Beyond Multiscale: A Neural-Network-Unified Molecular Model for Asymmetry-Free Dynamics Simulation
4 affiliations
- 1Department of Chemistry, Graduate School of Science, Kyushu University, 744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan
- 2Department of Chemistry, Faculty of Science, Kyushu University, 744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan
- 3Quantum Computing Center, Keio University, 3-14-1 Hiyoshi, Kohoku-ku, Yokohama 223-8522, Japan
- 4Quantum and Spacetime Research Institute, Kyushu University, 744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan
Abstract
Adaptive multiscale simulations, such as adaptive quantum mechanics (QM)/molecular mechanics (MM), are indispensable for investigating complex solution dynamics but historically suffer from a fundamental dilemma: the dynamic exchange of identical solvent molecules across different model resolutions introduces an energetic asymmetry under solvent exchange (specifically, a permutation asymmetry where physically identical molecules are treated indistinguishably yet subjected to different force fields depending on their instantaneous positions). This asymmetry forces a prohibitive trade-off between severe structural distortions and the violation of Hamiltonian conservation (continuous energy drift). To resolve this issue, we propose the “Integrated Molecular Model,” a paradigm-shifting framework that unifies multiple resolutions into a single, dynamically consistent potential. By employing a neural network that takes only the multiscale weight functions as input descriptors, our model learns to systematically cancel the nonphysical transition forces originating from spatial discontinuities. As a proof-of-concept, we validated this unified framework on bulk water, aqueous ionic solutions, and reactive hydronium ions. Our approach not only faithfully restores artifact-free solvation structures but also strictly conserves the Hamiltonian, enabling stable, long-term simulations under the microcanonical (NVE) ensemble. Consequently, the model successfully eliminates artificial thermostat-induced breathing modes, allowing for the rigorous evaluation of intrinsic dynamical properties such as diffusion coefficients. Furthermore, owing to its physically informed network design, the correction potential introduces less than 2% additional computational overhead and demonstrates remarkable zero-shot transferability across chemically similar ionic species. By eradicating the inherent asymmetry of adaptive multiscale modeling, this highly scalable framework establishes a robust foundation for high-fidelity simulations of complex condensed-phase phenomena, including reactive dynamics at solid–liquid interfaces and ion transport in battery electrolytes.
Related
- AutoSolvate 2.0: Automated Preparation of Chemically Diverse Explicit-Solvent Systems for Multiscale Molecular Simulation
- Multiscale Modeling of Ion Transport in Nanopores: Fitting Implicit-Water Radial Diffusion Profiles to Explicit-Water Molecular Dynamics
- Broadband molecular dynamics simulation of confined Brownian motion
- Nonadiabatic molecular dynamics simulation with external field
- Large integration time-step in molecular dynamics simulation artificially enhances the strength of hydrophobic interaction
BibTeX
@article{ikeda2026beyond,
title = {{Beyond Multiscale: A Neural-Network-Unified Molecular Model for Asymmetry-Free Dynamics Simulation}},
author = {Takuma Ikeda and Hiroshi C. Watanabe and Haruyuki Nakano},
journal = {Journal of Chemical Theory and Computation},
year = {2026},
month = oct,
doi = {10.1021/acs.jctc.6c01037},
url = {https://doi.org/10.1021/acs.jctc.6c01037},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Beyond Multiscale: A Neural-Network-Unified Molecular Model for Asymmetry-Free Dynamics Simulation
AU - Takuma Ikeda
AU - Hiroshi C. Watanabe
AU - Haruyuki Nakano
PY - 2026
DA - 2026/10/02
JO - Journal of Chemical Theory and Computation
DO - 10.1021/acs.jctc.6c01037
UR - https://doi.org/10.1021/acs.jctc.6c01037
AB - Adaptive multiscale simulations, such as adaptive quantum mechanics (QM)/molecular mechanics (MM), are indispensable for investigating complex solution dynamics but historically suffer from a fundamental dilemma: the dynamic exchange of identical solvent molecules across different model resolutions introduces an energetic asymmetry under solvent exchange (specifically, a permutation asymmetry where physically identical molecules are treated indistinguishably yet subjected to different force fields depending on their instantaneous positions). This asymmetry forces a prohibitive trade-off between severe structural distortions and the violation of Hamiltonian conservation (continuous energy drift). To resolve this issue, we propose the “Integrated Molecular Model,” a paradigm-shifting framework that unifies multiple resolutions into a single, dynamically consistent potential. By employing a neural network that takes only the multiscale weight functions as input descriptors, our model learns to systematically cancel the nonphysical transition forces originating from spatial discontinuities. As a proof-of-concept, we validated this unified framework on bulk water, aqueous ionic solutions, and reactive hydronium ions. Our approach not only faithfully restores artifact-free solvation structures but also strictly conserves the Hamiltonian, enabling stable, long-term simulations under the microcanonical (NVE) ensemble. Consequently, the model successfully eliminates artificial thermostat-induced breathing modes, allowing for the rigorous evaluation of intrinsic dynamical properties such as diffusion coefficients. Furthermore, owing to its physically informed network design, the correction potential introduces less than 2% additional computational overhead and demonstrates remarkable zero-shot transferability across chemically similar ionic species. By eradicating the inherent asymmetry of adaptive multiscale modeling, this highly scalable framework establishes a robust foundation for high-fidelity simulations of complex condensed-phase phenomena, including reactive dynamics at solid–liquid interfaces and ion transport in battery electrolytes.
SN - 1549-9618
SN - 1549-9626
ER -