Applied Energy 219 (2018) 93–104

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

An effective method of evaluating the device-level thermophysical T

properties and performance of micro-thermoelectric coolers
⁎ ⁎
Dongfang Suna, Limei Shenb, , Miao Sunb, Yu Yaob, Huanxin Chenb, , Shiping Jina
a
Department of Engineering Thermo-physics, Huazhong University of Science and Technology, Wuhan 430074, China
b
Department of Refrigeration & Cryogenics Engineering, Huazhong University of Science and Technology, Wuhan 430074, China

H I GH L IG H T S

• AThethree-dimensional numerical model of micro thermoelectric cooler is developed.

• A comparison
joint impact of boundary and size effects on TE properties is discussed.
• The impact of between the numerical and reported experimental results is conducted.
• The cooling capacity
interfacial resistances on cooling performance of TEC is discussed.
• and optimal working condition of the micro TEC is analyzed.

A R T I C LE I N FO A B S T R A C T

Keywords: Despite the success of achieving thermoelectric materials with high figure of merit, precisely evaluating the
Device level performance of micro-thermoelectric coolers remains challenging at the microdevice level because of various
Interfacial effects interfacial effects and device construction. This study develops a method for the effective evaluation of the
Size effect device-level thermophysical properties capturing various interfacial and size effects, and establishes a three-
Cooling performance
dimensional numerical model to evaluate the cooling performance of micro-thermoelectric coolers. The model is
Micro-thermoelectric cooler
validated by the reported experimental data. The impact of interaction between boundary and size effects is
captured in the investigation of Seebeck coefficient, thermal conductivity and electricity resistivity of the
thermoelectric materials at the device-level. Contact resistances are also considered in analyzing the cooling
performance. Results indicate that the device-level figure of merit decreases by 5–18.1% with decreased ther-
moelectric element thickness from 20 μm to 5 μm. The boundary effects considerably weaken the cooling per-
formance of the microdevice, and a higher heat flux corresponds to a greater impact of boundary effects. Cooling
temperature increases by 6.1 K due to the boundary effects when heat flux is 300 W/cm2, while the temperature
difference decreases by 17.1%. Finally, the three-dimensional numerical model is performed to evaluate the
cooling performance and optimal working condition of the micro-thermoelectric cooler. At heat flux of 300 W/
cm2 and 200 W/cm2, the minimum cold side temperatures of 310.7 K and 287.3 K are predicted to be achieved at
11 μm/20 mA (Hte/ I ), 15 μm/16 mA, respectively.

1. Introduction [5]. Recognizing the importance of the thermophysical properties of

materials evaluated by the figure of merit (ZT) to TECs, many re-
Thermoelectric cooler (TEC) is widely used in electronic devices, searchers have been devoting themselves to developing new thermo-
refrigerators, air conditioning systems, photovoltaic equipment, med- electric (TE) materials to improve ZT [6]. However, when TE materials
ical instruments and industrial fields due to its numerous advantages, with high ZT are fabricated into thin-film microdevices, the effective
such as compact size, no moving part, environmentally friendly and thermophysical properties at the device level are diminished, and the
temperature control capability [1–3]. Recent advance in fabrication practical cooling performance is severely limited by interfacial effects
techniques have enabled the fabrication of microdevices [4]. Thin-film [7]. For practical applications, the practical cooling performance of
micro-TEC has got more attention for its potential application in the on- micro-TECs has also been investigated [8].
demand localized cooling of microelectronic and optoelectronic devices The low efficiency of common bulk TE materials has limited the

⁎
Corresponding authors.
E-mail addresses: ep_shenlimei@hust.edu.cn (L. Shen), chenhuanxin@tsinghua.org.cn (H. Chen).

https://doi.org/10.1016/j.apenergy.2018.03.027
Received 21 November 2017; Received in revised form 5 March 2018; Accepted 17 March 2018
Available online 22 March 2018
0306-2619/ © 2018 Elsevier Ltd. All rights reserved.
D. Sun et al. Applied Energy 219 (2018) 93–104

Nomenclature R e,co electrical resistance of copper connector/lead [Ω]

R e,c1 electrical contact resistance between Cu and thermo-
ZT figure of merit electric element [Ω·m2]
k thermal conductivity [W/(m·K)] L Lorenz number [V2/K2]
T room temperature [K] Rk,c thermal contact resistance [K·m2/W]
Th0 temperature of hot surface of thermoelectric element [K] Re,c electrical contact resistance [Ω·m2]
Tc0 temperature of cold surface of thermoelectric element [K] Tl temperature of the lower Cu leads [K]
Te temperature of electron [K] ΔText external temperature difference, Tl−Tc0 [K]
Tp temperature of phonon [K] ΔT temperature difference, Th−Tc [K]
Hte thickness of thermoelectric element [m] Ioptimal optimal current achieving the minimum cold side tem-
I electrical current [A] perature [A]
Qk,h0 − c 0 heat flow across thermoelectric element [W] Tcmin minimum cold side temperature of micro-thermoelectric
kD device-level thermal conductivity of thermoelectric mate- cooler [K]
rial [W/(m·K)] ΔTmax maximum temperature difference of micro-thermoelectric
Ate cross area of thermoelectric element [m2] cooler [K]
ke thermal conductivity of electron [W/(m·K)]
kp thermal conductivity of phonon [W/(m·K)] Greek letters
je electrical current density [A/m2]
P strength of hole/electron–phonon interaction [W/(m3·K)] αs seebeck coefficient [V/K]
Rk,b,p thermal boundary resistance of phonon [K·m2/W] ρe electrical resistivity [Ω·m]
Rk,b,e thermal boundary resistance of electron [K·m2/W] α s,D device-level Seebeck coefficient of thermoelectric material
R e,b electrical boundary resistance [Ω·m2] [V/K]
Rk,b total thermal boundary resistance [K·m2/W] Δφh0 − c 0 electromotive force between cold and hot interfaces of
Re∗ dimensionless thermal boundary resistance of electron thermoelectric element [V]
Rp∗ dimensionless thermal boundary resistance of phonon α s,b boundary seebeck effect [V/K]
Rk∗ dimensionless thermal boundary resistance γ ratio of ke and kp
Hte∗ dimensionless thickness of thermoelectric element δ cooling length [m]
R e,D electrical resistance [Ω] ρe,D device-level electrical resistivity of thermoelectric mate-
R e,p peltier resistance [Ω·m2] rial [Ω·m]
ZD T device-level figure of merit of thermoelectric material αs,D,p − n total device-level Seebeck coefficient of p-type and n-type
Qc0 cooling capacity at the cold junction of thermoelectric elements [V/K]
elements [W] αs,D,p device-level Seebeck coefficient of p-type elements [V/K]
Q h0 heating rejection at the hot junction of thermoelectric αs,D,n device-level Seebeck coefficient of n-type elements [V/K]
elements [W] ψ change percentage
KD,p − n total thermal conductance of p-type and n-type elements βD device-level properties of thermoelectric material (e.g.,
[W/K)] α s,D , kD , ρe,D , ZD T )
R e,D,p − n total electric resistance of p-type and n-type elements [Ω] βm intrinsic properties of thermoelectric material (e.g., α s , k ,
Qc net cooling capacity of micro-thermoelectric cooler [W] ρe , ZT )
Qh net heat rejection of micro-thermoelectric cooler [W]
Subscripts
Tc temperature of cold surface of micro-thermoelectric cooler
[K]
h0 hot surface of thermoelectric element
Th temperature of hot surface of micro-thermoelectric cooler
c0 cold surface of thermoelectric element
[K]
D device-level properties of thermoelectric material
Rk,s thermal resistance per unit area of substrate [K/W]
te thermoelectric element
Rk,c u thermal resistance per unit area of copper connector/lead
k thermal properties
[K/W]
b boundary resistance
Rk,c1 thermal contact resistance between Cu and thermoelectric
co copper connector
element [K·m2/W]
l copper lead
Rk,c 2 thermal contact resistance between Cu and substrate
m base properties
[K·m2/W]
Acu cross area of copper connector/lead [m2]

applications of TECs [9]. Recent developments in nanotechnologies extraneous or parasitic factors. Despite the success of achieving TE
have led to significant ZT enhancement [10–12], and considerable ef- materials with high ZT, the ZT of micro- TECs made of high ZT mate-
forts have been made to explore low-dimensional TE materials with rials is typically lower than expected, presumably because of the fab-
high efficiency. Low-dimensional materials and nanostructures, such as rication of TE couple, various interfacial effects, and device construc-
quantum wells, super lattices (SLs), quantum wires, and quantum dots, tion [7,8,16,17].
offer new ways to improve ZT by manipulating the electron and phonon For practical applications, thin-film microdevices fabricated with
properties of materials [13,14]. Venkatasubramanian et al. [15] re- high ZT TE materials have also been investigated. Goncalves et al. [18]
ported that Bi2Te3/Sb2Te3 SLs with a period of 6 nm fabricated by presented flexible micro-TECs made of ultrathin (10 μm) Bi2Te3/Sb2Te3
MOCVD technique had a maximum ZT of 2.4 at 300 K. Harman et al. TE elements. However, the tested maximum temperature difference of
[16] presented PbTe/PbSeTe based quantum dots SLs with an intrinsic 5 K between the cold and hot sides of the device was much lower than
ZT value of 2.0 at 300 K. However, the measured operating device the simulated value (18 K). Bulman et al. [19] reported the experi-
thermoelectric ZT declined to 1.6 at the system level because of mental results of thin-film SL thermoelectric modules, demonstrating an

94
D. Sun et al. Applied Energy 219 (2018) 93–104

external temperature drop of 55 K under a heat pumping capacity of approximately 1 K. Owoyele et al. [21] proposed a novel TEC archi-
128 W/cm2. The measured average ZT300 for the best SL devices was tecture that employed thin-film TE elements on a plastic substrate in a
0.75, which was much lower than that of the materials (2.4 in p-type corrugated structure. They showed that parasitic heat transfer through
and 1.2 in n-type). Da Silva et al. [20] measured the performance of a the substrate in the C-TE module is a source of performance losses,
column-type micro-TEC made of Sb2Te3/Bi2Te3 with an average which can be minimized by proper thickness selection. To theoretically
thickness of 4.5 μm. The average temperature drop achieved was analyze the performance of micro-TECs, Ju et al. [22] proposed a

Fig. 1. Flowchart of the three-dimensional numerical model.

95
D. Sun et al. Applied Energy 219 (2018) 93–104

model based on the coupled Boltzmann equations, W–F law, and

boundary and size effects was firstly built to evaluate the thermal
thermophysical of TE materials at the microdevice level. Then, the
numerical model was introduced to a three-dimensional model of a
microcooler and solved in COMSOL Multiphysics software package. To
validate the developed numerical model, the numerical results were
compared with the reported experimental data. To assess its practical
application, the numerical model was also performed to evaluate the
cooling performance of the micro-TEC under various operating condi-
tions. Note that the boundary effects were considered only at the metal/
TE interfaces and adopted to TE elements, whereas contact resistances
were considered at all interfaces and presented as contact layers. And,
this study ignored the radiation from the surfaces to the ambient and
from one leg to the others or to the gap, as well as the radiation between
substrates.

Fig. 2. TE element structure of a micro-TEC. 2.1. Device-level thermophysical property

modified definition of ZT capturing the contact resistances and The performance of TE materials is generally evaluated using ZT. A
boundary Seebeck effect, in which the contact resistances were simply higher ZT corresponds to a better performance of a TE device. ZT is a
connected into TE elements in series. They found that contact re- dimensionless parameter defined by
sistances considerably influence the performance of TE refrigerators. Da α s2
Silva et al. [7] developed a one-dimensional theoretical model to ana- ZT = T
ρe k (1)
lyze the effects of electron and phonon boundary resistances on in-
creasing thermal conduction resistance and decreasing Seebeck coeffi- where α s , k and ρe are the Seebeck coefficient, thermal conductivity and
cient of TE elements, however, the effects of interfacial resistances on electrical resistivity of the TE materials, respectively and T is the tem-
the electrical resistance of TE elements was not yet considered. Kong perature (usually room temperature) [26].
et al. [4] performed a three-dimensional analysis of the cooling per- Given that the thickness of the considered TE element was ex-
formance of micro-TECs under controlled temperature differences tremely small, its thermophysical properties at the device level and
considering the non-equilibrium between electrons and phonons at the cooling performance can be influenced by boundary effects (e.g.,
TE/metal interfaces. Many researchers noticed that contact resistances boundary Seebeck effect, boundary resistances and electron–phonon
(originated in the fabrication process) exerted severe adverse effects on non-equilibrium) and size effect. The TE element structure of a micro-
the cooling performance of micro-TE devices [17,23–25]. Nevertheless, TEC is presented in Fig. 2. The device-level thermophysical properties
boundary resistances and boundary Seebeck effect have received lim- (e.g., Seebeck coefficient, thermal conductivity, electrical resistivity
ited attention. In view of the inevitable interfacial and size effects, and ZT) can be defined as follows.
numerical methods still face challenges in forecasting the performance The Seebeck coefficient of the TE material at the device level is
of micro-TECs. given by [12,27]
It can be noted from the above analysis that the joint impact of
interfacial and size effects on the thermophysical properties of TE ma- Δφh0 − c 0
α s,D =
terials at the device level and the cooling performance of microcoolers Th0−Tc 0 (2)
has remained largely unexplored. And, the numerical methods for es-
where Th0 and Tc0 are the temperatures of the hot and cold surfaces of TE
timating the performance of micro-TE devices still need to be further
element, respectively. Δφh0 − c 0 is the electromotive force between the
developed. Thus, this study developed a method based on coupled
cold and hot surfaces of the TE element due to both bulk and boundary
Boltzmann equations and Wiedemann–Franz (W-F) law to evaluate the
Seebeck effects, and can thus be obtained by [7]
device-level thermophysical properties and performance of micro-TECs
capturing various interfacial (e.g., contact, boundary) and size effects. Δφh0 − c 0 = α s,b [Th0−Te (−Hte/2)] + αs [Te (−Hte/2)−Te (Hte/2)]
First, the numerical results were validated using the reported experi-
mental data. Then, the joint influence of electron and phonon boundary + αs,b [Te (Hte/2)−Tc 0] (3)
resistances, boundary Seebeck and size effects on the Seebeck coeffi-
where α s,b is the boundary Seebeck effect, Te is the temperature of
cient, thermal conductivity and electricity resistivity of TE materials at electrons and Hte is the thickness of the TE element. Therefore, the
the device level were assessed. To investigate the influence of inter-
Seebeck coefficient of the TE material at the device level becomes
facial and size effects, sensitivity analyses of boundary resistances,
contact resistances and thickness of the TE element were also con- α s,D
ducted. Finally, a three-dimensional numerical model was performed to α s,b [Th0−Te (−Hte/2)] + αs [Te (−Hte/2)−Te (Hte/2)] + αs,b [Te (Hte/2)−Tc 0]
predict the cooling performance and optimal working condition of the =
Th0−Tc 0
micro-TEC. The contact resistances were considered in analyzing the
(4)
cooling performance of micro-TE devices under various operating
conditions. When the electrical current, I equals to zero, the heat flow across the
TE element can be expressed as [7,27]
2. Numerical modeling kDAte (Th0−Tc 0)
Qk,h0 − c 0 =
Hte (5)
This study aimed to develop an effective three-dimensional nu-
merical model for analyzing the device-level thermophysical properties where kD is the effective thermal conductivity of the TE material at the
and predicting the cooling performance of micro-TECs. The flowchart of device level, and Ate is the cross area of the TE element. Considering the
the three-dimensional numerical model is shown in Fig. 1. A numerical heat flow of phonons and electrons, it can be obtained that

96
D. Sun et al. Applied Energy 219 (2018) 93–104

∂T ∂Tp ⎤ ∂T ∂Tp ⎤ By analyzing Eqs. (13)-(16) combined with Eqs.(10) and (11) under
Q k,h0 − c 0 = Ate ⎡−ke e −kp = Ate ⎡−ke e −kp
⎢
⎣ ∂z ∂z ⎥⎦− H2te ⎢
⎣ ∂z ∂z ⎥⎦H2te the condition of electrical current, I equals to zero, it can be obtained
(6)
that
where Tp is the temperature of phonons, ke and kp are respectively the Th0 +Tc0
thermal conductivity of electrons and phonons. Then, the thermal A=
2 (17)
conductivity of the TE material at the device level can be given as
B = (Th0−Tc0) X (18)
kD =
Hte ⎡−ke ∂Te −kp ∂Tp ⎤
(Th0−Tc 0) ⎢ ∂z ∂z ⎥ C=0 (19)
⎣ ⎦Hte 2 (7)
(Th0−Tc0) Y
To calculate α s,D and kD , the unknown Te and Tp were obtained based D=
2sinhHte∗ (20)
on the kinetics of the electrons and phonons in an electric or a tem-
perature field, which can be described by coupled Boltzmann equations where the dimensionless parameters are expressed by
[4,28]. For a three-dimensional geometry, the equations can be ex- Rk,b ∗ Rk,b ∗ Rk,b Ate k ∗ H
pressed as Re∗ = ,Rp = ,Rk = ,Hte = te .
Rk,b,e Rk,b,p Hte 2δ (21)
−∇ (k e ∇Te ) = ρe je2 −P (Te−Tp) (8)
Re∗ Rp∗ (1 + γ )2
γHte∗ cothHte∗ + 2Rk∗
−∇ (kp∇Tp) = P (Te−Tp) (9) X=
Re∗ Rp∗ (1 + γ )2 ∗ cothH ∗
γHte
Re∗ + 2Rk∗
+ te
+ Rp∗ γ 2
where je is electrical current density, P is the strength of hole/electro- (1 + 2Rk∗)−1 (22)
n–phonon interaction, which can be obtained from the density of
electron/hole, the energy relaxation time and the Boltzmann constant γ 2Rp∗−γRe∗
Y=
Re∗ Rp∗ (1 + γ )2 ∗ cothH ∗
[4,28]. By regarding je , ρe and P as constants, Eqs. (8) and (9) can be Re∗ + +
γHte te
+ Rp∗ γ 2
2Rk∗ (1 + 2Rk∗)−1 (23)
solved by defining a center-of-thermal-conductivity temperature
kTetc = ke Te + kp Tp , where k = ke + kp [27]. The solutions are And Rk,b is the total thermal boundary resistance, which is given by
2
Rk,b,e Rk,b,p
z
ρe je2 ⎡
⎣
( ) −z ⎤⎦
Hte
2
2
ρe je2 δ 2 Rk,b =
Rk,kD = Xkb,e + Rk,b,p
Te = A + B + + (24)
Hte 2k γk
1 z z Substituting Eqs. (10), (11) and (17)-(20) into Eqs. (4) and (7)
− ⎡C cosh ⎛ ⎞ + Dsinh ⎛ ⎞ ⎤ produces that
γ⎣ ⎝δ ⎠ ⎝ δ ⎠⎦ (10)
Y Y
Hte 2 α s,D = ⎜⎛X − ⎟⎞ αs + ⎜⎛1 + −X ⎟⎞ αs,b
z
ρe je2 ⎡
⎣
( ) −z ⎤⎦
2
2
ρe je2 δ 2 z z ⎝ γ⎠ ⎝ γ ⎠ (25)
Tp = A + B + − + ⎡C cosh ⎛ ⎞ + Dsinh ⎛ ⎞ ⎤
Hte 2k k ⎣ ⎝δ ⎠ ⎝ δ ⎠⎦ kD = Xk (26)
(11)
With regard to the electrical resistivity of the TE material at the
where γ = ke / kp , and δ is the cooling length, which represents the device level, it is calculated by
distance from the boundary required for phonons and electrons to reach
equilibrium, and is given by R e,D Ate
ρe,D =
Hte (27)
ke kp
δ= where R e,D is the total electrical resistance expressed by
P (ke + kp) (12)
ρe Hte R e,b R e,p
The constants A, B, C and D can be obtained by adopting the R e,D = +2 +
Ate Ate Ate (28)
boundary conditions of heat flow for electrons and phonons as follows
For phonons, the boundary conditions are, at z= Hte /2 where R e,p is the Peltier resistance[29] given by

Tc0−Tp ∂Tp R e,p = Hte·λ (αs + αs,b)(αs−αs,b) T / ke (29)

= −kp
Rk,b,p Hte ∂z Hte
Here,
2 2 (13)
And z= −Hte /2 , γ (1 + γHte∗ th (Hte∗ )) + (1 + γHte∗2) th (Hte∗ )/ Hte∗
λ=
(1 + Hte∗ th (Hte∗ ))(1 + γHte∗2) + γ (1 + γHte∗ th (Hte∗ ))(1 + Hte∗2) (30)
Tp−Th0 ∂Tp
= −kp Substituting Eqs. (28) and (29) into Eq. (27), it becomes
Rk,b,p H
− te
∂z H
− te
2 2 (14)
2R e,b λ (αs2−αs2,b) T
ρe,D = ρe + +
where Rk,b,p is the thermal boundary resistance of phonons. Hte ke (31)
For electrons, the boundary conditions are, at z= Hte /2
Thus, ZT of the TE materials at the device level can be obtained by
Tc0−Te ∂T je2
= (α s−α s,b) je Te | Hte −ke e −R e,b 2
Rk,b,e Hte
2
2 ∂z Hte
2
2
(15)
ZD T =
2
α s,D
T=
( (X − ) α + (1 +Y
γ s
Y
γ
−X )α ) T
s,b

ρe,D kD 2R e,b λ (αs2 − αs2,b) T

And z= −Hte /2 , Xk (ρe + Hte
+ ke
) (32)
Te−Th0 ∂Te je2 Eqs. (21)-(31) shows that the thermophysical properties of the TE
= (α s−α s,b) je Te|− Hte −ke + R e,b
Rk,b,e H
− te
2 ∂z H
− te 2 material at the device level (e.g., αs,D , kD and ρe,D ) are lumped equiva-
2 2 (16)
lent parameters of intrinsic physical properties (e.g., αs , k and ρe ),
where Rk,b,e is the thermal boundary resistance of electrons and R e,b is geometrical dimensions (e.g., Ate and Hte), boundary resistances (e.g.,
the electrical boundary resistance. R e,b , Rk,b,e and Rk,b,p ) and boundary Seeebeck coefficient (αs,b ) of TE

97
D. Sun et al. Applied Energy 219 (2018) 93–104

elements. The presented method is still applicable and valid if the Table 1
structure, geometrical dimensions, or intrinsic physical properties were Geometrical dimensions of the micro-TEC.
changed.
Components Dimensions

Cu header 580 μm × 560 μm × 30 μm

2.2. Cooling performance of micro-TE device Sn solder 520 μm × 500 μm × 25 μm
TE elements 250 μm × 500 μm × 7.5 μm
Cu post 180 μm × 30 μm
A micro-TEC was built in COMSOL, and the governing equations in In solder 180 μm × 30 μm
this study are expressed in the following statement[5,30]: Cu leader 278 μm × 500 μm × 30 μm
The cooling capacity at the cold junction of the TE element can be
described by
3. Validation and analysis
1
Qc0 = αs,D,p − n Tc 0 I −KD,p −n (Th0−Tc 0)− I 2R e,D,p − n
2 (33) 3.1. Case study

The heat rejection at the hot junction of TE element is given by, To validate the presented model, a reported micro-TEC with ex-
perimental and numerical results was selected [19,30], and a model
1
Q h0 = αs,D,p − n Th0 I −KD,p −n (Th0−Tc 0) + I 2R e,D,p − n with the same structure and geometrical dimensions was built, as
2 (34)
shown in Fig. 3. It consists of a single couple of n-type and p-type Bi2Te3
where Qc0 and Q h0 are respectively the cooling capacity at the cold SL elements with in-plane dimensions of 250 μm × 500 μm and a
junction of the TE element and the heat rejection at the hot junction of thickness of 7.5 μm. The diameter of the circular contacts (Cu post)
TE element. αs,D,p − n = αs,D,p−αs,D,n , is the total Seebeck coefficient of p- between the TE elements and the metal leads is 180 μm. Each element
type and n-type elements (e.g., αs,D,p and αs,D,n ). has one such circular contact (1N-1P). The detailed geometrical di-
KD,p − n = kD,pAte / Hte + kD,nAte / Hte , is the total thermal conductance of mensions are provided in Table 1. As shown in Table 2, the electrical
p-type and n-type elements, and R e,D,p − n = R e,D,p + R e,D,n , is the total and thermal contact resistances of the TE element–solder interface
electrical resistance of p-type and n-type elements. (7 × 10−11 Ω·m2, 6.25 × 10−6 K·m2/W) and the metal–solder interface
The net cooling capacity, net heat rejection and thermal phenom- (3 × 10−12 Ω·m2, 4.16 × 10−7 K·m2/W) were also derived from [30].
enon of the micro-TE device are given by The adopted Seebeck coefficient is 301 μV/K for p-type and −301 μV/K
for n-type. Thermal conductivity and electrical resistivity of the SL TE
Tc−Tc 0 R e,c1 elements are 1.08 × 10−5 Ω·m and 1.2 W/(m·K), respectively [5,30].ke
Qc = = Qc0−I 2R e,cu−2I 2
Rk,s +
Rk ,c 2
+ Rk,cu +
Rk ,c1
Ate can be obtained according to W-F law ( ρe k e = LT ), where L is the
Acu 2Ate (35) Lorenz number, approximately 1.5 × 10−8 V2/K2 for non-degenerate
semiconductors at 300 K [31]. Thus it can be calculated that ke equals to
Th0−Th R e,c1 0.42 W/(m·K) and k p 0.78 W/(m·K) (k p = k−ke ). The boundary Seebeck
Qh = = Q h0 + I 2R e,cu + 2I 2
Rk,s +
Rk ,c 2
+ Rk,cu +
Rk ,c1
Ate coefficients for p-type and n-type are −187 μV/K and 187 μV/K, re-
A cu 2Ate (36) spectively. In addition, the boundary resistances can be obtained based
on diffuse mismatch model theory and W-F law as 9.2 × 10−8 K·m2/W
where Qc is the net cooling capacity at the cold surface of the micro-
(Rk,b,p), 3.5 × 10−7 K·m2/W (Rk,b,e), and 2.6 × 10−12 Ω·m2 (Re,b) [7].
TEC, and Q h is the net heat rejection at the hot surface of the micro-
The three-dimensional numerical model was performed under
TEC. Tc and Th are the temperatures of the cold and hot surfaces of the
adiabatic boundary condition. The radiation was disregarded and the
micro-TEC, respectively.Rk,s and Rk,c u are the thermal resistance per
temperature of the lower Cu leads, Tl , was held constant temperature at
unit area of the substrate and copper connector/lead, respectively.Rk,c1
300 K. Fig. 4 compares the obtained results against the steady state
is the thermal contact resistance between Cu and the TE element, Rk,c 2 is
experimental results in [19,30] and the numerical results in [30]. The
the thermal contact resistance between Cu and the substrate, R e,c u is the
external temperature difference ΔText is given by Tl−Tc0 , where Tc0 is the
electrical resistance of the copper connector/lead, R e,c1 is the electrical
average temperature of the cold surfaces of the TE elements. The ob-
contact resistance between Cu and the TE element. Acu is the cross area
tained results agree well with the experimental results, and the
of the copper connector/lead.

Fig. 3. Schematic of the 1N-1P micro-TEC (a) structure; (b) section view A-A [30].

98
D. Sun et al. Applied Energy 219 (2018) 93–104

Table 2 Table 3
Properties of the micro-TEC. Parameters of the unicouple microdevice.

Property Unit TE element Cu TE-solder Metal-solder Parameter Description Value

as μV/K 301 – – – Lte Length of TE elements 7 μm

k W/(m·K) 1.2 400 – – Wte Width of TE elements 7 μm
ρe Ω·m 1.08 × 10−5 1.7 × 10−8 – – Hte Thickness of TE elements 6 μm
Rk,c K·m2/W – – 6.25 × 10−7 4.16 × 10−7 Lco Length of connector 19 μm
Re,c Ω·m2 – – 7 × 10−11 3 × 10−12 Wco Width of connector 7 μm
as,b μV/K 187 – – – Hco Thickness of connector 2 μm
Rk,b,p K·m2/W 9.2 × 10−8 – – – Ls Length of substrate 29 μm
Rk,b,e K·m2/W 3.5 × 10−7 – – – Ws Width of substrate 17 μm
Re,b Ω·m2 2.6 × 10−12 – – – Hs Thickness of substrate 10 μm
P [4] W/(m3·K) 8.61 × 1013 – – – Ll Length of leads 9.5 μm
T K 300 Wl Width of leads 7 μm
Hl Thickness of leads 2 μm
Sp Space of TE elements 5 μm
deviations are within 5%, thereby confirming the validity of the model.
Moreover, the obtained results in this paper agree better with the ex-
perimental results than that of the modeling results in [30], which did βm (e.g., αs, k, ρe and ZT ) were considered as the intrinsic properties.
not consider the boundary resistances and boundary Seebeck effect. The effects of the boundary and size effects on the device-level
Therefore, the proposed method can significantly improve the perfor- Seebeck coefficient and thermal conductivity are shown in Fig. 5. It is
mance evaluation of micro-TECs. observed that the boundary and size effects weaken the Seebeck coef-
ficient and thermal conductivity of the TE materials at the device level,
particularly the former. A smaller thickness of the TE element corre-
3.2. Device-level thermophysical property sponds to a greater influence of the boundary effects. The device-level
Seebeck coefficient and thermal conductivity rapidly increase as the
As the thickness of thermoelectric elements considered in this study thickness increase at first and then gradually approach to the intrinsic
was very small, the thermophysical properties can be influenced by the values of the TE material (e.g., α s and k). For the boundary resistances
size and boundary effects. Thus, the three-dimensional numerical listed in Table 2 (e.g., Rk,b,p, Rk,b,e and Re,b), compared with the intrinsic
model of the TE element was built to study the influence of thickness values of the TE material, the decrease in device-level Seebeck coeffi-
and boundary resistances on the device-level thermophysical proper- cient and thermal conductivity varies in the ranges of [3.7%, 7%] and
ties, as shown in Fig. 2. The detailed properties and detailed geometry [1.8%, 3.5%], respectively, as the thickness of the TE element decreases
size are provided in Tables 2 and 3, respectively. from 10 μm to 5 μm. In comparison, the device-level Seebeck coefficient
To investigate the influence of boundary resistances and thickness and thermal conductivity respectively decrease from 3.7% to 0.7% and
of the TE element, a relative change ratio ψ , was introduced to reflect from 1.8% to 0.37%, when Hte increases from 10 μm to 50 μm. It in-
the variation between the device-level properties and intrinsic proper- dicates that the trend is more obvious for Hte smaller than 10 μm. The
ties of the TE material, as shown in Eq. (37), reason is that the boundary resistances are comparable to the intrinsic
resistances (e.g., Hte /(Ate k ) and ρe Hte / Ate ) ) of the TE elements and exert
|βD−βm |
ψ= a great adverse effect on the transport of carriers (e.g., phonons and
βm (37) electrons) when Hte is small. For larger Hte, the boundary resistances
and their adverse effects gradually become negligible. Thus, the device-
where βD represents the properties of the TE material at the device level
level Seebeck coefficient and thermal conductivity gradually approach
(e.g., α s,D , kD , ρe,D and ZD T ). Note that the properties of the TE material

Fig. 4. External temperature difference ΔText with respect

to the electrical current.

99
D. Sun et al. Applied Energy 219 (2018) 93–104

Fig. 5. Effects of boundary and size effects on the device-level Seebeck coefficient and thermal conductivity.

to the intrinsic values of the TE material as Hte becomes larger, as Fig. 6 further indicates that the thermal boundary resistance (Rk,b)
shown in Eqs. (21)-(26). In addition, the electrical boundary resistance exerts no influence on the electrical resistivity but weakens ZT. In
(Re,b) has no influence on the Seebeck coefficient and thermal con- comparison, the electrical boundary resistance (Re,b) not only increases
ductivity. Furthermore, the device-level Seebeck coefficient and the electrical resistivity, but also weakens ZT. This phenomenon be-
thermal conductivity significantly decrease with increasing Rk,b,e and comes more apparent when the thickness of the TE element becomes
Rk,b,p, when Hte is smaller than the critical value. For the case of Hte = 5 smaller. For the boundary resistances listed in Table 2 (e.g., Rk,b,p, Rk,b,e
μm, the Seebeck coefficient and thermal conductivity decrease in re- and Re,b), the increase in the device-level electrical resistivity changes
lative ranges of [4.2%, 15.9%] and [1.2%, 12%], respectively, when the from 9.7% to 4.8% when Hte increases from 5 μm to 10 μm. The de-
phonon thermal boundary resistance changes from 0.1 × Rk,b,p to crease in the device-level ZT varies from 18.1% to 5% when Hte in-
10 × Rk,b,p. And, when the electron thermal boundary resistance in- creases from 5 μm to 20 μm. It indicates that the device-level electrical
creases from 0.1 × Rk,b,e to 10 × Rk,b,e, the relative ranges are [1.3%, resistivity is significantly affected until Hte exceeds 10 μm, whereas the
11.9%] and [1.8%, 5.2%]. This finding suggests that the thermal device-level ZT is significantly affected until Hte exceeds 20 μm. This
boundary resistance of phonons, rather than that of electrons, is mainly reveals that size effect has a greater effect on ZT than that on the
responsible for the large decline in the Seebeck coefficient and thermal electrical resistivity. Moreover, when Hte = 5 μm, the decrease in ZT
conductivity. varies in the ranges of [15.7%, 27.1%] and [9.5%, 25.5%] when the

Fig. 6. Effects of boundary and size effects on the device-level electrical resistivity and ZT.

100
D. Sun et al. Applied Energy 219 (2018) 93–104

phonon thermal boundary resistance changes from 0.1 × Rk,b,p to the Peltier cooling effect of the micro-TEC. However, the Peltier cooling
10 × Rk,b,p and the electron thermal boundary resistance changes from effect is very weak at low currents. For higher currents, the Joule heat
0.1 × Rk,b,e to 10 × Rk,b,e, respectively. The decrease in ZT changes becomes significantly large to counter the Peltier cooling effect. Thus,
from 11% to 52.4% when the electrical boundary resistance changes the influence of the boundary effects is more obvious around the op-
from 0.1 × Re,b to 10 × Re,b. Comparison of the effects of the thermal timal current, indicating that the impact of the boundary effects on
boundary resistances (Rk,b,p, Rk,b,e) and the electrical boundary re- microcoolers can’t be ignored, particularly around the optimal current.
sistance (Re,b) on ZT reveals that Re,b has the greatest influence on ZT, For the case of I = 0.9 mA, the contact resistances reduce the tem-
followed by Rk,b,p. These results indicate that both boundary and size perature difference (ΔT = Th−Tc ) by 36.8%, and the boundary effects
effects should be considered in the design of microcoolers. further decrease it by 9.2%.

3.3. Temperature distribution of micro-TEC 4. Practical cooling performance of micro-TEC

A micro-TEC was built in COMSOL. The 8 × 8 TE elements shown in To ascertain the influence of contact resistances and boundary ef-
Fig. 2 were connected in series. The three-dimensional numerical model fects on the practical cooling performance of the microcooler, calcula-
was performed to investigate the effects of boundary effects and contact tions were carried out when the hot surface of the microcooler was set
resistances on the thermal transport in micro-TE device. The thickness constant temperature of 350 K, and typical heat flux (e.g., 100 W/cm2,
of the TE elements was set to 6 μm. A heat load of 2 mW was located on 200 W/cm2 or 300 W/cm2) were applied on the cold surface of the
the cold surface of the micro-TEC, and the heat convection coefficient microcooler.
between the hot surface and ambient was assigned value of 4252 W/ Fig. 9 shows the effects of contact resistances on the cooling per-
(K·m2)[30]. The thermal conductivity of the silicon substrate was set to formance at TE element thickness and heat flux respectively equaling to
163 W/(m·K)[4]. The electrical contact resistance on the Semi- 6 μm and 200 W/cm2. It can be observed that contact resistances con-
conductor–metal (SL–Cu) surface is assumed to be 1 × 10−11 Ω·m2, and siderably degrade the cooling performance of the cooler. And the
the thermal contact resistance at the Semiconductor–metal (SL–Cu and thermal contact resistance plays a more dominant role, compared with
Si–Cu) surfaces was set to 1 × 10−6 K·m2/W[5]. The other parameters the electrical contact resistance. For the case of I = 14 mA, the cooling
used are listed in Tables 2 and 3. temperature, Tc drops from 316.4 K to 280.2 K, when the thermal
The calculated results for three kinds of assumptions under various contact resisrance decreases from Rk,c to zero and the electrical contact
currents are present in Figs. 7 and 8. One of the assumptions was N-BR- reistance maintains at constant. As a contrast, Tc drops from 316.4 K to
N-CR, where neither the boundary effects nor the contact resistances 311.7 K, when the electrical contact resisrance decreases from Re,c to
were considered. Another assumption was N-BR-Y-CR, where only the zero and the thermal contact reistance maintains at constant. Fig. 10
contact resistances were considered, and the boundary effects were further shows the results under diffierent heat loads at Hte = 6 μm. It
ignored. The other assumption was Y-BR-Y-CR, where both the demonstrates that the boundary effects also considerably weaken the
boundary effects and the contact resistances were considered. Fig. 7 cooling performance of the micro-TE device. A higher heat flux corre-
illustrates that the sharp temperature jumping on the TE–Connector and sponds to a greater impact of the boundary effects. When I = 24 mA, Tc
Connector–Substrate surfaces for N-BR-Y-CR and Y-BR-Y-CR is caused of Y-BR-Y-CR increases by 5.1 K, 5.5 K, 6.1 K for heat flux of 100 W/
by the contact resistances. Comparison of N-BR-Y-CR and N-BR-N-CR cm2, 200 W/cm2 and 300 W/cm2, respectively, relative to that of N-BR-
reveals that the contact resistances significantly increase Tc0 , Th0 and Tc , Y-CR. Meanwhile, ΔT declines by a factor of 7.3%, 8.7%, and 17.1%,
whereas, Th is scarcely influenced. Figs. 7 and 8 show that the boundary respectively.
effects not only increase Tc0 and Tc , but also decline Th0 and Th . Tc un- Finally, the three-dimensional numerical model was performed to
dergoes more obvious changes when the input electrical current is predict the cooling performance and optimal working condition of the
around the optimal current for achieving the minimum Tc . The reason is micro-TEC under certain heat flux or thickness. Fig. 11 illustrates the
that the boundary effects decline the device-level ZT which is crucial to cold side temperature of the microdevice under various thicknesses and

Fig. 7. Temperature distribution across the microcooler.

101
D. Sun et al. Applied Energy 219 (2018) 93–104

Fig. 8. Temperature of the microcooler with respect to the current at Hte = 6 μm.

Fig. 9. Effects of contact resistances on the cooling performance.

currents for a heat flux of 300 W/cm2. It can be observed that Tc initially of 300 W/cm2 and 200 W/cm2, the minimum cold side temperatures of
decreases with increasing the TE element thickness and then increases 310.7 K and 287.3 K are predicted to be achieved at 11 μm/20 mA
with increasing the thickness under a certain current. The reason is that (Hte/ I ), 15 μm/16 mA, respectively.
the Peltier cooling effect initially increases with increasing Hte. When
Hte further increases, the cooling capacity is counteracted by the Joule 5. Conclusion
heat. To obtain a good cooling performance, the feasible scope is ex-
tracted to be with Hte from 7 μm to 15 μm and I from 16 mA to 24 mA. This paper developed an effective method for estimating the device-
Fig. 12 shows the optimal current and cooling performance. It can be level thermophysical properties and cooling performance of micro-TECs
observed that Ioptimal gradually decreases as the TE element thickness capturing interfacial and size effects. It found that both boundary and
increases. Ioptimal is the optimal current achieving the minimum cold size effects weaken the figure of merit of the thermoelectric material at
side temperature, Tcmin or the maximum temperature difference, ΔTmax the device level. And Re,b has the greatest effect on the figure of merit,
for a certain thickness of the TE element. The predicted optimal followed by Rk,b,p. This phenomenon is more obvious for smaller Hte,
structure and cooling capacity are also depicted in Fig. 12. For the cases which suggests that the thickness of thermoelectric element should be

102
D. Sun et al. Applied Energy 219 (2018) 93–104

Fig. 10. Cooling performance under diffierent heat loads and currents.

appropriately controlled in practical application. Moreover, the the minimum cold side temperature of the micro thermoelectric cooler
boundary effects substantially weaken the cooling performance of the is predicted to be 310.7 K, which can be achieved at 11 μm/20 mA
microdevice. And a higher heat flux corresponds to a greater impact of (Hte/ I ).
the boundary effects. The device-level figure of merit decreases by 5% This work provides a more appropriate method to evaluate the
to 18.1% when the thickness of thermoelectric element decreases from performance of micro thermoelectric coolers. It can provide guidance
20 μm to 5 μm. For the case of I = 0.9 mA, the contact resistances re- for the structure design and parameter optimization of micro thermo-
duce the temperature difference by 36.8%, and the boundary effects electric coolers for some specific cooling targets, such as chip cooling or
further decrease it by 9.2%. The cooling temperature increases by 6.1 K any other semiconductor devices that integrated with micro thermo-
due to the boundary effects when the heat flux is 300 W/cm2, while the electric coolers.
temperature difference decreases by 17.1%. At heat flux of 300 W/cm2,

Fig. 11. Temperature of the cold side with respect to the thickness and current.

103
D. Sun et al. Applied Energy 219 (2018) 93–104

Fig. 12. Optimal current and cooling performance.

Acknowledgement [15] Venkatasubramanian R, Siivola E, Colpitts T, et al. Thin-film thermoelectric devices

with high room-temperature figures of merit. Nature 2001;413(6856):597–602.
[16] Harman TC, Taylor PJ, Walsh MP, et al. Quantum dot superlattice thermoelectric
This work is jointly supported by the Natural Science Foundation of materials and devices. Science 2002;297(5590):2229–32.
China (Grant Nos. 51506060 and 51376068) and the Fundamental [17] Tian Z, Lee S, Chen G. Heat transfer in thermoelectric materials and devices. J Heat
Research Funds for the Central Universities (2016YXMS048). The Transfer 2013;135(6):061605.
[18] Goncalves LM, Rocha JG, Couto, et al. Fabrication of flexible thermoelectric mi-
supports are gratefully acknowledged. crocoolers using planar thin-film technologies. J Micromech Microeng
2007;17(7):S168–73.
References [19] Bulmana GE, Siivolab E, Shen B, et al. Large external ΔT and cooling power den-
sities in thin-film Bi2Te3-superlattice thermoelectric cooling devices. Appl Phys Lett
2006;89(12):122117.
[1] Shen LM, Chen HX, Xiao F, et al. The practical performance forecast and analysis of [20] Silva L, Kaviany M. Fabrication and measured performance of a first-generation
thermoelectric module from macro to micro. Energy Convers Manage microthermoelectric cooler. J Microelectromech Syst 2005;14(5):1110–7.
2015;100:23–9. [21] Owoyele O, Ferguson S, Connor BO. Performance analysis of a thermoelectric
[2] Cai Y, Liu D, Zhao FY, Tang JF. Performance analysis and assessment of thermo- cooler with a corrugated architecture. Appl Energy 2015;147:184–91.
electric micro cooler for electronic devices. Energy Convers Manage [22] Sungtaek JuY, Ghoshal U. Study of interface effects in thermoelectric micro-
2016;124:203–11. refrigerators. J Appl Phys 2000;88(7):4135–9.
[3] Dizaji HS, Jafarmadar S, Khalilarya S, et al. An exhaustive experimental study of a [23] Labounty C, Shakouri A, Bowers JE. Design and characterization of thin film mi-
novel air-water based thermoelectric cooling unit. Appl Energy 2016;181:357–66. crocoolers. J Appl Phys 2001;89(7):4059–64.
[4] Lee KH, Kim OJ. Analysis on the cooling performance of the thermoelectric micro- [24] Prabhakar Aparna, Murphy Michael C, Devireddy Ram V. Modeling of a bio-
cooler. Int J Heat Mass Transf 2007;50(9–10):1982–92. thermo-electric micro-cooler. Open Biotechnol J 2007;1:1–8.
[5] Chowdhury I, Prasher R, Lofgreen K, et al. On-chip cooling by superlattice-based [25] Ramírez MA, Castillo EE. Thermoelectric simulation using comsol multiphysics and
thin-film thermoelectrics. Nat Nanotechnol 2009;4(4):235–8. analysis of contact resistances effects. In: 13th LACCEI annual international con-
[6] Ibañez-Puy M, Bermejo-Busto J, Martín-Gómez C, et al. Thermoelectric cooling ference: “Engineering Education Facing the Grand Challenges, What Are We
heating unit performance under real conditions. Appl Energy 2017;200:303–14. Doing”; 2015.
[7] Da Silva LW, Kaviany M. Micro-thermoelectric cooler: interfacial effects on thermal [26] Chen WH, Liao CY, Hung CI. A numerical study on the performance of miniature
and electrical transport. Int J Heat Mass Transf 2004;47(10–11):2417–35. thermoelectric cooler affected by Thomson effect. Appl Energy 2012;89:464–73.
[8] Zheng XF, Liu CX, Yan YY, et al. A review of thermoelectrics research – recent [27] Bartkowiak M, Mahan GD. Chapter 8 Heat and electricity transport through inter-
developments and potentials for sustainable and renewable energy applications. faces. In: Tritt Terry M, editor. Semiconductors and Semimetals; 2001;
Renew Sustain Energy Rev 2014;32:486–503. 70(01):245–71.
[9] He W, Zhang G, Zhang X, et al. Recent development and application of thermo- [28] Gurevich YG, Mashkevich OL. The electron-phonon drag and transport phenomena
electric generator and cooler. Appl Energy 2015;143:1–25. in semiconductors. Phys Rep 1989;181(6):327–94.
[10] Chen G, Dresselhaus MS, Dresselhaus G, et al. Recent developments in thermo- [29] Gurevich YG, Logvinov GN. Thermo-EMF and thermoelectric current in unipolar
electric materials. Int Mater Rev 2013;48(1):45–66. semiconductors with finite dimensions. Soviet Phys Semicond 1992;26(11):1091–4.
[11] Alvarez-Quintana J. Impact of the substrate on the efficiency of thin film thermo- [30] Choday SH, Lundstrom MS, Roy K. Prospects of thin-film thermoelectric devices for
electric technology. Appl Therm Eng 2015;84:206–10. hot-spot cooling and on-chip energy harvesting. IEEE Trans Compon Packag Manuf
[12] Rowe DM. Thermoelectrics handbook: macro to nano. CRC/Taylor & Francis; 2005. Technol 2013;3(12):2059–67.
[13] Yang R, Chen G. Nanostructured thermoelectric materials: from superlattices to [31] Gayner C, Kar KK. Recent advances in thermoelectric materials. Prog Mater Sci
nanocomposites. Mater Integr 2005;18:31–6. 2016;83:330–82.
[14] Böttner H, Chen G, Venkatasubramanian R. Aspects of thin-film superlattice ther-
moelectric materials. Devices Applicat 2006;31(3):211–7.

104