0% found this document useful (0 votes)
24 views

Lecture 3

1) The inverse function theorem is extended from Rn to manifolds, stating that if the differential of a smooth map f between manifolds is invertible at a point x, then f is a local diffeomorphism near x. 2) A point x in the domain is called critical if the differential of f at x is not surjective, otherwise it is regular. A value whose preimage contains only regular points is called a regular value. 3) If the target has lower dimension than the domain, then the differential can never be surjective and all points are critical. 4) For manifolds of the same dimension, the preimage of a regular value consists of finitely many isolated

Uploaded by

ThetaOmega
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
24 views

Lecture 3

1) The inverse function theorem is extended from Rn to manifolds, stating that if the differential of a smooth map f between manifolds is invertible at a point x, then f is a local diffeomorphism near x. 2) A point x in the domain is called critical if the differential of f at x is not surjective, otherwise it is regular. A value whose preimage contains only regular points is called a regular value. 3) If the target has lower dimension than the domain, then the differential can never be surjective and all points are critical. 4) For manifolds of the same dimension, the preimage of a regular value consists of finitely many isolated

Uploaded by

ThetaOmega
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

LECTURE 3

Exercise 19 (Alternative definition). Given Ωx ⊂ Rn open around x and F : Ωx → Y


smooth (in the usual sense) with f|Ωx ∩X = F|Ωx ∩X , we can define Dfx : Tx X → Tf (x) Y as
the restriction of DFx : Rn → Rm to the vector space Tx X. Prove that it coincides with the
previous definition (which proves in particular that (DFx )|Tx X is the same for all extensions).

2.3 Inverse function theorem on manifolds


Let us start with the inverse function theorem in Rn :

Theorem 20 (Revision from Analysis II). Given U ⊂ Rn open, x ∈ U and f : U → Rn


smooth so that Dfx : Rn → Rn invertible, then f is a local diffeomorphism around x: there
is Ũ ⊂ U open containing x so that f|Ũ : Ũ → f (Ũ ) diffeomorphism.

We now extend this fundamental result to manifolds:

Theorem 21 (Inverse function theorem for manifolds). Given X ⊂ Rn and Y ⊂ Rm smooth


manifolds and f : X → Y smooth so that Dfx : Tx X → Tf (x) Y invertible for some x ∈ X,
then f is a local diffeomorphism: there is Ux ⊂ X open in X around x so that f|Ux : Ux →
f (Ux ) is a diffeomorphism.

Proof. Note first that the hypothesis implies dim X = dim Tx X = dim Tf (x) Y = dim Y so
the manifolds have same dimension k. Consider then parametrisations ϕ around x and ψ
around f (x) and the corresponding transition map θ. Then, given that the parametrisations
have invertible differential and Dθ0 = (Dψ0 )−1 ◦ Dfx ◦ Dϕ0 , we deduce that Dθ0 : Rk → Rk
is invertible and the inverse function theorem in Rk implies that θ is a local diffeomorphism
around 0, which implies f = ψ◦θ◦ϕ−1 is a local diffeomorphism around x by composition.

2.4 Regular values and Sard theorem


So far we have extended concepts to manifolds, we now move on to some new concepts.

Definition 22. Given X and Y manifolds and f : X → Y a smooth map, we say that x ∈ X
is a critical point of f if Dfx : Tx X → Tf (x) Y is not surjective. Otherwise it is a regular
point. We write C ⊂ X for the set of critical points. A critical value is the image of the
critical point. All points that are not critical values are called regular values.

Remark 23. Note that the image of a regular point is not necessarily a regular value: indeed
all pre-image of a regular value need be regular points. Note also that values in Y \ f (X) are
regular according to this definition. Note finally that if dim X < dim Y , the differential Dfx
is never surjective and C = X. Then the only regular values are Y \ f (X).

Theorem 24 (Pre-image theorem). Given X and Y manifolds of dimensions k and ℓ re-


spectively, with k ≥ ℓ, f : X → Y a smooth map, and y ∈ Y a regular value of f in f (X),
then f −1 (y) is a submanifold of X of dimension dim f −1 (y) = dim X − dim Y .

6
Proof. Let x ∈ f −1 (y) (it exists because we assume y ∈ f (X)). Since y is a regular value,
Dfx : Tx X → Tf (x) Y surjective. Let K = KerDfx . This is a subspace of Tx X with dimension
k − ℓ. Assume X ⊂ Rn and consider a linear map T : Rn → Rk−ℓ so that KerT ∩ K = {0}
in Rn (for instance one can choose T to be the identity on K and zero on K ⊥ ). Consider
then the map F : X → Y × Rk−ℓ defined by F (x) = (f (x), T (x)). It is smooth since each
component is, and the differential is DFx = (Dfx , T ) : Rk → Rℓ × Rk−ℓ (since T is linear,
DT = T ). The construction of T implies that DFx is injective so invertible, hence the inverse
function theorem implies that F is a local diffeomorphism around x. It maps a neighbourhood
U ⊂ X of x to a neighbourhood V ⊂ Y × Rk−ℓ of (y, T (x)). Therefore F −1 restricted to
V ∩ ({y} × Rk−ℓ ) provides a parametrisation of U ∩ f −1 (y), which shows that f −1 (y) is a
manifold of dimension k − ℓ.

Remark 25. Note that the previous proof shows in fact that at x ∈ Y := f −1 (y) we have
−1
Tx Y = KerDfx = K. Indeed F|V∩({y}×R k−ℓ ) is a local parametrisation around x, so

−1
Tx Y = DF(y,T (x)) ({0} × R
k−ℓ
) = (Dfx , T )−1 ({0} × Rk−ℓ ) = Ker(Dfx ).

The case where the dimensions are equal is particularly important:

Theorem 26 (Stack of records theorem). Given X and Y manifolds of same dimension,


with X compact, f : X → Y a smooth map, and y ∈ Y a regular value of f in f (X), then
f −1 (y)) is a finite collection of points, say f −1 (y) = {x1 , . . . , xI } with I ∈ N∗ . Moreover
there is V open around y in Y so that f −1 (V) is a disjoint union of Ui for i = 1, . . . , I, where
each Ui is open around xi and f maps each Ui diffeomorphically to V.

Remark 27. An important consequence of the stack of records theorem is that for f smooth
map between manifolds of same dimension, the number #f −1 (y) is locally constant as y
ranges over regular values of f (and the set of regular values is open by the inverse fct theo).

Proof. Applying the inverse function theorem near each point in the pre-image shows that
f −1 (y) is a collection of isolated points. If X is compact such collection is then finite due to the
compactness. (Note that it is compatible with the convention that a 0-dimensional manifold
means an isolated collection of points.) Let f −1 (y) = {x1 , . . . , xI } and for each 1 ≤ i ≤ I, by
the inverse function theorem we can pick Ũi open around xi so that f maps diffeomorphically
Ũi to the open set f (Ũi ). The subset X \ (∪Ii=1 Ũi ) is compact and f (X \ (∪Ii=1 Ũi )) is therefore
compact and does not contain y by construction. We define

V := ∩Ii=1 f (Ũi ) \ f (X \ ∪Ii=1 Ũi )

which is open around y in Y . Finally we take Ui := Ũi ∩ f −1 (V) which is open around xi .

We had seen two basic “lego” ways of building new manifolds: open subset and product.
The pre-image theorem gives a new more interesting procedure.
Pn+1 2
Example 28. Let f : Rn+1 → R defined by f (x) = i=1 xi . Then Dfx (y) = 2(x · y).
Therefore the only critical point is x = 0 and the corresponding critical value is y = 0. And
y = 1 is a regular value, so that the pre-image theorem shows that f −1 (1) = Sn is a smooth
manifold of dimension (n + 1) − 1 = n.

You might also like