This question is about knots and links in the 3-sphere. I want to find an example of a "large" knot or link with some special properties. I'm looking for some fairly specific examples, but I'm also interested in related examples that don't precisely address my concern, so "near answers" are welcome.

Here is an example of something I would call a "large" knot.

knot with a genus 2 incompressible surface

It is hyperbolic, but it also has an incompressible genus 2 surface in the knot complement. The genus 2 incompressible surface has a rather nice property, that it bounds a handlebody on one side (in $S^3$ rather than in the knot complement).

I'd like to find a similar example, but what I specifically want is a knot (or link) in $S^3$ which contains two genus 2 closed incompressible surface in the knot/link exterior, such that both surfaces bound handlebodies in $S^3$. The key thing I would like is for the two surfaces to intersect -- they can not be isotoped to be disjoint.

Is this possible? I presume it is, but I haven't seen examples of this type, and examples aren't rapidly coming to mind.

Other similar examples I'd like to see (although not "the" question here) would be knot or link exteriors that have infinitely many genus 2 closed incompressible surfaces -- preferrably bounding handlebodies on one side in $S^3$.

I'd also love to see a knot in $S^3$ with a genus 2 incompressible surface that does not bound a handlebody in $S^3$. Maybe this isn't possible?

  • $\begingroup$ Dear Ryan - Perhaps you could accept one of the answers? This question pops up as unanswered on my screen. :) $\endgroup$
    – Sam Nead
    Apr 30, 2016 at 7:35
  • $\begingroup$ Sam, the problem is my own example was an answer to my question and I had not realized it. I'll accept Allison's, although it was Neil's answer that made me realize what was going on. $\endgroup$ May 6, 2016 at 6:45

3 Answers 3


As Ian Agol mentioned, manifolds with Conway spheres will have genus 2 incompressible surfaces. I'd just like to point out that an easy source of knots whose complements contain Conway spheres are pretzel knots, or more generally, Montesinos knots.

Here is a specific example of the pretzel knot $P(5, -7, 5, -4)$ with two incompressible Conway spheres in blue and red. Two Conway Spheres

Form a pair of closed genus two surfaces by tubing along the arcs to join the boundary components pairwise on each Conway sphere. The resulting genus two surfaces each bound the handlebodies on the side indicated by the colors. (In the image, only one handle has been added to each Conway sphere.)

  • $\begingroup$ May I know how this particular image was generated? Looks quite appealing to my eyes :) $\endgroup$
    – Santiago
    Apr 16, 2019 at 5:30
  • $\begingroup$ I used Inkscape to make the image. $\endgroup$ Apr 22, 2019 at 21:32

A couple of remarks:

Manifolds with Conway spheres will have genus 2 incompressible surfaces, by Theorem 2.0.3 of the proof of the cyclic surgery theorem. When there is a meridional planar incompressible surface (such as a Conway sphere), then one may tube the boundary components together in pairs to obtain a closed incompressible surface. I suspect that if one has a knot with several Conway spheres, then tubing carefully should yield genus 2 surfaces which intersect non-trivially.

If one has infinitely many genus 2 incompressible surfaces, then the knot must have an essential torus (this may be shown using normal surface theory/incompressible branched surfaces). For example, if one has a satellite knot with a Conway sphere which intersects the satellite torus essentially, then one probably has infinitely many genus 2 incompressible surfaces.

If one has a genus 2 surface in R^3, then it must be compressible to one side or the other (say to the inside; an incompressible surface in a knot complement must be incompressible to the outside). Compress the surface to get a torus. If that torus is incompressible to the inside (otherwise, it bounds a handlebody), then it must compress to the outside, and one has a ball-with-knotted-hole. To reconstruct the genus 2 surface from a ball-with-knotted-hole, add a tube to the outside. In order to make the outside incompressible, this tube must run through the knotted hole in some non-trivial way. In this way, one can manufacture a genus 2 surface which does not bound a handlebody, and is incompressible to one side. Put the knot on the inside so it busts the compressing disk to get the desired example.

  • $\begingroup$ Thanks Ian. I seem to always forget about arborescent knots and Conway spheres. You've given me some things to think about. $\endgroup$ Sep 5, 2013 at 4:06

This should be a partial construction. Embed $K$ as a knot in a handlebody such that the knot crosses the neck of the handlebody in two places as in the figure ($T_1$ and $T_2$ are meant to represent sufficiently complicated tangles). We may assume this embedding is hyperbolic by Myers' "Simple Knots in Compact, Orientable 3-Manifolds". Then by the methods of Adams and Reid in "Quasifuchsian surfaces in hyperbolic knot complements", the handlebody can be embedded in $S^3$ such that the exterior is incompressible.

enter image description here

The next paragraph is basically a sallow-follow construction. Cutting the handlebody along its neck we get two solid tori $H_1$ and $H_2$ with an arc (coming from the knot) embedded in each one of them. In the knot exterior the boundary of $H_i$ is a twice punctured torus and the annulus that bounds a neighborhood of the arc in it. Let $S_{1,2}$ be the union the twice punctured torus of $\partial H_1$ and the neighborhood of the arc in $H_2$ and let $S_{2,1}$ be the analog with the indices reversed. Then $S_{1,2}$ and $S_{2,1}$ should be incompressible genus 2 surfaces in a knot complement with the desired properties.

  • $\begingroup$ Ha, so my picture is an example of what I was looking for. I only saw one incompressible genus 2 surface in the complement. Now I see there's at least three. $\endgroup$ Sep 6, 2013 at 0:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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