Revealing Crystallization Mechanism of Gallium Arsenide by Machine Learning Molecular Dynamics Simulation

Area of Science:

• Materials Science
• Computational Chemistry
• Solid State Physics

Background:

• Gallium arsenide (GaAs) is a crucial semiconductor material in microelectronics and optoelectronics.
• Its properties include high electron mobility, moderate band gap, radiation resistance, and high-frequency performance.

Purpose of the Study:

• To systematically investigate the crystallization mechanism of GaAs.
• To reveal phase competition rules and the dominant crystalline phase during GaAs formation.

Main Methods:

• Utilized machine learning molecular dynamics simulations.
• Employed unbiased molecular dynamics simulations combined with deep neural network potentials.
• Predicted the melting temperature of GaAs crystals.

Main Results:

• Undercooling conditions favor faster crystallization of hexagonal wurtzite (WZ) over cubic zinc blende (ZB).
• Cubic zinc blende (ZB) is the dominant phase due to superior thermodynamic stability.
• GaAs melts form mixed crystals primarily composed of the ZB phase.
• Short-range order is identified as the key driver for rapid crystal growth at the solid-liquid interface.

Conclusions:

• The microscopic mechanism of the solid-liquid phase transition in GaAs has been elucidated.
• Provides theoretical support for the design, preparation, and performance control of GaAs-based semiconductor materials.