122 CHAPTER 5. FUNCTIONS ON NORMED LINEAR SPACES

Proof: Let H = f−1 ([−1,−1/3]) ,K = f−1 ([1/3,1]) . Thus H and K are disjoint closedsubsets of M. Suppose first H,K are both nonempty. Then by Lemma 5.7.1 there exists gsuch that g is a continuous function defined on all of X and g(H) = −1/3, g(K) = 1/3,and g(X)⊆ [−1/3,1/3] . It follows ∥ f −g∥M < 2/3. If H = /0, then f has all its values in[−1/3,1] and so letting g≡ 1/3, the desired condition is obtained. If K = /0, let g≡−1/3.If both H,K = /0, let g = 0.

When M is symmetric and f is odd, g(x)≡ 13

dist(x,H)−dist(x,K)dist(x,H)+dist(x,K) . When x ∈H this gives

13−dist(x,K)dist(x,K) = − 1

3 . Then x ∈ K, this gives 13

dist(x,H)dist(x,H) =

13 . Also g(H) = −1/3, f (H) ⊆

[−1,−1/3] so for x ∈ H, |g(x)− f (x)| ≤ 23 . It is similar for x ∈ K. If x is in neither H

nor K, then g(x) ∈ [−1/3,1/3] and so is f (x) . Thus ∥ f −g∥M ≤ 23 . Now by assumption,

since f is odd, H =−K. It is clear that g is odd because

g(−x) =13

dist(−x,H)−dist(−x,K)

dist(−x,H)+dist(−x,K)=

13

dist(−x,−K)−dist(−x,−H)

dist(−x,−K)+dist(−x,−H)

=13

dist(x,K)−dist(x,H)

dist(x,K)+dist(x,H)=−g(x) . ■

Lemma 5.7.4 Suppose M is a closed set in X and suppose f : M→ [−1,1] is continuousat every point of M. Then there exists a function g which is defined and continuous on allof X such that g = f on M and g has its values in [−1,1] . If X is a normed linear spaceand f is odd, then we can also assume g is odd.

Proof: Using Lemma 5.7.3, let g1 be such that g1 (X)⊆ [−1/3,1/3] and ∥ f −g1∥M ≤23 . Suppose g1, · · · ,gm have been chosen such that g j (X)⊆ [−1/3,1/3] and∥∥∥∥∥ f −

m

∑i=1

(23

)i−1

gi

∥∥∥∥∥M

<

(23

)m

. (5.1)

This has been done for m = 1. Then∥∥∥( 3

2

)m(

f −∑mi=1( 2

3

)i−1gi

)∥∥∥M≤ 1 and so

(32

)m(

f −m

∑i=1

(23

)i−1

gi

)can play the role of f in the first step of the proof. Therefore, there exists gm+1 defined andcontinuous on all of X such that its values are in [−1/3,1/3] and∥∥∥∥∥

(32

)m(

f −m

∑i=1

(23

)i−1

gi

)−gm+1

∥∥∥∥∥M

≤ 23.

Thus∥∥∥( f −∑

mi=1( 2

3

)i−1gi

)−( 2

3

)mgm+1

∥∥∥M≤( 2

3

)m+1. It follows there exists a sequence

{gi} such that each has its values in [−1/3,1/3] and for every m 5.1 holds. Then letg(x) ≡ ∑

∞i=1( 2

3

)i−1gi (x) . It follows |g(x)| ≤

∣∣∣∑∞i=1( 2

3

)i−1gi (x)

∣∣∣ ≤ ∑mi=1( 2

3

)i−1 13 ≤ 1

and∣∣∣( 2

3

)i−1gi (x)

∣∣∣ ≤ ( 23

)i−1 13 so the Weierstrass M test applies and shows convergence

is uniform. Therefore g must be continuous by Theorem 3.9.3. The estimate 5.1 impliesf = g on M. The last claim follows because we can take each gi odd. ■

The following is the Tietze extension theorem.

122 CHAPTER 5. FUNCTIONS ON NORMED LINEAR SPACESProof: Let H = f~' ([-1,—1/3]),K =f7! ((1/3, 1]) . Thus A and K are disjoint closedsubsets of M. Suppose first H,K are both nonempty. Then by Lemma 5.7.1 there exists gsuch that g is a continuous function defined on all of X and g(H) = —1/3, g(K) = 1/3,and g(X) C [—1/3, 1/3]. It follows || f — g||,, < 2/3. If H = 9, then f has all its values in[—1/3,1] and so letting g = 1/3, the desired condition is obtained. If K = 0, let g = —1/3.If both H,K = 90, let g=0.When M is symmetric and f is odd, g(x) = } Tia rds)= 3 Gist(x H)tdist(e.K) When x € H this givesSe = —}- Then a € K, this gives S527} = 4. Also g(H) = —1/3, f(H) C[—1,—1/3] so for a € H,|g (x) — f (x)| < 2. It is similar for x € K. If & is in neither Hnor K, then g(a) € [—1/3, 1/3] and so is f (aw). Thus || f —g||y < 2. Now by assumption,since f is odd, H = —K. It is clear that g is odd because1 dist(—x,H)—dist(—a,K) 1 dist(—a,—K) — dist (—ax, —H)8(-@) = 3 dist(car,H) + dist(ca.K) ~ 3 dist(cx,—-K) dist (<a, -H)1 dist (a, K) — dist (x, H)3 dist w, K) +dist (a, H)=-g(a).Lemma 5.7.4 Suppose M is a closed set in X and suppose f : M — [1,1] is continuousat every point of M. Then there exists a function g which is defined and continuous on allof X such that g = f on M and g has its values in |—1,1]. If X is a normed linear spaceand f is odd, then we can also assume g is odd.Proof: Using Lemma 5.7.3, let g1 be such that g) (X) C [—1/3, 1/3] and || f — gully <3. Suppose g1,-++ ,8m have been chosen such that g;(X) C [—1/3, 1/3] andm 2 i-1 m1 -£(3) gi <( ) (5.1)i=lThis has been done for m = 1. Then 1 (3)" (f-ri1 (3) si)a) £0)"can play the role of f in the first step of the proof. Therefore, there exists g,,,1 defined andcontinuous on all of X such that its values are in [—1/3,1/3] andWINM| <1 and soM2\"! 24 » 3 Si &m+1 3°MThus | (f- ye AG yo ) — (3)" Sm-+1 |< Q@y"" . It follows there exists a sequence{g;} such that each has its values in [—1/3,1/3] and for every m 5.1 holds. Then let0 i—1 co i—1 i-1g(w) = 22, (3) | gi(w). It follows |g(a)| < [EF (3) gi(w)| EM (G) TE <1i—l i—1 . .and \(3)' gi (x)| < (3) ; so the Weierstrass M test applies and shows convergenceis uniform. Therefore g must be continuous by Theorem 3.9.3. The estimate 5.1 impliesf =g nM. The last claim follows because we can take each g; odd.The following is the Tietze extension theorem.