First prototype of superconducting neuromorphic network modelled at Lobachevsky University

Nizhny Novgorod physicists have proposed a design of a neuromorphic processor based on superconductors.  The proposed neural network is dozens of times superior to its semiconductor counterparts in terms of energy efficiency and fast performance. The chip can be used for high-speed processing of large data arrays, including broadband space signals. Superconductors made of aluminium or niobium operate at very low temperatures, which makes it possible to integrate superconductor-based neural networks into the equipment of space satellites, telescopes and stations.

"The advantage of our model is that such a neural network can work in both classical and quantum modes. For example, it can be used as a coprocessor for quantum information processing. Development of hybrid quantum-classical neural networks is one of the popular trends in quantum machine learning," explained Marina Bastrakova, the author of the study, Head of the Laboratory for Nanostructure Theory at the UNN Physics and Technology Research Institute.

The topology and design of the neural network proposed by Nizhny Novgorod researchers minimise spurious effects of signal loss due to the smaller number of transformers involved. The authors have also described the interaction of neurons and calculated the optimal parameters of the basic components for further experimental studies.

"We investigated in detail the prototype of a superconducting neuron - the basic element of this neural network - back in 2022. It was recently fabricated at the Institute of Solid State Physics of the Russian Academy of Sciences, and its first experimental characteristics were obtained. Our new research aims at scaling up the developed design. The proposed system can be implemented in hardware in the near future," said Marina Bastrakova.

The project was implemented with grant support from the Russian Science Foundation. The research was carried out by scientists from the Laboratory for Nanostructure Theory at the UNN Physics and Technology Research Institute with the participation of researchers from the Russian Quantum Centre.

The results were published in the international journal Nanomaterials in 2024.