GoKawiilDevices known as thermoelectric generators (TEGs) can convert waste heat directly into electricity without using moving parts or producing carbon dioxide emissions. From powering wearable devices to recovering heat produced by industrial processes, TEGs could have a pivotal role in addressing global energy challenges. However, optimizing TEG designs is a highly intricate task that has prevented these devices from reaching their full potential. Writing in Nature, Li et al.1 introduce TEGNet, a neural-network-based system that models TEG performance with greater than 99% accuracy while slashing computational time by around 10,000-fold compared with using conventional predictive systems. Not only does TEGNet accelerate TEG design, but it also produces material-specific models of TEG components that can be assembled virtually in a modular way. This enables rapid exploration of diverse device architectures.
Nature 652, 570-572 (2026)
doi: https://doi.org/10.1038/d41586-026-00907-z
References Li, A. et al. Nature 652, 643–649 (2026). Martinez, A. Energy Conv. Mgmt. 292, 117419 (2023). Reiser, P. et al. Commun. Mater. 3, 93 (2022). Snyder, G. J. Appl. Phys. Lett. 84, 2436–2438 (2004). Snyder, G. J. & Ursell, T. S. Phys. Rev. Lett. 91, 148301 (2003). Snyder, G. J. & Toberer, E. S. Nature Mater. 7, 105–114 (2008). Yan, Q. & Kanatzidis, M. G. Nature Mater. 21, 503–513 (2022). Sun, X. et al. Nature Commun. 16, 4220 (2025). Iwasaki, Y. et al. Sci. Rep. 9, 2751 (2019). Download references
Competing Interests The authors declare no competing interests.
Related Articles
Subjects