{"id":"dedc3ce0-e0d3-4ec2-bddf-b494176d7cda","arxiv_id":"1908.11856","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":2,"one_line_summary":"Three direct measurements of DC and AC flux crosstalk between tunable transmons are presented, along with a derived relation between AC crosstalk and parametric CZ gate infidelity.","lead":"This paper introduces three methods to directly measure magnetic flux crosstalk between tunable transmons, with sensitivities as fine as 0.001%, and derives a formula linking AC crosstalk to the infidelity of a parametric CZ gate. The authors validate the formula with quantum process tomography on a 16-qubit chip.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Re-expanding Appendix B gives the cross term + (1/10)(δωτ)(δΔτ), not − as in Eq. (B4); the Eq. (6) coefficient may be 11/80 rather than 27/80 for CZ02, shifting the 0.2% threshold.","rationale":"The three measurement methods (DC resonator, DC qubit, AC crosstalk) are well described, agree where they overlap (3.52±0.05% vs 3.55±0.04%), and are valuable engineering tools regardless of Eq. (6); this independent support is real. The load-bearing claim, however, is the conversion from crosstalk to a quantitative CZ infidelity. My concern is narrower than the reader's: I do not dispute the Ramsey-to-gate transfer or the neglect of δg_eff as the primary risk; instead, the displayed Appendix B derivation itself appears to contain an algebraic sign error that changes the prefactor. If the correct prefactor for the realized CZ02 gate is 11/80, the predicted infidelity is 41% of Eq. (6)'s value, and the experimental validation would need to be re-examined: the theory curves in Fig. 6(c) would lie below the data for worst-case phase, not above. Because the reader's CONDITIONAL verdict already requires scrutiny of the validation, I would keep the verdict unchanged but add the re-derivation of Eq. (B4) as the decisive test. I did not find grounds for rejecting the measurement sections, and no ad hominem is intended; this is a mathematical check on a displayed equation.","tokens_in":17409,"tokens_out":29011,"duration_ms":239680,"concrete_test":"Use a computer algebra system to expand |tr(U_CZ†U)|² exactly from Eq. (B3), or simulate the two-level Hamiltonian H=δΔ|Q⟩⟨Q|+(g+δg)(|11⟩⟨Q|+h.c.) for τ=π/g and compute r for α=β=0.01 and α=−β=0.01. If the cross term is +αβ/10, Eq. (B4) has a sign error; then determine δΔ for the gate's actual resonance by adding δω to the |11⟩→|Q⟩ energy gap and recompute the coefficient for CZ02/CZ20. This single check confirms Eq. (6) or replaces 27/80 with 11/80 (or vice versa), which also revises the 0.2% crosstalk threshold.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Equation (6) is the quantitative bridge between measured AC crosstalk and simultaneous-CZ infidelity, and it rests entirely on Appendix B. Taking the displayed rotated unitary U=diag{1,e^{-iδωτ},1,e^{-iδωτ}U11} with U11 from Eq. (B3), a straightforward second-order expansion of r=(d²−|tr(U_CZ†U)|²)/(d²+d) gives r=(1/5)α²+(1/10)αβ+(3/80)β²+(1/5)γ², with α=δωτ, β=δΔτ, γ=δgτ. Equation (B4) instead has −(1/10)αβ. The sign matters: the text obtains r02=27/80 by inserting β=−α into (B4), but the same insertion into the re-expanded formula gives 11/80; inserting β=+α gives 27/80. The physical sign of δΔ relative to δω for CZ02 vs CZ20 is not settled in the paper, and the stated relation δΔ=−δω is not derived from the resonance level scheme. If the realized CZ gate corresponds to the case with coefficient 11/80, Eq. (6) overestimates crosstalk infidelity by a factor 27/11 and the 99%-fidelity crosstalk threshold changes from 0.2% to about 0.31%. The experimental validation in Fig. 6(c) cannot adjudicate this because the phase between the gate and adversarial pulses was not set to the worst-case value and the measured fidelities have no error bars.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript addresses the problem of control crosstalk in flux-tunable transmon architectures. It defines a dimensionless crosstalk XΦ=dΦA/dΦB and presents three direct measurements: a resonator-spectroscopy DC method, a qubit-Ramsey DC method, and an AC method that measures the mean frequency shift of a qubit under a modulated flux pulse while an adversarial modulated pulse is applied to another qubit. The AC method is calibrated through a Bessel-function model of the average detuning. The paper then derives, in Appendix B, a leading-order expression r=(27π^2/20)(δf01 τ)^2 for the process infidelity of a parametrically activated CZ gate due to crosstalk, and uses quantum process tomography of a CZ gate in the presence of an adversarial pulse to test the prediction. It concludes that operating at the AC sweet spot suppresses crosstalk-induced infidelity to quartic order and that XΦ<0.2% is sufficient for 99% simultaneous gate fidelity.","tokens_in":17769,"tokens_out":11825,"duration_ms":97001,"significance":"If the central quantitative claim holds, the paper provides a practical and falsifiable bridge between a low-level crosstalk measurement and a high-level two-qubit gate fidelity: a Ramsey measurement of δf01 plus the calibrated gate time predicts an entangled-gate infidelity, and the AC sweet spot is identified as a design/operating principle for crosstalk robustness. The paper also contributes reproducible measurement methodologies—two independent DC methods that agree, a systematic AC crosstalk matrix, and frequency-dependent AC crosstalk data—with stated sensitivities. The Bessel-model fit for the AC calibration and the explicit Appendix B derivation are strengths; the derivation is independent of the fidelity data, and the comparison in Fig. 6(c) is not a fit to the theory.","major_comments":[{"comment":"Re-expanding the displayed final unitary U(τ)=diag{1,e^{-iα},1,e^{-iα}U11} with U11 from Eq. (B3) directly yields r=(1/5)α²+(1/10)αβ+(3/80)β²+(1/5)γ² to second order in α=δωτ, β=δΔτ, γ=δgτ. This has a positive αβ cross term, while Eq. (B4) has a negative one. Consequently, inserting δΔ=−δω into the re-expanded formula gives r02=(11/80)(δωτ)² rather than (27/80)(δωτ)²; the factor 27/11 discrepancy propagates to Eq. (6) and raises the 99%-fidelity crosstalk threshold from 0.2% to about 0.31%. Because the physical sign of δΔ relative to δω is stated without derivation, the coefficient in Eq. (6) is not established; the authors should either correct the expansion or justify the sign convention that makes Eq. (B4) valid.","section":"Appendix B, Eq. (B4) and Eq. (6)"},{"comment":"The experimental validation cannot resolve the coefficient ambiguity. The measured fidelities are plotted without error bars, the phase between the CZ pulse and the adversarial pulse was not set to the worst-case value used in the theory lines, and the paper itself notes that additional dephasing from the adversarial pulse may contribute. The data demonstrate a qualitative decrease away from the sweet spot and greater robustness at low adversarial amplitude, but they do not quantitatively confirm the coefficient in Eq. (6).","section":"Section V, Fig. 6(c)"}],"minor_comments":[{"comment":"The text says qubit A is modulated 'to the linear regime of its ¯Δ vs. flux amplitude curve,' but the figure shows a broad maximum; specify the numerical amplitude range used and how 'linear' is identified.","section":"Section IV, Fig. 4(b)"},{"comment":"The y-axis label 'Measured Fidelity (%)' and the legend entries in volts would be clearer if the adversarial amplitudes were converted to equivalent crosstalk XΦ using the calibration of Section IV, since the paper's metric is XΦ.","section":"Section V, Fig. 6(c)"},{"comment":"The first-order expansion in XΦ is written for the amplitude of the total flux; the paper should state explicitly that XΦ² terms are dropped and verify that this is negligible for the largest measured crosstalk values.","section":"Appendix B, Eq. (B1)"},{"comment":"The conclusion repeats the 0.2% threshold without the conditions used in Section V (equal modulation frequencies, worst-case phase, and negligible δgeff); these caveats should accompany the quoted number.","section":"Section VI, Conclusion"}],"recommendation":"major_revision","confidential_remarks":"The main issue is the sign of the cross term in Appendix B. If the authors can show that the negative sign follows from their actual phase convention (e.g., by deriving U11 from the Hamiltonian rather than quoting it), the paper would be close to acceptable; as written, Eq. (6)'s coefficient is not reliable. The experimental validation's lack of error bars and uncontrolled relative phase are additional concerns that the editor should weigh."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First the useful part: the three methods are a genuine toolkit. The resonator phase-offset method and the qubit slope method are simple, the cross-check on Q0/Q12 agrees, and the AC phase-interference method gives clean sinusoids with a sensitivity table. The full pairwise matrices and the frequency dependence plots are exactly the kind of data an engineering team needs to decide whether crosstalk is a design problem or a control problem. These parts deserve to be published.\n\nThe theory section is where I have a problem. Eq. (B4) in Appendix B has a sign error in the cross term. I expanded |tr(U_CZ† U)|² using their displayed U and U11 from (B3), and the cross term is +(1/10)(δωτ)(δΔτ), not −. With their CZ02 relation δΔ=−δω, the infidelity becomes 11/80 (δωτ)², not 27/80. So Eq. (6) should be (11π²/20)(δf τ)² unless the gate is actually CZ20, in which case 27/80 is right but the paper mislabels it. Either way, the Appendix B derivation as written does not produce the headline coefficient. The 99% / 0.2% crosstalk threshold changes to roughly 0.31% in the 11/80 case.\n\nThis is fixable, but it is load-bearing because Eq. (6) is the quantitative bridge to gate error. The experimental validation in Fig. 6(c) cannot settle the sign: no error bars on the QPT points, the adversarial phase was not held at the worst-case value, and the paper itself notes extra dephasing. The agreement is qualitative, which is fine for a methods paper, but the abstract should not present it as a confirmed prediction.\n\nCitation pattern is fine; the Bessel model and AC sweet spot come from the group's own published work, which is appropriate. No invented entities or hidden parameters beyond the standard voltage-to-flux conversions.\n\nFor whom: experimental groups working on tunable-transmon control, crosstalk mitigation, and simultaneous gate calibration. It deserves a serious referee; the measurements are valuable and the theory issue is fixable. I would ask the authors to show the expansion in Appendix B, state which CZ resonance was used, and either add error bars or soften the validation claim. Send to peer review.","headline":"Useful crosstalk-measurement toolkit with a sign error in the Appendix B infidelity derivation that flips the headline coefficient—fixable, but Eq. (6) should not be trusted as-is.","tokens_in":18324,"tokens_out":15399,"would_cite":true,"duration_ms":120525,"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":"Flux crosstalk error in two-qubit gates is now predictable from a Ramsey measurement.","keywords":["flux crosstalk","tunable transmon","parametric CZ gate","quantum process tomography","AC sweet spot","Ramsey measurement","superconducting qubit","gate fidelity"],"falsifier":"Run a CZ gate between two qubits while an adversarial flux pulse of fixed amplitude plays on a third qubit; measure the QPT process infidelity and the Ramsey-measured $\\delta \\bar{f}_{01}$, sweep $\\tau$, and check whether the infidelity tracks $(27\\pi^2/20)(\\delta \\bar{f}_{01}\\tau)^2$ to within calibration errors. At the AC sweet spot, verify the infidelity scales quarticly with the adversarial amplitude; a persisting quadratic term or a mismatch with the Ramsey-based prediction would show the formula or the assumed dominance of the frequency-shift channel is wrong.","tokens_in":17176,"feed_emoji":"⚛️","tokens_out":4914,"duration_ms":43421,"temperature":0.7,"pith_summary":"The paper tries to turn flux crosstalk between tunable transmons from a vague worry into a measurable specification. It presents three direct measurements—two for DC and one for AC flux crosstalk—with sensitivities as fine as about 0.001%, and defines crosstalk as a ratio of fluxes so values are comparable across devices. Its central theoretical result is an identity linking a Ramsey-measured average frequency shift to the process infidelity of a simultaneous parametrically activated CZ gate: $r = (27\\pi^2/20)(\\delta \\bar{f}_{01}\\tau)^2$. The paper argues that operating the gate at the AC sweet spot suppresses this error from quadratic to quartic order in crosstalk, keeping simultaneous gate fidelity above 99% for crosstalk below about 0.2%.","feed_headline":"Flux crosstalk error in CZ gates predicted by one formula","feed_subtitle":"Crosstalk measured to 0.001 percent now predicts simultaneous gate fidelity, with a sweet spot that suppresses the error.","key_machinery":"The central identity is the crosstalk ratio $X_\\Phi = d\\Phi_A/d\\Phi_B$ and the derived formula $r = (27\\pi^2/20)(\\delta \\bar{f}_{01}\\tau)^2$ for CZ infidelity. The ratio turns crosstalk into a device-comparable number, and the formula turns a Ramsey measurement of $\\delta \\bar{f}_{01}$ under a coherent adversarial pulse into a quantitative fidelity prediction without running two-qubit tomography. The argument runs through an effective interaction-picture Hamiltonian in which crosstalk adds an average-frequency-shift term $\\delta\\bar{\\Delta}$ and a coupling shift $\\delta g_{\\text{eff}}$; leading-order infidelity follows from the trace overlap of the resulting evolution with the ideal CZ unitary. The AC sweet spot enters as the operating point where $\\partial \\bar{f}_{01}/\\partial\\Phi = 0$, making the leading frequency shift second-order in crosstalk and the infidelity quartic.","core_discovery":"The central claim is that flux crosstalk between tunable transmons can be quantified directly with three low-level measurements, and that the AC version of those measurements predicts the fidelity of simultaneous two-qubit gates through a single formula. For a CZ gate of duration $\\tau$ whose tunable qubit suffers an average frequency shift $\\delta \\bar{f}_{01}$ from an adversarial flux pulse, the leading-order process infidelity is $r = (27\\pi^2/20)(\\delta \\bar{f}_{01}\\tau)^2$. The paper derives this from an effective Hamiltonian in which crosstalk shifts both the average qubit frequency and the effective coupling, keeps the frequency-shift term after calibrating away local phase errors, and validates it with quantum process tomography of a CZ gate run while an adversarial pulse is played. At the AC sweet spot—the flux-pulse amplitude where the average frequency shift is stationary—the linear frequency shift vanishes, so the leading infidelity becomes quartic in crosstalk; the paper finds this operating point sustains 99% simultaneous gate fidelity for crosstalk below roughly 0.2%.","pith_inferences":["If equation (6) survives further tests, the same Ramsey-interference measurement could serve as an acceptance test for chips: crosstalk below a specified $\\delta \\bar{f}_{01}$ at the operating pulse amplitude would guarantee simultaneous gate fidelity without process tomography.","A natural extension is to apply the phase-scan method to measure crosstalk from flux lines into fixed-frequency qubits or other circuit elements, using any frequency-dependent observable instead of $\\bar{f}_{01}$.","The quartic sweet-spot suppression implies there may be an optimal trade-off between error from crosstalk and error from increased gate time or reduced coupling when moving to the sweet spot."],"forward_implications":["A Ramsey measurement of $\\delta \\bar{f}_{01}$ plus the calibrated gate time $\\tau$ gives a direct prediction of simultaneous CZ infidelity, replacing two-qubit tomography as a crosstalk diagnostic.","Operating parametrically activated CZ gates at the AC sweet spot reduces the crosstalk contribution to infidelity from quadratic to quartic order, so 99% simultaneous gate fidelity requires crosstalk below roughly 0.2% on this architecture.","The DC resonator and qubit methods agree to within about 0.1% across pairs and can be used to build chip-wide crosstalk matrices that expose asymmetric, non-local coupling between tunable qubits.","AC crosstalk varies strongly with modulation frequency for some qubit pairs, meaning simultaneous-gate error can be mitigated by choosing pulse frequencies where crosstalk is small."],"supporting_citations":[{"why":"Supplies the Bessel-function model of the mean shift (Eq. 3) and the interaction-picture Hamiltonian (Eqs. 4-5) on which the CZ gate and the infidelity derivation build.","marker":"[28]"},{"why":"Establishes the native parametric CZ gate that the crosstalk measurement and tomography target.","marker":"[29]"},{"why":"Demonstrates the parametric entangling gate architecture and its calibration on a multi-qubit lattice.","marker":"[30]"},{"why":"Provides the background for the AC sweet spot concept and the parametrically activated gate protected from flux noise.","marker":"[1]"},{"why":"Defines the transmon and the asymmetric SQUID flux-tunable configuration used by the device.","marker":"[20]"}],"fun_headline_variants":["Three measurements quantify transmon flux crosstalk","AC crosstalk formula predicts CZ gate infidelity","Crosstalk sweet spot gives 99% two-qubit gate fidelity","Flux crosstalk measured to 0.001% predicts gate error","One formula links flux crosstalk to CZ gate error"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The prediction assumes that the average frequency shift measured with a Ramsey sequence while an adversarial pulse is played matches the average shift the qubit experiences during the simultaneous CZ gate, and that crosstalk infidelity is dominated by that frequency-shift term rather than by coupling changes, dephasing, or spurious resonances.","fun_headline_variants_meta":{"raw":{"variants":["Three measurements quantify transmon flux crosstalk","AC crosstalk formula predicts CZ gate infidelity","Crosstalk sweet spot gives 99% two-qubit gate fidelity","Flux crosstalk measured to 0.001% predicts gate error","One formula links flux crosstalk to CZ gate error"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001537,"raw_usage":{"total_tokens":6144,"prompt_tokens":932,"completion_tokens":5212,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":548,"completion_tokens_details":{"reasoning_tokens":5125}},"tokens_in":548,"tokens_out":5212,"duration_ms":30565,"temperature":1.0,"reasoning_tokens":5125,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T10:05:25.475014+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a CZ gate between two qubits while an adversarial flux pulse of fixed amplitude plays on a third qubit; measure the QPT process infidelity and the Ramsey-measured $\\delta \\bar{f}_{01}$, sweep $\\tau$, and check whether the infidelity tracks $(27\\pi^2/20)(\\delta \\bar{f}_{01}\\tau)^2$ to within calibration errors. At the AC sweet spot, verify the infidelity scales quarticly with the adversarial amplitude; a persisting quadratic term or a mismatch with the Ramsey-based prediction would show the formula or the assumed dominance of the frequency-shift channel is wrong.","supporting_citations":[{"cited_title":"Charge-insensitive qubit design derived from the cooper pair box,","cited_arxiv_id":null,"evidence_quote":"Defines the transmon and the asymmetric SQUID flux-tunable configuration used by the device."},{"cited_title":"Analytical modeling of parametrically-modulated transmon qubits","cited_arxiv_id":"1706.06566","evidence_quote":"Supplies the Bessel-function model of the mean shift (Eq. 3) and the interaction-picture Hamiltonian (Eqs. 4-5) on which the CZ gate and the infidelity derivation build."},{"cited_title":"Parametrically activated entangling gates using transmon qubits,","cited_arxiv_id":null,"evidence_quote":"Establishes the native parametric CZ gate that the crosstalk measurement and tomography target."},{"cited_title":"Demonstration of universal paramet- ric entangling gates on a multi-qubit lattice,","cited_arxiv_id":null,"evidence_quote":"Demonstrates the parametric entangling gate architecture and its calibration on a multi-qubit lattice."},{"cited_title":"Demonstration of a Parametrically-Activated Entangling Gate Protected from Flux Noise","cited_arxiv_id":"1901.08035","evidence_quote":"Provides the background for the AC sweet spot concept and the parametrically activated gate protected from flux noise."}],"review_version":1}