Canonical submersion theorem

Let f:M→N be a submersion at p∈M, then m=dimM≥n=dimN and there exist charts (φ1,U1,V1) around p and (ψ1,X1,Y1) around f(p) such that

ψ1∘f∘φ1−1=π|V1.

Canonical submersion theorem

REMARK: As can be deduced from the following proof, a submersion gives to M the structure of a product around p since we can construct a diffeomorphism

tp:p∈U2⟶f(U2)×Ipm−n

being Ip⊆R an interval depending on p.
Proof of the Canonical Submersion Theorem. Take a chart {φ,U,V} near p and a chart {ψ,X,Y} near f(p) so that f(U)⊂X. Since f is a submersion,

d(ψ∘f∘φ−1)φ(p)=dψq∘dfp∘dφφ(p)−1:Tφ(p)V=Rm→Tψ(q)Y=Rn

is surjective. Denote F=ψ∘f∘φ−1. Then the Jacobian matrix (∂Fi∂xj) is an n×m matrix of rank n at φ(p). By reordering the coordinates if necessary, we may assume the sub-matrix

(∂Fi∂xj),1≤i≤n,1≤j≤n

is nonsingular at φ(p). Note that this re-ordering procedure can be done by modifying (ψ,X,Y) to another chart (ψ1,X1,Y1), and thus we really have F=ψ1∘f∘φ−1. Define

G:V→Rm,(x1,⋯,xm)↦(F1,⋯,Fn,xn+1,⋯,xm).

Then obviously dGφ(p) is nonsingular. By the inverse function theorem, there is a neighborhood V0 of φ(p) so that G is a diffeomorphism from V0 to G(V0). Let H be the inverse of G on G(V0). Note that F=π∘G. Let U1=φ−1(V0), V1=G(V0), and φ1=G∘φ. Then (φ1,U1,V1) is a chart near p, and

ψ1∘f∘φ1−1=ψ1∘f∘(φ−1∘H)=F∘H=π∘G∘H=π.◻

In some sense this is "dual" to the canonical immersion theorem.

The proof use the inverse function theorem.