When do machine-learning potentials need a configuration-dependent kernel? Learning long-range interactions with Fourier neural operators
- 1Institute of Condensed Matter and Nanosciences (IMCN)
- 2UCLouvain
- 3WEL Research Institute
Abstract
Long-range interactions in machine-learning interatomic potentials are commonly treated by coupling learned sources through either a prescribed electrostatic kernel or a learned but configurationindependent one. Whether such a kernel suffices depends on the underlying screened interaction. We address this question by using Fourier neural operators (FNOs) to learn how latent source fields, built from the descriptors of a local backbone (here MACE), interact across the simulation cell. For a synthetic ionic system with tunable screening, a prescribed Coulomb kernel underestimates the longwavelength charge fluctuations by 43% under strong screening, whereas the FNO reproduces them within 3% and recovers 95–99% of the variance of the residual left by the local backbone. A linear FNO, which learns a configuration-independent kernel, performs equally well in the homogeneous case. Under inhomogeneous screening, rearranging the remote medium modulates the coupling between distant ions, a dependence that no configuration-independent kernel can represent. Among the models tested, only the nonlinear FNO responds to this modulation, and its force errors are 27–41% lower than those of configuration-independent models. In first-principles benchmarks of liquid water, lithium lanthanum zirconate, and Au 2 /MgO(001), MACE–FNO is comparable to a prescribed-kernel model for the bulk systems and more accurate for the interface. It further captures the reversal of the Au 2 wetting preference induced by remote Al dopants, with deviations below 5meV per cell from density functional theory. These results indicate that a configuration-dependent kernel becomes necessary once the screening medium varies, even though the energy remains quadratic in the charges for a fixed medium. Nonlinear FNOs, which learn such a kernel from energies and forces alone, offer a flexible route to environment-dependent long-range interactions.
Related
- Learning Memory-Dependent Neural Network Correlation Potentials for Accurate Electron Dynamics
- Concentration-Dependent Speciation in ZnCl2 Electrolyte Solutions: Transferability of Machine-Learning Interatomic Potentials and Speciation Sensitivity of EXAFS
- Universal Interatomic Potentials as Configuration-Space Generators for One-Shot and Iterative Fine-Tuning of Ab Initio-Accurate Material-Specific Models
- Performance of Highly Coarse-Grained Chain Models with Extremal Measures for a Configuration-Dependent Hydrodynamic Diameter in Rheological Flow Fields
- Electronic configuration vectors provide transferable elemental representations for machine learning across the periodic table
BibTeX
@misc{chen2026when,
title = {{When do machine-learning potentials need a configuration-dependent kernel? Learning long-range interactions with Fourier neural operators}},
author = {Wei Chen and Gian-Marco Rignanese},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15011048/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15011048/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - When do machine-learning potentials need a configuration-dependent kernel? Learning long-range interactions with Fourier neural operators
AU - Wei Chen
AU - Gian-Marco Rignanese
PY - 2026
DA - 2026/10/26
PB - ChemRxiv
DO - 10.26434/chemrxiv.15011048/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15011048/v1
AB - Long-range interactions in machine-learning interatomic potentials are commonly treated by coupling learned sources through either a prescribed electrostatic kernel or a learned but configurationindependent one. Whether such a kernel suffices depends on the underlying screened interaction. We address this question by using Fourier neural operators (FNOs) to learn how latent source fields, built from the descriptors of a local backbone (here MACE), interact across the simulation cell. For a synthetic ionic system with tunable screening, a prescribed Coulomb kernel underestimates the longwavelength charge fluctuations by 43% under strong screening, whereas the FNO reproduces them within 3% and recovers 95–99% of the variance of the residual left by the local backbone. A linear FNO, which learns a configuration-independent kernel, performs equally well in the homogeneous case. Under inhomogeneous screening, rearranging the remote medium modulates the coupling between distant ions, a dependence that no configuration-independent kernel can represent. Among the models tested, only the nonlinear FNO responds to this modulation, and its force errors are 27–41% lower than those of configuration-independent models. In first-principles benchmarks of liquid water, lithium lanthanum zirconate, and Au 2 /MgO(001), MACE–FNO is comparable to a prescribed-kernel model for the bulk systems and more accurate for the interface. It further captures the reversal of the Au 2 wetting preference induced by remote Al dopants, with deviations below 5meV per cell from density functional theory. These results indicate that a configuration-dependent kernel becomes necessary once the screening medium varies, even though the energy remains quadratic in the charges for a fixed medium. Nonlinear FNOs, which learn such a kernel from energies and forces alone, offer a flexible route to environment-dependent long-range interactions.
ER -