{"id":"428c3c05-0e6e-485d-86f7-680cc9adcded","arxiv_id":"1908.04257","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Adding lower-gap aluminum phonon-trap islands around granular aluminum resonators raises single-photon internal quality factors by up to threefold and halves quasiparticle burst rates.","lead":"Surrounding superconducting resonators with tiny aluminum islands that act as phonon traps improves their quality, reduces noise, and cuts quasiparticle bursts. If the effect is real, simple chip-level phonon engineering could reduce decoherence in quantum circuits.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central claim is confounded by a wafer-scale material gradient: chips with higher trap filling have lower grAl resistivity (6 to 4 mOhm cm, Suppl. III), which independently lowers kinetic inductance and can raise Qi, lower noise, and suppress burst detection at fixed quasiparticle density.","rationale":"The paper's central claim is that Al islands act as phonon traps and lower quasiparticle density. The data are internally consistent: Qi increases, noise decreases, and burst rate decreases with filling factor, while QP lifetimes remain unchanged. The phenomenological model is a reasonable extension of Ref. 43 and captures the shape of the trends. However, the weakest link is causal attribution. The authors openly report a resistivity gradient across the wafer but do not treat it as a confound. Because each filling factor is realized on a separate chip at a different wafer position, filling factor is not an independent variable. The reported 6-to-4 mOhm cm gradient changes kinetic inductance by roughly a third, directly affecting the observables used as evidence: higher Qi, lower noise, and fewer detectable bursts could all arise from reduced kinetic-inductance sensitivity rather than from phonon trapping. A same-wafer interleaved control, or a randomized block design, would settle the question. Without it, conditional acceptance is appropriate; the concern is load-bearing but not fatal, and the authors deserve credit for transparently reporting the gradient, which makes the proposed test feasible.","tokens_in":14306,"tokens_out":6272,"duration_ms":66800,"concrete_test":"Place several identical grAl resonators on a single 330-micrometer sapphire chip with F = 0 and F = 34% trap regions interleaved or randomized in a grid, and measure Qi, 1/f noise, and burst rate blind to position, while characterizing local resistivity and Tc at each device. If the F = 34% devices still show the factor-3 Qi gain, order-of-magnitude noise reduction, and factor-2 burst suppression, the material-gradient confound is ruled out.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The phonon-trap interpretation rests on the assumption that the only systematic difference between F = 0, 8.5, 19, and 34% chips is the areal coverage of Al islands. The supplement (Sec. III) documents that this is false: the same chips show a monotonic DC resistivity gradient from 6 mOhm cm at the F = 0 position to 4 mOhm cm at the F = 34% position, accompanied by a monotonic increase in resonant frequency. For a grAl film on the high-resistivity side of the Tc dome, lower normal-state resistivity implies both lower kinetic inductance (hence a smaller kinetic-inductance fraction) and a higher Tc/gap. The QP-induced loss rate and the fractional frequency shift per quasiparticle both scale with kinetic inductance fraction, so the F = 34% devices would show higher Qi, lower 1/f noise, and fewer bursts crossing a fixed detection threshold even if the quasiparticle density were identical. The global fits to Eqs. (1) and (2) with beta = 9 and Lambda = 0.18 are not an independent test: they impose the F-dependence of the phonon-trap model and can absorb any monotonic improvement. Because the trap filling factor is perfectly correlated with wafer position, the observed trends are consistent with the phonon-trap mechanism but do not uniquely establish it.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental study of granular aluminum (grAl) superconducting resonators surrounded by lithographically patterned thin-film aluminum islands intended to act as phonon traps. Four chips with filling factors F = 0, 8.5, 19, and 34% are measured in a reflection geometry at millikelvin temperatures. The authors report that as F increases, the single-photon internal quality factor Qi increases by up to a factor of three, the low-frequency noise amplitude decreases by about an order of magnitude, and the observed quasiparticle burst rate decreases by roughly a factor of two. They fit the Qi and burst-rate trends with a rate-equation model from their prior work, obtaining global parameters beta = 9 and Lambda = 0.18, and attribute the improvements to downconversion of pair-breaking substrate phonons by the lower-gap aluminum islands.","tokens_in":14665,"tokens_out":3802,"duration_ms":41292,"significance":"If the causal interpretation is correct, the work introduces a practical and scalable phonon-trapping strategy for improving high-kinetic-inductance superconducting circuits, complementing existing quasiparticle-mitigation approaches. The raw monotonic trends in Qi, noise, and burst rate are encouraging, and the model is physically motivated and visibly connected to prior work (Refs. 38 and 43). However, the central claim is weakened by a wafer-scale material gradient disclosed in the supplementary material: the chips with higher F also have lower grAl DC resistivity, which independently affects kinetic inductance and therefore the resonator's sensitivity to quasiparticles. Because the filling factor is perfectly correlated with wafer position, the data as presented do not uniquely establish the phonon-trap mechanism. The paper is significant as a demonstration of a promising technique, but the causal attribution needs additional support.","major_comments":[{"comment":"The causal claim is confounded by the wafer resistivity gradient reported in Suppl. III. The text states that the DC resistivity varies from 6 mOhm cm at label 0 (F = 0) to 4 mOhm cm at label 3 (F = 34%), and that the resonant frequency increases monotonically along the same direction. For a grAl film, lower normal-state resistivity implies lower kinetic inductance and a smaller kinetic-inductance fraction; the quasiparticle-induced loss rate and the frequency shift per quasiparticle both depend on this fraction. Thus higher-F chips would show higher Qi, lower 1/f noise, and fewer bursts crossing a fixed detection threshold even with an identical quasiparticle density. Since F is set by the lattice design and the chips come from one wafer with this gradient, the observed trends are consistent with, but do not uniquely establish, the phonon-trapping mechanism. The authors should either measure interleaved fill factors on the same wafer, or provide a quantitative correction for the kinetic-inductance difference and show that the phonon-trap model is still required.","section":"Supplementary Material III"},{"comment":"The fits to Eqs. (1) and (2) are not an independent test of the phonon-trap model. The functional forms and the underlying Rothwarf-Taylor rate equations are taken from the authors' prior paper (Ref. 43), and the parameters beta and Lambda are fitted to the very same data that the model is used to explain. A monotonic improvement with F from any cause, including the resistivity gradient, can be absorbed by these two-parameter forms. The paper reports beta = 9 and Lambda = 0.18 but does not report the fit uncertainties, the covariance, or a comparison against simpler alternative models (e.g., a linear or power-law dependence on F). Without such diagnostics, the agreement with Eqs. (1) and (2) does not provide mechanistic evidence for phonon trapping.","section":"Eqs. (1)-(2) and Supplementary Material II"},{"comment":"The experimental design has only one chip per filling factor and no repeated fabrication or cooldown for a given F. The missing B resonator at F = 34% further reduces the already small sample. The noise improvement by an order of magnitude is illustrated for a single pair of resonators (A, F = 0 vs. F = 19%), and the burst-rate reduction is based on ten-hour averages from single chips. The paper should address chip-to-chip and cooldown-to-cooldown variability, for instance by reporting multiple nominally identical chips or by explicitly bounding the systematic wafer-gradient contribution. This is necessary to support the claim that the filling factor, rather than wafer position, is the controlling variable.","section":"Fig. 2 and Fig. 3"}],"minor_comments":[{"comment":"The arXiv title page contains a typo: 'supercond ucting' should be 'superconducting'.","section":"Title (header)"},{"comment":"The square-root expression in Eq. (1) is easy to misread; adding a clear definition of the positive branch and its limiting behavior for F -> 0 and F -> 1 would improve readability.","section":"Eq. (1)"},{"comment":"The filling factor F is described as the fraction of substrate covered by traps, but later the text says 'traps covering as little as a third of one side of the substrate.' Please clarify whether F refers to one side only or to the total chip area, since the traps are patterned on one side.","section":"Main text, Fig. 1 caption"},{"comment":"In the derivation of Eq. (II.25), the order-of-magnitude estimate leading to the neglect of unity could be stated more explicitly; the notation tilde and the definition of the area AP are introduced in the preceding paragraphs, but a brief summary of the physical meaning of Lambda would help the reader.","section":"Supplementary Material II"},{"comment":"The fit curves in Figs. 2b and 3b are shown without confidence bands or parameter uncertainties; since the model is fitted to the same data, reporting the uncertainties of Qi0, beta, Gamma0, and Lambda is important for assessing the strength of the fit.","section":"Fig. 2b and Fig. 3b"}],"recommendation":"major_revision","confidential_remarks":"The wafer resistivity gradient is a genuine confound and the main reason this is not acceptable in its current form. The authors are transparent about the gradient in Supplement III, which is a point in their favor, and the raw trends are interesting. If they can provide an additional control (e.g., interleaved filling factors on the same wafer, or a quantitative model of the resistivity gradient's effect on Qi/noise/burst rates) and strengthen the statistical reporting of the fits, the paper could become suitable for publication. I do not see the problems as unfixable, but they are load-bearing for the central causal claim."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The phonon-trap idea is real, and this paper is a decent engineering demonstration. Applying lower-gap Al islands to high-impedance grAl resonators gives a monotonic improvement in single-photon Qi, a suppression of 1/f noise, and a reduction in QP burst rate as the trap filling factor grows. That scaling with F is new for this platform, and it extends earlier work from the same group and Karatsu et al. to a high-kinetic-inductance system. The raw data look careful: single-photon Qi, noise spectra, and long burst traces are all reported, and the trends are consistent with phonon trapping.\n\nThe soft spot is the confound the stress-test note identifies. The F=0 to F=34% chips lie along a wafer resistivity gradient from 6 to 4 mOhm cm, documented in Suppl. III. Lower resistivity means lower kinetic inductance and a larger gap, which by itself can raise Qi, reduce frequency noise, and reduce the number of bursts crossing a fixed threshold. Because trap filling factor and wafer position are perfectly correlated, the observed improvements do not uniquely fingerprint phonon trapping. The fits to Eqs. (1) and (2) do not rescue this: beta and Lambda are free parameters fitted to the same data, so agreement is not an independent test. This is a load-bearing confound, not a nitpick.\n\nMinor issues: the sample is small, resonator B at F=34% is missing, and the error bars are fitting uncertainties rather than device-to-device scatter. These are secondary. To the authors' credit, they report the resistivity gradient honestly in the supplement, but they use it to explain the frequency shifts rather than as a caveat on the QP results.\n\nWho is this for? Experimentalists working on high-impedance superconducting circuits and quasiparticle mitigation. It deserves a serious referee, but a referee should demand an interleaved or randomized trap layout on the same wafer, or some other direct control on material uniformity. Without that, the magnitude of the effect (factor 3 in Qi, order of magnitude in noise) is not established. The qualitative idea is plausible and likely correct, but this chip set cannot prove it. Send it to peer review, conditional on a controlled fabrication run.","headline":"A plausible and useful phonon-trap demonstration on grAl resonators, but the central causal claim is undercut by a wafer resistivity gradient that is perfectly correlated with trap filling factor.","tokens_in":15227,"tokens_out":2204,"would_cite":true,"duration_ms":25393,"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":"Surrounding superconducting resonators with lower-gap aluminum islands cuts quasiparticle noise tenfold and triples quality factors.","keywords":["quasiparticles","phonon traps","granular aluminum","superconducting resonators","kinetic inductance","internal quality factor","low-frequency noise","quasiparticle bursts"],"falsifier":"Fabricate $F = 0$ and $F = 34\\%$ chips from adjacent wafer regions with matched grAl resistivity, or remove the aluminum islands after the first measurement and remeasure the same resonators; if the quality-factor gain, noise drop, and halved burst rate disappear, the effect is a wafer-gradient artifact rather than phonon trapping.","tokens_in":14133,"feed_emoji":"🧊","tokens_out":4340,"duration_ms":40726,"temperature":0.7,"pith_summary":"This paper tries to show that the density of broken Cooper pairs (quasiparticles) that limit superconducting circuits can be lowered by coating part of the chip with a different superconductor that absorbs high-frequency phonons. The test devices are high-impedance resonators made of granular aluminum, which are unusually sensitive to quasiparticles. Surrounding them with thin-film aluminum islands, whose superconducting gap is smaller, improves single-photon internal quality factors by up to a factor of three, reduces low-frequency noise by an order of magnitude, and halves the rate of quasiparticle bursts. The authors attribute these gains to phonon trapping: energetic phonons in the sapphire substrate are downconverted by the aluminum islands before they can break Cooper pairs in the resonators. If the mechanism holds, it offers a materials-based route to quieter superconducting qubits and detectors.","feed_headline":"Phonon traps cut superconducting-circuit noise tenfold","feed_subtitle":"Aluminum islands with a lower gap drain excess quasiparticles, tripling resonator quality and cutting burst rates in half.","key_machinery":"The load-bearing element is the phonon trap: a lattice of thin-film aluminum islands with a superconducting gap smaller than that of the granular-aluminum resonator. High-energy phonons in the sapphire substrate break Cooper pairs in the aluminum, and the resulting quasiparticles relax by emitting lower-energy phonons that cannot break pairs in the higher-gap grAl; the trap thus downconverts pair-breaking phonons into harmless ones. The quantitative argument is a Rothwarf-Taylor rate-equation model for hot phonons and quasiparticles in the resonator and the traps, which yields the filling-factor scalings $1/Q_i = (1/Q_{i,0})\\sqrt{1+(\\beta F)^2 - \\sqrt{2(\\beta F)^2+(\\beta F)^4}}$ and $\\Gamma_B = \\Gamma_0 \\Lambda/(F+\\Lambda)$, with $\\beta = 9$ and $\\Lambda = 0.18$ fitted across all resonators.","core_discovery":"The central claim is that non-equilibrium phonons in the substrate, not just stray radiation or direct drives, are a major source of quasiparticles in high-kinetic-inductance circuits, and that these phonons can be removed by lower-gapped superconducting islands placed nearby. The paper demonstrates that increasing the surface coverage $F$ of 10 µm aluminum islands around granular-aluminum resonators monotonically improves all measured figures of merit, with the data following the phonon-trapping model of Eq. (1) using $\\beta = 9$ and Eq. (2) using $\\Lambda = 0.18$. Specifically, the single-photon internal quality factor rises by up to a factor of three, the $1/f$ noise amplitude drops by an order of magnitude, and the quasiparticle burst rate falls by about a factor of two. The islands are electromagnetically decoupled from the resonators, so the observed improvements are attributed to phonon frequency downconversion at the aluminum gap, well below the granular-aluminum gap.","pith_inferences":["A testable extension would be to etch away the aluminum islands after measurement on the same chip; if quality factor, noise, and burst rate revert to the $F = 0$ values, that would isolate the trap geometry from wafer-level material drift.","The reported wafer-scale gradient in grAl resistivity (6 to 4 mΩ cm) is a competing explanation for part of the trend, since lower resistivity changes kinetic inductance and quasiparticle sensitivity; an experiment with randomized trap placement across the wafer would settle how much of the gain is genuinely phononic.","The model's two fitted parameters, $\\beta$ and $\\Lambda$, could be extracted independently by varying trap thickness or material (for example, using an even lower-gap superconductor), which would test the predicted dependence of trapping efficiency on the gap difference.","One implication the authors do not develop is that phonon-trap lattices could be optimized as a phononic crystal, potentially engineering the substrate dispersion to further suppress the phonon energies that break pairs."],"forward_implications":["The fitted model extrapolates that a filling factor approaching $F = 1$ could raise single-photon internal quality factors by up to an order of magnitude, not just the factor of three demonstrated at $F = 34\\%$.","Because the burst rate follows $\\Gamma_B = \\Gamma_0 \\Lambda/(F+\\Lambda)$, complete trap coverage could in principle cut quasiparticle burst rates by $(1+\\Lambda)/\\Lambda \\approx 6$ relative to the untrapped chip.","The result identifies non-thermal substrate phonons as a controllable source of quasiparticle loss and noise in high-kinetic-inductance devices, pointing to phonon engineering as a complement to radiation shielding and quasiparticle trapping.","The same phonon-trapping strategy should transfer to other high-impedance elements, including protected qubits and kinetic-inductance detectors, where quasiparticles are especially damaging."],"supporting_citations":[{"why":"Supplies the phenomenological phonon-trapping model and the efficiency scaling with gap difference that this paper applies to grAl resonators.","marker":"Ref. 43"},{"why":"First demonstrated that surrounding kinetic inductance detectors with a lower-gapped superconducting film reduces quasiparticle bursts and noise equivalent power.","marker":"Ref. 38"},{"why":"Provides the granular-aluminum resonator baseline, measurement setup, and quasiparticle burst data that this work extends.","marker":"Ref. 26"},{"why":"Established quasiparticle generation-recombination as a source of low-frequency noise and characterized burst phenomenology in superconducting resonators.","marker":"Ref. 18"},{"why":"Reported comparable quasiparticle burst rates in untrapped aluminum devices, used as the comparison point for the burst-rate reduction.","marker":"Ref. 36"},{"why":"Gives the granular-aluminum resistivity versus critical-temperature dome used to justify the gap difference between grAl and aluminum.","marker":"Ref. 49"},{"why":"Provides the circle-fit method used to extract resonant frequency, internal quality factor, and coupling quality factor from reflection data.","marker":"Ref. 57"}],"fun_headline_variants":["Phonon traps quiet superconducting circuits","Aluminum islands drain quasiparticles from resonators","Lower-gapped phonon traps cut noise tenfold","Trap substrate phonons to boost resonator quality","Phonon sink cuts quasiparticle bursts by half"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper assumes the monotonic improvements come from phonon trapping, but the trapped chips were also made from lower-resistivity granular aluminum due to a wafer-position gradient, so material differences could account for part of the effect.","fun_headline_variants_meta":{"raw":{"variants":["Phonon traps quiet superconducting circuits","Aluminum islands drain quasiparticles from resonators","Lower-gapped phonon traps cut noise tenfold","Trap substrate phonons to boost resonator quality","Phonon sink cuts quasiparticle bursts by half"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000434,"raw_usage":{"total_tokens":2221,"prompt_tokens":967,"completion_tokens":1254,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":583,"completion_tokens_details":{"reasoning_tokens":1181}},"tokens_in":583,"tokens_out":1254,"duration_ms":12877,"temperature":1.0,"reasoning_tokens":1181,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:45:57.618111+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fabricate $F = 0$ and $F = 34\\%$ chips from adjacent wafer regions with matched grAl resistivity, or remove the aluminum islands after the first measurement and remeasure the same resonators; if the quality-factor gain, noise drop, and halved burst rate disappear, the effect is a wafer-gradient artifact rather than phonon trapping.","supporting_citations":[],"review_version":1}