Averages of multiplicative functions along equidistributed sequences (Q7019149)

A function \(f : \mathbb{N} \rightarrow \mathbb{C}\) is multiplicative if \(f(a b) = f(a) f(b)\) for all \(\gcd(a, b) = 1\). People have been interested in mean values of multiplicative functions. More recently, works were done on mean values of multiplicative functions over values of integer polynomials [\textit{M. Nair} and \textit{G. Tenenbaum}, Acta Math. 180, No. 1, 119--144 (1998; Zbl 0917.11048); \textit{K. Henriot}, Math. Proc. Camb. Philos. Soc. 152, No. 3, 405--424 (2012; Zbl 1255.11048)], binary forms [\textit{R. de la Bretèche} and \textit{T. D. Browning}, Acta Arith. 125, No. 3, 291--304 (2006; Zbl 1159.11035)], and even principal ideals [\textit{T. D. Browning} and \textit{E. Sofos}, Int. J. Number Theory 15, No. 3, 547--567 (2019; Zbl 1443.11196)].\N\NMore generally, \textit{D. Wolke} [J. Reine Angew. Math. 251, 54--67 (1971; Zbl 0234.10030)] worked on mean value of non-negative multiplicative functions over values of an increasing integer sequence \(\{ c_a \}_{a = 1}^{\infty}\):\N\[\N\sum_{a \in \mathbb{N} \cap [1, T]} f(c_a).\N\]\NIn this article under review, the authors extend Wolke's work to even more versatile settings and obtain a general upper bound. To state the bound, let us introduce some notation. Let \(\mathcal{A}\) be an infinite set and, for each \(T \ge 1\), let \(\chi_T : \mathcal{A} \rightarrow [0, \infty)\) be any function (some sort of counting function) for which\N\[\N\{a \in \mathcal{A} : \chi_T(a) > 0 \} \text{ is finite for every } T \ge 1\N\]\Nand\N\[\N\lim_{T \rightarrow \infty} \sum_{a \in \mathcal{A}} \chi_T(a) = \infty.\N\]\NThe authors consider the following class of ``general sub-multiplicative'' functions \(\mathcal{M}(A, \varepsilon, C)\) satisfying\N\[\Nf(a b) \le f(a) \min \{ A^{\Omega(n)}, C n^\varepsilon \}\N\]\Nwhere \(\Omega(n)\) counts the total number of prime factors of \(n\) with multiplicity. Moreover, one requires the following equidistribution assumption:\N\[\N\mathop{\sum_{a \in \mathcal{A}}}_{c_a \equiv 0 (\bmod{d})} \chi_T(a) = h_T(d) M(T) \biggl[ 1 + O \biggl( \mathop{\prod_{B < p \le M(T)}}_{p \nmid d} (1 - h_T(p))^2 \biggr) \biggr] + O(M(T)^{1-\xi})\N\]\Nfor \(T \ge 1\) and \(d \le M(T)^\theta\) with \(\max \{\theta, \xi\} < 1\). Here \(h_T(d) \in \mathcal{D}(\kappa, \lambda_1, \lambda_2, B, K)\), some set of multiplicative density functions over some positive real parameters \(\kappa, \lambda_1, \lambda_2, B, K\). The set of density functions \(\mathcal{D}\) satisfy the following sieve-like requirements:\N\[\N\mathop{\prod_{p \text{ prime}}}_{w \le p < z} (1 - h(p))^{-1} \le \Bigl( \frac{\log z}{\log w} \Bigr)^\kappa \Bigl(1 + \frac{K}{\log w} \Bigr),\N\]\N\[\Nh(p^e) \le \frac{B}{p} \; \text{ for } p > B, \; \; \text{ and } \; \; h(p^e) \le p^{-e \lambda_1 + \lambda_2}.\N\]\NThe main result (omitting some technical details) is the following upper bound:\N\[\N\sum_{a \in \mathcal{A}} \chi_T(a) f(c_a) \ll M \prod_{B < p \le M} (1 - h_T(p)) \sum_{a \le M} f(a) h_T(a).\N\]\NIts method of proof builds on works of \textit{D. Wolke} [J. Reine Angew. Math. 251, 54--67 (1971; Zbl 0234.10030)] and \textit{P. Shiu} [J. Reine Angew. Math. 313, 161--170 (1980; Zbl 0412.10030)] by breaking into four cases, depending on the smoothness of certain factorization and potential divisibility by a large prime power. Careful estimations are done through establishing progressively more refined and related lemmas. Rankin's trick and the fundamental lemma of sieve theory are applied.\N\NA matching lower bound is also established when \(f\) is not too close to \(0\). As a concrete corollary, the authors prove that, for any irreducible polynomial \(Q(x_1, x_2, \ldots, x_n) \in \mathbb{Z}[x_1, x_2, \ldots, x_n]\) and \(T \ge 2\), there exist constants \(C_1, C_2 > 0\) depending on \(Q\) and \(n\) such that\N\[\NC_1 T^n \log T \le \mathop{\sum_{\textbf{x} \in \mathbb{Z}^n, |x_i| \le T}}_{Q(\textbf{x}) \neq 0} \tau( |Q(\textbf{x})| ) \le C_2 T^n \log T\N\]\Nwhere \(\tau\) is the divisor function.