2260 CHAPTER 65. STOCHASTIC INTEGRATION

Now consider

P

(∣∣∣∣∫ T

0X[0,τnp]ΦndW −

∫ T

0X[0,τ p]ΦdW

∣∣∣∣2 > λ

)

≤E(∣∣∣∫ T

0 X[0,τnp]ΦndW −∫ T

0 X[0,τ p]ΦdW∣∣∣2)

λ

and so, there exists a subsequence, still denoted as n such that

P

(∣∣∣∣∫ T

0X[0,τnp]ΦndW −

∫ T

0X[0,τ p]ΦdW

∣∣∣∣2 > 1n

)< 2−k

It follows that N can be enlarged so that for ω /∈ Np∣∣∣∣∫ T

0X[0,τnp]ΦndW −

∫ T

0X[0,τ p]ΦdW

∣∣∣∣2 ≤ 1n

for all n large enough. Now obtain a succession of subsequences for p = 1,2, · · · , each asubsequence of the preceeding one such that the above convergence takes place and let Ninclude ∪pNp. Then for ω /∈ N, and letting n denote the diagonal sequence, it follows thatfor all p,

limn→∞

∣∣∣∣∫ T

0X[0,τnp]ΦndW −

∫ T

0X[0,τ p]ΦdW

∣∣∣∣= 0

For ω /∈ N, there is a p such that τ p = ∞. Then this means∫ T

0 ∥Φ∥2 ds < p. It follows that

the same is true for Φn for all n large enough. Hence τnp = ∞ also. Thus, for large enoughn, ∣∣∣∣∫ T

0ΦndW −

∫ T

0ΦdW

∣∣∣∣= ∣∣∣∣∫ T

0X[0,τnp]ΦndW −

∫ T

0X[0,τ p]ΦdW

∣∣∣∣and the latter was just shown to converge to 0.

65.12 The Holder Continuity Of The IntegralLet Φ ∈ L2

([0,T ]×Ω,L2

(Q1/2U,H

)). Then you can consider the stochastic integral as

described above and it yields a continuous function off a set of measure zero. What ifΦ ∈ L∞

([0,T ]×Ω,L2

(Q1/2U,H

))? Can you say more? The short answer is yes. You

obtain a Holder condition in addition to continuity. This is a consequence of the BurkolderDavis Gundy inequality and Corollary 65.11.1 above. Let α > 2. Let ∥Φ∥

∞denote the

norm in L∞([0,T ]×Ω,L2

(Q1/2U,H

)). By the Burkholder Davis Gundy inequality,∫

(∣∣∣∣∫ t

sΦdW

∣∣∣∣)α

dP≤

∫Ω

(sup

r∈[s,t]

∣∣∣∣∫ r

sΦdW

∣∣∣∣)α

dP≤C∫

(∫ t

s∥Φ∥2 dτ

)α/2

dP

2260 CHAPTER 65. STOCHASTIC INTEGRATIONT T 2(| [ Fey) PrndW [ io) PAW >)E (\i 0 yp] PndW — fo Zin 24W| ): Xrand so, there exists a subsequence, still denoted as n such that21> <2*n2 4<-nNow considerT T°(| [ Kio tap] Pn — [ io] PAWIt follows that N can be enlarged so that for @ ¢ NpT Ti Foam] PrdW ~ fF Zo.g) PAWfor all n large enough. Now obtain a succession of subsequences for p = 1,2,---, each asubsequence of the preceeding one such that the above convergence takes place and let NVinclude U,N>p. Then for @ ¢ N, and letting n denote the diagonal sequence, it follows thatfor all p,limn—yooT T[ Fogg) Pn — | Zine" =0For @ ¢ N, there is a p such that t, =e. Then this means [j ||®||’ ds < p. It follows thatthe same is true for ®, for all n large enough. Hence Ty) = - also. Thus, for large enough”" T T T T[ dW — [ edW| = | [ Ki ctup| PndW [ Zin)and the latter was just shown to converge to 0. I65.12 The Holder Continuity Of The IntegralLet ® € L? (0, T] xQ,L% (g'/ °U,H )). Then you can consider the stochastic integral asdescribed above and it yields a continuous function off a set of measure zero. What if® € L* ((0,7] x Q, ZH (Q'/?U,H))? Can you say more? The short answer is yes. Youobtain a Holder condition in addition to continuity. This is a consequence of the BurkolderDavis Gundy inequality and Corollary 65.11.1 above. Let a > 2. Let ||®||,, denote thenorm in L® ([0,7] x Q, % (o'/ ?U,H)). By the Burkholder Davis Gundy inequality,t aL (fel) dP<Q sr “ t a/2[o0w|) ap<c | (/ lar) dPs Q \Ys| supQ \reE{s,t]