Advancement in light-based computing shows capabilities of future smart cameras

Researchers developing the next generation of computing technology aim to shed some light on the field, literally. Optical computing, which relies on particles of light called photons, is expected to provide alternatives to traditional electronic approaches. Such systems, or lightweight components of hybrid systems that also retain electronic parts, could be faster, consume less power, and compute visual information more efficiently through simultaneous and parallel processing.

To date, optical computing has faced a limitation in achieving non-linear responses, which means producing signals that are not directly proportional to the input. Nonlinearity makes universal computing applications possible, including artificial intelligence.

Nonlinear materials and devices under development require a substantial amount of light to function. Previously, this required high-powered lasers that operated only in a narrow band of the electromagnetic spectrum; absorb light over time, which slows down processing; or the use of energy inefficient materials that absorb a lot of light but exclude applications that require light efficiency or transparency.

Now, a recent collaborative study by members of the California NanoSystems Institute at UCLA, or CNSI, has presented a device that overcomes these obstacles.

In an important step towards optical computing for processing visual information, CNSI researchers demonstrated that a small array of transparent pixels could produce a fast, non-linear, broadband response from low-power ambient light. The team also demonstrated an app that combines their device with a smartphone camera to reduce glare in images. The study was published in Nature Communications.

“Optical nonlinearities are far behind what we need for visual computing applications,” said co-author Aydogan Ozcan, Volgenau Professor of Engineering Innovation at UCLA’s Samueli School of Engineering. “We need fast, low-power, wide-bandwidth, and low-loss nonlinearities for optical systems to meet our visual computing needs. This work helps fill that gap.”

The technology’s potential applications, beyond the glare reduction validated in the study, span a variety of industrial and consumer uses: enhanced sensing for autonomous vehicles; cameras that recognize certain objects while hiding others; image encryption; and the efficient and effective detection of defects in robotic assembly lines, among many others.

The device could offer many advantages. For example, incoming images could be processed without conversion to a digital signal, which would speed up results and reduce the amount of data sent to the cloud for digital processing and storage. The researchers plan to link their technology with cheap cameras and compress data to produce images with much higher resolution than has been achieved before, and more precisely and accurately capture useful information about the arrangement of objects in space and their spectra. electromagnetic present in light.

“A cheap device measuring a couple of centimeters could make a low-power camera function as a super-resolution camera,” said Ozcan, a professor of electrical and computer engineering and of bioengineering at UCLA, as well as associate director of the CNSI. “That would democratize access to high-resolution images and sensors.”

The device under study is a transparent plane measuring 1 cm square. It uses a 2D semiconductor material, represented as a film just a few atoms thick, that was developed by co-author Xiangfeng Duan, a professor of chemistry and biochemistry at UCLA.

The thinness of the material makes it transparent, while retaining qualities that allow incoming photons to efficiently regulate electrical conductivity. The research team combined the 2D semiconductor with a liquid crystal layer and made it functional with a series of electrodes. The result is a smart filter comprising 10,000 pixels, each of which is capable of selectively and rapidly darkening in a non-linear manner when exposed to broadband ambient light.

“Basically, we want to use a material that doesn’t absorb a lot of light, but still produces enough signal that can be used to process light,” Duan said. “Each pixel can change from completely transparent to partially transparent and then opaque. It only takes a small number of photons to change transparency dramatically.”

“This unique opportunity led to a very, very interesting collaboration,” Duan said. “It’s really great to think outside of our comfort zones. It showed me that as a materials developer, I can benefit from going beyond a fundamental study or proof of concept to explore applications.

“We hope to continue down this path,” he added. “This is just the beginning. There is certainly much more to do.”

Other co-authors, all affiliated with UCLA, are doctoral students Dong Xu, Yuhang Li, Jingxuan Zhou, Yucheng Zhang, Boxuan Zhou, Peiqi Wang, and Ao Zhang; postdoctoral researchers Yi Luo, Jingtian Hu, Xurong Li and Huaying Ren; Bijie Bai, who earned a PhD in 2023; Mona Jarrahi, professor of electrical engineering at Northrop Grumman; and Yu Huang, professor and chair of materials science and engineering.