When we define a “computer” as any device that processes information through input and output, the question arises as to what objects can do these calculations and how small these computers can be. As transistors reach the miniaturization limit, answering these questions becomes crucial as they could lead to the development of a new computing paradigm.

In a new study published in the journal EPJ plus By researchers from Tulane University in New Orleans, Louisiana, Gerard McCall and his team show that atoms, one of the simplest building blocks of matter, can act as a reservoir for computation, where all input and output processes are optical.

“We have the idea that computability is a universal property that all physical systems share, but there’s a huge amount of structure within this paradigm to how you can actually perform computations,” McCall says.

He adds that one of the most important of these constructs is neuromorphic, or reservoir computing, using a neuromorphic computer that aims to simulate the brain. This concept underlies the explosive development of machine learning and artificial intelligence over the past few decades, leading to a potentially non-linear computer where output is not linearly proportional to input. This is desirable because it could lead to a computational architecture flexible enough to achieve any output with a suitable input.

“So if we want to achieve a certain computational result, we are guaranteed to have certain computational inputs that will enable us to achieve it,” McCall says. “This is impossible if our system only exhibits a linear response!”

The team proposed a nonlinear monoatomic computer with input directly coded into light and output in the form of light. The calculation is then determined by the filters through which the light passes.

“Our work confirmed that this approach works in principle, and also confirmed that the system performs better when the input light is designed to result in a higher degree of non-linearity in the system,” McCall says. “I would argue that what we’re probably trying to do with this work is to emphasize that there is a minimal system capable of computation at a single atomic level, and that the computation can be done purely by optical processes.”