Optical memristors set to make impact in computing and AI

07 Jun 2023

University of Pittsburgh survey indicates how the technology can reach its full potential.

The development of computers with functions modelled on those of the human brain and nervous system, termed neuromorphic computing, has been a long standing goal of researchers.

Architectures to achieve it have employed inventive systems design, such as NIST's development of circuits that behave much like a biological synapse yet use just single photons to transmit and receive signals.

One valuable building block is the memristor, a non-volatile electronic memory in which resistance to current is programmed and subsequently remembered, so the device's behavior remains stable even after an interruption to its power supply.

Optical memristors where this ability is controlled by light should provide synergy with photonics-based applications such as optical computing and telecommunications, and open up potential neuromorphic operations not previously possible.

This could be particularly true in AI applications, as demonstrated by research at the University of Vienna building optical networks with quantum memristors for both classical and quantum tasks, potentially "a missing link between artificial intelligence and quantum computing."

A project including the University of Pittsburgh, University of Oxford, Heidelberg University, and the University of Maryland has now carried out a review of optical memristors, to explore their potential impact in computing and the challenges that still remain. The study and its findings were published in Nature Photonics.

"Researchers are truly captivated by optical memristors because of their incredible potential in high-bandwidth neuromorphic computing, machine learning hardware, and artificial intelligence," commented Nathan Youngblood from the University of Pittsburgh. "Imagine merging the incredible advantages of optics with local information processing. It's like opening the door to a whole new realm of technological possibilities that were previously unimaginable."

Searching for the ideal materials and devices

The survey assessed the benefits of ultrafast, high-bandwidth optical communication with local information processing that neuromorphic computing can deliver, but also confirmed that scalability is the most pressing issue that future research should address.

"Scaling up in-memory or neuromorphic computing in the optical domain is a huge challenge," noted Youngblood. "Having a technology that is fast, compact, and efficient makes scaling more achievable and would represent a huge step forward."

One example quoted involves the challenge of implementing a relatively simple network-on-a-chip architecture using phase change materials, which currently have the highest storage density for optical memory. The result "would take a wafer the size of a laptop to fit all the memory cells needed," according to Youngblood.

“Size matters for photonics, and we need to find a way to improve the storage density, energy efficiency, and programming speed to do useful computing at useful scales."

Dynamic memristors with nonvolatile storage and nonlinear output would be one route to replicating the long-term plasticity of synapses in the brain, and research to scale-up and improve optical memristor technology could unlock unprecedented possibilities for high-bandwidth neuromorphic computing, machine learning hardware, and artificial intelligence, according to the study.

"We looked at a lot of different technologies, and we are still far away from the target of an ideal optical memristor: something that is compact, efficient, fast, and changes the optical properties in a significant manner," commented Nathan Youngblood. "We're still searching for a material or a device that actually meets all these criteria in a single technology, in order for it to drive the field forward."