Machine-learning-accelerated nonadiabatic dynamics at metal surfaces

Nonadiabatic effects that arise from the concerted motion of electrons and atoms at comparable energy and time scales are omnipresent in thermal and light-driven chemistry at metal surfaces. Short-lived excited (hot) electrons can measurably affect molecule-metal reactions by introducing energy dissipation, dynamical steering effects, and by contributing to state-dependent reaction probabilities. Hot electrons, created by plasmonic excitation upon light exposure, can selectively activate chemical reactions at metal catalyst surfaces. I will present our recent efforts to establish molecular dynamics methods able to capture nonadiabatic and quantum effects at metal surfaces. We employ a range of methods to capture hot electron effects such as molecular dynamics with electronic friction and surface hopping dynamics. By combining linear response electronic structure calculations with high-dimensional machine-learning representations, we are able to perform ensemble-averaged nonadiabatic dynamics simulations at surfaces. I will showcase this for examples such as the vibrational state-to-state scattering of NO on Au(111) and light-driven hydrogen evolution on copper. I will also provide a detailed analysis of the limitations of the existing approach and our ongoing efforts to include memory effects and explicit excited-state effects to capture the dynamics of light-driven hot-electron chemistry.