AI-enhanced prediction of high-pressure phase transformations induced by electron-phonon interaction using random structure searching simulations

Non-perturbation density functional theory calculations of electron-phonon interactions revealed the quantum degeneracy effect on phase transformations under high pressures. The degenerate energy levels generally occur around symmetry points and lines of its band structure in a solid, rendering reactive centers of phase transformations in momentum space. Immersive visualization of Lewis structures as a virtual reality approach enhances random structure searching simulations using machine learning for efficiently predicting high-pressure phase transformations. Such immersive visualization of Lewis structures furnishes a chemical bonding space to supervise random structure searching simulations using machine learning, providing both chemical and symmetry constraints to determine reactive centers of phase transformations. Further non-perturbation calculations of electron-phonon interactions can show details of the quantum degeneracy effects on high-pressure phase transformations. As an application, we performed predictive simulations of the high-pressure phase transformations of hydrazine borane induced by electron-phonon interaction. The high-pressure crystal structure of hydrazine borane obtained by simulation agrees well with experimental data.