Engineers have developed a compact, rapid, and highly efficient photonic switch that applies quantum mechanics principles, which could enhance various applications such as streaming and AI training by optimizing data centers.
Every second, vast amounts of data—equivalent to the simultaneous downloading of thousands of movies—are transmitted globally as light through fiber-optic cables, akin to cars on a high-speed highway. When this data arrives at data centers, a switching system, much like traffic lights for vehicles, is necessary for orderly processing.
Previously, the photonic switches used to direct optical signals faced a significant challenge—a compromise between size and speed. Larger switches could support greater speeds and more data but at the expense of higher energy use, increased space requirements, and greater costs.
Accelerating the Information Superhighway
A recent study published in Nature Photonics by researchers from the University of Pennsylvania School of Engineering and Applied Science (Penn Engineering) outlines a groundbreaking photonic switch that tackles this size-speed dilemma. Measuring only 85 by 85 micrometers, this new switch is even smaller than a grain of salt.
By efficiently manipulating light on a nanoscale, this innovative switch accelerates the transfer of data onto and off the extensive network of fiber-optic cables worldwide. “This could speed up everything from watching movies to training AI,” states Liang Feng, a professor in Materials Science and Engineering (MSE) and Electrical and Systems Engineering (ESE), and the paper’s senior author.
Quantum Mechanics Meets Light
The new switching mechanism utilizes non-Hermitian physics, a quantum mechanics field that examines the peculiar behaviors of specific systems, providing better control over light behavior. “We can adjust the material’s gain and loss to steer the optical signal to the proper exit on the information highway,” explains Xilin Feng, a doctoral candidate in ESE and the paper’s primary author. Essentially, this distinctive physics allows researchers to manage light flow on the minuscule chip, granting precise control over any light-based network’s connectivity.
This advancement means the switch can redirect signals in trillionths of a second with minimal power usage. “It’s about a billion times quicker than a blink of an eye,” says Shuang Wu, a doctoral student in MSE and co-author of the paper. “Older switches were either compact or quick, but achieving both properties at once has been very challenging.”
Leveraging Silicon for Scalability
This new switch is also noteworthy for its partial use of silicon, a cost-effective and readily available material. “Non-Hermitian switching had never been demonstrated within a silicon photonics framework before,” Wu states. Combining silicon into the switch potentially makes it easier to scale for mass production and widespread industry use. Silicon is a fundamental component in various technologies, from computers to smartphones; utilizing it ensures compatibility with existing silicon photonic factories, which produce advanced chips for devices like graphics processing units (GPUs).
From Idea to Reality
The switch comprises a silicon layer topped with a semiconductor known as Indium Gallium Arsenide Phosphide (InGaAsP), which is particularly adept at manipulating infrared light wavelengths, such as those transmitted through undersea fiber-optic cables.
Combining these two layers was a complex task, involving many trials before achieving a working prototype. “It’s like making a sandwich,” says Xilin Feng, referring to layering the materials. However, if even a tiny misalignment occurs, the whole assembly could fail. “The alignment needs nanometer precision,” Wu adds.
Revolutionizing Data Centers
Ultimately, the researchers believe this new switch will benefit not only academic physicists studying non-Hermitian physics but also companies operating and constructing data centers and the billions of users who depend on them. “Data can only move as fast as we can control it,” Liang Feng remarks. “Our experiments have shown that our system’s speed limit is just 100 picoseconds.”
This research was conducted at the University of Pennsylvania School of Engineering and Applied Science and received support from the Army Research Office (ARO) (W911NF-21-1-0148 and W911NF-22-1-0140), the Office of Naval Research (ONR) (N00014-23-1-2882), and the National Science Foundation (NSF) (ECCS-2023780, DMR-2326698, DMR-2326699, and DMR-2117775).
Other co-authors include Tianwei Wu, Zihe Gao, Haoqi Zhao, and Yichi Zhang from Penn Engineering, as well as Li Ge from the City University of New York.
