Thermoelectric (TE) materials have been attracting attention due to their potential for recycling energy using exhausted heat through the thermal-electric conversion effect, generating clean energy without polluting the environment. Recently, several oxides have been recognized as potential thermoelectric materials     . TE conversion efficiency is represented by a dimensionless figure of merit, ZT = S2σT/κ, where S, σ, κ, and T are the Seebeck coefficient, electrical conductivity, thermal conductivity, and absolute temperature, respectively. From these equations, large S and σ values and a low κ are necessary for high TE performance.
It is known that typical oxides have a low mobility (m) and high κ (particularly, κph, which is the thermal conductivity mediated by phonons), originating from their ionic bonding between light atoms and high electronegativity of oxygen. Different from these oxides, silver antimonite (AgSbO3), which has a defect pyrochlore structure composed of linear chains of AgO6 and SbO6, possesses rather high m originating from its highly-dispersed valence band and conduction band composed of Ag 5s and Sb 5s orbitals, respectively   . In addition, AgSbO3 has low stacking density and thus its κph is significantly lower than those of other oxides   . So, AgSbO3 has been investigated as a candidate n-type TE material, e.g., by addition of CuO  and by using a spark plasma sintering (SPS) method to prepare dense AgSbO3  . According to Wiggers et al., aggregated Ag islands are formed in the thermally treated AgSbO3, causing the increase in s by the electron hopping between such metallic islands  . However, the previous studies, regarding TE properties of AgSbO3, did not mention the existence of metallic Ag. Moreover, the valency of Sb (Sb3+ (4d10), Sb5+ (4d105s2)) should affect s due to generation of oxygen vacancy in the presence of Sb3+; however such discussion has not been performed. Thus, in the present paper, we characterize the AgSbO3 in details, prepared by the combination of a solid state reaction (SSR) method and a nitric acid (HNO3) treatment  . In addition, the existence of metallic silver and the ratio of Sb3+/Sb5+ in AgSbO3 are discussed in connection with σ, S, and κ.
AgSbO3 was synthesized by SSR using Ag2O (>99.0%, Kanto Kagaku) and Sb2O3 (>98.0%, Kanto Kagaku) as raw materials. The raw materials were weighed to the stoichiometric molar ratio, and were mixed by ball milling (200 rpm, milling rate) for 20 h, and then the mixed powder was calcined at 900˚C for 2 h. The partial calcined powder was uniaxially pressed into a rectangular pellet with the dimensions ~ 4 ´ 5 ´ 20 mm, followed by calcination in an electric furnace (HPM-1N, AS-ONE) in air at 900˚C for 2 h (denoted by NS). The remaining calcined powder was washed by HNO3 (5.0 mol/L, Kanto Kagaku), then dried at 90˚C for 20 h. The obtained powder was pressed into a pellet followed by calcination in the electric furnace under the same conditions, and obtained sample was denoted by NAT-NS.
The obtained calcined powders and sintered bodies, after grounding powders and pulverizing bodies, respectively, into fine powders were characterized by conventional X-ray diffraction analysis (XRD) with CuKα radiation (PANalytical, PW-1700). Synchrotron radiation powder diffraction patterns were obtained at BL02B2 in SPring-8 after pulverizing bodies into fine powders. The patterns were collected in the 2θ range of 3˚ - 73˚ in a step scanning mode, with a step length of 0.01˚. The Rietveld analyses using a RIETAN-FP software (a multipurpose pattern-fitting system) were performed on the synchrotron radiation diffraction patterns. For the analysis, the space groups Fd3m (cubic, No. A-227) and Fm-3m (cubic, No. A-225) were used as the starting structural model of AgSbO3 (a = 10.23 nm) and Ag (a = 4.0857 nm). A scanning electron microscope (SEM; JSM-6500F, JEOL) was used to observe the morphology of cross-sectional surfaces of the samples. The valency of Ag and Sb ions was evaluated by X-ray photoelectron spectroscopy (XPS; Axis-Ultra, Shimadzu).
To measure σ and S, either two or four Pt wires, respectively, were attached by an Ag-conductive region to the prepared rectangular pellet samples. The temperature dependence of S (Yokogawa, Model 7563) with a temperature gradient of 5 - 20 K and that of σ (Agilent Technologies, E5273A) were measured from 473 K to 873 K by conventional two-probe steady-state and four-probe methods, respectively. We then calculated the power factor (PF), as expressed by S2σ, for each sample. A modified Harman method was used to measure ZT at 600˚C  and then we calculated κ by using the equation, ZT = S2σT/κ.
3. Results and Discussion
The observed diffraction patternsfor NS and NAT-NS by synchrotron radiation are shown in Figure 1(a) and Figure 1(b), respectively, and indicated it adopted double phases of cubic AgSbO3 and Ag. Not shown here, but calcined AgSbO3 indicated the homogeneous defect pyrochlore structure by the conventional XRD. The detection of Ag was reasonably expected as Ag2O is thermally decom-
Figure 1. Measured synchrotron radiation diffraction patterns and patterns calculated by Rietveld analysis for the prepared samples, (a) NS and (b) NAT-NS. The dotted red plots, black lines, and blue lines represent measured patterns, calculated patterns, and the differences between the measured and calculated patterns, respectively. Crystallographic data determined by the Rietveld analysis are shown in Table 1.
posed to Ag at ~280˚C. Figure 1 also includes the calculatedpattern involving AgSbO3 and Ag, and the difference between the observed and calculated values. The molar percentages of AgSbO3 and Ag in both samples, their lattice parameters, and their compositions were determined by Rietveld refinement and are summarized in Table 1. In the NS sample, the Ag/Sb ratio is unity, in good agreement with the stoichiometric ratio (Ag1.00Sb1.00O3), although Ag exists separately in spite of the stoichiometric ratio of starting materials, Ag2O and Sb2O3. This is reasonable because Sb element volatilized during sintering   . In contrast, in the NAT-NS sample, the Ag/Sb ratio is less than unity (Ag0.98Sb1.00O3), reflecting the HNO3 treatment which extracted Ag from the AgSbO3lattice, causing the Ag defect in AgSbO3. The amounts of the separated Ag in both samples were observed to be similar. The relative densities of NS and NAT-NS pellets were also quite similar, 58.9 and 56.6%, respectively (Table 2). Those values were small, ~60% of the theoretical ones, which is considered to be attributable to the evaporation of Sb  .
SEM images of the fracture surfaces of NS and NAT-NS are shown in Figure 2(a) and Figure 2(b), respectively. The NS and NAT-NS powders, with particle sizes of ~100 - 200 nm in diameter, appeared to be aggregated, indicating that both of them were porous. The binding energies corresponding to Ag 3d5/2 of Ag0 (Ag), Ag+ (Ag2O), and Ag2+ (AgO) are quite similar, and are 368.0 - 368.3, 367.6 - 367.8, and 367.3 - 367.4 eV, respectively  . Since all the energies are in such a narrow range, it is quite difficult to perform a quantitative analysis of the Ag oxidation states from a deconvolution analysis of XPS spectra with high certainty. However, qualitatively, as shown in Figure 3(a), the spectrum peak of HNO3-treated sample (NAT-NS) shifted to the higher binding energy side (inset in Figure 3(a)), and also the spectrum had tailing in the lower binding energy region. These observations indicate that the HNO3-treated sample has more Ag
Table 1. Crystallographic data determined by the Rietveld analysis.
a. Estimated from the occupancies of Ag and Sb in AgSbO3.
Table 2. Characterization data of NS and NAT-NS, SPS samples.
Figure 2. Cross-sectional SEM images of (a) NS and (b) NAT-NS.
Figure 3. XPS spectra for (a) Ag 3d (Ag 3d3/2, Ag 3d5/2), Sb 3d3/2 of (b) NS and (c) NAT-NS. The plotted experimental data and the fitted curves (black lines) closely coincided. The deconvolution curves of Sb3+ (red lines) and Sb5+ (blue lines) are included. The Sb 3d peaks were calibrated using the C 1s peak derived from the hydrocarbon surface contaminant with a binding energy of 284.6 eV. The Sb3+/Sb5+ ratios obtained from the peak deconvolution are described in Table 2.
in AgSbO3 including Ag0 and Ag2+ oxidation states. The Sb 3d3/2 XPS spectra of NS and NAT-NS are shown in Figure 3(b) and Figure 3(c), respectively. As for the oxidation states of Sb, the peaks at 540.7 and 539.7 eV are assigned to the 3d3/2 orbital of Sb5+ and Sb3+, respectively  . To quantitatively evaluate the Sb3+/Sb5+ atomic ratios, peak deconvolution was performed using a Gaussian lineshape. The component Sb3+ and Sb5+ spectra determined by deconvolution and fitted peaks are also shown in Figure 3(b) and Figure 3(c). The atomic ratios of Sb3+/Sb5+, which were determined by the area of Sb 3d3/2 for Sb3+ divided by thatfor Sb5+, are also shown in Table 2. When we compare the Sb3+/Sb5+ ratio between NS and NAT-NS, the amount of Sb5+ increased after HNO3 treatment. This is also reasonable because HNO3 acts as an oxidant. So, it could be considered that Ag2+ also increased, and then Ag0 in AgSbO3 might increase to maintain charge neutrality (Figure 2 & Figure 3).
3.2. Thermoelectric Properties
Figure 4(a) shows the temperature dependence of σ for NS and NAT-NS. The σ values of both samples increased with increasing temperature, which is typical semiconductor behavior. In Figure 4(a), NS containing larger amount of Sb3+ with the similar amount of the separated Ag, compared to NAT-NS, indicated the higher σ. So, producing Sb3+ is the key to increase σ. This could be plausible because generating Sb3+ caused the generation of more oxygen vacancies in the material, and thus the corresponding donor levels are created in the band gap. This can provide more conduction electrons than in the case of no donor levels in the gaps. It should be noted that the Sb3+/Sb5+ ratio did affect σ more greatly than the Ag/Sb ratio. As mentioned above, NAT-NS containing Ag-defect (Ag0.98Sb1.00Oy) has more amounts Ag0 (4d105s1) and Ag2+ (4d9), meaning the increase in σ due to their open electronic-shell configurations than NS (Ag0.98Sb1.00Oy, Ag+ (4d10), closed electronic-shell configuration). However, the opposite experimental results were observed.
Temperature dependence of S for NS and NAT-NS is shown in Figure 4(b). S values of both samples were negative, indicating the occurrence of n-type conduction. The trend of the absolute values of S almost covers the Ioffe’s theory, that is, the absolute value of S decreases when σ increases. So, NAT-NS had the higher absolute values of S. Figure 4(c) shows the temperature dependence of power factor (PF), which was calculated from the formula σ ´ S2. The PF observed in this study (~1 ´ 10−5 W∙m−1∙K−2) was obtained for NS at 600˚C.
Finally, we examined the value of ZT for the NS and NAT-NS samples at 600˚C using the modified Harman method and then κ was calculated. The ZT values for NS and NAT-NS were 0.009 and 0.007, respectively. The κ value for NS was calculated to be 0.42 W∙m−1∙K−1, which is reasonably consistent with the previously reported values of 1.1 and 0.8 W∙m−1∙K−1 at room temperature and
Figure 4. Temperature dependence of (a) electrical conductivity, σ and (b) Seebeck coefficient, S, and (c) power factor (PF) for NS and NAT-NS samples.
400˚C, respectively  . The κ value for NAT-NS was 0.38 W∙m−1∙K−1, which is slightly smaller than that for NS. This might be attributable to the presence of Ag defects, causing the phonon scattering by the defect site. Although these ZT values need to be increased by around two orders of magnitude to be suitable for practical use, this study provides that the introduction of Sb3+ in AgSbO3 enhanced the thermoelectric property.
We prepared AgSbO3-based thermoelectric materials with and without the HNO3 treatment. Then the sample that had higher σ indicated the higher Z, attributable to the generation of Sb3+, in spite of smaller S and larger κ. The ZT value for NS was observed to be ~0.009 at 600˚C, which is two orders of magnitude smaller than the minimum value considered being required for practical use, which is greater than 1. To achieve a higher ZT value, we are now trying to prepare dense AgSbO3 with and without HNO3 treatment using the SPS method because it is well-known that the SPS method allows the production of dense samples, resulting in the enhanced TE properties.
The experiments at SPring-8 were performed with the approval of the Japan Synchrotron Radiation Research Institute (JASRI) (Proposal No. 2014A1008).