{"id":"6e564ddf-c9de-44e7-9a86-93a6e4ce6b12","arxiv_id":"2411.17147","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Critical local well-posedness for NLS with non-algebraic nonlinearity |u|^a u is proved on tori for a > 4/d, using a new function space and bilinear estimate.","lead":"This paper proves local well-posedness for nonlinear Schrödinger equations on tori at critical regularity for a wide range of small nonlinearities, including new cases in high dimensions. It introduces a new function space and bilinear estimate that overcome obstacles in earlier techniques, advancing the theory of periodic dispersive equations.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.10's summation step loses control of R^θ; without a time-frequency assumption on A, the bilinear estimate (3.26) is unjustified.","rationale":"The reader's weakest_assumption correctly points to the bilinear estimate (3.25) and the delicate interpolation underlying Lemma 3.9. My stress-test agrees that the bilinear layer is the most fragile part of the proof, but I locate the concrete gap one step later, in Lemma 3.10's summation over R. The displayed replacement of R^θ||A_R|| by ||A||_{B^{1/r−1/q}B^θ} is not justified by the definitions in §2.2 unless A possesses additional time-frequency structure. This is load-bearing because Lemma 3.10 is used to prove Proposition 3.11, which in turn yields the key nonlinear estimates (4.8) and (4.9) and the uniqueness/continuous-dependence arguments. The issue is not an external-consensus conflict but an internal consistency question: either the summation step needs a missing argument, or the statement of Lemma 3.10 must be narrowed to functions whose time Besov regularity at spatial frequency R is R^θ-compatible. Because the argument may be repairable with an additional structural hypothesis, I recommend a conditional verdict rather than outright rejection: acceptance should be contingent on a rigorous justification of the R^θ summation, ideally with a precise statement of the Besov product notation used.","tokens_in":40504,"tokens_out":57990,"duration_ms":449102,"concrete_test":"Verify the summation inequality in Lemma 3.10 with the explicit family A_R(t,x)=a_R(x)h(t), where h has Fourier support in [−1,1] and a_R is a fixed spatial-frequency-R wave packet (e.g., a_R=e^{iR·x} localized at frequency R), and u a free evolution at frequency N≫R. Compute the scaling in R of both (3.28) and the claimed norm ||A||_{B^{1/r0−1/q0}_{r0,r0}B^θ_{r0,∞}}. If the ratio grows like R^θ, the step fails for general A. If the authors intended A to satisfy a Schrödinger-type time-frequency relation, check whether the Besov norm as defined in §2.2 encodes that relation; if not, the lemma requires an explicit additional assumption.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The proof of Lemma 3.10 contains an unjustified summation step after (3.28): it replaces ∑_R R^θ ||A_R||_{B^{1/r0−1/q0}_{r0,r0}L^{r0}} by ||A||_{B^{1/r0−1/q0}_{r0,r0}B^θ_{r0,∞}}. For a general A in this Besov space, the R^θ factor is not controlled. For example, take A_R(t,x)=a_R(x)h(t) with a_R at spatial frequency R and h(t) at time frequency 1. Then the left-hand side after (3.28) is roughly R^{θ+1/r0−1/q0}||a_R||_{L^{r0}}||h||, while the claimed right-hand side is R^{1/r0−1/q0}||a_R||_{L^{r0}}||h||, a discrepancy of R^θ. The proof gives no additional time-frequency localization for A (e.g., time frequencies of A_R at scale R^2) that would absorb this factor. Since Lemma 3.10 feeds directly into Proposition 3.11 and hence into the nonlinear estimates (4.8)–(4.9) and the bootstrap in Theorem 1.1, the central well-posedness claim is not fully supported as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves local well-posedness for the nonlinear Schr\\\"odinger equation |u|^a u on T^d at the scaling-critical regularity s=d/2-2/a under the assumptions a>4/d and s<1+a (Theorem 1.1), together with a Lipschitz well-posedness result for a larger range of a (Theorem 1.2) and a sharp failure of H\\\"older continuity of the solution map for small a (Theorem 1.3). The method introduces a new function space Z^s, a bilinear Strichartz estimate (Lemma 3.9), and a package of Besov-space embeddings, which are then combined with Bony linearizations and a weak-limit construction. I checked the summation step in Lemma 3.10 that was flagged as questionable: the R^\\theta factor is part of the spatial Besov norm B^\\theta_{r_0,\\infty}, and the dyadic sum in R converges because the kernel (\\langle N/R\\rangle^{-\\sigma_1}+R^{-2\\sigma_1}) is summable in R for each fixed N. Thus the specific objection about a missing time-frequency assumption on A does not land. My remaining concerns are about technical points in the function-space framework rather than about the main bilinear summation.","tokens_in":40748,"tokens_out":35037,"duration_ms":331463,"significance":"If correct, Theorem 1.1 is a substantial advance: it treats non-algebraic nonlinearities with small a and includes new energy-critical cases in d\\ge5, a regime where previous atomic-space methods required algebraic nonlinearities or a\\ge2. The construction of Z^s, which is not based on conventional atomic spaces and shrinks on short time intervals, is an interesting contribution in its own right. The negative result Theorem 1.3 also gives a useful almost-sharp limitation on the regularity of the solution map. The paper is detailed and many of the intermediate estimates are proven step by step, including the new bilinear estimate and the embedding properties of Z^s. The manuscript would benefit from a careful revision addressing the Besov-space technicalities and the definitional identity Z^s=\\ell^2_s Z^0.","major_comments":[{"comment":"The proof uses Besov spaces B^{1/r_0-1/q_0}_{r_0,r_0} and B^\\theta_{r_0,\\infty} with r_0=(2+\\sigma_3+\\sigma_2)/(d+2), which is smaller than 1 for d\\ge2. However, the interpolation and product rules stated in Section 2, in particular Proposition 2.2 and Corollary 2.4 (2.13), are formulated only for p,q\\in[1,\\infty]. The step (3.31) explicitly invokes (2.13) with a target Besov exponent r_0<1, which is outside the stated hypotheses. Since (3.31) is the bridge from the bilinear estimate (3.26) to the nonlinear estimate (3.29) used in Proposition 3.11 and hence in the proof of Theorem 1.1, this is a load-bearing technical point. Please provide a proof or a precise reference for the quasi-Banach Besov product and interpolation rules needed here, or adjust the exponents so that r_0\\ge1.","section":"Section 3.3, Lemma 3.9 and Proposition 3.11"},{"comment":"The identity Z^s=\\ell^2_s Z^0 is asserted in (3.11) and used to justify the retarded estimate K^+: (Z^{-s})'\\to Z^s and the bootstrap in (4.14), but it is not immediate from Definition 3.4. In the definition, the maxima over q and \\alpha sit outside the \\ell^2_s sums, the first term is weighted by s-\\sigma, while Z^0 already contains an \\ell^2 summation in N with weight -\\sigma. As written, Definition 3.4 and the standard meaning of \\ell^2_s Z^0 are not visibly equivalent. Please clarify the convention: either define Z^s as \\ell^2_s Z^0 and prove equivalence with (3.10), or prove directly that the two norms are comparable. This is needed because the linear estimates and duality arguments in Section 4 rely on (3.11).","section":"Section 3.2, Definition 3.4 and Eq. (3.11)"},{"comment":"The interpolation step from (3.21) and (3.22) to (3.23) is described in a single sentence and does not specify the interpolation functor or the precise way in which the three parameters (1/q,1/r,\\rho) run over the open tetrahedron. Since the endpoints have different q and r, and since one of the spaces may be quasi-Banach when r<1, this interpolation deserves a full proof or a precise reference. This is particularly important because Lemma 3.8 feeds directly into Lemma 3.9, the key bilinear estimate.","section":"Section 3.3, Lemma 3.8"}],"minor_comments":[{"comment":"In the passage from (3.28) to the following display, the authors replace R^\\theta\\|A_R\\|_{B^{s}_{r,r}L^r} by \\|A\\|_{B^{s}_{r,r}B^\\theta_{r,\\infty}} without comment. I verified this is valid, since R^\\theta\\|P_R A\\|_{L^r} is controlled by the B^\\theta_{r,\\infty} norm and the dyadic sum in R converges by the decay factors in (3.25). Nevertheless, adding the one-line justification would remove the appearance that the R^\\theta factor is dropped.","section":"Section 3.3, after Eq. (3.28)"},{"comment":"The notation B^s_{p,q}E is introduced for q\\in[1,\\infty], but Lemma 3.9 and Proposition 3.11 use exponents r_0<1. Please state explicitly that quasi-Banach Besov spaces are allowed in those instances, or modify the notational convention in Section 2.","section":"Section 2.1"},{"comment":"There is a typo in the abstract: \"equa tion\" should be \"equation\".","section":"Abstract"},{"comment":"The displayed uncovered band in Remark 1.5 is missing the word \"and\" between the two alternatives; it should read \"a < ... or a > ...\".","section":"Remark 1.5"}],"recommendation":"major_revision","confidential_remarks":"The paper is a serious and potentially important contribution. The main mathematical idea is sound and the new bilinear estimate appears to be a genuine advance. However, the manuscript currently relies on quasi-Banach Besov spaces and on a duality identity for Z^s that are not adequately documented. These are not merely cosmetic issues: they occur at the point where the nonlinear estimate and the bootstrap are derived. I would be willing to accept a revised version that supplies the missing technical foundations, but I cannot recommend acceptance in the present state."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my honest read. The paper has a genuinely new idea: a Z^s space and a bilinear Strichartz estimate with a decay factor that could push the local well-posedness range from Lee's a>=2 down to a>4/d. The main theorems are new, the proofs are detailed, and the negative Holder result in Theorem 1.3 looks like a nice complement. I found no issues with the early machinery (Lemmas 3.2, 3.3, etc.). But there is a real gap in Lemma 3.10, and it is load-bearing.\n\nThe problem is in the summation step right after (3.28). The inequality (3.25) comes with a factor R^theta on the A_R term. When the authors sum over R, they replace the sum over R of R^theta ||A_R||_{B^{1/r-1/q}_{r,r}L^r} by the single Besov norm ||A||_{B^{1/r-1/q}_{r,r} B^theta_{r,infty}}, which does not control the R^theta factor. A concrete example: take A_R(t,x)=a_R(x)h(t) with a_R at spatial frequency R and h(t) at time frequency 1. Then the norm on the right is R^{1/r-1/q}||a_R||||h|| while the left sum has an additional R^theta. The proof gives no time-frequency localization of A that would absorb this. Since Lemma 3.10 feeds directly into Proposition 3.11 and the nonlinear estimates (4.8)-(4.9), the central well-posedness claim is not fully supported as written.\n\nIs this repairable? Possibly. The A that actually appears in the application comes from the nonlinearity via Bony linearization, and those A's may carry extra structure (e.g., time frequencies tied to spatial frequencies). But the lemma is stated for arbitrary A, and the proof does not use such structure. So the gap is real as written.\n\nMy advice: send it to a serious referee, and instruct the referee to focus on Lemma 3.10. The paper deserves referee time because the idea is important and the rest of the architecture may well be correct. But I would not cite it until the gap is fixed.","headline":"Strong ideas and genuinely new results, but Lemma 3.10's summation step leaves the central bilinear estimate unjustified.","tokens_in":41379,"tokens_out":4312,"would_cite":false,"duration_ms":38258,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q55","35A01","35B65"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves local well-posedness of the critical nonlinear Schrödinger equation on tori for every power $a > 4/d$, a range that includes previously open energy-critical cases in dimensions $d \\ge 5$.","keywords":["nonlinear Schrödinger equation","critical regularity","torus","local well-posedness","non-algebraic nonlinearity","bilinear Strichartz estimate","Z^s function space","Hölder ill-posedness"],"falsifier":"Take $u = e^{it\\Delta}\\delta_N$, $v = e^{it\\Delta}\\delta_R$, and $A_R = \\chi_{[0,1]}\\delta_R$ in dimension $d = 5$ with $N/R$ ranging over powers of two. If the optimal constant in inequality (3.25) decays only like $(\\log(N/R))^{-C}$ instead of a positive power of $N/R$, then the frequency summation in Corollary 4.3 would diverge and the theorem's stated exponent range could not hold.","tokens_in":52,"feed_emoji":"🌀","tokens_out":8344,"duration_ms":133987,"temperature":0.7,"pith_summary":"This paper establishes local well-posedness for the nonlinear Schrödinger equation $i u_t + \\Delta u = \\pm |u|^a u$ on the torus $\\mathbb{T}^d$ at its critical Sobolev regularity, for every $a > 4/d$ with $s < 1 + a$, where $s = d/2 - 2/a$. Previously, non-algebraic nonlinearities (where $a$ is not an even integer, in particular small $a < 2$) were only treatable by contraction methods requiring $a \\ge 2$; the paper removes that barrier. The advance rests on a new bilinear Strichartz estimate with a positive frequency-decay factor and a newly designed function space $Z^s$ whose norms shrink on short time intervals, allowing large-data bootstrap arguments. As corollaries, the energy-critical cases in dimensions $d \\ge 5$ fall into the well-posed range, the flow is Lipschitz for $a > \\max\\{4/d, 1\\}$ with $s < a$, and for $0 < a < 1$ the solution map fails to be $\\alpha$-Hölder for every $\\alpha > a$.","feed_headline":"New proof widens NLS well-posedness on tori to low powers","feed_subtitle":"For a>4/d, critical Sobolev solutions exist uniquely, covering new energy-critical cases in d≥5.","key_machinery":"The engine of the paper is a bilinear estimate, Lemma 3.9 (inequality (3.25)), controlling products of high-frequency and low-frequency pieces of solutions: for $N \\ge 32 R$, the spacetime pairing of $\\psi^2 u_N \\overline{v} A_R$ is bounded by $\\|u\\|_{Z^0}\\|v\\|_{Z^0}(\\langle N/R\\rangle^{-\\sigma_1} + R^{-2\\sigma_1}) R^{\\theta} \\|A_R\\|_{B^{1/r_0 - 1/q_0}_{r_0,r_0} L^{r_0}}$ with $\\sigma_1 > 0$. The positive decay factor in $N/R$ is what makes the frequency summation converge when the nonlinearity has small power $a < 2$. To carry this estimate, the paper introduces a function space $Z^s$ on $\\mathbb{R} \\times \\mathbb{T}^d$ whose norm measures, after localizing in time and in dyadic spatial frequency $N$, the $L^q L^r$ size and a time-Besov size of Galilean-shifted frequency cubes; unlike atomic spaces $U^p, V^p$, the $Z^s$ norm of a free evolution restricted to a short interval tends to zero as the interval shrinks (property (3.13)), which is what makes large-data bootstrap bounds possible. Together with the Bony linearization (4.1), which decomposes the nonlinearity into $u_N$ times a derivative factor, the bilinear estimate yields the nonlinear estimates (4.8) and (4.9).","core_discovery":"The central claim is that the Cauchy problem for $|u|^a u$ on $\\mathbb{T}^d$ is locally well-posed in $H^s(\\mathbb{T}^d)$ at the scaling-critical regularity $s = d/2 - 2/a$ whenever $a > 4/d$ and $s < 1 + a$. This range includes all mass-supercritical exponents and, for $d \\ge 5$, the energy-critical case $s = 1$, which was previously unresolved on pure tori. The proof does not rely on a contraction mapping when $a$ is small, because the paper simultaneously proves that the solution map is not Lipschitz there; instead it constructs a priori bounds in a new space $Z^s$, extracts a weak limit, establishes uniqueness by a difference estimate in lower regularity, and proves continuous dependence via a weighted high-frequency decay estimate. The same package yields Lipschitz well-posedness for $a > \\max\\{4/d, 1\\}$, $s < a$, and a Hölder-ill-posedness theorem showing the failure of $\\alpha$-Hölder continuity for $0 < a < 1$ and $\\alpha > a$.","pith_inferences":["Beyond the paper's claims, the same bilinear-decay mechanism could apply to other periodic dispersive equations whose Strichartz estimates lose derivatives, such as higher-order Schrödinger or wave equations on tori.","A natural testable extension is to replace the single power $|u|^a u$ by a nonlinearity with several powers; the $Z^s$ machinery suggests the smallest power controls well-posedness and the regularity of the solution map.","The boundary $a = 1$ is a plausible next target: the paper leaves Lipschitz continuity inconclusive in dimension $6$ at $a = 1$, and the Hölder exponent $\\alpha > a$ suggests interpolation might determine the optimal modulus exactly there."],"forward_implications":["For every dimension $d \\ge 5$, the energy-critical case $s = 1$ now lies in the local well-posedness range, which was previously open on pure tori.","The proof uses no number-theoretic property of $\\mathbb{Z}^d$, so the same argument works on irrational tori $\\mathbb{R}^d/(\\theta_1 \\mathbb{Z} \\times \\cdots \\times \\theta_d \\mathbb{Z})$.","The solution map is locally Lipschitz for $a > \\max\\{4/d, 1\\}$ with $s < a$, and this is nearly sharp: for $0 < a < 1$, no $\\alpha$-Hölder modulus with $\\alpha > a$ is possible.","In dimensions $d \\le 7$, the condition $s < 1 + a$ is automatic, so Theorem 1.1 covers every mass-supercritical power $a > 4/d$.","The new $Z^s$ space provides an alternative to atomic $U^p, V^p$ spaces for critical dispersive problems, with the key property that time localization shrinks the norm."],"supporting_citations":[{"why":"Supplies the kernel estimate and the $L^4_{t,x}$ Strichartz bound on $\\mathbb{T}$, the base linear estimate feeding the time-Besov Strichartz lemmas.","marker":"[6]"},{"why":"Gives the $\\ell^2$ decoupling Strichartz estimate with arbitrarily small loss, sharpened to scale-invariant form on rational tori and used for Proposition 2.12.","marker":"[7]"},{"why":"Provides the earlier bilinear estimate (1.5) with decay factor and the atomic $X^s, Y^s$ framework that the new $Z^s$ space replaces.","marker":"[10]"},{"why":"Establishes the scale-invariant Strichartz estimate on general tori and the prior critical well-posedness range that Theorem 1.1 extends.","marker":"[15]"},{"why":"Introduces the Bony linearization for non-algebraic $|u|^a u$ on $\\mathbb{T}^3$; its restriction $a \\ge 2$ is the barrier Theorem 1.1 removes.","marker":"[18]"},{"why":"Sets the energy-critical well-posedness framework on partially periodic domains that the new $d \\ge 5$ torus result extends.","marker":"[11]"}],"fun_headline_variants":["NLS on tori: critical well-posedness down to a>4/d","Torus NLS: new local well-posedness for a>4/d","Critical NLS on tori covers mass-supercritical and energy-critical","Bilinear estimate unlocks NLS well-posedness on tori","New function space proves critical NLS well-posed on tori"],"cache_read_input_tokens":43392,"weakest_assumption_plain":"The proof's load-bearing premise is that when a high-frequency wave interacts with a low-frequency one, the interaction shrinks by a positive power of the frequency ratio; without that shrinkage, the sums over frequencies in the proof would diverge logarithmically and the construction of solutions would fail.","fun_headline_variants_meta":{"raw":{"variants":["NLS on tori: critical well-posedness down to a>4/d","Torus NLS: new local well-posedness for a>4/d","Critical NLS on tori covers mass-supercritical and energy-critical","Bilinear estimate unlocks NLS well-posedness on tori","New function space proves critical NLS well-posed on tori"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001033,"raw_usage":{"total_tokens":4347,"prompt_tokens":936,"completion_tokens":3411,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":552,"completion_tokens_details":{"reasoning_tokens":3312}},"tokens_in":552,"tokens_out":3411,"duration_ms":25158,"temperature":1.0,"reasoning_tokens":3312,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T12:28:20.536124+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take $u = e^{it\\Delta}\\delta_N$, $v = e^{it\\Delta}\\delta_R$, and $A_R = \\chi_{[0,1]}\\delta_R$ in dimension $d = 5$ with $N/R$ ranging over powers of two. If the optimal constant in inequality (3.25) decays only like $(\\log(N/R))^{-C}$ instead of a positive power of $N/R$, then the frequency summation in Corollary 4.3 would diverge and the theorem's stated exponent range could not hold.","supporting_citations":[{"cited_title":"G lobal well-posedness of the energy- critical nonlinear Schrödinger equation with small initia l data in H 1(T3)","cited_arxiv_id":null,"evidence_quote":"Provides the earlier bilinear estimate (1.5) with decay factor and the atomic $X^s, Y^s$ framework that the new $Z^s$ space replaces."},{"cited_title":"S trichartz estimates for partially pe- riodic solutions to Schrödinger equations in 4d and applications","cited_arxiv_id":null,"evidence_quote":"Sets the energy-critical well-posedness framework on partially periodic domains that the new $d \\ge 5$ torus result extends."}],"review_version":1}