Journal of Power Sources 343 (2017) 316e321

Contents lists available at ScienceDirect

Journal of Power Sources

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

An additive approach to low temperature zero pressure sintering of

bismuth antimony telluride thermoelectric materials
Glenn C. Catlin*, Rajesh Tripathi, Geoffrey Nunes Jr., Philip B. Lynch, Howard D. Jones,
Devin C. Schmitt
Materials Science, Central Research & Development, DuPont Experimental Station, 200 Powder Mill Rd., Wilmington, DE, 19083, USA

h i g h l i g h t s g r a p h i c a l a b s t r a c t

An additive-based approach for pro-

ducing screen-printable thermoelec-
trics is outlined.

Low pressure and low thermal
budget thermoelectric materials
processing.

Thermoelectric paste is used to pro-
duce printed thermoelements with a
ZT of 0.74.

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

Article history: This paper presents an additive-based approach to the formulation of thermoelectric materials suitable
Received 10 October 2016 for screen printing. Such printing processes are a likely route to such thermoelectric applications as
Received in revised form micro-generators for wireless sensor networks and medical devices, but require the development of
20 December 2016
materials that can be sintered at ambient pressure and low temperatures. Using a rapid screening
Accepted 23 December 2016
Available online 21 January 2017
process, we identify the eutectic combination of antimony and tellurium as an additive for bismuth-
antimony-telluride that enables good thermoelectric performance without a high pressure step. An
optimized composite of 15 weight percent Sb7.5Te92.5 in Bi0.5Sb1.5Te3 is scaled up and formulated into a
Keywords:
Thermoelectrics
screen-printable paste. Samples fabricated from this paste achieve a thermoelectric figure of merit (ZT) of
Power generation 0.74 using a maximum processing temperature of 748 K and a total thermal processing budget of 12 K-
Solid state cooling hours.
Printable electronics © 2017 Elsevier B.V. All rights reserved.
Composites
Seebeck coefficient

1. Introduction applications as cooled seats in luxury automobiles [2]. The history

Thermoelectric materials act as solid state transducers that can
convert the flow of thermal energy into electrical energy, or vice
versa. Use of thermoelectrics as a solid state heat pump for
refrigeration is well known [1], and has even been extended to such
* Corresponding author.
E-mail address: glenn.c.catlin@dupont.com (G.C. Catlin).
of thermoelectrics in power generation is equally long. NASA, for
example, has used radioisotope-powered thermoelectric genera-
tors to power satellites since the 1960s [3] and a similar generator is
currently powering the Curiosity rover on Mars [4]. Recently, there
has been renewed interest in thermoelectrics for waste heat re-
covery, for boosting the efficiency of microelectronics, and as power
sources for wireless sensor networks, medical devices, and wear-
able electronics [2] and [5].
These newer applications require thinner, more flexible form

http://dx.doi.org/10.1016/j.jpowsour.2016.12.092
0378-7753/© 2017 Elsevier B.V. All rights reserved.
 G.C. Catlin et al. / Journal of Power Sources 343 (2017) 316e321
factors [6] and [7], in contrast to more traditional rigid, block-like
designs. Peltier coolers with a thin form factor can offer signifi-
cant increases in cooling power over more conventional designs,
provided the thermoelectric material retains a substantial fraction
of the bulk material performance [8] and [9]. Interest in these novel
form factors is driving interest in alternative deposition techniques.
Recent work has employed dispenser [10] and [11], stencil [12],
screen [13] and [14], and spray [15] printing as the deposition
technology, with an eye to the development of flexible energy
harvesters [14] [16], and [17]. These approaches face a common
problem: the as-deposited material is a powder, often in a liquid
paste form, that must be consolidated at atmospheric pressure. In
addition, the thermal budget (processing temperature integrated
over time) needs to be compatible with flexible, i.e., polymeric
substrates. Reported consolidation methods include laser heating
of the printed material [10] and a high temperature sintering step
in which additional Te powder is introduced into the heating
chamber to limit the loss of vaporized tellurium from the Sb2Te3
thermoelectric [18]. Evans and co-workers have added selenium
[16] and additional tellurium [19] to their powders to lower the
sintering temperature and improve the electrical properties of their
dispenser-printed formulations.
In the above mentioned works, the challenge of fabricating a
consolidated material from a powder is met primarily by adjusting
the processing conditions. In the work reported here, we address
this challenge through a materials chemistry approach. In partic-
ular, we have investigated the use of low-melting point, tellurium-
based compounds as additives to Bi0.5Sb1.5Te3 (BST) thermoelectric
material. These additives act as low temperature sintering aids for
zero pressure processing of powders into consolidated thermo-
electric elements. This study was enabled by the development of a
rapid screening technique based on reusable graphite molds. The
molds allowed us to process small amounts of material directly into
bar-shaped samples suitable for electrical measurement. The
combination of sample consolidation during heating and graphite's
exfoliating properties allowed for a clean and easy sample release
after heat treatment (during the development of this screening
tool, we learned that Talapin et al. [20] developed a similar
approach for shaping thermoelectric composite materials). Using
these molds enabled us to quickly investigate a substantial number
of additives over a range of concentrations.
One of the main advantages of the paste formulation in this
work is that the paste can be used in the screen-printing process,
which is a high throughput manufacturing technique and is
commonly used in the production of photovoltaic devices.
Furthermore, the screening technique and formulation developed
in this work can be adapted for use in other additive manufacturing
techniques, such as 3D-printing and stereolithography [21]. These
manufacturing techniques enable the production of electronic el-
ements with complex geometries, which can increase the perfor-
mance and efficiency of energy devices [22].
As a result of these investigations, we were able to identify the
eutectic compound Sb7.5Te92.5 as the best candidate additive, scale
up a composite material based on that additive, and formulate a
screen-printable thermoelectric paste. Parts printed and processed
from that paste were found to have a thermoelectric figure of merit,
calculated from individual measurements of Seebeck coefficient,
thermal conductivity, and electrical conductivity, of ZT ¼ 0.74.
2. Experimental
2.1. Synthesis and sample preparation
Bismuth antimony telluride (BST) powders were generated
through mechanical alloying of elemental Bi, Sb, and Te powders in
317
a 0.5/1.5/3.01 M ratio. The ball-to-material ratio (BMR) was kept
between 2.0 and 3.5. The additives were generated through melt
synthesis in a quartz tube furnace under an argon atmosphere. As a
standard practice, the furnace was purged (by evacuation and back-
filling with process gas) a minimum of 3 times prior to each use.
The additives were then cryogenically ball-milled for particle size
reduction. Composite materials were created by cryogenic milling
BST and additive powders together in 5 mL stainless steel vials with
a BMR of 1.0 and a powder loading of 2.0 g per vial. A scaled-up
process for larger batches of composite used a 50 mL stainless
steel milling jar with a BMR of 2.0 and a powder loading of 20.0 g
per jar.
Samples of thermoelectric material were formed by heat treat-
ing the composite powders in graphite molds. The molds were
fabricated by cutting 2 mm wide, 11.2 mm long, 1.2 mm deep
channels in blocks of graphite. The molds were prepared by over-
filling each channel to form a small mound, and then hand-
compressing the mound using the flat bottom of a quartz cruci-
ble. A second iteration of this mounding/compaction process was
carried out to ensure homogeneous material distribution within
the channel. The three channels in each mold were always filled
with the same composite. The filled molds were then heat treated
in a quartz tube furnace under forming gas. All molded samples
were held at the maximum temperature indicated in Table 2 for
1 min. After heat treatment, the samples were removed by simply
inverting the molds (Fig. 1).
Baseline comparison samples were formed by hot pressing in a
standard plate press using 12.7 mm diameter stainless steel dies.
Additives were hot pressed at 120 ± 20 MPa with the platens held at
613 K for a minimum of 30 min. A BST control sample was hot
pressed in an argon atmosphere using a graphite die at a temper-
ature of 723 K for 20 min.
Screen printable pastes were formulated by blending BST (or
BST composite) powder (89.4 wt%), ethyl cellulose binder (0.7 wt%),
and Texanol™ organic vehicle (9.9 wt%) in a speedmixer. Texanol™
is an ester alcohol and has a boiling point of 254  C (at 760 mm Hg),
a molecular weight of 216.321 g mol1, and is a trademark of
Eastman Chemical Company. The resulting pastes were then
Fig. 1. (a) Graphite mold used in processing the samples. (b) Typical sample pellet after
release from the mold. (c) Arrangement of the samples in the tube furnace for heat
treatment.
318 G.C. Catlin et al. / Journal of Power Sources 343 (2017) 316e321
printed using a 60 mesh screen in 2 mm  5 mm bars for electrical
measurement and 15.2 mm squares for thermal conductivity
measurements. Printed parts were dried in air at 333 K for 30 min
prior to thermal processing.
2.2. Characterization
X-ray diffraction was performed on select powdered samples
(Supplementary Materials, Figs. 6e9) for phase identification. The
selected samples were ground using an agate mortar & pestle and
mounted on a low background silicon specimen holder as a thin
layer of powdered material mixed with a 1:2 collodion/pentyl ac-
etate solution. Cu-Ka diffraction patterns were obtained with a
Philips X’PERT Model 3040 and X'Celerator RTMS detector in
continuous scan mode. Data were collected over the range
4 <2q < 80 with an equivalent step size of 0.2 and a count time of
80 s. JADE 9.1.1 software was used for search/match in conjunction
with the ICDD PDF-4þ 2012 database. For multi-phase samples, the
weight percent of each phase was determined through least-
squares fits to the entire measured pattern.
Densities of graphite-molded materials were calculated from
the weight and external dimensions of rectangular parallelepipeds
cut from larger samples.
The powder melting points were determined by differential
scanning calorimetry (DSC) measurements with a TA Instruments
SDT Q600, scanned at a rate of 10 K per minute under flowing ni-
trogen gas (Supplementary Materials, Figs. 1e5).
2.3. Physical property measurement
Electrical conductivity and Seebeck coefficient measurements
were made on the molded samples without further processing. The
hot- and cold-pressed comparison samples were cut into long,
narrow bars using a wire saw. The electrical conductivity was
measured using a four-point probe station and a source/measure-
ment unit (SMU). The current of the SMU was ramped, and the
conductivity determined from the slope of the voltage vs. current
curve to eliminate thermoelectric effects and other DC offsets.
Seebeck coefficients were determined by electrically and thermally
anchoring the samples to a pair of Cu blocks separated by a
5e8 mm air gap; the setup is shown schematically in Fig. 2. One
block remained at room temperature while the temperature of the
other was stepped in 1 K increments from 297 K to 300 K. The
voltage across the sample was measured using a high input
impedance differential amplifier, and the Seebeck coefficient
determined from the slope of the voltage vs. temperature curve.
The thermal diffusivity a was determined using a Netzch LFA 447
Nano Flash. The heat capacity Cp was obtained using a TA In-
struments Q2000 DSC using the three run method. The thermal
conductivity k was calculated from k ¼ arCp, where the bulk density
r was determined from the sample mass and external dimensions.
Fig. 2. Schematic representation of the Seebeck coefficient measurement device.
3. Results and discussion
Additives to enable the low-temperature sintering of thermo-
electric composite materials must satisfy a number of stringent
requirements. They should themselves have a low melting tem-
perature. They must have a Seebeck coefficient of the same polarity,
and preferably similar magnitude, as the host material in order to
avoid diode-like inter particle effects. Finally, their electrical con-
ductivity should not significantly exceed that of the host material so
that, while electrically and mechanically joining the host particles,
they do not shunt the current around them.
Our approach to selecting additives for p-type bismuth anti-
mony telluride that satisfy these requirements has been to focus on
eutectics of tellurium. The use of eutectics ensures a low melting
temperature. Basing the eutectics on tellurium increases the like-
lihood of polarity matching, since Te itself is an intrinsically p-type
conductor. The candidate additives, eutectics of Te with Bi, Sb, Cu,
Ag, and Ge, are listed in Table 1, together with their melting points.
The electrical conductivities s and Seebeck coefficients S given in
the table were made on samples cut from hot-pressed pellets. For
comparison purposes, the properties of a hot pressed pellet of the
BST material are also shown.
As described in the Experimental Section, composite materials
were made by cryogenic co-milling of these additives with the BST
powder. Cryogenic milling was used to prevent alloying of the ad-
ditives with the BST. The goal was a well-mixed powder, but with
each material retaining its chemical identity. The resulting com-
posite powders were then heated in the graphite molds to form
sintered solids. In evaluating the resulting composites, we sought
materials in which the electrical conductivity was significantly
higher than that of the pure BST material, indicating good sintered
contact between the BST particles. In addition, we sought com-
posites in which the Seebeck coefficient was not significantly
reduced from the BST value.
Table 2 summarizes the electrical conductivity, Seebeck coeffi-
cient and power factor (PF) for additives at selected weight percent
loadings. Except as noted, the reported values represent the
average across the three samples from a single mold (a complete
listing can be found in the Supplementary Materials). For each
additive, the processing temperature was chosen to be approxi-
mately 25e30 K above its melting point (Supplementary Materials,
Figs. 1e4) to ensure liquid phase sintering. Where we combined
additives with different melting temperatures, the processing
temperature corresponded to the higher melting additive. All the
loading levels for a given additive were processed together in a
single run of the quartz tube furnace (see Fig. 1). Also included in
Table 2 are the results for control samples made from BST powder
without additive. Depending on the processing temperature, these
controls were either treated together with composites, as indicated
in Fig. 1, or in a separate run. For later reference, Table 2 also in-
cludes the results for a powder composed entirely of the Sb-Te
eutectic additive.
For a coarse-grained powder under low compression, inter-
particle sintering typically begins at approximately 80% of the
Table 1
Measured properties of eutectic additives.
Material Melting T (K) s (S cm1) S (mV K1)
Bi10Te90 688 126 64
Sb7.5Te92.5 690 408 135
Cu29Te71 621 1432 6
Ag33.3Te66.7 625 5 437
Ge15Te85 656 303 78
BST 883 476 236
 G.C. Catlin et al. / Journal of Power Sources 343 (2017) 316e321 319

Table 2
Measured properties of zero-pressure samples processed in graphite molds.

Additive Wt% Processing T (K) s (S cm1) S (mV K1) PF (mW cm1 K2)

Cu29Te71 5 643 375(28) 111(2) 4.6(4)

Cu29Te71 10 643 526(23) 97(9) 4.9(9)
Ag33Te67 5 643 610(32) 72(1) 3.3(2)
Ag33Te67 10 643 516(23) 60(2) 1.9(1)
Ge15Te85 5 688 129(5) 159(8) 3.3(4)
Ge15Te85 10 688 59(7) 59(3) 0.20(3)
Ge15Te85/Cu29Te71 10 688 448(35) 86(5) 3.3(5)
Ge15Te85/Ag33Te67 10 688 176(16) 59(1) 0.6(1)
Sb7.5Te92.5 20 718 276(12) 206(3) 11.7(6)
Bi10Te90 20 718 88(4) 86(11) 0.7(2)
BST control 0 643 6(10) da da
BST control 0 683 10(1) 237b 0.6(1)
BST control 0 718 97(24) 234(2) 5(1)
BST control 0 733 122(27) 234(7) 7(2)
BST control 0 748 148(65) 218(7) 7(3)
Sb7.5Te92.5 100 718 433(38) 101(1) 4.4(4)
a
The samples lacked mechanical integrity for measurement.
b
Measurement on a single sample.

melting point of a material [23], and the BST powders appear to
conform to this behavior. The electrical conductivity of the BST
control samples increases by an order of magnitude between the
683 K and 718 K processing temperatures, and the latter temper-
ature represents 81% of the BST powder's 883 K melting tempera-
ture (see Supplementary Materials, Fig. 5). We therefore conclude
that the electrical conductivity in the Cu and Ge composites arises
primarily from the melting of the additive, with little contribution
from the BST particles. Something more complex appears to govern
the conductivity in the case of the Ag composite, an aspect of the
results which we leave as the subject of future work. In the case of
the Bi and Sb containing additives, direct sintering of the BST par-
ticles becomes possible, so that the resulting material can be
modeled as a composite thermoelectric.
An examination of the results in Table 2 shows the expected
trade-off between an increase in the electrical conductivity and a
decrease in the Seebeck coefficient for the Cu and Ag based addi-
tives. The Ge and Bi based additives do not offer much of an in-
crease in the conductivity over the BST controls, but have a lower
Seebeck coefficient and a concomitantly small power factor. In an
attempt to raise the conductivity and gain back some of the loss in
Seebeck coefficient, we tested 50-50 mixtures of the Ge-Te eutectic
with the Cu-Te and Ag-Te eutectics. As can be seen from Table 2,
these mixtures did not offer a significant improvement, and the
power factor of the Ge-Te/Ag-Te mixture was particularly low.
Table 2 also shows the basic value of our approach, independent
of the performance of any particular additive. The spread in
measured conductivities, as indicated by the uncertainties in pa-
rentheses, for the BST controls is quite large, reflecting the fact that
without additives, these samples were poorly consolidated. The
samples with additive, however, show an improved conductivity
(with a few material specific exceptions) and, more importantly, a
reduced spread in the measurements.
From a practical perspective, these results show that the
Sb7.5Te92.5 eutectic offers the most promise. The power factor for
the composite is more than twice the power factor for the BST
control processed at the same temperature. For that reason, we
selected the antimony-based additive for further study as a print-
able thermoelectric composite material.
In this study, summarized in Fig. 3, we varied both the amount of
additive and the processing temperature. Fig. 3a shows that the
Seebeck coefficient is independent of the processing temperature,
but shows a steady decrease as a function the volume fraction of
additive. Also shown in the figure is the composite thermoelectric
model of Bergman and coworkers [24] and [25]. The solid line is
calculated assuming the additive properties are those of the
pressed pellet reference sample (Table 1), while the dashed line
assumes the properties of the graphite-molded material (Table 2). It
is clear that while neither case precisely captures the observed
behavior, the general downward trend with increasing additive is
not unexpected.
Fig. 3b shows the measured electrical conductivity compared
with a scaled version of the composite model. As in the case of the
Seebeck coefficient, there is no significant difference between
materials processed at different temperatures. Instead, the results
are heavily influenced by the density of the samples. The powders
are hand packed into the graphite mold and are not heated to the
melting point of the BST. Therefore, we expect that the resulting
solids are not fully dense, but rather have an open, porous structure,
and a correspondingly reduced electrical conductivity. The
measured densities, based on the sample masses and external di-
mensions, are shown in Fig. 3d. Also shown is the “density fraction,”
the ratio of the observed density to what would be expected from a
fully dense material. This density fraction has been used in Fig. 3b
to scale the conductivity calculated from the Bergman composite
model. As in Fig. 3a, the solid and dashed lies represent the results
using the additive properties from pressed pellets and molded
samples, respectively.
Although the graphite mold technique offers a direct route to
optimizing powders for zero-pressure processing, it does have a
drawback that is evident in the large uncertainties in Fig. 3b. Ex-
amination of the samples under a microscope revealed the pres-
ence of small, micron-scale cracks predominantly perpendicular to
the long axis of the samples. These cracks are therefore perpen-
dicular to the electrical current path when measuring the sample
conductivity. Although we did not attempt to quantify the rela-
tionship, we generally observed that samples with a greater degree
of cracking had a higher resistance. The origin of these cracks has
not been conclusively identified, but may be the result of thermal
stresses during sample cooling.
Fig. 3d shows the power factor computed from the data in
Fig. 3aeb, together with the results from the composite model.
Although an important result from that model is the possibility of a
maximum in the power factor as a function of composition [25] and
[26], the broad maximum seen in the figure is unrelated. Rather, it
is driven by the sample density, as discussed above. Note that the
power factor is essentially independent of the amount of additive in
the composite above an additive fraction of 15%.
320 G.C. Catlin et al. / Journal of Power Sources 343 (2017) 316e321

Fig. 3. Properties of graphite mold processed bismuth-antimony-telluride thermoelectric materials as a function of volume percent Sb7.5Te92.5 additive and processing temperature.
(a) Seebeck coefficient. (b) Electrical conductivity. (c) Power factor. (d) Density. The fully dense curve is an interpolation between the densities of Bi0.5Sb1.5Te3 and Sb7.5Te92.5. The
solid and dashed lines are the composite model of Bergman et al. [24] and [25], as described in the text.

The goal of this work is to develop a solid-state chemistry
approach to the design of screen printable thermoelectric mate-
rials. In order to validate our approach, we scaled up the composite
with 15 wt% additive and formulated it as a paste for screen
printing. The reasoning behind this choice of additive fraction was
straightforward. The experiments on the graphite-molded samples
showed no strong dependence of the power factor on the additive
concentration. By choosing the 15% additive level, we chose to keep
the Seebeck coefficient of the composite as high as possible, so that
any improvements in conductivity (achieved through better
densification of the printed parts) would result in an overall
improvement in performance.
Using a standard 60 mesh screen, we printed 2 mm by
5 mm bars of the composite paste, and a control paste formulated
from pure BST powder, onto standard alumina substrates. The wet
thickness of the printed samples was 95 mm. After a solvent-drying
step, the printed samples were heat treated under forming gas in
the tube furnace at 748 K. We used a heating profile of 10 K min1,
followed by a hold at the maximum temperature for 1 min, and a
5 K per minute cool-down to room temperature.
Electrical conductivity measurements were performed directly
on the as-printed bars. For Seebeck measurements, it was necessary
to carefully separate the bars from the alumina substrates. There
was no indication of preferred orientation in the samples; thus,
thermal conductivity was performed in the through-plane direc-
tion using the laser flash method. The results are summarized in
Table 3
Thermoelectric properties of printed and graphite mold processed BST materials.
Material s (S cm1) S (mV K1) k (W m1 K1)
BST-only 99 248 0.31
BST w/additive 332 211 0.60
Table 3, which includes data on a BST-only paste for comparison.
As we observed with our mold-processed materials, the BST
paste without additive exhibits a Seebeck coefficient similar to that
of the hot-pressed material (Table 1), but has a significantly smaller
electrical conductivity. The resulting power factor is
6.1 mW cm1 K2. For the printed material with eutectic additive,
the Seebeck coefficient is nearly identical to what was observed for
the mold processed material, but the electrical conductivity is
about 50% higher. The resulting power factor of 14.8 mW cm1 K2 is
2.4 times larger than that of the printed material without additive.
The thermal conductivity of the two printed materials also dif-
fers greatly. The thermal conductivity of the BST-only material is
0.31 W m1 K1, similar to previous reports of printed BST [10] and
[27], while the thermal conductivity of the 15 wt% additive material
is nearly doubled at 0.60 W m1 K.
The increased thermal conductivity is likely due to the additive
matrix and will be the topic of a future investigation. The dimen-
sionless thermoelectric figure of merit for the 15 wt% additive
printed material, calculated from ZT ¼ TS2 sk1 and assuming
T ¼ 300 K, is 0.74, about 25% higher than for the BST-only printed
paste.
Since sintering is an Arrhenius-type process in which the
dominant effects occur at the highest temperatures, a reasonable
estimate of the thermal budget for a manufacturing process is the
product of the maximum temperature with the time at that tem-
perature. Table 4 lists, to the best of our knowledge, the highest
reported ZT values for printed bismuth telluride-based materials
found in the literature, their maximum processing temperatures,
and the required thermal budgets. Bismuth telluride-based printed
ZT values in excess of 0.4 have been previously reported. In
ZT (300 K) particular, We et al. [28] have reported a ZT of 0.61 for screen
0.59 printed n-type Bi2Te3 consisting of Bi and Te powders, glass, organic
0.74 binder, and solvent. The reported ZT was obtained using a
 G.C. Catlin et al. / Journal of Power Sources 343 (2017) 316e321 321

Table 4 thermoelectric systems, Science 80 (321) (2008) 1457e1461.

ZT, maximum processing temperature and thermal budget of the highest performing [2] H. Zervos, P. Harrop, R. Das, Energy harvesting and storage for electronic
printed bismuth telluride-based thermoelectric materials. devices 2014-2024, IDTechEx, Rep. (2014), 1e443.
[3] C. Wood, Materials for thermoelectric energy conversion, Rep. Prog. Phys. 51
Material (Polarity) ZT Max temp (K) Thermal budget (K h) Ref. (1988) 459e539.
[4] A.D. Lalonde, Y. Pei, H. Wang, G.J. Snyder, Lead telluride alloy thermoelectrics,
Bi2Te3 (n) 0.31 623 7476 [16]
Mater. Today 14 (2011) 526e532.
Bi2Te3 (n) 0.61 773 193 [28] [5] Global Energy Harvesting Devices Market 2014-2018, Technavio, Toronto,
Bi0.4Sb1.6Te3 (p) 0.2 423 71 [27] Ontario, 2013.
Sb2Te3 (p) 0.41 350 7476 [10] [6] A. Yadav, K.P. Pipe, M. Shtein, Fiber-based flexible thermoelectric power
Sb2Te3 (p) 0.32 773 387 [18] generator, J. Power Sources 175 (2008) 909e913.
BST 0.59 748 12 a [7] L. Francioso, C. De Pascali, I. Farella, C. Martucci, P. Cretì, P. Siciliano,
BST w/additive 0.74 748 12 a A. Perrone, Flexible thermoelectric generator for ambient assisted living
a
wearable biometric sensors, J. Power Sources 196 (2011) 3239e3243.
Current work. [8] J.-P. Fleurial, G.J. Snyder, J.A. Herman, P.H. Giauque, W.M. Phillips, M.A. Ryan,
P. Shakkottai, E.A. Kolawa, M.A. Nicolet, Thick-Film thermoelectric micro-
devices, in: 18th Int. Conf. Thermoelectr, 1999, pp. 294e300.
temperature of 773 K and a thermal budget of 193 K h. By com- [9] H. Bottner, J. Nurnus, A. Schubert, Miniaturized thermoelectric converters, in:
D.M. Rowe (Ed.), Thermoelectr. Handb. Macro to Nano, CRC/Taylor & Francis,
parison, the BST with eutectic additive in this work achieves a New York, 2006.
higher material ZT with significantly reduced maximum processing [10] D. Madan, A. Chen, P.K. Wright, J.W. Evans, Dispenser printed composite
temperature and overall thermal budget, both of which are thermoelectric thick films for thermoelectric generator applications, J. Appl.
Phys. 109 (2011), 34904e1e6.
compatible with polyimide flexible substrates. It is worth empha- [11] Z. Cao, J.J. SHi, R.N. Torah, M.J. Tudor, All dispenser printed flexible 3D
sizing that we have kept the stoichiometry of the BST constant structured thermoelectric generators, J. Phys. Conf. Ser. 660 (2015) 12096.
throughout this study. Further improvements are likely possible [12] C. Navone, M. Soulier, M. Plissonnier, A.L. Seiler, Development of (Bi,Sb)
2(Te,Se)3-based thermoelectric modules by a screen-printing process,
through optimization of the Bi/Sb ratio in the BST starting material.
J. Electron. Mater 39 (2010) 1755e1759.
[13] H.B. Lee, J.H. We, H.J. Yang, K. Kim, K.C. Choi, B.J. Cho, Thermoelectric prop-
4. Conclusions erties of screen-printed ZnSb film, Thin Solid Films 519 (2011) 5441e5443.
[14] Z. Cao, E. Koukharenko, M.J. Tudor, R.N. Torah, S.P. Beeby, Screen printed
flexible Bi2Te3-Sb2Te3 based thermoelectric generator, J. Phys. Conf. Ser. 476
We have used a materials chemistry-based approach for the (2013), 12031e1e5.
development of bismuth-antimony-telluride materials that can be [15] C. Navone, M. Soulier, J. Testard, J. Simon, T. Caroff, Optimization and fabri-
formulated into pastes suitable for screen printing. The key aspect cation of a thick printed thermoelectric device, J. Electron. Mater 40 (2011)
789e793.
of our approach is the replacement of a high-pressure processing [16] D. Madan, Z. Wang, A. Chen, R. Juang, J. Keist, P.K. Wright, J.W. Evans,
step with a low melting point eutectic additive. Using a rapid Enhanced performance of dispenser printed MA n-type Bi2Te3 composite
screening process based on graphite molds, we were able to quickly thermoelectric generators, ACS Appl. Mater. Interfaces 4 (2012) 6117e6124.
[17] S.E. Jo, M. Kim, M.S. Kim, Y.J. Kim, Flexible thermoelectric generator for human
explore a range of additives, additive concentrations, and pro- body heat energy harvesting, Electron. Lett. 48 (2012) 1015e1017.
cessing temperatures. This screening work showed that for [18] S.J. Kim, J.H. We, J.S. Kim, G.S. Kim, B.J. Cho, Thermoelectric properties of P-
Bi0.5Sb1.5Te3 p-type material, the most suitable additive was type Sb2Te3 thick film processed by a screen-printing technique and a sub-
sequent annealing process, J. Alloys Compd. 582 (2014) 177e180.
Sb7.5Te92.5, which was found to be effective when added at levels as [19] D. Madan, Z. Wang, A. Chen, P.K. Wright, J.W. Evans, High-performance
low as 15 wt%. When scaled to a larger batch size and formulated dispenser printed MA p-type Bi0.5Sb1.5Te3 flexible thermoelectric generators
for screen printing, this material exhibited a power factor of for powering wireless sensor networks, ACS Appl. Mater. Interfaces 5 (2013)
14.8 mW cm1 K2 and a thermal conductivity of 0.60 W m1 K1, 11872e11876.
[20] J.S. Son, H. Zhang, J. Jang, B. Poudel, A. Waring, L. Nally, D.V. Talapin, All-
resulting in a calculated figure of merit ZT ¼ 0.74. The thermal inorganic nanocrystals as a glue for BiSbTe grains: design of interfaces in
budget for processing this material, 12 K h, is considerably below mesostructured thermoelectric materials, Angew. Chem. Int. Ed. Engl. 53
what has been previously reported in the literature. (2014) 7466e7470.
[21] M.E. Oruc, A.V. Desai, R.G. Nuzzo, P.J.A. Kenis, Design, fabrication, and char-
acterization of a proposed microchannel water electrolyzer, J. Power Sources
Acknowledgements 307 (2016) 122e128.
[22] L. Van Dijk, U.W. Paetzold, G.A. Blab, R.E.I. Schropp, M. Di Vece, 3D-printed
external light trap for solar cells, Prog. Photovolt. Res. Appl. (2016) 623e633.
X-ray analysis was performed by J. Galperin. P. Badrinarayanan, [23] F. Thummler, R. Oberacker, Introduction to Powder Metallurgy, The Institute
H. Salerno, and R. Pryor assisted with the thermal measurements. of Materials, London, 1993.
This research did not receive any specific grant from funding [24] D.J. Bergman, O. Levy, Thermoelectric properties of a composite medium,
J. Appl. Phys. 70 (1991) 6821e6833.
agencies in the public, commercial, or not-for-profit sectors. [25] D.J. Bergman, L.G. Fel, Enhancement of thermoelectric power factor in com-
posite thermoelectrics, J. Appl. Phys. 85 (1999) 8205e8216.
Appendix A. Supplementary data [26] J.P. Heremans, C.M. Jaworski, Experimental study of the thermoelectric power
factor enhancement in composites, Appl. Phys. Lett. 93 (2008), 122101e1e3.
[27] K. Kato, H. Hagino, K. Miyazaki, Fabrication of bismuth telluride thermoelec-
Supplementary data related to this article can be found at http:// tric films containing conductive polymers using a printing method, J. Electron.
dx.doi.org/10.1016/j.jpowsour.2016.12.092. Mater 42 (2013) 1313e1318.
[28] J.H. We, S.J. Kim, G.S. Kim, B.J. Cho, Improvement of thermoelectric properties
of screen-printed Bi2Te3 thick film by optimization of the annealing process,
References J. Alloys Compd. 552 (2013) 107e110.

[1] L.E. Bell, Cooling, heating, generating power, and recovering waste heat with