New Light-Powered Chip Handles Generate, Steer and Read All on One Device: A Step Toward Photonic AI

[ProperMadlad20]

Scientists published results in June describing a tiny chip that can generate, steer and read light-based information all within a single device, using atomically thin two-dimensional materials and nanoscale engineering to combine functions that have previously required separate components. The achievement is described as a major step toward ultra-fast, energy-efficient computing that uses photons rather than electrons for information processing, an approach that sidesteps the energy and heat constraints that are increasingly limiting further scaling of conventional electronic computing.

Current photonic computing systems are typically built from separate components: a light source, a set of optical waveguides or modulators that steer and process the light, and a photodetector that converts the optical signal back to an electrical one for output. Each interface between components introduces loss and latency, and the physical separation means the devices cannot be as small or as integrated as modern electronic chips. Combining all three functions, generation, steering and detection, in a single device using the same material system eliminates those interface penalties and opens the path to photonic integrated circuits that approach the density of electronic ones.

The use of atomically thin materials, likely transition metal dichalcogenides given the research group's background, is the enabling technology. These materials have optical and electronic properties that vary dramatically with the number of atomic layers and with applied electric fields, allowing a single material to function as emitter, modulator and detector depending on how it is configured locally. The result is more analogous to how electronic chips use a single silicon process to create transistors, capacitors and interconnects from the same material than to traditional optics where different materials optimised for different functions are assembled together.

Quantum Computing Newswww.sciencedaily.com