New Curved Kakeya Estimates

Arian Nadjimzadah · 2025

Read New Curved Kakeya Estimates by Arian Nadjimzadah (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-15760.txt.

Introduction

Arian Nadjimzadah, writing alone in 2025, proves new size estimates for curved Kakeya sets in three dimensions. A Kakeya set contains a unit segment in every direction; the curved version replaces straight segments with curves from a family of phase functions. How small such a set can be is a famous open problem.

For an open class of translation-invariant phase functions, the author shows the Hausdorff dimension is at least about 2.348, beating the previous two-and-a-third bound. The proof combines Wolff's hairbrush argument with a new estimate for curve families satisfying two stability conditions he calls coniness and twistiness.

Catalog record

CREATORArian Nadjimzadah
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE142 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-15760

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

New Curved Kakeya Estimates Thanks: The author was supported in part by NSF grant DMS-2347850. Arian Nadjimzadah Abstract . We prove that for an open class of translation-invariant Hörmander-type phase functions, the corresponding curved Kakeya sets in ℝ 3 \mathbb{R}^{3} have Hausdorff dimension at least ( 13 − 13 ) / 4 = 2.348 ​ … (13-\sqrt{13})/4=2.348\ldots . Previous methods, such as polynomial partitioning, have attained at best dimension 2 + 1 / 3 2+1/3 for curved Kakeya sets in ℝ 3 \mathbb{R}^{3} . The proof combines Wolff’s classical hairbrush argument with a new Kakeya estimate for 3-parameter families of curves satisfying stable geometric conditions that we call coniness and twistiness. The latter estimate extends ideas of Katz, Wu, and Zahl from the study of SL 2 \mathrm{SL}_{2} -Kakeya sets. 1. Introduction 1.1. Curved Kakeya in \mathbb{R}^{3} beyond the 2+1/3 threshold As first discovered by Bourgain [ 4 ] , curved Kakeya problems arise naturally from the wavepacket geometry of variable-coefficient oscillatory integral operators. This connection was further developed by Wisewell in [ 23 ] , and we refer the reader there for details. Let \phi:\mathbb{R}^{3}\times\mathbb{R}^{2}\to\mathbb{R} be a smooth phase function satisfying Hörmander’s non-degeneracy conditions [ 12 ] , and write \mathbf{x}=(x,t)\in\mathbb{R}^{2}\times\mathbb{R} . For \epsilon_{\phi}>0 sufficiently small and (\xi,v)\in B_{\epsilon_{\phi}}^{2}\times B_{\epsilon_{\phi}}^{2} , the associated \phi -curve is (1.1) \displaystyle\ell_{\xi,v}:=\{\mathbf{x}\in B_{\epsilon_{\phi}}^{3}:\nabla_{\xi}\phi(\mathbf{x},\xi)=v\}. We write \mathcal{C}(\phi) for this 4-parameter family and call \xi the direction parameter. In the translation-invariant case (1.2) \displaystyle\phi(x,t,\xi)=\langle x,\xi\rangle+\psi(t,\xi), Hörmander’s non-degeneracy conditions are equivalent to \partial_{t}\nabla_{\xi}^{2}\psi(t,\xi)\neq 0 , and the curves have the explicit parameterization (1.3) \displaystyle\ell_{\xi,v}(t)=(v-\nabla_{\xi}\psi(t,\xi),t). By a change of variables, we may assume that a translation-invariant phase function \phi(\mathbf{x},\xi)=\langle x,\xi\rangle+\psi(t,\xi) satisfies one of the two alternatives: (1.4) \displaystyle\partial_{t}\nabla_{\xi}^{2}\psi(0,0) \displaystyle=I,\ (positive-definite) (1.5) \displaystyle\partial_{t}\nabla_{\xi}^{2}\psi(0,0) \displaystyle=I_{-}:=\begin{pmatrix}1&0\\ 0&-1\end{pmatrix}. (non-positive-definite) A compact set K\subset\mathbb{R}^{3} is a \phi - Kakeya set if, for every \xi\in B_{\epsilon_{\phi}}^{2} , K contains a curve \ell_{\xi,v}\in\mathcal{C}(\phi) . Let d_{\mathrm{set}}(\phi) denote the infimum of the Hausdorff dimensions of such sets. A first step to understanding the L^{p} estimates for Hörmander-type oscillatory integral operators is to better understand the behavior of \phi -Kakeya sets. The straight-line Kakeya problem and the restriction problem in \mathbb{R}^{3} are obtained from the phase \phi_{\mathrm{parab}}(\mathbf{x},\xi)=\langle x,\xi\rangle+t|\xi|^{2} . Recently Wang and Zahl proved d_{\mathrm{set}}(\phi_{\mathrm{parab}})=3 , resolving the Kakeya conjecture in \mathbb{R}^{3} [ 22 ] . Much less is known in the curved setting, though. There are certain \phi so that \phi -Kakeya sets can lie in a surface, giving the worst possible bound d_{\mathrm{set}}(\phi)=2 , and for generic \phi , small portions of \delta -discretized \phi -Kakeya sets can compress near a surface [ 4 ] . There has been a long line of work in quantifying how poorly \phi -Kakeya sets can behave, in order to better understand the corresponding oscillatory integral operator, including but certainly not limited to [ 15 , 19 , 10 , 9 , 5 , 6 , 7 ] . In all cases, excluding those \phi for which \mathcal{C}(\phi) is simply diffeomorphic to a family of lines, the best unconditional estimates available were d_{\mathrm{set}}(\phi)\geq 2+1/3 . For the very special class of phase functions satisfying Bourgain’s condition , the threshold 2+1/3 can be obtained by a polynomial partitioning argument [ 11 , 6 ] . See [ 16 ] for an example of a (non-translation-invariant) phase function satisfying Bourgain’s condition whose curves are not diffeomorphic to a family of lines. Earlier, Sogge attained Minkowski dimension 2+1/3 for certain families of curves by using geometric methods [ 19 ] . Theorem 1 surpasses the 2+1/3 threshold for curved Kakeya sets in \mathbb{R}^{3} . Theorem 1 (Main Theorem on \phi -curved Kakeya) . Let \phi(x,t,\xi)=\langle x,\xi\rangle+\psi(t,\xi) be a translation-invariant non-positive-definite phase function. Define (1.6) \displaystyle B \displaystyle=\tfrac{1}{2}\partial_{t}^{2}\nabla_{\xi}^{2}\psi(0,0), (1.7) \displaystyle C \displaystyle=\tfrac{1}{6}\partial_{t}^{3}\nabla_{\xi}^{2}\psi(0,0). Suppose that the matrices B and C satisfy (1.8) \displaystyle|B_{12}(B_{22}-B_{11})+C_{12}|>\frac{1}{2}|(B_{22}+B_{11})(B_{22}-B_{11})+C_{11}+C_{22}|. Then (1.9) \displaystyle d_{\rm{set}}(\phi)\geq\frac{13-\sqrt{13}}{4}\geq 2.348. In fact the following stronger discretized statement holds. For all \epsilon>0 there is an M=M(\epsilon) and \delta_{0}=\delta_{0}(\epsilon) so that the following holds for all \delta\in(0,\delta_{0}] . Let L\subset C(\phi) be a \delta -direction separated collection of curves. Let E\subset\mathbb{R}^{3} be a union of \delta -cubes and suppose that |\ell\cap E|\geq\lambda for each \ell\in L . Then (1.10) \displaystyle|E|\geq\delta^{\epsilon}\lambda^{M}\delta^{\frac{a+1}{2}}(\delta^{2}\#L)^{3/4}, where a=(\sqrt{13}-3)/2\approx 0.303 is the positive root of a^{2}+3a-1 . The hypothesis on \phi in Theorem 1 is stable under translation-invariant perturbations. A simple example of a phase function satisfying the hypothesis is (1.11) \displaystyle\phi(x,t,\xi)=\langle x,\xi\rangle+\frac{1}{2}\langle\left(tI_{-}+t^{2}\begin{pmatrix}0&1\\ 1&1\end{pmatrix}\right)\xi,\xi\rangle. The proof of Theorem 1 combines Wolff’s hairbrush argument [ 24 ] with a new stable curved Kakeya estimate for 3-parameter families of curves (Theorem 2 ) in order to estimate the hairbrushes of curves. We will see that ( 1.8 ) is a natural condition that guarantees the hairbrushes of curves satisfy the conditions of Theorem 2 : coniness and twistiness . 1.2. The 3-parameter curved Kakeya estimate The main new input is an incidence estimate for general 3-parameter families of curves. Let (1.12) \displaystyle X:B_{2}^{3}\times B_{2}^{1}\to\mathbb{R}^{2},\qquad\ell_{\mathbf{p}}(t):=(X(\mathbf{p},t),t), and write \mathcal{C}(X)=\{\ell_{\mathbf{p}}:\mathbf{p}\in B_{2}^{3}\} . We defer the formal definitions used in the theorem to Section 2 , but give their geometric idea here. We say that \mathcal{C}(X) is coney (formally Condition 2 ) if the directions of the 1-parameter family of curves in \mathcal{C}(X) through a point are not contained in a plane. We now describe the twistiness condition (formally Condition 3 ). The curves intersecting a fixed curve in \mathcal{C}(X) (at a relatively small angle) fill out a certain twisting grain . The grain direction can be projected out to give a plane curve. Applying the same procedure for each fixed curve in \mathcal{C}(X) gives a 3 -parameter family of plane curves. We say that \mathcal{C}(X) is twisty if this family of curves forbids 2nd-order tangency in the sense of [ 25 ] . Twistiness is motivated by the structure of \mathrm{SL_{2}} -Kakeya sets in [ 13 ] . Theorem 2 (Coney-Twisty 3-parameter Kakeya) . Suppose that X is \mathfrak{c} -coney, \tau -twisty, and satisfies the basic conditions. For all \epsilon>0 there is M=M(\epsilon) and \delta_{0}=\delta_{0}(\epsilon,C_{0},\ldots,C_{O_{\epsilon}(1)}) so that the following holds for all \delta\in(0,\delta_{0}] . Suppose that L\subset\mathcal{C}(X) satisfies the two-dimensional ball condition (1.13) \displaystyle\#(L\cap B)\leq(r/\delta)^{2}\text{ for all $r\in[\delta,1]$, and all balls $B$ of radius $r$.} Let E\subset\mathbb{R}^{3} be a union of \delta -cubes, and suppose |\ell\cap E|\geq\lambda for each \ell\in L . Then (1.14) \displaystyle|E|\geq\delta^{\epsilon}\mathfrak{c}^{2}(\tau\lambda)^{M}\delta^{a}(\delta^{2}\#L), where a=(\sqrt{13}-3)/2\approx 0.303 is the positive root of a^{2}+3a-1 . Theorem 2 has implications to restricted projections, which we record in Section 2 . The new feature of Theorem 2 is that its hypotheses are stable under small perturbations of the family of curves. The proof uses the techniques in [ 13 ] when the typical angle between curves are small, and the curved multilinear Kakeya theorem [ 2 , Theorem 6] when the typical angle is large. Below we give a sketch of the proof of Theorem 1 by using Theorem 2 . 1.3. Sketch of the proof of Theorem 1 We give a sketch of the proof of Theorem 1 by tracking the example \phi_{A}(\mathbf{x},\xi)=\langle x,\xi\rangle+\tfrac{1}{2}\langle A(t)\xi,\xi\rangle , where \det A^{\prime}(0)\neq 0 . There are two main ingredients. (1) Wolff’s classical hairbrush argument lets us extract a hairbrush configuration H\subset L to study. (2) Theorem 2 lets us estimate the union of

Read the complete text