{"id":"2082418a-2f94-409b-b639-2c93e0086616","arxiv_id":"2608.12454","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Dynamical freezing in a driven fluxonium circuit does not suppress quasiparticle-induced decay, and the paper identifies drive-frequency operating windows that balance freezing quality against pair-breaking and tunneling losses.","lead":"Periodic driving can freeze a superconducting fluxonium circuit into a nearly harmonic oscillator, but the paper shows this does not automatically stop quasiparticle-induced energy loss. It maps drive frequencies that avoid the worst loss resonances, giving practical design targets for experiments.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The quantitative predictions and proposed safe operating frequencies assume drive-generated quasiparticles are promptly removed from the junction; if they linger, local x_qp rises and the quoted 14 μs / 0.953 numbers are not guaranteed.","rationale":"The paper's central claim—that dynamical freezing does not generically suppress quasiparticle-induced dissipation—is well supported by the exact-diagonalization rate maps and by the parameter-free reduced-Sambe predictions of the resonance ridges, including the quantitative two-channel splitting (predicted 0.8408 GHz versus roughly 0.8 GHz read from the figure). My review focused on what would have to be true for the quantitative operating recommendations to hold. The paper explicitly assumes generated quasiparticles are promptly removed from the junction. This assumption is load-bearing because pair generation is a source of quasiparticles at a rate comparable to the T1 rate; if those quasiparticles linger near the junction, the local x_qp entering the tunneling rate is no longer the fixed 1e-6, and the T1 and fidelity numbers in Fig. 4 shift downward. The same fragility is visible in the Supplemental Material's finite-width analysis, where the branch-0 lifetime changes by a factor of three at low frequencies depending on quasiparticle spectral details. I considered whether the branch-labeling ambiguity near resonances or the 25 MHz detuning margin was a more central flaw; neither threatens the qualitative conclusion that resonances must be avoided, and both are secondary calibration issues. The qualitative claim and the resonance-based framework are secure; the specific safe-frequency intervals are conditional on quasiparticle transport behavior. This matches the reader's weakest assumption, so I recommend no change to the CONDITIONAL verdict.","tokens_in":18723,"tokens_out":14112,"duration_ms":134087,"concrete_test":"Build a minimal kinetic model for the quasiparticle population near the junction: source rate = 2 Γ_pair_0(f_d) for pair creation, with sinks from diffusion out of the junction volume, recombination, and trapping using parameters typical of aluminum fluxonium (e.g., recombination time ~100 μs, diffusion length ~100 μm). Solve for the steady-state x_qp_local at f_d = 21.6785 GHz and 21.7290 GHz, then recompute Γ_tunnel with x_qp = x_qp_background + x_qp_local and re-extract T1,01 and F01. If x_qp_local remains ≲ 2×10^-6, the 14 μs operating points survive within a factor of two; if x_qp_local rises above ~10^-5, the proposed frequencies are not safe and the quantitative claim requires additional quasiparticle mitigation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing assumption is the paper's explicit statement in the Discussion: 'we have assumed that generated quasiparticles are promptly removed from the junction.' This is not a minor footnote: the pair-generation channel Γ_pair_0 acts as a source of new quasiparticles at a rate comparable to the inverse T1 (tens of kHz), and the tunneling channel Γ_tunnel is linear in the local dimensionless density x_qp. If generated quasiparticles are not promptly removed, the steady-state x_qp near the junction rises above the representative value of 1e-6 used in Fig. 4, so the computed T1 ≈ 14 μs, fidelity ≈ 0.953, and the proposed operating frequencies 21.6785 GHz and 21.7290 GHz are not guaranteed. The same fragility is already visible in the Supplemental Material's finite-width analysis, where the branch-0 lifetime changes by a factor of three at low frequency (2.03 μs to 6.06 μs at f_d = 11.8 GHz) depending on the quasiparticle spectral distribution. Importantly, the resonance landscape of Figs. 2–3—the 2Δ/n pair-breaking thresholds and the Magnus-organized tunneling ridges—is independent of this assumption, so the qualitative central claim that dynamical freezing does not generically suppress quasiparticle dissipation is not at stake. What is at stake is the quantitative embodiment of the claim and the actionable advice to experimentalists.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a Floquet theory of quasiparticle-induced dissipation in a periodically driven fluxonium circuit operating at the frozonium freezing point. Using Floquet Fermi's Golden Rule with exact-diagonalization matrix elements, the authors compute two rate channels: drive-assisted Cooper-pair breaking and tunneling of pre-existing quasiparticles. They identify gap-breaking thresholds at 2Δ/n, multiphoton-resonance enhancements at lower drive frequencies, and connected resonance ridges in the tunneling rate that are organized by the harmonic Floquet-Magnus spectrum. A reduced-Sambe model predicts the hybridized resonance loci with a parameter-free 0.8408 GHz splitting consistent with the 0.8 GHz read from the figure. The paper closes by proposing drive frequencies near 21.7 GHz at the freezing amplitude, with T1 ≈ 14 μs and fidelity ≈ 0.953 for x_qp = 10^-6, and emphasizes that dynamical freezing does not generically suppress quasiparticle poisoning.","tokens_in":18950,"tokens_out":7649,"duration_ms":75360,"significance":"If the results hold, this is the first quasiparticle-loss theory for the frozonium regime and provides a practical framework for choosing drive parameters away from harmful resonances. The numerical work is unusually transparent: Hilbert-space dimension N=200, Nt=512 time samples, solver tolerances 10^-12, and explicit sideband sets are stated. The reduced-Sambe prediction is parameter-free and checked against exact diagonalization, and the branch-labeling algorithm is clearly specified. The central qualitative claim---that dynamical freezing does not generically suppress quasiparticle dissipation---is supported by the resonance landscape and is independent of the prompt-removal assumption. The quantitative operating-window predictions, however, depend on x_qp and on the assumption that drive-generated quasiparticles are promptly removed from the junction, which limits the actionable conclusions.","major_comments":[{"comment":"The paper explicitly assumes in the Discussion that 'generated quasiparticles are promptly removed from the junction.' This assumption is load-bearing for the quantitative claims: the pair-generation channel Γ_pair_0 is a source of quasiparticles at a rate comparable to the inverse T1, while Γ_tunnel_0 is linear in the local dimensionless density x_qp. If generated quasiparticles linger or diffuse back, the steady-state local density rises above the assumed x_qp = 10^-6, so the quoted T1 ≈ 14 μs, the fidelity ≈ 0.953, and the proposed operating frequencies 21.6785 GHz and 21.7290 GHz in Fig. 4(b) are not guaranteed. The resonance landscape in Figs. 2 and 3 is unaffected, but the actionable operating-window advice depends on this assumption. The Supplemental Material's finite-width analysis already shows a factor-of-three lifetime change at f_d = 11.8 GHz depending on the quasiparticle spectral distribution, underscoring the sensitivity of the quantitative rates. Please either add a kinetic or transport estimate that closes this loop or explicitly restate the operating-frequency proposal as conditional on prompt removal.","section":"Discussion; Implications for experiments (Fig. 4)"},{"comment":"The reduced-Sambe model successfully predicts the two-channel avoided-crossing splitting, but the text notes that the inner dressed state of the three-channel family in Fig. 3(d) is shifted by about 0.15 GHz by the additional |M_36,-4⟩ Sambe state. This is a stated limitation of the truncation, not an error, but it should be reflected in the main-text claim that the ridge structure is quantitatively organized by the harmonic Magnus spectrum. Please either include the fourth Sambe state in the reported comparison or present the inner-locus prediction as approximate rather than fully determined.","section":"End Matter, Magnus Resonance Theory"}],"minor_comments":[{"comment":"The Pareto-front construction in Fig. 4(a) is described in one sentence but not defined algorithmically; please specify how the front was computed over the frequency scan.","section":"Fig. 4"},{"comment":"The solver tolerances are stated, but the QuSpin propagation method used to solve Eq. (S4) is not named; please identify the solver and report a convergence check in N and N_t.","section":"Supplemental Material"},{"comment":"The prefactor 16 E_J/h in the pair-generation rate deserves a one-line derivation or a reference, since the standard Floquet Golden Rule prefactor is not immediately obvious from the definition of E_J.","section":"End Matter"},{"comment":"The notation M_{n/p} is used for p-photon resonances between Magnus levels M_n and M_0; please define it once in a prominent place, such as the first mention in the main text or the Fig. 3 caption.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid contribution to the theory of driven superconducting circuits, and the resonance-physics core is convincing. My main editorial concern is the distance between the robust qualitative message (resonance landscape exists) and the quantitative operating-frequency proposal, which depends on the prompt-removal assumption that the authors themselves flag but do not model. A revision that either adds a minimal kinetic estimate or carefully re-scopes the operating-window claims would resolve this."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Solid theory paper that answers a real question about the frozonium regime. The headline result is that dynamical freezing at the Josephson null does not by itself protect the qubit from quasiparticle-induced dissipation. They calculate both drive-assisted pair breaking and tunneling of pre-existing quasiparticles with Floquet Fermi's Golden Rule, and they identify the structures that matter: 2Δ/n gap-breaking thresholds at high frequency and multiphoton resonances at lower frequency. What's genuinely new is the consequence of the harmonic Magnus spectrum at the freezing point: the resonance family M n/1, M 2n/2, M 3n/3... becomes degenerate, so the tunneling rate landscape develops connected avoided-crossing ridges. That is a specific, falsifiable prediction absent from the transmon literature.\n\nThe numerics are careful: N=200, Nt=512, tolerances 10^-12, and the reduced-Sambe prediction of a 0.8408 GHz splitting against 0.8 GHz read from the figure is strong evidence the resonance loci are computed correctly. No free constants are used to get that agreement. The citation pattern is appropriate and builds directly on the frozonium and driven-quasiparticle work.\n\nThe soft spot is where they move from landscape to operating advice. The recommended frequency windows near 21.7 GHz, the T1≈14 μs figure, and the fidelity 0.953 depend on two assumptions: x_qp=10^-6 (standard and reasonable) and prompt removal of drive-generated quasiparticles from the junction (load-bearing). If those quasiparticles linger, the local density rises and the quantitative numbers are not guaranteed. The authors are transparent about this in the Discussion and call for a transport model, but it means the practical design rules are provisional. The finite-width correction in the Supplement shows a factor-of-three lifetime change at low frequency, which is a lesser effect but points the same way. None of this undermines the central qualitative claim that freezing does not generically suppress quasiparticle loss; that part is well supported.\n\nThis paper is for theorists working on Floquet superconducting circuits and experimentalists seeking drive parameters for fluxonium. It deserves a serious referee. I would send it, but I would expect referees to ask for the prompt-removal assumption to be relaxed or strongly defended before the specific operating frequencies are treated as safe.","headline":"Careful Floquet theory showing dynamical freezing does not generically suppress quasiparticle loss; the resonance landscape is solid, but the quantitative operating windows hinge on prompt quasiparticle removal.","tokens_in":19527,"tokens_out":3409,"would_cite":true,"duration_ms":29220,"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":"Dynamically freezing a fluxonium circuit does not suppress quasiparticle loss; the drive frequency must be chosen around Floquet resonances.","keywords":["frozonium","dynamical freezing","Floquet Fermi golden rule","quasiparticle poisoning","fluxonium qubit","multiphoton resonance","Cooper-pair breaking","superconducting circuit relaxation"],"falsifier":"Measure the relaxation time $T_1$ of an aluminum fluxonium driven at the freezing amplitude $\\phi_{\\rm ac}\\simeq 2.4048$ while sweeping the drive frequency across 21.7 GHz: the paper predicts a sharp $T_1$ dip at the resonance near 21.704 GHz and broad safe intervals around 21.6785 GHz and 21.7290 GHz with $T_1\\approx 14$ $\\mu$s. Seeing no such frequency structure, or a $T_1$ far below 14 $\\mu$s because generated quasiparticles persist, would refute the quantitative predictions.","tokens_in":18456,"feed_emoji":"⚛️","tokens_out":8131,"duration_ms":70925,"temperature":0.7,"pith_summary":"The paper asks whether dynamical freezing—driving a fluxonium circuit at a special amplitude where the Josephson nonlinearity cancels, leaving a nearly harmonic Floquet oscillator—also protects the circuit from quasiparticle-induced loss. It argues it does not, generically: drive-assisted Cooper-pair breaking is controlled by $2\\Delta/n$ multiphoton thresholds at high drive frequency and by multiphoton resonances at low frequency, while tunneling of pre-existing quasiparticles shows resonance ridges inherited from the harmonic Magnus spectrum near the freezing point. The practical consequence is a frequency-selection problem: experiments must place the drive away from both the gap-breaking thresholds and the Floquet resonances. For a representative aluminum device the paper identifies two operating frequencies near 21.7 GHz, each roughly 25 MHz from the nearest resolved resonance, with a relaxation time of about 14 microseconds and a wave-function fidelity near 0.953.","feed_headline":"Freezing a qubit's anharmonicity does not stop quasiparticle loss","feed_subtitle":"Pair-breaking thresholds and multiphoton resonances set safe drive windows near 21.7 GHz, giving a ~14 μs lifetime.","key_machinery":"The machinery is Floquet Fermi's Golden Rule applied to the microscopic tunneling Hamiltonian $\\hat H_T(t)=\\sum \\tau_{kk'} e^{i\\hat\\varphi_J(t)/2}c^\\dagger_{Rk\\sigma}c_{Lk'\\sigma}+{\\rm h.c.}$, written in the same irrotational gauge as the driven fluxonium Hamiltonian $H_{\\rm mov}(t)=4E_C\\hat n^2+(E_L/2)\\hat\\varphi^2-E_J\\cos(\\hat\\varphi-\\varphi_{\\rm ext}-\\Theta(t))$, with $\\Theta(t)=\\phi_{\\rm ac}\\sin(\\omega t)$. The BCS coherence factors enter through two channels, $\\cos(\\hat\\varphi_J/2)$ and $\\sin(\\hat\\varphi_J/2)$; the sideband amplitudes $M^{(q)}_{p,\\beta\\alpha}$ feed the rates and enforce energy conservation through $f^{(q)}_{\\alpha\\beta}=(\\epsilon_\\alpha-\\epsilon_\\beta)/h+q f_d$. The reference for interpreting the rates is the leading Magnus Hamiltonian $H_M^{(0)}$ and its amplitude-continuous eigenstates; resonances are identified independently through period-averaged energies, and their loci are predicted by the reduced Sambe matrix $C^{(n,p)}_{jk}$ whose generalized eigenvalues give the dressed resonance frequencies.","core_discovery":"The central discovery is that the freezing points of the Floquet-Magnus expansion—amplitudes $\\phi_{\\rm ac}$ where $J_0(\\phi_{\\rm ac})=0$ and the leading effective Hamiltonian becomes $H_{\\rm quad}=4E_C\\hat n^2+(E_L/2)\\hat\\varphi^2$ with frequency $\\nu=\\sqrt{8E_CE_L}/h$—do not single out quasiparticle processes. The pair-generation rate $\\Gamma^{\\rm pair}_0$ inherits discontinuous openings at the sideband frequencies $f_d \\approx 2\\Delta/(n h)$; the Magnus reference reproduces the rate down to roughly 20 GHz, and the freezing point brings no special suppression of the pair rate. At lower drive frequencies, spikes in $\\Gamma^{\\rm pair}_0$ coincide with avoided crossings of period-averaged energies, i.e., multiphoton resonances $E^M_\\alpha-E^M_\\beta = n h f_d$, whose density grows as $f_d$ falls. Tunneling of pre-existing quasiparticles lacks the $2\\Delta/n$ thresholds but shows connected bright ridges formed when the harmonic Magnus spectrum makes the commensurate families $M_{n}/1, M_{2n}/2, M_{3n}/3,\\dots$ degenerate at the freezing point; resonant hybridization splits these into avoided-crossing loci, quantitatively captured by a reduced Sambe description. Operating near the freezing point therefore requires choosing $f_d$ that balances the approach to the harmonic limit against these Floquet-enhanced loss channels.","pith_inferences":["Editorial inference: the same resonance-assisted poisoning mechanism should appear in any driven Josephson circuit whose Floquet spectrum contains near-degenerate multiphoton levels; fluxonium is a clean test bed because its charge-insensitivity avoids the transmon's sector-structure complication.","Editorial inference: the quantitative lifetime estimates depend on prompt quasiparticle removal; if experiments show $T_1$ below the predicted 14 $\\mu$s at the proposed frequencies, the likely cause is local quasiparticle accumulation rather than a failure of the resonance map.","Editorial inference: the avoided-crossing ridges in the tunneling rate could be used as a spectroscopic probe of the Floquet-Magnus spectrum, since the rate map reveals level crossings that are difficult to see in energy measurements alone.","Editorial inference: a dedicated experiment sweeping $\\phi_{\\rm ac}$ at fixed low $f_d$ should observe rate ridges along the predicted $M_{n}/p$ loci, providing a direct test of the harmonic-Magnus organization of the loss channels."],"forward_implications":["At the freezing amplitude, quasiparticle pair generation is not suppressed: the rate still opens at the $2\\Delta/n$ sideband thresholds, so choosing $\\phi_{\\rm ac}\\approx 2.4048$ is not by itself a protection strategy.","The usable drive frequencies at the freezing point are interrupted by resonance ridges; scanning $f_d$ near 21.7 GHz reveals a resonance at about 21.704 GHz separating two safe intervals centered at 21.6785 GHz and 21.7290 GHz.","At those frequencies a representative aluminum device reaches $T_{1,01}\\approx 14\\,\\mu$s and average state fidelity $\\bar F_{01}\\approx 0.953$, roughly 25 MHz away from the nearest resolved resonance.","Lowering $f_d$ has competing effects: it suppresses direct multiphoton pair breaking but increases the density of Floquet resonances, so the optimal frequency is an intermediate compromise.","Raising the superconducting gap shifts the pair-breaking thresholds upward and widens the range of frequencies where high oscillator fidelity is possible, making gap and drive frequency coupled design parameters."],"supporting_citations":[{"why":"Introduces the frozonium regime, the freezing points where $J_0(\\phi_{\\rm ac})=0$ makes the effective Hamiltonian harmonic, and the irrotational gauge used here.","marker":"[15]"},{"why":"Supplies the Floquet Fermi Golden Rule used to compute drive-assisted quasiparticle transition rates.","marker":"[45]"},{"why":"Provides the companion Floquet quasiparticle-rate formalism for microwave-driven superconducting qubits, including sideband-assisted pair breaking and tunneling.","marker":"[46]"},{"why":"Justifies the coordinated charge-plus-flux drive and co-moving-frame gauge transformation that produce the phase-driven Hamiltonian.","marker":"[47]"},{"why":"Gives the BCS coherence-factor rate formulas, the cold-edge tunneling form factors, and the $x_{\\rm qp}$ normalization used for quasiparticle densities.","marker":"[26]"},{"why":"Provides the Floquet-Magnus high-frequency expansion whose multiphoton resonance condition organizes the rate ridges.","marker":"[22]"}],"fun_headline_variants":["Freezing fluxonium doesn't stop quasiparticle loss","Frozen anharmonicity, alive quasiparticle loss","Drive resonances poison frozen fluxonium","Freezing point fails to block quasiparticles","Quasiparticle loss survives fluxonium freezing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The calculation assumes that quasiparticles generated by the drive are promptly removed from the junction, keeping the local density at the assumed level of $x_{\\rm qp}=10^{-6}$; if they linger, the predicted lifetimes, fidelities, and safe frequency windows are not guaranteed.","fun_headline_variants_meta":{"raw":{"variants":["Freezing fluxonium doesn't stop quasiparticle loss","Frozen anharmonicity, alive quasiparticle loss","Drive resonances poison frozen fluxonium","Freezing point fails to block quasiparticles","Quasiparticle loss survives fluxonium freezing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000261,"raw_usage":{"total_tokens":1645,"prompt_tokens":1048,"completion_tokens":597,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":664,"completion_tokens_details":{"reasoning_tokens":524}},"tokens_in":664,"tokens_out":597,"duration_ms":5489,"temperature":1.0,"reasoning_tokens":524,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:08:11.515052+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the relaxation time $T_1$ of an aluminum fluxonium driven at the freezing amplitude $\\phi_{\\rm ac}\\simeq 2.4048$ while sweeping the drive frequency across 21.7 GHz: the paper predicts a sharp $T_1$ dip at the resonance near 21.704 GHz and broad safe intervals around 21.6785 GHz and 21.7290 GHz with $T_1\\approx 14$ $\\mu$s. Seeing no such frequency structure, or a $T_1$ far below 14 $\\mu$s because generated quasiparticles persist, would refute the quantitative predictions.","supporting_citations":[],"review_version":1}