{"id":"51c654ef-38fe-4c1c-94e1-97ddf2f3d8e7","arxiv_id":"2608.13218","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A single silicon-nitride chip now generates, routes, and detects a 34-mode squeezed quantum microcomb with about 3 dB of raw squeezing.","lead":"A silicon-nitride photonic chip now generates squeezed light, routes it, and detects it with balanced homodyne detection all on the same chip, reporting about 3 dB of squeezing across 34 quantum modes. The work points toward scalable, fully integrated quantum photonic circuits for sensing and continuous-variable quantum computing.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Shot-noise reference at the 500-MHz-detuned LO is the unvalidated linchpin; if the leaked-pump floor drifts or the detuned vacuum differs from on-resonance shot noise, the reported 3 dB could be inflated.","rationale":"The reader's weakest assumption identifies the same load-bearing point: the shot-noise reference is measured at a 500-MHz-detuned LO and relies on subtracting a leaked-pump-plus-dark floor from both traces. I agree that this is the place where the 3 dB claim could be overstated. The paper's own Methods acknowledge the leakage floor depends on pump resonance, and the manuscript does not describe an active lock or repeated floor measurement during the traces, so non-stationarity is a concrete risk rather than a hypothetical. The LO-power scaling in Fig. 2d is real evidence that the detector itself is shot-noise limited, but it does not certify equality between the detuned vacuum reference and the on-resonance reference. The theory comparison in the Results is a genuine strength: it uses independently measured escape efficiency, dispersion, and quantum efficiency rather than fitting the squeezing result, so the claim is not circular. The uncorrelated-pair check and the observation of 17 pairs are also supportive. Other limitations flagged in the text—the unexplained variation of squeezing with mode number ('Further investigation is needed'), the visible phase jitter, and data/code available only on request—reinforce the need for calibration transparency but do not by themselves invalidate the demonstration. Since the reader already assigned CONDITIONAL and the required condition is precisely an independent validation of the shot-noise calibration, my stress-test does not move the verdict. I would keep CONDITIONAL and make the interleaved calibration check an explicit acceptance condition.","tokens_in":8404,"tokens_out":14493,"duration_ms":144448,"concrete_test":"Re-measure qumode pair (-5,5) with an interleaved calibration protocol: for each of at least 10 cycles, record (1) the squeezing trace on resonance, (2) the 500-MHz-detuned LO shot-noise trace, and (3) the dark-plus-leaked-pump floor, all with identical LO powers and no changes to the heater or pump lock. In one cycle, add a fourth reference by heating the filter off-resonance so the BHD sees on-resonance vacuum plus leaked pump instead of squeezed modes. If the calibrated squeezing changes by more than 0.3 dB between the detuned and on-resonance references, or if the cycle-to-cycle spread exceeds 0.3 dB, the reported 3 dB is not robust against the calibration assumption.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—3 dB raw two-mode squeezing with 34 qumodes detected on-chip—rests entirely on a comparison between quadrature noise and a shot-noise reference. The Methods (Shot noise calibration) define that reference by tuning the LO 500 MHz away from the squeezing frequency and subtracting a leaked-pump-plus-dark-noise floor from both the signal and shot-noise traces. This is the least secure step in the argument. First, the subtracted floor is assumed additive and stationary across the two traces, yet the Methods themselves state that the leaked-pump floor depends on whether the pump is on resonance with the squeezer. No active cavity lock or repeated floor calibration during the measurement is described, so thermal drift or heater settling could change the floor between traces and bias the apparent squeezing. Second, the validation in Fig. 2d (noise power linear in LO power) shows the BHD is shot-noise-limited, but it does not show that the 500-MHz-detuned vacuum reference equals the on-resonance shot-noise level, nor that the same background subtraction applies. The only prior evidence cited for this equivalence is reference 22, not a measurement in this paper. The 2.7 MHz versus 3.5 MHz offset inconsistency in the text further prevents independent reproduction. With no error bars or quantitative anti-squeezing report, a 0.3–0.5 dB calibration error would be invisible yet would change the headline from 3 dB to 2.5 dB or less.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a silicon-nitride photonic chip that integrates a Kerr microresonator squeezed-light source, a tunable ring filter for qumode routing, a 50/50 MMI coupler, and a pair of heterogeneously integrated MUTC photodiodes for balanced homodyne detection (BHD). The central claim is that two-mode squeezed vacuum is generated, routed, and detected entirely on chip, with approximately 3 dB of raw squeezing for the (-5,5) mode pair and 17 pairs (34 qumodes) measured across the comb. The paper also reports a shot-noise-limited BHD (noise power linear in LO power), an uncorrelated-pair control, and a theoretical estimate of 3.1 dB squeezing based on independently measured escape efficiency (75%), total on-chip quantum efficiency (72%), and pump power (0.2 dB below threshold). The authors conclude that heterogeneous integration resolves the material conflict between low-loss quantum state generation and efficient photodetection.","tokens_in":8583,"tokens_out":7954,"duration_ms":90409,"significance":"If validated, this is a significant step toward fully integrated continuous-variable quantum photonic systems. The demonstration of on-chip generation, routing, and BHD without the squeezed light leaving the chip, together with the 17-pair measurement, is a natural and important integration milestone. The paper's strengths include that the 3.1 dB theoretical estimate is not a fit to the squeezing data but uses separately characterized component parameters; the shot-noise-limited behavior is checked by a LO-power linearity measurement; and a control measurement of an uncorrelated pair is provided. The main uncertainties are in the calibration of the shot-noise reference and in the internal consistency of reported measurement conditions, which affect the headline squeezing value.","major_comments":[{"comment":"The shot-noise reference is the linchpin of the 3 dB claim, but its validity is not demonstrated in this work. The reference is obtained by detuning the LOs 500 MHz away from the squeezing frequency, and the authors cite their earlier work (ref. 22) for the statement that this yields the shot-noise level. However, the present chip, detector, and pump-leakage conditions differ from ref. 22, and the Methods also state that the leaked-pump noise floor depends on whether the pump is on resonance with the squeezer. No measurement of the stability of this floor over the duration of the squeezing traces, and no repeated floor calibration, is reported. If the detuned vacuum reference is not identical to the on-resonance shot noise, or if the subtracted floor drifts between the signal and reference traces, the reported 3 dB would be systematically biased. Please provide a direct validation on this chip (e.g., a noise plateau versus LO detuning, or a calibration with a known coherent state), report the subtraction procedure explicitly (linear power subtraction), and give an uncertainty budget for the squeezing value.","section":"Methods (Shot noise calibration)"},{"comment":"The frequency at which the squeezing traces were recorded is stated inconsistently: the text in Results says 3.5 MHz offset, while the Fig. 3c caption says 2.7 MHz. This is not a purely cosmetic error, because the noise background, the photodiode response, and the relationship to the 500-MHz-detuned shot-noise reference can all depend on the RF analysis frequency. Please correct the inconsistency and confirm that all traces in Figs. 3 and 4 were acquired under identical analysis conditions, specifying which frequency applies to which trace.","section":"Results and Fig. 3c caption"},{"comment":"The abstract and introduction claim that the post-squeezer circuit has 1.1 dB total optical loss is inconsistent with the measured on-chip total quantum efficiency of 72% used in the squeezing measurement; 72% corresponds to 1.43 dB loss, whereas 1.1 dB corresponds to the 78% efficiency that would be obtained if one photodiode's quantum efficiency were not deliberately reduced for LO balancing. Since the 72% value is the operating point for the reported squeezing and for the 3.1 dB theory estimate, the paper should quote the operating loss as 1.43 dB, or clearly separate passive circuit loss from detection quantum efficiency.","section":"Abstract and Results (component characterization)"},{"comment":"The headline value of 'approximately 3 dB' is presented without error bars, confidence intervals, or a quantitative report of the anti-squeezing level. The visible phase jitter in the traces affects the inferred squeezing depth, and a 0.3-0.5 dB systematic error would change the result from 3 dB to below 2.5 dB. Please provide repeated measurements or a statistical summary, and state the anti-squeezing value so that its consistency with the expected reciprocal relation to the squeezing level can be checked.","section":"Results (Squeezing measurement)"}],"minor_comments":[{"comment":"There is a typo: 'frquency' should be 'frequency'.","section":"Figure 2 caption"},{"comment":"There is a typo: 'quadrture' should be 'quadrature'.","section":"Figure 3c text"},{"comment":"'parallelled' is nonstandard; use 'parallel' or 'parallelized'.","section":"Figure 1c caption"},{"comment":"The derivation of the splitting ratio and quantum-efficiency ratio from four photocurrent measurements is sound, but the assumption that the MMI is unitary should be stated in the main text as well as in the Methods.","section":"Methods (MMI splitting ratio measurement)"},{"comment":"The right-side coupling facet loss is not reported, although the probe laser is coupled from the right; please clarify whether the quoted 1.5 dB facet loss is for the left facet only and how it affects the probe-based alignment.","section":"Results (experimental setup)"}],"recommendation":"major_revision","confidential_remarks":"The integration result is timely and, if the calibration holds, would be an important advance. The main risk is the 500-MHz-detuned shot-noise reference, whose validation is delegated to earlier work and whose stationarity is not checked; I would ask for a direct on-chip calibration and a proper uncertainty budget before acceptance. The 2.7/3.5 MHz inconsistency and the 1.1/1.43 dB loss discrepancy also need to be resolved. The disclosure of related independent work (ref. 44) is appropriate, but priority and novelty should be assessed in light of that work and of ref. 36."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is the first demonstration I know of where squeezed-light generation and balanced homodyne detection share a single chip, and the result—about 3 dB of raw two-mode squeezing across 34 qumodes—is credible. The system integration is the contribution; the individual components (SiN Kerr squeezer, tunable filter, MMI, MUTC photodiodes) have all been shown before, and the paper says so. What is new is putting them together without wrecking the quantum states: the 1.1 dB post-squeezer loss and the improvement from their own off-chip 1.1 dB to 3 dB on-chip tell that story well.\n\nThe theory comparison is honest. They predict 3.1 dB from measured escape efficiency, quantum efficiency, and pump power; they don't fit the squeezing result. That is exactly the right way to do a consistency check.\n\nThe soft spots are mostly about the measurement chain. The shot-noise reference is a 500 MHz-detuned LO with a subtracted leaked-pump/dark-noise floor. The stress-test worry is that the floor could drift or that the detuned vacuum might not equal the on-resonance shot noise. On reading the Methods, the paper already addresses part of this: they explicitly measure the shot noise with the pump on resonance, because the leaked-pump floor depends on that, and they cite prior work verifying the detuned-LO convergence. LO-power linearity is checked. What's missing is any stability check on the subtracted floor during the squeezing trace, and no error bars. So the concern is legitimate but not damning; it's a calibration-detail question rather than a demonstrated flaw.\n\nThe direct inconsistency between the Fig. 3 caption (2.7 MHz) and the text (3.5 MHz) is embarrassing and needs fixing. Data and code only on request is also a bit weak for a quantum photonics paper with a headline claim.\n\nOverall: the central claim holds up to the evidence presented. This deserves serious peer review, and I'd want the referees to push for a cleaner calibration description, uncertainty estimates, and public data. I'd bring it to reading group.","headline":"First heterogeneous integration of squeezed-light generation and on-chip balanced homodyne detection, with a credible 3 dB/34-mode demonstration; the shot-noise calibration is the main thing to probe.","tokens_in":9237,"tokens_out":2283,"would_cite":true,"duration_ms":32060,"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":"This paper demonstrates that squeezed light can be generated, routed, and measured by balanced homodyne detection on a single silicon-nitride chip, with about 3 dB of raw two-mode squeezing across 34 qumodes.","keywords":["squeezed light","quantum microcomb","balanced homodyne detection","heterogeneous integration","silicon nitride photonics","continuous-variable quantum information","photodetectors","quantum photonic integrated circuit"],"falsifier":"Block the pump while keeping the local oscillators at the squeezed-mode frequency and measure the noise variance; if it does not match the 500-MHz-detuned vacuum reference within the experimental uncertainty, the raw squeezing is overestimated. A second check is to vary the detuning (e.g., 300 MHz and 700 MHz) and see whether the subtracted noise floor converges to a stable shot-noise level.","tokens_in":8129,"feed_emoji":"⚛️","tokens_out":6037,"duration_ms":46342,"temperature":0.7,"pith_summary":"This paper reports the first single-chip demonstration in which squeezed light is generated, routed, and measured by balanced homodyne detection without leaving the chip. The authors show that the usual conflict between quantum-state generation and measurement can be circumvented by heterogeneous integration: low-loss silicon-nitride waveguides carry the fragile squeezed light, while photodiodes that absorb light efficiently are bonded onto the same chip. They measure roughly 3 dB of raw two-mode squeezing across 34 qumodes, arranged as 17 pairs, with a total post-squeezer circuit loss of 1.1 dB. The significance is that quantum measurements can now be treated as on-chip active operations rather than external readout, which matters for continuous-variable quantum computing and sensing.","feed_headline":"Squeezed light generated and detected on one chip","feed_subtitle":"A silicon-nitride circuit with integrated photodiodes shows 3 dB squeezing across 34 quantum modes.","key_machinery":"The load-bearing mechanism is heterogeneous wafer bonding of modified uni-traveling-carrier (MUTC) photodiodes—photodetectors with a thick absorber and thin collector that absorb light efficiently—onto a low-loss silicon-nitride waveguide circuit. Around that detector sits a racetrack squeezer cavity with a loaded Q of 0.89 million and 75% escape efficiency, a tunable racetrack filter whose FSR is set to twice the squeezer FSR (within 4 MHz) so it drops odd comb lines while rejecting the pump by about 27 dB, and a multimode-interference coupler that combines the routed squeezed modes with a local oscillator. The photodiodes are connected in an anti-parallel RF configuration for balanced detection, and the p-metal profile is shaped to minimize absorption loss. The shot-noise calibration detunes the local oscillator 500 MHz away from the squeezing frequency and subtracts a leaked-pump plus dark-noise floor; this step carries the quantitative weight of the reported squeezing level.","core_discovery":"The central claim is that low-loss quantum-state generation and high-efficiency photodetection, previously thought to impose incompatible material requirements, can coexist on a single photonic chip through heterogeneous integration. The demonstration combines a high-Q Kerr microresonator that produces a two-mode squeezed quantum microcomb, a tunable ring filter whose free spectral range is twice that of the squeezer so it routes only the odd comb lines, and a balanced homodyne receiver made of a 50/50 multimode-interference coupler and a pair of heterogeneously integrated modified uni-traveling-carrier photodiodes. With the pump at 95% of the OPO threshold, the authors observe about 3 dB of raw squeezing for the (-5,5) mode pair and measure squeezing across 17 pairs spanning 34 qumodes. The measured value is close to the analytical prediction of 3.1 dB when the 72% total quantum efficiency after the squeezer is taken into account, and it improves on the group's earlier 1.1 dB off-chip result despite a lower resonator escape efficiency, because the on-chip circuit eliminates 2.3 dB of waveguide-to-fiber coupling loss and 2 dB of off-chip filter loss.","pith_inferences":["One testable extension is to measure squeezing versus pump power across the threshold; the authors' analytical model makes a specific prediction that would separate escape-efficiency losses from detector-imbalance losses.","Because the MMI imbalance accounts for about 0.8 dB of the 72% efficiency, rebalancing the coupler alone could recover most of the gap to the loss budget without changing the photodiodes.","The independent silicon-photonics result noted in the paper implies that heterogeneous III-V-on-SiN and monolithic silicon approaches will compete on loss, efficiency, and scalability; a direct loss-budget comparison between the two would settle which path reaches higher squeezing first."],"forward_implications":["Squeezed-light sources and their detectors can be unified on one chip, eliminating the waveguide-to-fiber and off-chip-filter losses that previously cut measured squeezing from 3 dB to 1.1 dB.","The same chip can address 34 qumodes as 17 two-mode squeezed pairs, so the architecture already provides a multimode resource for continuous-variable processing.","Near-term component improvements the authors identify, such as raising escape efficiency to 90% and post-squeezer quantum efficiency to 80%, would push expected measured squeezing to about 4.9 dB.","The heterogeneous-integration approach extends naturally to single-photon and photon-number-resolving detectors, which would let non-Gaussian operations happen on the same platform."],"supporting_citations":[{"why":"Supplies the squeezer design and the earlier off-chip-detection baseline of 1.1 dB that this work improves on.","marker":"[23]"},{"why":"Provides the heterogeneous photodiode-on-silicon-nitride integration method used to bring balanced detection onto the chip.","marker":"[37]"},{"why":"Verifies that detuning the local oscillator away from the squeezing frequency yields a shot-noise reference, which the calibration relies on.","marker":"[22]"},{"why":"Motivates the p-metal taper design that keeps photodiode absorption loss low while maintaining high quantum efficiency.","marker":"[39]"},{"why":"Supplies the non-destructive loss characterization used to measure facet coupling loss and on-chip quantum efficiency.","marker":"[40]"},{"why":"Documents the prior separation of squeezed-light generation and detection, the gap this paper closes.","marker":"[36]"}],"fun_headline_variants":["One chip: squeezed light generation and detection","Squeezed light and detection on a single chip","Heterogeneous chip unifies squeezed light and sensing","3 dB squeezing on a chip with integrated detection","Squeezed-light system on one photonic chip"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reported 3 dB of squeezing depends on the assumption that measuring vacuum noise 500 MHz away from the squeezed mode, after subtracting a leaked-pump and dark-noise floor, gives exactly the same shot-noise reference as the shot noise at the squeezed-mode frequency itself.","fun_headline_variants_meta":{"raw":{"variants":["One chip: squeezed light generation and detection","Squeezed light and detection on a single chip","Heterogeneous chip unifies squeezed light and sensing","3 dB squeezing on a chip with integrated detection","Squeezed-light system on one photonic chip"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00015,"raw_usage":{"total_tokens":1200,"prompt_tokens":951,"completion_tokens":249,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":567,"completion_tokens_details":{"reasoning_tokens":176}},"tokens_in":567,"tokens_out":249,"duration_ms":3276,"temperature":1.0,"reasoning_tokens":176,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:12:09.512565+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Block the pump while keeping the local oscillators at the squeezed-mode frequency and measure the noise variance; if it does not match the 500-MHz-detuned vacuum reference within the experimental uncertainty, the raw squeezing is overestimated. A second check is to vary the detuning (e.g., 300 MHz and 700 MHz) and see whether the subtracted noise floor converges to a stable shot-noise level.","supporting_citations":[],"review_version":1}