Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore foundational models for materials chemistry in this DDPS seminar featuring Professor Gabor Csanyi from the University of Cambridge. Learn about the revolutionary computational breakthrough achieved over the past 15 years in analytically fitting the Born-Oppenheimer potential energy surface for extended material systems, resulting in reactive, many-body force fields with evaluation costs of 0.1-10 ms/atom/cpu core and accuracies reaching a few meV/atom. Discover how the latest architectures leverage many-body symmetric descriptions of local geometry and equivariant message passing networks to achieve unprecedented performance. Examine the surprising stability of models trained on diverse datasets across the entire periodic table, with particular focus on the MACE-MP-0 model trained on approximately 150,000 real and hypothetical small inorganic crystals that demonstrates remarkable generalization capabilities for stable molecular dynamics simulations across crystals, liquids, surfaces, clusters, molecules, and their combinations. Understand how these foundation models are democratizing quantum-accurate large-scale molecular simulations by lowering barriers to entry in the field, and explore applications in organic chemistry where training on small molecules enables accurate condensed phase system simulations and first-principles prediction of properties like hydration free energies. Gain insights into the integration of machine learning techniques with materials modeling, statistical challenges in molecular dynamics, and enhanced sampling algorithms for exploring configuration spaces of materials and molecules from a leading expert in atomistic simulation and multiscale modeling.
