Exercises for Chapter 2
3.6 Separable Spaces
Remark18.Corollary 3.23 is the main reason whyreflexive spacesandconvex func- tionsare so important in many problems occurring in thecalculus of variationsand inoptimization.
Theorem 3.24.LetEandF be two reflexive Banach spaces. LetA:D(A)⊂E→ F be an unbounded linear operator that is densely defined and closed. ThenD(A) is dense inF. ThusAis well defined(A:D(A)⊂E →F)and it may also be viewed as an unbounded operator fromEintoF. Then we have
A=A.
Proof.
1. D(A)is dense inF. Let ϕ be a continuous linear functional on F that vanishes onD(A). In view of Corollary 1.8 it suffices to prove thatϕ ≡0 onF. SinceF is reflexive,ϕ∈F and we have
(6) w, ϕ =0 ∀w∈D(A).
Ifϕ =0 then[0, ϕ]∈/ G(A)inE×F. Thus, one may strictly separate[0, ϕ]and G(A)by a closed hyperplane inE×F; i.e., there exist some[f, v] ∈E×Fand someα∈Rsuch that
f, u + v, Au< α <v, ϕ ∀u∈D(A).
It follows that
f, u + v, Au =0 ∀u∈D(A) and
v, ϕ =0.
Thusv∈D(A), and we are led to a contradiction by choosingw=vin (6).
2.A=A. We recall (see Section 2.6) that I[G(A)] =G(A)⊥ and
I[G(A)] =G(A)⊥. It follows that
G(A)=G(A)⊥⊥=G(A), sinceAis closed.
3.6 Separable Spaces 73 Many important spaces in analysis are separable. Clearly, finite-dimensional spaces are separable. As we shall see in Chapter 4 (see also Chapter 11),Lp(andp) spaces are separable for 1≤p <∞. AlsoC(K),the space of continuous functions on a compact metric spaceK, is separable (see Problem 24). However,L∞and∞ arenotseparable (see Chapters 4 and 11).
Proposition 3.25.LetEbe a separable metric space and letF ⊂Ebe any subset.
ThenF is also separable.
Proof. Let(un)be a countable dense subset ofE. Let(rm)be any sequence of positive numbers such thatrm→0. Choose any pointam,n∈B(un, rm)∩F whenever this set is nonempty. The set(am,n)is countable and dense inF.
Theorem 3.26.Let E be a Banach space such that E is separable. Then E is separable.
Remark19.The converse is not true. As we shall see in Chapter 4, E = L1 is separable but its dual spaceE =L∞isnotseparable.
Proof. Let(fn)n≥1be countable and dense inE. Since fn = sup
x∈E x≤1
fn, x,
we can find somexn∈Esuch that
xn =1 andfn, xn ≥ 1 2fn.
Let us denote byL0the vector space overQgenerated by the(xn)n≥1; i.e.,L0consists of all finitelinear combinations with coefficients in Q of the elements (xn)n≥1. We claim thatL0 is countable. Indeed, for every integer n, let n be the vector space overQgenerated by the(xk)1≤k≤n. Clearly,nis countable and, moreover, L0=
n≥1n.
LetLdenote the vector space overRgenerated by the(xn)n≥1. Of course,L0is a dense subset ofL. We claim thatLis a dense subspace ofE—and this will conclude the proof(L0will be a dense countable subset ofE). Letf ∈ E be a continuous linear functional that vanishes onL; in view of Corollary 1.8 we have to prove that f =0. Given anyε >0, there is some integerN such thatf−fN< ε. We have
1
2fN ≤ fN, xN = fN−f, xN< ε
(sincef, xN =0). It follows thatf ≤ f−fN + fN<3ε. Thusf =0.
Corollary 3.27.LetEbe a Banach space. Then
[Ereflexive and separable] ⇔ [Ereflexive and separable].
Proof. We already know (Corollary 3.21 and Theorem 3.26) that [Ereflexive and separable]⇒ [Ereflexive and separable].
Conversely, ifEis reflexive and separable, so isE=J (E); thusE is reflexive and separable.
Separabilityproperties are closely related to themetrizabilityof the weak topolo- gies. Let us recall that a topological spaceX is said to bemetrizableif there is a metric onXthat induces the topology ofX.
Theorem 3.28.LetEbe a separable Banach space. ThenBE is metrizable in the weaktopologyσ (E, E).
Conversely, ifBEis metrizable inσ (E, E), thenEis separable.
There is a “dual” statement.
Theorem 3.29.Let E be a Banach space such thatE is separable. ThenBE is metrizable in the weak topologyσ (E, E).
Conversely, ifBEis metrizable inσ (E, E), thenEis separable.
Proof of Theorem3.28.Let(xn)n≥1be a dense countable subset ofBE. For every f ∈Eset
[f] = ∞ n=1
1
2n|f, xn|.
Clearly, [ ] is a norm on E and[f] ≤ f. Let d(f, g) = [f −g] be the corresponding metric. We shall prove that the topology induced byd onBE is the same as the topologyσ (E, E)restricted toBE.
(a) Letf0 ∈BE and letV be a neighborhood off0forσ (E, E). We have to find somer >0 such that
U = {f ∈BE;d(f, f0) < r} ⊂V . As usual, we may assume thatV has the form
V = {f ∈BE; |f −f0, yi|< ε ∀i=1,2, . . . , k}
withε >0 andy1, y2, . . . , yk ∈E. Without loss of generality we may assume that yi ≤1 for everyi=1,2, . . . , k. For everyithere is some integernisuch that
yi−xni< ε/4 (since the set(xn)n≥1is dense inBE).
Chooser >0 small enough that
2nir < ε/2 ∀i=1,2, . . . , k.
We claim that for suchr,U⊂V. Indeed, ifd(f, f0) < r, we have
3.6 Separable Spaces 75 1
2ni|f −f0, xni|< r ∀i=1,2, . . . , k and therefore,∀i=1,2, . . . , k,
|f −f0, yi| = |f−f0, yi−xni + f −f0, xni|< ε 2 +ε
2. It follows thatf ∈V.
(b) Letf0 ∈ BE. Given r >0, we have to find some neighborhoodV of f0for σ (E, E)such that
V ⊂U= {f ∈BE; d(f, f0) < r}. We shall chooseV to be
V = {f ∈BE; f −f0, xi|< ε ∀i=1,2, . . . , k} withεandkto be determined in such a way thatV ⊂U. Forf ∈V we have
d(f, f0)= k n=1
1
2n|f −f0, xn| + ∞ n=k+1
1
2n|f −f0, xn|
< ε+2 ∞ n=k+1
1
2n =ε+ 1 2k−1.
Thus, it suffices to takeε= r2andklarge enough that 2k1−1 < r2.
Conversely, suppose BE is metrizable inσ (E, E)and let us prove thatEis separable. Set
Un= {f ∈BE; d(f,0) <1/n}
and letVnbe a neighborhood of 0 inσ (E, E)such thatVn⊂Un. We may assume thatVnhas the form
Vn= {f ∈BE; |f, x|< εn ∀x∈n} withεn>0 andnis a finite subset ofE. Set
D= ∞ n=1
n,
so thatDis countable.
We claim that the vector space generated byDis dense inE(which implies that Eis separable). Indeed, supposef ∈Eis such thatf, x =0 ∀x∈D. It follows thatf ∈Vn ∀nand thereforef ∈Un ∀n, so thatf =0.
Proof of Theorem3.29.The proof of the implication
[Eseparable]⇒ [BEis metrizable inσ (E, E)]
is exactly the same as above—just change the roles ofEandE. The proof of the converse is more delicate (find where the proof above breaks down); we refer to N. Dunford–J. T. Schwartz [1] or Exercise 3.24.
Remark20.One should emphasize again (see Remark 3) that in infinite-dimensional spaces the weak topologyσ (E, E)(resp. weak topologyσ (E, E))onallofE (resp.E) isnotmetrizable; see Exercise 3.8. In particular, the topology induced by the norm [ ] on all ofEdoes not coincide with the weaktopology.
Corollary 3.30.Let E be a separable Banach space and let (fn) be a bounded sequence inE. Then there exists a subsequence(fnk)that converges in the weak topologyσ (E, E).
Proof. Without loss of generality we may assume thatfn ≤1 for alln. The setBE is compact and metrizable for the topologyσ (E, E)(by Theorems 3.16 and 3.28).
The conclusion follows.
We may now return to the proof of Theorem 3.18:
Proof of Theorem3.18.Let M0 be the vector space generated by the xn’s and let M = M0. Clearly,M is separable (see the proof of Theorem 3.26). Moreover,M is reflexive (by Proposition 3.20). It follows thatBM is compact and metrizable in the weak topologyσ (M, M), sinceM is separable (we use here Corollary 3.27 and Theorem 3.29). We may thus find a subsequence(xnk)that converges weakly σ (M, M), and hence (xnk)converges also weaklyσ (E, E)(as in the proof of Proposition 3.20).