The first factor of the class number of the \(p\)-th cyclotomic field (Q2447687): Difference between revisions

Let \(p\) be an odd prime. Let \(h_{p}\) denote the class number of the \(p\)-th cyclotomic field \(\mathbb{Q}(\zeta_{p})\) and let \(h_{p}^{+}\) denote the class number of its maximal totally real subfield \(\mathbb{Q}(\zeta_{p}+\zeta_{p}^{-1})\). It is well-known that \(h_{p}^{+}\) divides \(h_{p}\). The quotient \(h_{p}^{-}:=h_{p}/h_{p}^{+}\) is called the relative class number of \(\mathbb{Q}(\zeta_{p})\), and is also sometimes referred to as the first factor of \(h_{p}\). Let \(L(s,\chi)\) denote the Dirichlet \(L\)-function of a Dirichlet character \(\chi\) of conductor \(p\) and recall that such a character is said to be odd if \(\chi(-1)=-1\). From the analytic class number formula, one can deduce that \( h_{p}^{-} = G(p) \prod_{\chi} L(1,\chi) \) where \( G(p) = 2p \left(\frac{p}{4\pi^{2}}\right)^{\frac{p-1}{4}} \) and the product is taken over all odd Dirichlet characters \(\chi\) of conductor \(p\). The guess that \(h_{p}^{-}\) is asymptotically equivalent to \(G(p)\) is known as Kummer's Conjecture. \textit{A. Granville} [Invent. Math. 100, No. 2, 321--338 (1990; Zbl 0701.11051)] has shown that Kummer's conjecture is incompatible with two well-known and widely-believed conjectures in analytic number theory, and therefore is unlikely to be true. A Siegel zero is a type of potential counterexample to the generalized Riemann hypothesis; more precisely, it is a hypothetical zero of \(L(s,\chi)\) that is very close to \(1\) in a quantifiable sense. If \(L(s,\chi)\) has such a zero then \(\chi\) is necessarily quadratic; furthermore, a Siegel zero is necessarily real and simple. Note that if \(p \equiv 1\bmod 4\), the odd characters are not quadratic and hence their corresponding Dirichlet \(L\)-functions have no Siegel zero. The paper under review gives new bounds for the error term in Kummer's conjecture as follows. If no Siegel zero is present among the odd Dirichlet \(L\)-functions of conductor \(p\), then the relative class number of \(\mathbb{Q}(\zeta_{p})\) satisfies \[ |\log(h_{p}^{-}/G(p))| \leq 2 \log_{2}(p) + O(\log_{3}(p)). \] If there is a Siegel zero \(\beta\) present among the odd Dirichlet \(L\)-functions of conductor \(p\), then \[ |\log(h_{p}^{-}/G(p))- \log(1-\beta)| \leq 4 \log_{2}(p) + O(\log_{3}(p)). \] As a corollary, the author obtains the following result: for \(p > 9649\) we have \[ h_{p}^{-} \leq 2p\left(\frac{p}{39}\right)^{\frac{p-1}{4}}, \] where the upper bound is an approximation to \(G(p)\) with \(4\pi^{2}\) replaced by the smaller constant \(39\).