Super-approximation. II: The \(p\)-adic case and the case of bounded powers of square-free integers (Q2002838)

A family of Cayley graphs \(\mathrm{Cay}(G_n,S_n)\) forms an expander family if there is a constant \(c>0\) such that the difference between the largest and the next eigenvalues of the adjacency matrix of \(\mathrm{Cay}(G_n,S_n)\) is at least \(c\) for almost all \(n\). Margulis proved long back that the family of Cayley graphs of finite quotients of a finitely generated group with Kazhdan's property T is a family of expanders. Two important examples of expander families due to Bourgain and Gamburd are the following. They showed that for a fixed prime \(p\), and a fixed finite, symmetric subset \(S\) of \(\mathrm{SL}(2,\mathbb{Z})\) generating a Zariski-dense subgroup, the family of graphs \(\mathrm{Cay}(\mathrm{SL}(2,\mathbb{Z}/p^n \mathbb{Z}), S\bmod p^n)\) forms an expander family. They also showed that the family \(\mathrm{Cay}(\mathrm{SL}(2,\mathbb{Z}/p \mathbb{Z}), S\bmod p)\) as the prime \(p\) varies is an expander family. Now, since strong approximation of a Zariski dense subgroup \(\Gamma\) of \(\mathrm{SL}(2, \mathbb{Z})\) amounts to its surjectivity onto \(\mathrm{SL}(2,\mathbb{Z}/p \mathbb{Z})\) for almost all primes \(p\), the property of being an expander family has come to be called by some mathematicians as superstrong approximation or super-approximation by the present author. The interest in expander graphs is due to the fact that they are simultaneously sparse as well as highly connected. In Part I of this paper [Int. Math. Res. Not. 2017, No. 23, 7190--7263 (2017; Zbl 1405.22025)], the author had proved the following theorem: Let \(S\) be a finite, symmetric subset of \(\mathrm{GL}(N,k)\), where \(k\) is a number field. Consider the set \(C\) of finite places \(P\) of \(k\) such that \(S \subset \mathrm{GL}(N,O_P)\). Let \(\Gamma\) be the subgroup generated by \(S\). Assume that the Zariski closure of \(\Gamma\) is semisimple. Then, the family of graphs \(\mathrm{Cay}(\pi_{P^n}(\Gamma), \pi_{p^n}(S))\) as \(P\) varies over \(C\) and \(n\) varies over positive integers, forms an expander family. Here \(\pi_{P^n}\) is the natural map \(\mathrm{GL}(N,O_P) \rightarrow \mathrm{GL}(N,O_p/P^nO_P)\). In the present paper, the author proves several results including a generalization of the above theorem where semisimplicity is replaced by the weaker assumption of perfectness. More precisely, he proves: Let \(S\) be a finite, symmetric subset of \(\mathrm{GL}(N_0,\mathbb{Z}[1/q_0])\), and let \(M_0\) be a fixed positive integer. Assume that the group \(\Gamma\) generated by \(S\) is infinite. Consider the families of graphs \(\mathrm{Cay}(\pi_{n}(\Gamma), \pi_{n}(S))\) where \(n\) varies either over prime powers of primes not dividing \(q_0\) or over positive integers whose \(p\)-adic valuation is at the most \(M_0\) for every prime \(p\). Then, the families are expanders if and only if the connected component of the Zariski closure of \(\Gamma\) is a perfect group. Unlike earlier works, the author needs to use new ideas and a combination of techniques from \(p\)-adic analytic pro-\(p\) groups like Howe's Kirillov theory of compact \(p\)-adic groups, \(p\)-adic analysis and noncommutative algebra. In particular, he uses Bruhat-Tits theory also. The author's result here is the first super-approximation theorem for powers of primes beyond the semisimple case. It should be noted that the main result is strong because it is a necessary and sufficient condition. There are many subsidiary results and applications to the largeness of the images of \(l\)-adic Galois representations in this rich paper.