Estimation of electron-phonon coupling via moving least squares averaging: a method for fast-screening potential thermoelectric materials

In this communication, we present a method of predicting the Seebeck coefficient and electrical conductivity of inorganic semiconductors by estimating the electron-phonon (el-ph) coupling from the first principles for fast-screening potential thermoelectric (TE) materials. The method we propose, i.e. the electron-phonon averaged via moving least squares (EPA-MLS) method, combines the EPA method and the MLS averaging strategy. To demonstrate the performance of the EPA-MLS method, the Seebeck coefficient and electrical conductivity of a half-Heusler compound, i.e. HfCoSb, were computed with the EPA-MLS method and compared with the results from the EPA method and comparable experimental data sets. The results show that the EPA-MLS method exhibits several advantages over the original EPA method. The smoother interpolation reduces the risk of spurious numerical behaviors. The EPA-MLS method also requires less human intervention for tuning numerical parameters, since the calculation result of the EPA-MLS method exhibits robustness against the change of associated numerical parameters. The method may even reduce the overall computational cost by allowing the employment of a coarser resolution. All these advantages make the EPA-MLS method a suitable tool for fast-screening potential TE materials. One more example of an archetypical Skutterudite, i.e. CoSb₃, is also provided for showing that the method can be used for TE materials with more complex structures.