Electronic Journal of Differential Equations, Vol. 2004(2004), No. 31, pp. 114.

ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu

ftp ejde.math.txstate.edu (login: ftp)

N-TH ORDER IMPULSIVE INTEGRO-DIFFERENTIAL

EQUATIONS IN BANACH SPACES
MANFENG HU & JIANG ZHU

Abstract. We investigate the maximal and minimal solutions of initial value

problem for N-th order nonlinear impulsive integro-differential equation in Banach space by establishing a comparison result and using the upper and lower
solutions methods.

1. Introduction
The theory of impulsive differential equations in Banach spaces has become an
important area of investigation in recent years. In [2], the existence of solution of
initial value problem for second order nonlinear impulsive integro-differential equation in Banach space was studied by establishing a comparison result and using
the upper and lower solutions methods. Now, in this paper, we shall investigate
the existence of solution of initial-value problem (IVP) for N-th order nonlinear
impulsive integro-differential equation in Banach space by establishing a new comparison result and using the upper and lower solutions methods. Consider the IVP
for impulsive integro-differential equation in a Banach space E:
u(n) = f (t, u(t), u0 (t), . . . , u(n1) (t), (T u)(t)),
u|t=ti =

t J, t 6= ti

L0i u(n1) (ti )

u0 |t=ti = L1i u(n1) (ti )

...

(1.1)

u
|t=ti = Ln2
u(n1) (ti )
i
u(n1) |t=ti = Ln1
u(n1) (ti )
i
(n2)

u(0) = u0 , u0 (0) = u1 , . . . , u(n1) (0) = un1

where i = 1, 2, . . . , m, J = [0, a](a > 0), uj E(j = 0, 1, 2, . . . , n 1), f
C[J E E E, E], 0 < t1 < < ti < < tm < a, Lji (i = 1, 2, . . . , m;
j = 0, 1, . . . , n 1) are constants, and
Z t
(T u)(t) =
k(t, s)u(s)ds, t J ,
(1.2)
0

2000 Mathematics Subject Classification. 45J05, 47H07, 34A12.

Key words and phrases. Integro-differential equations in Banach spaces, cone,
partial ordering, monotone iterative technique.
c
2004
Texas State University - San Marcos.
Submitted August 14, 2003. Published March 3, 2004.
1

MANFENG HU & JIANG ZHU

EJDE-2004/31

k C[D, R+ ], D = {(t, s) J J : t s}, R+ is the set of all nonnegative real

numbers, and k0 = max{k(t, s) : (t, s) D}. u(j) |t=ti denotes the jump of u(j) (t)
at t = ti , i.e.

(j)
u(j) t=ti = u(j) (t+
i ) u (ti ),
(j)
where u(j) (t+
i ) and u (ti ) represent the right-hand limit and left-hand limit of
(j)
u (t) at t = ti respectively. In (1.1) and the following, u(n1) (ti ) is understood as
u(n1) (t
i ).
Let P C[J, E] = {u : u is a map from J into E such that u(t) is continuous at
j
t 6= ti , left continuous at t = ti and u(t+
i ) exist for i = 1, 2, . . . , m}, P C [J, E] =
j1
(j)
{u P C
[J, E]: u (t) is continuous at t 6= ti , left continuous at t = ti and
n1
u(j) (t+
[J, E] = {u
i ) exist for i = 1, 2, . . . , m}(j = 1, 2, . . . , n 2) and P C
n2
(n1)
(n1) +
(n1)
PC
[J, E] : u
(t) is continuous at t 6= ti , and u
(ti ),u
(ti ) exist for
i = 1, 2, . . . , m}. Evidently, P C[J, E] is a Banach space with norm

kukpc = sup ku(t)k.

tJ

It is clear that P C j [J, E] is a Banach space with norm

kukj = max{kukpc , ku0 kpc , . . . , ku(j) kpc },

(j = 1, 2, . . . , n 1)

Let J 0 = J\{t1 , . . . , tm }, = max{ti ti1 : i = 1, 2, . . . , m + 1}, (where t0 =

0, tm+1 = a), J0 = [0,Tt1 ], J1 = (t1 , t2 ], . . . , Jm1 = (tm1 , tm ], Jm = (tm , a]. A
map u P C n1 [J, E] C n [J 0 , E] is called a solution of (1.1) if it satisfies (1.1).
2. Comparison Result
Let E be partially ordered by a cone P of E, i.e. x y if and only if y x P .
P is said to be normal if there exists a positive constant N such that x y
implies kxk N kyk, where denotes the zero element of E, and P is said to be
regular if x1 x2 xn y implies kxn xk 0 as n for some
x E. It is well known that the regularity of P implies the normality of P . For
details on cone theory, see [1].
T
Lemma 2.1 (Comparison result). Assume that p P C n1 [J, E] C n [J 0 , E] satisfies
p(n) (t) M0 p M1 p0 M2 p00 Mn1 p(n1) N T p,

t J, t 6= ti

p|t=ti = L0i p(n1) (ti )

p0 |t=ti = L1i p(n1) (ti )
...
p(n2) |t=ti = Ln2
p(n1) (ti )
i
p(n1) |t=ti Ln1
p(n1) (ti ),
i
p

(n1)

(0) p

(j)

(0) ,

(i = 1, 2, . . . , m)

j = 0, 1, , n 2,
(2.1)

where Mj 0, Lji 0 (j = 0, 1, . . . , n 1; i = 1, 2, . . . , m) are constants and

m
X
i=1

Ln1
+ (m + 1)M0 1
i

(2.2)

EJDE-2004/31

N-TH ORDER IMPULSIVE INTEGRO-DIFFERENTIAL EQUATIONS

where
M0 = Mn1 +

n2
X

c?i + k1? a +

i=0

ci =

n2
X

(Mj

j=0

m
X

n2
X

Lji ) +

i=1

(d?j

j=0

m
X

Lji )a

(2.3)

i=1

X aij
ai+1
Mj +
N k0 , i = 0, 1, . . . , n 2 ,
(i j)!
(i + 1)!
ji

an2
an3
k0 N an1
k1 =
M0 +
M1 + + Mn2 +
,
(n 2)!
(n 3)!
(n 1)!
d0 = N k0 ,
d?1 = N k0 a + M0 ,
N k0 aj
M0 aj1
+
+ + Mj2 a + Mj1 , j = 2, 3, . . . , n 2,
j!
(j 1)!

dj =

then p(j) (t) , t J, (j = 0, 1, . . . , n 1), where p(0) (t) = p(t).

T
Proof. Let p1 (t) = p(n1) (t), t J. Then p1 P C[J, E] C 1 [J 0 , E] and
Z t
X
(n2)
p(n2) (t) = p(n2) (0) +
p1 (s)ds +
[p(n2) (t+
(ti )],
i )p
0

(n3)

(t) = p

(n3)

0<ti <t

(0) + tp

(n2)

Z
(0) +

Z
ds1

p1 (s2 )ds2

0
t

Z
+

s1

(n2)
[p(n2) (t+
(ti )]ds
i )p

0 0<t <s
i

(n3)
[p(n3) (t+
(ti )]
i )p

0<ti <t

...
tn3 (n2)
p
(0)
(n 3)!

p0 (t) = p0 (0) + tp00 (0) + +

Z

s1

ds1
0

Z
+

sn3

ds2 . . .
0

Z
ds1

p1 (sn2 )dsn2
0

s1

Z
ds2 . . .

+ +

sn4

(n2)
[p(n2) (t+
(ti )]dsn3
i )p

0<ti <sn3

00
[p00 (t+
i ) p (ti )]ds +

0 0<t <s
i

0
[p0 (t+
i ) p (ti )]

0<ti <t

tn2 (n2)
p
(0)
(n 2)!
Z t
Z s1
Z sn2
+
ds1
ds2 . . .
p1 (sn1 )dsn1
0
0
0
Z t
Z s1
Z sn3
X
(n2)
+
ds1
ds2 . . .
[p(n2) (t+
(ti )]dsn2
i )p

p(t) = p(0) + tp0 (0) + +

0<ti <sn2

MANFENG HU & JIANG ZHU

t

Z
+ +

0
[p0 (t+
i ) p (ti )]ds +

0 0<t <s
i

EJDE-2004/31

[p(t+
i ) p(ti )] .

0<ti <t

It is easy to see by induction that for m = 1, 2, . . . , n 1,

Z

s1

sm1

A(sm )dsm =

ds2 . . .

ds1
0

1
(m 1)!

(t s)m1 A(s)ds,

where A denotes an integrable function. So, we have

p(n2) (t) = p(n2) (0) +

p(n1) (t) = p1 (t)

t
X
(n2)
p1 (s)ds +
[p(n2) (t+
(ti )]
i )p

0<ti <t

Z t
p(n3) (t) = p(n3) (0) + tp(n2) (0) +
(t s)p1 (s)ds
0
Z t X
(n2)
+
[p(n2) (t+
(ti )]ds
i )p
0 0<t <s
i

(n3)
[p(n3) (t+
(ti )]
i )p

0<ti <t

...
Z t
t
1
p(n2) (0) +
(t s)n3 p1 (s)ds
(n 3)!
(n 3)! 0
Z t
X
(n2)
(t s)n4
[p(n2) (t+
(ti )]ds
i )p
n3

p0 (t) = p0 (0) + tp00 (0) + +

1
(n 4)! 0
0<ti <s
Z t X
X
0
00
+ +
[p00 (t+
[p0 (t+
i ) p (ti )]ds +
i ) p (ti )]
+

0 0<t <s
i

0<ti <t

Z t
1
t
(n2)
p(t) = p(0) + tp (0) + +
p
(0) +
(t s)n2 p1 (s)ds
(n 2)!
(n 2)! 0
Z t
X
1
(n2)
+
(t s)n3
[p(n2) (t+
(ti )]ds
i )p
(n 3)! 0
0<ti <s
Z t X
X

0
+ +
[p0 (t+
[p(t+
i ) p (ti )]ds +
i ) p(ti )]
n2

0 0<t <s
i

0<ti <t

(2.4)
Substituting (2.4) into (2.1), we get
p01 (t) Mn1 p1 (t) c0 (t)p(0) c1 (t)p0 (0) cn2 (t)p(n2) (0)
Z t
n2
X
X
(j)

k1 (t, s)p1 (s)ds

Mj
[p(j) (t+
i ) p (ti )]
0

j=0

n2
XZ t
j=0

dj (t, s)

X
0<ti <s

0<ti <t

(j)
[p(j) (t+
i ) p (ti )]ds,

t J

(2.5)

EJDE-2004/31

N-TH ORDER IMPULSIVE INTEGRO-DIFFERENTIAL EQUATIONS

Where
Z

c0 (t) = M0 + N

k(t, s)ds,
Z t
c1 (t) = tM0 + M1 + N
sk(t, s)ds,
0

...
Z t
tn2
tn3
N
M0 +
M1 + + Mn2 +
sn2 k(t, s)ds,
(n 2)!
(n 3)!
(n 2)! 0
(t s)n2
(t s)n3
k1 (t, s) =
M0 +
M1
(n 2)!
(n 3)!
Z t
N
+ + Mn2 +
(r s)n2 k(t, r)dr,
(n 2)! s
d0 (t, s) = N k(t, s),
Z t
d1 (t, s) = N
k(t, r)dr + M0 ,
s
Z
N t
d2 (t, s) =
k(t, r)(r s)dr + M0 (t s) + M1 ,
1! s
...
Z t
N
(t s)n3
dn2 (t, s) =
k(t, r)(r s)n3 dr + M0
(n 3)! s
(n 3)!
+ + Mn4 (t s) + Mn3 .
T
For g P , the dual cone of P , let v(t) = g(p1 (t)), then v P C[J, R] C 0 [J 0 , R].
By (2.5) and (2.1), we have
cn2 (t) =

v (t) Mn1 v(t)

n2
X

cj (t)g(p

(j)

Z
(0))

Mj

n2
XZ t
j=0

g(p

(j)

(t+
i )

(j)
g(p(j) (t+
i ) p (ti ))

0<ti <t

j=0

k1 (t, s)v(s)ds
0

j=0
n2
X

dj (t, s)

(j)
g(p(j) (t+
i ) p (ti ))ds,

0<ti <s

(ti )) = Lji v(ti ) (j =

v|t=ti Ln1
v(ti )
i

(j)

v(0) g(p

(2.6)
t J

(j)

(0)) 0,

0, 1, 2, . . . , n 2; i = 1, 2, . . . , m)
(i = 1, 2, . . . , m)
j = 0, 1, , n 2,

We now show that

v(t) 0,

t J.

(2.7)

Assume that (2.7) is not true, i.e. there exists a 0 < t a such that v(t ) > 0.
Let t Jj = (tj , tj+1 ] and inf 0tt v(t) = . We have 0. Assume that
there exist a Jk = (tk , tk+1 ](k j) such that v(t ) = hold for some t Jk or
+
v(t+
k ) = . We may assume that v(t ) = , since for the case v(tk ) = the

MANFENG HU & JIANG ZHU

EJDE-2004/31

proof is similar. By (2.6), we have

0

v (t) [Mn1 +

n2
X

cj (t) +

j=0

M0 ,

k1 (t, s)ds +
0

j=0
n2
XZ

(dj (t, s)

n2
X

(Mj

j=0

m
X

Lji )

i=1

(2.8)

Lji ds]

0<ti <s

t [0, t ],

where M0 is given by (2.3),

v|t=ti Ln1
v(ti ) Ln1
i
i

(i = 1, 2, . . . , m)

(2.9)

v(0) 0
Now, the mean value theorem implies
0

v(t ) v(t+
j ) = v (j )(t tj ),
0
v(tj ) v(t+
j1 ) = v (j1 )(tj tj1 ),

tj < j < t ;
tj1 < j1 < tj ;

...
v(tk+2 )

v(t+
k+1 )

(2.10)

k+2

= v (k+1 )(tk+2 tk+1 ),

0

v(tk+1 ) v(t ) = v (k )(tk+1 t ),

tk+1 < k+1 < t

t < k < tk+1 .

By (2.9) we have
n1
v(t+
,
i ) = v(ti ) + v|t=ti v(ti ) + Li

ti t .

(2.11)

By (2.8), (2.10), (2.11), we obtain

v(t ) v(tj ) Ln1
M0
j
v(tj ) v(tj1 ) Ln1
j1 M0
...
v(tk+2 ) v(tk+1 )

(2.12)
Ln1
k+1

M0

v(tk+1 ) + M0 .
Adding these inequalities, we obtain
v(t ) +

j
X

Ln1
(j k + 1)M0
i

i=k+1

and so
0 < v(t ) +

m
X

Ln1
+ (m + 1)M0
i

i=1

Evidently 6= 0, so, > 0, then, we have

1<

m
X

Ln1
+ (m + 1)M0 ,
i

i=1

which contradicts (2.2), hence v(t) 0, t J. Since g P is arbitrary, it implies

that p(n1) (t) for all t J. By p(n2) (0) , p(n2) |t=ti = Ln2
p(n1) (ti )
i
; this implies p(n2) (t) for all t J. Continuing in this manner, p(i) (t)
for all t J, i = 0, 1, . . . , n 3. The proof is complete.


EJDE-2004/31

N-TH ORDER IMPULSIVE INTEGRO-DIFFERENTIAL EQUATIONS

Lemma 2.2. Assume P C[J, E], and Mj , NT

, Lji (j = 0, 1, 2, . . . , n 1; i =
n1
1, 2, . . . , m) are constants, then u P C
[J, E] C n [J 0 , E] is a solution of the
linear IVP
u(n) =

n1
X

Mj u(j) N T u + (t),

t J, t 6= ti

j=0

u(j) |t=ti = Lji u(n1) (ti ),

(j = 0, 1, . . . , n 2)

u(n1) |t=ti = Ln1

u(n1) (ti ),
i
0

u(0) = u0 , u (0) = u1 , . . . , u

(2.13)

(i = 1, 2, . . . , m)

(n1)

(0) = un1 .

if and only if u P C n1 [J, E] is a solution of the linear impulsive integral equation

u(t) = u0 + tu1 +
+
+

1
(n 1)!
X

t2
tn1
u2 + +
un1
2!
(n 1)!
Z t
n1
X
(t s)n1 [
Mj u(j) (s) N (T u)(s) + (s)]ds
0

(2.14)

j=0
n1

0<ti <t

n2

(t ti )
(t ti )
Ln1 +
Ln2
(n 1)! i
(n 2)! i

+ + (t ti )L1i + L0i ]u(n1) (ti )

t J .

The proof of this lemma is similar to the proof of Lemma 3 in [2]; therefore, we
omit it.
Lemma 2.3. Let P C[J, E], Mj 0, N 0, Lji 0 (j = 0, 1, 2, . . . , n 1; i =
1, 2, . . . , m) be constants. Assume
Pn1
m
Mi + N k0 a nj X (a ti )nj1 n1
j = i=0
a
+
[
L
(n j)!
(n j 1)! i
i=1
(2.15)
+ + (a ti )Lj+1
+ Lji ] < 1
i
= max{j }
j

where j = 0, 1, . . . , n 1). Then the impulsive integral equation (2.14) has a unique
solution in P C n1 [J, E].
Proof. Define operator F by
t2
tn1
u2 + +
un1
2!
(n 1)!
Z t
n1
X
1
+
(t s)n1 [
Mj u(j) (s) N (T u)(s) + (s)]ds
(n 1)! 0
j=0

(F u)(t) =u0 + tu1 +

0<ti <t

(t ti )n1 n1 (t ti )n2 n2
L
+
L
(n 1)! i
(n 2)! i

+ + (t ti )L1i + L0i ]u(n1) (ti )

(t J).

MANFENG HU & JIANG ZHU

EJDE-2004/31

Then for all t J 0 , j = 1, 2, . . . , n 1,

(F u)(j) (t) = uj + tuj+1 + +
1
+
(n j 1)!

0<ti

and F : P C

(t s)nj1 [

n1
X

Mj u(j) (s) N (T u)(s) + (s)]ds

j=0

(t ti )nj1 n1
[
Li
+ + (t ti )Lj+1
+ Lji ]u(n1) (ti )
i
(n

1)!
<t

+
n1

tnj1
un1
(n j 1)!

[J, E] P C n1 [J, E]. For u, v P C n1 [J, E], by (2.12) we have

k(F u)(j) (t) (F v)(j) (t)k

Pn1
Z t
m
X
Mi + N k0 a
(t ti )nj1 n1
ku vkn1
(t s)nj1 +
L
i=0
[
(n j 1)!
(n j 1)! i
0
i=1
(t ti )nj2 n2
L
+ + (t ti )Lj+1
+ Lji ]ku vkn1
i
(n j 2)! i
j ku vkn1 (t J, j = 0, 1, . . . , n 1)
+

and
kF u F vkn1 ku vkn1 ,

u, v P C n1 [J, E]

(2.16)

where j , is defined by (2.15), The Banach fixed point implies that F has a unique
fixed point in P C n1 [J, E], and the lemma is proved.

3. Main Theorem
Let us list some conditions used for stating the main result.
T
(H1) There exist v0 ,w0 P C n1 [J, E] C n [J 0 , E] with v0 (t) w0 (t)(t J)
such that
(n)

v0

(j)
v0 |t=ti

(n1)

f (t, v0 , v00 , . . . , v0

(n1)
Lji v0
(ti ),

(n1)
(0)
u j , v0

t J, t 6= ti

(j = 0, 1, . . . , n 2; i = 1, 2, . . . , m)

(n1)
|t=ti
v0
(j)
v0 (0)

, T v0 ),

(j)
v0 (0)

(n1)

Ln1
v0
i

(ti )

un1 uj

(j = 0, 1, 2, . . . , n 1),

and
(n)

w0
(j)

(n1)

f (t, w0 , w00 , . . . , w0
(n1)

w0 |t=ti = Lji w0

(ti ),

(n1)
w0
|t=ti
(j)
w0 (0)

uj ,

(n1)
w0
(0)

, T w0 ),

t J, t 6= ti

(j = 0, 1, . . . , n 2; i = 1, 2, . . . , m)
(n1)

Ln1
w0
i

(j)
w0 (0)

un1 uj ,

(ti )
(j = 0, 1, 2, . . . , n 1),

where Lji 0, (i = 1, 2, . . . , m; j = 0, 1, . . . , n 1), v0 and w0 are lower and

upper solution of (1.1) respectively.
(H2) There exist constants Mi 0 (i = 0, 1, . . . , n 1) and N 0 such that
f (t, u0 , u1 , u2 , . . . , un1 , v) f (t, u
0 , u
1 , u
2 , . . . , u
n1 , v)

EJDE-2004/31

N-TH ORDER IMPULSIVE INTEGRO-DIFFERENTIAL EQUATIONS

n1
X

Mj (uj u
j ) N (v v),

t J

j=0
(j)

(j)

v0 u
j uj w0 , (j = 0, 1, . . . , n 1)
T v0 v v T w0
(j)

(j)

Let [v0 , w0 ] = {u P C n1 [J, E] : v0 (t) u(j) (t) w0 (t), t J, j = 0, 1, . . . , n

1}
Theorem 3.1. Let cone P be regular and f be uniformly continuous on J Br
Br Br for any r > 0, where Br = {x E : kxk r}. Suppose that
conditions (H1) and (H2) are satisfied, Lji 0(j = 0, 1, . . . , n 1; i = 1, 2, . . . , m)
and inequalities (2.2), (2.15) hold. Then (1.1) has minimal and maximal solutions
u
and u in [v0 , w0 ]; Moreover, there exist monotone sequences {vk (t)} and {wk (t)}
(j)
(j)
such that {vk (t)},{wk (t)}(j = 0, 1, 2, . . . , n 1) converge uniformly on Jj (j =
(j)
0, 1, . . . , m) to the u
(t) and (u )(j) (t)(j = 0, 1, 2, . . . , n 1) respectively, and
(j)

(j)

(j)

(j) (t)
v0 (t) v1 (t) vk (t) u
(j)

(j)

(j)

u(j) (t) (u )(j) (t) wk (t) w1 (t) w0 (t)

(3.1)

for all t J, j = 0, 1, . . . , n 1, where u(t) is any solution of (1.1) in [v0 , w0 ].

Proof. For [v0 , 0 ], consider the linear problem (2.13) with
(t) = f (t, (t), 0 (t), . . . , (n1) (t), (T )(t)) +

n1
X

Mj (j) (t) + N (T )(t)

(3.2)

j=0

By Lemma 2.3, (2.13) has aTunique solution u P C n1 [J, E]. Let u = A. Then
A : [v0 , w0 ] P C n1 [J, E] C n [J 0 , E] P C[J, E], we now show that
(j)

(a) v0 (t) (Av0 )(j) (t), (Aw0 )(j) (t) (w0 )(i) (t), t J, j = 0, 1, 2, . . . , n 1
(j)
(j)
(b) 1 , 2 [v0 , w0 ], 1 2 implies (A1 )(j) (A2 )(j) , t J, j =
0, 1, 2, . . . , n 1.
To prove (a), we set v1 = Av0 and p = v0 v1 . From (2.13) and (3.2), we have
(n)

v1

(n1)

= f (t, v0 , v00 , . . . , v0

, T v0 ) +

n1
X

(j)

Mj v0 + N (T v0 )

j=0

n1
X

(j)

Mj v1 N (T v1 ),

t J, t 6= ti

j=0
(j)

(n1)

v1 |t=ti = Lji v1
(n1)
v1
|t=ti

v1 (0) = u0 ,

(ti ),

(j = 0, 1, . . . , n 2)

(n1)
Ln1
v1
(ti ),
i

v10 (0)

(i = 1, 2, . . . , m)

(n1)
u1 , . . . , v1
(0)

= un1

so, by (H1),
p(n) (t)

n1
X

Mj p(j) (t) N (T p)(t),

t J, t 6= ti ,

j=0

p(j) |t=ti = Lji p(n1) (ti ),

(j = 0, 1, . . . , n 2; i = 1, 2, . . . , m),

10

MANFENG HU & JIANG ZHU

p(n1) |t=ti Ln1

p(n1) (ti ),
i
p

(n1)

(0) p

(j)

(0) ,

EJDE-2004/31

(i = 1, 2, . . . , m),

(j = 0, 1, 2, . . . , n 2),

which implies by virtue of Lemma 2.1 that p(j) (t) (j = 0, 1, . . . , n 1) for t J,

(j)
i.e. v0 (t) (Av0 )(j) (t), for all t J, j = 0, 1, 2, . . . , n 1. Similarly, we can show
that (Aw0 )(j) (t) (w0 )(j) (t) for all t J, j = 0, 1, 2, . . . , n 1.
(j)
(j)
To prove (b), let 1 , 2 [v0 , w0 ], such that 1 2 and p = A1 A2 .
Then, from (2.13) and (H2), we have

p(n) (t)

n1
X

Mj p(j) (t) N (T p)(t),

t J, t 6= ti ,

j=0

p(j) |t=ti = Lji p(n1) (ti ),

p

(n1)

|t=ti =

(j)

(j = 0, 1, . . . , n 2; i = 1, 2, . . . , m),

Ln1
p(n1) (ti ),
i

(i = 1, 2, . . . , m),

(j = 0, 1, 2, . . . , n 1).

(0) = ,

So, Lemma 2.1 implies (b). Let

vk = Avk1 ,

wk = Awk1 ,

k = 1, 2, . . . ,

(3.3)

By (a) and (b) above, we have

(j)

(j)

(j)

(j)

(j)

(j)

v0 (t) v1 (t) vk (t) wk (t) w1 (t) w0 (t),

(3.4)

for all t J, j = 0, 1, 2, . . . , n 1. On account of the definition of vk , we have

vk (t) = u0 + tu1 +
+
+

1
(n 1)!
X

tn1
t2
u2 + +
un1
2!
(n 1)!
Z t
n1
X
(j)
(t s)n1 [
Mj vk (s) N (T vk )(s) + k1 (s)]ds
0

(3.5)

j=0
n1

0<ti <t

n2

(t ti )
(t ti )
Ln1 +
Ln2
(n 1)! i
(n 2)! i
(n1)

+ + (t ti )L1i + L0i ]vk

(ti ),

(t J, k = 1, 2, 3, . . . )

where
(n1)

0
k1 (t) = f (t, vk1 (t), vk1
(t), . . . , vk1 (t), (T vk1 )(t))

n1
X
j=0

(j)

Mj vk1 (t) + N (T vk1 )(t),

t J,

k = 1, 2, 3, . . .

(3.6)

EJDE-2004/31

N-TH ORDER IMPULSIVE INTEGRO-DIFFERENTIAL EQUATIONS

11

so,
(j)

vk (t)
= uj + tuj+1 + +
+
+

1
(n j 1)!
X

tnj1
un1
(n j 1)!

(t s)nj1 [

n1
X

(j)

Mj vk (s) N (T vk )(s) + k1 (s)]ds

j=0
nj1

0<ti <t

(t ti )
(t ti )nj2 n2
Ln1
+
L
i
(n j 1)!
(n j 2)! i
(n1)

+ + (t ti )Lj+1
+ Lji ]vk
i

(ti ),

(3.7)
for all t J 0 , j = 1, 2, . . . , n 1, k = 1, 2, 3, . . . . Similar to the (2.16), for k, i =
1, 2, . . . , we can obtain
kvk+i vk kn1 kvk+1 vk kn1 + kk+i1 k1 kpc ,
where is defined by (2.15) and
= max{

an an1
a2
,
, . . . , , a} .
n! (n 1)!
2

(3.8)

Hence, for k, i = 1, 2, . . . ,
kvk+i vk kn1

kk+i1 k1 kpc .
1

(3.9)

Since the regularity of P implies the normality of P , we see from (3.4) that Vj =
(j)
{vk : k = 0, 1, 2, . . . } (j = 0, 1, . . . , n 1) is a bounded set in P C j [J, E]. It is easy
to show that the uniform continuity of f on J Br Br Br implies the
boundedness of f on J Br Br Br , so by (3.6),there is a constant b > 0
such that
kk1 kpc b (k = 1, 2, . . . )
(j)

and therefore, from (3.7) we know that functions {vk (t)} (j = 0, 1, . . . , n 2) are
equicontinuous on each Ji (i = 0, 1, . . . , m). From (3.4) and the regularity of P , we
(j)
can infer that {vk (t)} converges uniformly to u
(j) (t) P C[J, E] in J; i.e.,
(j)

kvk u
(j) kpc 0

(k )

(3.10)

From (3.6),(3.10) and the uniform continuity of f on J Br Br Br , we

get
kk1
kpc 0 (k )
where

(t) = f (t, u
(t), u
0 (t), . . . , u
(n1) (t), (T u
)(t)) +

n1
X
j=0

for all t J. Taking limits in (3.5),we obtain

u
(t) = u0 + tu1 +

t2
tn1
u2 + +
un1
2!
(n 1)!

Mi u
(j) (t) + N (T u
)(t),

12

MANFENG HU & JIANG ZHU

+
+

1
(n 1)!
X

(t s)n1 [

n1
X

EJDE-2004/31

Mj u
(j) (s) N (T u
)(s) + k1 (s)]ds

j=0
n1

0<ti <t

(t ti )
(t ti )n2 n2
Ln1
+
L
i
(n 1)!
(n 2)! i

u(n1) (ti )
+ + (t ti )L1i + L0i ]
T
which by Lemma 2.2 implies u
P C n1 [J, E] C n [J 0 , E] and u
(t) is a solution of
(1.1).
In the same
we can show that kwk u kn1 0(k ) for some u
T way,
n1
n 0
PC
[J, E] C [J , E] and u (t)
T is a solution of (1.1).
Finally, let u P C n1 [J, E] C n [J 0 , E] be any solution of (1.1) in [v0 , w0 ]
(j)
(j)
satisfying v0 (t) u(j) (t) w0 (t), for all t J, j = 0, 1, . . . , n 1. Assume
(j)
(j)
that vm1 (t) u(j) (t) wm1 (t) for all t J, j = 0, 1, . . . , n 1. Then by
(j)
(j)
(H2) and Lemma 2.1, we can infer that vm (t) u(j) (t) wm (t) for all t J,
(j)
(j)
j = 0, 1, . . . , n 1. Hence, by induction, vk (t) u(j) (t) wk (t) for all t J,
(j)
j = 0, 1, . . . , n 1, k = 1, 2, . . . , which implies that u
(t) u(j) (t) (u )(j) (t)
for all t J, j = 0, 1, . . . , n 1. Hence (3.1) follows from (3.4). The proof is
complete.

Remark 3.2. The condition that P is regular will be satisfied if E is weakly
complete (reflexive, in particular) and P is normal (see [4, theorem 2]).
4. An Example
Consider the initial-value problem infinite system for scalar third order integrodifferential equations
u(3)
n =

1
[(t un )2 + t2 un+1 + (u02n )2 + (t u00n )2 ]
100n2
Z t
t
+
(t
ets un (s)ds)2 , 0 t 2, t 6= 1
800n3
0
1
un |t=1 = u00n (1)
3
1
0
un |t=1 = u00n (1)
2
1
00
un |t=1 = u00n (1)
60
un (0) = u0n (0) = u00n (0) = 0, n = 1, 2, . . .

(4.1)

Claim: The system (4.1) admits minimal and maximal solutions which are continuously differentiable on [0, 1) (1, 2] and satisfy
(
t3 /n2 ,
if 0 t 1
0 un (t)
3
2
2
(t + t + t)/n , if 1 < t 2
0

u0n (t)

(
3t2 /n2 ,
if 0 t 1

2
2
(3t + 2t + 1)/n , if 1 < t 2

EJDE-2004/31

N-TH ORDER IMPULSIVE INTEGRO-DIFFERENTIAL EQUATIONS

u00n (t)

13

(
6t/n2 ,
if 0 t 1

2
(6t + 2)/n , if 1 < t 2

where n = 1, 2, 3, . . . .
P
Proof. P
Let E = l1 = {u = (u1 , u2 , . . . , un , . . . ) : n=1 |un | < } with the norm

kuk = n=1 |un | and let P = {u = (u1 , u2 , . . . , un , . . . ) l1 : un 0, n = 1, 2, . . . }.

Then P is a normal cone in E. Since l1 is weakly complete, from remark 3.1 we
know that P is regular. System (4.1) can be regarded as an IVP of form (1.1),
where
a = 2, k(t, s) = ets ,
v = (v1 , v2 , . . . , vn , . . . ),
z = (z1 , z2 , . . . , zn , . . . ),

u = (u1 , u2 , . . . , un , . . . ),
w = (w1 , w2 , . . . , wn , . . . ),
f = (f1 , f2 , . . . , fn , . . . ),

in which
1
t
[(t un )2 + t2 un+1 + (v2n )2 + (t wn )2 ] +
(t zn )2
2
100n
800n3
(4.2)
1
and m = 1, t1 = 1, L01 = 31 , L11 = 12 , L21 = 60
, u0 = u1 = u2 = (0, 0, . . . , 0, . . . ).
Evidently, f C[J E E E E, E] (J = [0, 2]). Let v0 (t) = (0, 0, . . . , 0, . . . ),
for 0 t 2 and
(
(t3 , . . . , t3 /n2 , . . . ),
if 0 t 1;
w0 (t) =
3
2
+t
(t3 + t2 + t, . . . , t +t
,
.
.
.
),
if 1 < t 2.
2
n
fn (t, u, v, w, z) =

We have v0 C 3 [J, E], w0 P C 2 [J, E] C 3 [J 0 , E], where J 0 = J\{1} = [0, 1)

(1, 2], v0 (t) w0 (t) (t J) and
(
2
(3t2 , . . . , 3t
if 0 t 1
2 , . . . ),
0
n
w0 (t) =
3t2 +2t+1
2
(3t + 2t + 1, . . . ,
, . . . ),
if 1 < t 2
n2
(
(6t, . . . , n6t2 , . . . ),
if 0 t 1
w000 (t) =
(6t + 2, . . . , 6t+2
,
.
.
.
),
if 1 < t 2
n2
(3)

w0 = (6, . . . ,

6
, . . . ),
n2

0 t 2.

It is clear that
v00 (t) w00 (t), v000 (t) w000 (t),

t J

v0 (0) = w0 (0) = (0, 0, . . . , 0, . . . ) = u0 ,

v00 (0)
v000 (0)

v0 (0) = w00 (0) w0 (0) = u1 u0 = (0, 0, . . . , 0, . . . )

v00 (0) = w000 (0) w00 (0) = u2 u1 = (0, 0, . . . , 0, . . . )
(3)

t J
1
v0 |t=1 = (0, 0, . . . , 0, . . . ) = v000 (1)
3
1
v00 |t=1 = (0, 0, . . . , 0, . . . ) = v000 (1)
2
1
00
v0 |t=1 = (0, 0, . . . , 0, . . . ) = v000 (1)
60
v0 (t) = (0, 0, . . . , 0, . . . ),

14

MANFENG HU & JIANG ZHU

2
1
, . . . ) = w000 (1)
2
n
3
3
1
0
w0 |t=1 = (3, . . . , 2 , . . . ) = w000 (1)
n
2
2
1
00
w0 |t=1 = (2, . . . , 2 , . . . ) > w000 (1)
n
60
2t2
t3
(3)
0
00
fn (t, v0 (t), v0 (t), v0 (t), (T v0 )(t)) =
+
0 = v0 (t),
100n2
800n3
When 0 t 1,we have

EJDE-2004/31

w0 |t=1 = (2, . . . ,

t J.

fn (t, w0 (t), w00 (t), w000 (t), (T w0 )(t))

1 
t3
t3
3t2 2
6t 
=
(t 2 )2 + t2
+(
) + (t 2 )2
2
2
2
100n
n
(n + 1)
(2n)
n
Z t
3
t
s
+
(t
ets 2 ds)2
800n3
n
0
1
t5
9t4
t3
6
2
2.
(t +
+ 2 + t2 ) +

2
2
100n
(n + 1)
4n
800n3
n
When 0 < t 2, we have
fn (t, w0 (t), w00 (t), w000 (t), (T w0 )(t))
3
2
t3 + t 2 + t 2
3t2 + 2t + 1 2
6t + 2 2 
1 
2t + t + t
(t

)
+
t
+
(
) + (t
)
=
2
2
2
2
100n
n
(n + 1)
(2n)
n2
Z t
t
s3 + s2 + s 2
+
(t

ds)
ets
3
800n
n2
0
t5 + t4 + t3
(t2 + 2t + 1)2
t3
6
1
(t2 +
+
+ t2 ) +
2.

2
2
2
3
100n
(n + 1)
4n
800n
n
Hence v0 , w0 satisfy (H1). On the other hand, for t J
v0 (t) u u w0 (t),
v000 (t) w w w00 (t),

v00 (t) v v w00 (t),

(T v0 )(t) z u (T w0 )(t),

we have
t3 + t2 + t
3t2 + 2t + 1
,
0

,
n
n
n2
n2
6t + 2
3t4 + 4t3 + 6t2
0 wn wn
, 0 z n zn
,
2
n
12
n = 1, 2, . . . . Therefore, by (4.2),
0 un un

fn (t, u, v, w, z) fn (t, u, v, w, z)

1 

(t un )2 (t un )2 + (t wn )2 (t wn )2
2
100n

t 
+
(t zn )2 (t z n )2
3
800n
1
2t2
1

[2t(un un ) + 2t(wn wn ]
(zn z n ) (un un )
2
100n
800n3
25
1
1
(wn wn )
(zn z n ).
25
100

EJDE-2004/31

N-TH ORDER IMPULSIVE INTEGRO-DIFFERENTIAL EQUATIONS

15

Consequently, (H2) is satisfied for M0 = 1/25 = M2 , M1 = 0, N = 1/100. It is

clear that k0 = 1 and = 1, and it is easy to verify that inequalities (2.2) and
(2.15) hold. Hence, our conclusion follows from Theorem 3.1.

References
[1] Guo Dajun and Lakshmikantham.V., Nonlinear problems in abstract cones, Academic Press,
New York, (1998).
[2] Guo Dajun, Initial value problems for second impulsive integro-differential equations in Banach spaces. Chin. Ann. Of Math. 18B:4(1997), 439-448.
[3] Guo Dajun, Integro-differential equations on unbounded domains in Banach spaces, Chin.
Ann. Of Math. 20B:4(1999), 435-446.
[4] Du, Y., Fixed points of increasing operators in ordered Banach spaces and applications, Appl.
Anal. 38(1990), 1-20.
Manfeng Hu
Department of Science, Jiangnan University, Wuxi 214000, China, and Department of
Mathematics, Xuzhou Normal University, Xuzhou 221116, China
Jiang Zhu
Department of Mathematics, Xuzhou Normal University, Xuzhou 221116, China
E-mail address: [email protected]