Atomic Modelling of Oxidized Crystalline InAs(100) Surface with First Principles and Bayesian Optimization

Abstract

This study examines the surface reconstruction of the InAs(100)-ζ(4×2), the initial oxygen adsorption sites on the surface and their influence on the reconstructed optimized surface structure using a combined methodology of Density Functional Theory (DFT) calculations and Bayesian Optimization Structure Search (BOSS). A detailed computational model of the InAs(100)-ζ(4×2) surface was systematically reconstructed and optimized across slab thicknesses ranging from 8 to 16 atomic layers, establishing 12 layers as the converged, stable thickness. Subsequently, an oxygen atom was placed near the reconstructed surface, and BOSS was employed to effectively map the potential energy landscape and predict the optimal adsorption configurations. A one-dimensional BOSS exploration optimized the oxygen atom’s vertical position above the surface, and a two-dimensional BOSS exploration sampled possible lateral adsorption coordinates across the surface unit cell. Predicted atomic adsorption sites of the oxygen atoms identified by BOSS were fully relaxed with DFT to confirm stability and obtain precise adsorption energies. The results reveal ten energetically favourable oxygen adsorption configurations on InAs(100)-ζ(4×2) reconstruction. In the most stable structures, oxygen binds between the surface indium-arsenic atoms, consistent with the general tendency of oxygen to preferentially oxidize group III-V elements. The BOSS methodology successfully located these adsorption sites with far fewer energy evaluations than a conventional potential energy surface search, demonstrating its efficiency for complex surface systems. In future studies, these findings can be used as the foundation for further oxygen adsorption research on the surface to improve passivation strategies and support the development of more accurate models for InAs/oxide interfaces in electronic and optoelectronic devices.