{"id":"445680b3-1396-413d-8a7d-e055e07ee481","arxiv_id":"2412.20149","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"Pontryagin-optimized pulses can move a damped resonator to a target coherent state with lower energy than shortcut-to-adiabatic pulses and provide high readout SNR at large critical photon numbers.","lead":"This paper uses Pontryagin's maximum principle to design energy-efficient and fast pulses for a lossy microwave resonator, and applies them to qubit readout in circuit quantum electrodynamics. The authors report signal-to-noise ratios comparable to longitudinal coupling readout at high critical photon numbers, but a few derivation steps and one modeled regime need scrutiny.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (7) does not solve the stated time-optimal control problem: substituting φ_opt=ω_r t+θ into Eq. (1) gives a different trajectory, and Eq. (7) cannot reach |α|=10 for the Fig. 3 parameters.","rationale":"The central claim of the paper is not only that the energy-optimal Gramian pulse is correct; it is the broader claim that PMP supplies both energy- and time-optimal analytical pulses for the dissipative LC system, with Fig. 3(a) showing t_min=1.01 μs at ε_max=10 MHz. That time-optimal claim rests entirely on Eq. (7). The derivation of Eq. (7) is wrong in a way that is easy to check: the control phase derived from the adjoint equations is φ_opt(t)=ω_r t+θ, i.e. ε(t)=ε_max e^{i(ω_r t+θ)}, but Eq. (7) is not the solution of Eq. (1) for that ε(t). The exact solution retains a factor e^{iω_r t} that Eq. (7) replaces by 1. With the paper's own Fig. 3 parameters, Eq. (7) gives max |α|≈5.3, so the boundary condition |α(t_f)|=10 is impossible; the reported t_min=1.01 μs is the value for a different rotating-frame constant-drive model. This is an internal inconsistency, not a disagreement with consensus. The reader's weakest_assumption concerned the n_crit=1 dispersive readout, which is a real limitation the paper itself concedes; I agree with that concern, but it affects only one of the three readout cases and the conclusion already restricts the SNR claim to large critical photon numbers. The Eq. (7) inconsistency, by contrast, invalidates a headline contribution advertised in the abstract and conclusion. A second abstract claim, 'exponential scaling in energy cost,' is also not supported by Eq. (6), but the time-optimal error is the more directly checkable single point of failure. I therefore move the verdict from CONDITIONAL to REJECT for the current version: the time-optimal analytical pulse should be re-derived in a consistent frame, or withdrawn, before the paper's central claims can be evaluated.","tokens_in":19639,"tokens_out":27100,"duration_ms":277431,"concrete_test":"One concrete check: substitute ε(t)=ε_max e^{i(ω_r t+θ)} and α(0)=0 into Eq. (1) and compare the exact solution with Eq. (7); the two differ by e^{iω_r t} versus 1. Numerically, for ω_r=2π×0.3 MHz, κ=2π×10 kHz, ε_max=10 MHz, plot max_{t∈[0,10 μs]} |α(t)| from Eq. (7); it is approximately 5.3, so α(t_f)=10e^{iϑ} is unattainable and t_min=1.01 μs is not a solution of the stated boundary-value problem. Additionally, run a direct ODE solver for Eq. (1) under the reported phase-optimal control and verify the terminal state; if it lands at |α|≈4.3 at t=1.01 μs rather than 10, the time-optimal claim is refuted.","verdict_should_be":"REJECT","load_bearing_attack":"Eq. (7) is not the solution of Eq. (1) for the control derived in the same paragraph. The text sets ε1=ε_max cos φ, ε2=ε_max sin φ and obtains φ_opt(t)=ω_r t+θ, so ε(t)=ε_max e^{i(ω_r t+θ)}. Solving Eq. (1) with α(0)=0 gives α(t)=[2ε_max e^{iθ}/(κ+4iω_r)][e^{-(κ/2+iω_r)t}−e^{iω_r t}], whereas Eq. (7) contains [e^{-(κ/2+iω_r)t}−1]. The printed expression is equivalent to a constant lab-frame drive, not to the phase-modulated pulse φ_opt(t)=ω_r t+θ. The discrepancy is quantitative: for the Fig. 3 parameters (ω_r=2π×0.3 MHz, κ=2π×10 kHz, ε_max=10 MHz, |α(t_f)|=10), the maximum of |α(t)| from Eq. (7) is about 5.3, so the target |α|=10 is never reached. The quoted t_min=1.01 μs matches the rotating-frame constant-drive formula α(t)=−(2ε_max/κ)(1−e^{−κt/2})e^{iθ}, a different model. Thus the claimed analytical time-optimal pulses and Fig. 3(a) do not follow from the stated control problem.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a Pontryagin Maximum Principle (PMP) framework for driving a dissipative LC resonator modeled by a linear Langevin equation. It derives energy- and time-optimal pulses for coherent-state preparation and then applies these pulses to dispersive qubit readout in circuit QED, reporting SNR curves for three critical photon numbers. The central advertised results are analytical energy-optimal pulses with exponential energy scaling, microsecond-scale time-optimal pulses, and SNR comparable to longitudinal-coupling readout at large critical photon numbers.","tokens_in":19922,"tokens_out":11428,"duration_ms":113296,"significance":"If the derivations were correct, the paper would provide a useful analytically solvable example of PMP-based control for an open linear system, with a direct application to qubit readout. The comparison with counter-diabatic driving and the robustness analysis are valuable, and the energy-optimal construction appears to follow the standard Gramian/adjoint formalism. However, the time-optimal part contains a load-bearing integration error, and the low-critical-photon-number readout is modeled in a regime the paper itself says is dominated by nonlinearities; these issues materially affect the claims.","major_comments":[{"comment":"Equation (7) does not solve the stated time-optimal control problem. With ε₁(t)=ε_max cosφ(t), ε₂(t)=ε_max sinφ(t) and φ_opt(t)=ω_r t+θ, the drive is ε(t)=ε_max e^{i(ω_r t+θ)}. Direct integration of Eq. (1) from α(0)=0 gives α(t)= [2ε_max e^{iθ}/(κ+4iω_r)] [e^{-(κ/2+iω_r)t} − e^{iω_r t}], not the expression in Eq. (7), which contains [e^{-(κ/2+iω_r)t}−1] and corresponds to a different drive. The discrepancy is quantitative: for the Fig. 3 parameters (ω_r=2π×0.3 MHz, κ=2π×10 kHz, ε_max=10 MHz) the amplitude |α(t)| from the correct solution is bounded by approximately 2ε_max·2/|κ+4iω_r| ≈ 5.3, so the claimed target |α(t_f)|=10 is never reached. The quoted t_min^f=1.01 μs and the time-optimal curves in Figs. 3 and S1 therefore do not follow from the model in Eq. (1).","section":"Minimization of pulse energy and time, Eq. (7)"},{"comment":"The adjoint equations used for the time-optimal derivation are internally inconsistent. The main text states ˙p_j = (−1)^{j+1}ω_r p_k − κ p_j/2, while the supplementary material gives ˙p_1 = κ/2 p_1 − ω_r p_2 and ˙p_2 = ω_r p_1 + κ/2 p_2. Neither matches the PMP canonical equation ˙p = −A^T p for the matrix A in Eq. (4), which would read ˙p_1 = κ/2 p_1 + ω_r p_2 and ˙p_2 = −ω_r p_1 + κ/2 p_2. Moreover, the stated solution p_j(t)=e^{-κt/2}[p_j(0)cos(ω_r t)+(−1)^{j+1}p_k(0)sin(ω_r t)] does not solve either form. These errors undermine the derivation of φ_opt(t)=ω_r t+θ and the associated minimal-time formula, and they also propagate to Eq. (12) and the time-optimal readout results.","section":"Minimization of pulse energy and time (adjoint equations); Supplementary Eqs. (32)–(33)"},{"comment":"The readout results for n_crit=1 are computed with the linear Langevin equation in which ω_r is replaced by χ_z, but the paper itself states in the discussion of Fig. 4(a) that for g=2π×1 GHz 'nonlinearities, such as Kerr effect, dominate the dynamics'. Using a linear dispersive model to generate the IQ trajectory and SNR in that regime is therefore not justified. The n_crit=1 SNR curve and the 'exotic trajectory' in Fig. 4(a) are outputs of a model that the authors acknowledge is invalid there; either the calculation must be replaced by a nonlinear treatment (e.g., including the Kerr term), or the claims should be restricted to the regimes where the linear model applies.","section":"Qubit-resonator interaction, Fig. 4 and Eqs. (11)–(13)"}],"minor_comments":[{"comment":"The text calls Eq. (1) the Langevin equation 'in the rotating frame', but the equation contains the explicit term −iω_r α. Please clarify the frame convention, because the sign and meaning of the drive phase φ(t) depend on it.","section":"Eq. (1) and surrounding text"},{"comment":"The claim of 'exponential scaling' of the energy cost is imprecise: Eq. (6) grows as 1/(1−e^{−κt_f}) ≈ 1/(κt_f) for large t_f, which is algebraic, not exponential, in that limit. Please state the asymptotic behavior more carefully.","section":"Fig. 2(a) and Eq. (6)"},{"comment":"In Eq. (13) the noise operator M_N is already defined as the centered operator M−⟨M⟩, so writing ⟨M_N²|ℓ⟩ for the variance is acceptable, but the notation ⟨M²_N|ℓ⟩ in the denominator is confusing because it suggests a second centering. Please clarify.","section":"Eq. (13)"},{"comment":"The expression for α_out(t) in the supplementary material appears to have a sign that is inconsistent with the input-output relation a_out(t)=a_in(t)+√κ a(t) used in the main text. Please check whether the sign in Eq. (49) should be positive or negative, since it feeds into the SNR calculation.","section":"Supplementary Eq. (49)"}],"recommendation":"major_revision","confidential_remarks":"The time-optimal solution error and the invalid use of the linear model at n_crit=1 are load-bearing, but both are potentially fixable within the scope of the manuscript if the authors re-derive Eq. (7) and restrict the readout claims to valid parameter regimes. The energy-optimal construction and the comparison with CD driving are likely salvageable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the application idea is fine, but I checked the two central formulas by substitution and neither solves the stated Langevin equation. The paper needs major correction before it can be reviewed seriously.\n\nWhat is genuinely useful here is the framing: treat the Langevin equation for a driven-damped resonator as a linear optimal-control problem, use the standard minimum-energy Gramian solution, and then apply those pulses to dispersive readout in cQED. The connection to counter-diabatic driving and the robustness study are the right kind of comparison. If the formulas were fixed, the ncrit=10 and 100 readout discussion might survive.\n\nThe problems start with the time-optimal section. The paper derives phi_opt(t)=omega_r t+theta and then writes Eq. (7) as the trajectory. Substituting that phase into Eq. (1) gives a different expression, with e^{i omega_r t} factors and a denominator kappa+4i omega_r; it is not the printed Eq. (7). The quoted t_min=1.01 us matches the rotating-frame constant-drive formula alpha=-(2 epsilon_max/kappa)(1-e^{-kappa t/2})e^{i theta}, not Eq. (7). The phase sign is also wrong: the resonant drive should be e^{-i omega_r t}, not e^{+i omega_r t}. So Fig. 3(a) and all time-optimal claims do not follow from the text.\n\nThe energy-optimal formula has a similar problem, which the reader's report missed. Set omega_r=0, kappa=t_f=1, alpha_f=1. The correct minimum-energy control is epsilon(t)= -kappa e^{kappa(t_f+t)/2} alpha_f/(e^{kappa t_f}-1), and it gives x(t_f)=1. The paper's Eq. (5) gives epsilon(t)= -2kappa e^{kappa(t+t_f)} alpha_f/(e^{kappa t_f}-1), which yields x(t_f) about 4.45, not 1. The energy cost is also off by a factor of 4, and the claimed exponential scaling in t_f is not what the formula gives; it is roughly 1/t_f at short times.\n\nFinally, the ncrit=1 readout case is an acknowledged inconsistency: the paper says Kerr nonlinearities dominate at g=2pi GHz, then uses the linear model to compute SNR. These are not subtle limitations; they are load-bearing.\n\nWho is this for? Someone who wants to see textbook linear control applied to cQED readout might get ideas, but as it stands the central results are not reliable. I would not send this to referees in current form. If the authors correct the equations and re-run the simulations, a revised version could warrant review.","headline":"The readout idea has some merit, but neither central analytical formula solves the stated problem, so the paper is not ready for referees.","tokens_in":20471,"tokens_out":27708,"would_cite":false,"duration_ms":243368,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A driven-dissipative LC resonator can be controlled by closed-form Pontryagin-optimal pulses whose energy cost saturates at a constant, and the same pulses give dispersive qubit readout with SNR comparable to longitudinal coupling at…","keywords":["quantum optimal control","Pontryagin maximum principle","open quantum systems","Langevin equation","circuit quantum electrodynamics","dispersive qubit readout","signal-to-noise ratio","shortcuts to adiabaticity"],"falsifier":"A calibrated homodyne measurement of the output field from a driven damped cavity, with $\\kappa$, $\\omega_r$, and $\\varepsilon(t)$ independently verified, should reproduce the analytic $\\alpha_{\\rm out}(t)$ from the input-output relation; a systematic discrepancy that grows with the qubit-resonator coupling $g$—already visible by $\\bar n_{\\rm crit}=1$—would falsify the linear-model readout claim.","tokens_in":19424,"feed_emoji":"⚛️","tokens_out":10971,"duration_ms":97418,"temperature":0.7,"pith_summary":"The paper tries to show that the driven-dissipative dynamics of an LC resonator coupled to a transmission line is not just numerically optimizable: its Langevin equation is a linear control system, and the Pontryagin Maximum Principle solves it in closed form. The energy-optimal pulses reach a target coherent state with an energy cost that saturates at $4\\kappa|\\alpha(t_f)|^2$ in the long-time limit, while standard Hahn and counter-diabatic pulses keep paying an energy cost that grows with time. The time-optimal pulses reach microsecond-scale transfer times under a maximum-amplitude constraint, which matters for practical drive-line limits. Reusing the same pulses in a dispersively coupled qubit-resonator system produces a readout signal-to-noise ratio that, for large critical photon numbers, is comparable to longitudinal-coupling readout within a time shorter than the cavity decay time. The paper itself flags that at $\\bar n_{\\rm crit}=1$ the coupling is so strong that Kerr nonlinearities dominate, so the linear model behind that case's IQ trajectory and SNR is not justified.","feed_headline":"Optimal pulses rival longitudinal readout at high photon numbers","feed_subtitle":"Pontryagin pulses steer dissipative cavities at fixed energy cost and give fast qubit readout.","key_machinery":"The machinery is the linear-control interpretation of the quantum Langevin equation, $\\dot x=\\hat A x+\\hat B u$, with state $x=(x_1,x_2)=(\\mathrm{Re}\\,\\alpha,\\mathrm{Im}\\,\\alpha)$, matrix $\\hat A$ containing the damping $-\\kappa/2$ and the rotation $\\omega_r$, and control $u=(\\varepsilon_1,\\varepsilon_2)$. The Pontryagin Hamiltonian for energy, $H_c=u^Tu+p^T\\dot x$, gives the stationarity condition $u=\\hat B^T p$, and the adjoint equation $\\dot p=-\\hat A^T p$ turns the boundary-value problem into the explicit pulse $u_{\\rm opt}(t)=\\hat B^T e^{\\hat A^T(t_f-t)}p(0)$, where $p(0)$ is fixed by the target state through a Gramian matrix that enforces the final boundary condition. For time minimization, the same Hamiltonian with cost $J_T=t_f$ and the amplitude constraint yields the phase $\\phi_{\\rm opt}(t)=\\omega_r t+\\theta$ and the closed trajectory for $\\alpha_{\\rm opt}(t)$. For readout, the substitution $\\omega_r\\to\\chi_z$ with $\\chi_z=\\pm\\chi$ carries the whole single-resonator solution over to the dispersive qubit-cavity system, so the pointer-state separation and the SNR are computed from the same analytic formulas.","core_discovery":"On the paper's own terms, the central discovery is that a dissipative cavity in circuit QED is not merely optimizable numerically: its Langevin equation is a linear control system, and the Pontryagin Maximum Principle solves that system in closed form for both cost functions. The energy-optimal drive $u_{\\rm opt}(t)=\\hat B^T e^{\\hat A^T(t_f-t)}p(0)$ has cost $J_E^{\\rm opt}(t_f)=4\\kappa|\\alpha(t_f)|^2/(1-e^{-\\kappa t_f})$, which stays at the finite value $4\\kappa|\\alpha(t_f)|^2$ in the adiabatic limit, whereas the Hahn and counter-diabatic reference pulses keep paying a growing energy cost. The time-optimal drive uses a maximum-amplitude constraint $|\\varepsilon(t)|\\le\\varepsilon_{\\max}$ and reaches a target coherent state in about a microsecond at MHz amplitudes. Reusing the same analytic solution with $\\omega_r$ replaced by the qubit-state-dependent shift $\\chi_z=\\pm\\chi$ gives the pointer-state dynamics of dispersive readout, and for $\\bar n_{\\rm crit}=10$ and $100$ the resulting SNR exceeds one on timescales shorter than the cavity decay time, which the authors state is comparable to longitudinal-coupling readout. The paper explicitly registers that at $\\bar n_{\\rm crit}=1$ the coupling is strong enough for Kerr nonlinearities to dominate, which invalidates the linear model behind that case's IQ trajectory and SNR curve.","pith_inferences":["Editorial inference: because the PMP solution only uses the linear form of the dynamics, any driven-dissipative system with the same structure—spin-qubit resonator readout, optomechanical cooling, or linearized flux qubits—can inherit the same closed-form pulses with an appropriate substitution for $\\hat A$.","Editorial inference: the comparison suggests the practical advantage of the energy-optimal pulse is largest when operations run for many cavity lifetimes; at very short $t_f$ all compared schemes pay exponential energy costs, so the benefit there is only the exact constraint satisfaction, not the asymptotic scaling.","Editorial inference: the authors' own validity bound implies the readout advantage should be tested in the large-$\\bar n_{\\rm crit}$ regime; a direct experiment at $\\bar n_{\\rm crit}=1$ would likely see the claimed SNR degraded by the Kerr-induced trajectory distortion the paper describes."],"forward_implications":["In the long-time limit the energy-optimal pulse costs a constant $4\\kappa|\\alpha(t_f)|^2$ to prepare a target coherent state, while the Hahn and counter-diabatic pulses keep adding energy linearly in $t_f$, so the optimal pulse becomes increasingly favourable the longer the operation.","Time-optimal pulses reach microsecond-scale transfer times at MHz drive amplitudes, making the scheme compatible with the amplitude ceilings of practical cQED drive lines.","For $\\bar n_{\\rm crit}=10$ and $100$, the analytic readout pulses give a signal-to-noise ratio that reaches order unity on timescales shorter than $\\kappa^{-1}$, which the paper reports as comparable to longitudinal-coupling readout.","The maximal SNR grows monotonically with $\\bar n_{\\rm crit}$ because larger target photon numbers produce larger pointer-state displacements and better state discrimination.","At $\\bar n_{\\rm crit}=1$ the paper's own analysis shows Kerr nonlinearities dominate, so the linear-model IQ trajectory and SNR for that parameter set are not physically predictive."],"supporting_citations":[{"why":"Supplies the open-system counter-diabatic driving and cQED benchmark that the optimal pulses are designed to beat.","marker":"[15]"},{"why":"Supplies the linear-control and inverse-engineering formulation from which the analytic Pontryagin pulses are derived.","marker":"[34]"},{"why":"Defines the longitudinal-coupling readout scheme whose signal-to-noise ratio this paper compares against.","marker":"[36]"},{"why":"Provides a shortcut-to-adiabaticity readout benchmark whose SNR is also compared against the optimal pulses.","marker":"[37]"},{"why":"Contains the derivations of the Langevin equation, input-output relation, PMP solution, and SNR comparisons used in the main text.","marker":"[40]"},{"why":"Supplies the dispersive Hamiltonian (10) that defines the qubit-resonator readout model.","marker":"[45]"},{"why":"Introduces the dispersive readout framework in circuit QED that motivates the $\\omega_r\\to\\chi_z$ substitution.","marker":"[46]"},{"why":"Defines the critical photon number $\\bar n_{\\rm crit}$ used to set the target states and readout limits.","marker":"[47]"},{"why":"Used to ground the claim that Josephson nonlinearities such as the Kerr effect dominate resonator dynamics at strong coupling.","marker":"[49]"},{"why":"Used with [49] to support the statement that Jaynes-Cummings nonlinearity dominates at high drive, motivating the linear-model caveat.","marker":"[50]"}],"fun_headline_variants":["Pontryagin pulses tame dissipative cavities at fixed energy","Closed-form control for open circuit QED pulses","Optimal pulses rival longitudinal readout at high photon number","Energy-saturating pulses beat shortcut-to-adiabaticity","Fast qubit readout via Pontryagin-optimal cavity drive"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the dispersive Hamiltonian and the linear Langevin equation with $\\omega_r$ replaced by the Stark shift $\\chi_z$ stay valid all the way up to the stated critical photon numbers, and the paper's own discussion of Fig. 4(a) says this fails at $\\bar n_{\\rm crit}=1$, where $g=2\\pi\\times 1$ GHz makes Kerr nonlinearities dominate.","fun_headline_variants_meta":{"raw":{"variants":["Pontryagin pulses tame dissipative cavities at fixed energy","Closed-form control for open circuit QED pulses","Optimal pulses rival longitudinal readout at high photon number","Energy-saturating pulses beat shortcut-to-adiabaticity","Fast qubit readout via Pontryagin-optimal cavity drive"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000344,"raw_usage":{"total_tokens":1901,"prompt_tokens":965,"completion_tokens":936,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":581,"completion_tokens_details":{"reasoning_tokens":853}},"tokens_in":581,"tokens_out":936,"duration_ms":8725,"temperature":1.0,"reasoning_tokens":853,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T23:31:40.718207+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A calibrated homodyne measurement of the output field from a driven damped cavity, with $\\kappa$, $\\omega_r$, and $\\varepsilon(t)$ independently verified, should reproduce the analytic $\\alpha_{\\rm out}(t)$ from the input-output relation; a systematic discrepancy that grows with the qubit-resonator coupling $g$—already visible by $\\bar n_{\\rm crit}=1$—would falsify the linear-model readout claim.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the open-system counter-diabatic driving and cQED benchmark that the optimal pulses are designed to beat."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the linear-control and inverse-engineering formulation from which the analytic Pontryagin pulses are derived."},{"cited_title":"Ansel, E","cited_arxiv_id":null,"evidence_quote":"Defines the longitudinal-coupling readout scheme whose signal-to-noise ratio this paper compares against."},{"cited_title":"Didier, J","cited_arxiv_id":null,"evidence_quote":"Provides a shortcut-to-adiabaticity readout benchmark whose SNR is also compared against the optimal pulses."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Contains the derivations of the Langevin equation, input-output relation, PMP solution, and SNR comparisons used in the main text."},{"cited_title":"Mandelstam and I","cited_arxiv_id":null,"evidence_quote":"Supplies the dispersive Hamiltonian (10) that defines the qubit-resonator readout model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the dispersive readout framework in circuit QED that motivates the $\\omega_r\\to\\chi_z$ substitution."},{"cited_title":"Blais, R.-S","cited_arxiv_id":null,"evidence_quote":"Defines the critical photon number $\\bar n_{\\rm crit}$ used to set the target states and readout limits."},{"cited_title":"Shillito, A","cited_arxiv_id":null,"evidence_quote":"Used to ground the claim that Josephson nonlinearities such as the Kerr effect dominate resonator dynamics at strong coupling."},{"cited_title":"Siddiqi, R","cited_arxiv_id":null,"evidence_quote":"Used with [49] to support the statement that Jaynes-Cummings nonlinearity dominates at high drive, motivating the linear-model caveat."}],"review_version":1}