Search papers, labs, and topics across Lattice.
The paper introduces DMTS-NC, a distilled multi-time-step (DMTS) strategy incorporating non-conservative (NC) forces to accelerate molecular dynamics simulations using neural network potentials. This approach couples an accurate conservative potential with a simplified distilled representation optimized for non-conservative force production within a dual-level RESPA formalism. The method enforces physical priors like equivariance and force component cancellation to improve distillation robustness and simulation stability, achieving 15-30% speedups over the conservative DMTS approach.
Molecular dynamics simulations just got a whole lot faster: a new method achieves 15-30% speedups by cleverly using non-conservative forces within a multi-time-stepping scheme.
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. 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 the "holes" in the simpler potential, 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 finetuning 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. As for DMTS, DMTS-NC is applicable to any neural network potential.
