See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/352879108

APPROXIMATE CONTROLLABILITY OF NON-INSTANTANEOUS IMPULSIVE

FRACTIONAL EVOLUTION EQUATIONS OF ORDER 1 < α < 2 WITH STATE-
DEPENDENT DELAY IN BANACH SPACES

Preprint · June 2021

CITATIONS READS
0 143

3 authors:

Sumit Arora Manil T. Mohan

Indian Institute of Technology Roorkee Indian Institute of Technology Roorkee
13 PUBLICATIONS   54 CITATIONS    120 PUBLICATIONS   509 CITATIONS   

SEE PROFILE SEE PROFILE

Jaydev Dabas
Indian Institute of Technology Roorkee
50 PUBLICATIONS   599 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Large deviations of stochastic differential and partialdifferential equations View project

Random Dynamics of Partial Differential Equations View project

All content following this page was uploaded by Manil T. Mohan on 01 July 2021.

The user has requested enhancement of the downloaded file.

 APPROXIMATE CONTROLLABILITY OF NON-INSTANTANEOUS
IMPULSIVE FRACTIONAL EVOLUTION EQUATIONS OF ORDER
1 < α < 2 WITH STATE-DEPENDENT DELAY IN BANACH SPACES

S. ARORA1 , MANIL T. MOHAN2 AND J. DABAS3*

arXiv:2106.15122v1 [math.OC] 29 Jun 2021

Abstract. The current article examines the approximate controllability problem for non-
instantaneous impulsive fractional evolution equations of order 1 < α < 2 with state-
dependent delay in separable reflexive Banach spaces. In order to establish sufficient con-
ditions for the approximate controllability of our problem, we first formulate the linear-
regulator problem and obtain the optimal control in feedback form. By using this optimal
control, we deduce the approximate controllability of the linear fractional control system of
order 1 < α < 2. Further, we derive sufficient conditions for the approximate controllability
of the nonlinear problem. Finally, we provide a concrete example to validate the efficiency
of the derived results .

1. Introduction
α
Fractional evolution equations involve the derivatives of the form dtd α , where α > 0 is not
necessarily an integer, have received much attention from researchers. This rising interest is
motivated both by important applications of the theory, and by the difficulties involved in the
mathematical structure. Fractional evolution equations appear in many physical phenomena
arising from various scientific fields including analysis of viscoelastic materials, electrical
engineering, control theory of dynamical systems, electrodynamics with memory, quantum
mechanics, heat conduction in materials with memory, signal processing, economics, and
many other fields (cf. [21, 23, 44, 48, 52], etc). For an extensive study of the fractional
differential equations (FDEs), we refer the interested readers to [23, 48], etc.
The real-world phenomena such as natural disasters, harvesting, shocks, etc, experience
unpredictable changes in their states for negligible time duration. These sudden changes are
relevant and interesting, and well approximated by instantaneous impulses. Such processes
are conveniently described by instantaneous impulsive evolution equations. The theory of

1,3
Department of Applied Sciences and Engineering, Indian Institute of Technology Roorkee-IIT
Roorkee, Roorkee, 247667, India.
e-mail1 : sumit@as.iitr.ac.in.
e-mail3 : jay.dabas@gmail.com.
2
Department of Mathematics, Indian Institute of Technology Roorkee-IIT Roorkee, Roorkee, 247667,
India.
e-mail2 : maniltmohan@ma.iitr.ac.in, maniltmohan@gmail.com.
*
Corresponding author.
Key words: approximate controllability, non-instantaneous impulses, state-dpendent delay, fractional
derivative, Mainardi’s Wright-type function.
Mathematics Subject Classification (2020): 34K06, 34A12, 37L05, 93B05.

1
2 S. ARORA, M. T. MOHAN AND J. DABAS

instantaneous impulsive evolution equations has found various applications such as opti-
mal control models in economics, threshold, pharmacokinetics, bursting rhythm models in
medicine and frequency modulated systems etc (cf. [8, 28, 35, 63] etc). Moreover, the
evolution processes in pharmacotherapy, such as the consequent absorption of the body, in
hemodynamical steadiness of a person and the insulin distribution in the bloodstream are
moderate and continuous. Hence, the dynamics of such phenomena cannot be interpreted
by the instantaneous impulsive systems. Therefore, this situation can be appropriately esti-
mated as an impulsive action that starts suddenly and continuously active over a finite time
interval. Thus, the dynamics of such processes are suitably described by non-instantaneous
impulsive systems.
Hernández et. al. [20], first introduced a class of impulses that are suddenly active for
an arbitrary fixed point and remains in action on a finite time interval. Such impulses are
called non-instantaneous impulses. Then, they developed the existence of solutions for non-
instantaneous impulsive differential equations. Later, Wang et. al. [57, 60] extended this
model to two general classes of impulsive differential equations, which are very useful to
characterize certain dynamics of evolutionary processes in pharmacotherapy. For the study
of non-instantaneous impulsive systems, one can refer to [58, 61], etc, and the references
therein.
On the other hand, there are several real-world phenomena in which the current state of
a system depends on the past states. The dynamics of such types of processes are suitably
analyzed by delay (functional) evolution system, which naturally occurs in logistic reaction-
diffusion processes with delay, ecological models, neural network, inferred grinding models,
etc (cf. [9, 30, 36], etc). Moreover, the evolution equations with state-dependent delays are
more frequent and widely applicable, see for example [4, 9], etc, and the references therein.
The notion of controllability is one of the fundamental concepts in mathematical control
theory and play a significant role in many control problems such as, irreducibility of transi-
tion semigroups [12], the optimal control problems [6], stabilization of unstable systems via
feedback control [7] etc. Controllability (exact and approximate) means that the solution
of a considered dynamical control system can steer from an arbitrary initial state to a de-
sired final state by using some suitable control function. For the infinite dimensional control
systems, the problem of approximate controllability is more substantial and is having broad
range of applications, see for instance, [33, 55, 56, 67], etc. In the past few decades, the prob-
lem of approximate controllability for different kinds of dynamical systems (in Hilbert and
Banach spaces) such as fractional evolution equations, Sobolev type systems, delay (func-
tional) differential equations, impulsive systems, stochastic differential equations, etc, has
been widely studied by many researchers and they have produced excellent results, see for
instance, [1, 3, 16, 33, 46, 64] etc, and the references therein.
It is well known that the slow diffusion processes are described by first-order evolution
equations while the fast diffusion processes are characterized by second-order evolution sys-
tems. Similarly, FDEs of order α ∈ (0, 1) stand for slow anomalous diffusion processes
(subdiffusion process), while the fast anomalous diffusion processes (superdiffusion process)
are suitably modeled by FDEs of order α > 1. Therefore, by this reason, the theory for frac-
tional evolution equations of order α ∈ (0, 1) is discrete from that for fractional evolution
equations of order α > 1, and it inspired the researchers to study both theories separately
(see [26, 27], etc for more details). In the past two decades, a good number of results came out
on the existence of solutions and approximate controllability for fractional systems of order
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 3

α ∈ (0, 1) in Hilbert and Banach spaces, see for example, [10, 11, 15, 25, 31, 32, 59, 64], etc,
and the references therein. Recently, many authors established the existence of mild solu-
tions and approximate controllability for fractional evolution equations of order α ∈ (1, 2) by
invoking the theory of sectorial operators via suitable fixed point theorems, see for instance,
[40, 47, 49, 50] etc.
Recently, a few developments have been reported on the existence and controllability of
fractional evolution equations of order 1 < α < 2 by applying the cosine and sine family of
bounded linear operators (cf. [19, 42, 66] etc). A new concept of mild solutions for fractional
evolution systems with order α ∈ (1, 2) in Banach spaces by using the Laplace transform and
Mainardi’s Wright-type function is introduced in [66]. Henrı́quez et.al [18] investigated the
existence of a classical solution for the fractional homogeneous and inhomogeneous abstract
Cauchy problems of order α ∈ (1, 2) by using the α-resolvent family corresponding to the
homogeneous Cauchy problem. Raja et. al. [43] formulated sufficient conditions for the
approximate controllability of fractional differential equation of order 1 < α < 2 in Hilbert
spaces by applying the concept of cosine and sine family of operators via Schauder’s fixed
point theorem. Raja and Vijaykumar [41] derived sufficient conditions for the approximate
controllability of fractional integro-differential system of order α ∈ (1, 2) in Hilbert spaces.
Later, Vijaykumar et.al. in [62] demonstrated the approximate controllability of Sobolev-
type Volterra-Fredholm integro-differential equations of order 1 < α < 2 in Hilbert spaces
by invoking the Krasnoselskii fixed point theorem.
To the best of our knowledge, the approximate controllability for impulsive fractional
differential equations of order 1 < α < 2 with state-dependent delay in Banach spaces
by using the idea of cosine and sine family of operators is not investigated yet, and it
motivated us to encounter this problem. In this paper, we consider semilinear impulsive
fractional differential equations of order 1 < α < 2 with non-instantaneous impulses and
state-dependent delay in separable reflexive Banach spaces. We derive sufficient conditions
for the approximate controllability of the considered system by applying the concept of
α/2-resolvent family (defined in (2.2), which is related to the strongly continuous consine
family generated by the operator A), resolvent operator condition and Schauder’s fixed point
theorem. Moreover, we discuss the construction of a feedback control and the approximate
controllability of the linear fractional control system of order 1 < α < 2 in details (see,
Section 3), which lacks in the available literature. Furthermore, due to the observation
related to characterization of the phase space for impulsive systems discussed in [17], we
replace the uniform norm on the phase space by the integral norm (see Example 2.11),
Let us consider the following non-instantaneous impulsive fractional order system with
state-dependent delay:
p

[
C α




 D0,t x(t) = Ax(t) + Bu(t) + f (t, x̺(t,xt ) ), t ∈ (sj , τj+1 ] ⊂ J = [0, T ],

 j=0

x(t) = hj (t, x(τj− )), t ∈ (τj , sj ], j = 1, . . . , p, (1.1)

x′ (t) = h′j (t, x(τj− )), t ∈ (τj , sj ], j = 1, . . . , p,





x0 = ψ ∈ B, x′ (0) = η,



where
C
• Dα0,t is the Caputo fractional derivative of order α with 1 < α < 2,
4 S. ARORA, M. T. MOHAN AND J. DABAS

• the operator A : D(A) ⊂ X → X is linear (not necessarily bounded), where X is a

separable reflexive Banach space having a strictly convex dual X∗ ,
• the linear operator B : U → X is bounded with kBkL(U;X) = M̃ and the control
function u ∈ L2 (J; U) (admissible control class), where U is a separable Hilbert
space,
• the appropriate function f : J × B → X, where B is a phase space satisfying some
axioms which will be provided in next section,
• the functions hj : [τj , sj ] × X → X and their derivatives h′j : [τj , sj ] × X → X for
j = 1, . . . , p, denote the non-instantaneous impulses and the fixed points sj and τj
satisfy 0 = τ0 = s0 < τ1 ≤ s1 ≤ τ2 < . . . < τp ≤ sp ≤ τp+1 = T , the values x(τj+ ) and
x(τj− ) stand for the right and left limit of x(t) at the point t = τj , respectively and
satisfy x(τj− ) = x(τj ), for j = 1, . . . , p,
• the function xt : (−∞, 0] → X with xt (θ) = x(t + θ) and xt ∈ B, for each t ∈ J. The
function ̺ : J × B → (−∞, T ] is continuous.
The rest of the article is structured as follows: In section 2, we recall some basic definitions
of fractional calculus, property of cosine family and the phase space axioms. Moreover,
we also provide the assumptions and important results which is useful in the subsequent
sections. Section 3 is reserved for the approximate controllability of the linear fractional
control system of order 1 < α < 2. For this, first we formulate the linear regulator problem,
show the existence of an optimal control (Theorem 3.1), and then we derive an explicit
form of the optimal control in the feedback form in Lemma 3.3. Further, we examine the
approximate controllability of the linear fractional control system (3.2), by using the optimal
control (Lemma 3.5). In section 4, we first prove the existence of a mild solution for the
control system (1.1) by applying Schauder’s fixed point theorem (Theorem 4.2), and then
demonstrate the approximate controllability of the system (1.1) in Theorem 4.3. In the final
section, we present an example to support the application of the developed theory.

2. Preliminaries
In the present section, we introduce some standard notations, basic definitions and facts
about fractional calculus. Moreover, we also recall some important properties of cosine and
sine families, and also provide the axioms of phase space.
Let AC([a, b]; R) denote the space of all mappings from [a, b] to R, which is absolutely
continuous and ACn ([a, b]; R), for n ∈ N, represent the space of all functions f : [a, b] → R,
which have continuous derivatives up to order n − 1 with f (n−1) ∈ AC([a, b]; R).
First, we recall some basic definitions from fractional calculus and the details can be found
in [23, 48], etc.

Definition 2.1. The fractional integral of order β > 0 for a function f ∈ L1 ([a, b]; R),
a, b ∈ R with a < b, is given as
t
1 f (s)
Z
Iaβ f (t) := ds, for a.e. t ∈ [a, b],
Γ(β) a (t − s)1−β
R∞
where Γ(α) = 0
tα−1 e−t dt is the Euler gamma function.
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 5

Definition 2.2. The Riemann-Liouville fractional derivative of order β > 0 for a function
f ∈ ACn ([a, b]; R) is defined as
1 dn t
Z
L β
Da,t f (t) := (t − s)n−β−1 f (s)ds, for a.e. t ∈ [a, b],
Γ(n − β) dtn a
where n − 1 < β < n.
Definition 2.3. The Caputo fractional derivative of a function f ∈ ACn ([a, b]; R) of order
β > 0 is defined as
" n−1 (k)
#
C β
X f (a)
Da,t f (t) := L Dβa,t f (t) − (x − a)k , for a.e. t ∈ [a, b].
k=1
k!

Remark 2.4 ([48]). If a function f ∈ ACn ([a, b]; R), then the Caputo fractional derivative
of f can be written as
Z t
C β 1
Da,t f (t) = (t − s)n−β−1 f (n) (s)ds, for a.e. t ∈ [a, b], n − 1 < β < n.
Γ(n − β) a
2.1. The strongly continuous cosine family: In this subsection, we first construct a
strongly continuous cosine family {C(t) : t ∈ R} in X generated by A, and then review some
important properties of the cosine and sine families. In order to construct a strongly continu-
ous cosine family {C(t) : t ∈ R}, we impose the following assumptions on the linear operator
A : D(A) → X.
Assumption 2.5. The operator A fulfills the following conditions:
(R1) The linear operator A is closed with dense domain D(A) in X.
(R2) For real λ, λ > 0, λ2 is in the resolvent set ρ(A) of A. The resolvent R(λ2 ; A) exists,
strongly infinitely differentiable, and satisfies
 
d k Mk!
(λR(λ2 ; A)) ≤ k+1 ,

dλ λ


L(X)

for k = 0, 1, 2, . . . .
(R3) R(λ2 ; A) is compact for some λ with λ2 ∈ ρ(A).
Definition 2.6 ([53]). A family of bounded linear operators {C(t) : t ∈ R} in a Banach
space X is called a strongly continuous cosine family if
(1) C(s + t) + C(s − t) = 2C(s)C(t), for s, t ∈ R;
(2) C(0) = I, where I represents the identity operator;
(3) C(t)x is continuous in t on R, for each fixed x ∈ X.
The family {S(t) : t ∈ R} on X defined by
Z t
S(t)x = C(s)xds, x ∈ X, t ∈ R.
0
is called strongly continuous sine family.
The infinitesimal generator A of a strongly continuous cosine family {C(t) : t ∈ R} is
defined by
d2 
Ax = 2 C(t)x , x ∈ D(A),

dt t=0
6 S. ARORA, M. T. MOHAN AND J. DABAS

where
n o
D(A) = x ∈ X : C(t)x is two times continuously differentiable function with respect to t ,

equipped with the graph norm

kxkD(A) = kxkX + kAxkX , x ∈ D(A).
We also define the set
n o
E = x : C(t)x is continuously differentiable function with respect to t ,

with the norm

kxk1 = kxkX + sup kAS(t)xkX , x ∈ E,
0≤t≤1

forms a Banach space, see [24].

Lemma 2.7 (Proposition 2.7, [54]). If the assumptions (R1)-(R2) hold true. Then the
operator A generates a strongly continuous cosine family {C(t) : t ∈ R}, which is uniformly
bounded, that is, there exists a constant M ≥ 1 such that kC(t)kL(X) ≤ M.
Lemma 2.8 (Proposition 5.2, [54]). Suppose that the operator A generates a strongly con-
tinuous cosine family {C(t) : t ∈ R} on X. Let {S(t) : t ∈ R} be the associated sine family
on X. Then the following are equivalent:
(i) The operator S(t) is compact for every t ∈ R.
(ii) Assumption (R3) holds.
Lemma 2.9 (Propositions 2.1 and 2.2, [53]). Let the operator A generates a strongly con-
tinuous cosine family {C(t) : t ∈ R}, which is uniformly bounded in X, and let {S(t) : t ∈ R}
be the associated sine family on X. Then the following are true:
(i) kS(t2 ) − S(t1 )kL(X) ≤ M|t2 − t1 |, for all t1 , t2 ∈ R;
(ii) If x ∈ E, then S(t)x ∈ D(A) and dtd C(t)x = AS(t)x.
2.2. Phase space: Let us provide the definition of the phase space B in axiomatic form (cf.
[22]) and appropriately modify for the impulsive systems (cf. [37]). Specifically, the phase
space B is a linear space of all mapping from (−∞, 0] into X equipped with the seminorm
k·kB and satisfying the following axioms:
(A1) If x : (−∞, ν + σ) → X, σ > 0, such that xν ∈ B and x|[ν,ν+σ] ∈ PC([ν, ν + σ]; X),
then for every t ∈ [ν, ν + σ], the following conditions hold true:
(i) xt is in B;
(ii) kxt kB ≤ P(t − ν) sup{kx(s)kX : ν ≤ s ≤ t} + Q(t − ν)kxν kB , where P, Q :
[0, ∞) → [0, ∞) such that P is continuous, Q is locally bounded, and both P, Q
are independent of x(·).
(A2) The space B is complete.
For any ψ ∈ B, the function ψt , t ≤ 0, is defined by ψt (θ) = ψ(t + θ), θ ∈ (−∞, 0]. If the
function x(·) fulfills the axiom (A1) with x0 = ψ, then we can extend the mapping t 7→ xt
by taking xt = ψt , t ≤ 0, to the whole interval (−∞, T ]. Moreover, let us define a set
R(̺− ) = {̺(s, ψ) : ̺(s, ψ) ≤ 0, for (s, ψ) ∈ J × B},
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 7

where the function ̺ is the same as the one defined in section 1. Further, we assume that
the function ψt : R(̺− ) → B is continuous in t, and there exists a continuous and bounded
function Θ ψ : R(̺− ) → (0, ∞) such that
kψt kB ≤ Θ ψ (t)kψkB .
Let us define the set
n
PC(J; X) := x : J → X : x|t∈Ii ∈ C(Ii ; X), Ii := (τi , τi+1 ], i = 0, 1, . . . , p,
o
x(τi+ ) and x(τi− ) exist for each i = 1, . . . , p, and satisfy x(τi ) = x(τi− ) ,

with the norm kxkPC(J;X) := supkx(t)kX .

t∈J

Lemma 2.10 (Lemma 2.3, [46]). Let x : (−∞, T ] → X be a function such that x0 = ψ and
x|J ∈ PC(J; X). Then
kxs kB ≤ K1 kψkB + K2 sup{kx(θ)kX : θ ∈ [0, max{0, s}]}, s ∈ R(̺− ) ∪ J,
where
K1 = sup Θ ψ (t) + sup Q(t), K2 = sup P(t).
t∈R(̺−) t∈J t∈J

Example 2.11. Let B = PC g (X) be the space consisting of all mappings φ : (−∞, 0] → X
R 0 kφ(θ)k
such that φ is left continuous, φ|[−q,0] ∈ PC([−q, 0]; X), for q > 0, and −∞ g(θ) X dθ < ∞.
The norm k·kB is defined as
Z 0
kφ(θ)kX
kφkB := dθ, (2.1)
−∞ g(θ)
where the function g : (−∞, 0] → [1, ∞) satisfies the following conditions:
(1) g(·) is continuous with g(0) = 1, and lim g(θ) = ∞,
θ→−∞
g(t+θ)
(2) the function O(t) := sup g(θ)
is locally bounded for t ≥ 0.
−∞<θ≤−t

2.3. Mild solution, resolvent operator and assumptions. In order to define the mild
solution of the system (1.1), we first consider Mainardi’s Wright-type function Mυ (cf.
[34],[39])
∞
X (−z)k
Mυ (z) = , υ ∈ (0, 1), z ∈ C.
k=0
k!Γ(1 − υ(n + 1))
Note that, for any t > 0, Mυ (t) ≥ 0 and it satisfies the following:
Z ∞
Γ(1 + δ)
sc Mυ (s)ds = , for − 1 < c < ∞.
0 Γ(1 + υδ)
For any x ∈ X and γ = α2 , we define
Z ∞
Cγ (t) := Mγ (θ)C(tγ θ)dθ, (2.2)
0
Z t
Tγ (t) := Cγ (s)ds, (2.3)
0
8 S. ARORA, M. T. MOHAN AND J. DABAS
Z ∞
Sγ (t) := γθMγ (θ)S(tγ θ)dθ. (2.4)
0
Next, we discuss some important properties of the operators Cγ (t), Tγ (t) and Sγ (t) for t ≥ 0
in the following lemma.
Lemma 2.12. ([66]) The operators Cγ (t), Tγ (t) and Sγ (t) have the following properties:
(i) The operators Cγ (t), Tγ (t) and Sγ (t) are linear. Moreover, for all t ≥ 0,
M γ
kCγ (t)kL(X) ≤ M, kTγ (t)kL(X) ≤ Mt, kSγ (t)kL(X) ≤ t .
Γ(2γ)
(ii) For all t ≥ 0, the operator Cγ (t) is strongly continuous and the operators Tγ (t), Sγ (t)
and tγ−1 Sγ (t) are uniformly continuous.
(iii) If S(t) is compact for t ≥ 0, then the operator Sγ (t) is also compact for t ≥ 0.
Definition 2.13. A function x(·; ψ, η, u) : (−∞, T ] → X is called a mild solution of (1.1),
if x(t) = ψ(t) on t ∈ (−∞, 0] (where ψ ∈ B) and x(·) satisfies the following:
 Z t
(t − s)γ−1 Sγ (t − s) Bu(s) + f (s, x̺(s,xs ) ) ds,
 



 Cγ (t)ψ(0) + Tγ (t)η +

 0




 t ∈ [0, τ1 ],
hj (t, x(τj− )),




 t ∈ (τj , sj ], j = 1, . . . , p,
 − ′
Cγ (t − sj )hj (sj , x(τj )) + Tγ (t − sj )hj (sj , x(τj− ))

x(t) = Z sj (2.5)

(sj − s)γ−1 Sγ (sj − s) Bu(s) + f (s, x̺(s,xs) ) ds
 
−






 0

 Z t
γ−1
  


 + (t − s) Sγ (t − s) Bu(s) + f (s, x̺(s,x s ) ) ds,
0



t ∈ (sj , τj+1 ], j = 1, . . . , p.


Definition 2.14. The system (1.1) is said to be approximately controllable on J, if for any
initial function ψ ∈ B, xT ∈ X and given ǫ > 0, there exist a control function u ∈ L2 (J; X)
such that
kx(T ) − xT kX ≤ ǫ,
where x(·) is the mild solution of the system (1.1) with the control u(·).
To investigate the approximate controllability of the system (1.1), we first introduce the
following operators:
 Z T
T



 L0 u := (T − t)γ−1 Sγ (T − t)Bu(t)dt,

 0
Z T
(2.6)

 ΦT0 := (T − t)γ−1 Sγ (T − t)BB∗ Sγ (T − t)∗ dt,


 0
R(λ, ΦT0 ) := (λI + ΦT0 J )−1 , λ > 0,


where B∗ and Sγ (t)∗ represent the adjoint operators of B and Sγ (t), respectively. It is
straightforward that the operators LT0 and ΦT0 are linear and bounded. Moreover, the map-
∗
ping J : X → 2X is defined as
J [x] = {x∗ ∈ X∗ : hx, x∗ i = kxk2X = kx∗ k2X∗ }, for all x ∈ X,
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 9

which is known as the duality mapping. The notation h·, ·i denotes the duality pairing
between X and its dual X∗ . Since X is a reflexive Banach space, then X∗ becomes strictly
convex (see [5]), which ensures that the mapping J is strictly monotonic, bijective and
demicontinuous, that is,
w
⇀ J [x] in X∗ as k → ∞.
xk → x in X implies J [xk ] −
Remark 2.15. If X is a separable Hilbert space, which is identified by its own dual, then
the resolvent operator is defined as R(λ, ΦT0 ) := (λI + ΦT0 )−1 , λ > 0.
Lemma 2.16 (Lemma 2.2 [33]). For λ > 0 and every h ∈ X, the equation
λzλ + ΦT0 J [zλ ] = λh, (2.7)
has a unique solution which is given as
zλ (h) = λ(λI + ΦT0 J )−1 (h) = λR(λ, ΦT0 )(h)
and satisfies the following
kzλ (h)kX = kJ [zλ (h)]kX∗ ≤ khkX . (2.8)

Proof. Since the operator ΦT0 is linear and bounded, then proceeding similar way as in the
proof of Lemma 2.2 [33], one can complete the proof. 
We need to impose the following assumptions to obtain the approximate controllability of
the system (1.1).
Assumption 2.17. (H0) For every h ∈ X, zλ = zλ (h) = λR(λ, ΦT0 )(h) → 0 as λ ↓ 0 in
strong topology, where zλ (h) is a solution of the equation (2.7).
(H1) (i) Let x : (−∞, T ] → X be such that x0 = ψ and x|J ∈ PC(J; X). The mapping
t 7→ f (t, x̺(t,xt ) ) is strongly measurable on J and the function f (t, ·) : B → X is
continuous for a.e. t ∈ J. Also, the map t 7→ f (s, xt ) is continuous on R(̺− )∪J,
for every s ∈ J.
(ii) For each positive integer r, there exists a constant δ ∈ [0, γ] and a function
1
φr ∈ L δ (J; R+ ), such that
sup kf (t, ψ)kX ≤ φr (t), for a.e. t ∈ J and ψ ∈ B,
kψkB ≤r

with
kφr k 1
L δ (J;R+ )
lim inf = ζ < ∞.
r→∞ r
(H2) The impulses hj : [τj , sj ] × X → X, for j = 1, . . . , p, are such that
(i) the impulses hj (·, x) : [τj , sj ] → X are continuously differentiable for each x ∈ X,
(ii) for all t ∈ [τj , sj ], the impulses hj (t, ·) : X → X and their derivatives h′j (t, ·) :
X → X are completely ′ continuous,

(iii) khj (t, x)kX ≤ κj , hj (t, x) X ≤ ϑj , for all x ∈ X, t ∈ [τj , sj ], j = 1, · · · , p,
where κj ’s and ϑj ’s are positive constants.
The following version of the discrete Gronwall-Bellman lemma (cf. [65]) is used in the
sequel.
10 S. ARORA, M. T. MOHAN AND J. DABAS

Lemma 2.18. If {fn }∞ ∞ ∞

n=0 , {gn }n=0 and {wn }n=0 are non-negative sequences and
X
fn ≤ gn + wk fk , for n ≥ 0,
0≤k<n

then !
X X
fn ≤ gn + gk wk exp gj , for n ≥ 0.
0≤k<n k<j<n

3. Fractional order linear control problem

This section deals with the linear fractional control problem and investigate the approxi-
mate controllability of the linear system. To achieve this goal, first we consider the linear-
regulator problem with the cost functional defined as
Z T
2
G (x, u) = kx(T ) − xT kX + λ (T − t)γ−1 ku(t)k2U dt, (3.1)
0
where γ = α2 with α ∈ (1, 2), λ > 0, xT ∈ X and x(·) is the solution of the linear fractional
control system: (
C α
D0,t x(t) = Ax(t) + Bu(t), t ∈ J,
(3.2)
x(0) = v, x′ (0) = w,
with the control u(·) ∈ U. Since Bu ∈ L1 (J; X) for any u ∈ L2 (J; U) = Uad (admissible
control class), then, (t − s)γ−1 Sγ (t − s)Bu(s) is integrable. Hence, the above linear system
possess a unique mild solution x ∈ C(J; X) (see section 4.2, Chapter 4, [38]) given by
Z t
x(t) = Cγ (t)v + Tγ (t)w + (t − s)γ−1 Sγ (t − s)Bu(s)ds, t ∈ J. (3.3)
0
Next, we define the admissible class Aad for the optimal control problem (3.1)-(3.2) as
Aad = {(x, u) : x is the unique mild solution of (3.2) with the control u ∈ Uad }.
For any u ∈ L2 (J; U), the existence of a unique mild solution ensures that the set Aad is
nonempty. Our next aim is to prove the existence of an optimal pair, which minimize the
cost functional (3.1).
Theorem 3.1. For given v, w ∈ X, there exists a unique optimal pair (x0 , u0) ∈ Aad of the
problem:
min G (x, u). (3.4)
(x,u)∈Aad

Proof. Let us assume

L := inf G (x, u).
u∈Uad
Since 0 ≤ L < +∞, there exists a minimizing sequence {un }∞
n=1 ∈ Uad such that

lim G (xn , un ) = L,
n→∞

where (xn , un ) ∈ Aad , for each n ∈ N. Moreover, xn (·) satisfies

Z t
n
x (t) = Cγ (t)v + Tγ (t)w + (t − s)γ−1 Sγ (t − s)Bun (s)ds, t ∈ J. (3.5)
0
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 11

Since 0 ∈ Uad , without loss of generality, we may assume that G (xn , un ) ≤ G (x, 0), where
(x, 0) ∈ Aad . Using the definition of G (·, ·), we obtain
Z T Z T
2 2
T λγ−1 n n
ku (t)kU dt ≤ kx (T ) − xT kX + λ (T − t)γ−1 kun (t)k2U dt
0 0
2
≤ kx(T ) − xT kX ≤ 2 kx(T )k2X + kxT k2X < +∞.


(3.6)
The above estimate guarantees the existence of a large R > 0 (independent of n) such that
Z T
kun (t)k2U dt ≤ R < +∞. (3.7)
0

Using the relation (3.5), we estimate

Z t
n
γ−1 n

kx (t)kX ≤ kCγ (t)vkX + kTγ (t)wkX +
(t − s) Sγ (t − s)Bu (s)ds

0 X
Z t
M M̃
≤ MkvkX + MtkwkX + (t − s)2γ−1 kun (s)kU ds
Γ(2γ) 0
 1 Z t  21
M M̃ T 4γ−1 2

n 2
≤ MkvkX + MT kwkX + ku (s)kU ds
Γ(2γ) 4γ − 1 0
1
M M̃ T 4γ−1 R 2

≤ MkvkX + MT kwkX + < +∞,
Γ(2γ) 4γ − 1
for all t ∈ J. Since the space L2 (J; X) is reflexive, then the Banach-Alaoglu theorem yields
the existence of a subsequence {xnk }∞ n ∞
k=1 of {x }n=1 such that
w
⇀ x0 in L2 (J; X), as k → ∞.
xnk − (3.8)
The estimate (3.7) implies that the sequence {un }∞
n=1 is uniformly bounded in the space
2
L (J; U). Once again, by applying the Banach-Alaoglu theorem, we can find a subsequence,
say, {unk }∞ n ∞
k=1 of {u }n=1 such that
w
⇀ u0 in L2 (J; U) = Uad , as k → ∞.
u nk −
Since the linear operator B is bounded (continuous) from U to X, then we have
w
⇀ Bu0 in L2 (J; X), as k → ∞.
Bunk − (3.9)
Moreover, by
R · using γ−1
the above convergences together with the compactness of the operator
(Qf )(·) = 0 (· − s) Sγ (· − s)f (s)ds : L2 (J; X) → C(J; X) (see Lemma 3.4 below), we
obtain
Z t Z t
γ−1 n γ−1

(t − s) Sγ (t − s)Bu (s)ds − (t − s) Sγ (t − s)Bu(s)ds
→ 0,
k

0 0 X
as k → ∞, for all t ∈ J. We now estimate
kxnk (t) − x∗ (t)kX
Z t Z t
0
γ−1 nk γ−1

= (t − s) Sγ (t − s)Bu (s)ds −
(t − s) Sγ (t − s)Bu (s)ds

0 0 X
→ 0 as k → ∞, for all t ∈ J, (3.10)
12 S. ARORA, M. T. MOHAN AND J. DABAS

where
Z t
∗
x (t) = Cγ (t)v + Tγ (t)w + (t − s)γ−1 Sγ (t − s)Bu0 (s)ds, t ∈ J.
0

The above expression guarantees that the function x∗ ∈ C(J; X) is the unique mild solution
of the equation (3.2) with the control u0 ∈ Uad . Using the convergences (3.8) and (3.10) and
the fact that the weak limit is unique, we deduce that x∗ (t) = x0 (t), for all t ∈ J. Therefore,
the function x0 is the unique mild solution of the system (3.2) with the control u0 ∈ Uad , so
that the whole sequence xn → x0 ∈ C(J; X). Hence, we have (x0 , u0 ) ∈ Aad .
Next, we show that the functional G (·, ·) attains its minimum at (x0 , u0 ), that is, G =
G (x0 , u0). Note that the cost functional G (·, ·) given in (3.1) is continuous and convex (see
Proposition III.1.6 and III.1.10, [13]) on L2 (J; X) × L2 (J; U), it follows that G (·, ·) is weakly
lower semi-continuous (Proposition II.4.5, [13]). That is, for a sequence
w
⇀ (x0 , u0) in L2 (J; X) × L2 (J; U), as n → ∞,
(xn , un ) −
we have
G (x0 , u0) ≤ lim inf G (xn , un ).
n→∞

Hence, we obtain
L ≤ G (x0 , u0 ) ≤ lim inf G (xn , un ) = lim G (xn , un ) = L,
n→∞ n→∞

and thus (x0 , u0 ) is a minimizer of the problem (3.4). Since the cost functional given in (3.1)
is convex, the constraint (3.2) is linear and the class Uad = L2 (J; U) is convex, then the
optimal control obtained above is unique. 
Remark 3.2. Since X is a separable reflexive Banach space with strictly convex dual X∗ ,
then the fact 8.12 in [14] ensures that the norm k · kX is Gateaux differentiable. Moreover,
the Gateaux derivative ∂x of the function ϕ(x) = 21 kxk2X is the duality map, that is,
ϕ(x + ǫy) − ϕ(x) 1d 2

h∂x ϕ(x), yi = lim = kx + ǫykX  = hJ [x], yi,
ǫ→0 ǫ 2 dǫ ǫ=0

for y ∈ X. Furthermore, as we know that U is a separable Hilbert space (identified with its
own dual), then the space U admits a Fréchet differentiable norm (cf. Theorem 8.24, [14]) .
The expression for the optimal control in the feedback form is obtained in the following
lemma:
Lemma 3.3. Assume that (x, u) is the optimal pair for the problem (3.4). Then the optimal
control u is given by
1
u(t) = B∗ Sγ (T − t)∗ J R(λ, ΦT0 )ℓ(x(·)) , t ∈ [0, T ), λ > 0, < γ < 1,
 
2
with
ℓ(x(·)) = xT − Cγ (T )v − Tγ (T )w.
Proof. Let us consider the functional
L (ε) = G (xu+εz , u + εz),
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 13

where (x, u) is the optimal solution of (3.4) and z ∈ L2 (J; U). Also the function xu+εz is the
unique mild solution of (3.2) with the control u + εz. Then, it is immediate that
Z t
xu+εz (t) = Cγ (t)v + Tγ (t)w + (t − s)γ−1 Sγ (t − s)B(u + εz)(s)ds.
0

It is clear that the functional L (ε) has a critical point at ε = 0. We now calculate the first
variation of the cost functional G (defined in (3.1)) as
 Z T 
d  d 2 γ−1 2
L (ε) = kxu+εz (T ) − xT kX + λ (T − t) ku(t) + εz(t)kU dt

dε ε=0 dε 0 ε=0

d
= 2 hJ (xu+εz (T ) − xT ), (xu+εz (T ) − xT )i
dε
Z T 
γ−1 d
+ 2λ (T − t) (u(t) + εz(t), (u(t) + εz(t)))dt
0 dε ε=0
 Z T 
= 2 J (x(T ) − xT ), (T − t)γ−1 Sγ (T − t)Bz(t)dt
0
Z T
+ 2λ (T − t)γ−1 (u(t), z(t))dt.
0

By taking the first variation zero, we deduce that

 Z T  Z T
γ−1
0 = J (x(T ) − xT ), (T − t) Sγ (T − t)Bz(t)dt + λ (T − t)γ−1 (u(t), z(t))dt
0 0
Z T Z T
= (T − t)γ−1 hJ (x(T ) − xT ), Sγ (T − t)Bz(t)idt + λ (T − t)γ−1 (u(t), z(t))dt
0 0
Z T
= (T − t)γ−1 (B∗ Sγ (T − t)∗ J (x(T ) − xT ) + λu(t), z(t))dt, (3.11)
0

where (·, ·) is the inner product in the Hilbert space U. Note that the function z ∈ L2 (J; U)
is an arbitrary (one can take z to be B∗ Sγ (T − t)∗ J (x(T ) − xT ) + λu(t)), it is immediate
that
u(t) = −λ−1 B∗ Sγ (T − t)∗ J (x(T ) − xT ), (3.12)
for a.e. t ∈ [0, T ]. Along with the relations (3.11) and (3.12), it is clear that u ∈ C([0, T ]; U).
Using the above expression of control, we find
Z T
x(T ) = Cγ (T )v + Tγ (T )w − λ−1 (T − s)γ−1 Sγ (T − s)BB∗ Sγ (T − s)∗ J (x(T ) − xT )ds
0
−1
= Cγ (T )v + Tγ (T )w − λ ΦT0 J [x(T ) − xT ]. (3.13)
Let us assume
ℓ(x(·)) := xT − Cγ (T )v − Tγ (T )w. (3.14)
Further, by (3.13) and (3.14), we have
x(T ) − xT = −ℓ(x(·)) − λ−1 ΦT0 J [x(T ) − xT ]. (3.15)
14 S. ARORA, M. T. MOHAN AND J. DABAS

From (3.15), one can easily get

x(T ) − xT = −λI(λI + ΦT0 J )−1 ℓ(x(·)) = −λR(λ, ΦT0 )ℓ(x(·)). (3.16)
Finally, from (3.12), we find the optimal control as
u(t) = B∗ Sγ (T − t)∗ J R(λ, ΦT0 )ℓ(x(·)) , for all t ∈ [0, T ],
 

which completes the proof. 

R·
In the next lemma, we discuss the compactness of the operator (Qf )(·) = 0 (·−s)γ−1 Sγ (·−
s)f (s)ds : L2 (J; X) → C(J; X), where γ ∈ ( 21 , 1) and X is a general Banach space.
Lemma 3.4. Suppose that the operator Sγ (t) is compact for t ≥ 0. Let the operator Q :
L2 (J; X) → C(J; X) be defined as
Z t
1
(Qf )(t) = (t − s)γ−1 Sγ (t − s)f (s)ds, t ∈ J, < γ < 1. (3.17)
0 2
Then the operator Q is compact.
A proof of the above lemma is a straightforward adaptation of the proof of Lemma 3.2,
[2].
Lemma 3.5. The linear control system (3.2) is approximately controllable on J if and only
if Assumption (H0) holds.
A proof of the above lemma can be obtained by proceeding similarly as in the proof of
Theorem 3.2, [3].
Remark 3.6. Note that the operator ΦT0 is positive if and only if Assumption (H0) holds
(see, Theorem 2.3, [33]). The positivity of ΦT0 is equivalent to
hx∗ , ΦT0 x∗ i = 0 ⇒ x∗ = 0.
Using the definition of ΦT0 , we have
Z T
∗ T ∗
hx , Φ0 x i = (T − t)γ−1 kB∗ Sγ (T − t)∗ x∗ k2X∗ dt. (3.18)
0
The above fact and Lemma 3.5 guarantee that the approximate controllability of the linear
system (3.2) is equivalent to the condition
B∗ Sγ (T − t)∗ x∗ = 0, 0 ≤ t < T ⇒ x∗ = 0.

4. Approximate Controllability of the Fractional order Impulsive System

The present section is reserved for the approximate controllability of the semilinear impul-
sive fractional control system (1.1). For this, we first verify the existence of a mild solution
of the system (1.1) with the control defined as
p
X 1
uγλ (t) = uγj,λ(t)χ[sj ,τj+1 ] (t), t ∈ J, < γ < 1, (4.1)
j=0
2
where
h i
uγj,λ(t) ∗ ∗τj+1
= B Sγ (τj+1 − t) J R(λ, Φsj )gj (x(·)) ,
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 15

for t ∈ [sj , τj+1], j = 0, 1, . . . , p, with

Z τ1
g0 (x(·)) = ξ0 − Cγ (τ1 )ψ(0) − Tγ (τ1 )η − (τ1 − s)γ−1 Sγ (τ1 − s)f (s, x̺̃(s,x̃s ) )ds,
0
−
gj (x(·)) = ξj − Cγ (τj+1 − Tγ (τj+1 − sj )h′j (sj , x̃(τj− ))
sj )hj (sj , x̃(τj )) −
j−1
Z sj " #
X
+ (sj − s)γ−1 Sγ (sj − s) f (s, x̺̃(s,x̃s) ) + B uγk,λ(s)χ[sk ,τk+1 ) (s) ds
0 k=0
Z τj+1
− (τj+1 − s)γ−1 Sγ (τj+1 − s)f (s, x̺̃(s,x̃s ) )ds
0
Z sj j−1
X
− (τj+1 − s) γ−1
Sγ (τj+1 − s)B uγk,λ(s)χ[sk ,τk+1 ) (s)ds, j = 1, . . . , p,
0 k=0

the function x̃ : (−∞, T ] → X is given by

x̃(t) = ψ(t), t ∈ (−∞, 0], x̃(t) = x(t), t ∈ J,
and for arbitrary ξj ∈ X for j = 0, 1, . . . , p.
τ
Remark 4.1. Since the operator Φsj+1j , for each j = 0, . . . , p, is linear, bounded and non-
τ
negative, Lemma 2.16 is valid for each Φsj+1
j , for j = 0, . . . , p.
In the following theorem, we obtain the existence of a mild solution of the system (1.1)
with the control given in (4.1).
Theorem 4.2. If Assumptions (R1)-(R3) and (H1)-(H2) hold true. Then for fixed ξj ∈ X,
for j = 0, 1, . . . , p and every λ > 0, the system (1.1) with the control (4.1) has at least one
mild solution on J, provided
p−1
( )
2MT 2γ−δ K2 ζ (p + 1)(p + 2)R p(p + 1)R2 X (p+k)(p−k−1)R
1+ + e 2 < 1, (4.2)
Γ(2γ)c1−δ 2 2 k=0
3γ
 2
where R = 2T 3γλ
M M̃
Γ(2γ)
and c = 2γ−δ
1−δ
.

Proof. Let us take a set Z := {x ∈ PC(J; X) : x(0) = ψ(0)} with the norm k·kPC(J;X) . Next,
we consider a set Er = {x ∈ Z : kxkPC(J;X) ≤ r}, for each r > 0.
For λ > 0, we define an operator Fλ : Z → Z such that
 Z t
(t − s)γ−1 Sγ (t − s) Buγλ (s) + f (s, x̺̃(s,x̃s ) ) ds,
 



 Cγ (t)ψ(0) + Tγ (t)η +

 0




 t ∈ [0, τ1 ],
hj (t, x̃(τj− )),




 t ∈ (τj , sj ], j = 1, . . . , p,

Cγ (t − sj )hj (sj , x̃(τj )) + Tγ (t − sj )hj (sj , x̃(τj− ))
− ′

(Fλ x)(t) = Z sj (4.3)

(sj − s)γ−1 Sγ (sj − s) Buγλ (s) + f (s, x̺̃(s,x̃s ) ) ds
 
−






 0

 Z t
 γ−1
 γ 


 + (t − s) Sγ (t − s) Buλ (s) + f (s, x̺̃(s,x̃s ) ) ds,
0



t ∈ (sj , τj+1 ], j = 1, . . . , p.

16 S. ARORA, M. T. MOHAN AND J. DABAS

where uγλ (·) is given in (4.1). It is clear form the definition of Fλ that the system (1.1) has
a mild solution, if the operator Fλ has a fixed point. The proof that the operator Fλ has a
fixed point is divided into the following steps.
Step (1): Fλ (Er ) ⊂ Er , for some r. We achieve this goal by contradiction. To prove
this, we assume that for any λ > 0 and every r > 0, there exists xr (·) ∈ Er such that
k(Fλ xr )(t)kX > r, for some t ∈ J (t may depend upon r). First, we compute

kg0 (x(·))kX ≤ kξ0 kX + kCγ (τ1 )ψ(0)kX + kTγ (τ1 )ηkX

Z τ1

+ (τ1 − s)γ−1 Sγ (τ1 − s)f (s, x̺̃(s,x̃s) ) X ds
0
Z τ1
M
≤ kξ0 kX + Mkψ(0)kX + MtkηkX + (τ1 − s)2γ−1 φr′ (s)ds
Γ(2γ) 0
≤ kξ0 kX + Mkψ(0)kX + MT kηkX
Z τ 1−δ Z τ1 δ
M 2γ−1 1
+ (τ1 − s) 1−δ ds (φr′ (s)) δ ds
Γ(2γ) 0 0

Mτ12γ−δ
≤ kξ0 kX + Mkψ(0)kX + MT kηkX + kφr′ k δ1
Γ(2γ)c1−δ L ([0,τ1 ];R+ )

2MT 2γ−δ
≤ kξ0 kX + Mkψ(0)kX + MT kηkX + kφr′ k δ1 =: N0 , (4.4)
Γ(2γ)c1−δ L (J;R+ )

2γ−δ
where c = 1−δ
and r ′ = K1 kψkB + K2 r. Next, we estimate

kgj (x(·))kX ≤ kξj kX + Cγ (t − sj )hj (sj , x̃(τj− )) X + Tγ (t − sj )h′j (sj , x̃(t−

j )) X

Z sj

+ (sj − s)γ−1 Sγ (sj − s)f (s, x̺̃(s,x̃s) ) X ds
Z0 τj+1

+ (τj+1 − s)γ−1 Sγ (τj+1 − s)f (s, x̺̃(s,x̃s) ) X ds
0
Z sj j−1

X
γ−1 γ
+ (sj − s) Sγ (sj − s)B uk,λ(s)χ[sk ,τk+1 ] (s) ds

0
k=0

X
Z sj j−1

X
+ (τj+1 − s)γ−1 Sγ (τj+1 − s)B uγk,λ(s)χ[sk ,τk+1 ] (s) ds

0
k=0

X
Z sj
M
≤ kξj kX + Mκj + M(t − sj )ϑj + (sj − s)2γ−1 φr′ (s)ds
Γ(2γ) 0
Z τj+1 !2 j−1
M 1 M M̃ X
+ (τj+1 − s)2γ−1 φr′ (s)ds + kgk (x(·))kX
Γ(2γ) 0 λ Γ(2γ) k=0
Z τk+1
(sj − s)2γ−1 + (τj+1 − s)2γ−1 (τk+1 − s)γ ds
 
×
sk
Z τj+1
2M
≤ kξj kX + Mκj + MT ϑj + (τj+1 − s)2γ−1 φr′ (s)ds
Γ(2γ) 0
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 17
!2 j−1 τk+1
2 M M̃ X Z
+ kgk (x(·))kX (τj+1 − s)2γ−1 (τk+1 − s)γ ds
λ Γ(2γ) k=0 sk
Z τj+1
2M
≤ kξj kX + Mκj + MT ϑj + (τj+1 − s)2γ−1 φr′ (s)ds
Γ(2γ) 0
!2 j−1 Z τk+1
2 M M̃ X
+ kgk (x(·))kX (τj+1 − s)3γ−1 ds
λ Γ(2γ) sk
k=0
2γ−δ
2Mτj+1
≤ kξj kX + Mκj + MT ϑj + kφr′ kL([0,τj+1 ];R+)
Γ(2γ)c1−δ
!2 j−1
2 M M̃ X (τj+1 − sk )3γ − (τj+1 − τk+1 )3γ
+ kgk (x(·))kX
λ Γ(2γ) 3γ
k=0
2MT 2γ−δ
≤ kξj kX + Mκj + MT ϑj + kφr′ kL(J;R+ )
Γ(2γ)c1−δ
!2 j−1
2T 3γ M M̃ X
+ kgk (x(·))kX
3γλ Γ(2γ)
k=0
j−1
X
≤ Nj + R kgk (x(·))kX ,
k=0

 2
2T 3γ M M̃ 2M T 2γ−δ
where R = 3γλ Γ(2γ)
and Nj = kξj kX + Mκj + MT ϑj + Γ(2γ)c 1−δ kφr ′ kL(J;R+ ) , for j =

1, . . . , p. Using the discrete Gronwall-Bellman lemma (Lemma 2.18), we obtain

j−1
X (j+k)(j−k−1)R
kgj (x(·))kX ≤Nj + R Nk e 2 =: Cj , for j = 1, . . . , p. (4.5)
k=0

Taking t ∈ [0, τ1 ] and using the estimates (2.8), (4.4),(4.5), Lemma 2.12 and Assumption
2.17 (H1)-(H2), we evaluate

r < k(Fλ xr )(t)kX

Z t
γ−1
 γ 
= Cγ (t)ψ(0) + Tγ (t)η + (t − s) Sγ (t − s) Buλ (s) + f (s, x̺̃(s,x̃s) ) ds

0 X
Z t
M M̃
≤ Mkψ(0)kX + MtkηkX + (t − s)2γ−1 kuγλ (s)kU ds
Γ(2γ) 0
Z t
M
(t − s)2γ−1 f (s, x̃ρ(s,x̃s ) ) X ds

+
Γ(2γ) 0
Z τ1
M M̃
≤ Mkψ(0)kX + Mτ1 kηkX + (τ1 − s)2γ−1 kuγλ (s)kU ds
Γ(2γ) 0
Z τ1
M
(τ1 − s)2γ−1 f (s, x̃ρ(s,x̃s ) ) X ds

+
Γ(2γ) 0
18 S. ARORA, M. T. MOHAN AND J. DABAS
Z τ1
M M̃
(τ1 − s)2γ−1 uγ1,λ (s) U ds

≤ Mkψ(0)kX + Mτ1 kηkX +
Γ(2γ) 0
Z τ1
M
+ (τ1 − s)2γ−1 φr′ (s)ds
Γ(2γ) 0
!2 Z τ1
1 M M̃
≤ Mkψ(0)kX + Mτ1 kηkX + kg0 ((·))kX (τ1 − s)3γ−1 ds
λ Γ(2γ) 0
Z τ1
M
+ (τ1 − s)2γ−1 φr′ (s)ds
Γ(2γ) 0
!2 Z
τ1
N0 M M̃
≤ Mkψ(0)kX + Mτ1 kηkX + (τ1 − s)3γ−1 ds
λ Γ(2γ) 0
Z τ1 1−δ Z τ1 δ
M 2γ−1 1
+ (τ1 − s) 1−δ φr′ (s)ds
δ

Γ(2γ) 0 0
!2
N0 τ13γ M M̃ Mτ12γ−δ
= Mkψ(0)kX + Mτ1 kηkX + + kφr′ k δ1
3γλ Γ(2γ) Γ(2γ)c1−δ L ([0,τ1 ];R+ )
!2
N0 T 3γ M M̃ MT 2γ−δ
≤ Mkψ(0)kX + MT kηkX + + kφr′ k δ1
3γλ Γ(2γ) Γ(2γ)c1−δ L (J;R+ )
!2 p
2T 3γ M M̃ X 2MT 2γ−δ
< Mkψ(0)kX + MT kηkX + Ck + 1−δ
kφr′ k δ1
3γλ Γ(2γ) k=0
Γ(2γ)c L (J;R+ )

p
X 2MT 2γ−δ
= Mkψ(0)kX + MT kηkX + R Ck + kφr′ k 1δ , (4.6)
k=0
Γ(2γ)c1−δ L (J;R+ )

where C0 = N0 . For t ∈ (τj , sj ], j = 1, . . . , p, we obtain

r < k(Fλ xr )(t)k ≤ hj (t, x̃(τ − ))

X j X
p 2γ−δ
X 2MT
≤ κj < κj + R Ck + kφr′ k 1δ (J; R+ ). (4.7)
k=0
Γ(2γ)c1−δ L

Taking t ∈ (sj , τj+1], j = 1, . . . , p, we compute

r < k(Fλ xr )(t)kX

− ′ −

= Cγ (t − sj )hj (sj , x̃(τj )) + Tγ (t − sj )hj (sj , x̃(τj ))
Z sj
(sj − s)γ−1 Sγ (sj − s) Buγλ (s) + f (s, x̺̃(s,x̃s ) ) ds
 
−
0
Z t
γ−1
 γ 
+ (t − s) Sγ (t − s) Buλ (s) + f (s, x̺̃(s,x̃s) ) ds
0 X
≤ Cγ (t − sj )hj (sj , x̃(τj− )) X + Tγ (t − sj )h′j (sj , x̃(t−

j ))
X
Z sj
γ
+ (sj − s)γ−1 kSγ (sj − s)Buλ (s)kX ds
0
APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 19
Z sj
+ (sj − s)γ−1 Sγ (sj − s)f (s, x̺̃(s,x̃s) ) X ds
0
Z t
+ (t − s)γ−1 kSγ (t − s)Buγλ (s)kX ds
0
Z t
+ (t − s)γ−1 Sγ (t − s)f (s, x̺̃(s,x̃s) ) X ds
0
Z sj
M M̃
≤ Mκj + M(t − sj )ϑj + (sj − s)2γ−1 kuγλ (s)kU ds
Γ(2γ) 0
Z sj Z t
M M M̃
2γ−1
(t − s)2γ−1 kuγλ (s)kU ds

+ (sj − s) f (s, x̺̃(s,x̃s ) ) X +
Γ(2γ) 0 Γ(2γ) 0
Z t
M
(t − s)2γ−1 f (s, x̺̃(s,x̃s ) ) X ds

+
Γ(2γ) 0
Z sj j−1
M M̃ X
≤ Mκj + MT ϑj + (sj − s)2γ−1 uγk,λ(s)χ[sk ,τk+1 ] (s) ds

Γ(2γ) 0
k=0 U
Z sj
M
+ (sj − s)2γ−1 φr′ (s)ds
Γ(2γ) 0
Z τj+1 j
M M̃ X
+ (τj+1 − s)2γ−1 uγk,λ(s)χ[sk ,τk+1 ] (s) ds

Γ(2γ) 0
k=0

U
Z tj+1
M
+ (tj+1 − s)2γ−1 φr′ (s)ds
Γ(2γ) 0
!2 j−1 Z τk+1
1 M M̃ X
≤ Mκj + MT ϑj + kgk (x(·))kX (sj − s)2γ−1 (τk+1 − s)γ ds
λ Γ(2γ) k=0 sk

Ms2γ−δ
j Mt2γ−δ
j+1
+ kφ r ′ k 1 + kφr′ k δ1
Γ(2γ)c1−δ L δ ([0,sj ];R+ ) Γ(2γ)c1−δ L ([0,τj+1 ];R+ )
!2 j Z τk+1
1 M M̃ X
+ kgk (x(·))kX (sj − s)2γ−1 (τk+1 − s)γ ds
λ Γ(2γ) k=0 sk
!2 j−1
T 3γ M M̃ X 2MT 2γ−δ
≤ Mκj + MT ϑj + kgk (x(·))kX + 1−δ
kφr′ k δ1
3γλ Γ(2γ) k=0
Γ(2γ)c L (J;R+ )

!2 j
T 3γ M M̃ X
+ kgk (x(·))kX
3γλ Γ(2γ) k=0
!2 j
2T 3γ M M̃ X 2MT 2γ−δ
≤ Mκj + MT ϑj + kgk (x(·))kX + kφr′ k 1δ
3γλ Γ(2γ) k=0
Γ(2γ)c1−δ L (J;R+ )

!2 p
3γ
2T M M̃ X 2MT 2γ−δ
≤ Mκj + MT ϑj + kgk (x(·))kX + 1−δ
kφr′ k 1δ
3γλ Γ(2γ) k=0
Γ(2γ)c L (J;R+ )
20 S. ARORA, M. T. MOHAN AND J. DABAS
!2 p
2T 3γ M M̃ X 2MT 2γ−δ
≤ Mκj + MT ϑj + Ck + kφr′ k δ1
3γλ Γ(2γ) k=0
Γ(2γ)c1−δ L (J;R+ )

p
X 2MT 2γ−δ
= Mκj + MT ϑj + R Ck + kφr′ k δ1 . (4.8)
k=0
Γ(2γ)c1−δ L (J;R+ )

Using Assumption 2.17 (H2 )(ii), we easily obtain

 
kφr′ k 1δ kγ r ′k 1
r ′
L (J;R+ ) L α1 (J;R+ )
lim inf = lim inf  ′
×  = K2 ζ.
r→∞ r r→∞ r r

Thus, dividing by r in the expressions (4.6), (4.7), (4.8) and then passing r → ∞, we obtain
p−1
( )
2MT 2γ−δ K2 ζ (p + 1)(p + 2)R p(p + 1)R2 X (p+k)(p−k−1)R
1+ + e 2 > 1,
Γ(2γ)c1−δ 2 2 k=0

which is a contradiction to (4.2). Therefore, for some r > 0, Fλ (Er ) ⊂ Er .

Step (2): Next, we verify that the operator Fλ is compact. To prove this claim, we use
the infinite-dimensional version of the Ascoli-Arzela theorem (see, Theorem 3.7, Chapter 2,
[29]). According to this theorem, it is enough to verify the following:
(i) the image of Er under Fλ is uniformly bounded (which is proved in Step (1)),
(ii) the image of Er under Fλ is equicontinuous,
(iii) for any t ∈ J, the set V(t) = {(Fλ x)(t) : x ∈ Er } is relatively compact.
First, we show that the map Fλ is equicontinuous, that is, the image of Er under Fλ is
equicontinuous. For t1 , t2 ∈ [0, τ1 ] such that t1 < t2 and x ∈ Er , we evaluate the following:

k(Fλ x)(t2 ) − (Fλ x)(t1 )kX

≤ k[Cγ (t2 ) − Cγ (t1 )]ψ(0)kX + k[Tγ (t2 ) − Tγ (t1 )]ηkX
Z t2
γ−1

+ (t 2 − s) S γ (t2 − s)f (s, x̺̃(s,x̃ s ) )ds

t1 X
Z t2
(t2 − s)γ−1 Sγ (t2 − s)Buγλ (s)ds

+
t1 X
Z t1
γ
 γ−1 γ−1

+ (t2 − s) − (t1 − s) Sγ (t2 − s)Buλ (s)ds
0
Z t1 X
γ
γ−1

+ (t 1 − s) [S γ (t2 − s) − S γ (t1 − s)]Bu λ (s)ds

0 X
Z t1
 γ−1 γ−1

+ (t2 − s) − (t1 − s) Sγ (t2 − s)f (s, x̺̃(s,x̃s ) )ds
0
Z t1 X
γ−1

+ (t 1 − s) [S γ (t2 − s) − S γ (t1 − s)]f (s, x̺̃(s,x̃ s ) )ds

0 X
≤ kCγ (t2 )ψ(0) − Cγ (t1 )ψ(0)kX + kTγ (t2 ) − Tγ (t1 )kL(X) kηkX
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 21
!2 Z
t2 t2
M N M M̃
Z
0
+ (t2 − s)2γ−1 φr′ (s)ds + (t2 − s)2γ−1 (τ1 − s)γ ds
Γ(2γ) t1 λ Γ(2γ) t1
!2 Z
t1 
N0 M M̃ 
(t2 − s)γ−1 − (t1 − s)γ−1 (t2 − s)γ−1 (τ1 − s)γ ds
+
λ Γ(2γ) 0
2 Z t1
M M̃ N0
+ (t1 − s)γ−1 kSγ (t2 − s) − Sγ (t1 − s)kL(X) (τ1 − s)γ ds
λΓ(2γ) 0
Z t1
M  
(t2 − s)γ−1 − (t1 − s)γ−1 (t2 − s)γ φr′ (s)ds
+
Γ(2γ) 0
Z t1
+ (t1 − s)γ−1 kSγ (t2 − s) − Sγ (t1 − s)kL(X) φr′ (s)ds
0
≤ kCγ (t2 )ψ(0) − Cγ (t1 )ψ(0)kX + kTγ (t2 ) − Tγ (t1 )kL(X) kηkX
Z t2 δ
M (t2 − t1 )2γ−δ 1
+ (φr′ (s)) δ ds
Γ(2γ) c1−δ t1
!2 Z
t2
N0 M M̃
+ (t2 − s)2γ−1 (τ1 − s)γ ds
λ Γ(2γ) t1
!2 Z
t1 
N0 M M̃ 
(t2 − s)γ−1 − (t1 − s)γ−1 (t2 − s)γ−1 (τ1 − s)γ ds
+
λ Γ(2γ) 0
Z t1
M M̃ 2 N0
+ sup kSγ (t2 − s) − Sγ (t1 − s)kL(X) (t1 − s)γ−1 (τ1 − s)γ ds
λΓ(2γ) s∈[0,t1 ] 0
Z t1
M  
(t2 − s)γ−1 − (t1 − s)γ−1 (t2 − s)γ φr′ (s)ds
+
Γ(2γ) 0
Z t1
+ sup kSγ (t2 − s) − Sγ (t1 − s)kL(X) (t1 − s)γ−1 φr′ (s)ds. (4.9)
s∈[0,t1 ] 0

Similarly for t1 , t2 ∈ (sj , τj+1 ], j = 1, . . . , p with t1 < t2 and x ∈ Er , one can compute

k(Fλ x)(t2 ) − (Fλ x)(t1 )kX

≤ Cγ (t2 − sj )h(sj , x(τ − )) − Cγ (t1 − sj )h(sj , x(τ − ))

j j X
Z t2 δ
M (t2 − t1 )2γ−δ 1
+ kTγ (t2 − sj ) − Tγ (t1 − sj )kL(X) ϑj + (φr′ (s)) δ ds
Γ(2γ) c1−δ t1
!2 Z
t2
Cj M M̃
+ (t2 − s)2γ−1 (τj+1 − s)γ ds
λ Γ(2γ) t1
Z t1
M M̃ (t2 − s)γ−1 − (t1 − s)γ−1 (t2 − s)γ kuγ (s)k ds
 
+ λ U
Γ(2γ) 0
Z t1
+ M̃ sup kSγ (t2 − s) − Sγ (t1 − s)kL(X) (t1 − s)γ−1 kuγλ (s)kU ds
s∈[0,t1 ] 0
22 S. ARORA, M. T. MOHAN AND J. DABAS

t1 
M
Z

+ (t2 − s)γ−1 − (t1 − s)γ−1 (t2 − s)γ φr′ (s)ds
Γ(2γ) 0
Z t1
+ sup kSγ (t2 − s) − Sγ (t1 − s)kL(X) (t1 − s)γ−1 φr′ (s)ds
s∈[0,t1 ] 0

≤ Cγ (t2 − sj )h(sj , x(τ − )) − Cγ (t1 − sj )h(sj , x(τ − ))

j j X
Z t2 δ
M (t2 − t1 )2γ−δ 1
+ kTγ (t2 − sj ) − Tγ (t1 − sj )kL(X) ϑj + (φr′ (s)) δ ds
Γ(2γ) c1−δ t1
!2 Z
t2
Cj M M̃
+ (t2 − s)2γ−1 (τj+1 − s)γ ds
λ Γ(2γ) t1

j−1
!2 Z
τk+1 
X Ck M M̃ 
(t2 − s)γ−1 − (t1 − s)γ−1 (t2 − s)γ (τk+1 − s)γ ds
+
k=0
λ Γ(2γ) sk
!2 Z
s1 
Cj M M̃ 
(t2 − s)γ−1 − t1 − s)γ−1 (t2 − s)γ (τj+1 − s)γ ds
+
λ Γ(2γ) sj
j−1
Ck M M̃ 2 τk+1
X Z
+ sup kSγ (t2 − s) − Sγ (t1 − s)kL(X) (t1 − s)γ−1 (τk+1 − s)γ ds
s∈[0,s1 ] k=0
λ Γ(2γ) sk

Cj M M̃ 2 t1
Z
+ sup kSγ (t2 − s) − Sγ (t1 − s)kL(X) (t1 − s)γ−1 (tj+1 − s)γ ds
s∈[0,s1 ] λ Γ(2γ) sj
Z t1
M  
(t2 − s)γ−1 − (t1 − s)γ−1 (t2 − s)γ φr′ (s)ds
+
Γ(2γ) 0
Z t1
+ sup kSγ (t2 − s) − Sγ (t1 − s)kL(X) (t1 − s)γ−1 φr′ (s)ds. (4.10)
s∈[0,t1 ] 0

Moreover, for t1 , t2 ∈ (τj , sj ], j = 1, . . . , p with t1 < t2 and x ∈ Er , we have

k(Fλ x)(t2 ) − (Fλ x)(t1 )kX ≤ hj (t2 , x̃(τj− )) − hj (t1 , x̃(τj− )) X .

(4.11)

From the facts that the operator Cγ (t)x, x ∈ X is uniformly continuous for t ∈ J, the
operators Tγ (t) and Sγ (t) are uniformly continuous for all t ∈ J and the impulses hj (·, x)
are continuous for j = 1, . . . , p and each x ∈ X, we conclude that the right hand side of the
expressions (4.9), (4.10) and (4.11) converge to zero as |t2 − t1 | → 0. Hence, the image of Er
under Fλ is equicontinuous.
The set V(t) = {(Fλ x)(t) : x ∈ Er }, for each t ∈ J, is relatively compact in Er follows
from the facts that the operator Sγ (t) is compact for t ≥ 0, the impulses hj (t, ·), h′j (t, ·),
for
R · t ∈ [τ j , sj ], j = 1, . . . , p, are completely continuous and also the operator (Qf )(·) =
γ−1
0
(· − s) Sγ (· − s)f (s)ds is compact (see Lemma 3.4). Hence, for each t ∈ J, the set
V(t) = {(Fλ x)(t) : x ∈ Er }, is relatively compact.
Thus, by applying the Arzela-Ascoli theorem, we obtain that the operator Fλ is compact.
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 23

Step (3): In the final step, we prove that the operator Fλ is continuous. In order to prove
this, we consider a sequence {xn }∞ n
n=1 ⊆ Er such that x → x in Er , that is,

lim kxn − xkPC(J;X) = 0.

n→∞
From Lemma 2.10, we have

x˜n − x˜s ≤ K2 sup x˜n (θ) − x̃(θ) = K2 kxn − xk
s B X PC(J;X) → 0 as n → ∞,
θ∈J

for all s ∈ R(̺− ) ∪ J. Since ̺(s, x˜ks ) ∈ R(̺− ) ∪ J, for all k ∈ N, then we conclude that

˜n
x ̺(s,x˜k ) − x̺̃(s,x˜k ) → 0 as n → ∞, for all s ∈ J and k ∈ N.

s s B
In particular, we choose k = n and use the above convergence together with Assumption
2.17 (H1) to obtain

f (s, x˜n ̺(s,x˜n ) ) − f (s, x̺̃(s,x˜s) )
s X

˜n
≤ f (s, x ̺(s,x˜ns ) ) − f (s, x̺̃(s,x˜ns) ) X + f (s, x̺̃(s,x˜ns ) ) − f (s, x̺̃(s,x˜s ) ) X

→ 0 as n → ∞, uniformly for s ∈ J. (4.12)
Using the convergence (4.12) and the dominated convergence theorem (DCT), we conclude
kg0 (xn (·)) − g0 (x(·))kX
Z τ1

(τ1 − s) Sγ (τ1 − s) f (s, x˜n ̺(s,x˜ns ) ) − f (s, x̺̃(s,x˜s ) ) ds
γ−1

≤
0 X
Z τ1
M
(τ1 − s)2γ−1 f (s, x˜n ̺(s,x˜ns ) ) − f (s, x̺̃(s,x˜s ) ) X ds

≤
Γ(2γ) 0
→ 0 as n → ∞. (4.13)
Invoking the above convergences and the relation (2.8), we estimate
kR(λ, Φτ01 )g0 (xn (·)) − R(λ, Φτ01 )g0 (x(·))kX
1
= kλR(λ, Φτ01 )(g0 (xn (·)) − g0 (x(·)))kX
λ
1
≤ kg0 (xn (·)) − g0 (x(·))kX
λ
→ 0 as n → ∞.
By using the demicontinuous property of the duality mapping J : X → X∗ , one can easily
obtain
w
⇀ J [R(λ, Φτ01 )g0 (x(·))] as n → ∞ in X∗ .
J [R(λ, Φτ01 )g0 (xn (·))] − (4.14)
From Lemma 2.12, we infer that the operator Sγ (t) is compact for t > 0. Consequently, the
operator Sγ (t)∗ is also compact for t > 0. Thus, by using this compactness and the weak
convergence (4.14), one can easily conclude that
n,γ
u (t) − uγ (t)|U

0,λ 0,λ
≤ kB Sγ (τ1 − t)∗ (J [R(λ, Φτ01 )g0 (xn (·))] − J [R(λ, Φτ01 )g0 (x(·))])kU
∗

≤ M̃ kSγ (τ1 − t)∗ (J [R(λ, Φτ01 )g0 (xn (·))] − J [R(λ, Φτ01 )g0 (x(·))])kX
→ 0 as n → ∞, uniformly for t ∈ [0, τ1 ]. (4.15)
24 S. ARORA, M. T. MOHAN AND J. DABAS

Similarly, for j = 1, we evaluate

kg1 (xn (·)) − g1 (x(·))kX
≤ Cγ (τ2 − s1 ) h1 (s1 , x˜n (τ1− )) − h1 (s1 , x̃(τ1− )) X
 

+ Tγ (τ2 − s1 ) h′1 (s1 , x˜n (τ1− )) − h′1 (s1 , x̃(τ1− )) X

 
Z s1
γ−1

(s1 − s) Sγ (s1 − s) f (s, x˜n ̺(s,x˜ns ) ) − f (s, x̺̃(s,x̃s) ) ds

+
0
Z τ2 X

γ−1 ˜

+ (τ − s) S (τ − s) f (s, x n ) − f (s, x̃ ) ds
2 γ 2 ˜
n
̺(s,xs ) ̺(s,x̃ s )
Z0 s1 X

(s1 − s)γ−1 Sγ (s1 − s)B un,α

 α

+ 0,λ (s) − u0,λ (s) ds

0
Z s1 X
n,α
γ−1
 α

+ (τ 2 − s) S γ (τ 2 − s)B u 0,λ (s) − u 0,λ (s) ds

0 X
≤ M h1 (s1 , x˜n (τ1− )) − h1 (s1 , x̃(τ1− )) X

+ M(τ2 − s) h′1 (s1 , x˜n (τ1− )) − h′1 (s1 , x̃(τ1− )) X

Z s1
M
(s1 − s)2γ−1 f (s, x˜n ̺(s,x˜ns ) ) − f (s, x̺̃(s,x̃s ) ) X ds

+
Γ(2γ) 0
Z τ2
M
(τ2 − s)2γ−1 f (s, x˜n ̺(s,x˜ns ) ) − f (s, x̺̃(s,x̃s ) ) X ds

+
Γ(2γ)) 0
Z s1
M M̃
(s1 − s)2γ−1 un,γ γ

+ 0,λ (s) − u0,λ (s) U ds

Γ(2γ) 0
Z s1
M M̃
(τ2 − s)α−1 un,α
α

+ 0,λ (s) − u0,λ (s) U ds

Γ(2γ) 0
→ 0 as n → ∞, (4.16)
where we have used the convergences (4.12), (4.15), Assumption 2.17 (H2 ) and DCT. More-
over, similar to the convergence (4.15), one can prove
n,γ
u (t) − uγ (t)|U → 0 as n → ∞, uniformly for t ∈ [s1 , τ2 ].

1,λ 1,λ

Further, employing a similar analogy as above for j = 2, . . . , p, one obtains

n,γ
u (t) − uγ (t)|U → 0 as n → ∞, uniformly for t ∈ [sj , τj+1 ], j = 2, . . . , p.

j,λ j,λ

Hence, we have
kun,α α
λ (t) − uλ (t)||U → 0 as n → ∞, uniformly for t ∈ [sj , τj+1 ], j = 0, 1, . . . , p. (4.17)
Using the convergences (4.12), (4.17) and DCT, we arrive at
k(Fλ xn )(t) − (Fλ x)(t)kX
Z t
≤ (t − s)γ−1 kSγ (t − s)B[un,α γ
λ (s) − uλ (s)]kX ds
0
Z t

(t − s)γ−1 Sγ (t − s) f (s, x˜n ̺(s,x˜ns ) ) − f (s, x̺̃(s,x̃s) ) X ds
 
+
0
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 25
Z t
M M̃
≤ (t − s)2γ−1 kun,γ γ
λ (s) − uλ (s)kU ds
Γ(2γ) 0
Z t
M
(t − s)2γ−1 f (s, x˜n ̺(s,x˜ns ) ) − f (s, x̺̃(s,x̃s ) ) X ds

+
Γ(2γ)) 0
→ 0 as n → ∞, uniformly for t ∈ [0, t1 ].

Similarly, for t ∈ (sj , τj+1 ], j = 1, . . . , p, we deduce that

k(Fλ xn )(t) − (Fλ x)(t)kX

≤ Cγ (t − sj ) hj (sj , x˜n (τj− )) − hj (sj , x̃(τj− )) X
 

+ Tγ (t − sj ) h′j (sj , x˜n (τj− )) − h′j (sj , x̃(τj− )) X

 
Z sj

(sj − s)γ−1 Sγ (sj − s) f (s, x˜n ̺(s,x˜ns ) ) − f (s, x̺̃(s,x̃s) ) X ds
 
+
Z0 sj
+ (sj − s)γ−1 kSγ (sj − s)B[un,γ γ
λ (s) − uλ (s)]kX ds
0
Z t
+ (t − s)γ−1 kSγ (t − s)B[un,γ γ
λ (s) − uλ (s)]kX ds
0
Z t

(t − s)γ−1 Sγ (t − s) f (s, x˜n ̺(s,x˜ns ) ) − f (s, x̺̃(s,x̃s) ) X ds
 
+
0
≤ M hj (sj , x˜n (τj− )) − hj (sj , x̃(τj− )) X

+ M(t − sj ) h′j (sj , x˜n (τj− )) − h′j (sj , x̃(τj− )) X

Z sj
M
(sj − s)2γ−1 f (s, x˜n ̺(s,x˜ns ) ) − f (s, x̺̃(s,x̃s ) ) X ds

+
Γ(2γ) 0
Z sj
M M̃
+ (sj − s)2γ−1 kun,γ γ
λ (s) − uλ (s)kU ds
Γ(2γ) 0
Z t
M M̃
+ (t − s)2γ−1 kun,γ γ
λ (s) − uλ (s)kU ds
Γ(2γ) 0
Z t
M
(t − s)2γ−1 f (s, x˜n ̺(s,x˜ns ) ) − f (s, x̺̃(s,x̃s ) ) X ds

+
Γ(2γ) 0
→ 0 as n → ∞, uniformly for t ∈ (sj , τj+1 ].

Moreover, for t ∈ (τj , sj ], j = 1, . . . , p, using Assumption 2.17 (H2 ), we obtain

k(Fλ xn )(t) − (Fλ x)(t)kX ≤ hj (t, x˜n (τj− )) − hj (t, x̃(τj− )) X → 0 as n → ∞.

Therefore, it follows that Fλ is continuous.

Hence, by the application of Schauder’s fixed point theorem, we conclude that the operator
Fλ has a fixed point in Er , or the system (1.1) has a mild solution. 

In order to verify the approximate controllability of the system (1.1), we replace the
assumption (H2 ) of the function f (·, ·) by the following assumption:
26 S. ARORA, M. T. MOHAN AND J. DABAS

(H3) The function f : J × B → X satisfies Assumption (H1 )(i) and there exists a function
1
φ ∈ L δ (J; R+ ) with δ ∈ [0, γ] such that
kf (t, ψ)kX ≤ φ(t), for all (t, ψ) ∈ J × B.
Theorem 4.3. Let Assumptions (R1)-(R3), (H0)-(H1), (H3) and the condition (4.2) of
Theorem 4.2 be fulfilled. Then the system (1.1) is approximately controllable.
Proof. Theorem 4.2, ensures that for every λ > 0 and ξj ∈ X, for j = 0, 1, . . . p, the system
(1.1) has a mild solution, say, xλ ∈ Er with the control uγλ same as given in (4.1). Then, it
is immediate that
 Z t h i

C (t)ψ(0) + T (t)η + (t − s) γ−1
S (t − s) Bu γ
(s) + f (s, ˜
xλ ) ds,
̺(s,x˜λ
 γ γ γ

 λ s)


 0
t ∈ [0, τ1 ],





hj (t, x˜λ (τj− )),




 t ∈ (τj , sj ], j = 1, . . . , p,

− sj )hj (sj , x˜λ (τj− )) + Tγ (t − sj )h′j (sj , x˜λ (τj− ))

xλ (t) = Cγ (t Z (4.18)
 sj h i
(sj − s)γ−1 Sγ (sj − s) Buγλ (s) + f (s, x˜λ ̺(s,x˜λ ) ) ds



 −
 s

 0
t

 Z h i
γ ˜
 γ−1


 + (t − s) Sγ (t − s) Buλ(s) + f (s, x ̺(s,x˜λ ) ) ds, λ
 s


 0
t ∈ (sj , τj+1 ], j = 1, . . . , p.


Next, we estimate
xλ (T ) = Cγ (T − sp )hp (sp , x˜λ (τp− )) + Tγ (T − sp )hp (sp , x˜λ (τp− ))
Z sp h i
− γ−1 γ ˜ λ
(sp − s) Sγ (sp − s) Buλ (s) + f (s, x ̺(s,x˜λ ) ) ds
s
0
Z T h i
+ (T − s)γ−1 Sγ (T − s) Buγλ (s) + f (s, x˜λ ̺(s,x˜λ ) ) ds
s
0

= Cγ (T − sp )hp (sp , x˜λ (τp− )) + Tγ (T − sp )hp (sp , x˜λ (τp− ))

Z sp h i
− (sp − s)γ−1 Sγ (sp − s) Buγλ (s) + f (s, x˜λ ̺(s,x˜λ ) ) ds
s
0
Z T Z sp
+ (T − s)γ−1 Sγ (T − s)f (s, x˜λ ̺(s,x˜λ ) ) + (T − s)γ−1 Sγ (T − s)Buγλ (s)ds
s
0 0
Z T h i
+ (T − s)γ−1 Sγ (T − s)BB∗ Sγ (T − s)∗ J R(λ, ΦTsp )gp (xλ (·)) ds
sp

= Cγ (T − sp )hp (sp , x˜λ (τp− )) + Tγ (T − sp )hp (sp , x˜λ (τp− ))

Z sp h i
− (sp − s)γ−1 Sγ (sp − s) Buγλ (s) + f (s, x˜λ ̺(s,x˜λ ) ) ds
s
0
Z T Z sp
+ (T − s)γ−1 Sγ (T − s)f (s, x˜λ ̺(s,x˜λ ) ) + (T − s)γ−1 Sγ (T − s)Buγλ (s)ds
s
0 0
h i
T T λ
+ Φsp J R(λ, Φsp )gp (x (·))
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 27

= ξp − λR(λ, ΦTsp )gp (xλ (·)). (4.19)

The sequence xλ (t) is bounded in X, for each t ∈ J, follows from the argument xλ ∈ Er .
Then by the Banach-Alaoglu theorem, we can find a subsequence, still labeled as xλ such
that
w
⇀ z(t) in X as λ → 0+ , t ∈ J.
xλ (t) −
Using the condition (H2 ) of Assumption 2.17, we have the following convergence:
hp (t, xλ (τp− )) → hp (t, z(τp− )) in X as λ → 0+ , for all t ∈ J, (4.20)
h′p (t, xλ (τp− )) → h′p (t, z(τp− )) in X as λ → 0+ , for all t ∈ J. (4.21)
Moreover, by using Assumption (H3), we obtain
Z t2 2 Z t2 Z t2 2δ
˜λ
f (s, x ̺(s,x˜λ ) ) ds ≤
2
φ (s)ds ≤
1
φ (s)ds
δ (t1 − t2 )1−2δ < +∞,
s X
t1 t1 t1

for any t1 , t2 ∈ [0, T ] with t1 < t2 . The above estimate guarantees that the sequence
{f (·, x˜λ ̺(s,x˜λ ) ) : λ > 0} in L2 ([t1 , t2 ]; X) is bounded. Once again by applying the Banach-
s

Alaoglu theorem, we can extract a subsequence still denoted by{f (·, x˜λ ˜λ ) : λ > 0} suchρ(s,x s)
that
f (·, x˜λ ρ(s,x˜λ s ) ) −
w
⇀ f (·) in L2 ([t1 , t2 ]; X) as λ → 0+ . (4.22)
Next, we compute
Z sp
kuγλ (s)k2U ds
0
Z τ1 Z s1 Z τ2 Z sp
2 2 2
= γ
kuλ (s)kU ds + γ
kuλ (s)kU ds + γ
kuλ (s)kU ds + · · · + kuγλ (s)k2U ds
0 τ1 s1 τp
Z τ1 Z τ2 Z τp
γ 2 γ 2 γ 2
= u (s) ds +
0,λ U
u (s) ds + · · · +
1,λ U
u
p−1,λ (s) U ds

0 s1 sp−1
!2 p−1
2γ+1 X
M M̃ T
≤ Cj2 = C,
λΓ(2γ) 2γ + 1 j=0

where Cj , for j = 1, . . . , p − 1 are the same as given in (4.5) and C0 = N0 given in (4.4).
Moreover, the above fact implies that the sequence {uγλ(·) : λ > 0} in L2 ([0, sp ]; U) is bounded.
Further, by an application of the Banach-Alaoglu theorem, we obtain a subsequence, still
denoted by {uγλ (·) : λ > 0} such that
w
uγλ (·) −
⇀ uγ (·) in L2 ([0, sp ]; U) as λ → 0+ . (4.23)
Next, we compute

gp (xλ (·)) − ω
X

˜λ − −
≤ Cγ (T − sp )(hp (sp , x (τp )) − hp (sp , z(τp )))

X

′ ˜ − ′ −
+ Tγ (T − sp )(hp (sp , xλ (τp )) − hp (sp , z(τp )))

X
28 S. ARORA, M. T. MOHAN AND J. DABAS
Z sp
γ
γ−1 γ

+ (s p − s) Sγ (s p − s)B[u λ (s) − u (s)]ds

Z0 sp h
X
i
˜
γ−1

+ λ
(sp − s) Sγ (sp − s) f (s, x ̺(s,x˜λ ) ) − f (s) ds
s
Z0 sp X

(T − s)γ−1 Sγ (T − s)B[uγλ (s) − uγ (s)]ds

+
0 X
Z T h i
˜
γ−1

+ (T − s) Sγ (T − s) f (s, xλ ̺(s,x˜λ ) ) − f (s) ds
s
0 X
λ − −

≤ Cγ (T − sp )(hp (sp , x (τp )) − hp (sp , z(τp ))) X

+ Tγ (T − sp )(h′p (sp , xλ (τp− )) − h′p (sp , z(τp− ))) X

Z sp
γ
γ−1 γ

+ (s p − s) Sγ (s p − s)B[u λ (s) − u (s)]ds

0 X
Z sp h i
˜
γ−1

+ λ
(sp − s) Sγ (sp − s) f (s, x ̺(s,x˜λ ) ) − f (s) ds s
0 X
2γ−1 2γ−1
Z sp  21
T − (T − sp ) 2
+ kSγ (T − s)B[uγλ (s) −u γ
(s)]kX ds
2γ − 1 0
T
Z h i
˜
γ−1

+ (T − s) λ
Sγ (T − s) f (s, x ̺(s,x˜λ ) ) − f (s) ds
s
0 X
+
→ 0 as λ → 0 , (4.24)
where
ω = ξp − Cγ (T − sp )hp (sp , z(τp− )) − Tγ (T − sp )h′p (sp , z(τp− ))
Z T Z sp
γ−1
− (T − s) Sγ (T − s)f (s)ds + (sp − s)γ−1 Sγ (sp − s)[Buγ (s) + f (s)]ds
Z0 sp 0

− (T − s)γ−1 Sγ (T − s)Buγ (s)ds.

0

In (4.24), we have used the convergences

R· (4.20), (4.21), (4.22), (4.23), DCT and the com-
pactness of the operator (Qf )(·) = 0 (· − s)γ−1 Sγ (· − s)f (s)ds : L2 (J; X) → C(J; X) (see
Lemma 3.4). Finally, by using the equality (4.19), we estimate
λ
x (T ) − ξp ≤ T
λR(λ, Φsp )gp (x(·))

X X

T
≤ λR(λ, Φsp )(gp (x(·)) − ω) + λR(λ, ΦTsp )ω

X X

≤ λR(λ, ΦTsp ) kgp (x(·)) − ωkX + λR(λ, ΦTsp )ω . (4.25)

L(X) X

Using the above inequality, (4.24) and Assumption 2.17 (H0 ), we obtain
x (T ) − ξp → 0 as λ → 0+ .
λ
X

Hence, the system (1.1) is approximately controllable on J. 

 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 29

5. Application
In this section, we investigate the approximate controllability of the fractional order wave
equation with non-instantaneous impulses and delay:
Example 5.1. Let us consider the following fractional system:
Z t
∂ 2 v(t, ξ)
 α
∂ v(t, ξ)

 = + w(t, ξ) + b(t − s)v(s − β(kv(t)k), ξ)ds,
∂tα ∂ξ 2



 −∞
m



 [



 t∈ (sj , τj+1 ] ⊂ J = [0, T ], ξ ∈ [0, π],
j=0




v(t, ξ) = hj (t, v(t− j , ξ)), t ∈ (τj , sj ], j = 1, . . . , m, ξ ∈ [0, π], (5.1)
 −

 ∂v(t, ξ) ∂hj (t, v(τj , ξ))
= , t ∈ (τj , sj ], j = 1 . . . , m, ξ ∈ [0, π],


∂t ∂t




v(t, 0) = 0 = v(t, π), t ∈ J,





 v(θ, ξ) = ψ(θ, ξ), ∂v(0, ξ) = ζ (ξ), ξ ∈ [0, π], θ ≤ 0,



0
∂t
where α ∈ (1, 2) and the function w : J × [0, π] → [0, π] is square integrable in t and ξ, and
the functions β : [0, ∞) → [0, ∞) and b : [0, ∞) → R are also continuous. The functions
ψ(·, ·) and hj for j = 1, . . . , p satisfy suitable conditions, which will be specified later.

Step 1: Strongly continuous families and phase space: Let Xp = Lp ([0, π]; R), for p ∈ [2, ∞),
and U = L2 ([0, π]; R). Note that Xp is a separable reflexive Banach space with strictly convex
p
dual X∗p = L p−1 ([0, π]; R) and U is separable. We define the operator Ap : D(Ap ) ⊂ Xp → Xp
as
Ap f (ξ) = f ′′ (ξ), where D(Ap ) = W2,p ([0, π]; R) ∩ W01,p ([0, π]; R). (5.2)
Moreover, the spectrum of the operator Ap is given by σ(Ap ) = {−n2 : n ∈ N}. Then, for
every f ∈ D(Ap ), the operator Ap can be written as
X∞ Z π
2
Ap f = −n hf, wn iwn , hf, wn i := f (ξ)wn (ξ)dξ,
n=1 0
q
where wn (ξ) = π2 sin(nξ) are the normalized eigenfunctions (with respect to the X2 norm)
of the operator Ap corresponding to the eigenvalues −n2 , n ∈ N. The operator Ap satisfies
all the conditions (R1 )-(R3 ) of Assumption 2.5 (see application section of [51]). By applying
Lemma 2.7, we deduce the existence of a strongly continuous cosine family Cp (t), t ∈ R in
Xp . The compactness of the associated strongly continuous sine family Sp (t), t ∈ R, follows
by Lemma 2.8. The strongly continuous families {Cp (t) : t ∈ R} and {Sp (t) : t ∈ R} can be
written as
X∞
Cp (t)f = cos(nt)hf, wn iwn , f ∈ Xp ,
n=1
∞
X 1
Sp (t)f = sin(nt)hf, wn iwn , f ∈ Xp .
n=1
n
30 S. ARORA, M. T. MOHAN AND J. DABAS

Next, we define the operator

Z ∞ ∞
X 1
Sγ,p (t)f = γθMγ (θ) sin(ntγ θ)hf, wn iwn dθ, f ∈ Xp ,
0 n=1
n
where γ = α2 .
Phase space: The space B = PC g (Xp ) (see, Example 2.11) is a phase space satisfying the
axioms (A1) and (A2) with P(t) = t and Q(t) = O(t), (see, section 5 of [51]). Specifically,
one can choose the function g(θ) = eaθ , for a < 0. In this case, we assume the following
condition:
(C1) The function ψ ∈ PC g (X), we assume L := ess sup |b(−θ)|g(θ).
θ∈(−∞,0]

Step 2: Approximate controllability. Let us define

x(t)(ξ) := v(t, ξ), for t ∈ J and ξ ∈ [0, π],
and the operator B : U → Xp as
Z π
Bu(t)(ξ) := w(t, ξ) = K(ζ, ξ)u(t)(ζ)dζ, t ∈ J, ξ ∈ [0, π],
0
where K ∈ C([0, π] × [0, π]; R) is a symmetric kernel, that is, K(ζ, ξ) = K(ξ, ζ), for all
ζ, ξ ∈ [0, π]. We assume that the operator B is one-one. Hence, the operator B is bounded
(see application section of [2]). The symmetry of the kernel ensures that the operator B is
self-adjoint, that is, B = B∗ . For example, one can take K(ξ, ζ) = e(ξ+ζ) , for all ξ, ζ ∈ [0, π].
The function ψ : (−∞, 0] → X is given as
ψ(t)(ξ) = ψ(t, ξ), ξ ∈ [0, π].
Next, the functions f, ̺ : J × B → X are defined as
Z 0
f (t, ψ)ξ := b(−θ)ψ(θ, ξ)dθ,
−∞
̺(t, ψ) := t − β(kψ(0)kX ),
for ξ ∈ [0, π]. It is easy to verify that the function f is continuous and uniformly bounded by
L. Hence, the function f fulfills the condition (H1) of Assumption 2.17 and the condition
(H3).
Moreover, the impulse functions hj : [τj , tj ] × Xp → Xp , for j = 1, . . . , m, are defined as
Z π
hj (t, x)ξ := ρj (t, ξ, z) cos2 (x(τj− )z)dz, for t ∈ (τj , sj ],
0
1 2
where, ρj ∈ C (J × [0, π] ; R). Clearly, the impulses hj for j = 1, . . . , m, satisfy the condition
(H2) of Assumption 2.17.
Using the above transformations, the system (5.1) can be rewritten in the form (1.1)
which satisfies Assumption 2.17 (H1)-(H2) and Assumption (H3 ). Next, we show that
the corresponding linear fractional control system of the equation (1.1) is approximately
controllable. To achieve this, we assume that
B∗ Sγ,p (T − t)∗ x∗ = 0, for any x∗ ∈ X∗p , 0 ≤ t < T.
Since the operator B∗ is one-one, we get that
Sγ,p (T − t)∗ x∗ = 0, for all t ∈ [0, T ). (5.3)
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 31

Let us compute
R∞ P∞ 1
Sγ,p (T − t)∗ x∗ 0
γθMγ (θ) sin(n(T − t)γ θ)hx∗ , wn iwn dθ
n=1 n
lim = lim
t↓T (T − t)γ t↓T (T − t)γ
∞ Z ∞
sin(n(T − t)γ θ)
 
X 1
= lim γ θMγ (θ) γ
dθhx∗ , wn iwn
t↓T
n=1
n 0 (T − t)
∞ ∞
sin(n(T − t)γ θ)
 
1
X Z
= γ θMγ (θ) lim γ
dθhx∗ , wn iwn
n=1
n 0 t↓T (T − t)
X∞ Z ∞
= γ θ2 Mγ dθhx∗ , wn iwn
n=1 0
∞
2γ X
= hx∗ , wn iwn , (5.4)
Γ(1 + 2γ) n=1
where γ ∈ (0, 1). Combining the estimates (5.3) and (5.4), we obtain
∞
X
hx∗ , wn iwn = 0,
n=1
∗
which implies that x = 0. Thus, by applying Lemma 3.5 and Remark 3.6, we deduce that
the corresponding linear fractional control system of (1.1) is approximately controllable.
Finally, by Theorem 4.3, we conclude that the semilinear fractional control system (1.1)
(equivalent to the system (5.1)) is approximately controllable.

6. Conclusions
In this manuscript, we considered the semilinear fractional control system (1.1) of order
1 < α < 2, where the operator A generates a strongly continuous cosine family {C(t) : t ∈
R}. We first formulated the linear regulator problem and obtained the optimal control in
the feedback form. With the help of this optimal control, we discussed the approximate
controllability of the linear control system (3.2). Then, we proved the existence of a mild
solution of the semilinear fractional control system (1.1) for the suitable control function
defined in (4.1) via the resolvent operator and the Schauder fixed point theorem. Then, we
derived sufficient conditions for the approximate controllability of the system (1.1), whenever
the corresponding linear control system is approximately controllable. Moreover, we modified
the axioms of phase space to deal with impulsive functional differential equations. Note that,
in this work, we have used the idea of α/2-resolvent family related to cosine family generated
by the operator A.
The existence of solutions and the approximate controllability results of the semilinear
fractional abstract Cauchy problem of order α ∈ (1, 2) by using the idea related to the α-
resolvent family introduced in [18] is not investigated so far, and it will be addressed in a
future work.
Acknowledgments: S. Arora would like first to thank the Council of Scientific and Indus-
trial Research, New Delhi, Government of India (File No. 09/143(0931)/2013 EMR-I), for
financial support to carry out his research work and also thank the Department of Math-
ematics, Indian Institute of Technology Roorkee (IIT Roorkee), for providing stimulating
scientific environment and resources. M. T. Mohan would like to thank the Department of
32 S. ARORA, M. T. MOHAN AND J. DABAS

Science and Technology (DST), Govt of India for Innovation in Science Pursuit for Inspired
Research (INSPIRE) Faculty Award (IFA17-MA110). J. Dabas would like to thank the
Council of Scientific and Industrial Research, New Delhi, Government of India, project (ref.
no. 25(0315)/20/EMR-II).

References
[1] S. Arora, M.T. Mohan and J. Dabas, Approximate controllability of a Sobolev type im-
pulsive functional evolution system in Banach spaces, Math. Control Relat. Fields, (2020),
https://doi:10.3934/mcrf.2020049.
[2] S. Arora, M.T. Mohan and J. Dabas, Approximate controllability of fractional order non-
instantaneous impulsive functional evolution equations with state-dependent delay in Banach spaces,
https://arxiv.org/abs/2106.02939v1.
[3] S. Arora, M.T. Mohan and J. Dabas, Approximate controllability of the non-autonomous impulsive
evolution equation with state-dependent delay in Banach spaces, Nonlinear Anal. Hybrid Syst., 39
(2021), 100989.
[4] O. Arino, K. Boushaba and A. Boussouar, A mathematical model of the dynamics of the phytoplankton-
nutrient system. Spatial hetrogeneity in ecological models, Nonlinear Anal. Real World Appl., 1 (2000),
69-87.
[5] E. Asplund, Averaged norms, Israel J. Math., 5 (1967), 227-233.
[6] V. Barbu, Analysis and Control of Nonlinear Infinite Dimensional Systems, Academic Press, New York,
1993.
[7] V. Barbu, Controllability and Stabilization of Parabolic Equations, Springer International Publishing
AG, 2018.
[8] M. Benchohra, J. Henderson and S. Ntouyas, Impulsive Differential Equations and Inclusions, Hindawi
Publishing Corporation, New York 2006.
[9] F. Chen, D. Sun and J. Shi, Periodicity in a food-limited population model with toxicants and state
dependent delays, J. Math. Anal. Appl., 288 (2003), 136-146.
[10] D.N. Chalishajar, K. Malar and K. Karthikeyan, Approximate controllability of abstract impulsive
fractional neutral evolution equations with infinite delay in Banach spaces, Electron. J. Differ. Equ.,
275 (2013), 1–21 .
[11] J. dabs, A. chauhan and M. Kumar, Existence of the mild solutions for impulsive fractional equations
with infinite delay, Int. J. Differ. Equ., 2011 (2011), 793023.
[12] G. Da Prato and J. Zabczyk, Ergodicity for Infinite Dimensional Systems, London Mathematical Society
Lecture Notes, Cambridge University Press, 1996.
[13] I. Ekeland and T. Turnbull, Infinite Dimensional Optimization and Convexity, Chicago press, London,
1983.
[14] M. Fabian et.al., Functional Analysis and Infinite Dimensional Geometry, CMS Books in Mathematics,
Springer-Verlag, New York, 2001.
[15] G.R. Gautam and J. Dabas, Mild solutions for class of neutral fractional functional differential equations
with not instantaneous impulses, Appl. Math. Comput., 259 (2015), 480-489.
[16] A. Grudzka and K. Rykaczewski, On approximate controllability of functional impulsive evolution in-
clusions in a Hilbert space, J. Optim. Theory Appl., 166 (2015), 414-439.
[17] L. Guedda, Some remarks in the study of impulsive differential equations and inclusions with delay,
Fixed Point Theory, 12 (2011), 349–354.
[18] H.R. Henrı́quez, J. G. Mesquita, J. C.Pozo, Existence of solutions of the abstract Cauchy problem of
fractional order, J. Funct. Anal., (2021), 109028, doi: https://doi.org/10.1016/j.jfa.2021.109028.
[19] J.W. He, Y. Liang, B. Ahmad and Y. Zhou, Nonlocal fractional evolution inclusions of order α ∈ (1, 2),
Mathematics, 209 (2019), 1–17 .
[20] E. Hernández and D. O’Regan, On a new class of abstract impulsive differential equations, Proc. Amer.
Math. Soc. 141 (2013), 1641–1649.
[21] R. Hilfer, Applications of Fractional Calculus in Physics, World Scientific: Singapore, 2000.
[22] Y. Hino, S. Murakami and T. Naito, Functional Differential Equations with Infinite Delay, Lecture
Notes in Mathematics, Springer-Verlag, Berlin, 1991.
 APPROXIMATE CONTROLLABILITY OF FRACTIONAL ORDER IMPULSIVE SYSTEMS 33

[23] A.A. Kilbas, H.M. Srivastava and J.J. Trujillo, Theory and Applications of Fractional Differential Equa-
tions, in: North-Holland Mathematics Studies, Elsevier Science B.V., Amsterdam, 2006.
[24] J. Kisyński, On cosine operator functions and one parameter group of operators, Studia Math., 49
(1972), 93–105.
[25] S. Kumar and N. Sukavanam, Approximate controllability of fractional order semilinear systems with
bounded delay, J. Differential Equations, 252 (2012), 6163–6174.
[26] J. Klafter and R. Metzler, The random walk’s guide to anomalous diffusion: a fractional dynamics
approach, Phys. Rep. 339 (2000), 1-77.
[27] J. Kemppainen, J. Siljander, V. Vergara and R. Zacher, Decay estimates for time fractional and other
non-local in time subdiffusion equations in Rd , Math. Annalen 366 (2016) 941-979.
[28] V. Lakshmikantham, D. D. Bainov and P.S. Simeonov, Theory of Impulsive Differential Equations,
World Scientific, Singapore (1989).
[29] X. Li. and J. Yong, Optimal Control Theory for Infinite Dimensional Systems, Birkhäuser Basel, 1995.
[30] A. Lunardi, On the linear heat equation with fading memory, SIAM J. Math. Anal., 21 (1990), 1213-
1224.
[31] N.I. Mahmudov, Approximate controllability of fractional neutral evolution equations in Banach spaces,
Abstr. Appl. Anal., 2013 (2013), 531894.
[32] N.I. Mahmudov, Approximate controllability of fractional Sobolev-type evolution equations in Banach
spaces, Abstr. Appl. Anal., 2013 (2013), 502839.
[33] N.I. Mahmudov, Approximate controllability of semilinear deterministic and stochastic evolution equa-
tions in abstract spaces, SIAM J. Control Optim., 42 (2003), 1604-1622.
[34] F. Mainardi, Fractional Calculus and Waves in Linear Viscoelasticity: An Introduction to Mathematical
Models, Imperial College Press, 2010.
[35] D. Nesic and A.R. Teel, Input-to-state stability of networked control systems, Automatica 40 (2004),
2121–2128.
[36] J.W. Nunziato, On heat conduction in materials with memory, Quart. Appl. Math., 29 (1971), 187-204.
[37] V. Obukhovski and J.C. Yao, On impulsive functional differential inclusions with Hille-Yosida operators
in Banach spaces, Nonlinear Anal., 73 (2010), 1715-1728.
[38] A. Pazy, Semigroup of Linear operators and Applications to partial equations, Springer-Verlag, New
York, 1983.
[39] I. Podlubny, Fractional Differential Equations, Academic Press, San Diego, 1999.
[40] H. Qin, X. Zuo, J. Liu and L. Liu, Approximate controllability and optimal controls of fractional
dynamical systems of order 1 < q < 2 in Banach spaces, Adv. Differ. Equ., 2015 (2015) 1–17.
[41] M.M. Raja and V. Vijayakumar, New results concerning to approximate controllability of fractional
integro-differential evolution equations of order 1 < r < 2, Numer. Methods Partial Differ. Eq., (2020),
1-16, https://doi.org/10.1002/num.22653.
[42] M.M. Raja, V. Vijayakumar, and R. Udhayakumar, Results on the existence and controllability of
fractional integro-differential system of order 1 < r < 2 via measure of noncompactness, Chaos, Solitons
& Fractals, 139 (2020): 110299.
[43] M.M. Raja, V. Vijayakumar, R. Udhayakumar and Y. Zhou, A new approach on the approximate
controllability of fractional differential evolution equations of order 1 < r < 2 in Hilbert spaces, Chaos,
Solitons & Fractals, 141 (2020): 110310.
[44] Y.A. Rossikhin and M.V. Shitikova, Application of fractional derivatives to the analysis of damped
vibrations of viscoelastic single mass systems, Acta Mech., 120 (1997), 109–125.
[45] S.G. Samko, A.A. Kilbas and O.I. Marichev, Fractional Integrals and Derivatives, Gordon and Breach
Science Publishers, Switzerland, 1993.
[46] R. Sakthivel and E.R. Anandhi, Approximate controllability of impulsive differential equations with
state-dependent delay, Internat. J. Control, 83 (2010), 387-393.
[47] R. Sakthivel, R. Ganesh, Y. Ren and S.M. Anthoni, Approximate controllability of nonlinear fractional
dynamical systems, Commun. Nonlinear Sci. Numer. Simul., 18 (2013), 3498–508.
[48] S.G. Samko, A.A. Kilbas and O.I. Marichev, Fractional Integrals and Derivatives, Gordon and Breach
Science Publishers, Switzerland, 1993.
 34 S. ARORA, M. T. MOHAN AND J. DABAS

[49] L. Shu, X.B. Shu and J. Mao, Approximate controllability and existence of mild solutions for Rie-
mann–Liouville fractional stochastic evolution equations with nonlocal conditions of order 1 < α < 2,
Fract. Calculus Appl. Anal., 22 (2019) 1086–1112 .
[50] X.B. Shu and Q. Wang, The existence and uniqueness of mild solutions for fractional differential equa-
tions with nonlocal conditions of order 1 < α < 2, Comput. Math. Appl., 64 (2012), 2100–2110.
[51] S. Singh, S. Arora, M. T. Mohan and J. dabas, Approximate controllability of second order impul-
sive systems with state-dependent delay in Banach spaces, Evol. Equ. Control Theory, (2020), doi:
10.3934/eect.2020103.
[52] M.S. Tavazoei, M. Haeri, S. Jafari, S. Bolouki and M. Siami, Some applications of fractional calculus in
suppression of chaotic oscillations, IEEE Trans. Ind. Electron., 55 (2008), 4094–4101.
[53] C.C. Travis and G.F. Webb, Cosine families and abstract nonlinear second order differential equations,
Acta Math. Hungar., 32 (1978), 75-96.
[54] C.C. Travis and G.F. Webb, Second order differential equations in Banach space, in Nonlinear Equations
in Abstract Spaces, Academic Press, 1978, 331-361.
[55] R. Triggiani, Addendum: A note on the lack of exact controllability for mild solutions in Banach spaces,
SIAM J. Control Optim., 18 (1980), 98.
[56] R. Triggiani, A note on the lack of exact controllability for mild solutions in Banach spaces,SIAM J.
Control Optim., 15 (1977), 407-411.
[57] J.R. Wang and M. Fečkan, A general class of impulsive evolution equations, Topol. Methods Nonlinear
Anal. 46 (2015), 915–933.
[58] J.R. Wang and M. Fečkan, Non-Instantaneous Impulsive Differential Equations, IOP, 2018.
[59] J. Wang and Y. Zhou, Existence and controllability results for fractional semilinear differential inclu-
sions, Nonlinear Anal. Real. World Appl., 12 (2011), 3642–3653.
[60] J.R. Wang, Y. Zhou and Z. Lin, On a new class of impulsive fractional differential equations, Appl.
Math. Comput. 242 (2014), 649–657.
[61] Y. Tian, J.R. Wang and Y. Zhou, Almost periodic solutions for a class of non-instantaneous impulsive
differential equations, Quaest. Math. 42 (2019), 885–905.
[62] V. Vijayakumar, C. Ravichandran, K.S. Nisar and K.D. Kucche, New discussion on approximate con-
trollability results for fractional Sobolev type Volterra-Fredholm integro-differential systems of order
1 < r < 2, Numer. Methods Partial Differ. Eq., (2021), 1–19, https://doi.org/10.1002/num.22772.
[63] T. Yang and L.O. Chua, Impulsive control and synchronization of nonlinear dynamical systems and
application to secure communication, Internat. J. Bifur. Chaos Appl. Sci. Engrg. 7 (1997), 645–664.
[64] Z. Yan, Approximate controllability of partial neutral functional differential systems of fractional order
with state-dependent delay, Internat. J. Control, 85 (2012), 1051-1062.
[65] C.C. Yeh, Discrete inequalities of the Gronwall-Bellman type in n independent variables, J. Math. Anal.
Appl., 105 (1985), 322-332.
[66] Y. Zhou and J.H. He, New results on controllability of fractional evolution systems with order α ∈ (1, 2),
Evol. Equ. Control Theory, (2020), https://doi.org/10.3934/eect.2020077.
[67] E. Zuazua, Controllability and observability of partial differential equations: some results and open
problems, in Handbook of differential equations: evolutionary equations, 3 (2007), 527-621.

View publication stats