{"id":"d78993c6-9d90-424b-a629-e57b4bbe20b3","arxiv_id":"2412.03738","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A cascade-interferometer scheme estimates the parameters of a wrapped-Gaussian phase-correlation model for gain-switched lasers, enabling computation of the security parameter q for arbitrary correlation lengths in decoy-state QKD.","lead":"This paper designs interferometer setups that measure how strongly the phase of a laser pulse depends on the phases of several earlier pulses, a correlation that can weaken quantum key distribution security. The method turns the measurement into an optimization over optical attenuators and phase shifters, yielding the security parameter required by a recent proof.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The method's q estimate is not robust to model mismatch: Assumption A4 fixes a phase-independent wrapped-Gaussian noise, while Appendix A concedes the conditional variance may depend on phase history; the measured visibility only fixes an average, so q may be overestimated.","rationale":"The reader's verdict is CONDITIONAL with moderate confidence, and I identify the same load-bearing weakness. The paper's contribution is a coherent extension of first-order visibility to arbitrary lc and a plausible interferometric reconstruction; Eqs. (C17)-(C18) are valid conditional on A4. My stress-test does not reveal an internal inconsistency in the derivation: the algebra of Eq. (24) follows from the stated model, and the reduction to a maximization over attenuators is sensible. The central claim, however, is stronger than what the evidence supports: it asserts the scheme certifies the security parameter for real gain-switched sources. That certification requires the true conditional phase distribution to be wrapped Gaussian with a constant variance; Appendix A concedes this may fail. The generalized visibility measures only the first circular moment of delta_phi_i, and for non-Gaussian or history-dependent noise infinitely many distributions share that moment while having different minimum densities, hence different q. The numerical simulation is the natural place to test this, yet it does not report a comparison between the estimated q and the empirical q computed from the simulated phase trajectories. Therefore a model-mismatch check is required before the scheme can be used as a certificate. This does not overturn the paper: under A4 the method stands, and the simulation trends are plausible. It does mean the claimed applicability to 'any gain-switched source' is not yet established, so the CONDITIONAL verdict is appropriate and should be retained.","tokens_in":21258,"tokens_out":7569,"duration_ms":77027,"concrete_test":"Re-run the Sec. VI simulation and, from the same simulated phase trajectories, empirically estimate the conditional density f(phi_i | phi_{i-1}, phi_{i-2}) (e.g., by kernel density) and compute q_true = 2*pi * min f over a fine grid. Compare with the q obtained from the paper's visibility method, Eqs. (19)-(22) and (13). Also bin the residuals phi_i - hat_phi_i^(2) by the phase separation phi_{i-1} - phi_{i-2} or by the coherent amplitude |hat_alpha_i^(2)|, and test whether the estimated sigma_2 varies by more than the statistical uncertainty. If q_est exceeds q_true in any bin, the certification is not conservative.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eqs. (19)-(22) and (C17)-(C18) identify max v^(lc) with exp(-sigma_lc^2/2) only under A4, i.e., delta_phi_i is iid wrapped Gaussian with constant sigma_lc, independent of the past phases that define epsilon'_i and of the coherent amplitude left in the cavity. Appendix A explicitly acknowledges that sigma_lc may depend on the alignment of previous phases, and the numerical simulations in Sec. VI never validate A4 against the empirical conditional phase distribution of the simulated laser. The security parameter q in Eq. (1) is a minimum over histories of a density ratio; a single effective sigma obtained from an average visibility cannot certify a lower bound if the true conditional distributions are a mixture of wrapped Gaussians with history-dependent widths. In that case the maximum of v can be close to exp(-<sigma^2>/2) or to a different average, and plugging the resulting sigma into Eq. (13) can give q_est above the true q. Since the paper's stated goal is to certify arbitrary gain-switched sources, this unvalidated parametric step is the load-bearing point.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper addresses the practical problem of certifying gain-switched laser sources for decoy-state QKD when phase correlations between consecutive pulses are present. It proposes an interferometric method, based on a cascade of delay lines with tunable attenuators and a phase shifter, to estimate the parameters of a wrapped-Gaussian phase-noise model: the conditional standard deviation σ_ℓc, the relative residual-photon amplitudes r_{i−k}, and the phase offset δφ. From these parameters, the security parameter q appearing in the recent security proof of Currás-Lorenzo et al. is computed numerically for arbitrary correlation length ℓc, via Eqs. (13), (24) and (25). The authors simulate a gain-switched laser using stochastic rate equations, apply their estimation scheme for ℓc = 1 and 2, and present resulting secret-key rates. The main claimed contribution is a practical recipe to quantify the impact of arbitrary-order phase correlations without modifying the source.","tokens_in":21532,"tokens_out":9587,"duration_ms":101989,"significance":"If the proposed method is valid, it would substantially increase the applicability of the security proof of Ref. [37] to high-speed gain-switched lasers, which is an important practical issue for decoy-state QKD. The manuscript contains a clear algebraic derivation of the generalized visibility, a self-contained derivation of the wrapped-Gaussian conditional distribution from Assumption (A4), and a numerical demonstration based on realistic laser rate equations. The extension to exponentially decaying correlations in Appendix D is also a useful contribution. The derivation of Eq. (C18) and the optimization procedure are, within the model, internally consistent. However, the significance is conditional: the method is a parametric characterization under Assumption (A4), and the paper does not yet provide evidence that this assumption holds for the simulated laser or that the resulting q is a conservative lower bound under model mismatch.","major_comments":[{"comment":"The central claim that the protocol certifies q for an arbitrary gain-switched source is not established because Assumption (A4) is not validated against the simulated laser, and Appendix A explicitly concedes that σ_ℓc may depend on the previous phase realisations. Equations (C17)-(C18) rely on the independence of δφ_i from the past phases to factor the average ⟨cos(δφ_i+φ) cos ε'_i⟩; if the conditional variance is history-dependent, the measured max_{φ,A} v^(ℓc) is an average of exp(−σ²(history)/2) over histories, not a value that bounds every history. Since q in Eq. (1) is a minimum over histories, plugging a single σ_ℓc obtained from this average into Eq. (13) can yield q_est above the true q. The simulations in Sec. VI should directly test Assumption (A4), for example by estimating the empirical conditional phase distribution of the simulated laser as a function of the previous phase-alignment variables and checking that the variance is constant and the shape is wrapped Gaussian. Alternatively, the estimation procedure should be modified to produce a conservative lower bound on σ for every history, for instance by binning histories and taking the worst-case conditional variance.","section":"Sec. VI and Appendix A, Eqs. (C17)-(C18) and (13)"},{"comment":"The paper does not provide a finite-sample or confidence analysis for the estimated security parameter. In Sec. VI, v^(2) is estimated from 10^4 pulses and then maximized over φ and A; the maximum of noisy sample averages is biased upward relative to the true maximum, and Tables I-II report point values of σ_ℓc and q without error bars. Because the security statement requires a certified lower bound on q in Eq. (1), the procedure as written is an estimator rather than a certification. The authors should state how many pulses are needed, provide concentration bounds or bootstrap confidence intervals, and explain how the reported q values should be interpreted conservatively in a security analysis.","section":"Secs. IV-VI, Tables I and II"}],"minor_comments":[{"comment":"The numerical minimization of Eq. (13) is not specified in terms of grid size or optimization algorithm; please provide these details so the results in Fig. 3 can be reproduced.","section":"Sec. IV, Fig. 3, Eq. (13)"},{"comment":"The generalized visibility expressions can have near-zero denominators when the interfering previous-pulse amplitudes are nearly opposite in phase, making the per-round cosine undefined or noisy; the manuscript should discuss how such rounds are treated in the average and whether this affects the estimate.","section":"Eqs. (19) and (24), Sec. VI"},{"comment":"There are minor typographical issues: 'Ito' should be 'Itô' and 'Mach-Zender' should be 'Mach-Zehnder'; also, the notation for the off-current I_off is typeset inconsistently in Tables I and II.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of the journal and addresses a timely problem in practical QKD security. The main risk is that the proposed characterization is presented as a certification while it is in fact conditional on a parametric model that the authors themselves acknowledge may fail. I would invite a revision that either validates Assumption (A4) in the simulations or reformulates the claims conservatively, and that adds a finite-statistics analysis. With those additions, the contribution would be solid."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: the paper is a solid, clearly written extension of the first-order analysis to arbitrary correlation length, and it fills a real gap—no recipe existed for estimating q for lc>1. The cascade interferometer and the generalized visibility v^(lc) are natural and well derived, and the math under Assumption A4 is consistent. The soft spot is that A4 is load-bearing and never validated against the simulated source. The paper is honest about the limitation in Appendix A, where it concedes that sigma_lc may depend on the alignment of previous phases. That is not just a technical caveat: the security parameter q is a minimum over histories of the conditional density. If the true noise is a mixture of wrapped Gaussians with history-dependent widths, the visibility only fixes an average, and the actual minimum can be far worse than the model predicts. Plugging the effective sigma into Eq. (13) can then overestimate q, which is unsafe for certification.\n\nWhat is genuinely new: the general recipe for arbitrary lc, the convergence of v^(lc) to the standard visibility at lc=1, and the numerical demonstration across realistic laser parameters. The optimization proof is fine—by a simple inequality, max v^(lc) is exactly exp(-sigma^2/2) under A4—so that concern from a first pass is not a real flaw.\n\nThe main weakness is that the virtual experiment in Sec. VI never computes the actual conditional phase distribution of the rate-equation laser, nor the true q, and compares with the estimated q. That is the obvious validation step, and its absence leaves the central claim conditional. Missing error bars and code are minor in comparison.\n\nThis paper is for people building practical decoy-state QKD with gain-switched sources, and for researchers who need to certify sources for security proofs. It deserves a serious referee: the idea is promising, the writing is clear, and the missing robustness check is addressable. I would send it to review with a request—probably a major revision—to either validate A4 against the simulation or provide a conservative bound on q under model mismatch. Until then, the method should not be used to claim security without additional work.","headline":"Fills the lc>1 gap with a clean cascade-interferometer recipe, but the security parameter q rests on an unvalidated constant-variance Gaussian assumption; the simulation never checks estimated q against the true q.","tokens_in":22053,"tokens_out":5951,"would_cite":false,"duration_ms":60192,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A tunable cascade interferometer can determine the phase-correlation security parameter q of a gain-switched laser, allowing decoy-state QKD with fast sources.","keywords":["quantum key distribution","gain-switched lasers","phase correlations","decoy-state method","phase randomisation","generalised visibility","wrapped Gaussian distribution","laser phase noise"],"falsifier":"A phase-resolved measurement of every pulse at a high repetition rate, binning the observed conditional distribution $f(\\varphi_i|\\varphi_{i-1},\\varphi_{i-2})$ over many alignments of the previous phases, would settle the central model: if the jitter spread changes with the alignment of previous phases or the conditional distribution departs from a wrapped Gaussian, the visibility-based estimates of $\\sigma_{\\ell_c}$ and $q$ would not be reliable lower bounds.","tokens_in":21064,"feed_emoji":"🔐","tokens_out":8505,"duration_ms":77661,"temperature":0.7,"pith_summary":"This paper provides an experimental recipe for determining how much phase randomness a gain-switched laser actually has when it is pulsed quickly enough that each pulse inherits light from earlier pulses. The central quantity is the security parameter $q$, the minimum uniform component of the conditional phase distribution, which a recent security proof needs to certify decoy-state QKD. The paper shows that a cascade interferometer with tunable attenuators and a phase shifter can be optimised over a generalised visibility $v^{(\\ell_c)}$ to estimate the phase-noise width $\\sigma_{\\ell_c}$, the residual-field ratios $r_{i-k}$, and the phase offset $\\delta\\varphi$. These estimates feed a numerical minimisation that yields $q$ for any finite correlation length. If this works, high-speed gain-switched sources can be certified for secure key distribution without changing the source.","feed_headline":"Interferometer certifies fast gain-switched lasers for QKD","feed_subtitle":"A tunable cascade measures q, the uniform part of pulse phases, closing the gap between security proofs and high-speed sources.","key_machinery":"The central object is the generalised visibility $v^{(\\ell_c)}$, an ensemble-averaged interference cosine in which the intensity measured at each round is rescaled so that the reconstructed combination intensity and the pulse phase are decoupled. It is measured in a cascade Mach-Zehnder interferometer with delay lines of lengths $0, T, 2T, \\ldots, \\ell_c T$, tunable attenuators $A_k$, and a phase shifter $\\varphi$. The optimisation $\\max_{\\varphi,\\{A_k\\}} v^{(\\ell_c)}$ returns $\\sigma_{\\ell_c}$ through Eq. (25), the residual-field ratios $r_{i-k}$, and the phase offset $\\delta\\varphi$; the matched attenuator settings make the interfering state $|\\chi\\rangle$ carry the same phase that the laser cavity would produce by coherently combining residual photons from the previous $\\ell_c$ pulses, so the remaining spread in the interference is the spontaneous-emission noise.","core_discovery":"The paper establishes that the security parameter $q$ of a gain-switched laser source with phase correlations of arbitrary finite range $\\ell_c$ can be obtained from interferometric measurements alone. The key claim is that a cascade of delay lines can be tuned so that the phase of a reconstructed coherent state $|\\chi\\rangle$ equals the conditional centre $\\hat{\\phi}^{(\\ell_c)}_i$ of the next pulse phase; the visibility of the interference between $|\\chi\\rangle$ and the actual pulse then measures the spontaneous-emission jitter. Maximising the generalised visibility $v^{(\\ell_c)}$ gives $\\sigma_{\\ell_c}$ through $\\max v^{(\\ell_c)} = \\exp(-\\sigma_{\\ell_c}^2/2)$, and the optimal attenuator settings give the residual-field ratios $r_{i-k}$ and the phase offset $\\delta\\varphi$. Feeding these into the wrapped-Gaussian conditional distribution and the law of total probability yields $q$. Numerical simulations with a stochastic laser model at 100 MHz to 10 GHz show that $q$ stays close to 1 at low rates and can fall below 0.001 at 5 GHz and above, where the paper applies the security proof to compute achievable secret key rates.","pith_inferences":["The same interferometric characterisation could serve as a continuous monitor of phase-randomisation quality for other phase-sensitive tasks, such as quantum random number generation or coherent-state communication, not just QKD.","If correlations decay exponentially, the feedback-loop variant of the scheme suggests a direct test: fit measured visibilities across several delay lengths to confirm that a single decay ratio $r_0$ captures the full history.","The strongest test of the model would be phase-resolved measurements of individual pulses at 5-10 GHz, comparing the observed conditional distribution with the predicted wrapped Gaussian; this would show whether the constant-spread assumption in the paper's central model holds."],"forward_implications":["A gain-switched laser with phase correlations of any finite length $\\ell_c$ can be certified for decoy-state QKD without changing the source hardware or the state preparation.","The measurement can run during key distribution because it does not alter the emitted states, so $q$ can be monitored in situ.","At a fixed phase-noise spread $\\sigma_{\\ell_c}$, the worst-case $q$ in the simulations occurs near $r_{i-2}=1$, so that value can serve as a conservative bound when the residual ratio is uncertain.","At low channel loss, a faster source with slightly lower $q$ can still produce more secure bits per second than a slower source with $q\\approx 1$.","For sources with $q=0$ the proof used here gives no positive key rate, and the paper notes that such non-random-phase sources require a different security analysis."],"supporting_citations":[{"why":"Supplies the security proof for decoy-state QKD with phase correlations and defines the parameter q that the paper's method estimates.","marker":"[37]"},{"why":"Introduces the phase-diffusion model and the first-order visibility relation V = exp(-sigma_1^2/2) that the paper generalises to higher orders.","marker":"[35]"},{"why":"Provides the 5 GHz experimental visibility value and evidence that high repetition rates create measurable phase correlations.","marker":"[36]"},{"why":"Gives the coherent-state description of how residual photons and spontaneous emissions combine in the cavity, on which the reconstructed-state argument rests.","marker":"[38]"},{"why":"Supplies the discrete-mode laser parameters used in the paper's stochastic simulations of gain-switched operation.","marker":"[41]"},{"why":"Together with [41], provides the simulated laser parameters and rate-equation model used for the virtual experiment.","marker":"[42]"}],"fun_headline_variants":["Tunable interferometer measures phase jitter for QKD security","Cascade interferometer certifies high-rate laser sources for QKD","Phase-correlation probe secures gain-switched lasers for QKD","Interferometric cascade extracts q parameter for QKD","Fast QKD lasers verified by interferometric phase measurement"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the phase jitter added at each pulse is independent and identically distributed with a fixed spread regardless of the actual previous phase values or how well they align, which is what lets visibility measurements calibrate $\\sigma_{\\ell_c}$ and feeds the $q$ calculation.","fun_headline_variants_meta":{"raw":{"variants":["Tunable interferometer measures phase jitter for QKD security","Cascade interferometer certifies high-rate laser sources for QKD","Phase-correlation probe secures gain-switched lasers for QKD","Interferometric cascade extracts q parameter for QKD","Fast QKD lasers verified by interferometric phase measurement"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001177,"raw_usage":{"total_tokens":4854,"prompt_tokens":922,"completion_tokens":3932,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":3847}},"tokens_in":538,"tokens_out":3932,"duration_ms":24278,"temperature":1.0,"reasoning_tokens":3847,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T22:08:14.830485+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A phase-resolved measurement of every pulse at a high repetition rate, binning the observed conditional distribution $f(\\varphi_i|\\varphi_{i-1},\\varphi_{i-2})$ over many alignments of the previous phases, would settle the central model: if the jitter spread changes with the alignment of previous phases or the conditional distribution departs from a wrapped Gaussian, the visibility-based estimates of $\\sigma_{\\ell_c}$ and $q$ would not be reliable lower bounds.","supporting_citations":[{"cited_title":"Kobayashi, A","cited_arxiv_id":null,"evidence_quote":"Supplies the security proof for decoy-state QKD with phase correlations and defines the parameter q that the paper's method estimates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the phase-diffusion model and the first-order visibility relation V = exp(-sigma_1^2/2) that the paper generalises to higher orders."},{"cited_title":"Lucamarini, K","cited_arxiv_id":null,"evidence_quote":"Provides the 5 GHz experimental visibility value and evidence that high repetition rates create measurable phase correlations."},{"cited_title":"Agulleiro, F","cited_arxiv_id":null,"evidence_quote":"Supplies the discrete-mode laser parameters used in the paper's stochastic simulations of gain-switched operation."},{"cited_title":"Pereira, G","cited_arxiv_id":null,"evidence_quote":"Together with [41], provides the simulated laser parameters and rate-equation model used for the virtual experiment."}],"review_version":1}