Convolution of periodic multiplicative functions and the divisor problem
The famous Erdős discrepancy problem asks whether for all arithmetic functions \(f : \mathbb{N} \rightarrow \{-1, 1\}\), we have the discrepancy\N\[\N\sup_{x, d} \Big| \sum_{n \le x} f(n d) \Big| = \infty.\N\]\NIn 2015, \textit{T. Tao} [Discrete Anal. 2016, Paper No. 1, 29 p. (2016; Zbl 1353.11087)] remarkably resolved this problem by reducing it to the class of completely multiplicative functions \(f\) taking values in the (complex) unit circle. Later, \textit{M. Aymone} [Bull. Braz. Math. Soc. (N.S.) 53, No. 4, 1317--1329 (2022; Zbl 1505.11123)] considered \(M\)-periodic multiplicative functions \(f\) allowing values outside the unit disk with \(f(M) \neq 0\) and proved that they has bounded partial sums if and only if the following three conditions are satisfied:\N\begin{align*}\N\text{i.} & \text{ For some prime } q \mid M, \, \sum_{k = 0}^{\infty} \frac{f(q^k)}{q^k} = 0. \\\N\text{ii.} & \text{ For each } p^a \| M, \, f(p^k) = f(p^a) \text{ for all } k \ge a. \\\N\text{iii.} & \text{ For each } \gcd(p, M) = 1, \, f(p^k) = 1 \text{ for all } k \ge 1.\N\end{align*}\NIn this nice article under review, the authors are interested in the situation where \(f\) is the convolution\N\[\Nf(n) = f_1 * f_2 (n) = \sum_{d \mid n} f_1(d) f_2(n / d)\N\]\Nfor periodic multiplicative functions \(f_1, f_2\) satisfying the above three conditions i, ii, iii. The authors establish the following \(\Omega\)-result: There exists some constant \(C > 0\) such that \N\[\N \Big| \sum_{n \le x} f_1 * f_2 (n) \Big| \ge C x^{1/4} \text{ for infinitely many positive integers } x.\tag{F}\N\]\NThe starting point is the following formula from [\textit{M. Aymone}, Bull. Braz. Math. Soc. (N.S.) 53, No. 4, 1317--1329 (2022; Zbl 1505.11123)]: If \(M_1\) and \(M_2\) are the periods of \(f_1\) and \(f_2\) respectively, then\N\[\N\sum_{n \le x} f_1 * f_2 (n) = \sum_{n \mid M_1 M_2} (f_1 * f_2 * \mu * \mu)(n) \Delta(x / n)\N\]\Nwhere \(\mu\) is the Möbius function and \(\Delta(x)\) is the classical error term in the Dirichlet divisor problem defined by\N\[\N\sum_{n \le x} \tau(n) = x \log x + (2 \gamma - 1) x + \Delta(x).\N\]\NThus, the partial sums of \(f_1 * f_2\) can be written as a finite linear combination of the quantities \((\Delta(x / n))_n\). Then (F) is established through studying the asymptotic of the following second moment\N\[\N\lim_{X \rightarrow \infty} \frac{1}{X^{3/2}} \int_{1}^{X} \Big| \sum_{n \le x} f_1 * f_2 (n) \Big|^2 dx\N\]\Nwhich, after expanding things out, can be expressed as a quadratic form with matrix \(\mathcal{M} := (c_{a, b})_{a, b \mid M_1 M_2}\) where\N\[\Nc_{a, b} := \lim_{x \rightarrow \infty} \frac{1}{X^{3/2}} \int_{1}^{X} \Delta(x / a) \Delta(x / b) dx.\N\]\NThese correlation coefficients, \(c_{a, b}\), are studied through a method inspired by \textit{K.-C. Tong} [Acta Math. Sin. 6, 139--152, 515--541 (1956; Zbl 0075.25003)]. The matrix \(\mathcal{M}\) is investigated via Selberg's diagonalization process and a clever change of basis relating it to a matrix with completely multiplicative coefficients. Sylvester's criterion is used to establish positive definiteness. As a by-product, the authors prove the following interesting result: For almost all irrational numbers \(\theta\) with respect to the Lebesgue measure, for all small fixed \(\varepsilon > 0\),\N\[\N\int_{1}^{X} \Delta(x) \Delta( \theta x ) dx = O( X^{3/2 - 1/18 + \varepsilon} ).\N\]
- Complex valued multiplicative functions with bounded partial sums
- Correlations of multiplicative functions and applications
- Einige Sätze über Kettenbrüche, mit Anwendungen auf die Theorie der Diophantischen Approximationen.
- Exponential Sums and Lattice Points III
- scientific article; zbMATH DE number 3125409 (Why is no real title available?)
- scientific article; zbMATH DE number 2005655 (Why is no real title available?)
- scientific article; zbMATH DE number 3057074 (Why is no real title available?)
- Mean square of the remainder term in the Dirichlet divisor problem
- On an approximate functional equation due to J. R. Wilton
- On certain integrals involving the Dirichlet divisor problem
- Pointwise and correlation bounds on Dedekind sums over small subgroups
- Pretentious multiplicative functions and an inequality for the zeta-function.
- Some unsolved problems
- Study of the multiplicative autocorrelation of the fractional part function
- The Erdös discrepancy problem
- The twisted moments and distribution of values of Dirichlet L-functions at 1
This page was built for publication: Convolution of periodic multiplicative functions and the divisor problem
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6893813)
