MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

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

Given a smooth function $f:\mathbb{R}^n \to \mathbb{R},$ how much is known about the cohomology of the interior of the zero set $f^{-1}\left(0\right)$? I know this can be quite wild, e.g. this amounts to computing the cohomology of the interior of any arbitrary closed subset of $\mathbb{R}^n,$ but is there any hope in describing this cohomology in terms of some properties of $f$?

share|cite|improve this question
Keep in mind that all closed sets are the zero set of a smooth function. – Otis Chodosh Nov 16 '11 at 18:57
Thanks Otis, but in fact I'm aware of this. See my comment to the answer of algori. – David Carchedi Nov 16 '11 at 22:15
the question is a bit too vague. of course a strong condition on f will say something about level and sublevel sets. e.g. if f is convex then the sublevel sets are convex and hence contractible. what do you really know about f in your situation? personally, I have used a variation of the above when f is partially k-convex which implied that sublevel sets have homotopy types of k-dimensional CW-complexes. – Vitali Kapovitch Nov 24 '11 at 21:26
@Vitali: This question is a bit vague on purpose. Essentially, if there happens to be a way to compute the cohomology of the interior of the zero set of each smooth function on R^n, then this would give a way of computing the cohomology of $B\Gamma^n$. – David Carchedi Nov 25 '11 at 16:04

When $f$ is a Morse-Bott function, you may be able to deduce information about the Betti numbers of $f^{-1}(0)$ if you understand the critical sets of $f$.

A bit more generally, when $f$ is minimally degenerate, a notion due to Frances Kirwan, then you can do the same thing.

This may not help, in that it requires understanding all of the other critical sets!

There is one nice class of examples where this variant of Morse theory does work very well: when $f$ is the norm-square of a momentum map for a Hamiltonian $G$ action on a manifold $M$, then $f$ is minimally degenerate and $f^{-1}(0)$ is the $0$-level set of the momentum map. The critical sets are subsets of fixed sets $M^T$ for certain abelian subgroups of $G$. When $0$ is a regular value of the momentum map, $G$ acts locally freely on the level set, and the cohomology of the quotient $f^{-1}(0)/G$ can be computable. I don't know if the cohomology of $f^{-1}(0)$ itself is computable, but when the $G$-action is free, you know that it is fiber bundle over the quotient $f^{-1}(0)/G$.

A reference: Frances Kirwan, Cohomology of quotients in symplectic and algebraic geometry. Mathematical Notes, 31. Princeton University Press, Princeton, NJ, 1984.

share|cite|improve this answer
Thanks Tara. One small thing is, (unfortunately) I need to compute this for the interior of the zero set. Maybe this destroys the nice properties... – David Carchedi Nov 25 '11 at 16:05

It is a theorem of Whitney's that every closed subset $X$ of a smooth manifold $M$ is the zero locus of a smooth function on $M$. The idea of the proof is as follows: cover the complement $M\setminus X$ by a locally finite system of open balls with compact closures in $M\setminus X$ and let $(f_k)$ be the corresponding partition of unity. Set $$g_k=f_k\frac{1}{2^k}max_{|\alpha|\leq k,x\in M}\left|\frac{\partial^\alpha}{\partial x^\alpha}f(x)\right|.$$ Then as $k\to \infty$, for each fixed order $n$, each $x\in M$ and each $\alpha$ with $|\alpha|=n$ we will have $\left|\frac{\partial^\alpha}{\partial x^\alpha}g_k(x)\right|\leq \frac{1}{2^k}$, so $f=\sum g_k$ has the required property.

share|cite|improve this answer
Dear Algori. Thanks, but in fact, this is why I said "this amounts to computing the cohomology of the interior of any arbitrary closed subset of $mathbb{R}^n$." I am well aware of this result, however, it does not mean necessarily that the properties of the function $f$ cannot be used to describe the cohomology. I'm guessing it still can't, but, I wanted to see what was known. – David Carchedi Nov 16 '11 at 22:15
Dear David -- you are right. A spectacular example of dyslexia on my part. – algori Nov 17 '11 at 2:05

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.