{"id":"8a1abf77-384b-4883-b3a9-af0cbe1d9172","arxiv_id":"2607.15461","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A monolithic SOI chip generated and detected 0.25(1) dB of squeezed vacuum entirely on-chip at room temperature — the first direct measurement of squeezing from silicon waveguides.","lead":"A single silicon photonic chip created squeezed light and measured it on the same chip, at room temperature, on a commercial foundry platform. The 0.25 dB of squeezing is small, but it is the first direct measurement of quantum light generated and homodyne-detected entirely inside a silicon circuit.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Stated efficiency correction 0.25→0.42 dB is inconsistent with Eq. 2 and η_HD=83%; SI-only calibration cannot be audited.","rationale":"The paper's primary empirical claim is the direct 0.25(1) dB measurement, which is plausible and presented with standard diagnostics (π-periodic variance modulation, power scaling, anti-squeezing optimization). The most load-bearing weakness is not the electronic-noise floor per se — that bias is small if noise is subtracted consistently from both signal and reference — but the stated correction from 0.25 dB to 0.42 dB. A direct calculation from Eq. 2 and the quoted η_HD = 83% gives ~0.30 dB, not 0.42 dB. This is a concrete internal inconsistency in the main text. It could be resolved if SI VII uses additional factors (mode overlap, photodiode QE, or a different η_HD definition), but the main text does not say so, making the quantitative analysis non-auditable. This does not invalidate the raw 0.25 dB observation, but it weakens the derived 0.83 dB generated squeezing and the fitted parameters. The reader's CONDITIONAL verdict is appropriate; I do not think the concern warrants rejection, but it does require the SI to be examined and the correction model to be stated explicitly. Hence the verdict remains unchanged.","tokens_in":14257,"tokens_out":14261,"duration_ms":146703,"concrete_test":"Recompute the inferred squeezing entering the detector from Eq. 2 using V_min = 10^{−0.25/10} = 0.944 and η_HD = 0.83 (and also η_HD = 0.785 for the 6.67 dB operating clearance). If the result is ~0.30–0.32 dB instead of the claimed 0.42(2) dB, require the SI VII text to state the full efficiency product (photodiode quantum efficiency, electronic-noise contribution, and LO–Schmidt-mode overlap) used in that step and re-derive 0.42 dB from the stated inputs. This single recalculation determines whether the paper's inference chain is internally consistent and whether the fitted η_L and γ values need revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim chains 0.25(1) dB detected squeezing to 0.42(2) dB entering the detector and 0.83(3) dB generated. The first step is directly testable from the main text and fails. From Eq. 2, V_min = 1−η+η e^{−g}. If V_min = 10^{−0.25/10} = 0.944 and η_HD = 0.83 (the stated detection efficiency), then the squeezing entering the detector is V_in = (V_min −1+η)/η = 0.9325, i.e. 0.303 dB — not 0.42 dB. Using η_HD = 0.785 (the effective efficiency at the stated 6.67 dB operating clearance) gives V_in ≈ 0.929, i.e. 0.32 dB. Reproducing 0.42 dB requires η ≈ 0.61, an additional unstated loss or mode-overlap factor. The LO–Schmidt-mode overlap of 0.88 has squared value 0.774, and 0.83×0.774 ≈ 0.64, which is closer but still not 0.42 dB; the main text does not state whether this overlap was included in the η_HD correction. This internal inconsistency means the SI VII correction model is not auditable from the provided text, and the derived quantities γ, η_L, and the 0.83 dB generated squeezing inherit this uncertainty. The reader's noise-floor concern is related but numerically weaker: if electronic noise is subtracted consistently from both the squeezed trace and the shot-noise reference, a 10% error in the electronic-noise estimate shifts the normalized variance ratio by only ~0.01 dB at 6.67 dB clearance. The load-bearing problem is the unstated/inconsistent efficiency model, not the noise-floor subtraction per se.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a monolithic silicon-on-insulator photonic integrated circuit that generates squeezed vacuum via spontaneous four-wave mixing in a spiral waveguide and detects it with on-chip balanced germanium photodiodes in a pulsed homodyne configuration. The authors directly measure 0.25(1) dB of squeezing, infer 0.42(2) dB entering the detector by correcting for detector inefficiency, and use a power-dependent fit to Eqs. (2)-(3) to extract the waveguide nonlinearity γ and linear loss η_L, from which they quote a generated squeezing level of 0.83(3) dB. They argue this is the first monolithic integration of a continuous-variable source and homodyne detector on a CMOS-compatible, commercially available platform, and that it provides direct evidence of squeezed-light generation from SOI waveguides. The paper also presents a Schmidt-mode analysis (K = 34.7), an LO mode-overlap optimisation, and a discussion of nonlinear-loss limits including a predicted 0.26 dB ceiling for this device.","tokens_in":14547,"tokens_out":10653,"duration_ms":105202,"significance":"If the calibration chain is sound, this is a significant advance for scalable continuous-variable quantum photonics: it removes off-chip coupling loss for squeezed-state detection and leverages a commercial MPW process, making replication accessible. The direct measurement of sub-shot-noise variance, the power-scaling data, and the honest treatment of nonlinear loss are strengths. However, the quantitative hierarchy 0.25 dB → 0.42 dB → 0.83 dB is not fully auditable from the main text, and the inferred values are model-dependent. The core integration result is plausible, but the calibration and fitting must be clarified before the quantitative claims can be accepted.","major_comments":[{"comment":"The quoted inference of 0.42(2) dB entering the detector is inconsistent with Eq. (2) and the stated η_HD = 0.83. For V_min = 10^{-0.25/10} = 0.944 and V_min = 1 − η + η V_in, using η = 0.83 gives V_in = 0.933, i.e. 0.30 dB. Reproducing 0.42 dB requires η ≈ 0.61. Including the squared LO–Schmidt overlap (0.88² = 0.774) with η_HD = 0.83 gives η ≈ 0.64 and V_in ≈ 0.40 dB, still not 0.42 dB. The main text does not state whether the LO mode overlap or the reduced 6.67 dB-clearance efficiency is included in η_HD. Because the 0.42 dB value anchors the generated-squeezing claim, the efficiency model in SI VII must be presented explicitly in the main text and the arithmetic reconciled.","section":"Squeezing estimation, Eq. (2)"},{"comment":"The values γ = 112.5(56) W⁻¹m⁻¹, η_L = 0.52(6), and the 'total generated squeezing level of 0.83(3) dB' are obtained by fitting the same power-dependent data they describe. They are therefore model-dependent extrapolations, not independent measurements. The model assumes a single-mode squeezed state and a mode-matched LO; with Schmidt number K = 34.7 and an LO overlap of 0.88, omission of the mode-overlap factor can bias the fitted η and g. The paper should label 0.83 dB and the 0.26 dB ceiling as 'model-inferred', report the fitted α_TPA used in Eq. (3), and provide a goodness-of-fit metric or an independent validation. This is load-bearing because the Discussion uses 0.83(3) dB as the headline generated squeezing.","section":"Fig. 3b and Eqs. (2)-(3)"},{"comment":"The main text does not specify the electronic-noise subtraction used to normalise the variances to shot noise. At the operating 6.67 dB clearance the electronic noise is ~21% of the total variance at the shot-noise level, so the 0.25 dB effect is only a 5.6% variance reduction. Although a random error in a consistently applied subtraction shifts the ratio by only ~0.01 dB, an inconsistent treatment between the squeezed trace and the shot-noise reference could produce an error comparable to the effect. Please state the subtraction formula and the uncertainty in the noise-floor estimate in the main text so that the direct 0.25 dB claim is fully auditable.","section":"Methods / SI VI"}],"minor_comments":[{"comment":"The axis label on the variance plot appears corrupted in the preprint rendering ('/uni00000013/...'). Please ensure the axis is correctly labelled, e.g. 'Quadrature variance (dB)'.","section":"Fig. 3a"},{"comment":"The sentence 'maximum shot noise clearance of 8.06 dB ... corresponding to a detection efficiency of 83%' does not explain how the clearance maps to a detection efficiency. A formula or reference should be given.","section":"Results, 'The photonic integrated circuit'"},{"comment":"The experimentally estimated nonlinear-loss coefficient α_TPA = β_TPA/A is said to be used in the fit, but its numerical value is not given in the main text. Please report it for reproducibility.","section":"Eq. (3)"},{"comment":"The uncertainty in the 0.25(1) dB value is not derived in the main text. State how many independent phase bins or runs contribute to the statistical error.","section":"Squeezing estimation"},{"comment":"The quoted LO–Schmidt mode overlap of 0.88 should specify whether it is an amplitude overlap or an intensity overlap, and clarify how it enters the detection-efficiency correction.","section":"Results, 'Generation and control of pump and LO fields'"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is honest about limitations and the direct sub-shot-noise measurement appears plausible. The main issue is the auditable calibration chain: the 0.42 dB and 0.83 dB claims need either correction or explicit framing as model-dependent estimates. A careful re-analysis of the efficiency model, including the mode-overlap factor and the operating shot-noise clearance, should be required before publication. The core integration result is likely to survive even if the inferred quantitative values change."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this is the first monolithic continuous-variable source and detector on an SOI chip, and the direct 0.25 dB squeezing measurement is credible. That headline result holds up. The variance modulation is π-periodic, the anti-squeezing scales with pump power, and the authors are honest about the data being fitted. They also clearly state the 0.26 dB ceiling from nonlinear loss, so they are not overselling the platform's current limit.\n\nThe soft spot is the inference chain. The step from 0.25 dB detected to 0.42 dB entering the detector does not reproduce from Eq. 2 with η_HD = 0.83. Using the numbers in the main text gives about 0.30 dB, not 0.42. To get 0.42 you need η ≈ 0.61, which is close to η_HD times the squared LO–Schmidt overlap (0.83 × 0.88² ≈ 0.64). So the SI likely includes a mode-overlap or additional loss term in the efficiency model, but the main text does not say that. As a result, the 0.42 dB and the derived 0.83 dB generated squeezing are not auditable from the provided text. The electronic-noise issue is real but secondary; at 6.67 dB clearance, even a 10% error in the noise floor shifts the normalized variance by only ~0.01 dB.\n\nSo the direct measurement is the result to trust; the inferred quantities are provisional. This is a platform demonstration with modest squeezing, aimed at the integrated-photonics community. It deserves a serious referee, with the SI carefully checked and the efficiency accounting clarified. I would bring it to reading group and cite it as the first monolithic CV integration, but I would not quote the 0.83 dB generated number without seeing the SI.","headline":"First monolithic CV source-plus-detector on SOI; direct 0.25 dB squeezing measurement is credible, but the inferred efficiency chain doesn't add up.","tokens_in":15234,"tokens_out":5129,"would_cite":true,"duration_ms":50519,"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 single monolithic silicon photonic chip, made on a commercial foundry process, generates and detects squeezed light entirely on-chip, directly measuring 0.25(1) dB of squeezing at room temperature.","keywords":["squeezed light","silicon photonics","homodyne detection","spontaneous four-wave mixing","continuous-variable quantum","monolithic integration","room-temperature quantum photonics","nonlinear loss"],"falsifier":"Measure the homodyne variance with the pump blocked (so no squeezing is generated) at the same local oscillator power and detector settings, sweep the local oscillator phase, and apply the same noise-subtraction and normalization routine; if the subtracted variance trace shows any phase-dependent dip below 1 (shot noise) or does not sit flat at 1 across the full ramp, the 0.25 dB squeezing claim is not supported.","tokens_in":13998,"feed_emoji":"🔬","tokens_out":4181,"duration_ms":42564,"temperature":0.7,"pith_summary":"The paper claims that a monolithic silicon-on-insulator photonic chip can both generate and detect squeezed light entirely on-chip. A spiral silicon waveguide produces squeezed vacuum by spontaneous four-wave mixing; cascaded interferometers strip away the pump, and the squeezed field is measured by an integrated balanced homodyne detector. The authors directly observe 0.25(1) dB of squeezing, infer 0.42(2) dB at the detector, and estimate 0.83(3) dB generated before losses. If correct, this is the first continuous-variable quantum photonic device with source and detector co-integrated on a single chip, pointing toward mass-manufacturable, room-temperature quantum technologies.","feed_headline":"One silicon chip makes and measures 0.25 dB of squeezed light","feed_subtitle":"Source, filters, and homodyne detector share a single 1.44 mm² room-temperature chip—a first for continuous-variable quantum photonics.","key_machinery":"The central object is the monolithic SOI photonic circuit: a spiral waveguide source using spontaneous four-wave mixing, cascaded asymmetric Mach-Zehnder interferometers as pump filters, and an on-chip balanced homodyne detector built from a tunable Mach-Zehnder beam splitter and two waveguide-coupled germanium photodiodes. The dual-pump scheme and spectrally shaped local oscillator select a single Schmidt mode from the highly multimode squeezed field (Schmidt number 34.7). The model ΔX² = 1-η+ηe^{±g} with a power-dependent gain g accounts for two-photon and cross-two-photon absorption, which set the ceiling on achievable squeezing.","core_discovery":"The authors report direct evidence of squeezed light generation from silicon-on-insulator waveguides, achieved by generating and detecting degenerate squeezed vacuum states within one photonic integrated circuit. A bichromatic pulsed pump coupled into a 1.1 cm spiral waveguide drives spontaneous four-wave mixing; two cascaded asymmetric Mach-Zehnder interferometers filter the pump, and the squeezed field is measured by an on-chip balanced homodyne detector with germanium photodiodes. The measured variance dips 0.25(1) dB below shot noise with the expected π-periodic phase dependence, and the squeezing scales with pump power as the model predicts. Removing detector inefficiency yields 0.42(2)","pith_inferences":["If the result is robust, the same integration recipe could be extended to multi-channel devices, where many squeezed sources and detectors are monolithically combined—a configuration that is impractical with fibre-coupled external detection.","The 0.25 dB figure includes all on-chip losses, so the nonclassicality generated at the source is several times larger; future design changes that reduce linear loss could translate directly into higher detected squeezing.","Because a single Schmidt mode is selected by shaping the local oscillator spectrum, the same chip could support mode-multiplexed quantum operations simply by reprogramming the local oscillator profile."],"forward_implications":["Silicon-on-insulator becomes a viable platform for continuous-variable quantum photonics: generation, filtering, and detection can all be done on a single commercial chip.","On-chip detection removes the chip-to-fibre coupling loss that dominates external squeezing measurements, so even modest source efficiency can yield measurable squeezing.","Nonlinear loss mechanisms cap on-chip measured squeezing at about 0.26 dB on this device, so higher squeezing requires longer pump wavelengths or other mitigation.","Room-temperature operation and a 1.44 mm² footprint open the way to portable quantum sensors and scalable arrays of continuous-variable quantum devices.","The demonstrated source-plus-detector integration is a step toward monolithic continuous-variable cluster states and GKP qubits, where many squeezed modes and homodyne measurements must coexist on one chip."],"fun_headline_variants":["Squeezed light born and measured on one silicon chip","On-chip squeezed light: generation and detection in one device","Single chip: squeezed light on silicon, room-temp","0.25 dB squeezed light: on-chip source and detector","Squeezed light made and measured on a single chip"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The result depends on the electronic noise floor being subtracted from the shot-noise and signal variances without bias; at the operating point the detector's shot-noise clearance is only 6.67 dB, so the claimed 0.25 dB squeezing (a ~6% variance reduction) is comparable in size to any small error in that subtraction.","fun_headline_variants_meta":{"raw":{"variants":["Squeezed light born and measured on one silicon chip","On-chip squeezed light: generation and detection in one device","Single chip: squeezed light on silicon, room-temp","0.25 dB squeezed light: on-chip source and detector","Squeezed light made and measured on a single chip"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000666,"raw_usage":{"total_tokens":2848,"prompt_tokens":687,"completion_tokens":2161,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":431,"completion_tokens_details":{"reasoning_tokens":2079}},"tokens_in":431,"tokens_out":2161,"duration_ms":16305,"temperature":1.0,"reasoning_tokens":2079,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T23:18:10.443254+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the homodyne variance with the pump blocked (so no squeezing is generated) at the same local oscillator power and detector settings, sweep the local oscillator phase, and apply the same noise-subtraction and normalization routine; if the subtracted variance trace shows any phase-dependent dip below 1 (shot noise) or does not sit flat at 1 across the full ramp, the 0.25 dB squeezing claim is not supported.","supporting_citations":[],"review_version":1}