2338 CHAPTER 68. A DIFFERENT KIND OF STOCHASTIC INTEGRATION

which clearly converges to 0 as m→ ∞ provided that n≥ 2. In case n = 1, all you have todo is approximate XA from something in E1 and of course you just use XA.

Let f ,g ∈ En. Then from Lemma 68.3.9,

E((In ( f −g))2

)= n!

∥∥ f̃ − g̃∥∥2

L2(T n)

∥∥ f̃∥∥

L2(T n)=

(∫T· · ·∫

T

∣∣ f̃ (t)∣∣2 dt)1/2

=

∫T· · ·∫

T

∣∣∣∣∣ 1n! ∑

σ

f(tσ(1), · · · , tσ(n)

)∣∣∣∣∣2

dt

1/2

≤ 1n! ∑

σ

(∫T· · ·∫

T

∣∣ f (tσ(1), · · · , tσ(n))∣∣2 dt

)1/2

=1n! ∑

σ

∥ f∥L2(T n) = ∥ f∥L2(T n)

Therefore,

E((In ( f −g))2

)= n!

∥∥ f̃ − g̃∥∥2

L2(T n)≤ n!∥ f −g∥2

L2(T n) . (68.3.20)

The following theorem comes right away from this and Lemma 68.3.11.

Theorem 68.3.12 The integral In defined on En extends uniquely to an integral In definedon L2 (T n) . This integral satisfies

In ( f ) ∈ L2 (Ω)

AlsoE (In ( f ) In (g)) = n!

(f̃ , g̃)

L2(T n)

Proof: This follows right away from the density of En in L2 (T n) and the inequality68.3.20.

Obviously one wonders whether linear combinations ∑n cnIn ( fn) are dense in L2 (Ω) . Itlooks like the important thing to notice is that for f ∈ En, In ( f ) is a polynomial in W

(Aik

)≡

W(XAik

). Recall the corollary above, Corollary 68.2.4,

Corollary 68.3.13 Let P0n denote all polynomials of the form

p(W (h1) , · · · ,W (hk)) , degree of p≤ n, some h1, · · · ,hk

Also let Pn denote the closure in L2 (Ω,F ,P) of P0n . Then

Pn =⊕ni=0Hi

2338 CHAPTER 68. A DIFFERENT KIND OF STOCHASTIC INTEGRATIONwhich clearly converges to 0 as m — © provided that n > 2. In case n = 1, all you have todo is approximate 2, from something in & and of course you just use 24.Let f,g € &,. Then from Lemma 68.3.9,E((n(f—8))”) =a) ||P allracrnyflac) = ([--[ weerta)- [of> \ 1/2dt1ad (tec to(n))an oo1 > \1?< a (fo fle Guyot) at)= BL Mlle = MillenTherefore,E (((f—8))?) =a! Fallon) <mUllf—allizern- (68.3.20)The following theorem comes right away from this and Lemma 68.3.11.Theorem 68.3.12 The integral I, defined on &, extends uniquely to an integral I, definedon L* (T") . This integral satisfiesIn(f) €L? (Q)Also -E(In(f)In(8)) = 2! (F,8) 2pm)Proof: This follows right away from the density of &, in L?(T") and the inequality68.3.20.Obviously one wonders whether linear combinations Y,, CnJn (fn) are dense in L? (Q) . Ilooks like the important thing to notice is that for f € &,, 1, (f) is a polynomial in W (Ai, )WwW (24,,)- Recall the corollary above, Corollary 68.2.4,ooCorollary 68.3.13 Let P denote all polynomials of the formp(W(h),---,W(hx)), degree of p <n, some hy,--- , hgAlso let P;, denote the closure in L? (Q,.F ,P) of Po. ThenPn = Dio Hi