International Journal of Heat and Mass Transfer 58 (2013) 135151

Contents lists available at SciVerse ScienceDirect

International Journal of Heat and Mass Transfer

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

Thermouidics and energetics of a manifold microchannel heat sink

for electronics with recovered hot water as working uid
Chander Shekhar Sharma a, Manish K. Tiwari a, Bruno Michel b, Dimos Poulikakos a,
a
b

Department of Mechanical and Process Engineering, ETH Zurich, 8092 Zurich, Switzerland
Advanced Thermal Packaging, IBM Research Zurich, 8803 Rueschlikon, Switzerland

a r t i c l e

i n f o

Article history:
Received 2 October 2012
Received in revised form 31 October 2012
Accepted 5 November 2012

Keywords:
Microchannel heat sink
Electronics cooling
Energy
Exergy
Efciency
Turbulent ow

a b s t r a c t
A detailed thermo-hydrodynamic analysis of a hot water cooled manifold microchannel heat sink for
electronic chip cooling is presented. The hot water cooling enables efcient recovery of heat dissipated
by the even hotter chip by using hot water recovered from a secondary application. Contrary to usual
expectation of laminar ow in electronic cooling, high ow rate and high uid temperatures result in turbulent ow conditions in the inlet and outlet manifolds of the heat sink with predominantly laminar ow
conditions in microchannels. To simulate these complex ow conditions, a three dimensional (3D) conjugate heat transfer model with turbulent ow is developed. Microchannel heat transfer structure is
modeled as porous medium with permeability parameters extracted from a 3D model for a single microchannel. The energetic performance of the heat sink is analyzed in terms of 2nd law efciency and
sources of exergy destruction are identied by detailed local entropy generation analysis at low and high
Reynolds number conditions of 2400 and 11200 respectively. This analysis shows that entropy generation
due to heat transfer dominates the net entropy generation in the heat sink for both conditions. Although
entropy generation due to viscous dissipation increases signicantly with increased Reynolds number, it
still contributes less than a third to the total entropy generated at high Reynolds numbers. Use of hot
water reduces the heat transfer component of entropy generation signicantly, thus leading to higher
2nd law efciency.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
Electronic chips have witnessed an exponential increase in circuit density in the past few decades. This increase has largely kept
pace with Moores law [1]. However, since about the year 2000,
this has resulted in an exponential increase in heat dissipation
from the chips due to lack of voltage scaling, thus bringing thermal
packaging of electronics into sharp focus. The ever-increasing heat
dissipation from chips has put much more pressure for the replacement of air cooling with liquid cooling. Single phase liquid cooling
for microprocessors has been long recognized as an effective method to replace conventional air-cooling to handle the increasing
heat densities of current and future microprocessors. The liquid
best suited thermally for single-phase cooling is water due to its
high specic heat and thermal conductivity, as well as high availability and environmental friendliness.
Apart from effective cooling of high heat dissipating electronic
chips, increasing energy consumption by large computing systems
such as data centers has also become an issue of concern. Direct
electricity consumption by data centers had already reached 1%
Corresponding author. Tel.: +41 44 632 27 38; fax: +41 44 632 11 76.
E-mail address: dimos.poulikakos@ethz.ch (D. Poulikakos).
0017-9310/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2012.11.012

of the total world electricity consumption by 2005 [2] due to increased demands for IT (Information Technology) related services
such as internet and telephony. This gure grew further, though
at a signicantly lower rate, to between 1.1% and 1.5% by 2010
[3].The introduction of the Green500 list for supercomputers
[4,5] emphasized that performance can no longer be the sole
motivation for development of microprocessors and that performance per unit energy consumption is the more appropriate metric for better computing. This becomes especially important in case
of large air-cooled data centers where energy spent for cooling
comprises almost half of the total energy consumed by such systems. This portion of energy use can be signicantly reduced by
switching to liquid cooling. This is because the much lower thermal resistance inherent in liquid use enables cooling above the free
cooling limit thereby eliminating the need for coolant chillers. The
free cooling limit represents the minimum temperature at which
the coolant can effectively transport heat from a chip to ambient
conditions without the need for an additional chiller.
Additionally, and perhaps more importantly, if hot water in the
temperature range of 5070 C is used to cool electronic chips,
direct utilization of the collected thermal energy for secondary
applications, like district heating or specic industrial applications,
becomes feasible [69]. In such a system, the hot water, after

136

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

Nomenclature
A
cp
closs
d
D
DHT
Ex
f
GCI
h
H
k
Kperm
L
_
m
n
N
p
P
Prt
q_
Q_
R
Re
s
S_ gen
T
T
T0
U
Us
V
W
_
W

area (m2)
specic heat at constant pressure (J/kg K)
coefcient of non-linear momentum loss term
length scale (lm)
diameter (lm)
porous medium size
ow exergy (W)
CFD solution on a grid, owrate (l/min)
grid convergence index
specic enthalpy (J/kg), grid spacing (lm)
height, thickness (lm)
turbulent kinetic energy (m2/s2)
permeability (m2/s2)
length (lm)
mass ow rate (kg/s)
number of microchannels
level of granularity
observed order of convergence
pressure (Pa)
turbulent Prandtl number
heat ux vector (W/m2)
heat dissipation (W)
thermal resistance (C cm2/W)
Reynolds number
specic entropy (J/kg K)
rate of entropy generation (W/m3 K)
temperature (C, K)
Reynolds averaged temperature (K)
uctuating temperature (K)
velocity vector (m/s)
supercial velocity vector in porous medium (m/s)
volume (m3)
width (lm)
power (W)

Greek letters
turbulent dissipation
efciency
d
kronecker delta

e
g

performing the task of chip cooling, is used for the secondary application where it gives off a part of its energy. The cooling loop then
recovers the water back, still hot but at a lower temperature, to the
hotter chip to serve as the coolant.
A signicant body of work has analyzed liquid cooled microchannel heat sinks. Tuckerman and Pease [10] were the rst to report that a heat sink with microchannels is ideal for liquid cooling
of electronic chips because the heat transfer coefcient scales inversely with the characteristic channel dimension for a fully developed laminar ow. A maximum power dissipation density of
790 W/cm2 with a thermal resistance of 0.1 C cm2/W but at the
expense of a high pressure drop of 2 bar was reported. Most of
the later studies have concentrated on traditional microchannel
heat sinks in which liquid enters axially at one end of long microchannels. These studies have utilized single microchannel models
to successfully capture the thermodynamic performance of such
heat sinks. Lee et al. [11] studied heat transfer in rectangular
microchannels and concluded that conventional numerical analysis can be used to model thermal performance of microchannels.
Modeling of entire heat sinks with limited number of channels
has also been reported [12,13].
Traditional liquid cooled microchannel heat sinks have been followed by the development of manifold microchannel (MMC) heat

u
c
k

l
lt
x
q
s

viscous dissipation (s2)

porosity
thermal conductivity (W/m K)
dynamic viscosity (Pa s)
eddy viscosity (Pa s)
turbulent frequency (s1)
density (kg/m3)
stress tensor (Pa)

Subscripts
avg
average
c
coldplate
ch
channel
D
dissipation
D
direct dissipation
D0
turbulent dissipation
eff
effective
elect
electrical
f
uid
hs
microchannel heat transfer structure
im
inlet manifold
in
inlet
intf
solidliquid interface
m
manifolds
max
maximum temperature at chip
meas
measured
min
minimum
om
outlet manifold
out
outlet
pump
pumping
pred
predicted
Q
heat transfer
Q
mean temperature gradient
Q0
uctuating temperature gradient
s
solid
th
thermal
tot
total
TIM
thermal interface material

sinks. These heat sinks differ from the traditional heat sinks as
the coolant uid is delivered and returned through alternating,
uniformly spaced slot nozzles across the chip plane. This hierarchical design greatly reduces the travel length of the coolant uid
through the microchannel heat transfer structure and hence, signicantly reduces the pressure drop for ow of coolant through
the heat sink, thus making MMC heat sinks a very viable electronic
cooling solution. Copeland et al. [14] and Ryu et al. [15] undertook
numerical studies of MMC heat sinks with multiple inlet and exit
manifolds supplying liquid to microchannels from vertical direction. Based on the geometric and ow symmetry, they used a single
microchannel model for analysis of the heat sink.
Modeling of ow and heat transfer in the area of electronic cooling is generally made simpler by the fact the ow is predominantly
laminar due to small physical dimensions of the heat sinks. Thus a
conjugate heat transfer analysis of a heat sink involves solving
laminar incompressible NavierStokes equations. Most of the studies reported so far have analyzed such scenarios. With rising chip
heat uxes, however, the ow rate through the heat sink is continuously being pushed and there is now emphasis on developing
heat sinks using scalable manufacturing approaches [16]. This,
along with the use of hot water for electronic cooling for efcient
chip heat recovery, pushes up the ow Reynolds number of coolant

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

uid in heat sink. As a result, uid ow can reach transitional and

turbulent ow regimes. It is possible to encounter Reynolds numbers beyond the laminar range of 1000 and thus laminar ow
equations are no longer sufcient to model the ow. There are a
few studies that report turbulent ow analysis in the area of electronic cooling. Dhindsa and Pericleous [17,18] have compared various turbulent models with regards to prediction of ow and
temperature proles in electronic cooling with air as the coolant
uid. They compared various turbulence models for the ow regimes encountered in electronic cooling. It was reported that xequation based turbulence models like kx turbulence model or
the Shear Stress transport (SST) turbulence models perform better
in such low Reynolds number ows as compared to the e-equation
based ke model that is designed typically for high Reynolds number ows. Kasten et al. [9] performed preliminary analysis of a
MMC heat sink using turbulence modeling to model the uid ow
in the heat sink.
Apart from consideration of turbulence in uid ow, numerical
modeling of MMC heat sinks also involves other challenges. Due to
the use of a manifold to supply coolant uid across the microchannel
heat transfer structure, such heat sinks can suffer from non-uniform
distribution of coolant uid [9,19]. This, in turn, can lead to nonuniformity in temperature distribution in the chip. As a result, the
single channel assumption, as used in some previous studies
[14,15], no longer sufces and modeling of the full MMC heat sink
becomes imperative. The need for such modeling has been highlighted by Sharma et al. [8] where it was concluded that it is not possible to closely predict the maximum temperature at the chip for
MMC heat sink by representing heat transfer in the manifold microchannel heat sink through a single microchannel model. This
becomes difcult in a single microchannel model due to nonuniform distribution of coolant and heat dissipation from chip.
Hence, a complete numerical analysis of the entire heat sink is
required.
Modern day MMC heat sinks can contain many tens of microchannels. Thus, unlike a few previous studies [12,13], it is not

137

possible to model each channel separately while considering a

numerical analysis of the entire manifold plus microchannel
system. In such a scenario, microchannels can be modeled as
porous medium. Kim and Kim [20] showed analytically that it is
possible to numerically analyze the thermal and hydrodynamic
performance of microchannel heat sinks by modeling the microchannel heat transfer structure as uid-saturated porous medium,
thus making the consideration of entire heat sink feasible. Several
studies have analyzed manifold microchannel heat sinks following
this approach. However, in these studies, either the ow was fully
laminar [19] or the analysis was limited to hydrodynamics without
consideration of the thermal aspects [9].
In this paper, we present a detailed computational uid dynamics (CFD) study of conjugate heat transfer inside a realistic MMC
heat sink cooled by hot water at 3060 C. This heat sink is being
used for a liquid cooled data center supercomputer that employs
hot water to recover energy dissipated from the chip for building
heating [16]. We adopt a hierarchical modeling approach by rst
analyzing the microchannel heat transfer structure at single microchannel level and using the information from that analysis to model the entire heat sink through the porous media approach. We
model the turbulent ow in the manifold structure (encountered
due to the large size of manifolds, high ow rates through the heat
sink and high uid temperatures). We also account for the possibility of non-uniformity in the heat dissipation in the chip, in order to
enable a closer prediction of the maximum temperature at chip.
The results from the CFD analysis are then utilized to characterize
the 2nd law efciency of heat recovery from chip. Finally, we analyze the local entropy generation in the heat sink to identify the
major sources of exergy destruction.
2. Manifold microchannel heat sink
A schematic of the manifold microchannel heat sink analyzed in
this study is shown in Fig. 1. The heat sink material was copper and
it was made by diffusion bonding the manifold and microchannel

Fig. 1. Schematic of the manifold microchannel heat sink (arrows indicate direction of coolant ow): (a) Isometric view and (b) top view. The shown half of the heat sink
comprises the computational domain using hydrodynamic and thermal symmetry.

138

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

layers together (Wolverine, Inc). This heat sink was manufactured

using a micromachining process rather than micro-fabrication and
thus is a completely scalable design. The heat sink consisted of one
inlet manifold and two outlet manifolds. Fig. 1 depicts half of the
heat sink with the ow path through the heat sink indicated by arrows. Coolant entered through the inlet pipe into the inlet manifold as a jet. The ow impinged on the microchannels through
the slot nozzles in the base of the inlet manifold and divided
equally into two streams. Each stream then travelled through the
microchannels absorbing heat coming from the chip underneath
the TIM. A commercial TIM with thermal conductivity of 2.7 W/
m K was used. Finally, the two streams of coolant entered the outlet manifold, merged and exited the heat sink via the outlet pipe.
3. Single microchannel model
As discussed earlier, to simulate the thermouidics of the full
heat sink, it was necessary that the microchannel heat transfer
structure was modeled as a uid saturated porous medium. This
approach requires specication of numerical coefcients for the
momentum loss term in the porous medium model (see Section 4
for further details). In order to extract the values of these coefcients, ow through a single microchannel of the heat transfer
structure was investigated in detail. The microchannels were designed to be rectangular. However, the channel walls can become
curved due to the nature of the micromachining-based manufacturing process. The curved shape of the machined microchannel
is shown in Fig. 2(b). In order to analyze and quantify any effects
of this curved shape on the hydrodynamic and thermal characteristics of the microchannel heat transfer structure, a single curved
microchannel was also investigated numerically together with
the rectangular microchannel, both kinds having the same average
width. The models set-up is similar to what was used in Ref. [8]
and is briey described below.
The computational domains, major dimensions and boundary
conditions for both the channels are shown in Fig. 2 and listed in
Table 1. The TIM thickness, HTIM, was small compared to the rest
of the channel and hence is not depicted in Fig. 2. At the inlet,
the coolant mass ux and inlet temperature were imposed. At
the outlet nozzle, an opening type of boundary condition (ac-

Table 1
Microchannel dimensions.
Symbol

Description

Value (lm)

Hch
Wch
Wn
Lch
Wnozzle,in
Wnozzle,out
Hc
HTIM
LTIM

Height of channel
Width of channel
Width of n (i.e. microchannel side wall)
Length of channel/n
Width of inlet nozzle
Width of outlet nozzle
Thickness of coldplate above TIM
TIM thickness
Heated length of TIM

1473
159
181
17000
2000
2000
1200
12
12550

counts for ow recirculation across the boundary) was imposed.

This takes care of both inow and outow of the uid at the outlet
boundary, because it lies in recirculating ow zone [21]. The
symmetric boundary conditions as shown in Fig. 2 were exploited
to minimize computational cost. Uniform heat ux was imposed at
the lower surface of the TIM and an adiabatic condition was
imposed at all external boundaries of the solid domain. A blockstructured, hexagonal, non-uniform Cartesian mesh, with a ner
mesh near the walls to adequately resolve the boundary layer,
was used to discretize the computational domain. Grid independence was checked using the Grid Convergence Index (GCI)
method. Grids consisting of 7.27 and 7.45 million cells were used
for the curved channel and the rectangular channel simulations
respectively. The GCI method is described in detail in Section 4.3.
Inside the channels, the ow is laminar and governed by the following 3D conservation equations for mass (Eq. (1)), momentum
(Eqs. (2) and (3)) and energy (Eq. (4) for uid and Eq. (5) for solid)
[22].

$  qU 0;
$  qU  U $P $  s;

sl


2
$U $UT  d$  U ;
3

$  qUh $  kf rT s : $U;
$  ks $T 0:

4
5

The model takes into account the temperature dependence of intensive water properties such as density (q), dynamic viscosity (l) and

Fig. 2. Computational domain and boundary conditions for (a) rectangular channel and (b) curved channel.

139

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

specic heat (cp) [23]. Continuity of temperature and heat ux were

imposed at the uidsolid interface, as shown in Eqs. (6) and (7),
where n is the direction normal to the interface.

 ks $Ts;intf  n kf $Tf ;intf  n;

Ts Tf :

The conjugate heat transfer problem was solved using the commercial solver Ansys CFX version 12.1. The solver uses the nite
volume approach to discretize the governing equations into mostly
second-order accurate, coupled linear algebraic equations. The discretized governing equations were solved in a coupled manner
using the algebraic multi-grid method. The steady-state equations
were solved using a pseudo-transient term to evolve the steadystate solution [22]. Convergence of the equations being solved
was tracked by monitoring the normalized equation residuals as
well as global imbalances for conserved quantities of mass,
momentum and energy [21,22]. We imposed 1  106 as the normalized residual target and 0.1% as the conservation target for
our simulations.
4. Full heat sink model
4.1. Computational domain and boundary conditions
The schematic shown in Fig. 1 forms the computational domain
for the analysis of the entire heat sink. The boundary conditions for
the heat sink model are shown in Fig. 3. The analysis involves conjugate heat transfer, turbulent ow and ow through a porous
medium. The microchannel heat transfer structure was modeled
as a porous medium [20] as mentioned earlier. The uid, solid
and porous domains are also outlined in Fig. 3. At the inlet,
mass-ow and inlet temperature and at outlet, atmospheric pressure (arbitrary choice of constant pressure boundary condition)
was imposed. Fully developed turbulent ow was assumed at the
inlet. Symmetry boundary condition for the ow was imposed as
shown. At the outer walls of heat sink, heat loss to environment,
as obtained from measurements, was imposed. Heat ux dissipated from the chip was imposed over a part of the lower face of
heat sink (chip area, as shown in Fig. 4(c)). The major dimensions
of the heat sink are shown in Fig. 4 and their numerical values used
in the model are listed in Table 2.
4.2. Governing equations
The ow in the two manifolds is beyond the laminar range,
because the Reynolds number based on the hydraulic diameter of
the inlet pipe, Rein, is in the ranges of 200011,000 for the
operating range of ow rates (0.31.0 l/min) and inlet water
temperature (3060 C). In addition, due to the large variation in

the characteristic length scale in the heat sink, the ow regime

changes from turbulent in the inlet manifold to predominantly
laminar inside the microchannels and then back to turbulent in
the outlet manifold (see Fig. 3). This becomes clear from the wide
variation in ow Reynolds number as the uid ows through the
heat sink. For example, for the operating condition corresponding
to the highest ow rate and uid inlet temperature, the Reynolds
number changes from 11200 at the inlet of the heat sink to less
than 600 inside the microchannels and then back to 11200 at the
outlet of the heat sink. This makes numerical analysis of the heat
sink especially challenging. The effect of turbulence is modeled
using the kx model. The x-equation based models are known
to perform better than the e-equation based models for low Reynolds number ows [18]. The kx model belongs to the class of
2-equation models, which are based on the eddy-viscosity assumption. This assumption is not strictly valid in case of ows with
strong streamline curvatures. Streamlines can have signicant curvature in a ow through manifold microchannel heat sinks and
thus a case can be made for use of Reynolds stress models. These
models do not depend upon the eddy-viscosity assumption and
solve six additional equations for the Reynolds stresses [22,24].
However, a higher number of equations increase the computational effort and achieving convergence with these models becomes difcult for complex ows. The same was observed during
the course of the present study. Hence, the basic kx model was
used to model the ow in this work. The conservation equations
for the full heat sink model are as described below [22].
4.2.1. Momentum and energy conservation in uid domain
The Reynolds averaged continuity equation for the manifold is
the same as that in laminar ow i.e. Eq. (1). In tensorial notation,
the Reynolds averaged momentum equation for the ith velocity
component with the eddy viscosity assumption is






@
@
2
@
@U i @U j
;
qU i U j 
p qk
l lt

@xj
@xi
3
@xj
@xj
@xi

where the turbulent viscosity is given by

lt q :
x

The two additional equations for the komega model are

@
@
qU j k
@xj
@xj



@
@
qU j x
@xj
@xj



lt @k
@U i @U j @U i
lt

 b0 qkx;
rk @xj
@xj @xi @xj


10

l @x
x
@U i @U j @U i
l t

 bqx2 :
a lt
rx @xj
k
@xj @xi @xj
11

The model constants have standard values: b0 = 0.09, a = 5/9,

b = 0.075, rk = 2 and rx = 2.
Reynolds averaged energy conservation equation is






@
1
@
@T lt @h
qU j h U k U k
kf

@xj
2
@xj
@xj Prt @xj


 

@
@U i @U j
2
 dij k ;
U i l lt

@xj
3
@xj @xi
12

Fig. 3. Boundary conditions for full heat sink model.

where the turbulent Prandtl number Prt is set to 0.9.

For modeling near wall ow and heat transfer, a modied formulation of the standard wall functions was used. This formulation
automatically switches from the wall function approach to the
low-Re formulation as the mesh is rened near the wall. Additionally, it utilizes scalable wall functions which allow y+ insensitive
mesh renement [22].

140

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

Fig. 4. Major dimensions of heat sink (a) top view, (b) view in symmetry plane, (c) bottom view and (d) end view. The color codes are same as Fig. 3. (For interpretation of the
references to color in this gure legend, the reader is referred to the web version of this article.)

Table 2
Heat sink dimensions.

Symbol

Description

Value (lm)

L1
L2
W1
W2
Wom
Wim
D
Hms
Hm
Lhs
Lchip
Wchip
n

Length of heat sink

Length of outlet manifold
Width of heat sink
Width of manifold structure
Width of outlet manifold
Width of inlet manifold
Diameter of inlet and outlet pipes
Height of manifold structure
Height of manifolds
Length of microchannel heat transfer structure
Length of chip
Width of chip
Number of microchannels

35500
25250
35500
23000
5000
5000
4000
7000
4500
21250
16886
12550
62

closs
U s;x  p qjUs jU s;x ;
K perm;x
l
closs
U s;y  p qjUs jU s;y ;
SM;y 
K perm;y
K perm;y
l
closs
SM;z 
U s;z  p qjUs jU s;z :
K perm;z
K perm;z

q
l lt
U  Us  r 
rUs rUs T SM  rp;
c2 s
c

13

where Us is the supercial velocity and is equal to c U with U being

the true velocity in the porous medium. Turbulence viscosity for
ow inside porous medium, lt, is obtained by solving the same turbulence equations as in uid domain (see Eqs. (10) and (11)). c is
the volume porosity and is dened as

14

SM represents the directional momentum loss. Due to the high

velocities encountered, this term takes into account non-linear effects through Forchheimers extension of the basic Darcy law [25]
and is formulated as below

SM;x 

4.2.2. Conservation equations in the porous domain

Since the microchannel heat transfer structure was modeled as
a porous medium with spatially uniform porosity, the continuity
equation inside the porous medium reduces to same equation as
Eq. (1). The following equation was solved for conservation of
momentum inside porous medium

r

nW ch
:
nW ch n 1W fin

K perm;x

15
16
17

Kperm,i represents the permeability along coordinate i. As shown

in Figs. 13, the uid enters the microchannel heat transfer
structure predominantly in the negative z-direction and ows
along the y-direction through the structure. In order to inhibit ow
in the x- direction, the permeability in x direction was assumed to
be 100 times smaller than in other two directions. The permeability in the y direction was assumed to be equal to that in the z
direction. Kim and Kim [20] have suggested an expression for
permeability applicable to microchannels. However, since ow
through microchannels in the heat sinks changes direction at the
slot nozzles, the permeability values in the ow direction are obtained by matching the pressure drop through the porous medium
with that in a single rectangular microchannel. This is explained
further in Section 5. The coefcient for the non-linear term in the
above equations is obtained as below [19,26]



dch
;
closs 0:55 1  5:5
DHT

18

141

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151
Table 3
Discretization error and grid independence test using DP as solution functional (for ow rate of 1 l/min and Tf,in of 60 C).
Mesh

Number of cells (million)

1
2
3

79.98
36.62
17.57

Mean grid spacing (lm)

Fluid

Solid

79.98
39.16
51.31

121.78
121.78
121.78

where the characteristic length scales for pore size (dch) and overall
porous medium size (DHT) are determined following the references
[27,19]

p
Hch W ch ;
p
Hch L:

dch

19

DHT

20

A 1-equation model for heat transfer was used for conservation

of energy inside the porous medium.



@
@
@T
l @h
:
qU j h
keff ;i
c t
@xj
@xi
@xi
Prt @xi

21

The effective thermal conductivities in the three energy directions

are calculated based on whether solid and uid thermal resistances
to heat conduction are in parallel or series. Accordingly

keff ;x

kf ks

cks 1  ckf
keff ;y keff ;z ckf 1  cks :

22
23

4.2.3. Momentum and energy conservation at domain interfaces

Continuity of temperature and heat ux is imposed at both
uidsolid and porous-solid interfaces similar to Eqs. (6) and (7)
with an equivalent conductivity used for porous medium.
Similarly, at the uid-porous interface, continuity of mass ux,
momentum, heat ux as well as temperature is imposed.
The discretized governing equations for the full heat sink model
were also solved using Ansys CFX version 12.1. For the full heat
sink model, we imposed 1  106 as the normalized residual target
and 0.1% as the conservation target for our simulations.
4.3. Mesh and grid independence
Flow modeling in general and turbulence modeling in particular
is sensitive to the mesh used for discretization of the computational domain. A block-structured, hexagonal, Cartesian mesh
was used for all the domains in the full heat sink model. The mesh
was non-uniform with ner mesh near the walls to adequately resolve the boundary layer. Grid independence was checked using
GCI method which is based on Richardson extrapolation [28] and
has been recommended as a uniform method for reporting of grid
convergence [29]. GCI is dened as

GCIfine



F s f2  f1 
;
r p21  1  f1 

24

where Fs is a factor of safety and is set to 1.25. The solutions on the

three grids are denoted by f1, f2 and f3 and the mean grid spacing by
h1, h2 and h3, respectively, with h1 being the nest and h3 being the
coarsest grid. The parameter r21 represents the grid renement
factor and is dened as the ratio of grid spacing h2 to h1. The
resulting order of convergence, p, is obtained by the solution of a
transcendental equation [29]. The p value so computed is valid only
if the grids are in asymptotic range i.e. the following condition is
satised [30]

f = DP (Pa)

GCIne

GCIcoarse

GCIcoarse
r p21 GCIfine

1.31
1.31

13223.8
12960.1
12181.3

4.01

0.0127

0.0384

1.0203

GCIcoarse
1:
rp21 GCIfine

25

As mentioned earlier, the ow eld to be modeled was turbulent.

The mesh density close to the wall was kept sufciently dense so
that y+ for wall adjacent mesh cells was below 10 on all walls. A
number of meshes were prepared with average grid size in uid
and porous domains ranging from 120.36 lm to 29.89 lm. A set
of three grids was selected that satised the asymptotic range test
of Eq. (25) as well as exhibited monotonic convergence. The three
grids and results are described in Table 3, wherein DP is used as
the solution functional (i.e. f) [30].
coarse
The ratio rGCI
was close to one indicating that three grids
p
GCI
21

fine

were in asymptotic range. The GCIne and the GCIcoarse were small
for both the solution functionals. The relatively large value of p,
as compared to the formal order of convergence of the CFD solver,
can be explained by the fact that the ratio of mean grid sizes in dif

h
ferent directions hyx ; hhxz was not constant for the three grids [31].
Following the modied procedure for calculation of p [31], a set of
ve grids was used and the resulting value of p came out to be
1.154, which agrees better with the less than second order convergence typical of CFD codes. Also, the relative change in solution
functional values was smaller between mesh 2 and mesh 1 as compared to the corresponding change between mesh 3 and mesh 2.
Therefore, mesh 2 was used for all further simulations.
5. Results and discussions
5.1. Extracting the permeability from single microchannel model
As already discussed in Section 2, single microchannel models
for curved and rectangular channels have been set up and used
to compare the hydrodynamic and thermal characteristics. The
simulations were performed for ow rates ranging from 0.3 to
1.0 l/min and uid inlet temperature of 60 C. The water temperature at the outlet slot-nozzle and the maximum temperature of the
cold plate are compared in Fig. 5(a). Fig. 5(b) compares pressure
drop from the inlet to the outlet slot-nozzles. As evident from
Fig. 5, the thermal performance for both channel types is very similar. In terms of the pressure drop, the curved channels, in fact, perform slightly better than the rectangular channels by guiding the
ow and, thereby, reducing the pressure drop due to uid impingement on the channel walls acting as ns. Hence, the curvature in
channels does not affect the performance of the microchannel heat
transfer structure.
The coefcients in the momentum loss term of Eqs. (15)(17)
require specication of the permeability in three directions Kperm,i.
We extract these permeability values by tuning the basic permeability values from Kim and Kim [20] so that the pressure drop
through the microchannel cooling structure matches the pressure
drop through a single microchannel. This is required to account
for the effect of ow turning by 90 degrees inside the porous medium, after entry through the inlet nozzle and before exit through
the outlet slot nozzle. To determine the permeability, we used a
reduced heat sink model consisting of only the inlet and outlet

142

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

Since the channel width varies in the actual heat sink, an average channel width was used in the rectangular microchannel model. By tuning the pressure drop through the microchannel structure
against the microchannel model, a quadratic t was obtained for
the permeabilities in y and z directions, as a function of the ow
rate, f in l/min, through heat sink as below

K perm;y K perm;z 6:5065  109 f 2 8:1561  109 f

7:715  1010 :

26

A constant low permeability was used in x direction and is given by

K perm;x

!
1
cd2ch
:
100 12

27

Pressure drop through the reduced heat sink model, by using

permeability values from Eqs. (26) and (27), is compared against
pressure drop from rectangular microchannel model in Fig. 7. As
evident, the tuned permeability helps to closely approximate the
pressure drop through the microchannel.
5.2. Heat sink model simulations

Fig. 5. Comparison of rectangular and curved channel (a) Fluid outlet temperature
(Tf,out) and maximum cold plate temperature (Tmax) and (b) Overall pressure drop
(DPoverall).

nozzles and the porous medium structure as shown in Fig. 6. At the

inlet, massow and at the outlet, atmospheric pressure were imposed. The symmetry boundary condition for ow was adopted
as shown in Fig. 6. The rest of the bounding surfaces were modeled
as no-slip walls. In this reduced model, the ow was assumed to be
isothermal and only the continuity and momentum equations
were solved.

Fig. 6. Boundary conditions for reduced heat sink model.

5.2.1. Model validation

Experimental measurements from Sharma et al. [8] for the
water outlet temperature (Tf,out) and the maximum temperature
at the base of TIM (Tmax) were used to validate the full heat sink
model. These measurements were available for heat sink operating
conditions of ow rates from 0.3 to 1 l/min in steps of 0.1 l/min and
at 4 different water inlet temperatures, Tf,in, between 30 C and
60 C. The heat dissipation from the chip, Q_ chip , was kept constant
at 100 W. Fig. 8 shows the validation of the model in terms of predictions for the water temperature at the heat sink outlet,Tf,out. As
evident from Fig. 8, the model is able to capture well the heat absorbed by the water as it ows through the microchannel heat sink.
It was reported previously that the microchannel only model of
the heat sink is not able to predict accurately the chip level maximum temperature, Tmax, with the predictions deviating from
experimental measurements by as much as 7.2 C, which amounts
to an error of nearly 19% [8]. This deviation is caused by the
unavoidable error introduced by two key assumptions in the single
microchannel model: that the coolant is distributed uniformly
among all the channels and that the heat dissipation from the chip
is also spatially uniform. As will be explained further in Section 5.2.2, the coolant uid enters the inlet manifold as a jet, which

Fig. 7. Matching the pressure drop between microchannel and reduced geometry
model by tuning the permeability.

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

143

Fig. 8. Model validation for uid outlet temperature (Tf,out). Legend subscripts meas and pred indicate measured and simulated values for Tf,out respectively.

leads to non-uniform distribution of coolant among the channels.

The full heat sink model captures this effect. In addition to this,
non-uniformity in heat dissipation from the chip is modeled by
assuming a checkerboard pattern of power granularity in the chip.
Yazdani et al. [32] have shown that a minimum chip power granularity needs to be modeled for accurate predictions of chip level
maximum temperatures. We investigated this aspect for the present heat sink model. We characterized the granularity by the number of alternating rows or columns in the checkerboard pattern (N)
for the heat ux shown in Fig. 9a. Each row or column of the pattern consists of rectangular blocks with alternating heat ux
boundary conditions. Adiabatic boundary condition is assumed
for the blue colored blocks while heat ux boundary condition is
assumed for red colored blocks, such that the total heat ux for
all red blocks is equal to the total heat dissipated by the chip. Only
odd integer values are chosen for N in order to satisfy the symmetric boundary condition shown in Fig. 3. It is observed that a pattern
corresponding to N = 3 is sufcient, as model predictions for Tmax
for N > 3 do not change signicantly. Fig. 9(a) depicts the imposition of the heat ux boundary condition for N = 3. Fig. 9(b) compares model predictions for Tmax against experimental
measurements. The full heat sink model predicts Tmax accurate to
within 2.7 C at all operating condition. This amounts to a maximum of 6.3% deviation of predictions from experiments.
5.2.2. Hydrodynamic performance of the heat sink
Fig. 10 shows the velocity eld in two perpendicular planes for
the operating condition corresponding to the highest Reynolds
number (ow rate of 1 l/min and Tf,in of 60 C). The geometry of
the heat sink presents a backward facing step at the inlet and a forward facing step at the outlet from the microchannel heat transfer
structure (see Fig. 10(b)). Fig. 10(a) clearly depicts that the ow enters the inlet manifold as a jet and impinges either directly on the
microchannel heat transfer structure through slot nozzle or on the
backward facing step forming the inlet slot nozzle. The jet-like ow
causes large ow separation and the backward facing step leads to
two independent recirculation zones in the upper and lower part of
the inlet manifold. Fig. 10(b) shows that the ow enters the channels as a slot jet, impinges on the channel oor and then reattaches
to the channel upper wall after travelling a signicant fraction of
the channel length. Hence, the ow is developing over most of
the channel axial length. In the outlet manifold, the ow coming
out of the slot nozzle ows over the forward facing step that leads
to the formation of a large vortex, that lls almost the entire outlet

manifold, along with a small, counter-rotating vortex in the upper

half of the manifold. Similar ow eld patterns are observed for
other operating conditions as well.
Most of the turbulence in the ow is generated in the inlet manifold and in the inlet slot nozzle. Fig. 11 shows the variation of the
turbulent kinetic energy (TKE) in the plane of uid symmetry and
the mid cross-sectional plane. In the inlet manifold, strong shear
develops between the jet and the large recirculation zone in the
upper part of the inlet manifold. Another source of large shear is
the uid impingement on the microchannel heat transfer structure
(visible in both the planes). Both these zones contribute most of the
TKE generation with the latter contributing signicantly more than
the former. Fig. 11(b) also shows that as the uid enters and ows
along the channels, most of that TKE is dissipated within a short travel length as the ow laminarizes due to small channel widths. In
the outlet manifold, although the ow enters by forming a large
and a small vortex, the shear zone between the two is rather limited
and it does not lead to any signicant TKE generation.
5.2.3. Thermal and energetic performance of the heat sink
The thermal performance of the heat sink can be analyzed in
terms of the net thermal resistance to heat ow (Rth) from chip
to coolant uid. The thermal resistance of the heat sink is dened
as

Rth

Achip T max  T f ;in

:
Q_ chip

28

Fig. 12(a) shows how Rth changes with the ow rate at different
uid inlet temperatures, Tf,in. As the ow rate increases, the heat
transfer coefcient values also increase across the channel length
because of the predominantly developing laminar ow regime over
most of the channel axial length [8]. Fig. 12(b) shows temperature
contours in the mid cross-sectional plane and demonstrates the
thermally developing nature of ow inside the channels. This explains the decreasing thermal resistance with increasing ow rates.
It was discussed in Section 1 that hot water is used for heat
recovery from the chip so that it can be utilized for other purposes,
such as building or district heating. After this utilization the still
warm but at lower temperature water returns to the (hotter) chip
and serves as the coolant. Based on this it becomes important that
the heat recovery takes place in an exergetically efcient manner.
The exergetic or the 2nd law efciency of the heat sink can be calculated as below [33]

144

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

Fig. 9. (a) Checkerboard pattern of heat ux boundary condition (for N = 3) and (b) model validation for maximum temperature at the base of TIM (Tmax).

gII

Exout
_ pump W
_ elect
Exin W

Exout
;
Exin;tot

29

where the ow exergy Exj at jth location, with j indicating either inlet or outlet, is dened as



1
_ hj  h0  T 0 sj  s0 jUj2 :
Exj m
2

30

Ambient reference conditions of T0 = 25 C and P0 = 1 atm were

used to evaluate ow exergy in Eq. (30). In essence, the control volume for exergy analysis includes both the heat sink and the electronic components generating heat. The variation of the 2nd law
efciency at various operating conditions is shown in Fig. 13. For
any given uid inlet temperature Tf,in, the temperature differential,
Tw,avg(x)  Tf,bulk(x) required for same amount of heat transfer from

solid to uid is reduced as the ow rate increases. Thus, exergy

destruction due to heat transfer over a nite temperature differential is also reduced [33]. On the other hand, the exergy loss due to
uid pressure drop increases at higher ow rates. However, in the
case of liquid cooling, the contribution of the pressure drop term to
_ pump , see
the exergy destruction (in the form of pumping power, W
Eq. (29)) is less than 4% for all operating conditions. Hence, as the
ow rate increases, the exergetic efciency increases at any uid
inlet temperature. The exergetic efciency also increases with increase in uid inlet temperature at any given ow rate.
5.2.4. Local entropy generation in the heat sink
The 2nd law efciency quanties the amount of exergy
destruction in a particular system. Exergy destruction is directly

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

145

Fig. 10. Velocity vectors for ow rate of 1 l/min and Tf,in of 60 C in (a) plane of uid symmetry and (b) mid cross-sectional plane (i.e. the plane perpendicular to plane of uid
symmetry at half length of heat sink).

Fig. 11. TKE for ow rate of 1 l/min and Tf,in of 60 C in (a) plane of uid symmetry and (b) mid cross-sectional plane.

146

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

proportional to entropy generation in the system via the Gouy

Stodola theorem [33]. Hence, major sources of exergy destruction
can be identied by analyzing the amount of entropy generated
in various parts of the system. This analysis can be performed by
utilizing the entropy transport equation [34] and it forms a part

of post-processing analysis of the CFD results. The net entropy generation per unit volume, for incompressible ow with no bulk heat
sources, is expressed as

2l/
q_
;
S_ gen  2  rT
T
T

31

where / represents the viscous dissipation and q, the local heat diffusion as per Fouriers law. The rst and the second terms in the
above equation quantify the local entropy generation due to heat
transfer S_ gen;Q and uid friction S_ gen;D respectively. For turbulent
ows, Eq. (31) needs to be Reynolds averaged. Kock and Herwig
[35,36] have presented the Reynolds averaging of the above equation and modeling of the resulting unclosed terms for high Reynolds
number ow using ke model. However, to the best of our knowledge, the corresponding extensions factoring the volume porosity
and the anisotropic thermal conductivity have not been reported
in the literature. Therefore, the equations from reference [36] were
modied to account for the volume porosity and the anisotropic
thermal conductivity inside the porous medium and the use of k
x model for turbulence. Reynolds averaged entropy generation
due to heat transfer, S_ gen;Q , can be calculated as below

S_ gen;Q S_ gen;Q S_ gen;Q 0 ;

32

S_ gen;Q represents entropy generation due to mean temperature gradients. For a material volume with anisotropic conductivity, S_ gen;Q is
given by

S_ gen;Q

2
!2 3
14
@T 5
:

keff ;i
@xi
T2

33

keff is equal to kf for manifolds and ks for the solid domain and for
porous domain, it is obtained from Eqs. (22) and (23). S_ gen;Q 0 represents entropy generation due to uctuating temperature gradients
and is given by

2
3
 0 2
c4
@T 5
_S
;
gen;Q 0 2 keff ;i
@xi
T

Fig. 12. (a) Thermal resistance of the heat sink and (b) Temperature contours in
mid cross-section plane for ow rate of 1 l/min and Tf,in of 60 C.

34

where c was included to account for volume porosity in porous domain and is equal to 1 for uid domain, 0 for solid domain and is
given by Eq. (14) for porous domain. Using the value of eddy viscosity from Eq. (9), S_ gen;Q 0 can be modeled as

Fig. 13. Variation of 2nd law efciency with change in operating conditions.

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

147

Fig. 14. S_ gen;D for (a) ow rate of 0.3 l/min and Tf,in of 40 C and (b) ow rate of 1 l/min and Tf,in of 60 C (both gures are plotted to same magnitude scale and use logarithmic
color scale). (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

148

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

Fig. 15. S_ gen;Q for (a) ow rate of 0.3 l/min and Tf,in of 40 C and (b) ow rate of 1 l/min and Tf,in of 60 C (both gures are plotted to same magnitude scale and use logarithmic
color scale). (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

149

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

cqcp k @T
S_ gen;Q 0
xPrt T 2 @xi

!2
35

The Reynolds averaged entropy generation due to uid friction,

S_ gen;D , is given by

S_ gen;D S_ gen;D S_ gen;D0 ;

36

where, S_ gen;D represents entropy generation due to direct dissipation

(i.e. due to gradient in mean velocity) and is given by

2 8

S_ gen;D

!2 9
!2
@U z =
@U x @U y
2

: @x
;
@z
@y
@x
T
!2
!2 3
@U x @U z
@U y @U z 5
:
37

@z
@x
@z
@y
!2
cl 4 < @U x

@U y

@y

!2

For calculating S_ gen;D inside the porous medium, true velocities are
used. S_ gen;D0 represents the entropy generation due to turbulent disTable 4
Aggregate entropy generation at low and high Rein condition.
Aggregate entropy generation (W/K)

Rein = 2420

Rein = 11200

S_ gen;D;tot

1.4  106

3.1  106

5

9.6  10

1.4  103

9.7  105

1.4  103

S_ gen;Q ;tot

1.5  103

6.3  104

S_ gen;Q 0 ;tot
S_ gen;Q ;tot solid

1.9  104

1.6  104

3

2.6  103

S_ gen;D0 ;tot
S_ gen;D;tot S_

gen;D;tot

S_ gen;D0 ;tot

3.2  10

S_ gen;Q ;tot S_ gen;Q ;tot S_ gen;Q 0 ;tot S_ gen;Q ;tot solid

3

4.9  10

3.4  103

S_ gen;tot S_ gen;D;tot S_ gen;Q ;tot

5.0  103

4.8  103

sipation (i.e. due to gradients in velocity uctuations) and is given

by

2 8

S_ gen;D0

9
!2
 0 2 =
0
@U z
@U 0x @U y
2

: @x
@z ; @y
@x
T
3
!2
 0
2
@U 0y @U 0z
@U x @U 0z
5:
38

@z
@x
@z
@y


cl 4 < @U 0x 2

@U 0y

@y

!2

Again, using the denition of viscosity from Eq. (9), S_ gen;D0 , can modeled as

cqkx
:
S_ gen;D0
T

39

Volume porosity is included in Eqs. (37)(39) to account for the fact

that true velocities are used for computation of entropy generation
in porous domain. Wall functions for entropy generation from reference [36] were not included in this analysis. This is because, as discussed further ahead, most of the entropy generation was found to
take place in TIM and solid domain of the heat sink.
Figs. 14 and 15 compare the Reynolds averaged entropy generation due to uid friction, and heat transfer for two operating conditions that correspond to the low and the high Reynolds number
ow investigated: ow rate of 0.3 l/min and Tf,in of 40 C (Rein =
2420) and ow rate of 1 l/min and Tf,in of 60 C (Rein = 11200). For
the sake of comparison, plots for both the operating conditions
are shown for same magnitude range. Since the entropy generation
levels traverse over a large range of order of magnitudes, the entropy generation plots use a logarithmic color scale for ease of
visualization.
Fig. 14(a) and (b) show that as the operating condition changes
from low to high Rein, S_ gen;D increases. This is expected because the

Fig. 16. jrTj for ow rate of 0.3 l/min and Tf,in of 40 C.

150

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151

ow becomes more turbulent with increase in Rein thus increasing

viscous dissipation in the ow. Fig. 15(a) and (b) show that S_ gen;Q
decreases as the operating condition is changed from low to high
Rein. This is because the high Rein condition corresponds to high
heat transfer coefcient in microchannel heat transfer structure
and thus low temperature gradients in the uid. The temperature
gradient in the solid part also decreases with increase in Tf,in and
thus S_ gen;Q decreases in the entire heat sink.
The change in the various components of the entropy generation with change in operating conditions can be analyzed by calculating the aggregate entropy generation over the entire heat sink.
This can be done by integrating the entropy generation components over the volume of the heat sink as shown below

S_ gen;D;tot

S_ gen;D dV

fluid;porous

S_ gen;Q ;tot

fluid;porous

S_ gen;Q dV

40
Z

S_ gen;Q dV

41

solid

Table 4 shows calculated values of aggregate entropy generation components at low and high Rein. At both the operating conditions, most of the entropy generation due to viscous dissipation,
S_ gen;D;tot is contributed by the turbulent dissipation component,
S_ gen;D0 ;tot and most of the entropy generation due to heat transfer,
S_ gen;Q ;tot is contributed by solid due to low thermal conductivity of
TIM.
As operating condition is changed from low to high Rein, S_ gen;D;tot
increases due to signicant increase in turbulent dissipation at
high Rein. On the other hand, S_ gen;Q ;tot is reduced due to lower overall
temperature gradients at higher Tf,in. However, the decrease in
S_ gen;Q ;tot more than compensates the increase in S_ gen;D;tot thus reducing S_ gen;tot S_ gen;D;tot S_ gen;Q ;tot . As a consequence, the 2nd law efciency also increases with increase in Rein.
Most of S_ gen;D takes place in regions of high turbulence at both
low and high Reinconditions. This becomes evident from comparison of Fig. 14(b) with Fig. 11. Most of S_ gen;Q occurs in regions of high
temperature gradients. This becomes evident by comparing
Fig. 15(a) with Fig. 16. Fig. 16 shows the net temperature gradient
in manifold and microchannels for the low Rein condition.
In addition, S_ gen;Q ;tot dominates S_ gen;D;tot at both low and high Rein
conditions. This indicates that viscous dissipation plays a lesser
role in the overall entropy generation in the heat sink. At low Rein
condition, S_ gen;Q ;tot is about an order of magnitude higher than
S_ gen;D;tot . At high Rein conditions, although S_ gen;D;tot increases signicantly; it contributes less than a third of the total entropy
generation.
6. Conclusions
The thermouidic and energetic performance of an MMC heat
sink, operated with hot recovered water for the cooling of even
hotter electronic chips, was investigated. 3D conjugate heat transfer models for microchannels, were developed and combined with
a full heat sink model.
In the full heat sink model, the uid turbulence was modeled
using the kx model, the microchannel system heat transfer was
modeled as a uid-saturated porous medium, and the heat dissipation from the chip was imposed as a realistic checkerboard pattern.
The model predicts Tmax to within 2.7 C of literature experiments,
which is signicantly better than the state of the art. This shows
that the full MMC heat sink needs to be modeled in order to closely
predict the chip level maximum temperature.
The energetic performance of the heat sink was analyzed using
the 2nd law of thermodynamics. Sources of exergy destruction in

the heat sink were identied by a detailed local entropy generation

analysis at low and high Rein operating conditions. The analysis
0
shows that the turbulent dissipation, S_
, contributes most of
gen;D ;tot

the entropy generation arising from the uid friction, S_ gen;D;tot . However, the total entropy generation due to heat transfer, S_ gen;Q ;tot , is
much higher than the total entropy generation due to viscous dissipation, S_ gen;D;tot , at both low and high Rein conditions. S_ gen;D;tot increases signicantly at high Rein but still contributes less than
one-third of the total entropy generated in the heat sink. Reduction
in S_ gen;Q ;tot with increase in Rein leads to reduction in overall entropy
generation S_ gen;tot . Our results conclusively show that the use of hot
water as coolant can lead to a high 2nd law efciency, which is
essential to successfully utilize the heat recovered from cooling
of electronic chips and data centers for secondary usage such as
building heating. Our results also show that moderately increased
investment of pumping energy, while increasing viscous dissipation can reduce the entropy generation due to heat transfer and
thus minimize the overall entropy generation.
Acknowledgments
The work was supported in part through AQUASAR project
funded by the Competence Center Energy and Mobility (CCEM).
The support is gratefully acknowledged. The authors thank Severin
Zimmermann (ETH Zurich) for providing experimental data on the
heat sink performance and Ingmar Meijer (IBM Research, Zurich)
for useful discussions and comments on this work.
References
[1] G.E. Moore, Cramming more components onto integrated circuits, Electronics
(1965) 114117.
[2] J.G. Koomey, Worldwide electricity used in data centers, Environ. Res. Lett. 3
(3) (2008) 034008.
[3] J.G. Koomey, Growth in Data Center Electricity Use 2005 to 2010, Analytics
Press, Oakland, CA, 2011.
[4] S. Sharma, H. Chung-Hsing, F. Wu-chun, Making a case for a Green500 list, in:
20th International Parallel and Distributed Processing Symposium, IPDPS
2006, 2006, p. 8.
[5] The Green500 list, 2011. <http://www.green500.org/>.
[6] G.I. Meijer, Cooling energy-hungry data centers, Science 328 (5976) (2010)
318319.
[7] T. Brunschwiler, B. Smith, E. Ruetsche, B. Michel, Toward zero-emission data
centers through direct reuse of thermal energy, IBM J. Res. Dev. 53 (3) (2009)
11:1111:13.
[8] C.S. Sharma, S. Zimmermann, M.K. Tiwari, B. Michel, D. Poulikakos, Optimal
thermal operation of liquid-cooled electronic chips, Int. J. Heat Mass Transfer
55 (78) (2012) 19571969.
[9] P. Kasten, S. Zimmermann, M.K. Tiwari, B. Michel, D. Poulikakos, Hot water
cooled heat sinks for efcient data center cooling: towards electronic cooling
with high exergetic utility, Front. Heat Mass Transfer 1 (2) (2010) 023006.
[10] D.B. Tuckerman, R.F.W. Pease, High-performance heat sinking for VLSI, IEEE
Electron Device Lett. 2 (5) (1981) 126129.
[11] P.-S. Lee, S.V. Garimella, D. Liu, Investigation of heat transfer in rectangular
microchannels, Int. J. Heat Mass Transfer 48 (9) (2005) 16881704.
[12] R. Chein, J. Chen, Numerical study of the inlet/outlet arrangement effect on
microchannel heat sink performance, Int. J. Therm. Sci. 48 (8) (2009) 1627
1638.
[13] B.J. Jones, P.-S. Lee, S.V. Garimella, Infrared micro-particle image velocimetry
measurements and predictions of ow distribution in a microchannel heat
sink, Int. J. Heat Mass Transfer 51 (7-8) (2008) 18771887.
[14] D. Copeland, M. Behnia, W. Nakayama, Manifold microchannel heat sinks:
isothermal analysis, IEEE Trans. Compon. Packag. Manuf. Technol. Part A 20 (2)
(1997) 96102.
[15] J.H. Ryu, D.H. Choi, S.J. Kim, Three-dimensional numerical optimization of a
manifold microchannel heat sink, Int. J. Heat Mass Transfer 46 (9) (2003)
15531562.
[16] S. Zimmermann, I. Meijer, M.K. Tiwari, S. Paredes, B. Michel, D. Poulikakos,
Aquasar: a hot water cooled data center with direct energy reuse, Energy 43
(1) (2012) 237245.
[17] C. Bailey, K. Dhindsa, K. Pericleous, Low Reynolds number turbulence models
for accurate thermal simulations of electronic components, in: 5th Int. Con5
on Thermal and Mechanical Simulation and Experiments in Micro-Electronics
and Micro-Systems, EuroSimE2004, 2004.

C.S. Sharma et al. / International Journal of Heat and Mass Transfer 58 (2013) 135151
[18] B.C. Dhindsa, K. Pericleous, Investigation into the performance of turbulence
models for uid ow and heat transfer phenomena in electronic applications,
IEEE Trans. Compon. Packag. Technol. 28 (4) (2005).
[19] W. Escher, B. Michel, D. Poulikakos, A novel high performance, ultra thin heat
sink for electronics, Int. J. Heat Fluid Flow 31 (4) (2010) 586598.
[20] S.J. Kim, D. Kim, Forced convection in microstructures for electronic equipment
cooling, J. Heat Transfer 121 (3) (1999) 639645.
[21] ANSYS CFX-Solver Modeling Guide, Release 12.1, ANSYS, Inc., Canonsburg, PA,
2009.
[22] ANSYS CFX-Solver Theory Guide, Release 12.1, ANSYS, Inc., Canonsburg, PA,
2009.
[23] CRC handbook of chemistry and physics, 2011. <http://www.hbcpnetbase.
com/>.
[24] S.B. Pope, Turbulent Flows, eighth ed., Cambridge University Press, 2000.
[25] P.H. Forchheimer, Wasserbewegun durch Boden, Z. Ver. Dtsch. Ing. 45 (1901)
17821788.
[26] G.S. Beavers, E.M. Sparrow, D.E. Rodenz, Inuence of bed size on the ow
characteristics and porosity of randomly packed beds of spheres, Trans. ASME.
E. J. Appl. Mech. 40 (3) (1973) 655660.
[27] Y.S. Muzychka, M.M. Yovanovich, Modeling friction factors in non-circular
ducts for developing laminar ow, in: Second AIAA Theoretical Fluid
Mechanics Meeting, Albuquerque, NM, 1998.

151

[28] P.J. Roache, Perspective: a method for uniform reporting of grid renement
studies, J. Fluids Eng. 116 (3) (1994) 405413.
[29] C.J. Roy, Review of code and solution verication procedures for computational
simulation, J. Comput. Phys. 205 (1) (2005) 131156.
[30] Examining spatial (grid) convergence, 2011. <http://www.grc.nasa.gov/
WWW/wind/valid/tutorial/spatconv.html>.
[31] M.D. Salas, Some observations on grid convergence, Comput. Fluids 35 (7)
(2006) 688692.
[32] K. Etessam-Yazdani, M. Asheghi, H.F. Hamann, Investigation of the impact of
power granularity on chip thermal modeling using white noise analysis, IEEE
Trans. Compon. Packag. Technol. 31 (1) (2008) 211215.
[33] M.J. Moran, H.N. Shapiro, Fundamentals of Engineering Thermodynamics, fth
ed., John Wiley & Sons, 2004.
[34] A. Bejan, Entropy Generation Minimization: The Method of Thermodynamic
Optimization of Finite-Size Systems and Finite-Time Processes, CRC Press,
1996.
[35] F. Kock, H. Herwig, Entropy production calculation for turbulent shear ows
and their implementation in cfd codes, Int. J. Heat Fluid Flow 26 (4) (2005)
672680.
[36] F. Kock, H. Herwig, Local entropy production in turbulent shear ows: a highReynolds number model with wall functions, Int. J. Heat Mass Transfer 47 (10
11) (2004) 22052215.