Title: How Well Does Kohn–Sham Regularizer Work for Weakly Correlated Systems?

Abstract

Kohn–Sham regularizer (KSR) is a differentiable machine learning approach to finding the exchange-correlation functional in Kohn–Sham density functional theory that works for strongly correlated systems. Here we test KSR for a weak correlation. We propose spin-adapted KSR (sKSR) with trainable local, semilocal, and nonlocal approximations found by minimizing density and total energy loss. We assess the atoms-to-molecules generalizability by training on one-dimensional (1D) H, He, Li, Be, and Be²⁺ and testing on 1D hydrogen chains, LiH, BeH₂, and helium hydride complexes. The generalization error from our semilocal approximation is comparable to other differentiable approaches, but our nonlocal functional outperforms any existing machine learning functionals, predicting ground-state energies of test systems with a mean absolute error of 2.7 mH.

Authors:

Kalita, Bhupalee  ^[1];

• Search DOE PAGES for author "Kalita, Bhupalee"
• Search DOE PAGES for ORCID "0000-0001-9967-1629"

Pederson, Ryan ^[1]; Chen, Jielun ^[1]; Li, Li ^[2]; Burke, Kieron  ^[1]

• Search DOE PAGES for author "Burke, Kieron"
• Search DOE PAGES for ORCID "0000-0002-6159-0054"

---

+ Show Author Affiliations

1. Univ. of California, Irvine, CA (United States)
2. Google Research, Mountain View, CA (United States)

Publication Date:

Mon Mar 14 00:00:00 EDT 2022

Research Org.:

Univ. of California, Irvine, CA (United States)

Sponsoring Org.:

USDOE Office of Science (SC); National Science Foundation (NSF)

OSTI Identifier:

1866248

Grant/Contract Number:  

SC0008696; CHE-1856165; DGE-1633631

Resource Type:

Accepted Manuscript

Journal Name:

Journal of Physical Chemistry Letters

Additional Journal Information:

Journal Volume: 13; Journal Issue: 11; Journal ID: ISSN 1948-7185

Publisher:

American Chemical Society

Country of Publication:

United States

Language:

English

Subject:

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CHEMISTRY