Density of classical points in eigenvarieties (Q2428830)

The paper under review addresses the question of Zariski-density of the set \(C\) of classical points in the eigenvariety \(\mathcal{E}^{\text{rig}}/\mathbb{Q}_p\) (of some tame level) parametrizing \(p\)-adic automorphic forms for \(\text{GL}_1\) over a number field \(K\). The elements of \(C\) correspond to the algebraic Grössencharacters of \(K\). Loeffler shows that if \(K\) is not totally real and if \(K\) does not contain a CM subfield, then \(C\) is not Zariski-dense in the rigid space \(\mathcal{E}^{\text{rig}}\), but rather contained in a one-dimensional closed rigid subspace. On the other hand, he shows that under these assumptions, \(\mathcal{E}^{\text{rig}}\) has a natural formal model \(\mathcal{E}\) of formally finite type over \(\mathbb{Z}_p\) such that \(C\) lies Zariski-dense with respect to the corresponding \textit{formal Zariski topology} on \(\mathcal{E}^{\text{rig}}\) (cf.\ below). In particular, a rigid-analytic function on \(\mathcal{E}^{\text{rig}}\) is \textit{not} determined by its values in \(C\), but a \textit{bounded} rigid-analytic function on \(\mathcal{E}^{\text{rig}}\) \textit{is} determined by these values. Loeffler also sketches the situation for the \(\text{GL}_2\) case, pointing out that the arguments given in [\textit{F. Calegari} and \textit{B. Mazur}, J. Inst. Math. Jussieu, 8, No. 1, 99--177 (2009; Zbl 1211.11065)] are insufficient to rule out the existence of irreducible components of Hida's ordinary eigenvariety (cf. [\textit{H. Hida}, Ann. Inst. Fourier 44, No. 5, 1289--1322 (1994; Zbl 0819.11017)]) with only finitely many classical points; the existence of such components was claimed in ([Zbl 1211.11065], Theorem 8.9). Let us describe the content of the paper in greater detail. Section 1 provides an introduction. In Section 2, the author gives, for affine formal schemes \(\text{Spf}\,A\) of formally finite type over \(\mathbb{Z}_p\), a definition of a \textit{formal Zariski topology} on the rigid generic fiber \((\text{Spf}\,A)^{\text{rig}}\) of \(\text{Spf}\,A\): it corresponds, under the natural bijection \((\text{Spf}\,A)^{\text{rig}}\cong\text{Max}\,A[1/p]\), to the Zariski topology on \(\text{Max}\,A[1/p]\). Using the language of uniformly rigid geometry as developed by the reviewer in [\textit{C. Kappen}, Algebra Number Theory 6, No. 2, 341--388 (2012; Zbl 1254.14025)], the formal Zariski topology of \((\text{Spf}\,A)^{\text{rig}}\) could also be described as the Zariski topology on the uniformly rigid generic fiber \((\text{Spf}\,A)^{\text{urig}}\) of \(\text{Spf}\,A\); this point of view has the advantage of generalizing to non-affine situations. In Section 3, Loeffler recalls the definitions of the weight space and of the eigenvariety for \(\text{GL}_1/K\). He first shows that for any compact abelian \(p\)-adic analytic group \(G\), the functor mapping an Artinian local \(\mathbb{Z}_p\)-algebra \(A\) to the set of continuous group homomorphisms \(G\rightarrow A^\times\) is pro-represented by the formal spectrum of the Iwasawa algebra \(\mathbb{Z}_p[[G]]\) of \(G\), which is of formally finite type over \(\mathbb{Z}_p\). He then fixes a number field \(K\) and first applies this result to \(G=\mathcal{O}_{K,p}^\times:=(\mathcal{O}_K\otimes_\mathbb{Z}{\mathbb{Z}_p})^\times\), thus obtaining an affine formal \(\mathbb{Z}_p\)-scheme \(\mathcal{W}\) of formally finite type, the \textit{formal weight space}. The points of \(\mathcal{W}^{\text{rig}}\) correspond to the \textit{\(p\)-adic weights} of \(K\), i.e.\ to the continuous homomorphisms \(\mathcal{O}_{K,p}^\times\rightarrow E^\times\), where \(E\) varies among the finite extensions of \(\mathbb{Q}_p\). Loeffler then fixes a tame level structure, i.e.\ an open compact subgroup \[ U\subseteq(\mathbb{A}^{p,\infty}_K)^\times\;, \] and he considers the formal eigenvariety \(\mathcal{E}(U)\) for \(\text{GL}_1/K\) of tame level \(U\) as it has been defined in [\textit{K. Buzzard}, in: Modular curves and abelian varieties. Based on lectures of the conference, Bellaterra, Barcelona, July 15--18, 2002. Basel: Birkhäuser. Prog. Math. 224, 23--44 (2004; Zbl 1166.11322)]; that is, \(\mathcal{E}(U)\) is the formal spectrum of the Iwasawa algebra of the compact abelian \(p\)-adic analytic group \[ \mathbb{A}_K^\times/\overline{K^\times\cdot U\cdot K^\circ_\infty}\;. \] There is a natural morphism \(\mathcal{E}(U)\rightarrow\mathcal{W}\) which is finite and flat onto a closed formal subscheme \(\mathcal{W}(U)\) of \(\mathcal{W}\) which is also the formal spectrum of a suitable Iwasawa algebra. Section 4 is the heart of the paper. Here the author assumes that \(K\) contains no CM subfield, which implies that the infinity type of every algebraic Grössencharacter of \(K\) is parallel. Loeffler first investigates classical points on the weight space. If \(\kappa\in\mathcal{W}\) is locally algebraic, he writes \(c(\kappa)\) for the smallest integer \(r\geq 0\) such that the restriction of \(\kappa\) to \(1+p^r\mathcal{O}_{K,p}\) is algebraic. He shows that for every \(N\in\mathbb{N}\), there exists a one-dimensional closed formal subscheme of \(\mathcal{W}\) containing all \(\kappa\in\mathcal{W}(U)\) with \(c(\kappa)\leq N\). Loeffler now assumes in addition that \(K\) is not totally real; then \(\mathcal{W}(U)\) has dimension \(>1\), and the above result shows that the locally algebraic characters \(\kappa\) with \(c(\kappa)\leq N\) are not dense for the formal Zariski topology on \(\mathcal{W}(U)^{\text{rig}}\) that is given by \(\mathcal{W}(U)\). Now the author considers the set of \textit{all} locally algebraic characters in \(\mathcal{W}(U)^{\text{rig}}\). He shows that the Zariski-closure of this set in \(\mathcal{W}(U)^{\text{rig}}\) is a closed rigid subspace of dimension \(1\), while its closure in the formal Zariski topology is the whole space \(\mathcal{W}(U)^{\text{rig}}\). (The idea of the proof of the first statement is to choose a \(\mathbb{Z}_p\)-basis \(\{u_i\}\) for the torsion-free part of the norm-one-sphere in \(\mathcal{O}_{K,p}^\times\) and then to consider the rigid analytic functions \(\kappa\mapsto\log\kappa(u_i)\) on \(\mathcal{W}(U)^{\text{rig}}\); they cut out a closed rigid subspace of dimension one, and all locally algebraic points must lie on this subspace.) Finally, using the fact that the natural morphism \(\mathcal{E}(U)\rightarrow\mathcal{W}(U)\) is finite and flat, Loeffler lifts all these statements from \(\mathcal{W}(U)\) to \(\mathcal{E}(U)\). In Section 5, Loeffler sketches the situation for \(\text{GL}_2\), assuming that \(K\) is an imaginary quadratic field in which \(p\) splits; he fixes an integral ideal \(\mathfrak{N}\) of \(\mathcal{O}_K\) that is prime to \(p\). Here the role of \(\mathcal{W}(U)\) is played by an equidimensional formal subscheme \(\text{supp}\,\mathbb{H}\subseteq\mathcal{W}\) of dimension 1 whose description is less explicit than that of \(\mathcal{W}(U)\). Loeffler shows that for a suitable choice of \((K,\mathfrak{N},p)\), \(\text{supp}\,\mathbb{H}\) has irreducible components \(X\) of dimension \(1\) such that for any \(N\in\mathbb{N}\), the set of arithmetical weights (which are locally algebraic weights \(x\mapsto x^a\overline{x}^b\varepsilon(x)\) with \(a,b\geq 2\) and \(\varepsilon\) of finite order) in \(X\) is finite; he also shows (using the norm-one-sphere argument mentioned above) that the set of all arithmetical weights in \(X\) is not Zariski dense in \(X^{\text{rig}}\). Loeffler does not know whether the arithmetical weights in \(X\) lie dense for the formal Zariski topology on \(X^{\text{rig}}\), i.e.\ whether there exist infinite many arithmetical weights in \(X\). The problem is that while the norm-one-sphere argument shows that the locally parallel weights in \(\mathcal{W}\) form a closed \textit{rigid} subspace of \(\mathcal{W}^{\text{rig}}\), they do \textit{not} form a closed \textit{uniformly rigid} subspace in \(\mathcal{W}^{\text{urig}}\) (cf. the reviewer [Zbl 1254.14025], Section 3.1). Finally, Loeffler points out that the arguments of Theorem 7.1 in [Zbl 1211.11065] do not show that the Galois-theoretic deformation space constructed in [ibid.] has finitely many specializations with parallel Hodge-Tate weights, but rather the weaker statement that it has finitely many parallel weight specializations which are crystalline at \(\mathfrak{p}\) and \(\overline{\mathfrak{p}}\) or become so over any fixed extension of \(\mathbb{Q}_p\).