Reconfigurable In-Sensor Computing Memristor for Olfactory SNN and Reservoir Hybrid Neuromorphic Computing

Traditional gas sensing systems are facing efficiency challenges due to physically separated von Neumann architectures, making the construction of in-sensor computing neuromorphic olfactory systems urgently needed for low-power and low-latency scenarios. In this study, a reconfigurable neuromorphic heterostructure memristor based on MXene@SnS₂@PANI and an in-sensor computing olfactory system were proposed. Notably, the reconfigurable neuromorphic olfactory electronics differ fundamentally from conventional sensors. Specifically, the memristor's circuit architecture supports both synaptic and neuronal computational functions, enabling reconfigurable responses to both electrical and gas stimuli within a single device, which substantially minimizes circuit complexity. Through modulation of the energy band under both gas and electrical signals, the device achieves reconfigurable neuromorphic computing features supporting both volatile and nonvolatile conductance updates. Under electrical stimulation, it demonstrates integrate-and-fire neuronal dynamics for gas flow recognition via a spiking neural network. Under gas exposure, neuromorphic synaptic behaviors are realized, enabling gas concentration identification through reservoir computing. The system has been successfully implemented for real-time hazardous gas monitoring and automated ventilation control, paving the way for next-generation neuromorphic intelligent sensing systems.