{"id":"74639b33-dc38-402c-8245-c80afeeac72f","arxiv_id":"2608.03547","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"An AOM-based control system combining two feedback loops and one feedforward branch achieves -155 dB/Hz relative intensity noise up to the MHz range.","lead":"This paper shows a laser power stabilizer using two acousto-optic modulators that reduces intensity noise to -155 dB/Hz up to hundreds of kilohertz. It combines feedback and feedforward control to reach bandwidths normally requiring electro-optic modulators, important for ultracold atom experiments.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Headline RIN may be detector-limited: PD-OUT floor is estimated at -158 dB/Hz, only 3 dB below the claimed -155 dB/Hz, with no independent out-of-loop verification.","rationale":"The reader's weakest assumption correctly identifies the out-of-loop PD-OUT measurement as the load-bearing element. My reading of Methods 8.3 and Fig. 3 shows that the claimed -155 dB/Hz sits only 3 dB above the authors' own estimate of the PD-OUT measurement floor. In a RIN measurement, a 3 dB margin is not sufficient to distinguish a real optical signal from a slightly underestimated detector floor, especially when the authors themselves invoke uncharacterized electronic noise from the summing amplifier to explain the discrepancy. This is not an accusation of fraud; it is a request for a standard experimental check. A cross-spectral measurement with a second independent photodiode, or a calibrated modulation injection, would settle whether PD-OUT is truly resolving optical intensity noise at -155 dB/Hz. The reader's conditional verdict is appropriate: the paper is promising and internally consistent, but the headline quantitative claim needs this additional verification before it can be accepted as a demonstrated optical RIN. I therefore keep the verdict unchanged at CONDITIONAL, with the condition being the independent out-of-loop validation described above.","tokens_in":12255,"tokens_out":9638,"duration_ms":116575,"concrete_test":"Add a second, independently powered and amplified photodiode sampling the same output beam (e.g., via a 50/50 beam sampler after the final PBS, or in the unused port of the final HWP-PBS). Record both PD-OUT and the new photodiode simultaneously with the loop engaged, and compute the cross-spectral density between the two channels. True optical intensity noise is common to both detectors and appears in the correlated part of the spectrum, while independent electronic noise floors do not. If the correlated RIN between 10 and 200 kHz is at or below -158 dB/Hz, then the reported -155 dB/Hz spectrum is an upper bound set by PD-OUT electronics and the headline optical RIN is unproven. If the correlated RIN is -155 dB/Hz, the claim is validated. A complementary check is to inject a calibrated sinusoidal intensity modulation of about -150 dB/Hz at 100 kHz and verify that PD-OUT reproduces it wi","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the output RIN reaches -155 dB/Hz rests entirely on the single out-of-loop photodiode PD-OUT (Sec. 2, Methods 8.3). The paper's own estimate places the PD-OUT measurement floor at S_min = -158 dB/Hz: shot noise is -162 dB/Hz, electronic NSD is 35 nV/Hz (about -167 dB/Hz at 8 V), and these are combined and doubled to -158 dB/Hz. The stabilized spectrum in Fig. 3 flattens at -155 dB/Hz from 10 to 200 kHz, only 3 dB above that floor. The authors attribute the 3 dB discrepancy to 'electronic noise induced by the summing amplifier or other electronic components involved.' That is the same mechanism that would raise the true detector floor to -155 dB/Hz. If the actual PD-OUT floor is -155 dB/Hz rather than -158 dB/Hz—from slightly underestimated electronic noise, ground-loop pickup, power-supply interference, or ADC limitations—the reported spectrum is simply the measurement noise floor, and the optical RIN could be much lower or higher. No second detector, calibrated intensity-modulation injection, or cross-spectral measurement is shown to demonstrate that PD-OUT is limited by optical intensity noise at -155 dB/Hz. Additionally, PD-OUT sits in one output port of the final HWP-PBS while PD-FB2 is in the other; polarization or pointing-induced differential noise would make the out-of-loop port unrepresentative of the stabilized output. Either failure would invalidate the headline number.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an AOM-based laser intensity stabilization system that combines a slow DC feedback branch with a fast AC branch implementing both feedback and feedforward control on a second AOM. The authors report a relative intensity noise (RIN) suppression down to -155 dB/Hz from 10 kHz to 200 kHz, with the spectrum remaining below -145 dB/Hz up to 1.2 MHz and a claimed closed-loop control bandwidth of approximately 1.5 MHz. They also characterize a gain-scheduling linearization of the DC branch and demonstrate dynamic intensity ramps with 10-50 ms time scales. The measurement relies on an out-of-loop photodiode (PD-OUT) whose estimated noise floor is -158 dB/Hz, only 3 dB below the claimed RIN level.","tokens_in":12678,"tokens_out":3884,"duration_ms":44515,"significance":"If the headline RIN value is robust, this is a significant advance for AOM-based intensity control, extending the usable bandwidth beyond the typical few hundred kHz and combining two feedback loops with feedforward on a single actuator. The paper is careful in several respects: it provides a noise-floor model for the photodiodes, measures step responses of the AOM and summing amplifier, releases data on Zenodo, and compares spectra with and without feedforward. However, the central quantitative claim of -155 dB/Hz rests on a single out-of-loop photodiode whose estimated floor is only 3 dB below the reported value, and no independent verification of the detector floor is presented. This gap must be addressed before the absolute RIN claim can be accepted.","major_comments":[{"comment":"The claimed RIN floor of -155 dB/Hz is only 3 dB above the estimated PD-OUT floor of -158 dB/Hz (S_min = 2(S_el + S_sn), Methods 8.3). The paper itself attributes the 3 dB discrepancy to electronic noise from the summing amplifier or other components, which is exactly the mechanism that would raise the actual detector floor to -155 dB/Hz. No independent out-of-loop verification is shown: there is no calibrated intensity-modulation injection, no second photodiode in the same beam, and no cross-spectral measurement. As the headline result is the absolute RIN level, the measurement must distinguish optical RIN from detector and electronics noise. I recommend adding an in-situ calibration of the measurement floor, e.g., by injecting a known intensity modulation and measuring the photodiode response, or by performing a two-detector cross-spectral measurement, and reporting the resulting confi","section":"Section 4 / Methods 8.3"},{"comment":"PD-OUT and PD-FB2 are placed in the two output ports of the final HWP-PBS (Figure 1a). If the control loop reduces common-mode intensity noise but differential polarization or beam-pointing fluctuations affect the transmitted and reflected ports unequally, the out-of-loop port may not represent the actual stabilized output. This is a load-bearing assumption for the absolute RIN measurement. The manuscript does not quantify the polarization or pointing stability of the output beam or compare the two ports simultaneously. A simple test would be to swap the roles of the two photodiodes or to record the RIN from both ports concurrently and show that they agree within the claimed accuracy.","section":"Section 2"},{"comment":"Figure 3 shows RIN spectra without error bars or repeated measurements. Given that the spectrum flattens at -155 dB/Hz, only 3 dB above the estimated detector floor, the lack of statistical uncertainty is particularly limiting. The flattening could be a detector-floor artifact rather than an optical noise floor. Please provide multiple traces or a statistical confidence interval, and preferably an in-situ floor measurement as described above, so that the reader can assess whether the -155 dB/Hz level is real optical RIN.","section":"Section 4 / Figure 3"}],"minor_comments":[{"comment":"The statement 'the closed-loop control bandwidth is approximately 1.5 MHz, with the servo bump located at around 4.8 MHz' needs a clear definition. Is this the unity-gain frequency, the -3 dB suppression point, or inferred from the step response? The step response in Methods 8.1 measures the AOM-plus-photodiode response, not the closed-loop transfer function. Please clarify how the bandwidth is determined.","section":"Section 4"},{"comment":"Adding a legend to the figure would improve readability. The colors are described in the caption, but a direct label on the figure would help.","section":"Figure 3"},{"comment":"Equation (2) defines S_sn(f) = 2e/I_p, which is white noise and does not depend on frequency. The notation S_sn(f) with a frequency argument is slightly misleading; consider writing S_sn = 2e/I_p and noting that it is frequency-independent.","section":"Methods 8.3"},{"comment":"There are minor typographical issues such as '0 th' instead of '0th' and 'P opt' inconsistently formatted. Also, 'F ALC' appears without a clear abbreviation expansion; consider defining it at first use.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The engineering is solid and the data release is commendable, but the central quantitative claim is not yet verified against the detector floor. The 3 dB margin is too thin to accept without an independent calibration or cross-check. I believe this is fixable within the scope of the manuscript, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Thanks for the report. My take: this is a genuinely useful experimental paper on AOM-based intensity stabilization, and the bandwidth extension into the MHz range is probably real. The main thing to know is that the headline -155 dB/Hz RIN may be detector-limited: their estimated PD-OUT floor is -158 dB/Hz, only 3 dB below the measured plateau. The authors' explanation of the 3 dB gap is plausible, but it means the same electronics could have pushed the true floor up to -155, in which case the spectrum is just the detector's own noise. I'd want a second, independent out-of-loop detector or an injected intensity-modulation calibration before trusting the absolute number.\n\nWhat's new: combining two feedback loops plus one feedforward branch on a single AOM, plus optimizing the beam waist to shorten the acoustic transit time. That combination looks novel relative to the cited AOM work, and the measured step response (65 ns dead time, 35 ns rise) supports the claimed ~1.5 MHz bandwidth. The paper is also unusually careful with methods: out-of-loop photodiode, separate summing-amplifier noise characterization, photodiode NSD measurement, and a Zenodo data link. The gain-scheduling linearization is a nice practical touch for AOM nonlinearity.\n\nSoft spots beyond the detector floor: they state the system is suitable for 'up to several watts,' but all data are taken at ~170 mW input and 13.5 mW on the photodiodes. That's an extrapolation, not a demonstration. The RIN spectra also lack error bars, and reference [32] is cited in two contradictory ways—as an EOM example and as evidence for AOM bandwidth limits. Those are minor, but the citation needs cleaning.\n\nWho it's for: AMO experimentalists building optical lattices or atom traps at wavelengths where commercial low-noise 1064 nm lasers aren't an option. For that audience, the engineering recipe is directly useful. The theory side is simple; no load-bearing math to check.\n\nMy recommendation: send to peer review. The bandwidth result and the control scheme are worth refereeing, and the detector-calibration issue is a legitimate but fixable measurement gap. Ask the referee to focus on whether the absolute RIN floor is actually verified, and ask the authors to either add a cross-check or state clearly that -155 dB/Hz is an upper bound.","headline":"Credible AOM-based intensity control pushes bandwidth into the MHz range, but the headline RIN sits only 3 dB above the detector floor—so the absolute number is an upper bound until cross-checked.","tokens_in":13128,"tokens_out":3797,"would_cite":true,"duration_ms":41470,"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":"By pairing two feedback loops with a feedforward branch on acousto-optic modulators, this paper achieves -155 dB/Hz relative intensity noise at hundreds of kHz and about 1.5 MHz control bandwidth.","keywords":["relative intensity noise","acousto-optic modulator","feedforward control","feedback control","laser intensity stabilization","optical lattice","MHz bandwidth","RIN suppression"],"falsifier":"Measure the controlled output with a second, independent low-noise detector chain (different photodiode and amplifier) while simultaneously recording beam pointing after the AOM; if the measured floor shifts by more than a few dB or tracks the substitute detector's own noise floor, the -155 dB/Hz level was not the true output RIN.","tokens_in":12206,"feed_emoji":"🎛️","tokens_out":8436,"duration_ms":89187,"temperature":0.7,"pith_summary":"The paper shows that an acousto-optic modulator (AOM), usually limited to a few hundred kHz of control bandwidth, can stabilize laser intensity up to the MHz range when two standard feedback loops are joined by a feedforward branch and the beam path inside the AOM crystal is shortened and focused. On a titanium-sapphire laser at 841 nm, the combined controller reduces relative intensity noise to about -155 dB/Hz between 10 kHz and 200 kHz, keeps noise below -145 dB/Hz up to 1.2 MHz, and reaches a closed-loop bandwidth of roughly 1.5 MHz. This matters because optical-lattice and quantum-gas experiments need low intensity noise at high Fourier frequencies, and AOMs are attractive because they handle high optical powers and simple drive electronics. The same system also ramps the output from zero to maximum in tens of milliseconds, so it can serve as both stabilizer and power modulator. The authors attribute the result to their particular combination of actuators and control topology, with feedforward supplying the largest share of noise suppression at high frequencies.","feed_headline":"AOM-based control hits -155 dB/Hz noise floor into MHz range","feed_subtitle":"Feedback plus feedforward on one AOM gives optical-lattice-grade intensity noise past 1 MHz.","key_machinery":"The core object is the two-branch control topology built around two AOMs: a slow DC branch with a digital PI controller and gain-scheduling linearization on a double-pass AOM handles average power and low-frequency drift; a fast AC branch combines an analog PID feedback signal with a feedforward signal on a second AOM, whose zeroth-order transmitted beam is used as the output. The feedforward branch measures noise before a fiber delay line and applies an inverted correction to the same actuator as the feedback, while the beam is focused to a ~50 µm waist near the acoustic transducer to cut the AOM response to 65 ns. The load-bearing relations are the AOM power-transfer response P_out/P_in ∝","core_discovery":"The paper's central claim is that an acousto-optic modulator, long considered too slow for high-frequency intensity stabilization, can be made to suppress laser relative intensity noise up to the MHz range. The authors build a two-stage controller: a slow digital feedback loop on a double-pass AOM sets and holds the average power, while a fast analog loop and a feedforward branch act together on a second AOM. The feedforward path picks off the noise before a 27 m fiber delay and applies an inverted correction, giving it a head start; the feedback path cleans what remains. Optimizing the beam through the second AOM—a ~50 µm waist close to the acoustic transducer—cuts the actuator response to","pith_inferences":["Because the architecture uses only free-space AOMs and a fiber delay, the same controller should transfer directly to other wavelengths and multi-watt powers, limited mainly by AOM damage thresholds and fiber choice; the paper demonstrates 170 mW at 841 nm but states the scaling rationale.","The feedforward branch's 23 dB suppression at 700 kHz suggests that a digitally adjustable delay line, instead of a fixed SMA cable, could tune the feedforward timing in situ and push the usable bandwidth even closer to the AOM's 65 ns response limit.","If the -155 dB/Hz floor is confirmed by an independent detector, the practical noise bottleneck moves to the photodiode; using lower-gain, higher-power detectors could approach the -158 dB/Hz estimate or lower.","For quantum-gas experiments, this single platform could replace the common pair of a slow AOM power servo and a separate fast EOM noise eater, simplifying the optical path and increasing transmission, at the cost of the added complexity of the combined controller."],"forward_implications":["AOM-based intensity control can reach about 1.5 MHz closed-loop bandwidth, with total noise reduction at or above 29 dB up to 700 kHz.","The feedforward branch delivers 23 dB of suppression at 700 kHz, so feedforward, not feedback alone, is the dominant high-frequency noise killer.","The RIN spectrum stays below -145 dB/Hz up to 1.2 MHz and at -155 dB/Hz in the 10-200 kHz band, meeting the demands of optical lattice experiments at wavelengths where 1064 nm commercial lasers are not usable.","The system can ramp the average output from zero to maximum in tens of milliseconds, and the shown instability on fast falling edges is a fixable control-signal saturation issue.","Photodiodes with lower electronic noise and shot noise could lower the achievable RIN below -155 dB/Hz."],"supporting_citations":[{"why":"Provides the AOM bandwidth limit and the argument that response time scales with sound-wave propagation across the beam, motivating the fast beam-path optimization.","marker":"[34]"},{"why":"Supplies the AOM-based power stabilization baseline and the photodiode noise model used to estimate the -158 dB/Hz detection floor.","marker":"[29]"},{"why":"Sets the application benchmark of low-noise optical lattices and contributes the high-power practice of removing photodiode glass covers.","marker":"[19]"},{"why":"Baseline for reducing laser intensity noise over 1 MHz with a different actuator approach; the system is compared against this target.","marker":"[32]"},{"why":"Provides the double-pass AOM configuration used in the slow DC branch to make output pointing independent of drive frequency.","marker":"[6]"},{"why":"Supplies the FPGA-based digital control platform on which the DC PI controller and gain-scheduling linearization are implemented.","marker":"[25]"},{"why":"Gives the shot-noise formula S_sn = 2e/I_p used to compute the photodiode-limited RIN floor.","marker":"[16]"},{"why":"Demonstrates feedforward intensity stabilization on an EOM platform, the high-bandwidth precedent the AOM feedforward branch extends.","marker":"[20]"},{"why":"Shows the 10 MHz EOM stabilization bandwidth that motivates the goal of pushing AOM-based control beyond a few hundred kHz.","marker":"[21]"}],"fun_headline_variants":["AOM-based loop hits -155 dB/Hz noise floor at MHz","Feedback plus feedforward: AOM tames MHz laser noise","Laser intensity noise down to -155 dB/Hz up to MHz","AOM control achieves ultra-low noise to 1 MHz and beyond","MHz-range laser noise tamed by AOM with feedforward"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The claimed -155 dB/Hz floor rests on the out-of-loop photodiode PD-OUT faithfully measuring the output intensity noise, with no correlated artifact from beam pointing, polarization drift, amplifier noise, or photodiode nonlinearity hiding or creating that level.","fun_headline_variants_meta":{"raw":{"variants":["AOM-based loop hits -155 dB/Hz noise floor at MHz","Feedback plus feedforward: AOM tames MHz laser noise","Laser intensity noise down to -155 dB/Hz up to MHz","AOM control achieves ultra-low noise to 1 MHz and beyond","MHz-range laser noise tamed by AOM with feedforward"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00058,"raw_usage":{"total_tokens":2554,"prompt_tokens":717,"completion_tokens":1837,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":461,"completion_tokens_details":{"reasoning_tokens":1746}},"tokens_in":461,"tokens_out":1837,"duration_ms":13561,"temperature":1.0,"reasoning_tokens":1746,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T16:57:34.787044+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the controlled output with a second, independent low-noise detector chain (different photodiode and amplifier) while simultaneously recording beam pointing after the AOM; if the measured floor shifts by more than a few dB or tracks the substitute detector's own noise floor, the -155 dB/Hz level was not the true output RIN.","supporting_citations":[{"cited_title":"Doctoral dissertation, Harvard University Graduate School of Arts and Sciences (2024)","cited_arxiv_id":null,"evidence_quote":"Provides the AOM bandwidth limit and the argument that response time scales with sound-wave propagation across the beam, motivating the fast beam-path optimization."},{"cited_title":"Review of Scientific Instruments89(11), 113112 (2018) https://doi.org/10.1063/1.5046852","cited_arxiv_id":null,"evidence_quote":"Supplies the AOM-based power stabilization baseline and the photodiode noise model used to estimate the -158 dB/Hz detection floor."},{"cited_title":"Review of Scientific Instruments90(3), 033101 (2019) https:// doi.org/10.1063/1.5066623","cited_arxiv_id":null,"evidence_quote":"Sets the application benchmark of low-noise optical lattices and contributes the high-power practice of removing photodiode glass covers."},{"cited_title":"Optics Express 28(21), 31209 (2020) https://doi.org/10.1364/OE.405002 16","cited_arxiv_id":null,"evidence_quote":"Baseline for reducing laser intensity noise over 1 MHz with a different actuator approach; the system is compared against this target."},{"cited_title":"Review of Scientific Instruments91(8), 083001 (2020) https://doi.org/10.1063/5.0009524","cited_arxiv_id":null,"evidence_quote":"Supplies the FPGA-based digital control platform on which the DC PI controller and gain-scheduling linearization are implemented."},{"cited_title":"Applied Physics B102(3), 515–522 (2011) https://doi.org/10.1007/ s00340-011-4399-1","cited_arxiv_id":null,"evidence_quote":"Gives the shot-noise formula S_sn = 2e/I_p used to compute the photodiode-limited RIN floor."},{"cited_title":"Optics Letters40(7), 1334 (2015) https://doi.org/10.1364/OL.40.001334","cited_arxiv_id":null,"evidence_quote":"Demonstrates feedforward intensity stabilization on an EOM platform, the high-bandwidth precedent the AOM feedforward branch extends."},{"cited_title":"IEEE Photonics Technology Letters20(18), 1542–1544 (2008) https://doi.org/10.1109/LPT.2008.928838 15","cited_arxiv_id":null,"evidence_quote":"Shows the 10 MHz EOM stabilization bandwidth that motivates the goal of pushing AOM-based control beyond a few hundred kHz."}],"review_version":1}