10
$\begingroup$

Suppose that $\Gamma$ is a finitely generated nonsolvable subgroup of $GL(n, R)$. Is it in the literature that $\Gamma$ has a nonsolvable finite quotient? I know how to prove it (the hardest ingredient is due to Nori), but would prefer to give a reference instead of a proof.

$\endgroup$
0

1 Answer 1

7
$\begingroup$

It is a straightforward consequence of the combination of two old results of Malcev and most likely Malcev was aware of this consequence.

1) The first is the well-known Malcev's residual finiteness result (every finitely generated linear group over, say, a field, is residually finite), or rather the stronger result that follows from the proof. Indeed, the main ingredient is the lemma saying that every finitely generated domain is residually a finite field. As a corollary, we have that for every $n$, every field $K$, every finitely generated subgroup of $\mathrm{GL}_n(K)$ is residually a subgroup of $\mathrm{GL}_n(F)$ where $F$ ranges over finite fields (with the same $n$).

2) The second is Malcev's result (1951, English translation 1956) that solvable subgroups of $\mathrm{GL}_n(K)$ have solvability length $\le r_n$ for some $r_n$, independently of the field $K$. This can be found (Theorem 3.21) in the book of D. Robinson Finiteness conditions and generalized soluble groups I, Theorem 3.21. Actually, Malcev proved the stronger result that for some $s_n$ independent of the field, every solvable subgroup of $\mathrm{GL}_n(K)$ has a subgroup of index $\le s_n$ that is conjugate, in $\mathrm{GL}_n(\hat{K})$, to the upper triangular group ($\hat{K}$ denoting an algebraic closure).


3) The combination of these two facts entails the result for linear groups over fields. Now suppose we have a non-solvable subgroup $\Gamma$ of $\mathrm{GL}_n(R)$ where $R$ is an arbitrary (associative unital) commutative ring. Let $N$ be the nilradical of $R$. It is no restriction to assume that $R$ is finitely generated. So the kernel of $\mathrm{GL}_n(R)\to\mathrm{GL}_n(R/N)$ is nilpotent; thus the image of $\Gamma$ in $\mathrm{GL}_n(R/N)$ is not solvable. Thus, we can suppose that $R$ is reduced (i.e., $N=0$). So $\{0\}$ is the intersection of the finitely many minimal prime ideals $P_i$ of $R$. It follows that the image of $\Gamma$ in at least one of the $\mathrm{GL}_n(R/P_i)$ is non-solvable. This reduces to the case of a field.

$\endgroup$
2
  • $\begingroup$ Yves, this argument is indeed simpler than the one I originally had in mind. I was using Nori's theorem that for infinitely many finite fields $F_p$ the group $\Gamma$ maps to a subgroup between $G(F_p)$ and $G(F_p)^+$, where $G$ is the group scheme of the Zariski closure of $\Gamma$; provided that $\Gamma< GL(N, {\mathbb Q})$. $\endgroup$
    – Misha
    Feb 5, 2016 at 1:36
  • $\begingroup$ Just a historical note: The upper bound on the derived length in Part (2) of Yves' answer is due to H.Zassenhaus, 1938. (It follows from Malcev's theorem too, of course.) $\endgroup$
    – Misha
    Sep 10, 2017 at 2:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.