{"id":"df784ad4-e494-4e6c-9044-339f83c715cd","arxiv_id":"2411.10021","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A fully fiberized feedforward phase noise eater reduces laser frequency noise by over 20 dB in the 1-10 MHz range and stabilizes a Ramsey signal on cold 87Rb atoms.","lead":"This paper builds a fiber-based device that measures and then cancels fast laser phase noise using a delay-line interferometer and a fast modulator. It reports more than 20 dB of noise suppression between 1 and 10 MHz, and shows the cleaned lasers improve a Ramsey measurement on cold rubidium atoms.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the reader's delay-line-noise concern is already bounded by the measured corrected PSD, so the central suppression claim stands; only reporting gaps remain.","rationale":"The paper's central claim is an instrument demonstration: a fiberized delayed-MZI feedforward loop suppresses laser frequency noise by >20 dB in 1–10 MHz, reaching ~31.5 dB at 3–4 MHz, with residual ~0.5 Hz^2/Hz. For this to fail, the feedforward measurement would have to be contaminated by noise not present on the main beam, or the characterization would have to be blind to the residual. The reader singled out the former: delay-line and MZI-path noise. I examined whether this could survive the paper's own data. The correction path's 10 m delay line and the first MZI's internal paths are upstream of the EOM; the second MZI is downstream. Any phase noise added by the delay line to the main beam, or imprinted by the first MZI's readout noise through the feedforward, becomes phase modulation on the beam entering the second MZI. The second MZI is sensitive to exactly such modulation in the 1–10 MHz band (first null at 44.6 MHz). Therefore the measured corrected PSD at 0.5 Hz^2/Hz is an upper bound on the sum of those noise contributions. Literature estimates for fiber acoustic noise are not even needed to bound this; the in-situ spectrum does it. I therefore disagree with the reader's weakest_assumption as load-bearing.\n\nThe more legitimate concerns are about the absolute noise floor. The abstract claims <0.1 Hz^2/Hz, but Sec. III.C.b quotes a photon-shot-noise-equivalent floor of 0.2 Hz^2/Hz at 200 µW and a transimpedance-amplifier electronic floor of 0.1 Hz^2/Hz; adding these would exceed the abstract value. This inconsistency does not threaten the suppression ratio (the same second MZI measures both with and without correction), but it does threaten the 'measurement noise floor' part of the abstract's headline. The absence of error bars on the 20/30 dB figures is a precision gap rather than a correctness gap, given the large dynamic range between the uncorrected 700 Hz^2/Hz and the corrected floor.\n\nThe model agreement in Fig. 4(b) is an independent check because it is built from measured component parameters rather than fit to the final curve, and the Ramsey atom experiment is an external, if qualitative, validation. On the basis of the text, the central claim is credible. The CONDITIONAL verdict should remain, but the conditions are about reporting precision (noise-floor reconciliation, error bars), not about a suspected flaw in the feedforward physics.","tokens_in":15852,"tokens_out":27729,"duration_ms":291204,"concrete_test":"Reconcile the noise-floor claim by re-deriving the photon-shot-noise-equivalent frequency noise floor from the stated optical power, photodetector responsivity, and MZI delay τ=22.4 ns, and by measuring the second MZI's dark floor with light injected; if the floor is above 0.1 Hz^2/Hz, the abstract's 'measurement noise floor of less than 0.1 Hz^2/Hz' should be revised, and the corrected PSD of 0.5 Hz^2/Hz should be re-evaluated relative to this floor.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After a good-faith reading, I do not find a load-bearing objection to the central claim. The reader's candidate weakest assumption is that the 10 m delay line and the first MZI paths add negligible phase noise at 1–10 MHz. This concern does not land, because the second MZI located after the EOM measures the actual corrected beam. If the delay line or the first MZI path noise were imprinted onto the main beam, it would appear in the corrected output PSD of Fig. 1(c). The reported residual of ≈0.5 Hz^2/Hz at 3–4 MHz therefore provides an in-situ upper bound on any such added noise at those frequencies, independent of the literature estimates cited in Sec. IV.C. The suppression claim is further corroborated by the component-level model of Fig. 4(b) and by the qualitative Ramsey-fringe revival in Sec. V. The remaining issues are reporting gaps: the noise-floor statements in the abstract and Sec. III.C.b are not reconciled (shot-noise-equivalent floor quoted as 0.2 Hz^2/Hz while the abstract claims <0.1 Hz^2/Hz), and the headline suppression numbers lack error bars. These affect precision and reproducibility, but they do not invalidate the demonstrated >20 dB suppression, so the central claim is not threatened.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a fully fiberized instrument that measures and corrects fast (sub-microsecond) laser phase noise. A fraction of the light is sent to a delayed Mach-Zehnder interferometer (MZI) used as a frequency discriminator; the resulting electronic signal is low-pass filtered, amplified, and fed forward to a fiber electro-optic modulator acting on the main beam, which is delayed in fiber to match the electronic latency. The authors report a reduction of the frequency-noise PSD of a DBR laser from about 700 Hz^2/Hz to about 0.5 Hz^2/Hz at 3-4 MHz (31.5 dB), more than 20 dB suppression over 1-10 MHz, and a detection noise floor below 0.1 Hz^2/Hz in some configurations. They also characterize the fast frequency noise of five laser types and show that the phase noise eater improves the contrast of a Ramsey sequence on cold 87Rb atoms driven on a Raman transition.","tokens_in":15933,"tokens_out":12423,"duration_ms":128077,"significance":"If the quantitative claims are reliable, the instrument provides a practical, waveguide-based solution to a recognized bottleneck for Rydberg and Raman quantum gates, with a simple calibration procedure and a component-level model (Eq. (7), Fig. 4) that reproduces the observed cancellation. The independent confirmation of the correction by a second MZI and by an atomic Ramsey signal is a notable strength, as is the direct comparison with a commercial phase-noise analyzer in Fig. 3(e). The paper should be useful to experimental groups seeking MHz-bandwidth phase-noise reduction without high-finesse cavity filtering.","major_comments":[{"comment":"The measurement-noise-floor claim is internally inconsistent. The abstract asserts a noise floor below 0.1 Hz^2/Hz, while Sec. III.C.b quotes an electronic floor of about 0.1 Hz^2/Hz for the transimpedance-amplified detector and a photon-shot-noise floor of 0.2 Hz^2/Hz, which combine to roughly 0.22 Hz^2/Hz in quadrature. The paper must specify which detector configuration produced the dark-blue floor in Fig. 1(c) and reconcile these numbers; if the floor in the feedforward characterization is ~0.2 Hz^2/Hz, the '<0.1 Hz^2/Hz' statement in the abstract should be qualified to the configuration in which it was actually achieved.","section":"Abstract and Sec. III.C.b"},{"comment":"The headline suppression values (31.5 dB at 3-4 MHz, 'more than 20 dB' over 1-10 MHz, 'up to 30 dB at 3 MHz') are quoted without error bars, confidence intervals, or repeatability information. These are the central quantitative claims of the paper. Please provide at least the shot-to-shot spread of the measured PSD ratio, or an uncertainty budget propagated from the calibration of Vpp, tau, and the spectrum-analyzer settings, so the reader can assess whether the 20 dB and 30 dB claims are statistically distinct.","section":"Sec. II and Fig. 1(c)"}],"minor_comments":[{"comment":"The sentence 'The noise is remains below 10 Hz^2/Hz over the entire 1 to 10 MHz region, with the correction decreasing outside this range' is garbled; it should read 'The noise remains below 10 Hz^2/Hz ... with the correction degrading outside this range.'","section":"Sec. II"},{"comment":"The sentence 'For all lasers, expect the VECSEL' contains a typo: 'expect' should be 'except'.","section":"Sec. III.D"},{"comment":"The statement 'Together, the electronic and photon shot noise add up to a detection limit ... giving 32 dB of signal-to-noise ratio' should specify the reference signal level and Fourier frequency used for the 32 dB number; otherwise it is ambiguous.","section":"Sec. III.C.b"},{"comment":"The text says 'correction C(f) phi applied by the phase modulator' but C(f) multiplies the Fourier component of phi; please use a tilde or state explicitly that the equation is in the Fourier domain.","section":"Sec. IV.A, Eq. (7)"},{"comment":"The 'sum of these' model curve should state whether the four independent imperfections are added in linear units, in quadrature, or in dB; the current wording does not specify the combination rule.","section":"Sec. IV.C and Fig. 4(b)"},{"comment":"The discussion of fiber delay-line noise would be clearer with an explicit statement of why the 10^2 Hz^2/Hz acoustic-fiber-noise levels from Refs. [42,43] do not apply at 1-10 MHz in the present setup.","section":"Sec. IV.C"},{"comment":"The Ramsey revival is compelling but qualitative; given the paper's own admission that the atom interrogation 'was not very sensitive', the text should state explicitly that the atom result is an illustrative demonstration rather than a quantitative fidelity measurement.","section":"Sec. V"}],"recommendation":"major_revision","confidential_remarks":"This is a useful and basically sound instrumentation paper. The feedforward concept, the component-level model, and the independent atomic check all point to a valid central result. My main concern is that the headline quantitative claims are not yet presented with the precision they deserve: the noise-floor numbers are mutually inconsistent and the suppression values lack error bars. The atom experiment is only qualitative, but the paper already acknowledges this. These issues are fixable and do not require new physics, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a solid instrument paper that delivers what it claims, and the central suppression result survives scrutiny. The genuinely new part is not the principle—delayed MZI plus feedforward EOM has been around, and the authors say so—but the fully fiberized implementation with a documented noise floor, >20 dB suppression over 1–10 MHz (31.5 dB at 3–4 MHz), and a Ramsey demonstration on 87Rb. That combination is absent from the cited literature and is what makes the paper worth publishing.\n\nThe paper does several things well. The calibration procedure (Vpp and tau from MZI nulls) is simple and reproducible; Eq. (7)'s feedforward transfer function is built from independently calibrated component responses, so there is no circularity. The model in Fig. 4(b) matches the measured suppression across the band, and the Ramsey revival in Sec. V is an external consistency check. The authors are also honest about what is not new and about the atom test requiring a deliberately worsened lock.\n\nSoft spots are real but not load-bearing. The headline numbers have no error bars; a single representative spectrum is shown. The abstract says <0.1 Hz^2/Hz measurement noise floor, while Sec. III.C.b quotes the transimpedance-amplifier floor at ~0.1 Hz^2/Hz and photon shot noise at 0.2 Hz^2/Hz; these are not reconciled, and a careful reader cannot tell which floor applies to the corrected PSD. The delay-line-noise worry (does the 10 m fiber add phase noise?) is not fatal: the second MZI measures the actual corrected beam, so Fig. 1(c) bounds any such imprint at 1–10 MHz; the literature estimates only matter below 100 kHz where the feedforward is intentionally inactive. The atom interrogation is qualitative—a useful sanity check, not a metrology-grade demonstration. There is no raw data or code, but the parts list and calibration procedure are enough for reproduction.\n\nWho is this for? AMO experimentalists building Raman or Rydberg gate systems with ECDLs; also laser developers who want a comparison of VECSEL, SIL, DBR, and ECDL noise floors around 1–10 MHz. A serious referee can add value on the noise-floor bookkeeping and error bars; I would not desk-reject. My verdict: send it to peer review. Conditional acceptance is fair if the reporting gaps are addressed.","headline":"Solid instrument paper: the fully fiberized phase noise eater achieves the claimed >20 dB suppression at 1–10 MHz and the central result survives scrutiny; only reporting gaps need fixing.","tokens_in":16643,"tokens_out":2641,"would_cite":true,"duration_ms":27618,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper reports a fully fiberized phase noise eater that measures fast laser phase noise with a delayed Mach-Zehnder interferometer and feedforward-corrects it with a fiber electro-optic modulator, cutting frequency-noise PSD by more…","keywords":["laser phase noise","frequency noise PSD","feedforward correction","Mach-Zehnder interferometer","electro-optic modulator","Rydberg quantum gates","Raman Ramsey interferometry","fiberized optics"],"falsifier":"Feed a laser whose own noise is far below the claimed floor (for instance a TiSapph laser) into the phase noise eater, or block the main beam while leaving the interferometer lit, and measure the output PSD with a second independent interferometer; if the delay line or the first interferometer adds noise at 1-10 MHz, the output PSD will rise above the 0.1 Hz²/Hz floor instead of staying at the detection limit.","tokens_in":15503,"feed_emoji":"⚛️","tokens_out":7083,"duration_ms":65292,"temperature":0.7,"pith_summary":"Lasers used to drive quantum gates in neutral atoms and trapped ions carry fast phase noise that standard feedback loops cannot remove, because loop latency limits correction to roughly a few MHz. This paper builds a fully fiberized 'phase noise eater' that measures the laser's frequency noise with a short-delay Mach-Zehnder interferometer and then applies an inverted copy of the noise to a fiber electro-optic modulator placed after a matching fiber delay. The result is a measured frequency-noise floor below 0.1 Hz²/Hz and a suppression of more than 20 dB in the 1 to 10 MHz Fourier range, with a peak of about 30 dB near 3 MHz where the noise is pushed to roughly 0.5 Hz²/Hz. The authors also show that the correction visibly improves a Ramsey interference signal on cold rubidium atoms driven by two such stabilized lasers. If the method holds up, it lets ordinary extended-cavity diode lasers serve in high-fidelity Rydberg and Raman gates without switching to far more expensive laser architectures.","feed_headline":"Feedforward loop cuts laser noise 20 dB at 1-10 MHz","feed_subtitle":"A fiber interferometer plus phase modulator pushes frequency noise below 1 Hz²/Hz, protecting Rydberg and Raman gates.","key_machinery":"The load-bearing object is the delayed Mach-Zehnder interferometer used as a frequency discriminator, whose output voltage is proportional to the phase difference $\\phi(t)-\\phi(t-\\tau)$ and, in the Fourier domain, has sensitivity $s(f)=s_0 \\sin(\\pi f \\tau)/(\\pi f \\tau) e^{-i\\pi f \\tau}$. A low-pass filter with cut-off $f_c$ converts the frequency-proportional signal into a phase-proportional signal, and the feedforward path applies it to a fiber electro-optic modulator after a roughly 10 m fiber delay that matches the electronics delay. The correction transfer function $C(f)$ is the product of the interferometer response, the filter, the modulator $V_\\pi$, the adjustable gain, and the adjustable delay; the noise PSD is reduced by $|1-C|^2$, so the whole scheme works by setting gain within about 3% and electrical delay within about 0.5 ns while keeping the Fourier frequency inside the band $30 f_c < f < 0.15/\\tau$, here roughly a 60 kHz low-pass cutoff and a 20-22 ns interferometer delay.","core_discovery":"The central claim is that a delayed Mach-Zehnder interferometer can act as both a sensitive fast frequency discriminator and the front end of a feedforward correction loop, and that the loop can cancel sub-microsecond laser phase noise at levels relevant to Rydberg and Raman gates. Operated at quadrature and calibrated through the interference fringe peak-to-peak voltage and the delay extracted from sensitivity nulls, the interferometer converts frequency fluctuations into a voltage with a known frequency-dependent response; a low-pass filter integrates that signal to recover phase fluctuations, and a fiber electro-optic modulator under a precisely matched gain and delay applies the correction. The paper reports the frequency-noise PSD of the corrected laser dropping from about 700 Hz²/Hz to as low as 0.5 Hz²/Hz at 3-4 MHz, a 31.5 dB reduction, with more than 20 dB suppression across 1-10 MHz, and it confirms the measured noise reduction matches the model built from the interferometer roll-off, filter dispersion, delay mismatch, and detection noise floor. On a Raman Ramsey sequence on 87Rb, activating the phase noise eater removes the periodic contrast collapse caused by a strong servo-bump in one laser.","pith_inferences":["A direct extension would be to use the same balanced interferometer output to cancel intensity noise at megahertz rates, since the paper already quantifies the residual amplitude modulation the electro-optic modulator adds.","With an automatic gain-control loop on the variable optical attenuator, the device could hold its optimum cancellation as laser power drifts, moving it from a laboratory setup toward a turnkey instrument.","The published acoustic-noise estimates for fibers leave open how far the approach can be pushed at lower Fourier frequencies; an in-situ measurement of the 10 m delay-line noise would settle the practical floor for metrology-style applications.","If the delay line is as quiet as the estimates suggest, cascading two feedforward stages could push cancellation beyond 30 dB or extend the corrected band above 10 MHz, where Rydberg-gate sensitivity falls but other fast processes may matter."],"forward_implications":["An extended-cavity diode laser equipped with the phase noise eater can deliver frequency noise below about 1 Hz²/Hz in the 1-10 MHz band, replacing more costly VECSEL or self-injection-locked lasers in demanding gate applications.","Combined with a standard Pound-Drever-Hall feedback loop, the phase noise eater covers a noise spectrum from a few Hz to 10 MHz, offering a single low-noise source for Rydberg and molecular-state manipulation.","The 0.1 Hz²/Hz measurement floor lets the same fiberized interferometer serve as a fast laser-noise diagnostic, ranking lasers whose white-noise floors span eight orders of magnitude.","The demonstrated Raman Ramsey improvement indicates that phase-noise-driven contrast loss in two-photon atomic transitions can be inverted by feedforward correction."],"supporting_citations":[{"why":"Supplies the linear-response model showing Rydberg gate fidelity is most sensitive to frequency noise at 1-10 MHz, defining the target band.","marker":"[11]"},{"why":"Demonstrates that suppressing diode-laser phase noise improves Rydberg excitation, the benchmark this work extends.","marker":"[13]"},{"why":"Shows how phase noise can be extracted from a Pound-Drever-Hall error signal, the alternative measurement the Mach-Zehnder interferometer is compared against.","marker":"[12]"},{"why":"Provides the numerical phase-noise reconstruction and gate-error estimates used to model Ramsey fringe contrast and noise budgets.","marker":"[14]"},{"why":"Reports a prior feedforward phase-noise suppression using a cavity-filtered heterodyne reference, a performance baseline for the fiberized version.","marker":"[32]"},{"why":"Demonstrates the same delayed-interferometer feedforward concept on a DBR laser, with 15 dB suppression up to 100 MHz, the closest earlier implementation.","marker":"[29, 30]"},{"why":"Achieves 30 dB feedforward cancellation using an adaptive Pound-Drever-Hall error signal, providing the main contemporary comparison.","marker":"[41]"},{"why":"Provides literature estimates of acoustic-induced fiber frequency noise used to argue the 10 m delay line is negligible in the 1-10 MHz band.","marker":"[42, 43]"},{"why":"Describes the 87Rb cold-atom platform on which the Raman Ramsey interrogation is performed.","marker":"[45]"}],"fun_headline_variants":["Laser phase noise eater: 20 dB suppression at MHz rates","Feedforward cancels laser noise 30 dB at 3 MHz","Rydberg gate lasers quieted 20 dB from 1-10 MHz","Fiber device cuts laser noise 20 dB for quantum gates","Fiber loop fixes sub-microsecond laser phase jitter"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole correction rests on the assumption that the phase noise measured by the first interferometer is the same as the noise carried by the main beam after the 10 m fiber delay, so that the correction signal does not inject extra noise from the interferometer paths or the delay line itself; the authors rely on literature estimates rather than an in-situ measurement of the delay-line noise.","fun_headline_variants_meta":{"raw":{"variants":["Laser phase noise eater: 20 dB suppression at MHz rates","Feedforward cancels laser noise 30 dB at 3 MHz","Rydberg gate lasers quieted 20 dB from 1-10 MHz","Fiber device cuts laser noise 20 dB for quantum gates","Fiber loop fixes sub-microsecond laser phase jitter"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001327,"raw_usage":{"total_tokens":5401,"prompt_tokens":950,"completion_tokens":4451,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":566,"completion_tokens_details":{"reasoning_tokens":4360}},"tokens_in":566,"tokens_out":4451,"duration_ms":30100,"temperature":1.0,"reasoning_tokens":4360,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T20:04:25.997999+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Feed a laser whose own noise is far below the claimed floor (for instance a TiSapph laser) into the phase noise eater, or block the main beam while leaving the interferometer lit, and measure the output PSD with a second independent interferometer; if the delay line or the first interferometer adds noise at 1-10 MHz, the output PSD will rise above the 0.1 Hz²/Hz floor instead of staying at the detection limit.","supporting_citations":[{"cited_title":"Levine, A","cited_arxiv_id":null,"evidence_quote":"Demonstrates that suppressing diode-laser phase noise improves Rydberg excitation, the benchmark this work extends."},{"cited_title":"de L´ es´ eleuc, D","cited_arxiv_id":null,"evidence_quote":"Shows how phase noise can be extracted from a Pound-Drever-Hall error signal, the alternative measurement the Mach-Zehnder interferometer is compared against."},{"cited_title":"Jiang, J","cited_arxiv_id":null,"evidence_quote":"Provides the numerical phase-noise reconstruction and gate-error estimates used to model Ramsey fringe contrast and noise budgets."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reports a prior feedforward phase-noise suppression using a cavity-filtered heterodyne reference, a performance baseline for the fiberized version."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Describes the 87Rb cold-atom platform on which the Raman Ramsey interrogation is performed."}],"review_version":1}