Applied Thermal Engineering 195 (2021) 117189

Contents lists available at ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier.com/locate/apthermeng

Three-dimensional heat transfer analysis of flat-plate oscillating heat pipes

Kimihide Odagiri a, b, Kieran Wolk c, Stefano Cappucci a, Stefano Morellina a, Scott Roberts a,
Andre Pate a, Benjamin Furst a, Eric Sunada a, Takuro Daimaru a, *
a
NASA Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Dr., Pasadena, CA 91101, USA
b
JAXA Institute of Space and Astronautical Science, 3-1-1 Yoshinodai, Chuo, Sagamihara 252-5210, Japan
c
Department of Mechanical and Aerospace Engineering, University of California Los Angeles, 420 Westwood Plaza, Los Angeles, CA 90095, USA

A R T I C L E I N F O A B S T R A C T

Keywords: This paper presents a three-dimensional heat transfer analysis of a flat-plate oscillating heat pipe (OHP). To
Flat-plate oscillating heat pipe enhance implementation of OHPs in various applications, it is important to understand the comprehensive
3D heat transfer analysis phenomena that include thermal diffusion in the whole structure and thermal hydraulics in the channel. We
Thermal diffusion
developed an OHP model that includes the effect of thermal diffusion and thermo-fluid behavior. The model was
Thermal hydraulics
validated with experimental results of a flat-plate aluminum OHP which has a size of 200 mm × 90 mm × 3.8
mm, a channel diameter of 1.0 mm, and a turn number of 42. The effect of surface roughness in the channel and
liquid film thickness on operating temperature is investigated. The relation between thermo-fluid behavior in the
channel and temperature distribution of the OHP is analyzed. For aluminum OHP, the temperature difference
across the thickness direction is 0.1 ◦ C which is relatively small, whereas, the moving hot spot: local high-
temperature region, is found in the planer surface. The maximum hot spot superheat temperature reached
5.4 ◦ C for 200 W. The novelties of this paper are to develop the three-dimensional comprehensive OHP model
and to reveal the combined effect of thermo-hydraulic phenomena and thermal diffusion in OHP structure.

processes.
The conventional OHPs have a meandering tubular configuration
1. Introduction typically made with metal tubes. Recently, several studies are conducted
on flat-plate OHPs, where the channels are engraved on a flat plate. This
Future planetary space science missions require thermal control feature leads further miniaturization of OHPs and application to the high
systems that can manage high heat dissipation from electronic compo­ heat flux devices [7–12]. In our group, a flat-plate OHP by additive
nents with high efficiency. Oscillating heat pipes (OHPs) and pulsating manufacturing (AM) using AlSi10Mg powder was developed. AM tech­
heat pipes (PHPs), invented by Akachi [1], are promising for future nology allows complex three-dimensional (3D) OHP channel geometry
thermal control systems for spacecraft [2–4]. Fig. 1 shows a schematic of and makes the sealing between channels easier. Moreover, the appli­
an OHP. It consists of a bent, narrow channel that goes back and forth cation of AM technology in fabricating various materials is rapidly
between a heating and a cooling section several dozen times [5,6]. growing. Employing AM in the fabrication of OHPs is expected to
Working fluids filled in the channel separates into liquid slugs and vapor enhance the possibility of implementing OHPs in various devices as a
plugs. A slug flow state is formed due to the capillary force in the thermal control system.
channels under nominal operating conditions. When heat is applied to So far, many researchers have conducted experimental and numer­
the heating section, the liquid slugs and vapor plugs start self-excited ical studies on OHPs [13–25]. Mameli et al. experimentally investigated
oscillations due to the pressure difference along the flow path. This thermal-hydraulic performance using PHP made of transparent glass
pressure difference is induced by evaporation and condensation of the tubes [13]. They measured the local heat transfer coefficient in the
working fluid in the heating and cooling sections, respectively. OHPs evaporator and explained the relation between them and the visualized
transport heat by both phase change in the working fluid and forced flow patterns. Spinato et al. reported on the visualization of flow regime
convection of liquid slugs. Consequently, compared with conventional in the PHP using the time-strip imaging process technique [14]. Flow
heat pipes, OHPs have many advantages, including higher heat trans­ regime, quantitative data of the liquid-vapor interface, and the
port capability and simple internal structures, which aid easy fabrication

* Corresponding author.
E-mail address: [email protected] (T. Daimaru).

https://doi.org/10.1016/j.applthermaleng.2021.117189
Received 12 November 2020; Received in revised form 31 May 2021; Accepted 1 June 2021
Available online 5 June 2021
1359-4311/© 2021 Elsevier Ltd. All rights reserved.
K. Odagiri et al. Applied Thermal Engineering 195 (2021) 117189

Nomenclature T temperature, K
u velocity, m/s
Acr cross-sectional area of flow path, m2 V volume, m3
Ca capillary number, –
cp specific heat at constant pressure, J/kg K Greek Symbols
cv specific heat at constant volume, J/kg K α thermal diffusivity, m2/s
D channel diameter, m μ viscosity, Pa s
h heat transfer coefficient, W/m2 K ρ density, kg/m3
HLG latent heat, J/kg σ surface tension, N/m
k thermal conductivity, W/m K Subscripts
l position, m cool cooling block
L length, m e evaporator
m mass, kg p casing plate
ṁ momentum, kg m/s f fluid
p perimeter, m in inner
P pressure, Pa l liquid
Q heat input, W sat saturated condition
q heat flux, W/m2 turn turn section
qapply applied heat flux, W/m2 v vapor
rc radius of maximum active cavity for boiling, m w channel wall
R gas constant, J/kg K τ friction
t time, s
tlf liquid film thickness, m

exchange between the vapor plug and pipe wall compared to the single-
flow vapor. This is because the mechanism of heat transfer in the thin
liquid film changes from forced convection to conduction. Numerical
results from this model are in good agreement with experimental results.
Bae et al. developed a numerical model that directly simulates the spatial
and temporal variations in the thickness of a thin liquid film [19]. The
dry out of the liquid film was also simulated by assuming it occurs when
the local film thickness becomes smaller than the surface roughness of
the tube wall. The model was validated by the experimental results for
horizontal and inclined PHPs. Daimaru et al. investigated the thermal
cycle in OHPs using a 1D slug flow model, fast Fourier transform (FFT),
and cross-correlation analysis [20]. They observed that pressure prop­
agation occurs in OHPs and the vapor plugs receive or exert energy in
the direction of the propagation. The pressure propagation was attrib­
uted to energy propagation in OHPs. Senjaya et al. proposed the first
Fig. 1. Schematic of an oscillating heat pipe.
nucleate-boiling model of OHPs [21,22]. The semi-empirical model was
developed based on experimental results. Implementing bubble gener­
temperature changes were compared and rationalized. Sun et al. worked ation in the model, it was obtained that the ratio of the heat transferred
on the visualization of the bubble motion and size distribution in the by latent heat to the total heat transfer is much higher than that in
PHP made of the glass tube [15]. The filling ratio of the working fluid conventional models (i.e., OHP model without bubble generation).
was also changed. It was found that the proportion of small bubbles Mameli et al. investigated the effects of local pressure losses due to the
increased with the filling ratio and applied power. Hemadri et al., presence of turns under various conditions [23]. The model showed that
investigated temperature distribution in a flat-plate PHP radiator by a local pressure losses affect heat transfer in OHPs operated at high heat
high-resolution infrared camera [16]. It was revealed that the degree of flux under the horizontal condition. The numerical results were vali­
isothermalization in the PHP radiators strongly depends on their effec­ dated with experimental data in the literature. Daimaru et al. developed
tive Biot number. In addition, it was mentioned that the relative a numerical model for CVOHP (OHP with check valves) [24]. The model
advantage of embedded PHP compared with normal plate became considers the boiling phenomena and pressure loss generated by the
higher in case that the thermal conductivity of the radiator plate ma­ check valves. In addition, it integrates the variable-length liquid-film
terial was lower. model with a single dry spot per vapor plug. Comparing the numerical
Shafii et al. studied a transient one-dimensional (1D) model for slug simulation results and the on-orbit experimental data, the startup
flows along the flow path in unlooped and looped OHPs [17]. The model behavior of CVOHP was investigated. Nekrashevych et al. studied a
solves the energy conservation in the vapor plugs and liquid slug, numerical model that includes the effect of heat conduction in the
respectively. Energy changes in the vapors cause changes in the tem­ channel walls, bubble generation, and thermal coupling between the
perature and pressure of the vapor plugs. This results in a local pressure fluid and the channel walls [25]. They found that the start-up of PHPs is
difference among the vapor plugs, which is the driving force for liquid impossible without boiling in a case where tube heat conduction occurs.
slug oscillation. This model has been improved over time. Das et al. In addition, they revealed that the mechanism of sustainable oscillation
studied the liquid film dynamics in a single-branch OHP [18]. They with a large amplitude is bubble generation. Whereas, the model did not
reported that a thin liquid film exists around the vapor bubbles due to consider the heat conduction between the parallel channels and in a
the strong shearing force in the model. This results in higher heat thickness direction. The effect of thermal diffusion in planar surface and

2
K. Odagiri et al. Applied Thermal Engineering 195 (2021) 117189

thickness direction on the flat-plate OHP performance was not clear. estimation of heat transfer coefficients in transient and turbulent flow
Many researchers have studied the complex thermo-fluid phenom­ and friction factor for liquid slugs in a laminar and turbulent flow.
ena in OHPs, including slug flow dynamics, liquid film dynamics, and For the 3D thermal diffusion in OHP casing plates, heat conduction
boiling heat transfer. However, the phenomenon in flat-plate OHPs (i.e., between discretized elements and heat exchange between discretized
the combined effect of thermo-fluid behavior in the channels and ther­ elements and the channels are considered. The OHP casing plate is
mal diffusion in the casing plate) is yet to be clarified. To the best of our discretized along the x, y, and z axes, respectively. Layers 1 and 5
knowledge, no study has been reported on the numerical investigation of partially have a boundary for heat input and cooling. Heat exchange
the coupled effect of two-phase thermal-hydraulics and 3D thermal between the casing plate and the channels is assumed to occur only in
diffusion. In this study, we developed a comprehensive numerical model layer 3. The model considers an artificial channel wall for easy imple­
that implements the combined effect of thermo-fluid phenomena in the mentation. The channel wall and discretized casing plate are assumed to
OHP channel and 3D thermal diffusion in the OHP casing plate. have sufficiently high thermal conductance. Hence, their thermal be­
Furthermore, we employed the model to investigate the heat transfer haviors are considered the same. In layers 2 and 4, it is assumed that
characteristics and heat propagation phenomenon of a flat-plate OHP. only heat conduction with other discretized elements occurs. The gov­
The originality of this paper is to numerically investigate comprehensive erning equations for the thermal diffusion in the OHP casing plate are as
phenomena that include thermal diffusion in the whole structure and follows:
thermal hydraulics in the channel. The model was validated with ( 2 )
experimental results of an additive-manufactured aluminum OHP. We ∂Tp ∂ Tp ∂2 Tp ∂2 Tp qapply
=α + + + for the heat input domain,
further investigated the effects of surface roughness of the channels and ∂t ∂x2 ∂y2 ∂z2 ρp cp,p dz
liquid film on the heat transfer performance of the OHP and heat (1)
propagation phenomena at startup and steady-state. ( ) ( )
∂Tp ∂2 Tp ∂2 Tp ∂2 Tp hcb Tp − Tcool
=α + 2 + 2 − for the cooling domain,
2. Numerical model ∂t ∂x2 ∂y ∂z ρp cp,p dz
(2)
Fig. 2 shows the schematic diagram of the developed model. For the ( )
thermo-fluid phenomena in the channel, a 1D model was employed ∂Tp
=α
∂2 Tp ∂2 Tp ∂2 Tp
+ 2 + 2 −
qp− w πD
for layer 3, (3)
along the flow path. The details of the model are presented in Ref. [24]. ∂t ∂x2 ∂y ∂z ρp cp,p dxdz
The model consists of the channel wall and a working fluid. The working
( )
fluid domain is composed of liquid slugs, vapor plugs, and liquid films, ∂Tp ∂2 Tp ∂2 Tp ∂2 Tp
=α + 2 + 2 for layer 2 and 4, (4)
as shown in Fig. 2. The channel wall and each liquid slug are discretized ∂t ∂x2 ∂y ∂z
along the flow path for temperature calculation. The discretized part of
the channel wall is fixed spatially, whereas the discretized element of the where Tp is the temperature of each discretized element of the OHP
liquids can move in the working fluid domain. The following assump­ casing plate, t the time, α, ρp , and cp,p the thermal diffusivity, density,
tions are applied to the model: (1) The slug flow that is uniform along and specific heat of the OHP casing plate, respectively. qapply is the
the radial direction occurs in the channel, (2) The liquid phase is applied heat flux, hcb the contact heat transfer coefficient between the
considered as incompressible flow, (3) The vapor phase is treated as an OHP and the cooling block, Tcool the cooling block temperature, D the
ideal gas, (4) Temperature and pressure in each vapor plug are uniform, channel diameter, and qp− w the heat flux between the channel and the
(5) Liquid films do not flow due to strong shear forces on the channel OHP casing plate.
inner surface. Friction between liquid films and vapor plugs is not For the liquid slug, the temperature change is calculated by Eq. (5).
considered, (6) Volume and mass changes in the liquid slugs caused by
the generation and absorption of liquid films and via phase change in a
liquid-vapor interface are not considered, (7) Correlations are used for

Fig. 2. Schematic diagram of the model.

3
K. Odagiri et al. Applied Thermal Engineering 195 (2021) 117189

[ ]
∂Tl ∂2 Tl ( )
cp,l ρl Acr = qw− f p + kl Acr 2 where, qw− f = hl,n Tw,k − Tl,n,i (5)
∂t ∂l n,i

Acr is the cross-sectional area of the flow path, kl is the thermal con­
ductivity of liquid, l is the position in the channel along the flow di­
rection, qw− f and hl,n are heat flux and heat transfer coefficient between
channel wall and liquid slug, respectively. Tw and Tl are the temperature
of the channel wall and the liquid slug, respectively. hl,n is calculated by
different equations depending on the flow regime that is determined by
the Reynolds number. The momentum and velocity are calculated by
Eqs. (6), (7).
[ ]
dml ul
= Acr (ΔPv − ΔPτ − ΔPturn ) , (6)
dt n

[ ]
ṁl
ul = , (7)
Acr Ll ρl
n

where ml and ul are mass and velocity of each liquid slug, ΔPv is the
pressure difference between adjacent vapor plugs, ΔPτ is the pressure
loss in the circular channel, ΔPturn is pressure loss in the turn sections.
For the vapor plug, the temperature change is calculated by Eq. (8).
In this equation, the changes of energy by means of sensible heat in the
dry region, phase change in the liquid film region, and work on vapor
plug for adjacent liquid slugs, are integrated.
[ ( )/ ]
dTv dVv
= Qv,sen + Qv,lat − Pv cv,v mv , (8)
dt dt j

where Tv , Pv , Vv , mv , and cv,v are temperature, pressure, volume, mass,

and specific heat at constant volume, respectively. Qv,sen is the total
amount of sensible heat, calculated by Eq. (9). Qv,lat is the total amount
of latent heat, calculated by Eq. (10). It is noted the mass change in Eq.
(10) is caused by the phase change.
[ ∫ dry ] { Fig. 3. Flowchart of calculation procedure.
( ) kl,k /tlf (Tw,k − Tsat,i )
Qv,sen = qw− f pin dl where, qw− f ,k = (9)
j
hv,j (Tw,k − Tv,i )
Table 1
[ ]
dmv Calculation condition.
Qv,lat = RTv (10)
dt j Thermal conductivity of AlSi10Mg k, W/m K 160
Density of AlSi10Mg ρ, kg/m3 2670
pin is the channel inner perimeter, tlf is liquid film thickness, Tsat is the Specific heat of AlSi10Mg cp, J/kg K 890
Time step dt, sec 0.0001
saturation temperature, hv is the heat transfer coefficient of forced Number of time steps for each heat load Ntime, – 3,000,000
convection that occurs in the region where the liquid film does not exist Computational time dt × Ntime, sec 300
around the vapor plug. Mesh size for channel dzchannel, mm 1
For the boiling phenomenon in the liquid phase, the superheat de­ Mesh size for plate dx, dy, dz, 3, 3, 0.8
mm
gree is used to determine the boiling criterion, as shown in Eqs. (11),
Initial temperature Tint, ◦ C 10
(12). Temperature of cooling block Tcb, ◦ C 10
( ) Contact heat transfer coefficient between plate and hcb,W/m2 K 2300
ΔTsat,k = Tw,k − Tsat Pl,n,i , (11) cooling block
Liquid film thickness tlf , m Variable
rc 2σTsat (Pl,n,i ) Radius of maximum active cavity for boiling rc, m Variable
ΔTsat,boiling = qex + , (12)
kl HLG ρv rc Initial number of liquid slugs Nl,int, – 30

where rc is the radius of the active cavity for boiling, σ is the surface
tension, HLG is latent heat. When the superheat degree is higher than 3. Experimental setup
ΔTsat,boiling calculated by Eq. (12), a new vapor plug is generated.
Fig. 4 shows the channel geometry of the OHP and a schematic di­
Fig. 3 shows a flowchart of the calculation procedure. The entire
agram of the experimental setup. The specifications of the OHP are listed
calculation loop is implemented using the fourth-order Runge-Kutta
in Table 2. The OHP was made of AlSi10Mg by AM technology using
method. Thermal diffusion in the OHP casing plate and heat exchange
direct metal laser sintering. The width, length, and thickness of the OHP
between the channel and the OHP casing plate are calculated before the
are 90, 200, and 3.8 mm, respectively. The channel has a circular cross-
thermo-fluid behavior in each time step. The thermal properties of the
section with a diameter of 1.0 mm. The total turn number is 42. R134a
working fluid are calculated by linking with REFPROP 9.0, which cal­
was used as the working fluid. The filling ratio of R134a was 45% of the
culates fluid properties as a function of temperature. The calculation
channel volume.
conditions are presented in Table 1.
The heat load was applied by a Kapton heater (OMEGA, KHA-303/5-
P) and a DC power supply (Sorensen XHR 300-2). The condenser was

4
K. Odagiri et al. Applied Thermal Engineering 195 (2021) 117189

Fig. 4. (a) Channel geometry of the OHP, (b) Schematic diagram of the experimental setup.

Table 2
Specification of the OHP.
OHP material AlSi10Mg
Inner channel diameter, mm 1.0
Radius of turns, mm 1.0
Number of turns, - 42
Total channel length, mm 7944
Width of OHP, mm 90
Length of OHP, mm 200
Thickness of OHP, mm 3.8
Working fluid R134a
Charging ratio, wt% 45

mounted to a cold plate that was connected to a chiller. The inlet tem­
perature of the cold plate was maintained at 10 ◦ C. To make the surfaces
of the plate smooth and enhance the contact heat transfer between the
OHP surfaces and heater/cold plate, the surfaces of the OHP plate were
machined. The OHP was covered with a thermal insulator. All experi­
ments were performed under a horizontal condition. The temperatures Fig. 5. Calculation results of ΔT at tlf = 5 μm and rc = 5, 10, and 15 μm.
of the evaporator, adiabatic section, and condenser were measured
using five E-type thermocouples. The temperature measuring points are
shown in Fig. 4. Temperature data were recorded every 2 s using the PC
connected to the data logger (keysight 34972A).

4. Result and discussion

4.1. Effect of liquid film thickness and surface roughness

The effects of two fitting parameters (the radius of the active cavity
for boiling, rc, and liquid film thickness, tlf) on the calculation results are
discussed in this section. Figs. 5–7 depicts the calculated the average
temperature difference between the evaporator and the condenser for
various active cavity radii and liquid film thicknesses. The surface
roughness varies and depends on AM process direction. Because the OHP
is manufactured in various process directions, the channel inner surfaces
have various roughness, however, the representative roughness is within
the range from 5 to 15 μm. These values were used as the cavity radius
for nucleate boiling. The cavity and nucleation location are distributed
uniformly in the channel. Thus, the distribution of the nucleation site is Fig. 6. Calculation results of ΔT at tlf = 10 μm and rc = 5, 10, and 15 μm.
assumed to be uniform in the channel. In this model, the liquid film
thickness, tlf is a fitting parameter because tlf is a function of the velocity μl ul
of liquid slug which changes temporally and spatially, and it is difficult Cal = (14)
σl
to predict the velocity. The film thickness can be expressed by Eqs. (13),
(14). [26,27] where D is channel diameter and Cal is the capillary number, μl is the
2/3 liquid viscosity, ul is the velocity of liquid phase, σl is the surface tension.
0.67Cal
tlf = D (13) In case that the liquid slug velocity ranges from 0.01 to 10 m/s, the
1 + 3.35Cal 2/3

5
K. Odagiri et al. Applied Thermal Engineering 195 (2021) 117189

Fig. 8. Correlated calculation and experimental results of evaporator and

Fig. 7. Calculation results of ΔT at tlf = 25 μm and rc = 5, 10, and 15 μm.
condenser average temperature.

liquid film thickness tlf is in the range from 2.7 to 116 μm based on the
layer 1 (i.e., heating surface) at the start-up in the case of 100 W. The
Eqs (13), (14). Based on the fitting, tlf = 5–25 μm are chosen as the
indicated range of temperature is from 10 to 45 ◦ C. The temperature of
appropriate values in this study. Figs. 5–7 show the change in temper­
the evaporator increased from the center to the outside at t = 1–5 s,
ature (ΔT) at tlf = 5, 10, and 25 μm, respectively, and rc = 5, 10, and 15
similar to thermal diffusion in a solid plate without OHP. At t = 5 s, the
μm.
temperature of the condenser increased nonuniformly in the x-axis. This
For tlf = 5 μm, the smallest ΔT was obtained at rc = 5 μm and the
is caused by the movement of liquid slugs and vapor plugs in the OHP.
largest at rc = 10 μm. The difference in ΔT in all the cases was relatively
After 10 s, areas with high local temperature appeared in the evaporator
small for all heat loads. Compared with the experimental results, ΔT was
and move around.
much higher. In the case of tlf = 10 μm, the difference in ΔT was small at
Fig. 10 and Video 2 show changes in the temperature and liquid­
50–100 W but increased at 150–200 W. ΔT was the smallest at rc = 10
–vapor distributions at steady-state in the case of 100 W. In the liquid­
μm and the largest at rc = 15 μm. In addition, ΔT was smaller than that of
–vapor distribution diagram, the blue, light blue, and white areas
the experiment at 50–150 W, however it became larger than the
represent the liquid slug, vapor plug with liquid film, and vapor plug
experimental values in the case of rc = 5 and15 μm at 200 W. In the case
without liquid film, respectively. At steady-state, areas with higher local
of tlf = 25 μm, the difference in ΔT was small at all rc and heat loads. ΔT
temperature are observed in both the evaporator and condenser. The
was the smallest at rc = 5 μm and the largest at rc = 15 μm.
temperature distribution was not uniform in the evaporator and
The liquid film thickness affects the thermal resistance between the
condenser. The temperature of the upper edge was slightly higher than
channel wall and the liquid–vapor interface when evaporation and
that of other areas in the evaporator. On the other hand, the temperature
condensation occur. The thermal resistance decreases as the film
of the lower edge was slightly lower than that of other areas in the
thickness decreases. Thus, the thinner liquid film results in smaller ΔT.
condenser. This trend indicates that the temperature distribution is
On the other hand, thinner liquid film also results in the generation of
determined by the combined effect of heat transfer and heat conduction
the local dry-out in the evaporator.
in the OHP casing plate. In this model, the temperature distribution
The radius of the active cavity, rc affects the generation of nucleate
along the thickness was also calculated. At 100 W, the average evapo­
boiling. As rc increases, the superheat required for the nucleation be­
rator temperatures of layers 1 and 5 were 29.4 ◦ C and 29.3 ◦ C, respec­
comes smaller. Larger rc generate more active boiling in the evaporator,
tively; i.e., a temperature difference of 0.1 ◦ C between the layers. The
and this results in the drastic growth of bubbles and movement of the
much smaller temperature difference along the thickness is attributed to
liquid slugs in the OHP. Furthermore, the drastic growth of bubbles and
the effect of heat transfer in the OHP and the high thermal diffusivity of
active oscillation results in frequent local dry-out in the evaporator. In
aluminum.
the case of tlf = 5 μm, the smallest ΔT between the evaporator and
Comparing the temperature and liquid–vapor distributions in the
condenser was obtained at the minimum rc. This is because small rc
evaporator at t = 202 sec, we found that the temperature partially
prevents the generation of local dry-out. This trend is consistent with
increased in the area containing vapor plugs without liquid film. On the
that reported by Kim et al. [28]. The four PHPs that have different cavity
other hand, the temperature slightly decreased in the area through
sizes such as 10, 20, 30, and 40 μm were tested. Large cavity size
which the liquid slug flows from the condenser. At t = 204 s, most of the
enhanced the boiling in the channel. On the other hand, the partial dry-
channels in the evaporator were filled with the vapor plug without
out was observed at lower heat flux for the case of large cavity size.
liquid film. Based on Video 2, this phenomenon occurs temporarily
Fig. 8 shows the average evaporator and condenser temperatures for
within the nominal OHP operation. The area where the temperature is
the correlated calculation and experimental results. Based on the cor­
locally higher than the other areas appears in this case. This is because
relation, the liquid film thickness and radius of the active cavity for
the thermal resistance between the channel wall and fluid is large in the
boiling were 18 and 5 μm, respectively. Hence, the model well repro­
case of vapor plug without liquid films, which gives rise to high heat
duced the operating temperatures of the experiment. The calculation
transfer coefficient. Notably, in some cases, the temperature of the
results at tlf = 18 μm, which is a relatively large film thickness, agree
evaporator became high even in the areas where the liquid slugs exist.
well with the experimental results at high heat loads. It is considered
This is because the temperature response of the OHP casing plate from
that at high heat loads, liquid films are thick and dry-out does not occur.
the channels is delayed. In addition, thermal diffusion in the OHP casing
plate reduces the effect of temperature changes caused by the liquid slug
4.2. Changes in temperature distribution at start-up and steady-state flowing in. It is found that the surface temperature distribution of the
OHP plate is determined by the combined effect of OHP thermo-fluid
Fig. 9 and Video 1 show changes in the temperature distribution of behaviors and thermal diffusion in the OHP casing plate. Considering

6
K. Odagiri et al. Applied Thermal Engineering 195 (2021) 117189

Fig. 9. Temperature distribution in layer 1 at start-up (Q = 100 W).

45
40

Temperature, ºC
35
30
25
20
15
10

Liquid Slug

Vapor Plug
w/ L.F.

Vapor Plug
w/o L.F.
L.F.
: Liquid Film

202 sec 204 sec 206 sec 208 sec 210 sec
Fig. 10. Temperature distribution and liquid–vapor distribution at steady-state (Q = 100 W).

the applications of flat-plate OHPs, it is necessary to consider the tem­ observe the behavior of the hot spots, the binarized figures with ΔTHS,
perature gradient in both the evaporator and condenser, even when the which is defined by Eq. (15), are shown in Fig. 11.
OHP exhibits extremely high heat transfer.
ΔTHS,i = Te,i − Te,ave (15)
As shown in Fig. 10, areas with high local temperature (herein
referred to as hot spots) were observed in the evaporator. To closely

7
K. Odagiri et al. Applied Thermal Engineering 195 (2021) 117189

Evaporator Hot spot

201 sec 202 sec 203 sec 204 sec

Condenser

205 sec 206 sec 207 sec 208 sec

Fig. 11. Behavior of hot spots at steady-state (Q = 100 W).

where Te,i is the evaporator temperature of each discretized element at temperature distribution in the planar surface because thermal resis­
each time step and Te,ave is the average temperature of the evaporator at tance between parallel channels becomes larger. The quantitative
each time step. In Fig. 11 and Video 3, the threshold of ΔTHS is set to 2 ◦ C evaluation of the effect of the channel density on the 3D temperature
for convenience to express the behavior of hot spots. ΔTHS can be defined distribution will be conducted as future work.
as desired for various cases. The black areas depict areas where ΔTHS is The application of this work is a tool for the prediction of heat
higher than 2 ◦ C. It is observed that the hot spots move around the OHP transfer performance, temperature distribution, and hot spot generation
plate surface. The mechanism of hot spot generation is explained by the for flat-plate OHPs. In recent years, AM technology has been remarkably
liquid–vapor distribution in the OHP. Comparing with Fig. 10, it is ob­ growing. Various materials such as aluminum, stainless steel, titanium,
tained that the hot spots are generated in the areas containing the vapor Inconel, and ceramics will be applied to the OHP material in near future.
plugs without the liquid film. Large maximum ΔTHS were observed even when the material has rela­
Fig. 12 shows the changes in the maximum ΔTHS and the ratio of ΔTHS tively high thermal diffusivity. The issue of moving hot spots thus be­
to the average evaporator temperature at various heat loads. The comes significant in case that materials with low thermal diffusivity are
maximum ΔTHS at 50, 100, 150, and 200 W were 1.8 ◦ C, 3.0 ◦ C, 3.9 ◦ C, employed to OHPs. This is attributed to the increase in maximum ΔTHS
and 5.4 ◦ C, respectively. The maximum ΔTHS increased with an increase as the thermal diffusivity decreases. In case that the low thermal diffu­
in the heat load. On the other hand, there was no much change in the sivity materials are used for the OHP casing, the effect of 3D thermal
ratio of ΔTHS; it ranged from 9 to 11%. diffusion becomes important and is required to be implemented in the
So far, it has been reported that the higher channel density (i.e., the model. The proposed model can be used for the optimization of the
ratio of the area occupied by channels to casing plate cross-sectional design parameters such as OHP thickness, number of turns, and thermal
area) enhances the OHP heat transfer capability and the effective ther­ diffusivity of the casing for various configurations. In addition, it will be
mal conductivity [29]. Lower channel density results in less isothermal useful for enhancing the feasibility of implementing the OHPs to devices
for space and terrestrial fields.
6 40
5. Conclusion
Ratio of ΔT to evaporator temperature, %

35
5
This paper presents three-dimensional heat transfer analysis for a
Maximum ΔT of hot spot, K

30
4 flat-plate oscillating heat pipe (OHP). The novelty of this paper is to
ΔTmax
25 numerically investigate the combined effect of three-dimensional ther­
3 20 mal diffusion in the casing and the thermo-fluid behaviors in the
channels. The conclusions are summarized as follows:
15
2
10 • A numerical model combining the thermo-fluid behavior in the
1 Ratio of ΔTmax channel and 3D thermal diffusion in OHP casing plates was devel­
5 oped. The calculation results agree well with the experimental results
0 0 on a multi-branch OHP having a channel diameter of 1.0 mm and a
0 50 100 150 200 250 turn number of 42.
Heat load, W • The effect of liquid film thickness and active cavity radius for boiling
on the heat transfer performance of the OHP was investigated. Nine
Fig. 12. Maximum ΔTHS of hot spot and ratio of ΔTHS to the average evaporator cases were considered with liquid film thicknesses of 5, 10, and 25
temperature.

8
K. Odagiri et al. Applied Thermal Engineering 195 (2021) 117189

μm and surface roughness of 5, 10, and 15 μm. For tlf = 5, 25 μm, the [9] K.K. Mehta, N.S. Mehta, Development of flat plate oscillating heat pipe as a heat
transfer device, Front. Heat Pipes 7 (6) (2016) 1–7, https://doi.org/10.5098/
smallest ΔT between the evaporator and condenser was obtained at
fhp.7.6.
the minimum rc. This is because small rc prevents the generation of [10] A. Yoon, S. Kim, Characteristics of oscillating flow in a micro pulsating heat pipe:
local dry-out. Fundamental-mode oscillation, Int. J. Heat Mass Transfer 109 (2017) 242–253,
• The relation between the temperature distribution of OHP surface https://doi.org/10.1016/j.ijheatmasstransfer.2017.02.003.
[11] J. Kim, S. Kim, Experimental investigation on the effect of the condenser length on
and the liquid–vapor behavior in the channels were explained. It was the thermal performance of a micro pulsating heat pipe, Appl. Therm. Eng. 130
found that the surface temperature distribution of the OHP was (2018) 439–448, https://doi.org/10.1016/j.applthermaleng.2017.11.009.
determined by the combined effect of thermo-fluid behaviors and the [12] W.W. Wits, G. Groeneveld, H.J. Gerner, Experimental investigation of a flat-plate
closed-loop pulsating heat pipe, J. Heat Transfer. 141 (9) (2019), 091807, https://
thermal diffusion in the casing plate. doi.org/10.1115/1.4042367.
• The generation of the moving hot spots, local high temperature areas [13] M. Mameli, M. Marengo, S. Khandekar, Local heat transfer measurement and
on the evaporator planar surface is quantitatively simulated. In the thermo-fluid characterization of a pulsating heat pipe, Int. J. Therm. Sci. 75 (2014)
140–152, https://doi.org/10.1016/j.ijthermalsci.2013.07.025.
case of the aluminum OHP, the ratio of the maximum superheat of [14] G. Spinato, N. Borhani, B.P. d’Entremont, J.R. Thome, Time-strip visualization and
hot spots to the average evaporator temperature became 9 – 11% at thermo-hydrodynamics in a Closed Loop Pulsating Heat Pipe, Appl. Therm. Eng. 78
various heat loads. (2015) 364–372, https://doi.org/10.1016/j.applthermaleng.2014.12.045.
[15] Q. Sun, J. Qu, X. Li, J. Yuan, Experimental investigation of thermo-hydrodynamic
behavior in a closed loop oscillating heat pipe, Exp. Therm. Fluid Sci. 82 (2017)
Declaration of Competing Interest 450–458, https://doi.org/10.1016/j.expthermflusci.2016.11.040.
[16] V.A. Hemadri, A. Gupta, S. Khandekar, Thermal radiators with embedded pulsating
heat pipes: Infra-red thermography and simulations, Appl. Therm. Eng. 3 (6–7)
The authors declare that they have no known competing financial
(2011) 1332–1346, https://doi.org/10.1016/j.applthermaleng.2011.01.004.
interests or personal relationships that could have appeared to influence [17] M.B. Shafii, A. Faghri, Y. Zhang, Thermal modeling of unlooped and looped
the work reported in this paper. pulsating heat pipes, J. Heat Transf. 123 (2001) 1159–1172, https://doi.org/
10.1115/1.1409266.
[18] S.P. Das, V.S. Nikolayev, F. Lefevre, B. Pottier, S. Khandekar, J. Bonjour, Thermally
Acknowledgement induced two-phase oscillating flow inside a capillary tube, Int. J. Heat Mass Transf.
53 (19-20) (2010) 3905–3913, https://doi.org/10.1016/j.
This research was conducted at the Jet Propulsion Laboratory, Cal­ ijheatmasstransfer.2010.05.009.
[19] J. Bae, S. Lee, S. Kim, Numerical investigation of effect of film dynamics on fluid
ifornia Institute of Technology, USA, under internal research and tech­ motion and thermal performance in pulsating heat pipes, Energy Convers and
nology development funding. Manage 151 (2017) 296–310, https://doi.org/10.1016/j.enconman.2017.08.086.
[20] T. Daimaru, S. Yoshida, H. Nagai, Study on thermal cycle in oscillating heat pipes
by numerical analysis, Appl. Therm. Eng. 113 (2017) 1219–1227, https://doi.org/
Appendix A. Supplementary material 10.1016/j.applthermaleng.2016.11.114.
[21] R. Senjaya, T. Inoue, Oscillating heat pipe simulation considering bubble
Supplementary data to this article can be found online at https://doi. generation part I: presentation of the model and effects of a bubble generation, Int.
J. Heat Mass Transf. 60 (2013) 816–824, https://doi.org/10.1016/j.
org/10.1016/j.applthermaleng.2021.117189. ijheatmasstransfer.2013.01.059.
[22] R. Senjaya, T. Inoue, Oscillating heat pipe simulation considering bubble
References generation part II: effects of fitting and design parameters, Int. J. Heat Mass Transf.
60 (2013) 825–835, https://doi.org/10.1016/j.ijheatmasstransfer.2013.01.058.
[23] M. Mameli, M. Marengo, S. Zinna, Numerical model of a multi-turn Closed Loop
[1] H. Akachi, U.S. Patent, Patent Number 4921041, May 1, 1990.
Pulsating Heat Pipe: Effects of the local pressure losses due to meanderings, Int. J.
[2] D. Mangini, M. Mameli, D. Fioriti, S. Filippeschi, L. Araneo, M. Marengo, Hybrid
Heat Mass Transfer 55 (4) (2012) 1036–1047, https://doi.org/10.1016/j.
Pulsating Heat Pipe for space applications with non-uniform heating patterns:
ijheatmasstransfer.2011.10.006.
Ground and microgravity experiments, Appl. Therm. Eng. 126 (2017) 1029–1043,
[24] T. Daimaru, H. Nagai, M. Ando, K. Tanaka, A. Okamoto, H. Sugita, Comparison
https://doi.org/10.1016/j.applthermaleng.2017.01.035.
between numerical simulation and on-orbit experiment of oscillating heat pipes,
[3] A. Okamoto, M. Ando, H. Sugita, Initial evaluation of on-orbit experiment of flat-
Int. J. Heat Mass Transfer 109 (2017) 791–806, https://doi.org/10.1016/j.
plate heat pipe, in: Proceedings of the 17th International Heat Pipe Conference,
ijheatmasstransfer.2017.01.078.
Kanpur, India, 2013.
[25] I. Nekrashevych, V.S. Nikolayev, Reprint of: Effect of tube heat conduction on the
[4] M. Ando, A. Okamoto, K. Tanaka, M. Maeda, H. Sugita, T. Daimaru, H. Nagai, On-
pulsating heat pipe start-up, Appl. Therm. Eng. 126 (2017) 1077–1082, https://
orbit demonstration of oscillating heat pipe with check valves for space
doi.org/10.1016/j.applthermaleng.2017.08.117.
application, Appl. Therm. Eng. 130 (2018) 552–560, https://doi.org/10.1016/j.
[26] F. Bretherton, The Motion of Long Bubbles in Tubes, J. Fluid Mech. 10 (1961)
applthermaleng.2017.11.032.
166–188.
[5] P. Charoensawan, S. Khandekar, M. Groll, P. Terdtoon, Closed loop pulsating heat
[27] P. Aussillous, D. Quéré, Quick deposition of a fluid on the wall of a tube, Phys.
pipes part A: parametric experimental investigations, Appl. Therm. Eng. 23 (16)
Fluid 12 (10) (2000) 2367–2371.
(2003) 2009–2020, https://doi.org/10.1016/S1359-4311(03)00159-5.
[28] W. Kim, S. Kim, Effect of reentrant cavities on the thermal performance of a
[6] S. Khandekar, P. Charoensawan, M. Groll, P. Terdtoon, Closed loop pulsating heat
pulsating heat pipe, Appl. Therm. Eng. 133 (2018) 61–69, https://doi.org/
pipes part B: visualization and semi-empirical modeling, Appl. Therm. Eng. 23 (16)
10.1016/j.applthermaleng.2018.01.027.
(2003) 2021–2033, https://doi.org/10.1016/S1359-4311(03)00168-6.
[29] J. Qu, Q. Sun, H. Wang, D. Zhang, J. Yuan, Performance characteristics of flat-plate
[7] S.M. Thompson, H.B. Ma, Effect of localized heating on three-dimensional flat-
oscillating heat pipe with porous metal-foam wicks, Int. J. Heat Mass Transf. 137
plate oscillating heat pipe, Adv. Mech. Eng. 2 (2010) 465153, https://doi.org/
(2019) 20–30, https://doi.org/10.1016/j.ijheatmasstransfer.2019.03.107.
10.1155/2010/465153.
[8] F.F. Laun, H. Lu, H.B. Ma, An experimental investigation of an oscillating heat pipe
heat spreader, J. Thermal Sci. Eng. Appl. 7 (2) (2015), 021005, https://doi.org/
10.1115/1.4026815.