Advancing Thermal Management with Machine-Learning Potentials on Boron Nitride (BN) and Other Group 13 Nitrides

Toachieve seamless heat dissipation, it is essentialto use materialswith high thermal conductivity to improve thermal management. In thisstudy, we have utilized ab initio and machine learning techniquesto systematically explore the lattice thermal conductivity of BN andother group 13 nitride based bulk and bilayer materials. By employingdata-driven training of potentials of different atomic configurationsat different time steps obtained from the AIMD data, we have demonstratedthe comparability of the results obtained from the machine-learnedpotentials compared to the density functional theory (DFT) based calculationson thermal conductivity. Furthermore, we examined the significanceof four phonon interactions in group 13 nitrides by comparing thecalculated values with the available experimental values. Notably,bilayer AlN exhibits a high thermal conductivity (881 W m(-1) K-1) due to its stronger covalent bonding, whichcontradicts the trend observed from B to In. Our study highlightsthat the machine learning potential-based approach can provide moreaccurate results than DFT, paving the way for future robust investigationsof materials using high-throughput screening techniques.

Automated summary

In order to enhance thermal management, materials with high thermal conductivity are crucial for achieving seamless heat dissipation. Through the utilization of ab initio and machine learning techniques, this study systematically explores the lattice thermal conductivity of different group 13 nitride based bulk and bilayer materials, demonstrating the comparability of results obtained from machine-learned potentials and density functional theory based calculations. Additionally, the importance of four phonon interactions in group 13 nitrides is examined by comparing calculated values with experimental data. The study highlights the potential of machine learning potential-based approach in providing more accurate results than DFT and paving the way for future materials investigations using high-throughput screening techniques.