arXiv:2610.35838
Network nonlocality with tripartite entangled sources
Sudip Chakrabarty, Amit Kundu
Abstract
Quantum networks are built from independent sources, and their nonlocality has mostly been studied with sources that emit bipartite entangled states. A source of three entangled qubits is richer, since its nonlocality can be tested by the Mermin or the more demanding Svetlichny inequality. We ask how network nonlocality relates to these two single-source notions and find that the answer depends on the wiring. When two central nodes without inputs share every source, Bell-state measurements turn every Svetlichny violation into a network violation, and even all entangled generalised Greenberger-Horne-Zeilinger states violate the network inequality. When a single central node is shared, the relation reverses, and a network violation requires a Mermin violation of the source. With more sources, the maximal violation depends on whether their number is even or odd, and we give tight bounds for both cases. These results show that the way multipartite sources are wired decides which kind of nonlocality a network inherits.
AI Summary
The paper asks how nonlocality of independent three-qubit sources relates to single-source Mermin and Svetlichny violations, and how this depends on network wiring. It studies two-correlator nonlinear inequalities for closed networks, where two input-free central nodes share every source, and open networks, where only one central node is shared. For two identical sources, Bell-state measurements at the central nodes and projective qubit measurements at the ends, the optimized closed-network value satisfies $\mathcal N_c(\rho)\ge\mathcal S(\rho)/4$, with network-local bound $1$. Thus Svetlichny violation is sufficient but unnecessary: every entangled generalized GHZ state achieves at least $\sqrt{1+\sin^2(2\alpha)}$, and some biseparable sources also violate. In the open network, $\mathcal N_o(\rho)\le\mathcal M(\rho)/2$, so network violation requires Mermin violation within this measurement class. Correlation-tensor formulas and geometric comparisons establish these bounds; numerical optimization supplies additional examples. For $n$ sources and arbitrary central measurements, tight closed-network maxima are $\sqrt2$ for even $n$ and $2^{1/2-1/(2n)}$ for odd $n$; open-network maxima are $2$ and $2^{1-1/(2n)}$. The proof uses anticommutation of end-node operator products, with GHZ constructions attaining the bounds. GHZ entanglement swapping is also analyzed: both central outcomes are needed to herald end-node entanglement. The single-source comparisons concern the specified inequalities and Bell-state measurements, not network nonlocality generally. The closed parity bound assumes qubit ends; more-correlator inequalities can avoid the parity effect. General measurement optimization and tests certifying genuine multipartite sources remain open.
Central claim may not hold: The abstract's unqualified claim that an open-network violation requires a single-source Mermin violation is false for general central-output processing: Bell-pair sources with a maximally mixed central qubit violate the stated network inequality when the central outcome is ignored, despite having zero Mermin value under the paper's definition.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 16 pages, 5 figures