{"id":"f9f270cf-a957-4476-9d3a-1560b4b25dd2","arxiv_id":"2502.02161","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A plug-and-play two-way free-space link characterization system reaches 2.0e-19 fractional frequency instability at 10 s averaging over a 3.4 km folded intra-city path, with 94% uptime over 15 hours.","lead":"This paper demonstrates a portable, rack-integrated system for characterizing two-way optical frequency transfer over free-space links. On a 3.4 km folded intra-city link, it measured a fractional frequency instability of 2.0e-19 at 10 s averaging, with 94% uptime over 15 hours, a step toward practical optical clock comparisons.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline 2e-19 instability is measured on a fully folded link where delay-induced non-reciprocity cancels by construction; it is a technical floor, not the atmospheric non-reciprocity of a separated point-to-point link, and the paper contains no test of the extrapolation to real clock…","rationale":"I agree with the reader's weakest_assumption. The folded geometry is exactly what makes the measured residual so small, and the paper's own estimate of the unfolded delay limit depends on the unverified representativeness of folded-path turbulence. The AM-PM fit and best-run selection are secondary; even with perfect AM-PM calibration, the result would not establish the point-to-point capability. Because the paper is transparent that the link is fully folded and because the measurement itself is internally consistent, the appropriate verdict remains conditional: accept that a low technical floor was demonstrated on a folded link, but require a point-to-point test before the clock-comparison implication is credited. Thus I recommend UNCHANGED relative to the reader's CONDITIONAL verdict.","tokens_in":11751,"tokens_out":11394,"duration_ms":121847,"concrete_test":"Replace the STW FOR with a second, identical plug-and-play transceiver/AOM+heterodyne chain so that the forward and backward beams propagate over two independent 1.7 km paths, and repeat the two-way measurement on a day with C_n^2 comparable to measurement 1. Compare the 10 s MDEV with the folded-link value and with the red-dashed delay limit computed from the simultaneously measured one-way phase noise. If the unfolded MDEV at 10 s is close to 2e-19, the folded-link floor is representative; if it is significantly larger, the headline number is folded-geometry-specific and the clock-comparison claim is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Load-bearing concern: The 2.0e-19 at 10 s is measured on the fully folded link described in Sec. 2. In that geometry, the forward and backward beams occupy the same air path, so Eq. (2) isolates only the residual Δν_fs^21−Δν_fs^12 after the delay-induced non-reciprocity cancels exactly; the paper states this cancellation explicitly. The quoted instability is therefore a bound on the co-located technical floor (shot noise plus AM-PM at the receivers), not a measurement of the non-reciprocity that a separated two-way point-to-point link would contribute to a clock comparison. The Sec. 3 estimate of the unfolded delay limit does not close this gap: it uses the free-running noise recorded on the folded path and asserts that the only added term in an unfolded link is the red-dashed delay curve. That assumes the amplitude-scintillation statistics and the receiver noise of a round-trip folded path are representative of two one-way paths measured by independent terminals, and that the technical noise at both ends is common-mode as in this co-located rack. Section 4 concedes that only adaptations to genuine point-to-point links are 'straightforward', i.e., the point-to-point case was not measured. Without a point-to-point test, the abstract's implied compatibility with optical clock comparisons is an extrapolation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a portable, rack-integrated two-way free-space optical frequency transfer characterization system based on a single laser split into two paths with different AOM shifts, and a fully folded 3.4 km intra-city free-space link formed by a retroreflector at the remote terminal. The key reported result is a fractional frequency instability of 2.0e-19 at 10 s averaging time (best run, measurement 1), with an uptime of 94% over 15 hours using a defined validity threshold. The authors analyze the residual non-reciprocity using Eq. (2) and attribute short-term instability to shot noise (via a parameter-free SNR estimate) and long-term instability to AM-PM conversion, calibrating the phase-versus-power response in the laboratory and applying it to measured amplitude fluctuations. The paper concludes that the system is suitable for optical clock comparisons over free space.","tokens_in":12108,"tokens_out":6786,"duration_ms":62253,"significance":"If the result is taken as a characterization of a technical noise floor, the paper provides a valuable, cleanly measured demonstration of how low the co-located residual noise of a two-way free-space link characterization system can be: the direct beat-note measurements, the short-fiber noise-floor control, and the parameter-free shot-noise estimate are all strengths. The measured 2.0e-19 at 10 s, even if it is the best of several runs, is a real and impressive experimental result. However, the significance for actual optical clock comparisons is conditional: because the fully folded geometry cancels delay-induced non-reciprocity exactly, the headline number does not directly measure the atmospheric non-reciprocity that a separated point-to-point link would contribute. The paper's extrapolation to that case rests on untested assumptions about the representativeness of the folded-path free-running noise and the common-mode nature of technical noise in two independent terminals. The AM-PM analysis is a plausible in-sample explanation but is not an independent validation. These limitations are acknowledged partially in Sec.","major_comments":[{"comment":"The abstract's headline claim of 2.0e-19 at 10 s is the best of six runs; the paper's own Sec. 3 states that the 10-s MDEV across the six longest runs varies between 2.0e-19 and 1.1e-18 (Fig. 3a). Presenting the best run without qualification in the abstract overstates the achieved performance. The abstract should report the range or explicitly label the value as the best continuous measurement.","section":"Abstract and Sec. 3 (Fig. 3a)"},{"comment":"In the fully folded geometry, the forward and backward beams traverse the same air path, so the delay-induced non-reciprocity terms in Eq. (2) cancel exactly, as the paper states. The measured 2.0e-19 is therefore a co-located technical noise floor (shot noise, AM-PM, and residual frequency-dependent phase shifts), not the atmospheric non-reciprocity that a separated point-to-point two-way link would contribute to a clock comparison. The Sec. 3 estimate of the delay limit for an unfolded link does not close this gap: it assumes the free-running noise recorded on the folded path and the co-located receiver noise are representative of two one-way paths measured by independent terminals. The paper should either provide a point-to-point test or explicitly frame the result as a technical-floor characterization, and temper the abstract and conclusion accordingly.","section":"Sec. 2, 'fully folded free-space link' paragraph; Eq. (2)"},{"comment":"The AM-PM explanation of the long-term instability is supported only by an in-sample consistency check: the phase-versus-power curve is fitted to laboratory data and then applied to the amplitude fluctuations of measurement 1 to reproduce the same run's MDEV. This does not independently validate the AM-PM mechanism, and the same method cannot be assumed to transfer to a point-to-point configuration where the two directions experience different amplitude statistics, as the paper itself notes in Sec. 4 that only 'adaptations' are 'straightforward'. The manuscript should acknowledge that the agreement is consistent with, but does not uniquely prove, the AM-PM mechanism.","section":"Sec. 3, Fig. 4(b) and Fig. 5"},{"comment":"The conclusion that 'adaptations for minimizing interferometric noise contributions in genuine point-to-point two-way free-space links ... are straightforward to implement' is not supported by any measurement in the manuscript. The system was tested only on a fully folded link and a short-fiber noise floor; no genuine point-to-point two-way free-space measurement is reported. This claim should be softened or supported by additional data, and the sentence should not imply that the reported stability directly characterizes a point-to-point comparison.","section":"Sec. 4, Conclusion"}],"minor_comments":[{"comment":"The paper says 'The entire setup is interconnected using polarization-maintaining (PM) fiber' but later describes a 5 m SM fiber connection to the optical terminal and a polarization controller at the PM-to-SM transition; please clarify which parts of the setup are PM and which are SM.","section":"Sec. 2, first paragraph"},{"comment":"The uptime is defined with a threshold of ±5e-17 for invalid data; the paper states the rationale for this threshold, but it would improve clarity to report the sensitivity of the 94% uptime figure to the threshold choice, for example by giving uptime for a range of thresholds.","section":"Sec. 3, Fig. 6"},{"comment":"The fitted coefficients a_L^Φ, b_L^Φ, c_L^Φ, d_L^Φ, a_H^Φ, b_H^Φ, and c_H^Φ are not listed; providing these numerical coefficients would aid reproducibility of the AM-PM calculation.","section":"Appendix D, Eq. (14)"},{"comment":"The phrase 'zero compatible operation' should be reworded to 'compatible with zero' for clarity.","section":"Sec. 3, near Fig. 4(b)"},{"comment":"The abstract's uptime claim of 94% over 15 hours depends on the validity threshold defined in Sec. 3; adding a qualifier such as 'using the ±5e-17 threshold' in the abstract or a footnote would avoid ambiguity.","section":"Abstract and Sec. 3, Fig. 3(b)"}],"recommendation":"major_revision","confidential_remarks":"The measurements are sound and the paper is honest about the folded geometry in Sec. 2, but the abstract and conclusion overstate the relevance of the 2.0e-19 result to point-to-point optical clock comparisons. The main revision should reframe the headline as a technical noise floor and either add a genuine point-to-point test or clearly state that the point-to-point case is an extrapolation. I do not think this requires rejection, because the experimental work is valuable and the central derivation (Eq. 2) is correct; the issue is the scope of the claim, which can be fixed in revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is worth your time if you work on free-space optical frequency transfer or chronometric geodesy. The genuinely new part is the rack-integrated, plug-and-play two-way CW system: a portable instrument with a clean noise budget, deployed on a 3.4 km folded intra-city link with adaptive-optics terminals. The shot-noise model matches the short-term MDEV, the short-fiber noise floor is a sensible control, and the AM-PM lab characterization plausibly explains the run-to-run spread in long-term stability. That is a genuine engineering contribution, and the paper is transparent about most of its own limitations.\n\nThe soft spots are real but not fatal. The abstract's headline 2.0e-19 at 10 s comes from the best of six runs; Fig. 3 shows the 10-s MDEV spanning 2.0e-19 to 1.1e-18 across runs. The paper does report that spread in the body, but the abstract overstates the typical performance. More important, the fully folded geometry cancels delay-induced non-reciprocity by construction, as the paper explicitly states. So the 2e-19 is a technical floor—shot noise plus AM-PM—not the atmospheric non-reciprocity of a separated point-to-point link. The authors estimate the delay limit for a hypothetical unfolded link and argue it would not dominate, but that is an extrapolation; no point-to-point measurement is reported. The stress-test note is correct on this: the relevance to real clock comparisons depends on the untested assumption that the folded-path noise statistics carry over to two independent one-way paths. The AM-PM explanation also has a post-hoc flavor: a phase-versus-power curve is fit to controlled lab data and then used to reproduce the same run's long-term MDEV. That is a reasonable diagnostic, but it is not a prediction. No raw data or code are public, which is standard for this community but limits independent verification.\n\nThe central engineering claim holds up: the system works, the noise floor is well characterized, and the uptime statistics are meaningful. The paper is honest about the folded geometry and the open questions. I would send it to a serious referee in applied optics, asking for an abstract that distinguishes the folded-link technical floor from point-to-point performance, and for an explicit discussion of what would be needed to close that gap. For readers building or using free-space two-way links, this is a useful reference. I would cite it for the system design and noise budget.\n\nRecommendation: engage with it; referee it, but push for the abstraction to be qualified and for a clearer separation between measured technical floor and estimated point-to-point performance.","headline":"Solid engineering demonstration of a two-way free-space link characterizer, but the headline 2e-19 is a best-run technical floor on a fully folded link, not a point-to-point atmospheric non-reciprocity measurement.","tokens_in":12645,"tokens_out":1997,"would_cite":true,"duration_ms":21479,"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 two-way free-space optical link hits 2e-19 fractional frequency instability at 10 seconds of averaging.","keywords":["two-way optical frequency transfer","free-space optical link","optical clock comparison","fully folded link","residual non-reciprocity","modified Allan deviation","atmospheric turbulence","AM-PM conversion"],"falsifier":"Replace the fiber-optic retroreflector at the far site with two separate terminals at both ends of a 3.4 km point-to-point link and measure the same beat-note combination. If the residual non-reciprocity stays at the $10^{-19}$ level at 10 s under comparable turbulence, the folded-link result represents a true point-to-point capability; if it rises by orders of magnitude toward the calculated delay limit, the reported number is specific to the folded geometry.","tokens_in":11598,"feed_emoji":"🔭","tokens_out":6312,"duration_ms":53469,"temperature":0.7,"pith_summary":"This paper reports a portable, rack-integrated system that characterizes the residual non-reciprocity of a free-space optical link used for optical clock comparisons. On a 3.4 km fully folded intra-city link, the system reaches a fractional frequency instability of $2.0 \\times 10^{-19}$ at 10 s averaging, with 94% uptime over 15 hours. The key point is that the fully folded geometry lets the forward and return beams share the same air path, so delay-induced non-reciprocity cancels and the measurement exposes the technical noise floor set by shot noise and amplitude-to-phase conversion. A sympathetic reader would care because this is the level of stability needed to compare state-of-the-art optical clocks, and the plug-and-play design makes the technique deployable in field experiments, including chronometric geodesy.","feed_headline":"Folded laser link hits 2e-19 frequency stability at 10 seconds","feed_subtitle":"Two-way free-space link stays stable enough for optical clock comparisons on a 3.4 km city path.","key_machinery":"The central object is the two-way free-space link characterization setup: a Michelson-type interferometer with acousto-optic modulators (AOMs) and fiber-optic retroreflectors, connected to a free-space path under test. The AOM shifts $\\nu_{AOM1}$, $\\nu_{AOM2}$ mimic separate lasers in a real point-to-point comparison, and the three beat notes A, B, C on each photodetector allow pairwise combinations. The load-bearing identity is Eq. (2), $(PD1_A-PD1_B)-(PD2_A-PD2_B)=(\\Delta\\nu^{21}_{fi}-\\Delta\\nu^{12}_{fi})-(\\Delta\\nu^{21}_{fs}-\\Delta\\nu^{12}_{fs})$, which cancels all interferometric noise terms and isolates the residual non-reciprocity of the free-space link. The fully folded link, formed by a fiber-optic retroreflector with an AOM at the far site, makes forward and backward beams traverse exactly the same air path so delay noise cancels; this turns the measurement into a characterization of the technical noise floor.","core_discovery":"The central claim is that a two-way continuous-wave optical frequency transfer scheme, free of interferometric noise and laser phase noise through post-processing, can measure the residual non-reciprocity of a free-space connection at a level compatible with optical clock comparisons. Concretely, combining the A and B beat notes from two balanced photodetectors according to Eq. (2) cancels all common-mode interferometric noise and leaves the difference of the free-space path noises; with the fully folded link the delay-induced part cancels exactly. The measured residual instability is $2.0 \\times 10^{-19}$ at 10 s, reaching $8.5 \\times 10^{-21}$ at 100 s, and the fractional frequency offsets are compatible with zero. The paper identifies shot noise as the short-term limit and AM-PM conversion through the detection electronics as the long-term limit.","pith_inferences":["The paper leaves implicit that the $2.0 \\times 10^{-19}$ figure is a floor for the instrument under folded-path conditions; whether it transfers to a separated point-to-point link depends on the unmeasured assumption that shot noise and AM-PM conversion still dominate once the delay-noise term is added. That assumption is testable by comparing the folded result with the calculated delay limit.","A testable extension is to quantify the AM-PM coefficients of individual receiver stages (limiting amplifier, mixer, counter input) and then verify that the predicted modified-Allan bump around 0.1 s disappears when those coefficients are reduced; the paper identifies the effect but not the specific dominant component.","For future links of tens of kilometers, the cubic scaling of residual noise with link length cited from the delay-limit model means a folded testbed no longer bounds point-to-point performance; a folded link with an inserted fiber delay line could isolate the electronic floor from the atmospheric term without building a second terminal."],"forward_implications":["The system can serve as a testbed for free-space optical clock links: at the reported residual non-reciprocity, 10-second cycle-slip-free intervals are already sufficient for accurate clock comparisons.","The identified limiting factors give a concrete improvement path: reducing AM-PM conversion in the detection electronics and increasing received optical power to lower shot noise would push the technical floor further down.","The plug-and-play interface, connecting through a single-mode fiber patch cable, lets the same rack be tested against different beam-stabilization terminals, so results transfer across link infrastructures.","The 94% uptime over 15 hours, with the turbulence dependence quantified through $C_n^2$, indicates the system can collect the long averages needed for chronometric leveling at the centimeter level."],"supporting_citations":[{"why":"Supplies the two-way optical phase comparison method that the paper's scheme builds on and claims improved noise rejection relative to active noise cancellation.","marker":"[25]"},{"why":"Previous 600 m test and partially folded configuration by the same group; the current fully folded setup extends this work.","marker":"[26]"},{"why":"Gives the delay-noise limit model used to calculate the theoretical residual noise for an unfolded point-to-point link.","marker":"[31]"},{"why":"Justifies the deadtime-free Λ-mode counting and the statistical treatment of frequency offsets and uncertainties.","marker":"[27]"},{"why":"Provides the shot-noise-limited SNR relation used to quantify the short-term white phase noise.","marker":"[32]"},{"why":"Used to convert the phase noise power spectral density into the modified Allan deviation shown in the results.","marker":"[33]"},{"why":"Supplies the comparison for statistical uncertainty: the achieved value is three orders of magnitude below typical optical clock comparison uncertainty over the same period.","marker":"[34]"}],"fun_headline_variants":["Two-way laser link hits 2e-19 stability over 3.4 km","Plug-and-play free-space link hits 2e-19 frequency precision","Portable rack system stabilizes city laser link to 2e-19 at 10s","Folded two-way link achieves 2e-19 instability in urban free-space","Compact two-way system offers 2e-19 stability for free-space clocks"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result depends on the fully folded link making the forward and return beams traverse exactly the same air path, so that delay-induced non-reciprocity cancels and the measured $2.0 \\times 10^{-19}$ is a technical noise floor rather than the atmospheric non-reciprocity of a separated point-to-point link.","fun_headline_variants_meta":{"raw":{"variants":["Two-way laser link hits 2e-19 stability over 3.4 km","Plug-and-play free-space link hits 2e-19 frequency precision","Portable rack system stabilizes city laser link to 2e-19 at 10s","Folded two-way link achieves 2e-19 instability in urban free-space","Compact two-way system offers 2e-19 stability for free-space clocks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000624,"raw_usage":{"total_tokens":2850,"prompt_tokens":868,"completion_tokens":1982,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":484,"completion_tokens_details":{"reasoning_tokens":1875}},"tokens_in":484,"tokens_out":1982,"duration_ms":11853,"temperature":1.0,"reasoning_tokens":1875,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T13:06:50.852613+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Replace the fiber-optic retroreflector at the far site with two separate terminals at both ends of a 3.4 km point-to-point link and measure the same beat-note combination. If the residual non-reciprocity stays at the $10^{-19}$ level at 10 s under comparable turbulence, the folded-link result represents a true point-to-point capability; if it rises by orders of magnitude toward the calculated delay limit, the reported number is specific to the folded geometry.","supporting_citations":[{"cited_title":"Frequency transfer via a two-way optical phase comparison on a multiplexed fiber network,","cited_arxiv_id":null,"evidence_quote":"Supplies the two-way optical phase comparison method that the paper's scheme builds on and claims improved noise rejection relative to active noise cancellation."},{"cited_title":"A two-way free-space link for optical frequency comparisons,","cited_arxiv_id":null,"evidence_quote":"Previous 600 m test and partially folded configuration by the same group; the current fully folded setup extends this work."},{"cited_title":"Tackling the limits of optical fiber links,","cited_arxiv_id":null,"evidence_quote":"Gives the delay-noise limit model used to calculate the theoretical residual noise for an unfolded point-to-point link."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the shot-noise-limited SNR relation used to quantify the short-term white phase noise."},{"cited_title":"The Enrico’s chart for phase noise and two-sample variances,","cited_arxiv_id":null,"evidence_quote":"Used to convert the phase noise power spectral density into the modified Allan deviation shown in the results."},{"cited_title":"Long-distance chronometric leveling with a portable optical clock,","cited_arxiv_id":null,"evidence_quote":"Supplies the comparison for statistical uncertainty: the achieved value is three orders of magnitude below typical optical clock comparison uncertainty over the same period."}],"review_version":1}