Transferable Force Field for Gallium Nitride Crystal Growth from the Melt Using On-The-Fly Active Learning

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Engineering

Background:

• Atomic-scale simulations of reactive processes are computationally limited by the lack of accurate semiempirical force fields and the high cost of ab initio methods.
• Simulating complex processes like gallium nitride (GaN) thin film growth via additive manufacturing requires efficient and accurate modeling tools.

Purpose of the Study:

• To develop a nonparameterized, machine learning-based force field for atomic-scale simulations.
• To enable accurate and efficient modeling of the gallium nitride (GaN) crystallization process in additive manufacturing.

Main Methods:

• Utilized an "on-the-fly" active learning technique to create a novel machine learning force field.
• Developed a single force field capable of modeling solid, liquid, and gas phases involved in GaN synthesis.
• Validated computational predictions against experimental measurements and ab initio calculations.

Main Results:

• The machine learning force field achieves accuracy comparable to density functional theory with the speed of classical molecular dynamics.
• Successfully modeled the crystallization of gallium nitride (GaN) from liquid and gaseous precursors.
• Demonstrated the force field's ability to simulate solid-liquid interfaces and GaN crystallization from the melt.

Conclusions:

• The developed transferable, nonparameterized force field significantly enhances the accuracy and efficiency of simulating reactive processes.
• This advancement opens new avenues for precisely modeling liquid-phase epitaxial growth and establishing robust additive manufacturing models for GaN thin films.