arXiv subjects

Thomas Plé

Publications and source records attributed to Thomas Plé.

13 recordsLinked to original sources

arXiv2026

Faster Molecular Dynamics with Neural Network Potentials via Distilled Multiple Time-Stepping and Non-Conservative Forces

Following our previous work (J. Phys. Chem. Lett., 2026, 17, 5, 1288-1295), we propose the DMTS-NC approach, a distilled multi-time-step (DMTS) strategy using non-conservative (NC) forces to further accelerate atomistic molecular dynamics simulations using foundation neural network models such as FeNNix-Bio1. There, a dual-level reversible reference system propagator algorithm (RESPA) formalism couples a target accurate conservative potential to a simplified distilled representation optimized for the production of non-conservative forces. Despite being non-conservative, the distilled architecture is designed to enforce key physical priors, such as equivariance under rotation and cancellation of atomic force components. These choices facilitate the distillation process and therefore improve drastically the robustness of simulation, significantly limiting abnormal discrepancies between the two models, thus achieving excellent agreement with the forces data. Overall, the DMTS-NC scheme is found to be more stable and efficient than its conservative counterpart with additional speedups reaching 15-30% over DMTS. Requiring no fine-tuning steps, it is easier to implement and can be pushed to the limit of the systems physical resonances to maintain accuracy while providing maximum efficiency. We obtain additional speedup by combining hydrogen mass repartitioning (HMR), High Hydrogen Friction (HHF) to further extended the largest timestep up to 10fs of our schemes while conserving stability and accuracy. As for DMTS, DMTS-NC is applicable to any neural network potential and can be applied to approaches that are computationally heavier than FeNNix-Bio1. We show a proof of principle applying the approach to the distillation of MACE-OFF23 with consequent speedups ranging from 3.66 to 5.64 compared to single timestep.

physics.chem-ph↗

arXiv2026

The Convergence Frontier: Integrating Machine Learning and High Performance Quantum Computing for Next-Generation Drug Discovery

Integrating quantum mechanics into drug discovery marks a decisive shift from empirical trial-and-error toward quantitative precision. However, the prohibitive cost of ab initio molecular dynamics has historically forced a compromise between chemical accuracy and computational scalability. This paper identifies the convergence of High-Performance Computing (HPC), Machine Learning (ML), and Quantum Computing (QC) as the definitive solution to this bottleneck. While ML foundation models, such as FeNNix-Bio1, enable quantum-accurate simulations, they remain tethered to the inherent limits of classical data generation. We detail how High-Performance Quantum Computing (HPQC), utilizing hybrid QPU-GPU architectures, will serve as the ultimate accelerator for quantum chemistry data. By leveraging Hilbert space mapping, these systems can achieve true chemical accuracy while bypassing the heuristics of classical approximations. We show how this tripartite convergence optimizes the drug discovery pipeline, spanning from initial system preparation to ML-driven, high-fidelity simulations. Finally, we position quantum-enhanced sampling as the beyond GPU frontier for modeling reactive cellular systems and pioneering next-generation materials.

quant-ph↗

arXiv2026

Accelerating Molecular Dynamics Simulations with Foundation Neural Network Models using Multiple Time-Step and Distillation

We present a distilled multi-time-step (DMTS) strategy to accelerate molecular dynamics simulations using foundation neural network models. DMTS uses a dual-level neural network where the target accurate potential is coupled to a simpler but faster model obtained via a distillation process. The 3.5 Å-cutoff distilled model is sufficient to capture the fast-varying forces, i.e., mainly bonded interactions, from the accurate potential allowing its use in a reversible reference system propagator algorithms (RESPA)-like formalism. The approach conserves accuracy, preserving both static and dynamical properties, while enabling to evaluate the costly model only every 3 to 6 fs depending on the system. Consequently, large simulation speedups over standard 1 fs integration are observed: nearly 4-fold in homogeneous systems and 3-fold in large solvated proteins through leveraging active learning for enhanced stability. Such a strategy is applicable to any neural network potential and reduces their performance gap with classical force fields.

physics.chem-ph↗

arXiv2025

Probing the partition function for temperature-dependent potentials with nested sampling

Thermodynamic properties can be in principle derived from the partition function, which, in many-atom systems, is hard to evaluate as it involves a sum on the accessible microscopic states. Recently, the partition function has been computed via nested sampling, relying on Bayesian statistics, which is able to provide the density of states as a function of the energy in a single run, independently of the temperature. This appealing property is lost whenever the potential energy that appears in the partition function is temperature-dependent: for instance, mean-field effective potential energies or the quantum partition function in the path-integral formalism. For these cases, the nested sampling must be carried out at each temperature, which results in a massive increase of computational time. Here, we introduce and implement a new method, that is based on an extended partition function where the temperature is considered as an additional parameter to be sampled. The extended partition function can be evaluated by nested sampling in a single run, so to restore this highly desirable property even for temperature-dependent effective potential energies. We apply this original method to compute the quantum partition function for harmonic potentials and Lennard-Jones clusters at low temperatures and show that it outperforms the straightforward application of nested sampling for each temperature within several temperature ranges.

physics.comp-ph↗

arXiv2025

The Q-AMOEBA (CF) Polarizable Potential

We present Q-AMOEBA (CF), an enhanced version of the Q-AMOEBA polarizable model that integrates a geometry-dependent charge flux (CF) term while designed for an explicit treatment of nuclear quantum effects (NQE). The inclusion of CF effects allows matching experimental data for the molecular structure of water in both gas and liquid phases, addressing limitations faced by most force fields. We show that Q-AMOEBA (CF) provides highly accurate results for a wide range of thermodynamical properties of liquid water. Using the computational efficiency of the adaptive Quantum Thermal Bath method, which accounts for NQE at a cost comparable to classical molecular dynamics, we evaluate the robustness and transferability of Q-AMOEBA (CF) by calculating hydration free energies of various ions and organic molecules. Finally, we apply this methodology to the alanine dipeptide and compute the corresponding dihedral angle potential of mean force and hydration free energy. Unexpectedly, the latter quantity displays significant NQE. These results pave the way to a finer understanding of their role in biochemical systems.

physics.chem-ph↗

arXiv2025

Pushing the Accuracy Limit of Foundation Neural Network Models with Quantum Monte Carlo Forces and Path Integrals

We propose an end-to-end integrated strategy to produce highly accurate quantum chemistry (QC) synthetic datasets (energies and forces) aimed at deriving Foundation Machine Learning models for molecular simulation. Starting from Density Functional Theory (DFT), a "Jacob's Ladder" approach leverages computationally-optimized layers of massively GPU-accelerated software with increasing accuracy. Thanks to Exascale, this is the first time that the computationally intensive calculation of Quantum Monte Carlo forces (QMC), and the combination of multi-determinant QMC energies and forces with selected-Configuration Interaction wavefunctions, are computed at such scale at the complete basis-set limit. To bridge the gap between accurate QC and condensed-phase Molecular Dynamics, we leverage transfer learning to improve the DFT-based FeNNix-Bio1 foundation model. The resulting approach is coupled to path integrals adaptive sampling quantum dynamics to perform nanosecond reactive simulations at unprecedented accuracy. These results demonstrate the promise of Exascale to deepen our understanding of the inner machinery of complex biosystems.

physics.chem-ph↗

arXiv2025

Velocity Jumps for Molecular Dynamics

We introduce the Velocity Jumps approach, denoted as JUMP, a new class of Molecular dynamics integrators, replacing the Langevin dynamics by a hybrid model combining a classical Langevin diffusion and a piecewise deterministic Markov process, where the expensive computation of long-range pairwise interactions is replaced by a resampling of the velocities at random times. This framework allows for an acceleration in the simulation speed while preserving sampling and dynamical properties such as the diffusion constant. It can also be integrated in classical multi-timestep methods, pushing further the computational speedup, while avoiding some of the resonance issues of the latter thanks to the random nature of jumps. The JUMP, JUMP-RESPA and JUMP-RESPA1 integrators have been implemented in the GPU-accelerated version of the Tinker-HP package and are shown to provide significantly enhanced performances compared to their BAOAB, BAOAB-RESPA and BAOAB-RESPA1 counterparts respectively.

physics.comp-ph↗

arXiv2024

FeNNol: an Efficient and Flexible Library for Building Force-field-enhanced Neural Network Potentials

Neural network interatomic potentials (NNPs) have recently proven to be powerful tools to accurately model complex molecular systems while bypassing the high numerical cost of ab-initio molecular dynamics simulations. In recent years, numerous advances in model architectures as well as the development of hybrid models combining machine-learning (ML) with more traditional, physically-motivated, force-field interactions have considerably increased the design space of ML potentials. In this paper, we present FeNNol, a new library for building, training and running force-field-enhanced neural network potentials. It provides a flexible and modular system for building hybrid models, allowing to easily combine state-of-the-art embeddings with ML-parameterized physical interaction terms without the need for explicit programming. Furthermore, FeNNol leverages the automatic differentiation and just-in-time compilation features of the Jax Python library to enable fast evaluation of NNPs, shrinking the performance gap between ML potentials and standard force-fields. This is demonstrated with the popular ANI-2x model reaching simulation speeds nearly on par with the AMOEBA polarizable force-field on commodity GPUs (GPU=Graphics processing unit). We hope that FeNNol will facilitate the development and application of new hybrid NNP architectures for a wide range of molecular simulation problems.

physics.chem-ph↗

arXiv2023

Force-Field-Enhanced Neural Network Interactions: from Local Equivariant Embedding to Atom-in-Molecule properties and long-range effects

We introduce FENNIX (Force-Field-Enhanced Neural Network InteraXions), a hybrid approach between machine-learning and force-fields. We leverage state-of-the-art equivariant neural networks to predict local energy contributions and multiple atom-in-molecule properties that are then used as geometry-dependent parameters for physically-motivated energy terms which account for long-range electrostatics and dispersion. Using high-accuracy ab initio data (small organic molecules/dimers), we trained a first version of the model. Exhibiting accurate gas-phase energy predictions, FENNIX is transferable to the condensed phase. It is able to produce stable Molecular Dynamics simulations, including nuclear quantum effects, for water predicting accurate liquid properties. The extrapolating power of the hybrid physically-driven machine learning FENNIX approach is exemplified by computing: i) the solvated alanine dipeptide free energy landscape; ii) the reactive dissociation of small molecules.

physics.chem-ph↗

arXiv2022

Routine Molecular Dynamics Simulations Including Nuclear Quantum Effects: from Force Fields to Machine Learning Potentials

We report the implementation of a multi-CPU and multi-GPU massively parallel platform dedicated to the explicit inclusion of nuclear quantum effects (NQEs) in the Tinker-HP molecular dynamics (MD) package. The platform, denoted Quantum-HP, exploits two simulation strategies: the Ring-Polymer Molecular Dynamics (RPMD) that provides exact structural properties at the cost of a MD simulation in an extended space of multiple replicas, and the adaptive Quantum Thermal Bath (adQTB) that imposes the quantum distribution of energy on a classical system via a generalized Langevin thermostat and provides computationally affordable and accurate (though approximate) NQEs. We discuss some implementation details, efficient numerical schemes, parallelization strategies and quickly review the GPU acceleration of our code. Our implementation allows an efficient inclusion of NQEs in MD simulations for very large systems, as demonstrated by scaling tests on water boxes with more than 200,000 atoms (simulated using the AMOEBA polarizable force field). We test the compatibility of the approach with Tinker-HP's recently introduced Deep-HP machine learning potentials module by computing water properties using the DeePMD potential with adQTB thermostating. Finally, we show that the platform is also compatible with the alchemical free energy estimation capabilities of Tinker-HP and fast enough to perform simulations. Therefore, we study how the NQEs affect the hydration free energy of small molecules solvated with the recently developed Q-AMOEBA water force field. Overall, the Quantum-HP platform allows users to perform routine quantum MD simulations of large condensed-phase systems and will participate to shed a new light on the quantum nature of important interactions in biological matter.

physics.chem-ph↗

arXiv2022

Improving Condensed Phase Water Dynamics with Explicit Nuclear Quantum Effects: the Polarizable Q-AMOEBA Force Field

We introduce a new parametrization of the AMOEBA polarizable force field for water denoted Q-AMOEBA, for use in simulations that explicitly account for nuclear quantum effects (NQEs). This study is made possible thanks to the recently introduced adaptive Quantum Thermal Bath (adQTB) simulation technique which computational cost is comparable to classical molecular dynamics. The flexible Q-AMOEBA model conserves the initial AMOEBA functional form, with an intermolecular potential including an atomic multipole description of electrostatic interactions (up to quadrupole), a polarization contribution based on the Thole interaction model and a buffered 14-7 potential to model van der Waals interactions. It has been obtained by using a Force Balance fitting strategy including high-level quantum chemistry reference energies and selected condensed phase properties targets. The final Q-AMOEBA model is shown to accurately reproduce both gas phase and condensed phase properties, notably improving the original AMOEBA water model. This development allows the fine study of NQEs on water liquid phase properties such as the average H-O-H angle compared to its gas phase equilibrium value, isotope effects etc... Q-AMOEBA also provides improved infrared spectroscopy prediction capabilities compared to AMOEBA03. Overall, we show that the impact of NQEs depends on the underlying model functional form and on the associated strength of hydrogen bonds. Since adQTB simulations can be performed at near classical computational cost using the Tinker-HP package, Q-AMOEBA can be extended to organic molecules, proteins and nucleic acids opening the possibility for the large scale study of the importance of NQEs in biophysics.

physics.chem-ph↗

arXiv2022

Scalable Hybrid Deep Neural Networks/Polarizable Potentials Biomolecular Simulations including long-range effects

Deep-HP is a scalable extension of the \TinkerHP\ multi-GPUs molecular dynamics (MD) package enabling the use of Pytorch/TensorFlow Deep Neural Networks (DNNs) models. Deep-HP increases DNNs MD capabilities by orders of magnitude offering access to ns simulations for 100k-atom biosystems while offering the possibility of coupling DNNs to any classical (FFs) and many-body polarizable (PFFs) force fields. It allows therefore to introduce the ANI-2X/AMOEBA hybrid polarizable potential designed for ligand binding studies where solvent-solvent and solvent-solute interactions are computed with the AMOEBA PFF while solute-solute ones are computed by the ANI-2x DNN. ANI-2X/AMOEBA explicitly includes AMOEBA's physical long-range interactions via an efficient Particle Mesh Ewald implementation while preserving ANI-2X's solute short-range quantum mechanical accuracy. The DNNs/PFFs partition can be user-defined allowing for hybrid simulations to include biosimulation key ingredients such as polarizable solvents, polarizable counter ions, etc... ANI-2X/AMOEBA is accelerated using a multiple-timestep strategy focusing on the models contributions to low-frequency modes of nuclear forces. It primarily evaluates AMOEBA forces while including ANI-2x ones only via correction-steps resulting in an order of magnitude acceleration over standard Velocity Verlet integration. Simulating more than 10 $μ$, we compute charged/uncharged ligands solvation free energies in 4 solvents, and absolute binding free energies of host-guest complexes from SAMPL challenges. ANI-2X/AMOEBA average errors are within chemical accuracy opening the path towards large-scale hybrid DNNs simulations, at force-field cost, in biophysics and drug discovery.

physics.chem-ph↗

arXiv2021

Nuclear Quantum Effects in liquid water at near classical computational cost using the adaptive Quantum Thermal Bath

We demonstrate the accuracy and efficiency of a recently introduced approach to account for nuclear quantum effects (NQE) in molecular simulations: the adaptive Quantum Thermal Bath (adQTB). In this method, zero point energy is introduced through a generalized Langevin thermostat designed to precisely enforce the quantum fluctuation-dissipation theorem. We propose a refined adQTB algorithm with improved accuracy and we report adQTB simulations of liquid water. Through extensive comparison with reference path integral calculations, we demonstrate that it provides excellent accuracy for a broad range of structural and thermodynamic observables as well as infrared vibrational spectra. The adQTB has a computational cost comparable to classical molecular dynamics, enabling simulations of up to millions of degrees of freedom.

physics.chem-ph↗