Morse lemma
Let M be a smooth n-dimensional manifold, and f:M→ℝ a smooth map. We denote by Crit(f) the set of critical points of f, i.e.
Crit(f)={p∈M|(f*)p=0} |
For each p∈Crit(f) we denote by f**:TpM×TpM→ℝ (or (f**)p if p need to be specified) the bilinear map
f**(v,w)=v(˜w(f))=w(˜v(f)),∀v,w∈TpM, |
where ˜v,˜w∈𝒯(M) are
smooth vector fields such that ˜vp=v and ˜wp=w.
This is a good definition. In fact p∈Crit(f) implies
v(˜w(f))-w(˜v(f))=(˜v(f),˜w(f))p=f*(˜v,˜w)p=0. |
In smooth local coordinates
x1,…,xn on a neighborhood U of p we have
f**(∂∂xi|p,∂∂xj|p)=∂2f∂xi∂xj(p). |
A critical point p∈Crit(f) is called non degenerate when the matrix
(∂2f∂xi∂xj(p))i,j∈{1,…,n} |
is non singular. We can equivalently
express this condition without the use of local coordinates saying
that p∈Crit(f) is non degenerate when for each v∈TpM∖{0} the linear functional f**(v,⋅)∈Hom(TpM,ℝ)
is not zero, i.e. there exists w such that f**(v,w)≠0.
We recall that the index of a bilinear functional H:V×V→ℝ is the dimension Index(H) of a maximal linear
subspace W⊆V such that H is negative definite
on
W×W.
Theorem 1 (Morse lemma)
Let f:M→R be a smooth map. For each non degenerate p∈Crit(f) there exists a neighborhood U of p and smooth coordinates x=(x1,…,xn) on U such that x(p)=0 and
f|U=f(p)-(x1)2-…-(xλ)2+(xλ+1)2+…+(xn)2, |
where λ=Index((f**)p).
Title | Morse lemma |
---|---|
Canonical name | MorseLemma |
Date of creation | 2013-03-22 13:53:12 |
Last modified on | 2013-03-22 13:53:12 |
Owner | matte (1858) |
Last modified by | matte (1858) |
Numerical id | 18 |
Author | matte (1858) |
Entry type | Theorem |
Classification | msc 58E05 |
Defines | non degenerate critical point |
Defines | index of a bilinear map |