Modeling Ferroelectric Thin-Film Transistor toward Neuromorphic Computing Application

Ferroelectric thin-film transistors (Fe-TFTs) are emerging as critical components for flexible electronics, non-volatile memory, and neuromorphic computing. In this work, we present a physics-based model for Fe-TFTs that captures both memory switching behavior and I-V characteristics through a stochastic multidomain approach and virtual source modeling. This dual approach allows for precise modeling of the complex electrostatic interactions between the ferroelectric layer and the semiconductor channel, accounting for both the polarization memory effects and carrier transport. By reducing the number of required parameters while maintaining high accuracy, our model provides a streamlined yet comprehensive framework for exploring Fe-TFT behavior. Our model also reveals the dependence of the memory window on applied voltage pulse amplitude and duration, offering a quantitative framework for programming Fe-TFTs. Additionally, we introduce a machine-learning method for generating optimized voltage pulses to achieve precise multi-level programming of intermediate ferroelectric states. This is essential for multi-bit memory storage and neuromorphic operations. To demonstrate the model’s versatility, we simulate a 4 4 Fe-TFT crossbar array circuit, achieving multiply-accumulate (MAC) operations with a performance of ∼1.28 tera operations per second (TOPS) and a power efficiency of ∼0.43 W/PetaOPS. This model is a valuable tool for advancing Fe-TFT-based flexible memory and computing technologies.