First, some terminology. Let $(S,x)$ be a compact surface of genus $g>0$. A standard collection of loops $\gamma_1,\ldots, \gamma_{2g}$ based at $x$ is a collection of loops that cuts $S$ into a topological disk.

Question. Let $(\Sigma,x)$ is a compact hyperbolic surface (without boundary) of genus $g$ and diameter $d$ with a marked point $x$. How to estimate from above in terms of $g$ and $d$ the number $A$ such that there exists a standard collection of geodesic loops on $\Sigma$ based at $x$, all shorter than $A$?

I would be especially grateful for a reference.


I think that $A=2d$ will work, basically by applying Morse theory to the distance function from $x$. Morse theory for distance functions was originally considered by Gromov (and then Cheeger). For the case of surfaces, see a paper of Gershkovich (also see Gershkovich-Rubinstein for a Morse theory of distance functions).

Imagine taking metric balls $B_r(x)$ of radius $r, 0\leq r\leq d$ about $x$, so that $B_d(x)=S$. These will be the image of a hyperbolic disk $D_r$ at $\tilde{x}$ of radius $r$ in the universal cover $\mathbb{H}^2 \overset{\rho}{\to} S$. For $r$ sufficiently close to 0 ($\leq$ the injectivity radius at $x$), $B_r(x)$ will be an embedded image of the disk $\rho(D_r)$. When $r=injrad_x(S)$, there will be some tangencies of pairs of points in $\rho(\partial D_r(x))$. Taking geodesics in $D_r$ from $\tilde{x}$ to these pairs of points in $\partial D_r(x)$, and projecting by $\rho $ to $S$, we see a collection of loops of length $2r$ based at $x$. Now, continue to let $r$ increase. Each time we see a pair of tangencies between pairs of points in $\partial B_r(x)$, create a "dual" loop. Once one reaches $r=d$, we will obtain a collection of loops cutting $S$ up into disks (one may argue this via Morse theory: self-tangencies are the only way to create 1-handles). Then taking a subset of $2g$ of these loops, one obtains a standard collection, each with length $\leq 2d$.

  • $\begingroup$ Thank you! It looks indeed that this works. When I was asking the question I had in mind that this standard collection is the one from the very standard picture i.e. in $\pi_1(S,x)$ we have $[\gamma_1,\gamma_2]\cdot ...=1$ (like here: math.stackexchange.com/questions/479371/…). I wonder if your argument can be modified to give a good bound for this more restricted type of standard collections as well. $\endgroup$ – aglearner Mar 2 '18 at 20:56
  • 1
    $\begingroup$ @aglearner: I think this might be more difficult (or rather, with worse bounds). Off the top of my head, I see how to get a bound which is of the form $dg^{Cg}$ for some constant $d$, which is superexponential. Extend the standard collection of loops to a triangulation with one vertex. Then one can change it to another triangulation with at most $C g\log g$ flips. Each flip changes the length by at most a factor of 2. See theorem 1.4 of this paper: arxiv.org/abs/1411.4285. I suspect one should be able to obtain better bounds though, maybe exponential (but I'm not conjecturing that)? $\endgroup$ – Ian Agol Mar 3 '18 at 1:42

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.