Exploring Gain-Doped-Waveguide-Synapse for Neuromorphic Applications: A Pulsed Pump-Signal Approach

TL;DR

This paper introduces a novel gain-doped waveguide synapse using pulsed pump signals to improve neuromorphic computing by enabling scalable, energy-efficient, and bio-inspired neuron-like responses.

Contribution

It pioneers the use of gain on waveguide dynamics with pulsed signals for neuromorphic synapses, addressing scalability and energy efficiency challenges in AI hardware.

Findings

Pulse amplitude and period influence spiking responses.

Material properties affect neuronal emulation.

Event-driven responses mimic natural neurons.

Abstract

Neuromorphic computing promises to transform AI systems by enabling them to perceive, respond to, and adapt swiftly and accurately to dynamic data and user interactions. However, traditional silicon-based and hybrid electronic technologies for artificial neurons constrain neuromorphic processors in terms of flexibility, scalability, and energy efficiency. In this study, we pioneer the use of Doped-Gain-Layer-on-Waveguide-Synapses for bio-inspired neurons, utilizing a pulsed pump-signal mechanism to enhance neuromorphic computation. This approach addresses critical challenges in scalability and energy efficiency inherent in current technologies. We introduce the concept of Gain on Waveguide Dynamics for synapses, demonstrating how non-linear pulse transformations of input probe signals occur under various pump-probe configurations. Our findings reveal that primarily properties of pulse amplitude, period as well material properties such as doping densities and population dynamics influence strongly the generation of spiking responses that emulate neuronal behaviour and effectively how computational logic is. By harnessing the complex interactions of asynchronous spiking pump techniques and ion densities in excited states, our method produces event-driven responses that mirror natural neuronal functions. This gain-enhanced environment supports short-term memory capabilities alongside essential characteristics like asynchronous spike generation, threshold operation, and temporal integration, foundational to brain-inspired spiking neural network paradigms.