Aspects of Iwasawa theory over function fields (Q2197706)

In this chapter 7 of the book, the authors present a detailed survey of ``aspects of Iwasawa theory over function fields'', but not including some new developments impulsed by the work of L. Taelman around 2012. Throughout $F$ will be a global function field, with characteristic $p$ and constant field $\mathbb{F}_F$. The discussion will be limited to abelian Galois extensions of $F$. In this setting, local groups of units are $\mathbb{Z}_p$-modules of infinite rank. In particular, the natural analogue of the (cyclotomic) $\mathbb{Z}_p$-extension of $\mathbb{Q}$ is the maximal $p$-adic abelian extension $\mathcal{F}/F$ unramified outside a fixed place, with Galois group $\Gamma\cong\mathbb{Z}_p^N$, so that the Iwasawa algebra $\Lambda=\mathbb{Z}_p[[\Gamma]]$ is not Noetherian. The present article is divided into two great sections: 1) In the first section, from a ``motivic'' perspective, the authors start from elliptic curves and abelian varieties over $F$ to study problems analogous to the number theoretic ones. The statement on the algebraic side of an Iwasawa Main Conjecture (IMC) requires the construction of an adequate $\Lambda$-torsion module and an associated (algebraic) $p$-adic $L$-function. To get around the non noetherianity of $\Lambda$, the authors propose to see it as an inverse limit of noetherian Iwasawa algebras $\Lambda_d$ attached to intermediary $\mathbb{Z}_p^d$-extensions. More precisely, let $A/F$ be an abelian variety, $A[p^n]$ the group scheme of $p^n$-torsion points and $A[p^\infty]=\lim_\to A[p^n]$. For any finite algebraic extension $L/F$, denote by $\text{Sel}_A(L)_p$ the $p$-part of the Selmer group of $A$ over $L$, defined in the usual way via flat cohomology of group schemes. Given a $\mathbb{Z}_p^d$-extension $\mathcal{F}_d/F$ (for $d<\infty$) with Galois group $\Gamma_d$ and Iwasawa algebra $\Lambda_d$, the Pontrjagin dual $\mathcal{S}(\mathcal{F}_d)$ of $\text{Sel}_A(\mathcal{F}_d)_p$ is a noetherian (and some times torsion) $\Lambda_d$-module, whose structure can be described using the following control theorem: ``For any finite subextension $L/F$ of $\mathcal{F}_d$, let $a_L:\text{Sel}_A(L)_p\to\text{Sel}_A(\mathcal{F}_d)_p^{\Gamma_L}$ be the natural restriction map. Assume that $\mathcal{F}_d/F$ is unramified outside a finite set of places of $F$ and that $A$ has good ordinary or split multiplicative reduction at all ramified places. Then $\text{Ker }a_L$ is finite (of bounded order if $d=1$) and $\text{Coker }a_L$ is a cofinitely generated $\mathbb{Z}_p$-module. Moreover if all places of bad reduction for $A$ are unramified in $\mathcal{F}_d/F$, then $\text{Coker }a_L$ is finite as well (of bounded order if $d=1$) [\textit{K.-S. Tan}, Trans. Am. Math. Soc. 362, No. 8, 4433--4450 (2010; Zbl 1239.11122)]. In this setting, $\mathcal{S}(\mathcal{F}_d)$ is $\Lambda_d$-noetherian; moreover, if all ramified primes are of good reduction and $\text{Sel}_A(F)_p$ is finite, then $\mathcal{S}(\mathcal{F}_d)$ is $\Lambda_d$-torsion. One would like to define the algebraic $L_p$-function attached to the Pontrjagin dual $\mathcal{S}(\mathcal{F})$ of $\text{Sel}_A(F)_p$ as being simply an inverse limit of characteristic ideals living in $\Lambda_d$, but the inverse limit of pseudo-null modules is not necessarily pseudo-null. Assuming that $A[p^\infty](\mathcal{F})=0$ or that $\text{Fitt}_{\Lambda(\mathcal{F}_d)}(\mathcal{S}(\mathcal{F}_d))$ is principal for all $\mathcal{F}_d$, the authors propose instead the candidacy of the pro-Fitting ideal $\widetilde{\text{Fitt}}_\Lambda\mathcal{S}(\mathcal{F}):=\underset{\leftarrow}{\lim}(\pi_d)^{-1}(\text{Fitt}_{\Lambda(\mathcal{F}_d)}(\mathcal{S}(\mathcal{F}_d))$, where $\pi_d$ is the natural projection $\Lambda\to\Lambda(\mathcal{F}_d)$. In the case where $A$ is an elliptic curve, the first hypothesis is satisfied if $j(A)\notin F^{\ast^p}$, and the second is satisfied in general by elementary $\Lambda(\mathcal{F}_d)$-modules or by modules having a presentation with the same number of generators and relations. On the analytic side, because of the complex Hasse-Weil $L$-function $L(A/F,s)$, the problem becomes to relate it to some element in an Iwasawa algebra. The authors briefly sketch how this can be done for so called ``modular abelian varieties of $\text{GL}_2$-type''. They also recall the existence of various different approaches by Longhi, Pal, Tan etc. 2) The second section is devoted to a recap of what is known about the IMC for class groups in the framework of this chapter. Recall that a version of the IMC overa global function field $F$ was proved by \textit{R. Crew} [Invent. Math. 88, 395--403 (1987; Zbl 0615.14013)], in a geometric setting and using sophisticated (étale and crystalline) cohomological machinery. Here is a brief account of a more algebraically oriented (and with different cohomological tools) approach by Kueh-Lai-Tan and D. Burns. In view of applications to Drinfeld-Hayes cyclotomic extensions (see below), the authors restrict themselves to the particularly simple situation when the $\mathbb{Z}_p^N$-extension $\mathcal{F}/F$ is unramified outside a single prime $\mathcal{P}$ which is totally ramified in $\mathcal{F}/F$ (see the end of the introduction). For any finite extension $L/F$, let $\mathcal{M}(L)$ be the $p$-adic completion of the group of divisor classes of $L$, and introduce the ``Picard group'' $\mathcal{A}(L)$ as the kernel of the ``degree'' map $\mathcal{M}(L)\to\mathbb{Z}_p$. For a given finite $d$, it is well known that $\mathcal{A}(\mathcal{F}_d)$ is a noetherian $\Lambda(\mathcal{F}_d)$-module. Under a certain technical condition on the filration $(\mathcal{F}_d)$, the inverse images of the characteristic series $Ch_{\Lambda(\mathcal{F}_d)}(\mathcal|{A}(\mathcal{F}_d))$ w.r.t. the canonical projections $\pi_d:\Lambda\to\Lambda(\mathcal{F}_d)$ constitute an inverse system, so one can define a pro-characteristic ideal $\widetilde{Ch}_\Lambda(\mathcal{A}(\mathcal{F})):=\underset{\leftarrow}{\lim}(\pi_d)^{-1}(Ch_{\Lambda(\mathcal{F}_d)}(\mathcal{A}(\mathcal{F}_d)))$. It remains to mimick the construction of an inverse system of Stickelberger elements, with inverse limit $\vartheta_{\mathcal{F}/F,P}$, using the techniques of Weil-Deligne-Tate exposed in \textit{J. Tate}'s book [Les conjectures de Stark sur les fonctions \(L\) d'Artin en \(s=0\). Basel: Birkhäuser/Springer (1984; Zbl 0545.12009)], to get the desired IMC, namely $\widetilde{Ch}_\Lambda(\mathcal{A}(\mathcal{F}))=(\vartheta_{\mathcal{F}/F,P})$. The final subsection deals with the ``cyclotomic'' theory of function fields obtained via Drinfeld-Hayes modules. For simplification (but leaving intact the main aspects of the theory), the authors limit themselves to a base field $F=\mathbb{F}_q(T)$, with the usual place at infinity $\infty$, so that the ring of elements regular outside $\infty$ is $A:=\mathbb{F}_[T]$. In this case the only Drinfeld-Hayes module is the Carlitz module $\phi:A\to A\{\tau\}$, $T\to\phi_T:=T+\tau$, where $\tau$ is the operator $x\to x^q$ and, if $R$ is an $\mathbb{F}_p$-algebra, $R\{\tau\}$ is the ring of skew polynomials with coefficients in $R$, with multiplication given by composition. Fix a prime $\mathcal{P}=(\pi)$ in $R$ (with a monic $\pi$) and consider the ring of local integers $O_{\mathcal{P}}$ in the $\mathcal{P}$-adic completion $A_{\mathcal{P}}$. Put $F_n=F(\phi[\mathcal{P}^n])$ and $K_n=F_{\mathcal{P}}(\phi[\mathcal{P}_n])$. Since $\phi$ has rank 1, the Tate module $T_{\mathcal{P}}(\phi)$ is a free $A_{\mathcal{P}}$-module of rank 1. Fix a generator $\omega=(\omega_n)$, i.e. a sequence $\{\omega_n\}$ s.t. $\phi_{\pi^n}(\omega_n)=0\neq\phi_{\pi^{n-1}}(\omega_n)$ and $\phi_\pi(\omega_{n+1})=\omega_n$. By definition $K_n=F_{\mathcal{P}}(\omega_n)$ and it plays the same role as $\mathbb{Q}(\mu_{p^n})$ in number theory. Let $O_n$ be the closure in $K_n$ of the ring of $A$-integers $B_n$ of $F_n$, and let $\mathcal{U}_n$ be the subgroup of 1-units in $O_n^\ast$. Then, denoting by $\widehat{(.)}$ the $p$-adic completion, $\widehat{\underset{\leftarrow}{\lim}K}_n^\ast=\omega^{\boldsymbol{Z}_p}\times\underset{\leftarrow}{\lim}\mathcal{U}_n$ w.r.t. norms. Define the group of ``cyclotomic'' units $\mathcal{C}_n$ as the intersection of $B_n^\ast$ with the subgroup of $F_n^\ast$ generated by $\sigma(\omega_n)$, $\sigma\in\text{Gal}(F_n/F)$, and let resp. $\mathcal{C}_n$ and $\mathcal{C}_n^1$ be the closure of $\mathcal{C}_n\cap O^\ast_n$ and $C_n\cap\mathcal{U}_n$. When taking inverse limits w.r.t. norms along the tower $(F_n)$, the analogues in Drinfeld-Hayes' cyclotomic theory of the Coleman power series and of the Coates-Wiles homomorphisms in the number field case produce a Galois theoretic interpretation of the quotient $\mathcal{U}_\infty/\mathcal{C}_\infty^1$. More precisely, let $F_n^+$ be the fixed field in $F_n$ of the inertia group $I_\infty(F_n/F)$, $\mathcal{X}_n=$ the Galois group over $F_n^+$ of the maximal abelian extension of $F_N^+$ unramified outside $\mathcal{P}$ and totally split at $\infty$. Then $\mathcal{U}_\infty/\mathcal{C}_\infty^1\cong \mathcal{X}_\infty:=\underset{\leftarrow}{\lim}\mathcal{X}_n$. In general one expects an IMC under the form of an equality of (yet to be defined) characteristic ideals of $\mathcal{X}_\infty$ and of $\mathcal{U}_\infty/\mathcal{C}_\infty$. For the entire collection see [Zbl 1441.14003].