8
$\begingroup$

I know that there exists a nice presentation (generators and relations) of the general linear group over a finite field (by Steinberg, I think). Is there also a nice presentation of $GL(n,\mathbb{Z}/m\mathbb{Z})$ for an arbitrary integer $m$? And perhaps also for the symplectic group over $\mathbb{Z}/m\mathbb{Z}$?

I want to do some calculations in these groups with a computer and my first problem was to determine these groups. One solution (my current) is of course to determine both by brute force. But already for small m,n this takes too much time.

$\endgroup$
3
  • $\begingroup$ Any finite group has a finite presentation - just take generators all elements of the group, and relations from the multiplication table. GAP probably computes directly with matrices. Maybe you should clarify your question - what sort of presentation are you looking for? $\endgroup$
    – Ian Agol
    Nov 22, 2009 at 16:30
  • $\begingroup$ Well, good point. I'm looking of course for a presentation with less elements than the group order. Something that makes it possible (at least for small n and m) to determine/handle these groups with a computer without doing this by brute force. $\endgroup$
    – user717
    Nov 22, 2009 at 17:17
  • 1
    $\begingroup$ This comment is woefully late, but let me make it anyway for the benefit of future readers: the integer symplectic groups Sp(2n,Z) are generated by 2 matrices for all n. References are Stanek (Math Review MR0153748) for n>3 and Ishibashi (MR1367845) for n=2,3. In my experience of computer experiments, it seems to speed things up to have as small a generating set as possible. $\endgroup$
    – user5117
    May 10, 2011 at 14:11

2 Answers 2

9
$\begingroup$

Sure, you can certainly work something out if you like. It may well be that an off-the-rack answer to your question is in the literature somewhere, but let me say how I would go about getting there from here.

Note that $\textrm{GL}(n,\mathbb{Z}/m\mathbb{Z})$ is the product over all prime powers $p^a$ exactly dividing $m$ of $\textrm{GL}(n,\mathbb{Z}/p^a\mathbb{Z})$, so you're reduced to the case where $m$ is a prime power.

Second, there is an exact sequence $1 \to U_1 \to \textrm{GL}(n,\mathbb{Z}/p^a\mathbb{Z}) \to \textrm{GL}(n,\mathbb{Z}/p\mathbb{Z}) \to 1$ where I'll use $U_i$ to denote all mod $p^a$ matrices that are congruent to the identity modulo $p^i$. Given the Steinberg generators for $\textrm{GL}(n,\mathbb{Z}/p\mathbb{Z})$, you're reduced to giving generators/relations for $U_1$ (as well as working out how lifts of the Steinberg generators act by conjugation on the generators for $U_1$, as well as what elements of $U_1$ you get by applying the Steinberg relations to the lifts of your generators).

Note that $U_1$ is a big $p$-group, and each $U_i/U_{i+1}$ up to $i=a-1$ is just isomorphic to $\text{M}_n(\mathbb{Z}/p\mathbb{Z})$, so again you can produce generators and relations by devissage. In practice it shouldn't require so many generators as you get out of this. For instance for $n=2$, the group $U_1$ is generated by four elements: the upper-triangular unipotent matrix with $p$ in the upper-right, its transpose, and diagonal matrices with diagonal entries $c,1$ and $1,c$, where $c$ is a generator for the multiplicative group $1 + p\mathbb{Z}/p^a\mathbb{Z}$.

$\endgroup$
9
$\begingroup$

Two things. I'll talk about SL_n rather than GL_n, but that's just a technicality.

  1. There is a short exact sequence

$1 \rightarrow SL_n(\mathbb{Z},m) \rightarrow SL_n(\mathbb{Z}) \rightarrow SL_n(\mathbb{Z}/m\mathbb{Z}) \rightarrow 1,$

where $SL_n(\mathbb{Z},m)$ is the level $m$ congruence subgroup of $SL_n(\mathbb{Z})$. There is a nice presentation $\langle S|R \rangle$ for $SL_n(\mathbb{Z})$ due essentially to Magnus; it can be found in Milnor's book on algebraic K-theory (look for the chapter that calculates $K_2(\mathbb{Z})$). The generating set $S$ is just the set of elementary matrices. To get from there to $SL_n(\mathbb{Z}/m\mathbb{Z})$, you just need to add a normal generator for $SL_n(\mathbb{Z},m)$ to $R$. By the solution to the congruence subgroup problem for $SL_n(\mathbb{Z})$ (due to Mennicke, but you are better off looking at Bass-Milnor-Serre's paper), this congruence subgroup is normally generated by the mth power of a single elementary matrix.

A similar idea also works for the symplectic group. To find a presentation for $Sp_{2g}(\mathbb{Z})$, look at Theorem 9.2.13 of Hahn and O'Meara's book "The Classical Groups and K-Theory".

  1. I also want to make a comment on DLS's answer. There is a related short exact sequence

$1 \rightarrow V \rightarrow SL_n(\mathbb{Z}/p^{k+1}\mathbb{Z}) \rightarrow SL_n(\mathbb{Z}/p^k\mathbb{Z}) \rightarrow 1.$

The group $V$ has a beautiful description, due essentially to Lee and Szczarba. Namely, it is isomorphic to the additive group underlying the special lie algebra over $\mathbb{Z} / p\mathbb{Z}$ (in particular, it is abelian). Moreover, the action of $SL_n(\mathbb{Z}/p^k\mathbb{Z})$ on $V$ factors through the adjoint representation of $SL_n(\mathbb{Z}/p\mathbb{Z})$ on the special lie algebra.

Lee and Szczarba did not write this down in quite this form, but I wrote out a similar result for the symplectic group in Lemma 3.1 of my paper "The Picard group of the moduli space of curves with level structures".

$\endgroup$

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.