Thermally Activated Negative Differential Resistance VOx Memristor with Switchable Rate and Leaky Integrate-and-Fire Spiking Dynamics

Abstract

Spiking neural networks (SNNs) require neuron devices that are both compact and capable of supporting continuous-time and event-based dynamics. Here, we demonstrate a VO_x-based threshold switching memristor (TSM) that intrinsically enables dual-mode operation, functioning as both a spiking encoder and a leaky integrate-and-fire (LIF) neuron. While such dual behavior is theoretically possible in volatile memristors, it has only been experimentally demonstrated in limited cases. It is achieved intrinsically in a single VO_x-based device, arising from a thermally driven insulator-to-metal transition (IMT) within the VO_x layer, confirmed by temperature-dependent XRD and correlated with snap-back negative differential resistance (NDR) observed in electrical measurements. Integrated into a passive circuit, the device generates high-frequency spike trains under analog input and tunable LIF responses under pulsed stimulation. The device achieves a maximum spiking frequency of 570 kHz, a time-to-first-spike (TTFS) of 1.6 μs, and energy consumption as low as 4.7 nJ per spike. These results highlight the strong coupling between structural phase dynamics and neuronal functions, positioning the VO_x TSM as a promising platform for scalable neuromorphic hardware with biologically realistic spiking capabilities.