| Popis: |
Many experiments with high-dimensional quantum systems aim at observing physical properties without having to assume that all measurement devices are perfectly characterised. To implement high-dimensional measurements, it is a common practice to separately project the system onto each of the possible eigenstates followed by post-processing. Here, we argue that such measurement binarisation procedures are flawed and open a loophole in black-box correlation experiments. We point out how binarised measurements can lead to meaningless results unless artificial assumptions are introduced. We then propose a method to faithfully analyse correlations from binarised measurements. We apply this approach to well-known examples of quantum correlations in Bell, prepare-and-measure and steering experiments. For the former two, we find that binarisation permits reduced-but-yet-significant quantum advantage, while for the latter we find that binarisation can sometimes lead to undiminished quantum correlations. Our discussion emphasises the importance of both more careful data analysis and for developing genuine multi-outcome quantum measurements. |
