Room-Temperature Thermoelectric Performance of n-Type Multiphase Pseudobinary Bi2Te3–Bi2S3 Compounds: Synergic Effects of Phonon Scattering and Energy Filtering

Bismuth telluride-based alloys possess the highest efficiencies for the low-temperature-range (<500 K) applications among thermoelectric materials. Despite significant advances in the efficiency of p-type Bi2Te3-based materials through engineering the electronic band structure by convergence of multiple bands, the n-type pair still suffers from poor efficiency due to a lower number of electron pockets near the conduction band edge than the valence band. To overcome the persistent low efficiency of n-type Bi2Te3-based materials, we have fabricated multiphase pseudobinary Bi2Te3–Bi2S3 compounds to take advantages of phonon scattering and energy filtering at interfaces, enhancing the efficiency of these materials. The energy barrier generated at the interface of the secondary phase of Bi14Te13S8 in the Bi2Te3 matrix resulted in a higher Seebeck coefficient and consequently a higher power factor in multiphase compounds than the single-phase alloys. This effect was combined with low thermal conductivity achieved through phonon scattering at the interfaces of finely structured multiphase compounds and resulted in a relatively high thermoelectric figure of merit of ∼0.7 over the 300–550 K temperature range for the multiphase sample of n-type Bi2Te2.75S0.25, double the efficiency of single-phase Bi2Te3. Our results inform an alternative alloy design to enhance the performance of thermoelectric materials.


INTRODUCTION
Two-thirds of the world's energy is wasted as heat, with lowgrade waste heat, up to 500 K, accounting for 60% of this lost energy. 1 Thermoelectric generators (TEGs) that convert thermal energy to electricity and vice versa are a pioneer recovery technique for low-grade waste heat; several promising materials have been explored for power generation applications including GeTe, 2 PbTe, 3 and silicide. 4 However, there are only a few thermoelectric materials that perform relatively well in TEGs within this temperature range, with a conversation efficiency of <10% over a narrow temperature range. 5 Therefore, further development is required to improve the conversion efficiency of current thermoelectric materials, where α, σ, κ e , and κ l are the Seebeck coefficient, the electrical conductivity, the electronic thermal conductivity, and the lattice thermal conductivity, respectively. The state-of-the-art thermoelectric material for the lowtemperature range (<500 K) is based on Bi 2 Te 3 alloys. Despite significant advances in the conversion efficiency of p-type Bi 2 Te 3 -based materials, upon reaching zT of ∼1.5 at room temperature, 6,7 the n-type pair still suffers from poor efficiency, and the zT of ∼0.9 is still the highest value at room temperature, 8 resulting in low generated power by the TEG. This is associated with the lower number of electron pockets near the conduction band edge of Bi 2 Te 3 than the valence band, failing to take advantages of the power factor, S 2 σ, enhancement through engineering the electronic band structure by convergence of multiple bands. 9 Therefore, a new approach is required to tackle its persistent low zT.
Recently, multiphase thermoelectric materials have attracted the attention of the thermoelectric research community 10 due to higher degrees of freedoms that these materials provide to design high-performance compounds through phonon scattering at interfaces, 11,12 energy filtering, 12,13 modulation doping, 14,15 and potentially benefiting from incorporating mag-netic interaction. 16,17 The energy filtering effect creates potential barriers in the electronic band structure of the matrix through interfaces with the secondary phases, resulting in an increase in the overall Seebeck coefficient. 13,18 The secondary phases with the larger band gaps and higher carrier concentrations than the matrix can increase the electrical conductivity of the multiphase compounds called modulation doping. 15,19 These strategies are adopted to improve the thermoelectric performance, mostly by enhancing the power factor (α 2 σ); however, the lattice thermal conductivity is influenced at the presence of secondary phases. 17,20 Here, we have selected a pseudobinary Bi 2 Te 3 −Bi 2 S 3 system, which contains multiphase compounds of Bi 2 Te 2 S and Bi 2 Te 3 or Bi 2 S 3 phases. Both phases of Bi 2 Te 2 S (∼0.28 eV) 21 and Bi 2 S 3 (1.3 eV) 22 have higher band gaps than Bi 2 Te 3 (∼0.15 eV). 23 We have shown that n-type multiphase pseudobinary Bi 2 Te 2 S−Bi 2 Te 3 compounds possess a higher power factor and simultaneously provide lower total thermal conductivities than single-phase counterparts, resulting in a relatively high thermoelectric figure of merit of ∼0.7 over the temperature range of 300−550 K, compared to ∼0.3 for the Bi 2 Te 3 sample. These new findings will inform the design of next-generation, multiphase compounds with enhanced thermoelectric performance for low-grade waste heat recovery applications.

Synthesis.
A set of polycrystalline Bi 2 Te 3−x S x (x = 0, 0.25, 0.5, 0.75, and 1.0) samples were synthesized by direct reaction of stoichiometric amounts of high-purity Bi (99.999%, Alfa Aesar), Te (99.999%) shots, and dried S (99.99%) powder in vacuum-sealed quartz ampoules in an Ar glove box. The ampoules were homogenized at 1123 K for 10 h, quenched in cold water, and annealed at 673 K for 72 h. The obtained ingots were hand-ground to fine powders in an agate mortar and pestle housed inside an Ar atmosphere glove box. The powders were then loaded into a graphite die and sintered under vacuum to produce rods of 6 mm in diameter and 12−14 mm in length, using spark plasma sintering at 633 K and an axial pressure of 50 MPa for 5 min. The density (ρ) of samples was measured via the Archimedes method such that the relative densities were ≥99% for all samples.
The (y = 0.005, 0.01, and 0.02) that are iodine-doped samples of multiphase Bi 2 Te 2.75 S 0.25 were fabricated to ingots by direct reaction of stoichiometric amounts of high-purity Bi (99.999%, Alfa Aesar), Te (99.999%) shots, dried S (99.99%) powder, and BiI 3 (99.999%, Alfa Aesar Puratronic), using the abovementioned heat treatment procedure. The hand-ground powders were then loaded into a graphite die and sintered under vacuum to produce pellets using spark plasma sintering (FCT System GmbH, KCE FCT-H HP D-25 SD, Rauenstein, Germany) at 673 K and an axial pressure of 50 MPa for 5 min, obtaining samples with 11 mm in diameter and ∼12 mm height. Although the same sintering parameters were used to sinter these samples, the change in the SPS equipment resulted in samples with densities (ρ) of >90% of the relative densities, lower that the densities of >99% of the theoretical density for previous samples.

Material Characterization.
The phase purity and crystal structure of sintered samples were characterized by powder X-ray diffraction (XRD) using a PANalytical X'Pert PRO X-ray diffractometer with Cu Kα radiation (λ = 1.5406 Å, 40 kV, 25 mA). The lattice parameters were determined by Rietveld refinement of the collected diffraction patterns using the FullProf program.
For the microstructure and chemical mapping, samples were cut from the sintered disk cross section of Bi 2 Te 3−x S x (x = 0 and 0.5) ingots such that the compression direction was parallel to the sample surface. The sample surface was polished up to 1 μm diamond and then subjected to Ar-ion milling on a Leica EM RES101 at 4 kV for 1 h. The grain and phase distribution and chemical composition of samples were characterized using a JEOL JSM-7001F field emission gun scanning electron microscope (SEM), equipped with an Oxford Instruments 80 mm 2 X-Max energy-dispersive spectroscopy (EDS) detector and a Nordlys-S(II) electron backscattering diffraction (EBSD) detector working concurrently with the OI Aztec acquisition software. Combined EBSD and EDS maps were collected from 480 × 360 μm 2 areas at an accelerating voltage of 15 kV, a probe current of ∼5.5 nA, and a step size of 0.24 μm. In the Bi 2 Te 3 and Bi 2 Te 2.5 S 0.5 samples, overall indexing rates of ∼91.5 and 92.2% were obtained such that most zero solutions were located at grain boundary regions. The maps were cleaned by removing wild orientation spikes and filling-in zero solutions via cyclic extrapolation down to five neighbors and processed using the OI HKL Channel-5 analytical software suite. The parameters used to collect and process the combined EBSD and EDS maps are detailed in our previous study. 24 2.3. Electronic Transport Property Measurements. The electrical conductivity and Seebeck coefficient were measured perpendicular and parallel to the compression/sintering direction of the pellet and rod samples, respectively, by cutting ∼2 × 2 × 9 mm 3 specimens from the pellet. The measurements were carried out from room temperature to 523 K under a helium atmosphere by using a Linseis LSR-3 apparatus.
The laser flash diffusivity method (Linseis LFA 1000) was used to measure the thermal diffusivity of the pellet and rod samples. The thermal diffusivity (D) of all samples was measured parallel to the compression/sintering direction. The thermal conductivities (κ) were calculated by κ = D × C p × ρ. The heat capacity (C p ) of all samples was measured according to the ASTM-E1269-11 standard test using a differential scanning calorimeter (PerkinElmer-DSC 8000). The density (ρ) of samples was measured via the Archimedes method.
The samples were cut with their long axis perpendicular to the sintering direction of samples to measure the Hall coefficient (R H ) from transverse electrical resistivity measurements using the DC transport option of a Quantum Design physical property measurement system (PPMS) at temperatures between 5 and 400 K under a ±2 T magnetic field. The Hall carrier concentration (n H ) and Hall carrier mobility (μ H ) were calculated using n H = 1/(R H ·e) and μ H = σ/(n H ·e), respectively. The room-temperature Hall coefficient (R H ) of doped samples was measured through van der Pauw method, using an ECOPIA-HMS-3000 apparatus to obtain the room-temperature Hall carrier concentrations and electronic mobilities.
Te 2 ] 2/3 sequence with the subscripts indicating the z translation of the blocks within the hexagonal unit cell. 26 In the case of Bi 2 Te 2 S, it was initially proposed that the S atoms substituted at Te sites. 27 However, this structure was unstable at that composition due to the high strains on the internal sulfur layer caused by the electronegative anion of sulfur and the very different ionic radii of Te and S, resulting in a large size mismatch of the hexagonal-close-packed Te and S layers. Therefore, a more complex formula, Bi 14 Te 13 S 8 , was suggested for the single-phase structure where S only occupies the Te 1 sites in the Bi 2 Te 3 structure. 26,28 This phase has an orthorhombic crystal structure, belongs to the R3̅ space group with lattice parameters of a = b = 11.24 Å and α = β = γ = 56.83°or a corresponding hexagonal structure with lattice parameters of a = 4.25 Å, c = 29.6 Å, α = β = 90°, and γ = 120°. 28 The stoichiometrically prepared sample of Bi 2 Te 2 S contains less sulfur than single-phase Bi 14 Te 13 S 8 , and therefore, as XRD results indicate, it contains a secondary Bi 2 Te 3 phase within the Bi 14 Te 13 S 8 matrix.
It is worth noting that this crystal structure has been indexed as Bi 2 Te 2 S in the XRD databases of the Inorganic Crystal Structure Database (ICSD) and Open Quantum Materials Database (OQMD). Consequently, we used this to identify phases in our previous work, 24 and in fact, the phase labeled the Bi 2 Te 2 S phase in our previous report 24 is the Bi 14 Te 13 S 8 phase. We acknowledge the incorrect labeling of the Bi 14 Te 13 S 8 phase as Bi 2 Te 2 S in our previous study and refer to it as Bi 14 Te 13 S 8 in this work. Rietveld refinement was also employed via the FullProf software suite to determine the proportion of phases, summarized in Table 1.
The electronic transport properties of bismuth chalcogenides are affected by the fraction of individual phases. 7,29 Therefore, we employed a combination of an EBSD technique and EDS mapping to discriminate the Bi 2 Te 3 and Bi 14 Te 13 S 8 phases. 24 Figure 3a,c shows the band contrast maps of the Bi 2 Te 3−x S x (x = 0 and 0.5) sample cross section, the surface parallel to the sintering direction, respectively. The low-angle grain boundaries (LAGBs) are defined as misorientations between 2 ≤ θ ≤ 15°in blue and high-angle grain boundaries (HAGBs) as misorientations >15°in black. Since the powder was handground from an annealed ingot in an agate mortar and pestle, it resulted in a large variation in grain sizes, with some grains exceeding 100 μm. In general, most large grains are elongated perpendicular to the direction of sintering. The latter is along the map vertically. Figure 3d shows the phase distribution maps of Bi 2 Te 3 (red) and Bi 14 Te 13 S 8 (blue) phases in the multiphase sample of Bi 2 Te 2.5 S 0.5 . Although the same fabrication method was employed to prepare these samples, the grain sizes in the multiphase Bi 2 Te 2.5 S 0.5 sample are much smaller than in singlephase Bi 2 Te 3 (compare Figure 3a,c). The phase diagram of Bi 2 Te 3 −Bi 2 S 3 25 (Figure 1) shows that there is a eutectic transformation between Bi 2 Te 3 and Bi 14 Te 13 S 8 , following which a lamellar structure is expected in the final ingot. This has resulted in much finer grain size in the multiphase sample of Bi 2 Te 2.5 S 0.5 than in the single-phase Bi 2 Te 3 sample.  Figure 4 shows the temperature-dependent thermoelectric properties of the undoped Bi 2 Te 3−x S x (x = 0, 0.25, 0.5, 0.75, and 1) samples between 300 K and 500 K. All thermoelectric properties of this set of samples were measured parallel to the sintering direction.
All samples show negative Seebeck coefficient values, which indicates that electrons compose most charge carriers. The Seebeck coefficient (Figure 4a) of all samples with x = 0.5 and 0.75 decreases with temperature, exhibiting a temperature dependence behavior typical of intrinsic semiconductors. The Bi 2 Te 2 S and Bi 2 Te 3 samples show a bipolar effect within the range of 400−500 K that is typical behavior of narrow band gap degenerate semiconductors 30 where the intrinsic carriers  are excited and compete with the extrinsic carriers, which in turn decreases the Seebeck coefficient with increasing the temperature as a result of the combined contribution of negative and positive charge carriers. The electrical conductivity (Figure 4b) of all samples, except for samples with x = 0 and 0.25, increases with temperature and exhibits a temperature dependence behavior, typical of intrinsic semiconductors. The highest value of power factors, ∼1.75 mW/m·K 2 (Figure 4c) was obtained in the multiphase Bi 2 Te 2.75 S 0.25 sample (containing roughly 16 wt % of Bi 14 Te 13 S 8 in the matrix of Bi 2 Te 3 ) at around room temperature with a Seebeck coefficient of ∼150 μV/K and above 1.5 mW/m·K 2 over the whole temperature range. This marks a significant increase from ∼1.25 mW/m·K 2 for the single-phase Bi 2 Te 3 .
The temperature-dependent Hall coefficient, R H , of , explaining the higher electrical conductivity and the behavior, typical of a degenerate semiconductor. The optimum thermoelectric performance of n-type Bi 2 Te 3 is usually achieved at carrier concentrations in the lower range of ×10 19 cm −3 , 31 similar to the single-phase sample of Bi 2 Te 3 and the multiphase sample of Bi 2 Te 2.75 S 0.25 . All samples of this study are undoped, and these results suggest that the variations in the charge carrier concentrations in these materials might not be solely due to a substitutional effect of Te with S in Bi 2 Te 3 but likely due to the random formation of antisites and vacancy defects and various ratios of each phase in these samples. Table 2 summarizes the Seebeck coefficient,  μV/K, much higher than the value of ∼120 μV/K for single-phase Bi 2 Te 3 at room temperature (Figure 4a). This can be explained by the energy filtering effect, where the potential barriers generated in the electronic band structure of the matrix through interfaces with the secondary phases increases the overall Seebeck coefficient. 13,18 In this effect, low energy carriers are scattered by potential barriers formed at the junction of the two phases. The barrier height can be approximated as a function of the difference between the electron affinity of two phases. The multiphase samples of Bi 14 Te 13 S 8 −Bi 2 Te 3 (Bi 2 Te 3−x S x (x = 0.25, 0.5, 0.75, and 1)) have an estimated barrier height of ≈ 0.7 eV, considering the electron affinity, χ, of ≈5.26 eV for Bi 14 Te 13 S 8 32 and ≈4.5 eV for Bi 2 Te 3 , 33 schematically demonstrated in Figure 6.   The temperature dependence of the total thermal conductivity, κ, between 300 and 500 K for all Bi 2 Te 3−x S x (x = 0, 0.25, 0.5, 0.75, and 1) samples is presented in Figure 4d. The total thermal conductivity of all samples increases with temperature, indicating low carrier concentrations in these samples. In general. all multiphase samples of Bi 2 Te 3−x S x (x = 0.25, 0.5, 0.75, and 1) show the total thermal conductivities much lower than the single-phase Bi 2 Te 3 sample. The total thermal conductivity of the highly conductive multiphase Bi 2 Te 2.75 S 0.25 sample, with similar charge carrier concentration to Bi 2 Te 3 , is below ∼1.1 W/m·K over the temperature range as a result of the finer microstructure ( Figure 3).
To understand the effect of the microstructure on the lattice thermal conductivity of these samples, a multiband parabolic model with one electron (n) band and one hole (p) band was employed to calculate the electronic and bipolar contributions of the thermal conductivity for all samples. 34 The multiband model considers both holes and electrons, and the overall Seebeck coefficient and electrical conductivity of samples can be described by where α n , α p , σ n , and σ p are the partial Seebeck coefficient and electrical conductivity of electrons and holes, respectively. The electronic (κ e ) and bipolar (κ b ) contributions of the thermal conductivity are given as (y = 0.005, 0.01, and 0.02), with various charge carrier concentrations. The thermoelectric performance of these samples was measured for both directions parallel and perpendicular to the sintering orientation.
3.3. Electronic Transport Properties of (y = 0.005, 0.01, and 0.02) Samples. The temperature-dependent Seebeck coefficient, electrical conductivity, total thermal conductivity, and figure of merit of iodine-doped multiphase Bi 2 Te 2.75 S 0.25 compound, measured parallel to the sintering direction at the temperature range of 300−530 K, are shown in Figure 9a−d, respectively. The thermoelectric performance of  these samples was also measured at the direction perpendicular to the sintering direction (detailed in the Supporting Information). The electronic transport properties of these samples are compared with the intrinsic sample, shown in Figure 4. The samples are named with their Hall carrier concentrations in Figure 9 for simplicity of comparison. The electrical conductivity has been increased and the Seebeck coefficient has been reduced by the increase in the carrier concentration, as expected; the intrinsic sample has shown only slightly lower charge carrier concentrations (1.2 × 10 19 cm −3 ) than the slightly doped sample (1.7 × 10 19 cm −3 ); however, it has higher electrical conductivity and higher thermal conductivity than the slightly doped samples. We have used the same fabrication method to prepare doped samples as the intrinsic ones; however, the SPS equipment used to sinter these samples was changed, which had lower accuracy in the pressure control system at lower ranges, resulting in samples with densities (ρ) more than 90% of the relative densities, much lower than the density of intrinsic samples (>99% of the theoretical density) that were prepared by different equipment. The porosities in the doped samples resulted in lower electrical conductivity of samples with the same chemistry and carrier concentrations due to scattering of electrons. 36 However, it also reduced the thermal conductivity (Figure 9c) of samples due to the scattering of phonons in the porous structure. This resulted in a similar figure of merit for both samples, showing that optimum charge carrier concentrations for this composition is also in the lower range of ×10 19 cm −3 , similar to the single-phase n-type Bi 2 Te 3 sample. 31 The intrinsic Bi 2 Te 3 sample prepared in this study (Table 2 and Figure 4) showed the charge carrier concentrations at the lower range of ×10 19  (y = 0.005, 0.01, and 0.02) samples between 300 and 500 K measured parallel to the sintering direction, compared with the intrinsic sample shown in Figure 4. cm −3 , proving a reasonable comparison of the thermoelectric performance of multiphase samples with the single-phase ntype Bi 2 Te 3 .
These results suggest that there is a possibility to obtain high room temperature and/or over a wide temperature range thermoelectric performance in the multiphase pseudobinary Bi 14 Te 13 S 8 −Bi 2 Te 3 alloys. These samples exhibit a high power factor and simultaneously provide low total thermal conductivities.

CONCLUSIONS
In summary, multiphase pseudobinary Bi 14 Te 13 S 8 −Bi 2 Te 3 compounds are promising high-performance thermoelectric materials around room temperature. We have shown that the finely distributed secondary phase of Bi 14 Te 13 S 8 with a larger band gap and electron affinity than the matrix Bi 2 Te 3 increased the Seebeck coefficient through the energy filtering effect. This resulted in a higher power factor of multiphase compounds relative to single-phase Bi 2 Te 3 . Simultaneously, the increased number of interfaces between the two phases formed through eutectic transformation during solidification of the compounds added additional scattering centers for phonons, leading to a decrease in the lattice thermal conductivity of multiphase samples. The combination of these beneficial effects resulted in a reasonably high figure of merit of ∼0.7 over the temperature range of 300−550 K for the multiphase sample of Bi 2 Te 2.75 S 0.25 , doubling the efficiency of single-phase Bi 2 Te 3 .
More detailed information including XRD results, heat capacities, electronic transport properties of samples perpendicular to the direction of sintering, and detailed multiband model used to define bipolar and lattice thermal conductivities of samples (PDF)