Machine learning dielectric screening for the simulation of excited state properties of molecules and materials*

Accurate and efficient predictions of absorption spectra of materials and molecules at finite temperature are essential for the understanding and rational design of broad classes of systems. We present an approach to improve the efficiency of first principles calculations of absorption spectra of complex materials at finite temperature, based on the solution of the Bethe-Salpeter Equation in finite-field (FF) [1]. We demonstrate that methods using convolutional neural networks (CNN) may be efficiently used to compute the screened Coulomb interaction and to predict finite-temperature absorption spectra of solids, liquids, nanoparticles, and heterogeneous systems, such as solid/liquid interfaces. In addition, we show that our approach may be used to derive model dielectric functions for complex systems [2].

[1] N. L. Nguyen, H. Ma, M. Govoni, F. Gygi, and G. Galli, Phys. Rev. Lett. 122, 237402 (2019).
[2] S. S. Dong, M. Govoni, and G. Galli, 2020 (preprint).