Organic ferroelectric transistors with composite dielectric for efficient neural computing

Abstract

Organic ferroelectric field-effect transistors (Fe-OFETs) exhibit exceptional capabilities in mimicking biological neural systems and represent one of the primary options for flexible artificial synaptic devices. Ferroelectric polymers, such as poly(vinylidene fluoride-trifluoroethylene) (P(VDF-TrFE)), given their strong ferroelectricity and facile solution processing, have emerged as the preferred choices for the ferroelectric dielectric layer of wearable devices. However, the solution processed P(VDF-TrFE) films can lead to high interface roughness, prone to cause excessive gate leakage. Meanwhile, the ferroelectric layer in neural computing and memory applications also faces a trade-off between storage time and energy for read/write operations. This study introduces a composite dielectric layer for Fe-OFETs, fabricated via a solution-based process. Different thicknesses of poly(N-vinylcarbazole) (PVK) are shown to significantly alter the ferroelectric hysteresis window and leakage current. The optimized devices exhibit synaptic plasticity with a transient current of 3.52 mA and a response time of approximately 50 ns. The Fe-OFETs with the composite dielectric were modeled and integrated into convolutional neural networks, achieving a 92.95% accuracy rate. This highlights the composite dielectric's advantage in neuromorphic computing. The introduction of PVK optimizes the interface and balances device performance of Fe-OFETs for neuromorphic computing.