Thermoelectric Properties of CoAsSb: An Experimental and Theoretical Study

Abstract

Polycrystalline samples of CoAsSb were prepared by annealing a stoichiometric mixture of the elements at 1073 K for 2 weeks. Synchrotron powder X-ray diffraction refinement indicated that CoAsSb adopts arsenopyrite-type structure with space group P2 1 /c. Sb vacancies were observed by both elemental and structural analysis, which indicate CoAsSb 0.883 composition. CoAsSb was thermally stable up to 1073 K without structure change but decomposed at 1168 K. Thermoelectric properties were measured from 300 to 1000 K on a dense pellet. Electrical resistivity measurements revealed that CoAsSb is a narrow-band-gap semiconductor. The negative Seebeck coefficient indicated that CoAsSb is an n-type semiconductor, with the maximum value of −132 μV/K at 450 K. The overall thermal conductivity is between 2.9 and 6.0 W/(m K) in the temperature range 300−1000 K, and the maximum value of figure of merit, zT, reaches 0.13 at 750 K. First-principles calculations of the electrical resistivity and Seebeck coefficient confirmed n-type semiconductivity, with a calculated maximum Seebeck coefficient of −87 μV/K between 900 and 1000 K. The difference between experimental and calculated Seebeck coefficient was attributed to the Sb vacancies in the structure. The calculated electronic thermal conductivity is close to the experimental total thermal conductivity, and the estimated theoretical zT based solely on electronic thermal conductivity agrees with experimental values in the high temperature range, above 800 K. The effects of Sb vacancies on the electronic and transport properties are discussed.

Electronic structure and thermoelectric properties of CoAsSb with post-DFT approaches

“…The Seebeck coefficient calculated with LQSGW P BE approach is presented in Figure 3 for various doping levels. Zero doping results in a positive Seebeck coefficient at low temperatures and a negative Seebeck coefficient at high temperatures in contrast to the consistent negative Seebeck coefficient from the experiment (Figure 7a in [18]). From a small increase in the doping, the LQSGW P BE curves suddenly become much more similar to the experiment and T max increases.…”

“…LDA resulted in a metallic band structure contradicting with experimental results. 18 Consequently there was a mismatch of the calculated transport properties in accordance with general tendencies discovered in high-throughput calculations. 8 To compare ways of resolving the bandgap issue, authors of the Ref.…”

“…LQSGW results in a band gap of 0.279 eV which is slightly above the experimental range 0.12-0.256 eV. 18 The range of exact exchange in PBE0 (10%-25%) results in bandgaps which cover the experimental uncertainty. For the optimized geometry, we obtained a tiny gap in our DFT calculation (0.005 eV), which can be at- tributed to using GGA instead of LDA as in [18].…”

“…18 The range of exact exchange in PBE0 (10%-25%) results in bandgaps which cover the experimental uncertainty. For the optimized geometry, we obtained a tiny gap in our DFT calculation (0.005 eV), which can be at- tributed to using GGA instead of LDA as in [18]. But with LQSGW, a noticeable overestimation of the band gap (0.441eV) is observed.…”

“…However, the above slightly more advanced method LQSGW P BE (as compared to LQSGW) still mimics true vertex-corrected GW calculations and improves the bandgap. CoAsSb has a large unit cell (12 atoms), and its symmetry group (P2 1 /c, arsenopyrite-type structure 18 ) has 4 operations. The evaluation of the band energies on sufficiently fine k-mesh in the Brillouin zone is necessary followed by subsequent interpolation to even finer k-mesh used to study transport properties.…”