{"id":"a3221560-6df1-4aa5-9102-5dc6ced9dcfa","arxiv_id":"2507.01175","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"In a low-frequency fluxonium qubit, flux noise grows approximately linearly with temperature and dielectric-loss charge noise grows as T^3, while weak in-plane magnetic fields increase dielectric loss.","lead":"A superconducting fluxonium qubit was used to measure how energy loss depends on temperature and magnetic field, revealing that magnetic flux noise rises roughly linearly with temperature and dielectric loss scales as temperature cubed. These empirical trends provide new benchmarks for theories of material defects that limit quantum device coherence.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The T^3 scaling of high-frequency charge noise (β2 = 2.9 in Eq. 8) is not securely established because the extraction assumes T_bath = T_MC and reports no uncertainty; a re-analysis with measured effective temperatures or bootstrap errors is needed.","rationale":"The paper is a solid experimental study: data are aggregated over three cooldowns, the N-level model is a thoughtful extension, and the magnetic-field analysis includes bootstrapped uncertainties and sensitivity checks. I agree with the reader that the weakest assumption is the equality T_bath = T_MC in the temperature sweep, and I would elevate it to the primary load-bearing concern because the T^3 exponent is the most quantitative and novel claim in the abstract. The authors themselves flag the elevated-bath-temperature alternative in Section IV and footnote [72], and they do not provide error bars on the Eq. 8 fit parameters, so the current evidence does not distinguish a genuine T^3 material property from a thermometry artifact. The flux-noise A_Φ ∝ T claim is somewhat better supported by the echo measurements, though it also assumes a temperature-independent α; the B-field dielectric-loss-increase claim has bootstrap and sensitivity support. Neither of those requires a verdict change. A targeted re-analysis or a short thermometry measurement would settle the T^3 question. Therefore the verdict should remain CONDITIONAL (unchanged from the reader), with the explicit condition that the T^3 scaling be re-derived under a measured or free effective temperature and with reported uncertainties.","tokens_in":38159,"tokens_out":10715,"duration_ms":115308,"concrete_test":"Re-analyze the Fig. 3 dataset by resampling (bootstrap, n=1000) the T1 measurements to obtain a 95% confidence interval on β2 in Eq. 8, and repeat the fit using Teff values extracted from measured qubit-state populations at f01 ≈ 200 MHz and ≈1 GHz for each T_MC setting (or with Teff as a free parameter). If the β2 CI includes values ≤ 2, or if an elevated-TeFF model fits the data with β2 near 1, the T^3 scaling is not supported and should be downgraded from a central result to a tentative observation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing concern is the extraction of the T^3 scaling for high-frequency charge noise (β2 = 2.9 in Eq. 8). This exponent comes from a four-parameter power-law fit to Γ1/|⟨0|n|1⟩|² data in which temperature is set to the mixing-chamber temperature T_MC. The paper reports no confidence intervals for α, β1, γ, or β2, unlike the bootstrap analysis used in the magnetic-field section. Critically, Section IV states that offset residuals in the N-level model 'could also suggest an elevated bath temperature relative to the MC stage,' and footnote [72] explicitly allows different effective temperatures for the low- and high-frequency defect baths. In the measured regime (f ≈ 0.1–1.6 GHz, T ≈ 35–100 mK), the physical dielectric-loss expression in Eq. 6 with coth(ℏω/2kBTeff) predicts a roughly linear T dependence; if the high-frequency bath temperature saturates or grows sublinearly with T_MC, the fitted β2 would be biased upward. Because no independent thermometry is reported at the high-frequency bias points, the headline T^3 scaling is a fit-dependent claim resting on an unverified identification of T_MC with the bath temperature.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a study of energy relaxation in a low-frequency fluxonium qubit, with the aim of characterizing the temperature and in-plane magnetic-field dependence of flux and charge noise. T1 is measured as a function of flux bias at base temperature, over a mixing-chamber temperature range of 35–100 mK, and for in-plane fields up to 100 G. The paper's central claims are: (i) the low-frequency flux-noise amplitude AΦ increases approximately linearly with temperature; (ii) the high-frequency charge-noise contribution scales approximately as T^3 (β2 = 2.9 in Eq. 8); and (iii) weak in-plane magnetic fields increase the dielectric-loss contribution to relaxation. The analysis combines circuit-level loss models with a multi-level rate-matrix treatment, and the magnetic-field analysis is supported by bootstrapped confidence intervals and a sensitivity study over fixed noise exponents.","tokens_in":38496,"tokens_out":11922,"duration_ms":129035,"significance":"If correct, these observations would constitute new empirical constraints on microscopic models of flux and charge noise in superconducting circuits: the linear AΦ(T) trend differs from earlier reports of temperature-independent flux noise, and the T^3 charge-noise scaling lies outside simple two-level dielectric-loss predictions. The experimental work has notable strengths: data from three cooldowns are consistent; the AΦ(T) trend is cross-checked with independent spin-echo dephasing data; the magnetic-field section uses empirical bootstrapping and a sensitivity analysis; and the N-level model is tested for convergence and exponentiality. The principal weakness is the provenance of the T^3 exponent, which rests on a fit that assumes T_bath = TMC and reports no uncertainties.","major_comments":[{"comment":"The reported T^3 scaling of the high-frequency charge noise (β2 = 2.9) is not securely established because the fit sets T = TMC and reports no confidence intervals for the four fitted parameters. The physical dielectric-loss expression in Eq. (6) contains a coth(ℏω/2kBTeff) factor that is a strong function of temperature in the measured range (f ≈ 0.1–1.6 GHz, T ≈ 35–100 mK), whereas Eq. (8) folds this factor into a pure power law and attributes all remaining temperature dependence to the noise amplitude. The manuscript itself (Section IV and footnote 72) states that the N-level model residuals 'could also suggest an elevated bath temperature relative to the MC stage' and allows different effective temperatures for the low- and high-frequency defect baths. If the high-frequency bath temperature saturates or grows sublinearly with TMC, the fitted β2 would be biased upward. A re-analysis with independently measured effective temperatures, or at least a sensitivity analysis over plausible T_bath(TMC) curves with bootstrap errors on β2, is required before the T^3 claim can be supported.","section":"Section IV, Eq. (8)"},{"comment":"The same phenomenological fit returns β1 = 0.32 for the flux-noise term, which the text immediately discounts because the second term in Eq. (8) goes to zero as ω → 0. Since β2 is extracted from the same correlated four-parameter fit, the paper should demonstrate that β2 is not similarly contaminated by the low-frequency degeneracy—for example, by refitting with a physically motivated low-frequency model or by excluding low-frequency points—and should reconcile β1 = 0.32 with the separately claimed AΦ ∝ T from Appendix E. Without this, the reader cannot judge whether the T^3 exponent is a stable property of the high-frequency noise or an artifact of the chosen fitting form.","section":"Section IV, Eq. (8) and Appendix E"}],"minor_comments":[{"comment":"The parameter bookkeeping for the magnetic-field fits is ambiguous: the text first states that tanδ0C = 4 × 10−6 and α = 0.62 are used to constrain the B = 0 fit, then states that α and ϵ are held fixed while AΦ and tanδ0C are fitted at each field. Please state explicitly which parameters are fixed and which are free at each step.","section":"Section III and Section V"},{"comment":"The exclusion of the stray TLS near 150 MHz is qualitative; please specify the frequency window excluded and the criterion used to identify it, since the high-frequency fit to Eq. (8) depends on this choice.","section":"Section IV"},{"comment":"The abstract claims a 'power-law dependence of dielectric loss T^3' without the 'approximately' qualifier and without mentioning that the fitted exponent is β2 = 2.9 with no reported uncertainty; the abstract should be qualified to match the body.","section":"Abstract and Section IV"},{"comment":"For TMC > 35 mK, the echo-dephasing parameters are obtained by maximizing the R² of the linearity plot in Fig. 9b; please report the R² values and the sensitivity of the extracted AΦ to this procedure.","section":"Appendix E"},{"comment":"The averaging of points at equal frequencies is described only in the caption; please state how many points are averaged and whether the averaging is performed before or after the model comparison, because it affects the apparent scatter in Fig. 3.","section":"Figure 3"}],"recommendation":"major_revision","confidential_remarks":"The paper is a careful experimental study, and the magnetic-field and AΦ(T) parts are reasonably supported. The T^3 headline, however, is currently a fit-dependent claim with no uncertainty and with an untested T_bath = TMC assumption. I would ask the authors to either provide a re-analysis with independent effective temperatures or soften the claim to, for example, 'a power-law temperature dependence consistent with T^3 under the assumption T_bath = TMC'."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my read. The paper reports three empirical scalings from a low-frequency fluxonium: flux noise magnitude AΦ growing roughly linearly with T, a T^3 power law for the high-frequency dielectric-loss charge noise, and an increase in dielectric loss with in-plane magnetic field. The first and third have real support. The T^3 claim is the shakiest.\n\nWhat's new and good: the device reaches f01 = 52 MHz, so T1 directly probes ~52 MHz noise without needing spin-locking or echo reconstruction. The N-level rate-matrix model is a genuine methodological step for fluxonium, and the authors show it changes the predicted T1 at intermediate flux biases. The AΦ(T) behavior is checked two independent ways: fits to T1 and a separate echo-dephasing sweep. The magnetic-field dataset includes two cooldowns, bootstrapped 95% confidence intervals, and a sensitivity analysis over α and ε. I find that part convincing. They also make a reasonable effort to rule out quasiparticles and vortex loss.\n\nSoft spots: the T^3 result comes out of a single four-parameter fit (Eq. 8) with T set to TMC, and no error bars are reported for β2. The paper itself flags, in Section IV and footnote [72], that the residual offsets could be explained by an elevated bath temperature, and that the low- and high-frequency baths may have different effective temperatures. If the high-frequency bath temperature saturates or lags TMC, β2 would be biased upward. This is exactly the stress-tester's concern, and it is not resolved in the current draft. Also, there is no data/code repository, which matters for a single-device study like this.\n\nBottom line: this is a serious experimental paper and it should go to peer review. The AΦ ∝ T observation alone is worth publishing, and the multi-level model is useful to the fluxonium community. The T^3 claim should either be softened or hardened with a direct test: measure the high-frequency bath temperature, or at least refit Eq. 8 with a floating bath temperature and report confidence intervals on β2. I'd engage with it, but I'd push on that exponent before believing it.","headline":"Careful fluxonium study with a solid AΦ∝T result and a convincing B-field effect; the T^3 charge-noise scaling is a fit-dependent claim that needs a direct bath-temperature test.","tokens_in":39101,"tokens_out":2608,"would_cite":true,"duration_ms":136248,"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":"A low-frequency fluxonium qubit maps how flux noise and dielectric loss respond to temperature and magnetic field, giving microscopic noise theories concrete scaling targets.","keywords":["fluxonium qubit","energy relaxation","flux noise","dielectric loss","two-level systems","temperature dependence","magnetic-field dependence","superconducting qubit coherence"],"falsifier":"Repeat the temperature sweep while independently measuring the effective temperature of the lossy defect bath (for example, through qubit population thermometry or through a calibrated TLS thermometer at the same frequencies); if the bath temperature does not track $T_{MC}$, the fitted $\\beta_2 \\approx 2.9$ no longer describes the material. Alternatively, extend the sweep below 35 mK: if the $T^3$ power law flattens or saturates, the attribution to a thermal defect bath fails.","tokens_in":37911,"feed_emoji":"⚛️","tokens_out":9698,"duration_ms":93112,"temperature":0.7,"pith_summary":"This paper reports temperature- and magnetic-field trends in the two dominant sources of energy decay in a low-frequency fluxonium qubit, and argues these trends should constrain microscopic theories of noise in superconducting circuits. Using a qubit whose transition frequency reaches 52 MHz, the authors measure $T_1$ as a function of flux bias while heating the mixing chamber to 100 mK and while applying in-plane fields up to 100 G. They find the low-frequency flux-noise magnitude scales roughly linearly with temperature, $A_\\Phi \\propto T$, and that the high-frequency charge noise from dielectric loss follows an approximate power law $T^3$ (fitted exponent $\\beta_2 = 2.9$). They also find that weak in-plane magnetic fields increase the dielectric-loss contribution to relaxation, suggesting the charge-coupled defects respond to magnetic field. The paper presents a multi-level decoherence model—a rate matrix over six fluxonium levels—that captures parts of the data a two-level model misses, particularly at intermediate flux biases.","feed_headline":"Flux noise grows with temperature; dielectric loss follows T^3","feed_subtitle":"A 52 MHz fluxonium qubit gives noise theories concrete scaling targets to hit.","key_machinery":"The carrier of the argument is the fluxonium qubit itself, used as a spectrometer of its own noise environment: with minimum frequency $f_{01} = 52$ MHz and widely tunable $|\\langle 0|\\hat{\\phi}|1\\rangle|$ and $|\\langle 0|\\hat{n}|1\\rangle|$ matrix elements, the device maps $T_1$ versus flux bias into the low-frequency flux-noise spectrum $S_\\Phi(\\omega) = A_\\Phi(2\\pi/\\omega)^\\alpha$ and the dielectric-loss spectrum $S_Q(\\omega)$ via Fermi's golden rule. The second piece of machinery is the $N$-level rate-matrix decoherence model ($N = 6$), which accounts for heating transitions out of the computational subspace and is needed to capture $T_1$ at intermediate flux biases, especially for dielectric loss. The third is the phenomenological power-law model of Eq. 8, which jointly fits flux- and charge-noise terms with independent frequency and temperature exponents and yields the headline values $\\alpha = 1.5$, $\\beta_1 = 0.32$, $\\gamma = 0.19$, $\\beta_2 = 2.9$.","core_discovery":"The central claim is that in a fluxonium qubit operated at low frequency, the intrinsic noise that limits $T_1$ has three empirically distinguishable behaviors: flux-noise amplitude grows approximately linearly with temperature ($A_\\Phi \\propto T$ for 35–100 mK, probed at 1–100 MHz through $T_1$ and echo dephasing); the dielectric-loss charge noise grows as roughly $T^3$, extracted from a phenomenological fit with fitted exponent $\\beta_2 = 2.9$; and in-plane magnetic fields up to 100 G increase the inferred dielectric loss while also reshaping the flux-noise contribution, so that the dominant loss mechanism crosses over from flux noise to charge noise at high field. The authors argue these scaling laws, taken with the multi-level rate-matrix corrections, mean that fluxonium coherence models should not assume temperature-independent noise spectra, and that any microscopic theory of surface spins and charge TLS must reproduce the observed temperature and field trends.","pith_inferences":["I would read the $T^3$ dielectric-loss term as a signature worth testing against a resonant-absorption mechanism in which the TLS lifetime itself scales as $T^\\beta$; the paper notes this could produce a power-law qubit rate, and a direct measurement of the TLS lifetime temperature dependence on the same device would discriminate it from relaxation absorption.","A natural next experiment is to separate frequency and temperature scaling by measuring $T_1$ at fixed frequencies rather than along the flux-bias sweep; fitting the phenomenological model with independent temperature sweeps at, say, 100, 300, and 800 MHz would pin down whether $\\beta_2$ is truly a material exponent or an artifact of the elevated-bath-temperature ambiguity the paper flags.","The apparent magnetic response of dielectric loss could be checked on a non-qubit device, such as a superconducting resonator without a flux-tunable element, at the same fields; if resonator loss rises with $B$, the effect is a property of the charge-TLS bath, while if only the qubit sees it, fit assumptions about fixed $\\alpha$ and $\\epsilon$ deserve scrutiny.","If $A_\\Phi \\propto T$ is confirmed as a general feature of low-frequency fluxonium, it may also explain part of the variation in reported flux-noise amplitudes across devices: measurements taken at different base temperatures would be compared only after rescaling to a common temperature."],"forward_implications":["Fluxonium coherence simulations should incorporate transitions beyond the $|0\\rangle \\leftrightarrow |1\\rangle$ manifold whenever transition energies approach $k_BT$; the $N$-level rate matrix materially changes predicted $T_1$ at intermediate flux biases, where the two-level model fails.","Microscopic theories of flux noise must allow for a temperature-dependent noise amplitude: $A_\\Phi \\propto T$ contradicts earlier observations of temperature-independent flux noise below about 100 mK in SQUIDs and flux qubits, so either device-specific or regime-specific mechanisms are required.","Dielectric-loss models for low-frequency qubits cannot assume a temperature-independent loss tangent: the $T^3$ growth implies existing two-level dielectric models underestimate how fast $T_1$ degrades above base temperature.","Magnetic-field studies of fluxonium coherence should treat the dielectric-loss channel as field-sensitive: the increase in inferred loss up to 100 G means charge-coupled defects, not only surface spins, respond to applied fields.","The reported scaling laws supply concrete benchmark targets—$A_\\Phi(T)$, $\\beta_2$, and the field-dependent crossover—against which future microscopic models of surface spins and charge TLS can be tested."],"supporting_citations":[{"why":"Provides the magnet setup and the prior observation of a similar field-induced jump in $\\Gamma_1$, against which the magnetic-field trends here are compared.","marker":"[36]"},{"why":"Supplies the spin-locking noise-spectroscopy technique used to independently estimate the flux-noise amplitude and exponent $\\alpha$.","marker":"[6]"},{"why":"Provides the $1/f^\\alpha$ flux-noise model and the frequency-scaling conventions adopted for $S_\\Phi(\\omega)$.","marker":"[7]"},{"why":"Documents the prior observation of temperature-dependent $1/f$ flux noise and the classical-quantum crossover that this paper contrasts with its linear $A_\\Phi \\propto T$ result.","marker":"[11]"},{"why":"Supplies the standard tunneling model whose resonant and relaxation absorption mechanisms are evaluated as candidates for the $T^3$ dielectric-loss scaling.","marker":"[19]"},{"why":"Provides the dimensional-transformation treatment of defect-induced loss that yields the $\\tan\\delta \\propto T^d$ relaxation-absorption scaling discussed for the charge noise.","marker":"[65]"},{"why":"Supplies the high-coherence fluxonium dielectric-loss parametrization ($\\tan\\delta_C^0$, frequency exponent $\\epsilon$) used to model charge noise.","marker":"[50]"},{"why":"Provides the fluxonium energy-relaxation analysis including quasiparticle and dielectric-loss channels that the baseline model extends with $N$ levels.","marker":"[54]"}],"fun_headline_variants":["Fluxonium qubit: flux noise ~T, dielectric loss ~T^3, field-sensitive","T scaling of fluxonium noise: flux ~T, dielectric loss ~T^3","Field tunes dielectric loss in fluxonium qubit; flux noise scales with T","Flux noise linear, dielectric cubic: fluxonium qubit noise laws"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central $T^3$ result rests on assuming the defect bath that causes the loss sits at the measured mixing-chamber temperature across the 35–100 mK sweep; the paper itself notes that an elevated bath temperature would alter or remove the $T^3$ conclusion.","fun_headline_variants_meta":{"raw":{"variants":["Fluxonium qubit: flux noise ~T, dielectric loss ~T^3, field-sensitive","T scaling of fluxonium noise: flux ~T, dielectric loss ~T^3","Field tunes dielectric loss in fluxonium qubit; flux noise scales with T","Flux noise linear, dielectric cubic: fluxonium qubit noise laws"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00119,"raw_usage":{"total_tokens":4893,"prompt_tokens":910,"completion_tokens":3983,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":526,"completion_tokens_details":{"reasoning_tokens":3894}},"tokens_in":526,"tokens_out":3983,"duration_ms":32738,"temperature":1.0,"reasoning_tokens":3894,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:58:27.966233+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the temperature sweep while independently measuring the effective temperature of the lossy defect bath (for example, through qubit population thermometry or through a calibrated TLS thermometer at the same frequencies); if the bath temperature does not track $T_{MC}$, the fitted $\\beta_2 \\approx 2.9$ no longer describes the material. Alternatively, extend the sweep below 35 mK: if the $T^3$ power law flattens or saturates, the attribution to a thermal defect bath fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the dimensional-transformation treatment of defect-induced loss that yields the $\\tan\\delta \\propto T^d$ relaxation-absorption scaling discussed for the charge noise."}],"review_version":1}