Interfacial Layer Selection Methodology for Customized Ferroelectric Memories

This study presents a material selection strategy for the interfacial layer (IL) in ferroelectric (FE) memory stacks. The nucleation-limited switching (NLS) model was applied to analyze the switching kinetics of the metal/FE/insulator/metal (MFIM) structure, where Hf0.5Zr0.5O2 (HZO) was used as the FE. Activation field ( <tex-math notation="LaTeX">Eₐ </tex-math> ) and characteristic switching time ( <tex-math notation="LaTeX">τ </tex-math> ) were extracted for various 1-nm-thick ILs, including those of SiO2, La2O3 (LaO), AlN, and Hf3N4 (HfN). The adaptation of HZO/LaO reduced the <tex-math notation="LaTeX">Eₐ </tex-math> by ~44% in relation to that of HZO without an IL (MFM-HZO), resulting in considerably faster switching in the low-electric-field ( <tex-math notation="LaTeX">E </tex-math> ) region (< 4 MV <tex-math notation="LaTeX">· </tex-math> cm <tex-math notation="LaTeX">^-1 </tex-math> )—a highly suitable criterion for applications in 1-bit nonvolatile memories. In contrast, HZO/AlN showed the broadest <tex-math notation="LaTeX">τ </tex-math> distribution due to the large <tex-math notation="LaTeX">Eₐ </tex-math> ( <tex-math notation="LaTeX">~ </tex-math> 200% of MFM-HZO), which led to the stabilization of multiple-intermediate polarization states. Promising potentiation and depression characteristics were obtained for multibit synapse applications when an incremental pulse time scheme with a step size of 10 ns was used.