Inversefunction THM

A PROOF OF THE INVERSE FUNCTION THEOREM
SUPPLEMENTAL NOTES FOR MATH 703, FALL 2005
pPn
First we fix some notation. For x ∈ Rn we denote by kxk = 2
i=1 |xi | the Euclidean
n m
norm of x. Let G ⊂ R be an open set and let f : G → R be differentiable at x0 ∈ G,
i.e., there exists a unique linear map Df (x0 ) : Rn → Rm such that
f (x0 + h) − f (x0 ) − Df (x0 )(h) = (h)khk,
where (h) → 0 in Rm as h → 0 in Rn . If we write f(x) = (f1 (x), · · · , fm (x)), then Df (x0 )
∂f
has with respect to the standard basis the matrix ∂xji (x0 ) .
Lemma 1. Let G ⊂n Rn be an open and convex set and leto f : G → Rm be differentiable

∂f
on G. Let M = sup ∂xji (x) : x ∈ G, 1 ≤ i ≤ n, 1 ≤ j ≤ m . Then
kf (x) − f (y)k ≤ mnM kx − yk
for all x, y ∈ G.
Proof. Let x, y ∈ G. For j = 1, · · · , m define gj (t) = fj (x + t(x − y)). Then by the Mean
Value Theorem there exists cj ∈ (0, 1) such that
fj (y) − fj (x) = gj (1) − gj (0) = gj0 (cj ).
Now by the chain rule we have gj0 (cj ) = ∇fj (x + cj (y − x)) · (y − x), so
n
X ∂fj
|gj (cj )| = (x + cj (y − x))(yi − xi )
∂xi
i=1
n
X
≤M |yi − xi | ≤ nM ky − xk.
i=1
Pm
Hence kf (x) − f (y)k ≤ j=1 |fj (x) − fj (y)| ≤ mnM kx − yk.
Now recall that if A : Rn → Rn is linear, then there exists a C such that kAxk ≤
Ckxk for all x. If A is non-singular, then this implies that there exists also a C1 such
that C1 kxk ≤ kAxk for all x (take y = Ax in the inequality kA−1 (y)k ≤ C 0 kyk). This
observation can be used to show that linear maps satisfy the following theorem.
Theorem 2 (Inverse Function Theorem). Let G ⊂ Rn be an open set and let f : G → Rm
be continuously differentiable on G (i.e., all the partials of f are continuous on G). Let
a ∈ G be such that det Df (a) 6= 0. Then there exist open sets U ⊂ G, V ⊂ Rm with
a ∈ U such that f (U ) = V and f −1 : V → U exists and is differentiable and Df −1 (y) =
(Df (f −1 (y)))−1 for all y ∈ V .
1
2 SUPPLEMENTAL NOTES FOR MATH 703, FALL 2005
Proof. Using the remark from above, that linear maps satisfy the theorem, we can replace
f by (Df (a))−1 f and assume that Df (a) = I. Then put g(x) = f (x)−x. Then Dg(a) = 0,
∂g
so there exists r > 0 such that B(a; r) ⊂ G and supx∈B(a;r) ∂xji (x) < 2n1 2 for all 1 ≤ i ≤ n
and all 1 ≤ j ≤ m. Now apply the above lemma to g to get
1
kf (x) − x − (f (y) − y)k ≤ M n2 kx − yk < kx − yk
2
for all x 6= y ∈ B(a; r). This implies that kf (x) − f (y)k ≥ 21 kx − yk for all x, y ∈ B(a; r).
This implies that f is one-to-one on B(a; r) and has a continuous inverse on the range of
this closed ball. By making r smaller if necessary, we can assume that det Df (x) 6= 0 on
B(a; r). We shall now show that f (B(a; r)) ⊃ V , where V is open with f (a) ∈ V . Let
S be the boundary of B(a; r). Then S is compact, so f (S) is compact and f (a) ∈ / f (S)
d
implies that d = dist (f (S), f (a)) > 0. Let V = {y : ky − f (a)k < 2 }. Then V is open
and we claim that V ⊂ f (B(a; r)). To see this, let y ∈ V and put h(x) = kf (x) − yk2 for
x ∈ B(a; r). Then h has a minimum on B(a; r), but this can’t be on the boundary S of
B(a; r) by the definition of d and V . Hence h has a minimum at x0 ∈ B(a; r). This implies
that the derivative Dh(x0 ) = 0, so
n
X ∂fi
−2(fi (x0 ) − yi ) (x0 ) = 0
∂xj
i=1
for j = 1, · · · , n. As det Df (x0 ) 6= 0 this implies that fi (x0 )−yi = 0 for i = 1, · · · , n. Hence
f (x0 ) = y. This shows V ⊂ f (B(a; r)). Let U = f −1 (V ). Then U open and f : U → V is
one-to-one and onto. Remains to show that f −1 is differentiable on V . Let u, v ∈ V and
put x = f −1 (u), y = f −1 (v). For v close enough to u we can write
f (y) − f (x) − Df (x)(y − x) = (y − x)kx − yk,
where (h) → 0 as h → 0. Put A = Df (x) = Df (f −1 (u)). Then we have
v − u − A(f −1 (v) − f −1 (u)) = (f −1 (v) − f −1 (u))kf −1 (v) − f −1 (u)k = 0 (v − u)kv − uk,
where we used that kf −1 (u) − f −1 (v)k ≤ 2ku − vk. Multiplying the above equation by A−1
we get
f −1 (v) − f −1 (u) − A−1 (v − u) = −A−1 (0 (v − u))kv − uk.
Now A−1 (0 (v − u)) → 0 as v → u implies that f −1 is differentiable at u with derivative
A−1 .
Corollary 3. Let G ⊂ C be open and f ∈ H(G) such that f 0 (z0 ) 6= 0. Then there exists
open sets U ⊂ G and V ⊂ C such that z0 ∈ U and f : U → V is one-to-one and onto and
f −1 : V → C is holomorphic.
Proof. Consider G as an open subset of R2 and define F : G → R2 by F (x, y) = (u(x, y), v(x, y))
where f = u + iv. Then det DF (x0 , y0 ) = |f 0 (z0 )|2 6= 0 (where z0 = x0 + iy0 ). The matrix
A PROOF OF THE INVERSE FUNCTION THEOREM 3
of DF (x0 , y0 ) is given by the skew symmetric matrix

∂u ∂v

∂x (x0 , y0 ) − ∂x (x0 , y0 ) .
∂v ∂u
∂x (x0 , y0 ) ∂x (x0 , y0 )
By the inverse function theorem there exist open sets U ⊂ G, V ⊂ R2 with (x0 , y0 ) ∈ U
such that F (U ) = V and F −1 : V → U exists and is differentiable and DF −1 ((x, y)) =
(DF (F −1 ((x, y))))−1 for all (x, y) ∈ V . Define g(z) = F −1 (x, y) for z = x + iy with
(x, y) ∈ V . Then g is the inverse of f and the real part and imaginary part of g are con-
tinuous and satisfy the Cauchy-Riemann equations on V , since the matrix DF −1 ((x, y)) =
(DF (F −1 ((x, y))))−1 is again skew symmetric for all (x, y) ∈ V . Hence g is holomor-
phic.
We conclude with the implicit function theorem. Assume we are given a system of m
equations
f1 (x1 , · · · , xn , y1 , · · · , ym ) = 0
f2 (x1 , · · · , xn , y1 , · · · , ym ) = 0
.. ..
. .
fm (x1 , · · · , xn , y1 , · · · , ym ) = 0
in n + m variables x1 , · · · , xn , y1 , · · · , ym . Then the implicit function theorem will give

sufficient conditions for solving y1 , · · · , ym in terms of x1 , · · · , xn .
Theorem 4 (Implicit Function Theorem). Let E ⊂ Rn+m be open and f : E → Rm a
∂f
continuously differentiable map. Let (x0 , y0 ) ∈ E such that f (x0 , y0 ) = 0 and det ∂yji 6= 0.
Then there exists an open set U ⊂ Rn with x0 ∈ U , and a continuously differentiable map
g : U → Rm such that f (x, g(x)) = 0 for all x ∈ U .
Proof. Define F : E → Rn+m by F (x, y) = (x, f (x, y)). Then F is continuously differ-
∂fj
entiable in a neighborhood of (x0 , y0 ) and det DF (x0 , y0 ) = det ∂yi 6= 0. Hence by
the Inverse Function Theorem there exists open U0 ⊂ E containing (x0 , y0 ) and open
V0 ⊂ Rm+n containing (x0 , 0) such that F : U0 → V0 is one-to-one and onto and such
that F −1 : V0 → U0 is differentiable. Let F −1 (x, y) = (G1 (x, y), G2 (x, y)). Then for all
(x, 0) ∈ V0 we have (x, 0) = F (G1 (x, 0), G2 (x, 0)) = (G1 (x, 0), f (G1 (x, 0), G2 (x, 0))). Hence
G1 (x, 0) = x and thus f (x, G2 (x, 0)) = 0 for all (x, 0) ∈ V0 . Let U = {x ∈ Rm : (0, x) ∈ V0 }
and define g(x) = G2 (x, 0) for x ∈ U . It is clear then that U and g satisfy the conclusion
of the theorem.

Inversefunction THM

Uploaded by

Copyright:

Available Formats

Inversefunction THM

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Inversefunction THM

Uploaded by

Copyright:

Available Formats

A PROOF OF THE INVERSE FUNCTION THEOREM

SUPPLEMENTAL NOTES FOR MATH 703, FALL 2005

Lemma 1. Let G ⊂n Rn be an open and convex set and leto f : G → Rm be differentiable

of DF (x0 , y0 ) is given by the skew symmetric matrix

in n + m variables x1 , · · · , xn , y1 , · · · , ym . Then the implicit function theorem will give

You might also like