Let $G = \operatorname{SL}_2(\mathbb R)$, and let $\Gamma$ be a discrete subgroup of $G$. When studying automorphic forms, one might look at the quotient space $\Gamma \backslash \mathbb H$, where $\mathbb H$ is the upper half plane, or perhaps even $L^2(\Gamma \backslash \mathbb H)$, or $L^2(\Gamma \backslash G)$ (the elements of $\Gamma \backslash \mathbb H$ identify with right $K$-invariant functions on $\Gamma \backslash G$, where $K = \operatorname{SO}(\mathbb R)$).

Typically there will be a fundamental domain $\mathcal F$ for the action of $\Gamma$ on $\mathbb H$, through which one gets a measure $d \dot \tau$ on $\Gamma \backslash \mathbb H$, such that an "unwinding formula" holds:

$$\int\limits_{\mathbb H} f(\tau) \frac{dxdy}{y^2} = \sum\limits_{\gamma \in \Gamma} \int\limits_{\Gamma \backslash \mathbb H} f(\gamma. \tau) d \dot \tau $$

The classical case is when $\Gamma$ is a congruence subgroup of $G$. In this case, $\Gamma \backslash \mathbb H$ and $\Gamma \backslash G$ both have finite measure, but are not compact.

The compact case is nicer in a lot of ways. $L^2(\Gamma \backslash G)$ and $L^2(\Gamma \backslash \mathbb H)$ decompose as a Hilbert space direct sum of irreducible representations of $G$, for example.

What are some nice examples of subgroups $\Gamma$ such that $\Gamma \backslash G$ (equivalently, $\Gamma \backslash \mathbb H$) is compact? What does the fundamental domain look like?

  • $\begingroup$ For nice examples over $\Bbb C$ see this MO-question. For $SL_2(\Bbb R)$ see Arithmetic Fuchsian groups, which gives many examples: "all orders in quaternion algebras (satisfying the above conditions) which are not $M_{2}(\mathbb {Q} )$ yield cocompact subgroups. $\endgroup$ – Dietrich Burde Dec 9 '19 at 20:02
  • $\begingroup$ For lots of pretty pictures of this, see chapters 17 and 18 of "The Symmetries of Things" (Conway, Burgiel, and Goodman-Strauss). $\endgroup$ – Ted Dec 11 '19 at 2:02

Many examples arise by application of the Poincare polygon theorem, which you can find for example in Theorem 11.2.1 of Ratcliffe's book "Foundations of Hyperbolic Manifolds".

Here's a general description. Consider a finite polygon $P$ with $2n$ sides. Let $f_1,...,f_n \in \text{Isom}(\mathbb H)$ be isometries satisfying the following conditions:

  1. For each $i=1,...,n$ there exist sides $a_i,a'_i$ such that $$P \cap f_i(P) = a_i = f_i(a'_i) $$
  2. The list of sides of $P$ is $a_1,a'_1,...,a_n,a'_n$.
  3. The total angle of each "vertex cycle" is $2\pi$ (more can be said here, but I'll refer to topology textbooks that cover the classification of surfaces).

The conclusion of the Poincare Polygon Theorem is that $f_1,...,f_n$ are generators of a discrete group $\Gamma$, $P$ is a fundamental polygon for $\Gamma$, the quotient map $\mathbb H \mapsto \mathbb H / \Gamma$ is a universal covering map, and the composition $$P \hookrightarrow \mathbb H \mapsto \mathbb H / \Gamma $$ is the quotient map obtained by gluing $a_i$ to $a'_i$ using $f_i$, for each $i=1,...,n$.

Let me just do one example to give the idea, the "standard gluing pattern" of an octagon. Let $P \subset \mathbb H$ be a regular octagon with angles $2\pi/8$, so all side lengths are equal. Starting from a vertex $v_0$ and going around in counterclockwise order, list the sides like this: $$a_1, a_2, a'_1, a'_2, a_3, a_4, a'_3, a'_4 $$ Choose $f_i$ to be the unique orientation preserving isometry satisfying the condition above with respect to $a_i$ and $a'_i$. Now one has to check that all 8 vertices are identified to a single point on the quotient (this is what one does by going around the "vertex cycle"). Since each of these has angle $2\pi/8$, their sum is $2\pi$ as required.

  • $\begingroup$ The defect implies for all $\alpha <3\pi/4$ there is a (regular) hyperbolic octogon whose interior angles are $\alpha$ ? $\endgroup$ – reuns Dec 9 '19 at 23:32
  • $\begingroup$ That's right, just an intermediate value theorem argument, as the vertices go out to infinity. $\endgroup$ – Lee Mosher Dec 10 '19 at 0:57
  • $\begingroup$ How is this a subgroup of $\operatorname{SL}_2(\mathbb R)$? $\endgroup$ – D_S Dec 10 '19 at 0:59
  • $\begingroup$ $SL_2(\mathbb R) = \text{Isom}_+(\mathbb H)$. $\endgroup$ – Lee Mosher Dec 10 '19 at 0:59
  • $\begingroup$ What is $\operatorname{Isom}_+(\mathbb H)$? $\endgroup$ – D_S Dec 10 '19 at 1:03

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.