SKM 2023 – wissenschaftliches Programm

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

MM: Fachverband Metall- und Materialphysik

MM 29: Data Driven Materials Science: Big Data and Work Flows – Electronic Structure

MM 29.2: Vortrag

Mittwoch, 29. März 2023, 12:00–12:15, SCH A 251

Predicting electron density using a convolutional neural network — •Jae-Mo Lihm^1,2,3, Wanhee Lee⁴, and Cheol-Hwan Park^1,2,3 — ¹Department of Physics and Astronomy, Seoul National University, Seoul, Korea — ²Center for Correlated Electron Systems, Institute for Basic Science, Seoul, Korea — ³Center for Theoretical Physics, Seoul National University, Seoul, Korea — ⁴Department of Applied Physics, Stanford University, California, USA

Machine-learning methods are being widely applied to computational materials science. Most applications of machine learning in electronic structure calculations focus on learning the total energy and forces, while few works study the prediction of electronic properties. In this work, we develop a machine-learning model for predicting electron density using the convolutional neural network, a standard machine-learning method for image processing. We train the neural network using the electron density calculated from density functional theory. We show that the trained neural network can successfully predict the electron density of systems that were not included in the training set, bypassing the need for the self-consistent density functional theory calculation.