International Journal of Heat and Mass Transfer 49 (2006) 1833–1844

www.elsevier.com/locate/ijhmt

A numerical model for heat sinks with phase change materials

and thermal conductivity enhancers
K.C. Nayak, S.K. Saha, K. Srinivasan, P. Dutta *

Department of Mechanical Engineering, Indian Institute of Science, Bangalore 560012, India

Received 23 March 2005

Available online 19 January 2006

Abstract

The effectiveness of thermal conductivity enhancers (TCEs) in improving the overall thermal conductance of phase change materials
(PCMs) used in cooling of electronics is investigated numerically. With respect to the distribution of TCE and PCM materials, the heat
sink designs are classified into two types. The first type of heat sink has the PCM distributed uniformly in a porous TCE matrix, and the
second kind has PCM with fins made of TCE material. A transient finite volume method is used to model the heat transfer; phase change
and fluid flow in both cases. A generalized enthalpy based formulation and numerical model are used for simulating phase change pro-
cesses in the two cases. The performance of heat sinks with various volume fractions of TCE for different configurations is studied with
respect to the variation of heat source (or chip) temperature with time; melt fraction and dimensionless temperature difference within the
PCM. Results illustrate significant effect of the thermal conductivity enhancer on the performance of heat sinks.
 2005 Elsevier Ltd. All rights reserved.

Keywords: Heat sink; Phase change material; Thermal conductivity enhancer; Modeling

1. Introduction as Eicosane) also possess the undesirable property of low

thermal conductivity, which results in chip overheating.
As each new generation of electronic products squeezes This phenomenon is also known as the ‘‘self-insulating’’
more power and performance into ever-small packages, the effect of PCMs. In order to make a PCM suitable for elec-
relative importance of thermal management within the tronics cooling applications, it becomes necessary to add
overall product design continues to increase. In recent some amount of high thermal conductivity material (such
years, PCMs have been widely used as an alternative cool- as aluminium or copper), which are known as thermal con-
ing method for various applications such as spacecraft and ductivity enhancers (TCEs).
avionics thermal control, personal computing and commu- Comprehensive descriptions of experimental and com-
nication equipment, wearable computers, power electronic putational assessment of PCM-based thermal management
equipment, portable phones etc. Passive thermal manage- for different electronic systems are presented in [2–7].
ment scheme of PCMs can be used in situations where heat Among the studies regarding enhancement of heat transfer
dissipation is periodic or sudden transient. The selected through PCMs using TCEs, Tong et al. [8] studied the pos-
PCM should have melting temperature below the device’s sibility of heat transfer enhancement of a low thermal con-
maximum operating temperature, a high latent heat of ductive PCM (water/ice system, in this case) by inserting an
fusion per unit mass, a high thermal conductivity, a high aluminum matrix. Humphries and Griggs [9], Witzman
specific heat and a small volume change on phase change et al. [3] and Snyder [5] examined the effect of using flat fins
[1]. However, most PCMs with low melting point (such for thermal enhancement. Snyder [5] modeled a finned
PCM unit by using effective thermo-physical properties
*
Corresponding author. Tel.: +91 80 2293 3225; fax: +91 80 2360 4536. and neglecting melt convection. Abhat [1] performed a
E-mail address: [email protected] (P. Dutta). combined experimental and computational study of a

0017-9310/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2005.10.039
1834 K.C. Nayak et al. / International Journal of Heat and Mass Transfer 49 (2006) 1833–1844

Nomenclature

A coefficient in Darcy source term Greek symbols

aP, a0P coefficients in discretization equation a thermal diffusivity
b constant in the porosity source term b thermal expansion coefficient
c specific heat q density
C inertia coefficient in Forchheimer’s extension / porosity of the metal matrix
d diameter of the TCE matrix l dynamic viscosity
Da Darcy number DE change in enthalpy
e sensible enthalpy h dimensionless temperature
E enthalpy s dimensionless time
fs solid fraction
fl liquid fraction Subscripts
g gravitational acceleration amb atmospheric
h heat transfer coefficient i tensor variable
H height of the thermal control unit max maximum value
k thermal conductivity min minimum value
K permeability melt value at melting temperature
l length of the thermal storage unit m TCE
L latent heat of fusion of PCM n level of iteration
M morphology constant p PCM
q_ heat generation rate 0 initial state
Q chip power old old iteration value
S source term ref reference state
t time
T temperature
u velocity
x, y, z coordinate axes

thermal control unit, where PCM was incorporated in hex- the melted PCM. Essentially, two types of TSU’s are mod-
agonal honeycomb cells aligned parallel to the heating sur- eled: one with the PCM distributed uniformly in a porous
face. Recently, Pal and Joshi [10] investigated melting of TCE matrix and the other with the PCM having fins of
PCM inside honeycomb cores mounted such that the cells TCE material. Heat transfer enhancement with various
are vertical. Numerical simulation was performed both for volume fractions of TCE for different configurations is
a single cell and at the system level, considering the effect of studied with respect to the variation of heat source (or
natural convection in the PCM. The system level simula- chip) temperature; melt fraction and dimensionless temper-
tion results were in good agreement with the experimental ature difference within the PCM, with time.
data. Alawadhi and Amon [11] investigated the effective-
ness of a thermal control unit (TCU) made of PCM and 2. Mathematical model
aluminum fins for portable electronic devices.
It may be observed from the above discussion that sub- The mathematical model is described with respect to two
stantial amount of work has been reported on thermal physical situations. For this purpose, two basic types of
management of electronics using PCM and metal fins. It TCE distribution have been studied. In the first kind (Case
is also apparent that there is a need for finding an effective 1), the heat sink is designed using a highly porous TCE
method for heat transfer enhancement of PCMs by insert- matrix homogeneously distributed within the entire PCM.
ing high thermal conductivity materials. The objective of In the second kind (Case 2), the TCE is arranged in the
the present work is to establish a numerical model for form of fins protruding out of the base of the heat sink,
studying the effectiveness of thermal storage units (TSUs) while the rest of the space is filled with the PCM. For the
with PCM and TCE arrangement. The model can be suit- latter arrangement, both plate-type and rod-type (pin) fins
ably applied to find the optimum proportion of TCE mate- are considered. The two arrangements are shown schemat-
rials for a given configuration of TSU. In this work, a ically in Fig. 1. In each of the above configurations, con-
generalized computational model is developed to assess duction and convection (as a result of melt flow) are
the heat transfer performance of some common types of taken as the mode of heat transfer within the composite
TCE distribution considering conduction heat transfer in heat sink. Mathematical models for the two cases are out-
the solid regions and natural convection heat transfer in lined below.
 K.C. Nayak et al. / International Journal of Heat and Mass Transfer 49 (2006) 1833–1844 1835

Fig. 1. Numerical models (a) matrix-type TSU, (b) plate-type TSU, (c) and (d) rod-type TSU. All dimensions are in mm.

2.1. Case 1: PCM with metal matrix (porous medium are assumed to be two-dimensional. Based on the volume-
analysis) averaging technique, a common set of governing conserva-
tion equations for both the solid and liquid regions of the
In this case, we consider a rectangular cavity filled with a porous medium can be written as follows:
highly porous and permeable TCE (aluminum) matrix and Conservation of mass
PCM, distributed homogeneously. The thermal storage oðqp ui Þ
unit (consisting of PCM and TCE) is kept on an aluminum ¼0 ð1Þ
oxi
substrate, which, in turn, seats over an electronic chip or a
heat source. Initially, the whole system is kept at ambient Conservation of momentum
temperature, Tamb. The porosity of the metal matrix, /m, The equivalent singe phase momentum conservation
is assumed to be equal to the volume fraction of the equation for the porous medium in the ith direction is given
PCM. The boundary conditions are such that the side by
and bottom faces are insulated and the top face is subjected oðqp ui Þ oðqp ui uj Þ op
to convective cooling. We consider constant heat genera- þ ¼ þ r  ðlrui Þ þ qp gi bðT  T ref Þ
ot oxj oxi
tion at the heat source, which is distributed uniformly  
throughout the chip. This heat source (chip) requires cool- l qC
 þ pffiffiffiffi jui j ui ð2Þ
ing. As heating goes on, melting begins within the PCM by K K
absorbing latent heat of fusion along with the sensible heat where gi is the acceleration due to gravity in the ith direc-
from the chip. But the TCE always remains in the solid tion, Tref is the reference value of temperature, C is the
state and absorbs only sensible heat. Local thermodynamic inertia coefficient in Forchheimer’s extension and K is per-
equilibrium between the PCM and the porous aluminium meability. The permeability, K, and porosity, /, are corre-
matrix is assumed. For the present study, Eicosane is lated approximately by the Blake–Kozeny model [8], and is
selected as the PCM material, primarily because of its given by
high value of latent heat of fusion. Thermo-physical prop-
d 2 /3
erties of the homogeneous mixture of PCM and TCE K¼ 2
ð3Þ
matrix are calculated using a volume-averaging technique 175ð1  /Þ
[12,13]. Because of the geometry and boundary conditions where / is the porosity and d is the TCE matrix diameter.
chosen, the heat transfer and fluid flow analyses in this case The value of / depends on the total solid fraction of the
1836 K.C. Nayak et al. / International Journal of Heat and Mass Transfer 49 (2006) 1833–1844

medium. When the PCM is in a complete molten state, which is approximated as a porous medium whose porosity
/ = /m, where /m is the porosity of the metal matrix. is a direct function of the local liquid fraction. In the
Otherwise, when the PCM is fully or partially solid, / is momentum conservation equations, a flow resistance
modified according to the enthalpy porosity formulation source term as a function of porosity (and hence of solid
for phase change, which is described subsequently. fraction) is included. This source term ensures zero velocity
The value of the inertia coefficient C has been obtained in the fully solid region and permits normal viscous liquid
experimentally by Ward [14]. Although C is a function of flow in the fully liquid region. For phase change of a pure
the microstructure of the porous medium [15], Ward [14] substance, however, the sharp solid–liquid interface is rep-
found that for a large variety of porous media, it could resented by a numerical mushy region having a thickness of
be taken as a constant (0.55). only one control volume, across which there is a smooth
The term juij used in Eq. (2) is defined as variation of velocity and thermo-physical properties. It
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi may be recognized here that flow in a pure substance phase
jui j ¼ u21 þ u22 ð4Þ change process can be modeled alternatively by an effective
viscosity method. The present method, however, offers the
In our study we have used dl  102 , where l is the length
added advantage of straightforward extension to binary
of thermal storage unit.
mixtures, which undergo phase change within a range of
temperatures.
2.1.1. Energy conservation
The enthalpy formulation for the porous media energy
The single-phase energy conservation equation for por-
equation (5) leads to the latent heat source term, Sh, which
ous media is given by
is given by
oðqceff T Þ oðqceff ui T Þ
þ ¼ r  ðk eff rT Þ þ S h ð5Þ o
ot oxi Sh ¼  ðqDEÞ ð8Þ
ot
where qceff and keff are the effective thermal capacitance
and effective thermal conductivity, respectively, of the por- where
ous medium. The term Sh is the latent heat source term, in DE ¼ 0; T < T melt
which the subscript ‘h’ represents enthalpy. The effective ð9Þ
thermal capacitance is calculated, in the following manner, ¼ /m L; T > T melt
using the well known volume-averaged technique:
where Tmelt is the melting temperature of PCM, and L is its
qceff ¼ qp cp /m þ qm cm ð1  /m Þ ð6Þ latent heat of fusion.
It is important to note that the latent enthalpy content
The effective thermal conductivity for the porous med-
of the control volume in the porous medium is /mL after
ium is determined, in the following manner, using Veinberg
melting. This is so because a fixed volume fraction
model [8,16]:
(1  /m) of TCE matrix is always present, which absorbs
 1=3
k eff only sensible heat. This enthalpy value is to be accurately
k eff þ ðk m  k p Þ/m  k m ¼ 0 ð7Þ updated in every iteration within a particular time step.
kp
Since the chip and substrate absorb only sensible heat,
where km is the thermal conductivity of TCE matrix and kp DE for these regions should be set to zero.
is the thermal conductivity of the PCM. Referring to the expression for permeability given in Eq.
(3), it may be noted that for the molten control volumes,
2.1.2. Enthalpy–porosity formulation for phase change liquid fraction is /m, while for fully solid zones this is equal
In this model, we use a single domain enthalpy–porosity to zero. Hence, the solid region permeability should ideally
technique (appropriately modified) for modeling phase become zero. In order to model this effect, the effective
change and fluid flow in the porous medium. The advan- porosity of the TCE matrix is decreased, depending on
tage of this model is that no explicit velocity or thermal the solid fraction of the PCM. The solid fraction, fs, of
conditions are to be satisfied at the solid liquid interface. the PCM can be expressed as
A detailed description of this approach is available in
[17]. Essentially, in this approach, the solid–liquid interface DE
fs ¼ 1  ðliquid fraction of PCMÞ ¼ 1  ð10Þ
in a phase change process is assumed to be a mushy region, /m L

Table 1
Thermo-physical properties of the PCM used
Eicosane Density Kinematic Thermal Specific heat Melting Latent b (1/K)
(kg/m3) viscosity (m2/s) conductivity (W/m K) (J/kg K) point (C) heat (J/kg)
Solid 810 0.39 1900
Liquid 770 4.5e6 0.157 2200
Value used 790 4.5e6 0.23 2050 37 241,000 103
 K.C. Nayak et al. / International Journal of Heat and Mass Transfer 49 (2006) 1833–1844 1837

Table 2 Hence, the effective porosity of the medium can be ex-

Thermo-physical properties of TCE (or substrate) and chip used pressed as
Density Thermal Specific heat  
(kg/m3) conductivity (W/m K) (J/kg K)
DE
/ ¼ /m  1  / ¼ /m  ð1  fs Þ/m ð11Þ
TCE/substrate 2712.9 179.6 960
/m L m
Chip 740 20 1830 As a consequence of this, the coefficient of the source
term in Eq. (3) becomes very large when the PCM is in

Fig. 2. Temperature and velocity field evolution of PCM melting with TCE matrix (/ = 0.85, Da = 4.74 · 105): (a) 2100 s; (b) 2700 s and (c) 3600 s at
4 W. Umax in above cases are (a) 3 · 106 m/s; (b) 2.98 · 104 m/s; (c) 2.73 · 104 m/s.
1838 K.C. Nayak et al. / International Journal of Heat and Mass Transfer 49 (2006) 1833–1844

1.0 1.0

0.8 0.8
Melt fraction

Melt fraction
0.6 0.6
No fin
0.4 0.4
1.96E-04 5%
0.2 7.84E-04 0.2 10%
2.E-03 15%
0.0 0.0
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350 400
(a) Dimensionless time (a) Dimensionless time
100
100
Dimensionless maximum
temperature difference

Dimensionless maximum
80

temperature difference
80
60
60

40 No fin
40
1.96E-04 5%
20 7.84E-04 10%
20
2.E-03
15%
0
0 100 200 300 0
0 50 100 150 200 250
(b) Dimensionless time
(b) Dimensionless time
80 80
Chip temperature (οC)

70
Chip temperature (οC)

60 60

50
No fin
40 1.96E-04
40 5%
7.84E-04 10%
30
2.E-03 15%
20 20
0 600 1200 1800 2400 3000 3600 0 600 1200 1800 2400 3000 3600
(c) Time (s) (c) Time (s)

Fig. 3. Effect of Darcy number (Da) on (a) melt fraction, (b) uniformity Fig. 4. Effect of TCE matrix amount on (a) melt fraction, (b) uniformity
and (c) chip temperature at 4 W. and (c) chip temperature (for d/l = 2 · 102) at 4 W.

the solid state. This will swamp out all terms in the govern- are similar to those of Case 1. However, because of the
ing momentum equations and force any velocity predic- structured geometry of the arrangement, this case can be
tions in the solid region effectively to zero. In the liquid modeled using separate thermo-physical properties of the
region / value depends on the volume fraction of TCE two substances. Hence, in this case, determination of effec-
matrix and is always less than one. Hence, the source term tive thermo-physical properties for the thermal storage unit
does not vanish even in the molten region. This induces is not required.
drag to the fluid flow. Its magnitude depends on permeabil- Two types of configurations are considered for Case 2.
ity K, qp, l and fluid velocity. As the magnitude of the The first type (Case 2a) deals with plate-type fins oriented
source term increases, the effect of convection diminishes. in a particular direction. For this case, neglecting end
At this point conduction becomes the dominating mode effects, lengthwise variations of temperature and fluid flow
of heat transfer. It may be noted here that the chip and sub- can be ignored. Hence, the heat transfer and fluid flow
strate are also included in the domain, by treating them as analyses for this situation are based on a two-dimensional
fully solid materials. To ensure zero velocity in the chip and approach. For the second configuration (Case 2b), the ther-
substrate, a high value of viscosity (1020 m2/s) is assigned mal storage unit is filled with PCM and rod-type TCE fins.
for these regions. Additionally, a heat source term q_ is The rest of the arrangement is similar to that of Case 2a.
included for the heater region. Since the fins in this case are of rod-type, the heat transfer
and fluid flow cannot be analyzed using a two-dimensional
2.2. Case 2: PCM with TCE fins approach. Hence, a three-dimensional study is carried out
for this case.
In Case 2, the thermal storage unit is filled with PCM The governing equations are written following a single
and TCE fins with a given volume fraction of fin material. domain approach, as in Case 1. However, in absence of
The rest of the configuration and the boundary conditions the porous TCE matrix, /m in the present case is equal
 K.C. Nayak et al. / International Journal of Heat and Mass Transfer 49 (2006) 1833–1844 1839

Fig. 5. Isotherms and velocity vectors for a power level of 4 W with (a) no fin and (b) 10% volume fraction of TCE; time = 2400 s.

to one. The TCE, chip and substrate materials are assigned study), e is the sensible enthalpy, c is the specific heat, F1
very high viscosity (1020 m2/s) and melting point, which is the inverse latent heat function, DV is the volume of the
ensures that they remain solid and fixed in the domain. computational cell centered around the grid point P, Dt is
The rest of the formulation and analysis are same as that the time step chosen. The physical meaning of the term
of Case 1, and are not repeated here. aP =a0P is described in [17]. Since the PCM used in the pres-
ent study is considered to be a pure substance, F1 will be
2.3. Updating nodal heat content equal to cTmelt.

An accurate prediction of the liquid fraction in the fixed-

2.4. Initial and boundary conditions
grid enthalpy-based procedure used here, it is necessary to
update the latent heat content of each computational cell.
As the initial state and boundary conditions are similar
This is done in conjunction with temperature values
in both Case 1 and Case 2, a common set of appropriate
predicted by the macroscopic conservation equations,
initial and boundary conditions are adopted.
during each iteration within a time-step. In the present
The initial conditions appropriate to the physical situa-
case, to update the nodal latent heat content of a computa-
tion are:
tional cell at Pth node point, an iterative updating
At t = 0, T = T0, fl = 0.
scheme proposed in [17] is preferred. This has the following
The boundary conditions adopted for solutions of con-
form:
servation equations are:
nþ1 n aP  
½DEP  ¼ ½DEP  þ 0 k feP gn  cF 1 fDEP gn ð12Þ
aP 1. No slip condition at the walls, i.e. u = v = w = 0.
where DEP is the latent heat content of the computational 2. Natural
 convection with the ambient at the top, i.e.
cell, n is the iteration number, a0P ¼ qDV =Dt, aP is the coef- k oT  ¼ hðT jy¼H  T amb Þ.
ox y¼H 
ficient of TP in the discretization equation of the governing 3. Insulated side walls  and bottom; i.e. oT  ¼ 0 along
oT 
ox x¼0;L
energy equation, k is the relaxation factor (set to 0.1 in this vertical walls and oy  ¼ 0 for bottom.
y¼0
1840 K.C. Nayak et al. / International Journal of Heat and Mass Transfer 49 (2006) 1833–1844

1.0 1.0

0.8 0.8
Melt fraction

Melt fraction
0.6 0.6

0.4 0.4
No fin
4W
0.2 10% 0.2 5W
30%
6W
0.0 0.0
0 50 100 150 200 250 300 350
0 50 100 150 200 250 300 350
(a) Dimensionless time (a) Dimensionless time
100 60
No fin
Dimensionless maximum

Dimensionless maximum
temperature difference

temperature difference
80 10%
30%
60 40

40
20 4W
20
5W
0 6W
0 50 100 150 0
(b) Dimensionless time 0 50 100 150
(b) Dimensionless time
80
120
Chip temperature (οC)

4W
Chip temperature (οC)

100 5W
60 6W
80

40 60
No fin
10% 40
30%
20 20
0 600 1200 1800 2400 3000 3600 0 600 1200 1800 2400 3000 3600
(c) Time (s) (c) Time (s)

Fig. 6. Timewise variation of (a) melt fraction, (b) maximum temperature Fig. 8. Timewise variation of (a) melt fraction, (b) maximum temperature
difference within the PCM and (c) chip temperature for various TCE difference within the PCM and (c) chip temperature at different chip power
volume fractions at 4 W. level for 10% volume fraction of TCE.

Numerical simulations are performed for both the cases,

140
namely PCM with TCE matrix (Case 1) and PCM with
No fin with convection
120 TCE fins (Case 2). The dimensions of heat sinks used for
Chip temperature (οC)

No fin without convection the case studies are given in Fig. 1. Since the top surface
100 is cooled by natural convection, a nominal value of
h = 10 W/m2 K is used. It is also observed that the results
80
are not sensitive to the value of h during the melting stage
60 as most of the heat released from the chip is absorbed in
the TSU as latent heat. A comprehensive grid-indepen-
40 dence study is undertaken to determine the appropriate
spatial discretization, temporal discretization and iteration
20
0 600 1200 1800 2400 3000 3600 convergence criteria to be used. The quantities examined in
Time (s) this study are maximum magnitudes of various scalar vari-
Fig. 7. Effect of melt convection on heater temperature at 4 W. ables (i.e. components of velocity and temperature). As an
outcome of this study, we have taken a 42 · 34 · 42 uni-
form grid in 3-D, and 42 · 34 uniform grid in 2-D as our
3. Results and discussion final simulation matrix. Also, the time steps are kept high
(2 s) at the initial stages of simulation in the conduction
The governing equations described in the previous sec- regime. Once the melting starts the time step is reduced
tion are solved iteratively using a pressure based finite vol- to 0.2 s for better convergence. Time steps may be
ume method according to the SIMPLER algorithm [18]. increased after melting is sufficiently developed.
 K.C. Nayak et al. / International Journal of Heat and Mass Transfer 49 (2006) 1833–1844 1841

1.0 plies less inclination for overheating. It also ensures

uniform melting and stabilized chip temperature.
0.8
Darcy number: This number is relevant only in Case 1,
Melt fraction

0.6 and is defined as

0.4 Da ¼ K=l2 ð15Þ

2 fins
0.2 4 fins
where K is the permeability of the porous media.
8 fins
0.0 3.1. Results for PCM with porous metal matrix (Case 1)
0 50 100 150 200 250 300 350
(a) Dimensionless time
The simulations are based on the physical properties of
60
solid PCM, liquid PCM and the aluminum matrix, as given
in Tables 1 and 2. The effective thermal conductivity of the
Dimensionless maximum

2 fins
temperature difference

4 fins porous medium is determined according to the Veinberg

40 8 fins model described in [16].
Fig. 2 shows the evaluation of flow and temperature
fields, for the case of heating of PCM with TCE. It is
20
observed that the heating process is initially dominated by
conduction. Once melting begins, convection sets in and
0 enhances the heat transfer rate from the base of the heat
0 30 60 90 120 150
(b) Dimensionless time
sink. Rayleigh–Benard convection cells are formed in the
molten PCM as the molten layer is heated from below, lead-
60 ing to a wavy liquid–solid interface. Similar simulations are
performed with various values of the Darcy number (Da)
Chip temperature (οC)

50 and the effect of convection is studied. It is observed that

the effect of convection is more pronounced at high Da,
40 implying that a metal matrix insert enhances conduction
and inhibits convection. The heat transfer mechanism
2 fins
30 changes from a conduction regime to a conduction–convec-
4 fins
tion regime, as Da is increased. If Da is reduced by decreas-
8 fins
20 ing the size of the spherical matrix, conduction becomes the
0 600 1200 1800 2400 3000 3600 dominant mode of heat transfer. It is shown in Fig. 3 that a
(c) Time (s) reduction of Da from 2.0 · 103 to 1.96 · 104 results in
Fig. 9. Timewise variation of (a) melt fraction; (b) dimensionless lower melting rate, higher chip temperature and less unifor-
maximum temperature difference within the PCM and (c) chip temper- mity of temperature within the PCM. When the Darcy
ature for different fin distribution at 20% volume fraction of TCE and number is high, there is less resistance to convective fluid
5 W. motion in the molten PCM layer. A stronger melt convec-
tion results in quicker heat transfer from the base plate to
The results are studied with the help of an appropriate the upper regions of the heat sink, resulting in an increased
set of dimensionless parameters, which are defined as PCM melting rate. Hence, one can conclude that for the
follows: same volume fraction of TCE matrix, the cooling perfor-
Dimensionless time: It is given by mance is better for a larger matrix size.
Fig. 4 shows the effect of TCE volume fraction on vari-
am t
s¼ 2 ð13Þ ous quantities such as melt fraction, uniformity of temper-
l ature in PCM and chip temperature. It is observed that the
where am is the thermal diffusivity of TCE, l is length of the melting duration is reduced by inserting a TCE matrix. It
thermal storage unit, and t is time. may be noted that the enhancement of melting is large at
Dimensionless maximum temperature difference within low volume fractions of the TCE matrix (5%). Subse-
the thermal storage unit: It is defined as quent increase in volume fraction does not increase melting
rate linearly (Fig. 4(a)). This is expected because at higher
hmax ¼ ðk m l2 =QH ÞðT max  T min Þ ð14Þ
volume fractions of the metal matrix, the metal network
where km is the thermal conductivity of the TCE, H is the provides more resistance to melt flow, resulting in a con-
height of the thermal storage unit, Q is the chip level duction dominated heat transfer. One can note from
power, Tmax is the maximum temperature and Tmin is the Fig. 4(a) that the melting duration for 15% of TCE volume
minimum temperature, respectively, within the thermal fraction is higher than that for 10% of TCE volume frac-
storage unit. hmax gives an idea about the range of temper- tion. This occurs because the conduction dominated heat
ature variation within the PCM. A lower value of hmax im- transfer delays melting of PCM.
1842 K.C. Nayak et al. / International Journal of Heat and Mass Transfer 49 (2006) 1833–1844

Fig. 10. Isotherms and velocity vectors for the system with rod-type aluminum fins at (a) z = 0.018 m, (b) y = 0.034 m for chip power of 4 W and
time = 3900 s.

It is also observed that uniformity of temperature within performance of the system are discussed. Finally, the ther-
the PCM is better at a low volume fraction (5%) of TCE. mal performances of heat sinks with plate-type and rod-type
With a high TCE volume fraction, the effect of convection fins are compared for a given set of parameters.
diminishes, resulting in non-uniformity of temperature Fig. 5 shows the temperature and velocity fields within
within the PCM (Fig. 4(b)). Also, the chip temperature is the TSU after 2400 s of heating with a chip power level
found to be lower for a smaller volume fraction of TCE. of 4 W, for the case of plate fin arrangement. Fig. 5(b)
With a high volume fraction of TCE, heat transfer is pri- show a significant advantage of adding plate-type TCE fins
marily by conduction (enhanced by TCE), resulting in a in comparison to the baseline case (Fig. 5(a)). Because of
thermal stratification within the TSU. Consequently, the low thermal conductivity of the PCM, heat from the chip
chip temperature will be higher. One can note that even is not easily dissipated. Hence, the chip temperature rises
at 15% volume fraction of TCE, chip temperature is higher within a short time interval, leading to local overheating
than that of the base case. of the PCM adjacent to the chip. But when certain amount
of TCE is added, the effective conductance is significantly
3.2. Results for PCM with metallic fins (Case 2) increased. The TCE fins absorb heat at a faster rate and
diffuse it away from the base. Hence, melting of the PCM
In this case, computation is carried out with Eicosane as is initiated at multiple locations, at the base as well as along
the PCM and with the TCEs arranged in the form of plates the fins. Accordingly, convection cells are formed within
or rods (or pin fins) as shown in Fig. 1. First, the two-dimen- the molten PCM, as shown in Fig. 5(b).
sional case of plate-type fins is considered. The property val- It is observed from Fig. 6(b) and (c) that a 10% volume
ues used in the computation are listed in Tables 1 and 2. The fraction of TCE gives temperature uniformity and chip
thermo-physical properties of the PCM and TCE are temperature which is comparable to that with 30% TCE.
assumed to remain constant within the operating tempera- However, higher TCE volume fraction implies less latent
ture range. The effects of various parameters on the thermal heat storage. Hence, an optimum TCE volume fraction
 K.C. Nayak et al. / International Journal of Heat and Mass Transfer 49 (2006) 1833–1844 1843

1.0 the PCM significantly. The reason is that the rate of tem-
Matrix type perature gradient and power level depends mainly on the
0.8 Plate type
thermal conductivity of the TSU material. At a high chip
Melt fraction

Rod type
0.6 power level, the temperature of the system as a whole
may rise, but the non-dimensional temperature difference
0.4
between any two points in the system does not alter much.
0.2 The chip temperature, however, rises faster with higher
chip power until melting is initiated (Fig. 8(c)).
0.0 The effect of fin number (or fin thickness) is also studied,
0 50 100 150 200 250 300 350 for a particular TCE volume fraction, TCE material, and
(a) Dimensionless time
chip power level. Studies are conducted by distributing
120 20% of TCE volume fraction in 2, 4 and 8 fins, and the
Matrix type
results are shown in Fig. 9. It is observed that the larger
Dimensionless maximum
temperature difference

100 Plate type

the number of fins, the more is the temperature uniformity.
80 Rod type
The increase in temperature uniformity is caused by the
60 higher thermal diffusivity of the TCE material. With more
40
TCE fins distributed throughout the heat sink, the heat
from the base plate diffuses quickly, thus delaying local
20
overheating near the base plate. Hence the PCM adjacent
0 to the base plate reaches the melting temperature at a later
0 50 100 150 200 250 time. Although the onset of PCM melting is delayed with
(b) Dimensionless time
more TCE fins, the melting process is initiated at several
80 locations and the entire PCM melts in a shorter time
Matrix type (Fig. 9(b)).
Chip temperature (οC)

Plate type
Computations are also performed for the case of PCM
60 Rod type
with rod-type TCE fins. For this case, a three-dimensional
conjugate heat transfer study is necessary, as discussed ear-
40 lier. Fig. 10 shows the velocity and temperature fields of the
system along two different cross-sectional planes. Fig. 10(a)
shows circular isotherms around each rod fin. From
20
0 600 1200 1800 2400 3000 3600 Fig. 10(b), one can note that the temperature is nearly uni-
(c) Time (s) form in the entire domain.
Fig. 11. Comparative study of thermal performance of TSUs having TCE
Finally, a comparative study is performed with regard to
of matrix-type; plate-type and rod-type at 4 W. the thermal performance of the three PCM-TCE arrange-
ments (Fig. 11). It is observed that the rod-finned arrange-
ment maintains best uniformity of temperature in the
of about 10% can give an ideal heat sink performance from system and keeps the chip temperature lowest. Fig. 11(a)
the point of view of chip temperature and latent heat reveals that the onset of melting is significantly delayed in
storage. the case of PCM with rod-type TCE fins, even in compari-
In order to bring out the effect of natural convection son to the case of plate-type fins. This phenomenon can be
within the melt region, the variation of chip temperature attributed to an efficient three-dimensional heat diffusion in
with time is plotted for the cases with and without melt this case, leading to slower temperature rise. However, once
convection (Fig. 7). Natural convection of molten PCM melting begins, the process is completed in a short time.
promotes good mixing and reduces thermal stratification
within the melt. Hence, it enhances the melting rate of solid 4. Conclusions
PCM while maintaining temperature uniformity within the
liquid phase. This stabilizes the chip temperature close to In the present work, some basic studies are carried out
the PCM melting temperature. Hence, the effect of natural to model the thermal performance of heat sinks with phase
convection of the molten PCM cannot be ignored for the change materials and thermal conductivity enhancers.
present arrangement. Three types of TCEs are studied, namely porous matrix,
The results for effect of chip power on various quantities plate-type fins and rod-type fins. A generalized enthalpy
of interest are shown in Fig. 8. As expected, melting dura- based mathematical formulation and numerical model is
tion reduces with increase of chip power. Hence, if one developed for all the cases. This enthalpy–porosity based
needs to operate the equipment at a high chip power level, model can be easily extended to analyze the performance
more quantity of PCM or a PCM with a high latent heat of PCMs, which are mixtures or binary substances.
will be required. Fig. 8(b) demonstrates that power level For the case of PCM with porous TCE matrix, it is
does not affect non-dimensional temperature field within found that inserting aluminum matrix into Eicosane can
1844 K.C. Nayak et al. / International Journal of Heat and Mass Transfer 49 (2006) 1833–1844

offer an order-of-magnitude increase in thermal conductiv- [5] K.W. Snyder, An investigation of using a phase change material to
ity and melting rate. The effect of melt convection is signif- improve the heat transfer in a small electronic module for an airborne
radar application, in: Proceeding of the International Electronics
icant at high permeability and Rayleigh number. Melt Packaging Conference, San Diego, CA, 1991, vol. 1, pp. 276–303.
convection has a considerable effect on the evolution of [6] M. Ishizuka, Y. Fukuoka, Development of a new high density
the solid–liquid interface. However, the effect of convection package cooling technique using low melting point alloys, in:
becomes insignificant beyond a certain volume fraction of Proceedings of the ASME/JSME Thermal Engineering Joint Confer-
TCE. For the case of PCM with TCE fins, too, convection ence 1991, vol. 2, pp. 375–380.
[7] D. Pal, Y.K. Joshi, Application of phase change materials to thermal
in the melt plays a significant role in temperature unifor- control of electronics modules: a computational study, Trans. ASME,
mity. The performance of the heat sink improves if the J. Electron. Packag. 119 (March) (1997).
TCE material is distributed in the form of thinner fins. It [8] X. Tong, J.A. Khan, M.R. Amin, Enhancement of heat transfer by
is also found that rod-type fins perform better than plate- inserting a metal matrix into a phase change material, Numer. Heat
type ones, as they are able to maintain better uniformity Transfer, Part A 30 (1996) 125–141.
[9] W.R. Humphries, E.I. Griggs, A design handbook for phase change
of temperature within the PCM leading to less chip thermal control and energy storage devices, NASA T.P. 1074, 1977.
temperature. [10] D. Pal, Y.K. Joshi, Thermal management of an avionics module using
solid–liquid phase materials, J. Thermophys. Heat Transfer 12 (2)
Acknowledgement (1998).
[11] E.M. Alawadhi, C.H. Amon, Thermal analysis of a PCM thermal
control unit for portable electronics devices: experimental and
The work reported in this paper is supported by a re- numerical analysis, in: Eight Intersociety Conference on Thermal
search grant from the Defence Research and Development and Thermomechanical Phenomena in Electronic Systems, San
Organisation, Government of India. Diego, California, May 29–June 1, 2002.
[12] C. Beckermann, R. Viskanta, Natural convection solid/liquid phase
change in porous medium, Int. J. Heat Mass Transfer 31 (1988) 35–
References 46.
[13] D.N. Nield, A. Bejan, Convection in Porous Media, second ed.,
[1] A. Abhat, Experimental investigation and analysis of a honeycomb- Springer-Verlag, New York, 1999.
packed phase change material device, AIAA Paper 1976, pp. 76–437. [14] J.C. Ward, Turbulent flow in Porous Media, J. Hydraul. Div. ASCE
[2] E.W. Bentilla, K.F. Sterrett, L.E. Karre, Research and development 90 (HY5) (1964) 1–12.
study on thermal control by use of fusible materials, Northrop Space [15] C. Beckermann, R. Viskanta, S. Ramadhyani, A numerical study of
Laboratories, Control No. NAS 8-11163, NASA Document No. N66- non-Darcian natural convection in a vertical enclosure filled with a
26691, 1966. porous medium, Numer. Heat Transfer 10 (1986) 557–570.
[3] S. Witzman, A. Shitzer, Y. Zvirin, Simplified calculation procedure of [16] J.A. Weaver, R. Viskanta, Freezing of liquid—saturated porous
a latent heat reservoir for stabilizing the temperature of electronic media, J. Heat Transfer 108 (1986) 654–659.
devices, in: S. Oktay et al. (Eds.), Proceedings of the Winter Annual [17] A.D. Brent, V.R. Voller, K.J. Reid, Enthalpy porosity technique for
Meeting of the ASME, Boston, MA, HTD, 1983, vol. 28, pp 29–34. modeling convection–diffusion phase change: application to the
[4] R. Chebi, P.A. Rice, J.A. Schwarz, Heat dissipation in micro- melting of pure metal, Numer. Heat Transfer 13 (1988) 297–318.
electronic systems using phase change materials with natural convec- [18] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere
tion, Chem. Eng. Commun. 69 (1988) 1–12. Publications, New York, NY, 1980.