Professional Documents
Culture Documents
1. Introduction
It is our purpose to study a topological structure of the solution set to the
impulsive Cauchy problem governed by a semilinear differential inclusion on
noncompact intervals.
For a fixed τ > 0 and a given piecewise continuous function x : [−τ, 0] →
E, where E is a Banach space, the problem we deal with is
⎧
⎨ ẏ(t) ∈ A(t)y(t) + F (t, yt ), for a.a. t ∈ [0, ∞), t = tk , k ∈ N,
y(t) = x(t), for t ∈ [−τ, 0], (1)
⎩ +
y(tk ) = y(tk ) + Ik (ytk ), for k ∈ N+ ,
where {A(t)}t∈[0,∞) is a family of linear operators in E generating an evolution
operator; F is an upper-Carathéodory map; yt (θ) = y(t + θ), θ ∈ [−τ, 0]; Ik are
impulse functions, k ∈ N, y(t+ ) = lims→t+ y(s) and the time sequence (tk )k∈N
is an increasing sequence of given points in [0, ∞) without accumulation points.
Hence yt (·) represents the history of the state from t − τ to the present time t.
This research was partially supported by the MNiSW scientific project no. N N201 395137.
G. Gabor and A. Grudzka NoDEA
For the first time differential equations with impulses were investigated
by Milman and Myshkis [1]. Impulsive differential equations and inclusions
have applications in biology, economics, medicine, physics and other fields.
These problems often describe phenomenas in which states are changing rap-
idly. One of the example is the motion of an elastic ball bouncing vertically on
a surface. The moments of the impulses are in the time when the ball touches
the surface and rapidly its velocity is changed. The moments of the impulse
effects for the impulsive problems can be chosen in various ways: randomly,
fixed beforehand, determined by the state of a system. For some recent works
on impulsive differential problems, concerning the aspects we deal with, we
refer to [2–6].
The solution sets for differential problems often correspond with fixed
point sets of multivalued operators in function spaces. In this paper we use
the inverse system method, which, in studying the topological structure of fixed
point sets of operators in function spaces, was initiated in [7]. This method was
developed in [8] and then also in [9]. It is observed that differential problems on
noncompact intervals can be reformulated as fixed point problems in Fréchet
(function) spaces which are inverse limits of Banach spaces that appear when
we consider these differential problems on compact intervals. Some interesting
properties of fixed point sets of limit maps become very useful.
The existence of mild solutions for problem (1) has been obtained in [2].
We state and prove the compactness of the solution set for this problem. Next
we prove that the set of solutions to problem (1) is an Rδ -set.
This gives an important information from the topological point of view.
The translation operator along trajectories which is often used to detect, for
instance, periodic solutions, being Rδ -valued can be checked to be an admis-
sible (in the sense of Górniewicz) multivalued operator, and the fixed point
theory methods can be applied.
The paper is organized as follows. In Sect. 2 we recall useful definitions
and preliminary theorems. In the main Sect. 3 we obtain new results. In
Theorem 3.4 the compactness of the solution set is proved on the half-line.
The result improves Theorem 4.2 in [2], where only the existence of solutions
was shown. Our proof is essentially shorter and it shows how one can effectively
use the inverse systems technique. Theorems 3.5 and 3.7 are the main results
of the paper. We prove the Rδ -structure of the solution set. Note that in [4] it
was shown that the solution set for the impulsive problem on compact intervals
is an Rδ -set if F is a σ-Ca-selectionable multivalued map and A(t) = A is the
infinitesimal generator of a C0 -semigroup. The problem is that it is not clear
if a sufficiently good σ-Ca-selectionability is possible in infinite dimensional
spaces. In fact, as we show in the proof of Theorem 3.5, we can approxi-
mate the right-hand side of the inclusion by maps which have noncompact
values and which are not k-set contractions. Therefore, we propose different
arguments to avoid the obstacles and to prove the Rδ -structure on compact
intervals. Finally, we combine the information on a topological structure of
solution sets on compact intervals with the inverse systems technique to obtain
an Rδ -structure on the half-line in Theorem 3.7. In this way we essentially
Structure of the solution set
develop some recent results in [5], where an Rδ -structure of the solution set
for the multivalued impulsive differential inclusion on the half-line is shown
only in a finite dimensional case, where the compactness properties become
much easier, and for the problem without any retard.
2. Preliminaries
Let X, Y be two topological vector spaces. We denote by P (Y ) the family of
all nonempty subsets of Y and put Pcl (Y ) = {A ∈ P (Y ), closed}, Pcl,cv (Y ) =
{A ∈ P (Y ), closed and convex}, Pcp (Y ) = {A ∈ P (Y ), compact}, Pcp,cv (Y ) =
{A ∈ P (Y ), compact and convex}.
A multivalued map F : X → P (Y ) is said to be upper semicontinuous
(for short u.s.c.) if F −1 (V ) = {x ∈ X | F (x) ⊂ V } is an open subset of X for
every open V ⊆ Y. A multivalued map F : X → P (Y ) is said to be lower
semicontinuous (for short l.s.c.) if F+−1 (V ) = {x ∈ X | F (x) ∩ V = ∅} is an
open subset of X for every open V ⊆ Y. We say that a multivalued map
F : X → P (Y ) is continuous provided it is both u.s.c. and l.s.c.
Let (X, d) be a metric space and BC(X) denote the family of all non-
empty closed bounded subsets of X. For given A, B ∈ BC(X) let:
dH (A, B) = inf{ > 0 | A ⊂ O (B) and B ⊂ O (A)},
where O (A) = {x ∈ X | dist(x, A) < }. Observe that
dH (A, B) = max{sup dist(a, B), sup dist(b, A)}.
a∈A b∈B
measurable and integrably bounded, i.e., there exists α ∈ L1 (I) such that
||F (t)|| := max{||y|| | y ∈ F (t)} ≤ α(t)
for a.e. t ∈ I, then there exists a Bochner integrable selector f of F, i.e.,
f (t) ∈ F (t) for a.e. t ∈ I.
We say that a family V ⊂ L1 (I, E) is integrably bounded if V : I → P (E)
given by V (t) = {v(t) | v ∈ V} is integrably bounded.
Theorem 2.1. (see [10], Proposition 4.2.1.) Let E be a Banach space and V ⊂
L1 ([a, b], E) be integrably bounded. Assume that the sets V (t) are relatively
compact for a.e. t ∈ [a, b]. Then V is weakly compact in L1 ([a, b], E).
We denote by C([−τ, 0], E) the space of piecewise continuous functions
c : [−τ, 0] → E with finite number of discontinuity points {t∗ } such that t∗ = 0
and all values
c(t+ −
∗ ) = lim c(t∗ + h) and c(t∗ ) = lim c(t∗ + h)
h→0+ h→0−
are finite. We consider the space C([−τ, 0], E) with the L1 -norm, i.e.,
0
||c||C = ||c(t)||dt.
−τ
We do not consider the space C([−τ, 0], E) with the uniform convergence norm,
because it creates problems: the function t ∈ [0, ∞) → yt is not continuous,
moreover, it is not necessarily measurable (see Example 3.1, [11]). As a conse-
quence, the multivalued superposition operator, which we will define in Sect. 3,
would not be well defined. This space of delays with the integral norm was con-
sidered in [12] (see also [6]).
We denote by P C([a, b], E) the space of piecewise continuous functions
c : [a, b] → E with finite number of discontinuity points {t∗ } and such that
c(t+
∗ ) = lim c(t∗ + h) and lim c(t∗ + h) = c(t∗ )
h→0+ h→0−
are finite. The space P C([a, b], E) is a Banach space with the norm:
||c||P C = sup{||c(t)|| | t ∈ [a, b]}
and a space of continuous functions C([a, b], E) is a closed subspace of it.
We denote by P C([a, ∞), E) the space of piecewise continuous functions
c : [a, ∞) → E with infinite number of discontinuity points t1 , t2 , . . . such
that limn→∞ tn = +∞. The values c(t− +
i ), c(ti ) for i = 1, 2, . . . are finite and
−
c(ti ) = c(ti ). The space P C([a, ∞), E) is a Fréchet space with the family of
seminorms {pn } given by:
pn (c) = ||c|[0,tn ] ||P C
and a metric:
∞
1 pn (c1 − c2 )
d(c1 , c2 ) = .
n=1
2n 1 + pn (c1 − c2 )
Let us recall that by a Fréchet space we mean a locally convex space which is
metrizable and complete. Every Banach space is a Fréchet space.
Structure of the solution set
lim S = (xα ) ∈ Xα | πα (xβ ) = xα for all α ≤ β .
β
←−
α∈Σ
Consider two inverse systems S = {Xα , παβ , Σ} and S = {Yα , παβ , Σ }. Let us
recall (see [8]) that by a multivalued map of the system S into the systemS ,
we mean a family {σ, ϕσ(α ) } consisting of a monotone function σ : Σ → Σ,
i.e. σ(α ) ≤ σ(β ) for α ≤ β , and of multivalued maps ϕσ(α ) : Xσ(α ) Yα
with nonempty values, defined for every α ∈ Σ and such that
σ(β )
παβ ϕσ(β ) = ϕσ(α ) πσ(α ) ,
for each α ≤ β .
G. Gabor and A. Grudzka NoDEA
Theorem 3.2. (see [10], Proposition 4.2.2.) Let the operator Gk satisfy con-
ditions (G1) and (G2) and let the set {fn }∞ n=1 be integrably bounded with
∞
the property β(fn (t)n=1 ) ≤ η(t) for almost every t ∈ [tk−1 , tk ], where η(·) ∈
L1 ([tk−1 , tk ]). Then
t
∞
β(G fn (t)n=1 ) ≤ 2D
k
η(s)ds, t ∈ [tk−1 , tk ],
tk−1
and for every bounded D ⊂ C([−τ, 0], E), D(θ) = {c(θ)|c ∈ D}.
The following problem on a compact interval
⎧
⎨ ẏ(t) ∈ A(t)y(t) + F (t, yt ), for a.e. t ∈ [0, tm ], t = tk , k < m,
y(t) = x(t), for t ∈ [−τ, 0], (2)
⎩ +
y(tk ) = y(tk ) + Ik (ytk ), for k < m
was considered in [2]. The authors proved the existence of a mild solution and
that the solution set Sm for this problem is compact.
A piecewise continuous function y : [−τ, tm ] → E is a mild solution for
the impulsive Cauchy problem (2) if
t
(a) y(t) = T (t, 0)x(0) + 0<tk <t T (t, tk )Ik (ytk ) + 0 T (t, s)f (s)ds, t ∈ [0, tm ],
where f ∈ L1 ([0, tm ], E), f (s) ∈ F (s, ys ) for almost every s ∈ [0, tm ],
(b) y(t) = x(t), t ∈ [−τ, 0],
(c) y(t+ k ) = y(tk ) + Ik (ytk ), k < m.
Let [0, tm ] be a fixed interval on the real line. Put Δm = {(t, s) ∈ [0, tm ]×
[0, tm ] | s ≤ t ≤ tm }. The authors of [2] obtained the compactness of the solu-
tion set of problem (2) under the following assumptions:
G. Gabor and A. Grudzka NoDEA
and for every bounded D ⊂ C([−τ, 0], E), D(θ) = {c(θ) | c ∈ D}.
For z ∈ P C([0, tm ], E) we can define zi ∈ C([ti , ti+1 ], E), i = 0, 1, . . . ,
m − 1 as zi (t) = z(t) on (ti , ti+1 ] and zi (ti ) = z(t+ i ). For every set K ⊂
P C([0, tm ], E) we denote by Ki , i = 0, 1, . . . , m − 1 the set Ki = {zi |z ∈ K}.
It is easy to see that
Proposition 3.3. A set K ∈ P C([0, tm ], E) is relatively compact in
P C([0, tm ], E) if and only if each set Ki , i = 0, 1, . . . , m−1 is relatively compact
in C([ti , ti+1 ], E).
For any z ∈ P C([0, tm ], E), or z ∈ P C([0, ∞), E), such that z(0) = x(0)
we define the function z[x] : [−τ, tm ] → E as
x(t), t ∈ [−τ, 0],
z[x] =
z(t), t ∈ [0, tm ].
where x : [−τ, 0] → E is the function from the initial condition in (1). We
denote Ω[x] = {z[x] | z ∈ Ω}.
For a given multivalued map F : [0, tm ] × C([−τ, 0], E) → Pcp,cv (E) sat-
isfying (F 1)m − (F 4)m we consider the multivalued superposition operator
PF : D → P (L1 ([0, tm ], E)) defined as
PF (z) = {f ∈ L1 ([0, tm ], E) | f (s) ∈ F (s, z[x]s ) for a.e. s ∈ [0, tm ]}. (3)
This multivalued superposition operator PF is well defined (see e.g. [10]).
Notice that the function s ∈ [0, tm ] → z[x]s ∈ C([−τ, 0], E) is continuous.
Now, we prove the compactness of the solution set of problem (1). Note
that we simultaneously obtain a nonemptiness of the solution set and our proof
is essentially shorter than the one in [2].
Theorem 3.4. Let hypothesis (A)∞ hold, let the multivalued map F : [0, ∞) ×
C([−τ, 0], E) → Pcp,cv (E) satisfy conditions (F 1)∞ –(F 4)∞ and maps
Ik : C([−τ, 0], E) → E, k ∈ N, be continuous. Then the solution set for problem
(1) is a nonempty and compact subset of P C([0, ∞), E)[x].
Structure of the solution set
1
| f ∈ L ([0, tm ], E), f (s) ∈ F (s, ys ) for a.e. s ∈ [0, tm ]
for t ∈ [0, tm ] and φm (y)(t) = x(t) for t ∈ [−τ, 0]. Now we consider the projec-
tions pm+1
m : Cm+1 → Cm , which are defined as follows pm+1 m (y) = y|[−τ,tm ] .
We have the equalities
t
m+1
φm pm (y)(t) = T (t, 0)x(0) + T (t, tk )Ik (ytk ) + T (t, s)f (s)ds
0<tk <t 0
t
pm+1
m φm+1 (y)(t) = T (t, 0)x(0) + T (t, tk )Ik (ytk ) + T (t, s)f (s)ds
0<tk <t 0
1
| t ∈ [0, tm ], f ∈ L ([0, tm+1 ], E), f (s) ∈ F (s, ys )for a.e.s ∈ [0, tm+1 ] ,
for t ∈ [0, tm ] and φ(y)(t) = x(t) for t ∈ [−τ, 0]. Note that S := F ix(φ) =
lim←− Sm is the solution set of problem (1). It is known (see [2], Theorem
3.7) that for every m ≥ 1 the solution set Sm to (2) is a nonempty and com-
pact subset of P C([0, tm ], E)[x]. Using Proposition 2.3 we see that the set S
is nonempty and compact.
G. Gabor and A. Grudzka NoDEA
where R̄ is a common upper bound for ||I1 (zt1 )||, where z is any solution on
[−τ, t1 ], which exists because the solution set for (4) is compact and I1 is
continuous.
So, we have
t 0
||y(t)|| ≤ M̄2 + Bm α(s) ||ys (θ)||dθ ds
t1 −τ
t
≤ M̄2 + Bm α(s)τ · sup ||ys (θ)||ds
t1 −τ ≤θ≤0
t
≤ M̄2 + Bm α(s)τ · sup ||y(s + θ)||ds
t1 −τ ≤θ≤0
t
≤ M̄2 + Bm α(s)τ · sup ||y(θ)||ds
t1 s−τ ≤θ≤s
t
≤ M̄2 + Bm α(s)τ · sup ||y(θ)||ds
t1 −τ ≤θ≤s
t t
−τ Bm −τ Bm
v (t)e
α(s)ds α(s)ds
t1
≤e t1
α(t)(M̄2 + τ Bm v(t)).
Hence
t t
−τ Bm α(s)ds −τ Bm α(s)ds
(v(t)e t1
) ≤e t1
α(t)M̄2 .
We integrate both sides of this inequality from t1 to t2 and we obtain
t2
−τ Bm tt2 α(s)ds −τ Bm tt α(s)ds
v(t2 )e 1 ≤ M̄2 α(t)e 1 dt,
t1
G. Gabor and A. Grudzka NoDEA
so
v(t2 ) ≤ K2 ,
t −τ Bm tt α(s)ds
M̄2 2 α(t)e 1 dt
t1
where K2 = t
−τ Bm t 2 α(s)ds
. The function v is nondecreasing, so
e 1
The function r : C([−τ, 0], E) → clB(0, K̄m ) ⊂ C([−τ, 0], E) given by the for-
K̄m c
mula r(c) = ||c|| C
for every c ∈ C([−τ, 0], E) with ||c||C > K̄m and r(c) = c for
c with ||c||C ≤ K̄m is a continuous retraction of C([−τ, 0], E) onto a closed ball.
Therefore F̃ (t, c) = F (t, r(c)) and F̃ has the same measurability and continu-
ity properties as F. For every m ≥ 1 and t ∈ [0, tm ] the following inequalities
hold
||F̃ (t, c)|| = ||F (t, c)|| ≤ α(t)(1 + ||c||C ) ≤ α(t)(1 + K̄m )
for every ||c||C ≤ K̄m , and
K̄m c
||F̃ (t, c)|| = F t,
||c||C
K̄m c
≤ α(t) 1 + ≤ α(t)(1 + K̄m )
||c||C C
for every ||c||C > K̄m . So we have:
||F̃ (t, c)|| ≤ α(t)(1 + K̄m ) ≡ ψm (t) ∈ L1 ([0, tm ]).
Now we consider an impulsive problem for fixed m with a multivalued
map F̃
⎧
⎨ ẏ(t) ∈ A(t)y(t) + F̃ (t, yt ), for a.e. t ∈ [0, tm ], t = tk , k < m,
y(t) = x(t), for t ∈ [−τ, 0], (8)
⎩ +
y(tk ) = y(tk ) + Ik (ytk ), for k < m.
Let S̃m be the solution set of problem (8). If y is a solution of (2), i.e., y ∈ Sm ,
then ||y|| ≤ K̄m . F and F̃ coincide on clB(0, K̄m ), so we have that y ∈ S̃m . If
y ∈ S̃m , then we can easily see that ||y|| ≤ K̄m and ẏ(t) ∈ A(t)y(t) + F (t, yt )
Structure of the solution set
for a.e. t ∈ [0, tm ], t = tk , k < m and that y(t) = x(t) for t ∈ [−τ, 0], so y ∈ Sm .
We have Sm = S̃m . Consequently, we can assume from now on, without any
loss of generality, that
(F 3 )m ||F (t, c)|| ≤ ψm (t)for every t ∈ [0, tm ], where ψm (t) ∈ L1 ([0, tm ]).
Step 2. Now we prove that there exists a sequence of multivalued maps
{Gn }∞
n=1 , Gn : [0, tm ] × C([−τ, 0], E) → Pcl,cv (E) such that:
(i) each multivalued map Gn (t, ·) : C([−τ, 0], E) → Pcl,cv (E), n ≥ 1 is con-
tinuous for a.e. t ∈ [0, tm ],
· · · ⊂ Gn+1 (t, c) ⊂ Gn (t, c) ⊂ convF (t, B3dn (c)), n ≥ 1,
(ii) F (t, c) ⊂
(iii) F (t, c) = n≥1 Gn (t, c),
(iv) for each n ≥ 1 there exists a selection gn : [0, tm ] × C([−τ, 0], E) → E of
Gn , such that gn (·, c) is measurable and gn (t, ·) is locally Lipschitz.
Consider the sequence dn = 31n , n ≥ 1. Let us cover C([−τ, 0], E) by the
open balls {Bdn (c)}c∈P C([−τ,0],E) . Since the space C([−τ, 0], E) is metric, there
exists a locally finite refinement {Vj }j∈J of the cover {Bdn (c)}c∈C([−τ,0],E) ,
Now, we can associate a locally Lipschitz partition of unity {pj }j∈J subor-
dinated to the open covering {Vj }j∈J . For every j ∈ J let cj be such that
Vj ⊂ Bdn (cj ) and define
Gn (t, c) = pj (c) · convF (t, B2dn (cj )).
j∈J
To prove (ii) and (iii) note that pj (c) > 0 implies that c ∈ Vj ⊂ Bdn (cj ), hence
B2dn (cj ) ⊂ B3dn (c) and therefore
for t ∈ [0, tm ], where fn ∈ L1 ([0, tm ], E) is such that fn (s) ∈ F (s, yns ) for
almost every s ∈ [0, tm ]. m
Let R > Bm = sup(t,s)∈Δm ||T (t, s)||L(E) be such that k=1 rk + R1 <
1
Bm . We know that for every bounded linear operator S : E → E we have the
property: β(SΩ) ≤ ||S||β(Ω) for every Ω ∈ B(E). From this property (here
S = T (t, s)) we have:
β({T (t, s)fn (s)}n≥1 ) ≤ Bm β({fn (s)}n≥1 ).
For any p ≥ 1 we obtain
β({fn (s)}n≥1 ) = β({fn (s)}n≥p ) ≤ β [F ({s} × B({yns }n≥p , 3dp ))]
≤ μ(s) · sup β (B({yn (s + θ)}n≥p + 3dp ))
−τ ≤θ≤0
≤ μ(s) max( sup β({x(σ)}), sup β({yn (σ)}n≥p )) + 3dp
−τ ≤σ≤0 0≤σ≤s
= μ(s)(ρ̄(s) + 3dp ),
where ρ̄(s) = sup0≤σ≤s β({yn (σ)}n≥1 ). We have
β T (t, tk )Ik (yntk ) =β T (t, tk )Ik (yntk )
0<tk <t n≥1 0<tk <t n≥p
≤ ||T (t, tk )|| · β({Ik (yntk )}n≥p )
0<tk <t
≤ Bm rk sup β({yntk (θ)}n≥p )
0<tk <t −τ ≤θ≤0
≤ Bm rk · max( sup β({x(σ)}), sup β({yn (σ)}n≥1 ))
0<tk <t −τ ≤σ≤0 0≤σ≤tk
= Bm rk · sup β({yn (σ)}n≥1 )
0<tk <t 0≤σ≤tk
= Bm rk · ρ(tk ).
0<tk <t
Structure of the solution set
Now
β({yn (t)}n≥1 )
t
=β {T (t, 0)x(0) + T (t, tk )Ik (yntk ) + T (t, s)fn (s)ds}n≥1
0<tk <t 0
t
≤ Bm rk · ρ(tk ) + 2Bm μ(s)(ρ̄(s) + 3dp )ds
0<tk <t 0
To prove that f0 (s) ∈ F (s, ys ) for a.e. s ∈ [0, tm ] it is sufficient to use the
convexity and closedness of values of F, an upper semicontinuity of F (t, ·) and
some standard procedures based on the Mazur lemma.
Step 4. From Step 3 it follows that sup{d(v, Sm ); v ∈ Sm n
} → 0 (an easy
proof by contradiction). Therefore sup{d(v, Sm ); v ∈ Sm } → 0, as well. Hence,
n
for every n ≥ 1.
Fix ȳ ∈ Smn . We divide the interval [0, 1] on m parts, so we have 0 < 1 <
m
2 1
m < · · · < m−1
m < 1. Let r ∈ (0, m ]. We consider the problem:
⎧
⎨ ẏ(t) = A(t)y(t) + gn (t, y t ), for a.e. t ∈ [tm − mr(tm − tm−1 ), tm ],
y(t) = ȳ(t), for t ∈ [−τ, tm − mr(tm − tm−1 )], (10)
⎩
y(tm−1 ) = ȳ(tm−1 ) + Im−1 (ȳtm−1 ).
Here gn is a measurable—locally Lipschitz selection of Gn from Step 2.
m
Let ỹn,r denote the unique solution of this problem. Then the function
m
yn,r defined as:
m ȳ(t), t ∈ [0, tm − mr(tm − tm−1 )],
yn,r (t) = m
ỹn,r (t), t ∈ (tm − mr(tm − tm−1 ), tm ],
m
satisfies yn,r ∈ Sm
n.
1 2
Next for r ∈ ( m , m ] we consider the problem:
⎧
⎪
⎪ ẏ(t) = A(t)y(t) + gn (t, yt ),
⎪
⎪ 1
⎨ for a.e. t ∈ [tm−1 − m(r − m )(tm−1 − tm−2 ), tm ],
1
y(t) = ȳ(t), for t ∈ [−τ, tm−1 − m(r − m )(tm−1 − tm−2 )], (11)
⎪
⎪
⎪
⎪ y(t+ ) = y(tk ) + Ik (ytk ), k = m − 1,
⎩ k
y(tm−2 ) = ȳ(tm−2 ) + Im−2 (ȳtm−2 ).
m−1
Let ỹn,r denote the unique solution of this problem. Then we have yn,r m−1
∈
n
Sm , where:
1
m−1 ȳ(t), t ∈ [0, tm−1 − m(r − m )(tm−1 − tm−2 )],
yn,r (t) = 1
ỹn,r (t), t ∈ (tm−1 − m(r − m )(tm−1 − tm−2 ), tm ].
m−1
1 1
Let ỹn,r denote the unique solution of this problem. Then the function yn,r
defined as:
1 ȳ(t), t ∈ [0, t1 − m(r − m−1
m )t1 ],
yn,r (t) = 1
ỹn,r (t), t ∈ (t1 − m(r − m−1
m )t1 , tm ],
n.
also belongs to Sm
Finally we consider the following function hn : [0, 1] × Sm
n → Sn :
m
⎧
⎪
⎪ ȳ, r = 0,
⎪
⎪ m
r ∈ (0, m1
⎪ y
⎨ n,r , ],
1 2
hn (r, ȳ) = y m−1
n,r , r ∈ ( m m ],
, (13)
⎪
⎪ ..
⎪
⎪ .
⎪
⎩ 1
yn,r , r ∈ ( m−1
m , 1].
1
Here the functions yn,rm m−1
, yn,r , . . . , yn,r are determined by the choice of ȳ ∈ Sm
n.
One can show that the function hn is continuous applying a standard method,
which use a continuous dependence on initial conditions, and remembering
that the maps Ik are continuous, when checking a continuity in r ∈ { mi ; i =
1, . . . , m − 1}. The function hn , as continuous on [0, 1] × Sm n , is a homotopy.
1 n
By definition we have hn (0, ȳ) = ȳ and hn (1, ȳ) = yn,1 , so Sm is a contractible
set for every n ∈ N. Therefore, from Theorem 2.2 the set Sm is an Rδ -set.
Theorem 3.5 enables us to examine a structure of the solution set on the
half-line. The assumption (I2) will be replaced by the following:
(I2)∞
m
1
rk < ,
Bm
k=1
∞
→ E, k ∈ N, are continuous and satisfy (I1 ) and (I2 ) . Then the solution set
for problem (1) is an Rδ -set in P C([0, ∞), E)[x].
Proof. We have proved in Theorem 3.5, that solution sets on compact intervals
are Rδ sets (that is, for problem (2)). Next we consider an inverse system like
in the proof of Theorem 3.4. Using Theorem 2.5 we obtain that the solution
set of problem (1) is an Rδ -set.
Acknowledgments
The authors are indebted to the referee for his valuable comments and remarks
on some very recent papers connected with the material presented above. With
his helpful suggestions the paper has become more complete and familiar for
the reader.
Open Access. This article is distributed under the terms of the Creative Commons
Attribution Noncommercial License which permits any noncommercial use, distrib-
ution, and reproduction in any medium, provided the original author(s) and source
are credited.
References
[1] Milman, V.D., Myshkis, A.: On the stability of motion in the presence of
impulses. Sib. Math. J. 1, 233–237 (1960, in Russian)
[2] Benedetti, I., Rubbioni, P.: Existence of solutions on compact and non-com-
pact intervals for semilinear impulsive differential inclusions with delay. Topol.
Methods Nonlinear Anal. 32, 227–245 (2008)
[3] Cardinali, T., Rubbioni, P.: On the existence of mild solutions of semilinear
evolution differential inclusions. J. Math. Anal. Appl. 308, 620–635 (2005)
[4] Djebali, S., Górniewicz, L., Ouahab, A.: Filippov-Ważewski theorems and struc-
ture of solution sets for first order impulsive semilinear functional differential
inclusions. Topol. Methods Nonlinear Anal. 32, 261–312 (2008)
[5] Djebali, S., Górniewicz, L., Ouahab, A.: Topological structure of solution sets
for impulsive differential inclusions in Fréchet spaces. Nonlinear Anal. 74,
2141–2169 (2011)
Structure of the solution set
[7] Gabor, G.: Acyclicity of solution sets of inclusions in metric spaces. Topol. Meth-
ods Nonlinear Anal. 14, 327–343 (1999)
[8] Andres, J., Gabor, G., Górniewicz, L.: Topological structure of solution sets to
multivalued asymptotic problems. Z. Anal. Anwendungen 19(1), 35–60 (2000)
[9] Andres, J., Pavlačková, M.: Topological structure of solution sets to asymptotic
boundary value problems. J. Differ. Equ. 248, 127–150 (2010)
[10] Kamenskii, M., Obukhovskii, V., Zecca, P.: Condensing Multivalued Maps and
Semilinear Differential Inclusions in Banach Spaces. De Gruyter Ser. Nonlinear
Anal. Appl., vol. 7. Walter de Gruyter, Berlin-New York (2001)
[11] Guedda, L.: Some remarks in the study of impulsive differential equations and
inclusions with delay. Fixed Point Theory 12(2), 349–354 (2011)
[12] Benedetti, I., Obukhovskii, V., Zecca, P.: Controllability for impulsive semilinear
functional differential inclusions with a non-compact evolution operator. Dis-
cussiones Math. Differ. Inclusions Control Optim. 31(1), 39–69 (2011)
Agata Grudzka
e-mail: agata33@mat.uni.torun.pl