Photonic Computing with Photorefractive GaAs PhC Nanobeam: Preliminary Analysis

Conventional electronics based neural networks have recently made rapid advancements, but remains limited by large energy consumption. To address these issues, neuromorphic photonics rises as a new research fields that promises high computational density at lower energy consumption. In this paper we propose to execute the synaptic operation in photonic computing exploiting the photorefractive effect in GaAs Photonic Crystal (PhC) nanobeams, where for the first time the photorefractive effect in a GaAs PhC is simulated. The combination of photorefractive GaAs with 1D PhC nanobeams is utilized to detune the resonant wavelength by changing the refractive index via the self-interference pattern of the optical mode, and the redistribution of electrons and ionized donors in the GaAs PhC nanobeams, resulting in a built-in space charge field. The detuning of the resonant wavelengths would represent the synaptic weights in a neural network. We demonstrate the simulation via a FEM simulation using COMSOL, with a simplified photorefractive model where acceptors are removed and implemented. The polarizations of the optical mode suggest that a <111> orientation must be used instead of <001>, because the photorefractive space charge field and polarizations are parallel. Additionally, the drift and diffusion processes are shown to separate electrons and ionized donors.