Modelling atomic and nanoscale structure in the silicon-oxygen system through active machine learning.

Area of Science:

• Materials Science
• Computational Chemistry
• Nanotechnology

Background:

• Silicon-oxygen compounds are fundamental in minerals, semiconductors, and catalysis.
• Understanding nanoscale heterogeneity in these materials is a key scientific challenge.
• Existing methods struggle to capture complexity beyond the atomic scale.

Purpose of the Study:

• To develop a unified computational method for describing the full silicon-oxygen system.
• To accurately model nanoscale heterogeneity in diverse silicon-oxygen materials.
• To demonstrate the power of active machine learning for materials discovery.

Main Methods:

• Utilized atomistic machine learning integrated with an active-learning workflow.
• Applied the method to various silicon-oxygen systems including high-pressure silica, surfaces, and aerogels.
• Investigated the structure of amorphous silicon monoxide.

Main Results:

• Achieved a unified computational description of the complex Si-O system.
• Successfully modeled nanoscale heterogeneity across different material types.
• Validated the efficacy of the active machine learning approach.

Conclusions:

• Atomistic machine learning provides a powerful tool for understanding complex materials.
• Active learning enables accurate modeling of nanoscale structural complexity in functional materials.
• This approach advances the study of silicon-oxygen compounds and beyond.