Advancements in photonic memory to boost optical computing

Technological advancements like autonomous driving and computer vision have led to a surge in demand for computational power. Optical computing, with its high throughput, energy efficiency and low latency, has garnered attention from academia and industry, but current optical computing chips face limitations in power consumption and size, which hinders the scalability of optical computer networks.

Due to the rise of non-volatile integrated photonics, optical computing devices can achieve in-memory computing while operating with zero static power consumption. Phase-change materials (PCMs) are promising candidates for achieving photonic memory and non-volatile neuromorphic photonic chips. PCMs offer high refractive index contrast between different states and reversible transitions, making them suitable for large-scale non-volatile optical computing chips.

Researchers from Zhejiang University, Westlake University and the Institute of Microelectronics of the Chinese Academy of Science have developed a 5-bit photonic memory capable of fast volatile modulation and have proposed a solution for a non-volatile photonic network supporting rapid training. This was made possible by integrating the low-loss PCM antimonite (Sb₂S₃) into a silicon photonic platform.

The photonic memory utilises the carrier dispersion effect of a PIN diode to achieve volatile modulation with a rapid response time of under 40 nanoseconds, preserving the stored weight information. After training, the photonic memory utilises the PIN diode as a microheater to enable multilevel and reversible phase changes of Sb₂S₃, allowing the storage of trained weights in the photonic computing network. This leads to an energy-efficient photonic computing process.

Using the demonstrated photonic memory and working principle, the researchers simulated an optical convolutional kernel architecture, achieving over 95% accuracy in recognising the MNIST dataset, and showcasing the feasibility of fast training through volatile modulation and weight storage through 5-bit non-volatile modulation. This research establishes a new paradigm for photonic memory and offers a solution for implementing non-volatile devices in fast-training optical neural networks.

The research findings have been published in Advanced Photonics.

Image credit: iStock.com/narong sutinkham