In silico explorations of matter at extreme conditions: high-throughput and machine-learning calculations for materials and planetary science

Atomistic simulations of materials provide valuable information for broad research areas, including materials science, physics, chemistry, and biology. Two developments in recent years have elevated the field to a new era. Firstly, high-throughput calculations such as crystal structure prediction and database mining, mostly in conjunction with electronic structure calculations, have turned atomistic simulations into a reliably predictive science, uncovering new classes of materials with desirable properties, and allowing targeted synthesis and characterization efforts. Secondly, machine learning techniques have sped up the simulations by orders of magnitude, by providing accurate yet computationally affordable interatomic potentials. In this talk, I will illustrate the use of these developments with a particular focus on matter at extreme pressure and temperature conditions. There, where chemical intuition and experimental capabilities are at their limits, calculations provide unique insights into matter’s exciting new properties. Amongst these, I will draw on our recent work on new collective states of matter in the elements, high-temperature polyhydride superconductors, and materials’ properties deep inside rocky and icy planetary bodies.