Applied Thermal Engineering 130 (2018) 408–417

Contents lists available at ScienceDirect

Applied Thermal Engineering

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

Research Paper

Development of a numerical method for the performance analysis of

thermoelectric generators with thermal and electric contact resistance
Chang Nyung Kim
College of Engineering, Kyung Hee University, Yongin 446-701, Republic of Korea

h i g h l i g h t s

A numerical modeling for the performance evaluation of a TEG is newly devised.

It is based on the conservation principle of the thermal energy and electric charge.

The performances of TEGs with thermal and electric contact resistance are analyzed.

The validation of the present modeling is performed against a mathematic result.

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

Article history: Lately, the development of thermoelectric (TE) materials with higher ZT values allows the fabrication of
Received 23 February 2017 thermoelectric devices with higher performance. Though several analytic analyses of the effect of contact
Revised 3 July 2017 resistance on the thermoelectricity have been reported, the numerical prediction of a thermo-electric
Accepted 29 October 2017
generator (TEG) with finite thermal and electric contact resistance has been rarely studied. In the present
Available online 31 October 2017
study, based on the conservation principle of the thermal energy and electric charge, newly devised is a
numerical modeling for the performance evaluation of a TEG with thermal and electric contact resistance
Keywords:
and without any simplification (or neglect) of Joule heating, Thomson heating and Peltier heating, with
Thermoelectric generator
Computational analysis
the use of temperature-dependent thermoelectric properties, which allows accurate solutions, and built
Performance is a computer program for the performance simulation. Here, investigated is the effect of pellet length on
Resistance the TEG performance in the presence of thermal and electric contact resistances. The present numerical
Efficiency modeling and the computer program adopting a new numerical solution method is validated against a
Effectiveness mathematic result. Also, the obtained performance of the TEG is compared with that acquired by an exist-
ing analytic model. In the present study, as an evaluation factor of a TEG performance, the effectiveness
(meaning the ratio of the performance of a TEG with resistances to that without any resistance) is newly
defined, and is examined for different values of thermal and electric resistance. The results show that
when the length of the TE materials is smaller, the effect of the contact resistances is notable, which
decreases the power output (and the effectiveness) of the TEG devise.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction simulation for the performance of thermoelectric generators in

Lately, there has been intensive interest in the development of a
thermoelectric generator (TEG) that uses waste heat available. Tar-
tarinov et al. [1] performed a modeling study for a TEG using the
waste heat of a car. Also, Reddy et al. [2] modeled and simulated
an automotive thermoelectric generator in a MATLAB/Simulink
environment.
Lazard [3] solved the heat transfer equation in a thermoelectric
leg by analytic and numerical method, where software FlexPde is
used for the simulation. Chen et al. [4] performed a numerical
E-mail address: cnkim@khu.ac.kr
fluid power systems with the use of FLUENT software. Yan et al.
[5] performed a time-dependent analysis of thermoelectric devices
with one-dimensional finite-volume model and three-dimensional
simulation with COMSOL software.
Picard et al. [6] numerically simulated the thermomechanical
behaviors of thermoelectric modules with segmented Bismuth-
Telluride-based legs with the use of ANSYS Workbench 13. Oki
[7] performed a numerical optimization of trapezoidal thermoelec-
tric elements for a double-pipe-shaped module, using FLUENT.
Hogblom and Andersson [8] analyzed the performance of thermo-
electric generators numerically and experimentally, where a finite
element method (with ANSYS) is employed.

https://doi.org/10.1016/j.applthermaleng.2017.10.158
1359-4311/Ó 2017 Elsevier Ltd. All rights reserved.
 C.N. Kim / Applied Thermal Engineering 130 (2018) 408–417
Several analytic studies about the effect of the contact resistance
on the performance of a TE device were performed. Ebling et al. [9]
considered the effect of contact resistance of thermoelectric gener-
ators, and obtained an equation for the ZT value of module,
ZTmodule = ZTmaterial/(1 + RC/Rlegs), where RC is the contact resistance
and Rlegs means the resistance of the legs. However, the above equa-
tion is valid only when an = ap , kn = kp andqn = qp , which are temper-
ature independent, where a, k and q are Seebeck coefficient,
thermal conductivity and specific electric resistivity, respectively,
and only when An = Ap, where A denotes the cross-section area of
a leg. Since the properties of TE materials are usually temperature
dependent, the above analysis may include some inaccuracy.
Reddy et al. [10] investigated the influence of contact resistance
on the performance of different thermoelectric devices. However,
the authors did not consider the Peltier heating term in their gov-
erning equation (Refer to Eq. (1) of Ref. [10]), while keeping Thom-
! presence of the contact resistances is examined numerically in
son heating (Tð J rTÞda=dT). Usually, the amount of Peltier
heating is quite larger than that of Thomson heating unless the
Seebeck coefficient is a very strong function of temperature in a
local region of a leg [11]. Also, the above authors assumed that s
defined by T(d a= dT) is uniform in a leg (Refer to Eq. (2) of Ref.
[10], see also Ref. [12]), which is far from physical appropriateness.
Bjork [13] established an analytic model for the influence of
contact resistance on thermoelectric efficiency. In his model, in
consideration of the heat flow passing through the interface of a
leg and an electrode, by adding the numerators and denominators,
the temperatures on the both sides of the interface were obtained
(See Eq. (2) of Ref. [13]). However, the author neglected the fact
that Peltier heating/cooling is observed at the interface. Since the
temperature of the both sides of the interface obtained in this
study is only valid when there is no heat generation/absorption
at the interface, the above analysis is regarded to have some inac-
curacy in recognition that the amount of Peltier heating/cooling
observed at an interface is considerable.
Since (i) the governing equation for the thermoelectricity is
nonlinear in terms !
of temperature and current density, (ii) Thom-
son heating (Tð J rTÞda=dT) is notable in the region of a pellet
where the spatial change in the Seebeck coefficient of a TE material
is notable with significant d a /dT so that Thomson ! heating is not
uniform usually in a leg, and (iii) Peltier heating (T J raj interface)
is a phenomenon observed only at the interface of an electrode and
a pellet, where the contact resistance is seen, it seems that an ana-
lytic solution has difficulty in predicting accurate solutions for the
effect of contact resistances on the performance of a TE device.
In the present study, newly devised and validated is a numerical
modeling, based on the conservation principle of the thermal
energy and electric charge, which can predict the performance of
a TEG with thermal and electric contact resistance, with a rigorous
physical model and without any simplification. More specifically,
Fig. 1. Geometry of a thermoelectric module (dimensions in ‘‘mm”).
409
the present numerical study solves the governing partial differen-
tial equations which are non-linear in terms of temperature and
electric current density, and takes into account Joule heating,
Thomson heating and Peltier heating without any simplification.
Also, the properties of the TEG materials in this study depend on
the temperature, which allows an accurate analysis of the thermo-
electricity. With the effectiveness newly defined in this study as
the ratio of the performance of a TEG with resistances to that with-
out resistance, the performance characteristics of the thermoelec-
tric module with different sizes of thermal and electric contact
resistance are examined, and are compared with those obtained
from a simple analysis.
Also, with recognition that the thermal and electrical contact
resistance can be important factors influencing the performance
of TE devices, especially in a case where the length of a pellet is
short, the effect of pellet length on the TEG performance in the
detail, and the performance of a TEG with the contact resistances
is compared with that without the contact resistances. Here, the
results of the above analyses can suggest a guideline for the size
of allowable thermal and electric contact resistance of TE devices.
2. Problem formulation and solution method
2.1. Problem formulation
A TEG module has an n-leg and a p-leg with one top and two
bottom electrodes, as shown in Fig. 1, where the dimension of
the TEG module is given, and the thickness is 2 mm. The governing
equations for steady-state thermoelectricity can be given in the
below:
Energy equation:
! !
r  ðkrTÞ þ J 2 =r  T J ra ¼ 0; ð1Þ
where ra ¼ ddTa rT þ rajinterface
Ohm’s law:
!
J ¼ rðr£  arTÞ ð2Þ
Conservation of charge:
!
r J ¼0 ð3Þ
where a, r and k means the Seebeck coefficient, electric conductiv-
ity and thermal conductivity, respectively.
The thermoelectric properties of TE materials in the p- and n- !
leg are temperature dependent. Therefore, the current density J
depends on the temperature, leading to the fact that Eq. (1) is non-
linear. In this study, high quality TEG material Bi0.5Sb1.5Te3 [14]
whose ZT value reaches around 1.8 at T = 330 K is employed. The
410 C.N. Kim / Applied Thermal Engineering 130 (2018) 408–417

Seebeck coefficient a, electric conductivity r; and thermal conduc-
tivity k of the above TEG material for the p- and n-leg are given in
the below (see also Fig. 2 for the temperature-dependent
properties);


ap ¼ 525:98585 þ 4:97271T  0:00958T 2 þ 4:8283  106 T 3 =106 V=K
ð4Þ


an ¼  417:48349 þ 3:94411T  0:00689T 2 þ 2:56705  106 T 3 =106 V=K
ð5Þ


rp ¼ 4145:9581  20:476T þ 0:03664T 2  2:14407  105 T 3 100 S=m
ð6Þ


rn ¼ 5655:47725  28:96099T þ 0:05496T 2  3:60488  105 T 3 100 S=m
ð7Þ

kp ¼ 5:03763  0:03172T þ 7:23229  105 T 2  4:95402

 108 T 3 W=mK ð8Þ

kn ¼ 14:11288  0:09569T þ 2:29075  104 T 2  1:72292

 107 T 3 W=mK ð9Þ
The electrode made of copper has the following properties

r ¼ 5:6  107 S=m ð10Þ

k ¼ 380 W=mK ð11Þ

As for the boundary conditions, the temperatures of 400 K and
300 K are given on the upper side of the top electrode and on the
lower side of the bottom electrodes, respectively, and the adiabatic
(a) electric conductivity condition is given on the other surfaces. Typical values of the See-
beck coefficient for the n-leg and p-leg for the temperature range
of 300 K–400 K are 225.7  106 V/K and 244.4  106 V/K, respec-
tively. With the temperature difference 100 K, in the n-leg the
maximum voltage difference D£ ¼ an DT ¼ (225.7  106 V/K)
(100 K) can be created, and in the p-leg D£ ¼ ap DT ¼ (244.4 
106 V/K) (100 K) can be considered. This means that the maxi-
mum voltage difference for the TEG module to operate is 4.70 
102 V. For the optimal operation condition where the internal
electric resistance of the TEG device is the same as the external
electric resistance, the voltage difference across the module is to
be 2.35  102 V. Therefore, the electric voltages of 0.0 V and
0.0235 V are applied on the left side of the bottom left electrode
and on the right side of the bottom right electrode, respectively.
And, on the other surfaces @£=@n = 0 is given, which means that
the current does not flow in the outward direction.

2.2Numerical method
(b) Seebeck coefficient
The substitution of Eq. (2) into Eq. (3) gives the following elec-
tric potential equation:

r  ðrr£Þ þ r  ðrarTÞ ¼ 0 ð12Þ

Therefore, two simultaneous partial differential equations (that

is, Eqs. (1) and (12)) are formulated that can be solved for the tem-
perature and electric potential, respectively.
Here, Eqs. (1) and (12) can be transformed, for the use of the
finite difference method, into the following equations,
respectively:
! !
kr2 T þ ðrkÞ  ðrTÞ þ J 2 =r  T J ðraÞ ¼ 0 ð13Þ

rr2 £ þ ðrrÞ  ðr£Þ þ ðrrÞ  ðarTÞ þ ðraÞ  ðrrTÞ þ rar2 T ¼ 0;

ð14Þ

(c) thermal conductivity where the thermoelectric properties of TE materials in the p- and n-
leg are temperature-dependent. Therefore, Eqs. (1), (12)--(14) are
Fig. 2. Properties of the pellets. nonlinear.
 C.N. Kim / Applied Thermal Engineering 130 (2018) 408–417 411

Since the use of the finite difference method based on Eqs. (13)
and (14) cannot guarantee the conservation principle for thermal
energy and electric charge, the current study adopts the finite vol-
ume method for the discretization of Eqs. (1) and (12), which
allows the conservation principle for the thermal energy and elec-
tric charge, even in the case where the properties are strongly
temperature-dependent.
The integration of Eqs. (1) and (12) for a finite volume (dx)(dy)
(dz) yields the following equations, with the notation given in
Fig. 3:
Energy equation
Z xþDx Z yþDy Z zþDz  ! !

r  ðkrTÞ þ J 2 =r  T J ra dxdydz ¼ 0
x y z

ð15Þ
For a two-dimensional formulation the energy equation can be
discretized as (here, Dz means the thickness of the TE G moduleÞ; Fig. 3. Information for the discretization based on finite volume method.
Z xþDx Z  yþDy    
@ @T @ @T
0¼ k þ k
x y @x @x @y @y In association with the information given in Fig. 3, the points P, N, S,
 
! @ a @ a E and W are the main points where the values of the temperature
þ J 2 =r  T J x þ Jy dxdyDz ð16Þ and electric potential are stored. Meanwhile, the points n, s, e and
@x @y
w are the face points where the values of the flux (that is, the heat
     
@T @T flux and electric current) are considered. The values of ke ,kw ,kn ,ks in
0¼ ke  kw DyDz Eq. (18) and of re ,rw ,rn ,rs ,ae ,aw ,an ; as in Eq. (21) should be updated
@x e @x w
     during an iterative solution procedure, with the use of the spatial
@T @T !
þ kn  ks DxDz þ ð J 2 =rÞP DxDyDz interpolation, since the properties are position-dependent (because
@y n @y s the properties are temperature-dependent). Since the information !
 T P fðJ x Þp ðae  aw ÞDyDz þ ðJ y Þp ðan  as ÞDxDzg ð17Þ for at the point P can be easily obtained from the Eq. (2), J 2 =r for
the point P can be evaluated. Therefore, the temperature and elec-
 
TE  TP TP  TW tric potential can be obtained from Eqs. (18) and (21), respectively,
0 ¼ ke  kw Dy Dz with an iterative solution method.
Dx Dx
  Since Eqs. (1) and (12) are 2nd order elliptic partial differential
TN  TP TP  Ts !
þ kn  ks DxDz þ ð J 2 =rÞP DxDyDz equations with a generic form for the temperature and electric
Dy Dy
n o potential, respectively, the equations can be solved accurately with
 T P ðJ x Þp ðae  aw ÞDyDz þ ðJ y Þp ðan  as ÞDxDz ð18Þ the boundary condition of the temperature (for Eq. (18)) and the
electric potential (for Eq. (21)), and can be treated without any
Electric potential equation non-physical approximation (see the approximation given in Eqs.
Z xþDx Z yþDy Z zþDz (3) and (6) of Ref. [15], for example). The above numerical method
0¼ fr  ðrr£Þ þ r  ðrarTÞgdxdydz ð19Þ conserves the thermal energy and electric charge in the calculation
x y z
domain with the use of finite volume method, as mentioned before.
For a two-dimensional formulation, the electric potential equa- Now, Eqs. (18) and (21) are solved with an algorithm such as
tion can be discretized as Point SOR (Successive Over-Relaxation), line-SOR and so on. Many
 
     grid systems ranging from 2000 to 6000 cells are tested in consid-
Z xþDx Z yþDy
@ @£ @ @£ @ @T eration of grid independency, and the results shown in Section 4
0¼ r þ r þ ra
x y @x @x @y @y @x @x are based on the grid system of 5200 cells.
       
@ @T @£ @£
þ ra dxdyðDzÞ ¼ re  rw DyDz
@y @y @x e @x w
     3. Validation
@£ @£
þ rn  rs DxDz
@y n @y s The validation of the current numerical modeling and the com-
     
@T @T puter program against a mathematical analysis of the TEG perfor-
þ re a e  rw aw DyDz
@x e @x w mance [16] is to be performed for a TEG device with uniform
     properties for each of n- and p- leg and for electrodes, which
@T @T
þ rn a n  rs as DxDz ð20Þ may allow the comparison for the TEG performances. As for the
@y n @y s
numerical validation of the benchmarking problem, the tempera-
  tures of 400 K and 300 K are given on the upper side of the top
£E  £P £  £W
¼ re  rw P DyDz electrode and on the lower side of the bottom electrodes, respec-
Dx Dx
  tively, and the adiabatic thermal condition is given on the other
£  £P £  £S
þ rn N  rs P DxDz surfaces. For the temperature range of 300 K–400 K the average
Dy Dy values of the electric conductivity, Seebeck coefficient, and thermal
 
TE  TP TP  TW conductivity for the n-leg/p-leg are 7.231  104 /5.642  104 S/m, -
þ re ae  rw aw DyDz
Dx Dx 225.7  106 /244.4  106 V/K, 1.342/0.691 W/mK, respectively
  (see Eqs. (4)--(9) and Fig. 2).
TN  TP TP  TS
þ rn an  rs as DxDz ð21Þ With the temperature difference 100 K, in the n-leg the maxi-
Dy Dy
mum voltage difference D£ ¼ an DT ¼ (225.7  106 V/K) (100 K)
412 C.N. Kim / Applied Thermal Engineering 130 (2018) 408–417

can be created, and in the p-leg D£ ¼ ap DT ¼ (244.4  106 V/K) a2pn DT 2 2

(100 K) can be considered for an open circuit. This means that
the maximum voltage difference for the TEG module to operate
is 4.70  102 V. For the optimal operation condition where the
internal electric resistance of the TEG device is the same as the
external electric resistance, the voltage difference across the mod-
ule is given to be 2.35  102 V. Therefore, the electric voltages of
0.0 V and 0.0235 V are applied on the left side of the bottom left
electrode and on the right side of the bottom right electrode,
respectively, and on the other surfaces @£=@n = 0.0 is given, where
n means the outward direction of the surface.
In the validation procedure, the electrode made of copper has
the following properties
r ¼ 5:6  107 S=m
k ¼ 380 W=mK
Since the electric and thermal conductivity of the TE materials are
quite smaller than those of the copper, finer grids are needed in
the TE materials. Here, two different mesh systems with 5,200
meshes (for each of n- and p-leg 20  80 meshes), called Grid A,
and 6,000 meshes (for each of n- and p-leg 20  100 meshes),
referred to Grid B, are used. Since, in the TE materials, the values
of the variables are changed much in the vertical direction (see
Fig. 1) rather than in the horizontal direction, it is recommended
to have finer grids in the vertical direction.
The numerical evaluations of the performance for the given sit-
uation of the optimal operation condition yield the followings
characteristics;
PTEG ¼ IðDVÞ ¼ 0:03497 W for Grid A
ð470  106 V=KÞ ð100 KÞ2
PTEG;max ¼ ¼ ¼ 0:03502 W
4RTEG 4ð0:01578 XÞ
ð31Þ
The comparison of the current numerical evaluations (Eqs. (24)
and (25)) and the mathematic expression (Eq. (31) in Ref. [16]) are
in excellent agreement to each other. Therefore, the validation of
the numerical modeling and the computer program is obtained.
4. Results and discussion
The thermal characteristics for the TEG module without any
resistance are shown in Fig. 4. Here, the spatial change in temper-
ð22Þ ature is clearly seen in the pellets, while the temperature is almost
uniform in each electrode since the electrode made of copper has
ð23Þ much higher thermal conductivity (Fig. 4a). Also, the heat flux
flowing downward is depicted in Fig. 4b.
Fig. 5 shows the electric features of the module. A notable
change in electrical potential is observed in the pellets, while the
electric potential is almost uniform in each electrode because of
much higher electrical conductivity therein (Fig. 5(a)). Here, the
current flows generally to the right (Fig. 5(b)).
The characteristics of Joule heating is presented in Fig. 6. Here,
as the electric conductivity of the p-leg (positioned on the right
side) is generally smaller than that of the n-leg (see Fig. 2(a)), the
size of Joule heating is larger in the p-leg. Especially, since the elec-
tric conductivity is smaller for higher temperature, the size of Joule
heating in the top region of the p-leg is larger. It can be noted that
the size of the current intensity is almost uniform in the n- and
ð24Þ

PTEG ¼ IðDVÞ ¼ 0:03529 W for Grid B ð25Þ

A mathematical analysis of the TEG output can be denoted in
the following way [16];

2
V TEG;oc  RTEGRLþRL a2pn DT 2 RL
PTEG ¼ ¼ ; ð26Þ
RL ðRTEG þ RL Þ2
where VTEG,oc is the maximum voltage available under an open cir-
cuit condition, RL is the external electric resistance, RTEG is the inter-
nal electric resistance of TEG device, and apn is the summation of the
absolute value of Seebeck coefficients of the p and n semiconduc-
tors. For the optimal operation condition where the internal electric
resistance of the TEG device is the same as the external electric
resistance (that is, RL, = RTEG) the power output of the TEG devise
can be expressed in Ref. [16] in the following expression;
(a) Temperature
a2pn DT 2
PTEG;max ¼ ð27Þ
4RTEG

Here, with the following evaluations of

apn ¼ 470  106 V=K ð28Þ

DT ¼ 100 K; ð29Þ

1 Ln 1 Lp 1 1 L
RTEG ¼ Rn þ Rp ¼ þ ¼ þ
r n An r p Ap rn rp A
ðwhere An ¼ Ap ¼ A; Ln ¼ Lp ¼ LÞ
 
1 1 2  103 m
¼ þ ¼ 0:01578 X
72310 S=m 56420 S=m 4  106 m2
ð30Þ (b) Heat flux
The evaluation of Eq. (27) gives the flowing expression; Fig. 4. Thermal features.
 C.N. Kim / Applied Thermal Engineering 130 (2018) 408–417 413

Fig. 7. Thomson heating.

(a) Electric potential
present problem, it is to be reminded that the Seebeck coefficients
for n- and p-leg shown in Fig. 2(b) are increasing with temperature
in the range 300 K < T < 360 K, meaning da=dT is positive, and are
decreasing with temperature in the range 360 K < T < 400 K, imply-
ing da=dT is negative. The Thomson term can be written as follows:
! !
T J ra ¼ T J fðda=dTÞðdT=dyÞjg ð32Þ

with an assumption that dT=dy  dT=dx, where x and y are the

coordinates introduced in Fig. 1, and j is the unit vector in y-
direction. In the
!
p-leg, the current moves
!
downward so that the cur-!
rent intensity J can be denoted as j J jj with an assumption that J
moves vertically in the pellet.
In a low temperature range (that is, 300 K < T < 360 K) in the p-
leg, da=dT is positive (see Fig. 2(b)) and the value of dT=dy is pos-
itive. Therefore, Eq. (32) can be written as
(b) Electric current ! !
T J fðda=dTÞðdT=dyÞjg ¼ Tðj J jjÞ  fðda=dTÞðdT=dyÞjg
Fig. 5. Electric features.
!
¼ Tj J jðda=dTÞðdT=dyÞ ð33Þ
!
Since the terms of T, j J j, da=dT, dT=dy are all positive, the
Thomson term in the lower part of p-leg is positive, implying the
heat generation. Meanwhile, since in a high temperature range
(that is, 360 K < T < 400 K) in the p-leg da=dT is negative (see
Fig. 2b), the Thomson term in the upper part of the p-leg is nega-
tive, meaning the heat absorption. Also, the Thomson term in the
n-leg can be examined in a similar way.
The Peltier heating in the TEG module is given in Fig. 8. At the
interfaces between the pellets and the bottom electrode the heat
generation is observed, and at the interfaces between the pellets

Fig. 6. Joule heating.

p-leg in consideration of the continuity of the electric charge. The

highest value of Joule heating in Fig. 6 is around 2.5  106 W/m3 in
the top region of the p-leg. The Joule heating in the n-leg can be
analyzed in a similar way.
The features of Thomson heating is illustrated in Fig. 7, where
heat generations are observed in the bottom region of the n- and
p-legs, while heat absorption is shown in the top region of the n-
and p-legs. The sizes of the heat generation and absorption associ-
ated with Thomson term are around 3.25  106 W/m3 and 2.75 
106 W/m3, respectively. For the analysis of Thomson heating of the Fig. 8. Peltier heating.
414 C.N. Kim / Applied Thermal Engineering 130 (2018) 408–417

and the top electrode

!
the heat absorption is seen. Peltier heating is a computational cell adjacent to the interface, k is the thermal con-
expressed by T J ra. As a position is moved in the y-direction ductivity of the electrode, A is the cross-sectional area of the heat
across the lower interface in the right part of the TEG module, flow, and h is the thermal conductance at the interface. Then, in a
the Seebeck coefficient increases suddenly from zero to aP . There- steady state the following expression can be held:
fore, the term ra can be written as ðaP =DtÞj, where Dt is an
Ta  Tb
infinitesimal vertical distance between the pellet and electrode Q ab ¼ ð36Þ
Dy=kA
(that is, the gap of the interface). Therefore, the Peltier term at
the lower interface in the right part of the TEG module can be writ-
Tb  Tc
ten as Q bc ¼ ; ð37Þ
1=hA
! ! !
T J ra ¼ Tðj J jjÞ  fðaP =DtÞjg ¼ Tj J jðaP =DtÞ ð34Þ where b is the point on the bottom side of the top electrode, and c is
!
the point on the top side of the p-leg.
Since the terms T, j J j and ðaP =DtÞ are all positive, the Peltier Since Q ab and Q bc are the same,
term at the lower interface in the right part of the TEG module
Ta  Tc
means heat generation. Meanwhile, as a position is moved in the Q ¼ Q ab ¼ Q bc ¼ ð38Þ
Dy=kA þ 1=hA
y-direction across the upper interface in the right part of the TEG
module, the Seebeck coefficient decreases suddenly from aP to Now, the expression for the heat transfer through the two ther-
zero. Therefore, the term ra can be expressed as ðaP =DtÞj, yield- mal resistances in Eq. (38) can be expressed in terms of equivalent
ing the Peltier term to be thermal conductivity kequi of the computational cell adjacent to the
! ! !
interface shown in Fig. 9 in the following way:
T J ra ¼ Tðj J jjÞ  fðaP =DtÞjg ¼ Tj J jðaP =DtÞ ð35Þ Ta  Tc Ta  Tc
¼ ð39Þ
The above expression means a negative value, implying the heat Dy=kA þ 1=hA Dy=kequi A
absorption. The Peltier term in the left part of the TEG module can
be investigated in a similar way.
The absolute values of the Peltier heating/cooling obtained in
the above seem to go to infinity, since Dtis very small. However,
the volume for the Peltier heating/cooling is infinitesimal since
the gap of the interface is very small. Therefore, the total Peltier
heating/cooling in a TEG device observed at the lower interface
of the right ! part of the TEG module will be
R xþDx R yþDy R zþDz
T J ra dxdydz ¼ TIaP , which is finite, since
RxxþDx RyzþDz !z
x z
j J jdxdz ¼ I (the total current). Therefore, the Peltier
heating/cooling is of the type of Dirac delta function. Also, the heat
generation/absorption at the interfaces in the left part of the TEG
module can be interpreted in a similar way.

4.1. Effect of the contact resistances

4.1.1. With thermal contact resistance and without electrical contact

resistance
In this study, the thermal contact resistance observed at the
interface between the electrode and pellet is taken into account
in the following manner. For the vertical downward heat flow
passing through the computational cell (adjacent to the interface) Fig. 10. Effect of thermal contact conductance on the performance
in the electrode and through the interface, the thermal conduction ( : from the numerical simulation, : from
resistance Dy=kA and the thermal interface resistance 1=hA can be the simple relationship of Eq. (32)).
considered as shown in Fig. 9, where Dy is the vertical thickness of

Fig. 9. Thermal contact conductance at the interface.

 C.N. Kim / Applied Thermal Engineering 130 (2018) 408–417
This means that the summation of the thermal contact resis-
tance observed at the interface and the thermal conductive resis-
tance for the single computational cell adjacent to the interface
can be expressed in terms of the equivalent thermal conductive
resistance of the computational cell adjacent to the interface.
In this study, the effectiveness is defined as the ratio of the TEG
power output with any resistance to that without resistance as
follows:.
TEG power output with contact resistance
Eff ¼
TEG power output without contact resistance
Therefore, the effectiveness indicates how much portion of the
potential performance is really achieved in the presence of any
resistance, while the efficiency is defined as the ratio of the TEG
power output to the heat transfer through the TEG device.
In Fig. 10, the effectiveness of the TEG module with respect to
the thermal contact conductance, in situations without electric
contact resistance, is numerically obtained (it will be denoted by
Eff ;nu ; where nu means ‘‘numerical”), and is compared with the
effectiveness expressed by the following simple relation (it will
be referred to Eff ;S;T , where S means ‘‘simple relation” and T means
‘‘temperature”)
L=ðKAÞ K=L
Eff ;S;T ¼ ¼ ;
L=ðKAÞ þ 2  Rc;t K=L þ 2=ht
where Rc;t ¼ ð1=ht AÞ is the thermal contact resistance at an
interface.
Generally, the effectiveness obtained by the current numerical
analysis Eff ;nu is lower than Eff ;S;T . Since Eq. (41) uses the
temperature-independent properties, and does not consider the
fact that any change in thermal situation has an influence on the
Table 1
TEG performance with thermal and electric contact resistance (Q_ ðWÞ is the thermal flow, g is the efficiency and Eff is the effectiveness).
415
electric current (this fact is shown in Eq. (1)), Eff ;S;T expressed by
Eq. (41) may include considerable error. In Fig. 10, as the thermal
contact conductance is reduced, the effectiveness is decreased. In
Case 6 given in Table 1 the effectiveness Eff ;nu is fairly high with a
value of around 0.9432 with the thermal contact conductance
105 W/m2 K, while it is decreased to 0.7118 with the thermal con-
tact conductance 104 W/m2 K (see also Fig. 10). The decrease in the
effectiveness would be more notable in a region where the thermal
contact conductance is less than 1  104 W/m2 K. Based on Fig. 10,
ð40Þ a reasonably allowed value of the thermal contact conductance can
be estimated to be greater than 2  104 W/m2 K (where the effec-
tiveness Eff ;nu is 0.8317), which is to be taken into account in the
manufacturing process of TEG devices. It can be reminded that,
as the size of the TEG pellet decreases, leading to the decrease in
the internal thermal resistance of the TEG module, the relative
effect of the thermal contact resistance on the performance would
increase.
However, it is to be noted that, since the contact resistances are
caused by the imperfect contact of the electrode and pellet, a situ-
ation with considerable thermal contact resistance and without
electric contact resistance may not be practical.
4.1.2. With electrical contact resistance and without thermal contact
ð41Þ resistance
How to treat the electric contact resistance can be explained in a
way very similar to how to deal the thermal contact resistance (or
thermal contact conductance). Here, the summation of the electri-
cal contact resistance observed at the interface and the electrical
resistance for the single computational cell adjacent to the interface
can be treated in the computer program with the equivalent electric
conductivity requi of the computational cell adjacent to the interface.
416 C.N. Kim / Applied Thermal Engineering 130 (2018) 408–417
Fig. 11. Effect of electric contact resistivity on the performance
( from the numerical simulation,
the simple relationship of Eq. (33)).
In Fig. 11 the effectiveness of the TEG module with the electric
contact resistivity, in situations without thermal contact resis-
tance, is numerically obtained and compared with the effective-
ness expressed by the following simple relation (it will be
denoted by Eff ;S;E , where S means ‘‘simple relation” and E meanse
‘‘lectricity”):
qL=A qL
Eff ;S;E ¼ ¼ ;
qL=A þ 2  Rc;e qL þ 2=he
where Rc;e ¼ 1=ðhe AÞ is the electric contact resistance.
Generally, the effectiveness obtained by the current numerical
analysis is lower than Eff ;S;E . Since the explanation about the limita-
tion of the use of Eff ;S;T given in the previous section is still valid in
this situation, Eff ;S;E given in Eq. (42) may include some inaccuracy.
Here, as the electric contact resistivity increases, the effectiveness
decreases. In Case 2 given in Table 1, with the electric contact resis-
tivity of 1010 Xm2 the effectiveness Eff ;nu is fairly high with a value
of around 0.9610, while with the electric contact resistivity 108
Xm2 it is decreased to 0.5816 (see also Fig. 11). The decrease in
the effectiveness would be steeper in the region where the electric
Fig. 12. The power output, heat flow, efficiency and effectiveness (
the case with thermal and electric contact resistance, while
resistance).
contact resistivity is greater than 10-8 Xm2. Based on Fig. 11, a rea-
sonably allowed value of the electric contact resistivity would be
smaller than 3.0  109 Xm2 (where the value of the effectiveness
Eff ;nu is 0.8088), which is to be considered in the manufacturing
process of TEG devices.
Also, it is to be reminded that a situation with considerable
electrical contact resistance and without thermal contact resis-
tance may not be practical, as explained in the previous section.
4.1.3. Both with thermal and electrical contact resistance
In this section the performance of the TEG both with thermal
and electrical contact resistance are considered. For different com-
binations of the value of thermal contact conductance and electri-
cal contact resistivity, the heat flow, efficiency and effectiveness
ðEff ;nu ) are obtained, which are listed in Table 1.
Generally speaking, the heat flow decreases as the thermal con-
tact conductance decreases and the electric contact resistivity
: from increases. Also, the efficiency decreases as the thermal contact con-
ductance decreases. However, as the electric contact resistivity
increases, the efficiency first increases in a region where the elec-
tric contact resistivity is smaller, then it decreases later in a region
where the electric contact resistivity is larger. When it is reminded
that the efficiency is the ratio of the power output to the heat flow,
the higher values of the efficiency obtained in Case 4, Case 9 and
Case 14 with smaller values of the heat flow are not awkward.
The effectiveness decreases monotonously as the thermal con-
tact conductance decreases and/or the electric contact resistivity
ð42Þ increases. Therefore, the effectiveness can be used as a meaningful
factor for the evaluation of the performance of a TEG device. Based
on whether the value of the effectiveness is greater than 0.8 or not,
a region of good performance of the TEG device is indicated in the
upper left region of Table 1, bounded by the dashed line.
4.2. Effect of the pellet length in the presence of contact resistances
In this section, examined is the effect of pellet length on the TEG
performance in the presence of the thermal and electric contact
resistances. Here, the situation with thermal contact conductance
of 2  104 W/m2 K and electric contact resistivity of 3  109
Xm2 is selected. It can be noted that the effectiveness Eff ;nu of the
power output, heat flow, efficiency, effectiveness for
power output, heat flow, efficiency for the case without contact
 C.N. Kim / Applied Thermal Engineering 130 (2018) 408–417
TEG module (having L = 2 mm for the pellets) with both the above
thermal and electric resistance is 0.6939 (Case 19 in Table 1).
The heat flow moving downward (obtained in the top elec-
trode), power output, efficiency and effectiveness of the TEG mod-
ule are presented in Fig. 12. In the case without contact resistances,
as the pellet length is decreased, both the heat flow and power out-
put are increased. However, as the pellet length is decreased, the
increase speed of the heat flow is higher than that of the power
output in a range of smaller pellet length, yielding a decrease in
the efficiency.
In the case with the contact resistance, the values of the heat
flow and power output are generally smaller than those in the case
without contact resistance, respectively. It can be noted that, as the
pellet length is decreased in the range of the pellet length of 0.2
mm < L < 0.6 mm (where L is the pellet length), the power output
is decreased, though the heat flow is increased, which yields a
drastic decrease in the efficiency.
Here, the efficiency in the case with the contact resistances is
higher in the range of 2.2 mm < L < 3.2 mm (g = 0.05553 when
the pellet length is 3.2 mm long) than that in the case without
any contact resistance (g = 0.05478 when the pellet length L is
3.2 mm long). This is because the heat flow is rather smaller in
the case with the contact resistance. However, in the range of
L < 2.2 mm, the situation of the efficiency is overturned. Therefore,
the efficiency may not be an ideal factor for the presentation of the
TEG module performance, and, therefore, the effectiveness can be
employed as a useful factor in a situation with contact resistances.
As the pellet length is decreased, the effectiveness is decreased
mildly when the pellet is longer. However, it is decreased drasti-
cally when the pallet is shorter. The value of the effectiveness
Eff ;nu in a situation when L = 0.2 mm is obtained to be around
0.09758.
5. Conclusion
In this study, a new numerical method for the analysis of ther-
moelectricity is developed, and the characteristics of the thermo-
electricity in a TEG module with thermal and electric contact
resistance are numerically investigated with home-made com-
puter code. In this study, as a factor presenting the performance
of a TEG device with thermal and electric resistance, the effective-
ness is newly defined by the ratio of the power output of a TEG
device with contact resistances to that without any contact resis-
tance, and is examined for different combinations of the values
of thermal contact conductance and electric contact resistivity.
Also, in the presence of thermal and electric contact resistance
the effect of the pellet length on the TEG performance is reviewed
in detail, presenting quantitatively that the performance of the TEG
417
with contact resistances is quite poor when the pellet length is
smaller.
This study does not assume any non-physical assumption in the
numerical procedure of the solution of the governing equations,
and therefore, can predict accurately the power output, Joule heat-
ing, Peltier heating and Thomson heating in a situation where the
thermoelectric properties depend on the temperature. The current
study suggests reasonable ranges of the values of thermal contact
conductance, electric contact resistivity and pellet length for a
meaningful design of a thermoelectric generator.
References
[1] D. Tartarinov et al., Optimized characterization of thermoelectric generators
for automobile application, J. Electron. Mater. 41 (6) (2012) 1706–1712.
[2] C.R. Reddy, S.S. Rao, V. Desai, K. Ramachandran, Modeling of an automotive
thermoelectric generator (ATEG), Internat. J. Sci. Res. (IJSR) 2 (5) (2013) 242–
246.
[3] M. Lazard, Heat transfer in thermoelectricity: modeling, optimization and
design, Proceedings of the 7th IASME/WSEA International Conference on Heat
Transfer, Thermal Engineering and Environment (HTE’09) PP 129-134
[4] M. Chen, L.A. Rosendahl, T. Condra, A three-dimensional numerical model of
thermoelectric generators in fluid power systems, Int. J. Heat Mass Transf. 54
(2011) 345–355.
[5] D. Yan, F.P. Dawson, M. Pugh, A.A. El-Deib, Time-dependent finite-volume
model of thermoelectric devices, IEEE Transact. Indust. Appl. 50 (1) (2014)
600–608.
[6] M. Picard, S. Turenne, D. Vasilevskiy, R.A. Masut, Numerical simulation of
performance and thermomechanical behavior of thermoelectric modules with
segmented Bismuth-Telluride-Based Legs, J. Electron. Mater. 42 (7) (2013)
2343–2349.
[7] S. Oki, K. Ito, S. Natsui, R.O. Suzuki, Numerical optimization of trapezoidal
thermoelectric elements for double-pipe-shaped module, J. Electron. Mater. 46
(3) (2016) 1358–1364.
[8] O. Hogblom, R. Andersson, Analysis of thermoelectric generator performance
by use of simulations and experiments, J. Electron. Mater. 43 (6) (2014) 2247–
2254.
[9] D. Ebling et al., Module geometry and contact resistance of thermoelectric
generator analyzed by multiphysics simulation, J. Electron. Mater. 39 (9)
(2010) 1376–1380.
[10] B.V.K. Reddy, Convective heat transfer and contact resistances effects on
performance of conventional and composite thermoelectric devices, J. Heat
Transfer 136 (2014), 010401-111.
[11] C.N. Kim, Numerical analysis of the performance prediction for a
thermoelectric generator, Korean J. Metals Mater. 53 (9) (2015) 666–669.
[12] B.V.K. Reddy et al., Thermoelectric performance of novel composite and
integrated devices applied to waste heat recovery, J. Heat Transfer 135 (2013),
031706-111.
[13] R. Bjork, An analytic model for the influence of contact resistance on the
thermoelectric efficiency, J. Electron. Mater. 45 (3) (2016) 1301–1308.
[14] S.I. Kim et al., Dense distribution arrays embedded in grain boundaries for
high-performance bulk thermoelectrics, Science 348 (109) (2015) 109–113.
[15] M. Chen, L.A. Rosendahl, T. Condra, A three-dimensional numerical model of
the thermoelectric generators in fluid power system, Int. J. Heat Mass Transf.
54 (2011) 345–355.
[16] A. Paraskevas, E. Koutroulis, A simple maximum power point tracker for
thermoelectric generators, Energy Convers. Manage. 108 (2016) 355–365.