Modelling Physical Phenomena in Alloys and Polymers with Quantum Mechanical Accuracy

Explore a Lennard-Jones Centre discussion group seminar on data-driven interatomic potentials and their applications in modelling physical phenomena in alloys and polymers. Delve into the Atomic Cluster Expansion (ACE) framework, which enables quantum mechanical accuracy at reduced evaluation times. Learn about Hyperactive Learning (HAL), a method for rapidly building ACE potentials from scratch. Discover how these techniques are applied to determine polymer density and predict alloy phase transitions, providing insights into precipitate formation and chemical ordering in alloys. Gain understanding of linear regression, Bayesian linear regression, and the comparison between hyperactive learning and active learning. Examine case studies involving longer molecules, titanium, and tungsten, concluding with a comprehensive summary of the seminar's key points.

Introduction
Linear regression
Bayesian linear regression
Hyperactive learning
Hyperactive learning vs active learning
Longer molecules
Alloys
Ace model
Nested sampling
Titanium
Tungsten
Summary