Quantum Repeater Networks Path Selection

• Path selection is the process of choosing optimal routes through quantum repeater networks to maximize end-to-end entanglement fidelity, secret-key rate, and throughput.
• Centralized and decentralized routing algorithms, such as quality-weighted Dijkstra and best-first search, are designed to tackle the non-isotonic and stochastic nature of quantum links.
• Multi-path routing and strategic upgrades of high-quality nodes enhance network performance by mitigating quantum memory decoherence and boosting entanglement rates.

Path selection in quantum repeater networks is the process of choosing routes through a graph of quantum repeaters and entangled links with the objective of maximizing performance metrics—most critically, end-to-end entanglement fidelity, secret-key rate, throughput, and operational reliability. Unlike classical networks, quantum path selection must address the stochastic, fragile, and non-Markovian nature of quantum state transmission, heterogeneous device efficiencies, probabilistic entanglement creation, Bell-state measurement errors, and quantum memory decoherence. The design of effective routing protocols requires integrating quantum information theory, network optimization, and physical-layer constraints, with a range of centralized and decentralized approaches supported by analytic bounds and large-scale simulation.

1. Quantum Network Models and Path Selection Objectives

Quantum repeater networks are formally described by undirected graphs $G=(V,E)$, where $V$ is the set of quantum repeater nodes and $E$ is the set of optical channels supporting both quantum and classical communication (Kumar et al., 2023). Each link $(u,v)\in E$ is characterized by physical parameters including photonic transmissivity, length, noise, and the ability to generate and store Bell pairs. Nodes are typically heterogeneous, with classifications such as high-quality (HQ) and low-quality (LQ) repeaters indexed by measurement efficiency $\eta$—the probability of correct Bell-state measurement (e.g., $\eta_h\approx0.999$, $\eta_l\approx0.8$).

Performance metrics guiding path selection include:

• End-to-end fidelity ($F_N$): Quality of the final shared entangled pair after a series of entanglement swapping operations, captured by $$F_N = \frac{1}{4} \left\{ 1 + 3 ⋅ \prod_{g=1}^G \left( \frac{4 η_g^2−1}{3} \right)^{N_g} ⋅ \left( \frac{4 F_0−1}{3} \right)^{N_g} ⋅ \left( \frac{4 F_0−1}{3} \right) \right\}$$ (Kumar et al., 2023).
• Secret-key rate (SKR): Product of entanglement-generation rate and the secret-key fraction, sensitive to Pauli error rates and often non-isotonic in path extensions (Tang et al., 25 Nov 2025).
• End-to-end capacity: Defined by channel and repeater physical characteristics, often reduced to minimizing the bottleneck value along the selected path or maximizing flow in multipath scenarios (Harney et al., 2022, Pirandola, 2016).

Path selection must further respect threshold constraints for fidelity or stability, memory decoherence rates, and quantum-classical communication latency.

2. Classical and Quantum Routing Algorithms

A key distinguishing factor of quantum networks is the non-additive and non-isotonic nature of quantum utility functions—adding a link to a path can decrease overall utility due to fidelity decay, probabilistic swaps, or decoherence (Tang et al., 25 Nov 2025). Classical algorithms such as Dijkstra’s shortest path are suitable only for additive link cost metrics (e.g., seconds per Bell pair of fidelity ≥$F$, $c_e(F)=1/R_e(F)$) and require “pre-equalization” of per-hop fidelity targets (Meter et al., 2012). For path selection based on multiplicative or non-isotonic metrics, specialized algorithms are required:

• Centralized Quality-Weighted Dijkstra: Allocates paths based on node (and/or link) weights that penalize LQ nodes, either minimizing hop count or maximizing overall expected fidelity. After serving a request, edges in the path are removed to prevent overlap (Kumar et al., 2023).
• Best-First Search (BeFS): Grows a priority queue of path prefixes ordered by a defined merit (e.g., admissible upper bound on SKR). Exact BeFS is globally optimal in swap-ASAP models, while heuristic BeFS trades solution quality for query-efficiency (Tang et al., 25 Nov 2025).
• Metaheuristic Algorithms: Simulated annealing (SA) and genetic algorithms (GA) efficiently approximate optimal paths when the utility function is non-isotonic or computational cost is critical (Tang et al., 25 Nov 2025).
• Decentralized Base-Graph Greedy Routing: Embeds the network in a lattice ($G^k$), where forwarding is performed locally to neighbors minimizing Manhattan distance to the destination. Path lengths scale as $O((\log n)^2)$ (Gyongyosi et al., 2018).
• Threshold-Adaptive Topology: Links and paths are filtered using critical stability thresholds, and shortest paths are found in the pruned overlay or base graph (Gyongyosi et al., 2018).

Summary of common algorithmic approaches for quantum path selection:

Algorithm Type	Applicability	Optimality
Weighted Dijkstra	Additive metrics, pre-equalized fidelity (Meter et al., 2012)	Yes (for additive, monotonic cost)
BeFS (Exact/Heur.)	Non-isotonic metrics, e.g. SKR (Tang et al., 25 Nov 2025)	Exact BeFS: optimal; Heuristic BeFS: near-optimal
Metaheuristics (SA/GA)	Arbitrary utility, complex networks (Tang et al., 25 Nov 2025)	Adjustable (solution quality vs. cost)
Decentralized Greedy	Local knowledge, base-graph (Gyongyosi et al., 2018)	Polylog path-lengths w.h.p.
Topology Adaption	Dynamic, threshold-based (Gyongyosi et al., 2018)	Efficient under link failure

3. Heterogeneous Efficiencies and Path Establishment Dynamics

Node and link heterogeneity critically shape path selection performance. In mixed-efficiency networks, the fraction $\xi$ of HQ nodes strongly determines the achievable end-to-end fidelity. Simulation studies demonstrate that below $\xi\approx0.32$–$0.36$ (grid/cylinder), average path fidelity is near zero regardless of routing, whereas pushing above $\xi\approx0.8$ yields $F_N>0.5$ on typical multi-hop paths (Kumar et al., 2023). Targeted upgrade of LQ nodes on bottleneck segments can yield abrupt fidelity improvements ($\sim$50%) at substantially lower cost than uniform blanket upgrades.

Path-establishment order further impacts resource contention and blocking rates: requests served early (low $\theta$) exploit the full network graph and available HQ nodes; later requests may be forced onto LQ-dominated paths with degraded fidelity or incur blocking. Efficiency-aware routing can reduce blocking probability by up to 30% compared to naive shortest-path methods under moderate fidelity thresholds.

A plausible implication is that network operators should maintain adaptive tiered services (high-fidelity for priority traffic, relaxed thresholds for bulk sessions) and dynamically protect key HQ nodes from early exhaustion.

4. Multi-Path Routing, Diversity, and Memory Constraints

Quantum networks can exploit multi-path routing to enhance entanglement rates, reduce latency, and mitigate probabilistic link failures (Pant et al., 2017, Milligen et al., 2023):

• Edge-Disjoint Paths: Multiplexing over $\theta$ shortest edge-disjoint paths achieves $R_\infty= p\sum_j q^{m_j-1}$ in the infinite-memory regime where $q$ is swap success and $m_j$ is path length.
• Time-Multiplexing: Repeaters generate links for $k$ time slots, then perform swaps. Increasing $k$ always improves entanglement rate until quantum memory decoherence induces a trade-off; optimal $k_{opt}\sim\mu/\tau$, where $\mu$ is mean memory lifetime and $\tau$ is slot duration (Milligen et al., 2023).
• Dynamic Local Protocols: In low-connectivity or suboptimal consumer placement, dynamic (distance-based) local routing outperforms static path assignment due to adaptive exploitation of available links.

Entanglement diversity—maintaining and using multiple prior entangled paths—enables selection of the fastest path or distillation for higher-fidelity pairs. Incorporation of prior entanglements (pre-existing Bell pairs on links) can allow a longer path with stored entanglement to outperform a fresh shorter path under certain probability regimes (e.g., low generation and high swap success rates) (Fayyaz et al., 4 May 2025).

5. Capacity Bounds, Routing Optimization, and Fundamental Limits

Ultimate routing performance is characterized by single-path and multi-path end-to-end capacity theorems. For distillable channels (pure-loss, quantum-limited amplifiers, dephasing, erasure), quantum network capacity equals the classical widest-path (single-path) or max-flow (multi-path) values:

• Single-Path (Widest Path): $$C_{single} = \max_{P} \min_{e\in P} I(\mathcal{E}_e^*)$$, with $I(\cdot)$ given by coherent or reverse-coherent information (Harney et al., 2022).
• Multi-Path (Flooding / Max-Flow): $$C_{multi} = \min_{C} \sum_{e\in \tilde C} I(\mathcal{E}_e^*)$$, i.e., minimum total capacity across any entanglement cut. Polynomial-time algorithms (modified Dijkstra, Edmonds-Karp, Orlin’s algorithm) (Pirandola, 2016, Harney et al., 2022).

Node-splitting techniques further generalize model fidelity, allowing routing optimization that incorporates internal loss and noise, with capacity bottlenecked by repeater sub-channel imperfections or storage losses (Harney et al., 2022).

For non-isotonic metrics such as secret-key rate, standard Dijkstra fails to guarantee optimality. Destination-aware best-first search and metaheuristics recover practical optimality at tractable cost even for networks up to 100 nodes (Tang et al., 25 Nov 2025).

6. Practical Guidelines and Deployment Strategies

Empirical and analytic studies inform several practical recommendations for quantum repeater network path selection:

• Upgrade only $\xi\sim0.8$–$0.9$ of repeaters to HQ to reach $F_N>0.5$–$0.6$ with diminishing returns for further upgrades (Kumar et al., 2023).
• Prefer strategic bottleneck upgrades over uniform improvements for cost efficiency.
• Adopt efficiency-aware weight maps in centralized schedulers for improved fidelity and lower blocking under moderate thresholds.
• Reserve strict fidelity guarantees for early, high-priority requests and allocate remaining resources to best-effort traffic.
• Use time-multiplexing with $k_{opt}\sim\mu/\tau$ to balance rate and decoherence; adjust $k$ according to network memory and link success rates (Milligen et al., 2023).
• Prioritize path selection policies that include prior stored entanglements to exploit temporal network dynamics (Fayyaz et al., 4 May 2025).
• In decentralized architectures, base-graph greedy forwarding with matched entanglement probabilities achieves $O((\log n)^2)$ path-lengths and scalable operation (Gyongyosi et al., 2018).
• Integrate multi-path and diversity routing for optimal rate and reliability, with model-driven thresholds to adapt to dynamic conditions.

7. Outstanding Issues and Future Directions

Quantum routing research continues to address network scaling, robustness to dynamic failures, multi-user resource contention, and the integration of quantum-classical control planes. Entanglement assignment is NP-complete, typically tackled by polynomial-time algorithms scoped by local node degree (Gyongyosi et al., 2018). Multipath and diversity strategies are essential for robust performance, especially as quantum networks transition to larger topologies and incorporate heterogeneous device technologies. Further examination of utility function properties, e.g., non-isotonicity, remains critical for advancing routing algorithms beyond classical analogs (Tang et al., 25 Nov 2025).

References: Routing in Quantum Repeater Networks with Mixed Efficiency Figures (Kumar et al., 2023) Routing in Non-Isotonic Quantum Networks (Tang et al., 25 Nov 2025) Path Selection for Quantum Repeater Networks (Meter et al., 2012) End-to-End Capacities of Imperfect-Repeater Quantum Networks (Harney et al., 2022) Routing entanglement in the quantum internet (Pant et al., 2017) Decentralized Base-Graph Routing for the Quantum Internet (Gyongyosi et al., 2018) Topology Adaption for the Quantum Internet (Gyongyosi et al., 2018) On Selecting Paths for End-to-End Entanglement Creation in Quantum Networks (Fayyaz et al., 4 May 2025) Entanglement Routing over Networks with Time Multiplexed Repeaters (Milligen et al., 2023) Capacities of repeater-assisted quantum communications (Pirandola, 2016)