Reconfigurable neuromorphic networks enabled by robust organic memristors with tunable plasticity

Abstract Ensuring long-term robustness and environmental stability remains a critical challenge for organic memristors, despite their potential as low-cost, energy-efficient neuromorphic components. Here, we report a solution-processed Ag/poly(benzimidazobenzophenanthroline) (BBL)/Au organic memristor with excellent electrical performance and dual-mode plasticity. Benefiting from the chemical and structural stability of the BBL, the devices exhibited high endurance (>10 4 cycles) and over 100 linearly tunable conductance states. Importantly, the device maintained stable performance between 300 K to 413 K and after long-term aqueous exposure, demonstrating outstanding thermal and environmental robustness. Furthermore, short-term plasticity (STP) and long-term plasticity (LTP) can be independently modulated via pulse schemes, enabling dynamic tuning of synaptic behaviors within a unified architecture. These functions support both reservoir computing (RC) and binarized spiking neural networks (BSNNs). The STP-enabled RC system achieves 91.7 % accuracy on MNIST with low training cost, whereas the resource-efficient BSNN implementation attains full-precision-like accuracy using LTP-based binary memristive states. These results suggest that BBL-based organic memristors are scalable, reconfigurable, and reliable candidates for edge-oriented, low-power neuromorphic computing.