The size and morphology of defect clusters formed during primary damage play an important part in the subsequent evolution of the micro-structure of irradiated materials. Molecular dynamics (MD) simulations of collision cascades in W have been carried out using five interatomic potentials (IAP) including the quantum accurate machine learned (ML) spectral neighbor analysis potential (WSNAP), the ML based tabGAP potential and three embedded atom method (EAM) based potentials. A total of 3500 MD simulations with the primary knock-on atoms (PKA) having energys 5, 10, 20, 50, 75, 100 and 150 keV were carried out. The PKA are launched in hundred random directions at each of the PKA energies to obtain statistically valid results. Analysis using CSaransh, a web based tool to analyze a large collision cascade database, was carried out to obtain the number of defects (individual and in clusters), the defect cluster morphologies, the defect cluster size distributions and the number of sub-cascades formed. It is seen that <1 1 1> clusters dominate across all the inter-atomic potentials (IAP), except for the WSNAP potential. The WSNAP potential shows significantly higher number of C-15 like rings across all PKA energies. The WSNAP potential also shows a higher fraction of in-cluster vacancies. It is also seen that the stiffness and range of IAP do not affect the number of defects, defect clustering or vacancy clusters, but shows a tendency to form more sub-cascades for PKA energies less than 60 keV.
Comment: 20 pages, 9 figures, 2 tables