AEM8553

The design and discovery of new molecules and materials is the foundation of nearly every technological revolution and represents one of the principal drivers for scientific pursuit. The development of quantum mechanics provided exact mathematical laws that in principle can predict the properties and behavior of any molecular and material system, but their direct application is far too complex for many real-world problems. As a result, the development of approximate practical methods that can capture the main features of molecular interactions has become a major research focus. In recent years, data-driven approaches based on machine learning (ML) together with robust uncertainty quantification (UQ) are revolutionizing the field by providing models with accuracy comparable to quantum methods at a significantly reduced computational cost.

This course covers the fundamentals of molecular simulation methods for materials and chemistry (molecular statics, molecular dynamics, Monte Carlo), with a special emphasis on data-driven ML and UQ approaches. The course will consist of a mixture of lectures, in-class exercises, and hands-on work with state-of-the-art software packages for developing physics-based and ML models for atomic interactions (i.e. interatomic potentials), and using them with high-performance molecular simulation codes and packages (LAMMPS, ASE, OpenKIM, ColabFit, KLIFF).

000356

03308

Periodic Fall & Spring