Geometry‐Based Neural‐Network Prediction of Electron Localization Function Topology in Dense Hydrogen
Abstract
We develop a machine‐learning framework to predict the electron localization function (ELF) of pure, dense hydrogen directly from atomic geometry, bypassing explicit electronic‐structure calculations. Trained on first‐principles data spanning multiple pressure regimes in dense fluid hydrogen, the model achieves high accuracy () and faithfully reproduces the global distribution of the ELF. A combined real‐ and reciprocal‐space analysis reveals that the residual error is dominated by smooth, long‐wavelength components with correlation lengths exceeding typical H─H bonding scales, and that the magnitude of these components increases systematically with pressure. Despite being trained exclusively on dense fluid hydrogen networks, the model transfers robustly to crystalline hydrogen configurations, preserving key features of ELF topology, including critical points and hydrogen‐network connectivity. Taken together, these results suggest a viable route toward geometry‐based, high‐throughput evaluation of hydrogen‐networking characteristics in both fluid and crystalline hydrogen.
Related
- WISP: Learning spin-resolved electron localization from atomic environments
- Topology-Guided Multi-Level Graph Learning: A Unified Framework for Molecular Property Prediction
- Cation Mixing Controls Photocatalytic Activity of Ionic Carbon Nitrides by Switching Exciton Character and Electron Localization
- Cage Geometry Reorganizes N···C Contacts While Preserving Nitrogen Spin Localization in N@C60 and N@C70: A DFT Study
- Quantifying the Coupled Role of Mie Resonance Order and Geometry in Hot-Electron Photodetection
BibTeX
@article{wang2026geometrybased,
title = {{Geometry‐Based Neural‐Network Prediction of Electron Localization Function Topology in Dense Hydrogen}},
author = {Xiaoyu Wang and Miriam Marqués and Sergio Gómez and Francesc Serratosa and Eva Zurek and Julia Contreras-García},
journal = {Chemistry - A European Journal},
year = {2026},
month = jul,
doi = {10.1002/chem.202600030},
url = {https://hal.science/hal-05740165v1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Geometry‐Based Neural‐Network Prediction of Electron Localization Function Topology in Dense Hydrogen
AU - Xiaoyu Wang
AU - Miriam Marqués
AU - Sergio Gómez
AU - Francesc Serratosa
AU - Eva Zurek
AU - Julia Contreras-García
PY - 2026
DA - 2026/07/06
JO - Chemistry - A European Journal
DO - 10.1002/chem.202600030
UR - https://hal.science/hal-05740165v1
AB - We develop a machine‐learning framework to predict the electron localization function (ELF) of pure, dense hydrogen directly from atomic geometry, bypassing explicit electronic‐structure calculations. Trained on first‐principles data spanning multiple pressure regimes in dense fluid hydrogen, the model achieves high accuracy () and faithfully reproduces the global distribution of the ELF. A combined real‐ and reciprocal‐space analysis reveals that the residual error is dominated by smooth, long‐wavelength components with correlation lengths exceeding typical H─H bonding scales, and that the magnitude of these components increases systematically with pressure. Despite being trained exclusively on dense fluid hydrogen networks, the model transfers robustly to crystalline hydrogen configurations, preserving key features of ELF topology, including critical points and hydrogen‐network connectivity. Taken together, these results suggest a viable route toward geometry‐based, high‐throughput evaluation of hydrogen‐networking characteristics in both fluid and crystalline hydrogen.
SN - 0947-6539
SN - 1521-3765
ER -