You are on page 1of 8

ARTICLES

PUBLISHED ONLINE: 10 JANUARY 2012 | DOI: 10.1038/NMAT3213

A new class of doped nanobulk high-figure-of-


merit thermoelectrics by scalable bottom-up
assembly
Rutvik J. Mehta1 , Yanliang Zhang2 , Chinnathambi Karthik1 , Binay Singh1 , Richard W. Siegel1 ,
Theodorian Borca-Tasciuc2 * and Ganpati Ramanath1 *

Obtaining thermoelectric materials with high figure of merit ZT is an exacting challenge because it requires the independent
control of electrical conductivity, thermal conductivity and Seebeck coefficient, which are often unfavourably coupled. Recent
works have devised strategies based on nanostructuring and alloying to address this challenge in thin films, and to obtain bulk
p-type alloys with ZT > 1. Here, we demonstrate a new class of both p- and n-type bulk nanomaterials with room-temperature
ZT as high as 1.1 using a combination of sub-atomic-per-cent doping and nanostructuring. Our nanomaterials were fabricated
by bottom-up assembly of sulphur-doped pnictogen chalcogenide nanoplates sculpted by a scalable microwave-stimulated
wet-chemical method. Bulk nanomaterials from single-component assemblies or nanoplate mixtures of different materials
exhibit 25–250% higher ZT than their non-nanostructured bulk counterparts and state-of-the-art alloys. Adapting our
synthesis and assembly approach should enable nanobulk thermoelectrics with further increases in ZT for transforming
thermoelectric refrigeration and power harvesting technologies.

H
igh-efficiency commercially viable thermoelectric materials (Figs 1 and 2). Bulk pellets were prepared by cold compaction
could revolutionize energy generation, usage and environ- and sintering of nanoplates synthesized by a scalable microwave-
mental impact. However, achieving high-ZT thermoelectric stimulated, surfactant-directed, wet-chemical approach. We show
materials has been a challenge because it requires a combina- that the surfactants serve as sulphur doping agents and oxidation
tion of low thermal conductivity κ and high power factor α 2 σ , inhibitors, in addition to guiding particle size and shape. Single-
that is, high electrical conductivity σ and Seebeck coefficient component pellets made from pnictogen chalcogenide nanoplates
α, although these properties often follow unfavourably opposing show twofold lower lattice thermal conductivity and high power
trends. To attack the problem, there is a great deal of interest factors in the 1.5–4.5 mW m−1 K−2 range, comparable to state-
in nanostructuring bismuth- and antimony-based chalcogenides, of-the-art bulk alloys1–3 . This remarkable charge-carrier-crystal
because these materials exhibit the highest room-temperature ZT and phonon-glass behaviour without alloying is made possible
in bulk1–3 and restricting their characteristic dimensions to be- by combining nanostructuring and sub-atomic-per-cent sulphur
low 10 nm offers the potential for obtaining further increases doping. We also report that multicomponent pellets fabricated
in ZT (ref. 4). Factorial increases in ZT have been shown5–7 by mixing nanoplates of different chalcogenides show ultralow
in nanostructured multilayer thin films by exploiting boundary- thermal conductivities of 0.2 < κ ≤ 0.5 Wm−1 K−1 and large Seebeck
scattering-induced κ decrease and high σ and α obtained from coefficients up to α ∼ 300 µV K−1 (Fig. 2c), and that the ZT is
quantum effects. However, this concept is yet to be successfully a strong nonlinear function of the single-component nanoplate
adapted for scalably obtaining bulk quantities of both p- and fractions (Fig. 2d). Our scalable synthesis approach thus opens up
n-type nanomaterials with high ZT. Top-down approaches to new pathways for producing bulk thermoelectric nanomaterials
nanostructuring, such as ball-milling6 and melt-spinning8 , can for transformative technologies in solid-state refrigeration and
yield bulk quantities of high-ZT p-type alloys. These processes are electricity generation from heat.
energy intensive, and adapting them for scalably obtaining high-ZT
n-type materials has been a challenge9 for bottom-up routes10–15 as Synthesis and assembly of nanoplates
well16–19 . Devising low-energy-intensity scalable methods to obtain Our bottom-up wet-chemical synthesis method uses inexpensive
high-ZT materials of both n and p type is crucial for fabricat- organic solvents and metal salts to rapidly sculpt large quantities
ing commercially viable thermoelectric devices and modules for of sulphur-doped pnictogen chalcogenide (V2 VI3 ) nanocrystals
transforming the fields of solid-state refrigeration and harvesting to create single- and multicomponent nanostructured bulk
electricity from heat. (‘nanobulk’) thermoelectric materials (Fig. 1). In this method,
Here, we report the first time realization of both p- and microwave stimulation activates the reaction between molecularly
n-type bulk nanomaterials (nanobulk) with ZT as high as 1.1 ligated chalcogen and pnictogen complexes with thioglycolic
obtained by bottom-up assembly of rapidly synthesized single- acid (TGA) in the presence of a high-boiling solvent. TGA
crystal nanoplates of sulphur-doped pnictogen chalcogenides serves as a shape-directing, oxide-inhibiting and sulphur-dopant

1 Materials
Science and Engineering Department, Rensselaer Polytechnic Institute, 110 8th St., Troy, New York 12180, USA, 2 Department of Mechanical,
Aerospace and Nuclear Engineering, Rensselaer Polytechnic Institute, 110 8th St., Troy, New York 12180, USA. *e-mail: borcat@rpi.edu; Ramanath@rpi.edu.

NATURE MATERIALS | VOL 11 | MARCH 2012 | www.nature.com/naturematerials 233


© 2012 Macmillan Publishers Limited. All rights reserved
ARTICLES NATURE MATERIALS DOI: 10.1038/NMAT3213

S
S
Doped Nanograins
nanoplates Nanopores
S

Bulk-nano
200 nm

1.2 Ours

Figure of merit ZT (300 K)


1.0 3 mm
0.8

p- and n-nano

p-nano
p- or n-bulk
0.6

n-nano
0.4

0.2
Microwave
0

Figure 1 | Schematic representation of the scalable synthesis used to obtain both n- and p-type bulk thermoelectric nanomaterials with high figures of
merit. Rapid (2–10 g min−1 ) microwave synthesis of sulphur-doped nanoplates followed by cold-compaction and sintering yields up to 92 ± 3% dense bulk
pellets with nanostructured grains. Sulphur doping provides control over the electrical conductivity, Seebeck coefficient and majority carrier type, while
nanostructuring results in very low thermal conductivity. The combined effect is applicable to multiple pnictogen chalcogenide systems and their
combinations. The graph compares the best ZT of our p- and n-type nanomaterials with those of the best p- and n-bulk materials, denoted as p- or
n-bulk1,2 , nanoparticle-dispersed n-bulk, referred to as n-nano33 , and a p-type ball-milled alloy, denoted as p-nano6 .

a b p-Bi0.5Sb1.5Te3
1.2 1.3

1.0 1.2
Figure of merit ZT (300 K)

Figure of merit ZT

0.8 p-Bi0.5Sb1.5Te3 1.1


n-Bi2Te3
1.0
0.6
0.9
0.4
n-Bi2Se3 0.8
n-Bi2Te3
0.2 p-Sb2Te3
0.7 p-Sb2Te3
0
0.6
20 40 60 80 100 120 140 160
Temperature (°C)
c d
300 p-type n-type p-type
1.2
Figure of merit ZT (300 K)

200 1.0
Seebeck coefficient α (μV K¬1)

0.8
100
0.6

0 0.4
0 20 40 60 80 100
0.2
Sb2Te3 nanoplate mole fraction
¬100
0
0 20 40 60 80 100
Sb2Te3 nanoplate mole fraction
¬200
n-type

Figure 2 | Figures of merit ZT for single-component and multicomponent bulk nanostructured pnictogen chalcogenides. a, Room-temperature ZT of the
best pellets of each pnictogen chalcogenide system. The ZT values of the corresponding bulk material of the same stoichiometry are indicated for
comparison. b, Temperature-dependent ZT of an n-type Bi2 Te3 , p-type Sb2 Te3 and p-type Bi0.5 Sb1.5 Te3 sintered pellet. c,d, Seebeck coefficient (c) and
room-temperature ZT (d) plotted as a function of Sb2 Te3 nanoplate mole fraction in Bi2 Te3 –Sb2 Te3 nanoplate mixtures. The dotted lines are to guide the
eye. The error bars denote the experimental measurement uncertainties for each sample; the spread in data points connotes sample-to-sample variations.

234 NATURE MATERIALS | VOL 11 | MARCH 2012 | www.nature.com/naturematerials


© 2012 Macmillan Publishers Limited. All rights reserved
NATURE MATERIALS DOI: 10.1038/NMAT3213 ARTICLES
a b c

2110
[01 10]
1 120 12 10
03 30
[21 10] 1120
1210
21 10
500 nm 200 nm [11 20]
[12 10]

d [11 20] e f

(0003)

200 nm 5 nm
200 nm

H
H
O H
O O O
g O h C O
C C
H C H
C 1s C H S 2s H C H
H
S
As prepared S H S

Se Bi
Annealed
As prepared
H Bi2Se3
O
Intensity (a.u.)

O
O C Annealed
Bi2Se3
H C H
C
As prepared
S Bi2Te3
O Te
Annealed
C Bi2Te3

Thioglycolic
acid

291 288 285 282 234 231 228 225 222


Binding energy (eV)

Figure 3 | Single-crystal hexagonal pnictogen chalcogenide nanoplates and mercaptan-mediated sulphur injection. a, Scanning electron micrograph
showing Bi2 Te3 nanoplates; more than 95% are hexagonal and less than 5% are triangular (see arrow). b, Bright-field TEM image of Bi2 Te3 nanoplate.
Inset: A high-resolution lattice image of the (0001) plane; the scale bar is 2 nm. c, Electron diffraction pattern from a nanoplate down the [0001] zone. The
faint spots (for an example see the arrow) are kinematically forbidden reflections arising from antisite defects. d, Bright-field TEM image from a Bi2 Se3
nanoplate with faceted islands (black arrows) that grow parallel to the parent plate face (white arrows). e, Edge-on bright-field image of a Bi2 Te3 nanoplate
showing ∼1 nm steps of (0003) planes. f, Moiré fringes and rotated spot patterns due to twist boundaries between overlapping Bi2 Se3 nanoplate crystals.
g,h, C 1s (g) and S 2s (h) core-level bands from annealed and unannealed pnictogen chalcogenide nanoplate films. A baseline spectrum from unligated
mercaptan groups is also shown for comparison.

delivery agent, whereas the use of tri-n-octylphosphine-chalcogen precipitates are truncated triangles, and the rod-like structures
complexes as microwave susceptors facilitates rapid high- seen in Fig. 3 are nanoplates oriented edge-on to the micrograph
temperature (∼180–250◦ C) liquid-phase synthesis. The synthesis plane. Increasing the microwave dose results in larger nanoplates
can be carried out in either single-mode or off-the-shelf multimode with little effect on plate thickness (Supplementary Fig. S1),
domestic microwave ovens, and yields a black precipitate in indicating that the nanoplate aspect ratio can be tuned by adjusting
∼10–15 s and completes the reaction in ∼30–60 s. the microwave dose. X-ray diffractometry and energy dispersive
The precipitates consist of 5- to 20-nm-thick hexagonal X-ray spectroscopy reveal that the black precipitates consist
nanoplates (Fig. 3a,b) with bounding edge dimensions ranging of phase-pure rhombohedral crystals of nearly stoichiometric
between 50 and 1,200 nm for all three pnictogen chalcogenides pnictogen chalcogenides doped with sulphur (Supplementary
investigated: Bi2 Te3 , Bi2 Se3 and Sb2 Te3 . Less than 5% of the Fig. S2). The near-100% yield of nanocrystals at 2–10 g min−1

NATURE MATERIALS | VOL 11 | MARCH 2012 | www.nature.com/naturematerials 235


© 2012 Macmillan Publishers Limited. All rights reserved
ARTICLES NATURE MATERIALS DOI: 10.1038/NMAT3213

indicates that our method is quick, efficient and amenable to 10–15% higher than the room-temperature values (Fig. 2b). For
scale-up and optimization. Sb2 Te3 , the ZT monotonically increases in the temperature range
Bright-field transmission electron microscopy (TEM) studies studied. The temperature-induced ZT increase in all the samples
of individual nanoplates reveal uniform contrast (Fig. 3b) and originates mainly owing to an increase in α, because σ decreases
spot diffraction patterns (Fig. 3c), indicating that each nanoplate with temperature, as expected in degenerate semiconductors,
is a single crystal with the flat faces parallel to the (0001) basal and lattice thermal conductivity κL remains virtually unchanged
plane and the bounding prismatic edges consisting of {101̄0} (Supplementary Fig. S3). The high ZT values arise from a
planes (Fig. 3b,d). Faint kinematically forbidden {101̄0} and {202̄0} unique combination of bulk-like power factor α 2 σ enabled by
diffraction spots arise from antisite defects and/or incomplete sulphur doping and very low κL due to nanostructuring, as
layers20 . Although stacking faults in the basal planes could also elaborated below.
give rise to such forbidden spots21 , no stacking faults were
observable in our microscopy studies. The relative crystal plane Sulphur doping and inhibited oxidation by thioligation
orientations with respect to the bounding edges of the nanoplates Core-level spectra acquired by X-ray photoelectron spectroscopy
imply a higher growth rate along high charge-carrier mobility (XPS) from nanoplate assemblies (Fig. 3g,h) before sintering
h112̄0i and h101̄0i in-plane directions than along low thermal reveal that TGA thioligation to the nanoplate surfaces results in
conductivity [0001] directions. Surface steps and hexagonally sulphur doping and inhibits oxidation. We observe two sulphur
faceted islands with the bounding ledges aligned to the nanoplate 2s sub-bands due to thioligated pnictogen and chalcogen states
facets (Fig. 3d–f) indicate layer-by-layer growth of the nanoplates. (Fig. 3h) that are distinct from a central 229 eV sub-band from
For example, the 1-nm-high ledges along h112̄0i parallel to the unligated TGA. The C 1s ester peaks at 288 and 286 eV (Fig. 3g),
bounding facets (Fig. 3e) correspond to (0003) planes consisting and infrared spectra showing C–S bands without detectable S–
of the five-layer stoichiometric stack of the rhombohedral crystal2 . H bands (Supplementary Fig. S4), corroborate TGA capping
The atomic-level steps also manifest themselves as bend-contour of the nanoplates. Assuming that S diffusivity in Bi2 Te3 is
discontinuities, whereas Moiré fringes and rotated spot patterns similar to that of Se (ref. 22), we expect that the S from the
(Fig. 3f) arise from twist boundaries within each plate or between adsorbed TGA is incorporated into the nanoplate crystal as
adjacently stacked plates. a substitutional impurity, thus providing a means for doping.
Such doping plays a pivotal role in determining the majority
Nanostructured bulk pellets carrier type and achieving a high power factor, as described
We fabricated single- and multicomponent bulk pellets by the later below. The retention of thioligated pnictogen and chalcogen
assembly of nanoplates of a given pnictogen chalcogenide, or signatures and the C 1s ester sub-bands, even on vacuum annealing
mixtures of nanoplates of two different pnictogen chalcogenides, at 250 ◦ C, indicates robust molecular capping, responsible for
respectively, by cold compaction and sintering 0.3–2 g of the curtailed oxidation (Fig. 3g,h and Supplementary Fig. S5). As
nanoplate powders. Our nanobulk pellets have up to 92 ± 3% a consequence, the nanoplate powders can be stored under
density of the respective bulk compounds after sintering. The ambient conditions for more than two years, and do not require
high density is attributed to the tendency of the nanoplates to oxygen-free processing.
pack efficiently through stacking. Wavelength dispersive X-ray
spectroscopy reveals that the sintered single-component pellets Thermoelectric properties of n- and p-type nanomaterials
of Bi2 Te3 are 1–2 at.% bismuth rich and Bi2 Se3 is 1–5 at.% We measured the thermoelectric properties of the nanobulk
bismuth deficient, whereas Sb2 Te3 is essentially stoichiometric, pnictogen chalcogenide pellets using well-established techniques23
all within experimental uncertainty of ±0.5 at.%. Pellets of all that were calibrated using commercial samples. Our measure-
three compounds contain 0.01–0.3 at.% sulphur (Supplementary ments were verified by independent tests on several samples in
Fig. S2), which is controllable by adjusting the synthesis conditions. collaboration with Marlow Industries and Boston College. Our
The nanoscale grain size and porosity are controlled by the initial measured values are within ±1% for α and σ , and ±5% for
nanoplate size and the compaction and sintering conditions. The κ, when compared with independent measurements. All values
average dimensions of the sintered nanograins are a factor of two to reported here are the more conservative ones. The essentially
three larger than those of the as-synthesized nanoplates. identical values (within experimental error) of α, σ and κ mea-
sured along mutually perpendicular axial and radial directions
Figure-of-merit enhancement of the same pellet, and X-ray pole figures showing a lack of
The highest room-temperature ZT we obtain is 1.1 for bulk preferred grain orientation (Supplementary Fig. S6), indicate that
nanostructured pellets of single-component n-type Bi2 Te3 and the pellets are isotropic.
multicomponent p-type Bi0.5 Sb1.5 Te pellets obtained from a 25:75
Bi2 Te3 :Sb2 Te3 mixture of nanoplates. Our measurements were Power-factor manipulation and sulphur doping
verified through independent measurements on blind and round- Our sulphur-doped single-component pellets exhibit power
robin samples with J. Sharp at Marlow Industries and Z. Ren’s factors comparable to those of bulk single crystals, which are
group at Boston College, MA. These results, summarized in Fig. 2a up to a hundred times higher than recently reported works on
and Table 1, represent a 25% improvement over conventional nanocrystalline pnictogen chalcogenide materials17–19 . This result
alloys1–3 and are comparable to those of nanostructured ball- combined with remarkably low thermal conductivity highlights
milled alloys6 . Furthermore, our studies show that the ZT values the heat-carrier glass charge-carrier crystal characteristics of
of bulk nanostructured materials of pnictogen chalcogenides, in our nanobulk materials. The electrical conductivities are in
general, are two to four times higher than their corresponding the 30 × 103 ≤ σ ≤ 250 × 103 1 m−1 range at room temperature
bulk counterparts of the same composition and stoichiometry, and (Table 1; Fig. 4a,b), and the higher-end values are ∼2.5 times
correspond to a 25–250% improvement over the highest values higher than those of state-of-art alloys1,2 . Hall measurements
reported for each material1,2 . This is perhaps best exemplified reveal carrier mobilities 60 ≤ µ ≤ 330 cm2 V−1 s−1 , corresponding
by the p-type Sb2 Te3 case, where we obtain a room-temperature to 1018 –1020 cm−3 charge-carrier concentrations (Supplementary
ZT ∼ 0.75, compared with a polycrystalline bulk ZT of less Fig. S6), which are comparable to those of single-crystal
than 0.3. For Bi2 Te3 and Bi0.5 Sb1.5 Te, the ZT increases with pnictogen chalcogenides1,2 . Our measurements reveal α < 0 for
temperature and peaks between 70 and 90 ◦ C at values that are the bismuth chalcogenides and α > 0 for antimony telluride:

236 NATURE MATERIALS | VOL 11 | MARCH 2012 | www.nature.com/naturematerials


© 2012 Macmillan Publishers Limited. All rights reserved
NATURE MATERIALS DOI: 10.1038/NMAT3213 ARTICLES
Table 1 | Thermoelectric properties and figures of merit of the best nanoplate bulk pellets measured.

Material and type σ α κ κL ZT at ZT increase


105 −1 m−1 (µV K−1 ) (W m−1 K−1 ) (W m−1 K−1 )* 300 K (%)†

n-Bi2 Te3 1.3 −185 1.23 0.59 1.1 26


n-Bi2 Te3 1.18 −140 0.80 0.38 0.87 148‡
n-Bi2 Te3 0.39 −193 0.51 0.38 0.86 146‡
p-Bi2 Te3 (no S) 0.01 40 - - - -
p-Sb2 Te3 2.5 116 1.36 0.35 0.74 51
p-Sb2 Te3 1.1 132 0.8 0.35 0.71 47
n-Bi2 Se3 0.77 −82 0.7 0.34 0.22 144
p-Bi0.5 Sb1.5 Te3 0.35 225 0.46 0.29 1.16 29
p-Bi0.5 Sb1.5 Te3 0.35 225 0.5 0.29 1.06 18

The measurement uncertainties in σ , α, κ and ZT are < ±1%, < ±1%, ∼ ±5% and ∼ ±7%, respectively, for each individual sample. All the values reported here are conservative ones, and are represented
in the graphs by the error bars. The spread in the data points connotes sample-to-sample variations. *Determined from the ordinate intercept of the plot of κ versus carrier concentration in Fig. 4a. † The
ZT increase cited is with respect to the highest ZT obtained in the pure bulk regardless of stoichiometry or anisotropy for each pnictogen chalcogenide material system1,2 . ‡ These percentage increases
are referenced to the pure bulk sample with the same stoichiometry.

a b 300
1.75 Bi2Te3

Bi2Te3 κ L = 0.59 W m¬1 K¬1 Sb2Te3


1.50 250
L Bi2Se3

1.25 Bi0.5Sb1.5Te3
κ (W m¬1 K¬1)

200
⎥α⎥ (μV K¬1)

1.00

α 2 σ (10¬3 W m¬1 K¬2)


150
5
0.75 4
Bi2Te3 κL = 0.38 W m¬1 K¬1 3
κL 100
0.50 2

Bi0.5Sb1.5Te3 κ L = 0.29 W m¬1 K¬1 1


0.25 50
0.5
0 0.5 1.0 1.5 2.0 2.5 0.5 1.0 1.5 2.0 2.5
σ (105 Ω¬1 m¬1) σ (105 Ω¬1 m¬1)
c 50 d
Undoped Bi2Te3
0 Sb2Te3

2.0
¬50
Intensity (a.u.)

α 2 σ (10¬3 W m¬1 K¬2)


α (μV K¬1)

¬100
1.3
Bi0.5Sb1.5Te3
¬150
S-doped Bi2Te3
1.8
¬200
1.3
¬250
¬0.1 0 0.1 0.2 0.3 0.4 2,460 2,470 2,480
Sulphur content (at.%) Binding energy (eV)

Figure 4 | Thermoelectric characterization of bulk-nanostructured pnictogen chalcogenides. a, Thermal conductivity κ of nanoplate pellets plotted as a
function of electrical conductivity σ . The slope is the Lorenz number L and the lattice thermal conductivity κL is the ordinate intercept. The error bars
denote the experimental uncertainties in the thermal conductivity measurements. b, Absolute value of the Seebeck coefficient α plotted as a function of σ .
Dotted curves denote different power factors α 2 σ . c, Seebeck coefficient versus the sulphur content for sulphur-doped and undoped Bi2 Te3 ; the error bars
for the sulphur content are the standard deviations for 10–12 wavelength-dispersive X-ray spectroscopy measurements from 10- to 15-µm-size regions on
each sample. d, Sulphur 1s core-level bands obtained using synchrotron radiation from sintered pellets of Sb2 Te3 and Bi0.5 Sb1.5 Te3 alloys, indicating a
correlation between high power factors and low-energy sulphur states.

−240 ≤ α ≤ −90 µV K−1 for Bi2 Te3 , −80 ≤ α ≤ −40 µV K−1 bismuth anti-site defects1,2 . This observed majority carrier reversal
for Bi2 Se3 and 105 ≤ α ≤ 135 µV K−1 for Sb2 Te3 (Fig. 4b). points to sulphur-doping-induced deviation from bulk defect
Although the Seebeck coefficient signs for Bi2 Se3 and Sb2 Te3 chemistry in Bi2 Te3 , as proposed earlier for 2 at.% sulphur
are consistent with the n- and p-type behaviours in the additions24 , except that in our case the sulphur levels are ten
respective bulk counterparts, that our Bi2 Te3 pellets show to a hundred times lower. Although sulphur doping regulates
α < 0 is significant because stoichiometric and bismuth-rich electronically active point-defect concentrations in pnictogen
Bi2 Te3 is known to be p-type owing to acceptor states from chalcogenides24,25 that can influence σ and α, further studies

NATURE MATERIALS | VOL 11 | MARCH 2012 | www.nature.com/naturematerials 237


© 2012 Macmillan Publishers Limited. All rights reserved
ARTICLES NATURE MATERIALS DOI: 10.1038/NMAT3213

are needed to reveal and exploit the nexus between processing and similar-size equiaxed grains forming high-angle boundaries
conditions, sulphur doping and dopant–defect interactions. The (Fig. 5c,d) with the stacks or nanopores (Fig. 5e), and 10–75 nm
high sensitivity of α to variations in sulphur doping (Fig. 4c) grains dispersed inside the larger grains (Fig. 5d,f inset). The grain-
suggests the alteration of density of states near the Fermi and pore-size distributions are shown in Fig. 5f. The intragranular
level, which creates attractive possibilities to further increase crystals are distinct from the 5–10 nm intragranular fringes (for
the power factor by controlling sulphur content and pnictogen example, seen in Fig. 5a) due to strain-field modulations28 . The
chalcogenide stoichiometry. plate-shaped grains are about five to ten times thicker and wider
To verify the role of sulphur doping, we synthesized Bi2 Te3 than the as-synthesized nanoplates, suggesting sintering-induced
nanoplates without TGA, by using bismuth neodecanoate—a coalescence in assemblies of nanoplate building blocks. The
liquid organic compound acting both as bismuth precursor and intragranular dispersions are probably a result of incomplete
surfactant without sulphur. This microwave synthesis results in coalescence. Although the observed stacking of plate-shaped
bismuth telluride nanoplates of equivalent size and shape to grains suggests texturing, X-ray pole figures (Supplementary
those obtained by TGA. However, the pellets fabricated from Fig. S6) show no observable preferred orientation, indicating
these nanostructures without sulphur show p-type behaviour with that our pellets consist of randomly oriented domains of
low α ∼ 40 µV K−1 and σ ∼ 103 −1 m−1 , yielding power factors plate-shaped grain stacks with submicrometre-scale texture
comparable to recent reports17–19 , but up to a hundred times restricted to each domain.
lower than our sulphur-doped pellets or state-of-the-art alloys. Simulations of κL using a modified Debye–Callaway29 model
Furthermore, core-level spectra obtained by synchrotron XPS reveal combined with a modified effective-medium formulation30
that high power factor correlates with the appearance of low- (Fig. 5g) fitted best with our experimentally determined κL for
binding-energy sulphur 1s states at ∼2,471 eV and ∼2,468 keV average grain and pore sizes of 30 nm and 10 nm, respectively,
(Fig. 4d). These results underscore the importance of sulphur consistent with the feature sizes determined by TEM (Fig. 5f). Thus,
doping for obtaining high power factors α 2 σ . By adjusting the equiaxed grains and porosity are the main contributors to the κL
TGA concentration during Bi2 Te3 synthesis, we have been able to decrease. Our simulations also reveal that phonon scattering at
obtain α 2 σ ∼ 4.5 mW m−1 K−2 , which is 5% higher than that of grain boundaries and nanopores account for about 65% of the κL
polycrystalline bulk samples1–3 . Thus, our approach offers potential decrease, with the rest attributable to ∼1020 cm−3 sulphur doping.
opportunities for Fermi-level engineering26 of nanobulk materials The weak temperature dependence of κL (Supplementary Fig. S3)
through controlled doping of the nanostructured building blocks. indicates the dominance of temperature-independent phonon
We anticipate that further increases in power factor are possible scattering at nanograin boundaries and nanopores. This inference is
by refining our synthesis and processing procedures guided by supported by our model showing less than 10% change in κL for the
the connection between α 2 σ and sulphur-doping-induced defect- entire range of sulphur doping levels explored in our experiments.
chemistry and electronic-structure changes. Our ability to decrease κL mainly by nanostructuring enables us to
increase ZT by independently tailoring α and σ through sulphur
Thermal conductivity doping or alloying.
Our pnictogen chalcogenide pellets exhibit thermal conductivities
in the 0.5 ≤ κ ≤ 1.4 W m−1 K−1 range (shown for Bi2 Te3 in Fig. 4a Multicomponent nanobulk pellets
and Table 1), representing a 40–75% diminution when compared We can also obtain κ in the 0.4–0.6 W m−1 K−1 range and ZT as high
with their respective single-crystal bulk κ values1,2,27 . We deter- as 1.1 by fabricating multicomponent pellets by mixing, compacting
mined the lattice thermal conductivity κL and Lorenz number L and sintering nanoplates of different pnictogen chalcogenides.
from the ordinate intercept and slope of the linear fit to κ–σ For example, pellets fabricated from mixtures of n-Bi2 Te3 and
plots for pellets with similar microstructures (described below) and p-Sb2 Te3 nanoplates show up to ∼40% lower κL than single-
different doping levels. The κL values are ∼60–75% lower (Fig. 4a) component pellets of either compound. The further κL decrease
than those of the bulk1–3 for the three single-component pnicto- is attributable to alloying, as supported by X-ray diffractograms
gen chalcogenides, and are similar to those of ball-milled nanos- showing Bragg-reflection shifts when compared with that of the
tructured alloys6 . The Lorenz number L = 1.66 × 10−8 W  K−2 pure pnictogen chalcogenide components (Supplementary Fig.
is ∼15% lower than the value assumed in recent reports5,8,16 , S7). Assuming that linear extrapolation is permitted, we obtain
indicating that our analysis yields a more conservative (higher) κL ∼ 0.29 W m−1 K−1 from the ordinate intercept of the κ–σ
estimate of κL because the electronic contribution of thermal plot for the Bi0.5 Sb1.5 Te3 alloy (Fig. 4a). The ultralow κL is close
conductivity κe = Lσ T , and κL = κ − κe . We note that our direct to the theoretical minimum for fully dense samples31 . Although
determination of κL and L from a κ–σ plot is possible because our pellets obtained by mixing n-Bi2 Te3 and p-Sb2 Te3 nanoplates
process enables the independent control of microstructure through decreases σ to ∼0.2–0.5 × 105 −1 m−1 , which is 40–50% lower
microwave dose and sintering treatments, and doping through than those of their single-component counterparts, α is enhanced
surfactant concentration. significantly for certain single-component nanoplate fractions
Considering that the sulphur content is less than 0.5 at.%, the (Fig. 2c). Consequently, the ZT increases substantially owing to the
low κL is largely attributable to nanostructuring, rather than to quadratic contribution of α (Fig. 2d).
alloying1–3 . This inference is corroborated by the tenfold variation in We also observe majority-carrier reversal in select component-
σ for the sulphur doping range explored compared to only a twofold fraction ranges, opening a pathway for obtaining both n- and
variation in κ due to κe increase because our pellets have nearly p-type nanomaterials with high ZT. For example, we observe n-type
identical κL values owing to similar microstructures. The strong behaviour with peak α ∼ −200 µV cm−1 at ∼12–15% mole fractions
dependence of κL on nanocrystal size enables κL tuning by adjusting of Sb2 Te3 and p-type behaviour with peak α ∼ 298 µV cm−1
the microwave dose during synthesis. For example, for Bi2 Te3 κL at ∼75–78% Sb2 Te3 (Fig. 2c), with a carrier-type crossover at
can be decreased from ∼0.59 to ∼0.38 W m−1 K−1 by decreasing the ∼18–22% Sb2 Te3 . Further studies are needed to unveil the mecha-
dose from ∼ 50–80 to 5–10 kJ g−1 . nism of the nonlinear increase in Seebeck coefficient that may arise
The κL decrease can be explained in terms of the 10–100 nm from local compositional gradients or heterostructuring32 at certain
grains and 5–50 nm inter- and intragranular pores (Fig. 5a–f) nanoplate fractions. We note that the baseline single-component
revealed by our TEM measurements. The pellets consist of pellets made with either unoptimized n-Bi2 Te3 or p-Sb2 Te3
stacks of 50- and 100-nm-thick plate-shaped grains (Fig. 5a,b), nanoplates exhibit ZT ∼ 0.5, underscoring the highly nonlinear

238 NATURE MATERIALS | VOL 11 | MARCH 2012 | www.nature.com/naturematerials


© 2012 Macmillan Publishers Limited. All rights reserved
NATURE MATERIALS DOI: 10.1038/NMAT3213 ARTICLES
a b d

c
e

[0001]

200 nm [1120]

f g 1.25
40
1.00
30

κ L (W m¬1 K¬1)
Frequency (%)

0.75 n-Bi2Te3

20
0.50

10
0.25
p-Bi0.5Sb1.5Te3
0
0
0 25 50 75 100 ¬150 ¬100 ¬50 0 50
Grain/pore size (nm) Temperature (°C)

Figure 5 | Diminution of lattice thermal conductivity in bulk-nanostructured chalcogenides due to nanoscale grain size and porosity. a–e, Bright-field
TEM micrographs from a Bi2 Te3 nanoplate pellet showing stacked plate-shaped grains, with intra-grain structural modulations observable at the
appropriate tilt angles28 with a [101̄0] zone electron diffraction from a grain (a); nanopores (red arrows) and stacked grains (b); equiaxed grains with
high-angle boundaries (c) and equiaxed grains containing smaller embedded nanograins and nanopores (d) (scale bars in b–d, 200 nm) and nanopores
(scale bar, 25 nm) (e). f, The overall grain-size (black circles) and pore-size (open green diamonds) distributions and the component distributions of
plate-shaped grains (blue circles) and equiaxed grains (red circles) measured by TEM image analyses. Inset: A bright-field TEM image of equiaxed
nanograins embedded in other grains; scale bar, 50 nm. g, Temperature dependence of the lattice thermal conductivity of the Bi2 Te3 pellets fitted to a
model combining a modified Debye–Callaway approach with a modified effective-medium formulation. The error bars denote the experimental
measurement uncertainties.

nature of ZT enhancement in pellets made with a mixture of the Methods


two types of nanoplates. Nanoplate synthesis. All chemicals were obtained from Sigma Aldrich and
used without further purification. In a typical small-scale synthesis, we added
Summary and conclusions ∼0.08 mmol of tellurium or selenium shot to 2–5 ml tri-n-octylphosphine (TOP)
and heated it in a multimode 1,250 W domestic microwave oven for 90–120 s to
We have demonstrated a readily scalable bottom-up approach to obtain a pale-yellow TOP–chalcogen complex. We added ∼0.04 mmol of BiCl3 or
create bulk thermoelectric nanomaterials of both n- and p-type SbCl3 to 2.5–10 ml of 1,5-pentanediol followed by sonication for 5 min to obtain
pnictogen chalcogenides with ZT > 1. Nanobulk thermoelectrics a pnictogen chloride solution. The bismuth chloride solution turns yellow on
were obtained by the assembly and sintering of nanoplate building addition of 100–350 µl of TGA owing to thioligated bismuth complex formation.
The solutions with TOP–chalcogen and thioligated bismuth (or antimony)
blocks synthesized by a microwave-stimulated wet-chemical were mixed and exposed to microwaves for ∼30–60 s in a multimode domestic
technique. The high power factor and low thermal conductivity microwave oven or a single-mode variable-power 300 W CEM microwave oven.
necessary for high ZT were realized through a combination of The whole procedure is simple and quick: it can be completed in ∼15 min.
nanostructuring and surfactant-induced doping. The shape- and The nanoplates were cleaned by repeated centrifugation and sonication with
size-directing surfactants used for nanoplate synthesis inhibit oxi- isopropanol and acetone, and left to dry under ambient conditions, to obtain
powders consisting of single-crystal nanoparticles.
dation and serve as a means for doping the nanoplates with sulphur.
The doping-induced alteration of the electronic structure of the Pellet fabrication. We fabricated pellets from dried nanoplate aggregates
pnictogen chalcogenides opens up an attractive new avenue for tun- through cold-compaction using a hydraulic press. The ∼60–70% dense ‘green’
ing the thermoelectric power factor and charge-carrier mobilities pellets were sintered in a 10−7 torr vacuum at 300–400 ◦ C for 60–120 min to
obtain up to ∼92 ± 3% density measured by the Archimedes method. For
by simple chemical control during scalable synthesis. Bulk pellets thermoelectric measurements, we used 6-mm-diameter cylindrical pellets
consisting of nanoscale grains obtained by assembly and sintering with 2–5 mm thickness obtained by cutting and polishing using standard
of the nanoplates result in very low thermal conductivities. Mixing metallography techniques.
doped nanoplates of different chalcogenides enables the adaptation
Material characterization. Films formed by drop-casting the nanoplates on
of this bottom-up approach to realize both n- and p-type alloys to glass slides, silicon wafer or TEM grids, and green and sintered pellets were
and nanocomposites with ZT > 1. This unique set of features of our characterized by X-ray diffractometry, scanning and transmission electron
scalable bottom-up nanostructure synthesis and assembly approach microscopy and diffraction, and XPS. Fourier transform infrared spectra were
makes it attractive for the manufacture of high-performance acquired from KBr-compacted pellets with the nanoplates. We measured the
thermoelectric devices and modules. Further understanding and sulphur content of the pellets by wavelength dispersive X-ray spectroscopy on an
electron microprobe. The reported values constitute an average of at least 10–12
refinement of our approach holds promise for development and use measurements from 10- to 15-µm-sized regions. XPS spectra were acquired either
of high-ZT nanobulk thermoelectrics for transformative solid-state with a Mg Kα beam or by using a 3.028 keV synchrotron beam at Brookhaven
refrigeration technologies and thermoelectric power harvesting. National Laboratory.

NATURE MATERIALS | VOL 11 | MARCH 2012 | www.nature.com/naturematerials 239


© 2012 Macmillan Publishers Limited. All rights reserved
ARTICLES NATURE MATERIALS DOI: 10.1038/NMAT3213

Thermoelectric transport characterization. The Seebeck coefficient α and 15. Yoo, B. et al. Electrodeposition of thermoelectric superlattice nanowires.
the thermal conductivity κ were simultaneously measured using a steady-state Adv. Mater. 19, 296–299 (2007).
instrumental setup23 (see Supplementary Fig. S8 for further details) calibrated 16. Scheele, M. et al. ZT enhancement in solution-grown
with commercial thermoelectric samples. The electrical conductivities σ of the Sb(2−x) Bix Te3 nanoplatelets. ACS Nano 4, 4283–4291 (2010).
pellets were determined by four-probe switching methods using a.c. (6 kHz) 17. Zhao, Y. X., Dyck, J. S., Hernandez, B. M. & Burda, C. Enhancing
and d.c. techniques, van der Pauw for radial, and linear four-probe methods thermoelectric performance of ternary nanocrystals through adjusting
for axial directions. All properties were measured on the same sample in two carrier concentration. J. Am. Chem. Soc. 132, 4982–4983 (2010).
mutually perpendicular, axial and radial pellet directions, and independently 18. Scheele, M. et al. Synthesis and thermoelectric characterization of Bi2 Te3
verified by J. Sharp at Marlow Industries and Z. Ren’s group at Boston College. nanoparticles. Adv. Funct. Mater. 19, 3476–3483 (2009).
Verification experiments included measurements of blind round-robin samples 19. Dirmyer, M. R., Martin, J., Nolas, G. S., Sen, A. & Badding, J. V. Thermal and
and our nanostructured bulk samples by linear-probe (σ and α) and laser-flash electrical conductivity of size-tuned bismuth telluride nanoparticles. Small 5,
(κ) techniques. The measurement uncertainties for σ are less than ±1% (van der 933–937 (2009).
Pauw) and ∼ ± 3% (linear four probe), less than ±1% for α and ∼ ± 5% for κ, 20. Cullity, B. D. & Stock, S. R. Elements of X-Ray Diffraction 3rd edn
for each individual sample, and are represented in the graphs by the error bars (Prentice, 2001).
on the data points. The spread in the data points connotes sample-to-sample 21. Reyes-Gasga, J., Gomez-Rodriguez, A., Gao, X. X. & Jose-Yacaman, M.
variations. All values reported in the text are conservative ones, that is, the ones On the interpretation of the forbidden spots observed in the electron
that yield the lower ZT. diffraction patterns of flat Au triangular nanoparticles. Ultramicroscopy 108,
We determined α by measuring the temperature difference and Seebeck 929–936 (2008).
voltage across the thickness of the pellets using E-type thermocouples and copper 22. Chitroub, M., Scherrer, S. & Scherrer, H. Anisotropy of the selenium diffusion
wires (Supplementary Fig. S8). We polished the pellets and coated the top and coefficient in bismuth telluride. J. Phys. Chem. Solids 61, 1693–1701 (2000).
bottom surfaces with 150-µm-thick indium layers to minimize the contact 23. Pal, S. K. et al. Thermal and electrical transport along MWCNT arrays grown
resistances Rc (kept at less than 3% of the measured thermal resistance R) and on Inconel substrates. J. Mater. Res. 23, 2099–2105 (2008).
uniformly spread the heat flux. The thermocouple and copper wires, all of 25 µm 24. Horak, J., Lostak, P., Koudelka, L. & Novotny, R. Inversion of conductivity
diameter, were embedded in the indium layers. The lower indium piece is in type in Bi2 Te3−x Sx crystals. Solid State Commun. 55, 1031–1034 (1985).
contact with the heat sink whereas the upper piece is in contact with a heater and 25. Chizhevskaya, S. N. & Shelimova, L. E. Electroactive and electroinactive
the sample is mechanically pressed between the two layers. To minimize heat dopants in Bi2 Te3 and their interaction with antisite defects. Inorg. Mater. 31,
losses, we placed a thick Teflon block above the heater and gently fastened the 1083–1095 (1995).
system before carrying out κ measurements in a 10−7 torr vacuum. With respect 26. Heremans, J. P. et al. Enhancement of thermoelectric efficiency in PbTe by
to the total heater power, convection losses were less than 0.1% in the 10−7 torr distortion of the electronic density of states. Science 321, 554–557 (2008).
vacuum and conduction losses were restricted to less than 1% through the use 27. Navratil, J. et al. Conduction band splitting and transport properties of Bi2 Se3 .
of fine thermocouples and copper wires, and Teflon insulation of the heater. We J. Solid State Chem. 177, 1704–1712 (2004).
determined the total heat losses from conduction and radiation, and accounted for 28. Peranio, N. & Eibl, O. Structural modulations in Bi2 Te3 . J. Appl. Phys. 103,
them in our data reduction. We extracted the pellet thermal conductivity κ from R. 024314 (2008).
For more details of the synthesis procedures and measurement techniques, please 29. Morelli, D. T., Heremans, J. P. & Slack, G. A. Estimation of the isotope effect on
refer to the sections in Supplementary Information. the lattice thermal conductivity of group IV and group III–V semiconductors.
Phys. Rev. B 66, 195304 (2002).
Received 10 May 2011; accepted 23 November 2011; 30. Minnich, A. & Chen, G. Modified effective medium formulation for the
published online 10 January 2012 thermal conductivity of nanocomposites. Appl. Phys. Lett. 91, 073105 (2007).
31. Vineis, C. J., Shakouri, A., Majumdar, A. & Kanatzidis, M. G. Nanostructured
References thermoelectrics: Big efficiency gains from small features. Adv. Mater. 22,
1. Rowe, D. M. (ed.) Thermoelectrics Handbook: Macro to Nano (CRC, 2005). 3970–3980 (2010).
2. Rowe, D. M. (ed.) CRC Handbook of Thermoelectrics (CRC, 1995). 32. Mehta, R. J. et al. Seebeck tuning in chalcogenide nanoplate assemblies by
3. Nolas, G. S., Sharp, J. & Goldsmid, H. J. Thermoelectrics: Basic Principles and nanoscale heterostructuring. ACS Nano 4, 5055–5060 (2010).
New Materials Developments (Springer, 2001). 33. Zhao, X. B. et al. Bismuth telluride nanotubes and the effects on the
4. Dresselhaus, M. S. et al. New directions for low-dimensional thermoelectric thermoelectric properties of nanotube-containing nanocomposites.
materials. Adv. Mater. 19, 1043–1053 (2007). Appl. Phys. Lett. 86, 062111 (2005).
5. Minnich, A. J., Dresselhaus, M. S., Ren, Z. F. & Chen, G. Bulk nanostructured
thermoelectric materials: Current research and future prospects. Acknowledgements
Energ. Environ. Sci. 2, 466–479 (2009). We gratefully acknowledge funding from the US Department of Energy, Office
6. Poudel, B. et al. High-thermoelectric performance of nanostructured bismuth of Basic Energy Sciences through the S3 TEC Energy Frontiers Research Center
antimony telluride bulk alloys. Science 320, 634–638 (2008). at MIT under Award DE-SC0001299, National Science Foundation grants DMR
7. Venkatasubramanian, R., Siivola, E., Colpitts, T. & O’Quinn, B. Thin-film 0519081, ECCS 1002282 and CBET 0348613, and a gift grant from IBM through the
thermoelectric devices with high room-temperature figures of merit. Nature Rensselaer Nanotechnology Center. We thank J. Woicik, B. Karlin and D. Fischer
413, 597–602 (2001). for help with setting up the photoemission experiments carried out at the National
8. Xie, W. J., Tang, X. F., Yan, Y. G., Zhang, Q. J. & Tritt, T. M. High Synchrotron Light Source at Brookhaven National Laboratory, supported under the US
thermoelectric performance BiSbTe alloy with unique low-dimensional Department of Energy contract DE-AC02-98CH10886. We thank J. Sharp at Marlow
structure. J. Appl. Phys. 105, 113713 (2009). Industries and Z. Ren at Boston College for independently verifying our measured
9. Yan, X. A. et al. Experimental studies on anisotropic thermoelectric thermoelectric property values.
properties and structures of n-type Bi2 Te2.7 Se0.3 . Nano Lett. 10,
3373–3378 (2010). Author contributions
10. Martin, J., Nolas, G. S., Zhang, W. & Chen, L. PbTe nanocomposites R.J.M. carried out experiments, synthesized and characterized the materials and wrote
synthesized from PbTe nanocrystals. Appl. Phys. Lett. 90, 222112 (2007). the paper with G.R. Thermoelectric measurements and modelling were carried out
11. Zhang, G. Q., Wang, W., Lu, X. L. & Li, X. G. Solvothermal synthesis of V–VI by Y.Z. Data interpretation and analysis was carried out collaboratively by R.J.M.,
binary and ternary hexagonal platelets: The oriented attachment mechanism. Y.Z., G.R. and T.B-T. C.K. and B.S. carried out transmission electron microscopy and
Cryst. Growth Des. 9, 145–150 (2009). X-ray photoelectron spectroscopy measurements, respectively. G.R. directed the project
12. Purkayastha, A. et al. Surfactant-directed synthesis of branched bismuth together with T.B-T. and R.W.S. All authors discussed the results and implications and
telluride/sulfide core/shell nanorods. Adv. Mater. 20, 2679–2683 (2008). commented on the manuscript at all stages.
13. Purkayastha, A., Lupo, F., Kim, S., Borca-Tasciuc, T. & Ramanath, G.
Low-temperature, template-free synthesis of single-crystal bismuth telluride Additional information
nanorods. Adv. Mater. 18, 496–500 (2006). The authors declare no competing financial interests. Supplementary information
14. Purkayastha, A. et al. Molecularly protected bismuth telluride nanoparticles: accompanies this paper on www.nature.com/naturematerials. Reprints and permissions
Microemulsion synthesis and thermoelectric transport properties. Adv. Mater. information is available online at http://www.nature.com/reprints. Correspondence and
18, 2958–2963 (2006). requests for materials should be addressed to T.B-T. or G.R.

240 NATURE MATERIALS | VOL 11 | MARCH 2012 | www.nature.com/naturematerials


© 2012 Macmillan Publishers Limited. All rights reserved

You might also like