Faster Molecular Dynamics with Neural Network Potentials via Distilled Multiple Time-Stepping and Non-Conservative Forces
Nicolaï Gouraud, Côme Cattin, Thomas Plé, Olivier Adjoua, Louis Lagardère, Jean-Philip Piquemal · 2026
Read Faster Molecular Dynamics with Neural Network Potentials via Distilled Multiple Time-Stepping and Non-Conservative Forces by Nicolaï Gouraud, Côme Cattin, Thomas Plé, Olivier Adjoua, Louis Lagardère, Jean-Philip Piquemal (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-14975.txt.
Description
Gouraud and colleagues speed up molecular dynamics run with neural network potentials. Their DMTS-NC scheme distills a fast, simplified model that yields non-conservative forces, coupled through a reversible reference propagator to an accurate conservative potential. The distilled design keeps key physical priors, such as rotation equivariance and cancellation of atomic force parts. That steadies the distillation and keeps the two models in close agreement on forces.
The scheme proves more stable and efficient than its conservative twin, with extra speedups of 15 to 30 percent over DMTS and no fine-tuning needed. Hydrogen mass repartitioning and high hydrogen friction stretch the largest step to 10 femtoseconds while holding accuracy. A proof of principle on MACE-OFF23 runs between three and six times faster than a single-step run.
Catalog record
| CREATOR | Nicolaï Gouraud, Côme Cattin, Thomas Plé, Olivier Adjoua, Louis Lagardère, Jean-Philip Piquemal |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 34 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-14975