Predicting phase behavior of grain boundaries with evolutionary search and machine learning

Area of Science:

• Materials Science
• Computational Materials Science
• Condensed Matter Physics

Background:

• Grain boundary phase transitions are crucial for material properties but historically challenging to study computationally.
• Existing atomistic modeling methods lack robust tools for predicting interface structures, hindering research.
• Understanding grain boundary polymorphism is key to controlling material behavior at the nanoscale.

Purpose of the Study:

• To develop a robust computational tool for efficient grand-canonical grain boundary structure search.
• To design an automated clustering analysis for identifying distinct grain boundary phases.
• To investigate the phase behavior and structural polymorphism of symmetric tilt boundaries in Copper (Cu).

Main Methods:

• Development of an evolutionary algorithm-based computational tool for grain boundary structure prediction.
• Implementation of a clustering analysis to automatically classify different grain boundary phases.
• Application of the developed tool to a model system of symmetric tilt boundaries in Cu.

Main Results:

• Discovery of a rich polymorphism in grain boundary structures for symmetric tilt boundaries in Cu.
• Identification of new ground and metastable grain boundary states by exploring varying atomic densities.
• Demonstration that grain boundaries across the entire misorientation range exhibit multiple phases and structural transitions.

Conclusions:

• Grain boundary phase behavior is likely a general phenomenon, extending beyond the studied model system.
• The developed computational tool significantly advances the atomistic modeling of grain boundary phase transitions.
• This work opens new avenues for exploring and controlling interface properties in materials.