{"id":"face5318-d8ed-42f6-b414-712196c6f860","arxiv_id":"2411.15069","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Periodic KdV is locally well-posed in H^s for s > -2/3 with small data, proved without using complete integrability.","lead":"This paper proves local well-posedness of the periodic KdV equation for initial data in H^s with s > -2/3, extending the previous non-integrable threshold of -1/2. The proof uses a modulation-restricted normal form with an initial-data-dependent phase, and requires the data to be sufficiently small.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The section 6 contraction map is not defined on Y1 x Z1: Gamma2 is written as a function of phi1 only even though (19) contains v, so the Banach fixed-point step in Theorem 1 is not justified by the text.","rationale":"The central claim is local well-posedness of periodic KdV for s > -2/3 without integrability. To carry the argument, the paper needs a genuine contraction in a data-dependent space. The most fragile step is not any single bilinear estimate; it is the logical bridge between the estimates and the fixed point. Section 5.3 explicitly assumes the relation u = h + v and defers the corresponding change in the fixed-point equation, and Section 6 writes Gamma with Gamma2 depending only on phi1 while equation (19) and the terms in (18) contain v. If the correct map really is Gamma2(phi1, phi2), the displayed definition is incomplete; if the intended map is Gamma2(phi1), the displayed contraction estimate using ||phi2 - psi2||_Z is unjustified and the role of the product space B is unclear. Either way, the proof of Theorem 1 as written does not close the contraction argument. The reader's weakest_assumption identified the same substitution/cutoff gap, and I agree that this is the load-bearing soft spot. The paper does contain substantial plausible material: Lemmas 5, 7, 9, 10 and the cancellation Lemma 4 are detailed and appear internally coherent, and the modulation-restricted normal form idea is a reasonable adaptation of prior mKdV work. But because the gap sits exactly at the point where multilinear estimates are converted into existence, the central claim is not fully demonstrated. The concrete test above, writing Gamma explicitly and verifying the contraction estimate, would settle whether the gap is merely expository or substantive.","tokens_in":1263,"tokens_out":1863,"duration_ms":239481,"concrete_test":"Write the full fixed-point map for the pair (phi1, phi2) by expanding equation (19) with u = phi1 and v = phi2 in all terms of N_R listed in (18), including R(phi1,phi1,phi1) - R(phi2,phi2,phi2), M(phi1,phi1,phi1), N^h(phi1,phi1), and R^h(phi1,phi1,phi1). Then verify whether the resulting Gamma2 depends on phi2. Compute Gamma2(phi1,phi2) - Gamma2(psi1,psi2) using Lemmas 6, 8, 9, and 10 and check whether it is controlled by a contraction factor times (||phi1-psi1||_{Y1} + ||phi2-psi2||_{Z1}) with factor < 1. If closure requires eliminating phi2 via u = h(u) + v, the substitution must be made before Lemma 8 and the resulting scalar fixed-point equation for u alone must be stated and contracted in Y1.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 5.3 states that the estimates of Lemma 8 are carried out under the assumption u = h + v and then says 'our fixed point equation will need to have this substitution reflected', before ignoring it for readability. In Section 6, the map Gamma is defined as Gamma(phi1, phi2) = (Gamma1(phi1), Gamma2(phi1)), with Gamma2 depending only on phi1. However, the RHS of equation (19) contains v_n explicitly in the integral term and also inside N_R through terms like R(u,u,u) - R(v,v,v), R^h(u,u,u), M(u,u,u), and N^h(u,u). Unless v is replaced via u = h + v, the displayed map is not the intended contraction on the product space Y1 x Z1. Moreover, the contraction estimate compares Gamma2(phi1) and Gamma2(psi1) but the displayed difference includes a term proportional to ||phi2 - psi2||_{Z1}, which is not justified by the definition as written. This is load-bearing because uniqueness and continuous dependence in Theorem 1 are obtained from Banach's fixed point theorem in the ball B subset Y1 x Z1.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript develops a modulation-restricted normal form method to prove local well-posedness of the periodic KdV equation in H^s(T) for s > -2/3, thereby extending the previously known threshold s >= -1/2 obtained via bilinear X^{s,b} estimates. After an L^2-based rescaling, the authors introduce a data-dependent phase φ_n and define associated Bourgain-type spaces Y and Z. They split the solution as u = T^ell(u,u) + v, derive a Duhamel equation (19) for v, and state a series of multilinear estimates (Lemmas 5-10) for the nonlinear terms. Theorem 1 then asserts existence, uniqueness, and continuity of the data-to-solution map for sufficiently small data in a data-dependent space Y, with the proof intended to follow from a Banach fixed-point argument in Section 6.","tokens_in":25016,"tokens_out":4887,"duration_ms":44644,"significance":"If the result is correct, it is significant: it would give the first local well-posedness result for periodic KdV below H^{-1/2} without using complete integrability, matching the mKdV analogue of Nakanishi, Takaoka, and Tsutsumi up to a small-data caveat. The modulation-restricted normal form and the spectral-phase-dependent Bourgain spaces are novel tools for KdV and are likely to be reusable. The paper contains a substantial amount of detailed symbolic and case-based analysis, and the data-dependent phase is derived from the resonant term rather than fitted to force the conclusion. However, as written, the central contraction argument has load-bearing gaps that prevent the theorem from being established.","major_comments":[{"comment":"The map Γ(φ1, φ2) is defined as (Γ1(φ1), Γ2(φ1)), with Γ2 depending only on φ1, but the right-hand side of equation (19) contains v_n explicitly, both in the integral term ∫_0^t Re(v_n N_R)(s)ds and in the terms of N_R listed in (18), such as R(u,u,u) - R(v,v,v). Since v is not replaced by u - h inside Γ2 as written, the displayed Γ2 is not a functional of φ1 only, and the subsequent contraction estimate comparing Γ2(φ1) with Γ2(ψ1) cannot produce the claimed factor ||φ2 - ψ2||_{Z1}. This gap undermines the Banach fixed-point step and hence the uniqueness and continuous-dependence conclusions in Theorem 1.","section":"Section 6, definition of Γ"},{"comment":"The proof of Lemma 8, Case 3, uses the decomposition u = h + v, and Remark 4 states that the full equation should contain M^{(1,2)}(u,u,u) + M^{(3)}(h+v,u,u) in place of M(u,u,u). However, this substitution is not implemented in the fixed-point equation of Section 6; Section 5.3 explicitly says 'We ignore this minor detail in favor of increased readability.' As a result, the estimates of Lemma 8 are not shown for the actual contraction map, and this is load-bearing because Lemma 8 is one of the principal estimates used to bound N_R in (30).","section":"Section 5.3, Remark 4"},{"comment":"The passage from the original equation (1) to the time-cutoff equation (31) is asserted rather than proved. The text states that η ∈ H^8_t and that 'all of our spaces are bounded with respect to this cutoff', allowing the authors to 'disregard this minor technical difficulty.' Similarly, the compatibility of the cancellation in Lemma 4 with the cutoff is hand-waved by saying that a reconstitution of Littlewood-Paley blocks 'reduces us purely to the spatial constraint of Case 3.' Since the fixed-point argument is run on the cutoff equation, these assertions need to be demonstrated, especially because the data-dependent phase φ_n and the resonance removal are sensitive to the exact form of the nonlinearity.","section":"Section 6, time cutoff"}],"minor_comments":[{"comment":"The substitution u ↦ <∇>^{-s}u and the subsequent renaming s -> |s| is confusing; the reader must keep track that after the substitution, positive s in [1/2, 2/3) corresponds to negative regularity in Theorem 1. It would help to state this correspondence explicitly.","section":"Section 2"},{"comment":"The functional }v}_X = min(}v}_\\tilde{X}, }v}_{X^{2s-1+, -1/2+}}) is not a norm, since the minimum of two norms is not generally a norm, yet it is used in Lemma 6 and in (30) as though it were.","section":"Section 5.2"},{"comment":"The notation }K}_X is used for a functional applied to K, and the text acknowledges that X lacks the triangle inequality but proceeds to treat it as one; this requires clarification or a reformulation as a genuine norm.","section":"Section 5.2, Lemma 6"},{"comment":"The heading contains a typo: 'nonlinearlity' should be 'nonlinearity'.","section":"Section 5.3 heading"},{"comment":"The passage from the ℓ2-decoupling estimate to the claimed embedding X^{0+,1/2+} ↦ L^6_{t,x} is not fully justified; the proof cites [25, Lemma 2.9] but does not show how the frequency-dependent phase φ_n fits into that lemma.","section":"Lemma 3"},{"comment":"The term N_R is defined only by the list in (18); it would help readability to write out the full expression for N_R in a displayed equation.","section":"Section 4, equation (19)"}],"recommendation":"major_revision","confidential_remarks":"The contraction-map gap is likely fixable by redefining Γ2 to depend on both φ1 and φ2 through the substitution v = φ2 (or u = h(φ1) + φ2) and then proving the contraction estimate in the product norm. The authors should also provide a rigorous treatment of the time cutoff. Given the novelty of the approach and the large amount of plausible multilinear analysis, a major revision seems appropriate rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the claimed result is genuinely new: local well-posedness for periodic KdV for s > -2/3 without complete integrability, extending CKSTT's s = -1/2 by 1/6 derivative, via a data-dependent Bourgain space and modulation-restricted normal form adapted from NTT's mKdV work. Second, the proof as written has a real hole in the final contraction argument: Section 6 defines Gamma(phi1,phi2) = (Gamma1(phi1), Gamma2(phi1)), so Gamma2 depends only on phi1, while equation (19) contains v_n explicitly on the right-hand side. The paper says in Section 5.3 that the substitution u = h + v is needed and then ignores it \"in favor of increased readability.\" That is not a minor detail; the Banach fixed-point step in Theorem 1 depends on making this substitution or redefining the map so that the second coordinate actually appears. The stress-test note is correct to flag this.\n\nWhat the paper does well: the cancellation lemma (Lemma 4) is clean and the derivation of the data-dependent phase phi_n from the resonant term is natural, not fitted. Lemma 8's Case 3 quadrilinear reduction is the technical heart and the symbol estimates look plausible; the paper is honest about what is deferred. The citation pattern is appropriate: Oh's earlier work and NTT's mKdV paper are directly relevant, and self-citation there is not a problem.\n\nSoft spots, in proportion: the fixed-point gap is load-bearing. The time-cutoff modifications in Section 6 are asserted rather than shown, and some frequency cases in Lemma 8 are omitted as \"easier\" or \"omitted for brevity.\" Those are less concerning if the main gap closes. The small-data restriction is acceptable for this type of argument, and the failure to recover the conditional uniqueness range of NTT is not a flaw.\n\nWho this is for: anyone working on low-regularity well-posedness of dispersive equations. The method is a useful template if the Section 6 gap is patched. It deserves a serious referee, but the referee should be asked to focus on the contraction map.\n\nRecommendation: send to peer review, with the expectation of substantial revision. I would not cite the current version as a proven theorem, but I would read a revised version carefully.","headline":"A new low-regularity range for periodic KdV (s > -2/3) with a promising data-dependent normal form method, but the final fixed-point contraction in Section 6 has a real gap: Gamma2 is written as a function of phi1 only while (19) still contains v, and the paper explicitly waves off the required substitution.","tokens_in":25617,"tokens_out":3624,"would_cite":false,"duration_ms":34001,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q53","37L50","42B37"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the periodic Korteweg-de Vries equation is locally well-posed in $H^s(\\mathbb{T})$ for $-2/3 < s \\le -1/2$ and small data, without using complete integrability.","keywords":["periodic KdV equation","local well-posedness","modulation-restricted normal form","data-dependent X^{s,b}-type space","low-regularity well-posedness","normal form transformation","cancellation structure","small data"],"falsifier":"Compute the quadrilinear symbol bound claimed in Lemma 8, Case 3B(i), for frequencies $n_1 = N$, $n_3 = -N$, $N \\gg |n_2|, |n_4|$: if the support of the difference $\\chi(n_2,\\tau_2;N) - \\chi(n_2,\\tau_2;-N)$ does not force $\\langle \\tau_2 - n_2^3 - \\varphi_{n_2}\\rangle \\gtrsim N^2|n_2|$ when $\\varphi_{n_2}$ is the data-dependent phase, then the claimed $\\langle n_{\\min}\\rangle^{-1/2-}$ estimate fails and the contraction argument does not close.","tokens_in":24547,"feed_emoji":"🌊","tokens_out":15308,"duration_ms":123588,"temperature":0.7,"pith_summary":"This paper proves local well-posedness of the periodic Korteweg-de Vries equation for real-valued mean-zero initial data in $H^s(\\mathbb{T})$ whenever $-2/3 < s \\le -1/2$ and the $H^s$ norm is sufficiently small. Previously, without using complete integrability, the best result was $s \\ge -1/2$, where the standard bilinear $X^{s,b}$ estimate fails below that threshold. The proof builds a solution space $Y$ that depends on the initial datum and applies a modulation-restricted normal form to remove the resonant part of the nonlinearity, converting the difficult derivative loss into a time integral that can be controlled. If correct, this narrows the gap toward the integrable threshold $s \\ge -1$ by a method that does not rely on inverse scattering.","feed_headline":"KdV is well-posed below H^{-1/2} on the torus for small data","feed_subtitle":"A normal-form argument reaches s > -2/3 without integrability, beating the old s ≥ -1/2 barrier.","key_machinery":"The mechanism is a modulation-restricted normal form $T^{\\ell}(u,v) = T(\\chi u, \\chi v)$, where $\\chi$ is a smooth Fourier cutoff that keeps only those bilinear interactions whose modulation $\\langle \\tau - n^3 \\rangle$ is comparable to the resonant phase $|n(n_1+n_2)n_3|$; this removes the dangerous quadratic term while leaving bounded parts of the nonlinearity intact, preventing higher-order terms from acquiring unbounded modulation/frequency interactions. The argument runs in an initial-data-dependent $X^{s,b}$-type space $Y$ with weights $w_Y(n,L) = L^{1/2+}$ for low modulation and $L^{1/3+}\\langle n\\rangle^{1/3-}$ for high modulation, together with a smoother space $Z$, and uses the $L^6_{t,x}$ embedding supplied by $\\ell^2$-decoupling. A cancellation lemma converts the resonant combination $\\operatorname{Re}(r_n\\overline{w_n}) + \\tfrac{1}{2}|w_n|^2$ into an integral in time, so the derivative loss $2s-1$ is paid through the smoother $Z^*$ norm rather than through the failed bilinear estimate.","core_discovery":"The central claim is Theorem 1: for real-valued mean-zero data $u_0 \\in H^s(\\mathbb{T})$ with $-2/3 < s \\le -1/2$ and $\\|u_0\\|_{H^s}$ sufficiently small, there is a space $Y = Y(u_0)$ and a time $T = T(\\|u_0\\|_{H^s}) > 0$ such that $u_t + u_{xxx} = (u^2)_x$ has a unique solution $u \\in C^0_t([0,T], H^s_x(\\mathbb{T})) \\cap Y$, and the data-to-solution map is continuous. The theorem is obtained without using the inverse-scattering structure of KdV, extending the previous non-integrable threshold $s \\ge -1/2$ to $s > -2/3$. After rescaling the equation into $L^2$ by $u \\mapsto \\langle\\nabla\\rangle^{-s}u$, the proof writes $u = T^{\\ell}(u,u) + v$ and uses a cutoff that restricts the normal form to modulations comparable to the resonant phase, so only an acceptable part of the quadratic nonlinearity is removed. The resonant remainder is absorbed into a modified linear propagator $W_t$ with phase $\\varphi_n = \\frac{2}{3}\\langle n\\rangle^{2s} n^{-1}|f_n|^2$, and the remaining nonlinearity is controlled in a pair of spaces $Y$ and $Z$.","pith_inferences":["The same modulation-restricted normal form could be tried on other quadratic dispersive equations on the torus whose resonant phase has a comparable factorization, potentially pushing their well-posedness thresholds below the $X^{s,b}$ bilinear endpoint.","Because the space $Y$ depends on the initial datum through $\\varphi_n$, a natural test is whether a datum-independent version exists, or whether the small-data hypothesis can be removed by a different decomposition of the resonant dynamics.","The paper's own remarks in Sections 5.3 and 6 identify a concrete checkpoint: if the Littlewood-Paley reconstitution of $\\eta T^{\\ell}(\\eta u, \\eta u)$ cannot be closed, Theorem 1 would describe the time-cutoff equation but not the original KdV flow."],"forward_implications":["Local well-posedness for periodic KdV now holds for every $s > -2/3$ (small data), breaking the $s \\ge -1/2$ barrier without using integrability.","The data-to-solution map is continuous, though by the known non-uniform-continuity result it cannot be uniformly continuous in this range.","The small-data restriction is part of the theorem; the fixed-point argument does not claim large-data well-posedness.","Uniqueness and continuous dependence hold inside the data-dependent space $Y$, so the solution map is well-defined from a small ball in $H^s$ into $C^0_t H^s_x \\cap Y$."],"supporting_citations":[{"why":"Supplies the mKdV local-well-posedness theorem at $s>1/3$ whose restriction idea and cancellation structure this paper adapts to KdV.","marker":"[18]"},{"why":"Introduces the initial-data-dependent $X^{s,b}$-type space and the cancellation idea on which the KdV argument is built.","marker":"[24]"},{"why":"Provides the $\\ell^2$-decoupling estimate that restores the $L^6_{t,x}$ embedding for the modified phase $\\xi^3+\\varphi(\\xi)$.","marker":"[21]"},{"why":"Establishes the previous best non-integrable threshold $s\\ge -1/2$, the benchmark this paper extends.","marker":"[6]"},{"why":"Supplies the KdV normal-form operator and the resonant phase-shift formula used in the Section 4 decomposition.","marker":"[19]"},{"why":"Proves the bilinear estimate and the earlier $s>-1/2$ result whose failure below $s=-1/2$ motivates the new normal-form approach.","marker":"[14]"},{"why":"Settles ill-posedness below $H^{-1}$, placing the lower endpoint of the gap this paper narrows.","marker":"[17]"},{"why":"Shows the integrable threshold $s\\ge -1$ via inverse scattering, the target this paper approaches without integrability.","marker":"[11]"},{"why":"Creates the $X^{s,b}$-type Fourier restriction space that the paper's data-dependent spaces generalize.","marker":"[3, 4]"}],"fun_headline_variants":["KdV well-posed down to s > -2/3 on torus, no integrability","Periodic KdV local well-posedness for s > -2/3, beating -1/2","Normal form trick pushes KdV well-posedness to s > -2/3","Modulation restricted normal form improves KdV to s > -2/3","KdV on torus: local well-posedness for s > -2/3 without integrability"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the formal substitution $u = T^{\\ell}(u,u) + v$ can be inserted into the Duhamel nonlinearity after the time cutoff $\\eta(t)$ is introduced, with the cutoff-induced extra terms either vanishing or falling into the already-controlled Case 3 estimates; Section 5.3 explicitly says this substitution detail is ignored for readability, and Section 6 asserts that the cutoff complications reduce to the spatial constraint of Case 3 without a full derivation.","fun_headline_variants_meta":{"raw":{"variants":["KdV well-posed down to s > -2/3 on torus, no integrability","Periodic KdV local well-posedness for s > -2/3, beating -1/2","Normal form trick pushes KdV well-posedness to s > -2/3","Modulation restricted normal form improves KdV to s > -2/3","KdV on torus: local well-posedness for s > -2/3 without integrability"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001311,"raw_usage":{"total_tokens":5350,"prompt_tokens":961,"completion_tokens":4389,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":577,"completion_tokens_details":{"reasoning_tokens":4278}},"tokens_in":577,"tokens_out":4389,"duration_ms":25281,"temperature":1.0,"reasoning_tokens":4278,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T14:31:54.084168+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the quadrilinear symbol bound claimed in Lemma 8, Case 3B(i), for frequencies $n_1 = N$, $n_3 = -N$, $N \\gg |n_2|, |n_4|$: if the support of the difference $\\chi(n_2,\\tau_2;N) - \\chi(n_2,\\tau_2;-N)$ does not force $\\langle \\tau_2 - n_2^3 - \\varphi_{n_2}\\rangle \\gtrsim N^2|n_2|$ when $\\varphi_{n_2}$ is the data-dependent phase, then the claimed $\\langle n_{\\min}\\rangle^{-1/2-}$ estimate fails and the contraction argument does not close.","supporting_citations":[{"cited_title":"L ocal well-posedness in low regularity of the mKdV equa- tion with periodic boundary condition","cited_arxiv_id":null,"evidence_quote":"Supplies the mKdV local-well-posedness theorem at $s>1/3$ whose restriction idea and cancellation structure this paper adapts to KdV."},{"cited_title":"Well-posedness of t he Cauchy problem for the modified KdV equation with periodic boundary condition","cited_arxiv_id":null,"evidence_quote":"Introduces the initial-data-dependent $X^{s,b}$-type space and the cancellation idea on which the KdV argument is built."},{"cited_title":"On strichartz estimates from ℓ2-decoupling and applications","cited_arxiv_id":null,"evidence_quote":"Provides the $\\ell^2$-decoupling estimate that restores the $L^6_{t,x}$ embedding for the modified phase $\\xi^3+\\varphi(\\xi)$."},{"cited_title":"Colliander, M","cited_arxiv_id":null,"evidence_quote":"Establishes the previous best non-integrable threshold $s\\ge -1/2$, the benchmark this paper extends."},{"cited_title":"Resonant phase-shift and global smoothing of the periodic Korteweg-de Vries equation in low regu- larity settings","cited_arxiv_id":null,"evidence_quote":"Supplies the KdV normal-form operator and the resonant phase-shift formula used in the Section 4 decomposition."},{"cited_title":"Kenig, Gustavo Ponce, and Luis V ega","cited_arxiv_id":null,"evidence_quote":"Proves the bilinear estimate and the earlier $s>-1/2$ result whose failure below $s=-1/2$ motivates the new normal-form approach."},{"cited_title":"Sharp ill-posedness results for the KdV an d mKdV equations on the torus","cited_arxiv_id":null,"evidence_quote":"Settles ill-posedness below $H^{-1}$, placing the lower endpoint of the gap this paper narrows."},{"cited_title":"Kappeler and P","cited_arxiv_id":null,"evidence_quote":"Shows the integrable threshold $s\\ge -1$ via inverse scattering, the target this paper approaches without integrability."}],"review_version":1}