Pith. sign in

REVIEW 3 major objections 3 minor 37 references

Error-Generator-Level Compression for Fast Quantum Gates

T0 review · 3 major / 3 minor · reviewed 2026-08-03 · deepseek-v4-flash

Pith's one-line read The paper claims that a three-parameter compressed Magnus pulse suppresses coherent errors in fast transmon gates by orders of magnitude, outperforms grid-optimized leading-order DRAG, and approaches a fully parameterized 17-parameter fourt

desk verdict New generator-level compression objective with a clean local link to fidelity error, but the headline 3-parameter advantage rests on an unquantified Magnus truncation that a referee should press on before the numbers are taken as physical. read the letter →

arxiv 2607.28922 v1 pith:6CT5NXQB submitted 2026-07-31 quant-ph

classification quant-ph PACS 03.67.Lx
keywords error-generatorcompressionMagnus-basedcontrolDRAGtransmongatesleakagesuppressionpulseoptimizationcoherenterrorsfinite-bandwidthrobustness
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that fast quantum gates do not have to choose between low-dimensional experimental pulses and high-dimensional error suppression. It introduces error-generator-level compression: keep the full projected error generator at a chosen Magnus order but restrict the waveform to a few independently adjustable coefficients, then minimize a spectral penalty over the residual eigenphases. For a four-level transmon model, a three-parameter pulse—two Fourier coefficients plus a constant detuning—achieves average fidelity errors orders of magnitude below the uncorrected pulse, beats a grid-optimized leading-order DRAG benchmark, and comes close to a fully parameterized 17-parameter fourth-order Magnus correction over the gate-time range studied. The compressed pulse is also substantially less sensitive to a Gaussian finite-bandwidth filter. The broader claim is that generator-level objectives, rather than final-time fidelity, are the right target for low-dimensional pulse compression.

What carries the argument

The central object is the projected truncated Magnus error generator E_rel^(n) = P_rel[-i Omega^(n)(t_f)], a Hermitian operator whose eigenvalues are the residual eigenphases of the truncated interaction-picture evolution. The cost function is Phi^(n) = Tr[exp(E_rel) + exp(-E_rel) - 2 1_supp] = 2 sum_j (cosh(lambda_j) - 1), which is convex in the generator, reduces to the squared Frobenius norm of E_rel near the perturbative origin, and penalizes large eigenphases away from it. The pulse coefficients enter through a Fourier decomposition of the two quadrature envelopes and a constant detuning; the generator is built using the paper's iterative singular fourth-order construction, keeping at m

What would settle it

Take the optimized three-parameter coefficients reported at |alpha_2| t_f = 5.74 (a_x/E_C = −0.017, b_y/E_C = 0.301, Δ/E_C = 0.065) and propagate the full four-level Hamiltonian using a Magnus generator that keeps three or more powers of the new fourth-order correction, or use a five-level transmon truncation; if the average fidelity error no longer lies below the grid-selected DRAG envelope over the same gate-time range, the central advantage is an artifact of the truncation.

Watch

Extended reading notes

Core claim

The central discovery is that a pulse restricted to just three experimentally natural parameters—two lowest-order Fourier corrections on the two quadratures and a constant detuning—can suppress all eleven projected coherent-error and leakage components of a fourth-order Magnus generator in a four-level transmon model. The compressed pulse reaches an average fidelity error of 10^-3 at a gate time about 32% shorter than the grid-selected DRAG benchmark, reduces error by roughly a factor of 42 relative to the uncorrected pulse and by about a factor of 4 relative to DRAG at the shortest gate time studied, and remains within about a factor of 5 of the fully parameterized 17-parameter correction.

Load-bearing premise

The load-bearing premise is that the truncated fourth-order Magnus generator—computed with the iterative singular construction and keeping at most two powers of the new fourth-order correction—faithfully represents the exact error dynamics at |alpha_2| t_f ≈ 5.7–11.5, and that the four-level transmon truncation (justified only by a self-cited companion paper) is adequate; if the discarded higher-order terms or higher transmon levels dominate at these drive strengths, the repo

Editorial extensions

If this is right

  • A three-parameter compressed pulse reaches an average fidelity error of 10^-3 at |alpha_2| t_f ≈ 7.8, versus ≈ 11.5 for grid-selected leading-order DRAG, a roughly 32% shorter gate time at fixed error.
  • At the shortest gate time shown (|alpha_2| t_f = 5.74), the compressed pulse reduces error by about a factor of 42 relative to the uncorrected pulse and about a factor of 4 relative to the grid-selected DRAG benchmark, while staying within about a factor of 5 of the fully parameterized 17-parameter Magnus correction.
  • Minimizing the generator-level spectral penalty, rather than the truncated residual unitary or a subset of error channels, is what prevents unitary-periodicity traps and large masked eigenphases.
  • The compressed pulse degrades under a Gaussian finite-bandwidth filter only near omega_BW/|alpha_2| ≈ 1, while the fully parameterized 17-parameter pulse starts degrading near ≈ 5, so the low-dimensional pulse is substantially more robust to the filtering model considered.
  • The method retains all eleven projected error components of the fourth-order generator even though the waveform uses only three parameters, showing that high-dimensional error models can be paired with low-dimensional waveforms.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the generator-level compression principle generalizes, the same three-parameter-style construction could be applied to other coherent-error models—such as crosstalk or two-qubit gates—whenever those errors can be represented in a truncated Magnus generator with a restricted control parametrization; the paper states the principle is not transmon-specific but does not demonstrate those cases.
  • Because the objective is convex in the generator but not necessarily in pulse-coefficient space, experimental calibration may need global search or warm-starting even though the low-dimensional pulse is simple to implement; the paper notes this nonconvexity but does not provide a calibration recipe.
  • The finite-bandwidth robustness suggests that compressed low-frequency pulses could reduce the need for detailed transfer-function correction in control hardware, a practical consequence the paper illustrates but does not develop into a full hardware model.
  • A natural testable extension is to compare the three-parameter pulse against higher-order or continuously optimized DRAG variants on real hardware with randomized benchmarking; the paper only benchmarks against grid-selected leading-order DRAG.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 3 minor

Summary. The paper introduces 'error-generator-level compression,' a Magnus-based control method that restricts the implemented waveform to a few adjustable Fourier/detuning parameters while evaluating a spectral penalty on the projected truncated error generator. For a four-level transmon model with EJ/EC=50, a three-parameter pulse (two Fourier coefficients plus a constant detuning) is optimized by minimizing the functional Φ^(4). The authors report that this pulse suppresses average fidelity errors by orders of magnitude, outperforms grid-optimized leading-order DRAG across the investigated gate-time range, approaches a fully parameterized 17-parameter fourth-order Magnus correction, and is less sensitive to an illustrative Gaussian finite-bandwidth filter. The central technical content is the derivation of the spectral penalty Φ, its local equivalence to the squared Frobenius norm of the projected generator, and its relation to the average fidelity error via bounds in Supplemental Eq. (S66).

Significance. If the numerical results are robust, the paper offers a practical design principle: low-dimensional, experimentally natural pulses can suppress many coherent-error channels simultaneously. The main comparison is strengthened by evaluating all reported fidelity errors via exact numerical propagation of the four-level model, rather than through the truncated Magnus generator used for optimization. The DRAG benchmark is also fairly constructed as the pointwise best over a discrete β grid. The derivation of the generator-fidelity relation in Supplemental Sec. V is clean and dimensionally consistent. The method's claimed advantage depends on the adequacy of the four-level truncation and on the quantitative faithfulness of the truncated fourth-order Magnus objective; these points need further support before the strongest claims are accepted.

major comments (3)
  1. [Supplemental Sec. III, Eq. (S16)] The objective Φ^(4) is minimized over a diagnostic generator that is truncated to terms containing at most two powers of W_eff,4. The magnitude of the discarded ≥3-power terms is never quantified, and ε=1 at |α2|tf ~5.7-11.5 is not an obviously perturbative operating point. The abstract's phrase 'complete projected error generator' is therefore overstated. Note that because the reported fidelity errors are computed by exact four-level propagation, the specific optimized pulses' performance is not an artifact of this truncation; however, the optimality of the coefficients and the adequacy of the 17-parameter reference are established only within the truncated model. Please quantify the neglected terms (e.g., their operator norms) or soften the completeness claim.
  2. [Main text, 'Transmon model and pulse parameterization'; SM Table I] The four-level truncation is justified only by self-cited Ref. [24]. No convergence check with five or more transmon levels is presented, and at |α2|tf ≈ 5.7 the drive is strong enough that higher-lying states could contribute through virtual processes. Please provide a numerical convergence test (e.g., comparing the reported error curves for 4, 5, and 6 retained levels) or an independent estimate of the neglected |4> and |5> population/amplitude. This is load-bearing because the physical significance of the reported advantage over DRAG depends on the four-level model being quantitatively accurate.
  3. [Main text Eqs. (12)-(13); Supplemental Eq. (S66)] The relation between the compressed Magnus functional and the average fidelity error is derived only at leading quadratic order. The paper never reports the residual eigenphases λ_j or the value of Φ^(4) at the operating points shown in Fig. 2, so the reader cannot verify that the optimized pulses lie in the regime where Eq. (S66) applies. Reporting these diagnostics would directly support the claim that minimizing the generator-level penalty controls the exact error and would also clarify whether the selected three-parameter coefficients are genuinely in the perturbative regime.
minor comments (3)
  1. [Abstract and Fig. 2(c)] The statement that the three-parameter pulse 'approaches' the 17-parameter correction is accurate only at the shortest gate times; at |α2|tf ≈ 11.5 the two curves differ by roughly two orders of magnitude. The abstract and Discussion should qualify this claim, e.g., 'at the fastest gate times studied.'
  2. [Main text, 'Compressed generator minimization'] The optimization of Φ^(4) over the three parameters is not described (algorithm, initial guesses, number of restarts). Since coefficient-space convexity is not guaranteed, a brief description of the optimization procedure would improve reproducibility.
  3. [General] Minor typographical issue: the text consistently renders 'coefficients' with a ligature ('coefficients'). Please fix throughout.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: reported fidelity improvements are evaluated by exact four-level propagation after optimizing a generator-level cost, and the cost–fidelity relation is a proven bound rather than an identity.

full rationale

The derivation chain is not circular. The paper defines a truncated Magnus error generator E^(n), projects it to E_rel^(n), and minimizes the spectral penalty Phi^(n) over a few pulse coefficients. The reported average fidelity error epsilon is then obtained by numerically propagating the full four-level transmon Hamiltonian, not by evaluating Phi^(n) alone. Supplemental Eq. (S66) gives only the local bounds Phi/4 ≲ epsilon ≲ Phi/3; this is an inequality proved from the fidelity expansion, not an equality imposed by construction. Thus the three-parameter pulse's advantage over DRAG and its proximity to the 17-parameter reference are computed through an independent propagation step, even though the objective was designed to control the same error channels. The four-level truncation is justified by self-cited Ref. [24], but that is a model-choice citation, not a theorem that makes the numerical conclusion equal to its premise. Similarly, the truncation in Eq. (S16) (keeping at most two powers of Weff,4) is an unquantified approximation that raises numerical-accuracy concerns, but it does not make the reported result definitionally identical to the input. No fitted parameter is renamed as a prediction, and no uniqueness claim from the authors' prior work is used to forbid alternatives. The central compression principle and its numerical demonstration are therefore self-contained rather than circular; any weakness lies in model validation, not in circular reasoning.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central calculation rests on a fixed four-level transmon model, a fourth-order Magnus truncation with a quadratic-in-W_eff,4 approximation for the reference, and a Gaussian filter model. The optimization variables are the three (or 17) pulse coefficients and the benchmark DRAG coefficient. No new physical entities are introduced; the projected error generator is a mathematical diagnostic, not a new mediator, force, or dimension.

free parameters (5)
  • ax (cosine Fourier coefficient of gx) = -0.017/EC at |alpha_2| t_f = 5.74
    One of the three optimized pulse parameters; chosen by minimizing Phi^(4) in Eq. (10).
  • by (sine Fourier coefficient of gy) = 0.301/EC at |alpha_2| t_f = 5.74
    One of the three optimized pulse parameters; chosen by minimizing Phi^(4).
  • Delta (constant detuning) = 0.065/EC at |alpha_2| t_f = 5.74
    Third control parameter; constant carrier detuning, optimized alongside the two envelope coefficients.
  • 17 parameters of fully parameterized reference correction = One detuning plus 16 Fourier amplitudes (values not tabulated)
    Reference pulse obtained by solving the iterative fourth-order cancellation conditions (SM Eq. S18); used as the comparison baseline.
  • beta_DRAG (benchmark only) = Selected per gate time from {0, 0.1, ..., 1.0}
    Grid-optimized leading-order DRAG coefficient used only as a benchmark, not part of the proposed method.
assumptions (5)
  • domain assumption The fourth-order truncated Magnus generator adequately captures the dominant error dynamics at epsilon = 1.
    Main text Eqs. (4)-(5) and SM Sec. IV; the authors note the method does not remove terms beyond the retained order, but the size of those terms is not quantified.
  • domain assumption The four-level transmon truncation is quantitatively accurate for the fast-gate regime studied.
    Main text after Eq. (14) and Ref. [24]; the three-level model is said to be inadequate, but the four-level sufficiency is justified only by a self-cited companion paper.
  • domain assumption The only available controls are the two microwave quadratures of the charge drive plus a constant detuning.
    Eqs. (15)-(16) and SM Eq. (S6); this defines the constrained-control problem and ignores other parasitic or dissipative channels.
  • domain assumption Discarding the Q-internal block and the common phase via the projection Prel does not affect the leading average fidelity error.
    SM Eqs. (S9), (S55)-(S66); shown to hold only to leading quadratic order in the residual generator.
  • domain assumption The Gaussian filter is a representative finite-bandwidth model.
    SM Sec. VI; the authors explicitly label it illustrative rather than a measured transfer function, so the bandwidth comparison is qualitative.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Error-Generator-Level Compression for Fast Quantum Gates." pith.science (2026). https://pith.science/paper/6CT5NXQB

@misc{pith2026260728922,
  author       = {Pith},
  title        = {Pith review of: Error-Generator-Level Compression for Fast Quantum Gates},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6CT5NXQB}},
  note         = {Machine review of arXiv:2607.28922}
}
read the original abstract

Fast quantum gates in multilevel systems require suppressing many coherent-error channels using pulses that remain simple to implement. We introduce error-generator-level compression, a Magnus-based control principle that minimizes the complete projected error generator at a chosen perturbative order while restricting the implemented waveform to a few independently adjustable coefficients. This construction preserves all modeled computational and leakage errors rather than truncating the pulse by error channel, while its generator-level objective directly penalizes residual eigenphases that can signal departure from the perturbative regime. For fast transmon gates, a three-parameter pulse suppresses errors by orders of magnitude, outperforms grid-optimized leading- order DRAG over the gate-time range studied, and approaches a fully parameterized 17-parameter Magnus correction. The compressed pulse is also substantially less sensitive to the illustrative finite-bandwidth filtering model considered here.

Figures

Figures reproduced from arXiv: 2607.28922 by the authors.

Figure 1
Figure 1. Schematic comparison between fidelity-based com￾pression and generator-level compression. Minimizing a final￾time unitary objective can favor nonperturbative solutions because the unitary is periodic in the residual eigenphases. In contrast, compressed Magnus control evaluates a spectral penalty that is convex in the Hermitian diagnostic genera￾tor and is minimized when the projected residual eigenphases vanish; the… view at source ↗
Figure 2
Figure 2. Average fidelity error ϵ versus dimensionless gate time |α2|tf for a target π/2 rotation about the x axis, com￾puted from the projected final-time evolution of the four-level transmon model. (a) Naive three-parameter truncation re￾taining only the largest error channels predicted by the fourth￾order generator (yellow). (b) Three-parameter minimization of the truncated residual unitary exp[Ωˆ(4)] without (light blue)… view at source ↗
Figure 3
Figure 3. compares the modified pulses at |α2|tf = 5.74. The fully parameterized pulse (green trace) contains vis￾ible higher-frequency structure across the control chan￾nels, whereas the three-parameter compressed pulse (or￾ange trace) is smooth and dominated by low-frequency components [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 1 canonical work pages

  1. [24]

    R. T. Stanescu and H. Ribeiro, Effective Hamiltoni- ans for Predictive Quantum Control, arXiv preprint 10.48550/arXiv.2607.27111 (2026)

  2. [1]

    Motzoi, J

    F. Motzoi, J. M. Gambetta, P. Rebentrost, and F. K. Wilhelm, Simple pulses for elimination of leakage in weakly nonlinear qubits, Phys. Rev. Lett. 103, 110501 (2009)

  3. [2]

    J. M. Gambetta, F. Motzoi, S. T. Merkel, and F. K. Wil- helm, Analytic control methods for high-fidelity unitary operations in a weakly nonlinear oscillator, Phys. Rev. A 83, 012308 (2011)

  4. [3]

    J. M. Chow, L. DiCarlo, J. M. Gambetta, F. Motzoi, L. Frunzio, S. M. Girvin, and R. J. Schoelkopf, Optimized driving of superconducting artificial atoms for improved single-qubit gates, Phys. Rev. A 82, 040305 (2010)

  5. [4]

    L. S. Theis, F. Motzoi, S. Machnes, and F. K. Wilhelm, Counteracting systems of diabaticities using DRAG con- trols: The status after 10 years, EPL 123, 60001 (2018)

  6. [5]

    Hyyppä, A

    E. Hyyppä, A. Vepsäläinen, M. Papič, C. F. Chan, S. Inel, A. Landra, W. Liu, J. Luus, F. Marxer, C. Ockeloen-Korppi, S. Orbell, B. Tarasinski, and J. Heinsoo, Reducing leakage of single-qubit gates for su- perconducting quantum processors using analytical con- trol pulse envelopes, PRX Quantum 5, 030353 (2024)

  7. [6]

    Khaneja, T

    N. Khaneja, T. Reiss, C. Kehlet, T. Schulte-Herbrüggen, and S. J. Glaser, Optimal control of coupled spin dynam- ics: Design of NMR pulse sequences by gradient ascent algorithms, J. Magn. Reson. 172, 296 (2005)

  8. [7]

    V. F. Krotov, Global Methods in Optimal Control Theory , Monographs and Textbooks in Pure and Applied Math- ematics, Vol. 195 (Marcel Dekker, New York, 1996)

Show all 37 references
  1. [8]

    Caneva, T

    T. Caneva, T. Calarco, and S. Montangero, Chopped random-basis quantum optimization, Phys. Rev. A 84, 022326 (2011)

  2. [9]

    Machnes, E

    S. Machnes, E. Assémat, D. Tannor, and F. K. Wilhelm, Tunable, flexible, and efficient optimization of control pulses for practical qubits, Phys. Rev. Lett. 120, 150401 (2018)

  3. [10]

    S. J. Glaser, U. Boscain, T. Calarco, C. P. Koch, W. Köckenberger, R. Kosloff, I. Kuprov, B. Luy, S. Schirmer, T. Schulte-Herbrüggen, D. Sugny, and F. K. Wilhelm, Training schrödinger’s cat: Quantum optimal control, Eur. Phys. J. D 69, 279 (2015)

  4. [11]

    M. M. Müller, R. S. Said, F. Jelezko, T. Calarco, and S. Montangero, One decade of quantum optimal con- trol in the chopped random basis, Rep. Prog. Phys. 85, 076001 (2022)

  5. [12]

    Werninghaus, D

    M. Werninghaus, D. J. Egger, F. Roy, S. Machnes, F. K. Wilhelm, and S. Filipp, Leakage reduction in fast super- conducting qubit gates via optimal control, npj Quantum Inf. 7, 14 (2021)

  6. [13]

    Y. Song, J. Li, Y.-J. Hai, Q. Guo, and X.-H. Deng, Opti- mizing quantum control pulses with complex constraints and few variables through autodifferentiation, Phys. Rev. A 105, 012616 (2022)

  7. [14]

    Genois, N

    É. Genois, N. J. Stevenson, N. Goss, I. Siddiqi, and A. Blais, Quantum optimal control of superconducting qubits based on machine-learning characterization, Phys. Rev. Applied 24, 034073 (2025)

  8. [15]

    M. A. Rol, L. Ciorciaro, F. K. Malinowski, B. M. Tarasin- ski, R. E. Sagastizabal, C. C. Bultink, Y. Salathé, N. Haandbæk, J. Sedivy, and L. DiCarlo, Time-domain characterization and correction of on-chip distortion of control pulses in a quantum processor, Appl. Phys. Lett. 1...

  9. [16]

    L.-L. Guo, P. Duan, S. Zhang, X.-X. Yang, C. Zhang, L. Du, H.-F. Zhang, H.-R. Tao, T.-L. Wang, Z.-L. Jia, Z.-Y. Chen, and G.-P. Guo, Universal scalable character- ization and correction of pulse distortions in controlled quantum systems, Phys. Rev. Applied 21, 064060 (2024)

  10. [17]

    W. Dai, S. Hazra, D. K. Weiss, P. D. Kurilovich, T. Con- nolly, H. K. Babla, S. Singh, V. R. Joshi, A. Z. Ding, P. D. Parakh, J. Venkatraman, X. Xiao, L. Frunzio, and M. H. Devoret, Characterization of drive-induced unwanted state transitions in superconducting circuits, Phys....

  11. [18]

    Schutjens, F

    R. Schutjens, F. Abu Dagga, D. J. Egger, and F. K. Wil- helm, Single-qubit gates in frequency-crowded transmon systems, Phys. Rev. A 88, 052330 (2013)

  12. [19]

    Ribeiro, A

    H. Ribeiro, A. Baksic, and A. A. Clerk, Systematic magnus-based approach for suppressing leakage and nonadiabatic errors in quantum dynamics, Phys. Rev. X 7, 011021 (2017)

  13. [20]

    Buterakos, S

    D. Buterakos, S. Das Sarma, and E. Barnes, Geometrical formalism for dynamically corrected gates in multiqubit systems, PRX Quantum 2, 010341 (2021)

  14. [21]

    Watanabe, Y

    S. Watanabe, Y. Tabuchi, K. Heya, S. Tamate, and Y. Nakamura, ZZ -interaction-free single-qubit-gate op- timization in superconducting qubits, Phys. Rev. A 109, 012616 (2024)

  15. [22]

    T. F. Roque, A. A. Clerk, and H. Ribeiro, Engineering fast high-fidelity quantum operations with constrained interactions, npj Quantum Inf. 7, 28 (2021)

  16. [23]

    See Supplemental Material at [URL will be inserted by publisher] for the transmon parameters, grid search over the DRAG coefficient, iterative fourth-order construc- tion, derivation and convexity properties of the com- pressed Magnus functional, its relation to the average fid...

  17. [25]

    J. Koch, T. M. Yu, J. Gambetta, A. A. Houck, D. I. Schuster, J. Majer, A. Blais, M. H. Devoret, S. M. Girvin, and R. J. Schoelkopf, Charge-insensitive qubit design de- rived from the cooper pair box, Phys. Rev. A 76, 042319 (2007)

  18. [26]

    Blais, A

    A. Blais, A. L. Grimsmo, S. M. Girvin, and A. Wallraff, Circuit quantum electrodynamics, Rev. Mod. Phys. 93, 025005 (2021)

  19. [27]

    Makhlin, G

    Y. Makhlin, G. Schön, and A. Shnirman, Quantum-state engineering with josephson-junction devices, Rev. Mod. Phys. 73, 357 (2001)

  20. [28]

    Error-Generator-Level Compression for Fast Quantum Gates

    Z. Chen, J. Kelly, C. Quintana, R. Barends, B. Camp- bell, Y. Chen, B. Chiaro, A. Dunsworth, A. G. Fowler, E. Lucero, E. Jeffrey, A. Megrant, J. Mutus, M. Nee- ley, C. Neill, P. J. J. O’Malley, P. Roushan, D. Sank, A. Vainsencher, J. Wenner, T. C. White, A. N. Korotkov, and J....

  21. [29]

    (S14) Here, ˆWeff,2,I(t; c⋆

    + ˆWeff,4,I(t; c4). (S14) Here, ˆWeff,2,I(t; c⋆

  22. [30]

    The new correction is implemented using the same physical control resources: The envelopes gx(t) and gy(t) and the constant detuning ∆

    is the fixed second-order correction, whereas ˆWeff,4,I(t; c4) is the additional correction to be deter- mined at fourth order. The new correction is implemented using the same physical control resources: The envelopes gx(t) and gy(t) and the constant detuning ∆. We compute th...

  23. [31]

    remains inside the Magnus expansion through fourth order, and ˆWeff,4,I(t; c4) is chosen to cancel, through order ε4, every independent component of the projected diagnostic generator in B ˆP . Terms quadratic in ˆWeff,4,I(t; c4) are essential in the ill-conditioned case becau...

  24. [32]

    Applying the same fixed projection as at second order gives the complete fourth-order projected diagnostic generator ˆE(4) rel,eff(tf; c4; c⋆

    = i ˆΩ(4) eff (tf; c4; c⋆ 2) ≤2 powers of ˆWeff,4 , (S16) where the vertical bar denotes the truncation just described. Applying the same fixed projection as at second order gives the complete fourth-order projected diagnostic generator ˆE(4) rel,eff(tf; c4; c⋆

  25. [33]

    = Prel h ˆE(4) eff (tf; c4; c⋆ 2) i = X ˆAν ∈B ˆP E(4) ν (tf; c4; c⋆

  26. [34]

    (S17) The fourth-order correction is determined by canceling every independent component of this projected generator, E(4) ν (tf; c4; c⋆

    ˆAν. (S17) The fourth-order correction is determined by canceling every independent component of this projected generator, E(4) ν (tf; c4; c⋆

  27. [35]

    (S18) The system in Eq

    = 0 , ˆAν 2 B ˆP . (S18) The system in Eq. ( S18) is nonlinear but at most quadratic in c4. This is the practical advantage of the iterative construction: The coefficients c⋆ 2 have already been chosen so that ˆE(2) rel,eff(tf; c⋆

  28. [36]

    (S19) Thus, every traceless error within the computational subspace and every coupling between the computational and leakage subspaces is canceled through second order

    = 0. (S19) Thus, every traceless error within the computational subspace and every coupling between the computational and leakage subspaces is canceled through second order. The solution c⋆ 4 then cancels every component of the same projected diagnostic generator through fourt...

  29. [37]

    + ˆWeff,4(t; c⋆ 4). (S20) For the transmon drive, this correction is implemented through the two charge-drive quadratures and a constant detuning, ˆW (4)(t) = h g(4) x (t) cos ω(4) d t + g(4) y (t) sin ω(4) d t i ˆn, ω(4) d = ω01 + ∆(4). (S21) The drive frequency is held const...

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.