Multiplicative functions of polynomial values in short intervals (Q4024734)

\textit{P. Erdős} [J. Lond. Math. Soc. 27, 7--15 (1952; Zbl 0046.04103)] showed that \(\sum_{n \leq x} d(| P(n)|) \ll x\log x\). Here \(d(n)\) is the divisor function, \(P(n)\) is an irreducible polynomial with integer coefficients, and the implied constant may depend on \textit{P. D. Wolke} [J. Reine Angew. Math. 251, 54--67 (1971; Zbl 0234.10030)] generalized and extended the result to deal with sums of the form \(\sum_{n \leq x}f(a_ n)\), where \(f\) is a suitable non-negative multiplicative function, and \((a_ n)\) is a sequence of positive integers with a structure amenable to the sieve method. The reviewer [J. Reine Angew. Math. 313, 161--170 (1980; Zbl 0412.10030)] applied the Erdős-Wolke method to give a uniform bound for the sum over arithmetic progressions, that is linear polynomials, enlarging the family of multiplicative functions \(f\), and also giving a short-interval result. The author extends such ideas further to bound sums of the form \(\sum_{x-y < n \leq x}f(| P(n)|)\), preserving sufficient uniformity with respect to the polynomial \(P\) to imply the reviewer's theorem when \(P(n) = kn + 1\), \((a,k) = 1\). In the general case, he also allows \(P(n)\) to be reducible. The main theorem is too complicated to be given in full and we give an example instead. Let \(d_ r(n)\) denote the number of ways of writing \(n\) as a product of \(r\) factors, \(\ell \geq 1\) and \(0 < \alpha,\beta < 1\). Then, as \(x \to \infty\), we have \[ \sum_{\displaystyle{x-y < n \leq x\atop n \equiv \alpha\pmod k}}d^ \ell_ r(| P(n)|) \ll {y\over k}\left({\phi(k)\over k}\log x\right)^{r^ \ell-1}, \] uniformly in \(a\), \(k\) and \(y\), with \(a \leq k\), \((k,P(a)) = 1\), \(x^ \alpha \leq y \leq x\) and \(k \leq y^{1- \beta}\), but the implied constant may depend on \(P\), \(r\), \(\ell\), \(\alpha\) and \(\beta\).