The Continuity of SDE With Respect to Initial Value in the Total Variation

Similar documents
Divergence theorems in path space II: degenerate diffusions

Regularity of the density for the stochastic heat equation

On the controllability of conservative systems

Quasi-invariant Measures on Path Space. Denis Bell University of North Florida

Spatial Ergodicity of the Harris Flows

for all f satisfying E[ f(x) ] <.

LAN property for sde s with additive fractional noise and continuous time observation

Divergence Theorems in Path Space. Denis Bell University of North Florida

BOOK REVIEW. Review by Denis Bell. University of North Florida

Citation Osaka Journal of Mathematics. 41(4)

A REPRESENTATION FOR THE KANTOROVICH RUBINSTEIN DISTANCE DEFINED BY THE CAMERON MARTIN NORM OF A GAUSSIAN MEASURE ON A BANACH SPACE

Large Deviations Principles for McKean-Vlasov SDEs, Skeletons, Supports and the law of iterated logarithm

On the Error Term for the Mean Value Associated With Dedekind Zeta-function of a Non-normal Cubic Field

Central limit theorems for ergodic continuous-time Markov chains with applications to single birth processes

Harnack Inequalities and Applications for Stochastic Equations

A Concise Course on Stochastic Partial Differential Equations

Fonctions on bounded variations in Hilbert spaces

STOCHASTIC DIFFERENTIAL EQUATIONS WITH INTERACTION AND THE LAW OF ITERATED LOGARITHM

b i (µ, x, s) ei ϕ(x) µ s (dx) ds (2) i=1

Some Properties of NSFDEs

-variation of the divergence integral w.r.t. fbm with Hurst parameter H < 1 2

A Note on the Central Limit Theorem for a Class of Linear Systems 1

WEYL S LEMMA, ONE OF MANY. Daniel W. Stroock

Stability of Stochastic Differential Equations

1 Brownian Local Time

Independence of some multiple Poisson stochastic integrals with variable-sign kernels

Some SDEs with distributional drift Part I : General calculus. Flandoli, Franco; Russo, Francesco; Wolf, Jochen

L -uniqueness of Schrödinger operators on a Riemannian manifold

Fast-slow systems with chaotic noise

Densities for the Navier Stokes equations with noise

THE MULTIPLICATIVE ERGODIC THEOREM OF OSELEDETS

Feller Processes and Semigroups

THE MALLIAVIN CALCULUS FOR SDE WITH JUMPS AND THE PARTIALLY HYPOELLIPTIC PROBLEM

{σ x >t}p x. (σ x >t)=e at.

Kolmogorov equations in Hilbert spaces IV

Some classical results on stationary distributions of continuous time Markov processes

LARGE TIME BEHAVIOR OF THE RELATIVISTIC VLASOV MAXWELL SYSTEM IN LOW SPACE DIMENSION

Stationary distribution and pathwise estimation of n-species mutualism system with stochastic perturbation

Non-degeneracy of perturbed solutions of semilinear partial differential equations

Fast-slow systems with chaotic noise

Strong Solutions and a Bismut-Elworthy-Li Formula for Mean-F. Formula for Mean-Field SDE s with Irregular Drift

OPTIMAL SOLUTIONS TO STOCHASTIC DIFFERENTIAL INCLUSIONS

Mean-square Stability Analysis of an Extended Euler-Maruyama Method for a System of Stochastic Differential Equations

Potential Theory on Wiener space revisited

ON SECOND ORDER DERIVATIVES OF CONVEX FUNCTIONS ON INFINITE DIMENSIONAL SPACES WITH MEASURES

DIVERGENCE THEOREMS IN PATH SPACE

A NOTE ON THE COMPLETE MOMENT CONVERGENCE FOR ARRAYS OF B-VALUED RANDOM VARIABLES

Functional central limit theorem for super α-stable processes

BV functions in a Gelfand triple and the stochastic reflection problem on a convex set

On the simplest expression of the perturbed Moore Penrose metric generalized inverse

Fixed points of the derivative and k-th power of solutions of complex linear differential equations in the unit disc

Pathwise uniqueness for stochastic differential equations driven by pure jump processes

Verona Course April Lecture 1. Review of probability

Boundedness of solutions to a retarded Liénard equation

Introduction to Infinite Dimensional Stochastic Analysis

Cores for generators of some Markov semigroups

arxiv: v1 [math.pr] 8 Jan 2015

Mean-field SDE driven by a fractional BM. A related stochastic control problem

Ollivier Ricci curvature for general graph Laplacians

Pointwise convergence rates and central limit theorems for kernel density estimators in linear processes

BLOWUP THEORY FOR THE CRITICAL NONLINEAR SCHRÖDINGER EQUATIONS REVISITED

On the martingales obtained by an extension due to Saisho, Tanemura and Yor of Pitman s theorem

Quasi-invariant measures on the path space of a diffusion

Applications of Ito s Formula

Proof. We indicate by α, β (finite or not) the end-points of I and call

THE L 2 -HODGE THEORY AND REPRESENTATION ON R n

Richard F. Bass Krzysztof Burdzy University of Washington

MIXED BOUNDARY-VALUE PROBLEMS FOR QUANTUM HYDRODYNAMIC MODELS WITH SEMICONDUCTORS IN THERMAL EQUILIBRIUM

Stochastic Hamiltonian systems and reduction

SDE Coefficients. March 4, 2008

Convergence of Feller Processes

SYMMETRY RESULTS FOR PERTURBED PROBLEMS AND RELATED QUESTIONS. Massimo Grosi Filomena Pacella S. L. Yadava. 1. Introduction

Introduction to Empirical Processes and Semiparametric Inference Lecture 08: Stochastic Convergence

Lecture 12: Detailed balance and Eigenfunction methods

Applied Mathematics Letters. Stationary distribution, ergodicity and extinction of a stochastic generalized logistic system

Strong uniqueness for stochastic evolution equations with possibly unbounded measurable drift term

ON THE FIRST TIME THAT AN ITO PROCESS HITS A BARRIER

The reflexive and anti-reflexive solutions of the matrix equation A H XB =C

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS

On density asymptotics for diffusions, old and new

Functional Limit theorems for the quadratic variation of a continuous time random walk and for certain stochastic integrals

Regularity for the optimal transportation problem with Euclidean distance squared cost on the embedded sphere

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS

HAIYUN ZHOU, RAVI P. AGARWAL, YEOL JE CHO, AND YONG SOO KIM

Linear Ordinary Differential Equations

Almost Sure Convergence of the General Jamison Weighted Sum of B-Valued Random Variables

Quantum stochastic calculus applied to path spaces over Lie groups

The Equivalence of the Convergence of Four Kinds of Iterations for a Finite Family of Uniformly Asymptotically ø-pseudocontractive Mappings

Uniqueness of Fokker-Planck equations for spin lattice systems (I): compact case

1 Math 241A-B Homework Problem List for F2015 and W2016

Weak convergence and large deviation theory

Obstacle problems and isotonicity

arxiv:math/ v1 [math.dg] 19 Nov 2004

ON THE PATHWISE UNIQUENESS OF SOLUTIONS OF STOCHASTIC DIFFERENTIAL EQUATIONS

Trotter s product formula for projections

BSDEs and PDEs with discontinuous coecients Applications to homogenization K. Bahlali, A. Elouain, E. Pardoux. Jena, March 2009

Brownian Motion. An Undergraduate Introduction to Financial Mathematics. J. Robert Buchanan. J. Robert Buchanan Brownian Motion

Patryk Pagacz. Characterization of strong stability of power-bounded operators. Uniwersytet Jagielloński

Iterative common solutions of fixed point and variational inequality problems

LONG TIME BEHAVIOUR OF PERIODIC STOCHASTIC FLOWS.

Transcription:

Ξ44fflΞ5» ο ffi fi $ Vol.44, No.5 2015 9" ADVANCES IN MATHEMATICS(CHINA) Sep., 2015 doi: 10.11845/sxjz.2014024b The Continuity of SDE With Respect to Initial Value in the Total Variation PENG Xuhui (1. College of Mathematics and Computer Science, Hunan Normal University, Changsha, Hunan, 410081, P. R. China; 2. Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing, 100190, P. R. China;) Abstract: Let X t(x) be the solution of stochastic differential equation driven by Brownian motion, here x is the initial value. If the Hörmander s condition holds and the solution globally exists, we prove that the law of X t(x) is continuous in variable x with respect to the total variation distance. Keywords: stochastic differential equations; Hörmander s condition; strong Feller property MR(2010) Subject Classification: 60H10; 60H07 / CLC number: O211.63 Document code: A Article ID: 1000-0917(2015)05-0783-06 0 Preinaries Let W be the space of all continuous functions from R + := [0, )tor m vanishing at starting point 0, which is endowed with the locally uniform convergence topology and the Wiener measure μ W so that the coordinate process W t (ω) =ω t is a standard m-dimensional Brownian motion. Let H W be the Cameron-Martin space consisting of all absolutely continuous functions with square integrable derivatives. The inner product in H is denoted by h 1,h 2 H := m i=1 0 ḣ i 1(s)ḣi 2(s)ds. The triple (W, H,μ W ) is also called the classical Wiener space. Let D be the Malliavin derivative operator. For k N and p 1, let D k,p be the associated Wiener-Sobolev space with the norm F k,p := F p + DF p + + D k F p, where p is the usual L p -norm. Let X : W R d be a smooth Wiener functional in k,p Dk,p and Σ X ij := DXi,DX j H be the Malliavin covariance matrix of X. Received date: 2014-02-26. Foundation item: This work was supported by the Youth Scientific Research Fund of Hunan Normal University (No. Math140650) and the Scientific Research Foundation for Ph.D Hunan Normal University (No. Math140675). E-mail: pengxuhui@amss.ac.cn

784 ν fl ff # 44ffl 1MainResults Let X n and X be d-dimensional random variables on (W, H,μ W ). In [2, Corollary 9.6.12] Bogachev proved if X n X in D 1,p (p d) and for almost all ω, {D h X(ω), h H} = R d, then the laws of X n converge to the law of X in total variation distance. In [5, Theorem 1.1], Dong et al. proved that if X n converge to X in probability, X n,x D 2,p (p>1), and sup X n 2,p <, Σ X is almost invertible, n 0 then the laws of X n converge to the law of X in total variation distance. In [5, Theorem 1.1], p>1 is needed. But in [2, Corollary 9.6.12], p>dis needed. There are also many other papers concerned about the convergence of random variables, for example [1][6] etc. In this article, we consider the following SDE: dx t = b(x t )dt + A(X t )dw t, X 0 = x, (1.1) where b : R d R d and A : R d R d m are locally Lipschitz and smooth. Consider the following vectors fields on R d associated with (b, A), A(x) =(a ij (x)), i =1, 2,,d, j =1, 2,,m, d A j = a ij (x), j =1, 2,,m, x i i=1 A 0 = b 1 2 m A l A l, l=1 here A l A l = d i,j=1 a il(x) xi a jl (x) x j. We say that (b, A) satisfies Hörmander s condition at point x R d if the vector space spanned by the vector fields {A j } j=1,2, m, {[A i,a j ] m i,j=0 }, {[[A i,a j ],A k ]} m i,j,k=0, at point x is R d.here[a i,a j ] denotes the Lie bracket between A i and A j. The generator of X t is given by where L = i b i (x) x i + 1 2 2 σ ij (x), x i x j i,j σ ij =(A A ) ij. Let H : R d R + be a C -function with x H(x) =, which is called a Lyapunov function. We assume that for some Lyapunov function H, LH ch. (1.2)

5» fiφ: The Continuity of SDE With Respect to Initial Value in the Total Variation 785 Our main work is to prove the following theorem by using [5, Theorem 1.1]. Theorem 1.1 If for any point x R d,(b, A) satisfies Hörmander s condition at point x and the solution to (1.1) globally exists, then the law of X t (x) is continuous in variable x with respect to the total variation distance. In particular, the semigroup (P t ) t>0 has the strong Feller property, i.e., for any t>0andf B b (R d ), x Ef(X t (x)) is continuous. Remark 1.1 If b, A C and (b, A) satisfies Hörmander s condition at any point x R d, then it is well known that P t is strong Feller. For example, see [7]. For the stochastic differential equations driven by degenerate subordinated Brownian motions, Zhang [9 10] made some researches on the existence and smoothness of the density of solutions. By [7, Theorem 5.9], if (1.2) holds for some Lyapunov function H and c>0, then the solution to (1.1) globally exists, so we have the following corollary. Corollary 1.1 If (b, A) satisfyhörmander s condition at any point x R d,andforsome Lyapunov function H and c>0, (1.2) holds, then the law of X t (x) is continuous in variable x with respect to the total variation distance. The article is organized as follows. In Section 2, we give the proof of Theorem 1.1, and in Section 3, we give an example. 2 Proof of Theorem 1.1 Before we give the proof of Theorem 1.1, we need some notations and a lemma. Let B M := {x R d : x <M}, B M := {x R d : x M},BM c = {x Rd : x M} for any M>0. For any n N, letχ n (x) be a cut-off function on [0, ) with χ n Bn =1, χ n B c n+1 =0, 0 χ n 1, and set We have b n (x) =b(x)χ n (x), A n (x) =A(x)χ n (x). b n,a n C b (R d ). Consider the following SDE: dx n t (x) =b n(x n t (x))dt + A n(x n t (x))dw t, X n 0 = x. (2.1) Define τ n (x) :=inf { t 0: X t (x) n }. By the uniqueness of the solution to SDE, for any x R d, n N, X n t (x) =X t (x), t<τ n (x) a.s. (2.2)

786 ν fl ff # 44ffl X n+1 t (x) =X n t (x), t<τ n(x) a.s. (2.3) Now we introduce a lemma which will be used in the proof of Theorem 1.1. The proof of this lemma is the same to [4, Lemma 4.1]. But for the convenience of reading, we prove it again. Lemma 2.1 For any n N and x B n, sup I {t τn(y)} I {t τn 1(x)} a.s. Proof Let Γ be a measurable set with P(Γ c )=0andsuchthatXs n (x, ω) is continuous with respect to s and x for ω Γ. For ω Γ, the conclusion is apparent if sup I {t τn(y)}(ω) =0 or I {t τn 1(x)}(ω) =1. Assume that sup I {t τn(y)}(ω) =1andI {t τn 1(x)}(ω) =0,then sup Xs n 1 (x, ω) n 1. (2.4) Furthermore, by (2.3), sup Xs n (x, ω) n 1. (2.5) Since sup I {t τn(y)}(ω) = 1, then there exist a sequence {y k } R d with y k x as k,andfork large enough, sup Xs n (y k,ω) n. (2.6) Since [0,t] B n [0, ) R d is a compact set, Xs n (x, ω) is an uniformly continuous function on [0,t] B n. Hence, for ε 0 = 1 2,thereexistsδ 0 > 0 such that for any y x δ 0 and any s [0,t], Xs n (y, ω) Xs n (x, ω) ε 0 = 1 2. That means when y x δ 0, sup Xs n (y, ω) sup Xs n (x, ω) + 1 2, which contradicts (2.5) and (2.6). Now we are in a position to give the proof of Theorem 1.1. ProofofTheorem1.1 Let f be a bounded measurable function. For any x, y B n, E[f(X t (x)) f(x t (y))] E[f(X t (x))i t<τn(x) f(x t (y))i t<τn(y)] + f P(t τ n (x)) + f P(t τ n (y)) E[f(X n t (x))i t<τn(x) f(x n t (y))i t<τn(y)] + f P(t τ n (x)) + f P(t τ n (y)) E[f(X n t (x)) f(x n t (y))] +2 f P(t τ n (x)) + 2 f P(t τ n (y)).

5» fiφ: The Continuity of SDE With Respect to Initial Value in the Total Variation 787 Since b n = b, A n = A on B n,(b n,a n ) also satisfy Hörmander s condition at points x and y. So the Malliavin matrices of Xt n(x),xn t (y) are almost invertible. Since b n,a n Cb (Rd ), then for any p>0, sup DXt n (u) 1,p <, u B n By [5, Theorem 1.1] or [2, Corollary 9.6.12], So, by (2.7) and Lemma 2.1, sup f 1 sup f 1 sup DXt n (u) 2,p <. u B n E[f(Xt n (x)) f(xn t (y))] =0. (2.7) E[f(X t (x)) f(x t (y))] Letting n in the above inequality, we obtain 3Example 2P(t τ n (x)) + 2 sup P(t τ n (y)) 2P(t τ n (x)) + 2P(t τ n 1 (x)). sup f 1 E[f(X t (x)) f(x t (y))] =0. In the end of this article, we give a example. The following stochastic oscillators are studied in [3, 8] etc. dz i (t) =u i (t)dt, i =1, 2,,d, du i (t) = zi H(z(t),u(t))dt, i =2, 3,,d 1, du i (t) = [ zi H(z(t),u(t)) + γ i u i (t)]dt + (3.1) T i dwt i, i =1,d, where d N, γ 1,γ d R, T 1,T d > 0, and H(z,u):= d i=1 ( ) 1 2 u i 2 + V (z i ) U(z i+1 z i ). d 1 + As that in [8], we assume (H1) Growth at infinity. U and V are C and there exist real constants l 1, k max{2,l}, a k,b l > 0 such that λ λ k U(λx) =a k x k, λ λ1 k U (λx) =ka k x k 1 sign(x), λ λ l V (λx) =a k x l, λ λ1 l V (λx) =ka l x l 1 sign(x), (H2) Non-degeneracy. For any q R, thereexistsn = n(q) 2 such that n U(q) 0. Proposition 3.1 Assume that V,U C (R) are nonnegative functions such that (H1) and (H2) hold, then the law of (z t,u t ) are continuous with the initial value in the total variation distance. i=1

788 ν fl ff # 44ffl Proof It can be directly obtained by Theorem 1.1 and [8, Theorem 1.1 and Section 3]. Acknowledgements The author is very grateful to Dong Zhao, Song Yulin and Zhang Xicheng for their quite useful conversations. References [1] Bally, V. and Caramellino, L., On the distance between probability density functions, Electron. J. Probab., 2014, 19: Article 110, 33 pages. [2] Bogachev, V.I., Differentiable Measures and the Malliavin Calculus, Math. Surveys Monogr., Vol. 164, Providence, RI: AMS, 2010. [3] Carmona, P., Existence and uniqueness of an invariant measure for a chain of oscillators in contact with two heat baths, Stochastic Process. Appl., 2007, 117(8): 1076-1092. [4] Dong, Z. and Peng, X.H., Malliavin matrix of degenerate SDE and gradient estimate, Electron. J. Probab., 2014, 19: Article 73, 26 pages. [5] Dong, Z., Peng, X.H., Song, Y.L. and Zhang, X.C., Strong Feller properties for degenerate SDEs with jumps, 2013, arxiv: 1312.7380. [6] Malicet, D. and Poly, G., Properties of convergence in Dirichlet structures, J. Funct. Anal., 2013, 264(9): 2077-2096. [7] Rey-Bellet, L., Ergodic properties of Markov processes, In: Open Quantum Systems II, Lecture Notes in Math., Vol. 1881, Berlin: Springer-Verlag, 2006, 1-39. [8] Rey-Bellet, L. and Thomas, L.E., Asymptotic behavior of thermal nonequilibrium steady states for a driven chain of anharmonic oscillators, Comm. Math. Phys., 2010, 215(1): 1-24. [9] Zhang, X.C., Densities for SDEs driven by degenerate α-stable processes, Ann. Probab., 2014, 42(5): 1885-1910. [10] Zhang, X.C., Derivative formulas and gradient estimates for SDEs driven by α-stable processes, Stochastic Process. Appl., 2013, 123(4): 1213-1228. 829/.)4,/'1>*A?6&(-7:,5<; CDB (1. μπλflνflψωßψ flfl!,,, 410081; 2. %± fl!νflψffρ flχffi!, fl, 100190) @= U X t (x) hmgyonanmetgqpjmw, x h^kr. cs Hörmander f VIZ`PJMWbXLp, Eiq] X t (x) MQGRnKr x pbfhodj[l. 03+ etgqpj; Hörmander fv; _ Feller k