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

net/publication/278675542

Asymptotic behavior of solutions to a class of differential variational

inequalities

Article in Annales Polonici Mathematici · June 2015

DOI: 10.4064/ap114-2-5

CITATIONS READS
34 339

2 authors:

Nguyen Thi Van Anh Ke Tran Dinh

Đại học Sư phạm Hà Nội Đại học Sư phạm Hà Nội
15 PUBLICATIONS 85 CITATIONS 58 PUBLICATIONS 672 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Ke Tran Dinh on 18 June 2015.

The user has requested enhancement of the downloaded file.

 ASYMPTOTIC BEHAVIOR OF SOLUTIONS TO A CLASS OF
DIFFERENTIAL VARIATIONAL INEQUALITIES

NGUYEN THI VAN ANH, TRAN DINH KE \

Abstract. We address some questions concerning a class of differential vari-

ational inequalities with finite delays. The existence of exponential decay so-
lutions and global attractor for the associated multi-valued semiflow is proved.

1. Introduction
We consider the following differential variational inequality (DVI)
x0 (t) = Ax(t) + h(x(t)) + B(x(t), xt )u(t), t ∈ J = [0, T ], (1.1)
hv − u(t), F (x(t)) + G(u(t))i ≥ 0, ∀v ∈ K, for a.e t ∈ J, (1.2)
x(s) = ϕ(s), s ∈ [−τ, 0], (1.3)
n m
where x(t) ∈ R , u(t) ∈ K with K being a closed convex subset in R , xt denotes
the history of the state function up to time t; A, B, F, G and h are given maps
which will be specified in the next section.
The notion of differential variational inequality can be traced back to Aubin
and Cellina [1] in 1984. In a later work of Avgerinous and Papageorgiou [3], this
concept was revisited and expanded. However, it is seen that DVIs were first
systematically studied by Pang and Stewart [14]. As mentioned in this paper,
DVIs are useful for representing models involving both dynamics and constraints
in the form of inequalities which arise in many problems in reality, for example,
mechanical impact problems, electrical circuits with ideal diodes, Coulomb friction
problems for contacting bodies, economical dynamics and related problems such as
dynamic traffic networks. In case K = Rm , system (1.1)-(1.3) becomes a differential
algebraic equation with the unknown y = (x, u), while if K is a cone, it is a
differential complementarity problem. Some existence results for DVIs can be found
in [8, 9, 12] and the references therein.
In the theory of differential equations, problem (1.1)-(1.3) is called a differential
system with unilateral constraints. As a matter of fact, it can be seen as a control
problem subject to constraints. In this work, we will look for admissible conditions
to gain the existence and behavior of solutions to this problem. Specifically, after
giving a short proof for the solvability of (1.1)-(1.3) on bounded intervals, we show
that this system has a compact set of exponentially decay solutions. In addition, the
multi-valued semiflow generated by (1.1)-(1.3) admits a compact global attractor
in C([−τ, 0]; Rn ). Up to our knowledge, no attempt has been made to study the

2010 Mathematics Subject Classification. 34D45, 47H08, 47H10.

Key words and phrases. differential variational inequality; measure of noncompactness; con-
densing map; fixed point; decay solution; m-semiflow; global attractor.
\ Corresponding author. Email: [email protected] (T.D.Ke).

1
2 N.T.V. ANH, T.D. KE

behavior of solutions to (1.1)-(1.3) in literature. Then the last two mentioned

studies are the main part of our work.
The rest of this paper is organized as follows. In the next section, we recall
the notion of measure of noncompactness (MNC) and construct a regular MNC
to determine the compactness in BC(0, ∞; Rn ). On the other hand, the theory
of global attractor for multi-valued semiflows introduced in [13] will be taken into
account. Section 3 shows the existence result for (1.1)-(1.3) on compact intervals.
In Section 4, we prove the existence of exponentially decay solutions to our problem
and Section 5 is devoted to showing that the multi-valued semiflow associated with
(1.1)-(1.3) has a compact global attractor in C([−τ, 0]; Rn ).

2. Preliminaries
2.1. Measure of noncompactness. Let E be a Banach space. Denote
P(E) = {B ⊂ E : B 6= ∅},
B(E) = {B ∈ P(E) : B is bounded}.
We recall the definition of measure of noncompactness introduced in [2].
Definition 2.1. A function β : B(E) → R+ is called a measure of noncompactness
(MNC) on E if
β(co Ω) = β(Ω) for every Ω ∈ B(E),
where co Ω is the closure of the convex hull of Ω. An MNC β is called
(i) monotone if Ω0 , Ω1 ∈ B(E), Ω0 ⊂ Ω1 implies β(Ω0 ) ⊂ β(Ω1 );
(ii) nonsingular if β({a} ∪ Ω) = β(Ω) for any a ∈ E, Ω ∈ B(E);
(iii) invariant with respect to union with compact set if β(K ∪ Ω) = β(Ω) for
every relatively compact set K ⊂ E and Ω ∈ B(E);
(iv) algebraically semi-additive if β(Ω0 + Ω1 ) ≤ β(Ω0 ) + β(Ω1 ) for any Ω0 , Ω1 ∈
B(E);
(v) regular if β(Ω) = 0 is equivalent to the relative compactness of Ω.
An important example of MNC is the Hausdorff MNC χ(·), which is defined as
follows: for Ω ∈ B(E),
χ(Ω) = inf{ > 0 : Ω has a finite  − net}.
In particular, it is known that the Hausdorff MNC on C([0, T ]; Rn ), the space of
continuous functions on [0, T ] taking values in Rn , is given by (see [2])
1
χT (D) = lim sup max kx(t) − x(s)k. (2.1)
2 δ→0 x∈D t,s∈[0,T ],|t−s|<δ
The measure χT (D) of a subset D can be seen as the modulus of equicontinuity of
a subset in C([0, T ]; Rn ).
Consider the space BC(0, ∞; Rn ) of bounded continuous functions on [0, ∞) tak-
ing value on Rn . Denote by πT the restriction operator on the space BC(0, ∞; Rn ),
that is πT (x) is the restriction of x on [0, T ]. Then
χ∞ (D) = sup χT (πT (D)), D ⊂ BC(0, ∞; Rn ), (2.2)
T >0

is an MNC. One can check that χ∞ satisfies all properties given in Definition
2.1, but regularity. Indeed, we will testify this claim by choosing the sequence
 ASYMPTOTIC BEHAVIOR FOR A CLASS OF DVIS 3

{fk } ⊂ BC(0, ∞; R) as follows


0,
 t 6∈ [k, k + 1],
fk (t) = 2t − 2k, t ∈ [k, k + 12 ],
−2t + 2k + 2, t ∈ [k + 12 , k + 1].



Then it is obvious that {πT (fk )} is compact (converging to 0 in C([0, T ]; R)) for
any T > 0. However, one sees that
sup |fk (t) − fl (t)| = 1 for k 6= l,
t≥0

and then {fk } is not a Cauchy sequence in BC(0, ∞; R). This fact tells us that
χT (πT ({fk }))) = 0 for any T > 0 and then χ∞ ({fk }) = 0, but {fk } is not relatively
compact.
We now construct a regular MNC on BC(0, ∞; Rn ). Recall the following MNCs
on BC(0, ∞; Rn ) (see [4]):
dT (D) = sup sup kx(t)k,
x∈D t≥T
d∞ (D) = lim dT (D).
T →∞

Define
χ∗ (D) = χ∞ (D) + d∞ (D). (2.3)
∗ n
By a simple check, we get that χ is an MNC on BC(0, ∞; R ).
Lemma 2.1. The MNC χ∗ defined by (2.3) is regular.
Proof. Let D ⊂ BC(0, ∞; Rn ) be a bounded set such that χ∗ (D) = 0. We show that
D is relatively compact. Let P BC(0, ∞; Rn ) be the space of piecewise continuous
and bounded functions on R+ , taking values in Rn . This is a Banach space with
the norm
kxkP BC = sup kx(t)k,
t≥0
n
and contains BC(0, ∞; R ) as a closed subspace.
For  > 0, since d∞ (D) = 0 one can choose T > 0 such that supt≥T kx(t)k <

, ∀x ∈ D. This means that
2

kx − πT (x)kP BC < , ∀x ∈ D,
2
here πT (x) agrees with a function in P BC(0, ∞; Rn ) by the following manner
(
x(t), t ∈ [0, T ],
πT (x) =
0, t > T.

Now since D is bounded and χT (D) = 0, by the Arzela-Ascoli theorem πT (D) is a

relatively compact set in C([0, T ]; Rn ), we can write
N
[ 
πT (D) ⊂ BT (xi , ) (2.4)
i=1
2
4 N.T.V. ANH, T.D. KE

where xi ∈ C([0, T ]; Rn ), i = 1, ..., N , the notation BT (x, r) stands for the ball in
C([0, T ]; Rn ) centered at x with radius r. Put
(
xi (t), t ∈ [0, T ],
x̂i (t) =
0, t > T,

then {x̂i }N n
i=1 belong to P BC(0, ∞; R ). We assert that
N
[
D⊂ B∞ (x̂i , ),
i=1

here B∞ (x, r) is the ball in P BC(0, ∞; Rn ) with center x and radius r. Indeed, let
x ∈ D then by (2.4), there is a number k ∈ {1, ..., N } such that

kπT (x) − xk kC < ,
2
here k · kC is the norm in C([0, T ]; Rn ). This implies

kπT (x) − x̂k kP BC < .
2
Then
kx − x̂k kP BC ≤ kx − πT (x)kP BC + kπT (x) − x̂k kP BC
 
≤ + = .
2 2
N
S
Thus x ∈ B∞ (x̂k , ). We have D ⊂ B∞ (x̂i , ), and hence D is relatively
i=1
compact in P BC(0, ∞; Rn ). In order to show that D is also relatively compact
in BC(0, ∞; Rn ), we observe that if {xn } ⊂ D then one can find a function
x ∈ P BC(0, ∞; Rn ) such that
lim kxn − xkP BC = lim sup kxn (t) − x(t)k = 0,
n→∞ n→∞ t≥0

up to a subsequence. This means that {xn } converges to x uniformly on R+ .

Noting that {xn } are continuous functions, we have x ∈ BC(0, ∞; Rn ). The proof
is complete. 

We also make use of some notions and facts of set-valued analysis. Let Y be a
metric space.
Definition 2.2. A multivalued map (multimap) F : Y → P(E) is said to be
(i) upper semicontinuous (u.s.c) if F −1 (V ) = {y ∈ V : F(y) ∩ V 6= ∅} is a
closed subset of Y for every closed set V ⊂ E;
(ii) weakly upper semicontinuous (weakly u.s.c) if F −1 (V ) is closed subset of Y
for all weakly closed set V ⊂ E;
(iii) closed if its graph ΓF = {(y, z) : z ∈ F(y)} is a closed subset of Y × E;
(iv) compact if F(B) is relatively compact in E for any bounded set B ⊂ Y ;
(v) quasicompact if its restriction to any compact subset A ⊂ Y is compact.
Lemma 2.2. [11, Theorem 1.1.12] Let G : Y → P(E) be a closed quasicompact
multimap with compact values. Then G is u.s.c.
 ASYMPTOTIC BEHAVIOR FOR A CLASS OF DVIS 5

Lemma 2.3. [5, Proposition 2] Let X be a Banach space and Ω be a nonempty

subset of another Banach space. Assume that G : Ω → P(X) is a multimap with
weakly compact, convex values. Then G is weakly u.s.c iff {xn } ⊂ Ω, xn → x0 and
{yn } ∈ G(xn ) implies yn * y0 ∈ G(x0 ) up to a subsequence.
In the sequel, we use the following fixed point principle, which is a special case
of [11, Corollary 3.3.1].
Theorem 2.4. Let M be a bounded convex and closed subset of a Banach space E
and let F : M → P(M) be a compact, u.s.c multimap with compact convex values.
Then Fix(F) := {x ∈ F(x)} is a nonempty compact set.
2.2. Multivalued semiflows and their attractors. We summarize some defini-
tions and results regarding the theory of global attractor of multivalued semiflows
(m-semiflows) given in [13]. Let Γ be a nontrivial subgroup of the additive group
of real numbers R and Γ+ = Γ ∩ [0, ∞).
Definition 2.3. The mapping G : Γ+ × E → P(E) is called an m-semiflow if the
following conditions are satisfied
(1) G(0, w) = {w}, for all w ∈ E.
(2) G(t1 + t2 , x) ⊂ G(t1 , G(t2 , x)), for all t1 , t2 ∈ Γ+ , x ∈ E,
where G(t, B) = ∪x∈B G(t, x), B ⊂ E.
It is called a strict m-semiflow if G(t1 + t2 , w) = G(t1 , G(t2 , w)) for all w ∈ E and
t1 , t2 ∈ Γ+ . G is said to be eventually bounded if for each bounded set B ⊂ E, there
is a number T (B) > 0 such that γT+(B) (B) is bounded. Here γT+(B) (B) is the set of
orbits after time T (B) : γT+(B) (B) =
S
G(t, B).
t≥T (B)

Definition 2.4. The set A is called a global attractor of the m-semiflow G if it

satisfies the following conditions:
(1) A is negatively semi-invariant, i.e. A ⊂ G(t, A), ∀t ∈ Γ+ ;
(2) A attracts any B ∈ B(E), i.e. dist(G(t, B), A) → 0 as t → ∞, for all
bounded set B ⊂ E, where dist(·, ·) is the Hausdorff semi-distance of two
subsets in E: dist(B1 , B2 ) = sup inf kx − yk.
x∈B1 y∈B2

Definition 2.5. The m-semiflow G is called pointwise dissipative if there is a

bounded set B0 attracting any point x ∈ E, i.e. there exists K > 0 such that for
w ∈ E, u(t) ∈ G(t, w), one has ku(t)kE ≤ K for t ≥ t0 (kwkE ).
Definition 2.6. The m-semiflow G is called asymptotically upper semicompact if
∀B ∈ B(E) such that for some T (B) ∈ Γ+ , γT+(B) ∈ B(E), any sequence ξn ∈
G(tn , B), tn → ∞, is precompact in E.
Definition 2.7. A bounded set B1 ⊂ E which has the property that, for any bounded
set B ⊂ E there exists τ = τ (B) ≥ 0 such that γτ+ (B) ⊂ B1 , is called an absorbing
set for m-semiflow G.
It is obvious that if m-semiflow G has an absorbing set then it is pointwise
dissipative and eventually bounded.
The following theorem gives a sufficient condition for the existence of a global
attractor for the m-semiflow G.
Theorem 2.5. [13] Assume that the m-semiflow G has the following properties:
6 N.T.V. ANH, T.D. KE

(1) G(t, ·) is u.s.c and has closed values for each t ∈ Γ+ ;

(2) G is pointwise dissipative;
(3) G is asymptotically upper semicompact.
If G is eventually bounded then it possesses a compact global attractor A in E.
Moreover, if G is a strict m-semiflow then A is invariant, that is A = G(t, A) for
any t ∈ Γ+ .

3. Existence result on compact intervals

Put
J = [0, T ], CT = C([0, T ]; Rn ), Cτ = C([−τ, 0]; Rn ), C = C([−τ, T ]; Rn ).
In what follows, we use the assumptions that:
(H1) A is a linear operator on Rn .
(H2) B : Rn × Cτ → Rn×m is a continuous map such that there exist positive
constants ηB , ζB verifying that:
kB(v, w)k ≤ ηB (kvk + kwkCτ ) + ζB ,
for all v ∈ Rn , w ∈ Cτ .
(H3) The function F : Rn → Rm is continuous and there is a positive number
ηF such that kF (v)k ≤ ηF for all v ∈ Rn .
(H4) G : K → Rm is a continuous function such that
(1) G is monotone on K, i.e.
hu − v, G(u) − G(v)i ≥ 0, ∀u, v ∈ K;
(2) there exists v0 ∈ K such that
hv − v0 , G(v)i
lim > 0.
v∈K,kvk→∞ kvk2
(H5) h : Rn → Rn is continuous such that there are positive constants ηh , ζh
verifying
kh(u)k ≤ ηh kuk + ζh , ∀u ∈ Rn .
We now give the definition of solution for DVI (1.1)-(1.3).
Definition 3.1. A continuous function x : [−τ, T ] → Rn is called a solution of
(1.1) − (1.3) iff there exists an integrable function u : J → K such that
Zt Zt
tA (t−s)A
x(t) = e ϕ(0) + e B(x(s), xs )u(s)ds + e(t−s)A h(x(s))ds, t ∈ J,
0 0
hv − u(t), F (x(t)) + G(u(t))i ≥ 0, for a.e t ∈ J, ∀v ∈ K,
x(s) = ϕ(s), s ∈ [−τ, 0].
We denote
SOL(K, Q) = {v ∈ K : hw − v, Q(v)i ≥ 0, ∀w ∈ K}, (3.1)
where Q : Rm → Rm is a given mapping.
Due to [14, Proposition 6.2], we get the following result.
 ASYMPTOTIC BEHAVIOR FOR A CLASS OF DVIS 7

Lemma 3.1. Let (H4) hold. Then for each z ∈ Rm , the solution set SOL(K, z +
G(·)) is non-empty, convex and closed. Moreover, there exists ηG > 0 such that
kvk ≤ ηG (1 + kzk), ∀v ∈ SOL(K, z + G(·)). (3.2)
In order to solve (1.1) − (1.3), we convert it to a differential inclusion. Let
U (z) = SOL(K, z + G(·)), z ∈ Rm .
Then we get that U : Rm → P(Rm ) has compact convex values, thanks to Lemma
3.1. Moreover, it is easy to verify that U is a closed map. By (3.2), we see that U
is locally bounded, then it is u.s.c.
Now we define Φ : Rn × Cτ → P(Rn ) as follows:
Φ(v, w) = {B(v, w)y + h(v) : y ∈ U (F (v))}. (3.3)
n
Since B(v, w) is a linear operator for each v ∈ R , w ∈ Cτ and U has compact
convex values, Φ also has compact convex values. Furthermore, thanks to the
continuity of B, F , h and the fact that U is u.s.c, the composition multimap Φ is
u.s.c as well.
Due to the above setting, DVI (1.1)-(1.3) is converted to the following differential
inclusion
x0 (t) ∈ Ax(t) + Φ(x(t), xt ), t ∈ J, (3.4)
x(t) = ϕ(t), t ∈ [−τ, 0]. (3.5)
Denote
PΦ (x) = {f ∈ L1 (J; Rn ) : f (t) ∈ Φ(x(t), xt )}, for x ∈ C. (3.6)
Then we deduce that a solution x ∈ C of DVI (1.1)-(1.3) is given by the following
formula
Zt
x(t) = etA ϕ(0) + e(t−s)A f (s)ds, f ∈ PΦ (x), t ∈ J, (3.7)
0
x(t) = ϕ(t), t ∈ [−τ, 0]. (3.8)
For y ∈ CT and ϕ ∈ Cτ , we define the function y[ϕ] ∈ C as follows
(
y(t), if t ∈ [0, T ],
y[ϕ](t) =
ϕ(t), if t ∈ [−τ, 0].
By defining the following operator
W : L1 (J; Rn ) → CT
Zt
W(f )(t) = e(t−s)A f (s)ds, (3.9)
0

we construct the solution operator F : CT → P(CT ) as follows

F(y)(t) = {etA ϕ(0) + W(f )(t) : f ∈ PΦ (y[ϕ])}, t ∈ J.
It is obvious that y ∈ CT is a fixed point of F iff y[ϕ] is a solution of (1.1)-(1.3).
Lemma 3.2. Under assumptions (H2)-(H5), PΦ is well-defined and weakly u.s.c.
8 N.T.V. ANH, T.D. KE

Proof. Based on the hypotheses and the result of Lemma 3.1, we get
kΦ(v, w)k := sup{kzk : z ∈ Φ(v, w)}
≤ kB(v, w)kηG (1 + kF (v)k) + kh(v)k
≤ ηG (1 + ηF )[ηB (kvk + kwkCτ ) + ζB ] + ηh kvk + ζh . (3.10)
Since Φ is u.s.c with compact convex values, the multimap Λ(t) = Φ(x(t), xt ) is
strongly measurable due to [11, Proposition 1.3.1]. Therefore it has a Castaing
representation (see [11, Definition 1.3.3]) and hence PΦ (x) 6= ∅ for x ∈ C.
We prove the second assertion by using Lemma 2.3. Let {xk } ⊂ C such that
xk → x∗ , fk ∈ PΦ (xk ). Then {fk (t)} ⊂ C(t) := Φ({xk (t), (xk )t }), and C(t) is a
compact set for each t ∈ J. Furthermore, by (3.10), {fk } is integrably bounded
(bounded by an integrable function). Thus {fk } is weakly relatively compact in
L1 (J; Rn ) (see [7, Corollary 2.6]). Let fk * f ∗ in L1 (J; Rn ), then by Mazur’s
lemma (see, e.g [5]) there are f¯k ∈ co{fi : i ≥ k} such that f¯k → f ∗ in L1 (J; Rn )
and then f¯k (t) → f ∗ (t) for a.e. t ∈ J, up to a subsequence. Observe that in our
case, the upper semicontinuity of Φ implies that for a given  > 0
Φ(xk (t), (xk )t ) ⊂ Φ(x∗ (t), x∗t ) + B for all large k,
here B is the ball in Rn centered at origin with radius . So
fk (t) ∈ Φ(x∗ (t), x∗t ) + B for a.e. t ∈ J,
and
f¯k (t) ∈ Φ(x∗ (t), x∗t ) + B for a.e. t ∈ J,
thanks to the convexity of Φ(x∗ (t), x∗t ) + B . The last inclusion deduces f ∗ (t) ∈
Φ(x∗ (t), x∗t ) + B for a.e. t ∈ J. Since  is arbitrary, one obtains f ∗ ∈ PΦ (x∗ ). The
lemma is proved. 
Lemma 3.3. The operator W defined by (3.9) is compact.
Proof. We show that W(Ω) is relatively compact in CT for any bounded set Ω ⊂
L1 (J; Rn ). Obviously, we get that W(Ω)(t) is bounded in Rn . In addition, W(Ω) is
equicontinuous thanks to the fact that S(t) = etA is a norm-continuous semigroup.
So we get the conclusion by using Arzelà-Ascoli theorem. 
Lemma 3.4. Let (H1)-(H5) hold. Then the solution operator F is compact and
has a closed graph.
Proof. Since W is compact, it is easy to check that F(B) is relatively compact for
any bounded set B ⊂ CT . So F is a compact multimap.
Now let {xk } ⊂ CT , xk → x∗ , yk ∈ F(xk [ϕ]) and yk → y ∗ . We will verify that
y ∈ F(x∗ ). By the formulation of F, one can take fk ∈ PΦ (xk [ϕ]) such that
∗

yk (t) = etA ϕ(0) + W(fk )(t), t ∈ J. (3.11)

Since PΦ is weakly u.s.c and {xk } is relatively compact one has that {fk } is
weakly relatively compact, so we can assume that fk * f ∗ in L1 (J; Rn ), up to
a subsequence. Moreover, f ∗ ∈ PΦ (x∗ [ϕ]). By the compactness of W, we obtain
W(fk ) → W(f ∗ ) in CT . Passing (3.11) into limits as k → ∞, we get
y ∗ (t) = etA ϕ(0) + W(f ∗ )(t), t ∈ J.
Thus y ∗ ∈ F(x∗ ). The proof is complete. 
 ASYMPTOTIC BEHAVIOR FOR A CLASS OF DVIS 9

Theorem 3.5. Assume (H1)-(H5). Then problem (3.4)-(3.5) has at least one so-
lution on [−τ, T ]. Moreover, the solution set is compact.

Proof. We first prove using Theorem 2.4 that Fix(F) 6= ∅. According to Lemma
3.4, it suffices to show that there exists a closed convex set M0 ⊂ CT satisfying that
F(M0 ) ⊂ M0 . Let y ∈ F(x). Then it follows from the definition of the solution
operator and the estimate (3.10) that there exists f ∈ PΦ (x[ϕ]) verifying

Zt
tA
ky(t)k = ke ϕ(0) + e(t−s)A f (s)dsk
0
Zt
≤ M kϕ(0)k + ke(t−s)A kkf (s)kds
0
Zt
≤ M kϕ(0)k + M [(η + ηh )kx(s)k + ηkx[ϕ]s kCτ + ηG (1 + ηF )ζB + ζh ] ds, ∀t ∈ J,
0

where M = sup ketA k, η = ηG (1 + ηF )ηB .

t∈J
On the other hand, due to the estimate

kx[ϕ]s kCτ = sup kx[ϕ](s + θ)k ≤ kϕkCτ + sup kx(ρ)k,

θ∈[−τ,0] ρ∈[0,s]

one has
Z t
ky(t)k ≤ M1 + M ((η + ηh )kx(s)k + η sup kx(ρ)k)ds
0 ρ∈[0,s]
Z t
≤ M1 + M (2η + ηh ) sup kx(ρ)kds,
0 ρ∈[0,s]

where M1 = M kϕ(0)k + M T [ηkϕkCτ + ηG (1 + ηF )ζB + ζh ]. Since the last term is

nondecreasing in t, we have

Zt
sup ky(ρ)k ≤ M1 + M (2η + ηh ) sup kx(ρ)k)ds. (3.12)
ρ∈[0,t] ρ∈[0,s]
0

Denote
M0 = {x ∈ CT : sup kx(s)k ≤ ψ(t), t ∈ [0, T ]},
s∈[0,t]

where ψ is the unique solution of the integral equation

Z t
ψ(t) = M1 + M (2η + ηh ) ψ(s)ds, t ∈ J.
0

It is clear that M0 is a bounded closed convex subset of CT and estimate (3.12)

ensures that F(M0 ) ⊂ M0 . The proof is complete. 
10 N.T.V. ANH, T.D. KE

4. Decay solutions
In this section, we consider the solution operator F on BC(0, ∞; Rn ). For a
positive number γ and ϕ ∈ Cτ , denote

Bϕγ (R) = {x ∈ C([0, ∞); Rn ) : x(0) = ϕ(0), eγt kx(t)k ≤ R for all t ≥ 0}.

Then Bϕγ (R) is a closed bounded convex subset of BC(0, ∞; Rn ). We need to

replace the assumptions (H1), (H2) and (H5) by stronger ones:
(H1*) A is a linear operator on Rn such that there exists a > 0 : h−Az, zi ≥ akzk2
for all z ∈ Rn .
(H2*) B satisfies (H2) with ζB = 0.
(H5*) h fulfills (H5) with ζh = 0.

Lemma 4.1. Under hypotheses (H1*), (H2*), (H3)-(H4) and (H5*), F(Bϕγ (R)) ⊂
Bϕγ (R) for some R > 0, provided that

ηG (1 + ηF )ηB (1 + eγτ ) + ηh + γ < a. (4.1)

Proof. By (H1*) we have

ketA k ≤ e−at , t ≥ 0. (4.2)
Assume opposite: for each n ∈ N there exists xn ∈ Bϕγ (n) and yn ∈ F(xn ) with
yn 6∈ Bϕγ (n). Then there exists fn ∈ PΦ (xn [ϕ]) such that

Zt
tA
yn (t) = e ϕ(0) + e(t−s)A fn (s)ds, ∀t ≥ 0,
0

Using (4.2) and estimate (3.10) we have

Zt
−at
kyn (t)k ≤ e kϕkCτ + ηG (1 + ηF )ηB e−a(t−s) (kxn (s)k + kxn [ϕ]s kCτ )ds
0
Zt
+ ηh e−a(t−s) kxn (s)kds (4.3)
0

Now one observes that eγt kxn (t)k ≤ n for all t ≥ 0. Then for all t ≥ τ , we get

eγt kxn [ϕ]t kCτ = eγt sup kxn (t + ρ)k

ρ∈[−τ,0]

= eγt sup e−γ(t+ρ) eγ(t+ρ) kxn (t + ρ)k

ρ∈[−τ,0]
γt −γ(t−τ )
≤e e sup eγ(t+ρ) kxn (t + ρ)k
ρ∈[−τ,0]
≤ neγτ .

On the other hand, for t ∈ [0, τ ] one has eγt kxn [ϕ]t kCτ ≤ eγτ kϕkCτ . Hence

eγt kxn [ϕ]t kCτ ≤ eγτ (n + kϕkCτ ) for all t ≥ 0.

 ASYMPTOTIC BEHAVIOR FOR A CLASS OF DVIS 11

So it deduces from (4.3) that

Zt
−(a−γ)t
γt
e kyn (t)k ≤ e kϕkCτ + [ηG (1 + ηF )ηB + ηh ] e−(a−γ)(t−s) eγs kxn (s)kds
0
Zt
+ ηG (1 + ηF )ηB e−(a−γ)(t−s) eγs kxn [ϕ]s kCτ ds
0
≤ kϕkCτ + {n[ηG (1 + ηF )ηB + ηh ] + (n + kϕkCτ )eγτ ηG (1 + ηF )ηB }I,
where Z t
1
I= e−(a−γ)(t−s) ds = (1 − e−(a−γ)t ).
0 a−γ
Therefore
1 1   C
sup eγt kyn (t)k ≤ ηG (1 + ηF )ηB (1 + eγτ ) + ηh + , (4.4)
n t≥0 a−γ n
where
1 
kϕkCτ ηG (1 + ηF )ηB eγτ .

C = kϕkCτ +
a−γ
Passing (4.4) into limits, we get a contradiction with (4.1). The proof is complete.


We now prove the main result of this section.

Theorem 4.2. Assume that (H1*)-(H2*), (H3)-(H4), (H5*) take place and there
exists γ > 0 such that
ηG (1 + ηF )ηB (1 + eγτ ) + ηh + γ < a.
Then the DVI (1.1)-(1.3) has a nonempty compact set of solutions on [−τ, ∞)
satisfying
eγt kx(t)k = O(1) as t → ∞.
Proof. By Lemma 4.1, one can consider F : Bϕγ (R) → P(Bϕγ (R)) for a number
R > 0. Due to Theorem 2.4 it remains to show that F is compact and u.s.c.
We first prove that F is a compact multimap. Let D ⊂ Bϕγ (R). Recall that
χ∗ (D) = χ∞ (D) + d∞ (D) where χ∞ and d∞ are defined by Section 2.
By the same arguments as in the proof of Lemma 3.4, we have
χT (πT (F(D))) = 0.
Therefore
χ∞ (F(D)) = 0. (4.5)
In the rest of proof, we will show that d∞ (F(D)) = 0. Let x ∈ D and y ∈ F(x).
Then in view of Lemma 4.1 we have
ky(t)k ≤ Ce−γt , ∀t ≥ 0,
where C = C(R, a, γ, ηB , ηF , ηG , ηh ). Thus for T > 0, one gets
sup ky(t)k ≤ Ce−γT , ∀y ∈ D.
t≥T
12 N.T.V. ANH, T.D. KE

This implies dT (D) ≤ Ce−γT , so d∞ (D) = lim dT (D) = 0. Combining this with
T →∞
(4.5) yields
χ∗ (F(D)) = 0.
Since the MNC χ∗ is regular, we conclude that F(D) is relatively compact.
To prove that F is u.s.c, it suffices to show that F has closed graph. But
this is done by the same arguments as in the proof of Lemma 3.4. The proof is
complete. 

5. Existence of global attractor

The m-semiflow governed by DVI (1.1) − (1.3) is defined as follows
G : R+ × Cτ → P(Cτ )
G(t, ϕ) = {xt : x[ϕ] is a solution of (1.1) − (1.3) on [−τ, T ] for any T > 0},
By the same argument as in [6], we see that
G(t1 + t2 , ϕ) = G(t1 , G(t2 , ϕ)), for all t1 , t2 ∈ R+ , ϕ ∈ Cτ .
For each ϕ ∈ Cτ we denote
Σ(ϕ) = {x ∈ C([0, ∞); Rn ) : x[ϕ] is a solution of (1.1)-(1.3)
on [−τ, T ] for any T > 0}.
It is clear to see that
πt ◦ Σ(ϕ) ⊂ S(·)ϕ(0) + W ◦ PΦ (πt ◦ Σ(ϕ)[ϕ]). (5.1)
In addition, we have G(t, ϕ) = {x[ϕ]t : x ∈ Σ(ϕ)}. On the other hand, by Theorem
3.5, πt ◦ Σ(ϕ) is a compact set in C([0, t]; Rn ) for any t > 0. It follows that G(t, ϕ)
is compact in Cτ , and then G(t, ·) has compact values. In fact, we have a further
result in the following lemma.
Lemma 5.1. Let the hypotheses (H1)-(H5) hold. Then G(t, ·) is a compact mul-
timap for each t > τ .
Proof. Let Ω ⊂ Cτ be a bounded set and {zn } ⊂ G(t, Ω) be a sequence. Then for
each n one can find a function ϕn ∈ Ω and xn ∈ Σ(ϕn ) such that zn = xn [ϕn ]t .
Since t > τ we have
zn = xn (t + ·) = e(t+·)A ϕn (0) + W(fn )(t + ·),
where fn ∈ PΦ (xn [ϕn ]). Since {ϕn (0)} ⊂ Rn is a bounded set, the set {e(t+·)A ϕn (0)}
is relatively compact in Cτ . On the other hand, it is easily seen that {xn } is a
bounded sequence and it turns out that {fn } is integrably bounded due to estimate
(3.10). Since W is a compact operator, we get that {W(fn )} is relatively compact in
Cτ as well. Thus {zn } is relatively compact as desired. The proof is complete. 
Corollary 5.1. Let the hypotheses (H1)-(H5) hold. Then the m-semiflow G is
asymptotically upper semicompact.
Proof. Taking t1 > τ , we have that G(t1 , ·) is compact, thanks to Lemma 5.1. Then
the conclusion follows by [13, Proposition 1]. 
Lemma 5.2. Let the hypotheses (H1)-(H5) hold. Then G(t, ·) is u.s.c for each
t ≥ 0.
 ASYMPTOTIC BEHAVIOR FOR A CLASS OF DVIS 13

Proof. Since G(t, ·) is a compact multimap with compact values, it suffices to prove
that G(t, ·) is closed for each t ≥ 0, thanks to Lemma 2.2. Let ϕn → ϕ∗ in Cτ and
zn ∈ G(t, ϕn ) such that zn → z ∗ . We show that z ∗ ∈ G(t, ϕ∗ ), i.e. z ∗ = x∗ [ϕ∗ ]t
for an x∗ ∈ Σ(ϕ∗ ). Taking xn ∈ Σ(ϕn ) such that zn = xn [ϕn ]t , one can find
fn ∈ PΦ (xn [ϕn ]) verifying
xn = e(·)A ϕn (0) + W(fn ). (5.2)
n
Since {ϕn } is bounded in Cτ , {xn } is a bounded sequence in C([0, T ]; R ) for
any T > 0. Thus {fn } is integrably bounded in L1 (0, T ; Rn ). The compactness
of W implies that {W(fn )} is relatively compact in C([0, T ]; Rn ). In addition,
{e(·)A ϕn (0)} is a convergent sequence in C([0, T ]; Rn ), then taking into account
(5.2) we see that {xn } has a convergent subsequence (still denoted by {xn }). Let
x∗ = lim xn in C([0, T ]; Rn ). Then xn [ϕn ] → x∗ [ϕ∗ ] in C([−τ, T ]; Rn ). Since PΦ
n→∞
is weakly u.s.c, we have fn * f ∗ ∈ PΦ (x∗ [ϕ∗ ]) up to a subsequence, thanks to
Lemma 2.3. Therefore one can pass (5.2) into limits to get
x∗ = e(·)A ϕ∗ (0) + W(f ∗ ),
for f ∗ ∈ PΦ (x∗ [ϕ∗ ]). That is, x∗ [ϕ∗ ] is a solution of (1.1)-(1.3) and then x∗ [ϕ∗ ]t ∈
G(t, ϕ∗ ). Obviously, we have zn = xn [ϕn ]t → z ∗ = x∗ [ϕ∗ ]t and z ∗ ∈ G(t, ϕ∗ ). The
proof is complete. 

In order to apply Theorem 2.5, it remains to show that G has an absorbing set
in Cτ . We make use of the following result (see [10]).
Proposition 5.3 (Halanay’s inequality). Let the function f : [t0 − τ, T ) → R+ , 0 ≤
t0 < T < +∞ satisfy the functional differential inequality
f 0 (t) ≤ −γf (t) + ν sup f (s),
s∈[t−τ,t]

for t ≥ t0 , where γ > ν > 0. Then

f (t) ≤ κe−`(t−t0 ) , t ≥ t0 ,
where κ = sup f (s) and ` is the solution of the equation γ = ` + νe−`τ .
s∈[t0 −τ,t0 ]

Using Halanay’s inequality, we obtain the following result.

Lemma 5.4. Let (H1*) and (H2)-(H5) hold. Then the m-semiflow G admits an
absorbing set, provided that 2ηB ηG (1 + ηF ) + ηh < a.
Proof. For t > 0 and ϕ ∈ Cτ , we consider the solution x[ϕ] given by
Zt
tA
x(t) = e ϕ(0) + e(t−s)A f (s)ds,
0

for f ∈ PΦ (x[ϕ]). Using (H1*) and estimate (3.10), we obtain

kx(t)k ≤ e−at kϕ(0)k
Zt
+ e−a(t−s) [(η + ηh )kx(s)k + ηkx[ϕ]s kCτ + ηG (1 + ηF )ζB + ζh ]ds,
0
14 N.T.V. ANH, T.D. KE

where η = ηB ηG (1 + ηF ). Since a − (2η + ηh ) > 0, one can choose R > 0 such

ηG (1 + ηF )ζB + ζh
that η + = d < a − (η + ηh ). Firstly, we prove that for ϕ ∈ Cτ
R
satisfying kϕkCτ ≤ C, the solution x[ϕ] has the following property: ∃t0 > 0 such
that kx[ϕ]t0 kCτ ≤ R. Assume to the contrary that for all t > 0, kx[ϕ]t kCτ > R.
Then
ηkx[ϕ]s kCτ + ηG (1 + ηF )ζB + ζh ≤ dkx[ϕ]s kCτ , ∀s ≥ 0.
Therefore we have
Zt
−at
kx(t)k ≤ e kϕ(0)k + e−a(t−s) [(η + ηh )kx(s)k + dkx[ϕ]s kCτ ]ds, t ≥ 0.
0

Let

t
e−at kϕ(0)k + R e−a(t−s) [(η + η )kx(s)k + dkx k ]ds, t ≥ 0,

h s Cτ
y(t) = 0
kx(t)k, t ∈ [−τ, 0].



Then kx(t)k ≤ y(t), ∀t ≥ −τ and the following estimate holds

y 0 (t) ≤ −[a − (η + ηh )]y(t) + d sup y(s).
s∈[t−τ,t]

Applying Halanay’s inequality yields

kx(t)k ≤ kϕkCτ e−`t ≤ Ce−`t , ∀t ≥ 0,
where ` is a positive number.
Then we have
R < kxt kCτ = sup kx(t + θ)k ≤ Ce`τ e−`t , ∀t ≥ 0.
θ∈[−τ,0]

This inequality indicates that kxt kCτ tends to zero as t → ∞, so one can find t1 > 0
such that kxt1 kCτ < R. We get a contradiction.
We have just proved that if kϕkCτ ≤ C, then there exists t0 > 0 such that
kxt0 kCτ ≤ R. In the sequel, we assert that kut kCτ ≤ R, ∀t ≥ t0 . Assume to the
contrary that there exists t1 ≥ t0 satisfying
kxt1 kCτ ≤ R but kxt kCτ > R for all t ∈ (t1 , t1 + θ),
where θ > 0. Regarding the solution x[ϕ] on [t1 , t1 + θ), we have
Zt
(t−t1 )A
x(t) = e x(t1 ) + e(t−s)A f (s)ds, f ∈ PΦ (x[ϕ]).
t1

Then
Zt
−a(t−t1 )
kx(t)k ≤ e kϕ(0)k + e−a(t−s) [(η + ηh )kx(s)k + dkxs kCτ ]ds, t ∈ [t1 , t1 + θ).
t1

Using the same arguments as above, we see that for all t ∈ [t1 , t1 + θ),
kx(t)k ≤ kxt1 kCτ e−`(t−t1 ) ≤ kxt1 kCτ ≤ R.
 ASYMPTOTIC BEHAVIOR FOR A CLASS OF DVIS 15

Hence for t ∈ [t1 , t1 + θ), we have

kxt kCτ = sup kx(t + s)k = sup kx(r)k
s∈[−τ,0] r∈[t−τ,t]

≤ sup kx(r)k
r∈[t1 −τ,t]

= max{ sup kx(r)k, sup kx(r)k}

r∈[t1 −τ,t1 ] r∈[t1 ,t]

= max{kxt1 kCτ , sup kx(r)k} ≤ R.

r∈[t1 ,t]

This is a contradiction. In summary, one can take a ball centered at origin with
radius R as an absorbing set for the m-semiflow G, where R is chosen such that
ηG (1 + ηF )ζB + ζh
R> . 
a − (2η + ηh )
Theorem 5.5. Let (H1*), (H2)-(H5) hold. Then the m-semiflow G generated by
(1.1)-(1.3) admits a compact global attractor provided that
2ηB ηG (1 + ηF ) + ηh < a.
Proof. The conclusion follows from Corollary 5.1, Lemma 5.2 and 5.4. 
Acknowledgment. The authors are grateful to anonymous referee for the careful
reading and constructive comments and suggestions, that help to improve the man-
uscript. The work was supported by Vietnam Ministry of Education and Training,
under project B2014-17-70.

References
[1] J.-P. Aubin, A. Cellina, Differential inclusions. Set-valued maps and viability theory,
Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical
Sciences], 264. Springer-Verlag, Berlin, 1984.
[2] R.R. Akhmerov, M.I. Kamenskii, A.S. Potapov, A.E. Rodkina, B.N. Sadovskii, Measures of
noncompactness and condensing operators, Birkhäuser, Boston-Basel-Berlin, 1992.
[3] E.P. Avgerinos, N.S. Papageorgiou, Differential variational inequalities in RN , Indian J.
Pure Appl. Math. 28 (1997), no.9, 1267-1287.
[4] J. Banas and L. Olszowy, On a class of measures of noncompactness in Banach algebras and
their application to nonlinear integral equations, J. Anal. Appl. 28 (2009), 475-498.
[5] D. Bothe, Multivalued perturbations of m-accretive differential inclusions, Israel J. Math. 108
(1998), 109-138.
[6] T. Caraballo, P.E. Kloeden, Non-autonomous attractors for integro-differential evolution
equations, Discrete Contin. Dyn. Syst. Ser. S 2 (2009), 17-36.
[7] J. Diestel, W. M. Ruess, W Schachermayer, Weak compactness in Ll (µ, X), Proc. Amer.
Math. Soc. 118 (1993), 447-453.
[8] J. Gwinner, On differential variational inequalities and projected dynamical systems - equiv-
alence and a stability result, Discrete Contin. Dynam. Syst. 2007, Dynamical Systems and
Differential Equations. Proceedings of the 6th AIMS International Conference, suppl., 467-
476.
[9] J. Gwinner, A note on linear differential variational inequalities in Hilbert Spaces, in: IFIP
Advances in Information and Communication Technology, Syst. Model. Optim. Vol. 391,
2013, 85-91.
[10] A. Hanalay, Differential equations, stability, oscillations, time lags, Academic Press, New
York, London 1996.
[11] M. Kamenskii, V. Obukhovskii, P. Zecca, Condensing multivalued maps and semilinear dif-
ferential inclusions in Banach spaces, in: de Gruyter Series in Nonlinear Analysis and Ap-
plications, vol. 7, Walter de Gruyter, Berlin, New York, 2001.
 16 N.T.V. ANH, T.D. KE

[12] Z. Liu, N.V. Loi, V. Obukhovskii, Existence and global bifurcation of periodic solutions to a
class of differential variational inequalities, Int. J. Bifur. Chaos 23 (2013) No.7, 1350125.
[13] V.S. Melnik, J. Valero, On attractors of multivalued semi-flows and differential inclusions,
Set-Valued Anal. 6 (1998), 83-111.
[14] J.-S. Pang, D.E.Stewart, Differential variational inequalities, Math. Program. Ser. A 113
(2008), 345-424.
[15] I. Ekeland, R. Temam, Convex analysis and variational problems, Society for Industrial and
Applied Mathematics (SIAM), Philadenphia, PA, 1999.

Nguyen Thi Van Anh and Tran Dinh Ke

Department of Mathematics, Hanoi National University of Education
136 Xuan Thuy, Cau Giay, Hanoi, Vietnam
E-mail address: [email protected] (N.T.V.Anh), [email protected] (T.D. Ke)

View publication stats