Programming light’s memory: A new step in Quantum Computing

• The study published in Nature Photonics shows, for the first time, how to exploit a quantum system of light equipped with controllable memory, enabling learning and prediction of complex signals.
• The research is the result of a collaboration between Laboratoire Kastler Brossel (LKB, Sorbonne Université) and the Institute for Cross-Disciplinary Physics and Complex Systems (IFISC, UIB–CSIC)

A team of researchers has experimentally demonstrated how to equip a photonic quantum system with controllable memory, exploiting the properties of quantum light to learn temporal patterns. The study, published in Nature Photonics, opens the door to new forms of quantum machine learning capable of analysing and predicting time dependent signals in real time. The work was carried out by scientists from the Institute for Cross-Disciplinary Physics and Complex Systems (IFISC, UIB–CSIC) and from the Laboratoire Kastler Brossel (LKB, Sorbonne Université), where the experimental platform was built.

From financial markets to climate data or brain activity, many real world problems depend on recognising patterns that evolve over time. This requires memory: the ability to retain information about the past while processing the present. “Memory is the key ingredient for learning from temporal signals”, explains Iris Paparelle, researcher at LKB and IFISC and first author of the study. “In our experiment, we show that it can be tuned directly at the level of the quantum optical process, without the need for complex classical architectures afterwards”.

The research focuses on quantum reservoir computing, a machine learning approach inspired by neural networks in which only the output layer is trained, while the physical system itself performs the main computational task. “Quantum reservoir computing is an efficient learning approach that allows us to exploit uniquely quantum properties, such as entanglement and squeezing, as computational resources”, says Roberta Zambrini, researcher at IFISC. “Our results show that these resources can be harnessed in a controllable and scalable way, opening the way for practical quantum enhanced learning”.

By shaping ultrafast laser pulses and measuring quantum correlations, the team created a reconfigurable photonic system capable of processing temporal information. To introduce memory, the researchers implemented a real time feedback mechanism so that previous inputs influence future responses. “We engineer what is known as fading memory”, explains Gian Luca Giorgi, researcher at IFISC. “The system keeps a trace of previous inputs, but this trace gradually weakens over time, which is exactly what is needed for temporal learning tasks”.

This platform allows researchers to benchmark problems that require both nonlinearity and memory, including the prediction of chaotic time series. “The capability to reproduce complex target patterns brings us closer to understanding how quantum systems can offer advantages in processing complex temporal data”, concludes Miguel C. Soriano from IFISC.

Beyond these demonstrations, the researchers highlight the broader implications of their experiment, showing the potential of photonics for quantum machine learning and forecasting. “Our setup operates at room temperature and relies on standard optical detection techniques”, notes Valentina Parigi, researcher at LKB. “This makes it a practical and scalable candidate for real time quantum enhanced information processing, capable of learning complex temporal patterns beyond classical approaches, as indicated by the improved scaling with the number of resources”.

Image: Conceptual illustration of temporal learning using quantum reservoir computing (left) and the photonic experimental platform used to implement controllable memory in quantum light at the Laboratoire Kastler Brossel (right).

Paparelle, I., Henaff, J., García-Beni, J. et al. Experimental memory control in continuous-variable optical quantum reservoir computing. Nat. Photon. (2026). https://doi.org/10.1038/s41566-026-01880-9

• Laboratoire Kastler Brossel
• El diari de la UIB
• Crónica Balear