A Monolithic Artificial Olfactory Neuron With Gas‐pulse‐driven Physical Reservoir Computing for Intelligent Gas Monitoring
Yi Lu, Ziyu Liu, Yueyue Li, Yilin Wang, Mingxue Zhang, Yueying Liu, Xishuang Liang, Geyu Lu, Fengmin LiuABSTRACT
Existing artificial olfactory neurons (AONs) largely rely on external pulse‐encoding modules to interconnect gas sensors and memristors, which results in significant hardware redundancy and obscures the inherent nonlinear dynamics of gas‐surface interactions. To address this, we developed a monolithic AON that integrates a low‐power PdCu‐SnO 2 hydrogen sensor with an Au/TiO 2 nanowires/FTO memristor on a single Al 2 O 3 substrate. Leveraging their dynamic impedance matching and physical coupling, it directly converts gas concentration pulses into synaptic plasticity. Without external encoding, it simulates synaptic characteristics such as short‐term potentiation (STP), long‐term potentiation (LTP) and paired‐pulse facilitation (PPF), thereby precisely mapping complex gas pulse timing features to a high‐dimensional state space. As independent units in a physical reservoir computing network, the AON achieves 92.9% H 2 classification accuracy and high‐precision excitatory postsynaptic current peak fitting with R 2 = 0.99. This work proposes a universal monolithic integration scheme for low‐power brain‐inspired olfactory edge devices, offering a novel path for intelligent gas safety monitoring.