{"id":"da7dab29-4379-48cf-b545-ac8b03b41069","arxiv_id":"2507.18672","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A numerical extrapolation method estimates the energy participation ratio of nanoscale surface oxides in transmon qubits, showing that larger pad gaps and substrate trenches lower RF surface losses.","lead":"The authors simulate microwave surface losses in transmon qubit antennas and propose a way to estimate losses in the ultra-thin oxide layers that dominate qubit decoherence. The method extrapolates simulations from thicker to real nanoscale oxides and suggests that larger pad spacing and substrate etching reduce surface loss.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central linear extrapolation from 20–30 nm to 3–5 nm oxide thickness is never validated at the target thickness, and the manuscript itself concedes sidewall/corner linearity is not always accurate.","rationale":"The reader's weakest assumption is exactly the load-bearing issue: the asymptotic method is only exercised at thicknesses above the HFSS meshing limit, and the extrapolation to the physical 3–5 nm range is asserted rather than demonstrated. The manuscript's own caveat about sidewall corners and the missing reference [6] for experimental agreement reinforce this concern rather than resolving it. I agree with the reader that the paper is plausible and useful but not fully validated, so the CONDITIONAL verdict stands unchanged. The proposed concrete test—directly computing EPR at 5 nm with an independent solver that can handle the singular corner—would settle whether the extrapolation error is acceptable or whether the Table 1 values and design ranking need revision.","tokens_in":5247,"tokens_out":8528,"duration_ms":92258,"concrete_test":"Using an independent 2D electrostatic solver that resolves the corner singularity (e.g., a high-order FEM with a graded mesh down to 1 nm, or an analytic wedge expansion), compute EPR(t) for t = 3, 5, 10, 20, and 30 nm for the V1, V2, and V3 sidewall cross-sections with ε = 10 and the same geometry as in Fig. 11. Compare the direct t = 5 nm values with the linear extrapolation obtained from the t = 20 and t = 30 nm points. If the relative difference exceeds about 10%—comparable to the ~13% V2–V3 EPR gap in Table 1—then the extrapolation step is not validated and the design ranking may change, requiring a revision or explicit error estimates.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the full-oxide EPR is affine in thickness down to the real 3–5 nm range, because HFSS is only run at 25 nm (and up to 100 nm in Fig. 11). The manuscript explicitly concedes (section 'Study of EPR Variation...', discussion of Fig. 11) that for a side wall with sharp corners 'this assumption is not always accurate due to local field enhancements.' Fig. 8 confirms the electrostatic singularity |E| ~ ρ^{-1/3} near a right-angle corner; in the singular limit the corner contribution to EPR scales as a fractional power of thickness (e.g., ∫|E|^2 dV over a region of linear size t gives t^{5/3}), not as t^1. Thus a straight-line fit to data at 20–30 nm can misestimate the t = 3–5 nm value even if the curve looks linear above 20 nm. The paper provides no direct simulation at the target thickness, no error bars on the extrapolation, and the claimed experimental agreement is uncited: reference [6] appears in the caption of Fig. 14 but is absent from the reference list. Since Table 1 EPR values drive the design conclusions (V1 vs V2 vs V3 ranking and the substrate-etching recommendation), an unquantified extrapolation error is load-bearing. The concern would be mitigated if the corner contribution is a small fraction of total EPR, but that fraction is not reported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents finite-element eigenmode simulations (ANSYS HFSS) of three planar transmon antenna designs, with the goal of estimating microwave surface losses arising from the natural oxide layers on the superconductor. Because HFSS meshing is difficult at the physical 3–5 nm oxide thickness, the authors propose an asymptotic method: simulate oxide thicknesses in the 25–100 nm range, verify that the energy participation ratio (EPR) becomes linear above roughly 20–30 nm, and linearly extrapolate to the few-nanometer scale. They report EPR values for three transmon variants (V1, V2, V3) with different pad spacings and aspect ratios, study the dependence of the EPR on oxide permittivity and substrate trench depth, and conclude that larger pad spacing and substrate etching reduce RF surface loss.","tokens_in":5566,"tokens_out":6922,"duration_ms":69757,"significance":"If the linear extrapolation is valid, the paper offers a practical methodology for estimating surface-loss-limited coherence in planar transmons without meshing sub-nanometer dielectric layers. The paper has several genuine strengths: a convergence study favoring 2nd-order mesh elements; a moving-mesh approach that keeps the geometry consistent across oxide thicknesses; an electrostatic benchmark for the field singularity at sharp corners; and EPR values that are direct simulation outputs rather than parameters fit to the conclusions. The design trends (lower EPR for larger pad spacing and for substrate trenching) are plausible and actionable. However, the central extrapolation from 20–30 nm to 3–5 nm is not validated at the target thickness, and the claimed agreement with experiment is unverifiable because reference [6] is missing. These gaps affect the quantitative design guidance in Table 1 and therefore need to be addressed before the central claim can be accepted.","major_comments":[{"comment":"The extrapolation from 20–30 nm to 3–5 nm oxide thickness is the load-bearing step of the proposed method, but it is not validated at the target thickness. The manuscript itself acknowledges that for a side wall with sharp corners 'this assumption is not always accurate due to local field enhancements' (discussion of Fig. 11). The electrostatic comparison in Fig. 8 shows |E| ~ ρ^{-1/3} near a right-angle corner, and for such a singular field the corner contribution to the EPR is not guaranteed to be affine in thickness; it can scale as a fractional power. No simulation is run at 3–5 nm, no error bars are given on the extrapolated values, and no estimate is provided of the fraction of EPR coming from corner regions. Because Table 1 and the design ranking (V1 vs V2 vs V3, trench etching) rest on these extrapolated numbers, this is a load-bearing gap. A direct simulation at the physical thickness for at least one geometry, or an analytic corner correction with a quantified error bound, would resolve the concern.","section":"Study of EPR Variation in Transmons with Different Antenna Geometries (first Fig. 8, Fig. 11, Table 1)"},{"comment":"The claim of agreement with experiment cannot be checked because reference [6], cited in both Fig. 14 and the Conclusion, does not appear in the reference list. Please add the reference or remove the claim. In addition, the Conclusion states that the advantage of large pad spacings and substrate etching is 'confirmed in experiments,' but the only cited experiment concerns the etching effect; no experimental comparison is shown for the pad-spacing trend.","section":"Conclusion and Fig. 14"},{"comment":"The EPR values in Table 1 are extrapolated estimates, but no uncertainty or confidence interval is reported. Since these values drive the quantitative comparison among V1, V2, and V3, even a rough uncertainty estimate (for example, from the scatter in Fig. 11 or from the difference between linear and alternative fits) should be included so that the reader can judge whether the ranking is statistically meaningful.","section":"Table 1"}],"minor_comments":[{"comment":"The figure numbering is inconsistent: two different figures are labeled Fig. 8 (the electrostatic comparison and the mesh-convergence study).","section":"Throughout"},{"comment":"Equation (1) is garbled in the typesetting; the intended relation appears to be Qi = 1/(epri · tan δ), but the expression should be rewritten clearly.","section":"Eq. (1)"},{"comment":"The header of Table 1 is ambiguous: the G-factor columns are labeled with '(ε = 10, 5 nm)' and '(ε = 10, 4 nm)', although the G-factor defined in Eq. (4) does not depend on permittivity or oxide thickness. Please clarify what these headers mean.","section":"Table 1"},{"comment":"The text alternates between 'interpolation' and 'extrapolation' for the step from 25–100 nm to a few nanometers; this should be made consistent, since the procedure is an extrapolation outside the simulated thickness range.","section":"Abstract/Conclusion"},{"comment":"The Conclusion says the method reduces oxide thickness 'from hundreds to tens of nanometers,' but the simulations in Fig. 11 run up to 100 nm; please align the wording with the actual simulated range.","section":"Conclusion"}],"recommendation":"major_revision","confidential_remarks":"The core methodology is potentially useful, and the paper has good internal checks (convergence, electrostatic benchmark, moving mesh). The main risk is the unvalidated linear extrapolation at the physical thickness, which all numerical results depend on. If the authors can provide at least one validation at 3–5 nm or a quantified corner-correction analysis, the paper would be substantially stronger. The missing reference [6] should be fixed as well."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a serious referee, but the referee should make them close one obvious gap before publication.\n\nWhat is actually new: they propose an asymptotic workaround for a real nuisance—HFSS cannot mesh a 3–5 nm oxide conformally on a 200 nm film—so they simulate at 25–100 nm and extrapolate EPR linearly downward. The three pad-layout comparisons (V1/V2/V3, spacings 20/70/150 µm) and the substrate-etching curves are new numerical results, and the moving-mesh idea (reassigning virtual objects rather than regenerating geometry) is a sensible practical trick. The field-map comparisons to the electrostatic ρ^{-1/3} corner singularity give a useful sanity check on the FEM.\n\nThe paper does not oversell the physics. It does not claim a new decoherence mechanism or a new analytic formula. The EPR values are direct simulation outputs, not fitted to conclusions. The authors correctly note that the linear extrapolation is not guaranteed near sharp corners; Fig. 11 even shows the sidewall-plus-corner EPR deviating from the dotted linear line at small thicknesses, which is exactly where they want to use it.\n\nThe soft spots, in proportion: (1) The central extrapolation is empirical. The manuscript concedes that for side walls with sharp corners 'this assumption is not always accurate.' The stress-test note's fractional-power argument (corner contribution scaling like t^{5/3} rather than t^1) is not in the paper, but the paper's own Fig. 11 shows a kink at low thickness. That makes the Table 1 numbers at 3–5 nm uncertain, and Table 1 drives the V1 vs V3 design conclusion. (2) The claimed agreement with experiment cites [6], but reference [6] is missing from the list. The text says 'good agreement with the experiment [6]' but no experiment is actually shown. That is an incomplete citation, not necessarily a fatal flaw, but it must be fixed. (3) No error bars on the EPR values, and no convergence data at the extrapolated thickness. (4) Minor: the paper treats the oxide and substrate permittivity both as 10 in Table 1, so the numbers are conditional on that choice; that is disclosed, so it is fine.\n\nOn the citation pattern: self-citations are reasonable here—their own group's experimental work on transmon performance is directly relevant, and citing it is not a problem when the cited results are externally verifiable. The missing [6] is the only actual citation issue.\n\nThis paper is for people doing transmon or CPW loss engineering, especially in the SRF community. It gives concrete design guidance (large pad spacing, substrate etching) and a repeatable workflow. It deserves peer review. I would send it out, with a request that the authors report corner/sidewall/top EPR fractions and provide at least one direct check at a true few-nm thickness, or a quantitative statement of extrapolation uncertainty. Those are addressable revisions, not core-argument killers.\n\nRecommendation: accept for peer review with requested major revision; the method is plausible, the results are new, and the missing validation is fixable.","headline":"Practical extrapolation scheme for transmon oxide-loss EPR, with one load-bearing gap: the linear-in-thickness assumption is never checked at the real 3–5 nm target.","tokens_in":6093,"tokens_out":789,"would_cite":true,"duration_ms":10376,"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 claims that RF surface losses in planar transmon qubits can be accurately predicted by simulating oxide layers tens of nanometers thick and linearly extrapolating the energy participation ratio down to the real few-nanometer…","keywords":["transmon qubit","RF surface loss","energy participation ratio","oxide thickness extrapolation","finite element method","substrate etching","T1 coherence time"],"falsifier":"Fabricate two or more of the studied transmon designs with the same measured oxide thickness and permittivity, measure their surface-loss-limited T1, and check whether the ratio of measured losses matches the ratio of extrapolated EPR values in Table 1; alternatively, run a solver capable of directly meshing a 3-5 nm oxide on the same geometry and compare its EPR to the extrapolated numbers for the same corner radii.","tokens_in":5080,"feed_emoji":"⚛️","tokens_out":4351,"duration_ms":47795,"temperature":0.7,"pith_summary":"This paper tries to establish that a practical numerical method can estimate the dominant dielectric surface loss in planar transmon qubits, without directly meshing the true 3-5 nm oxide layer. The proposed asymptotic method simulates oxide layers of order 20-30 nm and linearly extrapolates the energy participation ratio (EPR) down to nanometer thicknesses, claiming good accuracy above a threshold thickness. Applying this method to three transmon antenna designs, the paper finds that larger pad spacing and substrate etching reduce surface losses, with EPR values listed in Table 1. These results matter because surface oxide loss is widely considered the main limit on transmon coherence time, so reliable EPR estimates translate into concrete design guidance.","feed_headline":"Thick-oxide simulations predict thin-film qubit loss","feed_subtitle":"Extrapolating energy participation from 25 nm down to real oxide ranks transmon designs by surface loss.","key_machinery":"The central object is the energy participation ratio, EPR_i, defined as the fraction of total stored electromagnetic energy that resides in the i-th lossy dielectric layer, computed from the electric field distribution of the qubit eigenmode. In practice, the fields come from a finite-element eigenmode solver with fields normalized to 1 J of stored energy, with an explicit correction for energy stored in lumped inductors. The method that carries the argument is asymptotic thickness extrapolation: a moving-mesh scheme reuses one fixed mesh to simulate oxide thicknesses from tens to one hundred nanometers, and a linear extrapolation maps those EPR values down to the physical 3-5 nm range. Second-order mesh elements are identified as necessary for a converged field solution near sidewalls and corners.","core_discovery":"The central claim is that the energy participation ratio of a natural surface oxide, the quantity that controls dielectric loss, can be obtained by a two-step asymptotic procedure when direct simulation of the true oxide thickness is numerically impossible. First, simulate a thicker oxide layer, tens of nanometers, with a converged second-order finite-element mesh; second, linearly extrapolate the EPR to the real few-nanometer thickness. The paper demonstrates that for flat surfaces the EPR is exactly linear in thickness, and that for geometries with sharp corners the linear regime is reached above roughly 20-30 nm, the thickness range used for the simulations. Using this method on three transmon layouts with pad spacings of 20, 70, and 150 micrometers, the paper finds that the design with the largest pad spacing and rounded pads has the lowest oxide EPR, and that etching a trench into the substrate suppresses the field enhancement at the lower sidewall corner.","pith_inferences":["A natural testable extension is to run the same two-step extrapolation on a geometry with deliberately varying corner roundness, to map where the 20-30 nm linear threshold shifts and whether the method needs a higher simulation thickness for sharper corners.","The paper's method could be applied to other ultra-thin lossy interfaces, such as the tunnel barrier of the Josephson junction itself, where similar mesh-size obstacles prevent direct simulation.","If the extrapolated EPR ranking were combined with independent TLS-loss measurements on identically fabricated devices, the predicted order of T1 across the three designs could be checked without needing a solver that resolves the true oxide thickness.","The etching benefit is shown for one transmon variant; the expectation that it transfers to other pad spacings is plausible but remains an extrapolation from a single geometry."],"forward_implications":["If the extrapolation method is accurate, the EPR values in Table 1 directly translate a measured oxide loss tangent into a surface-loss-limited T1 for each transmon design.","Larger pad spacing reduces oxide EPR by roughly a factor of two between the 20 and 150 micrometer pad designs, giving a quantitative rationale for wide-pad transmon layouts.","Substrate etching to a depth of about 10 nm substantially reduces EPR, with diminishing returns once the trench depth approaches the film thickness, so shallow etches capture most of the benefit.","The same asymptotic procedure can be applied to the aluminum conductor oxide and to other planar superconducting circuit geometries without re-meshing the true nanoscale film.","The observation that EPR does not simply scale inversely with permittivity means that material-parameter uncertainty in thin oxide films must be propagated through EPR calculations, not assumed away."],"supporting_citations":[{"why":"Supplies the eigenmode solver and mesh infrastructure used for all field and EPR computations.","marker":"[1]"},{"why":"Provides the electrostatic power-law edge-field behavior used to validate the HFSS field enhancement near sharp corners.","marker":"[2]"},{"why":"Defines the three transmon variants with different pad spacings and aspect ratios that are simulated.","marker":"[3]"},{"why":"Provides the maximum permittivity value of amorphous niobium pentoxide, setting the upper end of the permittivity sweep.","marker":"[4]"}],"fun_headline_variants":["Thick-oxide trick predicts thin-film qubit loss","Extrapolation ranks transmon designs by surface loss","Linear EPR scaling from thick to thin oxides","Two-step method finds real oxide loss in qubits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that linearly extrapolating the energy participation ratio from simulated oxide thicknesses of 20-30 nm down to the real 3-5 nm thickness remains accurate near sidewall corners, where the paper itself acknowledges that local field enhancements make the linear assumption not always accurate.","fun_headline_variants_meta":{"raw":{"variants":["Thick-oxide trick predicts thin-film qubit loss","Extrapolation ranks transmon designs by surface loss","Linear EPR scaling from thick to thin oxides","Two-step method finds real oxide loss in qubits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00024,"raw_usage":{"total_tokens":1503,"prompt_tokens":912,"completion_tokens":591,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":528,"completion_tokens_details":{"reasoning_tokens":528}},"tokens_in":528,"tokens_out":591,"duration_ms":6434,"temperature":1.0,"reasoning_tokens":528,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:14:03.946662+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fabricate two or more of the studied transmon designs with the same measured oxide thickness and permittivity, measure their surface-loss-limited T1, and check whether the ratio of measured losses matches the ratio of extrapolated EPR values in Table 1; alternatively, run a solver capable of directly meshing a 3-5 nm oxide on the same geometry and compare its EPR to the extrapolated numbers for the same corner radii.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the eigenmode solver and mesh infrastructure used for all field and EPR computations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the electrostatic power-law edge-field behavior used to validate the HFSS field enhancement near sharp corners."},{"cited_title":"Disentangling the Impact of Quasiparticles and Two-Level Systems on the Statistics of Superconducting Qubit Lifetime","cited_arxiv_id":null,"evidence_quote":"Defines the three transmon variants with different pad spacings and aspect ratios that are simulated."},{"cited_title":"Electronic, dielectric, and optical properties of the b phase of niobium pentoxide and tantalum pentoxide by first-principles calculations","cited_arxiv_id":null,"evidence_quote":"Provides the maximum permittivity value of amorphous niobium pentoxide, setting the upper end of the permittivity sweep."}],"review_version":2}