I have a basic question about the Heegaard diagrams involved in providing a framework for calculation of Floer-Homology of three-manifolds.

Typically such diagrams look like Figure 1 and Figure 2 here or these two (Image1 or Image2) from researchgate network.

And I would thankful if anybody could try to explain how to "read" this diagrams to a non-topologist.

We see a compact surface, which is probably assumed to be the boundary of certain three-manifold, and theory of compact surfaces states that this surface is up to homeomorphism always a connect sum of $g$ tori for $g \ge 1$. Looking at images in Image2 in last two links we find two sets of $g$ disjoint curves ("team red" $\alpha_0,..., \alpha_g$ and "team blue" $\beta_0,..., \beta_g$).

Now how does this information provide instructions to build a three-manifold?

My non-expert guess is that this data tells us: Start with two identical disjoint three-manifolds which have two $g$-tori as surfaces and the data provided by these Heegaard diagrams is nothing else than instructions how to glue the two three-manifolds along the surfaces. The instruction says probably that the curve $\alpha_i$ of one surface has to be glued homeomorphically with $\beta_i$ for other surface. And seemingly if we know all pairs of curves $\alpha_i$ and $\beta_i$ are glued together, then the gluing of the two surfaces is already uniquely determined up to homeomorphism and therefore we know how to glue the two disjoint three-manifold also the boundary.

Is this exactly the correct way to read a Heegaard diagram? Does there exist a more conventional way? Sorry, if the question is too elementary, I'm not an algebraic topologist and the motivation of this question is pure curiosity.

  • 6
    $\begingroup$ Start with $\Sigma \times I$. For each $\alpha$ curve (which you ought to thicken into an annulus, which you can do in a way which is unique up too isotopy), you glue on a copy of $D^2 \times I$ by pasting $S^1 \times I$ along the thickened $\alpha$ curve inside of $\Sigma \times \{0\}$. Do the same thing for the $\beta$ curves inside of $\Sigma \times \{1\}$. What you have produced is a compact 3-manifold with boundary $S^2 \sqcup S^2$; glue on 3-balls to make this a closed manifold. Gompf and Stipsicz have a good discussion. $\endgroup$
    – mme
    Aug 27 '20 at 1:38
  • 7
    $\begingroup$ I’ve taken the liberty of editing your title, because I don’t think anyone who studies 3-manifolds calls them “3-folds”. $\endgroup$
    – HJRW
    Aug 27 '20 at 6:43
  • 2
    $\begingroup$ You may also look at the first three chapters of beautiful paper of P. Ozsváth and Z. Szabó: math.mit.edu/~petero/Introduction.pdf $\endgroup$ Aug 27 '20 at 17:01
  • $\begingroup$ @MikeMiller: You mean it in the sense that you glue along each thickened tubulus of every $\alpha_i \subset \Sigma \times {0}$ a $2$-handle and them the same game for $\beta_j \subset \Sigma \times {1}$. What I still not understand is why the $3$-manifold obtained by this construction has boundary $S^2 \sqcup S^2$? $\endgroup$ Aug 28 '20 at 20:42
  • $\begingroup$ You are at each stage performing a handle attachment, which changes the boundary by a surgery along the attaching sphere. Check that this process reduces the genus by 1 every time you do a handle attachment. You do it g times. Again, G&S is a good source. $\endgroup$
    – mme
    Aug 28 '20 at 21:18

Chapter four of "Knots, Links, Braids and 3-Manifolds" by Prasolov and Sossinsky gives a highly readable (and nicely illustrated) introduction to three-manifolds via Heegaard splittings. Another, more classical, reference is chapter two of "Three-manifolds" by Hempel. Note that Hempel calls handlebodies "cubes with handles".


You are probably familiar with definitions and theorems. But I prefer to write those for completeness. And also excuse for a paint-like drawing. I hope that they will be useful.

A handlebody of genus $g$ is a $3$-manifold constructed from the standart $3$-ball $B^3$ by adding $g$ copies of $1$-handles $B^2 \times B^1$. It is denoted by $H$ and $\partial H \approx \Sigma_g$ where $\Sigma_g$ is a genus $g$ surface, see the following figures.enter image description here

Let $Y$ be a $3$-manifold. A Heegaard splitting of $Y$ is a decomposition of $Y$ such that

  • $Y=H_0 \cup H_1$ where $H_0$ and $H_1$ are handlebodies,
  • $\partial H_0 = \partial H_1 = \Sigma_g$.

Theorem(Singer, 1933): Any closed oriented 3-manifold $Y$ admits a Heegaard splitting.

The genus $g$-surface $\Sigma_g$ is constructed from $S^2 = \mathbb{R}^2 \cup \{ \infty \}$ by attaching $g$ copies of $1$-handles, where we draw attaching spheres as pairs of matching disks.

enter image description here

So the followings are Heegaard splittings of $S^3$ and $S^1 \times S^2$ respectively:

enter image description here

The following is for a Heegaard diagram of lens space $L(5,2)$:

enter image description here

And the last scheme is for the famous Poincaré homology sphere $\Sigma(2,3,5)$:

enter image description here


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.