Content uploaded by Je-Hyeong Bahk
Author content
All content in this area was uploaded by Je-Hyeong Bahk on Oct 08, 2015
Content may be subject to copyright.
10362 |J. Mater. Chem. C, 2015, 3,10362--10374 This journal is ©The Royal Society of Chemistry 2015
Cite this: J.Mater. Chem. C, 2015,
3,10362
Flexible thermoelectric materials and device
optimization for wearable energy harvesting
Je-Hyeong Bahk,†*
a
Haiyu Fang,
b
Kazuaki Yazawa
a
and Ali Shakouri
a
In this paper, we review recent advances in the development of flexible thermoelectric materials and
devices for wearable human body-heat energy harvesting applications. We identify various emerging
applications such as specialized medical sensors where wearable thermoelectric generators can have
advantages over other energy sources. To meet the performance requirements for these applications,
we provide detailed design guidelines regarding the properties of the material, device dimensions, and
gap fillers by performing realistic device simulations with important parasitic losses taken into account.
For this, we review recently emerging flexible thermoelectric materials suited for wearable applications,
such as polymer-based materials and screen-printed paste-type inorganic materials. A few examples
among these materials are selected for thermoelectric device simulations in order to find optimal design
parameters for wearable applications. Finally we discuss the feasibility of scalable and cost-effective
manufacturing of thermoelectric energy harvesting devices with desired dimensions.
Introduction
Despite the explosive growth of wearable electronics and sensors
on the market in recent years, most of the wearable devices are
still powered by batteries that are subject to frequent recharging
and replacement.
1,2
Often these devices require energy autonomy
for an extended service time without the need for the user’s
intervention. Examples include preventive healthcare for elderly
people with wearable medical sensors that monitor the wearer’s
physiological parameters.
3
These medical sensors need to be
preferentially wireless, and operational during the patient’s daily
activities for a long time up to many years without maintenance
or the doctor’s direct assistance.
One possible solution for powering these wearable devices
without a battery is to harvest energy from the human body to
generate electricity using a thermoelectric generator. A thermo-
electric generator (TEG) is a solid-state device that can convert
heat into electricity.
4
When a TEG is attached directly onto the
skin, heat from the human body flows through the TEG due to
the temperature difference between the skin and the ambient
environment. This heat flow, or the temperature gradient, creates
a voltage in the TEG by the Seebeck effect,
5
which performs
useful work when connected to an external circuit.
Although heat dissipation from a human body largely varies
depending on the body location and surrounding conditions,
typically heat flow available from the skin under indoor sedentary
conditions is 1–10 mW cm
2
on average at 22 1C ambient tem-
perature, and a higher heat flow of 10–20 mW cm
2
is possible
on the wrist, where the heat-carrying radial artery is located near
the skin.
6,7
However, the power densities generated by the TEGs
made of state-of-the-art Bi
2
Te
3
-based inorganic materials have
been reported to be less than 60 mWcm
2
under indoor
conditions.
6–11
The limited power densities were mainly due
to the low efficiency of the materials used and the technical
difficulties in device manufacturing. Furthermore, the non-
flexibility of the inorganic materials and the expensive and
non-scalable manufacturing techniques have been major limit-
ing factors for the thermoelectric energy harvesting devices to
scale up in size, and increase the power generated.
Hence, there has recently been great interest in synthesizing
flexible thermoelectric materials with scalable approaches for
wearable energy harvesting applications. A high efficiency thermo-
electric material needs to be electrically highly conductive while
thermally poorly conductive, as represented in the material
figure of merit, ZT =S
2
sT/k, where Sis the Seebeck coefficient,
sis the electrical conductivity, Tis the absolute temperature, and
kis the thermal conductivity. The factor S
2
sin the numerator is
called the power factor. Recently, conjugated polymers have been
intensely studied for thermoelectric energy conversion because
of their intrinsically low thermal conductivities and easy
doping to achieve very high electrical conductivities, as well
as their own advantages such as flexibility, material abundance,
light-weight, and solution processability.
12,13
Although their
a
Birck Nanotechnology Center, Purdue University, West Lafayette, IN 47907, USA
b
Materials Research Laboratory, University of California, Santa Barbara,
CA 93106, USA
†Current address: Department of Mechanical and Materials Engineering, and
Department of Electrical Engineering and Computing Systems, University of
Cincinnati, Cincinnati, OH 45221, USA. E-mail: bahkjg@uc.edu
Received 4th June 2015,
Accepted 1st July 2015
DOI: 10.1039/c5tc01644d
www.rsc.org/MaterialsC
Journal of
Materials Chemistry C
APPLICATION
This journal is ©The Royal Society of Chemistry 20 15 J. Mater. Chem. C, 2015, 3,10362--10374 | 10363
ZT values are still lower than those of the inorganic thermo-
electric materials (ZT B1), there have been significant enhance-
ments in recent years in the thermoelectric performances of
organic semiconductors. ZT = 0.2–0.4 have been recently
reported for the poly(3,4-ethylenedioxythiophene) (PEDOT)
polymer system.
14,15
Polymer-based nanocomposites with carbon
nanotubes
16,17
and inorganic nano-structures
18,19
also showed
enhanced power factors. On the other hand, there have been
efforts to synthesize paste-type inorganic materials to make them
flexible and screen-printable.
20,21
In this paper, we first review the recent development
of wearable thermoelectric generators, and discuss potential
applications of these TEGs such as healthcare monitoring.
Then we review recent flexible thermoelectric materials and
the physics behind the ZT enhancement in these materials with
discussion on charge transport mechanisms. With selected
material properties, device simulations are performed to opti-
mize both the output voltage and power with device dimen-
sions and material properties. Finally, we discuss the important
recent advances in polymer material deposition and patterning
techniques that can be useful for cost-effective and scalable
manufacturing of flexible wearable thermoelectric devices.
Devices and applications
Earlier studies pioneered by IMEC, Belgium, on the develop-
ment of thermoelectric energy harvesting devices in the
past decade have successfully demonstrated the practicality of
hundreds of microwatt-level power generation from human
body heat.
6–11,22,23
Their first wearable TEGs were fabricated
in 2004 to be equipped on the wrist and serve as a power supply
for a low-power wireless sensor.
6
B250 mW power was gener-
ated from the TEG at an ambient temperature of 22 1C, but only
40% of the generated power (100 mW) was transferred to the
sensor node due to the low efficiency of the voltage boost
converter. In 2005–2006, watch-sized (6 cm
2
hot side plate with
B10 cm
2
heat sink) TEGs were fabricated to produce 200–300 mW
with a output voltage of B1.0 V.
8
The power was increased to
500–700 mW when the wearer walked slowly outdoors due to the
forced air convection on a walking person. In these TEGs,
thousands of Bi
2
Te
3
elements of 200 200 mm
2
in cross-sectional
area were used to create 4-stages of thermoelectric legs with a
total of 7 mm in thickness, in order to maintain a sufficiently
large temperature gradient across the TEG and a high open-
circuit voltage. Also, a relatively bulky fin heat sink was used to
enhance the natural air convection heat transfer at the cold
side.
8,9
This shows how important and difficult it is to maintain
a sufficiently large temperature difference across TE elements
with an unobtrusive heat sink for a thermoelectric body heat
harvester.
Later, TEGs were used as a wearable power supply for a pulse
oximeter on a finger.
24
The TEG generated power (B100 mW)
was enough to operate the pulse oximeter with an measure-
ment update every 15 s, which consumed B60 mW. About a half
of the consumed power was used for the signal processing,
and B20 mW was consumed by two LEDs, and only 3 mW was
used for the radio transmission since the signal processing was
performed on-board to minimize the amount of data trans-
mitted. A two-channel electroencephalography (EEG) system was
powered by TEGs.
25
Since the power consumption was much
higher (0.8 mW), totally 10 units of TEGs were used to increase the
surface area in contact with the forehead of the wearer (B64 cm
2
)
with 1–1.3 cm thick elements in each. The measured power at
22 1C of ambient temperature was about 2.5 mW, resulting in
30 mWcm
2
power density. Thin, light weight, modular TEGs
were integrated in a patient’s clothing to perform electrocardio-
graphy (ECG) monitoring.
6
A total of fourteen TEGs of 3 4cm
2
in size and 6.5 mm in thickness each were distributed in an
office shirt to generate totally 0.8–1 mW with 1 V of load-
matching during the wearer’s office activities while the ECG
system consumed 0.44–0.5 mW. Unlike the previous pulse
oximeter and EEG system, this ECG system had a secondary
battery that was constantly recharged during the operation to
compensate the irregular power generation from the distri-
buted TEGs. Later, 16 TEGs of 8 9mm
2
in area and 5 mm
in thickness each (with a hot plate of 3 cm in diameter and a
cold plate of 4 3cm
2
in area and 1 mm in thickness for each
TEG) were integrated into a shirt on the front side of the body,
and successfully generated 0.5–5 mW at ambient temperatures
of 27–15 1C, respectively during usual office activities.
11
If the thermoelectric generator is flexible, so that it can be
bent, and conformally wrapped onto the curved skin surface,
then the TEG may be able to utilize a much larger amount of
body heat from the enhanced contact with the skin over a larger
surface area to generate greater power. In principle, both the
output power and voltage are proportional to the surface area of
a TEG. An increased output voltage can eliminate the necessity
of a boost converter. Also, the distributed weight over a larger
area can enhance user comfort. Fig. 1 shows several types of
wearable sensor devices that are powered by flexible thermo-
electric energy harvesters. A number of small light-weight
medical sensors can be distributed on a patient’s body with
wireless communications to form a wireless body area network
(WBAN).
26,27
These devices are capable of performing health
monitoring activities such as EEG, ECG, and other vital sign
reading at various locations on the wearer’s body.
3
The collected
data are transmitted via short-range wireless communication
protocols such as Bluetooth, ANT, or Zigbee, and collected by a
personal server at a short distance such as a cell phone, which in
turn sends out the collected patient’s data through the internet
or a secure long-distance network to the health service provider
for real-time health monitoring. The short-range wireless com-
munication technologies provide efficient, ultra-low power
consumption suitable for energy harvesting devices. For example,
an experiment showed that Bluetooth Low Energy (BLE) con-
sumes less than 35 mWat3.3Vsupplywith120stransmission
intervals.
28
The physiological and physical parameters that can be non-
invasively monitored by wearable sensors for preventive healthcare
include blood pressure,
29
respiration rate,
30
oxygen saturation
(heart rate, pulse oximetry),
24,31
body temperature, and sleep
Application Journal of Materials Chemistry C
10364 |J. Mater. Chem. C, 2015, 3,10362--10374 This journal is ©The Royal Society of Chemistry 2 01 5
period (actigraphy).
32
In addition, an adhesive bandage-type
wearable sensor for monitoring electrolytes in the wearer’s
sweat has been recently developed using a battery-free passive
radio-frequency identification (RFID) and paper microfluidics
technologies.
33
Hydration and heat-stress monitoring by sensing
electrolytes such as Na
+
and K
+
in sweat are possible using the
bandage type device.
Technological advances in integrated circuits, wireless com-
munications, and medical sensors have enabled miniaturized,
light weight, ultra-low power, and wearable health monitoring
devices. Recently, an ultra-low TEG-powered body sensor node
SoC (system on a chip) has been fabricated in a commercial
130 nm CMOS technology for ECG, EEG, and EMG (electro-
myogram) applications.
34
This SoC integrates a low-voltage
boost converter, dynamic power management circuits, recon-
figurable bio-signal processing units, and a RF transmitter in a
2.5 3.3 mm
2
size chip. Only 19 mW power is consumed by the
chip for 0.013% transmission duty cycle with 14 mA current at
1.35 V. A commercially available TEG generating B60 mW with
30 mV output voltage was sufficient to power this sensor node
without a battery. The low voltage from the TEG was converted
up to 1.35 V by an efficient low-voltage boost converter with an
efficiency of 38% for the chip operation.
35
This boost converter
demonstrates a very high up-conversion efficiency greater than
60% for an input voltage as low as 50 mV and an input power as
low as 10 mW, thus well suited for low-power energy harvesting
devices.
Watkins et al.
36
suggested the use of a TEG with 70 mWor
higher power for very small temperature differences of 0.3–1.5 1C
in implanted medical devices such as pacemakers and defibril-
lators to avoid additional surgery needed to replace batteries in
these devices. Chandrakasan et al.
37
suggested to scale down
supply voltages to 0.5 V or below for micro-power systems
utilizing low-power energy harvesting technologies and dis-
cussed design challenges for the systems. Mateu et al.
38
generated
about 2 mW power using a TEG under human body heat condi-
tions and designed a power management circuit to power a
wireless communication module.
For fabrication of flexible TEGs, Weber et al.
39
sputtered
thermoelectric thin films through a shadow mask on a 1.8 m
long flexible polyimide foil substrate to form 900 pairs of trans-
verse TE elements, and coiled up the polyimide stripe to make a
coin-size TEG. By rolling up such a long stripe with a large
number of TE elements, a voltage higher than 0.8 V was
achieved for a small temperature difference like 5 1C in the
transverse direction. Kim et al.
20
printed paste-type thermo-
electric materials within holes in a polymer fabric using the
dispenser printing method, and used silver-plated conductive
fibers as electrodes connecting TE elements to fabricate highly
flexible wearable TEGs. This TEG generated 178 nW power at an
ambient temperature of 5 1C when worn on a human chest at
32 1C. More recently, Kim et al.
21
screen-printed paste-type
n-type Bi
2
Te
3
and p-type Sb
2
Te
3
materials (B500 mm thick) on a
glass fabric to fabricate wearable TEGs. Similarly screen-printed
flexible Cu electrodes were bound onto the TE materials with
the PDMS elastic polymer as a gap-filling material. 3 mW power
was generated from a small band-shape TEG made up of
11 TE element pairs on human skin at an ambient temperature
of 15 1C. The power level is expected to scale up with the
device size.
Flexible thermoelectric materials
Since the early boom of thermoelectrics research in the 1950–
1960s for applications in cooling and space missions, advances
in thermoelectric materials development had been slow and the
state-of-the-art figure of merit ZT had remained around unity
until recently because the constituting properties in ZT are
mutually coupled, i.e. there is a trade-off between the electrical
conductivity and the Seebeck coefficient in most materials.
40
Over the past few decades, nano-engineering approaches to
conventional inorganic thermoelectric materials have enabled
great enhancements in ZT.AZT of B2.2 at B900 K has been
reported recently for the spark plasma sintered Na-doped
PbTe:SrTe.
41
The ZT enhancement was attributed to the all-
scale hierarchical material structures from nano- to mesoscale
that scattered a broad range of phonon mean free paths
to significantly reduce the lattice thermal conductivity to
B0.5Wm
1
K
1
. More recently, the single-crystal SnSe has
been reported to have a ZT of B2.6 at 923 K in a crystallo-
graphic direction (b-axis) with the ultra-low lattice thermal
conductivity in that axis, B0.25 W m
1
K
1
.
42
However, later a
theoretical study
43
and experimental results on the polycrystal-
line SnSe
44,45
showed that the lattice thermal conductivity could
be higher than the reported value in ref. 42, prompting further
studies. Skutterudites and clathrates also show high ZT values
above unity with inherently low thermal conductivities in the
mid-temperature range between 600–900 K.
46,47
Fig. 1 Three types of wearable sensor nodes powered by thermoelectric
energy harvesters. The thermoelectric generators are preferably made of
flexible materials and substrates, so that they can be conformally attached
onto the various locations of the skin with enhanced thermal contact.
Monitored data are transmitted via a short-distance wireless communica-
tion protocol such as Bluetooth, ANT, or Zigbee to a portable personal
server such as a cell phone, and then to the remote healthcare service
provider via a long-distance network.
Journal of Materials Chemistry C Application
This journal is ©The Royal Society of Chemistry 20 15 J. Mater. Chem. C, 2015, 3,10362--10374 | 10365
In the low temperature range near room temperature suit-
able for wearable applications, Bi
2
Te
3
-based inorganic materials
remain the state-of-the-art thermoelectrics with high ZT values. The
nanostructured p-type Bi
0.5
Sb
1.5
Te
3
showed a ZT of B1.4 at 100 1C
due to the reduced thermal conductivity below 1.0 W m
1
K
1
by
extensive phonon scattering at interfaces, with an inherently high
powerfactorontheorderof4000mWm
1
K
2
.
48
For an n-type
material, the Bi
2
Te
3
alloy with Se (Bi
2
Se
0.3
Te
2.7
)showedaZT of
B1.0 at 125 1C.
49
At room temperature, these ZT values of both the
p-type and n-type Bi
2
Te
3
alloys slightly decreased to 1.0 and 0.8,
respectively, mainly due to the increase in lattice thermal conduc-
tivity at the lower temperature. Previously, the Bi
2
Te
3
/Sb
2
Te
3
super-
lattices
50
and PbTe/PbTeSe nano-dot superlattices
51
were reported
to show ZT B2.4 and 1.6, respectively, in the cross-plane
direction at room temperature, but these values have not been
reproduced to our best knowledge.
52
Furthermore, they were
grown by the expensive epitaxial growth techniques that are not
scalable to large-scale manufacturing.
All these aforementioned TE materials with high ZTs are
inorganic semiconductors, and thus are not flexible nor highly
cost-effective to manufacture. Recently, great attention has been
attracted to polymer-based thermoelectric materials because of
their unique advantages such as mechanical flexibility, light
weight, low-cost synthesis, and solution processability. In addi-
tion, their thermal conductivities are typically very low due to
the highly disordered structures, which is desirable for thermo-
electric applications. However, they have typically much lower
power factors than those of inorganic TE materials, which has
been the main reason that the conjugated polymers have not
been thoroughly studied for thermoelectrics thus far. Doping
mechanisms, particularly for p-type, have been relatively much
studied recently for various applications such as organic solar
cells and organic light emitting diodes.
53,54
Electrical conduc-
tivities as high as one thousand S cm
1
or even higher have
been easily achieved by oxidizing (p-type) or reducing (n-type)
the backbone polymer chains, and maintaining relatively high
mobility with high crystallinity or disordered aggregates with a
sufficiently large molecular weight.
55
The low Seebeck coeffi-
cient is the major factor limiting the power factor, and thus the
ZT value of conjugated polymers.
Over the past few years, there have been significant improve-
ments in the thermoelectric figure of merit of conjugated
polymer-based materials. Bubnova et al.
14
optimized the oxida-
tion level of poly(3,4-ethylenedioxythiophene) with tosylate
(PEDOT:Tos) using a reduction agent to achieve a large power
factor of B320 mWm
1
K
2
, which is about an order of
magnitude greater than that of freshly polymerized PEDOT:Tos,
although it is still much lower than those of the state-of-the-art
Bi
2
Te
3
-based inorganic materials B4000 mWm
1
K
2
. As a result,
ZT B0.25 was achieved with the estimated thermal conductivity
of B0.37 W m
1
K
1
. In 2013, Kim et al.
15
dedoped p-type
PEDOT with polystyrene sulfonate (PSS) in a way to minimize
the counter ion volume by partially removing unionized counter
ions, which do not contribute to the charge density, but adversely
reduce charge carrier mobility. Thus, a very high electrical
conductivity B900 S cm
1
and a reasonably high Seebeck
coefficient B72 mVK
1
were simultaneously achieved to result
in a very high power factor of B460 mWm
1
K
2
. Along with
the in-plane thermal conductivity of B0.34 W m
1
K
1
, this
power factor makes ZT B0.4 at room temperature, which is the
highest value for organic materials up to date.
However, very recently Weathers et al.
56
found that unlike
the results reported in ref. 15, the electronic thermal conduc-
tivity could be significant, even beyond the values predicted by
the Wiedemann–Franz relation, in those PEDOT:PSS samples
where the electrical conductivity was larger than B100 S cm
1
.
As a result, the total thermal conductivity was found to be as
high as 1.5 W m
1
K
1
for the PEDOT:PSS samples that have
electrical conductivity B500 S cm
1
. Weathers et al.
56
pointed
out that the electrical conductivity and the thermal conductivity
were measured from two different sets of samples in ref. 15,
which could have resulted in a large uncertainty in the ZT value.
Liu et al.
57
independently measured the in-plane thermal con-
ductivity of PEDOT:PSS using time-domain thermoreflectance
(TDTR) to find that the anisotropy in thermal conductivity can
be very large as 1 : 0.3 (in-plane : cross-plane) when the electrical
conductivity is higher than 500 S cm
1
, due to the significant
electronic contribution in the in-plane thermal conductivity,
qualitatively agreeing with Weathers et al.
56
Typically measure-
ment of in-plane thermal conductivity of organic materials is very
difficult because the substrate contribution must be removed.
For this, the variable-width 3omethod
58
is usually employed as
reported by Kim et al.,
15
where two heater lines with a large width
contrast are measured to isolate the contribution from the
in-plane thermal conduction. Another method is to remove the
substrate and have the thin film suspended between two mem-
branes todirectly measure the in-plane thermal conductivity as in
ref. 56. When the organic film to be measured can be vertically
embedded in a template, the TDTR method can be used to
measure the in-plane thermal conductivity as in ref. 57.
Recently, Bubnova et al.
59
reported a very high power factor
of B450 mWm
1
K
2
with the electrical conductivity as high as
1500 S cm
1
and the Seebeck coefficient of B55 mVK
1
for
PEDOT:Tos. The authors claimed that due to the high crystal-
linity in this polymer, a bipolaron band was created and over-
lapped with the valence band to form a semi-metallic band
structure, which breaks the trade-off between the electrical
conductivity and the Seebeck coefficient, and enhances the
two quantities simultaneously. Although a further study might
be necessary to confirm the detailed band structure with the
Fermi level position, bipolar transport effects and so on, it is
evident from this paper that the slope of the density of states
with respect to the carrier energy around the Fermi level have
been increased, which was responsible for the high Seebeck
coefficients. On the contrary, they also independently measured
PEDOT:PSS with varying doping level, but could not achieve the
Seebeck coefficient higher than B20 mVK
1
for PEDOT:PSS, in
contrast to the high values 470 mVK
1
reported in ref. 15.
There have been many other reports in recent years on the
enhanced power factors for polymer-based materials both
p-type and n-type as well as inorganic-based paste-type printable
materials. Fig. 2 summarizes the electrical conductivity,
Application Journal of Materials Chemistry C
10366 |J. Mater. Chem. C, 2015, 3,10362--10374 This journal is ©The Royal Society of Chemistry 20 15
Seebeck coefficient, and power factor of the recent key flexible
thermoelectric materials. For comparison, the properties of the
best bulk Bi
2
Te
3
alloys
48,49
are also shown in the figure. Aı
¨ch
et al.
60
studied a series of p-type poly(2,7-carbazole) derivatives
and reported a maximum power factor of 19 mWm
1
K
2
with
electrical conductivity 160 S cm
1
and Seebeck coefficient
34 mVK
1
for PCDTBT. In 2010, the Yu team reported an
enhanced power factor (B25 mWm
1
K
2
,sB400 S cm
1
,
and SB25 mVK
1
) for p-type PEDOT:PSS with carbon nanotubes
(CNTs) filling the space between polymer particles to enhance the
electrical conductivity.
16
Later the same team further optimized
PEDOT:PSS with CNTs and enhanced the power factor to
B160 mWm
1
K
2
, enhancing both sB1000 S cm
1
and
SB40 mVK
1
.
61
Culebras et al.
62
synthesized PEDOT doped
with bis(trifluoromethylsulfonyl)imide (BTFMSI) via electro-
chemical deposition to achieve a maximum power factor of
B150 mWm
1
K
2
(sB1100 S cm
1
and SB37 mVK
1
).
In order to enhance the low Seebeck coefficients of organic
semiconductors, hybrid nanocomposites with inorganic TE
materials have been synthesized. Zhang et al.
63
drop-casted
PEDOT:PSS on top of each n-type and p-type Bi
2
Te
3
film made
of ball-milled nanopowders to achieve enhancements in the
effective power factors for both types: B130 mWm
1
K
2
(sB60 S cm
1
and SB150 mVK
1
) for p-type, and
B86 mWm
1
K
2
(sB60 S cm
1
and SB120 mVK
1
) for
n-type. Although the PEDOT:PSS matrix was p-type, the influ-
ence of the n-type Bi
2
Te
3
film was significant enough to change
the carrier type overall and achieve relatively large magnitudes
of n-type Seebeck coefficients. They also tried to stir Bi
2
Te
3
nanopowders in PEDOT:PSS solution to disperse the nano-
particles in the polymer matrix, but the film was easily delami-
nated due to the large hydrophilic surface area of the Bi
2
Te
3
particles. Also, it was important to remove native oxide on Bi
2
Te
3
particles before drop-cast by dipping into diluted HCl in order to
maintain the high power factor. See et al.
18
synthesized p-type Te
nanorod-based nanocrystal films coated with PEDOT:PSS using
solution-processing. An average power factor of B50 mWm
1
K
2
with sB19 S cm
1
and SB160 mVK
1
has been achieved for
this hybrid nanocrystal. With the ultra-low thermal conductivity
around 0.22–0.3 W m
1
K
1
,aZT of B0.1 was achieved.
Du et al.
19
incorporated varying contents of exfoliated Bi
2
Te
3
nanosheets into PEDOT:PSS to optimize the power factor up to
B30 mWm
1
K
2
. Recently, Wang et al.
17
reported highly
flexible polyaniline (PANI) composites with double-walled carbon
nanotubes (DWCNTs) to have a very high optimal power factor of
B220 mWm
1
K
2
with sB610 S cm
1
and SB60 mVK
1
with 30 wt% DWCNTs. This power factor value was more than
two orders of magnitude higher than that of PANI doped with
camphorsulfonic acid (CSA) without carbon nanotubes. The
mobility of PANI-CSA was greatly enhanced from 0.15 to
7.3 cm
2
V
1
s
1
by about 50 times with the addition of DWCNTs
while the carrier concentration was dropped by a factor of
four only.
For n-type organic materials, there have been a less, but
increasing, number of reports on their thermoelectric proper-
ties than p-type materials mainly due to the difficulties in
n-type doping of organic semiconductors. In 1993, Wang et al.
64
studied potassium-doped n-type fullerenes K
x
C
70
and found that
K
4
C
70
showed sB550 S cm
1
and SB22 mVK
1
at room
temperature, which resulted in a power factor of B26 mWm
1
K
2
.
Menke et al.
65
reported the maximum power factor of
B12 mWm
1
K
2
for fullerenes C
60
doped with Cr
2
(hpp)
4
.In
2012, Sun et al.
66
investigated both n-type and p-type polymers
containing 1,1,2,2-ethenetetrathiolate (ett), poly[A
x
(M-ett)], and found
that poly[K
x
(Ni-ett)] showed the maximum n-type power factor of
B60 mWm
1
K
2
with sB40 S cm
1
and SB120 mVK
1
,
while poly[Cu
x
(Cu-ett)] showed the maximum p-type power factor
of B6.5 mWm
1
K
2
with sB10 S cm
1
and SB80 mVK
1
at
Fig. 2 (a) Electrical conductivity, (b) Seebeck coefficient, and (c) power
factor of selected key flexible thermoelectric materials, both p-type (left)
and n-type (right), measured at room temperature. Properties of bulk Bi
2
Te
3
alloys are also shown for comparison. Numbers in brackets are references.
Journal of Materials Chemistry C Application
This journal is ©The Royal Society of Chemistry 20 15 J. Mater. Chem. C, 2015, 3,10362--10374 | 10367
room temperature. However these polymers were not solution-
processable. Schlitz et al.
67
studied a high-electron mobility,
soluble, and air-stable n-type polymer, poly{N,N0-bis(2-octyl-dodecyl)-
1,4,5,8-napthalenedicarboximide-2,6-diyl-alt-5,50-(2,20-bithiophene)}
(P(NDIOD-T2)) doped with dihydro-1H-benzoimidazol-2-yl (N-DBI)
derivatives, to find that the electrical conductivity was limited
below 10
2
Scm
1
by the dopant solubility, while the Seebeck
coefficient was as high as 850 mVK
1
to achieve the maximum
power factor around 0.2–0.6 mWm
1
K
2
. Recently, Kim et al.
68
reduced the carbon nanotube films, which were originally
p-type after exposure to air, using multiple agents, polyethylene
imine (PEI) and diethylenetriamine (DETA), and subsequently
using NaBH
4
to make the films n-type. A maximum power factor
of B38 mWm
1
K
2
with sB52 S cm
1
,andSB86 mVK
1
was obtained for the reduced CNT films with transport optimiza-
tion at room temperature. The solution-processed n-type FBDPPV
doped with ((4-(1,3-dimethyl-2,3-dihydro-1H-benzoimidazol-2-yl)-
phenyl)dimethylamine) (N-DMBI) showed high electron mobility
and high doping efficiency to achieve the electrical conductivity
of B14 S cm
1
, and the power factor of B28 mWm
1
K
2
.
69
Recently the hybrid superlattices of alternating inorganic
TiS
2
monolayers and organic [(hexylammonium)
x
(H
2
O)
y
(DMSO)
z
]
layers showed a very high power factor ofB450 mWm
1
K
2
with
sB790 S cm
1
and SB78 mVK
1
at room temperature.
70
Extra electrons were injected into TiS
2
layers due to the non-
stoichiometry occurred during the single-crystal growth, and
stabilized by the intercalated organic cations, providing high
conductivity in-plane n-type transport. The in-plane thermal con-
ductivity was also measured to be extremely low B0.12 W m
1
K
1
,
so that a ZT of B0.28 has been achieved at room temperature,
which is comparable to those of the best p-type organic semi-
conductors based on PEDOTs.
In addition, there have been efforts to synthesize inorganic
materials as paste-type so they can be printed on a flexible
substrate while maintaining their high power factors. Kim
et al.
20
mixed the Bi
2
Te
3
-based powders in a ceramic binder
to make both p-type and n-type TE material pastes and printed
500 mm thick and 10 mm diameter TE elements on a flexible
polyethylene terephthalate (PET) substrate using the dispenser
printing method. The B. J. Cho team synthesized both n-type
Bi
2
Te
3
and p-type Sb
2
Te
3
pastes that are screen-printable on a
glass fabric.
21,71,72
Elemental powders in a desired atomic ratio
were mixed together with a glass powder and a binder in the
solvent in the ball miller for 24 h. The glass powder was added
to increase adhesion to the glass fabric, and the binder to
maintain sufficient viscosity of the pastes. Using these paste-
type inorganic materials, both n-type and p-type TE elements of
500 mm thickness were successfully screen printed on a glass
fabric to fabricate highly flexible and light-weight (B0.13 g cm
2
)
TEGs for wearable applications.
21
The printed n-type Bi
2
Te
3
paste
showed a power factor of B1200 mWm
1
K
2
with sB600 S cm
1
and SB140 mVK
1
, and the p-type Sb
2
Te
3
showed a similar
power factor of B1200 mWm
1
K
2
with sB1300 S cm
1
and
SB95 mVK
1
. These power factors are lower than those of the
bulk-grown inorganic counterparts by a factor of 3–4, but still
much higher than those of the best polymer-based TE materials
discussed earlier as shown in Fig. 2. The thermal conductivities
were measured to be 1.0 and 1.3 W m
1
K
1
for the n-type and
p-type pastes, respectively, which resulted in ZT B0.3 for both
materials.
Device optimization
Fig. 3 shows a schematic of a thermoelectric energy harvester
made of multiple n-type and p-type elements, and the thermal
and electric circuit models used for device simulation. Total
Npairs of n- and p-type elements are connected electrically in
series, so that the voltages induced in each element are all added
up to apply a sufficiently large voltage to the load (external
circuit). Since the temperature inside the body (B37 1C) is higher
than the ambient temperature (B22 1C), heat Qflows by con-
duction from inside the body, through the skin to the top side of
the TEG, where the TEG hot-side substrate is in contact with the
skin, and then through the TEG to the bottom, where a heat sink
dissipates the remaining heat to the ambient environment. When
it flows through the TEG, the heat is converted into electricity to
do useful work at the load. The ratio between the work done at
the load divided by the heat input Qis the efficiency of the TEG,
which is typically 0.1–0.5% for a good TEG made of the state-of-
the-art materials.
Fig. 3 (a) Schematic of a thermoelectric energy harvester, and (b) the
thermal and electric circuit models used for device performance simula-
tion. The symbol cdenotes a thermal resistance, and Rdenotes an electrical
resistance.
Application Journal of Materials Chemistry C
10368 |J. Mater. Chem. C, 2015, 3,10362--10374 This journal is ©The Royal Society of Chemistry 20 15
One can optimize the power and voltage output of a TEG for
given material properties and environmental conditions by
adequately designing the TE element dimensions, i.e. the cross-
sectional area, thickness, and the fill factor F(0 oFo1), which
is the fractional area coverage of the TE element. According to
Yazawa et al.,
73
a smaller fill factor lowers the optimum thickness
for the maximum power output. The odd fraction (1 F)inside
the TEG module that is not covered by TE elements needs to be
filled with another material called the gap filler for mechanical
support and material encapsulation, especially when the fill
factor is very small ({1). There can be radiative thermal cross
talk between the interim walls of the hot and cold side substrates,
as well as conductive heat transport across the air gap if the space
is not filled.
74
Since heat still can flow through the gap filler from
the hot side to cold as modeled as a parallel thermal resistance
c
filler
in the thermal circuit model in Fig. 3(b), the gap filler must
be a good thermal insulator to minimize the parasitic conduction
heat loss. A silica aerogel is an excellent candidate for a gap-filler
material, having extremely low thermal conductivity close to that
of air (B0.03 W m
1
K
1
) and light weight, and can be cast into a
wide range of sizes, and capable of suppressing convection with
its porous structure.
75
In steady-state, all the heat input and output are balanced
off at each node of the thermal circuit, i.e. at the two ends of a
TE element as shown in Fig. 3(b). When a TEG operates and
generates electricity allowing electric current to flow through
the circuit, Joule heating occurs inside the TE elements, at the
contacts between TE elements and electrodes due to the finite
contact resistance, as well as inside the electrodes. The Joule
heating contribution from electrodes is typically much less
than the Joule heating in the TE elements.
76
Also, we assumed
the contact resistance on the order of 10
5
to 10
6
Ohm cm
2
,
which also turns out to be much smaller than the resistances of
TE elements for the cross-sectional areas used in the simula-
tions. The Joule heating occurring inside a TE element is
divided equally to the both directions of the TE element, i.e.
Q
J,c
=Q
J,h
=1
2(I
2
R) in Fig. 3(b), where Ris the electrical resistance
of the TE element, and Iis the current. This discrete weighting
of the volume heating exactly matches the fully distributed
modeling as far as the condition is steady state and the material
of the TE element is uniform.
In addition, the Peltier effect occurs at each node, either
cooling or heating the junction depending on the current
direction, i.e. Q
P
=DSTI, where DSis the Seebeck coefficient
difference between the TE element and the electrode making a
junction at the node, and Tis the temperature at the node.
During power generation, the hot side end of a TE element is
always cooled by the Peltier effect, while the other cold end is
heated. Hence, these Peltier effects slightly reduce the tempera-
ture difference across the TE element. Usually these terminal
effects are small because the heat conduction is dominant over
the Peltier and Joule terms under the maximum power output
conditions. Nonetheless, these Peltier and Joule terms create
coupling between the thermal circuit and the electric circuit,
so the two circuits must be solved simultaneously to find the
temperatures at the nodes and the current I. Due to the coupling,
it is important to co-optimize the thermal and electrical impe-
dances of the system to achieve the maximum power output.
73
At the optimal design, DTapplied across the TE elements becomes
approximately a half of the total DT
total
(= T
s
T
a
)forlowZT
materials, i.e. ZT o1.
The TE device simulation tool used in this paper has been
published online at nanohub.org for public use.
77
This simula-
tion tool is capable of simulating a TEG system with temperature-
dependent TE material properties, which involves iterative
one-dimensional finite element simulations to numerically solve
for the temperature profile across TE elements. In wearable
applications, however, the temperature difference applied to
TE elements is small, only a few degrees, so that the assumption
of temperature-independent material properties is still reason-
able. On the other hand, temperature-dependency of material
properties must be taken into account in general cases of waste
heat recovery applications, where a large temperature difference,
100–200 1C or even larger, can be commonly applied across the
TE elements.
There have been several reports about the thermal resistance
per unit area of human body on various body locations.
7,11,78,79
Basically body thermal resistance varies widely depending
on the location on the body. For example, one of the lowest
thermal resistance values is measured on the radial artery on
the wrist to be B150 cm
2
KW
1
, and it is much higher on the
anterior leg to be B300 cm
2
KW
1
. In addition, these thermal
resistance values are a function of heat flux on the location,
decreasing with increasing heat flux,
11
which means that when
a TEG is equipped on that body location, the body thermal
resistance reduces because the TEG enhances heat conduction
on the skin. About 60 cm
2
KW
1
has been measured on the
radial artery when a TEG was worn on it with heat flow
above 100 mW cm
2
.
11
However, if the heat flow is higher than
25–30 mW cm
2
, the wearer would feel sensation of coldness,
which would cause discomfort when worn for a long time.
9
In
our simulations, we assume the body thermal resistance around
the wrist to be 200 cm
2
KW
1
, considering the relatively low heat
flow through the TEG. This value includes the interface thermal
resistance between the skin and the TEG.
At the cold side, we chose the heat sink thermal resistance
to be c
sink
= 1000 cm
2
KW
1
. Note that in ref. 8, a large-area
(B12 cm
2
), and obtrusive fin-type metallic heat sink was
attached at the cold side to enhance the heat conduction, and
its thermal resistance was estimated to be B700 cm
2
KW
1
.
It will be a challenge to develop a flexible, light-weight, and
unobtrusive heat sink having a comparable or smaller thermal
resistance than this value for flexible devices. It is important to
keep the heat sink thermal resistance as low as possible (as far
as it does not cause severe sensation of coldness) for two
reasons: first, it increases the heat input coming through the
TEG by reducing the total thermal resistance of the device for a
fixed total temperature difference available, (T
s
T
a
), which in
turn increases the power output. Second, it reduces the thick-
ness of TE elements to keep DTacross the TE elements at
optimal B1
2(T
s
T
a
), which can reduce the material cost as well
as the manufacturing cost.
Journal of Materials Chemistry C Application
This journal is ©The Royal Society of Chemistry 20 15 J. Mater. Chem. C, 2015, 3,10362--10374 | 10369
To understand the impact of the material properties, we used
two representative sets of material properties in our simulations as
showninTable1:thefirstonerepresents low electrical conducti-
vity, high Seebeck coefficient, and low thermal conductivity
material. The electrical conductivity and Seebeck coefficient values
were excerpted from the study by Zhang et al.,
63
properties for the
PEDOT:PSS polymer coated on an inorganic Bi
2
Te
3
film, while the
thermal conductivity is set to 0.3 W m
1
K
1
,atypicalvaluefor
highly disordered conjugated polymers. We call this imaginary
material the ‘‘Inorganic-polymer hybrid’’. In fact, the power factor
and the resulting ZT value for this material are much lower than
those already reported for highly conductive PEDOT-based poly-
mers. However, we wanted to keep the electrical conductivity below
100 S cm
1
since the recent papers showing the carefully measured
in-plane thermal conductivities
56,57
pointed out that the electronic
contribution to thermal conductivity could be significant for such a
highly conductive polymer. The second set of materials represents
high electrical conductivity, reduced Seebeck coefficient, and high
thermal conductivity. All the properties for the second set were
excerpted from the study by Kim et al.,
21
properties for the screen-
printed paste-type Bi
2
Te
3
and Sb
2
Te
3
materials. We call this
‘‘Screen-printed inorganic’’.
We performed device simulations with these two materials.
A wrist-band type device of 3 cm width and 15 cm length to cover
an average adult wrist perimeter was selected for the simulation,
and the cross-sectional area of each TE element was fixed to be
0.5 0.5 mm
2
. Note that a much larger cross-sectional area may
not achieve sufficient power output because DTacross the TE
elements would be too small due to the small thermal resistance
of TE elements with such a large cross-section. ThefillfactorFis
an independent variable, from which the total number of TE
element pairs was determined to be F(total module size)/
(2 (each element size)). The thickness of TE elements was also
varied as another independent variable. The thermal conductivity
of the gap filler was initially set to 0.03 W m
1
K
1
and varied later
to investigate its effect on the power output. For each calculation,
we matched the load resistance to the total electrical resistance of
the TEG for the maximum power output.
Fig. 4 shows the calculated voltage and power output from
the wrist-band TEGs as a function of TE element thickness for
several different fill factors. Overall, the power output from the
inorganic-polymer hybrid TEG is lower than that from the
screen-printed inorganic TEG because of its lower ZT (Table 1),
but a higher voltage output is achieved because of its higher
Seebeck coefficient and its lower thermal conductivity, which
created a larger DTfor a similar thickness. As shown in Fig. 4(b),
TE elements thicker than 4–5 mm are needed to obtain a power
output greater than 100 mW for the wrist-band TEG made of the
inorganic-polymer hybrid material, while only 1–2 mm thickness
is needed for the screen-printed inorganic material because of its
higher ZT. However, the screen-printed inorganic may need a
boost converter to achieve sufficiently high voltages of 1–3 V. This
could be overcome if a larger size of TEG is used because both the
voltageandpoweroutputareproportionaltothemodulesize.
Different behaviors of the two materials with varying fill
factor state an important design strategy. As one can see in
Fig. 3(b), higher fill factors (0.3 and 0.5) produce larger power
than lower fill factors for the hybrid material. This is because
when the fill factor becomes small, the parasitic heat conduc-
tion through the gap filler becomes significant, which reduces
the power output. Since the thermal conductivity of the hybrid
TE material was 0.3 W m
1
K
1
, ten times higher than that of the
gap filler (0.03 W m
1
K
1
), when the fill factor is 0.1, meaning
that only 10% of the total area is covered by the TE elements,
while 90% is covered by the gap filler, the thermal conductances
of the two parallel thermal paths become comparable with each
other, so a half of the heat energy is just lost by flowing through
the gap filler. Therefore, the power output is cut to about a half.
When the fill factor is sufficiently large, the heat loss through the
gap filler becomes negligible, so that a larger power can be
produced. On the other hand, since the printed inorganic material
has much larger thermal conductivity than that of the gap filler,
most of the heat flows through the TE elements, and the heat loss
through the gap filler is very small even for very small fill factors
like 0.05 as shown in Fig. 4(d). Therefore, the smaller the fill
factor, the larger the power output for the printed inorganic
material, due to the increased temperature difference across the
TE elements with increased thermal resistance values. Note that
there is a trade-off between the voltage output and power output
for the printed inorganic because a smaller fill factor reduces the
total number of TE elements for a fixed module size, and the total
voltage output is proportional to the number of TE elements. In
contrast, both voltage and power outputs were higher with a larger
fill factor for the hybrid material due to the reduced heat loss
through the gap filler.
We performed additional simulations with varying gap filler
thermal conductivity to see its impact on the power output of
the hybrid material. As shown in Fig. 5, a significant drop of
power output is observed when the gap filler thermal conduc-
tivity increases up to 0.1 W m
1
K
1
compared to the case with
no gap filler conduction assumed. This effect will be more
apparent when the fill factor is smaller. From this observation,
it is fair to say that having relatively high thermal conductivity
for the TE material (B1Wm
1
K
1
should be reasonably good)
Table 1 Two sets of room temperature material properties used for device simulations. The inorganic-polymer hybrid is based on the data from Zhang
et al.
63
for PEDOT:PSS on Bi
2
Te
3
film, and the screen-printed inorganic is based on the data from Kim et al.
21
Material type s(S cm
1
)S(mVK
1
)S
2
s(mWm
1
K
2
)k(W m
1
K
1
)ZT
Inorganic-polymer hybrid p-type 60 150 135 0.3 0.135
n-type 60 120 86 0.3 0.086
Screen-printed inorganic p-type 1300 95 1170 1.3 0.27
n-type 600 140 1180 1.0 0.35
Application Journal of Materials Chemistry C
10370 |J. Mater. Chem. C, 2015, 3,10362--10374 This journal is ©The Royal Society of Chemistry 20 15
and keeping the thermal conductivity of gap filler very low is
desirable to minimize the heat loss through the filler. But to
keep the power output high at the same time, a high ZT value is
necessary, meaning that a high power factor is crucial. On
another thought, as long as the heat loss through gap filler
can be kept small, a reasonably low thermal conductivity may
provide an opportunity to reduce the thickness of TEG elements,
which is desirable for cost-effective generators with reduced
material cost.
80,81
Thinner materials tend to be more flexible in
general, so further preferable in flexible electronics.
Fabrication and deposition methods
As discussed in the previous sections, fabricating flexible TEGs,
especially those capable of producing significant power on the
order of hundreds mW, is of vital importance. To realize the
Fig. 5 Power output with varying gap filler thermal conductivity for the
inorganic-polymer hybrid material (first row in Table 1) as a function of TE
element thickness. The fill factor was fixed to 0.1.
Fig. 4 Calculated (a) voltage output and (b) power output for the inorganic-polymer hybrid material (the first row in Table 1), and similarly, (c) voltage
output and (d) power output for the screen-printed inorganic material (the second row in Table 1) as a function of TE element thickness. The module size
is 3 cm 15 cm (wrist-band type), and the cross-sectional area of each TE element is 0.5 0.5 cm
2
. The number of TE elements was determined by the
fill factor. The thermal conductivity of the gap filler is 0.03 W m
1
K
1
. Note that these results are calculated for a fixed module size (3 cm 15 cm).
In principle both voltage and power outputs increase proportionally with module size.
Journal of Materials Chemistry C Application
This journal is ©The Royal Society of Chemistry 20 15 J. Mater. Chem. C, 2015, 3,10362--10374 | 10371
flexibility of a TEG, soft materials such as organics, hybrid
composites, and paste-type materials are preferred to rigid
inorganics. In addition, solution-processability is a key to scalable
and cost-effective fabrication of TEGs. In this section, we review
the techniques used to fabricate flexible thermoelectric devices,
summarize the performance of the devices reported in the
literature and discuss the potential challenges.
The techniques applied to fabricate flexible thermoelectric
devices in the literature include screen printing,
21,82–89
inkjet
printing,
20,90–95
molding,
96,97
and lithography.
14,98
Screen printing
is the most commonly used method due to its straightforward
process, which involves casting ink onto a flexible substrate
covered by a pre-patterned mask.
86
The potential of screen printing
can be magnified when coupled with the roll-to-roll (R2R) process
that is capable of continuously manufacturing meter-long
modules. Søndergaard et al.
83
developed an automatic R2R process
and screen printed as many as 18 000 TE elements composed of
PEDOT:PSS and Ag paste with an active area of 1 6cm
2
for each
pair of elements. Another interesting technique is inkjet printing,
where the original ink cartridge is replaced by the thermoelectric
material formulation, which is later dispensed onto substrates
following the patterns preset using a controlling computer.
99
The
advantages of inkjet printing include minimal human labor
requirement, high precision dispensing and little material waste.
Madan et al.
93–95
developed an epoxy embedded with percolated
Sb
2
Te
3
(p-type) or Bi
2
Te
3
(n-type) particles, which is then printed
onto flexible substrates to form tens of element pairs with planar
and circular patterns. Although printing is convenient and fast to
create complex patterns, most applications are limited to use
temperature gradient in the in-plane direction due to the micro-
meter thicknesses of the printed films. For wearable energy
harvesting, TEGs that take advantage of temperature gradient in
the cross-plane direction are suitable. Molding and lithography
provide solutions to this problem. Jo et al.
96
and Sheng et al.
97
molded PDMS membranes with arrays of cavities, which are then
filled with thermoelectric materials whose thicknesses reach milli-
meters, enough to maintain a significant temperature difference in
thecross-planedirection.Bubnovaet al.
14
applied the photolitho-
graphy technique to create cavities in a photoresist SU-850, and
then used the inkjet printing technique to dispense materials into
the cavities, which potentiates the large-scale manufacturing.
Besides the techniques reviewed above, there are also some efforts
usingfreestandingcarbonnanotube composite films that can be
easily cut into many pieces with scissors and each piece can be
usedasaTEelementinalargeTEG.
68,100,101
For example, Kim
et al.
68
made a TEG composed of stacks of 72 pieces of CNT
composite films using this method. Additionally, it is worth noting
that a handful of papers also demonstrated the potential of
vacuum deposition of pure inorganic thin films with micrometer
thicknesses to make flexible thermoelectric modules.
102–104
Table 2 A summary of materials and experimental performance results of flexible TEGs grouped by their fabrication methods
Methods
Materials
D
film
(mm) R
in
(O)DT(K) NV
oc
(mV) P
max
(mW) Ref.p-type n-type
Screen printing Sb
2
Te
3
/epoxy Bi
2
Te
3
/epoxy 500 o1 50 8 90 10.5 21*
PEDOT:PSS — 1.3 138 65 576 0.18 5.5 10
5
83*
— CNT composite 0.1 26 100 5 20 4 84
Sb
2
Te
3
/PEDOT:PSS Bi
2
Te
3
/PEDOT:PSS 40 145 50 7 85 12 85
PEDOT:PSS — 20 10 100 300 40 50 86
Sb
2
Te
3
/epoxy Bi
2
Te
3
/epoxy 65 800 20 4 25 0.19 87
CNT/polystyrene — 150 352 70 1985 305 66 88*
Sb
2
Te
3
/epoxy Bi
2
Te
3
/epoxy 60 7200 20 8 36 0.04 89
Inkjet printing Sb
2
Te
3
/epoxy Bi
2
Te
3
/epoxy 500 300 30 20 25 2 20*
poly[Cu
x
(Cu-ett)]/PVDF poly[K
x
(Ni-ett)]/PVDF 3 54 25 6 15 1 90
—Bi
2
Te
3x
Se
x
/epoxy 120 480 20 62 220 25 93
Bi
0.5
Sb
1.5
Te
3
/epoxy — 120 800 20 60 270 21 94
Bi
0.5
Sb
1.5
Te
3
/epoxy Bi/epoxy 120 100 70 10 210 130 95
Molding Bi
0.5
Sb
1.5
Te
3
/epoxy Bi
2
Te
3x
Se
x
/epoxy 4000 170 25 15 35 5 96*
Cu(I)-ett poly[K
x
(Ni-ett)] 5000 557 60 220 1510 1000 97*
Lithography PEDOT:Tos TTF-TCNQ 30 — 10 54 — 0.13 14*
Sb
2
Te
3
Bi
2
Te
3
0.7 2400 20 63 37 0.14 98
CNT composites CNT CNT/PEI — 16000 50 45 21 0.66 100
CNT/SDBS CNT/PEI 8 12 000 50 72 460 4.4 68
CNT/tpp CNT/TCNQ 80 82 20 3 6 0.11 101
Vacuum deposition (Bi
0.15
Sb
0.85
)
2
Te
3
— 200 77 34 24 130 55 102
Bi
0.4
Sb
1.6
Te
3
Bi
2
Te
2.7
Se
0.3
1 1200 130 18 600 100 103
Sb
2
Te
3
Bi
2
Te
3
16 8300 20 10 42 5.3 10
2
104
D
film
,R
in
,DT,n,V
oc
and P
max
stand for, respectively, the thickness of each element, total internal resistance of the TEG, temperature difference (the
highest achieved), number of TE element pairs, open circuit voltage, and maximum power output with matching load resistance. In the references
marked with *, the temperature gradient is in the cross-plane direction; otherwise, it is in the in-plane direction. In the materials columns, PVDF
is poly(vinylidene fluoride), TTF-TCNQ is tetrathiafulvalene-tetracyanoquinodimethane, SDBS is sodium dodecylbenzenesulfonate, and tpp is
triphenylphosphine.
Application Journal of Materials Chemistry C
10372 |J. Mater. Chem. C, 2015, 3,10362--10374 This journal is ©The Royal Society of Chemistry 20 15
Table 2 summarizes the materials and performance of flexible
TEGs reported in the literature. Intriguingly, around 60% of the
TEGs have power output larger than 1 mW with a temperature
difference of 100 K or less, some of which are enough to power a
wireless sensor,
94
an electrochromic sensor,
68
an LED
86
and even
acalculator.
97
Some of the TEGs reached a microwatt-scale power
output even at a small temperature difference of 20 K.
93,94
In
Table 2, the highest power output is 1 mW with an open circuit
voltage of 1.5 V achieved in a TEG made of 220 pairs of
ethylenetetrathiolate (ett)-based organic TE elements.
97
Herein, we observe some trade-offs and challenges for TEG
fabrication based on the data summarized in Table 2. First, a
TEG with a larger Ntends to produce better performance. It can
be understood through the formula of the open circuit voltage
(V
oc
=N(|S
p
|+|S
n
|)DT) and the maximum power output (P
max
=
N(|S
p
|+|S
n
|)
2
DT
2
/(4r
in
)), where r
in
is the resistance of a TE
element pair, that increasing Nwill improve the device performance
for a fixed DT. Aforementioned methods have good scalability and
over 100 element pairs were fabricated through screen printing
83,86,88
and molding
97
without any additional technological breakthrough. A
parameterthathasnotbeentaken into account is the internal
electrical resistance of the TEG, which includes the intrinsic resis-
tance of thermoelectricelementsaswellastheparasiticcontact
resistances and the series resistances from the electrodes. Although
some TEGs can produce a reasonable open circuit voltage, the large
internal resistance leads to a small working current, which limits the
power output. For instance, the TEGs with an internal resistance
larger than 2 kOin Table 2 have relatively small power outputs, some
of which are even under 1 mW.
68,98,100,104
Making thicker films by
printing multiple times or optimizing the geometrical ratio between
p- and n-type elements
85
can help reduce the resistance of thermo-
electric elements while the contact resistance can be minimized by
evaporating a good ohmic contact layer between TE elements and
electrodes. For example, Kim et al.
21
deposited Ni between Bi
2
Te
3
/
Sb
2
Te
3
epoxy and copper electrodes and achieved an internal
resistance less than 1 O. Furthermore, a poor thermal interface
between the thermoelectric elements and the thermal contacts can
also reduce the performance by restricting the apparent temperature
difference across the thermoelectric elements. For example, the
module composed of hundreds of PEDOT:PSS/Ag junctions made
by Søndergaard et al.
83
had an unusually small open circuit voltage
(0.18 mV at a temperature difference of 65 K in the cross-plane
direction), which is due to the major temperature drop occurred in
the 60 mm thick substrate instead of the 1.3 mmthickPEDOT:PSS
active layer. The most common materials Bi
2
Te
3
and Sb
2
Te
3
are
found to be compromised with insulating epoxy for flexibility. The
material often needs high temperature (250–350 1C) annealing,
which may require a high process cost. Development of new high
ZT organic materials with low temperature scalable processes still
remains a major challenge.
Conclusions
In this paper, we reviewed the recent advances in the flexible
thermoelectric materials and device development for wearable
energy harvesting applications. We identified various applica-
tions in health monitoring for thermoelectric energy harvesters.
Organic materials have shown great potential to be excellent
thermoelectric materials for these applications with their own
advantages in addition to very low thermal conductivity, i.e.
flexibility, light weight, material abundance, and low-cost
manufacturing. Recently their thermoelectric figures of merit
have shown significant enhancement due to the advances in
doping control, material synthesis and processing techniques.
Inorganic based flexible and printable materials have been also
developed with high ZTs. For sufficient power generation above
100 mW, a mm-level thickness of TE materials and a large device
size are required. Optimization with fill factor and gap filler
material is also essential. Finally, scalable additive manufactur-
ing such as screen printing, inkjet printing, and molding have
shown promising results for future low-cost, high performance
flexible TEG fabrication. Yet, further technological advances in
cost-effective manufacturing of flexible TE materials and
devices will be necessary to realize the first commercially
available wearable thermoelectric energy harvesters.
References
1 J.Hsu,‘‘Printed,flexible,andorganicwearablesensorsworth
$244 millions in 10 years,’’ in IEEE Spectrum, Sept. 29, 2014.
2 M. Stoppa and A. Chiolerio, Sensors, 2014, 14, 11957–11992.
3 S. Patel, H. Park, P. Bonato, L. Chan and M. Rodgers,
J. Neuroeng. Rehabil., 2012, 9, 21.
4 G. J. Snyder, ‘‘Small Thermoelectric Generators’’ Interface,
Fall, 2008, p. 54.
5Thermoelectric Handbook: Macro to Nano, ed. D. M. Rowe,
CRC Press, Boca Raton, 2005.
6 V. Leonov, in Wearable Monitoring Systems, ed. D. D. Rossi,
Springer, New York, 2010, ch. 2, pp. 27–49.
7 V. Leonov, ISRN Renewable Energy, 2011, 785380, DOI:
10.5402/2011/785380.
8 V. Leonov, T. Torfs, P. Fiorini and C. V. Hoof, IEEE Sens. J.,
2007, 7, 650–657.
9 V. Leonov and R. J. M. Vullers, J. Electron. Mater., 2009, 38,
1491–1498.
10 Z. Wang, V. Leonov, P. Fiorini and C. V. Hoof, Sens.
Actuators, A, 2009, 156, 95–102.
11 V. Leonov, IEEE Sens. J., 2013, 13, 2284–2291.
12 O. Bubnova and X. Crispin, Energy Environ. Sci., 2012, 5,
9345–9362.
13 Y. Chen, Y. Zhao and Z. Liang, Energy Environ. Sci., 2015, 8,
401–422.
14 O. Bubnova, Z. U. Khan, A. Malti, S. Braun, M. Fahlman,
M. Berggren and X. Crispin, Nat. Mater., 2011, 10, 429–433.
15 G.-H. Kim, L. Shao, K. Zhang and K. P. Pipe, Nat. Mater.,
2013, 12, 719–723.
16 D. Kim, Y. Kim, K. Choi, J. C. Grunlan and C. Yu, ACS Nano,
2010, 4, 513–523.
17 H. Wang, S.-I. Yi, X. Pu and C. Yu, ACS Appl. Mater.
Interfaces, 2015, 7, 9589–9597.
Journal of Materials Chemistry C Application
This journal is ©The Royal Society of Chemistry 20 15 J. Mater. Chem. C, 2015, 3,10362--10374 | 10373
18 K. C. See, J. P. Feser, C. E. Chen, A. Majumdar, J. J. Urban
and R. A. Segalman, Nano Lett., 2010, 10, 4664–4667.
19 Y. Du, K. F. Cai, S. Chen, P. Cizek and T. Lin, ACS Appl.
Mater. Interfaces, 2014, 6, 5735–5743.
20 M. K. Kim, M. S. Kim, S. E. Jo, H. L. Kim, S. M. Lee and
Y. J. Kim, Proc. 17th Int. Conf. Transducers & Eurosensors,
2013, pp. 1376–1379, DOI: 10.1109/Transducers.2013.6627034.
21 S. J. Kim, J. H. We and B. J. Cho, Energy Environ. Sci., 2014,
7, 1959–1965.
22 R. J. M. Vullers, R. van Schaijk, I. Doms, C. Van Hoof and
R. Mertens, Solid-State Electron., 2009, 53, 684–693.
23 V. Leonov and R. J. M. Vullers, J. Renewable Sustainable
Energy, 2009, 1, 062701.
24 T. Torfs, V. Leonov and R. J. M. Vullers, Sens. Transducers J.,
2007, 80, 1230–1238.
25 M. Van Bavel, V. Leonov, R. F. Yazicioglu, T. Torfs,
C. Van Hoof, N. E. Posthuma and R. J. M. Vullers, Sens.
Transducers J., 2008, 94, 103–115.
26 S. Ullah, H. Higgins, B. Braem, B. Latre, C. Blondia,
I. Moerman, S. Saleem, Z. Rahman and K. S. Kwak,
J. Medical Systems, 2012, 36, 1065–1094.
27 D. C. Hoang, Y. K. Tan, H. B. Chng and S. K. Panda,
Proc. Int. Conf. Power Electronics and Drive Systems, 2009,
pp. 1277–1282, DOI: 10.1109/PEDS.2009.5385814.
28 A. Dementyev, S. Hodges, S. Taylor and J. Smith, Proc. IEEE
Int. Wireless Symposium, 2013, pp. 1–4, DOI: 10.1109/IEEE-
IWS.2013.6616827.
29 H2: wearable blood pressure monitor, http://www.h2care.
com/.
30 J. H. Atkins and J. E. Mandel, Anesth. Analg., 2014, 119,
1307–1314.
31 C. M. Lochner, Y. Khan, A. Pierre and A. C. Arias, Nat.
Commun., 2014, 5, 5745.
32 S. Ancoli-Israel, R. Cole, C. Alessi, M. Chambers,
W. Moorcroft and C. P. Pollak, Sleep, 2003, 26, 342–392.
33 D. P. Rose, M. Ratterman, D. K. Griffin, L. Hou, N. Kelley-
Loughnane, R. R. Naik, J. A. Hagen, I. Papautsky and
J. Heikenfeld, IEEE Trans. Biomed. Eng.,2014,62, 1457–1465.
34 Y. Zhang, Y. Shakhsheer, J. D. Silver, A. Klinefelter,
M. Nagaraju, J. Boley, J. Pandey, A. Shrivastava, E. J.
Carlson, A. Wood, B. H. Calhoun and B. P. Otis, IEEE
J. Solid-State Circuits, 2013, 48, 199–213.
35 E. J. Carlson, K. Strunz and B. P. Otis, IEEE J. Solid-State
Circuits, 2010, 45, 741–750.
36 C. Watkins, B. Shen and R. Venkatasubramanian, Proc.
24th Int. Conf. Thermoelectrics, 2005, pp. 265–267, DOI: 10.
1109/ICT.2005.1519934.
37 A. P. Chandrakasan, D. C. Daly, J. Kwong and Y. K.
Ramadass, Proc. IEEE Symposium on VLSI Circuits, 2008,
pp. 2–5, DOI: 10.1109/VLSIC.2008.4585930.
38 L. Mateu, C. Codrea, N. Lucas, M. Pollak and P. Spies,
Proc. Int. Conf. Sensor Technologies and Applications, 2007,
pp. 366–372, DOI: 10.1109/SENSORCOMM.2007.4394949.
39 J. Weber, K. Potje-Kamloth, F. Haase, P. Detemple, F. Vo
¨lklein
and T. Doll, Sens. Actuators, A, 2006, 132, 325–330.
40 A. Majumdar, Science, 2004, 303, 777–778.
41 K. Biswas, J. He, I. D. Blum, C.-I. Wu, T. P. Hogan,
D. N. Seidman, V. P. Dravid and M. G. Kanatzidis, Nature,
2012, 489, 414–418.
42 L.-D. Zhao, S.-H. Lo, Y. Zhang, H. Sun, G. Tan, C. Uher,
C. Wolverton, V. P. Dravid and M. G. Kanatzidis, Nature,
2014, 508, 373–377.
43 J. Carrete, N. Mingo and S. Curtarolo, Appl. Phys. Lett.,
2014, 105, 101907.
44 S. Sassi, C. Candolfi, J.-B. Vaney, V. Ohorodniichuk,
P. Masschelein, A. Dauscher and B. Lenoir, Appl. Phys.
Lett., 2014, 104, 212105.
45 C.-L. Chen, H. Wang, Y.-Y. Chen, T. Day and G. J. Snyder,
J. Mater. Chem. A, 2014, 2, 11171–11176.
46 G. S. Nolas, J. Poon and M. Kanatzidis, Mater. Res. Soc.
Bull., 2006, 31, 199–205.
47 G. J. Snyder and E. S. Toberer, Nat. Mater., 2008, 7,105–114.
48 B. Poudel, Q. Hao, Y. Ma, Y. Lan, A. Minnich, B. Yu, X. Yan,
D. Wang, A. Muto, D. Vashaee, X. Chen, J. Liu, M. S.
Dresselhaus, G. Chen and Z. Ren, Science, 2008, 320,
634–638.
49 X. Yan, B. Poudel, Y. Ma, W. S. Liu, G. Joshi, H. Wang,
Y. Lan, D. Wang, G. Chen and Z. Ren, Nano Lett., 2010, 10,
3373–3378.
50 R. Venkatasubramanian, E. Siivola, T. Colpitts and
B. O’Quinn, Nature, 2001, 413, 597–602.
51 T. C. Harman, P. J. Taylor, M. P. Walsh and B. E. LaForge,
Science, 2002, 297, 2229–2232.
52 C. J. Vineis, A. Shakouri, A. Majumdar and M. G.
Kanatzidis, Adv. Mater., 2010, 22, 3970–3980.
53 Organic Photovoltaics: Materials, Device Physics, and Manu-
facturing Technologies, ed. C. Brabec, U. Scherf and
V. Dyakonov, Wiley-VCH Verlag, 2008.
54 Z. R. Li, Organic Light-Emitting Materials and Devices, CRC
Press, Boca Raton, 2nd edn, 2015.
55 R. Noriega, J. Rivnay, K. Vandewal, F. P. V. Koch,
N. Stingelin, P. Smith, M. F. Toney and A. Salleo, Nat.
Mater., 2013, 12, 1038–1044.
56 A. Weathers, Z. U. Khan, R. Brooke, D. Evans, M. T. Pettes,
J. W. Andreasen, X. Crispin and L. Shi, Adv. Mater., 2015,
27, 2101–2106.
57 J. Liu, X. Wang, D. Li, N. E. Coates, R. A. Segalman and
D. G. Cahill, Macromolecules, 2015, 48, 585–591.
58 C. Dames, Annual Review of Heat Transfer, 2013, ch. 2, vol.
16, pp. 7–49, DOI: 10.1615/AnnualRevHeatTransfer.v16.20.
59 O. Bubnova, Z. U. Khan, H. Wang, S. Braun, D. Evans,
M. Fabretto, P. Hojati-Talemi, D. Dagnelund, J.-B. Arlin,
Y. H. Geerts, S. Desbief, D. Breiby, J. Andreasen,
R. Lazzaroni, W. M. Chen, I. Zozoulenko, M. Fahlman,
P. J. Murphy, M. Berggren and X. Crispin, Nat. Mater.,
2014, 13, 190–194.
60 R. B. Aı
¨ch, N. Blouin, A. Bouchard and M. Leclerc, Chem.
Mater., 2009, 21, 751–757.
61 C. Yu, K. Choi, L. Yin and J. C. Grunlan, ACS Nano, 2011, 5,
7885–7892.
62 M. Culebras, C. M. Gomez and A. Cantarero, J. Mater.
Chem. A, 2014, 2, 10109–10115.
Application Journal of Materials Chemistry C
10374 |J. Mater. Chem. C, 2015, 3,10362--10374 This journal is ©The Royal Society of Chemistry 20 15
63 B. Zhang, J. Sun, H. E. Katz, F. Fang and R. L. Opila,
ACS Appl. Mater. Interfaces, 2010, 2, 3170–3178.
64 Z. H. Wang, K. Ichimura, M. S. Dresselhaus, G. Dresselhaus,
W.-T.Lee,K.A.WangandP.C.Eklund,Phys.Rev.B:Condens.
Matter Mater. Phys., 1993, 48, 10657–10660.
65 T. Menke, D. Ray, J. Meiss, K. Leo and M. Riede, Appl. Phys.
Lett., 2012, 100, 093304.
66 Y. Sun, P. Sheng, C. Di, F. Jiao, W. Xu, D. Qiu and D. Zhu,
Adv. Mater., 2012, 24, 932–937.
67 R. A. Schlitz, F. G. Brunetti, A. M. Glaudell, P. L. Miller,
M. A. Brady, C. J. Takacs, C. J. Hawker and M. L. Chabinyc,
Adv. Mater., 2014, 26, 2825–2830.
68 S. L. Kim, K. Choi, A. Tazebay and C. Yu, ACS Nano, 2014, 8,
2377–2386.
69 K. Shi, F. Zhang, C.-A. Di, T.-W. Yan, Y. Zou, X. Zhou,
D. Zhu, J.-Y. Wang and J. Pei, J. Am. Chem. Soc., 2015, 137,
6979–6982.
70 C. Wan, X. Gu, F. Dang, T. Itoh, Y. Wang, H. Sasaki,
M. Kondo, K. Koga, K. Yabuki, G. J. Snyder, R. Yang and
K. Koumoto, Nat. Mater., 2015, 14, 622–627.
71 J. H. We, S. J. Kim, G. S. Kim and B. J. Cho, J. Alloys Compd.,
2013, 552, 107–110.
72 S. J. Kim, J. H. We, J. S. Kim, G. S. Kim and B. J. Cho,
J. Alloys Compd., 2014, 582, 177–180.
73 K. Yazawa and A. Shakouri, Environ. Sci. Technol., 2011, 45,
7548–7553.
74 K. Yazawa and A. Shakouri, J. Mater. Res., 2012, 27, 1211–1217.
75 J. Sakamoto, J.-P. Fleurial, J. Snyder, S. Jones and T. Caillat,
NASA Tech Briefs, July 2006, NPO-40630.
76 K. Yazawa and A. Shakouri, MRS Proceedings, 2012,
1396, mrsf11-1396-o08-13, DOI: 10.1557/opl.2012.498.
77 J.-H. Bahk, K. Margatan, K. Yazawa and A. Shakouri, Advanced
thermoelectric system simulator for waste heat recovery and
energy harvesting, http://nanohub.org/tools/advte.
78 V. Leonov, Sens. Transducers J., 2011, 126, 1–10.
79 V. Leonov, J. Micromech. Microeng., 2011, 21, 125013.
80 K. Yazawa and A. Shakouri, J. Appl. Phys., 2012, 111,
024509.
81 K. Yazawa and A. Shakouri, J. Electron. Mater., 2012, 41,
1845–1850.
82 C. Navone, M. Soulier, M. Plissonnier and A. L. Seiler,
J. Electron. Mater., 2010, 39, 1755–1759.
83 R. R. Søndergaard, M. Ho
¨sel, N. Espinosa, M. Jørgensen
and F. C. Krebs, Energy Sci. Eng., 2013, 1, 81–88.
84 N. Toshima, K. Oshima, H. Anno, T. Nishinaka, S. Ichikawa,
A. Iwata and Y. Shiraishi, Adv. Mater.,2015,27, 2246–2251.
85 J. H. We, S. J. Kim and B. J. Cho, Energy, 2014, 73, 506–512.
86 Q. Wei, M. Mukaida, K. Kirihara, Y. Naitoh and T. Ishida,
RSC Adv., 2014, 4, 28802–28806.
87 C. Zhuo, E. Koukharenko, M. J. Tudor, R. N. Torah and
S. P. Beeby, J. Phys.: Conf. Ser., 2013, 476, 012031.
88 K. Suemori, S. Hoshino and T. Kamata, Appl. Phys. Lett.,
2013, 103, 153902.
89 Z. Cao, E. Koukharenko, R. N. Torah, J. Tudor and
S. P. Beeby, J. Phys.: Conf. Ser., 2014, 557, 012016.
90 F. Jiao, C.-a. Di, Y. Sun, P. Sheng, W. Xu and D. Zhu,
Philos. Trans. R. Soc., A, 2014, 372, 20130008, DOI: 10.1098/
rsta.2013.0008.
91 Z. Lu, M. Layani, X. Zhao, L. P. Tan, T. Sun, S. Fan, Q. Yan,
S. Magdassi and H. H. Hng, Small, 2014, 10, 3551–3554.
92 E. Drahi, A. Gupta, S. Blayac, S. Saunier and P. Benaben,
Phys. Status Solidi A, 2014, 211, 1301–1307.
93 D. Madan, Z. Wang, A. Chen, R.-c. Juang, J. Keist, P. K.
Wright and J. W. Evans, ACS Appl. Mater. Interfaces, 2012, 4,
6117–6124.
94 D. Madan, Z. Wang, A. Chen, P. K. Wright and J. W. Evans,
ACS Appl. Mater. Interfaces, 2013, 5, 11872–11876.
95 D. Madan, Z. Wang, A. Chen, R. Winslow, P. K. Wright and
J. W. Evans, Appl. Phys. Lett., 2014, 104, 013902.
96 S. E. Jo, M. K. Kim, M. S. Kim and Y. J. Kim, Electron. Lett.,
2012, 48, 1013–1015.
97 P. Sheng, Y. Sun, F. Jiao, C. Di, W. Xu and D. Zhu, Synth.
Met., 2014, 193, 1–7.
98 G. A. T. Sevilla, S. B. Inayat, J. P. Rojas, A. M. Hussain and
M. M. Hussain, Small, 2013, 9, 3916–3921.
99 M. A. Kamarudin, S. R. Sahamir, R. S. Datta, B. D. Long, M. F.
Mohd Sabri and S. Mohd Said, Sci. World J., 2013, 2013, 17.
100 C. A. Hewitt, D. S. Montgomery, R. L. Barbalace, R. D.
Carlson and D. L. Carroll, J. Appl. Phys., 2014, 115, 184502.
101 Y. Nonoguchi, K. Ohashi, R. Kanazawa, K. Ashiba, K. Hata,
T. Nakagawa, C. Adachi, T. Tanase and T. Kawai, Sci. Rep.,
2013, 3, 3344.
102 S. Baba, H. Sato, L. Huang, A. Uritani, R. Funahashi and
J. Akedo, J. Alloys Compd., 2014, 589, 56–60.
103 K. Kato, Y. Hatasako, M. Kashiwagi, H. Hagino, C. Adachi
and K. Miyazaki, J. Electron. Mater., 2014, 43, 1733–1739.
104 M. Mizoshiri, M. Mikami and K. Ozaki, Jpn. J. Appl. Phys.,
2013, 52, 06GL07.
Journal of Materials Chemistry C Application