{"id":"41db0659-628d-4aea-b0e9-45e9adb35b4a","arxiv_id":"2504.16731","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A 68 cm optical cavity with 300 microsecond storage time and active residual amplitude modulation cancellation at the 10^-7 level reduce RAM-induced fractional laser frequency instability to the 10^-19 range.","lead":"This paper reports a laser-reference cavity that stores light for 300 microseconds and an active method that suppresses a common error source, residual amplitude modulation, to the 10^-7 level. If the results hold, they remove a key technical obstacle for ultra-stable lasers used in atomic clocks and precision measurement.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed 10^-19 RAM-induced frequency instability is inferred via Eq. (1) from open-cavity voltage noise using a DC discriminator calibration that is never validated under the conditions of the RAM measurement.","rationale":"The reader's weakest assumption identifies Eq. (1) and the D calibration as the load-bearing conversion from measured voltage noise to frequency. I agree this is the most critical unvalidated link. The paper's direct measurements (ringdown, RAM cancellation level, discriminator calibration) are described carefully and appear internally consistent, and the claimed RAM cancellation factor is supported by the reported voltage ratios. However, the central quantitative claim that RAM-induced frequency instability reaches 3e-19 is only an inference: the numerator is measured in an open-cavity configuration and the denominator is a DC-calibrated discriminator coefficient, so no measurement directly confirms the frequency scale. The concern is not that the authors are wrong but that the headline number is one step removed from a frequency measurement. Since the reader's CONDITIONAL verdict already captures this and the additional wording issue with the thermal-noise floor, no verdict change is needed. The proposed beat-note or AC-calibration check would settle whether the conversion is valid.","tokens_in":9508,"tokens_out":8444,"duration_ms":81996,"concrete_test":"While the laser is locked to the 68 cm cavity with the RAM servo engaged, record both the beat frequency against an independent cavity-stabilised reference laser (known instability below the level to be tested) and the out-of-loop PDH error signal. Compute the RAM-induced frequency instability via Eq. (1) from the error signal and compare with the measured beat frequency at the same averaging times (1-100 s). If the two disagree beyond the reference uncertainty, the conversion is invalid. A lower-cost check: repeat the D calibration with a small sinusoidal offset injected at 0.01-1 Hz (rather than DC) and test whether the inferred D is frequency-independent; if it changes, Eq. (1) is not applicable to fluctuating RAM noise.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline frequency numbers (1.3e-18, 5e-19, 3e-19) are not measured as laser frequency; they are computed as (Delta V_PDH_RAM / V_PDH_PP)/D. D = (4.7 +/- 0.5)e11 is calibrated by adding a 10% static DC offset to the PDH error signal of a locked laser and measuring the resulting frequency shift against a reference laser. But the RAM fluctuations entering the numerator are measured with an open test cavity (beam stopper between mirrors) in which the PDH signal is not a cavity discriminator; the transfer from RAM-induced photocurrent imbalance to the recorded error signal may differ from the locked-cavity path. The DC calibration also does not establish that time-varying RAM noise at the Fourier frequencies of interest (0.01-1 Hz) couples through the same gain and phase as a static offset. If the RAM-induced voltage contains components that are not equivalent to a true PDH error-signal offset, Eq. (1) over- or under-reports the frequency instability. The technical achievements (295 us storage time, 10^-7 active RAM suppression, D calibration) are plausible and credit-worthy, but the 10^-19 laser-stabilisation claim is only as strong as this unvalidated conversion. The title is additionally in tension with the stated 2e-17 thermal noise floor, which caps total locked-laser stability regardless of RAM.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports two technical advances for cavity-stabilized lasers: a 68 cm optical reference cavity with a measured ringdown storage time of (295 ± 2) µs (finesse ≈ 410,000) and an active residual amplitude modulation (RAM) cancellation scheme using an annealed-proton-exchanged lithium-niobate waveguide EOM with bias-field feedback. The authors measure the RAM-induced PDH error-signal voltage fluctuations with the cavity in an open (beam-blocked) configuration, normalize to the PDH error-signal peak-to-peak voltage, and convert to fractional laser frequency instability using a directly measured discriminator coefficient D = (4.7 ± 0.5) × 10^11. With active cancellation they report RAM-induced fractional frequency instability of 1.3 × 10^-18 at 1 s and 3 × 10^-19 at 10–100 s, about two orders of magnitude below the cavity thermal noise floor of 2 × 10^-17. The paper concludes that RAM no longer limits the stability of such cavity-stabilized lasers.","tokens_in":9726,"tokens_out":12021,"duration_ms":112034,"significance":"If the conversion from open-cavity voltage noise to locked-laser frequency instability is valid, the result is significant: it demonstrates that the RAM contribution can be pushed far below the thermal noise floor, and it provides a state-of-the-art storage time for an optical reference cavity. The direct measurement of the discriminator coefficient D, the explicit noise-floor characterization (DVM, mixer, PD noise), and the careful ringdown measurement are strengths; the reported numbers are conservative if the open-cavity measurement is indeed an upper bound. The claim of 10^-19 laser frequency stabilisation, however, is not demonstrated in total, since the cavity thermal noise floor of 2 × 10^-17 caps the overall locked-laser instability.","major_comments":[{"comment":"The title states \"for 10^-19 laser frequency stabilisation\" and the abstract says \"unlocking 10^-19 fractional frequency laser stabilisation,\" but the paper only measures the RAM-induced contribution to the frequency instability, not the total stabilised laser instability. The text explicitly states (after Fig. 2(b)) that the cavity has an estimated thermal noise limit of 2 × 10^-17, which caps the total fractional frequency instability of any laser locked to this cavity. The abstract and title should be reworded to refer to \"RAM-induced fractional frequency instability at the 10^-19 level\" or to \"a step towards 10^-19 stabilisation,\" otherwise the central claim as presented is not supported by the measurements.","section":"Title and Abstract"},{"comment":"The conversion of the measured voltage noise into frequency instability uses ΔV_PDH_RAM measured with the beam stopper in the cavity (open test cavity) and V_PDH_PP measured from the cavity resonance sweep. The paper argues that this gives an upper limit because the reflection coefficient of a locked cavity near resonance is smaller than that of a mirror. However, Eq. (1) involves the ratio ΔV_PDH_RAM/V_PDH_PP, and if both quantities scale with the reflected carrier amplitude, the ratio would be approximately invariant rather than an upper bound. The authors should either provide a quantitative analysis of how the cavity reflection coefficient affects both numerator and denominator, or explicitly label the reported 1.3 × 10^-18 and 3 × 10^-19 values as conservative upper bounds. This is load-bearing because the headline numbers are obtained solely through this conversion.","section":"Eq. (1) and the paragraph beginning \"Traditionally, the effects of RAM\""}],"minor_comments":[{"comment":"The caption reports \"τ_cav = (295 ± 2) ms\" but the text gives (295 ± 2) µs; the caption unit should be corrected to µs.","section":"Fig. 2(b) caption"},{"comment":"The inset shows a ringdown in air with τ_cav = 14 µs, but the caption text mentions \"Chamber pressure 6 × 10^-8 mbar\"; clarify that this pressure applies to the vacuum measurement only.","section":"Fig. 2(b) inset"},{"comment":"The RAM-induced frequency instability values (1.3 × 10^-18, 5 × 10^-19, 3 × 10^-19) are reported without uncertainties; given the 10% uncertainty on D, a propagation of this uncertainty would be useful for readers.","section":"Paragraph reporting numerical values"},{"comment":"The power spectral density values are given as \"~10^-30 Hz^-1\" and \"~10^-35 Hz^-1\"; specify the frequency at which these are evaluated (the text says 10 s integration time, which corresponds to 0.1 Hz, but the axis should make this clear).","section":"Fig. 3(c) and text"},{"comment":"The phrase \"the demonstrated simplicity, robustness and effectiveness of this technique has the potential for a wide application\" is vague; consider specifying the intended applications (optical clocks, spectroscopy, gravitational-wave lasers) and any limitations.","section":"Conclusion"}],"recommendation":"major_revision","confidential_remarks":"The technical work is sound and the measurements are careful, but the title and abstract overstate the result by claiming 10^-19 laser frequency stabilisation when only the RAM-induced frequency instability is measured. The conversion via Eq. (1) also deserves a clearer upper-limit justification. Both issues are fixable within the scope of a revision; the paper is not a reject."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the technical core is real: a 68 cm cavity with 295 µs storage time (finesse ~410,000) and active RAM cancellation on an APE waveguide EOM achieving 10^-7 fractional voltage noise are credible, carefully characterized results that look like genuine advances for the ultrastable laser community. Second, the '10^-19 laser frequency stabilisation' claim is not supported by what is measured. They compute RAM-induced fractional frequency deviations by converting open-cavity PDH voltage noise through Eq. (1) using a directly measured discriminator D = (4.7 ± 0.5) × 10^11. That conversion is the load-bearing step, and it is the least validated one.\n\nWhat is good: the ringdown fit is direct and gives τ = (295 ± 2) µs, with a reasonable finesse estimate. The discriminator calibration is a real improvement over relying on inconsistent literature formulas: adding a 10% offset and measuring the frequency shift against a reference laser is a sensible direct method. The RAM cancellation is demonstrated with more than two orders of magnitude suppression, and the phase-sensitivity study (10 mrad degrades cancellation by a factor of two) is the kind of practical detail that makes the work useful. The technical noise breakdown (DVM, mixer, PD) is honest and useful.\n\nWhere it is soft: the open test cavity with a beam stopper is the right way to avoid higher-order mode contamination, and the argument that it gives an upper limit is plausible but not airtight. The bigger issue is that D is calibrated on a locked cavity with a static DC offset, while the RAM noise is measured on an open cavity as a time-varying signal. The paper does not show that the transfer function from RAM-induced photocurrent imbalance to error signal voltage is the same in both paths, nor does it validate the conversion with an independent frequency measurement of a locked laser with RAM present. So the headline numbers (1.3e-18, 5e-19, 3e-19) should be read as derived estimates, not measured laser stability. The title also sits oddly next to the stated 2e-17 thermal noise floor; even if RAM were perfectly suppressed, the total locked-laser instability is capped there unless the cavity is upgraded or cryogenic.\n\nMinor: a figure caption says 'ms' where the text says 'µs' for τ_cav — an obvious typo. Allan deviation error bars are not quantified. Data are 'on request,' which is fine for a Letter but limits independent verification.\n\nBottom line: this is a useful engineering/metrology result worth citing for the storage time and RAM cancellation technique. It deserves peer review, but I would ask the authors to soften the title to 'RAM-induced frequency deviations at the 10^-19 level' and to add either a direct beat measurement with RAM on/off or a validation of the D calibration path at the relevant Fourier frequencies. Serious thinker: yes.","headline":"Record 300 µs cavity storage and 10^-7 RAM cancellation are real, but the 10^-19 frequency-stability headline is an overreach: it is a derived estimate, not a measured lock.","tokens_in":10321,"tokens_out":2808,"would_cite":true,"duration_ms":22806,"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":"A 68 cm cavity with 300 µs storage time plus active RAM cancellation pushes RAM-induced laser frequency noise to $3\\times 10^{-19}$, below the thermal noise floor.","keywords":["residual amplitude modulation","RAM cancellation","Pound-Drever-Hall locking","optical reference cavity","cavity storage time","laser frequency stabilization","electro-optic modulator","fractional frequency instability"],"falsifier":"Inject a known oscillating offset into the PDH error signal at frequencies between 0.1 and 100 Hz, with amplitude set so that the predicted fractional frequency shift via $D$ is around $10^{-18}$, and measure the actual laser frequency shift against an independent reference laser; if the measured response deviates from the DC-calibrated prediction by more than the quoted $3\\times 10^{-19}$ level, the central claim fails.","tokens_in":9218,"feed_emoji":"","tokens_out":11561,"duration_ms":105358,"temperature":0.7,"pith_summary":"The paper claims that residual amplitude modulation (RAM) from the electro-optic modulator — the main technical noise that corrupts Pound-Drever-Hall locking signals — can be pushed so low that it no longer limits the stability of cavity-stabilized lasers. It does this by pairing a 68 cm optical cavity with a storage time of $\\tau_{\\mathrm{cav}}=(295\\pm 2)\\,\\mu$s, the longest reported for a cavity with state-of-the-art thermal noise, with an active RAM-cancellation servo on a waveguide modulator that holds the RAM-induced voltage fluctuation at the $10^{-7}$ level. The measured RAM-induced fractional frequency instability is $1.3\\times 10^{-18}$ at 1 s and $3\\times 10^{-19}$ at 10–100 s, almost two orders of magnitude below the cavity's thermal noise limit of $2\\times 10^{-17}$. If correct, RAM is no longer a barrier to $10^{-19}$ fractional frequency stabilization.","feed_headline":"RAM-induced laser noise falls to 3e-19 with 300 µs cavity","feed_subtitle":"At 10 to 100 seconds, the RAM term is nearly two orders below the thermal noise floor, removing a key limit on ultrastable lasers.","key_machinery":"The load-bearing object is the relation in Eq. (1): $\\Delta f_{\\mathrm{RAM}}/f_L = (\\Delta V_{\\mathrm{PDH\\,RAM}}/V_{\\mathrm{PDH\\,PP}})(1/D)$, with $D\\propto \\tau_{\\mathrm{cav}}$; it converts the RAM-induced voltage noise of the PDH error signal into a laser frequency instability. The second object is the active RAM-cancellation loop, which extracts a RAM error signal from a photodetector before the cavity and feeds a ~1 V bias voltage back to the EOM to cancel the amplitude modulation at its source. The third is the cavity itself: 68 cm length, $\\tau_{\\mathrm{cav}}=(295\\pm 2)\\,\\mu$s ringdown, finesse $\\simeq 410{,}000$, linewidth $\\simeq 540$ Hz, from which $D=(4.7\\pm 0.5)\\times 10^{11}$ was measured directly by adding a 10\\% DC offset to the PDH error signal and observing the resulting frequency shift.","core_discovery":"The central claim is that the RAM-induced fractional frequency shift of a locked laser is governed by Eq. (1), $\\Delta f_{\\mathrm{RAM}}/f_L = (\\Delta V_{\\mathrm{PDH\\,RAM}}/V_{\\mathrm{PDH\\,PP}})(1/D)$, with the normalized discriminator coefficient $D$ proportional to the cavity storage time $\\tau_{\\mathrm{cav}}$. The paper demonstrates this relation in a 68 cm cavity with $\\tau_{\\mathrm{cav}}=(295\\pm 2)\\,\\mu$s and an active bias-field servo on a fibre-coupled annealed-proton-exchanged lithium-niobate waveguide EOM, reducing the normalized RAM voltage fluctuation to $6\\times 10^{-7}$ at 1 s, $2.5\\times 10^{-7}$ at 10 s, and $1.5\\times 10^{-7}$ at 10–100 s. Using the directly measured $D=(4.7\\pm 0.5)\\times 10^{11}$, these correspond to RAM-induced laser fractional frequency instabilities of $1.3\\times 10^{-18}$, $5\\times 10^{-19}$, and $3\\times 10^{-19}$ at the same integration times. The claim is that RAM has therefore ceased to be the limiting noise source for this design.","pith_inferences":["Because $D$ grows with $\\tau_{\\mathrm{cav}}$, pushing storage times beyond 300 µs or increasing finesse would push the RAM-induced instability below $10^{-19}$, provided the thermal noise limit is lowered in parallel; combining this RAM servo with cryogenic crystalline cavities could make RAM irrelevant at the $10^{-18}$ total-instability scale.","The same bias-port cancellation on an annealed-proton-exchanged waveguide EOM should transfer to other modulation-based locking schemes (FM spectroscopy, other wavelengths), wherever RAM sets a noise floor.","The RAM values were measured with an open test cavity and a beam stop, which the paper argues is an upper limit; a locked-cavity measurement would likely give even lower in-loop RAM, a testable consequence.","Fully digitising the error-signal extraction after photodetection could remove the RF pick-up and ground noise that currently limit 10 s to 100 s averaging, leaving the RAM term even more dominant as the residual."],"forward_implications":["At 10–100 s averaging, the RAM-induced instability of $3\\times 10^{-19}$ sits almost two orders below the cavity thermal noise limit of $2\\times 10^{-17}$, so RAM no longer sets the frequency-stability floor for this design.","A 68 cm cavity with 300 µs storage time and a directly measured discriminator coefficient is a practical route to $10^{-19}$ fractional frequency stabilization; the next limits are thermal noise and other technical noises, not RAM.","The active RAM cancellation works with a simple low-voltage bias servo and no temperature stabilisation, so it can be adopted in existing ultrastable laser systems without major redesign.","Maintaining the cancellation requires stable RF phases: a 10 mrad phase drift at 20 MHz degrades the RAM level from $2.5\\times 10^{-7}$ to $5\\times 10^{-7}$ at 10 s, so the RF path lengths must be controlled."],"supporting_citations":[{"why":"Supplies the Pound-Drever-Hall locking method, whose error signal is the measurement channel for RAM-induced noise.","marker":"[8, 9]"},{"why":"Provides the reflection-coefficient behaviour that justifies the open test cavity with a beam stop for a pure RAM measurement.","marker":"[9]"},{"why":"Identifies residual amplitude modulation as the noise source in electro-optic phase modulators and introduces cancellation methods.","marker":"[10, 11]"},{"why":"Demonstrates prior active RAM compensation in a free-space lithium-niobate modulator, the baseline this work improves on.","marker":"[16]"},{"why":"Reports earlier fibre-coupled RAM reduction and the expression D = f_L/Γ used to cross-check the directly measured discriminator coefficient.","marker":"[18]"},{"why":"Documents the annealed-proton-exchanged and x-cut waveguide properties that the EOM design relies on to reduce RAM.","marker":"[23]"}],"fun_headline_variants":["300 µs cavity cuts RAM noise to 3e-19","Active RAM cancel hits 1e-7, laser stable to 1e-19","RAM cancellation at 1e-7 yields 1e-19 laser","300 µs storage time unlocks 1e-19 laser stability","1e-19 laser stability from 300 µs cavity and RAM cancel"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conversion of the measured RAM voltage noise into a frequency instability assumes that the discriminator coefficient $D$, calibrated by adding a 10% DC offset to the PDH error signal, applies unchanged to the oscillating RAM-induced noise; if RAM enters the loop through a different transfer function, the reported $10^{-19}$ values would not hold.","fun_headline_variants_meta":{"raw":{"variants":["300 µs cavity cuts RAM noise to 3e-19","Active RAM cancel hits 1e-7, laser stable to 1e-19","RAM cancellation at 1e-7 yields 1e-19 laser","300 µs storage time unlocks 1e-19 laser stability","1e-19 laser stability from 300 µs cavity and RAM cancel"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000647,"raw_usage":{"total_tokens":2983,"prompt_tokens":967,"completion_tokens":2016,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":583,"completion_tokens_details":{"reasoning_tokens":1916}},"tokens_in":583,"tokens_out":2016,"duration_ms":15689,"temperature":1.0,"reasoning_tokens":1916,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:57:24.986706+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Inject a known oscillating offset into the PDH error signal at frequencies between 0.1 and 100 Hz, with amplitude set so that the predicted fractional frequency shift via $D$ is around $10^{-18}$, and measure the actual laser frequency shift against an independent reference laser; if the measured response deviates from the DC-calibrated prediction by more than the quoted $3\\times 10^{-19}$ level, the central claim fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the reflection-coefficient behaviour that justifies the open test cavity with a beam stop for a pure RAM measurement."},{"cited_title":"Gillot, S","cited_arxiv_id":null,"evidence_quote":"Demonstrates prior active RAM compensation in a free-space lithium-niobate modulator, the baseline this work improves on."},{"cited_title":"Zhang, M","cited_arxiv_id":null,"evidence_quote":"Reports earlier fibre-coupled RAM reduction and the expression D = f_L/Γ used to cross-check the directly measured discriminator coefficient."},{"cited_title":"Wooten, K","cited_arxiv_id":null,"evidence_quote":"Documents the annealed-proton-exchanged and x-cut waveguide properties that the EOM design relies on to reduce RAM."}],"review_version":1}