Adventures and Challenges in Computation-Guided Design and Discovery of Thermoelectric Materials -- from Direct Physical Modeling to Language Models

My presentation will highlight some of our recent attempts to make the computation-experiment handshake applied to the discovery and optimization of thermoelectric materials. We will describe the importance and state-of-the-art of accurate simulations of both intrinsic material properties and dopability in computation-guided search for new materials, focusing on the design space of diamond like semiconductors, ordered vacancy compounds, binary chalcogenides, and half-heuslers. Finally, we'll discuss where computational approaches have success and what aspects are missing and should be realized to enable true computation-guided design and optimization -- focusing on strategies to optimize doping and carrier concentrations, and alloying strategies. Time-permitting, I will depart from direct physical modeling to show some of our recent work in the use of language-based representations in the search for thermoelectric materials and discuss some observations of the merits and possible shortcomings of data-based methodologies.