All-optical leaky integrate-and-fire neuron in a thin-film lithium niobate micro-ring

Optical neuromorphic computing provides a highly efficient information-processing paradigm but entails several physical trade-offs. Laser-based spiking neurons exhibit fast reset capabilities but suffer from low integration density. Meanwhile, micro-ring resonators or phase-change material-based devices show better integration but lack efficient intrinsic reset mechanisms. To address these limitations, an all-optical leaky integrate-and-fire neuron is proposed using a thin-film periodically poled lithium niobate racetrack micro-ring resonator. This architecture exploits cascaded second-order nonlinearities to physically map neuronal dynamics onto a passive device, where second harmonic generation helps with intracavity leaky integration, and optical parametric amplification serves as a threshold-triggering mechanism. Through dispersion engineering and numerical simulations, the waveguide geometry, quasi-phase-matching period, and dichroic coupling parameters are systematically optimized. The device is capable of step-like energy accumulation, sharp spike emission, and ultrafast picosecond-scale reset. Thus, this work establishes a theoretical foundation for integrated optical neuromorphic computing.