International Journal of Heat and Mass Transfer 83 (2015) 345–356

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Comparison of thermal performance between plate-fin and pin-fin heat

sinks in natural convection
Younghwan Joo a, Sung Jin Kim a,⇑
a
School of Mechanical, Aerospace, and Systems Engineering, Korea Advanced Institute of Science and Technology, 291 Daehak-ro, Daejeon 305-701, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: The thermal performance of optimized plate-fin and pin-fin heat sinks with a vertically oriented base
Received 29 October 2014 plate is compared analytically in natural convection. A new correlation of the heat transfer coefficient
Received in revised form 3 December 2014 is proposed and validated experimentally to optimize pin-fin heat sinks, while a correlation of the heat
Accepted 5 December 2014
transfer coefficient for plate-fin heat sinks is adopted from previous studies. The comparison is made
Available online 24 December 2014
under the same base-plate dimensions and fin height conditions. Two objective functions are used in
optimizing the thermal performance: the total heat dissipation and the heat dissipation per unit mass
Keywords:
for a given base-to-ambient temperature difference. When the total heat dissipation is used as an objec-
Natural Convection
Heat sink
tive function, the optimized plate-fin heat sinks dissipate a larger amount of total heat than do the opti-
Pin-fin mized pin-fin heat sinks in most practical applications. When the heat dissipation per unit mass is used as
Optimization an objective function, on the other hand, the optimized pin-fin heat sinks dissipate a larger amount of the
heat per unit mass than the optimized plate-fin heat sinks in most practical applications.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction
Heat sinks are typically divided into forced convection and nat-
ural convection heat sinks based on the operating conditions.
Forced convection heat sinks dissipate a larger amount of heat
due mainly to such flow inducing devices as fans, but their reliabil-
ity is lower than that of natural convection heat sinks because of
these additional devices. Therefore, natural convection heat sinks
are widely used in applications for which high reliability is
required, and low performance may be tolerated. Two common
types of natural convection heat sinks are plate-fin heat sinks
and pin-fin heat sinks. Plate-fin heat sinks are easy to design and
fabricate, so they are widely used in applications for which cost
reduction is a main issue. Pin-fin heat sinks have omnidirectional
performance because of their geometric characteristics, so they
are widely used in applications for which heat sinks are used in
various orientations. There have been many studies on these two
types of heat sink [1–9] because of their advantages over the other
types of heat sink. However, it is not yet known which type of heat
sink between the two has better thermal performance in the natu-
ral convection mode.
Some previous studies have tried to answer this question. Spar-
row and Vemuri [1] found an optimal fin population of pin-fin heat
⇑ Corresponding author. Tel.: +82 (42) 350 3043; fax: +82 (42) 350 8207.
E-mail address: sungjinkim@kaist.ac.kr (S.J. Kim).
sinks by changing the number of fins for the fixed values of the fin
diameter. The thermal performance of the pin-fin heat sink having
the optimal number of fins was compared to that of a plate-fin heat
sink under the constraint of the same surface area for both heat
sinks. Their results showed that the pin-fin heat sink had lower
thermal resistance than the plate-fin heat sink, by about 40%. How-
ever, the constraint of the same surface area places an undesirable
limit on the thermal performance because the optimum surface
area does not have to be the same for each type of heat sink. For
a more meaningful comparison, therefore, the constraint of the
same surface area needs to be removed.
Iyengar and Bar-Cohen [2] compared plate-fin heat sinks and
pin-fin heat sinks that had been optimized using the least-material
method. In this method, the optimum fin thickness (or fin diameter
for pin-fin heat sinks) is determined when the fin height is given.
Then, the optimum spacing between the adjacent fins is obtained
by maximizing the amount of heat dissipated from the array for
various values of the spacing. From their analytical results, they
found that the optimized pin-fin heat sinks dissipate a larger
amount of heat than do the optimized plate-fin heat sinks. How-
ever, there are some inherent limitations in the least-material
method. This method is effective at reducing the mass of a single
fin, but may not provide a mass-minimizing optimum design for
the whole array of the heat sink. Therefore, some different
approaches are needed to compare the thermal performance per
unit mass of both types of heat sink.

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2014.12.023
0017-9310/Ó 2014 Elsevier Ltd. All rights reserved.
346 Y. Joo, S.J. Kim / International Journal of Heat and Mass Transfer 83 (2015) 345–356

Nomenclature

A surface area [m2] Greek symbol

B base plate thickness [m] a thermal diffusivity [m2/s]
CD drag coefficient [-] b volumetric thermal expansion coefficient [1/K]
cp specific heat [kJ/kg-°C] e emissivity [–]
d fin diameter [m] g fin efficiency [–]
El Elenbaas number [–] l dynamic viscosity [N-s/m2]
g standard acceleration of gravity [m/s2] m kinematic viscosity [m2/s]
Gr Grashof number [–] q density [kg/m3]
H fin height [m] r Stefan–Boltzmann constant [W/m2 K4]
h convective heat transfer coefficient [W/m2 K] u porosity [–]
K permeability [m2]
k thermal conductivity [W/m-K] Subscripts
L heat sink length [m] 1 limiting case 1
M heat sink mass [kg] 2 limiting case 2
n number of fins [–] 3 limiting case 3
Nu Nusselt number [–] 4 limiting case 4
P pressure [Pa] array array
Pr Prantl number [–] b base
Q heat dissipation [W] d diameter
Ra Rayleigh number [–] eff effective
Re Reynolds number [–] f fluid
Rth thermal resistance [K/W] fin fin
S spacing of pin-fin array [m] h horizontal
T1 ambient temperature [K] L heat sink length
Tb base temperature [K] pin pin-fin heat sink
uD Darcian velocity [m/s] plate plate-fin heat sink
V pore velocity [m/s] ratio ratio
W heat sink width [m] Sparrow Sparrow and Vemuri
wc fin thickness of plate-fin [m] s solid
ww channel spacing of plate-fin [m] v vertical

The objective of the present study is to determine which type of
a heat sink performs better under fixed volume conditions
between plate-fin heat sinks and pin-fin heat sinks. The fixed vol-
ume condition means that, physically, the space specified by the
length, width, and fin height of a heat sink is fixed in the present
study. Two objective functions are used for the comparison: the
total heat dissipation and the heat dissipation per unit mass for a
given base-to-ambient temperature difference. When the first
objective function is maximized, the thermal resistance, which is
defined as the base-to-ambient temperature difference divided
by the total heat dissipation, is minimized. Thus, the higher value
of the first objective function means a greater heat dissipation
capability under a given volume of a heat sink. The second objec-
tive function estimates how efficiently a heat sink can dissipate
heat at relatively small mass. Therefore, the second objective func-
tion plays an important role in designing heat sinks in applications
for which the mass of a heat sink is an important factor. Each
objective function will be maximized to optimize each type of heat
sink analytically using the correlations of the convective heat
transfer coefficient. For plate-fin heat sinks, the correlation of the
heat transfer coefficient suggested by Bar-Cohen and Rohsenow
[3] is used. For pin-fin heat sinks, a new correlation of the heat
transfer coefficient will be proposed and validated experimentally.
This correlation will be used to study the effects of the diameter,
the horizontal spacing, and the vertical spacing on the heat transfer
coefficient. After each type of heat sink is optimized, the thermal
performances of the optimized plate-fin heat sinks and the opti-
mized pin-fin heat sinks are compared for each objective function.
Finally, region maps for the ratio of the total heat dissipation and
the heat dissipation per unit mass are suggested. The region maps
can help thermal engineers to determine which type of natural
convection heat sink is better than the others according to specific
constraints.
2. Correlation of the heat transfer coefficient for pin-fin heat
sinks
In optimizing the thermal performance of the plate-fin heat
sinks, a correlation of the heat transfer coefficient suggested by
Bar-Cohen and Rohsenow [3] can be used. In optimizing the
thermal performance of the pin-fin heat sinks, no reliable corre-
lation that can be applied in a wide range of the geometric
parameters is available. Hence, a new correlation of the heat
transfer coefficient for pin-fin heat sinks will be proposed in this
section.
In the present study, the asymptotic method proposed by Chur-
chill and Usagi [10] is used to propose a new correlation of the heat
transfer coefficient for pin-fin heat sinks. For plate-fin heat sinks,
there are only two limiting cases associated with channel spacing
(wc): small spacing and large spacing [3]. However, there are four
limiting cases for pin-fin heat sinks because there are two types
of fin spacing: horizontal spacing (Sh) and vertical spacing (Sv).
Limiting case 1 means densely-positioned pin-fins with small hor-
izontal spacing and small vertical spacing. Limiting case 2 means a
vertical single array of pin-fins with large horizontal spacing and
relatively small vertical spacing. Limiting case 3 means a horizon-
tal single array of pin-fins with large vertical spacing and relatively
small horizontal spacing. Limiting case 4 means an isolated hori-
zontal cylinder with large horizontal and vertical spacing. Each
limiting case is analyzed, and these cases are integrated through
the asymptotic method to predict the heat transfer coefficient in
the intermediate region for the four limiting cases.
 Y. Joo, S.J. Kim / International Journal of Heat and Mass Transfer 83 (2015) 345–356 347

In the present study, the definition of the average convective

heat transfer coefficient for a fin array is as follows:
Q array
hfin ¼ ð1Þ
Aarray gfin ðT b  T 1 Þ

The total heat dissipation from the heat sink and the thermal resis-
tance can be evaluated as
 
Q total ¼ Q base þ Q array ¼ hbase Abase þ hfin Aarray gfin ðT b  T 1 Þ ð2Þ

Tb  T1 1
Rth ¼ ¼ ð3Þ
Q total hbase Abase þ hfin Aarray gfin

tanhðmHÞ
gfin ¼ ð4Þ
mH
8 qffiffiffiffiffiffiffiffi
>
< 2hfin for plate-fins
ks ww
m ¼ qffiffiffiffiffiffiffi ð5Þ
>
: 4hfin for pin-fins
ks d

where hfin , Q array , Aarray , gfin , T b , T 1 , Q total , Q base , hbase , Abase , H, ks , ww ,

and d are average convective heat transfer coefficient of fin array,
heat dissipated by fin array, surface area of fin array, fin efficiency,
base temperature, ambient temperature, total heat dissipated from
heat sink, heat dissipated by un-finned base, average convective
heat transfer coefficient of un-finned base, surface area of un-finned
base, fin height, thermal conductivity of solid phase, channel spac-
ing of plate-fin, and fin diameter, respectively.

2.1. Limiting case 1

In limiting case 1, pin-fins are densely positioned and the pin-

fin array can be modeled as a porous medium. Fluid flows in the
x-direction, as depicted in Fig. 1, so the situation is assumed to
be 1-dimensional. Then, the governing momentum equation for a
porous medium, the so-called Darcy equation, is as follows [11]:

DP l f
¼ uD ð6Þ
Dx K
where P, lf , K, and uD are pressure, fluid viscosity, permeability, and
Darcian velocity, respectively.
In natural convection, the pressure gradient term can be
expressed as a temperature difference term using the Boussinesq
approximation, as follows:

DP
¼ qf gbgfin ðT b  T 1 Þ ð7Þ
Dx
where g and b are the standard acceleration of gravity and the vol-
umetric thermal expansion coefficient, respectively.
As in the analysis of densely positioned plate-fins in natural
convection [12], the exit temperature of air is assumed to be the
same as the average fin temperature in limiting case 1. Then, the
following equation is obtained from the energy balance:
Fig. 1. (a) Vertically oriented pin-fin heat sink and (b) parameters of pin-fin heat
sink.
_ p gfin ðT b  T 1 Þ ¼ hfin;1 Aarray gfin ðT b  T 1 Þ
Q ¼ mc ð8Þ

_ ¼ qf uD WH
m ð9Þ
Fig. 2 shows a unit cell representing the structure of a staggered
where m,_ cp , W, and H are the mass flow rate, specific heat, width of
pin-fin array. By dividing the total base area by the unit area occu-
heat sink, and fin height, respectively. pied by a single fin, the number of fins is obtained as follows:
Using Eqs. (6)–(8), the heat transfer coefficient can be obtained
as LW
nfin ¼ ð11Þ
Sh S v
2
WK qf cp gbgfin ðT b  T 1 Þ
hfin;1 ¼ ð10Þ where L, Sh , and Sv are length of heat sink, horizontal spacing, and
pdnfin lf
vertical spacing, respectively.
348 Y. Joo, S.J. Kim / International Journal of Heat and Mass Transfer 83 (2015) 345–356

2.2. Limiting case 2

In limiting case 2, horizontal cylinders are arranged in a single

Fig. 2. Unit area occupied by a single pin-fin in a staggered array.
vertical direction. The horizontal spacing is very large, so this case
can be modeled as an isolated vertical single array of horizontal
cylinders. There have been previous studies about vertically
arranged horizontal cylinders. Marsters [15] performed experi-
ments on heat transfer from uniformly heated horizontal cylinders
arranged in a vertical direction. Aihara et al. [4] performed similar
experiments with more cases and suggested a Nusselt number cor-
relation as follows:
hfin L
Nufin ¼ ¼ ½0:311 þ 0:454 lnðSv =dÞGr 1=4
L ð18Þ
kf

Eq. (10) includes the permeability (K). There are many perme- gbgfin ðT b  T 1 ÞL3
GrL ¼ ð19Þ
ability models, and one of them is the drag model for flow perpen- m2f
dicular to the array of cylinders as suggested by Happel and
Brenner [13]. However, this model tends to underestimate the per- where Nu, kf , Gr, and mf are Nusselt number, thermal conductivity of
meability, which leads to lower heat transfer coefficients than fluid, Grashof number, and kinematic viscosity of fluid, respectively.
experimental results. Therefore, a new model for evaluating the This correlation is applicable to Sv =d ¼ 1  4 and
permeability is proposed as follows. In the present study, the drag Gr L ¼ 106  108 . As Sv =d becomes small, the values in this case
model [14] is employed because the drag caused by horizontal cyl- approach those in limiting case 2. Therefore, only Sv =d ¼ 1  2 is
inders is considered to be a main component of the pressure drop. selected to cover limiting case 2. The function in the square bracket
In Fig. 2, it is assumed that the force applied to the unit cell is the in Eq. (18) is linearized in the Sv =d ¼ 1  2 region as follows:
same as the drag force experienced by a single pin-fin. The flow is  
hfin;2 L Sv
assumed to be laminar flow. Then, the following equation is Nufin ¼ ¼ 0:3669  0:0494 Gr1=4
L ð20Þ
kf d
obtained.

qf V 2 2.3. Limiting case 3

D P  Sh ¼ C D  d  ð12Þ
2
In limiting case 3, horizontal cylinders are arranged in the single
where C D and V are drag coefficient and pore velocity, respectively. horizontal direction. According to the Eq. (18), the vertical spacing
The drag coefficient for a cylinder in laminar flow can be is large enough for this case to be modeled as a vertically isolated
expressed as horizontal array when Sv =d > 4. To analyze this case, the correla-
tion form of a single horizontal cylinder in natural convection
24 24lf was employed because limiting case 3 approaches the situation
CD ¼ ¼ ð13Þ
Re qf Vd of a single horizontal cylinder as the horizontal spacing increases.
There is a simple correlation applicable to a single horizontal cylin-
On the other hand, the 1-dimensional pressure gradient for the Dar- der: Rad ¼ 102  104 [16].
cian flow is given as follows:
hfin d
Nud ¼ ¼ 0:85Ra0:188
d ð21Þ
dP DP lf lf kf
¼ ¼ uD ¼ V/ ð14Þ
dx Sv K K
3
gbgfin ðT b  T 1 Þd
where / is porosity. Rad ¼ ð22Þ
a f mf
The porosity can be calculated from Fig. 2.

pd2
/¼1 ð15Þ
4Sh Sv

By combining Eqs. (12)–(15), the permeability can be expressed as

follows:

2
4Sh Sv  pd
K¼ ð16Þ
48

By substituting Eqs. (11) and (16) into Eq. (10), the convective heat
transfer coefficient for limiting case 1 is expressed as follows:

2
Sh Sv 4Sh Sv  pd qf cp gbgfin ðT b  T 1 Þ
hfin;1 ¼ ð17Þ
pdL 48 mf
While Eq. (17) was obtained, the air inflow from the open front edge
was neglected. Even though Aihara et al. [4] observed this inflow by
flow visualization, the inflow does not seem to have a significant
influence on the thermal performance of a pin-fin heat sink. Fig. 3. Variation of f ðSh Þ in limiting case 3.
 Y. Joo, S.J. Kim / International Journal of Heat and Mass Transfer 83 (2015) 345–356 349

where Ra and af are Rayleigh number and thermal diffusivity of Table 2

fluid, respectively. Nusselt numbers for each experimental case of Sparrow and Vemuri [1].

Then, the following correlation is suggested for limiting case 3. nfin RaSparrow  106 NuSparrow, NuSparrow, Eq. NuSparrow, Eq.

experiment (26) (27)

kf    0:188
hfin;3 ¼ f ðSh Þ Rad ð23Þ 18 1.14 73.9 64.1 70.5
d 27 1.02 82.4 79.9 78.3
39 1.13 85.7 93.0 86.3
Sh 52 1.10 83.0 86.8 79.4
Sh ¼ 3=4 1=4
ð24Þ 68 1.17 64.6 77.9 71.7
d L
⁄
L ¼ W ¼ 0:0762 m; H ¼ 0:0254 m; d ¼ 6:35 mm.
where Sh is non-dimensional horizontal spacing.
Fig. 3 shows the variation of f ðSh Þ in terms of Sh values. Numer-
ical simulation is used to obtain the data in place of a large number
of Eq. (27) are smaller than those of Eq. (26), so Eq. (27) is chosen
of experiments. Details about the numerical simulation are cov-
as the final form of the correlation for pin-fin heat sinks.
ered in the Appendix A. The dimensions of the simulated arrays
are listed in Table 1. In Fig. 3, the small Sh region belongs to limiting
case 3. Results obtained for different diameters can be fitted into 3. Experiments
the linear function for Sh ¼ 0:2  0:4. By substituting the fitted lin-
ear function into Eq. (23), the convective heat transfer coefficient In the previous section, a new correlation of the convective heat
for limiting case 3 is expressed as follows: transfer coefficient was proposed. To use this correlation to opti-

mize the thermal performance of pin-fin heat sinks, the correlation
kf   gbgfin ðT b  T 1 Þ 1 0:188 needs to be validated for various configurations of pin-fin heat
hfin;3 ¼ 2:132Sh  0:4064 ð25Þ
d af mf d sinks. Therefore, experiments were performed with pin-fin heat
sinks for various values of fin diameter, horizontal spacing, and
vertical spacing.
2.4. Limiting case 4

Limiting case 4 means an isolated horizontal cylinder. Eq. (21), 3.1. Materials and methods
which is applicable to a single horizontal cylinder in
Rad ¼ 102  104 , was already introduced in limiting case 3. There- The problem under consideration in the present study is natural
fore, Eq. (21) is used for limiting case 4. convective flow through staggered circular pin-fins with a verti-
cally oriented base plate, as shown in Fig. 1. The direction of the
fluid flow is parallel to the x-axis. The bottom of the base plate is
2.5. Final form of the convective heat transfer coefficient for pin-fin
uniformly heated. Air passes through the heat sink as a coolant,
heat sinks
thus taking away the heat generated in components attached to
the bottom of the heat sink substrate.
By combining the equations for each limiting case, the following
The dimensions of the heat sinks are listed in Table 3. The tested
two forms of the heat transfer coefficient are obtained.
heat sinks are made of aluminum alloy 6061 (k ¼ 170 W=m K,
 2:1   2:1 2:1
1
e ¼ 0:08); no additional surface treatment is applied. Usually, heat
1 1 1 4:85 4:85 4:85
hfin ¼ hfin;1 þ hfin;2 þ hfin;3 þ hfin;4 ð26Þ loss through radiation heat transfer from a heat sink made of bare
aluminum is less than 5% of the total heat loss [17], so it can be
 81 safely neglected. Fig. 4 shows a schematic diagram of the experi-
 8
1:3 1:3 1:3 1:3 8 mental apparatus. All experimental apparatuses are located in an
hfin ¼ hfin;1 þ hfin;2 þ hfin;3 þ hfin;4 ð27Þ
adiabatic chamber that is isolated from the external environment
because natural convection is affected by even small disturbances.
Exponents are determined by least-square fit to the experimental The size of the chamber is 3 m  5 m  3 m. Fig. 5 shows the test
results of Sparrow and Vemuri [1]. The Nusselt numbers that were section assembly. The test section is fixed by a stand made of alu-
obtained experimentally by Sparrow and Vemuri are compared minum. The test section is located at 50 cm from the bottom of the
with the Nusselt numbers that are predicted by Eqs. (26) and (27) stand. The distance from the bottom of the stand to the test section
in Table 2. Only the amount of heat dissipated through convection is 5 times larger than the length of the heat sink, so inflow induced
heat transfer is considered. The Nusselt numbers and the Rayleigh by natural convection is not disturbed. The thermocouples on the
numbers in Table 2 are defined as follows: test section are connected to an Agilent 34972 DAQ to transport
Q total measured data to a PC. To measure the temperatures at the base
NuSparrow ¼ ð28Þ plate, nine K-type thermocouples are mounted through 1.5 mm
LðT b  T 1 Þkf
deep holes on the back side of the base plate, which is 5 mm thick.
The heat sinks are heated by electrical heaters fabricated using a
gbðT b  T 1 ÞL3
RaSparrow ¼ ð29Þ stainless steel sheet with 30 lm thickness and sandwiched by
af v f
Table 3
Both equations predict the same optimal number of fins as do the Dimensions of pin-fin heat sinks for experiment.
experimental results of Sparrow and Vemuri. However, the errors
Type of heat sink L (m) W (m) H (m) d (m) Sv (m) Sh (m)
Table 1 Pin-fin 1 0.002 0.015 0.004
Dimensions of pin-fin heat sinks for numerical simulation. Pin-fin 2 0.004 0.010 0.004
Pin-fin 3 0.004 0.019 0.006
Type of heat sink L (m) W (m) H (m) d (m) Sv (m) Sh (m)
Pin-fin 4 0.1 0.1 0.03 0.004 0.030 0.008
Group 1 0.003 0.03 0.00175–0.01225 Pin-fin 5 0.004 0.010 0.010
Group 2 0.1 0.1 0.05 0.004 0.03 0.00225–0.01225 Pin-fin 6 0.004 0.030 0.004
Group 3 0.005 0.03 0.00275–0.01325 Pin-fin 7 0.006 0.017 0.006
350 Y. Joo, S.J. Kim / International Journal of Heat and Mass Transfer 83 (2015) 345–356

Fig. 4. Experimental apparatus: (a) schematic diagram and (b) photo of the experimental apparatus.

Kapton films. Electrical current is applied to a heater connected to 4. Optimization

a DC power supply (Agilent N5767A). To reduce the heat loss from
the bottom surface of the heat sink to the environment, Teflon In the previous sections, the correlation of the convective heat
(k ¼ 0:25 W=m K) and PEEK (k ¼ 0:25 W=m K) blocks are used as transfer coefficient for pin-fin heat sinks was proposed and vali-
insulation material. To calculate the amount of heat loss, K-type dated experimentally. In this section, a pin-fin heat sink will be
thermocouples are positioned on both sides of the PEEK block. optimized using the new correlation, while a plate-fin heat sink
Temperatures are recorded every 10 s, and temperature measure- will be optimized using the existing correlation of Bar-Cohen and
ments are monitored by the PC. Ambient temperature is measured Rohsenow [3]. The optimization will be performed individually
by the thermocouples exposed to ambient. The temperature of for each objective function.
the base plate is measured until the change in the temperature is The parameters of the plate-fin heat sinks are shown in Fig. 7.
smaller than 0:3  C in 30 min. The heat transfer coefficient of the plate-fin heat sink is obtained
using the following correlation, suggested by Bar-Cohen and
3.2. Validation of the correlation for pin-fin heat sinks Rohsenow [3], as follows:
2 312
Fig. 6 provides a validation of Eq. (27) by comparing the results hfin wc 6 576 2:873 7
Nufin ¼ ¼4 þ qffiffiffiffiffiffiffiffiffiffiffi5 ð30Þ
based on Eq. (27) to the experimental data of Sparrow and Vemuri kf ðgfin ElÞ
2
gfin El
[1] and those of the present study. The experimental results are
within 13% of the predictions of the correlation. Eq. (27) is appli- gbðT b  T 1 Þw4c
cable for 3  102 m 6 L 6 1 m and 3  102 m 6 H 6 3  101 m. El ¼ ð31Þ
af mf L
The applicable range is validated by comparing the results based
on Eq. (27) to the results of the numerical simulation, as depicted where wc and El are the channel spacing of the plate-fin and the
in Fig. A3. Elenbaas number, respectively.
 Y. Joo, S.J. Kim / International Journal of Heat and Mass Transfer 83 (2015) 345–356 351

Fig. 5. Test section: (a) assembly of components and (b) bottom side of a base plate.

For the pin-fin heat sinks, the amount of heat dissipated
through the pin-fin array is obtained using Eq. (27). For both types
of heat sink, the amount of heat dissipated through the un-finned
base plate is evaluated using the same correlation. This correlation
for the un-finned base plate and the equation for the surface area
were suggested by Iyengar and Bar-Cohen [2]; both the correlation
and the equation are used in the present study.
4.1. Objective function 1: the total heat dissipation
When this objective function is maximized for a given base-to-
ambient temperature difference, the thermal resistance, which is
defined as the base-to-ambient temperature difference divided
by the total heat dissipation, is minimized. This implies that the
heat dissipation capability of a heat sink can be evaluated using
this objective function. The objective function of total heat dissipa-
tion is determined using Eq. (2).
Fig. 8 shows the results of the optimization for plate-fin heat
sinks. For a given volume of a heat sink, the configuration of a
plate-fin heat sink is determined by the fin thickness (ww)
and the channel spacing (wc). There exists a maximum point
on the surface of the total heat dissipation at which the opti-
mum thermal performance and the optimum configuration are
obtained.
Fig. 9 shows the results of the optimization for pin-fin heat
sinks. For a given volume of a heat sink, the configuration of a
pin-fin heat sink is determined by the fin diameter (d), the horizon-
tal spacing (Sh), and the vertical spacing (Sv). To optimize the three
parameters at once, the horizontal spacing and the vertical spacing
are optimized first for a given fin diameter, as shown in Fig. 9(a)
352 Y. Joo, S.J. Kim / International Journal of Heat and Mass Transfer 83 (2015) 345–356

Case 1
Case 2
Case 3
Case 4
Case 5 +13% 35
Case 6
10 Case 7
Sparrow and Vemuri [1]: nfin=18
30
Sparrow and Vemuri [1]: nfin=27
-13%
Rth, predicted (K/W)

Sparrow and Vemuri [1]: nfin=39

(W)
Sparrow and Vemuri [1]: nfin=52 25
Sparrow and Vemuri [1]: nfin=68

total
20

Q
15

10
3
2.5
15
2
13
1.5
11
1 9
10 0.5 7
ww (mm) 0 5
wc (mm)
Rth, experiment (K/W)
Fig. 8. Optimization of the total heat dissipation for plate-fin heat sinks at
Fig. 6. Validation of the proposed pin-fin correlation. L = W = 0.1 m, H = 0.08 m, DT = 30 K.

28

26

24
(W)

22
total

20
Q

18

16

14
25
22.5
10
20 9
17.5 8
7
15 6
5
Sv (mm) 12.5
10
4
Sh (mm)
3

(a)
27.5

27

26.5

Fig. 7. Vertically oriented plate-fin heat sink.

26
(W)
total

25.5
and this process is repeated for various fin diameters. Then, the
optimum fin diameter is obtained using Fig. 9(b), and the optimum
Q

25
horizontal spacing and the optimum vertical spacing can be
obtained from the contour plot similar to that shown in Fig. 9(a),
24.5
which is drawn for the optimum fin diameter.
For the same base-plate dimensions and fin height, the thermal
24
performance of the optimized heat sink suggested by Sparrow and
Vemuri [1] is compared to that of the optimized heat sink sug-
23.5
gested by this study. The Nusselt number of the optimized pin- 2 2.5 3 3.5 4 4.5 5 5.5 6

fin heat sink of Sparrow and Vemuri is 85.7 (d = 6.35 mm), as d (mm)

shown in Table 2. For the same Rayleigh number, the Nusselt num-
ber of the optimized pin-fin heat sink of the present study is 119
(b)
(d = 2 mm, Sh = 3 mm, and Sv = 19 mm). The optimized heat sink Fig. 9. Optimization of the total heat dissipation for pin-fin heat sinks at
of the present study shows an approximately 35% improved ther- L = W = 0.1 m, H = 0.08 m, DT = 30 K: (a) Qtotal surface at d = 0.0031 m (b) optimized
mal performance. Qtotal at each diameter.
 Y. Joo, S.J. Kim / International Journal of Heat and Mass Transfer 83 (2015) 345–356 353

4.2. Objective function 2: the heat dissipation per unit mass

This objective function estimates how efficiently a heat sink can

120
dissipate heat at relatively small mass. For the optimization, this
110
objective function is maximized for a given base-to-ambient tem-
100
perature difference under the fixed volume condition. The objec-

Q/M (W/kg)
90
tive function of the heat dissipation per unit mass is evaluated
80
using the following equation:
70
 
hbase Abase þ hfin Aarray gfin ðT b  T 1 Þ 60

Q =M ¼ ð32Þ 50
M total 40

The mass of a heat sink is obtained as follows: 30

35
30
M total ¼ Mbase þ Marray ð33Þ 25
20
15
20
15 10
M base ¼ qs WLB ð34Þ 5
Sv (mm) 10 Sh (mm)
5 0
(
qs nfin ww LH for plate-fins
Marray ¼ 2
qs nfin pd4 H for pin-fins
ð35Þ (a)
where M total , M base , Marray , qs , and B are total mass of heat sink, mass 120
of base plate, mass of fin array, density of heat sink material, and
base-plate thickness, respectively. 110

Fig. 10 shows the results of the optimization for the plate-fin

100
heat sinks. There exists a maximum point on the surface for the
heat dissipation per unit mass. The optimum fin thickness is smal- 90
ler than that obtained using objective function 1. As the fin thick-
Q/M (W/kg)

80
ness decreases, the amount of the heat dissipated through a fin
decreases because of the lowered fin efficiency. In this case, how- 70
ever, the mass of the fin also decreases as the fin thickness
decreases. Therefore, the amount of heat dissipation per unit mass 60

is not significantly reduced even with a smaller fin thickness.

50
Fig. 11 shows results of the optimization for the pin-fin heat
sinks. The three parameters are optimized using the same proce- 40
dure that was described in Section 4.1. The optimum fin diameter
is also smaller than that obtained using objective function 1 for the 30

same reason as was true for the plate-fin heat sinks optimized with
20
the heat dissipation per unit mass as an objective function. 0.5 1 1.5 2 2.5 3
d (mm)
5. Comparison of optimized plate-fin and optimized pin-fin heat (b)
sinks
Fig. 11. Optimization of the heat dissipation per unit mass for pin-fin heat sinks at
In this section, the optimized heat sinks are compared in order L = W = 0.1 m, H = 0.08 m, DT = 30 K: (a) Q/M surface at d = 0.0015 m (b) optimized
to make a recommendation on which type of heat sink performs Q/M at each diameter.

120

100
Q/M (W/kg)

80

60

40

20

0
1
0.8 20
0.6 15
0.4 10

ww (mm) 0.2 5
wc (mm)
0 0

Fig. 10. Optimization of the heat dissipation per unit mass for plate-fin heat sinks Fig. 12. Ratio of the total heat dissipation of the optimized plate-fin and pin-fin
at L = W = 0.1 m, H = 0.08 m, DT = 30 K. heat sinks at DT = 30 K.
354 Y. Joo, S.J. Kim / International Journal of Heat and Mass Transfer 83 (2015) 345–356

Table 4
Geometries and Qtotal of the Qtotal optimized heat sinks at DT = 30 K, W = 0.1 m.

Fin Length Type of Optimum fin diameter Optimum horizontal spacing Optimum Optimum total heat
height H L (m) heat sink (thickness) d (ww) (m) (channel spacing) Sh (wc) (m) vertical spacing dissipation Qtotal,opt (W)
(m) Sv (m)
0.03 0.03 Pin-fin 0.0015 0.003 0.0125 7.55
Plate-fin 0.001 0.0055 – 7.18
0.03 1 Pin-fin 0.002 0.009 0.023 48.89
Plate-fin 0.001 0.013 – 53.72
0.3 0.03 Pin-fin 0.006 0.005 0.0155 25.39
Plate-fin 0.0054 0.006 – 30.59
0.3 1 Pin-fin 0.008 0.014 0.019 225.6
Plate-fin 0.0042 0.0135 – 287.1
⁄
w/minimum constraint for d (ww): 1 mm.

better under fixed volume in natural convection. For each objective
function, a region map for the ratio of the objective function
between the optimized pin-fin heat sinks and the optimized
plate-fin heat sinks will be suggested.
5.1. Objective function 1: the total heat dissipation
A region map is presented in Fig. 12; this map depicts the ratio
of the total heat dissipation between the optimized pin-fin heat
sinks and the optimized plate-fin heat sinks as a function of the
fin height (H) and the heat sink length (L). The ratio of the total
heat dissipation is defined as Q total,pin/Q total,plate. In Fig. 12, in the
region in which the ratio is above 1, the total heat dissipation of
the optimized pin-fin heat sinks is larger than that of the optimized
plate-fin heat sinks. On the other hand, the total heat dissipation of
the optimized plate-fin heat sinks is larger than that of the opti-
mized pin-fin heat sinks when the ratio is below 1. The region
map indicates the following: The optimized plate-fin heat sinks
have larger total heat dissipation than do the optimized pin-fin
heat sinks for a given base-to-ambient temperature difference in
most practical regions. Especially, the ratio becomes smaller than
1 as both the heat sink length and the fin height increase. There-
fore, it is recommended to use plate-fin heat sinks when the total
heat dissipation for a given volume of a heat sink is to be maxi-
mized. The optimized configurations at certain specific points on
the region map are listed in Table 4.
5.2. Objective function 2: the heat dissipation per unit mass

Fig. 13 depicts the ratio of the heat dissipation per unit mass
between the optimized pin-fin heat sinks and the optimized
plate-fin heat sinks. The abscissa and ordinate are the same as
those used in Fig. 12. The ratio of the heat dissipation per unit mass
is defined as (Q/M)pin/(Q/M)plate. In regions in which the ratio is
above 1, the heat dissipation per unit mass of the optimized pin-
fin heat sinks is larger than that of the optimized plate-fin heat
sinks. Therefore, Fig. 13 indicates that the optimized pin-fin heat
sinks dissipate a larger amount of heat per unit mass in most prac-
tical regions. Especially, the difference becomes large as the heat
sink length decreases. Therefore, when the mass of a heat sink is
an important constraint in its design, it is recommended to use
pin-fin heat sinks. The optimized configurations at certain specific
points on the region map are listed in Table 5.

6. Conclusion

In the present study, the thermal performances of optimized

Fig. 13. Ratio of the amounts of heat dissipated per mass of the optimized plate-fin plate-fin and optimized pin-fin heat sinks were compared analyt-
and pin-fin heat sinks at DT = 30 K. ically under fixed volume condition. A new correlation of the heat

Table 5
Geometries and Q/M values of Q/M optimized heat sinks at DT = 30 K, W = 0.1 m.

Fin Length Type of Optimum fin diameter Optimum horizontal spacing Optimum vertical Optimum heat dissipation per
height H L (m) heat sink (thickness) d (ww) (m) (channel spacing) Sh (wc) (m) spacing Sv (m) unit mass Q/Mopt (W/kg)
(m)
0.03 0.03 Pin-fin 0.001 0.0025 0.013 150.4
Plate-fin 0.001 0.0065 – 92.1
0.03 1 Pin-fin 0.0015 0.0085 0.0235 33.7
Plate-fin 0.001 0.0145 – 27.2
0.3 0.03 Pin-fin 0.001 0.002 0.031 82.1
Plate-fin 0.001 0.0095 – 55.3
0.3 1 Pin-fin 0.0025 0.0115 0.0295 37.3
Plate-fin 0.001 0.019 – 30.8
⁄
w/minimum constraint for d (ww): 1 mm, qs ¼ 2702 kg=m3 .
 Y. Joo, S.J. Kim / International Journal of Heat and Mass Transfer 83 (2015) 345–356 355

transfer coefficient for pin-fin heat sinks was proposed and vali- (3) The direction of the gravitational acceleration is the negative
dated experimentally for the optimization, while a correlation x-direction.
for plate-fin heat sinks was adopted from previous studies. Two
objective functions were used to optimize the thermal perfor- The governing equations are as follows:
mance of the heat sinks. These functions are the total heat For the fluid phase,
dissipation and the heat dissipation per unit mass. For a given Continuity equation:
base-to-ambient temperature, both objective functions were
@u @ v @w
maximized to find the optimal configuration and the optimal þ þ ¼0 ðA1Þ
thermal performance. Using the optimized heat sinks of both @x @y @z
array types, a region map was suggested for each objective func- Momentum equations:
tion. From the region maps, it was found that the answer to the !
question ‘‘Which type of a heat sink performs better?’’ depends @u @u @u 1 @P @2u @2u @2u
on the objective function. When the objective function is the total u þv þw ¼ þm þ þ
@x @y @z q @x @x2 @y2 @z2
heat dissipation, the optimized plate-fin heat sinks perform
better than do the optimized pin-fin heat sinks in most practical  g ½1  bðT  T 0 Þ ðA2Þ
applications. When the objective function is the heat dissipation
!
per unit mass, on the other hand, the optimized pin-fin heat sinks @v @v @v 1 @P @2v @2v @2v
perform better than do the optimized plate-fin heat sinks in most u þv þw ¼ þm þ þ ðA3Þ
@x @y @z q @y @x2 @y2 @z2
practical applications.
!
Conflict of interest @w @w @w 1 @P @2w @2w @2w
u þv þw ¼ þm þ 2 þ 2 ðA4Þ
@x @y @z q @z @x2 @y @z
None declared.
Energy equation:
Acknowledgement !
@T @T @T @2T @2T @2T
u þv þw ¼a þ þ ðA5Þ
This work was supported by the National Research Foundation @x @y @z @x2 @y2 @z2
of Korea (NRF) grant which is funded by the Korean Ministry of
Education, Science, and Technology (MEST) (No. For the solid phase,
2012R1A3A2026427). Energy equation:

@2T @2T @2T

Appendix A. Numerical simulation of pin-fin heat sinks
ICEPAK, commercial software based on the finite volume
method and provided by ANSYS, Inc., is used as the simulation tool.
The geometrical configurations for the numerical model are shown
in Fig. A1. The model consists of a heat sink base and fins. For the
numerical analysis, the following assumptions are imposed.
(1) The flow is steady and three dimensional.
(2) The Boussinesq approximation is used for natural convec-
tion flows.
þ þ ¼0 ðA6Þ
@x2 @y2 @z2
The boundary conditions are indicated in Fig. A1. The boundary con-
dition for the min.-y side wall is a symmetric condition. The bound-
ary conditions for the other walls are opening conditions with
ambient temperature and ambient pressure. In the present study,
ambient temperature is assumed to be 20 °C. The boundary condi-
tion for the bottom of a heat sink base is a constant heat flux con-
dition. The convergence criteria for the momentum equations and
the energy equations are relative errors of 104 and 109 ,
respectively.

Fig. A1. Computational domain for numerical simulation.

356 Y. Joo, S.J. Kim / International Journal of Heat and Mass Transfer 83 (2015) 345–356

4.5 The mesh points are densely packed near the heat sink. Grid
Case 1
Case 2 tests are conducted. The dimensions of the tested heat sink are
Case 3 L = W = 0.1 m, H = 0.03 m, d = 0.004 m, Sh = 0.006 m, and
4.0 Case 4 +10%
Case 5 Sv = 0.019 m. The power supplied to the base plate is constant at
Case 6 25 W. The number of grids is changed from 192,605 to 700,702
Case 7
nodes. The differences in the average temperature are observed
3.5
to not exceed 0.05 °C after the number of grids exceeds about
Rth, numerical

-10% 450,000 nodes. Based on the results, the grid of 450,000 nodes is
chosen for the numerical simulation.
3.0
Fig. A2 provides a comparison between the numerical results
and the present experimental data. The dimensions of the com-
2.5
pared heat sinks are listed in Table 3. As can be seen in Fig. A2,
the present experimental results correspond well to the numerical
results.
2.0 Fig. A3 provides a comparison between the numerical results
2.0 2.5 3.0 3.5 4.0 4.5 and the data predicted by the correlations of the plate-fin and
Rth, experiment the pin-fin heat sinks for various sizes of heat sink. The dimensions
of the heat sinks are listed in Table 4. The predictions of the corre-
Fig. A2. Comparison of the experimental data and the numerical results. lation correspond well to the numerical results even for large heat
sink length and large fin height.

References
Plate, L = 0.03 m, H = 0.03 m
Plate, L = 1 m, H = 0.03 m
Plate, L = 0.03 m, H = 0.3 m [1] E.M. Sparrow, S.B. Vemuri, Orientation effects on natural convection/radiation
Plate, L = 1 m, H = 0.3 m heat transfer from pin-fin arrays, Int. J. Heat Mass Transfer 29 (3) (1986) 359–
Pin, L = 0.03 m, H = 0.03 m +13% 368.
Pin, L = 1 m, H = 0.03 m
Pin, L = 0.03 m, H = 0.3 m [2] M. Iyengar, A. Bar-Cohen, Least-material optimization of vertical pin-fin, plate-
Pin, L = 1 m, H = 0.3 m
-13% fin, and triangular-fin heat sinks in natural convective heat transfer, in: The
Sixth Intersociety Conference on Thermal and Thermomechanical Phenomena
1
Rth, predicted

in Electronic Systems, ITHERM’98, IEEE, 1998, pp. 295–302.

0.1
0.1 1 1986.
Rth, numerical
Fig. A3. Comparison of the proposed correlation and the numerical results for
various dimensions of heat sinks at DT = 30 K.
When a natural convection problem is considered, the compu-
tational domain size should be determined carefully because the
accuracy of the results is affected by the domain size. To determine
the domain size, the effects of the height, the width, and the length
of the domain on the numerical results are investigated. Increasing
the height (y-direction), the width (z-direction), and the length (x-
direction) beyond 10H, 5L, and 7L, respectively, changes the aver-
age heat sink base temperature by less than 0.05 °C. Therefore,
the domain size is determined to be 10H for the height, 5L for
the width, and 7L for the length.
[3] A. Bar-Cohen, W.M. Rohsenow, Thermally optimum spacing of vertical, natural
convection cooled, parallel plates, J. Heat Transfer 106 (1984) 116–123.
[4] T. Aihara, S. Maruyama, S. Kobayakawa, Free convective/radiative heat transfer
from pin-fin arrays with a vertical base plate (general representation of heat
transfer performance), Int. J. Heat Mass Transfer 33 (6) (1990) 1223–1232.
[5] W. Elenbaas, Heat dissipation of parallel plates by free convection, Physica 9
(1942) 1–28.
[6] D.W. Van de Pol, J.K. Tierney, Free convection Nusselt number for vertical U-
shaped channels, J. Heat Transfer 95 (1973) 542.
[7] A. Bilitsky, The effect of geometry on heat transfer by free convection from a fin
array (MS Thesis), Ben-Gurion University of the Negev, Beer Sheva, Israel,
[8] J.R. Culham, M.M. Yovanovich, S. Lee, Thermal modeling of isothermal cuboids
and rectangular heat sinks cooled by natural convection, IEEE Trans. Compon.
Packag. Manuf. Technol. A 18 (1995) 559–566.
[9] A. Bar-Cohen, M. Iyengar, A.D. Kraus, Design of optimum plate-fin natural
convective heat sinks, J. Electron. Packag. (Trans. ASME) 125 (2003) 208–216.
[10] S.W. Churchill, R. Usagi, A general expression for the correlation of rates of
transfer and other phenomena, AIChE J. 18 (6) (1972) 1121–1128.
[11] D.A. Nield, A. Bejan, Convection in Porous Media, third ed., Springer, 2006.
[12] A. Bejan, Convection Heat Transfer, third ed., Wiley, 2004.
[13] J. Happel, H. Brenner, Low Reynolds number Hydrodynamics, Martinus Nijhoff
Publishers, 1983.
[14] M. Kaviany, Principles of Heat Transfer in Porous Media, second ed., Springer,
1995.
[15] G.F. Marsters, Arrays of heated horizontal cylinders in natural convection, Int.
J. Heat Mass Transfer 15 (5) (1972) 921–933.
[16] F.P. Incropera, D.P. DeWitt, T.L. Bergman, A.S. Lavine, Principles of Heat and
Mass Transfer, seventh ed., Wiley, 2013.
[17] C.B. Sobhan, S.P. Venkateshan, K.N. Seetharamu, Experimental studies on
steady free convection heat transfer from fins and fin arrays, Warme-und
Stoffubertragung 25 (6) (1990) 345–352.