2608 CHAPTER 77. STOCHASTIC INCLUSIONS

ψ to the closed subspace W is at least 1/5. Now C ,∪{C ∪{ψ}} would violate maximalityof C . Hence W =V ′. Now it follows that C must be countable since otherwise, V ′ wouldfail to be separable. Let M be the rational linear combinations of D . It must be dense in V ′.Note that linear combinations of the φ i are uniquely determined because none is a linearcombination of the others. Now define a linear mapping on M which makes sense for (t,ω)on a certain set.

Definition 77.2.15 Let E be those points (t,ω) such that the following limit exists for eachφ ∈D

Λ(t,ω)φ ≡ limmφ→∞

∫ t

lmφ(t)⟨φ ,v(s,ω)⟩ds

The set of points where the limit of measurable functions exists is always measurable so Eis a measurable set. Extend this mapping linearly. That is, for ψ ∈M,ψ ≡ ∑i aiφ i,

Λ(t,ω)ψ ≡∑i

aiΛ(t,ω)φ i = ∑i

ai

(lim

mφ i→∞mφ i

∫ t

lmφ i(t)⟨φ i,v(s,ω)⟩ds

)Thus (t,ω)→ Λ(t,ω)ψ is product measurable, being the sum of limits of product measur-able functions. Let G denote those (t,ω) in E such that there exists a constant C (t,ω) suchthat for all ψ ∈M,

|Λ(t,ω)ψ| ≤C (t,ω)∥ψ∥

Lemma 77.2.16 G is product measurable.

Proof: This follows from the formula

E ∩GC = ∩n∪ψ∈M {(t,ω) ∈ E : |Λ(t,ω)ψ|> n∥ψ∥}

which is clearly product measurable because (t,ω)→ Λ(t,ω)ψ is. Thus, since E is mea-surable, it follows that E ∩G = G is also.

For (t,ω) ∈ G,Λ(t,ω) has a unique extension to all of V , the dual space of V ′, stilldenoted as Λ(t,ω). By the Riesz representation theorem, for (t,ω) ∈ G, there existsu(t,ω) ∈V,

Λ(t,ω)ψ = ⟨ψ,u(t,ω)⟩V ′,VThus (t,ω)→XG (t,ω)u(t,ω) is product measurable by the Pettis theorem. Let u = 0 offG. We know G is product measurable. For each ω,{t : (t,ω) ∈ G} has full measure. Thisinvolves the fundamental theorem of calclulus.

Fix ω . By the fundamental theorem of calculus,

limm→∞

m∫ t

lm(t)v(s,ω)ds = v(t,ω) in V

for a.e. t say for all t /∈ N (ω) ⊆ [0,T ]. Of course we do not know that ω → v(t,ω)is measurable. However, the existence of this limit for t /∈ N (ω) implies that for everyφ ∈V ′,

limm→∞

∣∣∣∣m∫ t

lm(t)⟨φ ,v(s,ω)⟩ds

∣∣∣∣≤C (t,ω)∥φ∥

2608 CHAPTER 77. STOCHASTIC INCLUSIONSy to the closed subspace W is at least 1/5. Now @,U{@ U { w}} would violate maximalityof @. Hence W = V’. Now it follows that @ must be countable since otherwise, V’ wouldfail to be separable. Let M be the rational linear combinations of Y. It must be dense in V’.Note that linear combinations of the @; are uniquely determined because none is a linearcombination of the others. Now define a linear mapping on M which makes sense for (t, @)on a certain set.Definition 77.2.15 Let E be those points (t,@) such that the following limit exists for eachGEDAro) 0 timmy | (@,v(s,@)) dsThe set of points where the limit of measurable ve rion exists is always measurable so Eis a measurable set. Extend this mapping linearly. That is, for wy © M, wy = Yai,A(t,@) v= Dad ( t,@) d; =Lai (_ me [ 1 ovlsanyas)mg;Thus (t,@) + A(t,@) w is product measurable, being the sum of limits of product measur-able functions. Let G denote those (t,@) in E such that there exists a constant C (t,@) suchthat for all y € M,|A(t,@) yw] <C(z,@) ||y|Lemma 77.2.16 G is product measurable.Proof: This follows from the formulaENG =MnUyem {(t,) € E: |A(t,@) w| > nly}which is clearly product measurable because (t,@) — A(t,@) y is. Thus, since E is mea-surable, it follows thatE 1G=Gisalso. JFor (t,@) € G,A(t,@) has a unique extension to all of V, the dual space of V’, stilldenoted as A(t,@). By the Riesz representation theorem, for (t,@) € G, there existsu(t, @) EV,A(t,@) v= (W,u(t,@))yryThus (t,@) > %G(t,@) u(t, @) is product measurable by the Pettis theorem. Let u = 0 offG. We know G is product measurable. For each @, {t : (t,@) € G} has full measure. Thisinvolves the fundamental theorem of calclulus.Fix @. By the fundamental theorem of calculus,tlim m v(s,@)ds =v(t,@) inVim, att) ) (t,@)for a.e. t say for all t ¢ N(@) C [0,7]. Of course we do not know that @ > v(t,@)is measurable. However, the existence of this limit for t ¢ N(q@) implies that for everyoeV’,limm—- oom [| (@v(s.0)) ds) <C(r0) ||