2D Janus Halogenated Silicene database

Description

Recent advances in theoretical computing and machine learning (ML) have facilitated the in-depth exploration of two-dimensional (2D) materials. We combines ML with 2D materials screening with the aim of exploring novel halogenated silicene-based materials for applications in photocatalysis and solar cells. We construct a data-driven framework that combines first-principles computation and self-contained databases for supervised learning to develop highly accurate predictive models. This dataset containing 286 halogenated silicene structures, which is 11.5% of the total database, is obtained after Density-functional theory (DFT) calculations.

Steps to reproduce

We proposed a datadriven framework that incorporates first-principles electronic structure calculations and 286 self-contained databases trained by supervised learning to accurately predict the structural stability and electronic properties of halogenated silicene-based compounds. Our model employs feature descriptors based on relative atomic positions in the periodic table and their distinct physical properties. In addition, the pseudopotential file POTCAR and information based on the elemental configuration of surfaces were incorporated into the feature set.After determining the initial feature values, the feature descriptors were normalized to reduce computational biases resulting from differences in feature scales.The initial feature screening process identified 33 and 20 features at the 0.9 and 0.8 thresholds, which indicate important correlations within the initial feature set.

Institutions

Categories

Funding

Licence