{"id":"560d7547-0057-4083-9935-6f0858536ef7","arxiv_id":"2506.02187","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Moderate magnetic fields (trapped up to 600 mG, applied up to 400 mG) reduced temporal T1 fluctuations by more than a factor of two in Nb/Ta transmon qubits without significantly lowering mean T1.","lead":"Researchers tested whether magnetic fields can stabilize superconducting qubits instead of harming them. They found that moderate trapped or applied fields reduced hour-scale fluctuations in the qubit relaxation time while leaving its average value nearly unchanged.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline claim depends on comparing T1 fluctuation amplitudes across separate cooldowns (Table I) and a monotonic field sequence (Fig. 4), with no repeated zero-field baselines; known nonstationary T1 noise is a plausible alternative explanation.","rationale":"The paper's central claim is causal: magnetic fields reduce temporal T1 fluctuations. The strongest evidence is Table I for trapped flux and Fig. 4 for applied fields. The load-bearing assumption is that the measured MAD differences reflect the magnetic field rather than nonstationary noise and cooldown-to-cooldown variation. The reader identified exactly this weakness, and I agree with that assessment. The design has no repeated zero-field baselines for the trapped-flux runs and a monotonic field sequence for the applied-field runs; given the known nonstationarity of T1, the causal identification is insecure. However, there is nontrivial internal consistency: q1 and q2 both show monotonic MAD reduction up to 600 mG, q3 shows monotonic suppression with applied field, and the Allan deviation analysis indicates reduced noise amplitudes. These observations make the claim plausible but not established. The proposed randomized crossover and zero-field return would settle the concern. Since the reader's conditional verdict already reflects this uncertainty, no change to the verdict is warranted.","tokens_in":9823,"tokens_out":4081,"duration_ms":40385,"concrete_test":"Perform a randomized crossover on one qubit (e.g., q2): in each of at least six cooldowns, apply B_trapped = 0 or 600 mG in randomized order (three runs each), measuring T1 for 24 h with the identical protocol. Compute MAD for each run and test whether MAD(600) is consistently lower than MAD(0) using a paired permutation test across cooldowns. If the effect does not survive this between-cooldown design, the suppression is not robust. Separately, for the applied-field claim on q3, repeat the field sequence with inserted zero-field blocks (e.g., 0-100-200-400-0-400-200-100-0) in one cooldown; if MAD decreases monotonically with field strength and recovers at zero regardless of position in the sequence, drift and order effects are ruled out.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section III A and Table I compare the mean absolute deviation (MAD) of T1 between B_trapped = 0, 400, and 600 mG, but each condition is a distinct cooldown: q1 was measured for 12 h and q2 for 24 h, with no repeated zero-field baseline and no randomization of condition order. T1 fluctuations in superconducting qubits are strongly nonstationary, with cooldown-dependent TLS configurations and quasiparticle environments that can change the apparent fluctuation amplitude by factors of two or more (Klimov et al., PRL 121:090502; Carroll et al., npj Quantum Inf 8:132). The observed reductions (q2: MAD from 38.0 to 16.7/13.7 μs; q1: from 10.1 to 7.2/5.2 μs) may therefore reflect which cooldown each run happened to be, not the effect of the magnetic field. The Allan deviation analysis in Table II is derived from the same single runs and has no reported uncertainties, so it inherits the same confound. For the actively applied field data on q3 (Section III C, Fig. 4), the measurement sequence is monotonic in field (0 → 100 → 200 → 400 mG) within one cooldown, with no return to zero field; a slow monotonic drift in T1 stability over the 15 h run would produce the same trend. Without repeated zero-field baselines interleaved with field conditions, or a randomized crossover design, the central causal claim that magnetic fields suppress T1 fluctuations is not identified.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports measurements on Nb/Ta transmon qubits showing that both magnetic flux trapped during cooldown (up to 600 mG) and static magnetic fields applied during operation (up to 400 mG) reduce the temporal fluctuations of T1, quantified by the mean absolute deviation (MAD), without significantly changing the mean T1 or qubit frequency. Higher fields (800–1000 mG) cause a sharp degradation of T1. The authors support the central claim with box plots for two qubits (q1, q2) under trapped-flux conditions, a third qubit (q3) under applied fields, and an Allan deviation analysis of one qubit (q2). They interpret the stabilization as resulting from paramagnetic impurity polarization, quasiparticle trapping in vortices, and partial saturation of two-level systems.","tokens_in":10293,"tokens_out":2601,"duration_ms":25258,"significance":"The reported effect, if robust, is significant: it challenges the general assumption that magnetic fields are always detrimental to superconducting qubits and could offer a practical route to stabilizing coherence in devices with Nb/Ta capacitor pads. The experiment benefits from a dedicated three-axis Helmholtz coil system with in-situ field monitoring, long-duration T1 tracking (12–24 h), and a quantitative noise analysis via Allan deviation. However, the causal claim that magnetic fields suppress T1 fluctuations is currently identified through comparisons across separate cooldowns and a monotonic field sequence, without statistical tests or repeated zero-field baselines, so the strength of the evidence is not yet commensurate with the breadth of the conclusion.","major_comments":[{"comment":"The comparison of T1 fluctuations between B_trapped = 0, 400, and 600 mG is confounded by cooldown-to-cooldown variability, because each trapped-flux condition for q1 and q2 was realized in a separate cooldown and no repeated zero-field baseline was measured. Given the well-documented nonstationarity of T1 fluctuations in superconducting qubits (e.g., Klimov et al., PRL 121:090502, and Carroll et al., npj Quantum Inf 8:132), the reduction in MAD from 38.0 μs to 16.7/13.7 μs for q2 might reflect the specific cooldown rather than the applied field. The authors should provide repeated zero-field cooldowns interleaved with the field conditions, or multiple cooldowns per condition, and report a statistical test (e.g., bootstrap confidence intervals for M and MAD).","section":"§III A, Table I"},{"comment":"The actively applied field experiment on q3 uses a monotonic field sequence (0 → 100 → 200 → 400 mG) within a single 15-hour run, with no return to zero field. A slow monotonic drift in qubit stability over the course of the run would produce exactly the observed trend of decreasing MAD with field. To support the causal claim, the field sequence should be randomized or interleaved with repeated zero-field segments, allowing the authors to separate a field effect from time-dependent environmental drift.","section":"§III C, Fig. 4"},{"comment":"The Allan deviation analysis is presented as strong evidence ('n0 drops by nearly an order of magnitude'), but the fitted noise amplitudes n0, n1, n2 are extracted from a single 24-hour run per field condition and no uncertainties or goodness-of-fit metrics are given. Without confidence intervals on these fitted parameters, the claimed suppression of white noise amplitude is not statistically supported. At minimum, the authors should report fit uncertainties and, ideally, repeat measurements or apply a bootstrap over the measured T1 time series.","section":"§IV, Table II and Fig. 5"}],"minor_comments":[{"comment":"The expression N = B × A / Φ0 does not specify whether A is the projected area of the superconducting film or the total surface area; please clarify the definition used for estimating vortex number.","section":"Eq. (1)"},{"comment":"The table lists M and MAD without any measure of uncertainty or number of T1 samples; adding the standard error or the number of measurements would help the reader judge the stability of the reported values.","section":"Table I"},{"comment":"The statement 'No correlation was observed between the magnetic field and qubit frequency shift or dephasing parameters' is made without supporting data; providing a plot or the numerical limits would make this claim verifiable.","section":"§III, opening paragraph"},{"comment":"The Allan deviation formula uses τ but the text does not define the range of averaging times over which the fit was performed; please specify the τ values or the fitting window.","section":"§IV"},{"comment":"There are minor formatting inconsistencies, such as 'FIG. 1a' versus 'FIG. 1a' and the use of 'T1' with and without a subscript; a careful copyedit would improve readability.","section":"Various"}],"recommendation":"major_revision","confidential_remarks":"The central observation is interesting but the experimental design does not yet rule out the dominant alternative explanation of nonstationary T1 noise. The authors should be encouraged to collect interleaved or repeated zero-field baselines and to add statistical tests to the reported M and MAD values. The paper would be a good fit for a quantum information or applied superconductivity journal once the causal identification is strengthened."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline is that a well-equipped experimental group reports a practically useful effect: moderate magnetic fields—trapped flux up to 600 mG and applied fields up to 400 mG—reduce temporal T1 fluctuations by more than a factor of two in high-coherence Nb/Ta transmons, without significantly degrading the mean T1. That is worth taking seriously because the setup is unusually careful: calibrated three-axis Helmholtz coils and in-situ fluxgate monitoring give the measurements a credibility that many coherence studies lack.\n\nThe newness is real but bounded. Earlier work (Refs. 28, 29) saw similar qualitative effects in low-T1 aluminum and titanium nitride qubits. This paper extends the observation to ~300 µs Nb/Ta devices and identifies a sharp threshold around 600–800 mG, which is a useful engineering constraint. The box plots and Allan deviation fits show consistent trends across three qubits, and the authors are appropriately cautious about mechanisms, offering three plausible explanations without overclaiming.\n\nThe soft spot is the statistical identification. Each trapped-flux condition for q1 and q2 was measured in a separate cooldown, with no repeated zero-field baseline and no randomization of condition order. T1 noise in superconducting qubits is known to be nonstationary on the time scale of these 12–24 hour runs (Klimov et al., Carroll et al.), so the apparent suppression could be cooldown-to-cooldown luck rather than the field. The applied-field data on q3 are within a single cooldown but the sequence is monotonic in field with no return to zero, so a slow drift over the 15-hour run would produce the same trend. There are no error bars or hypothesis tests on the headline MAD numbers, and the Allan deviation fits inherit the same confound.\n\nThat does not mean the claim is wrong. The reductions are large (q2 MAD from 38.0 to 13.7 µs; q1 from 10.1 to 5.2 µs), and the threshold behavior is visually sharp. But the evidence, as presented, is suggestive rather than conclusive. A serious referee should send this back for repeated zero-field cooldowns interleaved with field conditions, or at least a randomized crossover design on one or two qubits, plus uncertainty quantification on the MAD and Allan amplitudes.\n\nBottom line: this is a sincere paper from a strong group with a plausible effect and honest reporting. It deserves peer review, probably with major revisions requested on the experimental design. I would not yet build a calibration strategy on it, but I would want it in the literature once the confounds are addressed.","headline":"A credible but under-powered experimental report that moderate magnetic fields stabilize T1 in high-coherence Nb/Ta transmons; the effect is plausible and the engineering is solid, but the statistical identification is confounded by separate cooldowns and nonstationary T1 noise.","tokens_in":10729,"tokens_out":2706,"would_cite":true,"duration_ms":24476,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Moderate magnetic fields, whether trapped or applied, more than halve temporal T1 fluctuations in Nb/Ta transmon qubits without degrading average coherence.","keywords":["transmon qubits","T1 fluctuations","magnetic flux trapping","Helmholtz coil system","quasiparticle trapping","two-level systems","coherence stabilization","niobium-tantalum capacitors"],"falsifier":"Run a series of interleaved cooldowns on a single qubit: zero field, 600 mG, zero field, 600 mG, each followed by a 24-hour T1 trace; if the zero-field runs sometimes show MAD as low as the 600 mG runs, the suppression is not magnetic.","tokens_in":9643,"feed_emoji":"🧲","tokens_out":6206,"duration_ms":52887,"temperature":0.7,"pith_summary":"Energy relaxation time T1 in superconducting transmon qubits fluctuates over hours, complicating calibration and gate fidelity. This paper claims those fluctuations can be suppressed by engineered magnetic fields. Using a three-axis Helmholtz coil system, the authors cooled Nb/Ta transmon qubits in static fields up to 600 mG to trap flux, and applied fields up to 400 mG during measurement. In both cases T1's mean value stayed near its zero-field level while its temporal fluctuation, measured by mean absolute deviation and Allan deviation, dropped by more than a factor of two. If correct, magnetic field control becomes a practical tool for stabilizing superconducting qubits rather than only a disturbance to be shielded against.","feed_headline":"Magnetic fields can stabilize qubit T1 noise, new data show","feed_subtitle":"Trapped flux up to 600 mG and applied fields up to 400 mG more than halve T1 fluctuations without hurting average coherence.","key_machinery":"The experimental machinery is a three-axis Helmholtz coil assembly wound from copper-clad superconducting wire, with fields calibrated in situ by a fluxgate magnetometer; it lets the experimenters apply a known perpendicular field either during cooldown, to trap flux in the capacitor pads, or during T1 measurement at 8 mK. The analysis machinery is Allan deviation analysis, a time-domain statistic that decomposes the T1 time series into white, flicker, and random-walk noise amplitudes. The physical mechanism proposed is threefold: polarization of paramagnetic impurities such as O2, NbO, and TaNb; trapping of non-equilibrium quasiparticles in vortex cores; and saturation of high-frequency two-level-system loss channels.","core_discovery":"The central discovery is that there is a window of magnetic field strengths in which transmon coherence becomes more stable rather than worse. For a qubit with mean T1 of 142.7 microseconds at zero field, cooling in 600 mG reduced the mean absolute deviation from 10.1 microseconds to 5.2 microseconds while mean T1 stayed at 140.2 microseconds; for a second qubit, MAD fell from 38.0 microseconds to 13.7 microseconds at mean T1 near 291 microseconds. A third qubit cooled in zero field and measured under applied fields of 100 to 400 mG showed the same trend: mean T1 remained around 205 to 218 microseconds while MAD fell from 23.2 microseconds to 10.7 microseconds. Trapped fields of 800 mG and above caused sharp degradation, placing a threshold between 600 and 800 mG. Allan deviation fits show the suppression is largely in the white-noise amplitude n0, which drops by nearly an order of magnitude at 400 to 600 mG.","pith_inferences":["The mechanism list is not discriminated by this dataset; a testable separation would be to compare perpendicular versus in-plane fields, since paramagnetic polarization should be more isotropic while vortex trapping is direction-sensitive.","If quasiparticle trapping is the dominant term, the stabilization should weaken when quasiparticle density is independently reduced by normal-metal traps, a prediction the paper does not test.","The reported MAD reductions imply a direct operational benefit: fewer T1 recalibrations over a 12 to 24 hour experiment, which matters for automating large processors.","The sharp threshold between 600 and 800 mG suggests material and geometry dependence, so varying pad spacing and junction area could shift the window and make the effect tunable per device."],"forward_implications":["Cooldown protocols can be tuned to trap 400 to 600 mG of flux, more than halving T1 fluctuations without recalibrating qubit frequency or average T1.","Applied static fields up to 400 mG can stabilize a qubit during operation, offering a live knob for noise suppression.","Allan deviation data indicate the suppression targets white noise, with n0 dropping nearly tenfold, so short-timescale calibration stability should improve.","Fields above 600 to 800 mG must be avoided; the sharp threshold defines a clear operating window for magnetic field engineering.","The effect, if reproduced across devices, reduces the recalibration overhead caused by T1 drift in multi-qubit processors."],"supporting_citations":[{"why":"Supplies the Nb/Ta transmon devices and their baseline coherence, the platform all measurements use.","marker":"[18]"},{"why":"Quantifies trapped-flux surface resistance in niobium resonators at millikelvin temperatures, the loss framework the qubit comparison extends.","marker":"[23]"},{"why":"Shows vortices can act as quasiparticle sinks, one of the proposed stabilization mechanisms.","marker":"[28]"},{"why":"Reports weak-field coherence enhancement in a transmon, the prior observation this work extends to high-T1 qubits.","marker":"[29]"},{"why":"Provides vortex flux thresholds in aluminum and rhenium films, setting the comparison for the observed 600 to 800 mG threshold.","marker":"[30]"},{"why":"Documents T1 fluctuations in superconducting qubits, the noise phenomenon being suppressed.","marker":"[6]"},{"why":"Describes quasiparticle relaxation in the presence of flux, underlying the vortex-sink mechanism.","marker":"[32]"},{"why":"Introduces Allan deviation analysis, the method used to quantify noise suppression.","marker":"[35]"}],"fun_headline_variants":["Moderate magnetic fields slash T1 noise in transmon qubits","Magnetic fields tame T1 fluctuations in transmon qubits","Field engineering reduces T1 noise in superconducting qubits","Moderate fields stabilize transmon coherence, T1 noise halved"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The results assume that differences in T1 noise between field settings are caused by the field, not by the fact that each setting was measured in a separate cooldown; no zero-field baseline was repeated and no statistical test was applied.","fun_headline_variants_meta":{"raw":{"variants":["Moderate magnetic fields slash T1 noise in transmon qubits","Magnetic fields tame T1 fluctuations in transmon qubits","Field engineering reduces T1 noise in superconducting qubits","Moderate fields stabilize transmon coherence, T1 noise halved"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000198,"raw_usage":{"total_tokens":1405,"prompt_tokens":1020,"completion_tokens":385,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":636,"completion_tokens_details":{"reasoning_tokens":314}},"tokens_in":636,"tokens_out":385,"duration_ms":3753,"temperature":1.0,"reasoning_tokens":314,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:28:21.794676+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a series of interleaved cooldowns on a single qubit: zero field, 600 mG, zero field, 600 mG, each followed by a 24-hour T1 trace; if the zero-field runs sometimes show MAD as low as the 600 mG runs, the suppression is not magnetic.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Nb/Ta transmon devices and their baseline coherence, the platform all measurements use."},{"cited_title":"Bafia, B","cited_arxiv_id":null,"evidence_quote":"Quantifies trapped-flux surface resistance in niobium resonators at millikelvin temperatures, the loss framework the qubit comparison extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows vortices can act as quasiparticle sinks, one of the proposed stabilization mechanisms."},{"cited_title":"Schneider, T","cited_arxiv_id":null,"evidence_quote":"Reports weak-field coherence enhancement in a transmon, the prior observation this work extends to high-T1 qubits."},{"cited_title":"Song, T.W","cited_arxiv_id":null,"evidence_quote":"Provides vortex flux thresholds in aluminum and rhenium films, setting the comparison for the observed 600 to 800 mG threshold."},{"cited_title":"Klimov, M","cited_arxiv_id":null,"evidence_quote":"Documents T1 fluctuations in superconducting qubits, the noise phenomenon being suppressed."},{"cited_title":"Catelani, J","cited_arxiv_id":null,"evidence_quote":"Describes quasiparticle relaxation in the presence of flux, underlying the vortex-sink mechanism."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces Allan deviation analysis, the method used to quantify noise suppression."}],"review_version":1}