Primary radiation damage in Si1−xGex alloys: Molecular dynamics study with machine-learning interatomic potential

[Display omitted] •A machine learning potential of the Si1-xGex system is developed to perform MD simulation with DFT accuracy.•The potential offers a precise depiction of the static properties, defect properties, and phonon spectral for both Si and Ge.•The effect of PKAs type on displacement damage in Si1-xGex is discussed.•The displacement damage in regular and random Si1-xGex with different Ge concentrations is compared. We developed an accurate and efficient machine learning (ML) potential with Density Functional Theory (DFT) accuracy to investigate the primary radiation damage in Si1-xGex. The accuracy of the potential was verified by comparing the static properties, defect formation energy, defect migration energy, threshold displacement energy and phonon spectra of silicon and germanium with experimental and DFT data. In subsequent a single primary knock-on atom (PKA) simulations, we find that the ML potential produces a larger number of defects than the empirical potential, and that Ge PKA produces more and denser defects than Si PKA at the same energy. Finally, by simulating the displacement cascade process in regular and stochastic Si1-xGex alloys with different Ge concentrations, we found that the defects produced by the cascade show an increasing and then decreasing trend in the regular model and a monotonic decrease in the random model as the Ge concentration increases. The reason for the different trends has been attributed to the fact that the form of Ge distribution in the alloy affects the defect migration energy, which in turn affects the defect recombination rate due to the change in migration energy.