Title: Lowering the Exponential Wall: Accelerating High-Entropy Alloy Catalysts Screening using Local Surface Energy Descriptors from Neural Network Potentials

Abstract: Computational screening is indispensable for the efficient design of high-entropy alloys (HEAs), which hold great potential for catalytic applications. However, the chemical space of HEAs is exponentially vast with respect to the number of constituent elements, and even screening calculations using machine learning potentials can be enormously time-consuming. To address this challenge, we propose a method to rapidly construct models that predict the properties of HEAs from data on monometallic systems (or few-component alloys). The core of our approach is a newly-introduced descriptor called local surface energy ($LSE$), which reflects the local reactivity of solid surfaces at atomic resolution. We successfully created a model using linear regression to screen the adsorption energies of molecules on HEAs based on LSEs from monometallic systems. Furthermore, we made high-precision model development by employing both classical machine learning and quantum machine learning. Using our method, we were able to complete the adsorption energy calculations of CO molecules on 1000 patterns of quinary nanoparticles consisting of 201 atoms within a few hours. These calculations would have taken hundreds of years and hundreds of days using density functional theory and a neural network potential, respectively. Our approach allows accelerated exploration of the vast chemical space of HEAs facilitating the design of novel catalysts.