Reliable FeFET-based Neuromorphic Computing through Joint Modeling of Cycle-to-Cycle Variability, Device-to-Device Variability, and Domain Stochasticity

Abstract

Inherent domain stochasticity induced by the random spatial distribution of domains inside the ferroelectric (FE) layer along with Cycle-to-Cycle (CTC) and Device-to-Device (DTD) variability seriously hamper our ability to tune the FeFET threshold voltage ($V_{TH}$) towards the desired analog states required to realize neuromorphic computing. In this work, we present an framework that enables the joint modeling of Domain Stochasticity (DS), CTC, and DTD variability, through coupling TCAD models (accurately capturing the distributed channel of the underlying FET) with a multi-domain model that accurately captures the switching dynamics of the individual domains in the above FE layer. Hence, the hidden interaction between the spatial polarization fluctuation in FE and the distributed channel is unveiled. As a result, accurate modeling of $V_{TH}$ distribution under various write scenarios is obtained. This, in turn, provides designers with guidelines on how analog states can be reliably programmed towards engineering reliable neuromorphic on unreliable FeFET.