AC2[R+,R] - Hindawi

4 downloads 0 Views 2MB Size Report
aC[R+,R+], AC2[R+,R]," is a constant, f C[FtxR+,R], gEC[R+ xJ, R], er. [J,R], and the integral in (E) is a Stieltjes integral.Throughout this paper we assume that.
Journal of Applied Mathematics and Stochastic Analysis 9, Number 1, 1996, 21-31

OSCILLATION CRITERIA FOR NONLINEAR INI-IOMOGENEOUS IiYPERBOLIC EQUATIONS WlTtI DISTRIBUTED DEVIATING ARGUMENTS 1 XINZHI LIU and XILIN FU 2 University of Waterloo Department of Applied Mathematics Waterloo, Ontario, Canada N2L 3G1

(Received March, 1995; Revised July, 1995) ABSTRACT This paper investigates the oscillatory properties of solutions of nonlinear inhomogeneous hyperbolic equations with distributed deviating arguments subject to two different boundary conditions. Several oscillation criteria are establishing employing Green’s Theorem and certain differential inequalities. An example is also given.

Key words: Inhomogeneous.

Oscillation, Hyperbolic Equation with Deviating Arguments,

AMS (MOS) subject classifications: 35B05, 34K40, 34A40.

1. Introduction The oscillation theory of hyperbolic partial differential equations with delays has become an important area of investigation in the past few years, see [1]-[4] and references therein. Recently, some interesting oscillation conditions have been obtained in [1] for homogeneous hyperbolic equations with distributed deviating arguments. In this paper, we shall consider the following nonlinear inhomogeneous hyperbolic equation with distributed deviating arguments: b

(E)

a

=a(t)Au+f(x,t),

(x,t) EG,

R+

where G- f/x R+, f is a bounded domain in R n with piecewise smooth boundary 0, [0, + oo), u u(x, t), A is the Laplacian in R n, p G C[-I x R + x J, R + ], J [a, b], F C[R, R],

is a constant, f C[FtxR+,R], gEC[R+ xJ, R], er Stieltjes integral. Throughout this paper we assume that inf g(t,)respectively, with g(t,)< t for any and lim

aC[R+,R+], AC2[R+,R], " [J,R], and the integral in (E) is a g(t,) is nondecreasing in t and

+ C,

and that

r()

is nondecreasing in

.

t--

+ cx

EJ

We shall consider two kinds of boundary conditions:

1Research supported by NSERC-Canada. 20n leave from Shandong Normal University. Printed in the U.S.A.

(C)1996 by North

Atlantic Science Publishing Company

21

22

ON and

-

XINZHI LIU and XILIN FU

7(x, t)u

,

#(x, t), (x, t) E Oft R +,

(B)

(..,). (:.,) aa x R +.

(B2)

where N is the unit outnormal vector to Of, 3" E C[8ft x R +, R + ], #, C[8ft x R +, R]. A solution u(x,t) of equation (E) satisfying certain boundary conditions is said to be oscillatory if, for any positive number a there exists a point (xo, to) ftx [a, + c) such that u( 0, to) 0. The objective of this paper is to study the oscillatory properties of solutions of equation (E) subject to boundary conditions (B1) and (B2). In Section 2, we shall establish several oscillation criteria for boundary value problems (E)-(B1) and (E)-(B2), employing Green’s Theorem and certain differential inequalities. We shall then develop, in Section 3, some results on differential inequalities which, in addition to their independent value, enable us to obtain (in Section 4) further oscillation criteria regarding boundary value problems (E)-(B1) and (E)-(B2). An example is also given.

2. Oscillation Criteria Let us begin with listing the following notations and assumptions.

U(t)

/ u(x, t)dx

ft

/

where

(1)

P(t,)-min_p(x,t,), #

(t)

#(x, t)dS,

l_l

1 f (t)- lal

/ f(x t)dx

(4)

Rl(t) fi (t) + f (t), and

F(u)

is convex in

R+

F(- u)

and

F(u) > O,

u

R +,

(6)

where dS is the surface integral element on Oft. Theorem 1" Assume that condition deviating arguments

(6)

holds.

If the differential

inequalities with distributed

b

-2[U(t) + (t)U(t

P(t,)g(V[g(t,)])dff()_ Re(t

T)]

t-

a

(I1)

Oscillation Criteria

for Nonlinear Inhomogeneous

23

Hyperbolic Equations

b

a

of problem (E)-(B1) are oscillatory in G. of problem (E)-(B1). We may assume that is a positive number. Since lim t+ > a such that

have no eventually positive solutions, then all solutions Proof: Let u(x,t) be a nonoscillatory solution u(x,t)>O for (x,t) Ex[a,+cx), where a rain {g(t, )} + c it follows that there exists a t o j

u(x,t-r) > 0 Integrating both sides of equation

(E)

and u[x,g(t,)]

> 0, t > t0, E J.

with respect to x over domain f we obtain b

a(t) It is easy to

J

Au(x, t)dx +

./f(x, t)dx,

t

>_ t 0.

see that b

b

(9) Using Green’s Theorem, we have

Au(x, t)dx

(10)

-ffdS

In view of condition (6), it follows from Jensen’s inequality that

F(u[x,g(t,)])dx >_ Combining (8)-(11)

O

IF

(11)

[al

we have

d2[f o"(x’t)dx-(t)/’("t-’)dxl-a(t)/#(x’t)d’ft

oft

(12)

b

Thus we can see that the function defined by contradicts the condition of the theorem.

If

u(x, t) < 0

for

(x, t) @ x [c, + oo),

*(. t) Using F(- u) problem"

F(u),

for u

(1)

is a positive solution of inequality

(I1);

but this

then set

(.. t). (. t) e o [. +

R +, it is easy to check that u*(x, t) is

a

positive solution of the

XINZHI LIU and XILIN FU

24 b

u+A(t)u(x,t-)]+ p(x,t,)F(u[x,g(t,)])dr()-a(t)Au-f(x,t), (x,t) EG ,0u #(x, t), (x, t) e 0a [0, + oo); 0N + 7,x t)u a

and it satisfies

b

t o which also contradicts the assumption of the theorem. This completes the proof of Theorem 1.

The following fact and notations shall be used later in the proof of Theorem 2. Consider the Dirichlet problem

Au +,u

0 in ft,

loa-0, where ,- constant. It is well-known [3] that the smallest eigenvalue eigenfunction O(x) are positive. We define

"0

and the corresponding

(13)

v(t) where

u(x, t)

1

(14)

r (P(x)dx

is a solution of the problem

Theorem 2: Assume that condition deviating arguments

(E)-(B2). (6) holds. If the differential

inequalities with distributed

b

-[V(t) + ,(t)V(t

r)] + Aoa(t)V(t) +

P(t, )F(V[g(t, )])dr({) 0. If u(x,t)>O for (x,t) eax[a, +oc), then there exists a 0>c have no eventually positive solutions, then all solutions

such that

u(x,t-r) > 0 Multiplying both sides of domain f, we get

-

and u[x,g(t,)] > O, t

(E) by the eigenfunction (I)(z)

a(t)

(,)()d + (t)

G J.

> to,

and integrating with respect to x over the

(,t-

b

(15) Ft

a

j

+

j

>_ to,

From Green’s Theorem, it follows that

/(x, t)(P(x)dS-A /u(x, t)(P(x)dx, o

(16)

t>t o.

Using Jensen’s inequality, we obtain

Combining

J" F(u[x,

g(t, )])(x)dx >_

15)-(17),

we

S

a(x)dx, r

(1

(17)

f (x)dx

get

b

Thus we see that the function defined by contradicts the condition of the theorem.

If u(x, t) < 0 for problem

(x, t) E 2 x [ct, + oc),

(14)

(I3)

which

-u is a positive solution

or the

is a positive solution of the inequality

then the function u*

XINZHI LIU and XILIN FU

26 b

u

+ A(t)u(x,t- -r)] +

p(x,t,)F(u[x,g(t,)])dr() a(t)Au- f(x,t), (x,t)

G

a

e Oax[o, and it satisfies

In other words, the function,

. (P(x)dxl /u*(x,t)ap(x)dx

v * (t) is a positive solution of inequality of Theorem 2 is complete.

(I4)

for t

>_

0.

This provides a contradiction. Thus the proof

3. Delay Differential Inequalities From the discussion in Section 2, it follows that the problem of establishing oscillation

(E) can be reduced to the investigation of the properties of the solutions of delay inhomogeneous differential inequalities of the form

criteria for

b

)+ A(t)y(t- 7)] +

P(t, )F(y[g(t )])do’() < R(t)

t

> to,

(I5)

a

C[[t + oc), R],

C[[t0,

],

C[[t0, + co), R].

PE + oc) x J, R + F C[R, R], R Theorem3: Let A(t) >_ O for >_ o > O and F(y) > O for y R +. If

where A E

0,

liminf

/ (1-)R(s)ds-

(18)

for

every sufficiently large tl, then the inhomogeneous differential inequality ly positive solutions.

Proof." Suppose that y(t) is an eventually positive solution of such that

(I5).

> to

y(t) > O, y(t- "r) > 0 and y[g(t,)] > O, >_ tl,

J.

(I5)

has no eventual-

Then there exists a

Oscillation Criteria

Hence

for Nonlinear Inhomogeneous

Hyperbolic Equations

27

we obtain b

-[y(t) + $(t)y(t- 7)] _< R(t)-

P(t,)F(y[g(t,)])&r() a

__t 1. Integrating the above inequality twice over

It1, t], > tl,

get

we

y(t) + A(t)y(t- 7) tl)

,(t)y(- 7)] _< -y + c2 1

R(s)dsd,

1

we have

+

(1 -)R(s)ds.

get lim inf[y(t) + (t)y(t

7)] + A(t) >_ O, y(t) > 0 and y(t- 7) > 0 for t >_ tl,

On the other hand, since

(19)

c.

t--,

we have

liminf[y(t) + A(t)y(t- 7)] > 0 +c

t

(19). This completes the proof. When f(x,t)- O, #(x,t)-0 and (x,t)- O, the problem of establishing oscillation criteria (E) can be reduced to the investigation of the properties of the solutions of delay

which contradicts

for

homogeneous differential inequalities of the forms b

(16) a

and b

[y(t) + A(t)y(t- 7)]+

P(t,)F(y[g(t,)])dcr(() >_ O,

t>_t o.

(I7)

a

Along with (I6), (I7) we consider the delay differential equation b

[y(t) + 1(t)y(t- 7)]+

P(t,)F(y[g(t,)])dcr() a

where

,

E

C[[t 0 + cx), R], P E C[[to, + ) x J, R], F

Lemma 1" [5] Assume that

(I)

e

+

+ ];

C[R, R].

O,

(20)

_

28

XINZHI LIU and XILIN FU

(II) 0 < "1 ,(t) "2, t to, where 1, and 2 are constants; (III) y(t) + A(t)y(t- 7.) >_ kl, t >_ to, to, ]1 > O. Then there exists a closed and measurable set E C [t o + cx) and a constant k > 0

_(t) >_

and

,

t

such that

E

meas(EC[t,t+27])7", tt 0. Theorem 4: Assume that 0

< )1

that

F(y)

)(t)

t

to,

is a monotone increasing

If for any closed and measurable set E C [t o following condition holds

_

2,

that

F(y) > O,

F(- y)

function

in

_

y E R +, and

R +.

+ c) for which meas(E N [t,t + 27"]) >_ 7",

t

(21)

to,

the

b

a

E

then

() () (iii)

t1

the inequality (I6) has no eventually positive solutions; the inequality (I7) has no eventually negative solutions; all solutions of the equation (20) are oscillatory.

Proof: Let y(t) be an eventually positive solution of inequality t o such that

y(t) > O, y(t- 7") > 0 and y[g(t, )] > 0, _> tl, Setting

y(t) + A(t)y(t- 7"), t >_ tl,

z(t) we obtain from

(16)

that b

z"(t) _ 1 such that

[tl, -t-o). Suppose

that there

z’(t2)- < 0; then we have

z’(t) t. Integrating both sides of the inequality

(24)

from

2

to

z(t) t2)

(24) we obtain

Oscillation Criteria

for Nonlinear Inhomogeneous Hyperbolic

Hence limsupz(t)_< 0 which contradicts the assumption that y(t) is

29

Equations

an

eventually positive

solution. Thus we have

Z’(t) 2 O, k tl

(25)

which implies

Z(t)_> k 1 > 0

for t

>_

1.

From Lemma 1 it follows that there exists a closed and measurable set E C [t 1, q-oo) and constant k

> 0 such that via(t, )] >_

,

e,

t

a

a

and meas (E

By (21)

It, + 2r]) _> r,

t

>_ t 1.

we have

r(v[v(t, e)]) >_ r() > 0, e E, Integrating both sides of inequality

(23),

we

a.

get b

b

./" j .P(s, )d()ds _ t o > O. If

(26)

XINZHI LIU and XILIN FU

30

+ oo every sufficiently large t I >_ to, then all solutions of problem (E)-(B1) are oscillatory in G. Proof: Note that due to condition (26) we can see that the differential inequality (I1) has no eventually positive solutions by Theorem 3. Since from (27) it follows that

for

R l(s)]ds

1

lim inf

t--+

+ oo

1

liT sup t

+ oo

1

1

-

/l(S

differential inequality (I2) has no eventually positive solutions by Theorem 3 either. Hence all solutions of problem (E)-(B1) are oscillatory in G by Theorem 1. The proof is complete.

The following result can be proved similarly with the use of Theorem 3 and Theorem 2. Theorem 6: Assume that condition

(6)

holds and that

R2(s)ds

1

liT in]"

(t) >_ 0 for t >_ t o > O. If

1

and

limsu,

i (l-)R2(s)ds 1

for

A consequence

__

> to, then all solutions of problem (E)-(B2) are oscillatory in G. of Theorem 1 and Theorem 4 in the case of f(x, t)-0 and #(x, t)-0 is the

every sufficiently large t

following result.