Enhanced thermoelectric properties of AgSbTe2 obtained by controlling heterophases with Ce doping

We report the enhanced thermoelectric properties of Ce-doped AgSbTe2 (AgSb1−xCexTe2) compounds. As the Ce contents increased, the proportion of heterophase Ag2Te in the AgSbTe2 gradually decreased, along with the size of the crystals. The electrical resistivity and Seebeck coefficient were dramatically affected by Ce doping and the lattice thermal conductivity was reduced. The presence of nanostructured Ag2Te heterophases resulted in a greatly enhanced dimensionless figure of merit, ZT of 1.5 at 673 K. These findings highlight the importance of the heterophase and doping control, which determines both electrical and thermal properties.

Thermoelectric (TE) power generators can be used to recycle waste heat from automobiles and incinerators. Although these devices are reliable and compact, the low energy conversion efficiency is a main drawback for TE power generation 1,2 . The efficiency of TE power generators is limited by the temperature environment and TE figure of merit (ZT) of the materials used, as follows: ZT = S 2 T/ρк, where S is the Seebeck coefficient, ρ is the electrical resistivity, κ is the thermal conductivity, and T is temperature in Kelvin 3,4 . Several efforts have been made to improve ZT.
Recently, the compound AgSbTe 2 has been investigated as a potentially important component in high-performance bulk nanostructured TE materials, such as the LAST-m, ((PbTe) m (AgSbTe 2 )) 5 and TAGS-x ((GeTe) x (AgSbTe 2 ) 100−x ) [6][7][8][9] . There is much interest in nanocomposite materials comprising heterophases which are embedded in the matrix [10][11][12][13] . AgSbTe 2 is also important by itself due to its good TE properties combined with relatively low thermal conductivity (0.6-0.7 W/mK) [14][15][16] . Wang et al. reported that AgSbTe 2 has a high ZT = 1.59 at 673 K due to its extremely low thermal conductivity 16 . Also, some papers have reported good TE properties in the non-stoichiometric compositions formed by controlling the secondary phase, such as nanoscaled Ag 2 Te and Sb 2 Te 3 17, 18 . Others have tried to not only reduce the lattice thermal conductivity but also adjust the carrier concentrations using dopants such as Na, Se, and Mn in AgSbTe 2 19-21 ; by suppressing the formation of impurity phases, they enhanced the TE properties. It is clear from the research mentioned that large-scale precipitates have a deleterious impact on TE properties, whereas the formation of nanoscale dispersion enhances these properties. Interfaces within a TE material have been shown to reduce the thermal conductivity without degrading the electrical conductivity 22,23 .
In this work, we investigated the influence of Ce doping on the microstructure and TE properties of AgSb 1−x Ce x Te 2 as we varied the Ce concentration (x = 0-0.004). We found that the proportion of Ag 2 Te was significantly influenced by the Ce dopant, leading to improvements in the ZT value. To our knowledge, this is first observation of ZT enhancement arising from controlling the Ag 2 Te phase by doping. Du et al. reported on the effect of Ce on excess Te in AgSbTe 2.01 compounds 24 . However, they did not sufficiently explain their results in terms of their impact on TE properties, particularly in relation to the morphology, or the effects of doping the compound with Ce. In comparison to their results, our results show considerably different transport properties. In practice, the carrier concentration and mobility were considerably different in our work. This difference may be attributable to a different fabrication process and the composition of the samples, which may result in different microstructures. A spark plasma sintering method was used in ref. 24., while a hot-pressing method was employed here to solidify samples, as well as the sintering temperature and pressure were also different, resulting in development of different morphologies. However, they did not report on the microstructure of their samples. In a related development, Marin et al. 25 . reported that the physical properties of single-crystalline samples of this material vary as a result of micro-fluctuations in the chemical composition. We demonstrate experimentally that doping AgSbTe 2 systems with Ce is an effective approach for improving TE performance by reducing thermal conductivity and adjusting carrier concentrations; specifically, Ce doping enables control over the proportion of Ag 2 Te heterophases in the matrix AgSbTe 2 .

Results and Discussion
The samples were obtained from sintering powders with hot-pressing method, in which the powders were obtained from melting elemental materials in quartz tubes with the nominal composition of AgSb 1−x Ce x Te 2 (0.001 ≤ x ≤ 0.004). Figure 1 shows the X-ray diffraction (XRD) pattern in Ce-doped AgSbTe 2 compounds, obtained from the powder of the sintered samples. The phase of the AgSbTe 2 and the small amount of Ag 2 Te were detected in all samples. The crystal structures of AgSbTe 2 were identified as they caused disordered NaCl structures (space group, Fm3m), in which the Ag and Sb were located at the Na site. A small amount of Ag 2 Te secondary phases were identified along with the main cubic phase. It should be noted that the exact crystal structure of AgSbTe 2 is currently a subject of debate, so we do not know, for example, whether it is crystallized in R3m, P4/mmm and etc 26,27 . We attempted to refine the AgSbTe 2 structure based on the model of the disordered NaCl structure (Fm3m) and different models of ordered structures (R3m, P4/mmm and, Fd3m) using the Rietveld refinement method. Our Rietveld refinement analysis revealed that it was not easy to distinguish the crystal structure based on the statistical results from the refinement: the R wp and χ 2 values are 7.85/2.08 in Fm m 3 , 7.91/2.10 in R m 3 , 7.71/2.45 in P mmm 4/ , and 7.76/2.05 in Fd m 3 , respectively. Thus, we consider the space group of AgSbTe 2 to maintain consistency with many other studies.
The cell parameters obtained from the refinement gradually decreased as the proportion of Ce increased. The small amount of Ce led to small changes: 0.6079 nm at x = 0; 0.6078 nm at x = 0.001; 0.6072 nm at x = 0.002; 0.6068 nm at x = 0.003; and 0.6066 nm at x = 0.004. This reduction was not attributed to the substitution of Ce for Sb or Ag, because the atomic or ionic radius of Ce is larger than that of Sb and Ag. Notably, it was found that the lattice parameters decreased with the amount of Ag 2 Te 18 . We speculate that the reduction in the lattice parameters is due to the reduction in the amount of Ag 2 Te. This speculation is supported by other observations detailed in the rest of this article.
Quantitative analysis of Ag 2 Te was carried out using the Rietveld refinement method. The results are shown in Table 1. The Ag 2 Te ratio decreased from 9.5% to 2.0%; the peaks also broadened with Ce doping. The peak broadening was analyzed using the Williamson-Hall equation:   where β is the integral breadth of the diffraction peak, λ is the X-ray wavelength, d is the average grain size, ѳ is the Bragg diffraction angle, and ε is the microscopic strain 28 . The crystal size decreased as the amount of Ce increased: 71.8 nm at x = 0; 55.9 nm at x = 0.001; 47.3 nm at x = 0.002; 35.4 nm at x = 0.003; and 34.5 nm at x = 0.004.
To determine the microstructure of the compound, SEM images were obtained (Fig. 2). The continuously segregated precipitates of Ag 2 Te in the melted ingot at the grain boundaries are clearly shown in Fig. 2(a), in which the SEM and elemental mapping images of the ingots obtained after the melting-quenching cycle are shown. It is clear that the Ag 2 Te phase formed after the solidification of AgSbTe 2 . The segregated precipitates were crushed during pulverization of the ingot and remained as isolated precipitates in the sintered samples, as shown in Fig. 2(b). The light and dark regions were Ag 2 Te precipitates and AgSbTe 2 matrix, respectively. Nano-sized precipitates were not observed in the images. Additionally, the amount of Ag 2 Te precipitate decreased as the amount of Ce increased, as shown in Fig. 2(c-e). The size of each precipitate was also reduced, in agreement with our XRD analysis. The reduced amount of Ag 2 Te was also confirmed by differential scanning calorimetry (DSC) measurements. The endothermic peaks at 423 K can be observed in the DSC curves shown in Fig. 3. The peak at 423 K corresponds to the phase transition peaks from α-Ag 2 Te to β-Ag 2 Te 29 . The intensity of the peak at 423 K is distinctly lower for the x = 0.002 and the x = 0.004 samples. This is due to the decreased Ag 2 Te ratio. We confirmed the reduction in the proportion of Ag 2 Te in Ce-doped AgSbTe 2 using a variety of analyses. At this stage, one may argue that the reduction in the amount of Sb in the studied samples may lead to the reduction of Ag 2 Te. From the phase diagram between Ag 2 Te and Sb 2 Te3 30, 31 , we can easily see that the amount of Ag 2 Te is lower in Sb-rich AgSbTe 2 . Thus, the reduction in the proportion of Ag 2 Te can be attributed to the addition of Ce. The reduction in Ag 2 Te with such a small amount of Ce may be due to the following: It is noteworthy that the rock salt-structured compound exists in a range of stable compositions between Ag 0.76 Sb 1.16 Te 2.08 , and Ag 0.88 Sb 1.12 Te 2 and the cations are disordered in Sb-rich compositions 32 . First-principles calculations also showed that Ag Sb and Sb Ag antisite defects can be favorably formed in AgSbTe 2 33 . The Sb-rich composition of AgSbTe 2 is attributed to the formation of Sb Ag . The formation of antisite defects is dependent on the difference in the electronegativity between two elements; the values for Ag and Sb are 1.93 and 2.05, respectively 34,35 . If the averaged difference of the electronegativity of the cations changes, the behavior of the antisite defect formation will also change 35,36 . It is known that the electronegativity of Ce is 1.12. Hence, the averaged difference would change as Ce is added, resulting in Ag occupying more cation sites in AgSbTe 2 and a reduction in the proportion of Ag 2 Te. It is clearly shown that ratio of cation sites occupancies of Ag to Sb in AgSbTe 2 increased with Ce doping as indicated in Table 1. Figure 4(a) and (b) show the effect of temperature on the electrical resistivity (ρ) and Seebeck coefficient (S) of Ce-doped AgSbTe 2 compounds. It is well known that the electrical resistivity depends on the carrier concentration and mobility according to ρ = 1/enμ, where μ and e are the mobility and carrier charge, respectively. Table 2 shows the carrier concentration and the mobility of Ce-doped AgSbTe 2 compounds at room temperature. The positive Seebeck coefficient and Hall coefficient indicate that the major carriers are holes. The carrier concentration increased with the amount of Ce added, whereas the mobility of the Ce-doped compound was lower than that of the undoped compound. Thus, the reduced electrical resistivity of the Ce-doped compound, except for x = 0.001, was mainly caused by changes in the carrier concentration. The reduction in the Seebeck coefficient as Ce is added can also be attributed to the carrier concentration, because S depends on S ~ p −2/3 in highly degenerate semiconductors 3 . The reduced mobility is thought to be due to the finer crystallite size in the Ce-doped AgSbTe 2 . Ag 2 Te acts as a minor carrier in the p-type AgSbTe 2 matrix, so by decreasing the proportion of Ag 2 Te, the carrier concentration can increase 17,37 .
The broad bump in the temperature dependence of the ρ and S from AgSbTe 2 is interesting. It is known that the abrupt changes at ~423 K in ρ and S are related to a phase transition from α-Ag 2 Te to β-Ag 2 Te 18 . This bump was not clearly observed in the Ce-doped compound, which may be due to the reduction in the proportion of Ag 2 Te. The values of the electrical resistivity at high temperatures were smaller for the Ce-doped compound. This is also thought to be due to the reduction in Ag 2 Te, which can be a source of scattering.
The temperature dependence of the power factor (S 2 /ρ) is shown in the inset of Fig. 4(b). The maximum value of 1.55 mW/m·K 2 at 673 K was achieved in the x = 0.002 sample. The enlarged power factor at elevated temperatures for the Ce-doped compound was mainly due to the reduced electrical resistivity.
In Fig. 4(c), the solid shapes represent the temperature dependence of the total thermal conductivity (κ), while open/empty shapes are the lattice thermal conductivity (κ ph ) of the Ce-doped AgSbTe 2 compounds. The lattice thermal conductivity can be evaluated by subtracting the electronic contribution (κ e ) from the total thermal conductivity (κ), κ ph = κ − κ e . κ e can be estimated using the Wiedemann-Franz law, κ e = LT/ρ, in which L is the Lorenz number. The value of L is largely dependent on the position of the Fermi level and the scattering mechanism. It can be calculated using: where λ is the scattering factor (λ = 0 for acoustic phonon scattering and λ = 2 for ionized impurity scattering) and , where f 0 is the Fermi distribution, ξ i s the reduced energy of the carriers, and η is the reduced Fermi energy 35,38 . Assuming acoustic phonon scattering and determining the value of η from the measured Seebeck coefficient according to Eq. (3), we were able to calculate the value of L, which is shown in the inset of Fig. 4(c): (1 ) 1 All Ce-doped samples had lower total thermal conductivity and lower lattice thermal conductivity than the ternary compound. The lattice thermal conductivity for the x = 0.002 sample was ~0.44 W/m•K at room temperature and ~0.43 W/m•K at 673 K, which is ~30% lower than the lattice thermal conductivity of the ternary compound (κ ph = 0.66 and 0.6 W/m•K at room temperature and 673 K, respectively). The total thermal conductivity   Table 2. Carrier concentration and the mobility of Ce-doped AgSb 1−x Ce x Te 2 samples.
was in the range of 0.6-0.8 W/m•K, similar to that observed in a previous report 14 . In the inset of Fig. 4(c), the measured specific heat capacity (Cp) of the AgSbTe 2 samples was used to calculate the total thermal conductivity. The values are 0.217 J/g•K at room temperature and 0.251 J/g•K at 673 K; these values are significantly higher than the literature value (Cp = 0.205 J/g•K) of AgSbTe 2 14 . The reduction in κ as Ce was added was mainly due to the reduction in κ ph . To investigate the origin of the reduced κ ph , we observed the nanoscale microstructures of the samples using TEM. The images obtained are shown in Fig. 5. The nanostructures with a size of about 10 nm were identified as Ag 2 Te, which is expected from the d-spacing (0.32 nm) of (112) 32 . This finding contrasts with the image of AgSbTe 2 matrix. Only the Ag 2 Te monoclinic phase has the same d-spacing size in the possible phases like Ag 2 Te, Sb 2 Te 3 , and AgSbTe 2 in this system. Also, if the nanodots are the Sb 2 Te 3 phase, it will be shown brightly compared to AgSbTe 2 matrix in the bright field image of TEM 39 . The difference in the size of the nanostructures was insignificant after annealing at 673 K for 24 h, as shown in Fig. 5(c) and (d). It is well known that the nanostructures are major phonon scattering centers and   13, 40, 41, 42, 20, 17, 19, 16 and 43, respectively. lead to a reduction in the level of lattice thermal conductivity. For example, 80% of phonons in PbTe have a mean free path below 100 nm and are effectively scattered by nanostructures 10 . Thus, we think that the reduced κ ph of the Ce-doped samples was mainly caused by the existence of nanostructures, as nanostructured Ag 2 Te was not observed in the ternary compound. Furthermore, the finer crystallite sizes in the Ce-doped compound also contribute marginally to the reduction in κ ph . Figure 4(d) shows the effect of temperature on ZT. The value of ZT = 0.92 at 673 K for the undoped sample is lower than the value reported in the literature (ZT = ~1.2) 19,20,24 . The measured value of κ was larger than previously reported, although the values of S and ρ were similar. We speculate that the difference is due to the use of a much higher value of Cp, although we do not know what value was used in previous studies. The maximum value of ZT was 1.5 at 673 K. This was obtained using the x = 0.002 sample. This ZT value is ~39% higher than that of the undoped sample. The thermal stability of the TE performance was analyzed after annealing at 673 K for 24 h. The ZT value was almost sustained at 1.49 at 673 K. The reductions in κ ph and ρ contributed significantly to the enhancement in ZT. The obtained ZT is as high as the highest reported values, as shown in Fig. 6, even though the report had shown very low density, from which low κ ph was obtained, and no experimental value of the specific heat capacity 16,19 . It is noteworthy that the enhanced ZT value was achieved with high density (over 97%), which means AgSbTe 2 has a range of possible applications.

Summary.
We developed Ce-doped AgSbTe 2 compounds that exhibit enhanced TE performance. The addition of Ce influenced the formation of the heterophase Ag 2 Te in the AgSbTe 2 matrix. We used a variety of methods to confirm that the proportion of Ag 2 Te was reduced as Ce was added. The change in the proportion of Ag 2 Te with the addition of Ce was thought to be responsible for the variation in the electrical and thermal properties. We obtained reduced ρ and κ values. This resulted in a particularly high ZT = 1.5 at 673 K within highly densified samples. Hence, the Ce-doped AgSbTe 2 has promising practical applications. It is remarkable that both electrical and thermal properties can be optimized by controlling the proportion of heterophases, so that high ZT values can be achieved.

Method
Materials and characterization. The Ce-doped AgSbTe 2 compounds were fabricated by a conventional melting, quenching, and hot-pressing method 6,17,18 . Each element (Ce, 99.9%; Ag, 99.999%, Sb, 99.999% and Te, 99.999% in purity) was weighed in a glove box with the nominal composition of AgSb 1−x Ce x Te 2 (0.001 ≤ x ≤ 0.004). The admixture of elements was directly loaded into quartz tubes and was sealed under Ar atmosphere. The sealed ampoules were heated to 1233 K for 10 h in a rocking furnace and quenched in ice water. The solidified ingots were crushed, sieved to particles of <45 μm, and sintered by hot pressing in Ar for 20 min at 683 K with a pressure of 100 MPa. The samples of 12.7 mm in diameter and about 15 mm in height were cut by a diamond wire saw and polished. The real compositions of the sintered samples were examined using inductively coupled plasma (ICP).
The phases of the Ce-doped AgSbTe 2 compounds were analyzed with an X-ray diffractometer (x'pert pro, Analytical) using Cu-Kα radiation (λ = 0.15406 nm). The microstructure and local composition of the ingot were investigated by field-emission scanning electron microscopy (FE-SEM, S-4800, Hitachi) with energy dispersive X-ray spectroscopy (EDS) and field-emission transmission electron microscopy (FE-TEM, Tecnai G 2 F30 S-twin, FEI) using an electron microprobe. The electrical properties were analyzed in terms of the Seebeck coefficient and resistivity measurement system (ZEM-3, ULVAC-RIKO). The thermal diffusivity was measured using the laser flash method (LFA-457, NETZSCH). The heat capacity was obtained using a differential scanning calorimeter (DSC 404 C, NETZSCH). The method of Archimedes was applied to measure the density of samples. Thermal conductivity was calculated from the density (d), heat capacity (Cp), and thermal diffusivity (a), using the equation: κ = a • Cp • d. The Hall effect measurement, conducted with a magnet of 0.55 T, provided information on the carrier concentration and mobility.