24.11. SOME EMBEDDING THEOREMS 695

Let n be this large. Then let{

ukXGε∗ψn

}r

k=1be a η

8 net for Aεn in Lp(Gε ;W

).

Then consider the balls B(

ukXGε, η

4

)in Lp

(Gε ;W

). If wXGε

is in Aε , is it in some

B(

ukXGε, η

2

)? By what was just shown, there is k such that∥∥∥wXGε

∗ψn−ukXGε∗ψn

∥∥∥Lp(Gε ;W)

8

and also ∥∥∥wXGε−wXGε

∗ψn

∥∥∥Lp(Gε ;W)

8+

η

25∥∥∥ukXGε−ukXGε

∗ψn

∥∥∥Lp(Gε ;W)

8+

η

25

Thus, ∥∥∥wXGε−ukXGε

∥∥∥Lp(Gε ;W )

≤∥∥∥wXGε

−wXGε∗ψn

∥∥∥Lp(Gε ;W)

+∥∥∥wXGε

∗ψn−ukXGε∗ψn

∥∥∥Lp(Gε ;W)

+∥∥∥ukXGε

∗ψn−ukXGε

∥∥∥Lp(Gε ;W)

<3η

8+

25<

η

2

It follows that{

ukXGε

}r

k=1is a η/2 net for Lp

(Gε ;W

)contrary to the construction. Thus

A has an η net after all. ■In case Ω is a closed interval, there are several versions of these sorts of embeddings

which are enormously useful in the study of nonlinear evolution equations or inclusions.The following theorem is an infinite dimensional version of the Ascoli Arzela theorem.

It is like a well known result due to Simon [52]. It is an appropriate generalization whenyou do not necessarily have weak derivatives. I am giving another proof although Theorem24.11.3 given above is actually more general.

Theorem 24.11.6 Let q > 1 and let E ⊆W ⊆ X where the injection map is contin-uous from W to X and compact from E to W. Let S be defined by{

u such that ∥u(t)∥E ≤ R for all t ∈ [a,b] , and ∥u(s)−u(t)∥X ≤ R |t− s|1/q}.

Thus S is bounded in L∞ (a,b,E) and in addition, the functions are uniformly Holder contin-uous into X . Then S⊆C ([a,b] ;W ) and if {un}⊆ S, there exists a subsequence,

{unk

}which

converges to a function u ∈C ([a,b] ;W ) in the following way: limk→∞

∥∥unk −u∥∥

∞,W = 0.

Proof: First consider the issue of S being a subset of C ([a,b] ;W ) . Let ε > 0 be given.Then by Lemma 24.11.1, there exists a constant, Cε such that for all u ∈W

∥u∥W ≤ε

6R∥u∥E +Cε ∥u∥X .

Therefore, for all u ∈ S,

∥u(t)−u(s)∥W ≤ε

6R∥u(t)−u(s)∥E +Cε ∥u(t)−u(s)∥X

24.11. SOME EMBEDDING THEOREMS 695Let n be this large. Then let {um 2iGe* Vu} be a q net for .%, in L? (Ge;W).k=1Then consider the balls B (u 2G 1) in LP (Ge:W). If WAG. is in .%, is it in someB (u LG 1) ? By what was just shown, there is k such thateee Yn 2Ge* Vn i (Gew) > 5and also n 7bw 2G W2E-*W, L»(Gew) <3 + 35| Ur 2E, — Up 2G * V, vGew) < + xThus,2a — me Xe LP(Ge:W) S Pee Peet L? (Ge:W)+ | w2eee Wn —U2E* VW,+Up AG * W, — Up REL»(Gew) L»(Gew)It follows that {u Reb ' isa 1/2 net for L? (Ge; Ww) contrary to the construction. Thus@ has an 77 net after all.In case Q is a closed interval, there are several versions of these sorts of embeddingswhich are enormously useful in the study of nonlinear evolution equations or inclusions.The following theorem is an infinite dimensional version of the Ascoli Arzela theorem.It is like a well known result due to Simon [52]. It is an appropriate generalization whenyou do not necessarily have weak derivatives. I am giving another proof although Theorem24.11.3 given above is actually more general.Theorem 24.11.6 Ler q> land let E CW CX where the injection map is contin-uous from W to X and compact from E to W. Let S be defined by{u such that |\u(t)||, <R for allt € {a,b|, and ||u(s) —u(t)|ly <R|r—s|"/4h.Thus S is bounded in L® (a,b, E) and in addition, the functions are uniformly Holder contin-uous into X. Then S C C [a,b] ;W) and if {un} CS, there exists a subsequence, { Un, } whichconverges to a function u € C (|a,b];W) in the following way: lim, +. || ung — ul|., y=Proof: First consider the issue of S being a subset of C ([a,b];W). Let € > 0 be given.Then by Lemma 24.11.1, there exists a constant, Cz such that for all uc WElull S Gp lleelle + Ce llully -Therefore, for all u € S,Ilu(t)—u(s)llw < = I|u(t) —u(s) lle + Ce [lu (2) — u (9) lx