{"id":"911b6c2c-14bd-4912-8306-0d88512bfa84","arxiv_id":"2508.03973","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Charge-parity flips from quasiparticle tunneling measurably limit transmon dephasing at EJ/EC~50, and post-selecting steady-parity Ramsey shots recovers longer T2*.","lead":"This experiment shows that transmon qubits, long thought to be noise-immune to stray charges, can still lose quantum coherence when quasiparticles flip the device's charge parity. A new interleaved measurement protocol detects those flips shot-by-shot and recovers a longer dephasing time when only steady-parity data are kept.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Single-flip-per-shot assumption and uncalibrated parity readout could bias the post-selected T2*; need conditional acceptance","rationale":"The paper is an experimental study with a clear protocol. The central claim, that charge-parity noise is a dominant source of dephasing in this transmon, is supported by the internal consistency of the post-selection result and the simulation. However, the claim's strength depends on the correctness of the parity classification. The authors themselves note that parity flip rates of ~1 kHz are 'unlikely' to cause more than one flip, but they do not quantify the probability of misclassification or the error in the single-shot parity readout. This is a load-bearing assumption because a small fraction of misclassified parity flips could account for a significant portion of the 19.6 us T2* difference. The reader's weakest_assumption exactly matches this, so I agree. The requested concrete test—characterizing the parity readout confusion matrix and simulating the post-selection with realistic flips—would settle whether the difference is real or an artifact. Given that these are addressable gaps and there is honest acknowledgement of the 5 us offset, conditional acceptance is appropriate, but the current version does not need to be rejected out of hand.","tokens_in":9508,"tokens_out":1559,"duration_ms":16787,"concrete_test":"Quantify readout fidelity and misclassification by interleaving two parity-detection pulses separated by a short time (shorter than expected flip times) and computing the confusion matrix of the parity-readout. Then, simulate the Ramsey-post-selection with realistic single-shot error rates and include the possibility of 0, 1, and 2 flips per shot (e.g., using a Poisson process with p_flip = 1 kHz * shot_duration). Recompute the T2* difference (43 vs 23.4 us) under this model. If the inferred intrinsic T2* of the parity-stable subset drops by more than ~3 us once misclassification is accounted for, the central claim weakens and needs to be qualified.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim is that parity switching is the dominant mechanism reducing Ramsey T2* from 43 us (parity-stable) to 23.4 us (ungated). This rests on the post-selection being correct. Two unquantified error sources could corrupt it. First, the parity detection is a single-shot readout of the 1-2 transition with no quoted assignment fidelity or readout discrimination error. If a misclassification occurs even at a few percent, the 'unflipped' category includes shots that actually flipped, which would lower the extracted T2* of the 'parity-stable' subset; conversely, 'flipped' shots that did not flip would dilute the flipped category. Second, the protocol assumes at most one parity flip per Ramsey shot since only pi and pf are compared. The text argues 1 kHz flip rate makes multi-flip unlikely over ~100 us, but no quantitative bound is given and a 1 kHz rate implies ~0.1 flips per shot. More importantly, a flip during the initial parity detection (pi) or during the Ramsey delay followed by a flip before pf detection would be misclassified as 'no flip'. Given the average of ~0.1 flips per shot, such events are not negligible, and they would preferentially contaminate the 'unflipped' category, inflating the apparent improvement. The paper does not provide the conditional probability of misclassification as a function of delay time, nor does it correct for the finite parity-detection duration (~10 us each).","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents an experimental study of charge-parity (quasiparticle-poisoning) noise in a tantalum transmon with E_J/E_C ≈ 50. The authors resolve the two parity states through the 1-2 transition, interleave single-shot parity detection before and after each Ramsey interrogation, and post-select shots that did not undergo a parity flip. They report T2* = 23.4 ± 0.6 µs for unsorted Ramsey data versus T2* = 43.0 ± 1.0 µs for parity-stable shots, and show that Ramsey T2* decreases with increasing charge dispersion while spin-echo T2* remains flat. A Lindblad master-equation simulation with a random parity-flip time is used to compare with the measured dispersion dependence. The conclusion is that charge-parity noise, not the usual assumption of charge insensitivity, is a dominant source of T2* fluctuations in this transmon regime.","tokens_in":9879,"tokens_out":7091,"duration_ms":84917,"significance":"If the post-selection and simulation withstand scrutiny, the result is an important experimental demonstration that transmons at the standard E_J/E_C ≈ 50 operating point can still be limited by charge-parity noise. The protocol—embedding single-shot parity detection within Ramsey and spin-echo measurements—is direct and transferable, and the reported contrast between unsorted (23.4 ± 0.6 µs) and parity-stable (43.0 ± 1.0 µs) T2* is a striking falsifiable observation. The paper also states its own limitation: the effect is only visible when T2* is sufficiently long (Appendix A.2), and longer-term drifts remain unexplained. These admissions strengthen the credibility of the central claim rather than weakening it.","major_comments":[{"comment":"The post-selection step is load-bearing: the 43.0 vs 23.4 µs difference is what establishes parity flips as the dominant dephasing mechanism. The manuscript does not report the assignment fidelity of the single-shot parity readout, and it treats “one flip per shot” as the only classification error. At the quoted ~1 kHz flip rate, the total per-shot exposure includes two ~10 µs parity-detection/readout windows plus the Ramsey delay (~100 µs), so a non-negligible fraction of shots will have a flip inside a parity-detection window; such events are exactly those misclassified by the (pi, pf) comparison. Please quantify the parity-readout contrast/fidelity and the conditional misclassification probability as a function of delay, and show that the extracted T2* improvement survives a correction for these errors.","section":"Section 2.2, Fig. 3"},{"comment":"The Lindblad simulation is offered as confirmation, but the input values of T1 and Tphi are not stated. If Tphi (or the no-flip T2*) was taken from the same Ramsey data, the comparison is partly circular. Report the exact T1 and Tphi values, state whether they are independently measured (e.g., T1 from inversion recovery and Tphi from Hahn echo), and quantify the sensitivity of the simulated T2*(Δ01) curve to these inputs. In addition, the simulation averages over exactly one parity flip; shots with zero and two flips should be included with their Poisson weights to match the experiment at the quoted flip rates.","section":"Appendix A.1, Fig. 5(b)"}],"minor_comments":[{"comment":"The improved T2* for parity-stable shots is shown only for the f+12 subgroup. To rule out an asymmetry in the parity-detection protocol, show the f−12 subgroup or a combined analysis.","section":"Section 3.1, Fig. 4(b)"},{"comment":"The statement that 1 kHz flip rates make multi-flip events unlikely is heuristic. Replace it with a quantitative Poisson estimate that includes the finite duration of the parity-detection and reset windows.","section":"Section 2.2"},{"comment":"The axes in Fig. 5 are not fully defined: the bottom and top x-axes presumably correspond to offset charge and Δ01, respectively. Specify the conversion, the number of repetitions per point, and how error bars are obtained.","section":"Section 3.2, Fig. 5"},{"comment":"Define δω and the sign convention for α in the Hamiltonian, and specify whether α is the measured anharmonicity or the full 1-2 detuning.","section":"Appendix A.1, Eq. (3)"},{"comment":"Several device parameters are attributed to Ref. [8]. Clarify which values (e.g., charge dispersion of the 1-2 transition, T1) are measured in the present work and which are taken from prior LLNL publications.","section":"Section 2.1"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper makes a solid case that quasiparticle parity switches measurably dephase a transmon at EJ/EC~50, which cuts against the usual assumption that this regime is charge-insensitive. The embedded parity detection is a smart protocol, and the post-selected T2* = 43 us vs 23.4 us is a strong signal. The missing calibration of parity-readout fidelity is a real gap, but it's fixable.\n\nWhat's new: interleaving single-shot parity detection directly into Ramsey, then sorting shots. That's a direct measurement of the noise process, not an inference. The monotonic drop in T2* with charge dispersion, while spin-echo stays flat, is a clean demonstration that the dephasing comes from slow parity telegraph noise, not faster processes. The simulation is a forward model using measured inputs rather than a fit of the central curve, which is honest. The explicit note about a 5 us long-term drift that the mechanism cannot account for is a good sign.\n\nSoft spots: the parity readout fidelity is never reported. The paper argues that a 1 kHz flip rate makes multi-flip shots unlikely, but doesn't give a bound. However, the direction of any contamination is actually conservative: misclassified or double-flip shots end up in the 'no-flip' bin, which would lower the post-selected T2*, not raise it. So the 43 vs 23 us gap is not an artifact of these effects; if anything, it's an underestimate. The bigger question is how often the pi readout misassigns parity during the ~10 us detection itself; a few percent error can't explain a factor of two, but the protocol is proposed as a characterization metric, so the fidelity needs to be on the table. The simulation's T1 and Tphi values are not fully justified, but the qualitative trend is robust to those.\n\nWho it's for: people working on transmon coherence, quasiparticle poisoning, and device characterization. It deserves a serious referee. I'd accept it with a request for the readout calibration and a contamination calculation. The result is worth sharing even before those numbers exist, but they matter for the 'metric' claim.","headline":"A useful experimental result with a genuine gap: charge-parity noise measurably dephases a transmon at EJ/EC~50, and the protocol is clever, but the missing parity-readout fidelity and multi-flip bound should be requested before publication.","tokens_in":10356,"tokens_out":5464,"would_cite":true,"duration_ms":68084,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Lx","85.25.Cp"],"model":"deepseek-v4-flash","headline":"Ramsey measurements with embedded parity detection show quasiparticle-induced parity flips, not charge offset, set the dephasing limit of a tantalum transmon at EJ/EC≈50.","keywords":["transmon qubit","charge-parity noise","quasiparticle poisoning","dephasing","Ramsey interferometry","spin echo","charge dispersion","superconducting qubits"],"falsifier":"Run the same Ramsey sequence with a parity detector that can catch two or more flips per shot, e.g., repeated or continuous parity readout during the delay. If the apparent T2* difference between unsorted and single-parity subsets vanishes, or if single-parity shots are found to contain more than one flip at the claimed ~1 kHz rate, the central mechanism would be contradicted.","tokens_in":9454,"feed_emoji":"⚛️","tokens_out":6162,"duration_ms":72890,"temperature":0.7,"pith_summary":"The paper tries to show that transmons, even at the canonical EJ/EC≈50 operating point, are not as charge-insensitive as assumed: residual charge dispersion (~6 kHz on the 0-1 transition) plus quasiparticle-induced parity switches produce measurable T2* fluctuations. The authors embed single-shot parity detection into every Ramsey shot, categorizing shots as parity-flipped or not. Unflipped shots yield T2*=43.0±1.0 μs versus 23.4±0.6 μs for the unsorted set, and T2* decreases monotonically with charge dispersion while spin-echo times stay flat. This establishes charge-parity noise as a dominant decoherence channel in this regime and motivates parity flip rate as a characterization metric.","feed_headline":"Parity flips halve transmon dephasing time","feed_subtitle":"Embedded single-shot parity checks in Ramsey experiments reveal that a hidden noise source, not charge offset, dominates T2*.","key_machinery":"Embedded parity detection: a single-shot measurement of the 1-2 transition, using a frequency-selective long π12 pulse tuned to one parity band (f+12 or f−12), performed before and after each Ramsey shot, with the two outcomes (pi, pf) classifying each shot as flipped or unflipped. The supporting machinery is a Lindblad master-equation simulation in which the Hamiltonian switches between parity-dependent detunings ±2πΔ01 at a random time tf, averaged over realizations.","core_discovery":"The central discovery is that the dephasing of a transmon in the charge-insensitive regime is limited by fast parity switches (quasiparticle tunneling events) rather than by the slow charge-offset drift itself. By resolving the two parity bands on the 1-2 transition (≈150 kHz charge dispersion) and using long selective π12 pulses to read out parity at the start and end of each Ramsey shot, the authors directly observe individual parity flips. Post-selecting shots that stayed in one parity recovers a dephasing time T2*=43.0±1.0 μs, nearly double the 23.4±0.6 μs obtained when averaging over both parities. The monotonic decrease of T2* with charge offset, the absence of this trend in spin-echo,","pith_inferences":["The paper's protocol assumes one flip per shot; a natural extension would be repeated or continuous parity readout during the delay to measure flip statistics within a shot, which would validate or bound the correction.","If the readout fidelity of the parity detector were quantified and accounted for, the inferred 'ideal' T2* would likely shift, since misclassification would mix flipped and unflipped populations; this correction is absent from the reported values.","The same embedded-detection logic could be applied to other telegraphic noise sources, such as two-level-system defects, by choosing a probe transition whose frequency distinguishes the noise state, effectively turning Ramsey decay into a noise-state-resolved measurement.","Because the 1-2 parity readout itself is a measurement on the qubit's higher levels, one could test for backaction of the parity detection on 0-1 coherence by varying the detection parameters; the paper does not report such a control."],"forward_implications":["Apparent T2* of a transmon can understate the intrinsic coherence by up to a factor of two when parity flips are averaged in; parity-resolved post-selection recovers the single-parity value.","Parity flip rate (≈1 kHz here) should be reported as a device characterization metric alongside T1 and T2*, since it sets a dephasing floor even at EJ/EC≈50.","Spin-echo sequences screen out the low-frequency noise from millisecond-timescale parity switches, so a Ramsey-versus-echo comparison can serve as a diagnostic for charge-parity noise.","In higher-coherence devices where other decoherence channels are weaker, charge-parity noise will be relatively more important, not less.","Devices deeper into the transmon regime or with reduced quasiparticle poisoning should show suppressed parity-related T2* variation; the protocol gives a direct way to verify this."],"supporting_citations":[{"why":"Defines the transmon and the exponential suppression of charge dispersion with EJ/EC, establishing the operating point the paper tests.","marker":"[6]"},{"why":"Provides the measured ~150 kHz charge dispersion on the 1-2 transition and the correlated charge-noise context that the parity readout leverages.","marker":"[8]"},{"why":"Demonstrates millisecond charge-parity fluctuations and induced decoherence in a transmon, the direct precedent for linking parity flips to dephasing.","marker":"[10]"},{"why":"Supplies the quasiparticle-tunneling framework for hot nonequilibrium quasiparticles as the source of parity switching.","marker":"[11]"},{"why":"Shows that engineering can push parity switching rates down to ~1 mHz, providing the comparison that makes the ~1 kHz rate here significant.","marker":"[13]"},{"why":"Describes the phase method used for detuning-insensitive Ramsey and spin-echo measurements in the same tantalum transmon platform.","marker":"[19]"},{"why":"Explains noise-specific beating in Ramsey curves, which the paper invokes to interpret the two-frequency envelope in unsorted data.","marker":"[20]"},{"why":"Provides the Lindblad master-equation solver used in the parity-flip simulations that reproduce the T2* reduction.","marker":"[24]"},{"why":"Analyzes charge-parity switching effects and transmon design optimization, connecting the measured parity noise to device-design implications.","marker":"[14]"}],"fun_headline_variants":["Transmon dephasing tied to parity flips, not charge offset","Hidden noise source halves transmon coherence","Parity switches limit transmon T2 despite charge insensitivity","Fast parity flips dominate transmon dephasing","Transmon coherence limited by parity flips, not charge noise"],"cache_read_input_tokens":2816,"weakest_assumption_plain":"The entire parity classification rests on the assumption that at most one quasiparticle tunneling event happens during a single Ramsey shot and that the one-shot readout of the 1-2 transition never mislabels which parity is occupied; the paper does not quantify the readout fidelity.","fun_headline_variants_meta":{"raw":{"variants":["Transmon dephasing tied to parity flips, not charge offset","Hidden noise source halves transmon coherence","Parity switches limit transmon T2 despite charge insensitivity","Fast parity flips dominate transmon dephasing","Transmon coherence limited by parity flips, not charge noise"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000324,"raw_usage":{"total_tokens":1661,"prompt_tokens":758,"completion_tokens":903,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":502,"completion_tokens_details":{"reasoning_tokens":822}},"tokens_in":502,"tokens_out":903,"duration_ms":8783,"temperature":1.0,"reasoning_tokens":822,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T00:57:49.961681+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same Ramsey sequence with a parity detector that can catch two or more flips per shot, e.g., repeated or continuous parity readout during the delay. If the apparent T2* difference between unsorted and single-parity subsets vanishes, or if single-parity shots are found to contain more than one flip at the claimed ~1 kHz rate, the central mechanism would be contradicted.","supporting_citations":[{"cited_title":"Yu, Jay Gambetta, A","cited_arxiv_id":null,"evidence_quote":"Defines the transmon and the exponential suppression of charge dispersion with EJ/EC, establishing the operating point the paper tests."},{"cited_title":"Low-frequency correlated charge-noise measurements across multiple energy transitions in a tantalum transmon","cited_arxiv_id":null,"evidence_quote":"Provides the measured ~150 kHz charge dispersion on the 1-2 transition and the correlated charge-noise context that the parity readout leverages."},{"cited_title":"Rist` e, C","cited_arxiv_id":null,"evidence_quote":"Demonstrates millisecond charge-parity fluctuations and induced decoherence in a transmon, the direct precedent for linking parity flips to dephasing."},{"cited_title":"Hot nonequilib- rium quasiparticles in transmon qubits","cited_arxiv_id":null,"evidence_quote":"Supplies the quasiparticle-tunneling framework for hot nonequilibrium quasiparticles as the source of parity switching."},{"cited_title":"Phonon downconversion to suppress correlated errors in superconducting qubits","cited_arxiv_id":null,"evidence_quote":"Shows that engineering can push parity switching rates down to ~1 mHz, providing the comparison that makes the ~1 kHz rate here significant."},{"cited_title":"Beck, Vito Mar- iano Iaia, Anika Zaman, and Yaniv Jacob Rosen","cited_arxiv_id":null,"evidence_quote":"Describes the phase method used for detuning-insensitive Ramsey and spin-echo measurements in the same tantalum transmon platform."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Explains noise-specific beating in Ramsey curves, which the paper invokes to interpret the two-frequency envelope in unsorted data."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Lindblad master-equation solver used in the parity-flip simulations that reproduce the T2* reduction."},{"cited_title":"Charge- parity switching effects and optimisation of transmon-qubit design parameters","cited_arxiv_id":null,"evidence_quote":"Analyzes charge-parity switching effects and transmon design optimization, connecting the measured parity noise to device-design implications."}],"review_version":1}