Fundamental Factors Governing Stabilization of Janus 2D-Bulk Heterostructures with Machine Learning

Area of Science:

• Materials Science
• Condensed Matter Physics
• Computational Chemistry

Background:

• Over 6000 2D materials offer vast potential for functional heterostructures in nanoelectronics, sensing, and energy conversion.
• Understanding structure-property relationships in 2D material-bulk heterostructures is crucial for advancing these applications.

Purpose of the Study:

• To investigate nearly 1000 2D Janus and bulk material heterostructures using ab initio simulations and machine learning (ML).
• To deduce structure-property relationships at complex interfaces within these heterostructures.
• To develop predictive ML models for binding energy and interfacial separation.

Main Methods:

• High-throughput van der Waals-corrected density functional theory (DFT) simulations were performed on 51 Janus 2D materials and 19 bulk materials.
• Thermodynamic stability of 1147 resultant heterostructures was analyzed, identifying 828 stable configurations.
• ML models were trained on computed data to predict heterostructure properties.

Main Results:

• 828 out of 1147 simulated Janus 2D-bulk heterostructures were found to be thermodynamically stable.
• ML models accurately predicted binding energy (RMSE: 0.05 eV/atom) and z-separation (RMSE: 0.14 Å).
• Bulk material properties were identified as dominant factors influencing heterostructure energies and interfacial structures.

Conclusions:

• The study provides fundamental insights into 2D-bulk heterostructures, confirming the significant role of bulk material properties.
• Developed predictive ML models can accelerate the discovery and application of novel heterostructures.
• Freely available data in the Ab Initio 2D-Bulk Heterostructure Database (aiHD) supports further research in electronic, quantum computing, sensing, and energy fields.