The canonical fractional Galois ideal at \(s = 0\) (Q927725)

Let \(K/k\) be a finite abelian extension of number fields with Galois group \(G\). Let \(S\) be a finite set of places of \(k\) containing the infinite places. The author constructs a canonical \(\mathbb{Z}[G]\)-submodule \(\mathcal{J}(K/k,S)\) of \(\mathbb{Q}[G]\) which can be seen as an \(r=0\) analogue of the fractional ideals \(\mathcal{J}^{r}(K/k,S)\) defined for \(r<0\) by \textit{V. P. Snaith} in [Can. J. Math. 58, No. 2, 419--448 (2006; Zbl 1215.11111)]. Crucial ingredients in the construction are the leading terms of Artin \(L\)-functions \(L_{K/k,S}^{*}(0, \chi)\) and the Stark regulators \(R_{\chi}^{f}\) for each \(\chi \in \widehat{G}\). (Here \(f\) is a choice of \(\mathbb{Q}[G]\)-module isomorphism \(f:\mathcal{O}_{K,S}^{\times} \bigotimes_{\mathbb{Z}} \mathbb{Q} \to X \bigotimes_{\mathbb{Z}} \mathbb{Q}\) where \(X\) is the kernel of the augmentation map \(\mathbb{Z}S \to \mathbb{Z}\).) Indeed, the existence of \(\mathcal{J}(K/k,S)\) relies upon the truth of the classical Stark conjecture, though it is independent of the choice of \(f\). Let \(r(\chi)=r_{K/k,S}(\chi)\) denote the order of vanishing of the Artin \(L\)-function \(L_{K/k,S}(s, \chi)\) at \(s=0\). Recall that the Brumer-Stickelberger element is defined as \[ \theta_{K/k,S}(1) = \sum_{\chi \in \widehat{G}} L_{K/k,S}(0,\chi)e_{\bar{\chi}} \] (though some use different notation). Let \(e_{0}=e_{0}(K/k,S) \in \mathbb{C}[G]\) be the sum of idempotents corresponding to the characters \(\chi \in \widehat{G}\) with \(r(\chi)=0\). Then in fact \(e_{0} \in \mathbb{Q}[G]\) and we have \[ \mathbb{Z}[G] \theta_{K/k,S}(1) = e_{0} \mathcal{J}(K/k,S). \] It is therefore hoped that \(\mathcal{J}(K/k,S)\) will participate in annihilation relations generalising and improving those involving \(\theta_{K/k,S}(1)\), such as Brumer's Conjecture. Under certain hypotheses on \(S\) and the assumption that \(r(\chi)=1\) for \(\chi \in \widehat{G}\), the author shows that \[ \mathcal{J}(K/k,S) = \frac{1}{e} \text{ann}_{\mathbb{Z}[G]}(\mathcal{O}_{K,S}^{\times}/\mathcal{E}) \] where \(e=|\mu(K)|\) is the number of roots of unity in \(K\) and \(\mathcal{E}\) is the group of Stark units attached to \((K/k,S)\) (note that \(\mathcal{E}\) is known to exist in the situation under consideration). He then uses this result to give explicit descriptions \(\mathcal{J}(K/k,S)\) in three related cases and compares them with each other. Finally, the author proves the following result. Let \(\ell\) be an odd prime, let \(k=\mathbb{Q}\), and let \(K^{+}\) be the maximal totally real subfield of the cyclotomic field \(K=\mathbb{Q}(\zeta_{\ell^{n+1}})\) for some \(n \geq 0\). Let \(G^{+} = \text{Gal}(K^{+}/\mathbb{Q})\) and \(S=\{ \infty, \ell \}\). If \(\mathcal{E}^{+}(\ell)\) is the \(\mathbb{Z}_{\ell}[G^{+}]\)-submodule of \(U(\ell) = \mathcal{O}_{K^{+},S}^{\times} \bigotimes_{\mathbb{Z}} \mathbb{Z}_{\ell}\) generated by \((1-\zeta_{\ell^{n+1}})(1-\zeta_{\ell^{n+1}}^{-1})\), then \[ \text{ann}_{\mathbb{Z}_{\ell}[G^{+}]}(U(\ell)/\mathcal{E}^{+}(\ell)) \subseteq \text{ann}_{\mathbb{Z}_{\ell}[G^{+}]}(\text{Cl}(K^{+}) \bigotimes_{\mathbb{Z}} \mathbb{Z}_{\ell}). \] This easily leads to annihilation relations involving \(\mathcal{J}(K^{+}/\mathbb{Q},S)\) and \(\mathcal{J}(K/\mathbb{Q},S)\). Of course, Vandiver's conjecture predicts that the \(\ell\)-part of \(\text{Cl}(K^{+})\) is in fact trivial. The author further points out that the above result is a direct consequence of work of \textit{P. Cornacchia} and \textit{C. Greither} [J. Number Theory 73, No. 2, 459--471 (1998; Zbl 0926.11085)] and also follows from work of \textit{K. Rubin} [Invent. Math. 89, 511--526 (1987; Zbl 0628.12007)]. This notwithstanding, the theory presented in the paper under review is hoped to be applicable in a wider setting and as part of a wider principle.