Machine Learning Assisted Compact Modeling of Dynamic Switching and Transient Response in Ferroelectric Capacitor

This work introduces a machine learning-assisted compact model of a ferroelectric capacitor (FeCap), leveraging a time-dependent recurrent neural network (RNN)-based methodology. This RNN-based compact model captures the dynamic switching and transient response of FeCap. The switching behavior of a metal-ferroelectric-metal (MFM) capacitor is validated experimentally for varying voltage ramp rates and amplitudes. A Verilog-A simulation workflow is implemented to demonstrate the compatibility of this model with SPICE circuit simulators. This simulation workflow is applied to a metal-ferroelectric-insulator-semiconductor (MFIS) capacitor in ferroelectric field-effect-transistor (FeFET) technology in a 28 nm foundry platform, highlighting the versatility of the RNN-based model.