MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I've been struggling with this one all day, and I was wondering if someone can give me a hand with the proof. I'm not even sure if the group in question is finitely generated, so I would appreciate if anyone will tell me otherwise as well.

remark 1 I've posted this question on Math.stackexchange as well, and am posting here since I don't seem to get any replies. I don't know if this question is "MO standard", but I guess there's no harm in trying :-P..

remark 2 For infinite topological groups, I write $G$ is finitely generated to mean that $G$ has a dense subgroup which is finitely generated (i.e. $G$ is topologically finitely generated).

Let $F\supseteq \mathbb{Q}_p$ be a local field, and let $D$ be a finite dimensional division algebra over $F$, of degree $n$ (i.e $\dim_F D=n^2$).

It is known that $D$ contains an unramified extension $E$ of $F$, such that $E/F$ is cyclic Galois. It follows that $D$ contains an element $u\in D^\times$, and that there is some $\pi\in F$ such that $D=\bigoplus_{j=0}^{n-1}u^j E$, $u^n=\pi$ and such that the restriction of the map $x\mapsto u^{-1}xu$ to $E$ is a generator of the Galois group $\mathbf{G}(E/F)$.

Thus, $D$ embeds into the $F$ algebra $M_n(E)$ of $n\times n$ matrices over $E$, via the left regular action $x\mapsto \lambda_x$ where $\lambda_x:E^n\to E^n$ is defined by $$\lambda_x(y_0,\ldots,y_{n-1})=x\cdot(\sum_{j=0}^{n-1}u^jy_j).$$

Once this is done, one can define the reduced norm on $D$ by $$Nrd_{D/F}(x)=\det\lambda_x.$$

My interest specificaly is in the groups $G:=SL_1(D):=\lbrace x\in D\mid Nrd_{D/F}(x)=1\rbrace$. I'm trying to understand whether this group is finitely generated or not.

It is clear that the embedding defined above maps $G$ into the group $SL_n(E)$. Once we use the fact that the group $SL_n(E)$ is generated by transvections (i.e matrices with $1$s along the diagonal and at most one additional non-zero entry), it is not hard to use the local structure of $E$ to prove that $SL_n(E)$ is finitely generated.

Unfortunately, a subgroup of a finitely generated group need not be finitely generated in general. If anyone knows of a criterion that might prove why this case is special that would be a great help.

On additional possibility, is to use some heavy tools from asymptotic group theory (even though I must admit this turns out a bit cumbersome..):

We call a group $G$ positively finitely generate (abbreviated PFG), if there exists a number $k\in\mathbb N$ such that the probability that an arbitrary $k$-tuple $(x_1,\ldots, x_k)\in G^k$ generates $G$, is positive. Clearly, any PFG group is finitely generated (although the converse is not true, e.g the free group in two generators is not PFG).

For any $n\in\mathbb N$ we define $m_n(G)$ to be the number of maximal subgroups of $G$ with index $n$. It is a theorem of Avinoam Mann that the group $G$ is PFG if and only if $m_n(G)$ is bounded by some polynomial in $n$. So it would suffice to prove that the (maximal) subgroups of $G$ have polynomial growth.

As I said, this sort of proof seems a bit of an overkill, but if someone knows of a reference (or a direct proof) to why the group $G$ has polynomial (maximal) subgroup growth that will be swell :-).

Anyway, this question turns out to be a bit lengthy, so I'll stop here. I would very much appreciate If someone can offer any hints as to whether or not $SL_1(D)$ is finitely generated.

Thank you very much.

share|cite|improve this question
up vote 6 down vote accepted

Yes, this follows from finiteness properties of arithmetic groups and the strong approximation theorem for simply connected groups.

Choose a global field $K$ with a non-archimedean place $v$ such that $K_v \simeq k$; this can be found via elementary approximation arguments. By our knowledge of the Brauer groups of $K$ and $k$ via class field theory, there is a central division algebra $A$ over $K$ such that $A_v \simeq D$ over $k$.

Let $G$ be the absolutely simple semisimple $K$-group informally denoted as ${\rm{SL}}_1(A)$ (to be precise, $G$ represents the functor $R \rightsquigarrow {\rm{SL}}_1(A \otimes_K R)$ on the category of $K$-algebras), so $G(k) = {\rm{SL}}_1(D)$. Hence, it suffices to show that $G(k)$ contains a dense finitely generated subgroup.

By general "spreading out" arguments, for any connected semisimple $K$-group $H$ there is a finite set of places $S$ of $K$ containing the archimedean places and such that $H$ extends to a semisimple group scheme $\mathcal{H}$ over $O_{K,S}$, so $H$ is quasi-split at all places outside $S$ and hence (by the Borel--Tits structure theory) has positive rank at all places outside $S$. Thus, by enlarging $S$ to contain at least two such places, we can arrange that $$\sum_{w \in S} {\rm{rank}}(H_{k_w}) \ge 2.$$ We apply this with $H = G$ and therefore denote $\mathcal{H}$ as $\mathcal{G}$, and we enlarge $S$ if necessary so that $v \in S$.

Consider the arithmetic $S$-group $\Gamma = \mathcal{G}(O_{K,S})$ inside $G(K)$. In the number field case this is finitely generated, by general finiteness results for arithmetic $S$-subgroups of the group of rational points of absolutely simple semisimple group over number fields. In the function field case the same holds because of the rank-sum condition arranged above (by a theorem of Behr). So it suffices to prove that $\Gamma$ has dense image in $G(k) = G(K_v)$.

Let $S' = S - \{v\}$ (non-empty). Since $G$ is absolutely simple and simply connected, by the strong approximation theorem we know that $G(K)$ is dense in $G(\mathbf{A}_K^{S'}) = G(K_v) \times G(\mathbf{A}_K^S)$, where $\mathbf{A}_K^{S'}$ is the adele ring of $K$ with the $S'$-factors removed. Let $\widehat{O}_{K,S} \subset \mathbf{A}_K^S$ be the open subring $\prod_{w \not\in S} O_{K_w}$, so $\mathcal{G}(\widehat{O}_{K,S})$ is an open subset of $G(\mathbf{A}_K^S)$ that meets $G(K)$ in exactly $\mathcal{G}(O_{K,S}) = \Gamma$. Hence, if $U \subset G(K_v)$ is a non-empty open subset then $U \times \mathcal{G}(\widehat{O}_{K,S})$ is an open subset of $G(\mathbf{A}_K^{S'})$ whose intersection with $G(K)$ is identified (via ${\rm{pr}}_v$) with $U \cap \Gamma$ inside $G(K_v)$. But recall that $G(K)$ is dense in $G(\mathbf{A}_K^{S'})$, so its intersection with $U \times \mathcal{G}(\widehat{O}_{K,S})$ is non-empty. In other words, $U \cap \Gamma$ is non-empty inside $G(K_v)$. Since $U$ was arbitrary, this says that $\Gamma$ is dense in $G(K_v) = G(k)$ as desired.

This argument is quite robust with respect to the permitted generality of the initial connected semisimple $k$-group, the main point being the existence of the "globalization" of some $K$. One is then led to ask: if $G$ is a connected semisimple group over $k$ that is absolutely simple then does $G$ admit a "globalization" relative to some pair $(K,v)$ consisting of a global field $K$ and place $v$ such that $K_v \simeq k$ (and if so then can this be proved without case-checking). But this refinement was not part of the question, so I'll stop here.

share|cite|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.