Two photonic memory devices boost light-based computing

Researchers in China and Europe have built a photonic integrated circuit with two coupled quantum memristors that shows stronger nonlinearity and memory than a single device. The system, reported online July 30 in Opto-Electronic Advances, was tested on temporal series prediction and could help speed up low-energy photonic computing.

Why it matters: - The circuit points to a new way to process information with light instead of standard von Neumann computing. - Stronger memory and nonlinearity in photonic hardware could improve temporal series prediction, signal processing and control. - The approach could support lower-energy and lower-latency systems for tasks used in finance, weather forecasting, industrial monitoring, healthcare and robotics.

What happened: - A research team led by Lorenzo Pavesi at the University of Trento and Christophe Couteau at the University of Troyes proposed a new configuration for photonic quantum memristors. - The study was published online July 30, 2026, in Opto-Electronic Advances. - The team built a photonic integrated circuit with two coupled photonic quantum memristors and crossed feedback. - The system used a room-temperature single-photon source based on a silicon-vacancy color center in a nanodiamond. - The setup was used experimentally for temporal series prediction.

The details: - Photonic quantum memristors control photon flux by combining a linear interferometer with an electrical circuit. - The interferometer splits light into two paths and recombines it, so the output depends on the phase difference between the paths. - The electrical circuit changes that phase difference based on previous input fluxes, creating memory. - The new network showed stronger nonlinearity and stronger memory effects than a single photonic quantum memristor. - Those properties make the architecture well suited to reservoir computing. - The circuit was fabricated by an external company using standard techniques. - The phase-reconfiguration method was developed with optimization routines designed to enhance nonlinearity and memory. - Single-quanta inputs were generated by defects in nanodiamonds, and detectors measured individual arrivals of light quanta. - The paper reports DOI https://doi.org/10.29026/oea.2026.260048. - Funding came from Horizon Widera 2023, the EUR NANO-PHOT program, the Region Grand Est and the OQuLus PEPR project.

Between the lines: - Reservoir computing lets a system evolve with its own internal dynamics, rather than alternating tightly between processing and memory. - That structure can reduce computational cost because memory is intrinsic to the system. - The study suggests that larger photonic circuits with more memristors and more single photons could expand the number of computational nodes and interconnections. - That scale-up could improve performance on harder tasks such as real-time signal analysis and more complex temporal prediction. - The work also points to a hybrid model in which photonic hardware handles the dynamics and standard electronics read out the result.

What's next: - The researchers point to larger photonic circuits that combine multiple photonic quantum memristors and multiple indistinguishable single photons. - Future systems could pair many deterministic single-photon sources with more complex circuit designs to raise memory and nonlinear capacity. - That direction could lead to a new class of photonic processors for time-dependent data and dynamical systems.

The bottom line: - Two coupled photonic memristors have moved light-based computing closer to practical reservoir systems that remember, adapt and process time-varying data.