Gaussian approximation potentials for accurate thermal properties of two-dimensional materials

Area of Science:

• Materials Science
• Computational Physics
• Condensed Matter Physics

Background:

• Two-dimensional materials (2DMs) possess remarkable flexibility and thermal properties, crucial for advanced applications.
• Accurate simulation of these properties relies on precise interatomic interaction models.
• First principles methods are accurate but computationally intensive; classical force fields are efficient but less accurate.

Purpose of the Study:

• To develop a systematic procedure for creating accurate and efficient machine learning interatomic potentials for 2DMs.
• To validate the developed Gaussian Approximation Potentials (GAPs) against high-accuracy first principles calculations.
• To demonstrate the utility of GAPs in simulating thermal properties of 2DMs.

Main Methods:

• Development of Gaussian Approximation Potentials (GAPs) trained on Density Functional Theory (DFT) data for graphene, silicene, and h-XN (X = B, Al, Ga).
• Calculation of harmonic and anharmonic force constants (up to fourth order) using GAPs and DFT.
• High-throughput interactive platform for the hive (HIPHIVE) calculations to assess higher-order force constants.
• Molecular dynamics simulations using GAPs to compute phonon density of states and thermal conductivity.

Main Results:

• GAPs accurately reproduced phonon dispersion curves and lattice thermal conductivity compared to DFT.
• HIPHIVE calculations confirmed the first-principles level accuracy of GAPs for interatomic forces.
• Molecular dynamics simulations with GAPs showed excellent agreement with DFT for high-temperature behavior.

Conclusions:

• The developed GAPs provide a computationally efficient and accurate method for simulating thermal properties of 2DMs.
• This approach bridges the accuracy gap between first principles and classical force fields for materials simulations.
• The validated GAPs are suitable for high-temperature molecular dynamics simulations of 2D materials.