1

Consider a compact oriented surface of genus $g$ with $k$ boundary components, ordered from $1$ to $k$. Next remove $n_i\geq 0$ points from the $i$-th boundary component, for $i=1,...,k.$ For which sequences $(g,n_1,...,n_k)$, the above surface admits a hyperbolic metric whose $i$-th component is bounded by $n_i$ infinite geodesic arcs (going to $n_i$ points at infinity)? $n_i=0$ means that the $i$-th component is a geodesic loop (or, alternatively, a puncture).

The answer is yes for all $n_i\geq 3$ but I am not sure about other cases.

Lee Mosher
  • 120,280
Adam
  • 538
  • 2
  • 11

1 Answers1

1

There is indeed a pretty simple way to derive necessary and sufficient conditions on $(g,n_1,...,n_k)$ for existence of such a hyperbolic structure on your surface $S$.

To derive a necessary condition, let $S$ be your surface equipped with such a hyperbolic structure. Let $DS$ be obtained by doubling $S$ across the boundary, in other words $DS$ is the quotient of a pair of copies of $S$ by identifying the two copies of the boundary using the identity map. The surface $DS$ is a surface of some finite genus $g'$ and some finite number of punctures $p'$, and one can work out formulas for $g'$ and $p'$ in terms of $(g,n_1,...,n_k)$ (I'll indicate more details below of how one does this). Furthermore, by doubling the hyperbolic structure on $S$ one sees that $DS$ possesses a hyperbolic structure, i.e. a complete hyperbolic metric.

Just as for compact surfaces, existence of a complete hyperbolic metric on a punctured surface such as $DS$ is equivalent to saying that $DS$ has negative Euler characteristic: $$\chi(S) = 2 - 2g' - p' < 0 $$ It turns out that this necessary condition is also sufficient, because for any complete hyperbolic structure on $DS$, under the embedding $S \hookrightarrow DS$ one can straighten the boundary of $S$ to get the desired hyperbolic structure on $S$.

The number of punctures on $DS$ is simply $$p' = \sum_{i=1}^k n_i $$ because there is exactly one puncture on $DS$ for each boundary point that was removed to form $S$.

To compute genus, its easier to work with Euler characteristic, and we have $$\chi(S) = 2 - 2g - k - \sum_{i=1}^k n_i $$ The disjoint union of two copies of $S$ has twice the Euler characteristic: $$\chi(S \cup S') = 4 - 4g - 2k - 2 \sum_{i=1}^k n_i $$ Identifying two circle components of the boundaries does not change Euler characteristic, but identifying two arc components reduces the Euler characteristic by $1$: $$\chi(DS) = 4 - 4g - 2k - 3 \sum_{i=1}^k n_i + \#\{i \mid n_i=0\} $$ Since $\chi(DS) = 2 - 2g' - p'$, and since we have a formula for $p'$, we can solve for $g'$.

Lee Mosher
  • 120,280
  • Thanks, Lee! Of course... I feel silly not thinking of it. The only thing is that I got bit confused about your Euler characteristic computation. I would say that chi(S)=2-2g-k (I am not removing points from boundary here) and chi(DS)= 2(2-2g-k) -sum_{i=1}^k n_i. That seems to give a different value. – Adam Aug 20 '19 at 21:12
  • But your surface that supports the hyperbolic structure does have those points removed from the boundary, as you say in your second sentence. That's the surface on which you are asking about a hyperbolic structure, and that's the surface I am calling $S$. – Lee Mosher Aug 21 '19 at 00:04
  • I meant that for the sake of computing chi(DS) it may be easier to remove the points from the double rather then from the original surface (with the same final effect). For g=0,k=1,n_1=2, S is an ideal bigon and DS an annulus. My formula gives chi(DS)=0 and yours chi(DS)=4-40-21-3*2= -4. – Adam Aug 21 '19 at 00:50