Multipartite Quantum Nonlocality

• Nonlocality of multipartite systems refers to many-body quantum correlations that defy simulation by any combination of fewer-than-N-party local models.
• Advanced methods like LOSR models, Bayesian networks, and matrix inflation techniques rigorously certify nonlocal behavior via Bell-type inequalities and noise thresholds.
• Insights into multipartite nonlocality drive applications in device-independent cryptography, secret sharing, and resilient quantum network security.

Nonlocality of Multipartite Systems

The nonlocal correlations arising in multipartite quantum systems exhibit a hierarchy of structures and resource-theoretic stratifications far richer than the bipartite setting. Multipartite nonlocality is rigorously defined as the failure of any decompositional model wherein fewer-than-N-party nonlocal resources, together with arbitrary shared randomness and local operations, can simulate the observed correlations among N parties. These phenomena are central to foundational studies, device-independent cryptography, quantum networks, and information hiding.

1. Foundational Definitions and Causal Decomposition

Multipartite nonlocality generalizes the bipartite notion based on the local hidden-variable (LHV) model. For N parties, classical models posit

$$P(\mathbf{a}|\mathbf{x}) = \sum_\lambda q_\lambda \prod_{i=1}^N P_i(a_i|x_i,\lambda),$$

where $a_i$ is the measurement outcome and $x_i$ the setting of party $i$. Standard nonlocality witnesses refute such models via Bell inequalities.

Genuine multipartite nonlocality is defined by the impossibility of expressing the joint correlations as a convex mixture of resource correlations involving at most $k<N$ nonclassical subsystems:

$$P(\mathbf{a}|\mathbf{x}) = \sum_\lambda q_\lambda \prod_j P_j^{(k)}(\mathbf{a}_j|\mathbf{x}_j,\lambda),$$

where each $P_j^{(k)}$ involves at most $k<N$ parties (Mao et al., 2022). This "LOSR" (Local Operations and Shared Randomness) framework eliminates explanations in terms of stitched-together smaller nonlocal boxes. The approach is theory-agnostic: it excludes all hybrid models, including those using non-signalling multipartite resources of size $k<N$ glued by global randomness.

The causal structure underlying these decompositions has been formalized via Bayesian networks and IO-BDAGs (input-output Bayesian Directed Acyclic Graphs) (Chaves et al., 2016). This yields a partial order of possible relaxations, from fully local, through various pairings and communication patterns, to the nonsignalling extreme.

2. Bell-Type Inequalities and Matrix-Inflation Techniques

Certifying genuine multipartite nonlocality generally requires Bell-type inequalities constructed to exclude all $k$-party decompositions. A recent algebraic development introduces a matrix-based inflation technique (Mao et al., 2022): representing the sharing pattern of nonlocal resources by an $a_i$0 incidence matrix $a_i$1, the approach employs non-fan-out inflations, replicating the network and enforcing identity under device replication, no-signalling, and inflation constraints (causal compatibility of subnetworks).

For $a_i$2 parties each performing two dichotomic measurements, the canonical inequality is:

$a_i$3

under all LOSR models derived from $a_i$4-partite resources (Mao et al., 2022). Quantum mechanical predictions for GHZ and generalized GHZ states violate this bound, attaining

$a_i$5

Inclusion of white noise introduces a critical threshold $a_i$6 for the robustness of violation. This approach unifies inflation techniques and graph-theoretic representation, yielding optimal or near-optimal quantum-to-classical separations.

3. Hierarchies of Multipartite Nonlocality

Beyond genuine nonlocality, multipartite correlations admit further classification. There exists a strict hierarchy, ranging from full locality, through $a_i$7-local models (nonsignalling models built from $a_i$8 blocks), to the standard (in the sense of Hardy or Mermin) and genuine extremes (Wang et al., 2016). For an $a_i$9-party system, the concept of nonsignalling $x_i$0-locality gives rise to a series of Bell-type inequalities, each detecting the failure of models with up to $x_i$1 subsystems. Explicit thresholds for GHZ and W states under noise display the strictness of this hierarchy.

Device-independent witnesses such as Mermin, Svetlichny, and Hardy-type inequalities, as well as two-body correlation-based constructions (Tura et al., 2013), can be constructed to target various levels of the hierarchy. Distinctions between NS$x_i$2 (non-signalling hybrid), T$x_i$3 (time-ordered hybrid), and Svetlichny's S$x_i$4 definitions are essential for cryptographic security in the presence of adversaries constrained only by no-signalling or constrained communication (Bancal et al., 2011).

4. Strong Quantum Nonlocality, State Discrimination, and Information Hiding

A distinct line of inquiry examines orthogonal sets—product or entangled—whose local indistinguishability encodes nonlocality "without entanglement." Strong nonlocality is defined by local irreducibility: no orthogonality-preserving local measurement on any subset, or across any bipartition, can eliminate any state from the set (Shi et al., 2022). Construction of minimal strongly nonlocal sets in arbitrary $x_i$5 exploits orbit structures under cyclic permutations and Fourier-rotated superpositions to ensure full local irreducibility. For $x_i$6 and $x_i$7, all states in such sets can be genuinely entangled; for $x_i$8, explicit constructions remain nontrivial (Shi et al., 2022, Halder et al., 2018).

A separation exists between "distinguishability-based genuine nonlocality" (no LOCC discrimination across any cut) and strong nonlocality (local irreducibility in all bipartitions) (Xiong et al., 2022). The minimal size for distinguishability-based nonlocality can be as low as $x_i$9 in $i$0, whereas strongly nonlocal sets require $i$1. This suggests practical implications for quantum secret sharing and local information hiding protocols.

Multipartite information hiding leverages strongly nonlocal orthogonal product sets such that no coalition of $i$2 parties, regardless of LOCC access, can gain any information about the encoded message, certifying device-independent security (Shi et al., 2022, Zhou et al., 2022). Entanglement-assisted protocols may reduce the entanglement cost of discriminating such sets compared to naive teleportation.

5. Experimental Realizations and Noise Tolerance

Demonstrations of genuine multipartite nonlocality have been implemented in all-optical networks with GHZ and generalized GHZ states. For four-photon systems, violation of the LOSR inflation-based inequality by more than 30 standard deviations confirms the infeasibility of explanations via triseparable nonlocal resources (Mao et al., 2022). Quantum nonlocality has also been detected in multidimensional GHZ lattices and in sequential observer scenarios, where an unbounded number of measurement chains can sequentially extract nonlocality (via Mermin inequality violations) for $i$3 (Shen et al., 2024).

Noise thresholds for multipartite entanglement and nonlocality are highly state- and scenario-dependent. For example, under local $i$4-dephasing, $i$5-qubit GHZ states are nonlocal under CHSH-based witnesses for all $i$6, while the corresponding Mermin–Klyshko inequality thresholds decay with $i$7 (Chaves et al., 2013). In contrast, Dicke and W states typically require more robust multi-body or two-body inequalities for optimal certification (Tura et al., 2013).

Notably, in physically relevant systems such as gapped spin lattices or finite-temperature Gibbs states, genuine multipartite nonlocality can be suppressed by exponential clustering of correlations, with only biseparable quantum correlations persisting when measured regions are far apart (Yang et al., 2021).

6. Resource Theory, Network Nonlocality, and Equivalence Theorems

The resource-theoretic approach frames multipartite nonlocality as a monotone under free (classically simulatable) operations, with operational notions extending to general quantum networks (Lamas et al., 2022). In Clifford-stabilizer networks, pure stabilizer resources and Clifford gates cannot activate network nonlocality—even aggregating arbitrarily many parties—whereas mixtures enable network nonlocality, evincing convexity failures.

Network-inflation approaches have recently shown that, for any isolated pure $i$8-partite quantum state, genuine multipartite entanglement (GME), genuine multipartite EPR-steering (GMS), and genuine multipartite nonlocality (GMNL) are equivalent (Luo et al., 2023). The equivalence is witnessed via a single extended Bell test on an appropriately inflated network, reinforcing a device-independent unification for finite-dimensional systems.

A resource hierarchy persists: full locality $i$9 nonsignalling bilocal $k<N$0 (asymmetric) time-ordered $k<N$1 general bilocal $k<N$2 full nonsignalling sets, with refined classes relevant to operational tasks and cryptographic protocol design (Dutta et al., 2020, Chaves et al., 2016).

7. Applications, Open Problems, and Outlook

Multipartite nonlocality fundamentally underpins device-independent multiparty cryptography, secret sharing, certified randomness, and resilient quantum networks. Criteria for persistency (minimum party loss before entanglement or nonlocality disappears), symmetry-vs-persistency tradeoffs, and the impact of decoherence on nonlocal correlations are all now quantitatively characterized (Brunner et al., 2012, Ali, 2016).

Open problems include:

• Extension of causal-matrix inflation techniques to arbitrary graph topologies or higher-rank resource sharing.
• Analytical determination of optimal measurement settings for hierarchy-detecting inequalities at large $k<N$3 (Wang et al., 2016).
• Minimal-size constructions for strongly nonlocal sets in various dimensions and their operational advantage in cryptographic tasks (Zhen et al., 2022).
• Tight device-independent lower bounds on asymmetry and required entanglement for nonlocally indistinguishable sets.
• Theoretical and experimental scaling to larger $k<N$4 in noisy and naturally fluctuating environments, using state-of-the-art entanglement and correlation witnesses.

The unification of foundational, operational, and experimental viewpoints in multipartite nonlocality theory continues to elucidate the structure and capabilities of many-body quantum systems as resources for distributed information processing and secure communication.