Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

The following should be true: every normal subgroup of a non-Abelian Right Angled Artin Group should contain a free group on two generators. Is there a standard reference one can cite for this?

share|improve this question
$\mathbf{Z}\times\{1\}\subset\mathbf{Z}\times F_2$ gives a counterexample. –  YCor Oct 4 '12 at 19:21
Hmm. Maybe the right question should be: what is the right statement :( –  Igor Rivin Oct 4 '12 at 19:29
Probably the starting point should be Sageev and Wise, [The Tits alternative for CAT(0) cubical complexes][1]., Bull. London Math. Soc. 37 (2005), no. 5, 706–710 So the question is: when are there non-solvable normal subgroups? [1]: ams.org/mathscinet-getitem?mr=2164832 –  Lior Silberman Oct 4 '12 at 19:58
@Lior: actually the question is when are there any SOLVABLE normal subgroups (I am saying never, except in trivial examples, such as direct products with abelian groups). –  Igor Rivin Oct 4 '12 at 20:08
Igor, correct question, I think, is about center-less RAAGs. Proof should follow from Sageev-Wise with a bit of fiddling. –  Misha Oct 4 '12 at 22:38

2 Answers 2

up vote 10 down vote accepted

Igor, I assume the question is about centerless RAAGs. Then it follows from the classical result of A. Baudisch (see MR0634562): every 2-generated subgroup of a RAAG is abelian or $F_2$.

Indeed if $N$ is a non-trivial normal subgroup in a RAAG $G$, take any $x\in N$. Since $G$ is centerless there exists $g\in G$ that does not commute with $x$. Then $\langle x,g\rangle\cong F_2$ and hence $\langle x, x^g \rangle \cong F_2$.

share|improve this answer
@Denis: thanks much! The argument you give actually seems to give more than I asked for: a normal subgroup is either central or contains an $F_2.$ –  Igor Rivin Oct 5 '12 at 23:38

A right-angled Artin group $G$ is the fundamental group of the compact non-positively curved cube complex called the Salvetti complex $X$. Thus $\tilde{X}$ is a CAT(0) space, and the action of $G$ on $\tilde{X}$ is proper and cocompact.

Thus, by the Solvable Subgroup Theorem, every virtually solvable subgroup $H< G$ contains a finitely generated abelian subgroup of finite index. Moreover, every element in $G$ will be semi-simple by Prop. 6.10 (2) from Bridson-Haefliger. By Corollary 7.2, there is an $H$-invariant closed convex subspace of $\tilde{X}$ isometric to a product $Y\times \mathbb{E}^n$, so that $H$ acts as the identity on $Y$ and acts cocompactly on the $\mathbb{E}^n$ factor. Any isometry of $\tilde{X}$ which normalizes $H$ preserves $Y\times \mathbb{E}^n$ and its splitting. We conclude that $\tilde{X}=Y\times\mathbb{E}^n$ when $H$ is normal in $G$.

Now, I think for a CAT(0) cube complex $\tilde{X}$, the factorization $Y\times \mathbb{E}^n$ should actually be a product of cube complexes. This isn't quite right, since in the group $\mathbb{Z}^n$ there are clearly $\mathbb{R}^k$ factors of $\mathbb{R}^n$ which are not a factor of the cube complex structure. But I think if $H$ is assumed to be a virtually solvable normal subgroup of $G$ and maximal with respect to this property, then $\tilde{X}=Y\times \mathbb{E}^n$ should be a product of cube complexes. I think it should be possible to prove this by analyzing links of vertices of the Salvetti complex, and how the foliation by $\mathbb{E}^n$'s passes through them.

But this implies that $G=G_1\times \mathbb{Z}^n$ where $G_1$ is a right-angled Artin group, since each edge of a cube of the Salvetti complex corresponds to a generator, so a cube complex factor isometric to $\mathbb{R}^n$ will give generators commuting with every other generator.

As observed in one of the comments, $G$ satisfies the Tits alternative since it is linear, so a normal subgroup is either virtually solvable, and then $G$ splits off a $\mathbb{Z}^n$ factor, or the normal subgroup is not solvable and therefore must contain a free subgroup. However, one would need to fill in the above missing step to finish this argument.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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