Discrimination of quantum states under locality constraints in the many-copy setting

PDFHTML

We study quantum hypothesis testing between orthogonal states under restricted local measurements in the many-copy scenario. For testing arbitrary multipartite entangled pure state against its orthogonal complement state via the local operation and classical communication (LOCC) operation, we prove that the optimal average error probability always decays exponentially in the number of copies. Second, we provide a sufficient condition for the LOCC operations to achieve the same performance as the positive-partial-transpose (PPT) operations. We further show that testing a maximally entangled state against its orthogonal complement and testing extremal Werner states both fulfill the above-mentioned condition. Hence, we determine the explicit expressions for the optimal average error probability, the optimal trade-off between the type-I and type-II errors, and the associated Chernoff, Stein, Hoeffding, and strong converse exponents. Then, we show an infinite asymptotic separation between the separable (SEP) and PPT operations by providing a pair of states constructed from an unextendible product basis (UPB). The quantum states can be distinguished perfectly by PPT operations, while the optimal error probability, with SEP operations, admits an exponential lower bound. On the technical side, we prove this result by providing a quantitative version of the well-known statement that the tensor product of UPBs is a UPB.
Submitted 25 Nov 2020 to Quantum Physics [quant-ph]
Published 30 Nov 2020
Updated 28 Aug 2023
Author comments: Paper re-organized and revised
https://arxiv.org/abs/2011.13063
https://arxiv.org/pdf/2011.13063.pdf
https://arxiv-vanity.com/papers/2011.13063

View this paper on arXiv.wiki:
https://arxiv.wiki/abs/2011.13063

2 comments

Nengkun Yu Nov 30 2020 08:36 UTC (1 points)

There seems to be an issue with the arXiv tex engine.
No idea why the system accepted our submission in the first place,
when everything seemed OK, but cannot compile it now.

However, it _does_ produce perfectly good Postscript, which you
can then convert to pdf or any other format.

Nengkun Yu Nov 30 2020 19:36 UTC

Problem solved.