Eighty years after the creation of ENIAC, the world’s first general-purpose electronic computer, researchers at the University of Pennsylvania are exploring new ways to drive the future of computing. Rather than relying entirely on electrons, which have formed the backbone of computers since the 1940s, scientists are turning to light.
ENIAC was developed by Pennsylvania researchers J. Presper Eckert and John Mauchly and helped usher in the modern era of computing by using the flow of electrons to solve complex mathematical problems. The same electronic approach still powers today’s computers, smartphones, and AI systems. However, as the demands for artificial intelligence increase, it becomes harder to ignore the limitations of electronic-based hardware.
Why electrons are reaching their limit
Electrons carry an electrical charge, which creates several challenges within modern computer chips. As it moves through matter, it generates heat and encounters resistance that wastes energy. These problems become even more difficult as chips become more complex and process vast amounts of data for AI applications.
Researchers led by physicist Bo Zheng of the University of Pennsylvania’s College of Arts and Sciences think photons, the particles that make up light, may be able to solve some of these problems.
“Because photons are charge-neutral and have zero rest mass, they can rapidly carry information over long distances with minimal loss, making them dominant in communications technology,” explains Li He, co-lead author of the 2006 paper. physical review letter Former postdoctoral researcher at Zhen Lab. “But their neutrality means they interact very little with their environment, making them poor at the kind of signal-switching logic that computers rely on.”
In other words, light is great at transporting information quickly and efficiently, but it struggles with the switching operations required for computing.
Combining light and matter for AI computing
To overcome this problem, Zhen’s team developed a special quasiparticle called an exciton-polariton. Particles are formed when photons combine strongly with electrons within atomically thin semiconductor materials. This combination allows light to interact more effectively and perform the signal switching required for computing tasks.
This breakthrough could be especially important for artificial intelligence systems, which consume huge amounts of power.
Many experimental photonic AI chips already use light to speed up certain calculations. However, if these systems need to perform nonlinear startup steps, such as decision-making operations, the optical signals typically need to be converted back to electronic signals. That conversion slows down the process, increases energy usage, and reduces the benefits of photonic computing.
Researchers at Penn State have demonstrated all-optical switching using exciton polaritons while using only about 4 quadrillion joules of energy. That amount is very small, far less than the energy needed to power a small LED light for a short period of time.
Aiming for faster and more efficient AI chips
If this technology can be successfully scaled up, it could lead to photonic chips that can directly process information from cameras without having to repeatedly convert between light and electricity. This approach could reduce the huge energy demands of large-scale AI systems and also support basic quantum computing capabilities in future chips.
Bo Zhen is the Jin K. Lee Presidential Associate Professor in the Department of Physics and Astronomy in the College of Arts and Sciences at the University of Pennsylvania.
Li He was a postdoctoral researcher in the Zhen Lab at Penn State College of Arts and Sciences. He is currently an assistant professor at Montana State University.
Additional authors on the study include Zhi Wang and Bumho Kim from the University of Pennsylvania’s College of Arts and Sciences.
This research was supported by the U.S. Office of Naval Research (N00014-20-1-2325 and N00014-21-1-2703) and the Sloan Foundation.
