Stochastic Analysis

Andreas Eberle

July 13, 2015 Contents

Contents 2

1LévyprocessesandPoissonpointprocesses 7 1.1 Lévy processes ...... 8 Characteristicexponents ...... 9 Basicexamples ...... 10 CompoundPoissonprocesses...... 12 Exampleswithinﬁnitejumpintensity ...... 15 1.2 Martingales and Markov property ...... 18 Martingales of Lévy processes ...... 19 LévyprocessesasMarkovprocesses ...... 20 1.3 Poisson random measures and Poisson point processes ...... 23 ThejumptimesofaPoissonprocess ...... 24 ThejumpsofaLévyprocess ...... 26 Poissonpointprocesses...... 28 Constructionof compound Poisson processes from PPP ...... 31 1.4 Stochastic integrals w.r.t. Poisson point processes ...... 33 Elementaryintegrands ...... 34 Lebesgueintegrals ...... 36 Itô integrals w.r.t. compensated Poisson point processes ...... 38 1.5 Lévy processes with inﬁnite jump intensity ...... 39 ConstructionfromPoissonpointprocesses ...... 40 TheLévy-Itôdecomposition ...... 44

2 CONTENTS 3

Subordinators ...... 46 Stableprocesses...... 49

2TransformationsofSDE 52 2.1 Lévy characterizations and martingale problems ...... 54 Lévy’scharacterizationofBrownianmotion ...... 55 Martingale problem for Itô diffusions ...... 59 Lévycharacterizationofweaksolutions ...... 62 2.2 Random time change ...... 65 Continuous local martingales as time-changed Brownian motions . . . . 65 Time-change representations of stochastic integrals ...... 67 Time substitution in stochastic differential equations ...... 68 One-dimensionalSDE ...... 70 2.3 Change of measure ...... 73 Changeofmeasureonﬁlteredprobabilityspaces ...... 75 Girsanov’sTheorem...... 77 Novikov’scondition...... 79 ApplicationstoSDE ...... 81 Doob’s h-transform ...... 82 2.4 Path integrals and bridges ...... 83 Pathintegralrepresentation ...... 85 TheMarkovproperty ...... 86 Bridgesandheatkernels ...... 87 SDEfordiffusionbridges...... 90 2.5 Large deviations on path spaces ...... 91 TranslationsofWienermeasure ...... 91 Schilder’sTheorem ...... 93 Randomperturbationsofdynamicalsystems ...... 97

3ExtensionsofItôcalculus 100 3.1 SDE with jumps ...... 101 Lp Stability...... 103

University of Bonn Summer Semester 2015 4 CONTENTS

Existenceofstrongsolutions ...... 105 Non-explosioncriteria ...... 107 3.2 Stratonovich differential equations ...... 109 Itô-Stratonovich formula ...... 110 StratonovichSDE...... 111 Brownianmotiononhypersurfaces ...... 112 Doss-Sussmannmethod...... 115 WongZakaiapproximationsofSDE ...... 117 3.3 Stochastic Taylor expansions ...... 118 Itô-Taylorexpansions...... 119 3.4 Numerical schemes for SDE ...... 124 Strongconvergenceorder...... 125 Weakconvergenceorder ...... 129 3.5 Local time ...... 131 Localtimeofcontinuoussemimartingales ...... 132 Itô-Tanaka formula ...... 136 3.6 Continuous modifications and stochastic flows ...... 138 Continuous modifications of deterministic functions ...... 139 Continuousmodificationsofrandomfields ...... 141 Existenceofacontinuousflow ...... 142 Markovproperty ...... 145 Continuityoflocaltime...... 145

4VariationsofparametersinSDE 148 4.1 Variations of parameters in SDE ...... 149 Differentationofsolutionsw.r.t.aparameter ...... 150 DerivativeﬂowandstabilityofSDE ...... 151 Consequencesforthetransitionsemigroup ...... 155 4.2 Malliavin gradient and Bismut integration by parts formula ...... 156 Gradient and integration by parts for smooth functions ...... 158 Skorokhodintegral ...... 162 DeﬁnitionofMalliavingradientII ...... 162

Stochastic Analysis Andreas Eberle CONTENTS 5

Productandchainrule ...... 165 4.3 Digression on Representation Theorems ...... 165 Itôs Representation Theorem ...... 166 Clark-Oconeformula ...... 169 4.4 First applications to SDE ...... 169 4.5 Existence and smoothness of densities ...... 169

5Stochasticcalculusforsemimartingaleswithjumps 170 Semimartingalesindiscretetime ...... 171 Semimartingalesincontinuoustime ...... 172 5.1 Finite variation calculus ...... 175 Lebesgue-Stieltjesintegralsrevisited ...... 176 Productrule...... 177 Chainrule...... 180 Exponentialsofﬁnitevariationfunctions ...... 181 5.2 Stochastic integration for semimartingales ...... 189 Integrals with respect to bounded martingales ...... 189 Localization...... 195 Integrationw.r.t.semimartingales ...... 197 5.3 Quadratic variation and covariation ...... 199 Covariationandintegrationbyparts ...... 200 Quadratic variation and covariation of local martingales ...... 201 Covariationofstochasticintegrals ...... 207 The Itô isometry for stochastic integrals w.r.t.martingales ...... 209 5.4 Itô calculus for semimartingales ...... 210 Integrationw.r.t.stochasticintegrals ...... 210 Itô’sformula...... 211 ApplicationtoLévyprocesses ...... 215 Burkholder’sinequality ...... 217 5.5 Stochastic exponentials and change of measure ...... 219 Exponentialsofsemimartingales ...... 219 ChangeofmeasureforPoissonpointprocesses ...... 222

University of Bonn Summer Semester 2015 6 CONTENTS

ChangeofmeasureforLévyprocesses ...... 225 Changeofmeasureforgeneralsemimartingales ...... 227 5.6 General predictable integrands ...... 228 Deﬁnition of stochastic integrals w.r.t. semimartingales ...... 230 Localization...... 231 Propertiesofthestochasticintegral...... 233

Bibliography 236

Stochastic Analysis Andreas Eberle Chapter 1

Lévy processes and Poisson point processes

Awidelyusedclassofpossiblediscontinuousdrivingprocesses in stochastic differential equations are Lévy processes. They include Brownian motion, Poisson and compound Poisson processes as special cases. In this chapter, weoutlinebasicsfromthe theory of Lévy processes, focusing on prototypical examplesofLévyprocessesand their construction. For more details we refer to the monographs of Applebaum [5] and Bertoin [8]. Apart from simple transformations of Brownian motion, Lévy processes do not have continuous paths. Instead, we will assume that the paths are càdlàg (continue à droite, limites à gauche),i.e.,rightcontinuouswithleftlimits.Thiscanalwaysbeassured by choosing an appropriate modiﬁcation. We now summarize a few notations and facts about càdlàg functions that are frequently used below. If x : I R is a càdlàg function → deﬁned on a real interval I,ands is a point in I except the left boundary point, then we denote by

xs = lim xs ε − ε 0 − ↓ the left limit of x at s,andby

∆xs = xs xs − − 7 8CHAPTER1.LÉVYPROCESSESANDPOISSONPOINTPROCESSES

the size of the jump at s.Notethatthefunctions xs is left continuous with right #→ − limits. Moreover, x is continuous if and only if ∆x =0for all s.Let (I) denote the s D linear space of all càdlàg functions x : I R. → Exercise (Càdlàg functions). Prove the following statements: 1) If I is a compact interval, then for any function x (I),theset ∈D s I : ∆x >ε { ∈ | s| } is ﬁnite for any ε>0.Concludethatanyfunctionx ([0, )) has at most ∈D ∞ countably many jumps. 2) A càdlàg function deﬁned on a compact interval is bounded. 3) A uniform limit of a sequence of càdlàg functions is again càdlàg .

1.1 Lévy processes

Lévy processes are Rd-valued stochastic processes with stationary and independent increments. More generally, let ( t)t 0 be a filtration on a probability space (Ω, ,P). F ≥ A Definition. An ( ) Lévy process is an ( ) adapted càdlàg stochastic process Ft Ft X :Ω Rd such that w.r.t. P , t → (a) X X is independent of for any s, t 0,and s+t − s Fs ≥ (b) X X X X for any s, t 0. s+t − s ∼ t − 0 ≥ Any Lévy process (X ) is also a Lévy process w.r.t. the filtration ( X ) generated by the t Ft process. Often continuity in probability is assumed insteadofcàdlàgsamplepaths.It can then be proven that a càdlàg modification exists, cf. [36, Ch.I Thm.30]. Remark (Lévy processes in discrete time are Random Walks). Adiscrete-time process (Xn)n=0,1,2,... with stationary and independent increments is a Random Walk: n Xn = X0 + ηj with i.i.d. increments ηj = Xj Xj 1. j=1 − − Remark (Lévy! processes and infinite divisibility). The increments X X of a s+t − s Lévy process are infinitely divisible random variables, i.e., for any n N there ex- ∈ ist i.i.d. random variables Y ,...,Y such that X X has the same distribution as 1 n s+t − s Stochastic Analysis Andreas Eberle 1.1. LÉVY PROCESSES 9

n Yi.Indeed,wecansimplychooseYi = Xs+it/n Xs+i(t 1)/n.TheLévy-Khinchin i=1 − − formula! gives a characterization of all distributions of inﬁnitely divisible random variables, cf. e.g. [5]. The simplest examples of inﬁnitely divisible distributions are normal and Poisson distributions.

Characteristic exponents

We now restrict ourselves w.l.o.g. to Lévy processes with X0 =0.Thedistributionof the sample paths is then uniquely determined by the distributions of the increments X t − X = X for t 0.Moreover,bystationarityandindependenceoftheincrements we 0 t ≥ obtain the following representation for the characteristicfunctionsϕ (p)=E[exp(ip t · Xt)]:

Theorem 1.1 (Characteristic exponent). If (Xt)t 0 is a Lévy process with X0 =0 ≥ then there exists a continuous function ψ : Rd C with ψ(0) = 0 such that →

ip Xt tψ(p) d E[e · ]=e− for any t 0 and p R . (1.1) ≥ ∈ Moreover, if (X ) has ﬁnite ﬁrst or second moments, then ψ is 1, 2 respectively, and t C C 2 k l ∂ ψ E[Xt]=it ψ(0) , Cov[Xt ,Xt]=t (0) (1.2) ∇ ∂pk∂pl for any k, l =1,...,dand t 0. ≥ Proof. Stationarity and independence of the increments implies theidentity

ϕ (p)=E[exp(ip X )] = E[exp(ip X )] E[exp(ip (X X ))] t+s · t+s · s · · t+s − s = ϕ (p) ϕ (p) (1.3) t · s for any p Rd and s, t 0.Foragivenp Rd,rightcontinuityofthepathsand ∈ ≥ ∈ dominated convergence imply that t ϕ (p) is right-continuous. Since #→ t

ϕt ε(p)=E[exp(ip (Xt Xε))], − · − the function t ϕ (p) is also left continuous, and hence continuous. By (1.3) and since #→ t ϕ (p)=1,wecannowconcludethatforeachp Rd,thereexistsψ(p) C such that 0 ∈ ∈ University of Bonn Summer Semester 2015 10 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

(1.1) holds. Arguing by contradiction we then see that ψ(0) = 0 and ψ is continuous, since otherwise ϕt would not be continuous for all t. Moreover, if X is (square) integrable then ϕ is 1 (resp. 2), and hence ψ is also t t C C 1 (resp. 2). The formulae in (1.2) for the ﬁrst and second moment now follow by C C computing the derivatives w.r.t. p at p =0in (1.1).

The function ψ is called the characteristic exponent of the Lévy process.

Basic examples

We now consider ﬁrst examples of continuous and discontinuous Lévy processes.

Example (Brownian motion and Gaussian Lévy processes). A d-dimensional Brow- nian motion (Bt) is by deﬁnition a continuous Lévy process with

B B N(0, (t s)I ) for any 0 s

Moreover, Xt = σBt + bt is a Lévy process with normally distributed marginals for d d d any σ R × and b R .NotethattheseLévyprocessesarepreciselythedriving ∈ ∈ processes in SDE considered so far. The characteristic exponent of a Gaussian Lévy process is given by 1 1 ψ(p)= σT p 2 ib p = p ap ib p with a = σσT . 2| | − · 2 · − · First examples of discontinuous Lévy processes are Poisson and, more generally, compound Poisson processes.

Example (Poisson processes). The most elementary example of a pure jump Lévy process in continuous time is the Poisson process. It takes values in 0, 1, 2,... and { } jumps up one unit each time after an exponentially distributed waiting time. Explicitly, aPoissonprocess(Nt)t 0 with intensity λ>0 is given by ≥

∞ N Nt = I Sn t = ♯ n : Sn t (1.4) { ≤ } { ∈ ≤ } n=1 " where S = T + T + + T with independent random variables T Exp(λ).The n 1 2 ··· n i ∼ increments N N of a Poisson process over disjoint time intervals are independent t − s Stochastic Analysis Andreas Eberle 1.1. LÉVY PROCESSES 11 and Poisson distributed with parameter λ(t s),cf.[13,Satz10.12].Notethatby(1.4), − the sample paths t N (ω) are càdlàg. In general, any Lévy process with #→ t X X Poisson (λ(t s)) for any 0 s t t − s ∼ − ≤ ≤ is called a Poisson process with intensity λ,andcanberepresentedasabove.The characteristic exponent of a Poisson process with intensity λ is

ψ(p)=λ(1 eip). − The paths of a Poisson process are increasing and hence of ﬁnite variation. The compensated Poisson process

M := N E[N ]=N λt t t − t t − is an ( N ) martingale, yielding the semimartingale decomposition Ft

Nt = Mt + λt with the continuous ﬁnite variation part λt.Ontheotherhand,thereisthealternative trivial semimartingale decomposition Nt =0+Nt with vanishing martingale part. This demonstrates that without an additional regularity condition, the semimartingale decomposition of discontinuous processes is not unique. A compensated Poisson process is a Lévy process which has both a continuous and a pure jump part.

Exercise (Martingales of Poisson processes). Prove that the compensated Poisson process M = N λt and the process M 2 λt are ( N ) martingales. t t − t − Ft Any linear combination of independent Lévy processes is again a Lévy process:

Example (Superpositions of Lévy processes). If (Xt) and (Xt′) are independent Lévy Rd Rd′ processes with values in and then αXt + βXt′ is a Lévy process with values in n n d n d′ R for any constant matrices α R × and β R × .Thecharacteristicexponentof ∈ ∈ the superposition is

T T ψαX+βX′ (p)=ψX (α p)+ψY (β p).

For example, linear combinations of independent Brownian motions and Poisson processes are again Lévy processes.

University of Bonn Summer Semester 2015 12 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

Compound Poisson processes

Next we consider general Lévy processes with paths that are constant apart from a finite number of jumps in finite time. We will see that such processes can be represented as compound Poisson processes. A compound Poisson process isacontinuoustime Random Walk defined by

Nt X = η ,t 0, t j ≥ j=1 " with a Poisson process (Nt) of intensity λ>0 and with independent identically distributed random variables η :Ω Rd (j N) that are independent of the Poisson j → ∈ process as well. The process (Xt) is again a pure jump process with jump times that do not accumulate. It has jumps of size y with intensity

ν(dy)=λπ(dy),

where π denotes the joint distribution of the random variables ηj.

Lemma 1.2. AcompoundPoissonprocessisaLévyprocesswithcharacteristic exponent ip y ψ(p)= (1 e · ) ν(dy). (1.5) ˆ −

Proof. Let 0=t

Ntk X X = η ,k=1, 2,...,n, (1.6) tk − tk−1 j j=N +1 "tk−1 are conditionally independent given the σ-algebra generated by the Poisson process d (Nt)t 0.Therefore,forp1,...,pn R , ≥ ∈ n n E exp i p (X X (N ) = E[exp(ip (X X ) (N )] k · tk − tk−1 t k · tk − tk−1 | t k=1 k=1 # $ " % & ' ( & n Nt Nt − = ϕ(pk) k − k 1 , k(=1

Stochastic Analysis Andreas Eberle 1.1. LÉVY PROCESSES 13

where ϕ denotes the characteristic function of the jump sizes ηj.Bytakingtheexpec- tation on both sides, we see that the increments in (1.6) are independent and stationary, since the same holds for the Poisson process (Nt).Moreover,byasimilarcomputation,

E[exp(ip X )] = E[E[exp(ip X ) (N )]] = E[ϕ(p)Nt ] · t · t | s ∞ k λt (λt) k λt(ϕ(p) 1) = e− ϕ(p) = e − k! "k=0 for any p Rd,whichproves(1.5). ∈ The paths of a compound Poisson process are of finite variationandcàdlàg.Onecan show that every pure jump Lévy process with finitely many jumpsinfinitetimeisa compound Poisson process , cf. Theorem 1.15 below.

Exercise (Martingales of compound Poisson processes). Show that the following processes are martingales:

1 (a) Mt = Xt bt where b = yν(dy) provided η1 , − ´ ∈L 2 2 2 (b) Mt at where a = y ν(dy) provided η1 . | | − ´ | | ∈L We have shown that a compound Poisson process with jump intensity measure ν(dy) is aLévyprocesswithcharacteristicexponent

ip y d ψ (p)= (1 e · )ν(dy) ,pR . (1.7) ν ˆ − ∈ Since the distribution of a Lévy process on the space ([0, ), Rd) of càdlàg paths is D ∞ uniquely determined by its characteristic exponent, we can prove conversely:

Lemma 1.3. Suppose that ν is a ﬁnite positive measure on Rd 0 with total mass B \{ } λ = ν(Rd 0 ),and(X ) is a Lévy process with X =0and characteristic exponent \{ } t 0 $ % ψ ,deﬁnedonacompleteprobabilityspace(Ω, ,P).Thenthereexistsasequence ν A 1 (ηj)j N of i.i.d. random variables with distribution λ− ν and an independent Poisson ∈ Process (N ) with intensity λ on (Ω, ,P) such that almost surely, t A Nt

Xt = ηj . (1.8) j=1 "

University of Bonn Summer Semester 2015 14 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

Proof. Let (ηj) be an arbitrary sequence of i.i.d. random variables with distribution 1 d λ− ν,andlet(Nt) be an independent Poisson process of intensity ν(R 0 ),all ) \{ } defined on a probability space (Ω, , P ).ThenthecompoundPoissonprocessXt = ! A Nt * j=1 ηj is also a Lévy process with X0 =0and characteristic exponent ψν .Therefore, ) ) ) * the! finite dimensional marginals of (Xt) and (Xt),andhencethedistributionsof(Xt) ) d * and (Xt) on ([0, ), R ) coincide. In particular, almost every path t Xt(ω) has D ∞ * #→ only finitely many jumps in a finite time interval, and is constant inbetween. Now set * S0 =0and let

Sj = inf s>Sj 1 :∆Xs =0 for j N { − ̸ } ∈ denote the successive jump-times of (Xt).Then(Sj) is a sequence of non-negative random variables on (Ω, ,P) that is almost surely finite and strictly increasing with A lim S = .Definingη := ∆X if S < , η =0otherwise, and j ∞ j Sj j ∞ j N := s (0,t]:∆X =0 = j N : S t , t { ∈ s ̸ } { ∈ j ≤ } & & & & as the successive jump& sizes and the number of& jumps up& to time t,weconcludethat& almost surely, (N ) is finite, and the representation (1.8) holds. Moreover, for any j N t ∈ and t 0 , ηj and Nt are measurable functions of the process (Xt)t 0.Hencethejoint ≥ ≥ distribution of all these random variables coincides with the joint distribution of the random variables η (j N) and N (t 0),whicharethecorrespondingmeasurable j ∈ t ≥ functions of the process (Xt).Wecanthereforeconcludethat(ηj)j N is a sequence ∈ ) * 1 of i.i.d. random variables with distributions λ− ν and (Nt) is an independent Poisson * process with intensity λ.

The lemma motivates the following formal deﬁnition of a compound Poisson process:

Definition. Let ν be a finite positive measure on Rd,andletψ : Rd C be the ν → function defined by (1.7).

1) The unique probability measure π on (Rd) with characteristic function ν B

ip y d e · π (dy)=exp(ψ (p)) p R ˆ ν − ν ∀ ∈ is called the compound Poisson distribution with intensity measure ν.

Stochastic Analysis Andreas Eberle 1.1. LÉVY PROCESSES 15

2) A Lévy process (X ) on Rd with X X π for any s, t 0 is called a t s+t − s ∼ tν ≥ compound Poisson process with jump intensity measure (Lévy measure) ν.

K The compound Poisson distribution πν is the distribution of j=1 ηj where K is a Pois- Rd son random variable with parameter λ = ν( ) and (ηj) is! a sequence of i.i.d. random 1 variables with distribution λ− ν.ByconditioningonthevalueofK ,weobtainthe explicit series representation

∞ k λ λ k π = e− ν∗ , ν k! "k=0 k where ν∗ denotes the k-fold convolution of ν.

Examples with inﬁnite jump intensity

The Lévy processes considered so far have only a finite number of jumps in a finite time interval. However, by considering limits of Lévy processes with finite jump intensity, one also obtains Lévy processes that have infinitely many jumps in a finite time interval. We first consider two important classes of examples of such processes:

Example (Inverse Gaussian subordinators). Let (Bt)t 0 be a one-dimensional Brow- ≥ nian motion with B =0w.r.t. a right continuous filtration ( ),andlet 0 Ft T = inf t 0:B = s s { ≥ t } denote the first passage time to a level s R.Then(Ts)s 0 is an increasing stochastic ∈ ≥ process that is adapted w.r.t. the filtration ( Ts )s 0.Foranyω, s Ts(ω) is the gener- F ≥ #→ alized left-continuous inverse of the Brownian path t B (ω).Moreover,bythestrong #→ t Markov property, the process

B(s) := B B ,t 0, t Ts+t − Ts ≥ is a Brownian motion independent) of for any s 0,and FTs ≥ T = T + T (s) for s, u 0, (1.9) s+u s u ≥ (s) (s) (s) where Tu = inf t 0:B = u is the) ﬁrst passage time to u for the process B . ≥ t + , University) of Bonn ) Summer Semester 2015) 16 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

(s) Bt Bt

s + u ) s

Ts Ts+u

By (1.9), the increment T T is independent of ,and,bythereﬂectionprinciple, s+u − s FTs 2 u 3/2 u Ts+u Ts Tu x− exp I(0, )(x) dx. − ∼ ∼ √ π −2x ∞ 2 - .

Hence (Ts) is an increasing process with stationary and independent increments. The process (Ts) is left-continuous, but it is not difﬁcult to verify that

(s) Ts+ = lim Ts+ε = inf t 0:Bt >u ε 0 ≥ ↓ + , ) is a càdlàg modiﬁcation, and hence a Lévy process. (Ts+) is called “The Lévy subordinator”,where“subordinator”standsforanincreasingLévyprocess. We will see below that subordinators are used for random time transformations (“subordination”) of other Lévy processes.

More generally, if Xt = σBt + bt is a Gaussian Lévy process with coefﬁcients σ>0, b R,thentherightinverse ∈ T X = inf t 0:X = s ,s 0, s { ≥ t } ≥ is called an Inverse Gaussian subordinator.

Exercise (Sample paths of Inverse Gaussian processes). Prove that the process (Ts)s 0 ≥ is increasing and purely discontinuous,i.e.,withprobabilityone,(Ts) is not continuous on any non-empty open interval (a, b) [0, ). ⊂ ∞

Stochastic Analysis Andreas Eberle 1.1. LÉVY PROCESSES 17

Example (Stable processes). Stable processes are Lévy processes that appear as scaling n Rd limits of Random Walks. Suppose that Sn = j=1 ηj is a Random Walk in with i.i.d. increments ηj.Iftherandomvariablesηj are! square-integrable with mean zero then Donsker’s invariance principle (the “functional central limit theorem”) states that the 1/2 diffusively rescaled process (k− S kt )t 0 converges in distribution to (σBt)t 0 where ⌊ ⌋ ≥ ≥ (B ) is a Brownian motion in Rd and σ is a non-negative deﬁnite symmetric d d t × matrix. However, the functional central limit theorem does not apply if the increments

ηj are not square integrable (“heavy tails”). In this case, one considers limits of rescaled (k) 1/α Random Walks of the form Xt = k− S kt where α (0, 2] is a ﬁxed constant. It is ⌊ ⌋ (k) ∈ not difﬁcult to verify that if (Xt ) converges in distribution to a limit process (Xt) then

(Xt) is a Lévy process that is invariant under the rescaling, i.e.,

1/α k− X X for any k (0, ) and t 0. (1.10) kt ∼ t ∈ ∞ ≥

Deﬁnition. Let α (0, 2].ALévyprocess(X ) satisfying (1.10) is called (strictly) ∈ t α-stable.

The reason for the restriction to α (0, 2] is that for α>2,anα-stable process ∈ does not exist. This will become clear by the proof of Theorem 1.4 below. There is abroaderclassofLévyprocessesthatiscalledα-stable in the literature, cf. e.g. [28]. Throughout these notes, by an α-stable process we always mean a strictly α-stable process as deﬁned above.

For b R,thedeterministicprocessX = bt is a 1-stable Lévy process. Moreover, ∈ t 1 aLévyprocessX in R is 2-stable if and only if Xt = σBt for a Brownian motion (B ) and a constant σ [0, ).Characteristicexponentscanbeappliedtoclassifyall t ∈ ∞ α-stable processes:

Theorem 1.4 (Characterization of stable processes). For α (0, 2] and a Lévy pro- ∈ 1 cess (Xt) in R with X0 =0the following statements are equivalent:

(i) (Xt) is strictly α-stable.

(ii) ψ(cp)=cαψ(p) for any c 0 and p R. ≥ ∈ University of Bonn Summer Semester 2015 18 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

(iii) There exists constants σ 0 and µ R such that ≥ ∈

ψ(p)=σα p α(1 + iµ sgn(p)). | |

Proof. (i) (ii).Theprocess(X ) is strictly α-stable if and only if X α cX for ⇔ t c t ∼ t any c, t 0,i.e.,ifandonlyif ≥

α tψ(cp) ipcXt ipX α c tψ(p) e− = E e = E e c t = e− / 0 / 0 for any c, t 0 and p R. ≥ ∈ (ii) (iii).Clearly,Condition(ii) holds if and only if there exist complex numbers ⇔ z+ and z such that − α z+ p for p 0, ψ(p)= | | ≥ ⎧z p α for p 0. ⎨ −| | ≤ Moreover, since ϕ (p)=exp( tψ(p))⎩ is a characteristic function of a probability t − measure for any t 0,thecharacteristicexponentψ satisﬁes ψ( p)=ψ(p) and ≥ − (ψ(p)) 0.Therefore,z = z+ and (z+) 0. ℜ ≥ − ℜ ≥

Example (Symmetric α-stable processes). ALévyprocessinRd with characteristic exponent ψ(p)=σα p α | | for some σ 0 and a (0, 2] is called a symmetric α-stable process.Wewillseebelow ≥ ∈ that a symmetric α-stable process is a Markov process with generator σα( ∆)α/2.In − − particular, Brownian motion is a symmetric 2-stable process.

1.2 Martingales and Markov property

For Lévy processes, one can identify similar fundamental martingales as for Brownian motion. Furthermore, every Lévy process is a strong Markov process.

Stochastic Analysis Andreas Eberle 1.2. MARTINGALES AND MARKOV PROPERTY 19

Martingales of Lévy processes

The notion of a martingale immediately extends to complex or vector valued processes by a componentwise interpretation. As a consequence of Theorem 1.1 we obtain:

Corollary 1.5. If (Xt) is a Lévy process with X0 =0and characteristic exponent ψ, then the following processes are martingales:

(i) exp(ip X + tψ(p)) for any p Rd, · t ∈ (ii) M = X bt with b = i ψ(0), provided X 1 t 0. t t − ∇ t ∈L ∀ ≥ j k jk jk ∂2ψ 2 (iii) Mt M a t with a = (0) (j, k =1,...,d),providedXt t − ∂pj ∂pk ∈L t 0. ∀ ≥ Proof. We only prove (ii) and (iii) for d =1and leave the remaining assertions as an exercise to the reader. If d =1and (X ) is integrable then for 0 s t, t ≤ ≤

E[X X ]=E[X X ]=i(t s)ψ′(0) t − s |Fs t − s − by independence and stationarity of the increments and by (1.2). Hence M = X t t − itψ′(0) is a martingale. Furthermore,

M 2 M 2 =(M + M )(M M )=2M (M M )+(M M )2. t − s t s t − s s t − s t − s

If (Xt) is square integrable then the same holds for (Mt),andweobtain

E[M 2 M 2 ]=E[(M M )2 ]=Var[M M ] t − s |Fs t − s |Fs t − s |Fs =Var[Xt Xs s]=Var[Xt Xs]=Var[Xt s]=(t s)ψ′′(0) − |F − − − 2 Hence M tψ′′(0) is a martingale. t − Note that Corollary 1.5 (ii) shows that an integrable Lévy process is a semimartingale with martingale part Mt and continuous ﬁnite variation part bt.Theidentity(1.1)canbe used to classify all Lévy processes, c.f. e.g. [5]. In particular, we will prove below that by Corollary 1.5, any continuous Lévy process with X0 =0is of the type Xt = σBt +bt d d d with a d-dimensional Brownian motion (B ) and constants σ R × and b R . t ∈ ∈

University of Bonn Summer Semester 2015 20 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

Lévy processes as Markov processes

The independence and stationarity of the increments of a Lévyprocessimmediately implies the Markov property:

Theorem 1.6 (Markov property). ALévyprocess(Xt,P) is a time-homogeneous Markov process with translation invariant transition functions

p (x, B)=µ (B x)=p (a + x, a + B) a Rd, (1.11) t t − t ∀ ∈ 1 where µ = P (X X )− . t ◦ t − 0 Proof. For any s, t 0 and B (Rd), ≥ ∈B P [X B ](ω)=P [X +(X X ) B ](ω) s+t ∈ |Fs s s+t − s ∈ |Fs = P [X X B X (ω)] s+t − s ∈ − s = P [X X B X (ω)] t − 0 ∈ − s = µ (B X (ω)). t − s

Remark (Feller property). The transition semigroup of a Lévy process has the Feller property,i.e.,iff : Rd R is a continuous function vanishing at infinity then the same → holds for p f for any t 0.Indeed, t ≥ (p f)(x)= f(x + y) µ (dy) t ˆ t is continuous by dominated convergence, and, similarly, (p f)(x) 0 as x . t → | |→∞ Exercise (Strong Markov property for Lévy processes). Let (X ) be an ( ) Lévy t Ft process, and let T be a finite stopping time. Show that Y = X X is a process t T +t − T that is independent of ,andX and Y have the same law. FT n n Hint: Consider the sequence of stopping times defined by T =(k +1)2− if k2− n ≤ n T<(k +1)2− .NoticethatT T as n .Inafirststepshowthatforanym N n ↓ →∞ ∈ and t

E [f(X X ,...,X X )I ]=E [f(X ,...,X )] P [A]. Tn+t1 − Tn Tn+tm − Tn A t1 tm

Stochastic Analysis Andreas Eberle 1.2. MARTINGALES AND MARKOV PROPERTY 21

Exercise (AcharacterizationofPoissonprocesses). Let (Xt)t 0 be a Lévy process ≥ with X0 =0a.s. Suppose that the paths of X are piecewise constant, increasing, all jumps of X are of size 1, and X is not identically 0.ProvethatX is a Poisson process.

Hint: Apply the Strong Markov property to the jump times (Ti)i=1,2,... of X to conclude that the random variables Ui := Ti Ti 1 are i.i.d. (with T0 := 0). Then, it remains to − − show that U1 is an exponential random variable with some parameter λ>0.

The marginals of a Lévy process ((Xt)t 0,P) are completely determined by the char- ≥ acteristic exponent ψ.Inparticular,onecanobtainthetransitionsemigroupandits in- ﬁnitesimal generator from ψ by Fourier inversion. Let (Rd) denote the Schwartz space S d k α consisting of all functions f C∞(R ) such that x ∂ f(x) goes to 0 as x for ∈ | | | |→∞ any k N and derivatives of f of arbitary order α Zd .RecallthattheFourier ∈ ∈ + transform maps (Rd) one-to-one onto (Rd). S S

Corollary 1.7 (Transition semigroup and generator of a Lévy process).

(1). For any f (Rd) and t 0, ∈S ≥

tψ ˇ ptf =(e− fˆ)

d/2 ip x d/2 ip x where fˆ(p)=(2π)− e− · f(x) dx and gˇ(x)=(2π)− e · g(p) dp denote the Fourier transform´ and the inverse Fourier transform of functions´ f,g ∈ 1(Rd). L

(2). The Schwartz space (Rd) is contained in the domain of the generator L of the S d Feller semigroup induced by (pt)t 0 on the Banach space Cˆ(R ) of continuous ≥ functions vanishing at inﬁnity, and the generator is the pseudo-differential operator given by

Lf =( ψfˆ)ˇ. (1.12) −

University of Bonn Summer Semester 2015 22 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

Proof. (1). Since (ptf)(x)=E[f(Xt + x)],weconcludebyFubinithat

d ip x (pˆf)(p)=(2π)− 2 e− · (p f)(x) dx t ˆ t

d ip x =(2π)− 2 E e− · f(X + x) dx · ˆ t 4 5 ip Xt = E e · fˆ(p) · tψ(p) = e−# fˆ('p)

for any p Rd.TheclaimfollowsbytheFourierinversiontheorem,notingthat ∈ tψ e− 1. ≤ & & (2).& For f& (Rd), fˆ is in (Rd) as well. The Lévy-Khinchin formula that we will ∈S S state below gives an explicit representation of all possibleLévyexponentswhich shows in particular that ψ(p) is growing at most polynomially as p .Since | |→∞ tψ tψ e− fˆ fˆ e− 1 − + ψfˆ = − + ψ fˆ , and & t & t ·| | & & & & & & & & & & & & & t & & & t s tψ e− 1 1 sψ 1 2 rψ − + ψ = ψ e− 1 ds = ψ e− dr ds, t − t ˆ − t ˆ ˆ 0 $ % 0 0 we obtain

tψ e− fˆ fˆ − + ψfˆ t ψ2 fˆ 1(Rd), & t & ≤ ·| |·| |∈L & & & & and, therefore, & & & & (p f)(x) f(x) t − ( ψfˆ)ˇ(x) t − − d ip x 1 tψ(p) =(2π)− 2 e · e− fˆ(p) fˆ(p) + ψ(p)fˆ(p) dp 0 ˆ t − −→ - 6 7 . as t 0 uniformly in x.Thisshowsf Dom(L) and Lf =( ψfˆ)ˇ. ↓ ∈ −

By the theory of Markov processes, the corollary shows in particular that a Lévy process (X ,P) solves the martingale problem for the operator (L, (Rd)) deﬁned by (5.14). t S Stochastic Analysis Andreas Eberle 1.3. POISSON RANDOM MEASURES AND POISSON POINT PROCESSES 23

Examples. 1) For a Gaussian Lévy processes as considered above, ψ(p)= 1 p ap ib p 2 · − · where a := σσT .Hencethegeneratorisgivenby

1 Lf = (ψfˆ)ˇ = (a f) b f, for f (Rn). − 2∇· ∇ − ·∇ ∈S

2) For a Poisson process (N ), ψ(p)=λ(1 eip) implies t −

(Lf)(x)=λ(f(x +1) f(x)). −

3) For the compensated Poisson process M = N λt, t t −

(Lf)(x)=λ(f(x +1) f(x) f ′(x)). − −

4) For a symmetric α-stable process with characteristic exponent ψ(p)=σα p α for ·| | some σ 0 and α (0, 2],thegeneratorisafractionalpoweroftheLaplacian: ≥ ∈ Lf = (ψfˆ)ˇ = σα ( ∆)α/2 f. − − −

We remark that for α>2,theoperatorL does not satisfy the positive maximum principle. Therefore, in this case L does not generate a transition semigroup of a Markov process.

1.3 Poisson random measures and Poisson point processes

AcompensatedPoissonprocesshasonlyfinitelymanyjumpsinafinitetimeinterval. General Lévy jump processes may have a countably infinite number of (small) jumps in finite time. In the next section, we will construct such processes from their jumps. As apreparationwewillnowstudyPoissonrandommeasuresandPoisson point processes that encode the jumps of Lévy processes. The jump part of a Lévyprocesscanbe recovered from these counting measure valued processes by integration, i.e., summation of the jump sizes. We start with the observation that the jump times of a Poisson process form a Poisson random measure on R+.

University of Bonn Summer Semester 2015 24 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

The jump times of a Poisson process

For a different point of view on Poisson processes let

+(S)= δ :(y ) finite or countable sequence in S Mc yi i +" , denote the set of all counting measures on a set S.APoissonprocess(Nt)t 0 can be ≥ viewed as the distribution function of a random counting measure, i.e., of a random variable N :Ω +([0, )). →Mc ∞ Definition. Let ν be a σ-finite measure on a measurable space (S, ).Acollectionof S random variables N(B), B ,onaprobabilityspace(Ω, ,P) is called a Poisson ∈S A random measure (or spatial Poisson process) of intensity ν if and only if

(i) B N(B)(ω) is a counting measure for any ω Ω, #→ ∈ (ii) if B ,...,B are disjoint then the random variables N(B ),...,N(B ) are 1 n ∈S 1 n independent,

(iii) N(B) is Poisson distributed with parameter ν(B) for any B with ν(B) < . ∈S ∞ APoissonrandommeasureN with ﬁnite intensity ν can be constructed as the empirical measure of a Poisson distributed number of independent samples from the normalized measure ν/ν(S):.

K N = δ with X ν/ν(s) i.i.d.,K Poisson(ν(S)) independent. Xj j ∼ ∼ j=1 " If the intensity measure ν does not have atoms then almost surely, N( x ) 0, 1 for { } ∈{ } any x S,andN = x A δx for a random subset A of S.Forthisreason,aPoisson ∈ ∈ random measure is often! called a Poisson point process, but wewillusethisterminology differently below.

Areal-valuedprocess(Nt)t 0 is a Poisson process of intensity λ>0 if and only if ≥ t N (ω) is the distribution function of a Poisson random measure N(dt)(ω) on #→ t ([0, )) with intensity measure ν(dt)=λdt.ThePoissonrandommeasureN(dt) B ∞ can be interpreted as the derivative of the Poisson process:

N(dt)= δs(dt).

s:∆Ns=0 "̸ Stochastic Analysis Andreas Eberle 1.3. POISSON RANDOM MEASURES AND POISSON POINT PROCESSES 25

In a stochastic differential equation of type dYt = σ(Yt ) dNt, N(dt) is the driving − Poisson noise.

The following assertion about Poisson processes is intuitively clear from the interpretation of a Poisson process as the distribution function of a Poisson random measure. Compound Poisson processes enable us to give a simple proof ofthesecondpartofthe theorem:

Theorem 1.8 (Superpositions and subdivisions of Poisson processes). Let K be a countable set.

(k) 1) Suppose that (Nt )t 0, k K,areindependentPoissonprocesseswithintensi- ≥ ∈ ties λk.Then N = N (k) ,t 0, t t ≥ k K "∈ is a Poisson process with intensity λ = λ provided λ< . k ∞ ! 2) Conversely, if (Nt)t 0 is a Poisson process with intensity λ>0,and(Cn)n N is ≥ ∈ asequenceofi.i.d.randomvariablesC :Ω K that is also independent of n #→ (Nt),thentheprocesses

Nt (k) Nt = I Cj =k ,t 0, { } ≥ j=1 "

are independent Poisson processes of intensities qkλ,whereqk = P [C1 = k].

The subdivision in the second assertion can be thought of as colouring the points in the support of the corresponding Poisson random measure N(dt) independently with (k) random colours Cj,anddecomposingthemeasureintopartsN (dt) of equal colour.

Proof. The first part is rather straightforward, and left as an exercise. For the second part, we may assume w.l.o.g. that K is finite. Then the process N⃗ :Ω RK defined t → by Nt ⃗ (k) Nt := Nt = ηj with ηj = I k (Cj) k K { } k K ∈ j=1 ∈ 6 7 " $ % University of Bonn Summer Semester 2015 26 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES is a compound Poisson process on RK,andhenceaLévyprocess.Moreover,bythe proof of Lemma 1.2, the characteristic function of N⃗ for t 0 is given by t ≥

E exp ip N⃗ =exp(λt(ϕ(p) 1)) ,pRK , · t − ∈ / 6 70 where

ipk ϕ(p)=E [exp(ip η1)] = E exp i pkI k (C1) = qke . · { } 8 9 k K :; k K "∈ "∈

Noting that qk =1,weobtain ! E[exp(ip N⃗ )] = exp(λtq (eipk 1)) for any p RK and t 0. · t k − ∈ ≥ k K (∈ The assertion follows, because the right hand side is the characteristic function of a Lévy process in RK whose components are independent Poisson processes with intensities qkλ.

The jumps of a Lévy process

We now turn to general Lévy processes. Note ﬁrst that a Lévy process (Xt) has only countably many jumps, because the paths are càdlàg. The jumpscanbeencodedinthe counting measure-valued stochastic process N :Ω +(Rd 0 ), t →Mc \{ }

N (dy)= δ (dy),t 0, t ∆Xs ≥ s t ∆"X≤s=0 ̸ or, equivalently, in the random counting measure N :Ω + R (Rd 0 ) →Mc + × \{ } deﬁned by $ %

N(dt dy)= δ(s,∆Xs)(dt dy). s t ∆"X≤s=0 ̸

Stochastic Analysis Andreas Eberle 1.3. POISSON RANDOM MEASURES AND POISSON POINT PROCESSES 27

d R Xt

∆Xt

The process (Nt)t 0 is increasing and adds a Dirac mass at y each time the Lévy pro- ≥ cess has a jump of size y.Since(Xt) is a Lévy process, (Nt) also has stationary and independent increments:

N (B) N (B) N (B) for any s, t 0 and B (Rd 0 ). s+t − s ∼ t ≥ ∈B \{ } Hence for any set B with N (B) < a.s. for all t,theintegervaluedstochasticprocess t ∞ (Nt(B))t 0 is a Lévy process with jumps of size +1.ByanexerciseinSection1.1,we ≥ can conclude that (N (B)) is a Poisson process. In particular, t E[N (B)] is a linear t #→ t function.

University of Bonn Summer Semester 2015 28 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

Definition. The jump intensity measure of a Lévy process (Xt) is the σ-finite measure ν on the Borel σ-algebra (Rd 0 ) determined by B \{ } E[N (B)] = t ν(B) t 0,B (Rd 0 ). (1.13) t · ∀ ≥ ∈B \{ } It is elementary to verify that for any Lévy process, there is auniquemeasureν satisfying (1.13). Moreover, since the paths of a Lévy process are càdlàg, the measures Nt and ν are finite on y Rd : y ε for any ε>0. { ∈ | |≥ } Example (Jump intensity of stable processes). The jump intensity measure of strictly α-stable processes in R1 can be easily found by an informal argument. Suppose we rescale in space and time by y cy and t cαt.Ifthejumpintensityisν(dy)= → → f(y) dy,thenafterrescalingwewouldexpectthejumpintensitycαf(cy)cdy.Ifscale 1 α invariance holds then both measures should agree, i.e., f(y) y − − both for y>0 ∝| | and for y<0 respectively. Therefore, the jump intensity measure of a strictly α-stable process on R1 should be given by

1 α ν(dy)= c+I(0, )(y)+c I( ,0)(y) y − − dy (1.14) ∞ − −∞ | | $ % with constants c+,c [0, ). − ∈ ∞

If (Xt) is a pure jump process with finite jump intensity measure (i.e., finitely many jumps in a finite time interval) then it can be recovered from (Nt) by adding up the jump sizes: X X = ∆X = yN(dy). t − 0 s ˆ t s t "≤ In the next section, we are conversely going to construct moregeneralLévyjumppro- cesses from the measure-valued processes encoding the jumps. As a first step, we are going to define formally the counting-measure valued processes that we are interested in.

Poisson point processes

Let (S, ,ν) be a σ-ﬁnite measure space. S Stochastic Analysis Andreas Eberle 1.3. POISSON RANDOM MEASURES AND POISSON POINT PROCESSES 29

Deﬁnition. AcollectionN (B), t 0, B ,ofrandomvariablesonaprobability t ≥ ∈S space (Ω, ,P) is called a Poisson point process of intensity ν if and only if A (i) B N (B)(ω) is a counting measure on S for any t 0 and ω Ω, #→ t ≥ ∈

(ii) if B1,...,Bn are disjoint then (Nt(B1))t 0,...,(Nt(Bn))t 0 are indepen- ∈S ≥ ≥ dent stochastic processes and

(iii) (Nt(B))t 0 is a Poisson process of intensity ν(B) for any B with ν(B) < . ≥ ∈S ∞ APoissonpointprocessaddsrandompointswithintensityν(dt) dy in each time instant dt.ItisthedistributionfunctionofaPoissonrandommeasureN(dt dy) on R+ ×S with intensity measure dt ν(dy),i.e.

N (B)=N((0,t] B) for any t 0 and B . t × ≥ ∈S

B Nt(B)

The distribution of a Poisson point process is uniquely determined by its intensity measure: If (Nt) and (Nt) are Poisson point processes with intensity ν then

(Nt(B*1),...,Nt(Bn))t 0 (Nt(B1),...,(Nt(Bn))t 0 ≥ ∼ ≥ for any ﬁnite collection of disjoint sets B ,...,B ,and,hence,foranyﬁnite 1 * n ∈S * collection of measurable arbitrary sets B ,...,B . 1 n ∈S Applying a measurable map to the points of a Poisson point process yields a new Poisson point process:

University of Bonn Summer Semester 2015 30 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

Exercise (Mapping theorem). Let (S, ) and (T, ) be measurable spaces and let S T f : S T be a measurable function. Prove that if (N ) is a Poisson point process with → t 1 intensity measure ν then the image measures N f − , t 0,formaPoissonpoint t ◦ ≥ 1 process on T with intensity measure ν f − . ◦

An advantage of Poisson point processes over Lévy processes is that the passage from ﬁnite to inﬁnite intensity (of points or jumps respectively)isnotaproblemonthelevel of Poisson point processes because the resulting sums trivially exist by positivity:

Theorem 1.9 (Construction of Poisson point processes).

1) Suppose that ν is a ﬁnite measure with total mass λ = ν(S).Then

Nt = δηj j=1 "

is a Poisson point process of intensity ν provided the random variables ηj are 1 independent with distribution λ− ν,and(Kt) is an independent Poisson process of intensity λ.

2) If (N (k)), k N,areindependentPoissonpointprocesseson(S, ) with intensity t ∈ S measures νk then

∞ (k) Nt = Nt "k=1

is a Poisson point process with intensity measure ν = νk. ! The statements of the theorem are consequences of the subdivision and superposition properties of Poisson processes. The proof is left as an exercise.

Conversely, one can show that any Poisson point process with ﬁnite intensity measure ν can be almost surely represented as in the ﬁrst part of Theorem1.9,whereKt = Nt(S). The proof uses uniqueness in law of the Poisson point process,andissimilartothe proof of Lemma 1.3.

Stochastic Analysis Andreas Eberle 1.3. POISSON RANDOM MEASURES AND POISSON POINT PROCESSES 31

Construction of compound Poisson processes from PPP

We are going to construct Lévy jump processes from Poisson point processes. Suppose first that (N ) is a Poisson point process on Rd 0 with finite intensity measure ν. t \{ } Then the support of N is almost surely finite for any t 0.Therefore,wecandefine t ≥

Xt = yNt(dy)= yNt( y ), ˆRd 0 { } y supp(Nt) \{ } ∈ " d Theorem 1.10. If ν(R 0 ) < then (Xt)t 0 is a compound Poisson process with \{ } ∞ ≥ jump intensity ν.Moregenerally,foranyPoissonpointprocesswithﬁniteintensity measure ν on a measurable space (S, ) and for any measurable function f : S Rn, S → n N,theprocess ∈ N (f) := f(y)N (dy) ,t 0, t ˆ t ≥

1 is a compound Poisson process with intensity measure ν f − . ◦ Proof. By Theorem 1.9 and by the uniqueness in law of a Poisson point process with

Kt given intensity measure, we can represent (Nt) almost surely as Nt = j=1 δηj with i.i.d. random variables η ν/ν(S) and an independent Poisson process (K ) of inten- j ∼ ! t sity ν(S).Thus,

Kt N (f)= f(y)N (dy)= f(η ) almost surely. t ˆ t j j=1 " Since the random variables f(η ), j N,arei.i.d.andindependentof(K ) with distri- j ∈ t 1 bution ν f − , (N (f)) is a compound Poisson process with this intensity measure. ◦ t As a direct consequence of the theorem and the properties of compound Poisson processes derived above, we obtain:

Corollary 1.11 (Martingales of Poisson point processes). Suppose that (Nt) is a Pois- son point process with ﬁnite intensity measure ν.Thenthefollowingprocessesaremar- tingales w.r.t. the ﬁltration N = σ(N (B) 0 s t, B ): Ft s | ≤ ≤ ∈S 1 (i) Nt(f)=Nt(f) t fdν for any f (ν), − ´ ∈L University* of Bonn Summer Semester 2015 32 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

2 (ii) Nt(f)Nt(g) t fg dν for any f,g (ν), − ´ ∈L ipf (iii) exp* (ipN* t(f)+t (1 e ) dν) for any measurable f : S R and p R. ´ − → ∈ Proof. If f is in p(ν) for p =1, 2 respectively, then L

p 1 p x ν f − (dx)= f(y) ν(dy) < , ˆ | | ◦ ˆ | | ∞

1 1 xν f − (dx)= fdν, and xy ν (fg)− (dxdy)= fg dν. ˆ ◦ ˆ ˆ ◦ ˆ

Therefore (i) and (ii) (and similarly also (iii)) follow fromthecorrespondingstatements for compound Poisson processes.

With a different proof and an additional integrability assumption, the assertion of Corol- lary 1.11 extends to σ-ﬁnite intensity measures:

Exercise (Expectation values and martingales for Poisson point processes with in-

finite intensity). Let (Nt) be a Poisson point process with σ-finite intensity ν. a) By considering first elementary functions, prove that for t 0,theidentity ≥

E f(y)N (dy) = t f(y)ν(dy) ˆ t ˆ 4 5 holds for any measurable function f : S [0, ].Concludethatforf 1(ν), → ∞ ∈L the integral Nt(f)= f(y)Nt(dy) exists almost surely and deﬁnes a random variable in L1(Ω, ,P)´. A b) Proceeding similarly as in a), prove the identities

E[N (f)] = t fdν for any f 1(ν), t ˆ ∈L Cov[N (f),N(g)] = t fg dν for any f,g 1(ν) 2(ν), t t ˆ ∈L ∩L E[exp(ipN (f))] = exp(t (eipf 1) dν) for any f 1(ν). t ˆ − ∈L

c) Show that the processes considered in Corollary 1.11 are again martingales provided f 1(ν), f,g 1(ν) 2(ν) respectively. ∈L ∈L ∩L

Stochastic Analysis Andreas Eberle 1.4. STOCHASTIC INTEGRALS W.R.T. POISSON POINT PROCESSES 33

If (Nt) is a Poisson point process with intensity measure ν then the signed measure valued stochastic process

N (dy) := N (dy) tν(dy) ,t 0, t t − ≥ is called a compensated) Poisson point process .NotethatbyCorollary1.11andthe exercise, N (f)= f(y)N (dy) t ˆ t is a martingale for any f *1(ν),i.e.,(N ) is a measure-valued* martingale. ∈L t * 1.4 Stochastic integrals w.r.t. Poisson point processes

Let (S, ,ν) be a σ-finite measure space, and let ( ) be a filtration on a probability S Ft d space (Ω, ,P).OurmaininterestisthecaseS = R .Supposethat(Nt(dy))t 0 is an A ≥ ( ) Poisson point process on (S, ) with intensity measure ν.Asusual,wedenoteby Ft S N = N tν the compensated Poisson point process, and by N(dt dy) and N(dt dy) the t t− corresponding uncompensated and compensated Poisson random measure on R+ S. ) ) × Recall that for A, B with ν(A) < and ν(B) < ,theprocessesN (A), N (B), ∈S ∞ ∞ t t and Nt(A)Nt(B) tν(A B) are martingales. Our goal is to define stochastic integrals − ∩ ) ) of type ) )

(G N)t = Gs(y) N(ds dy), (1.15) • ˆ(0,t] S ×

(G N)t = Gs(y) N(ds dy) (1.16) • ˆ(0,t] S × ) ) respectively for predictable processes (ω,s,y) G (y)(ω) defined on Ω R S.In #→ s × + × particular, choosing Gs(y)(ω)=y,wewillobtainLévyprocesseswithpossiblyinfinite jump intensity from Poisson point processes. If the measure ν is finite and has no atoms, the process G N is defined in an elementary way as •

(G N)t = Gs(y). • (s,y) supp(N),s t ∈ " ≤

University of Bonn Summer Semester 2015 34 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

Definition. The predictable σ-algebra on Ω R S is the σ-algebra generated × + × P by all sets of the form A (s, t] B with 0 s t, A and B .Astochastic × × ≤ ≤ ∈Fs ∈S process defined on Ω R S is called ( ) predictable iff it is measurable w.r.t. . × + × Ft P It is not difficult to verify that any adapted left-continuous process is predictable:

Exercise. Prove that is the σ-algebra generated by all processes (ω,t,y) G (y)(ω) P #→ t such that G is measurable for any t 0 and t G (y)(ω) is left-continuous t Ft ×S ≥ #→ t for any y S and ω Ω. ∈ ∈

Example. If (Nt) is an ( t) Poisson process then the left limit process Gt(y)=Nt is F − predictable, since it is left-continuous. However, Gt(y)=Nt is not predictable. This is intuitively convincing since the jumps of a Poisson process can not be “predicted in advance”. A rigorous proof of the non-predictability, however, is surprisingly difﬁcult and seems to require some background from the general theory of stochastic processes, cf. e.g. [7].

Elementary integrands

We denote by the vector space consisting of all elementary predictable processes G E of the form n 1 m − Gt(y)(ω)= Zi,k(ω) I(ti,ti+1](t) IBk (y) (1.17) i=0 " "k=1 with m, n N, 0 t

n 1 m − (G N)t = Zi,k Nti t(Bk) Nti t(Bk) , (1.18) • +1∧ − ∧ i=0 k=1 " " $ % Notice that the summands vanish for ti t and that G N is an ( t) adapted process ≥ • F with càdlàg paths.

Stochastic integrals w.r.t. Poisson point processes have properties reminiscent of those known from Itô integrals based on Brownian motion:

Stochastic Analysis Andreas Eberle 1.4. STOCHASTIC INTEGRALS W.R.T. POISSON POINT PROCESSES 35

Lemma 1.12 (Elementary properties of stochastic integrals w.r.t. PPP). Let G . ∈E Then the following assertions hold: 1) For any t 0, ≥

E [(G N)t]=E Gs(y) ds ν(dy) . • ˆ(0,t] S 4 × 5 2) The process G N deﬁned by • (G N)t) = Gs(y) N(ds dy) Gs(y) ds ν(dy) • ˆ(0,t] S − ˆ(0,t] S × ×

is a square integrable) ( t) martingale with (G N)0 =0. F • 3) For any t 0, G N satisﬁes the Itô isometry ) ≥ • 2 2 E) (G N)t = E Gs(y) ds ν(dy) . • ˆ(0,t] S / 0 4 × 5 2 ) 2 4) The process (G N)t (0,t] S Gs(y) ds ν(dy) is an ( t) martingale. • − ´ × F Proof. 1) Since the processes) (Nt(Bk)) are Poisson processes with intensities ν(Bk), we obtain by conditioning on : Fti

E [(G N)t]= E Zi,k Nti t(Bk) Nti (Bk) • +1∧ − i,k:ti

= E Gs(y) ds ν(dy) . ˆ(0,t] S 4 × 5 2) The process G N is bounded and hence square integrable. Moreover, it is a martin- • gale, since by 1), for any 0 s t and A s, ) ≤ ≤ ∈F

E [(G N)t (G N)s; A]=E IA Gr(y) I(s,t](r) N(dr dy) • − • ˆ(0,t] S 4 × 5

= E IA Gr(y) I(s,t](r) dr ν(dy) ˆ(0,t] S 4 × 5

= E Gr(y) dr ν(dy) Gr(y) dr ν(dy); A ˆ(0,t] S − ˆ(0,s] S 4 × × 5

= E Gs(y) ds ν(dy) . ˆ(0,t] S 4 × 5 University of Bonn Summer Semester 2015 36 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

3) We have (G N)t = Zi,k ∆iN(Bk),where • i,k ! ) ∆iN(Bk) :=) Nti+1 t(Bk) Nti t(Bk) ∧ − ∧ are increments of independent) compensated) Poisson point) processes. Noticing that the summands vanish if t t,weobtain i ≥ 2 E (G N)t = E Zi,kZj,l∆iN(Bk)∆jN(Bl) • / 0 i,j,k,l" / 0 =2) E Z Z ∆ N(B ))E[∆ N(B) ) ] i,k j,l i k j l |Ftj icovariance E[∆ N(B )∆ N(B )] = δ ν(B )∆ t. i k Fti i k i l kl k i 4) now) follows similarly as 2), and is left as an exercise) to thereader.)

Lebesgue integrals

If the integrand Gt(y) is non-negative, then the integrals (1.15) and (1.16) are well- deﬁned Lebesgue integrals for every ω.ByLemma1.12andmonotoneconvergence, the identity

E [(G N)t]=E Gs(y) ds ν(dy) (1.19) • ˆ(0,t] S 4 × 5 holds for any predictable G 0. ≥ Now let u (0, ],andsupposethatG :Ω (0,u) S R is predictable and ∈ ∞ × × → integrable w.r.t. the product measure P λ ν.Thenby(1.19), ⊗ (0,u) ⊗

E Gs(y) N(ds dy) = E Gs(y) ds ν(dy) < . ˆ(0,u] S | | ˆ(0,u] S | | ∞ 4 × 5 4 × 5 + Hence the processes G N and G−N are almost surely finite on [0,u],and,correspond- • • + ingly G N = G N G−N is almost surely well-defined as a Lebesgue integral, and it • • − • satisfies the identity (1.19).

Stochastic Analysis Andreas Eberle 1.4. STOCHASTIC INTEGRALS W.R.T. POISSON POINT PROCESSES 37

Theorem 1.13. Suppose that G 1(P λ ν) is predictable. Then the following ∈L ⊗ (0,u) ⊗ assertions hold:

P 1) G N is an ( t ) adapted stochastic process satisfying (1.19). • F

P 2) The compensated process G N is an ( t ) martingale. • F ) 3) The sample paths t (G N)t are càdlàg with almost surely ﬁnite variation #→ •

(1) Vt (G N) Gs(y) N(ds dy). • ≤ ˆ(0,t] S | | ×

Proof. 1) extends by a monotone class argument from elementary predictable G to general non-negative predictable G,andhencealsotointegrablepredictableG.

2) can be veriﬁed similarly as in the proof of Lemma 1.12.

+ 3) We may assume w.l.o.g. G 0,otherwiseweconsiderG N and G−N separately. ≥ • • Then, by the Monotone Convergence Theorem,

(G N)t+ε (G N)t = Gs(y) N(ds dy) 0, and • − • ˆ(t,t+ε] S → ×

(G N)t (G N)t ε Gs(y) N(ds dy) • − • − → ˆ t S { }× as ε 0.Thisshowsthatthepathsarecàdlàg.Moreover,foranypartition π of [0,u], ↓

(G N)r′ (G N)r = Gs(y) N(ds dy) | • − • | ˆ ′ r π r π & (r,r ] S & "∈ "∈ & × & & & & Gs(y) N(ds dy) < & a.s. ≤ ˆ(0,u] S | | ∞ ×

Remark (Watanabe characterization). It can be shown that a counting measure valued process (N ) is an ( ) Poisson point process if and only if (1.19) holds for any t Ft non-negative predictable process G.

University of Bonn Summer Semester 2015 38 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

Itô integrals w.r.t. compensated Poisson point processes

Suppose that (ω,s,y) G (y)(ω) is a predictable process in 2(P λ ν) for #→ s L ⊗ (0,u) ⊗ some u (0, ].IfG is not integrable w.r.t. the product measure, then the integral G N ∈ ∞ • does not exist in general. Nevertheless, under the square integrability assumption, the integral G N w.r.t. the compensated Poisson point process exists as a square integrable • martingale. Note that square integrability does not imply integrability if the intensity ) measure ν is not ﬁnite.

To deﬁne the stochastic integral G N for square integrable integrands G we use the Itô • isometry. Let ) 2([0,u]) = M 2([0,u]) t M (ω) càdlàg for any ω Ω Md ∈M | #→ t ∈ denote the space of all square-integrable< càdlàg martingales w.r.t. the completed= ﬁltra- tion ( P ).RecallthattheL2 maximal inequality Ft 2 2 2 2 E sup Mt E[ Mu ] t [0,u] | | ≤ 2 1 | | ∈ - − . # ' holds for any right-continuous martingale in 2([0,u]).Sinceauniformlimitofcàdlàg M 2 functions is again càdlàg, this implies that the space Md ([0,u]) of equivalence classes of indistinguishable martingales in 2([0,u]) is a closed subspace of the Hilbert space Md M 2([0,u]) w.r.t. the norm

2 1/2 M 2 = E[ M ] . || ||M ([0,u]) | u| Lemma 1.12, 3), shows that for elementary predictable processes G,

G N M 2([0,u]) = G L2(P λ ν). (1.20) || • || || || ⊗ (0,u)⊗ 2 On the other hand, it can be) shown that any predictable process G L (P λ(0,u) ν) ∈ ⊗ ⊗ is a limit w.r.t. the L2(P λ ν) norm of a sequence (G(k)) consisting of elementary ⊗ (0,u) ⊗ predictable processes. Hence isometric extension of the linear map G G N can be #→ • 2 2 used to deﬁne G N Md (0,u) for any predictable G L (P λ(0,u) ν) in such a • ∈ ∈ ⊗ ⊗ ) way that ) G(k)N G N in M 2 whenever G(k) G in L2. • −→ • →

Stochastic Analysis) ) Andreas Eberle 1.5. LÉVY PROCESSES WITH INFINITE JUMP INTENSITY 39

Theorem 1.14 (Itô isometry and stochastic integrals w.r.t. compensated PPP). Suppose that u (0, ].ThenthereisauniquelinearisometryG G N from ∈ ∞ #→ • 2 2 L (Ω (0,u) S, ,P λ ν) to Md ([0,u]) such that G N is given by (1.18) for × × P ⊗ ⊗ • ) any elementary predictable process G of the form (1.17). )

Proof. As pointed out above, by (1.20), the stochastic integral extends isometrically to the closure ¯ of the subspace of elementary predictable processes in the Hilbert space E L2(Ω (0,u) S, ,P λ ν).Itonlyremainstoshowthatany square integrable × × P ⊗ ⊗ predictable process G is contained in ¯,i.e.,G is an L2 limit of elementary predictable E processes. This holds by dominated convergence for bounded left-continuous processes, and by a monotone class argument or a direct approximation forgeneralboundedpre- dictable processes, and hence also for predictable processes in L2.Thedetailsareleft to the reader.

The deﬁnition of stochastic integrals w.r.t. compensated Poisson point processes can be extended to locally square integrable predictable processes G by localization we refer − to [5] for details.

Example (Deterministic integrands). If H (y)(ω)=h(y) for some function h s ∈ 2(S, ,ν) then L S

(H N)t = h(y) Nt(dy)=Nt(h), • ˆ ) ) ) 1 i.e., H N is a Lévy martingale with jump intensity measure ν h− . • ◦ ) 1.5 Lévy processes with inﬁnite jump intensity

In this section, we are going to construct general Lévy processes from Poisson point processes and Brownian motion. Afterwards, we will considerseveralimportantclasses of Lévy jump processes with inﬁnite jump intensity.

University of Bonn Summer Semester 2015 40 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

Construction from Poisson point processes

Let ν(dy) be a positive measure on Rd 0 such that (1 y 2) ν(dy) < ,i.e., \{ } ´ ∧| | ∞ ν( y >ε) < for any ε>0, and (1.21) | | ∞ y 2 ν(dy) < . (1.22) ˆ y 1 | | ∞ | |≤ Note that the condition (1.21) is necessary for the existenceofaLévyprocesswithjump intensity ν.Indeed,if(1.21)wouldbeviolatedforsomeε>0 then a corresponding Lévy process should have inﬁnitely many jumps of size greaterthanε in ﬁnite time. This contradicts the càdlàg property of the paths. The squareintegrabilitycondition (1.22) controls the intensity of small jumps. It is crucial for the construction of a Lévy process with jump intensity ν given below, and actually it turns out to be also necessary for the existence of a corresponding Lévy process.

In order to construct the Lévy process, let N (dy), t 0,beaPoissonpointprocess t ≥ with intensity measure ν deﬁned on a probability space (Ω, ,P),andletN (dy) := A t Nt(dy) tν(dy) denote the compensated process. For a measure µ and a measurable − ) set A,wedenoteby µA(B)=µ(B A) ∩ A the part of the measure on the set A,i.e.,µ (dy)=IA(y)µ(dy).Thefollowingdecom- position property is immediate from the deﬁnition of a Poisson point process:

Remark (Decomposition of Poisson point processes). If A, B (Rd 0 ) are ∈B \{ } A B disjoint sets then (Nt )t 0 and (Nt )t 0 are independent Poisson point processes with ≥ ≥ intensity measures νA, νB respectively.

If A B (y)= for some ε>0 then the measure νA has ﬁnite total mass νA(Rd)= ∩ e ∅ ν(A) by (1.21). Therefore,

A A Xt := yNt(dy)= yNt (dy) ˆA ˆ is a compound Poisson process with intensity measure νA,andcharacteristicexponent

ψXA (p)= (1 exp(ip y)) ν(dy). ˆA − ·

Stochastic Analysis Andreas Eberle 1.5. LÉVY PROCESSES WITH INFINITE JUMP INTENSITY 41

2 On the other hand, if A y ν(dy) < then ´ | | ∞ A A Mt = y Nt(dy)= y Nt (dy) ˆA ˆ is a square integrable martingale. If both) conditions are) satisﬁed simultaneously then

A A Mt = Xt t yν(dy). − ˆA In particular, in this case M A is a Lévy process with characteristic exponent

ψM A (p)= (1 exp(ip y)+ip y) ν(dy). ˆA − · · By (1.21) and (1.22), we are able to construct a Lévy process with jump intensity measure ν that is given by

r Xt = yNt(dy)+ y Nt(dy). (1.23) ˆ y >r ˆ y r | | | |≤ for any r (0, ).Indeed,let) ) ∈ ∞ r Xt := yNt(dy)= yI y >r N(ds dy), and (1.24) ˆ y >r ˆ(0,t] Rd {| | } | | × ε,r Mt := y Nt(dy). (1.25) ˆε< y r | |≤ for ε, r [0, ) with ε

Theorem 1.15 (Existence of Lévy processes with inﬁnite jump intensity). Let ν be a positive measure on Rd 0 satisfying (1 y 2) ν(dy) < . \{ } ´ ∧| | ∞ r 1) For any r>0, (Xt ) is a compound Poisson process with intensity measure r ν (dy)=I y >r ν(dy). {| | } 0,r 2) The process (Mt ) is a Lévy martingale with characteristic exponent

ip y d ψr(p)= (1 e · + ip y) ν(dy) p R . (1.26) ˆ y r − · ∀ ∈ | |≤ Moreover, for any u (0, ), ∈ ∞ ε,r 0,r 2 E sup Mt Mt 0 as ε 0. (1.27) 0 t u | − | → ↓ 4 ≤ ≤ 5

University of Bonn Summer Semester 2015 42 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

0,r r r r 0,r 3) The Lévy processes (Mt ) and (Xt ) are independent, and Xt := Xt + Mt is aLévyprocesswithcharacteristicexponent ) ip y d ψr(p)= 1 e · + ip yI y r ν(dy) p R . (1.28) ˆ − · {| |≤ } ∀ ∈ $ % Proof. 1) is) a consequence of Theorem 1.10. 0,r 2) By (1.22), the stochastic integral (Mt ) is a square integrable martingale on [0,u] for any u (0, ).Moreover,bytheItôisometry, ∈ ∞ u 0,r ε,r 2 0,ε 2 2 M M M 2([0,u]) = M M 2([0,u]) = y I y ε ν(dy) dt 0 ∥ − ∥ ∥ ∥ ˆ0 ˆ | | {| |≤ } → as ε 0.ByTheorem1.10,(M ε,r) is a compensated compound Poisson process with ↓ t intensity I ε< y r ν(dy) and characteristic exponent { | |≤ }

ip y ψε,r(p)= (1 e · + ip y) ν(dy). ˆε< y r − · | |≤ ip y 2 As ε 0,ψε,r(p) converges to ψr(p) since 1 e · + ip y = ( y ).Hencethelimit ↓ 0,r 1/n,r − · O | | martingale Mt = lim Mt also has independent and stationary increments, and n characteristic function→∞

0,r 1/n,1 E[exp(ip Mt )] = lim E[exp(ip Mt )] = exp( tψr(p)). n · →∞ · −

3) Since I y r Nt(dy) and I y >r Nt(dy) are independent Poisson point processes, {| |≤ } {| | } 0,r r r 0,r r the Lévy processes (Mt ) and (Xt ) are also independent. Hence Xt = Mt + Xt is a Lévy process with characteristic exponent ) ip y ψr(p)=ψr(p)+ (1 e · ) ν(dy). ˆ y >r − | | )

Remark. All the partially compensated processes (Xr), r (0, ),areLévypro- t ∈ ∞ cesses with jump intensity ν.Actually,theseprocessesdifferonlybyaﬁnitedriftterm, ) since for any 0 <ε

ε r Xt = Xt + bt, where b = yν(dy). ˆε< y r | |≤ ) ) Stochastic Analysis Andreas Eberle 1.5. LÉVY PROCESSES WITH INFINITE JUMP INTENSITY 43

AtotallyuncompensatedLévyprocess

Xt = lim yNt(dy) n ˆ y 1/n →∞ | |≥ does exist only under additional assumptions on the jump intensity measure:

Corollary 1.16 (Existence of uncompensated Lévy jump processes). Suppose that (1 y ) ν(dy) < ,orthatν is symmetric (i.e., ν(B)=ν( B) for any B ∧| | ∞ − ∈ ´ (Rd 0 ))and (1 y 2) ν(dy) < .ThenthereexistsaLévyprocess(X ) with B \{ } ∧| | ∞ t characteristic exponent´

ip y d ψ(p) = lim 1 e · ν(dy) p R (1.29) ε 0 ˆ ↓ y >ε − ∀ ∈ | | $ % such that ε 2 E sup Xt Xt 0 as ε 0. (1.30) 0 t u | − | → ↓ 4 ≤ ≤ 5 Proof. For 0 <ε

ε r ε,r Xt = Xt + Mt + t yν(dy). ˆε< y r | |≤ As ε 0, M ε,r converges to M 0,r in M 2([0,u]) for any ﬁnite u.Moreover,underthe ↓ assumption imposed on ν,theintegralontherighthandsideconvergestobt where

b = lim yν(dy). ε 0 ˆε< y r ↓ | |≤ Therefore, (Xε) converges to a process (X ) in the sense of (1.30) as ε 0.The t t ↓ limit process is again a Lévy process, and, by dominated convergence, the characteristic exponent is given by (1.29).

Remark (Lévy processes with finite variation paths). If (1 y ) ν(dy) < then ´ ∧| | ∞ the process Xt = yNt(dy) is defined as a Lebesgue integral. As remarked above, in ´ that case the paths of (Xt) are almost surely of finite variation:

V (1)(X) y N (dy) < a.s. t ≤ ˆ | | t ∞

University of Bonn Summer Semester 2015 44 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

The Lévy-Itô decomposition

We have constructed Lévy processes corresponding to a given jump intensity measure ν under adequate integrability conditions as limits of compound Poisson processes or partially compensated compound Poisson processes, respectively. Remarkably, it turns out that by taking linear combinations of these Lévy jump processes and Gaussian Lévy processes, we obtain all Lévy processes. This is the content of the Lévy-Itô decomposition theorem that we will now state before considering in more detail some important classes of Lévy processes.

Already the classical Lévy-Khinchin formula for infinity divisible random variables (see Corollary 1.18 below) shows that any Lévy process on Rd can be characterized by three d d d quantities:anon-negativedefinitesymmetricmatrixa R × ,avectorb R ,anda ∈ ∈ σ-finite measure ν on (Rd 0 ) such that B \{ }

(1 y 2) ν(dy) < . (1.31) ˆ ∧| | ∞ Note that (1.31) holds if and only if ν is ﬁnite on complements of balls around 0, and 2 y 1 y ν(dy) < .TheLévy-Itôdecompositiongivesanexplicitrepresentation of | |≤ | | ∞ aLévyprocesswithcharacteristics´ (a, b, ν).

d d T Let σ R × with a = σσ ,let(B ) be a d-dimensional Brownian motion, and let (N ) ∈ t t be an independent Poisson point process with intensity measure ν.WedeﬁneaLévy process (Xt) by setting

Xt = σBt + bt + yNt(dy)+ y (Nt(dy) tν(dy)) . (1.32) ˆ y >1 ˆ y 1 − | | | |≤ The ﬁrst two summands are the diffusion part and the drift of a Gaussian Lévy process, the third summand is a pure jump process with jumps of size greater than 1,andthelast summand represents small jumps compensated by drift. As a sumofindependentLévy processes, the process (Xt) is a Lévy process with characteristic exponent

1 ip y ψ(p)= p ap ib p + (1 e · + ip yI y 1 ) ν(dy). (1.33) 2 · − · ˆRd 0 − · {| |≤ } \{ } We have thus proved the ﬁrst part of the following theorem:

Stochastic Analysis Andreas Eberle 1.5. LÉVY PROCESSES WITH INFINITE JUMP INTENSITY 45

Theorem 1.17 (Lévy-Itô decomposition). 1) The expression (1.32) deﬁnes a Lévy process with characteristic exponent ψ.

2) Conversely, any Lévy process (Xt) can be decomposed as in (1.32) with the Pois- son point process

N = ∆X ,t 0, (1.34) t s ≥ s t ∆"X≤s=0 ̸ d d d an independent Brownian motion (B ),amatrixσ R × ,avectorb R ,and t ∈ ∈ a σ-finite measure ν on Rd 0 satisfying (1.31). \{ } We will not prove the second part of the theorem here. The principal way to proceed is to define (Nt) via (1.31), and to consider the difference of (Xt) and the integrals w.r.t. (Nt) on the right hand side of (1.32). One can show that the difference is a continuous Lévy process which can then be identified as a GaussianLévyprocessbythe Lévy characterization, cf. Section 2.1 below. Carrying out the details of this argument, however, is still a lot of work. A detailed proof is given in [5].

As a byproduct of the Lévy-Itô decomposition, one recovers the classical Lévy-Khinchin formula for the characteristic functions of inﬁnitely divisible random variables, which can also be derived directly by an analytic argument.

Corollary 1.18 (Lévy-Khinchin formula). For a function ψ : Rd C the following → statements are all equivalent:

(i) ψ is the characteristic exponent of a Lévy process.

(ii) exp( ψ) is the characteristic function of an infinitely divisible random variable. − d d (iii) ψ satisfies (1.33) with a non-negative definite symmetric matrix a R × ,a ∈ vector b Rd,andameasureν on (Rd 0 ) such that (1 y 2) ν(dy) < . ∈ B \{ } ´ ∧| | ∞ Proof. (iii) (i) holds by the first part of Theorem 1.17. ⇒ (i) (ii): If (X ) is a Lévy process with characteristic exponent ψ then X X is an ⇒ t 1 − 0 infinitely divisible random variable with characteristic function exp( ψ). − (ii) (iii) is the content of the classical Lévy-Khinchin theorem,seee.g.[17]. ⇒

University of Bonn Summer Semester 2015 46 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

We are now going to consider several important subclasses of Lévy processes. The class of Gaussian Lévy processes of type

Xt = σBt + bt

d d d with σ R × , b R ,andad-dimensional Brownian motion (B ) has already been ∈ ∈ t introduced before. The Lévy-Itô decomposition states in particular that these are the only Lévy processes with continuous paths!

Subordinators

A subordinator is by deﬁnition a non-decreasing real-valued Lévy process. The name comes from the fact that subordinators are used to change the time-parametrization of a Lévy process, cf. below. Of course, the deterministic processes X = bt with b 0 are t ≥ subordinators. Furthermore, any compound Poisson process with non-negative jumps is a subordinator. To obtain more interesting examples, we assume that ν is a positive measure on (0, ) with ∞ (1 y) ν(dy) < . ˆ(0, ) ∧ ∞ ∞

Then a Poisson point process (Nt) with intensity measure ν satisﬁes almost surely

supp(N ) [0, ) for any t 0. t ⊂ ∞ ≥ Hence the integrals X = yN(dy) ,t 0, t ˆ t ≥ deﬁne a non-negative Lévy process with X0 =0.Bystationarity,allincrementsof(Xt) are almost surely non-negative, i.e., (Xt) is increasing. In particular, the sample paths are (almost surely) of ﬁnite variation.

Example (Gamma process). The Gamma distributions form a convolution semigroup of probability measures on (0, ),i.e., ∞ Γ(r, λ) Γ(s, λ)=Γ(r + s, λ) for any r, s, λ> 0. ∗ Stochastic Analysis Andreas Eberle 1.5. LÉVY PROCESSES WITH INFINITE JUMP INTENSITY 47

Therefore, for any a, λ> 0 there exists an increasing Lévy process (Γt)t 0 with incre- ≥ ment distributions

Γ Γ Γ(at, λ) for any s, t 0. t+s − s ∼ ≥ Computation of the Laplace transform yields

at u − ∞ uxy 1 λy E[exp( uΓt)] = 1+ =exp t (1 e− )ay− e− dy (1.35) − λ − ˆ0 − 6 7 - . for every u 0,cf.e.g.[28,Lemma1.7].SinceΓ 0,bothsidesin(1.35)havea ≥ t ≥ unique analytic extension to u C : (u) 0 .Therefore,wecanreplaceu by ip { ∈ ℜ ≥ } − in (1.35) to conclude that the characteristic exponent of (Γt) is

∞ ipy 1 λy ψ(p)= (1 e ) ν(dy), where ν(dy)=ay− e− dy. ˆ0 − Hence the Gamma process is a non-decreasing pure jump processwithjumpintensity measure ν.

Example (Inverse Gaussian processes). An explicit computation of the characteristic function shows that the Lévy subordinator (Ts) is a pure jump Lévy process with Lévy measure 1/2 3/2 ν(dy)=(2π)− y− I(0, )(y) dx. ∞

More generally, if Xt = σBt + bt is a Gaussian Lévy process with coefﬁcients σ>0, b R,thentherightinverse ∈ T X = inf t 0:X = s ,s 0, s { ≥ t } ≥ is a Lévy jump process with jump intensity

1/2 3/2 2 ν(dy)=(2π)− y− exp( b y/2)I(0, )(y) dy. − ∞

Remark (Finite variation Lévy jump processes on R1).

Suppose that (Nt) is a Poisson point process on R 0 with jump intensity measure ν \{ } (0, ) ( ,0) satisfying (1 y ) ν(dy) < .ThenthedecompositionNt = Nt ∞ + Nt −∞ into ´ ∧| | ∞ University of Bonn Summer Semester 2015 48 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

the independent restrictions of (Nt) to R+, R respectively induces a corresponding − decomposition

(0, ) ( ,0) X = X↗ + X↘ ,X↗ = yN ∞ (dy) ,X↘ = yN−∞ (dy), t t t t ˆ t t ˆ t of the associated Lévy jump process Xt = yNt(dy) into a subordinator Xt↗ and a ´ decreasing Lévy process Xt↘.Inparticular,weseeoncemorethat(Xt) has almost surely paths of ﬁnite variation.

An important property of subordinators is that they can be used for random time transformations of Lévy processes:

Exercise (Time change by subordinators). Suppose that (Xt) is a Lévy process with Laplace exponent η : R R,i.e., X + → E[exp( αX )] = exp( tη (α)) for any α 0. − t − X ≥

Prove that if (Ts) is an independent subordinator with Laplace exponent ηT then the time-changed process X := X ,s 0, s Ts ≥ is again a Lévy process with Laplace) exponent

η(p)=ηT (ηX (p)).

The characteristic exponent can) be obtained from this identity by analytic continuation.

Example (Subordinated Lévy processes). Let (Bt) be a Brownian motion.

1) If (Nt) is an independent Poisson process with parameter λ>0 then (BNt ) is a compensated Poisson process with Lévy measure

1/2 2 ν(dy)=λ(2π)− exp( y /2) dy. − 2) If (Γ ) is an independent Gamma process then for σ, b R the process t ∈

Xt = σBΓt + bΓt

Stochastic Analysis Andreas Eberle 1.5. LÉVY PROCESSES WITH INFINITE JUMP INTENSITY 49

is called a Variance Gamma process.ItisaLévyprocesswithcharacteristic exponent ψ(p)= (1 eipy) ν(dy),where ´ − 1 λy µ y ν(dy)=c y − e− I(0, )(y)+e− | |I( ,0)(y) dy | | ∞ −∞ with constants c, λ, µ > 0.Inparticular,aVarianceGammaprocesssatisﬁes$ % X =Γ(1) Γ(2) with two independent Gamma processes. Thus the increments of t t − t (Xt) have exponential tails. Variance Gamma processes have been introduced and applied to option pricing by Madan and Seneta [31] as an alternative to Brownian motion taking into account longer tails and allowing for a wider modeling of skewness and kurtosis. 3) Normal Inverse Gaussian (NIG) processes are time changes of Brownian motions with drift by inverse Gaussian subordinators [6]. Their increments over unit time intervals have a normal inverse Gaussian distribution,whichhasslowerde- caying tails than a normal distribution. NIG processes are applied in statistical modelling in ﬁnance and turbulence.

Stable processes

We have noted in (1.14) that the jump intensity measure of a strictly α-stable process in R1 is given by

1 α ν(dy)= c+I(0, )(y)+c I( ,0)(y) y − − dy (1.36) ∞ − −∞ | | $ % with constants c+,c [0, ).Notethatforanyα (0, 2),themeasureν is ﬁnite on − ∈ ∞ ∈ R 2 ( 1, 1),and [ 1,1] y ν(dy) < . \ − ´ − | | ∞ We will prove now that if α (0, 1) (1, 2) then for each choice of the constants c ∈ ∪ + and c ,thereisastrictlyα-stable process with Lévy measure (1.36). For α =1this − is only true if c+ = c ,whereasanon-symmetric1-stable process is given by Xt = bt − with b R 0 .Todeﬁnethecorrespondingα-stable processes, let ∈ \{ } ε Xt = yNt(dy) ˆR [ ε,ε] \ − where (N ) is a Poisson point process with intensity measure ν.Setting X = t || ||u 2 1/2 E[supt a Xt ] ,anapplicationofTheorem1.15yields: ≤ | | University of Bonn Summer Semester 2015 50 CHAPTER 1. LÉVY PROCESSES AND POISSON POINT PROCESSES

Corollary 1.19 (Construction of α-stable processes). Let ν be the probability measure on R 0 deﬁned by (1.36) with c+,c [0, ). \{ } − ∈ ∞

1) If c+ = c then there exists a symmetric α-stable process X with characteristic − exponent ψ(p)=γ p α, γ = (1 cos y) ν(dy) R,suchthat X1/n X | | − ∈ || − ||u → 0 for any u (0, ). ´ ∈ ∞ 2) If α (0, 1) then (1 y ) ν(dy) < ,andX = yN(dy) is an α-stable ∈ ∧| | ∞ t t process with characteristic´ exponent ψ(p)=z p α, z =´ 1 eiy ν(dy) C. | | ´ − ∈ 3) For α =1and b R,thedeterministicprocessX = bt is$α-stable% with charac- ∈ t teristic exponent ψ(p)= ibp. − 4) Finally, if α (1, 2) then ( y y 2) ν(dy) < ,andthecompensatedprocess ∈ ´ | |∧| | ∞ Xt = y Nt(dy) is an α-stable martingale with characteristic exponent ψ(p)= z p α´, z = (1 eiy + iy) ν(dy). ·| | ) ´ − Proof.)By Theorem) 1.15 it is sufﬁcient to prove convergence of the characteristic exponents

ψ (p)= 1 eipy ν(dy)= p α 1 eix ν(dx), ε ˆ − | | ˆ − R [ ε,ε] R [ εp,εp] \ − $ % \ − $ % ψ (p)= 1 eipy + ipy ν(dy)= p α 1 eix + ix ν(dx) ε ˆ − | | ˆ − R [ ε,ε] $ % R [ εp,εp] $ % ) \ − \ − to ψ(p), ψ(p) respectively as ε 0.Thisiseasilyveriﬁedincases1),2)and4)by ↓ ix ix 2 ix noting that 1 e +1 e− =2(1 cos x)= (x ), 1 e = ( x ),and ) − − − O − O | | 1 eix + ix = ( x 2). − O | | Notice that although the characteristic exponents in the non-symmetric cases 2), 3) and 4) above take a similar form (but with different constants), the processes are actually very different. In particular, for α>1,astrictlyα-stable process is always a limit of compensated compound Poisson processes and hence a martingale!

Example (α-stable subordinators vs. α-stable martingales). For c =0and α − ∈ (0, 1),theα-stable process with jump intensity ν is increasing, i.e., it is an α-stable

Stochastic Analysis Andreas Eberle 1.5. LÉVY PROCESSES WITH INFINITE JUMP INTENSITY 51 subordinator.Forc =0and α (1, 2) this is not the case since the jumps are − ∈ “compensated by an inﬁnite drift”. The graphics below show simulations of samples from α-stable processes for c =0and α =3/2, α =1/2 respectively. For α (0, 2), − ∈ asymmetricα-stable process has the same law as (√2BTs ) where (Bt) is a Brownian motion and (Ts) is an independent α/2-stable subordinator.

100

500 1000 1500 2000 2500 3000

!100

!200

5 " 106

4 " 106

3 " 106

2 " 106

1 " 106

500 1000 1500 2000 2500 3000

University of Bonn Summer Semester 2015 Chapter 2

Transformations of SDE

Let U Rn be an open set. We consider a stochastic differential equation of the form ⊆

dXt = b(t, Xt) dt + σ(t, Xt) dBt (2.1) with a d-dimensional Brownian motion (B ) and measurable coefficients b : [0, ) t ∞ × n n d U R and σ : [0, ) U R × .Inapplicationsoneisoftennotinterestedinthe → ∞ × → random variables X :Ω R themselves but only in their joint distribution. In that t → case, it is usually irrelevant w.r.t. which Brownian motion (Bt) the SDE (2.1) is satisfied. Therefore, we can “solve” the SDE in a very different way: Instead of constructing the solution from a given Brownian motion, we first construct a stochastic process (Xt,P) by different types of transformations or approximations, and then we verify that the process satisfies (2.1) w.r.t. some Brownian motion (Bt) that is usually defined through (2.1). Definition (Weak and strong solutions). A (weak) solution of the stochastic differential equation (2.1) is given by

(i) a “setup” consisting of a probability space (Ω, ,P),aﬁltration( t)t 0 on A F ≥ d (Ω, ) and an R -valued ( t) Brownian motion (Bt)t 0 on (Ω, ,P), A F ≥ A (ii) a continuous ( ) adapted stochastic process (X ) where S is an ( ) stopping Ft t t

~~52 53~~

It is called a strong solution w.r.t. the given setup if and only if (Xt) is adapted w.r.t. the B,P ﬁltration σ t ,X0 t 0 generated by the Brownian motion and the initial condition. F ≥ Here q $ ([0$ ,S), Rn)%%consists of all Rn valued processes (ω,t) H (ω) deﬁned for La,loc #→ t t

Remark. Asolution(Xt)t 0 is a strong solution up to S = w.r.t. a given setup if ≥ ∞ and only if there exists a measurable map F : R Rn C R , Rd Rn, (t, x ,y) + × × + → 0 #→ n Ft(x0,y),suchthattheprocess(Ft)t 0 is adapted w.r.t. the ﬁltration (R ) t, t = ≥ $ % B ⊗B B σ(y y(s):0 s t),and #→ ≤ ≤ X = F (X ,B) for any t 0 t t 0 ≥ holds almost surely. Hence strong solutions are (almost surely) functions of the given Brownian motion and the initial value!

There are SDE that have weak but no strong solutions. An example is given in Section 2.1. The deﬁnition of weak and strong solutions can be generalized to other types of SDE including in particular functional equations of the form

~~dXt = bt(X) dt + σt(X) dBt where (b ) and (σ ) are ( ) adapted stochastic processes deﬁned on C(R , Rn),aswell t t Bt + as SDE driven by Poisson point processes, cf. Chapter 4.~~

~~Different types of transformations of a stochastic process (Xt,P) are useful for constructing weak solutions. These include:~~

Random time changes: (Xt)t 0 (XTa )a 0 where (Ta)a 0 is an increasing stochas- • ≥ → ≥ ≥ tic process on R such that T is a stopping time for any a 0. + a ≥ Transformations of the paths in space: These include for example coordinate changes • (X ) (ϕ(X )),randomtranslations(X ) (X +H ) where (H ) is another adapted t → t t → t t t process, and, more generally, a transformation that maps (Xt) to the strong solution (Yt)

~~University of Bonn Summer Semester 2015 54 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~of an SDE driven by (Xt).~~

Change of measure: Here the random variables X are kept fixed but the underlying • t probability measure P is replaced by a new measure P such that both measures are mutually absolutely continuous on each of the σ-algebras t, t R+ (but usually not on ) F ∈ ). F∞ In Sections 2.2, 2.3 and 2.4, we study these transformations as well as relations between them. For identifying the transformed processes, the Lévy characterizations in Section 2.1 play a crucial rôle. Section 2.5 contains an application to large deviations on Wiener space, and, more generally, random perturbations of dynamical systems. Section 3.2 fo- cusses on Stratonovich differential equations. As the Stratonovich integral satisfies the usual chain rule, these are adequate for studying stochasticprocessesonRiemannian manifolds. Stratonovich calculus also leads to a tranformation of an SDE in terms of the flow of a corresponding ODE that is useful for example in theone-dimensionalcase. The concluding Section 3.4 considers numerical approximation schemes for solutions of stochastic differential equations.

~~2.1 Lévy characterizations and martingale problems~~

~~Let (Ω, ,P,( )) be a given ﬁltered probability space. We ﬁrst note that Lévy pro- A Ft cesses can be characterized by their exponential martingales:~~

Lemma 2.1. Let ψ : Rd C be a given function. An ( ) adapted càdlàg process → Ft X :Ω Rd is an ( ) Lévy process with characteristic exponent ψ if and only if the t → Ft complex-valued processes

~~Zp := exp ip X + tψ(p) ,t 0, t · t ≥ $ % are ( ) martingales, or, equivalently, local ( ) martingales for any p Rd. Ft Ft ∈~~

~~Proof. By Corollary 1.5, the processes Zp are martingales if X is a Lévy process with characteristic exponent ψ.Conversely,supposethatZp is a local martingale for any~~

Stochastic Analysis Andreas Eberle 2.1. LÉVY CHARACTERIZATIONS AND MARTINGALE PROBLEMS 55 p Rd.Then,sincetheseprocessesareuniformlyboundedonﬁnitetime intervals, ∈ they are martingales. Hence for 0 s t and p Rd, ≤ ≤ ∈ E exp ip (X X ) =exp((t s)ψ(p)), · t − s Fs − − which implies that# X $ X is independent%& ' of with characteristic function equal to t − s & Fs exp( (t s)ψ). − − Exercise (Characterization of Poisson point processes). Let (S, ,ν) be a σ-ﬁnite S measure space. Suppose that (Nt)t 0 on (Ω, ,P) is an ( t) adapted process taking ≥ A F + values in the space Mc (S) consisting of all counting measures on S.Provethatthe following statements are equivalent:

(i) (Nt) is a Poisson point processes with intensity measure ν. (ii) For any function f 1(S, ,ν),therealvaluedprocess ∈L S N (f)= f(y) N (dy),t0, t ˆ t ≥ 1 is a compound Poisson process with jump intensity measure µ f − . ◦ (iii) For any function f 1(S, ,ν),thecomplexvaluedprocess ∈L S M [f] =exp(iN (f)+tψ(f)),t0,ψ(f)= 1 eif dν, t t ≥ ˆ − is a local ( ) martingale. $ % Ft Show that the statements are also equivalent if only elementary functions f L1(S, ,ν) ∈ S are considered.

~~Lévy’s characterization of Brownian motion~~

d By Lemma 2.1, an R -valued process (Xt) is a Brownian motion if and only if the processes exp ip X + t p 2/2 are local martingales for all p Rd.Thiscanbe · t | | ∈ Rd applied to prove$ the remarkable fact% that any continuous valued martingale with the right covariations is a Brownian motion:

~~Theorem 2.2 (P. Lévy 1948). Suppose that M 1,...,Md are continuous local ( ) mar- Ft tingales with [M k,Ml] = δ tP-a.s. for any t 0. t kl ≥ Then M =(M 1,...,Md) is a d-dimensional Brownian motion.~~

~~University of Bonn Summer Semester 2015 56 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~The following proof is due to Kunita and Watanabe (1967):~~

Proof. Fix p Rd and let Φ := exp(ip M ).ByItô’sformula, ∈ t · t 1 d dΦ = ip Φ dM Φ p p d[M k,Ml] t t · t − 2 t k l t k,l"=1 1 = ip Φ dM Φ p 2 dt. t · t − 2 t | | Since the ﬁrst term on the right hand side is a local martingaleincrement,theproduct rule shows that the process Φ exp( p 2 t/2) is a local martingale. Hence by Lemma t · | | 2.1, M is a Brownian motion.

~~Lévy’s characterization of Brownian motion has a lot of remarkable direct consequences.~~

~~Example (Random orthogonal transformations). Suppose that X :Ω Rn is a t → solution of an SDE~~

dXt = Ot dBt,X0 = x0, (2.2) w.r.t. a d-dimensional Brownian motion (Bt),aproduct-measurableadaptedprocess n d n (t, ω) O (ω) taking values in R × ,andaninitialvalex R .WeverifythatX is #→ t 0 ∈ an n-dimensional Brownian motion provided

~~O (ω) O (ω)T = I for any t 0, almost surely. (2.3) t t n ≥ Indeed, by (2.2) and (2.3), the components~~

~~d t i i ik k Xt = x0 + Os dBs ˆ0 "k=1 are continuous local martingales with covariations~~

~~[Xi,Xj]= Oik Ojl d[Bk,Bl]= Oik Ojk dt = δ dt. ˆ ˆ ij "k,l "k Applications include inﬁnitesimal random rotations (n = d)andrandomorthogonal projections (n~~

~~Stochastic Analysis Andreas Eberle 2.1. LÉVY CHARACTERIZATIONS AND MARTINGALE PROBLEMS 57~~

Example (Bessel process). We derive an SDE for the radial component R = B of t | t| Brownian motion in Rd.Thefunctionr(x)= x is smooth on Rd 0 with r(x)= | | \{ } ∇ 1 e (x),and∆r(x)=(d 1) x − where e (x)=x/ x .ApplyingItô’sformulato r − ·| | r | | d functions r C∞(R ), ε>0,withr (x)=r(x) for x ε yields ε ∈ ε | |≥ d 1 dRt = er(Bt) dBt + − dt for any t

t Wt := er(Bs) dBs,t0, ˆ0 · ≥ is a one-dimensional Brownian motion deﬁned for all times (the value of er at 0 being irrelevant for the stochastic integral). Hence (Bt) is a weak solution of the SDE d 1 dRt = dWt + − dt (2.4) 2Rt up to the ﬁrst hitting time of 0.Theequation(2.4)makessenseforanyparticulard R ∈ and is called the Bessel equation.MuchmoreonBesselprocessescanbefoundin Revuz and Yor [37] and other works by M. Yor.

Exercise (Exit times and ruin probabilities for Bessel and compound Poisson processes). a) Let (Xt) be a solution of the Bessel equation d 1 dXt = − dt + dBt,X0 = x0, − 2Xt where (Bt)t 0 is a standard Brownian motion and d is a real constant. ≥ i) Find a non-constant function u : R R such that u(X ) is a local martingale up → t to the ﬁrst hitting time of 0.

~~ii) Compute the ruin probabilities P [T 0.~~

~~University of Bonn Summer Semester 2015 58 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~i) Determine λ R such that exp(λX ) is a local martingale up to T . ∈ t 0 ii) Prove that for a<0,~~

~~P [Ta < ] = lim P [Ta~~

~~Whyisit notas easy as aboveto computethe ruinprobability P [Ta~~

~~The next application of Lévy’s characterization of Brownianmotionshowsthatthere are SDE that have weak but no strong solutions.~~

~~Example (Tanaka’s example. Weak vs. strong solutions). Consider the one dimensional SDE~~

~~dXt =sgn(Xt) dBt (2.5)~~

+1 for x 0, where (Bt) is a Brownian motion and sgn(x) := ≥ .Notetheunusual ⎧ 1 for x<0 ⎨− convention sgn(0) = 1 that is used below. We prove the following statements: ⎩ 1) X is a weak solution of (2.5) on (Ω, ,P,( )) if and only if X is an ( ) Brown- A Ft Ft ian motion. In particular, aweaksolutionexistsand its law is uniquely determined

~~by the law of the initial value X0.~~

~~2) If X is a weak solution w.r.t. a setup (Ω, ,P,( t), (Bt)) then for any t 0, A X ,P F ≥ B B is measurable w.r.t. the σ-algebra | | = σ( X : s t)P . t − 0 Ft | s| ≤~~

~~3) There is no strong solution to (2.5) with initial condition X0 =0.~~

4) Pathwise uniqueness does not hold: If X is a solution to (2.5) with X0 =0then X solves the same equation with the same Brownian motion. − The proof of 1) is again a consequence of the ﬁrst example above: If X is a weak solution then X is a Brownian motion by Lévy’s characterization. Conversely, if X is an ( ) Brownian motion then the process Ft t Bt := sgn(Xs) dXs ˆ0

Stochastic Analysis Andreas Eberle 2.1. LÉVY CHARACTERIZATIONS AND MARTINGALE PROBLEMS 59 is a Brownian motion as well, and t t 2 sgn(Xs) dBs = sgn(Xs) dXs = Xt X0, ˆ0 ˆ0 − i.e., X is a weak solution to (2.5). For proving 2) , we approximate r(x)= x by symmetric and concave functions r | | ε ∈ C∞(R) satisfying r (x)= x for x ε.Thentheassociativelaw,theItôisometry, ε | | | |≥ and Itô’s formula imply t t

Bt B0 = sgn(Xs) dXs = lim rε′′(Xs) dXs ˆ ε 0 ˆ − 0 ↓ 0 1 t = lim rε(Xt) rε(X0) rε′′(Xs) ds ε 0 2 ˆ ↓ − − 0 $ 1 t % = lim rε( Xt ) rε( X0 ) rε′′( Xs ) ds ε 0 ˆ ↓ | | − | | − 2 0 | | with almost sure convergence along$ a subsequence ε 0. % n ↓

Finally by 2), if X would be a strong solution w.r.t. a Brownian motion B then Xt X ,P would also be measurable w.r.t. the σ-algebra generated by and | | .Thisleads F0 Ft to a contradiction as one can verify that the event X 0 is not measurable w.r.t. this { t ≥ } σ-algebra for a Brownian motion (Xt).

~~Martingale problem for Itô diffusions~~

~~Next we consider a solution of a stochastic differential equation~~

dXt = b(t, Xt) dt + σ(t, Xt) dBt,X0 = x0, (2.6) defined on a filtered probability space (Ω, ,P,( )).Weassumethat(B ) is an ( ) A Ft t Ft Brownian motion taking values in Rd,b,σ ,...,σ : R+ Rn Rn are measurable 1 d × → and locally bounded (i.e., bounded on [0,t] K for any t 0 and any compact set K × ≥ ⊂ Rd)time-dependentvectorfields,andσ(t, x)=(σ (t, x) σ (t, x)) is the n d matrix 1 ··· d × with column vectors σ (t, x).Asolutionof(2.6)isacontinuous( P ) semimartingale i Ft (Xt) satisfying

~~t d t k Xt = x0 + b(s, Xs) ds + σk(s, Xs) dBs t 0 a.s. (2.7) ˆ0 ˆ0 ∀ ≥ "k=1 University of Bonn Summer Semester 2015 60 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~If X is a solution then~~

~~i j i k j l [X ,X ]t = σk(s, X) dB , σl (s, X) dB ˆ ˆ t k,l / 0 " t t i j k l ij = (σk σl )(s, X) d[B ,B ]= a (s, Xs) ds ˆ0 ˆ0 "k,l ij i j where a = k σkσk,i.e.,~~

~~! T n n a(s, x)=σ(s, x)σ(s, x) R × . ∈~~

~~Therefore, Itô’s formula applied to the process (t, Xt) yields~~

∂F 1 d ∂2F dF (t, X)= (t, X) dt + F (t, X) dX + (t, X) d[Xi,Xj] ∂t ∇x · 2 ∂xi∂xj i,j=1 " ∂F =(σT F )(t, X) dB + + F (t, X) dt, ∇x · ∂t L 6 7 for any F C2(R Rn),where ∈ + × 1 d ∂2F d ∂F ( F )(t, x)= aij(t, x) (t, x)+ bi(t, x) (t, x). L 2 ∂xi∂xj ∂xi i,j=1 i=1 " " We have thus derived the Itô-Doeblin formula

~~t t T ∂F F (t, Xt) F (0,X0)= (σ F )(s, Xs) dBs + + F (s, Xs) ds − ˆ0 ∇ · ˆ0 ∂t L $ % (2.8)~~

The formula provides a semimartingale decomposition for F (t, Xt).Itestablishesacon- nection between the stochastic differential equation (2.6)andpartialdifferentialequa- tions involving the operator . L Example (Exit distributions and boundary value problems). Suppose that F ∈ C2(R Rn) is a classical solution of the p.d.e. + × ∂F (t, x)+( F )(t, x)= g(t, x) t 0,x U ∂t L − ∀ ≥ ∈ on an open subset U Rn with boundary values ⊂ F (t, x)=ϕ(t, x) t 0,x ∂U. ∀ ≥ ∈ Stochastic Analysis Andreas Eberle 2.1. LÉVY CHARACTERIZATIONS AND MARTINGALE PROBLEMS 61

~~Then by (2.8), the process~~

t Mt = F (t, Xt)+ g(s, Xs) ds ˆ0 is a local martingale. If F and g are bounded on [0,t] U,thentheprocessM T stopped × at the ﬁrst exit time T = inf t 0:X / U is a martingale. Hence, if T is almost { ≥ t ∈ } surely ﬁnite then

~~T E[ϕ(T,XT )] + E g(s, Xs) ds = F (0,x0). ˆ0 / 0 This can be used, for example, to compute exit distributions (for g 0) and mean exit ≡ times (for ϕ 0, g 1)analyticallyornumerically. ≡ ≡~~

Similarly as in the example, the Feynman-Kac-formula and other connections between Brownian motion and the Laplace operator carry over to Itô diffusions and their generator in a straightforward way. Of course, the resulting partial differential equation L usually can not be solved analytically, but there is a wide range of well-established numerical methods for linear PDE available for explicit computations of expectation values.

Exercise (Feynman-Kac formula for Itô diffusions). Fix t (0, ),andsuppose ∈ ∞ that ϕ : Rn R and V : [0,t] Rn [0, ) are continuous functions. Show that if → × → ∞ u C2((0,t] Rn) C([0,t] Rn) is a bounded solution of the heat equation ∈ × ∩ × ∂u (s, x)=(u)(s, x) V (s, x)u(s, x) for s (0,t],x Rn, ∂s L − ∈ ∈ u(0,x)=ϕ(x), then u has the stochastic representation

t u(t, x)=E ϕ(X )exp V (t s, X ) ds . x t − ˆ − s 4 - 0 .5 Hint: Consider the time reversal uˆ(s, x) := u(t s, x) of u on [0,t].Showﬁrstthat − r ˆ Mr := exp( Ar)ˆu(r, Xr) is a local martingale if Ar := 0 V (s, Xs) ds. − ´ University of Bonn Summer Semester 2015 62 CHAPTER 2. TRANSFORMATIONS OF SDE

Often, the solution of an SDE is only deﬁned up to some explosion time ζ where it diverges or exits a given domain. By localization, we can apply the results above in this case as well. Indeed, suppose that U Rn is an open set, and let ⊆ U = x U : x 1/k ,kN. k { ∈ | | } ∈

Then U = Uk.LetTk denote the ﬁrst exit time of (Xt) from Uk.Asolution(Xt) of the SDE (2.6) up to the explosion time ζ =supTk is a process (Xt)t [0,ζ) 0 such that > ∈ ∪{ } for every k N, T <ζalmost surely on ζ (0, ) ,andthestoppedprocessXTk is ∈ k { ∈ ∞ } asemimartingalesatisfying(2.7)fort T .ByapplyingItô’sformulatothestopped ≤ k processes, we obtain:

Theorem 2.3 (Martingale problem for Itô diffusions). If X :Ω U is a solution of t → (2.6) up to the explosion time ζ,thenforanyF C2(R U) and x U,theprocess ∈ + × 0 ∈ t ∂F Mt := F (t, Xt) + F (s, Xs) ds, t < ζ, − ˆ0 ∂t L 6 7 is a local martingale up to the explosion time ζ,andthestoppedprocessesM Tk , k N, ∈ are localizing martingales.

Proof. We can choose functions F 2([0,a] U), k N, a 0,suchthatF (t, x)= k ∈Cb × ∈ ≥ k F (t, x) for t [0,a] and x in a neighbourhood of U .Thenfort a, ∈ k ≤ t Tk ∂Fk Mt = Mt Tk = Fk(t, Xt Tk ) + Fk (s, Xs Tk ) ds. ∧ ∧ − ˆ0 ∂t L ∧ 6 7 By (2.8), the right hand side is a bounded martingale.

~~Lévy characterization of weak solutions~~

~~Lévy’s characterization of Brownian motion can be extended to solutions of stochastic differential equations of type~~

~~dXt = b(t, Xt) dt + σ(t, Xt) dBt (2.9) driven by a d-dimensional Brownian motion (Bt).Asaconsequence,onecanshow that a process is a weak solution of (2.9) if and only if it solves the corresponding~~

Stochastic Analysis Andreas Eberle 2.1. LÉVY CHARACTERIZATIONS AND MARTINGALE PROBLEMS 63 martingale problem. As above, we assume that the coefﬁcients b : R Rd Rd and + × → d d d σ : R R R × are measurable and locally bounded, and we set + × → 1 d ∂2 d ∂ = aij(t, x) + bi(t, x) (2.10) L 2 ∂xi∂xj ∂xi i,j=1 i=1 " " where a(t, x)=σ(t, x)σ(t, x)T .

Theorem 2.4 (Weak solutions and the martingale problem). If the matrix σ(t, x) is 1 invertible for any t and x,and(t, x) σ(t, x)− is a locally bounded function on #→ R Rd,thenthefollowingstatementsareequivalent: + × (i) (X ) is a weak solution of (2.9) on the setup (Ω, ,P,( ), (B )). t A Ft t (ii) The processes M i := Xi Xi t bi(s, X ) ds, 1 i d,arecontinuouslocal t t − 0 − 0 s ≤ ≤ ( P ) martingales with covariations´ Ft t i j ij [M ,M ]t = a (s, Xs) ds P -a.s. for any t 0. (2.11) ˆ0 ≥

(iii) The processes M [f] := f(X ) f(X ) t( f)(s, X ) ds, f C2(Rd),are t t − 0 − 0 L s ∈ continuous local ( P ) martingales. ´ Ft (iv) The processes Mˆ [f] := f(t, X ) f(0,X ) t ∂f + f)(s, X ) ds, t t − 0 − 0 ∂t L s f C2(R Rd ,arecontinuouslocal( P )´martingales. ∈ + × Ft $ Proof. (i) (iv) is a consequence% of the Itô-Doeblin formula, cf. Theorem2.3above. ⇒ (iv) (iii) trivially holds. ⇒ (iii) (ii) follows by choosing for f polynomials of degree 2.Indeed,forf(x)=xi, ⇒ ≥ we obtain f = bi,hence L t i i 0 i [f] Mt = Xt Xt b (s, Xs) ds = Mt (2.12) − − ˆ0 is a local martingale by (iii). Moreover, if f(x)=xixj then f = aij + xibj + xjbi by L the symmetry of a,andhence

~~t i j i j [f] ij i j j i Xt Xt X0X0 = Mt + a (s, Xs)+Xs b (s, Xs)+Xs b (s, Xs) ds. (2.13) − ˆ0 $ % University of Bonn Summer Semester 2015 64 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~On the other hand, by the product rule and (2.12),~~

t t i j i j i j j i i j Xt Xt X0X0 = Xs dXs + Xs dXs +[X ,X ]t (2.14) − ˆ0 ˆ0 t i j j i i j = Nt + Xs b (s, Xs)+Xs b (s, Xs) ds +[X ,X ]t ˆ0 $ % with a continuous local martingale N.Comparing(2.13)and(2.14)weobtain

~~t i j i j ij [M ,M ]t =[X ,X ]t = a (s, Xs) ds ˆ0 since a continuous local martingale of ﬁnite variation is constant.~~

~~(ii) (i) is a consequence of Lévy’s characterization of Brownian motion: If (ii) holds ⇒ then~~

~~dXt = dMt + b(t, Xt) dt = σ(t, Xt) dBt + b(t, Xt) dt~~

1 d t 1 where Mt = Mt ,...,Mt and Bt := 0 σ(s, Xs)− dMs are continuous local martin- Rd 1 ´ gales with values$ in because% σ− is locally bounded. To identify B as a Brownian motion it sufﬁces to note that

~~t k l 1 1 i j [B ,B ] = σ− σ− (s, X ) d[M ,M ] t ˆ ki lj s 0 i,j t " $ % 1 1 T = σ− a(σ− ) kl (s, Xs) ds = δkl t ˆ0 $ % for any k, l =1,...,dby (2.11).~~

Remark (Degenerate case). If σ(t, x) is degenerate then a corresponding assertion still holds. However, in this case the Brownian motion (Bt) only exists on an extension of the probability space (Ω, ,P,( )).Thereasonisthatinthedegeneratecase,the A Ft Brownian motion can not be recovered directly from the solution (Xt) as in the proof above, see [38] for details.

The martingale problem formulation of weak solutions is powerful in many respects: It is stable under weak convergence and therefore well suitedforapproximationargu- ments, it carries over to more general state spaces (including for example Riemannian manifolds, Banach spaces, spaces of measures), and, of course, it provides a direct link

Stochastic Analysis Andreas Eberle 2.2. RANDOM TIME CHANGE 65 to the theory of Markov processes. Do not miss to have a look at the classics by Stroock and Varadhan [40] and by Ethier and Kurtz [16] for much more on the martingale problem and its applications to Markov processes.

~~2.2 Random time change~~

Random time change is already central to the work of Doeblin from 1940 that has been discovered only recently [3]. Independently, Dambis and Dubins-Schwarz have developed a theory of random time changes for semimartingales in the 1960s [25], [37]. In this section we study random time changes with a focus on applications to SDE, in particular, but not exclusively, in dimension one. Throughout this section we ﬁx a right-continuous ﬁltration ( ) such that = P Ft Ft F for any t 0.Right-continuityisrequiredtoensurethatthetimetransformation is ≥ given by ( ) stopping times. Ft

~~Continuous local martingales as time-changed Brownian motions~~

Let (Mt)t 0 be a continuous local ( t) martingale w.r.t. the underlying probability mea- ≥ F sure P such that M0 =0.OuraimistoshowthatMt can be represented as B[M]t with aone-dimensionalBrownianmotion(Ba).Forthispurpose,weconsidertherandom time substitution a T where T = inf u :[M] >a is the ﬁrst passage time to the #→ a a { u } level u.Notethata T is the right inverse of the quadratic variation t [M] ,i.e., #→ a #→ t [M] = a on T < , and, TA { a ∞} T = inf u :[M] > [M] =supu :[M] =[M] [M]t { u t} { u t} 1 by continuity of [M].If[M] is strictly increasing then T =[M]− .Byright-continuity of ( ), T is an ( ) stopping time for any a 0. Ft a Ft ≥ Theorem 2.5 (Dambis, Dubins-Schwarz). If M is a continuous local ( ) martingale Ft with [M] = almost surely then the time-changed process Ba := MTa , a 0,isan ∞ ∞ ≥ ( ) Brownian motion, and FTa M = B for any t 0, almost surely. (2.15) t [M]t ≥

~~University of Bonn Summer Semester 2015 66 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~The proof is again based on Lévy’s characterization.~~

~~Proof. 1) We ﬁrst note that B[M]t = Mt almost surely. Indeed, by deﬁnition, B[M]t = M .ItremainstoverifythatM is almost surely constant on the interval T[M]t~~

~~[t, T[M]t ].Thisholdstruesincethequadraticvariation[M] is constant on this interval, cf. the exercise below.~~

~~2) Next, we verify that Ba = MTa is almost surely continuous. Right-continuity holds since M and T are both right-continuous. To prove left-continuity note that for a>0,~~

~~lim MTa−ε = MTa− for any a 0 ε 0 ↓ ≥~~

~~by continuity of M.ItremainstoshowMTa = MTa almost surely. This again − holds true by the exercise below, because Ta and Ta are stopping times, and −~~

~~[M]Ta− = lim [M]Ta−ε = lim (a ε)=a =[M]Ta ε 0 ε 0 ↓ ↓ − by continuity of [M].~~

3) We now show that (B ) is a square-integrable ( ) martingale. Since the random a FTa variables T are ( ) stopping times, (B ) is ( ) adapted. Moreover, for any a, a Ft a FTa Ta the stopped process Mt = Mt Ta is a continuous local martingale with ∧

~~Ta E [M ] = E [M]Ta = a< . ∞ ∞ Hence M Ta is in#M 2 [0, ' ] ,and # ' c ∞ 2 $ 2 % Ta 2 E[Ba]=E[MTa ]=E[(M ) ]=a for any a 0. ∞ ≥~~

~~This shows that (Ba) is square-integrable, and, moreover,~~

E[B ]=E[M ]=M = B for any 0 r a a|FTr Ta |FTr Tr r ≤ ≤ by the Optional Sampling Theorem applied to M Ta . Finally, we note that [B] = B = a almost surely. Indeed, by the Optional Sampling a ⟨ ⟩a Theorem applied to the martingale (M Ta )2 [M Ta ],wehave − E B2 B2 = E M 2 M 2 a − r |FTr Ta − Tr |FTr # ' = E#[M]T [M]T 'T = a r for 0 r a. a − r |F r − ≤ ≤ # ' Stochastic Analysis Andreas Eberle 2.2. RANDOM TIME CHANGE 67

Hence B2 a is a martingale, and thus by continuity, [B] = B = a almost surely. a − a ⟨ ⟩a We have shown that (B ) is a continuous square-integrable ( ) martingale with a FTa [B]a = a almost surely. Hence B is a Brownian motion by Lévy’s characterization.

Remark. The assumption [M] = in Theorem 2.5 ensures Ta < almost surely. ∞ ∞ ∞ If the assumption is violated then M can still be represented in the form (2.15) with a Brownian motion B.However,inthiscase,B is only deﬁned on an extended probability space and can not be obtained as a time-change of M for all times, cf. e.g. [37].

Exercise. Let M be a continuous local ( ) martingale, and let S and T be ( ) stop- Ft Ft ping times such that S T .Provethatif[M] =[M] < almost surely, then M ≤ S T ∞ is almost surely constant on the stochastic interval [S, T ].Usethisfacttocompletethe missing step in the proof above.

~~We now consider several applications of Theorem 2.5. Let (Wt)t 0 be a Brownian ≥ motion with values in Rd w.r.t. the underlying probability measure P .~~

~~Time-change representations of stochastic integrals~~

By Theorem 2.5 and the remark below the theorem, stochastic integrals w.r.t. Brownian motions are time-changed Brownian motions. For any integrand G 2 (R , Rd), ∈La,loc + there exists a one-dimensional Brownian motion B,possiblydeﬁnedonanenlarged probability space, such that almost surely,

Gs dWs = B t G 2 ds for any t 0. ˆ 0 s 0 · ´ | | ≥ Example (Gaussian martingales). If G is a deterministic function then the stochastic integral is a Gaussian process that is obtained from the Brownian motion B by a deterministic time substitution. This case has already been studied in Section 8.3 in [14].

~~Doeblin [3] has developed a stochastic calculus based on timesubstitutionsinsteadof Itô integrals. For example, an SDE in R1 of type~~

~~t t Xt X0 = σ(s, Xs) dWs + b(s, Xs) ds − ˆ0 ˆ0~~

~~University of Bonn Summer Semester 2015 68 CHAPTER 2. TRANSFORMATIONS OF SDE can be rephrased in the form~~

Xt X0 = B t σ(s,X )2 ds + b(s, Xs) ds 0 s ˆ − ´ 0 with a Brownian motion B.Theone-dimensionalItô-Doeblinformulathentakesthe form t ∂f f(t, Xt) f(0,X0)=B t σ(s,X )2 f ′(s,X )2 ds + + f (s, Xs) ds − 0 s s ˆ ∂s L ´ 0 - . 1 2 with f = σ f ′′ + bf ′. L 2

~~Time substitution in stochastic differential equations~~

~~To see how time substitution can be used to construct weak solutions, we consider at ﬁrst an SDE of type~~

~~dYt = σ(Yt) dBt (2.16) in R1 where σ : R (0, ) is a strictly positive continuous function. If Y is a weak → ∞ solution then by Theorem 2.5 and the remark below,~~

t 2 Yt = XAt with At =[Y ]t = σ(Yr) dr (2.17) ˆ0 and a Brownian motion X.NotethatA depends on Y ,soatfirstglace(2.17)seemsnot 1 to be useful for solving the SDE (2.16). However, the inverse time substitution T = A− satisfies 1 1 1 T ′ = = = , A T σ(Y T )2 σ(X)2 ′ ◦ ◦ and hence a 1 Ta = 2 du. ˆ0 σ(Xu) Therefore, we can construct a weak solution Y of (2.16) from a given Brownian motion 1 X by first computing T ,thentheinversefunctionA = T − ,andfinallysettingY = X A.Moregenerally,thefollowingresultholds: ◦ Theorem 2.6. Suppose that (X ) on (Ω, ,P,( )) is a weak solution of an SDE of the a A Ft form

~~dXa = σ(Xa) dBa + b(Xa) da (2.18)~~

~~Stochastic Analysis Andreas Eberle 2.2. RANDOM TIME CHANGE 69~~

d d d d d with locally bounded measurable coefﬁcients b : R R and σ : R R × such → → 1 d that σ(x) is invertible for almost all x,andσ− is again locally bounded. Let ϱ : R → (0, ) be a measurable function such that almost surely, ∞ a Ta := ϱ(Xu) du < a (0, ), and T = . (2.19) ˆ0 ∞∀∈ ∞ ∞ ∞ Then the time-changed process deﬁned by

~~1 Yt := XAt ,A:= T − , is a weak solution of the SDE σ b dYt = (Yt) dBt + (Yt) dt. (2.20) √ϱ ϱ - . - . We only give a sketch of the proof of the theorem:~~

Proof of 2.6. (Sketch). The process X is a solution of the martingale problem for the 2 operator = 1 a (x) ∂ + b(x) where a = σσT ,i.e., L 2 ij ∂xi∂xj ·∇ a ! [f] Ma = f(Xa) F (X0) ( f)(Xu) du − − ˆ0 L is a local ( ) martingale for any f C2.Therefore,thetime-changedprocess Fa ∈ At [f] MAt = f(XAt f(XA0 ) ( f)(Xu) du − − ˆ0 L t

= f(Yt) f(Y0) ( f)(Yr)Ar′ dr − − ˆ0 L is a local ( ) martingale. Noting that FAt 1 1 1 Ar′ = = = , T ′(Ar) ϱ(XAr ) ϱ(Yr) we see that w.r.t. the ﬁltration ( ),theprocessY is a solution of the martingale FAt problem for the operator 1 1 a ∂2 b = = ij + . L ϱ L 2 ϱ ∂xi∂xj ϱ ·∇ i,j " ) a σ σT Since ϱ = ϱ ϱ ,thisimpliesthatY is a weak solution of (2.20).

~~University of Bonn Summer Semester 2015 70 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~In particular, the theorem shows that if X is a Brownian motion and condition (2.19) 1/2 holds then the time-changed process Y solves the SDE dY = ϱ(Y )− dB.~~

~~Example (Non-uniqueness of weak solutions). Consider the one-dimensional SDE~~

dY = Y α dB ,Y=0, (2.21) t | t| t 0 with a one-dimensional Brownian motion (Bt) and α>0.Ifα<1/2 and x is a a Brownian motion with X0 =0then the time-change Ta = 0 ϱ(Xu) du with ϱ(x)= 2α ´ x − satisﬁes | | a a 2α E[Ta]=E ϱ(Xu) du = E[ Xu − ] du ˆ0 ˆ0 | | / a 0 2α α = E[ X1 − ] u− du < | | · ˆ0 ∞

1 for any a (0, ).Hence(2.19)holds,andthereforetheprocessY = X , A = T − , ∈ ∞ t At is a non-trivial weak solution of (2.21). On the other hand, Y 0 is also a weak t ≡ solution. Hence for α<1/2,uniquenessindistributionofweaksolutionsfails.For α 1/2,thetheoremisnotapplicablesinceAssumption(2.19)isviolated. One can ≥ prove that in this case indeed, the trivial solution Y 0 is the unique weak solution. t ≡

~~Exercise (Brownian motion on the unit sphere). Let Yt = Bt/ Bt where (Bt)t 0 is a | | ≥ Brownian motion in Rn, n>2.Provethatthetime-changedprocess~~

~~t 1 2 Za = YTa ,T= A− with At = Bs − ds , ˆ0 | |~~

n 1 n is a diffusion taking values in the unit sphere S − = x R : x =1 with generator { ∈ | | } 2 1 ∂ f n 1 ∂f n 1 f(x)= ∆f(x) x x (x) − x (x),xS − . L 2 − i j ∂x ∂x − 2 i ∂x ∈ 9 i,j i j : i i " " One-dimensional SDE

~~By combining scale and time transformations, one can carry out a rather complete study of weak solutions for non-degenerate SDE of the form~~

~~dXt = σ(Xt) dBt + b(Xt) dt, X0 = x0, (2.22)~~

~~Stochastic Analysis Andreas Eberle 2.2. RANDOM TIME CHANGE 71~~

on a real interval (α,β).WeassumethattheinitialvalueX0 is contained in (α,β),and b, σ :(α,β) R are continuous functions such that σ(x) > 0 for any x (α,β).We → ∈ ﬁrst simplify (2.22) by a coordinate transformation Yt = s(Xt) where

~~s :(α,β) s(α),s(β) →~~

~~2 $ % is C and satisﬁes s′(x) > 0 for all x.Thescalefunction~~

z y 2b(x) s(z) := exp 2 dx dy ˆx − ˆx σ(x) 0 6 0 7 1 2 has these properties and satisﬁes 2 σ s′′ + bs′ =0.HencebytheItô-Doeblinformula, the transformed process Yt = s(Xt) is a local martingale satisfying

~~dYt =(σs′)(Xt) dBt, i.e., Y is a solution of the equation~~

~~dYt = σ(Yt) dBt,Y0 = s(x0), (2.23)~~

~~1 where σ := (σs′) s− .TheSDE(2.23)istheoriginalSDEin“naturalscale”.Itcan) ◦ be solved explicitly by a time change. By combining scale transformations and time change) one obtains:~~

~~Theorem 2.7. The following statements are equivalent:~~

~~(i) The process (X ) on the setup (Ω, ,P,( ), (B )) is a weak solution of (2.22) t t<ζ A Ft t deﬁned up to a stopping time ζ.~~

~~(ii) The process Yt = s(Xt), t<ζ,onthesamesetupisaweaksolutionof(2.23)up to ζ.~~

(iii) The process (Yt)s<ζ has a representation of the form Yt = BAt ,whereBt is a 1 one-dimensional Brownian motion satisfying B0 = s(x0) and A = T − with ) ) r ) 2 Tr = ϱ Bu du, ϱ(y)=1/σ(y) . ˆ0 $ % ) ) University of Bonn Summer Semester 2015 72 CHAPTER 2. TRANSFORMATIONS OF SDE

Carrying out the details of the proof is left as an exercise. The measure m(dy) := ϱ(y) dy is called the “speed measure” of the process Y although Y is moving faster if m is small. The generator of Y can be written in the form = 1 d d ,andthe L 2 dm dy generator of X is obtained from by coordinate transformation. For a much more L detailed discussion of one dimensional diffusions we refer to Section V.7 in [38]. Here we only note that 2.7 immediately implies existence and uniqueness of a maximal weak solution of (2.22):

~~Corollary 2.8. Under the regularity and non-degeneracy conditions on σ and b imposed above there exists a weak solution X of (2.22) deﬁned up to the ﬁrst exit time~~

ζ = inf t 0 : lim Xt a, b ≥ s t ∈{ } ? ↑ @ from the interval (α,β).Moreover,thedistributionofanytwoweaksolutions(Xt)t<ζ ¯ R and (Xt)t<ζ¯ on u>0 C([0,u), ) coincide. Remark. We> have already seen above that uniqueness may fail if σ is degenerate. For example, the solution of the equationdY = Y α dB , Y =0,isnotuniquein t | t| t 0 distribution for α (0, 1/2). ∈

Example (Bessel SDE). Suppose that (Rt)t<ζ is a maximal weak solution of the Bessel equation d 1 1 dRt = dWt + − dt, W BM(R ), 2Rt ∼ on the interval (α,β)=(0, ) with initial condition R = r (0, ) and the pa- ∞ 0 0 ∈ ∞ 1 d 1 rameter d R.TheODE s = s′′ + − s′ =0for the scale function has a strictly ∈ L 2 2r increasing solution 1 2 d 2 d r − for d =2, s(r)= − ̸ ⎧ ⎨log r for d =2 (More generally, cs + d is a strictly increasing solution for any c>0 and d R). ⎩ ∈ Note that s is one-to-one from the interval (0, ) onto ∞ (0, ) for d<2, ∞ (s(0),s( )) = ⎧( , ) for d =2, ∞ ⎪ ⎪ −∞ ∞ ⎨( , 0) for d>2. −∞ ⎪ ⎪ Stochastic Analysis ⎩ Andreas Eberle 2.3. CHANGE OF MEASURE 73

~~By applying the scale transformation, we see that s(r ) s(a) P T R~~

~~1 for d 2, ≤ R R ⎧ P lim inf Rt =0 = P Ta 2, ⎪ ⎪ ⎩ 1 for d 2, ≥ R R ⎧ P lim sup Rt = = P Tb~~

~~1 ds(R)=σ s(R) dW with σ(y)=s′ s− (y) .~~

$ % y $ % 2 2y In the critical case d =2, s)(r) = log r, σ(y)=e− ,andhence) ϱ(y)=σ(y)− = e . 2y Thus the speed measure is m(dy)=e dy,andlog Rt = BT −1(t),i.e., ) a ) Rt =expBT −1(t) with Ta = ) exp 2Bu du ˆ0 $ % $ % and a one-dimensional Brownian) motion B. )

~~) 2.3 Change of measure~~

~~In Section 2.3, 2.4 and 2.5 we study connections between two different ways of transforming a stochastic process (Y,P):~~

~~1) Random transformations of the paths: For instance, mapping a Brownian motion~~

~~(Yt) to the solution (Xt) of s stochastic differential equation of type~~

~~dXt = b(t, Xt) dt + dYt (2.24)~~

~~University of Bonn Summer Semester 2015 74 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~corresponds to a random translation of the paths of (Yt):~~

~~t Xt(ω)=Yt(ω)+Ht(ω) where Ht = b(Xs) ds. ˆ0~~

2) Change of measure: Replace the underlying probability measure P by a modified probability measure Q such that P and Q are mutually absolutely continuous on for any t [0, ). Ft ∈ ∞ In this section we focus mainly on random transformations of Brownian motions and the corresponding changes of measure. To understand which kind of results we can expect in this case, we first look briefly at a simplified situation:

~~Example (Translated Gaussian random variables in Rd). We consider the equation~~

X = b(X)+Y, Y N(0,I ) w.r.t. P, (2.25) ∼ d for random variables X, Y :Ω Rd where b : Rd Rd is a “predictable” map → → in the sense that the i-th component bi(x) depends only on the ﬁrst i 1 components − i i 1 X ,...,X − of X.Thepredictabilityensuresinparticularthatthetransformation deﬁned by (2.25) is invertible, with X1 = Y 1 + b1, X2 = Y 2 + b2(X1), X3 = Y 3 + 3 1 2 n n n 1 n 1 b (X ,X ), ... ,X = Y + b (X ,...,X − ).

Arandomvariable(X, P) is a “weak” solution of the equation (2.25) if and only if Y := X b(X) is standard normally distributed w.r.t. P ,i.e.,ifandonlyifthedistribution − 1 P X− is absolutely continuous with density ◦ ∂(x b(x)) f P (x)=f P x b(x) det − X Y − ∂x d/2 x b&(x) 2/2 & =(2π$)− e−| −% & | & & & x b(x) b(x) 2/2 d = e · −| | ϕ (x), where ϕd(x) denotes the standard normal density in Rd.Thereforewecanconclude:

(X, P) is a weak solution of (2.25) if and only if X N(0,I ) w.r.t. the unique proba- ∼ d bility measure Q on Rd satisfying P Q with ≪ dP =expX b(X) b(X) 2/2 . (2.26) dQ · −| | $ % Stochastic Analysis Andreas Eberle 2.3. CHANGE OF MEASURE 75

~~In particular, we see that the law µb of a weak solution of (2.25) is uniquely determined, and µb satisﬁes~~

b 1 1 0 µ = P X− Q X− = N(0,I )=µ ◦ ≪ ◦ d with relative density b dµ x b(x) b(x) 2/2 (x)=e · −| | dµ0 The example can be extended to Gaussian measures on Hilbert spaces and to more general transformations, leading to the Cameron-Martin Theorem (cf. Theorem 2.17 below) and Ramer’s generalization [1]. Here, we study the more concrete situation where Y and X are replaced by a Brownian motion and a solution of the SDE (2.24) respectively. We start with a general discussion about changing measure on ﬁltered probability spaces that will be useful in other contexts as well.

~~Change of measure on ﬁltered probability spaces~~

Let ( ) be a ﬁltration on a measurable space (Ω, ),andﬁxt (0, ).Weconsider Ft A 0 ∈ ∞ two probability measures P and Q on (Ω, ) that are mutually absolutely continuous A on the σ-algebra with relative density Ft0 dP Zt0 = > 0 Q-almost surely. dQ t0 &F Then P and Q are also mutually& absolutely continuous on each of the σ-algebras , & Ft t t ,withQ-andP -almost surely strictly positive relative densities ≤ 0 dP dQ 1 Zt = = EQ Zt0 t and = . dQ t F dP t Zt &F &F & # & ' & The process (Zt)t t0 is a martingale w.r.t. &Q,and,correspondingly,(1/Zt)t t0 is a mar- ≤ & & ≤ tingale w.r.t. P .Fromnowon,wealwayschooseacàdlàgversionofthesemartingales.

~~Lemma 2.9. 1) For any 0 s t t ,andforany -measurable random vari- ≤ ≤ ≤ 0 Ft able X :Ω [0, ], → ∞ E [XZ ] E [XZ ] E [X ]= Q t|Fs = Q t|Fs P -a.s. and Q-a.s. (2.27) P |Fs E [Z ] Z Q t|Fs s~~

~~University of Bonn Summer Semester 2015 76 CHAPTER 2. TRANSFORMATIONS OF SDE~~

2) Suppose that (Mt)t t0 is an ( t) adapted càdlàg stochastic process. Then ≤ F (i) M is a martingale w.r.t. P M Z is a martingale w.r.t. Q, ⇔ · (ii) M is a local martingale w.r.t. P M Z is a local martingale w.r.t. Q. ⇔ · Proof. 1) The right hand side of (2.27) is -measurable. Moreover, for any A , Fs ∈Fs E [E [XZ ]/Z ; A]=E [E [XZ ]; A] P Q t|Fs s Q Q t|Fs = EQ[XZt ; A]=EQ[X ; A].

2) (i) is a direct consequence of 1). Moreover, by symmetry, itisenoughtoprove the implication " "in(ii).HencesupposethatM Z is a local Q-martingale with ⇐ · Tn localizing sequence (Tn).WeshowthatM is a P -martingale, i.e.,

~~EP [Mt Tn ; A]=EP [Ms Tn ; A] for any A s, 0 s t t0. (2.28) ∧ ∧ ∈F ≤ ≤ ≤ To verify (2.28), we ﬁrst note that~~

EP [Mt Tn ; A Tn s ]=EP [Ms Tn ; A Tn s ] (2.29) ∧ ∩{ ≤ } ∧ ∩{ ≤ } since t T = T = s T on T s .Moreover,oneveriﬁesfromthedeﬁnitionof ∧ n n ∧ n { n ≤ } the σ-algebra s Tn that for any A s,theeventA Tn >s is contained in s Tn , F ∧ ∈F ∩{ } F ∧ and hence in t Tn .Therefore, F ∧

~~EP [Mt Tn ; A Tn >s ]=EQ[Mt Tn Zt Tn ; A Tn >s ] (2.30) ∧ ∩{ } ∧ ∧ ∩{ }~~

~~= EQ[Ms Tn Zs Tn ; A Tn >s ]] = EP [Ms Tn ; A Tn >s ] ∧ ∧ ∩{ } ∧ ∩{ } by the martingale property for (MZ)Tn ,theoptionalsamplingtheorem,andthefactthat~~

P Q on t Tn with relative density Zt Tn .(2.28)followsfrom(2.29)and(2.30). ≪ F ∧ ∧ If the probability measures P and Q are mutually absolutely continuous on the σ-algebra dP t,thentheQ-martingale Zt = dQ of relative densities is actually an exponential F t martingale. Indeed, to obtain a corresponding&F representation let & & t 1 Lt := dZs ˆ0 Zs − Stochastic Analysis Andreas Eberle 2.3. CHANGE OF MEASURE 77 denote the stochastic "logarithm" of Z.Hereweareusingstochasticcalculusfor càdlàg semimartingales, cf. Chapter 5 below. This can be avoided if one assumes that

Q-almost surely, t Zt is continuous, i.e., Zt = Zt for t 0.Inanycase,the #→ − ≥ process (Lt)t t0 is a well-deﬁned local martingale w.r.t. Q since Q-a.s., (Zt) is càdlàg ≤ and strictly positive. Moreover, by the associative law,

~~dZt = Zt dLt,Z0 =1, − so Zt is the stochastic exponential of the local Q-martingale (Lt):~~

~~Z = L. t Et~~

~~In particular, if (Zt) is continuous then~~

~~Lt [L]t/2 Zt = e − .~~

~~Girsanov’s Theorem~~

We now return to our original problem of identifying the change of measure induced by a random translation of the paths of a Brownian motion. Suppose that (Xt) is a d Brownian motion in R with X0 =0w.r.t. the probability measure Q and the ﬁltration ( ),andﬁxt [0, ).Let Ft 0 ∈ ∞ t Lt = Gs dXs,t 0, ˆ0 · ≥ 2 R Rd t 2 with G a,loc +, .Then[L]t = 0 Gs ds,andhence ∈L ´ | | $ % t t 1 2 Zt =exp Gs dXs Gs ds (2.31) ˆ0 · − 2 ˆ0 | | 6 7 is the exponential of L.Inparticular,sinceL is a local martingale w.r.t. Q, Z is a non- negative local martingale, and hence a supermartingale w.r.t. Q.ItisaQ-martingale for t t if and only if E [Z ]=1: ≤ 0 Q t0

~~Exercise (Martingale property for exponentials). Let (Zt)t [0,t0] on (Ω, ,Q) be a ∈ A non-negative local martingale satisfying Z0 =1.~~

~~University of Bonn Summer Semester 2015 78 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~a) Show that Z is a supermartingale.~~

~~b) Prove that Z is a martingale if and only if EQ[Zt0 ]=1. In order to use Z for changing the underlying probability measure on we have to t0 Ft0 assume the martingale property:~~

Assumption. (Zt)t t0 is a martingale w.r.t. Q. ≤ Theorem 2.11 below states that the assumption is satisﬁed if E exp 1 t G 2 ds < 2 0 | s| .Iftheassumptionholdsthenwecanconsideraprobabilitym/ easure6 ´P on with70 ∞ A dP = Zt0 Q-a.s. (2.32) dQ t0 &F Note that P and Q are mutually& absolutely continuous on for any t t with & Ft ≤ 0 dP dQ 1 = Zt and = dQ t dP t Zt &F &F both P -andQ-almost& surely. We are now ready to& prove one of the most important & & results of stochastic analysis:

Theorem 2.10 (Maruyama 1954, Girsanov 1960). Suppose that X is a d-dimensional 2 R Rd Brownian motion w.r.t. Q and (Zt)t t0 is defined by (2.31) with G a,loc( +, ).If ≤ ∈L EQ[Zt0 ]=1then the process t Bt := Xt Gs ds, t t0, − ˆ0 ≤ is a Brownian motion w.r.t. any probability measure P on satisfying (2.32). A Rd Proof. By Lévy’s characterization, it suffices to show that (Bt)t t0 is an -valued P - ≤ martingale with [Bi,Bj] = δ tP-almost surely for any i, j 1,...,d .Furthermore, t ij ∈{ } by Lemma 2.9, and since P and Q are mutually absolutely continuous on ,thisholds Ft0 i j true provided (BtZt)t t0 is a Q-martingale and [B ,B ]=δijtQ-almost surely. The ≤ identity for the covariations holds since (Bt) differs from the Q-Brownian motion (Xt) only by a continuous finite variation process. To show that B Z is a local Q-martingale, · we apply Itô’s formula: For 1 i d, ≤ ≤ d(Bi Z)=Bi dZ + ZdBi + d[Bi,Z] (2.33) = BiZG dX + ZdXi ZGidt + ZGi dt, · −

~~Stochastic Analysis Andreas Eberle 2.3. CHANGE OF MEASURE 79 where we have used that~~

~~d[Bi,Z]=ZG d[Bi,X]=ZGi dt Q-almost surely. · The right-hand side of (2.33) is a stochastic integral w.r.t.theQ-Brownian motion X, and hence a local Q-martingale.~~

~~The theorem shows that if X is a Brownian motion w.r.t. Q,andZ deﬁned by (2.31) is a Q-martingale, then X satisﬁes~~

~~dXt = Gt dt + dBt. with a P -Brownian motion B.ItgeneralizestheCameron-MartinTheoremtonon- deterministic adapted translation~~

~~t Xt(ω) Xt(ω) Ht(ω),Ht = Gs ds, −→ − ˆ0 of a Brownian motion X.~~

~~Remark (Assumptions in Girsanov’s Theorem). t 1) Absolute continuity and adaptedness of the “translation process” Ht = 0 Gs ds are essential for the assertion of Theorem 2.10. ´~~

2) The assumption EQ[Zt0 ]=1ensuring that (Zt)t t0 is a Q-martingale is not always ≤ satisfied asufficientconditionisgiveninTheorem2.11below.If(Z ) is not a martin- − t gale w.r.t. Q it can still be used to define a positive measure Pt with density Zt w.r.t. Q on each σ-algebra .However,inthiscase,P [Ω] < 1.Thesub-probabilitymeasures Ft t Pt correspond to a transformed process with finite life-time.

~~Novikov’s condition~~

~~To verify the assumption in Girsanov’s theorem, we now deriveasufﬁcientcondition for ensuring that the exponential~~

Z =expL 1/2[L] t t − t $ % of a continuous local ( ) martingale is a martingale. Recall that Z is always a non- Ft negative local martingale, and hence a supermartingale w.r.t. ( ). Ft University of Bonn Summer Semester 2015 80 CHAPTER 2. TRANSFORMATIONS OF SDE

~~R Theorem 2.11 (Novikov 1971). Let t0 +.IfE[exp [L]t0 /2 ] < then (Zt)t t0 is ∈ ∞ ≤ an ( ) martingale. Ft $ % We only prove the theorem under the slightly more restrictivecondition~~

E [exp(p[L] /2)] < for some p>1. (2.34) t ∞ This simpliﬁes the proof considerably, and the condition is sufﬁcient for many applications. For a proof in the general case and under even weaker assumptions see e.g. [37].

Proof. Let (Tn)n N be a localizing sequence for the martingale Z.Then(Zt Tn )t 0 is a ∈ ∧ ≥ martingale for any n.Tocarryoverthemartingalepropertytotheprocess(Zt)t [0,t0],it ∈ N is enough to show that the random variables Zt Tn , n ,areuniformlyintegrablefor ∧ ∈ 1 1 each ﬁxed t t .However,forc>0 and p, q (1, ) with p− + q− =1,wehave ≤ 0 ∈ ∞

E[Zt Tn ; Zt Tn c] ∧ ∧ ≥ p p 1 = E exp Lt Tn [L]t Tn exp − [L]t Tn ; Zt Tn c (2.35) ∧ − 2 ∧ 2 ∧ ∧ ≥ 2 # $ p % 1/p$ %p 1 ' 1/q E exp pLt Tn [L]t Tn E exp q − [L]t Tn ; Zt Tn c ≤ ∧ − 2 ∧ · · 2 ∧ ∧ ≥ # $p %'1/q # $ % ' E exp [L]t ; Zt Tn c ≤ 2 ∧ ≥ # N $ % ' for any n .HerewehaveusedHölder’sinequalityandthefactthatexp pLt Tn ∧ p2 ∈ p − [L]t Tn is an exponential supermartingale. If exp [L]t is integrable then the right 2 ∧ 2 $ hand side of (2.35) converges to 0 uniformly in n as c ,because % $ →∞% 1 1 P [Zt Tn 0] c− E[Zt Tn ] c− 0 ∧ ≥ ≤ ∧ ≤ −→ N uniformly in n as c .Hence Zt Tn : n is indeed uniformly integrable, and →∞ { ∧ ∈ } thus (Zt)t [0,t0] is a martingale. ∈ t Example (Bounded drifts). If Lt = 0 Gs dXs with a Brownian motion (Xt) and ´ · an adapted process (Gt) that is uniformly bounded on [0,t] for any ﬁnite t then the quadratic variation [L] = t G 2 ds is also bounded for ﬁnite t.Henceexp(L 1 [L]) t 0 | s| − 2 is an ( ) martingale for t´ [0, ). Ft ∈ ∞ AmorepowerfulapplicationofNovikov’scriterionisconsidered in the beginning of Section 2.4.

~~Stochastic Analysis Andreas Eberle 2.3. CHANGE OF MEASURE 81~~

~~Applications to SDE~~

~~The Girsanov transformation can be used to construct weak solutions of stochastic differential equations. For example, consider an SDE~~

dX = b(t, X ) dt + dB ,X= o, B BM(Rd), (2.36) t t t 0 ∼ where b : R Rd Rd is continuous, and o Rd is a fixed initial value. If the drift + × → ∈ coefficient is not growing too strongly as x ,thenwecanconstructaweaksolu- | |→∞ tion of (2.36) from Brownian motion by a change of measure. To this end let (X, Q) be an ( ) Brownian motion with X = oQ-almost surely, and suppose that the following Ft 0 assumption is satisfied: Assumption (A). The process t 1 t Z =exp b(s, X ) dX b(s, X ) 2 ds ,t0, t ˆ s · s − 2 ˆ | s | ≥ - 0 0 . is a martingale w.r.t. Q. We will see later that the assumption is always satisfied if b is bounded, or, more generally, growing at most linearly in x.If(A)holdsthenE [Z ]=1for any t 0,and,by Q t ≥ Kolmogorov’s extension theorem, there exists a probabilitymeasureP on (Ω, ) such A that dP = Zt Q-almost surely for any t 0. dQ t ≥ &F By Girsanov’s Theorem,& the process & t Bt = Xt b(s, Xs) ds, t 0, − ˆ0 ≥ is a Brownian motion w.r.t. P ,i.e.(X, P) is a weak solution of the SDE (2.36). More generally, instead of starting from a Brownian motion, we may start from a solution (X, Q) of an SDE of the form

dXt = β(t, Xt) dt + σ(t, Xt) dWt (2.37) where W is an Rd-valued Brownian motion w.r.t. the underlying probability measure Q. We change measure via an exponential martingale of type t 1 t Z =exp b(s, X ) dW b(s, X ) 2 ds t ˆ s · s − 2 ˆ | s | - 0 0 . University of Bonn Summer Semester 2015 82 CHAPTER 2. TRANSFORMATIONS OF SDE

n n n n d where b, β : R R R and σ : R R R × are continuous functions. + × → + × → Corollary 2.12 (Drift transformations for SDE). Suppose that (X, Q) is a weak solution of (2.37). If (Zt)t 0 is a Q-martingale and P Q on t with relative density Zt ≥ ≪ F for any t 0,then(X, P) is a weak solution of ≥ dX =(β + σb)(t, X ) dt + σ(t, X ) dB ,BBM(Rd). (2.38) t t t t ∼ Proof. By (2.37), the equation (2.38) holds with

~~t Bt = Wt b(s, Xs) ds. − ˆ0 Girsanov’s Theorem implies that B is a Brownian motion w.r.t. P .~~

~~Note that the Girsanov transformation induces a corresponding transformation for the martingale problem:If(X, Q) solves the martingale problem for the operator~~

~~1 ∂2 = aij + β ,a= σσT , (2.39) L 2 ∂xi∂xj ·∇ i,j " then (X, P) is a solution of the martingale problem for~~

~~= +(σb) = + b σT . L L ·∇ L · ∇ This “Girsanov transformation) for martingale problems” carries over to diffusion processes with more general state spaces than Rn.~~

~~Doob’s h-transform~~

The h-transform is a change of measure involving a space-time harmonic function that applies to general Markov processes. In the case of Itô diffusions, it turns out to be a special case of the drift transform studied above. Indeed, suppose that h C1,2(R ∈ + × Rn) is a strictly positive space-time harmonic function for the generator (2.39) of the Itô diffusion (X, Q),normalizedsuchthath(0,o)=1:

~~∂h + h =0,h(0,o)=1. (2.40) ∂t L Stochastic Analysis Andreas Eberle 2.4. PATH INTEGRALS AND BRIDGES 83~~

~~Then, by Itô’s formula, the process~~

Z = h(t, X ),t0, t t ≥ is a positive local Q-martingale satisfying Z0 =1Q-almost surely. We can therefore try to change the measure via (Zt).Tounderstandtheeffectofsuchatransformation, we write Zt in exponential form. By the Itô-Doeblin formula and (2.40),

dZ =(σT h)(t, X ) dW . t ∇ t · t Hence Z =exp(L 1 [L] ) where t t − 2 t t t 1 T Lt = dZs = (σ log h)(s, Xs) dWs ˆ0 Zs ˆ0 ∇ · is the stochastic logarithm of Z.Thusif(Z, Q) is a martingale, and P Q with local ≪ dP T densities = Zt then (X, P) solves the SDE (2.37) with b = σ log h,i.e., dQ t F ∇ & dX =(β +& σσT log h)(t, X ) dt + σ(t, X ) dB ,B BM(Rd) w.r.t. P. (2.41) t ∇ t t t ∼ The proces (X, P) is called the h-transform of (X, Q).

~~Example. If Xt = Wt is a Brownian motion w.r.t. Q then~~

dX = log h(t, X ) dt + dB ,BBM(Rd) w.r.t. P. t ∇ t t ∼ For example, choosing h(t, x)=exp(α x 1 α 2t),α Rd, (X, P) is a Brownian · − 2 | | ∈ motion with constant drift α,i.e.,dXt = αdt + dBt.

~~2.4 Path integrals and bridges~~

One way of thinking about a stochastic process is to interpretitasaprobabilitymea- sure on path space. This useful point of view will be developedfurtherinthisandthe following section. We consider an SDE

dW = b(W ) dt + dB ,W= o, B BM(Rd) (2.42) t t t 0 ∼ University of Bonn Summer Semester 2015 84 CHAPTER 2. TRANSFORMATIONS OF SDE with initial condition o Rd and b C(Rd, Rd).Wewillshowthatthesolutioncon- ∈ ∈ structed by Girsanov transformation is a Markov process, andwewillstudyitstransition function, as well as the bridge process obtained by conditioning on a given value at a ﬁxed time. W d Let µo denote the law of Brownian motion starting at o on (Ω, ) where Ω=C(R+, R ) F∞ and Wt(x)=xt is the canonical Brownian motion on (Ω,µo).Let t 1 t Z =exp b(W ) dW b(W ) 2 ds . (2.43) t ˆ s · s − 2 ˆ | s | - 0 0 . Note that if b(x)= H(x) for a function H C2(Rd) then by Itô’s formula, −∇ ∈ 1 t Z =exp H(W ) H(W )+ ∆H H 2 (W ) ds . (2.44) t 0 − t 2 ˆ −|∇ | s - 0 . $ % This shows that Z is more robust w.r.t. variations of (Wt) if b is a gradient vector ﬁeld, because (2.44) does not involve a stochastic integral. This robustness is crucial for certain applications, see the example below. Similarly as above, we assume:

~~Assumption (A).Theexponential(Zt)t 0 is a martingale w.r.t. µo. ≥ We note that by Novikov’s criterion, the assumption always holds if~~

~~b(x) c (1 + x ) for some ﬁnite constant c>0: (2.45) | |≤ · | |~~

Exercise (Martingale property for exponentials). Prove that (Zt) is a martingale if ε 2 (2.45) holds. Hint: Prove ﬁrst that E[exp 0 b(Ws) ds] < for ε>0 sufﬁciently ´ | | ∞ small, and conclude that E[Zε]=1.ThenshowbyinductionthatE[Zkε]=1for any k N. ∈ If (A) holds then by the Kolmogorov extension theorem, there exists a probability mea- b W b sure µo on such that µo and µo are mutually absolutely continuous on each of the F∞ σ-algebras W , t [0, ),withrelativedensities Ft ∈ ∞ b dµo = Zt µo-a.s. W dµo t &F & Girsanov’s Theorem implies: & Corollary 2.13. Suppose that (A) holds. Then:

~~Stochastic Analysis Andreas Eberle 2.4. PATH INTEGRALS AND BRIDGES 85~~

~~b 1) The process (W, µo) is a weak solution of (2.36).~~

2) For any t [0, ),thelawof(W, µb) is absolutely continuous w.r.t. Wiener ∈ ∞ o measure µ on W with relative density Z . o Ft t The ﬁrst assertion follows since B = W t b(W ) ds is a Brownian motion w.r.t. µb, t t − 0 s o b ´ 1 b and the second assertion holds since µ W − = µ . o ◦ o

~~Path integral representation~~

b Corollary 2.13 yields a rigorous path integral representation for the solution (W, µo) of b,t Rd b the SDE (2.36): If µo denotes the law of (Ws)s t on C [0,t], w.r.t. µo then ≤ t 1 t $ % µb,t(dx)=exp b(x ) dx b(x ) 2 ds µ0,t(dx). (2.46) o ˆ s · s − 2 ˆ | s | o - 0 0 . By combining (2.46) with the heuristic path integral representation

~~t 0,t 1 1 2 “ µ (dx)= exp x′ ds δ (dx ) dx ” o −2 ˆ | s| 0 0 s - 0 . 0~~

~~t b,t 1 1 2 “ µ (dx)= exp x′ b(x ) ds δ (dx ) dx ”(2.47) o −2 ˆ | s − s | 0 0 s - 0 . 0~~

xs′ = b(s, xs) obtained by setting the noise term in the SDE (2.36) equal to zero. Of course, these arguments do not hold rigorously, because I(x)= for µ0,t-andµb,t-almosteveryx. ∞ o o Nevertheless, they provide an extremely valuable guidelinetoconclusionsthatcanthen be veriﬁed rigorously, for instance via (2.46).

~~University of Bonn Summer Semester 2015 86 CHAPTER 2. TRANSFORMATIONS OF SDE~~

Example (Likelihood ratio test for non-linear filtering). Suppose that we are observ- d ing a noisy signal (xt) taking values in R with x0 = o.Weinterpret(xt) as a realization of a stochastic process (Xt).Wewouldliketodecideifthereisonlynoise,orifthe signal is coming from an object moving with law of motion dx/dt = H(x) where −∇ H C2(Rd).ThenoiseismodelledbytheincrementsofaBrownianmotion(white ∈ noise). This is a simplified form of models that are used frequently in nonlinear filtering (in realistic models often the velocity or the acceleration is assumed to satisfy a similar equation). In a hypothesis test, the null hypothesis and the alternative would be

~~H0 : Xt = Bt,~~

H1 : dXt = b(Xt) dt + dBt, where (B ) is a d-dimensional Brownian motion, and b = H.Inalikelihoodratio t −∇ test based on observations up to time t,theteststatisticwouldbethelikelihoodratio b,t 0,t dµo /dµo which by (2.44) can be represented in the robust form b,t t dµo 1 2 (x)=expH(x0) H(xt)+ (∆H H )(xs) ds (2.48) dµ0,t − 2 ˆ −|∇ | o - 0 . The null hypothesis H0 would then be rejected if this quantity exceeds some given value c for the observed signal x,i.e.,if t 1 2 H(x0) H(xt)+ (∆H H )(xs) ds > log c. (2.49) − 2 ˆ0 −|∇ | Note that the robust representation of the density ensures that the estimation procedure is quite stable, because the log likelihood ratio in (2.49) is continuous w.r.t. the supremum norm on C([0,t], Rd).

~~The Markov property~~

b W Recall that if (A) holds then there exists a (unique) probability measure µo on (Ω, ) F∞ such that µb[A]=E [Z ; A] for any t 0 and A W . o o t ≥ ∈Ft Here Ex denotes expectation w.r.t. Wiener measure µx with start in x.ByGirsanov’s b Theorem, the process (W, µo) is a weak solution of (2.42). Moreover, we can easily b verify that (W, µo) is a Markov process:

~~Stochastic Analysis Andreas Eberle 2.4. PATH INTEGRALS AND BRIDGES 87~~

~~b Theorem 2.14 (Markov property). If (A) holds then (W, µo) is a time-homogeneous Markov process with transition function~~

pb(x, C)=µb [W C]=E [Z ; W C] C (Rd). t x t ∈ x t t ∈ ∀ ∈B Proof. Let 0 s t,andletf : Rd R be a non-negative measurable function. ≤ ≤ → + Then, by the Markov property for Brownian motion,

~~Eb[f(W ) W ]=E [f(W )Z W ]/Z o t |Fs o t t|Fs s t t 1 2 W = Eo f(Wt)exp b(Wr) dWr b(Wr) dr s ˆs · − 2 ˆs | | F / - . & 0 b & = EWs [f(Wt s)Zt s]=(pt sf)(Ws) − − − &~~

~~b b b µo-andµo-almost surely where Ex denotes the expectation w.r.t. µx.~~

b Remark. 1) If b is time-dependent then one veriﬁes in the same way that (W, µo) is a time-inhomogeneous Markov process. 2) It is not always easy to prove that solutions of SDE are Markov processes. If the solution is not unique then usually, there are solutions thatarenotMarkovprocesses.

~~Bridges and heat kernels~~

We now restrict ourselves to the time-interval [0, 1],i.e.,weconsiderasimilarsetup as before with Ω=C([0, 1], Rd).Notethat W is the Borel σ-algebra on the Banach F1 b space Ω.Ourgoalistoconditionthediffusionprocess(W, µo) on a given terminal value W = y, y Rd.Moreprecisely,wewillconstructaregular version y !→ µb of the 1 ∈ o,y b conditional distribution µo[·|W1 = y] in the following sense:

~~(i) For any y Rd, µb is a probability measure on (Ω),and µb [W = y]=1. ∈ o,y B o,y 1 (ii) Disintegration: For any A (Ω),thefunctiony µb [A] is measurable, and ∈B #→ o,y~~

~~b b b µo[A]= µo,y[A] p1(o, dy). ˆRd~~

~~(iii) The map y µb is continuous w.r.t. weak convergence of probability measures. #→ o,y University of Bonn Summer Semester 2015 88 CHAPTER 2. TRANSFORMATIONS OF SDE~~

Example (Brownian bridge). For b =0,aregularversiony µ of the condi- #→ o,y tional distribution µ [ W = y] w.r.t. Wiener measure µ can be obtained by linearly o ·| 1 o transforming the paths of Brownian motion, cf. Theorem 8.11 in [14]: Under µo,the process Xy := W tW + ty, 0 t 1, t t − 1 ≤ ≤ y is independent of W1 with terminal value y,andthelawµo,y of (Xt )t [0,1] w.r.t. µo is ∈ aregularversionofµ [ W = y].Themeasureµ is called “pinned Wiener mea- o ·| 1 o,y sure”. The construction of a bridge process described in the exampleonlyappliesforBrown- ian motion and other Gaussian processes. For more general diffusions, the bridge can not be constructed from the original process by a linear transformation of the paths. For perturbations of a Brownian motion by a drift, however, we canapplyGirsanov’sThe- orem to construct a bridge measure. We assume for simplicity again that b is the gradient of a C2 function:

~~b(x)= H(x) with H C2(Rd). −∇ ∈~~

Then the exponential martingale (Zt) takes the form 1 t Z =expH(W ) H(W )+ (∆H H 2)(W ) ds , t 0 − t 2 ˆ −|∇ | s - 0 . cf. (2.44). Note that the expression on the right-hand side isdeﬁnedµo,y-almost surely for any y.Therefore,(Zt) can be used for changing the measure w.r.t. the Brownian bridge.

~~Theorem 2.15 (Heat kernel and Bridge measure). Suppose that (A) holds. Then:~~

~~b 1) The measure p1(o, dy) is absolutely continuous w.r.t. d-dimensional Lebesgue measure with density~~

pb (o, y)=p (o, y) E [Z ]. 1 1 · o,y 1 2) A regular version of µb[ W = y] is given by o ·| 1 p (o, y) exp H(o) 1 1 µb (dx)= 1 exp (∆H H 2)(x ) ds µ (dx). o,y pb (o, y) exp H(y) 2 ˆ −|∇ | s o,y 1 - 0 .

~~Stochastic Analysis Andreas Eberle 2.4. PATH INTEGRALS AND BRIDGES 89~~

~~b The theorem yields the existence and a formula for the heat kernel p1(o, y),aswellasa b path integral representation for the bridge measure µo,y:~~

1 1 µb (dx) exp (∆H H 2)(x ) ds µ (dx). (2.50) o,y ∝ 2 ˆ −|∇ | s o,y - 0 . Proof of 2.15. Let F :Ω R and g : Rd R be measurable functions. By the → + → + disintegration of Wiener measure into pinned Wiener measures,

~~Eb[F g(W )] = E [Fg(W )Z ]= E [FZ ] g(y) p (o, y) dy. o · 1 o 1 1 ˆ o,y 1 1 Choosing F 1,weobtain ≡~~

~~g(y) pb (o, dy)= g(y) E [Z ] p (o, y) dy ˆ 1 ˆ o,y 1 1 for any non-negative measurable function g,whichimplies1). Now, choosing g 1,weobtainby1)that ≡~~

~~b Eo,y[FZ1] b Eo[F ]= Eo,y[FZ1] p1(o, y) dy = p1(o, dy) (2.51) ˆ ˆ Eo,y[Z1] = Eb [F ] pb (o, dy) (2.52) ˆ o,y 1~~

This proves 2), because W = yµb -a.s., and y µb is weakly continuous. 1 o,y #→ o,y Remark (Non-gradient case). If b is not a gradient then things are more involved because the expressions for the relative densities Zt involve a stochastic integral. In principle, one can proceed similarly as above after making sense of this stochastic integral for µo,y-almost every path x.

Example (Reversibility in the gradient case). The representation (2.50) immediately implies the following reversibility property of the diffusion bridge when b is a gradient: d d If R : C([0, 1], R ) C([0, 1], R ) denotes the time-reversal deﬁned by (Rx)t = x1 t, → − b 1 then the image µ R− of the bridge measure from o to y coincides with the bridge o,y ◦ b measure µy,o from y to o.Indeed,thispropertyholdsfortheBrownianbridge,andthe relative density in (2.50) is invariant under time reversal.

~~University of Bonn Summer Semester 2015 90 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~SDE for diffusion bridges~~

~~An important application of the h-transform is the interpretation of diffusion bridges by b achangeofmeasurew.r.t.thelawoftheunconditioneddiffusion process (W, µo) on C([0, 1], Rd) satisfying~~

dWt = dBt + b(Wt) dt, W0 = o, with an Rd-valued Brownian motion B.Weassumethatthetransitiondensity(t, x, y) #→ pb(x, y) is smooth for t>0 and bounded for t ε for any ε>0.Thenfory R, t ≥ ∈ pb( ,y) satisﬁes the Kolmogorov backward equation t · ∂ pb( ,y)= bpb( ,y) for any t>0, ∂t t · L t · where b = 1 ∆+b is the corresponding generator. Hence L 2 ·∇

b b h(t, z)=p1 t(z, y)/p1(o, y),t<1, − is a space-time harmonic function with h(0,o)=1.Sinceh is bounded for t 1 ε ≤ − b b for any ε>0,theprocessh(t, Wt) is a martingale under µo for t<1.Nowletµo,y be the measure on C([0, 1], Rd) that is absolutely continuous w.r.t. µb on W with relative o Ft density h(t, Wt) for any t<1.Thenthemarginaldistributionsoftheprocess(Wt)t<1 b b under µo, µo,y respectively are

b b b k b (Wt1 ,...,Wtk ) pt1 (o, x1)pt2 t1 (x1,x2) pt t − (xk 1,xk) λ (dx) w.r.t. µo, ∼ − ··· k− k 1 − b b b b pt1 (o, x1)pt2 t1 (x1,x2) pt t − (xk 1,xk)p1 t (xk,y) − ··· k− k 1 − − k k b b λ (dx) w.r.t. µo,y. ∼ p1(o, y)

This shows that y µb coincides with the regular version of the conditional distribu- → o,y b b tion of µo given W1,i.e.,µo,y is the bridge measure from o to y.Hence,byCorollary 2.12, we have shown:

~~b Theorem 2.16 (SDE for diffusion bridges). The diffusion bridge (W, µo,y) is a weak solution of the SDE~~

~~b dWt = dBt + b(Wt) dt +( log p1 t( ,y))(Wt) dt, t < 1. (2.53) ∇ − ·~~

~~Stochastic Analysis Andreas Eberle 2.5. LARGE DEVIATIONS ON PATH SPACES 91~~

b Note that the additional drift β(t, x)= log p1 t( ,y)(x) is singular as t 1.Indeed, ∇ − · ↑ if at a time close to 1 the process is still far away from y,thenastrongdriftisrequired to force it towards y.Ontheσ-algebra W ,themeasuresµb and µb are singular. F1 o o,y Remark (Generalized diffusion bridges). Theorem 2.16 carries over to bridges of diffusion processes with non-constant diffusion coefﬁcients σ.Inthiscase,the SDE (2.53) is replaced by

T dWt = σ(Wt) dBt + b(Wt) dt + σσ log p1 t( ,y) (Wt) dt. (2.54) ∇ − · $ % The last term can be interpreted as a gradient of the logarithmic heat kernel w.r.t. the T 1 Riemannian metric g =(σσ )− induced by the diffusion process.

~~2.5 Large deviations on path spaces~~

~~In this section, we apply Girsanov’s Theorem to study random perturbations of a dynamical system of type~~

ε ε ε dXt = b(Xt ) dt + √εdBt,X0 =0, (2.55) asymptotically as ε 0.Weshowthatontheexponentialscale,statementsaboutthe ↓ probabilities of rare events suggested by path integral heuristics can be put in a rigorous form as a large deviation principle on path space. Before, we give a complete proof of the Cameron-Martin Theorem.

~~Let Ω=C ([0, 1], Rd) endowed with the supremum norm ω =sup ω(t) : t [0, 1] , 0 || || {| | ∈ } let µ denote Wiener measure on (Ω),andletW (ω)=ω(t). B t~~

~~Translations of Wiener measure~~

~~For h Ω,weconsiderthetranslationoperatorτ :Ω Ω, ∈ h →~~

~~τh(ω)=ω + h,~~

~~1 and the translated Wiener measure µ := µ τ − . h ◦ h University of Bonn Summer Semester 2015 92 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~Theorem 2.17 (Cameron, Martin 1944). Let h Ω.Thenµ µ if and only if h is ∈ t ≪ contained in the Cameron-Martin space~~

~~2 d H = h Ω:h is absolutely contin. with h′ L ([0, 1], R ) . CM ∈ ∈ < = In this case, the relative density of µh w.r.t. µ is~~

t t dµh 1 2 =exp hs′ dWs hs′ ds . (2.56) dµ ˆ0 · − 2 ˆ0 | | 6 7 Proof. “ ”isaconsequenceofGirsanov’sTheorem:Forh H ,thestochastic ⇒ ∈ CM 1 2 integral h′ dW has ﬁnite deterministic quadratic variation [ h′ dW ] = h′ ds. · · 1 0 | | Hence by´ Novikov’s criterion, ´ ´

t t 1 2 Zt =exp h′ dW h′ ds ˆ0 · − 2 ˆ0 | | 6 7 is a martingale w.r.t. Wiener measure µ.Girsanov’sTheoremimpliesthatw.r.t.the measure ν = Z µ,theprocess(W ) is a Brownian motion translated by (h ).Hence 1 · t t

~~1 1 µ = µ (W + h)− = ν W − = ν. h ◦ ◦~~

“ ”Toprovetheconverseimplicationleth Ω,andsupposethatµ µ.SinceW ⇐ ∈ h ≪ is a Brownian motion w.r.t. µ, W h is a Brownian motion w.r.t. µ .Inparticular,it − h is a semimartingale. Moreover, W is a semimartingale w.r.t. µ and hence also w.r.t. µh. Thus h = W (W h) is also a semimartingale w.r.t. µ .Sinceh is deterministic, this − − h implies that h has ﬁnite variation.Wenowshow:

1 2 Rd Claim. The map g 0 g dh is a continuous linear functional on L ([0, 1], ). #→ ´ · The claim implies h H .Indeed,bytheclaimandtheRieszRepresentationTheo- ∈ CM rem, there exists a function f L2([0, 1], Rd) such that ∈ 1 1 g dh = g fds for any g L2([0, 1], Rd). ˆ0 · ˆ0 · ∈

2 d Hence h is absolutely continuous with h′ = f L ([0, 1], R ).Toprovetheclaim ∈ 2 d let (g ) be a sequence in L ([0, 1], R ) with g 2 0.ThenbyItô’sisometry, n || n||L → Stochastic Analysis Andreas Eberle 2.5. LARGE DEVIATIONS ON PATH SPACES 93

g dW 0 in L2(µ),andhenceµ-andµ -almost surely along a subsequence. Thus n → h ´also g dh = g d(W + h) g dW 0 ˆ n · ˆ n · − ˆ n · −→ µ-almost surely along a subsequence. Applying the same argument to a subsequence of (g ),weseethateverysubsequence(g ) has a subsequence (ˆg ) such that gˆ dh 0. n n n n· → This shows that g dh converges to 0 as well. The claim follows, since´ (g ) was an n · n arbitrary null sequence´ in L2([0, 1], R)d).

~~AﬁrstconsequenceoftheCameron-MartinTheoremisthatthesupport of Wiener mea- d sure is the whole space Ω=C0([0, 1], R ):~~

Corollary 2.18 (Support Theorem). For any h Ω and δ>0, ∈ µ ω Ω: ω h <δ > 0. { ∈ || − || } # ' Proof. Since the Cameron-Martin space is dense in Ω w.r.t. the supremum norm, it is enough to prove the assertion for h H .Inthiscase,theCameron-MartinTheorem ∈ CM implies

~~µ W h <δ = µ h W <δ > 0. || − || − || || # ' # ' as µ[ W <δ] > 0 and µ h µ. || || − ≪ Remark (Quantitative Support Theorem). More explicitly,~~

µ W h <δ = µ h W <δ || − || − || || 1 1 # ' # ' 1 2 = E exp h′ dW h′ ds ;maxWs <δ − ˆ0 · − 2 ˆ0 | | s 1 | | / 6 7 ≤ 0 where the expectation is w.r.t. Wiener measure. This can be used to derive quantitative estimates.

~~Schilder’s Theorem~~

~~We now study the solution of (2.55) for b 0,i.e., ≡ Xε = √εB,t[0, 1], t t ∈ University of Bonn Summer Semester 2015 94 CHAPTER 2. TRANSFORMATIONS OF SDE~~

~~with ε>0 and a d-dimensional Brownian motion (Bt).Pathintegralheuristicssuggests that for h H , ∈ CM~~

~~ε h I(h/√ε) I(h)/ε “ P [X h]=µ W e− = e− ” ≈ ≈ √ε ∼ / 0 where I :Ω [0, ] is the action functional deﬁned by → ∞~~

1 1 2 2 0 ω′(s) ds if ω HCM, I(ω)= ´ | | ∈ ⎧+ otherwise. ⎨ ∞ The heuristics can be turned into⎩ a rigorous statement asymptotically as ε 0 on the → exponential scale. This is the content of the next two resultsthattogetherareknowas Schilder’s Theorem:

~~Theorem 2.19 (Schilder’s large derivation principle, lower bound).~~

1) For any h H and δ>0, ∈ CM lim inf ε log µ √εW B(h, δ)] I(h). ε 0 ↓ ∈ ≥− # 2) For any open subset U Ω, ⊆ lim inf ε log µ √εW U inf I(ω). ε 0 ω U ↓ ∈ ≥−∈ # ' Here B(h, δ)= ω Ω: ω h <δ denotes the ball w.r.t. the supremum norm. { ∈ || − || } Proof. 1) Let c = 8I(h).Thenforε>0 sufﬁciently small, D µ √εW B(h, δ) = µ W B(h/√ε,δ/√ε) ∈ ∈ # ' = µ# h/√ε B(0,δ/√ε) ' − 1 1 # 1 ' 1 2 δ = E exp h′ dW h′ ds ; B 0, − √ε ˆ0 · − 2ε ˆ0 | | √ε / 6 7 6 70 1 c 1 δ exp I(h) µ h′ dW c B(0, ) ≥ − ε − √ε ˆ0 · ≤ ∩ √ε 6 7 /+ , 0 1 1 8I(h) exp I(h) ≥ 2 9−ε − E ε :

~~Stochastic Analysis Andreas Eberle 2.5. LARGE DEVIATIONS ON PATH SPACES 95 where E stands for expectation w.r.t. Wiener measure. Here we have used that~~

~~1 1 2 2 2 µ h′ dW > c c− E h′ dW =2I(h)/c 1/4 ˆ0 · ≤ ˆ0 · ≤ / 0 /6 7 0 by Itô’s isometry and the choice of c.~~

2) Let U be an open subset of Ω.Forh U H ,thereexistsδ>0 such that ∈ ∩ CM B(h, δ) U.Henceby1), ⊂ lim inf ε log µ[√εW U] I(h). ε 0 ↓ ∈ ≥− Since this lower bound holds for any h U H ,andsinceI = on U H ,we ∈ ∩ CM ∞ \ CM can conclude that

~~lim inf ε log µ[√εW U] inf I(h)= inf I(ω). ε 0 h U HCM ω U ↓ ∈ ≥−∈ ∩ − ∈~~

To prove a corresponding upper bound, we consider linear approximations of the Brow- nian paths. For n N let ∈ W (n) := (1 s)W + sW t − k/n k+1/n whenever t =(k + s)/n for k 0, 1,...,n 1 and s [0, 1]. ∈{ − } ∈ Theorem 2.20 (Schilder’s large deviations principle, upper bound).

1) For any n N and λ 0, ∈ ≥ lim sup ε log µ[I(√εW (n)) λ] λ. ε 0 ≥ ≤− ↓ 2) For any closed subset A Ω, ⊆ lim sup ε log µ[√εW A] inf I(ω). ε 0 ∈ ≤−ω A ↓ ∈ Proof. 1) Let ε>0 and n N.Then ∈ n (n) 1 2 I(√εW )= ε n (Wk/n W(k 1)/n) . 2 − − "k=1 University of Bonn Summer Semester 2015 96 CHAPTER 2. TRANSFORMATIONS OF SDE

~~Since the random variables ηk := √n (Wk/n W(k 1)/n) are independent and standard · − − normally distributed, we obtain~~

~~µ[I(√εW (n)) λ]=µ η 2 2λ/ε ≥ | k| ≥ / 0 exp("2λc/ε) E exp c η 2 , ≤ − | k| / 6 70 where the expectation on the right hand side is ﬁnite for c<1/2.Henceforany" c<1/2,~~

~~lim sup ε log µ[I(√εW (n)) λ] 2cλ. ε 0 ≥ ≤− ↓ The assertion now follows as c 1/2. ↑ 2) Now ﬁx a closed set A Ω and λ~~

lim sup ε log µ[√εW A] λ. (2.57) ε 0 ∈ ≤− ↓ By the Theorem of Arzéla-Ascoli, the set I λ is a compact subset of the Banach { ≤ } space Ω.Indeed,bytheCauchy-Schwarzinequality, t ω(t) ω(s) = ω′(u) du √2λ √t s s, t [0, 1] | − | ˆs ≤ − ∀ ∈ & & holds for any ω Ω satisfying& I(ω) λ.Hencethepathsin& I λ are equicontinu- ∈ & ≤ & { ≤ } ous, and the Arzéla-Ascoli Theorem applies. Let δ denote the distance between the sets A and I λ w.r.t. the supremum norm. { ≤ } Note that δ>0,becauseA is closed, I λ is compact, and both sets are disjoint by { ≤ } the choice of λ.Henceforε>0,wecanestimate

~~µ[√εW A] µ[I(√εW (n)) >λ]+µ[ √εW √εW (n) >δ]. ∈ ≤ || − ||sup The assertion (2.57) now follows from~~

lim sup ε log µ[I(√εW (n)) >λ] λ, and (2.58) ε 0 ≤− ↓ (n) lim sup ε log µ[ W W sup >δ/√ε] λ. (2.59) ε 0 || − || ≤− ↓ The bound (2.58) holds by 1) for any n N.Theproofof(2.59)reducestoanestimate ∈ of the supremum of a Brownian bridge on an interval of length 1/n.Weleaveitasan exercise to verify that (2.59) holds if n is large enough.

~~Stochastic Analysis Andreas Eberle 2.5. LARGE DEVIATIONS ON PATH SPACES 97~~

Remark (Large deviation principle for Wiener measure). Theorems 2.19 and 2.20 show that 1 µ[√εW A] exp inf I(ω) ∈ ≃ − ε ω A 6 ∈ 7 holds on the exponential scale in the sense that a lower bound holds for open sets and an upper bound holds for closed sets. This is typical for largedeviationprinciples, see e.g. [10] or [11]. The proofs above based on “exponential tilting” of the underlying Wiener measure (Girsanov transformation) for the lower bound, and an exponential estimate combined with exponential tightness for the upper bound are typical for the proofs of many large deviation principles.

~~Random perturbations of dynamical systems~~

~~We now return to our original problem of studying small randomperturbationsofa dynamical system~~

~~ε ε ε dXt = b(Xt ) dt + √εdBt,X0 =0. (2.60)~~

~~This SDE can be solved pathwise:~~

Lemma 2.21 (Control map). Suppose that b : Rd Rd is Lipschitz continuous. Then: → 1) For any function ω C([0, 1], Rd) there exists a unique function x C([0, 1], Rd) ∈ ∈ such that t x(t)= b(x(s)) ds + ω(t) t [0, 1]. (2.61) ˆ0 ∀ ∈ The function x is absolutely continuous if and only if ω is absolutely continuous, and in this case,

~~x′(t)=b(x(t)) + ω′(t) for a.e. t [0, 1]. (2.62) ∈~~

~~2) The control map : C([0, 1], Rd) C([0, 1], Rd) that maps ω to the solution J → (ω)=x of (2.61) is continuous. J~~

~~University of Bonn Summer Semester 2015 98 CHAPTER 2. TRANSFORMATIONS OF SDE~~

Proof. 1) Existence and uniqueness holds by the classical Picard-Lindelöf Theorem. 2) Suppose that x = (ω) and x = (ω) are solutions of (2.61) w.r.t. driving paths J J ω,ω C[0, 1], Rd).Thenfort [0, 1], ∈ ∈ ) ) t ) x(t) x(t) = (b(ω(s)) b(ω(s))) ds + √ε(ω(t) ω(t)) | − | ˆ0 − − & & & t & ) &L ω(s) ω(s)) ds + √ε (t)m ω(t) ). & ≤ ˆ0 | − | | − | where L R is a Lipschitz constant for b.Gronwall’sLemmanowimplies) ) ∈ +

x(t) x(t) exp(tL) √ε ω ω t [0, 1], | − |≤ || − ||sup ∀ ∈ and hence ) ) x x exp(L) √ε ω ω . || − ||sup ≤ || − ||sup This shows that the control map is even Lipschitz continuous. ) J )

~~For ε>0,theuniquesolutionoftheSDE(2.60)on[0, 1] is given by~~

~~Xε = (√εB). J~~

Since the control map is continuous, we can apply Schilder’s Theorem to study the J large deviations of Xε as ε 0: ↓ Theorem 2.22 (Fredlin & Wentzel 1970, 1984). If b is Lipschitz continuous then the large deviations principle

~~ε lim inf ε log P [X U] inf Ib(x) for any open set U Ω, ε 0 x U ↓ ∈ ≥−∈ ⊆~~

~~ε lim inf ε log P [X A] inf Ib(x) for any closed set A Ω, ε 0 x A ↓ ∈ ≥−∈ ⊆ holds, where the rate function I :Ω [0, ] is given by b → ∞~~

~~1 1 2 2 0 x′(s) b(x(s)) ds for x HCM, Ib(x)= ´ | − | ∈ ⎧+ for x Ω H . ⎨ ∞ ∈ \ CM ⎩ Stochastic Analysis Andreas Eberle 2.5. LARGE DEVIATIONS ON PATH SPACES 99~~

~~Proof. For any set A Ω,wehave ⊆~~

~~ε 1 1 P [X A]=P [√εB − (A)] = µ[√εW − (A)]. ∈ ∈J ∈J~~

1 If A is open then − (A) is open by continuity of ,andhence J J lim inf ε log P [Xε A] inf I(ω)) ε 0 ∈ ≥−ω −1(A) ↓ ∈J 1 by Theorem 2.19. Similarly, if A is closed then − (A) is closed, and hence the corre- J sponding upper bound holds by Theorem 2.20. Thus it only remains to show that

inf I(ω) = inf Ib(x). ω −1(A) x A ∈J ∈ 1 To this end we note that ω − (A) if and only if x = (ω) A,andinthiscase ∈J J ∈ ω′ = x′ b(x).Therefore, − 1 1 2 inf I(ω) = inf ω′(s) ds −1 −1 ω (A) ω (A) HCM 2 ˆ | | ∈J ∈J ∩ 0 1 2 = inf x′(s) b(x(s)) ds = inf Ib(x). x A HCM x A ∈ ∩ 2| − | ∈

Remark. The large deviation principle in Theorem 2.22 generalizes tonon-Lipschitz continuous vector ﬁelds b and to SDEs with multiplicative noise. However, in this case, there is no continuous control map that can be used to reduce the statement to Schilder’s Theorem. Therefore, a different proof is required, cf. e.g. [10].

~~University of Bonn Summer Semester 2015 Chapter 3~~

~~Extensions of Itô calculus~~

This chapter contains an introduction to some important extensions of Itô calculus and the type of SDE considered so far. We will consider SDE for jumpprocessesdriven by white and Poisson noise, Stratonovich calculus and Brownian motion on curved surfaces, stochastic Taylor expansions and numerical methods for SDE, local times and asingularSDEforreﬂectedBrownianmotion,aswellasstochastic ﬂows.

We start by recalling a crucial martingale inequality that wewillapplyfrequentlyto p derive L estimates for semimartingales. For real-valued càdlàg functions x =(xt)t 0 ≥ we set ⋆ ⋆ xt := sup xs for t>0, and x0 := x0 . s

~~100 3.1. SDE WITH JUMPS 101~~

~~In these notes, we only prove an easy special case of the Burkholder-Davis-Gundy inequality that will be sufﬁcient for our purposes: For any p [2, ), ∈ ∞~~

~~E[(M ⋆ )p]1/p e/2 pE[[M]p/2]1/p (3.2) T ≤ T D This estimate also holds for càdlàg local martingales and is proven in Theorem 5.24.~~

~~3.1 SDE with jumps~~

Let (S, ,ν) be a σ-ﬁnite measure space, and let d, n N.Supposethatonaproba- S ∈ bility space (Ω, ,P),wearegivenanRd-valued Brownian motion (B ) and a Poisson A t random measure N(dt dy) over R+ S with intensity measure λ(0, ) ν.Let( t) × ∞ ⊗ F denote a complete ﬁltration such that (B ) is an ( ) Brownian motion and N (B)= t Ft t N((0,t] B) is an ( ) Poisson point process, and let × Ft

~~N(dt dy)=N(dt dy) λ(0, )(dt) ν(dy). − ∞~~

If T is an ( ) stopping) time then we call a predictable process (ω,t) G (ω) or Ft #→ t (ω,t,y) G (y)(ω) defined for finite t T (ω) and y S locally square integrable #→ t ≤ ∈ iff there exists an increasing sequence (T ) of ( ) stopping times with T =supT n Ft n 2 such that for any n,thetriviallyextendedprocessGtI t Tn is contained in (P { ≤ } L ⊗ λ), 2(P λ ν) respectively. For locally square integrable predictable integrands, L ⊗ ⊗ t the stochastic integrals 0 Gs dBs and (0,t] S Gs(y) N(ds dy) respectively are local × martingales defined for t´ [0,T). ´ ∈ ) In this section, we are going to study existence and pathwise uniqueness for solutions of stochastic differential equations of type

~~dXt = bt(X) dt + σt(X) dBt + ct (X, y) N(dt dy). (3.3) ˆy S − ∈~~

n n n )n d Here b : R (R , R ) R , σ : R (R , R ) R × ,andc : R + ×D + → + ×D + → + × (R , Rn) S Rn are càdlàg functions in the ﬁrst variable such that b , σ and c D + × → t t t University of Bonn Summer Semester 2015 102 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS are measurable w.r.t. the σ-algebras := σ(x x : s t), respectively for Bt #→ s ≤ Bt ⊗S any t 0.Wealsoassumelocal boundedness of the coefﬁcients, i.e., ≥

sup sup sup ( bs(x) + σs(x) + cs(x, y) ) < (3.4) st.Corresponding statements hold for σt and ct.Condition(3.4)impliesinparticularthatthejumpsizesare locally bounded. Locally unbounded jumps could be taken intoaccountbyextending the SDE (3.3) by an additional term consisting of an integral w.r.t. an uncompensated Poisson point process.

~~Deﬁnition. Suppose that T is an ( ) stopping time. Ft 1) A solution of the stochastic differential equation (3.3) for t~~

~~t t Xt = X0 + bs(X) ds + σs(X) dBs + cs (X, y) N(ds dy) (3.5) ˆ0 ˆ0 ˆ(0,t] S − × ) holds for any t~~

~~2) A solution (Xt)t~~

For a strong solution, Xt is almost surely a measurable function of the initial value X0 and the processes (Bs)s t and (Ns)s t driving the SDE up to time t.InSection2.1,we ≤ ≤ saw an example of a solution to an SDE that does not possess thisproperty.

Remark. The stochastic integrals in (3.5) are well-deﬁned strict local martingales. Indeed, the local boundedness of the coefﬁcients guaranteeslocalsquareintegrabil- ity of the integrands as well as local boundedness of the jumpsfortheintegralw.r.t.

~~Stochastic Analysis Andreas Eberle 3.1. SDE WITH JUMPS 103~~

~~N.Theprocessσs(X) is not necessarily predictable, but observing that σs(X(ω)) =~~

~~σs (X(ω)) for P λ almost every (ω,s),wemaydeﬁne )− ⊗~~

~~σs(X) dBs := σs (X) dBs. ˆ ˆ −~~

~~Lp Stability~~

In addition to the assumptions above, we assume from now on that the coefﬁcients in the SDE (3.3) satisfy a local Lipschitz condition: Assumption (A1). For any t R,andforanyopenboundedsetU Rn,there 0 ∈ ⊂ exists a constant L R such that the following Lipschitz condition Lip(t ,U) holds: ∈ + 0 2 bt(x) bt(x) + σt(x) σt(x) + ct(x, ) ct(x, ) L2(ν) L sup xs xs | − | || − || ∥ • − • ∥ ≤ · s t | − | ≤ R Rn for any t [0),t0] and x, x () +, ) with xs, xs ) U for s t0. ) ∈ ∈D ∈ ≤ We now derive an a priori estimate for solutions of (3.3) that is crucial for studying ) ) existence, uniqueness, and dependence on the initial condition:

Theorem 3.1 (Aprioriestimate). Fix p [2, ) and an open set U Rn,andlet ∈ ∞ ⊆ T be an ( ) stopping time. Suppose that (X ) and (X ) are solutions of (3.3) taking Ft t t values in U for t2,theproofcanbecarriedout in a similar way by relying on Burkholder’s inequality instead of Itô’s isometry. Clearly, (3.7) follows from (3.6) by Gronwell’s lemma. To prove (3.6), note that t t Xt = X0 + bs(X) ds+ σs(X) dBs + cs (X, y) N(ds dy) t

~~2 E[I ]=ε0, and~~

t T t 2 2 ∧ ⋆ 2 2 E[II ] L tE (X X)s ds L t εs ds. ≤ ˆ0 − ≤ ˆ0 / 0 Denoting by Mu and Ku the stochastic integrals) in III and IV respectively, Doob’s inequality and Itô’s isometry imply

2 ⋆ 2 2 E[III ]=E[Mt T ] 4E[Mt T ] ∧ ≤ ∧ t T t ∧ 2 2 =4E σs(X) σs(X) ds 4L εs ds, ˆ0 || − || ≤ ˆ0 / 0 ) 2 ⋆ 2 2 E[IV ]=E[Kt T ] 4E[Kt T ] ∧ ≤ ∧ t T t ∧ 2 2 =4E cs (X, y) cs (X,y) ν(dy) ds 4L εs ds. ˆ0 ˆ | − − − | ≤ ˆ0 / 0 The assertion now follows since by (3.8), )

~~⋆ 2 2 2 2 2 εt = E (X X)t T ] 4 E[I + II + III + IV . − ∧ ≤ · # ' )~~

The a priori estimate shows in particular that under a global Lipschitz condition, solutions depend continuously on the initial condition in mean square. Moreover, it implies pathwise uniqueness under a local Lipschitz condition:

~~Stochastic Analysis Andreas Eberle 3.1. SDE WITH JUMPS 105~~

~~Corollary 3.2 (Pathwise uniqueness). Suppose that Assumption (A1) holds. If (Xt) and (Xt) are strong solutions of (3.1) with X0 = X0 almost surely then~~

) P Xt = Xt for any t) =1. # ' Proof. For any open bounded set U ) Rn and t R ,theaprioriestimateinTheorem ⊂ 0 ∈ + 3.1 implies that X and X coincide almost surely on [0,t T c ) where T c denotes the 0 ∧ U U ﬁrst exit time from U. )

~~Existence of strong solutions~~

~~To prove existence of strong solutions, we need an additionalassumption:~~

~~Assumption (A2). For any t R , 0 ∈ + 2 sup ct(0,y) ν(dy) < . t~~

~~2 2 2 ⋆ 2 sup bt(x) + σt(x) + ct(x, y) ν(dy) C(t0) (1 + xt0 ) . (3.9) t~~

Theorem 3.3 (Itô). Let ξ :Ω Rn be a random variable that is independent of the → Brownian motion B and the Poisson random measure N. 1) Suppose that the local Lipschitz condition (A1) and (A2) hold. Then (3.1) has a

~~strong solution (Xt)t<ζ with initial condition X0 = ξ that is deﬁned up to the explosion time~~

ζ =supT , where T = inf t 0: X k . k k { ≥ | t|≥ } 2) If, moreover, the global Lipschitz condition Lip(t , Rn) holds for any t R , 0 0 ∈ + then ζ = almost surely. ∞ University of Bonn Summer Semester 2015 106 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS

Proof of 3.3. We first prove existence of a global strong solution (Xt)t [0, ) assuming ∈ ∞ (A2) and a global Lipschitz condition Lip(t , Rn) for any t R .Thefirstassertion 0 0 ∈ + will then follow by localization. For proving global existence we may assume w.l.o.g. that ξ is bounded and thus square integrable. We then construct a sequence (Xn) of approximate solutions to (3.1) by a Picard-Lindelöf iteration,i.e.,fort 0 and n Z we define inductively ≥ ∈ +

0 Xt := ξ, (3.10) t t n+1 n n n Xt := ξ + bs(X ) ds + σs(X ) dBs + cs (X ,y) N(ds dy). ˆ0 ˆ0 ˆ − (0,t] S × ) Fix t [0, ).WewillshowbelowthatAssumption(A2)andtheglobalLipschitz 0 ∈ ∞ condition imply that

~~(i) for any n N, Xn is a square integrable ( 0) semimartingale on [0,t ] (i.e., ∈ Ft 0 the sum of a square integrable martingale and an adapted process with square integrable total variation), and~~

~~(ii) there exists a ﬁnite constant C(t0) such that the mean square deviations~~

~~∆n := E[(Xn+1 Xn)⋆ 2]. t − t of the approximations Xn and Xn+1 satisfy~~

~~t n+1 n ∆t C(t0) ∆s ds for any n 0 and t t0. ≤ ˆ0 ≥ ≤~~

~~Then, by induction, tn ∆n C(t )n ∆0 for any n N and t t . t ≤ 0 n! t ∈ ≤ 0~~

n In particular, ∞ ∆ < .AnapplicationoftheBorel-CantelliLemmanowshows n=1 t0 ∞ n that the limit Xs = limn Xs exists uniformly for s [0,t0] with probability one. ! →∞ ∈ Moreover, X is a fixed point of the Picard-Lindelöf iteration, and hence a solution of the SDE (3.1). Since t0 has been chosen arbitrarily, the solution is defined almost surely on [0, ),andbyconstructionitisadaptedw.r.t.thefiltration( 0). ∞ Ft Stochastic Analysis Andreas Eberle 3.1. SDE WITH JUMPS 107

We now show by induction that Assertion (i) holds. If Xn is a square integrable ( 0) semimartingale on [0,t ] then, by the linear growth condition (3.9), the process Ft 0 b (Xn) 2 + σ (Xn) 2 + c (Xn,y) 2 ν(dy) is integrable w.r.t. the product measure | s | || s || | s | P λ .Therefore,byItô’sisometry,theintegralsontherightha´ nd side of (3.10) ⊗ (0,t0) all deﬁne square integrable ( 0) semimartingales, and thus Xn+1 is a square integrable Ft ( 0) semimartingale, too. Ft Assertion (ii) is a consequence of the global Lipschitz condition. Indeed, by the Cauchy- n Schwarz inequality, Itô’s isometry and Lip(t0, R ),thereexistsaﬁniteconstantC(t0) such that

∆n+1 = E Xn+2 Xn+1 ⋆ 2 t − t t t /$ % 0 2 2 3tE b (Xn+1) b (Xn) ds +3E σ (Xn+1) σ (Xn) ds ≤ ˆ s − s ˆ s − s 4 0 5 4 0 5 &t & F F +3E & c (Xn+1,y) c& (Xn,y) 2 ν(dy) dsF F ˆ ˆ s − s 4 0 5 t & & n & & C(t0) ∆s ds for any n 0 and t t0. ≤ ˆ0 ≥ ≤ This completes the proof of global existence under a global Lipschitz condition. Finally, suppose that the coefficients b, σ and c only satisfy the local Lipschitz condition (A1). Then for k N and t R ,wecanfindfunctionsbk, σk and ck that are globally ∈ 0 ∈ + Lipschitz continuous and that agree with b, σ and c on paths (xt) taking values in the ball B(0,k) for t t .ThesolutionX(k) of the SDE with coefficients bk, σk, ck is then ≤ 0 asolutionof(3.1)uptot T where T denotes the first exit time of X(k) from B(0,k). ∧ k k By pathwise uniqueness, the local solutions obtained in thiswayareconsistent.Hence they can be combined to construct a solution of (3.1) that is defined up to the explosion time ζ =supTk.

~~Non-explosion criteria~~

Theorem 3.3 shows that under a global Lipschitz and linear growth condition on the coefﬁcients, the solution to (3.1) is deﬁned for all times with probability one. How- ever, this condition is rather restrictive, and there are much better criteria to prove that

University of Bonn Summer Semester 2015 108 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS the explosion time ζ is almost surely inﬁnite. Arguably the most generally applicable non-explosion criteria are those based on stochastic Lyapunov functions.Considerfor example an SDE of type

~~dXt = b(Xt) dt + σ(Xt) dBt (3.11)~~

n n n n d where b : R R and σ : R R × are locally Lipschitz continuous, and let → → 1 n ∂2 = a (x) + b(x) ,a(x)=σ(x)σ(x)T , L 2 ij ∂x ∂x ·∇ i,j=1 i j " denote the corresponding generator.

Theorem 3.4 (Lyapunov condition for non-explosion). Suppose that there exists a function ϕ C2(Rn) such that ∈ (i) ϕ(x) 0 for any x Rn, ≥ ∈ (ii) ϕ(x) as x ,and →∞ | |→∞ (iii) ϕ λϕ for some λ R . L ≤ ∈ + Then the strong solution of (3.1) with initial value x Rn exists up to ζ = almost 0 ∈ ∞ surely.

Proof. We ﬁrst remark that by (iii), Z := exp( λt)ϕ(X ) is a supermartingale up to t − t the ﬁrst exit time T of the local solution X from a ball B(0,k) Rn.Indeed,bythe k ⊂ product rule and the Itô-Doeblin formula,

λt λt λt dZ = λe− ϕ(X) dt + e− dϕ(X)=dM + e− ( ϕ λϕ)(X) dt − L − holds on [0,Tk] with a martingale M up to Tk. Now we ﬁx t 0.Then,bytheOptionalStoppingTheoremandbyCondition(i), ≥

ϕ(x0)=E [ϕ(X0)] E [exp( λ(t Tk)) ϕ(Xt T )] ≥ − ∧ ∧ k E [exp( λt) ϕ(X ); T t] ≥ − Tk k ≤ exp( λt) inf ϕ(y) P [Tk t] ≥ − y =k ≤ | | for any k N.Ask , inf y =k ϕ(y) by (ii). Therefore, ∈ →∞ | | →∞

~~P [sup Tk t] = lim P [Tk t]=0 k ≤ →∞ ≤ for any t 0,i.e.,ζ =supT = almost surely. ≥ k ∞ Stochastic Analysis Andreas Eberle 3.2. STRATONOVICH DIFFERENTIAL EQUATIONS 109~~

By applying the theorem with the function ϕ(x)=1+ x 2 we obtain: | | Corollary 3.5. If there exists λ R such that ∈ + 2x b(x)+tr(a(x)) λ (1 + x 2) for any x Rn · ≤ · | | ∈ then ζ = almost surely. ∞ Note that the condition in the corollary is satisﬁed if x b(x) const. x and tr a(x) const. x 2 x · ≤ ·| | ≤ ·| | | | for sufﬁciently large x Rn,i.e.,iftheoutwardcomponentofthedriftisgrowingat ∈ most linearly, and the trace of the diffusion matrix is growing at most quadratically.

~~3.2 Stratonovich differential equations~~

Replacing Itô by Statonovich integrals has the advantage that the calculus rules (product rule, chain rule) take the same form as in classical differential calculus. This is useful for explicit computations (Doss-Sussman method), for approximating solutions of SDE by solutions of ordinary differential equations, and in stochastic differential geometry. For simplicity, we only consider Stratonovich calculus for continuous semimartingales, cf. [36] for the discontinuous case. Let X and Y be continuous semimartingales on a filtered probability space (Ω, ,P,( )). A Ft Definition (Fisk-Stratonovich integral). The Stratonovich integral X dY is the ◦ continuous semimartingale defined by ´ t t 1 Xs dYs := Xs dYs + [X, Y ]t for any t 0. ˆ0 ◦ ˆ0 2 ≥ Note that a Stratonovich integral w.r.t. a martingale is not alocalmartingaleingeneral. The Stratonovich integral is a limit of trapezoidal Riemann sum approximations:

~~Lemma 3.6. If (π ) is a sequence of partitions of R with mesh(π ) 0 then n + n → t Xs + Xs′ t Xs dYs = lim ∧ (Ys′ t Ys) in the ucp sense. ˆ ◦ n 2 ∧ − 0 →∞ s πn "s~~

~~University of Bonn Summer Semester 2015 110 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS~~

~~t ′ Proof. This follows since 0 XdY =ucplim s~~

For Stratonovich integrals w.r.t. continuous semimartingales, the classical chain rule holds: Theorem 3.7. Let X =(X1,...,Xd) with continuous semimartingales Xi.Thenfor any function F C2(Rd), ∈ d t ∂F F (X ) F (X )= (X ) dXi t 0. (3.12) t − 0 ˆ ∂xi s ◦ s ∀ ≥ i=1 0 " Proof. To simplify the proof we assume F C3.Underthiscondition,(3.12)isjusta ∈ reformulation of the Itô rule d t ∂F 1 d t ∂2F F (X ) F (X )= (X ) dXi + (X ) d[Xi,Xj] . t − 0 ˆ ∂xi s s 2 ˆ ∂xi∂xj s s i=1 0 i,j=1 0 " " (3.13) 2 ∂F Indeed, applying Itô’s rule to the C function ∂xi shows that ∂F ∂2F (X )=A + (X ) dXj ∂xi t t ˆ ∂xi∂xj s s j " for some continuous ﬁnite variation process A.Hencethedifferencebetweenthe Statonovich integral in (3.12) and the Itô integral in (3.13)is 2 1 ∂F i 1 ∂ F j i (X),X = (Xs) d[X ,X ]s. 2 ∂xi t 2 ˆ ∂xi∂xj j / 0 "

Remark. For the extension of the proof to C2 functions F see e.g. [36], where also a generalization to càdlàg semimartingales is considered. The product rule for Stratonovich integrals is a special caseofthechainrule: Corollary 3.8. For continuous semimartingales X, Y , t t XtYt X0Y0 = Xs dYs + Ys dXs t 0. − ˆ0 ◦ ˆ0 ◦ ∀ ≥ Exercise (Associative law). Prove an associative law for Stratonovich integrals.

~~Stochastic Analysis Andreas Eberle 3.2. STRATONOVICH DIFFERENTIAL EQUATIONS 111~~

~~Stratonovich SDE~~

Since Stratonovich integrals differ from the correspondingItôintegralsonlybytheco- variance term, equations involving Stratonovich integralscanberewrittenasItôequa- tions and vice versa, provided the coefﬁcients are sufﬁciently regular. We consider a Stratonovich SDE in Rd of the form

~~d dX = b(X ) dt + σ (X ) dBk,X= x , (3.14) ◦ t t k t ◦ t 0 0 "k=1 with x Rn,continuousvectorﬁeldsb, σ ,...,σ C(Rn, Rn),andanRd-valued 0 ∈ 1 d ∈ Brownian motion (Bt).~~

~~Exercise (Stratonovich to Itô conversion). 1) Prove that for σ ,...,σ C1(Rn, Rn), 1 d ∈ the Stratonovich SDE (3.14) is equivalent to the Itô SDE~~

d k dXt = b(Xt) dt + σk(Xt) dBt ,X0 = x0, (3.15) "k=1 ) where 1 d b := b + σ σ . 2 k ·∇ k "k=1 ) 2) Conclude that if b and σ1,...,σd are Lipschitz continuous, then there is a unique strong solution of (3.14). ) Theorem 3.9 (Martingale problem for Stratonovich SDE). Let b C(Rn, Rn) and ∈ 2 n n σ1,...,σd C (R , R ),andsupposethat(Xt)t 0 is a solution of (3.14) on a given ∈ ≥ setup (Ω, ,P,( ), (B )).ThenforanyfunctionF C3(Rn),theprocess A Ft t ∈ t F Mt = F (Xt) F (X0) ( F )(Xs) ds, − − ˆ0 L 1 d F = σ (σ F )+b F, L 2 k ·∇ k ·∇ ·∇ "k=1 is a local ( P ) martingale. Ft

~~University of Bonn Summer Semester 2015 112 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS~~

~~Proof. By the Stratonovich chain rule and by (3.14),~~

t F (Xt) F (X0)= F (X) dX − ˆ0 ∇ ·◦ t t k = (b F )(X) dt + (σk F )(X) dB . (3.16) ˆ0 ·∇ ˆ0 ·∇ ◦ "k By applying this formula to σ F ,weseethat k ·∇ (σ F )(X )=A + σ (σ F )(X) dBl k ·∇ t t ˆ l ·∇ k ·∇ "l with a continuous ﬁnite variation process (At).Hence

~~t t k k k (σk F )(X) dB = (σk F )(X) dB + [(σk F )(X),B ]t ˆ0 ·∇ ◦ ˆ0 ·∇ ·∇ t = local martingale + σk (σk F )(X)dt. ˆ0 ·∇ ·∇ (3.17)~~

~~The assertion now follows by (3.16) and (3.17).~~

The theorem shows that the generator of a diffusion process solving a Stratonovich SDE is in sum of squares form. In geometric notation, one brieﬂy writes b for the derivative b in the direction of the vector ﬁeld b.Thegeneratorthentakestheform ·∇ 1 = σ2 + b L 2 k "k Brownian motion on hypersurfaces

One important application of Stratonovich calculus is stochastic differential geometry. Itô calculus can not be used directly for studying stochasticdifferentialequationson manifolds, because the classical chain rule is essential forensuringthatsolutionsstay on the manifold if the driving vector ﬁelds are tangent vectors. Instead, one considers Stratonovich equations. These are converted to Itô form whencomputingexpectation values. To avoid differential geometric terminology, we only consider Brownian motion on a hypersurface in Rn+1,cf.[38],[20]and[22]forstochasticcalculusonmoregeneral Riemannian manifolds.

~~Stochastic Analysis Andreas Eberle 3.2. STRATONOVICH DIFFERENTIAL EQUATIONS 113~~

n+1 Let f C∞(R ) and suppose that c R is a regular value of f,i.e., f(x) =0for ∈ ∈ ∇ ̸ 1 any x f − (c).Thenbytheimplicitfunctiontheorem,thelevelset ∈ 1 n+1 M = f − (c)= x R : f(x)=c c ∈ is a smooth n-dimensional submanifold of R

~~For x Mc,thevector ∈ f(x) n(x)= ∇ Sn f(x) ∈ |∇ | is the unit normal to Mc at x.Thetangent space to Mc at x is the orthogonal complement~~

~~T M =spann(x) ⊥ . x c { } Let P (x):Rn+1 T M denote the orthogonal projection onto the tangent space w.r.t. → x c the Euclidean metric, i.e.,~~

~~P (x)v = v v n(x) n(x),vRn+1. − · ∈~~

~~η(x)~~

~~Mc′ TxMc~~

~~University of Bonn Summer Semester 2015 114 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS~~

~~For k 1,...,n+1 ,wesetP (x)=P (x)e . ∈{ } k k~~

~~Deﬁnition. A Brownian motion on the hypersurface Mc with initial value x M is 0 ∈ c asolution(Xt) of the Stratonovich SDE~~

~~n+1 dX = P (X ) dB = P (X ) dBk,X= x , (3.18) ◦ t t ◦ t k t ◦ t 0 0 "k=1 n+1 with respect to a Brownian motion (Bt) on R .~~

We now assume for simplicity that Mc is compact. Then, since c is a regular value of f,thevectorfieldsPk are smooth with bounded derivatives of all orders in a neigh- n+1 bourhood U of Mc in R .Therefore,thereexistsauniquestrongsolutionoftheSDE (3.18) in Rn+1 that is defined up to the first exit time from U.Indeed,thissolutionstays on the submanifold Mc for all times:

~~Theorem 3.10. If X is a solution of (3.18) with x M then almost surely, X M 0 ∈ c t ∈ c for any t 0. ≥ The proof is very simple, but it relies on the classical chain rule in an essential way:~~

~~Proof. We have to show that f(Xt) is constant. This is an immediate consequence of the Stratonovich formula:~~

~~t n+1 t k f(Xt) f(X0)= f(Xs) dXs = f(Xs) Pk(Xs) dBs =0 − ˆ0 ∇ ·◦ ˆ0 ∇ · ◦ "k=1 since P (x) is orthogonal to f(x) for any x. k ∇~~

Although we have deﬁned Brownian motion on the Riemannian manifold Mc in a non- intrinsic way, one can verify that it actually is an intrinsicobjectanddoesnotdependon n+1 the embedding of Mc into R that we have used. We only convince ourselves that the corresponding generator is an intrinsic object. By Theorem 3.9, the Brownian motion

~~(Xt) constructed above is a solution of the martingale problem fortheoperator~~

1 n+1 1 n+1 = (P )P = P 2. L 2 k ·∇ k ·∇ 2 k "k=1 "k=1 1 From differential geometry it is well-known that this operator is 2 ∆Mc where ∆Mc denotes the (intrinsic) Laplace-Beltrami operator on Mc.

~~Stochastic Analysis Andreas Eberle 3.2. STRATONOVICH DIFFERENTIAL EQUATIONS 115~~

Exercise (Itô SDE for Brownian motion on Mc). Prove that the SDE (3.18) can be written in Itô form as 1 dX = P (X ) dB κ(X )n(X ) dt t t t − 2 t t 1 n where κ(x)= n div (x) is the mean curvature of Mc at x.

~~Doss-Sussmann method~~

~~Stratonovich calculus can also be used to obtain explicit solutions for stochastic differ- n ential equations in R that are driven by a one-dimensional Brownian motion (Bt).We consider the SDE~~

dX = b(X ) dt + σ(X ) dB ,X= a, (3.19) ◦ t t t ◦ t 0 where a Rn, b : Rn Rn is Lipschitz continuous and σ : Rn Rn is C2 with ∈ → → bounded derivatives. Recall that (3.19) is equivalent to theItôSDE 1 dX = b + σ σ (X ) dt + σ(X ) dB ,X¸ = ¸a. (3.20) t 2 ·∇ t t t 0 $ % We ﬁrst determine an explicit solution in the case b 0 by the ansatz X = F (B ) ≡ t t where F C2(R, Rn).BytheStratonovichrule, ∈

~~dX = F ′(B ) dB = σ(F (B )) dB ◦ t t ◦ t t ◦ t provided F is a solution of the ordinary differential equation~~

~~F ′(s)=σ(F (s)). (3.21)~~

~~Hence a solution of (3.19) with initial condition X0 = a is given by~~

Xt = F (Bt,a) where (s, x) F (s, x) is the flow of the vector field σ,i.e.,F ( ,a) is the unique #→ · solution of (3.21) with initial condition a. Recall from the theory of ordinary differential equations that the flow of a vector field σ

University of Bonn Summer Semester 2015 116 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS as above deﬁnes a diffeomorphism a F (s, a) for any s R.Toobtainasolutionof #→ ∈ (3.19) in the general case, we try the “variation of constants” ansatz

Xt = F (Bt,Ct) (3.22) with a continuous semimartingale (Ct) satisfying C0 = a.Inotherwords:wemakea time-dependent coordinate transformation in the SDE that isdeterminedbytheflowF and the driving Brownian path (Bt).Byapplyingthechainruleto(3.22),weobtain ∂F ∂F dX = (B ,C) dB + (B ,C) dC ◦ t ∂s t t ◦ t ∂x t t ◦ t ∂F = σ(X ) dB + (B ,C ) dC t ◦ t ∂x t t ◦ t where ∂F (s, ) denotes the Jacobi matrix of the diffeomorphism F (s, ).Hence(X ) is ∂x · · t asolutionoftheSDE(3.19)provided(Ct) is almost surely absolutely continuous with derivative d ∂F 1 C = (B ,C)− b(F (B ,C)). (3.23) dt t ∂x t t t t For every given ω,theequation(3.23)isanordinarydifferentialequationfor Ct(ω) which has a unique solution. Working out these arguments in detail yields the following result: Theorem 3.11 (Doss 1977, Sussmann 1978). Suppose that b : Rn Rn is Lipschitz → continuous and σ : Rn Rn is C2 with bounded derivatives. Then the flow F of the → vector field σ is well-defined, F (s, ) is a C2 diffeomorphism for any s R,andthe · ∈ equation (3.23) has a unique pathwise solution (Ct)t 0 satisfying C0 = a.Moreover, ≥ the process Xt = F (Bt,Ct) is the unique strong solution of the equation (3.19), (3.20) respectively.

We refer to [25] for a detailed proof. Exercise (Computing explicit solutions). Solve the following Itô stochastic differential equations explicitly: 1 dX = X dt + 1+X2 dB ,X=0, (3.24) t 2 t t t 0 2 2 dXt = Xt(1 + XtD) dt +(1+Xt ) dBt,X0 =1. (3.25) Do the solutions explode in ﬁnite time?

~~Stochastic Analysis Andreas Eberle 3.2. STRATONOVICH DIFFERENTIAL EQUATIONS 117~~

~~Exercise (Variation of constants). We consider nonlinear stochastic differential equations of the form~~

~~dXt = f(t, Xt) dt + c(t)Xt dBt,X0 = x, where f : R+ R R and c : R+ R are continuous (deterministic) functions. × → → Proceed as follows :~~

~~a) Find an explicit solution Z of the equation with f 0. t ≡ b) To solve the equation in the general case, use the Ansatz~~

X = C Z . t t · t Show that the SDE gets the form dC (ω) t = f(t, Z (ω) C (ω))/Z (ω); C = x. (3.26) dt t · t t 0 Note that for each ω Ω,thisisadeterministic differential equation for the ∈ function t C (ω).Wecanthereforesolve(3.26)withω as a parameter to ﬁnd #→ t Ct(ω).

~~c) Apply this method to solve the stochastic differential equation 1 dXt = dt + αXt dBt ; X0 = x>0 , Xt where α is constant.~~

~~d) Apply the method to study the solution of the stochastic differential equation~~

~~γ dXt = Xt dt + αXt dBt ; X0 = x>0 ,~~

~~where α and γ are constants. For which values of γ do we get explosion?~~

~~Wong Zakai approximations of SDE~~

~~AnaturalwaytoapproximatethesolutionofanSDEdrivenbyaBrownian motion is to replace the Brownian motion by a smooth approximation. Theresultingequationcan~~

University of Bonn Summer Semester 2015 118 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS then be solved pathwise as an ordinary differential equation. It turns out that the limit of this type of approximations as the driving smoothed processes converge to Brownian motion will usually solve the corresponding Stratonovich equation.

d Suppose that (Bt)t 0 is a Brownian motion in R with B0 =0.Fornotationalconve- ≥ nience we define Bt := 0 for t<0.WeapproximateB by the smooth processes 2 (k) 1/2 t B := B⋆ϕ ,ϕ(t)=(2πε)− exp . 1/k ε − 2ε Other smooth approximations could be used as well, cf. [25] and [23].$ Let%X(k) denote the unique solution to the ordinary differential equation d d X(k) = b(X(k))+σ(X(k)) B(k),X(k) = a (3.27) dt t t t dt t 0 n n n n d with coefficients b : R R and σ : R R × . → → Theorem 3.12 (Wong, Zakai 1965). Suppose that b is C1 with bounded derivatives and σ is C2 with bounded derivatives. Then almost surely as k , →∞ X(k) X uniformly on compact intervals, t −→ t where (Xt) is the unique solution of the Stratonovich equation dX = b(X ) dt + σ(X ) dB ,X= a. ◦ t t t ◦ t 0 If the driving Brownian motion is one-dimensional, there is asimpleproofbasedon the Doss-Sussman representation of solutions. This shows that X(k) and X can be (k) (k) (k) represented in the form Xt = F (Bt ,Ct ) and Xt = F (Bt,Ct) with the flow F of the same vector field σ,andtheprocessesC(k) and C solving (3.23) w.r.t. B(k), B respectively. Therefore, it is not difficult to verify that almost surely, X(k) X → uniformly on compact time intervals, cf. [25]. The proof in the more interesting general case is much more involved, cf. e.g. Ikeda & Watanabe [23, Ch. VI, Thm. 7.2].

~~3.3 Stochastic Taylor expansions~~

In the next section we will study numerical schemes for Itô stochastic differential equations of type d k dXt = b(Xt) dt + σk(Xt) dBt (3.28) "k=1 Stochastic Analysis Andreas Eberle 3.3. STOCHASTIC TAYLOR EXPANSIONS 119 in RN , N N.Akeytoolforderivingandanalyzingsuchschemesarestochastic ∈ Taylor expansions that are introduced in this section.

We will assume throughout the next two sections that the coefﬁcients b, σ1,...,σd are N 1 d C∞ vector ﬁelds on R ,andB =(B ,...,B ) is a d-dimensional Brownian motion. Below, it will be convenient to set

~~0 Bt := t.~~

~~Asolutionof(3.28)satisﬁes~~

t+h d t+h k Xt+h = Xt + b(Xs) ds + σk(Xs) dBs (3.29) ˆt ˆt "k=1 for any t, h 0.Byapproximatingb(X ) and σ (X ) in (3.29) by b(X ) and σ (X ) re- ≥ s k s t k t spectively, we obtain an Euler approximation of the solutionwithstepsizeh.Similarly, higher order numerical schemes can be obtained by approximating b(Xs) and σk(Xs) by stochastic Taylor approximations.

~~Itô-Taylor expansions~~

~~N Suppose that X is a solution of (3.28), and let f C∞(R ).ThentheItô-Doeblin ∈ formula for f(X) on the interval [t, t + h] can be written in the compact form~~

d t+h k f(Xt+h)=f(Xt)+ ( kf)(Xs) dBs (3.30) ˆt L "k=0 for any t, h 0,where B0 = t, a = σσT , ≥ t 1 N ∂2f f = aij + b f, and (3.31) L0 2 ∂xi∂xj ·∇ i,j=1 " f = σ f, for k =1,...,d. (3.32) Lk k ·∇ By iterating this formula, we obtain Itô-Taylor expansions for f(X).Forexample,a ﬁrst iteration yields

~~d t+h d t+h s k l k f(Xt+h)=f(Xt)+ ( kf)(Xt) dBs + ( l kf)(Xr) dBr dBs . L ˆt ˆt ˆt L L "k=0 k,l"=0 University of Bonn Summer Semester 2015 120 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS~~

The ﬁrst two terms on the right hand side constitute a ﬁrst order Taylor expansion for f(X) in terms of the processes Bk, k =0, 1,...,d,andtheiteratedItôintegralinthe third term is the corresponding remainder. Similarly, we obtain higher order expansions in terms of iterated Itô integrals where the remainders are given by higher order iterated integrals, cf. Theorem 3.14 below. The next lemma yields L2 bounds on the remainder terms:

~~Lemma 3.13. Suppose that G :Ω (t, t + h) R is an adapted process in 2(P × → L ⊗ λ(t,t+h)).Then~~

2 t+h s1 sn−1 n+m(k) kn k2 k1 h 2 E Gsn dBsn dBs2 dBs1 sup E Gs ˆt ˆt ···ˆt ··· ≤ n! s (t,t+h) 8- . ; ∈ # ' for any n N and k =(k ,...,k ) 0, 1,...,d n,where ∈ 1 n ∈{ } m(k) := 1 i n : k =0 |{ ≤ ≤ i }| denotes the number of integrations w.r.t. dt.

~~Proof. By Itô’s isometry and the Cauchy-Schwarz inequality,~~

t+h 2 t+h E G dBk E G2 ds for any k =0,and ˆ s s ≤ ˆ s ̸ 8- t . ; t t+h 2 t+h # ' E G ds h E G2 ds. ˆ s ≤ ˆ s 8- t . ; t # ' By iteratively applying these estimates we see that the second moment of the iterated integral in the assertion is bounded from above by

~~t+h s1 sn−1 m(k) 2 h E[Gsn ] dsn ds2 ds1. ˆt ˆt ···ˆt ···~~

The lemma can be applied to control the strong convergence order of stochastic Taylor expansions. For k N we denote by Ck(R) the space of all Ck functions with bounded ∈ b k derivatives up to order k.NoticethatwedonotassumethatthefunctionsinCb are bounded.

~~Stochastic Analysis Andreas Eberle 3.3. STOCHASTIC TAYLOR EXPANSIONS 121~~

~~Deﬁnition (Stochastic convergence order). Suppose that Ah, h>0,andA are random variables, and let α>0.~~

~~2 1) Ah converges to A with strong L order α iff~~

~~1/2 E A A 2 = O(hα). | h − | # ' 2) Ah converges to A with weak order α iff~~

~~α 2(α+1) E [f(A )] E [f(A)] = O(h ) for any f C⌈ ⌉(R). h − ∈ b~~

Notice that convergence with strong order α requires that the random variables are de- ﬁned on a common probability space. For convergence with weakorderα this is not necessary. If Ah converges to A with strong order α then we also write

~~A = A + (hα). h O~~

Examples. 1) If B is a Brownian motion then Bt+h converges to Bt almost surely as h 0.Bythelawoftheiteratedlogarithm,thepathwiseconvergence order is ↓ 1/2 1 B B = O(h log log h− ) almost surely. t+h − t On the other hand, the strong L2 order is 1/2,andtheweakorderis1 since by Kol- mogorov’s forward equation,

t+h 1 h E[f(Bt+h)] E[f(Bt)] = E[ ∆f(Bs)] ds sup ∆f − ˆt 2 ≤ 2 for any f C2.Theexercisebelowshowsthatsimilarstatementsholdformore general ∈ b Itô diffusions. 2) The n-fold iterated Itô integrals w.r.t. Brownian motion considered in Lemma 3.13 have strong order (n + m)/2 where m is the number of time integrals.

Exercise (Order of Convergence for Itô diffusions). Let (Xt)t 0 be an N-dimensional ≥ stochastic process satisfying the SDE (3.28) where b, σ : RN RN ,k=1,...,d,are k → bounded continuous functions, and B is a d-dimensional Brownian motion. Prove that as h 0, ↓ University of Bonn Summer Semester 2015 122 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS

~~2 1) Xt+h converges to Xt with strong L order 1/2.~~

~~2) Xt+h converges to Xt with weak order 1.~~

Theorem 3.14 (Itô-Taylor expansion with remainder of order α). Suppose that α = j/2 for some j N.IfX is a solution of (3.28) with coefﬁcients b, σ1,...,σd 2α ∈ 2α+1 ∈ C⌊ ⌋(RN , RN ) then the following expansions hold for any f C⌊ ⌋(RN ): b ∈ b

f(X )= − f (X ) (3.33) t+h Lkn Lkn 1 ···Lk1 t × n<2α k:n+m(k)<2α " " $ % t+h s1 sn−1 kn k2 k1 α Gsn dBsn dBs2 dBs1 + (h ), × ˆt ˆt ···ˆt ··· O hn E [f(X )] = E [( nf)(X )] + O(hα). (3.34) t+h L0 t n! n<α " Proof. Iteration of the Itô-Doeblin formula (3.30) shows that (3.33) holds with a remainder term that is a sum of iterated integrals of the form

t+h s1 sn−1 kn k2 k1 kn kn−1 k1 f (Xsn ) dBsn dBs2 dBs1 ˆt ˆt ···ˆt L L ···L ··· $ % with k =(k1,...,kn) satisfying n + m(k) > 2α and n 1+m(k1,...,kn 1) < 2α. − − By Lemma 3.13, these iterated integrals are of strong L2 order (n + m(k))/2.Hence the full remainder term is of the order (hα). O Equation (3.34) follows easily by iterating the Kolmogorov forward equation

t+h E [f(Xt+h)] = E [f(Xt)] + E [( 0f)(Xs)] ds. ˆt L Alternatively, it can be derived from (3.33) by noting that all iterated integrals involving at least one integration w.r.t. a Brownian motion have mean zero.

~~Remark (Computation of iterated Itô integrals). Iterated Itô integrals involving only asingleonedimensionalBrownianmotionB can be computed explicitly from the Brownian increments. Indeed,~~

~~t+h s1 sn−1~~

~~dBsn dBs2 dBs1 = hn(h, Bt+h Bt)/n!, ˆt ˆt ···ˆt ··· −~~

~~Stochastic Analysis Andreas Eberle 3.3. STOCHASTIC TAYLOR EXPANSIONS 123~~

where hn denotes the n-th Hermite polynomial, cf. (5.57). In the multi-dimensional case, however, the iterated Itô integrals can not be represented in closed form as functions of Brownian increments. Therefore, in higher order numerical schemes, these integrals have to be approximated separately. For example, the second iterated Itô integral h s h kl k l k l Ih = dBr dBs = Bs dBs ˆ0 ˆ0 ˆ0 kl lk k l of two components of a d dimensional Brownian motion satisﬁes Ih + Ih = BhBh. Hence the symmetric part can be computed easily. However, theantisymmetricpart Ikl Ilk is the Lévy area process of the two dimensional Brownian motion (Bk,Bl). h − h The Lévy area can not be computed explicitly from the increments if k = l.Controlling ̸ the Lévy area is crucial for a pathwise stochastic integration theory, cf. [18, 19, 29].

Exercise (Lévy Area). If c(t)=(x(t),y(t)) is a smooth curve in R2 with c(0) = 0, then t t t A(t)= (x(s)y′(s) y(s)x′(s)) ds = xdy ydx ˆ0 − ˆ0 − ˆ0 describes the area that is covered by the secant from the origin to c(s) in the interval

~~[0,t].Analogously,foratwo-dimensionalBrownianmotionBt =(Xt,Yt) with B0 =0, one deﬁnes the Lévy Area~~

~~t t At := Xs dYs Ys dXs . ˆ0 − ˆ0 1) Let α(t), β(t) be C1-functions, p R,and ∈ α(t) V = ipA X2 + Y 2 + β(t) . t t − 2 t t~~

~~Vt $ % 2 2 Show using Itô’s formula, that e is a local martingale provided α′(t)=α(t) p − and β′(t)=α(t).~~

~~2) Let t [0, ).Thesolutionsoftheordinarydifferentialequationsforα and β 0 ∈ ∞ with α(t0)=β(t0)=0are~~

~~α(t)=p tanh(p (t t)) , · · 0 − β(t)= log cosh(p (t t)) . − · 0 − University of Bonn Summer Semester 2015 124 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS~~

~~Conclude that ipA 1 E e t0 = p R . cosh(pt0) ∀ ∈ # ' 3) Show that the distribution of At is absolutely continuous with density 1 fAt (x)= πx . 2t cosh( 2t )~~

~~3.4 Numerical schemes for SDE~~

~~Let X be a solution of the SDE~~

d k dXt = b(Xt) dt + σk(Xt) dBt (3.35) "k=1 where we impose the same assumptions on the coefficients as in the last section. By applying the Itô-Doeblin formula to σk(Xs) and taking into account all terms up to strong order (h1),weobtaintheItô-Taylorexpansion O d X X = b(X ) h + σ (X )(Bk Bh) (3.36) t+h − t t k t t+h − t "k=1 d t+h s + (σ σ )(X ) dBl dBk + h3/2 . l ·∇ k t ˆ ˆ r s O k,l=1 t t " $ % Here the first term on the right hand side has strong L2 order (h),thesecondterm O (h1/2),andthethirdterm (h).Takingintoaccountonlythefirsttwotermsleadsto O O the Euler-Maruyama scheme with step size h,whereastakingintoaccountalltermsup to order (h) yields the Milstein scheme: O Euler-Maruyama scheme with step size h • d Xh Xh = b(Xh) h + σ (Xh)(Bk Bh)(t =0,h,2h, 3h, . . .) t+h − t t k t t+h − t "k=1 Milstein scheme with step size h • d d t+h s h h h h k h h l k Xt+h Xt = b(Xt ) h + σk(Xt )(Bt+h Bt )+ (σl σk)(Xt ) dBr dBs − − ·∇ ˆt ˆt "k=1 k,l"=1

~~Stochastic Analysis Andreas Eberle 3.4. NUMERICAL SCHEMES FOR SDE 125~~

The Euler and Milstein scheme provide approximations to the solution of the SDE (3.35) that are deﬁned for integer multiples t of the step size h.Forasingleapproxi- mation step, the strong order of accuracy is (h) for Euler and (h3/2) for Milstein. O O To analyse the total approximation error it is convenient to extend the deﬁnition of the approximation schemes to all t 0 by considering the delay stochastic differential ≥ equations

h h h k dXs = b(X s ) ds + σk(X s ) dBs , (3.37) ⌊ ⌋h ⌊ ⌋h k " s dXh = b(Xh ) ds + σ (Xh )+(σ σ )(Xh ) dBl dBk (3.38) s s h k s h l k s h ˆ r s ⌊ ⌋ ⌊ ⌋ ∇ ⌊ ⌋ s h "k,l - ⌊ ⌋ . respectively, where s := max t hZ : t s ⌊ ⌋h { ∈ ≤ } denotes the next discretization time below s.Noticethatindeed,theEulerandMilstein scheme with step size h are obtained by evaluating the solutions of (3.37) and (3.38) respectively at t = kh with k Z . ∈ +

~~Strong convergence order~~

Fix a RN ,letX be a solution of (3.28) with initial condition X = a,andletXh be ∈ 0 acorrespondingEulerorMilsteinapproximationsatisfying(3.37),(3.38)respectively h with initial condition X0 = a. Theorem 3.15 (Strong order for Euler and Milstein scheme). Let t [0, ). ∈ ∞

~~1) Suppose that the coefﬁcients b and σk are bounded and Lipschitz continuous. Then the Euler-Maruyama approximation on the time interval [0,t] has strong L2 order 1/2 in the following sense:~~

h 1/2 sup Xs Xs = (h ). s t − O ≤ & & & & 2 2) If, moreover, the coefﬁcients b and σk are C with bounded derivatives then the Milstein approximation on the time interval [0,t] has strong L2 order 1,i.e.,

~~Xh X = (h). t − t O & & University of Bonn & & Summer Semester 2015 126 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS~~

AcorrespondinguniformintimeestimatefortheMilsteinapproximation also holds but the proof is too long for these notes. The assumptions on the coefficients in the theorem are not optimal and can be weakened, see e.g. Milstein and Tretyakov [33]. However, it is well-known that even in the deterministic case a local Lipschitz condition is not sufficient to guarantee convergence of the Euler approximations. The iterated integral in the Milstein scheme can be approximated by a Fourier expansion in such a way that the strong order (h) still holds, cf. Kloeden and Platen [26,33]XXX O Proof. For notational simplicity, we only prove the theorem in the one-dimensional case. The proof in higher dimensions is analogous. The basic idea is to write down an SDE for the approximation error X Xh. − h 1) By (3.37) and since X0 = X0,thedifferenceoftheEulerapproximationandthe solution of the SDE satisfies the equation t t h h h Xt Xt = b(X s h ) b(Xs) ds + σ(X s h ) σ(Xs) dBs. − ˆ0 ⌊ ⌋ − ˆ0 ⌊ ⌋ − This enables us to estimate$ the mean square% error $ %

h h 2 ε¯t := E sup Xs Xs . s t − 4 ≤ 5 & & By the Cauchy-Schwarz inequality and by Doob’s& L2 inequality,& t t h h 2 h 2 ε¯t 2t E b(X s h ) b(Xs) ds +8 E σ(X s h ) σ(Xs) ds ≤ ˆ0 ⌊ ⌋ − ˆ0 ⌊ ⌋ − /& t & 0 /& & 0 & 2 h & 2 & & (2t +8) L E X s h Xs ds (3.39) ≤ · · ˆ0 ⌊ ⌋ − t/& & 0 (4t +16) L2 ε¯&h ds + C h& , ≤ · · ˆ s t - 0 . where t C is an increasing real-valued function, and L is a joint Lipschitz constant #→ t for b and σ.Here,wehaveusedthatbythetriangleinequality,

h 2 h h 2 h 2 E X s Xs 2 E X s Xs +2E Xs Xs , ⌊ ⌋h − ≤ ⌊ ⌋h − − /& & 0 /& & 0 /& & 0 and the ﬁrst& term representing& the additional& error& by the time& discretization& on the interval [ s , s + h] is of order O(h) uniformly on ﬁnite time intervals by a similar ⌊ ⌋h ⌊ ⌋h argument as in Theorem 3.14. By (3.39) and Gronwall’s inequality, we conclude that

ε¯h (4t +16)L2C exp (4t +16)L2t h, t ≤ t · · $ % Stochastic Analysis Andreas Eberle 3.4. NUMERICAL SCHEMES FOR SDE 127 and hence ε¯h = O(√h) for any t (0, ).ThisprovestheassertionfortheEuler t ∈ ∞ scheme. D 2) To prove the assertion for the Milstein scheme we have to argue more carefully. We will show that h h 2 εt := sup E Xs Xs s t − ≤ /& & 0 is of order O(h2).Noticethatnowthesupremumisinfrontoftheexpectation,& & i.e., we are considering a weaker error than for the Euler scheme. We ﬁrst derive an equation (and not just an estimate as above) for the mean square error. By (3.38), the difference of the Milstein approximation and the solution of the SDE satisﬁes

~~t h h Xt Xt = b(Xs) b(X s h ) ds (3.40) − ˆ0 − ⌊ ⌋ $ t % h h + σ(Xs) σ(X s h ) (σσ′)(X s h )(Bs B s h ) dBs. ˆ0 − ⌊ ⌋ − ⌊ ⌋ − ⌊ ⌋ $ % By Itô’s formula, we obtain~~

~~t h 2 h h h Xt Xt =2 (X X ) d(X X )+[X X ]t | − | ˆ0 − − − t t t h h h h h 2 =2 (Xs Xs ) βs ds +2 (Xs Xs ) αs dBs + αs ds ˆ0 − ˆ0 − ˆ0 | |~~

h h h h h where βs = b(Xs) b(X s ) and αs = σ(Xs) σ(X s ) (σσ′)(X s )(Bs B s h ) − ⌊ ⌋h − ⌊ ⌋h − ⌊ ⌋h − ⌊ ⌋ are the integrands in (3.40). The assumptions on the coefﬁcients guarantee that the stochastic integral is a martingale. Therefore, we obtain

t t h 2 h h h 2 E Xt Xt =2 E (Xs Xs ) βs ds + E αs ds. (3.41) | − | ˆ0 − ˆ0 | | # ' # ' # ' We will now show that the integrands on the right side of (3.41)canbeboundedbya h 2 constant times εs + h .TheassertionthenfollowssimilarlyasabovebyGronwall’s inequality.

~~In order to bound E[ αh 2] we decompose αh = αh + αh where | s | s s,0 s,1~~

h αs,1 = σ(Xs) σ(X s h ) (σσ′)(X s h )(Bs B s h ) − ⌊ ⌋ − ⌊ ⌋ − ⌊ ⌋ University of Bonn Summer Semester 2015 128 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS is an additional error introduced in the current step, and

h h h αs,0 = σ(X s h ) σ(X s )+ (σσ′)(X s h ) (σσ′)(X s ) (Bs B s h ) ⌊ ⌋ − ⌊ ⌋h ⌊ ⌋ − ⌊ ⌋h − ⌊ ⌋ is an error carried over from previous$ steps. By the error estimate% in the Itô-Taylor expansion, αh is of strong order (h) uniformly in s,i.e., s,1 O E[ αh 2] C h2 for some ﬁnite constant C . | s,1| ≤ 1 1 B Furthermore, since Bs B s h is independent of s , − ⌊ ⌋ F⌊ ⌋h h 2 2 h 2 2 h E[ αs,0 ] 2(1 + h)L E X s h X s 2(1 + h)L εs , | | ≤ | ⌊ ⌋ − ⌊ ⌋h | ≤ and hence # '

E[ αh 2] C (h2 + εh) for some ﬁnite constant C . (3.42) | s | ≤ 2 s 2 It remains to prove an analogue bound for E[(X Xh) βh].Similarlyasabove,we s − s s h h h decompose βs = βs,0 + βs,1 where

~~h h h βs,0 = b(X s h ) b(X s ) and βs,1 = b(Xs) b(X s h ). ⌊ ⌋ − ⌊ ⌋h − ⌊ ⌋ By the Cauchy-Schwarz inequality and the Lipschitz continuity of b,~~

~~1/2 1/2 E[(X Xh) βh ] εh E βh 2 Lεh. (3.43) s − s s,0 ≤ s | s,0| ≤ s $ % # ' Moreover, there is a ﬁnite constant C3 such that~~

h h h B E (X s h X s ) βs,1 = E (X s h X s ) E b(Xs) b(X s h ) s ⌊ ⌋ − ⌊ ⌋h ⌊ ⌋ − ⌊ ⌋h − ⌊ ⌋ |F h 1/2 2 h # ' C3#h ε C3 (h#+ ε ). ''(3.44) ≤ s ≤ s Here we have used that by Kolmogorov’s$ equation,% s B B E b(Xs) b(X s h ) s = E ( 0b)(Xr) s dr, (3.45) ⌊ ⌋ − |F ˆ s h L |F # ' ⌊ ⌋ # ' and b is bounded by the assumptions on b and σ. L0 h h h Finally, let Zs := (Xs Xs ) (X s h X s ).By(3.40), − − ⌊ ⌋ − ⌊ ⌋h s s h h h Zs = βr dr + αr dBr, and ˆ s ˆ s ⌊ ⌋h ⌊ ⌋h Stochastic Analysis Andreas Eberle 3.4. NUMERICAL SCHEMES FOR SDE 129

s s h 2 h 2 h 2 2 h E Zs 2h E βr dr +2 E αr dr C4 h (h + εs ). | | ≤ ˆ s | | ˆ s | | ≤ ⌊ ⌋h ⌊ ⌋h # ' # ' h h h# ' Here we have used the decomposition βs = βs,0 + βs,1 and (3.42). Hence

~~h h h 2 h 1/2 2 h E[Zs βs,1] Zs 2 b(Xs) b(X s h ) 2 C5 h (h + εs ) 2 C5 (h + εs ). ≤ L − ⌊ ⌋ L ≤ ≤ F F F F By combiningF thisF estimateF with (3.44) andF (3.4), we eventually obtain~~

~~E[(X Xh) βh] C (h2 + εh) for some ﬁnite constant C . (3.46) s − s s ≤ 6 s 6~~

~~Weak convergence order~~

~~We will now prove under appropriate assumptions on the coefﬁcients that the Euler scheme has weak convergence order h1.Let~~

1 N ∂2f f = aij + b f L 2 ∂xi∂xj ·∇ i,j=1 " denote the generator of the diffusion process (Xt).Weassumethatthecoefﬁcients b, σ ,...,σ are in C3(RN , RN ).Itcanbeshownthatundertheseconditions,forf 1 d b ∈ 3 RN Cb ( ),theKolmogorovbackwardequation ∂u (t, x)=( u)(t, x),u(0,x)=f(x), (3.47) ∂t L has a unique classical solution u : [0, ) RN R such that u(t, ) C3(RN ) for any ∞ × → · ∈ b t 0,cf.XXX.Moreover,if(X ) is the unique strong solution of (3.28) with X = a, ≥ t 0 then by Itô’s formula,

~~E[f(Xt)] = u(t, a).~~

Theorem 3.16 (Weak order one for Euler scheme). Suppose that b, σ ,...,σ 1 d ∈ 3 RN RN h Cb ( , ),andlet(Xt) and (Xt ) denote the unique solution of (3.28) with X0 = a and its Euler approximation, respectively. Then

~~E[f(Xh)] E[f(X )] = O(h) for any t 0 and f C3(RN ). t − t ≥ ∈ b~~

~~University of Bonn Summer Semester 2015 130 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS~~

Proof. Fix t 0.Thekeyidea(thatiscommonwithmanyotherproofs)istoconsider ≥ the “interpolation” A := u(t s, Xh) for s [0,t]. s − s ∈ h h Notice that At = u(0,Xt )=f(Xt ) and A0 = u(t, a)=E[f(Xt)],whence

E[f(Xh)] E[f(X )] = E[A A ]. (3.48) t − t t − 0 We can now bound the weak error by applying Itô’s formula. Indeed, by (3.37) and (3.47) we obtain t ∂u h h h At A0 = Mt + (t s, Xs )+( s u)(t s, X0:s) ds − ˆ0 − ∂t − L − t 4 5 h h h = Mt + ( s u)(t s, X0:s) ( u)(t s, Xs ) ds. ˆ0 L − − L − # ' Here Mt is a martingale, Y0:t := (Ys)s [0,t],and ∈ N 2 h 1 ij ∂ f ( t f)(x0:t)= a (x t h ) (xt)+b(x t h ) f(xt) L 2 ⌊ ⌋ ∂xi∂xj ⌊ ⌋ ·∇ i,j=1 " is the generator at time t of the delay equation (3.37) satisfied by the Euler scheme. h Note that t (x0:t) is similar to (xt) but the coefficients are evaluated at x t h instead L L ⌊ ⌋ of xt.Takingexpectationsweconclude t h h h E[At A0]= E ( s u)(t s, X0:s) ( u)(t s, Xs ) ds. − ˆ0 L − − L − # ' Thus the proof is complete if we can show that there is a finite constant C such that

( hu)(t s, Xh ) ( u)(t s, Xh) Ch for s [0,t] and h (0, 1]. (3.49) Ls − 0:s − L − s ≤ ∈ ∈ & & This& is not difﬁcult to verify by the assumptions& on the coefﬁcients. For instance, let us assume for simplicity that d =1and b 0,andleta = σ2.Then ≡ ( hu)(t s, Xh ) ( u)(t s, Xh) Ls − 0:s − L − s 1 h h h & E a(Xs ) a(X s ) u′′(t & s, Xs ) & ≤ 2 − ⌊ ⌋h −& 1 & #$ h h % B '& h &E E a(Xs ) a(X s ) s u′′(t &s, X s ) ≤ 2 − ⌊ ⌋h |F⌊ ⌋h − ⌊ ⌋h &1 # # h h ' h '& h +& E a(Xs ) a(X s ) u′′(t s, Xs ) u′′(t & s, X s ) . 2 − ⌊ ⌋h − − − ⌊ ⌋h & #$ %$ %'& Stochastic Analysis& Andreas& Eberle 3.5. LOCAL TIME 131

Since u′′ is bounded, the ﬁrst summand on the right hand side is of order O(h),cp. (3.45). By the Cauchy-Schwarz inequality, the second summand is also of order O(h). Hence (3.49) is satisﬁed in this case. The proof in the generalcaseissimilar.

~~Remark (Generalizations).~~

~~1) The Euler scheme is given by~~

~~∆Xh = b(Xh) h + σ(Xh)∆B , ∆B independent N(0,hI ),t hZ . t t t t t ∼ d ∈ +~~

~~It can be shown that weak order one still holds if the ∆Bt are replaced by arbitrary~~

~~i.i.d. random variables with mean zero, covariance hId,andthirdmomentsof order O(h2),cf.[26].~~

~~2) The Milstein scheme also has weak order h1,soitdoesnotimproveonEuler w.r.t. weak convergence order. Higher weak order schemes areduetoMilstein and Talay, see e.g. [33].~~

~~3.5 Local time~~

The occupation time of a Borel set U R by a one-dimensional Brownian motion (B ) ⊆ t is given by t U Lt = IU (Bs) ds. ˆ0 Brownian local time is an occupation time density for Brownian motion that is informally given by t a “ Lt = δa(Bs) ds ” ˆ0 for any a R.Itisanon-decreasingstochasticprocesssatisfying ∈

U a Lt = Lt da. ˆU We will now apply stochastic integration theory for general predictable integrands to a R deﬁne the local time process (Lt )t 0 for a rigorously for Brownian motion, and, ≥ ∈ more generally, for continuous semimartingales.

~~University of Bonn Summer Semester 2015 132 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS~~

~~Local time of continuous semimartingales~~

~~Let (Xt) be a continuous semimartingale on a ﬁltered probability space. Note that by Itô’s formula,~~

t 1 t f(Xt) f(X0)= f ′(Xs) dXs + f ′′(Xs) d[X]s. − ˆ0 2 ˆ0 Informally, if X is a Brownian motion then the last integral on the right hand side a should coincide with Lt if f ′′ = δa.Aconvexfunctionwithsecondderivativeδa is + f(x)=(x a) .Notingthattheleftderivativeoff is given by f ′ = I(a, ),this − − ∞ motivates the following deﬁnition:

~~Deﬁnition. For a continuous semimartingale X and a R,theprocessLa deﬁned by ∈ t + + 1 a (Xt a) (X0 a) = I(a, )(Xs) dXs + Lt − − − ˆ0 ∞ 2 is called the local time of X at a.~~

Remark. 1) By approximating the indicator function by continuous functions it can be easily veriﬁed that the process I(a, )(Xs) is predictable and integrable w.r.t. X. ∞ 2) Alternatively, we could have deﬁned local time at a by the identity

t + + 1 â (Xt a) (X0 a) = I[a, )(Xs) dXs + Lt − − − ˆ0 ∞ 2 involving the right derivative I[a, ) instead of the left derivative I(a, ).Notethat ∞ ∞ t a â Lt Lt = I a (Xs) dXs. − ˆ0 { } This difference vanishes almost surely if X is a Brownian motion, or, more generally, acontinuouslocalmartingale.Forsemimartingales,however, the processes La and Lâ may disagree, cf. the example below Lemma 3.17. The choice of La in the definition of local time is then just a standard convention that is consistent with the convention of considering left derivatives of convex functions.

~~Lemma 3.17 (Properties of local time, Tanaka formulae).~~

~~Stochastic Analysis Andreas Eberle 3.5. LOCAL TIME 133~~

1) Suppose that ϕ : R [0, ), n N,isasequenceofcontinuousfunctionswith n → ∞ ∈ ϕn =1and ϕn(x)=0for x (a, a +1/n).Then ´ ̸∈ t a Lt =ucp lim ϕn(Xs) d[X]s. n ˆ − →∞ 0 a In particular, the process (Lt )t 0 is non-decreasing and continuous. ≥ 2) The process La grows only when X = a,i.e.,

~~t a I Xs=a dLs =0 for any t 0. ˆ0 { ̸ } ≥ 3) The following identities hold:~~

t + + 1 a (Xt a) (X0 a) = I(a, )(Xs) dXs + Lt , (3.50) − − − ˆ0 ∞ 2 t 1 a (Xt a)− (X0 a)− = I( ,a](Xs) dXs + Lt , (3.51) − − − − ˆ0 −∞ 2 t a Xt a X0 a = sgn (Xs a) dXs + Lt , (3.52) | − |−| − | ˆ0 − where sgn(x) := +1 for x>0,andsgn(x) := 1 for x 0. − ≤ Remark. Note that we set sgn(0) := 1.Thisisrelatedtoourconventionofusingleft − derivatives as sgn(x) is the left derivative of x .ThereareanalogueTanakaformulae | | for Lˆa with the intervals (a, ) and ( ,a] replaced by [a, ) and ( ,a),andthe ∞ −∞ ∞ −∞ sign function deﬁned by sgnˆ (x) := +1 for x 0 and sgn(x) := 1 for x<0. ≥ − x y 2 Proof. 1) For n N let fn(x) := ϕn(z) dz dy.Thenthefunctionfn is C ∈ ´−∞ ´−∞ with fn′′ = ϕn.ByItô’sformula, t 1 t fn(Xt) fn(X0) fn′ (Xs) dXs = ϕn(Xs) d[X]s. (3.53) − − ˆ0 2 ˆ0

~~As n , fn′ (Xs) converges pointwise to I(a, )(Xs).Hence →∞ ∞ t t~~

~~fn′ (Xs) dXs I(a, )(Xs) dXs ˆ0 → ˆ0 ∞ in the ucp-sense by the Dominated Convergence Theorem 5.34. Moreover,~~

~~f (X ) f (X ) (X a)+ (X a)+. n t − n 0 → t − − 0 − University of Bonn Summer Semester 2015 134 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS~~

The first assertion now follows from (3.53). R a 2) By 1), the measures ϕn(Xt) d[X]t on + converge weakly to the measure dLt with a distribution function L .HencebythePortemanteauTheorem,andsinceϕn(x)=0for x (a, a +1/n), ̸∈ t t a I Xs a >ε dLs lim inf I Xs a >ε ϕn(Xs) d[X]s =0 ˆ {| − | } n ˆ {| − | } 0 ≤ →∞ 0 for any ε>0.ThesecondassertionofthelemmanowfollowsbytheMonotone Convergence Theorem as ε 0. ↓ 3) The first Tanaka formula (3.50) holds by definition of La.Moreover,subtracting (3.51) from (3.50) yields

~~t (Xt a) (X0 a)= dXs, − − − ˆ0 which is a valid equation. Therefore, the formulae (3.51) and(3.50)areequivalent. Finally, (3.52) follows by adding (3.50) and (3.51).~~

~~Remark. In the proof above it is essential that the Dirac sequence (ϕn) approximates~~

δa from the right.IfX is a continuous martingale then the assertion 1) of the lemma also holds under the assumption that ϕn vanishes on the complement of the interval (a 1/n, a+1/n).Forsemimartingaleshowever,approximatingδ from the left would − a lead to an approximation of the process Lˆa,whichingeneralmaydifferfromLa.

Exercise (Brownian local time). Show that the local time of a Brownian motion B in a R is given by ∈ t a 1 Lt =ucp lim I(a ε,a+ε)(Bs) ds. ε 0 ˆ − − → 2ε 0 Example (Reﬂected Brownian motion). Suppose that X = B where (B ) is a one- t | t| t dimensional Brownian motion starting at 0.ByTanaka’sformula(3.52),X is a semimartingale with decomposition

~~Xt = Wt + Lt (3.54)~~

~~Stochastic Analysis Andreas Eberle 3.5. LOCAL TIME 135~~

t where Lt is the local time at 0 of the Brownian motion B and Wt := 0 sgn(Bs) dBs. By Lévy’s characterization, the martingale W is also a Brownian motion,´ cf. Theorem X 2.2. We now compute the local time Lt of X at 0.By(3.51)andLemma3.17,2), t 1 X Lt = Xt− X0− + I( ,0](Xs) dXs (3.55) 2 − ˆ0 −∞ t t t = I 0 (Bs) dWs + I 0 (Bs) dLs = dLs = Lt a.s., ˆ0 { } ˆ0 { } ˆ0

~~X t i.e., Lt =2Lt.Herewehaveusedthat I 0 (Bs) dWs vanishes almost surely by Itô’s 0 { } isometry, as both W and B are Brownian´ motions. Notice that on the other hand,~~

t 1 ˆX Lt = Xt− X0− + I( ,0)(Xs) dXs =0 a.s., 2 − ˆ0 −∞ so the processes LX and LˆX do not coincide. By (3.54) and (3.55), the process X solves the singular SDE 1 dX = dW + dLX t t 2 t driven by the Brownian motion W .ThisjustiﬁesthinkingofX as Brownian motion reﬂected at 0.

~~The identity (3.54) can be used to compute the law of Brownian local time:~~

~~Exercise (The law of Brownian local time).~~

a) Prove Skorohod’s Lemma:If(yt)t 0 is a real-valued continuous function with ≥ y =0then there exists a unique pair (x, k) of functions on [0, ) such that 0 ∞ (i) x = y + k, (ii) x is non-negative, and

~~(iii) k is non-decreasing, continuous, vanishing at zero, and the measure dkt is carried by the set t : x =0 . { t }~~

~~The function k is given by kt =sups t( ys). ≤ −~~

~~b) Conclude that the local time process (Lt) at 0 of a one-dimensional Brownian~~

motion (Bt) starting at 0 and the maximum process St := sups t Bs have the ≤ same law. In particular, L B for any t 0. t ∼| t| ≥ University of Bonn Summer Semester 2015 136 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS

~~c) More generally, show that the two-dimensional processes ( B ,L) and (S B,S) | | − have the same law.~~

Notice that the maximum process (St)t 0 is the generalized inverse of the Lévy subor- ≥ dinator (Ta)a 0 introduced in Section 1.1. Thus we have identiﬁed Brownian local time ≥ at 0 as the inverse of a Lévy subordinator.

~~Itô-Tanaka formula~~

Local time can be used to extend Itô’s formula in dimension onefromC2 to general convex functions. Recall that a function f : R R is convex iff → f(λx +(1 λ)y) λf(x)+(1 λ)f(y) λ [0, 1],x,y R. − ≤ − ∀ ∈ ∈ For a convex function f,theleftderivatives

~~f(x) f(x h) f ′ (x) = lim − − − h 0 h ↓ exist, the function f ′ is left-continuous and non-decreasing, and − b f(b) f(a)= f ′ (x) dx for any a, b R. − ˆa − ∈~~

~~The second derivative of f in the distributional sense is the positive measure f ′′ given by~~

f ′′([a, b)) = f ′ (b) f ′ (a) for any a, b R. − − − ∈ We will prove in Theorem 3.24 below that there is a version (t, a) La of the local #→ t time process of a continuous semimartingale X such that t La is continuous and #→ t a La is càdlàg. If X is a local martingale then La is even jointly continuous in t and #→ t t a.Fromnowon,weﬁxacorrespondingversion.

Theorem 3.18 (Itô-Tanaka formula, Meyer). Suppose that X is a continuous semimartingale, and f : R R is convex. Then → t 1 a f(Xt) f(X0)= f ′ (Xs) dXs + Lt f ′′(da). (3.56) − ˆ0 − 2 ˆR

~~Stochastic Analysis Andreas Eberle 3.5. LOCAL TIME 137~~

~~Proof. We proceed in several steps: 1) Equation (3.56) holds for linear functions f. 2) By localization, we may assume that X~~

supp(f ′′) [ C, C]. ⊆ − Moreover, by subtracting a linear function and multiplying f by a constant, we may even assume that f vanishes on ( ,C],andf ′′ is a probability measure. Then −∞ x f ′ (y)=µ( ,y) and f(x)= µ( ,y) dy (3.57) − −∞ ˆ −∞ −∞ where µ := f ′′. + 3) Now suppose that µ = δa is a Dirac measure. Then f ′ = I(a, ) and f(x)=(x a) . − ∞ − Hence Equation (3.56) holds by definition of La.Moregenerally,bylinearity,(3.56) holds whenever µ has finite support, since then µ is a convex combination of Dirac measures. 4) Finally, if µ is a general probability measure then we approximate µ by measures with finite support. Suppose that Z is a random variable with distribution µ,andlet n n µn denote the law of Zn := 2− 2 Z .By3),theItô-Tanakaformulaholdsforthe x ⌈ ⌉ functions fn(x) := µn( ,y) dy,i.e., −∞ −∞ ´ t 1 a fn(Xt) fn(X0)= fn′ (Xs) dXs + Lt µn(da) (3.58) − ˆ0 − 2 ˆR for any n N.Asn , µ ( ,X ) µ( ,X ),andhence ∈ →∞ n −∞ s → −∞ s t t

fn′ (Xs) dXs f ′ (Xs) dXs ˆ0 − → ˆ0 − in the ucp sense by dominated convergence. Similarly, f (X ) f (X ) f(X ) n t − n 0 → t − f(X ).Finally,therightcontinuityofa La implies that 0 #→ t a a Lt µn(da) Lt µ(da), ˆR → ˆR since Zn converges to Z from above. The Itô-Tanaka formula (3.56) for f now follows from (3.58) as n . →∞ University of Bonn Summer Semester 2015 138 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS

Clearly, the Itô-Tanaka formula also holds for functions f that are the difference of two convex functions. If f is C2 then by comparing the Itô-Tanaka formula and Itô’s a formula, we can identify the integral Lt f ′′(da) over a as the stochastic time integral t ´ 0 f ′′(Xs) d[X]s.Thesameremainstruewheneverthemeasuref ′′(da) is absolutely ´ continuous with density denoted by f ′′(a):

Corollary 3.19. For any measurable function V : R [0, ), → ∞ t a Lt V (a) da = V (Xs) d[X]s t 0. (3.59) ˆR ˆ0 ∀ ≥ Proof. The assertion holds for any continuous function V : R [0, ) as V can be → ∞ represented as the second derivative of a C2 function f.Theextensiontomeasurable non-negative functions now follows by a monotone class argument.

~~Notice that for V = IB,theexpressionin(3.59)istheoccupationtimeofthesetB by~~

~~(Xt),measuredw.r.t.thequadraticvariationd[X]t.~~

~~3.6 Continuous modiﬁcations and stochastic ﬂows~~

~~d Let Ω=C0(R+, R ) endowed with Wiener measure µ0 and the canonical Brownian motion Wt(ω)=ω(t).WeconsidertheSDE~~

~~dXt = bt(X) dt + σt(X) dWt,X0 = a, (3.60)~~

~~n n n d with progressively measurable coefﬁcients b, σ : R C(R , R ) R , R × respec- + × + → tively satisfying the global Lipschitz condition~~

~~b (x) b (x) + σ (x) σ (x) L (x x)⋆ t, x, x (3.61) | t − t | || t − t || ≤ − t ∀ R for some ﬁnite constant) L +,aswellas) ) ) ∈~~

sup bs(0) + σs(0) < t. (3.62) s [0,t] | | || || ∞∀ ∈ $ % Then by Itô’s existence and uniqueness theorem, there existsauniqueglobalstrong a Rn solution (Xt )t 0 of (3.60) for any initial condition a .Ournextgoalistoshow ≥ ∈ Stochastic Analysis Andreas Eberle 3.6. CONTINUOUS MODIFICATIONS AND STOCHASTIC FLOWS 139 that there is a continuous modiﬁcation (t, a) ξa of (Xa).Theproofisbasedonthe #→ t t multidimensional version of the Kolmogorov-Centsovˇ continuity criterion for stochastic processes that is signiﬁcant in many other contexts as well. Therefore, we start with a derivation of the Kolmogorov-Centsovˇ criterion from a corresponding regularity result for deterministic functions.

~~Continuous modiﬁcations of deterministic functions~~

Let x : [0, 1)d E be a bounded measurable function from the d-dimensional unit → cube to a separable Banach space (E, ).Intheapplicationsbelow,E will either be ∥·∥ Rn Rn or C([0,t], ) endowed with the supremum norm. The average of x =(xu)u [0,1)d ∈ over a smaller cube Q [0, 1)d is denoted by x : ⊆ Q 1 xQ = xu du = xu du. Q vol(Q) ˆQ

d n n Let be the collection of all dyadic cubes Q = [(k 1)2− ,k2− ) with Dn i=1 i − i k ,...,k 1, 2,...,2n .Foru [0, 1)d and n N,wedenotetheuniquecube 1 d ∈{ } ∈ ∈G in containing u by Q (u).Noticethatu x is the conditional expectation Dn n #→ Qn(u) of x given σ( ) w.r.t. the uniform distribution on the unit cube. By the martingale Dn convergence theorem,

~~d xu = lim xQn(u) for almost every u [0, 1) , n →∞ ∈ where the limit is w.r.t. weak convergence if E is inﬁnite dimensional.~~

~~Theorem 3.20 (Besov-Hölder embedding). Let β>2d and q 1,andsupposethat ≥ q 1/q xu xv Bβ,q := ∥ − ∥ du dv (3.63) β ˆ[0,1)d ˆ[0,1)d ( u v /√d) - | − | . is ﬁnite. Then the limit~~

~~xu := lim xQn(u) n →∞ exists for every u [0, 1)d,andx is a Hölder continuous modiﬁcation of x satisfying ∈ )~~

~~8 β (β 2d)/q xu xv ) Bβ,q u v − . (3.64) | − ∥≤log 2 β 2d | − | − ) ) University of Bonn Summer Semester 2015 140 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS~~

β d For s = −q < 1,theconstantBβ,q is essentially a Besov norm of order (s, q, q), or equivalently, a Sobolev-Slobodecki norm of order (s, q).Theassertionofthetheo- rem says that the corresponding Besov space is continuously embedded into the Hölder space of order (β 2d)/q,i.e.,thereisaﬁniteconstantC such that −

~~x Höl((β 2d)/q) C x Besov((β d)/q,q,q). ∥ ∥ − ≤ ·∥ ∥ − Proof. Let e(Q) denote) the edge length of a cube Q.Thekeystepintheproofisto show that the inequality~~

~~4 β (β 2d)/q x x ˆ B e(Qˆ) − (3.65) ∥ Q − Q∥≤log 2 β 2d β,q − holds for arbitrary cubes Q, Qˆ (0, 1]d such that Q Qˆ.Thisinequalityisprovenby ⊆ ⊆ a chaining argument:Let~~

Qˆ = Q Q Q = Q 0 ⊃ 1 ⊃···⊃ n be a decreasing sequence of a subcubes that interpolates between Qˆ and Q .Weassume that the edge lengths ek := e(Qk) satisfy 1 1 eβ/q = eβ/q for k 1, and eβ/q eβ/q. (3.66) k+1 2 k ≥ 1 ≥ 2 0

d Since vol(Qk)=ek and u v √dek 1 for any u, v Qk 1,weobtain | − |≤ − ∈ − 1/q x x = (x x ) du dv x x q du dv Qk Qk−1 u v u v − F Qk Qk−1 − F ≤ 9 Qk Qk−1 ∥ − ∥ : F F F F F F 1/q F F F q F F xu xv F d/q d/q β/q ∥ − ∥ du dv ek− ek− 1 ek 1 ˆ ˆ √ β − − ≤ 9 Qk Qk−1 ( u v / d) : | − | (β 2d)/q (β 2d)/q (β 2d)k/β 2 B e − 4 B e(Q) − 2− − . ≤ β,q k ≤ β,q

In the last two steps, we have used (3.66) and ek 1 ek.Notingthat − ≥ ∞ ak a 2− =1/(2 1) 1/(a log 2), − ≤ "k=1 n Equation (3.65) follows since x x ˆ x x − . ∥ Q − Q∥≤ k=1 ∥ Qk − Qk 1 ∥ ! Stochastic Analysis Andreas Eberle 3.6. CONTINUOUS MODIFICATIONS AND STOCHASTIC FLOWS 141

Next, consider arbitrary dyadic cubes Q (u) and Q (v) with u, v [0, 1)d and n, m n m ∈ ∈ N.ThenthereisacubeQˆ [0, 1)d such that Qˆ Q (u) Q (v) and ⊆ ⊃ n ∪ m n m e(Qˆ) u v +2− +2− . ≤| − | By (3.65) and the triangle inequality, we obtain

x x x x ˆ + x ˆ x (3.67) ∥ Qn(u) − Qm(v)∥≤∥Qn(u) − Q∥ ∥ Q − Qm(v)∥ 8 β n m (β 2d)/q B u v +2− +2− − . ≤ log 2 β 2d β,q | − | − $ % Choosing v = u in (3.67), we see that the limit xu = limn xQn(u) exists. Moreover, →∞ for v = u,theestimate(3.64)followsasn, m . ̸ →∞) Remark (Garsia-Rodemich-Rumsey). Theorem 3.20 is a special case of a result by

Garsia, Rodemich and Rumsey where the powers in the deﬁnitionofBβ,q are replaced by more general increasing functions, cf. e.g. the appendix in [19]. This result allows to analyze the modulus of continuity more carefully, with important applications to Gaussian random ﬁelds [4].

~~Continuous modiﬁcations of random ﬁelds~~

~~The Kolmogorov-Centsovˇ continuity criterion for stochastic processes and random ﬁelds is a direct consequence of Theorem 3.20:~~

Theorem 3.21 (Kolmogorov, Centsovˇ ). Suppose that (E, ) is a Banach space, || · || C = d I is a product of bounded real intervals I ,...,I R,andX :Ω E, k=1 k 1 d ⊂ u → u C,isanE-valued stochastic process (a random ﬁeld) indexed by C.Ifthereexists ∈ G constants q, c, ε R such that ∈ + E X X q c u v d+ε for any u, v C, (3.68) || u − v|| ≤ | − | ∈ # ' then there exists a modiﬁcation (ξu)u C of (Xu)u C such that ∈ ∈ q ξu ξv E sup || − α|| < for any α [0,ε/q). (3.69) u=v u v ∞ ∈ /6 ̸ | − | 7 0 In particular, u ξ is almost surely α-Hölder continuous for any α<ε/q. #→ u University of Bonn Summer Semester 2015 142 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS

~~Adirectproofbasedonachainingargumentcanbefoundinmanytextbooks,see e.g. [37, Ch. I, (2.1)]. Here, we deduce the result as a corollary to the Besov-Hölder embedding theorem:~~

Proof. By rescaling we may assume w.l.o.g. that C = [0, 1)d.Forβ>0,theassumption (3.68) implies q Xu Xv d+ε β E ∥ − ∥ du dv c u v − du dv (3.70) ˆ ˆ u v β ≤ ˆ ˆ | − | 4 C C | − | 5 C C √d d+ε β d 1 const. r − r − dr. ≤ ˆ0 Hence the expectation is ﬁnite for β<2d + ε,andinthiscase, X X q ∥ u − v∥ du dv < almost surely. ˆ ˆ u v β ∞ C C | − |

Thus by Theorem 3.20, ξu = lim supn XQn(u) deﬁnes a modiﬁcation of (Xu) that →∞ is almost surely Hölder continuous with parameter (β 2d)/q for any β<2d + ε. − Moreover, the expectation of the q-th power of the Hölder norm is bounded by a multiple of the expectation in (3.70).

Example (Hölder continuity of Brownian motion). Brownian motion satisfies (3.68) with d =1and ε = γ 1 for any γ (2, ).Lettingγ tend to ,weseethatalmost 2 − ∈ ∞ ∞ every Brownian path is α-Hölder continuous for any α<1/2.Thisresultissharpinthe 1 sense that almost every Brownian path is not 2 -Hölder-continuous, cf. [14, Thm. 1.20]. In a similar way, one can study the continuity properties of general Gaussian random fields, cf. Adler and Taylor [4]. Another very important application of the Besov-Hölder embedding and the resulting bounds for the modulus of continuity are tightness results for families of stochastic processes or random random fields,seee.g.StroockandVarad- han [40]. Here, we consider two different applications that concern the continuity of stochastic flows and of local times.

~~Existence of a continuous ﬂow~~

~~We now apply the Kolmogorov-Centsovˇ continuity criterion to the solution a (Xa) #→ s of the SDE (3.60) as a function of its starting point.~~

~~Stochastic Analysis Andreas Eberle 3.6. CONTINUOUS MODIFICATIONS AND STOCHASTIC FLOWS 143~~

~~Theorem 3.22 (Flow of an SDE). Suppose that (3.61) and (3.62) hold.~~

1) There exists a function ξ : Rn Ω C(R , Rn), (a, ω) ξa(ω) such that × → + #→ a a Rn (i) ξ =(ξt )t 0 is a strong solution of (3.60) for any a ,and ≥ ∈ (ii) the map a ξa(ω) is continuous w.r.t. uniform convergence on ﬁnite time #→ intervals for any ω Ω. ∈

2) If σ(t, x)=σ(xt) and b(t, x)=b(xt) with Lipschitz continuous functions n n d n n d σ : R R × and b : R R × then ξ satisﬁes the cocycle property → ) → ) a ) ξa (ω)=) ξξt (ω)(Θ (ω)) s, t 0,a Rn (3.71) t+s s t ∀ ≥ ∈

~~for µ0-almost every ω,where~~

~~Θ (ω)=ω( + t) C(R , Rd) t · ∈ + denotes the shifted path, and the deﬁnition of ξ has been extended by~~

ξ(ω) := ξ(ω ω(0)) (3.72) − to paths ω C(R , Rd) with starting point ω(0) =0. ∈ + ̸ Proof. 1) We fix p>d.BytheaprioriestimateinTheorem3.1thereexistsafinite constant c R such that ∈ + E[(Xa X"a)⋆p] c ect a a p for any t 0 and a, a Rn, (3.73) − t ≤ · | − | ≥ ∈ where Xa denotes a version of the strong) solution of (3.60) with initial) condition a. Now fix t R .WeapplytheKolmogorov-Centsovˇ Theorem with E = C([0,t], Rn) ∈ + endowed with the supremum norm X = X⋆.By(3.73),thereexistsamodificationξ || ||t t a a of (Xs )s t,a Rn such that a (ξs )s t is almost surely α-Hölder continuous w.r.t. t ≤ ∈ #→ ≤ ∥·∥ p n a for any α< − .Clearly,fort t ,thealmostsurelycontinuousmap(s, a) ξ p 1 ≤ 2 #→ s constructed on [0,t ] Rn coincides almost surely with the restriction of the correspond- 1 × ing map on [0,t ] Rn.HencewecanalmostsurelyextendthedefinitiontoR Rn 2 × + × in a consistent way.

~~University of Bonn Summer Semester 2015 144 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS~~

~~2) Fix t 0 and a Rn.Thenµ -almost surely, both sides of (3.71) solve the same ≥ ∈ 0 SDE as a function of s.Indeed,~~

t+s t+s a a a a ξt+s = ξt + b(ξu) du + σ(ξu) dWu ˆt ˆt s s a )a a = ξt + b(ξt+r) dr + σ(ξ)t+r) d (Wr Θt) , ˆ0 ˆ0 ◦ ) s ) s a a a ξt a ξt ξt ξs Θt = ξt + b ξr Θt dr + σ(ξr Θt) d(Wr Θt) ◦ ˆ0 ◦ ˆ0 ◦ ◦ $ % hold µ0-almost surely for any s )0 where r Wr Θt = Wr+t is again a Brownian a ≥a #→ ◦ ) ξt ξt (ω) motion, and ξr Θ (ω) := ξr (Θ (ω)).Pathwiseuniquenessnowimplies ◦ t t $ % a ξa = ξξt Θ for any s 0, almost surely. t+s s ◦ t ≥ Continuity of ξ then shows that the cocycle property (3.71) holds with probability one for all s, t and a simultaneously.

Remark (Extensions). 1) Joint Hölder continuity in t and a: Since the constant p in the proof above can be chosen arbitrarily large, the argument yields α-Hölder continuity of a ξa for any α<1.ByapplyingKolmogorov’scriterionindimensionn+1,itisalso #→ possible to prove joint Hölder continuity in t and a.InSection4.1wewillprovethat under a stronger assumption on the coefficients b and σ,theflowisevencontinuously differentiable in a. 2) SDE with jumps: The first part of Theorem 3.22 extends to solutions of SDE of type (3.3) driven by a Brownian motion and a Poisson point process.Inthatcase,undera global Lipschitz condition the same arguments go through if we replace C([0,t], Rn) by the Banach space ([0,t], Rn) when applying Kolmogorov’s criterion. Hence in spite D of the jumps, the solution depends continuously on the initial value a ! 3) Locally Lipschitz coefficients: By localization, the existence of a continuous flow can also be shown under local Lipschitz conditions, cf. e.g. [36]. Notice that in this case, the explosion time depends on the initial value.

~~Above we have shown the existence of a continuous ﬂow for the SDE (3.60) on the canonical setup. From this we can obtain strong solutions on other setups:~~

~~Stochastic Analysis Andreas Eberle 3.6. CONTINUOUS MODIFICATIONS AND STOCHASTIC FLOWS 145~~

~~Exercise. Show that the unique strong solution of (3.60) w.r.t. an arbitrary driving a a Brownian motion B instead of W is given by Xt (ω)=ξt (B(ω)).~~

~~Markov property~~

~~In the time-homogeneous diffusion case, the Markov propertyforsolutionsoftheSDE (3.60) is a direct consequence of the cocycle property:~~

Corollary 3.23. Suppose that σ(t, x)=σ(xt) and b(t, x)=b(xt) with Lipschitz contin- Rn Rn d Rn Rn a uous functions σ : × and b : .Then(ξt )t 0 is a time-homogeneous → → ≥ W,P ) ) ( t ) Markov process with transition function F ) ) p (a, B)=P [ξa B],t0,a Rn. t t ∈ ≥ ∈ Proof. Let f : Rn R be a measurable function. Then for 0 s t, → ≤ ≤ Θ (ω)=ω(t)+ ω(t + ) ω(t) , t · − $ % and hence, by the cocycle property and by (3.72),

~~a f(ξa (ω)) = f ξξt (ω) ω(t + ) ω(t) s+t s · − $ $ %% for a.e. ω.Sinceω(t+ ) ω(t) is a Brownian motion starting at 0 independent of W,P , · − Ft we obtain~~

~~W,P a E f(ξa ) (ω)=E f(ξξt (ω)) =(p f)(ξa(ω)) almost surely. s+t |Ft s s t # ' # '~~

~~Remark. Without pathwise uniqueness, both the cocycle and the Markovpropertydo not hold in general.~~

~~Continuity of local time~~

The Kolmogorov-Centsovcontinuityˇ criterion can also be applied to prove theexistence of a jointly continuous version (a, t) La of the local time of a continuous local #→ t University of Bonn Summer Semester 2015 146 CHAPTER 3. EXTENSIONS OF ITÔ CALCULUS martingale. More generally, recall that the local time of a continuous semimartingale X = M + A is deﬁned by the Tanaka formula

t t 1 a + + Lt =(X0 a) (Xt a) I(a, )(Xs) dMs I(a, )(Xs) dAs (3.74) 2 − − − − ˆ0 ∞ − ˆ0 ∞ almost surely for any a R. ∈ Theorem 3.24 (Yor). There exists a version (a, t) La of the local time process that #→ t is continuous in t and càdlàg in a with

t a a Lt Lt − =2 I Xs=a dAs. (3.75) − ˆ0 { } In particular, (a, t) La is jointly continuous if M is a continuous local martingale. #→ t Proof. By localization, we may assume that M is a bounded martingale and A has (1) + bounded total variation V (A).Themap(a, t) (Xt a) is jointly continuous in t ∞ #→ − and a.Moreover,bydominatedconvergence,

~~t a Zt := I(a, )(Xs) dAs ˆ0 ∞ is continuous in t and càdlàg in a with~~

~~t a a Zt Zt − = I a (Xs) dAs. − − ˆ0 { } Therefore it is sufﬁcient to prove that~~

t a Yt := I(a, )(Xs) dMs ˆ0 ∞ a R Rn has a version such that the map a (Ys )s t from to C([0,t], ) is continuous for #→ ≤ any t [0, ). ∈ ∞ Hence ﬁx t 0 and p 4.ByBurkholder’sinequality, ≥ ≥ s p ⋆p E Y a Y b = E sup I (X) dM (3.76) − t ˆ (a,b] 4 s

~~holds for any a~~

t t I(a,b](X) d[M]= I(a,b](X) d[X] ˆ0 ˆ0 t = 2 f(Xt) f(X0) f ′(X) dX − − − ˆ0 - t . 2 (1) (b a) +2 f ′(X) dM + b a V (A) ≤ − ˆ | − | t & 0 & & & & & 2 Here we have used in the last step that f ′ b a and f (b a) /2.Combining | |≤&| − | | |&≤ − this estimate with 3.76 and applying Burkholder’s inequality another time, we obtain

t p/4 a b ⋆p p/2 2 E Y Y C (p, t) b a + E f ′(X) d[M] − t ≤ 2 | − | ˆ 9 8- 0 . ;: /$ % 0 C (p, t) b a p/2 (1 + [M]p/4) ≤ 2 | − | t with a ﬁnite constant C (p, t).Theexistenceofacontinuousmodiﬁcationofa 2 #→ a ˇ (Ys )s t now follows from the Kolmogorov-Centsov Theorem. ≤

a Remark. 1) The proof shows that for a continuous local martingale, a (Ls )s t is #→ ≤ α-Hölder continuous for any α<1/2 and t R . ∈ + a ˆa 2) For a continuous semimartingale, Lt − = Lt by (3.75).

~~University of Bonn Summer Semester 2015 Chapter 4~~

~~Variations of parameters in SDE~~

In this chapter, we consider variations of parameters in stochastic differential equations. This leads to a ﬁrst introduction to basic concepts and results of Malliavin calculus. For amorethoroughintroductiontoMalliavincalculuswereferto [35], [34], [41], [23], [32] and [9].

Let µ denote Wiener measure on the Borel σ-algebra (Ω) over the Banach space Ω= B C ([0, 1], Rd) endowed with the supremum norm ω =sup ω(t) : t [0, 1] ,and 0 || || {| | ∈ } consider an SDE of type

dXt = b(Xt) dt + σ(Xt) dWt,X0 = x, (4.1) driven by the canonical Brownian motion Wt(ω)=ω(t).Inthischapter,wewillbein- terested in dependence of strong solutions on the initial condition and other parameters. The existence and uniqueness of strong solutions and of continuous stochastic ﬂows has already been studied in Sections 3.1 and 3.6. We are now going to prove differentiability of the solution w.r.t. variations of the initial condition and the coefﬁcients, see Section 4.1. A main goal will be to establish relations between different types of variations of (4.1):

Variations of the initial condition: x x(ε) • → Variations of the coefﬁcients: b(x) b(ε, x),σ(x) σ(ε, x) • → → Variations of the driving paths: W W + εH , (H ) adapted • t → t t t 148 4.1. VARIATIONS OF PARAMETERS IN SDE 149

Variations of the underlying probability measure: µ µε = Zε µ • → · Section 4.2 introduces the Malliavin gradient which is a derivative of a function on Wiener space (e.g. the solution of an SDE) w.r.t. variations of the Brownian path. Bis- mut’s integration by parts formula is an infinitesimal version of the Girsanov Theorem, which relates these variations to variations of Wiener measure. After a digression to representation theorems in Section 4.3, Section 4.4 discusses Malliavin derivatives of solutions of SDE and their connection to variations of the initial condition and the coefficients. As a consequence, we obtain first stability results for SDE from the Bismut integration by parts formula. Finally, Section 4.5 sketches briefly how Malliavin calculus can be applied to prove existence and smoothness of densitiesofsolutionsofSDE.This should give a first impression of a powerful technique that eventually leads to impres- sive results such as Malliavin’s stochastic proof of Hörmander’s theorem, cf. [21], [34].

~~4.1 Variations of parameters in SDE~~

~~We now consider a stochastic differential equation~~

d ε ε ε k ε dXt = b(ε,Xt ) dt + σk(ε,Xt ) dWt ,X0 = x(ε), (4.2) "k=1 on Rn with coefﬁcients and initial condition depending on a parameter ε U,where ∈ U is a convex neighbourhood of 0 in Rm, m N.Hereb, σ : U Rn Rn are ∈ k × → functions that are Lipschitz continuous in the second variable, and x : U Rn.We → ε already know that for any ε U,thereexistsauniquestrongsolution(Xt )t 0 of (4.2). ∈ ≥ For p [1, ) let ∈ ∞ 1/p ε ε p X p := E sup Xt . || || t [0,1] | | / ∈ 0 Exercise (Lipschitz dependence on ε). Prove that if the maps x, b and σk are all Lip- schitz continuous, then ε Xε is also Lipschitz continuous w.r.t. ,i.e.,thereexists #→ ||·||p aconstantL R such that p ∈ + Xε+h Xε L h , for any ε, h Rm with ε,ε + h U. || − ||p ≤ p | | ∈ ∈

~~University of Bonn Summer Semester 2015 150 CHAPTER 4. VARIATIONS OF PARAMETERS IN SDE~~

~~We now prove a stronger result under additional regularity assumptions.~~

~~Differentation of solutions w.r.t. a parameter~~

Theorem 4.1. Let p [2, ),andsupposethatx, b and σ are C2 with bounded ∈ ∞ k derivatives up to order 2.Thenthefunctionε Xε is differentiable on U w.r.t. , #→ || · ||p ε dXε and the differential Y = dε is the unique strong solution of the SDE

∂b ∂b dY ε = (ε,Xε)+ (ε,Xε)Y ε dt (4.3) t ∂ε t ∂x t t - . d ∂σ ∂σ + k (ε,Xε)+ k (ε,Xε)Y ε dW k, ∂ε t ∂x t t t k=1 ε " 6 7 Y0 = x′(ε), (4.4) that is obtained by formally differentiating (4.2) w.r.t. ε.

~~∂ ∂ Here and below ∂ε and ∂x denote the differential w.r.t. the ε and x variable, and x′ denotes the (total) differential of the function x.~~

ε ε Remark. Note that if (Xt ) is given, then (4.3) is a linear SDE for (Yt ) (with multiplicative noise). In particular, there is a unique strong solution. The SDE for the derivative process Y ε is particularly simple if σ is constant: In that case, (4.3) is a deterministic ODE with coefﬁcients depending on Xε.

Proof of 4.1. We prove the stronger statement that there is a constant M (0, ) such p ∈ ∞ that Xε+h Xε Y εh M h 2 (4.5) − − p ≤ p | | && && holds for any ε, h Rm&&with ε,ε + h U,where&& Y ε is the unique strong solution of ∈ ∈ (4.3). Indeed, by subtracting the equations satisﬁed by Xε+h, Xε and Y εh,weobtain for t [0, 1]: ∈

~~t d t ε+h ε ε k Xt Xt Yt h I + II ds + IIIk dW , − − ≤|| ˆ0 | | ˆ0 k=1 & & & & " & & & & & & Stochastic Analysis Andreas Eberle 4.1. VARIATIONS OF PARAMETERS IN SDE 151 where~~

~~I = x(ε + h) x(ε) x′(ε)h, − − ε+h ε ε h II = b(ε + h, X ) b(ε,X ) b′(ε,X ) , and − − 9Y εh:~~

~~ε+h ε ε h IIIk = σk(ε + h, X ) σk(ε,X ) σk′ (ε,X ) . − − 9Y εh:~~

~~Hence by Burkholder’s inequality, there exists a ﬁnite constant Cp such that~~

t d E (Xε+h Xε Y εh)⋆p C I p + E II p + III p ds . (4.6) − − t ≤ p · | | ˆ | | | k| 9 0 k=1 : # ' # " ' 2 Since x, b and σk are C with bounded derivatives, there exist ﬁnite constants CI, CII,

~~CIII such that~~

I C h 2, (4.7) | |≤ I| | ∂b II C h 2 + (ε,Xε)(Xε+h Xε Y εh) , (4.8) | |≤ II| | ∂x − − 2 & ∂σk ε ε+h ε ε & IIIk CIII h +& (ε,X )(X X Y h&) . (4.9) | |≤ | | ∂x − − & & Hence there exist ﬁnite constants Cp, C&ˆp such that &

d p E[ II p + III p] ) C h 2p + E Xε+h Xε Y εh , | | | k| ≤ p | | − − k=1 " $ #& & '% and thus, by (4.6) and (4.7), ) & & t ε+h ε ε ⋆p ˆ 2p ε+h ε ε ⋆p E (X X Y h)t Cp h + CpCp E (X X Y h)s ds − − ≤ | | ˆ0 − − # ' # ' for any t 1.Theassertion(4.5)nowfollowsbyGronwall’slemma.) ≤

~~Derivative ﬂow and stability of stochastic differential equations~~

We now apply the general result above to variations of the initial condition, i.e., we consider the ﬂow d x x x k x dξt = b(ξt ) dt + σk(ξt ) dWt ,ξ0 = x. (4.10) "k=1 University of Bonn Summer Semester 2015 152 CHAPTER 4. VARIATIONS OF PARAMETERS IN SDE

~~2 Assuming that b and σk (k =1,...,d)areC with bounded derivatives, Theorem 4.1 shows that the derivative ﬂow~~

x ∂ x,l ′ Yt := ξt(x)= k ξt ∂x 1 k,l n - . ≤ ≤ x exists w.r.t. p and (Yt )t 0 satisﬁes the SDE || · || ≥ d x x x x x k x dYt = b′(ξt ) Yt dt + σk′ (ξt ) Yt dWt ,Y0 = In. (4.11) "k=1 Note that again, this is a linear SDE for Y if ξ is given, and Y is the fundamental solution of this SDE.

~~Remark (Flow of diffeomorphisms). One can prove that x ξx(ω) is a diffeomor- #→ t phism on Rn for any t and ω,cf.[27]or[15].~~

~~In the sequel, we will denote the directional derivative of the ﬂow ξ in direction v Rn t ∈ by Yv,t: x x x Yv,t = Yv,t = Yt v = ∂vξt .~~

~~(i) Constant diffusion coefficients. Let us now first assume that d = n and σ(x)=In for any x Rn.ThentheSDEreads ∈ x x x dξ = b(ξ ) dt + dW, ξ0 = x; and the derivative flow solves the ODE~~

~~x x dY = b′(ξ )Ydt, Y0 = In.~~

~~This can be used to study the stability of solutions w.r.t. variations of initial conditions pathwise:~~

~~Theorem 4.2 (Exponential stability I). Suppose that b : Rn Rn is C2 with bounded → derivatives, and let~~

~~κ =supsup v b′(x)v. x Rn v Rn · ∈ v∈=1 | | Then for any t 0 and x, y, v Rn, ≥ ∈ ∂ ξx eκt v , and ξx ξy eκt x y . | v t |≤ | | | t − t |≤ | − |~~

~~Stochastic Analysis Andreas Eberle 4.1. VARIATIONS OF PARAMETERS IN SDE 153~~

~~The theorem shows in particular that exponential stability holds if κ<0.~~

~~x x Proof. The derivative Yv,t = ∂vξt satisﬁes the ODE~~

~~dYv = b′(ξ)Yv dt.~~

~~Hence 2 2 d Y =2Y b′(ξ)Y dt 2κ Y dt, | v| v · v ≤ | v| which implies~~

~~∂ ξx 2 = Y x 2 e2κt v 2, and thus | v t | | v,t| ≤ | | 1 x y (1 s)x+sy κt ξt ξt = ∂x yξt − ds e x y . | − | ˆ0 − ≤ | − | & & & & & &~~

~~n n Example (Ornstein-Uhlenbeck process). Let A R × .ThegeneralizedOrnstein- ∈ Uhlenbeck process solving the SDE~~

~~dξt = Aξt dt + dWt~~

n 1 is exponentially stable if κ =sup v Av : v S − < 0. { · ∈ } (ii) Non-constant diffusion coefficients. If the diffusion coefficients are not constant, the noise term in the SDE for the derivative flow does not vanish. Therefore, the derivative flow can not be bounded pathwise. Nevertheless, we can still obtain stability in an L2 sense.

Lemma 4.3. Suppose that b, σ ,...,σ : Rn Rn are C2 with bounded derivatives. 1 d → Then for any t 0 and x, v Rn,thederivativeﬂowY x = ∂ ξx is in L2(Ω, ,P), ≥ ∈ v,t v t A and d E[ Y x 2]=2E[Y x K(ξx)Y x ] dt | v,t| v,t · t v,t where d 1 T K(x)=b′(x)+ σ′ (x) σ′ (x). 2 k k "k=1

~~University of Bonn Summer Semester 2015 154 CHAPTER 4. VARIATIONS OF PARAMETERS IN SDE~~

~~(k) Proof. Let Yv denote the k-the component of Yv.TheItôproductruleyields~~

d Y 2 =2Y dY + d[Y (k)] | v| v · v v "k (4.11) k 2 =2Y b′(ξ)Y dt +2 Y σ′ (ξ) dW + σ′ (ξ)Y dt. v · v v · k | k v| "k "k Noting that the stochastic integrals on the right-hand side stopped at T = inf t 0: Y n are martingales, we obtain n { ≥ | v,t|≥ }

t Tn 2 2 ∧ E Yv,t Tn = v +2E Yv K(ξ)Yv ds . | ∧ | | | ˆ0 · # ' / 0 The assertion follows as n . →∞ Theorem 4.4 (Exponential stability II). Suppose that the assumptions in Lemma 4.3 hold, and let κ := sup sup v K(x)v. (4.12) x Rn v Rn · ∈ v∈=1 | | Then for any t 0 and x, y, v Rn, ≥ ∈

~~E[ ∂ ξx 2] e2κt v 2, and (4.13) | v t | ≤ | |~~

~~E[ ξx ξy 2]1/2 eκt x y . (4.14) | t − t | ≤ | − | Proof. Since K(x) κI holds in the form sense for any x,Lemma4.3implies ≤ n d E[ Y 2] 2κE[ Y 2]. dt | v,t| ≤ | v,t|~~

(4.13) now follows immediately by Gronwell’s lemma, and (4.14) follows from (4.13) y 1 (1 s)x+sy x − since ξt ξt = 0 ∂x yξt ds. − ´ − Remark. (Curvature) The quantity κ can be viewed as a lower curvature bound − w.r.t. the geometric structure deﬁned by the diffusion process. In particular, exponential stability w.r.t. the L2 norm holds if κ<0,i.e.,ifthecurvatureisboundedfrombelow by a strictly positive constant.

~~Stochastic Analysis Andreas Eberle 4.1. VARIATIONS OF PARAMETERS IN SDE 155~~

~~Consequences for the transition semigroup~~

~~x We still consider the ﬂow (ξt ) of the SDE (4.1) with assumptions as in Lemma 4.3 and Theorem 4.4. Let~~

~~p (x, B)=P [ξx B],xRn,B (Rn), t t ∈ ∈ ∈B denote the transition function of the diffusion process on Rn.Fortwoprobabilitymea- sures µ, ν on Rn,wedeﬁnetheL2 Wasserstein distance~~

2 1/2 2(µ, ν) = inf E[ X Y ] W (X,Y ) | − | X µ,Y ν ∼ ∼ as the infimum of the L2 distance among all couplings of µ and ν.Hereacouplingofµ and ν is defined as a pair (X, Y ) of random variables on a joint probability space with distributions X µ and Y ν.Letκ be defined as in (4.12). ∼ ∼ Corollary 4.5. For any t 0 and x, y Rn, ≥ ∈ p (x, ),p(y, ) eκt x y . W2 t · t · ≤ | − | Proof. The flow defines a coupling$ between%p (x, ) and p (y, ) for any t, x and y: t · t · ξx p (x, ),ξy p (y, ). t ∼ t · t ∼ t · Therefore, p (x, ),p(y, ) 2 E ξx ξy 2 . W2 t · t · ≤ | t − t | The assertion now follows$ from Theorem% 4.4. # '

Exercise (Exponential convergence to equilibrium). Suppose that µ is a stationary distribution for the diffusion process, i.e., µ is a probability measure on (Rn) satisfying B µp = µ for every t 0.Provethatifκ<0 and x 2 µ(dx) < ,thenforany t ≥ | | ∞ x Rd, p (x, ),µ 0 exponentially fast with rate´ κ as t . ∈ W2 t · → →∞ Besides studying$ convergence% to a stationary distribution,thederivativeﬂowisalso useful for computing and controlling derivatives of transtion functions. Let

~~(p f)(x)= p (x, dy)f(y)=E[f(ξx)] t ˆ t t denote the transition semigroup acting on functions f : Rn R.Westillassumethe → conditions from Lemma 4.3.~~

~~University of Bonn Summer Semester 2015 156 CHAPTER 4. VARIATIONS OF PARAMETERS IN SDE~~

Exercise (Lipschitz bound). Prove that for any Lipschitz continuous function f : Rn R, → p f eκt f t 0, || t ||Lip ≤ || ||Lip ∀ ≥ where f =sup f(x) f(y) / x y : x, y Rn s.t. x = y . || ||Lip {| − | | − | ∈ ̸ } For continuously differentiable functions f,weevenobtainanexplicitformulaforthe gradient of ptf:

Corollary 4.6 (First Bismut-Elworthy Formula). For any function f C1(Rn) and ∈ b t 0, p f is differentiable with ≥ t x n v p f = E Y x f x, v R . (4.15) ·∇x t v,t ·∇ξt ∀ ∈ # ' x Here p f denotes the gradient evaluated at x.NotethatY x f is the directional ∇x t t,v ·∇ξt x derivative of f in the direction of the derivative ﬂow Yt,v.

Proof of 4.6. For λ R 0 , ∈ \{ } λ (ptf)(x + λv) (ptf)(x) 1 x+λv x 1 x+sv − = E f(ξ ) f(ξ ) = E Y x+sv f ds. t t v,t ξt λ λ − λ ˆ0 ·∇ # ' # ' The assertion now follows since x ξx and x Y x are continuous, f is continuous #→ t #→ v,t ∇ and bounded, and the derivative ﬂow is bounded in L2.

The ﬁrst Bismut-Elworthy Formula shows that the gradient of ptf can be controlled by the gradient of f for all t 0.InSection4.4,wewillseethatbyapplyinganintegration ≥ by parts on the right hand side of (4.15), for t>0 it is even possible to control the gradient of ptf in terms of the supremum norm of f,providedanon-degeneracycondition holds, cf. (??).

~~4.2 Malliavin gradient and Bismut integration by parts formula~~

d Let Wt(ω)=ωt denote the canonical Brownian motion on Ω=C0([0, 1], R ) endowed with Wiener measure. In the sequel, we denote Wiener measure by P ,expectation values w.r.t. Wiener measure by E[ ],andthesupremumnormby . · || · || Stochastic Analysis Andreas Eberle 4.2. MALLIAVIN GRADIENT AND BISMUT INTEGRATION BY PARTS FORMULA 157

Deﬁnition. Let ω Ω.AfunctionF :Ω R is called Fréchet differentiable at ω iff ∈ → there exists a continuous linear functional d F :Ω R such that ω → F (ω + h) F (ω) (d F )(h) = o( h ) for any h Ω. || − − ω || || || ∈ If a function F is Fréchet differentiable at ω then the directional derivatives ∂F F (ω + εh) F (ω) (ω) = lim − =(dωF )(h) ε 0 ∂h → ε exist for all directions h Ω.Forapplicationsinstochasticanalysis,Fréchetdiffer- ∈ entiability is often too restrictive, because Ω contains “too many directions”. Indeed, solutions of SDE are typically not Fréchet differentiable asthefollowingexampleindi- cates:

1 1 2 1 2 Example. Let F = 0 Wt dWt where Wt =(Wt ,Wt ) is a two dimensional Brownian motion. A formal computation´ of the derivative of F in a direction h =(h1,h2) Ω ∈ yields 1 1 ∂F 1 2 1 2 = ht dWt + Wt dht . ∂h ˆ0 ˆ0 Clearly, this expression is NOT CONTINUOUS in h w.r.t. the supremum norm.

~~Amoresuitablespaceofdirectionsforcomputingderivatives of stochastic integrals is the Cameron-Martin space~~

~~d 2 d H = h : [0, 1] R : h =0,habs. contin. with h′ L ([0, 1], R )] . CM → 0 ∈ + , Recall that HCM is a Hilbert space with inner product~~

~~(h, g)H = ht′ gt′ dt, h, g HCM. ˆ0 · ∈~~

~~2 d The map h h′ is an isometry from H onto L ([0, 1], R ).Moreover,H is #→ CM CM continuously embedded into Ω,since~~

1 1/2 h =supht ht′ dt (h, h)H || || t [0,1] | |≤ˆ0 | | ≤ ∈ for any h H by the Cauchy Schwarz inequality. ∈ CM University of Bonn Summer Semester 2015 158 CHAPTER 4. VARIATIONS OF PARAMETERS IN SDE

As we will consider variations and directional derivatives in directions in HCM,itis convenient to think of the Cameron-Martin space as a tangent space to Ω at a given path ω Ω.WewillnowdefineagradientcorrespondingtotheCameron-Martin inner ∈ product in two steps: at first for smooth functions F :Ω R,andthenforfunctions → that are only weakly differentiable in a sense to be specified.

~~h 0 ω~~

~~Gradient and integration by parts for smooth functions~~

Let C1(Ω) denote the linear space consisting of all functions F :Ω R that are b → everywhere Fréchet differentiable with continuous boundedderivativedF :Ω Ω′, → ω d F .HereΩ′ denotes the space of continuous linear functionals l :Ω R #→ ω → endowed with the dual norm of the supremum norm, i.e.,

~~l ′ =supl(h):h Ω with h 1 . || ||Ω { ∈ || || ≤ } Deﬁnition (Malliavin Gradient I). Let F C1(R) and ω Ω. ∈ b ∈~~

~~H 1) The H-gradient (D F )(ω) is the unique element in HCM satisfying~~

~~∂F (DHF )(ω),h = (ω)=(d F )(h) for any h H . H ∂h ω ∈ CM $ % (4.16)~~

~~2) The Malliavin gradient (DF)(ω) is the function t (D F )(ω) in L2([0, 1], Rd) #→ t deﬁned by~~

~~d (D F )(ω)= (DH F )(ω)(t) for a.e. t [0, 1]. (4.17) t dt ∈~~

In other words, DH F is the usual gradient of F w.r.t. the Cameron-Martin inner product, and (DF)(ω) is the element in L2 [0, 1], Rd identiﬁed with (DH F )(ω) by the canoni- $ % Stochastic Analysis Andreas Eberle 4.2. MALLIAVIN GRADIENT AND BISMUT INTEGRATION BY PARTS FORMULA 159

~~2 d cal isometry h h′ between H and L ([0, 1], R ).Inparticular,foranyh H #→ CM ∈ CM and ω Ω, ∈~~

∂F H (ω)= h, (D F )(ω) =(h′, (DF)(ω)) 2 ∂h H L $ 1 % = ht′ (DtF )(ω) dt, (4.18) ˆ0 · and this identity characterizes DF completely. The examples given below should help to clarify the deﬁnitions.

Remark. 1) The existence of the H-gradient is guaranteed by the Riesz Representation The- orem. Indeed, for ω Ω and F C1(Ω),theFréchetdifferentiald F is a ∈ ∈ b ω continuous linear functional on Ω.SinceHCM is continuously embedded into

Ω,therestrictiontoHCM is a continuous linear functional on HCM w.r.t. the H- H norm. Hence there exists a unique element (D F )(ω) in HCM such that (4.16) holds. 2) By deﬁnition of the Malliavin gradient,

~~1 H 2 2 D F (ω) H = DtF (ω) dt. || || ˆ0 | |~~

~~3) Informally, one may think of DtF as a directional derivative of F in direction~~

~~I(t,1],because~~

~~1 d H H “ DtF = D F (t)= (D F )′ I(′t,1] = ∂I(t,1] F ”. dt ˆ0~~

~~Of course, this is a purely heuristic representation, since I(t,1] is not even continuous.~~

~~Example (Linear functions on Wiener space).~~

1) Brownian motion:ConsiderthefunctionF (ω)=W i(ω)=ωi ,wheres (0, 1] s s ∈ and i 1,...,d .Clearly,F is in C1(Ω) and ∈{ } b 1 ∂ i d i i i Ws = Ws + εhs ε=0 = hs = ht′ ei I(0,s)(t) dt ∂h dε ˆ0 · $ %& University of Bonn & Summer Semester 2015 160 CHAPTER 4. VARIATIONS OF PARAMETERS IN SDE

~~for any h H .Therefore,bythecharacterizationin(4.18),theMalliavin ∈ CM gradient of F is given by~~

D W i (ω)=e I (t) for every ω Ω and a.e. t (0, 1). t s i (0,s) ∈ ∈ $ % Since the function F :Ω R is linear, the gradient is deterministic. The H- → i gradient is obtained by integrating DWs :

~~t t H i i Dt Ws = DrWs dr = ei I(0,s) =(s t) ei. ˆ0 ˆ0 ∧~~

~~2) Wiener integrals:Moregenerally,let~~

~~1 F = gs dWs ˆ0 ·~~

~~where g : [0, 1] Rd is a C1 function. Integration by parts shows that → 1~~

~~F = g1 W1 gs′ Ws ds almost surely. (4.19) · − ˆ0 ·~~

~~The function on the right hand side of (4.19) is deﬁned for every ω,anditis Fréchet differentiable. Taking this expression as a pointwise deﬁnition for the stochastic integral F ,weobtain~~

~~∂F 1 1 = g1 h1 gs′ hs ds = gs hs′ ds ∂h · − ˆ0 · ˆ0 ·~~

~~for any h H .Therefore,by(4.18), ∈ CM t H DtF = gt and Dt F = gs ds. ˆ0~~

Theorem 4.7 (Integration by parts, Bismut). Let F C1(Ω) and G 2(Ω ∈ b ∈La × [0, 1] Rd,P λ).Then → ⊗ 1 1 E DtF Gt dt = E F Gt dWt . (4.20) ˆ0 · ˆ0 · / 0 / 0 Stochastic Analysis Andreas Eberle 4.2. MALLIAVIN GRADIENT AND BISMUT INTEGRATION BY PARTS FORMULA 161

t To recognize (4.20) as an integrationbyparts identityon Wiener space let Ht = 0 Gs¸ds. Then ´ 1 H DtF Gt dt = D F, H H = ∂H F. ˆ0 · Replacing F in (4.20) by F F with F, F $ C1(Ω),weobtaintheequivalentidentity% · ∈ b 1 E[F∂H F ]=) E[∂)H F F ]+E F F Gt dWt (4.21) − ˆ0 · / 0 by the product rule for the) directional derivative.) )

~~Proof of Theorem 4.7. The formula (4.21) is an inﬁnitesimal version of Girsanov’s The- orem. Indeed, suppose ﬁrst that G is bounded. Then, by Novikov’s criterion,~~

~~t 1 t ε ε 2 Zt =expε Gs dWs Gs ds ˆ0 · − 2 ˆ0 | | 6 7 R t is a martingale for any ε .HenceforHt = 0 Gs ds, ∈ ´ ε E[F (W + εH)] = E[F (W )Z1].~~

The equation (4.21) now follows formally by taking the derivative w.r.t. ε at ε =0. Rigorously, we have F (W + εH) F (W ) Zε 1 E − = E F (W ) 1 − . (4.22) ε ε / 0 / 0 As ε 0,therighthandsidein(4.22)convergestoE F (W ) t G dW ,since → 0 · 1 1 ´ 1 ε ε # '2 (Z1 1) = Z G dW G dW in L (P ). ε − ˆ0 · −→ ˆ0 · Similarly, by the Dominated Convergence Theorem, the left hand side in (4.22) converges to the left hand side in (4.21): 1 ε E (F (W + εH) F (W )) = E (∂H F )(W + sH) ds E[(∂H F )(W )] ε − ˆ0 −→ / 0 / 0 as ε 0 since F C1(Ω).Wehaveshownthat(4.21)holdsforboundedadaptedG. → ∈ b Moreover, the identity extends to any G 2(P λ) because both sides of (4.21) are ∈La ⊗ continuous in G w.r.t. the L2(P λ) norm. ⊗ Remark. Adaptedness of G is essential for the validity of the integration by parts identity.

~~University of Bonn Summer Semester 2015 162 CHAPTER 4. VARIATIONS OF PARAMETERS IN SDE~~

~~Skorokhod integral~~

~~The Bismut integration by parts formula shows that the adjoint of the Malliavin gradient coincides with the Itô integral on adapted processes. Indeed, the Malliavin gradient~~

D : C1(Ω) L2(Ω, ,P) L2(Ω [0, 1] Rd, ,P λ), b ⊆ A −→ × → A⊗B ⊗ F (DtF )0 t 1, #−→ ≤ ≤ is a densely deﬁned linear operator from the Hilbert space L2(Ω, ,P) to the Hilbert A space L2(Ω [0, 1] Rd, ,P λ).Let × → A⊗B ⊗ δ :Dom(δ) L2(Ω [0, 1] Rd, B,P λ) L2(Ω, ,P) ⊆ × → A⊗ ⊗ −→ A denote the adjoint operator (i.e., the divergence operator corresponding to the Malliavin gradient). By (4.21), any adapted process G 2(Ω [0, 1] Rd, ,P λ) is ∈L × ∈ A⊗B ⊗ contained in the domain of δ,and

~~1 2 δG = Gt dWt for any G a. ˆ0 · ∈L~~

Hence the divergence operator δ deﬁnes an extension of the Itô integral G 1 G dW #→ 0 t· t to not necessarily adapted square integrable processes G :Ω [0, 1] ´Rd.This × → extension is called the Skorokhod integral .

Exercise (Product rule for divergence). Suppose that (Gt)t [0,1] is adapted and bounded, ∈ 1 and F Cb (Ω).Provethattheprocess(F Gt)t [0,1] is contained in the domain of δ, ∈ · ∈ and 1 δ(FG)=Fδ(G) DtF Gt dt. − ˆ0 ·

~~Deﬁnition of Malliavin gradient II~~

~~So far we have deﬁned the Malliavin gradient only for continuously Fréchet differentiable functions F on Wiener space. We will now extend the deﬁnition to the Sobolev spaces D1,p, 1~~

~~In particular, we will be interested in the case p =2where~~

~~1 2 2 2 F 1,2 = E F + DtF dt . || || ˆ0 | | / 0 Theorem 4.8 (Closure of the Malliavin gradient).~~

~~1) There exists a unique extension of DH to a continuous linear operator~~

~~DH : D1,p LP (Ω H, P) −→ → .~~

~~2) The Bismut integration by parts formula holds for any F D1,2. ∈~~

~~1,2 Proof for p =2. 1) Let F D and let (Fn)n N be a Cauchy sequence w.r.t. the (1, 2) ∈ ∈ 1 2 norm of functions in Cb (Ω) converging to F in L (Ω,P).Wewouldliketodeﬁne~~

H H D F := lim D Fn (4.23) n →∞ w.r.t. convergence in the Hilbert space L2(Ω H, P).Thenon-trivialfacttobe → shown is that DH F is well-deﬁned by (4.23), i.e., independently of the approximating sequence. In functional analytic terms, this is the closability of the operator DH .

To verify closability, we apply the integration by parts identity. Let (Fn) and (Fn) be 2 approximating sequences as above, and let L = lim Fn and L = lim Fn in L (Ω,P). ) We have to show L = L.Tothisend,itsufﬁcestoshow ) ) (L L,) h) =0 almost surely for any h H. (4.24) − H ∈ Hence ﬁx h H,andlet) ϕ C2(Ω).Thenby(4.21), ∈ ∈ b

~~E[(L L, h)H ϕ] = lim E[∂h(Fn Fn) ϕ] n − · →∞ − · 1 ) = lim E (Fn F)n)ϕ h′ dW E (Fn Fn)∂hϕ n ˆ →∞ − 0 · − − + # 0 / 0, =0 ) )~~

University of Bonn Summer Semester 2015 164 CHAPTER 4. VARIATIONS OF PARAMETERS IN SDE since F F 0 in L2.AsC1(Ω) is dense in L2(Ω, ,P) we see that (4.24) holds. n − n → b A D1,2 2) To extend) the Bismut integration by parts formula to functions F let (Fn) be ∈ 1 an approximating sequence of Cb functions w.r.t. the (1, 2) norm. Then for any process 2 t G a and Ht = 0 Gs ds,wehave ∈L ´ 1 1 H E DtFn Gt dt = E (D Fn,H)H = E Fn G dW . ˆ0 · ˆ0 · / 0 / 0 / 0

~~Clearly, both sides are continuous in Fn w.r.t. the (1, 2) norm, and hence the identity extends to F as n . →∞~~

~~The next lemma is often useful to verify Malliavin differentiability:~~

~~2 1,2 Lemma 4.9. Let F L (Ω, ,P),andlet(Fn)n N be a sequence of functions in D ∈ A ∈ converging to F w.r.t. the L2 norm. If~~

~~H 2 sup E[ D Fn H ] < (4.25) n N || || ∞ ∈ D1,2 then F is in ,andthereexistsasubsequence(Fni )i N of (Fn) such that ∈~~

~~1 k F F w.r.t. the (1,2) norm. (4.26) k ni → i=1 " The functional analytic proof is based on the theorems of Banach-Alaoglu and Banach- Saks, cf. e.g. the appendix in [30].~~

H 2 Proof. By (4.25), the sequence (D Fn)n N of gradients is bounded in L (Ω H; P ), ∈ → which is a Hilbert space. Therefore, by the Banach-Alaoglu theorem , there exists a H weakly convergent subsequence (D Fki )i N.Moreover,bytheBanach-SaksTheorem, ∈ H there exists a subsequence (D Fni )i N of the ﬁrst subsequence such that the averages ∈ 1 k DH F are even strongly convergent in L2(Ω H; P ).Hencethecorrespond- k i=1 ni → ing averages 1 k F converge in D1,2.ThelimitisF since F F in L2 and the ! k i=1 ni ni → D1,2 2 norm is stronger! than the L norm.

~~Stochastic Analysis Andreas Eberle 4.3. DIGRESSION ON REPRESENTATION THEOREMS 165~~

~~Product and chain rule~~

~~Lemma 4.9 can be used to extend the product and the chain rule tofunctionsinD1,2.~~

~~Theorem 4.10. 1) If F and G are bounded functions in D1,2 then the product FGis again in D1,2,and~~

~~D(FG)=FDG+ GDF a.s.~~

~~2) Let m N and F (1),...,F(m) D1,2.Ifϕ : Rm R is continuously differen- ∈ ∈ → tiable with bounded derivatives then ϕ(F (1),...,F(m)) is in D1,2,and~~

m ∂ϕ Dϕ(F (1),...,F(m))= (F (1),...,F(m)) DF (i). ∂x i=1 i " Proof. We only prove the product rule, whereas the proof of the chain rule is left as 1 an exercise. Suppose that (Fn) and (Gn) are sequences of Cb functions converging to F and G respectively in D1,2.IfF and G are bounded then one can show that the approximating sequences (Fn) and (Gn) can be chosen uniformly bounded. In particular, F G FGin L2.BytheproductrulefortheFréchetdifferential, n n → DH (F G )=F DH G + G DH F for any n N, and (4.27) n n n n n n ∈ DH (F G ) F DHG + G DH F . || n n ||H ≤|n||| n||H | n||| n||H

~~H 2 Thus the sequence (D (FnGn))n N is bounded in L (Ω H; P ).ByLemma4.9,we ∈ → conclude that FGis in D1,2 and~~

~~k H 2 1 H D (FG)=L - lim D (Fni Gni ) k k →∞ i=1 " for an appropriate subsequence. The product rule for FGnow follows by (4.27).~~

~~4.3 Digression on Representation Theorems~~

~~We now prove basic representation theorems for functions andmartingalesonWiener space. The Bismut integration by parts identity can then be applied to obtain a more~~

University of Bonn Summer Semester 2015 166 CHAPTER 4. VARIATIONS OF PARAMETERS IN SDE explicit form of the classical Itô Representation Theorem. Throughout this section, W (ω)=ω denotes the canonical Brownian motion on Wiener space (Ω, ,P),and t t A = σ(W : s [0,t])P ,t0, Ft s ∈ ≥ is the completed ﬁltration generated by (Wt).

~~Itôs Representation Theorem~~

Itô’s Representation Theorem states that functions on Wiener space that are measurable w.r.t. the Brownian filtration = W,P can be represented as stochastic integrals: Ft Ft Theorem 4.11 (Itô). For any function F 2(Ω, ,P) there exists a unique process ∈L F1 G L2(0, 1) such that ∈ a 1 F = E[F ]+ Gs dWs P -almost surely. (4.28) ˆ0 · An immediate consequence of Theorem 4.11 is a corresponding representation for martingales w.r.t. the Brownian filtration = W,P : Ft Ft Corollary 4.12 (Itô representation for martingales). For any L2-bounded ( ) mar- Ft 2 tingale (Mt)t [0,1] there exists a unique process G La(0, 1) such that ∈ ∈ t Mt = M0 + Gs dWs P -a.s. for any t [0, 1]. ˆ0 · ∈ The corollary is of fundamental importance in financial mathematics where it is related to completeness of financial markets. It also proves the remarkable fact that every martingale w.r.t. the Brownian filtration has a continuous modification!Ofcourse,this result can not be true w.r.t. a general filtration.

~~We ﬁrst show that the corollary follows from Theorem 4.11, andthenweprovethe theorem:~~

2 Proof of Corollary 4.12. If (Mt)t [0,1] is an L bounded ( t) martingale then M1 ∈ F ∈ 2(Ω, ,P),and L F1 M = E[M ] a.s. for any t [0, 1]. t 1|Ft ∈ Stochastic Analysis Andreas Eberle 4.3. DIGRESSION ON REPRESENTATION THEOREMS 167

~~Hence, by Theorem 4.11, there exists a unique process G L2(0, 1) such that ∈ a 1 1 M1 = E[M1]+ G dW = M0 + G dW a.s., ˆ0 · ˆ0 · and thus~~

~~t Mt = E[M1 t]=M0 + G dW a.s. for any t 0. |F ˆ0 · ≥~~

~~Proof of Theorem 4.11. Uniqueness. Suppose that (4.28) holds for two processes G, G ∈ 2 La(0, 1).Then 1 1 ) G dW = G dW, ˆ0 · ˆ0 · and hence, by Itô’s isometry, )~~

~~G G L2(P λ) = (G G) dW =0. || − || ⊗ ˆ − · L2(P ) && && ) && ) && Hence Gt(ω)=Gt(ω) for almost every&&(t, ω). &&~~

Existence. We prove the existence of a representation as in (4.28) in several steps ) − starting with “simple” functions F . 1. Suppose that F =exp(ip (W W )) for some p Rd and 0 s t 1.By · t − s ∈ ≤ ≤ ≤ Itô’s formula,

t 1 2 1 2 1 2 exp(ip Wt + p t)=exp(ip Ws + p s)+ exp ip Wr + p r ip dWr. · 2| | · 2| | ˆs · 2| | · $ % Rearranging terms, we obtain an Itô representation for F with a bounded adapted integrand G. n

2. Now suppose that F = Fk where Fk =exp ipk (Wtk Wtk−1 ) for some k=1 · − d n N, p1,...,pn R ,and0G t0 t1 tn $1.DenotingbyGk the% bounded ∈ ∈ ≤ ≤ ≤···≤ ≤ adapted process in the Itô representation for Fk,wehave

~~n tk+1 F = E[F ]+ Gk dW . k ˆ tk · k(=1 6 7 University of Bonn Summer Semester 2015 168 CHAPTER 4. VARIATIONS OF PARAMETERS IN SDE~~

n Weshowthattherighthandsidecan bewrittenasthesumof k=1 E[Fk] and a stochastic integral w.r.t. W .Forthispurpose,itsufﬁcestoverifythattheproductoftwG ostochas- tic integrals X = t G dW and Y = t H dW with bounded adapted processes G t 0 · t 0 · and H is the stochastic´ integral of a process´ in L2(0, 1) provided 1 G H dt =0.This a 0 t · t holds true, since by the product rule, ´

~~1 1 1 X1Y1 = XtHt dWt + YtGt dWt + Gt Ht dt, ˆ0 · ˆ0 · ˆ0 · and XH + YGis square-integrable by Itô’s isometry.~~

~~3. Clearly, an Itô representation also holds for any linear combination of functions as in Step 2.~~

4. To prove an Itô representation for arbitrary functions in 2(Ω, ,P),weﬁrstnote L F1 that the linear combinations of the functions in Step 2 form a dense subspace of the Hilbert space L2(Ω, ,P).Indeed,ifϕ is an element in L2(Ω, ,P) that is orthogonal F1 F1 to this subspace then

n E ϕ exp(ip (W W )) =0 k · tk − tk−1 / (k=1 0 for any n N, p ,...,p Rd and 0 t t t 1.ByFourierinversion, ∈ 1 n ∈ ≤ 0 ≤ 1 ≤···≤ n ≤ this implies E[ϕ σ(W W :1 k n)] = 0 a.s. | tk − tk−1 ≤ ≤ for any n N and 0 t t 1,andhenceϕ =0a.s. by the Martingale ∈ ≤ 0 ≤···≤ n ≤ Convergence Theorem. Now ﬁx an arbitrary function F L2(Ω, ,P).ThenbyStep3,thereexistsasequence ∈ F1 (F ) of functions in L2(Ω, ,P) converging to F in L2 that have a representation of n F1 the form 1 (n) Fn E[Fn]= G dW (4.29) − ˆ0 · with processes G(n) L2(0, 1).Asn , ∈ a →∞

~~F E[F ] F E[F ] in L2(P ). n − n −→ − Stochastic Analysis Andreas Eberle 4.4. FIRST APPLICATIONS TO SDE 169~~

~~Hence, by (4.29) and Itô’s isometry, (G(n)) is a Cauchy sequence in L2(P λ ). ⊗ (0,1) Denoting by G the limit process, we obtain the representation~~

~~1 F E[F ]= G dW − ˆ0 · by taking the L2 limit on both sides of (4.29).~~

~~Clark-Ocone formula~~

~~If F is in D1,2 then the process G in the Itô representation can be identiﬁed explicitly:~~

Theorem 4.13 (Clark-Ocone). For any F D1,2, ∈ 1 F E[F ]= G dW − ˆ0 · where G = E[D F ]. t t |Ft Proof. It remains to identify the process G in the Itô representation. We assume w.l.o.g. that E[F ]=0.LetH L1([0, 1], Rd).ThenbyItô’sisometryandtheintegrationby ∈ a parts identity,

1 1 1 1 E Gt Ht dt = E G dW HdW = E DtF Ht dt ˆ0 · ˆ0 · ˆ0 ˆ0 · / 0 / 1 0 / 0 = E E[DtF t] Ht dt ˆ0 |F · / 0 for all Setting H := G E[D F ] we obtain t t − t |Ft G (ω)=E[D F ](ω) P λ a.e. t t |Ft ⊗ −

~~4.4 First applications to stochastic differential equations~~

~~4.5 Existence and smoothness of densities~~

~~University of Bonn Summer Semester 2015 Chapter 5~~

~~Stochastic calculus for semimartingales with jumps~~

Our aim in this chapter is to develop a stochastic calculus forfunctionsofﬁnitelymany (1) (2) (d) real-valued stochastic processes Xt ,Xt ,...,Xt .Inparticular,wewillmakesense of stochastic differential equations of type

~~d (k) dYt = σk(t, Yt ) dXt − "k=1~~

n n with continuous time-dependent vector fields σ1,...,σd : R+ R R .Thesample (k) × → paths of the driving processes (Xt ) and of the solution (Yt) may be discontinuous, but we will always assume that they are càdlàg,i.e.,right-continuouswithleftlimits.In most relevant cases this can be assured by choosing an appropriate modification. For example, a martingale or a Lévy process w.r.t.a right-continuous complete filtration always has a càdlàg modification, cf. [37, Ch.II, §2] and [36, Ch.I Thm.30].

An adequate class of stochastic processes for which a stochastic calculus can be developed are semimartingales,i.e.,sumsoflocalmartingalesandadaptedfinitevariation processes with càdlàg trajectories. To understand why this is a reasonable class of processes to consider, we first briefly review the discrete time case.

~~170 171~~

~~Semimartingales in discrete time~~

~~If ( ) is a discrete-time ﬁltration on a probability space (Ω, ,P) then any Fn n=0,1,2,... A ( ) adapted integrable stochastic process (X ) has a unique Doob decomposition Fn n~~

X = X + M + A↗ A↘ (5.1) n 0 n n − n into an ( ) martingale (M ) and non-decreasing predictable processes (A↗) and (A↘) Fn n n n such that M0 = A0↗ = A0↘ =0,cf.[14,Thm.2.4].Thedecompositionisdetermined by choosing

~~Mn Mn 1 = Xn Xn 1 E[Xn Xn 1 n 1], − − − − − − − |F −~~

~~+ An↗ An↗ 1 = E[Xn Xn 1 n 1] , and An↘ An↘ 1 = E[Xn Xn 1 n 1]−. − − − − |F − − − − − |F −~~

~~In particular, (Xn) is a sub- or supermartingale if and only if An↘ =0for any n,or~~

~~An↗ =0for any n,respectively.Thediscretestochasticintegral~~

(G X)n = Gk (Xk Xk 1) • − − "k=1 of a bounded predictable process (Gn) w.r.t. (Xn) is again a martingale if (Xn) is a martingale, and an increasing (decreasing) process if G 0 for any n,and(X ) n ≥ n is increasing (respectively decreasing). For a bounded adapted process (Hn),wecan deﬁne correspondingly the integral

(H X)n = Hk 1 (Xk Xk 1) −• − − − "k=1 of the predictable process H =(Hk 1)k N w.r.t. X. − − ∈ The Taylor expansion of a function F C2(R) yields a primitive version of the Itô ∈ formula in discrete time. Indeed, notice that for k N, ∈ 1 F (Xk) F (Xk 1)= F ′(Xk 1 + s∆Xk) ds ∆Xk − − ˆ0 − 1 s 2 = F ′(Xk 1)∆Xk + F ′′(Xk 1 + r∆Xk) dr ds (∆Xk) . − ˆ0 ˆ0 −

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 172 JUMPS~~

where ∆Xk := Xk Xk 1.Bysummingoverk,weobtain − − n 1 s 2 F (Xn)=F (X0)+(F ′(X) X)n + F ′′(Xk 1 + r∆Xk) dr ds (∆Xk) . −• ˆ0 ˆ0 − "k=1 Itô’s formula for a semimartingale (Xt) in continuous time will be derived in Theorem 5.22 below. It can be rephrased in a way similar to the formula above, where the last term on the right-hand side is replaced by an integral w.r.t. the quadratic variation process

~~[X]t of X,cf.(XXX).~~

~~Semimartingales in continuous time~~

In continuous time, it is no longer true that any adapted process can be decomposed into a local martingale and an adapted process of ﬁnite variation (i.e., the sum of an increasing and a decreasing process). A counterexample is given by fractional Brownian motion, cf. Section 2.3 below. On the other hand, a large classofrelevantprocesseshas acorrespondingdecomposition.

Deﬁnition. Let ( t)t 0 be a ﬁltration. A real-valued ( t)-adapted stochastic process F ≥ F (Xt)t 0 on a probability space (Ω, ,P) is called an (Ft) semimartingale if and only ≥ A if it has a decomposition

~~X = X + M + A ,t0, (5.2) t 0 t t ≥ into a strict local ( )-martingale (M ) with càdlàg paths, and an ( )-adapted process Ft t Ft (At) with càdlàg ﬁnite-variation paths such that M0 = A0 =0.~~

Here a strict local martingale is a process that can be localized by martingales with uniformly bounded jumps, see Section 2.2 for the precise definition. Any continuous local martingale is strict. In general, it can be shown that if the filtration is right continuous and complete then any local martingale can be decomposed intoastrictlocalmartingale and an adapted finite variation process (“Fundamental Theorem of Local Martingales”, cf. [36]). Therefore, the notion of a semimartingale defined above is not changed if the word “strict” is dropped in the definition. Since the non-trivial proof of the Fundamental Theorem of Local Martingales is not included in these notes, we nevertheless stick to the definition above.

~~Stochastic Analysis Andreas Eberle 173~~

Remark. (Assumptions on path regularity). Requiring (At) to be càdlàg is just a standard convention ensuring in particular that t A (ω) is the distribution function of #→ t asignedmeasure.Theexistenceofrightandleftlimitsholdsforanymonotonefunction, and, therefore, for any function of finite variation. Similarly, every local martingale w.r.t.a right-continuous complete filtration has a càdlàg modification.

Without additional conditions on (At),thesemimartingaledecompositionin(5.2)isnot unique,seetheexamplebelow.Uniquenessholdsif,inaddition,(At) is assumed to be predictable, cf. [7, 36]. Under the extra assumption that (At) is continuous, uniqueness is a consequence of Corollary 5.15 below.

~~Example (Semimartingale decompositions of a Poisson process). An ( ) Poisson Ft process (Nt) with intensity λ has the semimartingale decompositions~~

~~Nt = Nt + λt =0+Nt into a martingale and an adapted finite) variation process. Only in the first decomposition, the finite variation process is predictable and continuous respectively.~~

~~The following examples show that semimartingales form a sufﬁciently rich class of stochastic processes.~~

Example (Stochastic integrals). Let (B ) and (N ) be a d-dimensional ( ) Brownian t t Ft motion and an ( ) Poisson point process on a σ-ﬁnite measure space (S, ,ν) respec- Ft S tively. Then any process of the form

t t Xt = Hs dBs+ Gs(y)N(ds dy)+ Ks ds+ Ls(y)N(ds dy) (5.3) ˆ0 · ˆ(0,t] S ˆ0 ˆ(0,t] S × × is a semimartingale provided the integrands) H, G, K, L are predictable, H and G are (locally) square integrable w.r.t. P λ, P λ ν respectively, and K and L are ⊗ ⊗ ⊗ (locally) integrable w.r.t. these measures. In particular,bytheLévy-Itôdecomposition, every Lévy process is a semimartingale.Similarly,thecomponentsofsolutions of SDE driven by Brownian motions and Poisson point processes are semimartingales.More generally, Itô’s formula yields an explicit semimartingaledecompositionoff(t, Xt) for an arbitrary function f 2 (R Rn) and (X ) as above, cf. Section 5.4 below. ∈C + × t

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 174 JUMPS~~

Example (Functions of Markov processes). If (X ) is a time-homogeneous ( ) t Ft Markov process on a probability space (Ω, ,P),andf is a function in the domain A of the generator ,thenf(X ) is a semimartingale with decomposition L t t f(X )=local martingale + ( f)(X ) ds, (5.4) t ˆ L s 0 cf. e.g. [12] or [16]. Indeed, it is possible to deﬁne the generator of a Markov process L through a solution to a martingale problem as in (5.4).

~~Many results for continuous martingales carry over to the càdlàg case. However, there are some important differences and pitfalls to be noted: Exercise (Càdlàg processes).~~

1) A stopping time is called predictable iff there exists an increasing sequence (Tn) of stopping times such that T 0 and T =supT .Showthatfor n { } N acàdlàgstochasticprocess(Xt)t 0,theﬁrsthittingtime ≥ T = inf t 0:X A A { ≥ t ∈ } of a closed set A R is not predictable in general. ⊂ 2) Prove that for a right continuous ( t) martingale (Mt)t 0 and an ( t) stopping F ≥ F time T ,thestoppedprocess(Mt T )t 0 is again an ( t) martingale. ∧ ≥ F 3) Prove that a càdlàg local martingale (Mt) can be localized by a sequence (Mt Tn ) ∧ of bounded martingales provided the jumps of (Mt) are uniformly bounded, i.e.,

~~sup ∆M (ω) : t 0,ω Ω < . {| t | ≥ ∈ } ∞ 4) Give an example of a càdlàg local martingale that can not be localized by bounded martingales.~~

Our next goal is to deﬁne the stochastic integral G X w.r.t. a semimartingale X for • the left limit process G =(Ht ) of an adapted càdlàg process H,andtobuildupa − corresponding stochastic calculus. Before studying integration w.r.t. càdlàg martingales in Section 5.2, we will consider integrals and calculus w.r.t. ﬁnite variation processes in Section 5.1.

~~Stochastic Analysis Andreas Eberle 5.1. FINITE VARIATION CALCULUS 175~~

~~5.1 Finite variation calculus~~

In this section we extend Stieltjes calculus to càdlàg paths of ﬁnite variation. The results are completely deterministic. They will be applied later to the sample paths of the ﬁnite variation part of a semimartingale.

Fix u (0, ],andletA : [0,u) R be a right-continuous function of ﬁnite variation. ∈ ∞ → In particular, A is càdlàg. We recall that there is a σ-ﬁnite measure µA on (0,u) with distribution function A,i.e.,

~~µ ((s, t]) = A A for any 0 s t~~

~~The function A has the decomposition~~

~~c d At = At + At (5.6) into the pure jump function d At := ∆As (5.7) s t "≤ and the continuous function Ac = A Ad.Indeed,theseriesin(5.7)convergesabso- t t − t lutely since~~

~~∆A V (1)(A) < for any t [0,u). | s|≤ t ∞ ∈ s t "≤~~

~~The measure µA can be decomposed correspondingly into~~

~~µA = µAc + µAd where~~

~~µ d = ∆A δ A s · s s (0,u) ∈ ∆"As=0 ̸ is the atomic part, and µAc does not contain atoms. Note that µAc is not necessarily absolutely continuous!~~

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 176 JUMPS~~

~~Lebesgue-Stieltjes integrals revisited~~

Let 1 ([0,u),µ ) := 1 ([0,u), µ ) where µ denotes the positive measure with Lloc A Lloc | A| | A| distribution function V (1)(A).ForG 1 ([0,u),µ ),theLebesgue-Stieltjesintegral t ∈Lloc A of H w.r.t. A is deﬁned as u Gr dAr = Gr I(s,t](r) µA(dr) for 0 s t

~~µI(dr)=Gr µA(dr) with density G w.r.t. µA.Thishasseveralimportantconsequences: 1) The function I is again càdlàg and of ﬁnite variation with~~

~~t t (1) (1) Vt (I)= Gr µA (dr)= Gr dVr (A). ˆ0 | || | ˆ0 | | 2) I decomposes into the continuous and pure jump parts~~

~~t t Ic = G dAc ,Id = G dAd = G ∆A . t ˆ r r t ˆ r r s s 0 0 s t "≤ 3) For any G 1 (µ ), ∈Lloc I t t ) Gr dIr = GrGr dAr, ˆ0 ˆ0 i.e., if “dI = GdA”thenalso“) GdI= GG dA) ”.~~

Theorem 5.1 (Riemann sum approximations) for) Lebesgue-Stieltjes integrals). Sup- pose that H : [0,u) R is a càdlàg function. Then for any a [0,u) and for any → ∈ sequence (π ) of partitions with mesh(π ) 0, n n → t lim Hs(As′ t As)= Hs dAs uniformly for t [0,a]. n ∧ − ˆ − ∈ →∞ s πn 0 "s

~~Stochastic Analysis Andreas Eberle 5.1. FINITE VARIATION CALCULUS 177~~

~~Remark. If (At) is continuous then~~

t t Hs dAs = Hs dAs, ˆ0 − ˆ0 t because 0 ∆HsdAs = s t ∆Hs∆As =0for any càdlàg function H.Ingeneral, ´ ≤ however, the limit of the! Riemann sums in Theorem 5.1 takes themodiﬁedform t t c Hs dAs = Hs dAs + Hs ∆As. ˆ − ˆ − 0 0 s t "≤ Proof. For n N and t 0, ∈ ≥

~~′ Hs(As t As)= Hs dAr = H r n dAr ∧ − ˆ ′ ˆ ⌊ ⌋ s πn s πn (s,s t] (0,t] "s~~

~~t t~~

~~sup H r n dAr Hr dAr H r n Hr µA (dr) 0 t a ˆ0 ⌊ ⌋ − ˆ0 − ≤ ˆ(0,a] ⌊ ⌋ − − | | → ≤ & & & & & & as n & by dominated convergence.& & & →∞& &~~

~~Product rule~~

~~The covariation [H, A] of two functions H, A : [0,u) R w.r.t. a sequence (π ) of → n partitions with mesh(π ) 0 is deﬁned by n →~~

~~[H, A]t = lim (Hs′ t Hs)(As′ t As), (5.8) n ∧ − ∧ − →∞ s πn "s~~

~~Lemma 5.2. If H and A are càdlàg and A has ﬁnite variation then the covariation exists and is independently of (πn) given by~~

~~[H, A]t = ∆Hs∆As 0~~

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 178 JUMPS~~

~~Proof. We again represent the sums as integrals: t ′ ′ (Hs t Hs)(As t As)= (H r n t H r n ) dAr ∧ − ∧ − ˆ ⌈ ⌉ ∧ − ⌊ ⌋ s πn 0 "s~~

Remark. 1) If H or A is continuous then [H, A]=0. 2) In general, the proof above shows that t t Hs dAs = Hs dAs +[H, A]t, ˆ0 ˆ0 − i.e., [H, A] is the difference between limits of right and left Riemann sums.

Theorem 5.3 (Integration by parts, product rule). Suppose that H, A : [0,u) R → are right continuous functions of ﬁnite variation. Then t t HtAt H0A0 = Hr dAr + Ar dHr +[H, A]t for any t [0,u). (5.9) − ˆ0 − ˆ0 − ∈ In particular, the covariation [H, A] is a càdlàg function of ﬁnite variation, and for a

~~d(HA)r = Hr dAr + Ar dHr + d[H, A]r. − − As special cases we note that if H and A are continuous then HA is continuous with~~

~~d(HA)r = Hr dAr + Ar dHr, and if H and A are pure jump functions (i.e. Hc = Ac =0)thenHA is a pure jump function with~~

~~∆(HA)r = Hr ∆Ar + Ar ∆Hr +∆Ar∆Hr . − − Stochastic Analysis Andreas Eberle 5.1. FINITE VARIATION CALCULUS 179~~

~~In the latter case, (5.9) implies~~

~~H A H A = ∆(HA) . t t − 0 0 r r t "≤ Note that this statement is not completely trivial, as it holds even when the jump times of HA form a countable dense subset of [0,t]!~~

~~Since the product rule is crucial but easy to prove, we give twoproofsofTheorem5.3:~~

~~Proof 1. For (π ) with mesh(π ) 0,wehave n n →~~

~~HtAt H0A0 = (Hs′ tAs′ t HsAs) − ∧ ∧ − s πn "s~~

= Hs(As′ t As)+ As(Hs′ t Hs)+ (As′ t As)(Hs′ t Hs). ∧ − ∧ − ∧ − ∧ − " " " As n ,(5.9)followsbyTheorem5.1above.Moreover,theconvergence of the →∞ covariation is uniform for t [0,a], a

~~s>r~~

(Ht H0)(At A0)= µH (dr) µA(ds) − − ˆ(0,t] (0,t] × is the area of (0,t] (0,t] w.r.t.the product measure µ µ .Bydividingthesquare × H ⊗ A (0,t] (0,t] into the parts (s, r) sr and the diagonal (s, r) s = r × { | } { | } { | } we see that this area is given by

~~t t + + = (Ar A0) dHr + (Hs H0) dAs + ∆Hs∆As, ˆ ˆ ˆ ˆ − − ˆ − − sr s=r 0 0 s t "≤ The assertion follows by rearranging terms in the resulting equation.~~

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 180 JUMPS~~

~~Chain rule~~

~~The chain rule can be deduced from the product rule by iteration and approximation of 1 functions by polynomials: C~~

Theorem 5.4 (Change of variables, chain rule, Itô formula for ﬁnite variation functions). Suppose that A : [0,u) R is right continuous with ﬁnite variation, and let → F 1(R).Thenforanyt [0,u), ∈C ∈

~~t F (At) F (A0)= F ′(As ) dAs + (F (As) F (As ) F ′(As )∆As) , − ˆ − − − − − 0 s t "≤ (5.10) or, equivalently,~~

~~t c F (At) F (A0)= F ′(As ) dAs + (F (As) F (As )) . (5.11) − ˆ − − − 0 s t "≤~~

~~If A is continuous then F (A) is also continuous, and (5.10) reduces to the standard chain rule t F (At) F (A0)= F ′(As) dAs. − ˆ0~~

~~If A is a pure jump function then the theorem shows that F (A) is also a pure jump function (this is again not completely obvious!) with~~

F (At) F (A0)= (F (As) F (As )) . − − − s t "≤ Remark. Note that by Taylor’s theorem, the sum in (5.10) converges absolutely when- 2 ever s t(∆As) < .ThisobservationwillbecrucialfortheextensiontoItô’s ≤ ∞ formula! for processes with ﬁnite quadratic variation, cf. Theorem 5.22 below.

Proof of Theorem 2.4. Let denote the linear space consisting of all functions F A ∈ 1(R) satisfying (5.10). Clearly the constant function 1 and the identity F (t)=t are in C Stochastic Analysis Andreas Eberle 5.1. FINITE VARIATION CALCULUS 181

~~.Wenowprovethat is an algebra: Let F, G .Thenbytheintegrationbyparts A A ∈A identity and by (5.11),~~

~~(FG)(A ) (FG)(A ) t − 0 t t = F (As ) dG(A)s + G(As ) dF (A)s + ∆F (A)s∆G(A)s ˆ − ˆ − 0 0 s t ≤ t " c = (F (As )G′(As )+G(As )F ′(As )) dAs ˆ0 − − − −~~

+ (F (As )∆G(A)s + G(As )∆F (A)s +∆F (A)s∆G(A)s) − − s t ≤ t" c = (FG)′(As ) dAs + ((FG)(As) (FG)(As )) ˆ − − − 0 s t "≤ for any t [0,u),i.e.,FGis in . ∈ A Since is an algebra containing 1 and t,itcontainsallpolynomials.Moreover,ifF A is an arbitrary 1 function then there exists a sequence (p ) of polynomials such that C n p F and p′ F ′ uniformly on the bounded set A s t .Since(5.11)holds n → n → { s | ≤ } for the polynomials pn,italsoholdsforF .

~~Exponentials of ﬁnite variation functions~~

~~Let A : [0, ) R be a right continuous finite variation function. The exponen- ∞ → tial of A is defined as the right-continuous finite variation function (Zt)t 0 solving the ≥ equation~~

~~dZt = Zt dAt ,Z0 =1 , i.e., − t Zt =1+ Zs dAs for any t 0. (5.12) ˆ0 − ≥~~

~~If A is continuous then Zt =exp(At) solves (5.12) by the chain rule. On the other hand, if A is piecewise constant with ﬁnitely many jumps then Zt = s t(1 + ∆As) solves ≤ (5.12), since G~~

~~Zt = Z0 + ∆Zs =1+ Zs ∆As =1+ Zs dAs. − ˆ − s t s t (0,t] "≤ "≤ In general, we obtain:~~

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 182 JUMPS~~

~~Theorem 5.5. The unique càdlàg function solving (5.12) is~~

~~Z =exp(Ac) (1 + ∆A ), (5.13) t t · s s t (≤ where the product converges for any t 0. ≥ Proof. 1) We ﬁrst show convergence of the product~~

~~Pt = (1 + ∆As). s t (≤ Recall that since A is càdlàg, there are only ﬁnitely many jumps with ∆A > 1/2. | s| Therefore, we can decompose~~

Pt =exp⎛ log(1 + ∆As)⎞ (1 + ∆As) (5.14) · s t s t ≤ ≤ ⎜ ∆A"s 1/2 ⎟ ∆A(s >1/2 ⎜| |≤ ⎟ | | ⎝ ⎠ in the sense that the product Pt converges if and only if the series converges. The series converges indeed absolutely for A with finite variation, since log(1+x) can be bounded by a constant times x for x 1/2.ThelimitS of the series defines a pure jump | | | |≤ t function with variation V (1)(S) const. V (1)(A) for any t 0. t ≤ · t ≥ 2) Equation for P :Thechainandproductrulenowimplyby(5.14)thatt P is also t #→ t afinitevariationpurejumpfunction.Therefore,

t d Pt = P0 + ∆Ps =1+ Ps ∆As =1+ Ps dAs, t 0, − ˆ − ∀ ≥ s t s t 0 "≤ "≤ (5.15) d i.e., P is the exponential of the pure jump part At = s t ∆As. ≤ c c 3) Equation for Zt:SinceZt =exp(At )Pt and exp(A!) is continuous, the product rule and (5.15) imply

~~t t c c As As c Zt 1= e dPs + Ps e dAs − ˆ0 ˆ0 − t t c As d c = e Ps d(A + A )s = Zs dAs. ˆ0 − ˆ0 −~~

~~Stochastic Analysis Andreas Eberle 5.1. FINITE VARIATION CALCULUS 183~~

~~4) Uniqueness:SupposethatZ is another càdlàg solution of (5.12), and let Xt :=~~

~~Zt Zt.ThenX solves the equation − ) t ) Xt = Xs dAs t 0 ˆ0 − ∀ ≥ with zero initial condition. Therefore,~~

~~t Xt Xs dVt MtVt t 0, | |≤ˆ0 | −| ≤ ∀ ≥~~

~~(1) where Vt := Vt (A) is the variation of A and Mt := sups t Xs .Iteratingtheestimate ≤ | | yields t 2 Xt Mt Vs dVs MtVt /2 | |≤ ˆ0 − ≤ by the chain rule, and~~

t Mt n Mt n+1 N Xt Vs dVs Vt t 0,n . (5.16) | |≤ n! ˆ0 − ≤ (n +1)! ∀ ≥ ∈ Note that the correction terms in the chain rule are non-negative since V 0 and t ≥ [V ] 0 for all t.Asn ,therighthandsidein(5.16)convergesto0 since M and t ≥ →∞ t V are ﬁnite. Hence X =0for each t 0. t t ≥ From now on we will denote the unique exponential of (A ) by ( A). t Et Remark (Taylor expansion). By iterating the equation (5.12) for the exponential, we obtain the convergent Taylor series expansion

~~n A =1+ dA dA dA + R(n), t ˆ ˆ ˆ sk sk−1 s1 t E (0,t] (0,s1) ··· (0,sn−1) ··· "k=1 where the remainder term can be estimated by~~

~~R(n) M V n+1/(n +1)!. | t |≤ t t If A is continuous then the iterated integrals can be evaluated explicitly:~~

~~dA dA dA =(A A )k/k!. ˆ ˆ ˆ sk sk−1 s1 t 0 (0,t] (0,s1) ··· (0,sk−1) ··· −~~

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 184 JUMPS~~

~~If A is increasing but not necessarily continuous then the right hand side still is an upper bound for the iterated integral.~~

~~We now derivea formula for A B where A and B are right-continuous ﬁnite variation Et ·Et functions. By the product rule and the exponential equation,~~

t t A B A B B A A B t t 1= s d s + s d s + ∆ s ∆ s E E − ˆ E − E ˆ E − E E E 0 0 s t ≤ t " A B A B = s s d(A + B)s + s s ∆As∆Bs ˆ E −E − E −E − 0 s t ≤ t " A B = s s d(A + B +[A, B])s ˆ0 E −E −

for any t 0.Thisshowsthatingeneral, A B = A+B. ≥ E E ̸ E Theorem 5.6. If A, B : [0, ) R are right continuous with ﬁnite variation then ∞ → A B = A+B+[A,B]. E E E Proof. The left hand side solves the deﬁning equation for the exponential on the right hand side.

In particular, choosing B = A,weobtain: − 1 A+[A] = − A E E Example (Geometric Poisson process). A geometric Poisson process with parameters λ>0 and σ,α R is deﬁned as a solution of a stochastic differential equationoftype ∈

dSt = σSt dNt + αSt dt (5.17) − w.r.t.a Poisson process (Nt) with intensity λ.GeometricPoissonprocessesarerelevant for ﬁnancial models, cf. e.g. [39]. The equation (5.17) can beinterpretedpathwiseas the Stieltjes integral equation

~~t t St = S0 + σ Sr dNr + α Srdr , t 0. ˆ0 − ˆ0 ≥~~

~~Stochastic Analysis Andreas Eberle 5.1. FINITE VARIATION CALCULUS 185~~

~~Deﬁning At = σNt + αt,(5.17)canberewrittenastheexponentialequation~~

~~dSt = St dAt , − which has the unique solution~~

S = S A = S eαt (1 + σ∆N )=S eαt(1 + σ)Nt . t 0 ·Et 0 · s 0 · s t (≤ Note that for σ> 1,asolution(S ) with positive initial value S is positive for all t, − t 0 whereas in general the solution may also take negative values. If α = λσ then (A ) − t is a martingale. We will show below that this implies that (St) is a local martingale.

~~Indeed, it is a true martingale which for S0 =1takes the form~~

~~Nt λσt St =(1+σ) e− .~~

~~Corresponding exponential martingales occur as “likelihood ratio” when the intensity of a Poisson process is modiﬁed, cf. Chapter 2 below.~~

Example (Exponential martingales for compound Poisson processes). For compound Poisson processes, we could proceed as in the last example. To obtain a different point of view, we go in the converse direction: Let

Xt = ηj j=1 " be a compound Poisson process on Rd with jump intensity measure ν = λµ where λ ∈ (0, ) and µ is a probability measure on Rd 0 .Hencetheη are i.i.d. µ,and(K ) is ∞ \{ } j ∼ t an independent Poisson process with intensity λ.Supposethatwewouldliketochange the jump intensity measure to an absolutely continuous measure ν¯(dy)=ϱ(y)ν(dy) with relative density ϱ 1(ν),andletλ¯ =¯ν(Rd 0 ).Intuitively,wecouldexpect ∈L \{ } that the change of the jump intensity is achieved by changing the underlying probability measure P on X with relative density (“likelihood ratio”) Ft

~~Kt (λ λ¯)t (λ λ¯)t Zt = e − ϱ(ηj)=e − ϱ(∆Xs). j=1 s t ( ∆(X≤s=0 ̸ University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 186 JUMPS~~

In Chapter 2, as an application of Girsanov’s Theorem, we willproverigorouslythat this heuristics is indeed correct. For the moment, we identify (Zt) as an exponential martingale. Indeed, Z = A with t Et A =(λ λ¯)t + (ϱ(∆X ) 1) t − s − s t ∆"X≤s=0 ̸ = (λ¯ λ)t + (ϱ(y) 1) N (dy). (5.18) − − ˆ − t

Kt Here Nt = j=1 δηj denotes the corresponding Poisson point process with intensity measure ν.Notethat! (At) is a martingale, since it is a compensated compound Poisson process A = (ϱ(y) 1) N (dy) , where N := N tν. t ˆ − t t t −

~~By the results in the next section, we) can then conclude that) the exponential (Zt) is a local martingale. We can write down the SDE~~

~~t Zt =1+ Zs dAs (5.19) ˆ0 − in the equivalent form~~

t Zt =1+ Zs (ϱ(y) 1) N(ds dy) (5.20) ˆ(0,t] Rd − − × ) where N(ds dy) := N(ds dy) ds ν(dy) is the random measure on R+ Rd with − × d N((0,t] B)=Nt(B) for any t 0 and B (R ).Indifferentialnotation,(5.20)is )× ≥ ∈B an SDE driven by the compensated Poisson point process (Nt): ) )

dZt = Zt (ϱ(y) 1) N(dt) dy). ˆy Rd − − ∈ Example (Stochastic calculus for finite Markov chains) ). Functions of continuous time Markov chains on finite sets are semimartingales with finite variation paths. There- fore, we can apply the tools of finite variation calculus. Our treatment follows Rogers &Williams[38]wheremoredetailsandapplicationscanbefound. Suppose that (X ) on (Ω, ,P) is a continuous-time, time-homogeneous Markov pro- t A cess with values in a finite set S and càdlàg paths. We denote the transition matrices by

~~Stochastic Analysis Andreas Eberle 5.1. FINITE VARIATION CALCULUS 187~~

1 pt and the generator (Q-matrix) by =( (a, b))a,b S.Thus = limt 0 t− (pt I), L L ∈ L ↓ − i.e., for a = b, (a, b) is the jump rate from a to b,and (a, a)= b S,b=a (a, b) is ̸ L L − ∈ ̸ L the total (negative) intensity for jumping away from a.Inparticular,!

( f)(a) := (a, b)f(b)= (a, b)(f(b) f(a)) L L L − b S b S,b=a "∈ ∈"̸ for any real-valued function f =(f(a))a S on S.Itisastandardfactthat((Xt),P) ∈ solves the martingale problem for ,i.e.,theprocess L t [f] Mt = f(Xt) ( f)(Xs) ds , t 0, (5.21) − ˆ0 L ≥ is an ( X ) martingale for any f : S R.Indeed,thisisadirectconsequenceofthe Ft → Markov property and the Kolmogorov forward equation, which imply

t [f] [f] X E[Mt Ms s ]=E[f(Xt) f(Xs) ( f)(Xr) dr s] − |F − − ˆs L |F t =(pt sf)(Xs) f(Xs) (pr s f)(Xs) ds =0 − − − ˆs − L for any 0 s t.Inparticular,choosingf = I b for b S,weseethat ≤ ≤ { } ∈ t b Mt = I b (Xt) (Xs,b) ds (5.22) { } − ˆ0 L is a martingale, and, in differential notation,

~~b dI b (Xt)= (Xt,b) dt + dMt . (5.23) { } L Next, we note that by the results in the next section, the stochastic integrals~~

t a,b b Nt = I a (Xs ) dMs ,t 0, ˆ0 { } − ≥ are martingales for any a, b S.Explicitly,foranya = b, ∈ ̸ a,b Nt = I a (Xs ) IS b (Xs )I b (Xs) I b (Xs )IS b (Xs) { } − \{ } − { } − { } − \{ } s t ≤ "t $ % I a (Xs) (Xs,b) ds , i.e., − ˆ0 { } L

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 188 JUMPS~~

N a,b = J a,b (a, b) La (5.24) t t −L t a,b where Jt = s t : Xs = a, Xs = b is the number of jumps from a to b until |{ ≤ − }| time t,and t a Lt = Ia(Xs) ds ˆ0 is the amount of time spent at a before time t (“local time at a”). In the form of an SDE, dJ a,b = (a, b) dLa + dN a,b for any a = b. (5.25) t L t t ̸ More generally, for any function g : S S R,theprocess × →

~~[g] a,b Nt = g(a, b)Nt a,b S "∈ is a martingale. If g(a, b)=0for a = b then by (5.24),~~

t [g] T Nt = g(Xs ,Xs) ( g )(Xs,Xs) ds (5.26) − − ˆ L s t 0 "≤ Finally, the exponentials of these martingales are again local martingales. For example, we ﬁnd that αN a,b a,b a =(1+α)Jt exp( α (a, b)L ) Et − L t is an exponential martingale for any α R and a, b S.Theseexponentialmartingales ∈ ∈ appear again as likelihood ratios when changing the jump rates of the Markov chains.

Exercise (Change of measure for ﬁnite Markov chains). Let (X ) on (Ω, ,P,( )) t A Ft be a continuous time Markov chain with ﬁnite state space S and generator (Q-matrix) ,i.e., L t [f] Mt := f(Xt) f(X0) ( f)(Xs) ds − − ˆ0 L is a martingale w.r.t. P for each function f : S R.Weassume (a, b) > 0 for a = b. → L ̸ Let g(a, b) := (a, b)/ (a, b) 1 for a = b, g(a, a) := 0, L L − ̸ where is another Q-matrix. L ) Stochastic) Analysis Andreas Eberle 5.2. STOCHASTIC INTEGRATION FOR SEMIMARTINGALES 189

1) Let λ(a)= b=a (a, b)= (a, a) and λ(a)= (a, a) denote the total jump ̸ L −L −L intensities at!a.Wedeﬁnea“likelihoodquotient”forthetrajectoriesofMarkov ) ) chains with generators and by Z = ζ /ζ where L L t t t t ζt =exp) λ(Xs) ds) (Xs ,Xs), − ˆ0 L − - . s t:Xs−=Xs ≤ (̸ ) ) ) [g] and ζt is deﬁned correspondingly. Prove that (Zt) is the exponential of (Nt ),and

conclude that (Zt) is a martingale with E[Zt]=1for any t. 2) Let P denote a probability measure on that is absolutely continuous w.r.t. P on A t with relative density Zt for every t 0.Showthatforanyf : S R, F ) ≥ → t [f] Mt := f(Xt) f(X0) ( f)(Xs) ds − − ˆ0 L

is a martingale w.r.t.*P .Henceunderthenewprobabilitymeasure) P , (Xt) is a Markov chain with generator . L ) [f] ) Hint: You may assume without proof that (Mt ) is a local martingale w.r.t. P if [f] ) and only if (ZtMt ) is a local martingale w.r.t. P .Aproofofthisfactisgivenin * ) Section 3.3. *

~~5.2 Stochastic integration for semimartingales~~

Throughout this section we fix a probability space (Ω, ,P) with filtration ( t)t 0.We A F ≥ now define the stochastic integral of the left limit of an adapted càdlàg process w.r.t. a semimartingale in several steps. The key step is the first, where we prove the existence for the integral Hs dMs of a bounded adapted càdlàg process H w.r.t. a bounded − martingale M. ´

~~Integrals with respect to bounded martingales~~

~~P Suppose that M =(Mt)t 0 is a uniformly bounded càdlàg ( t ) martingale, and H = ≥ F P (Ht)t 0 is a uniformly bounded càdlàg ( t ) adapted process. In particular, the left limit ≥ F process~~

H := (Ht )t 0 − − ≥ University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 190 JUMPS is left continuous with right limits and ( P ) adapted. For a partition π of R we con- Ft + sider the elementary processes

π ′ π ′ Ht := Hs I[s,s )(t), and Ht = Hs I(s,s ](t). − s π s π "∈ "∈ The process Hπ is again càdlàg and adapted, and the left limit Hπ is left continuous and − (hence) predictable . We consider the Riemann sum approximations

~~π ′ It := Hs(Ms t Ms) ∧ s π − "s~~

~~t to the integral Hs dMs to be deﬁned. Note that if we deﬁne the stochastic integral 0 − of an elementary´ process in the obvious way then~~

t π π It = Hs dMs . ˆ0 − We remark that a straightforward pathwise approach for the existence of the limit of Iπ(ω) as mesh(π) 0 is doomed to fail, if the sample paths are not of ﬁnite variation: → Exercise. Let ω Ω and t (0, ),andsupposethat(π ) is a sequence of partitions ∈ ∈ ∞ n of R+ with mesh(πn) 0.Provethatif s π hs(Ms′ t(ω) Ms(ω)) converges for → s

The assertion of the exercise is just a restatement of the standard fact that the dual space of ([0,t]) consists of measures with ﬁnite total variation. There are approaches to ex- C tend the pathwise approach by restricting the class of integrands further or by assuming extra information on the relation of the paths of the integrand and the integrator (Young integrals, rough paths theory, cf. [29], [19]). Here, following the standard development of stochastic calculus, we also restrict the class of integrands further (to predictable processes), but at the same time, we give up the pathwise approach. Instead, we consider stochastic modes of convergence.

~~For H and M as above, the process Iπ is again a bounded càdlàg ( P ) martingale as Ft is easily veriﬁed. Therefore, it seems natural to study convergence of the Riemann sum~~

~~Stochastic Analysis Andreas Eberle 5.2. STOCHASTIC INTEGRATION FOR SEMIMARTINGALES 191~~

2 2 approximations in the space Md ([0,a]) of equivalence classes of càdlàg L -bounded ( P ) martingales deﬁned up to a ﬁnite time a.Thefollowingfundamentaltheorem Ft settles this question completely:

Theorem 5.7 (Convergence of Riemann sum approximations to stochastic integrals). Let a (0, ) and let M and H be as deﬁned above. Then for every γ>0 ∈ ∞ there exists a constant ∆ > 0 such that

~~π π" 2 I I 2 <γ (5.27) || − ||M ([0,a]) holds for any partitions π and π of R+ with mesh(π) < ∆ and mesh(π) < ∆.~~

) ) The constant ∆ in the theorem depends on M,H and a.Theproofofthetheoremfor discontinuous processes is not easy, but it is worth the effort. For continuous processes, the proof simpliﬁes considerably. The theorem can be avoidedifoneassumesexis- tence of the quadratic variation of M.However,provingtheexistenceofthequadratic variation requires the same kind of arguments as in the proof below (cf. [16]), or, alternatively, a lengthy discussion of general semimartingale theory (cf. [38]).

Proof of Theorem 5.7. Let C (0, ) be a common uniform upper bound for the pro- ∈ ∞ cesses (Ht) and (Mt).Toprovetheestimatein(5.27),weassumew.l.o.g.thatboth partitions π and π contain the end point a,andπ is a refinement of π.Ifthisisnot the case, we may first consider a common refinement and then estimate by the triangle inequality. Under) the additional assumption, we have )

" π π ′ Ia Ia = (Hs H s )(Ms Ms) (5.28) − − ⌊ ⌋ − s π "∈ where from now on, we only sum over partition points less than a, s′ denotes the suc- cessor of s in the ﬁne partition π,and

s := max t π : t s ⌊ ⌋ { ∈ ≤ } University of Bonn ) Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 192 JUMPS is the next partition point of the rough partition π below s.Nowfixε>0.By(5.28), the martingale property for M,andtheadaptednessofH,weobtain ) π π" 2 π π" 2 I I 2 = E (I I ) || − ||M ([0,a]) a − a 2 2 = E (Hs H s #) (Ms′ M's) (5.29) − ⌊ ⌋ − s π "∈ 2# 2 2 ' 2 ε E (M ′ M ) +(2C) E (M ′ M ) ≤ s − s s − s s π t π" s π "∈ "∈ τt(ε") ∈s< t # ' # ≤ ⌈ ⌉ ' where t := min u π : u>t is the next partition point of the rough partition, and ⌈ ⌉ { ∈ } τ (ε) := min s π, s > t : H H >ε t . t ) { ∈ | s − t| }∧⌈⌉ is the first time after t where H deviates substantially from Hs.Notethatτt is a random variable. The summands on the right hand side of (5.29) are now estimatedseparately.SinceM is a bounded martingale, we can easily control the first summand:

~~2 2 2 2 2 2 E (M ′ M ) = E M ′ M = E M M C . (5.30) s − s s − s a − 0 ≤ " " The second# summand is more' difﬁcult# to handle.' Noting# that '~~

~~2 2 2 E (M ′ M ) = E M ′ M on τ s , s − s |Fτt s − s |Fτt { t ≤ } we can rewrite# the expectation' value as # '~~

~~2 E E (M ′ M ) (5.31) s − s |Fτt t π" τt s< t "∈ # ≤"⌈ ⌉ # '' 2 2 2 = E E M t Mτt τt = E (M t Mτt ) =: B ⌈ ⌉ − |F ⌈ ⌉ − t π" t π" "∈ # # '' # "∈ '~~

~~Note that M t Mτt =0only if τt < t ,i.e.,ifH oscillates more than ε in the interval ⌈ ⌉ − ̸ ⌈ ⌉ [t, τt].WecanthereforeusethecàdlàgpropertyofH and M to control (5.31). Let~~

Dε/2 := r [0,a]: Hr Hr >ε/2 { ∈ | − −| } denote the (random) set of “large” jumps of H.SinceH is càdlàg, Dε/2 contains only ﬁnitely many elements. Moreover, for given ε, ε¯ >0 there exists a random variable δ(ω) > 0 such that for u, v [0,a], ∈ Stochastic Analysis Andreas Eberle 5.2. STOCHASTIC INTEGRATION FOR SEMIMARTINGALES 193

(i) u v δ H H ε or (u, v] D = , | − |≤ ⇒|u − v|≤ ∩ ε/2 ̸ ∅ (ii) r D ,u,v[r, r + δ] M M ε¯. ∈ ε/2 ∈ ⇒|u − v|≤ Here we have used that H is càdlàg, Dε/2 is ﬁnite, and M is right continuous. Let ∆ > 0.By(i)and(ii),thefollowingimplicationholdson ∆ δ : { ≤ }

τt < t Hτt Ht >ε [t, τt] Dε/2 = M t Mτt ε¯, ⌈ ⌉⇒| − | ⇒ ∩ ̸ ∅⇒|⌈ ⌉ − |≤ i.e., if τ < t and ∆ δ then the increment of M between τ and t is small. t ⌈ ⌉ ≤ t ⌈ ⌉ Now ﬁx k N and ε¯ >0.ThenwecandecomposeB = B + B where ∈ 1 2 2 2 B1 = E (M t Mτt ) ;∆ δ, Dε/2 k kε¯ , (5.32) ⌈ ⌉ − ≤ | |≤ ≤ t π" # "∈ ' 2 B2 = E (M t Mτt ) ;∆>δor Dε/2 >k ⌈ ⌉ − | | t π" # "∈ ' 2 2 1/2 1/2 E ( (M t Mτt ) ) P ∆ >δor Dε/2 >k (5.33) ≤ ⌈ ⌉ − | | t π" # "∈ ' # ' √6 C2 P ∆ >δ + P D >k 1/2 . ≤ | ε/2| In the last step we have used$ # the following' # upper bound'% for the martingale increments

ηt := M t Mτt : ⌈ ⌉ − 2 2 4 2 2 E ηt = E ηt +2E ηt ηu t π" t t u>t #$ "∈ % ' # " ' # " " ' 4C2E η2 +2E η2E η2 ≤ t t u |Ft t t u>t # " ' # " # " '' 6C2E η2 6C2E M 2 M 2 6C4. ≤ t ≤ a − 0 ≤ t # " ' # ' This estimate holds by the Optional Sampling Theorem, and since E[ η2 ] u>t u |Ft ≤ E[M 2 M 2 ] C2 by the orthogonality of martingale increments M M u − t |Ft ≤ ! Ti+1 − Ti over disjoint time intervals (Ti,Ti+1] bounded by stopping times. We now summarize what we have shown. By (5.29), (5.30) and (5.31),

π π" 2 2 2 2 I I 2 ε C +4C (B + B ) (5.34) || − ||M ([0,a]) ≤ 1 2 where B1 and B2 are estimated in (5.32) and (5.33). Let γ>0 be given. To bound the right hand side of (5.34) by γ we choose the constants in the following way:

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 194 JUMPS~~

1. Choose ε>0 such that C2ε2 <γ/4. 2. Choose k N such that 4√6 C4P D >k 1/2 <γ/4, ∈ | ε/2| 2 2 3. Choose ε¯ >0 such that 4C kε¯ <# γ/4,thenchoosetherandomvariable' δ depending on ε and ε¯ such that (i) and (ii) hold. 4. Choose ∆ > 0 such that 4√6 C4P ∆ >δ 1/2 <γ/4. Then for this choice of ∆ we finally obtain# ' π π" 2 γ I I 2 < 4 = γ || − ||M ([0,a]) · 4 whenever mesh(π) ∆ and π is a refinement of π. ≤ The theorem proves that the stochastic integral H M is well-defined as an M 2 limit of ) −)• the Riemann sum approximations:

Definition (Stochastic integral for left limits of bounded adapted càdlàg processes w.r.t. bounded martingales). For H and M as above, the stochastic integral H M is −• the unique equivalence class of càdlàg ( P ) martingales on [0, ) such that Ft ∞ πn 2 H M = lim H M in Md ([0,a]) −• [0,a] n −• [0,a] →∞ for any a (0, ) and& for any sequence (π ) of& partitions of R with mesh(π ) 0. ∈ ∞ & n & + n → Note that the stochastic integral is defined uniquely only up to càdlàg modifications. We will often denote versions of H M by • Hs dMs,butwewillnotalwaysdistinguish −• 0 − between equivalence classes and their representatives´ carefully. Many basic properties of stochastic integrals with left continuous integrands canbederiveddirectlyfromthe Riemann sum approximations: Lemma 5.8 (Elementary properties of stochastic integrals). For H and M as above, the following statements hold:

1) If t Mt has almost surely ﬁnite variation then H M coincides almost surely #→ −• with the pathwise deﬁned Lebesgue-Stieltjes integral • Hs dMs. 0 − 2) ∆(H M)=H ∆M almost surely. ´ −• − 3) If T :Ω [0, ] is a random variable, and H, H, M, M are processes as → ∞ above such that Ht = Ht for any t

~~Stochastic Analysis ) * Andreas Eberle 5.2. STOCHASTIC INTEGRATION FOR SEMIMARTINGALES 195~~

~~Proof. The statements follow easily by Riemann sum approximation. Indeed, let (πn) be a sequence of partitions of R such that mesh(π ) 0.Thenalmostsurelyalonga + n → subsequence (πn),~~

(H M)t = lim Hs(Ms′ t Ms) ) −• n ∧ − →∞ s t s"≤π"n ∈ w.r.t.uniform convergence on compact intervals. This proves that H M coincides −• almost surely with the Stieltjes integral if M has ﬁnite variation. Moreover, for t>0 it implies

∆(H M)t = lim H t n (Mt Mt )=Ht ∆Mt (5.35) −• n ⌊ ⌋ − − →∞ − almost surely, where t denotes the next partition point of (π ) below t.Sinceboth ⌊ ⌋n n H M and M are càdlàg, (5.35) holds almost surely simultaneously for all t>0.The −• third statement can be proven similarly. )

~~Localization~~

We now extend the stochastic integral to local martingales. It turns out that unbounded jumps can cause substantial difﬁculties for the localization. Therefore, we restrict ourselves to local martingales that can be localized by martingales with bounded jumps. Remark 2 below shows that this is not a substantial restriction.

Suppose that (Mt)t 0 is a càdlàg ( t) adapted process, where ( t) is an arbitrary filtra- ≥ F F tion. For an ( ) stopping time T ,thestoppedprocessM T is defined by Ft T Mt := Mt T for any t 0. ∧ ≥ Definition (Local martingale, Strict local martingale). A localizing sequence for M is a non-decreasing sequence (Tn)n N of ( t) stopping times such that sup TN = , ∈ F ∞ and the stopped process M Tn is an ( ) martingale for each n.TheprocessM is called Ft a local (Ft) martingale iff there exists a localizing sequence. Moreover, M is called a

~~Tn strict local (Ft) martingale iff there exists a localizing sequence (Tn) such that M has uniformly bounded jumps for each n,i.e.,~~

~~sup ∆M (ω) :0 t T (ω) ,ω Ω < n N. {| t | ≤ ≤ n ∈ } ∞∀∈~~

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 196 JUMPS~~

Remark. 1) Any continuous local martingale is a strict local martingale. 2) In general, any local martingale is the sum of a strict localmartingaleandalocal martingale of ﬁnite variation. This is the content of the “Fundamental Theorem of Local Martingales”, cf. [36]. The proof of this theorem, however, is not trivial and is omitted here. The next example indicates how (local) martingales can be decomposed into strict (local) martingales and ﬁnite variation processes:

Example (Lévy martingales). Suppose that X = y (N (dy) tν(dy)) is a compen- t t − sated Lévy jump process on R1 with intensity measure´ ν satisfying ( y y 2) ν(dy) < | |∧| | .Then(X ) is a martingale but, in general, not a strict local martingal´ e. However, ∞ t we can easily decompose Xt = Mt + At where At = yI y >1 (Nt(dy) tν(dy)) ´ {| | } − is a ﬁnite variation process, and Mt = yI y 1 (Nt(dy) tν(dy)) is a strict (local) {| |≤ } − martingale. ´

~~Strict local martingales can be localized by bounded martingales:~~

~~Lemma 5.9. M is a strict local martingale if and only if there exists a localizing se-~~

~~Tn quence (Tn) such that M is a bounded martingale for each n.~~

Proof. If M Tn is a bounded martingale then also the jumps of M Tn are uniformly bounded. To prove the converse implication, suppose that (Tn) is a localizing sequence such that ∆M Tn is uniformly bounded for each n.Then

~~S := T inf t 0: M n ,nN, n n ∧ { ≥ | t|≥ } ∈ is a non-decreasing sequence of stopping times with sup S = ,andthestopped n ∞ processes M Sn are uniformly bounded, since~~

~~Tn Mt Sn n + ∆MSn = n + ∆MS for any t 0. | ∧ |≤ | | | n | ≥~~

Deﬁnition (Stochastic integrals of left limits of adapted càdlàg processes w.r.t. strict P local martingales). Suppose that (Mt)t 0 is a strict local ( t ) martingale, and ≥ F Stochastic Analysis Andreas Eberle 5.2. STOCHASTIC INTEGRATION FOR SEMIMARTINGALES 197

P (Ht)t 0 is càdlàg and ( t ) adapted. Then the stochastic integral H M is the unique ≥ F −• equivalence class of local ( P ) martingales satisfying Ft H M = H M a.s., (5.36) −• [0,T ] −• [0,T ] & & whenever T is an ( P ) stopping& time, H is) a bounded*& càdlàg ( P ) adapted process Ft Ft P with H [0,T ) = H [0,T ) almost surely, and M is a bounded càdlàg ( t ) martingale with | | ) F M = M almost surely. [0,T ] [0,T)] * & & You& should* convince& yourself that the integral H M is well deﬁned by (5.36) because −• of the local dependence of the stochastic integral w.r.t. bounded martingales on H and

~~M (Lemma 5.8, 3). Note that Ht and Ht only have to agree for t~~

~~(n) Tn H M = H M on [0,Tn], −• −• and, by Lemma 5.8, ∆(H(n)M Tn )=H(n)∆M Tn is uniformly bounded for each n. −• −~~

~~Integration w.r.t. semimartingales~~

The stochastic integral w.r.t. a semimartingale can now easily be deﬁned via a semimartingale decomposition. Indeed, suppose that X is an ( P ) semimartingale with Ft decomposition X = X + M + A ,t 0, t 0 t t ≥ into a strict local ( P ) martingale M and an ( P ) adapted process A with càdlàg ﬁnite- Ft Ft variation paths t A (ω). #→ t University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 198 JUMPS

Deﬁnition (Stochastic integrals of left limits of adapted càdlàg processes w.r.t. semi- P martingales ). For any ( t ) adapted process (Ht)t 0 with càdlàg paths, the stochastic F ≥ integral of H w.r.t. X is deﬁned by

H X = H M + H A, −• −• −• where M and A are the strict local martingale part and the ﬁnite variation part in asemimartingaledecompositionasabove,H M is the stochastic integral of H −• − t w.r.t. M,and(H A)t = Hs dAs is the pathwise deﬁned Stieltjes integral of H −• 0 − − w.r.t. A. ´

Note that the semimartingale decomposition of X is not unique. Nevertheless, the integral H X is uniquely deﬁned up to modiﬁcations: −• Theorem 5.10. Suppose that (π ) is a sequence of partitions of R with mesh(π ) 0. n + n → Then for any a [0, ), ∈ ∞

~~(H X)t = lim Hs(Xs′ t Xs) −• n ∧ − →∞ s πn "s~~

continuous w.r.t. P then each version of the integral (H X)P deﬁned w.r.t. P is −• aversionoftheintegral(H X)Q deﬁned w.r.t. Q. −• The proofs of this and the next theorem are left as exercises tothereader.

Theorem 5.11 (Elementary properties of stochastic integrals). P 1) Semimartingale decomposition:TheintegralH X is again an ( t ) semi- −• F martingale with decomposition H X = H M + H A into a strict local mar- −• −• −• tingale and an adapted ﬁnite variation process.

~~Stochastic Analysis Andreas Eberle 5.3. QUADRATIC VARIATION AND COVARIATION 199~~

~~2) Linearity:Themap(H, X) H X is bilinear. #→ −• 3) Jumps: ∆(H X)=H ∆X almost surely. −• − 4) Localization:IfT is an ( P ) stopping time then Ft~~

~~T T (H X) = H X =(H I[0,T )) X. −• −• · −•~~

~~5.3 Quadratic variation and covariation~~

From now on we fix a probability space (Ω, ,P) with a filtration ( ).Thevector A Ft space of (equivalence classes of) strict local ( P ) martingales and of ( P ) adapted Ft Ft processes with càdlàg finite variation paths are denoted by Mloc and FV respectively. Moreover,

= M +FV S loc denotes the vector space of ( P ) semimartingales. If there is no ambiguity, we do not Ft distinguish carefully between equivalence classes of processes and their representatives. The stochastic integral induces a bilinear map , (H, X) H X on the S×S→S #→ −• equivalence classes that maps M to M and FV to FV. S× loc loc S× Asuitablenotionofconvergenceon(equivalenceclassesof)semimartingalesisuniform convergence in probability on compact time intervals:

~~Deﬁnition (ucp-convergence). AsequenceofsemimartingalesX converges to a n ∈S limit X uniformly on compact intervals in probability iff ∈S~~

~~n P R sup Xt Xt 0 as n for any a +. t a | − | −→ →∞ ∈ ≤~~

By Theorem (5.10), for H, X and any sequence of partitions with mesh(π ) 0, ∈S n → the stochastic integral H dX is a ucp-limit of predictable Riemann sum approxima- − tions, i.e., of the integrals´ of the elementary predictable processes Hπn . − University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 200 JUMPS

~~Covariation and integration by parts~~

The covariation is a symmetric bilinear map FV.Insteadofgoingonce S×S→ more through the Riemann sum approximations, we can use what we have shown for stochastic integrals and deﬁne the covariation by the integration by parts identity

~~t t XtYt X0Y0 = Xs dYs + Ys dXs +[X, Y ]t. − ˆ0 − ˆ0 − The approximation by sums is then a direct consequence of Theorem 5.10.~~

~~Deﬁnition (Covariation of semimartingales). For X, Y , ∈S~~

~~[X, Y ] := XY X0Y0 X dY Y dX. − − ˆ − − ˆ −~~

Clearly, [X, Y ] is again an ( P ) adapted càdlàg process. Moreover, (X, Y ) [X, Y ] Ft #→ is symmetric and bilinear, and hence the polarization identity 1 [X, Y ]= ([X + Y ] [X] [Y ]) 2 − − holds for any X, Y where ∈S [X]=[X, X] denotes the quadratic variation of X.Thenextcorollaryshowsthat[X, Y ] deserves the name “covariation”:

~~Corollary 5.12. For any sequence (π ) of partitions of R with mesh(π ) 0, n + n →~~

~~[X, Y ]t =ucplim (Xs′ t Xs)(Ys′ t Ys). (5.37) − n ∧ − ∧ − →∞ s πn "s~~

~~Stochastic Analysis Andreas Eberle 5.3. QUADRATIC VARIATION AND COVARIATION 201~~

~~Proof. (5.37) is a direct consequence of Theorem 5.10, and 1) followsfrom(5.37)and the polarization identity. 2) follows from Theorem 5.11, which yields~~

~~∆[X, Y ]=∆(XY ) ∆(X Y ) ∆(Y X) − −• − −• = X ∆Y + Y ∆X +∆X∆Y X ∆Y Y ∆X − − − − − − =∆X∆Y.~~

~~3) follows similarly and is left as an exercise and 4) holds by (5.37) and the Cauchy- Schwarz formula for sums.~~

~~Statements 1) and 2) of the corollary show that [X, Y ] is a ﬁnite variation process with decomposition c [X, Y ]t =[X, Y ]t + ∆Xs∆Ys (5.38) s t "≤ into a continuous part and a pure jump part.~~

~~If Y has ﬁnite variation then by Lemma 5.2,~~

~~[X, Y ]t = ∆Xs∆Ys. s t "≤ Thus [X, Y ]c =0and if, moreover, X or Y is continuous then [X, Y ]=0.~~

~~More generally, if X and Y are semimartingales with decompositions X = M + A, Y = N + B into M,N M and A, B FV then by bilinearity, ∈ loc ∈~~

~~[X, Y ]c =[M,N]c +[M,B]c +[A, N]c +[A, B]c =[M,N]c.~~

~~It remains to study the covariations of the local martingale parts which turn out to be the key for controlling stochastic integrals effectively.~~

~~Quadratic variation and covariation of local martingales~~

~~If M is a strict local martingale then by the integration by parts identity, M 2 [M] is a − strict local martingale as well. By localization and stopping we can conclude:~~

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 202 JUMPS~~

Theorem 5.13. Let M and a [0, ).ThenM 2([0,a]) if and only if ∈Mloc ∈ ∞ ∈Md M 2 and [M] 1.Inthiscase,M 2 [M] (0 t a) is a martingale, and 0 ∈L a ∈L t − t ≤ ≤ 2 2 M 2 = E M + E [M] . (5.39) || ||M ([0,a]) 0 a Proof. We may assume M =0;otherwiseweconsider# ' #M = 'M M .Let(T ) be a 0 − 0 n joint localizing sequence for the local martingales M and M 2 [M] such that M Tn is * − bounded. Then by optional stopping,

~~2 N E Mt Tn = E [M]t Tn for any t 0 and any n . (5.40) ∧ ∧ ≥ ∈ Since M#2 is a submartingale,' # we have'~~

~~2 2 2 E[Mt ] lim inf E[Mt Tn ] E[Mt ] (5.41) n ∧ ≤ →∞ ≤ by Fatou’s lemma. Moreover, by the Monotone Convergence Theorem,~~

~~E [M]t = lim E [M]t Tn . n ∧ →∞ Hence by (5.41), we obtain # ' # '~~

E[M 2]=E [M] for any t 0. t t ≥ For t a,theright-handsideisdominatedfromaboveby# ' E [M] ,Therefore,if[M] ≤ a a 2 2 is integrable then M is in Md ([0,a]) with M norm E [M]a# .Moreover,inthiscase,' the sequence M 2 [M] is uniformly integrable for each t [0,a],because, t Tn t Tn n N # ' ∧ − ∧ ∈ ∈ $ 2 % 2 1 sup Mt [M]t sup Mt +[M]a , t a | − |≤ t a | | ∈L ≤ ≤ 2 Therefore, the martingale property carries over from the stopped processes Mt Tn ∧ − 2 [M]t Tn to Mt [M]t. ∧ − Remark. The assertion of Theorem 5.13 also remains valid for a = in the sense that ∞ 2 1 if M0 is in and [M] = limt [M]t is in then M extends to a square integrable L ∞ →∞ L martingale (Mt)t [0, ] satisfying (5.40) with a = .TheexistenceofthelimitM = ∈ ∞ ∞ ∞ 2 limt Mt follows in this case from the L Martingale Convergence Theorem. →∞ The next corollary shows that the M 2 norms also control the covariations of square integrable martingales.

~~Stochastic Analysis Andreas Eberle 5.3. QUADRATIC VARIATION AND COVARIATION 203~~

~~Corollary 5.14. The map (M,N) [M,N] is symmetric, bilinear and continuous on #→ 2 Md ([0,a]) in the sense that~~

~~E[sup [M,N] t] M M 2([0,a]) N M 2([0,a]). t a | | ≤|||| || || ≤ Proof. By the Cauchy-Schwarz inequality for the covariation (Cor. 5.12,4),~~

~~[M,N] [M]1/2[N]1/2 [M]1/2[N]1/2 t a. | t|≤ t t ≤ a a ∀ ≤ Applying the Cauchy-Schwarz inequality w.r.t.the L2-inner product yields~~

~~1/2 1/2 E[sup [M,N]t ] E [M]a E [N]a M M 2([0,a]) N M 2([0,a]) t a | | ≤ ≤|||| || || ≤ # ' # ' by Theorem 5.13.~~

Corollary 5.15. Let M and suppose that [M] =0almost surely for some ∈Mloc a a [0, ].Thenalmostsurely, ∈ ∞ M = M for any t [0,a]. t 0 ∈ In particular, continuous local martingales of ﬁnite variation are almost surely constant.

Proof. By Theorem 5.13, M M 2 = E [M] =0. || − 0||M ([0,a]) a The assertion also extends to the case when a is replaced# by' a stopping time. Combined with the existence of the quadratic variation, we have now proven:

~~»Non-constant strict local martingales have non-trivial quadratic variation«~~

Example (Fractional Brownian motion is not a semimartingale). Fractional Brow- nian motion with Hurst index H (0, 1) is deﬁned as the unique continuous Gaussian ∈ H process (Bt )t 0 satisfying ≥ 1 E BH =0 and Cov BH ,BH = t2H + s2H t s 2H t s t 2 −| − | # ' # ' $ % for any s, t 0.IthasbeenintroducedbyMandelbrotasanexampleofaself-similar ≥ process and is used in various applications, cf. [2]. Note that for H =1/2,thecovari- ance is equal to min(s, t),i.e.,B1/2 is a standard Brownian motion. In general, one can

University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 204 JUMPS prove that fractional Brownian motion exists for any H (0, 1),andthesamplepaths ∈ t BH (ω) are almost surely α-Hölder continuous if and only if α

V (1)(BH )= for any t>0 almost surely , and t ∞ 0 if H>1/2 , H H H 2 ⎧ [B ]t = lim Bs′ t Bs = t if H =1/2 , n ∧ − ⎪ →∞ s πn ⎪ "s1/2 and assume that there is a decomposition Bt = Mt + At into a continuous local martingale M and a continuous ﬁnite variation process A.Then

[M]=[BH ]=0 almost surely , so by Corollary 5.15, M is almost surely constant, i.e., BH has ﬁnite variation paths. Since this is a contradiction, we see that also for H>1/2, BH is not a continuous semimartingale,i.e.,thesumofacontinuouslocalmartingaleandacontinuous adapted ﬁnite variation process. It is possible (but beyond the scopeofthesenotes)toprovethat any semimartingale that is continuous is a continuous semimartingale in the sense above (cf. [36]). Hence for H =1/2,fractionalBrownianmotionisnotasemimartingaleand ̸ classical Itô calculus is not applicable. Rough paths theoryprovidesanalternativeway to develop a calculus w.r.t. the paths of fractional Brownianmotion,cf.[19].

~~The covariation [M,N] of local martingales can be characterized in an alternative way that is often useful for determining [M,N] explicitly.~~

Theorem 5.16 (Martingale characterization of covariation). For M,N M ,the ∈ loc covariation [M,N] is the unique process A FV such that ∈ (i) MN A M , and − ∈ loc (ii) ∆A =∆M ∆N,A0 =0 almost surely .

~~Proof. Since [M,N]=MN M0N0 M dN N dM,(i)and(ii)aresatisﬁed − − − − − for A =[M,N].NowsupposethatA is´ another process´ in FV satisfying (i) and (ii).~~

~~Stochastic Analysis ) Andreas Eberle 5.3. QUADRATIC VARIATION AND COVARIATION 205~~

Then A A is both in M and in FV,and∆(A A)=0almost surely. Hence A A − loc − − is a continuous local martingale of ﬁnite variation, and thus A A = A0 A0 =0 ) ) − − ) almost surely by Corollary 5.15. ) ) The covariation of two local martingales M and N yields a semimartingale decomposition for MN: MN = local martingale +[M,N]. However, such a decomposition is not unique. By Corollary 5.15 it is unique if we assume in addition that the ﬁnite variation part A is continuous with A0 =0(which is not the case for A =[M,N] in general).

Deﬁnition. Let M,N M .IfthereexistsacontinuousprocessA FV such that ∈ loc ∈ (i) MN A M , and − ∈ loc (ii) ∆A =0,A0 =0almost surely, then M,N = A is called the conditional covariance process of M and N. ⟨ ⟩ In general, a conditional covariance process as deﬁned aboveneednotexist.General martingale theory (Doob-Meyer decomposition) yields the existence under an additional assumption if continuity is replaced by predictability, cf.e.g. [36].Forapplicationsitis more important that in many situations the conditional covariance process can be easily determined explicitly, see the example below.

Corollary 5.17. Let M,N M . ∈ loc 1) If M is continuous then M,N =[M,N] almost surely. ⟨ ⟩ 2) In general, if M,N exists then it is unique up to modiﬁcations. ⟨ ⟩ 3) If M exists then the assertions of Theorem 5.13 hold true with [M] replaced by ⟨ ⟩ M . ⟨ ⟩ Proof. 1) If M is continuous then [M,N] is continuous. 2) Uniqueness follows as in the proof of 5.16. 3) If (T ) is a joint localizing sequence for M 2 [M] and M 2 M then, by monotone n − −⟨ ⟩ convergence,

E M t = lim E M t Tn = lim E [M]t Tn = E [M]t n ∧ n ∧ ⟨ ⟩ →∞ ⟨ ⟩ →∞ for any#t 0'.TheassertionsofTheorem5.13nowfollowsimilarlyasabov# ' # ' #e. ' ≥ University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 206 JUMPS

Examples (Covariations of Lévy processes). Rd k 1) Brownian motion:If(Bt) is a Brownian motion in then the components (Bt ) are independent one-dimensional Brownian motions. Therefore,theprocessesBkBl δ t t t − kl are martingales, and hence almost surely,

~~[Bk,Bl] = Bk,Bl = t δ for any t 0. t ⟨ ⟩t · kl ≥ 2) Lévy processes without diffusion part:Let~~

Xt = y Nt(dy) tI y 1 ν(dy) + bt ˆRd 0 − {| |≤ } \{ } $ % with b Rd,aσ-finite measure ν on Rd 0 satisfying ( y 2 1) ν(dy) < ,anda ∈ \{ } | | ∧ ∞ Poisson point process (N ) of intensity ν.Supposefirstthat´ supp(ν) y Rd : y ε t ⊂ ∈ | |≥ k for some ε>0.ThenthecomponentsX are finite variation processes,< and hence = [Xk,Xl] = ∆Xk∆Xl = ykyl N (dy). (5.42) t s s ˆ t s t "≤ In general, (5.42) still holds true. Indeed, if X(ε) is the corresponding Lévy process with (ε) (ε),k k intensity measure ν (dy)=I y ε ν(dy) then X X M 2([0,a]) 0 as ε 0 {| |≥ } || − || → ↓ for any a R and k 1,...,d ,andhencebyCorollary5.14, ∈ + ∈{ } k l (ε),k (ε),l k l X ,X =ucplim X ,X = ∆Xs ∆Xs. t ε 0 t ↓ s t # ' # ' "≤ On the other hand, we know that if X is square integrable then Mt = Xt it ψ(0) and 2 − ∇ M kM l t ∂ ψ (0) are martingales, and hence t t − ∂pk∂pl 2 k l k l ∂ ψ X ,X t = M ,M t = t (0). ⟨ ⟩ ⟨ ⟩ · ∂pk∂pl 3) Covariations of Brownian motion and Lévy jump processes:ForB and X as above we have

Bk,Xl =[Bk,Xl]=0 almost surely for any k and l. (5.43) ⟨ ⟩ Indeed, (5.43) holds true if Xl has finite variation paths. The general case then follows once more by approximating Xl by finite variation processes. Note that independence of B and X has not been assumed!WewillseeinSection3.1that(5.43)implies that a Brownian motion and a Lévy process without diffusion term defined on the same probability space are always independent.

~~Stochastic Analysis Andreas Eberle 5.3. QUADRATIC VARIATION AND COVARIATION 207~~

~~Covariation of stochastic integrals~~

We now compute the covariation of stochastic integrals. Thisisnotonlycrucialfor many computations, but it also yields an alternative characterization of stochastic integrals w.r.t. local martingales, cf. Corollary 5.19 below.

~~Theorem 5.18. Suppose that X and Y are ( P ) semimartingales, and H is ( P ) Ft Ft adapted and càdlàg. Then~~

H dX, Y = H d[X, Y ] almost surely. (5.44) ˆ − ˆ − Proof. 1. We# first note that' (5.44) holds if X or Y has finite variation paths. If, for example, X FV then also H dX FV,andhence ∈ ´ − ∈ H dX, Y = ∆(H X)∆Y = H ∆X∆Y = H d[X, Y ] . ˆ − −• − ˆ − " " The# same holds' if Y FV. ∈ 2. Now we show that (5.44) holds if X and Y are bounded martingales, and H is bounded. For this purpose, we fix a partition π,andweapproximateH by the elemen- − π ′ tary process H = s π Hs I(s,s ].Let − ∈ · ! π π ′ It = H dX = Hs(Xs t Xs). ˆ − ∧ − (0,t] s π "∈ We can easily verify that

[Iπ,Y]= Hπ d[X, Y ] almost surely. (5.45) ˆ − Indeed, if (π ) is a sequence of partitions such that π π for any n and mesh(π ) 0 n ⊂ n n → then ) ) ) π π ′ ′ ′ (Ir′ t Ir )(Yr t Yr)= Hs (Xr t Xr)(Yr t Yr). ∧ − ∧ − ∧ − ∧ − r π"n s π r π"n "r

~~Having veriﬁed (5.45) for any ﬁxed partition π,wechooseagainasequence(πn) of partitions with mesh(π ) 0.Then n →~~

~~H dX = lim Iπn in M 2([0,a]) for any a (0, ), ˆ − n →∞ ∈ ∞ and hence, by Corollary 5.14 and (5.45),~~

H dX, Y =ucplim [Iπn ,Y]= H d[X, Y ]. ˆ − n ˆ − →∞ # ' 3. Now suppose that X and Y are strict local martingales. If T is a stopping time such T T that X and Y are bounded martingales, and HI[0,T ) is bounded as well, then by Step 2, Theorem 5.11 and Corollary 5.12,

~~T T T T T H dX, Y = H dX ,Y = (H I[0,T )) dX ,Y ˆ − ˆ − ˆ −~~

# ' #$ % T 'T # T ' = H I[0,T ) d[X ,Y ]= H d[X, Y ] . ˆ − ˆ − $ % Since this holds for all localizing stopping times as above, (5.45) is satisﬁed as well. 4. Finally, suppose that X and Y are arbitrary semimartingales. Then X = M + A and Y = N + B with M,N M and A, B FV.Theassertion(5.44)nowfollowsby ∈ loc ∈ Steps 1 and 3 and by the bilinearity of stochastic integral andcovariation.

~~Perhaps the most remarkable consequences of Theorem 5.18 is:~~

~~Corollary 5.19 (Kunita-Watanabe characterization of stochastic integrals). P Let M Mloc and G = H with H ( t ) adapted and càdlàg. Then G M is the unique ∈ − F • element in Mloc satisfying~~

~~(i) (G M)0 =0, and • (ii) [G M,N]=G [M,N] for any N Mloc. • • ∈ Proof. By Theorem 5.18, G M satisﬁes (i) and (ii). It remains to prove uniqueness. Let • L M such that L =0and ∈ loc 0~~

~~[L, N]=G [M,N] for any N Mloc. • ∈ Stochastic Analysis Andreas Eberle 5.3. QUADRATIC VARIATION AND COVARIATION 209~~

~~Then [L G M,N]=0for any N Mloc.ChoosingN = L G M,weconclude − • ∈ − • that [L G M]=0.HenceL G M is almost surely constant, i.e., − • − •~~

~~L G M L0 (G M)0 =0. − • ≡ − •~~

~~Remark. Localization shows that it is sufﬁcient to verify Condition (ii) in the Kunita- Watanabe characterization for bounded martingales N.~~

The corollary tells us that in order to identify stochastic integrals w.r.t. local martingales it is enough to “test” with other (local) martingales via the covariation. This fact can be used to give an alternative deﬁnition of stochastic integrals that applies to general pre- P dictable integrands. Recall that a stochastic process (Gt)t 0 is called ( t ) predictable ≥ F iff the function (ω,t) G (ω) is measurable w.r.t. the σ-algebra P on Ω [0, ) → t × ∞ generated by all ( P ) adapted left-continuous processes. Ft Deﬁnition (Stochastic integrals with general predictable integrands). Let M M ,andsupposethatG is an ( P ) predictable process satisfying ∈ loc Ft t 2 Gs d[M]s < almost surely for any t 0. ˆ0 ∞ ≥

~~If there exists a local martingale G M Mloc such that conditions (i) and (ii) in Corol- • ∈ lary 5.19 hold, then G M is called the stochastic integral of G w.r.t. M. •~~

~~Many properties of stochastic integrals can be deduced directly from this deﬁnition, see e.g. Theorem 5.21 below.~~

~~The Itô isometry for stochastic integrals w.r.t. martingales~~

Of course, Theorem 5.18 can also be used to compute the covariation of two stochastic integrals. In particular, if M is a semimartingale and G = H with H càdlàg and − adapted then [G M,G M]=G [M,G M]=G2 [M]. • • • • •

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 210 JUMPS~~

Corollary 5.20 (Itô isometry for martingales). Suppose that M M .Thenalso ∈ loc 2 2 GdM G d[M] Mloc,and ´ − ´ ∈ $ % 2 a 2 a GdM = E GdM = E G2 d[M] a 0, a.s . 2 ˆ M ([0,a]) ˆ0 ˆ0 ∀ ≥ && && /6 7 0 / 0 && && 2 &Proof.& If M&& Mloc then G M Mloc,andhence(G M) [G M] Mloc.Moreover, ∈ • ∈ • − • ∈ by Theorem 5.13,

~~2 2 G M M 2([0,a]) = E [G M]a = E (G [M])a . || • || • • # ' # '~~

The Itô isometry for martingales states that the M 2([0,a]) norm of the stochastic integral GdMcoincides with the L2 Ω (0,a],P norm of the integrand (ω,t) G (ω), × [M] #→ t ´where P is the measure on Ω R given by [M] $ × + %

~~P[M](dωdt)=P (dω)[M](ω)(dt).~~

~~This can be used to prove the existence of the stochastic integral for general predictable integrands G L2(P ),cf.Section2.5below. ∈ [M]~~

~~5.4 Itô calculus for semimartingales~~

We are now ready to prove the two most important rules of Itô calculus for semimartingales: The so-called “Associative Law” which tells us how to integrate w.r.t. processes that are stochastic integrals themselves, and the change of variables formula.

~~Integration w.r.t. stochastic integrals~~

Suppose that X and Y are semimartingales satisfying dY = GdXfor some predictable integrand G,i.e.,Y Y0 = GdX.Wewouldliketoshowthatweareallowedto − ) multiply the differential equation´ formally by another predictable process G,i.e.,we ) ) would like to prove that GdY = GGdX: ´ ´ dY = GdX =) GdY = GGdX ⇒ Stochastic Analysis ) ) Andreas Eberle 5.4. ITÔ CALCULUS FOR SEMIMARTINGALES 211

~~The covariation characterization of stochastic integrals w.r.t.local martingales can be used to prove this rule in a simple way.~~

Theorem 5.21 (“Associative Law”). Let X .Then ∈S G (G X)=(GG) X (5.46) • • • holds for any processes G = H and G = H with H and H càdlàg and adapted. − ) − ) Remark. The assertion extends with) a similar) proof to more) general predictable integrands.

Proof. We already know that (5.46) holds for X FV.Therefore,andbybilinearityof ∈ the stochastic integral, we may assume X M .BytheKunita-Watanabecharacteri- ∈ loc zation it then sufﬁces to “test” the identity (5.46) with local martingales. For N M , ∈ loc Corollary 5.19 and the associative law for FV processes imply

~~[G (G X),N]=G [G X, N]=G (G [X, N]) • • • • • • =(GG) [X, N] = [(GG) X, N]. ) ) • ) • Thus (5.46) holds by Corollary 5.19. ) )~~

~~Itô’s formula~~

We are now going to prove a change of variables formula for discontinuous semimartingales. To get an idea how the formula looks like we first briefly consider a semimartingale X with a finite number of jumps in finite time. Suppose that ∈S 0

~~[Tk 1,Tk), X is continuous. Therefore, by a similar argument as in the proof of Itô’s − formula for continuous paths (cf. [14, Thm.6.4]), we could expect that~~

~~F (Xt) F (X0)= F (Xt T ) F (Xt T − ) − ∧ k − ∧ k 1 "k t Tk $ t Tk % ∧ − 1 ∧ − = F ′(Xs ) dXs + F ′′(Xs ) d[X]s + F (XT ) F (XT ) ˆ − 2 ˆ − k − k− Tk−1~~

c where Xt = Xt s t ∆Xs denotes the continuous part of X.However,thisformula − ≤ does not carry over! to the case when the jumps accumulate and the paths are not of ﬁnite variation, since then the series may diverge and the continuous part Xc does not exist in general. This problem can be overcome by rewriting (5.47) in the equivalent form

~~t t 1 c F (Xt) F (X0)= F ′(Xs ) dXs + F ′′(Xs ) d[X]s (5.48) − ˆ0 − 2 ˆ0 −~~

~~+ F (Xs) F (Xs ) F ′(Xs )∆Xs , − − − − s t "≤ $ % which carries over to general semimartingales.~~

1 d Theorem 5.22 (Itô’s formula for semimartingales). Suppose that Xt =(Xt ,...,Xt ) with semimartingales X1,...,Xd .ThenforeveryfunctionF C2(Rd), ∈S ∈ d d 2 ∂F i 1 ∂ F i j c F (Xt) F (X0)= (Xs ) dXs + (Xs ) d[X ,X ]s − ˆ ∂xi − 2 ˆ ∂xi∂xj − i=1 i,j=1 "(0,t] "(0,t] d ∂F i + F (Xs) F (Xs ) (Xs )∆Xs (5.49) − − − ∂xi − s (0,t] i=1 "∈ $ " % for any t 0,almostsurely. ≥ i Remark. The existence of the quadratic variations [X ]t implies the almost sure absolute convergence of the series over s (0,t] on the right hand side of (5.49). Indeed, a ∈ Taylor expansion up to order two shows that

d ∂F i i 2 F (Xs) F (Xs ) (Xs )∆Xs Ct ∆Xs | − − − ∂xi − |≤ · | | s t i=1 s t i "≤ " "≤ " C [Xi] < , ≤ t · t ∞ i " where Ct = Ct(ω) is an almost surely ﬁnite random constant depending only on the maximum of the norm of the second derivative of F on the convex hull of X : s [0,t] . { s ∈ } It is possible to prove this general version of Itô’s formula by a Riemann sum approximation, cf. [36]. Here, following [38], we instead derive the “chain rule” once more from the “product rule”:

~~Stochastic Analysis Andreas Eberle 5.4. ITÔ CALCULUS FOR SEMIMARTINGALES 213~~

~~Proof. To keep the argument transparent, we restrict ourselves to the case d =1.The generalization to higher dimensions is straightforward. Wenowproceedinthreesteps:~~

~~1. As in the ﬁnite variation case (Theorem 5.4), we ﬁrst prove that the set consisting A of all functions F C2(R) satisfying (5.48) is an algebra, i.e., ∈~~

~~F, G = FG . ∈A ⇒ ∈A~~

~~This is a consequence of the integration by parts formula~~

~~t t F (Xt)G(Xt) F (X0)G(X0)= F (X ) dG(X)+ G(X ) dF (X) − ˆ0 − ˆ0 − c + F (X),G(X) + ∆F (X)∆G(X), (5.50) (0,t] # ' " the associative law, which implies~~

~~1 c F (X ) dG(X)= F (X )G′(X ) dX + F (X )G′′(X ) d[X] ˆ − ˆ − − 2 ˆ − −~~

~~+ F (X )(∆G(X) G′(X )∆X), (5.51) − − − " the corresponding identity with F and G interchanged, and the formula~~

c c [F (X),G(X)] = F ′(X ) dX, G′(X ) dX (5.52) ˆ − ˆ − / c 0 c = F ′(X )G′(X ) d[X] = (F ′G′)(X ) d[X] ˆ − − ˆ − 6 7 for the continuous part of the covariation. Both (5.51) and (5.52) follow from (5.49) and the corresponding identity for G.Itisstraightforwardtoverifythat(5.50),(5.51)and (5.52) imply the change of variable formula (5.48) for FG,i.e.,FG .Therefore, ∈A by induction, the formula (5.48) holds for all polynomials F .

2. In the second step, we prove the formula for arbitrary F C2 assuming X = M +A ∈ with a bounded martingale M and a bounded process A FV.Inthiscase,X is ∈ uniformly bounded by a ﬁnite constant C.Therefore,thereexistsasequence(pn) of

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 214 JUMPS~~

~~polynomials such that p F , p′ F ′ and p′′ F ′′ uniformly on [ C, C].For n → n → n → − t 0,weobtain ≥~~

F (Xt) F (X0) = lim pn(Xt) pn(X0) n − →∞ − t t Xs y $ 1 % c = lim pn′ (Xs ) dXs + pn′′ (Xs ) d[X]s + pn′′ (z) dz dy n ˆ − 2 ˆ − ˆ ˆ →∞ 0 0 s t Xs− Xs− 6 "≤ 7 t t Xs y 1 c = F ′(Xs ) dXs + F ′′(Xs ) d[X]s + F ′′(z) dz dy ˆ − 2 ˆ − ˆ ˆ 0 0 s t Xs− Xs− "≤ w.r.t.convergence in probability. Here we have used an expression of the jump terms in (5.48) by a Taylor expansion. The convergence in probabilityholdssinceX = M + A,

t t 2 E pn′ (Xs ) dMs F ′(Xs ) dMs ˆ0 − − ˆ0 − /& t & 0 & 2 & 2 =& E (pn′ F ′)(Xs ) d[M]s & sup pn′ F ′ E [M]t ˆ0 − − ≤ [ C,C] | − | · / 0 − # ' by Itô’s isometry, and

Xs y 1 2 (p′′ F ′′)(z) dz dy sup p′′ F ′′ (∆X ) . ˆ ˆ n − ≤ 2 | n − | s s t Xs− Xs− [ C,C] s t & ≤ & − ≤ & " & " 3. Finally,& the change of variables formula& for general semimartingales X = M + A with M M and A FV follows by localization. We can ﬁnd an increasing se- ∈ loc ∈ quence of stopping times (T ) such that sup T = a.s., M Tn is a bounded martingale, n n ∞ Tn and the process A − deﬁned by

~~At for t~~

~~n Tn Tn is a bounded process in FV for any n.Itô’sformulathenholdsfor⎩ X := M + A − for every n.SinceXn = X on [0,T ) and T a.s., this implies Itô’s formula for n n ↗∞ X.~~

Note that the second term on the right hand side of Itô’s formula (5.49) is a continuous ﬁnite variation process and the third term is a pure jump ﬁnitevariationprocess.More- over, semimartingale decompositions of Xi, 1 i d,yieldcorrespondingdecomposi- ≤ ≤ tions of the stochastic integrals on the right hand side of (5.49). Therefore, Itô’s formula

~~Stochastic Analysis Andreas Eberle 5.4. ITÔ CALCULUS FOR SEMIMARTINGALES 215~~

~~1 d can be applied to derive an explicit semimartingale decomposition of F (Xt ,...,Xt ) for any C2 function F .Thiswillnowbecarriedoutinconcreteexamples.~~

~~Application to Lévy processes~~

~~We ﬁrst apply Itô’s formula to a one-dimensional Lévy process~~

X = x + σB + bt + y N (dy) (5.53) t t ˆ t with x, σ, b R,aBrownianmotion(B ),andacompensatedPoissonpointprocess) ∈ t N = N tν with intensity measure ν.Weassumethat ( y 2 y ) ν(dy) < .The t t − | | ∧| | ∞ only restriction to the general case is the assumed integrab´ ility of y at ,whichen- ) | | ∞ sures in particular that (Xt) is integrable. The process (Xt) is a semimartingale w.r.t. the ﬁltration ( B,N) generated by the Brownian motion and the Poisson point process. Ft We now apply Itô’s formula to F (X ) where F C2(R).SettingC = y N (dy) we t ∈ t t ﬁrst note that almost surely, ´ ) 2 2 2 [X]t = σ [B]t +2σ[B,C]t +[C]t = σ t + (∆Xs) . s t "≤ Therefore, by (5.54),

F (X ) F (X ) t − 0 t t 1 c = F ′(X ) dX + F ′′(X ) d[X] + F (X) F (X ) F ′(X )∆X ˆ − 2 ˆ − − − − − 0 0 s t ≤ t t " $ % 1 2 = (σF ′)(Xs ) dBs + (bF ′ + σ F ′′)(Xs) ds + F ′(Xs ) y N(ds dy) ˆ0 − ˆ0 2 ˆ − (0,t] R × ) + F (Xs + y) F (Xs ) F ′(Xs )y N(ds dy), (5.54) ˆ − − − − − (0,t] R × $ % where N(ds dy) is the Poisson random measure on R R corresponding to the Pois- + × son point process, and N(ds dy)=N(ds dy) ds ν(dy).Here,wehaveusedarulefor − evaluating a stochastic integral w.r.t.the process Ct = y Nt(dy) which is intuitively ) clear and can be veriﬁed by approximating the integrand´ by elementary processes. Note ) University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 216 JUMPS

also that in the second integral on the right hand side we couldreplaceXs by Xs since − almost surely, ∆Xs =0for almost all s. To obtain a semimartingale decomposition from (5.54), we note that the stochastic integrals w.r.t. (Bt) and w.r.t. (Nt) are local martingales. By splitting the last integral on the right hand side of (5.54) into an integral w.r.t. N(ds dy) (i.e., a local martingale) and an ) integral w.r.t. the compensator ds ν(dy),wehaveproven: ) Corollary 5.23 (Martingale problem for Lévy processes). For any F C2(R),the ∈ process t [F ] Mt = F (Xt) F (X0) ( F )(Xs) ds, − − ˆ0 L 1 ( F )(x)= (σF ′′)(x)+(bF ′)(x)+ F (x + y) F (x) F ′(x)y ν(dy), L 2 ˆ − − is a local martingale vanishing at 0.ForF 2$(R), M [F ] is a martingale, and% ∈Cb 1 ( F )(x) = lim E F (Xt) F (X0) . t 0 L ↓ t − [F ] # ' Proof. M is a local martingale by the considerations above and since Xs(ω)= 2 Xs (ω) for almost all (s, ω).ForF b , F is bounded since F (x + y) F (x) − ∈C L − − 2 [F ] F ′(x)y = ( y y ).HenceM is a martingale in this case,& and O | |∧| | & &1 1 t & E F (Xt) F (X0) = E ( F )(Xs) ds ( F )(x) t − t ˆ0 L → L / 0 as t 0 by right# continuity of'( F )(X ). ↓ L s The corollary shows that is the inﬁnitesimal generator of the Lévy process. The L martingale problem can be used to extend results on the connection between Brownian motion and the Laplace operator to general Lévy processes andtheirgenerators.Forex- ample, exit distributions are related to boundary value problems (or rather complement value problems as is not a local operator), there is a potential theory for generators of L Lévy processes, the Feynman-Kac formula and its applications carry over, and so on.

Example (Fractional powers of the Laplacian). By Fourier transformation one veri- ﬁes that the generator of a symmetric α-stable process with characteristic exponent p α | | is = ( ∆)α/2.Thebehaviourofsymmetricα-stable processes is therefore closely L − − linked to the potential theory of these well-studied pseudo-differential operators.

~~Stochastic Analysis Andreas Eberle 5.4. ITÔ CALCULUS FOR SEMIMARTINGALES 217~~

~~Exercise (Exit distributions for compound Poisson processes). Let (Xt)t 0 be a com- ≥ pound Poisson process with X0 =0and jump intensity measure ν = N(m, 1),m>0.~~

~~i) Determine λ R such that exp(λX ) is a local martingale. ∈ t ii) Prove that for a<0,~~

~~P [Ta < ] = lim P [Ta~~

~~Why is it not as easy as for Brownian motion to compute P [Ta~~

~~Burkholder’s inequality~~

As another application of Itô’s formula, we prove an important inequality for càdlàg local martingales that is used frequently to derive Lp estimates for semimartingales. For real-valued càdlàg functions x =(xt)t 0 we set ≥

~~⋆ ⋆ xt := sup xs for t>0, and x0 := x0 . s~~

~~Theorem 5.24 (Burkholder’s inequality). Let p [2, ).Thentheestimate ∈ ∞~~

E[(M ⋆ )p]1/p γ E[[M]p/2]1/p (5.55) T ≤ p T holds for any strict local martingale M such that M =0,andforanystopping ∈Mloc 0 time T :Ω [0, ],where → ∞ (p 1)/2 1 − √ γp = 1+ p 1 p/ 2 e/2 p. − ≤ 6 7 D Remark. The estimate does not depend on the underlying ﬁltered probability space, the local martingale M,andthestoppingtimeT .However,theconstantγ goes to p ∞ as p . →∞

Notice that for p =2,Equation(5.55)holdswithγp =2by Itô’s isometry and Doob’s L2 maximal inequality. Burkholder’s inequality can thus be used to generalize arguments based on Itô’s isometry from an L2 to an Lp setting.

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 218 JUMPS~~

Proof. 1) We ﬁrst assume that T = and M is a bounded càdlàg martingale. Then, ∞ by the Martingale Convergence Theorem, M = lim Mt exists almost surely. Since ∞ t p 2 →∞ p 2 the function f(x)= x is C for p 2 with ϕ′′(x)=p(p 1) x − ,Itô’sformula | | ≥ − | | implies

~~p ∞ 1 ∞ c M = ϕ′(Ms ) dMs + ϕ′′(Ms) d[M]s | ∞| ˆ0 − 2 ˆ0~~

+ (ϕ(Ms) ϕ(Ms ) ϕ′(Ms )∆Ms, ) , (5.56) − − − − s " where the ﬁrst term is a martingale since ϕ′ M is bounded, in the second term ◦ ⋆ p 2 ϕ′′(Ms) p(p 1)(M ) − , ≤ − ∞ and the summand in the third term can be estimated by

1 2 ϕ(Ms) ϕ(Ms ) ϕ′(Ms )∆Ms sup(ϕ′′ M)(∆Ms) − − − − ≤ 2 ◦ 1 ⋆ p 2 2 p(p 1)(M ) − (∆Ms) . ≤ 2 − ∞ Hence by taking expectation values on both sides of (5.56), weobtainforq satisfying 1 1 p + q =1:

E[(M ⋆ )p] qp E[ M p] ∞ ≤ | ∞| p p(p 1) ⋆ p 2 c 2 q − E (M ) − [M] + (∆M) ≤ 2 ∞ ∞ p(p 1) / p−62 p "2 70 qp − E[(M ⋆ )p] p E[[M] 2 ] p ≤ 2 ∞ ∞ by Doob’s inequality, Hölder’s inequality, and since [M]c + (∆M)2 =[M] .The ∞ ∞ p p 1 2 inequality (5.55) now follows by noting that q p(p 1) = q − p . − ! 2) For T = and a strict local martingale M M ,thereexistsanincreasing ∞ ∈ loc Tn sequence (Tn) of stopping times such that M is a bounded martingale for each n. Applying Burkholder’s inequality to M Tn yields

⋆ p Tn,⋆ p p Tn p/2 p p/2 E[(MTn ) ]=E[(M ) ] γp E[[M ] ]=γp E[[M]Tn ]. ∞ ≤ ∞ Burkholder’s inequality for M now follows as n . →∞ 3) Finally, the inequality for an arbitrary stopping time T can be derived from that for T = by considering the stopped process M T . ∞ Stochastic Analysis Andreas Eberle 5.5. STOCHASTIC EXPONENTIALS AND CHANGE OF MEASURE 219

~~For p 4,theconverseestimatein(3.1)canbederivedinasimilarway: ≥ Exercise. Prove that for a given p [4, ),thereexistsaglobalconstantc (1, ) ∈ ∞ p ∈ ∞ such that the inequalities~~

~~1 p/2 p p/2 cp− E [M] E [(M ∗ ) ] cp E [M] ∞ ≤ ∞ ≤ ∞ # ' # ' with Mt∗ =sups~~

~~Exercise. Let M be a continuous local martingale satisfying M0 =0.Showthat x2 P sup Ms x ;[M]t c exp s t ≥ ≤ ≤ − 2c / ≤ 0 6 7 for any c, t, x [0, ). ∈ ∞~~

~~5.5 Stochastic exponentials and change of measure~~

Achangeoftheunderlyingprobabilitymeasurebyanexponential martingale can also be carried out for jump processes. In this section, we ﬁrst introduce exponentials of general semimartingales. After considering absolutely continuous measure transformations for Poisson point processes, we apply the results to Lévy processes, and we prove a general change of measure result for possibly discontinuous semimartingales. Finally, we provide a counterpart to Lévy’s characterization of Brownian motion for general Lévy processes.

~~Exponentials of semimartingales~~

~~If X is a continuous semimartingale then by Itô’s formula, 1 X =expX [X] Et t − 2 t 6 7 is the unique solution of the exponential equation~~

~~d X = X dX, X =1. E E E0 University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 220 JUMPS~~

In particular, X is a local martingale if X is a local martingale. Moreover, if E n ∂ 2 hn(t, x)= exp(αx α t/2) (5.57) ∂αn − α=0 & denotes the Hermite polynomial of order n and X =0then & 0 & n Ht = hn [X]t,Xt (5.58) solves the SDE $ % n n 1 n dH = nH − dX, H0 =0, for any n N,cf.Section6.4in[14].Inparticular,Hn is an iterated Itô integral: ∈ t sn s2 n Ht = n! dXs1 dXs2 dXsn . ˆ0 ˆ0 ··· ˆ0 ··· The formula for the stochastic exponential can be generalized to the discontinuous case:

Theorem 5.25 (Doléans-Dade). Let X .Thentheuniquesolutionoftheexponen- ∈S tial equation t Zt =1+ Zs dXs,t0, (5.59) ˆ0 − ≥ is given by 1 Z =expX [X]c (1 + ∆X )exp( ∆X ). (5.60) t t − 2 t s − s s (0,t] 6 7 ∈( Remarks. 1) In the ﬁnite variation case, (5.60) can be written as 1 Z =expXc [X]c (1 + ∆X ). t t − 2 t s s (0,t] 6 7 ∈( In general, however, neither Xc nor (1 + ∆X) exist. 2) The analogues to the stochasticG polynomials Hn in the discontinuous case do not have an equally simply expression as in (5.58) . This is not toosurprising:Alsofor continuous two-dimensional semimartingales (Xt,Yt) there is no direct expression for the iterated integral t s dX dY = t(X X ) dY and for the Lévy area process 0 0 r s 0 s − 0 s ´ ´ t s ´ t s At = dXr dYs dYr dXs ˆ0 ˆ0 − ˆ0 ˆ0 in terms of X,Y and their covariations. If X is a one-dimensional discontinuous semimartingale then X and X are different processes that have both to be taken into account − when computing iterated integrals of X.

~~Stochastic Analysis Andreas Eberle 5.5. STOCHASTIC EXPONENTIALS AND CHANGE OF MEASURE 221~~

~~Proof of Theorem 5.25. The proof is partially similar to the one given above for X ∈ FV,cf.Theorem5.5.Thekeyobservationisthattheproduct~~

~~P = (1 + ∆X )exp( ∆X ) t s − s s (0,t] ∈( exists and deﬁnes a ﬁnite variation pure jump process. This follows from the estimate~~

~~log(1 + ∆X ) ∆X const. ∆X 2 const. [X] | s − s|≤ · | s| ≤ · t 0~~

S = (log(1 + ∆X ) ∆X ),t0, t s − s ≥ s t ∆X"s≤ 1/2 | |≤ defines almost surely a finite variation pure jump process. Therefore, (Pt) is also a finite variation pure jump process. Moreover, the process G =exp X 1 [X]c satisfies t t − 2 t 6 7 G =1+G dX + (∆G G ∆X) (5.61) ˆ − − − " by Itô’s formula. For Z = GP we obtain

~~∆X ∆X ∆Z = Z e (1 + ∆X)e− 1 = Z ∆X, − − − 6 7 and hence, by integration by parts and (5.61),~~

Z 1= P dG + G dP +[G, P ] − ˆ − ˆ − = P G dX + (P ∆G P G ∆X + G ∆P +∆G ∆P ) ˆ − − − − − − − " = Z dX + (∆Z Z ∆X)= Z dX. ˆ − − − ˆ − " This proves that Z solves the SDE (5.59). Uniqueness of the solution follows from a general uniqueness result for SDE with Lipschitz continuouscoefﬁcients,cf.Section 3.1.

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 222 JUMPS~~

Example (Geometric Lévy processes). Consider a Lévy martingale Xt = y Nt(dy) ´ where (Nt) is a Poisson point process on R with intensity measure ν satisfying ( y )| |∧ y 2) ν(dy) < ,andN = N tν.WederiveanSDEforthesemimartingale´ | | ∞ t t − )Z =exp(σX + µt),t0, t t ≥ where σ and µ are real constants. Since [X]c 0,Itô’sformulayields ≡

σ∆X Zt 1=σ Z dX + µ Z ds + Z e 1 σ∆X (5.62) − ˆ − ˆ − − − − (0,t] (0,t] "(0,t] 6 7 σy = σ Zs y N(ds dy)+µ Zs ds + Zs e 1 σy N(ds dy). ˆ − ˆ − ˆ(0,t] R − − − (0,t] R (0,t] × 6 7 × ) If e2σy ν(dy) < then (5.62) leads to the semimartingale decomposition ´ ∞ σ dZt = Zt dMt + αZt dt, Z0 =1, (5.63) − − where M σ = eσy 1 N (dy) t ˆ − t 6 7 is a square-integrable martingale, and )

~~α = µ + (eσy 1 σy) ν(dy). ˆ − −~~

~~In particular, we see that although (Zt) again solves an SDE driven by the compensated process (Nt),thisSDEcannotbewrittenasanSDEdrivenbytheLévyprocess (Xt).~~

~~) Change of measure for Poisson point processes~~

Let (Nt)t 0 be a Poisson point process on a σ-finite measure space (S, ,ν) that is de- ≥ S fined and adapted on a filtered probabilityspace (Ω, ,Q,( )).Supposethat(ω,t,y) A Ft #→ H (y)(ω) is a predictable process in 2 (Q λ ν).Ourgoalistochangetheunder- t Lloc ⊗ ⊗ lying measure Q to a new measure P such that w.r.t. P , (Nt)t 0 is a point process with ≥ intensity of points in the infinitesimal time interval [t, t + dt] given by

~~(1 + Ht(y)) dt ν(dy).~~

~~Stochastic Analysis Andreas Eberle 5.5. STOCHASTIC EXPONENTIALS AND CHANGE OF MEASURE 223~~

Note that in general, this intensity may depend on ω in a predictable way. Therefore, under the new probability measure P ,theprocess(Nt) is not necessarily a Poisson point process. We deﬁne a local exponential martingale by

~~L Zt := t where Lt := (H N)t. (5.64) E • Lemma 5.26. Suppose that inf H> 1,andletG := log (1 + H)).Thenfort 0, − ≥~~

~~L t =exp Gs(y) N(ds dy) (Hs(y) Gs(y)) ds ν(dy) . E ˆ(0,t] S − ˆ(0,t] S − 6 × × 7 Proof. The assumption inf H> 1)implies inf ∆L> 1.Since,moreover,[L]c =0, − − we obtain~~

~~L L [L]c/2 ∆L = e − (1 + ∆L)e− E =expL +( (log(1 + ∆L) ∆L) − 6 " 7 =expG N + (G H) ds ν(dy) . • ˆ − 6 7 Here we have used that )~~

~~(log(1 + ∆L) ∆L)= log (1 + H (y)) H (y) N(ds dy) − ˆ s − s " $ % holds, since log(1 + H (y)) H (y) const. H (y) 2 is integrable on ﬁnite time | s − s |≤ | s | intervals.~~

~~The exponential Z = L is a strictly positive local martingale w.r.t. Q,andhencea t Et supermartingale. As usual, we ﬁx t R ,andweassume: 0 ∈ +~~

Assumption. (Zt)t t0 is a martingale w.r.t. Q,i.e.EQ[Zt0 ]=1. ≤ Then there is a probability measure P on such that Ft0 dP = Zt for any t t0. dQ t ≤ &F & In the deterministic case Ht(&y)(ω)=h(y),wecanprovethatw.r.t.P , (Nt) is a Poisson point process with changed intensity measure

~~µ(dy)=(1+h(y)) ν(dy):~~

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 224 JUMPS~~

Theorem 5.27 (Change of measure for Poisson point processes). Let (Nt,Q) be a Poisson point process with intensity measure ν,andletg := log (1+h) where h 2(ν) ∈L satisﬁes inf h> 1.Supposethattheexponential −

" Z = N(h) =expN (g)+t (g h) dν (5.65) t Et t ˆ − 6 7 ) dP is a martingale w.r.t. Q,andassumethatP Q on t with relative density dQ = Zt ≪ F t for any t 0.Thentheprocess(N ,P) is a Poisson point process with& intensityF ≥ t & measure &

~~dµ =(1+h) dν.~~

~~Proof. By the Lévy characterization for Poisson point processes (cf. the exercise below Lemma 2.1) it sufﬁces to show that the process~~

M [f] =expiN (f)+tψ(f) ,ψ(f)= 1 eif dµ, t t ˆ − $ % $ % is a local martingale w.r.t. P for any elementary function f 1(S, ,ν).Further- ∈L S more, by Lemma 2.9, M [f] is a local martingale w.r.t. P if and only if M [f]Z is a local martingale w.r.t. Q.Thelocalmartingalepropertyfor(M [f]Z, Q) can be veriﬁed by a computation based on Itô’s formula.

Remark (Extension to general measure transformations). The approach in Theo- rem 5.27 can be extended to the case where the function h(y) is replaced by a general predictable process Ht(y)(ω).Inthatcase,oneveriﬁessimilarlythatunderanewmea- sure P with local densities given by (5.64), the process

M [f] =expiN (f)+ (1 eif(y))(1 + H (y)) dy t t ˆ − t 6 7 is a local martingale for any elementary function f 1(ν).Thispropertycanbeused ∈L as a deﬁnition of a point process with predictable intensity (1 + Ht(y)) dt ν(dy).

~~Stochastic Analysis Andreas Eberle 5.5. STOCHASTIC EXPONENTIALS AND CHANGE OF MEASURE 225~~

~~Change of measure for Lévy processes~~

Since Lévy processes can be constructed from Poisson point processes, a change of measure for Poisson point processes induces a correspondingtransformationforLévy processes. Suppose that ν is a σ-ﬁnite measure on Rd 0 such that \{ } ( y y 2) ν(dy) < , and let ˆ | |∧| | ∞ µ(dy)=(1+h(y)) ν(dy).

~~Recall that if (Nt,Q) is a Poisson point process with intensity measure ν,then~~

X = y N (dy), N = N tν, t ˆ t t t − is a Lévy martingale with Lévy measure) ν w.r.t. Q). Corollary 5.28 (Girsanov transformation for Lévy processes). Suppose that h ∈ 2(ν) satisﬁes inf h> 1 and sup h< .IfP Q on with relative density Z L − ∞ ≪ Ft t for any t 0,whereZ is given by (5.65),thentheprocess ≥ t X = y N (dy), N = N tµ, t ˆ t t t − is a Lévy martingale with Lévy measure µ w.r.t. P ,and

X = X + t yh(y) ν(dy). t t ˆ Notice that the effect of the measure transformation consists of both the addition of a drift and a change of the intensity measure of the Lévy martingale. This is different to the case of Brownian motion where only a drift is added.

~~Example (Change of measure for compound Poisson processes). Suppose that (X, Q) is a compound Poisson process with ﬁnite jump intensity measure ν,andlet~~

~~h Nt = h(∆Xs) s t "≤ with h as above. Then (X, P) is a compound Poisson process with jump intensity measure dµ =(1+h) dν provided~~

~~" dP N(h) t hdν = t = e− ´ (1 + h(∆Xs)). dQ t E F s t & ≤ & ( & University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 226 JUMPS~~

~~Lévy’s characterization for Brownian motion has an extension to Lévy processes, too:~~

d d Theorem 5.29 (Lévy characterization of Lévy processes). Let a R × , b R, ∈ ∈ and let ν be a σ-ﬁnite measure on Rd 0 satisfying ( y y 2) ν(dy) < .If \{ } | |∧| | ∞ X1,...,Xd :Ω R are càdlàg stochastic processes such´ that t t → (i) M k := Xk bkt is a local ( ) martingale for any k 1,...,d , t t − Ft ∈{ } (ii) [Xk,Xl]c = akl t for any k, l 1,...,d , and t ∈{ }

(iii) E s (r,t] IB(∆Xs) r = ν(B) (t r) almost surely ∈ F · − for/ any 0 r t and& for0 any B (Rd 0 ), ! & ≤ ≤ & ∈B \{ } 1 d then Xt =(Xt ,...,Xt ) is a Lévy process with characteristic exponent

~~1 ip y ψ(p)= p ap ip b + (1 e · + ip y) ν(dy). (5.66) 2 · − · ˆ − ·~~

For proving the theorem, we assume without proof that a local martingale is a semimartingale even if it is not strict, and that the stochastic integral of a bounded adapted left-continuous integrand w.r.t. a local martingale is again a local martingale, cf. [36].

~~Proof of Theorem 5.29. We ﬁrst remark that (iii) implies~~

t E Gs f(∆Xs) r = E Gs f(y) ν(dy) ds r , a.s. for 0 r t · F ˆr ˆ · F ≤ ≤ / s (r,t] & 0 / & 0 "∈ & & (5.67) & & for any bounded left-continuous adapted process G,andforanymeasurablefunction f : Rd 0 C satisfying f(y) const. ( y y 2).Thiscanbeveriﬁed \{ }→ | |≤ · | |∧| | by ﬁrst considering elementary functions of type f(y)= ci IBi (y) and Gs(ω)= A (ω) I (s) with c R, B (Rd 0 ), 0 t

k l c k l c kl Noting that [M ,M ]t =[X ,X ]t = a t by (ii), Itô’s formula yields t t 1 kl Zt =1+ Z ip dM + Z (ψ(p)+ip b pkpla ) dt (5.68) ˆ0 − · ˆ0 − · − 2 "k,l ip ∆X + Z e · 1 ip ∆X . − − − · "(0,t] 6 7 ip y 2 By (5.67) and since e · 1 ip y const. ( y y ),theseriesontherighthand | − − · |≤ · | |∧| | side of (5.68) can be decomposed into a martingale and the ﬁnite variation process

~~t ip y At = Zs (e · 1 ip y) ν(dy) ds ˆ0 ˆ − − − · Therefore, by (5.68) and (5.66), (Z ) is a martingale for any p Rd.Theassertionnow t ∈ follows by Lemma 2.1.~~

~~An interesting consequence of Theorem 5.29 is that a BrownianmotionB and a Lévy process without diffusion part w.r.t. the same ﬁltration arealwaysindependent,because [Bk,Xl]=0for any k, l.~~

Exercise (Independence of Brownian motion and Lévy processes). Suppose that ′ B :Ω Rd and X :Ω Rd are a Brownian motion and a Lévy process with t → t → ip y characteristic exponent ψ (p)= ip b + (1 e · + ip y) ν(dy) deﬁned on the X − · − · same ﬁltered probability space (Ω, ,P,( ))´.Assumingthat ( y y 2) ν(dy) < , A Ft | |∧| | ∞ d d′ ´ prove that (Bt,Xt) is a Lévy process on R × with characteristic exponent

~~1 2 d d′ ψ(p, q)= p d + ψ (q),pR ,q R . 2| |R X ∈ ∈ Hence conclude that B and X are independent.~~

~~Change of measure for general semimartingales~~

~~We conclude this section with a general change of measure theorem for possibly discontinuous semimartingales:~~

Theorem 5.30 (P.A. Meyer). Suppose that the probability measures P and Q are equiv- dP alent on t for any t 0 with relative density dQ = Zt.IfM is a local martingale F ≥ t w.r.t. Q then M 1 d[Z, M] is a local martingale&F w.r.t. P . Z & − ´ & University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 228 JUMPS

The theorem shows that w.r.t. P , (Mt) is again a semimartingale, and it yields an explicit semimartingale decomposition for (M,P).Fortheproofwerecallthat(Zt) is a local martingale w.r.t. Q and (1/Zt) is a local martingale w.r.t. P .

Proof. The process ZM [Z, M] is a local martingale w.r.t. Q.HencebyLemmy2.9, − the process M 1 [Z, M] is a local martingale w.r.t. P .Itremainstoshowthat 1 [Z, M] − Z Z 1 differs from Z d[Z, M] by a local P -martingale. This is a consequence of the Itô product rule:´ Indeed,

1 1 1 1 [Z, M]= [Z, M] d + d[Z, M]+ , [Z, M] . Z ˆ − Z ˆ Z Z − # ' The ﬁrst term on theright-handside isa local Q-martingale, since 1/Z is a Q-martingale. 1 The remaining two terms add up to Z d[Z, M],because ´ 1 1 , [Z, M] = ∆ ∆[Z, M]. Z Z # ' "

~~1 Remark. Note that the process Z d[Z, M] is not predictable in general. For a predictable counterpart to Theorem 5.30´ cf. e.g. [36].~~

~~5.6 General predictable integrands~~

So far, we have considered stochastic integrals w.r.t. general semimartingales only for integrands that are left limits of adapted càdlàg processes.Thisisindeedsufﬁcient for many applications. For some results including in particular convergence theorems for stochastic integrals, martingale representation theorems and the existence of local time, stochastic integrals with more general integrands areimportant.Inthissection, we sketch the deﬁnition of stochastic integrals w.r.t. not necessarily continuous semimartingales for general predictable integrands. For details of the proofs, we refer to Chapter IV in Protter’s book [36].

Throughout this section, we ﬁx a ﬁltered probability space (Ω, ,P,( )).Recallthat A Ft the predictable σ-algebra on Ω (0, ) is generated by all sets A (s, t] with P × ∞ × Stochastic Analysis Andreas Eberle 5.6. GENERAL PREDICTABLE INTEGRANDS 229

A and 0 s t,or,equivalently,byallleft-continuous( ) adapted processes ∈Fs ≤ ≤ Ft (ω,t) G (ω).Wedenoteby the vector space consisting of all elementary pre- #→ t E dictable processes G of the form n 1 − Gt(ω)= Zi(ω)I(ti,ti+1](t) i=0 " with n N, 0 t

P[M](dωdt)=P (dω)[M](ω)(dt) is the Doléans measure of the martingale M on Ω R .TheItôisometryhasbeen × + derived in a more general form in Corollary 5.20, but for elementary processes it can easily be veriﬁed directly (Excercise!). In many textbooks, the angle bracket process M is used instead of [M] to deﬁne ⟨ ⟩ stochastic integrals. The next remark shows that this is equivalent for predictable integrands: Remark ([M] vs. ⟨M⟩). Let M M 2(0, ).Iftheangle-bracketprocess M exists ∈ d ∞ ⟨ ⟩ then the measures P[M] and P M coincide on predictable sets. Indeed, if C = A (s, t] ⟨ ⟩ × with A and 0 s t then ∈Fs ≤ ≤ P (C)=E [[M] [M] ; A]=E [E[[M] [M] ]; A] [M] t − s t − s|Fs = E [E[ M t M s s]; A]=P M (C). ⟨ ⟩ −⟨ ⟩ |F ⟨ ⟩

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 230 JUMPS~~

Since the collection of these sets C is an -stable generator for the predictable σ- ∩ algebra, the measures P[M] and P M coincide on . ⟨ ⟩ P Example (Doléans measures of Lévy martingales). If M = X E[X ] with a square t t − t integrable Lévy process X :Ω R then t →

P[M] = P M = ψ′′(0) P λ(0, ) ⟨ ⟩ ⊗ ∞ where ψ is the characteristic exponent of X and λ(0, ) denotes Lebesgue measure on ∞ R+.HencetheDoléansmeasureofageneralLévymartingalecoincides with the one for Brownian motion up to a multiplicative constant.

~~Deﬁnition of stochastic integrals w.r.t. semimartingales~~

We denote by 2 the vector space of all semimartingales vanishing at 0 of the form H X = M + A with M M 2(0, ) and A FV predictable with total variation ∈ d ∞ ∈ (1) 2 2 V (A)= ∞ dAs L (P ).Inordertodeﬁneanormonthespace ,weas- ∞ 0 | |∈ H sume without´ proof the following result, cf. e.g. Chapter IIIinProtter[36]: Fact. Any predictable local martingale with ﬁnite variation pathsisalmostsurelycon- stant. The result implies that the Doob-Meyer semimartingale decomposition

X = M + A (5.70) is unique if we assume that M is local martingale and A is a predictable ﬁnite variation process vanishing at 0.Therefore,weobtainawell-deﬁned norm on 2 by setting H 2 2 2 (1) 2 ∞ X 2 = M M 2 + V (A) L2 = E [M] + dA . || ||H || || || ∞ || ∞ ˆ | | 8 - 0 . ; Note that the M 2 norm is the restriction of the 2 norm to the subspace M 2(0, ) H ∞ ⊂ 2.Asaconsequenceof(5.69),weobtain: H Corollary 5.31 (Itô isometry for semimartingales). Let X 2 with semimartingale ∈H decomposition as above. Then

~~G X 2 = G X for any G , where || • ||H || || ∈E Stochastic Analysis Andreas Eberle 5.6. GENERAL PREDICTABLE INTEGRANDS 231~~

2 2 2 ∞ G := G 2 + G dA . X L (P[M]) 2 || || || || ˆ0 | || | L (P ) 2 && && Hence the stochastic integral : , &X&(G)=G X,hasauniqueisometric&& X J E→H J&& • && extension to the closure of w.r.t. the norm in the space of all predictable E E || · ||X 2 processes in L (P[M]). Proof. The semimartingale decomposition X = M + A implies a corresponding decomposition G X = G M + G A for the stochastic integrals. One can verify that • • • 2 for G , G M is in Md (0, ) and G A is a predictable finite variation process. ∈E • ∞ • Therefore, and by (5.69), 2 2 2 (1) 2 2 G X 2 = G M M 2 + V (G A) L2 = G L2(P ) + G dA . || • ||H || • || || ∞ • || || || [M] ˆ | || | L2(P ) && && && && && && X The Itô isometry yields a definition of the stochastic integral G X for G .For • ∈ E G = H with H càdlàg and adapted, this definition is consistent with the definition − given above since, by Corollary 5.20, the Itô isometry also holds for the integrals defined X above, and the isometric extension is unique. The class of admissible integrands is E already quite large: X Lemma 5.32. contains all predictable processes G with G < . E || ||X ∞ Proof. We only mention the main steps of the proof, cf. [36] for details: 1) The approximation of bounded left-continuous processes by elementary predictable processes w.r.t. is straightforward by dominated convergence. || · ||X 2) The approximability of bounded predictable processes by bounded left-continuous processes w.r.t. can be shown via the Monotone Class Theorem. || · ||X n 3) For unbounded predictable G with G X < ,theprocessesG := G I G n , || || ∞ · { ≤ } n N,arepredictableandboundedwith Gn G 0. ∈ || − ||X →

~~Localization~~

2 Having deﬁned G X for X and predictable integrands G with G X < ,the • ∈H || || ∞ next step is again a localization. This localization is slightly different than before, because there might be unbounded jumps at the localizing stopping times. To overcome

University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 232 JUMPS this difﬁculty, the process is stopped just before the stopping time T ,i.e.,atT .How- − ever, stopping at T destroys the martingale property if T is not a predictable stopping − time. Therefore, it is essential that we localize semimartingales instead of martingales!

~~T For a semimartingale X and a stopping time T we deﬁne the stopped process X − by~~

Xt for t0, t ⎪ T ⎪ − ≥ ⎨0 for T =0. ⎪ The deﬁnition for T =0is of course rather⎩⎪ arbitrary. It will not be relevant below, since we are considering sequences (T ) of stopping times with T almost surely. We n n ↑∞ state the following result from Chapter IV in [36] without proof.

~~Fact. If X is a semimartingale with X0 =0then there exists an increasing sequence~~

Tn 2 (T ) of stopping times with sup T = such that X − for any n N. n n ∞ ∈H ∈ Now we are ready to state the deﬁnition of stochastic integrals for general predictable integrands w.r.t. general semimartingales X.BysettingG X = G (X X0) we may • • − assume X0 =0.

~~Deﬁnition. Let X be a semimartingale with X0 =0.ApredictableprocessG is called integrable w.r.t. X iff there exists an increasing sequence (Tn) of stopping times such~~

Tn 2 that sup T = a.s., and for any n N, X − and G Tn− < . n ∞ ∈ ∈H || ||X ∞ If G is integrable w.r.t. X then the stochastic integral G X is deﬁned by • t t Tn N (G X)t = Gs dXs = Gs dXs − for any t [0,Tn),n . • ˆ0 ˆ0 ∈ ∈ Of course, one has to verify that G X is well-deﬁned. This requires in particular a • locality property for the stochastic integrals that are usedinthelocalization.Wedonot carry out the details here, but refer once more to Chapter IV in[36].

Exercise (Sufﬁcient conditions for integrability of predictable processes). 1) Prove that if G is predictable and locally bounded in the sense that GTn is bounded for a sequence (T ) of stopping times with T ,thenG is integrable w.r.t. any n n ↑∞ semimartingale X . ∈S Stochastic Analysis Andreas Eberle 5.6. GENERAL PREDICTABLE INTEGRANDS 233

2) Suppose that X = M + A is a continuous semimartingale with M loc and ∈Mc A FV .ProvethatG is integrable w.r.t. X if G is predictable and ∈ c t t 2 Gs d[M]s + Gs dAs < a.s. for any t 0. ˆ0 ˆ0 | || | ∞ ≥

~~Properties of the stochastic integral~~

Most of the properties of stochastic integrals can be extended easily to general predictable integrands by approximation with elementary processes and localization. The proof of Property (2) below, however, is not trivial. We refertoChapterIVin[36]for detailed proofs of the following basic properties:

~~(1) The map (G, X) G X is bilinear. #→ • (2) ∆(G X)=G∆X almost surely. •~~

~~T T (3) (G X) =(GI[0,T ]) X = G X . • • •~~

T T (4) (G X) − = G X −. • • (5) G (G X)=(GG) X. • • • In all statements,) X is) a semimartingale, G is a process that is integrable w.r.t. X, T is a stopping time, and G is a process such that GG is also integrable w.r.t. X.Westatethe formula for the covariation of stochastic integrals separately below, because its proof is ) ) based on the Kunita-Watanabe inequality, which is of independent interest.

Exercise (Kunita-Watanabe inequality). Let X, Y ,andletG, H be measurable ∈S processes deﬁned on Ω (0, ) (predictability is not required). Prove that for any × ∞ a [0, ] and p, q [1, ] with 1 + 1 =1,thefollowinginequalitieshold: ∈ ∞ ∈ ∞ p q a a 1/2 a 1/2 G H d[X, Y ] G2 d[X] H2 d[Y ] , (5.71) ˆ0 | || || |≤ ˆ0 ˆ0 6 7 6 7 a a 1/2 a 1/2 E G H d[X, Y ] G2 d[X] H2 d[Y ] . p q ˆ0 | || || | ≤ ˆ0 L ˆ0 L / 0 &&6 7 && &&6 7 && && && && &(5.72)& Hint: First consider elementary processes&& G, H. && && &&

~~University of Bonn Summer Semester 2015 CHAPTER 5. STOCHASTIC CALCULUS FOR SEMIMARTINGALES WITH 234 JUMPS~~

~~Theorem 5.33 (Covariation of stochastic integrals). For any X, Y and any pre- ∈S dictable process G that is integrable w.r.t. X,~~

GdX,Y = Gd[X, Y ] almost surely. (5.73) ˆ ˆ / 0 Remark. If X and Y are local martingales, and the angle-bracket process X, Y exists, ⟨ ⟩ then also GdX,Y = Gd X, Y almost surely. ˆ ˆ ⟨ ⟩ N O Proof of Theorem 5.33. Weonlysketch themainstepsbrieﬂy, cf.[36]fordetails. Firstly, one veriﬁes directly that (5.73) holds for X, Y 2 and G .Secondly,for ∈H ∈E X, Y 2 and a predictable process G with G < there exists a sequence ∈H || ||X ∞ (Gn) of elementary predictable processes such that Gn G 0,and || − ||X →

Gn dX, Y = Gn d[X, Y ] for any n N. ˆ ˆ ∈ / 0 As n , Gn dX GdXin 2 by the Itô isometry for semimartingales, and →∞ → H hence ´ ´ Gn dX, Y GdX,Y u.c.p. ˆ −→ ˆ / 0 / 0 by Corollary 5.14. Moreover,

Gn d[X, Y ] Gd[X, Y ] u.c.p. ˆ −→ ˆ by the Kunita-Watanabe inequality. Hence (5.73) holds for G as well. Finally, by localization, the identity can be extended to general semimartingales X, Y and integrands G that are integrable w.r.t. X.

~~An important motivation for the extension of stochastic integrals to general predictable integrands is the validity of a Dominated Convergence Theorem:~~

Theorem 5.34 (Dominated Convergence Theorem for stochastic integrals). Suppose that X is a semimartingale with decomposition X = M + A as above, and let Gn, n N,andG be predictable processes. If ∈ Gn(ω) G (ω) for any t 0, almost surely, t −→ t ≥ Stochastic Analysis Andreas Eberle 5.6. GENERAL PREDICTABLE INTEGRANDS 235 and if there exists a process H that is integrable w.r.t. X such that Gn H for any | |≤ n N,then ∈ Gn X G X u.c.p. as n . • −→ • →∞ If, in addition to the assumptions above, X is in 2 and H < then even H || ||X ∞ n G X G X 2 0 as n . || • − • ||H −→ →∞ Proof. We may assume G =0,otherwiseweconsiderGn G instead of Gn.Now − suppose ﬁrst that X is in 2 and H < .Then H || ||X ∞ 2 n 2 ∞ n 2 ∞ n G X = E G d[M]+ G dA 0 || || ˆ0 | | ˆ0 | || | −→ / 6 7 0 as n by the Dominated Convergence Theorem for Lebesgue integrals. Hence by →∞ the Itô isometry, Gn X 0 in 2 as n . • −→ H →∞ The general case can now be reduced to this case by localization, where 2 convergence H is replaced by the weaker ucp-convergence.

We ﬁnally remark that basic properties of stochastic integrals carry over to integrals with respect to compensated Poisson point processes. We refer to the monographs by D.Applebaum [5] for basics, and to Jacod & Shiryaev [24] for a detailed study. We only state the following extension of the associative law, which has already been used in the last section:

Exercise (Integration w.r.t. stochastic integrals based on compensated PPP). Sup- pose that H :Ω R S R is predictable and square-integrable w.r.t. P λ ν, × + × → ⊗ ⊗ and G :Ω R R is a bounded predictable process. Show that if × + →

~~Xt = Hs(y) N(ds dy) ˆ(0,t] S × then ) t Gs dXs = GsHs(y) N(ds dy). ˆ0 ˆ(0,t] S × Hint: Approximate G by elementary processes. )~~

~~University of Bonn Summer Semester 2015 Bibliography~~

~~[1]~~

~~[2]~~

~~[3] Sur l’équation de Kolmogoroff, par W. Doeblin.ÉditionsElsevier,Paris,2000.C. R. Acad. Sci. Paris Sér. I Math. 331 (2000), Special Issue.~~

~~[4] Robert J. Adler and Jonathan E. Taylor. Random ﬁelds and geometry.Springer Monographs in Mathematics. Springer, New York, 2007.~~

~~[5] David Applebaum. Lévy Processes and Stochastic Calculus.CambridgeUP,Cam- bridge, 2004.~~

~~[6] Ole E. Barndorff-Nielsen. Processes of normal inverse Gaussian type. Finance Stoch.,2(1):41–68,1998.~~

~~[7] Richard F. Bass. Stochastic Processes.CambridgeUniversityPress,Cambridge, 2011.~~

~~[8] Jean Bertoin. Lévy Processes.CambridgeUP,Cambridge,1998.~~

~~[9] Jean-Michel Bismut. Large deviations and the Malliavin calculus.Birkhaeuser, Boston, 1994.~~

~~[10] Amir Dembo and Ofur Zeitouni. Large Deviations Techniques and Applications. Springer, 1998, Berlin.~~

~~[11] Frank den Hollander. Large Deviations.AmericanMathematicalSociety,Provi- dence, 2008.~~

~~236 BIBLIOGRAPHY 237~~

~~[12] Andreas Eberle. Markov processes. http://www.uni-bonn.de/~eberle/MarkovProcesses/MPSkript.pdf, 2009.~~

~~[13] Andreas Eberle. Wahrscheinlichkeitstheorie und Stochastische Prozesse. http://www.uni-bonn.de/~eberle/StoPr/WthSkriptneu.pdf, 2010.~~

~~[14] Andreas Eberle. Introduction to Stochastic Analysis. http://www.uni-bonn.de/~eberle/StoAn1011/StoAnSkriptneu.pdf, 2011.~~

~~[15] David Elworthy. On the Geometry of Diffusion Operators and Stochastic Flows. Springer, 1999, Berlin.~~

~~[16] Stewart N. Ethier and Thomas G. Kurtz. Markov Processes.Wiley-Interscience, Hoboken, NJ, 2005.~~

~~[17] William Feller. Introduction to Probability Theory and Its Applications.Wiley, New York, 1957.~~

~~[18] Peter K. Friz and Martin Hairer. Acourseonroughpaths.Universitext.Springer, Cham, 2014. With an introduction to regularity structures.~~

~~[19] Peter K. Friz and Nicolas Victoir. Multidimensional Stochastic Processes as Rough Paths.CambridgeUP,Cambridge,2010.~~

~~[20] W. Hackenbroch and A. Thalmaier. Stochastic Analysis.Teubner,1991.~~

~~[21] Martin Hairer. On malliavin’s proof of hörmander’s theorem. Bull. Sci. Math., 135(6-7):650–666, 2011.~~

~~[22] Elton Hsu. Stochastic Analysis on Manifolds.OxfordUniversityPress,Oxford, 2002.~~

~~[23] Nobuyuki Ikeda and Takehiko Ed. Watanabe. Stochastic Differential Equations and Diffusion Processes.North-Holland,Amsterdam,1981.~~

~~University of Bonn Summer Semester 2015 238 BIBLIOGRAPHY~~

~~[24] Jean Jacod and Albert Shiryaev. Limit Theorems for Stochastic Processes. Springer, Berlin, 2003.~~

~~[25] Ioannis Karatzas and Steven E. Shreve. Brownian motion and stochastic calculus. Springer, New York, 2010.~~

~~[26] Peter E. Kloeden and Eckhard Platen. Numerical solution of stochastic differential equations,volume23ofApplications of Mathematics (New York).Springer- Verlag, Berlin, 1992.~~

~~[27] Hiroshi Kunita. Stochastic Flows and Stochastic Differential Equations.Cam- bridge University Press, Cambridge, 1997.~~

~~[28] Andreas Kyprianou. Introductory Lectures on Fluctuations of Lévy Processes with Applications.Springer,Berlin,2006.~~

~~[29] Terry J. Lyons. Differential Equations Driven by Rough Paths, Ecole d’Eté de Probabilités de Saint-Flour XXXIV.Springer,2004.~~

~~[30] Zhi-Ming Ma and Michael Röckner. Introduction to the Theory of (Non- Symmetric) Dirichlet Forms.Springer,Berlin,1992.~~

~~[31] Dilip B. Madan and Eugene Seneta. The variance gamma model for share market returns. Journal of Business,63(4):511–524,1990.~~

~~[32] Paul Malliavin. Stochastic Analysis.Springer,Berlin,2002.~~

~~[33] G. N. Milsteinand M. V.Tretyakov. Stochastic numerics for mathematical physics. Scientiﬁc Computation. Springer-Verlag, Berlin, 2004.~~

~~[34] David Nualart. The Malliavin Calculus and Related Topics.Springer,Berlin, 2006.~~

~~[35] Giuseppe Da Prato. Introduction to Stochastic Analysis and Malliavin Calculus. Edizioni della Normale, Pisa, 2009.~~

~~[36] Philip E. Protter. Stochastic Integration and Differential Equations.Springer, Berlin, 2005.~~

~~Stochastic Analysis Andreas Eberle BIBLIOGRAPHY 239~~

~~[37] Daniel Revuz and Marc Yor. Continuous martingales and Brownian motion. Springer, Berlin, 2005.~~

~~[38] L. C. G. Rogers and David Williams. Diffusions, Markov processes and martingales, Vol. 2: Ito calculus.CambridgeUP,Camebridge,2008.~~

~~[39] Steven E. Shreve. Stochastic calculus for ﬁnance. II.SpringerFinance.Springer- Verlag, New York, 2004. Continuous-time models.~~

~~[40] Daniel W. Stroock and S.R. S. Varadhan. Multidimensional Diffusion Processes. Springer, Berlin, 2005.~~

~~[41] Shinzo Watanabe. Lectures on stochastic differential equations and Malliavin calculus.Springer,Berlin,1984.~~

~~University of Bonn Summer Semester 2015~~