Fiber surfaces in three-manifolds minimize genus

Here's a cool fact.  If you have a fibered three-manifold, i.e. your three-manifold is a fiber bundle with fiber a surface and base S^1, then the fiber surface minimizes genus among all surfaces in its homology class.  (If the fibered three-manifold has boundary, necessarily a union of tori, then we think about the homology class of the surface rel boundary.)   This says that for fibered knots in S^3, the fiber surface is a minimal genus Seifert surface!  Just for funsies I wanted to put a proof of this here, because I think it's a nice idea that shows up in various guises in low-dimensional topology.  

If you haven't thought much about fibered three-manifolds, some good examples to keep in mind are: \Sigma^2 \times S^1, the complement of a torus knot (here is an awesome visualization), the complement of the figure eight knot, or 0-surgery on any fibered knot in the three-sphere.      

Let Y be a fibered three-manifold with a fiber \Sigma.  Form the infinite cyclic cover of Y, denoted \widetilde{Y}, by cutting Y along \Sigma and gluing infinitely many copies together end to end.  (A fancier way of constructing this is that the Poincare dual of \Sigma is an element of H^1(Y), i.e. a map from \pi_1(Y) to \mathbb{Z}, and this cover is the one corresponding to the kernel of this map.)  Since \Sigma is a fiber, Y - \Sigma is just \Sigma \times I and so \widetilde{Y} is \Sigma \times \mathbb{R}.  Let \Sigma' be any surface in Y which is homologous to \Sigma.  Then, you can check that \Sigma' actually lifts to the cover \widetilde{Y} and in the cover \Sigma and \Sigma' are homologous.  (Here's one perspective, loosely sketched.  Since [\Sigma] = [\Sigma'], their Poincare duals are the same, so the covers associated to each are the same.  But the latter cover is the one where we cut along \Sigma' and stack together infinitely many copies, which \Sigma' certainly lifts to.)  Now, define a map from \Sigma' to \Sigma as follows: include \Sigma' into \Sigma \times \mathbb{R} and project onto \Sigma; because \Sigma' and \Sigma are homologous in \Sigma \times \mathbb{R}, this map induces an isomorphism on H_2.  Get excited because that means this is a degree one map from \Sigma' to \Sigma, and so that means b_1(\Sigma') \geq b_1(\Sigma)!  (See the previous post on degree one maps.)  In other words, g(\Sigma') \geq g(\Sigma).  

This idea is pretty useful in other settings.  Here's an example of a well-known application to Floer homology.  Let Y be a three-manifold and suppose that Z is a homology sphere that embeds in Y \times S^1.  Then, Y is a homology sphere and \dim F(Z) \geq \dim F(Y), where F is basically any Floer homology you can think of.  As before, build the associated infinite cyclic cover of Y \times S^1 which is Y \times \mathbb{R}.  Again, Z will lift and thus embeds in Y \times \mathbb{R} such that Z and Y are homologous (as elements of H_3), and we can get a degree one map from Z to Y which means \pi_1(Z) surjects \pi_1(Y).  Since H_1(Z) = 0, this means that H_1(Y) = 0 as well, so Y is a homology three-sphere.  Now, for the Floer homology part.  By compactness, let's really think of Z as embedding in Y \times I inside Y \times \mathbb{R}; so we can think of Y \times I as the union of a cobordism W_1 from Y to Z and another W_2 from Z to Y.  Now, the map on Floer homology for Y \times I is the identity on F(Y).  But, the functoriality for cobordism maps in Floer homology says that F(Y \times I) = F(W_2) \circ F(W_1).  That means the identity map has to factor through F(Z).  This is only possible if \dim F(Z) \geq \dim F(Y)!  

Comments

Popular posts from this blog

Negotiating job offers

Weird hair and mathematicians

Comparison inequalities