Low Off-State Current Density in Vertical Organic Field-Effect Transistors via AgNWs/TiO₂ Interface Modification for Neuromorphic Devices

Abstract

With the rapid advancement of artificial intelligence and the Internet of Things, neuromorphic computing, which mimics the efficient and low-power information processing of biological brains, has emerged as a pivotal direction for transcending conventional computing paradigms. Notably, high on/off current ratio in OFET-based neuromorphic devices provides a broader, more precise synaptic weight dynamic range and distinguishable conductance states, enabling fine weight modulation and improved complex pattern recognition accuracy. The vertical organic field-effect transistor (VOFET) is regarded as a promising device architecture for building high-density, lowpower neuromorphic systems, owing to its inherent nanoscale vertical channel.However, in VOFETs, the three-dimensional interface and the ultrashort channel formed between the mesh source electrode and the semiconductor layer often lead to high off-state current density and poor output saturation characteristics, which fundamentally limit device performance and neuromorphic application. Therefore, an interface optimization strategy by spray-depositing a titanium dioxide (TiO₂) modification layer onto the surface of silver nanowires (AgNWs) mesh source electrode is proposed to successfully fabricate VOFET devices with low off-state current density(5.0 × 10⁻⁷ mA cm⁻²), high on-off ratio(~2.8 × 10⁶) and well-saturated output characteristics. Furthermore, an optoelectronic neuromorphic device with an AgNWs/TiO₂ & PDVT-10/TiO₂ heterojunction structure is constructed. The device successfully emulates diverse synaptic behaviors and achieves an ultralow power consumption of 0.235 fJ, yielding a recognition accuracy of 93.64% for banana ripeness 3 classification based on a convolutional neural network. The VOFET interface optimization strategy provides a practical and feasible route toward high-performance neuromorphic devices, showing great promise for advancing the development of ultralow-power intelligent computing systems.