Applied Mathematics and Computation 218 (2011) 14991511

Contents lists available at ScienceDirect

Applied Mathematics and Computation

journal homepage: [Link]/locate/amc

Asymptotically almost periodic solutions of stochastic functional

differential equations
Junfei Cao a, Qigui Yang a,, Zaitang Huang a,b, Qing Liu a
a
b

Department of Mathematics, School of Science, South China University of Technology, Guangzhou 510640, Peoples Republic of China
School of Mathematics and Physics, Wuzhou University, Wuzhou 543002, Peoples Republic of China

a r t i c l e

i n f o

Keywords:
Asymptotically almost periodic solution
Stochastic functional differential equations
Exponential stability

a b s t r a c t
In this paper, we investigate a class of stochastic functional differential equations of the
form

dxt Axt Ft; xt; xt dt Gt; xt; xt  dWt:

Our main results concern the existence and exponential stability of quadratic-mean
asymptotically almost periodic mild solutions. An example is given to illustrate our results.
 2011 Elsevier Inc. All rights reserved.

1. Introduction
Founded between 1923 and 1926 by Bohr, the theory of almost periodicity had led to strong development of harmonic
analysis on groups and compact topological semigroups of linear operators. The theory had attracted many mathematicians
for decades. Some generalizations of the concept had been introduced successfully by Stepanov and Besicovitch. Almost periodic functions with values in Banach space were studied by Bochner and developed by several mathematicians including
Corduneanu, Fink, Goldstein, Prouse, Glicksberg, Deleeuw, Amerio, Zaidman, etc. Applications include ordinary, partial as
well as abstract differential equations, topological and smooth dynamical systems, statistics, etc. [19].
The concept of almost periodicity is also important in probability for investigating stochastic processes, and it is interesting for applications arising in mathematical physics and statistics. The concept of almost periodicity for stochastic processes
was rst introduced at the end of the 1930s by Slutsky [10] and developed by several mathematicians including Udagawa
[11], Kawata [12], Swift [13], etc.
Recently, in [14], the concept of quadratic-mean almost periodicity was introduced and studied by Bezandry and Diagana.
In particular, such a concept was utilized to study the existence and uniqueness of quadratic-mean almost periodic solution
to a class of stochastic differential equations of the form

dxt Axtdt Ft; xtdt Gt; xt  dWt;

t 2 R:

More recently, Huang and Yang [15] presented some new criteria ensuring the existence and uniqueness of quadratic-mean
almost periodic solution as well as global exponential stability of the almost periodic solution for stochastic cellular neural
networks with delay of the form

dxt Axt Bf t; xt Cf t; xt  rdt Gt; xt  dWt;

t 2 R:

These criteria are important in signal processing and the design of networks. As other work we refer the readers to [1620]
and the references therein.
Corresponding author.
E-mail addresses: jfcaomath@[Link] (J. Cao), qgyang@[Link] (Q. Yang), zaitanghuang@[Link] (Z. Huang), bnumlq@[Link] (Q. Liu).
0096-3003/$ - see front matter  2011 Elsevier Inc. All rights reserved.
doi:10.1016/[Link].2011.06.033

1500

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

The asymptotically almost periodic functions were rst introduced by Frchet. In the modern theory of differential equations, many authors [35] applied successfully the asymptotic property to determine the existence of almost periodic solutions. Along with the development of such equations as the evolution partial differential equations, functional differential
equations, and so forth, where the phase spaces are innite, the theory of Banach valued asymptotically almost periodic
functions had been developed [2125]. Some techniques in functional analysis and harmonic analysis were applied to such
equations, for example, in [1,2], the authors applied spectrum theory to get almost periodic solutions for some linear abstract
evolution differential equations.
In the present paper, we introduce and develop the notion of p-mean asymptotically almost periodicity for stochastic processes. Among others, it will be shown that each p-mean asymptotically almost periodic stochastic process is stochastically
bounded, and the collection of all p-mean asymptotically almost periodic stochastic processes is a Banach space when it is
equipped with some norm obtained through the norm of Lp P; B, where P is the probability measure of the probability
space X; F; P and B; k  k is a Banach space. The previous basic results are subsequently applied to study the existence
and uniqueness of quadratic-mean asymptotically almost periodic mild solution to stochastic functional differential equations on L2 P; H; under a hyperbolic and analytic semigroup {T(t)}tP0, where H is a Hilbert space. Furthermore, we prove
that the asymptotically almost periodic solution is global exponentially stable provided the semigroup {T(t)}tP0 is exponentially stable.
The rest of this paper is organized as follows. In Section 2, some concepts, the related notations and some useful lemmas
are introduced. In Section 3, we present some criteria ensuring the existence and uniqueness of a quadratic-mean asymptotically almost periodic mild solution. In Section 4, we prove some results on exponential stability of the asymptotically
almost periodic mild solution. An example is given to illustrate our results in Section 5.
2. Preliminaries
This section is concerned with some notations, denitions, lemmas and preliminary facts which are used in what follows.
Let X; F; P be a probability space. For a Banach space B; k  k and p P 1, denote by Lp P; B the Banach space of all Bvalue random variable Y such that

kYkLp P;B

Z

kYkp dP

1p
< 1:

The following useful Denitions 2.12.3 come from [14].

Denition 2.1
(1) A stochastic process X : R ! Lp P; B is said to be continuous provided that for any s 2 R;

lim EkXt  Xskp 0:

t!s

(2) A stochastic process X : R ! Lp P; B is said to be stochastically bounded provided that

lim sup PfkXtk > Ng 0:

N!1 t2R

By CUBR; Lp P; B we denote the collection of all continuous and uniformly bounded stochastic processes from R to
Lp P; B: It is easy to check that CUBR; Lp P; B is a Banach space when it is equipped with the norm

n
o
1
kXk1 sup EkXtkp p ;
t2R

see, e.g. [14].

Denition 2.2. A continuous stochastic process X : R ! Lp P; B is said to be p-mean almost periodic provided that for each
e > 0, the set

JX; e : fs : sup EkXt s  Xtkp < eg

t2R

is relatively dense in R, i.e., there exists a constant l = l(e) > 0 such that J(X, e) \ [t, t + l] ; for any t 2 R.
By APR; Lp P; B we denote the set of all p-mean almost periodic stochastic processes, it follows from Lemma 2.5 of [14]
that APR; Lp P; B is a closed subspace of CUBR; Lp P; B: Therefore, APR; Lp P; B is a Banach space when it is equipped
with the norm k  k1 (see, e.g. [14]).
Let B1 ; k  k1 ; B2 ; k  k2 be two Banach spaces, and Lp P; B1 ; Lp P; B2 be their corresponding Lp-spaces respectively.
Denition 2.3. A continuous function F : R  Lp P; B1 ! Lp P; B2 is called p-mean almost periodicity in t, uniformly for x in
compact subset K of Lp P; B1 ; provided that for each e > 0, the set

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

JX; K; e :

1501

s : sup EkFt s; x  Ft; xkp2 < e

t2R

is relatively dense in R; i.e., there exists a constant l = l(e) > 0 such that JX; K; e \ t; t l ; for any t 2 R.
The collection of such functions will be denote by APR  Lp P; B1 ; Lp P; B2 :
The space of the restrictions of all p-mean almost periodic stochastic processes (p-mean almost periodic function in t, uniformly for x in compact subset K of Lp P; B1 ) on R is denoted by APR ; Lp P; BAPR  Lp P; B1 ; Lp P; B2 .
Denote by C 0 R ; Lp P; B the space of all continuous stochastic processes Z : R ! Lp P; B such that
limt!1 EkZtkp 0, and denote by C 0 R  Lp P; B1 ; Lp P; B2 the space of all continuous functions Z : R  Lp P; B1
! Lp P; B2 such that limt!1 EkZt; xkp2 0; uniformly for x in any compact subset of Lp P; B1 .
Now, we introduce a new concept called p-mean asymptotically almost periodic stochastic process.
Denition 2.4. A stochastic process X : R ! Lp P; B is called p-mean asymptotically almost periodicity if there exist two
stochastic processes Y 2 APR ; Lp P; B and Z 2 C 0 R ; Lp P; B such that X = Y + Z.
By AAPR ; Lp P; B we denote the collection of all p-mean asymptotically almost periodic stochastic processes.
Denition 2.5. A continuous function X : R  Lp P; B1 ! Lp P; B2 is called p-mean asymptotically almost periodicity in t,
uniformly for x in compact subset K of Lp P; B1 ; if there exist two continuous functions Y 2 APR  Lp P; B1 ; Lp P; B2 and
Z 2 C 0 R  Lp P; B1 ; Lp P; B2 such that X = Y + Z.
By AAPR  Lp P; B1 ; Lp P; B2 we denote the collection of such functions.
Lemma 2.1 (See [14]). If X belongs to APR; Lp P; B; then
(i) the mapping t ! EkXtkp is uniformly continuous;
(ii) there exists a constant M > 0 such that EkXtkp 6 M for each t 2 R;
(iii) X is stochastically bounded.
Lemma 2.2. If X belongs to AAPR ; Lp P; B; then we have
(i) there exists a constant M > 0 such that EkXtkp 6 M for each t 2 R ;
(ii) X is stochastically bounded.
Proof
(i) Since X 2 AAPR ; Lp P; B; by Denition 2.4, there exist Y 2 APR ; Lp P; B and Z 2 C 0 R ; Lp P; B such that
X = Y + Z. According to Lemma 2.1, there exists a constant M1 > 0 such that

EkYtkp 6 M 1

for each t 2 R :
p

Since limt!1 EkZtk 0; then for some given e > 0, there exists a constant T > 0 such that

EkZtkp < e for each t 2 T; 1:

From Lemma 2.1, it follows that EkZtkp is uniformly continuous on [0, T], therefore there exists a constant M2 > e such that

EkZtkp 6 M 2

for each t 2 0; T:

Let M = 2 (M1 + M2), thus

EkXtkp 6 2p EkYtkp kZtkp 6 2p M 1 M 2 M:

(ii) To prove (ii), using Chebychev inequality, one obtains

!
p

sup PfkXtk P Ng 6

sup EkXtk

t2R

t2R

=N P 6 M=NP ;

which implies

sup PfkXtkp P Ng ! 0 as N ! 1:

t2R

Remark 2.1. It is easy to see that AAPR ; Lp P; B APR ; Lp P; B  C 0 R ; Lp P; B. Then the space AAPR ; Lp P; B of
p-mean asymptotically almost periodic stochastic processes is a Banach space when it is equipped with the norm kk1.

1502

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

Lemma 2.3. Assume that X 2 AAPR ; Lp P; B admits a decomposition X = Y + Z, where Y 2 APR ; Lp P; B and Z 2
C 0 R ; Lp P; B. Then

fYt : t 2 R g  fXt : t 2 R g:
Proof. If fYt : t 2 R g is not a subset of fXt : t 2 R g; then there exists t0 2 R such that inf s2R kYt0  XskLp P;B > e:
Since Y(t) is continuous at t0, there exists d > 0 such that for 0 < t < d,

inf kYt t0  Xs t 0 kLp P;B > e:

s2R

Since Y(t + t0) is in APR ; Lp P; B, by Denition 2.2, for the e above, there exists a constant l
length l 2e contains at least s satisfying

kYs t t0  Yt t0 kLp P;B <

e
2

e
2

> 0 such that any interval of

for every t 2 R :

If t 2 (0, d), then t + s 2 (s, s + d). Therefore

kZt t 0 skLp P;B kXt t 0 s  Yt t0 skLp P;B

P kXt t 0 s  Yt t 0 kLp P;B  kYt t 0  Yt t0 skLp P;B
P inf kXs t 0  Yt t 0 kLp P;B  kYt t0  Yt t0 skLp P;B >
s2R

e
2

This means that limt!1 kZt t 0 skLp P;B > 0, a contradiction. h

Lemma 2.4. Let X 2 AAPR ; Lp P; B, then the range RX fXt : t 2 R g is relatively compact in Lp P; B.
Proof. Suppose that X = Y + Z, where Y 2 APR ; Lp P; B and Z 2 C 0 R ; Lp P; B. Taking a sequence ftn g  R , if {tn} is
bounded, then there exists a subsequence {tk} and t0 P 0 such that limk?+1tk = t0. Since Z is continuous, one has
limk?+1Z(tk) = Z(t0). If {tn} is unbounded, then there exists a subsequence {tm} such that limm?+1tm = 1. So limm?+1Z(tm) = 0.
Thus, RZ fZt : t 2 R g is relatively compact in Lp P; B. Recalling that RY fYt : t 2 R g is relatively compact in
Lp P; B, we conclude RX fXt : t 2 R g is relatively compact in Lp P; B. h
Denote by Cr; 0; Lp P; B1 the space of all continuous functions from [r, 0] into Lp P; B1 with the sup-norm

kwkCr;0;Lp P;B1 supfkwskLp P;B1 : w 2 C; r 6 s 6 0g:

Let K  Lp P; B1  Cr; 0; Lp P; B1 ; T  R. Denote by C K T  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2 the set of all the
functions F : T  Lp P; B1  Cr; 0; Lp P; B1 ! Lp P; B2 satisfying F(t, , ) is uniformly continuous on K uniformly for
t 2 T, i.e., "e > 0, $d > 0 such that for all (x1, y1), (x2, y2) 2 K with

kx1  x2 kpLp P;B1 ky1  y2 kpCr;0;Lp P;B1 < d

and all t 2 T,

kFt; x1 ; y1  Ft; x2 ; y2 kpLp P;B2 < e:

In the proof of our main results, the following useful lemmas will be needed.
Lemma 2.5. Let x 2 APR ; Lp P; B1 ; K fxt; xt : t 2 R g, where xt = {x(t + h) : r < h < 0} is regarded as a Cr; 0;
Lp P; B1 -valued stochastic process. Moreover

F 2 APR  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2

and

F 2 C K R  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2 :
Then F; x; x 2 APR ; Lp P; B2 .
Proof. From x 2 APR ; Lp P; B1 and

F 2 APR  Lp P; B1  Lp P; B1 ; Lp P; B2 ;
K is compact in Lp P; B1  Lp P; B1 , it follows that for each e > 0, there exists a constant l(e) > 0 such that any interval of
length l(e) contains at least s satisfying

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

Ekxt s  xtkp1 < e;

1503

8t 2 R ;

EkFt s; x; y  Ft; x; ykp2 < e=2p ;

8t 2 R ;

x; y 2 K:

Thus one has

kxt s  xtkpLp P;B1 kxts  xt kpCr;0;Lp P;B1 < 2e;

p

8t 2 R :

Since F 2 C K R  L P; B1  L P; B1 ; L P; B2 , then "e > 0, $d > 0 such that for all (x1, y1), (x2, y2) 2 K with

kx1  x2 kpLp P;B1 ky1  y2 kpLp P;B1 < d

and all t 2 T  R ,

kFt; x1 ; y1  Ft; x2 ; y2 kpLp P;B2 < e:

Let d = 2e, one has

kFt s; xt s; xts  Ft s; xt; xt kpLp P;B2 < e=2p ;

for 8t 2 R ; x; y 2 K. Therefore,

EkFt s; xt s; xts  Ft; xt; xt kp2

1p


1
6 EkFt s; xt s; xts  Ft s; xt; xt kp2 p

1
EkFt s; xt; xt  Ft; xt; xt kp2 p
<

e
2

e
2

e:

Lemma 2.6. Let x 2 AAPR ; Lp P; B1 ; K fxt; xt : t 2 R g, where xt = {x(t + h) : r < h < 0} is regarded as a Cr; 0;
Lp P; B1 -valued stochastic process. Moreover

F 2 AAPR  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2

and

F 2 C K R  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2 :
Then F; x; x 2 AAPR ; Lp P; B2 .
Proof. From x 2 AAPR ; Lp P; B1 and

F 2 AAPR  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2 ;

it follows that there exist

k 2 APR  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2 ;

h 2 C 0 R  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2
and

y 2 APR ; Lp P; B2 ;

z 2 C 0 R ; Lp P; B

such that F = k + h and x = y + z. Then,

Ft; xt; xt Ft; xt; xt  Ft; yt; yt kt; yt; yt ht; yt; yt I1 t I2 t I3 t;
where

I1 Ft; xt; xt  f t; yt; yt ;

I2 kt; yt; yt ;

I3 ht; yt; yt :

By Lemmas 2.3 and 2.4, K is compact and (y(t), yt) 2 K for all t 2 R . Combining

lim Ekztkp lim Ekut  ytkp 0

t!1

t!1

with F(t, y(t), yt) is uniformly continuous on K uniformly for t 2 R , one gets

lim EkI1 tk 0:

t!1

From the denition of C 0 R  Lp P; B1 ; Lp P; B2 , it is not difcult to show that

h 2 C K R  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2 :

1504

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

Thus

k 2 C K R  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2 :
Let K 1 fyt; yt : t 2 R g, then K1  K is compact and

k 2 C K 1 R  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2 :
By Lemma 2.5, one has

I2 t 2 APR ; Lp P; B2 :
For I3, from

h 2 C 0 R  Lp P; B1  Cr; 0; Lp P; B1 ; Lp P; B2 ;
it follows that

lim EkI3 tk 0:

t!1

Hence, we have Ft; ut; ut 2 AAPR ; Lp P; B2 . h

3. Existence of asymptotically almost periodic solutions
For convenience, throughout this section, let H; k  k be a real separable Hilbert space, and X; F ; P be a complete probability space equipped with a normal ltration fF t : t 2 Rg; that is, a right-continuous, increasing family of sub r-algebras
of F : Let LL2 P; H be the space of all bounded linear operators from L2 P; H into L2 P; H; and C Cr; 0; L2 P; H be
the space of all continuous functions from [r, 0] into L2 P; H with the sup-norm

kwkC supfkwskL2 P;H : w 2 C; r 6 s 6 0g:

For any continuous F t -adapted L2 P; H-valued stochastic process xt : X ! L2 P; H; t P r; we associate with a continuous F t -adapted C-valued stochastic process xt : X ? C, t P 0, by setting xt(s)(x) = x(t + s)(x), s 2 [r, 0].
In this section, we study the existence and uniqueness of quadratic-mean asymptotically almost periodic mild solution to
stochastic functional differential equations of the form

dxt Axt Ft; xt; xt dt Gt; xt; xt  dWt;

t 2 R;

3:1

where A : DA  L2 P; H ! L2 P; H generates a strongly continuous semigroup {T(t)}tP0,W(t) is a certain Q-Wiener process

with covariance operator Q and takes its values on L2 P; H; F : R  L2 P; H  C ! L2 P; H and G : R  L2 P; H  C !
L2 P; LL2 P; H are two continuous mappings.
Throughout this section, we require the following assumptions.
(H1) The semigroup {T(t)}tP0 is hyperbolic, i.e., there exist a projection P and constants M, M0, d > 0 such that each T(t) commutes with P, ker(P) is invariant with respect to Tt; Tt : RQ ! RQ is invertible and

kTtPxk 6 M expdtkxk;

for t > 0;

kTtQxk 6 M 0 expdtkxk;

for t 6 0;

where Q I  P; kerA; RA stand for the kernel, range of A, and for t 6 0, T(t) = (T(t))1;
(H2) F 2 APR  L2 P; H  C; L2 P; H: Furthermore, there exists a constant K1 > 0 such that



k2L2 P;H 6 K 1 kx  xk2L2 P;H ky  y
k2C ;
kFt; x; y  Ft; x; y

 2 L2 P; H  C; t 2 R;
for all stochastic processes x; y; 
x; y
(H3) G 2 APR  L2 P; H  C; L2 P; LL2 P; H: Furthermore, there exists a constant K2 > 0 such that



k2L2 P;LL2 P;H 6 K 2 kx  xk2L2 P;H ky  y
k2C ;
kGt; x; y  Gt; x; y

 2 L2 P; H  C; t 2 R:
for all stochastic processes x; y; 
x; y
Remark 3.1. It is easy to see that (H2) and (H3) imply F 2 C K R  L2 P; H  C; L2 P; H and G 2 C K R  L2 P; H
C; L2 P; LL2 P; H:

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

1505

Denition 3.1. An F t -progressively process fxtgt2R is called a mild solution of Eq. (3.1) on R if

xt Tt  sxs

Tt  rFr; xr; xr dr

Tt  rGr; xr; xr  dWr;

3:2

for all t P s, for each s 2 R:

In the following, we state our rst main result.
Theorem 3.1. Let (H1), (H2) and (H3) be satised. Then Eq. (3.1) has a unique quadratic-mean asymptotically almost periodic mild
solution, which can be explicitly expressed as follows

xt

Tt  sPFs; xs; xs ds

1

Tt  sPGs; xs; xs  dWs 

Tt  sQFs; xs; xs ds

1

Tt  sQGs; xs; xs  dWs

3:3

for each t 2 R, provided

s


2K 1 K 2
H 2 M 2 M 20

< 1:
d
d2
Proof
Step 1. We show that Eq. (3.1) exists a mild solution. Note that (3.3) is well dened for each t 2 R; and satises (3.2) for all
t P s, for each s 2 R, hence x(t) given by (3.3) is a mild solution of Eq. (3.1).
Step 2. We construct the following operator on CR ; L2 P; H

Lxt :

Tt  sPFs; xs; xs ds

1

Tt  sPGs; xs; xs  dWs 

1

Tt  sQFs; xs; xs ds

Tt  sQGs; xs; xs  dWs

and further show that L maps AAPR; L2 P; H into itself.

Dene

/1 xt :

Tt  sPFs; xs; xs ds;

1

/2 xt :

Tt  sQFs; xs; xs ds;

t

w1 xt :

Tt  sPGs; xs; xs  dWs;

1

w2 xt :

Tt  sQGs; xs; xs  dWs:

t

First, we show that /1(x()) and /2(x()) are quadratic-mean asymptotically almost periodic whenever x is.
In fact, assuming that x is quadratic-mean asymptotically almost periodic, using (H2) and Lemma 2.6, one can easily see
that s F(s, x(s), xs) is quadratic-mean asymptotically almost periodic. Therefore, there exist an almost periodic function
k(s, x(s), xs) and a function h 2 C 0 R  L2 P; H; L2 P; H such that F = k + h. Furthermore, one observes that

/1 xt

Tt  sPks; xs; xs ds

1

Tt  sPhs; xs; xs ds at bt:

1

We claim that at 2 APR; L2 P; H. In fact, for each e > 0, there exists a constant l(e) > 0 such that any interval of length l(e)
contains at least s satisfying

Ekks s; xs s; xss  ks; xs; xs k2 <

for each s 2 R: Then

d2 e
M2

3:4

1506

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

Z

kat s  atk 


ts

Tt s  sPks; xs; xs ds 

1

1



Tt  sPks; xs; xs ds


Z t

Z t




Tt

sPks

s
;
xs

s
;
x
ds

Tt

sPks;
xs;
x
ds
ss
s


1
1
Z t
6
kTt  sPks s; xs s; xss  ks; xs; xs kds
1

6M

edts kks s; xs s; xss  ks; xs; xs kds;

1

and hence, using CauchySchwarz inequality and (3.3), one has

Z
Ekat s  atk2 6 M 2 E
Z
M2 E

2
edts kks s; xs s; xss  ks; xs; xs kds

1
t

e

1

Z
6 M2 E
M

Z

e

dts
2

dts

Z
ds

d2


edts kks s; xs s; xss  ks; xs; xs k2 ds

1
t

edts Ekks s; xs s; xss  ks; xs; xs k2 ds

1

2
edts ds sup Ekks s; xs s; xss  ks; xs; xs k2
t2R

1

M2

2
kks s; xs s; xss  ks; xs; xs kds

Z
edts ds

1

Z

1

6 M2

dts
2

sup Ekks s; xs s; xss  ks; xs; xs k2 < e:

t2R

Next, let us show that bt 2 C 0 0; 1; L2 P; H. In fact

Z

Ekbttk2 E

2

t
1

M E
M2

2
Z

2
Tt  sPhs; xs; xs ds
 6M E

Z

Z

d2

dts
 2

khs; xs; xs kds

1
t

1

M2

dts
 2

Z
edts ds

2

2
edts khs; xs; xs kds

1

6 M2 E

Z

Z
edts ds

1


Z
edts Ekhs; xs; xs k2 ds 6 M 2

1


edts khs; xs; xs k2 ds

1

1

2
edts ds sup Ekhs; xs; xs k2
t2R

sup Ekhs; xs; xs k2 :

t2R

Since h 2 C 0 0; 1; L2 P; H, then we deduce that limt!1 Ekbtk2 0.

Therefore, /1(x()) is quadratic-mean asymptotically almost periodic.
Similar to the proof given for /1(x()), one can prove that /2(x()) is quadratic-mean asymptotically almost periodic.
Secondly, we show that w1(x()) and w2(x()) are also quadratic-mean asymptotically almost periodic whenever x is.
Of course, this is more complicated than the previous case because of the involvement of the Brownian motion W. To
f dened by
overcome such a difculty, we make extensive use of the Ito,s isometry identity and the properties of W
f
f
W : Ws s  Ws; for each s. Note that W is also a Brownian motion and has the same distribution as W. Assuming that
x is quadratic-mean asymptotically almost periodic, using (H3) and Lemma 2.6, one can easily see that s G(s, x(s), xs) is
quadratic-mean asymptotically almost periodic. Therefore, there exist an almost periodic stochastic process u(s, x(s), xs) and a
stochastic process v 2 C 0 0; 1  L2 P; H; L2 P; H such that G = u + v. Furthermore, one observes that

w1 xt

Tt  sPus; xs; xs  dWs

1

Tt  sPv s; xs; xs  dWs Ut Vt:

1

We claim U 2 APR; L2 P; H. In fact, for each e > 0, there exists a constant l(e) > 0 such that any interval of length l(e)
contains at least s satisfying

Ekus s; xs s; xss  us; xs; xs k2 <

for each s 2 R: Then

d2 e
M2

3:5

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

Ut s  Ut

ts

Tt s  sPus; xs; xs  dWs 

Tt  sPus; xs; xs  dWs

1

1
t

Tt  sPus s; xs s; xss  dWs 

1
Z t

Tt  sPus; xs; xs  dWs

1

Tt  sPus s; xs s; xss  us; xs; xs   dWs:

^s isometry identity that

Let s = l  s, it follows from It o

2
Z t


f s
EkUt s  Utk2 E 
Tt  sPus s; xs s; xss  us; xs; xs   d W


1
Z t

E
kTt  sPus s; xs s; xss  us; xs; xs k2 ds
1

6M

6 M2

e2dts Ekus s; xs s; xss  us; xs; xs k2 ds

1
Z t


e2dts ds sup Ekus s; xs s; xss  us; xs; xs k2
s2R

1

M2

sup Ekus s; xs s; xss  us; xs; xs k2 < e:

2d s2R
Next, let us show that V 2 C 0 0; 1; L2 P; H. In fact

Z t
2
2
Z t




f s
 E 
EkVtk2 E
Tt

sP
v
s;
xs;
x
ds
Tt

sP
v
s;
xs;
x


d
W
s
s




1
1
Z t
6 M2
e2dts Ekus s; xs s; xss  us; xs; xs k2 ds
6M

1
t

Z


e2dts ds sup Ekus s; xs s; xss  us; xs; xs k2
s2R

1
2

M
sup Ekus s; xs s; xss  us; xs; xs k2 :
2d s2R

Since v 2 C 0 0; 1; L2 P; H, then we deduce that limt!1 EkVtkp 0.

Therefore, /1(x()) is quadratic-mean asymptotically almost periodic.
Similarly, we can prove that w2(x()) is also quadratic-mean asymptotically almost periodic.
In view of the above, it is clear that L maps AAPR; L2 P; H into itself.
Step 3. We prove that L is a contraction mapping and has a unique xed-point.
Clearly,

Z

kLxt  Lytk 


Tt  sPFs; xs; xs  Fs; ys; ys ds 

1

Tt  sQ Fs; xs; xs  Fs; ys; ys ds

Tt  sPGs; xs; xs  Gs; ys; ys  dWs

1



Tt  sQ Gs; xs; xs  Gs; ys; ys  dWs


edts kFs; xs; xs  Fs; ys; ys kds M 0

6M

1

edst kFs; xs; xs  Fs; ys; ys kds

1



Tt  sPGs; xs; xs  Gs; ys; ys   dWs




Tt  sQ Gs; xs; xs  Gs; ys; ys   dWs
:

Since (a + b + c + d)2 6 4a2 + 4b2 + 4c2 + 4d2, one can write

1507

1508

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

Z
EkLxt  Lytk2 6 4M 2 E

2
edts kFs; xs; xs  Fs; ys; ys kds

1

Z
4M 20 E

2
edst kFs; xs; xs  Fs; ys; ys kds

2
Z t



4E 
Tt

sPGs;
xs;
x


Gs;
ys;
y


dWs
s
s


1
2
Z 1


4E 
Tt  sQ Gs; xs; xs  Gs; ys; ys   dWs


t

U1 U2 U3 U4 :
We rst evaluate U1 as follows

Z
U 1 4M 2 E

2
edts kFs; xs; xs  Fs; ys; ys kds

t
1

Z
6 4M2 E
4M

Z
edts ds

1

Z

t
dts

1

6 4M2 K 1

Z

Z
ds


edts kFs; xs; xs  Fs; ys; ys k2 ds

1
t


edts EkFs; xs; xs  Fs; ys; ys k2 ds

1

2
Z


edts ds sup kxs  ysk2L2 P;H kxs  ys k2C 6 8M 2 K 1

s2R

1

1

2
edts ds kxs  ysk21

8M K 1

kxs 

d2

ysk21 :

Similar to the discussion given for U1, for U2, one has

Z
U 2 4M 20 E

1
t

2
8M 20 K 1
edst kFs; xs; xs  Fs; ys; ys kds 6
kxs  ysk21 :
d2

^s isometry identity, one obtains

For U3, using It o

Z

U3 E 

6 M2

1
t

e2dts EkGs; xs; xs  Gs; ys; ys k2 ds 6 M 2 K 2

1

6 2M2 K 2

2
Z

Tt  sPGs; xs; xs  Gs; ys; ys dWs

Z

Z

kTt  sPGs; xs; xs  Gs; ys; ys k2 ds

1

1




e2dts ds sup kxs  ysk2L2 P;H kxs  ys k2C


M2 K 2
e2dts ds kxs  ysk21
kxs  ysk21 :
d
1

t2R

Similar to the discussion given for U3, for U4, one has

Z

U 4 E


2

M20 K 2
2
Tt  sQ Gs; xs; xs  Gs; ys; ys   dWs
 6 d kxs  ysk1 :

Thus, by combining U1, U2, U3 with U4, it follows that



2K
K2
1
EkLxt  Lytk2 6 4 M2 M20

kxs  ysk21 ;
d
d2

which implies

s


2K 1 K 2
kLxt  Lytk1 6 2 M2 M20

kxs  ysk1 Hkx  yk1 :

2
d
d

3:6

Since H < 1, thus by (3.6), we know that L is a contraction mapping. Hence, by the contraction mapping principle, L has a
unique xed point x(t), which is obviously the unique quadratic-mean asymptotically almost periodic mild solution of
Eq. (3.1). h

4. Exponential stability
In this section, we consider the particular case where the semigroups {T(t)}tP0 is exponentially stable, furthermore, we
have the following result on the stability of the unique quadratic-mean asymptotically almost periodic mild solution.

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

1509

Theorem 4.1. Let (H2) and (H3) be satised. Suppose that A generates a uniformly exponentially stable semigroup {T(t)}tP0, that is,
there exist constants M0 > 0, d0 > 0 with
0

kTtk 6 M 0 ed t ;

t P 0:

4:1

Then, Eq. (3.1) has a unique quadratic-mean asymptotically almost periodic mild solution x(t), provided

s


1
2K 1 K 2
H2
M 02 02
0 < 1:
02
d
d
M
Furthermore, x(t) is global exponentially stable.
Proof. From
0

for t P 0; Tt Tt1 ;

kTtk 6 M 0 ed t ;

for t 6 0;

it follows that

kTtxk 6 M 0 expd0 tkxk; for t > 0;

1
kTtxk 6 0 expd0 tkxk; for t 6 0:
M
Thus, from Theorem 3.1 one can immediately obtain the existence and uniqueness of a quadratic-mean asymptotically almost periodic mild solution of Eq. (3.1). In the following, we discuss the stability of the quadratic-mean asymptotically almost periodic mild solution.
By the assumptions, one can choose a positive constant a such that 0 < a < d0 . It follows from (3.3) that

Z

eat kxtk2L2 P;H 6 4eat E


1

2
Z


at 
Tt  sPFs; xs; xs ds

4e
E



2

Tt  sQFs; xs; xs ds


Z t
2



4eat E
Tt

sPGs;
xs;
xs

r


dWs


1
Z 1
2


4eat E
Tt  sQGs; xs; xs  dWs


t

V 1 V 2 V 3 V 4:

4:2

Estimating the terms on the right-hand side of (4.2) yields

Z

V 1 4eat E


1

4eat M 02 K 1
6 4eat M 02 K 1

2
Z

at 02
Tt  sPFs; xs; xs ds
 6 4e M K 1

1

Z
Z

d0 ts
2

1
t

d0 ts

1

t
1


12 2
kxsk2L2 P;H kxs k2C ds

Z t




0
1
0
ed ts kxsk2L2 P;H kxs k2C ds 4M 02 K 1 eat
ed ts kxsk2L2 P;H kxs k2C ds
d
1
1


0
ed as eas kxsk2L2 P;H kxs k2C ds:

ed ts ds

1
0
4M02 K 1 0 ed at
d


2
0
ed ts kxsk2L2 P;H kxs k2C ds

For any xt 2 L2 P; H and any e > 0, according to (4.1), there exists a constant t1 > 0 such that eas kxs  rk2L2 P;H < e for
s P t1. Thus

V 1 64M 02 K 1

1 d0 at
e
d0

t1



0
ed as eas kxsk2L2 P;H kxs k2C ds

Z t1


1
0
0
ed as eas kxsk2L2 P;H kxs k2C ds
4M K 1 0 ed at
d
1
Z

1 d0 at t1 d0 as as 
1
02
e
e kxsk2L2 P;H kxs k2C ds 8M02 K 1 0 0
e:
64M K 1 0 e
d
d d  a
1
02

4:3

From ed at ! 0 as t ? 1, it follows that there exists a constant t2 P t1 such that for any t P t2,

4M 02 K 1

1 d0 at
e
d0

t1
1



0
ed as eas kxsk2L2 P;H kxs k2C ds < e  8M 02 K 1

Therefore, from (4.3) and (4.4), one obtains for any t P t2,

1
e:
d0 d0  a

4:4

1510

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

Z

4e E


at

1

2

Tt  sPFs; xs; xs ds
 < e;

which implies

Z

V 1 4eat E


1

2

Tt  sPFs; xs; xs ds
 ! 0 as t ! 1:

4:5

Similar to the discussion of the rst term on the right-hand side of (4.2), for any xt 2 L2 P; H; t 2 r; 1; one has

Z

V 2 4eat E


2

Tt  sQFs; xs; xs ds
 ! 0 as t ! 1:

4:6

Estimating the third term on the right-hand side of (4.2), for any xt 2 L2 P; H; t 2 r; 1; one has

Z

V 3 4eat E


1

6 4eat M 02 K 2

2
Z

at 02
Tt  sQGs; xs; xs  dWs
6
4e
M


1

t0

e2d ts kGs; xs; xs k2L2 P;H ds



0
e2d ts kxsk2L2 P;H kxs k2C ds:

Similar to the proof of the rst term on the right-hand side of (4.2), for any xt 2 L2 P; H; t 2 r; 1; one obtains

Z

V 3 4eat E


2

Tt  sPGs; xs; xs  dWs
 ! 0 as t ! 1:

4:7

Similarly, one can prove that for any xt 2 L2 P; H; t 2 r; 1;

Z

V 4 4eat E


2

Tt  sQGs; xs; xs  dWs
 ! 0 as t ! 1:

4:8

Thus, from (4.5)(4.8), one knows that eat kxtk2L2 P;H ! 0 as t ? 1. Therefore, Eq. (3.1) has a unique quadratic-mean asymptotically almost periodic mild solution, which is exponentially stable. h

5. Applications
In this section, we provide an example to illustrate our main results.
Consider the following semilinear stochastic partial functional differential equations with Dirichlet conditions

"

#
@2
dut; x
ut; x v1 t; ut; x; ut x dt v2 t; ut; x; ut x  dwt;
@x2
ut; p 0;

t 2 R;

x 2 0; p;

t 2 R;

ut; 0
5:1

where w(t) is a standard one-dimensional Wiener process. Let

us; x us; x;

u; x 2 Cr; 0; R1 ;

us;  2 L2 0; p; s 2 r; 0; x 2 0; p:

2

@
Denote X L2 P; L2 0; p and dene A : D(A)  X ? X given by A @x
2 with the domain

DA fu 2 X : u00 2 X; u0 2 X is absolutely continuous on 0; p; u0 up 0g:

It is well known that a strongly continuous semigroup {T(t)}tP0, generated by the operator A, satises kT(t)k 6 et for
t P 0. Taking

Ft; u; ut  v1 t; ut; ; ut ;

Gt; u; ut  v2 t; ut; ; ut ;

then Eq. (5.1) takes the following abstract form

dut Aut Ft; ut; ut dt Gt; ut; ut  dwt:

From Theorem 4.1, it follows that the following proposition holds.
Proposition 5.1. Let v1, v2 satisfy (H2) and (H3), then Eq. (5.1) has a unique global exponentially stable quadratic-mean
asymptotically almost periodic mild solution provided

p
2 4K 1 2K 2 < 1:

J. Cao et al. / Applied Mathematics and Computation 218 (2011) 14991511

1511

Acknowledgements
This work is supported by the National Natural Science Foundation of China (No. 10871074) and the Youth Science Foundation of Guangxi (No. 2011GXNSFB018065).
References
[1] A. Fink, Almost Periodic Differential Equations, Lecture Notes in Mathematics, vol. 37, Springer-Verlag, Berlin, 1974.
[2] W. Arendt, C. Batty, Almost periodic solutions of rst and second order Cauchy problems, J. Differential Equations 137 (1997) 363383.
[3] Y. Hino, S. Murakami, T. Yoshizawa, Almost periodic solutions of abstract functional differential equations with innite delay, Nonlinear Anal. 30
(1997) 853864.
[4] B. Levitan, V. Zhikov, Almost Periodic Functions and Differential Equations, Cambridge University Press, Cambridge, 1982.
[5] T. Yoshizawa, Stability theory and the existence of periodic solutions and almost periodic solutions, Appl. Math. Sci., vol. 14, Springer-Verlag, New York,
Heidelberg, 1975.
[6] J. Prss, Evolutionary Integral Equations and Applications, Birkhuser, Basel, 1993.
[7] C. Zhang, Almost Periodic Type Functions and Ergodicity, Science Press/Kluwer Acad. Publ., Beijing/Dordrecht, 2003.
[8] K. Engel, R. Nagel, One-Parameter Semigroups for Linear Evolution Equations, Springer, Berlin, 1999.
[9] A. Pazy, Semigroups of linear operators and applications to partial differential equations, Applied Mathematical Science, vol. 44, Springer-Verlag,
Berlin, New York, 1983.
[10] E. Slutsky, Sur les fonctions alatoires presque priodiques et sur la decomposition des functions alatoires, Actualits Sceintiques et industrielles, 738,
Herman, Paris, 1938. pp. 3355.
[11] M. Udagawa, Asymptotic properties of distributions of some functionals of random variable, Rep. Statist. Appl. Res. Union Jap. Sci. Eng. 2 (1952) 198.
[12] T. Kawata, Almost periodic weakly stationary processes, in: G. Kallianpur, P.R. Krishnaiah, J.K. Ghosh (Eds.), Statistics and Probability: Essays in
Honour of C.R. Rao, North-Holland, Amsterdam, 1982, pp. 383396.
[13] R. Swift, Almost periodic harmonizable processes, Georgian Math. J. 3 (1996) 275292.
[14] P. Bezandry, T. Diagana, Existence of almost periodic solutions to some stochastic differential equations, Appl. Anal. 117 (2007) 110.
[15] Z. Huang, Q. Yang, Existence and exponential stability of almost periodic solution for stochastic cellular neural networks with delay, Chaos Solitons
Fract. 42 (2009) 773780.
[16] T. Christopher, H. Baker, E. Buckwar, Exponential stability in pth mean of solutions, and of convergent Euler-type solutions, of stochastic delay
differential equations, J. Comput. Appl. Math. 184 (2005) 404427.
[17] H. Bao, J. Cao, Existence and uniqueness of solutions to neutral stochastic functional differential equations with innite delay, Appl. Math. Comput. 215
(2009) 17321743.
[18] Y. Ren, N. Xia, A note on the existence and uniqueness of the solution to neutral stochastic functional differential equations with innite delay, Appl.
Math. Comput. 214 (2009) 457461.
[19] J. Luo, A note on exponential stability in pth mean of solutions of stochastic delay differential equations, J. Comput. Appl. Math. 198 (2007) 143148.
[20] Y. Ren, N. Xia, Existence, uniqueness and stability of the solutions to neutral stochastic functional differential equations with innite delay, Appl. Math.
Comput. 210 (2009) 7279.
[21] W. Ruess, W. Summers, Asymptotic almost periodicity and motions of semigroups of operators, Linear Algebra Appl. 84 (1986) 335351.
[22] W. Ruess, Q. Vu, Asymptotically almost periodic solutions of evolution equations in Banach spaces, J. Differ. Eqs. 122 (1995) 282301.
[23] W. Arendt, C. Batty, Asymptotically almost periodic solutions of inhomogeneous Cauchy problems on the half-line, Bull. London Math. Soc. 31 (1999)
291304.
[24] N. Minh, G. NGurkata, R. Yuan, Lectures on the Asymptotically Behavior of Solutions of Differential Equations, Science Publishers, Inc, 2008.
[25] Q. Liu, R. Yuan, Asymptotic behavior of solutions to abstract functional differential equations, J. Math. Anal. Appl. 356 (2009) 405417.