23.8. A FUNCTION WITH VALUES IN SMALLER DIMENSIONS 787

algebra. It is done in [38]. Here we will be a little less ambitious. The following lemmawill be useful when extending the degree to finite dimensional normed linear spaces andfrom there to Banach spaces. It is motivated by the following diagram.

θ−1 (y) θ

−1← y

↑ θ−1 ◦g◦θ ↑ g

θ−1 (Ω)

θ→ Ω

Lemma 23.8.2 Let y /∈ g(∂Ω) and let θ be an isomorphism of Rn. That is, θ is one to oneonto and linear. Then

d(θ−1 ◦g◦θ ,θ−1 (Ω) ,θ−1y

)= d (g,Ω,y)

Proof: It suffices to consider g ∈C2(Ω̄;Rn

)for which y is a regular value because you

can get such a ĝ with ∥ĝ−g∥∞< δ where

B(y,2δ )∩g(∂Ω) = /0.

Thus B(y,δ )∩ ĝ(∂Ω) = /0 and so d (g,Ω,y) = d (ĝ,Ω,y) . One can assume similarly that∥ĝ−g∥

∞is sufficiently small that

d(θ−1g◦θ ,θ−1 (Ω) ,θ−1y

)= d

(θ−1ĝ◦θ ,θ−1 (Ω) ,θ−1y

)because both θ and θ

−1 are continuous. Thus it suffices to consider at the outset g ∈C2(Ω̄;Rn

). Then from the definition of degree for C2 maps,

d(θ−1 ◦g◦θ ,θ−1 (Ω) ,θ−1y

)= lim

ε→0

∫θ−1

φ ε

((θ−1 ◦g◦θ

)(z)−θ

−1y)

detD(θ−1 ◦g◦θ

)(z)dz

Now D(θ−1 ◦g◦θ

)(z) = θ

−1 D(g◦θ)(z) = θ−1Dg(θ (z))θz. Changing the variables

x = θz,z = θ−1x, this last integral equals∫

φ ε

((θ−1g◦θ

)(θ−1 (x)

)−θ

−1y)

detDg(x) |detθ |∣∣detθ

−1∣∣2 dx

=∫

φ ε

(θ−1g(x)−θ

−1y)∣∣detθ

−1∣∣detDg(x)dx

Recall that φ ε is a mollifier which is nonzero only in B(0,ε). Now

g−1 (y) = {x1, · · · ,xm}=(θ−1g)−1 (

θ−1y)

and so g(xi)= y and θ−1g(xi)= θ

−1y. By the inverse function theorem, there exist disjointopen sets Ui with xi ∈Ui, such that θ

−1g is one to one on Ui with det(D(θ−1g)(x))=

det(θ−1)detDg(x) having constant sign on Ui and θ

−1g(Ui) is an open set containingθ−1y. Then let ε be small enough that B

(θ−1y,ε

)⊆ ∩m

i=1θ−1g(Ui) and let

Vi ≡(θ−1g)−1 (

B(θ−1y,ε

))∩Ui

23.8. A FUNCTION WITH VALUES IN SMALLER DIMENSIONS 787algebra. It is done in [38]. Here we will be a little less ambitious. The following lemmawill be useful when extending the degree to finite dimensional normed linear spaces andfrom there to Banach spaces. It is motivated by the following diagram.-1a7! (y) “iy+ @-logo@ te@-!(Q) 4 9Lemma 23.8.2 Let y ¢ g(dQ) and let 0 be an isomorphism of R". That is, 8 is one to oneonto and linear. Thend(@-!0go0,0~!(Q),0-'y) =d(g,Q,y)Proof: It suffices to consider g € C” (Q; R") for which y is a regular value because youcan get such a & with || — g||,, < 6 whereB(y,28) Ng (dQ) =0.Thus B(y, 6) N&(0Q) = O and so d(g,Q,y) = d(,Q,y). One can assume similarly that||& — g||,, is sufficiently small thatd(@-'g0o6,@-'(Q),@ 'y) =d(0-'800,0 | (Q),6"'y)because both @ and @~' are continuous. Thus it suffices to consider at the outset g €C? (Q;R") . Then from the definition of degree for C? maps,d(@-'0go0,@ | (Q),0-'y)= lim oft ((0~! ogo@) (z)—@~'y) detD (07! 0g06) (z) dzé0 Jg—!Now D(@~!0go@) (z) = @' D(go@)(z) = @ 'Dg(@ (z)) @z. Changing the variablesx = 6z,2= 0 'x, this last integral equals[ $e ((0-'g08) (8! (x)) — @~!y) detDg(x) |det 6| |det @~! |” axQ= | 6, (@ 'g(x)—@ 'y) |det 0~' | det Dg (x) dxQRecall that @, is a mollifier which is nonzero only in B(0,€). Nowg-! (y) ={x1,-+: Xm} = (@7'g) | (@7!y)and so g(x;) =y and @~!g(x;) = @~'y. By the inverse function theorem, there exist disjointopen sets U; with x; € U;, such that @~'g is one to one on U; with det (D (6~'g) (x)) =det (0-') det Dg (x) having constant sign on U; and @~'g(U;) is an open set containing@~'y. Then let € be small enough that B(@~'y,e) CM", 'g(U;) and letVj= (@-'g) (B(0~'y,€)) NU;