thermoelectric property studies on ... · bi 2 te 3-based alloyed thermoelectric materials have...

FULL P

APER

www.MaterialsViews.comwww.advenergymat.de

Wei-Shu Liu, Qinyong Zhang, Yucheng Lan, Shuo Chen, Xiao Yan, Qian Zhang, Hui Wang, Dezhi Wang, Gang Chen,* and Zhifeng Ren*

Thermoelectric Property Studies on Cu-Doped n-type Cu x Bi 2 Te 2.7 Se 0.3 Nanocomposites

Combining high energy ball-milling and hot-pressing, signifi cant enhance-ments of the thermoelectric fi gure-of-merit ( ZT ) have been reported for p-type Bi 0.4 Sb 1.6 Te 3 nanocomposites. However, applying the same technique to n-type Bi 2 Te 2.7 Se 0.3 showed no improvement on ZT values, due to the anisotropic nature of the thermoelectric properties of n-type Bi 2 Te 2.7 Se 0.3 . Even though texturing was effective in improving peak ZT of Bi 2 Te 2.7 Se 0.3 from 0.85 to 1.04, reproducibility from batch to batch remains unsatisfactory. Here, we show that good reproducibility can be achieved by introducing an optimal concentration of 0.01 copper (Cu) per Bi 2 Te 2.7 Se 0.3 to make Cu 0.01 Bi 2 Te 2.7 Se 0.3 samples. A peak ZT value of 0.99 was achieved in Cu 0.01 Bi 2 Te 2.7 Se 0.3 samples without texturing. With texturing by re-pressing, the peak ZT was increased to 1.06. Aging in air for over 5 months did not deteriorate but further improved the peak ZT to 1.10. The mechanism by which copper improves the reproducibility, enhances the carrier mobility, and reduces the lattice thermal conductivity is also discussed.

1. Introduction

Thermoelectric effects involve direct conversion between thermal and electrical energy by employing electrons and holes as energy carriers. Such effects are potentially useful for waste heat recovery and for thermal management of microelectronics and biological systems. [ 1 ] The energy conversion effi ciency of thermoelectric devices is governed by the dimensionless thermoelectric fi gure-of-merit ( ZT ) defi ned as ZT = ( S 2 σ / κ ) T , where S , σ , κ , and T are the Seebeck coeffi cient, electrical con-ductivity, thermal conductivity, and absolute temperature. Since S , σ , and the electronic contribution to κ are coupled via band structures (energy gap E g , charge carriers’ effective mass m ∗ , etc.) and scattering mechanisms, it is diffi cult to control the

© 2011 WILEY-VCH Verlag GmbH & Co. KGaA, WeinheAdv. Energy Mater. 2011, 1, 577–587

Dr. W.-S. Liu , Dr. Q. Y. Zhang , Dr. Y. C. Lan , Dr. X. Yan , Dr. Q. Zhang , H. Wang , Dr. D. Z. Wang , Prof. Z. F. Ren Boston CollegeChestnut Hill, MA 02467, USAE-mail: [email protected] Dr. S. Chen , Prof. G. Chen Massachusetts Institute of TechnologyCambridge, MA 02139, USAE-mail: [email protected]

DOI: 10.1002/aenm.201100149

parameters independently. [ 2 ] Therefore, ZT ∼ 1 has been regarded as a benchmark for many thermoelectric materials for a long time.

Bi 2 Te 3 -based alloyed thermoelectric materials have been studied extensively since the 1960s. [ 3 ] Single-crystal p-type Bi 2-

x Sb x Te 3 alloys and n-type Bi 2 Te 3−x Se x alloys are still the best commercially available thermoelectric materials for applications at or near room temperature. Neverthe-less, the lamellar structure and weak van der Waals bond between the two quintets makes them susceptible to easy cleavage along their basal planes, perpendicular to the c -axis, and hence impart very poor mechanical properties. For this reason, their polycrystalline counterparts are also used in commercial devices, despite their worse thermoelectric performance. [ 4–6 ] Recently, ZT values of 1.3 using elemental

chunks [ 7 ] and 1.4 using crystalline ingots [ 8 ] as the feedstocks were achieved in nanostructured p-type Bi 2−x Sb x Te 3 nanocom-posites by combining high-energy ball milling (BM) and direct current hot-pressing (dc-HP). These values correspond to a 30 ∼ 40% improvement in the peak ZT values compared with their single-crystal counterpart. Microstructure analysis by transmission electron microscopy (TEM) demonstrated that the high ZT values resulted from a signifi cantly decreased lattice thermal conductivity arising from the scattering of grain bound-aries and nanoprecipitates. [ 9 ] Nanoengineering has now become the preferred approach in enhancing thermoelectric materials’ performance. [ 10–12 ] This simple BM-HP route was also applied to fabricate polycrystalline n-type Bi 2 Te 3−x Se x nanocomposites. However, no obvious improvement in ZT value was obtained since the gain from the decreased lattice thermal conductivity was offset by the reduced power factor due to the decreased car-rier mobility and the increased carrier thermal conductivity due to the increased carrier concentration. It is reported that the n-type Bi 2 Te 3−x Se x ingots have similar high power factor to that of p-type Bi-Sb-Te ingots. [ 13 ] However, the power factor of n-type Bi 2 Te 3−x Se x single crystals is more sensitive to the lattice direc-tions. [ 14 ] The power factor along the basal plane is much higher than that perpendicular to the basal plane. This high electrical anisotropy results in a much lower average power factor for polycrystalline samples with randomly oriented grains than that of single crystals grown along the basal plane. Since the p-type

im 577wileyonlinelibrary.com

57

FULL

PAPER


Bi 2-x Sb x Te 3 single crystals are less sensitive to lattice directions, the polycrystals have a similar power factor to the single crys-tals. It has been experimentally confi rmed that reorientation of randomly oriented grains can improve the carrier mobility of n-type Bi 2 Te 3−x Se x polycrystalline samples and hence enhance their ZT values. [ 15–18 ] However, the signifi cantly increased car-rier concentration is probably related to the lattice defects gen-erated from the heavy deformation during the BM process, such as Te vacancies ( V Te ) that can provide two free electrons. [ 19 , 20 ] Recently, Jiang et al. [ 21 ] conducted an experiment in which they crushed Bi 2 Te 3−x Se x ingots fabricated by zone-melting into pow-ders, sieved them into three grades of size 96–120, 120–180 and 180–380 μ m, and fi nally sintered them into bulk samples. The results showed that the samples formed from fi ner particles had higher electron concentrations. Jiang et al . attributed this increased carrier concentration from reduced starting particle size to the increase of the V Te concentration. Another possible explanation for increased carriers is defect states related to dan-gling bonds at the grain boundary of Bi 2 Te 3−x Se x . Donor-like defect states increase the carrier concentration, and therefore decrease the Seebeck coeffi cient. Furthermore, these donor-like defects are extremely diffi cult to control, and hence probably lead to the serious irreproducibility problem related to the fab-rication parameters, such as BM time, BM energy, HP tempera-ture, heat treatment, etc. For industrial large scale production, good process reproducibility is equally as important as the high ZT value.

The effect of copper as a donor in Bi 2 Te 3 -based alloys has been widely investigated in both single-crystal bulk (via direct addition [ 22 , 23 ] and electrochemical intercalation) [ 24 , 25 ] and poly-crystalline samples. [ 26 , 27 ] Increased carrier mobility [ 23 ] and aging problems [ 27 ] have been noted in copper-doped Bi 2 Te 3 -based alloys. Our investigation concerned the effect of copper on reproducibility and thermoelectric properties of Cu x Bi 2 Te 2.7 Se 0.3 nanocomposite fabricated by combining BM and dc-HP. Re-orientation of random grains was also conducted to further improve the ZT value of Bi 2 Te 3 -based nanocomposites, as dem-onstrated previously, [ 18 ] and the effect of aging time on thermo-electric properties was also investigated. We also try to explain why copper can simultaneously improve the reproducibility, increase the carrier mobility, and decrease the lattice thermal conductivity.

2. Results

2.1. Processing Reproducibility

Figure 1 (a,b) shows the temperature-dependent electrical resis-tivity and Seebeck coeffi cient of ten batches of Bi 2 Te 2.7 Se 0.3 fabricated by ball-milling and hot-pressing methods under the same fabrication conditions. It clearly shows that both the electrical resistivity and Seebeck coeffi cient are highly irre-producible from batch to batch. It seems that some batches show typical semiconductor behavior while others behave as semimetals, which indicates that the carrier concentration of the as-fabricated Bi 2 Te 2.7 Se 0.3 samples changes from batch to batch. After extensive experimentation, we found that the irreproducibility issue is related to the ball-milling rather than

8 © 2011 WILEY-VCH Verlag Gwileyonlinelibrary.com

the hot-pressing process. It has been reported that the grain size reduction and solid-state reaction during the ball-milling process are driven by the accumulation of mechanical energy as a result of random ball-to-ball and ball-to-wall collisions. [ 28 ] Ball-milled powders are homogenous on the macroscale, but inhomogenous in microscale due to a broad particle size distri-bution, presence of nanoscale inclusions, and atomic defects. The uncontrollable atomic defects in Bi 2 Te 2.7 Se 0.3 arising from mechanical deformation are therefore considered a direct reason for the carrier concentration fl uctuation from batch to batch. The most common defects in Bi 2 Te 3 -based alloys include antisite defects of Bi in Te-sites ( Bi Te , contributes one hole per defect), vacancies at the Te-sites ( V Te , contributes two electrons per defect), and vacancies at Bi-sites ( V B i , contributes three holes per defect). Since the energy of evaporation for Te (52.55 kJ/mol) is much lower than that of Bi (104.80 kJ/mol), the evaporation of Te is much easier than that of Bi. The evap-oration of each Te leaves one Te vacancy ( V Te ) with two free electrons, as indicated in Equation (1).

Bi2Te3 = 2Bi×Bi + (3 − x) Te×

Te

+xTe (g ) ↑ +xV 2+Te + 2xe−

(1)

Because of the small difference in electronegativity between Te (2.1) and Bi (2.02), Bi can easily jump from Bi-sites to Te-sites to form antisite defects, contributing one hole as a free carrier. Most Bi 2 Te 3 single crystals or ingots with large grains are therefore intrinsically p-type. For fi ne-grained polycrystal-line samples, the dangling bonds at grain boundaries due to Te defi ciencies can also be considered as fractional-V Te , and also work as n-type doping in the same manner as the whole-V Te defects inside the grains. We also noted that vacancies at the Bi-sites and Te-sites can cancel each other out at a ratio of V Bi /V Te = 2:3, and contribute to zero net free charge. We therefore arrive at a defect formation equation, Equation (2), where vacan-cies at the Te-sites are totally neutralized by antisites atTe-sites and vacancies at Bi-sites.

Bi2Te3 =(

2 − 2

5x

)Bi×

Bi + (3 − x) Te×Te

+xTe (g ) ↑ +(

2

5xV 3−

Bi + 3

5xV 2+

Te

)×

+2

5xBi−

Te + 2

5xh+

(2)

Finally, we fi nd that the free carrier concentration when Te vacancies are the dominant type is fi ve times higher than when Bi antisite defects are dominant, even with the same concen-tration of Te evaporation. Hence, a small random fl uctuation of missing Te will generate more serious irreproducibility prob-lems in n-type Bi 2 Te 3 than in p-type Bi 2 Te 3 . It is worth noting that the situations described by Equation (1) for n-type materials and Equation (2) for p-type materials are ideal cases. In a real case, there are minor acceptor-like Bi Te or Sb Te antisite defects as well as the major donor-like V Te or V Se vacancies in n-type Bi 2 Te 3 alloying, while there are also minor donor-like V Te or V Se vacancies alongside the major acceptor-like Bi Te or Sb Te antisite defects in p-type Bi 2 Te 3 . Alloying with Sb usually increases

mbH & Co. KGaA, Weinheim Adv. Energy Mater. 2011, 1, 577–587

FULL P

APER


Figure 1 . Temperature dependence of electrical resistivity (a) and Seebeck coeffi cient (b) of ten batches of Bi 2 Te 2.7 Se 0.3 samples fabricated by the BM plus dc-HP method. Temperature dependence of electrical resistivity (c) and Seebeck coeffi cient (d) of eight batches of Cu 0.01 Bi 2 Te 2.7 Se 0.3 samples made by the same method and under the same conditions as shown in (a) and (b). (e) Coeffi cient of variation for room temperature thermoelectric properties of ten batches of Bi 2 Te 2.7 Se 0.3 samples and eight batches of Cu 0.01 Bi 2 Te 2.7 Se 0.3 samples, and (f) Seebeck coeffi cient at room temperature as a function of natural logarithm electrical conductivity for ten batches of Bi 2 Te 2.7 Se 0.3 samples and eight batches of Cu 0.01 Bi 2 Te 2.7 Se 0.3 samples.

the concentration of antisite defects at Te-sites (Sb Te ) and hence results in more holes, due to the smaller electronegative difference between Sb (2.05) and Te (2.10) than between Bi (2.02) and Te (2.10). The alloy with Se usually increases the concentration of vacancies at Te-sites (V Se ) and hence gives more electrons because Se has a lower energy of evaporation (37.70 kJ/mol) than Te (52.55 kJ/mol). The concentration of vacancies (V Te and V se ) in n-type Bi 2 Te 3-x Se x will be higher than that in p-type Sb 2-x Bi x Te 3 , which is another reason why we have

© 2011 WILEY-VCH Verlag GmAdv. Energy Mater. 2011, 1, 577–587

a more serious reproducibility problem in n-type Bi 2 Te 3-x Se x than in p-type Sb 2-x Bi x Te 3 .

The key to improving the reproducibility of n-type Bi 2 Te 2.7 Se 0.3 is to suppress the generation of Te vacancies, both whole-V Te and fractional-V Te . Reducing the energy of the ball-milling process, for example by decreasing ball-milling rotation speed or adjusting the ball-milling medium fi lling parameter, is one method of decreasing the mechanical deformation and thus reducing Te vacancy generation. However, to achieve fi ne

579bH & Co. KGaA, Weinheim wileyonlinelibrary.com

580

FULL

PAPER


Figure 2 . (a) Schematic of Cu x Bi 2 Te 3 lattice structure, (b) XRD patterns of Cu x Bi 2 Te 2.7 Se 0.3 , and (c) lattice parameters of Cu x Bi 2 Te 2.7 Se 0.3 .

grains, high-energy ball-milling is essen-tial. Since Bi 2 Te 3 -based alloys have a rhom-bohedral crystal structure with the space group R3̄m , it has been reported that copper can easily diffuse into Bi 2 Te 3 single crystal along the basal plane direction during an electro-deposition process, improving their mechanical properties. [ 31 ] This result sug-gests that copper may be a good choice to raise the formation energy of V Te and sup-press escaping Te. Figure 1 (c,d) presents the temperature dependence of electrical resis-tivity and Seebeck coeffi cient of eight batches of Cu 0.01 Bi 2 Te 2.7 Se 0.3 . The reproducibility of the Cu added n-type Cu 0.01 Bi 2 Te 2.7 Se 0.3 is obviously improved. In order to quantify the scattering behavior of the thermoelectric properties from batch to batch, a coeffi cient of variation (C v ) is calculated. Here, the C v is defi ned as the root of mean-square-deviation normalized by the mean value (

∧2 ) of scat-

tering data 2 ,

Cv =

√√√√E

((∧2 −2

)2)

/∧2

,

where E is an operator to solve the mean value. The coeffi cient of variation of electrical resis-tivity, Seebeck coeffi cient, and power factor for Cu 0.01 Bi 2 Te 2.7 Se 0.3 are 1.92%, 1.00% and 0.98%, respectively, much lower than 13.23%,
6.50% and 3.47% for Bi 2 Te 2.7 Se 0.3 , as shown in Figure 1 (e).
Figure 1 (f) shows the Seebeck coeffi cient at room tem-perature as a function of the natural logarithm of electrical conductivity of ten batches of Bi 2 Te 2.7 Se 0.3 and eight batches of Cu 0.01 Bi 2 Te 2.7 Se 0.3 prepared under identical ball-milling and hot-pressing conditions. It is interesting to note that the Seebeck coeffi cient of the ten batches of Bi 2 Te 2.7 Se 0.3 spans a wide range, while falling into a linear relationship with nat-ural logarithm electrical conductivity. The Seebeck coeffi cient of eight batches of Cu 0.01 Bi 2 Te 2.7 Se 0.3 is less scattered and behaves similarly in a linear relationship with natural loga-rithm electrical conductivity. This linear relationship between the Seebeck coeffi cient and the natural logarithm electrical conductivity indicates more fl uctuating carrier concentration and less varying carrier mobility. By using a simple model, [ 32 ] the weighted mobility was calculated, as 240 ± 6 and 258 ± 6 cm 2 V − 1 s − 1 for Bi 2 Te 2.7 Se 0.3 and Cu 0.01 Bi 2 Te 2.7 Se 0.3 , respec-tively. This indicates that the addition of copper not only sup-presses the generation of Te vacancy and hence improves the reproducibility, but also slightly enhances the carrier mobility. A similarly enhanced carrier mobility due to copper doping has also been observed in Bi 2 Te 3 -based alloy by Svechnikova. [ 23 ] The possible reason is that copper locates at the interstitial site between the two quintets and works as an electrical con-nection between them, leading to an improvement of electron transfer across the basal plane and therefore an enhancement of mobility.

© 2011 WILEY-VCH Verlag Gwileyonlinelibrary.com

2.2. Optimization of Cu Doping Concentration

Figure 2 (a) shows the typical lattice structure of Bi 2 Te 3 -based alloys in a hexagonal cell, which is characterized by the stacking layers perpendicular to the c -axis in a sequence, -Te (1) -Bi-Te (2) -Bi-Te (1) . Here the superscripts refer to the type of Te with dif-ferent chemical bonding environments. The interstitial sites formed by four-Te (1) atoms are also highlighted in Figure 2 (a). Figure 2 (b) shows the X-ray diffraction (XRD) patterns of Cu x Bi 2 Te 2.7 Se 0.3 for different copper contents. The XRD spectra demonstrates that the as-pressed Cu x Bi 2 Te 2.7 Se 0.3 compounds produced by ball-milling and hot-pressing posses a pure Bi 2 Te 3 phase, and no special preferred orientation is observed in com-parison with the standard spectra of random powder samples, indicating the randomness of the grains. The lattice constants of Cu x Bi 2 Te 2.7 Se 0.3 compounds are calculated by Rietveld refi ne-ment and are shown in Figure 1 (c). The lattice parameter c of Bi 2 Te 3 -phase increases from 30.364 to 30.402Å with increasing Cu content from x = 0 to x = 0.03 in Cu x Bi 2 Te 2.7 Se 0.3 . The most probable reason for the increased c -axis value is that Cu atoms likely enter into the interstitial site formed by four-Te (1) atoms, and increase the distance between the weak van der Waals force bounded layers. This structure has been observed in Bi 2 Te 3 single crystals with electrochemical intercalated copper. [ 24 , 25 ]

The Cu x Bi 2 Te 2.7 Se 0.3 samples consist of nanograins with diameter of less than one micrometer, as shown in Figure 3 (a).


FULL P

APER


Figure 3 . (a) A typical TEM image of Cu x Bi 2 Te 2.7 Se 0.3 samples. (b) Black and white stripes vertical to < 001 > direction with a periodicity of 30 nm, observed under the diffraction condition shown as the inset. (c) A gray scale along the dotted line in the HRTEM image, which identify a clear inter face between a black stripe (left) and white stripe (right).

Figure 3 (b) shows a typical TEM images containing parallel black and white stripes observed under the diffraction condi-tions shown in the inset. These stripes are horizontal across the < 001 > direction and their periodicity is about 30 nm. Previous HRTEM work indicated that a strain fi eld induces black and white stripes. [ 29 , 30 ] For inherent strain fi elds caused by disloca-tion or composition fl uctuation, the periodicity of the induced black and white stripes is 10 nm and those stripes are parallel to (1010) in single crystals, [ 29 ] hot-pressed Bi 2 Te 3 nanocompos-ites, [ 9 ] and Bi 2 Te 3 superlattices. [ 30 ] Here the observed black and white stripes are 30 nm wide and perpendicular to the < 001 > direction. Therefore the black and white stripes observed here should be caused by another kind of strain fi eld.

In order to fi nd the origin of the strain fi eld, EDS is meas-ured from several pairs of black and white stripes. EDS signals were collected from small regions with a diameter of 5 nm. The


calculated average concentration of copper in the black stripes at different locations across many grains is 0.4 at.% higher than that in the adjacent white stripes. It is probable that the intersti-tial copper in the lattice causes the strain fi eld, resulting in the intensity contrast between the black stripes and white stripes.

Figure 3 (c) shows a line profi le along the dotted line in the HRTEM images, which indentifi es a clear interface between a black stripe (left segment) and a white stripe (right segment). The copper concentration in the black stripe is 0.8 at.% higher than that in the white stripe. The average width of 15 fringes is d = 3.09 nm in black stripe while d = 3.04 nm in the white stripe. Obviously the lattice is expanded along the c -axis in the copper-rich regions. The lattice parameter c is 0.05 nm longer in the black stripe than that in the white stripe. The expansion along the c axis should come from the interstitial copper in Cu x Bi 2 Te 2.7 Se 0.3 . It is worth pointing out that the expansion of


5

FULL

PAPER


Figure 4 . Temperature dependence of thermoelectric properties of as-pressed Cu x Bi 2 Te 2.7 Se 0.3 , (a) electrical resistivity, (b) Seebeck coeffi cient, (c) power factor, (d) carrier concentration as a function of Cu concentration, (e) thermal conductivity, and (f) ZT of Cu x Bi 2 Te 2.7 Se 0.3 . The ZT of as-pressed Bi 2 Te 2.7 Se 0.3 from Ref. 18 was also plotted for comparison.

the fringe width depends on the copper concentration. When the copper concentration in the black stripes is close to that in the white stripes, likely 0.2 at.% difference, the change of the fringe width is too small to be measured from HRTEM images.

We also noted that the lattice expansion along < 001 > direc-tion observed by HRTEM is much larger than that observed by XRD. This is attributable to the difference between a local expansion in each unit cell and an average expansion of the bulk. Since the copper content in our sample Cu x Bi 2 Te 2.7 Se 0.3 is only 1–3 at.%, the large local expansion in the region with Cu is reduced by the non-expansion region without Cu, resulting in


a smaller average expansion. This lattice distortion is likely to work as new phonon scattering center.

Figure 4 shows the temperature-dependent thermoelec-tric properties of Cu x Bi 2 Te 2.7 Se 0.3 with different copper con-tents. The data for the copper-free sample used here is from a sample fabricated at the same time as the copper-doped sam-ples. The decreased electrical resistivity and Seebeck coeffi cient with increased copper content clearly demonstrates a donor behavior, as shown in Figure 4 (a,b). The positive temperature-dependent behavior of electrical resistivity demonstrates that all the copper-doped samples are degenerated semiconduc-tors except the copper-free sample. We also note that the peak


FULL P

APER


Seebeck coeffi cient shifts from 25, to 75, 125 and 175 ° C as the copper content increases from x = 0 to 0.01, 0.02, and 0.03 in Cu x Bi 2 Te 2.7 Se 0.3 , respectively. This is a typical behavior in ther-moelectric materials since the increased external major carriers suppress the generation of minor carriers and hence increase the onset temperature of bipolar effect. The power factor (S 2 / ρ ) is calculated from the measured electrical resistivity and Seebeck coeffi cient, which is shown in Figure 4 (c). The optimized content of copper is x = 0.01 in Cu x Bi 2 Te 2.7 Se 0.3 , which shows the highest power factor of 3150 μ Wm − 1 K − 2 . This value is much higher than the 2060 μ Wm − 1 K − 2 of the copper-free sample Bi 2 Te 2.7 Se 0.3 . It is noted that the as-pressed Bi 2 Te 2.7 Se 0.3 is not optimized for carrier concentration due to the irreproducibility issue. The maximum power factor of the copper-free Bi 2 Te 2.7 Se 0.3 with an optimized carrier concentration is ∼ 2500 μ Wm − 1 K − 2 , about 20% less than that of the optimized Cu 0.01 Bi 2 Te 2.7 Se 0.3 . The room temperature mobility, calculated from the Hall coeffi cient and electrical con-ductivity, increases from 173, to 245, 245, and 184 cm 2 V − 1 s − 1 as copper content increases from x = 0 to 0.01, 0.02, and 0.03, respec-tively. It clearly demonstrates that the enhancement in power factor is due to the improved charge carrier mobility. In order to quantify the number of free electrons contributed from each copper, the carrier concentration of as-pressed Cu x Bi 2 Te 2.7 Se 0.3 (x = 0, 0.01, 0.02, 0.03) is plotted as a function of copper con-centration, as shown in Figure 4 (d). The carrier concentration of the as-pressed Cu x Bi 2 Te 2.7 Se 0.3 is about (1–7) × 10 19 cm − 3 , depending on the copper concentration. Calculating the slope of carrier concentration as a function of copper concentra-tion, we obtained a value of 0.3 electrons/copper. This value is comparable to 0.4 electrons/copper obtained in copper-doped Bi 2 Te 2.7 Se 0.3 single crystals by direct copper addition, [ 22 ] but less than the 0.65 electrons/copper obtained in copper-doped Bi 2 Te 3 by electrochemical intercalation. [ 24 ]

Figure 4 (e,f) shows the temperature-dependent thermal con-ductivity and dimensionless thermoelectric fi gure-of-merit ( ZT ) of Cu x Bi 2 Te 2.7 Se 0.3 for different copper contents. The thermal conductivity of Cu x Bi 2 Te 2.7 Se 0.3 at room temperature is 0.846, 1.041, 1.156 and 1.574 Wm − 1 K − 1 , as Cu content increases from x = 0 to 0.01, 0.02, and 0.03 respectively, It is well known that the total thermal conductivity κ tot comprises three parts, i.e., lattice thermal conductivity κ lat , carrier thermal conductivity κ car , and bipolar thermal conductivity κ bipolar . For most heavily doped semiconductors, κ bipolar near room temperature is negli-gible, and κ lat is therefore estimated by directly subtracting κ car from κ tot . However, κ tot - κ car would be a very arbitrary estimation for κ lat in an intrinsic excitation region. The reported increase in κ tot - κ car with increasing temperatures, in a large amount of references, is a typical phenomenon that bipolar effect starts to contribute to the thermal conduction. The direct calculation of κ bipolar is diffi cult because it involves many band structure parameters, such as mobility μ and effective mass m ∗ for both valence band and conduction band, and forbidden bandgap E g , that cannot be directly measured. [ 33 , 34 ] For semiconductors with large band gap or heavy doping, κ tot - κ car can be used as a good estimation of κ lat in a wide temperature range. As a result, κ lat in intrinsic excitation region can be estimated by extrapolating the linear relationship in extrinsic region between κ lat and T − 1 when phonon-phonon scattering is the dominant scattering mechanism, fi nally κ tot - κ car - κ lat can be considered as an indirect


evaluation of κ bipolar in intrinsic excitation region. [ 34 ] Neverthe-less, both methods are not accurate enough for the estimation of κ lat for as-pressed Cu x Bi 2 Te 2.7 Se 0.3 in the intrinsic excitation region, so only κ lat near room temperature is calculated from κ lat = κ tot - κ car . Here the κ car is calculated by the Wiedemann-Franz law, i.e., κ car = L σ T, where L is the Lorenz number. For free electrons, L = 2.45 × 10 − 8 V 2 K − 2 . However, for most thermo-electric materials, the real Lorenz number is lower than 2.45 × 10 − 8 V 2 K − 2 , depending on the reduced Fermi energy ξ = E f / k B T and scattering parameter r as shown in the following, [ 33 ]

L =(

kB

e

)2 ((r + 7 /2)Fr + 5 /2(> )

(r + 3 /2)Fr + 1 /2 (> )

−[

(r + 5 /2) Fr + 3 /2 (> )

(r + 3 /2) Fr + 1 /2 (> )

]2)

(3)

where F n ( ξ ) is the Fermi integration

Fn (> ) =

∫ ∞

0

Pn

1 + eP−>dP

(4)

The reduced Fermi energy can be derived from both the car-rier concentration and Seebeck coeffi cient on the basis of single band approximation,

n = 4B

(2m∗kB T

h2

)3 /2

F1 /2 (> )

(5)

S = ±

kB

e

((r + 5 /2) Fr + 3 /2 (> )

(r + 3 /2) Fr + 1 /2 (> )− >

)

(6)

It is noted that the carrier concentration is usually obtained through the Hall measurement, by the relationship n = ϕ n H = ϕ /(eR H ) , where R H is the Hall coeffi cient and e is free charge, and ϕ is a parameter related to the scattering parameter r and ξ . It is clearly shown that the evaluation of ξ from the measured S is only related to scattering parameter r while the estima-tion of ξ from the measured n is associated with two unknown para meters m ∗ and r . So the calculation of ξ is derived from the measured S by using Equation (4) and (6), as shown in Table 1 . Here, acoustic phonon scattering has been assumed as the main carrier scattering mechanism near room temperature, i.e., r = –0.5. By applying the calculated ξ into Equation (3), the Lorenz number is obtained as 1.54 × 10 − 8 , 1.63 × 10 − 8 , 1.73 × 10 − 8 , and 1.84 × 10 − 8 V 2 K − 2 as the copper content increases from x = 0 to 0.01, 0.02, and 0.03, respectively. The κ lat of copper-free Bi 2 Te 2.7 Se 0.3 is 0.728 Wm − 1 K − 1 which is consistent with our pre-vious data. [ 18 ] There is a considerable decrease in κ lat at room temperature from 0.728 to 0.606, 0.546, and 0.501 Wm − 1 K − 1 when copper content increases from x = 0 to 0.01, 0.02, and 0.03, respectively. Our observation of nano scale strain domains (i.e., nano stripes) due to composition fl uctuation of copper, as shown in Figure 3 , could be the most possible reason for this signifi cant reduction in thermal conductivity. As a result, an enhancement in ZT is observed due to the addition of copper, as shown in Figure 4 (f). The sample Cu 0.01 Bi 2 Te 2.7 Se 0.3 shows an optimized ZT values with peak ZT value of 0.99 at 175 ° C, which is higher than 0.73 at 25 ° C for the copper free sample.


584

FULL

PAPER


Table 1. Volume density, carrier concentration n , mobility μ , effective mass m ∗ , Fermi energy E f , Lorenz number L , and lattice thermal conductivity κ lat for as-pressed Cu x Bi 2 Te 2.7 Se 0.3 at room temperature.

Samples Volume density [gcm − 3 ] n [ × 10 19 cm − 3 ] μ [cm 2 V − 1 s − 1 ] m ∗ [m 0 ] E f [10 − 1 eV] L [ × 10 − 8 V 2 K − 2 ] κ lat [Wm − 1 K − 1 ]

x = 0 7.699 0.92 173 1.12 –0.294 1.54 0.728

x = 0.01 7.770 2.28 245 0.97 0.077 1.63 0.606

x = 0.02 7.748 3.02 245 0.71 0.275 1.73 0.546

x = 0.03 7.706 6.47 182 0.53 0.527 1.84 0.510

This value is also higher than 0.85 at 150 ° C of the copper free sample with an optimized carrier concentration. [ 18 ] This enhanced ZT value suggests that slight copper addition is ben-efi cial to the Bi 2 Te 2.7 Se 0.3 system. This also explains why bar-rier layers are needed to prevent excessive diffusion of Cu from the circuit contacts into the thermoelectric legs for commercial thermoelectric cooling modules. [ 35 ]

2.3. Further Enhancement of ZT by Texturing

Our previous work has shown that partial texturing of random grains of as-pressed Bi 2 Te 2.7 Se 0.3 nanocomposites by a re-pressing process signifi cantly enhanced the ZT value in the per-pendicular direction due to enhanced electrical conductivity. [ 18 ] Thus, similar texturing fabrication processes were applied to the as-pressed Cu 0.01 Bi 2 Te 2.7 Se 0.3 samples. Firstly, an as-pressed sample with dimensions 19.05 mm in diameter and 22.7 mm in thickness was pressed at 500 ° C, and then subjected to re-pressing at 530 ° C under the protection of fl owing nitrogen gas into a bulk sample with dimensions of 25.4 mm in diameter and 12.7 mm in thickness, allowing us to conduct an anisotropy investigations on the thermoelectric properties for the same sample. The XRD pattern shows that a similar degree of ori-entation was achieved as before. [ 18 ] For simplicity, the detailed XRD patterns are not given here. Figure 5 shows the compar-ison of temperature-dependent thermoelectric properties of as-pressed and re-pressed Cu 0.01 Bi 2 Te 2.7 Se 0.3 samples in both parallel (//) and perpendicular ( ⊥ ) directions. A small anisot-ropy behavior is observed in the as-pressed Cu 0.01 Bi 2 Te 2.7 Se 0.3 sample between parallel and perpendicular directions, which is consistent with the Cu-free samples. [ 18 ] After re-pressing, an obvious ( 00l )-texture is obtained, and hence higher ani-sotropies in both electrical and thermal transport are observed. In order to quantify anisotropy in electrical transport, the ratio of electric conductivity between the perpendicular direc-tion and parallel direction σ ⊥ / σ // is calculated. The anisotropy ratios of σ ⊥ / σ // are 1.08 and 4.15 for as-pressed and re-pressed Cu 0.01 Bi 2 Te 2.7 Se 0.3 samples, respectively, compared to 1.12 and 4.40 for as-pressed and re-pressed Bi 2 Te 2.7 Se 0.3 samples calcu-lated according to the reported data in our previous work. [ 18 ] The decreased anisotropy ratios of σ ⊥ / σ // due to the addition of copper in return support our previous conclusion, i.e., copper increases the electrical bonding between the van der Waals force weakly bound layers and hence improves the electronic transport along the direction perpendicular to the basal plane. Peak ZT values of 0.94 and 0.99 were obtained in as-pressed Cu 0.01 Bi 2 Te 2.7 Se 0.3 polycrystalline samples along the parallel and

© 2011 WILEY-VCH Verlag Gwileyonlinelibrary.com

perpendicular directions, respectively, which is higher than 0.85 of Bi 2 Te 2.7 Se 0.3 with an optimized carrier concentration in our previous work. [ 18 ] However, only a small increase in ZT value from 0.99 to 1.06 is obtained in the re-pressed Cu 0.01 Bi 2 Te 2.7 Se 0.3 sample along the perpendicular ( ⊥ ) direction, which means that a signifi cant cast saving for large quantity production without the need of re-pressing.

2.4. Stability upon Aging

Deterioration of thermoelectric properties due to aging has been reported in Cu or Cu-halide doped Bi 2 Te 3 -based alloys, since copper can easily diffuse out from the interstitial position. Oxidization of copper upon exposire to air also leads to a deteri-oration of ZT values. [ 27 ] The effect of aging time on the stability of our re-pressed samples of Cu 0.01 Bi 2 Te 2.7 Se 0.3 has also been investigated. Shown in Figure 6 is the comparison of tempera-ture-dependent thermoelectric properties between a re-pressed Cu 0.01 Bi 2 Te 2.7 Se 0.3 sample that was measured immediately fol-lowing sample preparation and then measured again after being stored in air for fi ve months. The slightly increased electrical resistivity and Seebeck coeffi cient of the aged sample indicate a slightly decreased carrier concentration and possible copper dif-fusion. However, no obvious deterioration of the power factor is observed. Interestingly, the thermal conductivity measured fi ve months later is slightly lower due to the decreased contri-bution from the carriers. As a result, a slightly increased ZT value is obtained in the whole temperature region, with a peak ZT = 1.10 at 100 ° C. The temperature-dependent ZT of the re-pressed Bi 2 Te 2.7 Se 0.3 sample reported in Ref. [ 18 ] is also shown in Figure 6 (d) for comparison. The ZT value of the re-pressed Cu 0.01 Bi 2 Te 2.7 Se 0.3 in this study is almost the same as that of the re-pressed Bi 2 Te 2.7 Se 0.3 at T < 125 ° C. However, an enhanced ZT value is obtained in Cu 0.01 Bi 2 Te 2.7 Se 0.3 at temperatures higher than 125 ° C. At 250 ° C, the ZT value of Cu 0.01 Bi 2 Te 2.7 Se 0.3 is still 0.79, 38% higher than 0.57 forBi 2 Te 2.7 Se 0.3 . The higher average ZT value in the temperature range from 25 to 250 ° C is useful for thermal-to-electrical conversion effi ciency.

3. Conclusions

We have achieved a signifi cant improvement on reproducibility from batch to batch processing of Bi 2 Te 2.7 Se 0.3 by adding a small amount of Cu to fabricate Cu 0.01 Bi 2 Te 2.7 Se 0.3 . The reason for such improvement is that Cu atoms locate at interstitial sites, suppressing the escape of Te atoms, and as a consequence,


FULL P

APER


Figure 5 . Comparison of temperature-dependent thermoelectric properties of as-pressed and re-pressed Cu 0.01 Bi 2 Te 2.7 Se 0.3 in both parallel (//) and perpendicular ( ⊥ ) directions. (a) electrical resistivity, (b) Seebeck coeffi cient, (c) specifi c heat, (d) thermal diffusivity, (e) thermal conductivity, and (f) ZT.

reducing the concentration of Te vacancies. The copper atoms located at the interstitial sites between the van der Waals bound layers enhance the electrical bonding of the layers and improve the electronic transport along the direction perpendicular to the basal plane. Some parallel black and white stripes perpen-dicular to the < 001 > direction with periodicity of about 30 nm were observed in TEM images of the Cu doped Cu x Bi 2 Te 2.7 Se 0.3 , which is found to be related to the copper content fl uctuation on the nanoscale and responsible for the reduced lattice thermal conductivity. As a result, an enhanced peak ZT value of 0.94 and 0.99 were obtained in as-pressed Cu 0.01 Bi 2 Te 2.7 Se 0.3 samples along the parallel and perpendicular directions, respectively,


higher than teh value of 0.85 for Bi 2 Te 2.7 Se 0.3 with an optimized carrier concentration. The peak ZT of Cu 0.01 Bi 2 Te 2.7 Se 0.3 is fur-ther increased from 0.99 for the as-pressed sample to 1.06 for the re-pressed sample, and to 1.10 for the sample after fi ve-month aging in air at room temperature.

4. Experimental Section Synthesis : The fabrication process is similar to that in previous

reports. [ 7 , 17 , 18 ] Appropriate amounts of Cu (99.999%, Alfa Aesar), Bi (99.999%, Alfa Aesar), Te (99.999%, Alfa Aesar), and Se (99.999%, Alfa Aesar) were weighted according to the stoichiometric ratio


58

FULL

PAPER


Figure 6 . Comparison of the temperature-dependent thermoelectric properties of re-pressed Cu 0.01 Bi 2 Te 2.7 Se 0.3 samples that were measured imme-diately after preparation with those measured again after being stored in air for 150 days, (a) electrical resistivity, (b) Seebeck coeffi cient, (c) thermal conductivity, and (d) ZT. The ZT of re-pressed Bi 2 Te 2.7 Se 0.3 from Ref. 18 was also plotted for comparison.

Cu x Bi 2 Te 2.7 Se 0.3 ( x = 0, 0.01, 0.02, and 0.03), and loaded into the stainless steel ball-milling jar in a glove-box under argon atmosphere. The jar was then subjected to ball-milling for 20 h at 1200 rpm. The phase composition and microstructure of the ball-milled powders were studied by X-ray diffraction and transmission electron microscopy, respectively, which confi rmed that the ball-milled powders are fully alloyed fi ne grains with grain sizes ranging 20–50 nm. The ball-milled powders were then loaded into a graphite die with an inner diameter of 12.7 mm in a glove-box. The graphite die with the loaded powder was removed from the glove-box, and immediately sintered by dc-HP at 500 ° C for 2 min into a rod with a height of 12–13 mm. These dimensions allow us to carry out the thermal and electrical conductivity measurements along the same direction.

Phase identifi cation and microstructure characterization : X-ray diffraction measurements were conducted on a PANalytical multipurpose diffractometer with an X’celerator detector (PANalytical X’Pert Pro). The lattice parameters of Bi 2 Te 3 phase were calculated from Rietveld refi nement, which was performed using X’Pert HighScore Plus software (PANalytical, X’Pert Pro). The Cu x Bi 2 Te 2.7 Se 0.3 samples were cut by a diamond saw, mechanically polished, and ion-milled for TEM specimens. The microstructure was investigated using a transmission electron microscope (JSM2010F, JEOL).

Thermoelectric transport property measurement : The electrical resistivity ( ρ ) was measured by a reversed dc-current four-point method, while the Seebeck coeffi cient was determined by the slope of the voltage difference versus temperature-difference curve based on a static temperature difference method. The simultaneous measurement of electrical resistivity and Seebeck coeffi cient was conducted on a commercial


system (ZEM-3, ULVAC). The thermal conductivity was calculated from the relationship κ = DC p d , where D , C p , and d are the thermal diffusivity, heat capacity, and volume density, respectively. The thermal diffusivity was measured by a laser fl ash method with a commercial system (LFA447, Netzsch). The heat capacity was determined by a differential scanning calorimeter (DSC200-F3, Netzsch). The volume density was measured by an Archimedes method. The Hall coeffi cient R H measurement of the sample was carried out on a commercial system (7600, LakeShore), with a magnetic fi eld of 2 T and an electrical current of 30 mA.

In order to differentiate the properties along different directions, we use the hot-pressing direction as the reference: parallel direction (//) is defi ned as that all the properties are measured along the hot pressing direction whereas the perpendicular direction ( ⊥ ) is defi ned as that all the properties are measured in the plane that is perpendicular to the hot-pressing direction.

Acknowledgements This work is supported by “Solid State Solar-Thermal Energy Conversion Center (S 3 TEC)”, an Energy Frontier Research Center founded by the U.S. Department of Energy, Offi ce of Science, Offi ce of Basic Energy Science under award number DE-SC0001299/DE-FG02-09ER46577 (GC and ZFR).

Received: March 20, 2011 Revised: May 4, 2011

Published online: June 9, 2011


FULL P

APER


[ 1 ] L. E. Bell , Science 2008 , 321 , 1457 . [ 2 ] P. Pichanusakorn , P. Bandrau , Mater. Sci. Eng. R 2010 , 67 , 19 . [ 3 ] H. J. Goldsmid , J. Appl. Phys. 1961 , 32 , 2198 . [ 4 ] J. Y. Yang , T. Aizawa , A. Yamamoto , T. Ohta , J. Alloys & Compounds

2000 , 312 , 326 . [ 5 ] T. S. Oh , D. B. Hyun , N. V. Kolomoets , Scripta Mater. 2000 , 42 , 849 . [ 6 ] J. Jiang , L. D. Chen , S. Q. Bai , Q. Yao , Q. Wang , Scripta Mater. 2005 ,

52 , 347 . [ 7 ] Y. Ma , Q. Hao , B. Poudel , Y. C. Lan , B. Yu , D. Z. Wang , G. Chen ,

Z. F. Ren , Nano Lett. 2008 , 8 , 2580 . [ 8 ] B. Poudel , Q. Hao , Y. Ma , Y. C. Lan , A. Minnich , B. Yu ,

X. Yan , D. Z. Wang , A. Muto , D. Vashaee , X. Y. Chen , J. M. Liu , M. S. Dresselhaus , G. Chen , Z. F. Ren , Science 2008 , 320 , 634 .

[ 9 ] Y. C. Lan , B. Poudel , Y. Ma , D. Z. Wang , M. S. Dresselhaus , G. Chen , Z. F. Ren , Nano Lett. 2009 , 9 , 1491 .

[ 10 ] D. L. Medlin , G. Snyder , J. Current Opinion in Colloid & Interface Sci-ence 2009 , 14 , 226 .

[ 11 ] A. J. Minnich , M. S. Dresselhaus , Z. F. Ren , G. Chen , Energy & Environ. Sci. 2009 , 2 , 466 .

[ 12 ] J. F. Li , W. S. Liu , L. D. Zhao , M. Zhou , NPG Asia Mater. 2010 , 2 , 152 .

[ 13 ] O. Yamashita , T. Ochi , H. Odahara , Mater Res. Bulletin 2009 , 44 , 1352 .

[ 14 ] H. Scherrer , S. Scherrer , in CRC Handbook of Thermoelectrics , Rowe D. M. Ed.; CRC Press : Boca Raton, FL , 1995 .

[ 15 ] J. Seo , C. Lee , K. J. Park , Mater. Sci. 2000 , 35 , 1549 . [ 16 ] S. J. Hong , S. H. Lee , B. S. Chun , Mater. Sci. Eng. B 2003 , 98 , 232 . [ 17 ] L. D. Zhao , B. P. Zhang , J. F. Li , H. L. Zhang , W. S. Liu , Solid State

Sci. 2008 , 10 , 651 . [ 18 ] X. Yan , B. Poudel , Y. Ma , W. S. Liu , G. Joshi , H. Wang , Y. C. Lan ,

D. Z. Wang , G. Chen , Z. F. Ren , Nano Lett. 2010 , 10 , 3373 .


[ 19 ] J. M. Schultz , J. P. McHugh , W. A. Tiller , J. Appl. Phys. 1962 , 33 , 2443 .

[ 20 ] D. B. Hyun , J. S. Hwang , J. D. Shim , T. S. Oh , J. Mater. Sci. 2001 , 36 , 1285 .

[ 21 ] J. Jiang , L. D. Chen , S. Q. Bai , Q. Yao , Q. Wang , Mater. Sci. Eng. B 2005 , 117 , 334 .

[ 22 ] O. Beckman , P. Bergvall , Arkiv För Fysik 1963 , 24 , 113 . [ 23 ] T. E. Svechnikova , P. P. Konstatinov , G. T. Alekseeva , Inorgan. Mater.

2000 , 36 , 556 . [ 24 ] T. A. McCarthy , H. J. Goldsmid , J. Phys. D: Appl. Phys. 1970 , 3 , 697 . [ 25 ] J. Bludska , S. Karamazov , J. Navratil , I. Jakubec , J. Horak , Solid State

Ionics 2004 , 171 , 251 . [ 26 ] J. L. Cui , L. D. Mao , W. Yang , X. B. Xu , D. Y. Chen , W. J. Xiu , J. Solid

State Chem. 2007 , 180 , 3583 . [ 27 ] S. Fujimoto , S. Sano , T. Kajitami , J. Alloys & Compounds 2007 , 443 ,

182 . [ 28 ] W. S. Liu , B. P. Zhang , J. F. Li , J. Liu , Acta Phys. Sinica 2006 , 55 ,

0567 . [ 29 ] D. Maier , Solid State Commun. 2002 , 122 , 565 . [ 30 ] N. Peranio , O. Eibl , J. Nurnus , J. Appl. Phys. 2006 , 100 , 114306 . [ 31 ] M. A. Korzhuev , S. N. Chizhevskaya , T. E. Svechnikova ,

O. G. Karpinskii , A. V. Arakcheeva , G. U. Lubman , A. N. Milykh , Inorganic. Mater. 1992 , 28 , 1093 .

[ 32 ] W. S. Liu , B. P. Zhang , L. D. Zhao , J. F. Li , Chem. Mater. 2008 , 20 , 7526 .

[ 33 ] D. M. Rowe , C. M. Bhandari , Moden thermoelectric , Reston Publishing Company, Inc. : Reston, Virginia , 1983 .

[ 34 ] W. S. Liu , B. P. Zhang , J. F. Li , H. L. Zhang , L. D. Zhao , J. Appl. Phys. 2007 , 102 , 103717 .

[ 35 ] Y. C. Lan , D. Z. Wang , G. Chen , Z. F. Ren , Appl. Phys. Lett. 2008 , 92 , 101910 .


thermoelectric property studies on ... · bi 2 te 3-based alloyed thermoelectric materials have...

Documents