Fully Integrated Artificial Olfactory Memristive Neuron for Event-Driven Toxic Gas Alarm

Toxic gas concentration monitoring is essential for avoiding chronic respiratory and cardiovascular diseases from prolonged exposure to toxic gases. Traditional CMOS based toxic gas detection and alarm systems, due to the separated sensing and processing modules, face great challenges in both area and power consumption. Here, we present a compact artificial olfactory neuron, fully integrating the ZnO gas sensor for sensing and the NbO2 memristor for processing. The artificial neuron outputs spike signals only when gas concentration exceeds the threshold, endowed with event-driven features. It could encode analog gas concentration information into spike frequency in real time. By varying source voltages of the neuron, the alarm concentration varies, making it adaptable to different alarm scenes. Experimental results show that the neuron’s static and dynamic power consumption could be as low as 243.84 μW and 327.8 μW, respectively, with a threshold concentration of 20 ppm. These results highlight the potential of the proposed neuron as a compact and energy-efficient solution for intelli-gent olfactory systems.