Strain–Stress\nImpact on Ferroelectric Devices:\nA Multilayer Analysis and Optimization Strategy for Neural Networks

We investigate the effects of strain stress on the hafnium\nzirconium\noxide ferroelectric tunnel junctions (FTJs). The impact of strain\nstress on each layer of the FTJ is investigated depending on the thickness\nof the metal capping layer and the post-metal-annealing temperature.\nIt is revealed that, for the insulator layer, an increase in strain\nlead to an increased off-current in the FTJs. In contrast, increased\nstrain stress in the ferroelectric layer directly increases the trap\ndensity, leading to an increase in the on-current of the FTJs. Furthermore,\nwe analyze how these strain-induced changes affect the performance\nand reliability of FTJs in neuromorphic systems. We propose optimization\nstrategies for strain stress in FTJs based on the frequency of neural\nnetwork updates, highlighting the critical balance between achieving\na large dynamic range and ensuring device endurance, aligning device\nperformance with the specific demands and conditions of neural network\napplications.