The distribution of sums and products of additive functions

DOI10.5802/JTNB.1113MaRDI QIDQ2199674zbMATH OpenOpenAlexFDO

Authors Greg Martin, Lee Troupe

Publication date 14 September 2020

Published in Journal de Théorie des Nombres de Bordeaux (Search for Journal in Brave)

Copyright license Creative Commons Attribution-NoDerivs 4.0 International

Full work available at URL https://arxiv.org/abs/1811.04297

zbMATH Keywords

additive functions Erdős-Kac theorem

Mathematics Subject Classification ID

Asymptotic results on arithmetic functions (11N37) Distribution functions associated with additive and positive multiplicative functions (11N60)

Abstract: The celebrated ErdH{o}s--Kac theorem says, roughly speaking, that the values of additive functions satisfying certain mild hypotheses are normally distributed. In the intervening years, similar normal distribution laws have been shown to hold for certain non-additive functions and for amenable arithmetic functions over certain subsets of the natural numbers. Continuing in this vein, we show that if

g_{1} (n), l d o t s, g_{k} (n)

is a collection of functions satisfying certain mild hypotheses for which an ErdH{o}s--Kac-type normal distribution law holds, and if

Q (x_{1}, l d o t s, x_{k})

is a polynomial with nonnegative real coefficients, then

Q (g_{1} (n), l d o t s, g_{k} (n))

also obeys a normal distribution law. We also show that a similar result can be obtained if the set of inputs

n

is restricted to certain subsets of the natural numbers, such as shifted primes. Our proof uses the method of moments. We conclude by providing examples of our theorem in action.

Review text

The celebrated Erdős-Kac theorem says that the values of an additive function satisfying certain mild hypothesis are uniformly distributed. More formally, if \(g\) is strongly additive, \(|g(p)|\le 1\) for all primes \(p\), defining \[ A_g(x):=\sum_{p\le x}\frac{g(p)}{p}\quad {\text{and}}\quad B_g(x):=\left(\sum_{p\le x} \frac{g^2(p)}{p}\left(1-\frac{1}{p}\right)\right)^{1/2}, \] and assuming that \(B_g(x)\) is unbounded, then one has that \[ \lim_{x\to \infty} \frac{1}{x}\#\{n\le x: g(n)<A_g(x)+uB_g(x)\}=\frac{1}{\sqrt{2\pi}} \int_{-\infty}^u e^{-t^2/2} dt. \] Let us say that \(g\) satisfies an Erdős-Kac law with mean \(A_g\) and variance \(B_g^2 \) if the above limit is satisfied for \(g\). The main result of the paper is that if \(Q(T_1,\ldots,T_\ell)\) is a polynomial with nonnegative real coefficients and \(g_1,\ldots,g_{\ell}\) are nonnegative strongly additive functions such that \begin{itemize} \item[(i)] \(g_j(p)\ll_j 1\) for all primes \(p\), \item[(ii)] for each \(j\in \{1,\ldots,\ell\}\) the series \(\sum_p g_j^2(p)/p\) diverges, \end{itemize} defining \[ \mu(g_j):=\sum_{p\le x}\frac{g_j(p)}{p},\quad \kappa(p_i,p_j):=\sum_{p\le x} \frac{g_i(p)g_j(p)}{p}\left(1-\frac{1}{p}\right),\quad \sigma^2(g_i):=\kappa(g_i,g_i), \] and assuming \(\sigma^2(g_i)\gg \mu(g_j)\) for all sufficiently large \(x\) and all \(1\le i,j\le \ell\), then \(Q(g_1,\ldots,g_\ell)\) satisfies an Erdős-Kac law with mean \(A(\mu(g_1),\ldots,\mu(g_\ell))\) and variance given explicitly in terms of the partial derivatives of \(Q\) and the quantities \(\kappa(g_i,g_j)\) for \(1\le i,j,\le \ell\). As the authors remark, the unnatural condition on the relative sizes of \(\sigma^2(g_i)\) versus \(\mu(g_j)\) is necessary for the proof in its generality but one can get away without these conditions in specific situations. They also show that under the same hypotheses, the same conclusion holds if \(n\) is allowed to vary only in the set of shifted primes, that is the set of numbers of the form \(p-a\) where \(p\) runs through primes and \(a\) is a fixed nonzero integer. The proof uses sieve methods and the Bombieri-Vinogradov theorem.

Recommendations

Cites work

Cited in

(5)

This page was built for publication: The distribution of sums and products of additive functions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2199674)