See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/321225222

Wearable thermoelectric generator for harvesting heat on the curved human

wrist

Article  in  Applied Energy · November 2017

CITATIONS READS

9 1,137

4 authors, including:

Yancheng Wang Yaoguang Shi

Zhejiang University Xidian University
81 PUBLICATIONS   582 CITATIONS    9 PUBLICATIONS   105 CITATIONS   

SEE PROFILE SEE PROFILE

Deqing Mei
Zhejiang University
85 PUBLICATIONS   1,125 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Wearable flexible thermoelectric generator for harvesting human heat View project

All content following this page was uploaded by Yaoguang Shi on 27 November 2017.

The user has requested enhancement of the downloaded file.

 Applied Energy 205 (2017) 710–719

Contents lists available at ScienceDirect

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

Wearable thermoelectric generator for harvesting heat on the curved human MARK
wrist
⁎
Yancheng Wanga,b, , Yaoguang Shib, Deqing Meia,b, Zichen Chenb
a
State Key Laboratory of Fluid Power and Mechatronic Systems, School of Mechanical Engineering, Zhejiang University, Hangzhou 310027, China
b
Key Laboratory of Advanced Manufacturing Technology of Zhejiang Province, School of Mechanical Engineering, Zhejiang University, Hangzhou 310027, China

H I G H L I G H T S

• ARadii
numerical model to study the performance of wearable TEG on curved surface is developed.
• The TILof with
curvature of the curved surface have great effects on the heat distribution in the TEG.
• thin thickness and high thermal conductivity is beneficial to enhance the voltage generation.

A R T I C L E I N F O A B S T R A C T

Keywords: Wearable electronics and sensors for health monitoring are becoming increasingly popular as their functionality
Thermoelectric generator (TEG) continues to grow. Wearable thermoelectric generators (TEGs) are attracting interest due to their ability to self-
Numerical model power these electronic devices or sensors by harvesting human body heat. For wearable TEGs, a flexible thermal
Body-fitted coordinate interface layer (TIL) is used underneath the TEG for wearing on the human body. The large thermal resistance
Thermal interface layer (TIL)
induced at the interface between the skin and the TEG currently limits improvements in the performance of
Curved surface
Wearable
wearable TEGs and needs to be evaluated. This paper develops a numerical model to investigate the performance
of wearable TEGs on the curved human wrist. The TEG and bottom TIL are meshed using rectangular grids and
the body-fitted coordinate (BFC) transformation, respectively. Using the finite volume method (FVM), the
proposed model is calculated, and the temperature and voltage distributions in the TEG and bottom TIL are
analyzed. The effects of the radii of curvature of the curved surface, the material properties, and the thicknesses
of the TIL are investigated both numerically and experimentally. The results obtained in this research can be
utilized for optimal structural designs for wearable TEGs and for material selection of the TIL to enhance the
power generation for self-powered electronics.

1. Introduction development of new thermoelectric materials with a high figure of

merit (ZT) [4,5], novel fabrication processes [6,7], optimal structural
With the advancement of microelectromechanical systems (MEMS), designs [8–10], and so on.
miniaturized flexible electronics and sensors can be worn on the human Wearable TEGs are designed to be worn on the human body, e.g.,
body and used for health or environmental monitoring. The power the forehead, chest, forearm, wrist, legs, etc. For these TEGs, rigid
supply demands of these miniaturized devices are typically in the range thermoelectric legs are usually designed on a flexible substrate [11,12].
of microwatts to milliwatts [1]. Wearable thermoelectric generators The selection of material for the substrate is critical as it affects the
(TEGs) have been shown to be a promising strategy for powering these performance of the TEG, because it will introduce a large thermal re-
devices, as they can convert human body heat into electricity [2,3]. The sistance at the interface between the skin and TEG and in turn lower the
wearable TEGs that have been developed usually feature the ad- voltage and power generation. Sakamoto et al. [13] presented an ex-
vantages of small compact size and light weight, while the low effi- perimental method to evaluate thermal resistance at the interface layer
ciency and low power density limit TEGs’ application when the func- when the TEG was on a flat surface. The thermal interface substrate
tionality of the devices is expanded. To enhance the performance of with a thinner layer and higher thermal conductivity could enhance the
TEGs, considerable researches have been carried out in the performance of the TEG. When the TEG is worn on skin that is curved,

⁎
Corresponding author at: State Key Laboratory of Fluid Power and Mechatronic Systems, School of Mechanical Engineering, Zhejiang University, Hangzhou 310027, China.
E-mail address: [email protected] (Y. Wang).

http://dx.doi.org/10.1016/j.apenergy.2017.08.117
Received 15 May 2017; Received in revised form 1 July 2017; Accepted 12 August 2017
0306-2619/ © 2017 Elsevier Ltd. All rights reserved.
Y. Wang et al. Applied Energy 205 (2017) 710–719

Nomenclature m, n, l the coefficient in body fitted coordinates (BFC) transfor-

mation
Roman
Greek
W width of thermoelectric leg (mm)
H height of thermoelectric leg (mm) α Seebeck coefficient (VK−1)
L1 distance between N- and P- type thermoelectric legs in one γ electrical conductivity (Sm−1)
unit (mm) ξ, η axes in BFC
L2 half-distance between N- and P- type thermoelectric legs δ thickness of the TIL (mm)
in adjacent unit (mm)
T temperature (K) Subscripts
V voltage (V)
k thermal conductivity (W/m·K) e, w, n, s control interfaces
h convection coefficient (W/m2·K) C, E, W, N, S control volumes
J Jacobian factor OC open circuit

the flexible substrate, e.g., the thermal interface layer (TIL), is bent and parameter to affect performance. For wearable TEGs, poly-
deformed. The effects of bending around the curved surface and of the dimethylsiloxane (PDMS) has been utilized as a flexible insulating and
deformed TIL on the heat flowing through the TEG for heat harvesting supporting material for the encapsulation of rigid thermoelectric legs
needs to be investigated and become one goal of this research. [16–18]. The effects of PDMS filler in the TEG on a flat surface have
The performance of TEGs worn on the human body has been studied been studied via the development of a numerical model [19], and re-
by numerical modeling and experimental tests. Leonov [14] developed sults showed that the PDMS can improve the flexibility of the TEG
a model to study the effects of localized properties of the body (like skin without sacrificing power generation. However, the model developed
temperature and heat flow) on the thermal properties of the TEG. The previously does not consider the effects of a curved surface and the TIL,
curved shape of the skin and its effects were not included. The results which often occur in wearable applications. Here, we propose a nu-
showed that the radiator on the cold-side of the TEG can reduce the merical model that aims to investigate the performance of a wearable
thermal resistance at that part of the human body and, in turn, this TEG on curved skin surface. The influence of the deformation of the TIL
generates greater power. Suarez et al. [15] presented a quasi-three-di- induced on heat flow, temperature and voltage distributions in the TEG
mensional (3D) model to study the design criteria for rigid and flexible will be analyzed.
TEGs for wearable applications on the human body. For wearable TEGs, In this paper, we present a numerical model to investigate the
they identified the significance of thermal conductivity over Seebeck performance of a wearable TEG on a curved surface. The physical do-
coefficient and electrical resistivity of the thermoelectric material in mains of the rigid thermoelectric legs and deformed TIL are meshed
enhancing voltage and power generation. The thermal conductivity of using rectangular grids and body-fitted coordinate (BFC) transforma-
the filler material in the TEG has also been demonstrated as a critical tions, respectively. Then, this model is solved through the finite volume

Fig. 1. Structural design of a wearable TEG on the wrist (a);

cross-sectional view of the TEG on skin that is a flat surface
(b), and skin that is a curved surface (c).

711
Y. Wang et al. Applied Energy 205 (2017) 710–719

method (FVM) and utilized to study the heat transfer and temperature temperature (298.15 K), respectively.
and voltage distributions in the TEG. The effects of the radii of curva-
ture of the surface, thermal conductivity, and thickness of the TIL are
also analyzed. This is followed by experimental validation and a dis- 3. Modeling approach
cussion of the results.
3.1. Numerical modeling
2. Wearable TEG
When the TEG wearing on the curved skin of the wrist, the TIL
Fig. 1(a) shows a schematic view of a wearable TEG on a human substrate will bend to conform the curved skin. Thus, most deformation
wrist. The main structure of the TEG has several N-type and P-type occurs in the region of TIL substrate, as shown in Fig. 1(c). The heat
thermoelectric legs, copper strips, and the encapsulating PDMS around spreading from the wrist skin to ambient air, and heat transfer through
the thermoelectric legs. A thin layer of TIL substrate made of silicone the deformed TIL need to be investigated. The copper strips and solder
rubber can facilitate wearing the TEG on curved skin. The N-type and P- material utilized in this study have much higher γ and k than those of
type thermoelectric legs having the same cubic shape are connected by thermoelectric legs, PDMS filler, and TIL, as listed in Table 1. Therefore,
copper strips electrically in series and thermally in parallel. These both the electrical and thermal contact resistance at the solder and
thermoelectric legs are encapsulated using the soft PDMS to increase copper strip layers can be ignored. The deformed PDMS far from the
the flexibility of the TEG. thermoelectric legs is not involved in the modeling as it has little impact
When the TEG is placed on a flat surface (as in Fig. 1(b)), the TEG on the heat transfer. For the TEG on a flat surface, the differential
will be in its initial state. The encapsulating PDMS and TIL substrate are equations to calculate thermal and electrical distributions of the ther-
not deformed. When the TEG is worn on the wrist, the flexible TIL is moelectric legs in a Cartesian coordinate system are [19]
deformed to suit the rigid thermoelectric legs and the curved skin.
Fig. 1(c) shows one pair of thermoelectric legs with the deformed TIL ∂ ⎛ ∂T ⎞ ∂ ⎛ ∂T ⎞
k + k ⎜ =0 ⎟
substrate on the curved wrist skin; the encapsulating PDMS, the rigid ∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎠ (1)
thermoelectric legs, and the copper strips can be assumed to have re-
latively small deformations. At room temperature (25 °C), the surface of
the wrist skin typically has a temperature of 30 °C. It can be seen as the ∂ ⎛ ∂VOC ⎞ ∂ ⎛ ∂VOC ⎞ ∂ ⎛ ∂T ⎞ ∂ ⎛ ∂T ⎞
γ + γ ⎜ + γα
⎟ + γα =0 ⎜ ⎟

heat source for the TEG. Therefore, the bottom surface of the TIL, in ∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎠ ∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎠ (2)
contact with the skin, is treated as the hot-side with a temperature of Th.
The top surface of the TEG is exposed to ambient air and can be seen as where T is absolute temperature, VOC is open-circuit voltage. In Eq. (2),
the cold-side at Ta = 25 °C (298.15 K). The convection coefficient (h) of the Seebeck coefficient (α) is involved to convert the temperature dif-
the ambient air is 20 W/m2·K [20]. The temperature differences be- ference (ΔT) into VOC.
tween the hot and cold-sides will lead to a steady heat flux flowing The temperature distributions in the PDMS, TIL, solder layers and
through the TEG, which in turn generates the voltage for self-powered copper strip regions can be calculated by using Eq. (1). Because the
electronics. PDMS and TIL substrate can be seen as the electrical insulator, there
Since the TEG is made up of multiple thermoelectric units, each unit will be no voltage generated in these regions. Compared to P-type and
has one pair of N-type and P-type thermoelectric legs. The structural N-type thermoelectric legs, both the copper strip and solder layers have
dimensions of one thermoelectric unit are given in Fig. 1(b). The width extremely low α. The induced voltages via Seebeck effect in these re-
and height of the cubic-shaped thermoelectric legs are W and H, re- gions would be small [25]. Therefore, to calculate the voltage dis-
spectively. The distance between the N-type and P-type legs in one unit tributions in the region of copper strip and solder layers, the α in Eq. (2)
is L1. Thus, the upper copper strip has a length (2W + L1). The bottom can be neglected as
copper strip, with a longer length (L2 + W) than that of the thermo-
electric legs, can reduce the bending stress at the soldering interface ∂ ⎛ ∂VOC ⎞ ∂ ⎛ ∂VOC ⎞
γ + γ ⎜ =0 ⎟

between the thermoelectric legs and the copper strips and therefore ∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎠ (3)
improves the reliability of the TEG.
For wearable TEGs, several technologies, such as dispenser printing The TIL will be deformed when the TEG is mounted on a curved
[21,22] and screen printing [23], have been proposed to fabricate surface. For numerical modeling, the rectangular grids (e.g. control
thermoelectric powder onto a flexible substrate. It is difficult to control volumes (CVs)) cannot accurately mesh the deformed TIL in Cartesian
the organic contamination during these processes, the physical and coordinate system. Therefore, non-rectangular grids will be utilized and
chemical defects in the powder-based thermoelectric materials during need to be transformed into rectangular grids to facilitate calculation.
the fabrication of the thermoelectric legs. Thus, the thermoelectric legs Here, the body fitted coordinate (BFC) transformation is adopted [30]
made of powder-based polymer mixtures usually have a relatively and shown in Fig. 2.
lower figure of merit (ZT) and larger inner resistance [21]. To increase For the BFC, the coordinate axes are defined as ξ and η. The trans-
the ZT, Bi2Se0.5Te2.5 and Bi0.5Sb1.5Te3 powders are selected [19] and formations from Cartesian coordinates to BFC are [30]
fabricated into bulk thermoelectric legs utilizing a semisolid powder
forming process [24].
Then, the rigid thermoelectric legs are soldered onto the copper Table 1
strips to improve the mechanical and electrical stability. Here, Seebeck coefficient (α), electrical resistivity (γ), and thermal conductivity (k) of the
Sn96.5Ag3Cu0.5 solder is utilized to weld the N-type and P-type ther- materials utilized in wearable TEG (at 298.15 K).
moelectric legs onto the copper strips. The thicknesses of the copper
Materials α (μV/K) k (W/m·K) γ (S/m)
strips and solder layers are controlled to about 0.3 and 0.1 mm, re-
spectively. For the bottom TIL, the initial thickness is δT. On the curved N-type leg (Bi2Se0.5Te2.5) −215 1.1 6.9 × 104
surface, the TIL will be deformed with a radius of curvature (R) the P-type leg (Bi0.5Sb1.5Te3) 186 1.2 1.14 × 105
same as that of the curved surface, as shown in Fig. 1(c). Copper strips 1.8 [25] 400 [26] 5.96 × 107 [26]
PDMS filler – 0.27 [27] 0
The properties of the materials utilized in the wearable TEG are
Thermal interface layer – kT –
listed in Table 1. Using these parameters, the ZT values for the N-type Sn96.5Ag3Cu0.5 solder – 64 [28] 8 × 106 [29]
and P-type thermoelectric legs can be calculated as 1.0 and 0.9 at room

712
Y. Wang et al. Applied Energy 205 (2017) 710–719

distribution in the TIL will have the same form as Eq. (1). Combining
Eqs. (1)–(4), the 2D numerical model of the temperature and voltage
distribution in the TEG with TIL on a curved skin surface is built.

3.2. Model calculation

For the developed model, Eqs. (1)–(4) need to be solved iteratively

as the properties of thermoelectric materials are considered to be a
function of temperature [19]. Here, we utilized the FVM [26] to solve
these equations and to derive the voltage and temperature distributions
in the TEG. Compared with finite element modeling (FEM), the ad-
vantages of using FVM can be describes as: (1) For FEM, when wearable
TEG on the curved skin, the deformed TIL substrate requires large
Fig. 2. Body-fitted coordinate transformation.
amount of meshes or mesh refinement to avoid mesh convergence [31],
and limits the application of FEM in wearable TEG. (2) In FVM, the
⎧ ξ xx + ξ yy = P (ξ ,η) physical domain of the TEG can be divided by non-overlapping CVs, as
⎪ η + η = Q (ξ ,η) shown in Fig. 3(a). The transported quantities of these CVs are con-
⎪ xx yy
⎪ x ξ x ξξ + yξ yξξ served. Therefore, Eqs. (1)–(4) can be easily discretized by using a 2nd-
2 2
⎨ P (ξ ,η) = − xξ2 + yξ2 (ξx + ξ y ) order central differencing scheme.
⎪ x x +y y As in Fig. 3(a), the thermoelectric legs, copper strips, solder, and
⎪Q (ξ ,η) = − η ηη2 η2 ηη (η 2 + η 2)
⎪ x η + yη x y PDMS filler are meshed by rectangular grids in Cartesian coordinate
⎩ (4)
system. For the deformed TIL, the meshed CVs are generated based on
where ξx, ξy, ηx, and ηy are the first partial derivatives of ξ and η with BFC transformation. The CVs in the regions of the thermoelectric legs,
respect to x and y, respectively; and xξ, yξ, xη, and yη are the first partial copper strips, solder, and PDMS filler have the same length (Δx) in the
derivatives of x and y with respect to ξ and η, respectively. ξxx, ξyy, ηxx, x-axis and different lengths (Δy) in the y-axis. For example, the solder
and ηyy are the second partial derivatives of ξ and η with respect to x layer and copper strips have greater Δy than that in the regions of the
and y, respectively; while xξξ, xηη, yξξ, and yηη are the second partial thermoelectric legs and PDMS filler. These are due to the copper and
derivatives of x and y with respect to ξ and η, respectively. P and Q are solder layers having much higher k and γ than those of PDMS and
the control functions to adjust the distribution and orthogonality of the thermoelectric legs, as in Table 1. In the BFC, the CVs of the TIL region
grid meshes, respectively. have the same length in both the ξ- and η-axes and are marked as Δξ and
Therefore, the deformed TIL in Cartesian coordinates is transferred Δη, respectively, as shown in Fig. 3(b). For the deformed TIL, the in-
into BFC through Eq. (4), and the differential equation of temperature ternal and upper CVs near the copper strip, and the CVs at the bottom

Fig. 3. (a) Non-uniform grids with 3948 CVs for the TEG on a curved surface in Cartesian coordinate; (b) CVs of the TIL in Cartesian coordinate system and BFC transformation; (c)
internal and upper CVs, and (d) bottom CVs of the TIL in BFC.

713
Y. Wang et al. Applied Energy 205 (2017) 710–719

side near the skin in the BFC are shown in Fig. 3(c) and (d), respec- Fig. 3(d)) can also be calculated.
tively. To simulate the skin, the additional CVs with constant hot-side For the deformed TIL, the temperature values calculated in the BFC
temperature (Th) are set as shown in Fig. 3(d). The values in thermal need to be converted into those of the Cartesian coordinate system, and
and electrical fields (T, V) of each CV are evaluated at the centroid the inverse transformation is derived as
(point C) with the adjacent CVs. 2
Thus, for a specific case, the W and H of the thermoelectric legs are ⎧ mx ξξ −nx ξη + lx ηη = −J (Px ξ + Qx η)
2
⎨ myξξ −nyξη + lyηη = −J (Pyξ + Qyη)
1.15 mm and 1.2 mm, respectively. L1 and L2 are set as 1.2 mm and ⎩ (6)
0.6 mm, respectively. The original thickness of the TIL (δT) is 0.5 mm
and the radius of curvature of the curved skin R is 30 mm. There are where m, n, l are the coefficients; and J is the Jacobian factor. These
3948 CVs in total for the physical domain of the TEG, shown in parameters can be calculated as
Fig. 3(a), and about 940 CVs are generated in the deformed TIL. 2 2
⎧ m = x η + yη
The discrete equations for solving temperature and voltage dis- ⎪n = x x + y y
tributions in the thermoelectric legs and surrounding PDMS filler are ⎪ ξ η ξ η

studied in our previous work [19]. The discretized equation to calculate ⎨ l = xξ2 + yξ2
⎪
the temperature for the CVs in the BFC is ⎪ J = x ξ yη −x η yξ
⎩ (7)
ACT TC = AeT TE + AwT TW + AnT TN + AsT TS Therefore, by using Eqs. (1)–(7), the temperature and voltage dis-
T tributions in the wearable TEG can be obtained. The calculation pro-
⎧ Ae = ke Δη / δξ
⎪ AT gram and codes are written with MATLAB software.
= k w Δη / δξ
⎪ w
AT = kn Δξ / δη
⎨ n 4. Results and discussion
⎪ AsT = ks Δξ / δη
⎪ T
⎩ AC = AeT + AwT + AnT + AsT (5) 4.1. Temperature and voltage distributions

where δξ and δη are the distances between point C and the centroids of Fig. 4(a) and (b) show the typical temperature and voltage dis-
adjacent CVs in the BFC, respectively. Using the same method, the tributions in the TEG from the wrist skin to ambient air, where
temperature values of the boundary CVs on the bottom side (in Th = 303.15 K, Ta 298.15 K, and h 20 W/m2·K. Because the copper

Fig. 4. (a) Temperature and (b) voltage distributions in the TEG when mounted on a curved surface; (c) close-up view of the temperature distribution in the deformed TIL; temperature
differences, (d) between hot and cold-sides of the TEG, (e) in region D, the thermoelectric legs and central PDMS, and (f) in the deformed TIL.

714
Y. Wang et al. Applied Energy 205 (2017) 710–719

strips and solder layers have small thicknesses and much higher k, a the TEG will be transferred to thermoelectric legs, only a small portion
small fraction of the temperature and voltage differences drop across of heat loss occurred in the TIL and affect the temperature distribution,
these regions. The voltage is only generated where there are non-in- as in Fig. 4(c).
sulated components, such as the thermoelectric legs, copper strips, and Under these three types of mounted surfaces, the voltage (VOC) is
solder layers. Here, the generated voltage difference of the TEG is about calculated as 0.067 mV. We can also conclude that the effect of R on
0.067 mV, as shown in Fig. 4(b). VOC is also low. If we increase the ΔT or choose an improper δT and kT of
For the deformed TIL, the temperature distribution is not uniform the TIL, the effects of R on the temperature distribution and generated
because the deformed TIL has a variable thickness, as shown in VOC will be more significant. In this study, if the designed wearable TEG
Fig. 4(c). In Fig. 4(d), the temperature differences are symmetrically has sufficient flexibility to conform to the curved surface, the radii of
distributed along the x-axis of the TEG, and the maximum temperature curvature of the curved surfaces will not affect the performance of the
difference (ΔT) is 0.32 K and occurs in the region of PDMS at both sides TEG.
by thermoelectric legs. In region D (including the P- and N-type ther-
moelectric legs, and central PDMS), there is a relatively low ΔT with a 4.3. The effect of the TIL substrate
value of 0.19 K because the upper copper strip has a higher k
(= 400 W/m·K) than that of the side PDMS (0.27 W/m·K). Thus, most From the above results, we know that the temperature distributions
of the heat will be transferred through the copper strips. A close-up in the deformed TIL region are non-uniform. The material selection and
view of the temperature differences in region D is shown in Fig. 4(e); ΔT properties of the TIL, such as thickness and thermal conductivity, may
is in the range of about 0.18 to 0.19 K. As for the central PDMS, there is affect the performance of the wearable TEG. Furthermore, to improve
a slightly higher value for ΔT than that in the region of the thermo- the performance of the wearable TEG on a curved surface, the heat loss
electric legs, due to its lower k. Thus, the larger thermal resistance of in the TIL substrate should be reduced. In this study, two parameters
the PDMS filler will lead to most of the heat flowing through the (thickness δT and thermal conductivity kT) are selected to study the
thermoelectric legs, which is beneficial for heat harvesting in wearable effects of the TIL. Here, the TIL with three different thicknesses
TEGs. (δT = 0.3, 0.5, and 1.0 mm) and thermal conductivities (kT = 4.0, 1.6,
Fig. 4(f) shows the temperature difference along the x-axis in the and 0.27 W/m·K) are investigated, respectively. For the TIL material,
deformed TIL. The ΔT in the TIL region in contact with the PDMS is PGS-7000 and PSG-1000 ceramic-filled silicone rubber (Shenzhen
much lower (about 0.004 K), while in the TIL region in contact with the Maoyuan, New Materials Technology Co. Ltd) with different con-
copper strips, ΔT is about 3.25 times higher with 0.013 K. Thus, the centrations feature kT of 4.0 and 1.6 W/m·K; the PDMS (Sylgard 184,
large thermal resistance of the encapsulating PDMS will also block the Dow Corning Corp.) has a kT of 0.27 W/m·K [27].
heat transfer though the PDMS and accounts for this low ΔT value in the Fig. 6(a)–(b) show the effects of the δT on the calculated ΔT in the
region where the TIL is in contact with the central PDMS. Compared to TEG and TIL, respectively. The correspondingly generated VOC is shown
the thermoelectric leg region, we can see that about 7% of ΔT is in- in Fig. 6(c). The thermal conductivity of the TIL material kT is set as
troduced due to the deformed TIL substrate. This temperature differ- 4.0 W/m·K. Generally, we can see that the ΔT increased as the thickness
ence in the TIL cannot be utilized for heat harvesting and needs to be
reduced. Hence the results obtained in this research indicated that the
developed model can be utilized to investigate the temperature and
voltage distributions of wearable TEG on curved skin surface, and the
effects of the deformed TIL substrate can also be studied.

4.2. The effect of curved surface

Because the wearable TEG is designed for wearing on human skin,

the body surface shape and local skin area may also affect the perfor-
mance of the TEG. For example, human subject of different genders and
age groups may have different radii of curvature in the forehead, chest,
wrist, and hand. To conform to these curved skin surfaces, the TEG may
be bent to the same radii of curvature as that of the curved skin sur-
faces. Thus, the TEG, especially the TIL substrate should have different
deformations for these applications. In this study, we utilized the pro-
posed model to analyze the effects of the radii of curvature of the
curved surfaces. For a comparison, surfaces with three radii of curva-
tures are studied: R = 10 mm, R = 30 mm, and a flat surface (R = ∞).
The thickness and thermal conductivity of the TIL is set as δT = 0.5 mm
and kT = 4.0 W/m·K.
Fig. 5(a) shows the temperature difference (ΔT) between the hot-
and cold-sides of the TEG along the x-axis under different radii of
curvature. We can see that the TEG has an almost constant ΔT when
worn on different curved surfaces. On the flat surface, ΔT has a slightly
higher value of 0.187 K. As R decreases to 10 mm, ΔT reduces to about
0.185 K. For a close-up view of the TIL region, the same trend in ΔT can
be observed as shown in Fig. 5(b). Generally, there is no significant
difference on the ΔT among R = 10 mm, 30 mm, and flat surface. Two
reasons can be explained as: firstly, for human wearing application, the
ΔT between the skin and ambient air is relatively low, typically in the
range of about 3–5 K. Secondly, compared to thermoelectric legs, the
TIL has a small thickness of δT = 0.5 mm and high thermal conductivity Fig. 5. Effects of the mounted surfaces with different radii of curvatures on ΔT: (a) the
TEG and (b) the TIL.
of kT = 4.0 W/m·K. Therefore, most of heat from the bottom surface of

715
Y. Wang et al. Applied Energy 205 (2017) 710–719

Fig. 6. Effects of the δT on (a) ΔT in the TEG, (b) ΔT in the TIL, (c) VOC; and effects of the kT on (d) ΔT in the TEG, (e) ΔT in the TIL, and (f) VOC.

increased, as shown in Fig. 6(a). For δT = 0.3 mm, the ΔT is about harvesting by the thermoelectric legs. Therefore, the temperature dif-
0.18 K. As for δT = 1.0 mm, the value of the generated ΔT increased to ferences in the TEG are mainly dominated by the temperature differ-
0.20 K. The increase in δT of the TIL will enlarge the thermal resistance, ences in the TIL substrate, which has a slight effect on the voltage
and lead to more heat loss in the substrate. In this respect, it will lower generation. Combined with the effects of δT on ΔT and VOC, there
the temperature difference for heat harvesting. Yet, the thicker TIL will should be an optimal thickness value for the selection of the TIL to
extend the distance for heat flow from the skin on the bottom to the obtain a relatively low heat loss in the substrate and high VOC for
ambient air, and the expected ΔT between the hot and cold-sides of the wearable TEGs.
TEG will be increased which is beneficial for heat harvesting. For a Due to the original thickness of the TIL (δT) was set as 0.5 mm
close-up view of the ΔT in the deformed TIL, the same trend can be during the model calculation, here we also set the δT as 0.5 mm, the
observed. As in Fig. 6(b), a larger δT will generate greater ΔT. When δT effects of kT on ΔT and VOC are shown in Fig. 6(d)–(f). In Fig. 6(d), we
is 0.3 mm, the maximum ΔT in the TIL region is 0.008 K. As δT in- can see that ΔT in the TEG is increased along with the reduction of kT.
creased to 1.0 mm, the maximum ΔT increased about 3.13 more times When kT = 4.0 W/m·K, the ΔT is 0.18 K. For kT = 0.27 W/m·K, the ΔT
(0.025 K). The calculated VOC of the TEG with three different thick- is increased 1.89 times with a value of 0.34 K. For the TIL region
nesses are shown in Fig. 6(c). Under these three TIL thicknesses, the (Fig. 6(e)), almost the same trend can be observed, ΔT increases as there
TEG generates an almost constant VOC of about 0.067 mV. At is a reduction of kT. When kT = 4.0 W/m·K, the ΔT in the TIL substrate
δT = 0.3 mm, VOC is 0.0672 mV; when δT is thickened to 1.0 mm, the is 0.01 K. For kT = 0.27 W/m·K, the ΔT is increased to 0.18 K. In-
VOC is weakened to 0.0668 mV. As for the reasons, increasing the creasing the kT decreases the thermal resistance in the TIL, and this will
thickness of the TIL will enlarge the thermal resistance and in turn leads reduce the heat loss and temperature difference in the region of TIL.
to higher temperature differences in the region of the TIL substrate. Thus, most of heat and temperature difference can be used to generate
However, these temperature differences cannot be utilized for heat greater voltage via Seebeck effect. For the TEG with three different kT,

716
Y. Wang et al. Applied Energy 205 (2017) 710–719

the calculated VOC is shown in Fig. 6(f). When kT = 0.27 W/m·K, the greater thermal resistance at the substrate layer, more heat loss will
VOC equals to 0.064 mV. When kT = 4.0 W/m·K, the VOC is increased to exist and in turn reduce the generated VOC and enhance the experi-
0.067 mV. Therefore, for the structural design of the wearable TEG, the mental and calculated discrepancy.
TIL with a higher thermal conductivity will has a positive effect on the
voltage generation and heat harvesting performance improvement. 5.2. Wrist wearing experiment
Compared to the effects of radii of curvature, the significance of the
thickness and thermal conductivity of the TIL on improving the TEG The fabricated wearable TEG is mounted onto the human wrist to
voltage generation can be identified. A TIL with appropriate thickness harvest body heat, as shown in Fig. 8(a). PGS-7000 silicone rubber with
and thermal conductivity can reduce the heat loss in the TIL region and 0.3 mm thickness is utilized as the TIL substrate, and the thermal
help to generate greater voltages for the application. As for the optimal conductivity is 4.0 W/m·K. During the experiment, two conditions are
parameters of the TIL, optimized structural design of wearable TEG conducted: (1) standing without motion, where the convection coeffi-
including the TIL needs to be conducted in future work. cient (h) at the cold-side of the TEG is 20 W/m2·K [20]. (2) Walking
with an arm swing, where the convection coefficient is 75 W/m2·K [32].
5. Experiments The generated voltage is recorded using an oscilloscope (Agilent, MSO-
X 2002A). The temperatures of the wrist skin surface and ambient air
5.1. Model validation are measured by surface thermocouple and an air thermometer with
values of 27 °C and 19 °C, respectively.
To validate the model developed here, a wearable TEG with 24 pairs When the operator stands and keeps his hand still without motion,
of thermoelectric legs is fabricated. The dimensions of the thermo- the TEG on the wrist skin can generate about 3.0 mV. When the op-
electric legs, copper strips, PDMS, and solder in one unit are the same as erator is walking and his hand keeps swinging, the generated voltage
the model calculation (in Section 3.2), the soft extra PDMS between can be increased to 11.2 mV. When the hand swings, the cold-side of
units has the width of 2 mm. For the TIL, ceramic-filled silicone rubber the TEG has a higher convection coefficient, so it can enlarge the
with a thermal conductivity of 4.0 W/m·K (PGS-7000) and thickness of temperature difference for heat harvesting and generate a higher VOC.
0.5 mm is utilized. Fig. 7(a) shows the experimental setup of the TEG on Thus, the results indicate that the wearing conditions and body motions
the curved surface. This TEG is mounted onto a hollow aluminum pipe will affect the performance of a wearable TEG. For this millivolt level
with R = 30 mm. This aluminum pipe is used as the heat source, where voltage, a DC-DC boost circuit [33] will be utilized for the powering of
a silicone-based rubber heating sheet is placed inside with the tem- wearable electronics and/or devices.
perature controlled by a PID controller. The temperature of the outer
surface of the pipe as the hot-side temperature (Th) is measured by a 6. Conclusions
surface thermocouple (SA1XL-K, OMEGA Engineering Inc.). The cold-
side of the TEG is exposed to ambient air with a temperature of This paper presents a numerical model to investigate the perfor-
298.15 K and convection coefficient of 20 W/m2·K. During the experi- mance of a wearable TEG on a curved skin surface. The BFC transfor-
ment, Th is changed from 303.15 K to 338.15 K. The generated voltage mation and inverse transformation are utilized to mesh the deformed
is measured using a digital multimeter (34401A, Keysight Technolo- TIL for numerical modeling. By using the FVM, the temperature and
gies). For a comparison, the performance of the TEG on a flat surface is voltage distributions of the TEG on curved surfaces are analyzed. In
also measured. addition, the effects of the radii of curvature of the surfaces, and the
Fig. 7(b) shows the measured VOC compared to the calculated results
of the model when the TEG is worn on curved and flat surfaces. Gen-
erally, the experimental measured results matched well with the model
calculation, and the calculated values are a little higher than those of
the measured results. This is due to the thermal and electrical contact
resistances at the interface layers (such as the solder and copper strips)
that exist, and they will reduce the voltage generation. The maximum
discrepancy in the VOC is about 13% which occurs at ΔT = 25 K. When
worn on human skin, the ΔT is only 2–3 K, thus the discrepancy in the
VOC is quite small and demonstrates that thermal and electrical contact
resistances at the interface layers are relatively small and can be ig-
nored during the modeling. Based on the comparison results, we can
conclude that the model developed in this paper can be utilized to ac-
curately predict the performance of a TEG worn on curved surfaces.
To validate the effects of the TIL, experimental tests of the TEG with
three different thicknesses (δT = 0.3, 0.5, 1.0 mm) and three different
types of TIL material (PGS-7000 silicone rubber, PSG-1000 silicone
rubber, and PDMS) are carried out. These three types of TIL materials
have 4.0, 1.6, and 0.27 W/m·K thermal conductivities, respectively.
During the experiment, Th is set at 303.15 K with a room temperature of
298.15 K. The results comparison of the experimental tests and nu-
merical modeling are summarized in Table 2. The calculated VOC also
matched well with the experimental measured values, with the average
discrepancy at about 10%. Reducing the thickness and increasing the
thermal conductivity of the TIL will leads to greater VOC. This trend is
the same as that in the numerical modeling in Section 4.3. Along with
the increase in δT, the discrepancies in the experimental and calculated
VOC are also enlarged. While for kT, the increase in kT has a lower
discrepancy with the experimental and calculated VOC. Because the TIL Fig. 7. (a) Measurement setup for a wearable TEG mounted on a curved surface, (b)
comparison of the VOC.
with greater thickness and lower thermal conductivity will lead to

717
Y. Wang et al. Applied Energy 205 (2017) 710–719

Table 2
Experimental and calculated VOC for different types of TIL.

Different parameters kT = 4.0 W/m·K δT = 0.5 mm

0.3 mm 0.5 mm 1.0 mm 0.27 W/m·K 1.6 W/m·K 4.0 W/m·K

Measured VOC (mV) 1.55 1.50 1.44 1.36 1.43 1.5

Calculated VOC (mV) 1.612 1.609 1.605 1.543 1.600 1.609
Discrepancy (%) 4.0 7.3 11.5 13.5 11.9 7.3

thermoelectric generator for human body applications. Sens Actuators A

2009;156:95–102.
[2] Im H, Moon HG, Lee JS, Chung IY, Kang TJ, Kim YH. Flexible thermocells for uti-
lization of body heat. Nano Res 2014;7:443–52.
[3] Madan D, Wang Z, Wright PK, Evans JW. Printed flexible thermoelectric generators
for use on low levels of waste heat. Appl Energy 2015;156:587–92.
[4] Fu CG, Bai SQ, Liu YT, Tang YS, Chen LD, Zhao XB, et al. Realizing high figure of
merit in heavy-band p-type half-Heusler thermoelectric materials. Nat Commun
2015;6(8144):1–7.
[5] Mei DQ, Li Y, Yao ZH, Wang H, Zhu TJ, Chen SC. Enhanced thermoelectric per-
formance of n-type PbTe bulk materials fabricated by semisolid powder processing.
J Alloy Compd 2014;609:201–5.
[6] Lu Z, Zhang H, Mao C, Li CM. Silk fabric-based wearable thermoelectric generator
for energy harvesting from the human body. Appl Energy 2016;164:57–63.
[7] Rojas JP, Conchouso D, Arevalo A, Singh D, Foulds IG, Hussain MM. Paper-based
origami flexible and foldable thermoelectric nanogenerator. Nano Energy
2017;31:296–301.
[8] Menon AK, Yee SK. Design of a polymer thermoelectric generator using radial ar-
chitecture. J Appl Phys 2016;119(055501):1–11.
[9] Ouyang ZL, Li DW. Modelling of segmented high-performance thermoelectric gen-
erators with effects of thermal radiation, electrical and thermal contact resistances.
Sci Rep 2016;6(24123):1–12.
[10] Fateh H, Baker CA, Hall MJ, Shi L. High fidelity finite difference model for ex-
ploring multi-parameter thermoelectric generator design space. Appl Energy
2014;129:373–83.
[11] Jo SE, Kim MK, Kim MS, Kim YJ. Flexible thermoelectric generator for human body
heat energy harvesting. Electron Lett 2012;48:1015–7.
[12] Kim SJ, We JH, Cho BJ. A wearable thermoelectric generator fabricated on a glass
fabric. Energy Environ Sci 2014;7:1959–65.
[13] Sakamoto T, Iida T, Sekiguchi T, Taguchi Y, Hirayama N, Nishio K, et al. Selection
and evaluation of thermal interface materials for reduction of the thermal contact
resistance of thermoelectric generators. J Electron Mater 2014;43:3792–800.
[14] Leonov V. Thermoelectric energy harvester on the heated human machine. J
Micromech Microeng 2011;21(125013):1–8.
[15] Suarez F, Nozariasbmarz A, Vashaee D, Ozturk MC. Designing thermoelectric
generators for self-powered wearable electronics. Energy Environ Sci
Fig. 8. Wrist wearing experiment: (a) experimental setup, (b) measured results. 2016;9:2099–113.
[16] Kim SJ, Lee HE, Choi H, Kim Y, We JH, Shin JS, et al. High-performance flexible
thermoelectric power generator using laser multiscanning lift-off process. ACS Nano
thickness and thermal conductivity of the TIL are investigated both 2016;10:10851–7.
[17] Hyland M, Hunter H, Liu J, Veety E, Vashaee D. Wearable thermoelectric generators
numerically and experimentally. The significance of the thickness and
for human body heat harvesting. Appl Energy 2016;182:518–24.
thermal conductivity of the TIL over the radii of curvature of the curved [18] Suarez F, Parekh DP, Ladd C, Vashaee D, Dickey MD, Öztürk MC. Flexible ther-
surface in improving the performance of the wearable TEG is identified. moelectric generator using bulk legs and liquid metal interconnects for wearable
The TIL with thinner thickness and higher thermal conductivity could electronics. Appl Energy 2017;202:736–45.
[19] Shi YG, Wang YC, Mei DQ, Chen ZC. Numerical modeling of the performance of the
improve the heat transfer and voltage generation of the wearable TEG, thermoelectric module with PDMS packaging. Int J Energy Res 2017 [Unpublished
and is recommended for the structural design of the TEG for body heat results].
harvesting. [20] Chen YJ, Young TF. Thermal stress and heat transfer characteristics of a Cu/dia-
mond/Cu heat spreading device. Diam Relat Mater 2009;18:283–6.
Future work will focus on the optimal structural design of the [21] Siddique ARM, Rabari R, Mahmud S, Heyst BV. Thermal energy harvesting from the
wearable TEG by using the developed numerical model and FVM. The human body using flexible thermoelectric generator (FTEG) fabricated by a dis-
fabrication method of the optimized wearable TEG, and its application penser printing technique. Energy 2016;115(Part 1):1081–91.
[22] Kim MK, Kim MS, Lee S, Kim C, Kim YJ. Wearable thermoelectric generator for
for self-powered electronics and/or sensor by harvesting human body harvesting human body heat energy. Smart Mater Struct 2014;23(105002):1–7.
heat will also be carried out. [23] Kim SJ, Choi H, Kim Y, We JH, Shin JS, Lee HE, et al. Post ionized defect en-
gineering of the screen-printed Bi2Te2.7Se0.3 thick film for high performance flexible
thermoelectric generator. Nano Energy 2017;31:258–63.
Acknowledgments
[24] Mei DQ, Wang H, Li Y, Yao ZH, Zhu TJ. Microstructure and thermoelectric prop-
erties of porous Bi2Te2.85Se0.15 bulk materials fabricated by semisolid powder pro-
This work was supported by the National Natural Science cessing. J Mater Res 2015;30:1–8.
[25] Kasap SO. Thermoelectric effects in metals: thermocouples. Department of
Foundation of China (Grant No. 51575485), Zhejiang Province Natural
Electrical Engineering: University of Saskatchewan, Canada; 2001.
Science Foundation of China (Grant No. LY16E05002), Fund for [26] Oliveira KSM, Cardoso RP, Hermes CJL. Numerical assessment of the thermo-
Creative Research Groups of the National Natural Science Foundation dynamic performance of thermoelectric cells via two-dimensional modelling. Appl
of China (Grant No. 51521064), and Fundamental Research Funds for Energy 2014;130:280–8.
[27] Dow Corning. Product information Dow Corning® 184 silicone elastomer; 2013.
the Central Universites (2017XZZX001-02A). http://www.dowcorning.com/applications/search/defaultaspx?R=131EN.
[28] Amin NAAM, Shnawah DA, Said SM, Sabri MFM, Arof H. Effect of Ag content and
References the minor alloying element Fe on the electrical resistivity of Sn–Ag–Cu solder alloy.
J Alloy Compd 2014;599:114–20.
[29] Liou BH, Chen CM, Horng RH, Chiang YC, Wuu DS. Improvement of thermal
[1] Wang Z, Leonov V, Fiorini P, Van Hoof C. Realization of a wearable miniaturized management of high-power GaN-based light-emitting diodes. Microelectron Reliab

718
Y. Wang et al. Applied Energy 205 (2017) 710–719

2012;52:861–5. coefficient of human body under strong forced convective flow. J Wind Eng Ind
[30] Thompson JF, Warsi ZUA, Mastin CW. Numerical grid generation: foundations and Aerodyn 2014;126:107–17.
applications vol. 45. Amsterdam: North-Holland; 1985. [33] Thielen M, Sigrist L, Magno M, Hierold C, Benini L. Human body heat for powering
[31] Babuska I, Szabo B. On the rates of convergence of the finite element method. Int J wearable devices: from thermal energy to application. Energy Convers Manage
Numer Meth Eng 1982;18:323–41. 2017;131:44–54.
[32] Li C, Ito K. Numerical and experimental estimation of convective heat transfer

719

View publication stats