Vanadium-free colusites Cu26A2Sn6S32 (A Nb, Ta) for environmentally friendly thermoelectrics

Research Institute for Energy Conservation, Science and Technology (AIST), Tsukuba, michihiro@aist.go.jp Department of Applied Science for Elect Graduate School of Engineering Sciences, K 8580, Japan Department of Quantum Matter, Graduate Hiroshima University, Higashi-Hiroshima, H Institute for Advanced Materials Research, H Hiroshima 739-8530, Japan † Electronic supplementary informa 10.1039/c6ta05945g Cite this: J. Mater. Chem. A, 2016, 4, 15207


Introduction
In aiming at mitigating climate change, thermal management should be improved in various elds such as vehicles, industrial processes, and fossil fuel combustion, because more than 50% of the primary energy consumed is wasted in the form of heat. 1 Solid-state devices based on thermoelectrics can directly generate electrical energy from the waste heat, and thus provide a new strategy for reducing and managing energy consumption. [2][3][4][5] For a thermoelectric device, the efficiency in converting thermal energy to electrical energy depends on the thermoelectric gure of merit ZT of the material, described as: where S is the Seebeck coefficient, r is the electrical resistivity, k el is the electronic thermal conductivity, k lat is the lattice thermal conductivity, and T is the absolute temperature. The ZT is improved through the enhancement of thermoelectric power factor S 2 /r and reduction of k lat . The potential materials for high performance thermoelectrics are Bi 2 Te 3 for near-room-temperature applications and PbTe for intermediate-temperature ($700 K) applications. 6 In particular, the ZT of PbTe-based materials has been dramatically enhanced by nanostructuring. [7][8][9][10][11][12][13][14][15] Recently, high ZT values were also reported for new class thermoelectric materials such as SnSe 16,17 and MgAgSb. 18 Industrial application of thermoelectric devices requires high-ZT materials composed of less-toxic and cost effective elements. This requirement is, however, not satised for conventional thermoelectric materials such as Bi 2 Te 3 and PbTe, because they are composed of toxic (Pb) and expensive (Te) elements. On the other hand, copper (Cu) and sulfur (S) are earth abundant and less-toxic elements; therefore, copper-containing suldes have been intensively studied in the past several years. 19 Binary copper suldes Cu 1.97 S have been reported to have ZT over $1.9 at 973 K. 20,21 In these systems, liquid-like Cu ions travel freely within the sulfur sublattice, reducing the k lat to 0.6 W K À1 m À1 . However, this Cu migration drives the physical degradation of materials. 22 Tetrahedrites Cu 12Àx Tr x Sb 4 S 13 (Tr ¼ Mn, Fe, Co, Ni, and Zn) have been demonstrated to possess promising thermoelectric properties over the intermediate temperature range of 600-700 K. The high p-type ZT ($1.0 at 700 K for Cu 10.5 Ni 1.0 Zn 0.5 Sb 4 S 13 ) arises from the combination of a degenerate semiconducting-like band structure and an extremely low k lat ($0.4 W K À1 m À1 at 700 K for Cu 10.5 Ni 1.0 Zn 0.5 Sb 4 S 13 ). [23][24][25][26][27][28][29][30][31] However, tetrahedrites contain a somewhat toxic element (Sb) which prevents them from industrial applications. Cu 26 V 2 E 6 S 32 (E ¼ Ge, Sn) are a family of colusites without Sb, therefore highly promising for practical thermoelectric devices operating in the intermediate temperature range of 600-700 K. [31][32][33][34][35] The crystal structure of colusites is shown in Fig. 1. The unit cell is composed of 66 atoms and belongs to a cubic system with a P 43n space group. [35][36][37] In the crystal structure, there are large void spaces at interstitial sites, such as 6b (0, 0, 1/2). It has been reported that the density of states (DOS) calculations show unoccupied states in the valence band above the fermi level, allowing p-type conduction. 32,33 This electron-decient character is consistent with the electron count for formal valences such as Cu + (3d 10 ), V 5+ (3d 0 ), E 4+ , and S 2À . 32,33 We have demonstrated relatively high p-type ZT values (0.6-0.7 at 663 K) owing to the low k lat ($0.5 W K À1 m À1 at 300-670 K). The low k lat is attributed to the large number of atoms in the simple cubic crystal structure. Moreover, the electronic band structure calculation revealed that the high power factor of $0.6 mW K À2 m À1 at 663 K originates from the Cu-3d and S-3p hybridized orbitals close to the Fermi level. 32,33 Although colusites could pave the way for environmentally friendly and cost-effective thermoelectric generation, there are concerns regarding the oxidation of V in the production processes and practical use because V-oxides are toxic. 38,39 Therefore, an attempt was made to substitute less toxic elements Nb and Ta for V in colusites. Isoelectronic elements V, Nb and Ta are expected to lead to a similar electronic band structure in the colusite system; therefore, the value of S 2 /r for Cu 26 Nb 2 Sn 6 S 32 and Cu 26 Ta 2 Sn 6 S 32 could be high, like that for Cu 26 V 2 Sn 6 S 32 . Furthermore, the large number of atoms in the unit cell of Cu 26 Nb 2 Sn 6 S 32 and Cu 26 Ta 2 Sn 6 S 32 would yield a low k lat . Bearing this in mind, we have synthesized Cu 26 A 2 Sn 6 S 32 (A ¼ V, Nb, Ta) and compared their thermoelectric properties.

Synthesis and sintering
Elemental copper (Cu; 99.99%), vanadium (V; 99.9%), niobium (Nb; 99.9%), tantalum (Ta; 99.9%), tin (Sn; 99.9999%), and sulfur (S; 99.9999%) supplied from Kojundo Chemical Laboratory were used as starting materials without further purication. These elements were mixed in the stoichiometric ratio of Cu 26 A 2 Sn 6 S 32 (A ¼ V, Nb, Ta). For example, 1.321 g of Cu, 0.289 g of Ta, 0.569 of Sn, and 0.820 g of S were used to prepare 3 g of Cu 26 Ta 2 Sn 6 S 32 . A mixture of approximately 3 g was loaded into fused quartz tubes of 12 mm outer diameter Â 8 mm inner diameter. The tubes were evacuated to a pressure of $5 Â 10 À3 Pa and then ame-sealed. The mixtures were heated to 1323 K for 60 h, held at 1323 K for 12 h, cooled to 673 K in 19 h, and then to room temperature over 5 h.
The obtained ingots of Cu 26 A 2 Sn 6 S 32 were hand-ground into powders, which were placed into graphite dies with 10 or 15 mm-inner diameter for hot pressing. The graphite die was inserted into a hot-press furnace (FUT-17000, TOKYO VACUUM). Sintering was performed at 1023 K for 1 h under 70 MPa uniaxial pressure in an Ar gas ow atmosphere. The heating and cooling rates were 10 K min À1 and 20 K min À1 , respectively. The density of the sintered compacts was determined using the gas pycnometer method (AccuPyc II 1340, Micromeritics). The sintered compacts of 15 mm diameter were cut into rods of $2 Â $2 Â $7 mm for the measurements of the Seebeck coefficient and electrical resistivity, and squares of $10 Â $10 Â $2 mm for thermal diffusivity measurements. The sintered compacts of 10 mm diameter were cut into plates of $5 Â $5 Â $0.2 mm for Hall measurements.

Powder X-ray diffraction and scanning electron microscopy
Crystal structures of the as-prepared ingots and sintered compacts were examined by powder X-ray diffractometry (XRD; MiniFlex600, Rigaku) with Cu Ka radiation over the 2q range 10-100 at room temperature. The powder XRD patterns were rened by the Le Bail method using the JANA2006 40 to calculate the lattice parameters. The crystal structures were drawn with the VESTA soware. 41 The microstructures and chemical compositions of the sintered compacts were investigated by scanning electron microscopy (SEM; 15 kV, Miniscope TM3030Plus, Hitachi High-Technologies) with energy dispersive X-ray spectroscopy (EDX; Quantax70, Bruker).

Thermogravimetric analysis
The thermogravimetry curves of the as-prepared ingots and sintered compacts were obtained using a DTG-60 (Shimadzu). The powders of 5 mg mass were put in a boron nitride crucible and heated to 1023 K under an Ar gas ow atmosphere (100 ml min À1 ) at a rate of 5 K min À1 . The a-Al 2 O 3 powder was used as a reference.

Electrical transport measurement
The Seebeck coefficient S and electrical resistivity r of the sintered compacts were simultaneously measured using temperature-differential and four-probe methods, respectively, under a He atmosphere over the temperature range of 300-670 K (ZEM-3, ADVANCE RIKO). The values of S and r were reproducible over heating and cooling cycles for all samples. The relative uncertainty of the measurements was estimated to be 5%.
The Hall coefficient R H was measured at room temperature by the Van der Pauw method with a home-built system under a magnetic eld of 2.3 tesla. The Cu contact wires were attached to the sintered samples using In-rich In-Ga paste.  36,37 The crystal structure is simple cubic (space group: P 43n) containing 66 atoms in a unit cell. The empty 6b site is located between two [CuS 4 ] tetrahedra.

Thermal transport measurement
The total thermal conductivity k total of the sintered compacts was calculated from the density (d), heat capacity C p , and thermal diffusivity D using the expression k total ¼ dC p D. The thermal diffusivity was directly measured and the heat capacity was indirectly derived using a standard sample (Pyroceram 9606, Netzsch) using the laser ash method (LFA 457 MicroFlash, Netzsch) under an Ar gas ow atmosphere at 100 ml min À1 over the temperature range of 300-670 K. Graphite spray coating was applied on the sample surface in order to improve the emission and absorption properties. The heat capacity and thermal diffusivity were measured over heating and cooling cycles. The heat capacity and thermal diffusivity are provided in Fig. S1 in the ESI. † The relative uncertainty of the thermal conductivity is estimated to be smaller than 8% from the uncertainties of d, C p , and D. Thus, the combined relative uncertainty for the calculation of ZT is approximately 12%. Fig. 2(a) shows the powder X-ray diffraction (XRD) patterns of the as-prepared ingots of Cu 26 A 2 Sn 6 S 32 (A ¼ V, Nb, Ta) over the 2q range of 10-100 at room temperature. Most of the reections are those of the cubic colusite-type structure with a P 43n space group, 36,37 while weak peaks due to impurities are observed. Moreover, the main peaks are split as shown in the inset (see Fig. 2(b)), which cannot be accounted for by using the different wavelengths of Cu Ka 1 and Ka 2 . In our previous work, the scanning electron microscopy (SEM) examination with wavelength dispersive X-ray spectroscopy (SEM-WDX) has revealed that the sintered compact with A ¼ V consists of two colusite phases with slightly different chemical compositions (Sn poor and Sn rich phases). [32][33][34] Moreover, the Sn-poor and Sn rich colusite phases have been formed in natural colusites. 36 In our as-prepared samples, the splitting of the main peaks on the XRD patterns is probably due to the presence of two colusite phases. This experimental observation likely originates from an exsolution phenomenon. 35 Fig. 3(a) and S2 in the ESI † show the XRD patterns of the sintered compacts of Cu 26 A 2 Sn 6 S 32 (A ¼ V, Nb, Ta) with simulated ones over the 2q range of 10-60 and 2q range of 10-100 , respectively, at room temperature. Unlike the as-prepared ingots (Fig. 2), no impurity peaks are observed in the sintered compacts within the detection limits of the XRD. Furthermore, the XRD peak splitting disappears as shown in Fig. 3(b). In our previous study, XRD peak splitting remained in the A ¼ V system aer the sintering with a holding time of 5 min to 10 min. [32][33][34] The longer holding time (1 h) in the sintering process causes an annealing effect, promoting the change from two colusite phases to a single colusite phase.

Synthesis and sintering
The rened lattice parameters of the two colusite phases for the as-prepared ingots are denoted as a 1 and a 2 (with a 1 > a 2 ) and that of the single colusite phase for the sintered compacts is denoted as a. The a 1 , a 2 and a are listed in Table 1 for all A ¼ V, Nb, and Ta systems. The a 1 is $0.4% larger than the a 2 ; for example, the a 1 and a 2 of Cu 26 Ta 2 Sn 6 S 32 are 1.0865 nm and 1.0827 nm, respectively. The a is slightly larger than a 1 for all systems; for example, the a of the A ¼ Ta system (1.0885 m) is $0.2% larger than a 1 (1.0865 nm).
In the thermogravimetry (TG) curve for the as-prepared ingots, the weight drops sharply over the temperature of 600-700 K in an Ar ow atmosphere (Fig. 4(a)). The weight reductions are likely due to the evaporation of a part of sulfur, which has the lowest boiling point of 718 K among all the constituent elements. This result implies that the metal content of sintered compacts is larger than the stoichiometric value. This is in fact supported by the SEM-EDX as discussed below. No weight loss was detected in the TG curves up to 750 K for all sintered compacts, as shown in Fig. 4(b).
The SEM back scattered electron (BSE) images of the polished surface of the sintered compacts of Cu 26 A 2 Sn 6 S 32 (A ¼ V, Nb, Ta) are shown in Fig. 5. The SEM-EDX investigations reveal a high homogeneity for A ¼ Nb and V but the presence of secondary phases for the A ¼ Ta sintered compact. For the latter, microscale CuS-and Cu 2 S-based secondary phases were identied, whose actual compositions are listed in Table S1 in the ESI. † For the sintered compacts, the nominal composition, measured density, and calculated density are summarized in  Table 2. It should be noted that the measured density d is larger by $3% than the calculated one estimated from the lattice parameter a and nominal composition. This difference suggests that the actual composition deviates from the nominal one. The average values of the composition were calculated from ve spots EDX data on the surface and are listed in Table 2. We calculated the composition using two assumptions: (i) the sum of the number of cations (Cu, A, and Sn) is 34, and (ii) the number of S atoms in the unit cell is 32. The calculated density was estimated from the a and EDX composition. In the case of assumption (i), the calculated density was smaller than the measured one. For example, the calculated density of 4.4 g cm À3 for the A ¼ V sintered compact is 9% lower than its measured value of 4.73 g cm À3 at room temperature. On the other hand, assumption (ii) provides a reasonable agreement between the calculated and measured densities for all systems. In the A ¼ V sample, the calculated and measured densities are 4.80 g cm À3 and 4.73 g cm À3 , respectively. For all samples, the measured densities were estimated to be greater than 96% of the calculated densities. In assumption (ii), the composition of the cations is larger than the stoichiometric one. These extra cations are a result of the loss of sulfur during the sintering process. To understand the better agreement for (ii), it should Table 1 Lattice parameters refined for the as-prepared ingots (a 1 , a 2 ) and the hot-press sintered compacts (a) of Cu 26    be noted that relatively large voids are present in the colusite crystal structure, such as the interstitial site 6b (0, 0, 1/2) surrounded by four S atoms as shown in Fig. 1. 36,37 The extra cations could occupy these large voids. For example, in the A ¼ Ta system, the interatomic distance between the interstitial 6b site and the nearest S site is 0.2357 nm. This distance fairly agrees with the one between the Cu and S in the CuS 4 tetrahedron (0.2363 nm), suggesting that the interstitial 6b site could be partially lled with extra Cu atoms.

Thermoelectric properties
The temperature dependences of the Seebeck coefficient S and electrical resistivity r for the sintered compacts of Cu 26 A 2 Sn 6 S 32 (A ¼ V, Nb, Ta) over the temperature range of 300-670 K are shown in Fig. 6(a) and (b), respectively. Both r and S increase monotonically with temperature, which are the characteristics of a degenerate semiconductor. This trend is in agreement with previous studies performed on V-based colusite compounds. [32][33][34][35] The signs of S and Hall coefficient R H (see Table 3) are positive, conrming p-type carrier transport. The r for A ¼ Ta is lower than those for A ¼ V and Nb over the whole temperature range measured. At 660 K, the r for A ¼ V, Nb, and Ta is 63 mU m, 73 mU m, and 57 mU m, respectively. The secondary phases in the A ¼ Ta sample probably contribute p-type charge carriers, reducing the electrical resistivity in comparison to the A ¼ Nb and V samples. The value of r for A ¼ V is lower than that for the sample Cu 26 V 2 Sn 6 S 32 reported by Suekuni et al., 34 which reaches 110 mU m at 660 K. The former and latter sintered compacts consist of a single colusite phase and two colusite phases, respectively. Moreover, the former and latter values were measured for the actual compositions Cu 26 (20%), respectively. The differences in the crystal phases and chemical composition between two sintered compacts result in the difference in r. The S for A ¼ V, Nb and Ta increases with temperature reaching a value of $200 mV K À1 at 660 K. The results suggest that the chemical substitution of Ta and Nb for V has little effect on the electrical properties, but the secondary phases in the sample for A ¼ Ta affect the electrical properties. Fig. 6(c) shows the power factor (S 2 /r) of the Cu 26 A 2 Sn 6 S 32 . Owing to the lowest r for the sample Fig. 5 Backscattered electron images of the hot-press sintered compacts of Cu 26 A 2 Sn 6 S 32 (A ¼ (a) V, (b) Nb, (c) Ta). In (c) for A ¼ Ta, the gray and dark gray areas are CuS-and Cu 2 S-based secondary phases, respectively. Table 2 Chemical compositions of the matrix phases and measured and calculated densities of colusite systems Cu 26 A 2 Sn 6 S 32 (A ¼ V, Nb, Ta). The calculated density is estimated from the lattice parameter of the sintered compacts a (see Table 1) and the chemical composition with A ¼ Ta, the S 2 /r reaches the highest value of $0.69 mW K À2 m À1 at 660 K. We found that S and r are proportional to the hole carrier concentration p using the following relations: where k B is Boltzmann's constant, h is Planck's constant, e is the electronic charge and m is the electrical mobility. They are derived from the parabolic band model with acoustic phonon scattering assumption for degenerate semiconductors. 6,42 The p was estimated from the formula p ¼ 1/eR H , where e represents the electronic charge and R H is the Hall coefficient measured at room temperature. The values of p listed in Table 3 for all samples are in the range from 4.2 Â 10 20 cm À3 to 6.6 Â 10 20 cm À3 , which fall in the range expected for degenerate semiconductors. Similar values have been reported for synthetic colusites. 34 The value of p ($6.6 Â 10 20 cm À3 ) for Cu 26 Ta 2 Sn 6 S 32 is slightly higher than those for Cu 26 Nb 2 Sn 6 S 32 (4.7 Â 10 20 cm À3 ) and Cu 26 V 2 Sn 6 S 32 ($4.2 Â 10 20 cm À3 ). r is proportional to 1/p (eqn (3)); therefore, the r for A ¼ Ta is lower than those for A ¼ V and Nb. The SEM-EDX investigations reveal the formation of CuS-and Cu 2 S-based secondary phases in Cu 26 Ta 2 Sn 6 S 32 , indicating that the chemical composition of the colusite matrix deviates from the nominal one. The presence of secondary phases in the A ¼ Ta sample contributes p-type charge carriers, leading to a lower r and higher S 2 /r. The hole mobility m (Table 3) of Cu 26 A 2 Sn 6 S 32 with A ¼ V, Nb, and Ta is estimated to be $4.4 cm 2 V À1 s À1 , $3.5 cm 2 V À1 s À1 , and $3.6 cm 2 V À1 s À1 , respectively. The similar values of the m suggest that the microstructures of the CuS-and Cu 2 S-based secondary phases have only a small effect on the m of the systems. As already reported in a previous study, 34 the p for Snpoor sintered compacts (Cu 26 V 2 Sn 6Àx S 32 , x ¼ 0.5, 1.0) is higher than that for x ¼ 0, leading to a higher S 2 /r. Consequently, a ne-tuning of the p by changing the Sn content would further boost the S 2 /r of the A ¼ Nb and Ta systems.
The room-temperature values of m* were calculated from measured S and p and are given in Table 3. For the three samples, the values of m* lie in the range of 3.2m 0 to 4.3m 0 (m 0 is the free electron mass). From the relation of S f m* in eqn (2), the heavy m* should lead to a high S for the systems. The Mahan-Sofo theory shows that a local increase of DOS at the Fermi energy results in an increase of the m*, enhancing the S. 43,44 S is proportional to m* and (1/p) 2/3 (eqn (2)). The differences in m* and (1/p) 2/3 between the sintered compacts of A ¼ V, Nb, and Ta result in almost the same values of S for all the samples.
The temperature dependences of the total thermal conductivity k total and lattice thermal conductivity k lat for the sintered compacts of Cu 26 A 2 Sn 6 S 32 (A ¼ V, Nb, Ta) are shown in Fig. 7. The electronic thermal conductivity k el was estimated using the Wiedemann-Franz relation (k el ¼ LT/r), where L is the Lorenz number. Then the k lat was calculated by subtracting k el from k total . In a single parabolic band dominated by acoustic phonon scattering, L can be estimated as a function of reduced chemical potential (z*): 15,45-47 z* was obtained from the experimental S values using the following equation: The Fermi integrals F m (z*) are dened as: where x is the reduced energy of carriers. The temperature dependence of the estimated L is given in Fig. S3 in the ESI. † For example, the L at 660 K of A ¼ V, Nb and Ta is $1.6 Â 10 À8 Fig. 6 Temperature dependence of (a) Seebeck coefficient S, (b) electrical resistivity r, and (c) thermoelectric power factor S 2 /r for the sintered compacts of Cu 26 A 2 Sn 6 S 32 (A ¼ V, Nb, Ta) over the temperature range of 300-670 K. The S and r were measured during both heating and cooling cycles. W U K À2 . This procedure can be applied since colusites exhibit a degenerate semiconducting-like behavior. The chemical substitution of Ta and Nb for V has little effect on the thermal transport as in the electrical transport. The complex unit cell of colusite systems yields very low k lat values below 0.6 W K À1 m À1 , which gradually decrease with increasing temperature. This is consistent with the previous results on A ¼ V. 33,34 The lowest k lat for the A ¼ Ta sample would be provided by the strain and/or defect in the crystal structure induced by the secondary phases, which scatter effectively the heat-carrying phonons. The sintered compact of Cu 26 Ta 2 Sn 6 S 32 shows the lowest k lat ; for example, at 670 K the k lat of the A ¼ V, Nb, Ta systems is $0.47 W K À1 m À1 , $0.43 W K À1 m À1 , and $0.40 W K À1 m À1 , respectively. The very low value of k lat is of great advantage to further develop high-performance thermoelectric materials. Fig. 8 shows the dimensionless thermoelectric gure of merit ZT versus temperature. The ZT for the A ¼ V and Nb samples monotonically increases with temperature and achieves a value of 0.6-0.7 at 670 K. Comparable values were obtained by Suekuni et al. for Cu 26 V 2 E 6 S 32 (E ¼ Sn, Ge), with a ZT of 0.6 and 0.7 at 663 K for Cu 26 V 2 Sn 6 S 32 and Cu 26 V 2 Ge 6 S 32 , respectively. 33 Due to the lower r and k total , the A ¼ Ta sample reaches a higher value of 0.8 at 670 K. This value approaches the thermoelectric performances of tetrahedrite compounds at the same temperature (ZT $ 1.0). 27 We have successfully demonstrated the promising ZT for environmentally friendly colusites Cu 26 A 2 Sn 6 S 32 (A ¼ Nb, Ta) at intermediate temperature ($700 K).

Conclusion
We have successfully developed a new class of environmentally friendly and cost-effective thermoelectric materials based on colusites Cu 26 A 2 Sn 6 S 32 (A ¼ Nb, Ta). The as-prepared ingots consist of two colusite phases with different lattice parameters, which have been converted to single-phase samples by hot-press sintering. The chemical substitution of isoelectronic Nb and Ta for V has little effect on the thermoelectric properties. For the three systems, the heavy effective mass (3m 0 -4m 0 ) leads to a high power factor. Furthermore, the complex unit cell of colusites yields very low lattice thermal conductivity below 0.6 W K À1 m À1 over the temperature range of 300-670 K. For the A ¼ Ta sample, the CuS-and Cu 2 S-based secondary phases formed in the sintered compact effectively scatter the phonons, leading to a lower lattice thermal conductivity. Moreover, the carrier concentration in the system is the highest among all the samples, resulting in a higher power factor. Consequently, the A ¼ Ta sample shows an enhanced ZT value of $0.8 at 670 K.   7 Temperature dependence of the total thermal conductivity k total and its lattice contribution k lat for the sintered compacts of Cu 26 A 2 Sn 6 S 32 (A ¼ V, Nb, Ta) on heating and cooling cycles over the temperature range of 300-670 K.