Wavelength-Selective Bidirectional Photoconductance in a Fully Optically Modulated Organic Heterojunction Synapse for Energy-Efficient Neuromorphic Vision

Abstract Optoelectronic synaptic devices are promising hardware for biomimetic vision systems. Fully optical modulation with wavelength selectivity enables direct mapping from optical stimuli to synaptic responses for in-sensor neuromorphic processing. However, realizing controllable carrier transport for wavelength-dependent bidirectional photoconductance under low-energy-consumption constraints remains challenging. Here, we report a fully optically modulated artificial synapse based on a C8-BTBT/F16CuPc organic heterojunction that exhibits wavelength-selective bidirectional photoconductance. Under 350 nm illumination, the device shows positive photoconductivity (PPC), whereas 578 nm illumination induces negative photoconductivity (NPC), corresponding to synaptic potentiation and depression, respectively. This dual-mode behavior is attributed to spectrally selective excitation and wavelength-dependent interfacial charge redistribution enabled by the type-II heterojunction. The device operates at voltages down to 1 mV and achieves an electrical energy consumption of 220 fJ per synaptic event. By tuning optical pulse parameters, the device successfully emulates representative synaptic functions, including paired-pulse facilitation (PPF), multi-level conductance modulation, and pulse-duration-dependent relaxation dynamics. By incorporating the experimentally derived optical conductance modulation, the device supports both artificial neural network (ANN)- and convolutional neural network (CNN)-based image-recognition tasks, achieving performance close to that of the corresponding ideal synaptic-weight models. This work advances the development of wavelength-encoded, energy-efficient neuromorphic vision systems based on fully optical modulation.