Linking numbers and the tame Fontaine-Mazur conjecture (Q742919)

Let \(p\) be a prime number, \(K\) a number field, \(S\) a finite set of primes of \(K\) with residual characteristics distinct from \(p\). Denote by \(\Gamma_{S,K}\) the Galois group of the maximal algebraic extension of \(K\) unramified outside \(S\), and by \(G_{S }=G_{S,K}\) the maximal pro-\(p\)-quotient of \(\Gamma_{S,K}\). The tame Fontaine-Mazur conjecture asserts that every continuous homomorphism \(\rho : \Gamma_{S,K} \rightarrow \mathrm{GL}_n (\mathbb Z_p)\) must have finite image. Let \(\mathrm{GL}_n^{(1)}(\mathbb Z_p)\) be the subgroup of matrices of \(\mathrm{GL}_n (\mathbb Z_p)\) which are congruent to \(1\) mod \(p\). It is easily seen that the tame Fontaine-Mazur conjecture is equivalent to the following Conjecture 1.1: Every continuous homomorphism \(\rho : G_{S,K} \rightarrow \mathrm{GL}_n^{(1)}(\mathbb Z_p)\) must be trivial. In this paper the author proves some cases of Conjecture 1.1 when \(K = \mathbb Q\), \(p\) is odd and the primes \(q_i\) in \(S\) are congruent to \(1 \mod\, p\) (this is not a restriction), but not congruent to \(1 \mod\, p^2\) (this means that the abelianized group \(G_S^{\mathrm{ab}}\) is elementary). The main tool is the so called linking \(\mathbb F_p\)-Lie algebra \(\mathfrak l_S\) of \(S\), defined by generators \(\xi_1,\dots, \xi_d\) and relations \(\sigma_1,\dots, \sigma_d\), with \(\sigma_i = c_i \xi_i + \sum_{i\neq j} l_{ij} [\xi_i, \xi_j],\) where \(c_i = (q_i {- 1})/p\) and the linking number \(l_{ij}\) of \((q_i , q_j)\) is such that \(q_i \equiv {g_j}^{-l_{ij}} \mod q_j\), \(g_j\) being a primitive root mod \(q_j\). The main point is that there exists a map \(\ell : \Hom_{\mathrm{cont}} (G_{S}, \mathrm{GL}_n^{(1)}(\mathbb Z_p)) \rightarrow \Hom (\mathfrak l_S , {gl}_n(\mathbb F_p))\) such that \(\rho = 1\) if and only if \(\ell(\rho) = 0\). One can thus obtain sufficient conditions for Conjecture 1.1 to hold; for example: {\parindent=6mm \begin{itemize}\item[-] If the cup-product \(H^1(G_{S}, \mathbb F_p) \times H^1(G_{S}, \mathbb F_p)\rightarrow H^2(G_{S}, \mathbb F_{p})\) is null, then Conjecture 1.1 is valid for all \(n\). \item[-] If every \(k\)-dimensional representation of \(\mathfrak l_S\) is trivial (let us call \(FM(k)\) this condition), then Conjecture 1.1 is valid for \(n = k\). \end{itemize}} The author gives necessary and sufficient criteria in terms of linking numbers for condition \(FM(k)\) to hold or not when \(|S|< 3\) and \(k < p\). He concludes by related considerations on mild and Fontaine-Mazur pro-\(p\)-groups: {\parindent=6mm \begin{itemize}\item[-] There exists a finite set \(T\) containing \(S\) and consisting of primes \(q\) congruent to \(1 \mod\, p\), not congruent to \(1 \mod\, p^2\), such that \(G_{T }\) is mild (in the sense of [the author, J. Reine Angew. Math. 596, 155--182 (2006; Zbl 1122.11076)] and, if \(n < p\), the Lie algebra \(\mathfrak l_T\) has property \(FM(n)\) if \(\mathfrak l_S\) does. \item[-] A pro-\(p\)-group \(G\) is labelled Fontaine-Mazur if every continuous homomorphism \(G\rightarrow \mathrm{GL}_n (\mathbb Z_p)\) is finite. The author constructs by generators and relations a pro-\(p\)-group \(G\) such that every continuous homomorphism \(G\rightarrow \mathrm{GL}_n^{(1)}(\mathbb Z_p)\) is trivial if \(n < p\). \end{itemize}}