(66e) Characterizing Defect Dynamics in Silicon Carbide Using Symmetry-Adapted Collective Variables and Machine Learning Potentials

Solid-state spin defects in silicon carbide (SiC) offer promising platforms for quantum information technologies, including quantum sensing, communication, and computing. The divacancy (VV) in SiC, formed through vacancy diffusion and recombination, serves as a particularly attractive qubit candidate due to its long coherence times, optical addressability, and high-fidelity readout. Engineering and stabilizing these defects require precise control over their formation dynamics at elevated temperatures, which in turn demands accurate and efficient atomistic computational models. However, the high activation barriers of SiC defect dynamics pose a significant challenge for direct ab initio molecular dynamics simulations, while classical empirical force fields lack the necessary accuracy to capture defect behavior.

To address this, we develop machine learning interatomic potentials (MLIPs) that accelerate defect dynamics simulations while maintaining first-principles accuracy. The reliability of MLIPs depends on the diversity of atomic local environments in the training data, particularly for defect structures. We employ an active learning approach to systematically expand the training set, ensuring accurate sampling of defect configurations. Crucially, the indistinguishability of carbon and silicon atoms in defect environments necessitates the use of collective variables (CVs) that respect the crystal's fundamental symmetries. We achieve this by integrating Permutation Invariant Vector (PIV) features with regularized autoencoders in the Permutationally Invariant Networks for Enhanced Sampling (PINES) approach to construct symmetry-adapted CVs.

Using our trained force field, we achieve density functional theory (DFT)-level accuracy in predicting defect transition free energies and charge effects in defect dynamics. Furthermore, by employing Markov state modeling, we analyze the thermodynamics and kinetics of defect transitions across various temperatures, providing insights into the selective formation of divacancies. Our results contribute to the fundamental understanding of SiC defect behavior and offer a computational framework to guide the controlled synthesis of spin defects for quantum technology applications.