Recently I have calculated transport properties of an antiperovskite. Experimentally It has been found that the electrical conductivity shows metallic behavior. But when I have calculated the band structure by using 35x35x35 k-points in Wien2K, it shows semiconducting. However, by using 47X47X47 k-point, it shows metallic nature. Unfortunately, in both cases, the electrical conductivity has found from BoltzTraP data to increase with temperature. I have performed the calculations several times but the results are same. Finally, I have turned to Quantum espresso and BoltzTrap calculations. Unfortunately, I have failed and the electrical conductivity shows the same behavior. Much below the Fermi level, (actual 0.48 Ry to 0.31 Ry) it shows metallic conductivity. If anyone informs me where the problem, it would be a great pleasure and I will be greatly appreciated. Thanks to advances!
Hi Enamul,
Boltztrap uses relaxation time approximation, which is an efficient quick way, but with many limitations. I had that problem too, boltztrap clearly fail to predict the conductivity variation with temperature and changing parameters will not solve the issue. we use (opensource) package that included scattering mechanism, it predicts well the right slop of conductivity variation with temperature.
see: Code ab initio Model for Mobility and Seebeck coefficient using B...
BoltzTraP needs a very dense k-point sampling in order to sample the Fermi surface well, so you do have to be very careful with the k-point sampling. If this is a problem, you could investigate one of the Wannier-based methods such as BoltzWann:
Article BOLTZWANN: A code for the evaluation of thermoelectric and e...
(In this context, Wannier functions can be thought of as a way to take the ab initio electronic structure and resample it very accurately across the Brillouin zone.)
As for the temperature dependence, BoltzTraP doesn't know about the scattering time for the electrons, in particular the electron-phonon coupling, so if you really want to model the conductivities well you'll need to think about how to compute or model this.
Dear El Tayeb Bentria and Philip James Hasnip,
Thank you very much for good information. Usually, the constant relaxation time approximation (RTA) used in conductivity calculation gives the reasonable accuracy as found in many compounds. I have checked using different k-points but the results are same. May this be for defect in the experimental synthesized compound? (In the paper, they have not mentioned it.). Any idea? Please!
- Badges
- Science topic
- Similar topics
- Filing