Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

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

How can I prove that if $S,S'$ are two different finite generating sets of a group $G$ , then the metric spaces induced by the "word metric" are quasi-isometric?

The definition of quasi-isometry is: Let $X,Y$ be compact metric spaces, and $f:X \to Y$ a map. We say that $f$ is a quasi-isometry if there exist $L,A>0$ such that for every $x_1 ,x_2 \in X$ , $y \in Y$ : $ \frac{1}{L} d(x_1, x_2) -A \leq d(f(x_1), f(x_2)) \leq Ld(x_1,x_2) +A$ and $d(y, f(X))\leq A$ .

[The first condition is the problematic one... How can we find proper constants L, A ? ] Thanks in advance !

share|cite|improve this question
I like t.b.'s answer below, but another approach is to prove that if $X$ and $Y$ are two metric spaces on which $G$ acts freely, properly, and cocompactly by isometries then $X$ and $Y$ are quasi-isometric. $G$ acts on the Cayley graphs of $(G,S)$ and $(G,S')$ with one point quotients, so the result follows. – Paul Siegel May 20 '12 at 17:10
@Paul: I think you need to assume that $X$ and $Y$ are geodesic. Consider the natural action of $\mathbb{Z}$ given by $\begin{pmatrix} 1 & 1 \\ 0 & 1\end{pmatrix}$ on the "horosphere" $H_t = \{1+it\,:\,t \in \mathbb{R}\}$ in the upper half plane. But in any case, your argument works if you let $G$ act on the Cayley graph for a counterexample to your statement. .@joshua See here for a proof of the fact Paul is alluding to. – t.b. May 21 '12 at 9:06
Yeah, you're right. I've become accustomed to the metric geometers' convention that all spaces are length spaces unless otherwise specified. – Paul Siegel May 26 '12 at 21:06
up vote 4 down vote accepted

For compact spaces the notion of quasi-isometry is rather meaningless: simply take $A$ larger than the diameter of both $X$ and $Y$, so I'll assume that you don't want this answer...

I'm used to the convention that the word metric $d_S$ is obtained from the word length $$ \ell_S (x) = \min\left\{n \in \mathbb{N}\,:\, x= s_{1}\cdots s_n \text{ for some } s_{1},\ldots,s_n \in S^{\pm1}\right\} $$ (allowing the empty product, so that $\ell_S(e) = 0$) by setting $d_S(x,y) = \ell_S(x^{-1}y)$, but there are some variations which need some modifications of what I'm saying here, but nothing essential.

The subadditivity property $\ell_{S}(xy) \leq \ell_{S}(x) + \ell_{S}(y)$ of the length function $\ell_S$ is responsible for the triangle inequality of $d_S$.

Let $C' = \max\left\{\ell_{S}(s')\,:\,s' \in (S')^{\pm 1}\right\} \geq 1$. Then it follows from the definitions and subadditivity that $\ell_{S}(x) \leq C'\ell_{S'}(x)$, so $d_S(x,y) \leq C'd_{S'}(x,y)$.

Symmetrically $\ell_{S'}(x) \leq C \ell_{S}(x)$, so $d_{S'}(x,y) \leq Cd_{S}(x,y)$ which means that $$ \frac{1}{C'} d_{S}(x,y) \leq d_{S'}(x,y) \leq Cd_{S}(x,y), $$ so that the identity map $(G,d_{S}) \to (G,d_{S'})$ is a quasi-isometry (in fact a bi-Lipschitz map) with constants $L = \max\{C,C'\}$ and $A = 0$.

If instead you are interested in the Cayley graphs of $G$ with respect to $S$ and $S'$, you can take $A = 1$ and the same $L$ as above.

Also, notice that I didn't use that $G$ is finite (so that $(G,d_S)$ is compact), only that the generating sets $S$ and $S'$ are finite. For the reason I mentioned at the beginning of my answer, compactness isn't usually included in the definition of quasi-isometries, and it is common to require $L \geq 1$ and allow $A = 0$.

share|cite|improve this answer
Thanks a lot both of you ! – joshua May 20 '12 at 18:19

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.