{"id":"65c8ff97-1df9-4924-a5ea-2dba9416397e","arxiv_id":"2505.00769","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Drive photons can re-enable quasiparticle-induced errors in gap-engineered transmons, and multi-photon absorption can create fresh quasiparticles that also corrupt the qubit.","lead":"This paper derives formulas for two ways quasiparticles, tiny excitations left behind when Cooper pairs break, destroy quantum information in a driven superconducting qubit. It matters for quantum computing hardware because it predicts readout and gate errors that persist even in qubits designed to block quasiparticle tunneling.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Gate-fidelity claim rests on an unperformed regularization of the Eq. (6) resonance; until a Floquet or nonlinear calculation is shown, the gate part of the 'fundamental limitation' claim is unverified.","rationale":"I read the paper in good faith. The derivations of the off-resonant rates are coherent, and the two mechanisms are physically plausible; the diagrammatic Supplement gives a systematic route to multi-photon amplitudes. The weakest point is not the existence of the processes but the quantitative bridge from the perturbative rates to the resonant gate-fidelity limit. The single-photon amplitude in Eq. (6) diverges at omega_d = omega_q, and the paper's proposed regularization by replacing omega_d - omega_q with E_C/hbar is asserted rather than derived. Since the abstract and Discussion claim a fundamental limitation on gates, this missing calculation is load-bearing for that part of the central claim. The readout and pair-breaking statements are less exposed because they are evaluated away from the resonance, though they are still order-of-magnitude estimates. This is precisely the reader's weakest_assumption, so I agree with it. I would not change the verdict: CONDITIONAL remains appropriate because the off-resonant rates are plausible and a well-specified numerical or analytical test can settle the gate part. No code or data are shipped, so the proposed Floquet computation would be a new check rather than a rerun of existing materials.","tokens_in":16764,"tokens_out":12921,"duration_ms":144078,"concrete_test":"Implement a Floquet-Fermi-Golden-rule computation of Gamma_{1->0} for the driven transmon, keeping the full cosine potential in a truncated phase or charge basis (e.g., E_J/E_C = 50, omega_q/2pi = 6 GHz, delta/hbar = 2pi x 10 GHz) and treating QP tunneling to first order. Evaluate at omega_d = omega_q and at omega_d = omega_q +/- E_C/hbar, and compare with Eq. (7) after the proposed replacement omega_d - omega_q -> E_C/hbar, and with the claimed gate bound beta x_QP/(E_C t_gate/hbar). If the Floquet rate differs by more than an order of magnitude, the gate-fidelity bound in the Discussion is unsupported; if it agrees, the regularization is validated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that microwave drives re-enable QP-induced transitions in gap-engineered transmons, with rates given by Eqs. (7), (10), (11), and (13). The most load-bearing soft spot is the gate-fidelity bound derived from the one-photon rate near resonance. Eq. (6) has an amplitude proportional to a omega_d^2/(omega_d^2 - omega_q^2), and Eq. (7) inherits the divergence through the ac-Stark shift. The paper states that the divergence 'should be regularized by the qubit non-linearity, omega_d - omega_q -> E_C/hbar' and that 'a detailed calculation yields' 1 - F >= beta x_QP/(E_C t_gate/hbar), but no such calculation appears in the main text or in the Supplement [Ref. 35]. This is not a cosmetic gap: a resonant drive dresses the qubit into Floquet states with quasienergy splittings set by the Rabi frequency and anharmonicity, so a principled calculation must replace the perturbative Golden-rule expression with a Floquet or exact-diagonalization treatment. Simply moving the pole by E_C/hbar is an uncontrolled guess unless verified. The readout estimate (16) and the pair-breaking estimate (13) are less exposed because they are evaluated away from resonance, but they also rely on order-of-magnitude choices such as alpha ~ 1 and a specific |omega_ac|/omega_q.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper develops a microscopic theory of two quasiparticle-induced decoherence mechanisms in a driven flux-tunable transmon qubit. In the first mechanism, an existing quasiparticle absorbs one or more drive photons while tunneling across the Josephson junction, with the qubit making a transition; the one- and two-photon rates are given by Eqs. (7) and (10) for relaxation and Eq. (11) for excitation. In the second mechanism, n drive photons break a Cooper pair at the junction and induce a qubit transition, with rate given by Eq. (13). The authors derive structure factors from the quasiparticle energy distribution, identify frequency thresholds and flux dependence, and use the rates to claim fundamental fidelity limits for readout and gates of gap-engineered qubits.","tokens_in":17033,"tokens_out":13059,"duration_ms":130796,"significance":"If the results hold, this is an important contribution: it identifies an operational constraint that is not captured by idle-time coherence studies of gap-engineered transmons, and it provides concrete, falsifiable predictions (thresholds, flux dependence, and rate scalings with drive power and quasiparticle density). The rates are obtained from a microscopic Hamiltonian with no parameters fitted to the data; the Supplement gives a general diagrammatic framework and generalizes the main-text expressions to arbitrary flux and to arbitrary qubit transitions. The authors also disclose and compare with an independent numerical study (Ref. [48]). The main caveat is that the quantitative gate-fidelity claim is not actually derived, as detailed in the major comments.","major_comments":[{"comment":"The asserted bound 1 − F ≥ β x_QP/(E_C t_gate/ℏ) is not derived. The one-photon amplitude, Eq. (6), diverges at ω_d = ω_q, and the text states that the divergence should be regularized by the qubit nonlinearity, ω_d − ω_q → E_C/ℏ, and that a detailed calculation yields the bound, but no such calculation appears in the main text or in the Supplement. The Supplement's perturbative framework (Eqs. (S1)–(S8)) is built on the harmonic unperturbed Hamiltonian and therefore cannot justify a strong or near-resonant drive result. The regularization is an unverified assumption at the operating point where the predicted rate is largest. Because the abstract and conclusions claim a fundamental limitation on gate fidelity, this is a load-bearing gap. Please either provide a Floquet or exact nonlinear calculation, or explicitly restrict the fundamental-limitation claim to off-resonant operation and readout.","section":"Discussion and conclusions, gate-fidelity paragraph"},{"comment":"The two-photon excitation rate Γ^(2)_{0→1} in Eq. (S12) diverges at ω_d = ω_q through D(2ω_d − ω_q), as acknowledged after Eq. (S13). The Supplement states that the accuracy of the derivation does not capture the difference between ω_02 and 2ω_q, but it does not provide a regularized expression. Since the main text invokes excitation processes as a source of leakage errors after Eq. (11), this is a second place where a resonant-drive claim outruns the supplied perturbative calculation. A unified Floquet treatment, or at least a regularization with an explicit justification, is needed.","section":"Supplement, Sec. IV (Eqs. (S12)–(S14))"}],"minor_comments":[{"comment":"As typeset, Eq. (9) is ambiguous: the ±1/2 appears not to be an exponent, although the subsequent claim S_+ ≫ S_- for ℏω − δΔ ≪ Δ requires the exponent reading. Please clarify the notation.","section":"Eq. (9)"},{"comment":"The estimate 1 − F ≥ 0.05 for a 60 GHz readout tone should state the assumed readout duration and ac-Stark shift so that the reader can reproduce the number from Eq. (13).","section":"Final discussion, 60 GHz readout estimate"},{"comment":"Ref. [35] is cited as 'See supplmental materials,' but it is not resolvable as a numbered bibliography entry; please fix the citation.","section":"References"},{"comment":"Please specify whether x_QP in Eq. (7) and in the structure factors of Eq. (8) refers to the global quasiparticle density or to the density in the low-gap lead, and define the normalization of n_L(ε)/x_QP explicitly.","section":"Eqs. (7) and (8)"},{"comment":"In Eq. (16), the statement 'we can bound the fidelity by' with 1 − F ≥ Γ t_RO is a lower bound on the infidelity; consider rewording to avoid implying an upper bound.","section":"Eq. (16)"}],"recommendation":"major_revision","confidential_remarks":"The paper is a strong Letter, and the central off-resonant rate calculations appear sound and well supported by the Supplement. The main risk is that the gate-fidelity claim is advertised in the abstract before the required nonlinear or Floquet calculation is shown. If the authors supply that calculation or visibly soften the gate part of the fundamental-limitation claim, I would support publication. The concurrent work in Ref. [48] is properly acknowledged."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nHere's my take on 2505.00769. The paper does something real: it works out, from a microscopic Hamiltonian, the rates for a microwave drive to re-enable quasiparticle (QP) tunneling in gap-engineered transmons, plus multi-photon Cooper-pair breaking at the junction. The one-photon (Eq. 7), two-photon (Eq. 10), and n-photon pair-breaking (Eq. 13) rates, including the flux dependence and the parity effect in the pair-breaking thresholds, are not in the prior literature as far as I can tell. The diagrammatic supplement is careful and reproduces the main expressions. The agreement with the concurrent numerical work (Ref. 48) is a good sign.\n\nThe soft spots are real but localized. The resonant divergence in Eq. (6) — and the gate-fidelity bound built on it — is the weakest point. The paper says the divergence 'should be regularized by the qubit nonlinearity' and that 'a detailed calculation yields' the bound, but no such calculation is shown in the main text or the supplement. That is a gap, and it matters because a resonant drive dresses the qubit into Floquet states; you can't just move the pole by E_C/ℏ without checking. So the gate bound is an estimate, not a derived result, and the word 'fundamental' in the abstract overreaches slightly. The readout bound (Eq. 16) and the pair-breaking estimate are order-of-magnitude estimates too, though they are evaluated away from resonance and are less exposed.\n\nI wouldn't call this a load-bearing flaw for the core rates. The derivations of Eqs. (7), (10), (11), and (13) are standard Fermi's Golden rule with controlled approximations (cold QPs, E_J >> E_C, small drive). No free parameters are fitted; the rates depend only on physical inputs. The citation pattern is normal; self-citations to Refs. 12, 13, 41 provide input parameters, not back-fitting.\n\nWho is this for? Anyone working on quasiparticle-induced decoherence in superconducting qubits, gap engineering, or high-frequency readout. The quantitative floors on readout and gate fidelity under drive are directly actionable for experiment. It deserves a serious referee. The referee should push for either a proper Floquet calculation for the gate bound or a clear statement that it's an estimate with unknown constant, and for a sensitivity check on α ~ 1.\n\nOverall: new, clearly derived results with a real soft spot in one application-level claim. Worth engaging with.","headline":"Solid, novel rate theory for microwave-assisted quasiparticle decoherence in gap-engineered transmons; the gate-fidelity bound is a stated but unverified estimate and should be labeled as such.","tokens_in":17621,"tokens_out":1841,"would_cite":true,"duration_ms":17923,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Microwave drives reopen a quasiparticle-induced decoherence channel in gap-engineered transmon qubits.","keywords":["superconducting qubits","transmon","quasiparticle tunneling","gap engineering","microwave drive","Cooper pair breaking","readout fidelity","decoherence"],"falsifier":"Measure the qubit relaxation rate $1/T_1$ of a gap-engineered transmon while sweeping drive frequency and drive amplitude, with the quasiparticle density $x_{\\mathrm{QP}}$ independently characterized. Process 1 predicts a turn-on at $\\omega_d+\\omega_q(\\Phi)=\\delta\\Delta/\\hbar$ and a rate linear in $|\\omega_{\\mathrm{ac}}|x_{\\mathrm{QP}}$; process 2 predicts additional step-like turn-ons at $\\omega_d=(2\\Delta-\\hbar\\omega_{if})/n$. Observing neither threshold, or rates that do not scale with $x_{\\mathrm{QP}}$, would falsify the central claim.","tokens_in":16565,"feed_emoji":"⚛️","tokens_out":12316,"duration_ms":106689,"temperature":0.7,"pith_summary":"The paper argues that a microwave drive reopens quasiparticle-induced decoherence in gap-engineered superconducting qubits, the devices designed to be insensitive to quasiparticles when idle. Two mechanisms are identified: an existing quasiparticle can tunnel across the Josephson junction while absorbing drive photons together with the qubit excitation, and several drive photons can break a Cooper pair at the junction, again changing the qubit state. The resulting rates depend on drive power through the ac-Stark shift, on drive frequency through sharp thresholds, and linearly on quasiparticle density. If the theory is right, readout and gate fidelities in these qubits have a quasiparticle-set floor that gap engineering alone cannot remove.","feed_headline":"Microwaves revive quasiparticle losses in gap-engineered qubits","feed_subtitle":"Two mechanisms tie readout and gate errors to quasiparticles even when the qubit is gap engineered.","key_machinery":"The machinery is a diagrammatic perturbation theory in which the microwave drive enters the Hamiltonian twice, once in the qubit phase potential and once in the quasiparticle tunneling operator. Expanding the tunneling operator in the drive amplitude $a$ and the qubit phase $\\hat\\phi$, and summing the interfering contributions, yields the renormalized single-photon amplitude of Eq. (6) with the factor $\\omega_d^2/(\\omega_d^2-\\omega_q^2)$; the same expansion generates two-photon amplitudes whose particle-hole channel interference is constructive at $\\Phi=0$, and $n$-photon Cooper-pair-breaking amplitudes at order $a^n$. The rates are assembled with Fermi's Golden rule and expressed through structure factors $S_\\pm[\\omega]$, which carry the dependence on the quasiparticle energy distribution and produce the threshold steps at $\\hbar\\omega=\\delta\\Delta$.","core_discovery":"The central claim is that a microwave drive can reanimate quasiparticle transitions in a flux-tunable transmon whose two superconducting gaps differ. For an existing quasiparticle, the single-photon-assisted relaxation rate is $\\Gamma^{(1)}_{1\\to0} = (|\\omega_{\\mathrm{ac}}|x_{\\mathrm{QP}}/4\\pi)\\{S_+[\\omega_d+\\omega_q(\\Phi)](\\omega_q^2(0)/\\omega_q^2(\\Phi)-1)+S_-[\\omega_d+\\omega_q(\\Phi)](\\omega_q^2(0)/\\omega_q^2(\\Phi)+1)\\}$, with structure factors $S_\\pm$ that turn on at $\\omega_d+\\omega_q(\\Phi)>\\delta\\Delta/\\hbar$ for cold quasiparticles. The two-photon variant, Eq. (10), is not suppressed at zero flux because its particle-hole interference is constructive. Independently, when $n\\hbar\\omega_d$ exceeds the sum of the gaps, the drive breaks a Cooper pair at the junction and drives qubit transitions at the rate of Eq. (13), with thresholds at $\\omega_d=(2\\Delta-\\hbar\\omega_{if})/n$. The paper concludes that microwave operations, readout and gates, face a quasiparticle-imposed fidelity bound: $1-F\\gtrsim\\alpha|\\omega_{\\mathrm{ac}}|t_{\\mathrm{RO}}x_{\\mathrm{QP}}$ for readout, and roughly $\\beta x_{\\mathrm{QP}}/(E_C t_{\\mathrm{gate}}/\\hbar)$ for gates.","pith_inferences":["A clean experimental discriminator is the predicted threshold turn-on of the relaxation rate at $\\omega_d+\\omega_q=\\delta\\Delta/\\hbar$; measuring $1/T_1$ versus drive frequency across this threshold in a device with characterized $x_{\\mathrm{QP}}$ would test process 1 directly.","Because the pair-breaking thresholds sit at $\\omega_d=(2\\Delta-\\hbar\\omega_{if})/n$ with parity-dependent behavior, the theory predicts discrete steps in qubit error rates as a function of drive frequency, a signature that could be used to infer the local gap sum $2\\Delta$.","The leakage rate $\\Gamma_{1\\to2}\\approx2\\Gamma_{0\\to1}$ suggests that repeated gates could populate non-computational states and then return through the same microwave-assisted channels, producing correlated, gate-frequency-dependent error patterns that a single-pulse fidelity measurement would miss.","Extending the calculation to finite drive amplitude beyond the weak-drive expansion would replace the resonant divergence in Eq. (6) with a width set by $E_C$, turning the gate bound into a quantitative prediction that could be compared with error-per-gate measurements."],"forward_implications":["Readout fidelity in a gap-engineered transmon has a floor $1-F\\gtrsim\\alpha|\\omega_{\\mathrm{ac}}|t_{\\mathrm{RO}}x_{\\mathrm{QP}}$; with $|\\omega_{\\mathrm{ac}}|t_{\\mathrm{RO}}\\sim100$ and post-radiation $x_{\\mathrm{QP}}\\sim10^{-4}$, this floor exceeds $10^{-2}$.","At zero flux bias, destructive interference suppresses the one-photon process, but the two-photon-assisted process remains active and can be comparable in rate at realistic readout parameters.","High-frequency readout cannot be raised without limit: for a 5 GHz aluminum qubit read out at 60 GHz, two-photon Cooper-pair breaking caps fidelity at $1-F\\gtrsim0.05$.","Increasing the gap difference pushes single- and multi-photon thresholds to higher drive frequencies, and quasiparticle traps near the junction reduce $x_{\\mathrm{QP}}$, so the error floor can be mitigated but not eliminated by hardware design.","The same mechanisms should appear in other microwave-driven superconducting circuits beyond transmon qubits, wherever drives couple to Josephson junctions and quasiparticles are present."],"supporting_citations":[{"why":"Supplies the gap-engineering baseline: a gap difference blocks cold quasiparticle tunneling in the absence of a drive.","marker":"[13]"},{"why":"Demonstrates that gap engineering suppresses radiation-impact effects in qubit arrays, the performance the driven mechanisms would undercut.","marker":"[27]"},{"why":"Provides the single-photon Cooper-pair-breaking result that the paper's Eq. (13) generalizes to n photons.","marker":"[36]"},{"why":"Sets the high-frequency readout setup whose multi-excitation-free range the pair-breaking mechanism limits.","marker":"[41]"},{"why":"Independently treats similar mechanisms numerically and is reported to agree with the analytic rates.","marker":"[48]"},{"why":"Defines the transmon model and its charge-insensitive spectrum on which the driven-qubit calculation is built.","marker":"[29]"},{"why":"Characterizes cold quasiparticles and the gap-engineering condition used in the threshold analysis.","marker":"[12]"}],"fun_headline_variants":["Drive reopens quasiparticle loss in gap-engineered qubits","Quasiparticle pairs from drive cap qubit fidelity","Microwave photons resurrect quasiparticle decoherence","Drive-assisted quasiparticle paths cap qubit fidelity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the weak-drive, weakly nonlinear oscillator treatment of the qubit remains valid in the near-resonant and strong-drive regimes where the predicted rates are largest; the single-photon amplitude diverges at $\\omega_d=\\omega_q$, and the paper defers its regularization by qubit nonlinearity to an unperformed calculation.","fun_headline_variants_meta":{"raw":{"variants":["Drive reopens quasiparticle loss in gap-engineered qubits","Quasiparticle pairs from drive cap qubit fidelity","Microwave photons resurrect quasiparticle decoherence","Drive-assisted quasiparticle paths cap qubit fidelity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000534,"raw_usage":{"total_tokens":2585,"prompt_tokens":982,"completion_tokens":1603,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":598,"completion_tokens_details":{"reasoning_tokens":1536}},"tokens_in":598,"tokens_out":1603,"duration_ms":12583,"temperature":1.0,"reasoning_tokens":1536,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T04:35:55.904186+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the qubit relaxation rate $1/T_1$ of a gap-engineered transmon while sweeping drive frequency and drive amplitude, with the quasiparticle density $x_{\\mathrm{QP}}$ independently characterized. Process 1 predicts a turn-on at $\\omega_d+\\omega_q(\\Phi)=\\delta\\Delta/\\hbar$ and a rate linear in $|\\omega_{\\mathrm{ac}}|x_{\\mathrm{QP}}$; process 2 predicts additional step-like turn-ons at $\\omega_d=(2\\Delta-\\hbar\\omega_{if})/n$. Observing neither threshold, or rates that do not scale with $x_{\\mathrm{QP}}$, would falsify the central claim.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the gap-engineering baseline: a gap difference blocks cold quasiparticle tunneling in the absence of a drive."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates that gap engineering suppresses radiation-impact effects in qubit arrays, the performance the driven mechanisms would undercut."},{"cited_title":"Houzet, K","cited_arxiv_id":null,"evidence_quote":"Sets the high-frequency readout setup whose multi-excitation-free range the pair-breaking mechanism limits."},{"cited_title":"Riwar and G","cited_arxiv_id":null,"evidence_quote":"Independently treats similar mechanisms numerically and is reported to agree with the analytic rates."},{"cited_title":"Catelani, R","cited_arxiv_id":null,"evidence_quote":"Defines the transmon model and its charge-insensitive spectrum on which the driven-qubit calculation is built."}],"review_version":1}