Photons in optical circuits replicate brain-like memory behavior

Yesterday 08:50

By: Dakir Madiha

Photons in optical circuits replicate brain-like memory behavior

A team of Italian scientists has shown that photons traveling through optical circuits can spontaneously behave like neural networks, imitating how the human brain stores and retrieves memories. The study, published in Physical Review Letters on February 18, demonstrates that identical photons can reproduce the characteristics of a Hopfield Network, a model central to understanding associative memory in both neuroscience and machine learning.

The research was carried out by teams from the Institute of Nanotechnology of Italy’s National Research Council (CNR-Nanotec), the Italian Institute of Technology (IIT), the University of Salento, and Sapienza University of Rome. Using quantum interference within photonic chips, the researchers found that photons themselves function as the active components of information processing, effectively acting as neurons. According to Marco Leonetti, senior researcher at CNR-Nanotec and IIT, “photons are no longer just carriers of data; they become the memory elements themselves.”

The experiment involved identical photons distributed across multiple optical modes, guided by phase-shifters and interferometers. This structure produced output patterns described by a generalized Hopfield Hamiltonian, revealing that the system has a finite memory capacity similar to that of biological brains. When limited data were stored, the quantum coherence of photons allowed precise retrieval. As the memory load increased, the system transitioned into a disordered, spin-glass phase, losing its ability to recall stored information.

The discovery bridges two Nobel Prize domains. John Hopfield received the 2024 Nobel Prize in Physics for theoretical work underpinning modern machine learning, while Giorgio Parisi earned the 2021 award for his research on disordered systems. Both intellectual lineages intersected at Sapienza University, where this new quantum study took place.

Co-author Luca Leuzzi, of CNR-Nanotec and Sapienza, noted that such light-based systems could dramatically reduce the energy demands of artificial intelligence. With global data centers projected to consume over 500 terawatt-hours by 2026, optical computing could offer a sustainable alternative to energy-intensive processors. Fabrizio Illuminati, director of CNR-Nanotec, added that “light becomes a miniature laboratory for exploring the same complex behaviors seen in natural and artificial systems, from climate dynamics to biological networks.”