{"id":"53f2a395-c903-4e81-85b5-8d970847fdc5","arxiv_id":"2509.03482","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Noise in multimode fiber output can be suppressed by more than 10 dB below linear attenuation by optimizing the input wavefront, reaching near-shot-noise levels.","lead":"A team shows that shaping the input beam of a multimode optical fiber with a spatial light modulator can dramatically reduce the intensity noise of the output, achieving levels close to the quantum limit even when the input light is very noisy. This suggests a way to build high-power or broadband laser sources that operate at quantum-limited noise, beyond what simple attenuation can achieve.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Single-frequency PSD optimization may not demonstrate broadband shot-noise-limited output; full noise spectrum needed.","rationale":"The reader's weakest assumption already flags that 'if the 10 MHz PSD does not faithfully represent the quantum noise of the optical field, the central claim weakens.' My analysis sharpens this into the most load-bearing concern: the optimization metric is the 10 MHz PSD, so the system is being trained to minimize exactly that quantity. The claim of near-shot-noise operation is then circular unless it is verified that the noise reduction is not frequency-specific. This is a concrete, testable issue that directly addresses whether the output is genuinely at the quantum limit. The reader's CONDITIONAL verdict is appropriate; the paper should provide broadband noise spectra or a clear justification for why the 10 MHz PSD is representative. My verdict remains UNCHANGED because the reader already identified the core vulnerability and the required condition.","tokens_in":11140,"tokens_out":7675,"duration_ms":85475,"concrete_test":"Measure the intensity-noise PSD of the optimized output state and of the linear-attenuation baseline across a broad frequency range (e.g., 1 MHz to 1 GHz) using the same photodiode and a spectrum analyzer. Compare to the shot-noise level calculated from the DC photocurrent (S_I = 2eI) or independently calibrated with a known shot-noise-limited source. If the optimized PSD is within ~1 dB of the shot-noise level across the entire band, the claim is supported. If it dips to shot noise only near 10 MHz but rises elsewhere, the central claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the output reaches near-shot-noise levels rests on the power spectral density (PSD) of the photocurrent measured at a single frequency, 10 MHz (Figs. 2c, 3b). The theory (Eq. 1) is formulated for the full photon-number variance, which corresponds to the frequency-integrated PSD. The footnote justifies using a single-frequency PSD for shot-noise comparison by assuming white noise, but this is not established for the optimized states. Because the optimization algorithm uses the 10 MHz PSD as its feedback signal, it could converge to states that specifically minimize noise at that frequency while redistributing excess noise to other frequencies, where it remains above the shot-noise level. If so, the output is not truly quantum-noise-limited; the claim would be an artifact of the measurement bandwidth. The absence of broadband noise spectra makes this scenario indistinguishable from genuine shot-noise-limited operation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports experiments and simulations showing that the input wavefront to a nonlinear multimode fiber, together with programmable output spatial filtering, can strongly suppress the intensity noise of a selected output region. The central experimental claim is that the optimized states reduce the beam noise by about 10-12 dB relative to linear attenuation and approach the quantum shot-noise limit, despite a laser source that is ~30 dB above shot noise. A theoretical expression (Eq. 1) decomposes the output photon-number variance into a shot-noise term, a mode-sensitivity term, and an excess-input-noise term, and a simulation framework is used to support the optimization results and to identify the Kerr-induced conversion of input intensity noise into intermodal phase noise as the dominant noise-generation mechanism.","tokens_in":11414,"tokens_out":6735,"duration_ms":74759,"significance":"If the central claim holds, this is a significant advance: it offers a practical route to producing near-shot-noise-limited light from noisy sources, and it introduces programmable wavefront shaping as a tool for controlling quantum noise in complex multimode nonlinear systems. The experimental demonstration that wavefront shaping reduces noise well below the linear-attenuation limit is convincing and the combination of input SLM and output DMD is genuinely novel. The theoretical framework (Eq. 1) is a useful first-order description and, importantly, is not a fit to the optimization result. The paper also makes a clear falsifiable mechanistic prediction, namely that noise patterns are dominated by intermodal phase fluctuations, which is supported by the reduced model in Fig. 5. However, the headline claim of reaching the quantum shot-noise limit rests on a single-frequency (10 MHz) PSD measurement with an absolute calibration deferred to the SI, and this requires additional evidence.","major_comments":[{"comment":"The central claim that optimized output states reach near the shot-noise limit is based on the photocurrent PSD at 10 MHz, while Eq. (1) describes the frequency-integrated photon-number variance. Because the optimizer uses the 10 MHz PSD as its feedback signal, the measured reduction could in principle reflect spectral reshaping of excess noise away from 10 MHz rather than broadband removal. The footnote [39] argues that single-frequency PSD is the standard comparison to shot noise, but this does not rule out the possibility that the 10 MHz component is minimized at the expense of other frequencies. The manuscript should provide broadband noise spectra (or an estimate of the full variance) for the random and optimized states, or a quantitative argument that the Kerr nonlinearity and detection preserve whiteness across the relevant band. This issue is load-bearing for the headline claim o","section":"§2 (Optimization of noise via input shaping), Figs. 2c and 3b, footnote [39]"},{"comment":"The statement that the low-noise states are 'within a few decibels of the shot noise level' depends on a shot-noise calibration that is not described in the main text. The reader cannot assess how the shot-noise level was determined, how the detector noise floor was subtracted, or how the 10 MHz PSD is converted to a photon-number variance. Please include the calibration procedure and its estimated uncertainty in the main text or, at minimum, state explicitly that this is the sole absolute calibration and show its key result. This point, together with the previous comment, determines whether the experiment actually reaches the SQL.","section":"§3 (Fig. 3c and 'fit to the shot-noise level based on experimental data (see SI)')"},{"comment":"The simulations used to support the optimization are continuous-wave, while the experiment uses femtosecond pulses. The paper argues in a parenthetical remark that this does not impact the main conclusions because (1) the lower bound for noise minimization is approximately the shot noise and is the same for CW and pulsed waves, and (2) the experimentally measured spectrum shows 'limited spectral dynamics' (SI). However, the lower-bound argument does not imply that the optimization landscape or the noise-cancellation mechanism is identical for pulsed and CW light, and 'limited spectral dynamics' is not quantified in the main text. The agreement in Fig. 3c is encouraging, but the simulation suite as presented cannot fully rule out pulse-specific noise redistribution. Please either show a representative pulsed simulation or discuss the limitations of the CW approximation more explicitly.","section":"§2 (simulation of continuous-wave propagation, Fig. 2b/d)"}],"minor_comments":[{"comment":"The abstract states '12 dB' of noise reduction beyond linear attenuation, while the main text says 'more than 10 dB' and '20 dB improvement compared to the initial random state.' Please reconcile these numbers and specify the exact conditions (e.g., which experimental run, which transmission).","section":"Abstract and §2"},{"comment":"Please define all symbols in the main text (Φ, δF_in, u_m^(0), U_m) and state the validity conditions of the expansion. Currently these definitions are only implicit or deferred to the SI, which makes the equation difficult to interpret.","section":"Eq. (1)"},{"comment":"The paper lists the free parameters in the simulation (noise floor, excess input noise, DMD loss, power scale) but does not give their values or how they were estimated. A table or paragraph in the SI with these values and uncertainties would aid reproducibility.","section":"§3, simulation free parameters"},{"comment":"No data or code availability statement is included. Given the novelty of the simulation framework, providing access to the simulation code and experimental data would strengthen the paper.","section":"General"},{"comment":"The footnote is unusually long and contains a substantive argument about why a single-frequency PSD is used. Consider moving this argument to the main text or expanding it into a short 'Methods' paragraph, because it is central to the interpretation of the results.","section":"Footnote [39]"}],"recommendation":"major_revision","confidential_remarks":"The paper is likely publishable in a strong journal after the single-frequency/calibration concerns are addressed. The authors should be asked to provide broadband noise spectra for optimized versus random states, a clear description of the shot-noise calibration, and a more explicit discussion of the CW simulation approximation. The fit between the topic and the journal's scope is good, and I see no evidence of circularity in the optimization procedure."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth reading: this is the first convincing experimental evidence I know of that wavefront shaping can control quantum intensity noise in a nonlinear multimode system, not just classical instabilities or SBS. The central comparison—random input wavefronts versus optimized ones—shows a 20 dB drop in measured noise at the target pixel, and the optimized state goes more than 10 dB below the linear-attenuation baseline. That is a direct, feedback-driven demonstration, and the simple mechanism they identify—input intensity noise gets converted to intermodal phase noise via cross-phase modulation, which then creates spatial noise patterns—is plausible and consistent with their power-dependent images.\n\nThe theory in Eq. (1) is a first-order perturbation expression that separates shot noise, sensitivity to initial fields, and excess input noise. It is not fitted to the result; it explains the optimization after the fact. That is honest and useful. The simulation framework also captures the main features, though it uses several fitted nuisance parameters (noise floor, DMD loss, excess noise, power scale). Those are not circular in a damaging way, but they do limit how much quantitative weight the simulation can carry.\n\nThe soft spots are real but not fatal. The headline claim—near shot-noise-limited output—rests on a calibration that is only in the SI, and on a PSD measurement at a single frequency, 10 MHz. The optimization uses that 10 MHz value as feedback, so it could in principle redistribute excess noise to other frequencies where it remains above the shot-noise level. The authors address this in a footnote, arguing the Kerr nonlinearity is quasi-instantaneous and citing limited spectral dynamics in the SI, but the broadband spectrum is not shown in the main text. A referee should ask for that spectrum and the calibration procedure before accepting the quantum-noise-limited phrasing. I would also want error bars or repeat statistics on the optimized noise values; the histograms are suggestive but not quantitative.\n\nThe paper is honest about its own limitations: it notes the simulations are continuous-wave for optimization, that random mode coupling is absent, and that the squeezing prediction is small (0.5 dB) and only in simulation. The citations to prior wavefront-shaping work (refs 13, 50-53) are appropriate and the distinction is clear.\n\nWho gains: experimentalists in multimode fiber, quantum optics, and high-power laser noise control. The paper deserves a serious referee. I would send it to review with a request for the SI calibration details, the broadband noise spectrum, and uncertainty estimates. If those check out, the central claim stands.","headline":"A genuinely new demonstration that wavefront shaping can suppress intensity noise in a multimode fiber to near shot-noise levels, but the shot-noise claim rests on a single-frequency PSD and SI calibration that need to be checked.","tokens_in":11853,"tokens_out":2908,"would_cite":true,"duration_ms":32155,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["42.50.Lc","42.65.-k"],"model":"deepseek-v4-flash","headline":"Active wavefront shaping can push noisy multimode light down to quantum shot-noise levels, even when the input laser carries large excess noise.","keywords":["wavefront shaping","multimode fiber","quantum noise","shot noise","intensity noise","Kerr nonlinearity","spatial filtering","Fano factor"],"falsifier":"Measure the optimized output's noise with an independent shot-noise calibration, e.g., balanced homodyne detection against a local oscillator, over a range of frequencies and powers; if the 10 MHz point or other frequencies lie more than a few dB above the calibrated shot-noise floor, the central claim fails.","tokens_in":11106,"feed_emoji":"💡","tokens_out":6349,"duration_ms":58145,"temperature":0.7,"pith_summary":"Nonlinear multimode systems are usually expected to scramble light and amplify its noise, but this paper argues that those same dynamics can be steered to do the opposite. By actively shaping the wavefront of light entering a multimode fiber, the authors find input states that make a chosen output region almost insensitive to fluctuations in the source. In experiments, this reduces intensity noise by about 12 dB relative to simply attenuating the beam, pushing selected regions near the quantum shot-noise limit even when the input laser carries roughly 30 dB of excess noise. The accompanying variance formula explains the effect: optimization suppresses the term that couples output intensity fluctuations to input excess noise. The result matters because it suggests noisy, high-power amplified sources can be made to behave like quiet, quantum-limited light.","feed_headline":"Wavefront shaping quiets noisy fiber light to quantum limit","feed_subtitle":"Programmed input wavefronts make nonlinear multimode fibers output near-shot-noise light even from noisy lasers.","key_machinery":"The load-bearing object is a variance formula for the photon number in a measured pixel, Eq. (1): (Δn)^2 = n(1-Φ) + Σ_m |∂n/∂u_m^(0)|^2 + δF_in |Σ_m U_m ∂n/∂u_m^(0)|^2. The first term is shot noise, the second is the intrinsic sensitivity of the output to each input modal field, and the third — proportional to the input excess-noise parameter δF_in — is the amplified source noise that dominates for random initial conditions. The optimization works by minimizing this third term, i.e., finding input wavefronts for which the collective sensitivity vector is nearly orthogonal to the normalized input field. The paper also identifies a second mechanism: Kerr-induced coupling of input intensity noi","core_discovery":"The paper's central claim is that controlling the initial wavefront of light entering a nonlinear multimode system can suppress intensity noise in a chosen output quantity all the way down to the quantum shot-noise limit, despite large excess noise in the input. The authors demonstrate this in a multimode fiber using a spatial light modulator to shape the input and a digital micromirror device to select the output region of interest, with a gradient-free optimization on the measured intensity-normalized noise. The optimized states show noise more than 10 dB below the level of linear attenuation, and combined input shaping with programmable output spatial filtering yields regions whose fluctu","pith_inferences":["If the 10 MHz power-spectral-density point is a faithful proxy for the quantum noise of the field, the same shaping strategy could be extended to suppress noise at other frequencies or in other quadratures; the paper measures only one frequency.","Because the mechanism is generic (Kerr-induced phase-noise conversion), the wavefront-shaping protocol should transfer to other nonlinear multimode platforms such as integrated photonic circuits, where the optimization variables are on-chip phase shifters rather than an SLM.","A natural next experiment would start from a source already near the shot-noise limit and check whether optimized states show sub-shot-noise behavior; the paper's outlook suggests squeezing, but its experiments stop at the SQL.","The authors found random mode coupling in the fiber helpful, implying that deliberately engineered mode-coupling statistics could make low-noise states easier to discover — a design principle not tested here."],"forward_implications":["Nonlinear multimode propagation does not have to amplify noise: with programmable input wavefronts, output regions can be made quieter than linear attenuation of the same beam.","Highly amplified, noisy laser sources could be turned into near-shot-noise-limited sources for applications such as interferometry, microscopy, and spectroscopy, without requiring exotic quiet lasers.","Programmable spatial filtering of the output adds a second handle: correlated noise between pixels can be harnessed to reach quantum-level fluctuations at higher transmitted power than single-pixel shaping alone.","The derived variance formula gives a practical optimization target and a fast simulation route for noise control in highly multimode nonlinear systems, where full quantum simulations are intractable.","With lower linear loss, the same control scheme should produce weakly squeezed light below the shot-noise limit, a step toward tailored quantum states from multimode nonlinear devices."],"supporting_citations":[{"why":"Supplies the wavefront-shaping principle used to control the input phase profile.","marker":"[47]"},{"why":"Shows feedback-based wavefront shaping can control mean-field properties of multimode fibers; the optimization protocol extends this to noise.","marker":"[48]"},{"why":"Demonstrates wavefront shaping to suppress nonlinear instabilities in multimode fibers, a related approach the paper contrasts with its noise-targeted method.","marker":"[13]"},{"why":"Defines the standard quantum limit and shot noise that serve as the paper's reference floor.","marker":"[36]"},{"why":"Documents excess noise in amplified laser sources, the input-noise model the paper tries to decouple.","marker":"[37]"},{"why":"Shows how nonlinear dynamics amplify shot noise into large excess noise, motivating the need for noise control.","marker":"[38]"},{"why":"Justifies using the power spectral density at 10 MHz as the noise metric that can be compared directly with shot noise.","marker":"[39]"}],"fun_headline_variants":["Wavefront shaping quiets fiber noise to near shot-noise limit","Preshape input to make multimode output quantum quiet","Programmable wavefronts cut nonlinear fiber noise 12 dB","Shaped light beats linear attenuation to hit quantum noise floor","Wavefront control makes noisy lasers produce quantum-clean light"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"Everything hinges on the calibration that maps the measured 10 MHz photodetector power spectral density to the absolute quantum shot-noise level, together with the model that treats all input excess noise as a single parameter; if that calibration or model is off, the near-shot-noise conclusion would weaken.","fun_headline_variants_meta":{"raw":{"variants":["Wavefront shaping quiets fiber noise to near shot-noise limit","Preshape input to make multimode output quantum quiet","Programmable wavefronts cut nonlinear fiber noise 12 dB","Shaped light beats linear attenuation to hit quantum noise floor","Wavefront control makes noisy lasers produce quantum-clean light"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000641,"raw_usage":{"total_tokens":2770,"prompt_tokens":711,"completion_tokens":2059,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":455,"completion_tokens_details":{"reasoning_tokens":1977}},"tokens_in":455,"tokens_out":2059,"duration_ms":15169,"temperature":1.0,"reasoning_tokens":1977,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T10:51:53.085112+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the optimized output's noise with an independent shot-noise calibration, e.g., balanced homodyne detection against a local oscillator, over a range of frequencies and powers; if the 10 MHz point or other frequencies lie more than a few dB above the calibrated shot-noise floor, the central claim fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the wavefront-shaping principle used to control the input phase profile."},{"cited_title":"Tzang, A","cited_arxiv_id":null,"evidence_quote":"Shows feedback-based wavefront shaping can control mean-field properties of multimode fibers; the optimization protocol extends this to noise."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates wavefront shaping to suppress nonlinear instabilities in multimode fibers, a related approach the paper contrasts with its noise-targeted method."},{"cited_title":"Loudon, The Quantum Theory of Light, third edition ed","cited_arxiv_id":null,"evidence_quote":"Defines the standard quantum limit and shot noise that serve as the paper's reference floor."},{"cited_title":"Bachor and T","cited_arxiv_id":null,"evidence_quote":"Documents excess noise in amplified laser sources, the input-noise model the paper tries to decouple."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows how nonlinear dynamics amplify shot noise into large excess noise, motivating the need for noise control."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Justifies using the power spectral density at 10 MHz as the noise metric that can be compared directly with shot noise."}],"review_version":1}