). Jet Impingement Boiling, Advances in Heat
Transfer, vol. 23, (March 1993), pp 1-132, ISSN 0065-2717.
Zareifard, M.R;, Marcotte, M. & Dostie, M. (2006). A method for balancing heat fluxes
validated for a newly designed pilot plant oven. Journal of Food Engineering, vol. 76,
no. 3, (October 2006), pp 303-312, ISSN 0260 8774.
Zheleva, I & Kambourova, V. (2005). Identification of heat and mass transfer processes in
bread during baking, Thermal Science, vol. 9, no. 2, (September 2005), pp 73-86, ISSN
0354-9836.
Zhou, W. & Therdthai, N. (2007). Three-dimensional CFD modeling of a continuous
industrial baking process, in: Da-Wen Sun (Ed.), Computational Fluid Dynamics in
Food Processing, pp 287-312, ISBN 0849392861, retrieved from
12
Unsteady Mixed Convection Flow in the

Stagnation Region of a Heated Vertical Plate
Embedded in a Variable Porosity
Medium with Thermal Dispersion Effects
S. M. Alharbi1 and I. A. Hassanien2
1Departmentof Mathematics, Faculty of Science,
Umm AL-Qura University, P.O. 8140, KSA
2Department of Mathematics, Faculty of science, Assiut University, Assiut,
Egypt
1. Introduction
The mixed convection flow finds applications in several industrial and technical processes
such as nuclear reactors cooled during emergency shutdown, solar central receivers exposed
to winds, electronic devices cooled by fans and heat exchanges placed in a low-velocity
environment. The mixed convection flow becomes important when the buoyancy forces
increase due to the temperature difference between the wall and the free stream. The mixed
convection flow in the stagnation region of a vertical plate has been investigated by
Ramachandra et al. [16].
When there is an impulsive change in the velocity field the inviscid flow is developed
instantaneously, but the flow in the viscous layer near the wall is developed slowly which
becomes fully developed steady flow after a while. For small period the flow is dominated
by the viscous forces and the unsteady acceleration, but for runtime it is dominated by the
viscous forces, the pressure gradient and the convective acceleration. The unsteady mixed
convection flow in the stagnation region of a heated vertical plate due to impulsive motion
has been studied by Schadri et al. [17]. The boundary layer flow development of a viscous
fluid on a semi-infinite flat plate due to impulsive motion of the free stream have been
investigated by Hall [5], Dennis [3] and Watkins [22]. The corresponding problem over a
wedge has been studied by Simth [18], Nanbu [11] and Williams & Rhyne [23].
The problem of unsteady free convection flow in the stagnation-point region of a rotating
sphere embedded in a porous medium has been analyzed by Hassanien et al. [7]. The
unsteady flow and heat transfer of a viscous fluid in the stagnation region of a three-
dimensional body embedded in a porous medium was investigated by Hassanien et al. [8].
The problem of thermal radiation and variable viscosity effects on unsteady mixed
convection flow in the stagnation region on a vertical surface embedded in a porous
medium with surface heat flux has been studied by Al-Arabi and Hassanien [6].
Motivated by all of the above referenced work and the significant possible applications of
porous media in industries, it is of interest in this paper to consider the unsteady mixed
convection flow in the region of a heated vertical plate embedded in a porous medium
having porosity distribution in the presence of the thermal dispersion with the effect of the
buoyancy force. The unsteadiness in the flow field is caused by impulsively creating motion
in the free stream and at the same time by suddenly increase in the surface temperature. The
partial differential equations governing the flow and the heat transfer have been solved
numerically using the finite difference scheme by Pereyra [14]. Particular cases of the
present results are compared with previously numerical work by Ramachandra et al. [16]
and Scshadri et al. [17]. The problem is formulated in such way that it represented by
Rayleigh type of equation at t=0 and for t → ∞ it represented by type of Hemennz equation.
2. Mathematical analysis
Let us consider a semi-infinite vertical plate embedded in a variable porosity porous
medium with thermal dispersion effect and uniform temperature T∞. At t = 0.0 the ambient
fluid is impulsively moved with a velocity Ue and at the same time the surface temperature
is suddenly raised. Figure (1) shows a flow field over a heated vertical surface where the
upper half of the field is assisted by the buoyancy force, but the lower part is opposed by the
buoyancy force. The surface of the plate is assumed to have an arbitrary temperature. All the
physical properties of the fluid are assumed to be constant except the density variation in
the buoyancy force term. Both the fluid and the porous medium are assumed to be in local
thermal equilibrium. Under above assumptions along with Boussinesq approximation, the
unsteady laminar boundary layer equations governing the mixed convection flow are given
by Vafai and Tien [20].
u g
v
Buoyancy assisting region
x
U∞ (x)
T∞ Tw(x)
y
Buoyancy assisting region
Fig. 1. The flow model and the coordinate system
∂u ∂v
+ =0 (1)
∂x ∂y
1 ∂u 1 ∂u ∂u ∂U e ν ∂ 2 u ν
+ 2 (u + v ) = U e + + gβ (T − T∞ ) + (ue − U e ) (2)
ε ∂t ε ∂x ∂y ∂x ε ∂y 2 k( y )
∂T ∂T ∂T ∂ ∂T
+u +v = (α c ) (3)
∂t ∂x ∂y ∂y ∂y
Unsteady Mixed Convection Flow in the Stagnation Region of a Heated
Vertical Plate Embedded in a Variable Porosity Medium with Thermal Dispersion Effects 219
The initial conditions are given by
u( x , y ) = v( x , y ) = 0, T ( x , y ) = T∞ , for t < 0. (4)
The boundary conditions for t ≥ 0 are given by
u( x ,0) = v( x ,0) = 0, u( x , ∞ ) = U e = ax , a > 0,

(5)
T ( x , ∞ ) = T∞ , T ( x ,0) = Tw ( x ) = bx n , b > 0, n ≥ 0.
The indices n = 0 and n = 1.0 correspond the constant surface temperature and the linear
surface temperature respectively. The variable x is measured along the surface and y is
measured normal to it. The fluid velocity u, v is in x and y direction respectively as shown in
figure (1). The fluid density, the fluid dynamical viscosity, the gravitational fluid acceleration,
the thermal expansion coefficient and the temperature will be denoted by ρ , μ , g , β , T
respectively. K(y) is the porous medium permeability, α c is the effective thermal diffusivity
and ε is the porous medium porosity. Equations (1) through (3) are supplemented by
constitutive equations for the variations of the porosity permeability and thermal
conductivity of the porous medium. It has been shown by Vafai [21] that the results
obtained experimentally by Nithiarasu et al. [12 ] in their study on void fraction distribution
in packed beds gives the functional dependence of the porosity on the normal distance from
the boundary and so the porosity can be represented by the exponential form
ε = ε 0 (1 + b exp ( −cy / d )) (6)
where ε 0 is the free-stream porosity, d is the particle diameter and b, c are empirical
constants that depend on the ratio of the bed to particle diameter. The values for ε 0 , b and c
chosen to be 0.38, 1, and 2 respectively. These values were found to give good approximation
to the variable porosity data given by Nithiarasu et al. [12] for a particle diameter d=5 mm.
The type of decay of porosity as the normal distance increases given by Equation (4) is well
established and has been used extensively in studies on flow in porous media with variable
porosity. It is also established that k(y) varies with the porosity as follows
d 2ε 3 (7)
K( y ) =
150(1 − ε )2
The effective thermal conductivity of the porous medium is given by Al-Arabi and Hassanien
[6]
αc = αm + γ d u (8)
where α m and γ are the molecular thermal diffusivity and mechanical dispersion coefficients,
respectively. Equations (1) through (3) can be transformed into a set of ordinary differential
equations by using the following transformations given by Williams and Rhyne [23].
η = ( a /ν )1/2 y ξ −1/2 , ξ = 1 − exp( −t *),
t * = at , a > 0, u( x , y , t ) = a x f ′(η , ξ ),
v( x , y , t ) = −( aν )1/2 ξ 1/2 f (η , ξ ),
T ( x , y , t ) = T∞ + (Tw ( x ) − T∞ )θ (η , ξ ),
Pr = ν / α , λ = Grx / Re 2x ,
Grx = gβ (T − T∞ )x 3 /ν 2 , Re x = ax 2 /ν . (9)
By using the transformations (9), equations (2-5) may be transformed to
ξ ξ 150(1 − ε )2
f ′′′ + 2 −1η (1 − ξ ) f ′′ + ff ′′ + (1 − f ′2 ) − Da 2 (1 − f ′)
ε ε ε (10)
ξ ∂f
− (1 − f ′2 ) + λξεθ = εξ (1 − ξ ) ,
ε ∂ξ
∂θ ′
Pr −1 θ ′′ + +Ds( f 'θ ')' + 2 −1η (1 − ξ )θ ′ + ξ ( f θ ′ − nf ′θ ) = ξ (1 − ξ ) , (11)
∂ξ
where Da = ν is the Darcian parameter , Ds = γ d is the dispersion parameter , λ > 0 for

ad 2 a
the buoyancy assisting flow and λ < 0 for the buoyancy opposing flow. The transformed
form of the variable porosity function becomes
ε = ε 0 (1 + b exp( −cη ξ Da))
The boundary conditions (5) may be reduce to
f (0, ξ ) = f ′(0, ξ ) = 0, f ′( ∞ , ξ ) = 1.0,

(12)
θ (∞ , ξ ) = 0, θ (0, ξ ) = 1.0.
It may be noted that the buoyancy parameter λ is a function of streetwise distance x unless
the surface temperature (Tw = T∞ ) for n = 1.0 , λ is constant. Equations (10) and (11) are
coupled nonlinear partial differential equations, but for ξ = 0, (t ∗ = 0) and ξ = 1.0, (t∗ → ∞ )
and they can be reduced to ordinary differential equations. For the case ξ = 0.0 , these
equations take the form
f ′′′ + 2 −1η f ′′ = 0, (13)
θ ′′ + 2 −1 Prηθ ′ = 0. (14)
For the case ξ = 1 , equations (7), (8) can be reduced to
f ′′′ + ff ′′ + (1 − f ′2 ) − γ (1 − f ′) − Δ(1 − f ′2 ) + λθ = 0, (15)
Pr −1 θ ′′ + ( f θ ′ − nf ′θ ) = 0. (16)
For the above two special cases the boundary conditions (5) may be reduce to
f (0) = f ′(0) = 0, f ′(∞ ) = 1.0,

(17)
θ (∞ ) = 0, θ (0) = 1.0
Equations (13), (14) are uncoupled linear equations while equations (15), (16) are coupled
nonlinear equations. Equations (13) and (14) under conditions (17) admit closed form
solutions which are given by,
f = η erfc(η / 2) − (π )1/2 [1 − exp( −η 2 / 4)], (18)
θ = erfc(Pr 1/2 η / 2) (19)

hence we have
f ′′(0) = (π )1/2 , θ ′(0) = −(Pr/ π )1/2 . (20)
Equations (15), (16) do not admit closed form solutions. Equations (10), (11) under conditions
(12) for ξ = 1 (steady case) and non porous media are identical to those of Ramachandra et
al. [16] in nonporous medium ( ε = 1 . Also equations (10-11) under the condition λ = 0
(forced convection flow) and ε = 1 are the same as that of Williams and Rhyne [23] if we put
m=1 in their equation. The physical quantities of interest in this problem are the skin friction
coefficient and the Nusselt number, which are defined by Pop et al. [15].
C f = 2τ w / ρU∞2 , Nu = xq w / α (Tw − T∞ ), (21)
where τ w = μ (∂u / ∂y )y = 0 and q w = −α ( ∂T / ∂y )y = 0 . Using equations (9), the quantities in Eqs.

(21) can be expressed in the form
C f = 2ξ −1/2 Re −x 1/2 f ′′(ξ ,0), ξ > 0. (22)
−1/2
Nu = − Re 1/2
x ξ θ ′′(ξ ,0), ξ > 0. (23)
3. Method of solution
We are going now discuss the local non-similarity method to solve equations (10), (11). Since
it was already seen by Pereyra [14], and Sparrow et al. [19] that for the problem of coupled
local non-similarity equations, the considerations of equation up to the second level of
truncation gives almost accurate results comparable with the solutions from other methods.
We will consider here the local non-similar equations (10), (11) only up to the second level of
truncation. To do this, we introduce the following new functions
g = ∂f / ∂ξ , φ = ∂θ / ∂ξ (24)
Introducing these functions into equations (10) and (11) we get
ξ 150(1 − ε )2 ξ
f ′′′ + 2 −1η (1 − ξ ) f ′′ + ff ′′ + Da 2 (1 − f ′) + (1 − f ′2 ) + λξεθ = εξ (1 − ξ ) g , (25)
ε ε2 ε
Pr −1 θ ′′ + +Ds( f 'θ ')' + 2 −1η (1 − ξ )θ ′ + ξ ( f θ ′ − nf ′θ ) = ξ (1 − ξ )φ ′ (26)
Differentiating the above equations with respect to ξ one may easily neglect the terms
involving the derivative functions of g and φ with respect to ξ as follows
⎡
ε ε ff ′′ε ξ ⎤
g′′′ + 2 −1 γ [(1 − ξ ) g′′ − f ′′] + ⎢ε fg′′ + ε gf ′′ + ff ′′ − ⎥+
ε2
⎣ ε ⎦
ε ⎡ 2 (1 − f ′2 )ε ξ ⎤
⎢(1 − f ′ + ε ( −2 f ′
g ) − ⎥+
ε 2 ⎢⎣ ε ⎥⎦
(27)
1 ⎡ ⎧⎪ (1 − ε )2 ε ξ ⎫⎪⎤
150Da 2 2 ⎢(1 − ε 2 )2 ( − g′) + (1 − f ′) ⎨(2)(1 − ε )( −ε ξ ) − 2 ⎬⎥
ε ⎢⎣ ⎩⎪ ε ⎭⎪⎥⎦
1⎡ (1 − f )ε ξ ⎤
− ⎢(1 − f ′2 ) + ξ 2 f ′g′ − ξ ⎥ + λεξφ + λθε + λξθε ξ = εφ ( 1 − 2ξ ) + ξ ( 1 − ξ ) ε ξ φ
ε⎣ ε ⎦
pr −1φ ′′ + ps( g′θ ′′ + θ ′g′′) + 2 −1 γ (1 − ξ )φ ′

−2 −1 γθ ′ + ( f θ ′ − nf ′θ ) + ξ ( f φ ′ + gθ ′ − ng′θ − nf ′φ )
= (1 − 2ξ )φ ′
where ε ξ is the derivative of ε with respect to ξ . The appropriate boundary conditions

satisfied by the above equations are given by
f (0, ξ ) = f ′(0, ξ ) = 0, f ′(∞ , ξ ) = 1.0,

θ (∞ , ξ ) = 0, θ (0, ξ ) = 1.0.
g(0, ξ ) = g′(0, ξ ) = φ (0, ξ ) = 0,

(28)
g′(∞ , ξ ) = φ (∞ , ξ ) = 0.
4. Results and discussion

In order to validate our numerical solutions, we have compared the surface shear stress
f ′′(ξ ,0) and the surface heat transfer −θ ′(ξ ,0) for the prescribed surface temperature with
those of Ramachandra et al. [16] and Scshadri et al. [17]. The results are found to be almost
compatible to a reasonable degree.. The comparison is shown in Figures (2) and (3), which
corresponding Figures (2) and (3) in Scshadri [17].
The variation of the surface shear stress f ′′(ξ ,0) , the surface heat transfer rate −θ ′(ξ ,0)
with time ξ (0 ≤ ξ ≤ 1) for the Darcy parameter Da , the dispersion parameter Ds in the
presence of the buoyancy assisting flow (λ = 1) and buoyancy opposing flow (λ = −1) for
the non-isothermal surface (n = 1) are shown in Figures (2) through (6).
At the start of motion (ξ = 0) , the buoyancy force parameter (λ ) , the Darcy parameter
Da and the dispersion parameter Ds have no effects on both the surface shear stress and
the surface heat transfer and these effects become pronounced with increasing time ξ .
The steady state is reached at ξ = 1 (t * → ∞ ) . The surface shear stress f ′′(ξ ,0) and surface
heat transfer −θ ′(ξ ,0) decrease with the Darcy and dispersion parameters increasing for
the two cases (the buoyancy assisting flow (λ = 1) and the buoyancy opposing flow
(λ = −1) ). It is also clear from these figures that the surface shear stress and heat transfer
for buoyancy assisting flow are greater than those of the buoyancy opposing flow. Also,
the surface shear stress and the heat transfer rate increase with increasing the Darcy
parameter Da and the dispersion parameter Ds.
Fig. 2. Variation of shear stress f ′′ (ξ ,0 ) with time ξ
Fig. 3. Variation of surface heat transfer θ ′ (ξ ,0 ) with time ξ

Fig. 4. Variation of surface shear stress with time for ε 0 = 0.38, b=1.0, c= 2.0, Ds=0.0 and
Pr=0.7
Fig. 5. Variation of the surface heat transfer rate θ ` (ξ,0) with time ξ for ε0= 0.38, b=1.0,
c= 2.0, d=5.0, Ds=0.0 and Pr=0.7
Fig. 6. Variation of the surface heat transfer rate θ ` (ξ,0) with time ξ for ε0 = 0.38, b = 1.0,
c = 2.0, d = 5.0, Ds = 0.0 and Pr = 0.7
Figures (5), (6) display the variation of the surface shear stress and the surface heat transfer
with time ξ for the prescribed surface temperature when the buoyancy parameter
λ = 1.0, − 1.0, pr = 0.7, n = −0.5 and for the Darcy parameter Da = 0.0, 0.3, and the dispersion
parameter Ds = 0.0, 0.8. The surface shear stress and the heat transfer increase with λ
because positive buoyancy force acts like favorable pressure gradient which accelerates the
motion and reduces both momentum and thermal boundary layers. Hence both the surface
shear stress and the surface heat transfer are increased.
5. Acknowledgement
This research has been supported by a grant from the Institute of Scientific Research at
Umm Al-Qura University, Saudi Arabia.
6. References
[1] Amin N. and Riley N., (1995), "Mixed convection at a stagnation point", Quart. J. Mech.
App. Maths. Vol. 48, pp. 111-121.
[2] Brown S. N. and Riley N, (1973), "Flow past a suddenly heated vertical plate", J. Fluid
Mech. Vol. 59, pp. 225-237.
[3] Dennis S. C. R., (1972), "The motion of a viscous fluid past an impulsively started semi-
infinite flat plate", IMA J. Appl. Math. Vol. 10, pp. 105-117.
[4] Ece M. C., (1992), "An initial boundary layer flow past a translating and spinning
rotational symmetric body", J. Eng. Math. Vol. 26, pp. 415-428.
[5] Hall M. G., (1969), "The boundary layer over an impulsively started flat plate",
Proceedings of the Royal Society of London, Series A, Mathematical and physical
science, Vol. 310A, pp. 401-414.
[6] Hassanien, I.A. and Al-Arabi, T.H., (2008), "Thermal Radiation and variable viscosity
effects on unsteady mixed convection flow in the stagnation region on a vertical
surface embedded in a porous medium with surface heat flux", Far East journal of
Mathematical sciences (FJMS), Volume 29, pp. 187 – 207.
[7] Hassanien, I.A., Ibrahim, F.S., Omer, Gh.M., (2004), "Unsteady free convection flow in the
stagnation-point region of a rotating sphere embedded in a porous medium",
Mechanics and Mechanical Engineering,Vol.7, No.2, pp. 89-98.
[8] Hassanien, I.A., Ibrahim, F.S., Omer, Gh.M., (2006), "Unsteady flow and heat transfer of a
viscous fluid in the stagnation region of a three-dimensional body embedded in a
porous medium", Journal of Porous Media,Vol.9,No. 4, pp. 357-372.
[9] Ingham D. B., (1985), "Flow past a suddenly heated vertical plate", Proceedings of the
Royal Society of London, Series A, Mathematical and physical science, Vol. 402A,
pp. 109-134.
[10] Kumari M., (1997), "Development of flow and heat transfer on a wedge with a magnetic
field", Archives of Mechanics. Vol. 49, No.5, pp. 977-990.
[11] Nanbu K., (1971), "Unsteady Falkner Skan flow", Zeitschrift Fur Angewandte
Mathematik und Physik ( ZAMP) , Vol. 22, No. 6 , pp. 1167-1172.
[12] Nithiarasu, P., Setharamu, K. N., and Sundararajin, T., (1997),"Natural convection heat
transfer in a fluid saturated variable porosity medium", Int. J. Heat Mass Transfer,
(1981), vol. 40, pp. 3955-3967.
[13] Ozturk A. and Ece M. C., (1995), "Unsteady forced convection heat transfer from a
translating and spinning body", ASME Journal of Energy Resources Technology,
Vol. 117, No. 4, pp. 318-323.
[14] Pereyra V., PASVA3, (1978), "An adaptive finite difference Fortran program for first
order non-linear boundary value problems", Lecture Note in Computer Science,
vol. 76, Springer, Berlin.
[15] Pop I., Gorla R. S. and Rashidi M., (1992), "The effect of variable viscosity on flow and
heat transfer to a continuous moving flat plate", International Journal of
Engineering and science, vol. 30, No. 1, pp. 1-6.
[16] Ramachandra N., Chen T., and Armaly B. F., (1988), "Mixed convection in the
stagnation flows adjacent to vertical surface", J. Heat Transfer vol. 110, pp. 173-177.
[17] Scshadri R., Srccshylan N. and Nath G., (2002), "Unsteady mixed convection flow in the
stagnation region of a heated vertical plate due to impulsive motion", Int. J. Heat
and Mass Transfer vol. 45 pp. 1345-1352.
[18] Smith S. H., (1957), "The impulsive motion of a wedge in a viscous fluid", Zeitschrift Fur
Angewandte Mathematik und Physik (ZAMP) Vol. 18, No.4pp. 508-522.
[19] Sparrow E.M., Quack H. and Boerner J., (1970), "Local non-similarity boundary
solutions, AIAA vol. 8, pp. 1936-1942.
[20] Vafai, K. and Tien, C.L., "Boundary and inertia effects on flow and mass transfer in
porous media", Int. J. Heat Mass Transfer, (1981), vol. 24, pp. 492-484.
[21] Vafai, K., "Heat transfer in variable porosity media", J. Fluid Mech., (1984), vol. 147, pp.
233-259.
[22] Watkins C. B., (1975), "Heat transfer in the boundary layer over an impulsively started
flat plate", J. Heat Transfer vol. 97, pp. 492-484.
[23] Williams J. C. and Rhyne T. H., (1980), "Boundary layer development on a wedge
impulsively set into motion", SIAM J. Appl. Math. Vol. 38, pp. 215-224.
13
Heat Generation and

Transfer on Biological Tissues
Due to High-Intensity Laser Irradiation
Denise M. Zezell1, Patricia A. Ana2,
Thiago M. Pereira1, Paulo R. Correa1 and Walter Velloso Jr.3
1Centro de Lasers e Aplicacoes, Instituto de Pesquisas Energeticas e Nucleares,
IPEN – CNEN/SP, Sao Paulo- SP
2Centro de Engenharia, Modelagem e Ciencias Sociais Aplicadas,
Universidade Federal do ABC, Santo Andre- SP

3Faculdade de Zootecnia e Engenharia de Alimentos de Pirassununga,
Universidade de Sao Paulo, Pirassununga- SP

Brazil
1. Introduction
The use of high intensity laser irradiation is already used clinically for many dental
procedures, since the literature findings confirm its efficacy. However, for the establishment
of a laser procedure for any clinical application, it is strongly necessary to study the thermal
effects on irradiated tissues, considering that most of biological tissues is highly connective
and can easily transmit the received heat (Nyborg & Brännström, 1968; Baldissara et al, 1997;
Brown et al, 1970).
The interactions of high intensity laser irradiation on dental hard tissues are commonly
induced by thermal action (Seka et al, 1996; Niemz, 1997). For the clinical application of
lasers, a precise irradiation parameter must be chosen in order to avoid morphological
damage, such as surface carbonization or cracking, which could produce structural, esthetic
damages and post-operative complaints. Moreover, the energy densities used must be safe
as regards pulp and periodontal tissue vitality (Ana et al, 2006; Nammour et al, 2004; Ana et
al, 2007).
Considering a clinical application of high intensity lasers, parameters such as wavelength,
energy density, intensity, peak power, average power, repetition rate and pulse length are
extremely important to heat generation due to irradiation on any biological tissue. The
amount of heat inside the tissue is highly dependent of its optical properties, such as
absorption and scattering coefficients. Also, the heat transfer is dependent of some
properties of biological tissues, such as thermal diffusivity, thermal conductivity and others
(Brown et al, 1970; Seka et al., 1996; Yu et al., 1993).
The knowledge of some basic concepts of temperature and heat are necessary to understand
and to evaluate the effects of lasers on biological tissues. These concepts are described
below.
1.1 Temperature
The thermal state of a body normally is described by subjective words like “hot”, “cold” or
“warm”. Therefore, the temperature is a physical parameter which is used to characterize
the thermal state of the body in a less subjective and more quantitative way; in other words,
assigning values.
The temperature is defined as the measurement of the thermal vibration (average kinetic
energy) of the body particles. Considering that the thermal vibration of atoms and molecules
is difficult to measure, it is necessary to assign numeric values to some macroscopic
properties of the body. For example, the Celsius scale was assigned 100 for water boiling
temperature and zero to water solidification temperature at the sea level. From this scale it is
possible to graduate and assign values to the different thermal states.
1.2 Heat
Heat is the energy associated to the temperature variation. When a hot body gets together
with a cold body, the hot body transfers heat (thermal energy) to the cold one. The energy
flow stops when both bodies have the same temperature, reaching thermal equilibrium.
Microscopically the hot body molecules vibrate more intensely than the cold body,
transferring this vibration (energy) to the cold body molecules, until both bodies reach the
equilibrium temperature which means they have the same mean kinetic energy. The
equation that related the heat with temperature variation is:
ΔQ = m.c.ΔT (1)
where ΔQ is the heat amount of energy (cal) acquired or lost. In this equation,
( )
ΔT = T f − Ti is the temperature variation ( T f = final temperature, Ti = initial temperature);
m is the mass (g) and c is the specific heat (cal/g°C).
1.3 Heat propagation

The heat is energy transferred in association to a temperature variation. The heat
propagation occurs spontaneously from the hot body to cold body. There are three different
mechanisms for heat propagation into materials:
a. Conduction: this is the typical heat propagation mechanism that occurs in solids. The
vibrating particles inside the body transmit part of their kinetic energy to their
neighbors. As a consequence, he heat flows from the higher temperature regions (where
molecules have higher mean kinetic energy) to the low temperature ones (where
molecules have lower mean kinetic energy). This process, defined as heat flow, is
directly proportional to the cross-sectional area, to the temperature difference between
the two points and inversely proportional to the thickness (Figure 1).
The heat conduction equation is also known as Fourier´s law. For the one-dimensional
plane wall shown in Figure 1, the heat flow can be defined as:
ΔQ A ⋅ (Ti − T f )
heat flow = = k⋅ (2)
Δt Δz
where ΔQ is the heat amount that propagates through the area A and thickness Δz ,
Δt is the time interval, Ti e Tf are temperatures initial and final and k is a
proportionality constant, called thermal conductivity coefficient. This coefficient is a
Heat Generation and Transfer on
Biological Tissues Due to High-Intensity Laser Irradiation 229
characteristic of each material and expresses the facility that the heat flows through the
material. The materials with a high k value are called good thermal conductors and with
low k is a good thermal insulating.
b. Convection: this mechanism describes how the heat propagates in fluids (liquids or
gases). When heating a container with a fluid, the part closer to the heat source became
warmer, decreasing its density. Whence hot fluid goes up and the colder one goes
down, and these movements are called convection currents.
c. Irradiation: this heat propagation mechanism corresponds to the emission or
absorption of electromagnetic heat waves; for instance, the same way that the Sun heats
the Earth.
Fig. 1. Schematic diagram of heat propagation in a block having cross section area A and
thickness Δz
1.4 Heat generation in a laser-tissue interaction

Considering the interaction of a laser source with the biological tissue, there are some
models that should be described: the microscopic and the macroscopic model.
a. Microscopic model
When a photon with energy hv reach a molecule A, the energy could be absorbed, bringing
A to excited state A*:
hv + A → A∗ (3)
In this way, the molecule A* suffers an inelastic collision (equal energy loss) with the
neighbor B (an electron, atom or a molecule), transferring part of its energy and decay to an
energy lower state:
A∗ + B(ε cin ) → A + B(ε cin + Δε ) (4)
This process changes part of the laser energy (high frequency electromagnetic waves) into
thermal energy (low frequency electromagnetic waves).
The amount (ε cin + Δε ) represents the increase of the thermal vibration of the particle B and,
as a consequence, there is a microscopic temperature increment by the energy absorbed
from the photon.
The macroscopic effect of the temperature increase is observed due to the high number of
type-A molecules, which absorbed the energy from a high number of photons presented in a
laser beam, and transform this energy in thermal vibration. Figure 2 represents the
phenomenon described above.
Fig. 2. Schematic diagram of the heat generation in a biological tissue: (a) a photon with the
energy hv is applied in the molecule A, then goes to an excited state A*. (b) The molecule A*
collides with B. (c) A* transfers its energy to B; B becomes B(ε cin + Δε ) and starts to vibrate
more intensely
Fig. 3. Schematic diagram showing the laser-tissue interaction: reflection, scattering,

absorption and transmission
b. Macroscopic model
In a macroscopic approach it could be observed that the heat generated is directly related
with the laser propagation in the tissue. For that, it is convenient to remember how the heat
propagation occurs. When a laser beam irradiates a sample (Figure 3), a part of the beam is
reflected and the other part penetrates in the surface. That part which penetrates is
attenuated mainly in two different ways: the absorption and scattering - as long as the beam
penetrated in the sample.
The absorption and the scattering are characterized for absorption coefficient (μa) and
scattering coefficient (μs), which represents, respectively, the rate of radiation energy loss
per penetration length unit, due the absorption and the photons scattering. These two
coefficients are specific to each tissue and depend on the laser wavelength.
Fig. 4. Laser beam attenuation as a function of penetration length

To simplify, initially consider an absorber and not scattering sample. In this case, the beam
attenuation is described by the Beer’s law (Figure 4):
I ( z) = I 0 ⋅ e − μa ⋅z (5)
where I is the beam intensity that depends on the penetration length z and I0 is the intensity
for z = 0. The inverse of the absorption coefficient is defined as the optical absorption length
(L) (Figure 4):
1 (6)
L=
μa
The generated heat per area unit and per time unit, in a very small thickness ∆z, is given by:
I ( z ) − I ( z + Δz) ∂I ( z)
S( z ) = =− = μa ⋅ I ( z) (7)
Δz ∂z
The equation 7 expresses that the generated heat in the tissue is equal to the absorbed
energy and can be described as the absorption coefficient multiplied by the local intensity.
In the most cases, the light is both absorbed and scattered into the sample simultaneously.
The beam attenuation continues to be described by a similar law from the Beer’s Law, but
now the attenuation coefficient is the sum of the absorption and scattering coefficients,
which is called total attenuation coefficient ( μT = μ a + μs ) .
1.5 Heat propagation in biological tissues

The heat conduction equation in a material medium is given by:
∂T k S
= ∇ 2T + (8)
∂t ρ ⋅ c ρ ⋅c
where T is the temperature (°C), t is the time (s), k is the thermal conductivity, ρ is the tissue
density (g/cm3), c is the specific heat (cal/g.°C) and S is the generated heat per area and per
time (cal/s.cm2).
This equation can be deduced from the diffusion general equation, but it requires a specific
Physics and Mathematical knowledge. Therefore it is important to know that it describes a
⎛ ∂T ⎞
strong correlation among the temperature temporal variation ⎜ ⎟ , the temperature spatial
⎝ ∂t ⎠
variation (∇ 2T ) and the laser source S. It is also important to say that this same equation
also works when the sample is not being irradiated. In order to calculate how the heat
propagates after an exposure time, when the laser beam is off, it is only necessary to solve
the equation 8 with S = 0
There are some other thermal parameters related to the heat propagation. The thermal
penetration length is a parameter that describes the propagation extension per time, and it is
given by:
zthermal (t ) = 4 ⋅ α ⋅ t (9)
k
where α = is the tissue thermal diffusivity and t is the time. For instance, the thermal
ρ ⋅c
diffusivity of water is α = 1.4 × 10 −7 m2 s .
Other important parameter is the thermal relaxation time, which is obtained mathematically
correlating the optical penetration length with thermal penetration length:
L = zthermal
1
= 4 ⋅ α ⋅ τ thermal
μa
1
τ thermal = (10)
μ a2 ⋅ 4 ⋅ α
The thermal relaxation time (equation 10) describes the necessary time to the heat
propagates from the surface of irradiation until the optical penetration length and is
particularly important when the intention is to cause a localized thermal damage, with
minimal effect in adjacent structures. This parameter can be interpreted as follows: if the
time of the laser pulse is smaller than the relaxation time, the heat would not propagate
until a distance given by the optical penetration length L. So the thermal damage will
happen only in the first layer where the heat is generated. On the other hand, if the time of
the laser pulse is higher than the relaxation time, the heat would propagate for multiple of
the optical penetration length, resulting in a thermal damage in a bigger volume to the
adjacent structures.
2. Characteristics of dental tissues and their influence on heat propagation

The tooth is composed basically for enamel, dentin, pulp and cementum. Enamel, dentin
and cementum are called “dental hard tissues”, and the main constituent is represented by
the hydroxyapatite (Chadwick, 1997; Gwinnett, 1992) (Figure 5). Dentin and cementum have
higher water and organic compound percentage when compared to the enamel and, due to
this composition, they are more susceptible to heat storage than the enamel. Dentin has low
thermal conductivity values and offers more risk when lasers irradiate in deeper regions,
considering that dentinal tubules area and density increase at deepest regions, and
subsequently, can easily propagate the generated heat (Srimaneepong et al., 2002). As an
example, considering the use of CO2 lasers in dentistry (wavelength of 9.6 µm or 10.6 µm),
the absorption coefficient for dentin tissue is lower than enamel due to its low inorganic
content; also, the thermal diffusivity is approximately three times smaller, which can lead a
less heat dissipation amount and, as a consequence, can induce higher pulp heating (Fried et
al., 1997).
Fig. 5. Representation of a molar tooth, evidencing the macroscopic structures

Dental pulp is a connective and vital tissue, and the higher vascularization makes this tissue
strong susceptible to thermal changes. The minimal change in pulp temperature (ΔT ≤ 5 °C)
is sufficient to alter the microvascularization, the cellular activation and their capacity of
hydratation and defense (Nyborg & Brännström, 1968; Zach & Cohen, 1965).
The majority of high intensity lasers used for dental hard tissues cause photothermical and
photomechanical effects. Photons emitted at wavelength of visible and near infrared regions
of electromagnetic spectrum are poorly absorbed by dental hard tissues (Seka et al., 1996)
and, due to this fact, the heat diffusion to the pulp is easy. In this way, in order to choose a
parameter of laser irradiation for a clinical application, it is necessary to establish limit energy
densities that promote a significant temperature increment on enamel and dentin surface, in
order to produce mechanical and/or thermal effects on these structures (Ana et al., 2007).
Also, the temperature increment inside the pulp tissue must be bellowing a temperature
threshold.
Previous studies have indicated that temperature increments above 5.6 °C can be considered
potentially threatening to the vitality of the pulp (Zach & Cohen, 1965) and increments in
excess of 16 °C can result in complete pulpal necrosis (Baldissara et al., 1997). Further studies
showed levels of 60% and 100% of pulp necrosis when pulp tissue was heated about 11 °C
and 17 °C, respectively (Powell et al., 1993). The pulpal temperature rise due to laser-tissue
interaction has also been investigated and most of lasers systems promoted an increase in
pulpal temperature dependent on the power setting (Ana et al., 2007; Yu et al., 1993; Zezell
et al., 1996; Boari et al., 2009).
As well as the knowledge of laser wavelength, energy density and pulse duration, another
point to be considered in heat transfer is the tissue characteristics and the influence of the oral
environment. Although the calculation of heat transmission and dissipation is performed
using hole sound teeth at in vitro studies, in clinical situations several characteristics of
tissue can change, such as the type of teeth, the remaining thickness, the presence of saliva
and the presence of demineralization (Ana et al., 2007; Powell et al., 1993). For instance, due
to the great amount of water in carious lesions, the heat transfer to the pulp can be more
excessive in decayed teeth. Relating the influence of tissue thickness, White et al.(1994)
determined that Nd:YAG laser irradiation with a power output of 0.7 W (approximately 87
J/cm2) induces an increase of 43.2 °C in a remaining dentin thickness of 0.2 mm and induces
an increment of 5.8 °C in a dentin thickness of 2.0 mm. Considering that the human teeth
present a big variation in volume and weight, and taking into account the low thermal
conductivity of dentin, the operator must judge the physical conditions of dental hard tissue
in order to adequate the exposition time to avoid dangerous thermal effect on pulp.
3. Changes in tissue thermal characteristics during laser irradiation

Considering the laser irradiation in dental hard tissues, it is necessary to know and to
understand the thermal behavior of these tissues when submitted to heating. For that, the
evaluation of the heat conduction phenomenon is extremely necessary.
Teeth are mainly composed by hydroxyapatite that, in principle, has high heat capacity
value and low heat conduction value (Pereira et al., 2008). The main reason of the changes of
thermal parameters of hydroxyapatite can be explained by the complexity of the photon
diffusion into the material due to the ionic bond between the chemical elements.
Several studies about thermal parameters measurement in hard dental tissues have been
published (Brown et al, 1970; Incropera et al., 2006). Results of these studies are summarized
in table 1.
Thermal parameter Enamel Dentin Water
0.71 1.59 4.18

Specific Heat (J/g°C)
(Brown et al, 1970) (Brown et al, 1970) (Incropera et al., 2006)
Thermal conductivity 9.34 5.69 6.1

(10-3W/cm °C) (Brown et al, 1970) (Brown et al, 1970) (Incropera et al., 2006)
Thermal diffusivity 4.69 1.86 1.3

(10-3cm2/s) (Brown et al, 1970) (Brown et al, 1970) (Incropera et al., 2006)
Table 1. Thermal parameters of dental hard tissues (enamel and dentin) and water
Although these thermal values are well-established in literature and can be used for
supporting clinical applications, it is important to consider that all parameters were
measured at room temperatures. In the moment of laser irradiation of dental hard tissues,
the temperature increase can lead several chemical and ultra-structural changes on enamel
and dentin (Bachmann et al., 2009; Fowler & Kuroda, 1986); as a consequence, the tissue
thermal characteristics of tissue may change during laser irradiation.
Several studies have been developed in order to propose theoretical models of heat
propagation in dental hard tissues (Craig R.G & Peyton, 1961; Braden et al., 1964). These
models assumed that thermal parameters are constant in function of temperature, which
seems to be not true according to the discussed above. Thus, we have to assume that the
determination of laser irradiation parameters based only by theoretical calculation that
consider thermal properties as constant can be wrong. Figure 6 shows experimental data
(Pereira et al., 2008), obtained by infrared thermography, of the thermal diffusivity changes
as function of temperature changes.
Fig. 6. Thermal diffusivity of dentin as function of temperature (Pereira et al., 2008)

Figure 7 shows the changes on heat penetration on dentin in function of time of exposure. It
can be seen that data obtained vary among the related studies due to the fact that some of
them consider the thermal diffusivity values always constant, while the present study
(Pereira et al., 2008) consider the changes in thermal diffusivity according to the temperature
(Figure 7). This fact has significant relevance mainly for clinical procedures using laser
irradiation, when it is necessary temperature increases up to 800 °C for cutting dental hard
tissues and for caries prevention, for example (Fried et al., 1996; Ana et al., 2007). When a
tooth is submitted to this temperature elevation, the heat spreads more quickly than
calculated by theoretical models that considered thermal diffusivity values as constant,
which can represent a problem mainly for the deeper tissues (pulp tissue).
0,30 Craig
Braden
pereira300
0,25 pereira500
thermal depth (cm)
0,20
0,15
0,10
0,05
0,00
0 2 4 6 8 10
time (s)
Fig. 7. Calculated thermal depth (cm) in function of time (t) for dentin tissue, obtained by
four different literature studies (Pereira et al., 2008)
4. Considering temperature to determine clinical protocols using lasers

As it was stated previously, for the determination of clinical protocols it is demanding to
consider the safety and efficacy of lasers, also the characteristics and properties of target
tissues. Besides that, literature studies clearly show that laser features, such as wavelength,
mode of operation (continuous versus pulsed modes), temporal pulse length and repetition
rate are characteristics directly related with pulp heating. In this way, among the optical
properties, the transmission is the most important property to be considered for preserving
pulp vitality.
Among high intensity lasers with high absorption and low transmission through enamel
and dentin, erbium lasers seems to be the most appropriated wavelength to be used in
dentistry. However, some studies point out that, even with this laser, the repetition rate and
pulse duration are decisive on determining clinical parameters; for example, the longer
pulse duration is, the higher is the heat generated in pulp (Yu et al., 1993).
Taking into account the clinical application of high intensity lasers on dental hard tissues,
some strategies may be useful to control the heat generation and transmission on these
tissues. In order to restrict the heat dissipation through the teeth tissues, the application of a
photosensitizer is frequently applied over the enamel and dentin surfaces before laser
irradiation, and this application can avoid pulpal damages even when laser irradiation
occurs with high energy densities (Tagomori & Morioka, 1989; Jennett et al., 1994). The
application of a photosensitizer before laser irradiation is commonly used in order to
enhance surface tissue absorption in the near-infrared range for ablation and caries
prevention actions in dental tissues, considering that some lasers, such as Nd:YAG and
Ho:YAG, are poorly absorbed by enamel and dentin. The absorption of the laser beam is
increased at the surface of the enamel and the heat produced due to laser absorption in the
coating material is transmitted into the adjacent enamel. This technique certifies the deposit
of a short laser pulse energy to a small volume of tissue, avoiding the excessive laser beam
penetration in deeper dental structures and consequently with less risk of damages in dental
pulp (Boari et al., 2009).
The use of Indian Ink is a well-recognized and efficient technique to reduce beam
transmission on dental hard tissues. However, because of the difficulty in its removal, which
can prejudice the aesthetics of remaining teeth, it has been suggested the application of a
coal paste, a mixture of triturated vegetal coal in 50% ethanol, which is biocompatible, easy
to remove and presented important results in previous in vitro (Boari et al., 2009) and in vivo
(Zezell et al., 2009) studies. In an in vitro study performed by our group, it was
demonstrated that the enamel recovering with the coal paste promoted an increase of
surface temperatures, which confirmed the absorption of laser beam at the surface (Ana et
al., 2007) (Figure 8). Also, the coal paste significantly decreased the heat transfer into the
teeth when enamel was irradiated with Nd:YAG and Er,Cr:YSGG lasers, and can assure the
pulpal safety when laser irradiation is performed for a long period of time. The
Fig. 8. Surface temperature increase on enamel surface during Er,Cr:YSGG (λ = 2078 nm)
laser irradiation with and without the application of coal paste (Ana et al., 2007). It can be
noted that, even at three different average powers, the presence of the photosensitizer
significantly increased the surface temperature during laser irradiation. Bars mean standard
deviation
morphological changes promoted on enamel surface are similar than those promoted by the
recovering with Indian Ink, showing evidences of surface heating that promoted melting
and recrystallization of enamel (Boari et al., 2009) (Figure 9).
(a)
(b) (c)
Fig. 9. Scanning electron micrography of dental enamel after irradiation with Nd:YAG
(λ = 1064 nm) laser irradiation at energy density of 84.9 J/cm2 after surface recovering with
Indian Ink (a) or coal paste (b) or no recovering (c) (Boari et al., 2009). It is possible to note
the presence of melting and recrystallization of enamel after recovering with coal paste and
Indian Ink. These characteristics are not observed when enamel is irradiated without the
presence of a photosensitizer. Original magnification = 3500 X. (a) enamel + Indian Ink;
(b) enamel + coal paste; (c) sound enamel
The presence of air-water spray during laser irradiation is another strategy used for
clinicians to avoid excessive heat generation on the pulp. The water coolant allows the
cleaning of surfaces to be irradiated and increases the efficacy of ablation phenomenon, in a
process called “water augmentation” (Fried et al., 2002). When dental hard tissues are
irradiated with Er:YAG in addiction to a thin water layer, studies relate that the cutting
efficiency increases at the same time that the pulp temperature decreases. However, the
thickness of water layer should be well-controlled, considering that erbium lasers interacts
primary with water and an thick water layer over the tissue can restrict the laser interaction
with the enamel bellow it and, as a consequence, the absorption by the target tissue can
decrease.
5. How to determine temperature variations in biological tissues?

Among physical methods to determine the temperatures on materials, the thermocouples
(Ana et al., 2007; Boari et al., 2009), elliptical mirrors, HgCdZnTe detectors (Fried et al., 1996)
and infrared cameras (Ana et al., 2007) are the most used ones to measure temperature
changes in pulp, periodontal tissues and dental hard tissues surfaces. These techniques
present good accuracy and efficacy, and can be easily adapted to experimental conditions.
However, it should be considered that all experimental methods present some difficulties,
such as sample standardization (considering the large variation in volume, size, thickness,
and hydratation degree of tissues), the exact duplication of the thermal load, accuracy,
availability and cost of equipments (Ana et al., 2008).
The finite element method model (FEM model) is another method that has been popular
among researchers, taking into account that this technique is a good analytical tool to model
and simulate the thermal or mechanical behavior of dental structures (Toparli et al., 2003).
The FEM model can be used to simulate the effects of laser on enamel and dentin, but not on
gums or inside the pulp cavity, which is filled with blood vessels and innervated tissues,
since these materials are soft and highly inhomogeneous. However, the effects of laser
irradiation could be difficult to simulate even on the dental hard tissues (only enamel and
dentine), since the thermal characteristics of these materials may not have been well
determined.
It must be pointed out that all the in vitro methods do not reproduce exactly all the
interferences of the oral environment in the photothermal response of enamel and dentin
tissues, such as the influence of surrounding saliva, pulpal and periodontal tissues, presence
of biofilm, body temperature and other factors (Ana et al., 2008).
Fig. 10. FEM model used to simulate the heat generation and transmission at dental pulp
during Er,Cr:YSGG laser irradiation (Ana et al., 2008)
Fig. 11. Calculated temperature distribution: The gray scale represents the temperature in
Celsius. The sequence of images illustrates the effects of irradiation with a laser beam and
the propagation of heat through the tooth after the beam is turned off
6. Pre-clinical studies
As affirmed previously, the pre-clinical studies are necessary to predict the biological effects
of high-intensity lasers and to establish possible parameters and conditions for a future
application in dental practice. For preventing dental caries, for example, infrared lasers such
as Nd:YAG and Er,Cr:YSGG can be indicated.
The Er,Cr:YSGG laser is emitted in 2.78 µm wavelength, which is better absorbed by water
and OH¯ contents of hydroxyapatite (Seka et al., 1996), and promotes surface temperatures
up to 800 °C at the ablation threshold (Fried et al., 1996). Due to this fact, Er,Cr:YSGG laser is
applied for cutting of enamel, dentin and root surfaces, and also for caries prevention.
In a first in vitro study, surface temperature measurements were performed in order to
verify if this laser had potential to promote chemical and crystalline changes on dental
enamel, which can occur in temperatures above 100 °C (Bachmann et al., 2009). For that, the
temperature changes in enamel surface during and immediately after laser irradiation were
monitored using an infrared high resolution fast thermographic camera (ThermaCam FLIR
SC 3000 Systems, USA), which stores infrared images and data at rates up to 900 Hz. This
experiment was performed at a controlled room temperature of 24.6 °C, 47 % air relative
humidity and considering teeth emissivity as 0.91. The thermographic camera was
positioned at 0.1 m distance of samples and the obtained infrared images were recorded at
rates of 900 Hz for later analysis (Ana et al, 2007).
For laser irradiation, laser handpiece was positioned at focused beam, at 1 mm distance
from the enamel surface. This assembly was kept in optical supports and the area of interest
was isolated at a focal length of 0.1 m using an internal macro lens.
The results of surface temperature obtained in this study (Ana et al., 2007) are shown in
Figure 12. It is possible to evidence that the surface temperature rises with the increase of
energy density, and the presence of photosensitizer (coal paste) propitiated higher
temperature values when compared to the surfaces in which were not previously recovered
with the coal paste. In this way, even using laser wavelengths highly absorbed by dental
enamel, the application of a photosensitizer in the enamel surface can potentiate the
absorption phenomenon. This can reflect on temperature rise and, in this way, crystalline
changes at this superficial enamel may occur and can favor the caries preventive effect.
Another point to be considered is that the temperature rise of 247.6 °C found when teeth
were irradiated with 8.5 J/cm2 is lower than that temperature reported by literature studies,
who found approximately 400 °C measured by an elliptical mirror and a HgCdZnTe
detector with a time resolution of 1 µs (Fried et al., 2006). In this way, the temperature
elevation during Er,Cr:YSGG laser irradiation could be higher than those detected by
infrared camera. Taking into account that the pulse width of Er,Cr:YSGG laser is 140 μs,
even the 900 Hz recording rate of the infrared thermographic method seems to be unable to
detect the highest temperature peaks during laser irradiation. In this way, the infrared
thermographic camera gives an idea of average temperature changes when teeth are
irradiated with high intensity lasers. However more accurate systems are required to
precisely determine the maximum temperature peaks, such as the use of integrating sphere.
240 2
2.8 J/cm without coal
220 2
5.6 J/cm without coal
2
200 8.5 J/cm without coal
2
180
2.8 J/cm with coal
2
5.6 J/cm with coal
160 2
Temperature ( C)
8.5 J/cm with coal

o
140
120
100
80
60
40
20
-2 0 2 4 6 8 10 12 14 16 18 20 22 24
Time (s)
Fig. 12. Dental enamel surface temperature measurement, by infrared thermography,

during Er,Cr:YSGG laser irradiation at parameters aimed at caries prevention (Ana et al.,
2007)
For measurement of heat transfer to the pulp chamber, the use of fast-response
thermocouples seems to be more accurate since it is not possible to see the heat transfer from
enamel to pulp by infrared thermography, unless the teeth are half-sectioned.
In this way, calibrated K-type chromel-alumel thermocouples (Omega Engineering,
Stanford, USA) were inserted inside the pulp chamber of sound molar human teeth, which
were previously filled with a thermally-conductive paste (thermal conductivity of 0.4 cal s-1
m-1 K-1 – Implastec, Votorantim, Brazil) in order to keep thermal contact between the probe
end and dentin surface. These thermocouples had 0.05 mm diameter probe and were
sensitive to temperature variations between 0.1 °C and 100 °C. The temperature sensitive
end of the probe was placed at the closest distance to the area to be irradiated, and its
location was controlled radiographically for each sample (Romano et al., 2011). The
thermocouple apparatus was connected to an analogue-to-digital converter (SR lock-in
amplifier, Stanford Research System, USA) linked to a computer, and time and temperature
data were recorded at sampling rate of 20 Hz, with temperature resolution of 0.1o C. During
laser irradiation, samples were fixed and immersed in a water-filled heating circulator at
standardized temperature of 37 °C, with only the coronal part of the tooth not being
submerged in order to simulate body temperature in the oral environment.
2
2 .8 J /c m w ith o u t c o a l
5 .0 5 .6 J /c m
2
w ith o u t c o a l
2
8 .5 J /c m w ith o u t c o a l
4 .5 2
2 .8 J /c m w ith c o a l
2
4 .0 5 .6 J /c m w ith c o a l
2
8 .5 J /c m w ith c o a l
3 .5
3 .0
Δ T ( C)
2 .5
o
2 .0
1 .5
1 .0
0 .5
0 .0
0 10 20 30 40 50 60
t (s )
Fig. 13. Pulp chamber temperature variation during enamel surface irradiation with
Er,Cr:YSGG laser, measured by fast-response thermocouples (Ana et al., 2007)
Figure 13 shows the results of temperature evaluation inside the pulp chamber during
enamel irradiation by Er,Cr:YSGG laser. It is possible to observe that the device can detect
minimal temperature variations and, although surface temperatures detected increased up
to 230 °C, the pulp temperature variations were up to 4.5 °C. This fact evidences that dental
enamel and dentin are good thermal insulating tissues. Also, the presence of the
photosensitizer on enamel surface was important to effectively reduce the heat transfer
through the pulp chamber, increasing the safety of a future clinical procedure. It must to be
emphasized that the time of exposure is important and further histological in vivo studies
are also necessary to confirm this hypothesis.
All studies described previously were performed to evaluate just one possible clinical
application of Er,Cr:YSGG laser. Considering that this laser can be used for multiple
applications, the development of a FEM model could be a faster tool to evaluate the heat
generation and transfer to dental hard tissues. In this way, a further study was performed to
develop this model and to compare the simulation results with those obtained
experimentally (Ana et al., 2008).
For that, a geometric FEM model of a half-sectioned double rooted molar tooth was
constructed using a typical profile of a tooth root. Since the goal was to calculate the
temperature distribution on the surface and inside the tooth, in each element of the model
the total heat is given by the internal heat, determined by the material density (ρ) and
specific heat (c), and the heat flux, determined by the element material thermal conductivity
(α). In this study, the external heat source is due to almost instantaneous light absorption
converted into heat. Moreover, in spite of not considering wavelength dependence with
absorption, reflection, transmission and scattering, the FEM model predicts accurate values
for temperatures inside the teeth, mainly because 2.78 µm is strongly absorbed by the dental
hard tissue. The optical penetration is very small (to the order of few micrometers) since the
optical absorption coefficient of enamel is about 7000 cm-1.
The in vitro experiment aimed to compare the results of FEM simulation was performed
using half-sectioned teeth, irradiated with Er,Cr:YSGG laser on enamel surface and
monitored by infrared imaging. This set-up allowed the visual evaluation of temperature
changes and heat diffusion from enamel into dentin and the pulp chamber. Taking into
account that the variation in temperature is also dependent on the volume and weight of a
tooth, the use of sectioned teeth does not correspond to an in vivo condition inside the
mouth, but gives a reasonable idea of heat transfer inside the tooth and gives the exact value
of surface temperature during laser irradiation. Moreover, in a clinical protocol, laser
irradiation is performed by scanning all over the enamel surface, and it was not possible to
reproduce this situation in the present study. However, in a clinical application of lasers, the
time of irradiation in just one region of tissue is always less than the time considered during
experiments, which increases the safety of evaluated parameters.
Energy Presence of Infrared thermographic Finite element method

Density photosensitizer camera model
Surface Pulp Surface Pulp
yes 79.6 °C 0.5 °C 94.2 °C 0.95 °C
2.8 J/cm2 no 42.7 °C 1.0 °C 66.0 °C 0.9 °C
yes 184.1 °C 1.5 °C 178.0 °C 1.9 °C
5.6 J/cm2 no 136.1 °C 1.2 °C 136.0 °C 1.4 °C
yes 247.6 °C 2.1 °C 231.0 °C 2.1 °C
8.5 J/cm2 no 211.8 °C 1.9 °C 188.0 °C 2.3 °C
Table 2. Comparison of means of temperature rise on surface and in pulp obtained by FEM
model and by thermographic camera (Ana et al., 2008)
Table 2 shows the results of temperature monitoring by infrared thermography and the
simulation by FEM model. It is possible to observe a good correlation between the in vitro
and computational method, indicating that FEM model can be used as an alternative to
determine heat generation on the enamel surface as well as inside the pulp chamber
according to the several laser energies and the presence or not of a photosensitizer.
Although it is not possible to simulate the influence of oral environment on FEM model,
under the given conditions the simulated model was shown to have a good approximation
to the physical reality.
7. Conclusion
The effect of high intensity lasers irradiation on biological tissues and consequently their
clinical uses are based on heat generation, which is necessary to assure effective clinical
procedures such as faster cutting, good homeostasis and desired chemical changes on target
tissues. The comprehension of heating generation and transmission through these tissues is
essential to determine safe irradiation parameters of lasers and, for that, the knowledge of
optical and thermal properties of tissues and their changes due to heat are strongly
necessary. Also, the interaction of laser wavelength with the tissues is necessary to avoid
deleterious effects in target tissues, as well in surrounding ones.
The pre-clinical experiments give us important information about laser-tissue interaction,
and help to suggest laser parameters and conditions for development of a further clinical
protocol. For that, methods such as thermocouples, infrared thermography and finite
element simulation are good tools that demonstrated to be useful on predicting the clinical
results.
8. References
Ana PA, Bachmann L, Zezell DM. Laser effects on enamel for caries prevention. Laser Phys
2006; 16(5): 865 – 875.
Ana PA, Blay A, Miyakawa W, Zezell DM. Thermal analysis of teeth irradiated with
Er,C:YSGG laser at low fluences. Laser Phys Lett 2007; 4: 827 - 830.
Ana PA, Velloso WF Jr, Zezell DM. Three-dimensional finite element thermal analysis of
dental tissues irradiated with Er,Cr:YSGG laser. Rev Sci Instrum. 2008; 79(9): 093910.
Bachmann L, Rosa K, Ana PA, Zezell DM, Craievich AF, Kellermann G. Crystalline
structure of human enamel irradiated with Er,Cr:YSGG laser. Laser Phys Lett 2009;
6: 159-162.
Baldissara P, Catapano S, Scotti R. Clinical and histological evaluation of thermal injuty
thresholds in human teeth: a preliminary study. J Oral Rehabilitation 1997; 24: 791-
801.
Boari HGD, Ana PA, Eduardo CP, Powell GL, Zezell DM. Absorption and thermal study of
dental enamel when irradiated with Nd:YAG laser with the aim of caries
prevention. Laser Phys 2009; 19(7): 1463-1469.
Braden M. heat conduction in teeth + effect of lining material. J Dent Res 1964, 43 (3): 315.
Brown WS, Dewey WA, Jacobs HR. Thermal properties of teeth. J Dent Res 1970; 49(4): 752-
755.
Chadwick DJ, Cardew G. Dental Enamel. London UK: Chichester-Wiley 1997.
Craig R.G, peyton F.A. Thermal conductivity of Tooth Structure, Dental cements, and
amalgam. J Dent Res 1961, 40 (3): 411.
Fowler BO, Kuroda S. Changes in heated and in laser-irradiated human tooth enamel and
their probable effects on solubility. Calcif Tissue Int 1986; 38: 197-208.
Fried D, Feathestone JDB, Visuri SR, Seka WW. Walsh JT The caries inhibition potential of
Er:YAG and ErCr:YSGG laser irradiation. In: Wigdor HA, Featherstone JDB,
White JM, Neev J. Lasers in Dentistry II Bellinghan WA Proc SPIE 1996; 2672: 73-
77.
Fried D, Zuerlein MJ, Featherstone JDB, Machule D. Thermal and chemical modification of
dentin by pulsed CO2 laser irradiation at 9-11 µm. In: Wigdor HA, Featherstone
JDB, Rechmann P. Lasers in Dentistry III Bellinghan WA: Proc SPIE 1997; 2973: 94-
100.
Fried D, Ashouri N, Breunig T, Shori R. Mechanism of water augmentation during IR laser
ablation of dental enamel. Lasers Surg Med 2002; 31: 186 – 193.
Gwinnett AJ. Structure and composition of enamel. Oper Dent 1992; Supl 5: 10-17.
Incropera F P., DeWitt D P, Bergman T L, Lavine A S. Fundamentals of Heat and Mass
Transfer. 6th ed., New York: Wiley, 2006.
Jennett E, MotameIi M, Rastegar S, Frederickson C, Arcoria C, Powers JM. Dye-enhanced
ablation of enamel by pulsed lasers. J Dent Res 1994; 73: 1841-1847.
Nammour S, Kowalyk K, Valici C, Zeinoun T, Rocca JP, Powell GL, Van Reck J. Safety
parameters for pulp temperature during selective ablation of caries by KTP laser in
vitro. J Clin Laser Med Surg 2004; 22(2): 99-104.
Niemz MH. Laser-tissue interactions: fundamentals and applications. 3rd ed., Berlin:
Springer, 1997.
Nyborg H, Brännström M. Pulp reaction to heat. J Prosthet Dent 1968; 19: 605 – 612.
Pereira T. M.; Miyakawa, W; Goulart VP; Zezell, DM. In-plane thermal diffusivity of dentin
by infrared thermography. In: VII Congresso da SBPMAT, 2008.
Powell GL, Morton TH, Whisenant BK. Argon laser oral safety parameters for teeth. Lasers
Surg Med. 1993; 13(5):548-52.
Romano AC, Aranha AC, Silveira BL, Baldochi SL, Eduardo CP. Evaluation of carbon
dioxide laser irradiation associated with calcium hydroxide in the treatment of
dentinal hypersensitivity. A preliminary study. Lasers Med Sci. 2011; 26(1): 35-
42.
Seka W, Featherstone JDB, Fried D, Visuri SR, Walsh JT. Laser ablation of dental hard tissue:
from explosive ablation to plasma-mediated ablation. In: Wigdor HA, Featherstone
JDB, White JM, Neev J. Lasers in Dentistry II Bellinghan WA. SPIE 1996; 2672: 144-
158.
Srimaneepong V, Palamara JEA, Wilson PR. Pulpal space pressure and temperatura changes
from ND:YAG laser irradiation of dentin. J Dent 2002; 30: 291-296.
Tagomori S, Morioka T. Combined effects of laser and fluoride on acid resistance of human
dental enamel. Caries Res 1989; 54: 15-17.
Toparli M, Sasaki S. Finite element analysis of the temperature and thermal stress in a
postrestored tooth. J Oral Rehabil. 2003; 30(9): 921-6.
White JM, Mark CF, Goodis HE. Intrapulpal temperature during pulsed Nd:YAG laser
treatment on dentin in vitro. J Periodontol 1994; 65: 255-259.
Yu D, Powell GL, Higuchi WI, Fox JL. Comparison of three lasers on dental pulp chamber
temperature change. J Clin Laser Med Surg 1993; 11(3): 119-122.
Zach L, Cohen G. Pulp response to externally applied heat. Oral Surg 1965; 19: 515-530.
Zezell DM, Cecchini SCM, Pinotti M, Eduardo CP. Temperature changes under Ho:YLF
irradiation. In: Wigdor H, Featherstone JDB, White JM, Neev J. Lasers in Dentistry II
Bellinghan WA: Proc SPIE 1996; 2672: 34-39.
Zezell DM, Boari HG, Ana PA, Eduardo CP, Powell GL. Nd:YAG laser in caries prevention:
a clinical trial. Lasers Surg Med. 2009; 41(1): 31-5.
14
Entransy Dissipation Theory

and Its Application in Heat Transfer
Mingtian Xu
Shandong University
P. R. China
1. Introduction
In this chapter, our recent work on the entransy dissipation theory and its application in
heat convection and the heat exchanger design is reviewed. In that work, the
thermodynamic basis of the entransy dissipation theory is established. It is shown that the
entransy is a state variable and the second law of thermodynamics can be described by the
entransy and entransy dissipation. Then this entransy dissipation theory is applied to the
convective heat transfer and heat exchanger design. The local expression of the entransy
dissipation rate is obtained. The extremum principle of entransy dissipation rate is
proposed. The expressions of the entransy dissipation caused by heat conduction and fluid
friction in heat exchanger are derived. The optimization design of heat exchanger is
discussed.
1.1 The second law of thermodynamics in terms of entransy

The second law of thermodynamics is one of the most important fundamental laws in
physics, which originates from the study of the efficiency of heat engine and places
constraints upon the direction of heat transfer and the attainable efficiencies of heat engines
(Kondepudi & Prigogine, 1998). The concept of entropy introduced by Clausius for
mathematically describing the second law of thermodynamics has stretched this law across
almost every discipline of science. However, in the framework of the classical
thermodynamics the definition of entropy is abstract and ambiguous, which was noted even
by Clausius (Clausius, 1865). This has induced some controversies for statements related to
the entropy. Recently, Bertola and Cafaro found that the principle of minimum entropy
production is not compatible with continuum mechanics (Bertola & Cafaro, 2008). Herwig
showed that the assessment criterion for heat transfer enhancement based on the heat
transfer theory contradicts the ones based on the second law of thermodynamics (Herwig,
2010). The entropy generation number defined by Bejan (Bejan, 1988) is not consistent with
the exchanger effectiveness which describes the heat exchanger performance (Guo et al.,
2009b). Shah and Skiepko (2004) found that the heat exchanger effectiveness can be
maximum, minimum or in between when the entropy generation achieves its minimum
value for eighteen kinds of heat exchangers, which does not totally conform to the fact that
the reduction of entropy generation leads to the improvement of the heat exchanger
performance. These findings signal that the concepts of entropy and entropy generation
may not be perfect for describing the second law of thermodynamics.
Although there has been effort to modify the expression of the second law of
thermodynamics (Bizarro, 2008; Ben-Amotz & Honig, 2003, 2006) and to improve the
classical thermodynamics by considering the Carnot construction cycling in a finite time
(den Broek, 2005; Esposito & Lindenberg, 2009; Esposito et al., 2010), the eminent position of
entropy in thermodynamics has not been questioned. Recently, Guo et al. (2007) defined two
new physical quantities called entransy and entransy dissipation for describing the heat
transfer ability and irreversibility of heat conduction, respectively. Guo et al. (2009a) have
introduced a dimensionless method for the entransy dissipation and defined an entransy
dissipation number which can serve as the heat exchanger performance evaluation
criterion. Based on the concept of entransy dissipation, an equivalent thermal resistance of
heat exchanger was defined which is consistent with the exchanger effectiveness (Guo et
al., 2010). Cheng and Liang (2011) defined the entransy flux and entransy function for the
thermal radiation in enclosures with opaque surfaces, and the minimum principle of
radiative entransy loss was established. Chen et al. (2011) proposed an entransy
dissipation rate minimization approach for the disc cooling system and the influence of
various system parameters on the entransy dissipation rate of the cooling system has been
investigated.
Although the concepts of entransy and entransy dissipation have been applied to heat
transfer and demonstrate some advantages in comparison with the entropy and entropy
generation, how to define these concepts from the thermodynamic point of view is still an
open question. In this section we place the concepts of entransy and entransy dissipation on
the solid thermodynamic basis.
1.2 Carnot’s theorem in terms of entransy

We start with the Carnot cycle. In this cycle, the heat engine absorbs heat Q1 from the hot
reservoir with the temperature T1 (absolute temperature is always assumed in the following
discussion), converts part of heat to work W and discards the rest of the heat to the cold
reservoir with the temperature T2 . For this cycle, Carnot’s theorem states that (Kondepudi
& Prigogine, 1998)
Q1 / T1 ≤ Q2 / T2 (1)
where the equality and inequality correspond to the reversible and irreversible heat engines,
respectively. The efficiency of a reversible engine is defined as
T2
η =1− (2)
T1
Carnot’s theorem dictates that reversible engines have the maximum efficiency (Kondepudi
& Prigogine, 1998).
Equivalently, Inequality (1) can be rewritten as follows
Q1 (T1 − T2 ) ≥ (Q1 − Q2 )T1 (3)
Inspired by Inequality (3), we define E = Q1 (T1 − T2 ) as the entransy gained by the heat
engine from the hot reservoir in the Carnot cycle. From this definition, one can see that the
larger the amount of heat Q1 and the temperature difference between the hot and cold
Entransy Dissipation Theory and Its Application in Heat Transfer 249
reservoir, the greater the entransy obtained by the heat engine in the Carnot cycle.
Obviously, the larger entransy means higher ability for heat engine to perform work. Guo et
al. (2007) shown that the entransy can describe the heat conduction ability. Therefore, we
may say that the entransy defined here quantifies the energy transfer ability including the
ability to deliver work and transfer heat.
To obtain the efficiency of a reversible heat engine, firstly we know that W = Q1 − Q2 , which
is the statement of the law of conservation of energy. We define EW = (Q1 − Q2 )T1 = WT1 and
regard it as the entransy consumed by the heat engine for delivering work W to the
system’s exterior. For the reversible heat engine, Inequality (3) reduces to:
Q1 (T1 − T2 ) = (Q1 − Q2 )T1 = WT1 (4)
which says that the entransy gained by the reversible heat engine in the Carnot’s cycle is
completely converted to work W. While for the irreversible heat engine, Inequality (3) becomes
Q1 (T1 − T2 ) > (Q1 − Q2 )T1 (5)
which tells us that only part of the entransy obtained by the heat engine in the Carnot’s cycle
is utilized to deliver work W, the rest is consumed by the irreversibility in the heat engine.
In order to quantify the irreversibility occurring in the heat engine, we define the entransy
generation in parallel with the entropy generation in the following way:
Eg = (Q1 − Q2 )T1 − Q1 (T1 − T2 ) (6)
While Ediss = −Eg is called the entransy dissipation which represents the entransy consumed
by the irreversibility in heat engine. Since the reversible heat engine converts the total
entransy gained during the Carnot cycle to perform work, we may define the efficiency of
the heat engine in terms of entransy as follows
WT1 E
ηE = = W (7)
Q1 (T1 − T2 ) E
Notice that the reversible heat engine achieves the maximum value of ηE which is equal to
100%. This is the statement of Carnot’s theorem in terms of entransy.
1.3 The second law of thermodynamics

Carnot’s theorem has played a pivotal role on the development of the classical
thermodynamics. By generalizing Inequality (1) to an arbitrary cycle, Clausius introduced
the concept of entropy which is a physical quantity as fundamental and universal as energy.
In this section, we aim to generalize Inequality (3) to an arbitrary cycle. Firstly, Inequality (3)
can be rewritten as follows:
Q1ΔT + T1ΔQ ≥ 0 (8)
where ΔT = T1 − T2 , ΔQ = −(Q1 − Q2 ) . Note that since the system delivers energy to its
exterior, therefore ΔQ < 0 .
An arbitrary cycle can be decomposed into a group of Carnot’s cycles denoted as
C i (i = 1, 2," n) (Kondepudi & Prigogine, 1998). Applying Inequality (8) to the i-th Carnot’s
cycle yields:
Qa ,i ΔTi + ΔQiTh ,i ≥ 0 (9)
where Qa ,i is the heat absorbed from the hot reservoir with the temperature Th ,i ,
ΔQi = −(Qa , i − Qr ,i ) ( Qr ,i is the heat discarded into the cold reservoir), ΔTi = Th ,i − Tc ,i ( Tc ,i
is the temperature of the cold reservoir). If the number of Carnot’s cycles under
consideration tends to the infinity, the temperature difference between the hot and cold
reservoirs approaches to infinitesimal, Inequality (9) becomes
v∫ QdT + Tδ Q ≥ 0 (10)
For a reversible cycle, there is no heat conduction, Inequality (10) reduces to
v∫ QdT + Tδ Q = 0
Equation (11) suggests that we can define a quantity E called the entransy as follows
dE = QdT + Tδ Q (11)
which only depends on the initial and final states of a reversible process. Thus the entransy
is a state variable. If EA and EB are values of this variable in the initial state A and final
state B, respectively, we have
B
EB − EA = ∫ QdT + Tδ Q (12)
A
If the temperature remains fixed, it follows from Eq. (12) that for a reversible flow of heat
δ Q , the change in entransy is Tδ Q , while the change in entropy is δ Q / T . For this case we
get the following relationship between the entransy and entropy,
dE = T 2 dS (13)
Consider an irreversible process I which starts from the equilibrium state A and ends at the
equilibrium state B. In order to form a cycle, we add a reversible process R from the state B
to state A. Then from Inequality (10), we have
BI BR
∫A QdT + Tδ Q >∫A
I R
QdT + Tδ Q (14)
The subscripts R and I represent the reversible and irreversible processes, respectively. The
application of Eq. (12) on Inequality (14) yields
BI
∫A QdT + Tδ Q >EB − EA
I
(15)
Thus in parallel with the entropy generation we can define the entransy generation Eg of
the irreversible process as follows
BI
Eg = (EB − EA ) − ∫ QdT + Tδ Q < 0 (16)
AI
Note that for the reversible process Eg = 0 . We define Ediss = −Eg as the entransy
dissipation. Therefore, the entransy dissipation quantifies the entransy consumed by the
irreversibility in the irreversible process. For an irreversible process, the second law of
thermodynamics states that the entransy generation is always negative or the irreversible
process always decreases the system’s ability to do work and transfer heat. From the
entransy’s definition (12), it is evident that the entransy is an extensive quantity.
Subsequently, Inequality (16) is also valid under the local equilibrium assumption.
We may express the system’s change in entransy as a sum of two parts
dE = deE + di E , and di E ≤ 0 (17)
In which deE is the entransy change due to exchange of matter and energy with the exterior of
the system and di E is the entransy generation produced by the irreversible processes
occurring in the system. For the closed systems that exchange energy, but not matter, we have
de E = QdT + Tδ Q and di E ≤ 0 (18)
For open systems that exchange both matter and energy:
de E = QdT + Tδ Q + ( deE)matter and di E ≤ 0 (19)
Finally, for the isolated systems, we have
de E = 0 , and di E ≤ 0 (20)
Therefore, in the isolated systems the entransy never increases, namely the energy transfer
ability can not increase.
1.4 Entransy dissipation due to heat conduction

Now let us consider the heat conduction process discussed by Kondepudi & Prigogine
(1998). The system under consideration is an isolated system and consists of two parts, each
having a well-defined temperature. Let the temperatures of two parts be T1 and T2
( T1 > T2 ), respectively. δ Q is the amount of heat flow from the hotter part to colder part in a
time period dt. Since this isolated system does not exchange entransy with its exterior,
de E = 0 . Assume the volume of each part is constant, thus dW = 0 . The energy change in
each part is solely due to the flow of heat: dU i = δ Qi (i = 1, 2) . In accordance with the first
law, the heat gained by one part is equal to the heat lost by the other. Therefore,
−δ Q1 = δ Q2 = δ Q . The total change in entransy di E of the system is the sum of the changes
of entransy in each part due to the flow of heat
di E = −T1δ Q + T2δ Q = −(T1 − T2 )δ Q (21)
In terms of the rate of flow of heat δ Q / δ t , the rate of entransy generation can be written as
follows
di E δQ (22)
= −(T1 − T2 )
dt δt
The rate of heat flow JQ = δ Q / δ t is given by the Fourier law of heat conduction
JQ = k(T1 − T2 ) , in which k is the coefficient of heat conductivity, therefore
di E
= − k(T1 − T2 )2 ≤ 0 (23)
dt
Accordingly, the rate of entransy dissipation is written as
di Ediss
= k(T1 − T2 )2 ≥ 0 (24)
dt
This equation is consistent with the expression of the entransy dissipation function obtained
by Guo et al. (2007) from the heat conduction equation. Due to the flow of heat from the hot
part to cold part, the temperatures of both parts eventually become equal, and the entransy
dissipation rate tends to zero. Then the system reaches the equilibrium state. Therefore, the
entransy dissipation rate must vanish at the state of equilibrium, which is called as the
principle of minimum entransy dissipation rate. This is the counterpart of the principle of
the minimum entropy production. Generally, for three-dimensional steady heat conduction
without heat source, the entransy dissipation rate is expressed as
di Ediss 1
= ∫ k(∇T )2 dV (25)
dt V
2
where V is the volume of the heat conduction medium. The principle of minimum entransy
dissipation rate is mathematically formulated as
1
δ ∫ k(∇T )2 dV = 0 (26)
V
2
which is consistent with the least entransy dissipation principle established by Guo et al.
(2007). By the variational method, Eq. (26) is equivalent to the following Euler-Lagrange
equation:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
∂ ⎜ ∂F ⎟ ∂ ⎜⎜ ∂F ⎟⎟ ∂ ⎜ ∂F ⎟ (27)
⎜ ⎟+ + ⎜ ⎟=0
∂x ⎜ ∂( ∂T ) ⎟ ∂y ⎜ ∂( ∂T ) ⎟ ∂z ⎜ ∂( ∂T ) ⎟
⎜ ⎟ ⎜ ∂y ⎟ ⎜ ⎟
⎝ ∂x ⎠ ⎝ ⎠ ⎝ ∂z ⎠
with
1
F= k ( ∇ T )2 (28)
2
Substituting Eq. (28) into Eq. (27) gives
∂ 2T ∂ 2T ∂ 2T (29)
+ + =0
∂x 2 ∂y 2 ∂z 2
which is exactly the governing equation of the steady heat conduction without heat sources
based on the Fourier law. While Bertola and Cafaro (2008) found that the principle of
minimum entropy production can not lead to the governing equation of the steady Fourier
heat conduction. In this point, the principle of minimum entransy dissipation rate
demonstrates an obvious advantage.
1.5 Concluding remarks

Note that the concept of entropy can be replaced with entransy for describing the second
law of thermodynamics. In comparison with entropy, the entransy has a definite physical
meaning and the principle of minimum entransy dissipation rate avoids the contradiction
between the principle of minimum entropy production and the classical Fourier heat
conduction theory. Therefore, we anticipate that the concepts of entransy and entransy
dissipation may help us to gain more profound insight on thermodynamics in particular
and on science in general.
2. The application of entransy dissipation theory in heat convection

In this section, the entransy dissipation theory developed in Section 1 are applied to analyze
the heat convection.
2.1 Introduction
The entropy and entropy generation can help us to deeply understand the momentum and
heat transfer (Bejan, 1982; Herwig, 2010). Bejan (1982) realized that in order to improve the
performance of the heat transfer enhancement or thermal insulation equipments, one need
to reduce the entropy generation rate. Similarly, according to the definition of entransy
dissipation given in Section 1.3, it is required to minimize the entransy dissipation rate for
achieving the best heat transfer enhancement and thermal insulation. Therefore, it is of great
value to derive the expression of the local rate of entransy dissipation rate for heat
convection. Xu et al. (2009) have managed to get an expression of the local rate of entransy
dissipation rate for heat convection. However, it lacks the theoretical basis in this derivation.
In Section 2.2 with the help of the second law of thermodynamics in terms of entransy and
entransy dissipation established in Section 1.3 we will make the derivation more rigorously.
2.2 Local thermodynamic entransy dissipation in heat convection

The infinitesimal element as shown in Fig. 1 is an open thermodynamic system, where
[ vx , vy ]T is the velocity, [ qx , q y ]T is the heat flux. For this system, we assume that the
thermodynamic state is irrelevant with the position, but relevant with time. By the second
law of thermodynamics for the open system expressed as Eqs. (17) and (19) the rate of local
thermodynamic entransy dissipation generation E g per unit volume in the infinitesimal
element is expressed as
∂T ∂T
E g dxdy = (T + dx − T )qx dy + (T + dy − T )q y dx
∂x ∂y
∂q ∂qy
+ T (qx − qx − x dx )dy + T (q y − q y − dy )dx
∂x ∂y
∂e ∂v ∂ρ
+ ( e + dx )( vx + x dx )( ρ + dx )dy (30)
∂x ∂x ∂x
∂e ∂vy ∂ρ
+ ( e + dy )( vy + dy )( ρ + dy )dx
∂y ∂y ∂y
∂( ρ e )
− ρ vx edy − ρ vy edx + dxdy
∂t
∂q y ∂vy
qy + dy vy + dy
∂y ∂y
y + dy ∂vx
vx + dx
vx ∂x
qx ∂qx
qx + dx
∂x
y qy vy
x x + dx
Fig. 1. Two-dimensional infinitesimal element for heat convection
where e is the specific entransy, ρ is the density of fluid. The first four terms in the right side
of Eq. (30) account for the entransy exchanged with the environment through the element
boundary, the following four terms represent the entransy transported by convection, the last
term stands for the variation of the entransy with respect to time. Rearranging Eq. (30) yields,
∂T ∂T ∂q ∂q y ∂e ∂e ∂e
E g = (qx + qy ) − T( x + ) + ρ ( + vx + vy )
∂x ∂y ∂x ∂y ∂t ∂x ∂y
(31)
∂ρ ∂v ∂vy ∂ρ ∂ρ
+ e[ + ρ( x + ) + vx + vy ]
∂t ∂x ∂y ∂x ∂y
Making use of the following mass balance equation,
∂ρ ∂ρ ∂ρ ∂v ∂vy
+ vx + vy + ρ( x + )=0 (32)
∂t ∂x ∂y ∂x ∂y
where ρ is the density of fluid, Eq. (31) is simplified to
De
E g = q ⋅ ∇T − T ∇ ⋅ q + ρ (33)
Dt
where D / DT = ∂ / ∂t + vx ∂ / ∂x + vy ∂ / ∂y is the material derivative. By the following
canonical relation of thermodynamics,
De Du PT Dρ (34)
ρ = −T ρ +
Dt Dt ρ Dt
where u is the specific internal energy, P is the pressure,
and the energy balance equation for heat convection
Du (35)
ρ = −∇ ⋅ q − P(∇ ⋅ v) + μΦ
Dt
where μ is the viscosity, Φ is the dissipation function and can be expressed as
∂vx 2 ∂vy 2 ∂v ∂vy 2

Φ = 2[( ) +( ) ]+( x + ) (36)
∂x ∂y ∂y ∂x
By Eqs. (34) and (35), Eq. (33) becomes
E g = q ⋅ ∇T − μT Φ (37)
The application of the Fourier law, q = − k ⋅ ∇T , on Eq. (37) yields,
E diss = −E g = k(∇T )2 + μT Φ (38)
where k is the thermal conductivity. Thus we obtained the formula for calculating the local
entransy dissipation rate in heat convection. It is evident that the expression of the local
entransy dissipation rate in the three-dimensional case shares the same form as Eq. (38).
From Eq. (38), one can see that if the temperature and velocity fields in heat conduction
problems are known, the local entransy dissipation rate can be obtained exactly.
Unfortunately, it is seldom to obtain the exact solution of heat convection problems. In
many cases only the heat transfer correlations are available. Therefore it is of importance to
develop expressions of the entransy dissipation rate by the knowledge of heat tranfer
correlations for heat convection. In this section we attempt to establish such a formula for
internal flow in a duct with arbitrary cross section. Consider a flow passage enclosed by two
cross sections with distance dx. The temperatures of the duct wall and fluid inside the duct
are T + ΔT and T, respectively. The heat transfer rate from the duct wall to the fluid per unit
length is q . The fluid inside the duct flows with friction in x-direction, thus the pressure
gradient dP / dx ≤ 0 . In this heat transfer process, assume there are two irreversibilities. One
is the heat conduction from duct wall to the fluid. The other irreversibility is flow friction.
As discussed in Section 1.4 the entransy dissipation rate per unit length (∂E diss / ∂x )T
induced by the heat conduction is
⎛ ∂E diss ⎞
⎜ ⎟ = q ΔT (39)
⎝ ∂x ⎠T
Assume the fluid is incompressible. The entransy dissipation rate per unit length
(∂E diss / ∂x ) f induced by the flow friction is expressed as
⎛ ∂E diss ⎞ ⎛ dP ⎞
mT
⎜ ⎟ =− (40)
⎝ ∂x ⎠ f ρ ⎜⎝ dx ⎟⎠
Where ρ is the density of the fluid, m is the mass flow rate. Thus the total entransy
dissipation rate per unit length ∂E diss / ∂x is
∂E diss ⎛ dP ⎞
mT
= q ΔT − (41)
∂x ρ ⎜⎝ dx ⎟⎠
The Stanton number correlation reads
q
St = (42)
pΔTc p R
with
R=m
/A (43)
where A is the cross sectional area of the duct, c p is the specific heat at constant pressure, p
is the wetted perimeter. And the friction factor correlation is
ρ D dP
f =− (44)
2 R 2 dx
Where D is the hydraulic diameter. By Eqs. (42) and (44), Eq. (41) arrives at
∂E diss q 2 D 2 mT f

(45)
= + 2
∂x p St
4mc ρ DA2
Eq. (45) is thus the formula for calculating the derivative of the entransy dissipation rate
with respect to the position variable for the forced convection in duct flow. Note that the
higher Standon number means the smaller entransy dissipation rate, while a high friction
factor leads to the increase of the entransy dissipation rate.
2.3 Entransy dissipation for external flow

Consider an external steady flow that a solid body with arbitrary shape is suspended in a
uniform stream with velocity U ∞ and absolute temperature T∞ . For the convenience of
analysis, the stream tube with cylindrical shape and large enough volume around the solid
body as shown in Fig.2 is taken as the thermodynamic system. The surface temperature of
the solid body is Tw . In the regions near the stream tube surface and the external part, the
influence of the solid body on the flow is neglected. There is no heat transfer between the
solid surface and its external environment. The irreversible dissipation is only caused by the
heat conduction and flow friction. For the considered open thermodynamic system, the
mass and energy conservations are expressed as,
in = m
m out = m
(46)
in + ∫∫ qds
mh − mh
out = 0 (47)
A
where the subscripts “in” and “out” respectively indicate the inlet and outlet, h is the
specific enthalpy, q is the local heat transfer rate between the solid body and the stream, A
is the surface area of the solid body, m is the mass flow rate. For the heat conduction across
the nonzero temperature difference Tw − T∞ , the entransy dissipation rate E diss ,T is written
as
E diss ,T = ∫∫ q(Tw − T∞ )dA

A
where A is the surface area of the solid body. For the flow friction, the entransy dissipation
rate E diss , P is
∞
mT
E diss , P = − ( Pout − Pin )
ρ∞
where ρ∞ is the density of the fluid, P is the pressure. Here we assume the fluid is
incompressible. Therefore, the total entransy dissipation rate is written as
∞
mT
E diss = ∫∫ q(Tw − T∞ )dA − ( Pout − Pin ) (48)
A
ρ∞
Since
= Atube ρ∞U ∞
m
and from the force balance on the control volume, we have
FD = Atube ( Pin − Pout )
where FD is the drag force, Atube is the cross sectional area of the cylindrical tube.
Therefore,
E diss = ∫∫ q(Tw − T∞ )dA + T∞ FDU ∞ (49)

A
If the body temperature is uniform, then the temperature difference Tw − T∞ is constant,

thus Eq. (49) becomes
E diss = (Tw − T∞ )∫∫ qdA

+ T∞ FDU∞ (50)
A
Since
∫∫ qdA
= Q = hA(Tw − T∞ )
A
where h is the average heat transfer coefficient based on A, Eq. (50) arrives at
E diss = hA(Tw − T∞ )2 + T∞ FDU∞ (51)
Note that when the body-ambient temperature difference is fixed, the only way to minimize
the heat transfer contribution to the entransy dissipation rate is by decreasing the thermal
conductance hA .
The other extreme is that the heat transfer rate q is uniform around the solid body. In this
case Eq. (50) becomes
1
E diss = q 2 ∫∫ dA + T∞ FDU∞ (52)
A
h
where h is the local heat transfer coefficient. Note that in order to reduce the heat transfer
entransy dissipation rate, we must increase the heat transfer coefficient. In other words, the
enhancement of the heat transfer from the solid body to the external fluid can decrease the
entransy dissipation rate.
Atube U ∞ T∞
m FD
inlet outlet
Fig. 2. Schematic diagram for external flow

In this section we have derived the expressions of the entransy dissipation rate for heat
convection. These results lay a foundation for the application of entransy dissipation theory
in the heat convection problems.
3. From the extremum principle of entransy dissipation to balance equations

of fluid mechanics
In this section, our work (Xu & Cheng 2010) is reviewed. In this work, the extremum
principle of entransy dissipation is developed by using the entransy dissipation theory
developed in Section 1. We further show that this principle is compatible with the Fourier
law and Newton’s law of viscosity.
3.1 Introduction
The modeling and optimization design of heat exchanger based on the second law of
thermodynamics has attracted lots of attention in past decades. In the early 1951,
McClintock (1951) recognized that the irreversibility in heat exchanger could not be
neglected in heat exchanger design. Bejan’s creative work led to the establishment of the
minimization entropy generation approach which is widely applied in optimization designs
of thermal equipments (Bejan, 1982, 1988, 1995, 1996). However some paradoxes and
contradiction appear when the minimization entropy generation approach is applied to the
heat transfer problems (Hesselgreaves, 2000; Bertola & Cafaro, 2008). Although a non-
dimensionalisation method for the entropy generation in heat exchanger was proposed by
Hesselgreaves (2000) that avoids the ‘entropy generation paradoxes’, it has induced other
contradictions.
In fact the thermodynamic basis of the minimization entropy generation approach is the
minimum entropy production principle which reads “A steady state has the minimum rate
of entropy production with respect to other possible states with the same boundary
condition”. Actually even for this principle itself, there are lots of controversies (Bertola &
Cafaro, 2008; Jaynes, 1980; Mamedov, 2003; Landauer 1975; Ziman, 1956). Landauer (1975)
found that the principle of the minimum entropy production is not valid even for some
simple thermodynamic systems such as electric resistances. A variational principle
developed by Ziman (1956) demonstrates that the entropy production assumes the
maximum value in the steady state for transport process. Bertola and Cafaro (2008) found
that generally the principle of minimum entropy production does not agree with the
conventional Fourier law as well as mass, energy and momentum balance equations in
continuum mechanics. In order to remedy the defects of the principle of the minimum
entropy production, Guo et al. (2007) developed the extremum principle of entransy
dissipation in which only the irreversibility due to heat conduction under finite temperature
difference was considered. In Section 3.2, we review the work which extends the current
extremum principle of entransy dissipation by taking into account the influence of fluid
friction and demonstrate the advantage of this principle over the principle of minimum
entropy production in dealing with fluid flow problems (Xu & Cheng, 2010).
3.2 Extremum principle over entransy dissipation and balance equations of fluid
mechanics
In order to seek some clues for establishing the extremum principle of entransy dissipation,
we first consider natural convection problems, such as the Rayleigh-Benard convection. In
this problem, when the temperature difference increases and reaches some critical point, the
stationary fluid loses its stability and begins to move. In terms of the entransy, the Rayleigh-
Benard convection can be viewed as the entransy related to conduction is converted to the
entransy for driving the movement of fluid. The entransy consumed by performing work
can be quantified as − μΦδ T according to the derivation in Section 2.2. The entransy
dissipation associated with the heat conduction in the Rayleigh-Benard convection is
quantified by −q ⋅ ∇T . Therefore, we obtain the following extremum principle of entransy
dissipation
∫ [ −q ⋅ δ (∇T ) −μΦδ T ]dV = 0 (53)

V
The application of the Fourier law on Eq. (53) gives
1
δ ∫ [ k(∇T )2 − μΦT ]dV = 0 (54)
V
2
This is the principle of minimum entransy dissipation. Note that in this principle, the
entransy dissipation comes from two irreversibilities: one is the heat conduction under finite
temperature difference, another is the work against the fluid friction.
According to fluid mechanics the governing equation of imcompressible fluid under steady
state with constant shear viscosity and thermal conductivity reads
∂vi
=0 (55)
∂xi
∂vi ∂p ∂ ∂vi ∂v j
ρvj + −μ ( + )=0 (56)
∂x j ∂xi ∂x j ∂x j ∂xi
∂T ∂ 2T 1 ∂v ∂v j ∂vi ∂v j
ρcv v j − k 2 = μ( i + )( + ) (57)
∂x j ∂x j 2 ∂x j ∂xi ∂x j ∂xi
where x = ( x1 , x2 , x3 )T is the position vector, v = ( v1 , v2 , v3 )T is the velocity field, p is the

pressure. The density of entropy production is given by:
k ∂T ∂T 1 μ ∂vi ∂v j 2 ∂vk ∂v ∂v j 2 ∂vk

σ= 2
+ ( + − δ ij )( i + − δ ij ) (58)
T ∂xi ∂xi 2 T ∂x j ∂xi 3 ∂x k ∂x j ∂xi 3 ∂xk
Now we consider the one-dimensional steady imcompressible shear flow, v = (0, v( x ),0)T ,
the continuity equation, momentum balance equation and energy conservative equation
reduce to
∂v (59)
=0
∂y
d2 v ∂p
(60)
μ 2
=−
dx ∂x
d 2T μ dv 2 (61)
+ ( ) =0
dx 2 k dx
From Eq. (59), one can see that the continuity equation is satisfied automatically for the flow
under consideration. By Eq.(54), we define the following function
1 ∂T ∂T μT ∂vi ∂v j ∂vi ∂v j ∂v (62)

F= k − ( + )( + ) − 2 λT i
2 ∂xi ∂xi 2 ∂x j ∂xi ∂x j ∂xi ∂xi
In order to satisfy the continuity equation, the Lagrange multiplier λ is introduced in Eq.
(62). From the variational method, Eq. (54) is equivalent to the following Euler-Lagrange
equations
⎛ ⎞
∂ ⎜ ∂F ⎟ ∂F
⎜ ⎟− =0 (63)
∂xi ⎜ ∂( ∂T ) ⎟ ∂T
⎜ ∂x ⎟
⎝ i ⎠
⎛ ⎞
⎜ ⎟
∂ ⎜ ∂F ⎟ ∂F
− =0 (64)
∂xi ⎜ ∂v j ⎟ ∂v j
⎜ ∂ ( ) ⎟
⎝ ∂xi ⎠
Substituting Eq. (62) into Eqs. (63) and (64) yields
∂ 2T μ ∂vi ∂v j ∂vi ∂v j ∂v (65)

k 2
+ ( + )( + ) + 2λ i = 0
∂xi 2 ∂x j ∂xi ∂x j ∂xi ∂xi
∂T ∂vi ∂v j ∂ 2 vi ∂2v j ∂T ∂λ
μ ( + ) + μT ( + ) = −λ − T (66)
∂xi ∂x j ∂xi ∂xi ∂x j ∂xi ∂x j ∂xi ∂xi
The application of the continuity equation (59) on Eq. (65) gives
∂ 2T μ ∂vi ∂v j ∂vi ∂v j
k + ( + )( + )=0 (67)
∂xi2 2 ∂x j ∂xi ∂x j ∂xi
For the one-dimensional steady shear flow, Eqs. (65) and (66) become
∂T ∂v ∂2v ∂T ∂λ (69)
μ + μT 2 = −λ − T
∂x ∂x ∂x ∂x ∂x
2
∂ 2T ⎛ ∂v ⎞ (70)
k 2
+ μ⎜ ⎟ = 0
∂x ⎝ ∂x ⎠
By setting λ = p = − μ dv / dx , Eqs. (69) and (70) reduce to Eqs. (60) and (61), respectively.
Therefore, the extremum principle of entransy dissipation yields the same governing
equation of the one-dimensional steady shear flow as given in fluid mechanics. Since Eqs.
(60) and (61) are determined by the Fourier law and Newton’s law of viscosity, we may
say that the extremum principle of entransy dissipation is compatible with these two
classical laws. While the principle of the minimum entropy production contradicts with
the standard balance equations of the one-dimensional steady shear flow (Bertola &
Cafaro, 2008). In this point, the extremum principle of entransy dissipation demonstrates
the obvious advantage.

In summary, a new type of extremum principle of entransy dissipation is established by
taking into account the viscosity resistance of the fluid. The analysis shows that this
extremum principle of entransy dissipation gives rise to the standard governing equations
of one-dimensional steady shear flow in fluid mechanics. Therefore, the entremum
principle of entransy dissipation is consistent with the Fourier law and Newton’s law of
viscosity.
4. The application of entransy dissipation theory in heat exchanger design

In this section the expressions of the entransy dissipation related to the irreversibility in heat
exchanger are developed. The entransy dissipation number which can be employed to
evaluate the performance of heat exchangers is defined by a non-dimensionalisation method
for the entransy dissipation. The optimization of heat exchanger design is analyzed.
4.1 Introduction
With the sharp decline of fossil fuels such as petroleum and coal, to use energy efficiently is
one of effective ways to face the increasing energy demand. Heat exchanger as an important
device in thermal system is widely applied in power engineering, petroleum refineries,
chemical industries, and so on. Hence, it is of great importance to develop technologies
which enable us to reduce the unnecessary energy dissipation and improve the performance
of heat exchanger.
The evaluation criteria for heat exchanger performance are generally classified into two
groups: the first is based on the first law of thermodynamics; the second is based on the
combination of the first and second law of thermodynamics. The heat transfer in heat
exchangers usually involves the heat conduction under finite temperature difference, the
fluid friction under finite pressure drop and fluid mixing. These processes are characterized
as irreversible non-equilibrium thermodynamic processes. Hence, in recent decades the
study of the second group has attracted a lot of attention (Yilmaz et al., 2001). Inspired by
the principle of the minimum entropy production advanced by Prigogine (1976), Bejan
(1982, 1995) developed the entropy generation minimization (EGM) approach to heat
exchanger optimization design. In this approach Bejan (1982) took into account two types of
the irreversibilities in heat exchanger, namely the heat conduction under the stream-to-
stream temperature difference and the frictional pressure drop that accompanies the
circulation of fluid through the apparatus. Therefore, the total entropy production rate
denoted by S gen is the sum of entropy productions associated with heat conduction and
fluid friction. However, among all the variational principles in thermodynamics, Prigogine’s
minimum entropy generation principle is still the most debated one as mentioned in
Section 3.1. Accordingly, the entropy generation minimization approach, widely applied to
modeling and optimization of thermal systems that owe their thermodynamic imperfection
to heat transfer, mass transfer, and fluid flow irreversibilities, demonstrates some
inconsistencies and paradoxes in applications of heat exchanger designs (Hesselgreaves,
2000).
Recently based on the concept of entransy Guo et al. (2010), the heat transfer efficiency and
equivalent thermal resistance in a heat exchanger were defined. Wang et al. (2009) derived
an entransy transfer equation describing the entransy transfer processes of a multi-
component viscous fluid subjected to heat transfer by conduction and convection, mass
diffusion and chemical reactions. Chen and Ren (2008) defined a ratio of temperature
difference to heat flux as the generalized thermal resistance of convective heat transfer
processes, and developed the minimum thermal resistance theory for convective heat
transfer optimization, it was found that the minimum thermal resistance principle is
equivalent to the entransy dissipation extremum principle. Xia et al. (2009) studied the
optimum parameter distributions in two-fluid flow heat exchanger by using optimal control
theory under the fixed heat load condition and taking the entransy dissipation minimization
as the optimization objective. Liu et al. (2009) investigated the applicability of the extremum
principles of entropy generation and entransy dissipation for heat exchanger optimization,
and found that the former is better for the heat exchanger optimization when it works in the
Brayton cycle, while the latter gives better results when heat exchanger is only for the
purpose of heating and cooling. Recently the entransy dissipation extremum principle was
extended to the radiative heat transfer by Wu and Liang (2008). Guo et al. (2009) introduced
a non-dimensionalisation method for the entransy dissipation in heat exchanger and an
entransy dissipation number which can be used to evaluate the heat exchanger performance
was defined.
However, all the work mentioned above is based on the entransy defined by Guo et al.
(2007) which is only related to the heat conduction in heat exchanger. In Section 1 we have
presented a new definition of entransy which not only describes the ability to transfer heat,
but also the ability to perform work. Subsequently, the entransy dissipation can quantify not
only the irreversibility associated with the heat conduction, but also the irreversibility
induced by fluid friction. In this section, we first derive the expressions of the entransy
dissipation induced by heat conduction, fluid friction and fluid mixing in heat exchanger.
Then a non-dimensionalisation of the total entransy dissipation will be introduced.
4.2 Entransy dissipation in heat exchanger

According to the definition of entransy, we have the following relationship between the
entransy and enthalpy
de = Tdh (71)
where e and h are the specific entransy and enthalpy, respectively. Assume that the heat
exchanger under consideration is an adiabatic open system. By making use of Eq. (71) and
the following relationship
h = mc
pT (72)
the entransy dissipation rate induced by heat conduction in the heat exchanger is written as
i i
E 1 = ∫ ( mTdh
)1 = ∫ mc (
pTdT )1 = 21 (mc
p )1 (T1,i 2 − T1, o 2 ) (73)
o o
where E 1 is the entransy dissipation rate of the fluid 1, c p is the specific heat of fluid at
constant pressure and assumed to be a constant, the subscript 1 denotes the fluid 1, the
subscripts i and o denote the inlet and outlet, respectively. Similarly, for the fluid 2 we have
i i
E 2 = ∫ ( mTdh
)2 = ∫ mc (
pTdT )2 = 21 (mc
p )2 (T2,i 2 − T2 , o 2 ) (74)
o o
Therefore, the total entransy dissipation rate induced by heat conduction in the heat
exchanger is expressed as
1 1
E T = E 1 + E 2 = (mc
p )1 (T1,i 2 − T1, o 2 ) + (mc
p )2 (T2, i 2 − T2 , o 2 ) (75)
2 2
Now we consider the entransy dissipation rate caused by flow friction in heat exchangers.
Assume that the fluid flow in the heat exchanger is driven by the finite pressure drop
between the inlet and outlet, and the flow is stationary and adiabatic. Then from the
definition of entransy, we have
de = −Tdp / ρ (76)
where p is the pressure. Then the entransy dissipation rate induced by flow friction is
written as
o o
E p = ∫ mdg
0 = − ∫ mT
/ ρ dp (77)
i i
If the fluid is the ideal gas, applying its state equation in

Eq. (77) gives
o
E p = − ∫ mT
2 R / pdp (78)
i
where R is the ideal gas constant. If replacing the temperature

in Eq. (78) with the logarithmic mean temperature, we obtain
o
E p = −Tlm 2 ∫ mR
/ pdp (79)
i
where Tlm = (To − Ti ) /(ln To − ln Ti ) . Similarly, for

the incompressible fluid we have
lm Δp
mT
E p = (80)
ρ
where Δp = pi − po . If both fluids in the heat exchanger
are idea gases, the entransy dissipation rate induced by
fluid friction is written as
T1, o − T1,i p1, o T2, o − T2 ,i p2, o

E p = −m
1( )2 R ln −m
2( )2 R ln (81)
ln T1, o − ln T1,i p1,i ln T2, o − ln T2 ,i p2 , i
When both fluids in the heat exchanger are incompressible, we obtain
m Δp T1, o − T1,i 2 Δp2

m T2 , o − T2, i
E p = 1 1 + (82)
ρ1 ln T1, o − ln T1,i ρ2 ln T2 , o − ln T2 ,i
When the fluid 1 is idea gas, fluid 2 is impressible, we have
T1, o − T1,i p1, o 2 Δp2

m T2, o − T2, i
E p = −m
1( )2 R ln + (83)
ln T1, o − ln T1,i p1,i ρ2 ln T2, o − ln T2, i
It is well known that when two or more streams of dissimilar fluids are mixing, the
irreversibility occurs, which is detrimental to the performance of heat exchanger. In the
cross flow exchanger, the mixing of thermal dissimilar fluids (with different temperatures)
frequently happens. In the following we focus on the entransy dissipation rate induced by
this type of mixing. Assume that n streams of thermal dissimilar fluids enter into the heat
exchanger from its inlet. Their temperatures are denoted as Tj ,i ( j = 1, 2," , n) . In the flow
process from the inlet to outlet, the streams of thermal dissimilar fluids are adequately
mixing. At the outlet, they have the same temperature To . Assume that the flow is
stationary and the mixing process is adiabatic. Then making use of the law of energy
conservation, the entropy balance equation and Eq. (71), we obtain
n
o = ∑m
m j ,i (84)
j =1
n n
o eo − ∑ ( me
E m = m ) j ,i = ∑ m
j Δe j (85)
j =1 j =1
The similar derivation as leading to Eq. (72) yields

n
1
E m = ∑ (mc p ) j (Tj ,i 2 − To 2 ) (86)
j =1 2
Finally, the total entransy dissipation rate caused by heat conduction, fluid friction and fluid
mixing is expressed as
E = E T + E p + E m (87)
4.3 Entransy dissipation number

In this section our work (Guo et al. 2009, Li et al. 2011) is reviewed. In this work, a
dimensionless method of the entransy dissipation of the heat exchanger is established.
Subsequently a quantity called the entransy dissipation number of heat exchanger is
defined. Guo et al. (2009) examined the validity of the entransy dissipation number as the
performance evaluation criterion of heat exchanger. Furthermore, for the water-water
balanced counter-flow heat exchanger Li et al. (2011) took as the minimum overall entransy
dissipation number as an objective function, under certain assumptions it was proved that
there is a corresponding optimum in duct aspect ratio or mass velocity since the variation in
the duct aspect ratio or mass velocity has opposing effects on the two types of entransy
dissipations caused by heat conduction under finite temperature difference and the flow
friction under finite pressure drop, respectively. We also develop analytically expressions
for the optimal duct aspect ratio and mass velocity of a heat exchanger that are useful for
design optimization.
Now we consider a water-water balanced counter-flow heat exchanger. Assume that both
the hot and cold fluids are imcompressible. The inlet temperatures and pressures of the hot
and cold fluids are denoted as T1,i , P1,i and T2 ,i , P2 ,i respectively. Similarly the outlet
temperatures and pressures are T1,o , P1,o and T2,o , P2,o . For the balanced heat exchanger,
the heat capacity rate ratio satisfies condition C ∗ = ( mc
)2 / ( mc
)1 = 1 (where m is the mass
flow rate). For the one-dimensional heat exchanger considered in this section, the usual
assumptions such as steady flow, no heat exchange with environment, and ignoring changes
in kinetic and potential energies as well as the longitudinal conduction are made. In the heat
exchanger, there mainly exist two kinds of irreversibility: the first is heat conduction under
finite temperature differences and the second is flow friction under finite pressure drops.
Note that according to the definition of entransy given in Section 1 the maximum entransy
in a heat exchanger is Q(T1,i − T2,i ) , where Q is the actual heat transfer rate, (T1,i − T2,i ) is the
maximum temperature difference in the heat exchanger. Therefore the entransy dissipation
rate caused by heat conduction under a finite temperature difference can be non-
dimensionalized in the following way (Guo et al. 2009):
E T E T
E T * = = (88)
)1 (T1,i − T2, i )
2
Q(T1,i − T2, i ) ε ( mc
where ε is the heat exchanger effectiveness which is defined as the ratio of the actual heat
transfer rate to the maximum possible heat transfer rate. As discussed by Guo et al (2009),
the entransy dissipation number can be employed as the performance evaluation criterion of
heat exchangers which avoids the “entropy generation paradox”.
From Eq. (82), the entransy dissipation due to flow friction under a finite pressure drop is
expressed as
m ΔP T1, o − T1, i 2 ΔP2

m T2 , o − T2 ,i
E P = 1 1 + (89)
ρ1 ln T1, o − ln T1,i ρ2 ln T2 , o − ln T2 ,i
where ΔP1 and ΔP2 refer to the pressure drops of the hot and cold water, respectively; ρ 1
and ρ 2 are their corresponding densities. Putting in dimensionless form leads to
−ΔP1 1
E P * = +
( ρc )1 (T1,i − T2,i ) T ⎡ T2 ,i − T1, i ⎤
ln 2, i ⎢ 1 − ( 1 − ε ) ⎥
T1,i ⎣ T2 ⎦
(90)
ΔP2 1
( ρ c )2 (T1,i − T2 ,i ) T1, i ⎡ T2 ,i − T1,i ⎤
ln ⎢1 + ( 1 − ε ) ⎥
T2 ,i ⎢⎣ T1,i ⎥⎦
which is called the entransy dissipation number due to flow friction. Assuming that the heat
exchanger behaves as a nearly ideal heat exchanger, then ( 1 − ε ) is considerably smaller
than unity (Bejan, 1982). For a water-water heat exchanger under usual operating
conditions, the inlet temperature difference between hot and cold water, ΔT = T1,i − T2 ,i , is
less than 100K, hence ΔT Ti < 100 / 273 ≈ 0.366 ( i = 1, 2) . Therefore, Eq. (90) can be simplified
to
ΔP1 1 ΔP2 1
E P * = + (91)
( ρc )1 (T1,i − T2,i ) ln T1,i ( ρc )2 (T1,i − T2 ,i ) ln T1,i
T2, i T2, i
Accordingly, the overall entransy dissipation number becomes
ΔP1 1 ΔP2 1
E * = E T * + E P * = ( 1 − ε ) + + (92)
( ρc )1 (T1,i − T2,i ) ln T1,i ( ρ c )2 (T1,i − T2,i ) ln T1,i
T2,i T2,i
For a typical water-water balanced heat exchanger, the number of heat transfer units Ntu
can be introduced, which approaches infinity as the effectiveness tends to unity. Since
C ∗ = 1 , the effectiveness is (Bejan, 1982):
Ntu
ε= (93)
1 + Ntu
where the number of heat transfer units is defined as
UA
Ntu =
p
mc
here U is the overall heat transfer coefficient, and A is the heat transfer area. Assume that the
heat conduction resistance of the solid wall can be neglected, compared with the convective
heat transfer, then it is appropriate to replace U with the convective heat transfer coefficient
h. Therefore,
1 1 1
= + (94a)
UA ( hA )1 ( hA )2
or
1 1 1
= + (94b)
Ntu Ntu1 Ntu2
where h1 and h2 are the convective heat transfer coefficients of the hot and cold fluids,
respectively, and Ntui = ( hA)i /(mc
)i (i = 1, 2) . In the nearly ideal heat exchanger limit,
Ntu >> 1 , that is (Bejan, 1982)
1
1−ε ≈ (95)
Ntu
From Eq.(95) the overall entransy dissipation number is
⎡ ⎤ ⎡ ⎤
⎢ ⎥ ⎢ ⎥
1 ΔP1 1 ⎥+⎢ 1 + ΔP2 1
E * = ⎢ + ⎥ (96)
⎢ Ntu1 ( ρ c ) (T1,i − T2,i ) T1,i ⎥ ⎢ Ntu2 ( ρ c ) (T1,i − T2,i ) T1,i ⎥
⎢ 1 ln ⎥ ⎢ 2 ln ⎥
⎣ T2,i ⎦ ⎣ T2 ,i ⎦
The two terms on the right of Eq. (96) correspond to the entransy dissipations of two sides of
heat transfer surfaces. For each side, the entransy dissipation number can be written as
follows:
1 Pi ΔPi
E i* = + ( i = 1, 2 ) (97)
Ntui T1,i Pi
( ρ c )i (T1,i − T2, i ) ln
T2 ,i
It is evident that the first term accounts for the entransy dissipation from the heat
conduction under finite temperature difference and the second for the entransy dissipation
from flow friction under finite pressure drop. For simplicity, we now use E * instead of E i*
to denote the entransy dissipation number for each side of the heat exchanger surface.
4.4 Parameter optimization

Theoretically, the exchanger effectiveness increases when the irreversible dissipation in the
heat exchanger decreases. From Section 1.3 one can see that the entransy dissipation can be
used to describe the irreversible dissipations, therefore we seek optimums in duct aspect
ratio and mass velocity by minimizing the entransy dissipation number E * based on Eq.
(97). Although the entransy dissipation number on one side of a heat transfer surface can be
expressed as the sum of the contributions of the heat conduction under the finite
temperature difference and flow friction under the finite pressure drop, the effects of these
two factors on heat exchanger irreversibility are strongly coupled through the geometric
parameters of the heat exchanger tube residing on that side. Therefore, based on entransy
dissipation minimization, it is possible to obtain optimal geometric parameters of the heat
exchanger such as the optimal duct aspect ratio.
( )
Recall the definition of the Stanton number St ( Re )D ,Pr and friction factor f ( Re )D :( )
4L
Ntu = St (98)
D
ΔP 4L G 2
=f (99)
P D 2ρP
where G = m / A is the mass velocity, L is the flow path length and D is the duct hydraulic
diameter. Introducing the dimensionless mass velocity, G* = G 2 ρ P
P
τ2 =
T1,i
( ρc ) (T1,i − T2 ,i ) ln
T2, i
and substituting Eqs. (98) and (99) into Eq. (97), we obtain
1 4L 2
E * = +τ 2 f G* (100)
4L D
St
D
Clearly, the duct aspect ratio 4L / D has opposing effects on the two terms of the right side
of Eq. (100). Therefore, there exists an optimal duct aspect ratio to minimize the entransy
number. When the Reynolds number and mass velocity are fixed, minimizing the entransy
dissipation number leads to the following expression for this optimum:
⎛ 4L ⎞ 1
⎜ ⎟ = (101)
⎝ D ⎠opt τ G* ( fSt )
12
The corresponding minimum entransy dissipation number is
12
⎛ f ⎞
E min
*
= 2τ G* ⎜ ⎟ (102)
⎝ St ⎠
From Eqs.(101) and (102), one can see that the optimal duct aspect ratio decreases as the
mass velocity G* increases, and the minimum entransy dissipation number is directly
proportional to the dimensionless mass velocity. Note that the minimum entransy
dissipation number is also dependent on the Reynolds number via f and St . However, the
impact of the Reynolds number on the minimum entransy dissipation number is very weak
since for many heat transfer surfaces the ratio of the friction factor to the Stanton number
does not have a significant change as the Reynolds number varies (Bejan, 1982). Therefore,
the minimum entransy dissipation number is mainly determined by the selected
dimensionless mass velocity. Obviously, the smaller the mass velocity, the longer the
working fluid remains on the heat transfer surface and the lower the irreversible
dissipations in the heat exchanger.
In designing a heat exchanger, the heat transfer area is an important consideration when it
accounts for most of the total cost of a heat exchanger. Thus in the following, we discuss
design optimization of the heat exchanger with a fixed heat transfer area.
From the definition of the hydraulic diameter, the heat transfer area for one side is
4L
A= Ac
D
where Ac is the duct cross-section area. This expression can be put in dimensionless form as
4L −1
A* = G* (103)
D
where A* is the dimensionless heat transfer area, A* = ( 2 ρ P )
12
. Substituting Eq. (103)
A /m
into Eq. (100) yields
1
E * = G* −1 + τ 2 fA*G* 3 (104)
A*St
Obviously, the dimensionless mass velocity has an opposing effect on the two terms of the
right side of Eq.(104). Thus there exists an optimal dimensionless mass velocity which
allows the entropy dissipation number to reach a minimum value when A* and Reynolds
number ( Re )D are given. Solving this optimization problem yields:
14
⎛ 1 ⎞
G*, opt = ⎜ (105)
⎜ 3 A 2τ 2 fSt ⎟⎟
⎝ * ⎠
14
⎛ τ2f ⎞
E min
*
= 4⎜ (106)
⎜ 27 A 2St 3 ⎟⎟
⎝ * ⎠
Eqs. (105) and (106) respectively give the optimal dimensionless mass velocity and the
minimum entransy dissipation number under fixed A* and Reynolds number ( Re )D . From
these two equations, the larger heat transfer area clearly corresponds to the smaller mass
velocity and lower entransy dissipation rates. Hence, to reduce the irreversible dissipation
occurring in a heat exchanger, the largest-possible heat transfer area should be adopted
under the allowable conditions.
If E * and ( Re )D are given, the minimum heat transfer area is
16 τ f12
A*,min = (107)
3 32
E *2St 3 2
with
⎛ 4L ⎞ 4 1
⎜ ⎟ = * (108)
⎝ D ⎠opt 3 E St
From Eqs. (107) and (108), one can see that a lower entransy dissipation rate corresponds to
larger heat transfer area or duct aspect ratio. Eqs. (106) and (107) are identical, providing an
expression for the minimum attainable value for the product A*1/2 E * under the given
Reynolds number.

As exemplified by the water-water counter-flow heat exchanger, the present work shows
that there exists an optimal duct aspect ratio for heat exchangers under the fixed Reynolds
number and mass velocity when the entransy dissipation number is taken as the
performance evaluation criterion. Furthermore, the formula for the optimal duct aspect ratio
was obtained analytically. Under constraints of the fixed heat transfer area (or duct volume)
and Reynolds number, it was shown that there is an optimal dimensionless mass velocity;
for which an analytical expression was also given. The results indicated that to reduce
irreversible dissipations in heat exchangers, largest-possible heat transfer areas and lowest-
possible mass velocities should be adopted. This conclusion is in agreement with numerical
results obtained by design optimization of the shell-and-tube heat exchanger based on the
entransy dissipation number as the objective function (Guo et al., 2010).
From the results obtained in this study, it can be seen that the traditional heat exchanger
design optimizations based on total cost as an objective function usually sacrifice heat
exchanger performance. This issue has been demonstrated by numerical results (Guo et al.,
2009). Guo et al. (2009) found that a little improvement in heat exchanger performance can
lead to large gains in terms of energy saving and environmental protection. Hence, in heat
exchanger design, reduction in total cost and improvement in heat exchanger performance
should be treated equally. The present work will be useful to drive new research in this
direction.
5. Acknowledgements
The support of our research by National Basic Research Program of China (Project No.
2007CB206900) is greatly appreciated.
6. References
Bejan, A. (1982). Entropy Generation through Heat and Fluid Flow. New York, Wiley
Bejan, A. (1988). Advanced Engineering Thermodynamics. New York, Wiley
Bejan, A. (1995). Entropy Generation Minimization. New York, CRC Press
Bejan, A. (1996). Entropy generation minimization: the new thermodynamics of finite-size
devices and finite-time processes. J Appl Phys, Vol. 79, pp.1191-1218
Ben-Amotz, D. & Honig, J. M. (2006). Average entropy dissipation in irreversible
mesoscopic processes. Phys Rev Lett, Vol. 96, pp. 020602.
Ben-Amotz, D. & Honig, J. M. (2003). Rectification of thermodynamic inequalities. J Chem
Phys, Vol.118, pp. 5932.
Bertola, V. & Cafaro, E. (2008). A critical analysis of the minimum entropy production
theorem and its application to heat and fluid flow. Int J Heat Mass Transfer, Vol. 51,
pp.1907-1912.
Bizarro, J. P. S. (2008). Entropy production in irreversible processes with friction. Phys Rev E,
Vol. 78, pp. 021137
Chen, L.; Wei, S. & Sun, F. (2011). Constructal entransy dissipation rate minimization of a
disc. Int J Heat Mass Transfer, Vol. 54, pp. 210.
Chen, Q. & Ren, J. X. (2008). Generalized thermal resistance for convective heat transfer and
its relation to entransy dissipation. Chinese Science Bulletin, Vol. 53, pp. 3753-3761
Cheng, X. & Liang, X. (2011). Entransy flux of thermal radiation and its application to
enclosures with opaque surfaces. Int J Heat Mass Transfer, Vol. 54, pp.269.
Cheng, X. G. (2004). Entransy and its Application in Heat Transfer Optimization. PhD thesis,
Tsinghua University, Beijing
Clausius, R. (1865). The Mechanical Theory of Heat-With its Applications to the Steam
Engine and to Physical Properties of Bodies. London: John van Voorst, 1
Paternoster Row, MDCCCLXVII.
den Broek, C. V. (2005). Thermodynamic efficiency at maximum power. Phys Rev Lett, Vol.
95, pp.190602.
Esposito, M. & Lindenberg, K. (2009). Universality of efficiency at maximum power. Phys
Rev Lett, Vol.102, pp.130602.
Esposito, M. et al. (2010). Efficiency at maximum power of low-dissipation Carnot engines.
Phys Rev Lett, Vol. 105, pp. 150603.
Guo, J. F.; Cheng, L. & Xu, M. T. (2009a). Entransy dissipation number and its application to
heat exchanger performance evaluation. Chinese Sci Bull, Vol. 54, pp. 2708.
Guo, J. F.; Xu, M. T. & Cheng, L. (2009b). The application of field synergy number in
shell-and-tube heat exchanger optimization design. Appl Energ, Vol. 86, pp.
2079-2087
Guo, J. F.; Li, M. X. & Xu M. T. (2010). The application of entransy dissipation theory in
optimization design of heat exchanger, Proceedings of the 14th International Heat
Transfer Conference, IHTC14-23348, Washington, DC, USA, August 8-13, 2010
Guo, Z. Y.; Zhu, H. Y. & Liang, X. G. (2007). Entransy-A physical quantity describing heat
transfer ability. Int J Heat Mass Transfer, Vol. 50, pp. 2545.
Guo, Z. Y. et al. (2010). Effectiveness-thermal resistance method for heat exchanger design
and analysis. Int J Heat Mass Transfer, Vol. 53, pp. 2877.
Hesselgreaves, J. E. (2000). Rationalisation of second law analysis of heat exchanger. Int. J.
Heat Mass Transfer, Vol. 43, pp. 4189-4204
Herwig, H. (2010). The Role of Entropy Generation in Momentum and Heat Transfer,
Proceedings of the 14th International Heat Transfer Conference, IHTC14-23348,
Washington, DC, USA, August 8-13, 2010
Jaynes, E. T. (1980). The minimum entropy production principle. Annu Rev Phys Chem, Vol.
31, pp. 579-601
Kondepudi, D & Prigogine, I. (1998). Modern Thermodynamics-From Heat Engines to Dissipative
Structures, John Wiley & Sons, Chichester
Landauer R. (1975). Inadequacy of entropy and entropy derivatives in characterizing the
steady state. Phys Rev A, Vol. 12, pp.636-630
Li, X. F.; Guo, J. F., Xu, M. T. & Cheng, L. (2011). Entransy dissipation minimization for
optimization of heat exchanger design. Chinese Sci. Bull. (Accepted)
Liu, X. B.; Meng, J. A. & Guo, Z. Y. (2009). Entropy generation extremum and entransy
dissipation extremum for heat exchanger optimization. Chinese Science Bulletin, Vol.
54, pp. 943-947
Mamedov, M. M. (2003). On the incorrectness of the traditional proof of the principle of

minimum production. Technical Physics Letters, Vol. 29, pp. 69-71
McClintock, F. A. (1951). The design of heat exchangers for minimum irreversibility. ASME
Paper, No. 51-A-108, 1951, presented at the 1951 ASME Annual Meeting
Prigogine, I. (1967). Introduction to Thermodynamics of Irreversible Processes, Third ed., Wiley,
New York, pp. 76-77.
Shah, R. K. & Skiepko, T. (2004). Entropy generation extrema and their relationship with
heat exchanger effectiveness-number of transfer unit behavior for complex flow
arrangement. ASME J Heat Transfer, Vol. 126, pp. 994.
Wang, S. P.; Chen, Q. L. & Zhang, B. J. (2009). An equation of entransy transfer and its
application, Chinese science bulletin, Vol.54, pp. 3572-3578.
Wu, J. & Liang, X. G. (2008). Application of entransy dissipation extremum principle in
radiative heat transfer optimization. Science in China Series E: Technological Sciences,
Vol. 51, pp. 1306-1314
Xia, S. J.; Chen, L. G. & Sun, F. R. (2009). Optimization for entransy dissipation minimizatino
in heat exchanger. Chinese Science Bulletin, Vol.54, pp. 3587-3595
Xu, M. T. & Cheng L. (2010). From the extremum principle of entransy dissipation to
steady balance equations of fluid mechanics, Proceedings of the 14th International
Heat Transfer Conference, IHTC14-23348, Washington, DC, USA, August 8-13,
2010
Xu, M. T.; Guo, J. F. & Chen, L. (2009). Application of entransy dissipation theory in heat
convection. Front. Energy Power Eng China, Vol.3, pp. 402-405
Xu, M. T. (2011). The thermodynamic basis of entransy and entransy dissipation. Energy,
Vol. 36, 4272-4277
Yilmaz, M.; Sara, O. N. & Karsli, S. (2001). Performance Evaluation Criteria for Heat
Exchangers Based on Second Law Analysis. Exergy, an International Journal, Vol.1,
pp. 278-294.
Ziman, J. M. (1956). The general variational principle of transport theory. Can J Phys, Vol. 34,
pp.1256-1273
15
Inverse Space Marching Method for

Determining Temperature and Stress
Distributions in Pressure Components
Jan Taler1, Bohdan Węglowski1,
Tomasz Sobota1, Magdalena Jaremkiewicz1 and Dawid Taler2
1Cracow University of Technology
2University of Science and Technology
Poland
1. Introduction
Thermal stresses can limit the heating and cooling rates of temperature changes. The largest
absolute value of thermal stresses appears at the inner surface. Direct measurements of these
stresses are very difficult to take, since the inner surface is in contact with water or steam
under high pressure. For that reason, thermal stresses are calculated in an indirect way
based on measured temperatures at selected points, located on an outer thermally insulated
surface of a pressure element. First, time-space temperature distribution in pressure element
is determined using the inverse space marching method.
High thermal stresses often occur in partially filled horizontal vessels. During operation
under transient conditions, for example, during power plant start-up and shut-down, there
are significant temperature differences over the circumference of the horizontal pressure
vessels (Fetköter et al., 2001; Rop, 2010). This phenomenon is caused by the different heat
transfer coefficients in the water and steam spaces. This takes place in large steam generator
drums, superheater headers and steam pipelines. High thermal stresses caused by
nonuniform temperature distribution on vessel circumference also occur in emergency
situations such as fire of partially filled fuel tanks. The upper part of the horizontal vessel is
heated much faster than the lower part filled with liquid.
Similar phenomenon occurs in inlet nozzles in PWR nuclear reactor, at which high
temperature differences on the circumference of the feed water nozzles are observed.
The study presents an analysis of transient temperature and stress distribution in a
cylindrical pressure component during start-up of the steam boiler and shut-down
operations. Thermal stresses are determined indirectly on the basis of measured
temperature values at selected points on the outer surface of a pressure element. Having
determined transient temperature distribution in the entire component, thermal stresses are
determined using the finite element method. Measured pressure changes are used to
calculate pressure caused stresses. The calculated temperature histories were compared with
the experimental data at selected interior points.
The presented method of thermal stress control was applied in a few large conventional
power plants. It can also be used successfully in nuclear power plants. The developed
method for monitoring thermal stresses and pressure-caused stresses is also suitable for
nuclear power plants, since it does not require drilling holes for sensors in pressure element
walls. Measurements conducted over the last few years in power plants demonstrate that
the presented method of stress monitoring can be applied in systems for automatic power
boiler start-up operations and in systems for monitoring the fatigue and creep usage factor
of pressure components.
2. Present methods used to determine temperature and thermal stress

transient in boiler pressure elements
The simplest and most frequently used method for reducing thermal stresses in pressure
elements is to limit heating and cooling rates. The allowable rates of fluid temperature
changes can be determined using German Boiler Codes TRD 301 (TRD 301, 2001) or
European Standard EN 12952-3 (EN 12952-3, 2001). The allowable heating rate of temperature
changes is determined from the following condition (EN 12952-3, 2001; TRD 301, 2001)
αm
( p − po ) dm + α vT s 2
Tφwφ f = σ min . (1)
2s a
The maximum allowable cooling rate is calculated in a similar way:
αm
( p − po ) dm + α vT s 2
Tφwφ f = σ max . (2)
2s a
In Equations (1-2), the following nomenclature is used: α = k/(c·ρ) – thermal diffusivity,
m2/s, rin, rout – inner and outer radius, m, dm = rout + rin – mean diameter, m, p – absolute
pressure, MPa, po – ambient pressure, MPa, s = rout - rin – thickness of cylindrical element, m,
vT – rate of temperature changes of fluid or pressure element wall, K/s, αm – pressure
caused stress intensity factor, αT – thermal stress intensity factor, σmin – allowable stress
during start-up (heating), MPa, σmax – allowable stress during shutdown (cooling), MPa.
Coefficients φw and φf are defined as follows
Eβ , (3)
φw =
1 −ν
φf =
( 3u 2
)( )
− 1 u2 − 1 − 4u 4 ln u
, (4)
( )
8 u − 1 ( u − 1)
2 2
where: E – Young's modulus, MPa, β – linear coefficient of thermal expansion, 1/K,

ν – Poisson’s ratio, u = rout/rin – ratio of outer to inner surface radius.
Equations (1-2) can also be used to calculate the maximum total stress at hole edges. The
heating or cooling rate of temperature changes in pressure elements vT can be calculated
using the moving average filter (Taler, 1995)
df 1
vT = = ( −63 f i − 4 + 42 f i − 3 + 117 f i − 2 + 162 f i −1 +
dt t = ti 693Δt (5)
+177 f 0 + 162 f i + 1 + 117 f i + 2 + 42 f i + 3 − 63 f i + 4 ) ,
where fi are medium or wall temperatures at nine successive time points with Δt time step.
Inverse Space Marching Method for Determining
Temperature and Stress Distributions in Pressure Components 275
Equations (1-2) and (5) are not only valid for a quasi-steady state (Taler, 1995), but also
when the temperature change rate vT is the function of time: vT = vT (t). For pressure
components with complex geometry, the stress value in the stress concentration areas can be
calculated using the finite element method (FEM). By determining the so called influence
function with the use of the FEM, one is subsequently able to carry out an on-line stress
calculation with known heat transfer coefficient and fluid temperature transient (Taler et al.,
2002).
This chapter formulates the problem of determining the transient temperature in a pressure
element as an inverse transient heat conduction problem. The temperature distribution is
determined on the basis of temperature histories measured at selected points at the outer
insulated surface of a pressure component. After determining transient temperature
distribution in the entire pressure component, thermal stresses are computed using the
FEM. Inverse problem is solved using the finite volume method (FVM). Thermal and
pressure caused stresses are calculated using the FEM.
3. Mathematical formulation of inverse problem

In the following, two dimensional inverse heat conduction problem (IHCP) will be solved
(Fig. 1). The analyzed domain is divided into two subdomains: direct and inverse. Boundary
and initial conditions are known for the direct region so that the transient temperature
distribution is obtained for the solution of the boundary-initial problem. The temperature
distribution on the inner closed surface Sm which is located inside the analyzed area (Fig. 1)
is known from measurements. Based on the solution of the direct problem the heat flux on
the boundary Sm can be evaluated. Thus, the two boundary conditions are known on the
surface Sm:
T ( s, t ) S = f ( s, t ) , (6)
m
∂T
−k = q ( s, t ) , (7)
∂n Sm
while on the inner surface Sin of the body, the temperature and heat flux are unknown.
In order to evaluate the transient temperature distribution in the inverse region, this region
is divided into control volumes (Fig. 2). The method marches in space towards the inner
surface of the body S by using the energy balance equations for the finite volumes placed on
the boundary Sm to determine the temperatures in adjacent nodes. In this way of proceeding
the time derivatives of the measured temperature changes have to be calculated. The
accurate calculation of the time derivatives of the measured temperature histories is difficult
since the measured temperature values are burdened with random measurement errors.
Thus, time-temperature charts have to be smoothed before evaluating the time derivative. In
the present section this was achieved by using the local polynomial approximation. The
successive nine temperature data points were approximated using the polynomial of the
3rd degree and then the derivatives in the middle of each interval (time coordinate of the
point 5) were calculated.
The space-marching method will be illustrated by an example showing the evaluation of the
temperature distribution in the cylindrical wall using the temperature measurement points
equally distributed on the surface Sm (Fig. 2). It is assumed that the temperature and heat
flux distributions are known on the surface Sm from temperature measurements at points 16-
22 and from the solution of the direct problem.
Fig. 1. The analyzed body divided into the inverse and direct regions
Fig. 2. Division of the inverse region into control volumes

For the ideally insulated outer surface the heat fluxes q16-q22 and their time derivatives are
equal to zero. The heat balance equation for the finite volume with the node (i, j) is as
follows (Fig. 3):
1 ⎡⎛
2 2⎤
Δr ⎞ ⎛ Δr ⎞ dTi , j
⎢⎜ ri , j +
2 ⎢⎣⎝
⎟ − ⎜ ri , j −
2 ⎠ ⎝ 2 ⎠
⎟
⎥⎦
( ) ( )
⎥ Δϕ c Ti , j ρ Ti , j
dt
=
⎛ Δr ⎞ ( ) ( )
k Ti , j + k Ti + 1, j Ti + 1, j − Ti , j
= ⎜ ri , j + ⎟ Δϕ +
⎝ 2 ⎠ 2 Δr
(8)
⎛ Δr ⎞ ( ) ( )
k Ti , j + k Ti − 1, j Ti − 1, j − Ti , j ( ) ( )
k Ti , j + 1 + k Ti , j Ti , j + 1 − Ti , j
+ ⎜ ri , j − ⎟ Δϕ + Δr +
⎝ 2 ⎠ 2 Δr 2 Δϕ ri , j
+Δr
( ) ( )
k Ti , j − 1 + k Ti , j Ti , j − 1 − Ti , j
,
2 Δϕ ri , j
where: c – specific heat, J/(kg·K), ρ – density, kg/m3, k – thermal conductivity, W/(m·K),

t – time, s, T – temperature, °C, Δr – space step in radial direction, m, Δφ – angular step, rad.
Transforming the heat balance Equation (8) for Ti-1,j, we obtain:
2 2
⎛ Δr ⎞ ⎛ Δr ⎞
Ti − 1, j = Ti , j + Δr ⎝
⎜ ri , j +
2
⎟ − ⎜ ri , j −
⎠ ⎝ 2 ⎠
⎟
⋅
( ) ( )
c Ti , j ρ Ti , j
⋅
dTi , j
−
ri , j −
Δr ( ) ( )
k Ti , j + k Ti − 1, j dt
2
Δr
ri , j + ( ) (
k Ti , j + k Ti + 1, j ) ( ) ( )
( Δr )2 ⋅ k Ti , j + 1 + k Ti , j ⋅ Ti , j + 1 − Ti , j −
− 2 ⋅
( )
⋅ Ti + 1, j − Ti , j − (9)
ri , j − ( ) (
Δr k T + k T
i, j )
i − 1, j ri , j − ( ) ( )
Δr k T + k T
i, j i − 1, j Δϕ 2 ri , j
2 2
−
( ) ( )
( Δr )2 ⋅ k Ti , j −1 + k Ti , j ⋅ Ti , j −1 − Ti , j .
ri , j − ( ) ( )
Δr k T + k T
i, j i − 1, j Δϕ 2 ri , j
2
Since Eq. (9) is nonlinear, the fixed-point iterative technique is used to determine the
temperature Ti,j−1:
2 2
⎛ Δr ⎞ ⎛ Δr ⎞
( n + 1)
Ti − 1, j = Ti , j + Δr
⎜ ri , j +
⎝ 2 ⎠ ⎝
⎟ − ⎜ ri , j −
2 ⎠
⎟ ( ) (Ti , j ) dTi , j −
c Ti , j ρ
ri , j −
Δr
2
( ) (Ti(−n1,) j ) dt
k Ti , j + k
Δr
ri , j + ( ) ( )
k Ti , j + k Ti + 1, j ( Δr )2 k (Ti , j +1 ) + k (Ti , j ) Ti , j +1 − Ti , j −
− 2
( )
Ti + 1, j − Ti , j − (10)
ri , j −
2
( ) ( )
Δr k T + k T ( n )
i, j i − 1, j ri , j −
Δr k
2
(Ti , j ) + k (Ti(−n1,) j ) Δϕ 2 ri , j
−
2
( ) ( )
( Δr ) k Ti , j −1 + k Ti , j Ti , j −1 − Ti , j .
ri , j −
2
( ) ( )
Δr k T + k T ( n )
i, j i − 1, j
Δϕ 2 ri , j
Fig. 3. Space marching in the inverse region
The iteration process continuous until the condition Ti(− 1, j ) − Ti(− 1,) j ≤ ε is satisfied. Symbol n
n+1 n
denotes the iteration number and ε is some small number (tolerance), for example, ε = 0.0001 K.
If the thermal properties are constant, the iterations are not required.
To determine the wall temperature at the node 1 (Fig. 2) we proceed as follows. First the
heat balance equations are written for the nodes 16 to 22 from which the temperature at the
nodes 9 to 13 is determined. Then the heat balance equations are set for the nodes 9 to 13
from which the temperature at the nodes 4 to 6 is evaluated. From the heat equations for the
nodes 4, 5 and 6, we can determine the temperature at node 1 located at the inner surface of
the pressure component. Based on the measured temperatures at the nodes 17 to 23, the
temperature at the node 2 is calculated in similar way. Using the measured temperature at
nodes 18 to 24 the temperature at the node 3 is estimated.
Repeating the procedure described above for all the nodes, located at the outer surface, at
which the wall temperature is measured, the temperature at the nodes situated at the inner
component surface are determined. After determining the temperature distribution at the
whole cross-section at the time point t, the temperature distribution at time t + Δt is
computed. After calculating the temperature distribution the thermal stresses were calculating
using the FEM. The finite element mesh is constructed so that the FEM nodes are coincident
with the nodes used in the finite volume method.
Random measurement errors of the temperature fj(t) have great influence on the estimated
temperature and thermal stress distributions. If the temperature data are burdened with
random errors, least squares smoothing is used to reduce the effect of the measurement
errors on the calculated time derivatives dfj/dt or dTi,j/dt. The Gram orthogonal
polynomials were used for smoothing the measured time-temperature history fj(t) and
estimated temperatures Ti−1,j(t) (Taler, 1995). For linear IHCP, when thermal properties are
temperature independent, node temperatures can be expressed in explicit form (Taler &
Zima, 1999; Taler et al., 1999). In the following subsection the linear IHCP will be presented
in detail.
4. Linear inverse problem

The typical way of evaluation of inverse method accuracy is to use the direct method to
generate the ”measured data” on the outer surface assuming the values of fluid temperature
and heat transfer coefficients on the inner surface. The generated data are then used as input
data for the inverse method. The results from the inverse solution are accurate when they
agree with assumptions made for the direct problem.
First, a direct problem of heat conduction, where the boundary conditions on the drum
inner and outer surfaces as well as initial conditions are known, will be discussed. Then the
temperature distribution will be determined on the basis of measured temperature changes
on the outer, thermally insulated surface of the drum with unknown boundary condition on
the inner surface (the inverse problem). The problem formulated in this way, as ill-posed, is
more difficult to solve than the first one.
Both direct and inverse problems of heat conduction allow the determination of temperature
distribution in the drum cross-section. However, the procedure for thermal stresses
determination is slightly different in the two cases. In the direct method both temperature
distribution and thermal stresses are calculated using FEM. In the inverse method, unlike
the direct problem solved by FEM, the initial and boundary conditions on the inner surface
are not set because the transient temperature distribution is obtained from temperature
measurements on the outer surface of the drum. Temperature values used for stress
calculations are obtained from the inverse solution presented in this section.
4.1 Direct problem of heat conduction

The direct problem has been solved with the following assumptions:
Temperature distribution is described by equation:
1 ∂ ⎛ ∂ T ⎞ 1 ∂ 2T 1 ∂ T
⎜r ⎟+ = (11)
r ∂ r ⎝ ∂ r ⎠ r 2 ∂ϕ 2 a ∂ t
with boundary conditions:
∂T
k =0, (12)
∂r r = ro
k
∂T
∂r r = rin
(
= hw Tw − T r = r
in
) (13)
0 ≤ ϕ ≤ ϕw,
k
∂T
∂r r = rin
(
= hs Ts − T r = r
in
) (14)
ϕw ≤ ϕ ≤ π,
and initial condition:
T t = 0 = T0 ( r ,ϕ ) . (15)
The symbols hw and hs stand for heat transfer coefficients at water and steam region,
respectively. The initial temperature T0(r, φ) is known from measurements or calculations at
steady state.
Since the temperature field is symmetrical with respect to the vertical plane passing through
the drum longitudinal axis, the temperature field is analyzed only in half of the drum. The
division of drum cross-section into control volumes and finite elements has been shown in
Figs. 4 and 5, respectively.
Fig. 4. Division of half of drum cross section into control volumes (52 control volumes)
Fig. 5. Division of half of drum cross-section into finite elements (125 nodes, 96 elements)
The direct problem will be solved separately by means of two methods: the method of
control volume (CVM) and finite elements. The following data have been used for
calculations: rin = 0.65 m, rout = 0.74 m, k = 47 W/(m·K), ρ = 7850 kg/m3, c = 500 J/(kg·K),
ϕw = π/2, hw = 1000 W/(m2·K) and hs = 2000 W/(m2·K). The changes of the fluid temperature
in time (Tw = Ts) have been shown in Fig. 6.
In the control volumes method, following planar discretization, 52 ordinary differential
equations are obtained, which have been integrated by Runge-Kutta method.
Fig. 6. Fluid temperature changes in drum as function of time and temperature at nodes on
the outer surface obtained from the solution of the direct problem using FEM; 1 – 6 numbers
of curves refer to the numbers of temperature measuring points on the drum outer surface
Fig. 7. Comparison of temperature changes at points 2 and 6 calculated by control volume

method and FEM
The calculation results have been presented in Fig. 6. From the comparison of temperature
changes on the inner surface at points ϕ2 = π/6 and ϕ6 = 5π/6 shown in Fig. 7 it can be seen
that the agreement of results obtained by FEM and those by means of control volumes
method is very good.
4.2 Inverse problem of heat conduction

The transient temperature distribution in the drum cross-section will be determined on
the basis of temperature measurements on the drum outer thermally insulated surface.
Thus we have to solve an inverse problem of heat conduction because there are two
known conditions on the outer surface: temperature and heat flux and none on the inner
surface.
A general inverse problem in cylindrical coordinates will be solved by means of control
volume method. A diagram of drum wall division into elementary cells has been shown in
Fig. 8a.
(a) (b)
Fig. 8. Division of drum wall into control volumes: (a) calculation pattern in inverse problem
to determine temperature at point 1 at the inner surface on the basis of temperature
measurement at five points: 9 ÷ 13 at the surface; (b) half of drum cross-section divided into
control volumes when solving the inverse problem
Temperature is determined only in three nodes in the radial direction: on the outer surface,
in the wall middle and on the inner surface. Unlike in the direct problems, so few nodes
are entirely sufficient for exact evaluation of temperature distribution (Taler & Zima, 1999).
The method has a local character because in order to determine the temperature changes at
point 1 on the inner surface it is enough to measure the temperature at five points 9 ÷ 13 on
the outer surface. It is assumed that besides the temperature, the heat flux is known at
points 9 ÷ 13. Nodes 9 ÷ 13 are located on easily accessible outer surface, where it is easy to
place thermocouples for temperature measurements. When surface r5 on which the
temperature is measured is thermally insulated, then q9 = q10 = q11 = q12 = q13 = q14 = q15 = 0.
Since the considered problem is inverse, it is described only by Eq. (11) and boundary
condition (12) on the outer surface. In order to determine the temperature distribution in the
whole cross-section including the inner surface the inverse space marching method is
presented. Using the control volume method the following heat balance equations for nodes
10, 11 and 12 on the outer surface are obtained:
Δϕ 2 T − T10 Δr T − T10 T − T10 Δr

cρ
2
(
r5 − r42 )
dT10
dt
=k 9
Δϕ r5 2
+k 4
Δr
Δϕ r4 + k 11
Δϕ r5 2
− q10 Δϕ r5 , (16)
Δϕ 2 T − T11 Δr T − T11 T − T11 Δr

cρ
2
(
r5 − r42 )
dT11
dt
= k 10
Δϕ r5 2
+k 5
Δr
Δϕ r4 + k 12
Δϕ r5 2
− q11Δϕ r5 , (17)
Δϕ 2 T − T12 Δr T − T12 T − T12 Δr

cρ
2
(
r5 − r42 )
dT12
dt
= k 11
Δϕ r5 2
+k 6
Δr
Δϕ r4 + k 13
Δϕ r5 2
− q12 Δϕ r5 . (18)
From Eqs. (16), (17) and (18) temperatures T4, T5 and T6 are determined, respectively
( )
Δr r52 − r42 dT ( Δr )2 T − 2T + T + r5Δr q ,
( 9 11 )
10
T4 = T10 + − 10 10 (19)
2 ar4 dt 2 r4r5 ( Δϕ )2 r4 k
( )
Δr r52 − r42 dT ( Δr )2 T − 2T + T + r5Δr q ,
( 10 12 )
11
T5 = T11 + − 11 11 (20)
2 ar4 dt 2 r4r5 ( Δϕ )2 r4 k
( )
Δr r52 − r42 dT ( Δr )2 T − 2T + T + r5Δr q .
( 11 13 )
12
T6 = T12 + − 12 12 (21)
2 ar4 dt 2 r4r5 ( Δϕ )2 r4 k
Next, marching in space towards the inner surface the heat balance equation for node 5 is
written as:
Δϕ 2 T − T5 T − T5 T − T5 T − T5
cρ
2
(
r4 − r22 )
dT5
dt
=k 4
Δϕ r3
Δr + k 1
Δr
Δϕ r2 + k 6
Δϕ r3
Δr + k 11
Δr
Δϕ r4 , (22)
from which temperature T1 is determined:
Δr 2 ( Δr ) T − 2T + T + T ⎛ 1 + r4 ⎞ − r4 T .
2
T1 = (
2 ar2
r4 − r22)dT5
−
dt r2 r3 ( Δϕ )2
( 4 5 6) 5⎜
⎝
⎟
r2 ⎠ r2
11 (23)
After the substitution of temperatures T4, T5 and T6 defined by Equations (19), (20) and (21)
respectively to (22), we obtain after transformations:
(
⎡ r42 − r22 ) 2 2 2
( 2 2
)(
⎛ 1 1 ⎞ ⎤ Δr dT11 ( Δr ) r5 − r4 r4 − r2 d 2T11 )
T1 = T11 + ⎢
⎢ r2
(
+ r52 − r42 ⎜ + ⎟ ⎥
⎝
)
r4 r2 ⎠ ⎥ 2 a dt
+
4 a2 r2 r4 dt 2
−
⎣ ⎦
−
( Δr ) 3 ⎡
⎢ 5
2
(
r − r42
+
2
) (2 ⎤
r4 − r2 ⎛ dT
⎥ 10
) dT dT ⎞
− 2 11 + 12 ⎟ −
2 ⎢ ⎥ ⎜
2 ar2 r4 ( Δϕ ) r3 2 r5 ⎝ dt dt dt ⎠
⎣ ⎦
( Δr )2 ⎛ 1 + 1 + 1 ⎞ T − 2T + T +
− ⎜ ⎟ ( 10 11 12 ) (24)
( Δϕ )2 ⎝ 2r4r5 2r2r5 r2r3 ⎠
( Δr ) 2 ⎡ ( Δr )2 ⎛1 1 ⎞ r5 Δr ⎤
+ ⎢ ⎜ T9 − 2T10 + 3T11 − 2T12 + T13 ⎟ − ( q10 − 2q11 + q12 ) ⎥ +
r2 r3 ( Δϕ ) ⎢⎣ r4 r5 ( Δϕ )
2 2
⎝ 2 2 ⎠ r4 k ⎦
+
r5 Δr ⎛ 1 1 ⎞
+ q +
( Δr ) r42 − r22 r5 dq11
2
.
( )
⎜ ⎟ 11
k ⎝ r4 r2 ⎠ 2 ar2 r4 k dt
In order to determine the heat flux in the node on the inner surface it is indispensable to
know the temperature at two adjacent points also on the inner surface. Assuming that the
temperatures in nodes 2 and 3 have been determined in a similar way as for node 1, we can
define the heat flux in node 2 from the equation of heat balance for this node:
Δϕ 2 T − T2 Δr T − T2 Δr T − T2
cρ
2
(
r2 − r12
dT2
dt
=k 1 )
r1Δϕ 2
+k 3
r1Δϕ 2
+k 6
Δr
r2 Δϕ + q 2 r1Δϕ , (25)
from which a formula for heat flux is obtained:
q2 = k ⎢
(
⎡ r22 − r12
dT2
−
) Δr
( T1 − 2T2 + T3 ) − 2 6
⎤
r (T − T2 ) ⎥
. (26)
⎢ 2 ar1 dt 2 r12 ( Δϕ ) 2
r1 Δr ⎥
⎣ ⎦
If the temperature of fluid T∞ is known, also heat transfer coefficient h2 on cell 2 border
contacting the liquid can be determined:
q2
h2 = . (27)
T∞ − T2
Formulae (19), (20) and (24) defining the temperature in nodes and in formula (26) for heat
flux include the time derivatives of the functions representing the changes of the measured
temperatures in time. It is difficult to calculate these derivatives due to random
measurement errors.
In the present section the measured temperature histories have been smoothed by local
approximation with Gram’s polynomials (Taler, 1995; Taler & Zima, 1999). The
approximating polynomial coefficients have been determined on the basis of successive
eleven measurement points. The smoothed value of measured temperature (approximating
polynomial) and its time derivatives are determined only in the middle of the interval, i.e. in
time : ti + 5(Δt), where ti is time coordinate of the first point in the analyzed time interval: ti,
ti +10(Δt).
4.3 Calculation examples-testing of inverse method

The division of the half of drum cross-section into control volumes to solve the inverse
problem has been shown in Fig. 8b.
To test the inverse method a typical procedure was applied. First the direct problem was
solved by means of control volume method, calculating the temperature changes at points
15 ÷ 21. Next the calculated temperatures were used as ”measurement data”. The inverse
problem was solved at time step Δt = 40 s, and the time changes of temperature and heat
transfer coefficient at points 1 ÷ 7 on the inner surface were determined. From the analysis of
Fig. 9 it can be seen that there is a good agreement between the temperatures obtained from
the solution of the direct problem (FEM) and those obtained from the solution of the inverse
problem. So it can be seen, that in spite of the little number of nodes in the radial direction
the inner surface temperature can be exactly determined based on the temperature
measurements on the outer surface.
Fig. 9. Comparison of temperature changes at points 2 and 6 obtained from inverse problem
solution and direct problem solved by FEM
The heat transfer coefficients determined at the same points 2 and 6 differ only slightly from
hw = 1000 W/(m2·K) and hs = 2000 W/(m2·K) taken when solving the direct problem (Fig.
10). The differences between temperatures and stresses distributions on the drum
circumference are a little larger. This is due to the small number of control volumes on the
drum circumference. Using the seven temperature measurement points on the outer surface
and assuming the step change of heat transfer coefficient from hw to hs, it is very difficult to
determine the temperature and stress distribution in the drum for ϕ ≅ π/2. It is caused by
small temperature differences at points 1 ÷ 3 and at points 5 ÷ 7 (Fig. 4). In order to obtain
better accuracy of the inverse solution, especially at the water-steam boundary where an
abrupt change of heat transfer coefficient occurs, the grid of control volumes should be
made finer in this area.
Fig. 10. Comparison of heat transfer coefficients determined from solution of inverse heat
conduction problem at point 6 (hs = 2000 W/(m2·K)) and a point 2 (hw = 1000 W/(m2·K)) (see
Fig. 9): 1 and 2 – given (exact) and calculated value of heat transfer coefficient in steam
space, respectively, (point 6); 3 and 4 – given (exact) and calculated value of heat transfer
coefficient in water space, respectively, (point 2)
5. Experimental validation of the developed method

In order to validate the developed method and to improve the computer system for
continuous monitoring the power boiler operation, modern experimental installation was
built (Fig. 11). During its construction, particular emphasis was placed on ensuring that the
conditions are similar to those that may occur in pressure components installed in power
plants. In addition to metal temperature, temperature, mass flow rate, and pressure of the
steam were measured as a function of time. To measure temperature distribution on the
outer surface of the steam header pre-calibrated thermocouples were used. The
experimental installation consists of the following elements:
• steam generation unit with steam output capacity 700 kg/h, equipped with a three-
stage oil burner,
• installation for boiler feed water treatment,
• tray-type (cascade-type) deaerator,
• blow down flash-vessel for heat recovery,
• steam pressure reduction station 10 bar /4 bar /2 bar,
• boiler control system,
• steam header made of martensitic high alloy P91 steel.
The measurements of temperature carried out on the outer surface of the steam header will
be used for the determination of the temperature and stress distributions at the wall cross-
section including the inner and outer surfaces of the steam header using the inverse heat
conduction methods.
Fig. 11. View of the experimental installation for testing the computer system for on-line
monitoring thick-wall pressure components
The steam header was made out of the martensitic high alloy P91 steel. The following
physical properties of the P91 steel have been adopted for calculations: specific heat,
c = 486 J/(kg·K), thermal conductivity, k = 29 W/(m·K), density, ρ = 7750 kg/m3, modulus of
elasticity, E = 2.28·1011 Pa, Poisson’s ratio, ν = 0.29, coefficient of thermal expansion
β = 0.098·10−4 1/K. The outer diameter dout, wall thickness δ and the length L are: 355 mm,
50 mm and 3765 mm, respectively. Thirteen thermocouples NiCr-NiAl (K-type) were
mounted every 15° on the half of the outer circumference at the distance of 2150 mm from
the inlet of steam (Fig. 12a). The actual temperatures measured at the outer surface of the
steam header were used in the analysis. To validate the inverse technique developed in the
chapter, four thermocouples were installed at the interior locations: δ1 = δ2 = dout – 0.50δ,
δ3 = dout – 0.75δ and δ4 = din + 6 mm (Fig. 12b), where the symbol δ denotes the tube
thickness. These measurements were carried out with sheathed thermocouples NiCr-NiAl
(K-type) with outer diameter of 3 mm. The temperatures measured at seven nodes 22-28 on
the outer surface of the steam header (Fig. 13) were used for the determination of the
temperature and stress distributions at the wall cross-section including the inner surface.
The agreement between the calculated and measured temperature values is satisfactory.
The small discrepancies are caused by the delayed and damped response of the
thermocouples.
The measured and calculated temperature histories are depicted in Figs. 14a and 14b,
respectively. The comparison of the calculated and measured temperature values at the
interior points is shown in Fig. 15.
After determining the temperature distribution the thermal stresses and stresses due to
inner pressure were calculated using the FEM. Circumferential and longitudinal stresses at
the inner and outer header surfaces as functions of time are shown in Figs. 16 and 17,
respectively.
(a)
(b)
Fig. 12. Location of temperature sensors on the steam header (a) thermocouples at the outer
surface on the steam header; (b) location of interior thermocouples used for experimental
validation of the inverse procedure
Fig. 13. Division of the steam header cross-section into finite volumes
The largest stresses (Figs. 16-18) occur at the beginning of the heating process, when the
ratio of wall temperature change is significant. The stresses at the outer surface are lower in
comparison to the inner surface. The greatest absolute value has the longitudinal stress at
the inner surface (Figs. 16-17). The influence of the inner pressure on the calculated stresses
is negligible since the inner pressure is very low (Fig. 14a).
6. Conclusions
The chapter presents a method of transient temperature field identification in the drum on
the basis of measured temperature changes at locations on the outer insulated surface of the
pressure component. The method’s accuracy was demonstrated by comparison of measured
and calculated temperature at a few internal points. In order to reduce the sensitivity of the
inverse method to random errors, the measured temperature histories were smoothed by
the moving average filter based on the Gram’s polynomials approximating nine successive
measurement points. The smoothed value of measured temperature and its time derivatives
are determined only in the middle of the interval, i.e., in time ti + 4(Δt), where ti is time
coordinate of the first point in the analyzed time interval: [ti, ti + 8(Δt)]. From the measured
temperature histories on the outer surface of a pressure component, the temperature
distribution can be determined at different time points in the whole cross section. Based on
nodal temperatures determined in such a way, thermal stresses are calculated using FEM.
Values of fluid temperature and heat transfer coefficient on the inner surface of the component
are not needed because the temperature distribution in the component cross-section is
determined by the developed method for solving the inverse heat conduction problem.
Thus, the developed method is more accurate than the direct temperature distribution
methods, because the fluid temperatures and heat transfer coefficients are unknown in
operating components.
(a)
(b)
Fig. 14. Time changes of the measured (a) and computed (b) temperature values
Fig. 15. Comparison of the calculated and measured time changes of the wall temperature at
the interior points
(a) (b)
Fig. 16. Circumferential stress as a function of time; (a) – inner surface, (b) – outer surface
(a) (b)
Fig. 17. Longitudinal stress as a function of time; (a) – inner surface, (b) – outer surface
(a) (b)
Fig. 18. Equivalent stress as a function of time; (a) – inner surface, (b) – outer surface
The method developed in the chapter can be used for the stress identification in monitoring
systems for the assessment of remnant lifetime of pressure components in conventional and
nuclear power plants.
7. References
European Standard EN 12952-3 (2001). Water-tube boilers and auxiliary installations – Part 3:
Design and calculation for pressure parts, European Committee for Standardization,
reu de Stassart, 36 B-1050 Brussels, ISBN 0-580-38354-7
Fetköter, J.M.; Lauer, D.J.; Wedemeyer, M. & Engelhard G. (2001). Replacement of
Shutdown Cooling System and Repair of Nozzle Welds of Reactor Pressure Vessel
in the Forsmark 1 + 2 Nuclear Power Plant, VGB PowerTech, Vol.81, No.6, (June
2001), pp. 62-64, ISSN 978-3-86875-347-9
Rop, P. (2010). Drum Plus: a drum type HRSG with Benson benefits. Modern Power Systems,
Vol.30, No.10, pp. 35-40, ISSN 0260-7840
Taler, J. (1995). Theory and Practice of Identification of Heat Transfer Processes, Ossolineum,
ISBN 83-04-04276-2, Wrocław-Warszawa-Kraków, Poland (in Polish)
Taler, J., Węglowski, B., Zima, W., Grądziel, S. & Zborowski, M. (1999). Analysis of Thermal
Stresses in a Boiler Drum During Start-Up, Transactions of the ASME, Journal of
Pressure Vessel Technology, Vol.121, No.1, (February 1999), pp.84-93, ISSN 0094-9930
Taler, J.; Zborowski, M. & Węglowski, B. (2002). Optimisation of Construction and Heating
of Critical Structural Components of Boiler Drums. VGB PowerTech, Vol.82, No.11,
(November 2002), pp. 19-24, ISSN 978-3-86875-347-9
Taler, J. & Zima, W. (1999). Solution of inverse heat conduction problems using control
volume approach, International Journal of Heat and Mass Transfer, Vol.42, No. 6,
(March 1999), pp. 1123-1140, ISSN 0017-9310
TRD 301 (2010). Zylinderschalen unter innerem Überdruck. Technische Regeln für
Dampfkessel (TRD), Heymanns Beuth Köln – Berlin, S. 143-185, ISBN
9783452273925
16
Experimental Prediction of Heat

Transfer Correlations in Heat Exchangers
Tomasz Sobota
Cracow University of Technology
Poland
1. Introduction
Heat exchangers is a broad term related to devices designed for exchanging heat between
two or more fluids with different temperatures. In most cases, the fluids are separated by a
heat-transfer surface. Heat exchangers can be classified in a number of ways, depending on
their construction or on how the fluids move relative to each other through the device. The
use of heat exchangers covers the following areas: the air conditioning, process, power,
petroleum, transportation, refrigeration, cryogenic, heat recovery, and other industries
applications. Common examples of heat exchangers in everyday use are air preheaters and
conditioners, automobile radiators, condensers, evaporators, and coolers. (Kuppan, 2000).
Many factors enter into the design of heat exchangers, including thermal analysis, size,
weight, structural strength, pressure drop, and cost. Cost evaluation is obviously an
optimization process dependent upon the other design parameters (Pitts & Sissom, 1998).
Economics plays a key role in the design and selection of heat exchanger equipment, and the
engineer should bear this in mind when taking up any new heat transfer design problem.
The weight and size of heat exchangers are significant parameters in the overall application
and thus may still be considered as economic variables (Holman, 2009; Shokouhmand et al.,
2008; Rennie & Raghavan, 2006).
Calculations of heat exchangers can be divided into two categories, namely, thermo-
hydraulic and mechanical design calculations. The subject of thermal and hydraulic
calculations is to determine heat-transfer rates, heat transfer area and pressure drops needed
for equipment sizing. Mechanical design calculations are concerned with detailed
equipment specifications, including stress analyses.
Heat exchanger problems may also be considering as rating or design problems. In a rating
problem, should be determined whether particular exchanger will perform a given heat-
transfer duty adequately. It is of no importance whether the exchanger physically exists or
whether it is specified only on paper. In a design problem, one must determine the
specifications for a heat exchanger that will handle a given heat-transfer duty. A rating
problem also arises when it is desired to use an existing heat exchanger in a new or
modified application (Serth, 2007). A particular application will dictate the rules that one
must follow to obtain the best design commensurate with economic considerations, size,
weight, etc. They all must bee considered in practice (Holman, 2009; Shokouhmand et al.,
2008; Rennie & Raghavan, 2006).
2. Thermal design of the heat exchangers

The suitable use of heat transfer knowledge in the design of practical heat transfer equipment
is an art. Designers must be constantly aware of the differences between the idealized
conditions under which the fundamental knowledge was obtained and the real conditions of
their design and its environment. The result must satisfy process and operational requirements
and do so cost-effectively. An important element of any design process is to consider and
compensate the consequences of error in the basic knowledge, in its subsequent incorporation
into a design method. Heat exchanger design is not a extremely accurate procedure under
the best of conditions (Shilling et al., 1999).
The design of a heat exchanger usually consists the subsequent steps:
1. Specification of the process conditions, e.g. flow compositions, flow rates, temperatures,
pressures.
2. Obtaining of the required physical properties over the temperature and pressure ranges
of interest obtained.
3. Choosing the type of heat exchanger that is going to be used.
4. An initial estimation of the size of the heat exchanger that is made, using a heat transfer
coefficient appropriate to the fluids, the process, and the equipment.
5. A first design is chosen, complete in all details necessary to carry out the design
calculations.
6. Evaluation of ability to perform the process specifications with respect to both heat
transfer and pressure drop as the design of heat exchanger is chosen.
7. Described in point above procedure can be repeated to new heat exchanger design if it
is necessary. The final design should meet process requirements within reasonable error
expectations.
The calculation of convective heat transfer coefficients constitutes a crucial issue in
designing and sizing any type of heat exchange device. Thus its correct determining permits
for the proper selection of heat transfer area during designing of heat exchangers and
calculation of the fluids outlet temperature. A lot efforts have been made during
experimental investigations of pressure drop and heat transfer in different types of heat
exchangers to obtain proper heat transfer correlation formulas.
2.1 The Wilson plot technique to determine heat transfer correlations in heat
exchangers
One of the widely used methods for calculations of heat transfer coefficient is the Wilson
plot technique. This approach was developed by E.E. Wilson in 1915 in order to evaluate the
heat transfer coefficients in shell and tube condensers for the case of a vapour condensing
outside by means of a cooling liquid flow inside (Viegas et al., 1998; Kumar et al., 2001; Rose,
2004; Fernández-Seara et al., 2007). It is based on the separation of the overall thermal
resistance into the inside convective thermal resistance and the remaining thermal
resistances participating in the heat transfer process. The overall thermal resistance Roveral of
the condensation process in a shell-and-tubes heat exchanger can be expressed as the sum of
three constituent thermal resistances: Rin – the internal convection, Rwall – the tube wall and
Ro – the external convection, presented in Eq. (1).
Rtotal = Rin + Rwall + Ro . (1)

Experimental Prediction of Heat Transfer Correlations in Heat Exchangers 295
The thermal resistances of the fouling in Eq. (1) was neglected. Employing the expressions
for the thermal resistances in Eq. (1), the overall thermal resistance can be rewritten as
follows:
⎛d ⎞
ln ⎜ o ⎟
Rtotal =
1
+ ⎝ din ⎠ + 1 . (2)
hin Ain 2πλwall Lwall ho Ao
where hin and ho is the internal and outer heat transfer coefficients, din and do – the inner and
outer tube diameters, λwall is the tube material thermal conductivity, Lwall is the tube length
and Ai and Ao are the inner and outer tube surface areas, respectively.
On the other hand, the overall thermal resistance can be written as a function of the overall
heat transfer coefficient referred to the inner or outer tube surfaces and the corresponding
areas. Assuming this the overall thermal resistance is expressed as a function of the overall
heat transfer coefficient referred to the inner or outer surface Uin/o and the inner or outer
surface area Ain/o (Eq. 3)
1
Rtotal = . (3)
Uin o Ain o
Taking into account the specific conditions of a shell and tube condenser Wilson assumed
that if the mass flow of the cooling liquid was modified, then the change in the overall
thermal resistance would be mainly due to the variation of the in-tube heat transfer
coefficient, while the remaining thermal resistances remained nearly constant. Therefore, as
specified in Eq. (4) the thermal resistances outside of the tubes and the tube wall could be
regarded as constant:
Rwall + Ro = C 1 . (4)
Wilson determined that for the case of fully developed turbulent flow inside a tube of
circular cross-section, the heat transfer coefficient was proportional to a power of the
reduced velocity wr which describes the variations of the fluid property and the tube
diameter. Thus, the heat transfer coefficient could be written in form:
hin = C 2 wrn , (5)
where C2 is a constant, wr – the reduced fluid velocity and n – velocity exponent. In this case
the convective thermal resistance related to the inner tube flow is proportional to 1 wrn .
Inserting Eqs. (4) and (5) into Eq. (1), the overall thermal resistance becomes the linear
function of 1 wrn , where C1 is the intercept and 1 (C 2 Ain ) is the slope of the straight line.
The overall thermal resistance can be calculated using experimental data using the following
formula:
Q = U o Ao ΔTlm . (6)
Substituting Eq. (3) into Eq. (6), and assuming Q = m 1c p 1 (T1outlet − T1inlet ) , where m
1 is the
mass flow rate of cooling liquid, cp1 – average specific heat of cooling liquid, and T1inlet,
T1outlet, are inlet and outlet temperatures of cooling liquid, respectively, yields to
ΔTlm . (7)
Rtotal =
1c p 1 (T1outlet − T1inlet )
m
As the constants C1 and C2 are determined from straight–line approximation of measured

data, to evaluation, for a given mass flow rate, the internal heat transfer coefficient can be
used Eq. (5) and internal heat transfer coefficient Eq. (8):
1 . (8)
ho =
Ao (C1 − Rwall )
The original Wilson plot technique depends on the knowledge of the overall thermal
resistance, that involves to remain of one fluid flow rate constant and varying flow rate of
the another fluid.
Approach of Wilson plot technique to determine constant in heat transfer correlation formula
for helically coiled tube-in-tube heat exchanger is presented by Sobota (Sobota, 2011).
3. Experimental prediction of heat transfer correlations in heat exchangers

In this chapter, the experimental and numerical investigations of helically coiled tube-in-
tube heat exchanger are presented. Calculations of unknown constants and exponents in
correlations formula for Nusselt number have been performed with least squares method
using Levenberg-Marquardt algorithm. Presented method allows for determining unknown
values of constants and exponents in correlation formulas for Nusselt number. This method
enables to obtain values of heat transfer coefficient on both sides of the barrier simultaneously
without earlier indirect calculations of the overall heat transfer.
3.1 Mathematical formulation of the inverse problem

The issue consisting of simultaneous determining of the heat transfer coefficient on the
cooling and heating liquid is ranked among inverse heat transfer problems (IHCP) (Beck et
al., 1985). In discussed methodology the knowledge of correlation formula form for heat
transfer coefficient on the both sides of the heat transfer surface, for counter-flow and
parallel-flow heat exchanger, was assumed to be known. An unknown value of the
parameters in correlation formulas was hidden in equations for outlet temperature of the
liquids (Nashchokin, 1980):
a) parallel-flow arrangement of heat exchanger
- heating liquid
⎛ W ⎞ k ⋅F
− ⎜ 1 + 1 ⎟⋅ c
1−e ⎝ W2 ⎠ W1
T1,′′ calc = T1,′ meas − (T1,′ meas − T2,′ meas ) (9)
W
1+ 1
W2
- cooling liquid
⎛ W ⎞ k ⋅F
− ⎜ 1 + 1 ⎟⋅ c
W W1
W 1−e ⎝ 2 ⎠
T2′′,calc = T2′ ,meas + (T1,′ meas − T2,′ meas ) 1 (10)
W2 W1
1+
W2
b) counterflow arrangement of heat exchanger

- heating liquid
⎛ W ⎞ k ⋅F
− ⎜ 1 − 1 ⎟⋅ c
1−e ⎝ W2 ⎠ W1
T1,′′ calc = T1,′ meas − (T1,′ meas − T2′ ,meas ) (11)
⎛ W1 ⎞ k ⋅Fc
W −⎜ 1− W2 ⎟⎠⋅ W1
1− 1 e ⎝
W2
- cooling liquid
⎛ W ⎞ k ⋅F
− ⎜ 1 − 1 ⎟⋅ c
W 1−e ⎝ W2 ⎠ W1
T2′′,calc = T2′ ,meas + (T1,′ meas − T2,′ meas ) 1 (12)
W2 ⎛ W1 ⎞ k ⋅Fc
W −⎜ 1− W2 ⎟⎠⋅ W1
1− 1 e ⎝
W2
′
where T1,meas and T2′ ,meas – measured temperature of the heating and cooling liquid at the
′′
inlet of the helically coiled heat exchanger respectively, °C; T1,calc and T2′′,calc – calculated
temperature of the heating and cooling liquid at the outlet of the helically coiled heat
exchanger respectively, °C; and expression W = V⋅ρ⋅cv is called as water equivalent.
The minimum of the square of the differences between measured and calculated from
analytical formula temperatures of the hot fluid and differences between measured and
calculated from analytical formula temperatures of the cold fluid at the outlet of heat
exchanger was searching for:
n
S (α ) = ∑ ⎡(T1,′′ meas − T1,′′ calc ) + (T2′′,meas − T2′′,calc ) ⎤ → min .
2 2
(13)
⎢⎣ ⎥⎦
i =1
In analysed example the solution of the nonlinear least square problem was searching for.
Determining the values of the constants, and indirectly the values of heat transfer
coefficient, was carried out using least squares method with modified Levenberg-Marquardt
algorithm (Visual Numerics, 2007; Press et al., 1996).
In Levenberg-Marquardt algorithm unknown are formed an column vector x = (x1, x2, ...,
xm)T, for which the sum becomes minimum
n
2
S ( x ) = ∑ ⎣⎡ ri ( x ) ⎦⎤ → min , (14)
i =1
where ri ( x ) = Tmeas
′′ − Tcalc
′′ .
The method performs the k-th iteration as
x(
k + 1)
= x ( ) + δ( ) ,
k k
(15)
where
−1
⎡ k
( ) ⎤
( J( ) )
T T
δ( ) = ⎢ J(m ) J(m ) + μ ( )In ⎥ ⎣⎡ f − Tm ( x ) ⎤⎦ ,
k k k k
k = 0, 1, ... . (16)
⎣ ⎦
The symbols f and Tm(x) stand for vector of measured and vector of computed temperature,
respectively.
Jacobian determinant is described by formula
∂rm ( x ) ⎡⎛ ∂r ( x ) ⎞ ⎤
Jm = = ⎢⎜ i ⎟⎥ , (17)
∂x T
⎢⎣⎜⎝ ∂x j ⎟⎠ ⎥⎦
mxn
where i = 1, …, n, j = 1, …, m, D(k) denotes diagonal matrix with positive elements. Quite

often D(k) = Im, where Im is identity matrix.
The value of the parameter μ(k) → 0 when x(k) → x*. In the proximity of minimum x* the
iteration step in the Levenberg-Marquardt method is almost the same as in the Gauss-
Newton method. The computation programs for solving the non-linear least square problem
by the Levenberg-Marquardt method are described in (Lawson & Hanson, 1974) and in the
IMSL Library (Visual Numerics, 2007).
3.2 Correlations for Nusselt number

Although curved pipes are used in a wide range of applications, flow in curved pipes is
relatively less well known than that in straight ducts. A helical coil can be geometrically
described by the coil radius R, the pipe radius r, and the coil pitch 2πb (Fig. 1).
Fig. 1. Schematic representation of a helical pipe with its main geometrical parameters:
r – tube radius; R – coil radius; 2πb – coil pitch
The observations on the complexity of a flow in such a channels allowed to notice the effect
of curvature on the fluid flow regime which occurs delaying the transition from laminar to
transitional flow to a higher Reynolds number with respect to straight pipes (Ito, 1959;
Schmidt, 1967). Using data from his own experiments as well as that from previous
investigations, Ito developed the following empirical relation to determine the critical
Reynolds number for the range of curvature ratios of 1/15 to 1/860:
( R)
0.32
Recrit = 20000 r , (18)
whilst Schmidt suggested the form of critical Reynolds number listed below:
⎡ ⎛r⎞ ⎤
0.45
Recrit = 2300 ⋅ ⎢1 + 8.6 ⎜ ⎟ ⎥ . (19)
⎢⎣ ⎝ R ⎠ ⎥⎦
For curvature ratios δ = r/R less than 1/860, the critical Reynolds number was found to
correspond with that of a straight pipe.
Equation for Nusselt number that are most commonly found in literature concerning heat
transfer in curved or helical tubes can be assumed formula developed by Schmidt (Schmidt,
1967):
- for laminar regime
⎡ ⎛r⎞
0.194 ⎤
⎡ ⎛r⎞ ⎤
0.9 ⎢0.5 + 0.2903⎜ ⎟
⎢ ⎝R⎠
⎥
⎥⎦
1
Nu = 3.65 + 0.08 ⋅ ⎢ 1 + 0.8 ⎜ ⎟ ⎥ ⋅ Re ⎣ ⋅ Pr 3 , (20)
⎣⎢ R
⎝ ⎠ ⎦⎥
where Reynolds number varies 100 < Re < Recrit, where Recrit is described by Eq. (19).
- for turbulent flow
⎡ 0.1 ⎤
⎡ 1⎤ ⎛r⎞
⎢0.8 + 0.22 ⋅⎜ ⎟ ⎥ 1
⎢ ⎛ r ⎞ ⎛ r ⎞3 ⎥ ⎢ ⎝R⎠ ⎥⎦
Nu = 0.023 ⋅ 1 + 14.8 ⋅ ⎜ 1 + ⎟ ⋅ ⎜ ⎟ ⋅ Re ⎣ ⋅ Pr 3 , Recrit < Re ≤ 22000, (21)
⎢ ⎝ R⎠ ⎝R⎠ ⎥
⎣⎢ ⎦⎥
⎡ ⎛ r ⎞⎛ r ⎞ ⎤
0.8
Nu = 0.023 ⋅ ⎢1 + 3.6 ⋅ ⎜ 1 − ⎟⎜ ⎟ ⎥ ⋅ Re0.8 ⋅ Pr 1 3 , 22000< Re ≤ 150000. (22)
⎢⎣ ⎝ R ⎠⎝ R ⎠ ⎥⎦
Another formulas for Nusselt number that is valid for turbulent flow was invented by Seban
and McLaughlin (Seban & McLaughlin, 1963):
0.1
⎛r⎞
Nu = 0.023 ⋅ ⎜ ⎟ ⋅ Re0.85 ⋅ Pr 0.4 (23)
⎝R⎠
and Rogers and Mayhew (Rogers & Mayhew, 1964)
0.1
⎛r⎞
Nu = 0.021 ⋅ ⎜ ⎟ ⋅ Re0.85 ⋅ Pr 0.4 . (24)
⎝R⎠
Eqs. (23) and (24) have simple structure that makes them easy to use.
Other widely used method is that of Seider and Tate, who recommended the following
expression for applications with large property variations from the bulk flow to the wall
temperature:
0.14
⎛μ ⎞
Nu = 0.027 ⋅ Re0.8 ⋅ Pr 1 3 ⎜ bulk ⎟ (25)
⎝ μ wall ⎠
for 0.7 < Pr < 16000, Re > 10000 and L/D > 10.
For more accurate calculations in fully developed turbulent flow it is recommended to use
Petukhov heat transfer correlation that is valid for 0.5 < Pr < 2000 and 10000 < Re < 5000000:
0.14
Nu =
( f 2 ) Re Pr ⎛ μbulk ⎞
, (26)
⎜ ⎟
1.07 + 12.7 ( f 2 ) ( Pr 2 3 − 1 ) ⎝ μ wall ⎠
0.5
where friction factor f can be obtained from the Moody diagram or from Petukhov’s friction
factor correlation that is valid for 3000 < Re < 5000000:
1
f = . (27)
1.58ln ( Re ) − 3.28
Another heat transfer correlation commonly used is that of Gnielinski (Smith, 1997), which
extends the Petukhov correlation down into the transition regime:
0.14
⎛ f ⎞ Re − 100 Pr ⎛ Pr
⎜ 2 ⎟( ) ⎜ Pr ⎞⎟ ⎛ μbulk ⎞
0.14
Nu = ⎝ ⎠ ⎝ w⎠
, (28)
⎜ ⎟
⎝ μ wall ⎠
0.5
⎛ f ⎞ ⎛ 23 ⎞
1 + 12.7 ⎜ ⎟ ⎜ Pr − 1 ⎟
⎝ 2⎠ ⎝ ⎠
where
0.27
0.3164 r ⎛ η wall ⎞
+ 0.03 ⋅⎜ ⎟
Re 0.25 R ⎝ ηbulk ⎠
fc = . (29)
4
Worth to be mentioned are the following heat transfer correlations:
- Mori and Nakayama (Manglik, 2003)
1
⎡ ⎛ ⎛ r ⎞2.5 ⎞
0.167 ⎤ 5
⎛ r ⎞ 12 ⎢ ⎥ Pr 0.4 Re 6
⎜ ⎟ ⎢ 1 + 0.061 ⎜ Re ⎟
⎝R⎠ ⎜ ⎜⎝ R ⎟⎠ ⎟ ⎥
⎣ ⎝ ⎠ ⎦
Nu = . (30)
41.0
- Jeschke, which is the oldest one (Rogers & Mayhew, 1964)
⎡ r⎤
Nu = 0.045 ⋅ ⎢1 + 3.54 ⎥ ⋅ Re0.76 ⋅ Pr 0.4 . (31)
⎣ R⎦
This correlation was developed as a result of transposition formula that was valid for air
flow through helical two loop heat exchanger into water.
- Kirpikov (Nashchokin, 1980)
0.21
⎛r⎞
Nu = 0.0456 ⋅ ⎜ ⎟ Re 0.76 ⋅ Pr 0.4 , 10000< Re ≤ 45000 (32)
⎝R⎠
and Mikheev (Nashchokin, 1980)
⎡ ⎛ r ⎞⎤
Nu = 0.021 ⋅ ⎢1 + 1.77 ⋅ ⎜ ⎟ ⎥ ⋅ Re 0.85 ⋅ Pr 0.43 . (33)
⎣ ⎝ R ⎠⎦
The discussed correlations can be helpful in selecting the form of the heat transfer
correlation in which certain coefficients and exponents are to be determined.
3.3 Experimental setup

To determine the Nusselt number correlation for forced convection in helically coiled tube-
in-tube heat exchanger an experimental setup was build. It consisted of copper made heat
exchanger, electric heater and circulating pumps.
3.3.1 Heat exchanger

Helical coil heat exchangers are one of the most common equipment found in many
industrial applications ranging from chemical and food industries, power production,
electronics, environmental engineering, air-conditioning, waste heat recovery and cryogenic
processes. Helical coils are extensively used as heat exchangers and reactors due to higher
heat and mass transfer coefficients, narrow residence time distributions and compact
structure. The modification of the flow in the helically coiled tubes is due to the centrifugal
forces (Dean, 1927, Dean, 1928). The curvature of the tube produces a secondary flow field
with a circulatory motion, which causes the fluid particles to move toward the core region of
the tube. The secondary flow increases heat transfer rates as it reduces the temperature
gradient across the cross-section of the tube. Thus there is an additional convective heat
transfer mechanism, perpendicular to the main flow, which does not exist in conventional
heat exchangers. An extensive review of fluid flow and heat transfer in helical pipes has
been presented in the literature (Kumar et al. 2008; Shah & Joshi, 1987).
The examined heat exchanger was constructed from copper tubing and typical connections
were made of copper also and consisted of 6.5 loop. The outer tube of the heat exchanger
had an outer diameter of 35 mm and a wall thickness of 1.5 mm. The inner tube had an outer
diameter of 22 mm with wall thickness of 1 mm.
Fig. 2. Schematic of the examined heat exchanger with basic geometry

Coil had a radius of curvature, measured from the centre of the inner tube, of 137.5 mm. Its
thermal power Qn was equal to 14 kW, pressure drop Δp = 0.32 bar and volumetric flow V =
2.3 m3/h. Calculated heat transfer area Fc of the heat exchanger was 0.3952 m2. The heat
exchanger was very carefully insulated with polyurethane foam to avoid heat losses to the
surroundings.
3.3.2 Experimental apparatus

The heat exchanger was tested in the setup presented in Fig. 3. This stand consisted of
electrical heater (21.6 kW of thermal power) equipped with circulating pump and expansion
vessel, hydraulic couple and examined helically coiled tube-in-tube heat exchanger (Fig. 2).
Fig. 3. Diagram of the experimental setup; 1 – helically coiled heat exchanger, 2 –water
heater, 3 – hydraulic couple, 4 – circulating pump, 5 – coolant circulating pump, 6 – cold
feed water tank, 7 – boiler circulating pump, FMH – hot water flowmeter, FMC – cooling
water flowmeter, TI – temperature sensors
Hydraulic couple divided hydraulic system into two independent circuits – heater’s and
heat sink’s. In heat sink circuit, which consisted of hydraulic couple and heat exchanger hot
water flow was forced by circulating pump. Nominal volumetric flow of the hot water
through the pump was equal to 1.8 m3/h and maximum head 2 m. Circulating pump of the
same type was used to pump mains cold water through annular tube of heat exchanger. To
provide steady flow of cooling water through the heat exchanger a compensation vessel was
mounted on the wall at a height of 2 m. The inlet and outlet temperatures of the hot and
cold water were measured using precalibrated K-type thermocouples with high accuracy.
The flow of hot water in channel of circular cross-section was controlled by axial turbine
flowmeter allowing flows to be measured between 2 and 40 l/min. The flow rate of cold
water in annular cross-section channel was controlled by an identical flowmeter. All the
fluids properties were assessed at the arithmetic mean temperature of the fluids (average of
inlet and outlet temperatures). Temperature and flow data was recorded using a data
acquisition system connected to a computer.
3.3.3 Experimental procedure

Volumetric flow rate of the cold water, in the annulus, was kept on constant level, while hot
water volumetric flow, in the inner tube of circular cross-section, was varied. The range of
hot water flow rates from 3.33 l/min to 20 l/min and cold water from 4 l/min to 8 l/min
were used. All possible combinations of these flow rates in both the annulus and the inner
tube were examined. These were done for both coils, and in parallel-flow and counter-flow
configurations. Temperature data was recorded every one second. For further numerical
calculations only results of measurements after the temperatures achieved steady values
were taken. Next, experimental data were used for simultaneous calculations of constants
and exponents in heat transfer correlation formulas for Nusselt number on both sides of
heat transfer surface.
3.4 Results
Investigations of helically coiled tube-in-tube heat exchanger were conducted in steady state
conditions for wide range of temperature and volumetric flow changes of working fluids.
The hot fluid, in the channel with circular cross-section, flows in turbulent regime. It was
assumed that in this case the dependence for Nusselt number formula will be described by
equation shown below (Rogers & Mayhew, 1964; Hewitt, 1994):
⎛ d ⎞
Nu1 = ⎜ 1 + 3.5 1in ⎟ 0.023Re10.8 Pr10.333 , (34)
⎝ D ⎠
where d1in – denotes inner diameter of the tube with circular cross-section, m; D = 2R – heat
exchanger coil mean diameter, m.
While in the case of laminar flow in annular channel the formula for Nusselt number has the
following form (Schmidt, 1967):
−0.8
d ⎞ ⎡ d ⎞ ⎤
0.33
⎛ ⎛ d ⎞ ⎛
Nu2 = ⎜ 1 + 3.5 ⋅ h ⎟ ⋅ ⎢ 3.66 + 1.2 ⋅ ⎜ 1 ⎟ + 1.6 ⋅ ⎜ Re 2 ⋅ Pr2 ⋅ h ⎟ ⎥ , (35)
⎝ D⎠ ⎢ ⎝ d2 in ⎠ ⎝ L⎠ ⎥
⎣ ⎦
where dh – denotes equivalent diameter of the annular channel, m; d1 – outer diameter of
the tube with circular cross-section, m; d2in – inner diameter of the tube with annular
cross-section, m; L – total length of the heat exchanger.
In Eqs. (34) and (35) components (1+3.5⋅(d1in/D)) and (1+3.5⋅(dh/D)) takes into account the
geometry of the helically coiled tube-in-tube heat exchanger. Expression (3.66+1.2⋅(d1/d2in)-0.8)
is a correction for fluid flow in annular channel in examined heat exchanger.
It was assumed that in the first stage of calculation the unknown parameters on the left side
of the Reynolds number in equation (34) and (35) will be searched for. All calculations will
be carried out on the both sides of the heat transfer surface simultaneously. After taking into
consideration the unknown parameters the formulas mentioned above have the form:
⎛ d ⎞
Nu1 = ⎜ 1 + 3.5 ⋅ 1in ⎟ ⋅ A1 ⋅ Re0.8 0.33
1 Pr1 (36)
⎝ D ⎠
and
−0.8
d ⎞ ⎡ d ⎞ ⎤
0.33
⎛ ⎛ d ⎞ ⎛
Nu2 = ⎜ 1 + 3.5 ⋅ h ⎟ ⋅ ⎢ 3.66 + 1.2 ⋅ ⎜ 1 ⎟ + A2 ⋅ ⎜ Re 2 ⋅ Pr2 ⋅ h ⎟ ⎥ . (37)
⎝ D⎠ ⎢ ⎝ d2 in ⎠ ⎝ L ⎠ ⎥
⎣ ⎦
The values of unknown parameters A1 and A2 (Table 1) were obtained as a result of the
performed calculations. And next were used for drawing distributions of the Nusselt
number as a function of Reynolds number for hot and cold fluid in counter flow (Fig. 4a)
and parallel flow (Fig. 4b).
Changes of Nusselt number in circular channel, expressed by formula (34) and Eq. (35) are
very much the same for both arrangement of helically coiled tube-in-tube heat exchanger as
it is shown in Fig. 4. In the following, Eq. (34) will be used for the calculation of the Nusselt
number in the circular duct.
Flow type in helically coiled tube-in-tube heat exchanger

Parallel flow Counter flow
Equation (36)
A1 = 0.0202 A1 = 0.0188
Equation (37)
A2 = 30.2301 A2 = 61.8249
Table 1. Values of the constants in Eq. (36) and Eq. (37) calculated with Levenberg-
Marquardt method
a) b)
Fig. 4. Distribution of the Nusselt number as a function of Reynolds number in circular
channel of the helically coiled tube-in-tube heat exchanger; a) parallel-flow, b) counterflow;
○ Eq. (34), ◊ Eq. (36)
For next stage of calculations Eq. (34) and (35) were modified. Constant Ai and exponent Bi,
where i = 1, 2 of Reynolds number were investigated in this case.
⎛ d ⎞
Nu1 = ⎜ 1 + 3.5 ⋅ 1in ⎟ ⋅ A1 ⋅ Re1B1 Pr10.33 (38)
⎝ D ⎠
and
−0.8
d ⎞ ⎡ d ⎞ 2⎤
B
⎛ ⎛ d ⎞ ⎛
Nu2 = ⎜ 1 + 3.5 h ⎟ ⋅ ⎢ 3.66 + 1.2 ⋅ ⎜ 1 ⎟ + A2 ⋅ ⎜ Re 2 ⋅ Pr2 ⋅ h ⎟ ⎥ . (39)
⎝ D⎠ ⎢ ⎝ d2 in ⎠ ⎝ L⎠ ⎥
⎣ ⎦
As a first constant A2 in formula (39) for Nusselt number in annular channel for parallel flow
in examined heat exchanger was determined and then constant A2 and exponent B2
simultaneously. During the calculations the flow in circular channel was described by
formula (34).
Identical calculations were carried out for Eq. (38), which was modified in order to adapt it
to describe heat transfer in annular channel:
⎛ d ⎞
Nu1 = ⎜ 1 + 3.5 ⋅ h ⎟ ⋅ A1 ⋅ Re 1B1 Pr10.33 . (40)
⎝ D⎠
Also in this case as a first was calculated constant A1, and next constant A1 and exponent B1
in Eq. (40).
This procedure was to test, which of the equations will be better to map the set of
experimental data of the working fluid in annular channel – less complicated and correct in
case of turbulent flow – Eq. (34) or Eq. (35) describing the heat transfer for fluid flow in
laminar range.
−0.8
Nu2 = ( 1 + 3.5 ⋅ dh D ) ⋅ ⎡⎢ 3.66 + 1.2 ⋅ ( d1 d2 in ) + A2 ⋅ ( Re 2 ⋅ Pr2 ⋅ dh L ) 2 ⎤⎥
B
⎣ ⎦
A2 = 19.7433
A2 = 3.5754 B2 = 0.8229
Nu1 = ( 1 + 3.5 ⋅ dh D ) ⋅ A1 ⋅ Re 1B1 Pr10.33
A1 = 0.0566
A1 = 0.0269 B1 = 0.8926
Table 2. Values of constant and exponents calculated with Levenberg-Marquardt for parallel
flow in helically coiled tube-in-tube heat exchanger
a) b)
Fig. 5. Comparison of Nusselt number changes in annular channel (parallel flow) for Eqs.
(39) and (40). Using Levenberg-Marquardt method was determined: a) one coefficient
and b) two coefficients , experimental points ◊
In case when examined helically coiled tube-in-tube heat exchanger was working as a
counter flow analogous calculations, using least squares method with modified Levenberg-
Marquardt, were carried out to determine constants and exponents in correlation formulas
for Nusselt number. Results are shown in Table 3 – forms of correlation formulas and
determined values of the constants and exponents. Fig. 6. shows changes of two different
formulas for Nusselt number as a function of Reynolds number in annular channel. Also in
this case as a first was determined value of a constant, and as a second value of constant and
exponent.
−0.8
Nu2 = ( 1 + 3.5 ⋅ dh D ) ⋅ ⎡⎢ 3.66 + 1.2 ⋅ ( d1 d2 in ) + A2 ⋅ ( Re 2 ⋅ Pr2 ⋅ dh L ) 2 ⎤⎥
B
⎣ ⎦
A2 = 25.397
A2 = 14.1959 B2 = 0.4997
Nu1 = ( 1 + 3.5 ⋅ dh D ) ⋅ A1 ⋅ Re 1B1 Pr10.33
A1 = 0.0827
A1 = 0.3787 B1 = 0.6052
Table 3. Values of constant and exponents calculated with Levenberg-Marquardt for counter
flow in helically coiled tube-in-tube heat exchanger
a) b)
Fig. 6. Comparison of Nusselt number changes in annular channel (counter flow) for Eqs.
(38) and (39). Using Levenberg-Marquardt method was determined: a) one coefficient
and b) two coefficients , experimental points
Analyzing changes of the Nusselt number in annular channel (hot fluid) shown on the Fig.
5a and Fig. 6a, drawn as dotted line, the best fit curves to experimental points can be noticed
when the Levenberg-Marquardt method is used to determine values of two unknown
parameters – constant and exponent in Eq. (39) than in case where only value of the one
parameter was investigated.
The differences between calculated values of constants and exponents in heat transfer
correlation formulas for counterflow and parallel-flow configuration of the examined heat
exchanger are result of the larger average temperature difference between the two fluids.
Comparison of the Nusselt number changes with calculated one and two unknown
parameters for turbulent (40) and laminar (39) form of the Nusselt formula leads to the
conclusion that Eq. (40) better describes the kind of the fluid flows in annular channel for
heat exchanger operating in parallel and counter flow.
4. Conclusion
In this paper methodology which allows for numerical determination unknown parameters
in correlation formulas for Nusselt number and heat transfer coefficients on the hot and cold
fluid side simultaneously was presented. Calculations were carried out on the basis of
gathered experimental data for parallel and counter flow of working fluid in helically coiled
tube-in-tube heat exchanger. The changes of Nusselt number in circular channel (turbulent
flow) and annular channel (laminar flow) as a function of Reynolds number were presented
also.
Described methodology for determining constants and exponents in correlation formulas for
Nusselt number can be used in designing of the different heat exchangers types and shapes
of heat transfer surface.
5. References
Beck, J. V.; Blackwell, B.; Clair, Ch. R. St. (1985). Inverse Heat Conduction. Ill-posed Problems.
John Wiley & Sons, ISBN 0-47-108319-4, New York, USA
Dean W.R. (1928). The Streamline Motion of Fluid in a Curved Pipe (second paper), Philos.
Mag.,Vol. 7, pp. 673–695, ISSN 1941-5982
Dean, W.R. (1927). Note on the Motion of Fluid in a Curved Pipe, Philos. Mag., Vol. 4, pp.
208–223, ISSN 1941-5982
Fernández-Seara, J.; Uhía F.J.; Sieres, J.; Campo, A. (2007). A General Review of the Wilson Plot
Method and its Modifications to Determine Convection Coefficients in Heat
Exchange Devices. Applied Thermal Engineering, Vol.27, pp. 2745–2757, ISSN 1359-4311
Hewitt, G.F.; Shires, G.L.; Bott T.R. (1994). Process Heat Transfer. Begell House Publishers,
ISBN 0-8493-9918-1, New York, USA
Holman, J.P. (2009). Heat Transfer. McGraw-Hill, ISBN 0-07-114320-3, New York, USA
Ito, H. (1959). Friction factors for turbulent flow in curved pipes, Transactions of ASME D 81,
pp. 123-132, ISSN 0021-9223
Kumar, R.; Varma, H.K.; Agrawal, K.N.; Mohanty, B. (2001). A Comprehensive Study of
Modified Wilson Plot Technique to Determine the Heat Transfer Coefficient During
Condensation of Steam and R-134a Over Single Horizontal Plain and Finned Tubes.
Heat Transfer Engineering, Vol.22, pp.3–12, ISSN 0145-7632
Kumar, V.; Faizee, B.; Mridha, M.; Nigam, K.D.P. (2008). Numerical Studies of a Tube-in-
Tube Helically Coiled Heat Exchanger. Chemical Engineering and Processing, Vol.47,
pp. 2287–2295, ISSN 0255-2701
Kuppan, T. (2000). Heat Exchanger Design Handbook, Marcel Dekker, Inc, ISBN 0-8247-9787-6,
New York, USA
Lawson, C.; Hanson R. (1995). Solving Least Squares Problems, ISBN 0-89871-356-0, Prentice
Hall, Englewood Cliffs, USA
Manglik, R.M. (2003). HeatTransfer Enhancement, In: Heat Transfer Textbook, A. Bejan and
Kraus A.D., (Eds.), John Wiley & Sons, ISBN 0-471-39015-1, New York, USA
Nashchokin, V.V. (1980). Engineering Thermodynamics and Heat Transfer, Central Books Ltd,
ISBN 0-71-471523-9, London, Great Britain
Pitts, D.R.; Sissom, L.E. (1998). Shaum’s Outline of Theory and Problems of Heat Transfer.
McGraw-Hill, ISBN 0-07-050207-2, New York, USA
Press, W.H.; Teukolsky, S.A.; Vetterling W.T.; Flannery B.P. (1996). Numerical Recipes in
Fortran 77: The Art of Scientific Computing. Cambridge University Press, ISBN 0-521-
43064-X, New York, USA
Rennie, T.J.; Raghavan, V.G.S. (2006). Effect of Fuid Thermal Properties on the Heat Transfer
Characteristics in a Double-Pipe Helical Heat Exchanger. International Journal of
Thermal Sciences, Vol.45, pp. 1158–1165, ISSN 1290-0729
Rogers, G.F.C.; Mayhew Y. R. (1964). Heat Transfer And Pressure Loss In Helically Coiled
Tubes With Turbulent Flow. International Journal of Heat Mass Transfer, Vol. 7, pp.
1207 – 1216, ISSN 0017-9310
Rose, J.W. (2004). Heat-Transfer Coefficients, Wilson Plots and Accuracy of Thermal
Measurements. Experimental Thermal and Fluid Science, Vol.28, pp.77–86, ISSN 0894-
1777
Schmidt, E.F. (1967). Wärmeübergang und Druckverlust in Rohrschlangen. Chemie Ingenieur
- Technik, 39 Jahrgang, Heft 13, pp. 781 – 789, ISSN 0009-286X
Seban,. R.A.; McLaughlin, E.F. (1963). Heat Transfer in Tube Coils with Laminar and
Turbulent Flow. International Journal of Heat Mass Transfer, Vol. 6, pp. 387 – 395,
ISSN 0017-9310
Serth, R.W. (2007). Process Heat Transfer. Principles and Applications. Academic Press, ISBN
978-0-12-373588-1, Burlington, USA
Shah, R.K.; Joshi S.D. (1987). Convective Heat Transfer in Curved Ducts, In: Handbook of
Single-Phase Convective Heat Transfer, S. Kakac, R.K. Shah and W. Aung (Eds.), ISBN
0-47-181702-3, John Wiley & Sons, New York, USA
Shilling, R.L; Bell, K.J.; Bernhagen, P.M.; Flynn, T.M.; Goldschmidt, V.M.; Hrnjak, P.S.;
Standiford F.C.; Timmerhaus K.D. (1999). Heat-Transfer Equipment, In: Perry’s
Chemical Engineers’ Handbook, R.H. Perry and Green D.W., (Eds.), McGraw-Hill,
ISBN 0-07-049841-5, New York, USA
Shokouhmand, H.; Salimpour, M.R., Akhavan-Behabadi, M.A. (2008). Experimental
Investigation of Shell and Coiled Tube Heat Exchangers Using Wilson Plots.
International Communications in Heat and Mass Transfer, Vol.35, pp. 84–92, ISSN 0735-
1933
Smith, E. M. (1997). Thermal Design of Heat Exchangers: A Numerical Approach: Direct Sizing
and Stepwise Rating, and Transients, John Wiley & Sons, ISBN 0-47-001616-7, New
York, USA
Sobota, T. (2011). Determinig of Convective Heat Transfer Coefficients in Heat Exchangers
Using the Wilson Plot Technique, In: Thermal and Flow Proesses in Large Steam
Boilers. Modeling and Monitoring, J. Taler, (Ed.), ISBN 978-83-01-16479-9, Warszawa,
Poland (in Polish)
Viegas, R.M.C.M.; Rodríguez, M.; Luque, S.; Alvarez, J.R.; Coelhoso, I.M.; Crespo, J.P.S.G.
(1998). Mass Transfer Correlations in Membrane Extraction: Analysis of Wilson-
Plot Methodology. Journal of Membrane Science, Vol.145, pp.129-142, ISSN 0376-7388
Visual Numerics Inc. (2007). IMSL Fortran and C Application Development Tools, Vol. 1 and 2.
Houston, Texas, USA
17
High Temperature
Thermal Devices for Nuclear
Process Heat Transfer Applications
Piyush Sabharwall and Eung Soo Kim
Idaho National Laboratory
United States of America
1. Introduction
Thermal device is a broad term used to describe a component whose main objective is to
transport thermal energy across a system. The two thermal devices discussed in this chapter
are heat exchangers and thermosyphons.
1.1 Heat exchangers

A heat exchanger is a component used to transfer heat from one medium to another. The
media may be separated by a solid wall, so they never mix, or they may be in direct contact
(Kakac and Liu 2002). Heat exchangers generally have no external heat nor work interactions
and are typically used in the following applications:
• Space heating
• Refrigeration
• Air conditioning
• Power plants
• Chemical plants
• Petrochemical plants
• Petroleum refineries
• Natural gas processing.
1.1.1 High temperature heat exchangers

The high temperature heat exchanger technology is emerging in many industrial
applications such as gas turbines, chemical plants, and nuclear power plants. For this
reason, many researchers have been studying the development of heat exchangers that can
operate at high temperatures. Unlike typical heat exchangers, high temperature heat
exchangers require and exhibit some special characteristics to be operated in severe
environments (Sunden 2005):
• The transfer of radiative heat may have a significant role in the heat transfer process.
• Tube diameters and pitch should be larger for lower pressure drop. Adding a fan or
blower to work at high temperatures might be prohibitive.
• Fins are generally not used in high temperature units because of suspended dirt
particles that will foul and low available pressure drops. The advantage of fins is
negligible.
• Ceramic and high-temperature alloys (such as: Alloy 800H, 617, 230) are used for
construction materials.
• The thickness and mechanical design of selected materials are mainly governed by
thermal stress, but the extent of the materials oxidation, thermal shock bearing
capability, and erosion from suspended dirt particles, fouling, and corrosion because of
metallic salts, sulfates, etc., also need to be considered.
• Differential expansion is an important factor in high temperature units and should be
accounted for by using either expansion bellows or by using bayonet-type units.
Floating tube sheets cannot generally be used, because sealing gaskets or packing
materials do not work effectively at such high temperatures.
• Heat losses from the outside surface to the environment have to be considered in the
mechanical design of the unit and design of the foundation.
• Gases, air, liquid metals, or molten salts are preferred over steam for high temperature
heat transfer, because the latter require a very thick shell and tubes to contain its high
pressure.
Therefore, the thermal stress during startup, shutdown, and load fluctuations can be
significant for high temperature heat exchangers. The heat exchanger must be designed
accordingly for reliability and long life. The thermal capacitance should therefore be
reduced for high temperature heat exchangers for shorter startup time. High temperature
heat exchangers also require costly materials contributing to the high cost of balance of
power plant. Heat exchanger costs increase significantly for temperatures above 675°C.
1.1.2 Heat exchanger types and classifications

A variety of heat exchanger types with various features are used in industry. This
subsection generally explains how to classify and categorize them. According to Kakac and
Liu (2002), heat exchangers can be generally classified as follows:
1. Recuperator/Regenerator
a. Recuperations
b. Regenerations
2. Transfer Process
a. Direct contact
b. Indirect contact
3. Geometry of Construction
a. Tubular heat exchanger
i. Double pipe heat exchanger
- High pressure (in both sides)
ii. Shell and Tube heat exchanger
iii. Spiral tube type heat exchanger
b. Plate heat exchanger
i. Gasketed plate heat exchanger
ii. Spiral plate heat exchanger
c. Extended surface heat exchanger
i. Plate-fin heat exchanger
High Temperature Thermal Devices for Nuclear Process Heat Transfer Applications 311
ii. Tubular-fin heat exchanger (Gas to Liquid)

4. Heat Transfer Mechanism
a. Single phase convection on both sides
b. Single phase convection on one side, two phase convection on other side
c. Two phase convection on both sides
5. Flow Arrangement
a. Parallel flow
b. Counter flow
c. Cross flow
Table 1 shows the principle features for several types of heat exchangers (Shah and Sekulic
2003). According to this table, shell-and-tube, Bavex (plate heat exchanger), printed-circuit,
and Marbond are available for high temperature applications above 700°C.
The shell and tube heat exchanger is the most common type found in industry. This
exchanger is generally built of a bundle of round tubes mounted in a cylindrical shell with
the tube axis parallel to that of the shell. One fluid flows inside the tubes and the other fluid
flows across and along the tubes. The major components of this exchanger are tubes (or tube
bundles), shell, front-end head, rear-end head, baffles, and tube sheets (Shah and Sekulic
2003). The diameter of the outer shell in a shell and tube heat exchanger is greatly increased,
and a bank of tubes rather than a single central tube is used, as shown in Figure 1 (Sherman
and Chen 2008). Fluid is distributed to the tubes through a manifold and tube sheet. To
increase heat transfer efficiency, further modifications to the flow paths of the outer and
inner fluids can be accomplished by adding baffles to the shell to increase fluid contact with
the tubes, and by creating multiple flow paths or passes for the fluid flowing through the
tubes (Sherman and Chen 2008). These heat exchangers are used for gas-liquid heat transfer
applications, primarily when the operating temperature and/or pressure is very high (Shah
and Sekulic 2003).
Tube Shell
Outlet Inlet Baffles
Shell Tube
Outlet Inlet
Fig. 1. Shell and tube heat exchanger with baffles (Sherman and Chen 2008)
The Bevax hybrid welded-plate heat exchanger is a plate type heat exchanger that deploys
metal plates arranged in a stack-wise fashion and sealed with welds as shown in Figure 2.
This heat exchanger is reported to be operational at 900°C with pressures to 6 MPa on the
Temperature Maximum
312
Compactness Range Pressure Cleaning Corrosion Multistream Multipass
HX Type (m2/m 3 ) System Types Material ( °C)a (bar) b Method Resistance Capability c Capability d
Shell and Tube ~100 Liquid/Liquid, Stainless steel ~ +900 ~ 300 Mechanical, Good No Yes
Gas/Liquid, 2Phase (s/s), Ti, Inoloy, Chemical
Hastellroy,
graphite, polymer
Plate-and-frame ~200 Liquid/Liquid, s/s, Ti, Inoloy, -35 ~ +200 25 Mechanical Good Yes Yes
(gaskets) Gas/Liquid, 2Phase Hastellroy,
graphite, polymer
Partially ~200 Liquid/Liquid, s/s, Ti, Inoloy, -35 ~ +200 25 Mechanical, Good No Yes
welded plate Gas/Liquid, 2Phase Hastellroy Chemical
Fully welded ~200 Liquid/Liquid, s/s, Ti, Ni alloys -50 ~ + 350 40 Chemical Excellent No Yes
plate (Alfa Rex) Gas/Liquid, 2Phase
Brazed plate ~200 Liquid/Liquid, 2Phase s/s -195 ~ +220 30 Chemical Good No No
Bavex plate 200 ~ 300 Gas/Gas, Liquid/Liquid, s/s, Ni, Cu, Ti, -200 ~ +900 60 Mechanical, Good Yes Yes
2Phase special steels Chemical
Platular plate 200 Gas/Gas, Liquid/Liquid, s/s, Hastelloy, Ni ~700 40 Mechanical Good Yes Yes
2Phase alloys
Packinox plate ~300 Gas/Gas, Liquid/Liquid, s/s, Ti, Hastelloy, -200 ~ +700 300 Mechanical Good Yes Yes
2Phase Inconel
Spiral ~200 Liquid/Liquid, 2Phase s/s, Ti, Incoloy, ~400 25 Mechanical Good No No
Hastelloy
Brazed plate fin 800 ~ 1500 Gas/Gas, Liquid/Liquid, Al, s/s, Ni alloy ~650 90 Chemical Good Yes Yes
2Phase
Diffusion 700 ~ 800 Gas/Gas, Liquid/Liquid, Ti, s/s ~500 >200 Chemical Excellent Yes Yes
bonded plate 2Phase
fin
Printed circuit 200 ~ 5000 Gas/Gas, Liquid/Liquid, Ti, s/s -200 ~ +900 > 400 Chemical Excellent Yes Yes
2Phase
Polymer (e.g. 450 Gas/Liquid PVDF, PP ~150 6 Water Wash Excellent No No
channel plate)
Plate and shell — Liquid/Liquid s/s, Ti ~350 70 Mechanical, Excellent Yes Yes
Chemical
Marbond ~10,000 Gas/Gas, Liquid/Liquid, s/s, Ni, Ni alloys, -200 ~ +900 >400 Chemical Excellent Yes Yes
2Phase Ti
a. Heat exchanger operational temperature ranges.
Table 1. Principal features of several types of heat exchangers (Shah and Sekulic 2003)
b. Heat exchanger maximum applicable pressure.
c. Capability to connect several independent flow loops in a single heat exchanger.
d. Capability to split flow into several paths in the heat exchanger.
other flows between the plates from baffled plenums inside a pressure boundary (Fisher and
plate side. It is called a hybrid because one fluid is contained inside the plates while the
Sindelar 2008). It is reminiscent of a shell and tube arrangement with substantially greater
surface area. Plates can be produced up to 0.35 m wide and 16 m long (Fisher and Sindelar
2008). Other variants of the welded plate-type heat exchanger are produced, some of which
do not require external shells.
Plate-side entry
Tube-side exit
Baffles
Core element
Tube-side entry
Plate-side exit
Fig. 2. Bevax welded-plate heat exchanger (Reay 1999)

The printed circuit heat exchanger (PCHE) is a relatively new concept that has only been
commercially manufactured by Heatric™ since 1985. PCHEs are robust heat exchangers that
combine compactness, low pressure drop, high effectiveness, and the ability to operate with
a very large pressure differential between hot and cold sides (Heatric™ Homepage 2011).
These heat exchangers are especially well suited where compactness is important. The
Heatric™ heat exchanger falls within the category of compact heat exchangers because of its
high surface area density (2,500 m2/m3) (Hesselgreaves 2001). As the name implies, PCHEs
are manufactured by the same technique used for producing standard printed circuit boards
for electronic equipment. In the first step of the manufacturing process, the fluid passages
are photochemically etched into the metal plate (See Figure 3). Normally, only one side of
each plate is etched-out. The etched-out plates are thereafter joined by diffusion bonding,
which is the second step and results in extremely strong all-metal heat exchanger cores.
Plates for primary and secondary fluids are stacked alternately and formed into a module.
Modules may be used individually or joined with others to achieve the needed energy
transfer capacity between fluids. The diffusion bonding process allows grain growth,
thereby essentially eliminating the interface at the joints, which in turn gives the parental
metal strength. Because of the use of diffusion bonding, the expected lifetime of the heat
exchanger exceeds that of heat exchangers that are based on a brazed structure (Dewson and
Thonon 2003).
(a) Photo of PCHE (b) Cross-sectional view of PHCE

Fig. 3. Printed circuit heat exchanger (Heatric™ Homepage 2011)
The Marbond heat exchanger is a type of compact heat exchanger based on a novel
combination of photochemical etching and diffusion bonding (Phillips 1996). The internal
construction of this heat exchanger comprises a stack of plates that are etched
photochemically to form a series of slots as shown in Figure 4. The plates are stacked
with high positional tolerance such that series of slots form discrete flow paths. Adjacent
flow paths are separated by means of intervening solid plates. Thus, two or more separate
flow paths may be formed across a group of plates, enabling different fluid streams.
Injecting a secondary reactant into the flow of the primary reactant may be achieved by
means of perforations in the solid separator plate that are aligned exactly with the flow
paths of the primary reactant. The use of a positive pressure differential between the
secondary and primary reactant streams ensures that the secondary reactants flow in the
desired direction.
Fig. 4. Marbond heat exchanger (Phillips 1996)

1.1.3 Heat exchanger fluid types and comparisons

A variety of heat transfer fluids are available for high temperature heat exchangers
including gases, liquid metals, molten salts, etc. The following lists some general
characteristics required for the heat transfer fluid:
• High heat transfer performance to achieve high efficiency and economics
• Low pumping power to improve economics through less stringent pump requirements
• Low coolant volume for better economics
• Low structural materials volume for better economics
• Low heat loss for higher efficiency
• Low temperature drop for higher efficiency.
Characteristics of heat transfer fluids have been extensively investigated by Kim,
Sabharwall, and Anderson (2011) for high temperature applications based on the following
Figures-of-Merit (FOMs):
• FOMht represents the heat transfer performance of the coolant. It measures the heat
transfer rate per unit pumping power for a given geometry.
• FOMp represents the pumping power of the coolant. It measures the pumping power
required to transport the same energy for a given geometry.
• FOMcv represents the volume of the coolant. It measures the coolant volume required
for transferring heat with the same heat and pumping power.
• FOMccv represents the volume of the structural materials. It measures the volume of the
coolant structural materials required for transferring heat with the same heat duty and
pumping power under given operating conditions (T and P).
• FOMhl represents the heat loss of the coolant. It measures the heat loss of the coolant
when it is transported the same distance with the same heat duty and pumping power.
• FOMdt represents the temperature drop in the coolant while transferring thermal
energy with a given heat duty and pumping power.
Table 2 shows the comparisons of the thermal-hydraulic characteristics of the various
coolants based on the estimated FOMs (Kim, Sabharwall, and Anderson 2011). In this
estimation, the water at 25°C and 0.1 MPa was selected to be the reference coolant. The
following summarizes the results:
• Higher FOMht is preferred for better heat transfer performance. According to the
comparisons, sodium shows the highest value (=19.05) and argon has the lowest value
(0.05). Overall, FOMht is the highest in liquid metal followed by liquid water, molten
salt, and gases, respectively.
• Lower FOMp is preferred for better efficiency and economics. According to the
comparisons, liquid water has the lowest value (=1.0) and argon has the highest value
(=72592). Overall, FOMp is the lowest in molten salt followed by liquid metals and
gases, respectively.
• Lower FOMcv is preferred because it requires less coolant volume for providing the
same amount of heat transfer performance under the same pumping power. According
to the comparisons, the liquid water has the lowest value (=1.0) and argon has the
highest (=101.44). Overall, FOMcv is the lowest in molten salt followed by liquid metals
and gases, respectively.
• Lower FOMccv is preferred because it requires less structural material volume for both
heat transfer pipes and components. Overall, the same result was obtained as the
FOMcv. The FOMccv is the lowest in molten salt followed by liquid metals and gases,
respectively.
Coolant FOMht FOMp FOMcv FOMccv FOMhl FOMdl

Water
Ref. 1.00 1.00 1.00 1.00 1.00 1.00
(25°C, 1 atm)*
He 0.12 25407.41 67.74 4741.80 0.40 0.40
Gas Air 0.07 40096.15 80.10 5607.14 0.26 0.26
(700°C, CO2 0.11 11390.17 47.19 3303.46 0.32 0.32
7 MPa) H2O (Steam) 0.11 10012.63 45.10 3157.12 0.32 0.32
Ar 0.05 72592.09 101.44 7100.53 0.20 0.20
LiF-NaF-KF 0.80 2.87 1.57 1.57 0.92 0.92
NaF-ZrF4 0.45 5.02 1.98 1.98 0.56 0.56
KF-ZrF4 0.38 8.69 2.49 2.49 0.51 0.51
LiF-NaF-ZrF4 0.40 5.36 2.05 2.05 0.50 0.50
LiCl-KCl 0.55 14.99 3.07 3.07 0.76 0.76
Molten Salt
LiCl-RbCl 0.47 23.03 3.66 3.66 0.70 0.70
(700°C)
NaCl-MgCl2 0.58 16.26 3.18 3.18 0.81 0.81
KCl-MgCl2 0.50 14.30 3.02 3.02 0.70 0.70
NaF-NaBF4 0.71 5.66 2.04 2.04 0.88 0.88
KF-KBF4 0.64 8.98 2.47 2.47 0.84 0.84
RbF-RbF4 0.54 14.61 3.01 3.01 0.75 0.75
Sodium 19.05 33.62 4.19 4.19 28.91 28.91
Liquid
Lead 6.05 111.64 6.90 6.90 10.82 10.82
Metal
Bismuth 6.61 100.69 6.60 6.60 11.66 11.66
(700°C)
Lead-Bismuth 4.86 142.94 7.65 7.65 8.95 8.95
Table 2. Principal features of several types of heat exchangers (Shah and Sekulic 2003)
Fluid h [W/m2K]
Gases (natural convection) 3–25
Engine Oil (natural convection) 30–60
Flowing liquids (nonmetal) 100–10,000
Flowing liquid metal 5000–25,000
Boiling heat transfer:
Water, pressure < 5 bars, dT<25K 5000–10,000
Water, pressure 5-100, dT = 20K 4000–15,000
Film boiling 300–400
Condensing heat transfer:
Film condensation on horizontal tubes 9000–25,000
Film condensation on vertical surface 4000–11,000
Dropwise condensation 60,000–120,000
Table 3. Order of magnitude of heat transfer coefficient (Kakac and Liu 2002)
• Lower FOMhl is preferred because it requires less insulation for preventing heat loss.
According to the comparisons, argon has the lowest value (0.2), and sodium has the
highest (28.9). Overall, the FOMhl is the lowest in gases followed by molten salt and
liquid metal, respectively.
• Lower FOMdt is preferred because more thermal energy can be transferred long
distances without much of a temperature drop. Same values were obtained for the
FOMdt as obtained from FOMhl.
In the heat exchanger design, the heat transfer coefficient is a very important parameter
because it determines overall heat exchanger sizes and performance. Table 3 lists some
coolant types and the ranges of their heat transfer coefficients (Kakac and Liu 2002). As can
be seen, water exhibits the highest heat transfer coefficient in the drop-wise condensation,
and gases exhibit the lowest in the natural circulation.
1.1.4 Heat exchanger materials and comparisons

Material selection is one of the most important things in the high temperature application.
There are four main categories of high temperature materials: high temperature nickel-
based alloy, high temperature ferritic steels and advanced carbon silicon carbide composite,
and ceramics (Sunden 2005).
Ohadi and Buckley (2001) extensively reviewed materials for the high temperature
applications. High temperature nickel-based material has good potential for helium and
molten salts up to 750°C. High temperature ferrite steels shows good performance under
fusion and fission neutron irradiation to around 750°C. Advanced carbon and silicon
carbide composite has excellent mechanical strength at temperatures exceeding 1000°C. It is
currently used for high temperature rocket nozzles to eliminate the need for nozzle cooling
and for thermal protection of the space shuttle nose and wing leading edges. Many options
are available that trade fabrication flexibility and cost, neutron irradiation performance, and
coolant compatibility. Table 4 compares the properties of most commonly used high
temperature materials (Ohadi and Buckley 2001). It includes nickel-based alloy, ceramic
materials. and carbon and SiC composites. Figure 5 shows the specific strength versus
temperature for various composite materials.
Fig. 5. Specific strength vs. temperature (Brent 1989)

High temp. Metallic Ni alloys Ceramics oxides of Carbon-carbon Carbon fiber-SiC

material/fabrication (Inconel 718) Al, Si, Sr, Ti, Y, Be, composite composite
technology Zr, B and SiN, AiN,
B4C, BN, WC94/C06
Temperature range 1200 – 1250 ◦C 1500 – 2500 ◦C 3300 ◦C (inert 1400 – 1650 ◦C
environment) 1400 –
1650 ◦C (with SiC
layer)
Density 8.19 g/cm3 1.8 – 14.95 g/cm3 2.25 g/cm3 1.7 – 2.2 g/cm3
Hardness 250 – 410 (Brinell) 400 – 3000 kgf/mm2 0.5 – 1.0 (Mohs) 2400 – 3500 (V)
(V)
Elongation < 15% N/A N/A -
Tensile strength 800 – 1300 MPa 48 – 2000 MPa 33 (Bulk Mod.) 1400 – 4500 MPa
Tensile modules 50 GPa 140 – 600 GPa 4.8 GPa 140 – 720 GPa
Strength of HE Strength – adequate, Strength – not Strength – poor, Highest due to
but limited due to adequate, low oxidation starts at carbon fiber and SiC
creep and thermal mechanical 300 ◦C
exp parameters for
stress. Good thermal
and electrical
parameters
Electrical 125 µΩ cm 2E-06 – 1E+18 Ω cm 1275 µΩ cm 1275 µΩ cm
conductivity
Thermal 11.2 W/m K 0.05 – 300 W/m K 80 – 240 W/m K 1200 W/m K
conductivity
Thermal expansion 13E-06 K-1 0.54 – 10E-06 K-1 0.6 – 4.3E-06K-1 -
Comments Metallic expansion Often very expensive Life-time is low even Comparatively less
joints are the weak fabrication cost for protected by SiC expensive, successful
link conventional (adhesion is poor) proprietary
applications. fabrication
Technology technologies
proprietary for the available.
most part.
Technologically hard
to produce
Table 4. Selected properties of most commonly used high-temperature materials and

fabrication technologies (Ohadi and Buckley 2001)
Dewson and Li (2005) carried out a material selection study of very high temperature
reactor (VHTR) intermediate heat exchangers (IHXs). They selected and compared the
following eight candidate materials based on ASME VIII (Boiler and Pressure Vessel Code):
Alloy 617, Alloy 556, Alloy 800H, Alloy 880HT, Alloy 330, Alloy 230, Alloy HX, and 253MA.
UNS Tmax S898°C UTS 0.2%PS El

Alloys No (°C) (MPa) (MPa) (MPa) (%) Nominal compositions (wt%)
617 N06617 982 12.4 655 240 30 52Ni-22Cr-13Co-9Mo-1.2Al
21Ni-30Fe-22Cr-18Co-3Mo-3W-
556 R30556 898 11.0 690 310 40
0.3Al
800HT N08811 898 6.3 450 170 30 33Ni-42Fe-21Cr
800H N08810 898 5.9 450 170 30 33Ni-42Fe-21Cr
330 N08330 898 3.3 483 207 30 Fe-35Ni-19Cr-1.25Si
57Ni-22Cr-14W-2Mo-0.3Al-
230 N06230 898 10.3 760 310 40
0.05La
HX N06002 898 8.3 655 240 35 47Ni-22Cr-9Mo-18Fe
253MA S30815 898 4.9 600 310 40 Fe-21Cr-11Ni-0.2N
Table 5. Candidate materials for VHTR IHXs (Dewson and Li 2005)
Table 5 lists the allowable design stress (S) at 898°C, minimum required mechanical
properties (ultimate tensile stress [UTS]), 0.2% proof stress (PS), and elongation (El) at room
temperature, together with the nominal compositions of the alloys. They extensively
compared the mechanical properties, physical properties, and corrosion resistance for the
candidate materials, and finally concluded that Alloy 617 and 230 are the most suitable
materials for an IHX.
1.1.5 General heat exchanger design methodology

Once the process requirements are given, the high temperature heat exchanger is designed
using the same methods used for the typical heat exchangers. This subsection summarizes
the basic logical structure of the process heat exchanger design procedure described by Bell
(2004).
Figure 6 shows the logical process for heat exchanger design. The fundamental goal of any
such process is to find the optimal design among the infinite set of designs that could satisfy
the thermal-hydraulic and mechanical requirements. Because of large number of qualitative
factors in the design process, optimal needs are to be considered broadly. Generally, the
design process aims at the least costly (which usually means the smallest) heat exchangers
that meet the required thermal duty within the allowed pressure drops and satisfy the
mechanical requirements.
Problem Identification
Selection of a Basic
Heat Exchanger Type
Selection of Tentative
Set of Exchanger
Design Parameters
Modification of the
Design Parameters
Rating of the Design:
Thermal Performance
Pressure Drop
Evaluation of the Design:

Q, dP Acceptance? unacceptable
acceptable
Mechanical Design,
Costing, Etc
Fig. 6. Basic logical structure of the process heat exchanger design procedure (Bell 2004)
First, a basic heat exchanger type is selected based on the given operational requirements
and problems identified. The basic criteria for heat exchanger selection are as follows:
• It must satisfy the process specifications (performance) for temperature and pressure
• It must withstand the service conditions of the plant environment (reliability) for
temperature and pressure
• It must be maintainable for cleaning or replacement of a special component
• It should be cost effective (installed operating and maintenance costs)
• It must meet the site requirements or limitations for diameter, length, weight, and tube
configurations, and lifting and servicing capability or inventory considerations.
Second, main heat exchanger design parameters are selected based on the followings main
design factors:
• Heat duty
• Materials
• Coolant
• Pressure drop
• Pressure level
• Fouling
• Manufacturing techniques
• Cost
• Corrosion control
• Cleaning (with ease).
Third, rating and evaluation of the heat exchanger designs are iteratively conducted to
thermally optimize the design. The main purpose of this is to find all possible configurations
to meet the process requirements. Generally, the optimum heat exchanger is determined
based on the cost.
Once the optimal thermal design is provided, a mechanical design is developed based on
the following factors:
• Plate, tube, shell, and header thickness and arrangements
• Corrosion resistance
• Manifold
• Location of pressure and temperature measuring device
• Thermal stress analysis under steady and transient
• Flow vibrations
• Level of velocity to eliminate fouling and erosion
• Maintenance.
Detailed cost analyses are also conducted at this stage including both capital and
operational costs. The capital cost includes materials, manufacturing, testing, shipment, and
installation. The operational cost includes pumping power, repair, and cleaning.
1.1.6 Heat exchanger cost analyses

The cost of a heat exchanger is an important factor for heat exchanger design and selection.
Generally, manufacturers have their own methods for cost estimation. This section
introduces a simple heat exchanger costing methodology based on empirical cost data of the
ESDU (1994) for various feasible heat exchanger types (Shah and Sekulic 2003). Detailed cost
will depend on the operating conditions and materials used. The decision variable in this
method is cost of a heat exchanger per unit of its thermal size defined by
cos t
CUA = (1)
q
ΔTm
Where:
cost = cost of the heat exchanger
q = thermal duty
∆Tm = mean temperature.
Table 6 shows a selection of the cost data represented by the values of CUA (Shah and Sekulic
2003). This table is prepared for an application between gas as a hot fluid at a medium
pressure of about 2 MPa and cold fluid as treated water. An extensive set of CUA data for
various heat exchangers can be found in ESDU (1994).
CUA ($/(W/K)) Welded Plate

Shell-and-
q/∆Tm Double Tube Printed Circuit Plate-Fin
Tube U CUA
U = 484 U = 1621 U = 491
U = 484 (W/m2 K) ($/(W K))
(W/m2 K) (W/m2 K) (W/m2 K)
(W/m2 K)
103 3.98 2.5 12 - 349 4.9
5x103 1.00 0.75 2.4 3.1 1187 1.22
3x104 0.29 0.31 0.6 0.513 1068 0.42
105 0.17 0.31 0.42 0.210 1112 0.28
106 0.106 0.31 0.28 0.115 1173 0.22
* Original cost data in ESDU are approximated to the US dollar value in 2000.
Table 6. Cost data CUA vs. UA for various heat exchanger types (Shah and Sekulic (2003),
ESDU (1994))
2. Thermal energy transfer for process heat application

Recent technological developments in next generation nuclear reactors have created
renewed interest in nuclear process heat for industrial applications. The Next Generation
Nuclear Plant (NGNP) will most likely produce electricity and process heat for hydrogen
production. Process heat is not restricted to hydrogen production, but is also envisioned for
various other technologies such as the extraction of iron ore, coal gasification, and enhanced
oil recovery. To utilize process heat, a thermal device is needed to transfer the thermal
energy from NGNP to the hydrogen plant in the most efficient way possible. There are
several options to transferring multi-megawatt thermal power over such a distance. One
option is simply to produce only electricity, transfer it by wire to the hydrogen plant, and
then reconvert the electric energy to heat via Joule or induction heating. Electrical transport,
however, suffers energy losses of 60 to70% because of the thermal-to-electric conversion
inherent in the Brayton cycle. A second option is to transport thermal energy via a single-
phase forced convection loop where a fluid is mechanically pumped between heat
exchangers at the nuclear and hydrogen plants. High temperatures, however, present
unique challenges for materials and pumping. Single phase, low pressure helium is an
attractive option for NGNP, but is not suitable for a single-purpose facility dictated to
hydrogen production because low pressure helium requires higher pumping power and
makes the process very inefficient. A third option is two-phase heat transfer utilizing a high-
temperature thermosyphon. Heat transport occurs via evaporation and condensation, and
the heat transport fluid is recirculated by gravitational force. Thermosyphons have the
ability to transport heat at high rates over appreciable distances, virtually isothermally, and
without any requirement for external pumping devices.
Heat pipes and thermosyphons have the ability to transport very large quantities of heat
over relatively long distances with small temperature loses. The applications of heat pipes
and thermosyphons require heat sources for heating and heat sinks for cooling. The
development of the heat pipe and thermosyphon was originally directed towards space
applications. However, the recent emphasis on energy conservation has promoted the use
of heat pipes and thermosyphons as components in terrestrial heat recovery units and
solar energy systems. Thermosyphons have less thermal resistance, wider operating limits
(the integrity of the wick material might not hold in heat pipes at very high temperatures),
and lower fabrication costs than capillary heat pipes, which makes a thermosyphon a better
heat recovery thermal device. Perhaps the most important aspect of thermosyphon
technology is that it can easily be turned off when required, whereas a heat pipe cannot be
turned off. This safety feature makes the licensing of NGNP process heat transfer systems
comparatively easier. This section describes the thermosyphon system and the potential
benefits of using it in order to transfer process heat from the nuclear plant to the hydrogen
production plant.
2.1 Thermosyphon design

Considerable effort has been invested in thermosyphon and heat pipe development,
resulting in broad applications. One significant advantage of heat transfer by thermosyphon
is the characteristic of nearly isothermal phase change heat transport, which makes the
thermosyphon an ideal candidate for applications where the temperature gradient is limited
and high delivery temperatures are required, as in the case of thermochemical hydrogen
production (Sabharwall and Gunnerson 2009; Sabharwall 2009). The nature of isothermal
heat transport results in an extremely high thermal conductance (defined as the heat transfer
rate per unit temperature difference). A schematic diagram of a thermosyphon system is
shown in Figure 7.
The controllable thermosyphon, conceptually illustrated in Figure 7, is a wickless heat
pipe with a separate liquid return line, which is an intriguing option to traditional pumped
fluid heat transfer. Thermosyphons rely on convection to transport thermal energy inside
pipes and high-temperature heat exchangers for the evaporation and condensation end
processes. Ideally, no pumping power is required in contrast to single-phase gas or liquid
loops that require compressors or pumps, both of which are problematic at very high
temperatures.
Heat is transported by saturated or superheated vapor expanded from an evaporative heat
exchanger, through a long pipe, to a condensation heat exchanger. Liquid condensate
returns to the evaporator assisted by gravity through a separate liquid return line with a
liquid return control valve. When the thermosyphon is started by applying power (process
heat from NGNP) to the evaporator, the working fluid is evaporated and the latent heat of
vaporization is transported (~isothermally) along the thermosyphon to the condenser
region. Expansion joints are added near the condenser section and at the inlet to the
condensate return line in order to accommodate the thermal expansion of the thermosyphon
piping at higher temperature. The condensate returns to the evaporator region through a
liquid bypass line containing a liquid storage reservoir and a control check valve as shown
in Figure 7. The storage reservoir and part of the liquid lines may incorporate electric
resistance heating if necessary in order to melt the working fluid and restart the
thermosyphon after a long shutdown period. Liquid from the storage reservoir passes into
the thermosyphon system evaporator through a control valve which, as needed, plays a role
in controlling the rate of heat transfer and shutting off or isolating the thermosyphon. In
order for the thermosyphon system to be shut and completely disabled from heat transfer,
the control valve is closed wherein all the working fluid is collected in the liquid storage
reservoir and the condensing-evaporating cycle is terminated. When it is desired to resume
the thermosyphon action, the control valve is opened to again allow the liquid to flow into
the evaporator region of the system. The heat input governs the rate of evaporation and the
subsequent rate of heat transfer. The rate of thermal energy exchange can be regulated over
a spectrum of conditions from “off” to “fully on,” hence the term controllable thermosyphon.
Fig. 7. Schematic of a simple controllable thermosyphon

Depending on the temperature and pressure of operation, favorable working fluids can be
identified. Alkaline metals, for example, may be suited for process heat transfer because
they have the characteristics of:
• High boiling temperature
• Availability and cost effectiveness
• Good heat transfer properties (latent and specific heat are both high)
• Typically good chemical compatibility (except Li).
Working fluids more suitable than alkali metals may exist, such as molten salts (Sohal et al.
2010). Corrosive behavior at high temperatures or lack of high temperature thermodynamic
properties, especially for superheated vapors, rule out fundamental analysis of many
possible thermosyphon working fluids.
2.2 Thermosyphon startup

The charging of a thermosyphon requires a transfer station wherein molten working fluid
under an inert environment or vacuum is transferred to the evacuated thermosyphon
pipe(s). Pure fluids without condensable gases are required for proper thermosyphon
operation. Otherwise, the impurities, which are more volatile than the fluid itself, will be
driven to the condenser section of the thermosyphon and less volatile impurities will be
collected in the evaporator causing hot spots and reducing heat transfer. Noncondensable
gases will accumulate within the condenser. Although conceptually simple, the startup of a
large, high temperature thermosyphon is difficult to accurately predict.
Figure 8 describes the procedure commonly practiced for filling up the thermosyphon. The
liquid sodium valve is opened till 20%-by-volume limit is reached for filling up the
thermosyphon. The 20%-by-volume is a good approximation for the coolant as described by
Gunnerson and Sanderlin (1994).
For a thermosyphon to startup efficiently and effectively, the working fluid has to initially
be in molten state. The liquid reservoir, as shown in Figure 8, has provisions for external
heating. During normal startup for both heat pipes and thermosyphons, the temperature of
the evaporator section increases by a few degrees until the thermal front reaches the end of
the condenser as described by Reay and Kew (2006). At this point, the condenser
temperature will increase until the pipe structure becomes almost isothermal.
2.3 Comparison of thermosyphon with convective loop

Alkaline metal thermosyphons and alkaline metal forced convective loops can both deliver
comparable rates of heat transfer through a given size pipe. This can be demonstrated by
considering the ideal rate of convective heat transport through a pipe without losses,
modeled in terms of enthalpy as:

m
Q′′ = ∆h ρ V = Δh (2)
A
where:
A -Cross-sectional flow area
∆h -Specific enthalpy change of the transport fluid
m -Mass flow rate
Q′′ -Rate of thermal energy transport per unit flow area

V -Average flow velocity
ρ -Density of the fluid
Δ -Difference.
LIQUID SODIUM
RESERVOIR
VALVE
THERMOSYPHON*
VALVE
VACUUM
PUMP
* 1/5 OR (20 %) BY VOLUME SODIUM BASED ON PRIOR STUDIES
Fig. 8. Na filling procedure for a thermosyphon

Two-phase heat transfer by a thermosyphon has the advantage of high enthalpy transport
when compared with single-phase forced convection. Vapor-phase velocities within a
thermosyphon can also be much greater than single-phase liquid velocities within a forced
convective loop.
Figure 9 exemplifies the enthalpy enhancement in heat transfer afforded by a two-phase
thermosyphon versus a single-phase convective loop with sodium as the working fluid as
shown by Sabharwall et al. 2009. The specific enthalpy (∆h) of saturated liquid and vapor,
relative to the solid at 298.15 K, is illustrated as a function of temperature. Assuming heat
transfer from a high temperature gas-cooled reactor to an industrial facility at 1223 K, the
maximum single-phase heat transfer is given by the enthalpy gain from points A to B in
Figure 9, or approximately 1,190 kJ for each kilogram of sodium. Compared with two-phase
heat transfer from points A to B to C, where the enthalpy gain is approximately 3,864 kJ per
kilogram with no vapor superheat, over three times more heat per kilogram of sodium is
needed than with the single-phase. The saturation pressure of sodium at 1223 K is only
0.188 MPa, thus minimizing pressure and stress forces. Vapor flow through a pipe is limited
by compressible choke flow when the vapor reaches its sonic velocity. The sonic velocity for
sodium vapor is approximately 737 m/s at 1223 K as given by Bystrov et al. (1990).
The limiting heat transfer rate for an ideal sodium thermosyphon operating at 1223 K can
therefore be estimated as
Q′′ = h fg ρ V VS = (3,864) kJ/kg (0.47)kg/m3(737)m/s = 1338 MW/m2
where:
hfg -Latent heat & vaporization
Similarly, single-phase liquid sodium could transport the same rate of thermal energy with
an average flow velocity of about 2.2 m/s, well within the capabilities of advanced liquid
metal pumps. This simple analysis (Sabharwall 2009) for sodium as the working fluid
theoretically illustrates that both a thermosyphon and a forced convective loop can deliver
comparable rates of heat transfer through comparable diameter pipes. The thermosyphon,
however, has the luxury of controllable heat transfer without the need for high temperature
pumping and can deliver the heat at the same approximate temperature as the source. The
enthalpy gain for sodium that can be achieved by two-phase heat transfer versus a single-
phase is about 3.7 times greater (Sabharwall 2009).
C
Vapor
Average
B Liquid
Temperature (K)
Fig. 9. Enthalpy for saturated sodium: liquid and vapor (thermodynamic data from Fink and
Leibowitz [1995]; Gunnerson, Sabharwall, and Sherman [2007])
2.4 Heat transport limitations

Depending on the operational conditions, the heat transport may be limited by one of the
following (Sabharwall et al. 2009), as described below:
2.4.1 Sonic limit (choking) of vapor flow

After continuum flow is established, the evaporator-condenser pressure difference accelerates
the vapor until it reaches a maximum velocity at the evaporator exit. The maximum vapor
velocity that can exist at the evaporator exit corresponds to sonic velocity, or Mach 1. This
choked flow condition is a fundamental limit on the axial vapor flow in a thermosyphon.
2.4.2 Viscous limit

At startup for liquid metals, the vapor pressure difference between the evaporator and the
condenser is zero or very small. In such cases, the viscous forces may be larger than the
vapor pressure gradients, thus preventing vapor flow.
2.5 Flow instabilities

Instabilities are common to both forced and natural circulation systems; the latter is more
unstable than forced circulation systems. Instabilities can cause operational problems in
process heat transfer. Thus, it is important to classify mechanisms that can lead to unstable
operational behavior of the thermosyphon. The following processes can lead to an unstable
behavior for the thermosyphon.
2.5.1 Surging (chugging) and geysering instability

Surging and geysering occur mainly because of liquid superheat. Surging occurs when
boiling is initiated in the evaporator, but because of nonuniformity in the temperature at the
wall and bulk fluid temperature, the vapor being generated becomes trapped, eventually
resulting in vapor expulsion as described by Bergles et al. (1981). Geysering is a similar
phenomenon that occurs when the heat flux is sufficiently high and boiling is initiated at the
bottom. In low pressure systems, this results in a sudden increase in vapor generation
because of the reduction in hydrostatic head, usually causing an expulsion of vapor.
2.5.2 Thermosyphon evaporator instability

If the evaporator section of the thermosyphon system is not sufficiently long for vapor
superheat, instability can occur such that the fluid at the outlet of the evaporator experiences
a static pressure decrease, leading to the onset of fluid condensation within the thermosyphon.
Slight vapor superheat from the evaporator should reduce this concern.
2.5.3 Fluid superheating (alkaline metals)

Alkaline metals have relatively high boiling temperatures at atmospheric pressure. If the
heater surface does not have enough active nucleation sites, boiling may not occur near
saturation temperature but rather require significant superheat. At high superheat
temperatures a vapor burst expulsion can be expected upon phase change, which could lead
to structural damage and flow excursions.
3. Summary
Thermal device is a component whose main objective is to transport thermal energy
across a system. In this chapter, two key thermal devices for high temperature heat
transfer applications were introduced and discussed in detail; (1) heat exchanger and
(2) thermosyphon. A heat exchanger is a key component in the thermal systems used for
transferring heat from one medium to another. Especially, the high temperature heat
exchanger technology is emerging in many industrial applications such as gas turbines,
chemical plants, and nuclear power plants with increasing demands. Unlike typical heat
exchangers, high temperature heat exchangers require and exhibit some special
characteristics to be operated in severe environments.
In this chapter, shell-and-tube, Bevax, PCHE, and Marbond heat exchangers are recommended
to be the possible option for high temperature applications.
Working fluids and structural materials are the important design parameters that should be
carefully considered for high temperature heat exchangers. A variety of heat transfer fluids
are available for high temperature heat exchangers including gases, liquid metals, molten
salts, etc. General characteristics required for heat transfer coolant are (1) high heat transfer
performance, (2) low pumping power, (3) low coolant volume, (4) low heat loss, and (5) low
temperature drop. In this chapter, six figures-of-merit (FOMs) for heat transfer coolants
were introduced for evaluating important coolant characteristics. Based on these FOMs,
various candidate coolants were compared extensively. Material selection is one of the most
important things in the high temperature application. There are four main categories of high
temperature materials, high temperature nickel-based alloy, high temperature ferritic steels
and advanced carbon silicon carbide composite, and ceramics. Basic design methods and
cost analysis for the heat exchangers are also briefly mentioned. Another thermal device are
heat pipes and thermosyphons, which have the ability to transport very large quantities of
heat over relatively long distances with small temperature loses. In this chapter, the
thermosyphon system and the potential benefits of using it were described in the heat
transfer perspective from the nuclear plant to the hydrogen production plant (also
applicable to any other process application industry). One significant advantage of heat
transfer by thermosyphon is the characteristic of nearly isothermal heat transport, which
makes the thermosyphon an ideal candidate for applications where the temperature
gradient is limited and high delivery temperatures are required, as in the case of
thermochemical hydrogen production.
Depending on the temperature and pressure of operation, favorable working fluids can be
identified. Alkaline metals, for example, may be suited for process heat transfer because of
high boiling temperature, availability and cost effectiveness, good heat transfer properties
(latent and specific heat are both high), and typically good chemical compatibility (except
for Li).
4. References
Bell, K.J., (2004). “Heat Exchanger Design for the Process Industries,” Journal of Heat Transfer,
Vol. 126, pp. 877-885, 2004
Bergles, A.E., Collier, J.G., Delhaye, J.M., Hewitt, G.F. and Mayinger, F., (1981). Two-Phase
Flow and Heat Transfer in the Power and Process Industries, Hemisphere Publishing
Corporation.
Brent, A., (1989). Fundamentals of Composites Manufacturing Materials Methods and
Applications, Society of Manufacturing Engineers, Dearborn, MI.
Bystrov, P.V., Kagan, D.N., Krechetova, G.A. and Shpilrain, E.E., (1990). Liquid- Metal
Coolants for Heat Pipes and Power Plants, Hemisphere Publishing Corporation.
Dewson, S., Li, X., 2005, “Selection Criteria for the High Temperature Reactor Intermediate
Heat Exchanger,” Proceedings of ICAPP'05, Seoul, Korea, May 15-19, 2005
Dewson, S.J., and Thonon, B., (2003). “The Development of High Efficiency Heat Exchangers
for Helium Gas Cooled Reactors,” Report No. 3213, International Conference on
Advanced Nuclear Power Plants (ICAP), Cordoba, Spain.
ESDU, (1994). “Selection and costing of heat exchangers,” Engineering Science Data, Item
92013, ESDU, Int. London, UK.
Fink, J.K. and Leibowitz, L., (1995). Thermodynamic and Transport Properties of Sodium
Liquid and Vapor, ANL/RE-95/2
Fisher, D.L., and Sindelar, R.L., (2008). Compact Heat Exchanger Manufacturing Technology
Evaluation, Savannah River Nuclear Solutions, SRNS-STI-2008-00014, Savannah
River Site, Aiken, South Carolina.
Gunnerson, F.S., Sabharwall, P., and Sherman, S., (2007). “Comparison of Sodium
Thermosyphon with Convective Loop,” Proceedings of the 2007 AIChE Conference,
Salt Lake City, November 2007.
Gunnerson, F.S., and Sanderlin, F.D., (1994). “A Controllable, Wickless Heat Pipe Design for
Heating and Cooling,” Fundamentals of Heat Pipes, ASME HTD, Vol. 278.
Heatric™ Homepage, (Jan 20, 2011). URL: www.heatric.com.
Hesselgreaves, J.E., (2001). Compact Heat Exchangers, Selection, Design and Operation, First
edition, Pergamon.
Kakaç, S. and Liu, H. (2002). Heat Exchangers: Selection, Rating and Thermal Design (2nd ed.).
CRC Press. ISBN 0849309026.
Kim, E.S., Sabharwall, P., Anderson, N., (2011). “Development of Figure of Merits (FOMs)
for Intermediate Coolant Characterization and Selection,” ANS Annual Meeting,
Hollywood Florida, USA, June 25–30, 2011.
Ohadi, M.M. and Buckey, S.G., (2001). “High temperature heat exchangers and microscale
combustion systems: applications to thermal system miniaturization,” Experimental
Thermal and Fluid Science, Vol. 25, pp. 207–217.
Phillips, C., (1996). Intensification of Chemical Batch Processes using Integrated Chemical Reactor -
Heat Exchangers, Final Report for JOULE II Contract No. JOU2-CT94-0425 (DG 12
WSME), June 1996.
Reay, D. and Kew, P., (2006). Heat Pipes: Theory Design and Applications, Fifth edition,
Butterworth-Heinemann, Elsevier.
Reay, D.A., (1999). “Learning from Experiences with Compact Heat Exchangers,” CADDET
Analyses Series No. 25, Centre for the Analysis and Dissemination of Demonstrated
Energy Technologies, Sittard, The Nederland.
Sabharwall, P., (2009). “Nuclear Process Heat Transfer for Hydrogen Production,” VDM
Verlag Publications, August 2009.
Sabharwall, P., and Gunnerson, F., (2009). “Engineering Design Elements of a Two-Phase
Thermosyphon for the Purpose of Transferring NGNP Thermal Energy to a
Hydrogen Plant,” Journal of Nuclear Engineering and Design, Vol. 239, June 2009
Shah, R. K., and Sekulic S. P., (2003). “Fundamentals of Heat Exchanger Design,” John Wiley
and Sons.
Sherman, S. R., and Chen Y., (2008). Heat Exchanger Testing Requirements and Facility
Needs for the NHI/NGNP Project, WSRC-STI-2008-00152, April 2008.
Sohal, M., Ebner, M., Sabharwall, P., and Sharpe, P., (2010). Engineering Database of Liquid
Salt Thermo-Physical and Thermo-Chemical Properties, INL/EXT-10-18297, Idaho
National Laboratory, Idaho, March 2010.
Sunden, B., (2005). “High Temperature Heat Exchangers (HTHE),” Proceedings of Fifth
International conference on Enhanced, Compact and Ultra-Compact Heat
Exchangers: Science, Engineering and Technology, Hoboken, NJ, USA, September
2005.
18
Flow Properties and Heat Transfer of

Drag-Reducing Surfactant Solutions
Takashi Saeki
Yamaguchi University
Japan
1. Introduction
The frictional resistance of fluids can be reduced by adding small amounts of certain
polymers, a phenomenon first reported (Toms, 1948) known as drag reduction or Toms
phenomenon. The added polymers might form thread-like structures in fluids, which
interact with turbulent eddies due to their viscoelasticity. Since the polymer synthesis
technology and cost effectiveness have been highly improved, polymer drag reduction has
been adopted widely in large pipeline systems for crude oils and refined petroleum
products. Presently, more than 40% of all the gasoline consumed in the United States has
polymer drag reducer in it (Motier, 2002). However, mechanical degradation of polymer
chains in high shear rate regions, such as pumps, is frequently observed, which lowers the
molecular weight and causes a loss of drag reduction. For that reason, polymer drag
reduction cannot be adopted for circulating flow systems.
Drag reduction caused by surfactant solutions was first reported by Gadd (Gadd, 1966).
Combinations of certain cationic surfactants with a suitable counter ion are often chosen as
the drag-reducing agents. Some nonionic surfactants also show the drag-reducing effects,
rendering the use of counter ions dispensable. A number of authors have pointed out that
the surfactant molecules come together to form rod-like micelles, which are necessary for
drag reduction. Figure 1 shows surfactant molecule and micelle structures. Drag-reducing
surfactants form rod-like micelles, and their aggregates might be present in a solution.
Figure 2 shows a transmission electron microscope (TEM) image of surfactant micelles
(Shikata et al., 1988). Again, aggregates of rod-like micelles might interact with turbulent
eddies and cause drag reduction. These aggregates suffer mechanical degradation in high
shear rate regions, which is then repaired in lower shear stress regions, such as in flow
through pipes.
hydrophobic
hydrophilic
Surfactant molecule Spherical micell Rod-like micell
Fig. 1. Surfactant molecule and micelle structures

Fig. 2. Rod-like structure of surfactant micelles. (CTAB/NaSal. CD=0.001 mol/L)

The drag reduction caused by surfactant solutions is considered to be an effective way to
reduce the pumping power in closed-loop district heating and cooling systems. In 1994, a
commercial application of surfactant drag reduction was first conducted in Japan by our
group for the air conditioning system in the Shunan Regional Industry Promotion Center
building (a two-story building with a total floor space of 2490 m2). We also developed
commercially available drag-reducing additives (LSP-01A/M, LSP Corporative Union)
based on the mixture of a cationic surfactant and corrosion inhibitors. Since 1995, LSP-01 has
been used practically at more than 150 sites in building air conditioning systems throughout
Japan, including office buildings, hotels, hospitals, supermarkets, airport facilities, and
industrial factories. Quantitative evaluations of the energy conservation rate were
conducted for each application project. Almost all of our drag-reducing projects showed
more than a 20% reduction in the pumping power required for circulating water;
furthermore, some air conditioning systems have obtained up to 50% energy savings with
LSP-01 (Saeki et al., 2002). Table 1 shows some examples of our projects.
In the following section, several drag-reducing surfactants and their flow properties are
summarized. Some studies have shown that heat transfer reduction occurs simultaneously
for drag-reducing flows. Therefore in the next section, heat transfer characteristics obtained
for our laboratory experiments are displayed for two commercially available drag-reducing
surfactants, that are used frequently in practical facilities. In the last section, drag reduction
and heat transfer data measured for our university library building are presented and
evaluated with regard to the heat transfer characteristics of a practical air conditioning
system before and after introducing surfactant drag reduction. We believe this information
will be useful for future designs incorporating this technology.
2. Several drag-reducing surfactants and their flow properties

Many studies of surfactant drag reduction have been conducted since the 1980s, including
investigations of the selection and optimization of additives, the drag reduction flow
properties, the mechanism of drag reduction, and so on. Cetyl-trimethyl-ammonium
chloride (CTAC, C16H33N+ (CH3)3Cl-) with sodium salicylate (HOC6H4COONa, called NaSal)
displays significant drag reduction qualities, and many researchers have used these
additives. However, it was found that the solution lost its solubility in water at temperatures
lower than 7 °C, and some ions dissolved in tap water affected the drag reduction caused by
CTAC. Therefore, screening tests of surfactants had been conducted (Ohlendorf et al., 1986;
Chou et al., 1989; Usui et al., 1998). At the present, oreyl-bishydroxyethyl-methyl-ammonium
Flow Properties and Heat Transfer of Drag-Reducing Surfactant Solutions 333
chloride (C18H35N+(C2H4OH)2CH3Cl- brand name Ethoquad O/12) and

oreyl-trishydroxyethylethyl-ammonium chloride (C18H35N+(C2H4OH)3Cl- brand name
Ethoquad O/13) are used as suitable drag-reducing surfactants in Japan. Both surfactants
have to be used in combination with a counter ion, NaSal. Stearyl-trimethyl-ammonium
chloride (STAC, C18H33N+(CH3)3Cl-) was also selected as a drag-reducing surfactant for
higher-temperature use, and a surfactant having an alkyl group carbon number of 22 shows
effective drag reduction even over 100 °C (Chou et al., 1989). The molecular weights of these
surfactants are several hundreds.
Operating Water Energy

time capacity saving
No. Facility Heat origin Pump
(hours/year) (m3) rate (%)
Home for
1 5.5kW×1,
the aged ACWGM(90RT) ×1 8,760 3 39
1.5kW×2
ACWGM(360RT) ×2
2 Factory 1
Turbo(400RT) ×1 45kW×3 4,380 25 21
3 Factory 2 ACWGM(160RT) ×2 22kW×2 2,100 10 27
Department ACWGM(450RT) ×2
4
store 1 Turbo(400RT) ×1 22kW×3 3,850 30 33
Department ACWGM(400RT) ×1
5
store 2 Turbo(315RT) ×1 37kW×1 4,420 7 29
55kW×1,
Department ACWGM(600RT) ×1,
6 22kW×2,
store 3 (150RT) ×1 2,640 15 54
18.5kW×2
22kW×2,
18.5kW×2,
7 Hotel 1 ACWGM(300RT) ×3
11kW×2, 8,760 50 41
7.5kW×2
8 Hotel 2 ACWGM(125RT) ×1 15kW×1 8,000 10 48
11kW×1,
9 Hotel 3 ACWGM(100RT) ×3 2,000 15 48
7.5kW×1
ACWGM: Absorption Cooling Water Generating Machine Turbo: Turbo refrigerator
Table 1. Drag-reducing project with LSP-01

Figure 3 presents the friction factor versus Reynolds number data for water with different
concentrations of LSP-01. LSP-01 contains 10% Ethoquad O/12, so in the case of 5000 mg/L
LSP-01, for example, the solution contains 500 mg/L Ethoquad O/12, 300 mg/L NaSal, and
corrosion inhibitors. Since drag-reducing solutions are non-Newtonian fluids, the viscosity
used for the calculation of Reynolds number is used as the property of water. The drag
reduction increased with the Reynolds number and reached the maximum point (maximum
drag reduction at a certain temperature). The maximum drag reduction percent (DR%) is
almost 70% for 5000 mg/L LSP-01. When the Reynolds number was further increased, the
drag reduction was lost rather abruptly. This phenomenon may be caused by the breakup of
the surfactant micelle structures due to high shear.
10-2 Blasius equation

f [-]
drag reducing
additive [mg/l]
2000 6000
3000 8000
4000 10000
5000 25℃
-3
10 10000 100000
Re [-]
Fig. 3. Drag reduction results for LSP-01 at different temperatures
Figure 4 shows photos of water and a drag-reducing solution stirred in a beaker. Water
shows a big eddy, while the drag-reducing solution shows no eddies due to the Weissenberg
effect.
Water LSP-01(5000mg/L)
Fig. 4. Photos of water and a drag-reducing solution stirred in a beaker

Saeki et al. (2000) measured the flow characteristics of a drag-reducing surfactant by using
particle tracer velocimetry (PTV) to study the mechanism of drag reduction. In the
experimental results, the intensity of the fluctuation in the axial direction was high, while
that in the radial direction was low but not zero. Despite the fluctuations occurring in the
drag-reducing surfactant solution flow, the Reynolds stress was shown to be zero. Hence the
drag-reducing flow is considered to be neither “turbulent” nor “laminar” and is assumed to
be a kind of “transitional flow.”
3. Heat transfer characteristics of drag-reducing flow

3.1 Review of related works
Not a few researchers have pointed out that heat transfer reduction occurs simultaneously
for drag-reducing flows. Usui and Saeki (1993) measured the heat transfer characteristic of
CTAC solution and reported that the analogy between momentum and heat transfer was
invalid for drag reduction flow and that the heat transfer reduction was as high as to be
even larger than the drag reduction rate. Aguilar et al. (2001) introduced both a maximum
heat transfer reduction asymptote (MHTRA) and a maximum drag reduction asymptote
(MDRA) for surfactant solutions. They indicated that the ratio of MHTRA and MDRA could
be expressed with a constant value of 1.06, independent of the Reynolds number. Steiff et al.
(1998) noted that the influence of drag-reducing additives on heat exchangers had to be
given particular attention and recommended several ways to improve the heat output
behavior of heat exchangers. Qi et al. (2001) reported the enhanced heat transfer of drag-
reducing surfactant solutions with a fluted tube-in-tube heat exchanger. In their
experiments, there was a surprising increase in the heat transfer reduction with Ethoquad
T13-50 solution at a temperature of 60 °C, even though the pressure drop in the solution at
this temperature is close to that of water in the fluted tube. They pointed out that this
temperature is near the upper temperature limit for drag reduction of the solution and
noted the possibility that the heat transfer reduction decreases slowly with the Reynolds
number.
In a normal air conditioning system, the temperature of circulating water is around 60 °C,
and LSP-01 shows a sufficient and stable drag-reducing rate at this temperature range.
However, it is expected that the temperature inside a heat exchanger will be higher, and it
should coincide with the upper limit temperature of drag reduction with LSP-01. In this
study, two drag-reducing additives were used, and both drag reduction and heat transfer
reduction were measured from room temperature to that beyond the upper limit
temperatures of drag-reducing solutions.
3.2 Experimental procedure and data reduction

The cationic surfactants used were Ethoquad O/12 (oreyl-bishydroxyethyl-methyl -
ammonium chloride) and Ethoquad O/13 (oreyl-trishydroxyethylethyl-ammonium
chloride) produced by Lion Akzo Corporation, Japan. The counter ion selected was sodium
salicylate (NaSal). In the case of Ethoquad O/12, the 1.23 mM (500 mg/l) surfactant was
mixed with 2.08 mM (300 mg/l) NaSal and water, a mixing rate that shows good drag
reduction. Since a similar mixing rate is suitable for Ethoquad O/13, the experiments were
conducted with the weight ratio of both surfactants and NaSal kept at 1.67. The
experimental apparatus used for drag reduction and heat transfer experiments was a
recirculation system, as shown in Figure 5. Drag-reducing agents were added to a tank, and
solutions were prepared. The temperature of the solutions was controlled by using a heater
with a temperature control device to allow experiments to run from room temperature to
80 °C. The flow rate was controlled by an inverter system installed on then system’s main
pump.
The pressure drop test section was a straight PVC pipe, 1600 mm long, 14.5 mm inside
diameter. Measurements were carried out using an electric differential manometer. The heat
transfer test section consisted of a copper tube, an electric heater, and heat insulators. The
copper tube was a practical heat exchanger tube, 1500 mm long, 14.5 mm inside diameter,
17.1 mm outside diameter. A cross-sectional view of the test section of the heat transfer
experiment is shown in Figure 5. An electric heater of constantan wire insulated with glass
fiber and calcium silicate insulator was used to obtain a constant heat flux condition during
the test section. The heated surface temperature was measured at a point 1500 mm from the
inlet by a K-type thermocouple buried in the copper pipe (depth =0.5 mm). The inlet and
outlet temperatures of the test section were also measured. The outlet temperature was
measured at a mixing box installed just after the test section. Three K-type thermocouples
were fixed at different positions inside the mixing box, and the average temperature was
calculated. All the data were obtained over 5 minutes and averaged. The difference
temperature between the inlet and outlet of the heat transfer test section was set to more
than 2 °C but less than 5 °C throughout the experiments.
heater flow meter electric differential manometer
tank
heat transfer test section mixing
1500 mm box
tc-1 tc-5 tc-2,3,4

tc-5
pump inverter calcium silicate
insulator
constantan wire
ceramic tape
polyimide film
copper pipe
Fig. 5. Experimental apparatus

During drag reduction tests, the temperature of the test solution was kept constant
throughout the test loop. The Reynolds number was calculated with the property of water.
For convenient comparison of the drag reduction results between water and test solutions,
the drag reduction rate, DR%, was defined as follows:
f w − f DR
DR% = × 100 (1)
fw
in which the friction factor of water, fw , was obtained by using the Blasius equation
(fw =0.0791Re-0.25), while that of surfactant solution, fDR , was measured. For the heat transfer
tests, both Reynolds and Prandtl numbers were calculated with the property of the solvent
at the average temperature of the inlet and outlet of the test section. The amount of heat
flow was calculated as follows:
Q = GC p (Tout − Tin ) (2)
where G, Cp, Tout, and Tin are the flow rate, specific heat, outlet temperature, and inlet
temperature of the test section, respectively. By using measured outer surface temperature,
the inner surface temperature could be calculated as follows:
Q ( r1 − r2 )
T2 = T1 − (3)
kAim
where r, k, and Alm are the tube radius, thermal conductivity, and logarithmic mean of heat
surface area, respectively. The subscripts of 1 shows outer surface, while 2 indicates inner
surface. By assuming that the temperature through the test section increased linearly from
the inlet to the outlet temperature, we were able to calculate a fluid temperature at the
position where the thermocouple was buried and the difference between the temperature
and T2 was defined as ΔT. Then, the heat transfer coefficient, h, was calculated as follows:
Q
h= (4)
ΔTA2
Finally, the Nusselt number (= hD/k) was obtained. Here, for a convenient comparison of the
heat transfer results between water and the test solutions, we defined the heat transfer
reduction rate, HTR%, as follows:
hw − hDR
HTR% = × 100 (5)
hw
3.3 Results and discussion

Figure 6 shows the friction factors, f, and the Reynolds numbers, Re, for water and Ethoquad
O/12 with NaSal solutions at different concentrations measured at 30 °C. All points for the
water test lay close to the Blasius equation, with a deviation of less than 5%. For surfactant
solutions, the drag reduction increased with the Reynolds number and reached a maximum
point. When the Reynolds number was further increased, the drag reduction was lost rather
abruptly. This phenomenon might be related to the breakup of the surfactant micelle
structures due to high shear. With increasing surfactant concentration, the maximum drag
reduction was shifted to a higher Reynolds number.
Figure 7 gives the corresponding heat transfer test results for the solutions. The figure
presents the relation between the Nu/Pr0.4 values and the Reynolds numbers. Again, all
points for the water test lay close to the Bittus-Boelter empirical equation for a Newtonian
fluid with a deviation of less than 5%. For surfactant solutions, the heat transfer reduction
increased with the Reynolds number. The 100 mg/l Ethoquad O/12 solution showed the
maximum value and reached the Newtonian line at the higher Reynolds number region;
however, the 300 mg/l and 500 mg/l solutions maintained a high level of heat transfer
reduction. There was a significant decrease of drag reduction with an increase in the
Reynolds number; however, the Nu/Pr0.4 kept almost constant values with the 300mg/l and
500mg/l solutions. Qi et al. (2001) reported similar results with Ethoquad T13-50 solution.
Ethoquad O/12
-0.25
f = 0.0791 Re 0 mg/L
100 mg/L
300 mg/L
10-2 500 mg/L
f [-]
30℃
-3
10
104 105
Re [-]
Fig. 6. Drag reduction results for Ethoquad O/12 and NaSal systems at different
concentrations at 30 °C
103
0.8 0.4
Ethoquad O/12 Nu = 0.02 Re Pr
0 mg/L
[-]
100 mg/L
300 mg/L
0.4
102 500 mg/L

Nu/Pr
101 30℃
104 105
Re [-]
Fig. 7. Heat transfer results for Ethoquad O/12 and NaSal systems at different
Figure 8 shows f vs. Re for Ethoquad O/12 solutions at 50 °C. Drag reduction endured to a
higher Reynolds number compared with the results at 30 °C. The 100 mg/L solution
showed significant drag reduction compared with the 300 mg/L and 500 mg/L solutions;
however, the drag reduction was lost rather abruptly when the Reynolds number exceeded
35,000.
Ethoquad O/12
-0.25
f = 0.0791 Re 100 mg/L
300 mg/L
500 mg/L
10-2
f [-]
50℃
-3
10
104 105
Re [-]
Fig. 8. Drag reduction results for Ethoquad O/12 and NaSal systems at different
Figure 9 shows the corresponding Nu/Pr0.4 vs Re at 50 °C. The reduction rate of Nu/Pr0.4 is
lower than that for the 300 mg/L and 500 mg/L solutions. Experimental results of drag
reduction and heat transfer characteristics varied with the concentration of Ethoquad O/12,
Reynolds number (flow velocity), and temperature of the solutions. As these results show
the analogy between momentum and heat transfer is invalid for drag-reducing flows.
103
0.8 0.4
Ethoquad O/12 Nu = 0.02 Re Pr
0 mg/L
[-]
100 mg/L
300 mg/L
0.4
102 500 mg/L

Nu/Pr
101 50℃
104 105
Re [-]
Fig. 9. Heat transfer results for Ethoquad O/12 and NaSal systems at different concentrations
at 50 °C
Figure 10 shows the relationship between DR% and solution temperature. Experiments
were conducted to increase the temperature of 500 mg/l Ethoquad O/12 solution gradually
from room temperature to 75 °C. The velocity at the test sections was kept constant at 1.5
m/s, corresponding to the typical velocity inside a practical heat transfer pipe. It is clear that
with increasing solution temperature, drag reduction increased and a maximum DR% of
75% was observed from 60 °C to 65 °C. Over this temperature range, drag reduction was
suddenly lost. In this paper, the upper limit temperature is defined as the temperature that
shows a 50 percent DR%, as shown in Fig. 10. Although the value, 50%, has no particularly
important meaning, the criteria can be applied in that the micelle structure of the surfactant
changes due to the heat of a solution.
80
60
DR% [%]
40
DR% = 50 %
20
0 upper limit temperature

= 69.0 ℃
-20
20 30 40 50 60 70 80
Temperature [℃]
Fig. 10. Temperature dependence of drag reduction with 500 mg/l Ethoquad O/12 and 300
mg/l NaSal
Figure 11 shows DR% and HTR% of Ethoquad O/12 at different concentrations and
temperature conditions. In the case of 100 mg/l solution, the upper limit temperature of
drag reduction is 52 °C. We note that HTR% is less than DR% when the temperature is
higher than the upper limit temperature. In the case of 300 mg/l solution, the upper limit
temperature is 68 °C, and a similar relation between DR% and HTR% can be observed.
100
DR%, HTR% [%]
50
Ethoquad O/12
100 mg/L
300 mg/L
0 Open : DR%
Closed : HTR%
20 30 40 50 60 70 80
Temperature [℃]
Fig. 11. DR% and HTR% of Ethoquad O/12 at different concentrations and temperatures.
Average velocity of the flow was set at 1.2 m/s
Previous works on heat transfer enhancement have focused on the significant heat transfer
reduction rather than drag reduction, but we want to stress that the suitable additive
condition for an air conditioning system exist for Ethoquad O/12, which shows higher DR%
with lower HTR%. It is also important to keep the concentration of a drag-reducing additive
at a suitable value that can restrain the heat transfer reduction inside the heat exchanger and
also provide enough drag reduction through the pipeline.
Figure 12 shows both the DR% and HTR% of Ethoquad O/13 in comparison with the results
with Ethoquad O/12. The results show that twice as much Ethoquad O/13 as Ethoquad
O/12 is needed to obtain the similar level of DR%.
100
Ethoquad O/12
500p mg/L
80 DR%
DR%, HTR%[%]
HTR%
60 Ethoquad O/13
1000mg/l
DR%
40 HTR%
Ethoquad O/13
500mg/l
20 average velocity DR%
=1.5m/s HTR%
0
20 30 40 50 60 70 80
Temperature[℃]
Fig. 12. DR% and HTR% of Ethoquad O/13 at different concentrations and temperatures.
Average velocity of the flow was set at 1.5 m/s
Since the production cost of Ethoquad O/12 is cheaper, the drag-reducing additive
produced with Ethoquad O/12 has better cost effectiveness. When the concentration of
Ethoquad O/13 decreased due to water leakage from the system during long-term
operation, it is likely that the drag reduction decreased significantly. The suitable additive
condition for Ethoquad O/13, which shows higher DR% with lower HTR%, was not found
in this study.
4. Evaluation of the heat transfer characteristics of a practical air

conditioning system before and after introducing surfactant drag reduction
4.1 Objective
In the preceding section, heat transfer reduction was presented for drag reduction flows,
especially in the case of unsuitable additive conditions. It is not easy to measure the heat
transfer rate accurately for practical air conditioning systems in usual operation; however,
according to our projects with LSP-01, no predicted serious problem of heat transfer has
been detected. It is necessary to obtain quantitative heat transfer characteristics for a
practical air conditioning system. In this study, drag reduction and heat transfer data
measured at our university library building is presented as an example of application in a
practical air conditioning system.
4.2 Experimental procedure

The library building of Yamaguchi University (Figure 13) is two stories with the total floor
space of 2,400 m2. Each floor has its own pipeline system with a heat pump and a chiller unit
(40 RT, 33 kW, Figure 14). Figure 15 shows a schematic diagram of the air conditioning
system. Produced cold or hot water is transported to an air conditioner at the first floor in a
machine room to exchange heat with the air. The temperatures of each position mentioned
in the figure were measured continuously by using resistance bulb thermometers. The water
flow rate was measured by an ultrasonic flow meter (Model UFP-20, Tohki Sangyo, Japan).
All the data were stored in a PC and analyzed.
Fig. 13. Yamaguchi University library building
Fig. 14. Heat pump - chiller units

Experiments were conducted for almost one month during both the summer and winter
seasons in 2007. In the summer season, operation conditions of the air conditioning system
without drag-reducing additives were as follows:
1. The system was started at 7:30 every morning. At this time, the temperature of the
circulating water was around 22 °C.
2. The chiller unit produced cold water. When the circulating water temperature became
11 °C, the unit stopped.
3. When the circulating water temperature became 18 °C, the chiller unit started
automatically.
Input water
Expansion tank
temp. of chiller
unit (2F)
Input water
temp. of chiller
unit (1F)
Chiller Chiller
unit (1F) unit (2F)
2F
Input water
temp. of
Output air temp. of AH(1F)
AH(1F)
Humidity
Output air temp.
１F AH (1F)
of AH(2F)
Outside temp.
Output water AH (2F)
temp. of AH(1F)
Input water
temp. of
AH(2F)
Flow rate (1F)
Fig. 15. Schematic diagram of the air conditioning system. (AH: Air Handling Unit)
After accumulating data for two weeks, we added LSP-01M, which included Ethoquad
O/12, NaSal, and sodium molybdate as a corrosion inhibitor, to the system (Figure 16). The
concentration of LSP-01M was 5000 mg/L, and thus the surfactant concentration was 500
mg/L.
4.3 Experimental results

The water flow rate increased from 5.7 L/s to 6.01 L/s with the addition of 5000 mg/L LSP-
01. To adjust the flow rate to the former value (5.7 L/s) by controlling an inverter system
installed on the main pump, we changed the electric current of the pump from 5.4 A to 4.5A,
and as a consequence, 16.7% reduction of energy consumption was obtained. In the analyses
of all the stored temperature data for a month, we found no significant difference between
before and after the addition of the drag-reducing additive. The operation of this air
conditioning system can be summarized as follows:
1. Measure the output water temperature of the chiller unit.

2. If the measured temperature is higher than an upper setting temperature, run the chiller
unit.
3. If the measured temperature is lower than a lower setting temperature, stop the chiller
unit.
Fig. 16. Addition of LSP-01 from the discharged side of the main pump
Output water temp. of the chiller unit (1F) ［℃］
25
Sep 11
Sep 12
20 Sep 8
Sep 7
Aug 31
Aug 10
15
Aug 9
Aug 8
Aug 2
10 July 28
0 20 40 60 80
Elapsed time [min]
Fig. 17. Output water temperature versus elapsed time for of the chiller unit (1F)
If the heat transfer reduction caused by the drag-reducing additive occurs, the operating
time of the chiller unit should be longer than the normal operation without the additive.
Figure 17 shows the output temperatures of the chiller unit (1F). The figure indicates that the
temperatures were around 20 °C just before the operation of the air conditioning system.
Then, the temperature decreased gradually after the chiller unit was run. After several tens
of minutes, the chiller stopped and the temperature suddenly increased. If the temperature
reached the upper setting temperature, the chiller started. Therefore, the temperature charts
with respect to the elapsed times became zigzag shapes.
To determine the operating intervals of the chiller unit, the following function was used to
fit the down side of the zigzag chart:
⎛ −t ⎞
y = y0 exp⎜⎜ ⎟⎟ (6)
⎝ tR ⎠
where y, y0, t, and tR are the output water temperature, constant, elapsed time, and
relaxation time, respectively. The relaxation time can be used to evaluate the operating
interval. Table 2 indicates the relaxation times for the down side of the first zigzag chart
of each day. The drag-reducing additive was introduced on August 21st; however,
the relaxation times did not show a significant change between before and after the
addition.
Date ｙ０ 1／tR tR [min]

July 28 20.0 0.0117 85.5
Aug 8 20.0 0.0155 64.5
Aug 9 19.6 0.0103 97.1
Aug 10 18.9 0.0093 107.5
Aug 31 18.0 0.0142 70.4
Sep 7 19.6 0.0189 52.9
Sep 8 18.1 0.0209 47.8
Sep 11 19.3 0.0317 31.5
Sep 12 17.4 0.0249 40.2
Table 2. Relaxation time for the down side of the first zigzag chart of each day
4.4 Discussion
In section 3, we described the measurement of the heat transfer characteristics for drag-
reducing flow by the laboratory experimental apparatus, and a significant amount of heat
transfer reduction was observed when the additive conditions were unsuitable. In section
4-3, we evaluated the possibility of a heat transfer reduction occurring and found no
evidence of a heat transfer reduction. One of the following may have caused the difference
in these results:
1) In the case of the laboratory experiment, the heat transfer characteristics were evaluated
based on the reduction of the heat transfer coefficient between the wall of the copper pipe
(heat exchanger) and water. Since it was impossible to measure the wall temperature of the
chilling unit, we used the temperature differences of the inlet and outlet of the chilling unit
to evaluate the heat transfer characteristics. In other words, the heat transfer characteristics
were evaluated based on the overall coefficient of heat transfer, U , which is determined the
following equation:
1 1 x 1
= + + (7)
U hw k ha
where hw, x, k, and ha are the heat transfer coefficient of water, the thickness of the heat
exchanger’s wall, the thermal conductivity of the wall, and the heat transfer coefficient of
air, respectively. Since hw is two orders of magnitude larger than ha, the value of U is not
significantly changed even if the value of hw is reduced due to the heat transfer reduction. To
took at a quantitative example, let us assume the following values:
hw=5800 W/m2·K for water flow
hw=1200 W/m2·K for drag-reducing flow (HTR%=80 %)
x=1 mm, k=350 W/mK, ha=60 W/m2·K
The calculated overall heat transfer coefficient for water flow is 59.4 W/m2·K, while that for
drag-reducing flow is 57.1 W/m2·K. As a result, the heat transfer reduction with respect to
the overall heat transfer coefficient is only a 3.9% reduction, even though the heat transfer
coefficient of fluid was reduced significantly.
2) In the case of the laboratory experiment, the constant heat flux conditions were
maintained during the experiment, which means the difference of the inlet and outlet
temperatures of the copper pipe did not affect the flow conditions. In this experiment, when
the heat transfer occurred, the wall temperature increased. On the other hand, for the
practical air conditioning system, the heat transfer phenomenon occurred unsteadily due to
the intermittent operation of the chilling unit and inadequate mixing of the fluid. Therefore,
even though the heat transfer reduction occurred in response to the drag-reducing
additives, it was not easy to monitor the influence of the reduction of heat transfer.
3) Finally, we want to stress that micelle aggregates of drag-reducing surfactants normally
suffer mechanical degradation in the chiller unit pipe of practical air conditioning systems.
The average velocity inside heat exchangers is usually designed to be in the range of 1.5 m/s
to 2.0 m/s. Also, practical pipelines have many elbows, branches, reduction or expansion
pipes, valves, and so on, in which the shear rates applied to the fluids are larger than the
rated in straight pipes. However, it is quite difficult to evaluate the drag reduction and the
heat transfer reduction in the practical heat exchangers while factoring in the associated
changes in the micelle structure of drag-reducing surfactants.
5. Conclusion
The drag-reducing technology will be applied to an ever increasing range of applications.
There is also enormous potential for application of this approach to other fluids besides
water, so we can expect a lot of research in this area. Finally, we observed certain
characteristic near-wall momentum and heat transfer behavior in flows that calls for further
investigation. This behavior is attributed to the effects of uneven transfer phenomena of
micelle motion in pipes for drag-reducing flows, and investigation of this phenomenon
should lead to new insights regarding the mechanism involved in drag-reducing flows.
6. Acknowledgment
The author would like to express his thanks to Mr. T. Matsumura, LSP Cooperative Union,
and Mr. K. Tokuhara, Shunan Region Local Industry Promotion Center Foundation, for
their valuable discussion on the drag reduction by surfactant solutions. The author also
would like to acknowledge the experimental assistance of a student colleague (Yusuke
Uchiyama).
7. References
Aguilar, G.; Gasljevic, K. & Matthys, E. J. (2001). Asymptotes of Maximum Friction and Heat
Transfer Reductions for Drag-reducing surfactant Solutions, Int. J. of Heat and Mass
Transfer, Vol.44, pp.2835-2843
Gadd, G. E. (1966). Reduction of Turbulent Friction in Liquids by Dissolved Additives,
Nature, Vol. 212, pp. 874-877
Ohlendorf, D.; Interthal, W. & Hoffmann, H. (1986). Surfactant systems for drag
reduction:physico-chemical properties and rheological behaviour, Rheology Acta,
Vol. 25, pp.468-486
Qi, Y.; Kawaguchi, Y.; Lin, Z.; Ewing, M.; Christensen, R. & Zakin, J. (2001). Enhanced Heat
Transfer of Drag Reducing Surfactant Solutions with Fluted Tube-in-tube Heat
Exchanger, Int. J. of Heat and Mass Transfer, Vol.44, pp.1495-1505
Saeki, T.; De Guzman, M.; Morishima, H.; Usui, H. & Nishimura, T. (2000). A Flow
Visualization Study on the Mechanism of Turbulent Drag Reduction by
Surfactants, Nihon Reoroji Gakkaishi, Vol. 28, No.1, pp.35-40
Saeki, T.; Tokuhara, K.; Matsumura, T. & Yamamoto, S. (2002). Application of Surfactant
Drag Reduction for Practical Air Conditioning Systems, Nihon Kikai Gakkai
Ronbunshu, B, Vol. 68, No. 669, pp.1482-1488 (in Japanese)
Shikata T. et al. (1988). Micelle formation of detergent molecules in aqueous media. II: Role
of free salicylate ions on viscoelastic properties of aqueous
cetyltrimethylammonium bromide-sodium salicylate solutions, Langmuir, Vol. 4,
pp. 354-359, 1988
Usui, H. & Saeki T. (1993). Drag Reduction and Heat Transfer Reduction by Cationic
Surfactants, J. of Chem. Eng. Japan, Vol.26, No.1, pp.103-106
Usui, H. et al. (1998). Effect of surfactant molecular structure on turbulent drag reduction,
Kagaku Kogaku Ronbunshu, Vol. 24, No.1, pp.134-137 (in Japanese)
Chou, L.-C.; Christensen, R. N. & Zakin, J. L. (1989). The influence of chemical composition
of quaternary ammonium salt cationic surfactants on their drag reducing
effectiveness, Proceedings of 4th Int. Conf. on Drag Reduction, Davos, IAHR/AIRH,
Ellis Horwood Pub., pp.141-148
Motier, J. F. (2002). Polymer advances and the dramatic growth in commercial pipeline drag
reduction, Proceedings of 12th European Drag Reduction Meeting, Herning, Denmark,
April 18-20, 2002
Steiff, A.; Klopper, K. & Weinspach, P. (1998). Influence of Drag Reducing Additives in Heat
Exchangers, Proceedings of 11th IHTC, 6, pp.317-322, Kyongju, Korea
Toms, B. A. (1948). Some observations on the flow of linear polymers solutions through
straight pipe at large Reynolds numbers. Proceedings of International Congress on
Rheology, Vol. 2, pp. 135-141, Scheveningen, Italy, 1948
Li, B.; Xu, Y. & Choi, J. (1996). Applying Machine Learning Techniques, Proceedings of ASME
2010 4th International Conference on Energy Sustainability, pp. 14-17, ISBN 842-6508-
23-3, Phoenix, Arizona, USA, May 17-22, 2010
19
Entransy - a Novel Theory in

Heat Transfer Analysis and Optimization
Qun Chen, Xin-Gang Liang and Zeng-Yuan Guo
Department of Engineering Mechanics, Tsinghua University
China
1. Introduction
It has been estimated that, of all the worldwide energy utilization, more than 80% involves
the heat transfer process, and the thermal engineering has for a long time recognized the
huge potential for conserving energy and decreasing CO2 release so as to reduce the global
warming effect through heat transfer efficiency techniques (Bergles, 1988, 1997; Webb, 1994;
Zimparov, 2002). In addition, since the birth of electronic technology, electricity-generated
heat in electronic devices has frequently posed as a serious problem (Arden, 2002; Chein &
Huang, 2004), and effective cooling techniques are hence needed for reliable electronic
device operation and an increased device lifespan. In general, approaches for heat transfer
enhancement have been explored and employed over the full scope of energy generation,
conversion, consumption and conservation. Design considerations to optimize heat transfer
have often been taken as the key for better energy utilization and have been evolving into a
well-developed knowledge branch in both physics and engineering.
During the last several decades and promoted by the worldwide energy shortage, a large
number of heat transfer enhancement technologies have been developed, and they have
successfully cut down not only the energy consumption, but also the cost of equipment
itself. However, comparing with other scientific issues, engineering heat transfer is still
considered to be an experimental problem and most approaches developed are empirical or
semi-empirical with no adequate theoretical base (Gu et al., 1990). For instance, for a given
set of constraints, it is nearly impossible to design a heat-exchanger rig with the optimal heat
transfer performance so as to minimize the energy consumption.
Therefore, scientists developed several different theories and methods to optimize heat
transfer, such as the constructal theory (Bejan, 1997) and the minimum entropy generation
(Bejan, 1982). Then heat transfer processes were optimized with the objective of minimum
entropy generation. Based on this method, several researchers (Nag & Mukherjee, 1987;
Sahin, 1996; Sekulic et al., 1997; Demirel, 2000; Sara et al., 2001; Ko, 2006) analyzed the
influences of geometrical, thermal and flow boundary conditions on the entropy generation
in various convective heat transfer processes, and then optimized them based on the
premise that the minimum entropy generation will lead to the most efficient heat transfer
performance. However, there are some scholars (Hesselgreaves, 2000; Shah & Skiepko, 2004;
Bertola & Cafaro, 2008) who questioned whether the entropy generation is the universal
irreversibility measurement for heat transfer, or the minimum entropy generation is the
general optimization criterion for all heat transfer processes, regardless of the nature of the
applications. For instance, by analyzing the relationship between the efficiency and the
entropy generation in 18 heat exchangers with different structures, Shah & Skiepko (2004)
demonstrated that even when the system entropy generation reaches the extremum, the
efficiency of the heat exchangers can be at either the maximum or the minimum, or anything
in between. In addition, the so-called “entropy generation paradox” (Bejan, 1996;
Hesselgreaves, 2000) exists when the entropy generation minimization is used as the
optimization criterion for counter-flow heat exchanger. That is, enlarging the heat exchange
area from zero simultaneously increases the heat transfer rate and improves heat exchanger
efficiency, but does not reduce the entropy generation rate monotonously - the entropy
generation rate increases at first then decreases. Therefore, it was speculated that the
optimization criterion of minimum entropy generation is not always consistent with the
heat transfer improvement.
Recently, Guo et al. (2007) introduced the concepts of entransy and entransy dissipation to
measure, respectively, the heat transfer capacity of an object or a system, and the loss of
such capacity during a heat transfer process. Moreover, Guo et al. (2007) proposed the
entransy dissipation extremum and the corresponding minimum entransy dissipation-based
thermal resistance as alternative optimization criterions for heat transfer processes not
involved in thermodynamic cycles, and consequently, developed the minimum entransy
dissipation-based thermal resistance principle to optimize the processes of heat conduction
(Guo et al., 2007; Chen et al., 2009a, 2011), convective heat transfer (Meng et al., 2005; Chen
et al., 2007, 2008, 2009b), thermal radiation (Cheng & Liang, 2011), and in heat exchangers
(Liu et al., 2009; Guo et al., 2010).
This chapter summarizes the entransy theory in heat transfer, such as the definitions of
entransy, entransy dissipation and its corresponding thermal resistance with multi-
temperatures, the minimum entransy dissipation-based thermal resistance principle for heat
transfer, etc., and introduces its applications in heat transfer optimization. Finally, we will
make comparisons between entransy optimization and entropy optimization to further
examine their applicability to heat transfer optimization in applications of different natures.
2. The origin of entransy

After an intensive study, we found that all transport processes contain two different types of
physical quantities due to the existing irreversibility, i.e., the conserved ones and the non-
conserved ones, and the loss or dissipation in the non-conserved quantities can then be used
as the measurements of the irreversibility in the transport process. Taking an electric system
as an example, although both the electric charge and the total energy are conserved during
an electric conduction, the electric energy however is not conserved and it is partly
dissipated into the thermal energy due to the existence of the electrical resistance.
Consequently, the electrical energy dissipation rate is often regarded as the irreversibility
measurement in the electric conduction process. Similarly, for a viscous fluid flow, both the
mass and the momentum of the fluid, transported during the fluid flow, are conserved,
whereas the mechanical energy, including both the potential and kinetic energies, of the
fluid is turned into the thermal energy due to the viscous dissipation. As a result, the
mechanical energy dissipation is a common measure of irreversibility in a fluid flow
process. The above two examples show that the mass, or the electric quantity, is conserved
during the transport processes, while some form of the energy associated with them is not.
This loss or dissipation of the energy can be used as the measurement of irreversibility in
Entransy - a Novel Theory in Heat Transfer Analysis and Optimization 351
these transport processes. However, an irreversible heat transfer process seems to have its
own particularity, for the heat energy always remains constant during transfer and it
doesn’t appear to be readily clear what the non-conserved quantity is in a heat transfer
process (Chen et al., 2011). Based on the analogy between electrical and heat conductions,
Guo et al. (2007) made a comparison between electrical conduction and heat conduction as
shown in table 1. It could be found in the table that there is no corresponding parameter in
heat conduction for the electrical potential energy in a capacitor, and hence they defined an
equivalent quantity, G, that corresponding to the electrical potential energy in a capacitor
1
G = QvhT , (1)
2
which is called entransy. They further derived Eq. (1) according to the similar procedure of
the derivation of the electrical potential energy in a capacitor. Entransy was originally
referred to as the heat transport potential capacity in an earlier paper by the authors (Guo et
al., 2003).
Electrical charge Electrical current Electrical resistance Capacitance

stored in a capacitor
Qve [C] I [C/s]=[A] Re [Ω] Ce = Qve/Ue [F]
Thermal energy stored Heat flow rate Thermal resistance Heat capacity
in a body
Qvh = McvT [J] Q h [W] Rh [K/W] Ch = Qvh/T [J/K]
Electrical potential Electrical current Ohm’s law Electrical potential

density energy in a capacitor
Ue [V] q e [A/m2] q e = -Ke∇Ue Ee= QveUe/2
Thermal potential Heat flux density Fourier’s law Thermal potential in

(temperature) a body
Uh = T [K] q h [W/m2] q h = -Kh∇T ?
Table 1. Analogy between electrical and thermal conductions (Guo et al., 2007)
Entransy represents the heat transfer ability of an object (Guo et al., 2007). It possesses both
the nature of “energy” and the transfer ability. If an object is put in contact with an infinite
number of heat sinks that have infinitesimally lower temperatures, the total quantity of
“potential energy” of heat which can be output is ½QvhT. Biot (1955) suggested a similar
concept in the derivation of the differential conduction equation using the variation method.
Eckert & Drake (1972) pointed out that “Biot in a series of papers beginning in 1955
formulated from the ideas of irreversible thermal dynamics a variational equivalent of the
heat-conduction equation that constituted a thermodynamical analogy to Hamilton’s
principle in mechanics and led to a Lagrangian formulation of the heat conduction problem
1
in terms of generalized coordinates…, Biot defines a thermal potential E = ∫∫∫ ρ cT 2 dV ….
2 Ω
The thermal potential E plays a role analogous to a potential energy…”. However, Biot did
not further explain the physical meaning of thermal potential and its application was not
found later except in the approximate solutions of anisotropic conduction problems.
Accompanying the electric charge, the electric energy is transported during electric
conduction. Similarly, along with the heat, the entransy is transported during heat transfer
too. Furthermore, when a quantity of heat is transferred from a high temperature to a low
temperature, the entransy is reduced and some of entransy is dissipated during the heat
transport. The lost entransy is called entransy dissipation. Entransy dissipation is an
evaluation of the irreversibility of heat transport ability. For instance, let’s consider two
bodies A and B with the initial temperatures TA and TB, respectively. Their initial entransy is
G1 =
1
2
(
M Ac vATA2 + M Bc vBTB2 , ) (2)
where M is the body mass, cv the specific heat at constant volume, and the subscripts A and
B represent the bodies A and B, respectively. When these two bodies contact each other,
thermal energy will flow from the higher temperature body to the lower temperature one.
After infinite long time, their temperature will be the same and the entransy becomes
1
G2 = ( M Ac vA + MBc vB )T22 , (3)
2
where T2 is the equilibrium temperature
M Ac vATA + MBc vBTB

T2 = . (4)
M Ac vA + M Bc vB
The entransy difference before and after equilibrium is
1 M A M Bc vAc vB (TA − TB )
2
G1 − G2 = >0. (5)
2 M Ac vA + MBc vB
Equation (5) proves that the total entransy is reduced after the two bodies getting
equilibrium, i.e. there is an entransy dissipation associating with the heat transport.
3. Entransy balance equation and optimization for heat conduction (Guo et

al., 2007; Chen et al., 2011)
3.1 Entransy balance equation for heat conduction
In a heat conduction process, the thermal energy conservation equation is expressed as
∂T
ρcv = −∇ ⋅ q + Q , (6)
∂t
where ρ is the density, t the time, q the heat flow density, and Q the internal heat source.
Multiplying both sides of Eq. (6) by temperature T gives an equation which can be viewed
as the balance equation of the entransy in the heat conduction:
∂T
ρ c vT = −∇ ⋅ ( qT
) + q ⋅ ∇T + QT
, (7)
∂t
that is,
∂g
= −∇ ⋅ ( qT
) − φh + g , (8)
∂t
where g=G/V=uT/2 is the specific entransy, V the volume, u the specific internal energy,
the entransy flow density, g the entransy change due to heat source, and φh can be taken
qT
as the entransy dissipation rate per unit volume, expressed as
φh = −q ⋅ ∇T (9)
The left term in either Eq. (7) or (8) is the time variation of the entransy stored per unit
volume, consisting of three items shown on the right: the first represents the entransy
transferred from one (or part of the) system to another (part), the second term can be
considered as the local entransy dissipation during the heat conduction, and the third is the
entransy input from the internal heat source. It is clear from Eq. (9) that the entransy is
dissipated when heat is transferred from high temperature to low temperature. Thus, heat
transfer is irreversible from the viewpoint of entransy, and the dissipation of entransy can
hence be used as a measurement of the irreversibility in heat transfer.
3.2 The entransy dissipation extremum principle for heat conduction

Heat transfer optimization aims for minimizing the temperature difference at a given heat
transfer rate,
δ ( ΔT ) = δ f ( x , y , z ,τ , T , k , q , ρ , c v , ") = 0 , (10)
Or maximizing the heat transfer rate at a given temperature difference,
δ ( Q ) = δ g ( x , y , z ,τ , T , k , q , ρ , c v , ") = 0 . (11)
In conventional heat transfer analysis, it is difficult to establish the relationship between the
local temperature difference, or local heat transfer rate, and the other related physical
variables over the entire heat transfer area, so the variational methods in Eqs. (10) and (11)
are not practically useable. However, the entransy dissipation in Eq. (9) is a function of the
local heat flux and local temperature gradient in the heat transfer area, and thus the
variational method will become utilizable if written in terms of the entransy dissipation
(Cheng, 2004).
Integrating the balance equation of the entransy Eq. (7) over the entire heat transfer area
gives:
∂T
∫∫∫ ρc vT ∂t dV = ∫∫∫ −∇ ⋅ ( qT
) dV + ∫∫∫ q ⋅ ∇TdV + ∫∫∫ QTdV
. (12)
Ω Ω Ω Ω
For a steady state heat conduction problem, the left term in Eq. (12) vanishes, i.e.
0 = ∫∫∫ −∇ ⋅ ( qT
) dV + ∫∫∫ q ⋅ ∇TdV + ∫∫∫ QTdV
. (13)
Ω Ω Ω
If there is no internal heat source in the heat conduction domain, Eq. (13) is further reduced
into:
∫∫∫ −∇ ⋅ ( qT
) dV = ∫∫∫ −q ⋅ ∇TdV . (14)
Ω Ω
By transforming the volume integral to the surface integral on the domain boundary
according to the Gauss’s Law, the total entransy dissipation rate in the entire heat
conduction domain is deduced as
Φ h = ∫∫∫ −q ⋅ ∇TdV = ∫∫ −qTdS

= ∫∫ q inTin dS − ∫∫ q outTout dS . (15)
Ω Γ Γ+ Γ−
where Γ+ and Γ- represent the boundaries of the heat flow input and output, respectively.
The continuity of the total heat flowing requires a constant total heat flow Q t ,
Q t = ∫∫ q in dS = ∫∫ q out dS . (16)
Γ+ Γ−
We further define the ratio of the total entransy dissipation and total heat flow as the heat
flux-weighted average temperature difference ΔT
Φh q in q
Q t Γ∫∫+ Q t
ΔT = = Tin dS − ∫∫ out Tout dS . (17)
− Qt

Γ
For one-dimensional heat conduction, Eq. (17) is reduced into ΔT = (Tin – Tout), exactly the
conventional temperature difference between the hot and cold ends. Using the heat flux-
weighted average temperature difference defined in Eq. (17) and applying the divergence
theorem, a new expression for optimization of a steady-state heat conduction at a given heat
flow rate can be constructed as:
Q tδ ( ΔT ) = δ ∫∫∫ −q ⋅ ∇TdV = δ ∫∫∫ k ∇T dV = 0 .

2
(18)
Ω Ω
It shows that when the boundary heat flow rate is given, minimizing the entransy
dissipation leads to the minimum in temperature difference, that is, the optimized heat
transfer. Conversely, to maximize the heat flow at a given temperature difference, Eq. (18)
can be rewritten as:
1 2
ΔTδ Q t = δ ∫∫∫ −q ⋅ ∇TdV = δ ∫∫∫ q dV = 0 , (19)
Ω Ω
k
showing that maximizing the entransy dissipation leads to the maximum in boundary heat
flow rate.
Likewise, for a steady state heat dissipating process with internal heat source in Eq. (13), the
total entransy dissipation rate in the entire heat conduction domain is derived as

Φ h = ∫∫∫ −q ⋅ ∇TdV = ∫∫∫ QTdV − ∫∫ q outTout dS . (20)
Ω Ω Γ−
Since the heat generated in the entire domain will be dissipated through the boundaries, i.e.,
Q t = ∫∫∫ QdV
= ∫∫ q out dS . (21)
Ω Γ−
Again the heat flux-weighted average temperature is defined as the entransy dissipation
over the heat flow rate
Φ Q q
ΔT = h = ∫∫∫ TdV − ∫∫ out Tout dS , (22)
Qt Ω
Q t Γ − Q t
and thus the optimization of the process is achieved when
Q tδ ( ΔT ) = δ ∫∫∫ −q ⋅ ∇TdV = δ ∫∫∫ k ∇T dV = 0 ,

2
(23)
Ω Ω
which means that in a heat dissipating process, minimizing the entransy dissipation leads to
the minimum averaged temperature over the entire domain.
Based on the results from Eqs. (18), (19) and (23), it can be concluded that the entransy
dissipation extremum (EDE) lead to the optimal heat transfer performance at different
boundary conditions. This extreme principle can be concluded into the minimum thermal
resistance principle defined by entransy dissipation. (Guo et al., 2007)
3.3 Application to a two-dimensional volume-point heat conduction

We will apply our proposed approach to practical cases where heat transfer is used for
heating or cooling such as in the so-called volume-point problems (Bejan, 1997) of heat
dissipating for electronic devices as shown in Fig. 1. A uniform internal heat source
distributes in a two-dimensional device with length and width of L and H, respectively. Due
to the tiny scale of the electronic device, the joule heat can only be dissipated through the
surroundings from the “point” boundary area such as the cooling surface in Fig. 1, with the
opening W and the temperature T0 on one boundary. In order to lower the unit temperature,
a certain amount of new material with high thermal conductivity is introduced inside the
device. As the amount of the high thermal conductivity material (HTCM) is given, we need
to find an optimal arrangement so as to minimize the average temperature in the device.
Fig. 1. Two-dimensional heat conduction with a uniformly distributed internal heat source
(Chen et al., 2011)
According to the new extremum principle based on entransy dissipation, for this volume-
point heat conduction problem, the optimization objective is to minimize the volume-
average temperature, the optimization criterion is the minimum entransy dissipation, the
optimization variable is the distribution of the HTCM, and the constraints is the fixed
amount of the HTCM, i.e.,
∫∫∫ k ( x , y ) dV = const . (24)

Ω
By variational method, a Lagrange function, Π, is constructed
2
(
Π = ∫∫∫ k ∇T + Bk dV .
Ω
) (25)
where the Lagrange multiplier B remains constant due to a given amount of thermal
conductivity.
The variation of Π with respect to temperature T gives
G
∫∫ k∇Tδ T ⋅ ndS − ∫∫∫ ∇ ⋅ ( k∇T )δ TdV = 0 . (26)
Γ Ω
Because the boundaries are either adiabatic or isothermal, the surface integral on the left
side of Eq. (26) vanishes, that is,
G
∫∫ k∇Tδ T ⋅ ndS = 0 . (27)
Γ
Moreover, due to a constant entransy output and a minimum entransy dissipation rate, the
entransy input reaches the minimum when

δ ∫∫∫ QTdV = ∫∫∫ Q δ TdV = 0 . (28)
Ω Ω
Substituting Eqs. (27) and (28) into Eq. (26) in fact gives the thermal energy conservation
equation based on Fourier's Law:
∇ ⋅ ( k∇T ) + Q = 0 . (29)
This result validates that the irreversibility of heat transfer can be measured by the entransy
dissipation rate. The variation of Π with respect to thermal conductivity k gives
2
∇T = −B = const . (30)
This means that in order to optimize the heat dissipating process, i.e. to minimize the
volume-average temperature, the temperature gradient should be uniform. This in turn
requires that the thermal conductivity be proportional to the heat flow in the entire heat
conduction domain, i.e. the HTCM be placed at the area with the largest heat flux.
As an example, the cooling process in low-temperature environment is analyzed here. For
the unit shown in Fig. 1, L = H = 5 cm, Q = 100 W/cm2, W = 0.5 cm and T0 = 10 K. The
thermal conductivity of the unit is 3 W/(m·K), and that for the HTCM is 300 W/(m·K)
occupying 10% of the whole heat transfer area.
Fig. 2(b) shows the distribution of the HTCM according to the extremum principle of
entransy dissipation, where the black area represents the HTCM - the same hereinafter. (The
implementary steps are as shown in the reference of Chen et al., 2011.) The HTCM with a
tree structure absorbs the heat generated by the internal source and transports it to the
isothermal outlet boundary - similar in both the shape and function of actual tree roots.
(a) Simple uniform HTCM arrangement (b) HTCM arrangement using the extremum
principle of entransy dissipation
Fig. 2. Different arrangements of HTCM (Chen et al., 2011)
(a) From the uniform HTCM arrangement (b) From the optimized arrangement of
in Fig. 2(a) HTCM in Fig. 2(b)
Fig. 3. The temperature fields obtained from different arrangements of HTCM (Chen et al.,
2011)
For a fixed amount of HTCM, Figs. 3(a) and 3(b) compare the temperature distributions
between a uniform distribution of HTCM shown in Fig. 2(a), and the optimized distribution
in Fig. 2(b) based on the extremum principle of entransy dissipation. The average temperature
in the first case is 544.7 K while the temperature in the second optimized case is 51.6 K, a
90.5 % reduction! It clearly demonstrates that the optimization criterion of entransy
dissipation extremum is highly effective for such applications. Furthermore, as shown in

Fig. 3(b), the temperature gradient field is also less fluctuating in the optimized case.
In addition, based on the concept of the entransy dissipation rate, we (Chen et al. 2009a)
introduced the non-dimensional entransy dissipation rate and employed it as an objective
function to analyze the thermal transfer process in a porous material.
4. Entransy balance equation and dissipation for convective heat transfer

4.1 Entransy balance equation for convective heat transfer (Cheng, 2004)
For a convection problem, the energy equation is
⎡ DT D ⎛ 1 ⎞⎤
ρ ⎢c p + P ⎜ ⎟ ⎥ = φμ − ∇ ⋅ q + Q , (31)
⎣ Dt Dt ⎝ ρ ⎠ ⎦
where P, t and φμ are the pressure, time, and viscous dissipation, respectively. Multiply Eq.
(31) by T and by some subsequent derivation (Cheng, 2004), there is
⎛ ∂g ⎞
ρ ⎜ + U ⋅ ∇g ⎟ = −∇ ⋅ ( qT
) + QT
+ φ T + q ⋅ ∇T ,
μ (32)
⎝ ∂t ⎠
where g=cpT2/2, is the entransy per unit volume of the fluid. On the left, the 1st term is time
variation of entransy, and the 2nd term is entransy variation accompanying fluid flow; on the
right, the 1st term is entransy variation due to boundary heat exchange, the 2nd term is the
entransy change due to heat source, the 3rd term is the entransy variation due to dissipation
and the 4th term is the entransy dissipation. It is clear from Eq. (32) that the heat transport
ability is reduced when heat is transferred from high temperature to low temperature.
4.2 The entransy dissipation extremum principle for convective heat transfer
Similar to the derivation of EDE principle in heat conduction, for a steady-state convective
heat transfer process of a fluid with constant boundary heat flux and ignoring the heat
generated by viscous dissipation, integrating Eq. (32), and transforming the volume integral
to the surface integral on the domain boundary yields:
⎛ T + Tout ⎞ 2
Q0δ ⎜ Tw − in ⎟ = δ ∫∫∫ k ∇T dV . (33)
⎝ 2 ⎠ Ω
Since the inlet and outlet temperatures, Tin and Tout, of the fluid are fixed for a given
boundary heat flux, Equation (33) means that, the minimum entransy dissipation in the
domain is corresponding to the minimum boundary temperature, i.e. the minimum
boundary temperature difference.
Similarly, from (32) the maximum entransy dissipation is obtain,
Q0
(Tw − Tin )δ Q0 − 2
δ Q0 = δ ∫∫∫ k ∇T dV . (34)
ρc p Ω
Because Eq. (33) illustrates that for the range of the boundary heat flux from 0 to its
maximum value, ρVcp(Tw - Tin), the entransy dissipation increases monotonically with the
increasing boundary heat flux, that means maximum entransy dissipation results in the
maximum boundary heat flux. Equation (33) together with Eq. (34) is called entransy
dissipation extreme (EDE) principle in convective heat transfer. Thus, the entransy
dissipation can be used to optimize the flow field with given viscosity dissipation so that the
heat transfer rate can be increased most with given cost of pressure lose (Guo et al., 2007;
Meng et al. 2005, Chen et al. 2007).
4.3 Application to laminar heat transfer (Meng et al., 2005)

For a steady laminar flow, the viscosity dissipation is
⎡ ∂u 2 ⎛ ∂v ⎞
2
∂w ⎞
2 ⎤
⎢ 2 ⎛⎜ ⎞⎟ + 2 ⎜ ⎟ + 2 ⎛⎜ ⎟ ⎥
⎢ ⎝ ∂x ⎠ ⎝ ∂y ⎠ ⎝ ∂z ⎠ ⎥
φm = μ ⎢ 2 2
⎥, (35)
2
⎢ ⎛ ∂u ∂v ⎞ ⎛ ∂u ∂w ⎞ ⎛ ∂v ∂w ⎞ ⎥
⎢+ ⎜ + ⎟ + ⎜ + ⎟ +⎜ + ⎟ ⎥
⎢⎣ ⎝ ∂y ∂x ⎠ ⎝ ∂z ∂x ⎠ ⎝ ∂z ∂y ⎠ ⎥⎦
where u, v and w are the velocity components along x, y and z directions, respectively. The
mechanical work maintaining the fluid flow equals to the integral of the viscous dissipation
function over the whole domain,
Πφ m = ∫∫∫ φm dV . (36)
Ω
The optimal flow field was established by configuring a Lagrange functional that includes
the objective and constraint functions. The established Lagrange function is,
⎣ ( )
∏ = ∫∫∫ ⎡ k∇T ⋅ ∇T + C 0φm + A ∇ ⋅ k∇T − ρ c pU ⋅ ∇T + B∇ ⋅ ρU ⎤ dV ,
⎦
(37)
Ω
where A, B and C0 are the Lagrange multipliers. Because of the different types of the
constraints, A and B vary with position, while C0 remains constant for a given viscous
dissipation. The first term on the right is the entransy dissipation, the second is the
constraint on prescribed viscous dissipation, the third and forth terms are the constraints on
energy equation and continuum equation. The variational of Eq. (37) with respect to velocity
U offers
ρc p 1
μ∇ 2U + A∇T + ∇B = 0 . (38)
2C 0 2C 0
The variational of Eq. (37) with respect to temperature T is
− ρ c pU ⋅ ∇A = ∇ ⋅ k∇A − 2∇ ⋅ k∇T , (39)
and the boundary conditions of the variable A are Ab = 0 for given boundary temperatures,
and ( ∂A ∂n )b = 2 ( ∂T ∂n )b for given boundary heat flow rates.
Comparing the momentum equations and Eq. (38), gives the following relations
B = −2C 0 P , (40)
F = C Φ A∇T + ρU ⋅ ∇U , (41)
where CΦ is related to the viscous dissipation as
ρc p
CΦ = . (42)
2C 0
Substituting Eqs. (41) and (42) into the momentum equation gives
ρU ⋅ ∇U = −∇P + μ∇ 2U + (C Φ A∇T + ρU ⋅ ∇U ) . (43)
This is the Euler’s equation, i.e. field synergy equation, governing the fluid velocity and
temperature fields with the EDE principle during a convective heat transfer process.
Equation (43) essentially the momentum equation with a special additional volume force
defined in Eq. (41), by which the fluid velocity pattern is adjusted to lead to an optimal
temperature field with the extremum of entransy dissipation during a convective heat
transfer process.
For a given CΦ, solving Eqs. (39) and (43), the continuity equation and the energy equation
in combination result in the optimal flow field with the extremum of entransy dissipation
with a specific viscous dissipation and fixed boundary conditions. Fig. 4 shows a typical
numerical result of the cross-sectional flow field (Re=400, CΦ=-0.01). Compared with the
fully-developed laminar convection heat transfer in a circular tube ((fRe)s=64, Nus=3.66), the
flow viscous dissipation is increased by 17%, the Nusselt number is increased by 313% in the
case of Fig. 4. Meanwhile, the numerical analysis shows that multiple longitudinal vortex
flow is the optimal flow pattern for laminar flow in tube. With the guide of this optimal flow
field, the discrete double-inclined ribs tube (DDIR-tube) is designed to improve heat transfer
in laminar flow. Fig. 5 is a typical cross-sectional vortex flow of a DDIR-tube and experiments
have demonstrated the DDIR-tube have better heat transfer performance but lower
resistance increase (Meng et al., 2005).
Fig. 4. Optimum flow field of laminar heat transfer in circular tube (Re=400) (Meng et al.,
2005)
5%U m
o
0
Fig. 5. Numerical solution of cross-sectional flow fields in the DDIR-tube (Meng et al., 2005)
4.4 Application to turbulent heat transfer (Chen et al., 2007)

For turbulent heat transfer, the entransy dissipation function is modified as:
2
φht = keff ∇T , (44)
where keff is the effective thermal conductivity during turbulent heat transfer.
The optimization objective of turbulent heat transfer is to find an optimal velocity field,
which has the extremum of entransy dissipation for a given decrement of the time-averaged
kinetic energy, i.e. a fixed consumption of pumping power. Meanwhile, the flow is also
constrained by the continuity equation and the energy equation. All these constraints can
also be removed by using the Lagrange multipliers method to construct a function
⎣ ( )
∏t = ∫∫∫ ⎡ keff ∇T ⋅ ∇T + C 0φmt + A ∇ ⋅ keff ∇T − ρ c pU ⋅ ∇T + B∇ ⋅ ρU ⎤ dV .
⎦
(45)
Ω
where φmt is the decrement function of the mean kinetic energy
⎡ ∂u 2 ⎛ ⎞
2 2 ⎤
⎢ 2 ⎛⎜ ⎞⎟ + 2 ⎜ ∂v ⎟ + 2 ⎛⎜ ∂w ⎞⎟ ⎥
⎢ ⎝ ∂x ⎠ ⎝ ∂y ⎠ ⎝ ∂z ⎠ ⎥
φmt = ( μ + μt ) ⎢ ⎥. (46)
2 2
⎢ ⎛ ∂u ∂v ⎞ ⎛ ∂u ∂w ⎞2 ⎛ ∂v ∂w ⎞ ⎥
⎢+ ⎜ + ⎟ +⎜ + ⎟ +⎜ + ⎟ ⎥
⎢⎣ ⎝ ∂y ∂x ⎠ ⎝ ∂z ∂x ⎠ ⎝ ∂z ∂y ⎠ ⎥⎦
The decrement of the mean kinetic energy equals to the viscous dissipation by the viscous
forces, plus the work of deformation of the mean motion by the turbulence stresses, which
transform the mean kinetic energy to the turbulence-energy. The turbulent viscosity, μt,
which is a function of the velocity, can be calculated using the Prandtl’s mixing-length
model, the Prandtl-Kolmogorov assumption or the k-ε model. However, the Prandtl-
Kolmogorov assumption and the k-ε model require additional differential equations for the
turbulent kinetic energy (k) and the turbulent dissipation rate (ε) to calculate the turbulent
viscosity, which severely complicate obtaining the variation of Eq. (45). Thus, a single
algebraic function, which has been validated and used for indoor air flow simulation (Chen
& Xu, 1998), is used to calculate the turbulent viscosity:
μt = 0.03874 ρ U l , (47)
where, l is the distance to the nearest wall. As with other zero-equation turbulent models,
Eq. (47) is not very theoretically sound, but yields some reasonable results for turbulent
flows.
The variational of Eq. (45) with respect to temperature T is:
− ρ c pU ⋅ ∇A = ∇ ⋅ keff ∇A − 2∇ ⋅ keff ∇T . (48)
The variational of Eq. (45) with respect to velocity component u is:
1 ∂B ρ c p A ∂T Γu ⎡ φ Γu
2C 0 ∂x
( )
+ ∇ ⋅ μ eff ∇u + −
2C 0 ∂x 2C 0 Prt ⎣
∇T + A∇ ⋅ ( ∇T ) ⎤ − mt
2
⎦ 2 μ eff
AΓ ∂T ⎡ u ∂l ⎛ ∂u ∂v ∂w ⎞ −2 ∂u ⎤
− − ⎜u + v + w ⎟U u+ ⎥
2C 0 Prt ∂x ⎢⎣ l ∂x ⎝ ∂x ∂x ∂x ⎠ ∂x ⎦
AΓ ∂T ⎡ u ∂l ⎛ ∂u ∂v ∂w ⎞ −2 ∂u ⎤
− ⎢ − ⎜u + v + w ⎟U u+ ⎥ . (49)
2C 0 Prt ∂y ⎣⎢ l ∂y ⎝ ∂y ∂y ∂y ⎠ ∂y ⎦⎥
AΓ ∂T ⎡ u ∂l ⎛ ∂u ∂v ∂w ⎞ −2 ∂u ⎤
− ⎢ − ⎜u + v + w ⎟U u+ ⎥
2C 0 Prt ∂z ⎣ l ∂z ⎝ ∂z ∂z ∂z ⎠ ∂z ⎦
⎛ ∂μ eff ∂u ∂μ eff ∂v ∂μ eff ∂w ⎞ 1 ⎛ AΓu ⎞
+ ⎜⎜ + + ⎟⎟ + ∇⋅⎜ ∇T ⎟ = 0
⎝ ∂x ∂x ∂y ∂x ∂z ∂x ⎠ 2C 0 ⎝ Prt ⎠
−1
where Γ = 0.03874 ρ l U , Prt is the turbulent Prantdl number with the value of 0.9. The
variationals of Eq. (45) with respect to the velocity components v and w are similar to Eq.
(49). There are four unknown variables and four governing equations including Eqs. (48),
(49), the continuity equation and the energy equation, so the unknown variables can be
solved for a given set of boundary conditions. Meanwhile, the flow must also satisfy the
momentum equation,
ρU ⋅ ∇U = −∇P + μ eff ∇ 2U + F . (50)
Comparison with Eq. (50), Eq. (49) is actually a momentum equation with a special
additional volume force, which is referred to as the field synergy equation for turbulent heat
transfer. For a given set of boundary conditions, the optimal velocity field, which has a
larger heat transfer rate than any other field flow, can be obtained by solving this field
synergy equation.
As an example, flow in a parallel plate channel is studied to illustrate the applicability of the
field synergy equation. For simplicity, a repeated segment with the height of 20 mm and the
length of 2.5 mm was chosen. Water flowing between the parallel plates is assumed
periodically fully developed with a Reynolds number of 20,000. The inlet water temperature
is 300 K and the wall temperature is 350 K.
It is well known that for original results before optimization, the velocity vectors and the
temperature gradients are nearly perpendicular to each other, leading to a small scalar
product between the velocity vector and the temperature gradient, i.e. the field synergy
degree (Meng et al. 2003) is relatively poor. In the computational domain, the heat transfer
rate is 1782 W and the decrement of the mean kinetic energy is 4.61×10-3 W. The thicknesses
of the laminar sublayer and the transition sublayer are 0.116 mm and 0.928 mm, respectively.
The optimized velocity and temperature fields near the upper wall for C0 = -1.5×107 are
shown in Figs. 6 (a) and (b). There are several small counter-clockwise eddies near the upper
wall. The distances between eddies centres is about 0.4 mm and the eddy heights
perpendicular to the primary flow direction are about 0.2 mm. There are also several small
clockwise eddies near the lower wall for symmetry. For this case, the heat transfer rate is
1887 W and the decrement of the mean kinetic energy is 5.65×10-3 W. Compared with the
original results before optimization, the heat transfer rate is increased by 6 %, while the
decrement of the mean kinetic energy is increased by 23 %.
For heat transfer in turbulent flow between parallel plates, the temperature gradients in the
laminar sublayer are two to three orders of magnitude larger than the gradients far from the
wall, which means that the thermal resistance in the laminar sublayer is the dominate
resistance in turbulent flow. The conventional heat transfer enhancement viewpoint is to
first reduce the dominate resistance to most effectively increase the heat transfer rate. Eddies
and disturbances near the wall will increase the velocities, reduce the thermal resistances,
and enhance the heat transfer. The results support that the tubes with micro fins effectively
enhance turbulent heat transfer, which is different from multiple longitudinal vertex
generation in laminar heat transfer, and further point out the optimal heights of the fins for
different Reynolds numbers should be half of the transition sublayer of turbulent flow,
which is also validated experimentally (Li et al., 2009).
(a) velocity vectors
(b) temperature contours

Fig. 6. Optimized results near the wall for turbulent heat transfer (Chen et al., 2007)
5. Entransy dissipation-based thermal resistance for heat exchanger design

In engineering, both designing and checking heat exchanger performance are generally used
such four approaches as the logarithmic mean temperature difference method (LMTD), the
heat exchanger effectiveness – number of transfer units method (ε-NTU), the P-NTU method,
and the ψ-P method. However, these methods are not very convenient when used in some
particular situations. For example, in the LMTD method, it is inevitable to introduce a
correction factor to adjust the effective temperature difference for cross-flow and multi-pass
exchangers. Meanwhile, the use of LMTD method for checking heat exchanger performance
has to involve tedious iterations to determine the proper outlet fluid temperatures and
thereafter the value of LMTD which satisfy the requirement that heat transferred in the heat
exchanger be equal to the heat carried out by the fluid. On the other hand, in the ε-NTU
method, the fluid with the minimum heat capacity rate has to be first taken as the
benchmark to calculate both heat exchanger effectiveness ε and NTU, so iterations are also
unavoidable for the design of fluid flow rates. Besides, different types, e.g. parallel, counter-
flow, cross-flow and shell-and-tube, of heat exchangers have their individual diverse and
complex relations between heat exchanger effectiveness ε and NTU, which are more or less
inconvenient for engineering applications. Therefore, it is necessary to develop a general
criterion for the evaluation of heat exchanger performance and, more importantly, develop a
common method for heat exchanger performance design and optimization.
For a heat transfer process in a parallel heat exchanger, the heat lost by the hot fluid over a
differential element should be the same as that gained by the cold fluid, which both equal to
the heat transferred through the elements, that is
dq = −mh dhh = mc dhc . (51)
where, m is the mass flow rate, h is the specific enthalpy, and the subscripts h and c
represent the hot and cold fluids, respectively.
Integrating Eq. (51), we will obtain the total heat transfer rate in the heat exchange from the
viewpoint of energy conservation
Q = mh ( hh , a − hh ,b ) = mc ( hc , d − hc ,c ) , (52)
For a heat exchanger without any phase-change fluid, if the fluid specific heats are constant,
Eq. (51) will be rewritten as
1
dTh = − dq , (53)
mh c p , h
1
dTc = dq , (54)
mc c p ,c
and Eq. (52) is rewritten as
Q = mhc p , h (Th , a − Th ,b ) = mc c p ,c (Tc , d − Tc ,c ) . (55)
Based on Eqs. (53) and (54), Fig. 7 gives the fluid temperature variations versus the heat
transfer rate q. As shown, the shaded area is:
dS = Th dq − Tc dq . (56)
where, the first term on the right-hand side represents the entransy output accompanying
the thermal energy dq flowing out of the hot fluid, while the second term represents the
entransy input accompanying the thermal energy dq flowing into the cold fluid. Therefore,
the shaded area exactly indicates the entransy dissipation rate during the heat transferred
from the hot fluid to the cold one:
dφh = (Th − Tc ) dq . (57)
The total entransy dissipation in the heat exchanger is deduced by integrating Eq. (57),
Φh Q (Th ,a − Tc ,c ) + (Th ,b − Tc ,d ) Q = ΔT
Φh = ∫ dφh = ∫ (Th − Tc ) dq = AM Q . (58)
0 0 2
where, ΔTAM is the arithmetical temperature difference between the hot and cold fluids in
the heat exchanger.
Fig. 7. Sketch of the fluid temperature variations versus the heat transfer rate in a parallel
heat exchanger
Substituting Eqs. (55), (58) and the heat transfer equation
Q = KA
(Th ,a − Tc ,c ) − (Th ,b − Tc ,d ) (59)
ln (Th , a − Tc ,c ) − ln (Th ,b − Tc , d )
into the definition of entransy dissipation-based thermal resistance, EDTR, for heat
exchangers (Liu et al., 2009), Rh = Фh/Q2 = ΔTAM/Q, we get the formula of such thermal
resistance for parallel heat exchangers:
Rh , p =
(
ξ p exp KAξ p + 1 ) , (60)
(
2 exp KAξ p − 1 )
⎛ 1 1 ⎞
where, ξ p = ⎜ + ⎟ , termed the arrangement factor for parallel heat exchangers.
⎜ mhc p , h mc c p ,c ⎟
⎝ ⎠
Similarly, for counter-flow or TEMA E-type heat exchangers, we can also deduce the
same formula of EDTR as shown in Eq. (60), except that the expression of the arrangement
⎛ 1 1 ⎞
factor ξ is diverse, i.e. ξc = ⎜ − ⎟ for counter-flow heat exchangers and
⎜ mhc p , h mc c p ,c ⎟
⎝ ⎠
1 1
ξs = + for TEMA E-type heat exchangers.
( ) ( )
2 2
mc c p ,c mhc p , h
This general expression is convenient for us to design the heat transfer performance of heat
exchangers. For example, a one shell and two tube passes TEMA E-type shell-and-tube
exchanger with the heat transfer coefficient K and area A of 300 W/(m2·K) and 50 m2,
respectively, is used to cool the lubricating oil from the initial temperature Th,in = 57 °C to the
desired temperature Th,out = 45 °C. The mass flow rate and specific heat of the oil are
mh = 10 kg/s and cp,h = 1.95 kJ/(kg·K), respectively. If the cooling water enters the heat
exchanger at the temperature of 33 °C, what are its heat capacity rate and outlet
temperature, and what is the rate of heat transfer by the exchanger.
According to the relation between EDTR and arithmetical mean temperature difference, the
total heat transfer rate in the exchanger is
ΔTAM (Th ,in + Th , out − Tc ,in − Tc , out ) ( exp ( KAξs ) − 1 )

Q= = . (61)
Rh ,s ξs ( exp ( KAξs ) + 1 )
Numerically solving Eq. (61) and the energy conservation equation (55) simultaneously, we
can easily obtain the heat capacity rate and the outlet temperature of the cooling water are
73.8 kJ/(s·K) and 36.17 °C, respectively, and the total heat transfer rate is 234 kJ.
In this problem, neither the heat capacity rate nor the exit temperature of the cooling water
are known, therefore an iterative solution is required if either the LMTD or the ε-NTU
method is to be used. For instance, when using the LMTD method, the detail steps are: (1)
obtain the required heat transfer rate of the exchanger Q1 from the energy conservation
equation of the oil; (2) assume a heat capacity rate of the cooling water, and then calculate its
exit temperature; (3) according to the inlet and outlet temperatures of both the oil and the
cooling water, obtain the logarithm mean temperature difference and the correction factor of
the shell-and-tube exchanger; (4) based on the heat transfer equation, derive another heat
transfer rate of the exchanger Q2. Because the heat capacity rate of the cooling water is
assumed, iteration is unavoidable to make the derived heat transfer rate Q2 in step 4 close to
the required one Q1. Thus, it is clear that the entransy dissipation-based thermal resistance
method can design heat exchanger performance conveniently.
6. Differences between entransy and entropy

Besides entransy, the concepts of entropy and entropy generation are considered to be two
important functions in thermodynamic and used for estimating the irreversibility of and
optimizing heat transfer based on the premise that the minimum entropy generation (MEG)
will lead to the most efficient heat transfer performance. Thus, there exist two optimization
principles, and it is highly desired to investigate the physical essentials and the applicability
of, as well as the differences between, them in heat transfer optimization.
The local entropy generation function induced by the heat transfer over finite temperature
difference is
2
k ∇T
sg = 2
. (62)
T
Let’s reconsider the same volume-point problems as shown in Fig. 1. What we seek is also
the optimal HTCM arrangement, except in this case that: 1. the optimization criterion is the
minimum entropy generation; 2. the corresponding energy conservation equation should be
added as a constraint, because it is not implied in the principle of minimum entropy
generation when the thermal conductivity is constant (Bertola & Cafaro, 2008).
Introducing the corresponding Lagrange function
⎡ ∇T 2 ⎤
(
Π ' = ∫∫∫ ⎢ k 2 + B ' k + C ' ∇ ⋅ k∇T + Q )⎥⎥ dV . (63)
Ω ⎢⎣ T ⎦
where B’ and C’ are also the Lagrange multipliers. The constraint of thermal conductivity is
the isoperimetric condition, and consequently B’ is a constant. C’ is a variable related to
space coordinates. The variation of Π’ with respect to temperature T gives
2
2 k ∇T 2Q
−∇ ⋅ ( k∇C ') = 3
+ . (64)
T T2
while the variation of Π’ with respect to thermal conductivity k yields
∇T 2
∇C '⋅ ∇T − = B ' = const . (65)
T2
Likewise, Eq. (65) gives the guideline for optimization based on the criterion of minimum
entropy generation. That is, the HTCM be placed at the area with the extreme in absolute
value of ∇C '⋅ ∇T − ∇T 2 T 2 .
Fig. 8(a) shows the distribution of HTCM based on the principle of minimum entropy
generation. Comparison of Figs. 2(b) and 8(a) shows that although the distributions of
HTCM are similar between the two results in most areas, the root-shape structure from the
minimum entropy generation principle is not directly connected to the heat flow outlet,
leaving some parts with the original material in between them so that the heat cannot be
transported smoothly to the isothermal outlet boundary. Fig. 8(b) gives the optimized
temperature distribution obtained by the minimum entropy generation. Because the low
thermal conductivity material is adjacent to the heat outlet, the temperature gradient grows
larger and thus lowers the entire heat transfer performance. The averaged temperature of
the entire area is 150.8 K - 99.2 K higher than that obtained by the extremum principle of
entransy dissipation. From the definition, it is easy to find that in order to decrease the
entropy generation, we have to both reduce the temperature gradient, and raise the
temperature, thus leading to the arrangement of HTCM showed in Fig. 8(a).
In addition, according to the principle of minimum entropy generation, the optimization

objective of a steady-state heat dissipating process can be expressed as:
2
⎛ 1⎞ k ∇T
Q δ ⎜ Δ ⎟ = δ ∫∫∫ dV = 0 . (66)
⎝ T ⎠m Ω T2
⎛ 1⎞ ⎛1 1 ⎞
where ⎜ Δ ⎟ = ⎜ − ⎟ , is the equivalent thermodynamics potential difference which
⎝ T ⎠m ⎝ T T0 ⎠m
represents the “generalized force” in the entropy picture for heat transfer. Thus, minimizing
the entropy generation equals to minimizing the equivalent thermodynamics potential
⎛ 1⎞
difference ⎜ Δ ⎟ , leading to the highest exergy transfer efficiency. That is, the minimum
⎝ T ⎠m
entropy generation principle is equivalent to the minimum exergy dissipation during a heat
transfer process.
(a) HTCM arrangement (b) Temperature field

Fig. 8. Optimized results using the minimum entropy generation principle (Chen et al., 2011)
To facilitate the comparison between the two results from Figs. 3(b) and 8(b), Table 2 lists
the key findings side by side, obtained respectively by the optimization criteria of the
minimum entropy generation and the entransy dissipation extremum. It indubitably shows
in the table that the proposed entransy based approach is more effective than the entropy
based one in heat transfer optimization, for the former leads to a result with significantly
reduced mean temperature than that by the latter (51.6 K vs. 150.8 K), and much lower
maximum temperature (83.0 K vs. 194.9 K). Whereas the entropy based approach is
preferred in exergy transfer optimization, as it results in a significantly lower equivalent
thermodynamic potential (7.1×10-3 /K vs. 2.2×10-2 /K).
Besides heat conduction, we (Chen et al. 2009b) also compared the two criteria in convective
heat transfer optimization. Our results indicate that both principles are applicable to
convective heat transfer optimization, subject however to different objectives. The minimum
entropy generation principle works better in searching for the minimum exergy dissipation
during a heat-work conversion, whereas the entransy dissipation extremum principle is
more effective for processes not involving heat-work conversion, in minimizing the heat-
transfer ability dissipation.
Optimization
results ⎛ 1⎞
Optimization
Φh /(W•K) Sgen/(W/K) Tm/K Tmax/K ⎜ Δ ⎟ /(1/K)
⎝ T ⎠m
Criterions
Extremum Entransy
5.5×104 100.7 51.6 83.0 2.2×10-2
Dissipation
Minimum Entropy
1.58×105 81.7 150.8 194.9 7.1×10-3
Generation
Table 2. Optimized results obtained by the optimization criterions of minimum entropy
generation and entransy dissipation extremum (Chen et al., 2011)
7. Conclusion
The entransy is a parameter that is developed in recent year. It is effective in optimization of
heat transfer. Entransy is an evaluation of the transport ability of heat. Both the amount of
heat and the potential contribute to the entransy. Entransy will be lost during the heat
transportation from a high temperature to lower one and entransy dissipation will be
produces. Based on the energy conservation equation, the entransy balance equations for
heat conduction and convective heat transfer are developed. The entransy dissipation
extreme principles are developed, that is, the maximum entransy dissipation corresponds to
the maximum heat flux for prescribed temperature difference and the minimum entransy
dissipation corresponds to minimum temperature difference for prescribed heat flux. This
extreme principle can be concluded into the minimum thermal resistance principle defined
by entransy dissipation.
The entransy dissipation-based thermal resistance of heat exchangers is introduced as an
irreversibility measurement, for parallel, counter-flow, and shell-and-tube exchangers,
which has a general expression, is the function of heat capacity rates and thermal
conductance, and may analyze, compare and optimize heat exchanger performance from the
physical nature of heat transfer. Besides, from the relation among heat transfer rate,
arithmetical mean temperature difference and EDTR, the total heat transfer rate may easily
be calculated through the thermal conductance of heat exchangers and the heat capacity
rates of fluids, which is convenient for heat exchanger design.
Finally, we compared the criteria of entransy dissipation extremum to that of entropy
generation minimization in heat transfer optimization, the results indicates that the
minimum entransy dissipation-based thermal resistance yields the maximum heat transfer
efficiency when the heat transfer process is unrelated with heat-work conversion, while the
minimum entropy generation leads to the highest heat-work conversion when such is
involved in a thermodynamic cycle.
8. Acknowledgments
This work was financially supported by the National Basic Research Program of China
(2007CB206901) and the National Natural Science Foundation of China (Grant No. 51006060).
9. References
Arden W. M. (2002). The International Technology Roadmap for Semiconductors -
Perspectives and Challenges for the Next 15 Years. Current Opinion in Solid Sate and
Materials Science, Vol.6, No.5, (October 2002), pp. 371-377, ISSN 1359-0286
Bejan A. (1982). Entropy Generation through Heat and Fluid Flow, John Wiley & Sons, ISBN 0-
471-09438-2, New York, USA.
Bejan A. (1996). Entropy Generation Minimization: the Method of Thermodynamic Optimization of
Finite-size Systems and Finite-time Processes, CRC Press, ISBN 0-849-39651-4, Boca
Raton, USA
Bejan A. (1997). Constructal-theory Network of Conducting Paths for Cooling a Heat
Generating Volume. International Journal of Heat and Mass Transfer, Vol.40, No.4,
(March 1997), pp. 799-811, ISSN 0017-9310
Bergles A. E. (1988). Some Perspectives on Enhanced Heat Transfer --- 2nd-generation Heat
Transfer Technology. Journal of Heat Transfer-Transactions of the Asme, Vol.110,
No.4B, (November 1988), pp. 1082-1096, ISSN 0022-1481
Bergles A. E. (1997). Heat Transfer Enhancement --- The Encouragement and Accommodation
of High Heat Fluxes. Journal of Heat Transfer-Transactions of the Asme, Vol.119, No.1,
(Feburary 1997), pp. 8-19, ISSN 0022-1481
Bertola V. & Cafaro E. (2008). A Critical Analysis of the Minimum Entropy Production
Theorem and Its Application to Heat and Fluid Flow. International Journal of Heat
and Mass Transfer, Vol.51, No.7-8, (April 2008), pp. 1907-1912, ISSN 0017-9310
Biot M. (1955). Variational Principles in Irreversible Thermodynamics with Application to
Viscoelasticity. Physical Review, Vol.97, No.6, (March 1955), pp. 1463-1469, ISSN
0031-899X
Chein R., & Huang G. (2004). Thermoelectric Cooler Application in Electronic Cooling.
Applied Thermal Engineering, Vol.24, No.14-15, (October 2004), pp. 2207-2217, ISSN
1359-4311
Chen Q., Ren J. X. & Meng J. A. (2007). Field Synergy Equation for Turbulent Heat Transfer
and Its Application. International Journal of Heat and Mass Transfer, Vol.50, No.25-26,
(December 2007), pp. 5334-5339, ISSN 0017-9310
Chen Q. & Ren J. X. (2008). Generalized Thermal Resistance for Convective Heat Transfer
and Its Relation to Entransy Dissipation. Chinese Science Bulletin, Vol.53, No.23,
(December 2008), pp. 3753-3761, ISSN 1001-6538
Chen Q., Wang M. R., Pan N. & Guo Z. Y. (2009a). Irreversibility of Heat Conduction in
Complex Multiphase Systems and Its Application to the Effective Thermal
Conductivity of Porous Media. International Journal of Nonlinear Sciences and
Numerical Simulation, Vol.10, No.1, (January 2009), pp. 57-66, ISSN 1565-1339
Chen Q., Wang M. R., Pan N. & Guo Z. Y. (2009b). Optimization Principles for Convective
Heat Transfer. Energy, Vol.34, No.9, (September 2009), pp. 1199-1206, ISSN 0360-
5442
Chen Q. & Xu W. (1998). A Zero-equation Turbulence Model for Indoor Airflow Simulation.
Energy and buildings, Vol.28, No.2, (October 1998), pp. 137-144, ISSN 0378-7788
Chen Q., Zhu H. Y., Pan N. & Guo Z. Y. (2011). An Alternative Criterion in Heat Transfer
Optimization. Proceedings of the Royal Society a-Mathematical Physical and Engineering
Sciences, Vol.467, No.2128, (April 2011), pp. 1012-1028, ISSN 1364-5021
Cheng X. G. (2004). Entransy and Its Applications in Heat Transfer Optimization. (Doctoral
Dissertation), Tsinghua University, Beijing, China
Cheng X. T. & Liang X. G. (2011). Entransy Flux of Thermal Radiation and Its Application to
Enclosures with Opaque Surfaces. International Journal of Heat and Mass Transfer,
Vol.54, No.1-3, (January 2011), pp. 269-278, ISSN 0017-9310
Demirel Y. (2000). Thermodynamic Analysis of Thermomechanical Coupling in Couette
Flow. International Journal of Heat and Mass Transfer, Vol.43, No.22, (November
2000), pp. 4205-4212, ISSN 0017-9310
Gu W. Z., Shen S. R. & Ma C. F. (1990). Heat Transfer Enhancement, Science Press, ISBN 7-03-
001601-7, Beijing, China
Guo Z. Y., Liu X. B., Tao W. Q. & Shah R. (2010). Effectiveness-thermal Resistance Method
for Heat Exchanger Design and Analysis. International Journal of Heat and Mass
Transfer, Vol.53, No.13-14, (June 2010), pp. 2877-2884, ISSN 0017-9310
Guo Z. Y., Zhu H. Y. & Liang X. G. (2007). Entransy --- a Physical Quantity Describing Heat
Transfer Ability. International Journal of Heat and Mass Transfer, Vol.50, No.13-14,
(July 2007), pp. 2545-2556, ISSN 0017-9310
Guo Z. Y., Cheng X. G. & Xia Z. Z. (2003). Least Dissipation Principle of Heat Transport
Potential Capacity and Its Application in Heat Conduction Optimization. Chinese
Science Bulletin, Vol.48, No.4, (Feburary 2003), pp. 406-410, ISSN 1001-6538
Hesselgreaves J. E. (2000). Rationalisation of Second Law Analysis of Heat Exchangers.
International Journal of Heat and Mass Transfer, Vol.43, No.22, (November 2000), pp.
4189-4204, ISSN 0017-9310
Ko T. H. (2006). Numerical Analysis of Entropy Generation and Optimal Reynolds Number
for Developing Laminar Forced Convection in Double-sine Ducts with Various
Aspect Ratios. International Journal of Heat and Mass Transfer, Vol.49, No.3-4,
(Febuary 2006), pp. 718-726, ISSN 0017-9310
Li X. W., Meng J. A. & Guo Z. Y. (2009). Turbulent Flow and Heat Transfer in Discrete
Double Inclined Ribs Tube. International Journal of Heat and Mass Transfer, Vol.52,
No.3-4, (January 2009), pp. 962-970, ISSN 0017-9310
Liu X. B. & Guo Z. Y. (2009). A Novel Method for Heat Exchanger Analysis. Acta Physica
Sinica, Vol.58, No.7, (July 2009), pp. 4766-4771, ISSN 1000-3290
Meng J. A. (2003). Enhanced Heat Transfer Technology of Longitudinal Vortices Based on Field-
Coordination Principle and Its Application. (Doctoral Dissertation), Tsinghua
University, Beijing, China
Meng J. A., Liang X. G. & Li Z. X. (2005). Field Synergy Optimization and Enhanced Heat
Transfer by Multi-longitudinal Vortexes Flow in Tube. International Journal of Heat
and Mass Transfer, Vol.48, No.16, (July 2005), pp. 3331-3337, ISSN 0017-9310
Nag P. K. & Mukherjee P. (1987). Thermodynamic Optimization of Convective Heat
Transfer through a Duct with Constant Wall Temperature. International Journal of
Heat and Mass Transfer, Vol.30, No.2, (Feburary 1987), pp. 401-405, ISSN 0017-9310
Sahin A. Z. (1996). Thermodynamics of Laminar Viscous Flow through a Duct Subjected to
Constant Heat Flux. Energy, Vol.21, No.12, (December 1996), pp. 1179-1187, ISSN
0360-5442
Sara O. N., Yapici S., Yilmaz M. & Pekdemir T. (2001). Second Law Analysis of Rectangular
Channels with Square Pin-fins. International Communications in Heat and Mass
Transfer, Vol.28, No.5, (July 2001). pp. 617-630, ISSN 0735-1933
Sekulic D. P., Campo A. & Morales J. C. (1997). Irreversibility Phenomena Associated with
Heat Transfer and Fluid Friction in Laminar Flows through Singly Connected
Ducts. International Journal of Heat and Mass Transfer, Vol.40, No.4, (March 1997),
pp.905-914, ISSN 0017-9310
Shah R. K. & Skiepko T. (2004). Entropy Generation Extrema and Their Relationship with
Heat Exchanger Effectiveness – Number of Transfer Unit Behavior for Complex
Flow Arrangements. Journal of Heat Transfer - Transactions of the ASME, Vol.126,
No.6, (December 2004), pp. 994-1002, ISSN 0022-1481
Webb R. L. (1994). Principles of Enhanced Heat Transfer, Taylor & Francis Group, ISBN 0-471-
57778-2, Wiley, New York, USA
Zimparov V. (2002). Energy Conservation through Heat Transfer Enhancement Techniques.
International Journal of Energy Research, Vol.26, No.7, (June 2002), pp. 675-696, ISSN
0363-907X
20
Transient Heat Transfer and

Energy Transport in Packed
Bed Thermal Storage Systems
Pei Wen Li1, Jon Van Lew1, Wafaa Karaki1,
Cho Lik Chan1, Jake Stephens2 and James. E. O’Brien3
1Department of Aerospace and Mechanical Engineering
The University of Arizona, Tucson, AZ 85721,

2US Solar Holdings LLC., 1000 E. Water Street, Tucson, AZ 85719,
3Idaho National Laboratory, Idaho Falls, Idaho 83415,
USA
1. Introduction
Compared to fossil fuel energy resources, the major types of renewable energy—such as
wind power, solar energy, ocean currents, and tidal energy—generally possess the innate
characteristics of intermittence of availability, fluctuation of magnitude, as well as low
energy density (Li, 2008). However, the utilization of energy and power in industry, living,
and working often requires high energy densities, and demand may be out of phase with
the period of availability of renewable energy. In other words, renewable energy is not
always load following. This variability creates a demand for energy storage when people
develop renewable energy technologies (Kolb, 1998).
Among the several types of renewable energy, solar energy has the largest proportion of the
total available and may be directly used as thermal energy in conventional thermal power
plants, or converted into electrical power directly using photovoltaic panels. Although
direct electrical energy storage in batteries or capacitors may have a high efficiency, it is still
very challenging and expensive—particularly when storing a large quantity of electrical
energy (Spiers, 1995). Electricity may be indirectly stored by pumping water to reservoirs, or
by compressing air, or by electrolyzing water and making hydrogen fuel, etc. However,
these methods often have low round-trip efficiency (from electricity to electricity), or are
restricted by the availability of geographical conditions or suitable locations. In comparison,
it has been recognized that direct solar thermal energy storage is relatively easy to approach
at a reasonably low cost and high efficiency, and the energy storage capacity can be much
larger than that of direct electricity storage (Price et al., 2002; Montes et al, 2009). Thermal
energy storage systems use materials that can be kept at high temperatures in insulated
containers. The heat retrieved can then be used in conventional thermal power plants for
power generation at times when sunlight is not available or when weather conditions are
not favorable (Singer et al, 2010; Laing et al, 2010).
Researchers worldwide have done a great amount of research and development on

concentrated solar thermal power generation technologies in the last ten years (Renewable
Energy Policy Network for the 21st Century, 2007). Particularly, with these efforts solar
trough and solar tower concentrated thermal power generation technologies have become
more and more reliable and matured, and the cost of concentrated solar power systems have
been significantly reduced due to increased productivity and demand (Pitz-Paal et al., 2007;
Herrmann & Kearney, 2002; Gil et al., 2010.).
It has been widely recognized that further cost reduction of electricity generation using
concentrated solar thermal power may be accomplished by adding solar thermal storage
systems. Storage provides the heat necessary for operation of thermal power plants when
sunlight availability is out of phase, and thus increases the operational capacity (in terms of
the daily operational time) of the power plants. In addition, the extended operation of solar
thermal power plants using stored thermal energy can significantly improve the power
dispatch ability (Herrmann & Kearney, 2002) of the power plant.
Other than electrical power generation, solar thermal energy can also be stored for sundry
applications such as house heating, hot water supply, industrial drying processes, as well as
heating for greenhouse agriculture and animal husbandry. From an energy efficiency
perspective, direct use of solar thermal energy for heating is much more efficient than using
electricity for heating, as electricity generation requires much more input of other types of
energy than the generated electricity. Therefore, although thermal energy storage is not a
new technology, it will receive more and more attention, particularly with the development
of renewable energy technologies.
2. Brief review of thermal storage techniques

The ideal scenario for thermal energy storage is such that the energy-carrying fluid is
stored in a thermal storage system and can be withdrawn at a temperature of no
degradation from that of when the fluid was stored. On the basis of this fundamental
distinction, one may classify thermal energy storage systems into two categories: (1) a
system with direct storage of heat transport fluid, which may approach the performance of
an ideal thermal storage system; (2) two-medium heat storage system, which has a fluid
serving as heat- carrying medium, and another medium, either solid materials or a liquid,
serving the purpose as of a primary thermal storage material. A two-medium heat storage
system uses a reduced amount of expensive heat transfer fluid while sacrificing the energy
storage efficiency.
A first generation direct heat transfer fluid storage system usually has two storage tanks, as
shown in Fig. 1(a), one for hot fluid and the other for cold fluid (Herrmann at al., 2004).
During the energy storage process, fluid from the cold tank is pumped to the solar field to
be heated and then stored in the hot fluid tank; while during the energy discharging
process, fluid from the hot fluid tank is pumped out to release heat to the power plant and
afterwards, flows back to the cold fluid tank. Although there are two tanks in such a thermal
storage system, the heat transfer fluid only occupies a volume equivalent to that of one tank
at any instant in time. This means that the elimination of one tank in the system is possible.
As a consequence, a second generation direct heat transfer fluid storage system has only one
tank, as shown in Fig. 1(b). A stratification of fluid, which maintains hot fluid on top of cold
fluid, is important to such a single tank thermal storage system (Michel et al., 2009; Abdoly
& Rapp, 1982; Krane R.J. & Krane M.J.M., 1992; Brosseau et al., 2005). During a heat charging
Transient Heat Transfer and Energy Transport in Packed Bed Thermal Storage Systems 375
process, hot fluid is injected from the top of the tank, whereas cold fluid is pumped out from
bottom of the tank. The opposite is true for the heat discharge process, during which hot
fluid is pumped out from top of the tank and cold fluid is injected into the tank from the
bottom (Canada et al., 2006).
Hot fluid
Hot fluid
QT QT QT
QT
Solar Solar
Cold fluid
Cold fluid
(a) (b)
Fig. 1. Thermal storage using heat transport fluid only
Two-medium heat storage systems must have a heat transport fluid and a primary thermal
storage material (Laing et al, 2006), either solid or liquid. Also, a two-medium heat storage
system typically uses only a single tank. Depending on the contact and heat transfer
interaction between the heat-carrying fluid and the primary energy storage material, the
storage system may have two types. The first type, as shown in Fig. 2(a), is one that includes
loosely packed solid materials (such as rocks, pebbles of metals, and capsules of phase
change materials (PCM), etc.) as a porous bed held in a container and through which the
heat transport fluid (HTF) flows and transports energy to or from the solid material. In this
type of thermal storage system, the heat transfer between thermal storage material and the
heat transport fluid is relatively efficient due to the ubiquitous contact between the fluid and
storage materials. The second type of two-medium heat storage system, as shown in Fig.
2(b), is such that the heat transfer fluid flows in tubes or pipes that run through thermal
storage material, either solid (such as, concrete, wax, sands, soil (Nassar et al., 2006), salts,
etc), or liquid (such as oil, or liquid salts, etc). Due to the smaller contact area between the
fluid and thermal storage material, the heat transfer between the fluid and the thermal
storage material in this case is worse compared to that of the first type.
Cost-effectiveness is always the dictating criterion for selecting a thermal storage system for
a specific application. However, there are many factors that can influence the cost of a
thermal storage system; for example, the cost of the heat transfer fluid is a key factor which
can determine whether a direct HTF storage system or a two-medium thermal storage
should be used. A high-pressure tank may be needed if the vapor pressure of the heat
transfer fluid is high, or a stainless steel tank may be needed to mitigate corrosion problems.
Possible chemical interaction between tank material, the heat transfer fluid, and primary
thermal storage material, if applicable, must also be prevented.
The following section provides a survey of suitable heat transport fluids and solid materials
for thermal energy storage applications.
Hot fluid
QT QT QT QT
Solar Solar
Cold fluid
(a) direct contact heat transfer (b) heat transfer through fluid pipes
Fig. 2. Thermal storage using a primary thermal storage material with heat transport fluid
3. Thermal storage materials and heat transfer fluids

The quantity of sensible thermal energy stored in a mass is given by the equation:
QT = V ρ C (TH − TL ) (1)
where V , ρ , and C are the volume of the mass, the average density of the mass, and the
heat capacity of the material, respectively.
Obviously, a high specific heat and a large density are both important for a substance to be a
good thermal energy storage material, for that will keep the volume of the storage container
small. Other required properties such as high thermal conductivity, low cost, low thermal
expansion coefficient, etc. are also important. For a liquid to serve as a heat transport fluid
or a thermal storage material, a low solidification point, high boiling point, and low vapor
pressure, are all important. Particularly, if the vapor pressure is high, it will require pressurized
storage tanks, which can significantly increase the cost of the thermal storage system.
3.1 Properties of solid thermal storage materials

Solid materials, suck as concrete (Zhang et al., 2004), sand, rock, brick, soil, graphite, silicon
carbide, taconite, cast iron, and even waste metal chips, have been considered or applied for
thermal energy storage purposes. Depending on the formation and granular size of these
solid materials, some of them can be used to form a packed bed storage system as illustrated
in Fig. 2(a) and others may only be suitable for use in a storage system as shown in Fig. 2(b).
The main properties of some of these materials (Tritt, 2005; Hasnain, 1998) are given in Table 1.
Compared to sensible heat storage, the high latent heat associated with phase change of a
material offers the potential for higher energy storage densities. There are two approaches
using phase change material (PCM) for thermal storage. One is encapsulation of small
amounts of PCM in spherical or cylindrical capsules (Wu et al., 2011), which can be
arranged to form a packed bed. Heat transport fluid can flow through the packed bed for
energy delivery and extraction. This approach needs encapsulation technologies, either
coating or canning. The other approach is to embed the PCM in a matrix made of another
solid material with high heat conduction, and HTF pipes run through the PCM matrix (Sari
& Kaygusuz, 2001; Regin et al., 2008). The use of a matrix material (e.g. graphite or metal
mesh) helps enhance heat conduction in the PCM.
melting(°C)
Medium ρ (kg/m3) C(kJ/kg·°C) ρ ·C(kJ/m3·°C) k(W/m·°C)
(or crumbles)
Aluminum 660 2700 0.92 2484.0 250
Brick (common) 1800 1920 1.0 1920.0 1.04
Fireclay 1800 2100-2600 1.0 2100-2600 1-1.5
Soil (dry) 1650 1200-1600 1.26 1512-2016 1.5
Granite 1215 2400 0.79 1896 1.7-4.0
Sand (dry) 1500 1555 0.8 1244 0.15-0.25
Sandstone 1300 2000-2600 0.92 1840-2392 2.4
Rocks 1800 2480 0.84 2086.6 2-7
Concrete 1000 (Crumbles) 2240-2400 0.75 1680-1800 1.7
Graphite 3500 2300-2700 0.71 1633-1917 85
Silicon carbide 2730 3210 0.75 2407.5 3.6
Taconite 1538 3200 0.8 2560 1.0-2.0
Cast iron 1150 7200 0.54 3888 42-55
Table 1. Properties of solid material suitable for thermal energy storage application
There are many suitable PCM for relatively high temperature thermal storage purposes,
with melting temperatures ranging from below 100 Celsius up to several hundred Celsius.
Paraffin is a typical PCM used for low-grade heat storage. For concentrated high
temperature solar thermal energy storage, alloy and molten salts are often considered. Table
2 includes the properties of some alloys and salts (Zalba et al., 2003).
Medium ρ Heat of fusion Melting temperature

kg/m3 (kJ/kg) (°C)
Mg/Cu/Ca (52/25/23) 2000 184 453
Mg/Cu/Zn (60/25/15) 2800 254 452
Al/Cu/Mg/Zn (54/22/18/6) 3140 305 520
Al/Mg/Zn (59/35/6) 2380 310 443
Al/Cu/Si (65/30/5) 2730 422 571
NaNO3 2260 172 307
KNO3 2110 266 333
K2CO3/Na2CO3(51/49) 2400 163 710
CaCl/NaCl(67/33) 2160 281 500
Table 2. Properties of alloy and salt PCM for thermal energy storage application
3.2 Heat transfer fluids for thermal storage application

Heat transport fluid (HTF), either used for direct fluid storage or as a heat-carrying medium,
must have favorable properties for heat transfer and at the same time, must be stable and
have a low vapor pressure. Molten salts and oils are the two major types of heat transport
fluids developed so far.
Most molten salts have low vapor pressures, which is an advantage to the large quantity
storage of thermal energy as no pressurized tanks are needed. However, many molten salts
freeze at relatively high temperatures (in the range of 120 oC to 250 oC), which can cause
severe problems if freezing or solidification occurs in a solar heat collection system. There
has been some research done worldwide to develop eutectic salt mixtures to lower the
freezing points of salts. If the freezing points of salts are reduced to about 100 oC the freezing
problem is expected to be much more manageable. Table 3 lists the properties of some
typical eutectic molten salts popular in concentrated solar thermal power plant applications
(Bradshaw & Siegel, 2009).
Medium (Company) Components Freezing/Maximum °C

Hitec XL (Costal Chemical) NaNO3 +KNO3 +Ca(NO3)2 140/500
Hitec (Same as above) NaNO3 +KNO3 +NaN O2 142/538
Hitec Solar Salt (Same as above) NaNO3 +KNO3 240/593
Table 3. Molten salts suitable as heat transfer fluids
Mineral oils, or synthetic oils, are popular HTFs in trough concentrated solar thermal power
plants. For large quantity thermal storage large containers are used. If the vapor pressure of
oil is high, vessels for thermal energy storage must withstand high pressure, which can
dramatically drive the cost up. Therefore, synthetic oil or mineral oil is usually used as HTF
but not for fluid that flows through heat storage system (Becker, 1980). Table 4 lists several
mineral oils and synthetic oils which are tipically used as HTF for concentrated solar
thermal power application (Therminol VP-1, 1999; Produc resources, Radco).
Medium (Company) Components Maximum/boiling °C

Xceltherm® series (Radco Industries, Inc) Synthetic oil ~310
Therminol® series (Solutia) synthetic oil ~400
Table 4. Major synthetic oils suitable for HTF in concentrated solar thermal power plant
4. Model of an ideal thermal energy storage system

In an ideal thermal storage system high temperature heat transfer fluid is stored, and when
it is withdrawn, there should be no temperature degradation. Such a system requires that
there be no heat loss and no heat transfer when the HTF is stored in or withdrawn from a
tank.
Contingent upon the thermal insulation being perfectly maintained, the two-tank heat
transfer fluid storage system in Fig. 1(a) can operate like an ideal thermal energy storage
system. It has been discussed before that a two-tank storage system can be replaced by a
single tank storage system as shown in Fig. 1(b), in which a stratification of fluid (hot on top
of cold), or a thermocline mechanism, must be maintained. However, even if the
thermocline is maintained, the heat conduction between hot fluid and cold fluid may cause
a temperature drop in the hot fluid, which will not allow for ideal thermal storage
performance in such a system. A modification proposed by the current authors (Van Lew et
al., 2009) used a thermal insulation baffle in the single tank, which separates the hot fluid
from the cold fluid, as shown in Fig. 3. In this system, if the floating thermal insulation baffle
prevents heat conduction from the hot fluid to the cold fluid, an ideal thermal storage
performance can be achieved. In respect to both the cost reduction and energy storage
performance, the single tank with floating thermal insulation baffle will be the ideal thermal
storage system considered in this book chapter.
Hot fluid
Floating
QT thermal QT
insulation
Solar baffle
Cold fluid
Fig. 3. Schematic illustration of a single tank ideal thermal storage system

Whereas physically an ideal thermal storage system has the clearly identifiable features
previously detailed, mathematically it should be described as a system that has an energy
storage efficiency of 1.0. With this in mind, the following definition of thermal energy
delivery efficiency is adopted for thermal energy storage systems:
tref , disch arg e
∫ [T f ( z = H ,t ) − TL ]dt
η= 0
(2)
(TH − TL ) ⋅ tref , disch arg e
where z is the vertical coordinate of the tank and H is the height of the tank. For adopting
this definition, we assume that the average heat capacity, Cp , and the mass flow rates of the
heat transfer fluid for the charging and discharging processes are the same. The integration
on the numerator of Eq. (2) is the energy discharge in an actual process, and the value on the
denominator is the ideal energy discharge.
For the ideal thermal storage system, it is assumed that the temperature of the hot fluid in a
charging process is kept constant at TH ; and in the discharging process the discharged fluid
keeps a constant temperature of TH as well. After releasing heat in a heat exchanger, the
fluid returns to the bottom of the storage tank at a constant temperature, TL . To substitute
these conditions from the ideal thermal storage system into Eq(2), the fluid temperature
T f ( z = H ,t ) during the discharge process in a time period of 0 to tref , disch arg e should be equal to
the high temperature, TH . This will make the energy delivery efficiency equal to η = 1.0 for
the ideal thermal storage system.
In a real thermal energy storage system, such as the systems shown in Fig. 2, it is easy to
understand that when cold fluid is pumped into the tank from the bottom, it will extract
heat from the solid thermal storage material and be warmed up when it flows out of the
tank. However, after a certain time, the cold fluid going into the tank may not be heated up
sufficiently before it flows out from top of the tank. Unfortunately, this temperature
degradation is inevitable due to the heat transfer between the solid thermal storage material
and the heat transfer fluid, even if initially the solid thermal storage material is fully
charged, or its temperature is exactly equal to TH.
Considering the need of heat transfer fluid in a power plant, it is always important that
during the required operational period of time, tref , disch arg e , the temperature of the heat
transfer fluid have minimum or no degradation from the temperature the fluid is stored at.
To meet this requirement in an actual thermocline storage system, one needs to first store a
sufficient amount of energy (more than the ideal amount) in the tank. This requires a storage
tank to have a sufficiently large thermal energy storage capacity as well as a sufficiently long
charge time that allows heat to be charged to the tank. Giving this requirement as a
mathematical expression,it is such that:
{ ⎡⎣ ρ C ( 1 − ε ) + ρ C ε ⎤⎦V } > ⎡⎣ ( ρ C )V
s s f f real f f ideal
⎤
⎦
(3)
In engineering reality, one needs to know, specifically, how large the real thermal storage
volume, Vreal , is and how long a charging time is necessary, if the assumed operation time
period of a power plant is tref , disch arg e . This must be addressed through mathematical
analysis.
In the following section, the modeling of the heat transfer and energy transport between the
solar thermal storage material and the heat transfer fluid will be described. The goal of the
modeling analysis is to predict the size of the storage tank and the period of time required to
charge the tank for a given subsequent period of heat discharge from the system, within
which minimum or no temperature degradation must be maintained.
5. Analysis of transient heat transfer and energy transport in a packed bed

5.1 Thermal storage/delivery process explained
The following qualitative analysis helps readers better understand the behavior of the fluid
temperature variation from a thermocline tank during a discharging operation. When a tank
of volume Videal is filled with thermal storage material at a void fraction of ε , the volume
for heat transfer fluid in the tank will be εVideal , and the volume of the primary thermal
storage material must be the remaining portion, (1 − ε )Videal . Due to the existence of solid
filler material in the tank, the heat transfer fluid velocity in the charge/discharge processes
is higher than that in an ideal thermocline tank as the same mass flow rate of heat transfer
fluid is considered. The mass flow rate is a condition determined by the required power
output of the power plant, regardless of what type of thermal storage system is used.
Assume that a thermocline tank is initially fully charged. During a thermal discharge
process the temperature of the fluid flowing out from top will decrease after a time when
the pre-existing hot fluid in the tank is completely discharged; from then on, the hot fluid
discharged out from the top is originated from cold fluid that picks up energy from the solid
material during the discharging process. The longer the discharge process progresses, the
more the temperature of the discharged fluid will decrease. This scenario is illustrated in
Fig. 4.
In order to avoid the temperature degradation as show in Fig. 4, one either needs to use an
ideal thermal storage system or to have a thermocline system that stores much more energy
than is needed so that during the required time period, tref , disch arg e , the temperature
degradation of the discharged fluid is minimal or ignorable.
Texit (top)
TH
TL
t
tref, discharge
Fig. 4. The temperature variation of heat transfer fluid during discharge from a tank having
filler material
5.2 A generalized mathematical modeling

In general, the thermocline thermal storage systems shown in Fig. 2(a) and 2(b) can be
regarded as systems that have porous medium, through which a fluid flows and heat
transfer between the solid and fluid occurs. Therefore, generalized governing equations for
the energy balance in the solid material and fluid can be constructed. As a representative
case of a tank filled with porous medium, the system shown in Fig. 2(a) is subjected to
analysis for the objective of formulating an analytical model.
Shown in Fig. 5 is a one-dimensional control volume of an element dz in the packed bed. For
convenience in analysis, the positive direction of coordinate z is set to be always identical to
the fluid flow direction. In the energy charge process hot fluid flows into the tank from the
top, and thus z = 0 is at the top of the tank. During heat discharge process, cold fluid flows
into the tank from bottom to extract heat from the solid material, and this makes z = 0 at the
bottom of the tank.
H dz
2R
Fig. 5. Schematic of a packed-bed thermal storage system and a control volume for analysis
Several modeling assumptions are made to reasonably simplify the analysis of heat transfer
between the heat transfer fluid and the solid packing material:
1. There is a uniform radial distribution of fluid flow and filler material throughout the
storage tank. This allows the model to be one-dimensional, only in the z direction.
2. Assume that the particles of filler material have only point contact and therefore axial
heat conduction between layers of filler material is negligible.
3. Heat conduction in the axial direction in the fluid is negligible compared to the
convective heat transfer.
4. The lumped heat capacitance method is applied to the transient heat conduction in the
filler material (particles of size of 0.25-5.0 cm in nominal diameter). When this method is
inadequate, due to the large size of solid filler material, a modified lumped capacitance
method will be used, which introduces a modified heat transfer coefficient for the
convection heat transfer between the fluid and the solid filler material.
5. Assume that there is no heat loss from the storage tank to the surroundings. This
assumption applies to both the processes of energy charge and discharge, as well as the
resting time between a charge and a discharge.
Assumption (3) is valid when the Peclet number (=RePr) in the HTF is sufficiently large,
which is satisfied for most thermal energy storage applications (Kays, 2005). Assumption (4)
is valid when the Biot number ( = hLp / ks ) for the thermal storage material is sufficiently
small (Incropera, 2002). If the Biot number is large, a correction to the heat transfer
accounting for the effects of an internal temperature gradient in the filler material will be
considered. Heat loss from a thermal storage tank is inevitable and should be considered.
However, from the design point of view, one needs to first decide on the dimensions of a
storage tank in order to find the heat loss. To compensate for the heat loss from the tank, a
larger volume heat storage tank and a longer heat charge period may be adopted. A simple
way of refining this design is to increase both the heat charge time and tank size with a
factor that is equal to the ratio of heat loss versus the projected heat delivery. To focus on the
main issues, the current work determines the dimensions of a storage tank without
considering heat loss. The assumption of no heat loss to the surroundings also provides a
basis for using the results from a heat charge process as the initial condition of the following
discharge process, and visa versa. By using the end results of one process as the initial
conditions of the following process, multiple cyclic energy charges and discharges in the
actual operation can be simulated relatively easily.
5.2.1 Energy balance in the heat transfer fluid

Based upon the above modeling assumption (1), the cross-sectional area of the tank seen by
the fluid flow is assumed constant at all locations along the axis of the tank, which gives:
a f = επ R 2 (4)
The thermal energy balance of the fluid in the control volume dz is:
∂T f
ρ f επ R 2U( = z − = z + dz ) + hSs (Ts − T f )dz = ρ f C f επ R 2 dz (5)
∂t
where the parameter Ss denotes the heat transfer surface area between the filler material
and the heat transfer fluid per unit length of the tank; U is the actual fluid velocity in the
packed bed:
m
U= (6)
ρ f af
The heat transfer coefficient h in Eq. (5) is for the convection between the heat transfer fluid
and the packing material. It can be different depending on the flow, packing condition of
thermal storage material, fluid properties, and the interaction between packed material and
heat transfer fluid, such as is shown in Fig. 2(a) and 2(b). Detailed discussions of Ss and h
will be presented following the modeling work.
Using the definition of enthalpy change and a Taylor’s series expansion,
= z + dz − = z = C f (∂T f / ∂z)dz , the energy balance equation for the heat transfer fluid becomes:
hSs ∂T f ∂T f
2
(Ts − T f ) = +U (7)
ρ f C f επ R ∂t ∂z
Introducing the following dimensionless variables,
θ f = (T f − TL ) (TH − TL ) (8.a)
θ s = (Ts − TL ) (TH − TL ) (8.b)
z* = z / H (8.c)
t * = t /( H / U ) (8.d)
The dimensionless governing equation for the heat transfer fluid is finally reduced to:
∂θ f ∂θ f 1
+ = (θs − θ f ) (9)
∂t * ∂z * τr
where
2
U ρ f C f επ R Cfm
τr = = (10)
H hSs H h Ss
The boundary condition for Eq. (9) is from the fluid inlet temperature, while the initial
condition is the temperature distribution in a tank before a charge or a discharge starts.
5.2.2 Energy balance in solid thermal storage material

For the energy balance of the filler material in a control volume dz as shown in Fig. 5, it is
understood that the filler material delivers or takes heat to or from the passing fluid at the
cost of a change in the internal energy of the filler. The energy balance equation is:
∂Ts
hSs (Ts − T f )dz = − ρ sC s (1 − ε )π R 2 dz (11)
∂t
By substituting in the dimensionless variables given in Eq. (8), the above governing equation
for filler material is reduced to:
∂θ s H (12)
= − CR (θs − θ f )
∂t *
τr
where
ρfCfε
HCR = (13)
ρsC s (1 − ε )
In the energy charge and discharge processes, the filler material and heat transfer fluid will
have a temperature difference at any local location. Once the fluid comes to rest upon the
completion of a charge or discharge process, the fluid will equilibrate with the local filler
material to reach the same temperature, T final . The energy balance of this situation at a local
location is:
ερ f C f T f − initial + (1 − ε )ρ sC sTs − initial = ερ f C f T final + (1 − ε )ρsC sT final (14)
Here, the initial temperatures of primary thermal storage material and HTF are from the
results of their respective charge or discharge processes. The final temperatures of the
storage material and the fluid are the same after their thermal equilibrium is reached.
According to the assumption of no heat loss from the storage tank, it can be seen that the
equilibrium temperature at the end of one process (charge or discharge) will necessarily be
the initial condition of the next process in the cycle. This connects the discharge and charge
processes so that overall periodic results can be obtained.
The initial temperatures of filler material and fluid in the storage tank should be known.
Also, the inlet fluid temperature is known as a basic boundary condition, with which the
filler temperature at inlet location z=0 can be easily solved mathematically from Eq. (12).
5.2.3 Energy delivery efficiency

With the solution of the governing equations for filler material and HTF, the discharged
fluid temperature from a storage tank can be obtained. With the required heat discharge
period being given as tref , disch arg e , an energy delivery effectiveness can be obtained from
Eq.(2) as discussed before. For convenience of expression, the dimensionless form of the
required time period of energy discharge is defined as:
tref,discharge
Πd = (15)
HU
Similarly, a dimensionless form of the time period of energy charge is defined as:
tcharge
Πc = (16)
HU
Substituting the dimensionless energy discharge period Π d into Eq. (2), we obtain:
Πd
1
∫ θ f ( z =1, t )dt
*
η= * * (17)
Πd 0
The energy discharge efficiency will obviously be affected by how much energy is charged
into the storage tank. Therefore, it should be noted that η is essentially the function of the
following four parameters— Π c / Π d , Π d , τ r , and HCR . By specifying the dimensionless
time period of the discharge process and the mass flow rate, the dimensionless time period
of the energy charge can be determined to achieve the objective value of η , which is always
desired to approach as close as possible to 1.0.
As has been discussed, a longer energy charging time than energy discharging time is
needed in order to achieve an energy delivery effectiveness of approximately 1.0 in a
packed-bed system. In addition, the energy storage capacity of a packed-bed tank must be
larger than that of an ideal thermal storage tank, as expressed in Eq. (3).
5.2.4 Heat transfer area Ss and heat transfer coefficient h in different types of storage
systems
As has been discussed above, the governing equations for the temperatures and energy
exchange between the primary thermal storage material and the HTF are generally the same
for all the thermal storage systems as schematically shown in Fig. 2. However, the heat
transfer coefficients and the heat transfer area between the primary thermal storage material
and HTF for different types of storage systems (for example, in Fig. 2(a) and Fig. 2(b)) can be
significantly different.
If we consider the use of rocks as a filler material, the heat transfer area between rocks and
fluid per unit length of tank was denoted as Ss . Therefore, the unit of Ss is in meters. For
spherical filler materials, Ss is obtained through the following steps:
1. The volume of filler material in a unit length Δz of tank is given as π R 2 Δz(1 − ε ) .
One sphere of rock has a volume of Vsphere = 4π r 3 / 3 , and therefore, in the length of Δz in
the tank, the number of rocks is π R2 Δz(1 − ε ) / Vsphere . The total surface area of rocks is then
determined to be π R 2 Δz(1 − ε ) × 4π r 2 / Vsphere , which becomes 3π R 2 (1 − ε )Δz / r after Vsphere
being substituted in.
2. Finally, the heat transfer area of rocks per unit length of tank is:
Ss = 3π R 2 (1 − ε ) / r (18)
The above discussion considers the actual volume (assuming ε is known) for solid ‘spherical
particles’ in a packed volume. Depending on the packing scheme, the void fraction ε in a
packed bed with spheres of a fixed diameter may range from 0.26 to 0.476 (Conway &
Sloane, 1998). The loosest packaging of spherical rocks in a volume is given by the case
where each sphere (of diameter 2r) is packed into a cube with side lengths of 2r. The densest
packing of spheres causes a void fraction of 0.26, which was due to Kepler’s conjecture
(Hales, 2006). Nevertheless, if the packed bed void fraction ε is known, Eq. (18) should be
used for finding Ss.
The heat transfer coefficient h(W/m2 °C) between the primary thermal storage material
(porous media) and HTF can be found from reference (Nellis & Klein, 2009):
f
mC
h = 0.191 Re −0.278 Pr −2/3 (19)
επ R2
where Re is Reynolds number (equal to 4G rchar / μ f ) for porous media, as defined by Nellis
(Nellis & Kline, 2009). The mass flux of fluid through the porous bed is G (equal to
m /(επ R 2 ) ), and rchar is defined as the characteristic radius of the filler material (Nellis &
Kline, 2009), which is equal to 0.25ε dr /(1 − ε ) for spherical solid filler. Here, dr is the
nominal diameter of a rock, if it is not perfectly spherical.
5.2.5 Modification of lumped capacitance method

The model and equations in section 5.2.1 and 5.2.2 use the lumped capacitance method to
determine the heat transfer inside the filler material. This method actually ignores the
resistance to heat conduction inside the filler material. This will result in the calculated
energy going into, or coming out from, a filler material being higher than that in the actual
physical process. It is known (Incropera, 2002) that when the Biot number of the heat
transfer of a particle is larger than 0.1, the lumped capacitance assumption will result in
increased inaccuracy. In order to correct the lumped capacitance approximation for a
spherical ‘particle’ in fluid, Bradshaw et al. (Bradshaw et al., 1970) and Jeffreson (Jeffreson,
1972) proposed to correct the convective heat transfer coefficient between the solid spherical
‘particle’ and the fluid. The modified heat transfer coefficient is then used in the equations
for the transient temperature in the ‘particle’ from the standard lumped capacitance method.
The modified heat transfer coefficient for a spherical ‘particle’ is:
1
hp = h (20)
1 + Bi / 5
The smaller the Bi number [= h( dr / 2) / ks ] the smaller the correction that is needed. The
modified heat transfer coefficient hp will be used in Eqs. (10), where h should be replaced
by hp when calculating τ r .
Figure 6 shows a comparison of the results of dimensionless energy (normalized by the ideal
maximum energy change of the ‘particle’) going in or out from a single spherical ‘particle’ to
a fluid during a transient heat transfer. The energy is quantified based on the local change of
internal energy in the particle, and the ideal energy is the internal energy change assuming
its temperature completely changed from an initially uniform temperature to the fluid
temperature around the particle. The Bi of the case shown in Figure 6 is 2.54, which is close
t*
Fig. 6. Jeffreson correction to lumped capacitance heat conduction in a sphere (here only in
this figure the dimensionless time is specifically defined as t * = t /[( dr / 2)2 / α s ] )
to the value seen for rocks in typical thermal storage systems. Results from an exact
transient heat conduction solution from reference (Incropera, 2002) are compared to the
results using the lumped capacitance method, as well as the results obtained using the
corrected lumped capacitance method by introducing the modified heat transfer coefficient
given in Eq. (20). The heat transfer coefficient used for the comparison was listed in Table 5.
Results presented in Fig. 6 indicate that the lumped capacitance method has appreciable
discrepancy compared to the exact analytical solution. Results from the Jeffreson correction
model agree with the exact transient heat conduction solution very well. Therefore, the
Jeffreson correction introducing the modified heat transfer coefficient in the heat transfer
model for thermal storage medium and heat transfer fluid is recommended. The Jeffreson
correction allows for the thermocline model to remain in a one-dimensional system yet
increases the accuracy of the results by accounting for the internal thermal gradient in the
packed bed filler material.
Parameters Value Unit

dr 0.04 m
h 355.1 W /(m2 ⋅ K )
ks 2.8 W /(m ⋅ K )
ρs 2630 kg / m3
Cs 775 J /( kg ⋅ K )
α s = ks /( ρsC s ) 1.374 ⋅ 10 −6 m2 / s
Table 5. Properties of the solid particle and heat transfer coefficient for the results in Fig. 6
5.2.6 Application of the model to the storage system as shown in Fig. 2(b)
The same governing equations in sections 5.2.1 and 5.2.2 are also applicable to the case
shown in Fig. 2(b). However, the porosity in the tank, the heat transfer coefficient h and the
heat transfer surface area per unit length, Ss , for the thermal storage tank shown in Fig. 2(b)
will be quite different from that in Fig. 2(a). A cross-section of the storage system in Fig. 2(b),
including the thermal storage filler material (either solid or liquid) and tubes for the heat
transfer fluid, are illustrated in Fig. 7.
An equivalent porosity in the tank is obtained as:
di2
ε =N (21)
D2
where N is the number of heat transfer fluid tubes in the storage tank; di and D are
indicated in Fig. 7.
The heat transfer surface area per unit length in the tank is obtained as:
Ss = Nπ di (22)
The convective heat transfer coefficients inside tubes for either turbulent or laminar fluid
flow can be easily obtained from heat transfer textbooks elsewhere. The Reynolds number
and Nusselt number for the heat transfer coefficient in tubes are obtained assuming a
uniform distribution of total fluid to all the heat transfer fluid tubes.
It is assumed that each heat transfer fluid tube exchanges heat with the storage material in
an equivalent control area of diameter Deq as shown in Fig. 7.
Deq
D=2R
di/do
Fig. 7. Schematic of HTF tube and its surrounding thermal storage material ( Deq is an
equivalent diameter based on the cross sectional area of the container divided by the
number of HTF tubes)
The concept of the Jeffreson correction for the lumped capacitance method using a modified
heat transfer coefficient is still applicable to the heat conduction in the thermal storage
material. However, since the material is not in the form of spherical particles but rather is an
integrated solid or liquid, the modified heat transfer coefficient will be different and must be
found through analysis.
The following analysis takes one HTF tube and its surrounding thermal storage material for
analysis, as shown in Fig. 8. In order to accurately determine the heat going into or out of
the filler material in a transient heat conduction process, a transient heat conduction
problem is analyzed. This will allow for a comparison of results respectively based on an
exact analytical solution, the lumped capacitance method, and the lumped capacitance
method with the adoption of a modified heat transfer coefficient.
Deq
rs
di/do
Fig. 8. Physical model of transient heat conduction in a thermally protected “washer”

The analytical solution is for a transient heat conduction problem in a “washer” as shown in
Fig. 8. The ‘washer’ has its outer surface and the two flat surfaces thermally insulated,
leaving only the inner surface at rs = di / 2 having a convective heat transfer flux.
The mathematical description of this one dimensional problem is
∂T 1 ∂ ∂T
= αs (rs ) (23)
∂t rs ∂rs ∂rs
which is subjected to boundary conditions of:
t = 0 T = TL (24. a)
∂T
t > 0 at rs = di / 2 ; ks = h(T − T f ) (24. b)
∂rs
∂T
t > 0 at rs = Deq / 2 ; =0 (24. c)
∂rs
Solutions for the energy going into the tube based on the standard lumped capacitance
method and solutions of the exact analytical model can be easily obtained. Using results
from the exact analytical solution, a modified heat transfer coefficient hp is introduced to
correct the lumped capacitance method for the ‘washer’.
⎛ 1 ⎞
hp = h ⎜ ⎟ (25)
⎝ 1 + Bi washer / wc ⎠
where Biwasher is a specially defined Biot number [ = h( di / 2) / ks ] for the “washer”; wc is a

coefficient which depends on the ratio of the washer diameters, Deq / di . For a value of
Deq / di = 6 we found wc = 0.83442 through analysis. Using data for typical thermal storage
materials as shown in Table 6, a comparison of results for the dimensionless heat stored in
the washer based respectively on the lumped capacitance method, the exact analytical
solution, and the modified lumped capacitance method, is shown in Fig. 9. The agreement
between the modified lumped capacitance method and the exact analytical solution is very
good.
Parameter Value Unit

di 0.05 m
Deq 0.3 m
h (laminar in tube) 6.88 W /(m2 ⋅ K )
ks 0.57 W /(m ⋅ K )
ρs 1730 kg / m3
Cs 1.47 kJ /( kg ⋅ K )
α s = ks /( ρsC s ) 2.24 ⋅ 10 −7 m2 / s
Table 6. Heat transfer coefficient of HTF-Therminol® VP-1 and properties of primary
thermal storage material (molten salt: Na2S2O3 ⋅ 5H 2O ) for the results in Fig. 9
In the energy storage model discussed in sections 5.2.1 and 5.2.2, axial heat conduction in
the filler material (solid spherical particles in magnitude of 1.0-10 cm in diameter) is
neglected. For integrated solid or liquid thermal storage materials, it is necessary to
investigate whether axial heat conduction will be significant. When axial heat conduction in
the thermal storage material is considered, the energy balance equation for the thermal
storage material is:
∂ 2Ts ∂T
hSs (Ts − T f )dz − (1 − ε )π R 2 ks 2
dz = − ρsC s (1 − ε )π R 2 dz s (26)
∂z ∂t
Its dimensionless form is:
∂θ s H ks ∂ 2θ s
= − CR (θ s − θ f ) + (27)
∂t *
τr ρsC s HU ∂z*2
Fig. 9. Comparison of dimensionless energy storage in the ‘washer’, normalized by the ideal
maximum energy change in the ‘washer’, due to different methods of solution
( Biwasher = h( di / 2) / ks =3.0; Deq / di = 6.0 ; t * = t /[( di / 2)2 / α s ] ; wc = 0.83442)
The parameter cluster of ks /( ρ sC s HU ) is a dimensionless term. If it is sufficiently large, the
axial conduction term in Eq. (27) may not be dropped off. A basic effect of significant axial
heat conduction is that it will destroy the thermocline effect—a temperature gradient with
hot material being on top of cold. This can lower the thermal storage performance in
general. Therefore, to take into account the axial heat conduction effect, a similar correction
via the introduction of another factor to the modified heat transfer coefficient is proposed.
This results in a new modified heat transfer coefficient of:
⎛ ⎞
⎛ 1 ⎞⎜ 1 ⎟
hp = h ⎜ ⎟⎜ ks
⎟ (28)
⎝ 1 + Biwasher / wc ⎠⎜ 1 + ⎟
⎜ ρ ⎟
⎝ s s HU
C ⎠
For most thermal storage materials, such as rocks, molten salts, concrete, soil, and sands, the
value of ks /( ρ sC s HU ) is very small (in the order of 1 × 10 −6 ); while other terms in Eq.(27)
are in the order of 1.0. Therefore, the axial heat conduction effect in the thermal storage
material in Eq. (27) is negligible.
5.2.7 Application of the model to the storage system with PCM

For thermal storage with phase-change involved, the PCM can be enclosed in capsules to
form a packed bed as shown in Fig. 2(a), or simply put in a storage tank that has heat
transfer tubes inside as shown in Fig. 2(b). The governing equations discussed above are still
applicable to the heat transfer at locations where either phase change has not yet occurred or
has already been completed. However, at locations undergoing phase change, the energy
equations must account for the melting or solidification process (Halawa & Saman, 2011;
Wu et al., 2011). The key feature in a melting or solidification process is that the temperature
of the material stays constant.
Considering the energy balance for the thermal storage material:
dΦ
hSs (Tm − T f ) = −Γρs (1 − ε )π R 2 (29)
dt
where Γ is the fusion energy of the material, and Φ is the ratio of the liquid mass to the
total mass in the control volume of dz. For melting, Φ increases from 0 to 1.0, while for
solidification it decreases from 1.0 to 0.
Considering the invariant of the temperature of the material during a phase change process,
the energy balance equation for HTF is:
hSs ∂T f ∂T f
(Tm − T f ) = +U (30)
2
ρ f C f επ R ∂t ∂z
Equation (29) and (30) can be reduced to dimensionless equations by introducing the same
group of dimensionless parameters:
1 ∂θ f ∂θ f
(θ m − θ f ) = + (31)
τr ∂t *
∂z *
HCR dΦ (32)
− ψ (θm − θ f ) =
τr dt *
where a new dimensionless parameter ψ = (TH − TL )C s / Γ is introduced. Since the phase
change temperature is known, Eqs. (31) and (32) can be solved separately.
5.3 Numerical methods and solution to governing equations

5.3.1 Solution for the case of no phase change
A number of analyses and solutions to the heat transfer governing equations of a working
fluid flowing through a packed-bed have been presented in the past (Schumann, 1929;
Shitzer & Levy, 1983; McMahan, 2006; Beasley, 1984; Zarty & Juddaimi, 1987). As the
pioneering work, Schumann (Schumann, 1929) presented a set of equations governing the
energy conservation of fluid flow through porous media. Schumann’s equations have been
widely adopted in the analysis of thermocline heat storage utilizing solid filler material
inside a tank. His analysis and solutions were for the special case where there is a fixed fluid
temperature at the inlet to the storage system. In most solar thermal storage applications this
may not be the actual situation. To overcome this limitation, Shitzer and Levy (Shitzer &
Levy, 1983) employed Duhamel’s theorem on the basis of Schumann’s solution to consider a
transient inlet fluid temperature to the storage system. The analysis of Schumann, and
Shitzer and Levy, however, still carry with them some limitations. Their method does not
consider a non-uniform initial temperature distribution. For a heat storage system,
particularly in a solar thermal power plant, heat charge and discharge are cycled daily. The
initial temperature field of a heat charge process is dictated by the most recently completed
heat discharge process, and vice versa. Therefore, non-uniform and nonlinear temperature
distribution is typical for both charge and discharge processes. To consider a non-uniform
initial temperature distribution and varying fluid temperature at the inlet in a heat storage
system, numerical methods have been deployed by researchers in the past.
To avoid the long mathematical analysis necessary in analytical solutions, numerical
methods used to solve the Schumann equations were discussed in the literature by
McMahan (McMahan, 2006, 2007), and Pacheco et al. (Pacheco et al., 2002), and
demonstrated in the TRNSYS software developed by Kolb and Hassani (Kolb & Hassani,
2006). Based on the regular finite-difference method, McMahan provided both explicit and
implicit discretized equations for the Schumann equations. Whereas the explicit solution
method had serious stability issues, the implicit solution method encountered an additional
computational overhead, thus requiring a dramatic amount of computation time. The
solution for the complete power plant with thermocline storage provided by the TRNSYS
model in Kolb’s work (Kolb & Hassani, 2006) cites the short time step requirement for the
differential equations of the thermocline as one major source of computer time
consumption. To overcome the problems encountered in the explicit and implicit methods,
McMahan et al. also proposed an infinite-NTU method (McMahan, 2006, 2007). This model
however is limited to the case in which the heat transfer of the fluid compared to the heat
storage in fluid is extremely large.
The present study has approached the governing equations using a different numerical
method (Van Lew et al., 2011). The governing equations have been reduced to
dimensionless forms which allow for a universal application of the solution. The
dimensionless hyperbolic type equations are solved numerically by the method of
characteristics. This numerical method overcomes the numerical difficulties encountered in
McMahan’s work—explicit, implicit, and the restriction on infinite-NTU method (McMahan,
2006, 2007). The current model yields a direct solution to the discretized equations (with no
iterative computation needed) and completely eliminates any computational overhead. A
grid-independent solution is obtained at a small number of nodes. The method of
characteristics and the present numerical solution has proven to be a fast, efficient, and
accurate algorithm for the Schumann equations.
The non-dimensional energy balance equations for the heat transfer fluid and filler material
can be solved numerically along the characteristics (Courant & Hilbert, 1962; Polyanin, 2002;
Ferziger, 1998). Equation (9) can be reduced along the characteristic t * = z * so that:
Dθ f 1
= ( θr − θ f ) (33)
Dt*
τr
Separating and integrating along the characteristic, the equation becomes:
1 *
∫ dθ f = ∫ τ ( θ r − θ f )dt (34)
r
Similarly, Eq.(12) for the energy balance for the filler material is reposed along the
characteristic z * =constant so that:
dθ r H
= − CR ( θ r − θ f ) (35)
dt*
τr
The solution for Eq. (35) is very similar to that for Eq. (33) but with the additional factor of
H CR . The term H CR is simply a fractional ratio of fluid heat capacitance to filler heat
capacitance. Therefore, the equation for the solution of θ r will react with a dampened speed
when compared to θ f , as the filler material must have the capacity to store the energy
being delivered to it, or vice versa. Finally, separating and integrating along the characteristic
for Eq.(35) results in:
CR H *
∫ dθ r = ∫ − τ ( θ r − θ f )dt (36)
r
There are now two characteristic equations bound to intersections of time and space. A
discretized grid of points, laid over the time-space dimensions will have nodes at these
intersecting points. A diagram of these points in a matrix is shown in Fig. 10. In space, there
* *
are i = 1, 2, . . . ,M nodes broken up into step sizes of Δz to span all of z . Similarly, in
time, there are j = 1, 2, . . . ,N nodes broken up into time-steps of Δt to span all of t* .
*
Looking at a grid of the ϑ nodes, a clear picture of the solution can arise. To demonstrate a
*
calculation of the solution we can look at a specific point in time, along z where there are
two points, ϑ1,1 and ϑ 2 ,1 . These two points are the starting points of their respective
characteristic waves described by Eq. (33) and (36). After the time Δt * there is a third point
ϑ 2 ,2 which has been reached by both wave equations. Therefore, Eq. (34) can be integrated
numerically as:
ϑ2 , 2 ϑ2 , 2 1 *
∫ϑ1,1 dθ f = ∫ϑ1,1 τ ( θ r − θ f )dt (37)
r
The numerical integration of the right hand side is performed via the trapezoidal rule and
the solution is:
1 ⎛⎜ θ r2 ,2 + θ r1 ,1 θ f 2 ,2 + θ f1 ,1 ⎞ *
⎟Δt
θ f 2 ,2 − θ f1 ,1 = − (38)
τ r ⎜⎝ 2 2 ⎟
⎠
where θ f1 ,1 is the value of θ f at ϑ1,1 , and θ f 2 ,2 is the value of θ f at ϑ 2 ,2 , and similarly so
for θ r .
The integration for Eq. (36) along z * =constant is:
ϑ2 , 2 ϑ2 , 2 H CR
∫ϑ 2 ,1
dθ r = ∫ [ −
ϑ2 ,1 τr
( θ r − θ f )] dt * (39)
The numerical integration of the right hand side is also performed via the trapezoidal rule
and the solution is:
H CR ⎛⎜ θ r2 ,2 + θ r2 ,1 θ f 2 ,2 + θ f 2 ,1 ⎞ *
⎟Δt
θ r2 ,2 − θ r2 ,1 = − − (40)
τ r ⎜⎝ 2 2 ⎟
⎠
Equations (38) and (40) can be reposed as a system of algebraic equations for two
unknowns, θ f 2 ,2 and θ r2 ,2 , while θ f and θ r at grid points ϑ1,1 and ϑ 2 ,1 are known.
⎡ Δt * ⎤ ⎡θ f 2 ,2 ⎤ ⎡
Δt * ⎛ Δt * ⎞⎟ Δt * ⎤
⎢ 1+ − ⎥⎢ ⎥ ⎢θ f1 ,1 ⎜⎜ 1 − ⎟
+ θ r1 ,1 ⎥
2τ r 2τ r ⎥=⎢ ⎝ 2τ r ⎠ 2τ r ⎥
⎢ ⎥⎢ (41)
⎢ H CR Δt * H Δt * ⎥⎢ ⎥ ⎢ ⎛ H CR Δt ⎞
* ⎛ * ⎞⎥
⎢− 1 + CR ⎥ ⎢θ ⎥ ⎢θ f 2 ,1 ⎜⎜ ⎟ + θ r ⎜ 1 − H CR Δt ⎟⎥
2τ r 2τ r ⎦ ⎣⎢ r2 ,2 ⎦⎥ ⎢ ⎟ 2 ,1 ⎜ ⎟⎥
⎣ ⎣ ⎝ 2τ r ⎠ ⎝ 2τ r ⎠⎦
Cramer’s rule (Ferziger, 1998) can be applied to obtain the solution efficiently. It is
important to note that all coefficients/terms in Eq.(41) are independent of z* , t* , θ f , and
θ r , thus they can be evaluated once for all. Therefore, the numerical computation takes a
minimum of computing time, and is much more efficient than the method applied in
references (McMahan, 2006, 2007).
Fig. 10. Diagram of the solution matrix arising from the method of characteristics
From the grid matrix in Fig.10 it is seen that the temperatures of the filler and fluid at grids
ϑi ,1 are the initial conditions. The temperatures of the fluid and filler at grid ϑ1,1 are the
inlet conditions which vary with time. The inlet temperature for the fluid versus time is
given. The filler temperature (as a function of time) at the inlet can be easily obtained using
Eq.(12), for which the inlet fluid temperature is known. Now, as the conditions at ϑ1,1 , ϑ1,2 ,
and ϑ 2 ,1 are known, the temperatures of the rocks and fluid at ϑ 2 ,2 will be easily calculated
from Eq.(41).
Extending the above sample calculation to all points in the ϑ grid of time and space will
give the entire matrix of solutions in time and space for both the rocks and fluid. While the
march of Δz * steps is limited to z* = 1 the march of time Δt * has no limitation.
The above numerical integrations used the trapezoidal rule; the error of such an
implementation is not straightforwardly analyzed but the formal accuracy is on the order of
O( Δt *2 ) for functions (Ferziger, 1998) such as those solved in this study.
5.3.2 Solutions for the case with phase change

For the governing equations of the phase change case, the adopted convention of having the
z-direction coordinate always follow the flow direction is preserved, such that for heat
charging, z=0 is for the top of a tank, and for heat discharging, z=0 is for the bottom of a
tank. The two governing equations (Eq. (31) and Eq.(32)) for the phase change process can
be discretized using finite control volume methodology:
* * * * * * *
1 *
t*
θ tf (+Δ
i)
t
− θ tf ( i ) θ tf (+Δ
i)
t
− θ tf (+Δ t
i − 1)
(θ m − θ tf (+Δ
i) )= + (42)
τr Δt * Δz*
*
+Δt * *
HCR * * Φti − Φ ti
− ψ (θm − θ tf (+Δt
i) ) = (43)
τr Δt *
* *
From Eq.(42) the fluid temperature θ tf (+Δ
i)
t
can be solved, which is then used in Eq. (43) to
t * +Δt *
solve for the fusion ratio Φ i .
The procedures for finding the solution of phase change problem are as follows:
1. Solve the non-phase-change governing equation analytically using Eq.(12) for the phase
change material for the inlet point.
2. Monitor the temperature at each time step as given by Eq.(12), and see if the
temperature at a time step is greater than the fusion temperature, if yes, the solution for
that and subsequent time steps are to be solved using the phase change equation (Eq.
(43)
3. For each time step solved using Eq (43), monitor the fusion ratio, Φ ; when it becomes
larger than 1.0 then the solution for that and subsequent time steps are to be solved
using the non-phase-change governing equation (Eq.(12)) for the remainder of the
required time.
4. March a spatial step forward and repeat all of the above steps. However, now in part (1)
of this procedure, Eq.(41) must be used to solve the temperatures of both the fluid and
PCM for time steps before phase change starts; and also in part (3) of this procedure
Eq.(41) should be used to solve the temperatures of the fluid and PCM for steps after
the phase change is over. The repetition of parts (1) to (3) of this procedure is to be
continued until all the spatial steps are covered.
6. Results from simulations and experimental tests

6.1 Numerical results for the temperature variation in a packed bed
The first analysis of the storage system was done on a single tank configuration of a chosen
geometry, using a filler and fluid with given thermodynamic properties. The advantage of
having the governing equations reduced to their dimensionless form is that by finding the
values of two dimensionless parameters ( τ r and HCR ) all the necessary information about
the problem is known. The properties of the fluid and filler rocks, as well as the tank
dimensions, which determined τ r and HCR for the example problem, are summarized in
Table 7.
The numerical computation started from a discharge process assuming initial conditions of
an ideally charged tank with the fluid and rocks both having the same high temperature
throughout the entire tank, i.e. θ f = θs = 1 . After the heat discharge, the temperature
distribution in the tank is taken as the initial condition of the following charge process. The
discharge and charge time were each set to 4 hours. The fluid mass flow rate was
determined such that an empty (no filler) tank was sure to be filled by the fluid in 4 hours.
With the current configuration, after five discharge and charge cycles the results of all
subsequent discharge processes were identical—likewise for the charge processes. It is
therefore assumed that the solution is then independent of the first-initial condition. The
data presented in the following portions of this section are the results from the cyclic
discharge and charge processes after 5 cycles.
ε τr HCR H R t
0.25 0.0152 0.3051 14.6 m 7.3 m
Fluid (Therminol® VP-1 ) properties:
TH=395 °C; TL=310 °C; ρ f =753.75 kg/m3; Cf=2474.5 J/(kg K);
kf =0.086W/(m K); =128.74 kg/s;
m μ f =1.8 ×10 −4 Pa ⋅ s ;
Filler material (granite rocks) properties:
ρ s = 2630 kg/m3; Cs=775 J/(kg K); ks=2.8 W/(m K); dr= 0.04 m;
Table 7. Dimensions and parameters of a thermocline tank (Van Lew et al., 2011)
Fig. 11. Dimensionless fluid temperature profile in the tank for every 0.5 hours
Shown in Fig. 11 are the temperature profiles in the tank during a discharge process, in
which cold fluid enters into the tank from bottom of the tank. The location of z * = 0 is at the
bottom of a tank for a discharge process. The temperature profile evolves as discharging
proceeds, showing the heat wave propagation and the high temperature fluid moving out of
the storage tank. The fluid temperature at the exit ( z * = 1 ) of the tank gradually decreases
after 3 hours of discharge. At the end of the discharge process, the temperature distribution
along the tank is shown in Fig. 12. At this time the fluid and rock temperatures, θ f and
θs respectively ( θ s is denoted by θr when the filler material is rock), in the region with z *
below 0.7 are almost zero, which means that the heat in the rocks in this region has been
completely extracted by the passing fluid. In the region from z * = 0.7 to z * = 1.0 the
temperature of the fluid and rock gradually becomes higher, which indicates that some heat
has remained in the tank.
Fig. 12. Dimensionless temperature distribution in the tank after time t* = 4 of discharge
(Here θ r is used to denote θs , as rocks are used as the storage material in the example).
Fig. 13. Dimensionless temperature distribution in the tank after time t* = 4 of charge
(Here θ r is used to denote θs , as rocks are used as the storage material in the example)
A heat charge process exhibits a similar heat wave propagation scenario. The temperature
for the filler and fluid along the flow direction is shown in Fig. 13 after a 4 hour charging
process. During a charge process, fluid flows into the tank from the top, where z * is set as
zero. It is seen that for the bottom region ( z * from 0.7 to 1.0) the temperatures of the fluid
and rocks decrease significantly. A slight temperature difference between heat transfer fluid
and rocks also exists in this region.
Fig. 14. Dimensionless temperature histories of the exit fluid at z* = 1 for charge and
discharge processes
The next plots of interest are the variation of θ f at z * = 1 as dimensionless time progresses
for a charging or discharging process. Figure 14 shows the behavior of θ f at the outlet
during both charge and discharge cycles. For the charge cycle, θ f begins to increase when
all of the initially cold fluid has been ejected from the thermocline tank. For the present
thermocline tank, the fluid that first entered the tank at the start of the cycle has moved
completely through the tank at t * = 1 , which also indicates that the initially-existing cold
fluid of the tank has been ejected from the tank. Similarly, during the discharge cycle, after
the initially-existing hot fluid in the tank has been ejected, the cold fluid that first entered
the tank from the bottom at the start of the cycle has moved completely through the tank at
t * = 1 . At t * = 2.5 , or t=2.5 hours, the fluid temperature θ f starts to drop. This is because
the energy from the rock bed has been significantly depleted and incoming cold fluid no
longer can be heated to θ f = 1 by the time it exits the storage tank.
The above numerical results agree with the expected scenario as described in section 4. To
validate the above numerical method, analytical solutions were obtained using a Laplace
Transform method by the current authors (Karaki, et al, 2010), which were only possible for
cases with a constant inlet fluid temperature and a simple initial temperature profile. Results
compared in Fig. 15 are obtained under the same operational conditions—starting from a
fully charged initial state and run for 5 iterations of cyclic discharge and charge processes.
The fluid temperature distribution along the tank ( z* =0 for bottom of the tank) from numerical
results agrees with the analytical results very well. This comparison essentially proves the
effectiveness and reliability of the numerical method developed in the present study.
Fig. 15. Comparison of numerical and analytical results of the temperature distribution in
the tank after time t* = 4 of a discharge (Here θr is used to denote θ s , as rocks are used as
the storage material in the example)
Based on the results shown in Fig. 15, the temperature distribution along z * at the end of a
charge is nonlinear. This distribution will be the initial condition for the next discharge
cycle. Similarly a discharge process will result in a nonlinear temperature distribution,
which will be the initial condition for the next charge. It is evident that the analytical
solutions developed by Schumann (Schumann, 1929) could not handle this type of situation.
Fig. 16. Comparison of dimensionless temperature distributions in the tank after time t* = 4
of discharge for different numbers of discretized nodes
Another special comparison was made to demonstrate the efficiency of the method of
characteristics at solving the dimensionless form of the governing equations. Shown in Fig.
16 are the temperature profiles at t * = 4 obtained by using different numbers of nodes (20,
100, and 1000) for z * . The high level of accuracy of the current numerical method, even with
only 20 nodes, demonstrates the accuracy and stability of the method with minimal
computing time.
6.2 Comparison of modeling results with experimental data from literature

6.2.1 Temperature variations in charge processes
The authors have conducted experimental tests (Karaki at al., 2011). The test conditions for a
given heat charge process are listed in Table 8, which also shows the dimensionless
parameters. As shown in Fig. 17, at the initial time the thermocline tank has a uniform
temperature equal to room temperature. The temperature readings from the thermocouples
at the top of the tank provide the inlet fluid temperatures in a charge process.
Tank Length 0.65(m) Initial temperature 21.9 (oC)

Tank inner diameter 0.241 (m) High temperature 79.82 (oC)
Rock nominal diameter 0.01 (m) Low temperature 21.9 (oC)
Oil flow rate 1.0 (Liter/min.) Porosity 0.324
Density of rocks 2632.8 kg/m3 Π c = tc /( H / U ) 7.2511
Charging period tc 70.08 (min.) τr 0.1262
Rock heat capacity 790 (J/(kg K)) HCR 0.4068
Time (s) 0 856 1713 2570 3427 4284
Dimensionless time t* 0 1.45 2.9 4.35 5.80 7.25
Table 8. Conditions of a heat charging test (HTF is Xceltherm 600 by Radco Industries)
90
80
70
TC-1
TC-2
60
Temperature ( C)
TC-3
o
TC-4
50
TC-5
TC-6
40
TC-7
TC-8
30 TC-9
TC-10
20 TC-11
TC-12
10 TC-13
TC-14
0
0 1000 2000 3000 4000 5000
Time (second)
Fig. 17. Temperatures in the center of the tank along the height of 65 cm for a charging
process (Thermocouples were set every 5cm)
1
t*=7.25
0.9
0.8
t*=5.80
0.7
0.6
t*=2.90 t*=4.35
θavg
0.5
θ f
0.4
0.3
0.2 t*=1.45
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z*∗
z
Fig. 18. The temperature distribution along the height in the tank at different time points
(Solid lines are from simulation results and dashed lines are from experimental tests)
Based on the temperature measurements, the temperature in the tank increases gradually
and at the end of the charging process, the temperatures at all the locations are sufficiently
high. The temperature distribution along the height in the tank at different times is shown in
Fig. 18. Obviously at the end of the charge, the temperature at the top of the tank (at z*=0) is
high. Using the initial temperature distribution and the inlet fluid temperature, together
with the properties listed in Table 8, numerical simulation results were obtained and are
shown in Fig. 18 for the average temperature of the fluid and rocks. The real time and the
dimensionless time are listed in Table 8. The agreement between the experimental data and
the modeling simulation is very satisfactory.
The thermal storage performance test results of a thermocline tank reported in the literature
(Pacheco et al., 2002) was also referenced to validate the current modeling work. The
experimental tests used eutectic molten salt (NaNO3-KNO3, 50% by 50%) as the heat transfer
fluid and quartzite rocks and silica sands as the filler material. Thermocouples in the test
apparatus were imbedded in the packed-bed. Temperatures in the tank at different height
locations were recorded during a two-hour heat discharge process after the tank was
charged for the same length of time. The storage tank dimensions, packed-bed porosity, and
properties of the fluid and filler material are listed in Table 9.
Using the modeling of section 5, the heat charge followed by heat discharge was simulated.
In Fig. 19, the predicted temperatures at several height locations of the tank at different time
instances during the discharge process were compared to the experimental data reported
(Pacheco et al., 2002). The trend of temperature curves from the modeling prediction and the
experiment is quite consistent. Considering the uncertainties in the experimental test and
the properties of materials considered, the agreement between the experimental data and
the modeling prediction is quite satisfactory. This comparasion firmly validates the current
modeling and its numerical solution method.
ε τr HCR H R t
0.22 0.0041 0.2733 6.1m 1.5m 2 hr
Molten salt NaNO3-KNO3 properties:
TH=396 °C; TL=290 °C; ρ f =1733 kg/m3; Cf=1550 J/(kg K);
kf =0.57 W/(m K); =7.0 kg/s;
m μ f = 0.0021 Pa ⋅ s ;
Quartzite rocks/sands mixture properties:
ρ s = 2640 kg/m3; Cs=1050 J/(kg K); ks=2.5 W/(m K); dr= 0.015 m;
Table 9. Dimensions and parameters of a thermocline tank for the test in literature (Pacheco
et al., 2002)
Tank Length measured 0.65(m) Initial temperature 129.0 (oC)

Tank inner diameter 0.241 (m) High temperature 129.0 (oC)
Rock nominal diameter 0.01 (m) Low temperature 56.0 (oC)
Oil flow rate 1.96 (Liter/min.) Porosity 0.324
Density of rocks 2632.8 kg/m3 Π d = td /( H / U ) 8.0596
Charging period tc 39.74 (min) τr 0.1044
Rock heat capacity 790 (J/(kg K)) HCR 0.4210
Time (s) 0 476.2 952.5 1428.7 1905.0 2381.2
Dimensionless time t* 0 1.61192 3.22384 4.83576 6.44768 8.0596
Table 10. Conditions of a discharge test
1
t*=0 Modeling
0.9 t*=0.767 Modeling
t*=1.534 Modeling
0.8
t*=2.300 Modeling
0.7 t*=3.067 Modeling

t*=0.0 Test
0.6 t*=0.767 Test
θ avg
t*=1.534 Test
0.5
t*=2.300 Test
0.4 t*=3.067 Test
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z*=z/H
Fig. 19. Comparison of modeling predicted results with experimental data from reference
(Pacheco et al., 2002)
6.2.2 Temperature variations in a discharge process

A heat discharge experiment was conducted under the conditions listed in Table 10, which
also includes the dimensionless parameters. Shown in Fig. 20 is the temperature variation
versus time at different locations in the tank along the tank height. The thermocouples at the
top (z*=1.0) of the tank measure the temperature of the discharged fluid. The degradation of
discharged fluid temperature is clearly shown in the figure.
140
TC-1 (bottom-in)
120 TC-2
TC-3
100 TC-4
TC-5
Temperature (oC)
TC-6
80
TC-7
TC-8
60 TC-9
TC-10
40 TC-11
TC-12
TC-13
20
TC-14 (top-out)
0
0 500 1000 1500 2000 2500 3000
Time (second)
Fig. 20. Temperatures in the center of the tank along the height of 65 cm for a discharging
process (Thermocouples were set every 5cm)
0.9 t*=0.0
0.8
t*=1.61
0.7
0.6
t*=3.22
0.5
θf
θavg
0.4
t*=4.84
0.3
0.2
t*=6.44
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z*
t*=8.06
Fig. 21. The temperature distribution along the height in the tank at different time points
(Solid lines are from simulation results and dashed lines are from experimental tests)
The temperature distribution along the height in the tank at different times is shown in Fig.
21. At the end of the discharge, the temperature on top of the tank (at z*=1) becomes low,
which means that stored energy has been discharged. Using the measured initial
temperature distribution and the inlet fluid temperature (at z*=1) together with the
properties listed in Table 10, numerical simulation results were obtained and are compared
with the test results in Fig. 21. The real time and the dimensionless time are listed in Table
10. Again, the agreement between the experimental data and the modeling simulation is
satisfactory.
6.3 Correlation of energy delivery effectiveness to dimensionless parameters

Based on the above discussion and the dimensionless governing equations obtained in
section 5.2.3, the energy delivery effectiveness, η , is a function of four dimensionless
parameters, Π c / Π d , Π d , τ r , and HCR . Solutions of the dimensionless governing
equations for energy charge and discharge allow us to develop a database so that a series of
charts and diagrams for η = f (Π c / Π d , Π d , τ r , HCR ) may be prepared for reference by
engineers in the design of thermocline storage tanks. Illustrated in Fig. 22 is a configuration
of a group of database charts for a given Π d , in which multiple graphs, each with a specific
τ r , may be provided. In each graph, multiple curves, each with a given HCR, for the energy
storage effectiveness η versus Π c / Π d are provided. To build a large database, more
graphs in the same configuration can be provided covering a large range of Π d values.
1.0 τr1
Asymptote
H CRn
HCR τr2
η 0.5 HCR1 τr3
τr4
τr5
0.0
0.5 1.0 2.0 τri
Πc/Πd
Fig. 22. Configuration of a group of database charts including multiple graphs of η versus
Π c / Π d at a fixed Π d
It is easy to understand that under the same mass flow rate and a desired Π d , the energy
delivery effectiveness will increase with the increase of Π c / Π d . The asymptote of the
energy delivery effectiveness is due to an ideal thermal storage system, in which η can be
exactly 1.0 if the energy charge period is equal to or larger than that of the discharge
( Π c / Π d ≥ 1.0 ). For any non-ideal thermocline system η can only approach 1.0.
Figure 23 shows four charts of η versus Π c / Π d at Π d =4.0. Each chart is for a specific τ r
with multiple curves for different HCR . All the data of energy delivery effectiveness were
obtained based on several cyclic operations of the energy charge and discharge, and the
results are consequently independent of the number of cycles. More charts with wide range
of Π d , τ r , and HCR may be developed. However, in actual design practice, the ranges of
Π d , τ r , and HCR are not very wide and specific charts can be easily prepared.
(a) τ r = 0.01
(b) τ r = 0.04
(c) τ r = 0.1
(d) τ r = 0.2
Fig. 23. Multiple graphs from modeling results for energy storage effectiveness versus
Π c / Π d at Π d =4.0
Observing the above four graphs one can easily draw the following conclusions:
1. The energy delivery effectiveness never reaches 1.0 if Π c / Π d < 1.0 . This proves that
only for an ideal thermocline storage tank can η = 1.0 at Π c / Π d = 1.0 .
2. With the decrease of τ r , η will increase. For example, at a ratio of Π c / Π d = 1.5 and
HCR = 0.25 , the energy delivery effectiveness approaches 1.0 when τ r changes from 0.2
to 0.01. This is because a decrease of τ r is due to an increase in the volume of the
storage tank.
3. It is understood that a small HCR corresponds to the case where ( ρC )s is relatively large
when compared to ( ρC ) f , and therefore the energy storage capability is improved, and
η can approach 1.0 easier. On the other hand, when the void fraction in a packed bed
approaches 1.0, it will make HCR → ∞ , and the thermal storage effectiveness can
approach that of an ideal case. However, in most practical applications, a low void
fraction in a thermocline tank is required for the purpose of using less heat transfer
fluid, and therefore a smaller HCR value is practical and preferable.
4. For cases where η could never approach 1.0, even at large Π c / Π d values, it is
obviously attributable to the fact that the storage tank is too small, and reselection of a
larger storage tank is needed.
Designers for a thermal storage system often need to calibrate or confirm that a given
storage tank can satisfy an energy delivery requirement. Under such a circumstance, the
dimensions of the storage tank and the power plant operational conditions are known,
which means the values of τ r , Π d , and HCR are essentially given. One can easily check
whether over a range of values of Π c / Π d the energy delivery effectiveness η can
approach 1.0.
7. Procedures of sizing and design of a thermal storage system

The required operational conditions of the power plant dictate the size of the thermocline
storage tank. The relevant operational conditions include: electrical power, thermal efficiency
of the power plant, the extended period of operation based on stored thermal energy, the
required high temperature of the heat transfer fluid from the storage tank, and the low
temperature of the fluid returned from the power plant, the specific choice of heat transfer
fluid and thermal storage material, as well as the packing porosity in a thermocline tank.
The design analysis using the general charts provided in the present study will include the
following steps:
1. Select a minimum required volume for a thermocline tank using Eqs. (1) and (3).
2. Choose a radius, R, and the corresponding height, H, from the minimum volume
decided in step (1). Using these dimensions, the parameters— Π d , τ r , HCR for a
thermocline tank with filler material, can be evaluated, where Π d is determined based
on the required operational time.
3. Look up the design charts (such as those in Fig. 23) and see if an energy delivery
effectiveness of 1.0 can be achieved. Often the energy delivery effectiveness will not
approach 1.0 for the first trial design. This is because the first trial uses a minimum
volume. However, with the results from the first trial one can predict the required
height or volume of the tank necessary to decrease τ r and Π d in the same proportion.
A couple of trail iterations may be needed to eventually satisfy the criterion of η
closing to 1.0.
If the energy delivery effectiveness from step (1) cannot approach 1.0 even if a large Π c / Π d
is chosen, one actually has two ways to improve the effectiveness during the second trial.
These are to decrease HCR , or decrease both τ r and Π d in the same proportion. However,
HCR is determined by properties of the fluid and filler material, which has very limited
options, and therefore a decrease of τ r and Π d is more practical. The decrease of τ r can be
done by an increase of the height of a storage tank. This means that to achieve an
effectiveness of 1.0 one has to increase the size of the storage tank. When the height of tank
is increased, Π d decreases accordingly since it depends on the height of the tank.
Occasionally, a calibration analysis requires a designer to find a proper time period of
energy charge that can satisfy the needed operation time of a power plant. The known
parameters will be the tank volume, τ r , as well as HCR at a required operation period of Π d .
The first step of the calibration should be the examination of the criterion given in Eq. (3),
from which a minimum tank volume can be chosen. If the minimum tank volume is
satisfied, the second step of calibration will be to find a proper Π c / Π d that can make the
energy delivery effectiveness approach 1.0. Graphs including curves at the required HCR
and the given τ r and Π d must be looked up. Conclusions can be easily made depending on
whether the energy delivery effectiveness can approach 1.0 for a particular value of Π c / Π d .
Two practical examples of thermocline thermal storage tank design are provided to help
industrial designers practice the design procedures proposed in this work. Readers can
easily repeat the design procedures while using their specific material parameters and
operational conditions.
7.1 Design example 1—a system as shown in Fig. 2(a)

A solar thermal power plant has 1.0 MW electrical power output at a thermal efficiency of
20%. The heat transfer fluid used in the solar field is Therminol® VP-1. The power plant
requires high and low fluid temperatures of 390 °C and 310 °C, respectively. River rocks are
used as the filler material and the void fraction of packed rocks in the tank is 0.33. The
required time period of energy discharge is 4 hours, the storage tank diameter is chosen to
be 8 m. The rock diameter is 4 cm.
The solution is discussed as follows:
1. Making use of Eqs. (1) to (3) and the above given details on the power plant, as well as
the properties of Therminol® VP-1, we can find a necessary mass flow rate of 25.34
kg/m3 and an ideal tank height of 9.59 m. The minimum volume of the storage tank can
be determined from Eq. (3). Here, the ideal volume is used in the first trial of the design.
Using Eqs. (19) and (20) we find the modified heat transfer coefficient to be 32.05
W/(m2 K). With this information, the values of HCR , Π d , and τ r are found to be 0.451,
3.03, and 0.0227, respectively. Given in Fig. 24 is a chart for Π d =3 and τ r =0.0227 at
various values of HCR and Π c / Π d . It is seen that on the curve of HCR =0.45, there is
no time ratio Π c / Π d that allows the energy delivery effectiveness to be close to 1.0.
Therefore, the ideal volume chosen will not satisfy the energy storage need.
2. One option to provide the ability to store and deliver more energy and approach an
effectiveness of 1.0 is to increase the height of the storage tank. When the height is
increased to 12 m, the values of Π d and τ r changed to 2.42 and 0.0181, respectively.
Figure 25 gives the chart for Π d =2.42 and τ r =0.0181. It is seen on the curve of HCR =
0.45 that at the time period ratio, Π c / Π d =1.2, the energy delivery effectiveness
approaches 0.99. This should be an acceptable design.
In this example, compared to an ideal thermal storage tank, the rock-packed-bed
thermocline tank uses about 40.0% of the heat transfer fluid. To avoid the temperature
degradation of the heat transfer fluid in a required discharge time period, a 20% longer
charging time than discharging time is always applied in every charge and discharge cycle.
Fig. 24. Energy delivery effectiveness versus Π c / Π d at Π d =3 and τ r =0.0227
Fig. 25. Energy delivery effectiveness versus Π c / Π d at Π d =2.42 and τ r =0.0181
7.2 Design example 2—a system as shown in Fig. 2(b)

For the same solar thermal power plant and operational conditions as in Example 1, the
thermal storage primary material is molten salt with properties of ρ = 1680 kg / m3 ,
C s = 1560 J /( kg ⋅ K ) , and ks = 0.61 W /(m ⋅ K ) . The heat transfer fluid, Hitec (Bradshaw &
Siegel, 2009), flows in multiple heat transfer tubes, in the same arrangement as shown in Fig.
2(b). The required time period of energy discharge is 4 hours, the storage tank diameter is 8
m. The study will find the storage tank height.
The solution and design procedures are as follows:
3. Following the same procedure as seen in Example 1, we determine the mass flow rate to
be 40.35 kg/m3, and the ideal tank diameter and height to be 8 m and 6.44 m respectively.
It is assumed that we have 8448 steel pipes (with inner diameters of 0.025m) in the
storage tank, and that the heat transfer fluid flows in all of the pipes with an equal flow
rate. The void fraction ε is found to be 0.33, and the heat transfer surface area per unit of
S
length of the tank is found to be s =1327 m. The modified heat transfer coefficient inside
a pipe for laminar flow heat transfer is used based on the correction using Eqs. (23) to
(25), where the correction coefficient for the washer is wc = 3.69894 for the ratio of
Deq / di = 1.74 . The minimum volume of the tank is used in the first trial of the design.
The dimensionless values of HCR , Π d , and τ r are 0.522, 3.032, and 0.221, respectively.
Given in Fig. 26 is a chart for Π d =3.032 and τ r =0.221 at various values of HCR and
Π c / Π d . The figure shows that at an HCR value of 0.50 (close to 0.522) it is impossible
to get an energy efficiency of 1.0 for attempted time ratios of Π c / Π d up to 2.0.

To increase the energy delivery efficiency, a new tank height of 2.1 times that of the ideal
volume tank is used. This makes the value of Π d and τ r be 1.444 and 0.105, respectively.
Figure 27 shows the curves for Π d =1.444 and τ r =0.105. It is seen that at an HCR value of
0.50 (close to 0.522) and a time period ratio, Π c / Π d , of 1.2, the energy delivery effectiveness
can reach 0.96.
Note that in the finalized storage tank (with a height of 13.5 m), the volume of heat transfer
fluid takes 69% of the volume of an ideal thermal storage tank (with a height of 6.44 m). The
energy delivery effectiveness reached 0.96 if the charging time period was kept 20% longer
than the required discharge time. Obviously, the weak heat transfer between the HTF and
the primary thermal storage material in Example 2 is responsible for the much larger ratio of
the actual volume compared to the volume of the ideal storage tank. In Example 1 the ratio
of the actual volume of the finalized storage tank is only 1.25 times that of the volume of its
corresponding ideal storage tank. Therefore, it is important that in order to improve the
energy delivery efficiency, the heat transfer between the fluid and thermal storage material
must be improved, for example, by using pipes with fins. Nevertheless, this type of thermal
storage system, as in Example 2, still saves 31% of the heat transfer fluid.

Thermal energy storage in soil (Nassar et al., 2006), concrete (Zhang et al., 2004; Laing at al.,
2006; Reuss et al., 1997), and in sands (Wyman, 1980) has similar features to that of the
Example 2. When the heat transfer performance (the multiplication of the heat transfer
coefficient and the heat transfer area) between the heat transfer fluid and thermal storage
material is poor, the energy delivery effectiveness can be rather low. If the temperature
degradation of the discharged fluid is a big concern, for example in a power plant, the
thermal storage system in Example 2 may need enhancement of the heat transfer
performance between the heat transfer fluid and thermal storage material.
8. Concluding remarks
Thermal energy storage is very important to the development of concentrated solar thermal
power technologies. Thermal energy storage is also particularly attractive for large capacity
energy storage. It is becoming more and more common for industrial engineers to be
required to size and design thermal storage systems. Therefore, it is necessary to provide a
general analytical tool for the sizing and design of thermal energy storage systems.
The details of the analytical model provided in this book chapter will help researchers and
industrial engineers to understand the behavior of the heat transfer and energy transport
between heat transfer fluid and thermal energy storage material. Two examples for the
practical design of the size of thermal storage systems were provided for readers to practice
the design process concretely. The authors believe that the currently developed charts,
relating the energy storage efficiency to the properties and parameters of a thermal storage
system, are of great potential to help industrial engineers performing sizing calculations for
thermal storage systems. In this regard, the publication of this chapter is a significant and
timely event.
Due to space limitations, this chapter did not provide a large number of the developed
charts for thermal storage system sizing and design. However, the authors are currently
working to provide a handbook including a large number of design charts covering a wide
range of design parameters that industry will need. It is hoped that the handbook will be
available to readers in the near future.
9. Acknowledgement
The authors are grateful for the support provided by the US Department of Energy and the
National Renewable Energy Laboratory, under DOE Award Number DE-FC36-08GO18155,
and the US Solar Thermal Storage LLC. The Support from the Idaho National Laboratory
under award number 00095573 for phase change thermal storage is also gratefully
acknowledged.
10. Nomenclature
af The cross section area of a storage tank (m2)
Bi Biot number ( = Lp h / ks )
C Heat capacity ( J / kg ⋅ oC )
dr Nominal diameter of a single filler ‘particle’ (rocks) (m)
D Diameter of a storage tank (m)
Deq Equivalent diameter (m)
= Enthalpy of fluid at a location along the axis ( J / kg )
h Heat transfer coefficient ( W / m2 oC )
hp Modified heat transfer coefficient ( W / m2 oC )
H Length or height of a storage tank (m)
HCR A dimensionless parameter
k Thermal conductivity ( W / m oC )
Lp Characteristic length of ‘particles’ for Biot number.
m Mass flow rate (kg/s)
N Number of tubes for heat transfer fluid in a storage tank
Pr Prandtl number
Q Thermal energy involved per unit of time (W)
Q Thermal energy involved ( J )
r Average radius of the filler material (rocks) (m)
R Radius of the storage tank (m)
Re Modified Reynolds number for porous media
Ss Surface area of filler material per unit length of the storage tank ( m )
t Time (sec)
TH High temperature of fluid from solar field (°C)
TL Low temperature of fluid from power plant (°C)
U Fluid velocity in the axial direction in the storage tank (m/s)

V Volume ( m 3 )
z Location of a fluid element along the axis of the tank (m)
Greek symbols
αs Thermal diffusivity (= ks /( ρ sC s )) [m2/s]
ε Porosity of packed bed in a storage tank.
ηs Thermal storage efficiency.
ηT Thermal efficiency of a solar power plant.
Γ Fusion energy of thermal storage material ( J / kg )
μ Dynamic viscosity ( Pa ⋅ s )
Π Dimensionless charge or discharge time
τr A dimensionless parameter
ρ Density ( kg / m3 )
θ Dimensionless temperature.
Subscript
c Energy charge process
d Energy discharge process
f Thermal fluid
ref A required reference value
r Rocks
s Filler material (rocks), the primary thermal storage material.
z Location along the axis of the tank
Superscript
* Dimensionless values
11. References
Abdoly, M.A., Rapp, D., 1982, Theoretical and experimental studies of stratified thermocline
storage of hot water, Energy Conversion and Management, Vol. 22, No. 3, 1982, pp.
275-285.
Beasley, D.E., and Clark, J.A., 1984, Transient response of a packed bed for thermal energy
storage, International Journal of Heat and Mass Transfer, 27(9), pp. 1659 –1669.
Becker, M., 1980, Comparison of heat transfer fluid for use in solar thermal power stations.
Electric Power Systems Research, 3 (1980) 139-150.
Bradshaw R.W., Siegel N.P., 2009, Development of molten nitrate salt mixtures for
concentrating solar power systems, paper no. 11538, proc. SolarPACES, Berlin,
Germany, 2009.
Bradshaw, A.V., Johnson, A., McLachlan, N.H., Chiu, Y-T., 1970, Heat transfer between air
and nitrogen and packed beds of non-reacting solids, Trans. Instn Chem. Engrs,
Vol. 48, 1970, pp. T77-T84.
Brosseau, D., Kelton, J.W., Ray, D., Edgar, M., Chisman, K., and Emms, B., 2005, Testing of
thermocline filler materials and molten-salt heat transfer fluids for thermal energy
storage systems in parabolic trough power plants, Journal of Solar Energy
Engineering, 127(1), pp. 109–116.
Canada, S., Brosseau, D. A., and Price, H., 2006, Design and Construction of the APS 1MWe
parabolic trough power plant, ASME Conference Proceedings, 2006(47454), pp. 91–
98.
Conway, J.H. & Sloane, N.J.H., 1998, Sphere Packings, Lattices and Groups, (Third Edition).
ISBN 0-387-98585-9.
Courant, R., Hilbert, D., 1962, Methods of Mathematical Physics, Volume II, Wiley-Interscience.
Ferziger, J.H., 1998, Numerical Methods for Engineering Applications. Wiley-Interscience.
Gil, A., Medrano, M., Martorell, I., Lazaro, A., Dolado, P., Zalba, B., and Cabeza, L.F., 2010,
State of the art on high temperature thermal energy storage for power generation,
part 1—concepts, materials and modellization, Renewable and Sustainable Energy
Reviews, 14(1): 31-55.
Halawa, E., Saman, W., 2011, Thermal performance analysis of a phase change thermal
storage unit for space heating, Renewable Energy 36 (2011) 259-264.
Hales, T.C., 2006, Historical overview of the Kepler conjecture, Discrete & Computational
Geometry. an International Journal of Mathematics and Computer Science 36 (1): 5–
20.
Hasnain S.M., 1998, Review on sustainable thermal energy storage technologies, part I: heat
storage materials and techniques. Energy Convers. Mgmt. 1998, Vol. 39: 1127–1138.
Herrmann, U., and Kearney, D.W., 2002, Survey of thermal energy storage for parabolic
trough power plants, Journal of Solar Energy Engineering, 124(2), pp. 145–152.
Herrmann, U., Kelly, B., and Price, H., 2004, Two-tank molten salt storage for parabolic
trough solar power plants, Energy, 29(5-6), pp. 883 – 893.
Incropera, F.P., and DeWitt, D.P., 2002, Introduction to Heat Transfer, fourth ed. John Wiley
and Sons, Inc.
Jeffreson, C.P., 1972, Prediction of breakthrough curves in packed beds: 1. applicability of
single parameter models, American Institute of Chemical Engineers, 18(2), pp. 409-
416.
Karaki, W., Van Lew, J.T., Li, P.W., Chan, C.L., Stephens, J., 2010, Heat transfer in
thermocline storage system with filler materials: analytical model, Proceedings of
the ASME 2010 4th International Conference on Energy Sustainability, ES2010-
90209, May 17-22, 2010, Phoenix, Arizona, USA.
Karaki, W., Li, P.W., Van Lew, J.T., Valmiki, M.M., Chan, C.L., Stephens, J., 2011,
Experimental investigation of thermal storage processes in a thermocline storage
tank, Submitted to ASME 5th International Conference on Energy Sustainability,
ESFuelCell2011, August 7-10, Washington DC, USA
Kays, W. M., Crawford, M. E., and Weigand, B., 2005, Convective Heat and Mass Transfer,
fourth ed. McGraw Hill.
Kolb, G.J., 1998, Economic evaluation of solar-only and hybrid power towers using molten-
salt technology, Solar Energy, Vol. 62 (1), January 1998, pp. 51-61.
Kolb, G.J., and Hassani, V., 2006, Performance analysis of thermocline energy storage
proposed for the 1 mw saguaro solar trough plant, ASME Conference Proceedings,
2006 (47454), pp. 1–5.
Krane, R.J., and Krane, M.J.M., 1992, The optimum design of stratified thermal energy
storage systems—part II: completion of the analytical model, presentation and
interpretation of the results, Journal of Energy Resources Technology, 114(3), pp.
204–208.
Laing, D., Steinmann, W. D., Viebahn, P., Gräter, F., and Bahl, C., 2010, Economic analysis
and life cycle assessment of concrete thermal energy storage for parabolic trough
power plants, Journal of Solar Energy Engineering, 132, 041013.
Laing, D, Steinmann, W.D., Tamme, R., Richter, C., 2006, Solid media thermal storage for
parabolic trough power plants, Solar Energy, Vol. 80, pp. 1283-1289.
Li, P.W., 2008, Energy storage is the core of renewable energy technologies, IEEE
Nanotechnology Magzine, December 2008, pp. 3-18.
Li, P.W., Van Lew, J.T., Chan, C.L., Karaki, W., Stephens, J., 2010, Similarity and generalized
analysis of efficiencies of thermal energy storage systems, Submitted to Journal of
Renewable Energy, 2010.
McMahan, A. C., 2006, Design and optimization of organic Rankine cycle solar-thermal
power plants, Master’s thesis, University of Wisconsin, Madison, Wisconsin.
McMahan, A.C., Klein, S. A., and Reindl, D. T., 2007, A finite-time thermodynamic
framework for optimizing solar-thermal power plants, Journal of Solar Energy
Engineering, 129(4), pp. 355–362.
Michel, Y., Haller, C.A., Cruickshank, W.S., Harrison, S.J., Andersen, E., Furbo, S., 2009,
Methods to determine stratification efficiency of thermal energy storage processes –
Review and theoretical comparison, Solar Energy, Vol. 83, Issue 10, October 2009,
pp. 1847-1860.
Montes, M.J., Abánades, A., Martínez-Val, J.M., and Valdés, M., 2009, Solar multiple
optimization for a solar-only thermal power plant, using oil as heat transfer fluid in
the parabolic trough collectors, Solar Energy, Vol. 83 (12), Dec. 2009, pp. 2165-2176.
Nassar, Y., ElNoaman, A., Abutaima, A., Yousif, A., Salem, A., 2006, Evaluation of the
underground soil thermal storage properties in Libya, Renewable Energy, Volume
31, Issue 5, April 2006, Pages 593-598.
Nellis, G., and Klein, S., 2009, Heat Transfer, Cambridge University Press.
Pacheco, J.E., Showalter, S.K., and Kolb, W.J., 2002, Development of a molten salt
thermocline thermal storage system for parabolic trough plants, Journal of Solar
Energy Engineering, 124(2), pp. 153–159.
Pitz-Paal, R., Dersch, J., Milow, B., Tellez, F., Ferriere, A., Langnickel, U., Steinfeld, A., Karni,
J., Zarza, E., Popel, O., 2007, Development steps for parabolic trough solar power
technologies with maximum impact on cost reduction, Journal of Solar Energy
Engineering, 131, pp. 371-377.
Polyanin, A. D., 2002, Handbook of Linear Partial Differential Equations for Engineers and
Scientists, Boca Raton: Chapman & Hall/CRC Press, ISBN 1-58488-299-9.
Price, H., Lupfert, E., Kearney, D., Zarza, E., Cohen, G., Gee, R., Mahoney, R., 2002,
Advances in parabolic trough solar power technology, Journal of Solar Energy
Engineering, Vol. 124, May 2002, pp. 109-125.
Product Resources, http://www.radcoind.com/index.html
Regin, A. F., Solanki, S.C., Saini, J.S., 2008, Heat transfer characteristics of thermal energy
storage system using PCM capsules: A review, Renewable and Sustainable Energy
Reviews 12 (2008) 2438–2458.
Renewables 2007, 2008, Global Status Report, Issued by Renewable Energy Policy Network
for the 21st Century (www.REN21.net).
Reuss, M., Beck, M., Muller, J.P., 1997, Design of a seasonal thermal energy storage in the
ground, Solar Energy, Vol. 59, No. 4-6, April-June 1997, pp. 247-257.
Sari, A., Kaygusuz, K., 2001, Thermal performance of myristic acid as a phase change
material for energy storage, application. Renewable Energy, 24 (2001) 303–317.
Schumann, T. E.W., 1929, Heat transfer: a liquid flowing through a porous prism. Journal of
the Franklin Institute, 208(3), pp. 405 – 416.
Shitzer, A., and Levy, M., 1983, Transient behavior of a rock-bed thermal storage system
subjected to variable inlet air temperatures: Analysis and experimentation. Journal
of Solar Energy Engineering, 105(2), May, pp. 200–206.
Singer, C., Buck, R., Pitz-Paal, R., and Müller-Steinhagen, H., 2010, Assessment of solar
power tower driven ultrasupercritical steam cycles applying tubular central
receivers with varied heat transfer media. Journal of Solar Energy Engineering, 132,
041010.
Spiers, D.J., and Rasinkoski, A.D., 1995, Predicting the service lifetime of lead/acid batteries
in photovoltaic systems, J. Power Sources, vol. 53, no. 2, pp. 245–253.
Therminol VP-1 heat transfer fluid by Solutia, 1999, Technical Bulletin 7239115B, Solutia,
Inc., 1999.
Tritt T.M., 2005, Thermal Conductivity: Theory, Properties, and Applications (Physics of
Solids and Liquids), Springer, 2005-05-13, ISBN: 0306483270.
Van Lew, J.T., Li, P.W., Chan, C.L., Karaki, W., Stephens, J., 2009, Transient heat delivery
and storage process in a thermocline heat storage system, IMECE2009-11701,
Proceedings of the ASME 2009 International Mechanical Congress and Exposition,
November 13-19, 2009, Lake Buena Vista, Florida, USA.
Van Lew, J.T., Li, P.W., Chan, C.L., Karaki, W., Stephens, J., 2011, Analysis of heat storage
and delivery of a thermocline tank having solid filler material, Journal of Solar
Energy Engineering, MAY 2011, Vol. 133/021003.
Wu, S.M., Fang, G.Y., Liu, X. 2011, Dynamic discharging characteristics simulation on solar
heat storage system with spherical capsules using paraffin as heat storage material,
Renewable Energy, 36 (2011) 1190-1195.
Wyman, C., Castle, J., Kreith, F., 1980, A review of collector and energy storage technology
for intermediate temperature applications, Solar Energy, Vol. 24, No. 6, 1980, pp.
517-540.
Yang, Z., Garimella, S.v., 2010, Molten-salt thermal energy storage in thermoclines under
different environmental boundary conditions, Applied Energy, Vol. 87, No. 11,
November 2010, pp. 3322-3329.
Zalba, B., Marin, J.M., Cabeza, L.F., Mehling, H., 2003, Review on thermal energy storage
with phase change: materials, heat transfer analysis and applications, Applied
Thermal Engineering, 23 (2003) 251–283.
Zarty, O., Juddaimi, A. E., 1987, Computational models of a rock-bed thermal storage unit,
Solar and wind Technology, 2(4), pp. 215- 218.
Zhang, D., Li, Z.J., Zhou, J.M., Wu, K.R., 2004, Development of thermal energy storage
concrete , Cement and Concrete Research, Vol. 34, No. 6, June 2004, pp. 927-934.
21
Role of Heat Transfer on

Process Characteristics
During Electrical Discharge Machining
Ahsan Ali Khan
International Islamic University Malaysia
Malaysia
1. Introduction
Different non-traditional machining techniques are getting more and more employed
in manufacturing of complex machine components. Among the non-traditional methods
of material removal processes, electrical discharge machining (EDM) has drawn a great
deal of researchers’ attention because of its broad industrial applications (Zarepour, et al.,
2007).
In this process material is removed by controlled erosion through a series of electric sparks
between the tool (electrode) and the workpiece (Ghosh & Malik, 1991). The sparks are
conducted in a dielectric medium. During the spark discharge the dielectric material is
ionized and allows the electrons to pass, but it should be deionized immediately after the
spark. In the present study kerosene was used as the dielectric medium. The sparks leads to
intense heat conductions onto workpiece causing melting and vaporizing of workpiece
material (Yan & Tsai, 2005). The temperature during EDM may rise from 8000 °C to 20,000 °C.
This high heat energy can melt almost any material. EDM has become one of the most
widely used non-traditional machining processes and can be used on any material that
conducts electricity. EDM is widely used in machining high strength steel, tungsten carbide
and hardened steel (Wang & Tsai, 2001). These materials are used in making dies and molds
and it is very difficult to machine them by conventional metal cutting techniques. Fig. 1
shows that die-sinking applications are typically dominated by plastic injection and various
other mold fabrications processes.
EDM may be used to cut and shape hard materials and complex shapes but can cause
surface damage as the electrical charge flows between the electrode and workpiece material.
Because there is no mechanical contact between the tool and the workpiece as in traditional
operation, the hardness, toughness and strength of the workpiece material has minimal
effect on the material removal rate (MRR).
EDM is gaining popularity due to the following reasons:
Machining of Complex Shapes: Complex cavities can often be machined without difficulties
by EDM. It can be used to machine material with small or odd shaped holes, a large number
of holes, holes having shallow entrance angles, intricate cavities, or intricate contour.
Adaptability to Micro-Machining: Micro-EDM has been adopted as one of the most valuable
techniques for micro-fabrication. Micro- EDM is accepted as an efficient method for the
fabrication of precise micro-metal components and holes because of its excellence minimum
machinable size and diameters of about 5 µm. EDM is the only process capable of machining
three dimensional micro workpiece of size as small as 0.01 mm. With development of the
technology micro workpieces find their applications in micro electro mechanical systems
(MEMS).
Misc. molds
Micromachining
Stamping
Al and plastic
extrusion
Powder
compacting
dies
Production
parts
Plastic injection
molds Forges
Misc.
Fig. 1. Application of EDM die-sinking (Peter Fonda et al., 2007)
High Degree of Automation: The high degree of automation and the use of tool and
workpiece changers allow the machines to work unattended for overnight or during the
weekends. CNC control with feedback is generally used and the automatic electrode and
pallet changers can be added to EDM machine; thus, the process lends itself to unattended
and safe operation.
Accuracy of the Process: EDM is a very accurate machining process. Especially in the case of
wire-EDM, where the electrode (wire) is constantly renewed with the continuous unwinding
of the wire (no effect of tool wear and other tool inaccuracies), very accurate structures can
be machined. In the case of workpiece with a higher thickness, the accuracy and the fine
surface quality remains the same over the whole thickness of the workpiece, due to the fact
that the EDM wire is machining with the same process conditions over the total workpiece
height (Ho and Newman, 2003).
Insensitive to the Hardness of the Material (Workpiece): The capability of machining hard
materials is one of the main advantages of EDM, as most of the tools for forging, extrusion,
die-casting dies and moulds are made of hard materials to increase their lifetime. The recent
developments in cutting tools for turning and milling and the processes of high speed
machining allow to machine harder materials than before, but EDM still remains the only
available process for machining many hard materials. One of the important areas of EDM
application is in mould and dies making industries. To have a mould with longer life span,
workpiece material must be very hard. The high hardness of the workpiece is usually
obtained by a heat treatment. After the treatment, most workpiece cannot be machined
by conventional processes; EDM is the appropriate technique to manufacture these
workpiece.
Role of Heat Transfer on Process Characteristics during Electrical Discharge Machining 419
Absence of mechanical forces on the workpiece: The EDM-process is based on a thermal

principle; no mechanical force is applied to the workpiece; it eliminates the mechanical
stresses, chatter and vibration problem during machining (Ho and Newman, 2003). This
allows the effective machining of very thin, delicate and fragile workpiece without
distortion. The large cutting force of the materials removal processes on the convectional
machines is absent in EDM.
The electrode is the main part of the EDM process, which is connected to the DC power
source and is immersed in the dielectric fluid. The two main types of electrode materials are
copper and graphite. Each of these materials has their own important and distinguishing
features. Graphite is a widely used electrode material, but unlike most metals it does not
melt but undergoes sublimation at around 3350°C. This is the reason why graphite shows
less wear than copper, as there is less thermal damage. Graphite also has superior fabrication
capabilities. Other materials used for making electrodes are brass, copper-tungsten,
tungsten carbide, zinc, aluminum, etc. (Kalpakjian 2001; Zaw, 1999). Efforts have been done
to minimize electrode wear (EW). A metal matrix composite (ZrB2-Cu) was developed
adding different amount of Cu to get an optimum combination of wear resistance, electrical
and thermal conductivity of electrodes (Khanra et al., 2007). It was reported that ZrB2-40
wt% Cu composite shows more MRR with less EW. The electrode material is to be selected
carefully depending on the work material to machine. The main important properties of
electrode materials are electrical conductivity and thermal conductivity. During EDM, the
main output parameters are MRR, EW, wear ratio (WR) and job surface finish (Ra)
(Marafona & Wykes, 2000; Wang &Tsai, 2001). The suitability of the electrodes is judged
considering the machining output parameters. EW is quite similar to the material removal
mechanism as the electrode and the workpiece are considered as a set of electrodes in EDM
(Ho & Newman, 2003). Due to this wear, electrodes lose their dimensions resulting inaccuracy
of the cavities formed (Khan & Mridha, 2006). Research works have been conducted to draw
the relationship of MRR with current, voltage, pulse duration, etc (Ramasawmy & Blunt,
2001). C. F. Hu et al., 2008 found that MRR was enhanced acceleratively with increasing
discharge current and work voltage, but increased deceleratively with pulse duration.
Manufacturing of electrodes of special composition is expensive and not always cost
effective. In order to maintain the accuracy of machining, EW compensation has been
reported to be an effective technique, where wear was continuously evaluated by sensors
and the compensation was made (Bleys et al., 2004). Some researchers have tried to develop
mathematical models to optimize the EW and MRR (Puertas et al., 2004; Kunieda et al.,
2004). It was reported that MRR can be substantially increased with reduced EW using a
multi-electrode discharging system. But again, a special electrode involves additional cost.
In the present study the most common and easily available electrode materials like copper
and brass were taken under consideration during machining of aluminum and mild steel.
Wear of the electrode along the direction of movement of the electrode can be compensated
by imparting additional movement of the electrode. But the wear along the cross-section of
the electrode cannot be compensated. This phenomenon results inaccuracy in the dimension
of the cavities made by die-sinking technique of EDM. In the present study an analysis has
been done to evaluate the EW along the cross-section of the electrode compared to the same
along its movement. An analysis has also been done on the comparative performance of
copper and aluminum as electrode materials.
2. Heat source models

A number of different heat input methods have been adopted by various researchers to
predict the temperature distribution in electrodes. One of the models is disk heat source
used to simulate heat input into the workpiece surface. The heat source was insulated at the
outer area and assumed to exist during the pulse on-time. The authors assumed that the
fraction of the energy transferred to the workpiece is 50%. Tariq & Pandey, 1984 developed
models assuming that the heat from the plasma channel is transferred to the workpiece or
electrode by conduction only. About 90% of the total energy liberated is conducted to the
discharge gap and it was distributed equally between the electrode and workpiece.
DiBitonto et al, 1989 approximated the heat source by a point instead of a disk for
conducting heat to the interior. The point heat source can result in pulse with high current
density and a very small hot spot on the electrode. Uniformly distributed heat source within
the spark was considered by most EDM researchers, this assumption is not in line with real
EDM process (Vinod Yadav et al., 2002). This point is buttressed by the actual craters shape
formed during EDM. Gaussian heat input model gives the best approximation of EDM
plasma shape, which accounts for its wide usage by EDM researchers. This is evident by the
works of Marafona & Chousal, 2006.
z
qm
R
rsp
Workpiece
Fig. 2. Gaussian heat source model

For a Gaussian heat distribution, if the maximum heat intensity qm (Fig. 2) is at the axis of a
spark and its radius (rsp) are known, the heat flux qf at radius R is given by:
⎡
⎛ ⎞
2⎤
⎢ ⎥
⎢ − 4.5 ⎜ R ⎟ ⎥
⎢ ⎜⎜ rsp ⎟⎟ ⎥
4.45 WM IV ⎢ ⎝ ⎠ ⎥
Heat flux q ( R) = × e⎣ ⎦ (1)
f π (rsp )2
where
qf = Heat flux (W/mm2); WM = Fraction of energy utilized by the material (Watt)
I = Pulse current (Amp); V = Gap voltage (Volt)
R = Radial distance from the axis of the spark (µm); rsp = Plasma channel radius (µm)
2.1 Plasma channel radius

The extreme high temperature gradient that exists between plasma channel and electrode-
workpiece interface enhances heat conduction through the contact surface. Plasma channel
with high-energy plasma exhibits spatial and sequential variation at plasma channel-
electrode-workpiece interface. This leads to variation in energy flux available for conduction
into the electrode and workpiece (Dibitonto et al., 1989). Plasma channel radius is proportional
to current, pulse duration, and the proportionality constants and indices depend on inter-
electrode gap, electrode and workpiece materials and as well as dielectric fluid (Marafona &
Chousal, 2006).
Measurement of plasma channel radius is extremely difficult due to very high pulse
frequencies (Vinod Yadav et al., 2002). There are number of different equations used to
calculate plasma channel radius according to different authors. It was proposed to use Eq 2
which depends on discharge current and time to calculate plasma channel radius.
R ( t ) = KQmt n (2)
where
R = Plasma channel radius (µm)
Q = Discharge current (Amp)
exponents m, n and K are empirical constant
Eq. 3 obtained by Patel et al, 1989 was used to calculate plasma channel radius in EDM
process. According to the authors, the equation is time dependent with the empirical
constants K and exponent n given as 0.788 and 0.75 respectively
R ( t ) = K tn (3)
In the work of Shuvra et al, 2003 it was pointed out that plasma radius varies with time as
per the relationship in Eq. 4
3
a = Rt4 (4)
where t is time in µs, a is radius in µm and R is the proportionality constant.

According to Shuvra et al, because they could not have a good estimated value of
proportionality constant R, they took plasma radius (a) to be 0.75 µm which is equal to the
radius of the crater. Pandey & Jilani, 1986 in their paper titled, “Plasma channel growth and
the resolidified layer in EDM”, used Eq. 5 to calculate the plasma radius R.
0.5
Eo R ⎡ 4 αt ⎤
Tb = tan−1 ⎢ ⎥ (5)
Kπ 0.5 ⎣ R2 ⎦
where Tb = boiling temperature (0C), Eo is the energy density (Watt/m2) and α is the thermal
diffusivity (m2/s). Salonitis et al, 2007 in their works used Eq. 6 to obtained equivalent heat
input radius which is dependent on the current intensity and pulse on-time duration.
rsp = 2040 × (I)0.43 × (ton )0.44 (6)

where rsp is heat input radius in µm, I is current intensity in A and ton is the spark time in µs.
In the work of Bulent et al., 2006 and Ozgedik & Cogun, 2006 they used Eq. 7 to estimate the
energy released due to a single spark.
Esp = I ×V × ton (7)
where
Esp = Energy released due to one spark (Watt)
I = Pulse current (Amp)
V = Gap voltage (Volt)
ton = on-time (µs)
According to Bulent Ekmekci et al, 2006 the total energy received by any material due to a
single spark depends on it thermal conductivity, and is given by Eq. 8.
EM = WM × I × V × ton
(8)
where
EM = Total energy received by the material (Watt)
WM = Fraction of energy received by the material (Watt)
3. Experimental details
3.1 Electrode and work materials
In the present study copper and brass were taken as the electrode materials. These materials
have high electrical and thermal conductivity. High electrical conductivity of these materials
facilitates conducting electrical sparks. Their high thermal conductivity allows the absorbed
heat to escape easily and thus reduces EW. Moreover, it is easy to machine these materials
by conventional machining techniques into complex shapes. The electrodes used in the
present study were 70 mm long with a cross-sectional dimension of 15 mm x 15 mm. The
major properties of the electrode materials are shown in Table 1.
Electrode Thermal Melting Electrical Specific heat

materials conductivity point (°C) resistivity capacity
(W/m-°K) (ohm-cm) (J/g-°C)
Copper 391 1083 1.69 0.385
Brass 159 990 4.7 0.38
Table 1. Major properties of electrode materials
The workpiece materials used in the present study were mild steel and aluminum. Their
chemical composition and major properties are shown in the Table 2 and Table 3
respectively.
Work materials Chemical composition

Aluminum Al: 99.9%, Cu:0.05%, Fe:0.4%, Mg:0.05%, Mn:0.05%, Si:0.25%, Zn:0.05%
Mild steel C: 0.14%-0.2%, Fe: 98.81-99.26%, Mn: 0.6%-0.9%, P: 0.04%, S: 0.05%
Table 2. Chemical composition of the work materials
Work Thermal Melting Electrical Specific Hardness Tensile Yield Percentage

materials conductivity point resistivity heat (HB) strength strength elongation
(W/m-°K) (°C) (ohm-cm) capacity (MPa) (MPa) at break
(J/g-°C) (%)
Aluminum 227 660 2.9 0.9 35 131 124 8
Mild steel 51.9 1523 1.74 0.472 143 475 275 38
Table 3. Major properties of work materials
3.2 EDM machine

The experimental work were conducted on a die sinking EDM machine of type Mitsubishi
EX 22 model C11E FP60E. The machine has a rectangular table for mounting the workpiece.
The electrode is fixed above the workpiece on the tool holder and can be given a vertical
feed to machine a particular cavity. Usually the electrode is connected to the negative
terminal and the workpiece is connected to the positive terminal of the power supply which
is called direct polarity. However, depending on the combination of the electrode and the
work materials, reverse polarity is also sometimes used. The machine used in the present
study is shown in Fig. 3.
Fig. 3. EDM die sinking machine
3.3 Dielectric material

The common properties of dielectric fluids are: The dielectric fluid should have sufficient
and stable dielectric strength; it should de-ionize rapidly after the spark discharge; it should
have low viscosity and good wetting capacity; its flash point should be sufficient high to
avoid any fire hazard; it should not emit any toxic vapours or have unpleasant odours; it
should maintain its properties under all working conditions; it should be chemically neutral
to the workpiece, electrode or the work table and it should be easily available at a
reasonable price. Considering the above factors, kerosene was selected as the dielectric fluid
in the present study.
The product of machining, the tiny debrtis should be removed from the machining zone in
order to avoid sparks between the electrode and the debrtis. This is done by effective
flushing system of the dielectric fluid to remove the debrits away from the machining zone.
Suction flushing of dielectric fluid is preferred when straight vertical walls of the machined
cavity are needed. The vertical surfaces are produced with a small taper angle of 0.005mm to
0.05mm per 10mm depth when pressure flushing is employed. However, for machining die
cavities a small taper angle is desired. In the present study the machine was incorporated
with circulating emission flushing system coupled with jets of three nozzles in order to
assure the adequate flushing of the debris from the gap between the tool and the electrode.
The experimental set-up is shown in Fig. 4. The current tried in the experiments were 2.5
amps, 3.5 amps and 6.5 amps. The level of voltage used was 10 volts and 5 volts.
1 – tool; 2 – dielectric fluid; 3 – workpiece

Fig. 4. Set-up for the experiments
Square holes of dimensions of 15 mm x 15 mm were machined with a depth of 3 mm. After
machining, the wear (Fig. 5) along the direction of the electrode movement (y-direction) and
across the electrode (x-direction) was measured using an optical microscope Mitutoyo
Hisomet II.
Fig. 5. Electrode wear in x and y directions

4. Results and discussions

4.1 Electrode wear rate
The present EDM technology allows for a very small level of EW because EW cannot be
eliminated completely. EW is a critical problem in EDM since the tool shape degeneration
directly affects the final shape of the product. In addition, the cost of a part manufactured by
EDM is directly determined by the tool cost, which consists of the raw material cost, the
production cost and the number of electrodes required for the machining. In most EDM
operations, the cost of the electrode contributes more than 70% to the total operation cost. Due
to these reasons, EW should be carefully taken into consideration in planning and designing
EDM operations. Electrode wear is affected by the precipitation of carbon from the
hydrocarbon dielectric onto the electrode surface during sparking. The thickness of the carbon
inhibitor layer made a significant improvement on the EWR with little effect on the MRR.
As it was mentioned before, during EDM the electrode is connected to the negative terminal
and the workpiece is connected to the positive terminal of the power supply. Generally, the
rate of material removal from the cathode is comparatively less than that from the anode
due to the following reasons (Ghosh & Malik, 1991):
i. The momentum with which the stream of electrons strikes the anode is much more than
that due to the stream of the positive ions impinging on the cathode though the mass of
an individual electron is less than that of the positive ions.
ii. The pyrolysis of the dielectric fluid (normally a hydrocarbon) creates a thin film of
carbon on the cathode.
iii. A compressive force is developed on the cathode surface.
Therefore, normally, the tool is connected to the negative terminal of the DC source.
The weight of the electrodes before and after machining gives EW. It can be expresses as
follows:
volume of material removed from the electrode
EWR =
machining time
EVB − EWA
EWR =
MT
where EVB and EWA are the volumes of the workpiece before and after machining and
MT is the machining time.
In Fig. 6 to Fig. 9 electrode wear along the cross-sectional direction and along the length of
the tool has been presented after a machining a cavity of 3 mm depth. Fig. 6 to Fig. 9 shows
that electrode wear rate increases both in x-direction and y-direction with increase in
current. It is obvious that a higher current will produce a stronger spark which would cause
more material to be eroded from the electrodes. It can also be observed that at a higher gap
voltage an electrode undergoes more wear compared to that at a low gap voltage.
For example, during machining of aluminum with brass electrode with a current of 3.5 A the
EW in x-direction is 0.64 mm at a gap voltage of 5 volts (Fig. 6), whereas, under the same
condition the wear in x-direction is 0.86 mm at a gap voltage of 10 volts (Fig. 8).
It can be observed that in all cases (Fig. 6 to Fig. 9) wear of the electrodes in y-direction (along
the length) is less than that in x-direction (along the cross-section) of the electrodes. As it
was mentioned before, the length of the electrode is 70 mm (in y-direction) and the length of
each side of the cross-sectional area is 15 mm (in x-direction). As a result the heat generated
during the spark cannot be transferred into the body of the electrode easily in x-direction
Electrode wear in x-direction (V=5 volts)

1.8
1.6 Cu-
Electrode wear (mm)

1.4 Aluminum
1.2
Cu-steel
1
0.8
0.6 Brass-
0.4 steel
0.2
Brass-
0
0 2 4 6 8 Aluminum
Current (Amp)
Fig. 6. Relationship of current with electrode wear along the cross-section (v=5 volts)
Electrode wear in y-direction (V=5 volts)

1.40
Cu-
1.20
Aluminum
Electrode wear (mm)
1.00
0.80 Cu-steel
0.60
0.40 Brass-
0.20 steel
0.00
Brass-
0 2 4 6 8 Aluminum
Current (Amp)
Fig. 7. Relationship of current with electrode wear along the length (v=5 volts)
Electrode wear in x-direction (V=10 volts)

2.00
Cu-
Aluminum
1.50
Electrode wear (mm)
Cu-steel
1.00
0.50 Brass-
steel
0.00 Brass-
0 2 4 6 8 Aluminum
Current (Amp)
Fig. 8. Relationship of current with electrode wear along the cross- section (v=10 volts)
Electrode wear in y-direction (V=10 volts)

1.80
1.60
Cu-
Electrode wear (mm) 1.40 Aluminum
1.20
1.00 Cu-steel
0.80
0.60
Brass-
0.40 steel
steel
0.20
0.00
Brass-
Brass-
0 1 2 3 4 5 6 7 Aluminum
Aluminum
Cur r ent ( Amp)
Fig. 9. Relationship of current with electrode wear along the length (v=10 volts)
compared to that in y-direction. This results more wear of electrodes in x-direction
compared to that in y-direction. For the same reason it can be expected that that an electrode
of a smaller cross-section permits poor heat transfer in x-direction and will undergo more
wear compared to that of a larger cross-section. Fig. 10 clearly shows that electrode wear
along the cross-section of the electrode is higher compared to the same along its length
during machining of aluminum using brass electrodes.
(a) current: 2.5 amps (b) current: 3.5 amps (c) current: 6.5 amps
Fig. 10. Wear of brass electrodes; electrode – brass, workpiece – aluminum
Results of the experiments show that copper electrodes undergo less wear compared to
brass electrodes. For example, during machining of steel at a current of 3.5 amps and a
voltage of 5 volts, the wear of copper electrode is 0.29 mm in x-direction, whereas the same
for brass electrode is 1.5 mm (Fig. 6). This is due to the fact that the thermal conductivity of
copper (391 W/m-K) is almost 2.5 times higher than that of brass (159 W/m-K). This
facilitates rapid heat transfer through the body of copper electrodes compared to brass
electrodes. It can also be noted that melting point of copper (1083°) is higher to that of brass
(990°) that causes less melting and wear of copper electrodes.
From Fig. 6 to Fig. 9 it can also be observed that during machining of steel both the copper
and the brass electrodes undergo more wear compared to the same during machining of
aluminum. It can be mentioned that thermal conductivity of aluminum (227 W/m-K) is
almost four times to that of steel (51.9 W/m-K). As a result the heat generated during each
spark is easily absorbed by aluminum compared to that absorbed by steel. This causes less
wear of electrodes during machining of aluminum compared to the same during machining
of steel. For example, during machining of steel with brass electrodes at a current of 6.5
amps and a voltage of 10 volts electrode wear in y-direction is 1.6 mm, while the same is 0.8
mm during machining of aluminum (Fig. 9).
4.2 Wear ratio

Wear ratio (WR) is calculated as the ratio of the material removed from the work to the
material removed from the electrode. It can be expresses as follows:
volume of material removed from the electrode

WR = × 100%
volume of material removed from the workpiece
EVB − EWA
WR = × 100%
WPVB − WPWA
where EVB and EWA are the volumes of the workpiece before and after machining.
Wear ratio vs current (V=5)

30
Cu-
25
Aluminum
20
Cu-steel
wear ratio
15
Brass-
10
steel
5
Brass-
0 Aluminum
0 2 4 6 8
current (Amp)
Fig. 11. Relationship of current with wear ratio (v=5 volts)
Wear ratio vs current (V=10)

20 Cu-
Aluminum
15
Cu-steel
wear ratio
10
Brass-
5 steel
0 Brass-
Aluminum
0 2 4 6 8
current (Amp)
Fig. 12. Relationship of current with wear ratio (v=10 volts)

Fig. 11 and Fig. 12 illustrate the relationship of WR with current and voltage. It is obvious
from the figures that WR decreases with increase in current. That means, though a higher
current causes more removal of work material and the electrode, but comparatively more
material is removed from the electrode. At a higher current a stronger spark is generated
producing more heat. The size of the workpiece is massive and heat is easily dissipated
through it. But since the electrode is a smaller one, heat is accumulated in it resulting high
temperature and consequently high EW.
It can also be observed that WR is higher at a lower voltage (Fig. 11) compared to that at a
higher voltage (Fig. 12). Highest WR of 25.00 was found during machining of aluminum
using a copper electrode at a current of 2.5 amps and a voltage of 5 volts. The total heat
generated during a spark is absorbed by the workpiece, electrode, dielectric fluid and the
machine parts. It is desirable that most of the heat should be absorbed by the work material,
while the least quantity of heat should be absorbed by the electrode. High thermal
conductivity of copper electrodes facilitates easy heat transfer and its high melting point
facilitates low melting of the electrode material. At the same time, high thermal conductivity
of aluminum facilitates easy heat absorption and its low melting temperature facilitates fast
removal of work material. These factors result the highest WR during machining of
aluminum using a copper electrode. The lowest WR of 2.00 was found during machining of
aluminum using brass electrodes at a current of 6.5 A and a voltage of 5 volts. Thermal
conductivity of brass is only 1.4 times higher than that of the aluminum workpiece and its
melting point is approximately 1.5 times higher than that of aluminum. As a result, material
removal from the aluminum workpiece is comparatively low and the material removed
from the electrode is comparatively high. Therefore, the WR is only 2.00. From Fig. 11 and
Fig. 12 it can be concluded that in order of high to low WR of electrode-work pair are:
copper-aluminum, copper-steel, brass-steel and brass-aluminum.
4.3 Material removal rate

The mechanism of material removal in EDM is a sudden violent splash of the molten
material coinciding with the collapse of the plasma channel. As soon as the spark collapses
and the hydrostatic pressure of the arc is released and dielectric rushes back to fill the void,
the pressurized molten metal splatters from the workpiece surface leaving a crater on the
surface and some splattered fragments around the crater (Shuvra et al, 2003). Material
removal mechanism can also be explained in terms of the migration of material elements
between the workpiece and electrode. This could take place in the form of elements
diffusing from the electrode to the workpiece and vice-versa.
The use of CNC in EDM has helped to explore the possibility of using alternative methods
of tooling to improve the MRR. EDM is commonly used for producing complex and deep 3-
D shaped cavity in hard materials which necessitate the use of 3-D profile electrodes, which
are costly and time-consuming to manufacture for the sparking process. However, in order
to produce a complex shape a lot of experimental works have been performed with different
types of electrode configurations to generating different path movement on workpiece
surfaces by means of controlling the electrode motion.
Flushing pressure has a considerable effect on MRR and EW. When the dielectric fluid is
forced at low velocity into the spark gap, short-circuiting becomes less pronounced as a
result of the accumulated particles. This helps in improving the efficiency and thereby
increases MRR. Higher flushing pressure hinders the formation of ionized bridges across the
gap and results in higher ignition delay and decrease discharge energy and reduces MRR.
It was found by many researchers that the influential machining factors on MRR are the
current intensity and voltage.
Usually EDM is carried out by electrical sparks between the electrode and the workpiece
using a single discharge for each electrical pulse. Some researchers have carried out
experiments using a multi-electrode discharging system, delivering additional discharge
simultaneously from a corresponding electrode connected serially. The design of electrode
was based on the concept of dividing an electrode into multiple electrodes, which are
electrically insulated. The energy efficiency were claimed to be better than the conventional
EDM without any significant difference in work suface finish.
Material removal rate is expressed as the ratio of the difference in volume of the workpiece
before and after machining to the machining time, i.e.:
volume of material removed from the workpiece

MRR =
machining time
WPVB − WPWA
MRR =
MT
where WPVB and WPWA are the volumes of the workpiece before and after machining
and MT is the machining time.
Material removal rate vs current
35
30
material removal
25
rate(mm3/min)
Cu-Al
20
Cu-Steel
15
Brass-Al
10
Brass-Steel
5
0
-5 0 5 10
current (A)
Fig. 13. Relationship between current and MRR

Relationship of MRR with current during machining of aluminum and steel using brass and
copper electrodes are illustrated in Fig. 13. It is to be noted that at a low current MRR is very
low, but with increase in current MRR increases sharply. At a low current, a small quantity
of heat is generated and a substantial portion of it is absorbed by the surroundings and the
machine components and the left of it is utilized in melting and vaporizing the work
material. But as the current is increased, a stronger spark with higher energy is produced,
more heat is generated and a substantial quantity of heat is utilized in material removal.
However, the highest material removal rate was observed during machining of aluminum
using copper electrodes. Comparatively low thermal conductivity of brass as an electrode
material doesn’t allow absorbing much of the heat energy and most of the heat is utilized in
removal of material from aluminum workpiece of low melting point. But during machining
of steel using copper electrodes, comparatively smaller quantity of heat is absorbed by the
work material due to its low thermal conductivity. As a result MRR becomes very low.
4.4 Micro cracks

During the spark discharge in EDM the temperature is usually in the range of 8,000°C to
20,000°C. After the spark the work surface is immediately cooled rapidly by the dielectric
fluid. Repeated heating to a very high temperature followed by rapid cooling develops
micro-cracks on the work surface. Micro-cracks on the work surface are a major problem in
EDM. They strongly influence on the fatigue strength of the part machined by EDM. Micro-
cracks in the surface and loose grains in the subsurface resulted from thermal shock causes
surface damage and leads to degradation of both strength and reliability. Comparing the
SEM images in Fig. 14 it can be observed that more micro-cracks were formed during EDM
with a higher current of 6.5 Amp as shown in Fig. 14 (a) compared to that with a low current
of 2.5 Amp as illustrated in Fig. 14 (b). More heat is developed during EDM at a higher
current heating the work surface to a higher temperature followed by rapid cooling. As a
result more micro-cracks are found at a higher current. A larger ton results more cracks as it
can be observed comparing the Figs. 14 (c) and 14 (d). However, it was suggested by Lee &
Tai, 2003 that when the pulse voltage is maintained at a constant value of 120 V, it is
possible to avoid the formation of cracks if machining is carried out with a current in the
range of 12-16 A together with pulse duration of 6-9 µs.
4.5 Recast layer

There are three layers created on the top of the base metal which are spattered EDM surface
layer, recast layer and Heat Affected Zone (HAZ). Recast layer consists of dielectric fluid,
molten electrode and molten workpiece that are melted during EDM machining and
solidified. Usually recast layer has a higher hardness when compared to the base metal. The
recast layer is also known as white layer because it often appears as a bright white layer in a
sectional view under magnification. It occurs as the second layer under the spattered EDM
surface layer. This layer is formed by the un-expelled molten metal solidifying in the crater.
The recast layer is usually very thin and it can be removed by finishing operations. Recast
layer can cause problems in some applications due to stress cracking or premature failure.
Recast structure greatly affects die fatigue strength and shortens its service life. This is
because the recast layers have micro-cracks and discharge craters that cause bad surface
quality. HAZ consists of two layers: a hardened layer and the annealed layer. The depth of
the hardened layer depends on the machining conditions. Usually the depth is 0.002mm for
finish cut and 0.012 mm for rough cut. Below the hardened layer there is a layer which was
cooled slowly and as a result, the layer is annealed. Its hardness is 2 to 5 points below the
same of the base metal. Its thickness may be 0.05mm for finish cut and 0.2mm for rough
cut.
(a) I=6.5 Amp; ton=10 µs
(b) I=2.5 Amp; ton=10 µs
(c) I=2.5 Amp; ton=10 µs

(d) I=2.5 Amp; ton=3 µs

Fig. 14. Influence of current and pulse-on time on micro cracks
As stated above, a higher current and a higher pulse-on time produce a spark with more
energy, melt more materials from the workpiece and the electrode. Consequently higher
thickness of recast layer is found at a current of 6.5 Amp, Fig. 15 (a) compared to that at a
current of 2.5 Amp, Fig. 15 (b). Similarly, thickness of the recast layer was found to be at a
higher pulse-on time, Fig. 15 (c) compared to that at a shorter pulse-on time, Fig. 15 (d).
Hwa-Teng Lee et al., 2004 also stated that Ra and average white layer thickness tend to
increase at higher values of pulse current and ton. However, they found that for extended
pulse-on duration MRR, Ra and crack density all decrease.
(a) tav 22.1 µm: I= 6.5 Amp; ton =10µs
(b) tav 18.1 µm: I= 2.5 Amp; ton=10 µs

(c) tav 21.3 µm: I= 2.5 Amp; ton=10 µs
(d) tav 12.5 µm: I= 2.5 Amp; ton=1.5 µs

Fig. 15. Thickness of recast layer at different machining conditions
5. Conclusion
From the above discussions the following conclusions can be drawn:
1. Electrodes undergo more wear along its cross-section compared to that along its length.
2. Electrode wear increases with increase in current and voltage. Wear of copper
electrodes is less than that of brass electrodes. This is due to the higher thermal
conductivity and melting point of copper compared to those of brass.
3. During machining of mild steel, electrodes undergo more wear than during machining
of aluminum. This is due to the fact that thermal conductivity of aluminum is higher to
that of mild steel which causes comparatively more heat energy to dissipate into the
electrode during machining of mild steel.
4. Wear ratio decreases with increase in current, but decreases with increase in gap
voltage. The highest wear ratio was found during machining of aluminum using a
copper electrode.
5. MRR increases sharply with increase in current. In the present study, highest MRR was
obtained during machining of aluminum using a brass electrode.
6. Micro cracks are found on the machined surface. The tendency of formation of micro
cracks increases during EDM with a higher current and larger pulse-on time.
7. A recast layer was found on the machined surface which consists of the molten
materials from the workpiece and the electrode that could not be flushed away
completely by the dielectric fluid. A thicker layer of recast layer was formed on the
work surface machined with a higher current and pulse-on time.
6. Acknowledgement
The author of this work is indebted to the Research Management Center, International
Islamic University Malaysia (IIUM) for its continuous help during the research work. Also,
the author likes to appreciate the help of the staff and the technicians of the Department of
Manufacturing and Materials Engineering, International Islamic University Malaysia.
7. References
Bleys, P.; Kruth, P. & Lauwers, B. (2004). Sensing and compensation of tool wear in milling EDM.
Journal of Materials Processing Technology, vol.149, No.1-3, pp. 139-146, ISSN 0924-0136
Bulent, E.; Erman, A. & Abdulkadir E. (2006). A semi-empirical approach for residual
stresses in electric discharge machining (EDM). International Journal of Machine Tool
& Manufacture. Vol.46, pp. 858-865, ISSN 0890-6955
Dibitonto, D.; Eubank, T.; Patel, R. & Barrufet, A. (1989).Theoretical models of the electro
discharge machining process–a simple cathode erosion model. Journal of Applied
Physics, vol.69, pp. 4095-4103, ISSN 0021-8979
Ghosh, A. & Mallik, K. (1991). Electrical Discharge Machining, In: Manufacturing Science,
383-403, Affiliated East- West Press Private Limited, ISBN 81-85095-85-X, New
Delhi, India.
Ho, H. & Newman, T. (2003). State of the art electrical discharge machining (EDM).
International Journal of Machine Tools and Manufacture, vol.43, No.13, pp. 1287-1300,
ISSN 0890-6955
Hu, F. Zhou, C. & Bao, W. (2008). Material removal and surface damage in EDM of Ti3SiC2
ceramic. Ceramics International, vol.34, issue 3, pp. 537-541, ISSN 0272-8842
Hwa-Teng, L.; Fu-ChuanHsu & Tzu-Yao, T. (2004). Study of surface integrity using the small
area EDM process with a copper–tungsten electrode, Materials Science and
Engineering A, vol. 364, issues 1-2, pp. 346-356, ISSN 0025-5416
Kalpakjian, S. Schmid, R. (2001). Electrical-discharge machining, in: Manufacturing Engineering
and Technology, 6th edition, Prentice Hall, 769-774, Singapore
Khan, A. & Mridha, S. (2006). Performance of copper and aluminum electrode during EDM
of stainless steel and carbide. International Journal of Manufacturing and Production,
vol. 7, No.1, pp. 1-7, ISSN 0793-6648
Khanra, K.; Sarker, R.; Bhattacharya, B.; Pathak, C. & Godkhindi, M. (2007). Performance of
ZrB2–Cu composite as an EDM electrode. Journal of Materials Processing Technology,
vol.183, No.1, pp. 122-126, ISSN 0924-0136
Kunieda, M. & Kobayashi, T. (2004). Clarifying mechanism of determining tool electrode
wear ratio in EDM using spectroscopic measurement of vapor density. Journal of
Materials Processing Technology, vol.149, No. 1-3, pp. 284-288, ISSN 0924-0136
Lee, T. & Tai, Y. (2003). Relationship between EDM parameters and surface crack formation.
Journal of Materials Processing Technology, vol.142, issue 3, pp. 676-683, ISSN 0890-6955
Marafona, J. & Wykes. C. (2000). A new method of optimizing material removal rate using
EDM with copper-tungsten electrodes. International Journal of Machine Tools and
Manufacture, vol.40, pp. 153-164, ISSN 0890-6955
Marafona, J., & Chousal, G. (2006). A finite element model of EDM based on the Joule effect.
International Journal of Machine Tools & Manufacture, vol.46, pp. 595-602, ISSN 0890-6955
Pandey, C. & Jilani, T. (1986). Plasma channel growth and the resolidified layer in EDM.
Precision Engineering, Vol.8, issue 2, pp. 104-110, ISSN 0141-6359
Ozgedik, A. & Cogun, C. (2006). An experimental investigation of tool wear in electric

discharge machining. International Journal of Advance Manufacturing Technology.
Vol.27, pp. 488-500, ISSN 0268-3768
Patel, R.; Barrufet, A.; Eubank, T. & DiBitonto, D. (1989). Theoretical models of the electrical
discharge machining process-II: the anode model. Journal of Applied Physics, vol.66,
pp. 4104-4111, ISSN 0021-8979
Peter, Fonda.; Zhigang, Wang.; Kazuo, Yamazaki. & Yuji, A. (2007). A fundamental study on
Ti-6Al-4V’s thermal and electrical properties and their relation to EDM
productivity. Journal of Materials Processing Technology,
doi:10.1016/j.jmatprotec.2007.09.060, ISSN 0924-0136
Puertas, I.; Luis, J. & Alvarez, L. (2004). Analysis of the influence of EDM parameters on
surface quality, MRR and EW of WC-Co. Journal of Materials Processing Technology,
vol.153-154, No.10, pp. 1026-1032, ISSN 0924-0136
Ramasawmy, H. & Blunt, L. (2001). 3D surface characterization of electropolished EDMed
surface and quantitative assessment of process variables using Taguchi
Methodology. International Journal of Machine Tools and Manufacture, vol.42, pp.
1129-1133, ISSN 0890-6955
Salonitis, K.; Stournaras, A.; Stavropoulos, P. & Chryssolouris, G. (2007). Thermal modeling
of the material removal rate and surface roughness for die-sinking EDM.
International Journal of Advance Manufacturing Technology. DOI 10.1007/s00170-007-
1327-y, ISSN 0268-3768
Shuvra, D.; Mathias, K. & Klocke, F. (2003). EDM simulation: finite element-based
calculation of deformation, microstructure and residual stresses. Journal of Materials
Processing Technology, vol.142, pp. 434-451, ISSN 0924-0136
Tariq, S. & Pandey, C. (1984). Experimental investigation into the performance of water as
dielectric in EDM, International Journal of Machine Tool Design and Research, vol.24,
pp. 31-43, ISSN 0020-7357
Thomas, N.; Shreyes, M.; Thomas, W.; Rosa, T. & Laura, R. (2009). Investigation of the effect
of process parameters on the formation and characteristics of recast layer in wire-
EDM of Inconel 718. Materials Science and Engineering: A, vol.513-514, pp. 208-215,
ISSN 0025-5416
Vinod, Y.; Vijay, K. & Prakash, M. (2002). Thermal stresses due to electrical discharge
machining, International Journal of Machine Tools & Manufacture. Vol.42, pp. 877-888,
ISSN 0890-6955
Wang, J. & Tsai, M. (2001a) Semi-empirical model on work removal and tool wear in
electrical discharge machining. Journal of materials processing technology, vol.114,
No.4, pp. 1-17, ISSN 0924-0136
Wang, J. & Tsai, M. (2001b). Semi-emperical model of surface finish on electrical discharge
machining. International Journal of Machine Tools and Manufacture, vol.41, pp. 1455-
1477, ISSN 0890-6955
Yan, H.; Tsai, C. & Huang, Y. (2005). The effect of EDM of a dielectric of a urea solution in
water on modifying the surface of titanium. International Journal of Machine Tools
and Manufacture, vol. 45, No.2, pp. 194-200, ISSN 0890-6955
Zarepour, H.; Tehrani, A.; Karim, D & Amini, S. (2007). Statistical analysis on electrode wear
in EDM of tool steel DIN 1.2714 used in forging dies. Journal of Material Processing
Technology, vol.187-188, pp. 711-714, ISSN 0924-0136
Zaw, M.; Fuh, H.; Nee, C. & Lu, L. (1999). Fabrication of a new EDM electrode material
using sintering techniques. Journal of Materials Processing Technology, vol.89-90, pp.
182-186, ISSN 0924-0136
22
Thermal Treatment of Granulated

Particles by Induction Thermal Plasma
M. Mofazzal Hossain1 and Takayuki Watanabe2
1Department of Electronics and Communications Engineering, East West University
2Department of Environmental Chemistry and Engineering
Tokyo Institute of Technology,

1Bangladesh
2Japan
1. Introduction
After the invention of induction plasma torch by [Reed, 1961], tremendous achievements
have been earned by the researchers in the field of thermal treatment of micro particles by
induction plasma torch. Induction thermal plasma (ITP) has become very popular in material
processing due to several of its inherent characteristics: such as contamination free (no
electrode), high thermal gradient (between torch and reaction chamber), wide pressure range
and high enthalpy. ITP have extensively been used for the synthesis and surface treatment
of fine powders since couple of decades as a clean reactive heat source [Fan, 1997], [Watanabe,
2004]. ITP technology may ensure essentially the in-flight one-step melting, short melting
time, and less pollution compared with the traditional technologies that have been using in
the glass industries for the vitrification of granulated powders. Moreover ITP technology
may be very effective in the thermal treatment of porous micro particles and downsizing the
particle size. During in-flight treatment of particles, it is rear to have experimental records of
thermal history of particles; only some diagnosis of the quenched particles is possible for the
characterization. Thus, the numerical analysis is the only tool to have comprehensive
characterization of the particle thermal history and energy exchange during in-flight
treatment. Thus, for numerical investigation it is the challenge to predict the trajectory and
temperature history of the particles injected into the ITP torch. Among others Yoshida et al
[Yoshida, 1977] pioneered the modeling of particle heating in induction plasmas; though
their work assumed the particle trajectory along the centerline of the torch only. Boulos
[Boulos, 1978] developed a model and comprehensively discussed the thermal treatment of
alumina powders in the fire ball of argon induction plasma. Later (Proulx et al) [Proulx,
1985] predicted the trajectory and temperature history of alumina and copper particles
injected into ITP torch and discussed the particle loading effects in argon induction plasma.
In this chapter we shall discuss the in-flight thermal treatment mechanism of soda-lime-
silica glass powders by ITP and to optimize the plasma discharge parameters, particle size
and feed-rate of input powders that affect the quenched powders size, morphology, and
compositions. The thermal treatment of injected particles depends mainly on the plasma-
particle heat transfer efficiency, which in turn depends to a large extent on the trajectory and
temperature history of the injected particles. To achieve that goal, a plasma-particle
interaction model has been developed for argon-oxygen plasma, including a nozzle inserted
into the torch for the injection of carrier gas and soda-lime-silica glass powders. This model
can be used to demonstrate the particle loading effects and to optimize the parameters that
govern the particles trajectory, temperature history, quenched particles size and plasma-
particle energy exchange efficiency. This model may be used to optimize the plasma and
particle parameters for any combination of plasma gases for example argon-oxygen or argon
nitrogen etc.
2. Modeling
2.1 Plasma model
The schematic geometry of the ITP torch is presented in Fig.1. The torch dimensions and
discharge conditions are tabulated in Table 1. The overall efficiency of the reactor is assumed
to be 50%, thus, plasma power is set to 10 kW. The torch dimension, power and induction
frequency may vary and can be optimized through the simulation. The model solves the
conservation equations and vector potential form of Maxwell’s equations simultaneously
under LTE (local thermodynamic equilibrium) conditions, including a metal nozzle inserted
into the torch. It is assumed that plasma flow is 2-dimensioanl, axi-symmetric, laminar,
steady, optically thin, and electromagnetic fields are 2-dimensional. Adding the source terms
to the conservation equations, the plasma-particle interaction and particle loading effects
have been taken into account. In this model, the conservation equations are as follows:
Mass conservation:
∇ ⋅ ρ u = SCp (1)
Momentum conservation:
ρ u ⋅ ∇u = −∇p + ∇ ⋅ μ∇u + J × B + SpM (2)
Distance to initial coil position (L1) 19 mm

Length of injection tube (Lt) 52 mm
Distance to end of coil position (L2) 65 mm
Torch length (L3) 190 mm
Coil diameter (dc) 5 mm
Wall thickness of quartz tube (Twall) 1.5 mm
Inner radius of injection tube (r1) 1 mm
Outer radius of injection tube (rt) 4.5 mm
Outer radius of inner slot (r2) 6.5 mm
Inner radius of outer slot (r3) 21.5 mm
Torch radius (r0) 22.5 mm
Coil radius (rc) 32 mm
Plasma power 10 kW
Working frequency 4 MHz
Working pressure 0.1 MPa
Flow rate of carrier gas (Q1) 4 ∼ 9 L/min of Argon
Flow rate of plasma gas (Q2) 2 L/min of Argon
Flow rate of sheath gas (Q3) 22 L/min Argon & 2 L/min Oxygen
Table 1. Torch dimensions & discharge conditions
Thermal Treatment of Granulated Particles by Induction Thermal Plasma 439
Energy conservation:
⎛ κ ⎞
ρ u ⋅ ∇h = ∇ ⋅ ⎜ ∇h ⎟ + J ⋅ E − Qr − SpE (3)
⎜ Cp ⎟
⎝ ⎠
Species conservation:
( )
ρ u ⋅ ∇y = ∇ ⋅ ρ Dm∇y + SCp (4)
Vector potential form of Maxwell electromagnetic field equation [Mostaghimi, 1998]:
∇ 2 A c = iμ0σωA c (5)
Q1 Q2 Q3
Twall
L1
Lt
L2
r1 dc
rt
r2
r3
r0 L3
rc
Fig. 1. Schematic geometry of induction thermal plasma torch

2.1.1 Boundary conditions

The boundary conditions for the mass, momentum, energy and species conservation
equations are: at the inlet, gas temperature was set to 300 K and uniform velocity profiles
are assumed based on the given flow rates; on the axis of symmetry, the symmetry
conditions are imposed; on the walls, no-slip condition is assumed; the outer wall
temperature is set to 350 K; and, at the exit, axial gradients of all fields are set equal to zero.
The inserted nozzle is assumed to be water cooled at 300 K. On the nozzle wall, the velocity
is set to zero. The boundary conditions for the vector potential form of Maxwell’s equation
are the same as those described in reference [Mostaghimi, 1998].
2.1.2 Computational procedure and thermophysical properties

The conservation equations, which are listed in previous section, are solved numerically
using the SIMPLER algorithm of Patankar [Patankar, 1980]. The algorithm is based on a
control-volume finite-difference scheme for solving the transport equations of
incompressible fluids. Calculations are performed for a 44 (in radial direction) by 93 (in axial
direction) non-uniform grid system.
Thermodynamic and transport properties of argon and oxygen gases required for the
simulation are mass density, specific heat at constant pressure, viscosity, electrical and
thermal conductivity and radiative loss coefficient. The transport properties, which are
function of temperature, are calculated under LTE conditions using Chapman-Enskog first
approximation to Boltzmann equation [Tanaka, 2000]. The effective diffusion coefficient of
species is calculated based on the following equations:
Dim =
( 1 − yi ) (6)
υ
x
∑ Di
j ≠ i , j = 1 ij
Dij = 2.628 × 10 −2
( Mi + M j ) T 1.5
(7)
Ω(ij )
2 Mi M j 1.1
The ambipolar diffusion coefficient for ions can be approximated as Da=Dion (1+Te/Tion). As
the thermal equilibrium condition i.e Th=Te=Tion was applied thus, Da ≅ 2Dion.
2.2 Particle model

The following assumptions are made in the analysis of plasma-particle interactions in the
ITP torch; the particle motion is two-dimensional, only the viscous drag force and gravity
affect the motion of an injected particle, the temperature gradient inside the particle is
neglected, and the particle charging effect caused by the impacts of electrons or positive ions
is negligible. The particle charging effects have not been intensively studied yet. However,
the electromagnetic drag forces caused by the particle charging of the injected particles are
negligible compared with those by neutrals and charged particles due to negligible electrical
conductivity of soda-lime-silica powders. Thus, the momentum equations for a single
spherical particle injected vertically downward into the plasma torch can be expressed as
follows:
dup ⎛ ρ ⎞
3
(
= − C D up − u U R ⎜
⎜
⎟+ g
⎟ ) (8)
dt 4 ⎝ ρ p dp ⎠
dν p ⎛ ρ ⎞
dt
3
4
(
= − CD ν p − ν U R ⎜ ⎟
⎜ ρ p dp ⎟ ) (9)
⎝ ⎠
( up − u ) + (ν p −ν )
2 2
UR = (10)
The particle temperature, liquid fraction and diameter are predicted according to the
following energy balances:
( )
Q = π dp2 hc T − Tp − π dp2σ sε Tp4 − Ta4 ( ) (11)
dTp 6Q
= for Tp < Tb (12)
dt πρ p dp3C pp
dχ 6Q
= for 1000 ≤ Tp ≤ 1600 (13)
dt πρ p dp3 H m
ddp 2Q
= for 1000 ≤ Tp ≤ 1600, Tp ≥ Tb (14)
dt πρ p dp2 Hν
Drag coefficient CDf is calculated using Eq. (15) and the property variation at the particle
surface layer and the non-continuum effects are taken into account by Eq. (16) and (17)
[Chen, 1983].
⎧ 24
⎪R Re ≤ 0.2
⎪ e
⎪ 24 ⎛ 3 ⎞
⎪ ⎜ 1 + Re ⎟ 0.2 < Re ≤ 0.2
R
⎪ ⎝ 16 ⎠
C Df =⎨ e (15)
⎪ Re (
⎪ 24 1 + 0.11R0.81
e ) 2.0 < Re ≤ 21.0
⎪
⎪R
⎩ e
(
⎪ 24 1 + 0.189 R 0.62
e ) 21.0 < Re ≤ 200
−0.45
⎛ρ μ ⎞
f1 = ⎜ ∞ ∞ ⎟ (16)
⎝ ρ s μs ⎠
−0.45
⎪⎧ ⎛ 2 − α ⎞ ⎛ γ ⎞ 4 ⎫⎪
f 2 = ⎨1 + ⎜ ⎟⎜ ⎟ Kn ⎬ , 10 − 2 < Kn < 0.1 (17)
⎩⎪ ⎝ α ⎠ ⎝ 1 + γ Pr
⎠ s ⎭⎪
C D = C Df f 1 f 2 (18)
To take into account the steep temperature gradient between plasma and particle surface,
the Nusselt correlation can be expressed by Eq. (19) [Lee, 1985]. The non-continuum effect is
taken into account by Eq. (20) [Chen, 1983].
0.6 0.38
1/3 ⎛ ρ ∞ μ∞ ⎞
⎛ C p∞ ⎞
(
Nuf = 2.0 + 0.6 Re1/2
f Prf ⎜ ) ⎟
⎝ ρ s μs ⎠
⎜ ⎟
⎜ C ps ⎟
(19)
⎝ ⎠
−1
⎧⎪ ⎛ 2 − α ⎞ ⎛ γ ⎞ 4 ⎫⎪
f 3 = ⎨1 + ⎜ ⎟⎜ ⎟ Kn ⎬ , 10 − 3 < Kn < 0.1 (20)
⎩⎪ ⎝ α ⎠ ⎝ 1 + γ ⎠ Prs ⎭⎪
The convective heat transfer coefficient is predicted as follows:
κf
hc = Nuf f 3 (21)
dp
2.2.1 Particle source terms

Let us assume Nt0 be the total number of particles injected per unit time, nd is the particle
size distribution, and nr is the fraction of Nt0 injected at each point through the injection
nozzle. Thus, the total number of particles per unit time traveling along the trajectory (l, k)
corresponding to a particle diameter dl injected at the inlet point rk is:
N ( l , k ) = ndl nrk N t0 (22)
For the sake of computation, the particle concentration nr in the inlet is assumed to be
uniform and to be separated into five injection points, which are at radial positions of 0.3,
0.45, 0.6, 0.75 and 0.9 mm. In the present computation the particles diameter distribution is
assumed to be Maxwellian (similar to experiment). The particle size and corresponding
distribution fraction are presented in Table 2. In the present computation, the powder is
assumed to be composed of seven size particles according to its diameter and deviation. The
average particle diameter is 58 μm and the maximum deviation is 67%. As a result, there are
35 different possible trajectories of the injected particles. The injection velocity of the
particles is assumed to be equal to the injection velocity of carrier gas.
Particle diameter (μm) 20 40 50 68 77 80 90

Fraction 0.03 0.07 0.1 0.6 0.1 0.07 0.03
Table 2. Particle size and corresponding distribution fraction
To take into account the particles loading effects, particles source terms for the mass,
momentum, energy and species conservation equations have been calculated in the same
fashion as described in reference [Proulx, 1985], using the Particle-Source-In Cell (PSI-CELL)
approach [Crowe,1977], where the particles are regarded as sources of mass, momentum
and energy. The source terms in the mass and species conservation equation, SCp is the net
efflux rate of particles mass in a computational cell (control volume). Assuming the particles
are spherical, the efflux rate of particle mass for the particle trajectory (l, k) that traverses a
given cell (i, j) is:
1
(
SCp ,(ijl , k ) = πρ p N ij( l , k ) dij3 ,in − dij3 , out
6
) (23)
The net efflux rate of particle mass is obtained by summing over all particles trajectories
which traverse a given cell (i, j):
SCp,ij = ∑∑ SCp ,(ijl , k ) (24)

l k
The source terms for momentum conservation equations are evaluated in the same fashion
as that of mass conservation equation. In this case, the efflux rate of particles momentum for
the particle trajectory (l, k) traversing a given cell (i, j) is:
1
(
SpM,ijz ( l , k ) = πρ p N ij( l , k ) uij , in dij3 ,in − uij , out dij3 , out
6
) (25)
1
(
SpM,ijr ( l , k ) = πρ p N ij( l , k ) ν ij , in dij3 , in − ν ij , out dij3 , out
6
) (26)
Thus, the corresponding source terms for axial and radial momentum conservation
equations are:
SpM,ijz = ∑∑ SpM,ijz ( l , k ) (27)

l k
SpM,ijr = ∑∑ SpM,ijr ( l , k ) (28)

l k
The source term for energy conservation equation SpE,ij consists of the heat given to the
particles Q(pl,,ijk ) , and superheat to bring the particle vapors into thermal equilibrium with the
plasma Qν( l,,ijk ) :
( )
τ out
Q(pl,,ijk ) = ∫ π dp2 hc Tij − Tp( l,ij, k ) dt (29)
τ in
π ⎛ ddp ⎞
Qν( l,,ijk ) = ∫
τ in
τ out
2
π dp2 ρ p ⎜⎜
⎝ dt
((l ,k )
⎟⎟ C pν Tij − Tp ,ij dt
⎠
) (30)
l k
(
SpE,ij = ∑∑ N ij( l , k ) Q(pl,,ijk ) + Qν( l,,ijk ) ) (31)
The calculation is started by solving the plasma temperature and flow fields without
injection of any particles. Using these conversed temperature and flow fields, particles
trajectories together with particle temperature and size histories are calculated. The particle
source terms for the mass, momentum and energy conservation equations for each control
volume throughout the torch are then predicted. The plasma temperature and flow fields
are predicted again incorporating these particle source terms. The new plasma temperature
and flow fields are used to recalculate the particles trajectories, temperature and size
histories. Calculating the new source terms and incorporating them into conservation
equations constitute the effects of plasma-particle interaction, thereby completing the cycle
of mutual interaction. The above computation schemes are repeated until convergence. The
physical properties of soda-lime-silica glass powders used in the present investigation are
listed in Table 3.
Mass density 2300 kg/m3

Specific heat at constant pressure 800 J/kg-K
Porosity 80%
Fusion temperature 1000~1600 K
Boiling temperature 2500 K
Latent heat of fusion 3.69×105 J/kg
Latent heat of vaporization 1.248×107 J/kg
Table 3. Physical properties of soda-lime-silica glass powders
3. Simulated results
The calculation has been carrier out for a plasma power of 10 kW, reactor pressure 0.1 MPa
and induction frequency 4 MHz. The discharge conditions are tabulated in Table 1. In this
study, attention is given to the plasma-particle interaction effects on individual particle
trajectory, velocity, and temperature history along the trajectories for different carrier gas
flow-rate and powder feed-rates. Attention also paid to investigate how the plasma-particle
energy exchange process is affected by the particle loading effects. Two aspects of the
thermal treatment are investigated: the behavior of the individual particles, and the global
effects of the particles on the plasma fields. The carrier gas flow-rate is very vital in
determining the individual particle trajectories, and the allowable powder feed-rate. Figure
2 shows the isotherms in the torch for a carrier gas flow-rate of 6 L/min argon and various
powder feed-rates. The other discharge conditions are the same as presented in Table 1. A
comparison among the isotherms clearly reveals the intense cooling around the torch
centerline that increases with powder feed-rate. However, the plasma temperature away
from the centerline of the torch remains almost unaffected by higher powder feed-rates. This
is because the
individual particle trajectories are not widely outbound in the radial direction; rather the
trajectories are very close to the torch axis. Thus, the plasma-particle interaction around the
centerline is very crucial at higher powder feed-rate. The same kind of arguments is
proposed by Ye et al [Ye, 2000] to explain the particle trajectories for alumina and tungsten
particles. The effects of carrier gas flow-rates on the individual particle trajectories are
presented in Fig. 3, for the particle diameter of 50 μm and a feed-rate of 5 g/min. It is
comprehended that the higher flow-rate of carrier gas enhances the axial velocity of the
particles, because the initial axial velocity of the particles depends on carrier gas flow rate;
as a result the trajectories become closer to the torch axis at higher flow-rate. The individual
particle temperature history along the trajectory is also influenced by the carrier gas flow-
rate and powder feed-rate. Figure 4 shows the effects of carrier gas flow-rate on the particle
temperature for a feed-rate of 5 g/min. It is found that the particle temperature along the
trajectory decreases at higher carrier gas flow-rate. The main reason is the cooling of plasma
at Fig. 2 Effects of powder loading on the isotherms for a carrier gas flow-rate of 6 L/min
higher carrier gas flow-rate that leads less heat transfer to particles. Figure 5 describes the
effects of powder feed-rate on the particle temperature along the trajectory. Like the flow-
rate of carrier gas, the higher feed-rate of powder also causes intense cooling of plasma;
thus, the heat transfer to particles decreases what results lower particle temperature. At this
stage of investigation, it is indeed necessary to discuss the energy transfer mechanism to
particles. The energy transfer is affected by the particles physical properties, plasma
temperature, and velocity. The last two parameters are affected to a large extent by the
carrier gas flow-rate and powder loading. The net energy transfer to particles is calculated
by integrating the energy transfer rate to the particles injected per unit time over the
residence time for all the particle trajectories. Mathematically the net energy transfer to
particles (Qnet) can be expressed as follows:
t = ts
∫ {π dp hc (T − Tp ) − π dpσ sε (Tp )}
2 2 4
Qt = − Ta4 dt (32)
t =0
Qnet = ∑∑ N ( l , k )Qt (33)

l k
Feed-rate: 0 Feed-rate: 5 Feed-rate: 10

0
20
40
60
Axial position [mm]
80
11kK
100 10kK
9kK
8kK
120
7kK
6kK
140 5kK
4kK
3kK
160
2kK
1kK
180 0K
0 10 20 0 10 20 0 10 20
Radius [mm] Radius [mm] Radius [mm]
Fig. 2. Effects of powder loading on the isotherms for a carrier gas flow-rate of 6 L/min
3.0 Particle diameter: 50 μm

Powder feed-rate: 5 g/min
Radial distance from centerline [mm]
2.5
Carrier gas flow-rate: 4
2.0 Carrier gas flow-rate: 6
1.5
1.0
0.5
0.0
0 20 40 60 80 100 120 140 160 180
Axial distance from top of the torch [mm]
Fig. 3. Effects of carrier gas flow-rate on the particle trajectories for a powder feed-rate of
5 g/min
Particle diameter: 50 μm
3000
Powder feed-rate: 5 g/min
2500
Particle temperature [K]
2000 Carrier: 4
Carrier: 6
1500 Carrier: 7
Carrier: 9
1000
500
0
0 20 40 60 80 100 120 140 160 180
Axial distance [mm]
Fig. 4. Dependence of particle temperature history along the trajectory on carrier gas flow-
rate
3000 Particle diameter: 50 μm

Carrier flow-rate: 6 lpm
2500
Particle temperature [K]
2000
Feed-rate: 5
1500 Feed-rate: 10
1000
500
0
0 20 40 60 80 100 120 140 160 180
Axial distance [mm]
Fig. 5. Dependence of particle temperature history along the trajectory on the powder feed-
rate
Figure 6 clearly presents how the net energy transfer to particles is affected by the carrier
gas flow-rate under powder loading conditions. Only 5 g/min of powder feeding decreases
Average diameter: 58 μm
5 Powder feed-rate: 5 g/min
4
No loading effect
Energy transfer [%]
With loading effect

3
0
3 4 5 6 7 8 9 10
Carrier gas flow-rate [lpm]
Fig. 6. Effects of powder loading and carrier gas flow-rate on the plasma-particle energy
transfer
Carrier gas flow-rate: 9 lpm

300
Without loading effect
250 With loading effect
Energy transfer [W]
200
150
100
50
4 6 8 10 12 14 16 18 20 22
Powder feed-rate [g/min]
Fig. 7. Particle loading effects on plasma-particle energy transfer at various powder feed-rate
the energy transfer to particle by about 44%. The powder loading effect and the dependence
of energy transfer to particles on the powder feed-rate is presented in Fig. 7, for a carrier gas
flow-rate of 9 L/min. It can be noticed that energy transfer to particles increases linearly
with feed-rate in the absence of particle loading effect; however, when particle loading effect
is taken into account, energy transfer to particles yet increases with feed-rate but with a
declined slop. The main reason is the intense local cooling of plasma around the torch
centerline under dense particle loading. It is also evident that the particle loading effect is
pronounced at higher powder feed-rate.
4. Experimental
4.1 Setup
The experimental setup consists of a plasma torch (Fig. 1), a reaction chamber, powder
feeder, and a power supply unit (4 MHz, 20 kW). The plasma torch consists of a water-
cooled co-axial quartz tube surrounded by a three-turn induction coil. The granulated soda-
lime-silica glass powders are prepared by spray-drying method from the reagents of
Na2CO3, CaCO3 and SiO2 with the composition of Na2O:16, CaO:10 and SiO2:74 in wt%. The
mean diameter and porosity of soda-lime-silica glass powders are 58 μm and 80%,
respectively. The plasma discharge conditions are the same as those described in Table 1 in
the modeling section. The soda-lime-silica glass powders are injected into ITP torch along
with the carrier gas at a rate of 5-20 g/min and the quenched powders are collected on a
water-cooled ceramic block at 340 mm from the nozzle exit.
4.2 Characterization of plasma-treated particles

The treatment quality of the powders is characterized by the vitrification degree, the surface
morphology, cross-sectional structure and composition of the quenched powders. The
105 Powder feed-rate: 5 g/min 50
Evaporation of Na2O [%]

Vitrification degree [%]
40
100
30
95 20
10
90 2 3 4 5 6 7
Carrier gas flow-rate [lpm]
(a)
105 Carrier gas flow-rate: 6 lpm 50
Evaporation of Na2O [wt%]

40
Vitrification degree [%]
100
30
95 20
10
90
4 6 8 10 12 14 16 18 20 22
Powder feed-rate [g/min]
(b)
Fig. 8. Effects of carrier gas flow-rate (a), and powder feed-rate (b) on the vitrification degree
and Na2O evaporation rate
(a)
(b)
Fig. 9. The SEM photographs of soda-lime-silica glass powders before (a), and after (b)
thermal treatment
vitrification degree is defined as the ratio of the converted crystalline phases of SiO2 in the
quenched powder to the crystalline SiO2 in the raw powders. The vitrification degree of
quenched powders is quantitatively determined by X-ray diffractometry (XRD) on Miniflex
(Rigaku) with Cu Kα radiation at 30 kV and 15 mA. The data are collected in the 2θ range
3-90º with a step size of 0.02º and a scan speed of 4º/min. The quenched powders collected
at the reaction chamber are examined by scanning electron microscopy (SEM) on JSM5310
(JEOL) to observe their surface morphologies and cross-sectional microstructures. The
composition of quenched powders is analyzed by inductively coupled plasma (ICP) on
ICP-8100 (SHIMADZU).

In the experiment soda-lime-silica glass powders are injected along with the carrier gas
through the nozzle inserted into the plasma torch. Thus, the initial particle velocity is the
same as that of carrier gas. When the particles come in contact to high temperature plasma
flame, they are heated and their temperature starts to rise. As the particle temperature
reaches to its melting temperature, particle porosity decreases drastically; as a result,
particle diameter shrinks. When particle temperature reaches to its boiling temperature,
vaporization takes place and particle diameter shrinks further. In order to investigate the
effects of carrier gas flow-rate and powder feed-rate on energy transfer to particles, the
vitrification degree and Na2O evaporation rate are estimated through XRD and ICP
spectrum analysis. Higher vitrification degree and Na2O evaporation rate indirectly indicate
the large energy transfer to particles. Figure 8 shows the XRD and ICP spectrum analysis
results. It can be noticed that both the evaporation rate of Na2O and the vitrification degree
decrease with the increase of both carrier gas flow-rate [Fig. 8(a)] and powder feed-rate
[Fig. 8(b)]. The vitrification degree and the evaporation of Na2O depend to a large extent on
the particle temperature. Higher carrier gas flow-rate and powder feed-rate cause lower
plasma temperature which causes less heat transfer to particles; as a result particle
temperature is lower. It is important to underline that after the thermal treatment, the size,
composition, and morphology of the particles have been changed significantly. The effects
of thermal treatment are visualized in the SEM photograph as shown in Fig. 9. From Fig. 9
(a) and (b), it can be noticed that after treatment, the particle size becomes smaller, quite
spherical, smoother and compact surface.
5. Discussions
To validate the modeling and simulated results, a comparative discussion between
simulated and experimental results are indeed necessary. From the experimental results it is
found that at higher carrier gas flow-rate and powder feed-rate, both the evaporation rate of
Na2O and the vitrification degree decrease. These results indicate that less heat transfer to
particles takes place at higher carrier gas flow-rate and powder feed-rate. From the
simulated results it is evident that at increased carrier gas flow-rate and powder feed-rate,
the energy transfer to particles decreases; as a result, particles temperature becomes lower. It
is convinced that the main reason of less heat transfer to particles is the severe local cooling
of plasma around the torch centerline at higher carrier gas flow-rate and powder feed-rate.
Thus, it may be argued that the simulated results well agree with the experimental findings.
6. Conclusions
In this chapter we basically, discussed the way of thermal treatment of any type of
granulated porous particles by induction thermal plasma. A general plasma-particle
interactive flow model has been discussed using what it is possible to simulate the particle
trajectories, temperature histories, plasma temperature contours etc. The described model
can be used to optimize the carrier gas flow-rate, particle size, and powder feed-rate to
achieve the maximum treatment efficiency during thermal treatment of granulated powders
by argon-oxygen induction thermal plasmas. Numerically, it is found that the heat transfer
to particles decreases at increased carrier gas flow-rate and powder feed-rate, and these
results well agree with those of experiment. Thus, it can be concluded that, efficient thermal
treatment of particles depends not only on the physical properties of the particles, but also
on the plasma discharge conditions and particle parameters. Therefore, for a particular type
of powder (certain physical properties) both carrier gas flow-rate and powder feed-rate
mainly govern the treatment quality.
7. Nomenclature
Ac Complex amplitude of vector potential
B Magnetic field vector
Cp Specific heat at constant pressure
Cpp Particle specific heat at constant pressure
dp Particle diameter
Dij Binary diffusion coefficient between species i and j
Dm Multicomponent diffusion coefficient
E Electric field vector
f Frequency
g Acceleration of gravity
h Enthalpy
hc Heat transfer coefficient
Hm Latent heat of melting
Hv Latent heat of vaporization
i Complex vector ( −1 )
J Current density vector
Kn Knudsen number
Mi Molecular weight of species i
Mj Molecular weight of species of j
N t0 Total number of particles injected per unit time
nd Particle size distribution
nr Fraction of Nt0 injected at each point
Nu Nusselt number
p Pressure
Pr Prandtl number
Q Net heat exchange between the particle and its surroundings
Qr Volumetric radiation loss
Re Reynold number
SCp Particle source term in continuity equation
SpM Particle source term in momentum equation
SpE Particle source term in energy equation
t Time
T Plasma temperature
Ta Ambient temperature
Tb Boiling point temperature of particles
Te Electron temperature
Th Heavy particle temperature
Tion Ion temperature
Tp Particle temperature
ts Residence time of particle in the plasma
u Velocity vector
up Axial velocity component of particle
UR Relative speed of particles with respect to plasma
vp Radial velocity component of particle
xi Mole fraction of species i
y Mass fraction
Greek symbols
Vector operator
κ Thermal conductivity
Mass density
Viscosity
σ Electrical conductivity
κ Thermal conductivity
μ0 Permeability of free space
ω Angular frequency (2πf)
Ω(1.1)
ij Collision integral between species i and j
ε Particle porosity
σs Stefan-Boltzmann constant
Liquid mass fraction of a particle
Thermal accommodation coefficient
γ Specific heat ratio
Subscripts
f Properties corresponding to film temperature
p Particle
s Properties corresponding to particle temperature
∞ Properties corresponding to plasma temperature
(i, j) Location of a control volume or cell
Superscripts
(l, k) Particles having an initial diameter dl, and injection point rk.
8. References
Boulos, M. I, 1978, “Heating of powders in the fire ball of an induction plasma,” IEEE Trans.
on Plasma Sci. PS-6, pp. 93-106.
Crowe, C. T., Sharma, M. P., and Stock, D. E., 1977, “The particle-source-in cell (PSI-CELL)
model for gas-droplet flows,” J. Fluid Eng., 99, pp. 325-332.
Chen, X., and Pfender, E., 1983, “Effects of Knudsen number on heat transfer to a particle
immersed into a thermal plasma,” Plasma Chem. Plasma Process., 3, pp. 97-113.
Fan, X., Ishigaki, T., and Sato, Y., 1997, “Phase formation in molybdenum disilicide powders
during in-flight induction plasma treatment,” J. Mater. Res., 12, pp. 1315-1326.
Mostaghimi, J., Paul, K. C., and Sakuta, T., 1998, “Transient response of the radio frequency
inductively coupled plasma to a sudden change in power,” J. Appl. Phys., 83, pp.
1898-1908.
Lee, Y. C., Chyou, Y. P., and Pfender, E., 1985, “Particle dynamics and particle heat and mass
transfer in thermal plasmas. Part II. Particle heat and mass transfer in thermal
plasmas,” Plasma Chem. Plasma Process., 5, pp. 391-414.
Patankar, S. V., 1980, Numerical fluid flow and heat transfer, Hemisphere, New York.
Proulx, P., Mostaghimi, J., and Boulos, M. I., 1985, “Plasma-particle interaction effects in
induction plasma modeling under dense loading conditions,” Int. J. Heat Mass
Transfer, 28, pp. 1327-1335.
Reed T. B. “Induction-Coupled Plasma Torch,” Journal of Applied Physics, 1961, 32(5), p.821
Tanaka, Y., Paul, K. C., and Sakuta, T., 2000, “Thermodynamic and transport properties of
N2/O2 mixtures at different admixture ratio,” Trans. IEE Japan, 120-B, pp. 24-30.
Watanabe, T., and Fujiwara, K., 2004, “Nucleation and growth of oxide nanoparticles
prepared by induction thermal plasmas,” Chem. Eng. Comm., 191, pp. 1343-1361.
Yoshida, T., and Akashi, K., 1977, “Particle heating in a radio-frequency plasma torch,”
J. Appl. Phys., 48, pp. 2252-2260.
Ye, R., Proulx, P., and Boulos, M. I., 2000, “Particle turbulent dispersion and loading effects
in an inductively coupled radio frequency plasma,” J. Phys. D: Appl. Phys., 33, pp.
2154-2162.
23
Method for Measurement of

Single-Injector Heat Transfer Characteristics
and Its Application in Studying
Gas-Gas Injector Combustion Chamber
Guo-biao Cai, Xiao-wei Wang and Tao Chen
School of Astronautics, Beijing University of Aeronautics and Astronautics
P. R. China
1. Introduction
In the development of a Liquid Propellant Rocket Engine (LPRE), injector is always the
element which requires the longest development period. An injector always especially in the
initial design phase needs hundreds and thousands of tests to get a choice. These numerous
tests, in turn, require reliable and accurate measurement method to give the basic support.
The strength, the life cycle, and the cooling system effectiveness are highly dependent on
heat transfer into and out of the system (Tramecourt et al., 2005). Heat transfer characteristics,
combined with combustion efficiency, and combustion instability, are the three key
parameters needed to be investigated in developing a new injector in a combustion
chamber. The heat transfer characteristics generally contain the temperature and heat flux
on the hot-gas-wall in the combustion chamber. The combustion efficiency and combustion
instability can be observed easily by measuring the chamber pressure. However, it is always
very hard to measure the temperature and heat flux on hot-gas-wall of the chamber in the
hot-tests due to the extreme environment in LPRE chambers (temperature > 3000K, pressure
> 10atm).
Additionally, considerable efforts have been dedicated to model combustion in combustion
chambers to understand and predict the heat transfer to the chamber walls (Zurbach, 2006),
whereas the validation of computational fluid dynamics (CFD) design tools requires reliable
experimental data assessment. This acquires a comprehensive set of data in the same facility,
including wall heat fluxes along with inflow measurements (Tucker et al., 2005). To obtain
the exact wall heat flux data, a reliable measurement method, in turn, is also required.
Though a number of efforts had contributed to design new injectors and study on heat
transfer of injectors, studies focusing on the measurement method are less well documented.
In recent years, Pennsylvania State University, University of Florida and NASA MSFC etc.
made many attempts on studying the heat transfer characteristics of some injectors in heat
sink chambers (Conley et al., 2007; Jones et al., 2006; Marshall et al., 2005; Santoro & Pal,
2005; Vaidyanathan et al., 2007, 2010b). They applied coaxial thermocouple to measure two
point temperatures at the same axial location and difference radial locations in a heat-sink
copper combustion chamber, and then used the temperatures to calculate the local heat flux
through a simple radial heat conduction analytical calculation. Coy summarized and
analyzed the inverse heat transfer methods, and described a method for resolving inverse
heat conduction problems using approximating polynomials (Coy, 2010), in which two
sensors are also needed in any axial measurement location. All of previous method tried to
get the exact time-dependent data of heat flux and surface temperature based on analytical
solutions. Though these measurement methods can get the exact real-time heat flux data if
right calculation methods are used, they always need two measurement points at the any
axial measurement location, which increases the failure probability twice at any axial
measurement location, and the current most popular method always put one measurement
point right on the inner wall, and the thermocouple plug must be chamber-side geometry
contoured to match the internal radius of curvature of the chamber, otherwise it will
influence the inner flowfield and could not get the right temperature data. A little influence
on the inner flowfield may induce a large deviation of the temperature on the measurement
point. The potential influence of the thermocouple plug on the inner flowfield always exists,
because the separated plug was put on the inner wall. In summary, all of the previous
methods tried to get the exact time-dependent data with sacrificing the success probability
and the measurement systems are complicated. In addition, they solved the axisymmetric
problems with 1-D assumptions or solved 3-D problem with 2-D assumption.
This study introduces a simple method to obtain the inner wall temperature and heat flux in
a heat sink combustion chamber. This method uses numerical calculation method to get the
ultimate time-dependent inner wall temperature and heat flux with numerical calculation
method. It considers the inner wall heat transfer coefficient as unchanged, which suffers a
little change in hot-test, but the change is always less than 5%. Only single temperature
measurement point at one axial location is just needed. In current method the measurement
points are not on the inner wall surface, thus the potential influence on the inner flowfield
doesn’t exist anymore and there is not any sealing problem which also may cause severe
measurement error. A reliable 2-D axisymmetric numerical calculation was applied to get
the ultimate data. The measurement method, the related data processing and error analysis
are demonstrated in detail with a specific hot-testing example of a gaseous
hydrogen/gaseous oxygen single-element heat sink combustion chamber using this method.
This method can be a feasible alternate one to get the inner wall temperature and heat flux,
and it was originally developed for single-element axisymmetric chamber, and also can be a
reference for non-axisymmetric chambers and multi-element injector chambers.
Furthermore, this method was used to investigate the heat transfer characteristics of a shear-
coaxial single-element gas-gas injector combustion chamber. The full flow stage combustion
(FFSC) cycle engine is preferable because of its high performance and high reliability, for
which gas-gas injector technology is a key technology for this kind of engine and has
received extensive studies (Archambault et al., 2002a, 2002b; Davis & Compbell, 1997;
Farhangi et al., 1999; Foust et al., 1996; Meyer et al., 1996; Schley et al., 1997; Tucker et al.,
1997; Santoro et al 2005; Tucker, 2007a ,2008b; Lin et al., 2005; Cai et al., 2008; Wang et al,
2009a, 2010b, 2010c etc.). Also, gas-gas injectors have been widely used on other engines and
combustion devices (Calhoon, 1973; Groot, 1997). In this sense, it is of great value to
investigate the heat transfer characteristics of a gas-gas injector combustion chamber. A
single-injector heat-sink chamber was designed and hot-fire tested for 17 times at chamber
pressure from 0.92MPa to 6.1MPa. Inner hot-gas-wall temperature and heat flux along with
the axial direction of the chamber were obtained. The heat flux results were compared with
each other qualitatively and quantitatively. The inner combustion flows were also numerically
Method for Measurement of Single-Injector Heat Transfer
Characteristics and Its Application in Studying Gas-Gas Injector Combustion Chamber 457
simulated with multi-species turbulence N-S equations at higher chamber pressure from
5MPa to 20MPa to extend the experimental results. Both the combustion flow structures and
heat flux profiles on inner wall were obtained and discussed.
2. Measurement method
The heat-sink chamber is preferable (Calhoon, 1973; Marshall et al., 2005; Santoro et al 2005;
Conley et al., 2007; Jones et al., 2006) in the initial design phase of a new injector or in the
study on the heat transfer characteristics of injector owing to its simple structure, low cost
and easy manufacture. Thus, heat-sink combustion chamber was selected in this study. And
the temperature measurement scheme is shown in Fig. 1.
T1 T2 T3 T4 T n -2 T n - 1 T n
Fig. 1. Temperature measurement scheme in the heat sink chamber

This method applies thermocouples to measure the temperatures of a series of points near
the inner wall along the axial direction in the heat-sink chamber. The distances between
these points and inner wall surface are the same. The basic theory of this method is stated
this way: with the development of combustion flow in the chamber, different heat fluxes are
produced on the inner wall axially, further resulting in different temperatures on these
measurement points. Therefore, the temperatures of these points can be utilized to obtain
the heat flux and temperature on the inner wall.
In view that a heat-sink chamber can not undertake a long hot test, the tests are always
completed within several seconds in which, fortunately, enough effective temperature
output curves can be produced. In light of the temperatures are measured at those points
near the inner surface of the chamber, the temperature profile along measurement point can
clearly reflect the inner surface heat flux profile. The longitudinal heat transfer effect will be
considered in the heat flux calculation.
This measurement method was employed in a single-element gaseous hydrogen/gaseous
oxygen shear-coaxial injector test. The chamber shown in Fig. 2 was a modular heat sink
design assembled with two OFHC (oxygen-free high conductivity) copper cylindrical barrel
spool sections and an ablation resistance material nozzle. The inner diameter was 26 mm
and the cylinder part was 255 mm in length.
Many small holes were designed at the axial locations of the measurement points, with
temperatures of those at the bottom measured by the thermocouples. Axial locations for
instrumentation are indicated in Fig. 2. The distances between all the measurement points
and the chamber inner wall were the same except the distance between the first point and
the inner wall. Multiple instrumentations (4 each) were presented at some axial locations,
and four thermocouples were separated at 90° to examine the non-axisymmetry which may
probably happen due to presumable manufacture and installation errors. Four
thermocouples were also fixed in the injector surface to observe the plate surface’s heat load.
Fig. 2. Schematic of the chamber and thermocouple locations

The timing of the tests was designed as follows: the igniter preceded the injector, and with
the chamber pressure got steady, the igniter was shut down. In the light of the fact that the
igniter mass flowrate was less than 8.0g/s, the total ignition gas temperature was less than
1000K, and the ignition time lasted less than 0.8s, the influence of the ignition on wall
temperature of the long cylinder part could be neglected here. A photo of the calorimeter
chamber is shown in Fig. 3.
Fig. 3. The photo of the calorimeter chamber
3. Key design parameters and thermocouples

The test time was set to be 3s for this heat sink chamber. Some key parameters of the testing
article, including the chamber wall thickness H, measurement hole diameter D and the
distance L from the measurement point to the inner wall surface are listed in Table 1.
The Type-K thermocouples used in this hot-testing example were designed and provided by
the Beijing West Zhonghang Technology Ltd. Their measurement ranges were all set to be
0-600°C. The temperature versus voltage response for the Type-K thermocouple is shown in
Fig. 4, which shows a near-linear response. Their response time constants were designed
less than 100ms and calibrated individually.
H/mm 48
D/mm 1.5
L/mm 8
Table 1. The values of key parameters
600
500
400
T, 0C
300
200
100
0
0 5 10 15 20 25
Voltage, Millivolts
Fig. 4. Type-K thermocouple response

The chamber was fixed on a testing platform at the FFSC Laboratory in BUAA, resulting
that the chamber was electronically grounded, which can produce strong interference signal
in the output data. Therefore, an isolation module transmitter was designed for each
thermocouple to eliminate the interference signal. The temperature responses of a
thermocouple without and with isolation module are shown in Fig. 5.
40 24
22
30
20
T, C
T, 0C
20 18
0
16
10
14
0 12
0 1 2 3 4 0 1 2 3 4
t, s t, s
a) b)
Fig. 5. Temperature outputs of transmitters a) without and b) with insulate module

4. Calculation of inner wall temperature and heat flux

Several hot-fire tests were carried out for this gaseous hydrogen/gaseous oxygen single-
element injector. A typical chamber pressure and one typical temperature curve are shown
in Fig. 6. The temperature curve is depicted at the measurement point which is located
100mm away from the injector plane axially.
180 3
160
T 2.5
140 Pc
2
120
Pc, MPa
T, C
100 1.5
0
80
1
60
0.5
40
20 0
4 6 8 10 12 14 16
t, s
Fig. 6. Typical chamber pressure and wall temperature curve

Two steps are involved in converting these temperature output data into temperature and
heat flux on the inner wall: 1. depicting the real measurement point temperature curves
based on the output temperature curves, 2. obtaining the temperature and heat flux on the
inner wall with 2-D axisymmetric heat transfer simulation. An error analysis will also be
conducted in the end for this method.
4.1 Temperature conversion

Any thermocouple has somewhat response delay, that is, when thermocouple of initial
temperature T0 is used to measure an object of constant temperature ( T1 ), the response
curve function can write into equation (1):
T1 − T 1 (1)
= exp(− τ )
T1 − T0 τc
where τ c is thermocouple delay constant and τ is time. The qualitative relationship
between the object temperature and thermocouple output is shown in Fig. 7.
The heat transfer was unsteady in the test. The wall temperature kept rising after the
combustion started. Obviously, there was some discrepancy between the real temperature
curve and the experimental output curve for the existence of response delay, and it
increased with the testing time, as shown in Fig. 8.
In this method, curve fitting and analytical calculation are employed to solve the real
temperature curve from the thermocouple output. An example is shown in Fig. 9, T1 is a
thermocouple output curve, and T2 is the biquadratic polynomial fitting curve of T1 (almost
overlapped with T1). T3 is the real temperature result considering the response delay effect,
which is higher than T1. The thermocouple delay constant was set 100ms in this case.
T
T1
τ
Fig. 7. The thermocouple characteristic curve schematic
T the real temperature
the measurement temperature
τ
Fig. 8. Comparison of the TC measurement temperature curve and real wall temperament
curve
80
70
60
T, 0C
50
T1
40
T2
T3
30
20
4.5 5 5.5 6 6.5 7
t, s
Fig. 9. Example of temperature curve conversion
4.2 Inner wall heat flux and temperature calculation

The influence of the hot-testing start-up on the wall temperature was very limited, which
can be observed from Fig. 10, thus, this influence was neglected here. When chamber
pressure reached 90% of the stable value was considered as the beginning of the heat
transfer and the calculation initial time, as t0 shown in Fig. 10. The end time of calculation
was chosen at the end of stable combustion, as t1 shown in Fig. 10. The heat transfer from t0
to t1 was calculated and the combustion was assumed to be stable in this period.
The inner wall heat flux can be defined with equation (2), in which the coefficient h can be
defined by the Bartz equation as show in equation (3)(4).
q = h(T * − Tw ) (2)
η cp
0.2 *
1 pc At
h = const . ) σ
0.8 0.9
0.2
( 0.6
)( *
) ( (3)
dt Pr c A
1 Twg κ −1 1 −0.68 κ −1 2 −0.12

(1 + Ma ) + ] (1 +
2
σ =[ *
Ma ) (4)
2T 2 2 2
*
Here, T and Tw are inner flow total temperature and the inner wall temperature. From
Equations (3)-(4), it can be seen that the coefficient h is almost unchanged when the
*
chamber pressure pC is constant. Figure 10 shows that the chamber pressure is steady and
constant in the test, thus, the coefficient h is considered as unchanged in the hot-test, i.e.
time-averaged heat transfer coefficient through the hot-test, which will be used as the
boundary condition in the following heat conduction numerical modeling. In fact, in a real
engine, the total combustion flow temperature, specific heat ratio and the Ma number are
about 3600K, 1.2, and 0.1, the inner wall temperature changes from 300K to 600K, the heat
transfer coefficient just changes less than 5% according to Equation (3)-(4).
180 3
t0 t1
160
2.5
140
2
120
Pc, MPa
T, 0C
100 1.5
T
80
PC 1
60
0.5
40
20 0
3 4 5 6 7 8 9 10
t, s
Fig. 10. The calculation time and temperature chosen from the output curve
The original thermocouple output data were obtained from the hot-fire test directly, and
then the temperature Tm at the end time of calculation (t1) were gotten. Then the real
temperatures Tr of each measurement point can be obtained through a conversion program
described above. Tm and Tr are shown in Table 2. The injector plate was set at zero point
axially.
Axial location/mm Tm/ºC Tr/ºC

-3 74.5 78.5
10 51.1 54.2
25 92.4 97.2
40 108.3 113.6
55 115.4 121.4
70 98.6 103.5
85 108.3 113.6
100 144.1 150.8
115 157.9 164.9
130 159.6 166.6
160 161.5 168.7
175 156.4 163.3
190 158.6 165.6
210 165.7 173
240 171.1 178.6
Table 2. The measurement and real temperatures
The calculation method for the inner wall temperature and heat flux from the real
temperatures at measurement points are introduced below. The whole calculation
procedure can be divided into two parts: 1-D and 2-D modelling. The real temperatures at
the axial measurement points were labeled as Trk (k=1, 2, …, n) for the sake of convenience.
4.2.1 1D axisymmetric modeling

Firstly, surface heat transfer coefficients at all measurement points were calculated with 1-D
axisymmetric model regardless of the longitudinal heat transfer effect. These 1-D results will
be used as the first iteration data for the 2-D calculation to accelerate the convergence. The
unsteady axisymmetric 1-D heat conduction equation is as follows:
∂T 1 ∂ ∂T
ρc = (λ r )+Φ (5)
∂t r ∂r ∂r
The total temperature of the inner combustion flow was chosen as the adiabatic combustion
flame temperature T*. Two empirical surface heat transfer coefficients hk1 and hk2 (hk1<hk2)
(k=1,2,…,n) were chosen firstly for each measurement point. Two temperatures Tk1 and Tk2
at time t1 shown in Fig. 10 were obtained from hk1 and hk2 with unsteady 1-D axisymmetric
heat transfer model respectively. The 1-D mesh contains 60 grid points for the 48mm
thickness chamber wall, and the stretching of the grid were used near the inner wall
boundary. Each unsteady calculation period is from t0 to t1 shown in Fig. 10 (~2.95s in this
case). Then the first iteration data hk for 2-D modeling can be obtained using the linear
interpolation with Equation (6). All the hk (k=1,2,…,n) are shown in Table 3.
Tk − Tk1
hk = hk1 + (hk2 − hk1 ) (6)
Tk2 − Tk1
4.2.2 2D axisymmetric modeling

2-D axisymmetric unsteady modeling was applied to calculate the real inner wall
temperature and the heat flux from t0 to t1, as shown in Fig. 11.
T1
T2
T3
Tn-1
Tn
Fig. 11. Schematic of continuous calculation boundary condition of 2-D modeling

In this step, several 2-D unsteady simulations are needed to get a converged inner wall heat
transfer coefficients. In each 2-D unsteady simulation, the inner wall of the chamber was
treated as the third boundary condition, the calculation time is from t0 to t1, and the initial
temperature at time t0 of the whole chamber was set at ambient temperature. The 2-D mesh
contains 500×60 grid points, and the stretching of the grid was also used near the inner wall
boundary. The unsteady axisymmetric heat conduction equation is as follows (Chapman
1987):
∂T 1 ∂ ∂T ∂ ∂T
ρc = (λ r ) + (λ
)+Φ (7)
∂t r ∂r ∂r ∂z ∂z
Source term Φ is set to h (T * −T ) on the inner surface, where h is heat transfer coefficient,
w
a series of heat transfer coefficients hk (k = 1,2,..,n) obtained from the 1-D modeling were
used for the first 2-D simulation at the corresponding inner wall grid points. Tw is the inner
wall temperature which keeps changing in each 2-D simulation process, ambient
temperature Tw 0 for the initial condition of all the 2-D simulations. After the first unsteady
2-D simulation, a series of measurement point temperatures Tk’ (k = 1,2,…,n) at the end of
calculation time (t1) were obtained. Generally, Tk’ (k = 1,2,...,n) differs from the real
temperatures Trk (k = 1,2,…,n). A correction of each heat transfer coefficients hk ' for the
next unsteady 2-D simulation was made with Equation(8). And the unsteady 2-D simulation
was repeated until the differences between the simulation results and the real temperatures
Trk (k = 1,2,...,n) reach the tolerance.
hk ' = hk ⋅ (Trk − T0 ) / (Tk '− T0 ) (8)
Axial location/mm h1D/MW/(m2K) h2D/MW/(m2K)
0 368 255
10 933 655
25 1701 1928
40 2038.5 2379
55 2175 2360
70 1838 1450
85 2019 1564
100 2831 3136
115 3028 3280
130 3062 2964
160 3100 3020
175 3000 2800
190 2840 2991
210 3180 2800
240 3296 3060
Table 3. Heat transfer coefficients on the hot-gas-wall
In addition, the temperature-dependent metal (OFHC copper) properties were also
considered in the modeling. Table 3 lists the surface heat transfer coefficients at the
corresponding axial locations from 1-D and 2-D calculations when the adiabatic flame
temperature was considered as the total temperature. Because in the 2-D equation
calculation, the inner wall heat flux is the source term and unique, if difference total
temperature of the inner flow was set, the coefficients will be changed, but it did not
influence the ultimate results of inner wall temperature and heat flux.
The inner wall temperature and the heat flux distributions at the end of calculation time are
shown in Fig. 12, where X-coordinate denotes the distance from injector face plate. The
results show that heat load on the injector face plate is very low. Temperature and heat flux
distributions have similar profiles: They both begin to rise from the injector plate and reach
the first peak at around 0.05m, and decrease until 0.075m, and then continue to climb
successively to get the maximum values at 0.120m, followed by some small fluctuations and
a little decrease in the end section. Though some small fluctuations exits for both the
temperature and heat flux at the end, they mainly keep stable. It should be mentioned here
that the fluctuations here were somewhat produced by using the same delay constant in the
conversion calculation process, which were actually not involved in all thermocouples.
This heat load profile in the chamber was directly produced by the inner combustion
flowfield structure. The first peak of the heat flux in the beginning is due to the existence of
the strong recirculation zone there. Then the heat flux gets up continuously is because the
inner mixing and combustion become more and more intense and sufficient, and the flow
becomes faster and faster downstream. When the combustion is mainly completed, the
flowfield temperature and velocity reach their maxima. As the flow moves further
downstream, the combustion heat release is generally finished, but the wall heat loss still
exists, which induces the heat flux goes down a little in the end.
650 9
8
600
7
550
2
6
q, MW/m
T, K
500
5
450 T
q 4
400
3
350 2
0 0.05 0.1 0.15 0.2 0.25
x, m
Fig. 12. The hot-gas wall temperature and heat flux distributions
4.3 Error analysis

Errors of this method mainly come from the temperature conversion process and the
assumption of the 2-D axisymmetric calculation. Detailed analysis will be conducted in
these two parts.
4.3.1 Temperature conversion errors

Errors from the temperature conversion process occurred in the polynomial curve-fitting
and application of the same response time constants for all thermocouples. For example as is
shown in Fig. 9, the curve-fitting variance was 90.0635K2 with 2800 temperature data points,
indicating that this kind of error was very small and acceptable.
The deviation of response time also caused some errors to the results. Though the response
time constants were provided by the manufacturer, analysis should be carried out to
estimate its influence on the results.
Fig. 13 shows the different converted temperature curve results from one measurement
output data with different response time constants of 100ms and 150ms respectively. Curve
T1 is the original output of thermocouple. T100 is the converted curve for response time of
100ms and T150 for 150ms. It can be seen that though the difference of the time constants is
even 50%, difference of the results is about 1.2 K, just 2% compared with the total
temperature increment 60K.
4.3.2 Calculation assumption error

2-D axisymmetric chamber physical model was employed in the heat conduction
simulation, whereas real chamber body is three dimensional containing the measurement
holes. The 2-D axisymmetric assumption produced somewhat new error, for which, a 3-D
calculation and a 2-D axisymmetric calculation for one chamber structure were carried out.
80
70
60
T, C
50
0
40 T1
T100
T150
30
20
4.5 5 5.5 6 6.5 7
t, s
Fig. 13. Comparison of the converted temperature curves for different response times
Real chamber body 3-D physical model for calculations is shown in Fig. 14, where 1/4 of a
chamber section with a measurement hole was selected. Uniform convection heat load was
imposed on the inner wall surface. Bartz equation was used to get the convection heat
coefficient h, 7500W/m2K in this case. Initial temperature was set at 300K. Another 2-D
axisymmetric simulation was conducted with a whole chamber body without the
measurement hole under the same boundary conditions. The results of these two cases after
3s unsteady calculations are shown in Table 4.
Fig. 14. The physical model of 3-D calculation

Table 4 indicates that the differences of temperature increment on the measurement hole
bottom (ΔTh), the heat flux (qi) and temperature increment (ΔTi) at the corresponding inner
wall surface were very low. The maximum error was just 3.6%. The reason is that the 3-D
character of this chamber body with this measurement method is not evident because
measurement hole was rather small compared with the chamber body. The result indicates
that for this measurement method, though the chamber structure is three dimensional, the
2-D axisymmetric calculation is feasible and the calculation cost can be sharply saved.
Parameters Chamber with holes Chamber without holes Errors

ΔTh/K 361 348 3.6%
qi/MK/m2 17.1 17.5 2.3%
ΔTi/K 675 665 1.5%
Table 4. Comparison of some key results with 2-D and 3-D calculations
5. Application in studying gas-gas injector combustion chamber

This measurement method was used to investigate the heat transfer characteristics of a
single-element gas-gas injector combustion chamber.
5.1 The testing article

The single-element shear-coaxial injector selected from the real engine shown in Fig. 15 was
hot-tested here. The configuration of the single-element chamber can be seen in Fig. 2. All of
the design parameters of this chamber were the nominal ones from the real engine. The
combustor contract ratio was 3.1 and the character length was 800mm. The inner diameter
was 26 mm and the cylinder part was 255 mm in length.
Fig. 15. Schematic of the injector

In the tests of different chamber pressures, with the heat-sink design of the chamber and the
heat protection in consideration, the durations of the steady hot-fires were set from 1.5s to
5s for different chamber pressure cases: 0.92MPa—5s, 1.83MPa—4s, 2.69MPa—4s,
3.63MPa—3s, 4.52MPa—2s, 5.42MPa—2s, 6.1MPa—1.5s.
5.2 Test conditions

A total of 17 hot-fire tests were conducted steadily. The operation conditions and typical
chamber pressure profiles are summarized in Table 5 and Fig. 16. The nominal Mixture
Ration (MR) of all conditions were 6.0, the uncertainties of the MR of tests were -2.8%-2.3%.
The injection velocities and the propellant temperatures for all the cases were kept
unchanged.
Chamber OX Fuel Fuel

Ox. injection Repeat
Pressure Flowrate Flowrate injection MR
velocity/(m/s) Times
(MPa) (g/s) (g/s) velocity/(m/s)
0.92 66.7 ~70 11.2 ~760 5.96 3
1.83 135.1 ~70 22.0 ~760 6.14 2
2.69 195.4 ~70 32.8 ~760 5.95 2
3.63 258.3 ~70 44.0 ~760 5.87 3
4.52 327.2 ~70 54.4 ~760 6.01 2
5.42 397.8 ~70 65.6 ~760 6.06 3
6.1 446.9 ~70 76.6 ~760 5.83 2
Table 5. Test conditions summary
Fig. 16. The typical chamber pressure profiles of 7 cases

The time traces of some thermocouples for a representative 2.69MPa chamber pressure test
are shown in Fig. 17. A total of eight sets of thermocouple temperature measurements are
shown. In terms of nomenclature in the figure, for example, the first trace labeled TC-10-00
denotes that the thermocouple was at the 10mm axial location, at 00 degrees (angle was
defined with respect to major array of thermocouple). Except the curve TC-25-00, it can be
seen that all temperature traces had the same response characteristic, were all well behaved
and not noisy. The TC-25-00 had an obvious longer response time than others, so it could
not be utilized.
During the steady state portion of the firing, the temperatures rose steadily owing to the
heat sink nature of the chamber design. The curves of two thermocouples located
respectively at 40mm and 100mm are nearly identical suggesting that the chamber flow was
concentric. According to theory of heat transfer, higher heat flux on the inner wall at axial
location of measurement point consequentially induces higher temperature raise at this
point. Picture of the raises of temperatures at these measurement points versus the axial
distance for 2.69MPa chamber pressure case is shown in Fig. 18, manifesting that the results
of 2 repetitive tests were nearly identical.
All temperature curves were obtained for all the pressure cases, and then an axisymmetric
heat conduction numerical calculation was conducted to obtain the hot-gas-wall heat flux
for each pressure case. Inspection of empirical heat transfer correlations available in the
literature such as the Bartz (Bartz, 1957), all the heat flux data were scaled by 1 / p 0.8 , and the
results are shown in Fig.19. It can be seen that all the heat flux distribution curves collapse to
a single profile, and all the cases show the same qualitative distribution trends and the
almost same quantitative local values, which means that the heat flux q of a gas-gas injector
combustor correlates well with the pressure p as q ~ p 0.8 . A valuable suggestion can thus
be drawn that the heat flux data at high pressure condition can be predicted from that at a
low pressure condition.
Fig. 17. Thermocouple temperature traces (representative) for a 2.69MPa test
220
200
180
160
T, C
0
140
120 Run 1 of 2
Run 2 of 2
100
80
0 50 100 150 200
x, mm
Fig. 18. Wall temperature versus axial distance for 2.69MPa chamber pressure
3.5
q, (MW/m )⋅(Pa ) 2.5

-0.8
0.92MPa
2
2
1.83MPa
2.69MPa
1.5 3.63MPa
4.52MPa
1 5.42MPa
6.1MPa
0.5
0 0.05 0.1 0.15 0.2
x, m
Fig. 19. Heat flux (scaled with respect to (1/Pc0.8)) versus axial distance for each chamber
pressure case
5.4 Numerical study

In order to investigate the heat transfer characteristics at the high pressure condition
unavailable in the experimental hot-test, and further examine the inner combustion
flowfields at different chamber pressures, numerical simulations were conducted on this
combustion chamber.
5.4.1 Numerical models

A great effort has been made to perform the CFD simulation of gas-gas combustion flow at
Pennsylvania State University, NASA Marshall Space Flight Center, University of Michigan
and Beihang University et al. (Foust et al., 1996; Schley et al., 1997; Lin et al., 2005; Tucker et
al., 2007a, 2008b; Cai et al., 2008; Sozer et al., 2009; Wang, 2009a, 2010b, 2010c) And the
results indicated that the steady Reynolds Average Navier-Stokes (RANS) method
combined with a k − ε turbulence model could effectively simulate the whole combustion
flow and obtain the statistical average solutions that can match the experimental results. In
reference (Wang, 2010), difference RANS models were used to simulate a hot-testing
chamber, and a feasible k − ε turbulence model was obtained. Here, the RANS method
combined with this k − ε turbulence model was used.
Constant pressure specific heat of each species was calculated as a function of temperature
C / R = a0 + a1T + a2T 2 + a3T 3 + a4T 4 + a5T 5 (9)
Coefficients of laminar viscosity and heat conduction of single component were calculated
by molecular dynamics. The compressibility of the gas propellants at high pressure was
considered. The R-K equation was substituted for the ideal state equation to take the real gas
effect into account.
5.4.2 Numerical method and boundary condition

The entire system was solved by a strongly coupled implicit time-marching method with
ADI factorization for the inversion of the implicit operator. Convective terms were 2-order
flux split upwinding differenced, whereas diffusion terms were centrally differenced.
The calculation domain only occupied half the chamber. The radial and axial stretchings of
the grid were used near the wall boundary and in the shear layer domain. The grid
consisted of 29,028 cells, and the grid of half the cylinder was 43×350.
The inlets were fixed mass flowrate, and the inlet turbulence intensities both set to be 5%.
The centerline was an axisymmetric boundary, and the nozzle exit was specified as a
supersonic outlet. Non-slip wall boundaries were used on the chamber walls. The temperature
of the combustor wall was set at environment temperature of 300K to achieve a steady heat
flux.

The dimensions of the chambers were kept unchanged, and a total of 4 numerical cases
under different pressures from 5MPa to 20 MPa were chosen and shown in Table 6. The
combustion flowfields and heat flux along with the combustor wall were obtained. The
temperature contours are shown in Fig. 20, which shows that all the temperature contours of
4 pressure conditions are similar. And the similarity of the inner combustion flowfield
structures leads to the same inner wall heat flux distribution shown in Fig. 21. From the
time-mean inner flowfield results, the wall heat flux distribution can be clearly explained.
The little peak of the heat flux in the beginning originates from the existence of the strong
recirculation zone there. Then the heat flux gets up continuously with the increasing
intensity and sufficiency of the inner mixing and combustion and the increasing velocity of
the downstream flow. With the combustion mainly completed at the end of the combustor,
the flowfield temperature and velocity both reach their maximum values. As the flow moves
further downstream, the combustion heat release is generally finished, but the wall heat loss
still exists, inducing a little downward movement of the heat flux in the end. In Fig. 21 all
the heat flux data were scaled by pc 0.8 . It can be seen that all the curves almost collapse to a
a)5MPa b)10MPa
c)15MPa d)20MPa
Fig. 20. Temperature contours of the five different pressure cases

single profile, which indicates that in the high pressure conditions, the heat flux in gas-gas
injector combustors of different pressures also have the same qualitative distribution, and in
a good agreement with q ~ pc 0.8 quantitatively.
H2 O2
Chamber H2 H2 O2 O2
injection Injection
pressure flowrate temperature flowrate temperature
velocity velocity
/MPa /(kg/s) /K /(kg/s) /K
/(m/s) /(m/s)
5 0.054 300 ~760 0.324 300 ~70
10 0.108 300 ~760 0.648 300 ~70
15 0.162 300 ~760 0.972 300 ~70
20 0.216 300 ~760 1.296 300 ~70
Table 6. Parameters of pressure scaling conditions
5MPa
6 10MPa
q / (MW/m2)*(Pa-0.8)
15MPa
20MPa
0.05 0.1 0.15 0.2 0.25

x/m
Fig. 21. Heat flux (scaled with respect to 1/Pc0.8) versus axial distance for four chamber
pressure cases
6. Conclusion
A method for measurement of single-injector heat transfer characteristics in a heat sink
chamber was expound in this chapter. A series of measurement points are designed in the
chamber with the same axial intervals and the same distance from the inner wall surface.
This method measures the temperatures at these measurement points and then converts
these temperatures into inner wall temperatures and heat flux with 2-D axisymmetric
calculation. A hot-testing of a single-element gas-gas shear-coaxial injector chamber
applying this method was introduced to explain this method. And the inner wall
temperature and heat flux for this case were obtained and demonstrated. The basic principle
and design, data processing and the corresponding error analysis were described in detail.
And the error analysis showed that the accuracy of this method is sufficient for engineering
application, and the 2-D axisymmetric calculation can substitute for the expensive 3-D
calculation with its cost-saving advantage. The method was originally developed for single-
element axisymmetric chamber, and can also serve as a reference for non-axisymmetric
chambers and multi-element injector chambers.
Furthermore, this method was used to investigate the heat transfer characteristics of a
single-element shear-coaxial gas-gas injector combustion chamber. A single-injector heat-
sink chamber was designed and hot-fire tested for 17 times at chamber pressure from
0.92MPa to 6.1MPa. Inner hot-gas-wall temperature and heat flux along with the axial
direction of the chamber were obtained. The results show that heat flux in gas-gas injector
combustors of different pressures not only have the same distribution qualitatively, also
show a good agreement with q ~ pc 0.8 quantitatively. The inner combustion flows were also
numerically simulated with multi-species turbulence N-S equations at higher chamber
pressure from 5MPa to 20MPa to extend the experimental results. Both the flows structures
and heat flux profiles on inner wall were obtained and discussed, and the results of
numerical simulations indicated that the combustion flowfield of different pressures are
similar and the heat flux is also proportional to pressure to the power 0.8.
7. Acknowledgments
The authors acknowledge the support of the state high-tech research and development fund.
The authors also thank W. Zhang and Sh. Li from Beijing West Zhonghang Technology Ltd.
for helps in designing the thermocouples. Finally, the authors thank all the people who
made contribution and gave much help to this paper.
8. References
Archambault, M. R., Peroomian, O., "Characterization of a Gas/Gas Hydrogen/Oxygen
Engine," AIAA Paper 2002-3594, 2002a.
Archambault, M. R., Talley,R. D., Peroomian, O., "Computational Analysis of a Single-
Element Shear-Coaxial GH2/GO2 Engine," AIAA Paper 2002-1088, 2002b.
Bartz, D.R., "A Simple Equation for Rapid Estimation of Rocket Nozzle Convective Heat
Transfer Coefficients," Jet Propulsion, Vol.27, No.1, Jan. 1957. pp: 49-51.
Cai G B, Wang X W, Jin P, Gao Y S. Experimental and Numerical Investigation of Large
Mass Flow Rate Gas-Gas Injectors .AIAA Paper 2008-4562.
Calhoon, D., Ito, J., and Kors, D., "Investigation of Gaseous Propellant Combustion and
Associated Injector-Chamber design Guide- lines," NAS 3-13379, Aerojet Liquid
Rocket Company, Sacramento, California, 1973.
Chapman A. J., Fundamentals of Heat Transfer, Macmillan, New York, 1987.
Conley, A., Vaidyanathan, A., and Segal, C., "Heat Flux Measurements for a GO2/GH2
Single-Element, Shear Injector," Journal of Spacecraft and Rockets, Vol. 44, No. 3, May-
June 2007. pp. 633-639.
Coy E., “Measurement of Transient Heat Flux and Surface Temperature Using Embedded
Temperature Sensors”, Journal of Thermophysics and Heat Transfer, Vol.24, No.1.
January–February 2010. pp. 77-84.
Davis, J. A., Campbell, R. L., "Advantages of A Full-flow Staged Combustion Cycle Engine
System", AIAA Paper 1997-3318, 1997.
Farhangi, S., Yu, T., Rojas, L., and Sprouse, K., "Gas-Gas Injector Technology for Full Flow
Stage Combustion Cycle Application," AIAA Paper 1999-2757, 1999.
Foust, M. J., Deshpande, M., Pal, S., Ni, T., Merkle, C. L., Santoro, R. J., "Experimental and
Analytical Characterization of a Shear Coaxial Combusting GO2/GH2 Flow field,"
AIAA Paper 1996-0646, 1996.
Groot, W., A., McGuire, T., J., and Schneider, S., J., "Qualitative Flow Visualization of an
110N Hydrogen/Oxygen Laboratory Model Thruster", AIAA Paper 1997-2847, 1997.
Jones G., Protz C., Bullard B., and Hulka J., "Local Heat Flux Measurements with Single
Element Coaxial Injectors," AIAA Paper No. 2006-5194, July 2006.
Lin, J., West, J. S., Williamst, R. W., and Tucker, P. K., "CFD Code Validation of Wall Heat
Fluxes for a GO2/GH2 Single Element Combustor," AIAA Paper 2005-4524, 2005.
Marshall W. M., Pal S., and Santoro R. J., "Benchmark Wall Heat Flux Data for a GO2/GH2
Single Element Combustor," AIAA Paper No. 2005-3572, July 2005.
Meyer, L., Nichols, J., Jones, J. M., "Integrated Powerhead Demonstrator (booster hydrogen
oxygen rocket engines)," AIAA Paper 1996-4264, 1996.
NASA Space Vehicle Design Criteria. "Liquid rocket engine injectors," NASA SP-8089, 1976.
Santoro R. J. and Pal S., "Validation Data for Full Flow Staged Combustion Injectors," Final
Report for NASA Contract Grant NAG8-1792, Pennsylvania State University, 2005.
Schley, C-A., Hagemann, G., Tucker, P. K., "Comparison of Calculation Codes for Modeling
Hydrogen-Oxygen Injectors," AIAA Paper 1997-3302, 1997.
Sozer E, Vaidyanathan A, Segal C, and Shyy W, Computational Assessment of Gaseous
Reacting Flows in Single Element Injector, AIAA Paper 2009-449.
Tramecourt, N., Masquelet, M., and Menon, S., "Large-Eddy Simulation of Unsteady Wall
Heat Transfer in a High Pressure Combustion Chamber," AIAA Paper No. 2005-
4124, July 2005.
Tucker, K., West, J., Williams, R., Lin, J., Rocker, M., Canabal, F., Robles, B., and Garcia, R.,
"Using CFD as a Rocket Injector Design Tool: Recent Progress at Marshall Space
Flight Center," NASA NTRS 20050217148, Jan. 2005.
Tucker, P. K., Klemt, M. D., and Smith, T. D., "Design of Efficient GO2/GH2 Injectors: a
NASA, Industry and University Cooperative Effort," AIAA Paper 1997-3350, 1997.
Tucker, P. K., Menon, S., Merkle, C. L., Oefelein, J. C., and Yang, V., "An Approach to
Improved Credibility of CFD Simulations for Rocket Injector Design," AIAA Paper
2007-5572, 2007.
Tucker, P. K., Menon, S., Merkle, C. L., Oefelein, J. C., and Yang, V., "Validation of High-
Fidelity CFD Simulations for Rocket Injector Design," AIAA Paper 2008-5226, 2008.
Vaidyanathan A., Gustavsson J., and Segal C., "Heat Fluxes/OH-PLIF Measurements in a
GO2-GH2 Single-Element Shear Injector," AIAA Paper No. 2007-5591, July 2007.
Vaidyanathan A., Gustavsson J. and Segal C., "One- and Three-Dimensional Wall Heat Flux
Calculations in a O2-H2 System," Journal of Propulsion and Power, Vol. 26, No. 1,
January-February 2010.
Wang X W, Cai G B, Gao Y S. Large Flow Rate Shear-Coaxial Gas-Gas Injector. AIAA Paper
2009-5042.
Wang X W, Cai G B, Jin P. Scaling of the flowfield in a combustion chamber with a gas-gas
injector. Chinese Physics B, Vol. 19, No.1 (2010). SCI DOI: 10.1088/1674-
1056/19/1/019401.
Wang X W, Jin P, Cai G B . Method for investigatio n of combustion flowfield characteristics

in single-element gas/gas injector chamber. Journal of Beijing University of
Aeronautics and Astronautics, 35(9), (2009). pp.1095-1099
Zurbach, S. (ed.), Rocket Combustion Modeling, 3rd International Symposium, Centre
National D'Etudes Spatiales, Paris, March 2006.
24
Heat Transfer Related to

Gas Hydrate Formation/Dissociation
Bei Liu, Weixin Pang, Baozi Peng, Changyu Sun and Guangjin Chen
State Key Laboratory of Heavy Oil Processing,
China University of Petroleum, Beijing 102249,
P. R. China
1. Introduction
Gas hydrates are ice-like crystalline compounds comprised of small guest molecules, such as
methane or other light hydrocarbons, which are trapped in cages of a hydrogen-bonded
water framework. It has drawn attention in the gas and oil industry since 1930s because it
was found that the formation of gas hydrates may block oil/gas pipelines (Sloan and Koh,
2007). However, with the gradual discovery of huge reserve of natural gas hydrates in the
earth as well as the understanding of the peculiar properties of gas hydrates, more and more
studies have focused on how to benefit from gas hydrates in recent decades. The most
important aspect of gas hydrates research is attributed to the exploration and exploitation of
natural gas hydrates. Additionally, people also try a lot in the development of novel
technologies based on hydrates, such as separation of gas mixture via forming hydrates,
storage of natural gas or hydrogen in the form of solid hydrates, and sequestration of CO2,
etc. As the formation of gas hydrates is an exothermic process, heat transfer always
accompanies hydrate formation or dissociation. The understanding of heat transfer
mechanism is critical to the modeling of formation/dissociation kinetic process of gas
hydrates, which favors the best exploitation of natural gas hydrates and the best design of
reactor for hydrate production or decomposer for hydrate dissociation with respect to
different kinds of hydrate application objects.
In recent years, a variety of experimental and theoretical works focused on heat transfer
involved in formation/dissociation of gas hydrates have been reported. They are summarized
in this chapter accompanying presentation of our new work relevant to this topic. This
chapter is organized as follows. In section 2, we present progresses in experimental
measurement of the thermal conductivities of different kinds of gas hydrates, including
pure gas hydrates and hydrate-bearing sediments. The achievements on mechanism and
modeling of heat transfer occurring in the growth of hydrate film at the guest/water
interface, as well as its influence upon the hydrate film growth rate are summarized in
section 3. Our new experimental study on heat transfer in stirring or flowing hydrate system
is given in section 4. Section 5 presents our recent work on the experimental and modeling
studies on heat transfer in quiescent reactors for producing or decomposing big blocks of
hydrates, and the formulation of the influence of heat transfer upon the hydrate
formation/dissociation rate. In section 6, the mechanism of heat transfer in hydrate
bearing-sediment are analyzed and discussed. Finally, some concluding remarks are given
in section 7.
2. Thermal conductivity of gas hydrate

Thermal conductivity is a kind of basic data for studying the heat transfer of hydrates
involved systems. In recent decades, a number of researchers have made their efforts to
measure the thermal conductivities of different types of gas hydrates at different conditions.
Regarding to measurement technique, the most widely adopted ones are standard needle
probe technique and transient plane source (TPS) technique (Gustafsson et al., 1979, 1986).
For example, thermal conductivity of methane hydrate has been determined by deMartin
(2001), Krivchikov et al. (2005), and Waite et al. (2007) using the needle probe technique.
With same technique, thermal conductivities of several other gas hydrates, such as
tetrohydrofuran (THF) hydrate (Cortes et al., 2009), xenon hydrate (Krivchikov et al., 2006),
HCFC-141b hydrate (Huang et al., 2004), and CFC-11 hydrate (Huang et al., 2004) have been
measured. Transient plane source (TPS) technique in double- and single-sided configurations
has been used more recently to measure thermal conductivity of gas hydrates (Huang and
Fan, 2004; Li et al., 2010; Rosenbaum et al., 2007). This technique is based on the transient
method and the needle probe, but it has a very small probe (Gustafsson et al., 1979, 1986). It
allows measurements without any disturbance from the interfaces between the sensor and
the bulk samples. In addition, it is possible to measure thermal conductivity, thermal
diffusivity, and heat capacity per unit volume simultaneously (Gustafsson et al., 1979). It is
hard to draw a definite conclusion that which technique is better for pure gas hydrate
samples synthesized in laboratory; however, for in-situ determination of the thermal
properties of hydrate-containing sediments, the single-sided TPS technique may be more
suitable as the needle probe and double-sided TPS techniques need the probe to be
surrounded by the hydrates (English and Tse, 2010).
There are several factors, such as the porosity of the samples, temperature, pressure, and
measurement time, that influence thermal conductivity of gas hydrates. As pointed out by
English and Tse (2010), for relatively pure hydrates, reducing the porosity of the samples
by compacting them is critical for obtaining the reliable thermal conductivity in the
intermediate temperature range. For hydrate-bearing sediments, Tzirita (1992) concluded
that porosity is also a critical factor in controlling the thermal conductivity. More recently,
Cortes et al. (2009) carried out a systematic measurement of the thermal conductivity of
THF-hydrate saturated sand and clay samples. They found the influence is a complex
interplay among particle size, effective stress, porosity, and fluid-versus-hydrate filled
pore spaces, not only porosity. With respect to temperature effect, many studies found
that hydrates exhibit a glass-like temperature dependence of thermal conductivity
(Andersson and Ross, 1983; Handa and Cook, 1987; Krivchikov et al., 2005, 2006; Ross et
al., 1981; Ross and Andersson, 1982; Tse and White, 1988). Among these studies, the
works of Krivchikov et al. (2005, 2006) are interesting as they found that both methane
and xenon hydrates show crystal-like temperature dependence below 90 K, while
exhibiting glass-like behavior above 90 K. The effect of pressure has also been
investigated by many groups (Andersson and Ross, 1983; Rosenbaum et al., 2007; Waite et
al., 2007). Only weak pressure dependency was observed by them. Finally, the
relationship between thermal conductivity and measurement time for methane hydrate
Heat Transfer Related to Gas Hydrate Formation/Dissociation 479
has been studied by Li et al. (2010) very recently. They found that in 24h, thermal
conductivity increases 5.45% at 268.15 K; however, at 263.15 K, the increment is 196.29%.
From their results we may say that measurement time needs to be considered for thermal
conductivity studies at relatively low temperatures.
To give readers a clear picture of measured thermal conductivities of different kinds of gas
hydrates, the results of pure gas hydrates and hydrate-bearing sediments are listed in
Table 1.
compound T/K P / MPa λ / W m-1 K-1 Ref

pure gas hydrates
methane 253 - 290 31.5 ~ 0.62 Waite et al., 2007
methane
263.05 – 277.97 6.6 ~ 0.57 Huang and Fan, 2004
(compacted samples)
methane
261.5 – 277.4 3.8 – 14.2 ~ 0.68 Rosenbaum et al., 2007
(compacted samples)
tetrahydrofuran 15 -100 0.04 - 0.12 Tse and White, 1988
tetrahydrofuran·17 H2O 261 0.5 Waite et al., 2005
tetrahydrofuran·17 H2O 261 0.05 - 1 0.58 Cortes et al., 2009
xenon 245 ~ 0.05 0.36 Handa and Cook, 1987
ethylene oxide 263 0.49 Cook and Laubitz, 1983
Andersson and Ross,
cyclobutanone 260 100 ~ 0.47
1983
Andersson and Ross,
1,3 – dioxolane 260 100 ~ 0.51
1983
propane 275 1 ~ 0.4 Stoll and Bryan, 1979
sodium sulphide·9 H2O 295 0.001 0.12 Lunden et al., 1986
HCFC-141b ~ 250 0.1 ~ 0.5 Huang et al., 2004
CFC-11 ~ 250 0.1 ~ 0.5 Huang et al., 2004
hydrate-bearing sediments
natural methane
263 -283 0.1 3.8 – 5.8 Yamamoto et al., 2008
hydrate-layer sand
sand with 100%
261 0.05 - 1 4.1 – 4.5 Cortes et al., 2009
THF·17 H2O hydrate
clay with 100%
261 0.05 - 1 2.8 – 3.0 Cortes et al., 2009
THF·17 H2O hydrate
Table 1. Thermal conductivities, λ, of pure gas hydrates and hydrate-bearing sediments

3. Heat transfer in growth of hydrate film

Generally, because most of hydrate formers (guests) are water insoluble, the initial
formation of hydrate occurs at the guest/water interface, taking the form of thin porous
crystalline film. The further growth of hydrate is controlled by mass transfer of water or
hydrate former through the film. Many experimental and/or theoretical studies on the
growth of hydrate film have been carried out by several groups (Freer et al., 2001; Ma et al.,
2002; Mochizuki and Mori, 2006; Mori, 2001; Ohmura et al., 2000, 2005; Peng et al., 2007,
2008, 2009; Saito et al., 2010, 2011; Sun et al., 2007; Taylor et al., 2007; Uchida et al., 1999,
2002), including the morphology of hydrate film, the growth rate of hydrate film, the
thickness of hydrate film, the mechanism of hydrate film growth, and so on. However, so far
it is still a controversial topic on the growth mechanism of hydrate film. Recently, more
attention has been paid to the mechanism of heat transfer on hydrate film growth at the
guest/water interface than intrinsic kinetic and mass transfer mechanisms. In this part,
different hydrate film growth models, especially heat transfer models that have been
developed by various research groups are summarized.
Experimental and molecular dynamic simulation studies on the initial formation of hydrate
at the guest/water interface suggest that the interface where there is a significant
concentration gradient is the place to initiate and sustain hydrate formation (Moon, et al.
2003; Vysniauskas and Bishnoi, 1983). Englezos et al. (1987a, 1987b) studied the kinetics of
formation of methane, ethane, and their mixture hydrates in a semi-batch stirred tank
reactor. They presented an intrinsic kinetic model for the hydrate particle growth and the
rate of growth per particle was given by:
⎛ dn ⎞ ∗
⎜ ⎟ = K AP f − f eq
⎝ dt ⎠ P
( ) (1)
with
1 1 1
= + (2)
K ∗ Kr K d
where n is the moles of gas consumed, t is the hydrate reaction time, K ∗ is the combined
rate parameter, AP is the surface area of particles, f is the gas fugacity, f eq is the
equilibrium fugacity, K r and K d are the reaction rate constant and mass transfer coefficient
around the particle, respectively. Similarly, based on the assumption that the intrinsic
kinetics is the control step of hydrate formation and growth, Ma et al. (2002) developed a
model to correlate the lateral growth rate of hydrate film. The model was formulated as the
following form:
r f = A × ⎡⎢ e (
B −Δg /( RT ))
− 1⎤⎥ (3)
⎣ ⎦
where r f is the lateral growth of hydrate film. Parameters A and B are system composition
dependent and were determined by fitting experimental data. The Gibbs free energy
difference ( Δg ) was selected as the driving force to describe the hydrate growth process.
The experimental results indicated that this model could correlate the lateral film growth
rate perfectly (Ma et al., 2002; Sun et al., 2007).
Except for the models described above, some researchers suggested that the growth rate
of hydrate film is controlled by heat diffusion and some models were developed
correspondingly. For example, Uchida et al. (1999) presented a model analysis of the two-
dimensional growth of a carbon dioxide hydrate film (Figure 1).
Teq
TB
Texp
rc CO2 liquid
h rv H2O
Fig. 1. Hydrate film model of Uchida et al (1999)

In this model, they assumed that one half of the film is in water phase, and the other half is
in guest phase. The hydrate film has a semicircular front and is uniform in thickness. In
addition, this model assumes that hydrate crystals successively form only at the front of the
hydrate film and the front is maintained at the three-phase (water/guest-fluid/hydrate)
equilibrium temperature. The heat released by the hydrate crystal formation is diffused
away from the film front and into the water and guest-fluid phase. Based on these
assumptions, they formulated the heat balance at the edge of the film as
v f ρ h Δhh = λw ΔT / rc (4)
where v f is the rate of linear growth of the film, ρ h is the mass density of the film, Δhh is
the heat of hydrate formation (per unit mass of hydrate), λw is the thermal conductivity of
water, ΔT is the difference between the temperature at the film edge, T eq , and the
undisturbed temperature in the fluid phases, T B , and rc is the radius of curvature of the
edge. Uchida et al. (1999) correlated their experimental data on v f versus ΔT by means of
a linear regression analysis as follows:
v f = (1.73 ± 0.16)ΔT (5)
In Uchida et al.’s model (Uchida et al., 1999), the conductive heat transfer from the film front
was deduced from the temperature gradient, which was deemed as with little physical
reasoning (Mochizuki and Mori, 2006; Mori, 2001). Mori (2001) presented an alternative
model of hydrate film growth based on the idea that the front of hydrate film, which grew
on the interface between stagnant water and guest fluid, could be viewed to be held in
stratified flow of the two fluids with the velocity which was opposite in sign but equal in
magnitude to the velocity of the hydrate film front. In his work, the heat removed from the
film front to the liquid phases was treated as a steady convective heat transfer and other
assumptions were same as those of Uchida et al. (1999). The heat balance over the
hemicircular front of the film was formulated as follows:
1
(
v f δρ g Δhh = πδ α w + α g ΔT
4
) (6)
Mori (2001) assumed that the heat transfer coefficients, α w and α g , could be given by the
simplest type of convective heat transfer correlation in a dimensionless form and deduced a
δ - v f relation correspondingly:
v f δ = C ΔT 3/2 (7)
where
3/2
⎡π A 1 ⎛ λ λg ⎞⎤
C=⎢ ⎜ 1/3w
Prw1/3 + 1/3 Prg1/3 ⎟ ⎥ (8)
⎢ 4 ρ h Δhh ⎜⎝ ν w νg ⎟⎥
⎠⎦
⎣
In Equation 8, λw and λg are the thermal conductivity of water and the hydrate former,
respectively, ν w and ν g are the kinematic viscosity of water and the hydrate former,
respectively, and Prw and Prg are the Prandtl number of water and the hydrate former,
respectively.
Mochizuki and Mori (2006) modified Mori’s model and presented another model, as shown
in Figure 2. They assumed that there is a transient two-dimensional conductive heat transfer
from the film front to the water and guest-fluid phases plus the hydrate film itself. In this
model, the hydrate film was assumed to exist on the water side of the water-guest fluid
interface and the interface infinitely extend. No convection occurs in either of the water and
guest-fluid and other assumptions were same as those of Uchida et al. (1999). The rate of
heat removal from the front to the surroundings is balanced by the rate of heat generation of
hydrate-crystal formation.
y
Hydrate film growth Water
Conduct ive heat transfer

Hydrate film
δ
Interface
x
x0
Hydrate-forming fl uid
xh
Fig. 2. Hydrate film model of Mochizuki and Mori (2006)

The linear growth rate of the hydrate film along the water/guest-fluid interface,
v f = dx h /dt , was given in the following equation:
π /2 ⎛ ∂T ∂T ⎞
ρhδΔhH v f = ∫ ⎜λ − λw ⎟rh dθ (9)
−π /2 ⎜ h ∂r ∂r ⎟
⎝ r = rh − r = rh + ⎠
where ∂T / ∂r r = r −
and ∂T / ∂r r = r + are the radial temperature gradients on the hydrate side
h h
and the fluid side, respectively, at r = rh (i.e., the x position of the hydrate-film front). It
should be pointed out that this model is computationally complicated and hence
cumbersome to use. In addition, the assumptions adopted in this model, that is, the film
front is in the water phase or one half in water phase and the other half in guest-fluid phase
are too arbitrary. Therefore, Peng et al. (2007) proposed another hydrate film model based
on Mori’s model. In their model, they assumed part of thickness x of hydrate film is in guest
phase and another part of thickness δ − x is in water phase, as shown in Figure 3. The value
of x is guest composition dependent.
Teq
Texp TB
x
rc Gas
δ- x
δ water
vf
Fig. 3. Hydrate film model of Peng et al. (2007)

In Peng et al.’s model (Peng et al., 2007), the thickness of hydrate film was assumed to vary
with driving force inversely, i.e.,
δ = k / ΔT (10)
The lateral rate of hydrate film was then correlated by the following equation:
C 5
vf = ΔT 2 (11)
k
where the constant C in Mori’s model was reformulated by a generalized expression, as
shown below:
3/2
⎡ 1 ⎛ 2/3 1/3 1/3 ⎛ 2x ⎞ 2x ⎞⎤
C=⎢ 2/3 1/3 1/3
⎜ λw ρ w c p , w ⎜ π − arccos(1 − ) ⎟ + λg ρ g c p , g arccos(1 − ) ⎟ ⎥ (12)
⎣ 6 ρ h Δh h ⎝ ⎝ δ ⎠ δ ⎠⎦
The modified convection heat transfer model presented by Peng et al.(2007), i.e., Equation
11, has been used to correlate the lateral growth rate of hydrate film of different crystal
structures in wide temperature and pressure ranges (Peng et al., 2007, 2009). It can be
concluded that validity of Equation 11 is independent of the composition of hydrate former
and the structure type of hydrates (Peng et al., 2007).
For investigating the growth mechanism of the hydrate film, Freer et al. (2001) also studied
methane hydrate film growth on the water/methane interface experimentally and proposed
a model of lateral hydrate film growth. They calculated the v f of methane hydrate film by
assuming that one dimensional conductive heat transferred from the film front to water. As
the calculated v f was much lower than the experimentally measured v f , they suggested
that the hydrate film growth was controlled by both intrinsic kinetics and heat transfer.
Their model was expressed as:
λh ρ h v f = K (T eq − T B ) (13)
1 1 1
= + (14)
K k h
where λh is the thermal conductivity of hydrate, K is the total resistance, h is the heat
transfer coefficient, and k is the methane hydrate kinetic rate constant.
Additionally, for investigating which step is the main contribution to the hydrate film
growth, Peng et al. (2008) also presented a model based on the assumption that hydrate
lateral film growth is controlled by both intrinsic kinetics and heat transfer, as shown in
Figure 4.
T
s
T
exp TB
T
Gas
rc H2O
δ vf
vf dt
Fig. 4. Hydrate film model of Peng et al. (2008)
This model is similar to the model of Uchida et al. (1999). However, in this model Peng et al.
kept the hydrate film at T S rather than at the three-phase (water/guest/hydrate)
equilibrium temperature and they proposed the intrinsic rate of hydrate film growth should
not be of a linear relation with the driving force. Therefore, the balance of the heat removed
from the film front with that generated by the hydrate formation was formulated by the
equation proposed by Freer et al. (2001),
v f ρ h Δhh = k1 (T S − T B ) (15)
and the intrinsic kinetic equation was adopted as following:
v f = k2 (T eq − T S )n (16)
From Equations 15 and 16, the following equation can be obtained:
1/n
ρ h Δhh ⎛ vf ⎞
v f + ⎜⎜ ⎟⎟ = T eq − T B = ΔT (17)
k1 ⎝ k2 ⎠
Based on their experiment data, Peng et al. calculated the temperature differences between
the hydrate film front and the bulk water at different driving forces, which were taken as an
important factor on judging the dominating contribution for hydrate film growth at the
gas/water interface. They found that the effect of heat transfer on hydrate film growth is
much smaller than that of intrinsic kinetics, and suggested that the intrinsic kinetic is the
main control step for hydrate film growth of methane and carbon dioxide hydrate.
It should be pointed out that for the models mentioned above, the parameters were obtained
by correlating with different experimental data set. As the experimental data were obtained
in different experiment apparatus and the stochastic induction time of hydrate nucleation
may also affect the measurement of hydrate film growth rate for different experiment
device, it is hard to draw a definite conclusion that which model is better. More efforts need
to be made on hydrate film growth in the future.
4. Heat transfer in stirring or flowing hydrate system

Stirring is an important technique that can enhance heat and mass transfer, and thus
accelerating the speed of hydrate formation/dissociation. The state of hydrates formed
under stirring is usually in slurry, which is also the case when hydrates are formed in gas-
oil-water multi-phase flowing systems containing hydrate anti-agglomerants (AA). As a
result, the determination of heat transfer coefficient of hydrate slurry is crucial for
investigating the heat transfer in hydrate forming/dissociating processes under stirring or
flowing. Unfortunately, there are very few publications up to date, and thus only some
results obtained by our group are introduced in this part.
4.1 Experimental apparatus

The experimental equipments adopted in our work are shown in Figure 5, which mainly
contain the reactor with stirrer, constant temperature water bath, and temperature/pressure
sensor.
Stirrer RTD
P
Water Bath
Hydrate Slurry
RTD
Fig. 5. The schematic outline of the experimental apparatus

4.2 Experimental principals and steps

In our work the measurements were performed via the following steps:
1. The equipments were washed three to four times using distilled water;
2. The reactor was wetted using the experimental liquid, then the experimental liquid was
added to the reactor until the stirrer reaches half of the height of liquid added;
3. The experimental temperature and pressure was set. We started stirring until the
hydrate slurry was formed completely, then we closed the stirrer;
4. When the hydrate slurry was formed completely and the temperature was constant, the
temperature of the water bath was quickly decreased by about 10 K to make a
difference in temperature between hydrate slurry and water bath;
5. The stirrer was open at a certain stirring speed and the temperatures of both water bath
and hydrate slurry were recorded at a certain interval. The stirring was stopped when
the change in temperature is very small in both the water bath and the reactor.
4.3 Experimental data analysis

(1) Calculation of the total heat transfer coefficient
If we assume the heat released by the slurry in the reactor is equal to that adsorbed by the
water bath, that is, we neglect the heat loss during the measurement, the amount of heat
transfer, Q, can be calculated with the following equation:
Q = cm( − dT
dt
) (18)
where m is the mass of the hydrate slurry, dT is the temperature increase/decrease per
dt
unit time, and c is the specific heat of the experimental liquid. The amount of heat transfer
can also be calculated using the following equation:
Q = Kt AΔTm (19)
where K is the total heat transfer coefficient, A denotes the total area of heat transfer, and
ΔTm is the temperature difference between the water bath and the experimental liquid. The
total heat transfer coefficient can be calculated from Equations 18 and 19:
dT
Kt = cm(- )/AΔTm (20)
dt
(2) Calculation of the heat transfer coefficient α 1 of hydrate slurry
The total heat transfer coefficient can be expressed by:
d2 d
1
Kt
= α1 d1
+ λb d 2 + α1 (21)
1 m 2
where d1 , d2 , dm represent the inner, outer, and the averaged radius of the reactor,
respectively, and λ is the heat conduction coefficient of the reactor. Since in each run,
b d2 + 1 was kept constant, and thus this term can be eliminated by measuring the total
λ dm α 2
heat transfer coefficient of pure water and hydrate slurry, and calculating the difference in
the reciprocal total heat transfer coefficients of pure water and the hydrate slurry. The heat
transfer coefficient of hydrate slurry can be calculated by Equation 22 then.
d2 d2
1
Kwater
−K 1 =α 1
d1
−α 1 (22)
slurry water slurry d1

The heat transfer coefficients for both pure water and diesel oil/hydrate slurry systems with
the volume fraction of hydrates of 5%, 10%, 15%, and 20% were measured at 270 K or so.
The results are shown in Figure 6.
140
Heat transfer coefficient, W/( m2.K)
120
100
80
60
40
20
5% 10% 15% 20%

Vol ume per cent age of hydr at e i n sl ur r y, %
Fig. 6. Variation of heat transfer coefficient of hydrate slurry of different hydrate volume
percentage
Figure 6 shows that the heat transfer coefficient decreases with increasing the content of
hydrates in slurry, which can be attributed to the fact that the heat conductivity of hydrates
is very small.
5. Heat transfer in quiescent hydrate formation/dissociation reactor

It has been well known that the hydrate formation rate can be increased drastically by
adding low dose of suitable surfactants, such as sodium dodecyl sulfate (SDS). This kind of
additives can enhance mass transfer involved in hydrate formation by decreasing gas/liquid
interfacial tension and increasing the solubility of gas in liquid water. Then it is possible to
produce gas hydrate rapidly without stirring (Lin et al., 2004; Xie et al., 2005; Zhong and
Rogers, 2000). The advantage of quiescent formation of hydrate is that the cost on
manufacture and maintenance of the reactor could be reduced largely. Although the mass
transfer has been enhanced satisfyingly by adding SDS to water, the heat transfer becomes a
serious limitation to the application of quiescent reactor as hydrate formation is an
exothermic process. Rogers et al. (2005) designed a scaled-up quiescent process to store 5000
scf of natural gas in a vessel. They thought that the primary challenge of the scale-up design
was to provide a surface area/volume ratio in the larger vessel. Therefore they devised an
arrangement of finned-tube heat exchanger inside the hydrate formation vessel. Their
elementary tests on this process indicated that the hydrate formation in the vessel lasted
more than 9 hours, which is still too long for real applications. In order to suit the large scale
industrial applications, we devised a multi-deck cell-type vessel as the internals of the
reactor to reduce or eliminate the scale-up effect, which is schematically shown in Figure 7
(Pang et al., 2007). The vessel basically consists of a series of uniform boxes stacked up
vertically and each box is divided into a series of uniform cells by metal plates. The metal
plates were welded on the heat transfer tubes; therefore they also became the cool solid
surface during the hydrate formation. The SDS aqueous solution was loaded in these cells
with the same level. There are interspaces between two neighboring boxes such that the
hydrate forming gas can flow into each deck of the vessel easily. The multi-deck cell-type
vessel was placed in the high pressure reactor so that hydrate can form in each cell of the
vessel uniformly and simultaneously. In this case, the reaction time depends mainly on the
cell volume and the quantity of water loaded, and little on the total volume of the vessel and
total quantity of water loaded. Thus the scale-up effect can be eliminated to a large extent as
concluded by Pang et al. (2007). Since then, we carried out a systematical study on heat
transfer in hydrate formation/dissociation process using this vessel. Experimental details
and most recent results obtained by our group which have not been published are introduced
in this part.
Inlet of Outlet of
the Coolant the Coolant
Metal
Plate
Interspace
Heat
Transfer
Tube
Stanchion
Vessel
(a) Outline (b) Cell-type inner structure

Fig. 7. The schematic outline of multi-deck cell-type vessel

In order to investigate the heat transfer performance of this kind of inner structure during
hydrate formation/dissociation, a middle scale reactor of a volume of 10 liter as well as an
inner multi-deck cell-type vessel suitable for this reactor were manufactured and an
experimental set-up, as shown in Figure 8, was established correspondingly. The reactor is
200 mm in diameter, 320 mm in height, and has a volume of 10 L. It was sealed with a blank
flange bolted to its top.
Computer
RTD DPT
RTD Mass Flow

Meter
RTD RTD
DPT
Cooling
Jacket
Coolant Hot Water Spinner
Bath Bath Flowmeter
RT D
Spinner
Flowmeter
Reactor Gas Cylinder Gas Booster
Fig. 8. Flow diagram of the experimental apparatus

Besides the stainless steel quiescent reactor, the primary components of the experimental
apparatus consist of a refrigeration system and a coolant or hot water recycling system. A
cooling/heating jacket was welded to the outside of the reactor and coiled copper tubes
were placed inside the multi-deck cell-type vessel uniformly. Coolant or hot water was
circulated through the copper tubes to cool or heat the reactant system for the hydrate
formation or dissociation. The flow rate of the coolant or hot water was measured using a
spinner flow meter. The coolant is a mixture of water and glycol and has a freezing point
lower than 253 K. The coolant bath was controlled by the refrigeration system within a
temperature range from 253 K to ambient with an uncertainty of 0.1 K. For the sake of
accuracy, another platinum resistance thermometer of ±0.1 K accuracy was placed behind
the outlet of the pump to measure the temperatures of the coolant. In addition, a gas recycle
system was used to recover the residual gas after the completion of hydrate formation as
well as the gas released during hydrate dissociation. Two platinum resistance thermometers
of ±0.1 K accuracy were placed into the reactor from the flange to measure the temperatures
of the vapor phase and the liquid (or hydrate) phase, respectively. The pressure in the
reactor and the gas cylinder was measured using two calibrated Heise pressure gauges with
a precision of 0.05 MPa. A mass flow meter and computer system were used to measure and
record the flow rate of methane gas during the process of hydrates dissociation at
atmospheric pressure. The accuracy of the flow meter is ±1% F.S and the interval of record
time is 5s.
5.2 Experimental procedures

5.2.1 Hydrate formation
Prior to any experiments the reactor was washed with water, loaded with 1920 g aqueous
solution of 2000 mg/L SDS, and evacuated to remove air. The refrigeration system and
coolant pump were then turned on to set the temperature of the aqueous solution in the
reactor to 276.15 K. After the desired temperature of the aqueous solution was achieved,
methane gas was charged into the reactor until the desired pressure was achieved. During a
hydrate formation process, the gas was charged into the reactor continuously to keep the
pressure within the range of 6.4 to 6.8 MPa. When the drop in system pressure was less than
0.1 MPa over 20 minutes, the hydrate formation process was assumed to be complete.
Subsequently, the hydrate was frozen for dissociation test. In the experiment, the temperature
or flux of coolant was changed to determine the effect of heat transfer on the hydrate
formation rate. In each experimental run, the change of the pressure and temperature of the
gas in the reactor with the elapsing time was recorded, and the mole number, Δni , of
methane consumed in the time period of Δti was then determined with the following
equation:
PV
t t
Pt +Δti Vt +Δti
Δni = − (23)
Zt RTt Zt +Δti RTt +Δti
where Vt , Tt , Pt , and Zt are the volume, temperature, pressure, and compressibility of the
gas phase in the reactor at time t , respectively. Vt +Δt , Tt +Δt , Pt +Δt , and Zt +Δt are the volume,
temperature, pressure, and compressibility of the gas phase in the reactor at the time t + Δt ,
respectively. It should be noted that no gas was charged into the reactor from the gas
cylinder during the time period of Δti . The cumulative moles of gas consumed could be
calculated readily through
Δn = ∑ Δni (24)
i
5.2.2 Hydrate dissociation

After hydrate samples were formed with the procedure described above and cooled down
to 268.15 K, the system was left for about 3 hours with a less than 0.1 K fluctuation in reactor
temperature. Next, the vent was opened slowly to reduce the system pressure gradually to a
bit above the equilibrium formation pressure of methane hydrate at the set temperature.
Subsequently, the system was depressurized rapidly to near the atmospheric pressure of 0.1
MPa and the vent was then turned off. The methane hydrate samples were then heated with
hot water at a fixed temperature to a complete dissociation. The history of pressure and
temperature of the reactor, and the temperatures of the inlet and vent hot water were
recorded with the elapsed time. During the dissociation process, the hot water was charged
only into the coiled copper tubes and the coolant in the cooling jacket was expelled by air
before the dissociation experiment.
The cumulative moles of methane dissociated at time t without considering the shrinking of
methane hydrates volume during the dissociation can be calculated with the following
equation (Pang et al., 2009)
PVt i PV
Δnt = − i i (25)
Zt RTt Zi RTi
where Tt , Pt , and Zt are the temperature, pressure, and compressibility, respectively, of

the gas phase in the reactor at time t . Pi , Vi , Ti , and Zi are the initial dissociation
pressure, initial volume, initial temperature, and initial compressibility, respectively, of the
gas phase in the reactor.
5.3 Modeling of heat transfer dependence of hydrate formation

Englezos et al. (1987a) proposed that the hydrate formation rate was proportional to its
driving force and reaction area. In our work we defined the supercooling, i.e., the difference
between the equilibrium temperature and the actual reaction temperature, as the driving
force, then the rate of gas consumed for hydrate growth could be formulated as
dn
= K d Agl (Te − Tw ) (26)
dt
where K d is the intrinsic hydrate formation rate constant, Agl is the formation area, and Te
and Tw are the equilibrium temperature and actual temperature of bulk water, respectively.
As an exothermic reaction, the temperature of the water solution will increase with the
formation of hydrate. Thus the metal plates were welded on the heat transfer tubes to
remove the formation heat in the multi-deck cell-type vessel. Because of the temperature
difference between the bulk water and the metal plates, the heat transfer happened and the
reaction heat of hydrate formation could be expressed by
dn d(mC pT )
ΔH = K h (Tw − Tl ) Am + (27)
dt dt
where ΔΗ is the reaction heat of hydrate formation, K h is the heat transfer coefficient, Tl is
the temperature of coolant, Am is the heat transfer area(it is assumed to be equal to the area
of the metal plates), m is the total quantity of water and hydrate in the vessel, C p is the heat
capacity of mixture of water and hydrate in the vessel, and T is the temperature of water
and hydrate in the vessel (here we assumed that the temperature of water and hydrate is
same).
As the change of temperature is not obvious during the hydrate formation, it can be
assumed that most of the formation heat of hydrate formation is removed by the coolant. So
Equation 27 can be simplified as
dn
ΔH = K h (Tw − Tl ) Am (28)
dt
Combining Equations 26 and 28 yields
⎛ ⎞
⎜ ⎟
dn ⎜ 1 ⎟ (T − T )
= (29)
dt ⎜ ΔH + 1 ⎟ e l
⎜K A ⎟
⎝ h m K d Agl ⎠
Considering that the gas/water interface area decreases with the proceeding of hydrate
formation, the formation area Agl in Equation 29 is empirically formulated as
b
⎛ n⎞
Agl = As ⎜ 1 − ⎟ (30)
⎝ ns ⎠
where As is the interface area of water and gas at the beginning of reaction, ns is the total
quantity of consumed gas at the end of the reaction, n is the quantity of consumed gas at
time t, and b is the reduced exponent of reaction area. In our work, the shape of vessel is
round, and the vessel was divided into a series of uniform cells by metal plates with a 25
mm span. So the shape of most of the cells is square except the cell at the edge of the vessel,
which is not regular. In our model, those irregular cells were transformed into the square
cell according to the volume ratio, and the reaction status of hydrate formation in every cell
was assumed to be same.
The dependence of K d on temperature is formulated with the following Arrhenius-type
equation,
ΔE
−
K d = k0 e RT (31)
where k0 is the pre-exponential coefficient and ΔE is the activation energy of intrinsic

hydrate formation.
5.4 Heat transfer dependence in quiescent hydrate formation

Heat transfer is very important for the hydrate formation. The heat of hydrate formation
must be removed in time; otherwise the hydrate formation could not proceed continuously.
There are two factors significantly affecting the heat transfer rate: one is the structure of the
reactor, i.e., total area for heat exchange and the other one is the coolant. The effect of the
structure of the reactor has been investigated by Pang et al. (2007). Our present work focuses
on the effect of the coolant on the hydrate formation, including the influence of the
temperature and flux of the coolant. To study the effect of the coolant on the hydrate
formation and ensure all of the hydrate formation experiments are performed on the same
basis, the initial temperature of reaction system, the pressure, the mass of water solution,
and the concentration of surfactant SDS were uniformly specified as 276.15 K, 6.4-6.8 MPa,
1920 g, and 2000 mg/L, respectively. By changing the temperature and flux of coolant, a
series of experiments have been performed, and the results are shown in Figures 9 and 10,
where the hydrate formation rate was manifested by the cumulative mole numbers of
consumed gas vs. elapsed time.
5.4.1 Influence of coolant temperature

The effect of the temperature of coolant on the hydrate formation rate is shown in Figure 9.
One can see that hydrate formation rate increases with the decreasing of the coolant
temperature. This is easy to understand as when the temperature of coolant is lower, the
formation heat can be more easily removed. As a result, the temperature of the water
solution keeps lower and the driving force of hydrate formation keeps larger, resulting in
the hydrate forms faster. However, when the temperature of coolant is low enough, the
consumed rate of methane gas changed little with the decreasing of the coolant temperature.
It means that the effect of coolant temperature on the hydrate formation rate is limited.
When the temperature of coolant is low enough, the further decreasing of temperature have
little effect on the hydrate formation rate. At that time, the hydrate formation rate is not
controlled by the heat transfer, but controlled by the intrinsic kinetics of hydrate formation.
For comparison, the calculated results by Equation 29 are also presented in Figure 9. One
can see that the agreement between experimental data and calculated results is satisfying.
The parameters of model were correlated using the experimental data and are shown in
Table 2, where the flux of the coolant was fixed at 5.5 L/min and the hydrate formation heat
was set to 54.20 KJ/mol (Makogon,1997).
20 20
18 18
16 16
14 14
Consumed gas (mol)
Consumed gas (mol)

12 12
10 10
8 275.15 K 8 275.15 K
6 273.15 K 6 273.15 K
272.15 K 272.15 K
4 4 271.15 K
271.15 K
2 Cal. 2 Cal.
0 0
0 25 50 75 100 125 150 175 200 225 250 275 300 325 0 25 50 75 100 125 150 175 200 225 250 275 300 325
Time (min) Time (min)
(a) (b)
Fig. 9. The effect of the coolant temperature on hydrate formation rate at different flux of
coolant: (a) 2.0 L/min and (b) 3.5 L/min
Temperature System Heat transfer Reaction

ΔE k0 mol·m-2·
of coolant temperature coefficient rate constant b
KJ·mol -1 K-1·s-1
K K KW·m-2·K-1 mol·m-2·K-1·s-1
275.15 281.4096 0.1662 0.0723 1.0030
273.15 280.8802 0.1582 0.0688 0.8052
272.15 280.1633 0.1808 0.0635 0.9670 66.467 1.5754×1011
271.15 279.3663 0.1784 0.0589 1.0938
270.15 279.3233 0.1987 0.0585 0.9145
Table 2. Model parameters in Equation 29 and the correlated value of ΔE and k0
5.4.2 Influence of coolant flux

Experimental results showing the influence of the coolant flux on the hydrate formation rate
are plotted in Figure 10. It could be seen that the hydrate formation rate increases with the
increasing of coolant flux when the temperature is fixed. The effect of coolant flux on the
formation rate is more significant when the temperature of coolant is higher. When the
temperature of coolant is low enough, the further increase of coolant flux has little effect on
increasing hydrate formation rate. In this case, hydrate formation is controlled not by the
heat transfer, but by the intrinsic kinetics of hydrate formation.
5.5 Heat transfer dependence in quiescent hydrate dissociation

We performed a series of experiments to reveal the effect of heating on methane hydrate
dissociation in the quiescent reactor (Pang et al., 2009). Representative profiles of pressure,
temperature, and cumulative mole number of evolved methane during hydrate dissociation
are depicted in Figure 11, where hydrate was formed from 1920 g water of 2000 mg/L SDS,
the initial temperature of hydrate was set to 268.15 K, the temperature of the input hot water
was set to 298.15 K, and the flow rate was set to 2.0 liter per minute (LPM). As shown in
Figure 11, at the beginning of hydrate dissociation, the temperature of hydrate decreased
drastically and rapidly, which indicates a brief-but-rapid dissociation induced by the rapid
depressurization in a short time period. During this initial degassing event, the temperature
of sample typically dropped 3 to 7 K due to adiabatic cooling of methane gas expansion as
well as heat absorption of hydrate dissociation (Stern et al., 2001).
18 20
18
16
16
14
14
Consumed gas (mol)
Consumed gas (mol)

12
12
10
10
8 7.0 L/min
7.0 L/min 8
6 5.5 L/min 5.5 L/min
6
3.5 L/min 3.5 L/min
4 4 2.0 L/min
2.0 L/min
2 2
0 0
0 25 50 75 100 125 150 175 200 225 250 275 0 25 50 75 100 125 150 175 200 225 250 275 300 325
(a) (b)
Fig. 10. The effect of the coolant flux on hydrate formation rate at different coolant
temperatures: (a) 273.15 K and (b) 271.15 K
310 5 18
305 16
300 4 14
Evolved methane (mol)
295
12
Temperature (K)
Pressure (MPa)
290 3
10
285
8
280 Temperature 2
Pressure 6
275
270 1 4
265 2
260 0 0
0 100 200 300 400 500 600 700 800 900 100011001200 0 100 200 300 400 500 600 700 800 900 100011001200
Time (min) time (min)
(a) (b)
Fig. 11. Typical profiles of pressure, temperature and cumulative mole number of evolved
methane during hydrate dissociation, heated by 298.15 K hot water of a flux of 2.0 LPM in a
closed system
The temperature of hydrate and pressure of system increased sharply after the vent was
turned off as the hydrates dissociated rapidly. This brief but rapid hydrate dissociation is
driven mainly by the thermodynamic driving force, i.e., the difference between the
equilibrium pressure and the present system pressure. Figure 12(a) shows the comparison of
the experimental dissociation temperature/pressure locus with the equilibrium
temperature/pressure curve of methane hydrate. It can be seen that the experimental
pressure is obviously lower than the equilibrium pressure at this stage, which implies the
driving force is large.
Then a stage of buffered dissociation occurred when the hydrates was heated to approach
the melting point of ice. During this period, the sample temperature remains within a
narrow range near the ice point, rises slowly with the proceeding of the hydrate
dissociation, and the corresponding dissociation rate is very low. As shown in Figure 12(a),
the temperature/pressure locus is very close to the equilibrium temperature/pressure
curve, which implies that the hydrate is closely at equilibrium state at this buffering stage
and the thermodynamic driving force for hydrate dissociation is very low. At this stage, the
ice coming from the hydrate dissociation is heated to produce liquid water and much larger
amount of heat is required for the thaw of ice. This phase transition process buffers the
system temperature at nearly fixed temperature and pressure and the rate of heat transfer is
the key rate-limiting factor for the hydrate dissociation. Circone et al. (2003, 2004) and Stern
et al. (2003) also found this kind of thermal buffering during decomposing methane
hydrates by heating them from low temperature through the melting point of ice. After the
buffered dissociation stage, the rate of methane hydrate dissociation increases with the
increasing of temperature of heating water. Therefore, the rate of heat transfer is an
important factor that controls the rate of hydrate dissociation, especially during the
buffering stage. In this buffering region, the hydrate dissociation rate is very low, which is
obviously unfavorable to both the recovery of gas from hydrate in the NGH technology and
the gas production from in situ natural gas hydrates with heating method. Measures should
be taken to eliminate this effect.
13
Experimental temperature / pressure locus 100
12
-------- Equilibrium temperature/pressure line 90
11
Dissociation percentage (%)
10 80
9 70
Pressure (MPa)
8
60
7 289.15 K
50 298.15 K
6
5 40 303.15 K
4 308.15 K
30
313.15 K
3
20
2
1 10
0 0
255 258 261 264 267 270 273 276 279 282 285 288 0 100 200 300 400 500 600 700 800 900 100011001200
Temperature (K) Time (min)
(a) (b)
Fig. 12. (a) Comparison of the experimental dissociation temperature/pressure locus with
the equilibrium temperature/pressure curve of methane hydrate, and (b) The influence of
the temperature of heating water upon hydrate dissociation rate
Our previous work (Pang et al., 2009) demonstrated that the buffering effect can be
eliminated and the dissociation rate can be improved by increasing the temperature of
heating water or reducing the dissociation pressure. Figure 12(b) shows the influence of the
temperature of heating water upon the dissociation rate of hydrate, where the dissociation
rate is manifested by changing of cumulative dissociation percentage of hydrate with the
elapsed time. It can be seen that the thermal buffering period can be shortened obviously by
increasing heating water temperature, which suggests the important role of heat transfer in
controlling the dissociation of hydrate.
6. Heat transfer in hydrate bearing sediment

6.1 Heat effect during hydrate formation in sediment
Hydrate formation process in sediment is an exothermic process. The variation of
temperature is related to the hydrate formation rate and amount. For hydrate formation in
large scale of sediment, the fluctuations of temperatures at different locations will be
therefore different but will coincide with specified rules. A three-dimensional apparatus has
been used by our group recently to study the heat transfer process during hydrate formation
in sediment (Yang et al., 2010). Sixteen thermocouples were inserted into the sediment to
detect the variation of temperature, which were placed in different radius and different
depth of the reactor. Figures 13 and 14 show the pressure and temperature distribution
during the hydrate formation using the three-dimensional apparatus. From point A to B in
Figure 13, the top pressure decreases gradually, showing the process of hydrate formation.
At the same time, the temperatures at different locations in the reactor first rise to a
maximum value and then decreases, as shown in Figure 14. At the initial period, the hydrate
formation rate is high and temperatures rise due to the exothermic reaction. With the
decline of hydrate formation rate due to the decrease of the pressure driving force and pore
space, the rising trend of temperature weakens and the temperatures even decrease
gradually because the exothermic effect of hydrate formation can not make up the
refrigeration effect of the water bath. From point B to C, the pores of the sediment are
packed with formed hydrate, which resulting in that gas at the top of the reactor can hardly
permeate into the inner of sediment. Therefore, little hydrate forms and pressure and
temperature nearly keep constant in this stage. From point C to D, hydrate continues to
form again, which can be implied from the magnitude of pressure decrease in Figure 13 and
the temperatures increase in Figure 14. After point D, no hydrate formed and the top
pressure and temperature kept constant gradually.
8 A
B
6
Pressure (MPa)
C
D
0 5 10 15 20 25 30
Time (h)
Fig. 13. Variation of pressure of the reactor versus time during hydrate formation
During the whole process of hydrate formation, there are two stages (from A to B and C to
D) of rapid increase of temperature caused by the exothermic effect of hydrate formation.
The heat transfer during hydrate formation in sediment can then be characterized from the
appearance sequence of the temperature peak and its magnitude.
276.0
T1
275.5
T2
275.0 T3
Temperature (K)
T4
274.5
274.0
273.5
273.0
272.5
0 5 10 15 20 25 30
Time (h)
Fig. 14. Temperature distribution during hydrate formation with the same depth but
different radius: T1, T2, T3, T4 were placed in different radius of 132 mm, 99 mm, 66 mm, and
33 mm, respectively
6.2 Heat effect during hydrate dissociation in sediment

Depressurization, to depressurize the gas hydrates reservoir below the equilibrium pressure
at specified temperature, has been considered as one of conventional methods of producing
gas from hydrate bearing sediment. Heat effect will arise from the Joule-Thomson effect and
endothermic process of hydrate dissociation. For a typical hydrate depressurizing
decomposition in a three-dimensional environment (Su et al., 2010), three steps were
adopted. First, free gas is discharged until the pressure decreases to a specified value, which
is close to the equilibrium pressure at the current temperature. Afterward, gas is discharged
rapidly and the system pressure attains the experimental dissociation value quickly. During
the rapid depressurizing stage, there exists Joule-Thomson effect, and remarkable
temperature drop was observed. Hereafter, hydrate begins to decompose. The temperature
begins to rise under the heat transfer of air bath and endothermic process of hydrate
dissociation. The extent of temperature variation can be used to reflect the quantity of
hydrate decomposed. The difference between temperature value before and after rapid
depressurizing, ΔT, can be used to analyze the dissociation process. For the value of ΔT at
different positions, both Joule-Thomson effect and thermal effect of hydrate decomposition
contribute to ΔT. Since both are endothermic process, the quantity of hydrate decomposition
can be indicated by ΔT value. If assuming that hydrate in the reactor decomposing at the
same drive force and at the same apparent decomposition rate, the position sequence of
hydrate decomposition can also be connected with ΔT value. That is, the larger the ΔT is, the
earlier decomposition reaction begins. Hydrate at the top and bottom of the reactor
decomposes earlier than that in the inner of the reactor.
Thermal stimulation, to produce gas from hydrates by increasing the temperature of the gas
hydrates reservoir above the equilibrium temperature at specified pressure, is also an
important method to produce gas from hydrate bearing sediment. Selim and Sloan (1990)
found that the dissociation rate was a strong function of the thermal properties of the system
and the porosity of the porous medium. Kamata et al. (2005) applied thermal recovery
method to dissociate methane hydrate in hydrate sediment sample by hot-water injection
from one side, gas production from another side. It was found that temperature and
pressure in the sample fluctuated between stability region and decomposition region of
methane hydrate sample when temperature of the hot water was high. We (Yang et al.,
2010) studied the temperature distributions in the reactor during gas production by cyclic
injecting hot water method. Figure 15 shows the variation of temperatures with time at
different locations during the first cycle and the second cycle. It can be found that the overall
temperature trend increases with hot-water injection and decreases with gas production. At
initial stage of water injection, the temperatures rise slowly due to much heat is consumed
to warm the flowing channel of hot-water. Afterward, the temperatures at different
locations rise rapidly. It is noted that for the local positions of hydrate sample which are
away from the injecting well and near the reactor wall, the temperature-jump phenomenon
is not observed. It is known that the hot-water can hardly spread to those locations in a
shorter time. Strong heat exchange with environment may also occur on those positions near
the reactor wall with high conductivity. In contrast, at other locations, in view of low
conductivity of hydrate sample, the fluctuation of temperature mainly depends on the
seepage rate of hot-water in sediments instead of thermal diffusion. That is to say, the
seepage flow controls the heat transport in this process at these locations. In addition, the
temperature distribution also implied the injected hot-water distribution in the reactor to a
certain extent. The location of well has an important effect on the temperature distribution
in hydrate sediment, thereby affecting the gas production.
308 315
T1 310
304 T1
T2
300 305 T2
T3
T3
Temperature (K)
296 T4 300
Temperature (K)
T4
292 295
288 290
284 285
280 280
276 275
-5 0 5 10 15 20 25 30 35 40 45 50 -5 0 5 10 15 20 25 30 35 40 45 50
(a) first cycle (b) second cycle

Fig. 15.Variation of temperature with time at different locations with the same depth during
the first cycle: T1, T2, T3, T4 were placed in different radius of 132 mm, 99 mm, 66 mm, and 33
mm, respectively
7. Conclusion
In this chapter we have discussed heat transfer involved in formation/dissociation of gas
hydrates. The state of the art for the studies of the experimental measurement of thermal
conductivities of different kinds of gas hydrates, the mechanism and modeling of heat
transfer occurring in the growth of hydrate film at the guest/water interface, our
experimental study on macroscopic heat transfer in stirring reactors or flowing pipes, the
experimental and modeling studies on the heat transfer in quiescent reactors, and the
mechanism of heat transfer in hydrate bearing-sediment are summarized. We believe this
chapter will be helpful for readers to understand the critical position of heat transfer in the
hydrate formation/dissociation kinetics, favor the ongoing hydrate formation/dissociation
kinetic research, and promote the application of gas hydrate.
8. Acknowledgement
The financial supports of the NSFC (Nos. 20925623, 21006126), the Research Funds of China
University of Petroleum, Beijing (BJBJRC-2010-01), and the National Basic Research Program
of China (2009CB219504) are greatly appreciated.
9. References
Andersson P. & Ross R. G. (1983). Effect of guest molecule size on the thermal conductivity
and heat capacity of clathrate hydrates. J. Phys. C, Vol. 16, pp. 1423-1432.
Circone S., Stern L. A., Kirby S. H., Durham W. B., Chakoumakos B. C., Rawn C. J.,
Rondinone A. J. & Ishii Y. (2003). CO2 hydrate: synthesis, composition, structure,
dissociation behavior, and a comparison to structure I CH4 hydrate. J. Phys. Chem.
B, Vol. 107, pp. 5529-5539.
Circone S., Stern L. A. & Kirby S. H. (2004). The role of water in gas hydrate dissociation. J.
Phys. Chem. B, Vol. 108, pp. 5747-5755.
Cook J. G. & Laubitz M. J. (1983). In Thermal Conductiuity; Hust J. G., Ed.; Plenum: New
York, Vol. 17, pp. 745.
Cortes D. D., Martin A. I., Yun T. S., Francisca F. M., Santamarina J. C. & Ruppel C. (2009).
Thermal conductivity of hydrate-bearing sediments. J. Geophys. Res., Vol. 114, pp.
B11103.
deMartin B. J. (2001). Laboratory measurements of the thermal conductivity and thermal diffusivity
of methane hydrate at simulated in-situ conditions. Georgia Institute of Technology:
Atlanta, GA, USA.
Englezos P., Kalogerakis N. E., Dholabhai P. D. & Bishnoi P. R. (1987a). Kinetics of formation
of methane and ethane gas hydrates. Chem. Eng. Sci., Vol. 42, pp. 2647-2658.
Englezos P., Kalogerakis N. E., Dholabhai P. D. & Bishnoi P. R. (1987b). Kinetics of gas
hydrate formation from mixtures of methane and ethane. Chem. Eng. Sci., Vol. 42,
pp. 2659-2666.
English, N. J. & Tse, J. S. (2010). Perspectives on hydrate thermal conductivity. Energies, Vol.
3, pp. 1934-1942.
Freer, E. M., Selim, M. S. & Sloan, E. D. (2001). Methane hydrate film growth kinetics. Fluid
Phase Equilib., Vol. 185, pp. 65-75.
Gustafsson S. E., Karawacki E. & Khan M. N. (1979). Transient hot-strip method for
simultaneously measuring thermal conductivity and thermal diffusivity of solid
sand fluids. J. Phys. D: Appl. Phys., Vol. 12, pp.1411.
Gustafsson S. E., Karawacki E. & Chohan M. A. (1986). Thermal transport studies of
electrically conducting materials using the transient hot-strip technique. J. Phys. D:
Appl. Phys., Vol. 19, pp. 727.
Handa Y. P. & Cook J. G. (1987). Thermal conductivity of xenon hydrate. J. Phys. Chem., Vol.
91, pp. 6327-6328.
Huang D. Z. & Fan S. S. (2004). Thermal conductivity of methane hydrate formed from
sodium dodecyl sulfate solution. J. Chem. Eng. Data, Vol. 49, pp. 1479–1482.
Huang D. Z., Fan S. S. & Shi L. (2004). Thermal conductivity of gas hydrate. Huaxuetongbao,
Vol. 10, pp. 737-742. (in Chinese)
Kamata Y., Ebinuma T., Omura R., Minagawa H., Narita H., Masuda Y. & Konno Y. (2005).
Decomposition experiment of methane hydrate sediment by thermal recovery
method. Proceedings of the 5th International Conference on Gas Hydrates, Trondheim,
Norway, pp. 81-85.
Krivchikov A. I., Gorodilov B. Y., Korolyuk O. A., Manzhelii V. G., Conrad H. & Press W.
(2005). Thermal conductivity of methane-hydrate. J. Low Temp. Phys., Vol. 139, pp.
693–702.
Krivchikov A. I., Gorodilov B. Y., Korolyuk O. A., Manzhelii V. G., Romantsova O. O.,
Conrad H., Press W., Tse J. S. & Klug D. D. (2006). Thermal conductivity of Xe
clathrate hydrate at low temperatures. Phys. Rev. B, Vol. 73, pp. 064203.
Li D. L., Liang D. Q., Fan S. S. & Peng H. (2010). Estimation of ultra-stability of methane
hydrate at 1 atm by thermal conductivity measurement. J. Nat. Gas. Chem., Vol. 19,
pp. 229-233.
Lin W., Chen G. J., Sun C. Y., Guo X. Q., Wu Z. K., Liang M. Y., Chen L. T. & Yang L. Y.
(2004). Effect of surfactant on the formation and dissociation kinetic behavior of
methane hydrate. Chem. Eng. Sci., Vol. 59, pp. 4449-4455.
Lunden A., Trolle U., Azoulay M. & de Pablo J. (1986). Determination of kinetics and
thermal conductivity of hydrates simultaneously by using thermogravimetry and
transient hot strip method. Thermochimica Acta, Vol. 105, pp. 369-373.
Ma C. F., Chen G. J. & Guo T. M. (2002). Kinetics of hydrate formation using gas bubble
suspended in water. Science in China (Series B), Vol. 45, pp. 208-215.
Makogon Y. F. (1997). Hydrates of Hydrocarbons. Tulsa, Oaklahoma: Penn Well books.
Mochizuki T. & Mori Y. H. (2006). Clathrate-hydrate film growth along water/hydrate-
former phase boundaries-numerical heat-transfer study. J. Crystal. Growth, Vol. 290,
pp. 642-652.
Moon C., Taylor P. C. & Rodger P. M. (2003). Molecular dynamic study of gas hydrate
formation. J. Am. Chem. Soc., Vol. 125, pp. 4706-4707.
Mori Y. H. (2001). Estimating the thickness of hydrate films from their lateral growth rates:
application of a simplified heat transfer model. J. Crystal. Growth, Vol. 223, pp. 206-
212.
Ohmura R., Kashiwazaki S. & Mori Y. H. (2000). Measurements of clathrate-hydrate film
thickness using laser interferometry. J. Crystal. Growth, Vol. 218, pp. 372-380.
Ohmura R., Matsuda S., Uchida T., Ebinuma T. & Narita H. (2005). Clathrate hydrate crystal
growth in liquid water saturated with a guest substance: observations in a methane
+ water system. Cryst. Growth Des., Vol. 5, pp. 953-957.
Pang W. X., Chen G. J., Dandekar A., Sun C. Y. & Zhang C. L. (2007). Experimental study on
the scale-up effect of gas storage in the form of hydrate in a quiescent reactor. Chem.
Eng. Sci., Vol. 62, pp. 2198-2208.
Pang W. X., Xu W. Y., Sun C. Y., Zhang C. L. & Chen G. J. (2009). Methane hydrate
dissociation experiment in middle-sized quiescent reactor using thermal method.
Fuel, Vol. 88, pp. 497-503.
Peng B. Z., Chen G. J., Sun C. Y., Yang L. Y. & Luo H. (2008). Is the gas hydrate film growth
controlled by intrinsic kinetic or heat transfer? Proceedings of the 6th international
conference on gas hydrates, Vancouver, British Columbia, Canada.
Peng B. Z., Dandekar A., Sun C. Y., Luo H., Ma Q. L., Pang W. X. & Chen G. J. (2007).
Hydrate film growth on the surface of a gas bubble suspended in water. J. Phys.
Chem. B, Vol. 111, pp. 12485-12493.
Peng B. Z., Sun C. Y., Chen G. J., Yang L. Y., Zhou W. & Pang W. X. (2009). Hydrate film
growth at the interface between gaseous CO2 and sodium chloride solution. Science
in China Series B: Chemistry, Vol. 52, pp. 676-682.
Rogers R. E., Zhong Y., Etheridge J. A., Arunkumar R., Pearson L. E. & Hogancamp T. K.
(2005). Micellar gas hydrates storage process. Proceedings of the 5th international
conference on gas hydrates, Trondheim, Norway, pp. 1361-1365.
Rosenbaum E. J., English N. J., Johnson J. K., Shaw D. W. & Warzinski R. P. (2007). Thermal
conductivity of methane hydrate from experiment and molecular simulation. J.
Phys. Chem. B, Vol. 111, pp. 13194–13205.
Ross R. G. & Andersson P. (1982). Clathrate and other solid phases in the tetrahydrofuran–
water system: thermal conductivity and heat capacity under pressure. Can. J. Chem.,
Vol. 60, pp. 881-892.
Ross R. G., Andersson P. & Backstrom G. (1981). Unusual PT dependence of thermal-
conductivity for a clathrate hydrate. Nature, Vol. 290, pp. 322.
Saito K., Kishimoto M., Tanaka R. & Ohmura R. (2011). Crystal growth of clathrate hydrate
at the interface between hydrocarbon gas mixture and liquid water. Crystal. Growth
& Design, Vol.11, pp. 295-301.
Saito K., Sum A. K. & Ohmura R. (2010). Correlation of hydrate-film growth rate at the
guest/liquid-water interface to mass transfer resistance. Ind. Eng. Chem. Res., Vol.
49, pp. 7102-7103.
Selim M. S. & Sloan E. D. (1990). Hydrate dissociation in sediments. SPE Reserv. Eng., Vol. 5,
pp. 245-251.
Sloan E. D. & Koh C. A. (2007). Clathrate hydrate of natural gases. 3rd ed, CRC Press: Boca
Raton, FL.
Stern L. A., Circone S., Kirby S. H. & Durham W. B. (2001). Anomalous preservation of pure
methane hydrate at 1 atm. J. Phys. Chem. B, Vol. 105, pp. 1756-1762.
Stern L. A., Circone S., Kirby S. H. & Durham W. B. (2003). Temperature, pressure, and
compositional effects on anomalous or “self” preservation of gas hydrates. Can. J.
Phys., Vol. 81, pp. 271-283.
Stoll R. D. & Bryan, G. M. (1979). Physical properties of sediments containing gas hydrates.
J. Geophys. Res., Vol. 84, pp. 1629-1634.
Su K. H., Sun C. Y., Yang X., Chen G. J. & Fan S. S. (2010). Experimental investigation of
methane hydrate decomposition by depressurizing in porous media with 3-
Dimension device. J. Nat. Gas Chem., Vol. 19, pp. 210–216.
Sun C. Y., Chen G. J., Ma C. F., Huang Q., Luo H. & Li Q. P. (2007). The growth kinetics of
hydrate film on the surface of gas bubble suspended in water or aquesou surfactant
solution. J. Crystal. Growth, Vol. 306, pp. 491-499.
Taylor C. J., Miller K. Y., Koh C. A. & Sloan E. D. (2007). Macroscopic investigation of
hydrate film growth at the hydrocarbon/water interface. Chem. Eng. Sci., Vol. 62,
pp. 6524-6533.
Tse J. S. & White M. A. (1988). Origin of glassy crystalline behavior in the thermal properties
of clathrate hydrates: A thermal conductivity study of tetrahydrofuran hydrate. J.
Phys. Chem., Vol. 92, pp. 5006-5011.
Tzirita A. (1992). In situ detection of natural gas hydrates using electrical and thermal properties.
Ph.D. thesis, Texas A&M Univ., College Station.
Uchida T., Ebinuma T., Kawabata J. & Narita H. (1999). Microscopic observations of
formation processes of clathrate-hydrate films at an interface between water and
carbon dioxide. J. Crystal. Growth, Vol. 204, pp. 348-356.
Uchida T., Ikeda I. Y., Yakeya S., Ebinuma T., Nagao J. & Narita H. (2002). CO2 hydrate film
formation at the boundary between CO2 and water: effects of temperature, pressure
and additives on the formation rate. J. Crystal. Growth, Vol. 237, pp. 383-387.
Vysniauskas Y. & Bishnoi P. R. (1983). A kinetic study of methane hydrate formation. Chem.
Eng. Sci., Vol. 38, pp. 1061-1072.
Waite W. F., Gilbert L. Y., Winters W. J. & Mason D. H. (2005). Thermal property measurements
in Tetrahydrofuran (THF) hydrate and hydrate bearing sediment between -25 and +4 0C,
and their application to methane hydrate. Paper presented at 5th International
Conference on Gas Hydrates, Tapir Acad., Trondheim, Norway.
Waite W. F., Stern L. A., Kirby S. H., Winters W. J. & Mason D. H. (2007). Simultaneous
determination of thermal conductivity, thermal diffusivity and specific heat in sI
methane hydrate. Geophys. J. Int., Vol. 169, pp. 767–774.
Xie Y. M., Guo K. H., Liang D. Q., Fan S. S. & Gu J.M. (2005). Steady gas hydrate growth
along vertical heat transfer tube without stirring. Chem. Eng. Sci., Vol. 60, pp. 777-
786.
Yamamoto Y., Ohtake M., Kawamura T., Tsuji T. & Tsukada Y. (2008). Thermal conductivity
measurement of methane hydrate with soil. Proceedings of the 6th international
conference on gas hydrates, Vancouver, British Columbia, Canada.
Yang X., Sun C. Y., Yuan Q., Ma P. C. & Chen G. J. (2010). Experimental study on gas
production from methane hydrate-bearing sand by hot-water cyclic injection.
Energy & Fuels, Vol. 24, pp. 5912–5920.
Zhong Y. & Rogers R. E. (2000). Surfactant effects on gas hydrate formation. Chem. Eng. Sci.,
Vol. 55, pp. 4175-4187.
25
Progress Works of High and

Super High Temperature Heat Pipes
Wei Qu
Institute of Engineering Thermophysics, Chinese Academy of Sciences
North 4th Ring Road No.11, Beijing 100190
China
1. Introduction
Heat pipe is a high efficiency heat transfer element, depends on the evaporation,
condensation and circulation of inside working fluid. The good performance of a heat pipe
is due to that the working fluid evaporation of latent heat is generally large, so there needs
no much working fluid and the circulation flow rate is usually small, and the flow resistance
is small, when a heat pipe works. The temperature and pressure inside a heat pipe are
nearly at uniform levels through the evaporator to the condenser, so a heat pipe has a high
heat transfer ability at smaller temperature difference. The heat pipe has many other
advantages, such as the heat fluxes of evaporator and condenser can be changed easily by
adjusting the heat transfer lengths and areas etc..
The working fluid of a high temperature heat pipe is generally alkali metals, and their
evaporations of latent heat are very large, and the saturation pressure is smaller at high
temperature, then the case strength is not a problem. The high and super high temperature
heat pipe can work over 500°C, 1000°C, 1500°C or much higher. Sodium heat pipe can
operate during 500~1100°C, and lithium heat pipe can work over 1200°C (Ma et al, 1983).
The high and super high temperature heat pipes have wide applications in space thermal
control, solar storage power plant, nuclear thermal control, the constant industry stoves etc..
Usually, the heat pipes between 500~1100°C use the sodium as working fluid, they are
called as the high temperature heat pipe; while the heat pipes over 1200°C, utilize lithium as
the working fluid, they are named as the super high temperature heat pipes. Commonly,
1000°C can be supposed to divide the boundary of high and super high temperature heat
pipes.
The first heat pipe in China was made in 1972, those stainless steel and sodium heat pipes
realized the high temperature stove to make the semiconductor technology reliable. As
shown in figure 1, the super high temperature heat pipe of W-Li by USA was tested in 1966
(Busse, 1992). The basic parameters and the startup process were given beside the figure,
different temperature level tells the performances of the lithium heat pipe.
For the nuclear reactor cooled by the high temperature heat pipes, the fuel bars and the high
temperature heat pipes are designed to be coupled together. By the high temperature heat
pipes, the heat from the pile core is transferred to the heat exchanger, the technology of high
temperature thermal radiation is realized (Zhang et al, 2009). The high temperature heat
pipes connect the nuclear power suppliers to the outside and transfer the large amount of
heat. When the space nuclear power suppliers work at full, the temperature of heat core can
reach 3000~3300K, however, the maximum temperature to be controlled is 800K. So, the
high temperature heat pipe is the key problem to be solved for the whole system. Such
aspect relates also the miniaturization, and performance optimum of the high temperature
heat pipe, and the shape adaptability and coupling with the high temperature radiator etc..
Tilt angle change:

1: -90°; -45°; 0°
2: 5°
3: 10°; 5°
Inner diameter: 8mm;
Outer diameter: 10mm;
Length: 115mm;
Groove number and size: 30,
0.4mm×0.4mm;
Filling amount: 0.8g;
Surplus liquid: 1.4cm3
Fig. 1. Temperature levels of W-Li heat pipes when the heat flux is kept constant (Busse,
1992)
For the past few decades, the sodium and lithium heat pipes were developed a lot. There
were different application in many fields as, the thermal control of super high Mach
spacecraft, the heat storage and heat transfer of solar power plant, the heat recycling and
application of high temperature gas discharge, the thermal control in nuclear energy, the
constant temperature of industry stove, etc., in some situations, they are necessary and
cannot be realized by other ways.
Composite material of
high temperautre
Cross-section A
Structure of A
Heat pipe cross-section
of“D”cross-section
Fig. 2. Heat protection of high temperature heat pipe( by NASA) (David et al, 1998)
For the super high Mach spacecraft, the conduction heat protection is a kind of initiative
method to realize thermal control, it is different from the heat protection of conventional
surface ablation, or by the new material enduring much higher temperature level (Boman et
al, 1990), as shown in figure 2. The “D” shape high temperature heat pipes were bended to
Progress Works of High and Super High Temperature Heat Pipes 505
couple the outside composite material of high temperature. During flight, the friction heat at
cross-section A outside the cover will be conducted to the non-stagnant point of the cover by
heat pipe. So the temperature of the stagnant points can be kept at lower level (Faghri et al,
1992, David et al, 1998).
There will be large aerodynamic heat at the leading edge of the flying supersonic aircraft
and space shuttle due to the friction between the super high Mach flight and the ambience,
especially for the stagnation point (Jiang et al, 2008).
In this chapter, the progress works of the high temperature and super high temperature heat
pipes are summed up. Owing to the different levels of temperature and pressure, the
strength of the heat pipe shell, the compatibility of material and wick should be considered
and designed. Generally, the working fluid is solid, there will be startup problem, the
startup is analyzed. The heat transfer limits are important to the heat pipe operation, the
possible limitations of high and super high temperature heat pipes are calculated and
discussed. The technology is the key problem to realize the high performance heat pipe, the
distillation and non-distillation methods are introduced. At last, the technology development
of heat pipe is introduced.
2. The compatibility of case material and working fluid

Because the material of high temperature is generally dense, and the fabricating of heat pipe
will guarantee the welding connection to have enough thickness and strength, the eligible
welding, it is not supposed to occur the reported “sodium and water reaction” (Zhuang et
al, 1998,2000). For the failure of high temperature heat pipe, firstly, the purity of material
and cleansing belong to the treatment method, no problem to be controlled, then, the real
problem turns to the key cause of the compatibility (Jacobson et al, 1984).
When the metal contacts the ambient medium, the metal erosion may occur due to the
chemical reaction and electrochemistry. The erosion mechanism of alkali metal heat pipe is
important to take measures to keep the performance of high temperature heat pipes and
reliability.
2.1 The high temperatue micro cell erosion

The erosions of oxidation and dissolution are introduced a lot in some reports or books (Ma
et al, 1984; Zhuang et al, 1998; Zhang et al, 1987; Jacobson et al, 1984). Here, there are two
points to be emphasized. (1) For the outside of the heat pipe, if the ambient oxygen
concentration is significant, the heat pipe surface will be oxidized directly, the higher
temperature is, the much more erodible is the surface. However, if a dense layer of oxides is
formed, the existence of oxide will increase the anti-erosion ability of the outside surface.
However, if there is a little alkali produced, they will dissolve the protection layer of the
base metal, then the reaction will be increased. (2) For the inside of the heat pipe, some
components of the composite metal may also dissolve in the liquid alkali metal. The inner
surface will form rust speckles, so the effective thickness of the heat pipe wall is decreased.
The alkali metal evaporates continuously, the dissolved speckles will deposit and jam up the
liquid flow passages. The failure reason is liquid return resistance is increased a lot instead
of wall metal consumption.
Here, another important erosion is put forward, it is the most possible in high and super
high temperature heat pipes. The components and purity of metal are important for the
selection of material. The metal in liquid working fluid is easy to be erodible due to
electrochemistry (Li et al, 1987). The reason is, the metal of industry purity contains the
impurity material, and they have higher electricity potential. When they contact the liquid
working fluid, there form many micro cathodes on the metal surface. The metal of lower
electricity potential becomes the anodes. Then there are many many micro cells to make the
metal erodible. For higher temperature level, such erosion is very remarkable, they are
called the micro cell erosion.
The micro cell erosion belongs to a kind of the electrochemistry erosion. Such process consists
of three parts. Firstly, the metal is dissolved to be erodible on anodes, the oxidation reaction
occurs. Secondly, the electrons flow from anodes to cathodes. Thirdly, the electron flow is
accepted by the material in the working fluids, the deoxidization reaction occurs on cathodes.
For the cathodes, there are many kinds of material on cathodes, such as micro amount of H+
and O2 etc.. This is the main reason to select and treat the working fluids to be purified.
Φca
Φca
0
i
Φan
Φan(φ-)
Fig. 3. Schematic of polarized curve for electrochemistry erosion of micro cell

The value of electrode potential stands for the activity of a metal. If “φ” expresses the
electrode potential at arbitrary concentration, “i” is the electricity current, the polarization
curve is shown in figure 3, the curve can help to denote the electrode reaction. Generally for
the anodes, the metal will lose the electrons and become the positive ionic, the polarization
is lower; while for the cathodes, the polarization is higher due to the vapor and (or) gas.
Generally, there are two reasons to trigger the micro cell erosion. One is that, the inner
resistance of the erodible cells will increase with the amount of the hardly dissolved salts,
and the erosion electricity decreases continuously, then the erosion speed is decreased. The
other is that, the erosion cells are polarized, and the electromotive potential is decreased.
In summary, increasing the inner resistance and polarization of the erosion micro cells, the
erosion speed will decrease. In addition, the higher erosion characteristics relates with the
impurity quantity and the surface smoothness. The more impurity quantity is, the worse
anti-erosion is; the smoother is the surface, the better anti-erosion is.
The impurity with higher electricity potential in a metal, or the contact with the metal of
higher electricity potential will increase the erosion. So the impurity rate and the cleanness
process, fabrication process etc. should be paid attention essentially for the high and super
high temperature heat pipes.
2.2 The property and strength of high temperature alloy

Table 1 and 2 list the physical and mechanical properties of some high temperature alloys.
highest operating
material melting point(°C) density(kg/m3)
temperature(°C)
GH3030 1374~1420 8.40 800
GH128 1340~1390 8.81 950
SS304 1454 7.85 800
Inconel600 1354~1413 8.47 1093
Table 1. Physical properties
pull properties
Test Anti-pull Yielding Elongating Enduring Enduring
material rate
temp.(°C) strength strength strength time
ób(MPa) ó0.2(MPa) δs(％) (MPa) (h)
GH3030 20 730 — 44 — —
700 266 — 72 103 100
GH128 20 891 350 54 — —
950 198 — 99 42 100
SS304 20 520 205 40 — —
— — — — — —
Inconel600 20 671 246 41 — —
1093 — — — 9.7 100
Table 2. Mechanical properties
The heat pipe shell strength is considered during operation. The heat pipe shell strength is
calculated by (GB 9222-88, 1988),
PDin (1)
bwa =
2[σ ]
The most pressure of the heat pipe shell is derived as,
bwa 2[σ ] (2)

P=
Din
In equation (2), P is the inner pressure of the heat pipe; Din expresses the inner diameter of
the heat pipe; bwa is the thickness of the thinnest wall; [σ] is the permitted stress.
The end cap strength is checked by the most permitted pressure, i.e. the gauge pressure as,
Sy 2[σ ]ϕ
P= (3)
YDn + Sy
In equation (3), Sy is the effective thickness of the end cap wall, the weakening coefficient of
the end cap is φ=0.9, Dn is the diameter of the heat pipe end cap. The shape coefficient Y is
calculated by,
1 D
Y = [2 + ( n )2 ] (4)
6 2 hn
In equation (4), hn is the highness inside the heat pipe end cap.
For the material of GH128, the pull strength is 850MPa at room temperature. While at
temperature 1000°C, the yielding strength is 350MPa; the pull strength is 140MPa, the
extension rate is 85%. And at this temperature, the enduring life is 200 hours at 30MPa, the
extension rate is 40%. For the operating temperature of 1000°C, the heat pipe shell and end
cap have the lowest pull strength and yielding strength, the inside alkali metal vapor
pressure is the highest. So, if the worst situation can satisfy the demand, then the lower
temperature can also meet the strength demand. The explosion coefficient is set 2.0, the
yielding coefficient is given as 1.5, the calculated results are shown in table 3. By the results,
the design can meet the strength demands.
Enduring most Enduring most Vapor

Inner groove [σ]
pressure pressure of end, pressure
height (mm) (MPa) shell/end cap (MPa) shell/end cap (MPa) (MPa)
φ8×1 0.5 70 10/2.86 15.75/4.5 2.71
φ8×2 0.5 70 23/6.67 31.50/9.0 2.71
Table 3. The strength of high temperature heat pipe GH128-Na (1000°C)
2.3 The material selection of super high temperature heat pipe

For the situations of coupling the heat pipe and the outside material, the expanding
characteristics of both them at operating higher temperatures can lead to the thermal stress
due to the material distortion. If this problem is not considered, the crevice between the heat
pipe and the outside material can worsen the heat transfer. The linear expanding coefficients
of some high temperature materials are listed by table 4.
material 20~100°C 20~400°C 20~600°C 20~800°C 20~1000°C

GH3030 12.8 15 16.1 17.5 —
GH128 11.2 12.8 13.7 15.2 16.3
304 18.2 — — — —
Inconel600 13.3 14.5 15.3 16.1 —
Table 4. Linear expanding coefficient (10-6/°C)
In order to decrease the crevice between the heat pipe and the outside material, the material
niobium is the best candidate, owing to its smallest linear expanding coefficient. Niobium
belongs to the VB list in the chemistry element periodic table. Its melting point is 2468°C, the
strength is high during 1093~1427°C. Its thermal expanding coefficient is 7.2×10-6/°C, the
density is similar to the steel, the strength can be kept at 1649.9°C, and it can endure a
certain mechanical deformation. Niobium is stable in eroding medium, and has good
thermal conductivity. Based on the good performance of physical and chemical properties,
the niobium base alloys are best candidate for the super high temperature heat pipes.
However, such alloy has lower anti-oxidization. As shown in figure 4, the niobium bars and
strips are products by a certain company.
Fig. 4. Niobium sticks and straps, made by a certain company

For the heat leading problem of super Mach flight, the used heat pipes need to be coupled
with the outside material. The thermal expanding coefficients of both them should be
considered. The contact thermal resistance due to crevice by the different expanding can fail
the heat leading design.
2.4 Alkali metal lithium

Alkali metals are commonly used to be the working fluid of high and super high
temperature heat pipes, such as lithium, sodium, potassium, and their alloy. There are many
advantages to select alkali metals. Alkali metals can work at high temperature, they have
high latent heat of vaporization, they have stable characteristics and lower saturation vapor
pressure and good thermal conductivity.
At room temperature, alkali metals are solid states. They will be oxidized in air. So the heat
pipe filling process should keep them from contacting the air. Additionally, the purity of
alkali metals should be guaranteed to fill the heat pipe, there is no other oxidization.
For the lithium, the obtained physical, thermodynamic properties, the erosion property etc.
are extremely not consistent, especially for the high temperature situations. The
experimental results are different for researchers, the tentative explanations can be owing to
the different purity of lithium materials. The other reason may be due to the different
accuracies used. The key reason may be that the lithium is not purified carefully before
filling. Additionally, lithium is active to react with some impurities. The impurities have
very effects on the properties.
Lithium is the lightest metal. Its density is from 0.531~0.534g/cm3, it can be believed to be
0.533g/cm3. The melting point is from 179~186 °C, it can be supposed to be 180.5±0.1°C. The
saturation pressure of solid and liquid lithium (Ps and Pl, Pa) can be obtained by theory and
experiments (Gelishen, 1966). Between the temperature T=298~453.7K, there is,
8551.18
lg( Ps 133.3) = 14.2121 − − 3.79295lg T + 6.7167 × 10 −3 T (5)
T
If the temperature T=453.7~1350K, then the relation is,
8442.53
lg( Pl 133.3) = 10.1184 − − 1.64098lg T + 2.5968 × 10 −4 T (6)
T
The simple equation of saturation pressure of liquid lithium is as,
8143
lg( Pl 133.3) = 8.0 − (7)
T
When the temperature is between 700~1400 K, equation (7) coincides with equation (6).
2.5 The compatibility

Eliminating the fabrication factors, the compatibility of high and super high temperature
heat pipes are due to the micro cell erosions, the shell and wick materials dissolve in the
working fluids. In addition, the micro cell erosion can make the inner surface be granulating
erosion and make the shell wall thinner. The temperature level will influence the
compatibility essentially. Busse found that for tungsten and rhenium alloy-lithium heat
pipe, the heat pipe longevity is several years, one year and one month respectively,
corresponding to the temperatures 1600°C, 1700°C and 1800°C (Busse 1992). The effects of
temperature level to the longevity are very obvious.
Table 5 shows the general results of compatibility, this is the basic principle to select the
couple of shell material and alkali metal.
SS Ni Ni alloy W Ta Mo Re Ti Nb
Lithium × × × √ √ √ √ × √
Sodium √ √ √ — — — — × √
√: compatibility tested; ×: non-compatibility tested
Table 5. General results of compatibility
Fig. 5. Positions on element periodic table of materials suitable for super high temperature
heat pipes
Considering all the important factors, the high temperature material, the above mentioned
mciro cell erosion, high temperature alloy properties, the compatibility with the alkali
metals, and the experimental results, figure 5 gives the selected heat pipe materials, signed
as “ellipse” in the chemistry element periodic table. From 4B~7B rows, tungsten, tantalum,
molybdenum, rhenium and niobium are good candidates to the high and super high
temperature heat pipes.
In the mean time, the material selection should consider the material machining properties
and availability, the price and other factors.
3. Startup analysis of alkali metal heat pipe

For alkali metal heat pipes, commonly, the working fluid in heat pipes is solid state. The
heat transfer into the evaporator make the solid working fluid melt, then one equivalent
heating section is formed. If the heat pipe is started up by heating one end, such as the heat
pipe has only one end heating section for the high mach stagnant point. As shown in figure
6, when the solid working fluid is melted completely the temperature distribution along the
heat pipe is given.
T
( °C )
Tend
Δt2
97.72
Δt1
25.00
0 x0 200 x, mm
Fig. 6. Temperature distribution along the heat pipe when the solid working fluid is melted
completely
For one concrete high temperature heat pipe, the cross-section area is 0.7536cm2. At 700°C,
the thermal conductivity of heat pipe shell material is 25W/(m.°C). If the startup power is
50W, the axial temperature difference will reach 26°C/mm by Fourier law. Obviously, there
will be bigger temperature difference along the axial direction. When the heat pipe is started
up, if the melted working fluid is not enough to make all the solid working fluid melt
wholly, then the startup will fail. In order to control the highest temperature of the stagnant
point, the heat applied should be lower than one level.
3.1 Analysis of startup time

For the horizontal sodium heat pipe, the ambience temperature is 25°C, then the sodium is
solid before startup. Between the temperature range 25~97°C, the thermal conductivity and
the heat capacity of sodium are considered as constants. And from 25°C to 800°C, the heat
pipe shell is taken as constant. The axial conduction of the wick is ignored.
The sodium should be heated to melt completely, from room temperature 25°C to the
melting point 97.72°C. The solid working fluid inside heat pipe is assumed to distribute
uniformly. The thermal capacity of the heat pipe is Ctube=53J/K, the thermal capacity of the
working fluid is CNa=3.14 J/K. For the sodium, the melting latent heat is Lmelt=113kJ/kg, the
latent heat of evaporation is hfg=4215kJ/kg. The heat to evaporate the working fluid should
be large than that, the heat to increase the temperature to the melting point and the heat to
melt the solid working fluid completely in the condenser, there is,
x ⎛ x ⎞
W1 = h fg m = ⎜1 − ⎟ ⎡(Ctube + C Na ) ΔT1 + mLmelt ⎤⎦ (8)
200 ⎝ 200 ⎠ ⎣
When the solid working fluid melt wholly, the temperature distribution is given in figure 4.
∆T1 is the temperature difference between the melting point to the room temperature. x0 is
the needed length to evaporate the working fluid. For the given sodium heat pipe A without
groove, x0 = 58.6mm; while for the heat pipe B with groove, x0 = 33.24mm by equation (8).
It is assumed that there is no heat transfer between the heat pipe condenser and the
ambience. Then the startup heat transfer is estimated as,
x ΔT
W2 = (Ctube + C Na ) ΔT1 + mLmelt + Ctube 2 (9)
200 2
∆T2/2 is the mean temperature difference between the evaporator temperature and the
sodium melting point.
For the evaporator, there is only the axial conduction along the heat pipe shell. The heat
transfer rate is given by Fourier law as
Q = −λ A dT dx (10)
Considering equations (8) and (10), the relation between the input power and the end
temperature can be obtained. The startup time is given also by equation (8) and (10) as,
t = W2 Q (11)
3.2 Results of startup time

From the data of the sodium melting heat with pressure and temperature, it is known that
the melting point changes little with the pressure. The melting point and melting heat are
taken for one standard atmosphere pressure to calculate the startup time. The startup power
should be controlled not to exceed one value, which can make the evaporator dryout before
all the solid working fluid is melted completely.
30 45
388s
40 176s
25 406s
35 189s
430s
20 30 204s
461s
25 225s
Q(W)
Q(W)
15 505s
20 255s
570s
10 15 299s
679s
10 372s
5 894s
516s
5
1528s
942s
0 0
100 200 300 400 500 600 700 800 900 1000 0 200 400 600 800 1000 1200
o o
T( C) T( C)
(a) (b)
Fig. 7. Relations of heat transfer rate, startup time and temperature. (a) with screen wick,
without groove (b) with screen wick and with grooves
For the sodium heat pipe A and B, the results are shown in figure 7. At the same startup
heat transfer rate, 20W, the startup time of A and B is 450S and 290S respectively.
4. Technology control of alkali metal heat pipe

The performance of heat pipe depends on the fabrication technology. The charging process
should guarantee the vacuum level before filling. The working fluid quantity charged
should be controlled. The working fluid has enough purity, the oxidization and impurities
are at the endurable levels, the seal is soldered and guaranteed etc..
4.1 The vacuum level control

Commonly, if the pressure is less one atmosphere, 1.01×105 Pa , the vacuum is divided by
several regions, as shown in table 6 (Zhang et al,1987).
Density of Mean free

Pressure
Region molecule number, journey,
(Pa)
n(cm-3) λ(cm)
Little vacuum 1.01×105~103 ~1018 10-4
Low vacuum 103~10-1 ~1015 10-1
High vacuum (HV) 10-1~10-6 ~1010 103
Super high vacuum (UHV) 10-6~10-12 ~106 109
Extreme high vacuum (XHV) <10-12 <102 >1012
Table 6. The vacuum region partitions

For the high and super high temperature heat pipes, the vacuum region had better reach the
levels of HV, if the vacuum is UHV or XHV, the technology will last long and the cost is
increased a lot. If the vacuum is little or low, then the heat pipes have worse performances
as shown in figure 8.
Fig. 8. The heat pipes have worse performances if the vacuums was low
4.2 The distillation technology

The distillation technology can make the alkali metal melt and evaporate. By controlling the
temperature of distillation, the alkali metal is purified a lot, then the liquid alkali metal is
charged into the candidate heat pipe. Such method keeps the system to be active vacuum,
the vacuum equipment works continuously. This can keep the alkali metal purity, not to be
oxidized by a little leakage air.
Probe of liquid
surface
Argon
Cooling water
Valve Distillation
Sodium tank tank
U tube Argon
Cooling trap Vacuum
Condensation
tube Connection equipment
Vent pipe
Heat pipe
Fig. 9. Distillation technology of high temperature heat pipe

As shown in figure 9, the charging system consists of sodium tank, U tube, distillation tank,
connection, cooling trap and vacuum equipment. The argon can protect the alkali metal in
the sodium tank. The charging process includes two steps. Firstly, the set amount of alkali
metal is filled from the sodium tank to the distillation tank. Secondly, the alkali metal is
distilled and charged into the heat pipe.
For sodium, after pumping the system and the vacuum is permitted, the distillation tank is
heated to a certain temperature, in the mean time the other part of the system has different
temperature, such as, the outlet of the condensation tube should be controlled between
150~200°C. The much higher temperature can make sodium vapor be pumped into the
vacuum equipment. The much lower temperature will lead to the higher viscosity of liquid
sodium, then the small vent pipe can be jammed. Based on the same reasons, the connections,
heat pipe, especially for the vent pipe also should be heated to about 200°C. After the other
parts of the system reach the set temperature, the distillation tank is heated to a temperature
between 480~500°C, and this temperature is kept constant to distill the sodium. The
temperatures at every part are monitored. If the sodium is vaporized totally, the temperature
of distillation tank will increase a lot, then stop heating the distillation tank.
Obviously, the distillation technology is complicated a little, and the consumptions of time,
water and power are large. Once, only one heat pipe can be charged. And the after-
treatment is also complicated, the sodium remains in the tubes are hard to be cleaned up.
4.3 The non-distillation technology

In order to make the charging process simple and several or many heat pipes can be filled
simultaneously, the three-path-equipment of alkali metal charging was invented. As shown
in figure 10.
The non-distillation charging system is composed of the vacuum equipment, the transparent
glove chamber with argon protection, the valves of super high vacuum and tubes. There are
three paths, can realize three alkali metal heat pipes charging simultaneously. For example,
the flange of the first path is disconnected, the empty lower tank is put downward into the
transparent glove chamber with argon protection. Also, the heat pipe end is inserted into
one small tube, by which the air is replaced by argon. In this glove chamber, the alkali metal
is cut, weighed and put into the tank, which outlet is set stainless steel screen. Then the tank
with alkali metal is lifted to couple the flange and the system is closed by bolts and valves.
During this process, the main tube of the system is also blowed by argon. The system is
pumped some time, and the argon in the system is evacuated as much as possible.
Here it is pointed out that the argon in the heat pipe is pumped out through the tank of alkali
metal, the argon will cross the alkali metal by the aperture passage. By the bypass designed
near the outlet of the alkali metal tank, the vacuum of heat pipe can be increased a lot.
Small valve of Big valve of

super high super high
Vacuum Vacuum
Transparent glove
chamber with
argon protection Vacuum
equipment
Jack
Tank of
Heat pipe alkali metal
Fig. 10. Non-distillation technology of alkali metal heat pipe
Fig. 11. Three-path-equipment of alkali metal charging (figure 10)

After that, the alkali metal tank, the lower connection and the heat pipe is heated by the
outside heaters. The temperature can reach 150~160°C or so for sodium. If the alkali metal
inside is melted completely, then the big valve of high vacuum is closed, the pump
equipment is cut off. The small valve of high vacuum is opened and the argon will push the
alkali metal into the heat pipe. Then the small valve is closed and the big valve is opened.
The system is evacuated again to a high vacuum level. Finally, the heat pipe is sealed by a
special plier, soldered by a welder. A heat pipe is charged successfully.
Figure 11 is the photo of three-path-equipment.
4.4 The technology monitoring

The monitor equipment of technology is shown in figure 12. The power increase can be set
to heaters. The thermal couples and resistances are connected to the equipment. The computer
and the inserted instruments are two-level system. The computer, digital instruments,
controllable silicon, switches, contactors, buttons etc., are installed into the instrumental
cabinet. By the computer, the technology process can be realized.
Fig. 12. Monitor equipment of heat pipe technology

The instruments and the sensors are connected to collect data and control the process. By the
RS485 communication bus, the computer can display the process on time, the interface is
displayed by Chinese. The data can be storaged in the computer.
The performance of heat pipe depends on the process technology essentially.
5. Experimental results of alkali metal heat pipes

5.1 Startup from ambience
The startup experiments can test the heat pipe performance before the heat pipe is applied.
As shown in figure 11(a), the evaporator is heated by high frequency heater, the heat pipe is
set inside the high frequency loops, and the thermocouples are set along the condenser,
which is in the ambient air. By the high frequency heater, the heat flux can be very large,
and the some dryout point may be displayed by thermocouples.
(a) (b) (c)

Fig. 13. Test rig of alkali metal heat pipe and startup photos of sodium heat pipes. (a) Test
rig of heat pipe (b) Startup experiment, 5° anti-gravity (c) Startup, horizontal
The experimental photos are shown in figure 13 (b) and (c). By the color of the condenser,
the sodium heat pipe can be started up successfully. The color is uniform along the
condenser, the isothermal performance of the heat pipes are good. The purity of working
fluid and the high vacuum technology can guarantee that there is no noncondensible gas in
the fabricated heat pipes.
As in figure 14, temperatures of three condenser points are demonstrated for horizontal
position. For about 100S, the heat pipe can be started up. For 10 degree tilt angle, the
evaporator is set lower, the results of sodium heat pipe startup is shown in figure 15. For
another power step, the heat pipe performance is also satisfied.
700
600
500
T(℃ )
400 T1
300
T2
T3
200
100
0 50 100 150 200 250 300
t (s)
Fig. 14. Startup of horizontal sodium heat pipe
700
600 T1
500 T2
T3
400
(℃)
300
T
200
100
0
0 100 200 300 400 500
t ( s)
Fig. 15. Startup and power increased test of 10° degree tilt angle
5.2 Experiments in wind tunnel of electricity arc

By the high and super high temperature heat pipes, the local higher heat flux is moved to
the lower heat flux region, the heat is moved from the “peak point” to the “valley region” by
heat pipe, then the highest temperature is decreased a lot.
In a wind tunnel of electricity arc, three heat transfer elements as CC material, high
conduction CC material and heat pipes are tested and compared, as shown in figure 16.
From figure 16, after 1200s, the heat pipe is started up successfully. The operation lasts
nearly 5 minutes.
The upper three curves are the temperature histories of the stagnant points. The
temperature of heat pipe stagnant point is lower than the other CC and high conduction CC
elements, 120°Cand 50°C lower respectively. The heat pipe behaves good performance.
CC stagnant point
High conduction CC
stagnant point
heat pipe
stagnant point
Fig. 16. Experimental temperatures of different method by arc tunnel heating (By China
Academy 11)
6. Limits of alkali metal heat pipes

6.1 Continuum flow limit
With the decreasing of dynamic diameter, the heat pipe vapor flow may transit from the
continuum flow to the free molecule flow. The continuum limit can be judged by the
Knudsen number as,
λ (12)
Kn =
D
Here, λ is the free length of molecules, D is the minimum size of the vapor flow in heat pipe.
If Kn≤0.01, the flow is continuum; if Kn>0.01, the flow belongs to the free molecules. For the
latter situation, the heat pipe may lose its performance.
Cao and Fahgri derived the transition temperature as (Faghri, 1992),
P0 ⎡ h fg ⎛ 1 1 ⎞⎤
Ttr = exp ⎢ − ⎜ − ⎟⎥ (13)
ρtr Rg ⎣⎢ Rg ⎝ Ttr T0 ⎠ ⎦⎥
Substitute Kn=0.01 into equation (13), the transition temperature can be obtained.
As shown in figure 17, for sodium heat pipe, change of transition temperature with the
dynamic diameter is given. If the dynamic diameter is decreased less than 1mm, the
continuum limit occurs. The transition temperature will increase with decreasing the
diameter. When the dynamic diameter is 50μm, the transition temperature will be 830°C.
This temperature is in the range of normal temperature 500~1100°C. Such results mean that
the heat pipe will work at much higher temperature than the designed, in the mean time,
the heat transfer rate decreases.
900
800
Working fluid: Sodium
700
C
o
T tr ,
600
500
0.000 0.001 0.002 0.003 0.004 0.005

d, m
Fig. 17. For sodium heat pipes, change of transition temperature with the dynamic diameter
6.2 Other possible limits

When the alkali metal heat pipes are started up from low temperature, the vapor density is
very small. The viscous resistance may dominate (Ma et al, 1983). At the end of the
condenser, the vapor pressure decreases to extreme low, nearly zero. Then the viscous limit
is reached as,
Dv 2 L (14)
Qvis = ρ v , o Pv , o
64 μ vleff
If the alkali metal heat pipes operate at low vapor pressure, the vapor density is small and
the velocity is big, then the sonic velocity may choke the heat transfer, the sonic limit is
expressed as,
Qs = 0.474LAv ρ v , o Pv , o (15)
When the vapor flow can entrain the liquid, the inertial force is bigger enough, the
entrainment limit is given as,
QE = Av 2πσρ v L2 λ (16)
The capillary limit can be calculated by,
⎛ σρ L ⎞ ⎛ KAw ⎞
Qcap = 2 ⎜ l ⎟ ⎜ ⎟ (17)
⎝ μl ⎠ ⎜⎝ releff ⎟⎠
The basic heat transfer limits are given in figure 18 for sodium heat pipe. If the temperature
is lower than 500°C, the sonic limit should be paid attention. When the temperature is
between 500°C and 900°C, the entrainment limit is easy to occur, the temperature 700°C
corresponds to 756W, then the heat flux is 246W/cm2.
2500 200
2000
150
qvis
1500
q (W/cm )
2
qs
q(W/cm )
qvis
2
1000 100
qe qs
500 qcap qe
50
0
-500 0
200 400 600 800 1000 100 200 300 400 500 600 700
o o
T( C) T( C)
(a) (b)
Fig. 18. Four limits of sodium heat pipes with temperature, (b) is the detail of local (a)
Fig. 19. Viscous and sonic limits of lithium heat pipe

The viscous and sonic limits of lithium heat pipes are illustrated in figure 19. The bottom
line stands for the sonic limit. The details are also shown in the figure. From the results, the
lithium should work at higher temperature and higher heat flux.
7. Chemical vapor deposit technology

The new material and new technology for high and super high temperature heat pipes are
developed in recent years. There are some new technologies about alkali metal heat pipes.
Here the chemical vapor deposition (CVD) is introduced (Fortini, et al, 2010). The CVD
method can be used to fabricate the heat pipes, the number of wicking grooves, their location,
the cross-sectional shape, and the overall geometry of the heat pipe are easily varied. The
integral grooves also eliminate the need for screens, thus allowing for greater design flexibility.
Figure 20 shows the process schematically. The manufacturing process starts with a mandrel
whose outer contour matches the desired inner contour of the finished product. For a heat
pipe with a simple circular cross section, a tubular mandrel can be used to define the vapor
channel, and smaller rods can be attached to the mandrel to define the liquid return arteries.
After assembling the mandrel, the part is coated by CVD. The final step is etching away by
mandrel with acid.
Molybdenum
Molybdenum bar Mandrel
Rhenium removed
wires
Vapor
Space
Arteries
Liquid
lithium
A: Mandrel B: CVD coating C: Heat pipe
Fig. 20. Schematic of heat pipe fabrication process by chemical vapor deposition (Fortini, et
al., 2010)
The advantage of CVD is that the grooves are incorporated into the machining grooves into
the pipe wall after the pipe itself is fabricated. It also provides excellent flexibility in that it
enables any number of grooves to be incorporated into the heat pipe. Furthermore, the
grooves don’t have to be evenly spaced; they can be concentrated on one side or the other if
desired. Screens can be used in conjunction with the arteries at the evaporator end to further
delocalize the liquid. Also, because CVD is not a line-of-sight process, the cross section of
the heat pipe can be of virtually any geometry. It is not limited to straight heat pipes, nor is
it limited to pipes of circular symmetry. Finally, for heat pipes that are reasonably straight,
multiple pipes can be fabricated simultaneously.
Another benefit of CVD manufacturing is that multiple materials can be used to fabricate
the pipe. For example, rhenium was chosen because of its chemical compatibility with
lithium, ductility, and strength at high temperature. Rhenium, however, is expensive. To
reduce the amount of rhenium used, one could apply a thin film of rhenium to the mandrel
and then switch to say tantalum. Tantalum is extremely ductile and is more than an order of
magnitude less expensive than rhenium. Tantalum also has good strength at high
temperature, though not as good as rhenium. So even though a somewhat thicker layer of
tantalum would be needed over the initial thin film of rhenium, the material cost would be
reduced by more than a factor of 10.
8. Summings-up
In this chapter, the progress works of high and super high temperature heat pipes are
introduced. The micro cell erosion mechanism to the high temperature heat pipe case and
the compatibility are given, the selections of the case material and the working fluid should
be coupled to satisfy with the compatibility. The technology is the key problem to realize a
high performance heat pipe, the alkali metal distillation and non-distillation technology are
innovated, and the technology monitor is important for fabrication. Generally the working
fluid is solid before high and super high temperature heat pipe startups, the startup
possibility and time are analyzed and experimented. The heat transfer should be designed
much smaller than those of the operating limits, the possible heat transfer limits of high and
super high temperature heat pipes are calculated and discussed. The experimental and
theoretical results show that the fabricated heat pipes have good performances. The new
CVD methods can be used to fabricate the heat pipes, the integral grooves also eliminate the
need for screens, thus allowing for greater design flexibility.
9. References
Ma T. Z., Hou Z.Q. and Wu G.W. (1983). Heat Pipe (in Chinese), Science Press, pp.277-282,
ISBN 7-03-002011-1
Busse C.A. (1992). Heat Pipe Science, Advances in Heat Pipe Science and Technology, Proc.
of 8th Int. Heat Pipe Conference, pp.3-8, Bejing, Int. Academic Publishers
Zhang J.X. and Ning X.W. (2009). Study of Thermal Control Technology on Space Reactor Power
Supplier (in Chinese), Proc. of 9th China Space Thermophysics Conference, pp.1-5, Beijing
Boman B.L., et al. (1990). Heat Pipes for Wing Leading Edges of Hypersonic Vehicles, NASA
CP-181922
Faghri A. and Cao Y. (1992). Numerical Analysis of Leading Edge and Nosecap Heat Pipes,
Advances in Heat Pipe Science and Technology, Int. Academic Publishers, pp.303-
308, Proc. of 8th Int. Heat Pipe Conference, Bejing
David E. Glas.(1998). Closed Form Equations for the Preliminary Design of a Heat Pipe-
Cooled Leading Edge, NASA CR-1998-208962
Jiang G.Q., Ai B.C., Yu J. J., Chen L.Z. (2008). Application of high temperature heat pipe in
the protection technology of heat conduction (in Chinese), pp.74-80, Proc. of 11th
China Heat Pipe Conference, Weihai, China
Zhuang J. and Zhang H.(2000). Heat Pipe Technology and Engineering Application, pp.166-173,
Chemical Industry Press
Zhuang J. and Zhang G.Y.(1998). Sodium and water reaction in sodium heat pipe, pp.151-
155, Proc. of 6th China Heat Pipe Conference,Wuyishan, China
Jacobson D.L. and Wang J.H.(1984). Failure Analysis of a Sodium, Inconel 617 Heat Pipe,
Preprints of Proceedings of 5th Heat Pipe Conference, Tsukuba, Japan, 121-125
Zhang S.H. and Shen H.J. (1987), Molecule Physics and Thermodynamics, Beijing Science and
Technology Press, pp.75-76, ISBN7-5304-0023-1/Z
Jacobson D.L. and Soundararajan P.(1984). Failure Analysis of a Sodium Heat Pipe with
Integral Lithium Fluoride Thermal Energy Storage, pp.115-120, Preprints of Proc. of
5th Heat Pipe Conference, Tsukuba, Japan
Li T.H. and Hua C.S. (1987). Heat Pipe Design and Application (in Chinese), pp.102-108,
Chemical Industry Press
GB 9222-88 (1988). Strength Calculation of Pressure Parts for Water-tube Boilers, pp.1-72,
National Standards Press of China
Gelishen, Gejizunuofu (Russian accent) et al. (1966). Properties of Lithium (Interpretation
from Russian) , pp.10-12, China Industry Press
Fortini A.J., Arrieta V.M. (2010). Rhenium Heat Pipes for Hypersonic Leading Edges,
Preprints of 15th Int. Heat Pipe Conference, pp.1-6, Clemson, USA
26
Design of the Heat Conduction Structure

Based on the Topology Optimization
Yongcun Zhang, Shutian Liu and Heting Qiao
Dalian University of Technology
China
1. Introduction
The progress toward smaller scales in electronics makes the cooling of integrate circuits
become an important issue. The conventional convective cooling method which is feasible
and often used to control the temperature of a system becomes impractical because the
channels of heat transfer take up too much space for high compacted integrate circuit.
Hence, it is necessary to build heat conduct structures with high conductivity materials so
that the heat can be collected, transferred and exchanged with external environment
automatically and rapidly[1-2]. A key problem is how to design the structures with a rational
distribution of high conductive materials, which not only benefits to the temperature control
but also can reduce material and manufacturing costs and bring possibilities for further
miniaturization.
Studies designing the optimal heat conduction structure have attracted much attention and
many achievements have been obtained [1-24], including mathematical models and the
corresponding solving methods. For example, Bejan and co-workers put forward a tree-like
network construction method based on the constructal theory [1-8], Guo and co-workers
proposed some practical design criteria and developed the corresponding optimization
methods for the heat conduction structure based on the least dissipation principle of heat
transport potential capacity [9-15]. The topology optimization method has also been applied
for heat conduction structural optimization [16-22]. In all these cases, the nature of
optimization design for heat conduction structures is to build a mathematical model that
maximizes or minimizes an objective function (e.g. the thermal performance index)
subjected to certain constrains. Thus, it is a key to define a suitable thermal performance
index in such an optimization model.
Statistical data show that the failure of real devices with a fraction of 55% is caused by the
high temperature and this fraction increases exponentially with increasing temperature [25,
26]. Thus, the highest temperature is a primary factor that induces the failure of practical
cooling structure and should be well controlled. In practice, it is natural to define the highest
temperature as an objective function of the optimization model. However, the location of the
highest temperature usually changes with the change of material distribution in the
topology optimization process and is a discontinuous function of design variables, which
may introduce numerical difficulties in optimization. Therefore, instead of a directing
optimization of the highest temperature, it is more convenient to define another proper
thermal performance index as the objective function in an optimization model to accomplish

indirectly the goal of minimizing the highest temperature.
In the optimization model of heat conduction structure, the objective function can be
selected as
1
f (X ) = ∫ ( −q ( X )∇T ( X ))dΩ (1)
Ω 2
where X is the design variable used to describe the distribution of material, q(X) is the
flux density and ∇T ( X ) is the temperature gradient. Using the finite element formulation,
Eq. (1) can be also written as
f ( X ) = T Τ K ( X )T (2)
where T is the global temperature vector and K(X) is the thermal conductivity matrix.
Generally, Eq. (1) is defined as the dissipation of heat transport potential capacity
(DHTPC)[11], and the least dissipation principle of heat transport potential capacity is
presented based on this definition; Eq. (2) is defined as the heat dissipation efficiency[17,18],
which is the objective function of the heat conduction topology optimization.
Using the DHTPC (or the heat dissipation efficiency) as a thermal performance index, some
good design results have been obtained. However, this index can only tell us the heat
dissipative capability rather than the highest temperature. How much difference between
DHTPC and the present design goal, that is, the control of the highest temperature? Is there
any better thermal performance index? Answers to these questions are the motivation of this
study.
Firstly, the difference between the DHTPC and the present design goal is evaluated by a
one-dimensional heat conduction problem for a planar plate exchanger. Then, the geometric
average temperature (GAT) is proposed as a new thermal performance index and the
corresponding heat conduction optimization model is developed, the validity of optimization
model is proved by two example. Finally, some useful conclusions are given.
2. Heat conduction optimization of the planner plate exchanger

In many practical cooling structures, a commonly used design criterion is that the highest
temperature must not exceed a specified value. However, the optimization objective in
many existing heat conduction optimization models is the DHTPC. To evaluate the quality
of these exiting models, we compare their results with those obtained from an optimization
model with the highest temperature as the objective function. For simplicity, the presented
example is a one-dimensional heat conduction problem for a planar plate, which can be
solved analytically.
2.1 Problem description

A rectangular planar plate exchanger, with length l, width W (W>>l) and thickness t, is
embedded in the heater. The heat generated by heater flows into the exchanger uniformly.
The heat flowing into the exchanger is q′′ per unit time and area. Only one side along the
width direction of exchanger contacts with a thermostat with a constant temperature T0 and
others are adiabatic. This problem can be described as a planar heat conduction model with
uniform heat source, as shown in Fig. 1. Furthermore, this model can be simplified into a
Design of the Heat Conduction Structure Based on the Topology Optimization 525
one-dimensional heat transfer problem because the thickness t and the internal heat source q
do not change along the width direction. The goal is to obtain the optimal heat conduction
performance by designing the thickness t along the length direction of exchanger.
Since thermal conductivity is proportional to thickness t, the thickness design can be
transformed into the conductivity field design. That is to say, the limitation of material,
l l
∫ t ( x)Wdx = const , can be written as the capability of conductivity, ∫ k ( x)dx = K
0 0 0 , where K0
is a constant. The governing equation of heat conduction in the exchanger can be described
as
dT dq
q( x) = −k ( x) , + q′′ = 0, 0 < x < l
dx dx (3)
T ( x = 0) = T0 , q( x = l ) = 0
where, k(x) is the thermal conductivity, q(x) is the heat flux density and T(x) is the
temperature. In addition, the heat flux is assumed to be positive along the x direction.
Solving equations (3), we can obtain
x x
q = −q′′(l − x), T ( x) = T0 + ∫ ∇T dx = T0 + q′′∫ (l − x) k ( x)dx . (4)
0 0
Then, the optimization design for the exchanger is to determine the optimal heat conduction
performance by designing the conductivity filed under a given integral of thermal
conductivity (or material volume) over the design domain. Let f ( k ) denotes a thermal
performance index. The heat conduction optimization problem can be formulated as
Find : k(x)
min : f (k ) (5)
l
s.t.: ∫0
k ( x)dx = K 0
Using the Lagrange multiplier method, the solution of the thermal conductivity field can be
determined by
l
δ k ( f (k )) + λ ∫ δ kdx = 0, δλ
0 ( ∫ kdx − K ) = 0
0
l
0 (6)
2.2 Minimization of the highest temperature

According to the heat conduction theory, the highest temperature is located on the
boundary of x = l and can be written as
l
Tmax ( k ) = T0 + q′′∫ (l − x) k ( x)dx (7)
0
Substituting Eq. (7) into Eq. (6), we have
l l
∫ ⎡⎣ −q′′(l − x)
0
k 2 ( x) + λ ⎤⎦ δ k ( x)dx = 0, ∫ kdx = K
0 0 (8)
The optimal thermal conductivity field kTmax ( x) can be obtained by solving Eq. (8), which is
3K 0
kTmax ( x) = (l − x)1/ 2 (9)
2l 3/ 2
and the corresponding temperature distribution is
4q ′′l 3 / 2 3 / 2
TTmax ( x) = T0 + (l − (l − x) 3 / 2 ) (10)
9K 0
Introducing a dimensionless parameter
~
x = x/l (11)
the conductivity field and the temperature distribution can be expressed in the dimensionless
space as
~ k ( x)
kTmax ( ~
x) = = 3(1 − ~
x )1 / 2 / 2 (12)
K0 / l
and
~ T ( x) − T0
TTmax ( ~
x) = = 4(1 − (1 − ~
x )3/ 2 ) / 9 (13)
q ′′l 3 / K 0
where subscript Tmax denotes that the optimization objective is to minimize the highest
temperature.
Uniform Heat Source

l
x
t( x)
o y
T0
Fig. 1. A theoretical model of a planar plate exchanger
2.3 Minimization of the dissipation of heat transport potential capacity

For the planar plate exchanger, the DHTPC can be expressed as
(q′′)2 (l − x) 2
l
f (k ) =
2 ∫0 k dx (14)
When the DHTPC is considered as an optimization objective function, the optimal thermal
conductivity field should obey the following necessary conditions
l ⎡ (q′′) 2 (l − x) 2 ⎤ l
∫ ⎢⎣−
0 2k 2 ( x)
+ λ ⎥ δ k ( x)dx = 0,
⎦
∫ kdx = K
0 0 (15)
The thermal conductivity field can be obtained by solving Eq. (15), which is
2K0
kdis ( x) = (l − x) (16)
l2
and the corresponding temperature distribution is
q′′l 2
Tdis ( x) = T0 + x (17)
2K0
The dimensionless thermal conductivity field and temperature distribution are
k ( x)
kdis ( x ) = = 2(1 − x ) (18)
K 0 /(l )
and
T ( x) − T0
Tdis ( x ) = = x / 2 (19)
q′′l 3 / K 0
where subscript dis denotes that the optimization objective is to minimize the DHTPC.
2.5 0.6
~ ~
kdis TTm ax
~ 0.5
2 Tav
~
kTmax
0.4
1.5
~ ~
k T
~ 0.3
Tdis
1
0.2
0.5
0.1
0 0
0 0.2 0.4 0.6 0.8 1
x~
Fig. 2. Comparisons of thermal conductivity fields and temperature distributions from
different optimization models. av: the uniform conductivity field; dis: the dissipation of heat
transport potential capacity; Tmax: the highest temperature
2.4 Comparisons of two different optimization models

The dimensionless thermal conductivity fields and the corresponding dimensionless
temperature distributions from the two different optimization models are shown in Fig. 2.
To facilitate the comparisons, the temperature distribution with uniformly distributed
thermal conductivity (denoted by ‘av’) is analyzed, which can be expressed as
x q′′l (lx − x 2 / 2)
Tav = T0 + ∫ ∇T dx = T0 + (20)
0 K0
and the corresponding dimensionless temperature distribution is
Tav = x − x 2 / 2 (21)
which is also plotted in Fig. 2. It can be found that the temperature distribution from the
model with an objective function of the DHTPC has an obvious reduction in the internal
exchanger when compared with the temperature field from the model with a uniform
thermal conductivity field. However, these two models give the same highest temperature.
In addition, when compared with the model with an objective function of the highest
temperature, large differences in thermal conductivity field can be found and the highest
temperature increases by 12.5%, which indicates that the optimization model with an
objective function of the DHTPC sometimes cannot fulfill the present design goal. Thus, it is
necessary to propose new thermal performance indexes for the optimization model.
3. Optimization model based on the geometric average temperature

3.1 Objective function and optimization model
As mentioned above, the optimal design by the optimization model with DHTPC as an
objective function sometimes introduces large errors compared with the present design goal.
Furthermore, since the highest temperature is a discontinuous function of design variables,
direct optimization of it will bring numerical difficulties. To achieve a good tradeoff between
the optimization performance and numerical cost, a new thermal performance index called
the geometric average temperature Tgeoav is proposed, which can be expressed as
1/ n
⎛ 1 ⎞
Tgeoav = ⎜⎜ ∫Ω (T ( x )) dx ) ⎟⎟ , x∈Ω
n
(22)
⎝Ω ⎠
Where Ω denotes the area (or volume) over the design region. Theoretically, the geometric
average temperature is close to the highest temperature when n is infinitely large,
n →∞
i.e. Tgeoav → Tmax . Thus, the geometric average temperature is an appropriate approximation
of the highest temperature. The new heat conduction optimization model can be written as
Find : X = k ( x), x ∈ Ω
1/ n
⎛ 1 ⎞
min : Tgeoav ( X ) = ⎜⎜ ∫Ω (T ( x )) dx ) ⎟⎟
n
(23)
⎝Ω ⎠
s.t.: ∫Ω
kdΩ = K 0 , K 0 = const
Here, the finite element method is used to solve the optimization problem. Suppose that the
material is uniformly distributed and has the same conductivity in each element. Then the
distribution of material can be described by the different thermal conductivity in each
element mesh, which can be expressed by the finite element method
k ( x ) = ke , x ∈ Ω e , e = 1, 2," , Ne (24)
where ke (e = 1, 2, " , Ne) denotes the thermal conductivity of the e-th element, Ω e ∈ Ω the
region of the e-th element and Ne the total number of elements. Then, the temperature and
its n power in an element can be written as
T ( x ) = [ N ( x )]{T }, T n ( x ) = [ N ( x )]{T n } (25)
{T } = (T , T , " , T
1 2 Np )Τ , {T } = [(T ) , (T ) , " , (T
n
1
n
2
n
Np )n )Τ (26)
where {T } and {T n } denote the temperature vector of nodes and the corresponding n
power, respectively. Tn (n = 1, 2, " N p ) is the temperature of the n-th node, Np is the total
number of nodes and [ N ( x)] is the shape function matrix. The node temperature can be
solved by the following governing equation
[ K ]{T } = {Q} (27)
where {Q} is the thermal flux vector and [ K ] is the thermal conductivity matrix which can
be assembled by the element thermal conductivity matrix
Ne
[ K ] = ∑ [ K e ], [ K e ] = ke [ K e0 ] (28)
e =1
0
where [ K e ] is the e-th element thermal conductivity matrix with a unit thermal conductivity.
The geometric average temperature can be rewritten as
(
Tgeoav = [ B]{T n } ) 1
1/ n
, [ B] =
Ω ∫
Ω
[ N ( x)]dx (29)
Therefore, the heat conduction optimization problem can be expressed as
Find : X = (k1 ,k2 ," k Ne )Τ
min : Tgeoav ( X ) = [ B]{T n } ( )

1/ n
(30)
Ne
s.t.: ∑k Ve =1
e e = K 0 , Ve = Ωe
The feasible direction method was employed to find the optimal solution. The sensitivity of
the objective function (the geometric average temperature) can be expressed as
∂Tgeoav ∂
1
(
[ B ]{T n } ) {T n }
− ( n −1) / n
= [ B] (31)
∂ke n ∂ke
where
∂ ⎧ ∂T ⎫
∂ke
{T n } = diag (nT1(n−1) , nT2(n−1) , ", nTN(pn−1) )⎨ ⎬ (32)
⎩ ∂ke ⎭
and
⎧ ∂T ⎫ ∂
⎨
−1
⎬ = −[ K ] ( [ K ]) {T } = −[ K ]−1[ K e0 ]{T } (33)
⎩ ∂ke ⎭ ∂ke
3.2 Example 1
The planar plate exchanger is analyzed again by the new optimization model (23), in which
the one-dimensional heat conduction element with two nodes was used to mesh the design
domain and the feasible direction method was employed to find the optimal solution. The
obtained thermal conductivity field (material distribution) and the corresponding
temperature distribution are shown in Fig. 3. To facilitate comparisons, the solutions of the
optimization models with the DHPC and the highest temperature as objective functions are
also shown in Fig. 3. It can be seen that the thermal conductivity field and the temperature
distribution are close to the present design goal (the solution of minimizing the highest
temperature) when the power index n is larger than 16. Thus, the geometric average
temperature is an ideal thermal performance index. The new optimization model with the
geometric average temperature as the objective function is a more accurate description for
the design goal than that with the DHTPC as the objective function. The results obtained
from the convectional optimization model with the DHTPC as the objective function is equal
to that from the new optimization model when the power index n is 1. With the increasing
2.5 0.6
~
n=1 TTmax
0.5
2.0 n=2
n= 4
n= 8 0.4
~ 1.5 n = 16
~
k ~ 0.3 T
~ kTmax
Tdis
1.0 n = 16
0.2
0.5 n=8
n=4 ~ 0.1
kdis
n=2
n=1
0.0 0
0 0.2 0.4
~
x
0.6 0.8 1
Fig. 3. The solutions of the optimization model based on the geometric average temperature
with different power indexes. Results from other models are also shown for comparison
of power index n, the thermal conductivity field and temperature field obtained by the new
model is rapidly close to the ideal design. The change of the corresponding highest
temperature with the increasing power index n is shown in Fig. 4. Since the approximate
level tends to stabilize with the increasing power index n, an appropriate value is required
to select for the power index n in a practical optimization process.
0.50
~
Tmax,dis
0.48
~
T
0.46
~
Tmax,Tmax
0.44
0 2 4 6 8 10 12 14 16 18 20
n
Fig. 4. The highest temperature obtained by the optimization model based on the geometric
average temperature versus the power index n. Tmax,dis and Tmax,Tmax denote the corresponding
highest temperatures of the optimal solution of optimization models with the DHTPC and
the highest temperature as the objective function, respectively
3.3 Example 2
To demonstrate the difference between these two objective functions used in the topology
optimization method, a thermal structure with five heat sources is presented as an example
in this section.
A square planar plate, with dimension 50mm×50mm, is meshed by 50×50 discrete rectangular
elements. Temperature is 0 centigrade around the boundary, and five heat sources are set
symmetrically in the centre and around the plate with heat flux 1kW/m2. Material with
thermal conductivity k p = 200W/ (m·K) is used to filled this structure, and the volume
fraction Vf of this high heat-conductivity material is given to be 0.35, as shown in fig. 5.
A model using minimum heat dissipation as its objective function is adopted to solve this
problem, and its structural topology result is shown in fig. 6.(a), corresponding the contours
of temperature distribution and temperature gradient are shown in fig. 6.(b) and fig. 6.(c)
respectively. Likewise, the topology result of a model using geometric average temperature
as its objective function is shown in fig. 7.
T=0
T=0 T=0
T=0
Fig. 5. Initial design domain with five heat sources
(a)
(b) (c)
Fig. 6. (a) The optimal design generated from the optimization model 1 (b) the temperature
distribution isoline map (c) the temperature gradient isoline map
(a)
(b) (c)
Fig. 7. (a) The optimal design generated from the optimization model 3 (b) the temperature
distribution isoline map (c) the temperature gradient isoline map
Optimization Heat Maximum Maximum

model dissipation temperature temperature
(°C) gradient
Model 1 5.43e5 37.81 4.24
Model 2 6.23e5 33.72 6.71
Table 1. Results of the different objective

As shown in Fig. 6-7, these two different topology optimization models obtain entirely
different topology results. Using minimum heat dissipation as objective function, the result
shows that the high-heat conductivity material connects the central heat source with
surrounding heat sources directly and then extend to the outer thermal edge, as shown in
fig. 4.(a). This kind of heat transfer path will cause the temperature of the central heat source
much higher than the temperature of the surrounding heat sources. On the contrary, using
geometric average temperature as objective function, the result shows that, instead of
connecting the central heat source with surrounding heat sources directly, the high-heat
conductivity material bypass the surrounding heat sources and then connect to the outer
thermal edge, as shown in fig. 6.(a). This kind of heat transfer path will cause the
temperature of the central heat source and the surrounding heat sources equal to the
maximum temperature simultaneously. According to the topology results of these two
different objectives as shown in table 1, heat dissipation and maximum temperature
gradient in model 1 is smaller than model 2, and the maximum temperature in model 2 is
smaller than model 1.
According to the analysis of this example, we can conclude that: the model using heat
dissipation as its objective function can be used in problem which considers the
homogenization of the temperature gradient; the mode using geometric average temperature
as its objective function can be used in problem which considers oversize maximum
temperature as its failure mode.
4. Conclusion
We have discussed how to minimize the highest temperature of a heat conduction structure
by designing the material distribution with a specified material volume (conductivity
ability). The large error sometimes occurs between the results by the usual optimization
model with an objective function of the DHTPC and the theoretical optimal design. A
geometric average temperature has been proposed, which is a better thermal performance
index as the objective function. The solution of the new model with the geometric average
temperature as the objective function is close to the theoretical optimal solution.
5. Acknowledgment
This work was supported by National Natural Science Foundation of China (10902019,
90816025), by the national key basic research program of china (Grant No.2011CB610304)
and by the fundamental research funds for the central universities (DUT11LK07).
6. Referencess
Bejan A. Constructal theory network of conducting paths for cooling paths for cooling a
heat generating volume. International Journal of Heat Mass Transfer1997; 40(4): 799-
816.
Bejan A. Shape and Structure, from Engineering to Nature. Cambridge: Cambridge University
Press, 2000.
Ledezma GA, Bejan A, Errera MR. Constructal tree network for heat transfer. Journal of
Applied Physics1997; 82(1): 89-100.
Almogbel M, Bejan A. Conduction trees with spacing at tips. International Journal of Heat
Mass Transfer 1999; 42 (20): 3739-56.
Almogbel M, Bejan A. Constructal optimization of nonuniformly distributed tree-shaped
flow structures for conduction. International Journal of Heat Mass Transfer 2001; 44
(22): 4185-94.
Rocha LAO, Lorente S, Bejan A. Constructal design for cooling a disc-shaped area by
conduction. International Journal of Heat and Mass Transfer 2002; 45(8): 1643-52.
Silva AKD, Vasile C, Bejan A. Disc cooling with high-conductivity inserts that extend
inward from the perimeter. International Journal of Heat and Mass Transfer 2004;
47(19-20): 4257-63.
Rocha LAO, Lorente S, Bejan A. Conduction tree networks with loops for cooling a heat
generating volume. International Journal of Heat and Mass Transfer 2006; 49(15-16):
2626-35.
Cheng XG, Xia ZZ, Li ZX, et al. Optimization of heat conduction: Thermal dissipation and
optimal thermal conductivity distribution. Journal of Engineering Thermophysics 2002;
23(6): 715-17 [in Chinese].
Xia ZZ, Guo ZY. Heat conduction optimization simulated by life evolution. Progress in
Natural Science 2001; 11(8): 845-52 [in Chinese].
Guo ZY, Cheng XG, Xia ZZ. Least dissipation principle of heat transport potential capacity
and its application in heat conduction optimization. Chinese Science Bulletin 2003;
48(4): 406-10.
Cheng XG, Li ZX, Guo ZY. Variational principles in heat conduction. Journal of Engineering
Thermophysics 2004; 25(3): 457-9 [in Chinese].
Guo ZY, Liang XG, Zhu HY. Entranspy - A new physical property of the overall capacity of
heat transport. Progress in Natural Science 2006; 16(10): 1288-96 [in Chinese].
Cheng XG, Meng JA, Guo ZY. Potential capacity dissipation minimization and entropy
generation minimization in heat conduction optimization. Journal of Engineering
Han GZ, Guo ZY. Two different thermal optimization objective functions: Dissipation of
heat transport potential capacity and entropy production. Journal of Engineering
Li Q, Steven GP, Xie YM, et al. Evolutionary topology optimization for temperature
reduction of heat conducting fields. International Journal of heat mass transfer 2004; 47
(23): 5071-83.
Jia HP. Topology optimization of structure and compliant mechanism. Doctor Dissertation,
Dalian University of Technology. 2004; 95-118 [in Chinese].
Zuo KT, Chen LP, Zhang YQ, et al. Structural optimal design of heat conductive body with
topology optimization method. Chinese Journal of Mechanical Engineering 2005;
41(4):13-21 [in Chinese].
Zhang YC, Liu ST. Design of conducting paths based on topology optimization. Heat and
Mass Transfer, online; DOI 10.1007/s00231-007-0365
Li JC, Ye BY, Tang Y, et al. Algorithm of topology optimization criteria for heat conduction
structure based on density approach. Journal of South China University of Technology
(Natural Science Edition) 2006; 34(2):27-32[in Chinese].
Xu XH, Liang XG, Xun RJ. Optimization of heat conduction using combinatorial
optimization algorithms. International Journal of Heat Mass Transfer 2007; 50 (9-10):
1675-82.
Francois MP, Louis G. Optimal conduction pathways for a heat generating body: a
comparison exercise. International Journal of Heat Mass Transfer, 2007; 50(15-16):
2996-3006.
Wu WJ, Chen LG, Sun FR. Improvement of tree-like network constructal method for heat
conduction optimization, Science in China: Series E Technological Sciences 2006; 49(3):
1-10.
Zhou Shengbin, Chen lingen, Sun Fengrui. Optimization of constructual volume-point
conduction with variable cross section conducting path. Energy Conversion and
Management 2007; 48(1):106-11.
Liu MA, Dong QX, Chen S. Investigation on thermal design for opening cabinet of electronic
equipment by numerical simulation. Journal of air force engineering university (natural
science edition) 2005; 6(2):62-5 [in Chinese].
Qi YQ, He YL, Zhang W, et al. Thermal analysis and design of electronic equipments.
Modern Electronic Technology 2003; 144:73-76 [in Chinese].
27
Thermal Modelling for

Laser Treatment of Port Wine Stains
Li Dong1, Wang Guo-Xiang2 and He Ya-Ling1
1State
Key Laboratory of Multi-Phase Flow in Power Engineering
Xi’an Jiaotong University; Xi’an,
2Department of Mechanical Engineering; The University of Akron; Akron,
1China
2USA
1. Introduction
Port Wine Stains (PWS) are congenital vascular birthmarks (see Fig. 1) that occur in
approximately 0.3% of children (Alper & Holmes, 1983). Clinically, PWS ranges in
appearance from pale pink to red to purple. Most of the lesions are pale pink at birth and
can progressively be darken and thicken with age. PWS can be associated with significant
cosmetic disfigurement and psychologic distress. Histopathological analysis of PWS reveals
a normal epidermis overlying an abnormal plexus of benign vascular malformations consist
of ectatic capillaries of diameters varying from 10 to 300 μm. Laser treatment of PWS started
in the late 1960s with continuous-wave lasers such as carbon-dioxide and argon lasers
(Gemert et al., 1987; Dixon et al., 1984). During the laser treatment, light is absorbed by
hemoglobin (Hb) and oxyhemoglobin (HbO2) (the principle chromophores for light
absorption in human tissue) within the blood vessels and then converted into heat which
damages the endothelium and surrounding vessel walls (Alora & Anderson, 2000).
Unacceptable side effects such as scarring and permanent dyspigmentation were the two
major shortcomings of the early laser treatment of PWS. Those side effects result from
unselective heating of both PWS and nearby healthy dermal tissues during laser irradiation.
Fig. 1. PWS before and after laser treatment (Curtsey of Drs. Wang and Ying at Laser Cosmetic
Centre of 2nd hospital of Xi’an Jiaotong University, Xi’an, China)
The principle of selective photothermolysis of Anderson and Parrish (Anderson & Parrish,
1983) revolutionized the laser treatment of vascular lesions such as PWS. Based on this
principle, one can selectively destruct the PWS while keep the surrounding tissues
untouched by properly choosing right laser wavelength and pulse duration. In the case of
PWS, the primary absorption peaks of oxyhemoglobin and deoxyhemoglobin fall in the
visible range (418, 542, or 577 nm) (see Fig. 2). For those wavelengths the water (another
major chromophore for light absorption within tissue) has almost no absorption. Therefore,
lasers with these wavelengths can effectively damage the PWS while does not cause any
irreversible effect on surrounding tissues. As shown in Fig. 2, however, the melanin (another
chromophore for light absorption) within epidermis also has strong absorption over the
same wavelength range. Strong absorption of melanin and resulting epidermal heating not
only lead to undesired skin injury but also reduce the heating effect of PWS which is buried
deep underneath the epidermis. As a result, long wavelength lasers, such as the pulsed
tunable dye lasers (PDL) with wavelengths of 595 – 600 nm, are now widely employed in
laser treatment of PWS due to relatively weak absorption of melanin at the chosen
wavelengths. In addition, the modern PDL lasers are often equipped with adjustable laser
pulse width from 0.45 ms to 40 ms as well as surface cooling devices (Kelly et al., 2005).
Fig. 2. Absorption spectra of major chromophores in skin (Anderson, 1997)

Surface cooling is critical in success of modern laser treatment of PWS (Kelly et al. 2005). The
objective of surface cooling is to increase the laser fluence while prevent epidermal heating
due to competitive absorption by epidermal melanin. Various cooling devices have been
developed from early ice cubes (Gilchrest et al., 1982), chilled sapphire windows cooled by
circulating refrigerant (Kelly et al., 2005), to present chilled air cooling (Hammes et al., 2005)
and cryogen spray cooling (Nelson et al., 1995). The cryogen spray cooling technique
provides spatially selective cooling of epidermis with pulse spray of volatile refrigerant
R134a (tetrafluoroethane [C2H2F4]; boiling point at −26.2°C). Due to its short spurt, the PWS
buried deep within dermis will be not affected. Consequently, PDL laser with cryogen spray
cooling becomes standard protocol of PWS treatment (see Fig. 3).
Thermal Modelling for Laser Treatment of Port Wine Stains 539
Fig. 3. Schematic of the laser treatment of PWS together with cryogen spray cooling (CSC).
(Lanigan, 2000)
The laser surgery process of PWS (Laser PWS) has been studied extensively through
mathematical modeling. Various models have been developed to better understand the
laser-tissue interaction, in particular the thermal response of biological tissues and
physiological mechanisms of thermal damage of PWS (see e.g., Pfefer et al., 2000; Pickering
et al., 1989; Shafirstein et al., 2004; Tunnell et al., 2003). The models have also been used for
optimization of the surgery parameters and for development of new treatment protocols (Jia
et al., 2006, 2007). To model laser PWS, various simple skin models have been proposed to
simplify the complex anatomic structure of the skin with PWS (Gemert et al., 1995).
Mathematic techniques have been developed to calculate the light propagation and energy
deposition within the PWS (Jacques & Wang, 1995). Energy absorbed by PWS can then be
incorporated into the temperature equation as a heat source to determine the corresponding
temperature change in the irradiated tissues. A thermal injury sub-model is also used to
quantify the thermal damage to the PWS (Pearce & Thomsen, 1995).
The propagation of light in turbid media such as skin can be described by the radiative
transport equation (Ishimaru, 1989) which cannot be solved analytically for tissue
geometries. Techniques such as diffusion approximation (Gemert et al., 1997) and Beer’s law
(Verkruysse et al., 1993) have been used in the literature but they are limited to highly
scattering or coherent radiance materials (Niemz, 1996). For skin tissue, the Monte-Carlo
(MC) method, which offers a flexible approach to track the photon transportation in a
biological tissue, is widely accepted now as an accurate method and has been used in
modeling of laser surgery (Keijzer et al., 1991; Lucassen et al., 1996; Pfefer et al., 1996; Smith
& Butler 1995; Wang et al., 1995 & Wilson & Adam, 1983).
The bio-heat transfer equation such as Penn equation based on the skin models, in consistent
with that used for light simulation, can be solved for tissue temperatures in laser PWS. Due
to limited irradiation time of pulsed laser in laser PWS, the perfusion effect in Penn equation
is usually neglected so that the bio-heat equation becomes a heat conduction equation. The
thermal properties of PWS skin are scarce and those for normal skin tissues and blood are
usually employed as an approximation. For laser PWS, cryogen spray cooling is widely
employed by spraying volatile refrigerant such as R134a on skin surface before laser
irradiation. Therefore, the cooling of skins before laser irradiation should be part of the
model of laser PWS (Jia et al., 2006; Li et al., 2007a,b)
In this chapter, we will first present a brief review of the development of various skin
models proposed in the literature for thermal modeling of laser PWS. Next, we explain the
basic principle of the multi-layer Monte-Carlo method which is widely used for simulation
of light propagation in the skin with PWS. Then we introduce a general relation developed
recently by the present authors to quantify the heat transfer of cryogen spray cooling in laser
PWS. Finally, results from a multi-layer, homogeneous model of laser surgery of PWS are
presented to illustrate typical thermal characteristics of laser surgery of PWS. The effect of
laser wavelength and laser pulse duration on the heating of the PWS layer is also examined
briefly. Future needs for modeling the laser surgery of PWS is then discussed in conclusion.
2. Skin models and numerical simulation of laser surgery of PWS

The human skin has three main layers, the epidermis, the dermis, and the subcutaneous
layer (fat) (Brannon, 2007). The epidermis is the outer layer of skin with its thickness varying
in different types of skin. It is made up of cells called keratinocytes, which are stacked on
top of each other, forming different sub-layers. In the stratum basale sub-layer, there are
cells called melanocytes that produce melanin which is one of three major chromophores for
light absorption. Melanin is a pigment that is absorbed into the dividing skin cells to help
protect them against damage from sunlight (ultraviolet light). The amount of melanin in
skin is determined by genes and by how much exposure to sunlight. Below the epidermis is
the layer called dermis, which is a thick layer of fibrous and elastic tissue (made mostly of
collagen, elastin, and fibrillin) that gives the skin its flexibility and strength. The dermis
contains nerve endings, sweat glands and oil glands, hair follicles, and blood vessels. The
blood vessels of the dermis provide nutrients to the skin and help regulate body
temperature. Port Wine Stains (PWS) are the result of malformation of significant amount
blood vessels within dermis. The subcutaneous tissue is a layer of fat and connective tissue
that houses larger blood vessels and nerves. This layer is important in the regulation of
temperature of the skin itself and the body. The size of this layer varies throughout the body
and from person to person.
Ideally, a real three-dimensional multi-component skin models with detailed anatomic
structure of PWS vessels should be used to evaluate the light distribution and temperature
variation during laser PWS (Pfefer et al., 1996). However, low resolution of noninvasive
imaging technique and extremely complex human tissue structure makes this impossible at
moment. Thus, simplified skin models have been developed to analyze the thermal response
of complex human skin with PWS under laser irradiation (Gemert et al., 1995). A simplified
skin model should represent important histological characters of target chromophores and
physical events during and after laser surgery and should be easily handled mathematically.
The simplest model for laser PWS is probably the two-layer homogeneous model proposed
first by Gemert and Hulsbergen (Gemert & Hulsbergen, 1981) who simplified the human
skin containing PWS blood vessels to two-layer structure that parallel to the skin surface. In
their model, the upper epidermis layer includes uniformly distributed melanin and the
lower dermis layer is mixed with extra blood representing PWS. Gemert and Hulsbergen
were the first who attempted to calculate the light distribution within PWS theoretically by
employing the Kubelka-Munk method which significantly simplifies the light transport
equation. Refined skin models as shown in Figure 4 with multi-layers (Gemert et al., 1982,
1995; Miller & Veith, 1993) have then subsequently been developed, all including an extra
blood layer buried between dermis to represent the PWS. The Kubelka-Munk method and
later the Monte-Carlo method were used to predict the light deposition within the PWS.
The multi-layer model with homogeneously distributed chromophore (blood) in the PWS
layer has been widely used in numerical simulation of laser treatment of skin lesions. For
example, Pickering and Gemert (Pickering & Gemert, 1991) used this model to theoretically
investigate the mechanism of superior treatment of PWS by the laser of 585nm over that of
577nm. They found that the 585nm laser produced a deeper depth of vascular injury. The
effects of various laser parameters such as pulsed duration, the repetition rate of pulses, the
beam spot size, and the radiant energy fluencies on the outcomes of the laser surgery in
PWS were systematically investigated (Gemert et al., 1995, 1997; Kienle & Hibst, 1995, 1997;
Verkruysse et al., 1993). In all these studies, the PWS layer is treated as homogeneous
mixture of dermal tissue and blood of a given volumetric fraction. The average optical and
thermal properties of the chromophores and non-chromophore tissues weighted by their
volumetric fractions are employed in the analysis. It is found that the calculated temperature
in the PWS layer seems to be lower than expected (Aguilar et al., 2002), probably due to the
use of the averaging properties in the calculations.
Epidermis
Dermis
Chromophores PWS layer
Fig. 4. Schematic of a multi-layer skin model including a PWS layer within dermis. The PWS
layer is made of normal dermal tissue mixed homogeneously with blood
The multi-layer skin model simply treats the PWS as a homogeneous mixture of dermal
tissue and blood, ignore the detailed structural characteristics of blood vessels in PWS. To
represent more realistic anatomic features of PWS, skin models that contain individual
blood vessels have been developed (Gemert et al., 1986; Lahaye & Gemert, 1985; Lucassen et
al., 1995; Pickering et al., 1989), as shown in Figure 5. For example, Anderson and Parrish
(Anderson & Parrish, 1983) performed a thermal analysis directly on an isolated blood
vessel that was buried in dermal tissue and under laser irradiation. This analysis led to their
famous “selective photothermolysis” theory that provides theoretical foundation for laser
design and selection and for development of clinic protocol of laser treatment of PWS and
other cutaneous diseases (Lanigan, 2000). With the rapid growth of the computing power,
skin models with discretely distributed individual blood vessels became the favored among
researchers. Skin models with the blood vessels that are parallel to the skin surface and
regularly (aligned or staggered) buried in the dermis underneath the epidermal layer have
been widely employed. Recently, models with randomly distributed blood vessels with
varying sizes have also been utilized. Occasionally, the model with a single blood vessel was
also used to test the clinic outcome of new treating protocols of laser PWS (Jia et al., 2006,
2007). Concerns have been raised for possible lack of the scattering and shielding effects of
numerous blood vessels in real PWS for those discrete blood vessel models (Tan et al., 1990;
Verkruysse et al., 1993).
The skin model with discrete blood vessels can provide direct visualization of heating of the
blood vessels in laser PWS (Baumler et al., 2005; Shafirstein et al., 2004, 2007; Tunnell et al.,
2003). For example, Smith and Butler (Smith & Butler, 1995) have shown that the blood
vessels are heated up first during laser irradiation, and the dermis surrounding the vessel
are then heated conductively by the high temperature vessels. The effect of local heating
within the vessel and the shadowing effects of front vessels over the bottom vessels have
also been demonstrated.
Epidermis
Dermis
Blood vessels
Fig. 5. Schematic of a skin model with discrete blood vessels buried in dermis
3. Multi-layer Monte-Carlo (MC) method for light propagation in skin tissue

All above skin models can be treated as a multi-layered structure, even for those with
discretely distributed blood vessels (see Figs. 4 & 5). The light propagation within such
multi-layered structure can be accurately simulated by the multi-layered Monte-Carlo
(MLMC) method developed by Wang et al., (Wang et al., 1995). Here a brief introduction of
the MLMC method is presented and the details can be found elsewhere (Jacques & Wang,
1995; Wang et al., 1995).
The MLMC method deals with the transport of a laser beam within a medium made of
layered tissues that are parallel to each other. In MLMC, a photon packet is first injected
perpendicular to the tissue surface with a unit weight (W = 1) with the position of the
photon packet on the surface (position O in Figure 6) is determined from the shape of the
beam profile (e.g., Gaussian). The propagation or movement of the photon packet in the
tissue is then tracked step by step. At each step, the step size between two collisions within
the tissue is determined from a logarithmic distribution function with a random number
generated by the computer. Since the photon packet may hit a boundary of the current layer,
before each photon movement, the distance between the current photon location and the
boundary of the current layer in the direction of the photon propagation needs to be
computed. If the step size is smaller than the calculated distance, the step will move within
the current layer (segment OA, see Figure 6). Otherwise, if the step size is greater than this
distance (segment AB), the photon packet will hit the boundary and the photon packet can
be either internally reflected or transmitted across the boundary. If the photon is internally
reflected, the photon packet continues propagation with an updated step size and changes
its direction as a mirror reflection (segment BC). If the photon is transmitted, it continues its
propagation with an updated step size but with the direction newly calculated according to
the Snell’s law (segment BD). At the end of each step, the photon packet will interact with
the tissue. A fraction of the photon weight (ΔW= μa/μt) is first absorbed at the interaction
site, where μa and μt are the absorption coefficient and the attenuation coefficient,
respectively, of the medium. The absorbed weight is scored into the absorption array A[i,j]
at the local volume element (i,j) in a two-dimensional grid system, where i and j, are the
indices for volume elements. After absorption, the photon packet updates its weight to W-
ΔW and is then scattered at the interaction site (segment DE) by choosing a new direction of
propagation according to a given phase function and another random number generated by
the computer.
O O1
C A
B
D
E
Fig. 6. A schematic showing movement of photon packets in multi-layer Monte Carlo

simulation
The weight of a launched photon packet is continuously decreased due to absorption by the
medium as the photon packet moves within the tissue step by step. The photon packet is
terminated naturally if it moves out of the tissue domain due to either reflection or
transmission. For a photon packet still propagating inside the medium, the Russian roulette
technique is used to terminate the photon packet when the photon weight W is reduced to
be lower than a given threshold value. Once the photon packet is terminated, a new photon
packet enters the skin domain, probably at a new position (for example O1 in Figure 6), and
the above process is then repeated.
After all photon packets terminate, the MLMC simulation completes and the final
absorption array A[i,j] is obtained for the entire domain. Then, the energy density deposited
in each grid element can be calculated as follows:
A[i, j]
Q[i, j] = E ⋅ π r 2 ⋅ (1)
N ⋅ dv
where the array Q[i, j] is the rate of energy storage in the volume element (i,j), E is the
incident energy of the laser, r is the spot radius of the incident laser, N is the total number of
photo packets used in the simulation, and dv is the volume of the element. Although the
light transport in the layered-skin structure is gridless, the selection of the size of the volume
element in the grid affects the simulation through Equation (1). The size of the volume
element should be selected appropriately for proper scoring of the absorbed energy within
the element by satisfying the statistical requirement of the MC simulation. There should
have enough photons that have arrived at any given element in order to rightly reflect the
photon absorption in the medium. If the size of the volume element becomes too small and
the total number of photons used in the simulation is limited, one may find the case that no
photons at all arrive at certain elements and no energy deposition is registered in those
elements. Recently, the present authors have developed criteria that can be used to choose
right element size and the total number of photon packets needed in the MC simulation (Li
et al., 2011).
4. Cryogen spray cooling (CSC)

During the laser treatment of PWS, a significant amount of laser energy is absorbed by
melanin within the basal layer of the epidermis (Tunnell et al., 2000). To allow higher laser
energy doses used in clinic, one needs to cool the epidermis prior to laser irradiation.
Cryogen spray cooling (CSC) is one of the techniques developed to protect the epidermis
from non-specific heating by pre-cooling the skin prior to laser irradiation (Nelson et al.,
1995). The cryogen utilized is 1,1,1,2 tetrafluoroethane, also known as R134a, with boiling
temperature of Tb = - 26.2 °C at atmospheric pressure. Cryogen is usually kept in a container
at saturation pressure, which is approximately 660 kPa at 25 °C (95.7 psi), and delivered
through a standard high pressure hose to an electronically controlled fuel injector, to which
a straight-tube nozzle is attached. As the cryogen reaches the tissue surface, tissue
experiences a quick cooling process to reach low temperatures (Kelly et al., 2005).
To model laser PWS with CSC, a convective thermal boundary condition is usually used
(Aguilar et al., 2002; Jia et al., 2007; Li et al., 2007a,b; Majaron et al., 2001; Pfefer et al., 2000).
Due to lack of experimental data, the early models of laser PWS have employed a constant
heat transfer coefficient to quantify the short-pulsed cooling process (Aguilar et al., 2002;
Majaron et al., 2001; Pfefer et al., 2000). Recently, extensive experimental and numerical
investigations have been conducted to characterize the cryogen spray (Pikkula et al., 2001;
Zhou et al., 2008a,b) and to quantify the convection heat transfer during spray (Aguilar et
al., 2003a,b; Franco et al., 2004, 2005; Jia et al., 2004, 2007). Based on these experimental data,
the present authors have developed a quantitative relation that can be used to estimate the
convection heat transfer coefficient, h(r,t), as a function of both the space (r on the skin
surface) and time (t) during CSC in laser PWS (Li et al., 2007a,b). The non-dimensional form
of the relation is given as follows:
⎧ ⎪⎧ h (τ )
*
0 ≤ r * ≤ 0.4
⎪h * (r * ,τ ) = ⎨ o τ ≤ 1.0
* *
⎪ ⎪⎩[5(1 − r )] /[3ho (τ )] 0.4 < r * ≤ 1.0
⎪
⎪ ⎧ho* (τ ) 0 ≤ r * ≤ 0.2(τ + 1)
⎪ * * ⎪ (2)
⎨h (r ,τ ) = ⎨ * (5r * − τ − 1) 1.0 < τ < 4
⎪ ⎪ho (τ )] + × [0.09(τ − 1) − ho* (τ )] 0.2(τ + 1) < r * ≤ 1.0
⎪ ⎩ (4 − τ )
⎪ * * *
0 ≤ r * ≤ 1.0
⎪h (r ,τ ) = ho (τ ) τ ≥ 4.0
⎪
⎩
with h*o(τ) as the non-dimensional heat transfer coefficient at the center of spray:
⎧τ τ ≤ 1.0
⎪
h (t ) ⎪1.0 − 0.35(τ − 1) 1.0 < τ ≤ 3.0
ho* (τ ) = o =⎨ (3)
ho ,max ⎪0.3 − 0.02(τ − 3) 3.0 < τ ≤ 8.0
⎪⎩0.2 − 0.0125(τ − 8) 8.0 < τ
where ho(t) is the dimensional heat transfer coefficient at the center of spray and ho,max is the
maximum value of ho at time tmax, both estimated from experimental data (Li et al., 2007a).
In Eqs. (2) and (3), the non-dimensional heat transfer coefficient, h*, the non-dimensional
radius coordinate, r*, and the non-dimensional time,τ, are defined as follows:
h * (r * ,τ ) = h(r , t ) / ho ,max , r * = r / rspray , and τ = t / tmax
with rspray the radius of spray spot on the skin surface.

Equations (2) and (3) can be employed in any thermal models for laser PWS before laser
irradiation. Figure 7 shows typical dynamic variations of the temperature in the skin during
CSC calculated by our own model (to be discussed in detail below). In Figure 7, the
temperature distributions within the skin are plotted at two time instants during spray. The
corresponding spray distance is 30mm and the spray spurt is 100 ms. One point to be
noticed is that the skin surface temperature drops to about -20 oC at 30 ms after the spray
starts (see Figure 7a). At this time, however, the temperature within the skin including
majority of the epidermal layer still remains the same initial value. After 60 ms of CSC (Fig.
7b), the entire epidermal layer is cooled down below -10 oC, but the temperature of the PWS
layer still remains almost the same initial temperature and is not affected by the spray. Such
an effect of the cryogen spray is desirable for laser PWS since the only purpose of cooling is
to prevent the skin overheating. Since the skin tissue is a very poor thermal conductor, the
cooling effect at the surface is hardly felt by the PWS buried deep in the dermis if the spray
spurt is kept short.
(a) t = 30 ms (b) t = 60 ms
Fig. 7. Calculated temperature distributions within skin during cryogen spray cooling at
times after 30ms (a) and 60ms (b). (Spray distance: 30 mm)
5. A model for laser treatment of PWS with CSC

To illustrate the thermal characteristics of the laser surgery process of PWS, we present here
a multi-layer thermal model of laser PWS with CSC. The skin model chosen consists of four
layers: an epidermal layer containing melanin, two dermal layers with a PWS layer
sandwiched in between. The PWS layer is assumed to be a homogeneous mixture of dermis
and blood of a given volumetric fraction. The bulk optical and thermal properties of the
epidermal layer and the PWS layer are determined based on the corresponding volume
contents of melanin and hemoglobin (Verkruysse et al., 1993), respectively. The laser beam
has a Gauss profile and the MLMC method discussed in Section 3 is used to quantify the
energy deposition in various layers. The resulting rate of energy deposition in the tissue,
given in Eq. (1), is included as the source term, Q, in the following heat conduction equation:
∂T ki ∂ ⎛ ∂T ⎞ ∂ 2T
ρic p ,i = ⎜r ⎟ + ki 2 + Qi (4)
∂t r ∂r ⎝ ∂r ⎠ ∂z
where the subscript i (= e, d, p) represents, respectively, the epidermal layer, the dermal
layer, and the PWS layer; ρ is the density, cp the specific heat, k the thermal conductivity, T
the temperature, and t the time. A cylindrical coordinate system as shown in Figure 8 is
used with the origin located at the center of the laser beam on the skin surface.
R134a Laser
0 r
Epidermis
Dermis
PWS layer
Dermis
Fig. 8. Schematic of the multi-layer skin model and the corresponding two-dimensional
cylindrical coordinate system
On the skin surface (z=0), a convection heat transfer condition is used to describe the cooling
effect of CSC and the heating effect of the environment afterward:
∂T
ke = hc ( r , t )[T (r ,0, t ) − Tc ] during CSC
∂z z=0
(5)
∂T
ke = ha [T (r ,0, t ) − Ta ] after CSC or without CSC
∂z z=0
where Tc and Ta are the temperatures of liquid cryogen on the skin surface and the
environment, respectively. ha is the convective heat transfer coefficient between the cold
surface and air, and is treated as a constant. hc(r, t) is the heat transfer coefficient between
the surface and the liquid cryogen on the surface during short-pulsed cryogen spray and is
determined from Eqs. (2) and (3).
The above model is use to simulate the thermal process in the laser treatment of PWS with
CSC. The laser spot diameter is chosen as 5 mm while the CSC spray spot size is fixed at 10
mm. For the skin model, the thickness of the epidermal layer is 50 μm with melanin of five
volumetric percent (5%); the thickness of the PWS layer is 200 μm with the volumetric
fraction of hemoglobin of 30%. Here a high volumetric fraction of hemoglobin is used to
simulate the mature PWS lesion (dark red or purple). The initial skin temperature is
assumed to be at 37 °C and the surrounding temperature Ta is 25 °C. Two typical wavelengths
(585 nm and 595 nm) of the pulsed-dye lasers are examined with laser pulse duration
varying from 1.5 ms to 40 ms, corresponding to the working range of the clinically-used
PDL lasers (Kelly et al., 2005).
5.1 Light and temperature distribution without CSC

A typical result from the multi-layer Monte-Carlo simulation is given in Figure 9a which
shows the distribution of the photon weight A(r, z) within the skin after irradiation of a PDL
beam of 585 nm. As expected, a strong absorption occurs within both the epidermal layer (0
< z < 50 μm) and the PWS layer (250 < z < 450 μm) due to the existence of melanin and
hemoglobin within the two layers, respectively. Less absorption in the PWS layer than that
in the epidermal layer is due to the screening effect of the epidermal layer. It is such
absorption by the melanin in the epidermal layer that prohibits the application of a high
energy dose in laser PWS. Comparing to these two layers, the dermal layer shows very little
absorption of light because of a low absorption coefficient of dermis at the given wavelength
(585 nm). Figure 9a also shows a decrease of the photon absorption along the radial
direction, due to the nature of the incoming laser beam which has a Gaussian profile.
The photon absorption given in Figure 9a can be converted into the rate of heat generation
within the skin in the energy equation (4). Solving the energy equation gives the temperature
distribution within the skin, as shown in Figure 9b. Figure 9b plots the temperature
distribution within the skin at the end of 1.5 ms laser pulse. In this case, no cryogen spray
cooling is used. The incident laser fluence was 4 J/cm2. It is found that both the epidermal
and the PWS layers experience high temperatures due to strong photon absorption of the
two layers. At the present level of laser fluence, the temperatures of both the epidermal and
PWS layers exceed the critical coagulation temperature of about 70 oC (Pearce & Thomsen,
1995). Although a high PWS temperature is desired, a high epidermal temperature causes
unspecified skin injury and should be avoided. The cryogen spray cooling technique can be
used to protect the skin from such overheating as will be shown below.
(a) (b)
Fig. 9. Calculated distributions of the photon absorption (a) and temperature (b) within the
skin at the end of 1.5 ms laser pulse of 585 nm. (Laser fluence E = 4 J/cm2, without CSC)
5.2 Thermal characteristics of laser treatment of PWS with CSC

The thermal characteristics of laser PWS with CSC are illustrated in Figure 10 which
presents the temperature distributions within the skin at four time instants: 100ms (a),
101.5ms (b), 121.5ms (c) and 301.5ms (d) after the CSC starts. Laser was fired after 100 ms
spurt cryogen spray. The laser fluence is 6 J/cm2 and the laser pulse duration is 1.5ms. At
the end of the CSC (Fig. 10a), the skin surface temperature decreases to below -10 oC while
the temperature of the PWS layer remains almost the same as the initial value. At the end of
laser irradiation (Fig. 10b at 101.5 ms), the temperatures of both the epidermal and the PWS
layers rise as a result of energy absorption, while the temperature of the dermal layer keeps
almost unchanged due to a low absorption of the dermis. After laser irradiation, the
temperatures in both the epidermal and the PWS layers decrease as a result of heat
conduction. As we can see from Fig. 10c (20 ms after laser irradiation), the peak
temperatures of both the epidermal and the PWS layers lower down while the temperature
of the dermal layer rises. As heat conduction continues, the entire skin layers approach
almost uniform temperature, see Fig. 10d at 200 ms after laser irradiation.
(a) t = 100 ms (b) t = 101.5 ms
(c) t = 121.5 ms (d) t = 301.5 ms

Fig. 10. Calculated temperature distributions within the skin during laser surgery of PWS
with CSC at times after 100ms (a), 101.5ms (b), 121.5ms (c) and 301.5ms. (CSC spurt
duration: 100ms, Spray distance: 30 mm, laser fluence: 6 J/cm2, and laser pulse duration:
1.5ms)
5.3 Effect of laser wavelength

Appropriate selection of the wavelength of the laser beam is a critical issue in laser PWS due
to the sensitivity of the light absorption of PWS to the wavelength. Figure 11 shows the
temperature distributions of the skin at the end of 1.5 ms pulsed laser irradiation under two
wavelengths, 585 (a) and 595nm (b), respectively, with all other conditions remain the same.
A comparison of the temperature profiles along the central axial direction corresponding to
Figure 11 is also given in Fig. 12a. Figure 12b plots similar temperature distributions for a
thicker PWS layer of 300 μm. Inspecting these figures, one finds that the laser wavelength
affects the highest temperature possible in the PWS layer and the evenness of the
temperature distribution over the PWS layer. A much higher possible temperature in PWS is
achieved for the shorter 585 nm laser than that for the longer 595 nm laser. The longer 595
nm laser, however, produces a much more even heating over the PWS layer. This can be
more clearly demonstrated in the case of a thicker PWS layer as shown in Figure 12b.
(a) Wavelength: 585nm (b) Wavelength: 595nm
Fig. 11. Temperature distributions within the skin at the end of 1.5 ms laser irradiation.
(PWS layer thickness: 200 μm, laser fluence: 6 J/cm2, with CSC)
120 120
100 585nm 100

Wavelength
595nm
Temperature ( oC)
Temperature ( oC)
585 nm
595 nm
80
80
60
60
40
40
20
20
0 100 200 300 400 500 600 0 100 200 300 400 500 600
Depth ( μm) Depth ( μm)
(a) 200 μm thick PWS layer (b) 300 μm thick PWS layer
Fig. 12. Temperature distribution along the tissue depth direction at the spray center for two
wavelengths (585 and 595nm): PWS layer thickness (a) 200 μm and (b) 300 μm. (Laser
fluence: 6 J/cm2, pulse duration: 1.5 ms, with CSC)
5.4 Effect of laser pulse duration

The pulse duration of the laser beam is another important parameter that needs to be
carefully chosen in clinic practice. Figure 13 shows the calculated temperature distributions
within the skin at the end of laser irradiation for three pulse durations: 1.5 ms (a), 10 ms (b)
and 40ms (c), respectively. The laser fluence is 6 J/cm2. The comparison of the central
temperature profile within skin for three cases is given correspondingly in Figure 13d.
(a) Pulse duration: 1.5 ms (b) Pulse duration: 10 ms
120
Pulse duration
100 1.5ms
10ms
Temperature ( oC)
40ms
80
60
40
20
0 100 200 300 400 500 600

Depth ( μm)
(c) Pulse duration: 40 ms (d) Central temperature profile

Fig. 13. Calculated temperature distributions within the skin at the end of laser irradiation
for three pulse durations: (a) 1.5 ms, (b) 10 ms and (c) 40 ms; (d) Comparison of the central
temperature profiles for three pulse durations. (Laser fluence: 6 J/cm2 with CSC)
Inspecting these plots finds immediately that the peak temperature of the PWS layer at the
end of laser irradiation shows a continuous reduction, from 105 °C to 73 °C as the pulse
duration increases from 1.5 ms to 40 ms. In the case of a short pulse duration (e.g., 1.5 ms),
the PWS layer is heated up quickly at the end of laser irradiation with little heating of the
neighbor dermal tissue. A significant portion of the PWS layer is heated up over the critical
coagulation temperature of 70 °C. When the pulse duration increases to 40 ms, not only the
peak temperature of the PWS layer reduces to a lower value, the percentage of the PWS
layer that is above the critical coagulation temperature is also significantly reduced to a
smaller portion. Meanwhile, the neighbor dermal tissue is significantly heated up to close to
the coagulation temperature, which may lead to the damage of the healthy tissues. Such a
variation in the peak temperature of the PWS with pulse duration can be understood by
considering the combined effect of the laser heating and heat conduction of the heated PWS
to the neighbor colder dermis. As the laser heating prolongs over a long pulse, the heat
conduction from the heated PWS layer to the surrounding colder dermal tissues prevents a
further increase in the PWS temperature and thus reduces the peak PWS temperature. A
longer laser heating time is also associated with more energy through conduction into the
surrounding tissues, leading to continuous increase in the temperature of the dermal tissues.
In clinic practice of laser PWS, a short laser pulse is usually preferred, except for the cases
with extremely large blood vessels. For that case, however, a better model is needed to
provide more quantitative description of the laser surgery process of PWS.
6. Conclusion and future work

In this chapter, we present a brief review of thermal modelling of the treatment of port wine
stains with the pulsed dye laser. We show that laser treatment of port wine stains is
primarily a thermal issue involving both radiative energy transport within the tissue during
laser irradiation and tissue heat conduction during and after laser irradiation. Based on
simplified skin models that reduce the complex anatomic structure of skins to simple layer
structures, the process can be successfully simulated by solving the corresponding radiative
energy transport with the multi-layer Monte-Carlo method and the heat conduction
equation with traditional numerical methods. We have used a simple multi-layer
homogeneous model to illustrate the basic thermal characteristics of laser treatment of PWS.
We also demonstrated that the model can be used to make selections of the laser parameters
such as wavelength and pulse width in clinical practice. Quantitative information for critical
surface cooling technique, CSC, is also presented and included in our model.
Although great progresses have been achieved in both clinic practice and physical
understanding of laser PWS after four decades’ efforts, many issues remain. Clinically, the
present protocol of PDL-based lasers could significantly eliminate the PWS vessels, but only
less than 20% of complete clearance of the PWS has been achieved (Kelly et al., 2005).
Recurrence has been observed with a rate up to 50% after five years (Orten et al., 1996). All
these suggest a lack of fundamental understanding of the PWS destruction mechanisms in
the present laser PWS process. From the modeling point of view, neither the multi-layer
homogeneous model nor the discrete blood vessel model provides accurate representation
of the real and complex anatomic configuration of the PWS vessels. Attempts to construct
realistic PWS structure based on computer-reconstructed biopsy from PWS patients had
only limited success (Pfefer et al., 1996). New models are desired that should combine the
simplicity of the multi-layer homogeneous model while take into account the detailed effect
of complex PWS configurations. In addition, quantitative predictions of the temperature
change of the PWS in the laser treatment require accurate optical and thermal properties of
PWS, which are scarce at the moment.
The ultimate objective of any model for laser PWS is to accurately predict the thermal
damage after the laser irradiation. The existing PWS damage model is a pure thermal model
based on simple Arrhenius rate process integral (Pearce & Thomsen, 1995). The model does
not take into account the photochemical and photomechanical effect of laser on skin tissues
and blood vessels. Recent experimental evidence suggests that the vessel damage in laser
PWS is a multi-time scale phenomenon. The collateral damage of blood vessels in laser PWS
is due to accumulative result of early photothermal effect and later photochemical and
photomechanical effect. The recurrence of PWS involves a time scale that may last to more
than five years. Active researches are being conducted to understand these long term
phenomena in laser PWS.
7. Acknowledgments
We like to acknowledge valuable discussions with Drs. Y.X. Wang and Z.Y. Ying at Laser
Cosmetic Centre of 2nd Hospital of Xi’an Jiaotong University. Special thanks to Prof. Guo
Lie-jin, Prof. Chen Bin, Prof. Wang Yue-she, Dr. Zhou Zhi-fu and Dr. Wu Wen-juan for their
help to the project. G.-X. Wang thanks the support of “Changjiang Scholar” program of
Education Ministry of China. The work is supported in part by the special fund from the
State Key Laboratory of Multiphase Flow in Power Engineering at Xi’an Jiaotong University.
8. References
Aguilar, G.; Diaz, S.H.; Lavernia, E.J. & Nelson, J.S. (2002). Cryogen spray cooling efficiency:
improvement of port wine stain laser therapy through multiple-intermittent
cryogen spurts and laser pulses. Lasers in Surgery and Medicine, Vol.31, No.4, (July
2002), pp. 27-35, ISSN 0196-8092
Aguilar, G.; Wang, G.-X. & Nelson, J.S. (2003a). Effect of spurt duration on the heat transfer
dynamics during cryogen spray cooling. Phys Med Biol, Vol.48, No.14, (July 2003),
pp. 2169-2181, ISSN 0031-9155
Aguilar, G.; Wang, G.-X. & Nelson, J.S. (2003b). Dynamic Cooling Behavior during Cryogen
Spray Cooling: Effects of Spurt Duration and Spray Distance. Lasers in Surgery and
Medicine, Vol. 32, No.2, (February 2003), pp. 152-159, ISSN 0196-8092
Anderson, R.R. (1997). Laser-tissue interaction in dermatology, In: Lasers in cutaneous and
aesthetic surgery, K.A. Arndt (Ed.), pp. (26-51), Lippincott Williams and Wilkins,
ISBN 0316051772, Boston, Massachusetts, USA
Anderson, R.R. & Parrish, J.A. (1983). Selective photothermolysis: precise microsurgery by
selective absorption of pulsed radiation. Science, Vol.220, No.4596, (April 1983), pp.
524–527, ISSN 0036-8075 (Print), 1095-9203 (Online)
Alora, M. B. T. & Anderson, R. R. (2000). Recent developments in cutaneous lasers. Lasers in
Surgery and Medicine. Vol.26, No.2, (April 2000), pp. 108–118, ISSN 0196-8092
Alper, J.C. & Holmes, L.B. (1983). The incidence and significance of birthmarks in a cohort of
4641 newborns. Pediatr Dermatol., Vol. 1, No.1, (July 1983), pp. 58-68, ISSN 1525-
1470 (Online)
Baumler, W.; Vogl, A.; Landthaler, M.; Waner, M. & Shafirstein, G. (2005). Port wine stain
laser therapy and the computer-assisted modeling of vessel coagulation using the
finite elements method. Medical Laser Application, Vol.20, No.4, (December 2005),
pp. 247-254, ISSN 1615-1615
Brannon, H. (April 2007). Skin Anatomy, In: About.com, 10.04.2011, Available from:
http://dermatology.about.com/cs/skinanatomy/a/anatomy.htm
Dixon, J.A.; Huether, S.; & Rotering, R. (1984). Hypertrophic scarring in argon laser
treatment of port wine stains. Plast Reconstruct Surg, Vol.73, No.5, (May 1984), pp.
771-780, ISSN 0032-1052 (Print), 1529-4242 (Online)
Franco W.; Wang, G.-X.; Nelson, J.S. & Aguilar, G. (2004). Radial Heat Transfer Dynamics
during Cryogen Spray Cooling. Proceedings of IMECE 2004, IMECE2004-59609,
ASME, ISBN 0-7918-4711-X
Franco, W.; Liu, J.; Wang, G.-X.; Nelson, J.S. & Aguilar, G. (2005). Radial and temporal
variations in surface heat transfer during cryogen spray cooling. Phys Med Biol,
Gemert, M.J.C. van & Hulsbergen, J.P.H. (1981). A model approach to laser coagulation of
dermal vascular lesions. Arch Dermatol Res, Vol.270, No.4, (February 1981), pp. 429-
439, ISSN: 0340-3696 (Printed) 1432-069X (Online)
Gemert, M.J.C. van; de Kleijn, W.J.A. & Hulsbergen J.P.H. (1982). Temperature behaviour of
a model port wine stain during argon laser coagulation. Phys Med Biol, Vol.27, No.9,
(September 1982), pp. 1089–1104, ISSN 0031-9155
Gemert, M.J.C. van; Welch, A.J.; & Amin, A.P. (1986). Is there an optimal laser treatment for
port wine stains? Lasers in Surgery and Medicine, Vol.6, No.1, (January 1986), pp. 76-
83, ISSN 0196-8092
Gemert, M.J.C. van; Welch, A.J.; Tan, O.T.; & Parrish, J.A. (1987). Limitations of carbon
dioxide lasers for treatment of port-wine stains. Arch Dermatol, Vol.123, No.1,
(January 1987), pp. 71-73, ISSN (Printed): 0340-3696 (Online): 1432-069X
Gemert, M.J.C. van; Welch, A.J.; Pickering, J.W. & Tan, O.T. (1995). Laser Treatment of Port
Wine Stains, In: Optical-Thermal Response of laser-Irradiated Tissue, A.J. Welch &
M.J.C. van Gemert, (Ed.), 789-829, Plenum Press, ISBN 0-306-44926-9, New York
Gemert, M.J.C. van; Smithies, D.J.; Verkruysse, W.; Milner, T.E. and Nelson, J.S. (1997).
Wavelengths for port wine stain laser treatment: influence of vessel radius and skin
anatomy. Phys Med Biol, Vol.42, No.1, (January 1997), pp. 41–50, ISSN 0031-9155
Gilchrest, B.A.; Rosen, S. & Noe, J.M. (1982). Chilling port wine stains improves the response
to argon laser therapy. J Plast Reconstr Surg, Vol.69, No.2, (February 1982), pp. 278-
283, ISSN 0032-1052 (Print), 1529-4242 (Online)
Hammes, S. & Raulin, C. (2005). Evaluation of different temperatures in cold air cooling
with pulsed-dye laser treatment of facial telangiectasia. Lasers in Surgery and
Medicine, Vol.36, No.2, (February 2005), pp. 136–140, ISSN 0196-8092
Ishimaru, A. (1989). Diffusion of light in turbid media. Appl Opt, Vol.28, No.12, (June 1989),
pp. 2210-2215, ISSN 1559-128X (Print) 2155-3165 (Online)
Jacques, S.L. & Wang, L.H. (1995). Monte Carlo Modeling of Light Transport in Tissues, In:
Optical-Thermal Response of laser-Irradiated Tissue, A.J. Welch & M.J.C. van Gemert,
(Ed.), 73-100, Plenum Press, ISBN 0-306-44926-9, New York
Jia, W.C.; Aguilar, G.; Wang, G.-X. & Nelson, J.S. (2004). Heat transfer dynamics during
cryogen spray cooling of substrate at different initial temperatures. Phys Med Biol,
Vol. 49, No.23, (December 2004), pp. 5295-5308. ISSN 0031-9155
Jia, W.; Aguilar, G.; Verkruysse, W.; Franco, W. & Nelson, J.S. (2006). Improvement of port
wine stain laser therapy by skin preheating prior to cryogen spray cooling: a
numerical simulation. Lasers in Surgery and Medicine, Vol.38, No.2, (February 2006),
pp. 155-162. ISSN 0196-8092
Jia, W.C.; Choi, B.; Franco, W.; Lotfi, J.; Majaron, B.; Aguilar, G.; & Nelson, J.S. (2007).
Treatment of cutaneous vascular lesions using multiple-intermittent cryogen spurts
and two-wavelength laser pulses: numerical and animal studies. Lasers in Surgery
and Medicine, Vol.39, No.6, (July 2007), pp. 494-503, ISSN 0196-8092
Keijzer, M.; Pickering, J.W. & van Gemert, M.J.C. (1991). Laser beam diameter for port wine
stain treatment. Lasers in Surgery and Medicine, Vol.11, No.6, (October 1991), pp. 601-
605, ISSN 0196-8092
Kelly, K.M.; Choi B.; McFarlane, S; et al. (2005). Description and analysis of treatments for
port-wine stain birthmarks. Arch Facial Plast Surg, Vol.7, No.5, (October 2005), pp.
287—94, ISSN 1521-2491
Kienle, A. & Hibst, R. (1995). A new optimal wavelength for treatment of port wine stains?
Phys Med Biol, Vol.40, No.10, (October 1995), pp. 1559-76, ISSN 0031-9155
Kienle, A. & Hibst, R. (1997). Optimal parameters for laser treatment of leg telangiectasia.
Lasers in Surgery and Medicine, Vol.20, No.3, (December 1998), pp. 346–353, ISSN
0196-8092
Lahaye C.T.W. & van Gemert, M.J.C. (1985). Optimal laser parameters for port wine stain
therapy: a theoretical approach. Phys Med Biol, Vol.30, No.6, (June 1985), pp.573-87,
ISSN 0031-9155
Lanigan, S. W. (2000). Lasers in Dermatology, Springer, London, ISBN 1852332778
Li, D.; He, Y.L.; Liu, Y.W. & Wang, G.-X. (2007a), Numerical analysis of cryogen spray
cooling of skin in dermatologic laser surgery using realistic boundary conditions.
Proceedings of 22th Int. Congress Refrigeration conference, (July) 2007, Beijing, China
Li, D.; He, Y.L.; Wang, G.-X.; Xiao, J. & Liu, Y.W. (2007b), Numerical analysis of cold injury
of skin in cryogen spray cooling for dermatologic laser surgery. Proceedings of 2007
ASME International Mechanical Engineering Congress, IMECE2007-43876, Nov. 2007,
Seattle, Washington, ISBN 0-7918-4302-5
Li, D.; Wang, G.-X. & He, Y.L. (2011), Criteria for section of voxel size and photon number in
Monte Carlo Simulation, to be submitted
Lucassen, G.W.; Svaasand, L.O.; Verkruysse, W. & van Gemert, M.J.C. (1995). Laser energy
threshold for thermal vascular injury in a port wine stain skin model. Laser Med Sci,
Vol.10, No.4, (December 1995), pp. 231-234, 0268-8921 (Print), 1435-604X (Online)
Lucassen, G.W.; Verkruysse, W.; Keijzer, M. & van Gemer, M.J.C. (1996). Light distribution
in a port wine stain model containing multiple cylindrical and curved blood
vessels. Lasers in Surgery and Medicine, Vol.18, No.4, (April 1996), pp. 345-357, ISSN
0196-8092
Majaron, B.; Verkruysse, W.; Kelly, K.M.; & Nelson, J.S. (2001). Er:YAG laser skin
resurfacing using repetitive long-pulse exposure and cryogen spray cooling: II.
theoretical analysis. Lasers in Surgery and Medicine, Vol.28, No.2, (February 2001),
pp. 131-137, ISSN 0196-8092
Miller, I.D. & Veith, A. R. (1993). Optical modelling of light distributions in skin tissue
following laser irradiation. Lasers in Surgery and Medicine, Vol.13, No.5, (May 1993),
pp. 565–571, ISSN 0196-8092
Nelson, J.S.; Milner, T.E.; Anvari, B.; Tanenbaum, B.S.; Kimel, S.; Svaasand, L.O.; & Jacques,
S.L. (1995). Dynamic epidermal cooling during pulsed laser treatment of port-wine
stain: A new methodology with preliminary clinical evaluation. Arch Dermatol,
Vol.131, No.6, (June 1995), pp. 695–700, ISSN 1538-3652 (Print), 0003-987X (Online)
Niemz, M.H.; (1996). Laser-Tissue Interactions: Fundamentals and Applications, Springer Berlin
Heidelberg, New York. Hardcover, ISBN 978-3-540-40553-5 Softcover, ISBN 978-3-
540-72191-8
Orten S.S.; Waner M.; Flock S.; Roberson P.K.; Kincannon J. (1996). Port wine stains: an
assessment of five years of treatment. Arch Otolaryngol Head Neck Surg, Vol.122,
No.11, (November 1996), pp. 1174-9. ISSN 0886-4470 (Print) 1538-361X (Online)
Pearce, J. & Thomsen, S. (1995). Rate Process Analysis of Thermal Damage, In: Optical-
Thermal Response of laser-Irradiated Tissue, A.J. Welch & M.J.C. van Gemert, (Ed.),
561-606, Plenum Press, ISBN 0-306-44926-9, New York
Pfefer, T.J.; Barton, J.K.; Chan, E.K.; Ducros, M.G.; Sorg, B.S.; Milner, T.E.; Nelson, J.S. &
Welch, A.J. (1996). A three dimensional modular adaptable grid numerical model
for light propagation during laser irradiation of skin tissue. IEEE Journal of Selected
in Quantum Electronics, Vol.4, No.4, (December 1996), pp. 934–942, ISSN 1077-260X
Pfefer, T.J.; Smithies, D.J.; Milner, T.E.; van Gemert, M.J.C.; Nelson, J.S. & Welch, A.J. (2000).
Bioheat transfer analysis of cryogen spray cooling during laser treatment of Port
Wine Stains. Lasers in Surgery and Medicine, Vol.26, No.2, (February 2000), pp. 145-
157, ISSN 0196-8092
Pickering, J.W.; Butler, P.H.; Ring, B.J. & Walker, E.P. (1989). Computed temperature
distributions around ectatic capillaries exposed to yellow (578 nm) laser light. Phys
Med Biol, Vol.34, pp. 1247-1258, ISSN 0031-9155
Pickering, J.W. & van Gemert, M.J.C. (1991). 585 nm for the laser treatment of port wine
stains: A possible mechanism. Lasers in Surgery and Medicine, Vol.11, No.6, (June
1991), pp. 616-618, ISSN 0196-8092
Pikkula, B.M.; Torres, J. H.; Tunnell, J.W. & Anvari, B. (2001). Cryogen spray cooling: effects
of droplet size and spray density on heat removal. Lasers in Surgery and Medicine,
Vol. 28, No.2, (February 2001), pp. 103-112, ISSN 0196-8092
Shafirstein, G.; Bäumler, W.; Lapidoth, M.; Ferguson, S.; North, P.E. & Waner, M. (2004). A
New Mathematical Approach to the Diffusion Approximation Theory for Selective
Photothermolysis Modeling and Its Implication In Laser Treatment of Port-Wine
Stains. Lasers in Surgery and Medicine, Vol. 34, No.4, (April 2004), pp. 335–347, ISSN
196-8092
Shafirstein, G. & Buckmiller, L.M.; Waner, M. & Bäumler, W. (2007). Mathematical modeling
of selective photothermolysis to aid the treatment of vascular malformations and
hemangioma with pulsed dye laser. Lasers Med Sci, Vol.22, No.2, (June 2007), pp.
111-118, ISSN 0268-8921(Print), 1435-604X (Online)
Smithies, D.J. & Butler, P.H. (1995). Modelling the distribution of laser light in port-wine
stains with the Monte Carlo method. Phys Med Biol, Vol.40, No.5, (May 1995), pp.
701-33, ISSN 0031-9155
Tan, O.T.; Morrison, P. & Kurban, A.K. (1990). 585 nm for the treatment of port-wine stains.
Plast Reconstr Surg Med, Vol.86, No.6, (December 1990), pp. 1112-1117, ISSN 0032-
1052 (Print), 1529-4242 (Online)
Tunnell, J.W.; Nelson, J.S.; Torres, J.H.; & Anvari, B. (2000). Epidermal protection with
cryogen spray cooling during high fluence pulsed dye laser irradiation: an ex vivo
study. Lasers in Surgery and Medicine, Vol.27, No.4, (April 2000), pp. 373-383, ISSN
0196-8092
Tunnell, J.W.; Wang, L.V. & Anvari, B. (2003). Optimum pulse duration and radiant
exposure for vascular laser therapy of dark portwine skin: A theoretical study. Appl
Opt, Vol.42, No.7, (July 2003), pp. 1367–1378, ISSN 1559-128X (Print), 2155-3165
(Online)
Verkruysse, W.; Pickering, J.W.; Beek, J.F.; Keijzer, M.; van Gemert, M.J.C. (1993). Modeling
the effect of wavelength on the pulsed dye laser treatment of port wine stains. Appl
Opt, Vol.32, No.4, (February 1993), pp. 393–8, ISSN 1559-128X (Print), 2155-3165
(Online)
Wang, L.H.; Jacques, S.L. & Zheng, L.Q. (1995). MCML-Monte Carlo modeling of light
transport in multi-layered tissues. Comp Meth Prog Biol, Vol.47, No.2, (July 1995),
pp. 131-146, ISSN 0169-2607
Wilson, B.C. & Adam, G. (1983). A Monte-Carlo model for the absorption and flux
distribution of light in tissue. Med Phys, Vol.10, No.6, (December 1983), pp. 824-830,
ISSN 0094-2405
Zhou, Z.; Xin, H.; Chen, B. & Wang, G.-X. (2008a). Theoretical Evaporation Model of a Single
Droplet in Laser Treatment of PWS in Conjunction with Cryogen Spray Cooling.
Proceedings of 2008 ASME Summer Heat Transfer Conference, HT2008-56063, Aug. 10-
14, 2008, Hyatt Regency Riverfront, Jackonsville, Florida, ISBN: 978-0-7918-4849-4
Zhou, Z.; Xin, H.; Chen, B. & Wang, G.-X. (2008b). Single Droplet Evaporation Model in
Laser Treatment of PWS in Conjunction with Cryogen Spray Cooling. Proceedings of
2008 International Conference on Bio-Medical Engineering and Informatics, Vol. 1, pp.
551-556, May 28-30, Sanya, Hanan, China, ISBN: 978-0-7695-3118-2
28
Study of the Heat Transfer

Effect in Moxibustion Practice
Chinlong Huang and Tony W. H. Sheu
Dept of Engineering Sciences and Ocean Engineering,
National Taiwan University, No. 1, Sec. 4, Roosevelt Road, Taipei, 10617
Taiwan
1. Introduction
“First you use the needle (acupuncture), then the fire (moxibustion), and finally the herbs”
(Tsuei, 1996) has been well known in traditional Chinese medicine (TCM). In fact, moxibustion
has played an important role in Asia for many years (Zhang, 1993). In Huang Di Nei Jing
(Maoshing (translator), 1995), we can find that when needle can’t do a job, moxa is a better
choice. Moxibustion rather than acupuncture was commonly known to be able to alleviate
pains due to some severe diseases, manifested by vacuity cold and Yang deficiency. In
clinical studies, many experiments have confirmed that moxibustion is capable of enhancing
immunity, improving circulation, accommodating nerve, elevating internal secretion and
adjusting respiration, digestion and procreation et al. (Wu et al., 2001; Liu, 1999). However,
moxibustion has not been accepted as the modern therapy because of the lack of standard
practice procedures. In addition, moxibustion is subject to the danger of scalding patients.
More effort needs therefore to be made so as to increase our knowledge about the moxibustion
and, hopefully, these research endeavors can be useful for the future instrumentation and
standardization of the moxibustion by some emerging modern scientific techniques.
The existing moxibustion techniques can be separated into the direct and indirect moxibustion
therapies. In direct moxibustion, the ignited cone-shaped moxa is normally placed on the
skin surface near acupoints (Fig. 1(a)). Direct moxibustion can be further categorized into
the scarring and non-scarring two types. During the scarring moxibustion, the ignited moxa
is placed on the top of an acupoint till a time it burns out completely. This moxibustion type
may lead to a localized scarring or blister. In non-scarring moxibustion, moxa cone is also
burned directly on the skin. Such an ignited moxa will be removed when it may cause an
intense pain (moxa temperature should be under 60 ºC). Usually, this treatment will result
in a small red circular mark on the local area of the skin surface.
Indirect moxibustion becomes more popular currently because of its lower risk of leading to
pain or burning. A common way of administering the therapeutic properties of moxibustion
is to place, for example, a piece of ginger, garlic, salt or pepper in between the burning moxa
and the skin. One can also ignite a moxa stick, which is placed at a location that is closed to
but not in contact with the proper acupoint (about 2cm to skin surface normally), for several
minutes until the color of the skin surface near this acupoint turns red.
burning
moxa
burning needle
moxa handle
needle
calf section calf

section
(a) (b)
Fig. 1. Schematic of the moxibustions. (a) Direct moxibustion therapy; (b) Warm needle
moxibustion therapy
Another indirect moxibustion involves the use of needles and moxa. One needle, along
which there is a moxa, is inserted into the skin near an acupoint. The moxa cone placed on
the inserted needle is then ignited. The heat generated from the burning moxa will propagate
through the needle and transfers to the acupoint by heat conduction. During the treatment,
a dried moxa is in contact with the handle of the acupuncture needle after the needle being
inserted into the acupuncture point. This is followed by igniting the moxa and keeps it
burning (Fig. 1(b)). Typically, the distance between the skin surface and the burning moxa
stick is about 2 cm. Heat will be conducted from the needle handle to the needle itself and,
finally, to the surrounding tissues. This acupuncture design with a burning moxa can result
in a certain temperature gradient across the needle and enhances thus the Seebeck effect
(Cohen, 1997). In Chinese medicine theory, this method is highly recommended for use to
the patients with vacuity cold and wind damp (Wiseman, 1998) because of its functions of
warming the meridians and promoting the qi- and blood-flow. This therapy is also
applicable to release the cold-damp syndrome for the patients with rheumatoid arthritis
(Li, 1999). The other technique is called as the fire needle, which involves holding the needle
in a lamp flame until it becomes very hot. Afterwards, the needle is inserted to the
appropriate depth in the body quickly and it will be removed later on (Unschuld, 1988). In
comparison with the fire needle, the warming needle permits a longer retention and a
gentler heating.
In the present study, our aim is to study two types of the moxibustion effect, which are the
direct moxibustion and the warm needle moxibustion therapy. The acupoint GB 38 shown
in Fig. 2 is one of the acupoints in gall bladder (GB) meridian, which has an association with
the hemicrania and joint ache. Figure 3 shows the axial image of the right leg for the GB38
acupoint (Courtesy of Yang (1997)). At the calf section, the number of capillaries near the GB
acupoints is greater than those at the other parts of the body (Fei, 2000). Also, the distances
between the three acupoints GB37-GB38 and GB38-GB39 have about an inch. Therefore, the
acupoint GB38 was also the focus of other investigations (Sheu and Huang, 2008; Huang
and Sheu, 2008; Huang and Sheu, 2009).
Study of the Heat Transfer Effect in Moxibustion Practice 559
Fig. 2. Schematic of the stomach, gall bladder and bladder meridians. Note that GB38 is the
acupoint under current investigation (Chang, 1999)
Fig. 3. One axial image of the right leg that contains the GB38 acupoint (Courtesy of Yang
1997). Heaven, man and earth are three depths beneath of the skin surface
2. Materials and methods

Heat transfer process will be modelled by solving the energy conservation equation. In this
study, the energy equation cast in the following form for the total enthalpy will be
employed
∂( ρ h0 )
= ∇ ⋅ ( k∇T ) (1)
∂t
In the above equation, h0, k, ρ and T denote the internal energy, thermal conductivity,
density, and temperature, respectively.
Simulation of equation (1) will be carried out by employing the commercially available finite
volume package, namely, CFDRC (CFD-ACE-GUI, 2003). This software package provides
the modules CFD-GEOM for grid generation, CFD-ACE+ for solution solver, and CFD-
VIEW for post-processing. A convenient graphical user interface (GUI) is also available for
us to specify the physical properties of the medium under investigation, and the
specification of the boundary and initial conditions. In CFD-ACE+ solver, the finite volume
method employed together with the algebraic multigrid method and the conjugate gradient
squared solution solver accelerates calculation. In this study, the central difference scheme is
chosen to approximate the parabolic type partial differential equation.
tibia
calf GB38 GB38
fibula
(a) (b)
Fig. 4. (a) The generated surface meshes for the investigated calf section; (b) The zoom-in
area around GB38
A hybrid mesh, containing the structured- and unstructured-type meshes, shown in Fig. 4, is
generated from a total number of 17,000 nodal points. These mesh points have been
properly distributed so that the predicted solutions can more accurately represent the
physical phenomenon. In all the investigations, calculation of the enthalpy will be

terminated when the residual norms fall below 10-15. As the number of the employed nodal
points (with 17,000 nodal points) is increased by 50% (25,500 nodal points), only a negligibly
small difference is seen in the simulated results. On the contrary, when the number of the
mesh points is decreased by 50% (or 8,500 nodal points) the computed difference shows an
apparent difference. Hence the mesh generated by 17,000 nodal points will be employed in
the present simulation.
In order to know the temperature distribution due to the ignited moxa, the IR images are
collected under the almost dark condition. Both of the camera and the subject are kept apart
from the external draft IR source. The room temperature and the relative humidity are kept
at 22 ± 0.1 ºC and 60 ± 7%, respectively. Thermography is taken using a calibrated IR camera
(ThermaCAM® SC500 from FLIR Systems TM) (ThermaCAM SC500 Operator’s Manual),
which is equipped with a 45°close-up optic. Sensitivity, accuracy and resolution of the
employed camera are kept at 0.07 ºC, ± 2 ºC and 320 × 240 pixels, respectively. The distance
between the camera and the subject under current investigation is 0.1 m. The infrared
images of the subject obtained at a sampling rate of 4 Hz will be directly recorded in the
computer’s hard disk.
BC II
tibia
BC III
(TE =22 o C ) BC I (T=37oC)
calf GB 38
BC I (T=60oC)
fibula
BC II
BC I: isothermal BC II: symmetry
BC III: external heat transfer (by convection)
Fig. 5. Schematic of the calf section around the acupoint GB 38 and the specified boundary
conditions
3. Results and discussion

In the present study, direct moxibustion and warm needle moxibustion treatments are
described below:
(A) Direct moxibustion therapy
In direct moxibustion therapy, a burned moxa (moxa cone of 1 cm diameter and 1cm length,
made by the dried Artemisia vulgaris leaves with a weight of 100 mg), was placed on
acupoint GB38 shown in Fig. 4. The boundary conditions applied at the calf section with
tibia and fibula are shown in Fig. 5, where BC I, II and III represent the isothermal,
symmetric and heat transfer (by convection) boundary conditions, respectively. In BC III,
the wall subtype is chosen to account for the transfer of heat to/from the external
environment (i.e., the area outside of the computational system) by heat convection. On the
boundary BC III, it is specified by the condition
q w = hc (TE − Tw ) (2)
where hc and TE denote the heat transfer coefficient and the environment temperature. The
wall temperature (Tw) is determined by balancing the heat fluxes between the environment
and the skin surface.
The heat transfer coefficient of the skin surface, specific heat and the density of tissues in the
investigated calf section are denoted as hc, Cpc and ρc, respectively. At the normal state, these
coefficients are prescribed respectively with hc = 3.7 W/m2ºC (Nishi and Gagge, 1970),
Cpc = 3,594 J/kgºC (Blake et al., 2000) and ρc = 1,035 kg/m3. The thermal conductivity of the
human tissues is assumed to change with the temperature (T) by the equation kc = 0.840419+
0.001403T W/mºC (Mura et al., 2006). The rest of the employed coefficients are tabulated in
the Table 1.
ks 16 W/m ºC Thermal conductivity of the stainless steel

kc 0.840419+0.001403T W/m ºC Thermal conductivity of the calf
ka 0.0299 W/m ºC Thermal conductivity of the air
Cps 460 J/kg ºC Specific heat of the stainless steel
Cpc 3,594 J/kg ºC Specific heat of the calf
Cpa 1,009 J/kg ºC Specific heat of the air
ρs 7,800 kg/m3 Density of the stainless steel
ρc 1,035 kg/m3 Density of the calf
ρa 1.0 kg/m3 Density of the air
hs 7.9 W/m2 ºC Heat transfer coefficient of the stainless steel
hc 3.7 W/m2 ºC Heat transfer coefficient of the calf
TE 22 ºC Environment temperature
Table 1. Summary of the coefficients and the prescribed temperatures in the current
simulation
The predicted temperature on the skin surface is plotted in Fig. 6. The moxa temperature on
the skin surface is specified at 60 ºC (non-scarring direct moxibustion), which is the highest
temperature that our skin can possibly endure, to avoid scarring. On the tibia and fibula,
their temperatures are given to be T= 37 ºC, which is the same as the normal human body
temperature. One can find from the upper and bottom planes of the simulated domain that
the temperature near tibia and fibula has a larger value (Fig. 6). Since moxibustion takes
place at a location near GB38, the area around this acupoint has a higher temperature.
Figure 7 shows the predicted skin surface temperature that is distributed in a form similar to
the experimentally measured temperature by IR image.
59
57
55
53
51
49
47
45
43
41
39
37
35
Fig. 6. The predicted skin surface temperature contours subject to the moxibustion
60
57
56
52
47 48
44
40
37 36
32
28
27
24
(a) (b)
Fig. 7. Comparison of the predicted and measured temperatures. (a) Measured by the IR
image recording system; (b) Predicted by the numerical simulation
heaven
man
330 earth
59
57
55
53
51
49 T 320
B A 47
45
43
41
39
37
35
310
0 0.02 0.04 0.06

depth (m)
A B
(a) (b)
Fig. 8. (a) The predicted temperature contours on the cutting plane that passes through the
acupoint GB38 when the moxa temperature on the skin surface is 60 ºC; (b) The predicted
temperature profile along the line connected by two points A and B shown in Fig. 8(a). The
“heaven”, “man” and “earth” represent three depths of the investigated acupoints,
respectively
In TCM, there is one concept, “When using acupuncture, shallow position treats diseases
nearby; deep position treats diseases far away”. Huang Di Nei Jing (Maoshing (translator),
1995) supported the similar concept that at the same acupoint shallow position treats mild
diseases; deep position treats severe diseases. This enlightens that at the same acupoint
different depths have different effects and even have an association with different diseases.
For this reason, this study makes an effort to get the temperature distributions at different
depths. Figure 8 (a) shows the temperature contours predicted at the plane of the acupoint
GB38 for the case that the moxa temperature on the skin surface is 60 ºC. From Fig. 8 (b), one
can see the predicted temperature profile along a line that is connected by two nodes A and
B. For the acupoints “heaven”, “man” and “earth”, they have three different depths at a
location in between the skin surface and the associated connective tissue. When the moxa
temperature is controlled at 60 ºC, which is the temperature considered in the case of non-
scarring moxibustion, the temperatures at the “heaven” (~ 0.5 cm beneath of the skin
surface), “man” (~ 1.0 cm beneath of the skin surface) and “earth” (~ 1.5 cm beneath of the
60 60 60
skin surface) are predicted as Theaven = 47.8 ºC, Tman = 41.7 ºC and Tearth = 39.0 ºC. From the
skin surface to “heaven”, we found that temperature decreases faster (12.2 ºC in between)
than those from the “heaven” to “man” (6.1 ºC in between) and from the “man” to “earth”
(2.7 ºC in between) as well because our body has a bigger thermal capacity than that of the
moxa. As a result, the temperature variation on the side of “heaven” is greater than that
along the “earth”. The relation between the predicted temperature (T) and the depth (x) can
be expressed as
T = 60 − 5.31 × 10 −2 x + 5 × 10 −5 x 2 − 3 × 10 −8 x 3 + 7 × 10 −12 x 4 − 1 × 10 −15 x 5 + 6 × 10 −20 x 6 (3)

heaven
107 man
97
earth
92
87
82 87
77
72
67
62 T
B 57
A 52 67
47
42
37
47
0 0.02 0.04 0.06

depth (m)
A B
(a) (b)
respectively
heaven
T man
197
187 earth
177 177
167
157
147
137
127
117
B A 107 T 127
97
87
77
67
57 77
47
37
0 0.02 0.04 0.06

depth (m)
A B
(a) (b)
respectively
We then consider the scarring moxibustion by specifying the moxa temperatures at two
higher temperatures 100 and 200 ºC. The temperatures at three locations are predicted to be
100 100 100 200 200 200
Theaven = 68.1ºC, Tman = 51.3ºC and Tearth = 43.5ºC; Theaven = 116.0ºC, Tman = 72.8ºC and Tearth = 53.4º,
respectively. The predicted temperature contours on the plane of acupoint GB38 and the
temperature profile along the line connected by nodes A and B are shown in Fig. 9 and
Fig. 10. The relation of the predicted temperature (T) and the depth (x) is expressed by the
equations given below
T = 100 − 1.374 × 10 −1 x + 1 × 10 −4 x 2 − 6 × 10 −8 x 3 + 2 × 10 −11 x 4 − 2 × 10 −15 x 5 + 1 × 10 −19 x 6 (4)
T = 200 − 3.48 × 10 −1 x + 3 × 10 −4 x 2 − 2 × 10 −7 x 3 + 4 × 10 −11 x 4 − 6 × 10 −15 x 5 + 3 × 10 −19 x 6 (5)

According to equations (3), (4) and (5) for the burning moxa cones with the temperatures at
60, 100 and 200 ºC, the temperatures at the “heaven” are decreased to 47.8, 68.1 and 116.0 ºC,
respectively. In the scarring type (moxa temperatures are 100 and 200 ºC), the temperature
decrease from the skin surface to the “heaven” is faster (31.9 and 84 ºC in between the skin
surface and the “heaven” position) than that of the non-scarring type (12.2 ºC in between the
skin surface and the “heaven” position) since our human body has a larger thermal capacity
than the moxa. The temperature, as a result, will quickly reach the human body’s
temperature (37 ºC). In summary, the higher the moxa temperature, the more rapid
temperature decrease will be.
Fig. 11. Acupuncture needle is composed of one-column needle (portion I) and one coil-like
handle (portion II), which covers the column needle
(B) Warm needle moxibustion therapy
In warm needle moxibustion, a needle assembly for the acupuncture includes a column
needle (portion I) with its head having a sharp end and one coil-like handle (portion II),
which covers the dull end of the column needle (Fig. 11). Between the column needle and
the coiled handle, there is a triangle air volume (portion III). Figure 11 shows the
investigated 1.5-inch acupuncture needle. A hybrid mesh system, containing both of the
structured- and unstructured-type meshes shown in Fig. 12, is generated from a total
number of 7,421 mesh points. The mesh density has been properly distributed so that the
predicted solutions can accurately represent the physical phenomenon.
Fig. 12. The surface mesh points generated on the needle for acupuncture use
BC Ib (T=37oC)
BC II
tibia
fibula BC III
o
GB38
GB38
needle
calf BC Ia (T=200oC)
BC Ia: isothermal (T=200 oC) BC Ib: isothermal (T=37 oC)

BC II: symmetry BC III: external heat transfer (by convection)
Fig. 13. Schematic of the boundary condition types for the acupuncture needle inserted to
the acupoint GB38
The boundary conditions are applied on the calf section with the needle shown in Fig. 13,
where BC I, II and III represent the isothermal, symmetric and convection types of boundary
conditions, respectively. BC Ia shows the boundary of needle handle which has a moxa. The
temperatures in the bone zone (BC Ib), including both fibula and tibia, are specified as 37 °C
to simulate the human body temperature. The symmetric conditions at the upper and
bottom sides of the calf section (BC II), which connects knee and foot, are specified to
account for the same structures of muscle tissues with knee and foot. BC III is normally
chosen to simulate the transfer of heat to/from the surrounding (i.e., the area outside of the
computational system) by convection. This subtype is used to fix either the wall temperature
or the heat flux. On the wall of BC III, it is prescribed by equation (2).
200 ºC 200 ºC
needle
133 ºC
133 ºC
27 ºC
27 ºC burning moxa
27 ºC
Numerical simulation Experimental measurement
Fig. 14. Comparison of the numerically predicted and the experimentally measured
temperatures for the burning moxa stick applied on the handle of the acupuncture needle.
The temperature of burning moxa is 200 ºC. Temperature gradient will be established across
the needle, starting from the needle handle and ending at the needle head
In TCM, acupuncture needle can be made from stainless steel, iron, copper, silver and
pottery etc. For the safety and cost effectiveness reasons, stainless steel is now more popular.
This study employs the normal stainless steel needle (1.5 inch needle, 7 cm total length with
3cm length in the needle handle part) in the experiment and numerical simulation. The
thermal conductivity, specific heat, density and heat transfer coefficient of the stainless steel
are denoted as ks, Cps, ρs and hs, respectively. At the normal state with the room temperature
(22 ºC in this study), these coefficients are prescribed with the values of ks = 16 W/m °C,
Cps = 460 J/kgºC, ρs = 7,800 kg/m3 and hs = 7.9 W/m2°C (The Engineering Tool Box,
http://www.engineeringtoolbox.com/), respectively. The other associated coefficients are
tabulated in Table 1. The predicted convergent temperature, which is plotted in Fig. 14, on
the needle surface is seen to distribute similarly to the experimentally measured values by
IR image.
GB38
Fig. 15. The predicted skin surface temperature contours for the warm needle applied at the
location near the acupoint GB38
Figure 15 shows the predicted temperature contours near the acupoint GB38 under the
warm needle moxibution. Around the skin surface of GB38 (Fig. 16), the surface temperature
near the needle is predicted to be about 36 ºC, which is 1 ºC higher than those predicted at
other skin surfaces (~ 35 ºC). Figure 17 shows the predicted temperature contours on the
cutting plane passing through the acupuncture needle and GB38 acupoint. The tissue
temperature around the needle is only 1 ºC higher than the others. Since the needle is
smaller in comparison with the human body, the effect of the burning moxa on the needle
handle is relatively less significant. In TCM, warm needle is involved in the acupuncture
with burning moxa. From Figs. 16 and 17, we can understand why this treatment isn’t called
as the hot needle, even the burning moxa can reach the temperature that is as high as 200 ºC.
Figure 18 shows the predicted temperature profile along the center line of the acupuncture
needle. From this profile, we can see that the temperature can be varied from 200 ºC
(burning moxa section) to 36 ºC (skin surface) in a short length of 2 cm. We can also observe
that the temperatures at locations beneath of the skin surface of depths 0 mm, 1 mm, 2 mm,
3 mm and 4 mm are 36.00 ºC, 35.63 ºC, 35.59 ºC, 35.62 ºC and 35.66 ºC, respectively.
GB38
Fig. 16. The predicted skin surface temperature contours very near the acupoint GB38
skin surface
depth 2 mm
needle
GB38
Fig. 17. The predicted temperature contours on the cutting plane, which passes through the
acupoint GB38
177
127
T
77
27
0 0.02 0.04 0.06
lenght of the needle
burning moxa section skin surface

Fig. 18. The predicted temperature distribution along the centerline of the acupuncture
needle for a warm needle applied at the acupoint GB38
4. Conclusions
In the present study, both of the experimental measurement and the numerical simulation
have been conducted to get the temperature distribution near the acupoint which is
subjected to a direct and a warm needle moxibustion treatments. The predicted temperature
contours on the skin and needle surface are shown to agree well with the experimentally
measured results. Moreover, the temperature contours of the tissues beneath of the skin
surface have been also numerically predicted. In direct moxibustion, even for the lower
temperature treatment type (non-scaring type, moxa temperature equal to 60 ºC), human
tissue temperature is still increased by 12 ºC at the heaven position. This explains why
moxibustion rather than acupuncture has a better ability to treat severe diseases. In warm
needle moxibustion, along the centerline of the acupuncture needle the temperature
decreases very rapidly from the burning moxa section (200 ºC) to the skin surface (36 ºC).
The temperature near the needle is only 1 ºC higher than those predicted at other places.
This predicted phenomenon explains why we call it as the warm needle rather than as the
hot needle treatment. These results are fundamentally important in the study of TCM.
However, the physiological effect of warm needle moxibustion should include the
contributions of acupuncture. For this reason, a better understanding of the total effect of the
human body needs more intensive studies.
5. Acknowledgments
The financial support by the National Science Council under Grants NSC 97-2221-E-002-250-
MY3 is gratefully acknowledged. The authors also will acknowledge their thanks to Prof.
Ping-Hei Chen who kindly provided us the FLIR system to perform temperature
measurement.
6. References
[1] Blake, A. S. T., Petley, G. W., and Deakin, C. D., Effects of changes in packed cell volume
on the specific heat capacity of blood: implications for studies measuring heat
exchange in extracorporeal circuits. British Journal of Anaesthesia, 84(1), 2000, pp.
28-32.
[2] CFD-ACE-GUI User Manual Volume II. CFD Research Corporation, 2003, pp. 85-94.
[3] Chang, K. W., Meridian Anatomy. Zhi-Yang Press, 1999, pp. 47.
[4] Cohen, M., Kwok, G., and Cosic, I., Acupuncture needles and the Seebeck effect: do
temperature gradients produce electro-stimulation? Acupuncture and Electro -
therapeutics Research, 22, 1997, pp. 9-15.
[5] Construction of ThermaCAM SC500 Operator’s Manual. FLIR Systems, 1999.
[6] Fei, L., Researches and developments of meridian physical basic and function experiments.
Chinese Science Bulletin, 43(6), 2000, pp. 658-672.
[7] Huang, V. C., and Sheu, T. W. H., On a dynamical view on the meridian transmission.
Journal of Accord Integrative Medicine, 4(2), 2008, pp.97-107.
[8] Huang, V. C., and Sheu, T. W. H., Tissue fluids in microchannel subjected to an
externally applied electric potential. International Journal for Numerical Methods
for Heat & Fluid Flow, 19(1), 2009, pp. 64-77.
[9] Li, C., Jiang, Z., and Li, Y., Therapeutic effect of needle warming through moxibustion at
twelve shu points on rheumatoid arthritis. Journal of Traditional Chinese Medicine,
19(1), 1999, pp. 22-26.
[10] Liu, G. J., Traditional Chinese Moxibustion. Nanchang, The Jiangxi Press of Science and
Technology, 1999, pp. 339-352.
[11] Maoshing, N., (translator) The Yellow Emperor’s Classic of Medicine: A New
Translation of the Neijing Suwen with Commentary. Boston, 1995.
[12] Murat, T., Unal, C., and Cem, P., The bio-heat transfer equation and its applications in
hyperthermia treatments. Engineering Computations, 23(4), 2006, pp. 451-463.
[13] Nishi, Y., and Gagge, A. P., Direct evaluation of convective heat transfer coefficient by
naphthalene sublimation. Journal of Applied Physiology, 29(6), 1970, pp. 830-838.
[14] Sheu, T. W. H., and Huang, V. C., Development of an electro osmotic flow model to
study the dynamic behavior in human meridian. International Journal for
Numerical Methods in Fluids, 56, 2008, pp. 739–751.
[15] The engineering toolbox. Tools and Basic Information for Design, Engineering and
Technical Applications, http://www.engineeringtoolbox.com/.
[16] Tsuei, J. J., The science of acupuncture - theory and practice. IEEE Engineering in
Medicine and Biology Magazine, May/June, 1996, pp. 52–57.
[17] Unschuld, P. U., Introductory Readings in Classical Chinese Medicine. Kluwer
Academic Publishers, Dordrecht, Holland, 1988.
[18] Wiseman, N., and Feng, Y., A Practical Dictionary of Chinese Medicine. Paradigm
Publications, Brookline, MA, 1998, pp. 664.
[19] Wu, M. X., Liu X. X., and Wu. B., The moxibustion influence of tendency in rehabilitation
for mouse osteoporosis fracture. Moderate Rehabilitation, 5(6), 2001, pp. 46-47.
[20] Yang, J. K., Meridian Cross-section Anatomy. Shang Hai Science Technology Press,
Chinese, 1997, pp. 76.
[21] Zhang, Q. W., Chinese Moxibustion Handbook. Tianjin, The Tianjin Press of Science
and Technology, 1993, pp. 1-5.
[22] Zhu, B., Scientific Foundations of Acupuncture and Moxibustion. Qingdao Press,
Qingdao, 1998.
29
Heat and Mass Transfer in

Jet Type Mold Cooling Pipe
Hideo Kawahara
Oshima National College of Maritime Technology
Japan
1. Introduction
Impinging jet is widely used in heating and cooling applications due to their excellent heat
transfer characteristics. To optimize heat transfer an understanding of the temperature field
as well as the velocity field is essential, in particular near the impingement surface where
the flow characteristics dominate the heat transfer process. Heat transfer distributions of jet
impingement and the effect of various geometric and flow parameters on heat transfer are
well documented, for example by Miranda & Campos (1999,2001), Kayansayan & Kucuka
(2001), Hrycak (1981), Gau & Chung (1991) and Lee et al. (1997).
Jet cooling pipes are used across a wide range of temperatures in die cooling, probes for
cryogenic surgery and other applications. Coolants are diverse, and range from liquid
nitrogen to water. In practice, these pipes are used especially with dies to improve quality,
by regulating die temperature and preventing sticking. Die cooling can be classified into
two basic types: straight-flow and jet. With the straight-flow type, lines for providing a flow
of cooling water are placed so as to follow the surface of the molded part, and this approach
is mainly used to uniformly cool the entire part. However, molded parts with a complex
form cannot be properly cooled with straight-flow only. Temperature is controlled to the
proper level at such points by inserting jet cooling pipes to provide spot cooling.
The problem with jet cooling pipes is that the heat transfer and flow pattern become unclear
near the end of the cooling pipe, i.e. in the region where the jet impinges on the cooling pipe.
At present, the heat transfer coefficient is calculated by assuming a double-pipe annular
channel and using the associated empirical equation. However, in cases where more
sensitive temperature control is necessary, accurate values must be measured and used.
Also, the cooling pipe is made of metal, and thus its internal flow has not been observed.
However, since heat transfer is closely related to the flow of the fluid acting as a coolant, it is
also important to observe that flow pattern.
This research examined transfer phenomena characteristics in jet cooling pipe, using
channels with the same shape as the actual cooling pipe. This was done by visualizing the
flow pattern, measuring the flow velocity distribution at the jet outlet, and measuring heat
transfer. Here, it is necessary to measure the local distribution and average as heat transfer,
but since the dimensions of the test section are small, and the shape is complex compared
with an impinging jet onto a flat plate etc., direct measurement of heat is difficult. Therefore,
mass transfer was measured using the electrode reaction method, which enables particularly
exact local measurement, and heat transfer was predicted based on the analogy between
mass transfer and heat transfer.
2. Experimental apparatus and method

Fig. 1 shows the jet type cooling pipe used in the experiment. The cooling pipe has a dual
structure, with an inner and outer pipe. The outer pipe of the actual cooling equipment is
(a) Flow visualization
(b) Point electrode (Limited part mass transfer measurement)
(c) Whole electrode

Fig. 1. Shape of the test section
Heat and Mass Transfer in Jet Type Mold Cooling Pipe 575
fashioned by drilling a hole in the die, but in this experiment it was fabricated with acrylic
resin to enable visualization. The inner diameter is do2=22mm, and the end is worked into a
hemispheric shape. To enable use of the electrode reaction method, one pipe was fabricated
with 9 embedded point electrodes, and another with the entire part near the end made into
an electrode. The inner pipe is made of stainless steel, and is the same as that used in the
actual cooling equipment. It has an inner diameter di of 8mm and an outer diameter do1 of
10mm. It is inserted into the outer pipe, and delivers the working fluid for cooling as a jet.
Distance from the end of the outer pipe to the jet outlet of the inner pipe was varied in 5mm
units in the range from 5 to 25mm.
Fig. 2 shows a schematic diagram of the experimental apparatus. Working fluid is delivered
from the tank by a centrifugal pump. Then it passes through the flow meter and test section,
and returns again to the tank. Temperature of the working fluid is controlled using a heat
exchanger in the tank. The flow rate range was set to 0.25-2.0ml/min. The Reynolds
numbers used were jet Reynolds numbers taking flow velocity ui at the inner pipe outlet as
a baseline, as defined below.
ui ⋅ di
Re j = (1)
ν
where ν is a kinematics viscosity. Their range is 75<Rej<5700. In this experiment, the
aluminium powder method was used as the method of flow visualization. LDV was used
for velocity measurement. With LDV, measurement was done at 0.5mm spacing in the
radial direction from the central axis of the jet, with the purpose of measuring the velocity
distribution at the inner pipe outlet.
Fig. 2. Schematic of experimental apparatus

2.2 Method of measuring mass transfer using electrode reaction method

The jet type cooling pipe is a device for achieving heat exchange, but in this research,
instead of directly measuring that heat transfer, mass transfer was measured since it has an
analogy to heat transfer. The electrode reaction method was used for measurement. The
following equation gives the amount of mass transfer in the vertical direction on the
electrode surface.
⎛ Z ⋅ F ⎞ ⎛ ∂Φ ⎞ ⎛ ∂C ⎞
J = −D ⋅ C ⎜ ⎟⎜ ⎟ − D⎜ ⎟ +C ⋅U (2)
⎝ R ⋅ T ⎠ ⎝ ∂y ⎠ ⎝ ∂y ⎠
where:
- J is the mass flux in distance y from electrode surface [mol/(cm2 s)];
- D is the diffusion coefficient [cm2/s];
- C is the concentration [mol];
- Z is the ion [-];
- F is the Faraday constant (=96500[c/mol]);
- R is the Gas constant [cm2/(s2K)];
- T is the fluid temperature [K];
⎛ ∂Φ ⎞
- ⎜ ⎟ is the potential gradient [-];
⎝ ∂y ⎠
⎛ ∂C ⎞
- ⎜ ⎟ is the concentration gradient [-];
⎝ ∂y ⎠
- U is the velocity in which the volume element in the solution moves along the y axis
[cm/s];
The terms on the right side of the above equation are, from the left: the electrophoresis term,
diffusion term and convection term. Flux at the electrode is as follows.
I
J= (3)
A⋅Z⋅F
where I and A are the current [A] and electrode surface area [cm2] respectively.
A mixture of potassium ferricyanide and potassium ferrocyanide of the same concentration
is used as the electrolyte solution. Therefore, the following reactions occur at the electrodes.
Cathode plane (Reductive reaction):
3− 4−
Fe (CN )6 + e − → Fe (CN )6 (4)
Anode plane (Oxidation reaction):
4− 3−
Fe (CN )6 → Fe (CN )6 + e − (5)
Since this is a oxidation-reduction reaction, and involves only the transfer of electrons, it can
be assumed that there is no flow in the vertical y direction on the electrode surface. Thus
U=0, i.e. the convection term in equation (2) vanishes.
Electrophoresis is a phenomenon whereby ions in solution are moved by electric force due
to an electric field. If an uncreative supporting electrolyte (NaOH) is added here, it will
cause the same kind of electrophoresis, and lessen electrophoresis of the electrolyte.
Electrophoresis of electrolyte is reduced by adding this in a large amount. Therefore, the
potential gradient of the electrolyte becomes 0, and the electrophoresis term in equation (2)
vanishes. As a result, only the diffusion term on the right side remains in equation (2).
⎛ ∂C ⎞
J = −D ⎜ ⎟ (6)
⎝ ∂y ⎠
J is proportional to the difference in concentration between the electrode surface and the
fluid body, and is given by the following equation.
J = kD ( C b − C w ) (7)
where kD is the mass transfer coefficient [cm/s], Cb and Cw are the fluid average
concentration [mol/cm3] and concentration of the electrode surface [mol/cm3], respectively.
The current obtained with diffusion control is defined as the limiting current Id. The reaction
in equation (4) is carried out extremely rapidly under diffusion control conditions, and thus
concentration of Fe(CN) at the anode surface can be regarded as Cw=0. Thus equation (7)
becomes:
J = kD ⋅ C b (8)
Substituting (3) into (8) and rearranging, the equation becomes:
Id (9)
kD =
A ⋅ Z ⋅ F ⋅ Cb
Since A, Z and F are constants, the mass transfer coefficient kD can be found by assaying
fluid concentration Cb and measuring the limiting current. The Schmidt number (Sc) which
corresponds to the Prandtl number (Pr) in heat transfer, and the Sherwood number (Sh)
which corresponds to the Nusselt number (Nu), can be found as follows.
ν (10)
Sc =
D
kD ⋅ di I d ⋅ di (11)
Sh = =
D A ⋅ Z ⋅ F ⋅ Cb ⋅ D
Next, the wall shear stress is defined. Since the electrode surface is circular, the velocity
gradient sr of the wall surface can be calculated according to the following equation from the
limiting current.
I d3
sr = 3.55 × 10 −15 (12)
D ⋅ C b3 ⋅ dE5
2
where dE is the point electrode diameter.

From Newton's law of viscosity, shear stress is:
du (13)
τ = −μ
dy
where μ is the fluid viscosity [g/(cm s)].

Therefore, wall surface shear stress can be determined as follows.
μ ⋅ I d3
τ w = μ ⋅ sr = 3.55 × 10 −15 (14)
D2 ⋅ C b3 ⋅ dE5
It is also possible to predict the distribution of local mass transfer amounts from the wall
surface shear stress when mass transfer occurs at the entire wall surface. If S is taken to be
the distance in the flow direction from the stagnation point at the wall surface, and sr (S) is
taken to be the velocity gradient of the wall surface, then the local concentration gradient
can be expressed as follows if separation does not occur in the flow.
12
⎛ ∂C ⎞ C b ⎣⎡sr ( S ) ⎦⎤
⎜ ⎟ = (15)
{ }
13
⎝ ∂y ⎠s Γ ⎛ 4 ⎞ s 12
⎜ ⎟ 9D∫0 ⎣⎡sr ( x ) ⎤⎦ dx
3
⎝ ⎠
⎛4⎞
where Γ ⎜ ⎟ is the gamma function.
⎝3⎠
Substituting equation (6) and equation (8) into equation (15),
12
kD ⋅ C b C b ⎡⎣sr ( S ) ⎤⎦
= (16)
{ }
13
D ⎛4⎞ s 12
Γ ⎜ ⎟ 9D∫ ⎡⎣sr ( x ) ⎤⎦ dx
⎝3⎠ 0
Thus, using the following equation, it is possible to calculate the flow direction distribution
of local Sherwood numbers, which are the dimensionless mass transfer amounts.
12
kD ⋅ di di ⎡⎣sr ( S ) ⎤⎦
ShL ( S ) = = (17)
{ }
13
D ⎛4⎞ 12
Γ ⎜ ⎟ 9D∫ ⎡⎣sr ( x ) ⎦⎤ dx
⎝3⎠
Rewriting this equation with dimensionless numbers, it becomes:
12
⎡ ⎤
13 2 3 ⎢ τw ⎥
Sc ⋅ Re j ⎢ ⎥
1
⎢ ρ ⋅ ui2 ⎥
ShL ( S ) = ⎣2 ⎦ (18)
13
⎧ ⎡ ⎤
12 ⎫
⎪
⎪ s⎢ τw ⎥ ⎪⎪
⎛4⎞
Γ ⎜ ⎟ ⋅ 181 3 ⎨ ∫ ⎢ ⎥ dx di ⎬
⎝3⎠ ⎪ ⎢ 1 ρ ⋅ ui2 ⎥
0
⎪
⎩⎪ ⎣ 2 ⎦ ⎭⎪
where ρ is the fluid density [g/cm3], and ui is the average velocity of the inner pipe nozzle
exit.
2.3 Analogy of mass transfer and heat transfer

The amount of mass transfer is proportional to the concentration gradient, and is given by
Fick's law.
dC
J = −D (19)
dy
The amount of heat transfer is proportional to the temperature gradient, and is given by
Fourier's law.
dT
q = −λ (20)
dy
where λ is the thermal conductivity, and T is the fluid temperature.

Due to these relationships, there is a similarity in transport phenomena due to molecular
motion between mass transfer and heat transfer, and an analogy holds between the two. In
this experiment, since the electrode reaction is a diffusion controlled reaction and convection
does not occur, the concentration boundary layer assumes a state very similar to a thermal
boundary layer in the case of heat transfer. Therefore, mass transfer and heat transfer have
the following sort of correspondence, and can be interchanged.
C ⇔ T , Sc ⇔ Pr , Sh ⇔ Nu
This experiment employed the Colburn j factor analogy, which is widely used in
experiments and particularly effective for turbulence (Kobayashi, 1996). The J factors for
mass transfer and heat transfer are, respectively:
Sh
jD = (21)
Re j ⋅ Sc 1 3
Nu
jH = (22)
Re j ⋅ Pr 1 3
Here, the relationship jD=jH holds, and thus the amount of heat transfer can be inferred by
analogy as follows:
h ⋅ di Sh
Nu = = ⋅ Pr 1 3 (23)
λ Sc 1 3
λ Sh ⋅ Pr 1 3
h= (24)
di Sc 1 3
where h is the heat transfer coefficient.
Heat transfer coefficients in the jet-type cooling pipe of the die are predicted using values
derived from an empirical equation (Mizushima et al., 1952) for the case where turbulence
actually flows through the annular channel of a dual pipe, and calculation is done using the
following equation.
Rej =100
Re j =200
Re j =300
(a) L/di=0.63
Rej =100
Re j =200
Re j =300
(b) L/di=1.88
Rej =100
Re j =200
Re j =300
(c) L/di=3.13
Fig. 3. Stream line distribution by numerical simulation
0.53
h ⋅ de ⎛d ⎞
= 0.02 ⎜ o 2 ⎟ Re∗0.8 ⋅ Pr 1 3 (25)
λ ⎝ di ⎠
Here, Re* are the Reynolds numbers for an annular channel, and are expressed as follows.
de ⋅ uo
Re∗ = (26)
ν
where de is the equivalent diameter, and uo is the cross section average velocity in the
annular channel. The equivalent diameter used hydraulic diameter shown in the following.
4 Ac (27)
de =
l
where Ac is the cross section of the annular channel, and l is the open verandah length.
3. Experimental results
3.1 Numerical simulation
In first, the numerical simulation before the experiment had forecast the flow in the jet flow
type coolant pipe. Fig. 3 shows the result. The distance between the nozzle and the coolant
pipe point has been changed with L/di=0.63, 1.88, and 3.13. The flow was assumed to be a
stationary flow as calculation condition, in addition, ranges of the Rej number are from 100
to 300.The jet flow that occurred from the tube nozzle in L/di=0.63 of Fig. 3(a) flows along
the wall, and forms the boundary layer flow, and the vortex is not generated, and even if the
Rej number is raised, is not observed the vortex. The boundary layer flow is formed with the
coolant pipe point for L/di=1.88 of Fig. 3(b) as well as Fig. 3(a), and the vortex occurs
between the nozzle and the coolant pipe point. This vortex grows as Rej increases, and, the
flow parts from the outside tube after it passes in this vortex, and bounds. This bounding
flow becomes remarkable as Rej increases. And, because the flow is decelerated in the
vicinity of the wall, the occurrence of the second vortex is forecast by flaking off. In the flow
for L/di=3.13 of Fig. 3(c), the vortex becomes long and slender in the case of Fig. 3(b), and
the bounding flow is unremarkable.
Therefore, in this experiment, the first vortex that occurs in the coolant pipe point is defined
as Vortex A. And, the second vortex to which generation is forecast in the downstream is
defined as Vortex B.
3.2 Flow visualization

Fig. 4 shows a flow visualization photograph for L/di=0.63. The white line in the figure
indicates the position of the point electrodes. In this experiment, the vortex that arises near
the cooling pipe tip is defined as a Vortex-A, and in addition, the vortex that arises in the
downstream is defined as a Vortex-B. With Rej=106, there is a flow along the wall surface,
and Vortex-A does not occur. However, when the Rej number is increased, Vortex-A occurs
along the side surface of the inner pipe, and accompanying that, Vortex-B also appears. The
flow for L/di=0.63 is steady, and Vortex-A and Vortex-B expand toward the downstream
side up to around Rej=1000. At Rej numbers above that, the flow gradually becomes
unsteady, but Vortex-A continues to exist. In the range 603<Rej<856, a vortex with rotation
opposite to Vortex-A occurs on the side wall surface at the end of the inner pipe. This vortex
is surrounded by Vortex-A, and here is called Vortex-C.
Rej=106 Rej=713
Rej=1141 Rej=2282
Fig. 4. Flow visualization photograph by the aluminium powder method for L/di=0.63
t = 0.00s t = 0.85s t = 1.70s
t=2.60s t=3.40s t=4.30s
t=5.10s t=6.00s t=6.80s

Fig. 5. Visualization photograph of the periodic fluctuation of the flow for L/di=1.88 and
Rej=302
Fig. 5 shows photographs of periodic fluctuation in the case where L/di=1.88 and Rej=302.
At t=0.0sec, Vortex-B is clearly visible on the downstream side of Vortex-A, but as time
passes, Vortex-B disappears, as though pushed out to the downstream side, and the flow
becomes parallel. However, Vortex-B appears again. Then this fluctuation is repeated with a
period of 6-7 seconds. The instability of flow was evaluated based on these results, and
Table 1 gives a summary. To the extent that the L spacing grows larger, the flow becomes
unsteady more quickly, and large differences in flow were not observed in the range of high
Rej numbers.
Rej L/di=0.63 L/di=1.25 L/di=1.88 L/di=2.50 L/di=3.13

105 ○ ○ ○ ○ ○
241 ○ ○ ○ ○ ○
272 ○ ○ Δ ○ ○
302 ○ ○ Δ ○ ×
362 ○ ○ Δ Δ ×
407 ○ ○ Δ × ×
483 ○ ○ × × ×
543 ○ ○ × × ×
603 ○ ○ × × ×
713 ○ × × × ×
856 ○ × × × ×
999 × × × × ×
1141 × × × × ×
1712 × × × × ×
1997 × × × × ×
2282 × × × × ×
2853 × × × × ×
5135 × × × × ×
Table 1. Evaluation for the instability of the flow by the visualization
3.3 Wall surface stress

Fig. 6 shows the results of measuring wall surface shear stress. Here, I indicate the region
where the flow is fluctuating periodically, and II indicates the transition zone from laminar
flow to turbulence. At L/di=0.63, the shear stress values can be classified into No.2-5 and No.
6-9. For No. 2-5, the τw value is greatly increased from the low Rej numbers, but the flow is
constricted by Vortex-A at the position of these electrodes and the visualization photo
shown in Fig. 4. Therefore, the rapid flow velocity is maintained. At Rej=1000 or higher, the
increase in τw slows down, and this may be due to the effects of changes in the flow velocity
distribution at the nozzle outlet. There is also a tendency for No. 4 and 5 to match No. 6-9 at
low Rej numbers, and this shows that Vortex-A has still not developed. At No. 6-9 with low
Rej numbers, τw does not increase very much, and then it increases greatly after Rej=1000 is
exceeded. These electrodes are positioned at a location where the flow constricted at Vortex-
A spreads out after passing Vortex-A, and since the cross-sectional area of the flow increases,
the flow velocity near the wall surface slows down. Therefore the τw value is small. The
fact that τ w increases above Re j =1000 may be because the flow becomes turbulent.
101
L/di=0.63 1
100
1.5
τw [N/m ]
2
10-1
No.2
No.3
No.4
No.5
10-2 No.6
No.7
No.8
II No.9
10 -3 //////
101 102 103 104 105
Rej
101
L/di=1.88 1
100
1.5
τw [N/m2]
10-1 No.2
No.3
No.4
No.5
10-2 No.6
No.7
No.8
I II No.9
10-3 1 //// /////
10 102 103 104 105
Rej
101
L/di=3.13 1
100
1.5
τw [N/m2]
10-1 No.2
No.3
No.4
No.5
10-2 No.6
No.7
No.8
II No.9
10-3 /////
101 102 103 104 105
Rej
Fig. 6. Relationship between wall shear stress and Reynolds number
At L/di=0.63, Vortex-B grows large, but at No. 7-9, τw decreases rapidly in the range
Rej=500-900, and thus entry into the Vortex-B region can be confirmed. No. 6 exhibits a
peculiar phenomenon where τw abruptly increases and decreases midway through. This
may be because Vortex-A expanded to the No. 6 position as Rej numbers increased, and then
shrank again as the flow became turbulent. No. 6 is between Vortex-A and Vortex-B.
The results for L/di=1.88 are basically same as for L/di=0.63. The major difference is
behavior relating to Vortex-B. Vortex B at L/di=1.88 fluctuates periodically, and thus does
not remain steady and grow large. Therefore the reducing effect on τw due to Vortex-B does
not appear to the same extent as at L/di=0.63. The phenomenon where τw decreases when
the electrode enters the Vortex-B region is not observed at No. 9, and thus it can be seen that
Vortex-B has not developed as far as No. 9. When L/di=3.13, the flow is more unstable than
L/di=1.88, and thus the slope changes at an even lower Rej number.
3.4 Mass transfer and heat transfer

Fig. 7 shows the results for average mass transfer measured using the overall electrodes No.
(1)-(3). Here, the stagnation point is the value measured at point electrode No. 1. Average mass
transfer measured at No. (1)-(3) is smaller than the value at the stagnation point. This is
because the concentration boundary layer has developed taking the stagnation point as its
base point, and its thickness is extremely small. Therefore, the boundary layer gradually
develops and increases in thickness at No. (1)-(3) on the downstream side of that, and thus the
Sh number becomes smaller. For the same reason, the Sh number gradually decreases moving
from upstream to downstream at No. (1)-(3). With L/di=0.63, Vortex-B grows large at No. (2)
and (3), and thus a large increase in the Sh number is not seen until close to Rej=1000.
The heat transfer coefficients were calculated from these Sh numbers using the analogy
equation for mass transfer and heat transfer in equation (24). Fig. 7 shows the average value
of heat transfer coefficients for each electrode or No. (1)-(3). The solid line is the turbulence
correlation equation (equation (25)) for an annular channel of a dual pipe. In order to
compare with the annular channel of the dual pipe, the Re* numbers in the annular channel
of the dual pipe were used as the Re numbers. All of the measured values are larger than the
correlation equation for the annular channel of a dual pipe, and thus this correlation
equation cannot be used for design of the equipment. Heat transfer increased locally,
particularly in the No. (1) region. At No. (2) and (3), the slope of the heat transfer coefficient
Fig. 7. Average mass transfer coefficient measured in the overall electrode

became equal to the slope of the correlation equation for the annular channel of a dual pipe,
from the region where Re* numbers increased and the flow began to become unsteady. Fig.
8 shows the average values of heat transfer coefficients for No. (1)-(3). As a result, the
average values of heat transfer coefficients form a straight line, as shown by the dotted line
in the figure. The correlation equation is as given below.
Nu = 0.3 ⋅ Re∗0.6 ⋅ Pr 1 3 (28)
Fig. 8. Heat transfer coefficient in each electrode

4. Conclusions
In this research, the aim was to elucidate transfer phenomena in a jet-type cooling pipe. This
was achieved by conducting flow visualization and measurement of mass transfer
coefficients using a cooling pipe with the same shape as that actually used for die cooling,
and predicting heat transfer based on analogy.
The findings of this research are indicated below.
1. Through visualization using aluminium powder, a primary vortex (Vortex-A) and
secondary vortex (Vortex-B) were observed on the downstream side from the
stagnation point of the impingement jet. These vortices vary greatly in size as the Rej
number increases. Also, the secondary vortex causes periodic fluctuations immediately
before the flow starts to turn turbulent.
2. Point electrodes were embedded in the outer pipe of the cooling pipe, and wall surface
shear stress was measured using the electrode reaction method. Those values varied in
accordance with the expansion and shrinkage of the primary and secondary vortices
observed through visualization. Thus, it was found that the primary vortex has the
effect of accelerating flow near the wall surface, and the secondary vortex has the effect
of decelerating flow near the wall surface.
3. The heat transfer coefficients were inferred from Sh numbers by using the analogy
between mass transfer and heat transfer. The heat transfer coefficients showed almost
the same behavior as the flow pattern, and tendencies similar to previous research were
exhibited at the stagnation point. Also, a correlation equation for finding heat transfer
coefficients was obtained from the average value for No. (1)-(3).
5. References
Miranda, J.M. & Campos, J.B.L.M. (1999), Impinging jets conical wall: laminar flow
predictions, AIChE Journal, 45(11), pp.2273-2285, ISSN 1547-5905
Miranda, J.M. & Campos, J.B.L.M.(2001), Impinging jets confined by a conical wall-High
Schmidt mass transfer in laminar flow, International Journal of Heat and Mass
Transfer, 44, pp.1269-1284, ISSN 0017-9310
Kayansayan, N. & Kucuka, S. (2001), Impingement cooling of a semi-cylindrical concave
channel by confined slot-air-jet, Experimental Thermal and Fluid Science, 25, pp.383-
396, ISSN 0894-1777
Hrycak, P. (1981), Heat transfer from a row of impinging jets to concave cylindrical surface,
International Journal of Heat and Mass Transfer, 24, pp.407-419, ISSN 0017-9310
Gau, C. & Chung, C.M. (1991), Surface curvature effect on slot-air-jet impinging cooling flow
and heat transfer process, Transaction of the ASME, Journal of Heat Transfer, 113,
pp.858-864, ISSN 0022-1481
Lee, D.H., Chung, Y.S. & Kim, D.S. (1997), Turbulent flow and heat transfer measurement on
a curved surface with a fully developed round impinging jet, International Journal of
Heat and Fluid Flow, 18, pp.160-169, ISSN 0142-727X
Kobayashi, S. & Iida (1996), K., Idou-ron, ISBN 978-4-254-13518-3 Asakura-shoten
Mizushima, T., Hara, K. & Kyuno, T., Heat and mass transfer coefficients in double tube
cooler condenser, Kagaku Kikai, 16, pp.338-344, 1952.
30
Thermal State and Human

Comfort in Underground Mining
Vidal F. Navarro Torres1 and Raghu N. Singh2
1Centre for Natural Resources and Environment of Technical University of Lisbon
2Nottingham Centre for Geomechanics, of University of Nottingham
1Portugal
2United Kingston
1. Introduction
The human metabolism is accompanied by heat generation, with the body temperature
remaining constant near 36.9°C and in contact with surrounding atmospheric temperature;
people have cooler or hotter sensations.
When people are exposed to a temperature greater than the threshold limits, it causes
physiological effects expressed as follows: loss of interest in people’s activities, taking
frequent rests or breaks, a desire to quickly complete the task, irritability, reduced
concentration and reduction in sensitivity.
A prolonged exposure of people to unfavourable thermal conditions inevitably leads to
increase in body temperature and consequently producing physiological effects that affect
the work efficiency. Figure 1 shows a relationship between work efficiency and effective
temperature and air, wet temperature and air velocity. It may be noted that the prolonged
exposure of a worker to temperature exceeding 42°C may even cause death.
100
Tw
80 Te
Work efficiency (%)
V = 4.06 m/s
60
40
V = 2.03 m/s
V: air velocity
Tw = wet temperature
20
V = 0.51 m/s
0
60 70 80 90 100 110 120
Effective temperature, Te , or Wet temperature, Tw (ºF)
Fig. 1. Effect of temperature on work efficiency (Navarro, 2003, Ramani, 1992)

The temperature of intake air due to its passage through an underground opening gradually
increases due to depth and the length of air travel through underground opening. The main
cause of heat transfer to air flow in underground atmosphere is due to thermal properties of
virgin rock, known as geothermal gradient. Other sources of heat to the air in underground
atmosphere are air auto-compression, diesel emission, explosive detonation, human
metabolism and influx of thermal water.
2. Mathematical model of heat transfer

The total variation of temperature in an underground environment Δttotal can be calculated
by including the variation of temperature from air auto-compression Δta, thermal properties
of rock Δtr, heat emission from diesel equipments Δtd, heat due to breaking of rocks with the
use of explosives Δte, human metabolism Δth and thermal water Δtw as outlined in equation
(1):
Δttotal = Δt a + Δtr + Δtd + Δte + Δth + Δtw (1)
With increasing mining depths, the influence of the thermal properties of the rock mass
becomes more important (Navarro et al, 2008). Based on equation (1), the total underground
atmosphere temperature T2, will be expressed by equation (2), as a function of surface
temperature ts or underground opening initial temperature T1.
T2 = ts + Δttotal or T2 = T1 + Δttotal (2)
2.1 Surface air temperature

It is well known that the surface air temperature varies with the seasons and is subjected to
regional variations according to local weather conditions, so that the temperature variation
is influenced by the ventilation current temperature in underground openings.
Fig. 2. Typical surface air dry temperature in Neves Corvo and San Rafael mines
Figure 2 indicates that average monthly surface temperature in Neves Corvo mine was
maximum 24.5ºC in July, minimum 9.0ºC in January and mean being 15.6ºC and in San Rafel
Thermal State and Human Comfort in Underground Mining 591
mines was maximum temperature was 8.8ºC in November, minimum 4ºC in July and mean
being 6.61ºC. Surface air temperature variation throughout the year can be better illustrated
with monthly average temperature measured in Neves Corvo and San Rafael mines (Figure 2).
Neves Corvo mine is located in Portugal in North Hemisphere at 20º latitude and at altitude
of 800m, but San Rafael mine located in Perú in South latitude at the altitude of about 4500
m. Temperature trend in Neves Corvo mine is similar to the metalliferous mines in Portugal
with maximum variation range of 15 °C and temperature tendency in San Rafael mine is
typical of the South American Andes with maximum variation range of 4.0 °C. Average dry
bulb temperature measured in Neves Corvo mine stopes located at depth between 750 m to
770m, compared with variation of external dry bulb temperature, observed clearly the
influence of outside temperature in underground openings (Figure 3). Therefore, the
maximum temperature on the hottest month will be critical.
Fig. 3. Surface air temperature influencing underground openings air temperature
Fig. 4. Underground temperature variations as a function of surface air temperature

The surface air temperature can influence the temperature of air flow in the atmosphere of
underground openings, since these are more than 7 °C, as in the Neves Corvo mine (Figure 4).
This result indicates that during winter times or in mines located at large altitudes, such as
the South America Andes, the outside temperature has little or no influence on the
temperature of underground openings.
Moreover, as a part of an environmental thermal comfort assessment in deep underground
mines, it is necessary to consider the surface temperature, because this is the initial
temperature, t1, of intake air to underground openings.
For similar conditions of Neves Corvo mine and at 750 m depth, variation in underground
openings temperature, Δts, will be calculate by equation (3), based on surface air
temperature, te.
Δts = 0.301t e − 2.107 (3)
2.2 Heat transfer due to air auto-compression in vertical underground openings

Auto-compression process occurs during the air descent through the underground openings
and due to its own compression. The mathematical model is deduced considering the
equilibrium condition, air properties and the influenced of by vertical forces (Figure 5)
expressed by air equilibrium condition as follows:
t1
Raise
α
L dp h
dh
t2
Fig. 5. Air auto-compression in inclined raise layout
g.dh − dp / ρ a = 0 (4)
Where, g is gravity, dh is depth differential, dp is pressure differential, ρa is air density. By

substituting specific gravity γ and specific volume v in equation (4) the following expression
is obtained:
dh = dp / γ = vdp (5)
k
In adiabatic process p.v = constant, when k is air adiabatic coefficient and differentiating
results in equation (6) as follows:
v.dp + k.p.dp = 0 (6)
Clapeyron equation p.v= R.t2, where R is universal gases constant and t2 is compressed air
temperature, the following differential equation results:
p.dv = R.dt2 − v.dp (7)

Using equations (5), (6) and (7) equation (8) is obtained as follows:
dh + k( R.dt2 − dh ) = 0 (8)
Integrating the equation (8) obtains the following expression (9) where C is constant:
(1 − k )∫ dh + k.R ∫ dt2 = (1 − k )h + k.R.t2 + C = 0 (9)
Rearranging equation (9), the temperature t2 is obtained as follows:
( k − 1)h (10)
t2 = −C
k .R
For initial values h=0 and t=t1, the constant C=0, then C=-t1, the adiabatic equation 9 result
in equation 11 as follows:
( k − 1)h (11)
Δt a = t2 − t1 =
k .R
With numerical values of constant of perfect gases (R=29.27 kgf-m/kg-ºK) and average air
adiabatic index (1.302) the final equation is obtained as follows:
t2 − t1 = 0.0098h (12)
In general condition depth h will be expressed as a function of underground opening length

L (m) and inclination α (º), as h=LSinα and finally, the temperature increase due air auto-
compression Δta (ºC) results in following equation (Navarro Torres, 2003):
Δt a = t2 − t1 = 0.0098L sin α (13)
That means, when α=90º (vertical raise) for each 100 m air temperature increases by in
0.98ºC, for 200 m 1.96ºC, for 300 m 2.94ºC, for 400 m 3.92ºC, for 500 m 4.9ºC, for 650 m
6.37ºC, for 800 m 7.84 ºC and for 1000 m 9.8ºC. Therefore when α=0º (horizontal
underground opening) auto-compression temperature is zero (Figure 6)
Fig. 6. Variation of auto-compression temperature with raise inclination

2.3 Heat transfer of thermal properties of rock mass to underground atmosphere

At a certain depth hn from the surface defined as the thermal neutral zone (15 m according
to Ramani 1992; 20 to 40 m indicated by Vutukuri & Lama, 1986) the temperature of rock
masses varies during the year as a function of the changes of surface air temperature. The
temperature of any rock mass at depth thr and underground atmosphere air temperature
variation Δtr can be calculated by the following equations:
( h − hn )
thr = tn + (14)
gg
h1 − hn ± L sin α
Δt g = (15)
gg
where; thr is the rock temperature at depth h (ºC), tn is the temperature of the rock mass
above the thermal neutral zone (ºC), h depth of mining excavation below the surface, hn is
the depth of the thermal neutral zone (m) and gg geothermal gradient of the rock mass
(m/ºC).
In order to obtain the mathematical model for the calculation of heat transfer of thermal
properties of rock mass, use of the heat transfer formulation of gas flow in pipes can be
applied to underground openings.
Heat spreads from one point to another one in three distinct ways: conduction, radiation
and convection. In most cases, the three processes occur simultaneously and therefore the
amount of heat “q” supplied to a body of mass “m” and specific heat Ce, when the
temperature increases from t1 to t2 is given by the general equation (16):
q = m.C e (t2 − t1 ) = m.C e .Δt (16)
For the air flowing in the underground openings this equation can be expressed in function
of the circulating air volume Q through:
qr = 1000 ρ a .C e.Q.Δtr = 1000.ρ a .C e .Q.(t2 − t1 ) (17)
Where qr is the heat received by the air from the rock mass (W), ρa the air density (kg/m3), Ce
the specific heat of air (kJ/m.ºC), Q the flow of air (m3/s) and Δtr the variation of
temperature from t1 to t2 (Fig. 7). The heat coming out of the rock mass and received by the
ventilation air in the underground environment can also be expressed in terms of coefficient
of heat transfer λ (Holman, 1983) according to the equation (17):
dq = λ .P.dx( Pp − Tm ) (18)
Where Tp and Tm are the temperatures of rock wall and air mixture in the particular position
x (ºC), λ is the coefficient of heat transfer between the rock mass and the air mixture
(W/m2.ºC) and P is the perimeter of the section of the underground opening (m). The total
heat qr transferred (W) can be calculated by using equation (19) as follows:
qr = λ .P.L.(Tp − Tm )average (19)

qr L
hn t2
Q
Thermal neutral zone
h1 A
B
dx 2
Undergorund h1, t1 α
opening
1
Fig. 7. Layer of rock influenced by external temperature and elementary parameters of an
underground opening
Using equation (19) the average temperature of the rock mass may be given by equation (20)
and (21):
1 ⎧⎪ ⎛ h1 − hn ± L.sin α ⎞ ⎫⎪
Tp = ⎨t1 + ⎜⎜ t1 + ⎟⎬
⎟⎪
(20)
2⎪ ⎝ gg ⎠⎭
⎩
t1 + t 2
Tm = (21)
2
By substituting equations (20) and (21) in equations (17) and (19) the following expression is
obtained:
⎡ λ .P.L ⎤ ⎡ h1 − hn ± L.sin α ⎤
⎢ 2 ⎥⎢ + t1 − t2 ⎥ = 1000.ρ a .C e .Q. ( t2 − t1 )
⎣ ⎦ ⎢⎣ gg ⎥⎦
Finally the variation of temperature from t1 to t2 (Δtr) may be expressed as follows:
λ .P.L.( h1 − hn ± L.sin α )
Δtr = t2 − t1 = (22)
g g (λ .P.L + 2000.ρ a .C e .Q )
Resulting equation (22) is an innovative mathematical model developed for heat transfer of
thermal properties of rock mass to underground openings (Navarro, 2003).
In raises or in any vertical underground openings, h1 = 0, and the length which influences
the temperature due to geothermal gradient is L Sinα - hn and α =90º, thus, resulting in the
following equation:
λ .P.(L − hn )2
Δtr = t2 − t1 = (23)
g g [ λ .P.( L − hn ) + 2000.ρ a .C e .Q ]
The coefficient of heat transfer λ is calculated as a function of the thermal conductivity K

(W/mºC) which is the non-dimensional coefficient of Dittus and Boelter Ndb and the
diameter of section d (m); for horizontal and inclined underground openings d = (B + A)/2,
where B is the width of the section (m) and A its height (m):
k.N db (24)
λ=
d
The relation of Dittus and Boelter co-efficient Ndb. (Holman, 1983) was studied in detail by
Petukohov for gases (air) that derived the following equation:
f
Re d .Pr
N db = 8 (25)
f
1.07 + 12.7( )0.5 (Pr 0.67 − 1)
8
Where Red is the Reynolds number (non-dimensional), given by:
V .d (26)
Red =
μ
in which V is the average velocity of air (m/s), d the underground opening diameter (m)
and μ the kinematic viscosity of air (Kg/m.s). In addition, f is the friction coefficient of the
underground opening walls (Kg/m3), Pr is the Prandtl number (non-dimensional) calculated
by:
ρ a .C e .μ (27)
Pr =
K
Air properties at atmospheric pressure will be determined based in temperatures (Table 1).
2.4 Heat transfer from diesel equipment

The equipments used in underground work generate the heat transfer to the ventilation
current in underground atmosphere as follows:
1. Mobile diesel and electrical equipments, such as jumbo drills, trucks, LHDs, pumps,
locomotives, etc.
2. Electrical and non-mobile equipments (fans, lighting, pumps, hoists, stations or
transformer substations, etc.).
For the mobile and non-mobile equipments used in underground work, diesel equipments
contributes significantly to heat transfer to the air flow in underground atmosphere. Diesel
engines fuel consumption for mining equipment is 0.24 kg/kWh, with a calorific value of 44
MJ/kg (Vutukuri & Lama, 1986), so the total energy released is 0.24x 44x103 KJ/kWh =
10560 kJ/kWh = 176 kJ/mink = 2.9 kJ/s.KW = 2.9 kW/kW. Of the total 1kW energy release,
(34%) is converted into mechanical energy and 1.9 kW (66%) is exhausted to air flow of
underground atmosphere. This energy is not totally transferred to the air flow, because it
depends to the effective time for which the equipment used, so it is different for each
condition of underground work and the value is around 0.9 kW (31%).
Diesel equipment heat exhaust qed (KW) can be expressed by equation (28) as follows:
q ed = f m . f t .qd .Pd (28)
Where qd is the equivalent energy released by diesel fuel (2.9 kW/kW),

Pd is the equipment engine (kW),
fm is mechanical efficiency and
ft is equipment utilization efficiency.
Ce ν(kg/m.s) μ(m2/s)x Dif.Térm.

T (ºK) ρ a(kg/m3) K(W/m.ºC) Pr
(KJ/kg.ºC) x10-5 10-6 (m2/s)x10-4
100 3.6010 1.0266 0.6924 1.923 0.009246 0.02501 0.770
150 2.3675 1.0099 1.0283 4.343 0.013735 0.05745 0.753
200 1.7684 1.0061 1.3289 7.490 0.01809 0.10165 0.739
250 1.4128 1.0053 1.488 9.49 0.02227 0.13161 0.722
300 1.1774 1.0057 1.983 16.84 0.02624 0.22160 0.708
350 0.9980 1.0090 2.075 20.76 0.03003 0.2983 0.697
400 0.8826 1.0140 2.286 25.90 0.03365 0.3760 0.689
450 0.7833 1.0207 2.484 31.71 0.03707 0.4222 0.683
500 0.7048 1.0295 2.671 37.90 0.04038 0.5564 0.680
550 0.6423 1.0392 2.848 44.34 0.04360 0.6532 0.680
600 0.5879 1.0551 3.018 51.34 0.04659 0.7512 0.680
650 0.5430 1.0635 3.177 58.51 0.04953 0.8578 0.682
700 0.5030 1.0752 3.332 66.25 0.05230 0.9672 0.684
750 0.4709 1.0856 3.481 73.91 0.05509 1.0774 0.686
800 0.4405 1.0978 3.625 82.29 0.05779 1.1951 0.689
850 0.4149 1.1095 3.765 90.75 0.06028 1.3097 0.692
900 0.3925 1.1212 3.899 99.30 0.06279 1.4271 0.696
950 0.3716 1.1321 4.023 108.2 0.06525 1.5510 0.699
1000 0.3524 1.1417 4.152 117.8 0.06752 1.6779 0.702
1100 0.3204 1.160 4.44 138.2 0.0732 1.969 0.704
1200 0.2947 1.179 4.69 159.1 0.0782 2.251 0.707
1300 0.2707 1.197 4.93 182.1 0.0837 2.583 0.705
1400 0.2515 1.214 5.17 205.5 0.0891 2.920 0.705
1500 0.2355 1.230 5.40 229.1 0.0946 3.262 0.705
1600 0.2211 1.248 5.63 254.5 0.1000 3.609 0.705
1700 0.2082 1.267 5.85 280.5 0.105 3.977 0.705
1800 0.1970 1.287 6.07 308.1 0.111 4.379 0.704
1900 0.1858 1.309 6.29 338.5 0.117 4.811 0.704
2000 0.1762 1.338 6.50 369.0 0.124 5.260 0.702
2100 0.1682 1.372 6.72 399.6 0.131 5.715 0.700
2200 0.1602 1.419 6.93 432.6 0.139 6.120 0.707
2300 0.1538 1.482 7.14 464.0 0.149 6.540 0.710
2400 0.1458 1.574 7.35 504.0 0.161 7.020 0.718
2500 0.1394 1.688 7.57 543.5 0.175 7.441 0.730
Table 1. Air properties at atmospheric pressure (Holman, 1983, Navarro, 2003)
Based on equation (28), the temperature variation of air due to exhaust from the diesel
equipment Δtd (ºC) can be quantified by the following equation:
f m . f t .qd .pd
Δtd = (29)
ρ a .C e .Q
It may be noted that the exhaust heat from the diesel engines to the underground
atmosphere is from the local equipment use only.
2.5 Heat transfer from explosive blasting

The blasting process of explosive in underground environment generates heat that is
transferred to the surrounding rock mass and to the ventilation current of the underground
atmosphere.
Heat released by blasting qe (kW) can be calculated by equation (30), based on calorific
energy of explosive Ee (kJ/kg), and explosive quantity daily used qe (Kg/day). For example,
the calorific energy of ANFO is 3900 kJ/kg and the dynamite 60% varying between 4030 to
4650 kJ/kg.
Ee .q e
qe = (30)
86400
The thermal influence due to blasting Δte (ºC) can be quantified by equation (31) as follows:
Ee .q e (31)
Δt e =
86400.ρ a .C e .Q
Similar to diesel exhaust heat, the heat due to explosive detonations influences the local
atmosphere only.
2.6 Heat transfer due to human metabolism

The heat transfer of human metabolism in not significant and can be ignored (Hartman et
al., 1997), for example 800 workers in normal working conditions leads to a total release of
192 kW (65000 BTU/hr), energy corresponding to each worker being 0.25 kW.
Thus, when the number of people or workers in an underground environment is large,
temperature increase by human metabolism Δth (ºC) can be expressed by equation (32),
where qh is the human heat release and it is a function of effective temperature (kW/person)
and, n is the total number of human involved.
qh .n
Δth = (32)
ρ a .C e .Q
2.7 Heat transfer from underground water

Two sources of water are encountered in mining: Groundwater or Mine water. All ground
water, especially from hot fissures and natural rock reservoirs, is a prolific source of heat in
mine workings. Since water and heat are both derived from the surrounding rock or
geothermal sources, the water temperature will approach or even exceed the rock
temperature.
The water transfers its heat to the mine air, mainly by evaporation increasing the latent heat
of the air.
The total heat gain from hot underground water in open channel flow qw (kW) can be
calculated from the equation (33):
q w = Ftw .c w (ttw − t a ) (33)
Where Ftw is weight flow rate of thermal water (kg/s)

Cw is specific heat of water (4.187 kJ/kg°C), and
ttw and ta are water temperatures at points of emission and exit from the mine
airway in (°C), respectively.
The thermal influence of underground ventilation air flow can be calculated by equation
(34) as follows:
Ftw . ( ttw − t a )
Δtw = 4.187 (34)
ρ a .C e .Q
3. Case studies of heat transfer in underground mining

3.1 Case study in Portuguese Neves Corvo mine
Vertical underground opening
The Neves Corvo mine is an operating underground copper and zinc mine in the western
part of the Iberian Pyrite Belt which stretches through southern Spain into Portugal. The
mine uses both bench and fill and drift and fills underground stoping methods. The copper
plant has treated a maximum of 2.0 mt per annum of ore and in 2007 it was upgraded to
treat up to 2.2 mt of ore per annum.
Fig. 8. Map of the Neves Corvo area showing the massive sulphide ore bodies Neves, Corvo,
Gracia and Zambjal, main faults and the exploratory boreholes (Moura, 2005)
The ore bodies of the underground Neves Corvo copper mine (Fig. 8) were formed in a
volcanic sedimentary submarine environment possibly linked with an intercontinental rift
and, third order pull apart basins, not far from the collision zone and located in geological
formations between Volcanic Lavas (V1) and Volcanic Sediments (V2). The V1 is composed
of black shale/schist and has same silicfication but generally less than V2 volcanic. The V2
has a compact vitreous due the high quantity of silica (Riolitic) showing schistosity and
alteration from Chlorite (Lobato, 2000).Mining areas are located between +200m and -450m,
and they are referred to 0 level, equivalent to 0 m datum and transport level equivalent to
-550m level (Fig. 9). The total length of underground vertical shafts, inclined and horizontal
openings is about 80 kilometres.
CPV4 CPV3 CPV1 CPV8 CPV2 CPV6 CPV5
-3
NEVES COLECTOR
GRAÇA COLECTOR
COLECTOR 850 NEVES Ore body

GRAÇA
GRAÇA RAMP
Ore body
NEVES
SANTA CASTRO RAMP
BÁRBARA RAMP 720 COLECTOR
SHAFT
Transport Level 700
Comminution Level 700 Transport Neves
LEGEND CORVO
Ore body
Ventilation shaft CPV
Ore pass CPM

Use diverse galleries Transport level 550
Use diverse ramps
Comminution
Use diverse inclined openings
Level 550
Fig. 9. Neves Corvo underground mine cross section (Navarro, 2003)

The air temperature in underground stopes of Neves Corvo mine is moderate averaging
between 20ºC to 33ºC and in isolated areas in critical condition reaching 42ºC
For applying the mathematical model a vertical underground opening the CPV3 shaft
shown in Figure 10, was selected (Navarro Torres, et al, 2008). This shaft was constructed
using a raise boring machine from the depth of 1222.40m level to 973.64m level with a length
of 248.76m and a diameter of 4.2 m (perimeter 13.19 m and 13.85 m2 in cross section area).
The wall friction factor corresponded to 0.0362 kg/m3 with an average air velocity measured
in July 2000 of 11.84 m/s and average exterior temperature of 24.61 ºC (Figure 2 and Figure 3).
Average airflow resulted in 164.03 m3/s calculated based on air velocity measured, as
indicated in Figure 11.
Air intake
Shaft
CPV3
Depth
Fig. 10. Scheme and photograph of shaft CPV3
28
27
26
25
24
23
22
21
Dry temprature - Air velocity
20
19
18
17
16 Dry temperature (ºC)
15 Air velocity (m/s)
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
4
9
4
4
4
9
:4
:4
:5
:5
:0
:0
:1
:1
:2
:0
:0
:1
:1
:2
:2
:3
:3
:2
:3
:3
:4
:4
:5
:5
:0
:0
:1
:1
:2
:2
:3
:3
:4
:4
:5
:5
15
14
14
14
14
14
14
14
14
14
14
14
15
15
15
15
12
12
12
12
12
12
12
13
13
13
13
13
13
13
13
13
13
13
13
14
Fig. 11. Intake dry temperature and air velocity measured by Data LOGGER DL20K
The air temperature in CPV3 shaft is not influenced by the temperature rise due to diesel
equipments (Δtd), explosives (Δte), thermal water (Δtw) and human metabolism (Δth).
Therefore, only auto-compression (Δta) and geo-thermal properties of rock (Δtr) were
considered for the validation of the proposed model.
The physio-chemical properties of air shown in Table 2 extracted from Table 1 for 24.4ºC
enabled the calculation of the Prandtl number Pr, Reynolds number Red, related to Dittus
and Boelter number Ndb and coefficient of heat transfer λ, by applying equations (27), (26),
(25) and (24), respectively as shown in Table 3.
ρ a (kg/m3) Ce (kJ/kg.ºC) μ (m2/s) K(W/m.ºC)

1.1888 1.0056 16.48 x10-6 0.026
Table 2. Physio-chemical air properties at 24.4 ºC (Navarro, 2008).
Pr Red Nbd λ (W/ m2.ºC)

0.709 3.02 x 106 5162.02 76.106
Table 3. Coefficient of heat transfer and previous values calculated
Finally using the geothermal gradient as 30.3 m/ºC for the rock mass (gg) for the Neves
Corvo mine (Fernández-Rubio. et al., 1990) and using 30.0m as the depth of thermal neutral
zone in the developed mathematical model in equation (23), the temperature rise of rock
mass (Δtr = t2 – t1) is calculated as 2.65ºC. Applying these values to equation (13), the
temperature increase due to air auto-compression can be obtained as 2.38ºC. Then the total
increase of the air temperature during the air flow in the shaft CVP3 results in 5.03ºC (Fig.
12), calculated by following simplified equation:
0.033(L − 30)2 (34)

Δt a + Δtr = 0.0098.L +
2.38.Q + L − 30
Obviously, when airflow is decreased, the total temperature increment (auto-compression +

geothermal properties of rock + temperature due to depth increase) significantly raises the
total temperature increment (Fig. 12).
By applying equation (2) to equation (34), the underground atmosphere’s temperature of
CPV3 shaft T2 in Neves Corvo mine as a function of airflow quantity Q is represented by
equation (35) given below and illustrated in Figure 13 for the average surface temperature
15.65oC (maximum 24.5ºC, minimum 9ºC and mean 15.65ºC).
1579.25
T2 = ts + 2.44 + (35)
2.38Q + 248.76
Underground atmosphere temperature ºC
35
30 Shaft CPV3
100 m
25
200 m
20 248.76 m
15 300 m
10 400 m
500 m
5
600 m
0 700 m
50 75 100 125 150 175 200 225 250
Airflow, m 3/s
Fig. 12. Total increases in temperature due to auto-compression and geothermal properties
of rock as a function of shaft depth and airflow in shaft CPV3
35
30
Measure airflow quantity
25
20
15
10
ts maximum 24.5 ºC
5 ts medium 15.65 ºC
ts minimum 9 ºC
0
50 75 100 125 150 175 200 225 250
Airflow quantity, m 3/s
Fig. 13. Underground atmosphere temperature influenced by auto-compression and thermal

properties of rock and airflow quantity in CPV3 shaft
It may be observed that a slight increase in the underground atmospheric temperature,

results in slight decrease in the air flow. The average values measurement with Data
LOGGER DL20K of ROTRONIC in the air shaft intake with a thermo/hygrometer Casella in
the shaft (Fig. 14) was 29.52ºC in the shaft bottom and 24.61ºC in the intake, therefore the
difference is 4.91ºC.The comparison the results show a total variation of temperature (Δttotal)
between the mathematical model and measured temperature in CPV3 shaft is only 0.12ºC.
Fig. 14. Measurements with Data LOGGER DL20K and thermo/hygrometer Casel
3.2 Case study in Peruvian San Rafael tin mine

Sub-horizontal underground opening
The San Rafael mine belongs to the Peruvian company MINSUR S.A. and is located
Southwest of the San Bartolomé de Quenamari mountain (altitude 5299 m), in the
Department of Puno in the Eastern Mountains of Southern Peru. It is geographically located
in the coordinates of 70°19' longitude West and 14°14' latitude south. This mine is the only
producer of tin in Peru and ore production is 2500 tons per day, with 5.23% of tin (Sn).
Geology of San Rafael mine involves silts and quartzite rocks of the tertiary Sandia
formation with the intrusion of two granites. In the neighborhood there are rocks of the
superior Paleozoic age. In the Sandia formation silts have dark gray colors with muscovite
in the cleavage plans and the quartzites are intercalated with silts (Fig. 15).
The mineralized veins and ore bodies are located in the intrusive ore body of San Rafael
along the NE – SW direction, with a length of 800 m to 1000 m, a width of 300 m and depth
of up to 2000 m. These ore bodies have widths of 4 m to 30 m, lengths of 30 m to 180 m and
heights of 60 m to 610 m, and in general have prismatic forms. The main existing minerals
are cassiterite, stanhite and chalcopyrite.
The main access from surface is through 4523m ramp that communicates to 3825m level,
constituting the principal infrastructure of underground transport, as well as the ventilation
circuit (Fig. 16).
Figure 17 presents the clean air temperatures with normal trend until the level 3950 m
(17ºC), but in 3850m level, where a variation is only 100 m, temperature increases to 34ºC,
thus showing the effect of thermal water. A forecast for air temperature in the 3850m level
without the influence of hot water leads to 20ºC for the air flow of 8.11m3/s.
Fig. 15. Geological section with zones of mining in San Rafael along the direction N 70º E
1
2
3
4
5
6
Fig. 16. General project of the underground environment of San Rafael mine (Navarro, 2001)
40
Trend influenced by thermal water located in 3850
35 Trend without thermal water influenced
30
Temperature (ºC)
25
20
15
10
0
3800 3900 4000 4100 4200 4300 4400 4500 4600
Fig. 17. Air temperature trends with and without the influence of thermal water (Navarro,
2003)
The thermal water temperature measured in the flowing water channel was 40ºC and the
influence of air temperature in underground opening was up to 34 ºC (15 m2 section). It may
be observed that an increase in temperature by 12 ºC is not a normal trend of air
temperatures, because it represents an increase of about 60%.
In San Rafael mine case study, the calculated geothermal gradient gg based on equations (1)
and (22) and measured data and equation (22) needed recalculation of temperatures
variation due to diesel exhaust, explosive detonations and thermal water, because in local
level 3850, the temperature increment due to auto-compression and human metabolism are
insignificant.
Diesel equipments heat transfer Δtd, is calculated by applying equation (29) based on
combined factors of mechanical efficiency and equipment utilization efficiency (fm.ft) of 0.005
for diesel equipments used in ramp 4523(level 3850) (Table 4), air density 1.2661 kg/m3 and
specific heat of air 1.0056 kJ/kg.ºC (Table 1) and for depth local level 3850 of ramp 4523,
temperature increase result is 0.85ºC.
Heat transfer Δte due to explosive detonation is determined by applying equation (31), based
on an average of 120 kg per day ANFO, air density and air specific heat (Table 1), resulting
in temperature rise of 0.52ºC.
The thermal water heat transfer calculated by applying equation (34), based on measured
flow rate of 4.93 l/s of thermal water in channel Ftw, was 40ºC water temperatures at points
of emission ttw and 34ºC at the exit from the mine airway ta; using air density and air
specific heat (Table 1), result being 12ºC.
Using these results and applying equation (1) based in measured 22ºC (34ºC-12ºC) total
temperature increase Δttotal the heat transfer of virgin rock Δtr results in 8.63ºC.
Finally, for the following conditions d = 4.5m, f = 0.0046kg/m3, V = 0.39m/s, P = 18m,
L = 7000m, h1 = 30m, hn = 30m, α = 7º, Q = 8.11m3/s (branch 6-5 Figure 15) and for physical-
chemical air conditions (Table 5) and calculated Prandtl number Pr, Reynolds number Red,
relation of Dittus and Boelter Ndb and coefficient of heat transfer λ, is calculated applying
equations (27), (26), (25) and (24), respectively as shown in Table 6, and applying equation
(22) the geothermal gradient result in 59.51m/ºC or 1.68 ºC for each 100 m.
Mining operations Equipments

Development and . 2 Jumbo Boomer H 282 of Atlas Copco, with 75 HP (55.93 KW)
prospection . 2 LHD de 5.5. Yd3 EJC, with 186.43 KW each.
. 1 Simba H-1354 de Atlas Copco, Cop – 1838, with 80 HP (59.66 KW)
Drilling long holes . 1 DTH Tunnel 60, Drillco Toolls, Topo 3
. 1 DTH Mustang A32 de Atlas Copco, with Drill Cop – 34
. 2 LHD de 6.5 yd3 ST100 Wagner, with 250 HP (186.43 KW) each.
Mucking ins stopes
. 1 LHD de 3.5 Yd3 Wagner (stand by), com 185 HP (137.9 KW)
Fragmentation . 4 hydraulic drills Kent
Ore transport by ramp . 6 Trucks Volvo NL12, de 15 m3, with 410 HP (305.73 KW) each.
Supervision . 27 Cars, with 89 HP (66.37 KW) each.
Table 4. Underground diesel equipment used in San Rafael mine
ρ a (kg/m3) Ce (kJ/kg.ºC) μ (m2/s) K(W/m.ºC)

1.26614 1.0056 14.07 x10-6 0.0248056
Table 5. Physical-chemical air properties
Pr Red Nbd λ (W/ m2.ºC)

0.710 0.125 x 106 45.08 0.248
Table 6. Coefficient of heat transfer and previous values calculated
Geothermal gradient trend of San Rafael mine based on calculated results as compared with
worldwide mining district trends is shown in Fig. 18.
80
Temperature virgin rock mass (ºC)
60
40
20
0
Depth (Km)
Fig. 18. Geothermal gradient of San Rafael mine compared of some worldwide mining
districts (compiled by Navarro, 2001)
In Rafael mine, the intake clean air bay drifts 4666, 4600 and ramp 4523, compared with the
mine deepest level at 3835 m, there is an 831 m level difference where temperature raises
21.5ºC, or about 1ºC for each 40m depth. This result indicates a variation of annual average
external temperature of 6.61 ºC and at level 3850 would reach 16.70 ºC (Fig. 19), which is
compatible with the trend of temperatures measured in ramp 4523. The thermal
conductivity of the rock mass of the San Rafael mine is estimated as 3.25 W/mºC.
20
18
16
14
Temperature (ºC)
12
10
4 Geothermal degree determined

Ramp 4523 air temperature measured
2 Linear (Ramp 4523 air temperature measured)
0
3800 3900 4000 4100 4200 4300 4400 4500 4600
Level (m)
Fig. 19. Geothermal gradient determined and air temperature measured Navarro, 2003
Thermal human comfort assessment in underground openings based on International
Standards Organization (ISO 7730) and American Society of Heating, Refrigerating and Air-
Conditioning Engineers (ASHRAE/55) standards, and the thermal human comfort defined
based in operative ranges from 22ºC to 28ºC as shown in Figure 20.
Fig. 20. ISO 7730 and ASHRAE/55 human thermal comfort standards
By applying equations (1) and (2), using the annual average external temperature as 6.61 ºC
(Fig. 2), without the influence of thermal water and the physio-chemical properties of air
conditions (Tables 5 and 6) the thermal condition of 4523m ramp in San Rafael mine was
assessed by using equation (36)..
0.544L2 11.745
Δttotal = + (36)
265.653L + 151539.88Q Q
In San Rafael mine without thermal water heat transfer, the temperature is lower than
thermal comfort standards (ISO 7730 and ASHRAE/55). Figure 21 illustrates the
underground opening temperatures for various air flow rates and underground opening
lengths. It can be observed that when airflow increases, the underground opening
temperature decreases, and on the other hand, when underground opening length increases,
the change in underground air temperature is moderate.
Fig. 21. Temperature total increment influenced by airflow quantity and underground
openings length in San Rafael mine
For thermal human comfort assessments in underground openings it is necessary to
determine the local thermal situation and that obtain increasing for the total temperature
increment Δttotal the initial temperature in the underground opening branch (t1). In San
Rafael mine measured average minimum temperature in July was 4°C and the maximum
temperature in February was 7.8ºC and the average temperature was 6.16ºC a shown in
Figure 2.
Underground atmosphere temperature in ramp 4523 (Level 385 m), influenced by thermal
water as a function of initial temperature in local branch t1, local length (349 m) and airflow
quantity Q, will be expressed by particular equation (37) as shown in Figure 22.
1 109.04
t 2 = t1 + + (37)
0.0868 + 0.142Q Q
In the locality of ramp 4523 (level 3835) where the average intake air temperature is 16.7ºC,
the temperature is influenced by the heat transfer due geothermal properties of rock, diesel
exhaust, explosive detonation and thermal groundwater and the human thermal comfort
will be obtained for airflow quantity between 6 and 8 m3/s (Fig. 20). When the airflow
quantity are smaller than 6 m3/s the underground atmosphere temperature increases
greatly and for the flow rates higher than 8 m3/s the surface temperature decreases
gradually.
50
45 Minimum 4 ºC
40 Medium 6.61 ºC
35 Maximum 7.8 ºC
30 Thermal comfort upper limit
25
20 Thermal comfort lower limit
15
10
5
0
0 10 20 30 40 50 60 70 80 90 100 110 120
Airflow quantity, m /s
Fig. 22. Human thermal comfort situation in ramp 4523m (level 3835) of San Rafael mine
It may be noted that without the presence of thermal water there is no risk of environmental
thermal discomfort, but in localities with thermal water there appears to be risk of thermal
discomfort.
3. Acknowledgements
Thanks are due to Dr A. S. Atkins, Faculty of Computing, Engineering and Technology,
Staffordshire University for his valuable suggestions to improve the quality of this book
chapter.
4. References
Fernández-Rubio, R., Carvalho P. and Real F. (1988). Mining-Hydrological characteristics of
the Underground Copper Mine of Neves-Corvo, Portugal. Third International Mine
Water Congress. Melbourne, Australia, pp. 49-63
Hartman H. L., Mutmansky J. M., Ramani R. V. and Wang Y. J. (1997). Mine Ventilation and
Air Conditioning, (3d ed), John Wiley & Sons, Inc, New York. USA, pp.730
Holman, J. P. ( 1983). Transferência de calor. McGraw Hill, São Paulo Brasil, pp. 194–203
Moura, A. (2005). Fluids from the Neves Corvo massive sulphide ores, Iberian Pyrite Belt,
Portugal. Chemical Geology. Volume 223, Issues 1-3, 22 November 2005, 153–169
Lobato, J. C. (2000). Evaluation of mining induced stress on support requirements at Neves Corvo,
MSc Thesis University of Exeter, UK, 21–31
Navarro Torres V. F. (2001). Thermal, dynamic and volumetric impact assessment of
underground environment of Peruvian San Rafael mine, Technical Report,
Geotechnical Center of Technical University of Lisbon, pp. 47
Navarro Torres V. F. (2003). Environmental Underground Engineering and Applications of
Portuguese and Peruvian Mines, Doctoral Thesis, Technical University of Lisbon, pp.
85-225
Navarro Torres V. F., Dinis da Gama C., Singh R. N. (2008). Mathematical modelling of
thermal state in underground mining, Acta Geodyn. Geomater., Vol. 5, No. 4 (152),
pp. 341–349
Ramani R. V. (1992). Personnel Health and Safety, Chapter 11.1 SME Mining Engineering
Hand Book, 2nd Edition Volume 1, H. L. Hartman Senior Editor, pp. 995 -1039
Vutukuri V. S. & Lama R. D. (1986). Environmental Engineering in Mines, Cambridge
University Press, 0521157390, Great Britain, 518 pp. 518
31
Heat Transfer in the Environment:

Development and Use of Fiber-Optic
Distributed Temperature Sensing
Francisco Suárez1, Mark B. Hausner1,
Jeff Dozier2, John S. Selker3 and Scott W. Tyler1
1University of Nevada, Reno,
2University of California, Santa Barbara,
3Oregon State University,
United States
1. Introduction
In the environment, heat transfer mechanisms are combined in a variety of complex ways.
Solar radiation warms the atmosphere, the oceans, and the earth’s surface, driving weather
and climate (Lean & Rind, 1998). Clouds and aerosols reflect a fraction of the incoming solar
radiation and partially absorb the infrared radiation that comes from the earth’s surface,
allowing the existence of acceptable temperatures for the biota and human survival
(Norand, 1920; Moya-Laraño, 2010). In water bodies, absorption and scattering of solar
radiation results in stratification of the water column (Branco & Torgersen, 2009). Cooling
conditions, e.g., convective night-time cooling, at the water surface can destroy the
stratification and thus, mix the water column (Henderson-Sellers, 1984). In open water
bodies solar radiation also induces evaporation: as water changes its phase, heat is
transferred from the water body into the atmosphere by the release of latent heat (Brutsaert,
1982). Within the earth, temperature increases with depth. The temperature at the earth’s
center is estimated to be on the order of 6000 °C (Alfe et al., 2002). An average geothermal
gradient of 25-30 °C km-1 (Fridleifsson et al., 2008) indicates that approximately 40 TW (4 ×
1013 W) flow from the earth’s interior to its surface (Sclater et al., 1981). Much of this heat is
the result of radioactive decay of potassium, uranium, and thorium (Lee et al., 2009). In the
shallow subsurface, this geothermal gradient can be disturbed by groundwater flow and
atmospheric conditions (Uchida et al., 2003; Bense & Kooi, 2004).
By measuring the temperature in the environment, it is possible to elucidate the main heat
transfer mechanisms controlling different environmental, ecological, geological or
engineering processes. Many of these processes span spatial scales from millimeters to
kilometers. This extreme range of spatial scaling has been a barrier limiting observation,
description, and modeling of these processes. In the past, temperature measurements have
been performed at small scales (spanning millimeters, centimeters, or a few meters) or at
large scales (spanning tens of meters or kilometers (Alpers et al., 2004)). However, for spatial
scales between these two disparate scales and in a variety of media, there is a lack of
methods that can accurately estimate temperatures. Fiber-optic distributed temperature

sensing (DTS) is an approach available to provide coverage in both space and time that can
be used continuously to monitor real-time data in different environments and at spatial
scales that range from centimeters to kilometers. DTS was first developed in the mid 1980s
(Dakin et al., 1985) and used in the oil and gas industry during the 1990s and early 2000s
(Kersey, 2000), but only since the middle of this decade have DTS instruments achieved
acceptable levels of spatial and temporal resolution, along with the high temperature
accuracy and resolution needed to observe environmental processes. With their spatial and
temporal coverage, DTS methods offer significant advantages over traditional measurement
systems in the environment. For instance, the fiber-optic cables that serve as the temperature
probe are low cost, with no issues of bias or fluid column disturbance, and variability due to
different sensors and sensor measurement scale can be avoided.
The objective of this chapter is to provide the reader with an overview of the theory of fiber-
optic DTS technology and a review of environmental applications to date, which will be
used to investigate the main heat transfer mechanisms occurring in different environments.
Important considerations, recent advances, and future trends are also discussed.
2. Fiber-optic distributed temperature sensing theory

Fiber-optic DTS technology uses Raman spectra scattering in an optical fiber to measure
temperature along its length, i.e., the fiber-optic cable is the thermometer, achieving
temperature resolutions as small as ±0.01 °C, and spatial and temporal resolutions of 1-2 m
and 1-60 s, respectively, for cables up to 10 km (Selker et al., 2006a). Raman scattering can
also be used to estimate temperatures in media other than optical fibers, such as
atmospheric LIDAR (Eichinger et al., 1993), but can only be used to measure atmospheric
temperatures and are far less widely used. To understand how fiber-optic DTS systems
work, we first present the background of Raman scattering, describing how this scattering
can be used to determine temperatures along the length of the optical fiber. Then, we
present the governing equations that DTS systems use.
To measure the temperature along an optical fiber, the DTS instrument emits laser pulses at
a known wavelength into an optical fiber. An optical fiber consists of a glass core
surrounded by a glass cladding with a different refractive index than the core (Fig. 1). As
light travels longitudinally along the fiber, a fraction of the incident light is scattered by
interactions between the light and the crystalline structure, and vibration frequency
(temperature) of the fiber itself (Hausner, 2010). Light scattering is classified as elastic or
inelastic. Elastic (or Rayleigh) scattering occurs when the kinetic energy of the incident
photons is conserved and thus, the frequency of the scattered photons is equal than that of
the incident light. On the other hand, when the kinetic energy of the incident photons is not
conserved, inelastic scattering occurs. As a result, the frequency of the incident and scattered
photons differs. In optical fibers, the inelastic scattering typically has two components:
Brillouin and Raman. The Brillouin scattering propagates as acoustic waves and is the result
of density shifts caused by interaction between pulsed and continuous light waves counter-
propagating in the optical fiber (Kurashima et al., 1990). Brillouin scattering occurs at a
predictable amplitude but variable frequency. The Raman scattering is produced by
interactions between the photons and vibrating molecules within the lattice of the glass
fiber. This interaction results in a predictable frequency shift. The scattered light shifted to
lower frequencies (longer wavelengths) than the incident light is termed Stokes, while the
Development and Use of Fiber-Optic Distributed Temperature Sensing 613
scattered light shifted to higher frequencies (shorter wavelengths) is called anti-Stokes. The
backscattered Stokes signal is produced when a photon excites a molecule at a base
vibrational state and the molecule returns to a slightly higher state (Fig. 2(b)). When the
incident photon hits a previously-excited molecule and this molecule returns to the base
state (Fig. 2(c)), the backscattered anti-Stokes signal is generated. The greater the
temperature of the fiber, the more frequently these previously-excited molecules will be
encountered (Smith & Dent, 2005).
Fig. 1. Internal refractions of light in a cross section of an optical fiber
Fig. 2. Interactions between light and glass structure (Smith & Dent, 2005)
Although the frequencies of the Raman signals are predictable, their intensities are not. The
intensity of the anti-Stokes scattering depends strongly on the temperature of the silica
molecules of the fiber, while the intensity of the Stokes backscattering depends weakly on
this temperature (Fig. 3). Because of this differential temperature dependence, the ratio of
the anti-Stokes and Stokes signals can be used to determine the temperature of the fiber at
the point of scattering (this is described below). The distance of the point of light scatter is
calculated by time-domain reflectometry using the speed of light in the glass fiber, which is
dependent on the frequency of the light and the index of refraction of the fiber itself. In
standard optical fibers, the speed of light ranges between 1.7x108 and 2.0x108 m s-1 (Hausner,
2010). Commercial DTS instruments typically use a 10 or 20 ns laser pulse to illuminate the
optical fiber. After emitting the laser pulse, the backscattered signals begin to return to the
DTS instrument, where they are collected in discrete time periods by the detector unit.
Because the light in the fiber travels approximately 2 m in 10 ns, the signals that return to
the DTS instrument between the 0-10 ns following the injected pulse come from the first
meter of fiber. If the backscattering detection unit is set to 10 ns, the DTS instrument will
return temperature readings integrated over 1 m of fiber. The DTS instrument repeats the
pulse and data collection continuously for temporal integration periods as specified by the
user. A diagram of a typical DTS system is depicted in Fig. 4.
Fig. 3. Diagram of Rayleigh, Brillouin, and Raman backscattering in optical fibers
Fig. 4. Diagram of a typical distributed-temperature-sensing system

The intensities of the Raman anti-Stokes and Stokes signals collected at the DTS instrument
can be expressed by (Rogers, 1999):
I aS ( z ) = I 0 exp ( −α 0 z ) exp ( −α aS z )℘aSΓ aS (1)
I S ( z ) = I 0 exp ( −α 0 z ) exp ( −α S z )℘S ΓS (2)
where IaS(z) and IS(z) are the Raman anti-Stokes and Stokes signals, respectively; I0 is the
intensity of the laser pulse emitted from the DTS instrument; α0, αaS and αS are the
attenuation coefficients of the emitted laser pulse, the backscattered anti-Stokes signal and
the backscattered Stokes signal, respectively; ℘aS and ℘S are the Bose-Einstein probability
distribution of phonons of the Raman anti-Stokes and Stokes signals; and ΓaS and ΓS are the
capture coefficients of the Raman anti-Stokes and Stokes signals, which represent the
fraction of the scattered light that is directed back towards the light source. The term
“exp(α0)” in equations (1) and (2) represents the attenuation of the emitted light as it travels
away from the DTS instrument. The terms “exp(αaS)” and “exp(αS)” represent the
attenuation of the anti-Stokes and Stokes signals, respectively, as light travels back into the
DTS instrument. These attenuations are different because the Raman backscattered signals
have different frequencies. The Bose-Einstein probability distribution of phonons can be
described by (Farahani & Gogolla, 1999):
( ) ( )
−1
℘aS = ⎡exp − ΔE ⎤ ⎡1 − exp − ΔE ⎤ (3)
⎣ kT ⎦ ⎣ kT ⎦
( )
−1
℘S = ⎡1 − exp − ΔE ⎤ (4)
⎣ kT ⎦
where ΔE represents the difference in molecular energy states that drive Raman scattering; k
is the Boltzmann constant; and T is absolute temperature. Substituting equations (3) and (4)
into equations (1) and (2), respectively, and taking the ratio between the anti-Stokes and
Stokes intensities, R(z), yields:
I aS ( z ) exp ( −α aS z )
R ( z) =
IS ( z )
=
Γ aS
ΓS
exp − ΔE(kT )
exp ( −α S z )
(5)
The ratio between the anti-Stokes and Stokes capture coefficients, ΓaS/ΓS, can be considered
as a calibration parameter, C, that depends on the wavelength and frequency of the incident
laser, the backscattered Raman signals, the instrument’s photon detector, and the operating
conditions of the DTS instrument. Defining the differential attenuation of the backscattered
Stokes and anti-Stokes as Δα = αS – αaS and γ = ΔE/k, equation (5) results in:
⎛ γ ⎞
R ( z ) = C exp ⎜ − ⎟ exp ( Δα z ) (6)
⎝ T⎠
and rearranging terms (Suárez et al., 2011):
γ
T ( z) = (7)
ln [C ] − ln ⎣⎡ R ( z ) ⎦⎤ + Δα z
Equation (7) describes the temperature along the entire optical fiber. Here, for the sake of
simplicity, the differential attenuation has been assumed to be constant along the fiber but in
reality, this may change along the fiber due to differences in manufacturing, strain or other
defects that scatter the two Stokes frequencies differently. From equation (7), it can be seen
that integrated measurements of temperature, in both space and time, along the fiber can be
estimated using the ratio of the anti-Stokes and the Stokes intensities. As explained before,
when determining the temperature along the optical fiber, the location of the point of scatter
must also be known. This location can be found by measuring the duration of the light
reflection in the optical fiber. The distance, z, traveled by the light can be estimated by
(Yilmaz & Karlik, 2006):
ct
z= (8)
2n
where c is the speed of light in vacuum, n is the refractive index of the fiber, and t is the
propagation time of light in the forward and backward directions. More details about the
theoretical basis of DTS system can be found in other investigations (Rogers, 1999; Selker et
al., 2006a).
3. Use of fiber-optic distributed temperature sensing in the environment

The use of fiber-optic DTS in the environment began during the 1990s (Hurtig et al., 1994;
Förster et al., 1997) when temperature in boreholes was monitored with resolutions of
±0.1 °C. In the middle of the 2000s, DTS systems achieved acceptable levels of spatial and
temporal resolution, along with high temperature accuracy and resolution to monitor the
environment. For example, when the systems are calibrated carefully, using integration
times longer than 1 h and cables shorter than 5 km, a precision of ±0.01 °C can be achieved
every 1 meter of cable (Selker et al. 2006a,b). This great ability to precisely observe
temperatures at thousands of locations is the main thrust of DTS systems, which holds
potential for transformative observation of many environmental processes. In this section,
we present an overview of the DTS applications that have been performed in different
environments. To show the promise of this technology, we present hydrological, ecological,
engineering, and atmospheric DTS applications that have been studied.
3.1 Fiber-optic distributed temperature sensing in snow

Measurement of the thermal regime of snow has significant importance to a wide range
of environmental processes, yet can be surprisingly challenging to measure. Snow depth
and snow temperature strongly control the thermal balance and temperature profiles in
the underlying soil (Lachenbruch, 1959) as well as heat transfer between the snow and
atmosphere. The vertical distribution of temperature strongly controls metamorphism in
the snowpack, important for both avalanche forecasting and snowmelt dynamics. In
addition, the horizontal variation of snow temperatures can control the distribution of
permafrost and frozen ground, both important factors for infiltration, runoff and many
sub-snow biological processes. However, in almost all cases, snow temperature
measurements are limited to one or at most a few points of measurement in space across the
landscape.
While temperature sensing systems have been available for centuries, snow represents
several unique challenges. In the past, measurements were often made by hand in snow
pits, using simple thermometers but providing, at best, a very limited time series of the
temperature evolution. Even with advances in continuous, low-cost, low-power thermal
data collectors (Lundquist & Lott, 2008), measurements of the thermal regime of snow packs
are still generally limited to a few points in the vertical dimension, and rarely more than a
few different locales in any given watershed. Even continuous vertical measurements are
challenging, as solar radiation and wind often cause melting, heat conduction and/or
differential snow accumulation around measurement devices. The lack of ability to measure,
continuously in time and space, the snow thermal environments has limited our
understanding of the energy budgets of snow covered areas, as well as the spatial
distribution of snow physical properties. Recently, Tyler et al. (2008) demonstrated the
utility of DTS for measuring the thermal evolution of the snow/soil interface temperatures.
In that work, standard telecommunication cables were deployed across several hundred
meters at two experimental watersheds and monitored to determine the thermal differences
between snow and bare ground, and also the effects of aspect on soil/snow interface
temperatures. In both watersheds, deployed fibers withstood winter conditions and
provided 300-400 individual measurements of snow/soil interface temperatures at temporal

scales from minutes to hours (Tyler et al., 2008).
Since basal snow temperatures are now proven to be easily measured using DTS, the
distribution of internal snowpack temperatures represents the next challenge in distributed
sensing. While it has long been recognized that vertical gradients in the snowpack
temperature lead to significant snow metamorphosis and instability, little study has been
made on the horizontal gradients in temperature that may exist due to differences in
shading and snow accumulation. Areas of steep horizontal gradient in snow temperatures
may result in horizontal differences in vapor diffusion, snow morphology and strength,
thereby producing local zones of weakness that may serve as avalanche trigger points.
Several methods have been proposed in the last few years to measure internal snowpack
temperatures using fiber-optic sensing in harsh conditions during winter and snow loading.
Researchers from École Polytechnique Fédérale de Lausanne and Oregon State University
have suspended and stretched fiber-optic cables at fixed elevations above the soil prior to
the onset of winter, and allowed the fiber-optic “fence” to be subsequently buried by snow.
Such a design provides temperature data at fixed elevations above the land surface and
precise knowledge of the cable location. Challenges to the fixed height method include cable
sagging due to snow load, stress points at any cable attachment point and alteration of the
snow deposition due to the presence of the pre-emplaced cable “fence”. Alternatively,
optical fiber can be deployed as snow accumulates during the winter season and allowed to
follow both the snow surface, and any compaction/consolidation that occurs. Researchers at
the University of Nevada, Reno and the University of California, Santa Barbara have used
this deployment strategy in the Sierra Nevada of California, where storm accumulations are
typical large and quickly bury the deployed fiber, and access to the site allows for easy
deployment prior to major storms. Far less precise spatial location (both horizontal and
depth) can be predicted for this method, as the optical fiber is deployed before the storm
cycle begins, the optical fiber can move due to wind during the storm, and blowing snow
may bury the cable differentially. Following the storm, the optical fiber should remain at the
interface of the old and new snow, but its height above the ground surface will decrease as
the snow consolidates through the season. This method of deployment therefore tracks
snow layer temperatures, rather than at fixed depths in the snowpack. While there is
potentially less strain on the fiber due to attachment points and less effect of the cable
placement on snow accumulation, both methods are likely to alter the snow melting
dynamics as the optical fiber approaches the snow surface during the melt season. While the
“snow fence” method requires significant work at the beginning of the season, the “pre-
storm” deployment requires significant work throughout the snow season and repeated
access to the study site.
At the CRREL-UCSB Eastern Sierra Snow Study Site (CUES, Painter et al., 2000; Bales et al.,
2006), fiber-optic DTS monitoring of basal and snowpack temperatures has been on-going
since 2007 (Tyler et al., 2008). During the 2009/2010 and 2010/2011 winter seasons, fiber-
optic cable (ADSS Flat Drop Cable, AFL Telecommunications, Duncan, SC) was deployed
across the CUES field site to measure internal snowpack temperatures. Prior to the first
snow, fiber-optic cable was deployed, including a 70m loop of cable south of the instrument
shelter, and three coils of approximately 70m of fiber to be deployed during the winter
season. Beginning on February 4, 2010, these coils were deployed on the snow surface over
one another on successive storms (February 4, 21, and March 2, respectively). The snow
depths at these times, as measured by an ultrasonic sounder, were 40 cm, 270 cm and
approximately 350 cm, respectively. Throughout the winter season, DTS temperature
measurements were made (Sensornet Halo, Sensornet Inc., Hertfordshire, UK) using 1
minute averaging intervals and post-processed using instrumented calibration baths (Tyler
et al., 2008).
Fig. 5. (a) Horizontal distribution of snowpack temperatures from the CUES site on March 3,
2010. The elevations shown in the figures represent initial position of the fiber. Note that
significant horizontal variation in temperature is visible in the 40 and 270 cm cables.
(b) Horizontal distribution of snowpack temperatures from the CUES site later in the season
(April 3, 2010) shows much warmer snowpack as well as smaller vertical gradients in snow
temperature. Horizontal differences continue to remain along each depth, and remain fixed
in space when compared to the March 3, 2010 data
Fig. 5 (a) shows the temperature distribution from the soil surface up through the snowpack
along the 70 meters of instrumented snow (Note that the y-axis is reversed to place the
uppermost coil, which is also the coldest, at the highest elevation in the snowpack). Vertical
thermal gradients, assuming the initial snow depths as representative lengths, range from
0.3 to approximately 8 °C m-1. Horizontal gradients are also quite apparent (maximum ~0.25
°C m-1), particularly in the 40 and 270 cm cables where two zones of colder conditions can be
seen between 5 and 20 m and, to a lesser degree, between approximately 35 and 55 m. Both
of these locations correspond with areas of snow ablation, with the zone between them
characterized by drifted snow. Fig. 5 (b) shows the same transect data from one month later
in the season (April 3, 2010). Snow temperatures are generally warmer, with much smaller
differences with depth. However, horizontal differences have persisted, and remain
stationary. At this time, vertical gradients have decreased to near zero in some areas, while
the horizontal gradients reach a maximum of approximately 1 °C m-1. Based upon visual
observation of the transect late in the season, the zones of largest horizontal gradients
correspond to rapid changes in snow depth, i.e., between drifts. In these zones, since the
fiber-optic cable follows snow interfaces, the zones of coldest measurement correspond to
areas of shallowest snow, where the fiber-optic cable is closer to the snow/air interface.
While the horizontal gradients are small, their persistence and their gain in importance
compared to the vertical gradient late in the snow season suggest that they may have
significance in late season snow melt and snow stability. These transitions in temperature
may lead to melt enhancement as well as serve as locations of weakness that could result in
avalanche triggers on the slopes in the same area. Such horizontal variations in snow
temperature are rarely recognized from individual snow pits or profiles and point to the
utility of fiber-optic DTS approaches for snow thermal measurement.
The use of DTS for snow thermal process monitoring is relatively new and shows significant
progress. Applications to snowmelt dynamics and frozen ground analysis are already
underway and provide unique remotely sensed data that are important for hydrologic and
environmental monitoring. Progress has been made in snowpack internal temperature
measurement, and preliminary data suggest that horizontal gradients in snow temperature
become significant late in the season as snowpacks approach melting conditions. Using
fiber-optic DTS provides the first and only reliable method in which the spatial variability of
snowpack temperatures can easily and remotely be measured. Measurement of both vertical
and horizontal gradients and their spatial variability may provide important insights into
snowpack dynamics, melting and avalanche susceptibility.
3.2 Fiber-optic distributed temperature sensing in ecohydrology

Devils Hole, a geothermally influenced groundwater-filled fracture in the carbonate aquifer
of the southern Mojave Desert, represents a unique intersection of geology, hydrology, and
ecology. The system offers a window into the carbonate aquifer of the Death Valley regional
flow system (Riggs & Deacon, 2002), and is home to the world’s only extant population of
the endangered Devils Hole pupfish (Cyprinodon diabolis). Thought to comprise the smallest
habitat containing the entire population of a vertebrate species (Moyle, 2002), Devils Hole
was severely impacted by the development of nearby groundwater resources in the 1960s
and early 1970s (Andersen & Deacon, 2001), leading to extensive litigation that to this day
influences the United States’ management of endangered species and water resources (Riggs
& Deacon, 2002). The scientific footprint of Devils Hole is as significant as its legal impact;
the system has been the site of pioneering work in palaeoclimate reconstructions (Winograd
et al., 2006), as well as ichthyology (Minckley & Deacon, 1973) and evolution (Miller, 1950;
Lema & Nevitt, 2006). In the late 1990’s, the population of C. diabolis experienced an
unexplained decline that has only recently stabilized, and a number of the hypotheses to
explain this decline posit changes to the physical habitat of the system, especially changes in
water temperatures.
Cyprinodon species tend to be very sensitive to water temperatures, and the 33-34 °C waters
in Devils Hole are near the upper tolerance of most species (Brown & Feldmeth, 1971). The
reproductive cycle of this fish is particularly temperature-dependent, and oogenesis in adult
females, the viability of fertilized eggs, and the development of hatched larvae can all be
severely retarded by exposure to higher temperatures (Shrode & Gerking, 1977). The
thermal tolerances of Cyprinodon species also appear to be related to the variability of the
water temperatures in which they live (Otto & Gerking, 1973) – the Devils Hole pupfish,
which lives its entire life in water is seldom outside 33-35 °C, is especially susceptible to
thermal stresses. Because the population of C. diabolis varies seasonally from highs of 200 to
lows of fewer than 70 individuals, small variations in water temperature can have enormous
impacts on the health of the population as a whole. Over the last three years, researchers
have been using DTS instrumentation to characterize the thermal regime of Devils Hole.
Devils Hole comprises a 3 m x 8 m shallow shelf, which provides a spawning ground for the
pupfish population, and a deep pool with a surface area of barely 30 m2. The fracture that
forms the deep pool strikes NE-SW and dips approximately 70°, with a uniform aperture of
approximately 4 m. Between the surface and a depth of 30 m, the system is approximately
15 m wide along the NE-SW axis; below 30 m it opens into a cavern approximately 100 m
wide, although the 4 m aperture remains. Devils Hole has been explored by divers to a
depth of 130 m, and the clear waters allow visibility to almost 150 m (Riggs & Deacon, 2002).
Below that point, the structure of the system is unknown. In January 2009, a fiber-optic cable
was permanently installed in the deep pool to observe vertical temperature profiles.
Fig. 6. Seasonal changes in the vertical temperature profiles of Devils Hole. The heavy lines
indicate the mean temperature over 48 hours, and the shaded areas indicate the root mean
square error of the calibrated DTS data (0.04° and 0.06° C in January and July, respectively)
Hausner et al. (2010) presented seasonal changes in the vertical temperature profiles that
revealed a previously unidentified cycle of convective mixing. Fig. 6 shows the vertical
temperature profiles observed in January and July, 2009. The near-uniform temperatures in
the January profile indicate a system in which convective mixing is the dominant mode of
heat transfer, while the constant temperature gradient observed in July fits a system
dominated by conduction. In the summer, the water on the surface is warm, and tends to
remain at the surface; the waters below the surface are warmed and stabilized by the natural
geothermal gradient in the area, and the inverted temperature profile results. When water
on the shallow shelf cools in the winter, the greater density causes it to plunge through the
stratified water below, and the system mixes. The seasonal mixing affects the oxygen
dynamics of the system, the winter distribution of allochthanous carbon (one of the
pupfish’s primary sources of food), and the availability of nutrients in the water column.
However, these patterns would likely not have been noticed without the use of DTS in
Devils Hole. This mixing model is based on observations of temperature gradients as small
as 0.005 °C m-1. Because a single common calibration is used to return DTS temperature
observations along the entire length of the cable, fine gradients such as the one shown in
Fig. 6 can be more easily observed and more confidently quantified with DTS instruments
than with traditional sensors. Using a thermistor string, for example, the summer
temperature gradient shown in Fig. 6 would likely have been lost in the noise generated by
comparing multiple, separately calibrated instruments.
3.3 Fiber-optic distributed temperature sensing in salt-gradient solar ponds

A salt-gradient solar pond is a non-traditional solar collector that can provide long-term
thermal storage and recovery for the collected energy. It is an artificially stratified water
body that consists of three distinct zones (Suárez et al., 2010a): the upper convective zone,
which is a thin layer of cooler and fresher water; the non-convective zone, comprised of a
salt-gradient to suppress global circulation within the pond; and the lower-convective zone,
in which salinity and temperature are the highest. The solar radiation that reaches the
bottom of the pond is transformed into thermal energy and warms the brine in the lower
convective zone. This warmer brine cannot rise beyond the lower convective zone because
the effect of salinity on density is greater than the effect of temperature. The heat stored in
the lower convective zone can only escape to the atmosphere by conduction, making the
thickness of the non-convective zone a critical operating parameter for efficient solar pond
operation (Suárez et al., 2010b). Because the brine has a low thermal conductivity, heat
losses by conduction are relatively small. The hot brine in the lower convective zone may
then be used directly for heating (Rabl and Nielsen, 1975), thermal desalination (Lu et al.,
2001; Suárez et al., 2010c), or for other low-temperature thermal applications (Kumar and
Kishore, 1999).
To investigate sustainable freshwater production using thermal desalination powered with
solar energy, Suárez et al. (2010a, 2011) constructed a 1.0-m depth experimental salt-gradient
solar pond. The pond was built inside a laboratory operated under controlled conditions,
and was initially exposed to artificial lights 12 h per day. After reaching a thermal quasi
steady-state, the pond was exposed continuously to the artificial lights until a new thermal
steady-state was reached. Then, heat was extracted at approximately 0.5 m depth and was
used to drive membrane distillation (Suárez, 2010). This pond was instrumented with a
variety of sensors, including a vertical high-resolution DTS system to accurately monitor the
thermal stratification and to investigate thermohaline circulation. This vertical high-
resolution DTS system was constructed by wrapping a fiber-optic cable around a polyvinyl
chloride pipe of approximately 1.3 m in length. This allowed temperature measurements
every 1.1 cm in the vertical direction. A detailed explanation of the experimental salt-
gradient solar pond and the characteristics of the vertical high-resolution DTS system are
presented by Suárez et al. (2011).
Fig. 7 (a) presents the thermal evolution of the experimental salt-gradient solar pond as
measured by the vertical high-resolution DTS system. Over the 15-days of the maturation
period, the brine of the lower convective zone (0.65 m to 1.0 m) was warmed approximately
18 °C and showed strong internal convection when the lights were on. When the lights were
off, the lower convective zone stratified due to the heat lost through the bottom and sides of
the pond as well as through the non-convective zone. The upper convective zone (0.0 m to
0.1 m) also showed a diurnal pattern of strong stratification (due to radiation absorption)
and subsequent mixing at night (due to penetrative convective mixing) during the first week
of the experiment. After this, the upper convective zone showed strong internal convection
during the day and night. This occurred because of the higher temperatures in the non-
convective zone, which warms the bottom of the upper convective zone and induce thermal
convection cells in this zone. On the other hand, the non-convective zone (0.1 m to 0.65 m)
showed only conductive heat transfer effectively isolating the warm layer below. The data
collected using the vertical high-resolution DTS system also allowed closing of the energy
budget in this rather unique thermohaline environment. Tyler et al. (2009a) estimated the
sensible heat flux (Fig. 7 (b)) at the surface of the pond by measuring the net radiation (at the
water surface), estimating the evaporation rate over the pond, and combining these
measurements with the change in heat storage in the pond, which was evaluated using the
DTS system.
Fig. 7. (a) Thermal evolution of the water column and the air above it in an experimental
salt-gradient solar pond as measured by a vertical high-resolution fiber optic distributed-
temperature-sensing system (Suárez et al., 2011). (b) Heat fluxes at the surface of the pond
were calculated by closing the energy balance (Tyler et al., 2009a)
Fig. 8. Temperature profile in the salt-gradient solar pond at the beginning and at the end of
a heat extraction experiment. Modified from Suárez et al. (2010a)
Suárez et al. (2010a) also presented results of heat extraction in the experimental salt-
gradient solar pond. As shown in Fig. 8, before heat was extracted from the pond, the
temperature in the lower convective zone was approximately 45 °C and the lower portion of
the non-convective zone was completely mixed (between 0.55 and 0.65 m). The temperature
step change of 0.5 °C observed in the interface between these zones suggests that transport
is occurring by double-diffusive convection (Turner, 1974). Fig. 8 also shows that heat was
extracted directly from the non-convective zone and indirectly from the lower convective
zone. As heat was extracted, the temperature at the heat extraction depth decreased, and
when it was cooler than a threshold, the fluid at this depth sank to the bottom of the pond.
As a result, the warmer brine from the lower convective zone rose and energy was now
extracted from this brine. After 45 hours of heat extraction, the temperature profile showed
a staircase shape below the heat extraction depth. This staircase is typical of double-
diffusive convective systems (Kelley et al., 2003) and occurred inside the salt-gradient solar
pond because of the method used to create the salt-gradient within the pond (Suárez et al.,
2010a).
3.4 Detection of illicit connections in storm water sewers

Storm water sewer systems discharge rain or storm waters into surface waters without
treatment. Illicit connections that introduce fouled water in storm water systems are very
problematic because they result in the release of untreated sewage in surface water bodies
such as rivers, lakes or even the sea. Recently, Hoes et al. (2009a) developed a searching
technique for detection of illicit connections in storm water system using DTS. Monitoring
the temperature in storm water sewers allows finding anomalous temperatures or
temperature variations at the illicit connections. Typically, the sewer temperatures are
influenced by surrounding air, soil, and sometimes by the temperatures of other types of
waters that enters the sewer (Dürrenmatt & Wanner, 2008). Temperatures in an approximate
range between 5 and 20 °C are expected for sewer conditions, and the variations within this
range can only occur on a daily and seasonal basis. On the other hand, domestic wastewater
usually shows a broader variation. For example, houses appliances and showers heat
domestic waters, and temperatures in the range 30–90 °C could be expected in the drain.
Even though the temperature drops as the domestic water travels before discharging into
the sewer system, temperatures warmer than 20 °C could be found at the illicit connection
(Hoes et al., 2009a).
Fig. 9 shows the DTS measurements performed by Hoes et al. (2009a) at an illicit connection
that was located using this new technique. The temperatures in the sewer were relatively
stable at approximately 12 °C, especially during the night. During the day and evening,
anomalous peaks of temperatures were observed at this specific location. On-site
verification by excavation and testing provided conclusive evidence that these temperature
anomalies resulted from illicit connections in the storm water system.
Fig. 9. DTS measurements performed in Korendijk (Netherlands) at the location of an illicit

connection of a storm water system. Modified from Hoes et al. (2009a)
3.5 Distributed temperature sensing as an indirect tool for soil moisture estimation
Soil moisture is a key-state variable in water and energy balances at the land surface, and
affects many different physical and environmental processes. The water content of a given
soil is a controlling factor for agriculture and crop production, biological activity within the
soil microbial community, and meteorological processes (Hausner, 2010). It also influences
how efficiently water and solutes can move through the soil matrix. Despite its widespread
influence and importance, measuring soil moisture is a difficult task, especially at field scale
when soil heterogeneity could be important. The spatial scales of water content
measurements present additional challenges. At small scales, single point measurements can
be taken at specific intervals. However, these measurements are typically expensive, time
consuming, and are difficult to extrapolate to field scale (e.g., over areas larger than 0.1
km2). Remote sensing technologies can estimate near-surface soil moisture over larger areas,
but there is a lack on methods to estimate in-situ water content at scales ranging from 0.1 to
80 km2 (Robinson et al., 2008). By burying fiber-optic cables and recording the spatial
distribution of soil temperatures over several days, Steele-Dunne et al. (2010) demonstrated
that DTS can be used to estimate the distribution of soil moisture within these spatial scales.
They first described heat transfer in a soil column using the diffusion equation:
∂T ∂ 2T κ (θ ) ∂ 2T
= D (θ ) 2 = (8)
∂t ∂z C (θ ) ∂z2
where T is temperature; t is time; θ is volumetric water content (or soil moisture); D(θ), κ(θ)
and C(θ) are the thermal diffusivity, thermal conductivity, and thermal heat capacity of the
bulk soil, respectively, which are each functions of soil moisture; and z is depth. By
measuring temperature at different depths, the soil thermal properties can be inferred by
inversion, then the soil moisture can be estimated using a representative relationship
between the soil thermal conductivity and soil moisture. Steele-Dunne et al. (2010) chose the
Campbell exponential model (Campbell, 1985):
κ (θ ) = a + bθ − ( a − d ) exp ⎡ − ( cθ ) ⎤
e
(9)
⎣ ⎦
where a, b, c, d and e are empirical parameters that depend on the volume fraction of soil
minerals (e.g., quartz), organic matter, and on the clay fraction.
Steele-Dunne et al. (2010) buried two fiber-optic cables at approximately 8 and 10 cm below
the surface of the ground to monitor the thermal response of the diurnal signal. A section of
approximately 50 m was used to measure the transient temperature field at each depth and
soil moisture was derived from these measurements. Even though they obtained reasonable
results for both the soil thermal diffusivity and relative saturation (Fig. 10), it must be
pointed out that their method relies on knowledge of the depth of the cable, and small
variations were shown to cause large errors in the estimation of the thermal properties of the
soil and thus, in the estimated values of soil moisture. The key of this method is the linkage
between soil moisture and soil temperatures. Thus, the relationship between soil moisture
and soil thermal conductivity must also be ascertained using in situ measurements over the
entire range of saturation values (Steele-Dunne et al., 2010).
Fig. 10. (a) Estimated thermal diffusivity (D) at a selected location along the fiber-optic cable.
(b) Mean value of relative saturation observed using ECH2O probes and inferred relative
saturation from the estimated thermal diffusivity using the Campbell model (1985).
Modified from Steele-Dunne et al. (2010)
Sayde et al. (2010) also demonstrated the feasibility of using DTS to obtain accurate
distributed measurements of soil water content. While Steele-Dunne et al. (2010) made
passive observation of the natural diurnal cycles of heating and cooling, Sayde et al. (2010)
actively heated a fiber-optic cable that was buried in variably saturated sand and then
correlated the cumulative temperature increase and dissipation at different depths with the
corresponding soil water content. Their correlations showed coefficients of determination
(R2) higher than 0.985, showing the promise of this method. Current efforts on spatially
distributed soil moisture estimation are trying to combine both the passive and active DTS
methods to obtain more precise measurements at field scale.
3.6 Atmospheric boundary layer estimation using distributed temperature sensing

The atmospheric boundary layer height, also called the mixed-layer or mixing height, is the
lower part of the atmosphere in which the influence of heating and cooling, and surface
friction is important (Brutsaert, 1982). The vertical mixing that occurs in this layer is driven
by turbulence produced primarily by wind shear and buoyancy. The height of the
atmospheric boundary layer is a key parameter for describing the physical state of the lower
troposphere because it allows the prediction of air pollution concentrations or surface
temperatures. For instance, air pollutants emitted at the ground surface will mix into this
layer and, depending on meteorological conditions, can reach elevations from meters to a
few kilometers (Brutsaert, 1982; Keller et al., 2011). Although it is one of the main
parameters in atmospheric applications, the weakest point in meteorology is still the
determination of the atmospheric boundary layer height (Builtjes, 2001).
Fig. 11. (a) Evolution of potential temperature in the lowest 100 m of the atmosphere and
mixing height as determined from exponential profile fits (bold-dashed line). (b) Example of
a thermal profile measured by the DTS system with the corresponding best fit of an
exponential curve. The mixing height is also shown. Modified from Keller et al. (2011)
Keller et al. (2011) presented a new method for measuring the air temperature profiles in the
atmospheric boundary layer. A tethered balloon lifting system was used to suspend the
optical fiber in the lowest 100 m of the atmosphere. As shown in Fig. 11, the temperature
data were used to estimate the height of the stable boundary layer during the night, which
varied between 5 and 50 m (September 11th data). They were also able to observe the
erosion of the stable atmospheric stratification that occurred shortly after sunrise when the
surface warmed. This resulted in an increase of turbulence, convection, and mixing due to
latent and sensible heat being transferred to the atmosphere.
3.7 Other applications of fiber-optic distributed temperature sensing

Since it is difficult to describe all the work that has been carried out using this technology, in
this section we present a brief summary of other environmental applications that were not
described before.
Surface water-groundwater exchange has been widely studied using DTS methods in
different environments such as estuaries (Henderson et al., 2009), rivers and streams (Lowry
et al., 2007; Westhoff et al., 2007; Vogt et al., 2010; Slater et al., 2010), and ditches and canals
(Hoes et al., 2009b). In general, these applications have used DTS technology to identify
gaining sections of rivers, i.e., sections of the river where groundwater enters into it. The
exception is the work of Vogt et al. (2010), where a vertical high-resolution DTS system was
installed to measure both stream and streambed-sediment temperatures, which also allowed
the estimation of the seepage rates from the river into the groundwater. Other
environmental applications related with stream dynamics include the processes that
controls thermal regime of saltmarsh channel beds (Moffett et al., 2008), development of
distributed stream temperature models (Westhoff et al., 2007), response of stream
temperatures in different riparian vegetation (Roth et al., 2010), and quantification of heat
retardation along streams (Westhoff et al., 2010). The effects of radiative heating on fiber-
optic cables used to monitor water temperatures have also been evaluated (Neilson et al.,
2010; Suárez et al., 2011).
Early DTS work in hydrogeology was focused on thermal monitoring of geothermal wells
and boreholes (Hurtig et al., 1994; Förster et al., 1997). Then, DTS systems were used to
analyze the dynamic subsurface thermo-hydraulic conditions in aquifers (Macfarlane et al.,
2002). Recently, Freifeld et al. (2008) developed a methodology to determine thermal
conductivity in boreholes by combining a fiber-optic DTS system with a resistance heater,
which created a controlled thermal perturbation in the borehole. The transient thermal data
is inverted to estimate the thermal conductivity profile along the length of a wellbore with a
spatial resolution equal to the spatial resolution of the DTS instrument. They also were able
to determine the baseline geothermal profile and the ground surface temperature history in
their study site (High Lake region, Nunavut, Canada).
4. Important considerations and future trends

Even though DTS systems have significant advantages over traditional measurement
systems (as shown above), they also have limitations that must be assessed to maximize the
potential of this technology. The performance of a DTS system is highly dependent upon the
design of the experiment, the DTS instrument, the fiber-optic cables and connectors, the
calibration, and the operating conditions. In this section we briefly discuss these factors in
order to obtain thermal measurements with improved accuracy and precision.
4.1 Distributed-temperature-sensing instruments

There are a wide variety of commercial DTS instruments that use different methods for
signal generation and data processing, which will affect the resolution and cost of the
system. The cost of a DTS instrument ranges from approximately $15,000 to more than
$150,000 (April 2011 U.S. Dollars). This cost depends on the features of each system (e.g.,
spatial and temporal resolution, ports to connect external thermocouples, number of
channels). For temporary installations, there is also the option to lease DTS instruments for a
lower cost (Tyler and Selker, 2009). The power requirements for DTS instruments range
from approximately 20 to more than 100 W. To select the most cost-effective system, the
required temporal and spatial repeatability must be evaluated by taking into account the
goals and characteristics of each installation. The temporal and spatial averaging need to be
long enough to provide suitable temperature resolutions, but short enough to observe the
transient phenomena and the physics behind the system that is being measured.
4.2 Fiber-optic cables and connectors

As the cable acts as a thermometer, the selection of the cable is crucial in the experimental
design. The cable typically includes a plastic jacket, tensile strength members, armoring to
protect the optical fiber, a water-tight barrier, and the optical fiber(s). The cost of the cable
can range from approximately $0.5 m-1 to more than $10 m-1 and cable weights can range
from <1 kg km-1 to more than 30 kg km-1. In general, the cost and weight of the cable are
determined more by the armoring and construction than by the optical fiber itself. As cables
get heavier, they are more difficult to handle and they can respond more slowly to
temperature changes. Localized strains and stresses over the cable can result in greater
signal losses, thus it is important to understand the environment of the installation before
selecting the cable. Another important factor is the exposure of the cable to solar radiation,
especially when used in the atmosphere or in shallow streams. Fiber-optic cables, as well as
other thermal sensors, can absorb solar radiation and monitor higher temperatures than
those of the surrounding ambient. This issue is more important when air or water velocities
are small, and when the magnitude of solar radiation is large (Neilson et al., 2010). The effect
of solar radiation absorption on cables can be reduced using reflective coatings or shielding
the cable (Suárez et al., 2011). In addition, in a long-term installation with a cable exposed to
sunlight, for example, it is important to select jackets that can withstand ultra-violet
radiation and that can minimize radiative heating.
Connectors can also have an impact on signal strength. The standard connectors in the DTS
industry typically produce a signal loss on the order of 0.1–0.2 dB (or even more if the
connector is not clean or is incorrectly aligned). This loss is approximately the same loss that
occurs in 300–700 m of fiber-optic cable, respectively (assuming an attenuation of ~0.3 dB
km-1). Ideally, the cable should only have connectors to physically connect the cable itself
with the DTS instrument. If fibers need to be joined, fusion splices are recommended, since a
properly fused fiber splice has a loss on the order of 0.01–0.03 dB.
4.3 Calibration and types of measurements

The calibration process is critical to achieve measurements with high resolution. An
optimum deployment should have at least three sections of known temperature, each one
with at least 10 sampling points within its length (e.g., 10 m when the spatial sampling
resolution is 1 m). As suggested by Suárez et al. (2011), two of these sections can be at the
same temperature and the other section needs to be at a different temperature. Ideally, the
known-temperature sections should bracket the expected observations in the corresponding

environment. If possible, the fiber-optic cable should have a loop to return the cable to the
instrument (see Suárez et al. (2011) for more details about calibration procedures). This
permits the DTS instrument to interrogate the fiber-optic from each end, i.e., allowing
single- or double-ended measurements. Single-ended measurements refer to temperatures
estimated from light transmission in only one direction along the optical fiber. These
measurements assume a uniform rate of differential attenuation (Δα) over the entire fiber,
and provide greater precision near the instrument, degrading with distance because of the
energy loss along the fiber length. Double-ended measurements refer to temperatures
estimated from light transmission in both directions along the optical fiber. In these
measurements, the temperature is estimated using single-ended measurements made from
each end of the fiber, and can account for spatial variation in the differential attenuation of
the anti-Stokes and Stokes backscattered signals, which typically occurs in strained fibers.
Double-ended measurement results in a signal noise more evenly distributed across the
entire length of the optical fiber, but uniformly greater than that obtained in a single-ended
measurement (Tyler et al., 2009b; Suárez et al., 2011). Single-ended calibrations are
encouraged for short cables (i.e., smaller than 1 or 2 km) since they provide more precision
near the instrument. However, sometimes strains or sharp bends in the deployed fiber-optic
cable yields large localized losses in the Stokes and anti-Stokes signals, which decrease the
magnitude of the signals and add noise to the temperature data. Because these localized
losses cannot be handled adequately by a single uniform value of the differential
attenuation, further calibration is sometimes required to translate the scattered Raman
signals into usable temperature data. In these cases, double-ended measurements are
recommended because they allow the calculation of the differential attenuation along the
entire length of the cable, and are much better able to handle the step losses introduced by
strains and bends.
4.4 Operating conditions

An issue that has been observed in DTS installations is drift of the instrument. This drift
typically occurs because of large variations in the instrument’s temperature, particularly
when the DTS instrument is subject to large daily temperature fluctuations in the field. The
best solution to minimize this drift is to put the instrument in a controlled environment if
possible. Other solution to minimize drift is to calibrate the DTS instrument at every
measurement (sometimes referred to as dynamic calibration).
4.5 Current and future trends

As previously described, the ability to precisely measure temperature at thousands of
locations is the main thrust of DTS systems. This capability has opened a new window for
observation of environmental processes. Typical DTS instruments currently used in
environmental applications can achieve temperature resolutions as small as ±0.01 °C, and
spatial and temporal resolutions of 1-2 m and 10-60 s, respectively. At present, there are
ongoing efforts to improve both spatial and temporal resolution of DTS systems. A high-
resolution DTS instrument (Ultima, Silixa, Hertfordshire, UK) with temporal and spatial
resolutions of 1 Hz and 12.5 cm, respectively, was recently commercialized and is under
testing in environmental applications. This instrument simultaneously improved temporal
precision by a factor of ten and spatial precision by a factor of four over previously available
units. It was first deployed for observation of turbulent and stable atmospheric processes
(http://oregonstate.edu/bmm/DONUTSS-2010/first-deployment-array), and it has also

been utilized during a borehole heat tracer experiment designed to identify zones of high
horizontal hydraulic conductivity and borehole through-flow. While this new DTS instrument
has opened many possibilities, observation of atmospheric processes, for example, still
needs improvement of temporal resolution to monitor turbulent processes. Instruments with
this improved resolution are expected to be available in the near future and definitively will
open new opportunities for observation of environmental processes.
5. Conclusion
In the environment, heat transfer mechanisms are combined in a variety of ways and span
spatial scales that range from millimeters to kilometers. This extremely wide spatial scaling
has been a barrier that limits observation, description, and modeling of environmental
processes. The introduction of fiber-optic DTS has contributed to fill the gap between these
two disparate scales. Fiber-optic DTS has proven effective to precisely observe temperatures
at thousands of locations at the same time, with no issues of bias, and avoiding variability
due to use of different sensors.
In this work, we have shown some of the environmental applications that have benefited
from DTS methods. For instance, using fiber-optic DTS provides the first and only reliable
method in which the spatial variability of snowpack temperatures can easily and remotely
be measured. Measurement of both vertical and horizontal gradients and their spatial
variability may provide important insights into snowpack dynamics, melting and avalanche
susceptibility. DTS methods also have improved thermal measurements in natural and
managed aquatic systems. For example, the hydrodynamic regimes in Devils Hole were
observed at resolutions smaller than 0.1 °C, allowing observation of temperature gradients
as small as 0.003 °C m-1. This resolution allowed the examination of seasonal oxygen and
nutrient distribution in the water column. In salt-gradient solar ponds, this temperature
resolution allowed observation of both mixing and stratification, which is important for
pond efficiency. In both Devils Hole and the solar pond, fiber-optic DTS provided high-
resolution thermal measurements without disturbance of the water column. DTS methods
also have been successfully utilized in other environments such as in atmosphere, streams,
boreholes, and in many applications to understand the interdependence between
groundwater and surface water. Novel extensions of DTS methods include spatially
distributed soil moisture estimation, detection of illicit connections in storm water sewers,
and there are many more to come in the near future, especially because the technology is
growing and improving the spatial and temporal resolutions of DTS instruments, which will
open new opportunities for environmental observations.
6. Acknowledgement
This work was funded by the National Science Foundation by Award NSF-EAR-0929638.
7. References
Alfe, D., Gillan, M.J., Vocadlo, L., Brodholt, J. & Price, G.D. (2002). The ab initio simulation
of the earth’s core. Philosophical Transaction of the Royal Society of London A, Vol.360,
No.1795, (June 2002), pp. 1227-1244, ISSN 1364–503X
Alpers, M., Eixmann, R., Fricke-Begemann, C., Gerding, M. & Höffner, J. (2004).
Temperature lidar measurements from 1 to 105 km altitude using resonance,
Rayleigh, and Rotational Raman scattering. Atmospheric Chemistry and Physics,
Vol.4, No.3, (2004), pp. 793-800, ISSN 1680-7316
Andersen, M.E & Deacon, J.E. (2001). Population size of Devils Hole pupfish (Cyprindon
diabolis) correlates with water level. Copeia, Vol.2001, No.1, ( February 2001), pp.
224-228, ISSN 0045-8511
Bales, R.C., Molotch, N.P, Painter, T.H, Dettinger, M.D., Rice, R. & Dozier, J. (2006).
Mountain hydrology of the western United States. Water Resources Research, Vol.42,
No.W08432, (2006), 13 pp., ISSN 0043-1397
Bense, V.F. & Kooi, H. (2004). Temporal and spatial variations of shallow subsurface
temperature as a record of lateral variations in groundwater flow. Journal of
Geophysical Research, Vol.109, No. B04103, (2004), 13 pp., ISSN 0148-0227
Branco, B.F. & Torgersen, T. (2009). Predicting the onset of thermal stratification in shallow
inland waterbodies. Aquatic Sciences, Vol.71, No.1, (March 2009), pp. 65-79, ISSN
1015-1621
Brown, J.H. & Feldmeth, C.R. (1971). Evolution in constant and fluctuating environments:
thermal tolerances of desert pupfish (Cyprinodon). Evolution, Vol.25, No.2, (June
1971), pp. 390-398, ISSN 0014-3820
Brutsaert, W. (1982). Evaporation into the atmosphere: theory, history and applications (1st
Edition), Springer, ISBN 9789027712479, London, England
Builtjes, P.J.H. (2001). Major twentieth century milestones in air pollution modelling and its
Applications, In: Air Pollution Modeling and its Applications XIV, Gryning, S.E. &
Schiermeier (eds.), Springer, pp.3-16, Kluwer Academic/Plenum Publishers, ISBN
0306465345, New York
Campbell, G.S. (1985). Soil physics with BASIC: transport models for soil-plant systems (3rd
Edition), Elsevier, ISBN 9780444425577, New York, USA
Dakin, J.P., Pratt, D.J., Bibby, G.W. & Ross, J.N. (1985). Distributed optical fiber Raman
temperature sensor using a semiconductor light-source and detector. Electronics
Letters, Vol.21, No.13, (1985), pp. 569-570, ISSN 0013-5194
Dürrenmatt, D.J. & Wanner, O. (2008). Simulation of the wastewater temperature in sewers
with TEMPTEST. Water Science and Technology, Vol.57, No.11, (2008), pp.1809-1815,
ISSN 0273-1223
Eichinger, W.E., Cooper, D.I., Parlange, M. & Katul, G. (1993). The Application of a
Scanning, Water Raman-Lidar as a Probe of the Atmospheric Boundary Layer. IEEE
Transactions on Geoscience and Remote Sensing, Vol.31, No.1, (January 1993), pp.70-79,
ISSN 0196-2892
Farahani, M.A. & Gogolla, T. (1999). Spontaneous Raman Scattering in Optical Fibers with
Modulated Probe Light for Distributed Temperature Raman Remote Sensing.
Journal of Lightwave Technology, Vol.17, No.8, (August 1999), pp.1379-1391, ISSN
0733-8724
Förster, A., Schrötter, J., Merriam, D.F. & Blackwell, D. (1997). Application of optical-fiber
temperature logging—An example in a sedimentary environment. Geophysics,
Vol.62, No.4, (July-August 1997), pp.1107-1113, ISSN 0016-8033
Fridleifsson, I.B., Bertani, R., Huenges, E., Lund, J.W., Ragnarsson, A. & Rybach, L. (2008).
The possible role and contribution of geothermal energy to the mitigation of
climate change, In: O. Hohmeyer and T. Trittin (Eds.) IPCC Scoping Meeting on
Renewable Energy Sources, Proceedings, pp. 20-25, Luebek, Germany, January 20-25,
2008
Freifeld, B.M., Finsterle, S., Onstott, T.C., Toole P. & Pratt, L.M. (2008). Ground surface
temperature reconstructions: using in situ estimates for thermal conductivity
acquired with a fiber-optic distributed thermal perturbation sensor. Geophysical
Research Letters, Vol.35, No.L14309, (2008), 5 pp., ISSN 0094-8276
Hausner, M.B. (2010). Estimating in situ integrated soil moisture content using fiber-optic
distributed temperature sensing (DTS) measurements in the field. M. Sc. Thesis,
University of Nevada, Reno, 124 pp.
Hausner, M.B., Tyler, S.W., Wilson, K.P., Gaines, D.B. & Selker, J.S. (2010). Devils Hole: a
window into the carbonate aquifer of the Death Valley regional flow system.
Abstract H31A-0977 presented at 2010 Fall Meeting, American Geophysical Union, San
Francisco, California, USA, December 2010
Henderson, R.D., Day-Lewis, F.D. & Harvey, C.F. (2009). Investigation of aquifer-estuary
interaction using wavelet analysis of fiber-optic temperature data. Geophysical
Research Letters, Vol.36, No.L06403, (2009), 6 pp., ISSN 0094-8276
Henderson-Sellers, B. (1984). Engineering Limnology (1st Edition), Pitman Published Limited,
ISBN 0273085395, London, England
Hoes, O.A.C, Schilperoort, R.P.S., Luxemburg, W.M.J., Clemens, F.H.L.R. & van de Giesen,
N.C. (2009a). Locating illicit connections in storm water sewers using fiber-optic
distributed temperature sensing. Water Research, Vol.43, No.20, (December 2009),
pp. 5187-5197, ISSN 0043-1354
Hoes, O.A.C., Luxemburg, W.M.J., Westhoff, M.C., van de Giesen, N.C. & Selker, J.S.
(2009b). Identifying seepage in ditches and canals in polders in the Netherlands by
distributed temperature sensing. Lowland Technology International, Vol.11, No.2,
(December 2009), pp. 21-26, ISSN 1344-9656
Hurtig, E., S. Großwig, S., Jobmann, M., Kühn, K. & Marschall, P. (1994). Fibre-optic
temperature measurements in shallow boreholes: experimental application for fluid
logging. Geotermics, Vol.23, No.4, (August 1994), pp. 355-364, ISSN 0375-6505
Kelley, D.E., Fernando, H.J.S., Gargett, A.E., Tanny, J. & Özsoye, E. (2003). The diffusive
regime of double-diffusive convection. Progress in Oceanography, Vol.86, No.3-4,
(March 2003), pp. 461-481, ISSN 0079-6611
Keller, C.A., Huwald, H., Vollmer, M.K., Wenger, A., Hill, M., Parlange, M.B. & Reimann, S.
(2011). Fiber optic distributed temperature sensing for the determination of the
nocturnal atmospheric boundary layer height. Atmospheric Measurement Techniques,
Vol.4, No.2, (2011), pp. 143-149, ISSN 1867-1381
Kersey, A.D. (2000). Optical fiber sensors for permanent downwell monitoring applications
in the oil and gas industry. IEICE Transactions on Electronics, Vol.E83c, No.3, (March
2000), pp. 400-404, ISSN 0916-8524
Kumar, A. & Kishore, V. (1999). Construction and operational experience of a 6000 m2 solar
pond at Kutch, India. Solar Energy, Vol.65, No.4, (March 1999), pp. 237-249, ISSN
0038-092X
Kurashima, T., Horiguchi, T. & Tateda, M. (1990). Distributed-temperature sensing using
stimulated Brillouin scattering in optical silica fibers. Optics Letters, Vol.15, No.18,
(1990), pp. 1038-1040, ISSN 0146-9592
Lachenbruch, A.H. (1959). Periodic heat flow in a stratified medium with applications to
permafrost problems. U.S. Geological Survey Bulletin, 1083-A, 36 pp.
Lean, J. & Rind, D. (1998). Climate Forcing by Changing Solar Radiation. Journal of Climate,
Vol.11, No.12, (December 1998), pp. 3069-3094, ISSN 0894-8755
Lee, K.K.M., Steinle-Neumann, G. & Akber-Knutson, S. (2009). Ab initio predictions of
potassium partitioning between Fe and Al-bearing MgSiO3 perovskite and post-
perovskite. Physics of the Earth and Planetary Interiors, Vol.174, No.1-4, (May 2009),
pp. 247-253, ISSN 0031-9201
Lema, S.C. & Nevitt, G.A. (2006). Testing an ecophysiological mechanism of morphological
plasticity in pupfish and its relevance to conservation efforts for endangered Devils
Hole pupfish. The Journal of Experimental Biology, Vol.209, No.18, (September 2006),
pp. 3499-3509, ISSN 0022-0949
Lowry, C.S., Walker, J.F., Hunt, J.H. & Anderson, M.P. (2007). Identifying spatial variability
of groundwater discharge in a wetland stream using a distributed temperature
sensor. Water Resources Research, Vol.43, No.W10408, (2007), 9 pp., ISSN 0043-1397
Lu, H. , Walton, J. & Swift, A. (2001). Desalination coupled with salinity-gradient solar
ponds. Desalination, Vol.136, No.1-3, (May 2001), pp. 13-23, ISSN 0011-9164
Lundquist, J.D. & Lott, F. (2008). Using inexpensive temperature sensors to monitor the
duration and heterogeneity of snow-covered areas. Water Resources Research, Vol.44,
No.W00D16, (2008), 6 pp., ISSN 0043-1397
Miller, R.R. (1950). Speciation in fishes of the genera Cyprinodon and Empetrichthys
inhabiting the Death Valley region. Evolution, Vol.4, No.2, (June 1950), pp. 155-163,
ISSN 0014-3820
Minckley, C.O & Deacon, J.E. (1973). Observations on the reproductive cycle of Cyprinodon
diabolis. Copeia, Vol.1973, No.3, (August 1973), pp. 610-613, ISSN 0045-8511
Moffett, K.B., Tyler, S.W., Torgersen, T., Menon, M., Selker, J.S. & Gorelick, S.M. (2008).
Processes controlling the thermal regime of saltmarsh channel beds. Environmental
Science and Technology, Vol.42, No.3, (January 2008), pp. 671-676, ISSN 0013-936X
Moya-Laraño, J. (2010). Can Temperature and Water Availability Contribute to the
Maintenance of Latitudinal Diversity by Increasing the Rate of Biotic Interactions?
The Open Ecology Journal, Vol.3, No.1, (2010), pp. 1-13, ISSN 1874-2130
Moyle, P.B. (2002). Inland Fishes of California: Revised and Expanded (1st edition), University of
California Press, ISBN 9780520227545, Berkeley, California, USA
Neilson, B.T., Hatch, C.E., Ban, H. & Tyler, S.W. (2010). Solar radiative heating of fiber-optic
cables used to monitor temperatures in water. Water Resources Research, Vol.46,
No.W08540, (2010), 17 pp., ISSN 0043-1397
Norand, C.W.B. (1920). Effect of High Temperature, Humidity, and Wind on the Human
Body. Quarterly Journal of the Royal Meteorological Society, Vol.46, No.193, (January
1920), pp. 1-14, ISSN 0035-9009
Otto, R.G. & Gerking, S.D. (1973). Heat tolerance of a Death Valley pupfish (genus
Cyprinodon). Physiological Zoology, Vol.46, No.1, (January 1973), pp. 43-49, ISSN
0031-935X
Painter, T.H., Donahue, D., Dozier, J., Li, W., Kattelmann, R., Dawson, D., Davis, R.E., Fiori,
J., Harrington, B. & Pugner, P. (2000). The Mammoth Mountain cooperative snow
study site: data acquisition, management, and dissemination. Proceedings of the
International Snow Science Workshop, Vol. ISSW2000, pp. 447-451, Big Sky, Montana,
USA, October 2000
Rabl, A. & Nielsen, C. (1975). Solar ponds for space heating. Solar Energy, Vol.17, No.1,
(April 1975), pp. 1-12, ISSN 0038-092X
Riggs, A. & Deacon, J.E. (2002). Connectivity in Desert Aquatic Ecosystems: The Devils Hole
Story, Proceedings of Spring-fed wetlands: important scientific and cultural resources of
the intermountain region, DHS Publication No. 41210, Las Vegas, Nevada, USA, May
2002, available from: http://www.dri.edu/spring-fed-wetlands
Robinson, D.A., Campbell, C.S., Hopmans, J.W., Hornbuckle, B.K., Jones, S.B., Knight, R.,
Ogden, F., Selker, J.S. & Wendroth, O. (2008). Soil moisture measurement for
ecological and hydrological watershed-scale observatories: a review. Vadose Zone
Journal. Vol.7, No.1, (February 2008), pp. 358-389, ISSN 1539-1663
Rogers, A. (1999). Distributed optical-fibre sensing. Measurement Science and Technology,
Vol.10, No.8, (August 1999), pp. R75-R99, ISSN 0957-0233
Roth, T.R., Westhoff, M.C., Huwald, H., Huff, J.A., Rubin, J.F.,Barrenetxea, G., Vetterli, M.,
Parriaux, A., Selker, J.S. & Parlange, M.B. (2010). Stream Temperature Response to
Three Riparian Vegetation Scenarios by Use of a Distributed Temperature
Validated Model. Environmental Science and Technology, Vol.44, No.6, (February
2010), pp. 2072–2078, ISSN 0013-936X
Sayde, C., Gregory, C., Gil-Rodriguez, M., Tufillaro, N., Tyler, S.W., van de Diesen, N.C.,
English, M., Cuenca, R. & Selker, J.S. (2010). Feasibility of soil moisture monitoring
with heated fiber optics. Water Resources Research, Vol.46, No.W06201, (2010), 8 pp.,
ISSN 0043-1397
Sclater, J.G, Parsons, B. & Jaupart, C. (1981). Oceans and Continents: Similarities and
Differences in the Mechanisms of Heat Loss. Journal of Geophysical Research, Vol.86,
No.B12, (1981), pp. 11535-11552, ISSN 0148-0227
Selker, J.S., Thevenaz, L., Huwald, H., Mallet, A., Luxemburg, W., van de Giesen, N.C.,
Stejskal, M., Zeman, J., Westhoff, M. & Parlange, M.B. (2006a). Distributed fiber-
optic temperature sensing for hydrologic systems. Water Resources Research, Vol.42,
No.W12202, (2006a), 8 pp., ISSN 0043-1397
Selker, J.S., van de Giesen, N.C., Westhoff, M., Luxemburg, W. & Parlange, M.B. (2006b).
Fiber optics opens window on stream dynamics. Geophysical Research Letters, Vol.33,
No.L24401, (2006), 4 pp., ISSN 0094-8276
Shrode, J.B. & Gerking, S.D. (1977). Effects of constant and fluctuating temperatures on
reproductive performance of a desert pupfish, Cyprinodon n. nevadensis.
Physiological Zoology, Vol.50, No.1, (January 1977), pp. 1-10, ISSN 0031-935X
Slater, L.D., Ntarlagiannis, D., Day-Lewis, F.D., Mwakanyamale, K., Versteeg, R.J., Ward, A.,
Strickland, C., Johnson, C.D. & Lane, J.W. (2010). Use of electrical imaging and
distributed temperature sensing methods to characterize surface water–
groundwater exchange regulating uranium transport at the Hanford 300 Area,
Washington. Water Resources Research, Vol.46, No.W10533, (2010), 13 pp., ISSN
0043-1397
Smith, E. & Dent, G. (2005). Modern Raman spectroscopy: a practical approach (1st Edition), John
Wiley and Sons, ISBN 978-0471497943, Sussex, England
Steele-Dunne, S.C., Rutten, M.M., Krzeminska, D.M., Hausner, M.B., Tyler, S.W., Selker, J.S.,
Bogaard, T.A. & van de Giesen, N.C. (2010). Feasibility of Soil Moisture Estimation
using Passive Distributed Temperature Sensing. Water Resources Research, Vol.42,

No.W03534, (2010), 12 pp., ISSN 0043-1397
Suárez, F. (2010). Salt-gradient solar ponds for renewable energy, desalination and reclamation of
terminal lakes. Ph. D. Thesis, University of Nevada, Reno, 195 pp.
Suárez, F., Childress, A.E. & Tyler, S.W. (2010a). Temperature evolution of an experimental
salt-gradient solar pond. Journal of Water and Climate Change, Vol.1, No.4, (2010), pp.
246-250, ISSN 2040-2244.
Suárez, F., Tyler, S.W. & Childress, A.E. (2010b). A fully coupled transient double-diffusive
convective model for salt-gradient solar ponds. International Journal of Heat and Mass
Transfer, Vol.53, No.9-10, (April 2010), pp. 1718-1730, ISSN 0017-9310.
Suárez, F., Tyler, S.W. & Childress, A.E. (2010c). A theoretical study of a direct contact
membrane distillation system coupled to a salt-gradient solar pond for terminal
lakes reclamation. Water Research, Vol.44, No.15, (August 2010), pp. 4601-4615, ISSN
0043-1354.
Suárez, F., Aravena, J.E., Hausner, M.B., Childress, A.E. & Tyler, S.W. (2011). Assessment of
a vertical high-resolution distributed-temperature-sensing system in a shallow
thermohaline environment. Hydrology and Earth System Sciences, Vol.15, No.3,
(March 2011), pp. 1081-1093, ISSN 1027-5606
Tyler, S.W., Burak, S.A., Mcnamara, J.P., Lamontagne, A., Selker, J.S. & Dozier, J. (2008).
Spatially distributed temperatures at the base of two mountain snowpacks
measured with fiber-optic sensors. Journal of Glaciology, Vol.54, No.187, (December
2008), pp. 673-679, ISSN 0022-1430
Tyler, S.W. & Selker, J.S. (2009). New user facility for environmental sensing, EOS,
Transactions, American Geophysical Union, Vol.90, No.50, (December 2009), pp. 483,
ISSN 0096-3941
Tyler, S.W., Hausner, M.B., Suárez, F. & Selker, J.S. (2009a). Closing the energy budget:
advances in assessing heat fluxes into shallow lakes and ponds, EOS, Transactions,
American Geophysical Union, Vol.90, No.52 (Fall Meet. Suppl.), ISSN 0096-3941, San
Francisco, California, USA, December 2009
Tyler, S.W., Selker, J.S., Hausner, M.B., Hatch, C.E., Torgersen, T., Thodal, C.E. & Schladow,
S.G. (2009b). Environmental temperature sensing using Raman spectra DTS fiber-
optic methods. Water Resources Research, Vol.45, No.W00D23, (2009), 11 pp., ISSN
0043-1397
Turner, J.S. (1974). Double-diffusive phenomena. Annual Review of Fluid Mechanics, Vol.6,
(January 1974), pp. 37-54, ISSN 0066-4189
Uchida, Y., Sakura, Y. & Taniguchi, M. (2003). Shallow subsurface thermal regimes in major
plains in Japan with reference to recent surface warming. Physics and Chemistry of
the Earth, Vol.28, No.9-11, (2003), pp. 457-466, ISSN 1474-7065
Vogt, T., Schneider, P., Hahn-Woernle, L. & Cirpka, O.A. (2010). Estimation of seepage rates
in a losing stream by means of fiber-optic high-resolution vertical temperature
profiling. Journal of Hydrology, Vol.380, No.1-2, (January 2010), pp. 154-164, ISSN
0022-1694
Westhoff, M.C., Savenije, H.H.G., Luxemburg, W.M.J., Stelling, G.S., van de Giesen, N.C.,
Selker, J.S., Pfister, L., Uhlenbrook, S. (2007). A distributed stream temperature
model using high resolution temperature observations. Hydrology and Earth System
Sciences, Vol.11, No.4, (2007), pp. 1469-1480, ISSN 1027-5606
Westhoff, M.C., Bogaard, T.A. & Savenije, H.H.G. (2010). Quantifying the effect of in-stream
rock clasts on the retardation of heat along a stream. Advances in Water Resources,
Vol.33, No.11, (November 2010), pp. 1417-1425, ISSN 0309-1708
Winograd, I.J., Landwehr, J.M., Coplen, T.B., Sharp, W.D., Riggs, A.C., Ludwig, K.R. &
Kolesar, P.T. (2006). Devils Hole, Nevada δ18O record extended to the mid-
Holocene. Quaternary Research, Vol.66, No.2, (September 2006), pp. 202-212, ISSN
0033-5894
Yilmaz, G. & Karlik, S.E. (2006). A distributed optical fiber sensor for temperature detection
in power cables. Sensors and Actuators A: Physical, Vol.125, No.2, (January 2006), pp.
148-155, ISSN 0924-4247
32
Prandtl Number Effect on Heat Transfer

Degradation in MHD Turbulent Shear
Flows by Means of High-Resolution DNS
Yoshinobu Yamamoto and Tomoaki Kunugi
Department of Nuclear Engineering, Kyoto University
Japan
1. Introduction
Estimation of the heat transfer degradation effected by Magneto-Hydro-Dynamics (MHD)
forces is one of the key issues of the fusion reactor designs utilized molten salt coolant.
FLiBe which is the molten salt mixture of LiF and BeF, is one of the coolant candidates in the
first wall and blanket of the fusion reactors, and has several advantages which are little
MHD pressure loss, good chemical stability, less solubility of tritium and so on. In contrast,
heat transfer degradation for the high Prandtl number, (Pr=ν/α, Prandtl number, ν is the
kinetic viscosity, α is the thermal diffusivity) characteristics caused by the low thermal
diffusivity and high viscosity (Sagara et al, 1995), was one of the issues of concern.
MHD turbulent wall-bounded flows have been investigated extensively by both
experimental and numerical studies (Blum, 1967, Reed & Lykoudis, 1978, Simomura, 1991,
Lee & Choi, 2001, Satake et al., 2006, Boeck et al, 2007, etc.) and much important information
such as the drag reduction, the turbulent modulation, similarity of velocity profile, and heat
transfer have been obtained.
On the other hands, MHD turbulent heat transfer in a high-Pr fluid has not been understood
well. The previous experimental and direct numerical simulation (DNS) studies still have
conducted for Prandtl number up to Pr=5.7. Therefore, the knowledge of the MHD heat
transfer on higher-Pr fluids such as FLiBe (Pr=20–40), is highly demanded to verify and
validate the MHD turbulent heat transfer models for the fusion reactor designs.
The objective of this study is to perform a direct numerical simulation of MHD turbulent
channel flow for Prandtl number up to Pr=25, where all essential scales of turbulence are
resolved. In this study, we report that the MHD turbulent heat transfer characteristics in
Pr=25 for the first time and discuss that the MHD pressure loss and heat transfer
degradation under the wide-range Pr conditions. The obtained database is of considerable
value for the quantitative and qualitative studies of the MHD turbulent heat transfer models
for the blanket design of a fusion reactor.
2. Target flow field and flow condition

The flow geometry and the coordinate system are shown in figure 1. The target flow fields
are the 2-D fully-developed turbulent channel flows imposed wall-normal magnetic field
and the streamwise and spanwise computational periods (Lx and Lz) are chosen to be 8h and
4h, where h (=Ly/2) denotes channel half height.
y x Δθ=θtop - θbottom=1
By
z θtop
Top wall
Ly=2h θbottom
Bottom wall Flow
Lx=8h
Lz=4h
Fig. 1. Flow geometry and coordinate system
Grid number Resolution

Reτ Ha Pr Nx,Ny,Nz Δx+,Δy+,Δz+
(Mx,Ny,Mz) (temperature)
CASE1
0,8,10,12 0.025
(Lithium)
72,182,72 16,7,0.25-2.0,8.3
CASE2
0,6,8,10,12 5.7
(KOH)
CASE2‘
150 0 5.0 432,182,216 2.8,0.25-2.0,2.8
(fine grid)
CASE3 72,370,72 16.7,0.05-1.0,8.3
0,8,10,12 25.0
(FLiBe) (320,370,160) (3.8,0.05-1.0,3.8)
CASE3‘
0 25.0 648,370,324 1.9,0.05-1.0,1.9
(fine grid)
Table 1. Numerical condition
Duo to the limitation of our utilizable computational resources, turbulent Reynolds number
(Reτ=uτh/ν, uτ: friction velocity) was limited to 150, and three thermal properties of the
Lithium (Pr=0.025), KOH solution (Pr=5.7), and FLiBe (Pr=25) were covered. The KOH
solution was used as the FLiBe simulant fluid in the previous experimental study
(Yokomine et al., 2007) and the Lithium is a typical liquid metal coolant in a blanket of
fusion reactors. To maintain the fully-developed turbulent status, Hartman number
(Ha=By2h(σ/ρν)1/2, By : wall-normal magnetic flux density, σ: electrical conductivity, ρ :
density ) was also limited around 12 in Reτ=150 (Lee & Choi, 2001, Yamamoto et al., 2008).
Numerical conditions are tableted in Table 1. Here, Nx(Δx) ,Ny(Δy), and Nz(Δz) are the grid
numbers (resolutions) in the streamwise, vertical, and spanwise directions, respectively. The
super-script + denotes the nondimensional quantities normalized by the friction velocity,
friction temperature and the kinematic viscosity. Mx and Mz are also the grid numbers in a
horizontal direction temperature as mentioned 3.2, in case of adapting a different grid
resolution for the flow and for the temperature field. In a wall-normal direction, the grid
resolution resolved the Batchelor scale is ensured for all cases.
Prandtl Number Effect on Heat Transfer Degradation in
MHD Turbulent Shear Flows by Means of High-Resolution DNS 639
3. Numerical procedures
3.1 Governing equation and boundary condition
Governing equations of the present DNS are the continuity equation (1), the momentum
equations (2) with the electric field described using the electrical potential approach
(Simomura, 1991), Poisson equation (3) of the electrical potential, and the energy equation (4).
∂ui* (1)
= 0,
∂xi
* *
∂ ui* ∂ ui u j Fi ∂ ⎛ p* ⎞ ∂ 2 ui* σ ⎛ ∂ϕ * ⎞
+ = δi1 − ⎜⎜ ⎟⎟ +ν + Bijk ⎜ − + ε jlmul* Bm ⎟ Bk , (2)
∂t ∂ xj ρ ∂xi ⎝ ρ ⎠ ∂ x j ∂ x j ρ ⎜ ∂ xj ⎟
⎝ ⎠
∂ 2φ *
=
∂
∂xi ∂xi ∂xi
( )
ε ijk u*j Bk , (3)
* *
∂θ * ∂θ u j ∂2 θ *
+ =α . (4)
∂t ∂ xj ∂ x j∂ x j
Here ui and xi are the streamwise (i=1), the vertical (i=2) and the spanwise (i=3) velocity and
direction, respectively. t is time, Fi is the i-th competent mean pressure gradient, p is the
pressure, φ is the electric potential, Bi=(0, By,0) is the Magnetic flux density, and θ is the
temperature. Super script * denotes instantaneous value and δij, εijk (i,j,k=1-3) is the
Kronecker delta and the Levi-Civita symbol, respectively.
Non-slip and periodic conditions are imposed for the boundary conditions of velocities and
the constant temperature at top and bottom boundaries (θtop> θbottom, θtop: top wall
temperature, θbottom: bottom wall temperature), and the periodic conditions are imposed for
the temperature field. In this study, temperature transport is treated as a passive scalar.
The non-conducting conditions of the electric potential are applied to all walls and the
periodic condition imposed on the horizontal directions. Total electric current in the
spanwise flow domain is kept zero.
3.2 Numerical procedures

A hybrid Fourier spectral and the second-order central differencing method (Yamamoto et
al, 2009) is used for the computations. The spectral method is used to compute the spatial
discretization in the stream (x) and spanwise (z) directions. Nonlinear terms are computed
with 1.5 times finer grids in horizontal (x and z) directions to remove the aliasing errors
(Padding method). The derivative in the wall normal (y) direction is computed by a second-
order finite difference scheme at the staggered grid arrangement (Satake et al, 2006). Time
integration methods of the governing equations are the 3rd-order Runge-Kutta scheme for
the convection terms, the Crank-Nicolson scheme for the viscous terms and the Euler
Implicit scheme for the pressure terms, respectively. The Helmholtz equation for the viscous
(diffusion) terms and the Poisson equations of the pressure and the electrical potential are
solved by a Tri-Diagonal Matrix Algorithm, TDMA in Fourier space.
In DNS of the flow field, the Kolmogorov length scale has to be resolved. On the other
hands, the length scales of the high-Pr temperature field are smaller than the smallest length
scales of the velocity fields (Batchelor, 1959). To reduce the numerical costs in DNS of the
high-Pr fluids, a different number of grid resolutions in the horizontal direction for velocity
and temperature fields is adapted. In computing the temperature convection terms in (4)
pseudo-spectrally, the grid points of velocities were expanded to the same grid points of the
high-Pr temperature, as follow,
⎪ ( kx , y , kz ) kx ≤ π N x / Lx , kz ≤ π N z / Lz
⎧Φ
Φ( k x , y , k z ) = ⎨ , kx ≤ π Mx / Lx , kz ≤ π M z / Lz . (5)
⎪⎩0 otherwise
Here, Φ denotes the velocities in Fourier space, kx and kz are the wavenumbers in the
streamwise and spanwise directions, respectively. The phase-shift method (Patterson &
Orszag, 1971) is used to remove the aliasing errors derived from the temperature convection
terms. As a consequence, the grid size corresponded to the Batchelor length scale is retained
for Pr=25.
Present DNS were calculated by using the T2K Open Supercomputer at Kyoto University.
Elapsed time per one time step was about 1.2 [s] when using 8nodes (128cores) in CASE3.
4. Validation of present DNS

At the beginning of this study, we demonstrate the adequacy of the present DNS.
3
-1
Ha=0 10
Ha=8
-2
Ha=12 10
2
Euu+
u rms+
-3
10
-4
1 10 y+=149
Ha= 0
-5
10 Ha= 8
Ha=12
-6
0 10 0 1
0 50 + 100 150 10 10
y kxh
(a) Streamwise turbulent intensity, (b) Streamwise energy spectra,
Fig. 2. MHD suppression effects on turbulence
Figures 2 shows the turbulent intensities and the streamwise energy spectra at the channel
center in Ha=0, 8, and 12. As well as the previous study (Lee & Choi, 2001), turbulent
intensity was suppressed with increase of Ha as shown in Fig. 2-(a). Figure 2-(b) gives
evidence that turbulent suppression effects can be remarkable in the high wave-numbers
turbulence. It is clear that the effects of the grid dependency would be the biggest in Ha=0.
Therefore, the convergences of the grid tendency were investigated in Ha=0, by using the
DNS data fully-resolved the Batchelor length scale for Pr=5 or 25 in Ha=0 as tabled CASE2’
and CASE3’ in Table 1.
4.1 Medium high-Pr case

According to Na & Hanratty, 2000, the use of a higher resolution in horizontal direction
does not produce significant changes to the first-order statistics from Pr=1 to 10. In this
study, we investigated the grid dependency effects on the higher-order statistics such as the
energy dissipation (=ε) and temperature energy dissipation (=εθ).
0.2 0.8
Pr=5.0, Ha=0 Pr=5.0, Ha=0
0.6
εθ+
+
0.1
ε
0.4
+
Δx+=2.8 +
Δx =16.8 0.2 Δx =2.8
+
+
Δx =20.0 Δx+=16.8
Δx =20.0
0 1 0
2
10 10 1 2
+ 10 10
y +
y
(a)energy dissipation, (b) temperature energy dissipation,
Fig. 3. Grid dependency on high-order statistics in medium high-Pr fluid
Figures 3 show the energy dissipation and temperature energy dissipation for Pr=5 with
change of the horizontal resolutions. The required horizontal resolution for the reproductively
of the energy dissipation and temperature dissipation, was estimated as Δx+=16.7, and
Δz+=8.3 in this medium high-Pr fluid.
4.2 High-Pr case

For Pr=25, DNS in Reτ=180 were conducted by means of a hierarchical algorithm in which
only the scalar fields were solved on the grid dictated by the Batchelor scale (Schwertfirm
&Manhart, 2007). However, the validation by using the different resolution for flow and
high-Pr temperature field has not been reviewed yet. In this study, the adequacy of DNS by
using a different resolution for flow and high-Pr temperature field is verified compared
with DNS data fully-resolved the Batchelor length scale in the same grid size for flow and
temperature.
(a) (b) (c)

Fig. 4. Flow visualization, CASE3, Ha=0, Pr=25, y+=149. (a) streamwise turbulent velocity,
-2 (black) < u+ < 2.0 (white), (b) turbulent temperature (coarse grid), -0.15 (black) < θ/Δθ <0.15
(white) (c) turbulent temperature (fine grid), -0.15(black)<θ/Δθ<0.15 (white)
Figure 4 shows the flow visualization in results of Ha=0, Pr=25. In this case, 72x72 grids for
flow (in Fig.4-(a)), 72x72 grids for the temperature filed (in Fig.4-(b)) and 320x160 grids for
the temperature field (in Fig.4-(c)), were used in horizontal directions, respectively. Despites
of the high wave-number flow fluctuations, the high wavenumber temperature fluctuation
can be computed as shown in Fig.4-(c).
Figure 5-(a) shows the temperature energy dissipation for Pr=25 with change of the
horizontal resolutions. The required horizontal resolution for the reproductively of the
temperature dissipation, was estimated as Δx+=8.3, and Δz+=4.2. This grid resolution is
equivalent to twice as high for Pr=5; it is proportional to square root of the Pr ratio
(=(25/5)1/2). The effects of using the different resolution for flow and temperature cannot be
found even in the temperature energy dissipation.
Figure 5-(b) shows the streamwise energy spectra near channel center for Pr=25. Compared
with CASE3 and CASE3’, there is ninefold grid resolution in flow, but the variance of the
spectra profile cannot be observed in this high-Pr temperature field. This indicates that the
high wave-number velocity fluctuations less than the Kolmogorov scale can be ignored in a
high-Pr passive scalar transport. As a consequence, we verify the adequacy of DNS by using
the different resolution for flow and high-Pr temperature field and numerical cost in DNS of
high-Pr fluids can be substantially reduced.
3
-3
Pr=25.0, Ha=0 10
2
2
Euu , Eθθ/Δθ
-8
10
εθ+
1 -13 Euu+, CASE3'

Δx+=8.3
Δx+=1.9
10 2
Eθθ/Δθ , CASE3'
+
Δx =16.8 Euu+, CASE3
Δx+=16.8 (temp. fieldΔx+ =3.8) -18 Eθθ/Δθ 2, CASE3
0 10
10 1 10 2 10
0
10
1
10
2
y+ kxh
(a) (b)
Fig. 5. Grid dependency and validation of different grid resolution for flow and high-Pr
temperature field. (a) Temperature dissipation, (b) Streamwise energy spectra, streamwise
velocity and temperature
5. MHD pressure loss and heat transfer

In this study, the friction drag confident (Cf) and Nusselt number (Nu) at the wall were
expressed by
Cf=2uτ2/Ub2, (6)
Nu=2h(dΘ/dy)wall/Δθ. (7)
Here, Ub and (dΘ/dy)wall denotes the bulk mean velocity and mean temperature gradient at
the wall.
Figure 6-(a) shows the friction drag coefficient as a function of the interaction parameter N
(=Ha2/Reb, Reb: Bulk Reynolds number=Ub2h/ν), where the friction drag coefficients were
normalized by that in Ha=0. The friction drag coefficients were monotonically decreased
with increase of Ha; MHD pressure loss is less than the turbulent drag reduction effected by
MHD. Therefore, all MHD cases of this study might be considered in a turbulent-laminar
transition status. We need the DNS data in more higher Re to discuss the general
relationships between MHD pressure loss and MHD turbulent drag reduction in turbulent
condition.
Figure 6-(b) shows the Nusselt number as a function of N, where the Nusselt number were
also normalized by that in Ha=0. Maximum heat transfer degradation in the low-Pr fluid
was no more than 5% of the non-MHD condition. The usability of a low-Pr fluid was no
doubt about heat transfer, however, Ha of Lithium was 700 times as large as one of FLiBe in
the same Reynolds number (Re) and magnetic flux density (By) conditions.
1.05 1
0.9
Nu/Nu Ha=0
Cf/Cf Ha=0
1
0.8
Pr=5.7
0.7 Pr=25
0.95
Pr=0.025
0.6
0.9 0.5
0 0.01 0.02 0.03 0 0.01 0.02 0.03
2
N=Ha /Re b 2 N=Ha /Re b
(a) friction drag coefficient, (b) Nusselt number,

Fig. 6. Friction drag coefficient and Nusselt number as a function N
On the other hands, heat transfer degradation in the high-Pr fluids (Pr=5.7 and 25) reached
up to 30% without depending on Pr. This indicated that similarity of heat transfer
degradation in high-Pr MHD flows might be existed.
1
δHa=0 /δ, Nu/Nu Ha=0
0.9
0.8 δHa=0 /δ, Pr=5.7

Nu/NuHa=0 Pr=5.7
δHa=0 /δ, Pr=25
Nu/Nu Ha=0 Pr=25
0 5 10 15
Ha
Fig. 7. Thermal viscosity thickness and Nusselt number as a function N
Figure 7 shows the thermal viscosity thickness (δ) and Nusselt number as a function of Ha in
the high-Pr fluids, where thermal viscosity thickness was defined as
δ=y+ at Θ+=0.99Pry+. (8)

Thermal viscosity thickness was normalized by those in Ha=0. Heat transfer degradation
was strongly correlated with change of the thermal viscosity thickness without depending
on Pr.
6. Turbulence statistics
Figure 8 shows the profiles of temperature turbulent intensities for Pr=5.7 and 25. With
increase of Ha, the peak position of turbulent intensity was shifted to the channel center side
and the scale of it was decreased in both cases. In either case, the peak position was located
below the wall-normal height y+=15; thermal boundary layers for Pr=5.7 and 25 were
thinner than the velocity boundary layer in the present MHD conditions.
15 40
Pr=5.7 Pr=25.0
Ha=0 Ha=0
Ha=8.0 Ha=8.0
Ha=12.0 30
10 Ha=12.0
θrms+
θrms+
20
5
10
0 0 1 2 0 -1 0 1 2
10 10 10 10 10 10 10
y+ y+
(a) Pr=5.7, (b) Pr=25,
Fig. 8. Turbulent temperature profiles
Pr=25.0, Ha=12.0
Pr=25, Ha=0
5 5
Gain
Gain
0 0
Loss
Loss
-5 0 1 2
-5 0 1 2
10 10 10 10 10 10
y+ y+
Production εθ Turbulent diff. Viscous diff. Residual
(a) Ha=0, Pr=25, (b) Ha=12, Pr=25,
Fig. 9. Budget of turbulent temperature energy
0.1 0.1
Loss
Loss
0.05 0.05
0 0
Gain
Gain
-0.05 -0.05
-0.1 -0.1
10 0 10 1 10 2 10 0 10 1 10 2
y+ y+
Budget of vθ Budget of vθ
Production TPG Turbulent diff. Viscous diff. εθv Residual

(a) Ha=0, Pr=25, (b) Ha=12, Pr=25,
Fig. 10. Budget of wall-normal turbulent heat flux
Figures 9 and 10 show the budget of turbulent temperature energy (Kθ) and wall-normal
turbulent heat flux(vθ) for Pr=25, Ha=0 and 12. Transport equations (9) and (10) of turbulent
temperature energy and wall-normal turbulent heat flux are expressed by
2
∂Θ 1 ∂ θθ v ∂ 2 Kθ ⎛ ∂θ ⎞
0 = − vθ − + α − α⎜ ⎟ , (9)
∂y 2 ∂y ∂y 2 ⎜ ∂ xj ⎟

⎝

⎠
Production Turbulent diff. Viscous diff. Dissipation:εθ
∂Θ ∂ θ vv ∂ ⎛ ∂v ∂θ ⎞ ∂θ ∂ pθ ∂θ ∂v .
0 = − vv − + ⎜⎜ν θ +α v ⎟ + p − − (ν + α ) (10)
∂y ∂y ∂y ⎝ ∂y ∂y ⎟⎠ ∂y ∂y ∂x j ∂x j

Production Turbulent diff. Viscous diff. Temp.Press-Grad. Dissipation
Here, over bar denotes quantities estimated by ensemble average. In Fig. 9-(a), around the
thermal buffer region (y+=5), both diffusion terms of turbulent and viscous exceeded
dissipation (εθ) term. Predominance of the diffusion terms in the high-Pr fluids (Pr>10) was
confirmed in the previous DNS (Schwertfirm &Manhart, 2007). In Ha=12, predominance of
diffusion terms was observed more clearly as shown in Fig. 9-(b). As well as turbulent
temperature energy, turbulent diffusion term in Fig. 10-(b) was dominant at y+=15-30 in
Ha=12, however, the predominance of viscous diffusion term was indistinct. Compared
with no-MHD case in Fig. 9-(a), the damping of turbulent diffusion term was small but the
others were suppressed by the MHD effects; effects of turbulent diffusion on the MHD heat
transfer were relatively larger with increase of Ha. These indicate that a sensitive model of
the turbulent diffusion would be required in the prediction of MHD heat transfer in high-Pr
fluids.
Figure 11 shows the turbulent Prandtl number (PrT) profiles for Pr=5.7 and 25. Turbulent
Prandtl number was defined as
PrT = uv / vθ . (11)
1.2 2
Pr=25.0
Ha= 0.0
1 Ha= 8.0
Ha=12.0
1.5
PrT
PrT
0.8
Pr=5.7
0.6 Ha= 0.0 1
Ha= 8.0
Ha=12.0
0.4 0.5
0 1 2 0 1 2
10 10 10 10 10 10
+
y+ y
(a)Pr=5.7, (b) Pr=25,
Fig. 11. Turbulent Prandtl number profiles
Na & Hanratty, 2000 and Schwertfirm & Manhart, 2007 pointed out that turbulent Prandtl
number close to the wall increases with increase of Pr. The turbulent Prandtl number
profiles in the non-MHD case were good agreements with the results of Schwertfirm &
Manhart, 2007, however, profiles in MHD case was decreased close to the wall for Pr=5.7
and 25 with increase of Ha. In Ha=12, the values of the turbulent Prandtl number in the
vicinity of the wall fell into 1 for Pr=5.7 and 25. It was suggested that there was no MHD
terms in balance of the heat transfer equation; turbulent effect on heat transfer might exceed
that on momentum transfer as the limiting case of a turbulent-laminar transition status in
Ha=12.
Figure 12 shows the time scale ratio for Pr=5.7 and 25. In non-MHD flow, time scale ratio
had the weak peak at the buffer region for Pr=25 and 49 (Schwertfirm & Manhart, 2007
pointed out that). Time scale ratio profiles in MHD cases clearly had the peak in increase of
Ha for Pr=5.7 and 25. At the buffer region, MHD effects on heat transfer might to be
corresponded to the heat transfer in a higher-Pr fluid as shown in Figs. 9 and 12. However,
these close to the wall might act on like a lower-Pr fluid as shown in Fig. 11.
1.5 1.5
Ha= 0.0 Pr=25.0
Ha= 8.0 Ha= 0.0
(Kθ /εθ )/(Pr*K/ε)
(Kθ /εθ )/(Pr*K/ε)
Ha=12.0 Ha= 8.0

1 Ha=12.0
1
0.5 0.5
0 0 1 2
0 0 1 2
10 10 10 10 10 10
+
y y+
(a) Pr=5.7, (b) Pr=25,
Fig. 12. Time scale ratio profiles
Since both turbulent Prandtl number and time scale ratio were one of the dominant
parameters in turbulent heat transfer modeling, change of profiles in increase of Ha might
be caused the aggravation of the prediction accuracy.
7. Conclusion
In this study, direct numerical simulation of MHD turbulent channel flow for Prandtl
number up to Pr=25 were performed. The adequacy of the present DNS data was verified by
comparison with the DNS data fully-resolved the Batchelor length scale. As the results, the
MHD turbulent heat transfer characteristics in Pr=25 were reported for the first time.
Maximum heat transfer degradation in the low-Pr fluid was no more than 5% of the non-
MHD condition. On the other hands, heat transfer degradation in the high-Pr fluids (Pr=5.7
and 25) reached up to 30%. The similarity of heat transfer degradation in high-Pr MHD
flows seemed be existed.
On the MHD heat transfer in high-Pr fluids, effects of turbulent diffusion were relatively
larger. Turbulent Prandtl number and time scale ratio were considerably changed with
increase of Ha.
The scaling of MHD heat transfer in high-Pr fluids was not understood yet. For the high-Ha
and Reτ condition (Ha>5, Reτ>250), Boeck et al. 2007 reported the similarity of MHD mean
velocity profiles on the parameter R(: Hartmann Reynolds number). To discuss the scaling
of MHD heat transfer, we need DNS data of higher-Re and Ha conditions. In such cases,
present DNS procedure by using a different resolution for flow and high-Pr temperature
field will demonstrate a great advantage.
8. Acknowledgment
Present DNS were conducted by using the T2K open supercomputer at ACCMS and IIMC,
Kyoto University. This study was supported by the Global COE program “Energy Science in
the Age of Global Warming” and a Grant-in-aid for Young Scientists (B), KAKENHI
(21760156) MEXT, Japan.
9. References
Batchelor, G.K., (1959), Small-scale variation of convected quantities like temperature in
turbulent fluid. Part 1. General discussion and the case of small conductivity,
Journal of Fluid Mechanics, Vol.5, pp.113–133.
Blum, E.YA., (1967), Effect of a magnetic field on heat transfer in the turbulent flow of
conducting liquid, High Temperature,Vol. 5,pp. 68-74.
Boeck, T., Krasnov, D., and Zienicke, E., (2007), Numerical study of turbulent
magnetohydrodynamic channel flow, Journal of Fluid Mechanics, Vol.572, pp.179-
188.
Lee, D. & Choi, H., (2001), Magnetohydrodynamic turbulent flow in a channel at low
magnetic Reynolds number, Journal of Fluid Mechanics, Vol.429, pp.367–394.
Na, Y. & T.J. Hanratty, T.J., (2000), Limiting behavior of turbulent scalar transport close to a
wall, International Journal of Heat and Mass Transfer, Vol.43 , pp.1749-1758.
Patterson, S. & Orszag, S.A., (1971), Spectral calculations of isotropic turbulence: Efficient
removal of aliasing interactions, Physics of Fluids, Vol.14, pp.2538-2541.
Reed, C.B. & Lykoudis, P.S., (1978), The effect of a transverse magnetic field on shear
turbulence Journal of Fluid Mechanics, Vol.89, pp.147-171.
Sagara. A., Motojima, O., Watanabe, K., Imagawa, S., Yamanishi, H., Mitarai, O., Sato, T.,
Chikaraishi, H. and FFHR Group, Design and development of the Flibe blanket for
helical-type fusion reactor FFHR, Fusion Engineering and Design, Vol.29, pp.51-56.
Satake, S., Kunugi, T., Takase, T., and Ose, Y., (2006), Direct numerical simulation of
turbulent channel flow under a uniform magnetic field for large-scale structures at
high Reynolds number, Physics of Fluids, Vol.18, 125106.
Schwertfirm, F. & Manhart, M., (2007), DNS of passive scalar transport in turbulent channel
flow at high Schmidt numbers, International Journal of Heat and Fluid Flow, Vol.28,
pp. 1204–1214.
Simomura Y., (1991), Large eddy simulation of magnetohydrodynamic turbulent channel
flows under a uniform magnetic field, Physics of Fluids A 3, pp.3098-3106.
Yamamoto, Y.,Kunugi, T., Satake, S., and Smolentsev, S. (2008), DNS and k–ε model
simulation of MHD turbulent channel flows with heat transfer, Fusion Engineering
and Design, Vol.83, pp.1309-1312.
Yokomine, T., Takeuchi, J., Nakaharai, H., Satake, S., Kunugi, T., Morley, N.B., and M. A.
Abdou, M.A., (2007), Experimental investigation of turbulent heat transfer of high
Prandlt number fluid flow under strong magnetic field, Fusion Science and
Technology, Vol.52, pp.625-629.
33
Effective Method of
Microcapsules Production for Smart Fabrics
Luz Sánchez-Silva, Paula Sánchez and Juan F. Rodríguez
Department of Chemical Engineering/University of Castilla-La Mancha
Spain
1. Introduction
Nowadays, the attempts in the textile and clothing industry have moved towards more
innovative and high quality products in order to differentiate themselves and be more
competitive. The new demand for innovative textiles is increasingly oriented to match
material innovation, new technologies and fashion. The new products are not only different
for their lines, patterns and volumes but also for what they can do. Recently, microcapsules
have been applied to many functional and technical textiles. Examples of this include
fragrances, aromatic deodorants, cosmetics, insect repellents, antibiotics, polychromic, drug
delivery for medical textiles and thermo-regulating systems. This kind of fabrics, that
introduces new functionalities without affecting the look and feel of the textile, is commonly
denominated as smart textiles.
Microencapsulated phase change materials (PCM) can be incorporated into textile structures
to produce fabrics of enhanced thermal properties. A thermo-regulating fabric is an intelligent
textile that has the property of offering suitable response to changes in external temperature
changes or to external and environmental stimuli. The level of thermal comfort depends on
the heat exchange between the human body and the environment that surrounds it.
Microcapsule production may be achieved by means of physical and chemical techniques.
The use of some techniques has been limited to the high cost of processing, regulatory
affairs, and the use of organic solvents, which are concern for health and the environment.
In this way, a method based on a suspension like polymerization process for the encapsulation
of phase change materials has been selected. This PCM encapsulation method is simply,
inexpensive and technically easy.
Suspension like polymerization involves the dispersion of the monomer or monomers,
mainly as a liquid in small droplets, into an agitated stabilizing medium consisting of water
containing small amounts of suspension agents and without using aqueous phase inhibitors
of secondary nucleation or modifiers. The initiator is dissolved in the monomer-PCM
mixture and PCM material does not take part on the polymerization kinetic. In the proper
conditions the polymer reacts mainly in the interface of the drop forming a shell around the
PCM core since this interface is the only locus of polymerization. However, not all the
recipes and conditions for suspension polymerization favour the formation of the polymer
at the PCM/water interface all around the drop as desired and microparticles without
complete phase separation into capsules are obtained.
Well-known PCM are linear chain hydrocarbons known as paraffin waxes (or n-alkanes),
hydrated salts, polyethylene glycols (PEGs), fatty acids and mixture or eutectics of organic
and non-organic compounds. PCM materials absorb energy during the heating process as
phase change takes place and release energy to the environment in the phase change range
during a reverse cooling process.
The required properties for a phase change materials depend on their specific application in
textile fields. A wide spectrum of phase change materials are available with different heat
storage capacity and phase change temperature. Different types of commercial PCMs can be
encapsulated by means of suspension polymerization process. Rubitherm® RT20,
Rubitherm® RT27, Rubitherm® RT31, Petrepar® n-C14 and Petrepar® n-C-13 have
demonstrate their capability to be encapsulated and their thermal abilities to absorb and
release energy. Their physical and chemical properties make them very attractive for
thermal storage.
Thermal properties, air permeability, moisture vapour permeability and moisture regain of
materials also influence the heat balance of the body and, consequently, affect clothing
comfort (Ren & Ruckman, 2004). The incorporation of PCM microcapsules to textiles can
affect other comfort-related properties and hand of the materials adversely, especially when
the topical application of microcapsules results in drastic changes in the surface
characteristics of materials. The extent of change in these properties depends on the loading
amount of PCM microcapsules (Shin et al., 2005).
Several methods of incorporating PCM microcapsules into a fibrous structure have been
developed. The microcapsules can be applied by stamping works, exhaustion dyeing,
impregnation, spraying and coating or by direct incorporation in the fibre without highly
modifying its touch and colour (Monllor et al., 2009; Dixit & Goal, 2007; Rodrigues et al.,
2009). In previous applications of PCM technology in the textile industry, for garments and
home furnishing products, microencapsulated PCM were incorporated into acrylic fibers
(Bryant & Colvin, 1988) or polyurethane foams (Colvin & Bryant, 1996) or were embedded
into a coating compound and topically applied to a fabric (Bryant & Colvin, 1994). Shin et
al., (2005) incorporated melamine-formaldehyde microcapsules containing eicosane on
polyester knit fabrics by means of a pad dry cure method with a polyurethane binder.
Mengjin et al., (2008) developed a new kind of thermo-regulating fiber based on PVA and
paraffin. Furthermore, Onder et al., (2008) studied the microencapsulation of three types of
paraffin waxes by complex coacervation to improve thermal performances of woven fabrics.
Recently, Koo et al. (2009) have attempted to demonstrate the application of PCM
microcapsules on waterproof nylon fabrics and to enhance thermal insulation effect with
ceramic materials (SiC) by using a dual coating method.
Binders play a crucial role in microcapsule coating formulation for various textile materials,
as they are required to fix microcapsules on textile supports permanently. To a large extend,
binders determine the quality, durability and washability of textile materials with
microencapsulated ingredients. Some of the most frequently used binders in textile are
water-soluble polymers, such as starch and modified starches, carboxymethyl cellulose;
synthetic latexes, such as styrene-butadiene, polyvinylacetate or acrylate latexes; and
aminoaldehyde resins (Boh & Knez, 2006).
In our previous work, the fixation of PCM microcapsules containing paraffin with a melting
point around 40ºC, into a cotton textile substrate by means of a coating technique were
carried out. Furthermore, the influence of different coating formulations and mass ratio of
microcapsules to coating formulation were evaluated in order to obtain an adequate textile
Effective Method of Microcapsules Production for Smart Fabrics 651
with thermo-regulating properties (Sánchez et al., 2010). The coating fabric with 35 wt.% of
microcapsules added related to commercial coating binder (WST SUPERMOR®) showed a
energy storage capacity of 7.6 J g−1, a high durability and an adequate stability after
washing, rub fastness and ironing treatments. A difference of 8.8°C for 6 s was observed for
textiles with thermo-regulating properties in comparison with a coated one without
microcapsules. The different application areas of textiles with thermo-regulating properties
imply the fixation to very different substrates. In this sense, there are few references in the
literature studying the influence of the kind of textile on the fixation of microcapsules (Koo
et al., 2009). In addition, the PCM microcapsules incorporation could degrade the original
functionalities of the textile such as soft touch, vapor or moisture permeability and wearing
comfort.
The aim of this work was to investigate the production of textiles with thermo-regulating
properties by using PCM microcapsules and a coating technique. The influence of the type
of used PCM on the heat capacity of microcapsules, the particle size distribution (PSD) and
the microcapsules yield of each experiment was studied. On the other hand, different type of
textile substrates depending on the field of their textile applications (apparel, blankets,
insulation, protective clothing) were evaluated. Furthermore, a study of thermoregulatory
effect of the coating fabrics produced was carried out using an infrared thermography
camera. Thermal properties of textile samples were examined by Differential Scanning
Calorimetry (DSC). Furthermore, Environmental Scanning Electron Microscopy (ESEM) and
Optical Microscopy (OM) techniques were used to check the presence, surface distribution,
preferred join position and to analyse the structure of microcapsules into the textile.
2. Experimental
2.1. Microcapsules synthesis
Styrene (99 wt.%) of reagent grade (Merck Chemical) previously purified by washing with
sodium hydroxide and dried with calcium chloride was used as the monomer. Benzoyl
peroxide (97 wt.%) was used as initiator (Fluka Chemical). PRS® paraffin wax, Rubitherm®
RT20, Rubitherm® RT27, Rubitherm® RT31, Petrepar® n-C14 and Petrepar® n-C-13 were
used as core materials. Polyvinylpyrrolidone (K30, Mw 40,000 gmol-1) of reagent grade
(Fluka Chemical) was used as stabilizer and methanol to pour the samples. All these
reagents were used as received. Water was purified by distillation followed by deionization
using ion-exchange resins. Nitrogen was of high-purity grade (99.999%).
A suspension like polymerization process was used for the microcapsules synthesis. A
tubular type Shirasu porous glass membrane was used for a better control of microparticle
size. Details of the synthesis process were previously described elsewhere (Sánchez et al.,
2008b).
2.2 Preparation of textiles with thermo-regulating properties

Microcapsules were fixed into seven fabrics by means of a coating technique, using a
motorized film applicator from Elcometer model 4340 according with ASTM D-823C (ASTM
D-823-C, 1997). WST SUPERMOR® (supplied by Minerva Color Ltd.) were used as
commercial coating binder. In a previous study (Sánchez et al., 2010), this binder was
selected due to allow an efficient fixation of the PCM microcapsules on the fabrics. Every
sample had 200 mm of wide and 290 mm of length due to requirements of the motorized
film applicator.
The coating formulation consisted of WST SUPERMOR® commercial binder and

Rubitherm® RT31 microcapsules (35 wt. % of the coating mixture).
The textile substrate was set on the motorized film applicator surface assuring the fabric
with clips. In this study, the thickness selection of the coating layer was 0.1 mm to obtain a
high thermal storage. The position of the motorized film applicator and the selection
thickness was carried out manually. A dragging speed of 5 mm s-1 was chosen to allow a
homogeneous coating along the film applicator.
Finally, the coated fabric was cured at 95 ºC for 11 minutes.
2.3 Characterization
2.3.1 Differential Scanning Calorimetry (DSC)
Measurements of melting point and latent heat storage capacities of different materials were
performed in a differential scanning calorimetry model DSC Q100 of TA Instruments
equipped with a refrigerated cooling system and nitrogen as the purge gas. Measurements
were carried out in the temperature range from -30ºC to 80ºC with heating and cooling rate
of 10 ºCmin-1.
Various samples of each experiment were analyzed at least three times and the average
value was recorded. DSC analyses of coating textiles from random areas were done.
Furthermore, the encapsulation ratio of the different PCM in the microcapsule was
calculated with the following equation based on enthalpy values:
% PCM content by weight = (ΔHm / ΔHpcm) x 100% (1)

where ΔHm is the enthalpy for the analysed microcapsules (Jg-1) and ΔHpcm is the enthalpy
of pure PCM.
In order to determine the thermal stability of the reversible phenomena of phase change, the
coated textiles were subjected to repeated cycles of melting and crystallization.
2.3.2 Environmental scanning electron microscopy (ESEM)

ESEM was used to analyze the morphological structure of the microcapsules and the
fixation and integrity of PCM microcapsules into the coating textile substrates. Textile
samples were observed by using XL30 (LFD) ESEM with a wolfram filament operating at a
working potential of 20 kV.
2.3.3 Calculation of number-average diameter and volume-average diameter

Particle size and particle size distribution (PSD) of microcapsules were determined on a
Malvern Mastersizer Hydro 2000 SM light scattering apparatus with dilute dispersions of
the particles in methanol.
2.3.4 Infrared thermography

The temperature distributions of the coated textiles with thermo-regulating properties were
evaluated by means of an infrared and visible camera Fluke Ti25. This dispositive allows to
obtain thermal and visual images in the range of temperatures from -20ºC to 250ºC with a
precision of ±2 ºC. The screen was observed from a distance of 30 cm at 24ºC. Images were
downloaded using Fluke SmartViewTM software for analysis. The coated fabrics were pre-
heated at 60ºC, time considered as zero, and then cooled to room temperature.
Thermal human comfort in summer conditions was tested, recording images from 25ºC to
outside temperature (35ºC), comparing a reference textile with a prototype textile with
thermo-regulating properties in contact with the body (shoulders in this specific case).
3. Results
3.1 Microencapsulation of different type of phase change materials (PCM)
The required properties for a phase change materials depend on their specific application in
textile fields. A wide spectrum of phase change material is available with different heat
storage capacity and phase change temperature. In this study different type of commercial
PCM were assayed in order to know what PCM materials are suitable to be encapsulated by
means of suspension polymerization. Thus, Rubitherm® RT20, Rubitherm® RT27,
Rubitherm® RT31, Petrepar® n-C14 and Petrepar® n-C-13 were assayed due to their physical
and chemical properties are very attractive for thermal storage (Table 1). All of them are
saturated hydrocarbons
Molecular Latent heat Melting Viscosity

PCM weight of fusion temperature (mm2s-1)
(gmol-1) (Jg-1) (ºC) at 98ºC
PRS® paraffin wax 168-240 206.8 40-45 2.43
Rubitherm® RT31 268 199.3 31 2.07
Rubitherm® RT27 258 214.6 28 1.64
Rubitherm® RT20 244 177.7 22 1.52
Petrepar® n-C14 198.4 225.0 3-7 0.99
Petrepar® n-C13 184.4 134.4 (-7)-(-5) 0.88
Table 1. Properties of different types of PCMs investigated

Figure 1 shows the particle size distributions (PSDs) in volume (Figure 1a) and in number
(Figure 1b) of microcapsules obtained after the polymerization process using these PCM
materials. It can be seen from Figure 1a that Petrepar® n-C14 and n-C13 exhibit bimodal
PSDs with particles sizes smaller than 115 μm. However, PRS® paraffin wax, Rubitherm®
RT31, RT27 and RT20 shows unimodal PSDs ranging in the interval between 149 to 251 μm.
In all experiments a big difference between the average particle size in volume and in
number was observed due to the heterogeneous sizes of obtained microcapsules. This
behaviour was reported in previous works (Sánchez et al., 2007; Sánchez et al., 2008a;
Sánchez et al., 2008b).
Table 2 reports average diameters (dp0.5) in volume and in number, storage energy
capacities and amount of PCM encapsulated of microcapsules produced using different
PCMs. The mean diameter in number of the microcapsules increases as described:
Rubitherm® RT27>PRS® paraffin wax>Rubitherm® RT31> Petrepar® C-14>Petrepar®
C-13>Rubitherm® RT20. However, the average diameter in volume increases in the following
way: Rubitherm® RT27>Rubitherm® RT31>Rubitherm® RT20>PRS® paraffin wax>Petrepar®
n-C14>Petrepar® n-C13. Therefore, the average diameter of the microcapsules depends on
phase change materials encapsulated. For PRS® paraffin wax, Petrepar® C-14 and Petrepar®
C-13, the average diameter of microcapsules enlarged with increasing the viscosity. This fact
could be due to the polymerization system is subject to nonnegliglible effects of the viscosity
ratio (dispersed phase/continuous phase viscosity). According to Hamielec & Tobita, (1992)
in suspension polymerization an increase of the forming droplets viscosity and the density
difference between the phases results in a tendency towards settling. However, this trend
was not observed for Rubitherm materials.
10
8
Volume (%)
0
1 10 100 1000 10000
Particle size (μm)
(a)
25
20
Number (%)
15
10
0
0.1 1 10 100 1000
Particle size (μm)
Rubitherm RT31 Rubitherm RT27 Rubitherm RT20 Petrepar n-C14 Petrepar n-C13 PRS paraffin
(b)
Fig. 1. Particle size distribution for microcapsules obtained using different PCM: (a) in
volume and (b) in number
PRS ® paraffin wax
Petrepar ® n-C14
Transmittance
Petrepar ® n-C13
3,600 3,100 2,600 2,100 1,600 1,100 600

Wavenumber (cm-1)
(a)
Rubitherm ® RT27
Transmittance
Rubitherm ® RT31
Rubitherm ® RT20
3,600 3,100 2,600 2,100 1,600 1,100 600

-1
Wavenumber (cm )
(b)
Fig. 2. FTIR spectra of: (a) PRS® paraffin wax, Petrepar® C-14 and Petrepar® C-13 and (b)
Rubitherm® RT27, Rubitherm® RT31 and Rubitherm® RT20
This behaviour can be attributed to chemical nature of these commercial PCMs according
to the FTIR spectra of the different PCMs obtained (Figure 2). In the spectrums of the
PRS® paraffin wax, Petrepar® C-14 and Petrepar® C-13 materials (Figure 2b), in contrast to
that of the Rubitherm products (Figure 2a), medium intensity bands in the 980-1,240 cm-1
were observed. These bands indicate the presence of ethyl and propyl groups branches.
Therefore, PRS® paraffin wax, Petrepar® C-14 and Petrepar® C-13 are not straight chain
alkanes.
dpv0.5 dpn0.5 Thermal storage energy PCM encapsulated

PCM
(μm)1 (μm)2 (Jg-1) (wt. %)
PRS® paraffin wax 149.0 4.53 104.7 50.1
Rubitherm® RT31 203.9 4.01 98.7 49.5
Rubitherm® RT27 250.9 6.02 100.2 46.7
Rubitherm® RT20 180.1 2.86 68.3 38.5
Petrepar® n-C14 110.7 3.97 79.0 35.1
Petrepar® n-C13 93.9 3.83 58.6 43.6
1 dpv0.5 represents 50% microcapsule particles whose mean volumetric diameter is less than this value.
2 dpn0.5 represents 50% microcapsule particles whose mean numeric diameter is less than this value.
Table 2. Average diameters and thermal properties of microcapsules containing different

PCM
DSC measurements confirm that all PCMs studied have been successfully encapsulated
inside the polystyrene microcapsules (Table 2). PRS® paraffin wax, Rubitherm® RT27 and
Rubitherm® RT31 allow to obtain microcapsules with energy storage capacities higher than
98 Jg-1 and approximately 50 wt.% of encapsulation efficiency.
Figure 3 shows the environmental scanning electron microscopy (ESEM) micrographs of
microcapsules prepared with different phase change materials. It can be seen from Figure 3
that microcapsules prepared with PRS® paraffin wax, Petrepar® C-14 and Petrepar® C-13
and Rubitherm® RT27 and Rubitherm® RT31 are regularly spherical shape and smooth
surface. However, the particles obtained with Rubitherm® RT20 have are irregular with
rough surface, which can be attribute to the polarity of this product (Sánchez et al., 2007).
Although microcapsules containing PRS® paraffin wax, Rubitherm® RT27 and Rubitherm®
RT31 have similar average thermal storage energies and regular spherical microcapsules
with smooth surface. The most convenient melting temperature for an effective utilization of
this kind of materials in textiles field was obtained using Rubitherm® RT31. Obviously, its
phase change temperature (31ºC) is comfortable for the human body, and a higher efficiency
of encapsulation (49.5%) than the others.
3.2 Influence of the kind of textile substrate on the production of smart fabric
Seven fabrics substrates for different textile applications using Rubitherm® RT31 as core
material were used. Their description and properties are shown in Table 3. Samples were
named A to G, according to the textile substrates employed.
(a) (b)
(e) (f)
(c) (d)
(e) (f)
Fig. 3. ESEM micrographs of microcapsules containing: (a) Petrepar® C-13, (b) Petrepar®
C-14, (c) Rubitherm® RT20, (d) Rubitherm® RT27, (e) Rubitherm® RT31 and (f) PRS® paraffin
wax
Thermal performance of different coated textiles with thermo-regulating properties with 35
wt. % of PCM microcapsules as a function of the kind of textile substrate was evaluated by
DSC analyses (Figure 4). It can be seen that all treated textile substrates allows to obtain
thermo-regulating properties with acceptable latent heat storage capacities. The same result
was observed by other authors (Salaün et al., 2010; Izzo Renzi et al., 2010). On the other
hand, no significant differences on the phase change transition temperature of coated

samples and the Rubitherm® RT31 microcapsules were observed. It was found that the
melting transition points in the coated fabrics from A to G changed 0.72, 0.01, 0.13, 0.37, 0.13,
-0.35 and 0.01 ºC, respectively. This indicated that the kind of textile substrate have not
significant effect on the microcapsules melting effect. In this sense, Koo et al., (2009)
observed changes up to 0.58 ºC by using a wet coating method.
Area
Thickness
Sample Composition Weight Uses
(mm)
(gm )
-2
Soft-Shell fabric with an

82% Polyester
A 296 1.50 intermediate polyurethane
18% Polyurethane
membrane for cold protection
11% Elastane Soft-Shell fabric with an
B 35% Polyamide 270 1.34 intermediate polyurethane
54% Polyester membrane for cold protection
C 100% Polyamide 202 0.48 Green fabric for military uses
100% Polyester
Yellow fabric for garments of
D coating with 121 0.27
high visibility
100% PVC
40% Polyester Openwork fabric in blue tone for
E 185 0.35
60% Cotton medical garments uses
Military printed fabric for
F 100% Polyamide 328 0.61
military uses
Fabric used for upholstery,
G 100% Cotton 158 0.30
sheets, curtains and garments
Table 3. Textile substrates characterization and their textile uses
In our previous paper, textile substrate G (100% cotton) was used to obtain the textiles with
thermo-regulating properties (Sánchez et al., 2010). In the present work the employed
experimental conditions were the same except for the thickness of the coating layer. A
thermal storage capacity of 14.4 Jg-1 was achieved using cotton textile substrate and a
coating thickness of 0.1 mm. Comparing this result with the value of 7.6 Jg-1 obtained in the
previous work with 0.01 mm of thickness, it is observed that the latent heat storage capacity
of the treated fabric increased as the thickness of the coating layer increased.
Table 4 summarizes the latent heat storage capacity, the necessary time to decrease the
temperature of coated textiles from 33 to 25 ºC, the latent heat accumulated in 1 m2 of fabric
substrate associated with each sample and the amount of PCM microcapsules added on
each textile substrate. There are not important differences in the latent heat storage capacity
and the amount of retained PCM microcapsules depending on the kind of used substrate
textile. Coated textiles A and B exhibit the highest latent heat and a long thermoregulatory
effect, due to the soft shell characteristics and the large thickness of these fabrics that allow
to accommodate a high amount of PCM microcapsules and improve the resistance of heat
transfer, respectively. Sample D with 16.3 Jg-1 shows a short time of the heat release, this
may be related to the small thickness of the used textile substrate (0.27 mm). For samples E
and G the latent heat storage capacity, the amount of PCM microcapsules added on the
textiles and the thickness of these textiles are similar, thus the duration of the thermal buffer
effect was quite similar. Furthermore, the lowest latent heat storage capacity for samples
having polyamide as textile substrate (samples F and C) suggests that this composition does
not allow the incorporation of a large amount of PCM microcapsules into the fabric.
Nevertheless, small differences of latent heat storage capacity were obtained. According to
these results, the heat transfer through fabric depends on the quantity of PCM
microcapsules added on the coating binder but also the textile composition and structure.
Consequently, numerous factors must be considered in the evaluation of the heat transfer
properties of fabrics (Koo et al., 2009).
0.22
C
Heat Flow (W/g)
-0.92
-20 0 20 40 60 80
Exo Up Temperature (°C) Universal V4.2E TA Instruments
Fig. 4. DSC thermograms of the different coated textiles with 35 wt.% of microcapsules
containing Rubitherm® RT31
Reliability tests to study the thermal performance of a textile with thermo-regulating
properties (sample A) during thermal cycling were carried out using DSC analysis (Figure
5). It can be observed that the latent heat storage of the sample does not change when
heating/cooling cycle is repeated (less than 2 % of latent heat variation). Furthermore,
melting and freezing transition points of the coated fabrics shift to higher temperature than
microcapsules containing Rubitherm® RT31. This fact could be attributed to the influence of
the polymeric binder on the thermal response of textile (Salaün et al., 2010).
Therefore, these results demonstrated that textile substrates with different characteristics
and applications are suitable for application of the microcapsules by means of a coating
method. Similar results of latent heat storage capacity using natural leather and 40 wt.% of
PCM microcapsules were obtained by Izzo Renzi et al., (2010).
Latent heat PCM

Duration of
accumulated in 1 microcapsules
the heat
Sample ΔH (Jg-1) m2 of fabric added on the
release from
substrate textile
33 to 25ºC (s)
(kJm-2) (wt. %)
A 19.4 79 5.7 25.6
B 18.1 62 4.9 23.9
C 13.5 65 2.7 17.8
D 16.3 28 4.4 21.5
E 14.3 42 2.6 18.9
F 11.1 67 3.6 14.7
G 14.4 48 2.3 19.0
Table 4. Thermal properties of the thermo-regulating textiles
0.6
Intensity (a.u)
-0.6
-30 -20 -10 0 10 20 30 40 50 60 70 80
Exo Up Temperature (°C) Universal V4.2E TA Instruments
Fig. 5. DSC curves of a textile with thermo-regulating properties (sample A), triple scan
Figure 6 shows ESEM micrographs of the surface and the cross-sectional morphology of a
representatively sample (coated textile A) in which PCM microcapsules were used as the
thermal insulation material. It can be seen, the successful fixation between textile substrate
and the microcapsules containing Rubitherm® RT31. The PCM microcapsules in the treated
samples were mainly located at the spaces between fibers and the fiber surface. As shown in
Figure 6b, the textile substrate A is remarkable thick leads to add a large amount of PCM
microcapsules on the textile substrate according to DSC analysis obtained.
(a)
(b)
Fig. 6. ESEM micrographs of a thermo-regulating textile (sample A): (a) surface and (b) cross
section
During the last years several testing methods have been developed for measuring the
temperature-regulating ability of PCM in fabrics (Bryant & Colvin, 1994; Mengjin et al.,
2008; Onder et al., 2008; Koo et al., 2009). In this work, the efficiency of the active thermal
insulation effect of the textiles with incorporated microcapsules has been evaluated using
thermal vision camera. Figure 7 shows a comparison of the maxima temperature reached on
the surface temperature distribution for the seven samples with the time. In this study, the
coated textiles with and without PCM microcapsules were cooled from 60ºC to room
temperature.
In all the cases, Rubitherm® RT31 buffer effect is observed. Therefore, during the cooling
process there is a temperature range when the temperature of coated textile with PCM
microcapsules is higher than the non-coated one. This indicates that the stored energy of
PCM is transferred to the environment in the phase change range during the reverse cooling
62 62
58 A 58
B
54 54
50 50
Tem perature (ºC)
Tem perature (ºC)

46 46
42 42
38 38
34 34
30 30
26 26
22 22
0 20 40 60 80 100 120 140 160 180 200 220 240 0 20 40 60 80 100 120 140 160 180 200
Time (s)
Time (s)
62 62
58
C 58
D
54 54
50 50
Tem perature (ºC)
46 Tem perature (ºC) 46
42 42
38 38
34 34
30 30
26 26
22 22
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160
Time (s) Time (s)
62 62
58
E 58
F
54 54
50 50
Tem perature (ºC)
Tem perature (ºC)
46 46
42 42
38 38
34 34
30 30
26 26
22 22
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 180 200
Time (s) time (s)
62
58
54
G
50
Tem perature (ºC)
46
42
38
34
30
26
22
0 10 20 30 40 50 60 70 80 90 100 110
Time (s)
Fig. 7. Temperature distribution as a function of time for different coated textiles by thermal
vision camera: ----- Textile without PCM microcapsules Textile with PCM microcapsules
process (Salaün et al., 2010). As it was previously mentioned, coated textile D exhibits the
highest rate of temperature decrease and samples A and B have a significant effect on the
thermal insulation.
With the aim to test the thermal comfort in summer conditions, the effect of a reference
textile and a prototype of a textile with thermo-regulating properties in contact with the
body (shoulders in this specific case) was visualized by IR thermography images (Figure 8).
The zero time was taken when the individual went out to the building, from 25 ºC (inside
temperature) to 35 ºC (outside temperature).
t= 10 s
Textile with microcapsules Textile without microcapsules
t= 30 s
t= 45 s
Ç´`
Fig. 8. Infrared camera images of coated textile A with and without PCM microcapsules at
different times
Thermal imagines show that textiles with thermo-regulating properties undergoes a heating
effect less marked than the textile without PCM microcapsules, as a consequence of the
melting process of encapsulated Rubitherm® RT31. Thermoregulatory effect observed for
textile with PCM microcapsules with respect to the normal fabric was 1, 0.9, 0.8 and 0.7 ºC
after 10, 30, 45 and 75 s.
Concluding, the thermal performance of textiles with thermo-regulating properties depends
on the energy storage capacity, PCM microcapsules content, textile substrate structure and
the correspondence between the phase change temperature and the application temperature
range. Consequently, thermal effects can be improved according to the final application
request using an appropriate combination of all these parameters.
4. Conclusion
A method based on a suspension free radical polymerization process has been used for the
encapsulation. Different phase change materials (PRS® paraffin wax, Petrepar® C-14 and
Petrepar® C-13 and Rubitherm® RT27 and Rubitherm® RT31) can be encapsulated by this
method and form a core-shell structure. Although microcapsules containing PRS® paraffin
wax, Rubitherm® RT27 and Rubitherm® RT31 have similar average thermal storage energies
and regular spherical microcapsules with smooth surface. The most convenient melting
temperature for an effective utilization of this kind of materials in textiles field was obtained
using Rubitherm® RT31. Its phase change temperature (31ºC) is comfortable for the human
body and it has a higher efficiency of encapsulation (49.5%) than the others.
Fabrics with thermo-regulating properties had heat storage capacities of 11.1-19.4 Jg-1, which
depended on the textile substrate. The DSC and the thermal temperature distribution
analyses indicated that all the coated textiles with 35 wt. % of microcapsules showed a high
latent heat storage capacities and significant thermoregulatory effect. Furthermore, DSC
curves demonstrated that the thermal energy storage of the textiles with thermo-regulating
properties does not change when heating/cooling cycle was repeated. A thermoregulatory
effect in contact with the body of 0.9 ºC after 45 s was observed by means of IR thermography
images.
5. Acknowledgment
Financial support from ASINTEC S.A. and Ref. PBC08-0243-1458 from Consejería de Ciencia
y Tecnología (JCCM) are gratefully acknowledged.
6. References
ASTM D 823-C. (1997). Standard practices for producing films of uniform thickness of paint,
varnish, and related products on test panels
Boh, B. & Knez, E. (2006). Microencapsulation of essential oils and phase change materials
for applications in textile products, Indian Journal of Fibre and Textile Research,
Vol.31, pp.72–82, ISSN 0971-0426
Bryant, Y.G. & Colvin, D.P. (1988). Fiber with reversible enhanced thermal storage
properties and fabric made therefrom, US Patent 4,756,958
Bryant, Y.G. & Colvin, D.P. (1994). Fabric with reversible enhanced thermal properties, US
Patent 5,366,807
Colvin, D.P. & Bryant, Y.G. (1996). Thermally enhanced foam insulation, US Patent 5,637,389
Dixit, S. & Goel, A. (2007). Microencapsulation in textile processing: an overview, Asian
Textile of Journal, Vol.16, pp. 83–86, ISSN 1819-3358
Hamielec, A.E. & Tobita, H. (2003). Ullmann´s Encyclopedia of Industrial Chemistry
(Weinheim: WILEY-VCH).
Izzo Renzi, A.; Carfagna, C. & Persico, P. (2010). Thermoregulated natural leather using
phase change materials: An example of bioinsipiration, Applied Thermal Engineering,
Vol.30, pp.1369-1376, ISSN 1359-4311
Koo, K.; Choe, J. & Park, Y. (2009). The application of PCMMcs and SiC by commercially
direct dual-complex coating on textile polymer, Applied Surface Science, Vol.255,
pp.8313-8318, ISSN 0169-4332
Monllor, P.; Sánchez, L.; Cases, F. & Bonet, M.A. (2009). Thermal behaviour of
microencapsulated fragrances on cotton fabrics, Textile Research Journal, Vol. 79, pp.
365-380., ISSN 0040-5175
Mengjin, J.; Xiaoqing, S.; Jianjun, X. & Guangdou, Y. (2008). Preparation of a new thermal
regulating fiber based on PVA and paraffin, Solar Energy Materials & Solar Cells
Solar,Vol.92, pp. 1657-1660, ISSN 0927-0248
Onder, E.; Sarier, N. & Cimen, E. (2008). Encapsulation of phase change materials by
complex coacervation to improve thermal performances of woven fabric,
Thermochimica Acta, Vol.467, pp. 63-72, ISSN 0040-6031
Pause, B. (2003). Nonwoven protective garments with thermo-regulating properties, Journal
of Industrial Textiles, Vol.33, No.2, pp. 93-99, ISSN 1528-0837
Ren, Y.J. & Ruckman, J.E. (2004). Condensation in Three-layer Waterproof Breathable
Fabrics for Clothing, International Journal of Clothing Science and Technology, Vol.16,
No.3, pp. 335-347, ISSN 0955-6222
Rodrigues, S.N.; Martins, I.M.; Fernandes, I.P.; Gomes, P.B.; Mata, V.G.; Barreiro, M.F. &
Rodrigues, A.E. (2009). Microencapsulated perfumes for textile application,
Chemical Engineering Journal, Vol.149, pp. 463-472, ISSN 1385-8947
Salaün, F.; Devaux, E.; Bourbigot, S. & Rumeau, P. (2010). Thermoregulating response of
cotton fabric containing microencapsulated phase change materials, Thermochimica
Acta. Vol. 506, No.1-2, pp. 82-93, ISSN 0040-6031
Sánchez, L.; Lacasa, E.; Carmona, M.; Rodríguez, J.F. & Sánchez, P. (2008b). Applying an
experimental design to improve the characteristics of microcapsules containing
PCMs for fabric use, Industrial & Engineering Chemistry Research, Vol.47, pp.9783-
9790, ISSN 1520-5045
Sánchez, L.; Sánchez, P.; Carmona, M.; de Lucas, A. & Rodríguez, J.F. (2008a). Influence of
operation conditions on the microencapsulation of PCMs by means of suspension-
like polymerization. Colloid and Polymer Science, Vol.286, No.8-9, pp.1019-1027, ISSN
0303-402X
Sánchez, L.; Sánchez, P.; De Lucas, A.; Carmona, M. & Rodríguez, J.F. (2007).
Microencapsulation of PCMs with a polystyrene shell. Colloid and Polymer Science,
Vol.285, No.12, pp.1377-1385, ISSN 0303-402X
Sánchez, P.; Sánchez-Fernández, M.V.; Romero, A. ; Rodríguez, J.F. & Sánchez-Silva, L.

(2010). Development of thermo-regulating textiles using paraffin wax microcapsules,
Thermochimica Acta, Vol.498, pp.16-21, ISSN 0040-6031
Shin, Y. ; Yoo, D.I. & Son, K. (2005). Development of Thermoregulating Textile Materials
with Microencapsulated Phase Change Materials (PCM). IV. Performance
Properties and Hand of fabrics treated with PCM microcapsules, Journal of Applied
Polymer Science, Vol.97, pp.910-915, ISSN 0021-8995
34
Heat Conduction in Nonlinear Media

Michael M. Tilleman
Elbit Systems of America (Kollsman), LLC.,
220 Daniel Webster Highway, Merrimack, NH 03054
USA
1. Introduction
The objective of this chapter is to demonstrate a closed form solution to a unique problem in
heat transfer, that of heat transfer in nonlinear media. For this purpose identified are cases
in which the nonlinear phenomenon dominates, where properties of a medium exhibit
nonlinear response to heating, and the solution methodology is described. Further in this
chapter presented is a set of analytical solutions applicable to various geometrical forms
including the cases of: infinite and finite cylinders with axisymmetrical source, finite
cylinder with axisymmetrical and axially varying sources, slender disk, infinite and finite
parallelepiped with centrally symmetric source, finite parallelepiped with axially varying
source and the resulting stress due to temperature distribution. In this abstract an example
is shown of a solution for a particular case, that of an infinite cylinder.
In many cases equations governing the phenomenon of heat transfer are solved assuming
constant physical properties of the media concerned. That is where a whole class of closed
form solutions is found, if regular boundaries and boundary conditions are provided.
However, at instances in which the coefficient of heat conduction, specific heat and thermal
diffusivity, are functions of temperature, those solutions are no longer applicable. For those
cases another family of closed form solutions is found and described herewith.
The analytical solution to the thermally nonlinear problem assumes a certain dependence of
the coefficient of thermal conductivity, k, on temperature. This case is usually found in
instances of large temperature gradients in media, for instance in active optical materials
with intense electro-optical fields. They are realized in laser gain media, nonlinear optical
crystals and saturable absorbers. One of the most frequently used host materials for lasers is
Yttrium Aluminum Garnet, YAG, in which an inverse proportionality to temperature well
approximates measured k values over a vast temperature range. What we applied for the
solution of such a problem is the Kirchoff’s transformation (Joyce, 1975), whereby the heat
equation can be linearized and solved. However, the use of this method is limited to
materials whose k is integrable in temperature, T. Only for these cases the linearization of
the heat equation can be made. For instance in the case of AgGaSe2, a nonlinear optical
crystal useful for harmonic generation, the dependence is k=A+B/T (Aggarwal & Fan, 2005)
that is an integrable function in T, therefore solvable by the present method. Evidence has it,
though, that most of the known optical crystals have similarly thermal coefficient enabling
the use of the present solution. Another strategy used in many of the present solutions is the
use of the Green’s function as kernel in the integral expression. Owing to the strong
dependence of additional properties of these materials, for instance thermal expansion

coefficient and thermal rate of refractive index (dn/dT) on temperature, the solution to the
heat transfer problem presents an important tool to understand the reaction of these media
to intense heating.
The analysis begins by defining the governing equation and the boundary conditions.
Governing is the Poisson equation expressed as:
∇ ⋅ k ( ∇T ) + Q = 0 (1)
where: k=k(T) is the thermal conductivity and Q is the heat source term defined as
deposited power per unit volume. To solve for cylinders or rectangular parallelepipeds
either a polar or Cartesian coordinate systems need be considered as necessary.
1.1 Coefficient of thermal conductivity

The coefficient of thermal conductivity is a product of the material density, thermal
diffusivity and specific heat. Approximation to theory and the fitting to a broad body of
measured values suggest an inverse linear approximation:
T0
k (T ) = k0 (2)
T
where k0 is the coefficient of thermal conductivity at T0 whose values for several materials
are summarized.(Aggarwal et al., 2005) This simple function, similar to that suggested for
Nd:YLF in (Pfistner et al., 1994) and (Hardman et al., 1999), closely fits theory and agrees
very well with data in the range between cryogenic temperature and 770K. It is worth
mentioning that though the specific approximation of Eq. 2 holds true for some materials,
there are other materials for which alternative approximations may become a better fit.
1.2 The nonlinear poisson equation for a cylinder with axisymmetrical source
Let us consider the Poisson equation of Eq. 1 in cylindrical coordinates, using the expression
for k of Eq. 2:
⎡ T ⎤
⎣ T
( ⎦
)
∇ r ,ϕ , z ⋅ ⎢ k0 0 ∇ r ,ϕ , zT ⎥ + Q ( r ,ϕ , z ) = 0 (3)
Consistent with Kirchoff’s transformation this equation can be rewritten as:
1 T
k0T0∇ r ,ϕ , z ⋅
T
( )
∇ r ,ϕ , zT + Q ( r ,ϕ , z ) = k0T0 ∇ r2,ϕ , z ln + Q ( r ,ϕ , z )
T0
(4)
then one arrives at:
K∇ r2,ϕ , zθ ( r , ϕ , z ) + Q ( r , ϕ , z ) = 0 (5)
Being a linear equation in θ where:
T ( r ,ϕ , z )
θ ( r ,ϕ , z ) = ln (6)
T0
Heat Conduction in Nonlinear Media 669
and K is a constant:
K = k0T0 (7)
The above notation is cylindrical. Notwithstanding, this linearization holds true for any
orthogonal coordinate system. It should be emphasized that Eq. 5 represents any steady-
state heat equation, where θ and K may be an arbitrary representation of any compound
temperature-heat conduction function. Ultimately this case degenerates to the linear case
where the heat conduction coefficient is independent of temperature.
Assuming a distributed heat source one has:
P (8)
Q(r ,ϕ , z) = f 0 (r ,ϕ , z)
π rP2 Leff
where P is the power deposited in the medium, rP is the effective radius of the heating zone
in the cylinder, Leff is the effective length of the medium which may be, for instance, a
cylindrical rod or parallelepiped and f0 is an arbitrary spatial distribution function of the
deposited power.
The boundary conditions for the problem are prescribed by a given cooling mechanism,
which may consist of a coolant fluid or a solid heat sink, and physical surroundings of the
device. Typically at least one side of the heated medium is held in good thermal contact
with the cooling mechanism. Being either kept in vacuum or exposed to gas the other sides
of the medium may be considered insulated due to lack of any considerable heat transfer. At
the areas of the rod through which heat flow occurs, say at r=rW, one may specify the
boundary by a known temperature, i.e. a Dirichlet condition:
T (rW ,ϕ , z ) = T0 (9)
Otherwise, in the case the heat flux out of the system is known, the boundary is specified by
a Neumann condition:
∂T (10)
k =q
∂nˆ
and on all the insulated sides the heat flux vanishes, thus:
∂T (11)
=0
∂nˆ
2. Cylindrical rods
Consistent with Kirchoff’s transformation (Joyce, 1975) Eq. 5 is expressed in cylindrical
coordinates as:
K ∂ ∂ K ∂2 ∂2
r θ ( r ,ϕ , z ) + 2 2 (
θ r ,ϕ , z ) + K 2 θ ( rϕ , z ) + Q ( r ,ϕ , z ) = 0 (12)
r ∂r ∂r r ∂ϕ ∂z
2.1 Rod of finite length

This case is illustrated in Figure 1(a). Sketched is a rod held by two conducting mounts at
the ends, surrounded on its perimeter by a source emitting electromagnetic radiation. Such a
source may be a black-body emitter, incandescent lamp, light emitting diode or laser, to
name a few. The rod absorbs the radiation converting it to heat. In order to approximate
cases of: 1) rod with evenly spaced side heating and varying axial heating rate or boundary
condition, and 2) end heating with arbitrarily distributed source or boundary conditions, it
is sufficient to assume an axisymmetrical case with a finite rod. Then the governing equation
becomes:
K ∂ ∂ ∂2
r θ ( r , z) + K 2 θ ( r , z) + Q ( r , z) = 0 (13)
r ∂r ∂r ∂z
Thin Disk
Heat
Induction
(a) (b) (c)

Fig. 1. Three cylindrical optical devices: a) rod held by heatsink mounts on both ends with
radially symmetrical side induction heaters, b) rod held by a heatsink along its length with
end induction heater and c) end induction heated thin disk (dark) with a cap
2.1.1 Side heating

The boundary conditions for a rod may model various cooling configurations including
conductive and convective cooling means. A possible configuration is conduction cooling,
where a heat sink with a mount holds the rod over part of its length. This configuration
justifies the assumption of either Dirichlet or Neumann boundary conditions or their
combination specified around the rod circumference:
∂ T (r , z) ∂θ ( r , z )
T ( rW , z ) = f 1 ( z ) or k (T ) =K = f2 ( z) (14)
∂ r r =r ∂ r r =r
W W
The Dirichlet boundary condition is particularly suitable for cases where the boundary is
held at a set temperature such as the case of heat sinking to Peltier junction or cryogenic
cooling. In terms of the function θ it becomes:
f1 ( z )
θ ( rW , z ) = ln (15)
T0
Then, the Neumann condition is suitable for cases where the boundary provides a certain
heat flux across it, as specified in Eq. 14. At the rod ends the facets are assumed insulated,
specified by Neumann condition, such that:
∂θ ∂θ
= =0 (16)
∂ z z = 0 ∂ z z = L /2
where the origin is at the rod center and z = L/2 is half the rod length. Modeling of the heat
source assumes a region confined both radially and axially inside the rod:
⎧ P l
⎪⎪ π r 2 l / 2 f 0 ( r ) ; 0≤z≤
2
Q(r , z) = ⎨ P (17)
⎪ l
0 ; <z
⎪⎩ 2
where l is the length of the source region in the rod such that l≤L.
To solve the set of equations 13 – 16 it is convenient to employ the Green’s function G(r,z) in
which case for the mixed boundary conditions the general solution becomes:
L /2 rW L /2
1
θ ( r , z ) = rW ∫ f 3 (ζ )V ( r , z , rW , ζ ) dζ + ∫ ∫ Q (ξ ,ζ )G ( r , z,ξ ,ζ )ξ dξ dζ (18)
0
K 0 0
where the function f3(ζ) assumes either f1(ζ) or f2(ζ) defined in Eq. 14 and the function
V(r,z,rW,ζ) is either ∂G(r,z,ξ,ζ)/∂ξ at ξ=rW or G(r,z,rW,ζ), depending on the type of boundary
condition around the rod circumference (Dirichlet or Neumann). Green’s function is
constructed by solving the homogeneous Eq. 13, or the Laplace equation, satisfying the
specified boundary conditions of Eqs.14 and 16, and by the function holding throughout the
domain.( Polianin, 2002) Thus G takes the form as demonstrated in (Polianin, 2002):
ζ⎞
2 ∞ ∞ ( ) ( ⎛
⎝
) z⎞
L⎠
⎛
J 0 μs ,mr J 0 μs ,mξ cos ⎜ 2 nπ ⎟ cos ⎜ 2 nπ ⎟
⎝ L⎠
Gs ( r , z , ξ , ζ ) = 2 ∑ ∑ (19)
rW L m = 1 n = 1 ⎡ 2 ⎛ 2 nπ ⎞
2⎤
⎢ μs , m + ⎜
⎢⎣
2
⎟ ⎥ J s μs ,mrW
⎝ L ⎠ ⎥⎦
( )
where the subscript s assumes the value of either 0 or 1 corresponding to a Neumann or
Dirichlet type boundary condition, respectively. The coefficients μm are the roots of the
equation:
d
J 0 ( μm r ) = 0 ; s=0
dr rW (20)
J 0 ( μmrW ) = 0 ; s=1
To present a complete solution one still needs to define the functions f0(r) used in Eq. 17. Let
two cases be considered:
1. Uniform heat source distribution where:
⎧1 ; 0 ≤ r ≤ rP
f0 ( r ) = ⎨
⎩0 ; rP < r
2. Gaussian heat source profile:

⎡ ⎛ r ⎞2 ⎤
f 0 ( r ) = exp ⎢ −2 ⎜ ⎟ ⎥
⎢ ⎝ rP ⎠ ⎥
⎣ ⎦
Considering a Dirichlet boundary condition for case 1 one may solve the problem for θ −θW,
whereby the first term in Eq. 18 becomes θW and the solution is expressed as:
rp l /2
2P
θ ( r , z ) = θW + ∫ ∫ G1 ( r , z ,ξ ,ζ ) dζξ dξ
π rP2 lK 0 0
P ∞ ∞
⎛ l⎞
( ) ⎛
J 1 μ1,mrp J 0 ( μ1,mr ) cos ⎜ 2 nπ ⎟
⎝
z⎞
L⎠
(21)
= θW + 2 ∑ ∑ sin c ⎜⎝ nπ L ⎟⎠ ⎡
π rW rP LK m = 1 n = 1 ⎛ 2 nπ ⎞ ⎤ 2
2
μ1,m ⎢ μ1,2 m + ⎜ ⎟ ⎥ J 1 ( μ1,mrW )
⎣⎢ ⎝ L ⎠ ⎦⎥
Considering a Neumann boundary condition for case (1), and assuming a case where the
heat is relieved from the rod by conductive mounts holding the rod perimeter between ±S/2
and ±L/2 Eq. 18, the solution is:
L /2 rp l /2
qB 2η P
θ (r , z) = rW ∫ G0 ( r , z , rW , ζ ) dζ + 2 abs ∫ ∫ G0 ( r , z,ξ ,ζ ) dζξ dξ
K S /2 π rP lK 0 0
qB S ∞ ∞
⎛ S⎞ μ1,m J 0 ( μ1,mr ) ⎛ z⎞
= ∑ ∑
K rW L m = 1 n = 1
sin c ⎜ nπ ⎟
⎝ L⎠⎡ 2 2⎤
cos ⎜ 2 nπ ⎟
⎝ L ⎠
(22)
⎛ 2 nπ ⎞
⎢ μ1,m + ⎜ ⎟ ⎥ J 1 ( μ1,mrW )
⎣⎢ ⎝ L ⎠ ⎦⎥
P ∞ ∞
⎛ l⎞
( ) ⎛
J 1 μ0,mrp J 0 ( μ0,mr ) cos ⎜ 2 nπ ⎟
⎝
z⎞
L⎠
+ 2 ∑ ∑ sin c ⎜ nπ ⎟
⎡
π rW rP LK m = 1 n = 1 ⎛ 2 nπ ⎞ ⎤ 2
2
⎝ L⎠
μ0,m ⎢ μ0,2 m + ⎜ ⎟ ⎥ J 0 ( μ0,mrW )
⎣⎢ ⎝ L ⎠ ⎦⎥
where qB is the heat flux through the cooling mounts. For a very long rod the solutions in
Eqs. 21 and 22 approach the asymptotic solution for a two dimensional, axisymmetrical
geometry, which for the Dirichlet boundary condition becomes:
⎧ P
⎪⎛ rW ⎞ 2π Leff K ⎡ P ⎛ r 2 ⎞⎤
⎪⎜ r ⎟ exp ⎢ ⎜ 1 − 2 ⎟⎟ ⎥
⎜ ; 0 ≤ r ≤ rP
⎛ P ⎞ ⎪ ⎢⎣ 4π Leff K ⎝ rP ⎠ ⎥⎦
T ( r ) = ⎜ T∞ + ⎟ ⎨⎝ P ⎠ (23)
⎜ 2π L hr ⎟
eff W ⎠ ⎪ P
⎝
⎪ ⎛ rW ⎞ 2π Leff K
⎜ ⎟ ; rP < r ≤ rW
⎪⎩ ⎝ r ⎠
For the heat source of case 2 with Gaussian radial distribution the integrand becomes a
product of Bessel function of the first kind of order zero and Gauss function. Then, the
integral in the second term of Eq. 18 becomes:
rW ⎡ ⎛ ξ ⎞⎤
∫0 J 0 ( μmξ ) exp ⎢ −2 ⎜ ⎟ ⎥ ξ dξ (24)
⎣⎢ ⎝ rP ⎠ ⎦⎥
which has a closed form solution strictly for the case of rW→∞, being (Abramowitz &
Stegun, 1964):
rP2 ⎡ ( μ r )2 ⎤
exp ⎢ − m P ⎥ (25)
4 ⎢⎣ 8 ⎥⎦
It turns out that for rW/ rP =1, 1.25, 1.5, 1.75 the above result has an error of 15%, 6%, 1.5%
and less than 1% relative to a numerical approximation, respectively. Because in real
scenarios rW/ rP >1, this result is applicable to the present solution assuming a Dirichlet
boundary condition, yielding:
⎛ z⎞
∞ ∞ ⎡ ( μ r )2 ⎤ J 0 ( μ1,mr ) cos ⎜ 2nπ ⎟
P ⎛ l ⎞ ⎝ L ⎠
θ ( r , z ) = θW ∑ ∑ exp ⎢⎢− 8 ⎥⎥ sin c ⎜⎝ nπ L ⎟⎠ ⎡
1, m P
+ 2 2⎤
(26)
4π rW LK m = 1 n = 1 ⎛ 2 nπ ⎞
⎣ ⎦ ⎟ ⎥ J 1 ( μ1,mrW )
2 2
⎢ μ1,m + ⎜
⎢⎣ ⎝ L ⎠ ⎥⎦
The temperature in the cylinder for each case is derived via Eq. 6 using the values of θ in
Eqs.21, 22 and 26.
2.1.2 End heating

This case is illustrated in Figure 1(b). Treating the problem of end heating is not much
different than the above model for side heating. The main difference is in defining the heat
source as an exponentially decaying function along the rod axis. Thus in end heating the
temperature becomes maximized at the rod facet. The temperature surge is milder if an end-
cap made up of a non-heating material, is bonded to the rod at the entrance. A non-heating
material may be one which does not contain a heat source. The non-heating end-cap
conducts the heat generated in the rod. Assumed in the present model is an identical
coefficient of thermal conductivity in the heating and non- heating materials. For the sake of
smooth transition from the previous section a notation is selected such where L/2 expresses
the entire rod length. In sum the heat source in the rod becomes:
⎧ L
⎪ 0 ; − zC < z
2
⎪
⎪ P ⎡ ⎛L ⎞⎤ L
Q(r , z) = ⎨ 2 α exp ⎢ −α ⎜ − z − zC ⎟ ⎥ f 0 ( r ) ; 0 ≤ z ≤ − zC ∩ 0 ≤ r ≤ rP (27)
π
⎪ P r ⎣ ⎝ 2 ⎠⎦ 2
⎪ 0 ; rP < r
⎪
⎩
where zC is the thickness of the end cap, becoming zero for the case with no cap. As in the
previous section also here considered are two cases of radial heat induction distribution: a
flat-top beam and Gaussian. Also cooling of the rod is modeled by assuming either setting
the radial rod wall at a constant temperature or by a conductive mount holding the rod
between the lengths of S/2 and L/2 on each side.
The solution for a rod with uniform heat source and Dirichlet conditions on the cylinder
circumference becomes:
2α 2 LP
θ ( r , z ) = θW +
π rP rW2 K
⎧ n⎡ ⎛ zC ⎞ 2 nπ ⎛ z ⎞⎤ ⎡ ⎛L ⎞⎤ ⎫ (28)
⎨( −1 ) ⎢ cos ⎜ 2 nπ ⎟+ sin ⎜ 2 nπ C ⎟ ⎥ − exp ⎢ −α ⎜ − zC ⎟ ⎥ ⎬
×∑∑⎩
∞ ∞
⎣ ⎝ L ⎠ α L ⎝ L ⎠ ⎦ ⎣ ⎝ 2 ⎠ J ( μ r
⎦ ⎭ 1 1,m P 0 1,m) J ( μ r ) ⎛ z
cos ⎜ 2 nπ ⎟
⎞
⎡ 2⎤
μ1,m J 12 ( μ1,mrW ) ⎝ L⎠
m=1 n=1 ⎡( 2 nπ )2 + (α L )2 ⎤ ⎢ μ 2 + ⎛ 2 nπ ⎞ ⎥
⎣ ⎦ ⎢ 1,m ⎜⎝ L ⎟⎠ ⎥
⎣ ⎦
Considering the Neumann boundary condition the solution is expressed as:
L /2 rp l /2
qB 2P
θ (r , z) = rW ∫ G0 ( r , z , rW , ζ ) dζ + 2 ∫ ∫ G0 ( r , z,ξ ,ζ ) dζξ dξ
K S /2 π rP lK 0 0
qB S ∞ ∞
⎛ S⎞ μ1,m J 0 ( μ1,mr ) ⎛ z⎞
= ∑ ∑
K rW L m = 1 n = 1
sin c ⎜ nπ ⎟
⎝ L⎠⎡ 2 2⎤
cos ⎜ 2nπ ⎟
⎝ L ⎠
(29)
⎛ 2 nπ ⎞
⎢ μ1,m + ⎜ ⎟ ⎥ J 1 ( μ1, mrW )
⎢⎣ ⎝ L ⎠ ⎥⎦
P ∞ ∞
⎛ l⎞ ( )
J 1 μ0,mrp J 0 ( μ0,mr ) ⎛ z⎞
+ ∑ ∑ sin c ⎜⎝ nπ L ⎟⎠
π rW2 rP LK m = 1 n = 1 ⎡
cos ⎜ 2 nπ ⎟
⎛ 2nπ ⎞ ⎤ 2
2 L
⎝ ⎠
μ0,m ⎢ μ0,2 m + ⎜ ⎟ ⎥ J 0 ( μ0,mrW )
⎣⎢ ⎝ L ⎠ ⎦⎥
Finally, the solution for a rod with a Gaussian heat source and Dirichlet conditions on the
cylinder circumference becomes:
α 2 LP
θ ( r , z ) = θW + 2
2π rW K
(30)
(
⎡ μ r
) ⎤
2
exp ⎢ −
1m p ⎥ ⎧ −1 n ⎡ cos ⎛ 2nπ zC ⎞ + 2 nπ sin ⎛ 2nπ zC ⎞ ⎤ − exp ⎡ −α ⎛ L − z ⎞ ⎤ ⎫
⎢ 8 ⎥ ⎨( ) ⎣⎢ ⎜
⎝
⎟
L ⎠ αL
⎜
⎝ L ⎠⎦
⎟⎥ ⎢
⎣ ⎝2
⎜ C ⎟⎥ ⎬
⎠⎦ ⎭ J μ r
⎥⎦ ⎩ 0 ( 1, m )
∞ ∞
×∑ ⎣⎢ ⎛ z⎞
∑ ⎡ 2 2 ⎤ J 12 ( μ1,mrW )
cos ⎜ 2 nπ ⎟
⎝ L⎠
m=1 n=1 ⎡( 2 nπ ) + (α L ) ⎤ ⎢ μ +
2 2 ⎛ 2 nπ ⎞
⎥
⎣ ⎦ ⎢ 1,m ⎜⎝ L ⎟⎠ ⎥
⎣ ⎦
2.2 Example for a cylindrical rod

Several cases are calculated showing the profiles of temperature in nonlinear materials. To
allow a reasonable comparison between the various cases the following parameters are set:
material Yb:YAG, rod diameter 5 mm, radiation absorbing zone diameter 2.5 mm and
cryogenic cooling at 77K. Assumed is a circumferential, radially directed radiation forming
a uniform heat zone.
To estimate the effect of the rod aspect ratio on the axial temperature distribution let a heat
density of 410 W/cm3 be set while varying the rod length. Plotted in Figure 2(a) is the axial
temperature for half a rod from center to facet with a varying length. Observe that for short
rods the temperature is relatively small, growing with length to an asymptotic value.
Further length increase causes the temperature profile to flatten out reaching an asymptotic
value set by an infinitely long rod. A length of 50 mm may be considered as the value at
which the rod is well approximated by an infinitely long rod. Then, the temperature profile
is calculated assuming power magnitudes of P=240, 400 and 800 W and a rod length of 50
mm, corresponding to heat density rates of 245, 410 and 815 W/cm3, respectively. Shown in
Figure 2(b) is the radial distribution through a rod center for using finite and infinite rod
models yielding essentially identical results. Also in Figure 2(b) a comparison is made with
a calculation based on the linear solution where k is assumed constant equal to that for the
median temperature in the rod. In Figure 2(c) plotted are the temperature radial profiles for
all three heat density levels, compared again to the linear approach, exhibiting a gradually
growing discrepancy between the two for large power levels. It follows that the linear
approach is unjustifiable for large heat loads, say above 200 W/cm3.
(a)
(b) (c)
Fig. 2. Temperature in a rod: a) axial distribution for heat density of 410 W/cm3, b) radial
distribution for heat density of 410 W/cm3, comparing rod with finite and infinite length
and with the linear solution, and c) radial distribution for heat density of 245, 410 and 815
W/cm3, comparing the nonlinear with linear solutions
2.3 Heated thin disk

Illustrated in Figure 1(c), the heated disk experiences greatly diminished radial temperature
gradients due to its large area at an axial end being cooled. The thermal problem is solved as
in the section above with the difference that here one of the axial facets is attached to a heat
sink thus setting the temperature. The other facet is assumed either insulated or bonded to a
non-heating layer, a cap. Specifying the boundary conditions for the disk radial wall is
twofold: it is insulated for a non-heating disk whereas for a capped disk it has a set
temperature, same as the heat-sunk end. Note that for the case of the uncapped end there is
no substantial advantage to adding thermal contact to the radial wall due to the large aspect
ratio of the disk. On the other hand, for the capped-end case radial cooling must be assumed
to render the cap useful.
Further assumed is a heat source with flat-top profile enveloped by a circular cylinder with
a radius of rP. Denoting the disk thickness as t the heat source density is expressed as:
⎧0 ; t − zC < z
⎪
⎪ P
Q (r ) = ⎨ 2 ; 0 ≤ r ≤ rp (31)
⎪ π rP t
⎪ 0 ; rp < r
⎩
Then, the boundary conditions are:
∂θ ( r , z ) ∂θ ( r , z )
=0 ; =0 for uncapped disk
∂ z z =t ∂ r r =r (32)
W
θ ( rW , z ) = θ W for capped disk

and the end cooled boundary condition becomes:
θ ( r ,0 ) = θ W (33)
Green’s function for the Neumann and Dirichlet cases denoted by the subscript s (=0, 1)
becomes:
⎡⎛ 1 ⎞ z⎤ ⎡⎛ 1⎞ ζ ⎤
2 ∞ ∞ ( ) ( ⎣⎝
)
J 0 μsmr J 0 μsmξ sin ⎢⎜ n + ⎟ π ⎥ sin ⎢⎜ n + ⎟ π ⎥
2 ⎠ t ⎦ ⎣⎝ 2 ⎠ t⎦
Gs ( r , z , ξ , ζ ) = 2 ∑ ∑ (34)
⎪⎧ 2 ⎡⎛ 1 ⎞ π ⎤ ⎪⎫ 2
2
rW t m = 1 n = 0
⎨ μsm + ⎢⎜ n + ⎟ ⎥ ⎬ J s μsmrW
⎣⎝ 2 ⎠ t ⎦ ⎪⎭
( )
⎪⎩
Thereby the solution takes the form:
rW t
1
θs ( r , z ) = θ W + ∫ ∫ Q (ξ ,ζ ) Gs ( r , z,ξ ,ζ ) dζξ dξ
K 0 0
⎡ 1 ⎞ zC ⎤ ⎡⎛ 1 ⎞ z ⎤ (35)
4P ∞ ∞ ( −1)n sin ⎢⎛⎜ n +
⎣⎝
⎟π
2 ⎠ t ⎥⎦
( )
− 1 J 1 μs ,mrp J 0 ( μs ,mr ) sin ⎢⎜ n + ⎟ π ⎥
⎣⎝ 2⎠ t⎦
= θW + 2 2 ∑ ∑
π rW rptK m = 1 n = 1 2n + 1 ⎪⎧ 1 ⎞ π ⎤ ⎫⎪
2
⎡⎛
μs ,m ⎨ μs2,m + ⎢⎜ n + ⎟ ⎥ ⎬ J 02 ( μs ,mrW )
⎣⎝ 2⎠ t ⎦ ⎪
⎩⎪ ⎭
with the physical stipulation that for s=0 zC=0. Thence the temperature is derived as:
⎧ ⎫
⎡⎛ 1 ⎞ zC ⎤ ⎡⎛ 1 ⎞ z⎤ ⎪
⎪
⎪⎪ 4 P ∞ ∞ ( −1 ) n
sin ⎢⎜
⎣⎝
n + ⎟ π
2 ⎠ t ⎥⎦
− 1 J 1 μ r( J
s ,m p 0 ( μ )
s ,m r ) sin ⎢⎜
⎣⎝
n + ⎟ π
2 ⎠ t ⎥⎦ ⎪⎪
Ts ( r , z ) = TW exp ⎨ 2 2 ∑ ∑ + ⎧ 2⎫ ⎬ (36)
⎪ π rW rptK m = 1 n = 1 2 n 1 ⎪ 2 ⎡⎛ 1 ⎞π ⎤ ⎪ 2 ⎪
⎪ μ μ
s ,m ⎨ s , m + ⎢⎜ n + ⎟ ⎥ ⎬ 0 ( s ,m W ) ⎪
J μ r
⎣ ⎝ 2 ⎠ t ⎦
⎪⎩ ⎩⎪ ⎭⎪ ⎪⎭
Evidently the temperature grows exponentially with the dissipated heat density.
(a) (b)
(c) (d)
(e) (f)
(g) (h)
Fig. 3. Radial and axial temperature profiles in a thin disk assuming a cap length of 0 (a,b),
0.25 mm (c,d), 0.75 mm (e,f) and 1.25 mm (g,h)
2.4 Example for a thin disk

To calculate the temperature in the disk the following parameters are set: material Yb:YAG,
disk diameter 5 mm, heat source diameter or thickness 2.5 mm, slab lengths 50 mm, disk
thickness 0.25 mm, heat density rate 100 W/cm3 and cryogenic cooling at 77K.
Consistent with most realistic scenarios consider end heating of the thin disk with uniform
radial distribution within a radius of rP of a disk attached to a heat sink. To enhance heat
dissipation assumed is also a non-heating cap in good thermal contact with the disk and a
heat sink around its perimeter. For simplicity the cap is considered possessing the identical
values of coefficient of thermal conduction of the absorbing disk. Thus attached to the disk
with the thickness of 0.25 mm is a cap with a varying thickness of 0, 0.25, 0.75 and 1.75 mm.
For a heat sink temperature of 77K, disk diameter of 5 mm, source diameter of 2.5 mm and
power P=400 W the temperature axial and radial profiles are plotted in Figure 3. On
inspecting the graphs (a) – (h) one finds a dramatic drop in the maximum temperature of
above 300K to below 90K resulting from increasing the cap thickness from nil to 1.75 mm.
Another interesting point is that the axial location of the hottest spot remains at roughly the
doped disk facet. Next, the axial and radial profiles of Δn radial component are plotted in
Figure 3. On inspecting the graphs (a) – (h) one finds a dramatic decrease in the Δn
magnitude from 2.5×10-3 to below 4×10-5 with increasing the cap thickness. Also the axial
location of the peak Δn remains at the doped disk facet.
3. Slab geometry
For active optical media having the shape of a parallelepiped it is most convenient to treat
Eq. 5 in Cartesian coordinates, where consistent with Kirchoff’s transformation (Joyce, 1975)
it is expresses as:
⎛ ∂2 ∂2 ∂2 ⎞
K ⎜ 2 + 2 + 2 ⎟θ ( x , y , z ) + Q ( x , y , z ) = 0 (37)
⎜ ∂x ∂y ∂z ⎟⎠
⎝
3.1 Slab of infinite length

The case of a very long slab relative to its lateral sides is well approximated by an infinitely
long slab is schematically shown in Figure 4(a). Here the slab is held on one side in thermal
contact with the heat sink, having its width in the z direction and thickness in x direction,
both much shorter than its length in the y direction. Heating sources such as lasers, emit
radiation which forms inside the slab a laser gain zone a fraction of which generates heat
due to the quantum defect and other non-radiative relaxation processes.
Considered is a side-heated model having a slander spot shape on the slab side. It has
predominantly a Gaussian intensity distribution in the fast-axis plane, i.e. along x, and a
uniform distribution along y.
I choose to specify the boundary conditions as:
∂T ( x , y ) ∂T ( x , y ) ∂T ( x , y ) (38)
= = =0
∂y y =0
∂ y y =b
∂x x = 0
and:
T ( a , y ) = TW (39)
where TW is the temperature of the heat-sink at the interface with the slab wall. In the notation
of the non-dimensional temperature θ the two-dimensional Poisson equation becomes:
⎛ ∂2 ∂2 ⎞ Q (x, y )
⎜⎜ 2 + 2 ⎟⎟θ ( x , y ) + =0 (40)
⎝ ∂x ∂y ⎠ K
with the boundary conditions thus converted to:
∂θ ( x , y ) ∂θ ( x , y ) ∂θ ( x , y ) (41)
= = =0 ; θ ( a, y ) = θW
∂y y = 0 ∂y y = b ∂x x = 0
For a constant θW this allows a solution for the function θ ( x , y ) − θ W with a zero boundary
value at x=a.
Assumed for this case is a heat source:
Pα
Q( x , y ) = exp ( −α y ) f 0 ( x ) (42)
2rP L
where α is the absorption coefficient of the laser radiation, rP is the small aperture of the
radiation beam, a, b and L are the x , y and z dimensions of the slab (a,b<<L). If the slab side
at y=b reflects the inducing beam then the heat source becomes:
Pα exp ( −α b )
Q( x , y ) = cosh ⎡⎣α ( b − y ) ⎤⎦ f 0 ( x ) (43)
rP L
The formal solution to the equation set 40 – 41 is:
ab
1
θ ( x , y ) = θW + Q (ξ ,η )G ( x , y , ξ ,η ) dξ dη
K ∫∫
(44)
00
where the Green’s function construction is stipulated by solving the homogeneous Eq. 40, or
the Laplace equation, by satisfying the boundary conditions and by its holding throughout
the domain. It results in:
4 ∞ ∞ cos ( pl y ) cos ( plη ) sin ( qm x ) sin ( qmξ )

G (ξ ,η , x , y ) = ∑∑ (45)
ab l = 1 m = 0 pl2 + qm2
where:
πl π⎛ 1⎞ (46)
pl = ; qm = ⎜m + ⎟
b a⎝ 2⎠
To present a complete solution one still needs to define the functions f0(x) used in Eq. 42. Let
two cases be considered:
1. uniform heat source distribution in the x axis where:
⎧ a a
⎪⎪1 − rP ≤ x ≤ + rP
;
f0 ( x ) = ⎨ 2 2
⎪ 0 a a
; x < − rP ∩ + rP < x
⎪⎩ 2 2
2. Gaussian heat source profile:
⎡ ⎛ x − a / 2 ⎞2 ⎤
f 0 ( x ) = exp ⎢ −2 ⎜ ⎟ ⎥
⎢ ⎝ rP ⎠ ⎥⎦
⎣
Considering case (1), Eq. 44 is integrated to the closed form solution:
⎡1 − ( −1 )l e −α b ⎤ sin ⎛ qm a ⎞ sin ( q r )
Pα 2 ∞ ∞ ⎣ ⎦ ⎜ 2 ⎟ m P
θ ( x , y ) = θW ⎝ ⎠
+4 ∑ ∑ cos ( pl y ) sin ( qm x ) (47)
rP abLK l = 0 m = 0 (
qm pl2 + α 2 pl2 + qm 2
)( )
when no electromagnetic radiation is reflected from slab end, and to:
⎛q a⎞
sinh( pl b )sin ⎜ m ⎟ sin ( qmrP )
Pα 2 exp ( −α b ) ∞ ∞
2
θ ( x , y ) = θW + 8 ∑ ∑ q p2 + α⎝ 2 p⎠2 + q 2 cos ( pl y ) sin ( qm x ) (48)
rP abLK l =0 m=0 m l ( l m )( )
when the electromagnetic radiation is fully reflected from slab end.
Then for case (2) the solution takes the form:
( qm rP )2
1 − ( −1 ) e −α b
l
Pα 2 ∞ ∞ − ⎛q a⎞
θ ( x , y ) = θ W − 4 2π ∑ ∑ e 8 sin ⎜ m ⎟
( 2 2
)(
abLK l = 1 m = 0 pm + α pl + qm2 2
) ⎝ 2 ⎠
⎡ ⎛ iq r 2 − 2 a ⎞ ⎛ iq r 2 − 2 a ⎞ ⎤
× ⎢ erf ⎜ m P ⎟ − erf ⎜⎜ m P cos ( pl y ) sin ( qm x )
⎜
⎢⎣ ⎝ 2 2rP ⎠ ⎟ ⎟⎟ ⎥
⎝ 2 2rP ⎠ ⎥⎦
( qm rP )2
(49)
1 − ( −1) e −α b
l
Pα 2 ∞ ∞ − ⎛q a⎞
≈ θW − 8 2π ∑ ∑ e 8 sin ⎜ m ⎟ cos ( pl y ) sin ( qm x )
2
( 2 2
abLK l = 1 m = 0 pm + α pl + qm)( 2
) ⎝ 2 ⎠
⎧ ⎛ a⎞
2
⎛ a⎞
2
⎪ −⎜ ⎟
r
−⎜ ⎟
r
1
− n2
∞ ⎡ a⎛
⎪ ⎛ a ⎞ e ⎝ P⎠ ⎝ P⎠
⎛ nqmrP ⎞ ⎞
( 1 − cos ( qm a ) ) + 2 e π e 4
× ⎨erf ⎜ ⎟ +
r a ∑ ⎢ 2 ⎜ 1 − cosh ⎜
r 2
⎝ 2 ⎟ cos ( qm a ) ⎟
⎠
⎪ ⎝ P ⎠ 2π n=1 2 ⎛ a ⎞ ⎣ P⎝ ⎠
⎪ rP n + 4⎜ ⎟
⎩ ⎝ rP ⎠
⎛ nq r ⎞ ⎤ ⎫⎪
+ n sinh ⎜ m P ⎟ sin ( qm a ) ⎥ ⎬
⎝ 2 ⎠ ⎦ ⎪⎭
where pm is defined in Eq. 46. On inspecting Eqs. 47 and 49 one observes that for a>>rP the
solution is insensitive to the magnitude of rP.
3.2 Slab of finite length

The heat conduction problem in a slab of finite length is three dimensional, covering the
cases illustrated by Figure 4(a) and (b), where a slab is held on one side in thermal contact
with a heat sink and is either side heated (a), or end heated (b). The coordinate system is
chosen to represent length in z axis and comparable width sizes in x and y axis. The
governing equation is the Poisson equation in Cartesian coordinates, with respect to all
three axes as given in Eq. 37. Several configurations of heat inductions are considered:
1) side heat induction in the y direction where the source has an oblong spot shape and a
Gaussian intensity distribution in x, 2) side heat induction with alternating directions of + y
and –y, and 3) end heat source in the z direction having a circular spot shape and Gaussian
intensity distribution. Also solved is a case for a heat source additional to that occurring in
the heating path, which originates from fluorescence and photon trapping in the slab bulk
and walls. Finally, several cases of slab cooling are considered: 1) contact with heat sink on
one of the slab sides, 2) contact with heat sink on two of the slab sides, and 3) contact with
heat sink on all four of the slab sides.
(a) (b)
Fig. 4. Two parallelepipeds: a) slab attached to a heatsink on one side, with a slender side
pump and, b) slab attached to heatsink on one side with an end pump
3.2.1 Side heating

In this case the crystal is considered three dimensional, in thermal contact with a heat sink
on one side, thus there are five Neumann and a single Dirichlet type boundary conditions.
They are expressed as:
∂θ ( x , y , z ) ∂θ ( x , y , z) ∂θ ( x , y , z) ∂θ ( x , y , z) ∂θ ( x , y , z)
= = = = =0 (50)
∂z z=0 ∂ z z=L ∂ y y =0
∂ y y =b
∂x x =0
and: θ '( a , y , z) = θ ( a , y , z) − θ W = 0
Consequently the Green’s function becomes:
1 ∞ ∞ cos ( pn x ) cos ( pnξ ) cos ( qm y ) cos ( qmη )

G (ξ ,η , ζ , x , y , z ) = ∑∑
ab n = 0 m = 1 β nm cosh ( βnmL )
H nm ( z , ζ ) (51)
where the function Hnm(z,ζ) is:
⎧⎪cosh ( β nmζ ) cosh ⎣⎡ βnm ( L − z ) ⎤⎦ ; L ≥ z >ζ ≥0

H nm ( z , ζ ) = ⎨ (52)
⎩⎪cosh ( β nm z ) cosh ⎣⎡ β nm ( L − ζ ) ⎤⎦ ; L ≥ζ > z≥0
and the coefficients pn, qm and βnm are:

⎛ 1⎞
⎜ n + ⎟π mπ
2⎠
pn = ⎝ ; qm = ; β nm = pn2 + qm2 (53)
a b
Further, allowing for a non-absorbing cap having thickness of bC on the radiation side the
heat source is defined as:
⎧ 0 ; y < bC
⎪
Q( x , y , z) = ⎨ Pα (54)
⎪ 2 r L exp ⎣⎡ −α ( y − bC ) ⎦⎤ f 0 ( x , z ) ; bC ≤ y ≤ b
⎩ P P
where LP is the radiation length and the term f0 defines the lateral profile of the source,
assumed either as flat top or Gaussian:
1. flat top:
⎧ a
⎪⎪ 1 ; 0 ≤ x − 2 ≤ rP ∩ 0 ≤ z ≤ LP
f0 ( x , z) = ⎨ (55)
⎪0 a a
; + rP < x , x < − rP or LP < z
⎪⎩ 2 2
2. Gaussian:
⎧ ⎡ ⎛ ⎞ ⎤
2
⎪⎪1 ; exp ⎢ −2 ⎜ x − a / 2 ⎟ ⎥ ∩ 0 ≤ z ≤ LP
f0 ( x , z) = ⎨ ⎢ ⎝ rP ⎠ ⎥⎦ (56)
⎪ ⎣
⎪⎩ 0 ; LP < z
The formal solution takes the shape:
abL
θ ( x , y , z ) = θ0 + ∫ ∫ ∫ Q (ξ ,η , ζ )G ( x , y , z , ξ ,η , ζ ) dξ dη dζ (57)
000
which after substituting the functions given in Eqs. 51 and 54 and carrying out the
integration becomes:
Pα 2 eα zc
θ ( x , y , z ) = θ0 +
π rP2 abK {1 − exp ⎣⎡ −α ( b − bC ) ⎦⎤}
⎡1 − ( −1 )m ⎤ sin ( p r ) cos ⎛ p a ⎞
∞ ∞ ⎣ ⎦ n P ⎜ n ⎟
×∑ ⎝ 2⎠
∑ cos ( pn x ) cos ( qm y ) (58)
n=0 m=1 (
pn α 2 + q m
2 2
βnm )
⎡ cosh β nm z ⎤
× ⎢ tanh β nmL − sinh β nm ( L − LP ) ⎥
⎣ cosh β nmL ⎦
Observe that for rP=a/2 the term sin ( pnrP ) cos ( pn a / 2 ) becomes ( −1 ) / 2 , and for L=LP the
n
term in the square parenthesis becomes merely tanh β nmL thus rendering the solution in Eq.
58 independent of z.
Often radiation sources are placed successively on both sides of a slab. In order to solve for
this case one needs only to add a second source placed on the other side of the slab, whereby
the governing equation and boundary conditions remain as stated above, however the heat
source is defined as:
⎧ 0 ; b − bC < y < bC
⎪
Q( x , y , z) = ⎨ P α exp ⎡⎣ −α ( b − bC − y ) ⎤⎦ (59)
⎪ 2r L 1 − exp ⎡ −α b − 2b ⎤ f 0 ( x , z ) ; bC ≤ y ≤ b − bC
⎩ P P ⎣ ( C )⎦
where the term f0 is assumed either flat top:
⎧ a
⎪⎪ 1 ; 0≤ x− ≤ rP ∩ LP ≤ z ≤ 2 L P
2
f0 ( x , z) = ⎨ (60)
⎪0 a a
; + rP < x , x < − rP or z < LP ; 2 LP < z
⎪⎩ 2 2
The solution to this problem becomes:
Pα 2 e −α B
θ ( x , y , z ) = θ0 +
{
2rP LP abK 1 − exp ⎡⎣ −α ( b − bC ) ⎤⎦ }
∞ ∞ ⎡ eα B cos qm B − eα bc cos qmbc ⎤ + qm ⎡ eα B sin qm B − eα bc sin qmbc ⎤
⎣ ⎦ α ⎣ ⎦
×∑ ∑
n=0 m=1 (
pn α 2 + q m
2
)2
βnm (61)
⎛ a⎞
× sin ( pnrP ) cos ⎜ pn ⎟ cos ( pn x ) cos ( qm y )
⎝ 2⎠
⎡ cosh β nm z cosh β nm ( L − z ) ⎤
× ⎢ tanh β nmL − sinh β nm ( L − 2 LP ) − sinh β nmLP ⎥
⎣ cosh β nm L cosh β nmL ⎦
The complete solution to the slab induced by two sources disposed successively on opposite
sides of the slab is the combination of Eqs.58 and 61. Note that by this technique one can to
arbitrarily add additional induction sources on the slab sides arriving at a closed form
solution for each configuration. Finally, using this solution the temperature is again derived
via Eq. 6.
3.2.2 End heating

Treating the problem of end heat induction is in fact exactly that as the above model for
side heat induction however with changing the characteristics of the source. The heat
source is now considered axisymmetrical with a Gaussian power distribution. In end
heating the temperature is maximized at the slab facet. This surge may be relieved by
bonding to the entrance facet of the slab a non-heating end-cap segment. Lacking a heat
source yet being heat conductive contributes to removing heat from the slab facet. In the
present model it is assumed that the coefficient of thermal conductivity in both materials
is identical. For this case with the radiator aligned to the slab centerline the heat source
term is rewritten as:
⎧0 ; z < zC
⎪
Q( x , y , z) = ⎨ P α exp ⎣⎡ −α ( z − zC ) ⎦⎤ (62)
⎪ f 0 ( x , y ) ; zC ≤ z ≤ L
⎩ π rP2 1 − exp ⎡⎣ −α ( L − zC )⎤⎦
where f0 is given by:
⎧ ⎡⎛ x − a / 2 ⎞ 2 ⎛ y − b / 2 ⎞ 2 ⎤ ⎫
⎪ ⎪
f 0 ( x , y ) = exp ⎨−2 ⎢⎜ ⎟ +⎜ ⎟ ⎥⎬ (63)
⎢
⎪⎩ ⎣⎝ rP ⎠ ⎝ rP ⎥
⎠ ⎦ ⎪⎭
Again, the heat source in the slab can be expressed by Eq. 57, which after substituting the
functions given in Eqs.51 and 54 and carrying out the integration becomes:
1 Pα 2 eα zc
θ ( x , y , z ) = θ0 +
{
4 abK 1 − exp ⎡⎣ −α ( b − bC ) ⎤⎦ }
∞ ∞ cos ( pn x ) cos ( qm y )
×∑ ∑ Fn ( pn , a ) Fm ( qm , b ) (64)
n=0 m=1 β nm cosh βnmL
1 ⎡ β nm −α z ⎤
× 2 ⎢
cosh β nm ( L − z ) − e −α L cosh β nm z − e sinh β nmL ⎥
α 2 − β nm ⎣ α ⎦
where:
⎧ ⎛ a ⎞
2
⎫
⎪ ⎛ a ⎞ −⎜
2 2 a ⎝ 2 rP ⎠ ⎟
⎪
⎪erf ⎜⎜ ⎟⎟ + e ⎪
⎪ ⎝ 2 r0 ⎠ π r0
⎪
⎡ ( p r )2 ⎤ ⎛ p a ⎞⎪ ⎪
Fn ( pn , a ) = exp ⎢ − n P ⎥ cos ⎜ n ⎟ ⎨ −
k2
⎬ (65)
⎢⎣ 8 ⎥ ⎝ 2 ⎠⎪ ∞ e 4 ⎡ ⎛ kp r ⎞ ⎤ ⎪
⎪ ∑ ( n )
⎦ × 1 − cos p a cosh n P
2 ⎢ ⎜ ⎟⎥ ⎪
⎪ k =0 k 2 + 2 ⎛ a ⎞ ⎣ ⎝ 2 2 ⎠⎦ ⎪
⎪ ⎜ ⎟ ⎪
⎩ r
⎝ P⎠ ⎭
Note that in the above equation the index n, coefficient pn and length a can be replaced by m,
coefficient qm and length b, respectively. Next, the function F can be further simplified
considering the values of the coefficients pn and qm as given in Eq. 53 provides the following
values:
⎛ p a⎞ 1 1 1 1
cos ⎜ n ⎟ = ,− ,− , ,... for n = 0,1, 2, 3,...
⎝ 2 ⎠ 2 2 2 2
cos ( pn a ) = 0
(66)
⎛q b⎞
cos ⎜ m ⎟ = 0, −1,0,1,... for m = 1, 2, 3, 4...
⎝ 2 ⎠
cos ( qmb ) = ( −1 )
m
It should be mentioned that the F function is a close approximation to a combination of error

functions of complex arguments as shown in (Abramowitz & Stegun, 1964).
3.2.3 Additional case – slab attached to a heat sink on two sides

Selecting next to carry out the solution for the case of a slab held in thermal contact with a
heat sink on two adjacent walls, the boundary conditions are specified as:
∂T ( x , y , z) ∂T ( x , y , z ) ∂T ( x , y , z) ∂T ( x , y , z )
= = = =0 (67)
∂z z=0 ∂z z=L ∂x x =0 ∂y y =0
and:
T ( a , y , z ) = T ( x , b , z ) = TW (68)
Relative to the dimensionless temperature θ the boundary conditions become:
∂θ ( x , y , z ) ∂θ ( x , y , z ) ∂θ ( x , y , z ) ∂θ ( x , y , z)
= = = =0 (69)
∂z z=0 ∂ z z =L ∂ y y =0
∂x x =0
and: θ ( a , y , z ) = θ ( x , b , z) = 0
The formal solution is given by Eq. 57 with the heat source Q given by Eq. 62. The Green’s
function is constructed according to that the rules explained in the previous section, thus
becoming as in Eqs.51 and 52 however with the set of coefficients pn, qm and βnm are
⎛ 1 ⎞π ⎛ 1 ⎞π
pn = ⎜ n + ⎟ : qm = ⎜ m + ⎟ ; βnm = pn 2 + qm 2 (70)
⎝ 2⎠ a ⎝ 2⎠b
Because both walls attached to the heat sink are held at the uniform temperature of TW, the
solution is identical to that in Eqs.64 and 65 with the difference of the coefficient qm and that
the summation over m begins at 0.
3.2.4 Additional case – slab attached to a heat sink on four sides

In comparison with a slab in contact with a heat sink on two sides a surrounding heat sink
on four sides is expected to further lower the temperature elevation due to heating. Unlike
in the case where the slab is held by two sides, where the boundary conditions have four
Neumann and two Dirichlet type conditions, for this case there are two Neumann and four
Dirichlet type conditions expressed as:
∂θ ( x , y , z) ∂θ ( x , y , z)
= =0 (71)
∂z z=0 ∂z z=L
and: θ (0, y , z ) = θ ( a , y , z) = θ ( x ,0, z ) = θ ( x , b , z) = 0

One may approach a solution using symmetry considerations solving in the x and y
coordinates instead of the domain 0→a and 0→b, respectively, in the domain 0→a/2 and
0→b/2. Since in this case the heat source Q given by Eq. 62 is slightly modified to conform
with the symmetry consideration such that the f0 function is expressed as:
⎧ ⎡⎛ x ⎞ 2 ⎛ y ⎞ 2 ⎤ ⎫
⎪ ⎪
f 0 ( x , y ) = exp ⎨−2 ⎢⎜ ⎟ + ⎜ ⎟ ⎥ ⎬ (72)
⎪⎩ ⎢⎣⎝ rP ⎠ ⎝ rP ⎠ ⎥⎦ ⎭⎪
In this domain the solution is identical to the case of a slab held by a heat sink on two sides,
for which the solution is given by Eq. 64 where the Fn function is expressed as:
⎧ ⎛ a ⎞
2
⎫
⎪ ⎛ a ⎞ 2 2 a −⎜⎝ 2 rP ⎟⎠ ⎪
⎪erf ⎜⎜ ⎟⎟ + e ⎪
2 rP ⎠ π r
⎡ ( p r )2 ⎤ ⎪⎪ ⎝ ⎪
0
⎪
Fn ( pn , a ) = exp ⎢ − n P ⎥ ⎨ −
k2
⎬ (73)
⎢⎣ 8 ⎥⎪ ∞
⎦ ×∑ e 4 ⎡ ⎛ kpnrP ⎞ ⎤ ⎪
⎪ 2 ⎢
1 − cos ( p n a ) cosh ⎜ ⎟⎥ ⎪
⎪ k =0 k 2 + 2 ⎛ a ⎞ ⎣ ⎝ 2 2 ⎠⎦ ⎪
⎪ ⎜ ⎟ ⎪
⎩ ⎝ rP ⎠ ⎭
Further, the set of coefficients pn, qm and βnm as expressed in Eq. 70.
3.3 Example for slab

To calculate the temperature in the slab the following parameters are set: material Yb:YAG,
square slab side 5 mm, heat source diameter or thickness 2.5 mm, slab lengths 50 mm, heat
density rate 100 W/cm3 and cryogenic cooling at 77K.
Assumed in this case is a Yb:YAG slab side having heat induced by a flat top beam, with a
reflecting opposite wall spaced by a varying width of 5, 10 and 20 mm from the first wall,
where the slab height is 5 mm, its length is 50 mm and the radiation footprint is 2.5×50 mm.
The assumed power is P=400 W and the slab is attached to a heat sink held at 77K by its
wide side. For the three slab widths the vertical and horizontal temperature distributions are
plotted in Figure 5(a) and (b). A considerable variation is found for the vertical temperature
whereas horizontal the temperature remains nearly unchanged.
4. Concluding remarks
Closed form solutions to the nonlinear steady-state heat equation in a cylindrical rod, thin
disk and parallelepiped are presented. Hinging on using Kirchoff’s transformation, this
solution is applicable to any material in which the coefficient of thermal conductivity is
integrable in temperature. In turn, these solutions enable solving further the equation of
elasticity and that of a propagating optical beam in an inhomogeneous medium where the
refractive index is radially modulated.
Exemplary calculations are made for a Yb:YAG crystal with temperature spanning the range
of 77K – 770K. It is shown that under the condition of large heating loads, say above 200
W/cm3, predicting the temperature according to a linear approximation underestimates the
temperature. Thus, in order to obtain accurate results one must use the nonlinear solution
arrived at in this study. The predicted temperatures escalate rapidly with thermal loading
reaching 120 K and almost 200 K on the optical path for a respective heating by 400 and 800
W, in a cryogenically cooled rod. If the cooling level is raised to room temperature, then by
heating the rod at 400 W, the temperature at the center will exceed 700K.
Under the heat induction condition of heating at 500 W and cryogenic cooling the maximum
temperature reached in a thin disk, having a thickness of ¼ mm, is less than 90K provided it
has a cap at least seven times the disk thickness. Without the cap the maximum would
exceed 300K. For a slab pumped with identical power and cryogenically cooled on one side
the maximal temperature varies between 125K and 500K, depending on the slab width
spanning the range from 20 to 5 mm, respectively.
(a)
(b)
Fig. 5. Plot of characteristics in a slab: a) vertical temperature distribution, and b) horizontal
temperature distribution
5. References
Abramowitz, M., and Stegun, I.A., (1964), Handbook of Mathematical Functions with Formulas,
Graphs, and Mathematical Tables, p.299, Dover, New York.
Aggarwal, R.L. and Fan, T.Y., (2005), “Thermal diffusivity, specific heat, thermal
conductivity, coefficient of thermal expansion, and refractive index change with
temperature in AgGaSe2 “, Appl.Opt.,44, (2005), 2673-2677.
Aggarwal, R.L., Ripin, D.J., Ochoa, J.R. and Fan, T.Y., (2005), “Measurement of Thermo-optic
Properties of Y3Al5O12, Lu3Al5O12, YAlO3, LiYF4, LiLuF4, BaY2F8, KGd(WO4)2,
KY(WO4)2 Laser Crystals in the 80-300 K Temperature Range”, J.Appl.Phys.98,
(2005), 103514.
Hardman, P.J., Clarkson ,W.A., Friel, G.J., Pollnau, M. and Hanna, D.C., (1999) , “Energy-
transfer upconversion and thermal lensing in high-power end-pumped Nd:YLF
laser crystals”, IEEE J.Quant.Electron.35, (1999), 647-655.
Joyce, W.B., (1975), “Thermal resistance of heat sinks with temperature-dependent

conductivity”, Solid-State Electronics 18, (1975), 321.
Pfistner, C., Weber, R., Weber, H.P., Merazzi, S., and Gruber, R., (1994) , “Thermal beam
distortions in end-pumped Nd:YAG, Nd:GSGG and Nd:YLF rods”, IEEE J.Quantum
Electron.30, (1994), 1605-1615.
Polianin, A.D., (2002), Handbook of linear partial differential equations for engineers and scientists,
Chapman & Hall/CRC.

(Barresi, Antonello Fissore, Davide Pisano, Robe

Uploaded by

Copyright:

Available Formats

(Barresi, Antonello Fissore, Davide Pisano, Robe

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Barresi, Antonello Fissore, Davide Pisano, Robe

Uploaded by

Copyright:

Available Formats

DEVELOPMENTS IN

Copyright © 2011 InTech

Publishing Process Manager Ivana Lorković

First published August, 2011

A free online edition of this book is available at www.intechopen.com

Developments in Heat Transfer, Edited by Marco Aurélio dos Santos Bernardes

Chapter 1 Thermal Effects in Optical Fibres 1

Chapter 2 Heat Transfer for NDE: Landmine Detection 21

Chapter 3 The Heat Transfer Enhancement

Chapter 4 Magneto Hydro-Dynamics and

Chapter 5 Thermal Anomaly and Strength of Atotsugawa Fault, Central

Chapter 6 Heat Transfer in Freeze-Drying Apparatus 91

Chapter 7 Radiant Floor Heating System 115

Chapter 8 Variable Property Effects in

Chapter 9 Bioheat Transfer 153

Chapter 10 The Manufacture of Microencapsulated Thermal

Chapter 11 Heat Transfer and Thermal Air Management

Chapter 12 Unsteady Mixed Convection Flow in the Stagnation

Chapter 13 Heat Generation and Transfer on Biological

Chapter 14 Entransy Dissipation Theory

Chapter 15 Inverse Space Marching Method for

Chapter 16 Experimental Prediction of

Chapter 17 High Temperature Thermal Devices for

Chapter 18 Flow Properties and Heat Transfer of

Chapter 19 Entransy - a Novel Theory in

Chapter 20 Transient Heat Transfer and Energy

Chapter 21 Role of Heat Transfer on Process

Chapter 22 Thermal Treatment of Granulated

Chapter 23 Method for Measurement of Single-Injector

Chapter 24 Heat Transfer Related to

Chapter 25 Progress Works of High and

Chapter 26 Design of the Heat Conduction Structure

Chapter 27 Thermal Modelling for

Chapter 28 Study of the Heat Transfer Effect

Chapter 30 Thermal State and

Chapter 31 Heat Transfer in the Environment: Development and

Chapter 32 Prandtl Number Effect on

Chapter 33 Effective Method of

Chapter 34 Heat Conduction in Nonlinear Media 667

Recent 40 years have witnessed considerable advances in experimental, theoretical and

Marco Aurélio dos Santos Bernardes

Thermal Effects in Optical Fibres

3. Fibre fuse effect

3.1 Experimental characterization of the fuse effect

Dummy fiber Test fiber

3.2 Theoretical model

Optical Power (W)

10000 propagation direction propagation direction

4. Optical fiber coating temperature increase

Fig. 13. Temperature values in the bent region along time

4.1 Modelling the fibre temperature

Qout = π dlh(T − T0 ) (5)

Pin = Pmax ( 1 − exp ( − β ΔP ) ) (6)

Hanafusa, H., Y. Hibino and F. Yamamoto (1985). Formation Mechanism of Drawing-

Rocha, A. M., B. Neto, A. Martins, G. Incerti, D. Forin, G. Bellefi, M. Facao, J. L. Pinto, A. L. J.

Heat Transfer for NDE: Landmine Detection

2. Infrared thermography for NDE

     

2.1 Thermal model of the soil

yδ − F [ pk(δ,yδ ) ] ≤ μδ < yδ − F [ pδk ] 0 ≤ k ≤ k(δ, yδ ) (18)

yδ − F [ pδk(δ,yδ ) ] ≤ 1.23 < yδ − F [ pδk ], 0 ≤ k ≤ k(δ, yδ ) (39)