{"id":"fff12775-cafe-4cf0-9594-780c3990bb14","arxiv_id":"2509.10445","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A CMOS-compatible silicon nitride platform generates two-mode squeezed light with 2.9-3.1 dB across eight dies on a 4-inch wafer, with less than 0.2 dB variation.","lead":"This paper reports wafer-scale fabrication of silicon nitride photonic chips that generate squeezed light, a quantum resource, with uniform 2.9-3.1 dB squeezing across eight dies on a 4-inch wafer. If reproduced, this is a step toward manufacturing scalable continuous-variable quantum processors and sensors.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The <0.2 dB uniformity claim rests on per-die best-of-four maxima, not on independent device statistics; the 2.9-3.1 dB range is a range of order statistics and could mask large per-circuit variation.","rationale":"The paper reports a plausible advance and includes independent classical characterization, such as Q-factor and escape-efficiency statistics, that is consistent with the reported squeezing level. The central weakness is not the homodyne calibration per se, but the inferential step from eight per-die maxima to 'excellent wafer-scale uniformity.' The text explicitly says each die contains four circuits with different filter parameters and that Fig. 2a labels the maximum per die. Since the abstract's flagship number is the tight 0.2 dB range across the wafer, using maxima rather than all measured circuits makes the uniformity claim overreach. This is a conditional concern: it can be settled by releasing the 32 per-die/per-circuit values. The reader's weakest-assumption list already names the best-of-four selection, so I agree with the reader; the verdict remains conditional.","tokens_in":12502,"tokens_out":5071,"duration_ms":47049,"concrete_test":"Request the raw squeezing values for all four integrated circuits on each of the eight dies (32 measurements), with the corresponding filter design labeled. Compute the mean, standard deviation, and range of these 32 values (or of the subset sharing one design). If all values fall within 2.9-3.1 dB, the uniformity claim is confirmed; if not, the abstract's '<0.2 dB variation' should be restated as a property of per-die maxima only, changing the strength of the wafer-scale claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.1 states that each die integrates four squeezed-light circuits implementing different pump-rejection filter parameters, and Fig. 2a labels the directly measured maximum squeezing level of each die. The central 2.9-3.1 dB / <0.2 dB claim is therefore computed from eight per-die maxima. Taking the maximum over four variants compresses the reported range relative to the underlying distribution of circuits: even if individual circuits varied by, say, 1 dB, the per-die maxima would show a much smaller spread. As written, the paper does not establish that all 32 circuits (or all dies for a fixed design) are within 0.2 dB; it establishes that every die has at least one circuit in that window. This is load-bearing because the headline claim is excellent uniformity across the wafer, which is a statement about the distribution of devices, not about the best device on each die. The shot-noise calibration, in contrast, is standard and, if imperfect, would tend to reduce rather than inflate the measured squeezing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports wafer-scale fabrication and measurement of two-mode squeezed vacuum sources on a CMOS-compatible Si3N4 platform. Eight dies on a 4-inch wafer are characterized; each die contains four circuit variants differing in pump-rejection filter parameters, and the directly measured maximum squeezing per die is reported to lie between 2.9 and 3.1 dB, i.e., <0.2 dB variation. The authors also present classical characterization of quality factors and escape efficiencies, a quantum coupled-mode model for the squeezing spectra, and a comparison of measured squeezing versus pump power with the model. The central claims are wafer-scale uniformity, reproducibility, and agreement with a first-principles theory parameterized by independently extracted device parameters.","tokens_in":12666,"tokens_out":2416,"duration_ms":23842,"significance":"If the uniformity and squeezing-level claims hold, this is a substantive advance: it would demonstrate a reproducible, foundry-compatible route to nonclassical light generation across a full wafer, which is a prerequisite for scalable continuous-variable processors and multiplexed entanglement sources. The paper's strengths include direct quadrature-noise traces for all eight dies, a physically grounded coupled-mode model (Eqs. 1-15) with a clear set of independently characterized inputs, and a credible loss budget identifying the dominant limitation (approximately 60% overall collection efficiency). The work also usefully compares with previous chip-level demonstrations and quantifies the improvement from integrated pump rejection.","major_comments":[{"comment":"The headline claim of 'device-to-device variation <0.2 dB' is not supported by the data as presented. Each die integrates four circuits with different pump-rejection filter parameters, and Fig. 2a labels only the directly measured maximum squeezing level of each die. The reported 2.9-3.1 dB range is therefore a range of per-die maxima, i.e., an order statistic, not a characterization of the distribution of all 32 circuits. Even if individual circuits varied by, say, 1 dB, the per-die maxima could still fall in a 0.2 dB window. Please report the squeezing level for every circuit on every die (or a statistical summary over all circuits), or alternatively restrict the claim to 'best circuit per die' and change the wording accordingly. This is load-bearing because the paper's main claim is wafer-scale device-to-device uniformity.","section":"Section 3.1, Fig. 2a"},{"comment":"The agreement between the measured squeezing/anti-squeezing data and the theoretical model is stated only qualitatively ('agree well'), and the data points in Fig. 4b have no error bars or repeated-measurement uncertainties. Since the manuscript emphasizes that the model is parameterized solely by independently extracted device parameters and experimental settings, please provide measurement uncertainties (e.g., from repeated phase scans or repeated power settings) and a quantitative agreement metric, such as residuals or a reduced chi-squared value. This would make the model-validation claim more robust and would also clarify how the 2.9-3.1 dB wafer-level values relate to the model's prediction.","section":"Section 3.2, Fig. 4b"}],"minor_comments":[{"comment":"The phrase 'squeezed light is extreme vulnerable' should read 'extremely vulnerable'.","section":"Section 1"},{"comment":"The histograms in Figs. 3(c)-(e) would benefit from stating the number of resonances/dies included and the uncertainties in the Lorentzian fits; currently the text gives only 'most probable values' without a measure of the sample size or the fit error.","section":"Fig. 3, Methods"},{"comment":"In the derivation of the model, the word 'conjuction' appears; it should be 'conjunction'.","section":"Methods"},{"comment":"The abstract says 'device-to-device variation <0.2 dB', while the Introduction states 'device-to-device variation <0.2 dB (2.9-3.1 dB)'. Please use a consistent term (e.g., 'per-die maximum variation') that matches the actual measured quantity.","section":"Abstract / Introduction"}],"recommendation":"major_revision","confidential_remarks":"The manuscript addresses an important scalability question in integrated quantum photonics, and the underlying experimental work appears genuine and technically sound. The main issue is that the uniformity claim is based on per-die maxima over four circuit variants, which does not establish device-to-device uniformity for all circuits. This should be fixable within the scope of the manuscript if the authors have the per-circuit data or are willing to weaken the claim. I saw no indication of methodological circularity or data fabrication; the model uses independently extracted device parameters and the measured squeezing values are direct experimental outputs."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about this one. First, the experimental core is solid: they fabricated eight dies on a 4-inch CMOS-compatible Si3N4 wafer, each with a high-Q microring squeezer plus integrated pump-rejection filters and edge couplers, and directly measured 2.9–3.1 dB of two-mode squeezing on every die. That is a genuine manufacturability step forward—prior chip demos were single devices, not wafer-scale. The paper also does something rare: the squeezing traces are real measured noise variances below shot noise, and the theoretical curves come from a standard coupled-mode model with independently extracted Q factors and a combined detection efficiency, so there is no obvious circular fitting. Credit is earned here.\n\nThe soft spots are real but not fatal. The headline uniformity claim—'<0.2 dB variation across the wafer'—is computed from one number per die: the maximum squeezing among four circuit variants on that die. That does not establish that all 32 circuits or even all copies of a single design fall within 0.2 dB; it only shows every die has at least one circuit in that window. If the other three variants are worse, the wafer-level reproducibility story is weaker. For a claim about manufacturability, the right statistic is per-design yield or at least the full distribution of measured circuits. There are also no error bars on the reported squeezing values, so the 2.9–3.1 range might be within measurement uncertainty anyway. And the abstract's '>3 dB' language is contradicted by their own numbers: 2.9 dB is below 3 dB. Minor wording, but it should be fixed.\n\nThe model agreement is stated qualitatively ('agree well') and Fig. 4b has no error bars, so I can't fully judge how tight the theory-data match is. The shot-noise calibration is standard and, if anything, would tend to understate squeezing, so I am not worried about inflation there.\n\nWho is this for? Researchers working on integrated CV quantum information or foundry-based quantum photonics will get real value: the co-packaged filters and couplers, the wafer-scale Q-factor statistics, and the measured squeezing levels are useful benchmarks. But the uniformity inference needs the per-circuit data before the headline claim can be taken at face value.\n\nRecommendation: send it to peer review. A serious referee should ask for per-design data and error bars, but this is a genuine experimental result with clear value—not a desk reject.","headline":"Wafer-scale squeezed-light demo is real and worth refereeing, but the <0.2 dB uniformity claim rests on one best-of-four number per die and the paper overstates its own >3 dB claim.","tokens_in":13241,"tokens_out":2387,"would_cite":true,"duration_ms":20427,"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":"Across a 4-inch wafer, all eight silicon nitride dies produce 2.9–3.1 dB of squeezed light with less than 0.2 dB variation.","keywords":["squeezed light","two-mode squeezed vacuum","silicon nitride photonics","wafer-scale fabrication","CMOS-compatible photonics","continuous-variable quantum information","microring resonator","quantum photonic integrated circuits"],"falsifier":"Re-measure squeezing on all 32 circuits (four per die across eight dies) using a shot-noise level set by an independently calibrated detector and compute the full distribution; if the wafer-wide spread of raw squeezing values exceeds 0.2 dB or any circuit falls below 2.9 dB, the central uniformity claim fails as stated.","tokens_in":12304,"feed_emoji":"⚛️","tokens_out":7745,"duration_ms":63767,"temperature":0.7,"pith_summary":"This paper reports that a 4-inch silicon nitride wafer can host squeezed-light sources that work uniformly: eight dies each directly measured 2.9–3.1 dB of quadrature squeezing, with less than 0.2 dB spread across the wafer. The result matters because continuous-variable quantum processing and entanglement swapping need reproducible squeezing near or above 3 dB, and until now this level had been shown only on individual chips, not across a full foundry-compatible wafer. The authors attribute the uniformity to a co-integrated design: ultralow-loss, strongly overcoupled microrings that generate two-mode squeezed vacuum, cascaded pump-rejection filters, and low-loss edge couplers. They also show that a first-principles quantum model, using only independently extracted device parameters, reproduces the measured squeezing and anti-squeezing.","feed_headline":"Wafer-scale squeezed light: 2.9–3.1 dB on all 8 dies","feed_subtitle":"CMOS-compatible chips now make uniform nonclassical light, a step toward scalable quantum processors.","key_machinery":"The load-bearing element is the strongly overcoupled silicon nitride microring resonator used as a squeezer below threshold. Its escape efficiency \\(\\eta = 1 - Q_L/Q_i\\), where \\(Q_i\\) is the intrinsic quality factor and \\(Q_L\\) the loaded quality factor, sets the maximum squeezing that can leave the resonator; with \\(Q_i \\approx 10 \\times $10^{6}$\\) and \\(Q_L \\approx 0.83 \\times $10^{6}$\\), the inferred \\(\\eta \\approx 91\\%\\) corresponds to about 10.5 dB of on-chip squeezing. The measured 2.9–3.1 dB is then set by the total detection efficiency after generation, which includes two cascaded add-drop pump-rejection filters (each with 30–40 dB extinction), an inverse-tapered edge coupler (about 75% chip–fiber coupling), free-space propagation (95%), interference visibility (98%), and photodiode quantum efficiency (88%). The quantum model, derived from coupled-mode equations linearized below threshold, uses these independently measured quantities as inputs and correctly predicts the pump-power dependence of squeezing and anti-squeezing.","core_discovery":"The paper's central claim is that wafer-scale, CMOS-compatible silicon nitride photonics can produce uniformly squeezed light. Across a single 4-inch wafer, all eight measured dies yield 2.9–3.1 dB of directly measured quadrature squeezing (average 2.96 dB, spread <0.2 dB), with four dies at or above 3.0 dB. Working below the parametric oscillation threshold, a continuous-wave pump drives non-degenerate four-wave mixing in a strongly overcoupled, high-Q microring with escape efficiency \\(\\eta = 1 - Q_L/Q_i > 91\\%\\), implying up to about 10.5 dB of squeezing available on-chip. The measured 3.0 dB squeezing (with 6.0 dB anti-squeezing at 50 mW pump power) is limited not by the chip but by the overall off-chip collection and detection efficiency of roughly 60%, consisting of edge-coupler loss, free-space propagation, interference visibility, and photodiode efficiency. The observed power dependence matches a quantum coupled-mode model parameterized solely by independently measured device parameters and experimental settings.","pith_inferences":["(Editorial inference) The paper reports the maximum squeezing per die, chosen from four circuit variants per die; a full yield analysis counting all circuits that meet the 2.9 dB bar would give a truer measure of manufacturability than the per-die best case.","(Editorial inference) If the same uniformity holds across multiple wafer runs, the dominant cost of quantum photonic hardware shifts from generating nonclassical light to integration and packaging, which would accelerate the roadmap for continuous-variable quantum computing.","(Editorial inference) A natural extension is to test whether the uniform fabrication translates to broadband squeezing across many mode pairs, which would directly support quantum microcomb sources for frequency-multiplexed cluster states.","(Editorial inference) The authors do not report temperature sensitivity of the squeezing; a systematic measurement of squeezing versus chip temperature would validate the thermo-optic stabilization strategy used for wafer-scale operation."],"forward_implications":["A single 4-inch wafer can now yield a uniform bank of on-chip squeezed-light sources, so arrays of continuous-variable quantum processors or multiplexed entanglement sources could be fabricated in a standard CMOS-compatible foundry flow.","The measured squeezing crosses the commonly cited 3 dB threshold for continuous-variable entanglement swapping and teleportation, moving from marginal noise reduction to operationally useful nonclassical resources.","Because the dominant remaining loss is off-chip (fiber coupling and photodiode efficiency), straightforward packaging and detection improvements should push measured squeezing substantially higher, approaching the roughly 10.5 dB available from the resonator itself.","The same silicon nitride platform can host reconfigurable interferometers, long delay lines, modulators, and detectors, so this wafer-scale source is a step toward single-chip continuous-variable quantum processors.","The agreement of the model with data, using only independent device parameters, makes the model a predictive tool for designing future squeezing circuits without ad hoc fitting."],"supporting_citations":[{"why":"Defines the escape efficiency \\(\\eta = 1 - Q_L/Q_i\\) that quantifies the maximum squeezing extractable from a resonator and underpins the device design.","marker":"[38]"},{"why":"Previous chip-level two-mode-squeezed-vacuum demonstration with off-chip pump filters; the measurement procedure follows it and the more than 2 dB improvement is quantified against it.","marker":"[44]"},{"why":"Establishes the non-degenerate four-wave-mixing mechanism for on-chip squeezing that the authors use to generate two-mode squeezed vacuum states.","marker":"[48]"},{"why":"Provides the canonical quantization of the coupled-mode equations and the below-threshold linearization that yields the theoretical squeezing spectrum.","marker":"[49]"},{"why":"Supplies the silicon hardmask etching process used to fabricate the ultralow-loss, dispersion-engineered silicon nitride wafer.","marker":"[53]"},{"why":"Recent demonstration of more than 3.5 dB squeezing in a single foundry-compatible Kerr microresonator, the per-device benchmark that the wafer-scale result extends.","marker":"[42]"},{"why":"Prior silicon-nitride chip-level squeezing result showing the device class's capability, providing context for the uniformity claim.","marker":"[40]"}],"fun_headline_variants":["Wafer-scale uniform 3 dB squeezing: all 8 dies","CMOS chips now emit uniformly squeezed light across a wafer","8 dies, 3 dB squeezing: wafer-scale nonclassical light","Wafer-scale quantum light: uniform squeezing on all dies"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim that the wafer is uniform at the 2.9–3.1 dB level assumes the shot-noise calibration used in balanced homodyne detection is accurate and that reporting each die's best of four circuit variants fairly represents the wafer; if the calibration is off or the selection hides defective circuits, the <0.2 dB uniformity claim does not hold.","fun_headline_variants_meta":{"raw":{"variants":["Wafer-scale uniform 3 dB squeezing: all 8 dies","CMOS chips now emit uniformly squeezed light across a wafer","8 dies, 3 dB squeezing: wafer-scale nonclassical light","Wafer-scale quantum light: uniform squeezing on all dies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000953,"raw_usage":{"total_tokens":4122,"prompt_tokens":1064,"completion_tokens":3058,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":680,"completion_tokens_details":{"reasoning_tokens":2998}},"tokens_in":680,"tokens_out":3058,"duration_ms":17495,"temperature":1.0,"reasoning_tokens":2998,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:54:10.437402+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-measure squeezing on all 32 circuits (four per die across eight dies) using a shot-noise level set by an independently calibrated detector and compute the full distribution; if the wafer-wide spread of raw squeezing values exceeds 0.2 dB or any circuit falls below 2.9 dB, the central uniformity claim fails as stated.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the escape efficiency \\(\\eta = 1 - Q_L/Q_i\\) that quantifies the maximum squeezing extractable from a resonator and underpins the device design."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Previous chip-level two-mode-squeezed-vacuum demonstration with off-chip pump filters; the measurement procedure follows it and the more than 2 dB improvement is quantified against it."},{"cited_title":"N., Liu, S","cited_arxiv_id":null,"evidence_quote":"Establishes the non-degenerate four-wave-mixing mechanism for on-chip squeezing that the authors use to generate two-mode squeezed vacuum states."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the canonical quantization of the coupled-mode equations and the below-threshold linearization that yields the theoretical squeezing spectrum."},{"cited_title":"& Zhang, Z","cited_arxiv_id":null,"evidence_quote":"Supplies the silicon hardmask etching process used to fabricate the ultralow-loss, dispersion-engineered silicon nitride wafer."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Recent demonstration of more than 3.5 dB squeezing in a single foundry-compatible Kerr microresonator, the per-device benchmark that the wafer-scale result extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Prior silicon-nitride chip-level squeezing result showing the device class's capability, providing context for the uniformity claim."}],"review_version":2}