Analog computers have their roots in antiquity. They passed the peak of their development from the 40s to the 60s of the XX century, after which they gave way to the "figure". Today, analog computing is once again beginning to challenge digital solutions in neuromorphic, quantum, and other applications and, as it turns out, can be much more productive in the process of extremely complex mathematical calculations.

So, in the journal Nature Nanotechnology, an article was published by researchers from the AMOLF Institute, the University of Pennsylvania and the City University of New York (CUNY), in which they talked about creating a nanostructured surface that can solve equations using light.

In fact, we are talking about the creation of optical metasurfaces, the incidence of light on which (and its reflection) occurs according to the usual laws of optics according to a physically prewired function. This method cannot be put on a par with digital systems for solving any problems, but for the implementation of specific and mathematically complex algorithms, everything is very, very promising.

Digital architectures in the form of processors no longer promise explosive performance growth and the same increase in energy efficiency, which is often formulated in such a laconic form as "Moore's law is dead." The transition to optical signals and structured metasurfaces solves both problems: it will increase the speed of calculations due to the work of light in nanoscale optical structures and reduce power consumption due to the absence of heat generation during calculations.

The idea of such calculations is simple. In the thinnest transparent layer of matter, transparent areas and a certain nanoscale relief under them are created. Simply passing through the layer and reflecting from it, the light performs the calculation according to the given formula. Researchers have shown that images can be processed in this way, for example. And if conventional computers spend many cycles on such operations, especially if matrix processing is included in the process, then optical systems perform calculations in one cycle.

One of the lead authors of the paper said: “We have demonstrated a powerful new alliance between nanotechnology and analog computing that could pave the way for hybrid optical and electronic computing circuits. Further development of our ideas will lead to the solution of problems of increased complexity with speed and efficiency that were previously unthinkable.”