A Comparative Study of Multiwalled WS2 Nanotubes Using Universal Machine Learning Interatomic Potentials and an Atomistic Force Field.

This study explores the properties of multiwalled WS2 nanotubes using the CHGNet universal machine-learning interatomic potential (uMLIP) and the atomistic force field previously developed by our team. For the first time, both the original ("vanilla") and fine-tuned uMLIPs were benchmarked against quantum chemical calculations and atomistic force field models for single-walled and multiwalled nanotubes. The comparison reveals excellent agreement in structural features between the approaches, although the CHGNet uMLIP tends to underestimate interlayer interaction contributions, affecting certain energy-related parameters. However, uMLIPs are a significantly less labor-intensive alternative to traditional force field engineering. With further refinement, they could be used to study multiwalled nanotubes.