Machine learning electronic structure and atomistic properties from the external potential
Jigyasa Nigam, Tess Smidt, Geneviève Dusson · 2026
Read Machine learning electronic structure and atomistic properties from the external potential by Jigyasa Nigam, Tess Smidt, Geneviève Dusson (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-15345.txt.
Description
Nigam and colleagues tackle electron calculations, a chokepoint in atom tests, with ML. Most methods map molecular geometries, graphs or encoded local data to molecular properties. Others use ML as a stand-in for electronic structure, aiming at the Fock or density matrices in an atomic orbital basis. Led by the Hohenberg-Kohn theorem, the team offers an operator-centered framework where the external potential in that basis is the model input.
From this operator, they build layered representations echoing popular atom-based descriptors. The matrix-valued external potential links to equivariant message-passing networks. Products of the external potential provide a scalable route to message passing and long-range effects. This method can model molecular properties, such as energies and dipole moments, from the external potential. It can learn operator-to-operator maps to the Fock matrix and the reduced density matrix, yielding multiple molecular readings.
Catalog record
| CREATOR | Jigyasa Nigam, Tess Smidt, Geneviève Dusson |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 76 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-15345