Structure and Energetics of Chemically Functionalized Silicene: Combined Density Functional Theory and Machine Learning Approach.

It is crucial to control and comprehend the interaction between elemental adsorbates and two-dimensional materials to drive future generations of electronic, sensing, and energy applications. One such material, particularly interesting from the perspective of tunability, is silicene-the silicon-based cousin of graphene. In this work, we investigate nearly 2000 atomic adsorption models on silicene via a combination of density functional theory (DFT) and machine learning (ML). Different systems with varied adsorption geometries, element identities, and surface coverages were optimized using spin-polarized DFT, and the most stable configurations were selected based on adsorption energy. This information was used to train various ML models, including tree-based models and artificial neural networks, to predict adsorption geometry (classification) and adsorption energy (regression). The current hybrid DFT + ML approach provides a transferable framework for high-throughput screening of element-functionalized silicene and other 2D surfaces, which is of immense importance in directing surface modification strategies in electronic and catalytic device engineering.