{"id":"cb663f71-655a-44c0-be5c-923738ea4635","arxiv_id":"1908.02672","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Temporal filtering of detection windows in a BB84 testbed with a quantum-dot single-photon source improves error rates and extends tolerable loss, with real-time g(2)(0) monitoring inside the channel.","lead":"Temporal filtering of single-photon pulses can optimize the performance of quantum key distribution systems based on quantum-dot single-photon sources. Experiments and simulations show that carefully choosing acceptance time windows improves secret key rates and extends tolerable channel losses, and that photon statistics can be monitored in real time during key generation.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline range-extension figures rest on Eq. (3), but the paper never reports the single-photon fraction β or the privacy-amplification function τ(e) used; a consistency check against the stated μ=0.0043 and pdc=1.22e-6 suggests the quoted tolerable losses may not be reproducible.","rationale":"The reader's weakest assumption points to the Waks et al. security model and the unclosed PNS/gating caveat. That caveat is disclosed and, for the main figures, the authors use the full-period g(2)(0), which is conservative for the multi-photon bound. The more concrete and load-bearing problem is that the quantitative implementation of Eq. (3) is under-specified: β and τ(e) are not given, and the internal Tmin inconsistency suggests the numerical rate-loss outputs should not be taken at face value. This does not overturn the qualitative optimization claim, which is independently supported by the measured QBER and fraction trends and by the simulations. It does mean the paper's headline range-extension numbers should be treated as conditional on a complete and reproducible security-analysis parameterization. The reader's CONDITIONAL verdict therefore remains appropriate; a full rejection would be too strong given the plausible qualitative core, while an unconditional acceptance would overstate the evidence for the quantitative figures.","tokens_in":16837,"tokens_out":17096,"duration_ms":197949,"concrete_test":"Recompute the Fig. 3 rate-loss curves with the Waks et al. formula using β = p_single/p_click, where p_single is the single-photon click probability and p_click includes dark counts, with the stated μ=0.0043, pdc=1.22e-6 at Δt=12.5 ns scaled linearly with Δt, the measured zero-loss p_click, the measured QBER(Δt,tc), and the reported error-correction efficiency f(e). Check whether the optimized (Δt,tc) windows still yield 35.15 dB and 37.98 dB and whether the 23% range extension survives. If the filtered-window maxima drop by more than about 1 dB, or fall below the unfiltered value, the central range-extension claim needs revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim, a 23-24% range extension from temporal filtering, is computed from Eq. (3): S = pclick/2 (β τ(e) − f(e)h(e)). The paper identifies β as the fraction of detection events caused by single photons, but it never reports how β is estimated or what value is used in Fig. 3. The Methods section, 'Estimation of expected secret key rates', lists μ=0.0043, pclick, and pdc=1.22e-6 at Δt=12.5 ns, but is silent on β and τ(e). This matters because β = p_single/p_click decreases as dark counts dominate at high channel loss, exactly the regime where the range-extension claim is made. If β was absorbed into pclick or set to unity, the tolerable-loss values would be inflated. An internal inconsistency supports the concern: the same paragraph quotes Tmin = 33.24 dB for the maximal acceptance window, while Fig. 3 gives 28.28 dB for the same window. The qualitative trade-off, that narrow windows reduce the dark-count contribution to QBER, is plausible and is supported by the measured QBER curves, but the specific 35.15 dB and 37.98 dB figures, and hence the headline range extension, are not verifiable from the information provided.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports an experimental BB84-type QKD testbed built around a triggered quantum-dot single-photon source and a four-state polarization receiver. The authors record photon arrival-time distributions at Bob's four detectors and study how the choice of acceptance time window (width Δt and center tc) affects QBER, sifted-key fraction, and g(2)(0). Using Eq. (3) from the Waks security model, they translate these measured quantities into expected secret-key rates as a function of channel loss, and claim that temporal filtering optimizes back-to-back key rate or extends the maximally tolerable channel loss by about 23-24%. They also demonstrate real-time g(2)(0) monitoring from Bob's time-tags, and extend the study with simulations for synthetic pulse shapes and noise levels.","tokens_in":17078,"tokens_out":9740,"duration_ms":99834,"significance":"The qualitative core of the paper—that temporal filtering trades QBER against sifted fraction and has an optimum that depends on pulse shape and noise—is convincingly supported by the time-resolved data and is a useful practical design rule for single-photon QKD. The channel-side, real-time g(2)(0) monitoring using the four receiver ports is a genuinely useful certification tool, and the authors are explicit that exploiting the improved filtered g(2)(0) requires Alice-side gating, which is not implemented. The strengths are therefore real, but the headline quantitative figures (maximally tolerable loss and range extension) are computed from an incompletely specified model and contain an internal numerical inconsistency; these need to be fixed before the central claim can be accepted.","major_comments":[{"comment":"The rate-loss curves in Fig. 3 are computed from Eq. (3), which contains β, the fraction of detection events caused by single photons, and τ(e), the privacy-amplification compression function. The Methods paragraph on secret-key estimation gives μ=0.0043, pclick, and pdc=1.22e-6 but never states the value or estimation procedure for β and τ(e). Because β = p_single/p_click is loss-dependent and becomes small when dark counts dominate at high loss, the tolerable-loss values (35.15 dB and 37.98 dB) and the associated 23-24% range extension cannot be reproduced or audited from the reported information. Please report β and τ(e) explicitly, specify whether β was held constant or recalculated at each loss value, and provide a sensitivity analysis showing how the tolerable-loss curves depend on these assumptions.","section":"Eq. (3) and Methods, 'Estimation of expected secret key rates'"},{"comment":"The text states that for the maximal acceptance window the critical photon number μc=0.0053 corresponds to Tmin=33.24 dB and refers to Fig. 3, but Fig. 3 shows a maximal tolerable loss of only 28.28 dB for the same window (Δt=12.50 ns, tc=0.00 ns). The text goes on to say that the actual μ=0.0043 leads to a 'slightly reduced maximal tolerable loss (cf. Fig. 3)', which is difficult to reconcile with a 5 dB gap between 33.24 dB and 28.28 dB. This inconsistency undermines confidence in the quantitative range-extension numbers; please clarify which model convention produces Tmin=33.24 dB and correct either the text or the figure accordingly.","section":"Results, 'Performance optimization via temporal filtering' (paragraph on μ_c and Tmin)"},{"comment":"The claim that a 60 s accumulation block 'should already be enough to allow for a secret key distillation incorporating finite-key size effects' is presented without any finite-key calculation or block-size analysis, and the experiment is an off-line post-processing of recorded time-tags rather than live key distillation. Please either provide the finite-key estimate or rephrase the statement as an asymptotic-parameter indication; otherwise the 'real-time security monitoring' claim exceeds what is demonstrated.","section":"Results, 'Real-time photon statistics monitoring'"}],"minor_comments":[{"comment":"There are typos in the text and figures: 'fiexd' should be 'fixed' in the Fig. 2 caption, and the horizontal axis label 'Accpetance window Δt (ns)' should read 'Acceptance window Δt (ns)'.","section":"Fig. 2 caption and axis labels"},{"comment":"Please clarify the conversion from the cumulative dark-count rate (<100 Hz) to the per-pulse probability pdc=1.22e-6 for Δt=12.5 ns; state explicitly whether pdc is per pulse and how the 80 MHz repetition rate enters the calculation.","section":"Methods, 'Estimation of expected secret key rates'"},{"comment":"In Eq. (4), the symbol f_rep is not defined, and the expression changes units relative to Eq. (3) (bits/pulse versus bits per second); please define f_rep and state the relationship between the two rate conventions.","section":"Methods, 'Comparison of Device Performance', Eq. (4)"},{"comment":"In the sentence describing over- and underestimation of g(2)(0), the second occurrence of 'overestimation' is apparently meant to be 'underestimation', since only underestimation can lead to information leakage; please correct this.","section":"Results, 'Real-time photon statistics monitoring'"},{"comment":"The paper quotes the range extension as both '23%' in the Results section and '24%' in the same section and abstract-level discussion; please use one value consistently or state the rounding convention.","section":"Results, 'Performance optimization via temporal filtering'"}],"recommendation":"major_revision","confidential_remarks":"The experimental work and the qualitative optimization concept are solid, and the missing information (β, τ(e), and the Tmin inconsistency) is fixable without new measurements. I recommend major revision rather than rejection. The authors should also consider toning down the abstract-level wording about achievable communication distance unless the asymptotic-model caveats are clearly attached to those numbers."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is a solid experimental study of temporal filtering for single-photon QKD. What's genuinely new is the systematic 2D (Δt, tc) optimization applied to a real QD-based testbed, and the demonstration of g(2)(0) monitoring from Bob's detection stream during key generation. The qualitative conclusion—that narrowing Bob's acceptance window improves QBER and can extend the tolerable loss—holds up well. The measured QBER and sifted-fraction curves support it, and the simulations usefully generalize the trade-off to other pulse shapes and noise levels.\n\nThe authors deserve credit for being honest about the main caveat: the filtered g(2)(0) improvement only helps if Alice actively gates her output, otherwise a photon-number-splitting backdoor opens. They flag this in both the main text and the supplement, and they clearly state that the experiment is a testbed, not a full QKD implementation.\n\nThe soft spots are real but not fatal. The headline range-extension numbers (35.15 dB, 37.98 dB, 24% extension) come from Eq. (3), the Waks security model, but the paper never reports the single-photon fraction β or the privacy-amplification function τ(e) used. Without those, the specific figures are not reproducible from the manuscript. The stress-test's apparent inconsistency between 33.24 dB and 28.28 dB is actually explainable—one is for the unity-efficiency limit, the other for the actual μ—but the text doesn't make that distinction clear enough. Also, the main text says the other channels show \"similar behavior\" to H, but the supplement's D-channel clearly does not: its QBER rises again for very narrow windows. That's a minor overstatement, and the supplement does disclose it.\n\nNo circularity: they apply an external security model to measured data. The data-availability statement is the weak \"on reasonable request\" standard, and no code is released, which makes it harder to check the model-dependent numbers.\n\nWho is this for? Experimentalists working on QKD with single-photon sources, especially those interested in source certification and practical receiver optimization. It is not a protocol paper. I would send it to peer review and ask for revision: report β and τ(e), clarify the 33.24 dB vs 28.28 dB distinction, and note the D-channel behavior in the main text. The core contribution is worth adding to the literature.","headline":"A careful experimental study of temporal filtering for single-photon QKD; the qualitative optimization is solid, but the headline key-rate numbers depend on model parameters the paper never reports.","tokens_in":17724,"tokens_out":3126,"would_cite":true,"duration_ms":34267,"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":"Temporal filtering of single-photon pulses can maximize QKD key rate or extend its reach.","keywords":["single-photon sources","quantum key distribution","temporal filtering","BB84","g(2)(0)","quantum dot","secure key rate","photon-number-splitting attack"],"falsifier":"Run a full QKD exchange with active gating at Alice and a finite-key analysis: if, after the gating's added loss and timing overhead, the maximum tolerable channel loss does not reach the predicted 35 dB for a source with $g^{(2)}(0)\\approx0.09$, the optimization claim is falsified. Alternatively, a simulation that lets an eavesdropper intercept the photons outside Bob's acceptance window and checks whether the filtered key remains secret would settle the security question directly.","tokens_in":16594,"feed_emoji":"🔐","tokens_out":7416,"duration_ms":70350,"temperature":0.7,"pith_summary":"Single-photon sources can outperform attenuated lasers in quantum key distribution, but only if the receiver uses the photons efficiently. This paper shows that temporally filtering the detected single-photon pulses, keeping only clicks inside a carefully chosen acceptance window, changes the trade-off between error rate and sifted key. A careful setting can either maximize the secure key rate for a given channel loss or push the maximum tolerable loss further out, extending the achievable distance. It also demonstrates real-time monitoring of the two-photon emission probability $g^{(2)}(0)$ inside the quantum channel during key generation, which is needed to certify security. The result is a practical optimization tool for QKD with realistic quantum light sources.","feed_headline":"Narrowing Bob's time window extends QKD reach by 23%","feed_subtitle":"Tuning acceptance windows trades key rate for distance and enables real-time security monitoring in a quantum-dot testbed.","key_machinery":"The central object is the two-dimensional temporal acceptance filter $(\\Delta t, t_c)$ applied to Bob's time-tagged detection events: it selects which photon clicks enter the sifted key, changing both the error rate and the key fraction. The performance estimate is carried by the secret-key expression $S = \\frac{p_{\\mathrm{click}}}{2}(\\beta\\tau(e) - f(e)h(e))$ together with the multi-photon bound $p_m \\le \\mu^2 g^{(2)}(0)/2$ from the security model underlying the paper. Here $g^{(2)}(0)$ is the two-photon emission probability, a measure of single-photon purity. The filter's effect is quantified through the QBER, the sifted fraction, and the $g^{(2)}(0)$ evaluated from the same time-tags, and the paper searches the $(\\Delta t, t_c)$ plane for the optimum; in simulations the same machinery is used with synthetic pulse shapes to predict performance for different source lifetimes and noise levels.","core_discovery":"The paper's central claim is that optimal QKD performance for a given triggered single-photon source is achieved not by changing the source, but by carefully choosing Bob's acceptance time windows, the width $\\Delta t$ and center $t_c$ of the temporal filter applied to the recorded click times. Narrowing the window improves the signal-to-noise ratio and lowers the quantum bit error rate, but discards part of the sifted key; the paper maps this trade-off in a two-dimensional parameter space and shows that the best operating point either maximizes the back-to-back secret key rate or extends the maximum tolerable channel loss. With their quantum-dot source and four-state BB84 receiver, the optimized filter raises the tolerable loss from 28.28 dB to 35.15 dB (about a 23% range extension), and to 37.98 dB with a very narrow window at the cost of a much smaller sifted fraction. The paper additionally shows that narrowing the acceptance window improves the measured $g^{(2)}(0)$ of the emitted pulses, from $0.104\\pm0.017$ at full width to $0.032\\pm0.007$ at 2.5 ns, and that this filtered value can only be used for security if Alice actively gates her source to prevent photon-number-splitting attacks outside the window; their quoted rates therefore use the unfiltered $g^{(2)}(0)$. Real-time monitoring of $g^{(2)}(0)$ from the receiver's four ports is demonstrated over 90 minutes, with 60-second blocks giving $g^{(2)}(0)$ to about 16% uncertainty.","pith_inferences":["Beyond the paper, if Alice-side active gating is added so the filtered $g^{(2)}(0)$ can be used in security, the same data predict a further increase in tolerable loss beyond 38 dB, since the narrow-window data in the supplement show both lower $g^{(2)}(0)$ and lower QBER.","The authors do not discuss it, but the optimal window width should shrink as detector timing jitter improves; with superconducting detectors, the high-noise regime where filtering matters most is replaced by a low-noise regime, shifting the optimum toward wider windows.","An implicit testable extension is to run the same 2D optimization for decoy-state weak-coherent-pulse systems; the paper's framework would predict a much smaller gain there, because Poissonian light has no antibunching to improve.","Beyond the paper's asymptotic analysis, narrowing the window cuts the sifted block size, so finite-key effects become more severe at the optimized operating points; a practical implementation would need to choose the window with block-size constraints in mind."],"forward_implications":["With an optimized acceptance window, a quantum-dot source with $g^{(2)}(0)=0.089$ can tolerate 35.15 dB of channel loss instead of 28.28 dB, a range extension of about 23% for the tested receiver.","At fiber attenuations of 0.31 dB/km (1310 nm) and 0.17 dB/km (1550 nm), the same optimization would extend secure communication distance by roughly 22 km and 40 km, to about 113 km and 207 km, respectively.","The improvement is largest in the high-noise, high-loss regime: simulations with synthetic pulses show secret-key gains of 184.5% and 148.3% when detector noise is high.","Real-time $g^{(2)}(0)$ monitoring inside the quantum channel makes it possible to detect an eavesdropper who injects light with different photon statistics or performs a photon-number-splitting attack, by comparing the statistics seen by Bob with those seen by Alice.","The optimization routine transfers to measurement-device-independent QKD and quantum repeater architectures based on sub-Poissonian sources, where the same temporal-filter trade-off applies."],"supporting_citations":[{"why":"Supplies the security model for sub-Poissonian sources, including the multi-photon bound and the secret-key formula used throughout.","marker":"[27]"},{"why":"Earlier single-photon QKD experiment whose source efficiency value is used as a comparison benchmark.","marker":"[19]"},{"why":"Decoy-state weak-coherent-pulse protocol whose performance is the baseline the paper argues sub-Poissonian sources can surpass at higher source efficiency.","marker":"[8]"},{"why":"Recent single-photon QKD implementation cited as the best current source-efficiency comparison point.","marker":"[33]"},{"why":"Shows detection-efficiency mismatch raises the tolerable QBER threshold, justifying the careful channel synchronization in the postprocessing.","marker":"[29]"},{"why":"Cited to justify that 60-second blocks with the measured $g^{(2)}(0)$ uncertainty should support secret-key distillation with finite-key effects.","marker":"[37]"},{"why":"State-of-the-art deterministic solid-state sources cited to show the efficiency threshold needed to outperform weak-coherent-pulse QKD is within reach.","marker":"[34]"}],"fun_headline_variants":["Quantum dot testbed tunes time windows to extend QKD by 23%","Temporal filtering of single photons boosts QKD loss tolerance by 23%","Choose Bob's timing window: QKD range extends 23%","Time-window tuning extends quantum key distribution reach by 23%","Optimal time filter widens QKD's distance limit by 23%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The rate and range figures depend on the assumption that the two-photon emission probability measured over the whole pulse period still bounds the information an eavesdropper can obtain about the temporally filtered key, even though the experiment lacks the active gating at Alice's side that would prevent attacks using the photons discarded by the filter.","fun_headline_variants_meta":{"raw":{"variants":["Quantum dot testbed tunes time windows to extend QKD by 23%","Temporal filtering of single photons boosts QKD loss tolerance by 23%","Choose Bob's timing window: QKD range extends 23%","Time-window tuning extends quantum key distribution reach by 23%","Optimal time filter widens QKD's distance limit by 23%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00079,"raw_usage":{"total_tokens":3573,"prompt_tokens":1128,"completion_tokens":2445,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":744,"completion_tokens_details":{"reasoning_tokens":2348}},"tokens_in":744,"tokens_out":2445,"duration_ms":18398,"temperature":1.0,"reasoning_tokens":2348,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:38:02.277270+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a full QKD exchange with active gating at Alice and a finite-key analysis: if, after the gating's added loss and timing overhead, the maximum tolerable channel loss does not reach the predicted 35 dB for a source with $g^{(2)}(0)\\approx0.09$, the optimization claim is falsified. Alternatively, a simulation that lets an eavesdropper intercept the photons outside Bob's acceptance window and checks whether the filtered key remains secret would settle the security question directly.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the security model for sub-Poissonian sources, including the multi-photon bound and the secret-key formula used throughout."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier single-photon QKD experiment whose source efficiency value is used as a comparison benchmark."},{"cited_title":"The upper bound for the multi-photon emission probability from [27] yields P (n = 1) ≥ µ−µ2g(2)(0)","cited_arxiv_id":null,"evidence_quote":"Decoy-state weak-coherent-pulse protocol whose performance is the baseline the paper argues sub-Poissonian sources can surpass at higher source efficiency."},{"cited_title":"Sch¨ oll, L","cited_arxiv_id":null,"evidence_quote":"Recent single-photon QKD implementation cited as the best current source-efficiency comparison point."},{"cited_title":"Gschrey, A","cited_arxiv_id":null,"evidence_quote":"Shows detection-efficiency mismatch raises the tolerable QBER threshold, justifying the careful channel synchronization in the postprocessing."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Cited to justify that 60-second blocks with the measured $g^{(2)}(0)$ uncertainty should support secret-key distillation with finite-key effects."},{"cited_title":"Takemoto, Y","cited_arxiv_id":null,"evidence_quote":"State-of-the-art deterministic solid-state sources cited to show the efficiency threshold needed to outperform weak-coherent-pulse QKD is within reach."}],"review_version":1}