Complexity Reduction in Density Functional Theory: Locality in Space and Energy

TL;DR

This paper advances large-scale hybrid Density Functional Theory calculations by integrating complexity reduction techniques, new algorithms for orbital energies, and analysis of system fragmentation, enabling efficient study of systems with thousands of atoms.

Contribution

It introduces two novel algorithms for computing Kohn-Sham orbital energies and applies a complexity reduction framework to large-scale DFT calculations on supercomputers.

Findings

Algorithms efficiently handle systems with thousands of atoms.

Fragmentation analysis reveals energy-dependent system properties.

Complexity reduction improves computational efficiency in DFT calculations.

Abstract

We present recent developments of the NTChem program for performing large scale hybrid Density Functional Theory calculations on the supercomputer Fugaku. We combine these developments with our recently proposed Complexity Reduction Framework to assess the impact of basis set and functional choice on its measures of fragment quality and interaction. We further exploit the all electron representation to study system fragmentation in various energy envelopes. Building off this analysis, we propose two algorithms for computing the orbital energies of the Kohn-Sham Hamiltonian. We demonstrate these algorithms can efficiently be applied to systems composed of thousands of atoms and as an analysis tool that reveals the origin of spectral properties.