arXiv:2610.06526
Classical Simulation of Lossy Gaussian Boson Sampling beyond Square-Root Regime
Youngrong Lim
Abstract
Photon loss makes Gaussian boson sampling easier to simulate classically. A standard approach replaces the lossy input by classical light, and it stays accurate while the mean number of surviving photons grows no faster than the square root of the number $N$ of squeezed inputs at fixed squeezing and accuracy. We prove that this scale is a limit of replacement by classical light. For explicit interferometers, no state with a nonnegative Glauber--Sudarshan $P$ representation, Gaussian or not and however correlated, keeps the error in the photon count distribution small beyond this scale. We then surpass this limit by approximating the detection instead of the input. The sampler evolves the state exactly between detections and, after each detection, replaces the conditional state by a Gaussian with the same covariance. For every interferometer its total variation error is $O(Nη^3)$ at polynomial expected cost, where $η$ is the photon survival probability. The mean surviving photon number can therefore grow as $N^{2/3}$ at fixed accuracy. A separate pair construction reaches $N^{3/4}$ when every input mode is equally squeezed.
AI Summary
The paper asks whether classical simulation of lossy Gaussian boson sampling can exceed the regime of $O(\sqrt N)$ surviving photons at fixed squeezing and accuracy. It separates limitations of classical-light replacement from limitations of classical computation. For an explicit interferometer family built from independent two-mode squeezed pairs, every state with a nonnegative Glauber–Sudarshan $P$ representation—including non-Gaussian, arbitrarily correlated states—has photon-count total variation error at least $1-e^{-c_rN\eta^2}$. A bounded count-probability witness and an affinity bound establish this worst-case barrier. The family itself is efficiently simulable, so this is not a computational hardness result. A different sampler retains the exact Gaussian input and no-click evolution, but replaces each photon-subtracted conditional state by a Gaussian with matching covariance. For every interferometer, including output collisions and additional vacuum modes, its error is $O_r(N\eta^3)$ at polynomial expected arithmetic cost, permitting $O(N^{2/3})$ surviving photons. A covariance-potential drift argument controls accumulated reset errors and rare trajectories. When all modes are equally squeezed ($M=N$), a separate pair-generation/amplification construction achieves error $O_r(\sqrt N\,\eta^2)$ under stated small-loss conditions, permitting $O(N^{3/4})$ surviving photons. These guarantees concern loss decreasing with system size, not fixed nonzero transmission; the general bound is trivial at the cited experimental parameters. Small-system simulations favor a combined displacement-noise/reset variant over classical Gaussian replacement, but no uniform error bound is proved for that variant. Typical-interferometer classical-light barriers and the ultimate efficient-simulation threshold remain open.
GPT-6 and Fable-5.1 were used under the author's direction to assist with exploring and refining mathematical arguments, checking calculations, numerical simulations, and revising the manuscript.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 6+12 pages, 2 figures