Morse Theory: Understanding Theorem on Page 25

What does that tell you about the function on open sets around (0, ..., 0)? What does that tell you about the function on a neighborhood of (0, ..., 0)?5. Show that if p is a saddle point of a function f, then there exist (x_1, ..., x_n) with f(x_1, ..., x_n) > f(0, ..., 0). Hint: use the same e_i as in 4. with e_i = 0 for i = 1, ..., n \ {1, ..., n} (e.g. e_j = 0 for i = 1, ..., n \ {j}). What does that tell you about the function on open
  • #1
daishin
27
0
I started reading a Morse theory by Milnor and am not understanding something.
I am reading the proof of Theorem on page 25:
Let M be a compact manifold and f be a differentiable function on M with only two critical points, both of which are non-degenerate, then M is homeomorphic to a sphere.
We may assume that 0 is mimimum and 1 is maximum of f.
In the proof he says that by Morse lemma for small epsilon f^-1[0, epsilon], f^-1[1-epsilon, 1] are closed n-cell.
I guess he is using the fact that on f^-1(1) and f^-1(0) the morse index is 0.
But Is the fact obvious?
I know 0 and 1 are minimum and maximum respectively. So Hessian is positive at f^-1(0) and negative at f^-1(1).
But why is morse index 0 at these points?
 
Physics news on Phys.org
  • #2
isnt it sort of obvious that if a linear combination of squares has both signs, it can't be an extreme point? i.e. a saddle surface does not have a max or min at the origin.

look up your second year calc criteria for local extrema. you look at more than the determinant of the matrix of second partials (what do you mean by hessian, matrix or determinant?), you look at whether or not it is definite.
 
  • #3
Morse

Could you explain why if a linear combination of squares has both signs, then it can't be an extreme point?
Hessian is a matrix of second partial derivatives.
I suppose you are saying that:
If the hessian is positive definite(i.e all the eigenvalues are positive)at some point p, then f achieves local minimum at p.
If the hessian is negative definite(i.e all the eigenvalues are negative)at some point p, then f achieves local minimum at p.
If the hessian has both positive and negative eigen values at p, then p is a saddle point.

Could you prove why those are true? Or is there any book that I can look up for the proof?
Thanks.
 
  • #4
well it is negative at some places and positive at others, but zero at the origin.those statements are called the second derivative test. they must be in every book.
e.g. p. 759, university calculus, hass, weir, thomas. (proved two sections later.)

but if not, the morse lemma, proved in the morse theory book you are reading by milnor, says that in some coordinates, a non degenerate functiion actually equals its approximating quadratic form. this proves those criteria.

i.e. those criteria are obvious for quadratic forms, and morse lemma (that if the second derivative is non degenrate at a singular point then the function, in some coordinates actually equals its second derivative), implies the criterion holds for the function as well.i think you should do a little exercise to convince yourself that those criteria do hold for quadratic forms, since if you cannot do that exercise then you are reading the wrong book.

if indeed you decide you are reading the wrong book, you might read guillemin and pollack instead, or andrew wallace, differential topology first steps. milnor is very advanced.
 
Last edited:
  • #5


daishin said:
Could you explain why if a linear combination of squares has both signs, then it can't be an extreme point?
Hessian is a matrix of second partial derivatives.
I suppose you are saying that:

If the hessian is positive definite(i.e all the eigenvalues are positive)at some point p, then f achieves local minimum at p.
If the hessian is negative definite(i.e all the eigenvalues are negative)at some point p, then f achieves local minimum at p.
If the hessian has both positive and negative eigen values at p, then p is a saddle point.

Could you prove why those are true? Or is there any book that I can look up for the proof?
Thanks.

1. Look at Morse's lemma (contained in Milnor's book at a page p < 25).

2. Show that x_1^2 + ... + x_n^2 is minimal at (x_1, ..., x_n) = (0, ..., 0).

Hint: the square of a non-zero real number is stricly positive.

3. Show that - (x_1^2 + ... + x_n^2) is maximal at (x_1, ..., x_n) = (0, ..., 0).

Hint: the square of a non-zero real number is stricly positive.

4. Show that e_1 x_1^2 + ... + e_n x_n^2 does not have a minimum/maximum at (x_1, ..., x_n) = (0, ..., 0), where the e_i are (non-zero) real numbers (for i = 1, ..., n) such that there exist 1 <= i, j <= n with e_i > 0 and e_j < 0.

Hint: look at the sign of partial derivatives i and j at (0, ..., 0). What does that tell you about the restriction of the function to the corresponding axis (stricly increasing, stricly decreasing)?
 

1. What is Morse Theory?

Morse Theory is a mathematical concept that studies the topology of manifolds, or higher-dimensional spaces, by using critical points of differentiable functions.

2. How does Morse Theory relate to topology?

Morse Theory provides a way to understand the topology of a manifold by analyzing the critical points of a function defined on the manifold. This allows for a deeper understanding of the structure of the manifold and its topological properties.

3. What is a critical point in Morse Theory?

In Morse Theory, a critical point is a point on a manifold where the derivative of a function is equal to zero. These points play a crucial role in understanding the topology of the manifold.

4. What is the Theorem on Page 25 of Morse Theory?

The Theorem on Page 25 of Morse Theory is a fundamental result that states that the number of critical points of a function on a closed manifold is bounded by the sum of the Betti numbers of the manifold. This theorem has many important applications in topology and geometry.

5. How is Morse Theory used in real-world applications?

Morse Theory has many real-world applications, including in physics, engineering, and computer science. It is used to study the behavior of dynamical systems, analyze the stability of solutions, and understand the geometry of data sets.

Similar threads

  • Differential Geometry
Replies
1
Views
1K
  • Differential Geometry
Replies
1
Views
2K
Replies
9
Views
891
Replies
4
Views
639
Replies
3
Views
2K
  • Programming and Computer Science
Replies
1
Views
1K
Replies
4
Views
2K
Replies
5
Views
312
Replies
2
Views
2K
Back
Top