arXiv:2609.40307
A fully Gaussian quantum Stein's lemma
Filippo Girardi, Jacopo Rizzo, Leah Turner, Gerardo Adesso, Ludovico Lami
Abstract
A central yet mysterious problem in quantum information is the asymptotic distinguishability of quantum states under restricted measurements. Here we look at asymptotic asymmetric hypothesis testing for general multi-mode bosonic Gaussian states restricted to arbitrary collective Gaussian measurements. We establish a single-letter formula for the Stein exponent for many pairs of states, featuring two distinct contributions, corresponding to first and second moments, respectively. Remarkably, the Gaussian-measured relative entropy exhibits genuine non-additivity, but the only effect of additivity violations is to asymptotically decouple first and second moments. We prove this by establishing sufficient conditions for additivity of the second-moment term via novel log-determinant matrix inequalities. Our results not only identify a single-letter expression for the Gaussian Stein exponent in a wide range of practical situations, but also rule out any possible advantage from collective Gaussian measurements when considering centred Gaussian states.
AI Summary
The paper studies asymmetric hypothesis testing between i.i.d. multimode bosonic Gaussian states when arbitrary collective Gaussian measurements and classical post-processing are allowed. It defines the Gaussian Stein exponent by taking the many-copy limit followed by the type-I error bound tending to zero, and identifies it with the regularised Gaussian-measured relative entropy. For arbitrary Gaussian hypotheses with means $s,t$ and covariance matrices $V,W$, it proves the exact decomposition $\mathrm{Stein}_{\mathrm G}(\rho\|\sigma)=\mathrm{Stein}_{\mathrm G}(\rho^0\|\sigma^0)+(s-t)^\top W^{-1}(s-t)$, where the superscript denotes centred states. Passive mixing across copies concentrates the displacement difference into one output copy without changing the covariances, allowing displacement and covariance discrimination to be optimised separately asymptotically. The covariance contribution is proved additive whenever a symplectic transformation satisfies $SWS^\top=wI$ and, with $\widetilde V=SVS^\top$, $[\widetilde V,\Omega\widetilde V\Omega^\top]=0$. This includes every single-mode pair and every pure multimode pair. For these classes, the exponent reduces to a single-copy covariance optimisation plus the explicit displacement term; centred states gain nothing from collective Gaussian measurements. The proof combines passive diagonalisation, reduction to pure measurement seeds, and a log-determinant inequality involving eigenvector overlaps. Single-mode examples identify homodyne or finitely squeezed general-dyne measurements as optimal. The general mixed multimode covariance problem remains unresolved: the criterion is sufficient, not necessary, and whether centred Gaussian states can exhibit genuine non-additivity is left open.
were used as research assistants to help explore extensions of the arguments, refine proofs of some of the general results, and improve parts of the presentation.
Central claim may not hold: The abstract rules out collective Gaussian measurement advantage for centred Gaussian states without qualification, whereas the theorem proves this only under explicit covariance conditions and the Discussion leaves the general case open. The abstract attributes all effects of non-additivity to moment decoupling, although the separation theorem retains a potentially non-additive covariance contribution whose general additivity remains unresolved.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 5 + 12 pages, 1 figure