Early stages of self-healing at tungsten grain boundaries from ab initio machine learning simulations

Nanostructured tungsten has been reported as a possible alternative plasma–facing material due to its potential ability to self–heal radiation-induced defects, a property that is attributed to its high density of grain boundaries (GB). Here, we study the initial stages of self-healing at tungsten interfaces with molecular dynamics simulations driven by a machine-learning interatomic potential tailored to one of the most common GBs found in experiments. Our model accurately reproduces the ab initio potential energy surface derived from density functional theory (DFT) calculations and outperforms previously reported empirical and machine learning interatomic potentials in predicting defect energetics. The simulations reveal low-temperature defect migration to GBs driven by rapid dumbbell-like ordering and subsequent accommodation along GB grooves. In contrast to empirical potentials, which predict unexpected GB degradation at high temperatures after defect migration, our model maintains stable GB motifs over the investigated temperature range. The temperature-dependent defect counts, evaluated using an Arrhenius-like fit, yield an average interstitial migration energy of 0.048 eV, in agreement with experiment. This work underscores the capabilities of ab initio machine learning simulations in accurately modeling defect-GB interactions and highlights their potential to contribute to the development of radiation tolerant materials. Nanostructured tungsten is being investigated for use as a plasma-facing material in the nuclear industry, due to the possibility of self-healing of radiation-induced defects by the high grain boundary content. Here, a machine-learning interatomic potential is used to investigate the early stages of this process.