Thermoelectrics and Waste Heat Recovery

Thermoelectric materials convert heat gradients directly into electricity, offering a route to recover waste heat from industrial processes, vehicles, and power generation. Achieving high performance requires optimizing three interrelated transport properties simultaneously: electrical conductivity, Seebeck coefficient, and thermal conductivity. These properties are physically coupled, making independent optimization impossible. The figure of merit ZT remains below 2.5 for most known materials at practical temperatures, and the materials that do reach high ZT often depend on expensive, toxic, or scarce elements. Progress demands a generative approach that navigates this coupled landscape rather than screening individual properties one at a time.

Traditional thermoelectric development uses Boltzmann transport theory and DFT band-structure calculations to evaluate ZT for known or proposed compositions, then tunes carrier concentration through doping studies. This evaluate-then-optimize loop can only assess candidates that researchers propose from chemical intuition. Worse, improving one transport metric (e.g., suppressing thermal conductivity through nanostructuring) often degrades another (e.g., carrier mobility). Machine-learning models trained on thermoelectric databases reproduce known correlations but offer little help in regions where conventional trade-offs might be circumvented.