SKM 2023 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

HL: Fachverband Halbleiterphysik

HL 6: Focus Session: Frontiers of Electronic-Structure Theory III (joint session O/HL)

HL 6.7: Vortrag

Montag, 27. März 2023, 12:30–12:45, TRE Ma

Predicting the response of the electron density to electric field using machine learning — •Alan Lewis and Mariana Rossi — MPI for Structure and Dynamics of Materials, Hamburg, Germany

The response of the electron density of a molecule or material to a homogeneous electric field defines its dielectric constant, along with its Raman and sum-frequency spectrum. We present a local and transferable machine learning approach capable of predicting the density response of molecules and periodic system on the same footing. This uses a very similar framework to that of the SALTED method recently introduced by these authors,[1,2] requiring only a small modification to the λ−SOAP descriptors used to represent the atomic environments. This allows us to predict the density response of liquid water to a field applied in each Cartesian direction from a single machine learning model. The tensorial dielectric constant can then be derived from this predicted density response, dramatically reducing the computational cost of calculating these properties relative to the standard approach of using density functional perturbation theory. We discuss the transferability of the model to different phases, and demonstrate the extrapolative power of this approach.

[1] Lewis, Grisafi, Ceriotti, Rossi, JCTC 17, 11, 7203 (2021)

[2] Grisafi, Lewis, Rossi, Ceriotti, accepted JCTC (2022)