{"id":"fb6f2324-3445-40bd-b2ea-3107fdf508e9","arxiv_id":"2507.06799","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Optimal control enables fast, low-excitation coherent splitting of a 1D Bose-Einstein condensate into a double well, preserving number squeezing down to 350 microsecond ramps.","lead":"Researchers used optimal control to split a cloud of ultracold atoms into two halves much faster than normal, with minimal sloshing or breathing motion and preserved quantum correlations. The technique could make it faster and easier to prepare entangled atomic states for quantum sensors and simulators.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central claim rests on the reduced quartic GPE model being valid across the full parameter range; the paper's own OC+ correction and deferred calibration leave this unvalidated, especially at large A.","rationale":"The paper's central claim is that optimal-control ramps computed from a reduced 1D GPE suppress classical excitations and preserve quantum correlations across a wide parameter range, enabling fast splitting. The strongest evidence is the experimental comparison of fringe spacing and width variances for OC vs linear ramps (Fig. 2), which shows clear suppression, and the two-step engineering results (Fig. 3). These are real, reproducible measurements with jackknife uncertainties, and I would not dispute them. The load-bearing assumption is the fidelity of the effective model used to generate the OC ramps. The model has only five parameters and a fixed quartic form; the calibration (Appendix D) uses trap frequencies at the minima and minima positions, which do not fully determine the potential in the barrier region. The manuscript itself concedes that performance degrades at larger A 'due to increased sensitivity to modeling inaccuracies' and introduces OC+ time rescaling and adjusted hold times to compensate. This is a red flag that the model is not uniformly valid, yet the companion paper containing the derivation and validation is unpublished. The reader's verdict of CONDITIONAL is appropriate: the specific demonstrations stand, but the generality ('wide parameter range', 'approaching speed limit', and the implied predictive power of the model) cannot be fully assessed until the model is validated against the full potential and, ideally, a broader experimental scan. My concern matches the reader's weakest assumption, so I recommend no change to the verdict.","tokens_in":8980,"tokens_out":8901,"duration_ms":100446,"concrete_test":"Compute the full RF-dressed adiabatic potential V_full(x,A) from the experimental static magnetic field and RF coupling (as in refs. [17,18]) for the same A values used in Fig. 2c, and compare to the calibrated V_model(x,A)=a2(A)x^2+a4x^4. Quantify the mismatch in barrier height V_full(0)-V_full(xm) and in the potential at the well positions. If the barrier height differs by more than ~10% for A>0.5 (the regime where OC+ was needed), or if the potential difference exceeds the local chemical potential anywhere along the splitting trajectory, then the OC ramps are optimized for the wrong potential and the 'wide parameter range' claim is unsupported. In that case the published OC+ correction is a symptom of model error, not a robustness feature.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The OC ramps are computed by solving a boundary-value problem for the reduced 1D GPE with V(x,A)=a2(A)x^2+a4x^4 (Appendix C/D). The calibration constrains only the local curvature at the minima and the minima positions as functions of A; it does not directly validate the potential away from the minima, in particular the central barrier height and the shape between the wells. During fast splitting the wavefunction is strongly excited and explores precisely this region. The paper states (Fig. 2c caption and main text) that 'for larger final dressing amplitudes, the performance is limited by the simplicity of the model' and introduces a post hoc linear time rescaling (OC+) with γ≪1 to restore performance, as well as 'adjusted hold times' in the engineering experiments. This shows that the model is not uniformly accurate over the claimed parameter range. Since the OC trajectories are entirely model-based, any quantitative error in V or g⊥ shifts the computed optimal ramp away from the true optimum. The companion paper [22] is cited for the derivation and calibration but is unpublished, so the central assumption—that the quartic model captures all relevant physics for A up to 0.6 and ramps as fast as 350 μs—is not independently checkable from the manuscript. If the barrier height or anharmonicity deviates from the quartic form in this regime, the claimed suppression of sloshing/breathing and the 'wide parameter range' generalization are not established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental implementation of optimal-control (OC) ramps for splitting a one-dimensional 87Rb Bose-Einstein condensate into a double-well potential created by RF dressing. The authors use a reduced 1D Gross-Pitaevskii equation with a quartic double-well potential V(x,A)=a2(A)x^2+a4x^4, calibrated from measured trap frequencies, to compute control ramps that suppress sloshing and breathing. They report that OC ramps outperform linear ramps across a range of final amplitudes and ramp durations, enable clean splitting in 350 us (approaching a simulated speed limit of about 200 us), and preserve number squeezing for fast ramps. They also demonstrate two-step protocols in which OC sequences remove excitations created by a preceding linear ramp.","tokens_in":9312,"tokens_out":3392,"duration_ms":41237,"significance":"If the reported results hold, the work offers a practical route to fast, low-excitation splitting of BECs, which is directly relevant to quantum simulation and metrology. The experimental comparisons are strengthened by quantitative variance observables with jackknife uncertainties and by the demonstration of OC engineering starting from non-adiabatic states. The main weakness is that the central ingredient—the reduced model and its calibration—is deferred to an unpublished companion paper, and the paper itself acknowledges the need for a post hoc time-rescaling correction at large final amplitudes. These issues affect the evidential weight of the 'wide parameter range' claim but not the validity of the core experimental comparison for the regimes where the model is adequate.","major_comments":[{"comment":"The reduced model V(x,A)=a2(A)x^2+a4x^4 is calibrated (Appendix D) using only trap frequencies extracted from sloshing oscillations, which constrain the local curvature at the minima and the minima positions. This does not directly validate the potential away from the minima, in particular the central barrier height and the anharmonic shape between the wells. During fast splitting, the wavefunction is strongly excited and explores exactly this region. The derivation and full validation are cited to the unpublished companion paper [22], leaving the regime of validity of the model—especially at larger A where the paper itself reports degraded performance—insufficiently established in the manuscript.","section":"Appendix C and D"},{"comment":"The OC+ scheme is introduced with a 'simple linear time rescaling' parameter γ after observing that OC performance degrades at larger final amplitudes. Since γ is chosen post hoc to match the experimental data, it acts as an additional fitted parameter, and its use weakens the claim that the OC ramps are purely model-predictive. The manuscript should specify how γ is selected (e.g., cross-validation on independent data) and report the sensitivity of the OC+ results to this choice. Without this, the statement that OC outperforms linear ramps 'for almost all scenarios' is not fully supported for the largest Afinal values.","section":"Fig. 2c and OC+"},{"comment":"The estimated speed limit tlim ≈ 200 us is obtained by instantaneously switching to an inverted parabolic potential and measuring the time for the density to reach the final well positions. This is a heuristic dynamical time, not a rigorous lower bound, and it depends on the same reduced model whose accuracy is questioned in the large-A regime. The phrase 'theoretical speed limit' should be qualified to avoid overstating the rigor of this estimate.","section":"Splitting-speed limit"}],"minor_comments":[{"comment":"The squeezing factor ξ^2_N− is shown without error bars or jackknife uncertainties; since the claim of preserved quantum correlations rests on these values, adding uncertainty estimates would make the claim more convincing.","section":"Fig. 4"},{"comment":"The phrase 'minimal classical excitations' is used several times; 'minimized' or 'strongly suppressed' would be more precise, since the data do not establish that excitations reach the theoretical minimum.","section":"Abstract and Introduction"},{"comment":"The fitting function f(x) is described with parameters σ, C, k0, and φ; it would help to state explicitly which parameters are free in the fit and how the fringe contrast C is obtained, as this affects the extraction of k0 and σ.","section":"Appendix A"},{"comment":"There is a typo in the caption ('T op row' should be 'Top row'), and the markers for linear, OC, and OC+ are not identified in the caption, making the figure harder to read.","section":"Fig. 2 caption"},{"comment":"Reference [22] is listed as 'in preparation'; for a self-contained letter, at least the key calibration equations and validation plots should be included in the supplemental material, especially because the main text relies heavily on this unpublished work.","section":"Reference [22]"}],"recommendation":"major_revision","confidential_remarks":"The experimental demonstration appears solid, and the comparison against linear ramps with jackknifed variances is a strength. My main concern is that the model, which is the backbone of the OC protocol, is calibrated and validated only in an unpublished companion paper, and the paper's own OC+ correction indicates that the model is not uniformly accurate over the claimed parameter range. I would recommend asking the authors to include the essential calibration and validation material in the main text or supplement, and to clarify the selection and statistical treatment of the OC+ rescaling parameter. If these points are addressed, the paper could be suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear X,\n\nShort version: this is a solid experimental letter, and the central claim—OC ramps suppress sloshing and breathing during fast BEC splitting much better than linear ramps—is supported by the data. It is not a conceptual revolution; optimal control on BECs and RF-dressed double wells are both established. The new piece is the systematic experimental implementation of OC-designed splitting ramps, the two-step excitation/suppression protocol, and reaching 350 µs splitting near a ~200 µs estimated speed limit. That is genuinely useful for atom-chip interferometry and squeezed-state preparation.\n\nThe comparison against linear ramps is the paper's backbone, and it is done honestly: variances of k0 and σ over the hold time, with jackknife uncertainties, and OC consistently wins across the scanned range. The calibration from measured trap frequencies is a reasonable approach, and the authors are transparent that performance degrades at larger A and that OC+ uses a few-percent time rescaling. The two-step engineering experiments also require adjusted hold times. These are real limitations, but the paper reports them rather than hiding them.\n\nThe soft spots are the ones the reader flagged. The five-parameter quartic model is calibrated to local curvature and minima positions, which does not directly validate the barrier height and shape between wells—exactly the region probed during fast splitting. The full derivation and calibration are in an unpublished companion paper [22], so a referee cannot currently check whether the model is adequate across the claimed parameter range. OC+ is a post hoc fix, and the 'wide parameter range' statement should be read with that in mind. The quantum-correlation claim is more modest than the abstract suggests: number squeezing is preserved but reduced for fast ramps, and phase squeezing is obscured by detection noise. That is consistent with expectations, not a problem, but it is not a demonstration of entanglement preservation.\n\nI think the paper deserves serious peer review. The experimental demonstration is the kind of thing the community will use, and the limitations are addressable with an expanded appendix, the companion paper, and raw data. My recommendation: send it to review, and require the calibration details and data availability before acceptance.","headline":"A solid experimental demonstration that OC-designed ramps suppress sloshing and breathing in fast BEC splitting, with honest reporting of model limitations; deserves peer review once calibration details and data are made available.","tokens_in":9863,"tokens_out":2018,"would_cite":true,"duration_ms":24223,"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":"Optimal-control ramps computed from a calibrated five-parameter model split a one-dimensional Bose-Einstein condensate into a double well in 350 microseconds with minimal sloshing and breathing, approaching the estimated speed limit near…","keywords":["Bose-Einstein condensate","optimal control","coherent splitting","double-well potential","radio-frequency dressing","sloshing","breathing","number squeezing"],"falsifier":"Measure the post-ramp hold dynamics for the 350 microsecond OC ramp into the coupled trap with $A_{\\mathrm{final}}=0.5$ and compare the variance of the fringe spatial frequency and the transverse cloud width against the linear-ramp values; if the variances are comparable, or the interference pattern is not stationary within the first millisecond, the calibrated-model chain would be refuted. A sharper check is to compare the model's predicted trap frequency at $A\\gtrsim0.6$ with a direct sloshing-frequency measurement, since that is the regime where the paper must invoke the OC+ time-rescaling.","tokens_in":8818,"feed_emoji":"⚛️","tokens_out":6875,"duration_ms":73276,"temperature":0.7,"pith_summary":"The paper reports a method for splitting a one-dimensional Bose-Einstein condensate into two halves fast enough to beat ordinary adiabatic ramps, without shaking the atoms. The central claim is that optimal control ramps computed from a five-parameter reduced Gross-Pitaevskii model can suppress the two dominant classical excitations, inter-well sloshing and transverse breathing, across a wide range of splitting speeds. The same control tool can remove motion that a prior fast ramp deliberately created, and number squeezing below the standard quantum limit survives even at the shortest demonstrated ramp time of 350 microseconds, approaching the estimated speed limit of about 200 microseconds. If this holds, fast splitting becomes a usable starting point for preparing entangled states.","feed_headline":"350 microseconds: optimal control splits a BEC without shaking it","feed_subtitle":"Control ramps computed from a five-parameter model suppress sloshing and breathing, preserving squeezing at the fastest ramps.","key_machinery":"The central object is a calibrated reduced Gross-Pitaevskii equation in the transverse splitting direction, with the effective double-well potential $V(x,A)=a_2(A)x^2+a_4x^4$ and free parameters $A_s, c, \\kappa_1, \\kappa_2, g_\\perp$. Here $A_s$ marks the transition from a single-well to a double-well geometry, $c$ sets the separation of the minima, $\\kappa_1$ and $\\kappa_2$ set the local curvature on the two sides of the transition, and $g_\\perp$ is the effective transverse interaction strength. The machinery works by fitting these parameters to measured oscillation frequencies of kicked condensates, using the fitted equation as a dynamic constraint in a boundary-value energy-minimization problem solved by indirect optimal control, and then applying a small linear time rescaling $A(t)\\mapsto A((1-\\gamma)t)$ with $\\gamma\\ll 1$ for very large final amplitudes where the model becomes less accurate.","core_discovery":"The paper's central claim is that optimal control turns coherent splitting of a 1D Bose-Einstein condensate into a shortcut to adiabaticity: the atoms end in the double-well ground state of motion even when the ramp is far too fast for a linear ramp. The control sequences are computed in the transverse direction from a reduced Gross-Pitaevskii equation whose effective potential is $V(x,A)=a_2(A)x^2+a_4x^4$, with five calibrated parameters: the splitting point $A_s$, a length scale $c$, two curvature scales $\\kappa_1$ and $\\kappa_2$, and the effective transverse interaction $g_\\perp$. The calibrated model is checked against measured trap frequencies, and the optimized ramps reduce the variance of the fringe spatial frequency and transverse cloud width nearly to zero over a range of ramp durations and final dressing amplitudes. Two-step protocols, in which a linear ramp first creates strong motion and an OC ramp then removes it, reach excitation levels comparable to a single direct OC ramp. The fastest demonstrated clean ramp is 350 microseconds, and at that speed the number-squeezing factor stays below the standard quantum limit, indicating that quantum correlations survive.","pith_inferences":["Editorial inference: the kick-and-measure calibration procedure could likely be reused for other atomchip potentials without solving the full three-dimensional potential, because it only requires local curvature and minimum position as functions of the control parameter.","Editorial inference: the monotonic reduction of squeezing with ramp duration makes ramp time a continuous in-situ dial for the amount of number squeezing, which the paper does not explicitly propose as a metrological resource.","Editorial inference: a natural test is to map the time-rescaling factor $\\gamma$ as a smooth function of the final amplitude; if such a function exists, the model could be corrected globally rather than patched ramp by ramp."],"forward_implications":["OC ramps split cleanly in 350 microseconds, while linear ramps of the same duration fail to localize the atoms in the two wells; linear ramps typically need about 10 milliseconds or longer to avoid excitations.","Suppression of sloshing and breathing holds across a broad range of both ramp duration and final splitting amplitude, not just at one operating point.","The two-step engineering protocol shows that OC ramps can remove pre-existing classical motion, so controlled excitation followed by controlled de-excitation is a viable building block.","The number-squeezing factor remains below the standard quantum limit at fast ramp times, and its magnitude decreases monotonically as the ramp shortens, consistent with less time for correlations to build.","The estimated ballistic speed limit of about 200 microseconds is nearly reached at 350 microseconds, suggesting that further shortening is possible but limited by atomic transport timescales."],"supporting_citations":[{"why":"Supplies the optimal-control method for Bose-Einstein condensates in magnetic microtraps that is adapted to the reduced model.","marker":"[7]"},{"why":"Derives the radio-frequency dressed adiabatic potentials that motivate the double-well model form.","marker":"[17]"},{"why":"Analyzes manipulation of ultracold atoms in dressed radio-frequency potentials, providing the basis for the effective potential parametrization.","marker":"[18]"},{"why":"Contains the full calibration procedure and the derivation of the five-parameter model coefficients.","marker":"[22]"},{"why":"Provides the numerical optimal-control approach for Bose-Einstein condensates used to solve the boundary-value problem.","marker":"[28]"},{"why":"Defines number squeezing relative to the standard quantum limit and its role in witnessing entanglement.","marker":"[11]"},{"why":"Shows how residual sloshing modulates tunnel coupling, motivating the suppression of classical motion.","marker":"[21]"},{"why":"Identifies the collective modes, sloshing and breathing, whose variances are the paper's control objectives.","marker":"[23]"}],"fun_headline_variants":["Optimal control splits a BEC in 350 µs with no shake","Optimal control suppresses motion for fast coherent BEC splitting","Optimal control yields a shortcut to adiabatic BEC splitting","Optimal control beats linear ramps for fast BEC splitting","Optimal control preserves squeezing in a 350-µs BEC split"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole scheme rests on the claim that a fixed five-parameter one-dimensional mean-field model with potential $V(x,A)=a_2(A)x^2+a_4x^4$ reproduces the real transverse splitting dynamics for every ramp the control algorithm proposes, including very fast ramps and large final dressing amplitudes.","fun_headline_variants_meta":{"raw":{"variants":["Optimal control splits a BEC in 350 µs with no shake","Optimal control suppresses motion for fast coherent BEC splitting","Optimal control yields a shortcut to adiabatic BEC splitting","Optimal control beats linear ramps for fast BEC splitting","Optimal control preserves squeezing in a 350-µs BEC split"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001268,"raw_usage":{"total_tokens":5149,"prompt_tokens":866,"completion_tokens":4283,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":482,"completion_tokens_details":{"reasoning_tokens":4192}},"tokens_in":482,"tokens_out":4283,"duration_ms":27053,"temperature":1.0,"reasoning_tokens":4192,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T18:54:01.295491+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the post-ramp hold dynamics for the 350 microsecond OC ramp into the coupled trap with $A_{\\mathrm{final}}=0.5$ and compare the variance of the fringe spatial frequency and the transverse cloud width against the linear-ramp values; if the variances are comparable, or the interference pattern is not stationary within the first millisecond, the calibrated-model chain would be refuted. A sharper check is to compare the model's predicted trap frequency at $A\\gtrsim0.6$ with a direct sloshing-frequency measurement, since that is the regime where the paper must invoke the OC+ time-rescaling.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the optimal-control method for Bose-Einstein condensates in magnetic microtraps that is adapted to the reduced model."},{"cited_title":"Hofferberth, I","cited_arxiv_id":null,"evidence_quote":"Derives the radio-frequency dressed adiabatic potentials that motivate the double-well model form."},{"cited_title":"Lesanovsky, T","cited_arxiv_id":null,"evidence_quote":"Analyzes manipulation of ultracold atoms in dressed radio-frequency potentials, providing the basis for the effective potential parametrization."},{"cited_title":"Zhang, M","cited_arxiv_id":null,"evidence_quote":"Contains the full calibration procedure and the derivation of the five-parameter model coefficients."},{"cited_title":"Shao and D","cited_arxiv_id":null,"evidence_quote":"Provides the numerical optimal-control approach for Bose-Einstein condensates used to solve the boundary-value problem."},{"cited_title":"Est` eve, C","cited_arxiv_id":null,"evidence_quote":"Defines number squeezing relative to the standard quantum limit and its role in witnessing entanglement."},{"cited_title":"Berrada, S","cited_arxiv_id":null,"evidence_quote":"Shows how residual sloshing modulates tunnel coupling, motivating the suppression of classical motion."},{"cited_title":"W¨ urkner, Y","cited_arxiv_id":null,"evidence_quote":"Identifies the collective modes, sloshing and breathing, whose variances are the paper's control objectives."}],"review_version":1}