A Statistical Approach for the Rapid Prediction of Electron Relaxation Time Using Elemental Representatives

Efficiency of any thermoelectric material relies on a combination of suitable electronic and thermal transport properties, which are governed by various scattering mechanisms. Explicit evaluation of temperature dependent scattering time or the electron relaxation time (τ_el) is thus necessary to assess the efficiency of thermoelectrics. Experimental or computational measurement of τ_el is very challenging due to the inherent time limitation and high computational cost. We present a machine learning based statistical model for a rapid prediction of experimental electrical conductivity (σ) and subsequent estimation of temperature dependent relaxation time (τ_el). A statistical machine learning model was trained on 124 experimental data points to predict electrical conductivities, which span 8 orders of magnitude. By utilizing a unique mean ranking method for feature selection, simple elemental properties such as the boiling point, melting point, molar heat capacity, electron affinity, and ionization energy are identified as the potential descriptors for σ. The developed model has very small root-mean-square error (rmse) of 0.22 S/cm and a high coefficient of determination (R2 ) of 0.98 for prediction of log-scaled σ. Utilizing the predicted σ values, τel has been calculated for a wide range of temperatures. ML predicted τel values outperforms the widely used deformation potential model for the similar set of compounds. Moreover, the estimated τel using experimentally measured σ includes all the possible scattering mechanisms, which can affect the transport properties. This highly accurate ML model for σ could fill the biggest gap in the estimation of performance of the thermoelectric materials. Our developed approach using only elemental descriptors highlights the advantages to screen desired materials without any expensive experimental measurement or first-principles calculations.