4
$\begingroup$
  1. Let $H^2=\{(x,y)\mid y>0\}$ be the hyperbolic upper-half plane.
  2. Let $K=Q(\sqrt{d})$ be a quadratic number field, and $\mathcal{O}_K$ be the ring of algebraic integers in it.
  3. Let $\Gamma=\Delta(p,q,r)$ be a triangle group of hyperbolic type, whose fundamental domain at least one ideal vertex on the boundary of $H^2$. That is, at least one of $p,q,r$ is $\infty$. $\Gamma$ gives a hyperbolic tiling of $H^2$ by non-compact triangles.

Here is an example: Let $\Gamma=(2,q,\infty)$ be a Hecke group where $q\geq3$ is an integer. The fundamental domain $D$ of $\Gamma$ has vertices at $i, e^{i\pi/q},\infty$. The triangles $\{w(D)\mid w\in\Gamma\}$ tile $H^2$. The cusps are the set $\{w(\infty)\mid w\in\Gamma\}$.

My question is this:

For which choices of $d$ does there exist a triangle group $\Gamma$ such that the cusps in the tiling are exactly $\mathcal{O}_K$? That is, $\mathcal{O}_K=\{w(\infty)\mid w\in\Gamma\}$?

$\endgroup$
9
  • 1
    $\begingroup$ I am confused. If $p$, $q$, $r < \infty$, then $\Gamma$ gives a tiling of $H^2$ by actual hyperbolic triangles, not by ideal triangles. Their vertices are in the interior of $H^2$, not on the boundary. For example, here is an image of the $(2,3,7)$ tiling en.wikipedia.org/wiki/(2,3,7)_triangle_group . What do you mean by an ideal vertex of a tiling like this? $\endgroup$ Commented Feb 25 at 16:02
  • $\begingroup$ The standard definition of an ideal hyperbolic triangle requires all three vertices to be at infinity. It seems that you are not following this convention. What is your definition? At least one vertex at infinity? $\endgroup$ Commented Feb 25 at 16:57
  • $\begingroup$ Sorry for the confusion! I mean that the fundamental domain has at least one ideal vertex. So at least one of $p,q,r$ should be $\infty$. For example the (2, 3, inf) tiling. I'll update in the post. $\endgroup$
    – zemora
    Commented Feb 26 at 3:09
  • $\begingroup$ Could you explain the simplest example, i.e. $q=3$? In this case, the cusps are precisely the rationals (and infinity, which I presume you want to ignore). This is not the ring of integers of any number field… $\endgroup$
    – HJRW
    Commented Feb 26 at 9:02
  • $\begingroup$ Yes, for $q=3$ the modular group does not have cusps at "quadratic integral domain". However, I doubt the case $q=4$ has cusps at $\mathcal{O}_K$ for $K=\mathbb{Q}(\sqrt{2})$. $\endgroup$
    – zemora
    Commented Feb 26 at 9:10

1 Answer 1

2
$\begingroup$

This is a partial answer to say that this won't happen very often.

Let's work with the triangle group where $s_1 s_2$ has order $p$, $s_2 s_3$ has order $q$ and $s_1 s_3$ has order $\infty$. Let $\omega_2$ be the ideal vertex of the base triangle, so $\omega_2$ is fixed by $s_1$ and $s_3$. Consider the four points $\omega_2$, $s_2 \omega_2$, $s_1 s_2 \omega_2$, $s_3 s_2 \omega_2$ in $\partial H^2 \cong \mathbb{RP}^1$. (Note that, if $p=2$, then $s_1 s_2 \omega_2 = s_2 s_1 \omega_2 = s_2 \omega_2$, so we only have three distinct points.) In this answer, I'll compute the cross ratio of these points and get that it is $\tfrac{\cos(\pi/q)}{\cos(\pi/p)}$. In particular, if we want these points to lie in $\mathbb{P}^1(\mathbb{Q}(\sqrt{d}))$, then we must have $\tfrac{\cos(\pi/q)}{\cos(\pi/p)} \in \mathbb{Q}(\sqrt{d}) \cup \{ \infty \}$. At some point I might come back and carefully do the Galois theory to work out all the cases where this happens but, for typical $(p,q)$, it certainly doesn't.

The computation: Let $A$ be the Cartan matrix: $$A = \begin{bmatrix} 2 & -2 \cos(\pi/p) & -2 \\ -2 \cos(\pi/p) & 2 & - 2 \cos(\pi/q) \\ -2 & -2 \cos (\pi/q) & 2 \\ \end{bmatrix}.$$ Let $\alpha_1$, $\alpha_2$, $\alpha_3$ be the simple roots, so we have $\langle \alpha_i, \alpha_j \rangle = A_{ij}$ for the invariant dot product. Let $\omega_1$, $\omega_2$, $\omega_3$ be the dual basis, so $\langle \alpha_i, \omega_j \rangle = \delta_{ij}$. The action of the Coxeter group in the dual basis is $$s_i \omega_j = \begin{cases} - \omega_i - \sum_{k \neq i} A_{ik} \omega_k & i=j \\ \omega_j & i \neq j \\ \end{cases}.$$ I compute that $$\begin{array}{rcrcrcr} \omega_2 &=& && \omega_2 && \\ s_2 \omega_2 &=& -A_{12} \omega_1 &-& \omega_2 &-& A_{23} \omega_3 \\ s_1 s_2 \omega_2 &=& A_{12} \omega_1 &+& (A_{12}^2-1) \omega_2 &-& (2 A_{12} + A_{23}) \omega_3 \\ s_3 s_2 \omega_2 &=& - (A_{12} + 2 A_{23})\omega_1 &+& (A_{23}^2-1) \omega_2 &+& A_{23} \omega_3 \\ \end{array}$$

Now, these coordinates are in $\mathbb{R}^3$, or I can think of them as homogeneous coordinates in $\mathbb{P}^2$. I want to compute their images under the isomorphism $\partial(H^2) \to \mathbb{RP}^1$. To do this, I don't actually need to compute the equation of the conic $\partial H^2$ in $\mathbb{P}^2$: For any conic $C$ in $\mathbb{P}^2$, and any point $\omega$ in $C$, the isomorphism $C \to \mathbb{P}^1$ is given by linear projection from $\omega$.

However, there is one subtlety: To figure out where the point $\omega$ maps to, we need to figure out the equation of the tangent line to $C$ at $\omega$. The tangent line to $C$ at $\omega$ is $\beta^{\perp}$ where the vector $\beta$ is determined by the properties that $\langle \beta, \beta \rangle = \langle \beta, \omega \rangle = 0$. Then the image of $\omega$ under the map $C \to \mathbb{P}^1$ is given by applying the linear projection from $\omega$ to a point on $\beta^{\perp}$ other than $\omega$.

We take the projection from $\omega_2$. Note that the vector $\beta:= \alpha_1 + \alpha_3$ obeys the conditions $\langle \beta, \beta \rangle = \langle \beta, \omega_2 \rangle = 0$, so $\beta^{\perp}$ is the tangent line to $\partial(H^2)$ at $\omega_2$. Note also that $\omega_1 - \omega_3$ is a point on the line $\beta^{\perp}$.

So, we take the linear projection $\mathbb{R}^3 \to \mathbb{R}^2$ which projects away the $\omega_2$ coordinate. We apply this projection to $\omega_1 - \omega_3$, $s_2 \omega_2$, $s_1 s_2 \omega_2$ and $s_3 s_2 \omega_2$ in order to get the points of $\mathbb{P}^1$ corresponding to these points. Concretely, these points are $$(1:-1),\ (-A_{12}: -A_{23}),\ (A_{12}: - 2 A_{12} - A_{23}),\ (-A_{12} - 2 A_{23} : A_{23}).$$ I compute that the crossratio is $$\frac{(A_{12}+A_{23})(2 A_{12} A_{23} + 2 A_{23}^2)}{(A_{12}+A_{23})(2 A_{12}^2+2 A_{12} A_{23} )} = \frac{A_{23}}{A_{12}} = \frac{\cos(\pi/q)}{\cos (\pi/p)}.$$

Sorry for the partial answer, but this seems long enough to be worth writing down by itself. If I cared enough about this problem, I'd compute a bunch more ideal points and their cross ratios, to get a better idea of what is going on.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .