Implicit
Implicit
Theorem 1.1 (Implicit Function Theorem I). Let m, n be positive integers. Let A
be an open subset of Rn+m , and let F : A → Rm be a continuously differentiable
function on A. Let (x0 , y0 ) ∈ A such that F (x0 , y0 ) = 0. Assume that DY F (x0 , y0 ) is
invertible1. Then there are open sets U ⊂ Rn and V ⊂ Rm such that x0 ∈ U , y0 ∈ V ,
and there is a function g : U → V differentiable at x0 such that (x, g(x)) ∈ A and
F (x, g(x)) = 0 for all x ∈ U . Moreover
−1
(1.1) g 0 (x0 ) = − DY F (x0 , y0 ) DX F (x0 , y0 ).
where In×n is the n × n identity matrix and 0n×m is the n × m zero matrix. One can
operate with block matrices like if we had a 2 × 2 lower triangular matrix with no
zeros in the diagonal3 and verify that
In×n 0n×m
−1
f 0 (z0 )
=
,
−(DY F (x0 , y0 ))−1 DX F (x0 , y0 ) (DY F (x0 , y0 ))−1
So f 0 (z0 ) is invertible and the inverse function theorem ensures that there exists an
open set U ⊂ E ⊂ Rn+m containing z0 , and an open set V ⊂ Rn+m containing f (z0 )
2Note that F 0 (x0 , y0 ) = [DX F (x0 , y0 ), DY F (x0 , y0 )] and DX F (x0 , y0 ) is an m × n matrix,
DY F (x0 , y0 ) is an m × m matrix.
3 In this case 1 6= 0 and c 6= 0: A = 1 0 −1 1 0
, A = .
b c −c−1 b c−1
4
(the image of the base point), such that f is a bijection from U to V . In particular
there is an inverse map f −1 : V → U . Moreover f −1 is differentiable at f (z0 ) and
−1
(2.2) (f −1 )0 (f (z0 )) = (f −1 )0 (x0 , 0) = f 0 (x0 , y0 ) .
Let us write f −1 in coordinates, f −1 = (h, g) where h = (h1 , h2 , . . . , hn ) and g =
(g1 , g2 , . . . , gm ), and hi , gi : V → R. Since (h(x, y), F (h(x, y), g(x, y))) = f (f −1 (x, y)) =
(x, y) then hi (x, y) = xi for all i = 1, 2, . . . , n, and F (h(x, y), g(x, y)) = F (x, g(x, y)) =
y. Also g is differentiable at (x0 , 0) ∈ U . We now define W ⊂ Rn and function
G : W → Rm as follows
W := {x ∈ Rn : (x, 0) ∈ U }, G(x) := g(x, 0) for x ∈ W.
n+m
Note that x0 ∈ W , and since U is open in R , W which is the projection of U onto
n n
R is also open in R , finally G so defined is differentiable at x0 . We now prove (??),
namely
A := {(x, y) ∈ V : F (x, y) = 0} = {(x, G(x)) : x ∈ W } =: B.
Assume (x, y) ∈ V and F (x, y) = 0 then f (x, y) = (x, 0) ∈ U therefore x ∈ W .
Applying f −1 , we see that (x, y) = f −1 (x, 0) in particular y = g(x, 0) = G(x) ,
therefore A ⊂ B. All the steps can be reversed since f is a bijection from U to V to
obtain the other set inequality.
By previous discussion F (x, G(x)) = 0 for all x ∈ W , F is differentiable at
(x0 , G(x0 )) = (x0 , y0 ) and G is differentiable at x0 , the formula for the derivative
follows by the chain rule:
DX F (x0 , y0 ) + DY F (x0 , y0 )G0 (x0 ) = 0,
−1
and by simple algebra we conclude that G0 (x0 ) = − DY F (x0 , y0 ) DX F (x0 , y0 ).
Remark 2.2. Tao only proves the implicit function theorem when m = 1, but basically
is the same argument I just presented, see Section 6.8 in Book II. With only the case
m = 1 he couldn’t go from implicit to inverse function theorem except in the case
n = m = 1, because one needs the implicit function theorem from R2n into Rn , that
is the case n = m.
Remark 2.3. Final comment, under the hypothesis that the functions are continuously
differentiable one gets more differentiability than just at the base point, because the
hypothesis of invertibility of the matrices f 0 (x0 ) or DY F (x0 , y0 ) will persist in a
neighborhood of x0 or (x0 , y0 ) respectively. The Theorems can be streghtened to say
that f −1 is continuously differentiable on a possibly smaller open set V , and G is
continuously differentiable on a possibly smaller open set W .