Enhanced Thermoelectric Properties of Solution Grown Bi2Te3−xSexNanoplatelet CompositesAjay Soni,† Zhao Yanyuan,† Yu Ligen,‡ Michael Khor Khiam Aik,*,‡ Mildred S. Dresselhaus,§

and Qihua Xiong*,†,∥

†Division of Physics and Applied Physics, School of Physical and Mathematical Sciences, Nanyang Technological University,Singapore 637371‡Division of Manufacturing Engineering, School of Mechanical and Aerospace Engineering, Nanyang Technological University,Singapore 639798§Department of Physics and Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology,Cambridge, Massachusetts 02139, United States∥Division of Microelectronics, School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798

*S Supporting Information

ABSTRACT: We report on the enhanced thermoelectric properties ofselenium (Se) doped bismuth telluride (Bi2Te3−xSex) nanoplatelet (NP)composites synthesized by the polyol method. Variation of the Se compositionwithin NPs is demonstrated by X-ray diffraction and Raman spectroscopy.While the calculated lattice parameters closely follow the Vegard’s law, adiscontinuity in the shifting of the high frequency (Eg

2 and A1g2) phonon modes

illustrates a two mode behavior for Bi2Te3−xSex NPs. The electrical resistivity(ρ) of spark plasma sintered pellet composites shows metallic conduction forpure Bi2Te3 NP composites and semiconducting behavior for intermediate Secompositions. The thermal conductivity (κ) for all NP composites is much smallerthan the bulk values and is dominated by microstructural grain boundaryscattering. With temperature dependent electrical and thermal transport measure-ments, we show that both the thermoelectric power S (−259 μV/K) and the figureof merit ZT (0.54) are enhanced by nearly a factor of 4 for SPS pellets of Bi2Te2.7Se0.3 in comparison to Bi2Te3 NP composites.Tentatively, such an enhancement of the thermoelectric performance in nanoplatelet composites is attributed to the energy filtering oflow energy electrons by abundant grain boundaries in aligned nanocomposites.

KEYWORDS: Thermoelectric figure of merit, nanocomposites, Bi2Te3−xSex nanoplatelets, energy filtering, polyol synthesis

Research in the field of thermoelectricity is full of fantasticideas;1 the materials having properties of an electron-

crystal-phonon-glass2 such as rattling semiconductors,3 alloys,4−6

heterostructures,7,8 and nanocomposites9−11 are proven to beefficient thermoelectrics. The efficiency is defined by thethermoelectric figure of merit ZT = S2σT/(κe + κl), where S isthe thermoelectric power, σ is the electrical conductivity, κ is thethermal conductivity (κe is the electronic and κl is the latticecontributions) and T is average absolute temperature. An idealthermoelectric material in principle should be electrically short-circuited and thermally open-circuited with a high thermoelectricpower. However, the interdependency of the three physicalparameters (S, σ, and κ) imposes a limitation on the efficiency ofthermoelectric materials, and hence a decoupling of theseparameters is required to improve the figure of merit close to 2at room temperature.1,12,13 The improvement of ZT above 1 hasremained a challenging task until theoretical prediction byDresselhaus on the utilization of quantum confinement effects innanomaterials14 followed by experimental demonstration byVenkatasubramanian on thin films12 and recent work by the

Ren and Chen group on ball-milled and hot-pressed nano-composites.9,15 Currently, the research is mostly driven in twodirection: (i) to improve the quality of the materials16,17 andsynthesis methods18,19 and (ii) to develop a module ordevice.1,13,20,21

Nanocomposites of chalcogenides (Te and Se) andpnictogen (Bi and Sb) materials have been preferably studiedfor their high-performance TE properties.9,22−24 While the sixvalley degeneracy and the narrow energy gap of Bi2Te3 (and thefamily of similar compounds) is attractive for good electricalconduction, its layered crystal structure leads to a poor thermalconductivity, making these layered materials most suitable forthermoelectric applications.25,26 Experimentally, it turns outthat the binary Bi2−xSbxTe3 and Bi2Te3−xSex compounds are thebest p-type and n-type materials, respectively, for thermo-electric refrigeration.1,2,6,15 There are also various arguments

Received: October 5, 2011Revised: December 20, 2011Published: February 1, 2012

Letter

pubs.acs.org/NanoLett

© 2012 American Chemical Society 1203 dx.doi.org/10.1021/nl2034859 | Nano Lett. 2012, 12, 1203−1209

proposed to get better performance of these compounds likethe inclusion of Sb (at Bi sites) and Se (at Te sites) createsacceptor and donor levels, respectively, in the bulk band gap ofBi2Te3 and in turn increases the electronic conduction (andthus the power factor S2σ).15,16,25 Similarly, for some othercubic PbTe1−xSex composites, it has been demonstrated that Sedoping leads to a convergence of the degenerate electronicbands for enhanced thermoelectric performance.19 Further-more, interface scattering in nanostructures filters the low energyelectrons and contributes to enhanced the thermoelectricproperties with a reduction in κ other than by alloying.11,27,28

In the specific case of Bi2Te3−xSex, due to the isomorphic crystalstructure of Bi2Te3 and Bi2Se3, the solubility of Se in Bi2Te3results in a modification of the crystalline structure andelectronic density of states (DOS), which is beneficial forinhibiting the onset of intrinsic conduction and is an impor-tant factor for the thermoelectric properties of Bi2Te3alloys.24,25,29−31 Additionally, the anisotropy of the layeredcrystal structure plays an important role, as reported for solidsolutions of Bi2Te3 and Bi2Se3,

32 and also for nanograins ofBi2Te3−xSex aligned by the repressing of compacted pellets.15

The materials preparation method is another strategy forenahnacing the performances of thermoelectric materials asdifferent synthesis methods show different physical properties;however a deep understanding of these systems is still lacking.This provides a platform to further study the properties ofdoped nanocomposites within and beyond the alloy limit,29,31

where modified phonon and electron transport properties areexpected.24

Recently, we reported on thickness dependent Raman shiftsin chemically synthesized atomically thin topological insulatingBi2Se3 NPs.

33 In the present report, we demonstrate temper-ature dependent electrical and thermoelectric transport studieson SPS pellets of chemically synthesized NP of Bi2Te3−xSex.The observed results are compared with NP composites ofBi2Te3 and Bi2Se3. The study reveals enhancement in thethermoelectric power and power factor of NPs with Se doping,while the thermal conductivity remains smaller than the bulkvalues. Thermoelectric figure of merit is found to be ZT ∼ 0.54for Bi2Te2.7Se0.3 at 300 K, which shows the potential applica-bility of SPS composites of n-type NPs grown by a chemicalmethod.NPs of Bi2Te3 and Se doped Bi2Te3−xSex were synthesized

using a diethylene glycol (DEG) mediated polyol method.33 Ina typical synthesis, a stoichiometric ratio of bismuth nitrate(Bi(NO3)3.5H2O), potassium telluride (K2TeO3.H2O), andsodium selenide (Na2SeO3) were dissolved in DEG withvigorous stirring followed by refluxing the mixture solution at240 °C. Here the solvent DEG also works as a reducing agentfor the elemental species. The reaction was stopped after nearly4−5 h and the mixture was allowed to cool down to roomtemperature. The precipitated powder was recovered aftercentrifuging and washing several times using isopropyl alcoholfollowed by acetone. The intermediate sonication helps toremove any unreacted chemicals as well as to clean solventresidue from the surface of NPs. The synthesized powder wasthen dried at 70 °C for 1 h. It has been demonstrated earlierthat the polyol method adopted here can produce high qualityatomically thin NPs with a high yield 80−85% of finalproduct.33 The as-grown and washed powder was dispersed inan isopropyl alcohol solution and was then drop casted ontosilicon substrates for characterization by scanning electronmicroscopy (FESEM, JEOL 7001F) and Raman spectroscopy.

Raman spectroscopy measurements were conducted on amicro-Raman spectrometer (Horiba-JY T64000) in a back-scattering configuration with an excitation wavelength of 532 nm.The backscattered signal was collected through a 100× objectiveand dispersed by a 1800 g/mm grating under a triple subtractivemode with a spectral resolution of ∼1 cm−1 and the lowest fre-quency of 5 cm−1.33 X-ray diffraction measurements on synthesizedpowder and SPS pellets were performed using a Bruker D8advanced diffractometer with Cu Kα radiation in the lockedcoupled geometry of the X-ray gun and detector.The dried powder was compacted into macroscopic

composite pellets using spark plasma sintering (SPS) in agraphite die, by applying a pressure of 40 MPa at 250 °C and avacuum of 0.1 mbar after purging the chamber with nitrogen toavoid oxidation during pelletization. In SPS, the high currentthrough the powder produces sparks within the grainboundaries and thus connects NPs to one another due tolocal heating at weak links, without much affecting themorphologies of the NPs. To avoid grain growth, thetemperature for the SPS was kept just above the growthtemperature (240 °C) during synthesis. Pellets of compositesare shiny gray in color with a mass density more than 87% ofthe bulk density (7.857 g/cm3).25 A 10 mm long bar of width2 mm was cut from a nearly 0.3−0.5 mm thick circular disk andused for further transport measurements using a QuantumDesign PPMS instrument. A standard four probe method wasused for in-plane electrical resistivity measurements followed byheating the sample from one side for thermopower and thermalconductivity measurements. For the out-of-plane thermopowermeasurement, a small piece of the sample is sandwichedbetween two gold discs; the voltage is measured whilemaintaining temperature gradient across the two discs. In thisout-of-plane configuration, the accuracy for electrical andthermal conductivity measurements is sacrificed as current andvoltage leads are shared at cold and hot ends of the sample, liketwo probe measurements. In a continuous-mode measurementwith a 0.5 K/minute ramping rate and a 200 s step pulse for theheater leading to more than 20 min for each data point. Themeasurements have been performed under a vacuum of10−5 mbar to avoid any other losses. The emissivity value of0.3 is used for radiations loss corrections as described in theinstrument manual for rough conducting surfaces.The as-grown samples were mostly nanoplatelets having

lateral dimensions of more than 2−3 μm and a thickness of afew tens of nm (Figure 1a−c). It was noted that the scaling-up

Figure 1. FESEM images of NPs dispersed on a Si substrate (a)Bi2Te3, (b) Bi2Te2.7Se0.3, (c) Bi2Se3, (d,e) are top and cross-sectionview of SPS pellets, and (f) is a photograph of a diced SPS pellet with acoin used for size comparison. Scale bars are 1 μm for a, b, c, and e and10 μm for d.

Nano Letters Letter

dx.doi.org/10.1021/nl2034859 | Nano Lett. 2012, 12, 1203−12091204

of the synthesis results in NPs with slightly higher thicknessesthan in our previous report.33 Comparatively a perfecthexagonal morphology for Bi2Se3 and a truncated-edgehexagonal morphology for Bi2Te3 or Bi2Te3−xSex NPs wereobserved from the SEM images (Figure 1a−c). This mightbe because of the slower reactivity of the Se ions, which allowsthe reaction to progress slowly and thus the inherent anisotropiclayered crystal structure of Bi2Se3 is maintained. On the otherhand, a relatively faster reactivity and a lower reductionpotential of the Te ions leads to thicker NPs of Bi2Te3. TheSEM images of the top (Figure 1d) and cross section (Figure 1e)views of a sintered pellet are also shown for comparison. Fromthe significant alignment of the NPs, it might be expected thatthe orientation of the NPs would affect the physical propertiesof the resulting nanocomposites. This will be discussed later infurther sections. Figure 1f shows a photograph of the com-pacted pellet and the rectangular bar used for the measure-ments. The Se composition of the NP has been estimated usingEDS measurements from an average of three different sites ofthe SPS pellets (Figure S1 and Table 1, Supporting Information),and consequently the compositions are assigned to thevarious samples.The phase purity and crystal structure of NP SPS pellets have

been identified by X-ray diffraction and the spectra are shownin Figure 2a. Compared with the JCPDS data card no. 89-2009for Bi2Te3 and no. 89-2008 for Bi2Se3, these compounds havebeen found to exhibit rhombohedral crystal geometry (Spacegroup R3m) with no detectable impurities of other phases. Thelattice parameters calculated from the peak positions of the 006,110, and 015 planes are found to be in close agreement with thereported values (Table 2 in Supporting Information).15,34 Asystematic shift of the most intense XRD peak for the 015planes has been observed (Figure 2b). The linear contraction ofthe Bi2Te3 unit cell volume with increased Se concentrationclosely follows the Vegard’s law (Figure S2 in SupportingInformation).34 Broadening and shifting of the XRD peaks forthe Bi2Te2.7Se0.3, Bi2Te2.2Se0.8, and Bi2Te1.2Se1.8 compositessuggests that Se doping introduces disorder into the crystalstructures. It is noteworthy here to mention about theprominence of the XRD peak for the 006 and 0015 planes,which indicates a significant orientation of the NPs along the(001) direction. Shown in Figure 2c is the comparison of theXRD spectra of the synthesized powder and a SPS pellet ofBi2Te3. The two spectra show that crystalline NPs are significantlyoriented in the 001 direction with a diminishing of 110 directions

(peak at 2θ value of 41.14°) in the case of the pellet after thepelletization. This is advantageous for employing the anisotropyof the crystal structure for transport measurements withouttedious efforts of repressing/aligning of the NPs.15

Recently, Raman spectroscopy has been used extensively tostudy the vibrational properties and electron−phonon couplingin layered Bi2Se3 and Bi2Te3.

33 We performed Ramanspectroscopy measurements here to investigate how the com-positional variations due to Se doping affect the lattice vibra-tions and the electron phonon interactions in NPs. The unitcell of Bi2Te3 and similar compounds have three quintuplelayers stacked by van der Waal’s forces and this stackingfacilitates an easy cleavage along the c axis.34 A quintuple layeris an alternate arrangements of five atomic layers, Te(1)−Bi-Te(2)−Bi−Te(1), in which Te atoms exhibit two differentchemical environments (assigned as Te(1) and Te(2); Figure 3b).

While chemical bonding between Bi−Te(2) is of a purecovalent nature, it is slightly ionic but still covalent in naturebetween Bi−Te(1).34,35 For the Bi2Te3−xSex compounds, it isunderstood that Se atoms, being more electronegative than Te,preferentially replaces Te at Te(2) sites first, followed by

Figure 2. (a) XRD patterns of SPS pellets of Bi2Te3, Bi2Te3−xSex (as mentioned) and Bi2Se3 NP composites, (b) expanded view showing asystematic shift of the 015 peak for Bi2Te3 (at 27.3°) and Bi2Se3 (at 29.3°), (c) comparison of the XRD spectra for synthesized NPs powder (bottomblack curve) and a SPS pellet (top red curve) of Bi2Te3. The two spectra reveal the typical orientation of the NPs after pelletization.

Figure 3. (a) Raman spectra of pure and doped Bi2Te3−xSex NPs, (b)schematic diagrams for the four Raman-active normal modes, and (c)phonon frequencies versus Se composition. Raman spectra onindividual NPs were recorded in the range of 20−600 cm−1; however,for clarity, Raman shifts are shown for 20−250 cm−1.

Nano Letters Letter

dx.doi.org/10.1021/nl2034859 | Nano Lett. 2012, 12, 1203−12091205

random replacements of Te at Te(1) sites.34,35 Hence, it isexpected that the Se doping will change the crystallinestructures and lattice dynamics in a systematic manner.With 5 atoms in the primitive cell, there are 15 dynamical

modes, at the center of Brillouin zone among which 12 are theoptical modes and other 3 are the acoustic modes. Theobserved peaks in the Raman spectra for Bi2Te3 (Figure 3a) areassigned to the vibrational modes Eg

1 (∼36 cm−1), A1g1 (62−

72 cm−1), Eg2 (102−162 cm−1), and A1g

2 (137−172 cm−1),respectively. Schematic diagram in Figure 3b depicts the fourRaman active vibrational modes A1g

1, A1g2, and the doubly

degenerate Eg1 and Eg

2 with the corresponding atomicdisplacements. More details of the analysis can be found inthe literature.33,36 The observed Raman shifts of the variousvibrational modes with Se(x) are summarized in Figure 3c.Here, low frequency modes Eg

1 and A1g1 have very small shifts

with Se doping since these modes arises due to in-phasevibrations of Bi−Te/Se(1). From the schematic in Figure 3b,it is clear that the central Te(2) atoms are the center of massof the Raman active vibrations. Consequently, for small dopingof Se atoms, which preferentially replaces Te(2) atoms, onlya slight shifting in the Raman modes is anticipated for Bi2Te2.9Se0.1,Bi2Te2.84Se0.16, and Bi2Te2.7Se0.3. For higher Se content NPs, theSe atoms continue to replace Te(1) atoms randomly. In ourexperiments a discontinuity in the Raman Shift with a splittingof the Ag

2 and Eg2 vibrational modes is observed. It is important

to note that Bi−Te(1)/Se(1) atoms move out of phase in theAg

2 and Eg2 vibrational modes, which explains the discontinuity.36

Moreover, the different ionic sizes for Te and Se atoms alsoaccount for this behavior for Bi2Te2.2Se0.8 and Bi2Te1.2Se1.8nanocomposites. The possibility of oxidation and changesin structure due to SPS in pellets has been discussed in theSupporting Information Figure S3.Having obtained a convincing characterization of the various

compositions, we performed temperature-dependent transport

measurements on rectangular bars of SPS pellets (shown inFigure 1f). The electrical resistivity of the nanocomposites(Figure 4a) shows a metallic behavior and increases for a criticalSe doping. From the data, it is observed that scattering of thecharge carriers is temperature and doping dependent. Forinstance, ρ increases as the temperatures decrease for Bi2Te2.2Se0.8nanocomposites. Moreover, for intermediate concentrations suchas x ∼ 0.16, 0.3, 0.8, and 1.8, ρ increases as the temperature dropsbelow 30 K, possibly due to the trapping of charge carriers in NPsor the scattering of electrons with a large number of interfacesor grain boundaries. The disorder in the nanocomposites is alsodocumented by XRD and Raman analysis. Considering thenarrow band gap for pure Bi2Te3 (0.15 eV), where donor andacceptor levels lie extremely close to the conduction andvalence band edges, a small amount of Se impurity would leadto a broadening of the impurity levels.24 The Se atom substitu-tion in Te(2) sites, being more electronegative, will increasethe ionic component of the bonding between the Bi and Seatoms at these sites. The bonding then weakens in the structureand determines the energy gap of the materials.24 Thus forBi2Te3−xSex nanocomposites, the bond strength and hence theenergy gap will increase with increasing Se, until x reaches 1(for Bi2Te2Se1), at which composition all the Te(2) sites areoccupied by Se.24,25,37 As the Se impurities increase further, thetrend is expected to be reversed and the energy gap will tend todecrease with an increasing concentrations of Se atoms (whenx > 1). Since now Se atoms will go into Te(l) sites and theytend to attract charge along the Bi−Te(1) site bonds, thereforethe bismuth atoms become more electronegative, which in turncauses charge to move toward the bismuth along the Bi−Te(2)site bonds. Thus, semiconducting behavior is observed inBi2Te2.2Se0.8 and the rise in resistivity below 30 K (in Figure 4a)is in line with the earlier reports on solid solutions and alloysof Bi2Te3 and Bi2Se3.

32,37,38 The magnitude of ρ for NPcomposites showing higher values relative to the earlier reports

Figure 4. Temperature dependence of the (a) electrical resistivity, (b) thermoelectric power, (c) lattice (κl) and electronic (κe) thermalconductivities (estimated from the Wiedemann−Franz law; see text for details), (d) thermoelectric figure of merit of Bi2Te3−xSex NP composites. Allthe figures share the same color notations as shown in (d).

Nano Letters Letter

dx.doi.org/10.1021/nl2034859 | Nano Lett. 2012, 12, 1203−12091206

on ball milled and hot pressed nanocomposites,9,15 leads to thecomparatively poor performance of the thermoelectric figure ofmerit of the present solution grown NP composites. However,in the present nanocomposites, the thickness of the NP is verymuch smaller in comparison to the micrometer size nanocryst-als in earlier reports.9,15,16 As mentioned earlier, the presentSPS nanocomposites have a mass density of ∼87%, whichmeans that there exists a large number of grain boundaries inthe nanocomposites. Here the grain boundaries also introducetrap-states, which immobilize the charge carriers and thusreduce the carrier mobility resulting into a higher electricalresistivity.The thermoelectric power, S, for all NP composites have

negative values down to T ∼ 5 K (Figure 4b), showing a typicaln-type behavior. The linear temperature dependence showsthat the dominant transport mechanism is diffusive transportunder the applied thermal gradients.26 The magnitude of S forpure Bi2Te3 and Bi2Se3 compounds (−85 and −80 μV/K,respectively), is smaller than that of for Bi2Te3−xSex composites,which is lined up with the good conductivity for the purecompounds. It is known that S is very sensitive to theasymmetry of the electronic density of states (DOS) at theFermi level (EF), and a small change in the DOS is reflected inthe magnitude of S, which in turn is related to the charge carrierdensity (n). From the linear temperature dependence of S, n canbe calculated using the semiclassical Mott-Jones formula39,40

= −π

| |∂ σ

∂= −

π *

π ℏ | |ST

ke

EE

k m

e n3ln ( )( ) (3 )

1

( )E

2B2 2

B2

2 2/3 2 2/3F (1)

where kB is the Boltzmann constant, e is electronic charge, n iscarrier density, σ(E) is conductivity, m* is the effective mass ofelectrons, T is temperature, EF is the Fermi energy, and ℏ is thePlanck constant. Assuming a simple parabolic band structure anda constant m*, the slope of the linear fit to S/T is used to extractthe carrier density in the semiconductors.25,26 Here an effectivemass of m* = 0.36me (me is the electron rest mass)25 is used forthe calculations and the estimated carrier density is found to bethe order of 1019 cm−3, which lies in the range of maximum ZTvalues for the high-performance thermoelectric materials.2,6 It isnoteworthy here that n decreases with Se doping therebyattaining a minimum value of 9.6 × 1020 cm−3 for Bi2Te2.7Se0.3and n increases further with increase in Se concentration for theBi2Te2.2Se0.8 and Bi2Te1.2Se1.8 compositions. We also quantify thecarrier concentration of Bi2Te2.7Se0.3 pellet using Hall measure-ment, which yielded n ∼7.8 × 1020 cm−3, in good agreement withthe estimation based on Mott-Jones formula. Enhancement in thethermopower for NP composites can be attributed to the tuning ofn and to the modification of the DOS due to Se doping in NPs.Interfaces and grain boundaries also play an additional role byintroducing barriers for filtering low energy charge carriers.11,27 Thischarge carrier filtering is more prominent for Bi2Te3−xSex NPcomposites than for pure phases. As discussed earlier, interfacescattering increases the resistivity of the NPs, and this in turn hereenhances the thermopower of the present NP composites. Energy-filtered enhancement of the thermopower has been reported forheterostructured superlattices where interfaces become impor-tant.27,28,41

One of the main advantages of utilizing nanomaterials forthermoelectric applications is the reduced thermal conductivity,κ.6,25,42,43 It has been very difficult to reduce the thermalconductivity of crystalline solids below that of an alloy withoutcreating defects, dislocations, and voids. However, it is believed

that the alloy limit can be beaten by nanostructuring.9,15 Themagnitude of κ is significantly lower than that for bulk values(1 W/m·K)9 in the temperature range of 5−300 K and isapproaching close to the κlmin proposed by Cahill,44 which is0.28 for superlattices of Bi2Te3.

12 For all NP composites, κdecreases with temperature and shows a small hump-likefeature at typical intermediate temperatures where theelectron−phonon interactions are dominant (Figure 4d).45

The electronic contribution to the total thermal conductivity isrelated to the electrical conductivity and is calculated from theWiedemann−Franz Law: κe = LσT, where L is the Lorenzfactor, 2.45 × 10−8 W·S/K2.2,25 The possibility of heat transferby bipolar charge carriers is ruled out from the negative sign ofS, which means that the majority charge carriers are electrons inthe present nanocomposites. The electronic κe and lattice (κl =κ − κe) contributions are separately presented in the Figure 4c.It is clearly seen here that κl dominates over κe for the wholetemperature range, illustrating the importance of the micro-structural aspects of NP composites. From the temperature-dependent transport studies, even though ρ is comparatively ona higher scale, the increase in S is significant, thus increasing thepower factor (S2/ρ). Furthermore, both the increased powerfactor and the reduced κ result in an enhanced figure of merit inthe present NP composites (Figure 4(d).

Now, we discuss the effect of Se doping on the physicalproperties of nanocomposites at room temperatures. As shownin Figure 5, the ρ increases with Se concentration and finds amaximum value for Bi2Te2.2Se0.8. The thermopower values changedrastically from −85 μV/K for Bi2Te3 to −259 μV/K forBi2Te2.7Se0.3 NPs, indicating an enhancement of almost 3 times.Both the physical parameters (σ and S) are mostly affected by the n,which is the number of charge carriers (DOS at EF). Thus, Sedoping in the NP composites tunes the n in the regime of 1019−1020 cm−3, which is supposed to be advantageous for thermo-electric applications.2,6 The magnitude of κ is reduced with Sedoping for Bi2Te2.7Se0.3 and further increases attaining amaximum value for Bi2Te1.8Se1.2 which is in close agreementwith the reports on alloys and solid solutions.29,46 With all theseparameters, the room temperature thermoelectric figure of

Figure 5. Room temperature (a) electrical resistivity, (b) thermo-electric power, (c) total thermal conductivity (κl + κe) and (d)thermoelectric figure of merit as a function of Se composition.

Nano Letters Letter

dx.doi.org/10.1021/nl2034859 | Nano Lett. 2012, 12, 1203−12091207

merit is ranging between 0.1 for Bi2Te3 and 0.54 forBi2Te2.7Se0.3.Antisite defects play an important role for chalcogenide

thermoelectric materials. This has been reported for Bi2Te3 thinfilms where the most stoichiometric sample has highcrystallinity, high thermopower, and high electron mobility.47,48

However, the c-axis lattice constant, thermopower, and electronmobility decrease with excess Te, and the crystallinity clearlydegrades. All of these observations have been explained asarising from antisite defects in which excess Te occupies the Bilattice sites and behaves as an n-type dopant and the antisitedefects will dominate over vacancies. Cho et al. concluded thatboth the reduced lattice constants and the reduced thermo-powers are attributed to heavy n doping via antisite defects.This is not the case in the present samples, as the compositesare of a crystalline nature with a systematic reduction in theunit cell volume. Moreover the thermopower is higher for thedoped sample than for the pure samples in contradiction to thereport by Cho et al.,47 where the thermopower is smaller forsamples with antisite defects.Anisotropy in bulk-layered chalcogenide structures is an

important factor, as important physical parameters can bedifferent along different directions. For single crystal solidsolutions of Bi2Te2.7Se0.3, the electrical conductivity (1/ρ) isfour times higher along the ab-planes, due to the same type ofatoms in the basal planes, than along the c axis; κ is two timessmaller along the c axis due to possible damage in the cleavageplanes.32 Yan et al.15 also demonstrated that repelletizingorients the nanograins in a typical crystal direction and thusimproves the ZT by 22%. However, it was reported that Sremains isotropic and the improvement in ZT is due toanisotropic conductivities.15 In our measurements, we haveobserved that S is anisotropic when measured with thetemperature gradient across the two faces (Figure 6 top right

inset) of the pellet. The room temperature S(a) value is muchlower than the S(b) values with the temperature gradient alongthe lateral side (Figure 6, bottom left inset) of the SPS pellet.This anisotropic S can be understood by modeling theconnectivity of the NPs and the density of the grain boundariesin the two directions. It is shown in Figure 2c that the SPSpelletization significantly aligns the NPs (XRD comparison of thepowder and SPS pellet). The alignment of these two-dimensionalNPs during SPS pelletization mounts-up NPs along the top-

bottom surfaces, where a comparatively large surface area finds arelatively good connectivity with only a small amount of grainboundaries. This would lead to a lower probability for filteringlow energy charge carriers, and thus S has relatively smaller values(Figure 6 curve a). On the other hand, the situation is morecomplicated for other directions with more lateral connectivitycoming from the NPs and yet more scattering coming from thelarge number of grain boundaries. The temperature dependenceof S is linear, demonstrating that the transport is diffusive in theplanar direction; however, it is nonlinear across the pellet. Thisdemonstrates that the scattering of the charge carriers is differentin the two directions. Recently, two-dimensional nanostructuresof Bi2Te3 and Bi2Se3 have also been studied for their topologicalproperties where surface states enhance their thermoelectricproperties.49,50 Further studies on the linear (in-plane) andnonlinear (out of plane) temperature dependence of S arerequired for these NPs.In summary, we have demonstrated that the facile and cost-

effective polyol method can be scaled up to synthesize highquality Bi2Te3−xSex NPs, which enables the preparation ofmacroscopic pellets by spark plasma sintering with preferentialalignment along the c axis. Highly anisotropic thermoelectricpower values are found for the in-plane and out-of-plane S(T)and this anisotropy is tentatively attributed to anisotropicenergy filtering in the two directions by grain boundaries. Byadjusting the Se composition, the optimal thermoelectricperformance is found for the Bi2Te2.7Se0.3 NP composite,which exhibits an S value of −259 μV/K and a figure of meritZT of 0.54 at room temperature. Further research is required toimprove the electrical conductivity of the nanocomposite whilestill keeping the thermal conductivity low either by usingdifferent composite processing parameters or by introducingother conducting polymers with intrinsic lower thermalconductivities.

■ ASSOCIATED CONTENT

*S Supporting InformationAdditional information and figures. This material is availablefree of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATION

Corresponding Author*E-mail: (M.K.K.A.) mkakhor@ntu.edu.sg; (Q.X.) Qihua@ntu.edu.sg.

■ ACKNOWLEDGMENTS

Q.X. acknowledges support from the Singapore NationalResearch Foundation through Singapore 2009 NRF fellowshipgrant (NRF-RF2009-06), the Singapore Ministry of Education viaa Tier 2 grant (MOE 2010-T2-2-059) the Startup grant support(M58110061), and the New Initiative Fund (M58110100) fromNanyang Technological University, Singapore. M.S.D. would liketo acknowledge S3TEC-EFRC program supported by DOE.

■ REFERENCES(1) Bell, L. E. Science 2008, 321 (5895), 1457−1461.(2) Snyder, G. J.; Toberer, E. S. Nat. Mater. 2008, 7 (2), 105−114.(3) Mahan, G.; Sales, B.; Sharp, J. Phys. Today 1997, 50 (3), 42−47.(4) Joshi, G.; Lee, H.; Lan, Y.; Wang, X.; Zhu, G.; Wang, D.; Gould,R. W.; Cuff, D. C.; Tang, M. Y.; Dresselhaus, M. S.; Chen, G.; Ren, Z.Nano Lett. 2008, 8 (12), 4670−4674.

Figure 6. Thermopower measured with temperature gradients (a)across and (b) in-plane of SPS pellet. The top right schematic showsthe cross plane configuration of the measurements and the bottom leftschematic shows the in-plane configuration for the measurements.Arrows represents the direction of heat flow.

Nano Letters Letter

dx.doi.org/10.1021/nl2034859 | Nano Lett. 2012, 12, 1203−12091208

(5) Zebarjadi, M.; Joshi, G.; Zhu, G.; Yu, B.; Minnich, A.; Lan, Y.;Wang, X.; Dresselhaus, M.; Ren, Z.; Chen, G. Nano Lett. 2011, 11 (6),2225−2230.(6) Minnich, A. J.; Dresselhaus, M. S.; Ren, Z. F.; Chen, G. EnergyEnviron Sci 2009, 2 (5), 466−479.(7) Pei, Y.; Lensch-Falk, J.; Toberer, E. S.; Medlin, D. L.; Snyder,G. J. Adv. Funct. Mater. 2011, 21, 241−249.(8) Zhang, G.; Yu, Q.; Wang, W.; Li, X. Adv. Mater. 2010, 22 (17),1959−1962.(9) Poudel, B.; Hao, Q.; Ma, Y.; Lan, Y.; Minnich, A.; Yu, B.; Yan, X.;Wang, D.; Muto, A.; Vashaee, D.; Chen, X.; Liu, J.; Dresselhaus, M. S.;Chen, G.; Ren, Z. Science 2008, 320 (5876), 634−638.(10) See, K. C.; Feser, J. P.; Chen, C. E.; Majumdar, A.; Urban, J. J.;Segalman, R. A. Nano Lett. 2010, 10 (11), 4664−4667.(11) Dresselhaus, M. S. Adv. Mater. 2007, 19, 1043−1053.(12) Venkatasubramanian, R.; Siivola, E.; Colpitts, T.; O’Quinn, B.Nature 2001, 413 (6856), 597−602.(13) DiSalvo, F. J. Science 1999, 285 (5428), 703−706.(14) Hicks, L. D.; Dresselhaus, M. S. Phys. Rev. B 1993, 47 (19),12727.(15) Yan, X.; Poudel, B.; Ma, Y.; Liu, W. S.; Joshi, G.; Wang, H.; Lan,Y.; Wang, D.; Chen, G.; Ren, Z. F. Nano Lett. 2010, 10 (9), 3373−3378.(16) Scheele, M.; Oeschler, N.; Veremchuk, I.; Reinsberg, K.-G.;Kreuziger, A.-M.; Kornowski, A.; Broekaert, J.; Klinke, C.; Weller, H.ACS Nano 2010, 4 (7), 4283−4291.(17) Xie, W.; He, J.; Kang, H. J.; Tang, X.; Zhu, S.; Laver, M.; Wang,S.; Copley, J. R. D.; Brown, C. M.; Zhang, Q.; Tritt, T. M. Nano Lett.2010, 10 (9), 3283−3289.(18) Zhao, Y.; Dyck, J. S.; Hernandez, B. M.; Burda, C. J. Phys. Chem.C 2010, 114 (26), 11607−11613.(19) Pei, Y.; Shi, X.; LaLonde, A.; Wang, H.; Chen, L.; Snyder, G. J.Nature 2011, 473 (7345), 66−69.(20) Majumdar, A. Science 2004, 303 (5659), 777−778.(21) Majumdar, A. Nat. Nanotechnol. 2009, 4 (4), 214−215.(22) Datta, A.; Paul, J.; Kar, A.; Patra, A.; Sun, Z.; Chen, L.; Martin,J.; Nolas, G. S. Cryst. Growth Des. 2010, 10 (9), 3983−3989.(23) Mehta, R. J.; Karthik, C.; Singh, B.; Teki, R.; Borca-Tasciuc, T.;Ramanath, G. ACS Nano 2010, 4 (9), 5055−5060.(24) Greenaway, D. L.; Harbeke, G. J. Phys. Chem. Solids 1965,26 (10), 1585−1604.(25) Nolas, G. S.; Sharp, J.; Goldsmid, H. J. Thermoelectrics: BasicPrinciples and New Materials Developments; Springer: New York, 2001.(26) Rowe, D. M. Thermoelectrics handbook: macro to nano; Rowe,D. M., Ed.; Boca Raton: CRC/Taylor & Francis: Boca Raton, FL, 2006.(27) Medlin, D. L.; Snyder, G. J. Curr. Opin. Colloid Interface Sci.2009, 14 (4), 226−235.(28) Zebarjadi, M.; Shakouri, A.; Esfarjani, K. Phys. Rev. B 2006,74 (19), 195331.(29) Mahan, G. D. Solid State Phys. 1998, 51, p81−157.(30) Yu, R.; Zhang, W.; Zhang, H. J.; Zhang, S. C.; Dai, X.; Fang, Z.Science 2010, 329 (5987), 61−64.(31) Wright, D. A. Nature 1958, 181 (4612), 834−834.(32) Carle, M.; Pierrat, P.; Lahalle-Gravier, C.; Scherrer, S.; Scherrer,H. J. Phys. Chem. Solids 1995, 56 (2), 201−209.(33) Zhang, J.; Peng, Z.; Soni, A.; Zhao, Y.; Xiong, Y.; Peng, B.;Wang, J.; Dresselhaus, M. S.; Xiong, Q. Nano Lett. 2011, 11 (6),2407−2414.(34) Wiese, J. R.; Muldawer, L. J. Phys. Chem. Solids 1960, 15 (1−2),13−16.(35) Drabble, J. R.; Goodman, C. H. L. J. Phys. Chem. Solids 1958,5 (1−2), 142−144.(36) Richter, W.; Becker, C. R. Phys. Status Solidi B 1977, 84 (2),619−628.(37) Fuschillo, N.; Bierly, J. N.; Donahoe, F. J. J. Phys. Chem. Solids1959, 8 (0), 430−433.(38) Wang, R. Y.; Feser, J. P.; Gu, X.; Yu, K. M.; Segalman, R. A.;Majumdar, A.; Milliron, D. J.; Urban, J. J. Chem. Mater. 2010, 22 (6),1943−1945.

(39) Jonson, M.; Mahan, G. D. Phys. Rev. B 1980, 21 (10), 4223.(40) Schiller, W.; Richter, J. Phys. Status Solidi B 1982, 114 (1), 151−160.(41) Zebarjadi, M.; Esfarjani, K.; Bian, Z.; Shakouri, A. Nano Lett.2010, 11 (1), 225−230.(42) Kanatzidis, M. G. Chem. Mater. 2010, 22, 648−659.(43) Scheele, M.; Oeschler, N.; Meier, K.; Kornowski, A.; Klinke, C.;Weller, H. Adv. Funct. Mater. 2009, 19 (21), 3476−3483.(44) Cahill, D. G.; Watson, S. K.; Pohl, R. O. Phys. Rev. B 1992,46 (10), 6131−6140.(45) Ravich, Y. I.; Efimova, B. A.; Smirnov, I. A. Semiconducting LeadChalcogenides; Plenum Press: New York, 1970.(46) Yim, W. M.; Rosi, F. D. Solid-State Electron. 1972, 15 (10),1121−1140.(47) Cho, S.; Kim, Y.; DiVenere, A.; Wong, G. K.; Ketterson, J. B.;Meyer, J. R. Appl. Phys. Lett. 1999, 75 (10), 1401−1403.(48) Horak, J.; Koudelka, L.; Klikorka, J. React. Solids 1988, 5 (4),351−359.(49) Ghaemi, P.; Mong, R. S. K.; Moore, J. E. Phys. Rev. Lett. 2010,105 (16), 166603.(50) Moore, J. E. Nature 2010, 464 (7286), 194−198.

Nano Letters Letter

dx.doi.org/10.1021/nl2034859 | Nano Lett. 2012, 12, 1203−12091209