Complex quantum network models from spin clusters

The paper addresses how complex network topology might emerge and function in a future quantum internet. While complex topologies are beneficial in classical networks (e.g., robustness and small-world properties), their suitability for quantum communication is uncertain due to constraints like photonic losses, limited transmission distances, low network density possibly limiting end-to-end quantum capacity, and vulnerability to targeted attacks. Existing quantum network proposals often impose simple grid-like structures and routing protocols assume regular topologies. The authors propose a model system to study the impact of complex topology in quantum networks by mapping interacting spin systems on a 2D lattice to quantum communication networks, using entangled spin clusters as links between spatially localized regions (nodes). The aim is to test whether such constructions can yield complexity comparable to the classical internet and to provide a platform to explore quantum networking protocols on complex topologies.

The introduction reviews key concepts and prior work: quantum repeater architectures where nodes (collections of qubits) share links (entangled pairs), with entanglement swapping enabling long-distance connectivity. Hallmarks of complex networks (heavy-tailed degree distributions, small-world diameters) are summarized, with evidence of their benefits in the classical internet (robustness to random failures). For quantum networks, prior studies suggest robustness to random failures in complex topologies and that satellite-based quantum networks are small-world, reducing entanglement swaps and resource usage. However, fiber-based networks may not be small-world due to losses; complex networks may suffer from low density limiting throughput and vulnerability to targeted attacks. Most routing protocols assume regular lattice or ring structures, with limited work on arbitrary topologies. The authors position their work as a generative model to explore complex topologies arising from disordered spin systems (notably the 2D RTIM) where strong disorder produces GHZ spin clusters in the ground state, inspired in part by prior work modeling superconductivity using cluster-based network abstractions.

• Physical model: Two-dimensional random transverse-field Ising model (RTIM) on an L×L square lattice with periodic boundaries. Hamiltonian H = -∑⟨ij⟩ J_ij σ_i^x σ_j^x - ∑_i h_i σ_i^z, with ferromagnetic couplings J_ij ≥ 0 drawn i.i.d. from Uniform(0,1), and typically fixed transverse field h_i = h (fixed-h model). Critical control parameter θ ≈ 0.17034(2); universal properties near criticality under strong disorder.
• Ground state computation: Strong disorder renormalization group (SDRG), asymptotically exact near criticality, using an O(N log N) algorithm (N ≈ L^2). SDRG decimates largest local terms successively, yielding independent ferromagnetic clusters each in a GHZ state (|↑…↑⟩ + |↓…↓⟩)/√2. Lattices up to L = 4096, with at least 16 disorder instances per size.
• Node definition: Spatially localized, connected regions on the lattice selected as discretized circles with radii sampled from a power-law distribution p(r) ∝ r^{-γ_radius}, γ_radius = 2.67 (values between 2 and 3 give similar results). Minimum radius 2 lattice units. Placement via an outward-spiraling, tangent circle-packing heuristic to promote adjacency; circles added until 30% of lattice sites are covered. This yields nodes with a broad boundary-length distribution; by the RTIM area law, expected degree scales with boundary length.
• Link rule: Add a quantum link between node regions A and B if there exists a GHZ spin cluster with at least one spin in A and one spin in B, and no spins in any other node (allowing spins outside nodes to remain unmanipulated). This relaxes the stricter entanglement-negativity condition (which would forbid any cluster sites outside A∪B entirely) while retaining operational usability for communication assuming control operations only within nodes.
• Network extraction: If the constructed graph is disconnected, retain the largest connected component (LCC) as the quantum network for analysis.
• Benchmarks and comparisons: Construct grid-like benchmark networks by using uniformly sized nodes arranged in hexagonal packing (approximating a triangular grid) at the same overall coverage (30%). Compare topology to autonomous-systems-level classical internet snapshots (1997–2000).
• Topological analyses: Compute degree distributions; average shortest path length d vs network size N; degree correlations (average nearest-neighbor degree vs degree; Pearson correlation coefficient r across edges); global clustering coefficient C = 3×(triangles)/(connected triplets); local clustering coefficient C_local,i and its scaling with degree to test hierarchical modularity (C_local ~ k^{-1}).
• Variants: Repeat construction for off-critical fixed-h RTIM (vary θ), box-h RTIM (h_i ∼ Uniform(0,h)), and diluted RTIM (bond percolation with J_ij = J with probability p, else 0). For each, examine LCC size vs control parameter (θ or p) and qualitative topology.

• Emergent complexity: Networks constructed from GHZ clusters between heterogeneous circular nodes exhibit hallmark complex-network features comparable to the classical internet.
• Heavy-tailed degree distribution: The degree distribution follows a power law with exponent close to the node-radius distribution exponent (γ_radius = 2.67), consistent with the area law expectation. Averaged distributions across 16 instances show clear power-law behavior. Example network sizes: heterogeneous-node network with N ≈ 3.7×10^3 nodes (L = 4096), classical internet snapshot N ≈ 3.0×10^3.
• Near small-world behavior: Average shortest path length d increases much more slowly than √N (grid benchmark) and only somewhat faster than ln N, indicating nearly small-world scaling. Quantum networks have larger d than classical internet but far smaller than grid-like benchmarks.
• Disassortativity: High-degree nodes tend to connect to low-degree nodes (negative slope of nearest-neighbor degree vs degree). Degree correlation coefficient r < 0 across sizes, becoming less negative with increasing N, akin to classical internet behavior. Grid benchmarks are approximately neutral (r ≈ 0) asymptotically.
• Clustering: Global clustering coefficients for quantum networks lie between those of grid-like benchmarks (below ideal triangular lattice C = 0.4) and classical internet ranges. Local clustering decays with degree as C_local ~ k^{-1}, consistent with hierarchical modularity (Ravasz–Barabási model), paralleling classical internet patterns and contrasting with hubless grids.
• Robustness of construction across models: Off-critical fixed-h RTIM and box-h RTIM produce sizeable networks with similar complexity features; LCC size typically maximized near θ ≈ 0 for fixed-h and box-h. Diluted RTIM (bond percolation clusters) yields large, complex networks with maximal size far from criticality, showing criticality is not required—what matters is abundant local clusters.
• Generality: The approach should extend to other disordered spin systems with factorized GHZ-like ground-state clusters (e.g., random Potts, clock, Ashkin–Teller models).

The study demonstrates that interpreting GHZ spin clusters in disordered quantum spin systems as entanglement channels between spatially localized regions yields quantum networks with complex topology—heavy-tailed degree distributions, near small-world scaling, disassortative mixing, and hierarchical clustering—resembling the classical internet more than grid-based designs. This addresses the need for model platforms to explore how complex topology impacts quantum communication, beyond idealized regular lattices. The area law links node boundary size to degree, enabling scale-free networks via a broad node-size distribution while relying only on short-range cluster connectivity. The framework is flexible (applicable off-criticality and to other disorder classes), offering a generative testbed for evaluating routing, robustness, and capacity questions pertinent to an emerging quantum internet. Extensions such as multilayer constructions, weighted links (multiplicity of shared clusters), and hypergraph formulations (multipartite clusters) could further increase realism and enable protocol studies leveraging higher-order entanglement.

The authors introduce a generative method to build model complex quantum communication networks from spin clusters in disordered lattice systems, exemplified by the 2D RTIM. By selecting heterogeneous spatial nodes and linking pairs that share GHZ clusters confined to those nodes, the resulting graphs show complex-network features similar to the classical internet. The construction is robust beyond the critical RTIM and generalizable to other models with cluster-factorized ground states. Future work includes: exploring multilayer and weighted networks for denser connectivity; treating multi-node-spanning clusters as hyperedges to study multipartite entanglement distribution; extending to 3D (with degree distributions tied to surface areas) where nonplanarity may affect clustering; and experimental realizations of small-scale, many-node complex quantum networks in magnetic solids. These directions can clarify how network topology influences quantum communication performance and resilience.

• Pairwise-only abstraction: Clusters intersecting more than two node regions are discarded to maintain a simple graph, omitting potential multipartite entanglement that could be represented in a hypergraph.
• Relaxed link criterion: The operational link rule allows clusters with spins outside the two linked nodes (provided no other node contains them), diverging from strict logarithmic-negativity conditions; this may slightly alter entanglement structure compared to a stricter definition.
• 2D, quasi-planar setting: Analyses focus on 2D lattices with quasi-planar networks and a specific node placement heuristic (tangent circle packing), which is not uniformly random and may affect degree exponents and connectivity; results may differ in other geometries or placement strategies.
• Focus on topology: The study emphasizes structural metrics (degree, path length, clustering, assortativity) rather than end-to-end quantum capacities, error rates, or protocol performance under noise.
• Largest component only: Analyses consider the largest connected component, potentially overlooking fragmentation characteristics relevant to network reliability.
• SDRG assumptions: Reliance on SDRG (asymptotically exact near criticality) and specific disorder distributions; finite-size and off-critical effects may influence cluster statistics and thus network properties.