Image super-resolution inspired electron density prediction

Creators: Li, Chenghan¹; Sharir, Or¹; Yuan, Shunyue¹; Chan, Garnet Kin-Lic¹

1. California Institute of Technology

Abstract

Predicting ground-state electron densities of chemical systems has recently received growing attention in machine learning quantum chemistry, given their fundamental importance as highlighted by the Hohenberg-Kohn theorem. Drawing inspiration from the domain of image super-resolution, we view the electron density as a 3D grayscale image and use a convolutional residual network to transform a crude and trivially generated guess of the molecular density into an accurate ground-state quantum mechanical density. Here we show that this model produces more accurate predictions than all prior density prediction approaches. Due to its simplicity, the model is directly applicable to unseen molecular conformations and chemical elements. We show that fine-tuning on limited new data provides high accuracy even in challenging cases of exotic elements and charge states.

Copyright and License

This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Acknowledgement

This work was primarily supported by the United States Department of Energy, Office of Science, Basic Energy Sciences, Chemical Sciences, Geosciences, and Biosciences Division, FWP LANLE3F2 awarded to Los Alamos National Laboratory under Triad National Security, LLC (‘Triad’) contract grant no. 89233218CNA000001, subaward C2448 to the California Institute of Technology for C.L., O.S., S.Y., and G.K.C. Additional support for G.K.C was provided by the Camille and Henry Dreyfus Foundation via a grant from the program “Machine Learning in the Chemical Sciences and Engineering". G.K.C is a Simons Investigator in Physics. The authors disclose the use of the GPT-4 (OpenAI) model during the writing of the first draft of the article. The artificial intelligence model was used to polish the language, and the generated text was carefully inspected, validated, and edited by the authors.

Data Availability

The DFT density data and molecular geometries are available at Refs. ^80,81,82. Source data for the Figures are provided with this paper. Source data are provided with this paper.

Code Availability

Implementation and minimal examples can be found at Ref. ⁸³ Pre-trained models can be found in https://doi.org/10.6084/m9.figshare.25365508.

Conflict of Interest

G.K.C. is a part owner of QSimulate Inc. The remaining authors declare no competing interests.

Supplemental Material

Supplementary Table

Name	Size	Download all
s41467-025-60095-8.pdf md5:e1370505f665b0fb16787323030ae96e	1.3 MB	Preview Download
41467_2025_60095_MOESM1_ESM.xlsx md5:e811efc39012385c924df7763be1eb9e	11.1 kB	Download
41467_2025_60095_MOESM2_ESM.pdf md5:b2cb853b545450bea1053a48961c0599	161.7 kB	Preview Download
41467_2025_60095_MOESM3_ESM.pdf md5:c2acda76eca8f949fe9c5a3b29b83db4	1.6 MB	Preview Download
41467_2025_60095_MOESM4_ESM.xlsx md5:866a0530c8893865479ddac8049f1f3f	420.3 kB	Download

	All versions	This version
Views	0	0
Downloads	0	0
Data volume	0 Bytes	0 Bytes