Deep learning potential model of displacement damage in hafnium oxide ferroelectric films

Abstract

Abstract A model for studying displacement damage in irradiated HfO2 ferroelectric thin films was developed using deep learning and a repulsive table, combining the accuracy of density functional theory with the efficiency of molecular dynamics. This model accurately predicts the properties of various HfO2 phases, such as PO (Pca21), T (P42/nmc), AO (Pbca), and M (P21/c), and describes the atom collision-separation process during irradiation. The displacement threshold energies for the Hf atoms, three-coordinated O atoms, and four-coordinated O atoms are 57.72, 41.93, and 32.89 eV, respectively. The defect formation probabilities (DFPs) for the O primary knock-on atoms (PKAs) and Hf PKAs increase with energy, reaching 1. Below 80.27 eV, the O PKAs are more likely to form point defects than the Hf PKAs. Above this energy, the Hf PKAs have a higher DFP because the O PKAs form replacement loops more easily, inhibiting the generation of point defects. This study provides a comprehensive understanding of defect formation, which is crucial for increasing the reliability of HfO2 ferroelectric devices under irradiation.