Distributed GHZ States in Quantum Networks

• Distributed GHZ states are N-qubit maximally entangled states distributed among separate network nodes, serving as a core resource in quantum information science.
• They underpin critical quantum network protocols including distributed sensing, conference key agreement, and multi-target fan-out for quantum computations.
• Efficient generation strategies employ minimal Steiner trees, linear-optical methods, and quantum memories to optimize resource use and ensure high-fidelity distribution.

A distributed Greenberger–Horne–Zeilinger (GHZ) state is an $N$-qubit maximally entangled state $$\ket{\mathrm{GHZ}_N} = \frac{1}{\sqrt{2}}\bigl( \ket{0}^{\otimes N} + \ket{1}^{\otimes N} \bigr)$$, where each qubit resides in a distinct network node or physical location. Distributed GHZ states are fundamental resources in quantum networks, distributed computation, quantum metrology, and multiparty quantum cryptography. Their nonlocal correlations underpin network primitives such as distributed consensus, conference key agreement, and global parity measurement. The efficient generation, distribution, certification, manipulation, and application of distributed GHZ states are central tasks in quantum information science.

1. Distributed GHZ State: Definitions and Resource Model

In a quantum network, a distributed GHZ state refers to a multipartite entangled state with each constituent qubit (or qudit) held at a separated node. Formally,

$$\ket{\mathrm{GHZ}_N} = \frac{1}{\sqrt{2}}\left( \ket{0}^{\otimes N} + \ket{1}^{\otimes N} \right)$$

is both a stabilizer state and a graph state corresponding to the star graph—the canonical choice for distributed multiparty entanglement (Meignant et al., 2018). Each network node is assumed capable of local Clifford operations and local measurements; quantum channels serve to establish Bell pairs (bipartite entanglement) between connected nodes, which serve as the fundamental resource for distributed GHZ state construction.

In a standard resource model, operations and communications are classified as follows:

• Quantum channel use: Each use creates a Bell pair $\ket{\Phi^+}_{AB}$ between nodes $A$ and $B$; usage is counted as the main resource cost.
• LOCC: Local operations and classical communication (including Clifford gates and Pauli measurements) are assumed free and instantaneous.
• Network topology: The network is specified as a graph $\mathcal{N}=(V_C, E_C)$ with nodes $V_C$ and edges $E_C$, each supporting Bell-pair generation.

2. Protocols for Distributed GHZ State Generation

2.1 Optimal Distribution over Arbitrary Networks

The minimal-cost protocol for GHZ state distribution given an arbitrary network proceeds as follows (Meignant et al., 2018):

• Compute a minimal Steiner tree $T$ spanning the set $W$ of $N$ target nodes. The number of required Bell pairs is $|E_T|$.
• Star-expansion subroutine: At each round, select a leaf $\ell$ of $T$, and at its neighbor $A$, perform a local expansion by entanglement swapping and local operations, creating graph-state edges from $\ell$ to all remaining nodes and consuming relevant Bell pairs.
• Parallel LOCC round: All star-expansion operations across the network are mutually commuting and can be implemented in a single LOCC step, with classical postprocessing to correct Pauli by-products.
• Optimality: The total Bell-pair usage is exactly $|E_T|$. No protocol can succeed with fewer Bell-pair uses, as each edge in $T$ is necessary for connectivity among all GHZ parties.

This protocol outperforms naive sequential approaches, which typically scale linearly with path distance to the central node and may use up to twice as many Bell pairs, both in resource usage and LOCC round complexity.

2.2 Fan-Out and Distributed Quantum Computing

Distributed GHZ states function as primitives for nonlocal multi-target “fan-out” operations, enabling, e.g., distributed control gates in a single quantum depth round (Loke, 21 Jan 2026). A distributed GHZ state supports fan-out of a control qubit to $n-1$ spatially separated targets via local CNOTs, $Z$-basis measurements, Hadamards, and subsequent classical corrections, yielding logarithmic or even constant quantum circuit depth—resources that are otherwise linear using only Bell-pair-based decompositions.

2.3 Linear-Optical and Loss-Tolerant Schemes

Linear-optical schemes enable deterministic or probabilistic distributed GHZ state generation among spatially separated atomic or solid-state nodes via interference and postselection on multiple photons (Vivoli et al., 2018). In the presence of loss, certain protocols achieve rate-loss scaling $R \propto \eta^{N/2}$ (with $\eta$ single-photon link transmission) without quantum repeaters, remaining feasible with current technology for moderate $N$ and distances; this is exponentially favorable over direct transmission scaling $R \propto \eta^N$ (Shimizu et al., 2024).

2.4 Quantum Memories and Polynomial Scaling

Atomic-ensemble quantum memories allow asynchronous generation and fusion of entangled pairs into distributed GHZ states, supporting polynomial (rather than exponential) scaling in generation time for large $N$ (Zhang et al., 2022). Divide-and-conquer memory-assisted protocols accumulate small multipartite blocks and connect them with Bell or GHZ projections. This enables practical GHZ state distribution over large, time-uncertain networks, even in the presence of stochastic photon losses.

2.5 Distillation and Error Correction

Protocols using non-perfect Bell pairs and non-local stabilizer measurements (possibly leveraging quantum error correcting codes) provide robust generation and distillation of distributed GHZ states with high final fidelity, supporting arbitrary network topologies (Bone et al., 2020, Rengaswamy et al., 2021). Heuristic dynamic programming identifies optimal combinations of fusion and stabilizer measurement steps, and code-based multipartite distillation leverages the symmetry of the "GHZ map" for high-yield outputs under noisy links.

3. Extraction, Transformation, and Simulation

3.1 Extraction from Cluster and Graph States

GHZ states can also be extracted from resource graph states such as linear clusters through sequences of local Clifford unitaries and Pauli measurements (Jong et al., 2022). For a linear cluster of $n$ qubits, the maximal extractable GHZ is of size $k\leq\lfloor (n+3)/2 \rfloor$. Extraction is limited by the combinatorial structure of unmeasured “islands” and the maximal number and positions of $X$/$Z$ measurements.

3.2 Local Unitary Decomposition for Qudits

Any multipartite stabilizer state distributed among tripartite parties can be locally reduced to tensor products of $p$-level (prime-power) GHZ states, $p$-level EPR pairs, and product qudits, using Clifford and scaling unitaries. The full entanglement structure can be extracted via linear algebraic normal forms (subsystem phase matrices) computed in polynomial time (Wong et al., 12 Jul 2025).

3.3 Exact Classical Simulation

The outcome statistics of distributed GHZ measurements can be simulated classically with finite expected communication cost, using convex decompositions of the measurement probabilities, distributed Bernoulli sampling, and rejection sampling. For general $n$-qubit GHZ states and local von Neumann measurements, the expected communication is $O(n^2)$ bits; for equatorial measurements, this reduces to $O(n\log n)$ (Brassard et al., 2013).

4. Verification, Certification, and Fidelity Estimation

Distributed GHZ states require robust verification under arbitrary noise models. Fidelity estimation schemes achieve minimum mean-squared error by randomizing measurement settings across $X$, $Y$, and $Z$ axes, followed by error-flag counting and linear estimation to quantify global state fidelity (Ruan, 2024). Such protocols work under both i.i.d. and correlated noise and do not require global collective measurements or prior knowledge of the error model.

Verification of GHZ-entanglement via Mermin-Ardehali-Belinskii-Klyshko (MABK) inequalities and compressed-sensing tomography provide experimental confirmation of multipartite entanglement in photonic, atomic, or solid-state networks (Marcellino et al., 3 Dec 2025, Zhang et al., 2022).

5. Applications: Distributed Sensing, Cryptography, and Computation

Distributed GHZ states serve as resources for several fundamental quantum network tasks:

• Distributed Quantum Sensing: Sequential probing with distributed GHZ states (parameterized over the decoherence-free subspace) achieves near-optimal multiparameter precision even under strong correlated dephasing noise, with a quantum Fisher information matrix optimal up to a known factor (Hamann et al., 2023).
• Conference Key Agreement and Secret Sharing: GHZ-based protocols provide enhanced rates and thresholds in conference key agreement, quantum secret sharing—with threshold quantum-bit error rates— and distributed voting with information-theoretic anonymity (Marcellino et al., 3 Dec 2025, Zhang et al., 2022).
• Distributed Parity/Control-Operations: GHZ states as fan-out primitives reduce the circuit depth for global parity and control operations in distributed quantum computation, supporting constant or logarithmic depth implementations compatible with large-scale architectures (Loke, 21 Jan 2026).
• Quantum Repeaters: Incorporation of 2D repeater architectures, combining short-range GHZ generation, memory-assisted fusion, and measurement-based projective protocols, enable long-distance maintenance of high-fidelity GHZ-type entanglement across networks at favorable scaling of resource overhead and success probability (Azari et al., 16 May 2025, Kuzmin et al., 2019).

6. Limitations, Challenges, and Scalability

Despite their utility, distributed GHZ states face significant challenges:

• Success probability in heralded schemes decays exponentially in $N$ in all-photonic implementations without repeaters, though ensemble-based and repeater-enhanced architectures mitigate this (Vivoli et al., 2018, Shimizu et al., 2024, Kuzmin et al., 2019).
• Extraction from cluster/graph state resources is topologically constrained; maximal GHZ size depends on specific network geometry (Jong et al., 2022).
• LOCC transformation and certification protocols scale favorably with $N$ only when leveraging quantum memories and nonlocal stabilizer operations (Bone et al., 2020, Zhang et al., 2022).
• Error robustness and fidelity certification in the presence of general, correlated, or adversarial noise remains an active area of investigation (Ruan, 2024).
• Classical simulation of GHZ measurement outcomes over distributed parties, while polynomial in $n$, still demands substantial communication and random bits for large-scale systems (Brassard et al., 2013).

7. Extensions and Outlook

Distributed GHZ states generalize to $d$-level (qudit) and high-dimensional settings, with local-unitary decompositions available for arbitrary prime-power local levels. Extensions include hybrid fusion protocols accommodating mixed architectures and universal decompositions for stabilizer resource states (Wong et al., 12 Jul 2025). Multiparty protocols leveraging distributed GHZ entanglement underpin the roadmap for scalable, fault-tolerant networked quantum computation and cryptography. Advances in quantum repeater networks, memory integration, and loss-tolerant architectures continue to push the feasibility boundary for large-$N$ distributed GHZ state generation and utilization across diverse physical platforms (Azari et al., 16 May 2025, Shimizu et al., 2024, Zhang et al., 2022, Kuzmin et al., 2019).