Interface percolation and random trap generation in ferroelectric memory: A two-step degradation mechanism explored through low-frequency noise spectroscopy

Abstract

Ferroelectric (FE) HfO₂-based devices are promising for next-generation memory and logic applications. However, their long-term reliability under repeated field-driven switching remains a major challenge. In this study, we show that the degradation of a metal-ferroelectric-insulator-semiconductor (MFIS) stack occurs via a two-step mechanism involving an initial soft breakdown in the dielectric (DE) layer followed by the accumulation of randomly distributed bulk traps in the FE layer. We investigate the interactions between layers using low-frequency noise (LFN) spectroscopy combined with multi-frequency conductance measurements. As a non-destructive technique, LFN spectroscopy analyzes the stochastic current response in the frequency domain to monitor cycling-induced evolution of trap densities at the FE/DE interface—offering insights that conventional electrical tests cannot capture. The formation of localized percolation paths in the DE (first step) leads to a soft breakdown that increases the electric field in the FE layer, thereby accelerating trap generation (second step) and degrading ferroelectric properties. By comparing MFIS stacks with different DE layers (SiO₂ and Al₂O₃), we demonstrate that this two-step degradation mechanism is generally applicable. These results provide new insights into the relationship between defect fluctuations, interface conduction, and ferroelectric switching, and suggest possible approaches for improving the reliability of FE-based devices.