Restriction estimates with sifted integers

Tanmoy Bera, G. K. Viswanadham · 2025

Read Restriction estimates with sifted integers by Tanmoy Bera, G. K. Viswanadham (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-2512-21640.txt.

Introduction

Tanmoy Bera and G. K. Viswanadham prove restriction estimates for sifted integers in analytic number theory. Restriction estimates bound the Fourier transform of a set of integers, controlling how concentrated it can be. Sifting means keeping only the integers that survive a sieve, here defined prime by prime through allowed residue classes.

Their result generalizes a theorem of Green and Tao. In the setup, one fixes a set of primes and, for each prime, a subset of residue classes modulo that prime. The integers up to a bound N are sifted by these local conditions up to a sieve level z. Bera and Viswanadham provide restriction estimates for the surviving set. Such estimates are a basic tool in additive combinatorics and sieve theory, where they underpin results about patterns in primes and almost-primes.

Catalog record

CREATORTanmoy Bera, G. K. Viswanadham
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-21640

Browse the full library · About the Tower of Babel

Full text

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

Restriction estimates with sifted integers Tanmoy Bera, G. K. Viswanadham Abstract . Let 𝒫 \mathcal{P} be a subset of primes and for each prime p ∈ 𝒫 p\in\mathcal{P} , consider a subset ℒ p \mathcal{L}_{p} of ℤ / p ​ ℤ \mathbb{Z}/p\mathbb{Z} . We provide restriction estimates with integers ≤ N \leq N sifted by ( ℒ p ) p ≤ z p ∈ 𝒫 (\mathcal{L}_{p})_{\begin{subarray}{c}p\leq z\\ p\in\mathcal{P}\end{subarray}} . This generalizes a result of Green-Tao [ 3 ] on the restriction estimates. 2020 Mathematics Subject Classification 11L07, 11N36 1. Introduction For a given real number \alpha\in[0,1] and positive integer N , define (1.1) S(\alpha,N)=\sum_{p\leq N}a_{p}e(p\alpha)\;, where the sum runs over primes p\leq N and (a_{p})_{p\leq N} is a sequence of complex numbers. Bourgain [ 1 ] obtained an upper bound for the L_{\ell} -norm of S(\alpha,N) for any \ell>2 . More precisely, he obtained that (1.2) \left(\int_{0}^{1}|S(\alpha,N)|^{\ell}d\alpha\right)^{1/\ell}\ll_{\ell}N^{-1/\ell}\left(\frac{N}{\log N}\sum_{p\leq N}|a_{p}|^{2}\right)^{1/2}\;. The above bound can be seen to be optimal, up to the constant. The estimates of this type are called restriction estimates . An alternative proof of the result of Bourgain [ 1 ] is given by Green (see [ 2 , Theorem 2.1] ) and used it to show that the primes enjoy the Hardy-Littlewood majorant property: (1.3) \int_{0}^{1}\left|\sum_{p\leq N}a_{p}e(p\alpha)\right|^{\ell}\ll_{\ell}\int_{0}^{1}\left|\sum_{p\leq N}e(p\alpha)\right|^{\ell}\; for any sequence (a_{p}) of complex numbers with |a_{p}|\leq 1 and \ell\geq 2 . These results were generalized to a wider range of arithmetic sets, instead of primes in ( 1.1 ), by Green-Tao [ 3 ] , using the enveloping sieve developed by Ramaré [ 7 ] , and Ramaré and Ruzsa [ 10 ] . In the other direction, restriction estimates with smooth numbers are obtained by Harper (see Theorem 2, [ 4 ] ). The restriction estimate ( 1.2 ) fails at the endpoint \ell=2 , which can be seen using Parseval’s identity. Ramaré [ 8 ] provided a smooth transition from the case \ell=2 to the case \ell>2 and obtained explicit values for the constants in the estimates ( 1.2 ) and ( 1.3 ). He showed that the constant in ( 1.3 ) can be chosen to be absolute, namely 10^{5} when N\geq 10^{6} . He also provided restriction estimates with primes in intervals. In a recent preprint, Ramaré [ 9 ] improved these explicit constants. The applications of restriction estimates to number theory was initiated by Bourgain [ 1 ] . Later, these estimates were found to be very useful in number theory, for example, see [ 2 , 4 ] . The aim of this article is to obtain restriction estimates with sifted sequences, which we will describe below. Let \mathcal{P} be a subset of primes, and 2\leq y_{0}\leq y be two real numbers. We define \mathcal{P}(y)=\prod_{\begin{subarray}{c}p<y\\ p\in\mathcal{P}\end{subarray}}p,\text{ and }\mathcal{P}(y_{0},y)=\mathcal{P}(y)/\mathcal{P}(y_{0}). For any p\in\mathcal{P} , let \mathcal{L}_{p} be a non empty subset of \mathbb{Z}/p\mathbb{Z} with cardinality \lambda(p) . We assume that there exists a positive constant M such that (H1) \lambda(p)\leq Mp^{1/4}\text{ for all }p\in\mathcal{P}. For primes p\notin\mathcal{P} we take \mathcal{L}_{p}=\emptyset. We then extend it for any squarefree integer \ell by defining \mathcal{L}_{\ell}=\prod_{p|\ell}\mathcal{L}_{p}\subseteq\mathbb{Z}/\ell\mathbb{Z},\;\;\text{ and }\;\;\;\lambda(\ell)=\prod_{p|\ell}\lambda(p). For any y\geq 2 , we denote V(y):=\prod_{\begin{subarray}{c}p<y\\ p\in\mathcal{P}\end{subarray}}\left(1-\frac{\lambda(p)}{p}\right)^{-1}. We also assume that there exist positive constants T,\kappa such that for any y\geq 2 (H2) V(y)\leq T\log^{\kappa}y. Let z be a real number. For any positive integer N, define \mathcal{S}(N):=\mathcal{S}(N,\mathcal{P},z):=\{n\leq N:n\left(\mathrm{mod}\,p\right)\notin\mathcal{L}_{p}\text{ for any }\;p\in\mathcal{P}\text{ and }p<z\}. The main theorem of this article is the following: Theorem 1.1 . Let z\leq N^{1/4} and let \mathcal{S}(N) be as defined above. Then for any complex numbers (a_{n})_{n\leq N} and \ell>2 , we have (1.4) \int_{0}^{1}\left|\sum_{n\in\mathcal{S}(N)}a_{n}e(n\alpha)\right|^{\ell}d\alpha\leq\frac{C}{N}\left(\frac{2N}{V(z)}\sum_{n\in\mathcal{S}(N)}|a_{n}|^{2}\right)^{\ell/2}\;, where C=C(\kappa,\ell,T,M) is a constant that can be computed explicitly. As one can see by Parseval identity, Equation ( 1.4 ) doesn’t hold for \ell=2 . The exact dependency of the constant on \ell is given in ( 6.2 ) and ( 6 ), and this provides a smooth transition from \ell>2 to \ell=2 . This striking feature comes from the method of Ramaré [ 8 ] . Let \delta>0 be a real number. A finite set \mathcal{B}\subseteq\mathbb{R}/\mathbb{Z} is said to be a \delta -well spaced set if \|x_{i}-x_{j}\|\geq\delta\text{ for all }x_{i},x_{j}\in\mathcal{B}, where \|x\|=\min_{k\in\mathbb{Z}}|x+k| denotes the distance from the closest integer to x . The above theorem is a simple application of the following theorem: Theorem 1.2 . Let z\leq N^{1/4} and let \mathcal{S}(N) be as defined above and \mathcal{B} be a \delta -well spaced subset of \mathbb{R}/\mathbb{Z}. . Then for any complex numbers (a_{n})_{n\leq N} and \ell>2 , we have (1.5) \sum_{b\in\mathcal{B}}\left|\sum_{n\in\mathcal{S}(N)}a_{n}e(nb)\right|^{\ell}\leq C\left(\frac{N+\delta^{-1}}{V(z)}\sum_{n\in\mathcal{S}(N)}|a_{n}|^{2}\right)^{\ell/2}\;, where C=C(\kappa,\ell,T,M) is a constant that can be computed explicitly. By taking \mathcal{B}=\{\beta+k/N;0\leq k\leq N-1\} and integrating over \beta in [0,1/N] , Theorem 1.1 follows. In the proof of the above theorem, we have provided the exact dependency of the constant on \ell (see ( 6.2 ) and ( 6 )). To prove Theorem 1.2 , we need a dual large sieve type inequality, which we will describe below. Let N\geq 10^{3} be a natural number, and (u_{p})_{p\leq N} be a sequence of complex numbers. In [ 8 ] , Ramaré proved that \sum_{b\in\mathcal{B}}\left|\sum_{p\leq N}u_{p}e(bp)\right|^{2}\leq 280\frac{N+\delta^{-1}}{\log N}\log(2|\mathcal{B}|)\sum_{p\leq N}|u_{p}|^{2}. The constant 280 is further improved to 4e^{\gamma}+o(1) by Ramaré [ 9 , Theorem 2.4] . We need an analogous estimate with \mathcal{S}(N) instead of primes p\leq N . Theorem 1.3 . Let N and z be positive real numbers such that z\leq N^{1/4}, and \delta>0. Let \mathcal{B} be a \delta -spaced subset of \mathbb{R}/\mathbb{Z}. Then for any non-zero function f on \mathcal{B}, we have \sum_{n\in\mathcal{S}(N)}\left|\sum_{b\in\mathcal{B}}f(b)e(nb)\right|^{2}\leq K\frac{N+\delta^{-1}}{V(z)}\log^{\kappa}\left(2\frac{\|f\|_{1}^{2}}{\|f\|_{2}^{2}}\right)\|f\|_{2}^{2} with some constant K that depends only on T , M and \kappa . This result is the heart of our paper. The proof of Theorem 1.3 relies on the enveloping sieve of Ramaré and Ruzsa (see Section 4 of [ 10 ] ) and follows the method of Ramaré [ 8 ] . The strategy is not to study the characteristic function of the set \mathcal{S}(N) directly, but rather to construct a majorant \beta , which has good Fourier properties, using the enveloping sieve. Using the theorem above, we can prove the following large sieve inequality with sifted integers by the standard argument. Theorem 1.4 . Let N and z be positive real numbers such that z\leq N^{1/4}, and \delta>0. Let \mathcal{B} be a \delta -spaced subset of \mathbb{R}/\mathbb{Z}. Then for any sequence of complex numbers (a_{n})_{n\leq N} , we have \sum_{b\in\mathcal{B}}\left|\sum_{n\in\mathcal{S}(N)}a_{n}e(nb)\right|^{2}\leq K\frac{(N+\delta^{-1})}{V(z)}\log^{\kappa}(2|\mathcal{B}|)\sum_{n\in\mathcal{S}(N)}|a_{n}|^{2}\;. Organization of the paper: In Section 2 , we provide some necessary lemmas. In Section 3 , we discuss the enveloping sieve. In sections 4 , 5 and 6 , we prove Theorems 1.3 , 1.4 and 1.2 respectively. We will discuss Hardy-Littlewood majorant property for \mathcal{S}(N) in Section 7 . In the last section, we provide some applications of our results. 2. Necessary Lemmas Let h be the non-negative multiplicative function supported on square-free integers defined by h(p)=\frac{\lambda(p)}{p-\lambda(p)} for any p\in\mathcal{P} , and 0 if p\notin\mathcal{P} . For any positive integer d , and real numbers y\geq 1,z_{0}\geq 2 , let \displaystyle G_{d}(y;z_{0})\>=\sum_{\begin{subarray}{c}\ell\leq y\\ (\ell,\;d\mathcal{P}(z_{0}))=1\end{subarray}}h(\ell)\;, and let G_{1}(y,z_{0})=G(y,z_{0}) , G(y)=G_{1}(y,2) . Lemma 2.1 . When d is coprime to \mathcal{P}(z_{0}) , we have G(yd;z_{0})\geq\left(\sum_{\delta\mid d}h(\delta)\right)G_{d}(y;z_{0})\geq G(y;z_{0})\;. Proof. We have \displaystyle G(y;z_{0})=\sum_{\delta\mid d}\sum_{\begin{subarray}{c}\ell\leq y\\ (\ell,\mathcal{P}(z_{0}))=1\\ (\ell,d)=\delta\end{subarray}}h(\ell)=\sum_{\delta\mid d}h(\delta)\sum_{\begin{subarray}{c}m\leq y/\delta\\ (m,d\mathcal{P}(z_{0})=1)\end{subarray}}h(m)\geq\left(\sum_{\delta\mid d}h(\delta)\right)G_{d}(y/d,z_{0})\;. Taking yd in

Read the complete text