{"id":"0ead22c3-3445-437a-a709-37c0ac42c663","arxiv_id":"2507.13740","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Local exact controllability of the periodic KdV equation is proved with control supported on arbitrary measurable space-time sets of positive measure, via a new observability strategy for dispersive equations on tori.","lead":"This paper proves that the KdV equation on a circle can be steered between nearby states by a control that acts only on a set of positive measure in both space and time. The result extends a classical controllability theorem to very irregular control regions and introduces a harmonic-analysis method that works for many dispersive equations.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The uniform multiplicity bound Θ≤2 in Proposition 3.5 is false: scalar cubic differences admit more than two representations (e.g. α=217 has four), so the high-frequency KdV observability proof has a load-bearing gap.","rationale":"I read the paper as a serious and novel contribution: the general observability framework in Section 2 is coherent, the augmented-observability induction is an interesting way to avoid compactness-uniqueness, and the fixed-point argument in Bourgain spaces follows a standard pattern. The central theorem, however, is proved through the mass-conserved KdV observability inequality in Proposition 3.4, whose high-frequency step is Proposition 3.5. The reader identified the weakest assumption there: the uniform multiplicity bound Θ≤2 for solutions of l³−k³=α. This is indeed false, and it is load-bearing. The Cauchy–Schwarz step in (3.12) needs a uniform bound on the number of high-frequency pairs contributing to each Fourier coefficient of 1_{E_T}; without it, the off-diagonal term is not controlled by the estimate displayed. The reader's example α=217 is concrete: the four pairs listed give Θ(217)≥4. I also checked that the bound Θ≤2 in Lemma 2.2 comes from fixing both components of λ_{k1}−λ_{k2}, which is a genuinely different quantity; the scalar difference m³−k³ does not fix m−k, so the analogy with Lemma 2.2 is invalid. I did not find a second independent gap of comparable severity. The manuscript may be repairable — the kernel B(k,m) has explicit structure that could yield a Schur-type estimate — but the argument as written does not contain that repair. Since the reader's verdict of CONDITIONAL already reflects exactly this situation, I do not recommend changing the verdict.","tokens_in":35884,"tokens_out":7739,"duration_ms":97124,"concrete_test":"Compute the scalar-cube multiplicity Θ(α)=#{(k,l)∈Z² : l³−k³=α} for α=217 by enumerating k in, say, [−100,100] and checking whether k³+217 is a perfect cube. If four or more pairs are found, the assertion 'Θ≤2' in Proposition 3.5 is false. To check whether the gap is otherwise repairable, re-derive the bound for the cross term in (3.12) directly from the explicit formula B(k,m)=(2π/|F|)(\\hat g(m−k)−2π\\hat g(k)\\hat g(m)) and verify whether the resulting kernel is bounded on l² without any uniform scalar-cube multiplicity assumption.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 3.5 controls the off-diagonal term (3.12) by Cauchy–Schwarz over pairs (k,m) with the same scalar frequency difference m³−k³, and invokes the bound Θ≤2 for the number of integer solutions to l³−k³=α. This bound is false for scalar cubic differences. For α=217 there are at least four ordered pairs (k,l) with l³−k³=217: (−9,−8), (−6,1), (−1,6), and (8,9). The bound Θ≤2 in the proof of Lemma 2.2 is for the full two-dimensional vector difference (k−l, p(k)−p(l)), where fixing both coordinates leaves at most d−1 solutions; it does not apply here because \\(\\widehat{1_{E_T}}(m³−k³)\\) fixes only the time frequency m³−k³, not the spatial frequency difference m−k. Since the l-sum over L(k,l)L(m,l) is bounded but not summed with any decay, the displayed estimate (3.12) does not follow as written, and hence Proposition 3.5, Proposition 3.4, and the HUM control construction built on them are not fully proved. A repair may be possible using the explicit kernel \\(B(k,m)=\\frac{2\\pi}{|F|}(\\hat g(m-k)-2\\pi\\hat g(k)\\hat g(m))\\) and Schur-type estimates, but no such argument appears in the manuscript.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies the periodic KdV equation on the torus with a control supported on the product of a measurable time set E_T⊂[0,T] and a measurable space set F⊂T, both of positive measure. The main result, Theorem 1.1, asserts local exact mass-conserved controllability around constant states: for any T>0 and any mean M, sufficiently nearby L² states of mean M can be joined by a trajectory of the forced equation with a control of the form L(h)1_{E_T×F}. The proof has three parts: (i) an observability inequality for general dispersive operators e^{itP(D)} from arbitrary space-time measurable sets (Section 2); (ii) a 'twisted' observability inequality for the linear KdV adjoint with the mass-conservation operator L (Section 3.1), followed by a Hilbert uniqueness method construction of the control operator; and (iii) a Bourgain-space fixed-point argument for the nonlinearity (Section 3.2). Section 4 applies the Section 2 observability to exponential stabilization with time-periodic or time-block-precompact damping. The Section 2 argument is self-contained and uses a high-frequency estimate together with a finite low-frequency induction rather than the moment method.","tokens_in":36171,"tokens_out":14742,"duration_ms":177936,"significance":"If correct, Theorem 1.1 would be a substantial extension of the classical Russell–Zhang and Laurent–Rosier–Zhang controllability results for the periodic KdV equation, since both the spatial and temporal control regions are merely measurable rather than open. The Section 2 observability for arbitrary positive-measure sets with polynomial phases is itself a clean contribution, and it is proved by an explicit high/low-frequency iteration with no fitted constants or ad hoc assumptions; the Bourgain-space contraction is standard and reasonably complete. The paper also contains a uniform resolvent estimate and an exponential stabilization result. The main reservation is that the KdV-specific high-frequency estimate in Proposition 3.5 rests on a false multiplicity bound; because this estimate feeds into Proposition 3.4 and the HUM controllability argument, the central claim is not fully established as written, although a repair appears possible.","major_comments":[{"comment":"The proof of Proposition 3.5 relies on the uniform multiplicity bound Θ≤2 for the number of integer pairs (k,l) with l³−k³=α. This bound is false: for α=217 the equation has four ordered pairs, (−9,−8), (−6,1), (−1,6), and (8,9), and there is no uniform bound of this type for scalar cubic differences. The Cauchy–Schwarz estimate in (3.12) needs a bound on the number of representations of m³−k³ in order to pass from a double sum over (k,m) to ∑|φ̂(k)|²; fixing only the time frequency does not fix the spatial frequency difference, so the vector-difference argument in Lemma 2.2, which fixes both coordinates, does not apply. Consequently (3.12), Proposition 3.5, and the observability (3.9) and HUM construction built on it are not proved as written. A repair using the explicit kernel in (3.5), for instance a Schur-type estimate for B(k,m), should be supplied.","section":"§3.1.2, Proposition 3.5, Eqs. (3.11)–(3.12)"},{"comment":"With the definition ⟨φ⟩_T = (1/|T|)∫_T φ dx and |T|=2π, the state u0 − 2M/π does not have mean zero when ⟨u0⟩_T=M; the correct shift is u0−M, and the linearized drift coefficient is M, not 2M/π. Unless a different normalization for the mean is explicitly adopted, the reduction of Theorem 1.1 to the M=0 case in Section 3.1 is inconsistent. This is easily repaired, but it must be fixed for the statement for arbitrary M to follow.","section":"§3.1, reduction to mean-zero states"},{"comment":"Lemma B.2 is stated only for T∈(0,1), but Theorem 1.1 claims controllability for every T>0. The contraction argument for Ψ in Section 3.2 invokes Lemma B.2 without any reduction or a version valid on arbitrary time intervals. Please add the standard rescaling or partition argument, or replace Lemma B.2 by a statement covering all T>0 with a constant depending on T.","section":"§3.2, Lemma B.2 and arbitrary T"}],"minor_comments":[{"comment":"The proof header in this subsection says 'Proof of Theorem 1.5'; it should refer to Proposition 1.5, which is the statement proved there.","section":"§2.2.2"},{"comment":"The reference [Bur25] is listed as 'Privite discussion, 2025', while the text credits 'Burq and Zhu' with a recent observability result; please provide a proper citation of the actual preprint or paper, or clearly mark it as a personal communication.","section":"References, [Bur25]"},{"comment":"The inequality in Theorem 4.1 should read ∥u(t,·)∥_{L²(T)} ≤ Ce^{−γt}∥u0∥_{L²(T)}, not Ce^{−γt}∥u0∥²_{L²(T)}, since the proof iterates the square of the L² norm.","section":"Theorem 4.1"},{"comment":"The formula 'Lψ(t)1_F(x) = ψ(t)1_F(x)L' is not meaningful as written; please clarify the intended action of the operator on the product ψ(t)1_F.","section":"§3.1.1, Lemma 3.1"},{"comment":"In the sentence beginning 'Since |l³−k³|...', the variable l is used in the displayed estimate and m in the following sentence; use a single pair (k,m) throughout to avoid confusion.","section":"§3.1.2, proof of Proposition 3.5"}],"recommendation":"major_revision","confidential_remarks":"The Section 2 observability framework appears sound and is a genuine strength of the manuscript. The main obstacle is Proposition 3.5: the false multiplicity bound is a real, load-bearing error, but it looks repairable via the explicit kernel in (3.5). I would encourage the editor to invite a revision rather than reject, provided the repair is convincing. Please also verify the Burq–Zhu reference: as currently listed it is a 'private discussion', not a citable article."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the observability framework in Section 2: proving space-time measurable observability for polynomial dispersive equations by a Miheev-style augmented induction. That part reads as essentially correct - the Theta <= d-1 bound for full vector differences holds, and the high/low frequency iteration is a real improvement over the moment method for rough control sets. This deserves serious attention from the control and dispersive PDE community.\n\nThe KdV mass-conserved application, however, has a load-bearing gap. Proposition 3.5 needs a uniform bound on the number of integer solutions to l^3 - k^3 = alpha to control the off-diagonal terms. The paper claims Theta <= 2, apparently transferring the reasoning from Lemma 2.2. That reasoning does not transfer: Lemma 2.2 fixes the full two-dimensional difference (k-l, p(k)-p(l)), which leaves at most d-1 solutions by the fundamental theorem of algebra. Here only the time frequency m^3 - k^3 is fixed; the spatial frequency difference is free, and the scalar cubic difference problem is much richer. The stress-test note is correct: alpha = 217 already has four representations, and cabtaxi numbers give more. As a result, the displayed estimate (3.12) does not follow, and with it the high-frequency estimate, its uniform translation version, and the HUM construction all lose their support.\n\nIs this fatal to the program? Not necessarily. The general observability theorem in Section 2 stands on its own and is likely correct. The KdV result may also be repairable - for instance, using the explicit kernel B(k,m) to get Schur-type decay instead of a uniform multiplicity bound - but such an argument is not in the manuscript. As written, the main theorem is not established.\n\nCredit where it is due: the paper is self-contained, the nonlinear step is a clean contraction argument, and there is no sign of retrofitted constants or circular reasoning. The mistake is specific, localizable, and honestly identifiable. The bibliography is appropriate.\n\nI would send this to a serious referee. The flaw is subtle, the methodology is valuable, and the general theorem alone justifies referee time. The referee should check Proposition 3.5 carefully and ask the authors to supply the repair if it exists. If they close the gap, this becomes a substantial paper. Recommended: engage with it, but do not accept as is.","headline":"The general observability theorem is a genuine advance, but the mass-conserved KdV proof has a false multiplicity bound that leaves the main result unproved as written.","tokens_in":36681,"tokens_out":2776,"would_cite":false,"duration_ms":32797,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q53","42A99","76B15","93C20"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves local exact controllability of the periodic KdV equation with controls on arbitrary positive-measure space-time sets.","keywords":["KdV equation","exact controllability","observability inequality","space-time measurable sets","mass conservation","Bourgain spaces","augmented observability","dispersive equations on torus"],"falsifier":"For α=217, the four integer pairs (m,k)=(9,8), (−8,−9), (6,−1), and (1,−6) all satisfy m³−k³=217, which contradicts the uniform bound Θ≤2 that Proposition 3.5 uses to dominate the off-diagonal part of the high-frequency estimate; a corrected proof would need a different bound on representations of integers as differences of cubes.","tokens_in":35681,"feed_emoji":"🌊","tokens_out":15205,"duration_ms":165033,"temperature":0.7,"pith_summary":"The paper sets out to prove that the periodic KdV equation can be steered exactly, near any constant state with fixed total mass, by a control supported on an arbitrary measurable set of positive measure in both space and time. Previous controllability results for KdV required the spatial control region to be open and the time actuation to cover the full interval; here both may be rough. The proof establishes a new observability inequality for free dispersive evolutions from positive-measure space-time sets, using a high-frequency/low-frequency iteration rather than the moment method or the compactness-uniqueness method. A controlled linear system is obtained through the Hilbert uniqueness method, and a Bourgain-space fixed-point argument extends the result to the nonlinear KdV equation.","feed_headline":"KdV is locally controllable from any measurable space-time set","feed_subtitle":"A new observability argument handles actuation regions that are only measurable in both time and space.","key_machinery":"The core object is the augmented observability inequality: once a set of Fourier modes Λ is observable from the positive-measure set G under small translations—meaning from G∩(G−h) for all small |h|—a single extra frequency can be added, and Λ∪{λ} becomes observable from G itself. Iterating this finitely many times upgrades a high-frequency estimate over {|k|>N} into an observability inequality over the whole spectrum. The iteration rests on two estimates: a high-frequency bound whose off-diagonal error is controlled by the decay of the Fourier coefficients of the measurable set, and an L⁴ Strichartz bound showing that the high-frequency sum has small L²-mass on the thin difference set G\\(G−h). In the mass-conserved KdV case the scheme runs on the doubled lattice {(k³,l)}; the mass-removing operator L is represented by matrix coefficients L(k,l)=ĝ(l−k)−2πĝ(−k)ĝ(l), and the coercivity ∥L($e^{{ikx}}$)∥²_{L²}>δ for k≠0 supplies the lower bound at each step.","core_discovery":"The central claim is Theorem 1.1: for every time horizon T>0 and every mean M, there is a δ>0 such that any two states with mean M whose deviations from that mean are L²-small can be joined by a solution of ∂tu+∂x³u+u∂xu = L(h)1_{E_T×F}, with h∈L²([0,T]×T) and E_T,F arbitrary positive-measure measurable sets. Mass is conserved by the special form of the control operator, which subtracts the mean of the forcing on F. The proof proceeds in three layers: a general observability theorem for dispersive evolutions on the torus, a linear observability inequality for the mass-conserved KdV of the form ∥φ∥²_{L²} ≲ ∫∫_{E_T×F}|L(S(t)φ)|² dxdt, and a contraction argument in Bourgain spaces for the nonlinearity. In the author's own framing, the novelty is a finite iterative scheme that adds low frequencies one at a time while keeping the high-frequency estimate stable under small translations of the observed set.","pith_inferences":["The counterexample to the Θ≤2 bound for m³−k³=α concerns only the mass-conserved KdV section; the general observability theorem for monic degree-d polynomials uses the true bound Θ≤d−1, so the general measurable-set scheme is not affected.","A repair of the mass-conserved argument could replace the false uniform bound by a divisor-counting estimate for m³−k³=(m−k)(m²+mk+k²), whose number of representations grows sub-polynomially in α.","The translation-stability idea suggests that controllability costs should depend on a density or thickness of the measurable set rather than on openness, which could be probed numerically on Galerkin truncations with fractal or sparse observation sets.","The exponential-stabilization theorem indicates that damping on a translate of a fixed positive-measure set in every time block, with phases changing arbitrarily from block to block, should still yield uniform decay rates."],"forward_implications":["Near any constant state with fixed mean, any two sufficiently small L² perturbations can be joined exactly in time T by a solution whose control is supported on E_T×F, with both sets merely measurable of positive measure.","The general observability theorem applies to every monic polynomial dispersion on the torus, not only to KdV, so the same iteration should produce local controllability for a broad class of constant-coefficient dispersive equations.","The method replaces the two classical tools—the moment method's biorthogonal family and the compactness-uniqueness method's uniqueness continuation—by translation-stable high-frequency estimates and finite low-frequency insertion.","For damped dispersive equations whose damping is positive on a positive-measure set in each time block and whose blocks form a precompact family, the same observability gives uniform exponential decay of the L² norm.","The nonlinear step is a contraction in Bourgain spaces, so the local controllability statement comes with a concrete bound on the control cost in terms of the data size."],"supporting_citations":[{"why":"Supplies the prior local exact mass-conserved controllability result for KdV with an open spatial control region and full time interval, the statement Theorem 1.1 extends.","marker":"[RZ96]"},{"why":"Provides the global mass-conserved controllability and stabilization framework on the periodic domain that the rough-setting version builds on.","marker":"[LRZ10]"},{"why":"Introduces the Bourgain Fourier-restriction norms and the bilinear estimate used in the nonlinear fixed-point argument.","marker":"[Bou93]"},{"why":"Gives the sharp global well-posedness and bilinear Bourgain-space estimates for KdV on the torus that the contraction argument relies on.","marker":"[CKS+03]"},{"why":"Classical argument for controlling high-frequency trigonometric sums on a measurable set; its Fourier-decay step is reused for the off-diagonal error.","marker":"[Zyg72]"},{"why":"Supplies the harmonic-analysis augmented-observability lemma whose finite iteration is the paper's main new mechanism.","marker":"[BD06]"},{"why":"Is the classical compactness-uniqueness observability proof for KdV that the new method is designed to bypass.","marker":"[Ros97]"}],"fun_headline_variants":["Local KdV control from any measurable time-space region","Measurable regions suffice for local KdV control","New observability proof enables local KdV control","Any positive-measure set drives local KdV controllability","Local exact KdV control via iterative observability"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The mass-conserved KdV observability proof assumes as a load-bearing premise that any integer can be written as a difference of two integer cubes in at most two ways; that premise is false, since 217 is both 9³−8³ and 6³−(−1)³ and actually has four such representations, and the high-frequency estimate that controls off-diagonal terms depends on it.","fun_headline_variants_meta":{"raw":{"variants":["Local KdV control from any measurable time-space region","Measurable regions suffice for local KdV control","New observability proof enables local KdV control","Any positive-measure set drives local KdV controllability","Local exact KdV control via iterative observability"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00025,"raw_usage":{"total_tokens":1489,"prompt_tokens":817,"completion_tokens":672,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":433,"completion_tokens_details":{"reasoning_tokens":598}},"tokens_in":433,"tokens_out":672,"duration_ms":7355,"temperature":1.0,"reasoning_tokens":598,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:22:11.717509+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For α=217, the four integer pairs (m,k)=(9,8), (−8,−9), (6,−1), and (1,−6) all satisfy m³−k³=217, which contradicts the uniform bound Θ≤2 that Proposition 3.5 uses to dominate the off-diagonal part of the high-frequency estimate; a corrected proof would need a different bound on representations of integers as differences of cubes.","supporting_citations":[],"review_version":1}