2712 CHAPTER 79. INCLUDING STOCHASTIC INTEGRALS

Then

P

(ω : sup

t∈[0,T ]

∥∥∥∥unk (t)+∫ t

0znk ds−

(u(t)+

∫ t

0zds)∥∥∥∥

U ′> 2−k

)

≤ 2k∫

∫ T

0

1nk

∣∣∣∣Funk

∣∣∣∣dsdP < 2−k

Therefore there is a subsequence still denoted with n such that

P

(ω : sup

t∈[0,T ]

∥∥∥∥un (t)+∫ t

0znds−

(u(t)+

∫ t

0zds)∥∥∥∥

U ′> 2−n

)≤ 2−n

It follows that there is a set of measure zero N such that if ω /∈ N, then ω is in only finitelymany of the above sets. That is, for ω /∈ N,

supt∈[0,T ]

∥∥∥∥un (t,ω)+∫ t

0zn (s,ω)ds−

(u(t,ω)+

∫ t

0z(s,ω)ds

)∥∥∥∥U ′

< 2−n

for all n large enough.

Lemma 79.4.6 There is a set of measure zero N, an enlargement of the earlier set suchthat for ω /∈ N, and a suitable subsequence, still denoted with n, such that

limn→∞

(un (t,ω)+

∫ t

0zn (s,ω)ds

)= u(t,ω)+

∫ t

0z(s,ω)ds in U ′ (79.4.51)

for each t ∈ [0,T ]. In addition to this, for ω /∈ N,∥∥∥∥un (t,ω)−u0 (ω)+∫ t

0(zn (s,ω)− f (s,ω))ds−

∫ t

0ΦdW

∥∥∥∥U ′→ 0 (79.4.52)

Proof: The formula 79.4.51 follows from the above discussion. The second claimfollows from the first and the equation satisfied by u 79.4.49.

In the situation of 79.4.44, can we conclude that for the subsequence of Lemma 79.4.6,

lim supn→∞

⟨zn,un−u⟩U ′,U ≤ 0? (79.4.53)

Let {ek} be a complete orthonormal set in H, dense in H, with each vector in U . Thus

un (t,ω) = ∑k(un (t,ω) ,ek)H ek, |un (t,ω)|2 = ∑

k(un (t,ω) ,ek)

2 . (79.4.54)

The following claim is the key idea which will yield 79.4.53.Claim: liminfn→∞ (un (t,ω) ,ek)

2 ≥ (u(t,ω) ,ek)2 for a.e.(t,ω).

Proof of claim: Let Bε be those (t,ω) such that

lim infn→∞

(un (t,ω) ,ek)2 ≤ (u(t,ω) ,ek)

2− ε.

2712 CHAPTER 79. INCLUDING STOCHASTIC INTEGRALSin (t)+ [ends (« w+ sts) . >)vT<2 / | © |\Fun||dsdP <2JQ SO kThenP| @: supte[0,7]Therefore there is a subsequence still denoted with n such thatt tP| @: sup ||un w+ [ Znds — (un [ cds) >27)<2"t€(0,7] 0 0 U!It follows that there is a set of measure zero N such that if @ ¢ N, then @ is in only finitelymany of the above sets. That is, for @ ¢ N,<2"U'supt€(0,7]tn (t,00)+ [ en(050)ds— (u(r.0)+ [(s.0)as)for all n large enough.Lemma 79.4.6 There is a set of measure zero N, an enlargement of the earlier set suchthat for @ ¢ N, and a suitable subsequence, still denoted with n, such thatt tlim (« (t,@) + [ zn(s, ads) = u(t,@) + [ z(s,@)ds in U' (79.4.51)n—soofor each t € [0,T]. In addition to this, for o ¢ N,Un (t, 0) — up (@) + [ "(en (s,00) — f(s,)) ds — [ ‘eaw|| 30 (794.52)U'Proof: The formula 79.4.51 follows from the above discussion. The second claimfollows from the first and the equation satisfied by u 79.4.49. JjIn the situation of 79.4.44, can we conclude that for the subsequence of Lemma 79.4.6,lim sup (Zn, Un — Ua) a <0? (79.4.53)n—ooLet {e,} be a complete orthonormal set in H, dense in H, with each vector in U . ThusUn (t,@) = Y (un (t,@) ek) 4 ek [Un (t,0) |” = V° (Un (t,@) ex)” (79.4.54)k kThe following claim is the key idea which will yield 79.4.53.Claim: liminf;—,.0 (Up (t, @) sek) > (u(t, @) sek) for a.e.(t,@).Proof of claim: Let Bz be those (¢,@) such thatlim inf (y(t, 0) ,ex)” < (u(t,@) ,ex)” —€.n—co