Material descriptors for predicting thermoelectric performance

In the context of materials design and high-throughput computational searches for new thermoelectric materials, the need to compute electron and phonon transport properties renders direct assessment of the thermoelectric figure of merit (zT) for large numbers of compounds untenable. Here we develop a semi-empirical approach rooted in first-principles calculations that allows relatively simple computational assessment of the intrinsic bulk material properties which govern zT. These include carrier mobility, effective mass, and lattice thermal conductivity, which combine to form a semi-empirical metric (descriptor) termed β_SE. We assess the predictive power of β_SE against a range of known thermoelectric materials, as well as demonstrate its use in high-throughput screening for promising candidate materials.