arXiv:2609.25673
Post-selected enhancement of phase sensitivity in a two-path interferometer
Ryuya Fukuda, Katsuki Tanaka, Masataka Iinuma, Holger F. Hofmann
Abstract
Post-selected metrology can exceed the standard quantum limit of single photon interference by concentrating all of the available Fisher information in a subset of post-selected outcomes. Here we apply this principle to a standard two-path interferometer, activating the Fisher information concentrated in the low probability output port by transferring the phase sensitivity to the circular polarizations of the photon. We obtain a phase gradient of the post-selected interference fringe that corresponds to that of a four-photon NOON state, providing a direct comparison of the performance of multi-photon interference with single-photon post-selection.
AI Summary
The paper asks how phase information concentrated near the dark output of a two-path interferometer can remain accessible after post-selecting that port. Projecting onto a single path-output state otherwise removes its phase dependence. The proposed scheme introduces an ancillary polarization degree of freedom through opposite polarization rotations $\pm\theta_0$ in the two arms, then measures circular polarization at the selected output. For the ideal state $\cos\theta_0|+\rangle|V\rangle+\sin\theta_0|-\rangle|H\rangle$, the conditional Fisher information at zero phase is $I_{\mathrm{ps}}=\cot^2\theta_0$. The enhancement is per post-selected detection, not per incident photon: the ideal success probability there is $\sin^2\theta_0$. The theoretical interpretation connects the enhancement to anomalous conditional path-number uncertainties and negative Kirkwood–Dirac quasiprobabilities. A Sagnac-type experiment uses an attenuated 808.5-nm laser and polarization-resolved photon counting. The measured rotation, $\theta_0=0.16848\pm0.00004$ rad, predicts ideal conditional Fisher information $34.56$. Finite-difference estimates of the measured polarization fringe yield a slope near $-2.028$ at equal circular-polarization probabilities, corresponding to $I_{\mathrm{ps}}\simeq16.5$ and an effective fringe gradient equivalent to approximately four-photon NOON-state interference. This compares local conditional sensitivities, rather than total photon-resource requirements. Experimental imperfections limit the enhancement: the measured visibility is $0.9266\pm0.0005$, below that expected from the intentional polarization rotations alone. The slope estimate also uses corrections for fringe curvature at finite phase steps. The demonstrated enhancement is local to the low-probability, near-destructive-interference operating region.
Error found: At the stated finite operating point, the ideal selected port retains 97.188% of the input Fisher information after weighting by its success probability, so the abstract's 'all' is an asymptotic idealization rather than an experimentally established equality. The pi/3 correction used to obtain the reported local phase gradient follows from neither the stated ideal post-selected fringe nor an ideal four-photon NOON fringe at the quoted sampling interval, leaving the precise corrected gradient and its uncertainty unsupported.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.