Benchmarking single-photon sources from an auto-correlation measurement

TL;DR

This paper introduces a practical benchmarking method for single-photon sources using auto-correlation measurements, enabling certification of single-photon purity and non-classicality even with experimental imperfections.

Contribution

It provides a simple, robust framework to bound single-photon probability and non-classicality witnesses from auto-correlation data, accounting for real-world imperfections.

Findings

Single-photon probability ≥ 55% in experiments

Wigner negativity measure ≥ 0.004 with high confidence

Method effectively certifies non-classicality in practical setups

Abstract

Here we argue that the probability that a given source produces exactly a single photon is a natural quantity to benchmark single-photon sources as it certifies the absence of multi-photon components and quantifies the efficiency simultaneously. Moreover, this probability can be bounded simply from an auto-correlation measurement -- a balanced beam splitter and two photon detectors. Such a bound gives access to various non-classicality witnesses that can be used to certify and quantify Wigner-negativity, in addition to non-Gaussianity and P-negativity of the state produced by the source. We provide tools that can be used in practice to account for an imperfect beam splitter, non-identical and non-unit detection efficiencies, dark counts and other imperfections, to take finite statistical effects into account without assuming that identical states are produced in all rounds, and optionally allow one to remove the detector inefficiencies from the analysis. We demonstrate the use of the proposed benchmark, non-classicality witness and measure using a heralded single-photon source based on spontaneous parametric down-conversion. We report on an average probability that a single photon is produced $\geq 55\%$ and an average measure of the Wigner negativity $\geq 0.004$ with a confidence level of $1-10^{-10}$.