Problem with understanding Morse's Lemma / Function.

111 Views Asked by At

https://math.stackexchange.com/a/398282/1257548

In this answer it is said that $f:S→\mathbb{R}, (x,y,z)↦y$ is Morse function but I don't see why.

As far as I understand, because function is defined on manifold $S = \{(x,y,z):z^4+(x^2+y^2-1)(2x^2+3y^2-1)=0\}$ then using implicit function theorem we have that $$\frac{dy}{dz} = -\frac{4z^3}{12y^3+10x^2y-8y}$$ and $$\frac{dy}{dx} = -\frac{8x^3+10xy^2-6x}{12y^3+10x^2y-8y}$$ so $$\frac{d^2y}{dz^2}(x,y,0)=0, \frac{d^2y}{dx^2}(0,y,0)=a,\frac{d^2y}{dxdz}(x,y,0)=0.$$ So Hessian of $f$ is degenerated in $(0,−1,0),(0,−\frac{1}{\sqrt{3}},0),(0,\frac{1}{\sqrt{3}},0),(0,1,0)$ thus it can't be Morse function. Even differentiating $f$ like $\frac{df}{dy}=1$ and $\frac{d^2f}{dy^2}=0$ shows that it's also degenerated. I don't know if I don't understand something or if the answer is wrong.

I know that taking the height function as Morse function works on torus and I see that taking $f(x,y,z)=y$ on S should work the same but I don't know why it doesn't work (or if I'm just doing something wrong). And if the function is wrong is it possible to correct it to proper Morse function?

1

There are 1 best solutions below

2
On BEST ANSWER

It helps to look at the picture in the linked answer, looking near $(0, -1, 0)$. You can see that $y$ grows very slowly as a function of either $x$ or $z$ (or even along the direction $x = z$). In fact, looking just in the plane $x = 0$, we have $$ 0 = z^4 + (y^2-1) (3y^2 - 1) $$ At the point $y = -1$, this looks like \begin{align} 0 &= z^4 + (y+1)(y-1)(3y^2 - 1)\\ 0 &\approx z^4 + (y- (-1))(-2)(2)\\ &= z^4 + 4y+4 \end{align} so it's approximately the curve $$ y = (z^4 - 4)/4 $$

The idea of a Morse function is that when you approximate like this, it's supposed to look like a quadratic in the remaining two coordinates. Indeed, if we take the best quadratic approximation to this curve, it's just $$ y = -1 $$ The same general idea applies in any other direction, too, and we see that the best quadratic approximation to this surface near $(0, -1, 0)$ is in fact the plane $y = -1 + 0x + 0z$...which has a Hessian that doesn't tell us anything about the shape.

So your analysis is correct --- it looks as if the other answerer was a little bit too glib.

On the other hand, it does look as if perhaps $y^2-1$ might be a good morse-function near that point ... but then you'd have a problem near $y = 0$.

Anyhow, it sure looks to me as if your conclusion here is correct and this isn't really a a good Morse-theory analysis. Too bad.