2300 CHAPTER 67. THE EASY ITO FORMULA

67.9 Some Representation TheoremsIn this section is a very interesting representation theorem which comes from the Ito for-mula. In all of this, W will be a Q Wiener process having values in Rn for which Q = I.Recall that, letting

Gt ≡ σ (W(s) : s≤ t)

the normal filtration determined by the Wiener process is given by

Ft ≡ ∩s>tGs

where Gs is the completion of Gs. In this section, the theorems will all feature the smallerfiltration Gt , not the filtration Ft . First here are some simple observations which tie thisspecialized material to what was presented earlier.

When you have f an Gt adapted function in L2 (Ω,Rn) , you can consider

fT ∈ L2([0,T ]×Ω;L2

(Q1/2Rn,R

))as follows. Letting {gi} be an orthonormal basis for the subspace Q1/2Rn in the norm ofQ1/2Rn,

∥f∥2L2(Q1/2Rn,R) ≡∑

i

(fT gi

)2< ∞

For simplicity, let Q = I. Then you have the simple situation that∥∥fT∥∥L2(Q1/2Rn,R) = ∥f∥

2Rn

In what follows Wt will be the Q Wiener process onRn where Q = I. Then the Ito isometryis nothing more than the following lemma.

Lemma 67.9.1 Let f be Ft adapted in the sense that every component is Ft adapted andf ∈ L2 (Ω;Rn). Here Ft is the normal filtration coming from the Wiener process. Then∣∣∣∣∣∣∣∣∫ T

0f(s)T dW

∣∣∣∣∣∣∣∣L2(Ω)

= ||f||L2(Ω×[0,T ];Rn) .

Lemma 67.9.2 Let X ≥ 0 and measurable and integrable. Also define a finite measure ν

on B (Rp) by

ν (B)≡∫

XXB (Y)dP

Then ∫Ω

g(Y)XdP =∫Rp

g(y)dν (y)

where here Y is a measurable function with values in Rp and g ≥ 0 is Borel measurable.Formally, XdP = dν .

2300 CHAPTER 67. THE EASY ITO FORMULA67.9 Some Representation TheoremsIn this section is a very interesting representation theorem which comes from the Ito for-mula. In all of this, W will be a Q Wiener process having values in R” for which Q = J.Recall that, lettingG =o (W(s):s5 <1)the normal filtration determined by the Wiener process is given byF,= Ass1%where &, is the completion of %,. In this section, the theorems will all feature the smallerfiltration Y, not the filtration ¥,. First here are some simple observations which tie thisspecialized material to what was presented earlier.When you have f an Y adapted function in L? (Q,1R”) , you can considerfe ((0. T)xQ:B (o'?R",R) Jas follows. Letting {g;} be an orthonormal basis for the subspace Q!/*R” in the norm ofQ!'/2R"2 _ T .\2If’, (o12R".R) = Li (fi gi) <-LFor simplicity, let Q = 7. Then you have the simple situation that|e" IPACHE) = lfllienIn what follows W, will be the Q Wiener process on R” where Q = I. Then the Ito isometryis nothing more than the following lemma.Lemma 67.9.1 Let f be ¥; adapted in the sense that every component is ¥, adapted andf € L? (Q;R"). Here F, is the normal filtration coming from the Wiener process. ThenFaw|| = Iilaxo.rien:L?(Q)Lemma 67.9.2 Let X > 0 and measurable and integrable. Also define a finite measure Von &(R?) by .= [ x%n(vyar[em )XdP = ep oy )dv(y)where here Y is a measurable function with values in R? and g > 0 is Borel measurable.Formally, XdP = dv.Then