Integrated Nonlinear Optics: A Promising Solution for Powerful, Efficient Computing

Computers aren’t getting more powerful quickly enough to keep pace with the complexity of problems they need to solve, but a potential solution lies in nonlinear optics, or using pulses of light to perform computations.

“We’ve kind of hit the end of the road with traditional computing based on silicon electronics,” said Tim McKenna, Principal Scientists at NTT Research, during his talk at the NTT Research Upgrade 2023 event.

While numerous large companies are working on various forms of quantum computing, they all have drawbacks, McKenna said. Superconducting circuits require extreme cooling, for example. Others suffer from “fundamental physics problems,” he said, including scalability, qubit lifetime, susceptibility to noise, and error correction issues.

Researchers in the NTT Research Physics and Informatics Laboratories (PHI) are working on alternative methods of using quantum physics to enable computation based on integrated, non-linear optical circuits. Using pulses of light to perform calculations instead of electrons delivers faster clock rates and improved efficiency. In fact, the faster the clock rate, the more efficient the computers get. “This is basically the opposite of electronics,” McKenna said.

Achieving good compute performance requires three fundamental elements: non-linear processing, memory, and gain. “In electronics, the transistor basically saves the day on all counts,” he said. But NTT Research has made devices that likewise check all the boxes, using lithium niobate instead of silicon.

Lithium niobate is a man-made crystal first synthesized at Bell Labs in 1965. “It has amazing physical properties not found in silicon,” McKenna said.

PHI Lab researchers have found they can apply a voltage across lithium niobate and it will modulate the speed of light of a laser beam traveling through it. “If you hook that into an interferometer, you have a data modulator,” he said. It’s possible to pack photonic circuits onto the material just as with silicon, but with far higher performance, with bandwidths far exceeding 100 gigahertz.

What’s more, it takes little power to operate, giving rise to the possibility of ultra-low power, ultra-high-capacity communication. In his talk, McKenna pointed to an example of a 2 terabit-per-second data stream on a single carrier.

He envisions a day when we’ll see “a photonic accelerator on a chip.”

“I do think they’ll have tremendous benefit as an accelerator, just like a GPU, that you plug into some heterogeneous compute architecture,” he said. “They’re going to excel at tasks in machine learning and linear algebra and optimization problems; some of the world’s biggest problems and the biggest of the big data tasks.”

Tim McKenna enjoys waves and particles of all kinds but focuses on light and microwaves as he probes the limits of on-chip integrated systems to perform tasks in the areas of computing, communications, and sensing. He feels driven to combine breakthroughs in quantum information science with advances in nanofabrication, thereby pushing forward the state of the art of information processing. Mr. McKenna believes the future of quantum technologies relies upon the scalable integration of systems on the surface of a chip.