Bourgain-type projection theorems over finite fields

Alex Rose · 2025

Read Bourgain-type projection theorems over finite fields by Alex Rose (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-2511-08757.txt.

Introduction

Alex Rose proves finite-field versions of Bourgain's projection theorem, a result about how sets look when projected onto lines and planes. Projection theorems sit at the meeting point of geometry and combinatorics, and moving them to finite fields changes the tools available and the statements that can hold.

The new results live in higher dimensions and cover a range of parameters where earlier bounds were weaker. In particular, Rose improves the exceptional set estimate due to Chen in every dimension and codimension within that range. Exceptional sets collect the directions where a projection behaves worse than expected, so shrinking their possible size strengthens the whole theorem. The paper is a technical contribution to finite-field incidence geometry, written for specialists in harmonic analysis and additive combinatorics.

Catalog record

CREATORAlex Rose
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-08757

Browse the full library · About the Tower of Babel

Full text

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

Bourgain-type projection theorems over finite fields Alex Rose Abstract . We prove finite-field analogs of Bourgain’s projection theorem in higher dimensions. In particular, for a certain range of parameters we improve on an exceptional set estimate by Chen in all dimensions and codimensions. 1. Introduction 1.1. Over real numbers Fractal dimensions of orthogonal projections over real numbers were extensively studied. For example, the following bounds on the dimension of the exceptional set were obtained in works by Kaufman [ 7 ] , Mattila [ 9 ] , Falconer [ 4 ] , Peres and Schlag [ 12 ] . Theorem 1.1 . Let K\subseteq\mathbb{R}^{n} be a Borel set. The following bounds hold: • If \dim K\leq m and 0<t\leq\dim K , then \dim\{W\in Gr(n,m)\mid\dim\pi_{W}(K)\leq t\}\leq m(n-m)-(m-t). • If \dim K>m , then \dim\{W\in Gr(n,m)\mid\mathcal{H}^{m}(\pi_{W}(K))=0\}\leq m(n-m)-(\dim K-m). • If \dim K>2m , then \dim\{W\in Gr(n,m)\mid Int(\pi_{W}(K))=\varnothing\}\leq m(n-m)-(\dim K-2m). Here Gr(n,m) denotes the set of m -dimensional subspaces of the n -dimensional space, \pi_{W} denotes the orthogonal projection onto W and Int denotes the interior of a set. In particular, in the plane the first bound above yields that \dim\{\theta\in S^{1}\mid\dim\pi_{\theta}(K)<t\}\leq t. It was conjectured by Oberlin [ 10 ] that for \dim K\leq 1 and t\in(\frac{1}{2}\dim K,\dim K) the sharp bound for the planar case is actually 2t-\dim K . Bourgain, in his groundbreaking paper [ 1 ] established the t=\dim K/2 case of this conjecture. Theorem 1.2 . For any \kappa\in(0,1) and \alpha\in(0,2) there exists \varepsilon=\varepsilon(\kappa,\alpha)>0 such that for any K with \dim K=\alpha the following holds: \dim\{\theta\in S^{1}\mid\dim\pi_{\theta}(K)<\frac{\alpha}{2}+\varepsilon\}<\kappa. Recently, the conjecture was proven, first for almost Ahlfors-David regular sets by Orponen and Shmerkin [ 11 ] , and then for all Borel sets by Ren and Wang [ 13 ] . In higher dimensions, He [ 6 ] obtained analogs of Bourgain’s projection theorem for all codimensions. Theorem 1.3 . For any \kappa\in(0,1) and \alpha\in(0,n) there exists \varepsilon=\varepsilon(\kappa,\alpha)>0 such that for any analytic set K\subseteq\mathbb{R}^{n} with \dim K=\alpha the following holds: the set \{W\in Gr(n,m)\mid\dim\pi_{W}(K)<\frac{m}{n}\alpha+\varepsilon\} does not support any non-zero measure \mu on Gr(n,m) with the following non-concentration property: \forall\rho>0,\forall V\in Gr(n,n-m):\mu(\{W\mid\angle(W,V)\leq\rho\})\leq\rho^{\kappa}. 1.2. Over finite fields In this paper we are concerned with finite-field analogs of the abovementioned results. We follow the notation of Chen [ 3 ] and define projections in terms of the kernel. Definition 1.4 . Let \mathbb{F} be a field. For a linear subspace W\subseteq\mathbb{F}^{n} we define the projection map by \pi^{W}(x):=x+W\in\mathbb{F}^{n}/W . In the case of real numbers, \pi^{W} coincides with \pi_{W^{\bot}} considered above. Chen in [ 3 ] initiated the study of exceptional sets for projections over finite fields. He obtained the following bounds: Theorem 1.5 (Corollary 1.3 in [ 3 ] ) . Let K\subseteq\mathbb{F}_{p}^{n} with |K|=p^{s} . Then • If s\leq m and t\in(0,s] , then |\{W\in Gr(n,n-m)\mid|\pi^{W}(K)|\leq p^{t}/10\}|\leq\frac{1}{2}p^{m(n-m)-(m-t)} • If s>m , then |\{W\in Gr(n,n-m)\mid|\pi^{W}(K)|\leq p^{m}/10\}|\leq\frac{1}{2}p^{m(n-m)-(s-m)} • If s>2m , then |\{W\in Gr(n,n-m)\mid|\pi^{W}(K)|\neq p^{m}\}|\leq 4p^{m(n-m)-(s-2m)} Here and throughout the paper \mathbb{F}_{p} denotes the field with p elements and p is assumed to be a prime number. Chen also conjectured an analog of the sharp projection theorem in \mathbb{R}^{2} . We slightly reformulate his conjecture below: Conjecture 1.6 . Let K\subseteq\mathbb{F}_{p}^{2} and E\subseteq Gr(2,1) . Assume that 0\leq|K|\leq p^{1-\delta} for some fixed \delta>0 and 2\leq|E|\leq|K| . Then there exists W\in E with |\pi^{W}(K)|\gtrsim_{\delta}|K|^{\frac{1}{2}}|E|^{\frac{1}{2}}. Bright and Gan in [ 2 ] showed that this lower bound can be attained. They also improved the bound on the size of the exceptional set when |K| is large (for arbitrary n ). Lund, Pham and Vinh in a recent paper [ 8 ] proved the following bound on the size of the projections in the plane, improving the results of Chen. Theorem 1.7 . Let K\subseteq\mathbb{F}_{p}^{2} and E\subseteq Gr(2,1) . Assume that |K|^{\frac{1}{2}}\leq|E|\leq|K|\leq p . Then there exists W\in E such that |\pi^{W}(K)|\gtrsim\max\{|K|^{\frac{1}{2}}|E|^{\frac{1}{6}},|K|^{\frac{2}{5}}|E|^{\frac{2}{5}},|E|\}. 1.3. Main results The first result of this paper proves an analog of the bound in Theorem 1.7 for one-dimensional projections in higher dimensions. We obtain the following Bourgain-type estimates: Theorem 1.8 . Consider n\geq 2 . Let K\subseteq\mathbb{F}_{p}^{n} and let E\subseteq Gr(n,n-1) be a set of hyperplanes that do not contain a common line (i.e. \bigcap_{W\in E}W=0 ). Assume that |K||E|^{\frac{2n+1}{4(n-1)}}\leq p^{n} . Then there exists W\in E such that |\pi^{W}(K)|\gtrsim\min\{|K|^{\frac{1}{n}}|E|^{\frac{1}{4n(n-1)}},|K|^{\frac{1}{n-1}}\}. Note that if the set E has cardinality larger than O(p^{n-2}) , then the condition that the elements of E do not contain a common line is automatically satisfied. In this case the lower bound |K|^{\frac{1}{n}}|E|^{\frac{1}{4n(n-1)}} is an improvement on the lower bound \gtrsim p^{2-n}|E| provided by the first estimate in Chen’s theorem when |E|\ll|K|^{\frac{4(n-1)}{4n(n-1)-1}}p^{\frac{4n(n-1)(n-2)}{4n(n-1)-1}} . To extend the theorem to other codimensions we will need the following definition. Definition 1.9 . A set E\subseteq Gr(n,m) is called non-degenerate if for every V\in Gr(n,n-m) there exists W\in E such that W\cap V=0 . We give several examples to illustrate the non-degeneracy condition. • Let v_{1},\ldots,v_{n} be a basis and let E\subseteq Gr(n,m) be a set containing the subspace \text{span}\{v_{i_{1}},\ldots,v_{i_{m}}\} for any m-subset \{i_{1},\ldots,i_{m}\}\subseteq\{1,\ldots,n\} . Then E is non-degenerate. This is because given any V\in Gr(n,n-m) one can construct the desired W\in E by starting with W=0 and then adding the elements of the basis to W one by one while keeping the intersection W\cap V trivial. • Take k sufficiently large depending only on n and let W_{1},\ldots,W_{k} be independent uniformly distributed elements of Gr(n,m) . Then with probability 1-o(1) the set \{W_{1},\ldots,W_{k}\} (and any set containing it) is non-degenerate. Indeed, a calculation shows that for any V\in Gr(n,n-m) the probability that it non-trivially intersects W_{1} is \mathbb{P}(W_{1}\cap V\neq 0)=O(p^{-1}) (see Proposition 6.2 ). Hence, the probability that W_{i}\cap V\neq 0 for all i is O(p^{-k}) , and if k is chosen larger than m(n-m) , then the probability that such V exists is at most O(p^{-k}|Gr(n,n-m)|)=o(1) . • Assume the set E satisfies the analog of the non-concentration assumption in Theorem 1.3 , that is, for any V\in Gr(n,n-m) we have |\{W\in E\mid W\cap V\neq 0\}|\leq p^{-\kappa}|E| . Then the set E is non-degenerate (trivially). Now we can state the analog of Theorem 1.8 for projections on higher-dimensional subspaces. Theorem 1.10 . Let n\geq 2 and let 0<\varepsilon<\varepsilon_{0}=\frac{1}{4n(n-1)(2n)^{n-2}} . Let K\subseteq\mathbb{F}_{p}^{n} and E\subseteq Gr(n,n-m) be a non-degenerate set. Assume that |K||E|^{\frac{2n+1}{4(n-1)n^{n-2}}}\leq p^{n} and |E|^{\frac{1}{4n^{n-2}}}\leq|K| . Then there exists a divisor d\neq 1 of n that depends only on K and if d does not divide m , then \exists W\in E\text{ such that }|\pi^{W}(K)|\gtrsim|K|^{\frac{m}{n}}|E|^{\varepsilon}. In particular, the estimate above holds if n and m are coprime. We believe that the methods of [ 6 ] can be extended over finite fields. In light of this the main advantages of Theorem 1.10 over (the finite field version of) Theorem 1.3 are: • The non-degeneracy assumption is much weaker than the non-concentration assumption. In fact, it suffices to produce finitely many elements of the set E that together form a non-degenerate set, while the remaining elements of E can be completely arbitrary. • The dependence of the size of the projection on the cardinality of E is explicit as a power of |E| . • Explicit value of \varepsilon_{0}=\varepsilon_{0}(n) . In Chen’s theorem (Theorem 1.5 ) the factor of p^{m(n-m)-m} in

Read the complete text