AI accelerates thermoelectric generator design

Thermoelectric generators convert waste heat directly into electricity, offering a promising path to sustainable energy. However, designing high-performance devices remains an engineering challenge. Although artificial intelligence (AI) has rapidly advanced materials science, applying it to optimize the thermoelectric generators has proved difficult. Conventional finite-element simulations offer a reliable route to optimization, but they require repeatedly solving coupled partial differential equations (PDEs), making the exploration of vast design spaces highly time-consuming.

The researchers showed that TEGNet achieves over 99% accuracy while requiring only 0.01% of the computational time used by commercial solvers. This efficiency allowed the team to evaluate thousands of design trade-offs and identify optimal configurations in seconds. Experimental tests confirmed the effectiveness of the model: the team fabricated devices with conversion efficiencies exceeding 9% for segmented legs and reaching 8.7% for n–p paired generators — among the highest reported for near-room-temperature systems.