Nonlinear Dynamics and Systems Theory, 11 (2) (2011) 165–172

Existence of Almost Automorphic Solutions of Neutral

Functional Differential Equation
I. Mishra 1 , D. Bahuguna 1∗ and S. Abbas 2

1
Department of Mathematics and Statistics, Indian Institute of Technology Kanpur,
Kanpur — 208016, India
2
School of Basic Sciences, Indian Institute of Technology Mandi,
Mandi, H.P., 175001, India.

Received: April 16, 2010; Revised: March 30, 2011

Abstract: In this work we use the theory of evolution semigroup of bounded linear
operators and fixed point theorem to establish the existence and uniqueness of a mild
solution of a neutral functional differential equation in a Banach space.

Keywords: almost automorphic function; evolution semigroup; neutral functional

differential equation; mild solution.

Mathematics Subject Classification (2000): 37L05, 34K06, 34A12.

1 Introduction

In 1964, S. Bochner introduced almost automorphic functions in one of his landmark pa-
per [10]. Almost automorphic functions are more general than almost periodic functions.
Many authors had established the almost periodic solution of differential equations in
abstract spaces ([8, 9, 13, 15], etc.). The theory has been generalized by many authors
for almost automorphic solutions ([11, 12, 14], etc.). Goldstein [14] has considered the
following differential equation in a Banach space X
dx(t)
= Ax(t) + f (t, x(t)), t ∈ R, (1)
dt
where A generates an exponentially stable C0 - semigroup and f be a jointly continuous
function and shown the existence of almost automorphic solution of the problem if f is
almost automorphic.
∗ Corresponding author: mailto:dhiren@iitk.ac.in

c 2011 InforMath Publishing Group/1562-8353 (print)/1813-7385 (online)/http://e-ndst.kiev.ua 165

166 I. MISHRA, D. BAHUGUNA AND S. ABBAS

These days, functional differential equations have been of very great interest, for many
mathematicians. Bahuguna [1] studied a class of partial functional differential equtations
and its application to population dynamics. Analytical techniques of semigroup theory
have been applied in [2], [3] and [4], which we are also going to use in this paper.
Bahuguna and Muslim [5] also considered the second order history valued delay dif-
ferential equations [4] and used evolution equations and semigroup theory to find approx-
imation of a solution. Recently, D.N. Pandey, A. Ujlayan and D. Bahuguna [6] proved
existence and uniqueness of a hyperbolic integrodifferential equation with a nonlocal
condition.
Abbas and Bahuguna [7] considered the following nonautonomous neutral functional
differential equations
d
(x(t) − F1 (t, x(t − g(t)))) = A(t)x(t) + F2 (t, x(t), x(t − g(t))), (2)
dt
where A(t) generates an exponentially stable evolution systems and g is a continuous
function. The authors have shown the existence of an almost periodic mild solutions
using Kransnoselskii’s fixed point theorem and theory of evolution operator. They also
assumed the well known Acquistapace–Terreni conditions which ensure the existence of
evolution family.
In the present work we study the existence of an almost automorphic solution of
equation (2) using the evolution semigroup and the Banach fixed point approach.

2 Preliminaries

Let X be a complex Banach space endowed with the norm k.kX . N, R and C stand for
Natural, Real and Complex numbers respectively. Let B(X) be a Banach space of all
bounded linear operators from X to itself; endowed with norm k.kB(X) given by

kLkB(X) = sup{kLxkX : x ∈ X and kxkX ≤ 1}.

Now, we will recall certain definitions to be used subsequently in this paper.

Definition 2.1 A continuous function f : R → X is said to be almost automorphic
if for every sequence {sn }n∈N of real numbers there exists a subsequence {τn }n∈N such
that limn→∞ f (t + τn ) = g(t) and limn→∞ g(t − τn ) = f (t) for all t ∈ R.
We denote by AA(X) the set of all such functions.
Definition 2.2 A continuous function f : R × X → X is said to be almost automor-
phic if f (t, x) is almost automorphic for each t ∈ R uniformly for all x ∈ Y , where Y is
any bounded subset of X.
Equivalently, for every sequence of real numbers {sn }n∈N we can extract a subse-
quence {τn }n∈N such that g(t, x) = limn→∞ f (t + τn , x) is well defined for all t ∈ R and
for all x ∈ Y and f (t, x) = limn→∞ g(t − τn , x) is well defined for all t ∈ R and for all
x∈Y.

Lemma 2.1 (AA(X), k.kAA(X) ) is a Banach space with supremum norm, given by
kf kAA(X) = supt∈R kf (t)k.

Lemma 2.2 If f : R → X is almost automorphic, then f is bounded.

 NONLINEAR DYNAMICS AND SYSTEMS THEORY, 11 (2) (2011) 165–172 167

For the proof of the above two lemmas, we refer to [12].

Lemma 2.3 Suppose Z and W are Banach spaces. Let F : R × Z → W be an almost
automorphic function in t ∈ R, for each z ∈ Z and assume that F satisfies Lipschitz
condition in z uniformly in t ∈ R. Let φ : R → Z be an almost automorphic function,
then the function Φ : R → W, defined by Φ(t) = f (t, φ(t)) is almost automorphic.
In [18], Acquistapace and Terreni gave conditions on A(t), t ∈ R, which ensure the
existence of unique evolution family {U (t, s) : t ≥ s > −∞} on X, such that
Z t
u(t) = U (t, 0)u(0) + U (t, ξ)f (ξ)dξ,
0

where u(t) satisfies

du(t)
= A(t)u(t) + f (t), t ∈ R.
dt
Lemma 2.4 ATC (Acquistapace–Terreni condition). Let
π
Sθ = {λ ∈ C : | arg λ| ≤ θ} ∪ {0} ⊂ ρ(A(t)), θ ∈ ( , π).
2
If there exist a constant K0 and a set of real numbers α1 , ..., αk , β1 , ...βk with 0 ≤ βi <
αi ≤ 2, i = 1, 2, ...k, such that

kA(t)(λ − A(t))−1 (A(t)−1 − A(s)−1 )kB(X) ≤ K0 Σki=1 (t − α)αi |λi |βi −1

for t, s ∈ R and λ ∈ Sθ \{0} and there exists constant M ≥ 0 such that

M
k(λ − A(t))−1 k ≤ , λ ∈ Sθ ,
1 + |λ|

then there exists a unique evolution family {U (t, s) : t ≥ s > −∞} on X.

These conditions resulting from Theorem 2.3 of [17] are known as ”Acquistapace–
Terreni conditions”.

Definition 2.3 A mild solution of (2) is a continuous function x : R → X, satisfying

x(t) − F1 (t, x(t − g(t)) = U (t, s)(x(s) − F1 (s, x(s − g(s))))

Z t
+ U (t, ξ)F2 (ξ, x(ξ), x(ξ − g(ξ)))dξ (3)
a

for t ≥ s all s ∈ R.

Note: We say, an evolution family {U (t, s)}t≥s>−∞ is exponentially stable, if ∃ M ≥

1 and δ > 0 such that kU (t, s)k ≤ M e−δ(t−s) for t ≥ s. When s → −∞ the above
equation takes the form
Z t
x(t) = F1 (t, x(t − g(t))) + U (t, ξ)F2 (ξ, x(ξ), x(ξ − g(ξ)))dξ.
−∞
168 I. MISHRA, D. BAHUGUNA AND S. ABBAS

Assumptions:
(C1 ) : F1 (t, x), F2 (t, x, y) are almost automorphic.
(C2 ) : F1 and F2 are Lipschitz continuous that is there exist positive numbers LF1 (t)
and LF2 (t) such that

kF1 (t, x) − F1 (t, y)k ≤ LF1 (t)kx − ykAA(X) ,

kF2 (t, x, u) − F2 (t, y, v)k ≤ LF2 (t)(kx − ykAA(X) + ku − vkAA(X) ).

(C3 ) : {U (t, s) : t ≥ s} is an exponentially stable evolution family on X.

(C4 ) : For every sequence {sn } of real numbers there exists a subsequence {τn } and for
any fixed s ∈ R , ǫ > 0, there exists N ∈ N such that for all n ≥ N, it follows that
δ
kU (t + τn , s + τn ) − U (t, s)k ≤ ǫe− 2 (t−s)

and
δ
kU (t − τn , s − τn ) − U (t, s)k ≤ ǫe− 2 (t−s) for all t ≥ s ∈ R.

3 Almost Automorphic Solution

We define the mapping F by

Z t
(F x)(t) = F1 (t, x(t − g(t))) + U (t, s)F2 (s, x(s), x(s) − g(s)))ds.
−∞

Lemma 3.1 For x(.) ∈ AA(X), we have F x is also almost automorphic.

Proof Since F1 is Lipschitz and F1 ∈ AA(R, X); by Lemma 2.3, we have

F1 (t, x(t − g(t))) = K(t) ∈ AA(X).

By (C2 ), we have F2 (., x(.), y(.)) ∈ AA(R × X × X, X), also we have assumed that F2 is
Lipschitz with respect to both variables x and y, further using the fact that X × X is
Banach space; hence from Lemma 2.3, one can easily see that F2 (., x(.), y(.)) ∈ AA(X).
Next, we define F2 (t, x(t), y(t)) = H(t), where H(.) ∈ AA(X). Now we show that

kF xkAA(X) < ∞,
Z t
kF x(t)kX ≤ kK(t)kX + kU (t, s)kkF2 (s, x(s), x(s − g(s)))kX ds
−∞
Z t
≤ M1 + M e−δ(t−s) kH(s)kX ds
−∞
M
≤ M1 + M2 < ∞. where sup kH(t)k = M2 .
δ t∈R

Thus, we have shown that F x is bounded.

Now, we show that (F x)(t) is almost automorphic with respect to t ∈ R. Since
H(.) ∈ AA(X) for all sequence {sn } of real numbers, there exists a subsequence {τn }
such that
 NONLINEAR DYNAMICS AND SYSTEMS THEORY, 11 (2) (2011) 165–172 169

(H1 ) : h(t) = limn→∞ H(t + τn ) is well defined for all t ∈ R.

(H2 ) : H(t) = limn→∞ h(t − τn ) is well defined for all t ∈ R.
As we are going to use Lebesgue dominated convergence theorem to show that
(F x)(t + τn ) → (Gx)(t) as n → ∞; we need to show |F x(t + τn )| < l(t) for all n ∈ N;
where l is some integrable function. Consider

(F x)(t + τn ) = F1 (t + τn , x(t + τn − g(t + τn ))

Z t+τn
+ U (t + τn , s)F2 (s, x(s), x(s − g(s)))ds.
−∞
= F1 (t + τn , x(t + τn − g(t + τn ))
Z t
+ U (t + τn , s + τn )F2 (s + τn , x(s + τn ), x(s + τn − g(s + τn )))ds.
−∞

Taking the norm on both sides, we have

Z t
k(F x)(t + τn )k ≤ kKkAA(X) + kU (t + τn , s + τn )kkH(s + τn )kds
−∞
M2 M
≤ M1 + (kHk ≤ M2 ).
δ
By (H1 ), for any fixed s ∈ R, ǫ > 0 there exists N1 ∈ N such that for all n > N1 we have

kH(s + τn ) − h(s)k ≤ ǫ.

In addition by (C4 ) for s and ǫ as above there exists N2 ∈ N such that for all n > N2

2 (t−s)
−δ
kU (t + τn , s + τn ) − U (t, s)k < ǫe .

Let N = max{N1 , N2 }, then

kU (t + τn , s + τn )H(s + τn ) − U (t, s)h(s)k

≤ kU (t + τn , s + τn ) − U (t, s)kkH(s + τn )k + kU (t, s)kkH(s + τn ) − h(s)k
2 (t−s) 2 (t−s)
−δ −δ
≤ M2 ǫe + M ǫe

⇒ U (t + τn , s + τn )H(s + τn ) → U (t, s)h(s)

as n → ∞ for all fixed s ∈ R and t ≥ s. Since K(.) ∈ AA(X), for any sequence {sn } of
real numbers there exists a subsequence {τn } such that

lim K(t + τn ) = k(t), lim k(t − τn ) = K(t).

n→∞ n→∞

Thus, we have K(t+τn ) → k(t) as n → ∞. By Lebesgue dominated convergence theorem

we get (F x)(t+τn ) → Gx(t) as n → ∞. In a similar way we can show that (Gx)(t−τn ) →
(F x)(t) as n → ∞ for all t ∈ R ⇒ F x ∈ AA(X).

Theorem 3.1 Let x(.) be an almost automorphic function and F1 , F2 and U (t, s) sat-
isfy all conditions from (C1 ) to (C4 ). Then equation (2) has unique almost automorphic
mild solution, whenever (LF1 + 2LF2 M δ ) < 1.
170 I. MISHRA, D. BAHUGUNA AND S. ABBAS

Proof It follows by Lemma 3.1, that F x ∈ AA(X), whenever x does. Let us assume
that
LF1 = sup LF1 (t), LF2 = sup LF2 (t).
t∈R t∈R

For x, y ∈ AA(X), we have:

k(F x)(t) − (F y)(t)k
≤ kF1 (t, x(t − g(t))) − F1 (t, y(t − g(t)))k
Z t
+ kU (t, s)F2 (s, x(s), x(s − g(s))) − U (t, s)F2 (s, y(s), y(s − g(s)))kds
−∞
≤ LF1 (s)kx − yk
Z t
+LF2 (s){kx(s) − y(s)k + kx(s − g(s)) − y(s − g(s))k} M e−δ(t−s) ds
−∞
Z t
≤ LF1 kx − ykAA(X) + 2LF2 kx − ykAA(X) M e−δ(t−s) ds
−∞
M
≤ LF1 kx − ykAA(X) + 2LF2 .
δ
By Banach contraction principle, F has a unique fixed point x ∈ AA(X) such that
F x = x.
Fixing s ∈ R, we have
Z t
x(t) = F1 (t, x(t − g(t))) + U (t, s)F2 (s, x(s), x(s − g(s)))ds.
−∞

Since U (t, s) = U (t, r)U (r, s) for t ≥ r ≥ s, let

Z ξ
x(ξ) = F1 (ξ, x(ξ − g(ξ))) + U (ξ, s)F2 (s, x(s), x(s − g(s)))ds
−∞
so
Z ξ
U (t, ξ)x(ξ) = U (t, ξ)F1 (ξ, x(ξ − g(ξ))) + U (t, s)F2 (s, x(s), x(s − g(s)))ds.
−∞

For t ≥ ξ,
Z t Z t
U (t, s)F2 (s, x(s), x(s − g(s)))ds = U (t, s)F2 (s, x(s), x(s − g(s)))ds
ξ −∞
Z ξ
− U (t, s)F2 (s, x(s), x(s − g(s)))ds
−∞
= x(t) − U (t, ξ)x(ξ) − F1 (t, x(t − g(t)))
+U (t, ξ)F1 (ξ, x(ξ − g(ξ))).
Hence we get
x(t) = F1 (t, x(t − g(t))) − U (t, ξ)F1 (ξ, x(ξ − g(ξ)))
Z t
+U (t, ξ)x(ξ) + U (t, s)F2 (s, x(s), x(s − g(s)))ds. (4)
ξ
 NONLINEAR DYNAMICS AND SYSTEMS THEORY, 11 (2) (2011) 165–172 171

Remark 3.1 Consider the following differential equation

Z t
d
(x(t) − F1 (t, x(t − g(t)))) = A(t)x(t) + F2 (t, x(t), G(t − s)f (s, x(s))ds), (5)
dt −∞

where G ∈ L1 (R) and f is almost automophic, Lipschitz with respect to second variable.
Now f ∈ AA(R × X, X) and f is Lipshitz by Lemma 2.3, we have f ∈ AA(X).
Let f (t, x(t)) = ψ(t).
Rt
If we can show −∞ G(t − s)f (s, x(s)) is almost automorphic, then as a consequence
of the above theorem, equation (5) has a unique almost automorphic solution.
As ψ is almost automorphic for every sequence of real numbers {tn } there exists a sub-
sequence {τn } such that limn→∞ ψ(t + τn ) = ψ1 (t) is well defined for all t ∈ R and
ψ(t) = limn→∞ ψ1 (t − τn ) is well defined for all t ∈ R.

Consider
Z t+τn Z t
k G(t + τn − s)ψ(s)ds − G(t − s)ψ1 (s)dsk
−∞ −∞
Z t Z t
=k G(t − s)ψ(s + τn )ds − G(t − s)ψ1 (s)dsk
−∞ −∞
Z t
≤ (kψ(s + τn ) − ψ1 (s)k) |G(t − s)|ds
−∞
≤ M ′ (kψ(s + τn ) − ψ1 (s)k)
Rt
for some M ′ < ∞ → 0 as n → ∞. Thus, −∞ G(t − s)ψ(s)ds is almost automorphic and
we have the result.

4 Example

Consider the following equation

d 1 √
u′′ + (ε2 u2 + 1)u′ + u = ε1 √ u2 (t − g(t)) − ε2 (cos t + cos 2t).



sin
dt sin t + sin 2t
Let u = u1 and u′1 = u2 , then we can write the above equation in matrix form as
follows
 ′     
u1 0 1 u1 d
= × + F1 (t, U (t − g(t))) + F2 (t, U (t), U (t − g(t))),
u′2 −1 −1 u2 dt
where  
u1
U= ,
u2

0

F1 (t, U (t − g(t))) = 1 √ ,
sin( sin t+sin 2t
)u21
 
0√
F2 (t, U (t), U (t − g(t))) = .
ε2 (cos t + cos 2t) − ε2 u21 u2
This is of the form (2). Thus we can apply our results to ensure the existence and
uniqueness of almost automorphic solutions.
172 I. MISHRA, D. BAHUGUNA AND S. ABBAS

Acknowledgment

The second author would like to acknowledge the financial help provided by the
Department of Science & Technology (DST), New Delhi, under its research project
SR/S4/MS:581/09.

References
[1] Bahuguna, D., Shukla, R.K. Partial functional differential equations and application to
population dynamics. Nonlinear Dyn. and Syst. Theory 5 (4) (2005) 345–356.
[2] Bahuguna, D., Muslim, M. A study of nonlocal history valued retarded differential equations
using analytic semigroups. Nonlinear Dyn. and Syst. Theory 6 (1) (2006) 63–75.
[3] Bahuguna, D., Dabas, J. Existence and uniqueness of a solution to a semilinear partial
delay differential equation with an integral condition. Nonlinear Dyn. and Syst. Theory 8
(1) (2008) 7–19.
[4] Bahuguna, D., Dabas, J. and Shukla, R. K. Method of lines to hyperbolic integro-differential
equations in Rn . Nonlinear Dyn. and Syst. Theory 8 (4) (2008) 317–328.
[5] Bahuguna, D., Muslim, M., Approximation of solutions to a class of second order history
valued delay differential equations. Nonlinear Dyn. and Syst. Theory 8 (3) (2008) 237–254.
[6] Pandey, D.N., Ujalayan, A. and Bahuguna, D. Semilinear hyperbolic integrodifferential
equations with nonlocal conditions. Nonlinear Dyn. and Syst. Theory 10 (1) (2010) 77–92.
[7] Abbas, S. and Bahuguna, D. Almost periodic solutions of neutral functional differential
equations. Comp. Math. Appl. (55-11) (2008) 2593–2601.
[8] Abbas, S. and Bahuguna, D. Almost periodic solutions of nonlinear functional differential
equations, Diff. Eqn. Dyn. Sys. 16 (3) (2008) 289–308.
[9] Abbas, S. and Bahuguna, D. Almost periodic solutions of functional differential equations
by monotone iterative method. Diff. Eqn. Dyn. Sys. (1-2) (2008) 47–62.
[10] Bochner, S., Continuous mapping of almost automorphic and almost automorphic func-
tions. Proc. Nat. Sci. 52 (1964) 907–910.
[11] Diagana, T. and NGuerekata, G. M. Almost automorphic solutionsnext term to some classes
of partial evolution equations. Appl. Maths. Letters 20 (4) (2007) 462–466.
[12] NGuerekata, G. M. Topics in almost automorphy. Springer, NY, 2005.
[13] Ezzinbi, K. and NGuerekata, G. M. Almost automorphic solutions for some partial func-
tional differential equations. Journal of Mathematical Analysis and Applications 328 (1)
(2007) 344–358.
[14] Goldstein, J. A. and N’guerekata, G. M. Almost automorphic solutions of semilinear evo-
lution equations. Proc. American Math. Soceity 133 (8) (2005) 2401–2408.
[15] Lizama, C. Mild almost periodic solutions of abstract differential equations. Journal of
Mathematical Analysis and Applications 143 (2) (1989) 560–571.
[16] Pazy, A. Semigroup of linear operatora and applications to partial differential equations.
Springer-Verlag, New York, 2005.
[17] Acquistapace, P., Evolution operators and strong solution of abstract linear parabolic equa-
tions. Differential Integral Equations (1) (1988) 433–457.
[18] Acquistapace, P. and Terreni, B. A unified approach to abstract linear parabolic equations.
Rend. Sem. Math. Uni. Padova 78 (1987) 47–107.