Experimental Studies on AnisotropicThermoelectric Properties and Structures ofn-Type Bi2Te2.7Se0.3

Xiao Yan,† Bed Poudel,‡ Yi Ma,‡ W. S. Liu,† G. Joshi,† Hui Wang,† Yucheng Lan,†Dezhi Wang,† Gang Chen,*,§ and Z. F. Ren*,†

†Department of Physics, Boston College, Chestnut Hill, Massachusetts 02467, ‡GMZ Energy, Inc., 11 Wall Street,Waltham, Massachusetts 02453, and §Department of Mechanical Engineering, Massachusetts Institute ofTechnology, Cambridge, Massachusetts 02139

ABSTRACT The peak dimensionless thermoelectric figure-of-merit (ZT) of Bi2Te3-based n-type single crystals is about 0.85 in the abplane at room temperature, which has not been improved over the last 50 years due to the high thermal conductivity of 1.65 W m-1

K-1 even though the power factor is 47 × 10-4 W m-1 K-2. In samples with random grain orientations, we found that the thermalconductivity can be decreased by making grain size smaller through ball milling and hot pressing, but the power factor decreasedwith a similar percentage, resulting in no gain in ZT. Reorienting the ab planes of the small crystals by repressing the as-pressedsamples enhanced the peak ZT from 0.85 to 1.04 at about 125 °C, a 22% improvement, mainly due to the more increase on powerfactor than on thermal conductivity. Further improvement is expected when the ab plane of most of the small crystals is reorientedto the direction perpendicular to the press direction and grains are made even smaller.

KEYWORDS Bi2Te2.7Se0.3, thermoelectric material, anisotropy, grain orientation

Solid-state thermoelectric (TE) converters are recentlyreceiving increasing attention due to their potentialto make important contributions to the effort on

reducing CO2 and greenhouse gas emission and providingcleaner forms of energy.1 Bismuth telluride based singlecrystal like bulk solid solutions, including p-type BixSb2-xTe3

and n-type Bi2Te3-ySey, still remain the best TE materialsused at near room temperature.2,3 One notable attributeabout the Bi2Te3-based single crystal bulks is the lamellarstructure and the weak van der Waals bonding betweenTe(1)-Te(1), which is responsible for the easy cleavage alongthe planes perpendicular to the c-axis.4,5 Originating fromthis unique structural anisotropy, thermoelectric propertiesof n-type Bi2Te3-ySey single crystal solid solutions preparedby traveling heater method shows strong anisotropy.6 Theelectrical and thermal conductivities along the cleavageplanes (perpendicular to the c-axis) are about four and twotimes larger than those along the c-axis, respectively. Eventhough the Seebeck coefficient is nearly isotropic, the ther-moelectric figure-of-merit Z along the cleavage planes isapproximately two times as large as that along the c-axis.At room temperature a maximum dimensionless thermo-electric figure-of-merit ZT was achieved at 0.85 (Z ) 2.9 ×10-3 K-1) for solid solutions with a 2.5 atom % Se replacingTe:Bi2Te2.925Se0.075 that has a power factor (defined as S2σ,where S is the Seebeck coefficient and σ the electrical

conductivity) of 47 × 10-4 W m-1 K-2 and thermal conduc-tivity of 1.65 W m-1 K-1 in which the lattice contribution is1.27 W m-1 K-1 (ref 6).

In principle, ZT could be greatly improved if we candecrease the thermal conductivity by breaking the singlecrystal into individual nanosize grains and thus increasephonon scattering due to the significantly increased grainboundaries of nanograins7,8 while maintaining the highpower factor by keeping the preferential orientation ofgrains.9,10 Ball milling and direct current hot pressing provesto be a successful approach to make nanostructured com-posite (nanocomposite)11-13 with a 40% peak ZT improve-ment from 1 to 1.4 in p-type nanostructured bulk bismuthantimony telluride by decreasing the thermal conductivity.11

Following this approach, we have successfully synthesizedn-type Bi2Te2.7Se0.3 bulk samples by ball milling and dc hotpressing and achieved a significantly lower thermal conduc-tivity of 1.06 W m-1 K-1 (with a lattice contribution of 0.7W m-1 K-1, much lower than the 1.27 W m-1 K-1 in singlecrystals) due to the increase of grain boundaries, in com-parison to the 1.65 W m-1 K-1 in the case of single crystalbulk samples. However, ZT was not enhanced at all becauseof a much lower power factor of 25 × 10-4 W m-1 K-2, incomparison to the 47 × 10-4 W m-1 K-2 in single crystalbulk samples. We suspect that the reason for the lowerpower factor is due to the randomness of the small grains.Therefore, the challenge is to improve the power factor tothe level close to that of the single crystal like bulks whilekeeping the low thermal conductivity owing to the finegrains. In literature, there have been a couple of methods

* To whom correspondence should be addressed, gchen2@mit.edu andrenzh@bc.edu.Received for review: 4/01/2010Published on Web: 07/30/2010

pubs.acs.org/NanoLett

© 2010 American Chemical Society 3373 DOI: 10.1021/nl101156v | Nano Lett. 2010, 10, 3373–3378

reported to prepare polycrystalline Bi2Te3-based alloy bulkswith preferred grain orientation, such as hot pressing14,15

and hot extrusion.16,17 However, those reports were notabout small grains of less than a couple of micrometers, butlarge grains of many micrometers. There were also reportsabout melt spinning to improve the ZT.18,19 However theimportant aspects of structural and property anisotropy werenot discussed in detail.

In this paper we report experimental studies on thephysical properties of samples with partial alignment ofthe small grains. To achieve grain alignment, we repressedthe as-pressed Bi2Te2.7Se0.3 bulk in a bigger diameter die ata higher temperature so that lateral flow of the small grainstakes place in the disk plane to achieve certain orientationof the ab plane in the disk plane. As a result, we increasedthe power factor by 40% in the disk plane direction from25 × 10-4 to 35 × 10-4 W m-1 K-2 without too much penaltyon the thermal conductivity, leading to a peak ZT enhance-ment of 22% from 0.85 to 1.04. By contrast, ZT along thepress direction is decreased, which does not affect theapplication of thermoelectric materials since only the direc-tion with the highest ZT is used in devices.

Experimental Section. To make alloyed powder, ap-propriate amounts of element Bi (99.999%), Te (99.999%),and Se (99.999%) from Alfa Aesar were weighed accordingto the stoichiometry Bi2Te2.7Se0.3, loaded into a stainless steeljar with stainless steel balls, and then subjected to ballmilling. After ball milling, the powders were alloyed and thesize of the alloyed Bi2Te2.7Se0.3 particles is about 20-50 nmby transmission electron microscope (TEM). The as-milledpowders were then loaded into a graphite die with an innerdiameter of 19.05 mm and pressed using a dc hot press,11-13

resulting in a cylinder of 19.05 mm in diameter and 22.7mm in thickness (“as-pressed”). The as-pressed bulk wasthen loaded in the center of a graphite die with an inner

diameter of 25.4 mm and then repressed in a furnace underprotection of flowing nitrogen gas. Eventually, disklikesamples of 25.4 mm in diameter and 12.7 mm in thicknesswere obtained.

X-ray diffraction (PANalytical X’Pert Pro) analysis with awavelength of 0.154 nm (Cu KR) was performed on the as-pressed and repressed samples in the plane both parallel andperpendicular to the press direction to investigate the grainorientation. The freshly fractured surfaces of samples wereobserved by scanning electron microscopy (SEM) (JEOL6340F).

To study the thermoelectric properties of as-pressed andrepressed samples in both directions parallel and perpen-dicular to the press direction, disks of 12.7 mm diameterand 2 mm thickness and bars of about 2 × 2 × 12 mm weremade from both directions of the as-pressed and repressedbulk samples. The bar-shaped samples were used to mea-sure the electrical conductivity and Seebeck coefficientsimultaneously, and the disk-shaped samples were used tomeasure the thermal conductivity. The electrical conductiv-ity of the samples was measured by a four-point dc current-switching technique, and the Seebeck coefficient was mea-sured by a static dc method based on the slope of the voltageversus temperature-difference curves using commercialequipment (ULVAC, ZEM3). The thermal diffusivity wasmeasured by the laser-flash method with a commercialsystem (LFA 447 Nanoflash, Netzsch Instruments, Inc.).Specific heat was determined by a commercial instrument(DSC 200-F3, Netzsch Instruments, Inc.). The volume densityof the samples was measured by the Archimedes method.Thermal conductivity was then calculated as the product ofthermal diffusivity, specific heat, and volume density. Theerrors are 3% for electrical conductivity, 5% for Seebeckcoefficient, 4% for thermal conductivity, 10% for powerfactor, and 11% for ZT.

FIGURE 1. XRD patterns of the as-pressed Bi2Te2.7Se0.3 bulks for the planes (a) perpendicular (⊥) and (b) parallel (|) to the press direction,and SEM images of the freshly fractured surfaces of as-pressed samples in the direction (c) perpendicular (⊥) and (d) parallel (|) to thepress direction.

© 2010 American Chemical Society 3374 DOI: 10.1021/nl101156v | Nano Lett. 2010, 10, 3373-–3378

Results and Discussion. Parts a and b of Figure 1 showthe X-ray diffraction (XRD) patterns of both the planesperpendicular (⊥) and parallel (|) to the press direction,respectively, and parts c and d of Figure 1 show the SEMimages of the freshly fractured surfaces in ⊥ and | directions,respectively. These two XRD spectra look the same for boththe 2θ position and intensity of each peak, indicating thatthere is no significant grain orientation anisotropy. The ratiosof the integrated intensity of planes (006) and (0015) to thatof the strongest peak (015) are shown in the insets. Theminor larger ratios of I(006)/I(015) (15% vs 12%) and I(0015)/I(015)

(12% vs 8%) for the plane that is ⊥ to the press direction(Figure 1a) mean that there is a little bit more ab orientationin the ⊥ direction, which is reflected on the physical proper-ties (discussed below). From the SEM images (Figure 1c,d),we can see that they are consistent with the XRD result(Figure 1a,b): the grains are random without a significantlypreferred crystal orientation. The average grain size is about1-2 µm even though there are many grains smaller than1 µm.

In Figure 2, we show the transport properties of the as-pressed samples in both the directions perpendicular (⊥) andparallel (|) to the press direction. The electrical conductivityσ⊥ is about 12-15% higher than σ| for the whole tempera-ture range (Figure 2a), which clearly shows that there is alittle bit more ab plane orientated in the ⊥ direction, con-sistent with the results shown by XRD (Figure 1a). TheSeebeck coefficients of the as-pressed sample are verysimilar in both directions (Figure 2b), consistent with the factthat Seebeck coefficient of single crystals is nearly isotropicin different crystal orientations.6 The corresponding powerfactors of the as-pressed sample in both directions are shownin Figure 2c. A power factor (S2σ) of 25 × 10-4 W m-1 K-2

at room temperature was obtained for the ⊥ direction anddrops down to 13 × 10-4 W m-1 K-2 at 250 °C while that inthe | direction is about 15-19% lower in the temperaturerange. It is very important to note that the power factor ofthe as-pressed sample in either direction is much lower thanthe 47 × 10-4 W m-1 K-2 in the ab plane of the singlecrystals,6 which led us to think the possibility of furtherimproving ZT by enhancing the power factor through align-ing ab planes into the disk plane while keeping the lowthermal conductivity. The thermal conductivity dependencesof temperature of the as-pressed samples in both directionsare shown in Figure 2d. Over the whole temperature range,the thermal conductivity is much lower (1.06 W m-1 K-1 inthe ⊥ direction and 0.92 W m-1 K-1 in the | direction witha lattice contribution of 0.7 and 0.6 W m-1 K-1, respectively)(a Lorenz number of about 1.6 × 10-8 J2 K-2 C-2 is used forthe nanosamples) than 1.65 W m-1 K-1 (with a latticecontribution of 1.27 W m-1 K-1 (ref 6)) in the ab plane ofthe single crystals, a clear indication of the benefits ofstronger phonon scattering resulting from the smaller grains.Below 150 °C, the thermal conductivity does not increasetoo much but very quickly after 150 °C due to the bipolar

FIGURE 2. (a) Electrical conductivity, (b) Seebeck coefficient, (c)power factor, (d) thermal conductivity, and (e) ZT dependence oftemperature of the as-pressed samples in directions perpendicular(⊥) and parallel (|) to the press direction.

© 2010 American Chemical Society 3375 DOI: 10.1021/nl101156v | Nano Lett. 2010, 10, 3373-–3378

effect, and the differences between the two directionsbecome smaller and smaller with increasing temperature.Similar to the electrical conductivity, the thermal conductiv-ity is also higher in the direction perpendicular to the pressdirection, which resulted in a similar ZT dependence oftemperature all the way up to 250 °C in both directions,indicating isotropic ZT (Figure 2e). The peak ZT is about 0.85,which is about the same with that of the single crystals atroom temperature.

Considering the potentially high power factor of 47 ×10-4 W m-1 K-2 in the ab plane of the single crystals, it ispossible to enhance ZT by orienting the ab planes of thesmall grains into the disk plane. One way to orient the grainsis to press the as-pressed samples in a bigger diameter dieso that lateral flow takes place.20

Figure 3 shows the XRD patterns of the repressed samplesin the planes perpendicular (⊥) (Figure 3a) and parallel (|)(Figure 3b) to the press direction and the SEM images of thefreshly fractured surfaces of the repressed samples in theplane perpendicular (⊥) (Figure 3c) and parallel (|) (Figure3d) to the press direction. From the much stronger diffrac-tion intensities of the (006) and (0015) peaks (Figure 3a) andmuch weaker intensities of the same peaks in Figure 3b, itis very clear that reorientation of ab planes of the grains intothe disk plane took place during the repressing process. Theratios of the integrated intensities of the (006) and (0015)peaks to that of the strongest peak (015) in the ⊥ directionare 55% (I(006)/I(015)) and 44% (I(0015)/I(015)) (inset in Figure 3a),which are much higher than those from the as-pressedsamples (inset in Figure 1a) and much stronger than the 6%and 3% for the plane parallel to the press direction (inset inFigure 3b), respectively. On one hand, the (006) and (0015)peak intensity increased a lot in the disk plane after repress-ing (Figure 3a vs Figure 1a), and on the other hand theintensity of the same peaks decreased a lot in the plane

parallel to the press direction (Figure 3b vs Figure 1b). Thesignificant diffraction peak intensity difference shows thatwe have successfully developed significant anisotropy in therepressed samples, which should clearly show up in themicrostructures and physical properties (discussed below).From the SEM images (Figure 3c,d), we can clearly see thatthe grains are platelike and with the ab plane preferentiallyoriented in the disk plane, which confirms the much in-creased (00l) peak intensities in the disk plane and decreased(00l) peak intensities in the plane parallel to the pressdirection. Such a reorientation of the ab planes is clearly theresult of the lateral flow during the repressing process.20 Thelateral flow also makes the plates thinner due to the shearforce. However, these plates are still quite large (up to 5 µm)and thick (up to 0.5 µm). Comparing to the size beforerepressing, we can also see there was a significant graingrowth (in plane). The detailed microstructure investigationsof these samples by TEM are in progress and will be reportedwhen the data are available.

In Figure 4, we show the results of temperature depen-dence of electrical conductivity (σ), Seebeck coefficient (S),power factor (S2σ), thermal conductivity (κ), and ZT of therepressed samples in both the directions perpendicular (⊥)and parallel (|) to the press direction. It is clearly shown thatthe electrical conductivity (σ⊥) of the disk plane is increasedwhile σ| is decreased in comparison with those of the as-pressed samples (Figure 2a). A ratio (σ⊥/σ|) about 2.3 of theelectrical conductivity along the two directions is maintainedfrom room temperature to 250 °C, which is however muchsmaller than the 4.3 in single crystals.6 As expected, theSeebeck coefficients of the repressed samples in both direc-tions are very similar (Figure 4b), confirming the fact thatSeebeck coefficient is basically isotropic like that in the singlecrystals.6 Due to the electrical conductivity improvement,power factor S2σ of the repressed samples in the ⊥ direction

FIGURE 3. XRD patterns of the repressed Bi2Te2.7Se0.3 bulks of the planes (a) perpendicular (⊥) and (b) parallel (|) to the press direction, andSEM images of the freshly fractured surfaces of repressed samples in the direction of (c) perpendicular (⊥) and (d) parallel (|) to the pressdirection.

© 2010 American Chemical Society 3376 DOI: 10.1021/nl101156v | Nano Lett. 2010, 10, 3373-–3378

is improved to 35 × 10-4 W m-1 K-2 at room temperatureand drops with temperature to 16 × 10-4 W m-1 K-2 at 250°C while that in the | direction is decreased to 15 × 10-4 Wm-1 K-2 at room temperature and 6 × 10-4 W m-1 K-2 at250 °C. With the changes in electrical conductivity in bothdirections, the thermal conductivity of the repressed samplesalso changed (Figure 4d): κ⊥ increased from 1.06 to 1.16 Wm-1 K-1 in that the lattice contribution remained at 0.7 Wm-1 K-1 and κ| decreased from 0.92 to 0.78 W m-1 K-1 inthat the lattice contribution is about 0.58 W m-1 K-1. A ratio(κ⊥/κ|) of 1.5 is observed, which is smaller than the 2 in singlecrystals.6 ZT of the repressed samples in the disk plane isabout 0.9 at room temperature and reaches the peak of 1.04at 125 °C (Figure 4e), which is about a 22% improvementover the peak ZT (0.85) of the as-pressed samples. Incontrast, ZT of the repressed samples in the | directionremains very low for all temperatures (Figure 4e): 0.6 atroom temperature and 0.3 at 250 °C.

It is very clear that repressing improves the ZT in thedirection perpendicular to the press direction by enhancingthe power factor due to the reorientation of the ab plane ofsome of the crystals. The as-pressed samples consist of smallcrystals; each of these crystals can be regarded as a singlecrystal and thus show anisotropic thermoelectric properties.The shape of each grain is flakelike because of its cleavagecharacteristics between Te(1)-Te(1) layers. During repressing,grains easily slip along the cleavage planes and their ab planetends to be preferentially perpendicular to the press direc-tion. The more the reorientation happens, the higher thepower factor and ZT since the thermal conductivity does notincrease so much due to the grain boundary scattering. Theoverall degree of reorientation is determined by mainly theamount of lateral flows during repressing. However, we havenot reached the limit yet since (1) the grains are still notcompletely reoriented to the disk planes resulting in a notoptimal power factor, and (2) the grains are still large andhence the lattice thermal conductivity of 0.7 W m-1 K-1 isstill relatively high.

The data we report here are the typical results from manyruns. The results are repeatable within 10% under thesimilar conditions from run to run. A few typical ZT vs Tcurves are shown in Figure 5 to demonstrate the repeat-ability. To further prove that the process is repeatable, wehave been using samples from different runs to make p andn unicouples to measure power generation efficiency. TheZT repeatability on samples from different locations of thesame run and from different runs is assured by the similarpower generation efficiency that is going to be publishedseparately.

With the crystal reorientation, the thermoelectric proper-ties are clearly improved. A possible concern is the mechan-ical properties especially the cleavage problems encounteredin single crystals due to the nature of the layered structureof Bi2Te3. Fortunately, we do not have such a problem sincethese bulk samples contain grains of still fairly small size,

FIGURE 4. (a) Electrical conductivity, (b) Seebeck coefficient, (c)power factor, (d) thermal conductivity, and (e) ZT dependence oftemperature of the repressed samples in both the directions per-pendicular (⊥) and parallel (|) to the press direction.

© 2010 American Chemical Society 3377 DOI: 10.1021/nl101156v | Nano Lett. 2010, 10, 3373-–3378

another benefit of the nanosize, even though they are notsmall enough for a much lower thermal conductivity yet.

Conclusions. In summary, we have achieved an about22% improvement in peak ZT from 0.85 to 1.04 at 125 °Cin n-type Bi2Te2.7Se0.3 by repressing the as-pressed samples.The main improvement is the large increase of electricalconductivity and small increase of thermal conductivitywhile the Seebeck coefficient does not change too much. Thereason of such improvement is the reorientation of ab planesof the small crystals into the disk plane to improve theelectrical conductivity. Further improvement in ZT is ex-pected by promoting more ab reorientation into the diskplane and reducing the crystal plate size and thickness. Eventhough the crystals are partially oriented and the propertiesare anisotropic, the mechanical properties are strong due tothe small grains and do not show any cleavage problemencountered in single crystals.

Acknowledgment. This work is supported by “Solid StateSolar-Thermal Energy Conversion Center (S3TEC)”, an En-ergy Frontier Research Center funded by the U.S. Depart-ment of Energy, Office of Science, Office of Basic EnergySciences under Award Number: DE-SC0001299 (G.C. andZ.F.R.).

Note Added after ASAP Publication. This paper pub-lished ASAP July 30, 2010 with the plane (0015) displayedincorrectly throughout the article. The correct version pub-lished on August 4, 2010.

REFERENCES AND NOTES(1) Bell, L. E. Science 2008, 321, 1457.(2) Goldsmid, H. J. Thermoelectric Refrigeration; Plenum Press: New

York, 1964.(3) Nolas, G. S; Sharp, J.; Goldsmid, H. J. Thermoelectrics: Basic

Principles and New Materials Developments; Springer Series inMaterial Science 45; Springer: Berlin, 2001; pp 111-130.

(4) Scherrer, H.; Scherrer, S. Bismuth Telluride, Antimony Telluride,and Their Solid Solutions. In CRC Handbook of Thermoelectrics;Rowe, D. M., Ed.; CRC Press: Boca Raton, FL, 1995.

(5) Drabble, J. R.; Goodman, C. H. L. J. Phys. Chem. Solids 1958, 5,142.

(6) Carle, M.; Pierrat, P.; Lahalle-Gravier, C.; Scherrer, S.; Scherrer,H. J. Phys. Chem. Solids 1995, 56, 201.

(7) Liu, W. S.; Zhang, B.-P; Li, J.-F.; Zhao, L. D. J. Phys. D: Appl. Phys.2007, 40, 566.

(8) Dresselhaus, M. S.; Chen, G.; Tang, M.; Yang, R.; Lee, H.; Wang,D. Z.; Ren, Z. F.; Fleurial, J. P.; Gogna, P. Adv. Mater. 2007, 19,1043.

(9) Ben-Yehuda, O.; Gelbstein, Y.; Dashevsky, Z.; Shuker, R.; Dariel,M. P. Proc. Int. Conf. Thermoelectr., 25th 2006, 492.

(10) Zhao, L. D.; Zhang, B.-P.; Li, J.-F.; Zhang, H. L.; Liu, W. S. SolidState Sci. 2008, 10, 651.

(11) Poudel, B.; Hao, Q.; Ma, Y.; Lan, Y. C.; Minnich, A.; Yu, B.; Yan,X.; Wang, D. Z.; Muto, A.; Vashaee, D.; Chen, X. Y.; Liu, J. M.;Dresselhaus, M. S.; Chen, G.; Ren, Z. F. Science 2008, 320, 634.

(12) Ma, Y.; Hao, Q.; Poudel, B.; Lan, Y. C.; Yu, B.; Wang, D. Z.; Chen,G.; Ren, Z. F. Nano Lett. 2008, 8, 2580.

(13) Zhu, G. H.; Lee, H.; Lan, Y. C.; Wang, X. W.; Joshi, G.; Wang, D. Z.;Yang, J.; Vashaee, D.; Guilbert, H.; Pillitteri, A.; Dresselhaus, M. S.;Chen, G.; Ren, Z. F. Phys. Rev. Lett. 2009, 102, 196803.

(14) Jiang, J.; Chen, L. D.; Bai, S. Q.; Yao, Q.; Wang, Q. Mater. Sci. Eng.,B 2005, 117, 334.

(15) Hong, S. J.; Lee, Y. S.; Byeon, J. W.; Chun, B. S. J. Alloys Compd.2006, 414, 146.

(16) Kim, S. S.; Yamamoto, S.; Aizawa, T. J. Alloys Compd. 2004, 375,107.

(17) Im, J. T.; Hartwig, K. T.; Sharp, J. Acta Mater. 2004, 52, 49.(18) Kim, T. S.; Chun, B. S. Mater. Sci. Forum 2007, 534-536, 161.(19) Tang, X. F.; Xie, W. J.; Li, H.; Zhao, W. Y.; Zhang, Q. J.; Niino, M.

Appl. Phys. Lett. 2007, 90, 012102.(20) Ren, Z. F.; Wang, J. H.; Miller, D. J.; Goretta, K. C. Physica C 1994,

229, 137.

FIGURE 5. Demonstration of repeatability of the process reflectedon ZT dependence of temperature from different locations of thesame batch and different batches.

© 2010 American Chemical Society 3378 DOI: 10.1021/nl101156v | Nano Lett. 2010, 10, 3373-–3378