79.4. STOCHASTIC INCLUSIONS WITHOUT UNIQUENESS ?? 2721

which is P measurable. Thus G is a measurable multifunction.Since (t,ω)→ G(t,ω) is measurable, there is a sequence {wn (t,ω)} of measurable

functions such that ∪∞n=1wn (t,ω) equals

G(t,ω) =

{w : ⟨w,u(t,ω)− y(t,ω)⟩ ≤ α (t,ω)+

1n

, (t,ω) /∈ Σ

}∩B(0,γ)

As shown above, there exists wt,ω in A(u(t,ω) , t,ω) as well as G(t,ω) . Thus there isa sequence of wr (t,ω) converging to wt,ω . Of course r will need to depend on t,ω . Since(t,ω)→ A(u(t,ω) , t,ω) is a P measurable multifunction, it has a countable subset of Pmeasurable functions {zk (t,ω)} which is dense in A(u(t,ω) , t,ω). Let Uk be defined as

Uk (t,ω)≡ ∪mB(

zm (t,ω) ,1k

)⊆ A(u(t,ω) , t,ω)+B

(0,

2k

)Now define A1k = {(t,ω) : w1 (t,ω) ∈Uk (t,ω)} . Then let

A2k = {(t,ω) /∈ A1k : w2 (t,ω) ∈Uk (t,ω)}

andA3k =

{(t,ω) /∈ ∪2

i=1Aik : w3 (t,ω) ∈Uk (t,ω)}

and so forth. Any (t,ω) ∈ Sγ must be contained in one of these Ark for some r since if notso, there would not be a sequence wr (t,ω) converging to wt,ω ∈ A(u(t,ω) , t,ω). TheseArk partition Sγ \Σ and each is measurable since the {zk (t,ω)} are measurable. Let

ŵk (t,ω)≡∞

∑r=1

XArk (t,ω)wr (t,ω)

Thus ŵk (t,ω) is in Uk (t,ω) for all (t,ω) ∈ Sγ and equals exactly one of the wm (t,ω) ∈G(t,ω).

Also, by construction, the ŵk (·, ·) are bounded in L∞(Sγ ;V ′

). Therefore, there is a sub-

sequence of these, still called ŵk which converges weakly to a function w in L2(Sγ ;V ′

).

Thus w is a weak limit point of co(∪∞

j=kŵ j

)for each k. Therefore, in the open ball

B(w, 1

k

)⊆ L2

(Sγ ;V ′

)with respect to the strong topology, there is a convex combination

∑∞j=k c jkŵ j (the c jk add to 1 and only finitely many are nonzero) which converges strongly

in L2(Sγ ;V ′

). Since G(t,ω) is convex and closed, this convex combination is in G(t,ω).

Off a set of P measure zero, we can assume this convergence of ∑∞j=k c jkŵ j as k→ ∞

happens pointwise a.e. for a suitable subsequence. However,∞

∑j=k

c jkŵ j (t,ω) ∈Uk (t,ω)⊆ A(u(t,ω) , t,ω)+B(

0,2k

).

Thus w(t,ω) ∈ A(u(t,ω) , t,ω) a.e. (t,ω) because A(u(t,ω) , t,ω) is a closed set. Sincew is the pointwise limit of measurable functions off a set of measure zero, it can be as-sumed measurable and for a.e. (t,ω), w(t,ω) ∈ A(u(t,ω) , t,ω)∩G(t,ω). Now denotethis measurable function wn. Then

wn (t,ω) ∈ A(u(t,ω) , t,ω) ,⟨wn (t,ω) ,u(t,ω)− y(t,ω)⟩ ≤ α (t,ω)+1n

a.e. (t,ω)

79.4. STOCHASTIC INCLUSIONS WITHOUT UNIQUENESS ?? 2721which is Y measurable. Thus G is a measurable multifunction.Since (t,@) — G(t,@) is measurable, there is a sequence {w, (t,@)} of measurablefunctions such that U*_, w, (t,@) equalsGE.) = {w: (mu(t,0)-y(1.0)) <a(,0)+ +, (0) ¢ ELBOWAs shown above, there exists w;@ in A (u(t, @),t,@) as well as G(t,@) . Thus there isa sequence of w, (t, @) converging to wy. Of course r will need to depend on t, @. Since(t,@) > A(u(t,@) ,t,@) isa Y measurable multifunction, it has a countable subset of Fmeasurable functions {z, (t, @)} which is dense in A (u(t, @) ,t,@). Let U; be defined asU; (t, 0) =UpB (<n (1,0), z) CA(u(t,@) ,t,0) +B (0 |Now define Aj, = {(t,@) : w1 (t,@) € Ux (t, @)}. Then letAx = {(t, ©) £ Aix : wa (t, @) € Ux (t, @)}andA3x = {(t,@) ¢ U2 Aig [W3 (t, @) EU; (t,@) }and so forth. Any (t,@) € Sy must be contained in one of these A,, for some r since if notso, there would not be a sequence w,(t,@) converging to w;,@ € A(u(t,@) ,t,@). TheseA, partition Sy \ & and each is measurable since the {z; (t, @) } are measurable. LetWr (t,@) = y? Kang (t,@) w, (t,@)r=1Thus w; (t,@) is in Ug (t,@) for all (t,@) € Sy and equals exactly one of the wm (t,@) €G(t,@).Also, by construction, the ¥, (-,-) are bounded in L® (Sy; v’). Therefore, there is a sub-sequence of these, still called ; which converges weakly to a function w in L? (Sy; y! ) .Thus w is a weak limit point of co Gru i) for each k. Therefore, in the open ballB(w,z) CL’ (Sy:V’) with respect to the strong topology, there is a convex combinationYin C jk; (the c jx add to 1 and only finitely many are nonzero) which converges stronglyin L? (Sy;V’). Since G(t,@) is convex and closed, this convex combination is in G(t, @).Off a set of A measure zero, we can assume this convergence of Vik CjKW] as k — 00happens pointwise a.e. for a suitable subsequence. However,— 2C30 (00) CU (10) A(u(t.0) 4.0) +B (0,2).j=kThus w(t,@) € A(u(t,@),t,@) ae. (t,@) because A (u(t, @),t,@) is a closed set. Sincew is the pointwise limit of measurable functions off a set of measure zero, it can be as-sumed measurable and for a.e. (t,@), w(t,@) € A(u(t,@) ,t,@)G(t,@). Now denotethis measurable function w,. ThenWn (t,@) €A (u(t, @),t,@), (Wa (t,@),u(t,@)—y(t,@)) < a(t,0)+— ae. (t,@)