{"id":"3a760a03-4ea0-41f4-b777-f26e2a6a5141","arxiv_id":"2506.10714","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A universal gate set with erasure conversion is demonstrated on the metastable fine-structure qubit in strontium-88, along with a state-resolved detection scheme.","lead":"This experiment demonstrates universal quantum gates on a qubit made from two metastable states of strontium-88, with single-qubit fidelity 0.993 and two-qubit fidelity 0.9945 after removing loss events. It also shows how to detect qubit loss mid-circuit, a step toward error-corrected quantum computers.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline two-qubit fidelity is obtained by excising loss events; correlation between loss and gate errors would bias it upward, and the model-vs-measurement gap (0.16% vs 0.55% loss-corrected infidelity) shows the correction is not fully understood.","rationale":"The reader's verdict of CONDITIONAL is appropriate, and the reader's weakest_assumption identifies the same load-bearing concern I found: the two-qubit fidelity after loss excision assumes loss and computational errors are independent. This is the most central issue because the headline quantum-information claim (F2q = 0.9945(6)) is the loss-corrected number, and the raw fidelity is substantially lower (0.9759(5)). The paper is honest about reporting both raw and corrected values, and the single-qubit gates, erasure conversion, and detection scheme are independently demonstrated. However, the loss-correction step is not validated quantitatively: the SI error budget predicts a loss-corrected infidelity of 0.16%, while the measured value is 0.55%, a factor of 3.4 discrepancy attributed to unspecified drifts. This gap is precisely where a correlated loss-error channel could hide, and it undermines the claim that the loss-corrected fidelity represents the gate quality once losses are eliminated. The concrete test of re-analyzing the SSB data with a three-outcome model would settle whether the excision is biased. I do not find a more severe or more fundamental objection; the universality construction (single-qubit rotations plus CZ) is standard, and the benchmarking protocols are appropriate. The verdict should remain CONDITIONAL, with the condition being that the loss-correction independence assumption and the error-budget discrepancy be addressed before the loss-corrected fidelity is taken as the platform's representative gate quality.","tokens_in":19795,"tokens_out":9495,"duration_ms":109750,"concrete_test":"Re-analyze the raw SSB data without discarding loss trials. Treat atom loss as a third measurement outcome and fit all trials (|0>, |1>, lost) with a model that includes both a depolarizing gate error and an independently characterized loss probability, using maximum likelihood to extract the gate fidelity conditional on survival. Compare this estimate with the loss-excised value of 0.9945(6). If the two differ by more than the quoted statistical uncertainty, the excision is biased. In the same fit, check whether the measured loss-corrected infidelity of 0.55% can be reproduced from the modeled non-loss error sources (which sum to 0.16%) plus plausible parameter drifts; if not, the error budget is incomplete.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim, F2q = 0.9945(6), is derived by discarding experimental trials where atom loss is detected via the state-resolved detection scheme. This post-selection assumes loss events are independent of the computational error processes. In this system the dominant loss channel is photoionization of 3P2 by the 316 nm gate laser, but the same laser also drives off-resonant scattering that populates non-qubit 3P2 sublevels, and Rydberg decay can produce both a computational error and an untrapped dark state. If a trial is excised because an atom was lost, any concurrent computational error that caused or accompanied that loss is also removed, so the survival-conditioned fidelity is biased upward. The SI error budget (Fig. S12) predicts a loss-corrected infidelity of 0.16%, while the measured value is 0.55% (1 - 0.9945), a factor of 3.4 discrepancy that the paper attributes to unspecified drifts. This unexplained gap means the correction procedure is not quantitatively validated, so the quoted loss-corrected fidelity cannot be taken as a representative gate quality in a hypothetical loss-free implementation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports universal single- and two-qubit gates for a metastable fine-structure qubit encoded in the 3P0 and 3P2,mJ=0 states of 88Sr. Single-qubit Clifford randomized benchmarking gives F=0.993(1) after erasure-excision; two-qubit CZ gates benchmarked with symmetric stabilizer benchmarking give F=0.9945(6) after excising trials with atom loss, with a raw fidelity of 0.9759(5). The paper also demonstrates mid-circuit erasure conversion of leakage to 1S0, a state-resolved detection scheme with detection fidelity above 0.993, and a Bell-state fidelity of 0.983(8) conditioned on survival. The SI presents an error budget predicting raw and loss-corrected CZ infidelities of 1.84% and 0.16%, respectively.","tokens_in":1538,"tokens_out":1641,"duration_ms":81296,"significance":"If the headline two-qubit fidelity is taken at face value, this is a valuable experimental contribution to the neutral-atom quantum computing toolkit: it combines a long-lived metastable qubit with a universal gate set, mid-circuit erasure conversion, and high-fidelity state-resolved detection. The measurements are direct, use standard CRB and SSB protocols with statistical error bars, and the authors are transparent about raw versus loss-corrected numbers. The central quantitative significance of the two-qubit result, however, rests on an unvalidated postselection correction, as detailed below; the same caveat applies to the Bell-state fidelity.","major_comments":[{"comment":"The headline two-qubit fidelity F2q=0.9945(6) is obtained by excising trials in which atom loss is detected, with the dominant loss being 3P2 ionization by the 316 nm gate laser. The validity of this postselection requires that loss events are independent of the computational error processes acting on the surviving qubits. This independence is not established. In this system the same 316 nm light drives off-resonant scattering into non-qubit 3P2 sublevels, and Rydberg decay can concurrently produce both a computational error and a dark-state loss, so excising a lost trial can also remove a simultaneous computational error and bias the survival-conditioned fidelity upward. The error budget in Fig. S12 predicts a loss-corrected infidelity of 0.16%, whereas the measured loss-corrected infidelity is 0.55% (1-0.9945), a factor-of-3.4 discrepancy attributed to unspecified drifts. This unexplained gap means the loss-correction model is not quantitatively validated, so the quoted loss-corrected fidelity is not a demonstrated property of a hypothetical loss-free implementation. I recommend either reporting the raw fidelity as the primary two-qubit result and presenting the loss-corrected value as a model-dependent projection, or adding a direct test of loss-error independence, for example by varying the ionization rate (Rydberg principal quantum number or UV power) and verifying that the loss-corrected fidelity remains constant, or by correlating loss events with simultaneous computational errors in the same trials.","section":"Main text, Fig. 3d; SI Error budget, Fig. S12"},{"comment":"The error budget predicts a raw CZ infidelity of 1.84%, in reasonable agreement with the measured raw infidelity of 2.41%, but the loss-corrected prediction of 0.16% disagrees with the measured 0.55% by a factor of 3.4. The manuscript attributes this residual to experimental drifts in Rabi frequency, UV pulse area, phase, and shape, but these drifts are not included in the model or quantified. Without a decomposition of this residual, the claim that the loss-corrected fidelity represents the true conditional gate quality is not quantitatively supported. Please either include a drift model with estimated parameters or explicitly state that the loss-corrected number is an upper-bound estimate.","section":"SI Error budget, Fig. S12"},{"comment":"The Bell-state fidelity of 0.983(8) is obtained with the same loss-excision procedure used for the CZ gate; the raw value is 0.9355(9). The manuscript should state clearly in the main text that this is a postselected fidelity, and should explain why the loss-independence assumption is more plausible here than in the gate-benchmarking sequence, or present the raw value as the headline Bell fidelity.","section":"Main text, Fig. 3b"}],"minor_comments":[{"comment":"The caption contains a typo: 'baesed' should be 'based'.","section":"Fig. S1 caption"},{"comment":"The table lists identical infidelity contributions of 0.076% for both 3P0 and 3P2 repumping; please clarify whether these are independent error sources and how they are combined in the total detection infidelity.","section":"Fig. 4d"},{"comment":"The phrase 'we propose and experimentally demonstrate an erasure-convertible qubit' is stronger than needed, since erasure conversion in metastable alkaline-earth qubits has been proposed and demonstrated in related species; please rephrase to credit prior work explicitly.","section":"Main text, Introduction"},{"comment":"The fitted proportionality constant A=610/µs would benefit from an uncertainty and a definition of the functional form; currently the reader cannot tell whether the quoted value includes systematic uncertainties in the UV intensity calibration.","section":"Fig. S11"}],"recommendation":"major_revision","confidential_remarks":"The paper is a serious experimental demonstration with transparent reporting of raw data, and the state-resolved detection and erasure-conversion results are valuable. My main concern is that the headline two-qubit fidelity is postselected on atom survival and the loss-correction model is contradicted by the measured loss-corrected infidelity by a factor of 3.4. This is fixable by re-framing the headline result or by adding an explicit validation of loss-error independence, so I recommend major revision rather than rejection. I would not require perfection, but the central quantitative claim should be either re-scaled to the raw fidelity or backed by a quantitative drift model."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this paper demonstrates something genuinely new—the first universal gate set on the strontium-88 fine-structure qubit—and it does so with unusually honest reporting. Prior work had shown coherent coupling and millisecond coherence (refs 45-47), but no one had benchmarked a two-qubit gate or demonstrated erasure conversion on this encoding. The single-qubit CRB fidelity of 0.993(1) is standard and solid. The state-resolved detection scheme (detection fidelity >0.993) and the mid-circuit erasure conversion using the 1S0 ground state are real technical achievements, well described in the SI.\n\nThe soft spot is the headline two-qubit number. The raw CZ fidelity is 0.9759(5); the quoted 0.9945(6) is obtained by excising trials where an atom is lost, mostly photoionization of 3P2 by the 316 nm gate laser. That is a large carve-out, and it is not benign. If loss events are correlated with computational errors—for example, via Rydberg decay to an untrapped dark state—then post-selecting on survival removes both, biasing the corrected fidelity upward. The paper does not quantify this correlation. More concretely, the SI error budget predicts a loss-corrected infidelity of 0.16%, while the measured value is 0.55%—a factor-of-three gap attributed to unspecified drifts in Rabi frequency, pulse area, and phase. That gap means the loss-corrected fidelity is not a validated prediction for a loss-free implementation; it is an optimistic estimate. The paper is transparent about raw vs corrected, which is why this is a revision issue rather than a fatal flaw.\n\nThe citation pattern is appropriate: they credit refs 45-47 for the earlier coupling and coherence work, and they don't overclaim novelty. The error budget is carefully constructed, and the data on Rydberg lifetime and branching are a useful contribution in their own right.\n\nWho this is for: anyone working on neutral-atom QEC, erasure qubits, or alkaline-earth atom arrays. I would bring it to reading group and I would cite it, but I would quote the raw number alongside the corrected one. Recommendation: send it to peer review—it deserves serious referee time. The main asks should be a quantitative treatment of loss-error correlation and a breakdown of the drift contribution to the loss-corrected infidelity. Neither is desk-reject material.","headline":"First universal gate set on Sr-88 fine-structure qubit; strong paper, but the loss-corrected two-qubit fidelity is an optimistic estimate until the error-budget gap (0.16% vs 0.55%) is explained.","tokens_in":20604,"tokens_out":3394,"would_cite":true,"duration_ms":38052,"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 universal gate set is demonstrated for the metastable fine-structure qubit in strontium-88, with single-qubit fidelity 0.993(1) and loss-corrected two-qubit fidelity 0.9945(6).","keywords":["strontium-88","fine-structure qubit","erasure conversion","Rydberg blockade","neutral atom arrays","quantum gates","state-resolved detection","randomized benchmarking"],"falsifier":"Run the two-qubit gate with a Rydberg-coupling scheme that does not ionize $^3\\text{P}_2$ and measure the CZ fidelity; if the loss-excision assumption is correct, the remaining infidelity should be close to the modeled 0.16%, not the measured 0.55%. Alternatively, measure the loss-excised fidelity while varying the UV laser intensity: if loss is independent of coherent errors, the fidelity of surviving atom pairs will not change as the ionization rate changes.","tokens_in":19600,"feed_emoji":"⚛️","tokens_out":7474,"duration_ms":73922,"temperature":0.7,"pith_summary":"This paper reports a complete set of high-fidelity quantum gates for a new kind of atomic qubit: the metastable fine-structure qubit in bosonic strontium-88, encoded between the $^3\\text{P}_0$ and $^3\\text{P}_2$ states. It shows single-qubit Clifford gates with fidelity 0.993(1) and a Rydberg-blockade controlled-Z gate with fidelity 0.9945(6) after correcting for atom-loss events during the gate. It also demonstrates mid-circuit erasure conversion, using fast imaging of the ground state to detect and discard shots with leakage errors, and a state-resolved detection scheme that tells the two qubit states apart with better than 0.993 fidelity. If these results hold, this qubit could support quantum error correction with erasure conversion while using less optical power and faster gates than the strontium clock qubit.","feed_headline":"Metastable strontium qubit hits 0.9945 two-qubit fidelity","feed_subtitle":"Erasure conversion and state-resolved detection put the 17 THz fine-structure qubit on a path to error correction.","key_machinery":"The load-bearing mechanisms are: (1) the qubit encoding in the metastable states $|0\\rangle={}^3\\text{P}_2,m_J=0$ and $|1\\rangle={}^3\\text{P}_0$, separated by 17 THz and held in a triple-magic optical trap; (2) two-photon Raman coupling via the $^3\\text{S}_1$ state for single-qubit rotations; (3) Rydberg blockade using the $47s\\,{}^3\\text{S}_1$ state, where one atom's Rydberg excitation shifts a neighbor's excitation energy and prevents double excitation, providing the two-qubit entangling interaction; (4) erasure conversion via fast imaging of the $^1\\text{S}_0$ ground state, which lies outside the qubit subspace; and (5) a state-resolved detection scheme that uses 496 nm repumping from $^3\\text{P}_2$ to $^3\\text{D}_2$, shelving into $^1\\text{S}_0$ for imaging, followed by slow imaging of the remaining $^3\\text{P}_0$ population. These mechanisms together turn leakage errors into detectable erasures and provide loss-resolved readout for both qubit states.","core_discovery":"The paper establishes that the 17 THz metastable fine-structure qubit in $^{88}\\text{Sr}$—with $|0\\rangle = {}^3\\text{P}_2(m_J=0)$ and $|1\\rangle = {}^3\\text{P}_0$—supports a universal gate set. Coherent single-qubit control is achieved by a two-photon Raman process through the $^3\\text{S}_1$ intermediate state, and a two-qubit controlled-Z gate is realized by Rydberg blockade: a tightly focused 316 nm laser couples $|1\\rangle$ to the Rydberg state $|47s\\,{}^3\\text{S}_1, m_J=-1\\rangle$, with an interaction shift of $2\\pi\\times 114$ MHz. Benchmarking with Clifford randomized benchmarking and symmetric stabilizer benchmarking gives single-qubit gate fidelity 0.993(1) and loss-corrected CZ fidelity 0.9945(6); Bell-state fidelity is 0.983(8) with loss excision. The paper further demonstrates mid-circuit erasure conversion of $^1\\text{S}_0$ leakage by fast 461 nm imaging, raising state-preparation fidelity to about 0.996, and a state-resolved detection scheme that reaches detection fidelity above 0.993.","pith_inferences":["If the loss-excision independence assumption holds, eliminating $^3\\text{P}_2$ ionization (for example, by coupling to the Rydberg state through a different UV scheme) should bring the two-qubit infidelity from the measured 0.55% closer to the modeled 0.16%, making the fine-structure qubit competitive with the best Rydberg gates.","The same erasure-conversion machinery should extend to fermionic strontium-87 nuclear-spin qubits or to other metastable encodings, since the $^1\\text{S}_0$ ground state remains outside the qubit subspace and is directly imageable.","The state-resolved detection scheme could serve as a loss-resolved readout for Rydberg quantum simulators, distinguishing decay into $^3\\text{P}_2$ versus $^3\\text{P}_0$ and reducing post-selection bias in entangled-state characterizations.","A direct test of the independence assumption is to measure the loss-excised CZ fidelity as a function of UV laser intensity: if loss is independent of coherent errors, the conditional fidelity of surviving atom pairs should remain flat as the ionization rate changes."],"forward_implications":["The fine-structure qubit performs single-qubit Clifford gates at 0.993(1) fidelity and CZ gates at 0.9945(6) fidelity (loss-corrected), making it competitive with leading neutral-atom qubit platforms.","Mid-circuit erasure conversion removes state-preparation errors and off-resonant scattering errors, raising state-preparation fidelity to about 0.996 without measurably degrading qubit coherence.","The state-resolved detection scheme identifies both qubit states and atom loss with fidelity above 0.993, providing a path to monitor atom loss during error-correction cycles.","Because the qubit uses a 17 THz splitting and larger intermediate-state detunings, gate times could be pushed into the sub-microsecond regime with higher power, an order of magnitude faster than the clock qubit with lower optical power requirements.","The qubit is compatible with coherent transport in reconfigurable tweezer arrays, opening the way to erasure-converted error correction in larger arrays."],"supporting_citations":[{"why":"Supplies the erasure-convertible qubit scheme in ytterbium-171 that this work adapts to strontium-88.","marker":"[7]"},{"why":"Provides the symmetric stabilizer benchmarking sequence used to extract the two-qubit gate fidelity.","marker":"[13]"},{"why":"Supplies the SSB benchmarking protocol and fidelity response theory used for gate characterization and phase-noise estimation.","marker":"[14]"},{"why":"Demonstrates erasure conversion in a strontium Rydberg setup and contributes the Rydberg decay and branching-ratio methodology.","marker":"[32]"},{"why":"Establishes high-fidelity Rydberg excitation and detection in alkaline-earth atoms, the basis for the Rydberg blockade gate and autoionization loss detection.","marker":"[44]"},{"why":"Realizes the triple-magic trapping configuration for the fine-structure qubit in strontium-88, providing the trap depths and coherence used here.","marker":"[47]"},{"why":"Provides the fast single-atom imaging technique that underlies the erasure imaging protocol.","marker":"[50]"},{"why":"Expands fast single-atom imaging for optical lattice arrays, the version of the protocol used in this work.","marker":"[51]"},{"why":"Supplies the time-optimal phase-modulated CZ gate construction used to implement the entangling gate.","marker":"[55]"}],"fun_headline_variants":["Strontium-88 metastable qubit achieves universal gates","Erasure-ready strontium qubit hits 0.9945 gate fidelity","Universal gate set on strontium-88 metastable qubit","Metastable strontium qubit: universal gates and erasure","Strontium-88 qubit: erasure conversion and universal gates"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The headline two-qubit fidelity assumes that the atoms lost during the gate—mostly because the 316 nm laser ionizes the $^3\\text{P}_2$ state—are lost independently of the actual gate errors, so that throwing away those runs does not hide real errors.","fun_headline_variants_meta":{"raw":{"variants":["Strontium-88 metastable qubit achieves universal gates","Erasure-ready strontium qubit hits 0.9945 gate fidelity","Universal gate set on strontium-88 metastable qubit","Metastable strontium qubit: universal gates and erasure","Strontium-88 qubit: erasure conversion and universal gates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000708,"raw_usage":{"total_tokens":3221,"prompt_tokens":1011,"completion_tokens":2210,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":627,"completion_tokens_details":{"reasoning_tokens":2114}},"tokens_in":627,"tokens_out":2210,"duration_ms":20159,"temperature":1.0,"reasoning_tokens":2114,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:19:56.350333+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the two-qubit gate with a Rydberg-coupling scheme that does not ionize $^3\\text{P}_2$ and measure the CZ fidelity; if the loss-excision assumption is correct, the remaining infidelity should be close to the modeled 0.16%, not the measured 0.55%. Alternatively, measure the loss-excised fidelity while varying the UV laser intensity: if loss is independent of coherent errors, the fidelity of surviving atom pairs will not change as the ionization rate changes.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the SSB benchmarking protocol and fidelity response theory used for gate characterization and phase-noise estimation."},{"cited_title":"Scholl, A","cited_arxiv_id":null,"evidence_quote":"Demonstrates erasure conversion in a strontium Rydberg setup and contributes the Rydberg decay and branching-ratio methodology."},{"cited_title":"Bergschneider, V","cited_arxiv_id":null,"evidence_quote":"Provides the fast single-atom imaging technique that underlies the erasure imaging protocol."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Expands fast single-atom imaging for optical lattice arrays, the version of the protocol used in this work."}],"review_version":1}