Monolithically Integrated VO$_2$ Mott Oscillators for Energy-Efficient Spiking Neurons

Abstract

Brain-inspired non-Boolean computing and sensing enable energy-efficient, error-tolerant, and highly parallel information processing, yet their deployment remains limited by the lack of compact, scalable spiking hardware. Mott phase-transition materials offer a promising route because their abrupt insulator-to-metal transitions enable neuron-like thresholding and oscillations. Among them, vanadium dioxide (VO$_2$) is particularly attractive owing to its near-room-temperature transition, fast switching, and scalability. However, existing VO$_2$ neuristors rely on discrete components, limiting integration density. Here, we report monolithic back-end-of-the-line (BEOL) integration of one-transistor-one-VO$_2$-memristor (1T-1MR) spiking neurons on a CMOS-compatible platform. VO$_2$ nanosheets are fabricated by pulsed-laser deposition atop dielectrically isolated silicon-on-insulator (SOI) p-type junctionless field-effect transistors (JLFETs) below 430 $^\circ$C. The architecture exhibits gate-tunable oscillations from 40 to 410 kHz in 60 nm-thin VO$_2$ devices with a 6 $\mu$m$^2$ active area, achieving 18 pJ per spike and 8 $\mu$W at room temperature, with potential for sub-3 $\mu$W operation. We uncover a non-monotonic dependence of oscillation frequency on bias current and temperature and analyze bias-dependent stochastic firing, revealing the nonlinear physics of integrated VO$_2$ thin-film memristors. Finally, we demonstrate voltage-controlled oscillator functionality and on-chip resistive coupling between two nano-oscillators mediated by a JLFET. These results establish a pathway toward dense, energy-efficient, monolithically integrated Mott neuromorphic hardware compatible with future computing and spiking sensing systems.