Machine Learning Interatomic Potentials (MLIPs) have positioned themselves as a key tool for atomistic modeling in materials science. MLIPs cover an expansive range of systems, taking advantage of the highly accurate electronic structure calculations based on quantum mechanics, but at a significantly lower computational cost. They allow to scale up atomistic simulations to larger systems, longer timescales, and more complex phenomena; they therefore significantly contribute to the acceleration of the discovery of novel structural and functional materials, and in the advancements in our understanding of matter. Ground-breaking bodies of work have been published since the seminal work of Behler and Parrinello in 2007 [1], transforming the field into a rapidly evolving research discipline [2-16]. However, alongside these advancements, a crucial challenge emerges: the need for standardized protocols for MLIP generation and storage, as well as comprehensive, accessible databases for ab initio datasets.

This workshop is organized as part of targeted project DIAMOND.

ORGANISERS

• Magali Benoit (CEMES, CNRS, Toulouse)
• Arthur France-Lanord (CNRS)
• Noel Jakse (Université Grenoble Alpes)
• Antonino Marco Saitta (IMPMC – Université Pierre et Marie Curie (UPMC) – Paris)