Structure of an Ultrathin Oxide on Pt₃Sn(111) Solved by Machine Learning Enhanced Global Optimization

Area of Science:

• Materials Science
• Computational Chemistry
• Surface Science

Background:

• Surface structure determination often relies on comparing experimental data with simulations.
• Complex surface structures require reliable, theory-based search algorithms, which are currently limited by computational cost and complexity.
• Machine learning (ML) offers a promising avenue to overcome these limitations.

Purpose of the Study:

• To develop and demonstrate a novel, efficient, and robust algorithm for solving unknown surface structures.
• To leverage machine learning for accelerated energy estimation and diverse population generation within an evolutionary algorithm framework.
• To provide a broadly applicable computational tool for surface science studies.

Main Methods:

• An evolutionary algorithm was employed for structure searching.
• Machine learning was integrated for accelerated energy estimation and diverse population generation.
• The algorithm was tested on determining the unknown (4x4) surface oxide structure on Pt3Sn(111) using limited experimental input.

Main Results:

• The developed algorithm successfully determined the complex (4x4) surface oxide structure on Pt3Sn(111).
• The machine learning-enhanced evolutionary approach proved efficient and robust in solving the surface structure.
• The method demonstrated the potential to replace manual, intuition-based model generation in surface studies.

Conclusions:

• Machine learning-accelerated evolutionary algorithms are highly effective for determining complex surface atomic structures.
• This computational approach significantly enhances the efficiency and reliability of surface structure analysis.
• The methodology is expected to find broad application in surface science research, advancing the field beyond traditional methods.