Stress-Driven Grain Boundary Structural Transition in Diamond by Machine Learning Potential

Understanding the structural dynamics of carbon grain boundaries, particularly in diamond, is essential for advancing next-generation device applications. Carbon's diverse allotropes, driven by its versatile chemical bonding, hold immense potential, yet analyzing these boundaries is challenging due to the limitations of experimental techniques and the computational demands of ab initio molecular dynamics simulations. In this study, a machine learning-based molecular dynamics potential, rigorously trained on ab initio data, that accurately predicts structural transitions in incoherent twin boundaries within diamond is introduced. This potential reveals the atomic-scale mechanisms driving these transitions and identifies an 80% reduction in interfacial thermal conductance during the grain boundary transition. These findings provide deep insights into the complex behavior of diamond grain boundaries, uncovering a novel mechanism that regulates thermal properties and paving the way for enhanced thermal management in diamond-based technologies.