On the number of abelian groups of given order. I (Q2649917)

The author improves a result of \textit{D. G. Kendall} and \textit{R. A. Rankin} [Q. J. Math. Oxf. 18, 197--208 (1947; Zbl 0031.15303)] about the average of the number of abelian groups. More precisely, let \(a(n)\) be the number of essentially different abelian groups of order \(n\). Then \[ \sum_{n\leq x} a(n)=c_1x+c_2x^{1/2}+c_3x^{1/3}+O(x^{3/10}\log^{9/10}x) \] where \(c_\mu=\displaystyle\prod_{\nu=1\atop \nu\neq\mu}^\infty\zeta\left(\frac{\nu}{\mu}\right)\). The result is deduced from the following theorem which is of interest in its own: Let \(\mu\) and \(\nu\) be real numbers, \(\mu>\nu>0\), then \[ \sum_{m^\mu n^\nu\leq x} 1=\zeta\left(\frac{\mu}{\nu}\right)x^{1/\nu}+\zeta\left(\frac{\nu}{\mu}\right)x^{1/\mu}+\begin{cases} O\left(x^{2/3(\mu+\nu)}\right)&\qquad \text{for } \mu<2\nu, \\ O\left(x^{2/9\nu}\log x\right)&\qquad \text{for } \mu=2\nu, \\ O\left(x^{2/(2\mu+5\nu)}\right)&\qquad \text{for } \mu>2\nu.\end{cases} \]