Loading...
News Article

Photonic chip advances quantum estimation

News

Researchers have demonstrated a quantum analogue-to-digital converter using a programmable photonic integrated circuit.

An international research team has demonstrated a quantum analogue-to-digital converter (QADC) for optical phase estimation using an integrated photonic processor.

Published in Nature Photonics, the work explores how quantum information processing can convert continuous optical phase shifts into digital information.

Unlike conventional approaches that estimate phase as a continuous quantity, the quantum converter extracts a binary representation of the measured phase.

The researchers evaluated its performance according to the amount of recoverable information, comparing it with classical interferometric methods using equivalent detected-photon resources.

At the centre of the experiment is an eight-mode programmable photonic integrated circuit fabricated using femtosecond laser writing.

The processor occupies an area of 80 mm² and incorporates 28 reconfigurable Mach–Zehnder interferometers controlled by 56 thermo-optic heaters.

The architecture allows the researchers to programme optical transformations for generating and manipulating quantum states.

Following calibration to account for thermal cross-talk, the circuit achieved an average transformation fidelity of 0.995 ± 0.002.

The experimental system uses single photons generated by a semiconductor quantum-dot source.

These are directed into the photonic processor, where four-photon, two-photon and single-photon states are used in successive stages to extract the binary components of an unknown optical phase.

Processed photons are measured using superconducting nanowire single-photon detectors.

The team tested the system across 99 phase values spanning a full 2π range to assess its accuracy and information recovery.

Although the quantum approach initially outperformed its classical counterpart, experimental imperfections reduced the advantage as measurement numbers increased.

The researchers identified optical losses, programming inaccuracies and photon distinguishability among the contributing factors.

To address these limitations, the team introduced a machine-learning framework incorporating a denoising autoencoder and a separate neural network for phase reconstruction.

The denoising approach improved the information recovered from experimental measurements, while the reconstruction network enabled phase estimation across the full 2π range.

The researchers report that the machine-learning-assisted quantum converter maintained an information-theoretic advantage over the classical benchmark across the tested measurement regimes.

The demonstration highlights the potential of programmable integrated photonics for quantum metrology and digital sensing.

However, the experimental system still relies on external photon-generation and detection hardware, as well as post-selection, leaving further integration and scalability challenges to be addressed.


Logo
x