Performance variability in ferroelectric tunnel junctions: Impact of intrinsic fluctuations and spatial phase randomness

Abstract

Ferroelectric tunnel junctions (FTJs) based on hafnium oxide are promising candidates for low-power nonvolatile memory and neuromorphic computing. However, the device-to-device variability caused by the intrinsic material stochasticity of ultrathin polycrystalline ferroelectric films remains insufficiently understood. This work employs multi-grain TCAD simulations to systematically quantify the influence of intrinsic fluctuation sources (including the spontaneous and residual polarization charge, the coercive field) and spatial phase randomness in hafnium-based metal–ferroelectric–semiconductor FTJs. The analysis reveals that polarization charge is the dominant contributor to variability due to its direct modulation of the tunneling barrier. An incomplete polarization results in a higher FTJ variability, whereas saturation polarization mitigates device variation. Especially when the Si layer is in the depletion region, the FTJ variability becomes more pronounced. On the other hand, an increase in the DE phase content due to the spatial ferroelectric–dielectric phase distribution reduces FTJ variability and narrows the memory window. Furthermore, the influence of device scaling, uniformity, and key design parameters on FTJ fluctuations is evaluated. These findings establish a mechanistic understanding of performance fluctuations in HfO2-based FTJs and provide guidelines for variability-aware device design and process optimization.