Nonauton. Dyn. Syst.

2020; 7:1–21

Research Article Open Access

Ravikumar Kasinathan, Ramkumar Kasinathan, Mahamat Hassan Mahamat Hamit,

and Mamadou Abdoul Diop*

Exponential behavior of neutral impulsive

stochastic integro-differential equations
driven by Poisson jumps and Rosenblatt
process
https://doi.org/10.1515/msds-2020-0001
Received September 11, 2019; accepted February 3, 2020

Abstract: In this article, we are concerned with the neutral impulsive stochastic integro-differential equations
driven by Poisson jumps and Rosenblatt process. By using resolvent operator and some analysis techniques,
we ensure existence and uniqueness of solutions. Further, we investigate exponential stability of mild solu-
tions. We have also given an example to illustrate our theoretical results.

Keywords: Existence results, Stability, Poisson jumps, Rosenblatt process, Resolvent operator

MSC: 60H15, 35R60

1 Introduction
In the past decades, the theory of nonlinear functional differential or integro-differential equations with re-
solvent operators has become an active research field due to their applications in many physical phenomena.
The resolvent operator is comparable to the semigroup operator for abstract differential equations in Banach
spaces. However, the resolvent operator does not satisfy semigroup properties. The study of deterministic
neutral functional differential equations was initiated by Hale and Mayer [17]. For more details on the theory
and their applications, we also refer the readers to Hale and Lunel [18], Kolmanovkii and Nosov [20] and so
on. Deterministic and stochastic differential equations have gained great popularity in the last few years due
to their use in many areas, such as physics, electronics, control theory, engineering and economics. Several
authors have considered the existence, uniqueness and asymptotic behavior of mild solutions, and many im-
portant theory and applications findings have been obtained. For more details we refer to the papers by Ali et
al. [4], El-Borai et al. [11], Gorec and Sathanantham [12], Gupta and Dabas [15], Gupta and al. [16], Laksmikan-
tha [22], Ahmed [2], Ahmed et al. [3], Arthi and Balachandran [5],Gupta et al. [14], Levin et al. [24].
On the other side, fractional Brownian motion was intensively explored due to their applications in vari-
ous domain. We point out that, a fractional Brownian motion (fBm) of Hurst index H ∈ (0, 1) (see [21, 26] for

Ravikumar Kasinathan: Department of Mathematics, PSG College of Arts and Science, Coimbatore, 641014, India, E-mail:
[email protected]
Ramkumar Kasinathan: Department of Mathematics, PSG College of Arts and Science, Coimbatore, 641014, India, E-mail:
[email protected]
Mahamat Hassan Mahamat Hamit: Université Gaston Berger de Saint-Louis, UFR SAT, Département de Mathématiques,
B.P234 Saint -Louis Sénégal, E-mail: [email protected]
*Corresponding Author: Mamadou Abdoul Diop: Université Gaston Berger de Saint-Louis, UFR SAT, Département de Mathé-
matiques, B.P234 Saint-Louis Sénégal, E-mail: [email protected](corresponding author)

Open Access. © 2019 Ravikumar Kasinathan et al., published by De Gruyter. This work is licensed under the Creative Commons
Attribution alone 4.0 License.
2 | Ravikumar Kasinathan et al.

more details) is a Gaussian process B H = {B H (t), t ≥ 0}, centered, with the covariance function
  1 
R H (t, s) = E B H (t)B H (s) = t2H + s2H − |t − s|2H .
2
We also mention that fBm is not a semimartingale and when H = 1/2 , the fBm becomes standard Brownian
motion . Further, if H > 1/2, the fBm B H have a long-memory and this property makes it an ideal process to
modeling in biology, mathematics finance [7, 8] etc.
Moreover, the fBm belongs to Hermite family processes, it’s selfsimilar which being defined as limits that
appear in the so-called Non-Central Theorem. For d ≥ 1, they have the following representation

Z Zt Y
 
Z d
H −1
Y tH,d = c(H0 ) ...  s − x j + 0 ds dBx1 ...dBx d , ∀t > 0,

R R 0 j=1

where {B x i : x i ∈ R, i = 1, ..., d} are some two-sided Brownian motions, c(H0 ) is a normalizing constant such
that
1 H−1 1
E|Y1H,d |2 = 1, H0 = + , H ∈ ( , 1).
2 d 2
When d = 1 the process become a fractional Brownian motion, thus Taqqu [32] have named the Rosenblatt
process when d = 2, which is not Gaussian process but they have stationary increments and long range-
dependency. Recently, the Rosenblatt process have attracted attention of many authors due to their proper-
ties. For example, Meajima and Tudor [29], Veillette and Taqqu [36] have given many important properties
of distributions, Bernet and Tudor [6], Viens and Tudor [35] established the construction of estimator for the
self-similarity parameter H.
Based on the above works, we investigate the following neutral stochastic functional integro-differential
equations with delay and impulses effects

Zt
 "

  
 



 d u(t) − q(t, u t ) = A u(t) − q(t, u t ) + Υ(t − s) u(s) − q(s, u s ) ds + f (t, u t )

0





 Zt # Z
dv) + σ(t)dZ QH (t), t ∈ [0, T], t ≠ t k , (1)

+ g(t, s, u s )ds dt + h(t, u t , v)N(dt, e


 0 θ
−

∆u(t) = I (u(t )), t = t , k = 1, 2, · · · ,



 k k k





= ϕ(t) ∈ PC [−τ, 0], H , −τ ≤ t ≤ 0,


u0 (t)


where A : D(A) ⊂ H → H is the infinitesimal generator of a C0 -semigroup (T(t))t≥0 of bounded linear oper-
ators in a Hilbert space H; for t > 0, Υ(t) a closed linear operator on H, with D(A) ⊂ D(Υ). The impulsive
moments t k satisfy the condition 0 < t1 < t2 < · · · < t k < · · · , limk→∞ t k = ∞, I k : H → H, ∆u(t k ) =
u(t+k ) − u(t−k ), u(t+k ) and u(t−k ) are the right and left limits of u(t) at t k , respectively which is the jump size of the
state u at t k . For ϕ ∈ PC, ||ϕ|| = sups∈[−r,0] ||ϕ(s)|| < ∞, where PC = ϕ : [−r, 0] → H, ϕ(t) is continuous


everywhere except a finite number of points t̃ at which ϕ(t̃−k ), ϕ(t̃+k ) exist and ϕ(t̃−k ) = ϕ(t̃) .
For any t ∈ [0, T] and any continuous function u, the element of PC is defined by u t (θ) = u(t + θ), −τ ≤ θ ≤
0.The functions q, f : [0, +∞) × PC → H, g : [0, +∞) × [0, +∞) × PC → H, h : [0, +∞) × H × θ → H and
σ : [0, +∞) → L0Q (Y , H) are appropriate functions, and Z QH is assumed to be a Rosenblatt process.
The particular case Υ(t) = 0 and h = 0 of Eq. (1) has been considered by Ma et al [25], where the authors
used an impulsive integral inequality to prove their result. It should be mentioned that there is no work yet
reported on the exponential stability of neutral impulsive stochastic integro-differential equations driven by
Poisson jumps and Rosenblatt process. Motivated by this facts, our main objective is to study the exponential
stability for a class of neutral impulsive stochastic integro-differential equations(1). In this paper, we derive
existence and exponential results for the system (1) with the help of resolvent operator and fixed point tech-
niques.
In the first result, we obtain the sufficient conditions proving existence and uniqueness of the mild solution
 Exponential behavior | 3

of (1) by utilizing Banach fixed point theorem under Lipschitz conditions on nonlinear terms. While in the
second result, we have proved the exponential stability of mild solution via an integral inequality. Our arti-
cle, expands the usefulness of stochastic integro-differential equations, since the literature shows results for
existence and exponential stability for such equations in the case of semigroup only (see [2, 3, 11, 14, 16] and
the references therein ). The results obtained improve, extend and complete many other important ones in
the literature.
The following is the organization of this paper. We recall some preliminary definitions and outcomes
in Section 2. Section 3 is devoted to investigate existence and uniqueness of mild solution. The exponential
stability for the mild solution is also discussed. We give an example in the fourth section to illustrate the
results. The last section is dedicated to conclude this paper.

2 Preliminaries
In this section, we provided some basic results about Poisson process, Resolvent operator and Rosenblatt
process.

2.1 poisson jumps process

Let Bσ (H) the Borel σ-algebra of H. Let (p(t)), (t ≥ 0) be an H-valued, σ-finite stationary =t -adapted Poisson
point process on (Ω, =, P). The counting random measure N defined by
X
N((t1 , t2 ] × θ)(w) = 1U (p(s)(w)),
t1 <s≤t2

for any U ∈ Bσ (H − {0}). where 0 ∉ θ̄ is called the Poisson random measure associated to Poisson point
process p. The following notation is used

N(t, θ) = N((0, t] × θ).

Then it is known that there exists a σ-finite measure θ such that

E(N(t, θ)) = ν(θ)t,

exp(−tν(θ))(tν(θ))k
P(N(t, θ) = k) = .
k!
This measure ν is said Levy measure. Then the measure N
e is defined by

N((0,
e t] × θ) = N((0, t] × θ) − tν(θ).

This measure N(dt,

e du) is called the compensated Poisson random measure, and dtν(θ) is called the com-
pensator (see [30]).

Definition 2.1. Let θ ∈ Bσ (H− {0}). P2 ([0, T]×θ; H) is the space of all predictable mappings h : [0, T]×θ×Ω →
H for which
ZT Z
2
E h(t, ν) dtλ(dν) < ∞.
0 θ

ZT Z
We may then define the H-valued stochastic integral L(t, ν)N(dt,
e dν), which is a centered square-
0 θ
integral martingale [30].
4 | Ravikumar Kasinathan et al.

2.2 Rosenblatt process

In this subsection, we recall some basic concepts on the Rosenblatt process as well as the Wiener integral
with respect to it. Consider (ξ n )n∈Z a stationary Gaussian sequence with mean zero and variance 1 such that
2H−2
its correlation function satisfies that R(n) := E(ξ0 ξ n ) = n k L(n), with H ∈ ( 12 , 1) and L is a slowly varying
function at infinity. Let g be a function of Hermite rank k, that is, if g admits the following expansion in
Hermite polynomials
X 1
g(x) = c j H j (x), cj = E(g(ξ0 H j (ξ0 ))),
j!
j≥0

then k = min j|c j ≠ 0 ≥ 1, where H j (x) is the Hermite polynomial of degree j given by H j (x) =
x2 2
j d j − x2
(−1) e 2
dx j
e . Then, the Non-Central Limit Theorem (see, for example, Dobrushin & Major [10]) says
1 P[nt]
nH j=1 g(ξ j ) converges as n → ∞, in the sense of finite dimensional distributions, to the process
Z Zt Y
 
k
k (− 12 + 1−H )
Z H (t) = c(H, k)  (s − y j )+ k  dsdB(y1 ) · · · dB(y k ), (2)
j=1
Rk 0

where the above integral is a Wiener-Itô multiple integral of order k with respect to the standard Brownian
motion (B(y))y∈R and c(H, k) is a positive normalization constant depending only on H and k. The process
k k d
(Z H (t))t≥0 is called as the Hermite process and it is H self-similar in the sense that for any c > 0, (Z H (ct)) =
H k
(c Z H (t)) and it has stationary increments.
The fractional Brownian motion (which is obtained from (2) when k = 1) is the most used Hermite process
to study evolution equations due to its large range of applications. When k = 2 in (2), Taqqu [32] named the
process as the Rosenblatt process. The stationarity of increments, self-similarity and long range dependence
(see Tindel, Tudor and Viens [33]) were made that the Rosenblatt process is very important in practical ap-
plications. However, it is noted that Rosenblatt process is not Gaussian. In fact, due to their properties (long
range dependence, self-similarity), the fractional Brownian motion process has large utilization in practi-
cal models, for instance in telecommunications and hydrology. So, many researchers prefer to use fractional
Brownian motion than other processes because it is Gaussian and it facilitate calculations. However in con-
crete situations when the Gaussianity is not plausible for the model, one can use the Rosenblatt process.
In recent years, there exists many works that investigated on diverse theoretical aspects of the Rosenblatt
process. For example, Leonenko and Ahn [23] gave the rate of convergence to the Rosenblatt process in the
Non-Central Limit Theorem and the wavelet-type expansion has been presented by Abry and Pipiras [1]. Tu-
dor [34] established, the representation as a Wiener-Itô multiple integral with respect to the Brownian motion
on a finite interval and developed the stochastic calculus with respect to it by using both pathwise type cal-
culus and Malliavin calculus (see also Maejima and Tudor [27]). For more details on Rosenblatt process, we
refer the reader to Maejima and Tudor [28, 29]), Pipiras and Taqqu [31] and the references therein.

Consider a time interval [0, T] with arbitrary fixed horizon T and let Z H (t), t ∈ [0, T] be a one-
dimensional Rosenblatt process with parameter H ∈ ( 21 , 1). According to the work of Tudor [34], the Rosen-
blatt process with parameter H > 21 can be written as
Zt Zt
 t 
Z H′ H′
∂K ∂K
Z H (t) = d(H)  (u, y1 ) (u, y2 )du dB(y1 )dB(y2 ), (3)
∂u ∂u
0 0 y1 ∨y2
H
where K (t, s) is given by
Zt
1
K H (t, s) = c H s 2 −H (u − s)H−3/2 u H−1/2 du for t > s,
s

with s
H(2H − 1)
cH = ,
β(2 − 2H, H − 12 )
 Exponential behavior | 5

β(., .) denotes the Beta function, K H (t, s) = 0 when t ≤ s, (B(t), t ∈ [0, T]) is a Brownian motion, H ′ = H+1
2 and
q
1 H

d(H) = H+1 2(2H−1) is a normalizing constant. The covariance of the Rosenblatt process Z H (t), t ∈ [0, T]
satisfy
1  2H 2H 
E(Z H (t)Z H (s)) = s + t − |s − t|2H .
2
The covariance structure of the Rosenblatt process allows to construct Wiener integral with respect to it. We
refer to Maejima and Tudor [27] for the definition of Wiener integral with respect to general Hermite processes
and to Kruk, Russo, and Tudor [19] for a more general context (see also Tudor [34]).
Note that
ZT ZT
Z H (t) = I(1[0,t] )(y1 , y2 )dB(y1 )dB(y2 ),
0 0

where the operator I is defined on the set of functions f : [0, T] → R, which takes its values in the set of
functions g : [0, T]2 → R2 and is given by

ZT ′ ′
∂K H ∂K H
I(f )(y1 , y2 ) = d(H) f (u) (u, y1 ) (u, y2 )du.
∂u ∂u
y1 ∨y2

Let f be an element of the set E of step functions on [0, T] of the form

n−1
X
f = a i 1(t i ,t i+1 ] , t i ∈ [0, T].
i=0

Then, it is natural to define its Wiener integral with respect to Z H as

ZT n−1
X ZT ZT
f (u)dZ H (u) := a i (Z H (t i+1 ) − Z H (t i )) = I(f )(y1 , y2 )dB(y1 )dB(y2 ).
0 i=0 0 0

Let H be the set of functions f such that

ZT ZT
kf k2H := 2 (I(f )(y1 , y2 ))2 dy1 dy2 < ∞.
0 0

It follows that (see Tudor[34])

ZT ZT
kf k2H = H(2H − 1) f (u)f (v)|u − v|2H−2 dudv.
0 0

It has been proved in Maejima and Tudor [27] that the mapping

ZT
f → f (u)dZ H (u)
0

defines an isometry from E to L2 (Ω) and it can be extended continuously to an isometry from H to L2 (Ω)
because E is dense in H. We call this extension as the Wiener integral of f ∈ H with respect to Z H . It is
noted that the space H contains not only functions
 but its elements could be also distributions. Therefore
 it
RTRT 2H−2
is suitable to know subspaces |H| of H : |H| = f : [0, T] → R| 0 0 |f (u)||f (v)|u − v| dudv < ∞ . The
space |H| is not complete with respect to the norm k.kH but it is a Banach space with respect to the norm

ZT ZT
kf k2|H| = H(2H − 1) |f (u)||f (v)|u − v|2H−2 dudv.
0 0
6 | Ravikumar Kasinathan et al.

As a consequence, we have
L2 ([0, T]) ⊂ L1/H ([0, T]) ⊂ |H| ⊂ H.
For any f ∈ L2 ([0, T]), we have
ZT
kf k2|H| ≤ 2HT 2H−1
|f (s)|2 ds
0
and
kf k2|H| ≤ C(H)kf k2L1/H ([0,T]) , (4)
for some constant C(H) > 0. Let C(H) > 0 stands for a positive constant depending only on H and its value
may be different in different appearances.
Define the linear operator K *H from E to L2 ([0, T]) by

ZT
∂K
(K *H f )(y1 , y2 ) = f (t) (t, y1 , y2 )dt,
∂t
y1∨y2

where K is the kernel of Rosenblatt process in representation (3)

Zt ′ ′
∂K H ∂K H
K(t, y1 , y2 ) = 1[0,t] (y1 )1[0,t] (y2 ) (u, y1 ) (u, y2 )du.
∂u ∂u
y1 ∨y2

Note that (K *H 1[0,t] )(y1 , y2 ) = K(t, y1 , y2 )1[0,t] (y1 )1[0,t] (y2 ). The operator K *H is an isometry between E to
L2 ([0, T]), which can be extended to the Hilbert space H. In fact, for any s, t ∈ [0, T] we have
D E
K *H 1[0,t] , K *H 1[0,s] = K(t, ., .)1[0,t] , K(s, ., .)1[0,s] L2 ([0,T])
L2 ([0,T])
Zt∧s Zt∧s
= K(t, y1 , y2 )K(s, y1 , y2 )dy1 dy2
0 0
Z t Zs
= H(2H − 1) |u − v|2H−2 dudv
0 0
= 1[0,t] , 1[0,s] H
.

Moreover, for f ∈ H, we have

ZT ZT
Z H (f ) = (K *H f )(y1 , y2 )dB(y1 )dB(y2 ).
0 0

Let {Z n (t)}n∈N be a sequence of two-sided one dimensional Rosenblatt process mutually independent on
(Ω, F, P). We consider a K-valued stochastic process Z Q (t) given by the following series:
∞
z n (t)Q1/2 e n ,
X
Z Q (t) = t ≥ 0.
n=1

Moreover, if Q is a non-negative self-adjoint trace class operator, then this series converges in the space K,
that is, it holds that Z Q (t) ∈ L2 (Ω, K). Then, we say that the above Z Q (t) is a K-valued Q- Rosenblatt process
with covariance operator Q. For instance, if {σ n }n∈N is a bounded sequence of non-negative real numbers
such that Qe n = σ n e n , by assuming that Q is a nuclear operator in K, then the stochastic process
∞ ∞
√
z n (t)Q1/2 e n =
X X
Z Q (t) = σ n z n (t)e n , t ≥ 0, (5)
n=1 n=1

is well-defined as a K-valued Q- Rosenblatt process.

 Exponential behavior | 7

Definition 2.2. (Tudor[34]). Let φ : [0, T] → L0Q (K, H) such that ∞ * 1/2
P
n=1 k K H (φQ e n )kL2 ([0,T];H) < ∞. Then,
its stochastic integral with respect to the Rosenblatt process Z Q (t) is defined, for t ≥ 0, as follows :

Zt ∞ Z t ∞ Zt Zt
1/2
(K *H (φQ1/2 e n ))(y1 , y2 )dB(y1 )dB(y2 ).
X X
φ(s)dZ Q (s) := φ(s)Q e n dz n (s) = (6)
0 n=1 0 n=1 0 0

P∞
Lemma 2.1. For ψ : [0, T] → L0Q (K, H) such that n=1 k ψQ
1/2
e n kL1/H ([0,T];H) < ∞ holds, and for any a, b ∈
[0, T] with b > a, we have
2
Zb ∞ Z b
2H−1
kψ(s)Q1/2 e n k2 ds.
X
E ψ(s)dZ Q (s) ≤ c(H)(b − a)
a n=1 a

If, in addition,
∞
kψ(t)Q1/2 e n k is uniformly convergent for t ∈ [0, T],
X

n=1

then, it holds that

2
Zb Zb
2H−1
E ψ(s)dZ Q (s) ≤ C(H)(b − a) kψ(s)k2L0 (K,H) ds.
Q
a a

Proof. Let {e n }n∈N be the complete orthonormal basis of K introduced above.

Applying (6) and Hölder inequality, we have
2 2
Zb ∞ Z b
1/2
X
E ψ(s)dZ Q (s) = E ψ(s)Q e n dz n (s)
a n=1 a
2
∞ Zb
ψ(s)Q1/2 e n dz n (s)
X
= E
n=1 a

∞ Zb Zb
kψ(s)Q1/2 e n kkψ(t)Q1/2 e n k|t − s|2H−2 dsdt
X
= H(2H − 1)
n=1 a a
2H
Zb

∞
kψ(s)Q1/2 e n k1/H ds
X
≤ C(H) 
n=1 a

∞ Z b
2H−1
kψ(s)Q1/2 e n k2 ds.
X
≤ C(H)(b − a)
n=1 a

Lemma 2.2. [37] Let a function Γ : [−r, +∞) → [0, +∞) be such that there exist positive constants ω >
0, α j (j = 1, 2, 3), and β i (i = 1, 2, · · · ) such that

α1 e−ωt for t ∈ [−r, 0],








 Zt
α1 e−ωt + α2 supθ∈[−r,0] Γ(t + θ) + α3 e−ω(t−s) sup Γ(t + θ)ds


Γ(t) ≤ θ∈[−r,0]


 0

β i e−ω(t−t i ) Γ(t−i ) for t ≥ 0.
 X
 +



t i <t
8 | Ravikumar Kasinathan et al.

+∞
X
If α2 + α3
ω + β i < 1, the Γ(t) ≤ Ne−γt for t ≥ −r, where Γ > 0 is the unique solution to the equation
i=1
+∞
α1 (ω−γ)
X
α3
α2 + (ω−γ)
eγr + β i = 1 and N = max{α1 , α3 eγr }) > 0.
i=1

2.3 Partial integro-differential equations in Banach spaces

In this section, we recall some fundamental results needed to establish our main results. For the theory of
resolvent operators we refer the reader to [13]. Throughout this paper, H is a Banach space, A and Υ(t) are
closed linear operators on H. Y represents the Banach space D(A) equipped with the graph norm defined by

|y|Y := |Ay| + |y| for y ∈ Y .

The notations C([0, +∞); Y), B(Y , H) stand for the space of all continuous functions from [0, +∞) into Y, the
set of all bounded linear operators from Y into H, respectively. We consider the following Cauchy problem

Zt


 ρ′ (t) = Aρ(t) + Υ(t − s)ρ(s)ds, for t ≥ 0,

(7)

 0
ρ(0) = v0 ∈ H.


Definition 2.3. ([13]). A resolvent operator for Eq. (7) is a bounded linear operator valued function Π(t) ∈ L(H)
for t ≥ 0, satisfying the following properties:
(i) Π(0) = I and |Π(t)| ≤ Me βt for some constants M and β.
(ii) For each x ∈ H, Π(t)x is strongly continuous for t ≥ 0.
(iii) Π(t) ∈ L(Y) for t ≥ 0. For x ∈ Y , Π(·)x ∈ C1 ([0, +∞); H) ∩ C([0, +∞); Y) and

Zt
′
Π (t)x = AΠ(t)x + Υ(s)(t − s)Π(s)xds
0
Zt
= Π(t)Ax + Π(t − s)Υ(s)xds for t ≥ 0.
0

Remark 2.1. There exist a constant M̂ > 0 such that ||Π(t)|| ≤ M̂, for t ∈ [0, T].

The following assumptions are imposed on the system under consideration:

(H1) A is the infinitesimal generator of a strongly C0 -semigroup {S(t)}t≥0 on H.

(H2) For all t ≥ 0, Υ(t) is a closed linear operator from D(A) to H, and Υ(t) ∈ B(Y , H). For any y ∈ Y, the
map t → Υ(t)y is bounded, differentiable and the derivative t → Υ ′ (t)y is bounded and uniformly
continuous on R+ .

Theorem 2.3. ([13, Theorem 3.7]) Assume that (H1)-(H2) hold. Then there exists a unique resolvent operator
for the Cauchy problem (7).
 Exponential behavior | 9

3 Main Results
3.1 Existence of mild solution

In this section, we present and prove the existence and uniqueness of mild solutions of Eq.(1) by means of
the theory of resolvent operator and contraction mapping principle. First of all, we begin with the definition
of mild solution for Eq.(1)

Definition 3.1. An H-valued stochastic process u(t), t ∈ [−τ, T], is called a mild solution of Eq.(1) if
1. u(·) ∈ PC([−τ, T], L2 (Ω, H)),
2. for t ∈ [−τ, 0], u(t) = ϕ(t),
3. for t ∈ [0, T] , u(t) satisfies the following integral equation:
 
u(t) = Π(t) ϕ(0) − q(0, ϕ) + q(t, u t )
Zt Zt Zs
+ Π(t − s)f (s, u s )ds + Π(t − s) g(s, η, u η )dηds
0 0 0
Zt Z
(8)
Π(t − t k )I k (u(t−k )) +
X
+ Π(t − s)h(s, u s , v)N(ds,
e dv)
0<t k <t 0 θ
Zt
+ Π(t − s)σ(s)dZ QH (s) P − a.s.
0

In order to attain the result, we impose the following assumptions:

(H3)(i) For all t ∈ [0, T], there exist constants 0 < K1 < 1 such that, for ψ j ∈ PC, j = 1, 2, the H-valued
function q : [0, +∞) × PC → H satisfies the condition

||q(t, ψ1 ) − q(t, ψ2 )|| ≤ K1 ||ψ1 − ψ2 ||.

Also, K̃1 = sup ||q(t, 0)||.

t∈[0,T]
(ii) The function q is continuous in the quadratic mean sense. ∀φ ∈ PC,

lim E||q(t, φ) − q(s, φ)||2 = 0.

t→s

(H4) There exists a constant K2 > 0 such that, for ψ j ∈ PC, j = 1, 2, the mapping f : [0, +∞) × PC → H
satisfies the following Lipschitz condition for all t ∈ [0, T]:

||f (t, ψ1 ) − f (t, ψ2 )|| ≤ K2 ||ψ1 − ψ2 ||.

Here K̃2 = sup ||f (t, 0)||.

t∈[0,T]
(H5) The mapping g : [0, +∞) × [0, +∞) × PC → H satisfies the following Lipschitz condition. For t ∈ [0, T] ,
there exists a constant K3 > 0 such that, for ψ j ∈ PC, j = 1, 2,

Zt
 
g(t, s, ψ1 ) − g(t, s, ψ2 ) ds ≤ K3 ||ψ1 − ψ2 ||.
0

Here K̃3 = T sup ||g(t, s, 0)||.

0≤s≤t≤T
 10 | Ravikumar Kasinathan et al.

(H6) The mapping h : [0, +∞) × θ × PC → H satisfies the following Lipschitz condition. For t ∈ [0, T], there
exists a constant K4 > 0 such that, for ψ j ∈ PC, j = 1, 2,
 1/2
Z Z
2 2
||h(t, ψ1 , v) − f (t, ψ2 , v)|| λ(dv) ∨  ||h(t, ψ1 , v) − f (t, ψ2 , v)|| λ(dv) ≤ K4 ||ψ1 − ψ2 ||.
θ θ
 1/2
Z
 ||h(t, ψ1 , v) − f (t, ψ2 , v)||2 λ(dv) ≤ K4 ||ψ1 − ψ2 ||.
θ

(H7) The impulsive function I k : H → H is continuous. There exists constants d k > 0 (k = 1, 2, ...) satisfying
+∞
X
d k < ∞, such that
k=1

||I k (ψ1 ) − I k (ψ2 )|| ≤ d k ||ψ1 − ψ2 ||, ||I k (0)|| = 0 for all ψ1 , ψ2 ∈ PC.

(H8) The function σ : [0, +∞) → L0Q (Y , H) satisfies

Zt
||σ(s)||2L0 ds < ∞, ∀ t ∈ [0, T].
Q
0

We have the following two conditions for the complete orthonormal basis {a n }n∈N in Y.
∞
X 1
(C.1) ||σQ 2 a n ||L2 ([0,T];H) < ∞,
n=1
∞
X 1
(C.2) ||σ(t)Q 2 a n ||H is uniformly convergent for all t ∈ [0, T].
n=1
We now establish the existence and uniqueness results for Eq.(1).

Theorem 3.1. Assume that hypotheses (H1)—(H7) are satisfied for all ϕ ∈ PC, T > 0, and

+∞
!2
4M 2 X
K1 + di < 1. (9)
(1 − K1 )
i=1

Then, the Eq.(1) has a unique mild solution on [−τ, T].

Proof. To begin, we introduce Λ T := PC([−τ, T], L2 (Ω, H)) the Banach space of all continuous functions
from [−τ, T] into L2 (Ω, H) equipped with the supremum norm ||ξ ||2Λ T = sups∈[−τ,T] (E||ξ (s)||2 ). Now consider
n o
the closed subset Λ̂ T = u ∈ Λ T : u(τ) = ϕ(τ) for τ ∈ [−τ, 0] of Λ T endowed with the norm || · ||Λ T . The
problem (1) is transformed into a fixed point problem. We define the operator Ψ : Λ̂ T → Λ̂ T by
(Ψu)(t) = ϕ(t), t ∈ [−τ, 0] and for t ∈ [0, T]
 
(Ψu)(t) = Π(t) ϕ(0) − g(0, ϕ) + g(t, u t )

Zt Zt Zs
+ Π(t − s)f (s, u s )ds + Π(t − s) h(s, η, u η )dηds
0 0 0

Zt
t k )I k (u(t−k )) + Π(t − s)σ(s)dZ QH (s) P − a.s.
X
+ Π(t −
0<t k <t 0
 Exponential behavior | 11

Now, to prove the existence result of mild solution of Eq.(1), it is sufficient to show that Ψ has a fixed point.
To this end we subdivide the proof into two steps.
Step 1 : First, we show that the map t → (Ψx)(t) is continuous on the interval [0, T]. Let l̃ be suffiicently
small, for u ∈ Λ̂ T and 0 < t < T. We get
  2
E||(Ψu)(t + l̃) − (Ψu)(t)||2 ≤ 12E +12E||q(t + l̃, u t+l̃ ) − q(t, u t )||2

Π(t + l̃) − Π(t) ϕ(0) − q(0, ϕ)

Zt   2 Zt+l̃ 2
+ 12E Π(t + l̃ − s) − Π(t − s) f (s, u s )ds +12E Π(t + l̃ − s)f (s, u s )ds
0 t
Zt   Zs 2
+ 12E Π(t + l̃ − s) − Π(t − s) g(s, η, u η )dηds
0 0

Zt+l̃ Zs 2
+ 12E Π(t + l̃ − s) g(s, η, u η )dηds
t 0
Zt Z   2
+ 12E Π(t + l̃ − s) − Π(t − s) h(s, u s , v)N(ds,
e du)
0 θ

Zt+l̃Z 2
+ 10E Π(t + l̃ − s)h(s, u s , v)N(ds,
e dv)
t θ
X   2
+ 12E Π(t + l̃ − t k ) − Π(t − t k ) I k (u(t−k ))
0<t k <t
2
Π(t + l̃ − t k )I k (u(t−k ))
X
+ 12E
t<t k <t+l̃

Zt Zt+l̃ 2
+ 12E Π(t + l̃ − s) − Π(t − s)σ(s)dZ QH (s) +12E Π(t + l̃ − s)σ(s)dZ QH (s)
0 t
12
X
≤ 12 Pj .
j=1

By Definition 2.3-(ii) , we have

lim P1 = 0.
l̃→0
By using (ii) of (H3) we obtain that
lim P2 = 0.
l̃→0
By using Hölder inequality we have:
Zt   2
P3 ≤ tE Π(t + l̃ − s) − Π(t − s) f (s, u s ) ds.
0

By Definition 2.3-(ii), we have

 
lim ( Π(t + l̃ − s) − Π(t − s) f (s, u s )) = 0
l̃→0

and by combining the assumption (H4) and Definition 2.3-(ii), we obtain

  2
Π(t + l̃ − s) − Π(t − s) f (s, u s ) ≤ 2M 2 (e2(t+l̃)β + e2tβ )(K2 ||u t ||2 + K˜2 ),
12 | Ravikumar Kasinathan et al.

The Lebesgue dominated theorem, implies that

lim P3 = 0.
l̃→0

Using Hölder inequality, we have

 2
Zt+l̃
P4 ≤ Π(t + l̃ − s)f (s, u s ) ds
 
E
t
Zt+l̃ Zt+l̃
2
≤ ||Π(t + l̃ − s)|| dsE ||f (s, u s )||2 ds
t t
Zt+l̃
≤ l̃ M̂ 2
E||f (s, u s )||2 ds.
t

Thus, we have
lim P4 = 0.
l̃→0

By using similar arguments to P2 and combining assumption (H5) we obtain

lim P5 = 0.
l̃→0

Also, with the same argument to P3 and using assumption (H5) we obtain that

lim P6 = 0.
l̃→0

Similarly, with the same argument to P3 and using assumption (H5) we obtain that

lim P11 = 0, lim P12 = 0.

l̃→0 l̃→0

For P7 , application of Lemma 2.1 gives

Zt  
P7 ≤ 2cH(2H − 1)t2H−1 || Π(t + l̃ − s) − Π(t − s) σ(s)||2L0 ds.
Q
0

By Definition 2.3-(ii) we have

 
Π(t + l̃ − s) − Π(t − s) σ(s) → 0, as l̃ → 0

and by Definition 2.3-(i) we have the inequality :

2
Π(t + l̃ − s) − Π(t − s)σ(s) ≤ 2M 2 (e2(t+l̃)β + e2tβ )||σ(s)||2L0 ,
Q
L0Q

and by the Lebesgue dominated convergence theorem, we get

lim P7 = 0.
l̃→0

By Lemma 2.1 we have

Zt+l̃
P8 ≤ 2cH(2H − 1)l̃ 2H−1
||Π(t − s)σ(s)||2L0 ds,
Q
t
 Exponential behavior | 13

since s ∈ [0, T], we have

Zt+l̃
P8 ≤ 2cH(2H − 1)l̃2H−1 M 2 ||σ(s)||2L0 ds,
Q
t

and regarding to (H7), we get

lim P8 = 0.
l̃→0

By using Definition 2.3-(i) we have

  2
Π(t + l̃ − t k ) − Π(t − t k ) I k (u(t−k )) ≤ 2M 2 (e2(t+l̃)β + e2tβ )d k ||u(t−k )||2 .

By combining assumption (H6) and Definition 2.3-(i), we have

lim P9 = 0.
l̃→0

By using assumption (H6) and Definition 2.3-(i), we get

lim P10 = 0.
l̃→0

Therefore,we can conclude

lim E||(Ψx)(t + τ) − (Ψx)(t)||2 = 0.
l̃→0

Hence, the above arguments imply that function t → (Ψx)(t) is continuous on the interval [0, T].
Step 2: In this part of the proof, we will verify that Ψ is contraction mapping in Λ̂ T1 with some T1 ≤ T to be
specified later. Let u, v ∈ Λ̂ T1 and t ∈ [0, T]. By virtue of elementary inequality we obtain
 t
 Z 2
2 1 2 4
 
||(Ψu)(t) − (Ψv)(t)|| ≤ K1 ||q(t, u t ) − g(t, v t )|| + 1−K1 Π(t − s) f (s, u s ) − f (s, v s ) ds

0
Zt Zs Zs
 
2
+ Π(t − s)  g(s, η, u η )dη − g(s, η, v η )dη ds
0 0 0
Zt Z h i 2
+ Π(t − s) h(s, u s , v) − h(s, v s , v)N(ds,
e dv) ds
0 θ 
2
Π(t − t k ) I k (x(t−k )) − I k (y(t−k ))
X  
+ .

0<t k <t

By using assumptions (H3)-(H6), Defintion 2.3 together with Hölder’s inequality, we get

Zt Zt
4M 2 K22 t 4M 2 K32 t
E||(Ψu)(t) − (Ψv)(t)||2 ≤ K1 E||u t − v t ||2 + 1−k1 E||u s − v s ||2 ds + E||u s − v s ||2 ds
1 − K1
0 0
Zt 2 ∞
!2
4M 2
K42 4M
E||u s − v s ||2 ds + E||u(t−k ) − v(t−k )||2 .
X
+ 1−k dj
1 − k1
0 i=1

Hence, we have
sup E||(Ψu)(s) − (Ψv)(s)||2 ≤ θ(t) sup E||u(s) − v(s)||2 ,
s∈[−τ,t] s∈[−τ,t]

where !2
+∞
K22 t + K32 t + K42 t
 
2 4M 2 X
θ(t) = K1 + 4M + di .
1 − K1 (1 − K1 )
i=1
 14 | Ravikumar Kasinathan et al.

By inequality (9), we have

+∞
!2
4M 2 X
θ(0) = K1 + di < 1.
(1 − K1 )
k=1

Then there exists 0 < T1 ≤ T such that 0 < θ(T1 ) < 1 and the operator Ψ is a contraction on Λ̂ T1 and hence it
has a unique fixed point on [−τ, T1 ], which is a mild solution of Eq.(1) on the interval [−τ, T1 ]. By repeating
a similar process the solution can be extended to the entire interval [−τ, T].

Remark 3.1. Notice that we can extend the solution for t ≥ T. Indeed, if we assume that the constants
K1 , K2 , K3 , K4 which appear in assumptions (H3)-(H6) are independent of T > 0, then the mild solution is
defined for all t ∈ [−r, T], for each T > 0. This will play a crucial role in our analysis of stability. Therefore, in
the next section we will assume that the solutions are defined globally in time (for instance, under the previous
assumptions).

3.2 Exponential stability

In this subsection, it is established the exponential stability in the mean square moment of the mild
solution for Eq.(1), we need to state the following additional assumptions.
(H9)The corresponding resolvent operator (Π(t))t≥ of Eq. (7) verifies the following : There exist γ > 0 and M > 0
such that ||Π(t)|| ≤ Me−γt , for all t ≥ 0.
(H10)There exist nonnegative real numbers R i ≥ 0 and continuous functions
ξ i : [0, +∞) → R+ , ξ i (t) ≤ α i e−γt , i = 1, 2, 3, 4, α i > 0 such that for all t ≥ 0 and ψ i ∈ PC

||q(t, ψ1 )||2 ≤ R1 ||ψ1 ||2 + ξ1 (t),

||f (t, ψ2 )||2 ≤ R2 ||ψ2 ||2 + ξ2 (t),
Zt 2
h(t, s, ψ3 )ds ≤ R3 ||ψ3 ||2 + ξ3 (t),
0
 1/2
Z Z
2 2
h(t, ψ4 , v) λ(dv) ∨  h(t, ψ4 , v) λ(dv) ≤ R4 ||ψ4 ||2 + ξ4 (t).
θ θ

(H11)The function σ : [0, +∞) → L0Q (Y , H) satisfies the following condition in addition to assumptions (C.1) and
(C.2):
Z+∞
eγs ||σ(s)||2L0 ds < ∞.
Q
0

Theorem 3.2. Assume that the conditions of Theorem 3.1 are satisfied and (H9)–(H11) are fulfilled. Then the
mild solution of Eq.(1) is exponentially stable in mean square moment provided

5{[M 2 (R2 + R3 + R4 )/γ 2 ] + [M 2 ( +∞ 2

P
k=1 d k ) ]}
2
< 1, (10)
(1 − k)
√
where k := R1 .

Proof. From (10), it is possible to find a suitable number l̃ > 0 small enough such that
P+∞
5M 2 (R2 + R3 + R4 ) 5M 2 ( k=1 d k )2
k+ + < 1.
γ (γ − l̃)(1 − k) (1 − k)
 Exponential behavior | 15

Let assume that µ = γ − l̃ and u(t) be a mild solution of Eq.(1). Then, from (8) we have

E||u(t)||2 1 2 6
||Π(t)[ϕ(0) − q(0, ϕ)]||2

≤ k E|| q(t, u t )|| + 1−k E
Zt 2 Zt 2

+ Π(t − s)f (s, u s )ds + Π(t − s)σ(s)dZ QH (s)

0 0
2
Zt Zs 2 Zt Z
+ Π(t − s) g(s, η, u η )dηds + Π(t − s)h(s, u s , v)N(ds,
e dv)
0 0  0 θ
2

Π(t − t k )I k (u(t−k ))
X
+

0<t k <t
7
X
≤ J j (t).
j=1

By assumption (H10) we get

1
J1 (t) = E||q(t, u t )||2
k

1 (11)
≤ {R E||u t ||2 + ξ1 (t)}
k 1

≤ kE||u t ||2 + λ1 e−γt ,

α1
where λ1 = .
k
According to (H9) and (H10) we obtain
10 10
J2 ≤ E||Π(t)ϕ(0)||2 + E||Π(t)q(0, ϕ)||2
1−k 1−k

10M 2 −2γt 10M 2 −2γt (12)

≤ e E||ϕ(0)||2 + e {R1 E||ϕ||2 + ξ1 (t)}
1−k 1−k

≤ λ2 e−µt ,

10M 2
where λ2 = [E||ϕ(0)||2 + {R1 E||ϕ||2 + α1 }].
1−k
Employing (H9) , (H10) and Hölder’s inequality, we get

Zt
5
J3 (t) = E|| Π(t − s)f (s, u s )ds||2
1−k
0
 t 2
Z
5 −γ(t−s) (13)
≤ E  Me ||f (s, u s )||ds
1−k
0
Zt
5M 2 R2
≤ e−γ(t−s) E||u s ||2 ds + λ3 e−µt
γ (1 − k)
0

5M 2 R3 α3
where λ3 = .
γ (1 − k) γ − µ
16 | Ravikumar Kasinathan et al.

Also with the same arguments, we have

 t 2
Z Zt
J4 (t) ≤ 5  Ne−γ(t−s) g(s, η, x η )dη ds
1−k E
0 0
(14)
Zt
5M 2 R3
≤ γ(1−k)
e−γ(t−s) E||u s ||2 ds + λ4 e−µt ,
0

5M 2 R4 α4
where λ4 = .
γ (1 − k) γ − µ
By Lemma 2.1 and (H11) we have
Zt
5
J5 ≤ M 2 c(H)t2H−1 e−2γ(t−s) ||σ(s)||2L0 ds
1−k Q
0

Zt
−µt 5M 2
≤ e c(H)t2H−1 e−l̃t eγs ||σ(s)||2L0 ds.
1−k Q
0

According to assumption (H11), there exist a constant γ5 > 0 such that, for all t ≥ 0,
Zt
5M 2
c(H)t2H−1 e−l̃t eγs ||σ(s)||2L0 ds ≤ λ5 . (15)
1−k Q
0

Thus
J5 (t) ≤ λ5 e−µt .
and
2
Zt Z

6
J6 (t) ≤ E h(s, u s , v)N(ds,
e dv) 
1−k
0 θ
Zt
6M 2
≤ R4 e−γ(t−s) E kuk2s ds + λ6 e−µt , (16)
γ (1 − k)
0
2
6M α5
where λ6 = γ(1−k) γ−µ .
By using (H7) and (H9) we have
+∞
!2
5M 2
e−2γ(t−t k ) E||u(t−k )||2
X
J7 (t) ≤ di
1−k
i=1

(17)
+∞ +∞
! !
5M 2
e−γ(t−t k ) E||u(t−k )||2 .
X X
≤ di di
1−k
i=1 i=1

The above inequalities (11)-(17) together with Lemma 2.2, imply that

E||u(t)||2 ≤ νe−µt for t ∈ [−τ, 0]

and
Zt
2
2
E||u(t)|| ≤ νe −µt
+ k sup E||u(t + θ)|| + κ̂ e−µ(t−s) sup E||u(t + θ)||2 ds
−τ≤θ≤0 −τ≤θ≤0
0
+∞
ω k e−µ(t−t k ) E||u(t−k )||2 ,
X
+ t≥0
i=1
 Exponential behavior | 17

where
6
!
5M 2 (R2 + R3 + R4 )
λ j , sup E||ϕ(θ)||2
X
κ̂ = and ν = max .
γ (1 − k) −τ≤θ≤0 i=1

And we observe that k + Mk̂ + +∞

P
i=1 ω k < 1.
Thus, the mild solution of Eq.(1) is exponentially stable in mean square moment, since k + µκ̂ + +∞
P
k=1 ω k < 1
and by Lemma 2.2 we have the existence of two positive constants C and r such that E||u(t)||2 ≤ Ce−rt for any
k
e rτ + +∞
P
t ≥ −τ, where θ > 0 is the unique solution to the equation k + µ−θ k=1 ω k = 1 and
C = max{ν, ν(µ−r) } > 0.
k̂e rτ
This completes the proof of the theorem.

4 Illustration
This part consist to make an application of the theory studied above. Consider the impulsive neutral
stochastic partial integrodifferential equations of the form
 "
∂2
d[w(t, ξ ) − β1 (t, w(t − r, ξ ))] = [w(t, ξ ) − β1 (t, w(t − r, ξ ))]


∂ξ 2





Zt


∂2




 + b(t − s) 2 [w(s, ξ ) − β1 (s, w(s − r, ξ ))]ds
∂ξ



0






 Zt #
+β2 (t, w(t − r, ξ )) + β3 (t, s, w(s − r, ξ ))ds dt

(18)


 Z 0
dz), +σ(t)dZ QH (t), 0 ≤ ξ ≤ π, t ≠ t k , t ∈ [0, T],

+ β (t, x(t − s), z) N(ds,

4

 e



θ



w(t, 0) = w(t, π) = 0, 0 ≤ t ≤ T,




β

∆w(t k , ·)(ξ ) = 25 w(t−k , ξ ),

t = t k , k = 1, 2, · · · ,


k



w(t, ξ ) = φ(t, ξ ) ∈ PC([−τ, 0], L2 [0, π]), −τ ≤ t ≤ 0,


where Z QH is Rosenblatt process define on probability space (Ω, F, P) , β1 , β2 : R+ × R 7→ R,

β3 , β4 : R+ × R+ × R 7→ R, b : R+ 7→ R are continuous functions, β5 ≥ 0 constant and
σ : [0, ∞) 7→ L0Q L2 ([0, π]), L2 ([0, π]) .


q
Let H = Y = L2 [0, π] with the norm k · k and e n (x) = 2π sin nx, n = 1, 2, · · · . Then (e n )n∈N is a complete
orthogonal basis in Y. In order to define the operator Q : Y → Y, we chose a sequence (σ n )n≥1 ⊂ R+ and set
P √
Qe n = σ n e n , and assume that ∞ H
n=1 σ n < ∞. Define the process Z by

∞
√
Z H (t) = σn zH
X
n en ,
n=1
n o
where H ∈ ( 21 , 1) and z H
n is a sequence of two-side one-dimensional Rosenblatt process mutually
n∈N
independent.
∂2 1 2
Define the operator A : D(A) ⊂ H → H by A = ∂τ 2 with domain D(A) = H 0 (0, π) ∩ H (0, π).

It is well-known that A is the infinitesimal generator of a strong continuous semigroup {T(t)}t≥0 on H,

which is given by
∞ +∞
2
e−n t hx, e n ie n and Ax = − n2 hx, e n ie n , x ∈ D(A).
X X
T(t)x =
n=1 n=1

We suppose that:
 18 | Ravikumar Kasinathan et al.

(i) There exist a positive constant l1 , 0 < πl21 < 1 such that

β1 (t, y1 ) − β1 (t, y2 ) ≤ l1 |y1 − y2 |, t ≥ 0, y1 , y2 ∈ R.

(ii) There exist a positive constant, l2 > 0 such that

β2 (t, y1 ) − β2 (t, y2 ) ≤ l2 |y1 − y2 |, t ≥ 0, y1 , y2 ∈ R.

(iii) There exist a positive constant, l3 > 0 such that

Zt
 
β3 (t, y1 ) − β3 (t, y2 ) ds ≤ l3 |y1 − y2 |, t ≥ 0, y1 , y2 ∈ R.
0

(iv) There exist a positive constant, l4 > 0 such that

Z
 
β4 (t, y1 , v) − β4 (t, y2 , v) N(ds,
e dv) ≤ l4 |y1 − y2 |, t ≥ 0, y1 , y2 ∈ R.
θ

(v) There exist nonegative real numbers Q1 , Q2 , Q3 , Q4 > 0 and functions ξ1 , ξ2 , ξ3 , ξ4 : [0, ∞) 7→ R+ with
ξ i (t) ≤ P i e−λt (i = 1, 2, 3, 4), P i > 0, such that
2
β1 (t, y) ≤ Q1 |y|2 + ξ1 (t),
2
β2 (t, y) ≤ Q2 |y|2 + ξ2 (t),
Zt 2
β3 (t, s, y)ds ≤ Q3 |y|2 + ξ3 (t),
Z0 2
β4 (t, v, y)N(ds,
e dv) ≤ Q4 |y|2 + ξ4 (t).
θ

(v) The function σ : [0, ∞) 7→ L0Q L2 ([0, π]), L2 ([0, π]) satisfies

Zt Z∞
kσ(s)k2L0 ds < ∞, t ∈ [0, T] and e λs kσ(s)k2L0 ds < ∞.
Q Q
0 0

Let Υ : D(A) ⊂ H → H the operator define by Υ(t)(z) = b(t)Az for t ≥ 0, z ∈ D(A). For ξ ∈ [0, π], we define
the operators q, f : [0, ∞) × H 7→ H, g : [0, ∞) × [0, ∞) × H 7→ H, h : [0, ∞) × θ × H 7→ H and I k : H 7→ H by

G(t, ϕ)(ξ ) = β1 (t, ϕ(ξ )),

F(t, ϕ)(ξ ) = β2 (t, ϕ(ξ )),
Zt Zt
H̃(t, s, ϕ)(ξ )ds = β3 (t, s, ϕ(ξ ))ds,
Z0 0

h(t, ϕ)(ξ ) = β4 (t, ϕ(ξ )),

θ
β5
I k (ϕ)(ξ ) = ϕ(ξ ) (k = 1, 2, · · · ).
k2
If we put (
u(t)(ξ ) = w(t, ξ ), for t ≥ 0, and ξ ∈ [0, π],
φ(θ)(ξ ) = u0 (θ, ξ ), for − τ ≤ θ ≤ 0, and ξ ∈ [0, π],
 Exponential behavior | 19

then, Eq.(18) takes the following abstract form




" Zt
 
 
d u(t) − q(t, u ) = A u(t) − q(t, u ) + Υ(t − s) u(s) − q(s, u s ) ds + f (t, u t )

t t






 0
Zt

 # Z

 + g(t, s, u s )ds dt + h(t, u t , v)N(dt, e dv) + σ(t)dZ QH (t), t ∈ [0, T], t ≠ t k ,

0


 θ
∆u(t) = I k (u(t−k )), t = t k , k = 1, 2, · · · ,




= ϕ(t) ∈ PC [−τ, 0], H , −τ ≤ t ≤ 0,


u0 (t)

(19)
Moreover , if b is bounded and C1 function such that b′ is bounded and uniformly continuous, then (H1)


and (H2) are satisfied, and hence, by Theorem 2.3, Eq.(7) has a resolvent operator Π(t) t≥0 on H. Using
Lemma 5.2 [9], let µ > δ > 1 and b(t) < exp(−βt), for all t ≥ 0. Then the above resolvent operator decays
exponentially to zero. Specifically kΠ(t)k ≤ exp(−at) where a = 1 − 1/δ.
For (t, s, ϕ i ) ∈ R+ × R+ × H, j = 1, 2, respectively by assumptions (i), (ii), (iii) and (iv), we get

kq(t, ϕ1 ) − q(t, ϕ2 )k ≤ l1 kϕ1 − ϕ2 k,

kf (t, ϕ1 ) − f (t, ϕ2 )k ≤ l2 kϕ1 − ϕ2 k,
Zt
 
g(t, s, ϕ1 ) − g(t, s, ϕ2 ) ds ≤ l3 kϕ1 − ϕ2 k,
Z0 h i
h(t, ϕ1 , v) − h(t, ϕ2 , v)N(dt,
e dv) ≤ l4 kϕ1 − ϕ2 k,
Z
kq(t, ϕ1 )k2 ≤ Q1 kϕ1 k2 + ξ1 (t),
kf (t, ϕ1 )k2 ≤ Q2 kϕ1 k2 + ξ2 (t),
Zt 2
g(t, s, ϕ1 )ds ≤ Q3 kϕ1 k2 + ξ3 (t),
Z0 2
h(t, ϕ1 , v)ds ≤ Q4 kϕ1 k2 + ξ4 (t).
θ

We have also that

I k (ϕ1 ) − I k (ϕ2 ) ≤ d k kϕ1 − ϕ2 k,
∞ ∞
β5
X X β5
where d k = k2
, k=1,2,. . . and dk = < ∞.
k2
k=1 k=1
Thus, all assumptions of Theorem 3.1 are fulfilled. Therefore, the existence of a mild solution of Eq.(18)
follows. In addition, by Theorem 3.2, we easily see that the mild solution of Eq.(18) is exponentially stable
in the 2nd moment.

5 Conclusion
In this article, we showed existence and unicity of mild solution to Eq.(1) by using Banach fixed point
theorem. Further, we investigated exponential stability of mild solutions. The last part of this paper is
devoted to an example to illustrate our results.
20 | Ravikumar Kasinathan et al.

References
[1] P. Abry, V. Pipiras, Wavelet-based synthesis of the Rosenblatt process. Signal Processing. 86(2006), 2326-2339.
[2] H. M. Ahmed, Fractional neutral evolution equations with nonlocal conditions. Advances in Difference Equations. (2013),
2013-117.
[3] H. M. Ahmed, M. M. El-Borai, Hilfer fractional stochastic integro-differential equations. Applied Mathematics and
Computation. 331(2018), 182-189.
[4] P. Ali, K. Shah, F. Jarad, V. Gupta, T. Abdeljawad, Existence and stability analysis to a coupled system of implicit type
impulsive boundary value problems of fractional-order differential equations. Advances in Difference Equations. (2019),
1-21.
[5] G. Arthi, K. Balachandran, Controllability of damped second-order neutral functional differential systems withimpulses.
Taiwanese Journal of Mathematics. 16(2012), 89-106.
[6] J. M. Bardet, C. A Tudor, A wavelet analysis of the Rosenblatt process: chaos expansion and estimation of the
self-similarity parameter. Stochastic Processes and their Applications. 120(2010), 2331-2362.
[7] S. Boudrahem, P. R. Rougier, Relation between postural control assessment with eyes open and centre of pressure visual
feed back effects in healthy individuals. Experimental Brain Research. 195(2009), 145-152.
[8] F. Comte, E. Renault, Long memory continuous time models. Journal of Econometrics. 73(1996), 101-149.
[9] M. Dieye, M. A. Diop, K. Ezzinbi, On exponential stability of mild solutions for some stochastics partial integrodifferential
equations. Statistics and Probability Letters. 123(2017), 61-76.
[10] R. L. Dobrushin, P. Major, Non-central limit theorems for non-linear functionals of Gaussian fields. Zeitschrift fur
Wahrscheinlichkeits theorie and Verwandte Gebiete. 50(1979), 27-52.
[11] M. M. El-Borai, O. L. Moustafa, H. M. Ahmed, Asymptotic stability of some stochastic evolution equations. Applied
Mathematics and Computation. 144(2003), 273-286.
[12] J. Golec, S. Sathananthan, Strong approximations of stochastic integrodifferential equations. Dynamics of Continuous,
Discrete and Impulsive Systems, Series B. 8(2001), 139-151.
[13] R. C. Grimmer, Resolvent opeators for integral equations in a Banach space. Transactions of the American Mathematical
Society. 273(1982), 333-349.
[14] V. Gupta, S. N. Bora, J. J. Nieto, Dhage iterative principle for quadratic perturbation of fractional boundary value problems
with finite delay. Mathematical Methods in the Applied Sciences. (2019), 1-12.
[15] V. Gupta, J. Dabas, Existence results for fractional order boundary value problem with integrable impulse, Dynamics of
Continuous, Discrete and Impulsive Systems Series A:Mathematical Analysis. 25(2018), 267-285.
[16] V. Gupta, J. Dabas, Michal Feckan, Existence results of solutions for impulsive fractional differential equations.
Nonautonomous Dynamical Systems. 5(2018), 35-51.
[17] J. K. Hale, K. R. Meyer, A class of functional equations of neutral type. Memoirs of the American Mathematical Society.
76(1967), 1-65.
[18] J. K. Hale, S. M. Verduyn Lunel, Introduction to Functional Differential Equations. Springer-Verlag, New York, 1993.
[19] I. Kruk, F. Russo, C. A. Tudor, Wiener integrals, Malliavin calculus and covariance measure structure. Journal of Functional
Analysis. 249(2007), 92-142.
[20] V. B. Kolmanovskii, V. R. Nosov, Stability of neutral type functional differential equations. Nonlinear Analysis. 6(1982),
873-910.
[21] A. N. Kolmogorov, The Wiener spiral and some other interesting curves in Hilbert space. Doklady Akademii Nauk SSSR.
26(1940), 115-118.
[22] V. Laksmikantham, M. R.M Rao, Theory of Integro-Differential Equations. Gordon and Breach Publishers, London, 1995.
[23] N. Leonenko, V. Ahn, Rate of convergence to the Rosenblatt distribution for additive functionals of stochastic processes
with long-range dependence. Journal of Applied Mathematics and Stochastic Analysis. 14(2001), 27-46.
[24] J. J. Levin, J. A. Nohel, The integro-differential equations of a class of nuclear reactors with delayed neutrons. Archive for
Rational Mechanics and Analysis. 31(1968), 151-172.
[25] Y. K. Ma, G. Arthi, S. M. Anthoni, Exponential stability behavior of neutral stochastic integrodifferential equations with
fractional Brownian motion and impulsive effects. Advances in Difference Equations. (2018), 2018-110.
[26] B. B. Mandelbrot, J. Van Ness, Fractional Brownian motions, fractional noises and applications. SIAM Review. 10(1968),
422-437.
[27] M. Maejima, C. A. Tudor, Wiener integrals with respect to the Hermite process and a non central limit theorem. Stochastic
Analysis and Applications. 25(2007), 1043-1056.
[28] M. Maejima, C. A. Tudor, Selfsimilar processes with stationary increments in the second Wiener chaos. Probability and
Mathematical Statistics. 32(2012), 167-186.
[29] M. Maejima, C. A. Tudor, On the distribution of the Rosenblatt process. Statistics and Probability Letters. 83(2013),
1490-1495.
[30] P. E Protter, Stochastic integration and differential equations. 2nd edition, Springer, New York, 2004.
 Exponential behavior | 21

[31] V. Pipiras, M. S. Taqqu, Regularization and integral representations of Hermite processes. Statistics and Probability
Letters. 80(2010), 2014-2023.
[32] M. Taqqu, Weak convergence to the fractional Brownian motion and to the Rosenblatt process. Zeitschrift fur
Wahrscheinlichkeits theorie and Verwandte Gebiete. 31(1975), 287-302.
[33] S. Tindel, C. A. Tudor, F. Viens, Stochastic evolution equations with fractional Brownian motion. Probability Theory and
Related Fields. 127(2003), 186-204.
[34] C. A. Tudor, Analysis of the Rosenblatt process. ESAIM:Probability and Statistics. 12(2008), 230-257.
[35] C. A. Tudor, F. G. Viens, Variations and estimators for self-similarity parameters via Malliavin calculus. The Annals of
Probability. 37(2009), 2093-2134.
[36] M. Veillette, M. S. Taqqu, Properties and numerical evaluation of the Rosenblatt distribution. Bernoulli. 19(2013),
982-1005.
[37] H. Yang, F. Jiang, Exponential stability of mild solutions to impulsive stochastic neutral partial differential equations with
memory. Advances in Difference Equations. 148(2013), 2013.