Development and application of Machine Learning Potentials - Analyzing Phase behavior in complex materials

This research focuses on the development and application of Machine Learning Potentials (MLPs) for analyzing and understanding the defining features of complex materials, particularly phase behavior.

Building on previous advances in the field, such as the Behler-Parrinello Neural Network (BPNN) and more recent approaches such as MACE (Multiscale Atomic Cluster Expansion), the goal of this project is to apply and further refine these methods to provide more accurate predictions and computationally efficient models. This will allow a deeper exploration of complex material properties beyond the capabilities of current ab initio methods, facilitating access not only to phase transitions but also to conductivity, energy landscapes and many other crucial features.

In particular, the project will investigate, among other things, the behavior of confined water, a system of particular interest due to its unique phase behavior and its relevance to several scientific fields.