Replacing electrons with light in the physical implementation of artificial neural networks offers hope for higher data rates, lower power consumption, and increased functionality in performing neuromorphic computation. By utilizing small amounts of photons to transmit, detect, and store information, this project aims to develop optoelectronic devices and photonic solutions based on silicon single-photon avalanche diodes (SPADs) for applications as optical synapses and neurons in optical neural networks.

Using the SPAD framework, this project will address the challenges in building optical neural networks: i) For optical transmission, defect distribution in SPADs will be optimized to ensure efficient light emission. ii) Internal photon emission in SPADs with nano-Schottky contacts will be employed for infrared light detection, while photomultiplier configurations will be used to realize optoelectrical neurons. iii) By coupling with detectors or waveguides, photochromic materials will be utilized for low-light information storage as synaptic weights. Optical neural networks will be constructed, demonstrating the developed advancements.

With a SPAD-based platform, this project aims to pave the way for low-light optical neural networks that are both scalable and CMOS compatible, thus catering to the needs of researchers and technological development in the fields of photonic systems and semiconductor technology for the advancement of novel optical computational systems.