{"id":"eb5e2e4e-04a2-4530-a69f-0a1deb6c0990","arxiv_id":"2607.24050","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"PDH feedforward cuts 132-kHz servo-bump phase noise by 15 dB on a 1762-nm barium-ion optical-qubit laser and improves Rabi coherence time from ~3.7 to ~21.7 µs.","lead":"The paper shows that feeding the residual error signal of a Pound-Drever-Hall-locked 1762-nm laser through a delayed phase modulator suppresses its servo-bump phase noise by 15 dB and lengthens the coherence time of a trapped-ion optical qubit about sixfold when the drive sits inside the noisy band. The relevance is practical: it removes a constraint that currently limits gate speeds and pushes trapped-ion groups toward cavity filtering or frequency avoidance.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Gate-fidelity improvement estimate hinges on a second-order cumulant model used at the edge of its stated validity; direct 15-dB and T2 measurements stand.","rationale":"The reader's weakest assumption identifies exactly the same load-bearing concern: the quantitative bridge from measured spectra to qubit fidelity rests on a second-order cumulant expansion that the paper itself marks as potentially invalid in the relevant noise regime, and the calibration is circular. My stress-test agrees and sharpens this with the internal slope discrepancy (0.084 vs 0.1) as corroborating evidence that the quantitative model is not self-consistent. The direct experimental findings — 15 dB suppression verified by DSHI and the 6× Rabi-coherence improvement — are independent of the model and remain credible. Therefore the CONDITIONAL verdict is appropriate; no change is needed. The proposed Monte-Carlo test would settle whether the fidelity-gain estimate survives outside the perturbative approximation.","tokens_in":19028,"tokens_out":3035,"duration_ms":30028,"concrete_test":"Replace the second-order TCL generator by direct Monte-Carlo propagation of the stochastic Bloch equation (Eq. 12) using the measured DSHI-derived detuning-noise PSD with the same fitted scale factor, for both feedforward-off and feedforward-on conditions. Compare the simulated Rabi-envelope T2 and π-gate infidelity at f_R = 20, 63, 132, and 345 kHz against the TCL predictions in Fig. 5. If the Monte-Carlo infidelity differs by more than ~0.05 (or the T2 by more than ~20%) at 63 kHz or 132 kHz, the perturbative model is unreliable in the claimed improvement region and ΔF_avg ≈ 0.22 must be re-quoted with uncertainties or removed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative payoff is the claimed ΔF_avg ≈ 0.22 phase-noise-limited π-gate improvement. This number is produced by the finite-time second-order time-convolutionless generator of App. A. Appendix A explicitly warns that the truncated map 'may lose complete positivity' when the accumulated cumulant is not small. The paper's own un-stabilized estimate reaches 1−F_avg ≈ 0.498 at f_R = 63 kHz (App. F/Fig. 5), a regime where second-order perturbation theory is not controlled by any error bound given in the paper. Furthermore, the model is calibrated 'using the measured Rabi decay to calibrate the overall noise scale' (Sec. V.B), so the same observable is used both to fit and to validate the model; the 0.22 gain is therefore not an independent prediction. The model-internal tension is visible in the log-linear scaling: App. G derives slope 0.100 per dB, while the measured fit in Fig. 4 gives 0.084 per dB with R² = 0.992 — a 16% discrepancy not explained in the text. These issues do not undermine the directly measured 15 dB DSHI suppression or the 6× T2 improvement, but they do mean the fidelity-gain headline is not yet quantitatively supported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports an experimental demonstration of Pound-Drever-Hall (PDH) feedforward phase-noise suppression for the 1762 nm laser driving the S1/2↔D5/2 optical-qubit transition in 138Ba+. The authors recycle the residual PDH error signal through a delay-matched electrical path to a fiber EOM, and measure, via delayed self-heterodyne interferometry, a ~15 dB suppression of the servo-bump phase noise near 132 kHz without measurable linewidth broadening (160±15 Hz by DSHI versus 156±16 Hz by ion spectroscopy, App. H). They then present a theoretical model of a resonantly driven qubit under stochastic detuning noise, based on a second-order time-convolutionless expansion and a Pauli-transfer-matrix average gate infidelity, and use DSHI spectra plus a spectral calibration factor to estimate a phase-noise-limited π-gate fidelity improvement ΔF_avg ≈ 0.22 near the servo-bump region. Trapped-ion Rabi measurements show that enabling feedforward increases the extracted coherence time from 3.69±0.35 µs to 21.73±0.87 µs for Rabi frequencies near the bump, and the D-state population at the first π time increases from 0.641±0.014 to 0.870±0.009. The paper also reports a log-linear relation between T2 and servo-bump suppression in dB.","tokens_in":19296,"tokens_out":4428,"duration_ms":42750,"significance":"If the direct measurements stand, the work has clear practical value: it provides a power-efficient alternative to cavity filtering that extends the usable Rabi-frequency range of optical-qubit gates without sacrificing optical throughput, and it is the first experimental integration of PDH feedforward with trapped-ion optical-qubit control. The 15 dB suppression, the unchanged linewidth, and the ~6x improvement in Rabi coherence time are direct, cross-checked measurements and are the strongest parts of the paper. However, the specific quantitative gate-fidelity gain (ΔF_avg ≈ 0.22) and the Fig. 5 infidelity curves depend on a perturbative finite-time model that the paper itself warns may lose complete positivity, and the model is calibrated using the very Rabi-decay observable it is used to explain. These issues do not invalidate the direct measurements, but they mean the gate-fidelity headline is not yet quantitatively supported. The paper includes detailed appendices and a clear discussion of the model's limitations, which is helpful, but the limitations are not reflected in the strength of the central fidelity claim.","major_comments":[{"comment":"The phase-noise-limited gate-fidelity estimate ΔF_avg ≈ 0.22 is produced by the second-order time-convolutionless generator of App. A. Appendix A explicitly warns (after Eq. A13) that for stronger noise or longer evolution times the truncated map 'may lose complete positivity' and that Monte-Carlo propagation should be used as the reference. The paper's own Fig. 5/App. F shows an un-stabilized infidelity of 1−F_avg ≈ 0.498 at f_R ≈ 63 kHz, which is essentially the maximum possible average infidelity for a unital single-qubit channel. No error bound, no comparison with Monte-Carlo simulation, and no finite-size check is given to establish that the second-order expansion is controlled in this regime. Therefore the quantitative fidelity improvement and the Fig. 5 curves are not reliably supported. The direct 15 dB and T2 measurements stand independently, but the gate-fidelity claim should b","section":"App. A, App. F, Fig. 5, Sec. V.B"},{"comment":"The model is calibrated using the measured Rabi decay: App. F states that 'a single spectral calibration factor, determined from the feedforward-off Rabi-decay measurement, is used for all Rabi frequencies and both feedforward conditions.' The same observable (Rabi decay / T2) is thus used both to fix the noise scale and to validate the model's predictions. This makes the ΔF_avg ≈ 0.22 estimate, and the per-dB T2 scaling in Fig. 4, model-dependent rather than independent predictions. I would like to see an out-of-sample check — e.g., calibrating on one Rabi frequency and predicting another, or predicting the feedforward-on data using only feedforward-off calibration — or a sensitivity analysis showing how the inferred fidelity gain changes with the calibration factor.","section":"Sec. V.B, App. F"},{"comment":"There is an unexplained quantitative discrepancy in the log-linear relation. Appendix G derives that if the dephasing rate is proportional to the phase-noise PSD at the Rabi frequency, then log10(T2) should increase with slope 0.100 per dB of suppression. The measured fit in Fig. 4 gives a slope of 0.084 per dB with R² = 0.992 — a 16% deviation. This discrepancy is not discussed in the text, yet the abstract and discussion quote the empirical '1.21x per dB' factor. Since the per-dB improvement is a central quantitative claim, the origin of this deviation (finite-time filter effects, non-Lorentzian noise shape, non-exponential decay, or fitting choices) should be addressed, or the theoretical slope should not be presented as the expected behavior without comment.","section":"App. G, Fig. 4"}],"minor_comments":[{"comment":"The caption states 'pi time used: 22.1 us,' which corresponds to a Rabi frequency of approximately 22.6 kHz, not the 132 kHz servo-bump frequency discussed elsewhere in the text. Please clarify whether the frequency scan was performed at a Rabi frequency away from the bump, and state explicitly how the off-resonant excitation shoulder at ±132 kHz is produced in that case.","section":"Fig. 3b"},{"comment":"The improvement in D-state probability at the first π time (0.641→0.870) is a single-state transfer measurement, not an average gate fidelity. The authors acknowledge this distinction in Sec. V.B, but the Conclusion states that the results show 'clear improvements in the coherence time and gate fidelity.' I recommend reserving 'gate fidelity' language for actual randomized or state-tomography-based fidelity measurements, or for the model-based estimate after it has been properly validated.","section":"Sec. V.B / Conclusion"},{"comment":"There are several typographical errors and slightly awkward phrasings, e.g., 'supressing' in the Conclusion, 'an qubit' in the Introduction, and 'without limited by the transmission optical power' in Sec. II.A. A careful proofread would improve clarity.","section":"Abstract / Introduction"},{"comment":"The statement 'available from the corresponding author upon reasonable request' is weaker than current best practice. Since the Figs. 2–5 and the calibration procedure are central to the quantitative claims, I encourage depositing the data and analysis scripts in a permanent repository (e.g., Zenodo or figshare) so that the DSHI spectra, Rabi fits, and calibration factor can be independently checked.","section":"Data Availability"},{"comment":"The text says the red points at ≈345 kHz are 'shown for comparison' and are not included in the fit. It would be helpful to state this directly in the figure caption and to report the confidence interval on the fitted slope, not only R².","section":"Fig. 4"}],"recommendation":"major_revision","confidential_remarks":"The experimental core — 15 dB DSHI servo-bump suppression, unchanged linewidth, and the 6x Rabi-coherence-time improvement — appears solid and suitable for a quantum-optics journal. The main weakness is that the quantitative gate-fidelity gain is produced by a perturbative model operating in a regime where the paper itself flags loss of complete positivity, and the model is calibrated on the same observable it predicts. I believe this can be fixed within the manuscript's scope by either (i) replacing or benchmarking the second-order model with Monte-Carlo propagation of the stochastic Bloch equation, (ii) adding an out-of-sample calibration check, or (iii) reframing the headline claim as a direct coherence-time/state-transfer improvement and demoting the ΔF_avg estimate to a clearly labeled, model-dependent illustration. Given that the direct measurements are publishable, I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the part that's solid: this is the first experimental demonstration of PDH feedforward applied to a trapped-ion optical-qubit laser. They measure ~15 dB suppression of the servo bump at ~132 kHz with DSHI, show the linewidth doesn't broaden (160±15 Hz vs 156±16 Hz from ion spectroscopy), and see a real coherence improvement — T2 goes from 3.69±0.35 µs to 21.73±0.87 µs when the Rabi frequency is in the bump band. The 300 m cable control showing incomplete cancellation is a good experimental touch. These direct measurements are credible and cross-checked.\n\nThe soft spot is the model that converts spectra to gate fidelity. The headline ΔF_avg ≈ 0.22 comes from a second-order time-convolutionless expansion that App. A itself warns can lose complete positivity at stronger noise — and their own un-stabilized estimate reaches 1−F ≈ 0.498 at 63 kHz, exactly that regime. The scale factor is calibrated using the feedforward-off Rabi decay, so the model is being fit and validated on the same data. There's also a 16% discrepancy between the measured log-linear T2 slope (0.084 per dB) and the model's predicted 0.100 per dB, and the fidelity gain is quoted without uncertainty. These don't undermine the direct measurements, but they mean the quantitative gate-fidelity claim is not yet supported.\n\nThe authors are open about several of these limitations, which is to their credit. The fix is straightforward: publish the spectra and T2 data with ΔdB coordinates, reconcile or explain the slope discrepancy, and re-quote fidelity gains with uncertainties.\n\nI'd send this to peer review — the experimental core is valuable and the technique is transferable to other optical qubits. I'd also bring it to a reading group; it's a good example of where direct measurements outrun the theory used to interpret them.","headline":"First demonstration of PDH feedforward on a trapped-ion optical-qubit laser: the 15 dB bump suppression and 6x T2 improvement are direct and credible; the ΔF≈0.22 fidelity estimate is model-bound and not yet supported.","tokens_in":19874,"tokens_out":2249,"would_cite":true,"duration_ms":21672,"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":"PDH feedforward suppresses servo-bump noise by 15 dB and lifts optical-qubit Rabi coherence sixfold.","keywords":["Pound-Drever-Hall feedforward","laser phase noise","servo bump","optical qubit coherence","delayed self-heterodyne interferometry","trapped-ion quantum gates","barium-138 quadrupole transition"],"falsifier":"Run a positivity-preserving Monte-Carlo simulation of the stochastic Bloch equation (rather than the second-order truncation) using the measured DSHI phase-noise spectrum, and compare its predicted π-pulse infidelity versus Rabi frequency to the paper's Fig. 5 curves; disagreement near f_R≈63 kHz, where the second-order estimate reaches the unital-channel ceiling of about 0.5, would show the calibrated infidelity model is unreliable in that regime.","tokens_in":18855,"feed_emoji":"🔬","tokens_out":5849,"duration_ms":50508,"temperature":0.7,"pith_summary":"This paper establishes that the residual Pound-Drever-Hall error signal of a cavity-locked laser can be recycled as a feedforward phase correction, canceling the servo-bump phase noise that standard PDH feedback cannot remove. Demonstrated on the 1762 nm quadrupole-transition laser of a trapped 138Ba+ optical qubit, the method suppresses bump noise by about 15 dB near the 132 kHz bump peak, leaves the laser linewidth unchanged, and improves the measured Rabi coherence time from 3.69±0.35 µs to 21.73±0.87 µs when the Rabi frequency lies inside the bump band. A finite-time noise-response model connects the measured phase-noise spectrum to a phase-noise-limited average π-gate infidelity, predicting a gain of about ΔF≈0.22. If correct, the technique extends the usable Rabi-frequency range of optical-qubit gates without the optical-power penalty of cavity filtering.","feed_headline":"PDH feedforward cuts servo bump 15 dB, boosts qubit coherence sixfold","feed_subtitle":"Recycling the lock error signal through a fiber EOM lifts Rabi T2 from 3.7 to 21.7 µs without broadening linewidth.","key_machinery":"The load-bearing object is the PDH error signal, which after tight locking is proportional to the residual instantaneous phase noise ϕ(t), including components beyond feedback bandwidth. Recycling it with inverted gain and matched delay into a fiber EOM gives ϕ_out≈ϕ(t)+G_ff ϕ(t−τ); with G_ff=−1 and τ tuned to the 132 kHz bump, the bump noise cancels. The setup uses a 325 m BNC delay line and 20 m fiber; DSHI with 15 m fiber resolves the bump spectrum and a 5 km fiber measures the linewidth. The qubit model is a second-order time-convolutionless expansion of the toggling-frame Bloch equation, yielding a finite-time Pauli-transfer matrix M_I(t), the infidelity ϵ=(3−Tr[M_I(t)])/6, and a scalar","core_discovery":"The central claim is that PDH feedforward—applying the locked laser's own residual error signal, after matched delay and inverted gain, to a fast fiber EOM—cancels the servo-bump phase noise of the PDH lock beyond feedback bandwidth. The authors verify this on a 138Ba+ S1/2↔D5/2 optical qubit: DSHI spectra show up to 15 dB suppression at the bump peak with no linewidth broadening, Rabi oscillations driven near the bump show a roughly sixfold increase in T2, and a calibrated finite-time dynamical map translates the measured spectrum into a phase-noise-limited π-gate fidelity improvement of about 0.22. They present this as the first experimental demonstration that PDH feedforward is compatible","pith_inferences":["One could use the same feedforward loop to suppress bumps at multiple or higher frequencies by choosing delay lines matched to several bump harmonics, effectively widening the usable gate-speed window beyond the single peak demonstrated here.","The finite-time Rabi-filter formalism suggests a design heuristic: to predict the coherence impact of a given locking servo, evaluate the overlap between the phase-noise PSD and the Rabi filter for the target pulse time, rather than simply reading the PSD at the Rabi frequency.","A direct extension would be to apply PDH feedforward to Mølmer–Sørensen two-qubit gates and measure the entangling-gate fidelity as a function of motional frequency; the single-qubit results imply a measurable improvement when a motional sideband coincides with the bump.","The model's reliance on a single spectral calibration factor could be sharpened by measuring Rabi decay at several Rabi frequencies and checking whether the predicted log-linear T2 scaling across the whole bump region is borne out independently of the fitted decay."],"forward_implications":["Single-qubit optical gates can be operated at Rabi frequencies inside the servo-bump band, removing a constraint on gate-speed choice imposed by PDH lock noise.","Because the technique recycles the already-available error signal, it adds no optical loss; it is a power-preserving alternative to high-finesse cleanup cavities when transmitted cavity power is only microwatts.","The measured log-linear relation (≈1.21× in T2 per dB of bump suppression) gives a quantitative rule for predicting coherence gain from spectral noise reduction.","The method is presented as readily transferable to other optical qubits and transitions, with the main hardware caveat being the power handling of the phase actuator at the operating wavelength.","For two-qubit gates, suppressing servo bumps near motional frequencies should mitigate fidelity loss in Mølmer–Sørensen gates, which the authors identify as a motivation for future work."],"fun_headline_variants":["PDH feedforward cancels servo bump, 15 dB quieter","Sixfold qubit T2 gain from PDH feedforward trick","First PDH feedforward kills servo bump on ion qubit","Servo-bump phase noise slashed 15 dB via EOM feedforward","Rabi T2 jumps 3.7 to 21.7 µs with PDH feedforward"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The quantitative bridge from spectra to qubit claims—specifically the ΔF≈0.22 fidelity gain—assumes the second-order time-convolutionless expansion is valid at the strong-noise operating points, even though the paper's Appendix A warns that the truncated map may lose complete positivity for stronger noise and the spectral scale factor is fitted to the very decay it explains.","fun_headline_variants_meta":{"raw":{"variants":["PDH feedforward cancels servo bump, 15 dB quieter","Sixfold qubit T2 gain from PDH feedforward trick","First PDH feedforward kills servo bump on ion qubit","Servo-bump phase noise slashed 15 dB via EOM feedforward","Rabi T2 jumps 3.7 to 21.7 µs with PDH feedforward"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000141,"raw_usage":{"total_tokens":1028,"prompt_tokens":801,"completion_tokens":227,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":545,"completion_tokens_details":{"reasoning_tokens":126}},"tokens_in":545,"tokens_out":227,"duration_ms":2813,"temperature":1.0,"reasoning_tokens":126,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T23:11:42.272001+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a positivity-preserving Monte-Carlo simulation of the stochastic Bloch equation (rather than the second-order truncation) using the measured DSHI phase-noise spectrum, and compare its predicted π-pulse infidelity versus Rabi frequency to the paper's Fig. 5 curves; disagreement near f_R≈63 kHz, where the second-order estimate reaches the unital-channel ceiling of about 0.5, would show the calibrated infidelity model is unreliable in that regime.","supporting_citations":[],"review_version":1}