Uniaxial strain-induced cascade defect evolution in molybdenum: An atomistic study with a new machine-learning potential

The evolution of irradiation defects in nuclear structural materials is profoundly influenced by mechanical stress. However, a predictive understanding of this coupling at the atomic scale remains challenging. In this work, we develop a machine-learning moment tensor potential (MTP) suitable for cascade simulations and widely validate its accuracy. Utilizing this potential, we systematically perform molecular dynamics simulations of collision cascades with a focus on uniaxial [111] strain across a wide range of temperatures and PKA energies. These are supplemented by simulations under [100] and [110] strains at 673 K and 10 keV to assess crystallographic orientation effects. Our results reveal that the thermal spike defect yield increases monotonically with tensile strain, regardless of temperature; the stable defect number only increases with strain changing from compressive to tensile at low temperatures, while the strain effect is not significant at high temperatures. Moreover, the orientational response of self-interstitial atoms to strain is found to be highly dependent on the crystallographic direction of the applied strain. Tensile strain along [111] dirction increases the lattice spacing and reduces the formation energy of interstitials, causing self-interstitial atoms to preferentially align along the 〈1 1 1〉 direction and significantly increasing the yield of large-sized clusters and dislocation loops. This work provides an atomic-scale basis for understanding the irradiation defect evolution of Mo under actual service stress.