The first level of \(\mathbb{Z}_p\)-extensions and compatibility of heuristics (Q6923994)

This paper compares the well-known Cohen-Lenstra heuristics (CL) with more recent heuristics which are due to Ellenberg, Jain and Venkatesh [\textit{J. S. Ellenberg} et al., Commun. Pure Appl. Math. 64, No. 9, 1243--1262 (2011; Zbl 1231.11127)] (EJV), and are formulated in an Iwasawa-theoretic context. More concretely, \(K\) is taken to be an imaginary quadratic field, in which a chosen odd prime \(p\) does not split. Then CL makes predictions how likely a given abelian \(p\)-group \(A\) is to appear as Cl\((K)\{p\}\), and in particular how likely it is that Cl\((K)\{p\}\) is cyclic as a group. On the other hand EJV makes a prediction how likely it is that \(\lambda=1\), where \(\lambda\) is the Iwasawa invariant attached to the \(p\)-cyclotomic extension \(K_\infty = \bigcup_n K_n\) of \(K\). The authors establish a compatibility result for the two heuristics, based on the fact that one can read off whether \(\lambda=1\) or not, just considering \(A_0\) and \(A_1\). (Here \(A_n = \text{Cl}(K_n)\{p\}\).) One should note here that in this paper ``cyclic module'' means ``nonzero module which can be generated by one element''. There is also an equivariant version of this compatibility, with weight factors, taking into account the structure as modules over \(R=\mathbb Z_p[G]\), where \(G=\langle\sigma\rangle\) is the Galois group of the first layer \(K_1/K_0\).\N\NThere are two essential points which make this possible. First, under the assumption that \(A_0\) is cyclic one can show that \(A_1\) is isomorphic to \(R/J\) for some principal ideal \(J\). Second, these principal ideals are given by generators, and these can in turn be described by certain pairs \((\pi^r, jp^m)\), where \(\pi=\zeta_p-1\) is a parameter of the DVR \(\mathbb Z_p[\zeta_p]\) and \(1\le j \le p-1\). From these pairs one can calculate the structure (in other words, the invariants) of the abelian \(p\)-group \(A_1\), and this information almost determines the pair. This involves fairly long and involved calculations involving matrices.\N\NThe paper also contains computational results, which seem to confirm the heuristics. They were done for \(p=3\) and \(p=5\); for \(p=3\) the calculations treat the cases \(A_0\cong \mathbb Z/p\) or \(A_0 \cong \mathbb Z/p^2\), and for understandable reasons for \(p=5\) only the former is done. The paper closes with an interesting short section on the function field case.