2484 CHAPTER 73. THE HARD ITO FORMULA, IMPLICIT CASE

for all k large enough. By the usual proof of completeness of Lp, it follows that X lnk(t)→

X (t) for a.e. t, this for each ω /∈ N, a similar assertion holding for X rnk

. We denote thesesubsequences as

{X r

k

}∞

k=1 ,{

X lk

}∞

k=1 .Now with this preparation, it is possible to show the desired convergence.

Lemma 73.6.3 In the above context, let X (s)−X lk (s)≡ ∆k (s) . Then the integral∫ t

0

(Z ◦ J−1)∗BX ◦ JdW

exists as a local martingale and the following limit is valid for the subsequence of Lemma73.6.2

limk→∞

P

([sup

t∈[0,T ]

∣∣∣∣∫ t

0

(Z ◦ J−1)∗B∆k ◦ JdW

∣∣∣∣≥ ε

])= 0.

That is,

supt∈[0,T ]

∣∣∣∣∫ t

0

(Z ◦ J−1)∗B∆k ◦ JdW

∣∣∣∣converges to 0 in probability.

Proof: In the argument τm will be defined in 73.6.20. Let

Ak ≡

{ω : sup

t∈[0,T ]

∣∣∣∣∫ t

0

(Z ◦ J−1)∗B∆k ◦ JdW

∣∣∣∣≥ ε

}

then

Ak ∩{ω : τm = ∞} ⊆

{ω : sup

t∈[0,T ]

∣∣∣∣∫ t

0

(Z ◦ J−1)∗B∆

τmk ◦ JdW

∣∣∣∣≥ ε

}By Burkholder Davis Gundy inequality,

P(Ak ∩{ω : τm = ∞}) ≤ Cε

∫Ω

supt∈[0,T ]

∣∣∣∣∫ t

0

(Z ◦ J−1)∗B∆

τmk ◦ JdW

∣∣∣∣dP

≤ Cε

∫Ω

(∫ T

0∥Z∥2

L2

∥∥B∆τmk

∥∥2 dt)1/2

dP

≤ Cε

(∫Ω

∫ T

0∥Z∥2

L2

∥∥B∆τmk

∥∥2 dtdP)1/2

Recall that if ⟨Bx,x⟩ ≤ m, then ∥Bx∥W ′ ≤ m1/2 ∥B∥1/2. Then the integrand is bounded fora.e. t by ∥Z∥2

L24m∥B∥ . Next use the result of Lemma 73.6.2 and the dominated con-

vergence theorem to conclude that the above converges to 0 as k → ∞. Then from theassumption that τm = ∞ for all m large enough,

P(Ak) =∞

∑m=1

P(Ak ∩ ([τm = ∞]\ [τm−1 < ∞]))

2484 CHAPTER 73. THE HARD ITO FORMULA, IMPLICIT CASEfor all k large enough. By the usual proof of completeness of L”, it follows that Xi, (t) >X (t) for a.e. t, this for each @ ¢ N, a similar assertion holding for X;,. We denote thesesubsequences as {Xp}, ; {Xi}, om |Now with this preparation, it is possible to show the desired convergence.Lemma 73.6.3 In the above context, let X (s) — X/ (s) = Ax (s). Then the integralt[ (ZoJ!)* BX oJdW0exists as a local martingale and the following limit is valid for the subsequence of Lemma73.6.2t oklim P| | sup [ @or) Bd, oJdW| > e =0.k-400 te[0,7] 170That is,tsup [ Gory’ Baxosawte[0,7] 170converges to 0 in probability.Proof: In the argument 7, will be defined in 73.6.20. LetA, = @: supte [0,7]t| Zor)" Bay osaW >|0thenAg N{@ : Tm =} C4 @: supte[0,7]t *| (ZoJ-')” BA,” osaW| > eS0By Burkholder Davis Gundy inequality,CcP(AgN{@: Tn =c}) < =] supE JQ 4€(0,7]Cc T 2 1/2< <[ (/ \ZI\%,, Bar| ar) dPC T 2 t 2 1/2<= Z BA," ||" dtdP~ € (Lf Al kRecall that if (Bx,x) < m, then ||Bx||y: <m'/2 ||B||'/?. Then the integrand is bounded forae. t by IZ, 4m ||B||. Next use the result of Lemma 73.6.2 and the dominated con-vergence theorem to conclude that the above converges to 0 as k — oo. Then from theassumption that 7, = ce for all m large enough,tZoJ~!)* BA™ oJdW| dP( H0©P(Ag) = Y) P(ARM ([tm = 29] \ [tm—1 <e]))m=1