Cross-Validation of Open-Source Quantum Network Simulators

AI summary

Overview

Research area: Quantum networking, specifically the validation and reproducibility of discrete-event quantum network simulators. Technical level: Intermediate — the paper includes a tutorial-style preliminaries section on entanglement generation, swapping and purification, but assumes familiarity with quantum repeaters and simulation. Scope: A cross-validation study that runs matched entanglement-distribution experiments on two open-source simulators, QuISP and SeQUeNCe, comparing predicted timing and fidelity of the distributed quantum states.

What This Paper Is About

Quantum networks are too complex to analyze purely on paper, so researchers rely on simulators to predict how protocols and hardware will behave. Different simulators make different design choices and simplifications, so their outputs may not agree even for the same nominal scenario — and it is hard to tell whether a disagreement reflects a real design difference, a legitimate modeling simplification, or a bug. This paper runs a common set of basic networking tasks in QuISP and SeQUeNCe and compares how long the tasks take and how good the resulting entanglement is.

Key Contributions

1. A four-experiment cross-validation benchmark of increasing complexity between QuISP and SeQUeNCe: a symmetric memory-interference-memory (MIM) link, an asymmetric MIM link, two-link entanglement swapping, and one round of BBPSSW entanglement purification.
2. Explicit theoretical timing models for both simulators, covering connection setup, classical message exchange, photon-train emission, and — for QuISP's purification case — a discrete-time Markov chain model of memory states, plus closed-form expected-time expressions for SeQUeNCe.
3. Identification and quantification of a constant-factor timing offset between the simulators, traced to their different connection models: QuISP negotiates a connection once and then streams photon trains, while SeQUeNCe re-negotiates for every Bell pair with a three-way handshake.
4. A demonstrated path to reproducibility: the exercise led to numerous fixes in both simulators and establishes a foundational task set for cross-validating other quantum network simulators.

Main Findings

• Timing differs by a constant factor: the two simulators predict different times to complete the same network tasks. The paper attributes this to their respective connection models and states that the offset is explained and correctly quantified using the design differences.
• Fidelity agrees: under identical error parameters, both simulators predict the same fidelity for the distributed resources, indicating that both implement realistic error models correctly.
• Connection setup is the root cause of the timing gap: QuISP's connection setup takes (2L + d)/c and is performed once, after which rounds are described by 2T_BSA + 10t_sep + (N_mem − 1)t_sep; SeQUeNCe performs a three-way handshake for each Bell pair, giving a round time of 4L/c.
• Purification timing differs qualitatively: QuISP frees measured memories immediately after a purification attempt, even before the purification result reaches both nodes, so generation and purification can overlap; SeQUeNCe keeps both measured and unmeasured memories locked until the purification result is exchanged, so the two phases never overlap.
• Real bugs were found: the cross-validation exercise "resulted in numerous fixes to both simulators."
• Specific measured values are not in the provided excerpt: the supplied content is truncated before Section VI (Results), so the numerical times, fidelities, throughputs and purification success probabilities obtained in the four experiments are not reported here.

Methodology in Plain English

Both simulators use discrete-event simulation: the system is modeled as a queue of timestamped events, and nothing changes between events. The authors designed four experiments that isolate one effect at a time and ran them in both tools.

• Experiment 1 (symmetric link): two nodes 20 km apart with the Bell-state analyzer at 10 km, varying the number of memories per node over 1, 2, 4, 8 and 16, and timing how long it takes to produce 1000 Bell pairs.
• Experiment 2 (asymmetric link): the same 20 km link but with the analyzer moved from 10 km to 20 km in 1 km steps, with one memory per node, again timing 1000 Bell pairs.
• Experiment 3 (entanglement swapping): 40 km end-to-end with a repeater in the middle and analyzers at 10 km, measuring the fidelity of the swapped pair while sweeping CNOT error and measurement error from 0 to 1 and using memory coherence times of 18 ms, 55 ms and infinity.
• Experiment 4 (purification): a 40 km link with the analyzer in the middle and two memories per node, running one round of BBPSSW (without the protocol's twirling step), sweeping initial fidelity from 0.6 to 0.9 in 0.05 steps and using the same three coherence times, and comparing throughput, output fidelity and purification success probability.

Shared settings across the experiments are 0.2 dB/km fiber loss, a 1 GHz repetition rate, and a speed of light in fiber of 2 × 10^5 km/s. Before running the experiments, the authors derived analytical expressions for expected completion times in each simulator so that the measured outputs could be checked against theory rather than only against each other.

Why This Matters

Simulators sit between mathematical analysis and real hardware; if two independently developed simulators disagree, users cannot tell which prediction to trust, and if they agree, that is meaningful evidence. The paper argues that analytical verification only works for basic cases, and real experiments are scarce, so cross-validation between simulators is the practical route to confidence. The work also produced concrete bug fixes and a reusable benchmark methodology.

Real-world applications of the underlying technology named in the paper:

• Quantum key distribution over fiber and free space.
• Enhanced interferometric measurements for telescopes.
• High-precision clock synchronization.
• Blind quantum computation and distributed quantum computation.
• A future quantum internet linking quantum processors across the globe.

Industry relevance: network architects and protocol designers use these simulators to make design decisions for real deployments, so an unexplained factor difference in predicted latency directly affects planning. QuISP is explicitly aimed at internetwork-scale studies (the stated goal is 100 networks of 100 nodes each on minimal resources), while SeQUeNCe targets high-fidelity, highly customizable photonic modeling — making tool choice a practical engineering decision that this comparison informs.

Future Directions

• Scale the benchmark beyond two simulators: the paper lists NetSquid, QuantumSavory, QuNetSim, QuNet, ReQuSim and Q2NS as candidates, and frames its methodology as a first step toward cross-validation across all quantum network simulators.
• Move past first-generation repeaters: both simulators currently model 1G repeater networks (entanglement distribution followed by swapping without error correction), leaving higher-generation architectures untested.
• Full protocol development: the authors describe the work as leading toward complete protocol implementations rather than isolated primitives, and note that the BBPSSW twirling step was not simulated.
• Reconcile the remaining modeling gaps: the differing treatment of memory locking during purification and of per-pair connection setup are identified but not unified, leaving open the question of which model better reflects real hardware.

Target Audience

Quantum networking researchers and graduate students who use simulation to study entanglement distribution protocols; developers and maintainers of quantum network simulators who want a template for cross-validation; and network architects or program managers who need to choose a simulator and understand the provenance of its predictions. The paper's step-by-step timing derivations also make it useful to readers who want a worked example of how classical control messaging dominates latency in first-generation quantum repeater networks.

Authors’ abstract

We present a comparative analysis of two open-source quantum network simulators, QuISP and SeQUeNCe, focusing on cross-validation of basic networking tasks to ensure consistency and accuracy in simulation outputs. Although both simulators are designed to study quantum network management policies and their dynamics under complex noise models, their differing underlying assumptions can lead to variations in simulation results. We highlight the discrepancies in how the two simulators handle connections, internal network node processing time, and classical communication, resulting in significant quantitative and qualitative differences in the time required to perform basic network tasks such as entanglement generation, purification, and swapping. We simulate common scenarios between the simulators to compare both the latency of resource distribution and the fidelity of the distributed resources. Our findings indicate that while the simulators differ in the time required to complete network tasks---a constant factor difference attributable to their respective connection models---they agree on the fidelity of the distributed resources under identical error parameters. This work demonstrates a crucial first step towards enhancing the reliability and reproducibility of quantum network simulations, as well as leading to full protocol development. Furthermore, our benchmarking methodology establishes a foundational set of tasks for the cross-validation of simulators to study future quantum networks.