Learning DFT

TL;DR

This paper introduces a deep learning approach to construct Kohn-Sham potentials from density matrices, enabling the development of density functional theory functionals and scaling to larger systems.

Contribution

It presents a novel neural network scheme to map local densities to Kohn-Sham potentials, extending DFT capabilities via deep learning.

Findings

Successful neural network training for potential mapping

Effective up-scaling to larger system sizes

Enhanced DFT functional construction

Abstract

We present an extension of reverse engineered Kohn-Sham potentials from a density matrix renormalization group calculation towards the construction of a density functional theory functional via deep learning. Instead of applying machine learning to the energy functional itself, we apply these techniques to the Kohn-Sham potentials. To this end we develop a scheme to train a neural network to represent the mapping from local densities to Kohn-Sham potentials. Finally, we use the neural network to up-scale the simulation to larger system sizes.