Emlinear Analysrr. Theorv.

Prmted m Great Britain.

Methods & Applrcanonr.

Vol. 7, No.

12. pp

1335-1345,

19X3.
6

0362-546X, 83 $03.00 + .I111

1983 Pergamon
Pres Ltd

ON THE EXISTENCE
AND CONTINUOUS
DEPENDENCE
SOLUTIONS
OF A CLASS OF NEUTRAL FUNCTIONALDIFFERENTIAL
EQUATIONS
N.
Department

of Mathematics,

OF

PARHI

Berhampur

University.

(Received

1 Jmuary

Key words and phrases: Neutral functional-differential

local and global existence, continuous
dependence.

Berhampur-760007,

Orissa,

India

1983)

equations,

admissible

functions,

unique

solution

1. INTRODUCTION
IN [l], THE existence
problem for a class of neutral functional-differential
equations
discussed. The study in [l] is applicable to the following class of equations:
0

0
i(t)

i(f)

has been

K(r,

I --T

0)x,(6)

d6 +

I --r

K *(t, 0) I&(@ I do + h(f),

K(t, 19)x,(O) d6

K*(t,

K,(t,

(1.2)

d% 2 + h(t),
I

t9)ic(0)

0).x$(t3)

Cl.11

de 2.

(1.3)

A close took at the above class of equations reveals that the derivative or the absolute value
of the derivative
of the unknown
function under the integral sign occurs linearly. It is clear
from the norm used in [l] that the study there is not applicable to equations of the type
0

i(t)

K(t, 0)x,( 0) de +

= f

(1.4)

J--r

J-r

i(t)

K *(I,
@i;'(0)
d6 + h(t),

_T

K(t, 0)x,(O) de +

i K*(t,d)i@)

(1.5)

de.

--r

In this paper, we consider a norm which is suitable for all above

metric is not suitable for the following neutral equations:

equations.

However,

this

J?(t) = x(t - r)

(1.6)

x(t) = Ix(t - r) 1.

(1.7)

The study in [4] is applicable

to (1.6) but not to any one of the equations
(l.l)-(1.5)
and
(1.7).
Following notations are used throughout
this paper. E" denotes the Euclidean
n-space and
/x/ stands for any Euclidean
norm of X. R and R' denote intervals
(- x, 00) and [0, W)
respectively.
For t E R, Z,= [t- z,t],
T> 0,AC(Z) denotes the space of absolutely continuous
1335

1336

N. PARHI

functions x : I-, E, where I is a sub-interval

of R, U(Z), p > 1, denotes the space of Lebesgue
measurable
functions x : IE such that Ix(t) (p is integrable.
Qp(Z). p > 1, denotes the space
of x E AC(Z) such that k E Lp(Z). For x E !?(Z),
we define

where Z is a closed and bounded sub-interval

of R. For any sub-interval
[a, b] of R and for
of the restriction
of x to the
any x E QP([a - t, b]), r> 0, we denote x, as the translation
interval Z,, t E [a, b]. Hence xI E SP(Zo) for x E SP([a - r, b]) and t E [a, b]. It may be noted
that, for t E [a, b] and x E QP([a - t, b]), (/x~(~I~,
= I/x //I, and for t E [a, 61 and x E @([a, b]),
]x(t) / c IIxI/[d#
In Section 2, we prove a theorem concerning
local existence and uniqueness
and give some examples.
A theorem on continuous
dependence
of solutions
Section 3. Section 4 deals with global existence of solutions.

2. EXISTENCE

We need the following

LEMMA

1.

AND

UNIQUENESS

OF SOLUTIONS

lemma.

The space QP([u, b]) with the norm

The proof is easy and hence

(1.8) is complete.

is omitted.

Remark. For p = 1, the study here will be same as the study in [I]. Hence,
we take p > 1.
Let D be an open
differential
equations:

of solutions
is proved in

set in Qp(Za).

We consider

the following

in what follows,

class of neutral

functional-

(2.1)
where

f: (a, h) X D +

E.

Definition. Let c~E D and to E (u, h). A function x E Qp([to - r, to + (Y)), (Y> 0 such that
to + a< b, is said to be a solution of (2.1) on [to, to + a) with initial function p at to if
(i)
(ii)
in
(7
Let

xf E D, t E [&, & + a),

xrO= v, and
x ( t ) sa tis fi es ( 2.1) for t E [to, to + a).
to E (a, b) and q E D. So there exists r > 0 such that N&r) L D, where
Q(Zl)) :IIv - dlk < Y}.Let Q: [to - T, b) + E be defined as follows:
cprr,= Y

N,(r)

= {q!~E

(2.2)

cp(t) = V(O), t E [to, b).

Clearly,

@E W([to - z, b)).

In what

follows,

we show

that

there

exists

a /3, fi> 0, to +

Solutions of a class of neutral functional-differential

/3 < b, such that

1337

equations

@ E D, t E [to, to + /3]. For this it is sufficient

to show that //I$~- I&

< I,

t E [to, to + fi].
Now, for t E [to, to + t],
II cpt- cplll = ;:I;I

I cp(s)
- cp(s
- 0 I + (lLr I@@I- ciie-4 lPh) lip

sup
I Q?(s)SE[fP-T.IO]

Q7b - 0 I +

sup I Q(s) IE[[email protected]]

90 - 4

+ [~~~16(~)-~(~-t)l~~+~~~l~(S)-pl(S-r)l~ill]ly
=s

sup
M-to)-&--)I+
rz[r-r.ro]
+

[j:j@!(s-

sup MO)-q?(s-l)I
sE[to.r]

to) - @(s - t)

dy + [J-I
l@(s -

Ip

t) p

d.y

r--T

=s

I ds - to)- ds - 0 I + .,E[f().f]
sup /01(O)
- ds - 0 I
sEsup
[I- r.to]
/

@(u + t - to) - Q(u)

u+ t -

lpdu ]lb+

[~,~_,l()lpd]p

to) - c+(u) Ip du ] + [ j-0o_ f /c+(u) jP du] I.

Since q is uniformly continuous

and {@}CLp( I) is relatively compact, it is possible to choose
/3, 0 < /3~ z, to + @< b such that, for t E [to, to + /3],I/c&- I&, < r/2.
Let 0 < r* < r/2 and Af(r*) = {x E !S([tO - t, to + 01): IIx - &_r.ln+pl < r*. We show that
xr E D, for x E &(?(r*) and t E [to, to + /3], where I$&r*) is the closure of N,?(P). Now, for
such an x and t, we have

IIXI- 91lhzzIlxt- @tllk+ II@- CFlll

=sIIX- ~l/[s~r,r+,rl
+ /l@r
- ~III,,
< r* + L < r.
2
This proves

our assertion.

Definition.
For u < ~1 < ~2 < b, a function
x( E D, for t E [A, 4.
THEOREM

2. Suppose

f(t, x,) is measurable

that the following

x: [pl - z, ,LQ]+ E is said

conditions

in t for each fixed admissible

If(t, x,)) s m(t), for all admissible

to be admissible

if

are satisfied:

CC>

x,

x, where m E LP((a,

b), R+),

(C2>

N.

1338

If@>Yl - f(k Y*>I szL IIY - Y*IIlo, t E

PARHI

(a, b) and I+!J,I/J*E D, where L > 0 is a constant.

Then, for to E (a, b) and 9 E D, there exists an a> 0 such that

solution x(t, to, q) on [to, CO+ LY]with initial function q at to.
Proof.

We define

(2.1)

admits

(C3)

a unique

M : [to - T, b) + R+ and M* : [to - Z, b) -+ R as follows:

functions

0, t E [to- Gto1
and

M*(t)

(2.3)

m(s) dr, t E [to, b).

s 10

m(s) ds, t E [kl, b)

Further,
we define a function
@ : [to - t, b) + E as in (2.2). Since D is open, there exists
an r > 0 such that IVY c D. From the above discussion
it follows that there exists a
p, 0</3<z,
to+B<b
such that @ED,
t~[to,
to+p] and x,ED,
for t~[t~,
to+/J]
and x E N,(r*), where 0 < Y* < r/2.
We choose an LYsuch that 0 < a< /3 and M(tO + LY)+ {M*(to + oc>}p< r*. Such a choice
is always possible, since M(to) = 0 = M *(to) and both M(t) and M*(t) are continuous
functions
of t. Define
x0(t) = gS(t), t E [to - 5, to + a],
Xk
f0 -- 47,

(2.4)

Now we prove that xks are well-defined.

Clearly x1 is defined. Suppose that
k-1,2,....
x 2 , x 3 , . . , xk are well-defined.
To show that x k+l is also defined, it is enough to prove that
xk is a restriction
to [to - r, to + ol] of a function in N;r(r*). We define jk : [to - t, to + /3] +
E as follows:

?k(f) =
Clearly,

ik E P([to

xk(t), t
i

[to-

5.

to +

xyto+ a), f E [to+

- 5, to + /3]). Now

< r*.
So jk E N@(r*) and hence xk

is defined.

a!,

aj
to + /3].

(2.5)

Solutions

In the following,

lbk+l-

of a class of neutral

functional-differential

we show that the sequence

(xk) is Cauchy

= &ut;+4 Ixk+l(t) -xk(t)/ +

~kll[k-~,~"+al

to +

SL

1339

equations

in @([to - r, to + a]). Now

[yim~ik+yt)
-iytp dty
01

(Y

llxk - xk-l 111,dt + L [ j--fo+w{~~xk

- x~~~I~,}~dr]

(2.6)

f0
(11~~- xk ~ III,}P dt] ,+,

since by Holders

inequality,

where

l/p + l/q = 1. Further,

sup Ix(@--x k-1(O)l+@ik(,

HE[fO.fl

llxk - xkml III, =

- xk~(~$

4 L(l + &q)

Since,

for t E [to, to + a],

-ri~+3~PdH]lp
(2.7)
dH]*.

for tE [to, to + a],

11x1
- xllI, = ,yz,,
S

[lI~x(O)
-i(O)

lx(O) -x(O) 1+

M(t0 + ctz)+ {M *(to +

Ipd8]

cY)}:>>~
< r*,

from (2.7) we get

lIxk _ xk-:

Hence

II,,

Lk-(1

+ d.lq)k-

r*

{(;k~$l]p,

(2.8)

(2.6) and (2.8) yield

S Lk(l
/Ixk+l - xkl/[ro_r,t,I+aj

ak l/P
.
i 1

+ ~&~)~r* z

This proves that (xk) is a Cauchy sequence.

From lemma 1 it follows that there exists an x E @([to - r, to + a]) such that

Ibk- ~ll[~~-~.~,~+~~+
0 as k-+ x.
Now we show that x is an admissible
i(t)

(2.9)

We define 2: [to --t, to + /3] + E as

function.

x(t),tE[tot x(to + a), t E

?to+ al

[to+ (Y, to +

/3].

Since Rk E NG (r*) and

llfk - ~ll[t~Lr+~,
= /IXk- ~ll,t~r.r+a],
from (2.9) it follows

that X E N&r*).

Hence

X is admissible.

Consequently,

x is admissible.

1340

N.

PARHI

It is easy to see that xf, = 9 and

x(t) = ($40) +

[f(s, x,) ch, t E [to, to + 4

i fn

Hence x(t) is a solution of (2.1).

That this solution is unique can be proved proceeding
as in [l]. This completes the proof
of the theorem.
Theorem 2 in [l], on extended existence, remains valid for (2.1).
In the following example we take DH in place of D, whete DH = {y E Q*( IO) : 11
I& < H}.
Example

1. We consider

for an admissible x. Here Ki(t, 13)and K2(t, 0) are defined for t E (a, b) and 0 E IO with values
in R and measurable
in (t, 0). Suppose that I Kl(t, 0) I s L1 and I Kz(t, 0) / s Lz. for t E (a, b)
and 8 E Zo. We denote by f(t, x,) the right-hand
side of (2.10). So f(t, x,) is measurable
as a
function of t for each fixed admissible x. Further, for any admissible x,
lf(r,x,)l

~~l~llxll,,+L*I1XII:,<(Ll~+L*~)~.

For vi and yt2 E DH and t E (a, b),

Hence

all the conditions

of theorem

5 are satisfied.

Example 2. Consider
dx
dt=

1 KI(t, ~9)ix,(e) I de + lo K2(t, @i:(O)

-r
--T

de,

(2.11)

for an admissible x. Here we take DH = {y, E Q3(Zo) : 11

IJII~~< H}. K,(t, 0) and K2(t, 0) in
(2.11) are same as in example 1 above. It is easy to verify that all conditions of theorem 5 are
satisfied.
3. CONTINUOUS

DEPENDENCE

OF SOLUTIONS

In the following,
we prove a theorem on continuous
dependence
of solutions on initial
functions.
Other continuous
dependence
theorems can be proved in a similar fashion.

Solutions of a class of neutral functional-differential

1341

equations

THEOREM 3. Suppose
that the conditions
of theorem
2 are satisfied. Let x(t, to. q) and
y(t, to, t/~)be unique solutions of (2.1) on [to, to + cr] and [to, to + &I, 0 < N, CP < r, respectively
with initial functions
q and v E D at to E (a, b). If b = min(a,
ai), then for every E > 0
there
exists a 6> 0 such that 11~ - tj& < 6 implies
Ix(t, to. v) - y(t, f~. tj~)l < F,

for t E [r0, t0 + PI).

Proof.

For t E [to, to + p],

by Holders

inequality,

where

function
of t,
l/p + l/q = 1. Since IIx -y/l ,, is a continuous
from Gronwalls inequality in Lp (see [5] or [6]), it follows

l/x- Y III,E LP([to,to+ /3]).Hence,

that

where

C = L(l

+ /3q) [l - (1 - e-~Lp(l-~y))liP].
lx(t) - ~(4 I c Ix(to) - I

+P-

Now, for t E [to, to + /3],

I + [;-lbXt)

-y(t)

/ dt

k +B

vllzo+ L I r,~ /Ix- Y Ilz,dt

c l1q- qllz,+ LPliq (jiaig{llx - Yllz,Ipdfj

++~,ll~-~l,.
The conclusion

of the theorem

follows

4. GLOBAL

We need the following lemma

Frechet-Kolmogorov
theorems.
LEMMA 4. A set Q in np([a,

from the above

EXISTENCE

in the sequel.

b]) is relatively

inequality.

OF SOLUTIONS

The proof

compact

follows

from Ascoli-Arzela

if and only if (i) it is uniformly

and

bounded,

1342

N. PARHI

(ii) it is equicontinuous

and

(iii)

lii

uniformly in x E Q, where,
,k(t+o)=Ofort+a<a.
We consider

ab Il;(t + a) - i.(t) IPdt = 0

for t E [a, b], we define .k(t + a) = 0 for t + u> b and

(4.1)
where f: R X @(lo)-,

E. For this section, we define admissible functions as follows.

Definition. For - ~0 < ~1 < ~2 < + ~0, a function x : [ ~1 - z, Q] ---, E is said to be admissible

ifx,E

~1.

Qp(ZO>, tE

[rll,

THEOREM5. Suppose that

(HI)
(Hz)
(Hj)
(H4)

f(t, x,) is measurable in t for each fixed admissible x,

1f(t, .x,)1 s m(t) for all admissible x, where m E Lp(R, R+),
f(t, IJJ) is continuous in 11,E SP( IO) for each t belonging to a bounded interval in R,
Liio Jab/f (t + (5, x,+ 0) - f (t, x,) Ip dr = 0 uniformly with respect to admissible x.

Then; for

to

and VE Qp(Zo), there exists a solution x(t, to, q) of (4.1) on [to, 00).

Note. We define f(t, I$) = 0 for 111

$AS?(ZO).
Proof. Let 0 < /3 < CO.Define
@: [to- z, to + /3] + E

as follows:

and
q,(t) = q(O), for to G t 6 to + p.
So @E SP([to - t, to + /I]) and hence @ is admissible. Define
ql(t), t [ ,!o - 5, to+
Xyq

= =

q(O) + y

B
k

(4.2)
f(s, xs) h, t E [h +{,

to + p]:

for k = 1, 2, 3, . . . . Clearly x(t) is defined on [to - r, to + /3] and is admissible. For k 2 2,

the first expression of (4.2) defines xk(t) on [to - r, to + p/k] and its second expression defines
xk(t) on [to + p/k, to + 2p/k]. Suppose that xk(t) is defined on [to - t, to + jp/k]. To show that
xk(t) is defined on [to - r, to + (j + 1)/3/k]. For this it is enough to show that x: E SP(Zo)
for t E [to, to + j#k]. Smce xk(t) is defined on [to - t, to + j/$k], x: E !?(Zo) for t E [to,

Solutions
to + (j -

C&;iie-

l)@].
l)#k,

of a class of neutral

equations

1343

From (4.2) it follows that x ,: IO-+ En is absolutely continuous for t E

to + j/3/k].Now, for t E (to + (j - 1)/?/k, to + j/3/k]such that t - t > to + /3/k,
/ I$(Wde=
--r

If to c t - ts

functional-differential

/I, lf(s

- f,+~)

I& c~ZI,~~~(S)

Q.

to + p/k, then we obtain

- BP

m(s) ds.

Ift-tcto,then

Consequently, x,k E Qp(Zo) for t E [to, to + j/3/k] and hence xk(t) is defined on [to - t,
to + (i + l>L@l. F rom induction hypothesis it follows that xk(t), k 2 2, is defined on [to - z,
to + /3] and admissible.
Clearly, (xk) C RP([to - z, to + /3]). Now

for t E [to - r, to + p], implies that the sequence (xk) is uniformly bounded. In the following,
we show that (xk) is equicontinuous.
Let tl and t2 E [to - t, to + /3]. If tl and t2 E [to - t,
to + p/k], then

IXk(h)- Xk(h)I = IQqh>

- Q?(t2)

I.

If tl E [to - t, to] and t:! E (to + /3/k,to + fi], then

bk(td-

XkG2)

I c

I &I> - @(to)+

s I@,(h>

It2-Bkxs>
I
If@,

I ds

to

rz

@(to)

m(s) ds.

fl

If tl E (to, to + p/k] and t2 E (to + p/k, to + p], then

IXk(h)-

xk(t2) I s yk
(0

If(s, x, > 1ds s j- - Pkm(s) ds.

to

We may note that tz - p/k - to S t2 - tl if and only if tl s to +/3/k. Lastly, if tl and t2 E

(to + P/k to +

PI, then

1344

N.PARHI

From the uniform continuity

of functions C$and J&T m(s) ds on [to - r, to + /3] it follows that
the sequence (a?) is equicontinuous.
Lastly we show that
h +

lim
li(t
o-to f_T
uniformly

+ a) - i.(t) lpdt = 0

in k. Let u > 0. For o < 0 the calculation

holds.

(4.3)

Now

,:I: I.P(t + CJ)- x(t) IP dt = II + I2 + 13 + 14 + 15,

j
where
I,

QI 0
= 1_T Ix@ + a) - P(t)
j
=
j

12 =

@P

IP dt

-,I @(t + a) - c+(t) IPdt 4 j0 i@(t + a) - c@(t)lp dt,

_T

r+!ikpoli(t + u) - i(t) lp dt d j 1Q(t) Ip dt,

j 1- u
-0

li(t

+ u) - ik(r)Pdt]

lip

Xk(t + u)lP dt

=s

by Minkowskis

inequality,
IXk(t + u) - i(t)

ZJ =

IPdr

tII* /i
8
j Ill

1f(t + u. xf;+ .) - f(t, x,) lp dr,

and

Since {QI}C W(Zo) is relatively compact, m E LP(R, R), @E @([to - z, to + /3]) and (H4)
holds, (4.3) is satisfied. From lemma 4 it follows that the sequence (xk) has a sub-sequence

Solutions

(,@I) which converges

c

of a class of neutral

1345

equations

to a function
2 E W([l

Clearly,

functional-differential

- z, to + in]).

for f E [to - r, to]

I Q?(f) - x0(4 I = IXk(4 - x(f) I =z II*1-

Sox(t)= q,(t)for
x0 is admissible.

xll,,-L,,+p,.

x E !Z([to - z, to + /3]) implies

t E [to - z, to], that isx: = ~1. Further,

Since, for t E [to, to + /3],

that

112;
- .dllI,,= Ibki- XI/L
=zIW - Xll,r,,-r.t+&
from assumption

(Hj) we obtain
1f(c dc> - f(t, xY> I + 0

as k,+

=.

Hence

Lebesgues

dominated
;ir

convergence

theorem

JI:f(% x.3 h = /A%

fll

implies

that

XP) ds.

for t E [co, to + /cl]. Further

IXkW- x0(t)I c II+ - X/I,l,i-r.r,,+~,~

for t E [to, to + p], and

Ir

f(s,x,k,) ds-0

f- /3ik,

as ki*

0~.

Now taking

limit in (4.2) as kj+

x(t)= do)

m we

obtain

I,,+x3h,

for t E [to, to + b]. So x0 is a solution of (4.1) on [to. to + b] with initial

/3 is arbitrary, x0 is a solution of (4.1) on [to, CC).
This completes the proof of the theorem.
Acknowledgement-The

author

is thankful

to the referee

for helpful

function

q at to. Since

remarks

REFERENCES
1. DAS P. C. & PARHI N., On a functional-differential
equation of neutral type. J. math. Analysis Applic. 35, 6782 (1971).
2. DRIVER R. D., Existence and continuous
dependence
of solutions of a neutral functional-differential
equation,
Archs Ration. Mech. Analysis 19, 149-166 (1965).
3. DUNFORD N. & SCHWARTZ J. T., Linear Operators, Part I, Interscience.
New York (1957).
4. H.&LE J. K. & CRUZ M. A., Existence, uniqueness
and continuous
dependence
for hereditary
systems. Anna/i mat.
pura appl. 85, 63-82 (1970).
5. WALTER W., Differential and Integral Inequalities, Springer. New York (1970).
6. WILLE~~ D., Nonlinear
vector integral equations
as contracting
mappings,
Archs Rarion. Mech. Analysis 15,
79-86 (1964).