{"id":"b338f9d4-ac1d-4732-858c-c9a58119c55c","arxiv_id":"2501.09343","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A spectrally tailored 4 W 698 nm clock laser, stabilized to two reference cavities, achieves 0.99964(3) single-qubit Clifford gate fidelity averaged over 3000 strontium atoms.","lead":"This paper reports a high-power 698 nm clock laser with very low frequency noise, used to drive quantum bits in 3000 strontium atoms with a record average gate fidelity of 99.964%. It also introduces a pulse sequence that lets the atoms themselves measure the laser's noise at the point of use, making the laser optimization directly verifiable.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline fidelity rests on a single-exponential depolarizing fit with fixed offset; a free-offset or multi-exponential re-fit of the RB data is needed to rule out bias from coherent or spatially inhomogeneous errors.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: the conversion of raw RB decay data into the headline fidelity depends on the specific model in Eq. (6). I would sharpen it: the issue is not merely that errors may be non-depolarizing—single-qubit RB is designed to handle fairly general gate-independent noise—but that the fixed functional form A = 1/2 and the assumption of a single exponential are unnecessarily restrictive for an ensemble of ~3000 atoms. The paper's spatial histogram is suggestive but not quantitative, and the numerical simulation uses the same RB fitting equation and omits coherent-error sources, so the agreement in Fig. 4(d) does not independently validate the model choice. The other concerns noted by the reader—unverified 4 W operation, inherited long-term stability not measured on the final beam, and the record claim lacking a quantitative baseline—are real but secondary; they affect packaging and extensibility claims, not the numerical value of F1^2 if the RB extraction is sound. Therefore I would keep the reader's CONDITIONAL verdict and make the re-fit of the RB decay the specific condition that would settle the concern.","tokens_in":25075,"tokens_out":15433,"duration_ms":199021,"concrete_test":"Request the raw (L, p_e, N) data behind Fig. 4(b) and re-fit the decay in three ways: (i) Eq. (6) as published; (ii) the unconstrained RB form F_L = A + B p^L with A and B free; (iii) a two-exponential mixture A_1 p_1^L + A_2 p_2^L + B. Compare the per-Clifford infidelity (1-p)/2 across fits and use an F-test or AIC to determine whether the two-exponential model is statistically preferred. If the extracted F1^2 shifts by more than the quoted 3e-5, or if the two-exponential fit is clearly preferred, then the depolarizing-model assumption is biasing the headline fidelity and the claim should be revised; if not, the concern is refuted.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Sec. V extracts F1^2 = 0.99964(3) by fitting RB survival data to Eq. (6), which fixes the baseline at 1/2 and lumps all state-preparation and measurement error into d_SPAM. The quoted uncertainty is purely statistical. Standard RB is robust to gate-independent Markovian noise, but Eq. (6) is not the general RB form F_L = A + B p^L; fixing A = 1/2 can bias p when SPAM is not purely depolarizing. More importantly, a single exponential is not guaranteed if spatial inhomogeneity in Rabi frequency or detuning, or other coherent errors, are present: the cloud-averaged signal is then a weighted sum of exponentials, and the fitted p is a model-dependent effective rate rather than the true average gate fidelity. The spatial histogram in Fig. 4(c) is intended to rule out inhomogeneity, but it gives no quantitative width and no comparison of single-exponential versus multi-exponential fits. The Appendix C simulation includes only laser phase noise plus independently measured decay rates; it omits intensity noise, detuning, and pulse-area inhomogeneity, so the agreement in Fig. 4(d) does not by itself certify the depolarizing-model fit. This is the most load-bearing assumption because it directly converts raw survival data into the headline number.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a 698 nm clock laser system producing up to 4 W, whose frequency noise is spectrally tailored by locking an ECDL to a room-temperature ULE cavity and transferring the long-term stability of a cryogenic silicon cavity via a frequency comb and a 500 Hz phase-locked loop. The authors introduce an atom-based spectrum analyzer based on a pulse sequence designed to suppress intensity-noise sensitivity, validate the laser noise model on-site, and demonstrate randomized benchmarking of single-qubit Clifford gates on ~3000 atoms, extracting an average gate fidelity F1^2 = 0.99964(3) at a Rabi frequency of 741.9(5) Hz. A master-equation simulation using the measured phase-noise spectrum reproduces the Rabi-frequency dependence of the infidelity.","tokens_in":25339,"tokens_out":5572,"duration_ms":52073,"significance":"The main technical achievements are a several-watt clock laser with a low noise floor across 10 Hz to 10 kHz, an atom-based noise characterization method with a tunable, singly-peaked spectral response, and a high-fidelity demonstration on a large atomic ensemble. The noise model is carefully constructed from independent cross-correlation and in-loop measurements, and the atom-based validation at the site of the atoms is a valuable addition. The numerical simulation provides a mechanistic link between the measured laser noise spectrum and the gate infidelity. If the reported fidelity is robust to the modeling assumptions discussed below, the result would represent a notable advance in optical qubit control for scalable neutral-atom systems.","major_comments":[{"comment":"The reported single-qubit fidelity F1^2 = 0.99964(3) is extracted by fitting the RB data to Eq. (6), which fixes the depolarizing baseline at 1/2 and attributes all state-preparation-and-measurement errors to d_SPAM. As the authors note, standard RB for gate-independent Markovian noise has the more general form F_L = A + B p^L; fixing A = 1/2 can bias p when SPAM errors are not purely depolarizing or when the initial state preparation is imperfect. Moreover, if the cloud-averaged signal contains a spread of Rabi frequencies or detunings, the survival probability would be a weighted sum of exponentials, so the fitted p is an effective rate rather than the true average gate fidelity. The histogram in Fig. 4(c) supports homogeneity but gives no quantitative width or a comparison of single- versus multi-exponential fits. I request that the authors (i) report a free-parameter fit F_L = A + B p^L and examine whether A is consistent with 1/2, (ii) test for multi-exponential decay (e.g., by fitting to two depolarizing channels weighted by a measured Rabi-frequency distribution), and (iii) quantify the resulting systematic uncertainty on F1^2. This is the load-bearing step that converts the raw survival data into the headline claim.","section":"Sec. V, Eq. (6)"},{"comment":"The abstract states that the high-power 698 nm clock laser exhibits a long-term instability of 3.5e-17 at 1 to 1000 s, but the text attributes this value to the thermal-noise floor of the Si3 cavity (Ref. [80]) and does not report a direct measurement of the spectrally tailored 698 nm laser or the high-power fiber laser at these time scales. The atom-based measurements in Fig. 3(e) probe Fourier frequencies from roughly 10 Hz to 1 kHz, which do not constrain the sub-Hz regime responsible for 1-1000 s instability. Please clarify whether the 3.5e-17 number was measured on the final laser or transferred from the Si3 cavity, and state the expected contribution of the comb transfer, the 500 Hz phase-locked loop, and the high-power phase lock to the low-frequency instability.","section":"Abstract and Sec. III"},{"comment":"The numerical simulation of the RB fidelity includes only laser phase noise and independently measured decay/decoherence rates; it omits intensity noise, detuning, and pulse-area inhomogeneity. Since the agreement between simulation and experiment is used to support the interpretation that the gate infidelity is dominated by laser frequency noise, the simulation should either include these additional error sources or provide a quantitative argument for their negligible contribution at the level of 3.6e-4 infidelity. Without this, the agreement in Fig. 4(d) cannot certify that the fitted F1^2 is unbiased by the omitted effects.","section":"Appendix C and Fig. 4(d)"}],"minor_comments":[{"comment":"The reported chi-squared value for the sensitivity-function fit (chi2 = 1.540) should include the number of degrees of freedom so that the goodness of fit can be assessed.","section":"Fig. 3(d)"},{"comment":"The sentence 'the improved sequence prevents an accumulation of pulse area errors over time' is too strong; the sequence cancels first-order pulse-area errors, but residual higher-order or amplitude-transient effects may remain. Please soften to 'suppresses' or quantify the residual sensitivity.","section":"Sec. IV"},{"comment":"The phase traces used in the simulation are re-used across all Clifford gate strings on the grounds of computational cost. This can introduce correlations between gate strings and may affect the estimated statistical uncertainty of the simulated fidelity. Please comment on the magnitude of this effect (e.g., by repeating the simulation with a subset of independent traces).","section":"Appendix C"},{"comment":"Define the notation F_1^2 more explicitly in the text: it is the average fidelity of a single Clifford gate, and the superscript 2 refers to the fidelity measure F^2(rho, sigma), not a square of a gate fidelity in the usual sense. This will avoid confusion with the abstract's notation.","section":"Sec. V"},{"comment":"The axis and sign conventions for the rotations in Table I should be stated explicitly (e.g., whether R_r(theta) is a right-handed rotation about r for positive theta).","section":"Table I"},{"comment":"The data availability statement could be strengthened by depositing the RB dataset, the noise PSDs, and the simulation code in a public repository.","section":"Data availability"}],"recommendation":"major_revision","confidential_remarks":"The paper is generally well written and the technical content is strong. The main concern is the RB modeling; if the authors provide the free-offset fit and a quantitative homogeneity test, I would be supportive of publication. The paper is within the scope of physics.atom-ph and would be of interest to the quantum information and precision metrology communities."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper delivers what it claims: a high-power 698 nm laser with low frequency noise out to a few kHz, and a convincing RB measurement of F1^2 = 0.99964(3) on ~3000 atoms. The system integration—cryogenic silicon cavity for long-term stability, ULE cavity for high-frequency noise, comb transfer, and a 400 kHz phase lock to a multi-watt laser—is new and practical. The atom-based noise spectroscopy sequence with intensity-noise rejection is the most original piece: it is validated with injected modulation and gives a clean, tunable spectral response. That alone is a useful contribution to the clock-laser and optical-qubit communities.\n\nThe core result is well supported. The noise budget is careful, the cross-correlation measurements are standard and appropriate, and the simulation of RB using the measured phase noise PSD reproduces the data across Rabi frequencies without free parameters beyond the noise model. The spatial histogram in Fig. 4(c) gives reasonable evidence of uniformity across the cloud. On the evidence in the paper, phase noise from the laser is the dominant error source, and the agreement between experiment and simulation is the strongest point in favor of that conclusion.\n\nThe soft spots are real but not fatal. The headline fidelity comes from fitting RB to Eq. (6) with the offset fixed at 1/2. The stress-test worry about bias from non-depolarizing or inhomogeneous errors is legitimate, and the paper does not show a free-offset or multi-exponential fit to bound that bias. I think the risk is minor: the beam is far larger than the atomic cloud, the histogram is narrowly peaked, and the simulation matches without invoking coherent errors. But a one-figure robustness check would settle it. The long-term instability of 3.5e-17 is inherited from the Si3 cavity and not directly measured on the final high-power beam; the 4 W output is not tested; and the record claim lacks a quantitative comparison to prior work. These are stated in the text, so not misleading, but the abstract overstates them slightly.\n\nThe paper deserves a serious referee and likely publication after minor revision. I would ask for the free-offset RB fit, a quantitative statement on the spatial fidelity spread, and a less assertive abstract. Public data and code would help independent verification, but the absence of them is not a fatal flaw for an experimental paper of this type.","headline":"A strong demonstration of a spectrally tailored high-power clock laser for optical qubit control; the headline fidelity is model-dependent, but the internal consistency between noise model, atom-based spectroscopy, and RB simulation carries the paper.","tokens_in":25913,"tokens_out":3050,"would_cite":false,"duration_ms":34875,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["42.62.Fi","06.30.Ft","32.80.Qk"],"model":"deepseek-v4-flash","headline":"A spectrally tailored 698 nm clock laser—combining a cryogenic silicon cavity, a ULE cavity, and a frequency comb—achieves an average single-qubit Clifford gate fidelity of 0.99964(3) while simultaneously driving about 3000 strontium…","keywords":["clock laser","optical lattice clock","quantum gate fidelity","randomized benchmarking","laser frequency noise","spectral tailoring","strontium-87","optical qubits"],"falsifier":"Run interleaved randomized benchmarking or gate-set tomography at the same Rabi frequency of 741 Hz and compare the extracted per-gate error; if the inferred fidelity differs from 0.99964(3) by more than the statistical uncertainty, the depolarizing model underlying Eq. (6) is not capturing the actual error process.","tokens_in":24829,"feed_emoji":"⚛️","tokens_out":7694,"duration_ms":72477,"temperature":0.7,"pith_summary":"The paper reports a high-power 698 nm clock laser whose frequency noise is deliberately shaped to the needs of quantum operations: the low-frequency stability of a cryogenic silicon cavity is combined with the high-frequency noise suppression of a room-temperature ULE cavity, with a frequency comb transferring stability between the two and a phase-locked loop boosting power to 4 W. Using this laser to drive the strontium clock transition in about 3000 atoms confined in a 3D optical lattice, the authors measure an average single-qubit Clifford gate fidelity of $F_1^2 = 0.99964(3)$ at a Rabi frequency of 741 Hz. They also develop an atom-based spectral analysis method, a pulse sequence that cancels intensity-noise sensitivity while keeping a tunable single-peak frequency response, which verifies the laser noise model on site. If correct, this is the highest single optical-qubit gate fidelity demonstrated for a large number of atoms, approaching the regime where error correction overheads become practical.","feed_headline":"Clock laser drives 3,000 atoms at 0.99964 fidelity","feed_subtitle":"Two reference cavities and an atom-based noise probe push optical single-qubit gates to a new mark.","key_machinery":"The load-bearing object is the spectrally tailored laser itself: a 698 nm external cavity diode laser phase-locked to a narrow-linewidth ULE cavity (1 MHz bandwidth servo) and phase-locked to a cryogenic silicon cavity-stabilized frequency comb with a 500 Hz transfer loop, then phase-locked to a 4 W fiber laser with a 400 kHz loop. Its noise spectrum is a weighted sum of the two references, Eq. (1), so the crossover frequency can be chosen. The supporting mechanism is the atom-based spectrum analyzer: a Ramsey-type sequence of alternating $\\pi$ rotations around $\\pm x$, which makes the phase-sensitivity function a windowed sine wave, yielding a singly peaked, tunable frequency response $|R(f)|^2$ centered at the Rabi frequency $\\Omega/(2\\pi)$ while canceling pulse-area errors from intensity noise.","core_discovery":"The central claim is that a laser's spectral noise can be engineered specifically for high-fidelity quantum state engineering, rather than optimized for general stability. By locking a 698 nm external cavity diode laser to a ULE cavity with a 1 MHz Pound-Drever-Hall loop, phase-locking that light to a cryogenic silicon cavity-stabilized comb at 500 Hz bandwidth, and transferring the phase to a high-power fiber laser, the noise power spectral density of the delivered light becomes a frequency-dependent blend of the best of both references: silicon-cavity stability below 500 Hz and ULE-cavity low noise above. The authors validate this model with an in-situ atomic spectrum analyzer that uses repeated $(x_\\pi, x_{-\\pi})$ rotations to suppress intensity noise while localizing frequency-noise sensitivity at the Rabi frequency. They then achieve $F_1^2 = 0.99964(3)$ for about 3000 atoms simultaneously, with the fidelity distribution across the atomic cloud narrowly peaked, confirming uniform high-fidelity control.","pith_inferences":["The same spectral-tailoring recipe—two cavities with complementary noise floors spliced by a comb—should transfer to other wavelengths and atomic species, so the fidelity gain is likely not tied to strontium's 698 nm transition.","The atom-based noise probe, being in-situ and intensity-noise-immune, could serve as a general diagnostic for any quantum platform whose gate errors are dominated by laser phase noise, including Rydberg and molecular systems.","A natural next test is to extend randomized benchmarking with interleaved or gate-set protocols; if coherent error components are exposed, composite pulses or pulse shaping should push the fidelity beyond the reported value.","The 500 Hz crossover between the two cavities is an operational robustness tradeoff; lowering it would improve long-term fidelity but reduce stability against environmental perturbations, so the optimal crossover may depend on the target Rabi frequency."],"forward_implications":["Optical single-qubit gates in neutral-atom systems reach a fidelity ($1-F_1^2=3.6\\times10^{-4}$) that approaches trapped-ion demonstrations, while addressing about 3000 atoms simultaneously.","The demonstrated gate fidelity and multi-second coherence time allow deep Clifford circuits on thousands of qubits, reducing the physical-qubit overhead needed for error correction.","The same high-power clock laser retains an instability of $3.5\\times10^{-17}$ at 1 to 1000 s, so the hardware serves both quantum-information and optical-clock metrology.","Atomic spectral analysis with the repeated-$\\pi$ sequence provides a general way to measure laser frequency noise at the qubit location, applicable to other quantum sensors."],"supporting_citations":[{"why":"Provides the cryogenic silicon cavity-stabilized comb noise model and the 3.5e-17 long-term instability that anchors the low-frequency performance.","marker":"[80]"},{"why":"Describes the 21 cm cryogenic silicon cavity at 124 K used as the primary long-term frequency reference.","marker":"[58]"},{"why":"Characterizes the ULE cavity whose thermal noise floor sets the high-frequency noise model.","marker":"[88]"},{"why":"The Pound-Drever-Hall technique used for locking the 698 nm laser to the ULE cavity.","marker":"[87]"},{"why":"The atomic spectrum analyzer concept that the on-site frequency noise measurement sequence extends and validates.","marker":"[91]"},{"why":"Supplies the single-qubit Clifford gate implementation and the depolarizing model, Eq. (6), from which the gate fidelity is extracted.","marker":"[100]"},{"why":"Fiber noise cancellation technique that delivers the stabilized light to the atoms without added phase noise.","marker":"[90]"},{"why":"Gives the Raman-scattering-limited excited state decay rate used in the numerical fidelity simulation.","marker":"[86]"}],"fun_headline_variants":["Tailored 4W clock laser hits 99.964% fidelity on 3,000 qubits","3,000 optical qubits at 0.99964 fidelity via spectral noise shaping","Record 0.99964 single-qubit gate fidelity on 3000 atoms","Laser noise engineering unlocks 3,000-qubit gate fidelity of 0.99964","4W 698nm laser: 0.99964 fidelity on 3,000 qubits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The fidelity result assumes the only gate errors are depolarizing; if residual coherent errors like detuning or Rabi inhomogeneity are significant, the reported $F_1^2$ is a model-dependent number rather than the true average gate fidelity.","fun_headline_variants_meta":{"raw":{"variants":["Tailored 4W clock laser hits 99.964% fidelity on 3,000 qubits","3,000 optical qubits at 0.99964 fidelity via spectral noise shaping","Record 0.99964 single-qubit gate fidelity on 3000 atoms","Laser noise engineering unlocks 3,000-qubit gate fidelity of 0.99964","4W 698nm laser: 0.99964 fidelity on 3,000 qubits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000832,"raw_usage":{"total_tokens":3643,"prompt_tokens":966,"completion_tokens":2677,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":582,"completion_tokens_details":{"reasoning_tokens":2558}},"tokens_in":582,"tokens_out":2677,"duration_ms":17956,"temperature":1.0,"reasoning_tokens":2558,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:06:29.738081+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run interleaved randomized benchmarking or gate-set tomography at the same Rabi frequency of 741 Hz and compare the extracted per-gate error; if the inferred fidelity differs from 0.99964(3) by more than the statistical uncertainty, the depolarizing model underlying Eq. (6) is not capturing the actual error process.","supporting_citations":[{"cited_title":"Lintz, D.-H","cited_arxiv_id":null,"evidence_quote":"Provides the cryogenic silicon cavity-stabilized comb noise model and the 3.5e-17 long-term instability that anchors the low-frequency performance."},{"cited_title":"de L´ es´ eleuc, D","cited_arxiv_id":null,"evidence_quote":"Describes the 21 cm cryogenic silicon cavity at 124 K used as the primary long-term frequency reference."},{"cited_title":"Sonderhouse, C","cited_arxiv_id":null,"evidence_quote":"Characterizes the ULE cavity whose thermal noise floor sets the high-frequency noise model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The Pound-Drever-Hall technique used for locking the 698 nm laser to the ULE cavity."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The atomic spectrum analyzer concept that the on-site frequency noise measurement sequence extends and validates."},{"cited_title":"Santarelli, C","cited_arxiv_id":null,"evidence_quote":"Supplies the single-qubit Clifford gate implementation and the depolarizing model, Eq. (6), from which the gate fidelity is extracted."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Fiber noise cancellation technique that delivers the stabilized light to the atoms without added phase noise."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the Raman-scattering-limited excited state decay rate used in the numerical fidelity simulation."}],"review_version":1}