Unlocking New Frontiers in Scalable Photonic Artificial Neurons: A Study on Gain-Doped Layer-on-Waveguide-Synapse

The exponential growth in computational and energy demands for neural networks has vastly outstripped improvements in electronic hardware, driving an urgent search for more efficient, scalable neuromorphic solutions [1], [2]. Neuromorphic photonics emerges as a compelling alternative, offering intrinsically high bandwidth and parallelism with reduced latency and energy consumption. However, current photonic implementations - whether based on Mach-Zehnder interferometers, microring resonators, or excitable semiconductor lasers - face persistent hurdles, including optical losses, difficulty in achieving energy efficient nonlinear computation, or challenges in compute size respectively [3], [4].