Artin \(L\)-functions of almost monomial Galois groups (Q776162)

The paper under review investigates certain properties of Artin \(L\)-functions associated to almost monomial Galois extensions. In particular, it proves that all these Artin \(L\)-functions are holomorphic at any given point \(s_0 \in \mathbb{C}\) which is not a common zero of two \(L\)-functions associated to two different irreducible characters. The last section of the paper is then devoted to the study of certain properties of almost monomial Galois groups, and provides a list of examples. Let us be more specific, and give some background to the interested reader. Let \(K \subseteq L\) be a finite extension of number fields, and let \(G := \operatorname{Gal}(L/K)\) be its Galois group. Recall that a function \(\chi \colon G \to \mathbb{C}\) is called a character if there exists a linear representation \(\rho \colon G \to \operatorname{GL}_n(\mathbb{C})\) such that \(\chi(g) = \operatorname{tr}(\rho(g))\) for any \(g \in G\), where \(\operatorname{tr} \colon \operatorname{GL}_n(\mathbb{C}) \to \mathbb{C}\) denotes the trace map. To any such character \(\chi\) one can associate a formal Dirichlet series \(L(s,\chi)\), defined by the identity \[ \log(L(s,\chi)) := \sum_\mathfrak{p} \sum_{n = 1}^{+\infty} \frac{\chi(\mathfrak{p}^n)}{n \cdot \operatorname{N}(\mathfrak{p})^{n s}}, \] where the first sum runs over all the prime ideals \(\mathfrak{p} \subseteq \mathcal{O}_K\) and \(\operatorname{N}(\mathfrak{p}) := \lvert \mathcal{O}_K/\mathfrak{p} \rvert\). Moreover, \(\chi(\mathfrak{p}^n) \in \mathbb{C}\) is defined as \[ \chi(\mathfrak{p}^n) := \frac{1}{\lvert I_{\mathfrak{P}} \rvert} \sum_{\tau \in \sigma_\mathfrak{P}^n \cdot I_{\mathfrak{P}}} \chi(\tau), \] where \(\mathfrak{P} \subseteq \mathcal{O}_L\) is any prime ideal such that \(\mathfrak{P} \cap \mathcal{O}_K = \mathfrak{p}\), to which one associates a decomposition group \(G_\mathfrak{P} := \{ g \in G \mid g(\mathfrak{P}) = \mathfrak{P} \}\), an inertia subgroup \(I_\mathfrak{P} := \{ g \in G_{\mathfrak{P}} \mid g(x) \equiv x \, (\mathfrak{P}), \ \forall{x} \in \mathcal{O}_L \}\) and a Frobenius element \(\sigma_\mathfrak{P} \in G_{\mathfrak{P}}\) which generates the cyclic quotient \(G_\mathfrak{P}/I_\mathfrak{P}\). Note that the definition of \(\chi(\mathfrak{p}^n)\) does not depend on the choice of the prime ideal \(\mathfrak{P} \subseteq \mathcal{O}_L\), because the decomposition group, the inertia group and the Frobeinus element change only by conjugation if one chooses another prime of \(\mathcal{O}_L\) lying above \(\mathfrak{p}\). The formal Dirichlet series \(L(s,\chi)\) converges for \(\Re(s) > 1\), and Brauer's induction theorem implies that it can be meromorphically continued to a function on the whole complex plane. This function, the Artin \(L\)-function associated to the character \(\chi\), is still denoted by \(L(s,\chi)\), and admits a functional equation relating \(L(s,\chi)\) to \(L(1-s,\chi)\). Since \(L(s,\alpha+\beta) = L(s,\alpha) + L(s,\beta)\) for any pair of characters \(\alpha, \beta \colon G \to \mathbb{C}\), the set \[ \operatorname{Art}(L/K) := \{L(s,\chi) \mid \chi \colon \operatorname{Gal}(L/K) \to \mathbb{C} \} \] is a semigroup under multiplication. This semigroup is clearly generated by the Artin \(L\)-functions associated to the irreducible characters \(\chi_1,\dots,\chi_r \colon G \to \mathbb{C}\), where \(r\) is the number of conjugacy classes of \(G\). Moreover, \textit{E. Artin} himself proved in Satz 5 of [Abh. Math. Semin. Univ. Hamb. 3, 89--108 (1923; JFM 49.0123.01)] that the group \(\operatorname{Art}(L/K)\) is factorial, which means that the \(L\)-functions \(L(s,\chi_1),\dots,L(s,\chi_r)\) are multiplicatively independent. Let now \(s_0 \in \mathbb{C}\), and define \(\operatorname{Hol}(s_0;L/K) \subseteq \operatorname{Art}(L/K)\) to be the subgroup of all Artin \(L\)-functions \(L(s,\chi) \in \operatorname{Art}(L/K)\) which are holomorphic at \(s_0\). Then Artin's holomorphy conjecture predicts that \(\operatorname{Hol}(s_0;L/K) = \operatorname{Art}(L/K)\) for all \(s_0 \in \mathbb{C}\). This conjecture is known to hold (without any additional assumptions) only in a few cases, for example when \(K \subseteq L\) is abelian by work of \textit{E.Artin} himself [Abh. Math. Semin. Univ. Hamb. 5, 353--363 (1927; JFM 53.0144.04)], and for some small non-abelian extensions by work of \textit{R. Langlands}, as explained in Theorem 3.3 of [Base change for \(\operatorname{GL}(2)\). Princeton University Press (1980; Zbl 0444.22007)], and by the work of \textit{J. Tunnell} [Bull. Am. Math. Soc., New Ser. 5, 173--175 (1981; Zbl 0475.12016)]. Moreover, Artin's conjecture would be true if one could relate irreducible Artin \(L\)-functions to \(L\)-functions of cuspidal automorphic representations, as predicted by \textit{R. Langlands} in [Lect. Notes Math. 170, 18-61 (1970; Zbl 0225.14022)]. We refer the reader interested in these works, as well as in more recent developments, to the survey by \textit{D. Prasad} and \textit{C. S. Yogananda} [Number theory. Basel: Birkhäuser. 301--314 (2000; Zbl 0982.11063)]. The second author of the paper under review proposed another way of attacking Artin's conjecture in [Monatsh. Math. 186, No. 4, 679--683 (2018; Zbl 1404.11130)]. More precisely, he conjectures that, for any finite extension of number fields \(K \subseteq L\) and any complex number \(s_0 \in \mathbb{C}\), one has \(\operatorname{Hol}(s_0;L/K) = \operatorname{Art}(L/K)\) if and only if \(\operatorname{Hol}(s_0;L/K)\) is a factorial semigroup. If this new conjecture was true, one could try to prove Artin's conjecture for non-abelian extensions without using the automorphic techniques mentioned above. In the same paper mentioned above, the second author of the paper under review proved that his conjecture is true for a fixed \(s_0 \in \mathbb{C}\) if there exists \(k \in \{1,\dots,r\}\) such that \(L(s_0,\chi_k) = 0\) and \(d_k \leq 2\), where \(d_1, \dots, d_r \in \mathbb{N}\) denote the powers appearing in the factorization \[ \zeta_L(s) = L(s,\chi_1)^{d_1} \cdots L(s,\chi_r)^{d_r} \] of the Dedekind \(\zeta\)-function of \(L\) in terms of irreducible Artin \(L\)-functions. Moreover, the second author of the paper under review proved in [loc. cit.] that his conjecture holds if \(\operatorname{Gal}(L/K)\) is an almost monomial group, and the main result of the paper under review is that, if \(\operatorname{Gal}(L/K)\) is an almost monomial group and \(\operatorname{Hol}(s_0;L/K) \neq \operatorname{Art}(L/K)\) (which would contradict Artin's conjecture) then \(L(s_0,\chi_k) = L(s_0,\chi_l) = 0\) for some \(1 \leq k < l \leq r\). Let us conclude this review by recalling the definition of almost monomial group, and by giving some examples of these objects. Let \(G\) be a finite group and \(\chi \colon G \to \mathbb{C}\) be a character. Then \(\chi\) is called monomial if there exists a subgroup \(H \subseteq G\) and a character \(\lambda \colon H \to \mathbb{C}\) which induces \(\chi\) and is linear, \textit{i.e.} such that \(\lambda(\operatorname{Id}_H) = 1\). Moreover, a finite group \(G\) is called monomial if all its irreducible characters are monomial. Then the work of Artin on abelian extensions mentioned above shows that Artin's conjecture holds for extensions \(K \subseteq L\) has monomial Galois group. The notion of almost monomial group was defined in [loc. cit.] by the second author of the paper under review as a generalisation of the notion of monomial group. More precisely, a finite group \(G\) is almost monomial if for every pair of distinct irreducible, linear, complex-valued representations \(\rho_1, \rho_2\) of \(G\) there exists a subgroup \(H \subseteq G\) and a one-dimensional representation \(\lambda \colon H \to \mathbb{C}^\times\) such that the induced representation \(\operatorname{Ind}_H^G(\lambda)\) contains \(\rho_1\) but not \(\rho_2\) as an irreducible factor. The third section of the paper under review proves that the symmetric group \(S_n\) is almost monomial for any \(n \geq 1\), which might be surprising if one compares it to Theorem 1.1 of the preprint [``Squarefree values of polynomial discriminants I'', arXiv:1611.09806] by \textit{M. Bhargava, A. Shankar, X. Wang}, which shows that a positive density of Galois extensions of \(\mathbb{Q}\) have Galois group isomorphic to \(S_n\), and hence almost monomial. However, the alternating group \(A_n\) is not almost monomial for some \(n \in \mathbb{N}\), for example \(n = 6,7,10,11,12\), many sporadic groups (like the Mathieu groups, Janko group \(J_2\) and the Higman-Sims group \(\mathrm{HS}\)) are not almost monomial, and also many of the general linear groups \(\operatorname{GL}_n(\mathbb{F}_q)\) and special linear groups \(\operatorname{SL}_n(\mathbb{F}_q)\) associated to finite fields are not almost monomial. Finally, the last results of the paper under review prove that any quotient of an almost monomial group is almost monomial, and that a product of groups is almost monomial if and only if each of the factors is.