{"id":"a932c7ed-1184-475a-84f7-56b02f58a897","arxiv_id":"2505.01013","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A reservoir-engineered cavity scheme with feedforward realizes a quantum speed meter that can beat the standard quantum limit for force sensing at low frequencies.","lead":"This paper proposes a new way to measure the velocity of a free mass without back-action noise, using engineered optical reservoirs instead of special nonreciprocal components. If it works, it offers a simpler route to beating the standard quantum limit in cavity optomechanical force sensors and gravitational-wave detectors.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Wiener filters in Eqs. (A3)-(A5) have 1/omega poles whose impulse responses are non-causal, so the real-time feedforward needed to reach Eq. (19) is not established.","rationale":"The reader identifies the same load-bearing gap: the Wiener filters in Appendix A contain 1/omega terms and their realizability is not established. I agree, and the concern is concrete because the impulse responses of these filters are non-causal: a 1/omega transfer function is the Hilbert transform kernel, whose time-domain response does not vanish for negative times. The paper's language of feedforward and its proposed Michelson implementation imply real-time processing, so the missing causality analysis is central rather than cosmetic. I considered a stronger verdict of REJECT, but the underlying speed-meter principle is standard and an offline post-processing interpretation of the Wiener estimate could still reproduce Eq. (19) as a spectral bound. Therefore the appropriate outcome is to keep the CONDITIONAL verdict and require a demonstration of causal realizability or an explicit statement that the filtering is performed offline over a finite time record. The authors' own limitation section mentions losses and non-Markovian effects but does not address causality or stability of the feedforward filters, which further supports this condition.","tokens_in":10167,"tokens_out":17828,"duration_ms":209074,"concrete_test":"Re-derive the optimal causal Wiener filters for the same MIMO cancellation problem by spectral factorization of the noise matrix built from Eqs. (A1)-(A2), and recompute the low-frequency force spectral density. If the causal filters preserve the limit S_F^SM/S_F^SQL -> 1/(2 gamma), the concern is resolved; if the causal limit diverges or exceeds the value in Eq. (19), the claimed feedforward cancellation cannot be implemented in real time.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central result Eq. (19) is obtained after subtracting g1 a2,out and g2 c1,out from the optimized B-output (Eq. A5). The filters g1 and g2 in Eqs. (A3)-(A4) each contain a pole at omega = 0; at low frequency g1 ~ i gamma/[sqrt(1+gamma^2) omega] and g2 ~ gamma/[sqrt(1+gamma^2) omega]. A transfer function proportional to 1/omega has an impulse response proportional to sgn(t), i.e., it is non-causal and has infinite support on t < 0. It therefore cannot be realized by a real-time feedforward of the homodyne currents, and a finite delay does not cure the problem because the negative-time support extends to minus infinity. The paper does not state that the processing is offline, nor does it provide a causal spectral-factorized Wiener filter or a finite-time approximation with a controlled error. Because the claimed sub-SQL sensitivity relies on exact cancellation of loop noise by feedforward, an unrealizable cancellation leaves the central claim unsupported for real-time operation. This is a realizability gap in the construction, not a disagreement with prior consensus.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a back-action-evading force sensor for a free mass in which an engineered reservoir realizes an effective nonreciprocal coupling between two optical modes A and B, producing a speed-meter-type interaction from reciprocal beam-splitter couplings plus conditional feedforward. Section II recalls the velocity-measurement principle; Section III presents the Langevin equations and the frequency-domain input-output relation; Section IV gives the filtered output (Eq. 18) and the force-noise ratio (Eq. 19), which at low frequency approaches 1/(2γ) and is below the SQL for γ>1/2; Section V sketches a Michelson-type implementation. Appendix A supplies the full multi-input, multi-output relations and the Wiener filters, and Appendix B derives the position-meter comparison.","tokens_in":10355,"tokens_out":20791,"duration_ms":213223,"significance":"If the derivation is correct, the scheme is an appealing alternative to Sagnac-type and polarization-based speed meters because it uses only reciprocal interactions and can be implemented with three cavity modes plus homodyne detection and feedforward. The paper is clearly organized, identifies the loop-noise problem explicitly, and provides a concrete interferometer sketch, all of which are useful. However, the quantitative claims are not yet fully supported: the central equations (18) and (19) disagree under the stated noise normalization, and the feedforward filters in Appendix A are non-causal, so the exact low-frequency sensitivity is not established for a real-time measurement. No code or machine-checked derivation is provided, which places additional weight on the Appendix A algebra.","major_comments":[{"comment":"I cannot reproduce Eq. (19) from Eq. (18) under the spectral-density convention of Eqs. (15)-(16). With vacuum inputs (S_in = 1/2), the four surviving noise coefficients in Eq. (18) give S_F/S_F^SQL = [(1+ω^2)^2 + γ^2 ω^4 (ω^2+2)^2]/(4γ); if one instead takes S_in = 1, the denominator becomes 2γ. Eq. (19) reads (1/2)[(1+ω^2)^2/γ + γ ω^4 (2+ω^2)], which differs from both expressions: the prefactor is off by a factor of two (low-frequency limit 1/(4γ) instead of 1/(2γ) in the S_in = 1/2 case) and the radiation-pressure term has (2+ω^2) instead of (2+ω^2)^2. Since Eq. (19) and Fig. 3 are the quantitative basis for the sub-SQL claim, please correct Eq. (18) or Eq. (19) and state the noise normalization explicitly.","section":"Section IV, Eqs. (18)-(19)"},{"comment":"The Wiener filters g1 and g2 in Eqs. (A3)-(A4) each have a pole at ω=0; at low frequency g1 ~ iγ/[sqrt(1+γ^2) ω] and g2 ~ γ/[sqrt(1+γ^2) ω]. A 1/ω transfer function has an impulse response proportional to sign(t), with infinite support on t<0, so the exact filtering in Eq. (A5) cannot be implemented as real-time feedforward of the homodyne currents a2,out and c1,out. The manuscript does not state that the processing is offline and does not provide a causal spectral-factorized filter or a finite-time approximation with an error bound. Because Eq. (18) relies on exact cancellation of the loop noise, the low-frequency floor of Eq. (19) is not established for a real-time measurement; at minimum the causality assumption must be stated and its effect on the sensitivity quantified.","section":"Appendix A, Eqs. (A3)-(A5)"},{"comment":"The full MIMO input-output relations (A1) and (A2), and the filter choices (A3)-(A4), are asserted without derivation in Appendix A. These equations are the only route from the Langevin equations (4) to the central filtered output (18), so the derivation should be provided (or a detailed supplementary note included), with the cancellation pattern in Eq. (A5) shown explicitly. Without this, the central result cannot be independently verified.","section":"Appendix A, Eqs. (A1)-(A2)"}],"minor_comments":[{"comment":"The sentence following Eq. (20) states that the low-frequency noise scales as 4γ/ω^2, but the displayed Eq. (20) gives 2γ/ω^2 in the same limit; please reconcile this with the noise normalization chosen after correcting the major issue above.","section":"Section IV, Eq. (20)"},{"comment":"The symbol γ is used both as the dimensionless pump parameter in Eq. (11) and as a dimensional amplitude decay rate in Eq. (B2); please use different symbols or state the mapping between the two contexts.","section":"Appendix B, Eq. (B2)"},{"comment":"The optimal homodyne angle φ_opt = tan^{-1} γ is stated without derivation; a short derivation or an explicit reference would help readers verify how the off-diagonal term in D_d is cancelled.","section":"Section IV, Eq. (17)"},{"comment":"Adding the low-frequency asymptotes of the corrected speed-meter and position-meter curves to Fig. 3 would make the comparison much easier to read and would directly display the claimed sub-SQL behavior.","section":"Section V, Fig. 3"}],"recommendation":"major_revision","confidential_remarks":"The qualitative idea is interesting, but the central algebra needs to be checked and presented carefully. I would like to see the derivation of Appendix A and a corrected, internally consistent version of Eqs. (18)-(20) before this paper is accepted; the non-causality issue also needs to be addressed honestly, either by reframing the processing as offline or by providing a causal version with a quantified sensitivity penalty."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear Colleague,\n\nYou should know this paper has a genuinely new idea but a load-bearing gap: the feedforward filters that make Eq. (19) work are non-causal as written, so the sub-SQL sensitivity is not achieved in real time unless the authors provide a causal version or explicitly say the processing is offline.\n\nThe new thing here is the combination of reservoir engineering à la Metelmann-Clerk with a double-pass speed meter and feedforward from two auxiliary outputs. That specific combination is not in the cited literature, and the algebra checks out: Eq. (19) does follow from Eq. (18) once the homodyne projection is included, and the low-frequency limit 1/(2γ) is right. The authors are also honest about the idealizations – they explicitly defer optical losses and non-Markovian effects to future work.\n\nThe soft spot is the realizability of the Wiener filters. g1 and g2 in Appendix A both scale as 1/ω at low frequency, which means their impulse responses have infinite anti-causal support. You cannot implement that in a real-time feedforward loop with a finite delay. The paper doesn't mention offline processing or a causal spectral-factorized approximation with a bounded error. Since the claim of sub-SQL performance rests on exact cancellation of the loop noise, this is a real gap rather than a nitpick. The position-meter comparison in Eq. (20) is fine, and the Michelson implementation sketch is plausible, but the missing causality discussion is the thing a referee should pin down.\n\nAlso minor: the transfer matrices in Appendix A are stated without derivation, and the sign in Eq. (17) for the optimal homodyne angle looks suspicious. These are secondary.\n\nIf the authors can show that the cancellation is achievable in real time with a causal filter, or that offline processing is acceptable for the intended application, this is a solid contribution to the speed-meter literature. As it stands, it's a clever proposal with an incomplete control-theoretic detail.\n\nI'd send it to peer review – the core algebra is worth checking, and the causality issue is exactly what referees are for. If you work on cavity optomechanics or GW detector concepts, you'll want this on your radar.\n\nBest,\n[signature]","headline":"Reservoir-engineering speed meter with feedforward is new and the algebra is consistent, but the non-causal Wiener filters leave the claimed sub-SQL sensitivity unsupported for real-time operation.","tokens_in":10921,"tokens_out":3438,"would_cite":true,"duration_ms":37128,"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":"Reservoir engineering can reproduce a double-pass speed meter using only reciprocal interactions, and the resulting force sensitivity falls below the standard quantum limit.","keywords":["cavity optomechanics","back-action evasion","speed meter","reservoir engineering","standard quantum limit","quantum nondemolition","nonreciprocal interaction","feedforward filtering"],"falsifier":"A direct check would be to compute the poles of the two feedforward filters $g_1$ and $g_2$ from Eqs. (A3)–(A4), or to simulate the full three-cavity system without adiabatic elimination: if any pole lies in the right half-plane, or if the filtered closed loop is unstable, the derived $1/(2\\gamma)$ floor cannot be reached in real time. A tabletop experiment could also measure the low-frequency force noise at the optimal homodyne angle and look for the predicted $1/\\gamma$ dependence.","tokens_in":9932,"feed_emoji":"⚛️","tokens_out":9216,"duration_ms":81309,"temperature":0.7,"pith_summary":"This paper claims that a quantum speed meter—an instrument that measures the velocity of a free mass rather than its position—can be assembled from ordinary reciprocal cavity couplings, without circulators or ring-cavity infrastructure. The mechanism adds a strongly damped auxiliary optical mode that acts as an engineered reservoir; with the correct coupling phase, the reservoir makes light effectively flow one way between two optical modes, reproducing the physics of a double-pass speed meter. A homodyne readout with feedforward from the auxiliary outputs removes the extra loop noise the reciprocal construction generates, leaving a force-noise spectrum that at low frequency is a constant $1/(2\\gamma)$ times the standard quantum limit. If correct, this means increasing the pump power (or injecting squeezed light into one mode) pushes the noise floor below the standard quantum limit without bound, giving a practical route to quantum back-action evasion in cavity optomechanics.","feed_headline":"Engineered reservoirs make a speed meter that beats the quantum limit","feed_subtitle":"A strongly damped auxiliary cavity turns ordinary couplings into a velocity sensor with sub-SQL force noise.","key_machinery":"The load-bearing object is the engineered reservoir: a strongly damped auxiliary cavity mode C that is coupled to both optical modes A and B by beam-splitter interactions. In the Markovian limit where C's damping is the fastest rate, C is adiabatically eliminated and, with the choice $J = i\\Gamma$, it converts the two reciprocal A–B couplings into a single directional flow from A to B. This directional flow is what creates the two opposite-sign interactions with the mechanical mass that define a speed meter. The second piece of machinery is the feedforward readout: homodyning the A and C output fields at fixed angles and subtracting them from the B output using the filters $g_1$ and $g_2$ cancels the loop noise that the reciprocal construction would otherwise leave in the spectrum.","core_discovery":"The central claim is that reservoir engineering can reproduce the double-pass speed meter using only reciprocal interactions. Two optical modes A and B both couple to a mechanical free mass, with opposite radiation-pressure signs, while a strongly damped auxiliary mode C couples to both A and B through beam-splitter Hamiltonians. When the coherent coupling is chosen as $J = i\\Gamma$, adiabatic elimination of C turns the two reciprocal couplings into one directional flow from A to B, satisfying the two requirements of a speed meter: two coherent interactions with the test mass and opposite signs. The authors derive the full input–output relation, then cancel the resulting loop noise by homodyning the A and C outputs at fixed angles and subtracting them from the B output with optimal feedforward filters, yielding the force spectral density $S_F^{\\mathrm{SM}}/S_F^{\\mathrm{SQL}} = \\frac{1}{2}\\left[\\frac{(1+\\omega^2)^2}{\\gamma} + \\gamma \\omega^4(2+\\omega^2)\\right]$. At $\\omega \\to 0$ this ratio is $1/(2\\gamma)$, so for $\\gamma > 1/2$ the speed meter surpasses the standard quantum limit and the noise floor falls as the pump power grows.","pith_inferences":["The paper does not address whether the feedforward filters $g_1$ and $g_2$ are causal and stable; a natural extension is to test how closely finite-order realizable filters can approximate them and what sensitivity penalty that costs.","The same reservoir-engineering trick could be applied to other quantum nondemolition observables or to directional routing between more than two optical modes, since the mechanism only needs a damped intermediary and an interference phase.","Because the advantage grows with the pump parameter $\\gamma$, the practical ceiling will be set by optical loss and by non-Markovian corrections when the reservoir's damping is finite; quantifying that tradeoff is a testable extension the authors list as future work."],"forward_implications":["With $\\gamma > 1/2$, the derived spectrum puts the low-frequency force noise below the standard quantum limit, and the floor scales as $1/\\gamma$, so turning up the pump power (or injecting squeezed light into mode A) improves sensitivity without a SQL bound.","Because the scheme needs only reciprocal interactions, it can be realized in a single optical cavity containing a nonlinear crystal pumped at three frequencies, avoiding the spatial or polarization infrastructure of earlier speed meters.","The speed meter beats a position meter with the same bandwidth and a doubled coupling strength at all frequencies $\\omega < 1$, pointing to a broad low-frequency sensing advantage.","The feedforward stage uses only fixed homodyne angles and the outputs of modes A and C, so no additional filter cavities are required for the noise cancellation."],"supporting_citations":[{"why":"Supplies the reservoir-engineering recipe that converts reciprocal beam-splitter couplings into directional transport.","marker":"[35]"},{"why":"Introduces the double-pass speed meter whose two opposite-sign interactions the scheme reproduces.","marker":"[30]"},{"why":"Gives the modern double-pass speed-meter design and the speed-meter sensitivity benchmark referenced by the authors.","marker":"[31]"},{"why":"Establishes the standard quantum limit that the derived sensitivity is designed to surpass.","marker":"[12]"},{"why":"Provides the speed-meter principle and the input-output formalism used in the derivation.","marker":"[36]"},{"why":"Defines the QND speed-meter concept and the two criteria of opposite-sign coherent interactions.","marker":"[29]"},{"why":"Describes the sloshing-type speed meter, the closest previous reciprocal-interaction implementation.","marker":"[41]"},{"why":"Supplies the optimal-feedforward filtering approach used to cancel the loop noise.","marker":"[43]"}],"fun_headline_variants":["Reservoir engineering yields a speed meter that beats the quantum limit","Speed meter from reciprocal couplings via reservoir engineering","Reservoir engineering beats SQL for force sensing without nonreciprocity","Engineered dissipation turns reciprocal couplings into a speed meter","Beating SQL: reservoir-engineered speed meter"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The predicted low-frequency sensitivity depends on two frequency-dependent feedforward filters that contain $1/\\omega$ terms; the paper does not prove that these filters are causal and stable, so a real-time implementation may not achieve the claimed $1/(2\\gamma)$ noise floor.","fun_headline_variants_meta":{"raw":{"variants":["Reservoir engineering yields a speed meter that beats the quantum limit","Speed meter from reciprocal couplings via reservoir engineering","Reservoir engineering beats SQL for force sensing without nonreciprocity","Engineered dissipation turns reciprocal couplings into a speed meter","Beating SQL: reservoir-engineered speed meter"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000886,"raw_usage":{"total_tokens":3799,"prompt_tokens":892,"completion_tokens":2907,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":508,"completion_tokens_details":{"reasoning_tokens":2827}},"tokens_in":508,"tokens_out":2907,"duration_ms":19929,"temperature":1.0,"reasoning_tokens":2827,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T04:30:25.890482+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct check would be to compute the poles of the two feedforward filters $g_1$ and $g_2$ from Eqs. (A3)–(A4), or to simulate the full three-cavity system without adiabatic elimination: if any pole lies in the right half-plane, or if the filtered closed loop is unstable, the derived $1/(2\\gamma)$ floor cannot be reached in real time. A tabletop experiment could also measure the low-frequency force noise at the optimal homodyne angle and look for the predicted $1/\\gamma$ dependence.","supporting_citations":[{"cited_title":"Braginsky and F","cited_arxiv_id":null,"evidence_quote":"Introduces the double-pass speed meter whose two opposite-sign interactions the scheme reproduces."},{"cited_title":"A new type of quantum speed meter interferometer: measuring speed to search for intermediate mass black holes","cited_arxiv_id":"1702.01029","evidence_quote":"Gives the modern double-pass speed-meter design and the speed-meter sensitivity benchmark referenced by the authors."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the standard quantum limit that the derived sensitivity is designed to surpass."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the QND speed-meter concept and the two criteria of opposite-sign coherent interactions."},{"cited_title":"Purdue and Y","cited_arxiv_id":null,"evidence_quote":"Describes the sloshing-type speed meter, the closest previous reciprocal-interaction implementation."},{"cited_title":"Nishino, S","cited_arxiv_id":null,"evidence_quote":"Supplies the optimal-feedforward filtering approach used to cancel the loop noise."}],"review_version":1}