On a Sharp Fourier Extension Inequality on the Circle with Lacunary Spectrum
Felipe Gonçalves, João Paulo Ferreira · 2025
Read On a Sharp Fourier Extension Inequality on the Circle with Lacunary Spectrum by Felipe Gonçalves, João Paulo Ferreira (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-2510-20934.txt.
Introduction
Felipe Gonçalves and João Paulo Ferreira prove a sharp Fourier extension inequality. An inequality is a statement that one amount bounds another. Their proof concerns the circle and an exponent named Tomas-Stein. It applies to functions whose spectrum is lacunary, meaning values spread out fast. More precisely, each next value is more than three times the last. A sharp result means the bound cannot be improved in general. The work sits in Fourier restriction theory, a branch of harmonic analysis.
Catalog record
| CREATOR | Felipe Gonçalves, João Paulo Ferreira |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 34 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-20934
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
On a Sharp Fourier Extension Inequality on the Circle with Lacunary Spectrum Felipe Gonçalves, João Paulo Ferreira Abstract . We prove a sharp Fourier extension inequality on the circle for the Tomas-Stein exponent for functions whose spectrum { ± λ n } \{\pm\lambda_{n}\} satisfies λ n + 1 > 3 λ n \lambda_{n+1}>3\lambda_{n} . 1. Introduction Fourier restriction theory has been established as one of the most relevant areas in modern harmonic analysis. The celebrated Tomas-Stein theorem lies at the heart of Fourier restriction theory; in the spherical case, its dual version states that (1) ||\widehat{f\sigma}||_{L^{\frac{2(d+1)}{d-1}}(\mathbb{R}^{d})}\leq C||f||_{L^{2}(\mathbb{S}^{d-1})}, where \sigma denotes the embedded surface measure on \mathbb{S}^{d-1} and \widehat{f\sigma}(x)=\int_{\mathbb{S}^{d-1}}f(w)e^{-ix\cdot w}d\sigma(w). Questions concerning the best constant in ( 1 ) are quite more recent and have been proven to be surprisingly challenging. Foschi, in his celebrated paper [ 11 ] , proved that when d=3 constant functions are extremizers for ( 1 ) and all extremizers are of form ce^{ix_{0}\cdot w} where c is a constant. Dimension d=3 is the only dimension where the best constant is known for the sphere, although other related sharp extension inequalities have been investigated (see [ 3 ] , [ 5 ] , [ 6 ] , [ 7 ] , [ 8 ] , [ 10 ] , [ 13 ] ). Sharp Tomas-Stein inequalities for the sphere and other conic sections can be tackled when the exponent is even, using the natural convolution structure that emerges from the problem. It is conjectured that constant functions are the only extremizers (modulo symmetries) for ( 1 ) for every d\geq 2 . For d=2 , in [ 4 ] , the conjecture was proven to be locally true, in the sense that constant functions are locally extremizers. Later in [ 12 ] , Gonçalves and Negro proved that constant functions are local extremizers of ( 1 ) for 2\leq d\leq 60 , which adds an evidence for the conjecture. The case d=2 is the last remaining even exponent coming from extension problems associated to conic sections (see [ 14 ] and [ 15 ] ), hence it naturally attracts a lot of attention. We let C_{\text{opt}}:=\sup_{\stackrel{{\scriptstyle f\in L^{2}(\mathbb{S}^{1})}}{{f\neq 0}}}\dfrac{||\widehat{f\sigma}||_{L^{6}(\mathbb{R}^{2})}^{6}}{||f||_{L^{2}(\mathbb{S}^{1})}^{6}}. As far as the L^{2}\to L^{6} estimate in dimension d=2 is concerned, much partial progress has been made; it has been long established in [ 17 ] that extremizers do exist. Moreover, they are even, can be taken as non-negative and smooth. We let J_{n} denote the Bessel function of first kind, defined as 2\pi(-i)^{n}z^{n}J_{n}(|z|)=|z|^{n}\widehat{(z^{n}\sigma)}(z) , where z=x+iy with x,y\in\mathbb{R} . Conjecture 1 (Case d=2 ) . For all f\in L^{2}(\mathbb{S}^{1})\backslash\{0\} , (2) \dfrac{||\widehat{f\sigma}||^{6}_{L^{6}(\mathbb{R}^{2})}}{||f||^{6}_{L^{2}(\mathbb{S}^{1})}}\leq C_{\text{opt}}=\dfrac{||\widehat{\sigma}||^{6}_{L^{6}(\mathbb{R}^{2})}}{||\boldmath{\boldsymbol{1}}||^{6}_{L^{2}(\mathbb{S}^{1})}}=(2\pi)^{4}\left(\int_{0}^{\infty}J_{0}^{6}(r)rdr\right). Ciccone and Gonçalves in [ 9 ] proved Conjecture 1 for the subspace of functions whose spectrum is under arithmetic constraints (see Definition 2 below for a precise definition), meaning that as long as the set \text{spec}(f):=\{k\in\mathbb{Z};\,\widehat{f}(k)\neq 0\} is sufficiently sparse, the conjecture holds. The arithmetic constraints that they impose on \text{spec}(f) is a generalizations of a Sidon set . Classically, a Sidon set A\subset\mathbb{Z} is a set for which each element of A+A has unique representation modulo permutation. Ciccone and Gonçalves in [ 9 ] impose that if A=\text{spec}(f) then each element x\in A+A+A either has unique representation modulo permutation or is trivially represented as x=x+a-a for some a\in A\cap(-A) . It turns out that the special case \text{spec}(f)=\{\pm\lambda_{n}\} , where \{\lambda_{n}\} is a lacunary sequence with \lambda_{n+1}/\lambda_{n}>q and q\geq 5 falls into these conditions. However, the spectrum such as \{0\}\cup\{\pm 5^{n}\}_{n\geq 0} or \{0\}\cup\{\pm 2^{n}\}_{n\geq 0} are not contemplated by their result. A natural question is: Can we prove Conjecture 1 for \text{spec}(f)\subset\{0\}\cup\{\pm q^{n}\}_{n\geq 0} for small q>1 ? More generally, can we prove Conjecture 1 when \text{spec}(f) is lacunary with small lacunary constant? Theorem 1 . Let f\in L^{2}(\mathbb{S}^{1}) be such that \text{spec}(f)\subset A_{\lambda,q}:=\{\pm\lambda_{n};n\geq 0\} , where \{\lambda_{n}\} is a lacunary sequence of non-negative integers satisfying \lambda_{n+1}/\lambda_{n}>3 , and \lambda_{0}:=0 . Then (3) ||\widehat{f\sigma}||^{6}_{L^{6}(\mathbb{R}^{2})}\leq(2\pi)^{4}\left(\int_{0}^{\infty}J_{0}^{6}(r)rdr\right)||f||^{6}_{L^{2}(\mathbb{S}^{1})}, and equality is attained if and only if \widehat{f}(k)=0 for every k\neq 0 . 1.1. Motivation, Notation and Strategy The novelty in the proof of Theorem 1 relies on the fact that, although lacunary sequences are not always in the conditions of [ 9 ] , as long as the lacunary constant is large enough, the points that fail to satisfy these conditions can be classified, allowing a precise decomposition of the norm ||\widehat{f\sigma}||_{L^{6}(\mathbb{R}^{2})}^{6} (see Lemma 4 ). Recall that a sequence \{\lambda_{n}\} of non-negative integers is a lacunary sequence if \lambda_{n+1}/\lambda_{n}>q for some q>1 . In this case, we say that \{\lambda_{n}\} has lacunary constant q . To verify inequality ( 2 ) for all L^{2} functions, one could prove ( 2 ) for all functions whose spectrum is lacunary. Similarly, one could prove ( 2 ) for all functions whose spectrum is finite, i.e., band-limited functions. The band-limited case has been considered in [ 16 ] , where they prove that if \text{spec}(f)\subset[-30,30] then inequality ( 2 ) holds, and later improved in [ 1 ] for functions with \text{spec}(f)\subset[-120,120] . To prove Theorem 1 , we use an explicit formula for ||\widehat{f\sigma}||_{L^{6}(\mathbb{R}^{2})}^{6} . Letting A=\text{spec}(f) we have (4) \begin{array}[]{rl}(2\pi)^{-7}||\widehat{f\sigma}||_{L^{6}(\mathbb{R}^{2})}^{6}&=\displaystyle{\sum_{D\in A^{3}}\sum_{\begin{subarray}{c}n_{1},\dots,n_{6}\in A\\ n_{1}+n_{2}+n_{3}=D\\ n_{4}+n_{5}+n_{6}=D\end{subarray}}\widehat{f}(n_{1})\widehat{f}(n_{2})\widehat{f}(n_{3})\overline{\widehat{f}(n_{4})}\overline{\widehat{f}(n_{5})}\overline{\widehat{f}(n_{6})}I(n_{1},\dots,n_{6})},\end{array} where throughout this paper we convention the notation A^{h}:=\{n_{1}+\dots+n_{h};\,n_{1},\dots,n_{h}\in A\} , and I(n_{1},\dots,n_{6}) denotes the integral I(n_{1},\dots,n_{6})=\int_{0}^{\infty}J_{n_{1}}(r)J_{n_{2}}(r)J_{n_{3}}(r)J_{n_{4}}(r)J_{n_{5}}(r)J_{n_{6}}(r)rdr. As it turns out, a direct application of Plancherel shows that ||f||_{L^{2}(\mathbb{S}^{1})}^{6} can be expanded in a way that resembles ( 4 ): (5) (2\pi)^{-3}||f||_{L^{2}(\mathbb{S}^{1})}^{6}=\displaystyle{\sum_{D\in A^{3}}\sum_{\begin{subarray}{c}n_{1},n_{2},n_{3}\in A\\ n_{1}+n_{2}+n_{3}=D\end{subarray}}|\widehat{f}(n_{1})|^{2}|\widehat{f}(n_{2})|^{2}|\widehat{f}(n_{3})|^{2}}. If a good decomposition of the set A^{3} is available, a direct comparison between ( 4 ) and ( 5 ) is possible, which is precisely the idea behind [ 9 ] . They prove that if A is a P (3)-set, among all L^{2} functions f with \text{spec}(f)\subset A , the ratio ||\widehat{f\sigma}||_{L^{6}(\mathbb{R}^{2})}/||f||_{L^{2}(\mathbb{S}^{1})} is maximized at constant functions. For the sake of completeness, here we recall the definition of a P (3)-set. Definition 2 . We say that a set A\subset\mathbb{Z} is a P (3)-set if for every D\in A^{3} , one and only one of the following holds: (i) There exists a unique triple (modulo permutation) \{n_{1},n_{2},n_{3}\}\subset A such that D=n_{1}+n_{2}+n_{3} ; (ii) D has trivial representation, meaning D=D+m-m , where m\in A\cap(-A) . Given A\subset\mathbb{Z} , we say that D\in A^{3} satisfies the \text{P}(3) property (or is a \text{P}(3) point) if D verifies the definition above, possibly even in the case where A is not a \text{P}(3) -set itself. Otherwise, we call D an exception. If D\in A^{3} is a \text{P}(3) point we say that D is unique if it satisfies (i) or trivial if it satisfies (ii). 1.2. Further Difficulties One of the main difficulties inherent to the problem of lacunary spectrum with lacunary constant q<3 is the fact that the symmetric set A=\{\pm\lambda_{n}\} will no longer be an \text{P}(2) -set. Definition 3 . A set A\subset\mathbb{Z} is said to be an \text{P}(2) -set if for every D\in A^{2} , one and only one of the following holds: (i) Either there exists a unique pair (modulo permutation) \{n_{1},n_{2}\}\subset A such that D=n_{1}+n_{2} ; (ii) Or D=0 , which means D=m-m for every m\in A\cap(-A) . As in the \text{P}(3) property, we also say that a point D\in A^{2} satisfying