Molecular dynamics study of monocrystalline silicon sputtering with low-energy argon ions: Interatomic potentials, current density, thermostat geometry, and ion energy effects

This paper presents molecular dynamics modeling of monocrystalline silicon sputtering by argon ions with energies below 1 keV. The key challenge is the correct combination of interatomic potentials for Si–Si, Si–Ar, and Ar–Ar interactions. Several classical potentials were tested, and the influence of simulation cell size, thermostat geometry, and effective current density was systematically investigated. Model accuracy was verified by comparing calculated sputtering yields with experimental data for (100) and (111) silicon orientations. All classical potentials systematically underestimate the sputtering yield by 40%–60%, providing only qualitative agreement with experiment. The best quantitative agreement was achieved using a machine-learned interatomic potential (MLIP) trained on density functional theory data. The MLIP reproduces the experimental sputtering yield dependence with high accuracy while maintaining computational efficiency comparable to classical potentials. The effective current density does not affect the sputtering yield, which is determined by the ballistic cascade stage. However, current density critically influences thermal conditions and defect evolution: Low density results in numerous small argon bubbles, while high density leads to large bubbles. The main outcome is a systematization of critical molecular dynamics modeling parameters for silicon ion sputtering, demonstrating that further progress requires developing MLIP potentials with representative training sets.