Mathematically Enhanced Photonic Chip Simplifies Data Processing

Researchers from Politecnico di Milano, Scuola Superiore Sant’Anna, the University of Glasgow, and Stanford University have created photonic chips that mathematically calculate the optimal shape of light to best pass through any environment, even one that is unknown or changing over time.

The devices developed in this research are small silicon chips that serve as smart transceivers: Working in pairs, they can automatically and independently calculate what shape a beam of light needs to be in order to pass through a generic environment with maximum efficiency. They can also generate multiple overlapping beams, each with its own shape, and direct them without them interfering with each other. In this way, the transmission capacity is greatly increased.

A photonic chip developed by a group of international researchers is able to calculate the optimal shape for light to pass through any environment, even one that is unknown or changing over time. Courtesy of Politecnico di Milano.

“Our chips are mathematical processors that make calculations with light very quickly and efficiently, almost with no energy consumption. The optical beams are generated through simple algebraic operations, essentially sums and multiplications, performed directly on the light signals and transmitted by micro-antennas directly integrated on the chips,” said Francesco Morichetti, head of the photonic devices lab at Politecnico di Milano.

The technology offers advantages including easy processing, high energy efficiency, and a bandwidth exceeding 5000 GHz, Morichetti said.

The processors consist of meshes of electrically tunable Mach-Zehnder interferometers in silicon photonics. The meshes can configure themselves based on simple power maximization or minimization algorithms, without external calculations or calibration or any prior knowledge of the optical system. The identification of the communication mode channels corresponds to a singular value decomposition of the entire optical system, autonomously performed by the photonic processors.

The researchers observed crosstalk below –30 dB between the optimized channels, even in the presence of distorting masks or partial obstructions.

“Today, all information is digital, but in fact, images, sounds, and all data are inherently analog. Digitization does allow for very complex processing, but as the volume of data increases, these operations become increasingly less sustainable in terms of energy and computation,” said Andrea Melloni, director of Polifab, Politecnico di Milano’s micro- and nanotechnology center. “Today, there is great interest in returning to analog technologies, through dedicated circuits (analog co-processors) that will serve as enablers for the 5G and 6G wireless interconnection systems of the future. Our chips work just like that.”

According to Marc Sorel, professor of electronics at the TeCIP Institute (Telecommunications, Computer Engineering, and Photonics Institute) of Scuola Superiore Sant’Anna, there are numerous application scenarios in which analog computing using optical processors is crucial. Those include mathematical accelerators for neuromorphic systems, high-performance computing and artificial intelligence, quantum computers and cryptography, advanced localization, positioning and sensor systems, and, in general, all systems where processing large amounts of data at very high speed is required.

The research was published in Nature Photonics (www.doi.org/10.1038/s41566-023-01330-w).