MathOverflow is a question and answer site for professional mathematicians. 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

I've heard this result from my differential manifold class, and I don't know how to prove it.

Does anyone know how to construct such diffeomorphism? Please tell me, thanks a lot.

Any comments are welcome.

share|cite|improve this question
You can add one more object to your list: take the space of subsets of $S^1$ with at most three elements. This is homeomorphic to $S^3$ and a free orbit of the $SO_2$ action is a trefoil. In your case, $SL_2 \mathbb R / SL_2 \mathbb Z$ has an $SO_2$ action, so it's Seifert fibred. Consider the orbit decomposition. That's my favourite route to a proof. I'm pretty sure there's several MO threads on this topic already. – Ryan Budney Apr 13 '12 at 8:23
This is fairly standard, and more appropriate for math.stackexchange. – BS. Apr 13 '12 at 8:56
@Ryan Thanks for your links, I've found that in John Baez mentioned that it was proved in Milnor's book 'Algebraic K-theory'. – Yuchen Liu Apr 13 '12 at 8:58
@BS: this is a bit advanced for math.stackexchange, and I think the question is interesting to the MO community. – Jim Conant Apr 13 '12 at 13:56
up vote 12 down vote accepted

Here is a sketch:

$SL(2,\mathbb R)/SL(2,\mathbb Z)$ can be identified with the space of lattices of covolume 1 in $\mathbb R^2$. Indeed, $SL(2, \mathbb R)$ can be identified with a couple of vectors, and multiplying this couple by an element of $SL(2, \mathbb Z)$ does not alter the lattice they generate.

Let $\Lambda$ be a lattice, $v$ be its shorter vector, $v'$ be the shortest vector non colinear with $v$ so that the basis $(v,v')$ is direct. Denote by $\theta$ the angle (mod $\pi$) made by $v$ with the horizontal direction. Then the couple $(v'/v, \theta)$ determines the lattice. It is an exercise to see that $v'/v$ belongs to the domain $\mathcal D:=\{z\in\mathbb C\big\vert -1/2\le\Re z\le 1/2, \vert z\vert\ge 1\}$. In order two have an isomorphism, it should be noted that the only repetitions in the code concern the points $(z,\theta)$ and $(-\bar z, -\theta)$ for $\vert z\vert=1$ which determine the same rhombical lattice, as well as the points $(z, \theta)$ and $(z+1, \theta)$ for $\Re z = -1/2$. Let $\mathcal D/\tilde{}$ denote the corresponding quotient of $\mathcal D$.

All in all this shows that $SL(2,\mathbb R)/SL(2,\mathbb Z)$ is a Seifert circle bundle over $\mathcal D/\tilde{}$, with one singular fiber of order 2 above the point $i$ (which corresponds to square lattices) and one singular fiber of order 3 above the point $j$ (which corresponds to hexagonal lattices).

Now cut $\mathcal D/\tilde{}$ along the two lines $\Re z=-1/4$ and $\Re z=1/4$. You obtain two disks $D_i$ and $D_j$, where $D_i$ contains a point of index 2, and $D_j$ contains a point of index 3, and both of them have one point removed on the boundary. Now there is only one Seifert fibered space over a disk with a singular pointof order $p$, namely a solid torus where each fiber except the central one crosses $p$ times every meridian disk. Thus our Seifert fibered space is made of two solid tori glued along their boundary, where we have to identify the fibers on the boundary. What we obtain is a Lens space with one fiber removed. It turns out that in this $(2,3)$-case, the obtained Lens space is $S^3$, and the regular fibers are trefoils.

share|cite|improve this answer
1 with further reference to John Milnor, Introduction to Algebraic K-theory, Annals of Math. Studies 72, Princeton U. Press, Princeton, New Jersey, 1971. Milnor credits it to Quillen. – Swiat Gal Apr 13 '12 at 22:00

There is a proof without words in Etienne Ghys' ICM address. and a short proof with words by Rubinstein and Gardiner. (Compositio, 1979)

share|cite|improve this answer

A different proof (due to Quillen) is given in Milnor's book Introduction to algebraic K-Theory, page 84.

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.