Many-body machine-learned potential for the hydrated proton at CCSD(T) accuracy with fast analytical gradient
4 affiliations
- 1Department of Chemistry, Shanghai Key Laboratory of Electrochemical and Thermochemical Conversion for Resources Recycling, State Key Laboratory of Porous Materials for Separation and Conversion
- 2Fudan University
- 3Department of Chemistry
- 4Emory University and Cherry L. Emerson Center for Scientific Computation
Abstract
The hydrated proton plays a central role in aqueous chemistry, biological proton transport, and electrochemical energy conversion. Reliable simulations of hydratedproton properties require an accurate reactive potential with low computational cost. Here, we report HP-AQUA, a high-level many-body machine-learned potential for the hydrated proton with fast analytical gradients. HP-AQUA combines the CCSD(T) quality q-AQUA-pol water potential and an accurate H3O + monomer potential with new permutationally invariant polynomial representations of the hydronium-water two-, three-, and four-body interactions. HP-AQUA demonstrates CCSD(T)-level accuracy across a series of gas-phase protonated water clusters for binding energies, harmonic frequencies, proton-transfer profiles, and configurations sampled along molecular dynamics trajectories. Analytical gradients implemented through reverse differentiation reduce the computational cost by more than an order of magnitude relative to our previous
Related
- Accuracy, Reliability, and Efficiency of a MACE Machine-Learning Interatomic Potential for Aspirin: A Reproducible Benchmark for Molecular Simulation
- Revealing the H2O/Na environments in Hydrated Manganese Iron Prussian White: a D/H exchange and machine-learning-assisted MD simulation study
- Accuracy and Efficiency Benchmarks of Pretrained Machine Learning Potentials for Molecular Simulations
- Low-rank approximation of Moment Tensor Potential enables reducing training set size without loss of accuracy
- Full-data accuracy with fewer labels for training and fine-tuning machine-learning force fields
BibTeX
@misc{zhang2026manybody,
title = {{Many-body machine-learned potential for the hydrated proton at CCSD(T) accuracy with fast analytical gradient}},
author = {Guorui Zhang and Ruitao Ma and Joel M. Bowman and Qi Yu},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15009682/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009682/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Many-body machine-learned potential for the hydrated proton at CCSD(T) accuracy with fast analytical gradient
AU - Guorui Zhang
AU - Ruitao Ma
AU - Joel M. Bowman
AU - Qi Yu
PY - 2026
DA - 2026/09/29
PB - ChemRxiv
DO - 10.26434/chemrxiv.15009682/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009682/v1
AB - The hydrated proton plays a central role in aqueous chemistry, biological proton transport, and electrochemical energy conversion. Reliable simulations of hydratedproton properties require an accurate reactive potential with low computational cost. Here, we report HP-AQUA, a high-level many-body machine-learned potential for the hydrated proton with fast analytical gradients. HP-AQUA combines the CCSD(T) quality q-AQUA-pol water potential and an accurate H3O + monomer potential with new permutationally invariant polynomial representations of the hydronium-water two-, three-, and four-body interactions. HP-AQUA demonstrates CCSD(T)-level accuracy across a series of gas-phase protonated water clusters for binding energies, harmonic frequencies, proton-transfer profiles, and configurations sampled along molecular dynamics trajectories. Analytical gradients implemented through reverse differentiation reduce the computational cost by more than an order of magnitude relative to our previous
ER -