Machine Learning Reveals New Phase of Solid Hydrogen

Researchers at the University of Illinois Urbana-Champaign have used a machine-learning technique to discover a previously unknown high-pressure solid phase of hydrogen, offering insights into the behavior of the universe's most abundant element under extreme conditions and the composition of gaseous planets like Jupiter and Saturn. The team developed a machine-learning model trained with quantum Monte Carlo simulations, which allowed them to simulate an unprecedented number of atoms and discover the new phase. The discovery has inspired experimentalists to revisit the problem and explore hydrogen's behavior under extreme conditions.