We are asking more and more electronics. Due to the developments of artificial intelligence and self-driving cars, the technology is becoming increasingly complex. Electronic devices that must provide computing power consume this power, and are limited in the speed at which they can process information.
Researchers from the Amsterdam Research Institute Amolf, the University of Pennsylvania and the City University of New York think they can ease the strain on electronics by performing certain calculations using light similarly. With light, you can roughly calculate the speed of light. It is more energy efficient than electronics because it does not generate any heat.
In 2019, part of this research group showed that they By shining light through what is called an epitaxial surface Detect edges of objects, such as buildings and people, in an image. This is an important first step in image recognition. Metasurfaces handle energy differently than “normal” surfaces; For example, light is refracted differently. Researchers now present a meta-surface that can mathematically solve more complex computations called matrix inversions. results Noon last week in Nature’s Nanotechnology.
An optical metasurface for image recognition works like this: you polish the image you want to analyze, for example from a self-driving car camera, onto a metasurface consisting of a cleverly designed structure of silicon particles on a transparent surface. This structure is a few thousandths of a millimeter in size. The light is scattered and reflected in that metamaterial in such a way that an image appears on the other side showing only the edges of the objects. With this information, the computer can then decide whether the self-driving car can continue or whether it should stop for a pedestrian. Performing computer edge recognition requires additional steps and more time and energy.
“We have now extended this concept so that we can use it for more complex mathematical calculations,” says Professor Albert Pullman at Amulef over the phone. “We now have a metasurface that allows us to perform matrix inversions. This computational process is often used to solve scientific problems. It also has applications in aircraft control systems, navigation systems, and computer graphics, among others.”
New silicone body
For the matrix reflections, they designed a new silicon structure for the metasurface. “The steps required to solve mathematical equations are built into that architecture,” Pullman says. The light that hits it as input is distributed in all directions by the structure. Since you have to repeat certain steps for this calculation, the design also contains a mirror through which light is bounced back and forth. The whole is designed in such a way that the structure, together with the reflection, produces pure light rays by which the answer to the equation can be read.
Nowadays, every math problem requires its own specially designed fable deck. “This is why we are now developing a structure in which we can change the properties of metamaterial by applying an electric potential to it.”
Replacing the entire computer with optical identification surfaces is not an option at this time. They can only take on specific tasks. Pullman: “I think the app is mainly for tasks that need to be done quickly, on the go, like image processing.”
A version of this article also appeared in the January 19, 2023 Journal