Microsize Lens Pushes Photonics Closer to an On-Chip Future

University of Delaware researchers have developed a “metasurface” atop a silicon chip that behaves as a lens for possible use in photonics chips

2 min read

Illustration of photonics chip
On-chip microsized lens from University of Delaware researchers uses gratings in a dielectric material on top of a silicon chip to guide and then focus light. A color illustration is superposed atop the photo to show simulations of how light responds to this "lens."
Illustration: Tingyi Gu/University of Delaware

Optical microcomputing, next-generation compact LiDAR units, and on-chip spectrometers all took a step closer to reality with the recent announcement of a new kind of optical lens.

The lens is not made of glass or plastic, however. Rather, this low-loss, on-chip lens is made from thin layers of specialized materials on top of a silicon wafer. These “metasurfaces” have shown much promise in recent years as a kind of new, microscale medium for containing, transmitting, and manipulating light.

Photonics at the macro-scale is more than 50 years old and has applications today in fields including telecommunications, medicine, aviation, and agriculture. However, shrinking all the elements of traditional photonics down to microscale—to match the density of signals and processing operations inside a traditional microchip—requires entirely new optical methods and materials.

A team of researchers at the University of Delaware, including Tingyi Gu, an assistant professor of electrical and computer engineering, recently published a paper in the journal Nature Communications that describes their effort to build a lens from a thin metasurface material on top of a silicon wafer.

Gu says that metasurfaces have typically been made from thin metal films with nanosized structures in it. These “plasmonic” metasurfaces offered the promise of, as a Nature Photonics paper from 2017 put it, “Ultrathin, versatile, integrated optical devices and high-speed optical information processing.”

The problem, Gu says, is that these “plasmonic” materials are not exactly transparent like windowpanes. Traveling just fractions of a micrometer can introduce signal loss of a few decibels to tens of dB.

“This makes it less practical for optical communications and signal processing,” she says.

Her group uses an alternate kind of metasurface made from etched dielectric materials atop silicon wafers. Making optical components out of dielectric metasurfaces, she says, could sidestep the signal loss problem. Her group’s paper notes that their lens introduces a signal loss of less than one dB.

Even a small improvement (and going from handfuls of dB down to fractions of a dB is more than small) would make a big difference, because a real-world photonics chip might one day have many such components in it. And the more lossy the photonics chip, the greater the amount of laser power needed to be pumped through the chip. More power means more heat and noise, which might ultimately limit the extent to which the chip could be miniaturized. But with her team’s dielectric metasurface lens, “We can make a device much smaller and more compact,” she says.

Her group's lens is made from a configuration of gratings etched in the metasurface — following a wavy pattern of vertical lines that looks a bit like the Cisco company logo. Gu’s group was able to achieve some of the familiar properties of lenses, including converging beams with a measurable focal length (8 micrometers) and object and image distance (44 and 10.1 µm).

The group further used the device's lensing properties to achieve a kind of optical signal Fourier Transform—which is also a property of classical, macroscopic lenses.

Gu says that next steps for their device include exploring new materials and to work toward a platform for on-chip signal processing.

“We’re trying to see if we can come up with good designs to do tasks as complicated as what traditional electronic circuits can do,” she says. “These devices have the advantage that they can process signals at the speed of light. It doesn’t need logic signals going back and forth between transistors. … It’s going to be fast.”

The Conversation (0)