Mapping The Configuration Space of Intrinsic Properties and Dopability of Half-Heusler Compounds

In this work, we utilize computational approaches to navigate the half-Heusler phase space to identify patterns in how properties relevant to thermoelectric (TE) materials vary with chemistry. Identifying novel TE materials is a challenge due to the unique, and typically disparate, set of intrinsic properties required (thermal and electronic conductivity) together with the need to access high p-type or n-type carrier concentrations. Although half-Heuslers have long been known as candidate TE materials, the hundreds of compositions that have been experimentally realized represent a small proportion of the possible chemistries. Of these possible chemistries, we focus on the 1,126 that are expected to be semiconductors based on the “18 valence electron rule”. In addition to fundamental understanding, design rules that link chemistry to indicators of TE performance such as carrier mobilities, band edge density of states, band edge degeneracy, and thermal conductivity are useful for accelerated materials discovery. Towards identifying these design rules, we evaluate candidate half-Heusler transport properties using first-principles simulations and semi-empirical models. This approach utilizes a filtering scheme that successively eliminates compounds based on their electronic structure, TE performance, stability, dopability, and transport properties. The analysis yields trends across chemical phase space. For instance, Rh and V-containing chemistries consistently have strong n-type TE performance, Ni and Sb-containing chemistries are consistently strong p-type performers, and Ta, or W-containing chemistries are good for both p and n-type performance. Furthermore, good n-type TE is found to arise from chemistries exhibiting high electronic carrier mobility as a result of low band mass, while for p-type performance a high band degeneracy at the valence band edge further enhances performance. Chemistries that exhibit good TE performance for both n and p-type were found to have comparatively low thermal conductivities relative to what is typical for half-Heuslers. This presentation will focus on the key results of the computational analysis and a comparison to experiment, the features of half-Heusler compounds that make them particularly suitable for TE applications, and ongoing work in stochastic assessment of trends using machine learning.