Advanced Calculus (II)
WEN-CHINGLIEN
Department of Mathematics National Cheng Kung University
2009
Ch11: Differentiability on R
nCh11.6: Inverse Function Theorem
Notation:
f :Rn→ Rn,
The Jacobian of f is
∆f(a) := det(Df (a)).
Let V be open inRn, f : V →Rn,a ∈ V, and r > 0 be so small that Br(a) ⊂ V . Suppose that f is continuous and 1-1 on Br(a), and its first-order partial derivatives exist at every point in Br(a). If ∆f 6= 0 on Br(a), then there is a ρ >0 such that Bρ(f (a)) ⊂ f (Br(a)).
Theorem (11.39)
Let V be open and nonempty inRn, and f : V →Rn be continuous. If f is 1-1 and has first-order partial
derivatives on V , and if ∆f 6= 0 on V , then f−1is continuous on f (V ).
Let V be open inRn and f : V →Rn be C1 on V . If
∆f(a) 6= 0 for some a ∈ V , then there is an r > 0 such that Br(a) ⊂ V , f is 1-1 on Br(a), ∆f(x) 6= 0 for all x ∈ Br(a), and
det ∂fi
∂xj
(cj)
n×n
6= 0 for allc1, . . . ,cn ∈ Br(a).
Theorem (11.41 Inverse Function Theorem) Let V be open inRn and f : V →Rn be C1 on V . If
∆f(a) 6= 0 for some a ∈ V , then there exists an open set W containinga such that
(i) f is 1-1 on W ,
(ii) f−1is C1on f (W ), and (iii) for eachy ∈ f (W ),
D(f−1)(y) = [Df (f−1(y))]−1,
where [ ]−1 represents matrix inversion (see Theorem C.5).
Let V be open inRn and f : V →Rn be C1 on V . If
∆f(a) 6= 0 for some a ∈ V , then there exists an open set W containinga such that
(i) f is 1-1 on W ,
(ii) f−1is C1on f (W ), and (iii) for eachy ∈ f (W ),
D(f−1)(y) = [Df (f−1(y))]−1,
where [ ]−1 represents matrix inversion (see Theorem C.5).
Theorem (11.41 Inverse Function Theorem) Let V be open inRn and f : V →Rn be C1 on V . If
∆f(a) 6= 0 for some a ∈ V , then there exists an open set W containinga such that
(i) f is 1-1 on W ,
(ii) f−1is C1on f (W ), and (iii) for eachy ∈ f (W ),
D(f−1)(y) = [Df (f−1(y))]−1,
where [ ]−1 represents matrix inversion (see Theorem C.5).
The hypothesis ”∆f 6= 0 in Theorem 11.39 can be relaxed.
Proof.
If f (x ) = x3, then f :R → R and its inverse f−1(x ) =√3 x are continuous onR, but ∆f(0) = f0(0) = 0
Remark (11.42)
The hypothesis ”∆f 6= 0 in Theorem 11.39 can be relaxed.
Proof.
If f (x ) = x3, then f :R → R and its inverse f−1(x ) =√3 x are continuous onR, but ∆f(0) = f0(0) = 0
The hypothesis ”∆f 6= 0 in Theorem 11.41 cannot be relaxed. In fact, if f : Br(a) → Rn is differentiable ata and its inverse f−1exists and is differentiable at f (a), then
∆f(a) 6= 0.
Proof.
Suppose to the contrary that f is differentiable ata but
∆f(a) = 0. By Exercise 8, p.338, and the Chain Rule, I =D(f−1◦ f )(a) = D(f−1)(f (a))Df (a).
Taking the determinant of this identity, we have 1 = ∆f−1(f (a))∆f(a) = 0, a contradiction.
Remark (11.43)
The hypothesis ”∆f 6= 0 in Theorem 11.41 cannot be relaxed. In fact, if f : Br(a) → Rn is differentiable ata and its inverse f−1exists and is differentiable at f (a), then
∆f(a) 6= 0.
Proof.
Suppose to the contrary that f is differentiable ata but
∆f(a) = 0.By Exercise 8, p.338, and the Chain Rule, I = D(f−1◦ f )(a)=D(f−1)(f (a))Df (a).
Taking the determinant of this identity, we have 1 = ∆f−1(f (a))∆f(a) = 0, a contradiction.
The hypothesis ”∆f 6= 0 in Theorem 11.41 cannot be relaxed. In fact, if f : Br(a) → Rn is differentiable ata and its inverse f−1exists and is differentiable at f (a), then
∆f(a) 6= 0.
Proof.
Suppose to the contrary that f is differentiable ata but
∆f(a) = 0. By Exercise 8, p.338, and the Chain Rule, I =D(f−1◦ f )(a) = D(f−1)(f (a))Df (a).
Taking the determinant of this identity, we have 1 = ∆f−1(f (a))∆f(a) = 0, a contradiction.
Remark (11.43)
The hypothesis ”∆f 6= 0 in Theorem 11.41 cannot be relaxed. In fact, if f : Br(a) → Rn is differentiable ata and its inverse f−1exists and is differentiable at f (a), then
∆f(a) 6= 0.
Proof.
Suppose to the contrary that f is differentiable ata but
∆f(a) = 0. By Exercise 8, p.338, and the Chain Rule, I = D(f−1◦ f )(a)=D(f−1)(f (a))Df (a).
Taking the determinant of this identity, we have 1 = ∆f−1(f (a))∆f(a) = 0, a contradiction.
The hypothesis ”∆f 6= 0 in Theorem 11.41 cannot be relaxed. In fact, if f : Br(a) → Rn is differentiable ata and its inverse f−1exists and is differentiable at f (a), then
∆f(a) 6= 0.
Proof.
Suppose to the contrary that f is differentiable ata but
∆f(a) = 0. By Exercise 8, p.338, and the Chain Rule, I = D(f−1◦ f )(a) = D(f−1)(f (a))Df (a).
Taking the determinant of this identity, we have 1 = ∆f−1(f (a))∆f(a) = 0, a contradiction.
Remark (11.43)
The hypothesis ”∆f 6= 0 in Theorem 11.41 cannot be relaxed. In fact, if f : Br(a) → Rn is differentiable ata and its inverse f−1exists and is differentiable at f (a), then
∆f(a) 6= 0.
Proof.
Suppose to the contrary that f is differentiable ata but
∆f(a) = 0. By Exercise 8, p.338, and the Chain Rule, I = D(f−1◦ f )(a) = D(f−1)(f (a))Df (a).
Taking the determinant of this identity, we have 1 = ∆f−1(f (a))∆f(a) = 0, a contradiction.
The hypothesis ”f is C1 on V ” in Theorem 11.41 cannot be relaxed.
Proof.
If f (x ) = x + 2x2sinx1, x 6= 0, and f (0) = 0, then f :R → R is differentiable on V := (−1, 1) and f0(0) = 1 6= 0.
However, since
f
2
(4k − 1)π
<f
2
(4k + 1)π
<f
2
(4k − 3)π
for k ∈N, f is not 1-1 on any open set that contains 0.
Therefore, no open subset of f (V ) can be closen on which f−1exists.
Remark (11.44)
The hypothesis ”f is C1 on V ” in Theorem 11.41 cannot be relaxed.
Proof.
If f (x ) = x + 2x2sinx1, x 6= 0, and f (0) = 0,then f :R → R is differentiable on V := (−1, 1) and f0(0) = 1 6= 0.
However, since f
2
(4k − 1)π
<f
2
(4k + 1)π
<f
2
(4k − 3)π
for k ∈N, f is not 1-1 on any open set that contains 0.
Therefore, no open subset of f (V ) can be closen on which f−1exists.
The hypothesis ”f is C1 on V ” in Theorem 11.41 cannot be relaxed.
Proof.
If f (x ) = x + 2x2sinx1, x 6= 0, and f (0) = 0, then f :R → R is differentiable on V := (−1, 1) and f0(0) = 1 6= 0.
However, since f
2
(4k − 1)π
<f
2
(4k + 1)π
<f
2
(4k − 3)π
for k ∈N, f is not 1-1 on any open set that contains 0.
Therefore, no open subset of f (V ) can be closen on which f−1exists.
Remark (11.44)
The hypothesis ”f is C1 on V ” in Theorem 11.41 cannot be relaxed.
Proof.
If f (x ) = x + 2x2sinx1, x 6= 0, and f (0) = 0, then f :R → R is differentiable on V := (−1, 1) and f0(0) = 1 6= 0.
However, since f
2
(4k − 1)π
<f
2
(4k + 1)π
<f
2
(4k − 3)π
for k ∈N,f is not 1-1 on any open set that contains 0.
Therefore, no open subset of f (V ) can be closen on which f−1exists.
The hypothesis ”f is C1 on V ” in Theorem 11.41 cannot be relaxed.
Proof.
If f (x ) = x + 2x2sinx1, x 6= 0, and f (0) = 0, then f :R → R is differentiable on V := (−1, 1) and f0(0) = 1 6= 0.
However, since f
2
(4k − 1)π
<f
2
(4k + 1)π
<f
2
(4k − 3)π
for k ∈N, f is not 1-1 on any open set that contains 0.
Therefore, no open subset of f (V ) can be closen on which f−1exists.
Remark (11.44)
The hypothesis ”f is C1 on V ” in Theorem 11.41 cannot be relaxed.
Proof.
If f (x ) = x + 2x2sinx1, x 6= 0, and f (0) = 0, then f :R → R is differentiable on V := (−1, 1) and f0(0) = 1 6= 0.
However, since f
2
(4k − 1)π
<f
2
(4k + 1)π
<f
2
(4k − 3)π
for k ∈N,f is not 1-1 on any open set that contains 0.
Therefore, no open subset of f (V ) can be closen on which f−1exists.
The hypothesis ”f is C1 on V ” in Theorem 11.41 cannot be relaxed.
Proof.
If f (x ) = x + 2x2sinx1, x 6= 0, and f (0) = 0, then f :R → R is differentiable on V := (−1, 1) and f0(0) = 1 6= 0.
However, since f
2
(4k − 1)π
<f
2
(4k + 1)π
<f
2
(4k − 3)π
for k ∈N, f is not 1-1 on any open set that contains 0.
Therefore, no open subset of f (V ) can be closen on which f−1exists.
Remark (11.44)
The hypothesis ”f is C1 on V ” in Theorem 11.41 cannot be relaxed.
Proof.
If f (x ) = x + 2x2sinx1, x 6= 0, and f (0) = 0, then f :R → R is differentiable on V := (−1, 1) and f0(0) = 1 6= 0.
However, since f
2
(4k − 1)π
<f
2
(4k + 1)π
<f
2
(4k − 3)π
for k ∈N, f is not 1-1 on any open set that contains 0.
Therefore, no open subset of f (V ) can be closen on which f−1exists.
Suppose that V is open inRn+p, and F = (F1, . . . ,Fn) : V →Rnis C1on V . Suppose further that F (x0,t0) = 0 for some (x0,t0) ∈V , wherex0∈ Rnandt0 ∈ Rp. If
∂(F1, . . . ,Fn)
∂(x1, . . . ,xn)(x0,t0) 6=0,
then there is an open set W ⊂Rp containingt0and a unique continuously differentiable function g : W →Rn such that g(t0) = x0, and F (g(t), t) = 0 for all t ∈ W .
Example (11.49)
Prove that there exist function u, v :R4→ R, continuously differentiable on some ball B centered at the point
(x , y , z, w ) = (2, 1, −1, −2), such that
u(2, 1, −1, −2) = 4, v (2, 1, −1, −2) = 3, and the equations
u2+v2+w2=29, u2 x2 + v2
y2 +w2 z2 =17 both hold for all (x , y , z, w ) in B.
Set n = 2, p = 4, and
F (u, v , x , y , z, w )= (u2+v2+w2− 29,u2 x2+v2
y2+w2
z2 − 17).
Then F (4, 3, 2, 1, −1, −2) = (0, 0), and
∂(F1,F2)
∂(u, v ) =det2u 2v
2u x2
2v y2
=4uv 1 y2 − 1
x2
.
This determinant is nonzero when u = 4, v = 3, x = 2, and y = 1. Therefore, such functions u, v exist by the Implicit Function Theorem.
Proof.
Set n = 2, p = 4, and
F (u, v , x , y , z, w ) = (u2+v2+w2− 29,u2 x2+v2
y2+w2
z2 − 17).
Then F (4, 3, 2, 1, −1, −2) = (0, 0), and
∂(F1,F2)
∂(u, v ) =det2u 2v
2u x2
2v y2
=4uv 1 y2 − 1
x2
.
This determinant is nonzero when u = 4, v = 3, x = 2, and y = 1. Therefore, such functions u, v exist by the Implicit Function Theorem.
Set n = 2, p = 4, and
F (u, v , x , y , z, w ) = (u2+v2+w2− 29,u2 x2+v2
y2+w2
z2 − 17).
Then F (4, 3, 2, 1, −1, −2) = (0, 0), and
∂(F1,F2)
∂(u, v ) =det2u 2v
2u x2
2v y2
=4uv 1 y2 − 1
x2
. This determinant is nonzero when u = 4, v = 3, x = 2, and y = 1. Therefore, such functions u, v exist by the Implicit Function Theorem.
Proof.
Set n = 2, p = 4, and
F (u, v , x , y , z, w ) = (u2+v2+w2− 29,u2 x2+v2
y2+w2
z2 − 17).
Then F (4, 3, 2, 1, −1, −2) = (0, 0), and
∂(F1,F2)
∂(u, v ) =det2u 2v
2u x2
2v y2
=4uv 1 y2 − 1
x2
. This determinant is nonzero when u = 4, v = 3, x = 2, and y = 1.Therefore, such functions u, v exist by the Implicit Function Theorem.
Set n = 2, p = 4, and
F (u, v , x , y , z, w ) = (u2+v2+w2− 29,u2 x2+v2
y2+w2
z2 − 17).
Then F (4, 3, 2, 1, −1, −2) = (0, 0), and
∂(F1,F2)
∂(u, v ) =det2u 2v
2u x2
2v y2
=4uv 1 y2 − 1
x2
. This determinant is nonzero when u = 4, v = 3, x = 2, and y = 1. Therefore, such functions u, v exist by the Implicit Function Theorem.
Proof.
Set n = 2, p = 4, and
F (u, v , x , y , z, w ) = (u2+v2+w2− 29,u2 x2+v2
y2+w2
z2 − 17).
Then F (4, 3, 2, 1, −1, −2) = (0, 0), and
∂(F1,F2)
∂(u, v ) =det2u 2v
2u x2
2v y2
=4uv 1 y2 − 1
x2
. This determinant is nonzero when u = 4, v = 3, x = 2, and y = 1.Therefore, such functions u, v exist by the Implicit Function Theorem.
Set n = 2, p = 4, and
F (u, v , x , y , z, w ) = (u2+v2+w2− 29,u2 x2+v2
y2+w2
z2 − 17).
Then F (4, 3, 2, 1, −1, −2) = (0, 0), and
∂(F1,F2)
∂(u, v ) =det2u 2v
2u x2
2v y2
=4uv 1 y2 − 1
x2
. This determinant is nonzero when u = 4, v = 3, x = 2, and y = 1. Therefore, such functions u, v exist by the Implicit Function Theorem.