Expository article: “Bounded orthogonal systems and the Λ ⁡ ( p ) \Lambda(p) -set problem” by Jean Bourgain

Hongki Jung, Bartosz Langowski, Alexander Ortiz, Truong Vu · 2025

Read Expository article: “Bounded orthogonal systems and the Λ ⁡ ( p ) \Lambda(p) -set problem” by Jean Bourgain by Hongki Jung, Bartosz Langowski, Alexander Ortiz, Truong Vu (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-2504-18533.txt.

Introduction

Hongki Jung, Bartosz Langowski, Alexander Ortiz, and Truong Vu explain a landmark paper by Jean Bourgain on special sets in harmonic analysis, in work from 2025. Bourgain settled a long open conjecture of Rudin about whether certain sets, called Lambda(p) sets, must exist. The question sits inside the theory of bounded orthogonal systems. The exposition walks through his argument step by step, in a form meant for readers new to the proof.

Catalog record

CREATORHongki Jung, Bartosz Langowski, Alexander Ortiz, Truong Vu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE132 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-18533

Browse the full library · About the Tower of Babel

Full text

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

Expository article: “Bounded orthogonal systems and the Λ ⁡ ( p ) \Lambda(p) -set problem” by Jean Bourgain Hongki Jung, Bartosz Langowski, Alexander Ortiz, Truong Vu Abstract . In this paper, we present an exposition of the work [ 3 ] by Jean Bourgain, in which he resolved the well known conjecture posed by Rudin regarding the existence of Λ ⁡ ( p ) \Lambda(p) -sets. 1. Introduction and statement of the main result The purpose of this article is to present an exposition of the paper [ 3 ] of Jean Bourgain, where he proves the conjecture posed by Rudin about the existence of \Lambda(p) -sets which are not \Lambda(q) -sets for any q<p . The methods used in his proof are deeply insightful and have far-reaching implications in analysis and related fields. His argument involves sophisticated probabilistic techniques, metric entropy methods and chaining arguments. We begin with a short introduction to \Lambda(p) -sets. Definition 1.1 . For 0<p<\infty and an integer d\geq 1 we say that S\subset\mathbb{Z}^{d} is a \Lambda(p) -set (equivalently: has the \Lambda(p) -property ) if there exists q<p such that (1.1) \displaystyle\|\sum_{j\in S}a_{j}e^{2\pi ij\cdot\theta}\|_{L^{p}(\mathbb{T}^{d})}\lesssim\|\sum_{j\in S}a_{j}e^{2\pi ij\cdot\theta}\|_{L^{q}(\mathbb{T}^{d})} for all finitely supported multiparameter sequences (a_{j})_{j\in\mathbb{Z}^{d}} . It is not difficult to see, with the aid of Hölder’s inequality, that the \Lambda(p) -property from Definition 1.1 depends only on p and not on q . In particular, if p>2 , then one can show that ( 1.1 ) is equivalent to (1.2) \displaystyle\|\sum_{j\in S}a_{j}e^{2\pi ij\cdot\theta}\|_{L^{p}(\mathbb{T}^{d})}\lesssim\|\sum_{j\in\mathbb{Z}^{d}}a_{j}e^{2\pi ij\cdot\theta}\|_{L^{2}(\mathbb{T}^{d})}. Note that due to Hölder’s inequality the L^{p}(\mathbb{T}^{d}) -norm dominates the L^{2}(\mathbb{T}^{d}) -norm if p>2 , so ( 1.2 ) can be thought of as a strengthening of Bessel’s inequality \displaystyle\|\sum_{j\in S}a_{j}e^{2\pi ij\cdot\theta}\|_{L^{2}(\mathbb{T})}\leq\|\sum_{j\in\mathbb{Z}}a_{j}e^{2\pi ij\cdot\theta}\|_{L^{2}(\mathbb{T})}. The simplest example of a \Lambda(p) -set is the set of dyadic numbers \{2^{n}\}_{n=0}^{\infty} which has the \Lambda(p) -property for all 0<p<\infty . More generally, any lacunary sequence (b_{n})_{n\in\mathbb{Z}}\subset\mathbb{Z} , i.e., a sequence satisfying for some \varepsilon>0 the growth condition \frac{b_{n+1}}{b_{n}}>1+\varepsilon,\qquad n\in\mathbb{Z}, forms a \Lambda(p) -set for any 0<p<\infty . In [ 10 ] Rudin showed the \Lambda(p) -property for p>1 for Sidon sets, that is subsets E of integers for which an estimate \displaystyle\sum_{j\in E}|\widehat{f}(j)|\lesssim\|f\|_{L^{\infty}(\mathbb{T})} holds for any bounded function f on the torus. Every lacunary set is a Sidon set. Another array of examples of \Lambda(p) -sets comes from discrete restriction theory. There has been substantial progress in this area due to the recent development of \ell^{2} decoupling by Bourgain and Demeter—see [ 4 ] . In particular, decoupling establishes that the set of integer-lattice points on an elliptic paraboloid \{(\mathbf{x},|\mathbf{x}|^{2}):\mathbf{x}\in([1,N]\cap\mathbb{Z})^{d-1}\} is a \Lambda(p) -set for 2\leq p\leq\frac{2(d+1)}{d-1} up to an N^{\varepsilon} –loss for any \varepsilon>0 . Decoupling for the moment curve, for example as proved by Bourgain, Demeter, and Guth [ 5 ] , allows one to show that the integer-lattice points on the moment curve \{(n,n^{2},\dotsc,n^{d}):n\in[1,N]\cap\mathbb{Z}\} form a \Lambda(p) -set for 2\leq p\leq d(d+1) up to N^{\varepsilon} –loss. The ranges of p in both these theorems are sharp. It is conjectured that a discrete sphere \{\mathbf{x}\in\mathbb{Z}^{d}:|\mathbf{x}|=N\} is a \Lambda(p) -set for 2\leq p\leq\frac{2d}{d-2} . The decoupling techniques only yield partial progress on this problem. The difficulty in handling the case of a discrete sphere lies in its arithmetic features, which seem to be beyond the reach of purely analytic methods. As for negative examples, one can test ( 1.1 ) with the Dirichlet kernel to see that the set of all integers is not a \Lambda(p) -set for any 0<p<\infty . More involved arguments are needed to show that the set of squares of integers is not a \Lambda(p) -set for p<2 or p\geq 4 . A famous conjecture of Rudin asserts that squares have the \Lambda(p) -property for p\in(2,4) . Sets with the \Lambda(p) -property have some interesting structural properties. It was shown by Rudin [ 10 ] that for p>2 a \Lambda(p) -set of size n cannot contain an arithmetic progression of length larger than Cn^{2/p} ; in particular a Sidon set (which, as we already mentioned, belongs to \bigcap_{p>1}\Lambda(p) ) of size n can have at most C\log n elements in an arithmetic progression. For more details in that direction and a nice summary of the \Lambda(p) -problem and its connections with Rudin’s conjecture we encourage the reader to visit the blog of Ioannis Parissis [ 9 ] . It is straightforward to see that if 0<p_{1}<p_{2} then every \Lambda(p_{2}) -set is also a \Lambda(p_{1}) -set, which we will write shortly as \Lambda(p_{2})\subset\Lambda(p_{1}) . That raises a natural question whether this inclusion is necessarily proper. This problem was first posed in the seminal paper of Rudin [ 10 ] . In the case p\in(1,2) the negative answer was provided by Bachelis and Ebenstein in [ 1 ] . More precisely, they showed that for any S\subset\mathbb{Z} the set \{p\in(1,2):S\in\Lambda(p)\} is an open interval. The situation turned out to be much more difficult in the case p>2 . It had been a long standing open problem until Bourgain showed in his celebrated paper [ 3 ] that the inclusion is indeed proper. Theorem 1.2 . For any p>2 there exists S\subset\mathbb{Z} which is a \Lambda(p) -set, but is not a \Lambda(r) -set for any r>p . Bourgain proved the above result in the more general setting of 1 -bounded orthogonal systems. As we will see momentarily, Theorem 1.2 can be derived from the following result. Theorem 1.3 . For n\in\mathbb{N} let \Phi=(\varphi_{j})_{j=1}^{n} be a sequence of mutually orthogonal real-valued functions on a probability space satisfying a uniform bound \|\varphi_{j}\|_{L^{\infty}}\leq 1 for each j\in[n]:=\{1,\dots,n\} . For any p>2 there exists a subset of [n] such that |S|\simeq n^{2/p} and satisfying for any sequence of real coefficients (a_{i})_{i=1}^{n} the estimate (1.3) \displaystyle\|\sum_{i\in S}a_{i}\varphi_{i}\|_{L^{p}}\lesssim\big(\sum_{i\in S}|a_{i}|^{2}\big)^{1/2}, with the implicit constant depending only on p . Bourgain’s main result in [ 3 ] , Theorem 1.2 , though stated and proved for real-valued functions, applies to the complex exponential system (e^{2\pi ik\theta})_{k=1}^{n} by taking real and imaginary parts. See Remark 5.10 of [ 7 ] for additional examples of complex exponential systems in the context of Fourier restriction to which Theorem 1.2 applies. We also refer the reader to the paper [ 11 ] of Talagrand, where Theorem 1.2 is proved in a more general setting. Unlike Bourgain’s proof, Talagrand’s approach avoids using the special properties of the function x\mapsto|x|^{p} appearing in the definition of the L^{p} norm. As a result, his argument applies to a broader class of Banach spaces beyond just L^{p} spaces. Remark 1.4 . The size of S in the above theorem is maximal, in the sense that if ( 1.3 ) holds for all (a_{i})_{i=1}^{n} then |S|\lesssim n^{2/p} . To see that this is indeed the case, take S\subset[n] and consider f(\theta)=\sum_{k\in S}e^{2\pi ik\theta} . Letting D_{n}(\theta)=\sum_{|k|\leq n}e^{2\pi ik\theta} be the classicial Dirichlet kernel we can use Hölder’s inequality to write \displaystyle|S|=f(0)=f\ast D_{n}(0)\leq\|f\|_{p}\|D_{n}\|_{p^{\prime}}. Using the well known estimate \|D_{n}\|_{p^{\prime}}\lesssim n^{1/p} and ( 1.3 ) we get |S|\lesssim|S|^{1/2}n^{1/p}, which is equivalent to |S|\lesssim n^{2/p}. The following result, shows that the proof of Theorem 1.2 reduces to proving Theorem 1.3 . Proposition 1.5 . Theorem 1.3 implies Theorem 1.2 . Proof. For each k\in\mathbb{N} consider the system \Phi_{k}=\{e^{2\pi ij\theta}:2^{k}\leq j<2^{k+1}\} on the 1-torus \mathbb{T}

Read the complete text