Thermoelectric energy conversion technology has successfully been applied in deep-space exploration for an electrical power supply,^[1] and is now attracting extensive attention to resolve the energy and environmental issues.^[2] Therefore, it is significant to develop efficient thermoelectric materials or devices, which can utilize waste heat in direct conversion. From the viewpoint of promising applications, thermoelectric materials with high dimensionless figure of merit (zT) include clathrates.^[3,4] Zintl phase,^[5,6] half-Heusler,^[7,8] filled skutterudites,^[9,10] chalcogenides,^[11–13] etc. Particularly, SnTe, as a typical representative in chalcogenides,^[14–16] is characterized by a two-valence-band structure.^[17] To some extent, such a band structure is beneficial to decouple the correlation among Seebeck coefficient S, electrical conductivity σ, and thermal conductivity κ.^[14] However, the large energy offset between upper light hole band (L band) and second lower-lying heavy hole band (Σ band),^[18] and relative higher lattice thermal conductivity restrict the improvement of thermoelectric performance in SnTe.^[19] In general, chemical doping of SnTe is an effective approach to enhance zT.^[20–26]Especially for Cd- and Hg-doping,^[21,26] which can effectively modulate the band structure of SnTe, thus enhancing zT. Taking into account the fact that Zn is the family element of Cd and Hg, the Zn-doping should also have an analogous effect. Recently, the first-principles calculation on Zn-doped SnTe indicate that Zn-doping converge the two valence bands, and distinctly reinforce S and σ.^[27] Unfortunately, the experimental study on Zn-doped SnTe does not exhibit this.^[28] Moreover, the underlying thermoelectric properties have not been clarified. Indisputably, it is urgent to systematically investigate the intrinsic thermoelectric feature of Zn-doped SnTe.

In this study, we provide direct evidences that Zn-doping in SnTe induces a noticeable enhancement of thermoelectric performance due to band convergence^[29] and precipitation. It is worth noting that the value of zT for the optimal sample with a 2% Zn content has improved by 100% and reaches ∼0.5 at 775 K. Combining the Hall measurement, we demonstrate that carrier concentration and effective mass dominate S and the power factor. The values of S increase in Zn-doped SnTe owing to the reduction of energy offset between L band and Σ band. Furthermore, the strengthened phonon scattering due to precipitates, effectively decreases the lattice thermal conductivity . Additionally, the observed weaker bipolar conduction reveals the broadening band gap. It is therefore reasonable to conclude that Zn-doped SnTe shows a great potential for thermoelectric applications.

Polycrystalline Sn_1−xZn_xTe (x = 0, 0.01, 0.02, 0.04) samples were synthesized by the conventional solid-state reaction method. Stoichiometric quantities of high purity elements ( ) were loaded into quartz tubes that were evacuated, flame-sealed, slowly heated to 1123 K, soaked at this temperature for 24 h and then subsequently quenched by water. The obtained ingot was crushed into powders and densified at 823 K under 60 MPa for 3 min by spark plasma sintering (SPS). All the disk-shaped samples obtained were 10 mm in diameter with density no less than 95% of theoretical density ( ). Powder x-ray diffraction (XRD) was carried out on a DX-2700 x-ray diffractometer using Cu Kα radiation. Structural refinements were performed by using Rietveld analysis. The composition and microstructure are determined by energy dispersive spectroscopy (EDS) and scanning electron microscopy (SEM). The temperature dependence of electrical conductivity σ and Seebeck coefficient S were carried out by using apparatus (LSR-3). The temperature dependence of thermal diffusivity and thermal conductivity κ were performed by the laser flash method (LFA 457, Netzsch). The Hall coefficient and carrier concentration n were determined by using the Van der Pauw technique.

The powder XRD patterns of Sn_1−xZn_xTe (x = 0, 0.01, 0.02, 0.04) are shown in Fig. 1(a). The main Bragg reflections of all samples confirm the rock-salt structure (space group ) as plotted in the inset of Fig. 1(a). Within the detection limit, no impurity phases were observed for . However, with x increasing to x=0.04, the peaks of the second phase arise (2θ=25.2°, 41.9°), which match with the powder XRD pattern of ZnTe. It is noted that the peak shifts of XRD are not very obvious due to the minor doping. It needs to utilize Rietveld refinement to obtain the variation of lattice parameters. Figure 1(b) shows the powder XRD and structural Rietveld refinement for x=0.02. The refined lattice parameter a of all samples are displayed in Fig. 1(c). The refinement error is ∼0.0003 Å. Apparently, the value of a monotonously decreases with increasing Zn content, in accordance with the reduction of ironic radius from Sn²⁺ to Zn²⁺. Interestingly, the variation of a follows the Vegard law well for , while it obviously deviates from this law for x=0.04, revealing that the solid solubility of Zn in SnTe is about 2%. To probe the microstructure of x=0.04, we carried out the SEM characterization. It is clearly seen that numerous precipitates exist (bright white region) embedded in the matrix (dark black region) as shown in Fig. 1(d). To determine the element distribution of precipitates, we performed the EDS analysis deriving from spots 1 and 2 in Fig. 1(e). The EDS results demonstrate Zn, Sn, and Te distribution, pointing to the chemical composition of precipitates which are rich in Zn (spot 1) while the compositions of the matrix are deficient in Zn (spot 2) as plotted in Fig. 1(f).

Fig. 1.

Figure Option ViewDownloadNew Window

Fig. 1. (color online) (a) Powder XRD patterns for Sn1−xZnxTe (x = 0, 0.01, 0.02, 0.04) at room temperature. Inset: Crystal structure of SnTe. (b) Powder XRD pattern with Rietveld refinement for the sample with x=0.02. (c) Lattice parameter a as a function of Zn content x. The solid line represents Vegardʼs law. (d) SEM image for the sample with x=0.04. (e) Magnified SEM image within the white square area in panels (d). (f) The corresponding EDS results for spot 1 and spot 2 in image (e), respectively.

Understanding the origin of the improvement of S is of extreme significance for rationally designing high-performance thermoelectric materials. The character of S can be approximated by in the degenerate case, where n is the carrier concentration, and is the effective mass of the carrier.^[26] It is surprising that Zn, although nominally isovalent with Sn, actually emerges to serve as an electron donor. The reason could be the reduction of Sn vacancies caused by Zn substitution, giving rise to the decrease of hole concentration n until x=0.02, as plotted in Fig. 2(c). Nevertheless, excessive Zn-driven precipitates increase n for x=0.04. As to effective mass , the values of for all Zn-doped samples rise. Especially for x=0.01, reaches the largest value , where is the bare electron mass.

Fig. 2.

Figure Option ViewDownloadNew Window

Fig. 2. (color online) Temperature-dependent (a) electrical conductivity σ and (b) Seebeck coefficient S for all samples. (c) Carrier concentration n and effective mass at room temperature as a function of Zn content x. (d) Room temperature Pisarenko plot, in comparison with reported data from Refs. [30] and [31]. The solid line is based on a two-band model.[30] (e) Temperature-dependent Hall coefficient and (f) power factor PF.

To gain more insights into the correlation between Seebeck coefficient S and hole concentration n, we performed the well-established Pisarenko plot obtained from a two-band model.^[30] As plotted in Fig. 3(d), the data points for pristine SnTe in this work, combined with previous reported I/Gd- and Bi-doped SnTe,^[30,31] follows the Pisarenko line well. However, the values of S for Zn-doped samples are higher than the theoretical prediction, which can be attributed to the valence band convergence induced by Zn-doping. To obtain more evidence for the band convergence, we probe the temperature-dependent Hall coefficient as shown in Fig. 2(e). With increasing the temperature, increases due to the band offset and the redistribution of carriers between L band and Σ band. The peak of , which is the feature of band convergence,^[24] shifts towards low temperature from ∼730 K (x = 0) to ∼680 K (x=0.04). Such a decreased temperature of peak indicates the reduction of . The power factor (PF) displays a large enhancement for Zn-doped samples [see Fig. 2(f)], which is determined by hole concentration n and effective mass .

Fig. 3.

Figure Option ViewDownloadNew Window

Fig. 3. (color online) Temperature-dependent (a) total thermal conductivity κ and (b) lattice thermal conductivity for all samples. (c) The as a function of 1000/T for pristine SnTe. The solid line represents linear fitting, and deviation of thermal conductivity indicates the contribution of bipolar diffusion . Inset: the as a function of Zn content x. (d) Room temperature as a function of x. The solid line corresponds to the prediction of according to the Debye–Callaway model. (e) and (f) SEM images for the samples with x=0.02 and 0.04, respectively.

The temperature dependence of total thermal conductivity κ is shown in Fig. 3(a). κ can be expressed by the sum of lattice component and mobile electronic component as . The value of can be estimated from Wiedemann–Franz law, (L is the Lorenz number estimated by a single Kane band model^[25]). The value of was calculated and plotted in Fig. 3(b). For pristine and Zn-doped samples, decreases with increasing the temperature due to phonon–phonon Umklapp scattering and then increases at elevated temperatures because of the bipolar effect. It is expected that the Umklapp scattering dominate phonon scattering above Debye temperature. Namely, is linearly proportional to 1/T. The plot of as a function of 1/T is present in Fig. 3(c). For x = 0, starts to deviate from the linear relationship as the temperature increases to ∼630 K, revealing the contribution of bipolar diffusion . The at 775 K as a function of x is illustrated in the inset of Fig. 3(c). It is noted that the contribution of gradually decreases from for x = 0 to for x=0.04. Usually, a high hole concentration n and large band gap are beneficial to suppress bipolar diffusion. The decrease of demonstrates the widened band gap due to Zn-doping, even as hole concentration reduces in this case.

The lattice thermal conductivity as a function of x at room temperature is displayed in Fig. 3(d). We analyzed the variation of on the basis of the Debye–Callaway model.^[18] At the beginning, increases for x=0.01 owing to the decrease of Sn vacancies. Amazingly, decreases rapidly for x=0.02 and deviates from the theoretical model, while increases again for x=0.04 falling on the vicinity of the prediction, indicating an extra mechanism of phonon scattering. To explore the inherent origin, we carried out the SEM analysis for x=0.02 and 0.04, as shown in Figs. 3(e) and 3(f). Dense precipitates embedded in the matrix are observed, which is responsible for the distinct decline of from (x = 0) to (x=0.02) due to the strong phonon scattering. We have also checked statistical analysis for the size distribution of precipitates. For x=0.02, majority sizes of precipitates are less than 200 nm, while for x=0.04, most sizes of precipitates are located in the range from 200 mm to 300 nm. The relative smaller size of precipitates cause the stronger phonon scattering, resulting in lower of x=0.02 than that of x=0.04.

Finally, the thermoelectric performance is evaluated by the dimensionless figure of merit , as shown in Fig. 4(a). The values of zT promptly increase with increasing the temperature. Particularly, for Zn-doped samples, the values of zT are substantially improved. The value of zT at 775 K for the pristine sample is ∼0.2; it is improved by 100% to be ∼0.5 for x=0.02. Figure 4(b) summarizes the schematic band diagram and mechanism of phonon scattering, illuminating the enhancement of thermoelectric performance. In addition, the decreased band offset and increased band gap due to Zn-doping play important roles in the improvement of Seebeck coefficient S power factor PF. Moreover, the precipitates embedded in the matrix contribute to the reduction of lattice thermal conductivity through additional phonon scattering.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. (color online) (a) Temperature-dependent dimensionless figure of merit zT for all samples. (b) Schematic diagram of band structure and schematic illustration of phonon scattering by precipitates embedded in a matrix derived from SEM images.

We have demonstrated the large enhancement of thermoelectric performance in SnTe by Zn-doping, which arises from the band engineering and embedded precipitates. The carrier concentration n and effective mass manipulate Seebeck coefficient S and power factor PF through Zn-doping. The strengthened phonon scattering by precipitates leads to the decrease of lattice thermal conductivity . The band gap widening suppresses the bipolar effect. The improved performance makes Zn-doped SnTe a promising candidate for thermoelectric applications.

The authors thank Prof. Yanzhong Pei from Tongji University for his support and discussion on Hall measurement.