{"id":"2545a88d-faae-4c06-a361-77842788b79c","arxiv_id":"2501.03211","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A detailed, high-yield fabrication process for flat vacuum-gap capacitors achieves mechanical quality factors up to 40 million and reproducible microwave and mechanical frequencies in superconducting circuit optomechanics.","lead":"EPFL researchers describe a cleanroom recipe for superconducting vacuum-gap capacitors whose moving aluminum drumheads stay flat at millikelvin temperatures, giving gap sizes around 150 nanometers. The process yields mechanical oscillators with quality factors up to 40 million and low microwave loss, and the same fabrication already underlies their published quantum optomechanics experiments.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The cryogenic gap-control claim rests on an unmeasured flattening step: the final 150 nm gap is never directly verified at 10 mK.","rationale":"The reader identifies the cryogenic flattening of the aluminum drumhead as the weakest assumption, and my reading agrees: the entire scalability argument depends on the final gap being set by lithographically defined trench depth and bottom-plate thickness, yet no measurement in this paper directly confirms that gap at 10 mK. The mechanical quality factors, yield statistics, and frequency reproducibility are strong and independently useful, and they support the process claim at the level of resonator performance. However, they do not test the absolute gap. I considered the absence of new microwave-loss measurements as an alternative concern, but the paper explicitly routes that claim to Ref. [3], so it is a presentation weakness rather than a load-bearing gap. The load-bearing gap is the unverified connection between the room-temperature fabrication geometry and the cryogenic electromechanical behavior. Because the reader's CONDITIONAL verdict already reflects this uncertainty, my read does not move the verdict; it reinforces the condition that direct low-temperature gap verification is needed before the precise-gap-control claim can be accepted.","tokens_in":26542,"tokens_out":6000,"duration_ms":67485,"concrete_test":"Use the open dataset [77] (or a fresh cooldown) to extract the actual cryogenic gap from the measured LC-resonator frequencies: for each device in Sec. 4, compute the total capacitance from omega_c = 1/sqrt(LC) with an independently simulated L, subtract the known parallel-plate area, and compare the inferred gap to the designed trench depth minus bottom-plate thickness across the chip. If the inferred gaps match the design value within the stated tolerance (e.g., +/-10 nm) over multiple chips, the flattening hypothesis is confirmed; if they deviate systematically or scatter beyond that tolerance, the precise-gap-control claim is falsified. This uses microwave data already collected and requires no new physics.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central fabrication claim is that CMP planarization plus differential thermal contraction makes the final gap exactly the trench depth minus the bottom-electrode thickness (Sec. 2, Fig. S19l). That mechanism converts a lithographic trench depth into a reproducible 150 nm gap and thereby underpins the stated <0.2% frequency disorder and scalable arrays. The evidence in this manuscript is indirect: room-temperature profilometry of buckled or compliant drums (Fig. 9), a low-temperature stress of sigma_Al = 350 MPa extracted by fitting drum frequencies to Eq. 2 (Fig. 13b), and batch-level frequency statistics (Fig. 14, Fig. S1). None of these measurements directly senses the plate-to-plate separation at cryogenic temperature. The mechanical frequency is insensitive to the gap, so the same data would be obtained if the drum flattened at a gap 10-20% different from design or with a small residual curvature. Moreover, the 'theory' curve in Fig. 13b uses the stress fitted from the same frequency data, so the agreement is not an independent check. A reported <0.2% frequency disorder constrains radius and stress uniformity, not the absolute gap. Therefore the load-bearing claim of precise, design-defined gaps at 10 mK is not yet established by the measurements shown in this manuscript. Companion papers [1-3] may contain supporting microwave data, but the present paper does not report it.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a fabrication process for superconducting vacuum-gap capacitors based on a planarized SiO2 sacrificial layer and a 'flat-geometry' aluminum drumhead. The central claims are that the process yields vacuum gaps on the order of 150 nm, mechanical quality factors up to 40 million, low microwave loss, and sufficiently low frequency disorder for scalable optomechanical arrays. The main supporting evidence is a detailed process flow with SEM/FIB cross-sections, room-temperature profilometry, ring-down measurements, and statistics over 119 devices from three batches, with quantum-level system results cited to companion papers [1-3].","tokens_in":26729,"tokens_out":4693,"duration_ms":44831,"significance":"If the gap-control claim holds, the process is a significant advance: it offers a design-defined vacuum gap with high mechanical coherence, small footprint, and batch reproducibility, which would enable large optomechanical lattices and qubit-mechanics interfaces. The manuscript's strengths include its unusually detailed process documentation, explicit parameter tables, SEM/FIB verification of each step, batch-to-batch statistics, and a public data repository (Zenodo), all of which are valuable to the community. However, two load-bearing claims are not directly substantiated in this manuscript: the absolute vacuum-gap size at 10 mK and the 'low microwave loss' assertion, which is only cited to a companion paper. These issues are addressable with additional data or rescoped claims.","major_comments":[{"comment":"The 'theory' curve is not an independent verification of the model. The stress value sigma_Al = 350 MPa is obtained by fitting the measured drumhead frequencies to Eq. (2), and the same fitted relation is then plotted as the red 'Theory' line. Reporting that the data scatter about this line is below 0.2% demonstrates self-consistency of the fit, not independent agreement with a theoretical prediction. To support the claim of 'perfect control and high reproducibility', an independent stress measurement (e.g., cryogenic wafer-curvature measurement) or a theory curve based on independently known parameters is needed.","section":"Sec. 4, Fig. 13b and Fig. 16c"},{"comment":"The claim that the final vacuum gap at 10 mK equals the trench depth minus the bottom-electrode thickness is not directly verified. The mechanical frequency in Eq. (2) has no dependence on the gap, so the frequency data, batch statistics, and room-temperature profilometry shown here cannot constrain the absolute plate separation at cryogenic temperature. The paper should either provide a direct cryogenic gap measurement (e.g., microwave-frequency versus designed-gap calibration, capacitance measurement, or low-temperature interferometry) or explicitly state that the exact gap is an assumption inherited from the flattening model, with a quantified uncertainty budget.","section":"Sec. 2 and Sec. 4"},{"comment":"The paper claims that the platform exhibits 'low microwave loss', but the actual microwave quality-factor data are not presented here. The text states that 'in [3] we report internal quality factors up to 10^5 ...' and does not include those measurements or a direct microwave characterization. Since the title and abstract list low microwave loss as a key property, the authors should either include the relevant microwave loss data or clearly re-scope the claim as a property demonstrated in the cited companion work, not in this fabrication study.","section":"Abstract and Sec. 4, microwave loss paragraph"}],"minor_comments":[{"comment":"The chemical reaction equation for HF vapor etching is garbled: 'SiO2+4HF−2+4C2H5OH+2 − →SiF4+2H2O+4C2H5OH' contains stray superscripts and an unbalanced notation; please rewrite it with correct stoichiometry and formatting.","section":"Sec. 3E, Eq. (1)"},{"comment":"The caption contains the typo 'compareed'; it should read 'compared'.","section":"Supplementary, Fig. S2 caption"},{"comment":"The sentence 'The frequency tolerance is normally defined by the desired mutual coupling between the identical LCs winch is typically designed greater than 100 MHz' contains the typo 'winch' instead of 'which'.","section":"Supplementary, Sec. 7C"},{"comment":"The reported mechanical quality factors, including the single-device Q = 4e7 value, are presented without error bars or a description of the fitting uncertainty. Please state the number of ringdown traces used, the fitting procedure, and how uncertainties in the extraction propagate to the quoted values.","section":"Sec. 4, Fig. 11b and Fig. 13a"},{"comment":"The inset of Fig. 5, showing the magnified final topography, lacks axis labels and units; adding them would make the 10-nm claim verifiable.","section":"Sec. 3C, Fig. 5"},{"comment":"The phrase 'theoretical fit' is imprecise. The red line is the 1/R scaling relation evaluated with sigma_Al = 350 MPa obtained from the fit in Fig. 13b, so it should be described as a fitted model, not an a priori theoretical prediction.","section":"Sec. 4, Fig. 16c caption"}],"recommendation":"major_revision","confidential_remarks":"This is a methods paper whose headline numbers (40 million Q, microwave loss, quantum control) largely derive from the companion papers [1-3]. The editor may wish to clarify whether the present manuscript is judged as a standalone fabrication report or as the methods companion to a published series. The referee's major comments are fixable by adding described measurements or by scoping the claims to what is actually shown; the underlying fabrication process appears credible and well documented."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know about this paper: it is a good, honest methods disclosure that should go to peer review, but the abstract oversells one point. The flat-geometry, CMP-planarized vacuum-gap capacitor recipe is described at a level that lets others reproduce it, with real statistics: 119 devices across three batches, mechanical Q up to 40 million, and frequency disorder below 0.2% in an array. The ring-down data and the SEM/FIB cross-sections are direct evidence, and the low-temperature yield-stress observation for thin aluminum films is a nice byproduct. This is the first time the full recipe is public, and that alone justifies publication.\n\nThe soft spots are real but not fatal. The stress parameter sigma_Al = 350 MPa is extracted by fitting drum frequencies to Eq. (2), then the same curve is plotted as 'theory' in Fig. 13b. That is not an independent check. The low microwave loss claim is cited from the group's prior Science paper rather than measured here; if you want to verify that part you have to go read Ref. [3]. And the abstract headlines the best single-device Q of 40 million against a 4.3 million average, which is a bit of cherry-picking; the distributions in Fig. 14 are the more honest picture.\n\nThe stress-test's central concern is fair: the final 150 nm gap at 10 mK is never directly measured. The paper infers it from room-temperature profilometry, the stress fit, and batch statistics. Mechanical frequency is insensitive to the gap, so those data would look the same if the drum were flattening at a gap 10–20% off design or with slight residual curvature. The authors' mechanism—differential thermal contraction making the plate uniformly tensile and flat—is plausible and supported by the stress estimate, but the absolute gap control is not directly verified here. That part of the claim is a reasonable inference, not a measured result.\n\nFor whom is this paper? Experimentalists working on circuit optomechanics, microwave electromechanics, or anyone building arrays of superconducting mechanical resonators. It is a solid process paper with reproducible enough detail that I would cite it. It deserves a serious referee; the fabrication community will get real value from the recipe, the troubleshooting sections, and the batch statistics, even if the authors should be pushed to tone down the micrometric claim and either measure the gap at low temperature or clearly mark it as inferred from frequency and stress data.\n\nMy recommendation: send it to review, and ask for a modest revision that separates fitted parameters from independent theory.","headline":"A genuinely useful process disclosure that deserves peer review; the cryogenic gap-control claim is partly inferred rather than directly measured, but the mechanical coherence and reproducibility results are real.","tokens_in":27392,"tokens_out":1496,"would_cite":true,"duration_ms":15651,"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 planarized-sacrificial-layer process makes aluminum vacuum-gap capacitors with 150 nm gaps, mechanical quality factors up to 40 million, and frequency disorder below 0.2 percent across arrays.","keywords":["vacuum-gap capacitor","circuit optomechanics","superconducting circuits","mechanical resonator","quality factor","fabrication process","chemical mechanical polishing","microwave loss"],"falsifier":"Cryogenic optical profilometry or capacitance-based gap extraction on a released chip: if a drum that buckled at room temperature is not flat at 10 mK, or if the measured gap differs from trench depth minus bottom-electrode thickness by more than the reported tolerance, the central assumption fails. A simpler check is a large-batch histogram: if mechanical frequency disorder across identical drums exceeds the reported 0.2 percent, the claimed lithographic gap control is not being achieved.","tokens_in":26230,"feed_emoji":"⚛️","tokens_out":7683,"duration_ms":72527,"temperature":0.7,"pith_summary":"This paper tries to establish that a specific fabrication sequence—etching a trench in silicon, filling it with a planarized silicon-dioxide sacrificial layer, depositing a suspended aluminum drumhead, and releasing it with HF vapor—turns the vacuum-gap capacitor from an irreproducible component into a scalable one. The central claim is that after cooldown to cryogenic temperature the aluminum drumhead becomes uniformly tensile and flat, so the capacitor gap is set lithographically by the trench depth rather than by random buckling. With this control, the authors report mechanical quality factors up to 40 million, an average of 4.3 million across 119 devices from three batches, microwave-frequency disorder below 0.2 percent in a 12-site array, and a compact footprint. A sympathetic reader would care because this is the missing fabrication reliability for circuit optomechanics: reproducible mechanical frequencies and low loss are exactly what is needed for mechanical arrays, qubit-mechanics interfaces, quantum memories, and tests of quantum mechanics at larger mass.","feed_headline":"Flat-drum process lifts mechanical Q to 40 million","feed_subtitle":"Sacrificial-layer planarization sets reproducible 150 nm gaps, unlocking arrays and qubit-mechanics interfaces.","key_machinery":"The load-bearing object is the flat-geometry vacuum-gap capacitor: a circular aluminum drum suspended over a bottom electrode inside a silicon trench, with no dielectric between the plates. The sequence that carries the argument is trench DRIE etching; bottom aluminum deposition and patterning; low-temperature oxide (LTO) sacrificial fill roughly six times the trench depth; CMP planarization to below 10 nm topography; ion-beam etch-back to the substrate; top aluminum evaporation; and HF-vapor release. Cooling to 10 mK converts room-temperature compressive stress in the aluminum into tensile stress, flattening the drum and making the gap equal to trench depth minus bottom-electrode thickness; the measured 350 MPa tensile stress then gives a dissipation-dilution factor $D_Q\\approx100$, which the authors use to explain the factor-1000 gap between the aluminum material quality factor and the measured mechanical quality factor.","core_discovery":"On its own terms, the paper's discovery is the \"flat-geometry\" vacuum-gap capacitor: a drumhead whose operating gap is determined by a lithographically etched trench and a planarized sacrificial layer, not by the uncontrolled buckling of a released thin film. The authors argue that at room temperature the aluminum top plate may buckle into a dome, but cooling to 10 mK induces enough tensile stress from differential thermal contraction that the plate flattens; the gap is then the trench depth minus the bottom-electrode thickness, about 150 nm. They support this with mechanical-frequency fits giving $\\sigma_{\\mathrm{Al}} = 350 \\pm 10\\%$ MPa, a <0.2% frequency-disorder figure in a 12-site array, release yields near 100% for gaps above 150 nm, and quality factors up to $4\\times10^7$, about 100 times higher than earlier vacuum-gap electromechanical devices. The same chips show microwave internal quality factors up to $2\\times10^6$ at high photon number, with argon-ion milling of the galvanic contact reducing cavity heating by an order of magnitude.","pith_inferences":["If the low-temperature flattening is as reliable as the batch statistics suggest, the same trench-plus-planarization route could be transferred to other superconducting metals, such as tantalum, or to sapphire once adhesion and planarization are solved, since the mechanism only requires a larger thermal contraction of the top plate than the substrate.","The reported frequency disorder is set by optical lithography; the supplementary material notes that e-beam or deep-UV lithography could reduce disorder to roughly 0.01-0.1%, which would make larger topological or many-body mechanical arrays viable.","The design deliberately exploits room-temperature buckling to aid release and then relies on cooldown for flatness, suggesting a quantitative trade-off between release yield (dome height) and operating gap (trench depth) that the paper does not fully formalize.","A direct cryogenic gap measurement, rather than inference from mechanical frequency, would convert the central assumption into a verified parameter and would be the natural next characterization step."],"forward_implications":["Gap size and mechanical frequency become design parameters set by lithography: trench radius sets the drum frequency ($\\Omega_m \\propto 1/R$) and trench depth sets the capacitor gap, so arrays of nearly identical electromechanical circuits can be fabricated without post-selection.","Mechanical coherence reaches the millisecond regime, making ground-state cooling, motion squeezing, and sideband-resolved quantum control routine on this platform, as the cited companion papers demonstrate.","Microwave loss is low enough for circuit QED uses: internal quality factors of $10^5$ at moderate photon number and $2\\times10^6$ at high photon number, compatible with vacuum-gap transmon qubits.","The process is scalable to lattices: the authors report a 12-site SSH array with frequency disorder below 0.2%, a 24-site topological lattice, and a mechanical hexamer with collective ground-state cooling.","Footprint shrinks compared with planar capacitors and with metalized dielectric membranes, because the vacuum gap stacks vertically instead of spreading the electrodes across the chip."],"supporting_citations":[{"why":"Supplies the headline platform results: mechanical quality factor of 40 million, ground-state cooling, motion squeezing, and millisecond-scale decoherence; ringdown data in this paper are adapted from it.","marker":"[1]"},{"why":"Demonstrates the 24-site topological optomechanical lattice used to argue that the fabrication process scales to arrays.","marker":"[2]"},{"why":"Reports the mechanical hexamer with collective quantum dynamics and the measured high internal microwave quality factor of the released capacitors.","marker":"[3]"},{"why":"The prior standard vacuum-gap electromechanical platform whose mechanical quality factor and unpredictable gap this work improves upon; it also establishes sideband cooling to the ground state as the benchmark.","marker":"[10]"},{"why":"Introduced vacuum-gap capacitors as low-loss components for superconducting resonant circuits, the design goal this process adopts.","marker":"[26]"},{"why":"Shows the conventional buckled-drum vacuum-gap electromechanical platform; the contrast with its few-micrometer unpredictable gaps motivates the flat geometry.","marker":"[37]"},{"why":"Supplies the dissipation-dilution formalism used to account for the high mechanical quality factors (dilution factor roughly 100) in stressed drum resonators.","marker":"[55]"},{"why":"Reports compact vacuum-gap transmon qubits, providing the comparison showing the microwave loss levels are compatible with qubit integration.","marker":"[53]"}],"fun_headline_variants":["Flat-drum vacuum gaps achieve 40M mechanical Q","Planarized gaps lift Q to 40M in vacuum-gap caps","Scalable flat-gap capacitors: low loss, 40M Q","Compact vacuum-gap caps: flat geometry, 40M Q","150nm gaps yield 40M Q and low microwave loss"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The process assumes that cooling from room temperature to 10 mK makes the aluminum drumhead uniformly tensile and flat enough that the final vacuum gap equals the trench depth minus the bottom-electrode thickness across the whole chip.","fun_headline_variants_meta":{"raw":{"variants":["Flat-drum vacuum gaps achieve 40M mechanical Q","Planarized gaps lift Q to 40M in vacuum-gap caps","Scalable flat-gap capacitors: low loss, 40M Q","Compact vacuum-gap caps: flat geometry, 40M Q","150nm gaps yield 40M Q and low microwave loss"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000893,"raw_usage":{"total_tokens":3882,"prompt_tokens":1006,"completion_tokens":2876,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":622,"completion_tokens_details":{"reasoning_tokens":2785}},"tokens_in":622,"tokens_out":2876,"duration_ms":19135,"temperature":1.0,"reasoning_tokens":2785,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:53:37.184918+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Cryogenic optical profilometry or capacitance-based gap extraction on a released chip: if a drum that buckled at room temperature is not flat at 10 mK, or if the measured gap differs from trench depth minus bottom-electrode thickness by more than the reported tolerance, the central assumption fails. A simpler check is a large-batch histogram: if mechanical frequency disorder across identical drums exceeds the reported 0.2 percent, the claimed lithographic gap control is not being achieved.","supporting_citations":[{"cited_title":"Cicak, D","cited_arxiv_id":null,"evidence_quote":"Introduced vacuum-gap capacitors as low-loss components for superconducting resonant circuits, the design goal this process adopts."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows the conventional buckled-drum vacuum-gap electromechanical platform; the contrast with its few-micrometer unpredictable gaps motivates the flat geometry."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the dissipation-dilution formalism used to account for the high mechanical quality factors (dilution factor roughly 100) in stressed drum resonators."},{"cited_title":"Zemlicka, E","cited_arxiv_id":null,"evidence_quote":"Reports compact vacuum-gap transmon qubits, providing the comparison showing the microwave loss levels are compatible with qubit integration."}],"review_version":1}