The conversion efficiency of thermoelectric material is quantified by the so-called ZT value defined as σS²T/(κ_lat + κ_ele), where σ, S, κ_ele, κ_lat, and T are the electrical conductivity, the Seebeck coefficient, electronic and lattice thermal conductivity, and absolute temperature, respectively^[1,2,3,4]. The most commonly pursued approaches to enable high ZTs include the band structure manipulation for high power factor σS² by resonant levels near the Fermi level^{[5,6,7,8,9,10,11]} or introducing band convergence^[12,13,14], and phonon scattering enhancement by alloying or nanostructuring to minimize the thermal conductivity^{[15,16,17,18,19,20,21,22,23,24,25,26,27,28,29]}.

Over the last few years, SnTe has stepped into sight as a promising thermoelectric candidate due to its less toxic and a potential substitute of PbTe^{[21, 30]}. But comparing to PbTe, SnTe suffers from a much lower ZT due to a smaller band gap (~0.18 eV) and the higher separation between the light-hole band at L and the heavy-hole band at ε (~ 0.3 eV)^[31]. In addition, the relatively high thermal conductivity also limits its thermoelectric performance. Recent reports showed that the Seebeck coefficient of SnTe could be improved considerably by Cd, Mg, Mn, or Hg doping to converge the two valence bands and enlarge the band gap^{[32,33,34,35,36,37,38,39,40,41,42,43]}. In doping in SnTe has also been found to create the resonant level inside the valence band and enhance the Seebeck coefficients around room temperature^[21], which is similar to the Tl-doped PbTe system^[5,11].

On the other hand, the remarkable decrease in the lattice thermal conductivity could also lead to obviously improved thermoelectric performance of SnTe via phonon scattering by solid solution point defects, secondary phase nanoprecipitates^{[27-28,44-45]}, and mesoscale grain boundaries^[45,46]. For example, some nanoscale secondary phases such as CdS, HgTe, SrTe, Cu₂Se and Cu₂S, was introduced to regulate the thermal transport of SnTe, and the heat-carrying phonons were strongly scattered as expected^{[33-35,47-49]}. Therefore, it is necessary to combine the Seebeck coefficient enhancement and designed phonon scattering to further improve the thermoelectric performance of SnTe materials.

In this work, we try to improve the thermoelectric performance of SnTe from the coexistence of resonant levels and secondary phases scattering. Firstly, the thermal conductivity of SnTe is decreased through secondary phases of SnS by S, Se doping. Then the optimized sample with 15mol% S and Se incorporated is used for the following In doping study. The thermoelectric properties are improved by multi-doping_.

Commercial elements Sn, Te, S, Se and In were used as starting materials. The materials with nominal compositions of SnTe_1-2xS_xSe_x (x=0, 0.05, 0.1, and 0.15), and Sn_1-yIn_yTe_0.7S_0.15Se_0.15 (y=0, 0.0025, 0.005, 0.01, and 0.015) were prepared by hot pressing^[50,51]. X-ray diffraction (XRD) analysis was performed in a reflection geometry on a Bruker D8 diffractometer. The microstructures were investigated by the high resolution transmission electron microscope (JEOL 2100F), attached with an energy dispersive spectrometer (EDS). The samples were cut into bars with 10mm×2mm×2 mm for electrical property measurement and ϕ10mm×2mm for thermal transport property measurement. The Seebeck coefficient S and electrical conductivity σ were simultaneously measured using a ZEM-3 (ULVAC) instrument under a low-pressure helium atmosphere from room temperature to 850 K. The thermal conductivity was calculated from κ=DρC_p, where D is the thermal diffusivity, ρ is the density, and C_p is the specific heat. D was measured using the laser flash diffusivity method in a Netzsch LFA-457, ρ was measured by using Archimedes principle, and C_p was calculated from others data and our prebvious work^[19,44,50]. The Hall coefficient R_H was measured at room temperature using a physical properties measurement system (Quantum Design, PPMS-9). The carrier concentration nand the hall mobility μ were estimated according to n=1/(eR_H), and μ=σR_H by using the single parabolic band approximation, respectively.

Figure 1 shows the total thermal conductivity κ_tot and lattice thermal conductivity of SnTe_1-2xS_xSe_x (x=0, 0.05, 0.1, and 0.15) samples. The lattice thermal conductivity κ_lat is calculated by subtracting electrical thermal conductivity κ_ele from the total thermal conductivity, and the electrical thermal conductivity was calculated using the Wiedemann-Frenz relationship κ_ele=LσT. It should be noted that the Lorenz number L is roughly obtained by fitting S using a single parabolic band model. The corresponding temperature dependent electrical conductivity σ and Seebeck coefficient S are summarized in Supporting Information.

As shown in Fig.1(b), the lattice thermal conductivities of SnTe_1-2xS_xSe_x (x=0.05, 0.1, and 0.15) samples decrease significantly compared to that of pristine SnTe, which may be due to the excess phonon scattering by the solid solution and point defects. Considering that SnTe_1-2xS_xSe_x with 15mol% S and Se exhibits the lowest κ_lat,x=0.15 is used for the following In doping investigation. Typically, the lattice thermal conductivity of SnTe_0.7S_0.15Se_0.15 sample is 0.99 W•m^-1•K^-1 at 300 K, mildly decreases to a minima of 0.72 W•m^-1•K^-1 at 509 K, and then slightly increases to 1.13 W•m^-1•K^-1 at 850 K.

Figure 2 presents the total and lattice thermal conductivity as a function of temperature of Sn_1-yIn_yTe_0.7S_0.15Se_0.15(y=0, 0.0025, 0.005, 0.01, and 0.015) samples. As shown, the total thermal conductivities all decrease with increasing temperature. With increasing In doping, the total thermal conductivity significantly decreases. For example, κ_tot decreases from 3.6 W•m^-1•K^-1 for y=0 to 2.1 W•m^-1•K^-1 for y=0.015 at room temperature, and decreases from 1.9 W•m^-1•K^-1 for y=0 to 1.7 W•m^-1•K^-1 for y=0.015 at 850 K.

As shown in Figure 2(b), the lattice thermal conductivity of In-doped Sn_1-yIn_yTe_0.7S_0.15Se_0.15 increases from 0.99 Wm^-1•K^-1 for y=0 to 1.48 W•m^-1•K^-1 for y=0.0025 at room temperature, and decreases from 1.13 W•m^-1•K^-1 for y=0 to 1.01 W•m^-1•K^-1 for y=0.01 at 850 K. It is worth mentioning that the sample with y=0.0025 shows less temperature dependent, and κ_lat increases at elevated temperature, which is probably due to the bipolar diffusion. Typically, κ_lat of Sn_1-yIn_yTe_0.7S_0.15Se_0.15 sample for y=0.0025 is 1.03 W•m^-1•K^-1 at 300 K, slightly decreases to a minima of 0.96 W•m^-1•K^-1 at 592 K, and then increases to 1.10 W•m^-1•K^-1 at 850 K.

To gain deep understanding of suppressed lattice thermal conductivity in multi-doped SnTe, the microstructure of the samples was observed by using transmission electron microscope (TEM). Fig. 3(a) exhibits a medium-magnification of the multi-doped SnTe matrix along [004] orientation. Small precipitates (red circles in Fig. 3(b)) are observed in the matrix , and only one set of Bragg diffraction spots is observed in the inset selected area diffraction (SAD) pattern of the corresponding area (bottom-right in Fig.3(a)). The composition contrast can be evidenced by the energy-dispersive X-ray spectroscopy (EDS) (Fig.3(b)) obtained from the precipitates that exhibit characteristic peaks for In, S and Se.

High-resolution TEM (HRTEM) image in Fig.3(c) shows the matrix SnTe (200) and the secondary phase SnS (021). Accordingly, the respective fast Fourier transformation (FFT) image (the top-right and bottom-right inset in Fig.3(c)) shows no obvious peak splitting and reflects an endotaxial relationship between the matrix and secondary phase. In Fig.3(d), the lattice distortion is more obvious to be seen in the inverse fast Fourier transformation (IFFT) image (the bottom-right inset).

To further analyze the possible strain around the precipitates and the connection between the lattice distortions, high-quality HRTEM images in Figure 3(e-f) were carried out by geometric phase analysis, which is a semi-quantitative lattice image-processing approach for revealing spatial distribution of relative elastic strain. As shown, the strains exist in the secondary phases and extensively concentrate around them. The strains mainly result from the lattice or orientation mismatch between the matrix and the secondary phase, and it is a pervasive defect effect associated with point defects and dislocation. These high density of strains around the secondary phase playsa significant role on increasing the phonon scattering. In addition, the atomic-scale point defects and dislocation could scatter short-wavelength phonon efficiently. These multiscale phonon scattering mechanisms contribute to the low lattice thermal conductivity.

The temperature dependence of electrical conductivity for Sn_1-yIn_yTe_0.7S_0.15Se_0.15 (y=0, 0.0025, 0.005, 0.01, and 0.015) samples are shown in Fig.4(a). It may be seen that the electrical conductivity decreases with increasing temperature from 300 K to 850 K, showing a degenerate semiconductor behavior. At a particular temperature, the electrical conductivity monotonously decreases with increasing In content. For example, the electrical conductivity at room temperature significantly decreases from ~ 3480 S•cm^-1 for y=0to~910 S•cm^-1 for y=0.015, which can be attributed to the reduced carrier mobility of sample as summarized in Table 1.

In Fig.4(b), the temperature dependent Seebeck coefficients of Sn_1-yIn_yTe_0.7S_0.15Se_0.15 (y=0, 0.0025, 0.005, 0.01, and 0.015) samples are presented. It should be noted that the Seebeck coefficients of In doped samples are enhanced with respect to the SnTe_0.7S_0.15Se_0.15 over the entire temperature range. Especially, the Seebeck coefficient at room temperature increases from 7.6 μV•K^-1 for y=0 to 71 μV•K^-1 for y=0.015, and the improvement is almost an order of magnitude. With increasing temperature, the enhancement slows down, the Seebeck coefficient of SnTe_0.7S_0.15Se_0.15In_0.015 at 850 K is 180 μV•K^-1 and slightly higher than 151 μV•K^-1 of SnTe_0.7S_0.15Se_0.15. Such a subdued enhancement of Seebeck coefficient at higher temperature is mainly due to the downshift of the Fermi level, where the introduced resonant state no longer works^[34,51-52].

The power factor as a function of temperature for Sn_1-yIn_yTe_0.7S_0.15Se_0.15 (y=0, 0.0025, 0.005, 0.01, and 0.015) samples is shown in Fig.4(c). Because of the increased Seebeck coefficients, the power factors of In-doped samples are obviously enhanced as compared with that of SnTe_0.7S_0.15Se_0.15. If we focus on SnTe_0.7S_0.15Se_0.15In_0.01, the power factor is seen to rise almost linearly from 4.9 μW•cm^-1•K^-2at room temperature to 16 μW•cm^-1•K^-2 at 850 K. Such enhancement of power factor is achieved mainly resulted from the resonant level effect in In-doped SnTe_0.7S_0.15Se_0.15. In Fig. 4(d), the ZT values are presented as a function of temperature for all In-doped SnTe_0.7S_0.15Se_0.15 samples. The highest ZT value reaches 0.8 at 850 K for 1mol% In-doped SnTe_0.7S_0.15Se_0.15, which shows the largest power factor and the lowest lattice thermal conductivity.

This work demonstrats that the ZT value of SnTe can reach0.8 at 850 K by well-designed multi-doping. The improved thermoelectric performance of Sn_0.99In_0.01Te_0.7S_0.15Se_0.15 is accomplished by combination of reduced lattice thermal conductivity and enhanced power factor. The former mainly results from the secondary phase scattering by alloying Se, S, while the later mainly results from the resonant state by In doping. This work suggests that SnTe-based materials are important candidates for thermoelectric power generation and the well-designed multi- doping in SnTe is a promising approach to optimize the thermoelectric properties.