J. Math. Anal. Appl.

388 (2012) 10901106

Contents lists available at SciVerse ScienceDirect

Journal of Mathematical Analysis and

Applications
www.elsevier.com/locate/jmaa

An extension of some theorems of L. Barreira and C. Valls for the

nonuniform exponential dichotomous evolution operators
Ciprian Preda , Petre Preda, Cristina Prata

West University of Timisoara,

Bd. V. Parvan, no. 4, Timisoara

300223, Romania

a r t i c l e

i n f o

a b s t r a c t

Article history:
Received 20 July 2011
Available online 11 November 2011
Submitted by P. Yao

The aim of this paper is to study the connection between the (non)uniform exponential
dichotomy of a non(uniform) exponentially bounded, strongly continuous evolution family
and the admissibility of some function spaces. Thus, there are extended recent results
established by L. Barreira and C. Valls (2010, 2011) in [10] and [11].
2011 Elsevier Inc. All rights reserved.

Keywords:
Non(uniform) exponentially bounded
Strongly continuous evolution family
Admissibility
(Non)uniform exponential dichotomy

1. Introduction
We consider the non-autonomous abstract Cauchy problem

x (t ) = A (t )x(t ),

x(s) = xs D A (s) ,

t  s, t , s J ,

(1)

on a Banach space X , where J = R or J = R+ . We assume that the above Cauchy problem is well-posed, i.e. there exists an
exponentially bounded, strongly continuous evolution family {(t , s)}t s (see Denition 1.1), such that

x() : t  (t , s)xs ,

t  s,

is differentiable for any given initial conditions x(s) = xs D ( A (s)), x(t ) D ( A (t )) and (1) holds.
Now let f be a locally integrable X -valued function on J and consider the inhomogeneous equation

x (t ) = A (t )x(t ) + f (t ),

t J.

(2)

A function u () is called a mild solution of (2) if

s
u (t ) = (t , s)u (s) +

(t , ) f ( ) d ,

t  s, t , s J .

(3)

We recall now the denition of a uniform exponentially bounded, strongly continuous evolution family.
Denition 1.1. A family of bounded linear operators acting on a Banach space X denoted by = {(t , s)}t s0 is called a
uniform exponentially bounded, strongly continuous evolution family if the following properties hold:

Corresponding author.
E-mail addresses: preda@math.uvt.ro (C. Preda), prata.elena@info.uvt.ro (C. Prata).

0022-247X/$ see front matter

doi:10.1016/j.jmaa.2011.10.062

2011 Elsevier Inc. All rights reserved.

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

(i)
(ii)
(iii)
(iv)

1091

(t , t ) = I (where I is the identity on X ), for all t  0;

(t , s, x)  (t , s)x is continuous for every t  s  0 and x X ;
(t , s) = (t , )( , s), for all t   s  0;
there exist two constants M , > 0 such that



(t , s)  Me (t s) ,

for all t  s  0.

 t +1

If { A (t )}t 0 is a family of linear and bounded operators that belongs to M 1 (i.e. supt 0 t
 A ( ) d < ), then the
above Cauchy problem (1) is well-posed (and moreover (t , s) is an invertible operator, for each t  s  0).
Some other cases when the problem (1) is well-posed, for certain unbounded operator family { A (t )}t 0 , were pointed
out by Chicone and Latushkin [12], Curtain and Prichard [13], Pazy [29], Levitan and Zhikov [21] or Tanabe [40].
The characterization of the uniform exponential dichotomy of (1) in terms of the solution of (3) has a fairly long history
and goes back to the work of O. Perron [30] in 1930. His result served as a starting point for many works on the qualitative
theory of the solutions of differential equations. Relevant results concerning the extension of Perrons problem in the more
general framework of the innite-dimensional Banach spaces (but with A (t ) bounded) were obtained by J.L. Daleckij and
M.G. Krein [14]. J.L. Massera and J.J. Schffer [22]. For certain cases of unbounded A (t ) we refer the reader to the work of
Levitan and Zhikov (see [21]).
More recently another approach uses frequently the so-called evolution semigroup T = { T (t )}t 0 on some suitable space
of X -valued functions induced by the evolution family (see [12,1820,3538,4143]). It is worth to note that most of the
results in this direction are restricted to the case where J = R. The case of the positive semiaxis has been considered in
[12] and [43].
Over the last decades it can be seen an increasing interest in the study of the asymptotic behavior of evolution equations
in abstract spaces. We refer the reader to [1,2,16,17,2325,3134,39]. Important contributions in the study of the existence
of an exponential dichotomy for evolution equations has been made and it is worth to note here few works by Nguyen
Thieu Huy (see [2528]) and Luis Barreira and Claudia Valls (see [310]).
In 2010, L. Barreira and C. Valls, in [10], used appropriate adapted norms (which can be seen as Lyapunov norms), to
show an equivalence between the admissibility of their associated L p spaces (p [1, ]) and the nonuniform exponential
stability of certain evolution families. The result is extended by the same authors in 2011 to the case of nonuniform exponential dichotomy where they also establish a collection of admissible Banach spaces for any given nonuniform exponential
contraction.
The study of evolution equations in the nonuniform setting has a strong motivation. Assume for instance that there exist
M > 0, < 0 and  0 such that



(t , s)  Me (t s)+ s ,

(4)

for every t , s  0, or more generally



(t , s)  M (s)e (t s) , for every t , s  0,
for some function M. The constant is an upper bound for the largest Lyapunov exponent, while or the function M
measure the nonuniformity of the exponential behavior. It turns out that the classical notion of (uniform) exponential behavior is very stringent for the dynamics and it is of interest to look for more general types of hyperbolic behavior. These
generalizations can be much more typical. This is precisely what happens with the notion of the so-called nonuniform exponential contraction. With this regard, the results obtained by L. Barreira and C. Valls are also a contribution to the theory
of nonuniform hyperbolicity and an important tool in the study of the stochastic behavior. We refer the reader to [3] and
[4] for details and references.
Examples of differential equations that dont posses a uniform exponential dichotomy but admit a nonuniform exponential dichotomy can be pretty easy to nd. We will recall below such an example.
Let > a > 0 and consider the differential equation in R2 given by

u  = ( at sin t )u  ,

v  = ( + at sin t ) v .

(5)

For details we refer the reader to [7, Proposition 3, p. 222].

In this paper we extend the work done by L. Barreira and C. Valls [10] to the case of nonuniform exponential dichotomy
by using the admissibility of the pair (L p ( X ), L q ( X )), with ( p , q) = (1, ).
2. Preliminaries
Let X be a Banach space and B ( X ) the Banach algebra of all linear and bounded operators acting on X (the norms on
both X and B ( X ) will be denoted by ).
Denition 2.1. An evolution family {(t , t 0 )}t t0 0 on R+ is a family of operators (t , t 0 ) B ( X ), t  t 0  0 satisfying:

(t , s)(s, t 0 ) = (t , t 0 ) and (t , t ) = I ,
for all t  s  t 0 , where I denotes the identity on X .

1092

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

It is called strongly continuous if the map (t , t 0 )  (t , t 0 )x is continuous for all x X and t  t 0  0. If (t , t 0 ) 
M (t 0 )e (t t0 ) , for some function M : R+ R+ and R and all t  t 0  0, then {(t , t 0 )}t t0 0 is (non)uniform exponentially bounded.
Following L. Barreira and C. Valls (see [10]) we introduce the following Lyapunov norms:



xt = sup e ( t ) ( , t )x,
t

x X and t R+ .

We can easily see that:

x  xt  M (t )x,

x X , t R+ .

(6)

Remark 2.1. If {(t , t 0 )}t t0 0 is an evolution family, then



(t , t 0 )x  e (t t0 ) xt ,
0
t

for every t  t 0  0 and x X .

Proof. See [10, Proposition 1, p. 2893].

As usual, we denote by L p (p  1) the space of equivalence classes of Lebesgue measurable functions f : R+ R, such
that




 f (t ) p < .
0

As known, L p endowed with the norm


1

p p


 f p =
f (t )
0

is a Banach space. We also set:





L p ( X ) = f : R+ X Bochner measurable: t   f (t )t L p .

Remark 2.2. L p ( X ) is a Banach space (p [1, ]) with the norm

 f p =  g  p ,

(7)

where g : R+ R+ , g (t ) =  f (t )t and g L . Moreover, if a sequence is convergent in L ( X ) then it is a.e. pointwise

convergent.
p

Proof. See [10, Theorem 3, p. 2893].

We list below few technical results:

Lemma 2.1. Let f : R+ R+ be a function with the property that there exist H , > 0 and > 1 such that:
(i) f (t )  H f (t 0 ), for all t [t 0 , t 0 + ], t 0  0;
(ii) f (t 0 + )  f (t 0 ), for all t 0  0.
Then there exist N , > 0 such that

f (t )  Ne (t t0 ) f (t 0 ),
Proof. See [22, 20C, p. 39].

for all t  t 0  0.

Lemma 2.2. Let f : R+ R+ be a function with the property that there exist H , > 0 and (0, 1) such that:
(i) f (t )  H f (t 0 ), for all t [t 0 , t 0 + ], t 0  0;
(ii) f (t 0 + )  f (t 0 ), for all t 0  0.

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

1093

Then there exist N , > 0 such that

f (t )  Ne (t t0 ) f (t 0 ),

for all t  t 0  0.

Proof. It follows easily by using the same argument as in the previous lemma.
q

We set now X 1 (t 0 ) = {x X: (, t 0 )x L [t ,) ( X )}. Throughout this paper, we will assume that X 1 (t 0 ) is closed and
0
q
q
q
that there exists a closed subspace X 2 (t 0 ) such that X = X 1 (t 0 ) X 2 (t 0 ), for all t 0  0.
q
q
q
q
Also, we will denote by { P 1 (t 0 )}t0 0 and { P 2 (t 0 )}t0 0 the corresponding families of projectors, i.e. Im P 1 (t 0 ) = X 1 (t 0 ),
q
q
q
q
q
2
Im P 2 (t 0 ) = X 2 (t 0 ). It is known that P i (t 0 ) B ( X ) and ( P i (t 0 )) = P i (t 0 ), for each t 0  0, i = 1, 2.
We will also put X 1 (t 0 ) for X 1 (t 0 ). As previously, we will assume that X 1 (t 0 ) is closed and that there exists X 2 (t 0 ) such
that X 2 (t 0 ) = X  X 1 (t 0 ). The associated families of projectors will be denoted by { P 1 (t 0 )}t0 0 and { P 2 (t 0 )}t0 0 .
If X = X 1 X 2 , we denote by

[ X1, X2] =



 x1
x2 
,



xi X i {0} x1 
x2  
inf

i = 1, 2.

Remark 2.3. By [22] we have that

Pi

 [ X1, X2] 

Pi

i = 1, 2,

where P 1 and P 2 are the projectors associated to the decomposition X = X 1 X 2 .

Denition 2.2. Let {(t , t 0 )}t t0 0 be a (non)uniform exponentially bounded, strongly continuous evolution family. We say
that {(t , t 0 )}t t0 0 admits a (non)uniform exponential dichotomy if there exist N 1 , N 2 , > 0 such that:

(t , t 0 )xt  N 1 e (t t0 ) xt0 , for all x X 1 (t 0 ) and t  t 0  0;

(t , t 0 )xt  N 2 e (t t0 ) xt0 , for all x X 2 (t 0 ) and t  t 0  0.
Remark 2.4. Let {(t , t 0 )}t t0 0 be a (non)uniform exponentially bounded, strongly continuous evolution family. If
{(t , t 0 )}t t0 0 admits a (non)uniform exponential dichotomy then there exist constants N 1 , N 2 , > 0 such that:

(t , t 0 )x  N 1 e (t t0 ) M (t 0 )x, for all x X 1 (t 0 ) and t  t 0  0;

M (t )(t , t 0 )x  N 2 e (t t0 ) x, for all x X 2 (t 0 ) and t  t 0  0.
Denition 2.3. Let {(t , t 0 )}t t0 0 be a (non)uniform exponentially bounded, strongly continuous evolution family. We say
that {(t , t 0 )}t t0 0 admits a uniform exponential dichotomy if there exist N 1 , N 2 , > 0 such that:

(t , t 0 )x  N 1 e (t t0 ) x, for all x X 1 (t 0 ) and t  t 0  0;

(t , t 0 )x  N 2 e (t t0 ) x, for all x X 2 (t 0 ) and t  t 0  0.
Remark 2.5. Let {(t , t 0 )}t t0 0 be a uniform exponentially bounded (i.e. supt0 0 M (t 0 ) < ), strongly continuous evolution family. If {(t , t 0 )}t t0 0 admits a nonuniform exponential dichotomy, then {(t , t 0 )}t t0 0 also admits a uniform
exponential dichotomy.
3. Results
Throughout this section we assume that p , q [1, ].
Denition 3.1. Let {(t , t 0 )}t t0 0 be a (non)uniform exponentially bounded, strongly continuous evolution family. The pair
(L p ( X ), L q ( X )) is said to be admissible to {(t , t 0 )}t t0 0 if for each f L p ( X ), there exists x X such that

t
x f () : R+ X ,

x f (t ) = (t , 0)x +

(t , ) f ( ) d
0

belongs to L q ( X ).
q

In what follows we only assume that X 1 (0) is closed and that there exists X 2 (0) such that X = X 1 (0) X 2 (0).

1094

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

q

Remark 3.1. If x X 2 (0), x = 0, then (t , 0)x = 0, for each t  0.

Proof. Assume for a contradiction that there exists t 0 > 0 such that (t 0 , 0)x = 0. Then

(t , 0)x = (t , t 0 )(t 0 , 0)x = 0,

It follows that x
q

q
X 1 (0),

for all t > t 0 .

q

but we assumed that x X 2 (0). Then we have that

x X 1 (0) X 2 (0) = {0},

and from here the contradiction.

Proposition 3.1. Let {(t , t 0 )}t t0 0 be a (non)uniform exponentially bounded, strongly continuous evolution family. If the pair
q
(L p ( X ), L q ( X )) is admissible to {(t , t 0 )}t t0 0 then for every f L p ( X ) there is a unique x X 2 (0) such that

t
x f () : R+ X ,

x f (t ) = (t , 0)x +

(t , ) f ( ) d
0

belongs to L q ( X ).
Proof. Let f L p ( X ). We have that there exists x X such that

t
x f () : R+ X ,

x f (t ) = (t , 0)x +

(t , ) f ( ) d
0

belongs to L q ( X ).
q
If we denote y (t ) = (t , 0) P 1 (0)x, it implies that y () L q ( X ) and also x f (t ) y (t ) L q ( X ), for all t  0.
Therefore
q

t

x f (t ) y (t ) = z(t ) = (t , 0) P 2 (0)x +

(t , ) f ( ) d ,
0

q
P 2 (0)x

which implies that z L ( X ) and z(0) =

X 2 (0).
q
For the uniqueness part, we will suppose that there exist u , v X 2 (0) such that
q

t
x1 f (t ) = (t , 0)u +

(t , ) f ( ) d ,

x1 f () L q ( X )

(t , ) f ( ) d ,

x2 f () L q ( X ).

and

t
x2 f (t ) = (t , 0) v +
0
q

Let w (t ) = x1 f (t ) x2 f (t ). It results that w L q ( X ) and w (t ) = (t , 0) w (0). So, w (0) X 1 (0). But w (0) = u v X 2 (0),
therefore w (0) = 0.
Thus we have that w (t ) = 0, for all t  0, and furthermore x1 f (t ) = x2 f (t ), for all t  0. 2
Theorem 3.1. Let {(t , t 0 )}t t0 0 be a (non)uniform exponentially bounded, strongly continuous evolution family. If the pair
(L p ( X ), L q ( X )) is admissible to {(t , t 0 )}t t0 0 then there is K > 0 such that



x f q  K  f t and x f (0)  K  f p ,
q

for all f L p ( X ), with x f (0) X 2 (0).

q

Proof. Let U : L p ( X ) X 2 (0) L q ( X ), dened by U f = (x f (0), x f ()).

It is obvious that U is a linear operator. We will show that U is also closed. Let { f n }nN be a sequence with f n L p ( X ),
f L p ( X ), g L q ( X ) such that

fn f

in L p ( X )

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

1095

and
q

U f n ( y , g ) in X 2 (0) L q ( X ).
Taking into account that

 t

t
t





 






(t , ) f n ( ) f ( ) t d  e (t )  f n ( ) f ( ) d
 (t , ) f n ( ) f ( ) d  


t

t
 e t



1
 f n ( ) f ( ) d  e t t 1 p  f n f  p 0,

for n we obtain that:

t

t
(t , ) f n ( ) d

(t , ) f ( ) d .
0

It results that

t
x f n (t ) (t , 0) y +

(t , ) f ( ) d ,
0

for n , for all t  0.

Since x f n () g, in L q ( X ), we have that there exists a subsequence (x f n ) (x f n ) such that x f n () g a.e.
k

Therefore U f = ( y , g ) in X 2 (0) L q ( X ). Thus U is a closed linear operator and by the Closed Graph Theorem it is also
bounded. It follows that there exists K > 0 such that

U f   K  f p ,
which implies



x f (0)  K  f 

and

for all f L p ( X ).

x f q  K  f p ,

2
q

We will denote by x f the function from L q ( X ) with x f (0) X 2 (0).

Lemma 3.1. Let {(t , t 0 )}t t0 0 be a (non)uniform exponentially bounded, strongly continuous evolution family. If the pair
(L p ( X ), L q ( X )) is admissible to {(t , t 0 )}t t0 0 , then there exists L > 0 such that



x f (t )  e ( K L + 1) f  ,
p
t

for all t  0.

Proof. Let t  1 and s [t 1, t ]. We have that

t
x f (t ) = (t , 0)x +

s
(t , ) f ( ) d = (t , s)(s, 0)x +

t
(t , s)(s, ) f ( ) d +

(t , ) f ( ) d
s

t
= (t , s)x f (s) +

(t , ) f ( ) d .
s

Furthermore, we have that



t









x f (t )  e x f (s) +  f ( ) d  e x f (s) +  f  p ,
t
s

s
s

for all s [t 1, t ].
By integrating the above inequality on the interval [t 1, t ] and by using Theorem 3.1 we get that:

 t

x f t  e
t 1





x f (s) ds +  f   e x f q +  f   e ( K + 1) f  .
p
p
p
s

1096

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

Let now t [0, 1). We have that



x f t  (t , 0)xt +

t



t




(t , ) f ( ) d  e x0 +  f ( ) d
t





 e M (0)x +  f p  e K M (0) + 1  f p ,
for all t [0, 1).
Setting L = max{1, M (0)} > 0 we get that



x f (t )  e ( K L + 1) f  ,
p
t

for all t  0 and we complete the proof.

Now we can state the main result of this paper:

Theorem 3.2. Let {(t , t 0 )}t t0 0 be a (non)uniform exponentially bounded, strongly continuous evolution family. If the pair
q
(L p ( X ), L q ( X )) is admissible to {(t , t 0 )}t t0 0 , ( p , q) = (1, ) and X 1 (t 0 ) = {x X: (, t 0 )x L q ( X )}, then:

X 1 (t 0 ) is a complemented subspace for all t 0  0;

q
q
q
X 2 (t 0 ) = (t 0 , 0) X 2 (0) is a complement of X 1 (t 0 ), for all t 0  0;
is exponentially dichotomic;
q
X 1 (t 0 ) = X 1 (t 0 ), for all t 0  0;
q
q
the operator (t , t 0 ) : X 2 (t 0 ) X 2 (t ) is invertible;
q
q
q
the family of projectors { P 1 (t 0 )}t0 0 associated to the decomposition X = X 1 (t 0 ) X 2 (t 0 ) has the following property:

t   P 1 (t )t : R+ R+
q

is bounded.

Proof. Let x X 2 (0) {0} and

f (t ) = [t0 ,t0 +1] (t )

(t , 0)x
,
(t , 0)xt

where [a,b] denotes the characteristic function (indicator) of some interval [a, b].
Since  f (t )t = [t0 ,t0 +1] (t ), for all t  0, it follows that f L p ( X ) and  f p = 1. We denote by


y (t ) =

[t0 ,t0 +1] ( )

t

( , 0)x


=

[t0 ,t0 +1] ( )

0

t

( , 0)x

 
= (t , 0) [t0 ,t0 +1] ( )
0

(t , 0)x +

[t0 ,t0 +1] ( )

0

( , 0)x

for all t  t 0 + 1. It results y L ( X ), y (0)

By Lemma 3.1 we get that
q

(t , 0)x

q
X 2 (0)

(t , )( , 0)x
d
( , 0)x

t

x +

(t , ) f ( ) d = 0,
0

and furthermore y = x f .



 y (t )  e ( K L + 1) f  = e ( K L + 1),
p
t

for all t  0. Therefore

t
0 +1

t0

( , 0)x



(t , 0)xt  e ( K L + 1),

for all 0  t  t 0 .

[t 0 , t 0 + 1] we get that






( , 0)x = ( , t 0 )(t 0 , 0)x  e (t 0 , 0)x ,

Taking into account that for

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

1097

and thus
t
0 +1

e (t 0 , 0)xt0

( , 0)x

t0

Therefore





(t , 0)x  e 2 ( K L + 1)(t 0 , 0)x ,
t
t

for all t  t 0 ,

which implies that



(t , 0)x 
t

1
e 2 ( K L

+ 1)



(t 0 , 0)x ,
t

for all t  t 0  0.

(8)

Let now > 0, t 0  0, x X 2 (0) {0}. Setting

g (t ) = [t0 ,t0 +] (t )

(t , 0)x
,
(t 0 + , 0)xt0 +
1

it follows that  g (t )t  e 2 ( K L + 1)[t0 ,t0 +] (t ), which implies that g L p ( X ) and  g p  p e 2 ( K L + 1).
We set


z(t ) =

[t0 ,t0 +] ( )
t

(t , 0)x

(t 0 + , 0)xt0 +

 
= (t , 0) [t0 ,t0 +] ( )
0

(t 0 + , 0)xt0 +

t

x +

(t , ) f ( ) d
0

0,
t  t 0 + ,

t ,0)x
(t 0 + t ) (t0(
,
t
[t 0 , t 0 + ),
+,0)xt0 +
=

(t ,0)x

t < t0 .
(t 0 +,0)xt + ,
0

It follows that z L ( X ) and z(0) X 2 (0). Using Lemma 3.1 we obtain that
q



1
 z(t )  e ( K L + 1) g   e 3 ( K L + 1)2 p ,
p
t

for all t  0.

Thus, we get that



 z(t ) =
t

1
(t , 0)xt
 e 3 ( K L + 1)2 p ,
(t 0 + , 0)xt0 +

for all 0  t  t 0 .

Putting t = t 0 in the above equality we obtain that

1
(t , 0)xt
 e 3 ( K L + 1)2 p ,
(t + , 0)xt +

for all t  0.

We have now two cases:

1. If p > 1, it results that



(t + , 0)x

t +

1
e 3 ( K L

Taking into account that lim



(t + 0 , 0)x

t +0

+ 1)2



1
1 p (t , 0)xt .
1

1
1
p
e 3 ( K L +1)2



 2(t , 0)xt ,

= , it follows that there exists 0 > 0 such that

for all t  0.

Denoting u (t ) = (t , 0)xt , we obtain by the above inequality that u (t + 0 )  2u (t ), for all t  0. Using now (1) we get
that u (s)  Hu (t ), for all s [t , t + 0 ] with H = e2 ( K1L +1) .
By Lemma 2.1 it follows now that there exist N 2 , 2 > 0 such that

u (t )  N 2 e 2 (t t0 ) u (t 0 ),

for all t  t 0  0,

which is equivalent with





(t , 0)x  N 2 e 2 (t t0 ) (t 0 , 0)x ,
t
t
0

for all t  t 0  0.

1098

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

2. If p = 1 we have that:

2
2

(t 0 , 0)xt0
=
(t 0 + , 0)xt0 +

t0 +

(t 0 + s)
t0

(t 0 , 0)xt0
ds
(t 0 + , 0)xt0 +

t0 +

e

( K L + 1)

(t 0 + s)
t0

(s, 0)xs
1 1
ds  e 2 ( K L + 1) zq q
(t 0 + , 0)xt0 +

1
1

2
 e 2 ( K L + 1) 1 q K  g p  e 2 ( K L + 1) 2 q K .

Therefore

2
2

1

2
(t 0 , 0)xt0
 e 2 ( K L + 1) K 2 q ,
(t 0 + , 0)xt0 +

or equivalently



(t 0 + , 0)x

t 0 +

1
2e 4 ( K L

+ 1)2 K


1
q (t 0 , 0)xt ,
0

for all t 0  0.
Since q < we have that

lim

2e 4 ( K L

+ 1)2 K

q = ,

which implies that there exists 0 > 0 such that



(t 0 + 0 , 0)x

t 0 +0



 2(t 0 , 0)xt ,

for all t 0  0.

As in the previous case, we use now Lemma 2.1 to get that there exist N 2 , 2 > 0 such that





(t , 0)x  N 2 e 2 (t t0 ) (t 0 , 0)x ,
t
t

for all t  t 0  0.

We will show now that X 2 (t 0 ) is a closed subspace. Taking z X 2 (t 0 ) we have that there exists xn X 2 (t 0 ) such that
xn z.
q
Taking into account that xn = (t 0 , 0)un (with un X 2 (0)) we have that



xn xm  = (t 0 , 0)(un um ) 


1
M (t 0 )

1
M (t 0 )

N 2 e 2 t0 un um 0 



(t 0 , 0)(un um ) 
t
0

1
M (t 0 )

1
M (t 0 )



(t 0 , 0)(un um )
t

N 2 e 2 t0 un um .

Since (xn )n is a Cauchy sequence in X 2 (t 0 ) we have that (un )n is also a Cauchy sequence in X 2 (0). Therefore un u
q
which implies that (t 0 , 0)un (t 0 , 0)u X 2 (t 0 ).
q
q
Thus we obtain that z = (t 0 , 0)u X 2 (t 0 ), which shows that X 2 (t 0 ) is a closed subspace.
q
q
Let now w X 2 (t 0 ). Then there exists x X 2 (0) such that w = (t 0 , 0)x and therefore

q
X 2 (0),







(t , t 0 ) w  = (t , 0)x  N 2 e 2 (t t0 ) (t 0 , 0)x = N 2 e 2 (t t0 )  w t ,
0
t
t
t
0

for all t  t 0  0 and w X 2 (t 0 ).

Thus, we get that



(t , t 0 ) w   N 2 e 2 (t t0 )  w t ,
0
t

(8 )

for all t  t 0  0 and w X 2 (t 0 ).

Let t 0  0, x X 1 (t 0 ) and h(t ) = [t0 ,t0 +1] (t )(t , t 0 )x. Then





h(t ) = [t ,t +1] (t )(t , t 0 )x  e [t ,t +1] (t )xt ,
0 0
0 0
0
t
t

which shows that h L p ( X ) and hp  e xt0 .

We denote by

t
u (t ) =

t
(t , )h( ) d =

(t , t 0 )x,

t  t 0 + 1,

0,

t < t0 .

[t0 ,t0 +1] ( ) d (t , t 0 )x = (t t 0 )(t , t 0 ), t [t 0 , t 0 + 1),

0

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

1099

Then u L q ( X ) and u (0) = 0 X 2 (0).

By Lemma 3.1 we have now that



u (t )  e ( K L + 1)h  e 2 ( K L + 1)xt ,
p
0
t

for all t  t 0  0.
If t  t 0 + 1, then by the above inequality we have that



(t , t 0 )x  e 2 ( K L + 1)xt .
0
t

If t [t 0 , t 0 + 1) it results that



(t , t 0 )x  e xt .
0
t

Setting L  = e 2 ( K L + 1) > 0 we have that



(t , t 0 )x  L  xt ,
0
t
q

for all t  t 0  0 and x X 1 (t 0 ).

q
We take now t 0  0, > 0, x X 1 (t 0 ) and we set

v (t ) = [t0 ,t0 +] (t )(t , t 0 )x.

It follows that





 v (t ) = [t ,t +] (t )(t , t 0 )x  [t ,t +] (t ) L  xt .
0 0
0 0
0
t
t
1

Thus v L p ( X ),  v p  p L  xt0 and

t
w (t ) =

t
(t , ) v ( ) d =

(t , t 0 )x,

t  t 0 + ,

0,

t < t0 .

[t0 ,t0 +] ( ) (t , t 0 )x = (t t 0 )(t , t 0 )x, t [t 0 , t 0 + ),

0
q

Then we get that w L q ( X ) and w (0) = 0 X 2 (0). By Lemma 3.1 it follows that



1
 w (t )  e ( K L + 1) v   e ( K L + 1) L  p xt ,
p
0
t

for all t  t 0  0.
If t = t 0 + we get that



(t 0 + , t 0 )xt

0 +

 e ( K L + 1) L  p xt0 ,

for all t 0  0 and x X 1 (t 0 ), or equivalently



(t 0 + , t 0 )x

t 0 +

 e ( K L + 1) L  p

xt0 ,

for all t 0  0 and x X 1 (t 0 ).

We have again two cases:
1. If p > 1 we have that
1
1
lim e ( K L + 1) L  p
= 0,

and therefore there exists 0 > 0 such that



(t 0 + 0 , t 0 )x

t 0 +0

 xt0 ,
2

for all t 0  0 and x X 1 (t 0 ).

Thus



(t + 0 , t )x

t +0

 xt ,
2

for all t  0 and x X 1 (t 0 ).

()

Let now w 1 (t ) = (t , t 0 )xt , with t  t 0 . By using () we obtain that

w 1 (t + 0 ) = (t + 0 , t 0 )xt + =
0

and

1
2


(t , t 0 )x = 1 w 1 (t ),
t

w 1 (s) = (s, t 0 )xs  L  (t , t 0 )xt = L  w 1 (t ),

for all t  0

for all s  0 and t [s, s + 0 ].

1100

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

By Lemma 2.2 we get that there exist N 1 , 1 > 0 such that

w 1 (t )  N 1 e 1 (t t0 ) w 1 (t 0 ),

for all t  t 0  0,

or equivalently



(t , t 0 )x  N 1 e 1 (t t0 ) xt ,
0
t

for all t  t 0  0 and x X 1 (t 0 ).

(9)

2. If p = 1 we get that


2 
(t 0 + , t 0 )x
t
2

t0 +
0



(s t 0 )(t 0 + , t 0 )xt

=
+

ds  L 
+

t0

= L

t0 +



(s t 0 )(s, t 0 )xs ds

t0
t0 +



1
1
1
 
 w (s) ds  L  1 q  w q  L  1 q K  v   L  2 2 q xt .
p
0
s

t0

Therefore,



(t 0 + , t 0 )x

t 0 +

 2 1
 2 L  K q xt0 ,

for all t 0  0.

Since q < and

 2

lim 2 L 

1q

= 0,

we get that there exists 0 > 0 such that



(t 0 + 0 , t 0 )x

t 0 +0

 xt0 ,

for all t 0  0 and x X 1 (t 0 ).

By using again Lemma 2.2, we obtain that there exist N 1 , 1 > 0 such that



(t , t 0 )x  N 1 e 1 (t t0 ) xt ,
0
t

(9 )

for all t  t 0  0 and x X 1 (t 0 ).

By using (8 ), (9) and (9 ) we obtain that the (nonuniform) evolution family {(t , t 0 )}t t0 0 is exponentially dichotomic.
q

We can easily see that X 1 (t 0 ) is a closed subspace. Indeed, if we take w X 1 (t 0 ), we get that there exists a sequence
q
xn X 1 (t 0 ) such that xn w. Thus



(t , t 0 )xn   N 1 e 1 (t t0 ) xn t ,
0
t

for all t  t 0  0 and n N.

Passing to the limit (i.e. n ) in the inequality above, it results that



(t , t 0 ) w   N 1 e 1 (t t0 )  w t ,
0
t


for all t  t 0  0.

q
(t , t 0 )xt dt

< , i.e. (, t 0 ) L q ( X ) which implies that w X 1 (t 0 ).

q
q
q
We will show now that X = X 1 (t 0 ) X 2 (t 0 ), for all t 0  0, where X 2 (t 0 ) = (t 0 , 0) X 2 (0).
Taking x X , t 0  0, and setting
Therefore

t0

f 1 (t ) = [t0 ,t0 +1] (t )(t , t 0 ) z

we have that  f 1 (t )t  e  zt0 [t0 ,t0 +1] (t ), which shows that f 1 L p ( X ).
q
Therefore there exists a unique x X 2 (0) such that

t
x f (t ) = (t , 0)x +

(t , ) f 1 ( ) d ,

for all t  t 0  0.

t0

From here we get that

x f (t ) = (t , t 0 ) (t 0 , 0)x z L q ( X ),
q

for all t  t 0 + 1 and thus (t 0 , 0)x z X 1 (t 0 ).

Since
q

z = z (t 0 , 0)x + (t 0 , 0)x X 1 (t 0 ) + X 2 (t 0 ),
we have that
q

X = X 1 (t 0 ) + X 2 (t 0 ),

for all t 0  0.

for all t 0  0,

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

q

1101

We take now w X 1 (t 0 ) X 2 (t 0 ). It results that there exists u X 2 (0) such that w = (t 0 , 0)u. From here we get that
q
q
(t , t 0 ) w = (t , 0)u, for all t  t 0  0, which shows that u X 1 (0) X 2 (0) = {0}.
q
q
q
q
Therefore, it results that X 1 (t 0 ) X 2 (t 0 ) = {0} and we now have that X = X 1 (t 0 ) X 2 (t 0 ), for all t 0  0.
q

We denote now X 1 (t 0 ) = X 1 (t 0 ) and we will show in the next that X 1 (t 0 ) = X 1 (t 0 ).

q
Taking x X 1 (t 0 ) we get that



(t , t 0 )x  N 1 e 1 (t t0 ) xt ,
0
t

for all t  t 0  0.

Thus we have that x X 1 (t 0 ), which shows that

q

X 1 (t 0 ) X 1 (t 0 ).

(10)

Choosing randomly x X 1 (t 0 ), x = x1 + x2 , where x1

q
X 1 (t 0 )

and x2

q
X 2 (t 0 ),

x2 = 0 we have that



 


(t , t 0 )x  (t , t 0 )x2  (t , t 0 )x1   N 2 e 2 (t t0 ) x2 t N 1 e 1 (t t0 ) x1 t ,
0
0
t
t
t

as t .

Therefore x
/ X 1 (t 0 ), and from here the contradiction.
q
Then x2 = 0 and x = x1 X 1 (t 0 ), which shows that
q

X 1 (t 0 ) X 1 (t 0 ).

(11)
q
X 2 (t 0 )

q
X 2 (t )

We will show below that (t , t 0 ) :

is invertible.
q
For this reason, we take x X 2 (t ), t  t 0  0 and we set f (s) = [t ,t +1] (s)(s, t )x.
Therefore  f (s)s  e xt [t ,t +1] (s), which implies that f L p ( X ).
q
It results that there exists a unique y X 2 (0) such that

s
x f (s) = (s, 0) y +

(s, ) f ( ) d = (s, 0) y (s, t )x,

t

for all s  t + 1, x f () L q ( X ). Thus

x f (s) = (s, t ) (t , 0) y x = (s, t ) (t , t 0 )(t 0 , 0) y x ,

for all s  t + 1. Thus we get that
q

(t , t 0 )(t 0 , 0) y x X 1 (t ) X 2 (t ) = {0}.
q

It results that (t , t 0 )(t 0 , 0) y = x and therefore (t , t 0 ) : X 2 (t 0 ) X 2 (t ) is invertible.

The only thing that we have to prove now is that

t   P 1 (t )t : R+ R+
q

Taking x1

q
X 1 (t 0 ),

is bounded.
q

with x1 t0 = 1, x2 X 2 (t 0 ), with x2 t0 = 1 and x = x1 + x2 , we have that



(t , t 0 )x  e (t t0 ) xt ,
0
t

for all t  t 0  0.
It follows that





xt0  e (t t0 ) (t , t 0 )xt = e (t t0 ) (t , t 0 )x1 + (t , t 0 )x2 t
 




 e (t t0 ) (t , t 0 )x2  (t , t 0 )x1   e (t t0 ) N 2 e 2 (t t0 ) N 1 e 1 (t t0 ) .
t

Setting t t 0 = in the above inequality we get that



xt0  e N 2 e 2 N 1 e 1 .
Since N 2 e 2 N 1 e 1 , while , we have that there exists 0 > 0 such that xt0  2e 0 , for all t 0  0.
If we denote 2e 0 = 0 > 0 we get that xt0  0 , for all t 0  0. It follows that [ X 1 (t 0 ), X 2 (t 0 )]  0 and furthermore

0 

for all t 0  0,

 q

 P (t 0 )  2 ,
1
t

for all t 0  0.

q
 P 1 (t 0 )t0

or equivalently

Using a similar argument we can show that  P 2 (t 0 )t0  2 , for all t 0  0.

0

1102

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

Remark 3.2. Let {(t , t 0 )}t t0 0 be a uniform exponentially bounded, strongly continuous evolution family. Then





L p ( X ) = f : R+ X: Lebesgue-measurable,  f (t ) =  f (t ),  f  L p .

Proof. Taking into account that



 


 f (t )   f (t )  M  f (t ),
t

for all t R+

we obtain that g (t ) =  f (t )t , g : R+ R+ ,

g Lp

or equivalently

 f  Lp,

1 p

and

 f  p   f p  M  f  p ,

for all f L p ( X ).

Denition 3.2. Let {(t , t 0 )}t t0 0 be a (non)uniform exponentially bounded, strongly continuous evolution family. We say
that {(t , t 0 )}t t0 0 is hyperbolic if there exists a projector-valued function P : R+ B ( X ) such that the function t  P (t )x
is continuous and bounded for each x X and there exist N 1 , N 2  1 and > 0 such that

P (t )(t , t 0 ) = (t , t 0 ) P (t );
(t , t 0 ) is invertible as an operator from Im( I P (t 0 )) to Im( I P (t ));
(t , t 0 ) P (t 0 )x  N 1 e (t t0 )  P (t 0 )x, for all t  t 0  0 and x X ;
(t , t 0 )( I P (t 0 ))x  N 2 e (t t0 ) ( I P (t 0 ))x, for all t  t 0  0 and x X .

Theorem 3.3. Let {(t , t 0 )}t t0 0 be a (non)uniform exponentially bounded, strongly continuous evolution family. Then
q
{(t , t 0 )}t t0 0 is hyperbolic if and only if X 1 (0) is closed and admits a direct complement, and the pair (L p ( X ), L q ( X )) is admissible to the process with ( p , q) = (1, ), 1  p  q  .
Proof. Necessity. Let f L p ( X ) and

t
y (t ) =


(t , ) P ( ) f ( )

1 ( , t ) Q ( ) f ( ) d ,

where Q (t ) = I P (t ). Then we have that


y (0) =

1 ( , 0) Q ( ) f ( ) d

and

t
(t , 0) y (0) =

(t , 0)


( , 0) Q ( ) f ( )
t

t
=


(t , ) Q ( ) f ( ) d

t

(t , ) Q ( ) f ( ) d


 1
(t , 0) ( , t )(t , 0)
Q ( ) f ( ) d

1 ( , t ) Q ( ) f ( ) d

t
=

t
(t , ) f ( ) d +


(t , ) P ( ) f ( ) d

Therefore

t
(t , 0) y (0) +

(t , 0) 1 ( , 0) Q ( ) f ( ) d

(t , ) f ( ) d = y (t ) = x f (t ).
0

Taking into account that

1 ( , t ) Q ( ) f ( ) d .

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106



 y (t )  N 1

t



e (t )  P ( ) f ( ) d +

N2



 N 1 sup P (t )
t 0

t

e (t )  f ( ) d +



1103

e ( t )  Q ( ) f ( ) d

sup Q (t )

N 2 t 0

e ( t )  f ( ) d ,

it follows by [15, Theorem 6.4, p. 477] that x f belongs to L ( X ), with p  q.

Suciency. From Theorem 3.2 it results that the evolution family {(t , t 0 )}t t0 0 is exponentially dichotomic and the
family of projectors { P (t )}t 0 is bounded. The continuity of the function t  P (t )x follows easily by using a similar argument with the one from [43], Lemma 4.2(b), (c). Therefore {(t , t 0 )}t t0 0 is hyperbolic. 2
q

We analyze in the next the case ( p , q) = (1, ).

Proposition 3.2. Let {(t , t 0 )}t t0 0 be a (non)uniform exponentially bounded, strongly continuous evolution family. If the pair
( L 1 ( X ), L ( X )) is admissible to {(t , t 0 )}t t0 0 , X 1 (0) is complemented, then X 1 (t 0 ) = X 1 (t 0 ) is closed and admits a direct complement X 2 (t 0 ) = (t 0 , 0) X 2 (0), for all t 0  0. Moreover (t , t 0 ) : X 2 (t 0 ) X 2 (t ) is invertible and there exist constants N 1 , N 2 > 0
such that



(t , t 0 )x  N 1 xt ,
0
t


(t , t 0 )x  N 2 xt ,
0
t

for all x X 1 (t 0 ) and t  t 0  0;

for all x X 2 (t 0 ) and t  t 0  0.

Proof. Let x X 2 (0) {0} and t 0  0. Setting

f (t ) = [t0 ,t0 +1] (t )

(t , 0)x
(t , 0)xt

it results that f L 1 ( X ) and  f 1 = 1.

We consider now


y (t ) =

[t0 ,t0 +1] ( )

t

(t , 0)x

( , 0)x

 
= (t , 0) [t0 ,t0 +1] ( )
0

( , 0)x

t

x +

(t , ) f ( ) d = 0,
0

for all t  t 0 + 1.
Thus y L ( X ) and by Lemma 3.1 we obtain that



 y (t )  e ( K L + 1) f  = e ( K L + 1),
1
t

for all t  0.

If t  t 0 we have that



 y (t ) =

t
0 +1

t0

( , 0)x



(t , 0)xt  e ( K L + 1).

Taking into account that







( , 0)x = ( , t 0 )(t 0 , 0)x  e (t 0 , 0)x ,

t
0

it follows that

t
0 +1

e (t 0 , 0)xt


0

t0

( , 0)x

Therefore

1
e (t 0 , 0)xt0



(t , 0)xt  e ( K L + 1),

1104

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

or equivalently,





1
(t 0 , 0)x 
(t 0 , 0)x ,
t0
t0

e ( K L + 1)
Setting N 2 =

1
e ( K L +1)

for all t  t 0 .

> 0 in the above inequality, we obtain that





(t , 0)x  N 2 (t 0 , 0)x ,
t
t
0

for all t  t 0  0.

We show now that X 2 (t 0 ) = (t 0 , 0) X 2 (0) is a closed subspace.

Taking x X 2 (t 0 ) we have that there exists a sequence (xn ) X 2 (0) such that (t 0 , 0)xn x.
It follows that





(t 0 , 0)xn (t 0 , 0)xm  = (t 0 , 0)(xm xn )  N 2 xn xm 0  N 2 M (0)xn xm  0.
t
t
0

Since X 2 (0) is a Banach space we obtain that xn u X 2 (0), which implies that

(t 0 , 0)xn (t 0 , 0)u = x X 2 (t 0 ),
and therefore x X 2 (t 0 ).
Thus, X 2 (t 0 ) = X 2 (t 0 ).
Let now z X 2 (t 0 ). Then there exists u X 2 (0) so that z = (t 0 , 0)u. We have that







(t , t 0 ) z = (t , t 0 )(t 0 , 0)u  = (t , 0)u   N 2 zt ,
0
t
t
t

for all t  t 0  0.

It follows that



(t , t 0 ) z  N 2 zt ,
0
t

for all t  t 0  0 and z X 2 (t 0 ).

Let now t 0  0, x X 1 (t 0 ) and g (t ) = [t0 ,t0 +1] (t )(t , t 0 )x.

Since



 g (t )  e xt [t ,t +1] (t )
0
0 0
t

it follows that g L 1 ( X ) and  g 1  e xt0 .

Set now

t
z(t ) =

t
(t , ) g ( ) d =

[t0 ,t0 +1] ( ) d (t , t 0 )x, for all t > t 0 .

t0

If t  t 0 + 1 we have that z(t ) = (t , t 0 )x, which implies that z L ( X ) and z(0) = 0 X 2 (0).
By Lemma 3.1 it results that



 z(t )  e ( K L + 1) g   e 2 ( K L + 1)xt ,
0
1
t

for all t  t 0 + 1,

or equivalently



(t , t 0 )x  e 2 ( K L + 1)xt ,
0
t

for all t  t 0 + 1.

If t [t 0 , t 0 + 1) it follows that



(t , t 0 )x  e xt .
0
t

Therefore



(t , t 0 )x  N 1 xt ,
0
t

for all t  t 0  0 and x X 1 (t 0 ),

where N 1 = e 2 ( K L + 1) > 0.
To show that X 1 (t 0 ) is a closed subspace we choose randomly y X 1 (t 0 ) and we have that there exists a sequence
( yn ) X 1 (t 0 ), such that yn y and from here we get that (, t 0 ) yn (, t 0 ) y.
Since



(t , t 0 ) yn   N 1  yn t ,
0
t

for all t  t 0  0 and n N.

We obtain, by passing to the limit (i.e. n ), that



(t , t 0 ) y   N 1  y t ,
0
t

for all t  t 0  0.

Taking into account that (, t 0 ) y L ( X ), we get that y X 1 (t 0 ).

Therefore X 1 (t 0 ) = X 1 (t 0 ).
We show now that X = X 1 (t 0 ) X 2 (t 0 ), for all t 0  0.

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

1105

Let z X , t 0  0 and f (t ) = [t0 ,t0 +1] (t )(t , t 0 ) z. Then





 f (t ) = [t ,t +1] (t )(t , t 0 ) z  e zt [t ,t +1] (t )
0 0
0
0 0
t
t

and from here we get that f L 1 ( X ) and  f 1  e  zt0 .

Therefore there exists a unique x X 2 (0) such that

t
x f (t ) = (t , 0)x +

(t , ) f ( ) d
0

belongs to L .
Since

t
x f (t ) = (t , 0)x +

(t , ) f ( ) d ,

for all t  t 0

t0

and

x f (t ) = (t , t 0 )(t 0 , 0)x (t , t 0 ) z,
we get that

x f (t ) = (t , t 0 ) (t 0 , 0)x z

for all t  t 0 + 1,

belongs to L ( X ), for all t  t 0 + 1.

Therefore (t 0 , 0)x z X 1 (t 0 ).
Taking into account that z = z (t 0 , 0)x + (t 0 , 0)x X 1 (t 0 ) + X 2 (t 0 ) it follows that

X = X 1 (t 0 ) + X 2 (t 0 ),

for all t 0  0.

Choose now w X 1 (t 0 ) X 2 (t 0 ). Then there exists u X 2 (0) such that w = (t 0 , 0)u.

Since w X 1 (t 0 ), we have that (t , t 0 ) w = (t , 0)u, for all t  t 0 . Thus, u X 1 (0).
Therefore u X 1 (0) X 2 (0) = {0}, which implies u = 0 and thus w = 0.
We show now that (t , t 0 ) : X 2 (t 0 ) X 2 (t ) is invertible.
Setting u X 2 (t ) and v (s) = [t0 ,t0 +1] (s)(s, t )u, we have that v L 1 ( X ).
Then there exists a unique x X 2 (0) such that

s
x v (s) = (s, 0)x +

s
(s, ) v ( ) d = (s, 0)x

belongs to L ( X ), x v (0) = x.
If s  t, we have that

{[t0 ,t0 +1]} ( ) d (s, t )z

0

x v (s) = (s, t )(t , 0)x (s, t )u = (s, t ) (t , 0)x u

belongs to L ( X ).
Then(t , 0)x u X 1 (t ) X 2 (t ) = {0}, and therefore (t , t 0 )(t 0 , 0)x = u, i.e. (t , t 0 ) is surjective.
Since (t , t 0 ) is also one-to-one, we obtain that (t , t 0 ) : X 2 (t 0 ) X 2 (t ) is invertible. 2
References
[1] A. Ben-Artzi, I. Gohberg, Dichotomies of systems and invertibility of linear ordinary differential operators, Oper. Theory Adv. Appl. 56 (1992) 142.
[2] A. Ben-Artzi, I. Gohberg, M.A. Kaashoek, Invertibility and dichotomy of differential operators on the half-line, J. Dynam. Differential Equations 5 (1993)
136.
[3] L. Barreira, Ya. Pesin, Lyapunov Exponents and Smooth Ergodic Theory, Univ. Lecture Ser., vol. 23, Amer. Math. Soc., 2002.
[4] L. Barreira, Ya. Pesin, Nonuniform Hyperbolicity, Encyclopedia Math. Appl., vol. 115, Cambridge University Press, 2007.
[5] L. Barreira, J. Schmeling, Sets of non-typical points have full topological entropy and full Hausdorff dimension, Israel J. Math. 116 (2000) 2970.
[6] L. Barreira, C. Valls, Optimal estimates along stable manifolds of non-uniformly hyperbolic dynamics, Proc. Roy. Soc. Edinburgh Sect. A 138 (2008)
693717.
[7] L. Barreira, C. Valls, Nonuniform exponential dichotomies and Lyapunov regularity, J. Dynam. Differential Equations 19 (2007) 215241.
[8] L. Barreira, C. Valls, Stability of Nonautonomous Differential Equations, Lecture Notes in Math., vol. 1926, Springer, 2008.
[9] L. Barreira, C. Valls, Nonuniform exponential contractions and Lyapunov sequences, J. Differential Equations 246 (2009) 47434771.
[10] L. Barreira, C. Valls, Admissibility for nonuniform exponential contractions, J. Differential Equations 249 (2010) 28892904.
[11] L. Barreira, C. Valls, Nonuniform exponential dichotomies and admissibility, Discrete Contin. Dyn. Syst. 30 (1) (2011) 3953.
[12] C. Chicone, Y. Latushkin, Evolution Semigroups in Dynamical Systems and Differential Equations, Math. Surveys Monogr., vol. 70, American Mathematical Society, Providence, RI, 1999.
[13] R. Curtain, A.J. Prichard, Innite-Dimensional Linear Systems Theory, Lecture Notes in Control and Inform. Sci., vol. 8, Springer, New York, 1978.

1106

C. Preda et al. / J. Math. Anal. Appl. 388 (2012) 10901106

[14] J.L. Daleckij, M.G. Krein, Stability of Differential Equations in Banach Space, Amer. Math. Soc., Providence, RI, 1974.
[15] P. Hartman Ordinary, Differential Equations, Wiley, New York/London/Sidney, 1964.
[16] M.A. Kaashoek, S.M. Verduyn Lunel, An integrability condition on the resolvent for hyperbolicity of the semigroup, J. Differential Equations 112 (1994)
374406.
[17] K.J. Engel, R. Nagel, One-Parameter Semigroups for Linear Evolution Equations, Grad. Texts in Math., vol. 194, Springer, 1999.
[18] Y. Latushkin, S. Montgomery-Smith, Evolutionary semigroups and Lyapunov theorems in Banach spaces, J. Funct. Anal. 127 (1995) 173197.
[19] Y. Latushkin, T. Randolph, Dichotomy of differential equations on Banach spaces and an algebra of weighted composition operators, Integral Equations
Operator Theory 23 (1995) 472500.
[20] Y. Latushkin, T. Randolph, R. Schnaubelt, Exponential dichotomy and mild solutions of non-autonomous equations in Banach spaces, J. Dynam. Differential Equations 3 (1998) 489510.
[21] B.M. Levitan, V.V. Zhikov, Almost Periodic Functions and Differential Equations, Cambridge Univ. Press, Cambridge, 1982.
[22] J.L. Massera, J.J. Schffer, Linear Differential Equations and Function Spaces, Academic Press, New York, 1966.
[23] Nguyen Thieu Huy, Nguyen Van Minh, Exponential dichotomy of difference equations and applications to evolution equations on the half-line, Comput.
Math. Appl. 42 (2001) 301311.
[24] Nguyen Thieu Huy, Nguyen Van Minh, Characterizations of dichotomies of evolution equations on the half-line, J. Math. Anal. Appl. 261 (2001) 2844.
[25] Nguyen Thieu Huy, Exponentially dichotomous operators and exponential dichotomy of evolution operators on the half-line, Integral Equations Operator Theory 48 (2004) 497510.
[26] Nguyen Thieu-Huy, Exponential dichotomy of evolution equations and admissibility of function spaces on a half-line, J. Funct. Anal. 235 (2006) 330
354.
[27] Nguyen Thieu-Huy, Stable manifolds for semi-linear evolution equations and admissibility of function spaces on a half-line, J. Math. Anal. Appl. 354
(2009) 372386.
[28] Nguyen Thieu-Huy, Invariant manifolds of admissible classes for semi-linear evolution equations, J. Differential Equations 246 (2009) 18201844.
[29] A. Pazy, Semigroups of Linear Operators and Applications to Partial Differential Equations, Springer-Verlag, New York, 1983.
[30] O. Perron, Die Stabilittsfrage bei Differentialgleichungen, Math. Z. 32 (1930) 703728.
[31] P. Preda, M. Megan, Nonuniform dichotomy of evolutionary processes in Banach spaces, Bull. Aust. Math. Soc. 27 (1983) 3152.
[32] P. Preda, A. Pogan, C. Preda, ( L p , L q )-Admissibility and exponential dichotomy of evolutionary processes on the half-line, Integral Equations Operator
Theory 49 (3) (2004) 405418.
[33] P. Preda, A. Pogan, C. Preda, Schffer spaces and uniform exponential stability of linear skew-product semiows, J. Differential Equations 212 (2005)
191207.
[34] P. Preda, A. Pogan, C. Preda, Schffer spaces and exponential dichotomy for evolutionary processes, J. Differential Equations 230 (2006) 378391.
[35] F. Rbiger, R. Schnaubelt, The spectral mapping theorem for evolution semigroups on spaces of vector valued functions, Semigroup Forum 48 (1996)
225239.
[36] F. Rbiger, R. Schnaubelt, Absorption evolution families with applications to nonautonomous diffusion processes, Tubingen Ber. Functionalanal. 5
(1995/1996) 334335.
[37] R. Rau, Hyperbolic evolution groups and dichotomic of evolution families, J. Dynam. Differential Equations 6 (1994) 107118.
[38] R. Schnaubelt, Sucient conditions for exponential stability and dichotomy of evolution equations, Forum Math. 11 (1999) 543566.
[39] R. Schnaubelt, Asymptotically autonomous parabolic evolution equations, J. Evol. Equ. 1 (2001) 1937.
[40] H. Tanabe, Equations of Evolution, Pitman, London, 1979.
[41] N. van Minh, Semigroups and stability of nonautonomous differential systems in Banach spaces, Trans. Amer. Math. Soc. 345 (1994) 223242.
[42] N. van Minh, On the proof of characterization of the exponential dichotomy, Proc. Amer. Math. Soc. 127 (1999) 779782.
[43] N. van Minh, F. Rbiger, R. Schnaubelt, Exponential stability, exponential expansiveness and exponential dichotomy of evolution equations on the
half-line, Integral Equations Operator Theory 32 (1998) 332353.