Machine-learning approach for local classification of crystalline structures in multiphase systems.

Area of Science:

• Materials Science
• Computer Science
• Physics

Background:

• Machine learning is a rapidly growing field with diverse applications.
• Accurate identification of crystal structures is crucial in various scientific domains.
• Existing methods for classifying crystalline structures can be limited, particularly in complex or disturbed systems.

Purpose of the Study:

• To propose a novel neural network-based method for the local identification of crystal structures.
• To classify face-centered cubic (fcc), hexagonal close-packed (hcp), and body-centered cubic (bcc) structures within mixed-phase systems.
• To evaluate the performance of the proposed method against established techniques.

Main Methods:

• Utilizing a neural network for local identification of crystal structures.
• Applying the method to a mixed-phase Yukawa system containing fcc, hcp, and bcc clusters, alongside disordered particles.
• Comparing the neural network approach with existing identification methods.

Main Results:

• The proposed neural network method significantly increases the quality of crystal structure identification.
• The technique demonstrates high efficacy even for highly disturbed crystalline lattices.
• The method provides a flexible and robust approach to classifying crystalline structures based solely on particle positions.

Conclusions:

• The developed neural network method offers a significant improvement for identifying crystal structures in complex and disturbed systems.
• This approach provides valuable insights into the nature of highly disturbed crystalline structures.
• The technique's robustness and flexibility make it a promising tool for materials science and physics research.