{"id":"a26339a7-81f5-4014-bb2a-0c3d0577eca9","arxiv_id":"2506.20318","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A superconducting bolometer with homodyne interference performs amplification-free Wigner function tomography of propagating microwave photons, demonstrated on Gaussian states at the single-photon level.","lead":"Researchers built a superconducting bolometer that measures the quantum state of travelling microwave photons without adding amplifier noise. It works like a CT scanner, taking projections of the light at different angles to reconstruct the Wigner function.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Both pillars of the tomography claim are soft: the histograms are assumed Gaussian, so the sparse and N=3 'model-free' reconstructions are Gaussian fits, and the variance calibration is validated only on thermal states, leaving the squeezing claim unanchored.","rationale":"In good faith, the paper does what it says within its stated scope: it builds a Gaussian-state generator, turns an SNS bolometer into a two-moment quadrature detector via two-field interference, and reconstructs Gaussian Wigner functions by Hilbert-transform CT. The internal consistency between the Fig. 3 quadrature extraction and the Fig. 4C reconstruction (n̄_T=0.71 vs 0.73; ζ=0.16 vs 0.19) is genuine evidence that the pipeline works, and the authors deserve credit for stating in Methods that the two-moment method is suitable for Gaussian states and for sketching the eight-moment extension. The bolometer's 'no added noise' property is inherited from a prior published correlation measurement (Ref. 30) and is not the fragile link.\n\nThe fragile link is the combination the reader identified, but I weight the two components differently. The Gaussian-histogram assumption is not a threat to the scoped demonstration (the states are Gaussian by construction) — it is a threat to the scope itself: because only two moments per angle are measured and the rest of the marginal is assumed, the whole demonstration lives inside the five-parameter Gaussian family, and the sparsity/CS/NN/N=3 results amount to Gaussian parameter estimation rather than tomography. This makes the title-level 'computed tomography of propagating microwave photons' materially broader than what is tested, and the N=3 'without compromising quality' claim is true only because every measured state is Gaussian.\n\nThe more dangerous premise for the claim as scoped is the variance-channel calibration: established and verified exclusively on thermal states, where ⟨(Δn)²⟩ is slaved to ⟨n⟩, and then applied to coherent and squeezed states whose only 'quantum' signature (V_min=0.85, 1.6 dB squeezing, 0.7-photon thermal floor) is quantitative and small. The Eq. (2) truncation acknowledged in Fig. 4B is a concrete, unquantified bias of the same scale. Because no measurement in the paper is anchored to an independent detector or an independent prediction of α or ζ from the generator settings, a state-dependent variance offset would change the main results without being detectable in the thermal validation.\n\nThe proposed cross-check — an independent noise-calibrated heterodyne measurement of the same squeezed state — settles both questions at once. If it agrees, the core claim stands and only the generality/phrasing claims need revision; if it disagrees, the central demonstration fails. That is exactly the conditionality the reader assigned, and my stress-test finds no reason to change it.","tokens_in":15762,"tokens_out":23127,"duration_ms":228308,"concrete_test":"Run an independent, noise-calibrated heterodyne measurement of the same squeezed and coherent states in the same cryostat (standard amplified quadrature detection at 8.43 GHz with a calibrated noise source), and compare the extracted V_min, V_max, and α with the bolometer values (0.85, 1.79, α from Fig. 3). Agreement within ~0.1 would validate the variance channel across state statistics; disagreement above the quoted accuracy (~0.2) would show the calibration is state-dependent and the reconstructed Wigner functions are biased. As a complementary analytical check, recompute ⟨(Δn_c)²⟩ with the exact Methods expression for the reported parameters (n̄_T=0.73, ζ=0.19, α=0.55+i0.25, |β|²=15.3, Γ=0.49), including the Γ²⟨(Δn_a)²⟩+Γ(1−Γ)⟨n_a⟩ and cross terms, and quantify the bias in ⟨(ΔX_φ)²⟩ at each angle.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim — amplification-free Wigner-function CT of propagating photons with a bolometer plus two-field interference — requires that the marginal histogram at each angle is Gaussian and fully captured by the measured mean and variance, and that those two numbers are extracted without state-dependent systematic error. Both links are softer than the text implies.\n\n(1) The histograms are never measured. The bolometer yields only ⟨n_c⟩ and ⟨(Δn_c)²⟩ (Eqs. 1–2); the marginal fed into the Hilbert transform is the Gaussian defined by those moments ('Projective bolometry' and 'Computed tomography', Methods). All demonstrated states are Gaussian by construction (linear TWPA squeezing and directional-coupler displacement of thermal states), so within the demonstrated class the procedure is self-consistent — but 'Wigner function CT' thereby reduces to estimating the five-parameter Gaussian family (n̄_T, ζ, α). The sparsity shown in Fig. 4D is the trivial reciprocal-space sparsity of a Gaussian; the N=3 'reconstructions' are linear least-squares fits to the Gaussian model and an NN trained on 32768 Gaussian states, i.e., Gaussian parameter estimation, not CT. The abstract's 'without compromising the reconstruction quality', the 'model-free' NN claim, and the framing as complete characterization of propagating photons overstate what the data test: no non-Gaussian state is attempted, so the tomography claim is never exercised where it is nontrivial.\n\n(2) The quantitative anchor is internal. Γ and η1 are adjusted so thermal-state data match theory (Fig. 2B); the σ²↔⟨(Δn_c)²⟩ calibration curve is established and verified on the same thermal contours (Fig. 2C–E), where variance and mean are locked together.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experiment in which a superconducting SNS bolometer, combined with a strong coherent homodyne field at a 90° hybrid, measures the mean photon number and variance of the combined field. From these, using Eqs. (1) and (2), the authors extract the quadrature mean and variance of the input propagating microwave field. Assuming Gaussian input states, they construct Gaussian marginal histograms and apply computed tomography (Hilbert transform) to reconstruct Wigner functions. They demonstrate the method on thermal, coherent, squeezed, and general Gaussian states, and also show compressed-sensing and neural-network reconstructions with as few as three projection angles. The central claim is that this provides amplification-free, broadband Wigner function tomography of propagating microwave photons.","tokens_in":16069,"tokens_out":6579,"duration_ms":68352,"significance":"If the central claims hold, this would be a significant advance: a passive bolometer would serve as a noiseless quadrature detector and reconstruct Wigner functions of propagating microwave photons without the amplification noise typical of conventional heterodyne detection, with broad bandwidth and potential for multiplexed readout. The experiment is carefully executed: the derivation of Eqs. (1)-(2) is standard beam-splitter physics, the calibration of the transmissivity and insertion losses is detailed, and the data for Gaussian states (thermal, coherent, squeezed) show good agreement with theory. The connection to computed tomography and compressed sensing is interesting, and the availability of data and code is a strength. However, the demonstration is strictly limited to Gaussian states, and this limitation is understated in the title and abstract relative to the generality of the tomography claims.","major_comments":[{"comment":"The reconstruction procedure never measures the marginal histograms h_phi(x_phi) directly; it measures only <n_c> and <(Delta n_c)^2> and then assumes each marginal to be Gaussian. For the Gaussian states studied, the Wigner function is fully determined by five parameters (bar n_T, zeta, alpha), so the Hilbert-transform CT reduces to estimating these parameters from the measured moments. The N=3 reconstructions in Fig. 4F are linear least-squares fits to a Gaussian model or the output of a neural network trained on 32768 simulated Gaussian states; they do not constitute model-free CT. Consequently, the abstract's 'complete characterization' and the text's 'model-free' NN claim overstate what is demonstrated. The authors should either restrict the central claim to Gaussian-state tomography or add a measurement of a non-Gaussian state (e.g., a Fock state) to justify the general tomography framing.","section":"Wigner function CT; Methods: Computed tomography"},{"comment":"The bolometer calibration curves (mu versus <n_c> and sigma^2 versus <(Delta n_c)^2>) are established solely from thermal-state data. Their use for coherent and squeezed states assumes that sigma^2 depends only on the photon-number variance and not on the underlying state statistics or higher moments. The comparisons in Figs. 3H and 3L use the same calibration to extract the quadrature variance, so they do not independently validate the calibration for non-thermal states. The reported squeezing of 1.6 dB and thermal population bar n_T = 0.73 in Fig. 3L are therefore not independently anchored. The authors should provide a direct calibration check with a non-thermal state of known variance, or explicitly state that the variance calibration is validated only for thermal states.","section":"Fig. 2C; Fig. 3H and 3L"},{"comment":"The exact expression for <(Delta n_c)^2> contains terms beyond the leading-order approximation in Eq. (2). The observed 360-degree-periodic oscillations of sigma^2 for the coherent state (Fig. 3G) demonstrate that these terms are non-negligible at the employed |beta|^2 values. The text states that these oscillations 'contribute marginally' to <(Delta X_phi)^2> as |beta|^2 increases, but no quantitative error analysis is provided. Since the quadrature variance extraction for the squeezed state in Fig. 3L relies on Eq. (2), a bias in the extracted squeezing level cannot be ruled out without a systematic-error estimate. Please quantify the error from the discarded terms over the parameter range used.","section":"Eq. (2); Methods: Projective bolometry"}],"minor_comments":[{"comment":"The phrase 'Hilbert transform with no fitting parameter' is misleading because the Gaussian marginals entering the transform are constructed from fitted calibration parameters and a Gaussian assumption; please rephrase to avoid implying that no model or calibration is used.","section":"Fig. 4C caption"},{"comment":"The sentence 'A possible advantage of the NN-approach is the model-free architecture' is inaccurate: the neural network used here is trained on 32768 simulated Gaussian states and is therefore a Gaussian-specific regression, not a model-free estimator. Please correct the wording.","section":"p. 17 (ultra-sparse sampling paragraph)"},{"comment":"The Gaussian-state assumption is introduced only parenthetically ('Assuming a Gaussian state of the input field, as it is the common choice...'). Given that this assumption is central to the reconstruction, it should be highlighted in the abstract and the conclusions.","section":"Main text, p. 5"},{"comment":"The phrase 'combing a blackbody radiator' appears to be a typo for 'combining a blackbody radiator'.","section":"Main text, p. 6"},{"comment":"The exact expression for <(Delta n_c)^2> is written as a multi-line inline equation that is difficult to parse; a displayed equation with terms explicitly grouped would improve readability.","section":"Methods: Projective bolometry"}],"recommendation":"major_revision","confidential_remarks":"This is a solid experimental paper from a reputable group, and the Gaussian-state demonstration appears technically sound. The main weakness is the overstatement of the tomography claim: the method, as demonstrated, is Gaussian-state parameter estimation rather than full Wigner-function CT, and the calibration is validated only on thermal states. These issues are fixable with revised scope, additional calibration checks, and a quantitative error analysis, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Punchline: the bolometer quadrature technique is a real advance for passive, broadband, noiseless detection of propagating microwave fields, but what they call computed tomography is actually Gaussian-state estimation because only the mean and variance of each marginal are measured. The N=3 'model-free' reconstructions are fits to a Gaussian model.\n\nCredit where due: The two-field interference idea is new in bolometry and the experiment is carefully done. They see the expected phase dependence for thermal, coherent, and squeezed states, extract squeezing and displacement parameters that are consistent with the generator settings, and provide data and code. The derivation of Eqs. (1)-(2) is standard, and the finite-|β| correction for the variance is a nice touch. The bolometer's broad bandwidth and lack of added noise are genuine advantages over qubit-based parity detectors.\n\nSoft spots: First, the histograms used for CT are never measured. The bolometer returns only <n_c> and <(Δn_c)^2>, and the paper assumes a Gaussian marginal to define the histogram. That is fine for the Gaussian states they generate, but it collapses 'Wigner function CT' to recovering at most five Gaussian parameters. The N=3 results are linear least squares or NN fits to Gaussian states, so claiming 'without compromising reconstruction quality' overstates what the data test. Second, the calibration of Γ, η1, and the variance response is fitted to thermal-state data. The thermal check in Fig. 2E is therefore partly circular. I disagree with the stress-test claim that the squeezing result is unanchored: the extracted 0.85/1.79 quadrature variances reproduce the expected 1.6 dB squeezing and are in line with the TWPA's known behavior, so the calibration is not just being fit to the answer. But a direct test of the variance calibration on a non-thermal state with known variance would have been stronger.\n\nThe authors are transparent about the Gaussian assumption in the main text and Methods, so this is not a hidden flaw. It is a framing problem. The abstract and the 'complete characterization' language oversell the method's scope.\n\nFor whom: experimentalists working on microwave quantum networks or single-photon detection will find this valuable. It deserves peer review; the core technique is solid and the overclaims are fixable with rewriting and a clearer statement that non-Gaussian states require higher moments (which they already mention). I would send it to review.","headline":"Genuinely new bolometer quadrature detection, honestly demonstrated for Gaussian states, but the CT and N=3 'model-free' framing overstates the method's scope.","tokens_in":16683,"tokens_out":3623,"would_cite":true,"duration_ms":40154,"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 passive cryogenic bolometer, combined with a strong reference tone and medical-imaging mathematics, reconstructs the full quantum state of traveling microwave photons without any amplifier noise.","keywords":["Wigner function tomography","propagating microwave photons","SNS bolometer","quadrature detection","computed tomography","Gaussian quantum states","compressed sensing","quantum state reconstruction"],"falsifier":"Generate a non-Gaussian propagating microwave state, such as a single-photon wave packet or a photon-subtracted squeezed state, run the same phase-sweep protocol, and compare the reconstructed Wigner function with one obtained from an independent photon-number-resolving or parity measurement; the two will disagree if the Gaussian-histogram assumption is the load-bearing premise.","tokens_in":15544,"feed_emoji":"📡","tokens_out":14126,"duration_ms":128010,"temperature":0.7,"pith_summary":"This paper reports a measurement technique that reconstructs the full quantum state—the Wigner function—of microwave photons traveling along a transmission line, without amplifying the signal at any stage. The authors show that by letting the incoming field interfere with a strong coherent reference field before it is absorbed, a superconductor–normal-metal–superconductor (SNS) bolometer—a cryogenic resistive heat detector—can be used as a noiseless quadrature detector. Sweeping the reference phase produces a set of quadrature histograms, and the mathematics of computed tomography (CT) recovers the two-dimensional phase-space distribution from these projections. The demonstration is carried out on Gaussian states at the single-photon level, and compressed sensing or a neural network reduces the required number of projections to three. If correct, this gives superconducting quantum networks a passive, broadband, full-duty-cycle way to characterize propagating microwave photons and to diagnose errors in real time.","feed_headline":"Bolometer maps traveling microwave photons without amplifier noise","feed_subtitle":"Passive cryogenic detector and CT math yield full quantum-state pictures at the single-photon level.","key_machinery":"The load-bearing mechanism is two-field interference in power detection. The input field $\\hat a$ is combined on a $90^\\circ$ hybrid with a strong coherent homodyne field $|\\beta|e^{i\\phi}$; the SNS bolometer then reads the photon-number mean and variance of the combined field, and in the large-$|\\beta|^2$ limit these readings are affine in the input quadrature mean and variance (Eqs. (1) and (2)). Because the input is assumed Gaussian, each quadrature marginal is fully specified by those two numbers, and sweeping $\\phi$ samples the Radon transform of the Wigner function. The Hilbert transform inverts the Radon transform in reciprocal space, giving a parameter-free reconstruction of $W(x,p)$ from the measured histograms.","core_discovery":"The central claim is that a passive SNS bolometer, which measures the mean photon number and photon-number variance of absorbed radiation, becomes a quadrature detector when the input field is mixed with a strong coherent homodyne field on a beam splitter. In the large-homodyne limit, the measured mean and variance of the combined field are linear in the input quadrature mean $\\langle \\hat X_{\\phi+90}\\rangle$ and variance $\\langle(\\Delta \\hat X_{\\phi+90})^2\\rangle$. Assuming the input state is Gaussian, these two numbers completely determine the marginal histogram at each projection angle, so sweeping the homodyne phase and applying the Hilbert transform reconstructs the Wigner function $W(x,p)$. The paper demonstrates this protocol on thermal, coherent, and squeezed states at the single-photon level and shows that compressed sensing or a neural network reconstructs the same Wigner function from as few as three projections.","pith_inferences":["The paper only validates the Gaussian-assumption protocol; a natural extension is to feed a non-Gaussian propagating state and fit the eight moments that enter the exact expressions for $\\langle \\hat n_c\\rangle$ and $\\langle(\\Delta \\hat n_c)^2\\rangle$, which would let the same bolometer reconstruct states with Wigner negativity.","If the bolometer calibration curves remain valid for non-Gaussian states and other phases—checked here only for Gaussian states—the technique becomes a direct real-time monitor of photon loss and thermal contamination in a quantum network link.","The reconstruction-quality saturation at about 12 projections suggests that the effective information content of these Gaussian states is small; testing how the saturation point moves with homodyne power and with non-Gaussian states would separate the role of the state's Gaussianity from that of measurement noise."],"forward_implications":["Propagating microwave photons can be fully characterized at millikelvin temperatures without any amplifier, removing the added noise that earlier propagating-photon tomography had to deconvolve statistically.","The detector is passive and broadband—here a 133 MHz window centered at 8.43 GHz—and measures with full duty cycle, unlike qubit-based counters that need dead time and dedicated control and readout circuitry.","Sweeping the homodyne phase over 0–360 degrees recovers the displacement and squeezing parameters of the state; reconstruction quality saturates at about N=12 projections, and compressed sensing gives comparable results from N=9 projections.","With the Gaussian-state model, three projective measurements suffice for reconstruction via linear least-squares fitting or a neural network trained on simulated states, pointing toward real-time Wigner function monitoring.","Because the homodyne field boosts the input power without adding noise, the same scheme could detect signals at the yocto- to zepto-joule scale and, through transduction, be extended to other particle types such as surface-acoustic-wave phonons."],"supporting_citations":[{"why":"Supplies the SNS bolometer that measures the mean photon number and variance of absorbed radiation; the quadrature scheme builds directly on this detector.","marker":"30"},{"why":"Provides the continuous-variable tomography formalism connecting quadrature marginals to the Wigner function.","marker":"31"},{"why":"Demonstrates quantum state tomography of an itinerant squeezed microwave field, the propagating-mode context this work extends.","marker":"32"},{"why":"Supplies the computed-tomography and Radon-transform framework adapted for Wigner function CT.","marker":"33"},{"why":"Provides the calibration and self-calibration approach for the beam splitter and blackbody radiator used to set the photon-number scale.","marker":"35"},{"why":"Supplies the compressed-sensing theory that enables Wigner function reconstruction from sparse projections.","marker":"40"},{"why":"Provides the recovery guarantees for sparse signal reconstruction used in the compressed-sensing implementation.","marker":"41"},{"why":"Documents the bolometer's detection principle and fabrication, supporting the claim of passive, broadband, noiseless operation.","marker":"26"}],"fun_headline_variants":["Passive bolometer does CT on flying photons","Noiseless tomography of traveling microwave photons","From three angles to full Wigner functions","Single-photon Wigner maps without amplifiers"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reconstruction assumes that every projection of the input field is Gaussian, so that measuring just the mean and variance of each quadrature histogram fully determines that histogram; for non-Gaussian light the two bolometer readings do not fix the state.","fun_headline_variants_meta":{"raw":{"variants":["Passive bolometer does CT on flying photons","Noiseless tomography of traveling microwave photons","From three angles to full Wigner functions","Single-photon Wigner maps without amplifiers"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000217,"raw_usage":{"total_tokens":1429,"prompt_tokens":931,"completion_tokens":498,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":547,"completion_tokens_details":{"reasoning_tokens":442}},"tokens_in":547,"tokens_out":498,"duration_ms":6208,"temperature":1.0,"reasoning_tokens":442,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:51:24.252310+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Generate a non-Gaussian propagating microwave state, such as a single-photon wave packet or a photon-subtracted squeezed state, run the same phase-sweep protocol, and compare the reconstructed Wigner function with one obtained from an independent photon-number-resolving or parity measurement; the two will disagree if the Gaussian-histogram assumption is the load-bearing premise.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the SNS bolometer that measures the mean photon number and variance of absorbed radiation; the quadrature scheme builds directly on this detector."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the computed-tomography and Radon-transform framework adapted for Wigner function CT."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the recovery guarantees for sparse signal reconstruction used in the compressed-sensing implementation."},{"cited_title":"E., Tan, K","cited_arxiv_id":null,"evidence_quote":"Documents the bolometer's detection principle and fabrication, supporting the claim of passive, broadband, noiseless operation."}],"review_version":1}