{"id":"70bec949-1e0d-448f-8d18-aa3db5d4fc8f","arxiv_id":"2508.03873","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A microwave study of granular-aluminum resonators extracts kinetic inductance from temperature shifts and finds two-level-system losses, providing design data for high-kinetic-inductance devices.","lead":"Researchers built and cooled tiny superconducting microwave circuits made from granular aluminum, a metal sprinkled with oxide grains, and measured how their resonance frequency and losses change with temperature and power. These measurements give designers numbers they need for building sensitive microwave detectors and quantum-circuit parts.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Mattis-Bardeen local-limit, single-gap extraction is the load-bearing step; the authors' own caveat and the unresolved α1≠α2 discrepancy leave the reported values unvalidated.","rationale":"The reader's weakest assumption is that Mattis-Bardeen local-limit surface impedance with a single BCS gap describes strongly disordered granular aluminium. This is exactly the load-bearing condition for the central claim: α1 is obtained by fitting the shape of the frequency shift to that theory, and all sheet inductances follow from α1. The concern lands because the paper itself admits the theoretical expressions are not justified for the extracted λ_eff values, and because the internal cross-checks do not resolve the discrepancy: α1 and α2 disagree by up to 22%, and the rigid-shift values differ from α1 by 9–25%. These discrepancies are larger than the quoted statistical errors, so the numerical values should not be accepted as definitive without an independent test of the MB kernel. The proposed free-gap refit is a concrete, minimal check: it directly exposes whether the assumed gap ratio is the source of the bias. The qualitative conclusions—kinetic inductance increases with oxygen content and TLS losses are present—are well supported by the raw frequency shifts and power dependence, so the paper's overall direction is not in doubt. I therefore keep the reader's CONDITIONAL verdict unchanged rather than escalating to rejection. Agreement is 'agree' because the reader's weakest_assumption names the same underlying model dependence, and the additional α1/α2 evidence is already noted in the reader's rationale.","tokens_in":17209,"tokens_out":7078,"duration_ms":93370,"concrete_test":"For each resonator, refit the measured Δf(T)/f0 with Eq. (7) treating Δ0/kBTc as a free parameter, and optionally adding the TLS frequency-shift term of Eq. (9) weighted by the operating photon number; report the joint confidence region for α1 and Δ0/kBTc. If the best-fit Δ0/kBTc deviates from 1.76 by more than ~10%, or if α1 shifts by more than the stated statistical errors, the single-gap Mattis-Bardeen assumption is biasing the extraction and the Table II values should be revised or presented with a model-dependent error bar.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The quantitative claims—α1 from 0.137 to 0.356, and sheet kinetic inductances from 1.8 to 6.9 pH/□—all depend on the assumption that the temperature dependence of the surface reactance of these disordered granular films is described by Mattis-Bardeen theory in the local limit with a single BCS gap Δ0 = 1.76 kBTc. Section V.1 states L□_k = μ0λ²/t, giving λ_eff ≈ 800 nm even for the smallest L□_k, and then explicitly says 'the use of the theoretical expressions to extract α is not justified and we take these values only as indicative.' Because the fit in Eq. (7) contains α only as a scale factor multiplying a theoretically prescribed ΔXs(T)/Xs(0), any error in the assumed gap ratio, gap distribution, or local-limit kernel is absorbed into α1. This is not a purely academic worry: fitting the same MB kernel to the quality-factor shift gives α2 that differs from α1 by up to 22%, which the authors attribute to unsaturatated TLS without removing that contribution from the frequency-shift fit. The independent rigid-shift estimates also disagree with α1 by 9–25% depending on whether simulated or conformal-mapping frequencies are used. The reported Table II sheet inductances are then obtained by matching α(L□_k) to α1, so the same potential bias propagates into every derived quantity. The central claim is therefore only as secure as the unverified BCS local-limit assumption, which the paper itself flags as not justified.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports microwave characterization of coplanar-waveguide quarter-wave resonators fabricated from four aluminum films with different oxygen content (Al, grAl-1, grAl-2, grAl-3). The authors extract the kinetic-inductance fraction by fitting the temperature dependence of the resonance frequency and quality factor to Mattis-Bardeen theory (α1 from Eq. (7), α2 from Eq. (8)), compare these values with rigid-shift estimates based on finite-element simulations and conformal mapping, derive sheet kinetic inductances by inversion of the conformal-mapping formula (Table II), and characterize two-level-system losses from the power dependence of the internal quality factor (Eq. (10)). The main reported results are α1 between 0.137 and 0.356, sheet kinetic inductances between 1.8 and 6.9 pH/□, and an average TLS loss factor F·δ_TLS0 of about 1.5×10⁻⁴. The paper is framed as a characterization study useful for MKIDs, high-impedance resonators, and superinductors.","tokens_in":17591,"tokens_out":6284,"duration_ms":72623,"significance":"If the quantitative extraction were validated, this would be a useful contribution: it provides a consistent set of kinetic-inductance and TLS parameters for granular aluminum films with different disorder levels, along with detailed fabrication and design information. Strengths include the careful multi-probe film characterization (TEM, XRD, profilometry), the explicit reporting of discrepancies among extraction methods, and the TLS power-dependence analysis yielding β close to 1, consistent with the standard TLS model. However, the central quantitative claims are currently contingent on an unvalidated Mattis-Bardeen local-limit, single-gap assumption that the authors themselves flag as not justified; the paper would be considerably strengthened by an independent validation route or a clear statement of model uncertainty on every derived quantity.","major_comments":[{"comment":"The central extraction assumes that the Mattis-Bardeen local-limit, single-gap surface impedance (Eqs. (5)–(7)) describes these strongly disordered granular films, but this assumption is neither tested nor independently supported. The authors themselves state in §V.1 that λ_eff ≈ 800 nm for 1 pH/□ and that 'the use of the theoretical expressions to extract α is not justified and we take these values only as indicative.' Since Eq. (7) multiplies the theoretically prescribed ΔX_s(T)/X_s(0) by a single scale factor α, any error in the gap ratio, gap distribution, or local-limit kernel is absorbed into α1 and propagates through Table II into L□_k. The paper should validate the Mattis-Bardeen assumption for these films (for example, by comparing against a direct low-temperature measurement of the geometric resonance frequency from the same geometry, or by modeling the disordered-film conductivity) or explicitly present α1 and all derived quantities as indicative values with a model uncertainty.","section":"§V.1, Eq. (7), Table II"},{"comment":"The unresolved discrepancy between α1 and α2 (1–22%) is attributed to unsaturated two-level systems, but the frequency-shift data used for α1 are not corrected for the TLS frequency shift described by Eq. (9). If TLS also shifts the resonance frequency at the operating photon number, then α1 is biased as well and cannot be considered more reliable merely because it comes from the frequency shift rather than the quality factor. A joint fit of Δf/f0(T) and 1/Qi(T) that includes both the Mattis-Bardeen term and the TLS term, or an explicit numerical estimate of the TLS-induced frequency shift at the measurement power, is needed before α1 is adopted as the reported kinetic-inductance fraction.","section":"§V.2, Fig. 4, Table II"},{"comment":"The sheet inductances in Table II are not an independent check of the Mattis-Bardeen extraction: they are obtained by solving α(L□_k) = α1 with the conformal-mapping expression of Eq. (B9), so any systematic error in α1 is transferred directly to L□_k. Moreover, the rigid-shift estimates α_SIM and α_CM are systematically 9–25% lower than α1, and the origin of this offset is not explained; the spread is comparable to or larger than the quoted statistical errors. The paper should either reconcile these estimates (including the effect of the coupling capacitance and finite film thickness on the geometric frequency f_g) or quote a model uncertainty that covers the full spread of values from the different extraction methods.","section":"§V.1, Fig. 3, Table II"},{"comment":"The internal inconsistency between the presentation of α1 and L□_k as extracted values and the statement that the theoretical expressions used for the extraction are 'not justified' should be resolved in the text. As written, the conclusion states that the authors 'were able to extract the kinetic inductance contribution' and provide the sheet kinetic inductance, while the body of the paper simultaneously warns that these values are only indicative. The authors should either strengthen the validation to justify the numbers or consistently frame all quantitative kinetic-inductance results as estimates with explicit error bars that include the model uncertainty.","section":"§V.1 and §VI"}],"minor_comments":[{"comment":"The theory name is spelled 'Mathis-Bardeen' at the end of the introduction; it should be 'Mattis-Bardeen' as used elsewhere.","section":"§I"},{"comment":"The typeset equations contain garbled exponential factors and bracket expressions in the provided text; please ensure the final formulas are correct and consistent with the standard Mattis-Bardeen and TLS expressions.","section":"Eqs. (5), (6), (10)"},{"comment":"The caption says the extracted α values are indicated in the legend, but the numerical values are not visible in the current figure; adding the fitted values to the caption would improve readability.","section":"Fig. 4 caption"},{"comment":"The phrase 'with two outliers resonators which show a larger participation ratio' is unclear; please specify which resonators are outliers and quantify the deviation from the average F·δ_TLS0.","section":"§V.2"},{"comment":"No data or code availability statement is provided; archiving the S21 datasets, the fitted resonance parameters, and the fit residuals would substantially improve reproducibility, since the central claims are based on fits to temperature and power sweeps.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is honest and contains useful experimental material, but its central quantitative claim is not yet secured. I would encourage the authors to add an independent validation route or to reframe the kinetic-inductance numbers as indicative with a model uncertainty, rather than rejecting the work outright. The main risk is the internal inconsistency between the reported extraction and the authors' own caveat that the theoretical expressions are not justified for these films."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Punchline: this is a dataset paper, not a methodological advance. The measured kinetic-inductance fractions and TLS parameters for four grAl films are genuinely new and will be handy for people designing MKIDs, high-impedance resonators, and superinductors. But the headline numbers are less secure than the presentation suggests, and the authors' own caveat about Mattis-Bardeen deserves more weight than it gets.\n\nWhat the paper does well: the materials characterization in the appendices is thorough (AFM, TEM, XRD, profilometry all cross-checked). The comparison of rigid-shift extraction using simulated versus conformal-mapping geometric frequencies against the Mattis-Bardeen fits is sensible and clearly plotted. The TLS power dependence is a clean signature, and the extracted parameters (beta close to 1, F*delta_TLS0 around 1.5e-4) are consistent with standard resonator-TLS phenomenology. The authors are also honest about the alpha1/alpha2 discrepancy and about the mismatch with the simple sheet-resistance estimate.\n\nSoft spots: the core extraction assumes local-limit, single-gap Mattis-Bardeen surface impedance with Delta0 = 1.76 kTc. In Sec. V.1 the authors state that for 1 pH/sq the effective penetration depth is about 800 nm and that \"the use of the theoretical expressions to extract alpha is not justified.\" That sentence sits right after the sheet-inductance numbers it is supposed to support. The alpha1/alpha2 disagreement (1-22%) is not resolved; attributing it to unsataturated TLS is plausible, but they did not remove that contribution from the frequency-shift fit. The conformal-mapping inversion for Lk_sheet is a reframing of the same alpha1 fit, not an independent check. There is also a hard internal inconsistency: Table II lists Lk_sheet for grAl-1 R1/R2 as 3.480/3.502 pH/sq, while the text reports the grAl-1 average as 2.33 +/- 0.02. Those numbers do not match. No data or code are released, which makes checking the fits harder.\n\nWho this is for: device engineers who need rough grAl numbers for design choices, not someone needing a validated theory of disordered-superconductor electrodynamics. The qualitative trends--kinetic inductance grows with oxidation, TLS losses are present--are solid. The quantitative values should be treated as indicative until the Mattis-Bardeen assumption is justified or replaced, and the grAl-1 table error is corrected.\n\nRecommendation: send it to peer review. It is a careful experimental paper with a useful dataset. A serious referee should push for either validation of the MB extraction (e.g., independent cross-checks or data release) or a reframing that drops the unjustified sheet-inductance claims. Conditional acceptance at most.","headline":"A useful grAl dataset paper whose quantitative claims are undercut by the authors' own caveat about Mattis-Bardeen and by an internal table/text inconsistency.","tokens_in":18125,"tokens_out":3770,"would_cite":false,"duration_ms":41394,"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":"This paper shows that the kinetic inductance fraction of granular aluminium microwave resonators can be extracted from the temperature dependence of the resonance frequency, with Mattis-Bardeen fits giving values from 0.137 to 0.356…","keywords":["granular aluminium","kinetic inductance","kinetic inductance fraction","coplanar waveguide resonators","Mattis-Bardeen theory","two-level systems","superconducting microwave devices","disordered superconductors"],"falsifier":"Measure the kinetic inductance of the same four films by a method independent of Mattis-Bardeen, such as the magnetic-field dependence of the resonance frequency or a lumped-element tank-circuit inductance measurement, and compare with the values derived here; disagreement outside the quoted errors, for example $L^\\square_k$ differing by more than about 20% for grAl-3, would falsify the claim that the temperature-shift fit returns the true kinetic inductance.","tokens_in":17037,"feed_emoji":"🌡️","tokens_out":8609,"duration_ms":89731,"temperature":0.7,"pith_summary":"This paper reports microwave measurements of coplanar waveguide resonators fabricated from granular aluminium films with four different oxygen levels, and argues that the temperature dependence of the resonance frequency can be used to extract the kinetic inductance fraction. By fitting the relative frequency shift to Mattis-Bardeen theory, the authors obtain kinetic inductance fractions from 0.137 for pure aluminium up to 0.356 for the most oxidized film, and sheet kinetic inductances between 1.8 and 6.9 pH per square. The same measurements give the two-level-system contribution to microwave loss, about $1.5\\times10^{-4}$ in $F\\delta_{\\mathrm{TLS}}^0$, with no clear trend versus oxygen content. The central point is that this characterization route, combining temperature sweeps with Mattis-Bardeen fits and conformal-mapping or simulated geometric frequencies, is simple enough to be useful for designing MKIDs, high-impedance resonators, and superinductors.","feed_headline":"Kinetic inductance of granular aluminium read from resonance shifts","feed_subtitle":"Fits to Mattis-Bardeen theory yield fractions up to 0.36 and sheet inductances to 6.9 pH/sq across four films.","key_machinery":"The load-bearing object is the Mattis-Bardeen complex conductivity in the local limit, $\\sigma = \\sigma_1 - i\\sigma_2$, used to compute the surface impedance $X_s(T)$ of a thin film, together with the perturbation relation $\\Delta f/f_0 = -(\\alpha/2)(\\Delta X_s/X_s(0))$. The kinetic inductance fraction $\\alpha$ enters as the only free parameter in the frequency-shift fit, while $T_c$ is fixed to the measured value of each film. Geometric resonance frequencies are obtained independently from finite-element simulations and from conformal-mapping expressions for a coplanar waveguide, including the kinetic-inductance geometric factor used to convert $\\alpha_1$ into sheet inductance. For two-level systems, the machinery is the resonant TLS model relating $1/Q_i$ to photon number through $F\\delta_{\\mathrm{TLS}}^0$, $n_c$, $\\beta$, and a loss floor $\\delta_1$.","core_discovery":"The authors establish that the kinetic-inductance fraction $\\alpha$ of a granular aluminium film can be estimated from the measured temperature dependence of the resonance frequency, without needing an absolute calibration of the geometric resonance frequency. The relation $\\Delta f/f_0 = -(\\alpha/2)(\\Delta X_s/X_s(0))$, with $X_s$ computed from Mattis-Bardeen complex conductivity in the local limit, reproduces the measured shift from base temperature up to $T_c$. From that fit, $\\alpha_1 = 0.137 \\pm 0.005$ for aluminium, $0.201 \\pm 0.001$ for grAl-1, $0.291 \\pm 0.007$ for grAl-2, and $0.356 \\pm 0.008$ for grAl-3, giving sheet kinetic inductances $L^\\square_k$ of 1.8 to 6.9 pH per square after matching conformal-mapping expressions to $\\alpha_1$. The authors report that these Mattis-Bardeen values are systematically about 25% larger than rigid-shift estimates using simulated geometric frequencies and about 9% larger than those using conformal mapping, and they adopt $\\alpha_1$ as the estimate because the frequency shift is more reliable than the quality-factor shift. The two-level-system analysis yields $F\\delta_{\\mathrm{TLS}}^0 \\approx 1.5\\times10^{-4}$, $\\beta$ close to 1, and non-TLS losses limiting the internal quality factor to about $5\\times10^4$.","pith_inferences":["Beyond the paper, the systematic offset between Mattis-Bardeen and rigid-shift values of $\\alpha$ suggests a temperature-independent bias in one of the two methods; a decisive test would be to extract $\\alpha$ from the magnetic-field dependence of the same resonators, since that route does not rely on Mattis-Bardeen local-limit assumptions.","Beyond the paper, the lack of correlation between TLS loss and oxygen content hints that the loss floor is set by the granular microstructure, such as grain boundaries and the aluminium-oxide matrix, rather than by the oxide concentration; if so, annealing or altered growth protocols could reduce TLS without changing $\\alpha$.","Beyond the paper, the reported sheet inductances up to 6.9 pH per square and quality factors near $10^4$ could be fed directly into MKID sensitivity estimates; a concrete extension would be to fabricate MKID pixels from these same films and compare photon-noise-limited performance with predictions based on the extracted parameters."],"forward_implications":["The kinetic inductance fraction can be read from a single temperature sweep, avoiding the need for an absolute geometric-frequency calibration, which makes the method applicable to films whose geometric resonance frequency is hard to simulate.","Controlling oxygen content tunes sheet kinetic inductance from 1.8 to 6.9 pH per square, and growing thinner films would push these values higher.","The Mattis-Bardeen-derived $\\alpha$ is systematically larger than the rigid-shift $\\alpha$, so designs based on rigid-shift estimates may underestimate kinetic inductance by roughly 25%.","Two-level-system losses in granular aluminium are approximately $F\\delta_{\\mathrm{TLS}}^0 = 1.5\\times10^{-4}$, independent of oxidation level, so further improvement in quality factor will have to target non-TLS losses, which currently cap $Q_i$ near $5\\times10^4$.","The same characterization flow can be reused for other disordered superconducting films, since it only needs $S_{21}(T)$ and a film thickness."],"supporting_citations":[{"why":"Supplies the Mattis-Bardeen complex-conductivity formulas ($\\sigma_1$, $\\sigma_2$) used to model the surface impedance and fit the temperature-dependent frequency shift.","marker":"[57]"},{"why":"Provides the surface-impedance and TLS relations connecting resonance frequency and quality factor to kinetic inductance and two-level systems.","marker":"[64]"},{"why":"Gives the conformal-mapping expression for coplanar-waveguide inductance including kinetic inductance, used to convert $\\alpha_1$ into sheet inductance.","marker":"[62]"},{"why":"Corrects the geometric factor in the kinetic-inductance expression, a correction required for the sheet-inductance extraction.","marker":"[87]"},{"why":"Provides the coplanar-waveguide capacitance and inductance per unit length and the impedance equations used for design and geometric-frequency estimates.","marker":"[55]"},{"why":"Supplies the $S_{21}$ fitting formula used to extract resonance frequency and quality factor from the measured transmission data.","marker":"[82]"},{"why":"Gives the BCS estimate of sheet kinetic inductance from sheet resistance, which serves as the comparison baseline for the measured values.","marker":"[26]"},{"why":"Reviews the TLS loss model and gives the standard power-dependence fit used to extract $F\\delta_{\\mathrm{TLS}}^0$ and related parameters.","marker":"[67]"}],"fun_headline_variants":["Kinetic inductance up to 0.36 in granular aluminium films","Temperature shifts pin down kinetic inductance in grAl films","Resonance frequency shifts measure grAl kinetic inductance","Granular aluminium resonators reveal high kinetic inductance","Sheet kinetic inductance to 6.9 pH/sq in granular aluminium"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The extraction assumes that Mattis-Bardeen local-limit surface impedance with a single BCS gap of $\\Delta_0 = 1.76 k_B T_c$ describes strongly disordered granular aluminium films, even though the authors themselves note that for 1 pH per square the effective penetration depth reaches about 800 nm, where the theoretical expression is not justified.","fun_headline_variants_meta":{"raw":{"variants":["Kinetic inductance up to 0.36 in granular aluminium films","Temperature shifts pin down kinetic inductance in grAl films","Resonance frequency shifts measure grAl kinetic inductance","Granular aluminium resonators reveal high kinetic inductance","Sheet kinetic inductance to 6.9 pH/sq in granular aluminium"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000607,"raw_usage":{"total_tokens":2848,"prompt_tokens":984,"completion_tokens":1864,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":600,"completion_tokens_details":{"reasoning_tokens":1783}},"tokens_in":600,"tokens_out":1864,"duration_ms":13662,"temperature":1.0,"reasoning_tokens":1783,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T04:17:15.717305+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the kinetic inductance of the same four films by a method independent of Mattis-Bardeen, such as the magnetic-field dependence of the resonance frequency or a lumped-element tank-circuit inductance measurement, and compare with the values derived here; disagreement outside the quoted errors, for example $L^\\square_k$ differing by more than about 20% for grAl-3, would falsify the claim that the temperature-shift fit returns the true kinetic inductance.","supporting_citations":[{"cited_title":"Interplay between kinetic inductance, non-linearity and quasiparticle dynamics in granular aluminum MKIDs","cited_arxiv_id":"1810.12341","evidence_quote":"Supplies the Mattis-Bardeen complex-conductivity formulas ($\\sigma_1$, $\\sigma_2$) used to model the surface impedance and fit the temperature-dependent frequency shift."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the conformal-mapping expression for coplanar-waveguide inductance including kinetic inductance, used to convert $\\alpha_1$ into sheet inductance."},{"cited_title":"Probst, F","cited_arxiv_id":null,"evidence_quote":"Corrects the geometric factor in the kinetic-inductance expression, a correction required for the sheet-inductance extraction."},{"cited_title":"Borisov, D","cited_arxiv_id":null,"evidence_quote":"Provides the coplanar-waveguide capacitance and inductance per unit length and the impedance equations used for design and geometric-frequency estimates."},{"cited_title":"Megrant, C","cited_arxiv_id":null,"evidence_quote":"Supplies the $S_{21}$ fitting formula used to extract resonance frequency and quality factor from the measured transmission data."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the BCS estimate of sheet kinetic inductance from sheet resistance, which serves as the comparison baseline for the measured values."},{"cited_title":"Watanabe, K","cited_arxiv_id":null,"evidence_quote":"Reviews the TLS loss model and gives the standard power-dependence fit used to extract $F\\delta_{\\mathrm{TLS}}^0$ and related parameters."}],"review_version":1}