Thermal Transport with Message Passing Neural Networks via the Green-Kubo Method

The Green-Kubo method combined with first-principles calculations provides an accurate and precise framework to obtain thermal conductivities for novel materials, including strongly anharmonic ones [1]. However, high computational cost associated with the long dynamics simulations in large supercells required for convergence limits its applicability for large-scale, high-throughput materials discovery. Machine learning potentials can significantly reduce this cost [2].

Message passing neural networks (MPNNs) are a promising, but for this task yet untested, class of models, as they can accommodate implicit long-range interactions and directional information. In this work, we adapt the heat flux definition for MPNNs, and present a systematic account of their performance and convergence behaviour for calculating the thermal conductivity of several solid semiconductors and insulators.

[1]: C. Carbogno, R. Ramprasad, and M. Scheffler, Phys. Rev. Lett. 118 175901 (2017)

[2]: P. Korotaev et al., Phys. Rev. B 100 144308 (2019); C. Mangold et al., J. Appl. Phys. 127, 244901 (2020); C. Verdi et al., NPJ Computer. Mat. 7 156 (2021)