{"id":"239590cf-99e1-452a-b592-d2a44b3ab255","arxiv_id":"1908.03598","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A two-mode superconducting bosonic processor sampled molecular vibronic spectra of H2O, O3, NO2, and SO2 and demonstrated single-shot photon-number-resolving detection up to 15 photons per mode.","lead":"A superconducting microwave chip ran a quantum simulation that reproduces the vibration patterns of small molecules when they gain or lose an electron. The chip also demonstrated a new detector that counts up to 15 microwave photons at once, a step toward quantum computers that simulate chemistry and materials.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The scalability/PNR claim is supported only by raw sampling data whose measurement errors are explicitly left uncharacterized; the paper's own text defers this to future work, so \"high-fidelity single-shot PNR up to 15 photons\" is not yet demonstrated.","rationale":"The core two-mode simulation is credible: single-bit FCFs match theory (D = 0.019–0.105), master-equation baselines are provided, and calibrations are detailed. The conditional verdict is warranted not by an identified mistake but by an unquantified error source that sits exactly on the headline claim. The reader's Eq. S29 concern is valid, but it applies only to the corrected single-bit branch; the sampling branch is the one used for the scalability claim and is explicitly uncorrected. Thus I recommend no change to the conditional verdict. If the confusion-matrix test passes, the claim could be upgraded to accept; if it fails, the scalable-sampling claim would need to be rejected or substantially narrowed.","tokens_in":38813,"tokens_out":5529,"duration_ms":62172,"concrete_test":"Prepare one cavity in each Fock state |n> (n = 0,...,15) with the other cavity in vacuum, skip the Doktorov gates, and run the full four-bit sampling sequence (parity pulses with feedforward resets) exactly as in the experiment; build the 16x16 confusion matrix C_{m,n} giving the probability of outcome m given preparation n. If the per-state success probabilities C_{n,n} are high (e.g. > 0.9 for all n), the \"high-fidelity PNR up to 15\" claim is confirmed. Then invert C (or apply iterative unfolding as in ref. 19) to the sampling data of Fig. 3 and recompute D; if the corrected D matches the single-bit values within statistical error, the scalable sampling implementation is validated, whereas large corrected-D residuals would show the reported sampling distances were dominated by measurement error.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest claim in the abstract—fulfilling the scalability requirement by demonstrating, for the first time in any platform, a high-fidelity single-shot photon-number-resolving scheme up to 15 photons—is not backed by a characterization of the sampling chain. Section IV.B introduces four sequential QND parity pulses with transmon resets and states: \"We leave the task of characterizing these errors in full detail... as the subject of future work.\" Section V then reports sampling distances D = 0.075–0.209 computed from raw sampled counts with \"no correction protocol\" (Supplementary VI), while Table I estimates measurement-induced error κτ_meas ~ 10^-2–10^-1. Because no confusion matrix for the parity pulses is given, the 15-photon resolving capability is an assertion, not a demonstrated fact, and the D values for the sampling scheme do not separate detector error from simulator error. The single-bit extraction data are corrected (Eq. S29) and support the two-mode simulation itself, so this concern specifically undercuts the scalable-sampling/PNR component of the central claim. A secondary related worry is the S29 assumption of photon-number-independent ancilla Rabi and decoherence rates; if violated, the corrected single-bit probabilities carry a bias, but that can be re-inverted once n-dependent f and t are measured. The missing sampling-chain characterization is the more load-bearing gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental implementation of the boson-sampling algorithm for molecular vibronic spectra on a two-mode superconducting bosonic processor. The authors prepare Fock states in two microwave cavities, apply displacement, squeezing, and beamsplitter operations corresponding to the Doktorov transformation for four triatomic molecules (H2O, O3, NO2, SO2), and detect the output photon-number distributions using two complementary readout schemes: a 'single-bit extraction' method based on selective ancilla pulses and a 'sampling' scheme based on sequential QND parity measurements that aim to resolve up to 15 photons per mode. The measured Franck-Condon factors are compared to ideal classical calculations, yielding distances D in the range 0.019-0.105 for the single-bit scheme and 0.075-0.209 for the sampling scheme. The central claims are (i) the demonstration of a scalable photon-number-resolving detection scheme and (ii) accurate simulation of molecular vibronic spectra in a superconducting platform.","tokens_in":39106,"tokens_out":3132,"duration_ms":36162,"significance":"If the claims are fully supported, this is an important experimental milestone in bosonic quantum simulation. The paper has clear strengths: the Doktorov parameters are computed with independent quantum-chemistry software (Gaussian/ezSpectrum) at the CCSD(T) level; the Gaussian operations are calibrated against physics-based models; the ideal Franck-Condon target distributions are computed classically without using the measured data; and the single-bit-extraction data agree with the ideal distributions with distances as low as 0.019, with full time-domain master-equation simulations (including self-Kerr and photon loss) provided as baselines. The demonstration of non-Gaussian input-state synthesis for vibrationally excited initial states is also a meaningful advance over previous trapped-ion emulations. However, the scalable sampling/PNR component, which is highlighted in the abstract, is not yet backed by a characterization of the measurement chain; this is the main gap that must be addressed in revision.","major_comments":[{"comment":"The claim of 'for the first time in any platform, a high-fidelity single-shot photon number resolving detection scheme capable of resolving up to 15 photons per mode' is not supported by data in the manuscript. Section IV.B states that the task of characterizing the errors of the four sequential QND parity pulses is left to future work, and no confusion matrix or assignment-error analysis for the parity measurements is provided. Consequently, the sampling distances D reported in Table V and Fig. 3 are computed from raw counts without any correction protocol (Supplementary VI), so they do not separate detector error from simulator error. The estimate of measurement-induced error in Table I (κτ_meas ~ 10^-2-10^-1) indicates that these errors are not negligible. To substantiate the headline claim, the authors should provide a characterization of the sampling chain, e.g., by calibrating the bit-assignment probabilities for each Fock state up to 15 and reporting the resulting corrected distribution, or by explicitly qualifying the claim as 'demonstrated at the level of uncharacterized raw sampling'.","section":"IV.B and abstract"},{"comment":"The correction of the single-bit-extraction data relies on the assumption that the ancilla Rabi and decoherence rates have no photon-number dependence. The manuscript does not provide a test of this assumption across the relevant range n=0-15. If the Rabi frequency or the decoherence rates vary with cavity photon number, then the corrected Franck-Condon probabilities obtained from Eq. (S29) carry an unquantified systematic bias, and the quoted distances D for the single-bit scheme (e.g., D=0.049 for H2O and 0.019 for SO2) would be distorted. This is load-bearing because the single-bit results are the primary evidence for the accuracy of the two-mode simulation. The authors should either present a calibration of f and t as a function of photon number, or provide a theoretical and experimental argument that the variation is negligible at the reported level of precision.","section":"Supplementary V, Eq. (S29)"}],"minor_comments":[{"comment":"The notation for the sampling measurement outcome, |n′⟩ = |b3,b2,b1,b0⟩, uses commas inside the ket, which is nonstandard and potentially confusing; the authors should define b_i as the bits of the binary decomposition and write e.g. |n′⟩ = |b3 b2 b1 b0⟩.","section":"Table I"},{"comment":"The caption states that statistical error bars for the sampling data are not visible; it would be helpful to report approximate error bar sizes in the caption or show them on a zoomed inset.","section":"Fig. 3 caption"},{"comment":"In the expression for χ_iC, the argument of the absolute value is written as |g_i/δ_i|^2 with a missing subscript on δ; the authors should write δ_i to be consistent with the rest of the notation.","section":"Supplementary II.A, Eq. (S17)"},{"comment":"The phrase 'for the first time in any platform' is a strong priority claim. The authors should verify that no prior work has demonstrated single-shot photon-number resolution up to 15 photons in any platform, and add a citation or explicitly scope the claim to the circuit-QED platform.","section":"Abstract"},{"comment":"The table of numerical data would be easier to use if the columns for 'Single-bit extraction' and 'Sampling' included the number of runs N_runs used for each measurement, since the reported standard errors depend on it.","section":"Supplementary VI"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a strong experimental contribution, but the most prominent claim in the abstract (the single-shot PNR demonstration) is currently not substantiated by a characterization of the measurement errors. The single-bit extraction results are more solid and could carry the simulation claim on their own. I would recommend major revision: the authors should either add a detailed error characterization for the sampling chain (which may require new experimental data) or carefully restate the claims to distinguish the demonstrated two-mode simulation from the scalable PNR scheme that is described but not yet fully validated. I have no concerns about circularity; the Doktorov parameters are independently computed, and the ideal targets are obtained classically."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: the single-bit extraction results are the real work, and they are good; the scalable-sampling/PNR fifteen-photon claim is not yet demonstrated at the claimed fidelity. The paper says so itself, in Section IV.B, where it defers full characterization of the binary-decomposition measurement chain to future work. The abstract still calls it \"high-fidelity single-shot PNR up to 15 photons,\" which overstates what is shown.\n\nWhat is genuinely new: a two-mode superconducting bosonic processor that implements the Doktorov unitary and samples FCFs for four molecules, including non-Gaussian Fock inputs up to n=1,m=2. The QND parity measurement scheme itself—four sequential pulses resolving the binary decomposition—is a new hardware capability in this platform, even if the error budget is incomplete. The single-bit extraction data, corrected via Eq. S29 and compared against master-equation simulations with self-Kerr and loss, match the ab initio FCFs with distances 0.019 to 0.105. The Doktorov parameters come from independent CCSD(T) calculations, so there is no obvious circularity. No fitted parameter is doing the work.\n\nSoft spots, in order of importance. First, the sampling chain: the D=0.075–0.209 values for the sampling scheme are raw, uncorrected, and include detector error indistinguishably. The paper directly says characterization is future work. So the headline scalability claim rests on an assertion, not a measurement. That should be fixed or the claim softened. Second, the correction model for single-bit extraction assumes the ancilla Rabi and decoherence rates are photon-number independent. That is plausible but untested over n=0..15; if wrong, the corrected probabilities carry a small unquantified bias. Minor relative to the first. Third, two modes is a small instance; the paper is honest about it.\n\nVerdict: worth a serious referee. The experimental core is solid, the supplementary data is unusually complete, and the single-bit demonstration alone supports the paper's main physical point. A referee should require the authors to either add the measurement characterization or revise the abstract's fidelity claim. The paper would be stronger with both.","headline":"The single-bit FCF measurements are solid; the fifteen-photon sampling claim is ahead of its evidence.","tokens_in":39697,"tokens_out":3183,"would_cite":true,"duration_ms":35137,"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 superconducting two-mode bosonic processor efficiently samples molecular vibronic spectra using single-shot photon-number-resolved detection.","keywords":["boson sampling","vibronic spectra","Franck-Condon factors","superconducting circuits","photon-number-resolving detection","Doktorov transformation","circuit quantum electrodynamics","Gaussian operations"],"falsifier":"Prepare a cavity state with a known photon-number distribution, such as a displaced squeezed state, measure it with the single-bit scheme for every photon number up to 15, apply the paper's correction, and compare with the exact distribution; if the corrected error grows with photon number, the assumption of photon-number-independent auxiliary-qubit rates is false. The same check can be done by measuring the auxiliary qubit's Rabi and T1 rates conditioned on $n = 0,1,\\dots,15$.","tokens_in":38595,"feed_emoji":"🧪","tokens_out":7016,"duration_ms":67797,"temperature":0.7,"pith_summary":"This paper tries to establish that a naturally bosonic quantum processor—two microwave cavities in a superconducting circuit—can efficiently simulate the vibrational spectra of molecules, a task that maps onto multiphoton boson sampling. The authors implement the Doktorov unitary that connects vibrational states across an electronic transition and, crucially, demonstrate a single-shot photon-number-resolving detector that resolves up to 15 photons per mode, which removes the exponential overhead of one-bit-at-a-time readout. They then reprogram the processor to reproduce Franck-Condon spectra for four molecular processes, obtaining distances to ideal distributions as low as 0.019. A sympathetic reader would take the paper to show that superconducting bosonic hardware is a practical and scalable platform for this class of bosonic quantum simulations, including vibrationally hot initial states prepared as Fock states.","feed_headline":"Microwave photons mimic molecular vibronic spectra","feed_subtitle":"A two-cavity processor resolves up to 15 photons per mode and reproduces Franck-Condon spectra for four molecules.","key_machinery":"The central object is the Doktorov transformation $\\hat U_{\\rm Dok} = \\hat D(\\alpha) \\hat S^\\dagger(\\zeta') \\hat R(U) \\hat S(\\zeta)$, which maps the pre-transition vibrational ladder operators to the post-transition ones. It is implemented by displacement, single-mode squeezing, and a two-mode beamsplitter produced by four-wave mixing with a coupler transmon. The other load-bearing mechanism is the QND photon-number-resolving detector: parity operators $\\hat P_k$ that read the $k$th bit of the binary decomposition of the photon number are mapped onto an auxiliary transmon by optimal-control pulses, so four sequential measurements resolve $n = 0,\\dots,15$ in each mode and return a single sample from the joint Fock-state distribution per run.","core_discovery":"The central claim is that molecular vibronic spectra—the Franck-Condon factors $\\left|\\langle \\vec n' | \\hat U_{\\rm Dok} | \\vec n \\rangle\\right|^2$—can be extracted from a two-mode superconducting bosonic processor by directly realizing the Doktorov transformation $\\hat U_{\\rm Dok} = \\hat D(\\alpha) \\hat S^\\dagger(\\zeta') \\hat R(U) \\hat S(\\zeta)$ as a sequence of displacements, single-mode squeezes, and one beamsplitter, each generated by four-wave mixing on a Josephson coupler. The paper reports the first single-shot QND photon-number-resolving detector, in any platform, that resolves up to 15 photons per mode by reading the binary decomposition of the photon number onto an auxiliary transmon; this turns each run into a direct sample of the joint Fock-state distribution rather than a single bit of one joint occupation. With these tools, the measured spectra for photoelectron processes in H$_2$O, O$_3$, NO$_2$, and SO$_2$ sit close to the ideal Franck-Condon distributions, with total-variation distances $D$ from 0.019 to 0.105 for the corrected single-bit-extraction scheme and 0.075 to 0.209 for the scalable sampling scheme.","pith_inferences":["The reported sampling distances are noisier than the corrected single-bit distances; the paper leaves open whether deconvolution of the bit-wise readout errors, for example by unfolding methods, can close that gap, which is a direct testable next step.","The weakest premise is the photon-number independence of the auxiliary qubit's Rabi and decoherence rates used in the correction; a dedicated calibration at each $n = 0,\\dots,15$ would either validate all reported $D$ values or expose a systematic bias in them.","The same hardware elements—Gaussian operations plus programmable self-Kerr nonlinearity—are enough to simulate time-domain vibrational dynamics and anharmonic bosonic models, though the paper only gestures toward those targets.","If the architecture scales as proposed, a many-cavity linear array would be a general-purpose simulator of harmonic vibrational structure for polyatomic molecules, with the measurement cost growing only logarithmically in the per-mode Hilbert-space dimension."],"forward_implications":["Single-shot photon-number-resolving measurement of up to 15 photons per mode makes the sampling scheme scalable: $N\\log_2(n_{\\max})$ binary QND measurements return one full sample from the joint distribution, instead of querying each of the exponentially many joint Fock states.","The native bosonic implementation of the Doktorov unitary costs $O(N^2)$ Gaussian operations and $O(N)$ circuit depth, compared with $O(N^2 n_{\\max}^2 \\log^3(1/\\varepsilon))$ gates for a qubit-based algorithm at fixed error $\\varepsilon$.","Because the same device reproduces the spectra of H$_2$O, O$_3$, NO$_2$, and SO$_2$ by changing only the Doktorov parameters, the processor is reprogrammable across molecules and initial vibrational states.","The ability to prepare non-Gaussian Fock states as inputs lets the simulator start from vibrationally excited ensembles, not just the ground state.","Adding a third cavity mode would extend the simulation to nonlinear triatomic molecules of $C_s$ symmetry, and a linear array of $N$ modes covers molecules with up to $3M-6$ vibrational degrees of freedom."],"supporting_citations":[{"why":"Supplies the algorithmic mapping from molecular transition data to a generalized boson-sampling problem that the experiment implements.","marker":"[14]"},{"why":"Supplies the Doktorov unitary factorization of vibronic transitions used as the target transformation.","marker":"[17]"},{"why":"Proposes that superconducting circuits can perform microwave boson sampling, motivating the architecture.","marker":"[13]"},{"why":"Demonstrates the programmable two-mode beamsplitter interaction in circuit QED used here.","marker":"[12]"},{"why":"Provides the theory of bilinear mode coupling via a driven transmon that underlies the squeezing and beamsplitter gates.","marker":"[20]"},{"why":"Earlier trapped-ion emulation of vibronic spectra whose detection scheme the paper's sampling measurement improves on.","marker":"[16]"},{"why":"Demonstrates resolving photon number states in a superconducting circuit, the basis of QND photon-number measurement.","marker":"[7]"},{"why":"Demonstrates QND detection of single microwave photons, the foundation of the readout.","marker":"[8]"},{"why":"Qubit-based algorithm for vibronic spectra used as the resource comparison that justifies the native bosonic advantage.","marker":"[21]"},{"why":"Decomposition of any unitary into two-mode beamsplitters, used for the rotation part of the Doktorov transformation.","marker":"[18]"}],"fun_headline_variants":["Superconducting boson sampling extracts molecular vibronic spectra","First 15-photon resolving detector powers molecular spectrum simulation","Bosonic processor simulates vibronic spectra with 15-photon resolution","15-photon resolution in a bosonic processor mimics molecular spectra","Microwave boson sampling reproduces molecular Franck-Condon spectra"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The correction applied to the most accurate data assumes that the auxiliary qubit's rotation speed and its decay and heating rates are the same whether the cavity holds 0, 1, or 15 photons; if those rates actually change with photon number, every corrected Franck-Condon probability carries an unquantified bias.","fun_headline_variants_meta":{"raw":{"variants":["Superconducting boson sampling extracts molecular vibronic spectra","First 15-photon resolving detector powers molecular spectrum simulation","Bosonic processor simulates vibronic spectra with 15-photon resolution","15-photon resolution in a bosonic processor mimics molecular spectra","Microwave boson sampling reproduces molecular Franck-Condon spectra"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000944,"raw_usage":{"total_tokens":4091,"prompt_tokens":1066,"completion_tokens":3025,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":682,"completion_tokens_details":{"reasoning_tokens":2938}},"tokens_in":682,"tokens_out":3025,"duration_ms":21770,"temperature":1.0,"reasoning_tokens":2938,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:08:53.293151+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Prepare a cavity state with a known photon-number distribution, such as a displaced squeezed state, measure it with the single-bit scheme for every photon number up to 15, apply the paper's correction, and compare with the exact distribution; if the corrected error grows with photon number, the assumption of photon-number-independent auxiliary-qubit rates is false. The same check can be done by measuring the auxiliary qubit's Rabi and T1 rates conditioned on $n = 0,1,\\dots,15$.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Doktorov unitary factorization of vibronic transitions used as the target transformation."},{"cited_title":"Peropadre, G","cited_arxiv_id":null,"evidence_quote":"Proposes that superconducting circuits can perform microwave boson sampling, motivating the architecture."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates the programmable two-mode beamsplitter interaction in circuit QED used here."},{"cited_title":"Zhang, B","cited_arxiv_id":null,"evidence_quote":"Provides the theory of bilinear mode coupling via a driven transmon that underlies the squeezing and beamsplitter gates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates resolving photon number states in a superconducting circuit, the basis of QND photon-number measurement."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates QND detection of single microwave photons, the foundation of the readout."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Qubit-based algorithm for vibronic spectra used as the resource comparison that justifies the native bosonic advantage."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Decomposition of any unitary into two-mode beamsplitters, used for the rotation part of the Doktorov transformation."}],"review_version":1}