| Gas-surface reactions play a fundamental role in various interfacial phenomena,such as interstellar/atmospheric chemistry,heterogeneous catalysis,and crystal growth.However,accurate modeling of gas-surface reaction dynamics requires a globally accurate reactive potential energy surface(PES),which traditionally specialized for one molecule-surface system,neglecting surface atom motion and electron-hole pairs(EHPs)effects,and without transferability to other surfaces.In this context,we report our works on the H/Cu system,demonstrating the capability of the embedded atom neural network(EANN)method in describing surface degrees of freedom and developing transferable PES for gas-surface reactions.Besides,through the study of the reaction of H+Ge(111)c(2 ×8),which was found with a non-adiabatic channel in a recent experiment,we show the potential of combining the EANN-PES with non-adiabatic dynamics method to describe systems beyond the Born-Oppenheimer approximation.Firstly,we propose a novel machine-learned EANN-PES to investigate the dissociative chemisorption of hydrogen molecules on multiple low-index copper surfaces.Trained with limited data,this PES enables a uniformly and chemically accurate description of dissociative adsorption of H2/D2 on Cu(111)/Cu(100)/Cu(110)and offers quantitative insights to the remarkable surface temperature effect.Moreover,this PES is transferable to describe the dynamics of H2 dissociation on Cu(211)without learning any data on that stepped surface,which can be further improved when adding only a small amount of points.Given the achievement of EANN-PES in chemically accurately describing multiple surfaces in the H/Cu system,next work focuses on investigating Eley-Rideal(ER)reactions in this system,which is the prototype reaction for studying ER reactions at gas-surface interfaces,i.e.,the reaction between incident H/D atoms and pre-covered D/H atoms on Cu(111)surface.Compared to molecular scattering at surfaces,the ER reaction spans a very large configuration space and poses a challenging task for theoretical studies for a long time.By developing a high-dimensional neural network PES,we can describe this ER reaction including all molecular and surface degrees of freedom.Thanks to the high efficiency of this PES,we are able to perform extensive quasi-classical molecular dynamics simulations with inclusion of excitation of low-lying EHPs,which yield generally good agreement with various experimental results.Besides,the isotopic and/or EHP effects in total reaction cross sections and distributions of the product energy,scattering angle,and individual ro-vibrational states have been more clearly shown and discussed.Finally,we perform the first principles non-adiabatic molecular dynamics(NAMD)simulations on energetic H atoms scattered from a reconstructed Ge(111)c(2×8)surface using a newly developed EANN-PES.By studying the characteristics of H collisions with the semiconductor Ge surface,we process the collision trajectories obtained from the adiabatic PES using a mixed quantumclassical method,which demonstrate the different dynamic behaviors of the system at different sites in real-time and qualitatively explain the non-adiabatic phenomenon of effectively promoting electrons into the conduction band during collisions.Through these studies on the H/Cu system,we demonstrate the powerful capability of the EANN method in constructing PESs applicable to gas-surface reactions.These PESs provide an accurate description of the dynamics and enable quantitative comparisons with experimental data,shedding valuable light on important interfacial phenomena in heterogeneous catalysis.Moreover,the transferability of the PESs from low to high index surfaces opens up new avenues for studying the dynamics of structureor step density-sensitive gas-surface reactions relevant to heterogeneous catalysis.Our study also deepens the understanding of the important prototype reaction in ER reactions from a dynamical perspective and opens a new way for further investigations of ER reactions with various initial conditions,surface temperatures,and coverages in the future.Furthermore,by studying the collision process between hydrogen atom and semiconductor,we explain this new theoretical challenge of electron non-adiabatic surface chemistry and demonstrate that electronic excitation dominates the dynamics in collisions of a simple atom with a semiconductor.This opens new horizons for research into non-adiabatic effects in surface chemistry and chemical sensors. |