Bounds for the coefficients of cyclotomic polynomials (Q1216332)

Let \(\displaystyle\Phi(z) = \prod_{\substack{{r=1} \\ (r,n)=1}}^n (z - e(r/n))\) \((e(\alpha) = e^{2\pi i\alpha})\) denote the \(n\)-th cyclotomic polynomial and put \(\displaystyle\Phi_n(z)= \sum_{m=0}^{\varphi(n)} a(m,n)z^m\). \textit{P. T. Bateman} [Bull. Am. Math. Soc. 55, 1180--1181 (1949; Zbl 0035.31102)] showed that \[ \vert a(m,n)\vert < \exp(\tfrac12 d(n) \log n), \tag{*} \] where \(d(n)\) is the number of divisors of \(n\). \textit{P. Erdős} [Port. Math. 8, 63--71 (1949; Zbl 0038.01004); Proc. Glasg. Math. Assoc. 3, 102--104 (1957; Zbl 0081.01703)] proved the existence of a positive \(c\) such that, for infinitely many \(n\), \[ \log \max_m a(m,n) > \exp \left(\frac{c \log n}{\log \log n}\right). \] He asked whether it is possible to take \(c\) arbitrarily close to \(\log 2\), which would imply that (*) is best possible. In the present paper it is shown that this is the case and in fact \(c = \log 2\) is permissible. Theorem 1. There are infinitely many \(n\) such that \[ \log \max_m a(m,n) > \exp\left(\frac{(\log 2)(\log n)}{\log \log n}\right). \] Theorem 2. There are infinitely many \(m\) such that \[ \log \max_n \vert a(m, n)\vert \gg \frac{m^{1/2}}{(\log m)^{1/4}}. \]