Chapter 66

The Integral∫ t

0 (Y,dM)HFirst the integral is defined for elementary functions.

Definition 66.0.1 Let an elementary function be one which is of the form

m−1

∑i=0

YiX(ti,ti+1] (t)

where Yi is Fti measurable with values in H a separable real Hilbert space for 0 = t0 <t1 < · · ·< tm = T.

Definition 66.0.2 Now let M be a H valued continuous local martingale, M (0) = 0. Thenfor Y a simple function as above,∫ t

0(Y,dM)≡

m−1

∑i=0

(Yi,M (t ∧ ti+1)−M (t ∧ ti))H

Assumption 66.0.3 We will always assume that d [M] is absolutely continuous with respectto Lebesgue measure. Thus d [M] = kdt where k ≥ 0 and is in L1 ([0,T ]×Ω). This is doneto avoid technical questions related to whether t→

∫ t0 d [M] is continuous and also to make

it easier to get examples of a certain class of functions.

This includes the usual stochastic integral M (t) =∫ t

0 ΦdW where [M] (t) =∫ t

0 ∥Φ∥2L2

dsso d [M] = ∥Φ∥2 dt.

Next is to consider how this relates to stopping times which have values in the {ti}. Letτ be a stopping time which takes the values {ti}m

i=0. Then∫ t∧τ

0(Y,dM)≡

m−1

∑i=0

(Yi,M (t ∧ ti+1∧ τ)−M (t ∧ ti∧ τ))H (66.0.1)

Now consider X[0,τ]Y. Is it also an elementary function?

X[0,τ]Y =m−1

∑i=0

X[0,τ] (t)YiX(ti,ti+1] (t)

To get the ith term to be non zero, you must have τ ≥ t and t ∈ (ti, ti+1]. Thus it must be thecase that τ > ti. Also, if τ > ti and t ∈ (ti, ti+1], then τ ≥ ti+1 because τ has only the valuesti. Hence also τ ≥ t. Thus the above sum reduces to

m−1

∑i=0

X[τ>ti] (ω)YiX(ti,ti+1] (t)

This shows that X[0,τ]Y is of the right sort, the sum of Fti measurable functions timesX(ti,ti+1] (t). Thus from the definition of this funny integral,

∫ t

0

(X[0,τ]Y,dM

)≡

m−1

∑i=0

Fti︷ ︸︸ ︷

X[τ>ti] (ω)Yi,M (t ∧ ti+1)−M (t ∧ ti)

H

(66.0.2)

2273

Chapter 66The Integral {; (Y,dM),First the integral is defined for elementary functions.Definition 66.0.1 Let an elementary function be one which is of the formm—1» YQ titi] (7)i=0where Y; is F;, measurable with values in H a separable real Hilbert space for 0 = to <th <0 <tn =T.Definition 66.0.2 Now let M be a H valued continuous local martingale, M (0) = 0. Thenfor Y a simple function as above,m—1[ (¥.dM) = ¥ (%.M(tAtixt) —M(tAti))yi=0Assumption 66.0.3 We will always assume that d [M] is absolutely continuous with respectto Lebesgue measure. Thus d{|M] = kdt where k > 0 and is in L' ({0,T] x Q). This is doneto avoid technical questions related to whether t —+ Jd |M] is continuous and also to makeit easier to get examples of a certain class of functions.This includes the usual stochastic integral M (t) = {j ®dW where [M] (t) = Jo |||, dsso d[M] = ||®||? dr.Next is to consider how this relates to stopping times which have values in the {7;}. LetT be a stopping time which takes the values {t;}\" 9. Thent\T m-1| (¥,dM) = Vo (%,M(tAtini At) M(t At) (66.0.1)0 i=0Now consider 279 7Y. Is it also an elementary function?m—|Lio a\¥ = d 20,1) (O)Yi 2 iti.) (t)To get the i” term to be non zero, you must have t >t and ¢ € (¢;,t;1]. Thus it must be thecase that tT > ¢;. Also, if t >t; and t € (¢;,t;41], then tT > t;,; because T has only the valuest;. Hence also t > t. Thus the above sum reduces tom—1y Lest] (a) YX iti] 2)i=0This shows that Zio qY is of the right sort, the sum of -¥;, measurable functions timesX n;,1| (t)- Thus from the definition of this funny integral,Frt ee[ (Fog ¥edM) = | Qe (@)VM (Ati) —M (An) (66.0.2)i=0H2273