The latest issue of Laser&Photonics Reviews contains an article co-authored by Jacek Gościniak of CoE ENSEMBLE³ . The paper was co-authored with Prof. Jacob Khurgin of Johns Hopkins University and Prof. Volker Sorger of George Washington University, where he heads the Institute for Photonics and Artificial Intelligence. At the same time, Prof. Sorger is the founder of the startup Optolligence, which aims to apply photonics to intelligent systems.

In the paper, the researchers propose a new type of nonlinear and bistable devices for the next generation of optical networks and photonic neural systems, which could find numerous applications in optical logic systems and information processing, for example. They are based on nonlinear, so-called epsilon-near-zero materials, which include transparent conducting oxides that have the added advantage of being both inexpensive to manufacture and easy to integrate with available silicon-based technology.

As a result, the switching speed between bistable states can be reduced to a few picoseconds or even less, with negligible power consumption. The device can be integrated on a chip and act as the analogue equivalent of a memristor or thyristor, enabling the construction of photonic neural networks modelled on the human brain. The device does not require signal conversion from the electrical to the optical domain, which means it can operate solely on the basis of a photonic system. It can, but it does not have to. It can work in two domains, optical and electrical, depending on the need.

Switching between bistable states results in either high signal transmission or low signal transmission, simply by adjusting the intensity of the light directed at the device.

This is the first stage of work, and further work will further reduce power consumption.

The article is available at this link: https://doi.org/10.1002/lpor.202200723