{"id":"193dc216-496a-440f-98a2-0c692582b343","arxiv_id":"2603.20718","paper_version":3,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.5,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Four-channel 10-Mbaud FDM-CV-QKD with TLO and homodyne detection yields a 3.7-fold back-to-back secret-key-rate gain and outperforms single-channel systems out to 41.1 km under finite-size analysis.","lead":"A four-channel frequency-division multiplexed continuous-variable quantum key distribution experiment packs low-rate Gaussian signals densely and multiplies the secret key rate. The approach offers a practical way to raise key rates on short-to-medium fiber links without OFDM complexity.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"Finite-size SKR curves rest on back-to-back excess noise inserted into a VOA-emulated loss model; real-fiber effects that grow with distance are never measured.","rationale":"The reader correctly isolates the VOA-emulation assumption as the weakest link supporting the distance-dependent part of the strongest claim. My stress test confirms that this is load-bearing: the entire family of SKR-vs-distance curves (and therefore the 41.1 km superiority statement) is an extrapolation that inserts a single B2B excess-noise value into a pure-loss model. No independent measurement of excess noise after real fiber is reported, so the claim that dense FDM remains advantageous out to tens of kilometres rests on an untested premise. The concrete spool test would settle the issue cleanly. Because the back-to-back gain itself is directly measured and the engineering demonstration of dense low-baud FDM is solid, the appropriate verdict remains CONDITIONAL rather than REJECT; the concern simply reinforces the reader’s already-correct assessment.","tokens_in":7909,"tokens_out":675,"duration_ms":6732,"concrete_test":"Replace VOA2/VOA3 with a single spool of standard SMF (e.g., 20 km and 40 km) while keeping the same TLO power and modulation variance; re-estimate ε(k) for every FDM channel with m=1.25e6 symbols and recompute the finite-size RT curves of Fig. 5(a). If the four-channel advantage over single-channel disappears or the crossover falls below ~30 km, the distance claim does not survive real fiber.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The strongest claim (3.7-fold B2B gain and superiority out to 41.1 km under N=10^7) is obtained by measuring excess noise only in the back-to-back configuration (m=1.25e6 symbols) and then substituting that fixed ε into Eqs. (1)–(2) while sweeping T_ch via VOAs (Section II, Fig. 5 caption: “based on the back-to-back excess noise … Markers represent experimental verification via emulated channel loss using VOAs”). Equation (2) already shows that detector noise is scaled by 1/T_ch under TLO; any additional distance-dependent excess noise (fiber Raman, residual dispersion converting laser frequency noise into quadrature noise, polarization drift between signal and LO paths) would raise ε(k) and therefore ξ_tot(k) at longer reaches. Because the multi-channel system already sits closer to the noise floor (Fig. 4(b)), even a modest rise in ε would shrink the 41.1 km crossover more than the single-channel baseline, undermining the claimed distance advantage. The abstract/body numerical mismatch (3.6-fold/26.8 km vs 3.7-fold/41.1 km) further indicates that the finite-size numbers are sensitive to the precise noise model.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript reports a proof-of-principle four-channel FDM-CV-QKD experiment using 10-Mbaud Gaussian-modulated coherent states, a transmitted local oscillator, pilot-clock basis selection, and homodyne detection. After establishing empirical design rules (minimum f_IF / SR_BB ≥ 6.4 and channel spacing ≈ 4 \times SR_BB) that keep excess noise from sideband overlap and crosstalk under control, the authors measure back-to-back excess noise versus channel count and insert those values into the standard finite-size SKR formula (Eqs. 1–2, N = 10^7, m = 1.25 \times 10^6). They claim a 3.7-fold back-to-back SKR gain relative to a single channel and a higher SKR than the single-channel baseline out to 41.1 km, while also showing that dense low-baud FDM outperforms a single higher-baud frequency-upconverted signal occupying the same electrical bandwidth.","tokens_in":8267,"tokens_out":1052,"duration_ms":8827,"significance":"If the distance-dependent claims hold under realistic fiber, the work supplies a concrete, experimentally validated design rule for packing multiple low-baud CV-QKD carriers inside the shot-noise-dominant region of a commercial balanced detector. That is a useful engineering contribution for multi-user or multi-carrier CV-QKD architectures that prefer independent subcarrier processing over OFDM. The experimental chain (Gaussian modulation, TLO, pilot clocks, LPF filtering, Leverrier finite-size analysis) is standard and correctly applied; excess-noise growth with channel count is measured and shown to saturate. The comparison against a single high-baud up-converted carrier is a clear, falsifiable demonstration of the spectral-efficiency argument.","major_comments":[{"comment":"Section II and Fig. 5: all finite-size SKR-versus-distance curves are generated by measuring excess noise only in the back-to-back configuration and then sweeping T_ch with VOAs while holding ε fixed. Equation (2) already scales detector noise by 1/T_ch under TLO; any additional distance-dependent excess noise (Raman scattering, residual dispersion converting laser frequency noise into quadrature noise, polarization drift between the separate signal and LO fibers) would raise ξ_tot more severely for the multi-channel system, which already sits closer to the noise floor (Fig. 4(b)). The claimed 41.1 km crossover and the superiority over a single high-baud carrier therefore rest on an untested assumption. At least one real-fiber data point (or a quantitative bound on the additional ε) is required before the distance claims can be regarded as established.","section":null},{"comment":"Abstract versus body numerical inconsistency: the abstract states a 3.6-fold gain and superiority up to 26.8 km (m = 1.25 \times 10^6), while the body and Fig. 5 report 3.7-fold and 41.1 km (N = 10^7). The discrepancy indicates that the finite-size numbers are sensitive to the precise noise model or block-size convention; the manuscript must adopt a single, self-consistent set of parameters and correct both abstract and body.","section":null}],"minor_comments":[{"comment":"Fig. 1(c) caption and surrounding text: the LPF is described as 10 MHz, yet the demultiplexer bandwidth is later given as 35.16 MHz; a short clarification of the filtering cascade would help reproducibility.","section":null},{"comment":"Eq. (1) uses both N and n without an explicit statement that n = N - m; a one-line definition would remove ambiguity.","section":null},{"comment":"The phrase “optimized channel spacing of low-symbol-rate signals” appears in the abstract but the optimization criterion is never stated formally; a sentence linking the 40 MHz choice to the measured excess-noise floor would strengthen the claim.","section":null},{"comment":"References [10] and [11] already treat multi-carrier CV-QKD; a brief sentence distinguishing the present FDM approach (independent subcarrier DSP, no FFT) from those OFDM results would better locate the novelty.","section":null}],"recommendation":"major_revision","confidential_remarks":"The core experimental result (excess-noise saturation under dense FDM and the spectral-efficiency comparison) is solid and publishable once the VOA-emulation limitation is either closed with real-fiber data or explicitly caveated. The abstract/body mismatch looks like an uncorrected draft inconsistency rather than a deeper problem. Scope is appropriate for a Photonics Technology Letters-style short communication."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The useful new result here is the experimental demonstration that packing four 10-Mbaud filtered Gaussian channels at 40 MHz spacing (first IF at 64 MHz) gives a nearly linear 3.7× back-to-back SKR gain under finite-size analysis (N=10^7) and still beats a single higher-baud frequency-upconverted signal that occupies the same detector bandwidth. That comparison is the real contribution; multi-carrier CV-QKD already exists in the cited literature, but the concrete spacing rules (IF/SR ≥ 6.4, Δf ≈ 4×SR) and the head-to-head against high-baud single-channel are new and cleanly measured.\n\nThey did the standard things correctly: TLO, pilot-clock basis selection, LPF side-lobe control, excess-noise extraction versus main-carrier power and channel count, and the Leverrier finite-size formula. Excess noise grows with channel number but saturates, which is the key practical observation that makes the dense packing viable. The math and citation pattern look ordinary and solid for a Photonics Technology Letters piece.\n\nThe soft spots are real but proportionate. All distance curves (Fig. 5) insert the back-to-back excess noise into the TLO noise model while sweeping T_ch with VOAs; real fiber Raman, residual dispersion converting laser frequency noise, and polarization drift between the two paths are never measured. Because the multi-channel system already sits closer to the noise floor, any distance-dependent rise in ε would shrink the 41.1 km crossover more than the single-channel baseline. There is also a small abstract/body numerical mismatch (3.6×/26.8 km vs 3.7×/41.1 km) that should be cleaned up. Neither issue kills the B2B result or the spectral-efficiency argument.\n\nThis is for people building multi-user or multi-carrier CV-QKD hardware who need concrete packing rules inside a fixed detector bandwidth. It deserves a serious referee; the experiment is careful enough and the engineering claim is falsifiable. I would engage with it, cite the spacing numbers if I were doing similar work, and send it out for review with a request for either real-fiber data or a clearer statement of the VOA limitation.","headline":"Solid experimental letter showing dense low-baud FDM beats a single high-baud channel for CV-QKD SKR; the 3.7× B2B gain is real under lab conditions, but distance claims rest on VOA-emulated loss with fixed back-to-back excess noise.","tokens_in":8902,"tokens_out":589,"would_cite":true,"duration_ms":5777,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Four low-baud FDM channels give a 3.7-fold back-to-back secret-key-rate gain in CV-QKD and beat a single channel out to 41.1 km under finite-size security.","keywords":["continuous-variable quantum key distribution","frequency-division multiplexing","Gaussian modulation","transmitted local oscillator","homodyne detection","finite-size secret key rate","spectral efficiency","excess noise"],"falsifier":"Replace the attenuators with a real multi-kilometer fiber spool of the same loss and re-measure the finite-size secret-key rates of the four-channel and single-channel systems; if the four-channel advantage disappears or the excess noise rises faster than predicted, the claim fails.","tokens_in":8812,"feed_emoji":"🔑","tokens_out":782,"duration_ms":7013,"temperature":0.7,"pith_summary":"Continuous-variable quantum key distribution (CV-QKD) encodes secret bits on the amplitude and phase of ordinary laser light so that ordinary telecom hardware can be used. The paper shows that packing several narrowband (10-Mbaud) Gaussian-modulated channels into the same detector bandwidth by frequency-division multiplexing (FDM) multiplies the total secret-key rate far more effectively than simply raising the symbol rate of a single channel. In a four-channel laboratory demonstration that used a transmitted local oscillator and homodyne detection, the back-to-back rate rose by a factor of 3.7 under realistic finite-size analysis; the multi-channel system remained faster than the single-channel baseline out to 41.1 km. The advantage comes from keeping every sub-channel inside the frequency window where detector shot noise still dominates electronic noise, while low-pass filtering and a 40-MHz channel spacing keep crosstalk under control. A sympathetic reader cares because the same spectral budget can therefore support either more users or a higher aggregate key rate without new optical hardware.","feed_headline":"Four narrow FDM channels triple CV-QKD key rate","feed_subtitle":"Dense low-baud packing beats a single wideband channel out to 41 km under finite-size security","key_machinery":"The FDM packing rule (first IF / baud rate ≥ 6.4 and channel spacing = 4 \times baud rate) together with fourth-order Bessel low-pass filters that suppress side-lobe overlap; this rule keeps every sub-channel inside the detector’s shot-noise-dominated band while limiting inter-channel excess noise growth to a factor of only 1.11 when the channel count doubles from two to four.","core_discovery":"Dense frequency-division multiplexing of low-symbol-rate Gaussian-modulated CV-QKD signals, with the first intermediate frequency at least 6.4 times the baud rate and channel spacing four times the baud rate, yields a nearly proportional secret-key-rate gain (3.7-fold for four 10-Mbaud channels under finite-size analysis with N=10^7) and outperforms both a single channel and a single higher-baud-rate frequency-upconverted signal that occupies the same total bandwidth.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["Four 10-Mbaud FDM channels deliver 3.6x CV-QKD secret key rate","Dense low-baud FDM packing yields 3.6-fold CV-QKD SKR gain","Optimized FDM spacing of Gaussian CV-QKD triples back-to-back rate","Four narrow FDM signals outperform single wideband CV-QKD to 26 km","Low-symbol-rate FDM-CV-QKD achieves nearly proportional SKR scaling"],"cache_read_input_tokens":128,"weakest_assumption_plain":"All channel loss is treated as pure attenuation that can be dialed in with variable optical attenuators, so that excess-noise numbers measured back-to-back remain valid at every distance; real fiber dispersion, polarization drift and Raman noise are ignored.","fun_headline_variants_meta":{"raw":{"variants":["Four 10-Mbaud FDM channels deliver 3.6x CV-QKD secret key rate","Dense low-baud FDM packing yields 3.6-fold CV-QKD SKR gain","Optimized FDM spacing of Gaussian CV-QKD triples back-to-back rate","Four narrow FDM signals outperform single wideband CV-QKD to 26 km","Low-symbol-rate FDM-CV-QKD achieves nearly proportional SKR scaling"]},"model":"grok-4.5","effort":"low","cost_usd":0.006732,"raw_usage":{"total_tokens":1648,"prompt_tokens":693,"num_sources_used":0,"completion_tokens":123,"cost_in_usd_ticks":67320000,"prompt_tokens_details":{"text_tokens":693,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":832,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":693,"tokens_out":123,"duration_ms":6877,"temperature":1.0,"reasoning_tokens":832,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-13T21:17:03.698157+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Replace the attenuators with a real multi-kilometer fiber spool of the same loss and re-measure the finite-size secret-key rates of the four-channel and single-channel systems; if the four-channel advantage disappears or the excess noise rises faster than predicted, the claim fails.","supporting_citations":[],"review_version":1}