MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $\gamma_0$ and $\gamma_1$ be two simple closed curves in a closed surface $S$.

What is the maximum Euler characteristic of a compact properly embedded surface $\Sigma \subset S\times [0,1]$ such that $\partial \Sigma = \gamma_0 \times \{0\} \cup \gamma_1 \times \{1\}$?

Of course, in order for such a surface $\Sigma$ to exist, the two curves $\gamma_0$ and $\gamma_1$ must represent the same class in $H_1(S,\mathbb Z_2)$. Note that $\Sigma$ may be non-orientable.

If $2n$ is the minimum geometric intersection number of $\gamma_1$ and $\gamma_2$, it is easy to construct a $\Sigma$ with $\chi(\Sigma)\geqslant \chi(S) - n$. Is there a converse estimate of this kind? Do we have $\chi(\Sigma) \leqslant -n$ when $S$ is a torus?

share|cite|improve this question
what is the "geometric intersection number" and why is its minimum even? – Vivek Shende Feb 14 '11 at 23:48
@Vivek : The geometric intersection number of two curves $x$ and $y$ on a surface is the minimum of the set of numbers $#(x' \cap y')$, where $x'$ and $y'$ range over curves homotopic to $x$ and $y$. In the case at hand, this has to be even because the mod $2$ algebraic intersection number of the two curves in question is $0$, so mod $2$ there have to be the same number of positive intersections as negative intersections. – Andy Putman Feb 15 '11 at 2:21
up vote 7 down vote accepted

Here is a graph $\mathcal{G}(S,x)$ associated to a surface and homology class (similar to the 1-skeleton of the curve complex): For a fixed class in $x\in H_1(S;\mathbb{Z}_2)$, consider isotopy classes of embedded multicurves representing the homology class $x$ (one may assume no components are parallel and there are no trivial components). Make this collection the vertices of the graph $\mathcal{G}(S,x)$. Connect two vertices $A, B$ to be adjacent in the graph $\mathcal{G}(S,x)$ if $A\cup B$ are disjointly embedded, and after removing all parallel curves of $A\cup B$, the remaining components bound a pair of pants or a twice-punctured projective plane (this second can only happen if $S$ is non-orientable). I'm not sure if such a complex has been defined before, but there is a somewhat analogous complex defined in the integral homology case by Bestvina, Bux, and Margalit. Also, this is related to a technique of Hatcher-Thurston to undertand surfaces in two-bridge knot complements.

I claim that the maximal Euler characteristic of a surface bounding $\gamma_0\times 0 \cup \gamma_1\times 1$ is the negative of the distance between $\gamma_0$ and $\gamma_1$ in $\mathcal{G}(S,x)$. Put the product metric on $S\times [0,1]$, and make $\Sigma$ into a minimal surface with respect to this metric (Theorem 6.12 of Hass-Scott). Then $S\times t, t\in [0,1]$ gives a foliation by totally geodesic surfaces, and by the maximum principle, they can be tangent to $\Sigma$ in only saddle tangencies (see an argument of Hass, we will assume things are perturbed to be generaic). Thus, for all but finitely many $t$, $S\times t$ meets $\Sigma$ in a finite collection of curves, giving rise to a vertex of $\mathcal{G}(S,x)$. As one passes through a tangency point $S\times t_0$ (assuming things are generic), the intersection changes by a saddle move, giving a surface in $S$ of Euler characteristic $-1$ bounding the curves before and after the tangency. There can never be a closed trivial curve occurring, because this would give rise to a center tangency. Thus, each saddle tangency gives an edge between the adjacent vertices of $\mathcal{G}(S,x)$, and therefore $\Sigma$ gives a path between $\gamma_0$ and $\gamma_1$ in $\mathcal{G}(S,x)$. Conversely, any such path gives rise to a surface.

Of course, there will be many geodesics connecting $\gamma_0$ and $\gamma_1$ in $\mathcal{G}(S,x)$, given by any Morse function on $(\Sigma, \gamma_0,\gamma_1)$ with only index 1 critical points, and I don't expect the distance function to be easy to compute (probably one should use normal surface theory to compute it).

share|cite|improve this answer
In your definition of the graph is it equivalent (and perhaps easier) to say that $A$ and $B$ and joined by an edge iff they are related by a saddle move? – Kevin Walker May 18 '11 at 21:59
yes, that's equivalent - in fact, as one varies the parameter t, one will see a movie isotoping multicurves with finitely many saddle moves. – Ian Agol May 18 '11 at 22:07
Thanks Ian. Does $\Sigma$ need to be $\partial$-incompressible to guarantee that one can find a minimal representative, or is incompressibility enough? In fact my question was motivated precisely by normal surface theory: in some cases normal surfaces of highest $\chi$ are "seen" by quantum invariants by using the techniques described in a (nice) paper of Frohman - Bartoszynska and I was wondering whether this minimal distance between curves could be computed by using some Turaev-Viro invariant. – Bruno Martelli May 19 '11 at 14:35
This is Theorem 6.12 of Hass-Scott:… – Ian Agol May 19 '11 at 15:55

For orientable surfaces of genus at least $2$, pretty sharp bounds are obtained in the unpublished PhD thesis of Ingrid Irmer, which is available here.

share|cite|improve this answer
Nice reference, thanks! – Bruno Martelli Feb 15 '11 at 8:24
The thesis deals with oriented curves and surfaces, but probably the same techniques apply in the non-oriented case. – Bruno Martelli Feb 15 '11 at 14:00

Let $a$, $b$ and $x$ be curves on $S$ such that

  • $a$ and $x$ intersect once,
  • $b$ and $x$ intersect once, and
  • $a$ and $b$ have a large algebraic intersection number.

(I assume that $S$ is orientable.) Then there is a surface $\Sigma_a$ connecting $x$ and $x+2a$ with $\chi(\Sigma_a) = -1$ and a similar surface $\Sigma_b$ connecting $x$ and $x+2b$. Combining $\Sigma_a$ and $\Sigma_b$ we get a surface of Euler characteristic $-2$ connecting $x+2a$ and $x+2b$. Thus the geometric intersection number of two curves does not give a lower bound on the complexity of a surface joining them.

(In your original question you ask if $\chi(\Sigma) \le n$, but $n$ is positive and $\chi(\Sigma)$ is negative, so I assume you meant $|\chi(\Sigma)|$.)

share|cite|improve this answer
Thank you (I have corrected the sign) – Bruno Martelli Feb 15 '11 at 8:24

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.