Given a compact oriented aspherical $3$--manifold $M$ with torus boundary $\partial M\simeq T^2$ (e.g. a knot complement), the condition that the images in $\pi_1 M$ of basis $x,y\in \pi_1 T^2$ under the inclusion generate the boundary of a compact oriented $3$-manifold with fundamental group $\pi_1 M$ (the $3$-manifold $M$ itself, to be precise) is algebraically encoded by the vanishing of their Pontryagin product $\langle x,y\rangle\in H_2 (\pi_1 M)$, which you can think of as the fundamental class of the torus under the inclusion. Recall that the Pontryagin product sends two commuting elements of a group to an element of $H_2$ of the group.

Question: Given a compact oriented aspherical $3$--manifold $M$ with boundary a closed surface $\partial M \simeq \Sigma$ of genus $g\geq 2$, how can I express the statement that $\Sigma$ bounds a $3$--manifold (the $3$-manifold $M$) on the level of fundamental groups? What condition must images under the inclusion of a basis for $\pi_1 \Sigma$ satisfy? How can I express the fundamental class of $\Sigma$ in terms of elements of $\pi_1$, if I know what $\Sigma$ is topologically?

I'm trying to identify some homomorphs of the fundamental group of a knotted theta, with peripheral data, and the answer to this question would, I presume, be something all such homomorphs would have to satisfy.

  • 3
    $\begingroup$ Isn't $\pi_1(M)$ non-abelian in general? If so, how do you define the Pontryagin product on its homology? It seems you are just looking at the image of the fundamental class of $\partial M$ under inclusion into $M$, is this true? $\endgroup$
    – Mark Grant
    Aug 3, 2012 at 9:33
  • 1
    $\begingroup$ Question edited to clarify this point. $\endgroup$ Aug 5, 2012 at 9:46
  • $\begingroup$ I don't really understand your question: if $M$ is e.g. a knot complement, then $H_2(\pi_1(M))=0$, so the Pontryagin product vanishes for any immersed torus. So this doesn't characterize the peripheral torus. $\endgroup$
    – Ian Agol
    Aug 6, 2012 at 17:03
  • $\begingroup$ Agol: True- for the fundamental group itself, it's not interesting. The interest comes when one looks at homomorphs G of pi_1(M). These might have non-trivian H_2, but still the image of the Pontryagin product of the images of x and y under the homomorphism will vanish (I think this is noted in "Homomorphs of Knot Groups" by Johnson, where the significance of this condition is discussed- Johnson-Livingston says if you add a condition, this characterizes knot homomorphs). I'm asking about the analogous statement for a higher genus surface. But you're right- it isn't clearly stated. $\endgroup$ Aug 6, 2012 at 18:04

2 Answers 2


An obvious necessary condition is that the pair $(M,\Sigma)$ satisfies Lefschetz duality. Since $M$ and $\Sigma$ are aspherical, this can be reformulated in terms of group cohomology.

Let $\Gamma=\pi_1(M)$ and $S=\pi_1(\Sigma)$. Assume that the inclusion $\Sigma\subset M$ is $\pi_1$-injective, so that $S$ is a subgroup of $\Gamma$. Then the pair $(\Gamma,S)$ is a Poincaré duality pair of dimension $3$. See Definition 4.6 of this survey by Davis, which refers to papers of Bieri and Eckmann.

This should place some pretty serious restrictions on how the homology of $\Sigma$ sits in the homology of $M$, from which you might be able to extract a useful answer to your question.


Is there such a statement? I don't think so. As long as your manifold supports homotopy equivalences that do not restrict to homotopy equivalences on the boundary you have this problem. So manifolds with compressible boundary are a problem like high genus handelbodies - the handelbody mapping class group is very different from the outer automorphism group of a free group. So there is no such result for these manifolds. You'll want a more restrictive class of three manifolds to get a statement.

  • $\begingroup$ Thanks for this! This is useful. I'm willing to restrict to knotted theta complements in $S^3$. Do you know whether one can write down, at least in this case, at least a necessary condition for "Σ is a boundary" (what the vanishing of the Pontryagin product says in the torus case) in terms of the basis for Σ inside pi_1 of the manifold? $\endgroup$ Aug 6, 2012 at 4:31
  • $\begingroup$ There certainly isn't such a statement for the trivial theta curve, since the complement is a handlebody. $\endgroup$ Aug 6, 2012 at 23:35

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.