First-Principles and Machine-Learned Modeling of Atomic Structure and Phase Stability
This community develops computational tools to predict how atoms arrange themselves and how materials behave under extreme pressure and temperature, using quantum mechanics and artificial intelligence to simulate physical properties without physical experiments.
The work centers on calculating electronic structure, phase transitions, and thermal expansion using density functional theory and molecular dynamics. A major current focus is the development of machine learning interatomic potentials, such as neural network models, which allow for large-scale simulations of crystal structures and grain boundaries. Researchers apply these methods to transition metals, hydrogen-helium mixtures, and thin films to map phase diagrams and understand stability under high-pressure conditions, often validated against x-ray diffraction and Raman spectroscopy data.
The largest share of the community's output is found in helium research, accounting for 1.4% of all tracked helium studies, followed by beryllium at 2.1% of its total research volume. Hydrogen research also features prominently, with 997 papers in this group.
The community comprises 27,447 papers, published primarily in Physical Review B, Journal of Applied Physics, and The Journal of Chemical Physics.
Recent work includes the development of universal neural network potentials for energetic materials and advanced materials modeling, as well as machine learning-accelerated discovery of electrocatalysts for the hydrogen evolution reaction.