{"id":"946dcd62-1c8f-4ba3-9c5c-b7174b7e2917","arxiv_id":"2607.28783","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Quantum optimal control is extended to multiple intermediate times via costate jumps, enabling controlled steering of revivals in a five-spin Heisenberg chain.","lead":"This paper shows how to steer a quantum system so that a chosen property is maximized at multiple moments in time, not only at the final moment, by adding controlled jumps in the optimization variables. The authors demonstrate it on a small spin chain, shaping the driving field to control when a quantum excitation reappears after the field is switched off.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sec. II.B.2 asserts that Krotov convergence for each finite-width window O_τ(t) transfers to the τ→0 Dirac-delta limit, but no exchange-of-limits proof or discrete-time monotonicity argument is provided; this gap is load-bearing because the paper's central claim is monotonic convergence for instanta","rationale":"The reader's weakest assumption identifies the same load-bearing concern: the transfer of Krotov's monotonic-convergence guarantee from finite-width regularized windows to the τ→0 discontinuous costate algorithm is asserted, not proved. My independent reading of Sec. II.B.2 confirms that the argument is a one-sentence limit claim, and Sec. III.A's numerical evidence is suggestive but explicitly non-uniform and incomplete. I considered whether there is a stronger objection, e.g., the transversality condition Eq. (5) appearing inconsistent with a final-time δ-objective, but the implementation appears to use the standard χ(T_end)=Oψ(T_end) condition, so this is a presentation issue rather than a load-bearing flaw. The variational derivation of the jump condition is internally consistent, the numerics demonstrate the expected trade-off, and the missing step is addressable by either a formal proof or a direct monotonicity/convergence test. Therefore the appropriate verdict remains CONDITIONAL, matching the reader's assessment; no change is needed.","tokens_in":12268,"tokens_out":7098,"duration_ms":83256,"concrete_test":"Instrument the Krotov iteration for the Dirac-delta costate update (Eq. 10) on the 5-spin example and record J^{(i)} = C_5(T_int) + C_5(T_end) after each iteration; verify J^{(i+1)} ≥ J^{(i)} for many iterations and several random initial fields. Separately, repeat the Sec. III.A finite-width study for τ = 0.1, 0.05, 0.01, 0.005, reporting max_t |C5^{(τ)}(t) − C5^{(0)}(t)| and the L2 difference between the optimal fields. If the errors do not decrease systematically with τ, or if the delta-update J^{(i)} is non-monotonic in any run, the claimed τ→0 convergence and monotonicity are not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim — monotonic optimization and convergence in the limit of zero-width measurement windows — rests on the argument in Sec. II.B.2. The paper says Krotov converges for each positive-definite O_τ(t), 'including τ→0'. But the cited theorem [34] applies to regular time-dependent objectives satisfying smoothness/boundedness conditions; a finite sum of Dirac deltas is not such an objective. 'Convergence for all τ>0' does not imply that the limiting update rule with the discontinuous costate jump in Eq. (10) inherits the monotonicity inequality. The standard Krotov proof relies on a non-negative cross-term involving the time-dependent objective; with delta sources, that term becomes a discrete sum whose sign is not established by taking τ→0 in the regularized statement. The numerical section supports the conclusion but does not close the gap: Figs. 1–2 show close agreement for τ = 0.5, 2, 5, 10, yet the text explicitly notes nonuniform and inexact convergence, attributing discrepancies to numerical instabilities. No iteration-level monotonicity data or systematic τ-scaling of the error is reported. Since the advertised novelty is exact instantaneous objectives, not merely finite-width approximations, this unproved transfer is load-bearing. The variational derivation and numerics are credible, so the issue is a proof gap rather than a demonstrated failure.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends quantum optimal control (QOC) to objectives that are superpositions of instantaneous expectation values at multiple intermediate times, represented by Dirac-delta window functions O(t)=Σ_k O_k δ(t−T_k). Starting from the variational QOC functional, the authors derive a jump condition for the costate at each measurement time (Eq. (7)). They implement this in a Krotov-based scheme, applying it to a five-spin Heisenberg XXX chain in the single-excitation manifold, with the goal of maximizing the population of a target Dicke state at one intermediate time and at the final time. They compare the Dirac-delta limit with finite-width window regularizations, study the resulting control fields and their power spectra, and apply the method to post-pulse, field-free revivals, demonstrating a trade-off between intermediate and final objectives.","tokens_in":12668,"tokens_out":7604,"duration_ms":82473,"significance":"If the convergence claim is correct, the framework is a useful extension of QOC to multi-time tracking and to control problems in which observables must be maximized at specified intermediate instants. The variational derivation of the jump condition is internally consistent, and the numerical results are physically plausible and demonstrate a clear effect of intermediate measurements on the optimized fields and populations. The relation to the single-measurement formalism of Ref. [26] is clearly acknowledged, and the multi-time extension together with the spin-chain applications is a reasonable incremental contribution. However, the advertised rigorous validation of the τ→0 limit is not actually supplied: the proof in Sec. II.B.2 is only a heuristic argument, and the numerics in Sec. III.A are qualitative. The central claim therefore needs either a direct convergence proof or a clearly softened statement supported by quantitative diagnostics.","major_comments":[{"comment":"The transfer of Krotov convergence from the regularized objective O_τ(t) to the τ→0 discontinuous problem is asserted rather than proved. The sentence 'Krotov algorithm converges for O_τ(t) for all τ, including τ→0' does not follow from the cited theorem [34], since that theorem's hypotheses (smoothness/boundedness of the time-dependent objective) fail for a finite sum of Dirac deltas, and monotone convergence for each τ>0 does not imply that the limiting algorithm with the jump update (Eq. (10)) inherits monotonicity. In the standard Krotov monotonicity proof the cross-term involves the time-dependent objective; with delta sources that term becomes a discrete sum, and its non-negativity is exactly what needs to be established. Because the abstract and Introduction advertise rigorous convergence, this gap is load-bearing. Please provide a direct proof of monotone convergence for the jump","section":"Sec. II.B.2, Eq. (11)"},{"comment":"The numerical evidence for the τ→0 limit is qualitative. The paper reports that even for τ=0.5 and 2 the population dynamics 'converge closely but do not coincide exactly,' and attributes the discrepancy to numerical instabilities. No iteration counts, no per-iteration monotonicity data for the optimized functional, and no quantitative error measure (e.g., sup-norm or L2 difference between finite-τ and τ→0 results as a function of τ) are given. Without these diagnostics, the claim of convergence in the zero-window limit rests on visual closeness in selected plots. Adding such metrics would also help distinguish genuine convergence from numerical artifacts.","section":"Sec. III.A, Figs. 1–2"}],"minor_comments":[{"comment":"There are typos: 'fullfilled' and 'fullfills' should be 'fulfilled' and 'fulfills'; 'aknowledged' should be 'acknowledged' in the Acknowledgments.","section":"Throughout"},{"comment":"The functions f_k(t−T_k,τ) are introduced in Sec. II.B.2 but only defined later in Sec. III.A, Eq. (16). Define them earlier or state explicitly that they are positive nascent deltas with unit integral.","section":"Eq. (11)"},{"comment":"The text says O_τ(t) is 'positive definite.' For projectors, the appropriate term is 'positive semidefinite.'","section":"Sec. II.B.2"},{"comment":"The sentence 'Specifically, we maximize the population of |5⟩ at T_int=100 and T_end=200 for FWHM τ=0.5, 2, 5, 10 at each measurement time' appears twice, once before and once after Eq. (15). Remove the duplication.","section":"Sec. III.A"},{"comment":"The one-sided limits in Eq. (7) are clear, but the update in Eq. (10) would benefit from explicit left/right notation, e.g., χ(T_k^-) ← χ(T_k^+) + O_k ψ(T_k), to remove any ambiguity about the direction of backward propagation.","section":"Eqs. (8)–(10)"},{"comment":"The captions could state more explicitly that the plotted quantity is the optimal control field. The y-axis ranges differ between panels in Fig. 6; a note explaining this would avoid misleading comparisons.","section":"Figs. 2 and 6"}],"recommendation":"major_revision","confidential_remarks":"The paper's central variational derivation is sound, and the numerical applications are interesting. The main concern is the unproved exchange of limits in the convergence argument, which is advertised as a central contribution. I do not see this as a fatal flaw — the numerical evidence is suggestive — but the manuscript should either supply a rigorous proof for the jump update or substantially moderate the claim and add convergence diagnostics. The novelty relative to Ref. [26] is modest but real; the multi-time extension is the distinguishing element."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is a clean, practical extension of the authors' earlier single-time costate-jump formalism (Ref. [26]) to multiple intermediate observables, implemented in Krotov and demonstrated on a 5-spin XXX chain. The variational derivation of Eq. (7) is internally consistent, and the numerics in Sec. III.A show the expected smooth approach to the Dirac-delta limit. The physical trade-off between intermediate targets and final population is clearly illustrated, and the post-pulse revival control is a nice application. Credit where due: the method works in practice, and the authors are honest about the cost of forcing revivals at low-baseline times.\n\nThe soft spots are real but not fatal. The convergence claim in Sec. II.B.2 is asserted, not proved: convergence for each finite-width O_τ(t) does not imply the τ→0 limit inherits monotonicity, and the cited theorem [34] does not apply to a sum of Dirac deltas. This is a formal gap, and the numerics are supportive but not conclusive—no iteration counts, no systematic τ-scaling of errors, no code released. The 'tracking' language in the abstract and conclusions is an overreach; these are open-loop optimizations at a handful of times, not continuous tracking. The citation to Ref. [26] is appropriate; self-citation is not a problem here.\n\nThe paper is not a paradigm shift, but it is a solid incremental contribution that QOC practitioners will find useful. The proof gap should be addressed in revision—either a real exchange-of-limits argument or a softened claim that the method converges in practice. With that, I'd take it as a solid paper. Deserves a serious referee.","headline":"Useful incremental extension of the costate-jump method to multiple intermediate times, with credible numerics; the zero-width convergence proof is a gap but not a fatal one.","tokens_in":13086,"tokens_out":2513,"would_cite":true,"duration_ms":28497,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Mid-pulse targets become reachable in quantum optimal control","keywords":["quantum optimal control","intermediate-time observables","costate jump","Krotov method","Dicke states","Heisenberg spin chain","revivals","finite-width measurement windows"],"falsifier":"Run the iterative algorithm with a positive-definite observable and record the objective after each iteration in the τ→0 limit: if the objective ever decreases from one iteration to the next, the monotonic-convergence claim is false. A direct numerical check on a system where the finite-width optima do not approach the delta-limit optimum would also test the limit-exchange assumption.","tokens_in":12173,"feed_emoji":"🎯","tokens_out":3834,"duration_ms":40790,"temperature":0.7,"pith_summary":"This paper extends quantum optimal control so that a single driving pulse can maximize expectation values of observables at several chosen times during the evolution, not only at the final time. The key move is to treat each intermediate measurement as an infinitesimally narrow time window; the variational equations then acquire a jump in the costate at each such time. The authors implement the jump in an iterative optimal-control algorithm, argue that the iteration is monotonic and converges to the instantaneous-measurement limit, and demonstrate on a five-spin Heisenberg chain that one pulse can simultaneously shape population at an intermediate and a final time. If correct, this gives a practical recipe for tracking, protecting, and reviving quantum states on demand.","feed_headline":"Mid-pulse targets become reachable in quantum optimal control","feed_subtitle":"Adding a costate jump at each measurement lets one pulse shape states at several times, boosting post-pulse revivals.","key_machinery":"The central object is the costate jump condition: each instantaneous measurement inserts a discontinuity into the Lagrange multiplier trajectory, with the jump update χ(T_k) ← χ(T_k) + O_k ψ(T_k). This jump is embedded in the standard backward-forward iterative optimal-control loop, while the finite-width window regularization O_τ(t) supplies the convergence rationale and the practical route to the zero-width limit.","core_discovery":"The paper claims that imposing an observable at an intermediate time T_k acts as a Dirac-delta term in the variational objective, and that this term forces the Lagrange multiplier (costate) to jump: the value just before T_k equals the value just after plus O_k times the state at T_k (Eq. 7). Inserting this jump into the standard backward-forward iteration makes the algorithm monotonic for positive-definite observables, and the delta limit is recovered by narrowing finite-width measurement windows. On a five-spin Heisenberg XXX chain restricted to one excitation, the method reaches about 0.999 population of the target Dicke state at the final time and about 0.993 at the intermediate time, an","pith_inferences":["The same jump-condition formalism should admit distinct observables at different times, not just repeated measurements of one operator; this would let a pulse track one quantity early and another late, a step the paper gestures at but does not test.","A formal proof of monotonic convergence in the zero-width limit is not given; one can test convergence numerically on more complex systems where the optimization landscape is not as forgiving.","Finite-width windows may serve as a built-in robustness knob: broader windows deliberately sacrifice peak value to tolerate timing jitter, which could be exploited in experiments with limited time resolution.","Applied to larger chains or interacting many-body systems, the revival-control scheme could be used to engineer dynamical decoupling or to prepare non-equilibrium states at prescribed times."],"forward_implications":["A single optimized pulse can maximize the same observable at two or more times, so intermediate and final objectives can be balanced rather than prioritized.","The instantaneous-measurement protocol is the limit of finite-width measurement windows, so finite temporal resolution can be modeled explicitly and the protocol inherits a convergence guarantee in that limit.","Post-pulse field-free revivals can be enhanced or created at chosen times; creating revivals at low-baseline times costs final fidelity, while enhancing existing revivals is cheaper.","After an intermediate measurement, the optimization exploits transitions among all populated eigenstates, not only ground-state pathways, as seen in the field power spectrum."],"fun_headline_variants":["Quantum control handles intermediate times via costate jumps","Spin-chain revivals steered by multi-time quantum control","Costate jumps enable simultaneous targets in optimal control","Quantum control tracks states and revivals with new algorithm","Intermediate measurements reshape spin-chain control fields"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The claim that the iterative algorithm converges in the zero-window limit transfers the proof of monotonic convergence from each finite-width window to the discontinuous limit without proving the exchange of limit and optimization.","fun_headline_variants_meta":{"raw":{"variants":["Quantum control handles intermediate times via costate jumps","Spin-chain revivals steered by multi-time quantum control","Costate jumps enable simultaneous targets in optimal control","Quantum control tracks states and revivals with new algorithm","Intermediate measurements reshape spin-chain control fields"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000204,"raw_usage":{"total_tokens":1202,"prompt_tokens":696,"completion_tokens":506,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":440,"completion_tokens_details":{"reasoning_tokens":434}},"tokens_in":440,"tokens_out":506,"duration_ms":7090,"temperature":1.0,"reasoning_tokens":434,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T00:24:38.009670+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the iterative algorithm with a positive-definite observable and record the objective after each iteration in the τ→0 limit: if the objective ever decreases from one iteration to the next, the monotonic-convergence claim is false. A direct numerical check on a system where the finite-width optima do not approach the delta-limit optimum would also test the limit-exchange assumption.","supporting_citations":[],"review_version":1}