10
$\begingroup$

I'm studying Robert Ghrist papers on integration against Euler Characteristic. I am particularly interested in the relation with Morse Theory. I am trying to understand the proof of Theorem 25.1 (page 52) of this preprint, which I reproduce in an extended version (the parts I understand):

Theorem. Suppose $h\colon M\to \mathbb{R}$ is a function integrable against Euler characteristic and a Morse function on the closed (compact without boundary) manifold $M$. Then:

$$\int_M [h] \: d \chi=\sum_{p\in Cr(h)}(-1)^{n-\mu(p)}h(p)$$

where $Cr(h)$ is the set of critical points of $h$.

Remark. We follow the usual conventions of Milnor's book in Morse theory.

The proof as I understand it. First we use a previous result (Proposition 24.8 of the same paper) which says that:

$$\int_M [h] \: d \chi=\int_{s=0}^{\infty}\chi\{h\ge s\}-\chi\{h<-s\}\: ds.$$

Now, using Theorem 3.2 from Milnor's book I know that $\chi\{h\le s\}$ is piecewise constant and only changes at critical values of $h$. Moreover, the change is due to the addition of a cell of the dimension of the Morse index of that critical value. So, the change in the Euler Characteristic at a critical value $s=h(p)$ is: $$\chi\{h\le s + \epsilon\}-\chi\{h\le s - \epsilon\}=(-1)^{\mu(p)}.$$

At this point I get lost.

So I insert as a picture the proof given in the paper, and a similar proof of the same result in another preprint(page 8): Proof in the first paper

Gap I don't catch in the second paper

Any help would be appreciated! By the way, there is a proof which involves Stratified Morse Theory, but I am interested in this one using elementary methods.

$\endgroup$
2
$\begingroup$

I glanced at the paper briefly. Let me try to explain what I understand. Let us suppose that M is compact and without boundary. Let $f: M \to \mathbb{R}$ be a Morse function. Let us further suppose for simplicity that $f$ is positive on $M$, that is it only takes positive values. We could always translate $f$ to make this the case, since $M$ is compact.

In that case $\{h< -s\} = \emptyset$ for all non-negative $s$. So the Euler characteristic of that portion is always zero, so by prop 24.8 we then have

$$\int_m h [d \chi] = \int_{s=0}^\infty \chi(h \geq s) ds$$

Let $p_i$ be the critical points of the Morse function. And assume for simplicity that these take distinct critical values $h(p_i)$. What does $\chi(h \geq s)$ look like? It starts at $s=0$ with the Euler characteristic of the manifold $\chi(M)$. Then it is locally constant with jumps at exactly the critical values $s = h(p_i)$. These jumps are by one unit up or one unit down according to the index via $(-1)^{\mu(p_i) + 1}$. The extra factor of $-1$ is because we are taking away the handle as $s$ increases.

In other words we can write this as

$$ \chi(h \geq s) = \chi(M) + \sum_{i, h(p_i) < s} (-1)^{\mu(p_i + 1)}g_{[h(p_i), \infty)} $$

where $g_{[h(p_i), \infty)}$ is the "jump function" which takes value one on the interval $[h(p_i),\infty)$ and zero otherwise.

It is useful to fix a large $T > \max\{h(p_i)\}$ and consider the definite integral, which we can compute:

$$\int_{s=0}^T \chi(h \geq s) ds = T \cdot \chi(M) - \sum_i (-1)^{\mu(p_i)} \int_{s=0}^T g_{[h(p_i), \infty)} ds $$ $$= T \cdot \chi(M) -\sum_i (-1)^{\mu(p_i)} (T - h(p_i))$$ $$ =T \cdot \chi(M) - T \cdot (\sum_i (-1)^{\mu(p_i)} ) + \sum_i (-1)^{\mu(p_i)} h(p_i) $$ The first two terms cancel and this gives the second claimed formula 25.2. The first one is the same argument with the function $h$ flipped over.

$\endgroup$
  • $\begingroup$ First of all, thank you very much for your time. Could you elaborate a bit more on flipping over the function $h$, please? It seems to me that if you flip the function, then you are not proving the result for your arbitrary chosen function $h$. Thanks in advance! And, I have thought about this and, It seems we are arriving to the results exchanged, I mean, with the expression for lower integral we arrive to the result for the upper one and the other way around. Am I right? $\endgroup$ – D1811994 Jun 4 '17 at 0:06

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.