{"id":"2bcc93f6-e7bb-4742-8476-f27c60fed2b1","arxiv_id":"2411.11708","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"Researchers demonstrate individually controlled single- and two-qubit gates on ytterbium-171 nuclear spin qubits, reaching 99.72% CZ fidelity with post-selection and introducing a Hessian-based calibration method.","lead":"This paper reports a universal set of high-fidelity quantum gates on arrays of trapped ytterbium-171 atoms, including a two-qubit entangling gate with 99.72% fidelity after post-selection. It also introduces a faster calibration method for multi-parameter gates, bringing neutral-atom quantum computers closer to error correction.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Per-CZ fidelity extraction in App. F appears to over-subtract 1Q error by using the per-Clifford 1Q CRB error as a per-pulse error, biasing the 99.72% headline.","rationale":"The paper is a credible experimental demonstration with multiple cross-checked benchmarks; the qualitative claim of a high-fidelity CZ gate above 99% is robust. The specific concern here is narrower than the reader's depolarizing-model worry: the per-CZ fidelity extraction in Appendix F appears to use the per-1Q-Clifford CRB error as if it were the per-native-pulse error, over-subtracting single-qubit errors and biasing the headline 99.72(3)% upward by an amount comparable to its quoted uncertainty. This is concrete and checkable from published numbers, but it does not overturn the central result; it means the stated precision is not yet fully established. The reader's CONDITIONAL verdict remains appropriate, hence UNCHANGED, with the added condition that the 1Q normalization be verified or corrected.","tokens_in":21434,"tokens_out":17307,"duration_ms":159415,"concrete_test":"Recompute the per-CZ fidelity using the correct per-pulse single-qubit error: F_CZ = 1 - [ε_2QCliff - (N_x / n_x) ε_1QCliff] / N_CZ, with reported post-selected ε_2QCliff = 0.60(3)%, ε_1QCliff = 0.037(2)%, N_x = 4.36, N_CZ = 1.51, and n_x obtained by counting Xπ/2 pulses in the Qiskit-generated 1Q Clifford circuits. If the corrected F_CZ deviates from 99.72% by more than 0.03%, the headline uncertainty is underestimated and the extraction procedure must be revised; an interleaved RB or direct process-tomography measurement of the CZ gate would provide an independent check.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Appendix F extracts the per-CZ fidelity from the two-qubit Clifford RB decay by subtracting the 'known 1Q CRB error' multiplied by the average number of 1Q gates per Clifford, 4.36 (Fig. 8). However, the quoted single-qubit CRB fidelity, 99.963(2)% (Sec. III), is the error per 1Q Clifford gate, not per native Xπ/2 pulse. Each 1Q Clifford is compiled from more than one Xπ/2 rotation (plus virtual Z rotations), so the per-pulse Xπ/2 error is smaller than the per-Clifford error by the average number of Xπ/2 pulses per 1Q Clifford, n_x > 1. The 4.36 gates counted in Fig. 8 are native Xπ/2 pulses. Thus the subtraction over-estimates the single-qubit contribution by the factor (1 - 1/n_x). For n_x in the typical range 1.25-1.5, this over-subtracts roughly 0.01-0.05% per 2Q Clifford, which divided by 1.51 CZ/Clifford biases the headline CZ fidelity 99.72(3)% upward by approximately 0.01-0.03%, comparable to or exceeding the reported 0.03% statistical error. The paper does not report n_x or the Clifford compilation counts for the 1Q RB, so this normalization cannot be verified from the text.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a universal gate set on the 171Yb ground-state nuclear-spin qubit in arrays of optical tweezers. The central results are a single-qubit Clifford randomized benchmarking fidelity of 99.963(2)% per Clifford gate and a two-qubit CZ fidelity of 99.72(3)% with post-selection and 99.40(3)% without post-selection, extracted from two-qubit Clifford RB using average native-gate counts of 1.51 CZ gates and 4.36 Xπ/2 pulses per Clifford. A symmetric-subspace CZ-GERB measurement gives 99.84(6)% (99.56(5)%) with (without) post-selection. The paper also introduces an eigenvector-based multi-parameter calibration of the Rydberg gate and provides a detailed error budget with separate accounting of leakage and loss.","tokens_in":21784,"tokens_out":14705,"duration_ms":151015,"significance":"If correct, the results are significant: the ground-state nuclear-spin qubit combines long coherence and insensitivity to trap light shifts, and the reported CZ fidelities exceed commonly cited fault-tolerance thresholds for the surface code. The paper is unusually transparent on several points: it reports pre-selected and post-selected values side by side, gives separate leakage and loss measurements, and explicitly acknowledges where the simulated error budget does not reproduce measured infidelities. The Hessian-eigenvector calibration strategy is a useful methodological contribution that goes beyond the usual brute-force multi-parameter scans. The manuscript does not ship machine-checked proofs or an independent code repository, but it does cite the Qiskit Experiments package used for circuit generation.","major_comments":[{"comment":"The extraction of the per-CZ error from the 2Q CRB decay subtracts the 1Q CRB error multiplied by 4.36, where 4.36 is the average number of native Xπ/2 pulses per 2Q Clifford. The 1Q CRB error quoted in Sec. III is the error per 1Q Clifford gate, not per native Xπ/2 pulse. If the average number of Xπ/2 pulses per 1Q Clifford is n_x > 1, this subtraction overestimates the single-qubit contribution by a factor 1 - 1/n_x, biasing the headline CZ fidelity upward. For n_x in the range 1.25-1.5, the bias is roughly 0.01-0.03 percentage points, which is comparable to or larger than the quoted 0.03% statistical uncertainty. The manuscript does not report n_x or the decomposition counts for the 1Q Clifford circuits, and the consistency statement in Appendix B (0.32(2)% per eight Xπ/2 pulses versus eight times the 1Q CRB error) appears to assume n_x = 1. Please report n_x, justify that the per-Clifford error equals the per-pulse error for the compilation actually used, or correct the extraction accordingly.","section":"Appendix F, Sec. IV.C, Fig. 8"},{"comment":"The per-CZ fidelities are obtained under a depolarizing, gate-independent error model, and only statistical uncertainties are quoted. The systematic uncertainty from model dependence is not quantified, although the paper itself provides a useful cross-check: the CZ-GERB estimate and the 2Q CRB estimate differ by more than the statistical errors. The authors attribute much of this difference to single-qubit phase errors, but a quantitative decomposition is not given. Since the headline 99.72(3)% includes a third decimal place, the authors should state the model-dependent systematic uncertainty explicitly, for example by treating the spread between CRB- and GERB-based estimates as a partial bound and by reporting the sensitivity of the extracted CZ error to plausible non-depolarizing noise components.","section":"Sec. IV.C and Appendix G"}],"minor_comments":[{"comment":"The sentence 'an additional readout step preceded by an clock repumping state' contains a grammar error; it should read 'preceded by a clock repumping step'.","section":"Appendix B"},{"comment":"The figure shows simulated optimization trajectories, but the caption does not explicitly state that the green and orange curves are simulation results rather than experimental data; this should be clarified.","section":"Figure 3(c)"},{"comment":"The conversion 1 - F = (1 - p)(1 - b) is stated without derivation; a brief derivation or a reference would help readers understand how the fixed baseline b enters the reported infidelities.","section":"Appendix F, Eq. (F1)"},{"comment":"The paper does not report the number of random circuits and repetitions for the CZ-GERB depth scan in the same detail as for the 2Q CRB scan; Appendix F mentions 10 circuits and about 20 repetitions, but the main text could state this more explicitly.","section":"Section IV.C"}],"recommendation":"major_revision","confidential_remarks":"The experimental work appears substantial and the central claim is likely correct in spirit, but the normalization issue in Appendix F directly affects a headline number and must be resolved. If the authors can show that their 1Q Clifford compilation has n_x = 1, or if they correct the extracted value and its uncertainty accordingly, the paper should be acceptable. I see no grounds for rejection, only for the requested quantitative clarification."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this is a real experimental milestone, not a hype paper. They demonstrate a universal gate set on 171Yb ground-state nuclear spin qubits with CZ fidelity around 99.7% post-selected, and they introduce a Hessian-eigenvector calibration method that is genuinely useful for multi-parameter gates. The benchmarking is thorough—single- and two-qubit Clifford RB, U-GERB, clock-GERB—and they report post-selected and un-post-selected numbers, leakage, and loss transparently.\n\nThe soft spot is the per-CZ fidelity extraction. In Appendix F they subtract the single-qubit error by multiplying the 'known 1Q CRB error' by the average number of Xπ/2 pulses per two-qubit Clifford (4.36). But the quoted 1Q CRB error is per Clifford gate, not per Xπ/2 pulse. Each 1Q Clifford is compiled from more than one Xπ/2 rotation, so this over-subtracts the 1Q contribution by a factor n_x (likely 1.25–1.5). That biases the extracted CZ fidelity upward by roughly 0.01–0.03%, which is comparable to the reported 0.03% statistical error. The paper does not report n_x or the 1Q Clifford compilation counts, so the normalization cannot be checked from the text. This is a small but real flaw in the headline number. The GERB measurements use the same subtraction, so they inherit the same bias; the qualitative conclusion—high-fidelity gates above typical thresholds—still holds.\n\nOther weaknesses are minor: the depolarizing model assumption is standard for RB and partially mitigated by the GERB cross-check, and the error budget does not fully reproduce the measured numbers (they acknowledge this and list uncharacterized effects). No raw data or analysis code are provided, which limits independent verification, but that is common for industry papers.\n\nThe paper is worth a serious referee. The calibration method alone is a useful contribution, and the experimental work is careful. I would ask the authors to clarify the 1Q gate-count normalization and re-extract the fidelities with the correct per-pulse error, and to report the average number of Xπ/2 gates per 1Q Clifford. After that, the quantitative claims should be solid. I'd bring it to reading group and would cite it for the Hessian calibration and the state of the art for Yb ground-state qubits.","headline":"Solid experimental milestone with a useful calibration tool, but the headline CZ fidelity may be biased upward by a small normalization error in the 1Q-error subtraction.","tokens_in":22572,"tokens_out":3698,"would_cite":true,"duration_ms":32689,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Lx"],"model":"deepseek-v4-flash","headline":"The paper reports a universal gate set on the 171Yb ground-state nuclear spin qubit, with a two-qubit CZ fidelity of 99.72(3)% (post-selected) measured by two-qubit Clifford randomized benchmarking.","keywords":["neutral atom quantum computing","171Yb nuclear spin qubit","Rydberg blockade","two-qubit CZ gate","randomized benchmarking","clock shelving","shaped composite pulses","optical tweezer arrays"],"falsifier":"Run a noise-characterization experiment that does not assume depolarizing errors, such as gate-set tomography of the CZ gate, or interleaved randomized benchmarking in which the random Clifford circuits are recompiled to use different ratios of CZ to single-qubit gates at fixed Clifford depth. If the inferred per-CZ fidelity changes with the circuit composition or shows a coherent component, the depolarizing assumption behind the headline number is wrong. A less expensive check already sits in the paper's data: the simulated error budget predicts 0.375% infidelity per CZ gate while the pre-selected Clifford RB measurement gives 0.60(3)%, so an experiment that locates the missing roughly 0.2% (for example, by testing whether pair loss during the gate exceeds the single-atom Rydberg lifetime prediction) would either close the budget or invalidate the error model.","tokens_in":21278,"feed_emoji":"⚛️","tokens_out":15762,"duration_ms":130182,"temperature":0.7,"pith_summary":"This paper reports a universal gate set for the ground-state nuclear spin qubit of tweezer-trapped $^{171}$Yb atoms, with a two-qubit CZ gate fidelity of 99.72(3)% with post-selection and 99.40(3)% without, extracted from two-qubit Clifford randomized benchmarking, and a single-qubit fidelity of 99.963(2)%. The result matters because these fidelities sit above the commonly cited 99% surface-code threshold, so neutral-atom arrays of this type can aim at error-corrected circuits rather than at further fundamental gate improvement. The CZ gate works by shelving qubit state $|1\\rangle$ into a long-lived clock state with shaped composite pulses, applying a global Rydberg pulse whose blockade mechanism imparts the entangling phase, and then unshelving; gates are individually addressed and can run in parallel on two pairs of atoms. The paper also contributes a calibration recipe that diagonalizes the simulated error landscape, so multi-parameter gates can be tuned with a few decoupled one-dimensional scans, and it closes with a component-level error budget. Together these pieces make the case that high-fidelity, flexible, and scalable neutral-atom computation is within reach.","feed_headline":"Two-qubit gate hits 99.72 percent in a neutral atom array","feed_subtitle":"Ytterbium nuclear-spin qubits pass the ~99 percent threshold often cited for error correction.","key_machinery":"The load-bearing mechanism is the sequential state-selective CZ gate built from three steps: a Blackman-shaped composite clock pulse $Y_{\\pi/2}-X_{\\pi}-Y_{\\pi/2}$ shelves $|1\\rangle$ into the long-lived clock state $|c\\rangle$; a global UV pulse drives the clock-to-Rydberg transition with a sinusoidal phase profile $\\phi(t) = A\\cos(\\omega t - \\phi) + \\Delta t$ that approximates a time-optimal gate, so that $|01\\rangle$ and $|10\\rangle$ return to the clock manifold while $|11\\rangle$ acquires the entangling phase $\\pi$ through Rydberg blockade; and a second composite pulse unshelves. The gate's five control parameters are calibrated by a new method: computing the Hessian of the gate infidelity from simulation, diagonalizing it, and scanning once along each nearly decoupled eigenvector, which reaches the optimum in a fixed number of one-dimensional scans. Fidelity extraction relies on two-qubit Clifford randomized benchmarking with individual addressing, which samples the full two-qubit Hilbert space; the per-CZ fidelity is obtained by subtracting the known single-qubit error using the average native-gate counts of 1.51 CZ and 4.36 $X_{\\pi/2}$ gates per Clifford under a depolarizing error model.","core_discovery":"The paper's central claim is that the $^{171}$Yb ground-state nuclear spin qubit now supports a universal, individually controlled gate set with fidelity high enough for fault-tolerant designs: a CZ gate measured at 99.72(3)% with post-selection and 99.40(3)% without, extracted from two-qubit Clifford randomized benchmarking circuits that run up to roughly 150 Clifford gates, or more than 200 CZ gates, on one or two atom pairs simultaneously, together with a single-qubit Clifford RB fidelity of 99.963(2)%. The entangling gate uses a sequential excitation scheme: a shaped composite clock pulse shelves $|1\\rangle$ into the metastable clock state $|c\\rangle$, a global 302 nm pulse couples the clock state to a Rydberg state with a phase profile chosen so that the $|11\\rangle$ component picks up a $\\pi$ entangling phase through Rydberg blockade, and a second composite pulse unshelves. A symmetric-subspace benchmark (CZ-GERB) that echoes away single-qubit phases measures 99.84(6)% with post-selection, and the paper attributes most of the gap between the two metrics to quasi-static clock-laser detuning, which appears as single-qubit phase error. The authors state that combining these gates with continuous loading, mid-circuit measurement, and erasure conversion is expected to enable complex error-corrected circuits.","pith_inferences":["A testable consequence of the paper's own numbers: the simulated error budget predicts 0.375% infidelity per CZ gate while pre-selected Clifford RB measures 0.60(3)%, so roughly 0.2% of the error is unaccounted for, and locating it (the authors suspect Rydberg pair-state dynamics and Doppler-sensitive motion) is the most direct route to sub-0.1% two-qubit gates.","Because the gap between the two fidelity metrics is attributed to quasi-static clock-laser detuning that appears as single-qubit phase error, active frequency stabilization or echo-based phase correction could plausibly push the practical two-qubit fidelity closer to the 99.84(6)% GERB number without any hardware change.","Since leakage into the clock state and atom loss are each measured per gate (about 0.12-0.14% per CZ), the platform is already set up for erasure-biased error correction; a natural next experiment is a small erasure-checking code whose logical error rate tracks the detected-leakage budget, using the mid-circuit measurement and state-selective readout the paper cites as existing."],"forward_implications":["With post-selected CZ fidelity of 99.72(3)% and single-qubit fidelity of 99.963(2)%, the demonstrated gates clear the commonly cited 99% surface-code threshold, so the remaining work toward error correction on this platform is integration (mid-circuit measurement, erasure conversion, rearrangement) rather than raising gate quality.","Because ground-state nuclear spin qubits have near-infinite lifetime and low sensitivity to trap light shifts, the same gate set should transfer to large rearranged arrays with flexible connectivity, which the paper argues enables efficient error-correction encodings.","The dominant identified error sources, clock-laser detuning drift and phase noise, produce mostly detectable leakage and loss rather than silent in-subspace errors, so state-selective readout can convert the main physical errors into erasure events that error correction handles more cheaply.","The Hessian-eigenvector calibration method reduces multi-parameter gate optimization to a fixed number of decoupled one-dimensional scans, a recipe the paper argues applies to any entangling gate with several interdependent control parameters."],"supporting_citations":[{"why":"Supplies the sinusoidal phase parametrization of the Rydberg pulse and the U-X-pi-U echoed benchmarking (GERB) structure used for clock and CZ characterization.","marker":"[13]"},{"why":"Establishes the Rydberg-blockade CZ approach this gate builds on, with parallel high-fidelity multiqubit gates in neutral atoms.","marker":"[12]"},{"why":"Demonstrates sequential resonant excitation through a long-lived intermediate state, the scheme adopted here for clock-to-Rydberg excitation.","marker":"[14]"},{"why":"Provides the individually addressed Raman beam scheme for single-qubit control of 171Yb nuclear spin qubits used as the one-qubit gate layer.","marker":"[10]"},{"why":"Gives the time-optimal Rydberg gate pulse family that the sinusoidal phase profile approximates.","marker":"[48]"},{"why":"Defines the two-qubit Clifford randomized benchmarking protocol whose decay curves yield the headline CZ fidelity.","marker":"[30]"},{"why":"Provides the randomized benchmarking fidelity-extraction formalism used to convert decay fits into per-gate fidelities.","marker":"[31]"},{"why":"Supplies the spin-locking noise spectroscopy used to bound clock laser frequency noise and informs the error budget.","marker":"[23]"}],"fun_headline_variants":["Ytterbium qubit gate set hits 99.72% fidelity","Neutral atom array achieves 99.72% CZ gate fidelity","Universal ytterbium nuclear spin gates reach 99.7%","99.72% two-qubit gate in ytterbium atom array","Yb nuclear spin qubits enable high-fidelity universal gates"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The headline 99.72% CZ number assumes that the noise in the benchmarking circuits is depolarizing and gate-independent, so that the per-gate error can be recovered by multiplying average native-gate counts; if the real errors are coherent, biased toward some states, or correlated between gates, the extracted fidelity could be off.","fun_headline_variants_meta":{"raw":{"variants":["Ytterbium qubit gate set hits 99.72% fidelity","Neutral atom array achieves 99.72% CZ gate fidelity","Universal ytterbium nuclear spin gates reach 99.7%","99.72% two-qubit gate in ytterbium atom array","Yb nuclear spin qubits enable high-fidelity universal gates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000422,"raw_usage":{"total_tokens":2214,"prompt_tokens":1034,"completion_tokens":1180,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":650,"completion_tokens_details":{"reasoning_tokens":1084}},"tokens_in":650,"tokens_out":1180,"duration_ms":8472,"temperature":1.0,"reasoning_tokens":1084,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T18:13:14.935809+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a noise-characterization experiment that does not assume depolarizing errors, such as gate-set tomography of the CZ gate, or interleaved randomized benchmarking in which the random Clifford circuits are recompiled to use different ratios of CZ to single-qubit gates at fixed Clifford depth. If the inferred per-CZ fidelity changes with the circuit composition or shows a coherent component, the depolarizing assumption behind the headline number is wrong. A less expensive check already sits in the paper's data: the simulated error budget predicts 0.375% infidelity per CZ gate while the pre-selected Clifford RB measurement gives 0.60(3)%, so an experiment that locates the missing roughly 0.2% (for example, by testing whether pair loss during the gate exceeds the single-atom Rydberg lifetime prediction) would either close the budget or invalidate the error model.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates sequential resonant excitation through a long-lived intermediate state, the scheme adopted here for clock-to-Rydberg excitation."},{"cited_title":"Barnes, P","cited_arxiv_id":null,"evidence_quote":"Provides the individually addressed Raman beam scheme for single-qubit control of 171Yb nuclear spin qubits used as the one-qubit gate layer."},{"cited_title":"Jandura and G","cited_arxiv_id":null,"evidence_quote":"Gives the time-optimal Rydberg gate pulse family that the sinusoidal phase profile approximates."}],"review_version":1}