Partial sums of typical multiplicative functions over short moving intervals (Q6154495)

For positive integer \(X\) let \([X]:= \{1,2,\dots, X\}\) and \(\mathcal{F}_X\) be the set of completely multiplicative functions \(f: [X]\to \{|z|=1\}\). In the paper under review, the authors prove that letting integer \(X\) be large and \(W(X)\) tend to infinity arbitrarily slowly as \(X\) tends to infinity, also, letting \(H: = H(X) \ll X(\log X)^{-W(X)}\) and \(H\to +\infty\) as \(X \to +\infty\), then, for almost all \(f\in \mathcal{F}_{X+H}\), as \(X\to +\infty\), \[ \frac{1}{\sqrt{H}} \sum_{x<n \le x+H} f(n) \xrightarrow{d} \mathcal{CN}(0,1), \] where \(x\) is chosen uniformly from \([X]\). Here, the notation \(\xrightarrow{d}\) to denote convergence in distribution, \(\mathcal{CN}(0,1)\) denotes the standard complex normal distribution, almost all means the total measure of such \(f\) is \(1-o_{X\to +\infty}(1)\) under \(\nu_{X+H}\), and for any given prime \(p\) letting \(\mu_p\) to be the uniform distribution on the set \(\{f(p)\} = \{|z|=1\}\), the product measure \(\nu_X\) defined on \(\mathcal{F}_X\) by \(\nu_X:= \prod_{p\le X}\mu_p\). To prove the above main result of the paper, the authors establish moment statistics in several situations. First, they show that the integer moments of random multiplicative functions \(f\) supported on suitable short intervals match the corresponding Gaussian moments. More precisely, letting \(\mathbb{E}_f\) to be ``average over \(f\in \mathcal{F}_X\) with respect to \(\nu_X\)'', they prove that \[ \mathbb{E}_f \biggl|\frac{1}{\sqrt{H}}\sum_{x<n\le x+H}f(n)\biggr|^{2k} = k! + O_k\biggl(H^{-1} + \frac{H^{1/2}}{\max(x, H)^{1/2}} + \frac{H\cdot (\log{x} + \log{H})^{E(k)}}{\max(x, H)}\biggr), \] where \(x,H,k\ge 1\) are integers, \(f\in \mathcal{F}_{x+H}\) and \(E(k) = 2k^2+2\). The second step towards the above main result of the paper is to establish concentration estimates for the moments of the random sums given by \[ A_H(f, x) := \frac{1}{\sqrt{H}}\sum_{x<n \le x+H} f(n), \] where \(x\) is uniformly chosen from \([X]\), behave like a complex standard Gaussian. To explain this critical step in details, let \(\mathbb{E}_x\) denotes ``expectation over \(x\) uniformly chosen from \([X]\)''. Also, let \(X,k\ge 1\) be integers with \(X\) large, and suppose that \(H:=H(X)\to +\infty \) as \(X\to +\infty\). Letting \(f\in \mathcal{F}_{X+H}\), the authors show that there exists a large absolute constant \(A>0\) such that the following holds as long as \(H\ll X(\log X)^{-C_k}\) with \(C_k = A k^{A k^{A k}}\), \[ \mathbb{E}_f \biggl( \mathbb{E}_x \biggl|\frac{1}{\sqrt{H}}\sum_{x<n \le x+H} f(n)\biggr|^{2k} -k!\biggr)^{\!2}=o_{X\to +\infty}(1). \] Furthermore, for any fixed positive integer \(\ell < k\), one has \[ \mathbb{E}_f \biggl|\mathbb{E}_x \biggl(\frac{1}{\sqrt{H}}\sum_{x<n \le x+H} f(n)\biggr)^{\!k} \biggl(\frac{1}{\sqrt{H}}\sum_{x<n \le x+H} \overline{f(n)}\biggr)^{\!\ell}\, \biggr|^{2}= o_{X\to +\infty}(1). \]