Loading [MathJax]/jax/output/SVG/jax.js
Special Issues

Averaging principle on infinite intervals for stochastic ordinary differential equations

  • Received: 01 August 2020 Revised: 01 January 2021 Published: 22 February 2021
  • 34C29, 60H10, 37B20, 34C27

  • In contrast to existing works on stochastic averaging on finite intervals, we establish an averaging principle on the whole real axis, i.e. the so-called second Bogolyubov theorem, for semilinear stochastic ordinary differential equations in Hilbert space with Poisson stable (in particular, periodic, quasi-periodic, almost periodic, almost automorphic etc) coefficients. Under some appropriate conditions we prove that there exists a unique recurrent solution to the original equation, which possesses the same recurrence property as the coefficients, in a small neighborhood of the stationary solution to the averaged equation, and this recurrent solution converges to the stationary solution of averaged equation uniformly on the whole real axis when the time scale approaches zero.

    Citation: David Cheban, Zhenxin Liu. Averaging principle on infinite intervals for stochastic ordinary differential equations[J]. Electronic Research Archive, 2021, 29(4): 2791-2817. doi: 10.3934/era.2021014

    Related Papers:

    [1] David Cheban, Zhenxin Liu . Averaging principle on infinite intervals for stochastic ordinary differential equations. Electronic Research Archive, 2021, 29(4): 2791-2817. doi: 10.3934/era.2021014
    [2] Shuguan Ji, Yanshuo Li . Quasi-periodic solutions for the incompressible Navier-Stokes equations with nonlocal diffusion. Electronic Research Archive, 2023, 31(12): 7182-7194. doi: 10.3934/era.2023363
    [3] Weiyu Li, Hongyan Wang . Dynamics of a three-molecule autocatalytic Schnakenberg model with cross-diffusion: Turing patterns of spatially homogeneous Hopf bifurcating periodic solutions. Electronic Research Archive, 2023, 31(7): 4139-4154. doi: 10.3934/era.2023211
    [4] Weiwei Qi, Yongkun Li . Weyl almost anti-periodic solution to a neutral functional semilinear differential equation. Electronic Research Archive, 2023, 31(3): 1662-1672. doi: 10.3934/era.2023086
    [5] Shufen Zhao, Xiaoqian Li, Jianzhong Zhang . S-asymptotically $ \omega $-periodic solutions in distribution for a class of stochastic fractional functional differential equations. Electronic Research Archive, 2023, 31(2): 599-614. doi: 10.3934/era.2023029
    [6] Shuang Wang, FanFan Chen, Chunlian Liu . The existence of periodic solutions for nonconservative superlinear second order ODEs: a rotation number and spiral analysis approach. Electronic Research Archive, 2025, 33(1): 50-67. doi: 10.3934/era.2025003
    [7] Chungen Liu, Yuyou Zhong . Infinitely many periodic solutions for ordinary $ p(t) $-Laplacian differential systems. Electronic Research Archive, 2022, 30(5): 1653-1667. doi: 10.3934/era.2022083
    [8] Zaitao Liang, Xiuqiang Zhang, Shengjun Li, Ziqing Zhou . Periodic solutions of a class of indefinite singular differential equations. Electronic Research Archive, 2023, 31(4): 2139-2148. doi: 10.3934/era.2023110
    [9] Xintao Li, Rongrui Lin, Lianbing She . Periodic measures for a neural field lattice model with state dependent superlinear noise. Electronic Research Archive, 2024, 32(6): 4011-4024. doi: 10.3934/era.2024180
    [10] Yi Wei . The Riccati-Bernoulli subsidiary ordinary differential equation method to the coupled Higgs field equation. Electronic Research Archive, 2023, 31(11): 6790-6802. doi: 10.3934/era.2023342
  • In contrast to existing works on stochastic averaging on finite intervals, we establish an averaging principle on the whole real axis, i.e. the so-called second Bogolyubov theorem, for semilinear stochastic ordinary differential equations in Hilbert space with Poisson stable (in particular, periodic, quasi-periodic, almost periodic, almost automorphic etc) coefficients. Under some appropriate conditions we prove that there exists a unique recurrent solution to the original equation, which possesses the same recurrence property as the coefficients, in a small neighborhood of the stationary solution to the averaged equation, and this recurrent solution converges to the stationary solution of averaged equation uniformly on the whole real axis when the time scale approaches zero.



    Highly oscillating systems may be "averaged" under some suitable conditions, and the evolution of the averaged system can reflect in some sense the dynamics of the original system. This idea of averaging dates back to the perturbation theory developed by Clairaut, Laplace and Lagrange in the 18th century, and is made rigorous by Krylov, Bogolyubov, Mitropolsky [1,2,18] for nonlinear oscillations. There are vast amount of works on averaging for deterministic systems which we will not mention here. Meantime, there are also many works on averaging principle for stochastic differential equations so far, see e.g. [4,5,12,14,15,29,30,31,32] among others. But to our best knowledge, except for the centre manifold approach to averaging (see e.g. [3,22]), almost all the existing works on stochastic averaging are concerned with the so-called first Bogolyubov theorem, i.e. the convergence of the solution of the original equation to that of the averaged equation on finite intervals.

    In the present paper, we establish an averaging principle on the whole real axis, i.e. the so-called second Bogolyubov theorem, for stochastic differential equations: if there exists a stationary solution for the averaged equation, then there exists in a small neighborhood (in the super-norm topology) a solution of the original equation which is defined on the whole axis and has the same recurrence property (in distribution sense) as the coefficients of the original equation. Furthermore, this recurrent solution is more general than the classical second Bogolyubov theorem, which only treats the almost periodic case. Note that the work [16] studies the averaging principle for stochastic differential equations with almost periodic coefficients, but they only show the convergence on the finite interval, not the super-norm topology on the whole axis.

    To be more precise, we investigate the semilinear stochastic ordinary differential equation with Poisson stable (in particular, periodic, quasi-periodic, Bohr almost periodic, almost automorphic, Birkhoff recurrent, Levitan almost periodic, almost recurrent, pseudo periodic, pseudo recurrent) in time coefficients. Under some suitable conditions, this equation has a unique L2-bounded solution which has the same recurrent properties as the coefficients, see [9,20] for details. In this paper, we show that this recurrent solution converges to the unique stationary solution of the averaged equation uniformly on the whole real axis when the time scale goes to zero.

    The paper is organized as follows. In the second section we collect some known notions and facts. Namely we present the construction of shift dynamical systems, definitions and basic properties of Poisson stable functions, Shcherbakov's comparability method, and the existence of compatible solutions for stochastic differential equations. In the third and fourth sections, we investigate the averaging principle on infinite intervals for linear and semilinear stochastic differential equations respectively.

    Let (X,ρ) be a complete metric space and (X,R,π) be a dynamical system (or flow) on X, i.e. the mapping π:R×XX is continuous, π(0,x)=x and π(t+s,x)=π(t,π(s,x)) for any xX and t,sR. The mapping tπ(t,x) is called the motion through x. Denote by C(R,X) the space of all continuous functions φ:RX equipped with the distance

    d(φ1,φ2):=k=112kdk(φ1,φ2)1+dk(φ1,φ2),

    where

    dk(φ1,φ2):=sup|t|kρ(φ1(t),φ2(t)),

    which generates the compact-open topology on C(R,X). The space (C(R,X),d) is a complete metric space (see, e.g. [25,26,27,28]).

    Remark 1. (ⅰ) Let φ,φnC(R,X) (nN). Then limnd(φn,φ)=0 if and only if limnmax|t|lρ(φn(t),φ(t))=0 for any l>0.

    (ⅱ) If there exists a sequence ln+ such that limnmax|t|lnρ(φn(t),φ(t))=0, then limnd(φn,φ)=0 and vice versa. See [28] for details.

    Let us now introduce two examples of shift dynamical systems which we will use later in this paper.

    Example 2.1. For given φC(R,X), we denote by φτ the τ-translation of φ, i.e. φτ(t)=φ(τ+t) for tR. Let σ:R×C(R,X)C(R,X) be a mapping defined by equality σ(τ,φ):=φτ for (τ,φ)R×C(R,X). Clearly σ(0,φ)=φ and σ(τ1+τ2,φ)=σ(τ2,σ(τ1,φ)) for φC(R,X) and τ1,τ2R. It is immediate to check (see, e.g. [7,25,26,28]) that the mapping σ:R×C(R,X)C(R,X) is continuous, and consequently the triplet (C(R,X),R,σ) is a dynamical system which is called shift dynamical system or Bebutov dynamical system.

    The hull of φ, denoted by H(φ), is the set of all the limits of φτn in C(R,X), i.e.

    H(φ):={ψC(R,X):ψ=limnφτn for some sequence {τn}R}.

    Note that the set H(φ) is a closed and translation invariant subset of C(R,X) and consequently it naturally defines on H(φ) a shift dynamical system – (H(φ),R,σ).

    Example 2.2. Like in [9], we denote by BUC(R×X,X) the space of all continuous functions f:R×XX which are bounded on every bounded subset from R×X and continuous in tR uniformly with respect to x on each bounded subset Q of X. We equip this space with the topology of uniform convergence on bounded subsets of R×X, which can be generated by the following metric

    d(f,g):=k=112kdk(f,g)1+dk(f,g), (1)

    where

    dk(f,g):=sup|t|k, xQkρ(f(t,x),g(t,x))

    with QkX being bounded, QkQk+1 and kNQk=X.

    For given fBUC(R×X,X) and τR, we denote by fτ the τ-translation of f, i.e. fτ(t,x):=f(t+τ,x) for (t,x)R×X. Note that the space BUC(R×X,X) endowed with the distance (1) is a complete metric space and invariant with respect to translations. Now we define a mapping σ:R×BUC(R×X,X)BUC(R×X,X), (τ,f)fτ. It is clear that σ(0,f)=f and σ(τ2,σ(τ1,f))=σ(τ1+τ2,f) for all fBUC(R×X,X) and τ1,τ2R. It is immediate to see (e.g. [7,ChI]) that the mapping σ is continuous and consequently the triplet (BUC(R×X,X),R,σ) is a dynamical system. Similar to Example 2.1, for given fBUC(R×X,X), the hull H(f) is a closed and translation invariant subset of BUC(R×X,X) and consequently it naturally defines on H(f) a shift dynamical system – (H(f),R,σ).

    Denote by BC(X,X) the space of all continuous functions f:XX which are bounded on every bounded subset of X and equipped with the distance

    d(f,g):=k=112kdk(f,g)1+dk(f,g),dk(f,g):=supxQkρ(f(x),g(x))

    where Qk are the same as above. Note that (BC(X,X),d) is a complete metric space. For given FBUC(R×X,X), define F:RBC(X,X),tF(t) by letting F(t):=F(t,):XX. Clearly, FC(R,BC(X,X)).

    Remark 2. (ⅰ) Define the mapping h:BUC(R×X,X)C(R,BC(X,X)) by equality h(F):=F. It is immediate to see that the mapping h is one-one and continuous, and its inverse is also continuous. So it establishes an isometry between BUC(R×X,X) and C(R,BC(X,X)).

    (ⅱ) By the definition of h we have h(Fτ)=Fτ for any τR and FBUC(R×X,X), i.e. the shift dynamical systems (BUC(R×X,X),R,σ) and (C(R,BC(X,X)),R,σ) are (dynamically) homeomorphic.

    Let us recall the types of Poisson stable functions to be used in this paper; we refer the reader to [25,26,27,28] for further details and the relations among these types of functions.

    Definition 2.3. A function φC(R,X) is called stationary (respectively, τ-periodic) if φ(t)=φ(0) (respectively, φ(t+τ)=φ(t)) for all tR.

    Definition 2.4. (ⅰ) Let ε>0. A number τR is called ε-almost period of the function φ:RX if ρ(φ(t+τ),φ(t))<ε for all tR. Denote by T(φ,ε) the set of ε-almost periods of φ.

    (ⅱ) A function φC(R,X) is said to be Bohr almost periodic if the set of ε-almost periods of φ is relatively dense for each ε>0, i.e. for each ε>0 there exists a constant l=l(ε)>0 such that T(φ,ε)[a,a+l] for all aR.

    (ⅲ) A function φC(R,X) is said to be pseudo-periodic in the positive (respectively, negative) direction if for each ε>0 and l>0 there exists a ε-almost period τ>l (respectively, τ<l) of the function φ. The function φ is called pseudo-periodic if it is pseudo-periodic in both directions.

    Remark 3. A function φC(R,X) is pseudo-periodic in the positive (respectively, negative) direction if and only if there is a sequence tn+ (respectively, tn) such that φtn converges to φ uniformly with respect to tR as n.

    Definition 2.5. (ⅰ) A number τR is said to be ε-shift of φC(R,X) if d(φτ,φ)<ε; denote T(φ,ε):={τ:d(φτ,φ)<ε}. A function φC(R,X) is called almost recurrent (in the sense of Bebutov) if for every ε>0 the set T(φ,ε) is relatively dense.

    (ⅱ) A function φC(R,X) is called Lagrange stable if {φτ:τR} is a relatively compact subset of C(R,X).

    (ⅲ) A function φC(R,X) is called Birkhoff recurrent if it is almost recurrent and Lagrange stable.

    Definition 2.6. A function φC(R,X) is called Poisson stable in the positive (respectively, negative) direction if for every ε>0 and l>0 there exists τ>l (respectively, τ<l) such that d(φτ,φ)<ε. The function φ is called Poisson stable if it is Poisson stable in both directions.

    In what follows, we denote as well Y a complete metric space.

    Definition 2.7. (ⅰ) A function φC(R,X) is called Levitan almost periodic if there exists a Bohr almost periodic function ψC(R,Y) such that for any ε>0 there exists δ=δ(ε)>0 such that T(ψ,δ)T(φ,ε).

    (ⅱ) A function φC(R,X) is said to be almost automorphic if it is Levitan almost periodic and Lagrange stable.

    Remark 4. Note that:

    1. every Bohr almost periodic function is Levitan almost periodic;

    2. the function φC(R,R) defined by equality

    φ(t)=12+cost+cos2t

    is Levitan almost periodic, but it is not Bohr almost periodic [19,ChIV].

    Definition 2.8. A function φC(R,X) is called quasi-periodic with the spectrum of frequencies ν1,ν2,,νk if the following conditions are fulfilled:

    1. the numbers ν1,ν2,,νk are rationally independent;

    2. there exists a continuous function Φ:RkX such that Φ(t1+2π,t2+2π,,tk+2π)=Φ(t1,t2,,tk) for all (t1,t2,,tk)Rk;

    3. φ(t)=Φ(ν1t,ν2t,,νkt) for tR.

    Let φC(R,X). Denote by Nφ (respectively, Mφ) the family of all sequences {tn}R such that φtnφ (respectively, {φtn} converges) in C(R,X) as n. We denote by Nuφ (respectively, Muφ) the family of sequences {tn}Nφ (respectively, {tn}Mφ) such that φtn converges to φ (respectively, φtn converges) uniformly with respect to tR as n.

    Remark 5. (ⅰ) The function φC(R,X) is pseudo-periodic in the positive (respectively, negative) direction if and only if there is a sequence {tn}Nuφ such that tn+ (respectively, tn) as n.

    (ⅱ) Let φC(R,X), ψC(R,Y) and NuψNuφ. If the function ψ is pseudo-periodic in the positive (respectively, negative) direction, then so is φ.

    Definition 2.9. ([23,26,27]) A function φC(R,X) is called pseudo-recurrent if for any ε>0 and lR there exists Ll such that for any τ0R we can find a number τ[l,L] satisfying

    sup|t|1/ερ(φ(t+τ0+τ),φ(t+τ0))ε.

    Remark 6. ([23,26,27,28])

    1. Every Birkhoff recurrent function is pseudo-recurrent, but the inverse statement is not true in general.

    2. If the function φC(R,X) is pseudo-recurrent, then every function ψH(φ) is pseudo-recurrent.

    3. If the function φC(R,X) is Lagrange stable and every function ψH(φ) is Poisson stable, then φ is pseudo-recurrent.

    Finally, we remark that a Lagrange stable function is not Poisson stable in general, but all other types of functions introduced above are Poisson stable.

    Definition 2.10. A function FBUC(R×X,X) is said to possess the property A in tR uniformly with respect to x on every bounded subset Q of X if the motion σ(,F) through F with respect to the Bebutov dynamical system (BUC(R×X,X),R,σ) possesses the property A. Here the property A may be stationary, periodic, Bohr/Levitan almost periodic etc.

    Remark 7. Note that a function φC(R,X) possesses the property A if and only if the motion σ(,φ):RC(R,X) through φ with respect to the Bebutov dynamical system (C(R,X),R,σ) possesses this property.

    Definition 2.11. A function φC(R,X) is said to be comparable (respectively, strongly comparable) by character of recurrence with ψC(R,Y) if NψNφ (respectively, MψMφ).

    Theorem 2.12. ([26,ChII], [24]) The following statements hold.

    1. MψMφ implies NψNφ, and hence strong comparability implies comparability.

    2. Let φC(R,X) be comparable by character of recurrence with ψC(R,Y). If the function ψ is stationary (respectively, τ-periodic, Levitan almost periodic, almost recurrent, Poisson stable), then so is φ.

    3. Let φC(R,X) be strongly comparable by character of recurrence with ψC(R,Y). If the function ψ is quasi-periodic with the spectrum of frequencies ν1,ν2,,νk (respectively, Bohr almost periodic, almost automorphic, Birkhoff recurrent, Lagrange stable), then so is φ.

    4. Let φC(R,X) be strongly comparable by character of recurrence with ψC(R,Y) and ψ be Lagrange stable. If ψ is pseudo-periodic (respectively, pseudo-recurrent), then so is φ.

    Lemma 2.13. ([9]) Let φC(R,X), ψC(R,Y). The following statements hold.

    1. If MuψMuφ, then

    (a) NuψNuφ;

    (b) if the function ψ is Bohr almost periodic, then so is φ.

    2. If NuψNuφ and ψ is pseudo periodic, then so is φ.

    Let B be a real separable Banach space with the norm ||, and L(B) be the Banach space of all bounded linear operators acting on the space B equipped with operator norm . Consider the linear homogeneous equation

    ˙x=A(t)x (2)

    on the space B, where AC(R,L(B)). Denote by U(t,A) the Cauchy operator (see, e.g. [10]) of equation (2).

    Definition 2.14. Equation (2) is said to be uniformly asymptotically stable if there are positive constants N and ν such that

    GA(t,τ)Neν(tτ)  for any tτ (t,τR), (3)

    where GA(t,τ):=U(t,A)U1(τ,A) for any t,τR.

    If AC(R,L(B)), then by H(A) we denote the closure in the space C(R,L(B)) of all translations {Ah:hR}, where Ah(t):=A(t+h) for tR. Denote by Cb(R,B) the Banach space of all continuous and bounded mappings φ:RB equipped with the norm φ:=sup{|φ(t)|:tR}. Note that if fCb(R,B) and ˜fH(f), then ˜ff.

    Lemma 2.15. [6,ChIII] Suppose that equation (2) is uniformly asymptotically stable such that inequality (3) holds. Then

    G˜A(t,τ)Neν(tτ)

    for any tτ (t,τR) and ˜AH(A).

    Let (Ω,F,P) be a probability space, and L2(P,B) be the space of B-valued random variables X such that

    E|X|2:=Ω|X|2dP<.

    Then L2(P,B) is a Banach space equipped with the norm X2:=(Ω|X|2dP)1/2.

    Let P(B) be the space of all Borel probability measures on B endowed with the β metric:

    β(μ,ν):=sup{|fdμfdν|:fBL1},for μ,νP(B),

    where f are bounded Lipschitz continuous real-valued functions on B with the norm

    fBL:=Lip(f)+f,with Lip(f):=supxy|f(x)f(y)||xy|,f:=supxB|f(x)|.

    Recall that a sequence {μn}P(B) is said to weakly converges to μ if fdμnfdμ for all fCb(B), where Cb(B) is the space of all bounded continuous real-valued functions on B. It is well-known (see, e.g. [13,ChXI]) that (P(B),β) is a separable complete metric space and that a sequence {μn} weakly converges to μ if and only if β(μn,μ)0 as n.

    Definition 2.16. A sequence of random variables {Xn} is said to converge in distribution to the random variable X if the corresponding laws {μn} of {Xn} weakly converge to the law μ of X, i.e. β(μn,μ)0.

    In the following, we assume that H is a real separable Hilbert space. We still denote the norm in H by || and the operator norm in L(H) by which will not cause confusion. Let us consider the stochastic differential equation

    dX(t)=(A(t)X(t)+F(t,X(t)))dt+G(t,X(t))dW(t), (4)

    where AC(R,L(H)) and F,GC(R×H,H). Here W is a two-sided standard one-dimensional Brownian motion defined on the probability space (Ω,F,P); that is,

    W(t)={W1(t), for t0,W2(t), for t<0,

    where W1 and W2 are two independent one-sided standard one-dimensional Brownian motions. We set Ft:=σ{W(u):ut}.

    Definition 2.17. An Ft-adapted process {X(t)}tR is said to be a mild solution of equation (4) on R if it satisfies the following stochastic integral equation

    X(t)=GA(t,s)X(s)+tsGA(t,τ)F(τ,X(τ))dτ+tsGA(t,τ)G(τ,X(τ))dW(τ)

    for all ts and each sR, recalling that GA is defined in Definition 2.14.

    Definition 2.18. We say that functions F and G satisfy the condition

    (C1) if there exists a constant M0 such that |F(t,0)||G(t,0)|M for tR;

    (C2) if there exists a constant L0 such that Lip(F)Lip(G)L, where Lip(F):=suptR,xy|F(t,x)F(t,y)||xy| and similarly for Lip(G);

    (C3) if F and G are continuous in t uniformly with respect to x on each bounded subset QH.

    Remark 8. (i) If F and G satisfy the conditions (C1)–(C3), then F,GBUC(R×H,H).

    (ii) If F and G satisfy (C1)–(C2) with the constants M and L, then every pair of functions (˜F,˜G) in H(F,G):=¯{(Fτ,Gτ):τR}, the hull of (F,G), also possess the same property with the same constants.

    Definition 2.19. Let {φ(t)}tR be a mild solution of equation (4). Then φ is called compatible (respectively, strongly compatible) in distribution if N(A,F,G)˜Nφ (respectively, M(A,F,G)˜Mφ), where ˜Nφ (respectively, ˜Mφ) means the set of all sequences {tn}R such that the sequence {φ(+tn)} converges to φ() (respectively, {φ(+tn)} converges) in distribution uniformly on any compact interval.

    Theorem 2.20. Consider the equation (4). Suppose that the following conditions hold:

    (a) suptRA(t)<+;

    (b) equation (2) is uniformly asymptotically stable such that (3) holds;

    (c) the functions F and G satisfy the conditions (C1) and (C2).

    Then the following statements hold:

    1. if L<νN2+ν, then equation (4) has a unique solution ξC(R,B[0,r]) which satisfies

    ξ(t)=tGA(t,τ)F(τ,ξ(τ))dτ+tGA(t,τ)F(τ,ξ(τ))dW(τ), (5)

    where

    r=NM2+ννNL2+ν (6)

    and

    B[0,r]:={xL2(P,H): x2r};

    2. if additionally F and G satisfy (C3) and L<ν2N1+ν, then

    (a) Mu(A,F,G)˜Muξ, where ˜Muξ means the the set of all sequences {tn}R such that the sequence {ξ(+tn)} converges in distribution uniformly on R;

    (b) the solution ξ is strongly compatible in distribution.

    Proof. The proof is analogous to Theorem 4.6 in [9].

    Corollary 1. Under the conditions of Theorem 2.20 the following statements hold.

    1. If the functions AC(R,L(H)) and F,G C(R×H,H) are jointly stationary (respectively, τ-periodic, quasi-periodic with the spectrum of frequencies ν1,,νk, Bohr almost periodic, almost automorphic, Birkhoff recurrent, Lagrange stable, Levitan almost periodic, almost recurrent, Poisson stable), then equation (4) has a unique solution φCb(R,L2(P,H)) which is stationary (respectively, τ-periodic, quasi-periodic with the spectrum of frequencies ν1,,νk, Bohr almost periodic, almost automorphic, Birkhoff recurrent, Lagrange stable, Levitan almost periodic, almost recurrent, Poisson stable) in distribution.

    2. If the functions AC(R,L(H)) and F,G C(R×H,H) are Lagrange stable and jointly pseudo-periodic (respectively, pseudo-recurrent), then equation (4) has a unique solution φCb(R,L2(P,H)) which is pseudo-periodic (respectively, pseudo-recurrent) in distribution.

    Proof. These statements follow from Theorems 2.12 and 2.20 (see also Remark 2).

    Let ε0 be some fixed positive number. Consider the equation

    dX(t)=ε(A(t)X(t)+f(t))dt+εg(t)dW(t), (7)

    where AC(R,L(H)), f,gC(R,L2(P,H)), 0<εε0 and W is a two-sided standard one-dimensional Brownian motion defined on the filtered probability space (Ω,F,P,Ft), where Ft:=σ{W(u):ut}.

    Definition 3.1. Let f:R×(0,ε0]B. Following [17] we say that f(t;ε) integrally converges to 0 if for any L>0 we have

    limε0sup|ts|L|tsf(τ;ε)dτ|=0.

    If additionally there exists a constant m>0 such that

    |f(t;ε)|m 

    for any tR and 0<εε0, then we say that f(t;ε) correctly converges to 0 as ε0.

    Remark 9. [17,ChIV] If fC(R,B) and

    limT+1T|t+Ttf(s)ds|=0

    uniformly with respect to tR, then f(t;ε):=f(tε) integrally converges to 0 as ε0. If additionally the function f is bounded on R, then f(t;ε) correctly converges to 0 as ε0.

    Let AL(H). Denote by σ(A) the spectrum of A. Below we will use the following conditions:

    (A1) AC(R,L(H)) and there exists ˉAL(H) such that

    limT+1Tt+TtA(s)ds=ˉA (8)

    uniformly with respect to tR;

    (A2) fC(R,L2(P,H)) and there exists ˉfL2(P,H) such that

    limT+1Tt+Tt[f(s)ˉf]ds2=0 (9)

    uniformly with respect to tR;

    (A3) gC(R,L2(P,H)) and there exists ˉgL2(P,H) such that

    limT+1Tt+TtE|g(τ)ˉg|2dτ=0 (10)

    uniformly with respect to tR.

    Denote by Ψ the family of all decreasing, positive bounded functions ψ:R+R+ with limt+ψ(t)=0.

    Lemma 3.2. Let l>0 and ψΨ, then

    limε0sup0τlτψ(τε)=0.

    Proof. Let ε and l be two arbitrary positive numbers, ν(0,1) and ψΨ, then we have

    sup0τlτψ(τε)sup0τεντψ(τε)+supεντlτψ(τε)ενψ(0)+lψ(εν1).

    Letting ε0 we obtain the required result.

    Remark 10. (ⅰ) By Lemma 2 in [8] equality (8) holds if and only if there exists a function ωΨ satisfying

    1Tt+TtA(s)dsˉAω(T)

    for any T>0 and tR.

    (ⅱ) Similarly equality (9) (respectively, equality (10)) holds if and only if there exists a function ω1Ψ (respectively, ω2Ψ) satisfying

    1Tt+Tt[f(s)ˉf]ds2ω1(T)

    (respectively, 1Tt+TtE|g(τ)ˉg|2dτω2(T)) for any T>0 and tR.

    Theorem 3.3. [17,ChIV] Suppose that ACb(R,L(B)) and

    limT+1Tt+TtA(s)ds=ˉA

    uniformly with respect to tR and the operator ˉA is Hurwitz, i.e. Re λ<0 for any λσ(ˉA).

    Then the following statements hold:

    1. there exists a positive constant αε0 such that the equation

    dx(τ)=Aε(τ)x(τ)dτ,

    where Aε(τ):=A(τε) for any τR, is uniformly asymptotically stable for any 0<εα. Moreover there are constants N>0 and ν>0 such that

    GAε(τ,τ0)Neν(ττ0)

    for any ττ0 and 0<εα;

    2. there exists γ0>0 such that

    limε0sup(ττ0;τ,τ0R)eγ0(ττ0)GAε(τ,τ0)GˉA(τ,τ0)=0 . (11)

    Remark 11. (ⅰ) Note that Theorem 3.3 was proved for finite-dimensional almost periodic equations (this means that the matrix-function A() is almost periodic). For the proof for infinite-dimensional almost periodic systems see [19,ChXI].

    (ⅱ) It is not difficult to show that Theorem 3.3 remains true in general case (see above) and can be proved with slight modifications of the reasoning from [17,ChIV].

    (ⅲ) Under the conditions of Theorem 3.3 there are positive constants α,N and ν so that

    GAε(t,τ),GˉA(t,τ)Neν(tτ) (12)

    for any 0<εα and tτ.

    Lemma 3.4. Let fεC(R,B) for ε(0,α] be functions satisfying the following conditions:

    1. there exists a positive constant A such that |fε(t)|A for any tR and ε(0,α];

    2. for any l>0

    limε0sup|s|l, tR|t+stfε(σ)dσ|=0. (13)

    Then for any ν>0 we have

    limε0sup tR|teν(tτ)fε(τ)dτ|=0.

    Proof. To estimate the integral

    I(t,ε):=|teν(tτ)fε(τ)dτ|,

    we make the change τt=s, then

    teν(tτ)fε(τ)dτ=0eνsfε(t+s)ds=0eνsdds(t+stfε(σ)dσ)ds. (14)

    Since

    |eνst+stfε(σ)dσ|Aeνs|s|

    for any s<0, we have

    limseνst+stfε(σ)dσ=0 . (15)

    Integrating by parts and taking into consideration (15) we obtain

    0eνsdds(t+stfε(σ)dσ)ds=0νeνst+stfε(σ)dσds=lνeνst+stfε(σ)dσds0lνeνst+stfε(σ)dσds . (16)

    Note that

    |lνeνst+stfε(σ)dσds|Aeνl(l+1ν) (17)

    and

    |0lνeνst+stfε(σ)dσds|(1eνl)sup|s|l, tR|t+stfε(σ)dσ|. (18)

    By (14)–(18) we get

    |teν(tτ)fε(τ)dτ|Aeνl(l+1ν)+(1eνl)sup|s|l, tR|t+stfε(σ)dσ|.

    Then

    suptRI(t,ε)Aeνl(l+1ν)+(1eνl)sup|s|l, tR|t+stfε(σ)dσ|. (19)

    Since

    limε0sup|s|l, tR|t+stfε(σ)dσ|=0

    for any l>0, letting ε0 in (19) we have

    lim supε0suptRI(t,ε)Aeνl(l+1ν).

    Since l is arbitrary, it follows that

    limε0suptRI(t,ε)=0.

    The proof is complete.

    Remark 12. If the function fCb(R,B) and ˉfB are such that

    limL+1Lt+Lt[f(s)ˉf]ds=0 (20)

    uniformly with respect to tR, then the function fε(σ):=f(σε)ˉf satisfies the conditions of Lemma 3.4. Indeed, note that

    t+stfε(σ)dσ=t+st[f(σε)ˉf]dσ=sεst/ε+s/εt/ε[f(˜σ)ˉf]d˜σ,

    so the condition (13) of Lemma 3.4 holds by (20). Similarly, if the function g in (A3) is L2-bounded, then the function fε(σ):=E|g(σε)ˉg|2 satisfies as well the conditions of Lemma 3.4.

    Let Wε(t):=εW(tε) for tR. Then Wε is also a Brownian motion with the same distribution as W.

    Theorem 3.5. Suppose that ACb(R,L(H)), f,gCb(R,L2(P,H)) and conditions (A1)–(A3) are fulfilled. Suppose further that ˉA in (A1) is Hurwitz such that (11)–(12) holds. Then we have the following conclusions:

    1. equation

    dXε(t)=(Aε(t)Xε(t)+fε(t))dt+gε(t)dWε(t) (21)

    has a unique bounded solution ψεCb(R,L2(P,H)) defined by equality

    ψε(t)=tGAε(t,τ)fε(τ)dτ+tGAε(t,τ)gε(τ)dWε(τ) 

    and it is strongly compatible in distribution (i.e. M(Aε,fε,gε)˜Mψε) and Mu(Aε,fε,gε)˜Muψε, where Aε(t):=A(tε), fε(t):=f(tε) and gε(t):=g(tε) for tR;

    2. equation

    dXε(t)=(Aε(t)Xε(t)+fε(t))dt+gε(t)dW(t)

    has a unique bounded solution ϕεCb(R,L2(P,H)) defined by equality

    ϕε(t)=tGAε(t,τ)fε(τ)dτ+tGAε(t,τ)gε(τ)dW(τ); (22)

    3.

    limε0suptRE|ϕε(t)ˉϕ(t)|2=0, (23)

    where ˉϕ is the unique stationary solution of equation

    dX(t)=(ˉAX(t)+ˉf)dt+ˉgdW(t); (24)

    4. for any ε(0,α] equation (7) has a unique solution φεCb(R,L2(P,H)) and it is strongly compatible in distribution (i.e. M(A,f,g)˜Mφε) and Mu(A,f,g)˜Muφε;

    5.

    limε0suptRβ(L(φε(tε)),L(ˉϕ(t)))=0,

    where L(X) denotes the law of random variable X.

    Proof. The first and second statements follow directly from Theorem 2.20.

    We now verify the third statement, i.e. the uniform convergence of the unique bounded solution ϕε to the unique stationary solution ˉϕ of the averaged equation. By Theorem 2.20 equation (24) has a unique bounded and stationary solution ˉϕ, which is given by the formula

    ˉϕ(t)=tGˉA(t,τ)ˉfdτ+tGˉA(t,τ)ˉgdW(τ), (25)

    where GˉA(t,τ)=exp{ˉA(tτ)} for t,τR. From (22) and (25) we get

    E|ϕε(t)ˉϕ(t)|2=E|tGAε(t,τ)fε(τ)dτ+tGAε(t,τ)gε(τ)dW(τ)tGˉA(t,τ)ˉfdτtGˉA(t,τ)ˉgdW(τ)|22(E|t[GAε(t,τ)fε(τ)GˉA(t,τ)ˉf]dτ|2+E|t[GAε(t,τ)gε(τ)GˉA(t,τ)ˉg]dW(τ)|2)=:I1(t,ε)+I2(t,ε).

    By equality (11) there exists a function N:(0,α)R+ such that N(ε)0 as ε0 and

    GAε(t,τ)GˉA(t,τ)N(ε)eγ0(tτ)

    for any tτ (t,τR).

    Note that

    I1(t,ε):=2E|t[GAε(t,τ)fε(τ)GˉA(t,τ)ˉf]dτ|2=2E|t[GAε(t,τ)fε(τ)GAε(t,τ)ˉf]+[GAε(t,τ)ˉfGˉA(t,τ)ˉf]dτ|24E|tGAε(t,τ)(fε(τ)ˉf)dτ|2+4E|t[GAε(t,τ)GˉA(t,τ)]ˉfdτ|2=:4(I11(t,ε)+I12(t,ε)).

    To estimate the integral

    I11(t,ε):=E|tGAε(t,τ)(fε(τ)ˉf)dτ|2,

    making the change of variable s:=τt we obtain

    tGAε(t,τ)(fε(τ)ˉf)dτ=0GAε(t,t+s)(fε(t+s)ˉf)ds=0GAε(t,t+s)dds(t+st[fε(σ)ˉf]dσ)ds. (26)

    Since

    GAε(t,t+s)t+st[fε(σ)ˉf]dσ22Nfeνs|s|

    for any s<0, we have

    limsGAε(t,t+s)t+st[fε(σ)ˉf]dσ=0.

    Consequently, integrating by parts from (26) we get

    0GAε(t,t+s)dds(t+st[fε(σ)ˉf]dσ)ds=0GAε(t,t+s)s(t+st[fε(σ)ˉf]dσ)ds.

    Note that

    GA(t,τ)τ=GA(t,τ)A(τ),

    so we have

    GAε(t,t+s)sNAeνs

    for any tR and s<0.

    Let now l be an arbitrary positive number, then we have

    0GAε(t,t+s)s(t+st[fε(σ)ˉf]dσ)ds=lGAε(t,t+s)s(t+st[fε(σ)ˉf]dσ)ds0lGAε(t,t+s)s(t+st[fε(σ)ˉf]dσ)ds

    and consequently

    0GAε(t,t+s)s(t+st[fε(σ)ˉf]dσ)ds2lGAε(t,t+s)s(t+st[fε(σ)ˉf]dσ)ds2+0lGAε(t,t+s)s(t+st[fε(σ)ˉf]dσ)ds2NA(2f|lseνsds|+|0leνsds|sup|s|l,tRt+st[fε(σ)ˉf]dσ2)NA(2feνl(l+1ν)+1ν(1eνl)sup|s|l,tRt+st[fε(σ)ˉf]dσ2).

    Letting ε0 in above inequality we get

    lim supε0suptR0GAε(t,t+s)s(t+st[fε(σ)ˉf]dσ)ds22NAfeνl(l+1ν).

    Since l is arbitrary, we get by letting l

    limε0suptRI11(t,ε)=0.

    Note by Theorem 3.3–(ⅱ) that

    I12(t,ε):=E|t[GAε(t,τ)GˉA(t,τ)]ˉfdτ|2(ˉf2N(ε)γ0)20

    as ε0. Consequently,

    limε0suptRI1(t,ε)=0.

    Similarly we can show that

    limε0suptRI2(t,ε)=0. (27)

    In fact, using Itô's isometry property, the Cauchy-Schwartz inequality and reasoning as above we get

    I2(t,ε)=2E|t[GAε(t,τ)gε(τ)GˉA(t,τ)ˉg]dW(τ)|2=2Et|GAε(t,τ)gε(τ)GˉA(t,τ)ˉg|2dτ4(Et|GAε(t,τ)(gε(τ)ˉg)|2dτ+Et|(GAε(t,τ)GˉA(t,τ))ˉg|2dτ)4(EtN2e2ν(tτ)|gε(τ)ˉg|2dτ+EtN(ε)2e2γ0(tτ)|ˉg|2dτ)=4(N2te2ν(tτ)E|gε(τ)ˉg|2dτ+(N(ε))2ˉg222γ0). (28)

    By Lemma 3.4 the integral

    te2ν(tτ)E|gε(τ)ˉg|2dτ (29)

    goes to 0 as ε0 uniformly with respect to tR.

    Passing to the limit in (28) and taking into account (29) we obtain (27), and consequently limε0suptRE|ϕε(t)ˉϕ(t)|2=0.

    To prove the fourth statement we note that the function φε(t):=ψε(εt) (for tR) is a bounded solution of equation (7) if ψε is a bounded solution of equation (21). The uniqueness follows from the fact if φi (i=1,2) are two different bounded solutions of equation (7), then ψi(t):=φi(tε),tR (i=1,2) are two different bounded solutions of equation (21), a contradiction to the first statement. It remains to show that M(A,f,g)˜Mφε and Mu(A,f,g)˜Muφε. Let {tn}M(A,f,g) (respectively, {tn}Mu(A,f,g)), then {εtn}M(Aε,fε,gε)˜Mψε (respectively, {εtn}Mu(Aε,fε,gε)˜Muψε) by Theorem 2.20. By the relation between φε and ψε, we have {tn}˜Mφε (respectively, {tn}˜Muφε).

    Now we are in the position to prove the last statement. Since the L2 convergence implies convergence in probability, it follows from (23) that

    limε0suptRβ(L(ϕε(t)),L(ˉϕ(t)))=0.

    On the other hand taking into consideration that L(W)=L(Wε), we have L(φε(tε))=L(ϕε(t)) for any tR, and consequently

    limε0suptRβ(L(φε(tε)),L(ˉϕ(t)))=0.

    The proof is complete.

    Corollary 2. Under the conditions of Theorem 3.5 the following statements hold:

    1. if the functions AC(R,L(H)) and f,g Cb(R, L2(P,H)) are jointly stationary (respectively, τ-periodic, quasi-periodic with the spectrum of frequencies ν1,,νk, Bohr almost periodic, almost automorphic, Birkhoff recurrent, Lagrange stable, Levitan almost periodic, almost recurrent, Poisson stable), then equation (7) has a unique solution φεCb(R,L2(P,H)) which is stationary (respectively, τ-periodic, quasi-periodic with the spectrum of frequencies ν1,,νk, Bohr almost periodic, almost automorphic, Birkhoff recurrent, Lagrange stable, Levitan almost periodic, almost recurrent, Poisson stable) in distribution;

    2. if the functions AC(R,L(H)) and f,g Cb(R, L2(P,H)) are Lagrange stable and jointly pseudo-periodic (respectively, pseudo-recurrent), then equation (7) has a unique solution φεCb(R,L2(P,H)) which is pseudo-periodic (respectively, pseudo-recurrent) in distribution;

    3.

    limε0suptRβ(L(φε(tε),L(ˉϕ(t)))=0 .

    Proof. These statements follow from Theorems 2.12 and 3.5.

    Remark 13. Note that the constant α>0 in Theorem 3.3, Remark 11, Lemma 3.4 and Theorem 3.5 can be chosen the same, e.g. we may choose the minimal one of them. But there is no restriction on ε0 which we fix at the beginning of this section.

    As before, let ε0 be some fixed positive constant. Consider the following stochastic differential equation

    dX(t)=ε(A(t)X(t)+F(t,X(t)))dt+εG(t,X(t))dW(t), (30)

    where AC(R,L(H)), F,GC(R×H,H), 0<εε0 and W is a two-sided standard one-dimensional Brownian motion defined on the filtered probability space (Ω,F,P,Ft), with Ft:=σ{W(u):ut}. Below we will use the following conditions:

    (G1) there exists a positive constant M such that

    |F(t,0)||G(t,0)|M

    for any tR;

    (G2) there exists a positive constant L such that

    |F(t,x1)F(t,x2)||G(t,x1)G(t,x2)|L|x1x2|

    for any x1,x2H and tR;

    (G3) there exist functions ω1Ψ and ˉFC(H,H) such that

    1T|t+Tt[F(s,x)ˉF(x)]ds|ω1(T)(1+|x|)

    for any T>0, xH and tR;

    (G4) there exist functions ω2Ψ and ˉGC(H,H) such that

    1Tt+Tt|G(s,x)ˉG(x)|2dsω2(T)(1+|x|2)

    for any T>0, xH and tR;

    (G5) AC(R,L(H)) and there exists ˉAL(H) such that

    limT+1Tt+TtA(s)ds=ˉA

    uniformly with respect to tR.

    Remark 14. Under the conditions (G1)(G4) the functions ˉF and ˉG also possess the properties (G1)(G2) with the same constants M and L.

    We consider as well the following equations

    dX(t)=(Aε(t)X(t)+Fε(t,X(t)))dt+Gε(t,X(t))dW(t) (31)

    and

    dX(t)=(Aε(t)X(t)+Fε(t,X(t)))dt+Gε(t,X(t))dWε(t), (32)

    where Aε(t):=A(tε), Fε(t,x):=F(tε,x) and Gε(t,x):=G(tε,x) for tR, xH and ε(0,ε0]. Here as before Wε(t):=εW(tε) for tR. Along with equations (31)–(32) we also consider the following averaged equation

    dX(t)=(ˉAX(t)+ˉF(X(t)))dt+ˉG(X(t))dW(t). (33)

    Lemma 4.1. Suppose F,GC(R×H,H) and that the conditions (G1)–(G2) hold. If φ is an L2-bounded solution (i.e. φ=suptRE|φ(t)|2<+) of the equation

    dX(t)=F(t,X(t))dt+G(t,X(t))dW(t).

    then there exists a constant C>0, depending only on M,L,φ, such that

    E|φ(t+h)φ(t)|2Ch

    and

    Esuptst+h|φ(s)|2C(h2+1)

    for any tR and h>0.

    Proof. Since

    φ(t+h)=φ(t)+t+htF(τ,φ(τ))dτ+t+htG(τ,φ(τ))dW(τ),

    by Cauchy-Schwartz inequality and Itô's isometry property we have

    E|φ(t+h)φ(t)|22(E|t+htF(τ,φ(τ))dτ|2+E|t+htG(τ,φ(τ))dW(τ)|2)2(Et+ht12dτEt+ht|F(τ,φ(τ))|2dτ+t+htE|G(τ,φ(τ))|2dτ)2(ht+htE|F(τ,φ(τ))|2dτ+t+htE|G(τ,φ(τ))|2dτ)4(ht+ht(M2+L2φ2)dτ+t+ht(M2+L2φ2)dτ)Ch.

    Employing the BDG inequality (see, e.g. [11,Theorem 4.36] on page 114), we have

    Esuptst+h|φ(s)|23E|φ(t)|2+3Esuptst+h|stF(τ,φ(τ))dτ|2+3Esuptst+h|stG(τ,φ(τ))dW(τ)|23φ2+3Esuptst+h|st(M+L|φ(τ)|)dτ|2+3CEt+ht|G(τ,φ(τ))|2dτ3φ2+3ht+ht2(M2+L2φ2)dτ+3Ct+ht2(M2+L2φ2)dτC(h2+1),

    where C denotes some positive constants which may change from line to line.

    Theorem 4.2. Suppose that the following conditions hold:

    (a) suptRA(t)<+;

    (b) the functions A, F, G satisfy the conditions (G1)(G5), and the operator ˉA in (G5) is Hurwitz, i.e. Re λ<0 for any λσ(ˉA);

    (c)

    L<ν3N2+ν,

    where α,N and ν are the numbers figuring in Remark 11-(ⅲ).

    Then there exists a positive constant ε1α such that for any 0<εε1

    1. equation (30) has a unique solution φεCb(R,L2(P,H)) and φεr, where r:=NM2+ννNL2+ν;

    2. equation (31) has a unique solution ϕεCb(R,L2(P,H)) and ϕεr;

    3. if additionally F and G satisfy (C3) and L<ν2N1+ν, then the solution φε of equation (30) is strongly compatible in distribution (i.e. M(A,F,G)˜Mφε) and Mu(A,F,G)˜Muφε, recalling that ˜Muφε means the set of all sequences {tn} such that φε(t+tn) converges in distribution uniformly with respect to tR;

    4.

    limε0suptRE|ϕε(t)ˉϕ(t)|2=0,

    where ˉϕ is the unique stationary solution of equation (33);

    5.

    limε0suptRβ(L(φε(tε),L(ˉϕ(t)))=0.

    Proof. By Theorem 3.3 (see also Remark 11-(ⅲ)) there exist positive constants N,ν and α such that equation

    dX(t)=Aε(t)X(t)dt

    is uniformly asymptotically stable for any 0<εα and

    GAε(t,τ)Neν(tτ)

    for any tτ. By Theorem 3.3-(ⅱ) there are γ0>0 and N:(0,α)R+ such that N(ε)0 as ε0 and

    GAε(t,τ)GˉA(t,τ)N(ε)eγ0(tτ)

    for any tτ.

    Since Lip(Fε)=Lip(F)L and Lip(Gε)=Lip(G)L, by Theorem 2.20 equation (30) (respectively, equation (31)) has a unique solution φε (respectively, ϕε) from Cb(R,L2(P,B)) with φεC(R,B[0,r]) (respectively, ϕεC(R,B[0,r])), where

    r:=NM2+ννNL2+ν;

    and the solution φε is strongly compatible in distribution and Mu(A,F,G)˜Muφε.

    Let ˉϕ be the unique stationary solution of equation (33). We now estimate E|ϕε(t)ˉϕ(t)|2. To this end, we note that

    E|ϕε(t)ˉϕ(t)|2=E|tGAε(t,τ)Fε(τ,ϕε(τ))dτ+tGAε(t,τ)Gε(τ,ϕε(τ))dW(τ)tGˉA(t,τ)ˉF(ˉϕ(τ))dτtGˉA(t,τ)ˉG(ˉϕ(τ))dW(τ)|22(E|t(GAε(t,τ)Fε(τ,ϕε(τ))GˉA(t,τ)ˉF(ˉϕ(τ)))dτ|2+E|t(GAε(t,τ)Gε(τ,ϕε(τ))GˉA(t,τ)ˉG(ˉϕ(τ)))dW(τ)|2)=:2(I1(t,ε)+I2(t,ε)). (34)

    Since

    I1(t,ε):=E|t(GAε(t,τ)Fε(τ,ϕε(τ))GˉA(t,τ)ˉF(ˉϕ(τ)))dτ|23(E|tGAε(t,τ)(Fε(τ,ϕε(τ))Fε(τ,ˉϕ(τ)))dτ|2+E|t(GAε(t,τ)GˉA(t,τ))Fε(τ,ˉϕ(τ))dτ|2+E|tGˉA(t,τ)[Fε(τ,ˉϕ(τ))ˉF(ˉϕ(τ))]dτ|2),

    using Cauchy-Schwartz inequality we get

    I1(t,ε)3(N2L2νteν(tτ)E|ϕε(τ)ˉϕ(τ)|2dτ+2N(ε)2γ0teγ0(tτ)(M2+L2ˉϕ2)dτ+E|tGˉA(t,τ)[Fε(τ,ˉϕ(τ))ˉF(ˉϕ(τ))]dτ|2)3(N2L2ν2suptRE|ϕε(t)ˉϕ(t)|2+2N(ε)2γ20(M2+L2ˉϕ2)+E|tGˉA(t,τ)[Fε(τ,ˉϕ(τ))ˉF(ˉϕ(τ))]dτ|2). (35)

    We will show that

    limε0suptRE|tGˉA(t,τ)[Fε(τ,ˉϕ(τ))ˉF(ˉϕ(τ))]dτ|2=0.

    To this end, making the change of variable s=τt and integrating by parts, we obtain for any l>0

    E|tGˉA(t,τ)[Fε(τ,ˉϕ(τ))ˉF(ˉϕ(τ))]dτ|2=E|0GˉA(t,t+s)[Fε(t+s,ˉϕ(t+s))ˉF(ˉϕ(t+s))]ds|2=E|0GˉA(t,t+s)dds(t+st[Fε(σ,ˉϕ(σ))ˉF(ˉϕ(σ))]dσ)ds|22E|0GˉA(t,t+s)s(t+st[Fε(σ,ˉϕ(σ))ˉF(ˉϕ(σ))]dσ)ds|24E|lGˉA(t,t+s)s(t+st[Fε(σ,ˉϕ(σ))ˉF(ˉϕ(σ))]dσ)ds|2+4E|0lGˉA(t,t+s)s(t+st[Fε(σ,ˉϕ(σ))ˉF(ˉϕ(σ))]dσ)ds|24E(l|t+st[Fε(σ,ˉϕ(σ))ˉF(ˉϕ(σ))]dσ|NˉAeνsds)2+4Esupls0|t+st[Fε(σ,ˉϕ(σ))ˉF(ˉϕ(σ))]dσ|2|0lNˉAeνsds|2=:J1+J2. (36)

    For J1, we have

    J1:=4E(l|t+st[Fε(σ,ˉϕ(σ))ˉF(ˉϕ(σ))]dσ|NˉAeνsds)24N2ˉA2leνsdslE|t+st[Fε(σ,ˉϕ(σ))ˉF(ˉϕ(σ))]dσ|2eνsds4N2ˉA2νeνllE|t+st[Fε(σ,ˉϕ(σ))ˉF(ˉϕ(σ))]dσ|2eνsds4N2ˉA2νeνllst+stE[Fε(σ,ˉϕ(σ))ˉF(ˉϕ(σ))]2dσeνsds4N2ˉA2νeνllst+st8(M2+L2ˉϕ2)dσeνsds32N2ˉA2ν(M2+L2ˉϕ2)eνlls2eνsds32N2ˉA2ν(M2+L2ˉϕ2)(l2ν+2lν2+2ν3)e2νl. (37)

    Divide [0,l] into intervals of size δ, where δ>0 is a fixed constant depending on ε. Denote an adapted process ˜ϕ such that ˜ϕ(σ)=ˉϕ(tkδ) for any σ(t(k+1)δ,tkδ]. By Lemma 4.1, we have

    Esupls0|t+st[Fε(σ,ˉϕ(σ))ˉF(ˉϕ(σ))]dσ|2=Esupls0|t+st[Fε(σ,ˉϕ(σ))Fε(σ,˜ϕ(σ))+Fε(σ,˜ϕ(σ))ˉF(˜ϕ(σ))+ˉF(˜ϕ(σ))ˉF(ˉϕ(σ))]dσ|26Esupls0|t+stL|ˉϕ(σ)˜ϕ(σ)|dσ|2+3Esupls0|t+st[Fε(σ,˜ϕ(σ))ˉF(˜ϕ(σ))]dσ|26Esupls0ltt+sL2|ˉϕ(σ)˜ϕ(σ)|2dσ+3Esupls0|t+st[Fε(σ,˜ϕ(σ))ˉF(˜ϕ(σ))]dσ|26L2l2Cδ+3Esupls0|t+st[Fε(σ,˜ϕ(σ))ˉF(˜ϕ(σ))]dσ|2=:6L2l2Cδ+i2.

    For i2, denote s(δ):=[|s|δ], we have

    i2:=3Esupls0|t+st(Fε(τ,˜ϕ(τ))ˉF(˜ϕ(τ)))dτ|2=3Esupls0|s(δ)1k=0t(k+1)δtkδ(Fε(τ,ˉϕ(tkδ))ˉF(ˉϕ(tkδ)))dτ+t+sts(δ)δ(Fε(τ,ˉϕ(ts(δ)δ))ˉF(ˉϕ(ts(δ)δ)))dτ|26[lδ]Esupls0s(δ)1k=0|t(k+1)δtkδ(Fε(τ,ˉϕ(tkδ))ˉF(ˉϕ(tkδ)))dτ|2+6Esupls0|t+sts(δ)δ(Fε(τ,ˉϕ(ts(δ)δ))ˉF(ˉϕ(ts(δ)δ)))dτ|2=:i12+i22. (38)

    For i12, by Lemma 4.1 we have

    i12:=6[lδ]Esupls0s(δ)1k=0|t(k+1)δtkδ(Fε(τ,ˉϕ(tkδ))ˉF(ˉϕ(tkδ)))dτ|26l2δ2Esupls0max0ks(δ)1|t(k+1)δεtkδε(F(τ,ˉϕ(tkδ))ˉF(ˉϕ(tkδ)))εdτ|212l2δ2Esupls0max0ks(δ)1δ2ω21(δε)(1+|ˉϕ(tkδ)|2)12l2(C+Cl2+1)ω21(δε). (39)

    For i22, we obtain

    i22:=6Esupls0|t+sts(δ)δ(Fε(τ,ˉϕ(ts(δ)δ))ˉF(ˉϕ(ts(δ)δ)))dτ|26Esupls0δts(δ)δt+s(Fε(τ,ˉϕ(ts(δ)δ))ˉF(ˉϕ(ts(δ)δ)))2dτ6δEsupls0ts(δ)δt+s8(M2+L2|ˉϕ(ts(δ)δ)|2)dτ6δttl8(M2+L2Esupσ[tl,t]ˉϕ(σ)2)dτ48(M2+L2C(l2+1))lδ. (40)

    Therefore, (38)–(40) imply

    i212(Cl4+(C+1)l2)ω21(δε)+48(M2+L2C(l2+1))lδ.

    Therefore, we have

    J24N2ˉA2ν2(1eνl)2[6L2l2Cδ+12(Cl4+(C+1)l2)ω21(δε)+48(M2+L2C(l2+1))lδ]. (41)

    Combing (36), (37) and (41), we have

    E|tGˉA(t,τ)[Fε(τ,ˉϕ(τ))ˉF(ˉϕ(τ))]dτ|232N2ˉA2ν(M2+L2ˉϕ2)(l2ν+2lν2+2ν3)e2νl+4N2ˉA2ν2(1eνl)2[6L2l2Cδ+12(Cl4+(C+1)l2)ω21(δε)+48(M2+L2C(l2+1))lδ]. (42)

    Taking δ=ε and letting ε0 in (42), we have

    lim supε0suptRE|tGˉA(t,τ)[Fε(τ,ˉϕ(τ))ˉF(ˉϕ(τ))]dτ|232N2ˉA2ν(M2+L2ˉϕ2)(l2ν+2lν2+2ν3)e2νl.

    Since l is arbitrary, by letting l we get

    limε0suptRE|tGˉA(t,τ)[Fε(τ,ˉϕ(τ))ˉF(ˉϕ(τ))]dτ|2=0. (43)

    From (35) and (43) it follows that there exists a function A:(0,ε0)R+ so that A(ε)0 as ε0 and

    I1(t,ε)3N2L2ν2suptRE|ϕε(t)ˉϕ(t)|2+A(ε) (44)

    for any tR and ε(0,ε0).

    Now we will establish a similar estimation for I2(t,ε). Since

    I2(t,ε):=E|t(GAε(t,τ)Gε(τ,ϕε(τ))GˉA(t,τ)ˉG(ˉϕ(τ)))dW(τ)|23(E|tGAε(t,τ)(Gε(τ,ϕε(τ))Gε(τ,ˉϕ(τ)))dW(τ)|2+E|t(GAε(t,τ)GˉA(t,τ))Gε(τ,ˉϕ(τ))dW(τ)|2+E|tGˉA(t,τ)[Gε(τ,ˉϕ(τ))ˉG(ˉϕ(τ))]dW(τ)|2),

    using Itô's isometry property we have

    I2(t,ε)3(N2L2te2ν(tτ)E|ϕε(τ)ˉϕ(τ)|2dτ+2N(ε)2te2γ0(tτ)(M2+L2ˉϕ2)dτ+N2te2ν(tτ)E|Gε(τ,ˉϕ(τ))ˉG(ˉϕ(τ))|2dτ)3(N2L22νsuptRE|ϕε(t)ˉϕ(τ)|2+N(ε)2γ0(M2+L2ˉϕ2)+N2te2ν(tτ)E|Gε(τ,ˉϕ(τ))ˉG(ˉϕ(τ))|2dτ). (45)

    Now we prove that

    limε0suptR|te2ν(tτ)E|Gε(τ,ˉϕ(τ))ˉG(ˉϕ(τ))|2dτ|=0.

    By Lemma 3.4, it suffices to show that

    limε0sup|s|l,tR|t+stE|Gε(τ,ˉϕ(τ))ˉG(ˉϕ(τ))|2dτ|=0.

    To this end, define an adapted process ˆϕ such that ˆϕ(σ)=ˉϕ(t+kδ) for any σ[t+kδ,t+(k+1)δ). We can assume s>0 without loss of generality, then we have by Lemma 4.1

    t+stE|Gε(τ,ˉϕ(τ))ˉG(ˉϕ(τ))|2dτt+stE|Gε(τ,ˉϕ(τ))Gε(τ,ˆϕ(τ))+Gε(τ,ˆϕ(τ))ˉG(ˆϕ(τ))+ˉG(ˆϕ(τ))ˉG(ˉϕ(τ))|2dτ3t+stE|Gε(τ,ˉϕ(τ))Gε(τ,ˆϕ(τ))|2dτ+3t+stE|Gε(τ,ˆϕ(τ))ˉG(ˆϕ(τ))|2dτ+3t+stE|ˉG(ˆϕ(τ))ˉG(ˉϕ(τ))|2dτ6L2lCδ+3t+stE|Gε(τ,ˆϕ(τ))ˉG(ˆϕ(τ))|2dτ=:6L2lCδ+3J3.

    For J3, we have

    J3:=Et+st|Gε(τ,ˆϕ(τ))ˉG(ˆϕ(τ))|2dτE(s(δ)1k=0t+(k+1)δt+kδ|Gε(τ,ˉϕ(t+kδ))ˉG(ˉϕ(t+kδ))|2dτ+t+st+s(δ)δ|Gε(τ,ˉϕ(t+s(δ)δ))ˉG(ˉϕ(t+s(δ)δ))|2dτ)=:J13+J23.

    Then

    J13:=E(s(δ)1k=0t+(k+1)δt+kδ|Gε(τ,ˉϕ(t+kδ))ˉG(ˉϕ(t+kδ))|2dτ)[lδ]max0ks(δ)1Et+(k+1)δt+kδ|Gε(τ,ˉϕ(t+kδ))ˉG(ˉϕ(t+kδ))|2dτ=[lδ]max0ks(δ)1Et+(k+1)δεt+kδε|G(τ,ˉϕ(t+kδ))ˉG(ˉϕ(t+kδ))|2εdτlω2(δε)(1+ˉϕ2)

    and

    J23:=Et+st+s(δ)δ|Gε(τ,ˉϕ(t+s(δ)δ))ˉG(ˉϕ(t+s(δ)δ))|2dτ8(M2+L2ˉϕ2)δ.

    Therefore we have

    sup|s|l,tR|t+stE|Gε(τ,ˉϕ(τ))ˉG(ˉϕ(τ))|2dτ|6L2lCδ+24(M2+L2ˉϕ2)δ+3lω2(δε)(1+ˉϕ2). (46)

    Taking δ=ε and letting ε0 in (46), we have

    limε0sup|s|l,tR|t+stE|Gε(τ,ˉϕ(τ))ˉG(ˉϕ(τ))|2dτ|=0. (47)

    From (45) and (47) it follows that

    I2(t,ε)3(NL)212νsuptRE|ϕε(t)ˉϕ(t)|2+B(ε), (48)

    where B(ε) is some positive constant such that B(ε)0 as ε0.

    Combing (34), (44) and (48), we have

    (13(NL)2(2ν2+1ν))suptRE|ϕε(t)ˉϕ(t)|22(A(ε)+B(ε)).

    Consequently

    limε0suptRE|ϕε(t)ˉϕ(t)|2=0

    because 13(NL)2(2ν2+1ν)>0.

    To finish the proof of the theorem we note that L2-convergence implies convergence in distribution, so

    limε0suptRβ(L(ϕε(t)),L(ˉϕ(t)))=0.

    Since L(φε(tε))=L(ϕε(t)), we get

    limε0suptRβ(L(φε(tε)),L(ˉϕ(t)))=0.

    The proof is complete.

    Corollary 3. Under the conditions of Theorem 4.2 the following statements hold:

    1. if the functions AC(R,L(H)) and F,G C(R×H,H) are jointly stationary (respectively, τ-periodic, quasi-periodic with the spectrum of frequencies ν1,,νk, Bohr almost periodic, almost automorphic, Birkhoff recurrent, Lagrange stable, Levitan almost periodic, almost recurrent, Poisson stable), then equation (30) has a unique solution φεCb(R,L2(P,H)) which is stationary (respectively, τ-periodic, quasi-periodic with the spectrum of frequencies ν1,,νk, Bohr almost periodic, almost automorphic, Birkhoff recurrent, Lagrange stable, Levitan almost periodic, almost recurrent, Poisson stable) in distribution;

    2. if the functions AC(R,L(H)) and F,G C(R×H,H) are Lagrange stable and jointly pseudo-periodic (respectively, pseudo-recurrent), then equation (30) has a unique solution φεCb(R,L2(P,H)) which is pseudo-periodic (respectively, pseudo-recurrent) in distribution;

    3.

    limε0suptRβ(L(φε(tε),L(ˉϕ(t)))=0,

    with ˉϕ being the unique stationary solution of the averaged equation (33).

    Proof. This statement follows from Theorems 2.12 and 4.2 (see also Remark 2).

    Remark 15. To simplify the notations and highlight the idea, we consider only the one-dimensional noise. Indeed, the main results of this paper remain hold if we replace the one-dimensional Brownian motion W in Sections 3 and 4 by a Q-Wiener process, which brings no essential but just notational difference; see e.g. [20,21] for details.

    Remark 16. In the present paper, we only consider the second Bogolyubov theorem for semilinear stochastic ordinary differential equations, i.e. the linear part A() is bounded operator valued. We will consider the case when A() is an unbounded operator in future work, which can be applied to related stochastic partial differential equations.

    We would like to thank Professor A. J. Roberts for drawing our attention to stochastic centre manifold approach for averaging and to references [3,22]. We are grateful to the anonymous referees for their careful reading of our paper and valuable suggestions.



    [1] N. N. Bogolyubov, On Some Statistical Methods in Mathematical Physics, Akademiya Nauk Ukrainskoǐ SSR, 1945. (in Russian)
    [2] N. N. Bogolyubov and Y. A. Mitropolsky, Asymptotic Methods in the Theory of Non-Linear Oscillations, Translated from the second revised Russian edition. International Monographs on Advanced Mathematics and Physics Hindustan Publishing Corp., Delhi, Gordon and Breach Science Publishers, New York, 1961.
    [3] A stochastic version of center manifold theory. Probab. Theory Related Fields (1989) 83: 509-545.
    [4] Averaging principle for a class of stochastic reaction-diffusion equations. Probab. Theory Related Fields (2009) 144: 137-177.
    [5] Averaging principle for nonautonomous slow-fast systems of stochastic reaction-diffusion equations: the almost periodic case. SIAM J. Math. Anal. (2017) 49: 2843-2884.
    [6] D. N. Cheban, Asymptotically Almost Periodic Solutions of Differential Equations, Hindawi Publishing Corporation, New York, 2009.
    [7] D. N. Cheban, Global Attractors of Nonautonomous Dynamical and Control Systems, 2nd edition, Interdisciplinary Mathematical Sciences, vol.18, River Edge, NJ: World Scientific, 2015. doi: 10.1142/9297
    [8] D. Cheban and J. Duan, Recurrent motions and global attractors of non-autonomous Lorenz systems, Dyn. Syst., 19 (2004), 41–59. doi: 10.1080/14689360310001624132
    [9] Periodic, quasi-periodic, almost periodic, almost automorphic, Birkhoff recurrent and Poisson stable solutions for stochastic differential equations. J. Differential Equations (2020) 269: 3652-3685.
    [10] Ju. L. Dalec'kiǐ and M. G. Kreǐn, Stability of Solutions of Differential Equations in Banach Space, Translated from the Russian by S. Smith. Translations of Mathematical Monographs, Vol. 43. American Mathematical Society, Providence, R.I., 1974.
    [11] (2014) Stochastic Equations in Infinite Dimensions. Cambridge: 2nd edition, Encyclopedia of Mathematics and its Applications, 152. Cambridge University Press.
    [12] J. Duan and W. Wang, Effective Dynamics of Stochastic Partial Differential Equations, Elsevier Insights. Elsevier, Amsterdam, 2014. doi: 10.1016/B978-0-12-800882-9.00001-9
    [13] (2002) Real Analysis and Probability. Cambridge: Revised reprint of the 1989 original. Cambridge Studies in Advanced Mathematics, 74. Cambridge University Press.
    [14] M. I. Freidlin and A. D. Wentzell, Random Perturbations of Dynamical Systems, Translated from the 1979 Russian original by Joseph Szücs. 3rd edition. Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences], 260. Springer, Heidelberg, 2012. doi: 10.1007/978-3-642-25847-3
    [15] R. Z. Has'minskiǐ, On the principle of averaging the itô's stochastic differential equations, Kybernetika (Prague), 4 (1968), 260–279. (in Russian)
    [16] Weak averaging of semilinear stochastic differential equations with almost periodic coefficients. J. Math. Anal. Appl. (2015) 427: 336-364.
    [17] M. A. Krasnosel'skiǐ, V. Sh Burd and Yu. S. Kolesov, Nonlinear Almost Periodic Oscillations, Nauka, Moscow, 1970 (in Russian). [English translation: Nonlinear Almost Periodic Oscillations. A Halsted Press Book. New York etc.: John Wiley & Sons; Jerusalem- London: Israel Program for Scientific Translations. IX, 326 p., 1973]
    [18] (1943) Introduction to Non-Linear Mechanics. Princeton, N. J.: Annals of Mathematics Studies, no. 11. Princeton University Press.
    [19] (1978) Almost Periodic Functions and Differential Equations. Moscow: Moscow State University Press.
    [20] X. Liu and Z. Liu, Poisson stable solutions for stochastic differential equations with Lévy noise, Acta Math. Sin. (Engl. Ser.), In Press. (also available at arXiv: 2002.00395)
    [21] Almost automorphic solutions for stochastic differential equations driven by Lévy noise. J. Funct. Anal. (2014) 226: 1115-1149.
    [22] Normal form transforms separate slow and fast modes in stochastic dynamical systems. Phys. A (2008) 387: 12-38.
    [23] B. A. Ščerbakov, A certain class of Poisson stable solutions of differential equations, Differencial'nye Uravnenija, 4 (1968), 238–243. (in Russian)
    [24] B. A. Ščerbakov, The comparability of the motions of dynamical systems with regard to the nature of their recurrence, Differencial'nye Uravnenija, 11 (1975), 1246–1255. (in Russian) [English translation: Differential Equations 11 (1975), 937–943].
    [25] G. R. Sell, Topological Dynamics and Ordinary Differential Equations, Van Nostrand-Reinhold, 1971.
    [26] B. A. Shcherbakov, Topologic Dynamics and Poisson Stability of Solutions of Differential Equations, Stiinca, Kishinev, 1972.
    [27] B. A. Shcherbakov, Poisson Stability of Motions of Dynamical Systems and Solutions of Differential Equations, Știinţa, Chişinǎu, 1985. (in Russian)
    [28] K. S. Sibirsky, Introduction to Topological Dynamics, Kishinev, RIA AN MSSR, 1970. (in Russian). [English translation: Introduction to Topological Dynamics. Noordhoff, Leyden, 1975]
    [29] A. V. Skorokhod, Asymptotic Methods in the Theory of Stochastic Differential Equations, Translated from the Russian by H. H. McFaden. Translations of Mathematical Monographs, 78. American Mathematical Society, Providence, RI, 1989. doi: 10.1090/mmono/078
    [30] On large deviations in the averaging principle for SDEs with a "full dependence''. Ann. Probab. (1999) 27: 284-296.
    [31] Weak averaging of stochastic evolution equations. Math. Bohem. (1995) 120: 91-111.
    [32] Average and deviation for slow-fast stochastic partial differential equations. J. Differential Equations (2012) 253: 1265-1286.
  • This article has been cited by:

    1. Mengyu Cheng, Zhenxin Liu, The Second Bogolyubov Theorem and Global Averaging Principle for SPDEs with Monotone Coefficients, 2023, 55, 0036-1410, 1100, 10.1137/21M1443698
    2. Mengyu Cheng, Zhenxin Liu, Michael Röckner, Averaging principle for stochastic complex Ginzburg-Landau equations, 2023, 368, 00220396, 58, 10.1016/j.jde.2023.05.031
    3. Ye-Jun Chen, Hui-Sheng Ding, Remotely almost periodicity for SDEs under the framework of evolution system, 2023, 23, 0219-4937, 10.1142/S0219493723500338
    4. David Cheban, Averaging principle on semi‐axis for semi‐linear differential equations, 2024, 0025-584X, 10.1002/mana.202300392
    5. Ye-Jun Chen, Asymptotically almost periodic solutions for stochastic evolution equations with Lévy noise, 2025, 0, 1534-0392, 0, 10.3934/cpaa.2025058
  • Reader Comments
  • © 2021 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(2962) PDF downloads(376) Cited by(5)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog