arXiv:2609.24835
Trade-offs and experimental feasibility of nonlocal polygamy with two-outcome Bell inequalities
Josep Batle, Tomasz Rybotycki, Tomasz Białecki, Piotr Gawron, Adam Bednorz
Abstract
Entanglement and Bell nonlocality have different sharing constraints across subsystems of a multipartite quantum system. We examine simultaneous violations of Bell inequalities with two-outcome observables and two or three measurement settings per party. We identify configurations whose simultaneous violations can be demonstrated on IBM Quantum hardware, and derive bounds and trade-off relations for several polygamous configurations. The results connect experimental capabilities of quantum computers with mathematical methods for studying the limits of quantum correlations.
AI Summary
The paper studies how Bell nonlocality can be shared among overlapping subsystems when each party has two-outcome measurements and two or three settings. It combines optimized constructions, trade-off bounds, semidefinite-programming (SDP) certificates, and IBM Quantum demonstrations. For four parties, Śliwa inequality 5 admits equal violations on all four three-party subsets: $3.220475$ against a local bound of $3$. Hardware measurements give $3.0384$–$3.1000$. For five parties, simultaneous four-party MABK violations yield hardware values $4.1159$–$4.1990$ against $4$; the stated ideal construction gives $5.05964$ per subset. Separability bounds show that simultaneous violations alone need not certify genuine multipartite entanglement. The theoretical analysis constructs explicit dual certificates for the Lovász-$\vartheta$ bound on squared Pauli correlations, improving on anticommuting-clique grouping in selected configurations. It proves the uniform bound $\sum_j(12\cdots\hat j\cdots n)\leq2^{n-2}$, where each term sums squared $X/Z$ correlations, and an exact odd-cycle bound $C_{2n+1}\leq4n$, established algebraically and saturated by products of GHZ-equivalent states. With three settings, cyclic CHSH expressions sharing one observable have optimal sum $4+2\sqrt2$, attainable by biseparable strategies; hardware violations are also reported. Two aligned $I_{3322}$ copies reach approximately $-8.200346$ against $-8$, with numerical NPA agreement but no exact global-optimality certificate. Three cyclic copies cannot jointly violate within the common-sign relabelling family studied. Limitations include dimension-dependent entanglement conclusions, potentially loose first-level SDP bounds, and experimental uncertainties reported as shot noise rather than a full systematic-error analysis.
Error found: The displayed three-dimensional biseparable I3322 state and observables attain approximately -4.862 for I_AB + I_AC even after optimizing the separated party's deterministic outputs, rather than reproducing the reported value near -8.12948.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 34 pages, 9 figures, 22 tables, assisted by ChatGPT 6, contains real IBM Quantum data