Central limit theorems for random multiplicative functions (Q6183789)

Two natural models for random multiplicative functions are (1) the Steinhaus model of a random completely multiplicative function \(f(n)\) where the values \(f(p)\) (on primes \(p\)) are chosen independently and uniformly from the unit circle, and (2) the Rademacher model of a multiplicative function \(f(n)\) taking values \(\pm 1\) on square-free integers and \(0\) on integers having a square factor, with \(f(p)\) chosen independently and uniformly from \(\{-1, 1\}\). One of the goals in the present paper is to explore the distribution of partial sums of random multiplicative functions when restricted to subsets \(A\) of \([1, N]\). For, the authors prove that if for large \(N\), the subset \(A\) of \([1, N]\) satisfies \(|A|\geq N\exp(-\frac{1}{3}\sqrt{\log N\log\log N})\), and that there exists \(S\subset A\) with size \(|S|=(1+o(1))|A|\) satisfying the multiplicative energy condition \[ E_\times(S):=|\{(s_1,s_2,s_3,s_4)\in S^4: s_1s_2=s_3s_4\}|=(2+o(1))|S|^2, \] then as \(f\) ranges over random multiplicative functions in the Steinhaus model, the average quantity \[ \frac{1}{\sqrt{|A|}}\sum_{n\in A}f(n) \] is distributed like a standard complex normal random variable with mean \(0\) and variance \(1\). The authors use the this result to give several new examples of subsets \(A\) where a central limit theorem holds. The first example addresses the set \(A\) as the integers in the interval \([x,x+y]\) such that \(x\) and \(y\) are large with \(y\leq x/(\log x)^{2\log 2-1+\varepsilon}\) where \(\varepsilon>0\) is arbitrary. The second example addresses the set \(A\) consisting of the integers that are the sum of two integers, belonging to the interval \([x,x+y]\) such that \(x\) and \(y\) are large with \(x^{1/3}<y=o(x)\). The third example shows that the central limit theorem also holds for the set of shifted primes \(p+k\), for any fixed non-zero integer \(k\). The forth example is motivated by the question that what is the largest subset \(A\) of \([1, N]\) for which one can establishes a central limit theorem for \(\sum_{n\in A}f(n)\)? The largest set \(A\) that the authors were able to find is related to the Erdős multiplication table problem. For, the authors show that for all large \(N\) there exits a subset \(A\) of \([1, N]\) with \(|A|\geq N(\log N)^{-\theta}(\log\log N)^{-7}\) where \(\theta=1-(1+\log\log 4)/\log 4\) such that, for random Steinhaus multiplicative functions \(f\), the quantity \[ \frac{1}{\sqrt{|A|}}\sum_{n\in A}f(n) \] is distributed like a standard complex normal random variable with mean \(0\) and variance \(1\). Finally, the authors consider the set \(A=[1,N]\) and prove that if \(\alpha\) an irrational number such that \(\min_{n\in\mathbb{Z}}|q\alpha-n|\geq C\exp(-q^{1/50})\) for some positive constant \(C=C(\alpha)\) and all \(q\in\mathbb{N}\), and if \(N\) is sufficiently large (in terms of \(\alpha\)), then as \(f\) varies over Steinhaus random multiplicative functions, the quantity \[ \frac{1}{\sqrt{N}}\sum_{n\leq N}e^{2\pi i n \alpha}f(n) \] is distributed like a standard complex normal random variable with mean \(0\) and variance \(1\). As some useful auxiliary results, the authors develop the martingale central limit theorem in a quantitative form, based on the work of \textit{D. L. McLeish} [Ann. Probab. 2, 620--628 (1974; Zbl 0287.60025)].