{"id":"368f9b2e-5dba-4052-9be6-0f13882861a1","arxiv_id":"2507.13320","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A decoherence-free-subspace logical qubit stored in two 171Yb+ ions in a cryogenic trap achieves a fitted coherence time above two hours after post-selecting out leakage events.","lead":"This paper reports a multi-ion quantum memory in a cryogenic trap that stores a logical qubit encoded in two entangled 171Yb+ ions, with a fitted coherence time above two hours after discarding detectable leakage events. The result extends previous single-ion hour-scale memories to a multi-ion, logically encoded memory, a step toward scalable quantum repeaters and networks.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The two-hour coherence claim rests on treating leakage detection as a clean erasure, but the 93.4% |1F> detection fidelity and the unresolved |F=4,mF=±1> leakage channel could bias the post-selected decay; no confusion-matrix calibration is shown.","rationale":"The paper aims to demonstrate a multi-ion DFS logical qubit memory with post-selected coherence time above two hours. This requires (1) that the DFS encoding suppresses global dephasing, (2) that leakage is the dominant raw error, and (3) that the multi-state detection provides a faithful erasure flag whose errors do not bias the post-selected decay. The first two are well supported: Fig. 3 shows DFS states outlive physical and non-DFS states, and the raw fidelity decay (~2000 s) matches a leakage-dominated model. The third is the weakest point. The detection fidelity for |1F> is only 93.4%, and the paper explicitly states that leakage from |1F> to |F=4,mF=±1> cannot be directly resolved. An uncalibrated confusion matrix means the post-selection could either retain leaked population (making tau conservative) or discard good states in a phase/T-dependent manner (potentially inflating tau). The claim that a fixed SPAM error does not affect the lifetime measurement is plausible only if the error is constant over the multi-hour measurement, which is not demonstrated. A confusion-matrix measurement and re-fit would settle this. The authors deserve credit for the cryogenic multi-ion setup, the DFS encoding with only two echoes, and the leakage-rate temperature dependence, which independently supports the collision mechanism. The reported point estimates are plausible, but the 'above two hours' headline is not fully secured by the 68% confidence intervals (lower bounds ~4.5e3 s). The reader's CONDITIONAL verdict is appropriate.","tokens_in":11973,"tokens_out":12609,"duration_ms":143277,"concrete_test":"Measure the full confusion matrix of the multi-state detection by preparing known states (|0F>, |1F>, |F=3,mF=±1>, |F=4,mF=±1>, and representative leaked levels) and recording classification probabilities at several storage times (e.g., T=0, 1000, 4000, 8000 s). Then simulate the post-selected fidelity using this matrix and the measured leakage rates, and refit tau_+ and tau_- with detection errors included. If the refitted tau values shift by more than the reported 68% confidence interval or change the lower bound below two hours, the headline claim is not robust to detection errors.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (coherence time above two hours for a DFS logical qubit) depends on the validity of post-selecting runs in which no leakage is detected. For this to be a clean erasure, the multi-state detection must classify each ion as |0F>, |1F>, or leaked, with errors independent of the stored qubit state and storage time T. The paper reports only a single number for |1F> detection fidelity (93.4%) and explicitly notes that leakage from |1F> to |F=4,mF=±1> cannot be directly distinguished from |1F> because of equal Zeeman splittings. If some leaked population is misclassified as |1F>, it is retained and accelerates the apparent decay, making the reported tau a conservative lower bound. Conversely, if |1F> states are misclassified as leaked in a phase- or T-dependent way, the retained ensemble is biased and tau_+ could be inflated. The statement that a fixed SPAM error does not affect the storage lifetime only holds if the error is truly constant; the paper provides no calibration of detection fidelity versus time or versus Zeeman level. A confusion-matrix model is needed to verify that the post-selection does not distort the fitted exponential decay.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a cryogenic trapped-ion quantum memory using three 171Yb+ ions in a dual-type (ground-state/metastable) encoding. Two ions serve as memory ions in the metastable F7/2 manifold, encoding a logical qubit in the decoherence-free subspace (DFS) spanned by |0F 1F> and |1F 0F>; the third ion provides sympathetic cooling without crosstalk. The authors measure storage fidelity as a function of hold time T, observe a raw fidelity decay time of about 2000 s dominated by leakage to nearby Zeeman levels, and then post-select on events where no leakage is detected. After this post-selection they fit exponential decay constants tau_+ = 7.9e3 s and tau_- = 8.0e3 s for the two entangled DFS states, and on this basis claim 'a coherence time above two hours' for the logical qubit. They also study the leakage mechanism and present evidence that it is caused by collisions with background gas molecules.","tokens_in":12209,"tokens_out":4130,"duration_ms":47570,"significance":"If the claimed storage time is sustained, this would be the longest reported coherence time for a multi-ion logical quantum memory and a notable demonstration of dual-type qubit storage in a DFS. The work combines several nontrivial elements: a cryogenic trap to suppress ion hopping, same-species sympathetic cooling, coherent conversion between S-type and F-type qubits, and multi-state detection to herald leakage. The comparison among DFS, physical, and non-DFS encodings in Fig. 3 is a useful controlled demonstration of the DFS advantage. The leakage-mechanism study (temperature correlation with collision rate) is a valuable diagnostic that points toward further improvements. However, the central headline claim is not fully supported by the reported confidence intervals, and the post-selection bias analysis is incomplete, so the significance is conditional on those points being resolved.","major_comments":[{"comment":"The central claim 'coherence time above two hours' is not supported by the reported 68% confidence intervals. For the entangled DFS states the fits give tau_+ = 7.9(+18.0,-3.4) x 10^3 s and tau_- = 8.0(+9.7,-3.1) x 10^3 s, so the one-sigma lower bounds are 4.5 x 10^3 s and 4.9 x 10^3 s, both below 7.2 x 10^3 s (two hours). At 68% confidence the coherence time could be as short as about 1.3 hours. The abstract and title should either state the point estimate with its uncertainty or provide a one-sided confidence level at which the claim 'above two hours' is actually established.","section":"Main text, storage lifetime and Fig. 2(b)"},{"comment":"The assertion that a fixed SPAM error does not affect the storage-lifetime measurement is not justified without a confusion-matrix calibration. The paper reports a |1F> detection fidelity of 93.4% and notes that leakage from |1F> to |F=4,mF=±1> cannot be directly distinguished because of equal Zeeman splittings. If some leaked population is misclassified as |1F>, it is retained in the post-selected ensemble and can either accelerate the apparent decay or, if the misclassification is state- or time-dependent, bias the fitted tau_+ and tau_-. A fixed detection offset would only rescale the exponential amplitude, but the paper does not show that the detection errors are truly constant over the storage times up to about 10^4 s and independent of the qubit state. Please provide a confusion matrix and a test of the post-selection robustness, for example by comparing the fitted tau with and without a model of the detection errors.","section":"Main text, multi-state detection paragraph and the sentence 'such a fixed SPAM error does not affect the measurement…"},{"comment":"The post-selection of no-leakage events is treated as a clean erasure detection, but the paper does not demonstrate that the leakage detection efficiency is independent of the stored logical state and the storage time. The normalization after discarding leakage events in Fig. 2(c) assumes that the discarded events carry no information about the remaining state. If, for instance, the detection of |0F> and |1F> has different efficiencies or if the leakage rate itself depends on the stored state, the post-selected fidelities can be distorted. A control measurement at short storage time T (e.g., T=1s) showing that the post-selected fidelity matches the directly prepared state fidelity, both with and without the leakage discard, would address this concern.","section":"Main text, post-selection procedure and Fig. 2(c)"}],"minor_comments":[{"comment":"The phrase 'after correcting the dominant leakage error' is imprecise: the experiment does not correct the leaked population but post-selects on no-leakage events. The wording should be changed to something like 'after post-selecting on no-leakage detection' to avoid implying that leakage is actively reversed.","section":"Abstract and conclusion"},{"comment":"The caption says the distributions are 'normalized after discarding the leakage events' but does not specify the normalization procedure. Please state whether the remaining populations are renormalized to sum to 1 and whether this renormalization is consistent with the post-selected fidelity computation.","section":"Fig. 2(c) caption"},{"comment":"The dashed curves are described as numerical simulation results based on a simplified error model, but the model is not summarized in the main text. A brief description of the leakage and dephasing rates used in the simulation would help the reader assess the agreement without going to the Supplemental Material.","section":"Fig. 2(a,b) and numerical simulation"},{"comment":"The inset showing the leakage probability of |1F> is a single time point (T=800s). To support the statement that the leakage rate of |1F> is 'similar' to that of |0F>, additional time points or a fitted rate would be more convincing, given that the |1F> channel cannot be fully resolved from the |F=4,mF=±1> levels.","section":"Fig. 4(a) inset"}],"recommendation":"major_revision","confidential_remarks":"The paper is experimentally impressive and the DFS storage concept is well motivated. The main issue is that the headline 'above two hours' is not supported by the one-sigma confidence intervals, and the SPAM/post-selection bias question is load-bearing for the central claim. I believe the authors can address this with a revised statistical statement and additional calibration data, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe short version: this group has likely produced the first multi-ion logical quantum memory with a storage time in the hour-plus range, and they did it with a scheme that removes the two big overheads of the previous record—the need for an ultrastable microwave reference and a long DD sequence. That is real and worth taking seriously. The paper deserves peer review, with a request for one missing calibration and a more careful statement of the headline number.\n\nWhat is genuinely new: combining DFS encoding with dual-type qubits in the same 171Yb+ species in a cryogenic trap, then showing that the DFS logical qubit coherence time is much longer than either the physical qubit or a non-DFS encoding. The control experiment in Fig. 3 makes the central claim persuasive. The leakage analysis is also solid: using a 30-ion chain to amplify leakage and correlating the leakage rate with trap temperature and dark-ion hopping rate gives credible evidence for the gas-collision mechanism. The post-selection is a standard erasure approach, and the contrast between the raw fidelity decay (~2000 s) and the post-selected recovery (~8000 s) is a clean qualitative effect.\n\nThe soft spots, in proportion. First, the headline 'above two hours' is not supported by the confidence intervals. The fitted tau_+ and tau_- are about 7.9e3 and 8.0e3 s, but the 68% CIs run down to 4.5e3 and 4.9e3 s—roughly 1.25 hours. The point estimate clears two hours; the data do not establish it at that confidence. The authors should either report the interval or soften the wording. Second, the paper asserts that the 93.4% detection fidelity on |1F> does not affect the storage lifetime, but no confusion-matrix calibration is shown, and the |F=4,mF=±1> leakage channel is not directly distinguished. If leaked population is misclassified as |1F>, the post-selected tau is a conservative lower bound; if |1F> is thrown away in a T-dependent way, the fit could be biased upward. The authors should show detection fidelity versus storage time and Zeeman level to rule out the second case. The stress-test note flags this correctly. Third, a minor wording issue: the abstract says 'after correcting' while the body says 'discarding' leakage events; those are not equivalent and should be reconciled.\n\nNone of these are load-bearing in the sense that the work collapses without them. The core experiment is convincing, and the likely direction of bias is towards making the post-selected lifetime conservative. But the paper overstates the two-hour number and under-specifies the post-selection calibration.\n\nThis is a strong experimental milestone for ion-trap quantum memories. A serious referee can get it into publishable shape with those two fixes. Send it to peer review and ask for the calibration data and corrected confidence-interval language. It is the kind of paper worth engaging with.","headline":"A real multi-ion DFS logical memory with hour-scale lifetime, but the 'above two hours' claim is a point estimate—the 68% CI lower bound is about 1.25 h—and the post-selection calibration needs to be shown.","tokens_in":12889,"tokens_out":3497,"would_cite":true,"duration_ms":39410,"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 logical qubit encoded in two entangled 171Yb+ ions keeps its coherence for more than two hours after leakage events are discarded.","keywords":["quantum memory","decoherence-free subspace","trapped ions","171Yb+","dual-type qubits","logical qubit","leakage error","erasure detection"],"falsifier":"Vary the fidelity of the leakage detection, for example by changing the duration of the 3432-nm transfer pulse, while keeping the storage protocol fixed, and check whether the post-selected decay constants $\\tau_+$ and $\\tau_-$ shift; a clean erasure post-selection should make the fitted lifetimes independent of the detection error budget.","tokens_in":11745,"feed_emoji":"⚛️","tokens_out":7152,"duration_ms":76263,"temperature":0.7,"pith_summary":"This paper reports a trapped-ion quantum memory that stores a logical qubit encoded in a decoherence-free subspace (DFS) of two entangled 171Yb+ ions, achieving a coherence time above two hours after discarding runs in which the ions leaked to nearby levels. The authors use the dual-type scheme, where memory ions live in long-lived metastable F7/2 states while a third ion of the same species in the ground state provides sympathetic cooling, avoiding the mass mismatch of earlier designs. They show that the remaining dominant error, leakage to neighboring Zeeman levels caused by collisions with background gas, can be detected and treated as an erasure, making the stored logical qubit heralded. Long-lived multi-ion storage matters because quantum repeaters and fault-tolerant computers need memories whose coherence time vastly exceeds elementary gate and network-operation times.","feed_headline":"Two-ion logical qubit stores coherence past two hours","feed_subtitle":"Decoherence-free encoding removes the need for an ultrastable clock; leakage is detected and discarded as erasure.","key_machinery":"The load-bearing object is the decoherence-free subspace spanned by $|0_F1_F\\rangle$ and $|1_F0_F\\rangle$ of two F7/2 memory ions, which is immune to global phase noise. The dual-type scheme maps memory qubits to metastable F7/2 levels and the coolant ion to the S1/2 ground state of the same species, enabling crosstalk-free sympathetic cooling with equal masses. Storage in the DFS is followed by two microwave spin echoes to cancel residual magnetic-field gradients, and leakage to nearby F7/2 Zeeman levels is detected by a multi-state detection sequence with fidelity above 99% for $|0_F\\rangle$ and 93.4% for $|1_F\\rangle$; runs with detected leakage are discarded, converting leakage into an erasure error with known location.","core_discovery":"The central claim is that combining dual-type qubits, a cryogenic trap, DFS encoding, and leakage post-selection makes multi-ion quantum storage practical at the two-hour scale. For the four DFS logical states $|0_L\\rangle$, $|1_L\\rangle$, $|+_L\\rangle$, and $|-_L\\rangle$, the fitted decay constants are $\\tau_0 = 3.3\\times 10^4\\,\\mathrm{s}$, $\\tau_1 = 2.9\\times 10^4\\,\\mathrm{s}$, $\\tau_+ = 7.9\\times 10^3\\,\\mathrm{s}$, and $\\tau_- = 8.0\\times 10^3\\,\\mathrm{s}$, while the raw fidelity decays with a time constant of about $2000\\,\\mathrm{s}$ due to leakage. The paper attributes the leakage to collisions with background gas molecules, evidenced by increasing leakage at higher trap temperature correlated with a higher ion hopping rate. The DFS logical states are shown to be far more robust than a single physical qubit or a non-DFS logical state, demonstrating the benefit of the DFS encoding itself.","pith_inferences":["If the erasure post-selection is clean, the same DFS block could serve as the physical layer under a small quantum error-correcting code that exploits known erasure locations, potentially pushing the logical memory beyond the vacuum-limited lifetime even without better vacuum.","The collision mechanism predicts a quantitative relation between background pressure and Zeeman leakage rate; measuring leakage under deliberately varied hydrogen pressure would test the model and guide vacuum specifications.","The two-ion DFS could be generalized to N-ion symmetric DFS states, trading encoding overhead for additional robustness, provided individual addressing and crosstalk-free sympathetic cooling persist.","One implication the authors state only implicitly is that the reported two-hour T2 is conditional on retaining runs with no detected leakage, so the heralding efficiency, the fraction of runs that survive post-selection, matters as much as the conditional coherence time for practical use."],"forward_implications":["A logical qubit, not just a single physical qubit, can be stored for over two hours, a time more than a million times longer than the elementary gate operations for these dual-type qubits.","DFS encoding removes the need for an ultra-stable microwave frequency reference and for long dynamical-decoupling sequences; only two spin echoes are required.","Because the memory and coolant ions have the same mass, the scheme is expected to scale to larger ion chains without the sympathetic-cooling inefficiency of mixed-species traps.","Detectable Zeeman leakage can be treated as an erasure error with known location, which is friendlier for quantum error correction than an unheralded error.","The raw fidelity decay is dominated by background-gas collisions, so improving vacuum quality should directly extend the unprotected storage time."],"supporting_citations":[{"why":"Prior single-qubit memory exceeding ten minutes, the baseline for long-lived ion storage that this work generalizes to multiple ions.","marker":"[29]"},{"why":"Prior single-qubit memory with estimated coherence exceeding one hour, whose ultrastable-clock and dynamical-decoupling requirements the DFS scheme avoids.","marker":"[30]"},{"why":"Introduces coherently convertible dual-type qubits with the same ion species, the scheme used here for memory and coolant ions.","marker":"[38]"},{"why":"Previous metastable-qubit memory with about 136 s lifetime after leakage discarding, the benchmark this work surpasses.","marker":"[43]"},{"why":"Measured F7/2 level lifetime of 171Yb+ exceeding years, justifying the metastable storage manifold.","marker":"[44]"},{"why":"First decoherence-free quantum memory using trapped ions, the encoding concept underlying the logical qubit.","marker":"[47]"},{"why":"Cryogenic trapped-ion system that suppresses random position hopping, enabling multi-ion storage.","marker":"[37]"},{"why":"Cold ion-atom collision theory used to attribute the Zeeman leakage to background-gas collisions.","marker":"[49–51]"},{"why":"High-fidelity light-shift entangling gate used to prepare the DFS entangled states.","marker":"[52]"}],"fun_headline_variants":["DFS logical qubit holds coherence for over two hours","Two-ion entangled qubit survives two hours in cryotrap","Dual-type memory stores DFS qubit past two-hour mark","Logical qubit with erasure correction exceeds two-hour storage"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that post-selecting away leakage events leaves the remaining state unbiased; if the leakage detector, which is only 93.4 percent reliable for one of the two qubit levels, misclassifies states in a way that correlates with decoherence, the fitted two-hour coherence time would be inflated.","fun_headline_variants_meta":{"raw":{"variants":["DFS logical qubit holds coherence for over two hours","Two-ion entangled qubit survives two hours in cryotrap","Dual-type memory stores DFS qubit past two-hour mark","Logical qubit with erasure correction exceeds two-hour storage"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00016,"raw_usage":{"total_tokens":1224,"prompt_tokens":931,"completion_tokens":293,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":547,"completion_tokens_details":{"reasoning_tokens":225}},"tokens_in":547,"tokens_out":293,"duration_ms":3822,"temperature":1.0,"reasoning_tokens":225,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:26:00.708511+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Vary the fidelity of the leakage detection, for example by changing the duration of the 3432-nm transfer pulse, while keeping the storage protocol fixed, and check whether the post-selected decay constants $\\tau_+$ and $\\tau_-$ shift; a clean erasure post-selection should make the fitted lifetimes independent of the detection error budget.","supporting_citations":[{"cited_title":"Single-qubit quantum memory exceeding ten-minute coherence time,","cited_arxiv_id":null,"evidence_quote":"Prior single-qubit memory with estimated coherence exceeding one hour, whose ultrastable-clock and dynamical-decoupling requirements the DFS scheme avoids."},{"cited_title":"Cryogenic trapped- ion system for large scale quantum simulation,","cited_arxiv_id":null,"evidence_quote":"Introduces coherently convertible dual-type qubits with the same ion species, the scheme used here for memory and coolant ions."},{"cited_title":"Eliminating qubit-type cross-talk in the omg protocol,","cited_arxiv_id":null,"evidence_quote":"Previous metastable-qubit memory with about 136 s lifetime after leakage discarding, the benchmark this work surpasses."},{"cited_title":"Long-lived metastable-qubit memory,","cited_arxiv_id":null,"evidence_quote":"Measured F7/2 level lifetime of 171Yb+ exceeding years, justifying the metastable storage manifold."},{"cited_title":"Decoherence-free subspaces for quantum computation,","cited_arxiv_id":null,"evidence_quote":"First decoherence-free quantum memory using trapped ions, the encoding concept underlying the logical qubit."},{"cited_title":"Character of motional modes for entanglement and sympathetic cool- ing of mixed-species trapped-ion chains,","cited_arxiv_id":null,"evidence_quote":"Cryogenic trapped-ion system that suppresses random position hopping, enabling multi-ion storage."},{"cited_title":"Spin-orbit interactions and quantum spin dynamics in cold ion-atom collisions,","cited_arxiv_id":null,"evidence_quote":"High-fidelity light-shift entangling gate used to prepare the DFS entangled states."}],"review_version":1}