Emulation of Entanglement Distribution Networks on a Quantum Computer

Ashley N. Tittelbaugh; Boulat A. Bash; Daniel C. Kilper; Jerry Horgan; Marco Ruffini; Rohan Bali; Shelbi L. Jenkins

REVIEW 3 major objections 3 minor 14 references

Emulation of Entanglement Distribution Networks on a Quantum Computer

T0 review · 3 major / 3 minor · reviewed 2026-08-02 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Mathematically identical ways of modeling noise in a quantum-network emulation produce different results on real hardware, and the quasi-probability implementation is the only one that stays close to ideal.

desk verdict Useful empirical caution about emulating network noise on hardware, but the 'profound differences' claim is undercut by calibration drift and one broken implementation; worth refereeing if the confound is addressed. read the letter →

arxiv 2607.14260 v2 pith:MWDKPM6I submitted 2026-07-15 quant-ph cs.NI

Ashley N. Tittelbaugh , Jerry Horgan , Rohan Bali , Marco Ruffini , Daniel C. Kilper , Shelbi L. Jenkins , Boulat A. Bash This is my paper

classification quant-phcs.NI MSC 81P6881P4081P45 PACS 03.67.-a03.67.Hk

keywords distributedquantumcomputingentanglementdistributioncutBellpairquasi-probabilitydecompositiondepolarizingchannelgraphstatesteleportationclassicalcommunicationdelay

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper uses a single quantum computer to emulate a distributed quantum network: it builds a ring graph state in which one edge is created through a teleported cut Bell pair instead of a physical connection. The authors model imperfect entanglement sources with three depolarizing channels that are mathematically equivalent (a Stinespring-dilated unitary, random Pauli errors, and a quasi-probability decomposition) and compare them in simulation and on hardware. The central finding is that the equivalence holds only in theory: hardware constraints flip the comparison, with the quasi-probability method performing best because it adds no gates, while Pauli-error depolarization fails to act as a true depolarizing channel. The paper also reports that successful graph-state creation on hardware requires pre-distributed entanglement fidelity of at least 90%, and that classical communication latency only becomes detrimental at tens of kilometers. These results matter because they tell algorithm designers which noise-emulation choices are trustworthy when developing distributed quantum computing protocols on today's devices.

What carries the argument

The load-bearing object is the cut Bell pair: a Bell state |Φ+⟩ written as a quasi-probability distribution (QPD) over separable states ρ+_k and ρ−_k with weights summing to one but taking negative values. Each cut pair is used to implement a teleportation-based virtual controlled-Z gate across a previously disconnected edge of a six-qubit linear graph state, turning it into a ring. Depolarizing noise is applied to the cut Bell pair in three equivalent forms—a unitary Stinespring dilation, randomly applied Pauli gates, and the QPD itself—and the witness-based fidelities of the resulting graph state are measured using 12 observable circuits combined over QPD terms (60–84 runs total). The mach

What would settle it

Run all three depolarization implementations interleaved within a single calibration snapshot (or averaged over many calibration cycles) on the same hardware and check whether the quasi-probability-vs-Pauli fidelity gap persists; if the gap disappears or shrinks below shot noise, the paper's 'profound differences' conclusion is an artifact of calibration timing rather than the method.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that the three depolarizing implementations are not interchangeable on a physical quantum processor. Despite being mathematically equivalent channels, the unitary, Pauli, and quasi-probability versions diverge because hardware transpilation adds different numbers of gates, changes qubit maps, and introduces calibration drift between runs. The quasi-probability decomposition, which adds no gates, consistently matches simulation and the ideal no-noise case, whereas the Pauli implementation requires extra Hadamard gates that break the depolarizing property, producing fidelities below the entanglement threshold. The paper directly claims that 'the more err

Load-bearing premise

The central comparison assumes that differences between the three depolarization methods are not caused by calibration drift between the separate hardware runs, even though the paper acknowledges 'the noise model may differ slightly between methods.'

Editorial extensions

If this is right

If the results hold, quantum-network emulation on current hardware should use gate-free depolarization (QPD) to avoid implementation artifacts.
Network-distributed entanglement fidelity must be ≥90% on near-term hardware for a usable teleportation-built graph state, versus only 60% in an ideal-noise simulation.
Data-center-scale interconnects (meters of fiber, ≤100 ns) are safe from classical-communication-induced dephasing; effects only appear at hundreds of meters and break down at tens of kilometers.
The directional asymmetry of the virtual CZ (control qubit 5, target qubit 0) consistently degrades the target edge more, a feature future distributed-graph-state protocols must budget for.
The framework extends a single-QPU cut-Bell-pair construction to a hardware benchmark, establishing a workflow for testing future network-enabled distributed quantum computing algorithms.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the paper's hardware-vs-simulation divergence is general, then any noise-model comparison in quantum-network emulation should be conducted under a single calibration snapshot or averaged over many, since the paper's own runs used separate snapshots that could bias the 'profound differences' conclusion.
The discrepancy between the Pauli and quasi-probability methods on hardware could be repurposed as a probe of transpilation-induced error: the extra gate depth is measurable, and process tomography of the effective channel would test whether it is truly non-depolarizing.
A natural testable extension is to repeat the fidelity sweep with standard error-mitigation techniques; if the 90% threshold drops, the requirement is a hardware-noise artifact rather than a fundamental property of teleportation-based graph-state construction.
The paper's latency results suggest that classical-communication delays are not the bottleneck for data-center-scale distributed quantum computing; the more pressing constraint is the quality of pre-distributed entanglement, shifting research attention to source fidelity and purification.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: the paper has a real and useful observation—that on actual hardware you cannot treat three 'equivalent' depolarizing implementations as interchangeable—but the headline 'profound differences' claim is weaker than it looks because the comparison runs were taken under different calibration snapshots and one of the methods stops being a depolarizing channel after transpilation.

What's actually new: this is the first side-by-side evaluation I know of that puts QPD, Pauli, and (in simulation) unitary depolarization through the same cut-Bell-pair graph-state emulation pipeline. The measured result that IQM Emerald needs network-distributed fidelity ≥90% to produce a ring graph state above the 0.5 witness threshold is a concrete data point for anyone building near-term DQC emulations. The delay sweep result—no appreciable degradation until ~1000 ns, failure only at tens of km—is also useful for data-center-scale planning.

What the paper does well: the experimental design is honestly documented. They tell you that each method gets a fresh calibration snapshot, that the Pauli transpilation no longer realizes a true depolarizing channel, and that transient hardware noise influences results. That transparency is worth preserving.

Soft spots, in order: (1) The central comparison is confounded. Separate calibration snapshots per method mean drift, not implementation, could drive the QPD-vs-Pauli spread in Fig. 4. The authors themselves point this out in Sections 3.1 and 4.1 and in the conclusion, so the abstract's 'profound differences' is not established. (2) The Pauli method after transpilation is not a depolarizing channel; comparing it to QPD is therefore comparing a broken implementation to a clean one. (3) QPD performing best is partly by construction since it introduces no gates; that's fine as an implementation note, but it isn't a discovery about the noise models. (4) No code/data released, and thermal relaxation parameters for the delay sweep aren't reported, so the thresholds are hard to reproduce.

None of these are fatal to the paper's usefulness as a cautionary empirical study. The right fix is to run methods with the same calibration snapshot or to explicitly sweep calibration, and to release the artifacts. As written, it deserves a serious referee, mainly because the confound is addressable and the measurement is freshly useful.

I'd want to see the revision before citing it.

Referee Report

3 major / 3 minor

Summary. The paper demonstrates a framework for emulating a network-distributed entanglement link on a single quantum processor. It realizes a virtual CZ gate and a ring graph-state edge using a cut Bell pair, following Carrera Vazquez et al., and compares three mathematically equivalent depolarizing-channel implementations (Stinespring unitary, random Pauli, and quasi-probability decomposition). The methods are benchmarked in simulation on an IBM Torino noise model, with the Pauli and QPD variants also run on IQM Emerald hardware. The paper sweeps the input Bell-pair fidelity and the classical communication delay, reporting lower-bound graph-state fidelities from entanglement witnesses. The central claim is that, despite mathematical equivalence of the noise models, hardware constraints cause 'profound differences' in results, with QPD depolarization performing best, while Pauli depolarization degrades significantly after transpilation. The paper also reports that on hardware, successful graph-state creation requires network-distributed fidelity of at least 90%, and that classical delays only become detrimental at tens of kilometers.

Significance. If the central claim were fully established, the paper would provide a useful, cautionary message for distributed-quantum-computing emulation: the choice of how to implement an abstract noise model can materially change conclusions drawn from hardware experiments, and QPD offers a low-gate-overhead alternative. The paper is also valuable for its concrete 60/84-circuit measurement workflow, its use of witness-based fidelity lower bounds, and its transparent admission of limitations. However, the main comparative claim is currently underdetermined by the reported data because of calibration-snapshot confounding and because the Pauli implementation is admitted not to realize a depolarizing channel after transpilation. These issues are load-bearing: without controlling them, the 'profound differences' headline reduces to a comparison of uncompiled vs. compiled operations under differing hardware conditions. The delay-sweep conclusion is also not reproducible because the thermal-relaxation parameters are not reported. The paper's significance is therefore conditional on additional controlled experiments or substantially weakened claims.

major comments (3)

[Sec. 3.1 and Sec. 4.1] The comparison of depolarization methods is confounded by the use of independent calibration snapshots. Sec. 3.1 states that 'separate runs for different depolarization methods each retrieve an independent calibration snapshot,' and Sec. 4.1 repeats that a depolarized run can outperform ideal because each run receives 'a fresh calibration snapshot.' The conclusion then admits 'the results here are influenced by transient hardware noise.' Since the QPD and Pauli runs are executed at different times, the between-method spread in Fig. 4 could be dominated by calibration drift rather than by the depolarization implementation. This directly undermines the central 'profound differences' claim. A control experiment with interleaved runs under a single calibration snapshot, or a paired statistical comparison with reported calibration data, is needed.
[Sec. 4.2] The Pauli 'depolarizing' implementation on hardware is no longer a depolarizing channel after transpilation. Sec. 4.2 explains that the required Hadamard operations create unequal gate counts and accumulated error across Pauli operators, so 'the implemented operation therefore no longer represents a true depolarizing channel.' Thus Fig. 4 does not compare two mathematically equivalent noise models: it compares QPD depolarization with an arbitrarily compiled Pauli circuit that has a different effective channel. The abstract's claim that 'mathematically equivalent' implementations diverge is therefore not supported by the hardware data as presented. Either the Pauli implementation should be compiled in a way that preserves the depolarizing structure, or the implemented channel should be characterized (e.g., by process tomography) and the claim reframed.
[Sec. 4.4 and Sec. 3.1] The classical-communication-delay sweep is not reproducible because the thermal-relaxation parameters T1 and T2 used in the simulation are not reported. Sec. 3.1 says that classical delays are modeled 'using thermal relaxation' and Sec. 4.4 claims delays become detrimental only at 'tens of kilometers.' This threshold depends exponentially on the assumed T1/T2 values and on the per-gate timing model. Without these parameters, the quantitative conclusion—and even the order-of-magnitude threshold—cannot be verified or compared with other hardware models. Please report the T1/T2 values, the gate-duration model, and the mapping from delay to thermal-relaxation error.

minor comments (3)

[Sec. 4.3 / Fig. 5] The text says that on hardware 'successful graph-state creation' occurs at network-distributed fidelities of 0.9 and 1, but Fig. 5b apparently shows only data points at those values. Clarify whether 1.0 is a physically meaningful input or a normalization point, and whether the statement refers to the lowest successful point or to all points above threshold.
[Table 2 / Sec. 3] The relationship between 'Runs per Observable' in Table 2 (0 for ideal/Pauli, 2 for QPD) and the total circuit counts (60 vs. 84) is not immediately clear. Define the counting convention explicitly, including how the 12 observables combine with QPD sampling terms.
[Sec. 3.2 / Sec. 4.2] The simulation uses an IBM Torino noise model while the hardware experiments use IQM Emerald. The architectural mismatch is acknowledged, but the paper would benefit from a table comparing gate sets, error rates, and connectivity for the two platforms, since the central claims depend on hardware-specific behavior.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central hardware comparison is an empirical measurement, and the self-citation to [13] is descriptive rather than load-bearing.

full rationale

The paper's central claim is an experimental comparison of three implementations of the same depolarizing channel. The equivalence of the channel representations (Stinespring dilation, random Pauli errors, and quasi-probability decomposition) is a standard mathematical fact, and the cut-Bell-pair decomposition is imported from external references [4] and [14], not from the authors' own prior work. The self-citation to [13] (the first author's thesis) appears repeatedly as the source of implementation details ('These methods are described in depth in [13]'; 'Further details on the QPD construction, depolarization methods applied to the cut Bell pairs, and the entanglement witness formalism used in our evaluation are given in [13]'), but it is not used to justify the central claim; the performance differences are measured directly in simulation and on hardware. The conclusion's statement that 'QPD ... performed the best ... is expected, as QPD introduces no additional gates' does not claim a derived prediction; it explicitly labels the advantage as expected, and the 'more significant finding' is the empirical hardware divergence. Section 4.2's admission that the Pauli operation 'no longer represents a true depolarizing channel' after transpilation and the conclusion's admission that 'the results here are influenced by transient hardware noise' are validity/confounding concerns for the hardware comparison, not circular reductions: they do not show that any output is equivalent to an input by construction. No equation in the paper is fitted to the quantity it is used to predict. Therefore no significant circularity is present.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard mathematical identities (QPD combination, depolarizing channel equivalence) and on device-noise assumptions. The free parameters are the unstated thermal-relaxation values and depolarizing strength for the 99.6% point; neither is derived in the text. No new physical entities are introduced.

free parameters (2)

Thermal relaxation time constants (T1/T2) for simulated qubits
Used to model classical communication delay in simulation (Sec. 3.1, Sec. 4.4). The values are not reported; the quantitative delay thresholds depend on them.
Depolarizing noise strength p for 99.6% input fidelity
The paper states it models 99.6% ZALM fidelity but does not give the mapping from fidelity to the depolarizing parameter for the three methods; circuits are ambiguous.

assumptions (4)

domain assumption Depolarizing channel equivalence: Stinespring dilation, Pauli randomization, and quasi-probability decomposition implement the same channel in the absence of hardware errors.
Abstract and Sec. 5; this is the premise of the central comparison.
domain assumption Thermal relaxation dominates during classical communication delay; other idle errors are negligible.
Sec. 3.1: 'classical communication delays are the only point where we apply thermal relaxation, as we assume the time is dominated by classical communication.'
standard math The Vidal–Tarrach decomposition (Eq. 1) with the QPD weights in Table 1 reproduces Bell-pair measurement statistics.
Eq. (1) and Table 1; prior literature [4,14].
domain assumption Entanglement witnesses (Jungnitsch et al.) give a valid lower bound on bipartite fidelity; fidelity > 0.5 defines successful graph-state creation.
Sec. 4.3 uses the 0.5 threshold for success; from [8].

how reviews work

0 comments

Cite this review

Pith. "Pith review of Emulation of Entanglement Distribution Networks on a Quantum Computer." pith.science (2026). https://pith.science/paper/MWDKPM6I

@misc{pith2026260714260,
  author       = {Pith},
  title        = {Pith review of: Emulation of Entanglement Distribution Networks on a Quantum Computer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MWDKPM6I}},
  note         = {Machine review of arXiv:2607.14260}
}

read the original abstract

We investigate how quantum computers can be used to emulate quantum networks and study their performance under practical impairments. In particular, we evaluate how degraded entanglement and communication latency affect teleportation-based distributed multipartite-entanglement-state construction. We model imperfect Bell-pair sources using depolarizing noise channels and classical communication delays using thermal relaxation. We implement the depolarization using Stinespring dilation, randomly applied Pauli errors, and quasi-probability decompositions, evaluating the latter two on IQM quantum hardware and all three in simulation. We then study the performance of the entanglement distribution under noise generated by the aforementioned models. Although these noise models are mathematically equivalent, we find that hardware constraints result in profound differences in the corresponding results, highlighting the importance of careful experiment design.

Figures

Figures reproduced from arXiv: 2607.14260 by the authors.

**Figure 2.** Cut graph-state circuit The base form of the cut graph state circuit runs in all experiments. given in [4]. Thus, for a single set of input parameters, a total of 60 circuits are executed (or 84 when using QPD-based depolarization [13]). This number could potentially be reduced by measuring commuting stabilizers simultaneously, at the cost of more complex post-processing. We evaluate the efficacy of each of the thre… view at source ↗

**Figure 3.** Simulation at 99.6% ZALM fidelity and 10 ns classical communication Individual graph state stabilizer values and a lower-bound for pair-wise graph state fidelities. For both stabilizers and fidelity, a unity value indicates perfect entanglement with no errors. Fig. 3b further illustrates that the lower-bounded fidelity across the cut edge (0,5) is the smallest, while the lower-bounded fidelities of neighboring edge… view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: 99.6% ZALM fidelity on hardware Individual graph state stabilizer values and a lower-bound for pair-wise graph state fidelities. For both stabilizers and fidelity, a unity value indicates perfect entanglement with no errors. evaluate both the QPD and Pauli depolarizing…

**Figure 5.** Figure 5: Fidelity sweep across depolarization methods Subfigures (a)–(b) show the lower-bounded graph state fidelities under QPD depolarization for simulation (10 ns delay) and hardware, respectively. Subfigure (c) shows the Pauli depolarization simulation, and subfigure (d) s…

**Figure 6.** Figure 6: Classical delay sweep Logarithmic simulated classical delay sweep assuming a 99.6% network distributed fidelity [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

14 extracted references · 1 linked inside Pith

[1]

arXiv preprint arXiv:2410.00916 (2024)

AbuGhanem, M.: Ibm quantum computers: evolution, performance, and future directions. arXiv preprint arXiv:2410.00916 (2024)

arXiv 2024
[2]

Amazon Web Services: Amazon braket launches new 54-qubit supercon- ducting quantum processor from iqm. AWS Quantum Technologies Blog (July 2025), https://aws.amazon.com/blogs/quantum-computing/amazon-braket- launches-new-54-qubit-superconducting-quantum-processor-from-iqm/, accessed: 2026-07-08

2025
[3]

IEEE Trans

Brenner, L., Piveteau, C., Sutter, D.: Optimal wire cutting with clas- sical communication. IEEE Trans. Inf. Theory71(10), 7742–7752 (2025). https://doi.org/10.1109/TIT.2025.3601047

arXiv 2025
[4]

Na- ture636(8041), 75–79 (Dec 2024)

Carrera Vazquez, A., Tornow, C., Ristè, D., Woerner, S., Takita, M., Egger, D.J.: Combining quantum processors with real-time classical communication. Na- ture636(8041), 75–79 (Dec 2024). https://doi.org/10.1038/s41586-024-08178-2, https://doi.org/10.1038/s41586-024-08178-2

doi:10.1038/s41586-024-08178-2 2024
[5]

Physical Review Applied (2023)

Chen, K.C., Dhara, P., Heuck, M., Lee, Y., Dai, W., Guha, S., Englund, D.: Zero- added-loss entangled photon multiplexing for ground- and space-based quantum networks. Physical Review Applied (2023)

2023
[6]

IEEE Trans

Fan, X., Zhan, C., Gupta, H., Ramakrishnan, C.: Optimized distribution of entan- glement graph states in quantum networks. IEEE Trans. Quantum Eng. (TQE) (2025)

2025
[7]

Hofmann, H.F.: How to simulate a universal quantum computer using negative probabilities. J. Phys. A: Math. Theor.42(27), 275304 (2009)

2009
[8]

Jungnitsch, B., Moroder, T., Gühne, O.: Entanglement witnesses for graph states: General theory and examples. Phys. Rev. A84(3), 032310 (2011)

2011

Show all 14 references

[9]

Mitarai, K., Fujii, K.: Constructing a virtual two-qubit gate by sampling single- qubit operations. New J. Phys.23(2), 023021 (2021)

2021
[10]

Peng, T., Harrow, A.W., Ozols, M., Wu, X.: Simulating large quantum circuits on a small quantum computer. Phys. Rev. Lett.125(15), 150504 (2020)

2020
[11]

IEEE Trans

Piveteau, C., Sutter, D.: Circuit knitting with classical communication. IEEE Trans. Inf. Theory70(4), 2734–2745 (2023)

2023
[12]

Steane, A.: Multiple-particle interference and quantum error correction. Proc. R. Soc. Lond. A Math. Phys. Eng. Sci.452(1954), 2551–2577 (1996)

1954
[13]

Master’s thesis, The University of Arizona

Tittelbaugh, A.: Modeling and Quantum-Hardware-Based Investigation of Entan- glement Distribution and Classical Communication in Distributed Quantum Com- puting. Master’s thesis, The University of Arizona. (2025)

2025
[14]

Vidal, G., Tarrach, R.: Robustness of entanglement. Phys. Rev. A59(1), 141 (1999)

1999

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.

[1] [1]

arXiv preprint arXiv:2410.00916 (2024)

AbuGhanem, M.: Ibm quantum computers: evolution, performance, and future directions. arXiv preprint arXiv:2410.00916 (2024)

arXiv 2024

[2] [2]

Amazon Web Services: Amazon braket launches new 54-qubit supercon- ducting quantum processor from iqm. AWS Quantum Technologies Blog (July 2025), https://aws.amazon.com/blogs/quantum-computing/amazon-braket- launches-new-54-qubit-superconducting-quantum-processor-from-iqm/, accessed: 2026-07-08

2025

[3] [3]

IEEE Trans

Brenner, L., Piveteau, C., Sutter, D.: Optimal wire cutting with clas- sical communication. IEEE Trans. Inf. Theory71(10), 7742–7752 (2025). https://doi.org/10.1109/TIT.2025.3601047

arXiv 2025

[4] [4]

Na- ture636(8041), 75–79 (Dec 2024)

Carrera Vazquez, A., Tornow, C., Ristè, D., Woerner, S., Takita, M., Egger, D.J.: Combining quantum processors with real-time classical communication. Na- ture636(8041), 75–79 (Dec 2024). https://doi.org/10.1038/s41586-024-08178-2, https://doi.org/10.1038/s41586-024-08178-2

doi:10.1038/s41586-024-08178-2 2024

[5] [5]

Physical Review Applied (2023)

Chen, K.C., Dhara, P., Heuck, M., Lee, Y., Dai, W., Guha, S., Englund, D.: Zero- added-loss entangled photon multiplexing for ground- and space-based quantum networks. Physical Review Applied (2023)

2023

[6] [6]

IEEE Trans

Fan, X., Zhan, C., Gupta, H., Ramakrishnan, C.: Optimized distribution of entan- glement graph states in quantum networks. IEEE Trans. Quantum Eng. (TQE) (2025)

2025

[7] [7]

Hofmann, H.F.: How to simulate a universal quantum computer using negative probabilities. J. Phys. A: Math. Theor.42(27), 275304 (2009)

2009

[8] [8]

Jungnitsch, B., Moroder, T., Gühne, O.: Entanglement witnesses for graph states: General theory and examples. Phys. Rev. A84(3), 032310 (2011)

2011