A Capacitorless Neuron With Volatile Analog Pt/C/NbOx/TiN Memristor

Artificial neurons, as key components for implementing spiking neural networks (SNNs), are essential for neuromorphic systems to achieve efficient decision-making. Yet, conventional implementations require bulky capacitors or reset circuits that limit integration density and reliability. In this work, a capacitorless artificial neuron based on volatile analogue Pt/C/NbO ${}_{\boldsymbol {x}}$ /TiN memristor is demonstrated, featuring highly uniform characteristics, excellent linear response, and fast relaxation process. The device inherently enables basic integration and leakage functions without additional capacitors. The neuron parameters can be flexibly adjusted by finely modulating input pulses to support different application scenarios. Furthermore, we constructed a two-layer SNN using time-to-first-spike (TTFS) coding, achieving about 98.9% accuracy on the MNIST dataset while reducing the average spike count by 93.8% compared to rate coding. The proposed scheme demonstrates great potential for achieving efficient and flexible neuromorphic systems.