{"id":"f0715aa6-d272-48d4-970a-cfb8ddcf2493","arxiv_id":"2504.15901","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":1,"one_line_summary":"Fluxonium qubit readout and reset are demonstrated without a resonator, using a planar filter and dissipation engineering, achieving QND readout with N_QND=1.6e3 and reset fidelity above 99.5% in 250 ns.","lead":"Researchers demonstrated a resonator-free way to read out and reset a superconducting fluxonium qubit on a chip, using a microwave filter to engineer where the qubit loses energy. The method shelves the qubit into a bright transition for fluorescence readout and pumps excitations to a fast-decaying level for reset.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified.","rationale":"The reader identified the Markovian filter treatment and the placement of the |g>-|h> transition in the passband as the weakest assumption. I find this assumption well supported: the reflection data fit Eq. (3) with a single external decay rate, and the reset data independently confirm that the |h>-to-|g> decay is fast enough for high-fidelity reset in 200 ns. The verifiable marker of a non-Markovian or hybridized filter, namely a reflection lineshape that deviates from the single-Lorentzian prediction, is not observed. The remaining limitations—the unconstrained quasiparticle density and the reset demonstration only from |e>—are real but minor: the quasiparticle fit is not used in the central figures of merit, and the by-construction argument for unconditional reset is robust because any dark state in the driven {|e>,|f>,|h>} manifold has a significant fast-decaying |h> component for the stated drive parameters. I therefore see no load-bearing concern that would change the ACCEPT verdict, and I disagree that the filter assumption is the weakest point. The proposed QND correlation test would be a worthwhile strengthening of the demonstration but is not required for the validity of the present claims.","tokens_in":16245,"tokens_out":26827,"duration_ms":248088,"concrete_test":"Run a two-measurement QND correlation sequence: prepare |g> or |e>, perform a 15-µs fluorescence readout and record the outcome, then perform a second identical readout and compute the conditional probability that the second outcome matches the first. If the match probability is consistent with exp(-τ/T1_meas) ≈ 0.72 for τ = 15 µs, the QNDness is limited only by the measured T1_meas; a significantly lower match probability would reveal additional measurement-induced backaction not captured by the population-decay characterization.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claims are supported by direct measurement. The Markovian treatment of the filter is validated by the clean Lorentzian fit of Eq. (3) and by the consistency of the reset dynamics with the predicted |h>-to-|g> decay. The QNDness metric N_QND = Γr T1_meas is built from two independent measurements (reflection fit and exponential decay), and the readout-induced decay time of 0.55 ms separates measurement backaction from the intrinsic T1. The main residual limitations are acknowledged in the manuscript: the quasiparticle-density fit in Appendix B 3 is not independently constrained, and the reset is explicitly demonstrated only from |e> with unconditionality argued by construction. Neither breaks the central claim. The reset protocol drives |e>-|f> and |f>-|h> with a fast |h>-to-|g> decay; the only possible coherent population-trapping state is a superposition of |e> and |h> with no |f> component, and for the stated drive parameters (|e>-|f> drive an order of magnitude larger than the readout drive) this dark state is dominated by |h> and therefore decays quickly. The experimental residual at 200 ns is already below 1%, consistent with this picture. I find no load-bearing flaw in the argument.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a resonator-free fluorescence readout and an all-microwave unconditional reset of a fluxonium qubit, enabled by a planar CPW filter that suppresses computational-basis decay while allowing fast decay of the readout and reset transitions. The authors measure Γr/2π = 5.4(1) MHz from the reflection of the |e⟩–|f⟩ transition, T1 = 51(1) µs, and T1^meas = 46(1) µs under the readout drive, yielding N_QND = 1.6×10^3, an order of magnitude beyond a previous fluorescence-readout implementation. They also demonstrate a two-tone reset protocol with corrected residual excitation of 0.7(2)% at 200 ns and report better than 0.5% at 250 ns. The paper includes a first-principles master-equation derivation of the reflection coefficient, a rate-equation model of the readout-induced decay, and a transparent correction for the state transition during readout.","tokens_in":16399,"tokens_out":17524,"duration_ms":151305,"significance":"If the results hold, this work provides a scalable alternative to dispersive readout and reset for fluxonium, with the advantages of eliminating the resonator degree of freedom and associated modeling complexity, and using only microwave drives at orders-of-magnitude lower power than existing all-microwave reset schemes. The experimental characterization is careful: the external decay rate of the readout transition is measured directly from a Lorentzian reflection fit, the intrinsic T1 and readout-induced decay are separately quantified, and the reset residual is corrected for the finite readout-induced decay. The paper also identifies the residual non-QNDness as due to quasiparticle tunneling with an estimated density within the literature range, while acknowledging that the physical mechanism is not fully understood.","major_comments":[{"comment":"The claim that the reset protocol is unconditional is not fully established. The two-tone drive scheme on the |e⟩–|f⟩ and |f⟩–|h⟩ transitions possesses a coherent dark state |D⟩ ∝ Ω_fh|e⟩ − Ω_ef|h⟩ that is decoupled from both drives; its decay is controlled by its |h⟩ component and the |h⟩–|g⟩ relaxation rate. The manuscript does not state the relative amplitudes of the two reset drives, and the statement that the optimal |e⟩–|f⟩ drive is an order of magnitude larger than the readout drive does not determine the ratio Ω_ef/Ω_fh. Since the experiment prepares only |e⟩, the assertion that any excitation in {|e⟩,|f⟩,|h⟩} is unconditionally reset is supported by construction rather than by a complete dynamical analysis. The authors should add an explicit dark-state analysis, report the measured drive amplitudes, and ideally demonstrate reset from at least one additional initial state (e.g., |f⟩ or a superposition) to substantiate the unconditionality claim.","section":"Sec. III C and Appendix C"}],"minor_comments":[{"comment":"Using the quoted central values T1 = 51 µs and T1^meas = 46 µs gives (1/T1^meas − 1/T1)^−1 ≈ 0.47 ms, not 0.55 ms; please correct the arithmetic or specify the values used in the calculation.","section":"Sec. III B, Eq. (6)"},{"comment":"The sentence 'The qubit is prepared in |e⟩ by calibrated 1-µs reset pulses, followed by a π pulse between |g⟩ and |e⟩' is confusing; presumably the reset pulses prepare |g⟩, after which the π pulse excites the qubit to |e⟩. Please rephrase for clarity.","section":"Sec. III C"},{"comment":"The corrected residual excitation for a 250 ns reset pulse is not reported numerically; the text says only that it is comparable to statistical errors (~0.3%). Since the abstract and conclusion claim below 0.5% at 250 ns, please provide the actual corrected value and its uncertainty.","section":"Sec. III C"},{"comment":"Please report the measured values of the optimized |e⟩–|f⟩ and |f⟩–|h⟩ drive amplitudes (for example in frequency units) so that the dark-state composition and the reset dynamics can be checked quantitatively.","section":"Appendix C"},{"comment":"The filter is designed as a band-pass filter with center 4.6 GHz and bandwidth 1.0 GHz, yet the |e⟩–|f⟩ transition at 5.369 GHz lies outside this nominal passband. The text mentions high-pass-like transmission and transmission at integer multiples of the center frequency; please clarify explicitly how the 5.369 GHz and 7.814 GHz transitions fall within the filter transmission bands.","section":"Sec. II and Fig. 2(d)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is technically sound and the experiments appear carefully executed. The main concern is the support for the 'unconditional' reset claim, which I believe can be addressed by adding a dark-state analysis and reporting the drive amplitudes; a demonstration from another initial state would further strengthen the paper. The numerical inconsistency in Eq. (6) suggests that a careful proofreading pass is needed before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague — this is a solid paper, and the reader's take is right: the central claims are supported by direct measurement. What's genuinely new is the combination of a planar on-chip filter with fluorescence readout on the |e>–|f> transition, which closes the shelving cycle and gives an order-of-magnitude improvement in QNDness over Cottet et al., together with an all-microwave reset that exploits the same engineered decay. No resonator, no flux pulses; that's a meaningful architectural step for fluxonium.\n\nThe paper does several things well. The reflection coefficient is derived from a first-principles master equation in Appendix B1 with only Γr as a fitted parameter, and the Lorentzian fit shown in Fig. 3 is clean. T1 = 51 µs and Γr/2π = 5.4 MHz are directly measured, so the ratio 1.7e3 is not a fitting artifact. N_QND = Γr T1_meas uses two independent measurements, and the correction for readout-induced decay is transparent. The reset protocol is simple and robust: two drives pump any population in {|e>,|f>,|h>} up to |h>, which then decays through the filter to |g>. The unconditionality argument by construction holds up—the only potential dark state is dominated by |h> for their drive parameters, and the measured sub-1% residual at 200 ns is consistent.\n\nSoft spots are minor. The quasiparticle density xqp = 4e-7 is a fitted parameter, not independently constrained; the authors acknowledge the tension with the T1-derived upper bound in the junction array and cite precedent for elevated single-junction densities. Plausible, but not proven. Second, the reset is demonstrated only from an initial |e> state; unconditionality is argued, not directly measured from |f> or |h>. Third, QNDness is inferred from T1_meas under readout rather than a direct two-measurement QND correlation sequence. None of these break the central claim. The readout speed is still modest (15 µs integration), but the paper is honest about that and suggests a path to increase Γr.\n\nThe paper deserves a serious referee and, I think, acceptance after minor revision. It will be of most value to experimental groups working on resonator-free fluxonium readout and reset. I'd bring it to the reading group and would cite it.","headline":"Solid resonator-free fluxonium readout and reset with clean derivations and honest caveats; the quasiparticle attribution and reset-from-|e>-only are the main soft spots.","tokens_in":17045,"tokens_out":4032,"would_cite":true,"duration_ms":34678,"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":"This paper demonstrates that a fluxonium qubit can be read out non-destructively and reset to better than 99.5% fidelity using only a planar filter and its own internal transitions, with no readout resonator.","keywords":["fluxonium","fluorescence readout","quantum non-demolition readout","unconditional reset","dissipation engineering","coplanar waveguide filter","superconducting qubits","all-microwave reset"],"falsifier":"Measure the $|h\\rangle$–$|g\\rangle$ decay rate directly by preparing $|h\\rangle$ and watching its fluorescence: the reset model predicts a lifetime of about $1/(2\\pi\\times6.5\\,\\mathrm{MHz}) \\approx 24$ ns, set by the filter passband; a large deviation, or a reset fidelity that does not track the filter transmission at 7.814 GHz, would break the filter-as-environment assumption.","tokens_in":15982,"feed_emoji":"⚛️","tokens_out":12507,"duration_ms":103931,"temperature":0.7,"pith_summary":"This paper claims that a superconducting fluxonium qubit can be read out and reset without the usual readout resonator, by engineering how fast each of its internal transitions decays. A single on-chip coplanar-waveguide filter lets the $|e\\rangle$–$|f\\rangle$ transition decay quickly for fluorescence readout while keeping the computational $|g\\rangle$–$|e\\rangle$ transition long-lived. The same filter makes the $|g\\rangle$–$|h\\rangle$ transition a fast drain, so two microwave tones pump any excitation into $|h\\rangle$ and let it decay to $|g\\rangle$, giving unconditional reset. If right, this offers a path to superconducting quantum computing whose readout and reset do not rely on dispersive qubit–resonator interaction and its failure modes.","feed_headline":"One filter gives resonator-free fluxonium readout and 99.5% reset","feed_subtitle":"An on-chip filter tailors decay rates so the qubit's own levels do readout and reset, skipping dispersive cavities.","key_machinery":"The load-bearing object is a single-stage coplanar-waveguide filter, a flat on-chip microwave transmission-line filter, whose power transmittance $T(\\omega)$ rescales every external decay rate as $\\Gamma_{\\rm ext}^{ij} \\propto |T(\\omega_{ij})\\omega_{ij}|\\,|\\langle i|\\hat{n}|j\\rangle|^2\\,[\\coth(\\hbar\\omega_{ij}/2k_BT)+1]$. The filter is placed between the qubit and the readout transmission line so $T$ is near unity at the 5.369 GHz $|e\\rangle$–$|f\\rangle$ transition and at the 7.814 GHz $|g\\rangle$–$|h\\rangle$ transition, while giving more than 30 dB attenuation below 1 GHz, where the 255 MHz $|g\\rangle$–$|e\\rangle$ transition lives. This contrast in decay rates is what lets the $|e\\rangle$–$|f\\rangle$ transition serve as a fast, bright readout channel without shortening $|e\\rangle$'s lifetime, and what makes $|h\\rangle$–$|g\\rangle$ a fast drain for reset. The closed readout cycle relies on the $|g\\rangle$–$|f\\rangle$ transition being dipole-forbidden at the sweet spot, so fluorescence from the readout transition does not leak population back into the computational basis.","core_discovery":"The paper's central claim is that a fluxonium qubit's internal level structure plus a planar filter can replace the resonator for both readout and reset. With the filter passing 5.369 GHz and 7.814 GHz while strongly attenuating the 255 MHz computational transition, the external decay rate of the readout transition is $\\Gamma_r/2\\pi = 5.4$ MHz while the qubit's energy-relaxation time is $T_1 = 51$ µs. Because the $|g\\rangle$–$|f\\rangle$ transition is dipole-forbidden at the sweet spot, driving $|e\\rangle$–$|f\\rangle$ forms a closed fluorescence cycle, giving a QND figure $N_{\\rm QND} = \\Gamma_r T_1^{\\rm meas} = 1.6\\times 10^3$. Two-tone driving of $|e\\rangle$–$|f\\rangle$ and $|f\\rangle$–$|h\\rangle$ then resets any population in the excited manifold to $|g\\rangle$, with corrected residual excitation below 1% after 200 ns and below 0.5% after 250 ns.","pith_inferences":["- A direct time-resolved measurement of the $|h\\rangle$–$|g\\rangle$ lifetime would test the paper's estimate of $\\Gamma_{hg}/2\\pi = 6.5$ MHz; the paper infers this rate from Eq. (2) and the measured $\\Gamma_r$ rather than observing it directly.","- The same dissipation-engineering idea could serve as a built-in leakage drain in error-corrected circuits: if the filter passband is placed on a high-lying transition, leakage population would decay away automatically without dedicated reset pulses, something the paper does not explore.","- The paper's own mention of lumped-element and spiral filters suggests the same physics could be packaged much more compactly, which would matter for multiplexed chips, but that integration step is not demonstrated here."],"forward_implications":["- Because $\\Gamma_r$ scales with the square of the qubit–waveguide coupling capacitance, increasing that capacitance tenfold would make fluorescence readout about 100 times faster, potentially reaching dispersive-readout speeds.","- The reset is all-microwave, uses no fast flux-bias pulse, and involves no second-order process, so it requires more than two orders of magnitude less microwave power than earlier resonator-based Raman reset schemes.","- If the $|f\\rangle$–$|g\\rangle$ quasiparticle decay is made smaller than the $|e\\rangle$–$|g\\rangle$ decay, the readout's QNDness can exceed the ordinary $T_1$ limit, so suppressing quasiparticles directly improves the non-demolition character.","- Removing the resonator eliminates a whole class of readout modeling difficulties and the ionization/measurement-induced state-transition failures that plague dispersive readout, simplifying both analysis and architecture design.","- One of the two reset drives is at the same frequency as the readout drive, so the reset protocol adds minimal microwave hardware over what readout already requires."],"supporting_citations":[{"why":"Earlier resonator-free fluorescence readout of a fluxonium; supplies the baseline QNDness of 1.1×10^2 and the closed-cycle problem this paper solves.","marker":"[33]"},{"why":"Design method for band-pass filters with arbitrary center frequency, bandwidth, and stage count using transmission lines and short-to-ground stubs; supplies the filter used here.","marker":"[47]"},{"why":"Formula for energy relaxation by quasiparticle tunneling in a single Josephson junction; used to attribute the residual non-QNDness and estimate x_qp = 4×10^-7.","marker":"[72]"},{"why":"Fluxonium behavior at the sweet spot and quasiparticle-density observations; supports the level-structure assumptions and the correction method for state transitions during readout.","marker":"[13]"},{"why":"Standard dispersive readout framework that the paper's resonator-free scheme is designed to replace; supplies the comparison for modeling complexity.","marker":"[20]"},{"why":"Purcell filter used with dispersive readout to protect the qubit; provides the architectural contrast showing previous filters accompany resonators.","marker":"[21]"},{"why":"Earlier all-microwave resonator-based reset schemes; the paper compares its first-order, lower-power reset against these baselines.","marker":"[39–41]"}],"fun_headline_variants":["One filter replaces cavity for fluxonium readout and reset","Resonator-free fluxonium: filter enables QND readout and fast reset","Dissipation-engineered fluxonium drops resonator for readout and reset","Planar filter gives fluxonium non-demolition readout and 99.5% reset"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the filter acts purely as a frequency-dependent energy-loss rate for the qubit, holding no energy itself and introducing no memory or new states; if it instead behaves like a small resonator, the extracted decay rates and the closed reset picture collapse.","fun_headline_variants_meta":{"raw":{"variants":["One filter replaces cavity for fluxonium readout and reset","Resonator-free fluxonium: filter enables QND readout and fast reset","Dissipation-engineered fluxonium drops resonator for readout and reset","Planar filter gives fluxonium non-demolition readout and 99.5% reset"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001025,"raw_usage":{"total_tokens":4328,"prompt_tokens":959,"completion_tokens":3369,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":575,"completion_tokens_details":{"reasoning_tokens":3287}},"tokens_in":575,"tokens_out":3369,"duration_ms":21831,"temperature":1.0,"reasoning_tokens":3287,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:16:46.639401+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the $|h\\rangle$–$|g\\rangle$ decay rate directly by preparing $|h\\rangle$ and watching its fluorescence: the reset model predicts a lifetime of about $1/(2\\pi\\times6.5\\,\\mathrm{MHz}) \\approx 24$ ns, set by the filter passband; a large deviation, or a reset fidelity that does not track the filter transmission at 7.814 GHz, would break the filter-as-environment assumption.","supporting_citations":[{"cited_title":"Ficheux, L","cited_arxiv_id":null,"evidence_quote":"Earlier resonator-free fluorescence readout of a fluxonium; supplies the baseline QNDness of 1.1×10^2 and the closed-cycle problem this paper solves."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Design method for band-pass filters with arbitrary center frequency, bandwidth, and stage count using transmission lines and short-to-ground stubs; supplies the filter used here."},{"cited_title":"Sunada, K","cited_arxiv_id":null,"evidence_quote":"Formula for energy relaxation by quasiparticle tunneling in a single Josephson junction; used to attribute the residual non-QNDness and estimate x_qp = 4×10^-7."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Fluxonium behavior at the sweet spot and quasiparticle-density observations; supports the level-structure assumptions and the correction method for state transitions during readout."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Standard dispersive readout framework that the paper's resonator-free scheme is designed to replace; supplies the comparison for modeling complexity."}],"review_version":1}