Sharp sign uncertainty for trigonometric polynomials

Tolibjon Ismoilov · 2026

Read Sharp sign uncertainty for trigonometric polynomials by Tolibjon Ismoilov (2026) 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-2606-02299.txt.

Introduction

Ismoilov studies sign uncertainty for trigonometric polynomials of fixed degree. A symmetric Borel measure on the unit circle is given. Polynomials have prescribed degree N in the study. The question is the smallest radius of the last sign change. That radius is sought for polynomials with non-positive integral against the measure.

Sharp results are determined for each such measure. Extensions go to polar measures on higher dimensional spheres. The extremal problem reduces to the one dimensional case there. Reduction runs through the polar part of the measure. A polynomial analogue on the interval zero to one is also established. Orthogonal polynomials on the real line supply that analogue. Uncertainty principles gain sharp forms in this setting. Harmonic analysts are the main audience.

Catalog record

CREATORTolibjon Ismoilov
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE82 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.02299

Browse the full library · About the Tower of Babel

Full text

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

Sharp sign uncertainty for trigonometric polynomials Tolibjon Ismoilov Abstract . We study sign uncertainty principles for trigonometric polynomials of prescribed degree N N with respect to a symmetric Borel measure μ \mu on the unit circle ℝ / ℤ \mathbb{R}/\mathbb{Z} . For each such measure, we determine the smallest radius of the last sign change for trigonometric polynomials with non-positive μ \mu -integral. We further extend these results to polar measures on higher-dimensional spheres 𝕊 d \mathbb{S}^{d} , showing that the extremal problem reduces to the one-dimensional case via the polar part of the measure, and we establish a polynomial analogue on [ 0 , 1 ] [0,1] using orthogonal polynomials on the real line. 2010 Mathematics Subject Classification 42A16, 41A55, 42C05, 42A05, 46E22 1. Introduction 1.1. Background The Fourier uncertainty principles in mathematics describe the inability to control a function and its Fourier transform simultaneously. A classical instance is the Heisenberg inequality, which asserts that the mass of a function f\in L^{2}(\mathbb{R}^{d}) and that of its Fourier transform \widehat{f} cannot be localized near the origin simultaneously; see [ 10 ] . Within the framework of Fourier uncertainty, there is a concept known as sign uncertainty . This refers to the idea that, under certain conditions, it is impossible for both a function and its Fourier transform to have negative mass concentrated near the origin. Such a phenomenon was first observed by Bourgain, Clozel, and Kahane [ 4 ] in 2010. To describe it concretely, we recall some terminology. Throughout this paper, we will use the following convention for defining eventually non-negative functions. Let \mathbb{X} be either the euclidean space \mathbb{R}^{d} , the sphere \mathbb{S}^{d} , the unit circle \mathbb{R}/\mathbb{Z} or the unit interval [0,1] equipped with its standard metric \mathbf{\mathsf{d}}(\cdot,\cdot) . Let x^{\ast}\in\mathbb{X} be a distinguished point to serve as the origin or the north pole. We say a function f:\mathbb{X}\to\mathbb{R} is eventually non-negative if there exists a radius r>0 such that f(x)\geq 0 for all x\in\mathbb{X}\setminus B_{r}(x^{\ast}) , where B_{r}(x^{\ast})\coloneqq\bigl\{y\in\mathbb{X}\;:\;\mathbf{\mathsf{d}}(x^{\ast},y)<r\bigr\} . The smallest such radius is called the radius of last sign change and denoted by (1.1) \operatorname{r}(f)\coloneqq\inf\bigl\{r>0\;:\;f(x)\geq 0\text{ for all }x\in\mathbb{X}\setminus B_{r}(x^{\ast})\bigr\}. Recall that for a function f\in L^{1}(\mathbb{R}^{d}) we define its Fourier transform as \widehat{f}({\bm{\xi}})\coloneqq\int_{\mathbb{R}^{d}}e^{-2\pi i{\bm{\xi}}\cdot{\bm{x}}}f({\bm{x}})\,\text{\rm d}{\bm{x}}. In [ 4 ] , the authors proved that if f:\mathbb{R}^{d}\to\mathbb{R} is integrable, with \widehat{f} real-valued and integrable, satisfies f(0)\leq 0 , \widehat{f}(0)\leq 0 , and both f and \widehat{f} are eventually non-negative, then the product of their radii of last sign change is bounded from below by a positive absolute constant: (1.2) \operatorname{r}(f)\cdot\operatorname{r}(\widehat{f})\geq\mathbb{A}(d)>0. Finding the sharp value of \mathbb{A}(d) in ( 1.2 ) is a notoriously difficult problem. The only dimension in which the exact sharp constant is known is d=12 , where Cohn and Gonçalves [ 8 ] proved \mathbb{A}(12)=\sqrt{2} , exploiting the machinery of Viazovska on modular forms [ 24 ] . The sign uncertainty phenomenon is not confined to the Euclidean setting. Gonçalves, Oliveira e Silva, and Ramos [ 12 ] vastly generalized the framework of Bourgain–Clozel–Kahane by establishing sign uncertainty principles for a broad collection of operators and function spaces. Their results cover Fourier and Dini series on the circle, the Hilbert transform, the discrete Fourier and Hankel transforms, spherical harmonics, and Jacobi polynomials, among others. Related quantitative aspects of the Euclidean sign uncertainty principle were further studied in [ 7 , 11 , 13 ] . A central observation motivating the present work is that the underlying measure plays a decisive role in determining whether sign uncertainty can occur at all. This was made precise in the recent work of Carneiro, Ramos, and the author in [ 5 ] , where they studied sign uncertainty for entire functions of exponential type with a competing condition expressed as integration with respect to an even Borel measure \nu on \mathbb{R} . They established that certain choices of \nu can eliminate the sign uncertainty phenomenon. Conversely, for a broad class of measures, they established sharp sign uncertainty bounds by bridging the problem and the theory of de Branges spaces of entire functions, where the reproducing kernel structure allows one to identify the sharp constants and classify the extremizers explicitly. The present article takes up the sign uncertainty phenomenon in the framework of [ 5 ] to the compact setting of the unit circle \mathbb{R}/\mathbb{Z} , where entire functions of exponential type are replaced by trigonometric polynomials of a prescribed degree, and the competing condition is given by a symmetric finite Borel measure \mu on \mathbb{R}/\mathbb{Z} . Our main results provide a complete characterization of the sign uncertainty phenomenon for each measure and determine the sharp constants in the cases where sign uncertainty holds, complementing the body of work described above. 1.2. An extremal problem for trigonometric polynomials We identify the circle \mathbb{S}^{1} and the quotient space \mathbb{R}/\mathbb{Z} in such a way that the distinguished point x^{\ast}\in\mathbb{S}^{1} corresponds to 0\in\mathbb{R}/\mathbb{Z} . Let \mu be a finite Borel measure on \mathbb{R}/\mathbb{Z} , which satisfies the symmetry condition (1.3) \mu(A)=\mu(-A),\text{ for all }A\subseteq\mathbb{R}/\mathbb{Z}. We consider the following problem: (P1) For each fixed N\geq 1 , find the exact value of \rho(\mu,N)\coloneqq\inf_{0\not\equiv f\in\mathcal{T}(\mu,N)}\operatorname{r}(f), where the infimum is taken over the following family of trigonometric polynomials \mathcal{T}(\mu,N)\coloneqq\left\{f(x)=\sum_{n=-N}^{N}\widehat{f}(n)\,e^{2\pi inx}\;\left\lvert\;\begin{aligned} &f:\mathbb{R}/\mathbb{Z}\to\mathbb{R}\text{ is eventually non-negative}\\ &\text{ and }\int_{\mathbb{R}/\mathbb{Z}}f(x)\,\text{\rm d}\mu(x)\leq 0\end{aligned}\right.\right\}. Note that the metrics on the circle \mathbb{S}^{1} and on \mathbb{R}/\mathbb{Z} differ by a factor of 2\pi , and in the above problem, we are dealing with the radius of the last sign change on \mathbb{R}/\mathbb{Z} . The following observations allow us to restrict our attention to a smaller family of trigonometric polynomials. Remark 0. (1) Note that \rho(c\mu,N)=\rho(\mu,N) for any c>0 and N\geq 1 . Thus, it suffices to work with probability measures; (2) The even part f_{e}(x)\coloneqq\tfrac{f(x)+f(-x)}{2} of a function f\in\mathcal{T}(\mu,N) satisfies f_{e}\in\mathcal{T}(\mu,N) and \operatorname{r}(f_{e})\leq\operatorname{r}(f) ; (3) If f\in\mathcal{T}(\mu,N) , then the function f_{0}(x)=f(x)-\int_{\mathbb{R}/\mathbb{Z}}f(t)\,\text{\rm d}\mu(t) also belongs to the class \mathcal{T}(\mu,N) and \operatorname{r}(f_{0})\leq\operatorname{r}(f) ; Consider the case where \text{\rm d}x is the Lebesgue measure on \mathbb{R}/\mathbb{Z} . Take a non-trivial function f\in\mathcal{T}(\text{\rm d}x,N) and let f_{\pm}(x)=\max\{\pm f(x),0\} denote the positive and negative parts of f . Since f has a non-positive integral, we have \begin{split}\lVert f\rVert_{L^{1}}\leq 2\int_{\mathbb{R}/\mathbb{Z}}f_{-}(x)\,\text{\rm d}x&=2\int_{-\operatorname{r}(f)}^{\operatorname{r}(f)}f_{-}(x)\,\text{\rm d}x\leq 4\operatorname{r}(f)\lVert f\rVert_{L^{\infty}}\\ &\leq 4(2N+1)\operatorname{r}(f)\max_{|n|\leq N}\lvert\widehat{f}(n)\rvert\leq 4(2N+1)\operatorname{r}(f)\lVert f\rVert_{L^{1}}.\end{split} It implies that \operatorname{r}(f)\geq 1/(4(2N+1)) for any f\in\mathcal{T}(\text{\rm d}x,N) . Clearly, it shows the presence of the sign uncertainty phenomenon for the Lebesgue measure. In fact, this particular case of the problem has been considered in works [ 3 , 25 ] . In this paper, we are going to establish a more general framework that allows us to identify the exact value of \rho(\mu,N) for any finite, even Borel measure \mu on \mathbb{R}/\mathbb{Z} . In particular, we will see that \rho(\text{\rm d}x,N)=1/(2N+2) . 1.3. Orthogonal polynomials on the unit circle To state our main results, we recall several important concepts from the classical theory of orthogonal polynomials on the unit circle (OPUC). The theory of OPUC will play a key role in our results, as the theory of de Branges spaces [ 9 ] was crucial in finding the smallest radius of last sign change for bandlimited functions [ 5 ] . Here we follow a classic reference [ 20 ] on the topic. A

Read the complete text