{"id":"78940a05-c8c0-42eb-a180-2f910632f805","arxiv_id":"2412.14853","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A 3D re-entrant cavity filter enables frequency-multiplexed readout of four superconducting qubits with 98.6% average fidelity in 1 microsecond and low crosstalk, with simulated intrinsic Purcell filtering.","lead":"Researchers built a small 3D metal cavity that sits above a superconducting qubit chip and filters the microwave signals used to read out the qubits. It lets four qubits be measured at once through one cable with 98.6% accuracy in a microsecond, without a parametric amplifier.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Purcell-filtering claim rests solely on simulated admittance; measured T1 values are far below the predicted unfiltered Purcell limit, so they cannot validate the suppression.","rationale":"The paper reports a credible four-qubit multiplexed-readout demonstration with 98.6% average assignment fidelity and low measured crosstalk. The central claim, however, is not just readout fidelity but the intrinsic Purcell filtering provided by the 3D re-entrant cavity and the additional interferometric suppression from multipath capacitance. Those claims are supported only by Fig. 2 simulations. The measured T1 values do not constrain the Purcell contribution because they are far shorter than the expected unfiltered Purcell limit, meaning the device would show the same T1 whether or not the filter worked. This is precisely the gap identified by the reader's weakest-assumption analysis: the load-bearing part of the architecture is simulation-only. The appropriate response is to keep the CONDITIONAL verdict and require either direct experimental evidence of the Purcell suppression (e.g., T1 versus qubit frequency, or a with/without-filter comparison) or a clear separation of simulated claims from measured claims. No additional concern rises to the level of changing the verdict.","tokens_in":10112,"tokens_out":4796,"duration_ms":45764,"concrete_test":"Take one flux-tunable qubit on the demonstrated device, sweep its frequency through the designed notch region (about 5.0–8.0 GHz) by flux bias, and measure T1 at each frequency. Compute the predicted Purcell contribution Gamma_P(omega_q) = Re[Y(omega_q)]/C_Sigma from the simulated admittance in Fig. 2(b) and compare with the measured T1(omega_q), accounting for flux-dependent two-level-system loss. A clear local maximum in T1 near the predicted 5.5 GHz and 7.7 GHz notch frequencies would validate the interferometric Purcell claim; absence of any such peak would show that the simulation-based mechanism is not experimentally supported. A control measurement on an identical chip with the cavity filter removed or detuned would provide the missing with/without-filter baseline.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest version of the paper's claim is that the 3D re-entrant cavity provides intrinsic Purcell protection. The only support is the finite-element/lumped-element admittance simulation in Fig. 2(b), including the assumed parasitic cross-capacitances Cx1 and Cxq responsible for the 5.5 GHz and 7.7 GHz notch frequencies. No experimental measurement of the qubit environment admittance, no with/without-filter Purcell baseline, and no T1-versus-frequency data are presented. This matters because the reported T1 values (47–52 us, Table I) cannot confirm the mechanism. Using the tabulated parameters (qubit frequencies 5.7–6.0 GHz, readout frequencies about 10 GHz, chi/2pi about -1.5 MHz, kappa/2pi about 1 MHz) gives an estimated unfiltered Purcell T1 of order Delta^2/(g^2 kappa) of a few milliseconds, much longer than the measured 50 us. The measured coherence is therefore dominated by other, faster loss channels, and it would look essentially identical whether the cavity filter suppressed Purcell decay or not. The multiplexed readout demonstration itself is credible, but the central architectural advantage claimed for the cavity filter is currently unverified by experiment.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a 3D re-entrant cavity multiplexer that capacitively couples out-of-plane to an array of on-chip readout resonators, with the goal of providing broadband bandpass filtering, intrinsic Purcell protection, and frequency-multiplexed readout without additional on-chip filter components. The authors demonstrate the concept on a four-qubit coaxial transmon device, reporting an average single-shot readout fidelity of 98.6% with a 1 µs integration time and no parametric amplifier, off-diagonal measurement-induced dephasing rates below 0.15 kHz, and a 16-state assignment matrix with average fidelity 94.2%. The Purcell-filtering and interferometric-suppression claims are supported by finite-element and lumped-element simulations, not by direct experimental isolation of the filter's effect.","tokens_in":10380,"tokens_out":7768,"duration_ms":50360,"significance":"If the Purcell-filtering and interferometric-suppression claims were experimentally validated, the out-of-plane 3D re-entrant cavity would be a useful hardware-efficient alternative to on-chip Purcell filters, particularly for scaling because it removes filter footprint from the qubit chip. The experimental readout results are credible and useful: the single-shot histograms, the 16-state assignment matrix, and the Hahn-echo dephasing measurements directly support the headline readout fidelity and off-diagonal crosstalk numbers. The paper also demonstrates simultaneous multiplexed readout with low crosstalk without a parametric amplifier, which is a practical contribution. The principal weakness is that the central Purcell-protection mechanism is asserted on the basis of simulation alone, and the measured coherence times cannot distinguish a filtered from an unfiltered device.","major_comments":[{"comment":"The claim that the re-entrant cavity provides 'intrinsic Purcell filtering' rests solely on the FEM/lumped-element admittance simulation in Fig. 2(b). There is no experimental measurement of the qubit-environment admittance, no with/without-filter comparison of T1, and no T1-versus-frequency data. This is not a minor omission: using the Table I parameters (ωq/2π ≈ 5.7–6.0 GHz, ωR/2π ≈ 10 GHz, χ/2π ≈ 1.5 MHz, κ/2π ≈ 1 MHz) gives an unfiltered Purcell T1 of roughly 0.4 ms, which is about eight times longer than the measured T1 ≈ 50 µs. The measured T1 is therefore dominated by other loss channels, and it would be essentially unchanged whether the filter provided the simulated suppression or not. The manuscript should either add a direct test of the Purcell suppression (for example, admittance measurement or T1 as a function of qubit frequency across a filter edge) or explicitly qualify the Purcell claim as simulation-based and not experimentally isolated in this work.","section":"Section II, Fig. 2(b); Section V"},{"comment":"The text states that the interferometric Purcell filter creates bandstop notches designed at 5.5 GHz and 7.7 GHz, but the qubit frequencies in Table I are 5.658–6.034 GHz. The 5.5 GHz notch is not at any qubit frequency, and 7.7 GHz is far from all qubits. As written, the demonstrated device does not implement interferometric suppression at the operating qubit frequencies, and the claim that the fabricated device includes such protection is unsupported. If the notches belong to a different or future geometry, the text and Fig. 2(b) caption need to say so; if the device was intended to have notches at the qubit frequencies, the quoted notch frequencies and the simulation need to be reconciled.","section":"Section II, Fig. 1(c) and Fig. 2(b)"}],"minor_comments":[{"comment":"The abstract states 'measurement-induced dephasing rates below 0.15 kHz' without the qualifier that this refers only to off-diagonal crosstalk between different qubit–resonator pairs; the diagonal self-dephasing rates in Fig. 6(b) are orders of magnitude larger (tens of MHz). Please qualify the claim in the abstract to avoid overstatement.","section":"Abstract and Section IV"},{"comment":"The two-port simulation used to extract the filter bandwidth replaces the single physical port with two ports displaced along y and mirrored; the caption should explain how this two-port representation corresponds to the actual single-port device and whether it affects the extracted external quality factor.","section":"Section II, Fig. 2(a)"},{"comment":"The closing sentence contains a grammatical error: 'the design could capacitively couples to larger arrays' should read 'the design could capacitively couple to larger arrays.'","section":"Section V"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the scope of a quantum-hardware journal. The readout and crosstalk measurements are solid and useful, and the 3D out-of-plane filter concept is attractive. The main issue is epistemic: the abstract and conclusions claim intrinsic Purcell filtering and interferometric suppression as demonstrated properties, but the only evidence is simulation, and the measured T1 values cannot validate the mechanism. I would not require a full Purcell-baseline experiment for acceptance, but the claims should be explicitly conditioned on simulation, and the 5.5/7.7 GHz notch frequencies versus the 5.66–6.03 GHz qubit frequencies must be resolved. The other technical content is sound and the writing is mostly clear."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nWorth a look if you work on multiplexed readout or 3D-integrated superconducting circuits. The new thing here is the 3D re-entrant cavity used as a bandpass Purcell filter, placed out of the chip plane and capacitively coupled to an array of readout resonators. That geometry is not in the cited prior art. The authors show a four-qubit device with average single-shot readout fidelity of 98.6% in a 1 µs integration time, no parametric amplifier, and off-diagonal measurement-induced dephasing rates under 0.15 kHz. Those numbers look solid: the histograms are bimodal, the assignment matrix is consistent with uncorrelated errors, and the crosstalk echo data are clean.\n\nThe multiplexing demonstration itself is credible and the characterization is honest about error sources (T1 decay during readout, thermal population, post-heralding). The paper also does a nice job extracting resonator linewidths from group delay.\n\nThe soft spot is the Purcell-filtering claim. Figure 2(b) is FEM simulation of the qubit-environment admittance, with and without the filter, plus a simulated interferometric notch. There is no direct experimental validation: no measurement of the environment admittance, no with/without-filter baseline, no T1-vs-frequency sweep. The stress-test note is right that the measured T1 of about 50 µs cannot validate the suppression. For these parameters, the unfiltered Purcell limit is on the order of milliseconds, so the measured T1 is dominated by other loss channels and would look the same with or without the filter. That means the central architectural advantage—intrinsic Purcell protection—is currently unverified by experiment.\n\nNone of this makes the paper a bad paper. The device is real, the multiplexed readout works, and the simulation is a reasonable design calculation. But the marketing should be separated from the measurement. If this goes to review, the authors should be asked to either provide direct evidence of Purcell suppression (e.g., T1 as a function of filter coupling, or a direct admittance probe) or explicitly label the filtering benefit as a design prediction. Also, no data or code is released, which would help with reproduction.\n\nI'd send it to peer review. The hardware demonstration deserves referee time even if the filtering advantage remains provisional. I might cite it for the 3D re-entrant cavity multiplexer concept, but not for the Purcell filtering evidence.","headline":"A credible four-qubit multiplexed readout demo with a novel 3D cavity filter, but the Purcell-filtering advantage rests entirely on simulation and is not confirmed by the measured T1.","tokens_in":10968,"tokens_out":3371,"would_cite":true,"duration_ms":28010,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["85.25.Cp","03.67.Lx"],"model":"deepseek-v4-flash","headline":"A single 3D re-entrant cavity, coupled out-of-plane to an array of on-chip readout resonators, acts as a broadband bandpass filter with intrinsic Purcell protection, achieving 98.6% average multiplexed readout fidelity in one microsecond…","keywords":["multiplexed readout","superconducting qubits","re-entrant cavity filter","Purcell filter","dispersive readout","3D integration","readout crosstalk","measurement-induced dephasing"],"falsifier":"Measure qubit T1 as the qubit frequency is swept across the predicted 5.5 GHz interference notch: if the interferometric model is correct, the Purcell-limited decay rate should drop sharply at the notch. Alternatively, measure the admittance seen from a qubit port with a network analyser and compare it to the finite-element prediction; a substantial mismatch would invalidate the claimed Purcell suppression.","tokens_in":9923,"feed_emoji":"📡","tokens_out":6128,"duration_ms":37102,"temperature":0.7,"pith_summary":"This paper argues that a machined three-dimensional cavity can replace the usual on-chip network of readout resonators and Purcell filters. The cavity, capacitively coupled out-of-plane to an array of on-chip readout resonators, acts as a wideband bandpass filter that also suppresses the qubit's spontaneous-emission channel (Purcell decay). The authors demonstrate the idea on a four-qubit device, reporting an average single-shot readout fidelity of 98.6% within a 1 microsecond integration time, measurement-induced dephasing rates below 0.15 kHz for untargeted qubits, and no parametric amplifier. If the approach scales, it would simplify the chip footprint of multiplexed readout in superconducting quantum processors.","feed_headline":"3D cavity filter reads four qubits at 98.6% fidelity","feed_subtitle":"One out-of-plane re-entrant cavity multiplexes readout and suppresses Purcell decay without a parametric amplifier.","key_machinery":"The central object is the 3D re-entrant cavity multiplexer: a rectangular cavity with a re-entrant section and four extruded pins that capacitively couple to readout resonators on the opposite side of the substrate. The cavity is a quarter-wave resonator, miniaturised by a shunt-to-ground capacitor Cshunt, and is read out through a single coaxial port; its low external quality factor (about 9) makes it a wideband bandpass filter with a 3 dB bandwidth of 1.6 GHz. The Purcell protection is engineered through the coupling network: parasitic and intentional cross-couplings (Cx1 and Cxq) create destructive-interference notches at the qubit frequencies, and the environment admittance seen by a qubit junction is shown by finite-element simulation to drop by an order of magnitude relative to the unfiltered case. This machinery carries the argument because it provides both the multiplexing bandwidth and the qubit protection in one passive, off-chip component.","core_discovery":"The central claim is that a re-entrant cavity filter, integrated in the 3D package rather than on the chip, can perform frequency-multiplexed dispersive readout while intrinsically suppressing Purcell decay. The cavity is a capacitively loaded quarter-wave resonator whose open end faces an array of readout resonators through extruded pins; the coupling is purely capacitive and out-of-plane, so no galvanic connections or on-chip filter elements are needed. With a 1.6 GHz passband centered at 9.8 GHz and an external quality factor of about 9, the cavity provides a wideband readout channel, while its coupling network — including cross-coupling capacitors Cx1 and Cxq — forms interferometric notches at 5.5 GHz and 7.7 GHz that eliminate transmission paths responsible for Purcell decay. On a four-qubit device the authors achieve an average assignment fidelity of 98.6% in one microsecond, a 16-state assignment fidelity of 94.2% close to the product of the individual fidelities, and off-diagonal measurement-induced dephasing rates below 0.15 kHz.","pith_inferences":["If the interferometric notch model holds, the same capacitive-cross-coupling trick could place suppression notches at arbitrary frequencies — for example, to filter control-line crosstalk or to isolate qubits from other noise sources, not just readout decay.","A direct with-and-without-filter comparison of qubit T1 on the same device would test the core Purcell claim; the present paper offers only simulation support for that part.","The 98.6% fidelity achieved without a parametric amplifier suggests the cavity's wide bandwidth already provides much of the signal-to-noise benefit; adding a quantum-limited amplifier would likely trade some of that simplicity for higher fidelity or shorter measurement time."],"forward_implications":["Adding more extruded pins and multi-pole filtering sections to the same cavity could extend multiplexed readout to tens or hundreds of qubits without enlarging the chip footprint.","Because the filter lives off-chip, the readout linewidth can be tuned by adjusting the pin-to-resonator gap rather than by redesigning on-chip components.","Measured off-diagonal dephasing below 0.15 kHz means simultaneous readout of all qubits introduces negligible crosstalk, supporting the use of multiplexed readout for error-correction feedback.","The 16-state assignment fidelity of 94.2%, nearly equal to the product of individual fidelities, indicates that readout errors are largely uncorrelated across qubits."],"supporting_citations":[{"why":"Establishes dispersive readout of superconducting qubits, the measurement method being multiplexed here.","marker":"[9]"},{"why":"Demonstrates fast readout and reset with an intrinsic Purcell filter that this design extends to a 3D cavity.","marker":"[14]"},{"why":"Identifies the Purcell decay channel that the filter must suppress.","marker":"[15]"},{"why":"Introduces Purcell filters for suppressing spontaneous emission of a superconducting qubit.","marker":"[16]"},{"why":"Shows fast, accurate state measurement with multiplexed readout and a Purcell filter.","marker":"[17]"},{"why":"Exemplifies rapid multiplexed readout with individual narrowband filters and supplies the measurement-induced dephasing model used here.","marker":"[20]"},{"why":"Recent CPW implementation of multiplexed readout with intrinsic Purcell and notch filtering, the direct alternative this 3D design competes with.","marker":"[22]"},{"why":"Establishes the double-sided coaxial circuit QED architecture used for the device.","marker":"[23]"},{"why":"Shows high coherence and low cross-talk in the tileable 3D integrated platform on which the device is built.","marker":"[24]"},{"why":"Gives the theory of measurement-induced dephasing used to quantify readout crosstalk.","marker":"[26]"}],"fun_headline_variants":["3D cavity filters readout and Purcell decay in one","Out-of-plane cavity multiplexes qubit readout without amplifiers","One cavity, four qubits: readout and Purcell filtering combined","Re-entrant cavity enables scalable multiplexed qubit readout","4-qubit readout at 98.6% with Purcell-suppressing cavity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The Purcell protection and the extra notch suppression are validated only by finite-element simulations of the fabricated geometry, including assumed parasitic capacitances; the paper does not directly measure the qubit's electromagnetic environment or compare T1 with and without the filter.","fun_headline_variants_meta":{"raw":{"variants":["3D cavity filters readout and Purcell decay in one","Out-of-plane cavity multiplexes qubit readout without amplifiers","One cavity, four qubits: readout and Purcell filtering combined","Re-entrant cavity enables scalable multiplexed qubit readout","4-qubit readout at 98.6% with Purcell-suppressing cavity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000325,"raw_usage":{"total_tokens":1797,"prompt_tokens":899,"completion_tokens":898,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":515,"completion_tokens_details":{"reasoning_tokens":805}},"tokens_in":515,"tokens_out":898,"duration_ms":7504,"temperature":1.0,"reasoning_tokens":805,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:51:14.602852+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure qubit T1 as the qubit frequency is swept across the predicted 5.5 GHz interference notch: if the interferometric model is correct, the Purcell-limited decay rate should drop sharply at the notch. Alternatively, measure the admittance seen from a qubit port with a network analyser and compare it to the finite-element prediction; a substantial mismatch would invalidate the claimed Purcell suppression.","supporting_citations":[{"cited_title":"Wallraff, D","cited_arxiv_id":null,"evidence_quote":"Establishes dispersive readout of superconducting qubits, the measurement method being multiplexed here."},{"cited_title":"Sunada, S","cited_arxiv_id":null,"evidence_quote":"Demonstrates fast readout and reset with an intrinsic Purcell filter that this design extends to a 3D cavity."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Identifies the Purcell decay channel that the filter must suppress."},{"cited_title":"Jeffrey, D","cited_arxiv_id":null,"evidence_quote":"Shows fast, accurate state measurement with multiplexed readout and a Purcell filter."},{"cited_title":"Heinsoo, C","cited_arxiv_id":null,"evidence_quote":"Exemplifies rapid multiplexed readout with individual narrowband filters and supplies the measurement-induced dephasing model used here."},{"cited_title":"Rahamim, T","cited_arxiv_id":null,"evidence_quote":"Establishes the double-sided coaxial circuit QED architecture used for the device."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows high coherence and low cross-talk in the tileable 3D integrated platform on which the device is built."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the theory of measurement-induced dephasing used to quantify readout crosstalk."}],"review_version":1}