3
$\begingroup$

I am finding the explicit description of genus $2$ surface as the upper half plane modulo group of Deck Transformation. I did't find it anywhere. Also, I found a similar question here. But, there is no answer. My question is same as the question given in the link. Please help me to find the answer explicitly.

Thanking in advanced.

$\endgroup$
4
  • 2
    $\begingroup$ What do you mean by "explicitly"? There are various possibilities... $\endgroup$
    – Sam Nead
    Commented Dec 2, 2020 at 18:17
  • 2
    $\begingroup$ @SamNead, Sir by the word "explicitly" I meant that the answer stated clearly, and in details. Please help me. $\endgroup$
    – user147962
    Commented Dec 2, 2020 at 18:26
  • 3
    $\begingroup$ Your notion of "explicit" might be OK for an undergraduate/graduate class but is utterly insufficient for a research-level question. Also, a quantifier is missing in front of the word "surface." Are you looking for a generating set in the case of just one Riemann surface of genus 2 or you are asking for "explicit" generating sets for every Riemann surface of genus 2. $\endgroup$ Commented Dec 2, 2020 at 19:29
  • 1
    $\begingroup$ Incidentally, Seppala gave an algorithm for numerical approximation of Fuchsian uniformization of hyperelliptic curves (which includes all genus 2 compact Riemann surfaces) in terms of defining polynomial equations, here. $\endgroup$ Commented Dec 3, 2020 at 1:16

2 Answers 2

9
$\begingroup$

Looking at the link you give, it seems that you want an explicit representation of $\pi_1 = \pi_1(S)$, the fundamental group of the genus two surface, into $\mathrm{PSL}(2, \mathbb{C})$ so that the upper half plane, modulo the induced action of $\pi_1$, is a "hyperbolic structure" on the topological surface $S$.

That is a long question (so perhaps I've guessed incorrectly). The answer is somewhat involved, and requires some hyperbolic trigonometry, which is why nobody is about to just write down the answer for you. I have done this at least once - but it is buried deep in a program that I wrote ages ago.

So here is a sketch -- you will have to work out the details for yourself.

Step one: Draw a (cartoon of a) regular octagon $R$ in the upper half plane model of the hyperbolic plane. We want the one that has angle $2\pi / 8$ at each of its vertices. Do this so that the centre is at $i$, and so that two opposite sides of $R$ are perpendicular to the imaginary axis. Let $a$ and $b$ be the midpoints of the two sides crossing the imaginary axis.

Step two: Chop $R$ into sixteen triangles. Each triangle meets the centre (the point $i$), a vertex, and an edge midpoint. All of these triangles have angles $(\pi/2, \pi/8, \pi/8)$ at their vertices; thus the triangles are all congruent (and isosceles!)

Step three: Do the hyperbolic trig to find the side-lengths of this triangle. What you really want is $\ell_{m}$, the hyperbolic distance from the edge midpoint (of the octagon) to the centre $i$.

Step four: Produce the Mobius transformation $\rho$ that preserves the imaginary axis and takes $a$ to $b$ (the edge midpoints). This is easy since you have its hyperbolic translation distance ($2 \ell_m$) in your hands. Excellent - you have your first generator.

Step five: Find the Moebius transformation $\sigma$ that rotates the hyperbolic plane about $i$ through an angle of $2\pi/8$.

Step six: Conjugate. That is, build $\rho_i = \sigma^{-i} \rho \sigma^{i}$. This gives you eight Moebius transformations. These are the images of the four generators of $\pi_1$ and their inverses.

Step seven: Check your work by checking that $\rho_i^{-1} = \rho_{i + 4}$. Also, show that the elements $\rho_i$ satisfy the defining relation of $\pi_1$.

$\endgroup$
10
  • 3
    $\begingroup$ Dear 454229 - If I have answered your original question, you should consider "accepting" the answer. You ask for "other ways"; it depends on what you want... what tools do you know? What kind of answer do you want? You ask "can I get Fuschian groups"; yes, the construction gives a fuchsian group. $\endgroup$
    – Sam Nead
    Commented Dec 5, 2020 at 19:52
  • $\begingroup$ I am trying to figure out the details your answer about the explicit description of the group of deck transformations. In step one, you have advised to take midpoints $a$ and $b$ of the two sides crossing the imaginary axis. Can you please give me more advise about it? $\endgroup$
    – user147962
    Commented Dec 20, 2020 at 10:16
  • $\begingroup$ In step one you are asked to draw a "cartoon" - that is, a picture that is topologically accurate, but not necessarily geometrically accurate. In this step we do not yet know the lengths or angles of the figure we are trying to draw. Nonetheless, we can use our knowledge of hyperbolic geometry to make the cartoon more accurate. Where are the vertices of the octagon with respect to the imaginary axis? If we extend the eight sides of the octagon, where do the geodesics land at infinity (the real axis)? $\endgroup$
    – Sam Nead
    Commented Dec 20, 2020 at 11:39
  • $\begingroup$ This is discussed in step four. Think - what examples of Moebius transformations do you have that (A) preserve the imaginary axis and (B) preserve its (upward) orientation? Once you have the general form of these, you want the one that sends $b$ to $a$ (and so its inverse sends $a$ to $b$). $\endgroup$
    – Sam Nead
    Commented Dec 24, 2020 at 9:50
  • 1
    $\begingroup$ Ah I see what I got wrong... math.stackexchange.com/a/3954296/66607 $\endgroup$
    – ah--
    Commented Jun 11, 2022 at 1:44
5
$\begingroup$

There are many details in this github repository: https://neilstrickland.github.io/genus2/. In particular, there is a monograph at https://neilstrickland.github.io/genus2/genus2.pdf and you can look in Section 4 for discussion of the relevant formulae. The most important are as follows. We fix $b\in(0,1)$ and put $b_+=\sqrt{1+b^2}$ and $b_-=\sqrt{1-b^2}$. We then define Mobius transformations $\beta_k$ for $k\in\mathbb{Z}/8$ by \begin{align*} \beta_0(z) &= \frac{b_+z+1}{z+b_+} \\ \beta_1(z) &= \frac{b_+^3z+(2+i)b^2-i}{((i-2)b^2+i)z+b_+^3} \\ \beta_{2n}(z) &= i^n \beta_0(z/i^n) \\ \beta_{2n+1}(z) &= i^n \beta_1(z/i^n). \end{align*} These generate a group $\Pi$, subject only to the relations $\beta_k\beta_{k+4}=1$ and $\beta_0\beta_1\beta_2\beta_3\beta_4\beta_5\beta_6\beta_7=1$. Moreover $\Pi$ acts freely on the open unit disc, and the quotient has genus two.

As well as the PDF monograph, the repository contains extensive Maple code for working with these surfaces.

$\endgroup$

You must log in to answer this question.