718 CHAPTER 26. INDEPENDENCE

In the case where Z = E, a separable Banach space, if X is measurable then X ∈L1 (Ω;E) if and only if the identity map on E is in L1 (E;λX) and∫

X (ω)dP =∫

ExdλX (x) (26.2)

Proof: The regularity claims are established above. It remains to verify 26.1.Since h ∈ L1 (Z,E) , it follows there exists a sequence of simple functions {hn} such

thathn (x)→ h(x) ,

∫Z∥hm−hn∥dλX → 0 as m,n→ ∞.

The first convergence above implies

hn ◦X → h◦X pointwise on Ω (26.3)

Then letting hn (x)=∑mk=1xkXEk (x) , where the Ek are disjoint and Borel, it follows easily

that hn ◦X is also a simple function of the form hn ◦X (ω) = ∑mk=1xkXX−1(Ek)

(ω) andby assumptionX−1 (Ek) ∈F . From the definition of the integral, it is easily seen∫

hn ◦XdP =∫

hndλX ,∫∥hn∥◦XdP =

∫∥hn∥dλX

Also, hn ◦X−hm ◦X is a simple function and so∫∥hn ◦X−hm ◦X∥dP =

∫∥hn−hm∥dλX (26.4)

It follows from the definition of the Bochner integral and 26.3, and 26.4 that h ◦X is inL1 (Ω;E) and ∫

h◦XdP = limn→∞

∫hn ◦XdP = lim

n→∞

∫hndλX =

∫hdλX .

Finally consider the case that E = Z for E a separable Banach space, and supposeX ∈ L1 (Ω;E). Then letting h be the identity map on E, it follows h is obviously separablyvalued and h−1 (U) ∈B (E) for all U open and so h is measurable. Why is it in L1 (E;E)?∫

E∥h(x)∥dλX =

∫∞

0λX ([∥h∥> t])dt ≡

∫∞

0P(X ∈ [∥x∥> t])dt

≡∫

0P([∥X∥> t])dt =

∫Ω

∥X∥dP < ∞

Thus the identity map on E is in L1 (E;λX) . Next let the identity map h be in L1 (E;λX) .Then X (ω) = h ◦X (ω) and so from the first part, X ∈ L1 (Ω;E) and from 26.1, 26.2follows. ■

26.2 Convergence in ProbabilityDefinition 26.2.1 { fn} is said to be Cauchy in probability if for each ε > 0,

limn,m→∞

P(∥ fn− fm∥> ε) = 0

This means: for each δ > 0 there exists kδ such that if m,n≥ kδ , then P(∥ fn− fm∥> ε)<δ .

718 CHAPTER 26. INDEPENDENCEIn the case where Z = E, a separable Banach space, if X is measurable then X ©L! (Q;E) if and only if the identity map on E is in L! (E;A x) and| X (@)dP = | adh x (w) (26.2)Q EProof: The regularity claims are established above. It remains to verify 26.1.Since h € L' (Z,E), it follows there exists a sequence of simple functions {h,} suchthatIn (aw) > h(a), [ Im — tall dx + O0-as m,n 0.ZThe first convergence above impliesh,o X — ho X pointwise on Q (26.3)Then letting h, (x) = VL, xe 2x, (x), where the E; are disjoint and Borel, it follows easilythat /, 0X is also a simple function of the form hyo X (@) = Lp | Be 2 x~1(~,) (@) andby assumption X~! (E,) € ¥. From the definition of the integral, it is easily seen[ine xXaP= | mddx. [\tnlloxaP =f \\mn\|da.xAlso, hyo X — hyo X is a simple function and so[tno X = hyo XP = | | hy In dx (26.4)It follows from the definition of the Bochner integral and 26.3, and 26.4 that ho X is inL! (Q;E) and[noxap= tim. [ hyo XdP = lim [inddx = [ hdax.. neon—-ooFinally consider the case that E = Z for E a separable Banach space, and supposeX €L!(Q;E). Then letting h be the identity map on E, it follows h is obviously separablyvalued and h~' (U) € @(E) for all U open and so h is measurable. Why is it in L! (E;E)?[We@janx = [rx (nl >aar= [PX e [lel > aha[ Pulxii>nar= [ |x\ar<«Thus the identity map on E is in L! (E;A x) . Next let the identity map h be in L' (E;A x).Then X (@) =ho X(q) and so from the first part, X € L'!(Q;E) and from 26.1, 26.2follows.26.2 Convergence in ProbabilityDefinition 26.2.1 {fn} is said to be Cauchy in probability if for each € > 0,lim P(\\tn — fill > E) =0nym—0oThis means: for each 6 > 0 there exists kg such that if m,n > kg, then P (|| fn — full > €) <6.