15
$\begingroup$

Like every other group also $\mbox{GL}(n,\mathbb{Z})$ acts on the set of all its subgroups, by conjugation: if $\phi \in \mbox{GL}(n,\mathbb{Z})$, then $\phi$ acts by $H \mapsto \phi H \phi^{-1}$, where $H \leq \mbox{GL}(n,\mathbb{Z})$.

A theorem by Jordan (and later Zassenhaus) implies that the number of conjugacy classes of finite subgroups of $\mbox{GL}(n,\mathbb{Z})$ is finite.

About this fact I have a few questions:

  1. Is the actual number of conjugacy classes known for each $n$? Maybe at least asymptotically?

  2. Is it known which of these classes are particularly big for every $n$? (Edit: Doesn't make sense in this context, see comments.)

$\endgroup$
  • 1
    $\begingroup$ I am confused by question (2). These conjugacy classes are infinite, so how are you measuring their size? $\endgroup$ – David E Speyer Sep 4 '12 at 14:19
  • $\begingroup$ If I counted right $GL_1$ has $2$ and $GL_2$ has $9$. The number should increase fairly rapidly with $n$. $\endgroup$ – Will Sawin Sep 4 '12 at 15:29
  • 4
    $\begingroup$ The largest finite subgroup is the orthogonal group of order $2^nn!$, at least for all $n > 10$. These questions have been discussed on MO before, but I do not believe that any asymptotic results on the number of classes are known. $\endgroup$ – Derek Holt Sep 4 '12 at 19:14
  • $\begingroup$ @David Speyer: You're right, it doesn't make sense to speak about big conjugacy classes. Since I was thinking about space group and their arithmetic equivalence for the last few days, things got mixed up in my head. Thanks for pointing that out. $\endgroup$ – Gregor Samsa Sep 4 '12 at 20:42
  • 1
    $\begingroup$ I think the number of Bieberbach groups gives an upper bound. For each finite subgroup $G< GL_n(\mathbb{Z})$, one obtains a Bieberbach group $G \ltimes \mathbb{Z}^n$. Moreover, if two Bieberbach groups are obtained this way, then by Bieberbach's theorem, they are affinely equivalent. Since the $\mathbb{Z}^n$ subgroup is a maximal subgroup of translations, this means the affine map must send $\mathbb{Z}^n$ to $\mathbb{Z}^n$, so it lies in $GL_n(\mathbb{Z})$. However, this will be an overcount, since there are Bieberbach groups which are not of this form (e.g. Klein bottle group), ie not split. $\endgroup$ – Ian Agol Sep 5 '12 at 12:42
6
$\begingroup$

A rough estimate can be obtained by comparison with the number of isomorphism classes of (symmorphic) space groups (see also Agol's comments): The map $Q \mapsto Q \ltimes\mathbb{Z}^n$ induces a bijection between the conjugacy classes $C_n$ of finite subgroups of $GL_n(\mathbb{Z})$ and the isomorphism classes of symmorphic space groups. A proof thereof can be found in my answer to this question:

Subgroups of the Euclidean group as semidirect products

In particular, $|C_n| = 73,\; 710,\; 6079,\; 85311$ for $n=3,4,5,6$ respectively:

http://www.math.ru.nl/~souvi/papers/acta03.pdf

The number of isomorphism classes of (all) space groups has been estimated (cf. Remark 5.5) in

http://www.unige.ch/math/folks/bucher/Affine/pdfAffine/BuserBieberbach.pdf

and yields the upper bound $|C_n| \le e^{\displaystyle e^{4n^2}}$.

Actually it is conjectured by Schwarzenberger in a 1974 paper that the number of isomorphism classes of space groups is asymptotically $O(2^{\displaystyle n^2})$. But I don't know if this has been proved in the meanwhile.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ All this is very interesting, thank you. $\endgroup$ – Gregor Samsa Sep 12 '12 at 12:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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