A frequency-comb sensor separates a weak signal from measured noise

A sensor is useful not simply when it responds strongly, but when its response to the thing being measured stands out from fluctuations. That distinction matters in photonics, where amplifying light can also amplify noise. In this experiment, researchers built a system intended to make a weak signal and environmental noise grow at different rates as the system became larger.

The idea draws on a theoretical model called a bosonic Kitaev chain. In the version relevant here, two linked chains allow different components of light—its two quadratures—to propagate preferentially in opposite directions. A weak perturbation can transfer part of the signal from one chain to the other, where it is amplified on its return. The proposed benefit is that a localized probe signal gains more than noise spread through the system.

Turning frequency into a lattice

Instead of arranging many resonators in a row, the team used resonant frequencies in an optical-fiber loop. Each frequency mode acted like a site in a one-dimensional lattice: neighboring modes were coupled by modulating the light circulating in the cavity. A second loop helped define finite groups of modes, and two such groups represented the coupled chains.

The researchers used electro-optic phase and intensity modulation together to create non-reciprocal propagation in frequency space. Propagation in one direction was stronger than in the reverse direction, rather than being strictly one-way. They added a small modulation tone to couple the chains and measured the outgoing light with heterodyne detection. This method mixes the device’s light with a frequency-shifted reference, allowing researchers to examine different frequency components.

To test environmental noise, they removed the coherent laser drive and measured amplified spontaneous emission from an amplifier inside the loop. This provided a broad, approximately uniform source of optical noise across the modes. They compared measurements with Langevin-equation simulations as they varied chain length, reaching 74 frequency modes per chain.

What grew—and what did not

With the weak coupling tone present, the measured signal increased rapidly with chain length. The noise increased too, but more slowly. The authors report an exponential increase in the signal-to-noise ratio (SNR) over the tested range, which they connect to increased quantum Fisher information per input photon. Quantum Fisher information is a theoretical measure of how much information a measurement could, in principle, provide about an unknown parameter.

The paper distinguishes the scaling of signal intensity from that of noise: the output signal intensity is reported to scale approximately as exp(2AN), while noise scales as exp(AN). Here, N is chain length and A describes the degree of non-reciprocity. The authors report that the resulting SNR gains an exponential factor with system size. At a chain length of 54 modes, they also varied perturbation strength: the signal rose roughly linearly, while the measured noise stayed small and nearly constant over the tested range.

This is not a claim that any optical amplifier automatically improves sensing. The result depends on how signal and noise enter the device. Here, the probe was localized, whereas amplifier noise entered across the frequency modes. The authors argue that this difference, combined with the two-chain architecture, explains the distinct scaling. The measured perturbation was an intentionally added modulation tone, not a weak microwave signal or another practical target. Detecting such targets is a possible future application, not what this experiment demonstrated.

Important caveats

The noise accounting has a significant boundary: the environmental-noise test did not include the input drive’s inherent shot noise, which the authors say would also produce amplified fluctuations. They note that squeezed light could potentially address this contribution, but maintaining squeezing in a lossy, non-Hermitian system is an open challenge. The experiment therefore establishes the claimed scaling for the noise sources it tested, not a complete noise budget for a deployed sensor.

The researchers also avoided extreme directionality to limit nonlinear effects in the fiber, and used a specific fiber-loop platform and amplifier. Performance in other devices, with different noise sources or practical measurement targets, remains to be established. The result does not mean precision grows exponentially with the number of photons; the paper distinguishes scaling with system size from scaling with photon number.

The study, Exponentially enhanced sensing through non-reciprocal light propagation, is an arXiv preprint. Preprints may not have undergone peer review, and the available evidence does not establish independent replication. The experiment offers a demonstration that engineered directional bias can produce different scaling for a signal and the measured environmental noise in a frequency-based optical lattice. Whether that advantage survives the additional noise and engineering constraints of practical sensing remains an open question.