Ultra‑robust machine‑learning models run stable molecular simulations at extreme temperatures

Researchers at The University of Manchester have created a physics‑informed machine‑learning model that can run molecular simulations for unprecedented lengths of time, even at temperatures as high as 1,000 Kelvin. The study, published in Communications Chemistry, explores the first AI-powered model that can keep molecular simulations running safely and smoothly, even when molecules are pushed to extreme conditions. In simple terms, this model stops molecules from "breaking apart" inside the simulation, allowing researchers to study how they behave over long periods and at very high temperatures.