Planar p-n Junction Engineering toward Reconfigurable Organic Synaptic Transistors for High-Accuracy Neuromorphic Recognition

Synaptic transistors are pivotal for hardware-level neuromorphic computing. However, the lack of switching behavior diversity has limited the implementation of advanced computing tasks, which are constrained by traditional interfacial or uncontrollable materials engineering. Here, a universal planar p-n junction structure is devised, with rational alignment of energy levels between crosslinkable p-type poly(indacenodithiophene-alt-benzothiadizole)-based conjugated polymer with hydroxyl groups at the ends of its side chains (OH-IDTBT-10%) and different n-type conjugated polymers, fabricated through efficient solution processing. This structure enables reconfigurable switching of p-type and n-type carrier transport by modifying the transistor architecture, along with significant non-volatile memory and synaptic plasticity. By strategically adjusting crosslinkers, a large memory window up to 48.5 V is achieved, sustained performance over 500 cycles, and a diverse array of synaptic behaviors modulated by electrical pulses. The underlying mechanism involves quantum well-like structures and discrete physical charge traps at the bilayer interface. The versatility of the strategy is proven across different n-type polymer systems. An artificial neural network (ANN) constructed by these devices affords a remarkably high facial recognition accuracy of 97.58% using the Yale Face Database with minimized training epochs of 200. This design provides an opportunity for high performance hardware with diverse synaptic behaviors in advanced neuromorphic computing.