Applied Thermal Engineering 29 (2009) 2731–2737

Contents lists available at ScienceDirect

Applied Thermal Engineering

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

Thermoelectric air-cooling module for electronic devices

Yu-Wei Chang a, Chih-Chung Chang a, Ming-Tsun Ke b, Sih-Li Chen a,*
a
Department of Mechanical Engineering, National Taiwan University, Taipei, Taiwan
b
Department of Energy and Refrigerating Air Conditioning Engineering, National Taipei University of Technology, Taipei, Taiwan

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

Article history: This article investigates the thermoelectric air-cooling module for electronic devices. The effects of heat
Received 19 September 2008 load of heater and input current to thermoelectric cooler are experimentally determined. A theoretical
Accepted 12 January 2009 model of thermal analogy network is developed to predict the thermal performance of the thermoelectric
Available online 19 January 2009
air-cooling module. The result shows that the prediction by the model agrees with the experimental data.
At a specific heat load, the thermoelectric air-cooling module reaches the best cooling performance at an
Keywords: optimum input current. In this study, the optimum input currents are from 6 A to 7 A at the heat loads
Thermoelectric air-cooling module
from 20 W to 100 W. The result also demonstrates that the thermoelectric air-cooling module performs
Thermoelectric cooler
Heat sink
better performance at a lower heat load. The lowest total temperature difference-heat load ratio is exper-
imentally estimated as 0.54 W K1 at the low heat load of 20 W, while it is 0.664 W K1 at the high heat
load of 100 W. In some conditions, the thermoelectric air-cooling module performs worse than the air-
cooling heat sink only. This article shows the effective operating range in which the cooling performance
of the thermoelectric air-cooling module excels that of the air-cooling heat sink only.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction spot effect also occurs on the heat sink base. Lee et al. [4] devel-
oped the spreading resistance model for this problem. According
With the advantages of small size, quietness, and reliability, to the study by Lee et al. [4], the convection resistance shows
thermoelectric cooler (TEC) is widely applied to military, aero- significant influence on the spreading resistance. For plate fin heat
space, instrument, and industrial products [1]. TEC consists of sink with impingement flow, Duan and Muzychka [5] developed a
p-type and n-type semiconductor pellets connected electrically in simply and accurate model to predict the convection performance.
series. Whenever direct current passes through the circuit, it This article develops the thermal analogy model to analyze the
causes temperature differential between TEC sides. As a result, thermal performance of the thermoelectric air-cooling module
one TEC face, which is called cold side, will be cooled while its from the existing correlations. The influences of input current
opposite face, which is called hot side, is simultaneously heated. and heat load are experimentally determined. The predictions from
In recent years, some researchers apply TEC to electronic cooling the theoretical model are compared with the experimental results,
and develop many thermoelectric cooling modules. Thermoelectric and the prediction agrees well with the experimental data. Being
air-cooling module is a typical type among these innovations. integrated with TEC does not promise to improve the cooling abil-
A thermoelectric air-cooling module, as shown in Fig. 1, com- ity of heat sink. This study finds out the operating range under
prises of a TEC and an air-cooling heat sink. For a systematical which applying TEC to the heat sink is effective.
design, Huang et al. [2] first developed a thermal design method
for thermoelectric air-cooling module. However, they did not con- 2. Thermal analogy network
sider the hot spot effect on the TEC cold side caused by the area
difference between heat source and TEC. A practical application Fig. 2 shows the thermal analogy network of thermoelectric air-
should consider the hot spot effect on the TEC cold side. Drabkin cooling module. It consists of TEC spreading resistance Rc,sp, thermo-
et al. [3] proposed the analytical solution for this issue. They electric cooler, constriction resistance Rcons, heat sink base spreading
simplified the physical system to a two-dimensional system and resistance Rb,sp, heat sink base resistance Rb, and convection resis-
derived the temperature distribution on the TEC cold side. The heat tance Rconv. Here, TEC acts as a thermal amplifier. When applying
sink above the TEC removes the heat rejection from the TEC hot DC power to TEC, the heat rejection on the hot side will be higher
side. It plays an important role on the overall performance. Hot than the heat absorption on the cold side, i.e., Qh is higher than Qc.
The thermal resistance between heater and TEC is defined as
* Corresponding author. Address: No. 1, Sec. 4, Roosevelt Road, Taipei 10617,
T d  T c
Taiwan. Tel.: +886 2 23631808; fax: +886 2 23631755. Rc;sp ¼ ; ð1Þ
E-mail addresses: d91522013@ntu.edu.tw, slchen01@ntu.edu.tw (S.-L. Chen). Qc

1359-4311/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.applthermaleng.2009.01.004
2732 Y.-W. Chang et al. / Applied Thermal Engineering 29 (2009) 2731–2737

Nomenclature

A area (m2) t thickness (m)

Dh hydraulic diameter (m)
H height of fin (m) Greek
I electricity current (A) Dn half length on the side on contact area between heater
L length of TEC side or heat sink (m) and TEC (m)
Nf number of fins a Seebeck coefficient (V K1)
Nu Nusselt number / as Eq. (8)
Pr Prandtl number k as Eq. (19)
Q heat transfer (W) gf fin efficiency
R thermal resistance based on or electrical resistance of
TEC (K W1 or X) Subscripts
R’ temperature difference-heat load ratio (W K1) HS heat sink
Re Reynolds number TE thermoelectric
T temperature (K) TEC thermoelectric cooler
T average temperature (K) a ambient
W heat sink width (m) b heat sink base
b pitch between fins (m) c TEC cold side
h heat transfer coefficient (W m2 K1) cons constriction
k thermal conductance of thermoelectric cooler conv convection
(W K1) d heater (die)
kb thermal conductivity of heat sink base (W m1 K1) f fin
kc thermal conductivity of ceramic substrate h TEC hot side
(W m1 K1) max maximum
sp spreading

temperature on the upper surface of heat sink base, and Ta is the

Ta
ambient temperature.
Tf In the application of thermoelectric cooling, it is interested in
h
Tb the relation of heat load and heater temperature. According to
Tb,max Fig. 2, the temperature difference between heater and ambient
Th can be described as

Tc T d  T a ¼ ðT d  T c Þ þ ðT c  T h Þ þ ðT h  T a Þ
c Td ¼ Q c Rc;sp þ ðT c  T h Þ þ Q h  ðRcons þ Rb;sp þ Rb þ Rconv Þ: ð3Þ
Fig. 1. The schematic of thermoelectric air-cooling module.
Eq. (3) divided by the heat load gives

Td  Ta T c  T h Q h
where Td is the heater temperature, and T c is the average tempera- ¼ Rc;sp þ þ ðRcons þ Rb;sp þ Rb þ Rconv Þ
Qc Qc Qc
ture on the TEC cold side. The thermal resistance between TEC and ð4Þ
Q
ambient are defined as ) R0total ¼ Rc;sp þ R0TEC þ h ðRcons þ Rb;sp þ Rb þ Rconv Þ
Qc
T h  T a T h  T b;max T b;max  T b T b  T f T f  T a
¼ þ þ þ where R0total is the total temperature difference-heat load ratio, and
Qh Qh Qh Qh Qh
R0TEC is the TEC temperature difference-heat load ratio. This article
¼ Rcons þ Rb;sp þ Rb þ Rconv ; ð2Þ calls R0total and R0TEC temperature difference-heat load ratios rather
than thermal resistances. Because TEC is a kind of heat engine, it
where T h is the average temperature on the TEC hot side, Tb,max is
is inappropriate to use thermal resistance to describe its behavior.
the maximum temperature on the heat sink base surface, T b is the
These two ratios might be negative due to the work done by TEC.
average temperature on the heat sink base surface, T f is the average
This article introduces the following correlations to predict the
performance of the thermoelectric air-cooling module. By assum-
ing constant thermal and electrical properties of thermoelectric
Ta material, Qc and Qh can be written as
Rconv
Tf 1
Rb Q c ¼ aIT c  I2 R  kðT h  T c Þ ð5Þ
Tb 2
Rb , sp
Tb ,max and
Rcons
Th 1
Q h ¼ aIT h þ I2 R  kðT h  T c Þ; ð6Þ
h
2
c
where a is the Seebeck coefficient of TEC, I is the input current to
Tc TEC, R is the electrical resistance of TEC, and k is the thermal con-
Rc , sp
Td ductance of TEC.
This paper derives Rc,sp from the analytical solution by Drabkin et al.
Fig. 2. Thermal analogy network. [3]. The physical system is shown in Fig. 3. The solution is given as
 Y.-W. Chang et al. / Applied Thermal Engineering 29 (2009) 2731–2737 2733

 i ¼ K i sinhðp Þ  1 2 cos p ð1  ni Þ sinh p Dni ;

1 2
Q c ðL1 L2 Þ I R þ kT h / i ¼ 1; 2: ð16Þ
Tc ¼ /1 /2 þ 2 ; ð7Þ pi i
pi i i
4Dn1 Dn2 ðaI þ kÞ aI þ k
where Then, TEC spreading resistance can be obtained with the above two
8   equation, it is
>
> K i coshðpi xÞ; x 2 0; ni  Dni
>
>  1/
ð/1;max /2;max  /  2 ÞL1 L2
> xÞ  coshðpi ðx  ni þ Dni ÞÞ þ 1;
< K i coshðp
>
 i 
Rc;sp ¼
4Dn1 Dn2 ðaI þ kÞ
: ð17Þ
/i ¼ x 2 ni  Dni ; ni þ Dni
  ; ð8Þ
>
>
>
>   ni þ Dni ÞÞ According to the model by Lee et al. [4], the spreading resis-
> K i coshðpi x Þ  coshðp ðx
>
:
i
  tance, Rb,sp, and the constriction resistance, Rcons, are modeled as
þ coshðp ðx  ni  Dni ÞÞ; x 2 ni þ Dni ; 1
i pffiffiffiffiffi pffiffiffiffiffiffiffiffiffi
Ab  ATEC kkb Ab R0conv þ tanhðktb Þ
Rb;sp ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  ; ð18Þ
2 sinhðpi Dni Þ coshðpi ð1  ni ÞÞ kb pAb ATEC 1 þ kkb Ab R0conv tanhðkt b Þ
Ki ¼ ; ð9Þ
sinhðpi Þ
and
Li 0 sffiffiffiffiffiffiffiffiffi!3=2 pffiffiffiffiffi pffiffiffiffiffiffiffiffiffi1
pi ¼ 1 ATEC Ab  ATEC
Liþð1ÞI1 Rcons ¼@ pffiffiffiffiffiffiffiffiffi 1   pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi A
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2kb ATEC Ab kb pAb ATEC
u 0 2  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 311
u
u sinh 2 B I1 kkb Ab R0conv þ tanhðkt b Þ
u B 6 iþð1Þ 7C  ; ð19Þ
u B
tBiþð1ÞI1 @1:5  4
6 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ 17 C
5A ; ð10Þ 1 þ kkb Ab R0conv tanhðktb Þ
2 Biþð1ÞI1
where

aI þ k p3=2 1
Bi ¼ L2i ; ð11Þ k ¼ pffiffiffiffiffi þ pffiffiffiffiffiffiffiffiffi : ð20Þ
kc t c Ab ATEC
In Eq. (17)–(19), kb is the thermal conductivity of heat sink base, tb
xi ni ¼ ni ; Dni
xi ¼ ; Dni ¼ ; i ¼ 1; 2: ð12Þ is the thickness of heat sink base, ATEC is the area of TEC, and Ab is the
Li Li Li
area of heat sink base. To simplify calculation, Rcons and Rb,sp can be
The maximum temperature occurs at (n1 ; n2 ), that is combined as one single thermal resistance ðT h  T b Þ=Q h . However,
this investigation compares the prediction result and experimental
Q c ðL1 L2 Þ I R þ kT h
1 2
T c;max ¼ /1;max /2;max þ 2 ; ð13Þ data and only measures single point temperature on the heat sink
4Dn1 Dn2 ðaI þ kÞ aI þ k base. Rcons is mathematically derived for comparison, that is
where
T h  T b T h  T b;max þ T b;max  T b T h  T b;max T b;max  T b
¼ ¼ þ
/i;max ¼ K i coshðpi ni Þ  coshðpi Dni Þ þ 1; i ¼ 1; 2: ð14Þ Qh Qh Qh Qh
The average temperature on TEC cold side is ¼ Rcons þ Rb:sp : ð21Þ
Z Z
1 L1 L2 Note that Tb,max is always higher than T b , therefore, Rcons is negative
T c ¼ T c dxdy because of its definition. Rb is derived by conduction resistance
L1 L2 0 0
1 2  model, that is
¼
Q c ðL1 L2 Þ
/  2 þ 2 I R þ kT h ;
 1/ ð15Þ
4Dn1 Dn2 ðaI þ kÞ ðaI þ kÞ tb
Rb ¼ : ð22Þ
k b Ab
where
The experiment adopts a plate fin heat sink in impingement
flow. The model proposed by Duan and Muzychka [5] is recom-
mended to calculate the heat transfer coefficient

0:49 ka
h ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; ð23Þ
c L=2 Dh
ReD PrDh
h
c

c where
2bH
Dh ¼ ; ð24Þ
h
bþH

V ch Dh
ReDh ¼ : ð25Þ
m
Refer to Fig. 4 for the geometry of heat sink. Rconv is written in
the following form:
2 2 2
1
Rconv ¼ ; ð26Þ
hNf 2LHgf þ hðNf  1ÞbL
1 where the fin efficiency, gf , is obtained by
1
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
tanh H 2h=kf tf
1 gf ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð27Þ
H 2h=kf tf
Fig. 3. Physical system for TEC spreading resistance [3].
2734 Y.-W. Chang et al. / Applied Thermal Engineering 29 (2009) 2731–2737

are connected to the data recorder. The thermocouples have a

f measurement error of ±0.5 °C. The data recorder has a measure-
ment error of ±0.1%. The input power has a measurement error
of ±0.5%. By performing error propagation of uncertainty, the
maximum error of thermal resistance is within ±5%. Another
ch experiment is performed to evaluate the air-cooling heat sink
f
resistance. The apparatus is the same with that for the TEC air-
cooling module besides the TEC is removed. The experimental
parameters are heat load of the heater. The heat load is from
20 W to 100 W with an increment of 20 W.
b The physical properties of TEC are essential for the prediction.
Fig. 6 shows the experimental apparatus for these physical proper-
ties. The TEC cold side is attached on the heater, and the hot side is
Fig. 4. Geometry of plate fin heat sink in impingement air flow [5]. attached on water jacket. The heating area of the heater is identical
with the area of TEC. The water flow rate is pumped up to
3. Experimental apparatus 12 cm3 s1. It keeps the temperature difference between inlet
and outlet water flow within ±0.2 °C. The hot side temperature is
Fig. 5 shows the experimental apparatus for the thermoelectric kept at 50 °C. The experimental heating powers are 30 W, 40 W,
air-cooling module. The heating area of the heater is 30 mm and 50 W. The experimental input currents are 0A, 2A, 4A, 6A,
square. It is attached on the TEC cold side with thermal grease. 8A, and 10A. Fig. 7 shows the experimental result for the three
The TEC is manufactured by Kryotherm. The ceramic substrates physical properties of TEC. This investigation performs regression
on the both sides are made of nitride alumina whose thermal con- analysis to obtain the physical properties. The regression is mod-
ductivity is 180 W m1 K1. The dimension of the TEC is 40 mm  eled into Eq. (5). For the TEC used in this study, the Seebeck coef-
40 mm  3.2 mm. The TEC hot side is attached to the air-cooling ficient is 0.0566 V K1, the thermal conductance is 1.392 KW1,
heat sink with thermal grease. The heat sink is made of aluminum and the electrical resistance is 1.857 ohm.
alloy, it size is 77 mm  68 mm  10 mm. The fins are 68 mm in
length, 25 mm in height, and 1 mm in thickness. There are 35 fins 4. Results and discussion
with the pitch of 1.2 mm. The air flow rate of the fan is 45.48 CFM.
The air cylinder presses 10 kgf on the top to fix the heater, the Rcons, Rb,sp, Rb, Rconv of the heat sink used in this study are calcu-
TEC, and the heat sink. Thermal resistance caused by thermal lated by performing Eq.(17)–(26). As the result of calculation, Rcons
grease exists between the TEC and the heat sink. Another experi- is 0.0194 K W1, Rb,sp is 0.0477 KW1, Rb is 0.0114 K W1,and
ment was done for the interface resistance under the same heat- Rconv is 0.1905 K W1. Since the calculation of R0TEC involves T h
ing size with TEC and pressure. The result shows the interface and T c , the prediction must perform iteration skill. For prediction,
resistance is 0.001 °C W1 which is within the measurement error. the heat load Qc, input current I, and ambient temperature Ta must
Therefore, this study neglects the interface thermal resistance. be given. At the beginning of calculation guessed temperatures are
There are three T-type thermocouples to measure the tempera- given. Temperature iteration is performed until the tolerance is
tures. One is mounted at the center of the heating area for the within 103. Once the temperature prediction is done, the thermal
temperature Td. Another thermocouple is mounted at the center resistances and temperature difference–heat load ratios can be
on the heat sink base surface for the maximum heat sink base immediately obtained.
temperature Tb,max. One thermocouple is located by the experi- The experiment only measures the heater temperature Td, the
mental system for ambient temperature Ta. The thermocouples maximum temperature on the heat sink base surface T b;max , and

Fig. 5. The experimental apparatus for thermoelectric air-cooling module.

 Y.-W. Chang et al. / Applied Thermal Engineering 29 (2009) 2731–2737 2735

Fig. 6. The experimental apparatus for TEC physical properties.

the ambient temperature Ta. To meet the experimental measure- Qh

R0TE ¼ Rc;sp þ R0TEC þ Rcons ð29Þ
ment, the total temperature difference–heat load ratio can be ex- Qc
pressed as
and
T d  T a T d  T b;max T b;max  T a Qh
R0total ¼ ¼ þ ¼ R0TE þ RHS ; ð28Þ R0HS ¼ ðRb;sp þ Rb þ Rconv Þ: ð30Þ
Qc Qc Qc Qc
where R0TE is called thermoelectric temperature difference-heat load Fig. 8 shows the prediction and experimental result of R0TE
ratio and R0HS is called heat sink temperature difference-heat load ra- versus input current at every heat load. R0TE decreases with increas-
tio. These two ratios can be obtained by rearranging Eq. (4), that ing input current at each heat load. When heat load is under 80 W,
gives it becomes negative if input current is large enough. Furthermore,
a higher heat load needs a higher input current for negative R0TE .
When heat load is 100 W, it can not achieve negative R0TE even
40 the input current is 10 A. R0TE represents the performance caused
by the thermoelectric effects. The Peltier effect grows with increas-
30 ing input current. When the Peltier effect is strong enough, the TEC
cold side temperature will be lower than the TEC hot side temper-
ature, which leads to a negative R0TE . If heat load is too high to be
20
overcome by the Peltier effect, R0TE will be positive. Also shown in
Fig. 8, the trend of prediction agrees with the experimental data.
10
The accuracy of prediction is higher at a higher heat load and a
Th -Tc (K)

α=0.0566 VK-1
k=1.392 WK-1 lower input current.
0 R=1.857 ohm Fig. 9 illustrates R0HS versus the input current at different heat
loads. The prediction trend agrees with the experimental result.
-10 Exp. Curve-fitting R0HS dramatically increases with increasing input current at every
30W heat load. On the contrary, it decreases with increasing heat load
-20 40W
when the input current is fixed. The variation of R0HS results from
50W
the ratio of Qh to Qc, which is observed from Eq. (28) because Rb,sp,
-30
Rb, and Rconv are constant values. As input current increases, Qh
increases because of the Joule heat generated by input current. It
-40
0 2 4 6 8 10 12 makes the ratio of Qh to Qc grow with increasing input current. A
I (A) larger Qc causes a smaller ratio of Qh to Qc. Hence, R0HS becomes low-
er at a higher Qc when applying the same input current.
Fig. 7. The regression analysis for the TEC physical properties.
2736 Y.-W. Chang et al. / Applied Thermal Engineering 29 (2009) 2731–2737

3 2
Exp. Model
Exp. Model
20W
2 100W
40W 1.5 80W
60W
60W
80W
1 40W
100W
1 20W

R'total (KW-1)
R'TE (KW-1)

0
0.5
-1

0
-2

-0.5
R'total,w/o TEC
-3

-4 -1
0 2 4 6 8 10 0 2 4 6 8 10
I (A) I (A)
Fig. 8. The prediction result and the experimental result of R0TE versus input current Fig. 10. The prediction result and experimental result of R0total versus input current
at different heat loads. at different heat loads.

Fig. 10 shows R0total versus the input current at different heat cated in Fig. 10. When R0total is higher than R0total; w=o TEC , it means that
loads. Because of the contrary trends of R0TE and R0HS on the input the thermoelectric air-cooling module performs worse than the
current, an optimum input current forms to achieve the lowest heat sink not integrated with TEC. With the aid of the present
R0total under each heat load. As input current is increasing, R0total de- model, this investigation finds out the operating situation under
creases first and then reverses at the optimum input current. The which applying TEC to the heat sink is useful. The shade area in
optimum input currents predicted by the model are 5 A from Fig. 11 is the effective operating range. When operating under
20 W to 40 W and 6 A from 60 W to 100 W, while the experimental the area, the thermoelectric air-cooling module performs better
result indicates 6 A from 20 W to 80 W and 7 A at 100 W. The than the heat sink only. As shown in the figure, the effective range
result also shows the performance of thermoelectric air-cooling of input current becomes shorter and shorter as heat load in-
module is better at low heat loads than at high ones. At the heat creases. When operating outside the area, integrating TEC with
load of 20 W, R0total is experimentally estimated as 0.54 K W1 at the heat sink is not recommended. In this case, the maximum
the optimum input current of 6 A, while it is 0.664 K W1 at 7 A. effective operating heat load is 50.5 W.
Note that the negative R0total means that the heater temperature is
lower than the ambient temperature. The trend of prediction for 5. Conclusion
R0total also agrees with that of experimental data. The root mean
square deviation of the prediction is 0.13 K W1. According to the thermal analogy network, the total tempera-
The heat sink resistance not integrated with TEC, R0total; w=o TEC , is ture difference–heat load ratio R0total , consists of the thermoelectric
experimentally determined. The value is 0.385 K W1. It is indi- temperature difference–heat load ratio R0TE , and the heat sink tem-

4
60
Exp. Model
20W
40W
3 50
60W
80W
100W
R'HS (KW-1)

40
Qc (W)

30

20

0
0 2 4 6 8 10 10
0 2 4 6 8 10
I (A)
I (A)
Fig. 9. The prediction result and experimental result of R0HS versus input current at
different heat loads. Fig. 11. Map of effective operating range.
 Y.-W. Chang et al. / Applied Thermal Engineering 29 (2009) 2731–2737 2737

perature difference–heat load ratio R0HS ; R0TE represents the per- References
formance caused by the Peltier effect. Increasing input current
can reduce R0TE . However, it also increases the Joule heat genera- [1] S.B. Riffat, X. Ma, Thermoelectrics: a review of present and potential
applications, Applied Thermal Engineering 23 (2000) 913–935.
tion of TEC, which induces a large R0HS . An optimum input cur- [2] B.J. Huang, C.J. Chin, C.L. Duang, A design method of thermoelectric coolers,
rent exists to reach the lowest total resistance at a specific International Journal of Refrigeration 23 (2000) 208–218.
heat load. The result also shows the system works better at a [3] I.A. Drabkin, L.B. Yershova, D.A. Kondratiev, G.G. Gromov, The effect of the
substrates two-dimensional temperature distribution on the TEC performance,
lower heat load. This article also develops a thermal resistance in: Proceeding of 8th European Workshop on Thermoelectrics, vol. 12, 2004.
model to predict the performance of thermoelectric air-cooling [4] S. Lee, S.S. Van Au, K.P. Moran, Constriction/spreading resistance model for
module. With the model, this study verifies the effective operat- electronics packing, in: ASME/JSME Thermal Engineering Conference, vol. 4,
1995, pp. 199–206.
ing range of the device. The result shows the cooling module can
[5] Z. Duan, Y.S. Muzychka, Experimental investigation of heat transfer in
reach a higher performance than that of the heat sink when the impingement air cooled plate heat sinks, Transactions of the ASME 128
heat load is below 50.5 W. (2006) 412–418.