Professional Documents
Culture Documents
Oscillation Theory For Second Order Linear, Half-Linear, Superlinear and Sublinear Dynamic Equations PDF
Oscillation Theory For Second Order Linear, Half-Linear, Superlinear and Sublinear Dynamic Equations PDF
by
Ravi P. Agarwal
Florida Institute ()j'Technology,
Melbourne, Florida, U.S.A.
Said R. Grace
Cairo University,
Orman, Giza, Egypt
and
(~f1reland,
A C.LP. Catalogue record for this book is available from the Library of Congress.
ISBN 978-90-481-6095-2
ISBN 978-94-017-2515-6 (eBook)
DOI 10.1007/978-94-017-2515-6
Contents
Preface
Chapter 1
1.0.
1.1.
1.2.
1.3.
1.4.
Preliminaries
Introduction
Continuability, Boundedness and Convergence to Zero
Some Useful Results from Analysis and Fixed Point Theorems
Notes and General Discussions
References
Chapter 2
2.0.
2.1.
2.2.
2.3.
2.4.
2.5.
2.6.
2.7.
2.8.
2.9.
xi
Introduction
Sturm Comparison Theorem
Nonoscillatory Characterizations and Comparison Theorems
Nonoscillation Criteria
Oscillation Criteria
Oscillation Criteria-Integral Averaging
Oscillation Criteria-Integrable Coefficients
Forced Oscillations
Notes and General Discussions
References
Chapter 3
1
2
10
12
12
13
13
15
25
35
40
70
82
90
91
3.0.
Introduction
93
3.1.
3.2.
3.3.
3.4.
Preliminaries
Sturm-Type Comparison Theorems
Levin-Type Comparison Theorems
Liapunov's Inequality
94
94
100
105
vi
An Oscillation Criterion for Almost-Periodic
Sturm-Liouville Equations
3.6. On Zeros of Solutions of a Singular
Half-linear Differential Equation
3.7. Nonoscillation Criteria
3.7.1. Nonoscillatory Characterization
3.7.2. Comparison Theorems
3.7.3. Sufficient Conditions for Nonoscillation of Equation (3.1.1)
3.7.4. Further Results on Nonoscillation of Equation (3.1.1)
3.8. Oscillation Criteria
3.9. Oscillation Criteria-Integral Averaging
3.9.1. Oscillation Theorems for Equation (3.1.1)
3.9.2. Interval Criteria for the Oscillation of Equation (3.1.1)
3.9.3. The Weighted-Average Oscillation Criteria
3.10. Oscillation Criteria-Integrable Coefficients
3.11. Oscillation of Damped and Forced Equations
3.11.1. Oscillation of Equation (3.11.1)
3.11.2. Oscillation of Equation (3.11.2)
3.12. Distance of Zeros of Oscillatory Solutions
3.13. Oscillation and Nonoscillation of Half-linear
Equations with Deviating Arguments
3.13.1. Some Useful Lemmas
3.13.2. Oscillation of Equation (3.13.1)
3.13.3. Some Extensions and More Oscillation Criteria
3.13.4. Classification of Nonoscillatory Solutions
3.13.5. A Comparison Theorem
3.13.6. Oscillation of Equation (3.13.2)
3.13.7. Nonoscillation of Solutions of Equation (3.13.127)
3.14. Notes and General Discussions
3.15. References
3.5.
Chapter 4
4.0.
4.1.
4.2.
107
113
140
140
147
148
154
163
168
168
179
188
196
205
206
207
210
218
219
223
232
237
251
253
259
267
268
Introduction
Superlinear Oscillation Criteria
Further Results on Superlinear Oscillations
273
273
296
vii
4.3.
4.3.1.
4.3.2.
4.3.3.
4.4.
4.5.
4.6.
Chapter 5
5.0.
5.1.
5.2.
5.3.
5.4.
5.5.
5.6.
5.7.
358
362
362
367
371
372
Introduction
Sublinear Oscillation Criteria
Further Sublinear Oscillation Criteria
Linearization of Sublinear Oscillation Theorems
Nonoscillation Theorems for Sub linear Differential Equations
Oscillation Criteria for Certain Nonlinear Differential Equations
Notes and General Discussions
References
Chapter 6
328
328
332
339
341
376
376
387
429
433
444
453
453
6.0.
6.1.
Introduction
Oscillation Criteria of Olech-Opial-Wazewski Type
457
457
6.1.1.
6.1.2.
6.1.3.
6.2.
6.2.1.
6.2.2.
458
459
463
466
467
469
476
6.3.
viii
6.3.1.
6.3.2.
6.4.
6.5.
Preliminaries
Oscillation Criteria
Notes and General Discussions
References
Chapter 7
7.0.
7.1.
7.1.1.
7.1.2.
7.1.3.
7.2.
7.3.
7.4.
7.5.
7.6.
7.7.
8.1.1.
8.1.2.
8.1.3.
8.1.4.
8.2.
Introduction
491
Oscillation of Nonlinear Two-dimensional Differential Systems 491
Superlinear Case
493
Sublinear Case
498
Further Oscillation Criteria
500
Oscillation Theorems for Linear Differential Systems
503
Oscillation Theorems for Differential Systems with
Functionally Commutative Matrix Coefficients
533
Comparison Theorems for Operator-Valued Linear
Differential Equations
540
Oscillation Results for Differential Systems
with Forcing Terms
549
Notes and General Discussions
552
References
554
Chapter 8
8.0.
8.1.
477
479
489
490
Introduction
Asymptotic Behavior of Solutions of
Nonlinear Differential Equations
Asymptotic Behavior of Nonoscillatory Solutions
Bounded Asymptotically Linear Solutions
Unbounded Asymptotically Linear Solutions
Further Extensions and Improvements
Asymptotic Behavior of Positive Solutions of
Singular Emden-Fowler-Type Equations
558
558
559
562
567
574
584
585
595
lX
Chapter 9
9.0.
9.1.
9.2.
9.3.
9.4.
9.5.
9.6.
614
622
624
Miscellaneous Topics
Introduction
Comparison and Nonoscillation Results
Nonoscillatory Solutions of Forced Differential Equations
Limit Circle Criteria and Related Properties
for Nonlinear Equations
Properties of Certain Differential Equations
Notes and General Discussions
References
625
625
632
638
650
654
654
Introduction
Preliminaries
Differential Inclusions
Neutral Inclusions
Notes and General Discussions
References
Subject Index
656
656
657
664
668
668
670-672
Preface
This book is devoted to the qualitative theory of second order dynamic
equations. In the last 50 years the Oscillation Theory of ordinary, functional, neutral, partial, and impulsive differential equations, and their discrete versions has attracted many researchers. This has resulted in hundreds of research papers in every major mathematical journal. There are
many books which deal exclusively with oscillation of solutions of differential equations. However, most of those books appeal only to researchers
who already are familiar with the subject. Thus, in an effort to bring oscillation theory to a new and wider audience, in this book we present a
compact, thorough, and self-contained account for second order dynamic
equations. An important feature of this book is that many examples of
current interest are given to illustrate the theory.
In Chapter 1, we introduce oscillation theory of second order differential
equations. Here also we provide conditions which guarantee that all solutions of a particular class of second order nonlinear differential equations
are continuable, bounded, and converge to zero. Our final section in this
chapter states several fixed point theorems which play an important role in
establishing existence criteria of nonoscillatory solutions.
It is natural to begin our discussion with second order linear differential
equations. The interest in second order linear oscillations is due, in a large
part, to the fact that many physical systems are modelled by such equations. We begin Chapter 2 with some of the most basic results in the theory of oscillations of linear ordinary differential equations of second order.
In particular, we present Sturm and Sturm-Picone comparison theorems.
Then, we provide some necessary and sufficient conditions for nonoscillation as well as some comparison theorems of Sturm type. Also we present
sufficiency criteria for nonoscillation of solutions. Next, we establish sufficient conditions for oscillation of second order differential equations with
alternating coefficients. Integral averaging techniques and interval criteria
are two of the most important concepts in oscillation theory and both are
discussed in Chapter 2. Criteria for oscillation of linear second order differential equations with integrable coefficients are also given. We conclude
this chapter by discussing the problem of forced oscillations.
xu
In recent years the study of half-linear differential equations has become
an important area of research. This is largely due to the fact that half-linear
differential equations occur in a variety of real world problems such as in the
study of p-Laplace equations, non-Newtonian fluid theory and the turbulent flow of a polytrophic gas in a porous medium. In Chapter 3, we present
oscillation and nonoscillation criteria for second order half-linear differential equations. We begin with some preliminaries on half-linear differential
equations, and then we present Sturm and Levin type comparison theorems
and a Liapunov type inequality. Next, we provide an oscillation criterion for
almost periodic Sturm-Liouville equations. In this chapter we also present
a systematic study of the zeros of solutions of singular half-linear differential equations, nonoscillation characterizations (necessary and sufficient
conditions), comparison results and sufficiency criteria for nonoscillation.
In addition we study oscillation by employing integral and weighted averaging techniques. Here, interval criteria for the oscillation of half-linear
equations will also be provided. Next, we discuss oscillation of half-linear
equations with integrable coefficients, and damped and forced equations.
We also derive lower bounds for the distance between consecutive zeros of
an oscillatory solution. Finally in this chapter, we present a systematic
study of the oscillation and nonoscillation of half-linear equations with a
deviating argument. Here, classifications of nonoscillatory solutions, and
existence results which guarantee that the solutions have prescribed asymptotic behavior will be given.
In Chapter 4, we present a oscillation and nonoscillation theory for
second order nonlinear differential equations of superlinear type. We begin
with the oscillation of superlinear equations with sign changing coefficients.
In particular, we discuss results which involve integrals and weighted integrals of the alternating coefficients, and then provide several criteria which
use average behaviors of these integrals. More general averages such as
'weighted averages' and 'iterated averages' will also be employed. Also,
we provide sufficient conditions which guarantee the existence of nonoscillatory solutions, and then we present necessary and sufficient conditions
for oscillation of superlinear equations. Oscillation results via comparison
of nonlinear equations of the same form as well as with linear ones of the
same order will also be presented. We conclude this chapter by discussing
oscillation and nonoscillation of forced and damped superlinear differential
equations.
In Chapter 5, we present oscillation and nonoscillation criteria for all
solutions of second order nonlinear differential equations of sublinear type
with alternating coefficients. Our results will involve integrals and weighted
integrals of alternating coefficients and in some results we use integral aver-
xiii
aging techniques. We also provide some new linearized oscillation results for
second order sublinear differential equations. Next, we study the EmdenFowler equations. These equations originated from the theory of gaseous
dynamics in astrophysics around the beginning of the last century, and they
also occur in the study of gas dynamics, fluid mechanics, relativistic mechanics, nuclear physics, and in chemically reacting systems. In this chapter we present criteria for nonoscillation of sublinear Emden-Fowler type
equations. Also we compare the oscillatory behavior of certain nonlinear
equations with related half-linear differential equations. We conclude this
chapter by discussing oscillation of general nonlinear differential equations.
In Chapter 6 some special techniques are used to obtain oscillatory criteria for differential equations. Firstly, we present oscillation and nonoscillation theorems for nonlinear second order differential equations by using
the method of Olech, Opial and Wazewski. Secondly, we use a variational
inequality to study the oscillation of half-linear second order differential
equations. Finally, we apply Liapunov second method to provide criteria
for oscillation of second order nonlinear equations.
In Chapter 7, we discuss the oscillatory behavior of nonlinear twodimensional differential systems and second order vector-matrix differential
equations. We begin with the oscillation of nonlinear two-dimensional differential systems. This includes the superlinear, linear, and sub linear cases.
Then, we discuss oscillation of linear second order differential systems. Two
approaches are presented. In the first approach the system considered will
be reduced to a certain scalar Riccati inequality, so that known results from
the literature can be applied. In the second approach, we will use Riccati
and variational techniques. Next, we discuss the oscillation of nonlinear
second order differential systems with functionally commutative matrix coefficients. We show that the oscillation theory of such systems can be
effectively reduced to the study of diagonal systems of scalar second order
differential equations. Then some comparison theorems of Hille-Wintner
type for second order operator-valued linear differential equations will be
presented. Finally, we discuss oscillation of second order differential systems with a forcing term.
The study of behavioral properties of solutions of differential equations
near infinity is of immense importance in applications. In Chapter 8, we
present some recent contributions on the asymptotic behavior of solutions
of second order differential equations as well as the behavioral properties
of positive solutions of singular Emden-Fowler-type equations. It is shown
that for a large class of differential equations, not only can the existence of
nonoscillatory solutions be proved, but also an explicit asymptotic form of
XlV
the nonoscillatory solutions may be provided. Next, by imposing more restrictions on the sign of the integrable coefficient of the equation, we obtain
necessary and sufficient conditions so that the solutions have a specified
asymptotic behavior as t --; 00, i.e., solutions which behave asymptotically like a nonzero constant and also those which behave asymptotically
like ct, c 1= o. For this, various averaging techniques of the type employed
in the previous chapters to study the oscillatory behavior of such equations
will be used. Then, we study existence, uniqueness and asymptotic behavior of positive solutions of singular Emden-Fowler-type equations. The
cases when the coefficient of the equation under consideration is of constant
sign, or of an alternating sign are systematically discussed. Finally, existence and nonexistence results for positive solutions of Emden-Fowler-type
systems will be given.
In Chapter 9, some special results are presented. We extend the SturmPicone theorem, obtain nonoscillation theorems for perturbed second order
nonlinear differential equations, and present a nonlinear Picone type identity which enables us to prove some Sturm-Picone type comparison theorems for nonlinear equations. Next, we study nonoscillatory solutions of
forced differential equations of second order. Then, we present some limit
cycle criteria and discllss its related properties for nonlinear second order
differential equations. Finally, we provide some properties of solutions of
very general second order differential equations.
In our final chapter, we introduce a nonoscillation theory for second
order differential and neutral inclusions. Here, the results rely on some
fixed point theorems for multivalued maps and on a compactness criterion.
We hope this book is timely and complements existing works in the
literature on oscillation theory of second order dynamic equations. It can
be used as a textbook at the graduate level and hopefully it will stimulate
further research in this area.
It is impossible to acknowledge individually colleagues and friends to
whom we are indebted for assistance, inspiration and criticism during the
preparation of this book. We must, however, express our appreciation and
thanks to Sadhna for typing and proof reading parts of the manuscript.
Ravi P Agarwal
Said R Grace
Donal O'Regan
Chapter 1
Preliminaries
1.0. Introduction
In this introductory chapter, we shall provide conditions which guarantee that all solutions of a particular class of second order nonlinear differential equations are continuable, bounded, and converge to zero. Then, we
shall state several fixed point theorems which play an important role in our
later chapters.
In what follows, we shall use the following notation: lN = {I, 2,-}
represents the set of positive integers, lR = (-00,00), lRo = [0,00), lR+ =
(0,00) and 1R - = (-00,0) are the usual sets of real, nonnegative, positive
and negative numbers respectively. Further, for a nonnegative integer m
and I ~ JR, we denote by C (I x lRm, lR) the space of all continuous
functions mapping I x JRm into JR, and for a fixed i E lN, Ci(I X JRm , JR)
represents the space of all i times continuously differentiable functions.
Consider the second order ordinary differential equation
(1.0.1 )
where FE C([to, 00) xlR3, lR). Bya solution of (1.0.1) we mean a function
x(t), t E [tx, 00) C [to,oo) which is twice continuously differentiable and
satisfies equation (1.0.1) on the interval [tx, 00). The number tx ::::: to::::: 0
depends on the particular solution x(t) under consideration.
A nontrivial solution x(t) of (1.0.1) is said to be oscillatory if it has
arbitrarily large zeros. Otherwise, x(t) is said to be nonoscillatory, i.e.,
x(t) is nonoscillatory if there exists a tl ::::: to such that x(t) i- 0 for
t ::::: tl. In other words, a nonoscillatory solution must be eventually positive
or negative. Equation (1.0.1) is said to be oscillatory if all its solutions are
oscillatory. In our study of oscillatory behavior of differential equations
explicit knowledge of solutions will not be needed, rather we shall provide
easily verifiable conditions on the known quantities which guarantee the
oscillation.
Chapter 1
(a(t)x'(t'
(1.1.1)
where a(t), q(t) E C([to, 00), lR+), f, 9 E C(lR, lR) and h, e E C([to, 00) x
lR2 , lR), and g(y) > 0 for all y E lR. It will be convenient to write
equation (1.1.1) as the system
{
X'
y'
(1.1.2)
le(t,x,y)1
yh(t, x, y)
r(t),
(1.1.3)
(1.1.4)
~ 0
Iyl
g(y) ~ m
+ nG(y).
(1.1.5)
Before we prove the continuability theorem, we state the following wellknown Gronwall's inequality.
Lemma 1.1.1.
suppose that
Let
u(t)
= [to, T)
~ c+
it
q(s)u(s)ds
for
tEl,
to
u(t)
c exp
(1:
q(s )dS)
for
tEl.
Theorem 1.1.1.
If conditions (1.1.3) - (1.1.5) hold,
t ~ to, F(x) is bounded from below, and G(y) --700 as
all solutions of equation (1.1.1) are defined for all t ~ to.
a'(t)
Iyl
~ 0 for
--700, then
Proof. Suppose (x(t), y(t is a solution of (1.1.2) with finite escape time,
i.e., there exists T ~ to such that limt->T- [lx(t)1 + ly(t)l] = 00. Since
Preliminaries
F(x) is bounded from below, F(x) 2': -K for some constant K > O.
Define
then
q'(t)
a'(t) y2
1
( y )
- q2(t) G(y) - a(t)q(t) g(y) - a(t)q(t) h(t, x, y) g(y)
V'(x, y, t)
+ a(t)q(t) e(t, x, y)
y) -
g(y)
<
<
[q'(t)]q2(t) G(y)
r(t)
a'(t)
a2(t) [F(x)
+ K]
r(t)
Integrating the above inequality from to to t and noting that r(t)/ (a(t)q(t))
is bounded on [to, TJ, we obtain for t E [to, T] that
q(t) G(y(t))
< Kl exp
:<:;
K1exp
r(s)])
(i [[q'(S)]q(s) + n a(s) ds
t
to
(iT
to
< K2 <
00,
Remark 1.1.1.
y2
g(y) :<:; MG(y)
for
Iyl 2': k.
(1.1.6)
The proof of this result involves more details than the proof of Theorem
1.1.1, so we omit it here. We also note that (1.1.6) implies (1.1.5).
Chapter 1
:s:
a2,
a2
t 2: to
is a constant,
(1.1. 7)
1 () q' (t)
Ie (t,x,y )I < -;a
t q(t)'
If
F(x)
--->
as
00
Ixl
--->
(1.1.8)
00,
V(x, y, t) =
we have
V'(t)
q'(t)
:s:
[F(x)
:s:
y )
:s:
a(t)
m+nV(to)+
t
to
+ nV(t)
q'(s) ( q(s)
ly(s)J )
q(s) na(s)[F(x(s))+K]+ g(y(s)) ds.
J!&2L
+ n q(t) [F(x(t)) + K]
g(y(t))
a(t)
<_ Kl exp
(1 q'(s)
dS)
q(s)
t
to
K q(t)
1 q(to) ,
Preliminaries
Theorem 1.1.3. Suppose (1.1.4), (1.1.6) and (1.1.8) hold, a(t):::; a2,
where a2 is a constant, and
00
to
[a'(s)]ds <
a ()
s
(1.1.9)
00
and le(t, x, y)1 :::; a(t)q'(t)j(Mq(t)). Then all solutions of equation (1.1.1)
are bounded.
~ < B+ y2
g(y) -
g(y)
for all
y.
V(x, y, t) =
V'(t)
: :;
a(t)
M g(y)
L]
g(y)
<
+ K] +
+ K] + A + MG(Y))
and hence
V(t)
r (q, (s) +
ltD q(s)
Kl exp (1: ~gjdS)
<
M [a' (s )]-)
= Kl
a(s)
qq(~~) ,
dS)
Kl is a constant.
Chapter 1
6
The boundedness of x(t) now follows as in Theorem 1.1.2.
If M
< 1, we define
q(t)
V(x, y, t) = a(t) [F(x)
+ K] + G(y) + M (A + B).
V'(t)
::;
<
( q, (t)
q(t)
(q(t) [F(x)
a(t)
+ K] + A + MG(Y))
+ [a'(t)]-) V(t).
a(t)
Now, from Theorems 1.1.2 and 1.1.3 we deduce the following corollary.
Corollary 1.1.1. If in addition to the hypotheses of either Theorem 1.1.2
or 1.1.3, q(t)::; q2, q2 is a constant, and G(y) ---+ 00 as Iyl ---+ 00, then
all solutions of the system (1.1.2) are bounded.
Proof. From the proof of Theorem 1.1.2, we have
~~j
V'(t) ::;
so
to
[: + V(t)] ,
q'(s)
( q(t) )
q to
m
n
::; K2
qq(~~)
<
00,
K2 is a constant.
The following lemma will be used to prove the convergence to zero of the
nonoscillatory solutions of equation (1.1.1). For this, we need the following
additional assumptions:
(Id xf(x) > 0 if x =I- 0 and f(x) is bounded away from zero if x is
bounded away from zero,
(h) condition (1.1.3) holds and r(t)/q(t)
---+
0 as t
---+ 00,
Preliminaries
tl ;::::
r(t) < ~A
cq(t)
4
and
k(t) < ~A
q(t)
4
(a( t ) x')'
e(t,x,x')
h(t,x,x')
() ( )
g(x') g(x') - q t f x
g(x')
<
A]
<
f
-"21 Aq()tort;::::
t3
~Aclt q(s)ds
2
-->
00
as
--> 00,
t3
t4' Hence,
The following two examples illustrate that condition (h) above is essential.
Example 1.1.1. Consider the equation
xl/(t)
+ tx'(t) + -x(t)
t
t> O.
(1.1.10)
Chapter 1
Here g(x') = 1 and h(t, x, x') = tx', and it is clear that all the hypotheses
of Lemma 1.1.2 are satisfied except (13) since we do not have Ih(t, x, x')1 :::::
k(t)g(x'). Equation (1.1.10) has a bounded nonoscillatory solution x(t) =
(t + l)/t which does not have lim inft-+oo x(t) = o.
Exrunple 1.1.2. The equation
satisfies all the conditions of Lemma 1.1.2 except (h). Here Ih(t, x, x')1 =
tlx'i ::::: t[l + (x'?] = tg(x') but k(t)/q(t) = t 2 f+ 0 as t ---> 00. Again
x(t) = (t + l)/t is a bounded nonoscillatory solution of equation (1.1.11).
In what follows it will be convenient to have the following notation.
Condition C. If x(t) is a nonoscillatory solution of equation (1.1.1),
then limt-+oo x(t) = O.
Also, we define
(t [q'(s)]- )
p(t) = exp ( - lto q(s) ds
and
b(t) = exp (-
lto
[a'(s)]a(s)
dS) .
1
1
[q'(s)]ds <
q()
S
00
to
00
to
r(s) ds <
q(s)
(1.1.12)
00,
00
(1.1.13)
o.
+ K] + qtt) G(Y)) ,
Let
Preliminaries
then
,
{-a'(t)
V (t) = b(t)p(t) a2(t) [F(x)
+ K]-
q'(t)
a'(t) y2
q2(t) G(y) - a(t)q(t) g(y)-
(g~)
_(_1_
[F(x) + K] + _1 G(Y) ([a'(t)]- + [q'(t)]-)}
a(t)
q(t)
a(t)
q(t)
a2(t)
[a'(t)]- G
- a(t)q(t) (y)
b
:S (t)p(t)
q2(t)
[a'(t)]- ( y2 )
g(y)
+ a(t)q(t)
a(t)q(t)
+ a(t)q(t) e(t, x, y)
Y )}
g(y)
{[a'(t)]- ( y2 )
r(t)
Iyl}
a(t)q(t) g(y) + a(t)q(t) g(y) .
Now conditions (1.1.9) and (1.1.12) imply that q(t) ~ ql > 0, p(t) ~ PI >
0, a(t) ~ al > 0 and b(t) ~ bl > 0 where ql, PI, al and bl are
constants. Also, Iyll g(y) :S NI for all y and some constant NI > O.
Thus, on integrating V'(t) from to to t, we get
V( t )
for all t
<
-
V()
to
+ -Nit
ql
to
[a'(s)]- d
NI i t r(s)d
s+-- s:S K I <
a(s)
al to q(s)
00
a(t)
a2
for
t ~ to,
and
10
Chapter 1
F(x(t
a2
q(t)
::; b(t)p(t) V(t) ::; b1P1 V(t)
a2 N
< -a2- V()
t1 + -
b1P1
b1P1 q1
it
limt-->oo x(t)
tl
[a'(s)J- dS
a( s)
1
+ -a2-Na1 b1P1
it
tl
r(s) ds
-<
q( s)
-->
t.
0 if and
It:'
Preliminaries
11
12
Chapter 1
functions such that limn->oo fn(x) = f(x) a.e. in E, and for every
n E N, Ifn(x)1 ::; g(x) a.e. in E, where g is integrable on E. Then
lim r fn(x)dx
n->ooJ
E
JrE f(x)dx.
1.4. References
1. R.P. Agarwal, M. Meehan and D. O'Regan, Fixed Point Theory and
Chapter 2
Oscillation and Nonoscillation
of Linear Ordinary Differential
Equations
2.0. Introduction
The oscillation and nonoscillation property of solutions of second order
linear differential equations is of special interest, and therefore, it has been
the subject of many investigations. The interest in second order linear oscillations is due, in a large part, to the fact that many physical systems are
modelled by such equations. In this chapter we shall discuss some of the
most basic results in the theory of oscillations of linear ordinary differential equations of second order. In Section 2.1, we shall present Sturm and
Sturm-Picone comparison theorems which are useful in oscillation theory.
In Section 2.2, we shall provide some necessary and sufficient conditions for
the nonoscillation as well as some comparison theorems of Sturm's type.
Sufficiency criteria for the nonoscillation are given in Section 2.3. In Section 2.4, we shall establish sufficient conditions for the oscillation of second
order differential equations with alternating coefficients. Integral averaging
techniques as well as interval criteria for the oscillation are discussed in Section 2.5. In Section 2.6, several criteria for oscillation of linear second order
differential equations with integrable coefficients are established. Finally,
in Section 2.7 we shall discuss the problems of forced oscillations.
x"(t)
+ q(t)x(t)
= 0
and
y"(t)
+ ql(t)y(t)
= 0
Chapter 2
14
t2
y(t)x'(t) - x(t)y'(t)l t2 =
t,
t,
[ql(S) - q(s)]x(s)y(s)ds,
from which one derives the desired contradiction to the assumption that
x(t) > 0 and y(t) > in (tl, t2)'
This result extends readily to the solutions of
(a(t)x'(t)'
+ q(t)x(t) =
and
(a(t)y'(t'
+ ql(t)y(t) =
0,
dt (y(t)a(t)x'(t) - x(t)a(t)y'(t
(ql(t) - q(tx(t)y(t).
and
(a(t)x'(t'
+ q(t)x(t) =
(2.1.1)
(al(t)y'(t'
+ ql(t)y(t) =
0,
(2.1.2)
where a(t), al(t) E C I ([tl,t2], JR+) and q(t), ql(t) E C([tb t2], JR) under
the hypotheses
# q(t
(2.1.3)
:t
(y(t)a(t)x'(t) - x(t)al(t)y'(t
(ql(t) - q(tx(t)y(t)
+(a(t) - al(tx'(t)y'(t),
is not useful since the signs of x'(t) and y'(t) are not known. However,
if y(t) # 0, then this identity modified by Picone [26] becomes
! (:g~
+ (a(t) - al(t(x'(t2
+ al(t)
(x'(t) _ :i:?y'(t) 2,
(2.1.4)
15
t2
[(a(s) _ aI(s))(x'(s))2
t,
+ (qI(S)
- q(s))x2(s)] ds 2:
then every solution yet) of (2.1.2) has one of the following properties:
u'(t)
u 2 (t)
aCt)
+ q(t)
eventually.
(a(t)x'(t))'
+ p(t)x'(t) + q(t)f(x(t)) =
0,
(2.2.1)
Chapter 2
16
Theorem 2.2.1. Assume that there exists a function p(t) E C2([to, 00),
JR+). If (2.2.1) is nonoscillatory, then the equation
(a(t)p(t)y' (t'
+ kQ*(t)y(t)
(2.2.2)
Q*(t)
and
h*(t)
(2.2.3)
= p(t)p(t) - a(t)p'(t).
(2.2.4)
2 k a(t)p(t)
= p(t)a(t)
w(t)
[f~~~t;
0.
Then, we have
w'(t)
j2(x(t
:S -p(t)q(t) - p(t)p(t)
f~~~~) + p(t)(a(t)h*(t' -
*) (a(t)p(t (f(x(t
x'(t)
+ ( P(t)
a(t) - 2kh (t)
= -Q*(t) - ka(t)p(t)
or
w'(t)
(f~~~;
+ h*(t) 2
+h
ka(t)p(t)
*)
(t)
= -Q*(t) -
0,
p(t)
(f~S;)
a(t~(t) w 2(t),
eventually.
Corollary 2.2.1. If there exists a function p(t) E C2([to, 00), JR+) such
that equation (2.2.2) is oscillatory, where the functions Q*(t), h*(t) are
defined in (2.2.3) and (2.2.4) respectively, then (2.2.1) is oscillatory.
Remark 2.2.1.
defined as
For the special case when f(x) = x and p(t) = 0, i.e., when (2.2.1)
is reduced to equation (2.1.1), we have the following interesting result.
17
(a(t)p(t)w'(t))'
+ Q(t)w(t)
(2.2.6)
= 0
is oscillatory, where
Q(t)
p(t) [q(t)
+ a(t)h 2(t)
and
h(t)
- (a(t)h(t))']
(2.2.7)
p'(t)
(2.2.8)
= - 2p(t)'
(a(t)x'(t))'
+ q(t)x(t)
0,
x(td
X(t2)
has a solution x(t) with x(t) -=I 0 on (tl' t2) and there exist two functions
p(t), PI (t) E C 2([tl' t2], JR+) such that
t2
{[a(s)p(s) - al(s)pds)](x'(s))2
+ (Ql(S) -
t\
x"(t)
+ t~x(t) =
(2.2.11)
(2.2.12)
18
Chapter 2
on the interval [1,00), where C and Cl are real constants. We know that
the Euler equation (2.2.11) is oscillatory if c > 1/4 and nonoscillatory
if c:-::; 1/4. Let p(t) = 1 and Pl(t) = 1/t2 for equations (2.2.11) and
(2.2.12) respectively. Then it follows from Corollary 2.2.2 that (2.2.12) is
oscillatory if Cl 2: C> 1/4 and nonoscillatory if Cl :-::; C :-::; 1/4.
Now, in what follows we shall assume that the functions p(t), PI (t) E
C([to, 00), lR+), Q, h, Ql and hI are defined in (2.2.7), (2.2.8), (2.2.9)
and (2.2.10) respectively,
00
/
-,-,.---..,.-ds
a(s)p(s)
00
(2.2.13)
(3(t) =
00
(31(t) =
00
(2.2.14)
(2.2.16)
(iv) There exists a function y(t) E C1([T,00),lR) for some T 2: to such
that for t 2: T
,
y2 (t)
(2.2.17)
Y (t) + Q(t) + a(t)p(t) :-::; O.
Proof (i) => (ii). Since equation (2.1.1) is
nonoscillatory by Lemma 2.2.2. Let w(t) be
any loss of generality, we may assume that
Define v(t) = a(t)p(t)w'(t)/w(t) for t 2: T.
,
V
(t)
v 2 (t)
+ Q(t) + a(t)p(t)
= 0
19
v(u) - v(t)
We claim for fixed
lu
t:::: T
lu
Q(s)ds +
v 2 (s)
a(s)p(s) ds = O.
that
00
v2(s)
--:--:-'--:--:-ds <
a(s)p(s)
v(u) for
r
iT1
u::::: T 1 ,
v 2 (s)
a(s)p(s) ds
(2.2.18)
= v(t) -
(2.2.19)
00.
Q(s)ds -
00
IT1
t
or equivalently,
(2.2.20)
v 2(u)
a(u)p(u)[l
v(u)
+ J;1 a(s2)~(~)dsl
> - --:--:'-':-:-
a(u)p(u)
w'(u)
w(u)
---
for
u:::: T 1 .
h~ a~2)~ls)dS)
::::
In(:~;/)
for
u::::T1 .
Using (2.2.20) in the above inequality, we obtain -w' (u) :::: w(Td(a(u)p(u))
for u:::: T 1 . Finally, integrating this last inequality, we find that w(u)----+
-00 as u ----+ 00, which contradicts the fact that w(t) > 0 for t:::: T.
Thus, (2.2.19) holds.
Now, letting u ----+ 00 in (2.2.18), it follows that v(u) ----+ v(oo) < 00.
Hence, in view of (2.2.19), v(oo) = O. This in (2.2.18) now implies (2.2.15),
i.e., (ii) holds.
(ii) => (iii). It is obvious.
(iii) => (iv). Suppose there exists a function v(t) E C([T, (0), JR) satisfying
(2.2.16). Define y(t) = (3(t) + JtOO v 2(s)j(a(s)p(s))ds. Then, Iv(t)l::::
ly(t)l, and
y' (t)
y2(t)
v 2 (t)
- Q(t) - a(t)p(t) < - Q(t) - a(t)p(t)'
20
Chapter 2
y(t)
~ ~(t) +
1= a~;;;~~)
y2(s)ds
for
~ T.
(2.2.21)
(iii) There exist T 2' to and a function z(t) E C([T, (0), JR) such that
z(t)
where
~(t) +
1 /-*
1
00
~(t) =
t]
a(s);(s) z2(s)ds
2' T,
f32(s)
a(s)p(s) M[S, t]ds
(1
and
for t
M[S, t] = exp 2
f3(T)
)
a(T)p(T) dT .
(2.2.22)
(2.2.23)
(2.2.24)
,2f3(t)
y2(t)
f32(t)
+ a(t)p(t) y(t) = - a(t)p(t) - a(t)p(t)'
Y (t)
(2.2.25)
(2.2.26)
The left-hand side of inequality (2.2.26) is independent of u, and hence
we may let u --> 00 in (2.2.26). This immediately leads to (2.2.21).
(ii) =} (iii). Suppose there exists a function y(t) E C([T, (0), JR) satisfying
(2.2.21). Define zo(t) = y(t), and inductively for n = 1,2""
(2.2.27)
21
Using (2.2.21) and (2.2.27), it is easy to show by induction that 0:::; ~(t) :::;
zn(t) :::; Zn-l(t) :::; yet). Thus, the sequence {zn(t)} has a pointwise
limiting function z(t) = limn~oo zn(t). Since the integrand in (2.2.27) is
nonnegative, it follows from the Lebesgue dominated convergence theorem
that z(t) satisfies the equation (2.2.22).
(iii) =? (i).
Assume that (2.2.22) has a solution z(t), then an easy
computation shows that vet) = z(t) + (3(t) defines a solution of (2.2.17).
As a consequence of Theorem 2.2.3, we conclude that equation (2.l.1) is
nonoscillatory.
00
y2(s)
1l*[S, tj a(s)p(s) ds,
(2.2.28)
(iii) There exist T;:;' to and a function z(t) E C([T,oo),lR) such that
z(t) = C(t)
00
z2(S)
1l*[S, tj a(s)p(s) ds,
where
00
C(t)
exp
(2.2.29)
e(s)
f1*[s, tj a(s)p(s)ds
(2.2.30)
(2.2.31)
and
11 [s, t] =
t;:;,
(3(T) + ~(T) )
a(T)p(T) dT .
y'(t) =
- a(t)p(t) [y2(t)
(2.2.32)
yet)
Il*[u,t]y(u)+
>
e(s)
Il*[s,t j a (s)p(s)ds+
*
~2(s)
11 [s,tja(s)p(s)ds+
y2(S)
Il*[s,t j a (s)p(s)ds
*
y2(s)
11 [s,tja(s)p(s)ds.
(2.2.33)
22
Chapter 2
Letting u
-->
(ii) =} (iii). Suppose there exists a function y(t) E C( [T, 00), JR) satisfying
(2.2.28). Define zo(t) = y(t), and inductively for n = 1,2, ..
Zn (t ) = ~ * ()
t
00
*[
fi s, t
z;
(2.2.34)
Using (2.2.27) and (2.2.34), it is easy to show by induction that O:s: C(t) :s:
Zn(t) :s: zn-dt) :s: y(t). Thus, the sequence {zn(t)} has a pointwise
limiting function z(t) = lim n ..... oo zn(t). Since the integrand in (2.2.34) is
nonnegative, it follows from the Lebesgue dominated convergence theorem
that z(t) satisfies (2.2.29).
(iii) =} (i).
Assume that (2.2.29) has a solution z(t), then an easy
computation shows that v(t) = z(t) + (3(t) + ~(t) defines a solution of
(2.2.17). As a consequence of Theorem 2.2.3, we conclude that equation
(2.1.1) is nonoscillatory.
Remark 2.2.2. It is clear from Theorems 2.2.4 and 2.2.5 that the process of generating higher order iterated Riccati integral equations can be
continued.
Now, using above nonoscillatory characterizations of solutions of equation (2.1.1), we shall derive the following comparison results.
Theorem 2.2.6. With respect to equations (2.1.1) and (2.1.2) assume
that there exist two functions p( t), pdt) E C 2([ to, 00), lR+) such that
(2.2.35)
If (2.1.2) is nonoscillatory, then equation (2.1. 1) is also nonoscillatory.
Proof. Since equation (2.1.2) is nonoscillatory, it follows from Theorem
2.2.3 that there exist T 2: to and a function v(t) E C([T,oo),lR) such
that
Iv( t)1
2:
1(31 (t) +
00
a1
~2)~:(S) ds 1
for
t 2: T.
(2.2.36)
Iv(t)1
2:
1(3(t) +
1= a~s~~ls)
dsl
for
t 2: T.
() = 0,
x /I (t) + a sin
t It
xt
(2.2.37)
i:- 0
+ 4t 2y(t) =
y"(t)
Let
p(t) = exp
Then as t
(2~
---4
It
23
co: "Is
dS),
17 =
t 2: 1.
0,
2~1
(2 + sgn
(2.2.38)
~)
and pdt)
1.
Thus, we have
as
---4
00
1
and (3I (t) = - .
4t
(32 (t)
a(t)p(t)
(3f(t)
<
al (t)PI (t)
(3(t)
(3I (t)
<
a(t)p(t)
al (t)PI (t)
for
for
t 2: to
t 2: to
(2.2.39)
(2.2.40)
y(t) = 6(t)
OO
MIls, t] 2
() ()y (s)ds,
al s PI s
where
6(t) =
(2.2.41)
and
Mds, t]
(2.2.42)
24
Chapter 2
From (2.2.39) and (2.2.40), we have 6(t)::::: ~(t) and p,Ils, t] ::::: p,[s, t],
where ~(t) and p,[s, t] are defined in (2.2.23) and (2.2.24) respectively.
Hence,
00
p,[s t]
a(s);(s)y2(s)ds.
<
~i(t)
a(t)p(t) and
al (t)Pl (t)
(2.2.43)
(2.2.44)
where ~(t), 6(t) and p,Ils, t] are defined in (2.2.23), (2.2.41) and
(2.2.42) respectively. If (2.1.2) is nonoscillatory, then equation (2.1.1) is
also nonoscillatory.
Proof. Since equation (2.1.2) is nonoscillatory, it follows from Theorem
2.2.5 that there exist T::::: to and a function y(t) E C([T, 00), JR) such
that
00
*
*
y2(S)
y(t) = ~1 (t) +
p,Ils, t] () () ds,
t
al s PI s
where
~;(t)
and
p,i[s,t] = exp
1
1
00
~i(s)
() ()P,ds,t]ds
al s PI s
(218 (/3r(T)+6(T))
dT).
al(T)p(T)
t
From (2.2.43) and (2.2.44), we have ~i(t)::::: C(t) and p,ils, t] ::::: p,*[s, t],
where C(t) and p,*[s, t] are defined in (2.2.30) and (2.2.31) respectively.
Hence,
1 ~:~~(;)
00
y2(s)ds,
t::::: T.
25
~(t)
Remark 2.2.3. It is clear from Theorems 2.2.7 and 2.2.8 that we can establish the higher order iterated comparison theorems by using the nonoscillatory characterizations.
q(t)
0 for
t ~ T.
q(t)
qt
+a
( ) ( x'(t)2
t
x (t)
x 2 (t)
q(t) - q(t) x 2(t)
o.
Conversely, if q(t)+a(t)h 2(t) - (a(t)h(t' ::; 0, then it follows from Corollary 2.2.2 that (2.2.6) is nonoscillatory because equation (a(t)p(t)w'(t' =
(-2 J~ h(s)ds).
Thus, equation
4c-1
--;w-.
x"(t)
= 0,
(2.3.1)
26
Chapter 2
Then, we have
and
1
4t 2
1 (1
1(1
t2
4c -1
4(tlnt)2
-"2
1)2
+ 4"
t + tInt
1)
(tlnt)2
1
(tlnt)2
+ t 2 lnt +
c:::; 4
when
to, and
t 2 q(t)
(2.3.2)
q(t)
+ a(t)h 2 (t)
4c -1
- (a(t)h(t))' :::; 4t;2a(t) - a'(t)h(t) :::; 0 for
t ~ T.
to, and
-~)
4t
< -41 ,
(2.3.3)
(4~2 + (t l~ t)2 )
for
~ T.
27
Let
h(t) =
- -1 (1-
h'(t) _
-
and
q(t)
(~
+ - 1) .
tIn t
_1_ + __1_)
(t t)2
t 2 + t 2 In t
In
- ~ C~
+ t2 ~nt +
(tl~t)2)] -
a'(t)h(t)
2: T.
In what follows, we shall assume that p(t) E C 2 ([to, 00), JR+) and
h(t) = -p'(t)/(2p(t)). Clearly, from Lemma 2.2.2 and Theorem 2.3.1 the
following result is immediate.
Theorem 2.3.2. Equation (2.1.1) is nonoscillatory if and only if there
exist T 2: to and a function v( t) E C 1 ([T, 00), JR) satisfying
Q(t)
0 for
2: T,
(2.3.4)
v(t) =
1
[
- 2a(t)p(t) 2'IjJ(t)
1 - 2k]
+ g(i)
.
28
Chapter 2
Then, we have
(a(t)p(t)v(t))'
-'l//(t)
p(t) [q(t)
(1
~:2~~r(t)
and
p(t) [q(t)
[2
1 - 2k] 2
+ ---g(i)
<
1 - 2k
)
2a(t)p(t)g2(t)
4(1-2k)
g(t) V;(t)
(1-2k)2
+ ---g(i)
2(1-2k)]
g2(t)
( ) (1 ) 2() [2
4k + 4k(1- 2k) + (1 - 2k) 2
- 2(1 - ]
2k)
4a t p t 9 t
4k -1
for t 2 T.
2 ( ) ::; 0
(
4a t ) p( t ) 9 t
x"(t)
where q(t) E C([to, 00), JR+),
result.
+ q(t)x(t) =
Proof.
(2.3.5)
'l/J(t)
0,
<
= ftoo q(s)ds.
v(t) =
- a(t)lp(t) ('lj;(t)
+ g~t))
29
Then, we have
- 'lj;'(t)
+ c:~~~~ >
Q(t)
+ a(t)p(~)g2(t)
and
Q(t)
+ a(t)p(t)v 2(t)
- (a(t)p(t)v(t))'
(g2~t) - g2~t))
::::: a(t)lp(t)
0 for
2: to
and
ft
OO
q(s)ds <
00,
Proof. Apply Corollary 2.3.6 with p(t) = 1, g(t) = 1 + ft~) l/a(s)ds and
'lj;(t)
= f t q(s)ds.
OO
h(t)
1(1 ).)
-2"
and
t+tlnt
= t(lnt)>.,
where
Q(t)
Let
g(t) =
1
1 A
1 _ ). (In t) and 'lj;(t) =
2), + 4c
A-I
4(1 _ ).) (In t)
.
).2 -
g(t)'lj;(t)
2), + 4c < ~
4(1 -).)2
- 4
).2 -
and
1
if c <-.
-4
Chapter 2
30
'IjJ'(t) >
g'(t)(m + 'IjJ(t))
g2(t)
- g(t)
Q(t)
(a(t)p(t)v(t))'
m + 'IjJ(t)
+ a(t)p(t)g2(t)
Q(t)
+ a(t)p(t)v2 (t)
- (a(t)p(t)v(t))'
< (m+'IjJ(t))2 _
-
a(t)p(t)g2(t)
m + 'IjJ(t)
m+'IjJ(t)
< m + 'IjJ(t)
a(t)p(t)g2(t)- -a(:-:t)-p(:-:t)'-'g2"""(---:-t) - a(t)p(t)g2 (t) =
o.
Corollary 2.3.10. Let p(t) E C 2([to, 00), JR+), g(t) E C1 ([to, 00), JR+)
and 'IjJ(t) E C 1([to, 00), JR). If limt->oo 'IjJ(t) exists, g'(t):s; -l/(a(t)p(t))
and 'IjJ'(t) 2": -g(t)Q(t) for t 2": to, then equation (2.1.1) is nonoscillatory.
lim
t->oo
(g' (t) -
()1 ( ))
a t pt
Q(t)
=
+ a(t)p(t)v(t) - (a(t)p(t)v(t))'
_1_
g2(t)
+ a(t)p(t)g2(t)
_ g'(t) = _1_ (1
g2(t)
g2(t)
+ a(t)p(t)
_, t ) < 0
9 ()
31
_ _ ~ (~
2p( t) 2 t
_ _ p'(t)
() -
_1_)
+ tin t
Q(t)
and
4c-1
4tlnt ~ O.
00
'lj;2(S)
1
a(s)p(s) ds ~ 4'lj;(t) ,
- a(t)p(t) 'lj;(t)
v(t)
<
'lj;2 (S )
00
a(s)p(s) ds .
and
- a(t)p(t)
1
= a(t)p(t)
+ 4'lj;2(t)/(a(t)p(t
Q(t)
+4
'lj;(t)
00
+4
[2
l'lj; (t)
a(t)lp(t) ['lj;2(t)
'lj;2()
s ds ] 2
a(s)p(s)
+ 8'lj;(t)
1t
4'lj;2(t)
---:-:-,
-'-:-'-:a(t)p(t)
roo 'lj;2(s) )2
Vt a(s)p(s) ds -
'lj;2(S)
(
a(s)p(s) ds + 16
00
4'lj; (t)J
for t 2: to.
= 0
h(t) = _p'(t) =
2p(t)
Let
_~ (~+~)
2
tint
4(1 _ ).)
4t
nt
)>---1
Chapter 2
32
Then, 'ljJ'(t) = -Q(t) and
00
'ljJ2(s)
a(s)p(s) ds =
2), + 4c
1
4(1 _ ).)2 'ljJ(t) :::; 4'ljJ(t).
).2 -
'ljJl(t) =
00
(IS
'ljJ2(S)
a(s)p(s) exp 2
'ljJ(r)
)
a(r)p(r) dr ds.
Then, we have
+ 4'ljJ(t)'ljJl(t) + 2'ljJ2(t)
a( t )p( t)
and
Q(t)
:::;
a(t)p(t) ['ljJ2(t)
:::;
(a(t)p(t)v(t))'
_ (4'ljJ(t)'ljJdt) + 2'ljJ2(t))
a(t)p(t)
0 for
t;::: to
and 'ljJl(t) be as in
=-
['ljJ(t)
1
Then, we have
Q(t)
33
to,
+ a(t)p(t)v 2(t)
- (a(t)p(t)v(t))'
+ 47/;r(t)]
1
[2
a(t)p(t)
7/;1 (t)
47/;12]
(t)
(~7/;l(t)) + 16 C167/;Ut))]
O.
L6(t)
+ c(t)1 ~
i(3(t)
00
for t
T,
(2.3.6)
00
for t ~ T
ci(t)
i{3(t)
00
(Ci(s))2 dsi
a(s)p(s)
for t
T,
34
Chapter 2
Corollary 2.3.17. If
00
rP(s)
1
a(s)p(s) ds :::; 4 1,8(t)l,
Now, let <f>(t) E C1([to, (0), JR) and p(t) E C 2 ([to, (0), JR+) satisfy
<f>'(t) = l/(a(t)p(t)), and let c(t) = (v'k/<l>(t)) - ,8(t), where k > 0 is a
constant. Then, limt--->oo <f>(t) = 00, ,8(t) + c(t) = v'k/<l>(t) and
1 ~(t)+IOO[,8(S)+C(s)j2dsl
a(s)p(s)
t
fJ
I~()
fJ
+ <f>(t)
I.
<f>(t),8(t) :::;
C(t) -
."
00
-~(t),
\,6(t)
and
[,6(t)
00
+ ,6(t)])1/2 -
+ ).~(t)1
+ c(t)]2 =
).,6(t)[).~(t)
35
+ ,6(t)]
~ [).~(t)
+ ,6(t)]2,
which implies (2.3.6). Now, it follows from Theorem 2.3.3 that (2.1.1) is
nonoscillatory.
00
e(t) ~
+ e(s)F
a(s)p(s) J.L[S, t]ds.
[~(s)
Theorem 2.3.5.
Let C (t) and J.L* [s, t] be as in Theorem 2.2.5.
Then equation (2.1.1) is nonoscillatory if and only if there exists a function
e*(t) E C([T, (0), lR) for some T ~ to such that
e*(t)
1= [C(~~s;p~:~S)]2
J.L*[S, t]ds.
Taking c(t) = ~(t) and c*(t) = ~*(t) in Theorems 2.3.4 and 2.3.5
respectively, we obtain the following corollary.
Corollary 2.3.21. Equation (2.1.1) is nonoscillatory if
1
4:~(t) ~
00
~2(s)
or
1
4:C(t) ~
1=
t
C 2 (s)
a(s)p(s) J.L*[S, t]ds.
(a(t)x'(t))'
where
+ q(t)f(x(t)) =
0,
and
(2.4.1)
and
f E
Chapter 2
36
0 for x#- O.
00
q(s)ds
(2.4.2)
00.
w'(t)
(2.4.3)
Itt,
J;
It:
a(t)x'(t)
a(T)x'(T) -
q(s)f(x(sds
a(T)x'(T) - f(x(t
f'(x(sx'(s)
(8
q(s)ds
q(U)dU) ds < a(T)x'(T).
x( t) ::; x(T)
t a(1s) ds
+ a(T)x' (T) iT
which is a contradiction.
--> -00
as t
--> 00,
37
Joo
and
1
--:---:---:-:-ds
a(s)p(s)
Joo Q(s)ds
(2.4.5)
00
(2.4.6)
00,
<
00,
if c >
Joo
and
p'(t)
2p(t)
---
h(t)
Joo
p(s) ds
Joo Q(s)ds
-l-ds =
s ns
00
4s In s ds
00
h'(s)] ds
If
c >
Thus all conditions of Theorem 2.4.2 are satisfied if c> 1/4, and hence
(2.3.1) is oscillatory if c > 1/4.
As an immediate consequence of Theorems 2.2.4 and 2.2.5, we have the
following interesting criterion for the oscillation of (2.1.1).
Corollary 2.4.1. Let ';(t) be defined in (2.2.23). If
(J2(s)
(Js
J oo a(s)p(s)
exp 2
or
(3(T)
)
a(T)p(T) dT ds =
e(s)
(2 JS (3(T)a(T)p(T)
+ ';(T) dT) ds
Joo a(s)p(s)
exp
00,
00,
38
Chapter 2
Example 2.4.2.
Consider the Euler differential equation (2.2.11) for
c> 1/4. We let p(t) = ta, where 1 - V4c - 1 ::; 0: < l. Then, we have
(3(t) =
0:
~ ~O:o:+ 4c t a - 1
J= (3:(~~)
and
exp (2
JS ~~~~ dT) ds =
00.
where
0:,
-y '" 0,
(0:
siwyt
t17
c) () _
+ -2
t
(2.4.7)
0,
Jt [~-
0:
~~:-YsJ dS),
exp
and
--+ 00,
2
(
2-2(7) [1 O( -2>-)]
(2Jt (3(S)d)
p(S) S =
2-y2(2A+2iJ-1)(2-2iJ)
+ t
(JS (3(T)
)
J=(32(S)
p(s)
p(T) dT ds =
0:
exp
exp
00.
=1
(0: 2/2-y2),
we let
<c<
l.
39
Then,
(3( t) -
2')'2(,),2 +,), + c) +
2')'2(1 + 2,),)
2 -2>'-1
0 ( -2>.-2)
t
p(s)
and
00
/
(/8 (3(T)
)
p(T) dT ds =
(32(s)
p(s) exp 2
as
-> 00,
(C 2>'-1)] as t -> 00
00.
u>
Case (iii). If
~2
~2'
~+
- 2
~}
4'
c-
Then,
(3(t)
(t- 2>.-2)
as
-> 00,
o (t- 2>.-1)]
(/8 (3(T)
)
p(T) dT ds =
(32(S)
p(S) exp 2
t->oo
as
00.
>
1 and
T 2': to.
t-Hx)
iT
00
(2.4.8)
iT
T 2': to
(2.4.9)
40
Chapter 2
Case 1.
{Tn};:"=l with
a(t)x'(t)
hmsup
t--->oo f(x(t))
< a(TN)x'(TN).
+ hmsup (- i t q(s)ds )
-
f(x(TN))
t--->oo
TN
< 0,
a(t)x'(t)
f(x(t))
--7-:~7-:-'-
a(t2)x'(t2)
f(X(t2))
t2
+ i t q (s )ds < 0 ,
t2
and from (2.4.8), we obtain lim inft--->oo a(t)x'(t)/ f(x(t)) = -00, which is
a contradiction.
Case 3. Suppose x'(t) < 0 for t ~ t3 ~ to. Condition (2.4.9) implies
that for any t4 ~ t3, there exists a T ~ t4 such that
q(s)ds ~ 0 for
all t ~ T. Choosing T ~ t3 as indicated, and then integrating (2.4.1), we
obtain
J;
q(s)f(x(s))ds
+ It f'(x(s))x'(s) is q(u)duds
J;
-->
tl.
00
as t
--> 00.
But,
-1
ds
00
(2.5.1)
t->oo litis
to to q(u)duds
lim -
(2.5.2)
00
41
it,t
w(s)ds
it is -(2-) duds
to to
w (u)
a u
(k - ! it is q(U)dUdS) ,
t
to
Thus for t
a u
to
to
for
to such that
t
~ tl.
tl it follows that
F(t)
to
a u
to
F2(t) <
(1:
(2.5.4)
W(S)dS) 2
(1~ a(S)ds)
-1
< F- 2(t)F'(t)
for
~ t2.
1: (1:
a(U)du) -1 ds
<
00,
(2.5.5)
Chapter 2
42
From Lemma 2.2.2 and Theorem 2.5.1 the following corollary is immediate.
Corollary 2.5.1. If there exists a function p(t) E C 2 ([to, 00), 1R+) such
that
j= (1: a(U)p(U)dU)
litis
and
lim -
t .....
=t
to
to
-1
ds
00
= 00,
Q(u)duds
(2.5.6)
(2.5.7)
litis
lim t .....
=t
to
til
Q*(u)dud.c;
00,
+ t12X'(t) + t~x(t) =
0,
t ~ 1
(2.5.8)
(~X'(t)' + t~ x(t)
= 0
(2.5.9)
(2.5.10)
43
and
1
lim sup -T
rt q(s )ds
iT
> - A
ftfs q( u )duds =
T
t-~oo
(2.5.12)
(2.5.13)
00,
Proof.
a(t)x'(t)
x(t)
rt () (x'(S))2
t ()d =
d
x(s)
s+ iT q s S
+ iT a s
a(T)x'(T)
x(T)
c. (2.5.14)
(t)
(a(t)) 1/2 X'
x(t) E
By Schwartz's inequality, we have
([
a(;~)(s) dS) ~
2
alt
It
r2 (T
)
L
,00.
(:(~}
a(s)
-k +
~
t
rt t
iT iT
q(u)duds
(2.5.15)
ds
k2t2
J;a(s)x'(s)/x(s)ds::.> -kt
c(t - T) < c,
t
Chapter 2
44
Case 3. Suppose there exists tl :::: T such that x'(t) < 0 for t:::: tl.
From (2.5.14) and (2.5.12), we have
_ a(t)x'(t) > _ (
')
_
c + /\
x (t )
it ()(X'(S))2
t,
a s
( )
X S
ds.
(2.5.16)
t2
t,
(X'(S))
a(s)
x(s)
1+
ds
+ A.
Multiplying (2.5.16) by
-(x'(t)/x(t))
-(c + A)
> In :((:/'
Remark 2.5.2. In the proof of Theorem 2.5.2 the three cases considered
depending on the sign of x' (t) can be replaced by only two cases depending
on the integral
ft
(x'(s)) 2
x(s)
ds
JT a(s)
11 1
liminf t-->oo
to
to
Q(u)duds > -
1
-
to
00
1t 1s
to
to
Q(u)duds <
00.
45
Theorem 2.5.3. If there exists a function pet) E C2([to, (0), JR+) such
that
lit
lim 2"
t--->oo t
to
(2.5.17)
a(s)p(s)ds = 0
and
-00
<
lil~~f~iti8 Q(u)duds
t
to
to
<
limsuP~iti8 Q(u)duds
t--->oo
to
:::::
00,
to
(2.5.18)
where Q(t) is defined in (2.2.7), then equation (2.1.1) is oscillatory.
Remark 2.5.3. In Theorem 2.5.3 if aCt) = pet) = 1, then (2.5.17) is
disregarded and (2.5.18) reduces to the well-known condition, namely,
-00
<
lit~~f ~t
it i8
to
to
it i 8
to
to
q( u )duds :::::
00.
(2.5.19)
The proof of Theorem 2.5.3 is contained in a more general result which
uses weighted-integral condition. For this, we assume that the function
pet) E C 2 ([to, (0), JR+) is given, h(t) = -p'(t)/(2p(t)) and Q(t) =
pet) [q(t) + a(t)h 2(t) - (a(t)h(t))'] .
Let n(p) be the set of all nonnegative locally integrable functions get)
on [to, (0) satisfying the condition
li~~p
(j 9(S)dS)
t
l-k
k E [0,1),
(2.5.20)
where
Gk(t) =
If Gk(oo) =
00,
g(s)
(f g(u)du)k
(f a(u)p(u)g2(u)du) ds.
(2.5.21)
then 9 E n(p).
Joo
a(s)p(s)g2(~)ds = O.
(t g(s)ds)
(2.5.22)
g(s)ds =
00.
(2.5.23)
Chapter 2
46
Members of the classes D(p) and Do(p) are called weighted functions with respect to p(t). It follows from (2.5.20) and (2.5.22) that
Do(p) c D(p).
For 9 E D = D(p), we define
_ I: g(u) IsuQ(T)dTdu
Is g(u)du
Ag(s, t) -
(2.5.24)
v (t)
v 2 (t)
+ Q(t) + a(t)p(t) =
(2.5.25)
t::=: T.
for
rt a(u)p(u)
v (u)
rt
du - is Q(u)du.
2
is
v(t) = v(s) -
(2.5.26)
(2.5.27)
00,
t-->oo
V2(S)
a(s)p(s) ds <
00.
v2(s)
a(s)p(s) ds
00.
(2.5.28)
(2.5.29)
tg (S)V(S)dS = V(T)
it -it is
i is
g(s)ds
g(s)
g(s)
to t, t ::=:
::=: T,
Q(u)duds
v2 (u)
a(u)p(u) duds
(i
g(s)ds -
it is
T
g(s)
v 2 (u)
a(u)p(u) duds.
(2.5.30)
47
r Q(s)ds - Jr
JT
v 2 (s)
a(s)p(s) ds.
I~ g(s)ds Ag(T, t) _
IT g(s)ds
IT g(s)ds
JT
+ 0(1)
Ag(T, t) - / : Q(s)ds
as
--> 00.
v 2 (s)
V(T) - Ag(T, t) = v(T) - Ag(T, t) - JT a(s)p(s) ds
+ 0(1)
as
--> 00.
(2.5.31 )
(2.5.32)
where k is as in (2.5.21). It follows from (2.5.27), (2.5.29) and (2.5.31)
that there exist two numbers tl and t2 with t2:;': tl :;,: T such that
Ittl
forall
t:;,:t2.
(2.5.33)
tl
V2 (u) d
u >
a(u)p(u)
-
y2 (s )
Itls a(u)p(u)g2(u)du .
it
tl
- F(t). (2.5.34)
Thus, we find
F'(t)
,\g(t)
g(t)y2(t)
a(s)p(s)g2(s)ds
(2.5.35)
(2.5.36)
Itl
and
o :::: ,\
jt g(s)ds
tl
Chapter 2
48
)-1
i t ) k (it
F k- 2(t)F'(t) 2: Fk(t)F'(t)
y2(t)
2: )..kg(t) ( t,g(s)ds
tla(s)p(s)l(s)ds
2: (1- k)
(1:
v2 (s)
( ) ( )ds.
T asp s
00
(2.5.37)
JTt 9 ()
s V (s )ds
I: g(s)ds
(2.5.38)
I: g(s)ds
. I: g(s) I: a(:)~~)duds
11m
t
< 00.
t-->oo
IT g( s )ds
Next, by Schwartz's inequality, we have
. II: g(s)v(s)dsl
o <
hm
t-->oo
<
hm
t-->oo
-'-----t:;:-----'-
. u:
IT g( s )ds
a(s)p(s)g2(s)ds) 1/2
t
U: a(s2)~(s)ds)I/2
ITg(s)ds
O.
t~oo
00.
(2.5.39)
49
Further, following Lemma 2.5.1 we can easily prove the following result.
Theorem 2.5.5. Assume that there exist two functions p(t) E C 2([to, 00),
JR+) and g(t) E n such that Iimt-+oo Gk(t) = 00 for some k E [0,1)
and limt-+oo A g (., t) = 00, then equation (2.1.1) is oscillatory.
h(t) =
- 2t - 2tlnt
4c - 1
Q(t) = 4tlnt
and
= tInt, then
t 2: T> 1.
for
= t-+oo
lim
t [Inlns -lnlnT]k-l
I
ds
T
S ns
= t-+oo
lim [lnlnt -lnlnT] =
00
and
lim Ag(T, t)
t-+oo
1) 1t 1
1)
4C -lim ( t-+oo
4
4C lim ( - t-+co
4
--->
00
as
--ds
sIn s
[Inlnt -lnlnT]
---> 00
when
> 4"
"
1
< hm
lilf k( t )
t-+()()
It IS
to
to
t-+oo k(t)
be such that
Q(u)
a (u ) P(u ) duds
1t IS
to
to
Q(u) duds
a(u)p(u)
k'(t)
(2.5.40)
:s:
00,
Chapter 2
50
then equation (2.1.1) is oscillatory.
Proof.
get)
no
Corollary 2.5.4. Suppose there exist two nonnegative bounded functions gl(t) and g2(t) on [T,oo), T 2': to satisfying Joo gl(s)ds =
00 = Joo g2(s)ds, and a(t)p(t)gl (t) and a(t)p(t)g2(t) are bounded. If
limt-->oo A g, (T, t) < limt->oo Ag2 (T, t), then equation (2.1.1) is oscillatory.
Proof. Let al and a2 be two numbers satisfying limt-->oo A g , (T, t) <
al < a2 < limt--> 00 A g2 (T,t). Let ret) = g2(t) for T::; t < tl, where tl
is such that Ar(T, it) > a2 and I~' r(s)ds 2': 1. Let ret) = gl(t) for
it ::; t < t2, where t2 is such that Ar(T, t 2) ::; al and I~2 r(s)ds > 2.
This is possible because
+ Itt,2 gl(s)ds
( J~29l(S)J;Q(U)dUdS)
J~2 gl(s)ds
A g , (T, t2)[1
+ 0(1)] + 0(1)
(
as
J~'
t
J~2gl(s)ds
00.
hm
t->oo
J;a(s)p(s)r2(s)ds
t
<ITr(s)ds)2
= o.
This implies that r(t) E no and limt->oo Ar(T, t) does not exist. Thus,
by Theorem 2.5.4, equation (2.1.1) is oscillatory.
Remark 2.5.5. It follows from the results presented above and Theorem
2.2.1 that in the hypotheses of each of the Theorems 2.5.2 - 2.5.5 the
function Q(t) can be replaced by Q*(t) defined in (2.2.3).
Example 2.5.3. Consider the differential equation
0,
7r
t>to>-2
(2.5.41)
51
(i)
(ii)
We take p(t) = 1/t 3 and g(t) = t, t ;:::: 7r/2. Then, we have h*(t) =
(p(t)p(t) - a(t)p'(t))/(2k a(t)p(t)) = 0, and hence Q*(t) = p(t)q(t)
(2 + cost - 2tsint)/(2v'i). Now, since
p(s)q(s)ds
i [0
i
~/2
to
sin s +
d(0(2
2yS
+ coss))
~/2
Vi(2
and
.t
ito g( s )ds
;::::
lt (l
g(s)
to
t -
+ cost) -
2(7r/2)1/2 >
Vi -
2(7r/2)1/2
p(u)q(U)dU) ds
to
1
(/2)2
7r
it;;:;
(sys-2(7r/2)
1/2
~/2
s)ds
00
as
t~oo
o ds/a(s) =
We note that in Theorems 2.5.2 - 2.5.5 condition
not required. We also note that condition (2.5.2) is equivalent to
11t
lim -
t-->oo
to
(t - s)q(s)ds =
t-->oo
tn
lt
to
00,
00
is
(2.5.42)
>1
(t - s)nq(s)ds =
00.
(2.5.43)
It is to be further noted that (d. Remark 2.5.3) in (2.5.42) the limit cannot
be replaced by an upper bound.
Now, we state the following oscillation criterion for the linear differential
equation
xl/(t) + q(t)x(t) = 0,
(2.5.44)
Chapter 2
52
where q(t) E C([to, (0), lR).
Theorem 2.5.6. If for some integer n
> 1,
limsup -1 i t (t - s)nq(s)ds
tn
t-+oo
(2.5.45)
00,
to
limsup
t-+oo
~n i \ t t
to
s)nq(s)ds <
00
n>
for some
l.
liminf
t-+oo
~
t(t-s)nq(s)ds ~
t n iT
n(T)
T~to
forevery
H(t, t)
0 for
to,
= {(t, s)
t ~ s ~ to}
t >s
-->
to
lR is a
(2.5.46)
for all
(t, s)
(2.5.47)
V.
If
lim sup H( 1
t-+oo
t, to
00,
53
C1([to, (0), IR+), and replace the functions a(t) and q(t) by a(t)p(t)
and Q(t) = p(t) [q(t) + a(t)h 2(t) - (a(t)h(t))'] respectively, where h(t) =
-p'(t)/(2p(t)).
Further extensions and improvements of these criteria to more general
equations of type (2.2.1) can be obtained by applying Theorem 2.2.1. For
this, we let p(t) E C1([to,00),IR+), and replace the functions a(t) and
q(t) by a(t)p(t) and kQ*(t) where Q*(t) = p(t)[q(t) + a(t)h 2*(t) (a(t)h*(t))'], h*(t) = (p(t)p(t) - a(t)p'(t))/(2ka(t)p(t)) and the constant
k is as in (i2), i.e., f'(x) 2 k > 0 for x # o.
In the following result we extend Theorem 2.5.8 to equation (2.1.1).
Theorem 2.5.9. Let the functions Hand h be as in Theorem 2.5.8
such that (2.5.46) and (2.5.47) hold. If
lim sup
t->oo
(1
H t, to
It
to
[H(t, s)q(s) -
~h2(t, s)a(s)]
ds
00,
(2.5.48)
rt
iT H(t, s)q(s)ds
rt
rt
v2 (s)
H(t, T)v(T) - [
H(t, T)v(T) -
11t
+-
iTt
[h(t, s)JH(t, s)
[(H(t
a(;)s
+ H(t, s) :(~)]
ds
h2(t, s)a(s)ds.
H(t, To)v(To)
Chapter 2
54
To it follows that
::;
in
~h2(t, s)a(s)l ds +
J
rt fH(t, s)q(s) -
JTJ
[in
~h2(t, s)a(s)l ds
J
Iq(s)lds + Iv(To)1 .
This gives
lim sup H( 1
t, to
t-->oo
to
Now, in view of Remark 2.5.6 we can state the following more general
results.
Theorem 2.5.10. Let H, h be as in Theorem 2.5.8 and let there exist
a function p(t) E C1([to,00),lFt+) such that
lim sup
t-->oo
(1
H t, to Jto
[H(t,S)Q(S) -
~h2(t,s)a(s)p(s)]
ds
4
= 00, (2.5.49)
lim sup H( 1
t-->oo
t, to Jto
(2.5.50)
where Q*(t) and k are defined in Remark 2.5.6, then equation (2.2.1) is
oscillatory.
hto
Let R(t) =
dsja(s). If there exists a function
p(t) E C1([to, 00), IR+) such that for some n> 1,
Corollary 2.5.5.
00,
(2.5.51 )
then equation (2.2.1) is oscillatory.
55
Proof.
lim sup H( 1
t, to
t-+CXJ
t-+oo
ltDt
[H(t,S)Q*(s) - 41ka(s)P(S)h2(t,s)] ds
ltD
00.
ltD
1, then we have
a(s)p(s)h2(t,s)ds
~(R(t) -
n- 1
R(to))n-l
for
~ to
Corollary 2.5.6. If
1 )
lim sup Rn(
t
t-+oo
it
to
(R(t) - R(s)tQ*(s)ds
00
t (R(t) -
ltD
R(s))nq(s)ds =
Let Rl(t)
Corollary 2.5.7.
n> 1,
00
for some
00,
t ~ to.
n> 1.
If for some
+ a(t)h 2(t) -
(a(t)h(t))']
q(t)Rdt) - 4a(t)Rl(t)
and
h(t,s)
n
(Rl(S))(n-2)!2
a(s)Rl(S) In Rl(t)
for
> s ~ to
Chapter 2
56
a(s)p(s)h2(t,s)ds
17,
to
to
2 (
( )n-2
1 R IS
1
d
n Rl(t)
a(s)Rl(S) s
2
(
R (
n-l <
_17,_
ln~
n-1
R 1 (t)
00
It is easy to check that (2.5.49) implies condition (2.5.53), and hence equation (2.1.1) is oscillatory.
I
Next, we let H(t, s) = (t - s)n for t 2: s 2: to, where n > 1 is a
constant. Then from Theorem 2.5.11 the following corollary is immediate.
Let n > 1 be a constant and there exist a function
Corollary 2.5.8.
it
to
7 , 2 ] ds
(t - s)n-2 (t - s)2Q*(S) - 1
-ka(s)P(s)
00,
(t 2x'(t' - tx'(t)
+ cf(x(t
0,
t 2: to 2: 1
(2.5.54)
where c > 0 is a constant and f(x) is as in Example 2.5.3 with f' (x) 2:
k > 0 for x ~ O. Let p(t) = lit, t 2: 1 and n> 1 a number. Then, we
find a(t)p'(t)-p(t)p(t) = 0, and hence h*(t) = 0 and Q*(t) = cit, t 2: 1.
Now,
limsup
t~oo
it [(t to
li~~p t~
it
[(t - s)n
(~) - ~~(t -
Since n> 1 it follows from Hardy et. al. [10] that (t - s)n 2: t n - nst n - 1
for t 2: s 2: 1. Thus,
lim sup n
C
t~oo
jt -(t - s)nds
1
tIS
jt tn _ 17,st
s
n(t -1)]
n- 1
climsup [lnt _
t~oo
ds
=
00.
Also, we have
n2
4(n -1) (t _l)n-l
57
and hence
lim sup ~ {t [~(t _ st _ n 2 s(t _ s)n-2] ds
t-HOC
tn ) 1 s
4k
=climsup [lnt_ n (t-1)]_
t
t~oo
(n )k =00
4 n-1
if c>O.
(tx'(t))' +x'(t)
+ "if(x(t)) =
0,
2: 1
(2.5.55)
Joo dsla(s)
= 00,
(II)
for some
It
to
n>
00
1.
t--->oo
to
~4
(2.5.56)
Cbapter 2
58
. { . . H(t,s)}
o < s~to
mf hm
mf H( t, to )
t--->oo
(2.5.57)
00.
Suppose there exist two functions p E C 1 ([to: oo),Rn and DE C([to, (0),
R) such that
it
i
lim sup H( 1 )
t--->oo
t, to
to
oo
to
(2.5.58)
00,
(2.5.59)
00
to,
(1 )
H t, T
s)a(s)p(s)] ds ~
iTrt [H(t, s)Q(s) - ~h2(t,
4
= max{D(t), O},
D(T),
(2.5.60)
then equation
v (t)
v 2 (t)
+ Q(t) + a(t)p(t)
= 0
t ~ To ~ to
for
H(:,T)
it
= v(T) - H(t, T)
for t
~a(s)p(S)h2(t,S)] ds
[H(t,S)Q(s) -
H t
(a(s~;;;))
1/2
v(s)
+ 2v'a(s)p(s)h(t, s)
2
]
ds
To. Consequently,
li~~p H(:, T)
it
=v(T)-h~~f
1
H(t,T)
..
[H(t, s)Q(s) -
~a(s)p(s)h2(t, s)] ds
2
H(t,s) 1/2
1
]
[(a(s)p(s)) v(s)+2v'a(s)p(s)h(t,s) ds
59
for all T:::: To. This shows that for T:::: To,
v(T) :::: Q(T)
(2.5.61)
and
t.-,oo
liminf
H t, T
r (()
'( ))
asp s
t
t~oo
<:::
H t, To
..
H (t s)
1/2
JT
Hence,
. . [ (1
hmmf
v(s)
+ -Va(s)p(s)h(t,s)
2
1t
v 2 (s()ds+
)
H(t,s) ()
(1 )
To
asp s
H t, To
JTo
1t
To
ds <
00.
h(t,s)VH(t,s)v(s)dsJ
[(H(t'S))1/2
1
a(s)p(s)
v(s)+ 2va(s)p(s)h(t, s)J ds <
00,
(2.5.62)
i.e., we have
liminf [U(t)
t ..... oo
where
1
U(t)
H(t, To)
and
W(t) =
1t
1t
(1 )
H t, To
roo
JTo
~)
To
+ W(t)] <
H(t,s)
(2.5.63)
00,
v 2 (s)
a(s)p(s)
ds,
t:::: To
t:::: To.
v 2 (s)
a(s)p(s) ds <
00.
v 2 (s)
a(s)p(s) ds =
00.
(2.5.64)
roo
JTo
(2.5.65)
s"2to
> T/ > O.
(2.5.66)
Chapter 2
60
h
t
To
Therefore, for t
U(t)
V2(S)
--:-:--,--:--:-ds
a(s)p(s)
T1 ,
1
H( t, 1'.)
0
>
To
H(t,s)d
for all
TJ
(h
T1 .
V2(T)
)
(
) ( )dT
a T p T
To
2(T) dT) ds
(r
iTo a(T)p(T)
2(T) dT) ds
rt -~H(t ' s) (r
iTo a(T)p(T)
H(t, To) iT! as
1
H(t, To)
>
ht
J.L
iTo
-~H(t
as
'
s)
( ~)
TJ
H(t, To)
iT!
-~H(t
as'
s)ds =
(~)
TJ
H(t, T 1)
H(t, To) .
By (2.5.66), there exists a T2 ~ Tl such that H(t, Td/ H(t, to) ~ TJ for
all t ~ T2, which implies that U(t) ~ J.L for all t ~ T2. Since J.L is
arbitrary
lim U(t) = 00.
(2.5.67)
t->oo
Next, consider a sequence {tk}k=l in (to, (0) with limk->oo tk = 00
satisfying limk->oo[U(tk) + W(tk)] = lim inft->oo[U(t) + W(t)]. In view of
(2.5.63), there exists a constant M such that
U(tk)
+ W(tk)
for
k = 1,2, .. '.
(2.5.68)
(2.5.69)
(2.5.70)
k->oo
00.
U(tk)
< ~ < ~
-
U(tk)
provided k is sufficiently large. Thus, W(tk)/U(tk) < -1/2 for all large
k, which by (2.5.70) ensures that
. w2(tk)
I1m
-::-::--:-'--:-'k->oo U(tk)
00.
(2.5.71)
61
But (2.5.66) guarantees that lim inft--->oo H(t, To)/ H(t, to) > TJ. This means
that there exists a T3 ~ To such that H(t, To)/ H(t, to) ~ TJ for every
t ~ T3. Thus, H(tk, To)/ H(tk, to) ~ TJ for all large k, and therefore,
W2(tk)
1
itk
U( ) ::;
H(
)
a(s)p(s)h2(tk' s)ds for all large n.
TJ
tk, to to
tk
It follows from (2.5.71) that
=
(2.5.72)
00.
00
To
(D+(s2
a(s)p(s)
~--:-'-7-'-:-- ds
<
00
To
v 2(s)
ds <
a(s)p(s)
00,
which con-
00,
limsuP~lt(t-s)n-2a(s)p(s)dS
n
t--+oo t
to
<
00
to,
2
Chapter 2
62
Theorem 2.5.13. Let Hand h be as in Theorem 2.5.8 and conditions (2.5.46) and (2.5.47) hold. Suppose there exist two functions p(t) E
C 1([to, 00), JR+) and D(t) E C([to, 00), JR) such that condition (2.5.59)
hold,
liminf
T:::::
t-+oo
(1
H t, to
to,
iT
t, T
t-->oo
(2.5.73)
to
~h2(t,
s)a(s)p(s)]
4
v(T) -
d,
Consequently,
sf
d,
,r
d'
for all T::::: To. Hence, (2.5.61) holds for all T::::: To and
.
t[(H(t'S))1
iTo
a(s)p(s)
/2
v(s)
+ 2yia(s)p(s)h(t, s)
]2 ds
::;
iI~~p
+ W(t)]
H(t,s) 1/2
H(t, To)ho (a(s)p(s)) v(s)
1
+ 2yia(s)p(s)h(t, s)
ds < 00,
(2.5.75)
63
where U(t) and Wet) are defined in the proof of Theorem 2.5.12. By
condition (2.5.74),
n(to)
< liminf H( 1 )
t--->oo
liminf H( 1
:::::
[H(t,S)Q(S) -
t, to ltD
t--->oo
~h2(t,s)a(s)p(s)]
4
ds
t H(t,s)Q(s)ds
t, to ltD
-~4 lim
inf H( 1 ) rt h 2(t, s)a(s)p(s)ds.
t--->oo
t, to lto
From the above inequality and (2.5.73), we have
liminf H( 1 )
t--->oo
t h2(t,s)a(s)p(s)ds <
t, to ltD
00.
k--->oo
tk, to lto
00
h 2(tk, s)a(s)p(s)ds
liminf H( 1
t--->oo
t, to lto
(2.5.76)
h2(t,s)a(s)p(s)ds <
00.
(t>'x'(t))'+w' cost)x(t)
for
t:2:to>O,
(2.5.77)
where A and I-l are constants, A::::: 1, - 1 < I-l ::::: 1 and 21-l + 1 :2: A.
Let pet) = 1 and H(t, s) = (t - s)2 for t:2: s :2: to. Then, we have
limsuPt--->oo(1/t 2 ) ft:(t - s)2sILcossds = -tbcosto < 00, and
liminf;
t--->oo t lT
64
Chapter 2
= -TP.
D(T)
sin T - kl
L
00
82
00
s2P.- A ds
(2n+I)7r+(7r /4)
N=n
>
12(n+I)-(7r /4)
82
1 2 (n+I)-(7r/4)
ds
00.
(2n+I)7r+(7r/4) S
N=n
Thus, all conditions of Theorem 2.3.13 are satisfied, and hence equation
(2.5.77) is oscillatory.
The above oscillation criteria as well as most of the known results in the
literature require information of equations (2.1.1) and (2.2.1) on the entire
half-line [to, 00). Now, motivated by the Sturm separation theorem, we will
present oscillation results for equations (2.1.1) and (2.2.1) which depend on
the behavior of the coefficients of these equations only on a sequence of
subintervals {[ai, bi]}~l of [to, 00), no matter how 'bad' the coefficients
are on the remaining parts of [to, 00 ).
In what follows, we say that the function H = H(t, s) E r if H is as
in Theorem 2.5.8, i.e., if HE C(V,lRf), where V = {(t, s): t ~ s ~ to}
which satisfies (2.5.46) and has partial derivatives oH/ot and oH/os on
V such that
otH(t,s) = hl(t,s)..jH(t,s)
osH(t,s)
and
,
v (t)
v 2 (t)
+ Q(t) + a(t)p(t)
= 0
for
t ~ T,
(2.5.79)
(i)
Ib
c
+~
4
Ib a(s)p(s)h~(b,
c
s)ds.
(2.5.80)
Ic
65
(ii)
1Ic
+4
jcbH(b, s)Q(s)ds
Jb
+~
jb H(b, s) a(s)p(s)
v2
(s)
ds
Jb h~(b, s)a(s)p(s)ds
c
(ii) Similar to part (i), we multiply (2.5.79) by H(s, a), integrate it with
respect to s from a to e and use (2.5.46) and (2.5.78), to get
Ic
H(s, a)Q(s)ds
- H(e, a)v(c)
=
Ic
c
+I
+4
1I
- H(e, a)v(e)
H(s, a)v'(s)ds -
v 2 (s)
H(s, a) a(s)p(s) ds
hies, a)a(s)p(s)ds
-t [C7~;~~)) 'I'
vi')
1Ic
a
v (s) ds
H(s, a) a(s)p(s)
Ic
Ic
+ ~v'a(S)p(')h'(S,a)r ds
hies, a)a(s)p(s)ds.
Corollary 2.5.11. Let wet), vet) and T be defined as above. Then for
any HEr and a, b, c E lR such that T:S a < e < b,
H(~, a)
<
H(s, a)Q(s)ds
~ [H(~,a) l
+ H(~, e)
lb
hi(s,a)a(s)p(s)ds +
H(b, s)Q(s)ds
H(~,c) lb h~(b'S)a(S)p(S)dS].
(2.5.82)
66
Chapter 2
1 a)
H(c,
H(s, a)Q(s)ds
1
> 4"1 [ H(c,a)
+ H(b,1 c)
lb
C
2
hl(s,a)a(s)p(s)ds
H(b, s)Q(s)ds
1
+ H(b,c)
lb
c
2
]
h2(b,s)a(s)p(s)ds
,
(2.5.83)
Proof. Without any loss of generality we assume that w(t) is an eventually positive solution of equation (2.2.6). Then from Corollary 2.5.11 there
exists T ~ to such that inequality (2.5.82) holds for any HEr and
a, b, c E lR satisfying T <::: a < c < b. But, this contradicts (2.5.83).
t H(s, r)Q(s)ds
and
IT hi(s, r)a(s)p(s)ds
1t H(t, s)Q(s)ds
IT h~(t, s)a(s)p(s)ds
to,
00
(2.5.84)
00,
(2.5.85 )
11
In (2.5.85) we choose
lb
= c.
hi(s, a)a(s)p(s)ds.
(2.5.86)
lib
h~(b, s)a(s)p(s)ds.
(2.5.87)
67
Theorem 2.5.16. If there exists HEr such that for any r 2: to,
and
li~~p
it
[H(t,s)Q(S) -
~hi(s,
r)a(s)p(s)]
4
ds > 0
(2.5.88)
~h(t'S)a(s)p(s)]
ds > 0,
(2.5.89)
[H(S, a)Q(s) -
~hi(s, a)a(s)p(s)] ds
> O.
(2.5.90)
Ib
[H(b,S)Q(S) -
~h~(b,S)a(S)p(s)] ds
> O.
(2.5.91 )
1ic
h2(s-a)[a(s)p(s)+a(2G~s)p(2c-s)]ds,
(2.5.92)
J: H(b -
lb
s)Q(s)ds =
J; H(s -
J:
J;
a(s)p(s)h2(b - s)ds =
Thus (2.5.92) implies that (2.5.83) holds for HEro, and therefore equation (2.1.1) is oscillatory.
68
Chapter 2
(I)
and
limsupit [(t - s)nQ(s) - n 2 (t - s)n-2 a(s)p(S)] ds > 0,
t--->oo r
4
or
(II)
t [ (s
- r)n(Q(s)
x (a(s)p(s)
n
+ Q(2t - s - -(s
- r)n-2
4
O.
I:
li~~p Rn-l(t)
it
n2
(R(s) - R(rnq(s)ds > 4(n - 1)
(2.5.93)
li~~p Rn~l(t)
it
4(nn~ 1)'
(2.5.94)
00
and
= 1.
Then, Q(t)
h 1 (t, s)
and
~(R(t) - R(s(n-2)/2
a(t)
~(R(t) - R(s(n-2)/2.
a(s)
69
Next, since
limsup
t~oo
]t l(R(s)-R(r))nq(s)
r.
n
- T(R(s)-R(r))n-l a(s)
]t (R(s) - R(r))nq(s)ds
Rn-l( )
1
= li~~p Rn-l(t)
= lim sup
t~oo
- lim
t~oo
~a(s)hi(s,
r)] ds
4
2
1 ]
ds
Rn
17,2
( _ )
4 n
Remark 2.5.8.
l. Special cases of the above oscillatory interval criteria for equation (2.l.1)
when a(t) = 1 and/or p(t) = 1 can be easily formulated. The details are
left to the reader.
2. Extensions of the above results to more general equations of type (2.2.1)
can be easily formulated (see Remark 2.5.6).
The following examples illustrate how the theory can be applied in practice.
x"(t)
+ q(t)x(t) =
0,
(2.5.95)
q(t) =
where m E IN. For any T 2:: 0 there exists m E lN such that 3m 2:: T.
Let a = 3m, c = 3m+l, H(t-s) = (t-s?, t 2:: s 2:: T and p(t) = 1 for
t 2:: T. Then, Q(t) = q(t) for t 2:: T. It is easy to check that inequality
(2.5.92) holds, and hence every solution of equation (2.5.95) is oscillatory
by Theorem 2.5.14. We note that in this equation fooo q(s)ds = -00.
70
Chapter 2
t:::: r
t~=
Rn
1 )
-l(
t
it
r
(R(s) - R(r)tq(s)ds
it
it
1
lim R n -l( t )
t~=
c lim -1-
rs n -
t~oo
tn -
and
1 )
lim R -l(
n
t
t~oo
it
r
c
lim - (s - rt 2s ds :::: c t~=
tn - 1
2 -
nTs n - 3 ] ds
tn -
it
r
c hm
--1
tn -
t~oo
tit] ds
t n-l [ - - s2
s2
ds
n- 1
t~oo
sn - nrs n - 1
-C-
(R(t) - R(s))nq(s)ds
. -1c hm
it
. [t-
c Inn
t~oo
it
r
tn
nst n s
t]
- 1 - n In r
ds
oc.
condition (2.5.93) holds if c/ (n-1) > n 2 /[4( n-1 )]. Also condition (2.5.94)
is automatically fulfilled. Thus, Corollary 2.5.13 implies that the equation
(2.2.11) is oscillatory when c> 1/4.
--:-,.....--:---:-ds
a(s)p(s)
and
f3(t)
loo
-p'(t)/(2p(t)).
(2.6.1)
00
+ a(t)h 2 (t)
(2.6.2)
Zn(t)
zo(t)
00
(Z~_1 (s))2
a(s)p(s) ds,
= 1,2,
71
where z+(t) = (z(t) + Iz(t)I)/2. Clearly, Zl(t) 2 zo(t) and this implies
that zi(t) 2 z6(t). Thus, by induction, we have
Zn+l(t) 2 zn(t),
(2.6.3)
t2to
n=1,2,,
for
n->oo
2 tl.
(2.6.4)
Proof. Suppose equation (2.1.1) is nonoscillatory. Then in view of Theorem 2.2.3 there exists v(t) E C([tl, 00), IR) such that
v(t) = f3(t) +
00
v2(S)
a(s)p(s) ds
on [tl,oo) for some tl 2 to Thus, v(t) 2 zo(t) for t 2 tl, and hence
v+(t) 2 z6(t) for t 2 tl. This implies
f3(t)
v(t)
>
+.t
zo(t)+
00
V2(S)
a(s)p(s) ds 2 zo(t) +
1=
t
(Z6(8))2
() ()d8 = Zl(t)
asp.5
00
for
(v+(s)J2
a(s)p(s) ds
t2tl.
v(t) 2 zn(t),
= 0,1,2,
for
(2.6.5)
t 2 tl.
{zn(t)}
is
(i)
there exists a positive integer m such that zn(t) is defined for
n= 1,2,,m-1, but zm(t) does not exist, or
(ii) zn(t) is defined for n = 1,2, .. , but for arbitrarily large T* 2 to
there is t* 2 T* such that
00,
(2.6.6)
72
Chapter 2
Theorem 2.6.3. Suppose zo(t):::: 1,B(t)1 for t:::: to. If there exists a
tl :::: to such that (2.6.4) holds, then equation (2.1.1) is nonoscillatory.
Proof. From (2.6.3) and (2.6.4) it follows that zn(t) ~ z(t), n = 0,1,
2, . .. for t:::: tl. Applying the monotone convergence theorem, we find
z(t) = zo(t)
Thus, we have
00
z+(t) = z(t)
>
00
00
(z+(s))2
a(s)p(s) ds :::: 0 for
(Z+(S))2
( ) (S ) ds
asp
00
+ Zo (t ) ::::
(z+(s))2 ds + ,B(t) I
a(s)p(s)
for
t:::: t I
(Z+(S))2
( ) (s ) ds
asp
+ I,B (t ) I
t:::: ti.
The next result provides necessary conditions for the oscillation of equation (2.1.1).
Theorem 2.6.4. Suppose zo(t):::: 1,B(t)1 for t:::: to. If equation (2.1.1)
is oscillatory, then either
(i)
there exists a positive integer m such that Zn (t) is defined for
n = 1,2,, m - 1, but zm(t) does not exist, or
(ii) zn(t) is defined for n = 1,2,, but for arbitrarily large T*:::: to
there is t*:::: T* such that (2.6.6) holds.
If ,B(t) :::: 0 for t:::: to,
corollaries are immediate.
Corollary 2.6.1. Suppose zo(t) = ,B(t) :::: 0 for t:::: to. Then equation
(2.1.1) is nonoscillatory if and only if there exists a h :::: to such that
(2.6.4) holds.
Corollary 2.6.2. Suppose zo(t) = ,B(t) :::: 0 for t:::: to. Then equation
(2.1.1) is oscillatory if and only if either
there exists a positive integer m such that zn(t) is defined for
(i)
n=1,2,,m-1, but zm(t) does not exist, or
(ii) zn(t) is defined for n = 1,2,, but for arbitrarily large T* :::: to
there is t*:::: T* such that (2.6.6) holds.
Example 2.6.1. Consider the differential equation
XIl(t)+~C4/3(2+cost+3tsint)x(t)
0,
t>O.
(2.6.7)
zo(t)
['>0 q(s)ds
f3(t)
73
= e l / 3 (2 + cost) > e l / 3 ,
t> O.
(z+(s))2
o
ds >
a(s)p(s)
-
00
and hence
2.6.2.
Tn
00
s-2/3ds =
00,
~ (e 3/4sin
x" (t) +
zo(t) = f3(t) =
~e5/4 cos
0)
> o.
t>
x(t)
0,
(2.6.8)
1. Then,
q(t) = Q(t) =
Now, we let
1
~ 1=
00
o.
q(s)ds
[S-3/4 sin v's +
l / 4 cos Vi.
zI(t) =
11 Vs1
1
> -
1[(2n+3)7r/2]2 [ 1
00
2n=k
00
[(4n-l)7r/2J2
r57r /2
L j:{37r/2
00
[(2n+l)7r/2J2
r;:;
VS
(cos 2
j.[(4n+l)7r/2J2 cos2 Vs
n=k
00
"2
Vs
ds
v's +
]
cos v'sl cos v'sl) ds + zo(t)
+ zo(t)
00.
n=k
Tn
= 1,
and
74
Chapter 2
Now, we shall consider equation (2.1.2). Suppose Qo(t) E C([to, (0), IR)
is a given function and there exists a function p( t) E C 2([ to, 00 ), IR+) such
that
Ih(t)
1=
Q1(s)ds
and
00,
Qn(t)
Qo(t)
00
(Q~_1(S))2
a1
{Qn(t)}~=o
()
() ds,
s PI s
17, = 1,2,,
t :;;, to.
(2.6.9)
Clearly, Ql(t):;;, Qo(t) and this implies that Qi(t) :;;, Qt(t), t:;;, to Now,
by induction, we have
t:;;, to
17, = 1,2,,
(2.6.10)
o<
Assume that
IfJl(t)l::;; f3(t)
t. (2.6.11)
If equation (2.1.1) is nonoscillatory, then (2.1.2) is nonoscillatory, or equivalently, if equation (2.1.2) is oscillatory, then (2.1.1) is oscillatory.
Proof. Let zo(t) = f3(t) and Qo(t) = 1f31(t)1 for t:;;, to. Suppose
equation (2.1.1) is nonoscillatory. It follows from Theorem 2.6.1 that there
exists a t1 2 to such that (2.6.4) holds. Clearly, by (2.6.11), Qo(t) =
1f31(t)1 ::;; f3(t) = zo(t), and hence Qt(t)::;; zt(t) for t 2 t 1. Thus, for
Qo(t)
< Qo(t)
1 t)
+ 1 a(~)p(s)
+
00
(Q+(S))2
()ds
a1
00
s PI s
(0+(S))2
ds
17,=0,1,2,,
t2 t 1
(2.6.12)
Q(t)
n----+CXl
lim zn(t)
n-+oo
= z(t) <
00,
t:;;, iI.
75
(j
(3+(s)
)
a(s)p(s) ds
<
(2.6.13)
00,
yl (t)
a(t)p(t)
We claim that
I
Y (t)
:::
((3(t) + y(t))2
a(t)p(t)
v 2 (t)
(3+ (t)
for
t 2 T.
(2.6.14)
t 2 T.
0,
ht ats~~~~)
y(t) =
00
v(t) = (3(t)
+ y(t) >
+ y(t) 2
v2(s)
a(s)p(s) ds >
-
This gives
zo(t)
00
t 2 T.
dS) ,
00
(2.6.15)
t) ()
(Z+(S))2
asps ds.
(z+(s))2
a(~)p(s) ds
Zl(t)
for
t2T
and by induction
v(t) = (3(t)
+ y(t) 2 zn(t),
n = 0, 1,2,,,,,
t 2 T.
(3+(s)
)
(-4 Jrt a(s)p(s)
ds ,
T
(2.6.16)
Chapter 2
76
and hence
( t
,8+(s)
)
(Zn(t) - ,8(t)) exp 4 iT a(s)p(s) ds
< y(T),
n=0,1,2,,
t ~ T.
lim (zn(t) -
n-->oo
(4 [
a~s~~~~) dS)
,8(t)) exp (4 t ~~(~)) dS) ::; y(T) <
iT asp s
00
for t
~ T.
(i)
(j
4
,8+(s)
)
a(s)p(s) ds
00,
or
(ii)
(4 j
,8+(s)
)
a(s)p(s) ds
00,
j oo
exp -4
j8 a(u)p(u)
,8+(u)
)
du ds <
00,
(2.6.17)
joo zm(s)ds
(2.6.18)
00,
(-4 [
a~s~~~~) dS),
n = 0,1,2"",
~ T.
(2.6.19)
JTroo zn(s)ds
:::; y(T)
we find
--+ 00,
f3+(u)
)
JTr= exp (-4 JtT a(u)p(u)
du
77
ds+
JTroo f3(s)ds
<
00,
Joo
J=
Theorem 2.6.9. If
J=
(J8
1
a(s)p(s) exp -4
f3+ (T)
)
a(T)p(T) dT ds <
00,
(2.6.20)
v(t)
f3(t)
1=
t
v2(s)
a(s)p(s) ds
for
t::>: T,
Joo
Jt
( 1 w(t))2
n w(T)
J=
(tJ
: :; (tJT
T
v(s) d)2
a(s)p(s) s
ds
)
k(s)a(s)p(s)
oo
ds
a(s)p(s)w2(s)::>:
which is a contradiction.
c6
J=
(tJT
k (S)v 2(s) )
a(s)p(s) ds ,
Iw (t) I :::;
1
a(s)p(s)ex P (- 2c IJlnk(s))ds
00,
Corollary 2.6.3. If
1
(Js a(T)p(T)
f3(T)
)
Joo a(s)p(s)
exp -6
dT ds
<
00
Chapter 2
78
1 2(A-1) t
Q(t) = (>.2 - A + c) n t '
fJ(t) =
00
1
T
A21-=-~: c ln2a - 1 t
Q(s)ds =
(l
>0
<
00
fJ(y) dy ) ds
a(y)p(y)
1 (2c-2A 2 )/(2A-1)
(ln 2(C+A 2 -A)/(2A-1)T)
n
s
s ds <
1
---:-:--exp
a(s)p(s)
00
- 2
roo
iT
00.
00),
l/(a(t)p(t)).
1
lim inf k(t)fJ(t) > -4'
(2.6.22)
t~oo
00
1 - exp
-a(s)p(s)
<
(-4l
fJ+(y) dy ) ds
a(y)p(y)
(l
S
1
1
exp -4),
dy ) ds
a(s)p(s)
T a(y)p(y)k(y)
4A
1
(k(T))
a(s)p(s)(k(s))4A ds < 00.
00
roo
iT
((tln 2 t)x'(t))'
+ ~x(t) =
for
t 2: 2,
(2.6.23)
79
where c> 1/4 is a constant. Let p(t) = InA t and k(t) = _In- 1- At/(l +
A), where V4c -1 < A < -1. Then, we have
f3( ) = _ A2
t
+ 2A + 4c I l+A
+ A) n t
..
()
and hR:~f k t f3(t)
4(1
A2 + 2A + 4c
1
4(1 + A)2 > 2'
Now, it follows from Corollary 2.6.4 that (2.1.1) is oscillatory if c> 1/4.
Corollary 2.6.5. Let k(t) E C1([to, (0), JR+) satisfy k'(t)
If
liminf 4 ( ~ ( ) [4f3+(t)
t-HXl
a t p t
+ (a(t)p(s))']
>
= l/(a(t)p(t)).
~,
4
(2.6.24)
00
1
--:-:--:--:exp
a(s)p(s)
<
(1
-4
f3+(T) dT ) ds
a(T)p(T)
roo a(s)p(s)
1
exp ( t [_ 4c + (a(T)p(T))'] d~) ds
J
T
a(T)p(T)
, .
JT
T4c
{(Xl
JT
a(T)p(T)
s- 4c ds < 00.
Example 2.6.5. Consider the Euler equation (2.2.11) with c> 1/4. Let
p(t) = tA, where A < 1 is a constant. Then, we have
f3(t) = A2 - 2A + 4c t A- 1
,
4(1 - A)
and
lil~~f 4a(t~p(t)
[4f3+(t)
t ;:::
+ (a(t)p(t))']
T> 0
4~~ =~)'
Now, it follows from Corollary 2.6.5 that (2.2.11) is oscillatory if (c(A/4))/(1- A) > 1/4.
In equation (2.2.6) if Q(t) > 0 for t;::: to, then we have the following
result.
Theorem 2.6.10.
Suppose Q(t) > 0 for t;::: to
C1([to,00),JR+) satisfy k'(t) = l/(a(t)p(t)). If
and let
k(t) E
(2.6.25)
80
Chapter 2
Jt
w'(t)
v(t) = a(t)p(t) w(t)
(2.6.26)
w(T)
w(t)
k(t )w' (t)
+ ItT
a(s)p(s)w'(s)ds
a(s)p(s)
k(t)w'(t)
k(t)w'(t)
k(t)
it follows that
.
k(t)
li~~p k(t),B(t) S; li~~p k(t)v(t) < h~~p k(t) _ k(T)
which contradicts (2.6.25).
Theorem 2.6.11. If
I,
00
,B(s) ds <
a(s)p(s)
00
(2.6.27)
00
(IS
t;, (T)
)
a(T)p(T) dT ds <
a(s)p(s) exp -4
00,
(2.6.28)
Joo
y' (t)
!too
z2 (t)
a(t)p(t)
1
Jt
2,B( t )y( t)
a(t)p(t)
- a(t)p(t) [Z2(t)
1
(2.6.29)
ro
r a(s)p(s)
y(s) d
s
iT
iTr (iTt
C1
iT
1
a(t)p(t) exp
C1
C2
iT
(j'l
)2
z(s)
a (s) p( s) ds::;
y(s)j(a(s)p(s))ds <
(3 (u)
) ) z2 (s )
-2.1to a(u)p(u) du dt a(s)p(s) I1[S, to]ds
k(s)z2(s)
a(s)p(s) I1[S, to]ds 2:
(j ry
(t
81
k(S)Z2(S)
a(s)p(s) ds
J; dtj(a(t)p(t))
k(S)z2(S) )
( ) ( ) ds
asps
(jTl
and I1[S, t]
1
)
( ) ( ) ( ) ds
ksasps
it follows that
iT
z(s)
a(s)p(s) ds ::; c3Vlnk (1])
for some
> O.
C3
(2.6.30)
(iTr
d)
v( s)
a(s)p(s) s
(iTr
::;
k(s)v 2 (s) ) (
a(s)p(s) ds
iT
1
)
k(s)a(s)p(s) ds
(2.6.31 )
and
iT
fry
+ k(T)z(T) + iT
z(s)
a(s)p(s) ds.
(2.6.32)
From (2.6.30) - (2.6.32) the following estimate on the nonoscillatory solution w(t) is immediate
w(1]) _
In w(T) -
r a(s)p(s)
v(s)
ds
iT
<
-
c4ln
3/4
(2.6.33)
k(1]),
C5
>
exp
(-2c
In 3 / 4 k(1]))
a(1])p(1])
exp (- In k (1] ) )
C6
a(1])p(1])
C6
k(1])a(1])p(1])
T.
As a
Chapter 2
82
for some positive constants
C5
1
d TJ
/ 00 a(TJ)p(TJ)w
2 (TJ)
---,--,---,.......,.---;:--:-c-
which is a contradiction.
and
>
C6.
C6
Thus,
/00
1
d TJ
a(TJ)p(TJ)k(TJ)
00,
2
/ 00 a(s)p(s) exp (/s
(3(T)
)
a(T)p(T) dT ds =
00,
(2.6.34)
<
00.
Jt
w'(t)
w(t)
dt In w(t)
v(t)
a(t)p(t)
---'a(:-7t)l---'p(-:-t) [(3(t)
100 at:)~ls)
(3( t)
?: a(t)p(t) .
dS]
This implies
/=
a(s)u 2 (s)ds =
?:
/00 a(s)p(s)w
/=
(s)ds
a(s)p(s) exp
(2 /s a(~~:iT) dT) ds
00.
(a(t)x'(t))'
+ p(t)x'(t) + q(t)x(t) =
e(t),
(2.7.1)
(2.7.2)
83
a(s)p(s)z2 (s)
where
p(t)
exp
(2.7.4)
1jJ(t) =
(2.7.5)
TheoreITl 2.7.1.
limsup 1jJ(t)
t-+oo
= -liminf 1jJ(t) =
t-+oo
(a(t)p(t)z2(t)y'(t))'
(2.7.6)
00.
y(t)
= CrK::2
It
to
( ) ( ) 2( )ds+
asp s z s
It
( ) ( ) 2( )
asp s z S
to
(2.7.7)
by integration of
(is e(u)p(u)z(u)du)ds,
to
-liminfy(t) =
t-+oo
00.
(2.7.8)
Because z(t) is nonoscillatory, z(t) of. 0 for t 2': t1 2': to. Now, (2.7.8)
implies that y(t) is oscillatory, and hence x(t) = y(t)z(t) is also oscillatory.
ExaITlple 2.7.1.
x"(t)
where
0:,
+ sinkt x(t) =
t
z"(t)
+ sin kt z(t) =
t
t 2': to > 0
t a sin(Jt,
(2.7.9)
(2.7.lO)
84
Chapter 2
and
'""
(1'+(1/2)
,=
(1'+(1/2)[1
+ 0(1)]
C'"Y+(1/2) [1
+ 0(1)]
as
as
t ~ 00
t ~ 00,
where
((k 2 - 2)/(4k2))1/2 > O. Hence, Zj(t) is a nonprincipal
solution. Substituting zJ(t) and tOO sin;3t in (2.7.5), we find that (2.7.6)
is satisfied if Cl' >
(1/2) and any ;3 =/: O. Because, < 1/2, this shows
that equation (2.7.9) is oscillatory for any Cl' > 1 and for all ;3, k =/: O.
,+
t 2 to 20
(2.7.11)
where e(t) E C([to, (0), JR). Here, a(t) = 1, p(t) = 0 and p(t) = l.
The linear unforced equation zl/(t) - z(t) = 0 is nonoscillatory. It has
two linearly independent solutions e~t and et . Clearly, z(t) = e t is a
nonprincipal solution.
Now, we consider the following two cases:
Let e(t) = sint Then, we have 1jJ(t) = te~2S (j'eTsinrdr)ds.
(i)
Clearly, 11jJ(t)1 < 00 for all t 2 0, and hence condition (2.7.6) is violated.
In this case the general solution of equation (2.7.11) is x( t) = C1e~t + c2e t (1/2) sin t, where Cj and C2 are arbitrary constants. It is nonoscillatory
for all sufficiently large t and C2 =/: o.
(ii)
Let e(t) = t 6e t sint or e<isint, where /5 > 0 and E> 1
are constants. Substituting z(t) = e t and e(t) in (2.7.5), we find that
(2.7.6) is satisfied. It follows from Theorem 2.7.1 that equation (2.7.11) is
oscillatory. Thus, the large size of the forcing term generates oscillation.
xl/(t)
+ 2x'(t) + x{t) =
sint,
t 2 to 2
o.
(2.7.12)
85
liminfit e(s)p(s)z(s)ds =
t----+oo
and
-00
to
t----+oo
!'IjI(t)! ~ m
and
lim
t->oo
lim sup
to
e(s)p(s)z(s)ds =
to
00,
(2.7.13)
to
it
( ) ( ) 2( )ds
asp
s z s
(2.7.14)
ds =
(2.7.15)
a(s)p(s)z2(s)
00,
where p(t) and 'IjI(t) are defined in (2.7.4) and (2.7.5) respectively, then
(2.7.1) is oscillatory.
Proof. Let x(t) be an eventually positive solution of (2.7.1),say, x(t) > 0
for t 2: to > O. The function y(t) defined by x(t) = y(t)z(t) is a
nonoscillatory solution of equation (2.7.7). Integrating (2.7.7) from T> to
to t, where T is sufficiently large, we get
ht
e(s)p(s)z(s)ds
(2.7.16)
-00.
Next,
(2.7.17)
Replacing T by tl in (2.7.16), integrating it from tl to t and using
(2.7.14) and (2.7.17), we obtain
y(t) ~ y(tl) - m
t,
( ) ( ) 2( . ds
asp s z s)
e3t sint,
t 2: O.
(2.7.18)
86
Chapter 2
= e2t
as a non principal
(a(t)x'(t))'
+ q(t)x(t)
(2.7.19)
e(t)
r::
Jp !;
+11" /,j";f
+(q(s) - c;)
sin 2 (
V4(t -
p;))] ds
)
(2.7.20)
2 0
for all n 2 no, where no is a fixed positive integer. Further, let the
function e(t) satisfy
> 0 for
t E
[p~,p~ + 7r/.Jc;fl
:::: 0
t E
e(t) { -
for
(2.7.21)
e(t)
{ <0
20
for
for
tE[sl,hl
t E
[S2, t2]'
Sl
< t1
:::: S2
<
(2.7.22)
If there exist
such that
(2.7.23)
for i
= 1,2
87
+ p(t)x'(t) + q(t)f(x(t))
(a(t)x'(t))'
e(t),
(2.7.24)
f(x) E
D(Si,ti)'
Theorem 2.7.5. Let
x -=I- 0,
(2.7.25)
and suppose for any T::::: 0, there exist T::; SI < tl ::; S2 < t2 such
that condition (2.7.22) holds. If there exist p(t) E Cl([Si,ti],lR+) and
u(t) E D(Si,ti) such that
Qi(U)
2 (S)
a~)
(u'(s)
+ 2a~\;(s)) 2]
ds ::::: 0
(2.7.26)
(2.7.27)
,
W
p(t)e(t)
(t) = p(t)q(t) - f(x(t))
l'(t)
w 2 (t)
(2.7.28)
,
w (t) ::::: p(t)q(t)
l'(t),
w 2 (t)
on
I.
(2.7.29)
u 2 (s)w'(s)ds
:::::
u 2 (s)p(s)q(s)ds
1a(~~;~s)
+ JI f
u 2 (s)w(s)ds
(2.7.30)
Chapter 2
88
Integrating the left-hand side of (2.7.30) by parts and using the fact that
u(sI) = U(t1) = 0, we obtain
u 2(s)p(s)q(s)ds
J'
1a(;~;~s)
u 2(s)w(s)ds
2
w 2 (s)
If (x(s))u (s) a(s)p(s) ds,
which is equivalent to
;: : J[ a(s)p(s)
f'(x(s)) u(s)w(s) + (u'(s) + ')'(S)U(S))
2a(s)p(s)
Q1(U)
a(s)p(s)]
f'(x(s)) dS+Q1(U).
Since
2:::
[_f'(x(t))X'(t)u(t)
f(x(t))
or
(2.7.31)
d ,
dt Y (t)
u'(t)
,),(t)u(t)
')'(t)
0,
2a(t)p(t)
,),(t)u(t)
+ 2a(t)p(t) =
+ 2a(t)p(t)y(t)
0,
+ 2a(t)p(t)
+ u'(t)] + ')'(t)u(t)
d ( u(t))
f(x(t)) dt f(x(t))
or
on
0,
I,
(2.7.32)
where yet) = u(t)/ f(x(t)), tEl. It follows from (2.7.32) that (e(t)y(t))'
on I, where
e(t) = exp
tEl.
on
89
Qi(U)
lSift;
( ) u(s) )
p(s) [q(s)u 2(s) - a(s) (
u'(s) +'
;p(:)
2] ds::::
(2.7.33)
for i = 1,2 then equation (2.7.19) is oscillatory.
ReIllark 2.7.1. Theorem 2.7.4 follows from Corollary 2.7.2 by letting
p(t) = 1. Also, if the weight function
(2.7.34)
then "((t) defined in (2.7.27) will be identically zero. Thus, the condition
(2.7.26) in Theorem 2.7.5 takes the form
Qi(U) =
1:;
a~) (U'(S))2] ds
:::: 0
(2.7.35)
for i = 1,2.
ExaIllple 2.7.5. Consider the forced equation
(ytx'(t))' +
~x'(t) + ~x(t)
sinVt,
> O.
(2.7.3b)
(n+1)7r V2
Chapter 2
90
C2([to,00),JR+) to be p(t)
3. Theorem 2.3.1, Corollaries 2.3.1, 2.3.4 and 2.3.7 are due to Wintner
[33], Kneser [15], Hille [13J and Moore [22], respectively. Corollary 2.3.6
with a(t) = 1 and c = 1/4 reduces to Wintner's criterion [34J. Also
Corollary 2.3.11 with a(t) = p(t) = 1 reduces to a result of Potter [27J.
Theorem 2.3.3 improves a result of Wong [35], whereas Corollary 2.3.17
improves a result of Wintner [33J. Corollary 2.3.19 is due to Moore [22J.
Corollary 2.3.20 with A = 2, p(t) = 1 and f3(t) ~ 0, reduces to Willett's
criterion [31J.
4. Theorems 2.4.1 and 2.4.3 are taken from Graef et. al. [9J. Theorem
2.4.1 extends Leighton's oscillation criterion established in [20J.
5. Theorem 2.5.1 is extracted from Grace and Lalli [8J, and it includes
Wintner result [32J. Theorem 2.5.2 is due to Graef et. al. [9J. Theorem
2.5.3 includes Hartman's oscillation criterion [11 J as a special case. Lemmas 2.5.1 and 2.5.2 and Theorem 2.5.4 extend Willett's oscillation result
[31]. Theorems 2.5.6 - 2.5.8 are taken from Kamenov [14], Yan [37J and
Philos [25], respectively. Corollaries 2.5.5 - 2.5.7 are due to Philos [23,24J.
Theorems 2.5.12 and 2.5.13 are the extensions of Philos results [25J and are
borrowed from Garce [7J. Lemma 2.5.3 and Theorems 2.5.14 - 2.5.17 are
the generalizations of the work originated by Kong [16].
6. Theorems 2.6.1- 2.6.4 extend the results ofYan [38J. Theorem 2.6.5
is due to Li [21]. Lemma 2.6.1 is taken from Coppel [5J. Corollary 2.6.3
includes Wintner's criterion [33J. Theorem 2.6.11 generalizes theorem 2 of
Wong [35J, and Theorem 2.6.12 improves Wintner result [32J.
7. Theorems 2.7.1 and 2.7.2 are the extensions of Rankin results [28J.
Theorem 2.7.3 is due to EI-Sayed [6], and Theorem 2.7.4 is taken from Wong
[36]. Theorem 2.7.5 improves Theorem 2.7.4, and is taken from Agarwal
and Grace [2J.
8. For some other related results of this chapter see Agarwal et. al.
[1,3J.
91
2.9. References
l. R.P. Agarwal and S.R. Grace, On the oscillation of certain second
92
Chapter 2
21. H.J. Li, Oscillation criteria for second order linear differential equations,
J. Math. Anal. Appl. 194(1995), 217-234.
22. R.A. Moore, The behavior of solutions of a linear differential equation of
second order, Pacinc J. Math. 5(1955), 125-145.
23. Ch.G. Philos, Oscillation of second order linear ordinary differential equations with alternating coefficients, Bull. Austral. Math. Soc. 27(1983),
307-313.
24. Ch.G. Philos, On a Kamenev's integral criterion for oscillation of linear
differential equations of second order, Utilitas Math. 24(1983),277-289.
25. Ch.G. Philos, Oscillation theorems for linear differential equations of second order, Arch. Math. 53(1989), 482-492.
26. M. Picone, Sui valorieccezionali di un parametro de cui dipends un equazioni differenziale lineare ordinaria del second ordine, Ann. Scuola Norm.
Pisa 11(1909), 1-141.
27. R.L. Potter, On self-adjoint differential equations of second order, Pacinc
J. Math. 3(1953),467-191.
28. S.M. Rankin, Oscillation results for a nonhomogeneous equation, Pacinc
J. Math. 80(1979), 237-243.
29. B. Sturm, Sur les equations differentielles lineaires due second ordre, J.
Math. Pures Appl. 1(1836), 106-186.
30. C.A. Swanson, Comparison and Oscillation Theory of Linear Differential
Equations, Academic Press, New York, 1968.
31. D.Willet, On the oscillatory behavior of the solutions of second order
linear differential equations, Ann. Polon. Math. 21(1969), 175 194.
32. A. Wintner, A criterion for oscillatory stability, Quart. Appl. Math.
7(1949), 115-117.
33. A. Wintner, On the nonexistence of conjugate points, Amer. J. Math.
73(1951), .368-380.
34. A. Wintner, On the comparison theorem of Kneser-Hille, Math. Scand.
5(1957), 255-260.
35. J.S.W. Wong, Oscillation and nonoscillation of solutions of second order linear differential equations with integrable coefficients, Trans. Amer.
Math. Soc. 144(1969), 197-215.
36. J .S.W. Wong, Oscillation criteria for a forced second order linear differential equation, J. Math. Anal. Appl. 231(1999), 235-240.
37. J. Yan, Oscillation theorems for second order linear differential equations
with damping, Proc. Amer. Math. Soc. 98(1986), 276-282.
38. J. Yan, Oscillation property for second order differential equations with
an 'integrably small' coefficient, Acta Math. Sinica 30(1987), 206-215.
Chapter 3
Oscillation and Nonoscillation
of Half-Linear Differential
Equations
3.0. Introduction
In this chapter we shall present oscillation and nonoscillation criteria
for second order half-linear differential equations. In recent years these
equations have attracted considerable attention. This is largely due to the
fact that half-linear differential equations occur in a variety of real world
problems; moreover, these are the natural generalizations of second order
linear differential equations. In Section 3.1, we shall provide some preliminaries for the study of half-linear differential equations. In Sections 3.2 and
3.3, respectively, Sturm's and Levin's type comparison theorems are developed. In Section 3.4, we shall establish a Liapunov type inequality. Section
3.5 presents an oscillation criterion for almost periodic Sturm-Liouville
equations. A systematic study on the zeros of solutions of singular ha1linear equations is made in Section 3.6. Nonoscillation characterizations
(necessary and sufficient conditions), comparison results as well as several
sufficient criteria for the nonoscillation are presented in Section 3.7. Section 3.8 is devoted to the study of oscillation of half-linear equations. In
Section 3.9, we shall establish oscillation criteria by employing integral and
weighted averaging techniques. Here, interval criteria for the oscillation
of half-linear equations are also provided. Section 3.10 deals with the oscillation of half-linear equations with integrable coefficients. Section 3.11
addresses the oscillation of damped and forced equations. In Section 3.12,
we shall derive lower bounds for the distance between consecutive zeros of
an oscillatory solution. Finally, in Section 3.13, we shall present a systematic study of the oscillation and nonoscillation of half-linear equations with
a deviating argument. Here, classifications of the nonoscillatory solutions,
and the existence results which guarantee that the solutions have prescribed
asymptotic behavior are also presented.
94
Chapter 3
3.1. Preliminaries
In this chapter we shall study half-linear differential equations of the
form
d
(3.1.1 )
dt (a(t)'l/J(x'(t) + q(t)'Ij;(x(t = 0,
dt (a(t)x'(t
+ q(t)x(t)
(3.1.2)
O.
:t (a(t)[x'(t)]"') + q(t)x"'(t) = 0,
or
(3.1.3)
dt (a(t)[-x'(t)]"') - q(t)x"'(t)
(3.1.4)
respectively.
The existence of solutions to initial value problems for (3.1.1) on [T,oo),
T ~ 0 has been addressed in [15]. We note that any constant multiple
of a solution of (3.1.1) is also a solution. A solution x(t) of (3.1.1) is
called oscillatory if for each tl > 0 there exists a t > tl such that
x(t) = 0; otherwise, it is said to be nonoscillatory. Equation (3.1.1) is
called oscillatory if all its solutions are oscillatory, and nonoscillatory if
all its solutions are nonoscillatory. The main objective of this chapter is
to discuss the oscillation and nonoscillation behavior of all solutions of
equation (3.1.1).
and
AA
+ (A -
+ (A -
l)BA - AAB A- 1 ~ 0
l)BA - AAB A- 1
:s
for
for
A> 1
< A < 1,
95
= B.
l!~~~f
ld
[a(s)ly'(s)I<>+l - q(s)y<>+l(s)] ds
(3.2.1)
(3.2.2)
0,
d --+ b-
(a(t)gU(t))'
+ aa(t)gu+l(t)
q(t)
on
(a,b).
(3.2.3)
Let
a1/(u+ll(t)y(t)g(t)
and
), = a
+1
Id
[a(s)ly'(s)I<>+l-q(s)y<>+l(s)] ds
~ a(d)y<>+l(d)g<>(d)--a(c)y<>+l(c)gC>(c).
<>
a(t)
0<
a(t),
Chapter 3
96
Proof. We shall prove only the case x'(a) #- O. Assume to the contrary
that x'(a) = O. Since any constant multiple of a solution of (3.1.1) is
also a solution of (3.1.1), we can assume without loss of generality that
x(t) > 0 on (a,b). It follows from equation (3.1.1) that a(t)'l/J(x'(t)) =
q(s)'l/J(x(s))ds on [a, bJ, which implies
-J:
x'(t)
'l/J-l ( - att)
it
q(S)'l/J(X(S))dS) ,
x(t)
< (t - a)'l/J-l
for
a::::
iT
(Mit
'l/J-l ( - atr)
Mit
q(S)'l/J(X(S))dS) dr
Iq(s)I'l/J(x(S))dS)
::::
Iq(s)I'l/J(x(s))ds
for
a:::: t :::: b.
Thus, it follows from Lemma 1.1.1 that 'l/J(x(t)) = 0 for each t E [a, b).
This implies that x(t) = 0 on [a, bJ, which contradicts the hypothesis
x(t) > 0 on (a,b). This contradiction completes the proof.
From Lemmas 3.2.2 - 3.2.4, we shall establish the following Wirtingertype inequality.
Lemma 3.2.5. Let x(t) be a solution of equation (3.1.1) on [a,b)
satisfying x(t) #- 0 on (a, b). Denote by Y the family
#- 0
on (a,b)}.
97
Proof. Let y E Y We shall prove only the case y(t) > 0 on (a, b),
since the proof of the case y(t) < 0 on (a, b) is similar. Since x(t) is a
solution of equation (3.1.1) on [a, b) satisfying x(t) -=I- 0 on (a,b), we
may assume without loss of generality that x(t) > 0 on (a, b). Now, we
separate our proof into the following six cases:
Case 1. Either (i). there is a number c, a < c < b such that x'(c)
and x'(t) 2: 0 on (a, c), or (ii). x'(t) > 0 on (a, b).
(i). It follows from y(a)
t~a+
=0
t~c-
x(t)
Ha+
t~a+ X
t~a+
. y(t))"
( t~~+
x(t)
(II) If x'(a)
imply
= 0, then x(a)
0=
[Y'(a)]"
x'(a)
x 0 = O.
x(t)
t~a+
=0
y(a) = 0
= O.
x(t)
He-
O.
a(s)ly'(s)I"+lds 2:
= y(b) = 0
t~a+
q(s)y"+l(s)ds.
a(s)ly'(s)I"+lds 2:
(3.2.7)
t~b-
lb
lb
q(s)y"+l(s).
= O.
(3.2.8)
(3.2.9)
98
Chapter 3
y<>+l (t)
Ha+
[_ x' (t)]
x(t)
<>
= 0, x'(c) = 0
a(t)
lim
=0
y<>+l (t)
He-
o.
(3.2.10)
This and Lemma 3.2.3 imply that (3.2.7) holds.
(iv). It follows from y(a) = y(b) = 0 and Lemma 3.2.4 that
lim y<>+l(t) [- XI(t)]<> a(t) =
x(t)
Ha+
Hb-
o.
HCci
Cl
<>
O.
Hc:;-
(3.2.11)
Case 4. If there are two numbers Co and Cl, a < Co <
x'(co) = X'(Cl) = 0 and x'(t) < 0 on (CO,Cl), then
t ..... cci
Cl
O.
t ..... c:;-
<>
= 0, x'(c) = 0
a(t)
Hc+
Hb-
=0
0:
a(t)
O.
99
(vi). It follows from y(a) = y(b) = 0 and Lemma 3.2.4 that (3.2.8)
holds. This and Lemma 3.2.2 imply that (3.2.9) holds.
Case 6. Either (vii). there is a number c, a < c < b satisfying x'(c) = 0
and x'(t) < 0 on (c,b], or (viii). x'(t) < 0 on (a,b].
(vii). It follows from y(b)
t--+c+
t--+b-
(al(t)1f;(y'(t)))'
+ ql(t)1f;(y(t)) =
y(a) = y(b) = 0,
has a solution y(t) with y(t)
lb
(3.2.13)
#- 0 on (a, b) satisfying
then every solution x(t) of (3.1.1) must have a zero in (a, b) unless x(t)
and y(t) are proportional.
Proof.
Suppose to the contrary that x(t) is a solution of equation
(3.1.1) such that x(t) #- 0 on (a, b). It follows from Lemma 3.2.5 that
(3.2.5) holds. Multiplying equation (3.2.13) by y(t) and integrating it by
parts from a to b, we get
al(s)IY'(s)I<>+lds =
ql(s)ly(s)I<>+lds.
Combining this with (3.2.5), we find
J:
lb
J:
Chapter 3
100
If a(t) = al(t)
following corollary.
and
q(t) = ql(t),
Corollary 3.2.1. The zeros of two linearly independent solutions of equation (3.1.1) sepiirate each other.
Remark 3.2.3. Corollary 3.2.1 implies that the oscillatory and nonoscillatoryequations (3.1.1) are mutually exclusive.
Corollary 3.2.2. Let a(t), al(t) E C1([to,oo),IR+) and q(t), ql(t) E
C([to,oo),IR). Suppose al(t) 2: a(t) and ql(t) S; q(t) on [to, (0). If the
differential equation
(3.2.15)
is oscillatory, then so is (3.1.1); or equivalently, the nonoscillation of (3.1.1)
implies that of equation (3.2.15).
xl/(t)
+ ql(t)X(t) =
and
yl/(t)
+ q2(t)y(t) =
x(t) ((al(t)'!j;(X'(t)))'
+ ql(t)'!j;(X(t))} <
(3.3.1)
(i)
(ii)
If
'!j;(x'(a))
-al(a) '!j;(x(a))
it
a
'!j;(y'(a))
ql(s)ds > I-a2(a) '!j;(y(a))
it
a
q2(s)ds
(3.3.3)
101
on [a, b], then y(t) does not vanish on [a, b], and
x(t)x'(t) < 0,
if (i) holds,
if (ii) holds
also
1jJ(x' (t))
-al(t) 1jJ(X(t)) >
If >
by
t E [a,b].
for all
1jJ(y(t))
(3.3.4)
2.
Proof. Since the proofs of (i) and (ii) are similar, we shall prove only case
(i). Since x(t) > 0 on [a,b], the continuous function
w(t)
satisfies
w'(t) 2 ql(t)
Thus, we find
w(t)
w(a)
w(a)
it
it
[a,b]
on
+ cw~l/"'(t)lw(t)l(o+l)/(l
ql (s)ds
it
(3.3.5)
2 ql(t).
oa-1/"'(s)lw(s)I(Q+l)/ads
(3.3.6)
ql(s)ds.
z(t)
= -
1jJ(y'(t))
a2(t) 1jJ(y(t))
where a
(3.3.7)
z(t) 2 z(a)
it
q2(s)ds.
(3.3.9)
z(t) 2 z(a)
it
-it
ql (s)ds
Chapter 3
102
Hence, w(t) > -z(t) on [a, e]. Next, we claim that w(t) > z(t) on [a, e].
For this, assume that there exists a tl E (a,e) such that w(tJ} = Z(tl)
and z(t) < w(t) on [a, tl]' By (3.3.10), we have 0 < jz(t)j < w(t) on
[a, tl]' Thus, it follows from (3.3.3), (3.3.6) and (3.3.8) that
Z(tl)
z(a)
< w(a)
w(t l ),
on
[a, e].
(3.3.11)
Next, we shall show that y(t) cannot vanish on [a, b]. Suppose the
first point to the right of a at which y(t) vanishes is t = t2 :s; b, i.e.,
y(t) > 0 on [a, t2) and y(tz) = O. Then from Lemma 3.2.4 it follows
that y'(t2) =I O. This means that the solutions of equation (3.3.2) have only
simple zeros. However, since jz(t)1 < w(t) on [a, tz) and w(t) is bounded
on [a,b], we get 00 = limsuPHc jz(t)j :s; lim Ht - w(t) = W(t2) < 00,
which is a contradiction. Hence, 2y(t) cannot va~ish on [a, b]. Thus,
(3.3.11) holds on [a, e] C [a, b] on which z(t) is continuous. But, this
implies that z(t) is continuous on the entire interval [a, b], since w(t) is
bounded on [a, b] and y(t) cannot vanish on [a, bj. Therefore, (3.3.11)
holds on the interval [a, bj.
Now, from (3.3.5), (3.3.7) and (3.3.11) it follows that y(t) 2': x(t) on
(i)
(ii)
If
'ljJ(x'(b))
al(b) 'ljJ(x(b))
r ql(s)ds
+ it
'ljJ(y'(b))
(3.3.12)
for all t E [a, b], then y(t) does not vanish on [a, b], and
x(t)x'(t) < 0,
also
'ljJ(x'(t))
al(t) 'ljJ(x(t)) >
'ljJ(y'(t)) I
az(t) 'ljJ(y(t))
for all
[a, b].
(3.3.13)
>
then
~,
103
in (3.3.13) should be
Proof. Define
Xl(t)
= x(a+b-t),
Al(t)
Yl(t)
al(a+b-t),
A2(t)
Ql(t) = ql(a+b-t),
Q2(t)
y(a+b-t)
= a2(a + b - t)
= q2(a + b - t).
Then, Xl(t) does not vanish on [a,b], Al(a) = al(b), Yl(a) = y(b)
x(b) = xl(a) > 0, or Yl(a) = y(b) ::::: x(b) = xl(a) < 0, and
'lj;(x~ (a))
-A1(a) 1p(xl(a))
a +b- t
Ib q2(s)ds I
1p(y(b)) +
Ia2(b) 'lj;(y'(b))
t
Ql(s)ds
=
'lj;(x' (b))
al(b) 1p(x(b))
I-A2(a) 'lj;(Yl(a))
'lj;(y~(a)) +
Ib
ql(s)ds,
I ba
Q2(s)ds .
'lj;(y'(t))
'lj;(x'(t))
a2(t) 'lj;(y(t)) - al(t) 'lj;(x(t))
'lj;(y'(a))
'lj;(x'(a))
> a2(a) 'lj;(y(a)) - al(a) 'lj;(x(a))
It
+
a
(3.3.14)
(ql(S) - q2(s))ds
and
'lj;(x'(t))
'lj;(y'(t))
al(t) 'lj;(x(t)) - a2(t) 'lj;(y(t))
'lj;(x'(b))
'lj;(y'(b))
> al(b) 'lj;(x(b)) - a2(b) 1p(y(b)) -
Ib
t
(3.3.15)
(ql(S) - q2(s))ds.
z(t) - z(a)
<
-it
q2(s)ds =
it
aa;-l!a(s)lz(s)l(a+l)!ads
it aa~l!Q(s)lw(s)I(Q+l)!Qds
w(t) - w(a)
-it
ql(s)ds.
This in view of
z(t) =
'lj;(y'(t))
- a2(t) 'lj;(y(t))
and
w(t)
'lj;(x'(t))
- al(t) 'lj;(x(t))
104
Chapter 3
T,
(i)
x(t) and y(t) are nontrivial solutions of (3.3.1) and (3.3.2), respectively, and either
(I)
(ii)
it
ql(s)ds
lit
for all t
Q2(s)dsl
x(t)x'(t) < 0,
also
and
and
~
x(t) >
x(t) <
on
on
[a, (0), or
[a, (0),
a.
a, and
<
if (I) holds,
if (II) holds
_al(t)1P(X'(t)) >
1p(X(t)) -
la (t)1j;(y'(t))
I
1j;(y(t))
for
~ a.
(al(t)1j;(X'(t)))'
+ Ql(t)1j;(X(t)) =
of the
(3.3.17)
105
+ q(t)x(t)
x"(t)
having consecutive zeros at t
holds
(b-a)
lb
q(s)ds > 4.
In this section, first we shall extend Liapunov's inequality to the halflinear differential equation
+ q(t)~/J(x(t)) =
('l/J(x'(t)))'
(3.4.1 )
Proof. Since for the equation (3.4.1) any constant multiple of a solution
x(t) is also a solution, we may assume without loss of generality that
x(t) > 0 on (a, b). Let M be the maximum of x(t) on [a, b] which is
attained at some point to = a + f,(b - a), 0 < f, < 1. Then, we have
'l/J(x'(a)) - 'l/J(x'(b))
d
- Jar dt'l/J(x'(t))dt
1
b
1
b
(3.4.2)
q(t)dt.
It follows from x(t) > 0 and q(t) :2': 0 that x(t) is concave. Hence, we find
x'(a) :2': M/[f,(b - a)] and x'(b):::; -M/[(1 - f,)(b - a)]. This and (3.4.2)
Chapter 3
106
f:
f:
Theorem 3.4.3. Let q(t) E C([a, b], JR). Suppose x(t) is a solution of
equation (3.4.1) having consecutive zeros at t = a and t = b. Then the
following inequality holds
2"'+1
< (b - a)'"
lb
q+(t)dt,
(3.4.3)
(1jJ(y'(t)'
+ q+(t)1jJ(y(t
(3.4.4)
O.
The existence theory ensures that the initial value problem (3.4.4), y(a) =
0, y'(a) = 1 has a solution y(t). Then from Theorem 3.2.1 it follows that
y(t) has at least one zero in (a, b]. Let ~ = min{ t E (a, b] : y(t) = O}. Thus,
from Theorem 3.4.2, we have (b - a)'"
q+(t)dt 2: (~- a)'" q+(t)dt >
f:
f:
k=1,2,,N.
(3.4.5)
107
Therefore, we have 2::=1 (ak - ak~l)~<> 2:: N<>+l (b - a)~<>. This and
(3.4.5) imply that
q+(t)dt > (2N)<>+!(b - a)~<>. From this inequality
the following corollary is immediate.
J:
Corollary 3.4.1. Let q(t) E C([a, b], lR) be not identically zero on [a, b].
Suppose x(t) is a nontrivial solution of equation (3.4.1) having consecutive
zeros at a = ao < a1 < ... < aN = b. Then the following inequality holds
lb
a)~<>.
Definition 3.4.1.
A function f E C(lR, lR) is said to be quickly
oscillatory if it is defined in a neighborhood of 00 and there exists a
sequence {tn}~=l with limn-+oo tn = 00 such that f(t n ) = 0, n =
1,2,, tn+ 1 > tn and limn-+oo(t n+ 1 - tn) = O.
As an application of Theorem 3.4.3, we shall prove the following result.
If equation (3.4.1) has a quickly oscillating solution,
Theorem 3.4.4.
then
(3.4.6)
and limsuPt-+oo q(t) =
00.
00
q+(t)dt >
to
It
q+(t)dt >
2<>+!(tn+1-tn)~<>
--+
00
as n--+oo.
tn
Hence (3.4.6) holds. Now, from the mean value theorem for integrals, we
q+(t)dt = q+(~n)(tn+1 - tn) > 2<>+1(tn+! - tn)~<>' where
have
tn < ~n < tn+!. This implies that q+(~n) > 2<>+!(tn+! - tn)~<>~l and
hence limsuPt-+oo q(t) = 00. This completes the proof.
JLn+l
IlqllB =
1
lim sup -2
t-+oo
jt
~t
Iq(s)lds,
108
Chapter 3
it
M{q} = lim -1
t .... oo
to
q(s + (3)ds = 0
to 2:
for some
o.
('ljJ(x'(t)))' - )..q(t)'ljJ(x(t))
where q(t) E fl, ).. E JR - {O} and
~/J(x)
0,
(3.5.1)
is as in equation (3.1.1).
0: =
x"(t) - )..q(t)x(t)
From Levin's comparison theorem with
immediate.
O.
0: =
(3.5.2)
1 the following result is
+ b)'Y >
a'Y
+ b'Y.
(3.5.3)
('ljJ(x'(t)))' - q(t)'ljJ(x(t))
on [to, (0), then
. lit (IX'(S)I)"'+l
Ix(s)1
hm -
Hoo
to
--
ds
o.
(3.5.4)
109
w'(t) - o:lw(t)l(o+l)/o
+ q(t)
lit Iw(s)l(o+l)/Ods
t
t;::: to.
It: Iw(s)l(o+l)/ods =
limsup t-+oo
for
>
to
(3.5.5)
O. Assume
o.
(3.5.6)
w(t)
w(to)
t
+~
t
it q(s)ds + ~t it Iw(s)l(o+l)/Ods
to
to
(3.5.7)
From (3.5.6), (3.5.7) and (3.5.3) there exist a positive constant m and an
increasing sequence {t n };:"=l of [to, (0) with limn -+ oo tn = 00 such that
(3.5.8)
Ii:
it
for all
q(s)dsl <
t ;::: t*.
(3.5.9)
it
q(s)ds =
tn
q(s)ds
to
_ltn
to
+ tn ) .
(3.5.10)
w(t n )
-it
tn
(3.5.11)
!('lj;(x~(t))) -
0:
(;)"+1 'lj;(Xn(t))
(3.5.12)
l10
Chapter 3
(3.5.15)
and
(3.5.16)
on h = [tn, tn +cnl c [t*, (0), where n is large enough so that wn(tn) >
(m/2)Q. Then, Yn(t n ) = wn(t n ), and it follows from Lemma 3.5.1 that
+ cn )-(Q+1)
[(tn - t + cn )-(Q+1) + (m/2)Q+1]
a(tn - t
y~(t)
<
a [(tn - t+ cn)-Q
- a(m/2)Q+1
+ (m/2)Q] (Q+1)/Q
aIYn(t)I(Q+l)/O' - a(m/2)Q+1
on
- a(m/2)Q+1
h.
Thus, we have
+ a(m/2)Q+1 < 0
- alw n (t)I(Q+1)/Q + a(m/2)0'+1
- aIYn(t)I(Q+1)/a
y~(t)
w~(t)
wn(t n )
that tn
+ Cn E I
111
limt-4(tn+cn)-
lim
t-4(t n +C n )-
wn(t) = 00.
(3.5.17)
~ It+li q(s)ds = _
w(t: 6)
+ w~t) +
JIt+li Iw(s)I(,,+1)/"ds
IlflIB'
lit
limsup t-400
to
(3.5.18)
essentially a
If(s)lds
(3.5.19)
Chapter 3
112
~
8t
it lsHIW(~)I(e>+l)/ad~ds
~ it f'5Iw(~ + s)l(a+l)/ad~ds ~ i Iw(~ + s)l(a+l)/adsd~
io
io
11
i
to
at
< -1
8t
tH Iw(s)l(a+l)/Qdsd~ =
-1
to
{6
8t
to
to
tO
to
+6 Iw(s)l(a+l)/ads,
(3.5.20)
for any fixed 8 > O. Now, from (3.5.20) and Lemma 3.5.2, we get
(3.5.21 )
Applying (3.5.21) and JJWIJB'
= 0 for all
<5 >
o.
(3.5.22)
Finally, since q(t) is Besicovitch almost periodic, it follows from [9J that
t 2: to 2: O.
(3.5.23)
M{q} =
and
M{JqJ} =
lim -1
t-+oo
1t
0
11t
lim -
t--+oo
Jq(s)Jds
q(s)ds
sin t
.
1I
mt
t-+oo
1
lim 2 ( )
t-+oo
n + 1 7r
2 (n+l)7f
217f/2
lim I cos slds =
n-+oo 7r
I cossJds
2
7r
> O.
113
Thus, it follows from Theorem 3.5.2 that for each A E IR - {O} equation
(3.5.23) is oscillatory at + 00 and - 00.
x"(t)
where a E (0,1]
integrable and
q(t)lx(t)IOlx'(t)l l -
sgn x(t),
(3.6.1)
q-(t)
(lq(t)l- q(t))/2.
Chapter 3
114
In other words, q f:- OQ((a, b)) if and only if there is no proper solution
x(t) of (3.6.1) satisfying for some al E [a, b) the conditions
x(al +) = 0,
x(b l -) = 0.
(3.6.3)
Definition 3.6.3. The function q(t) is said to belong to the set Uc>((a, b))
if for any al E [a,b) and bl E (al,b] the problem (3.6.1), (3.6.3) has no
non-zero (not necessarily proper) solution.
It is clear that if q E Uc>((a, b)), then q f:- Oc>((a, b)). In the case when
a = 1, or a E (0,1) and q(t):s;
for a < t < b, the converse is also
valid, i.e., if q f:- Uc>((a,b)), then q E Oc>((a,b)).
t--+a+
(a+b)/2
Iq(s)lds =
and
lim (b-t)c>
it
t--+b-
xI(a+) = 0,
x2(b-) = 0,
x~ (a+)
x~(b-) =
(a+b)/2
X2 (t)
Iq(s)lds = 0.
of the equation
(3.6.4)
- 1.
(3.6.5)
Let x(, T) E T . We shall show that this solution is defined in the interval
[T, b). Suppose that
x'(t, T)
and
x(t, T)
(t-T)+
It
115
(IXs(~iIO<)
Jr.
< 1 + y for
> 0,
(3.6.9)
1~(~il+lx'(t,T)I::;
Consequently, sup IT
obtain
2exP (2h(s-a)0<Iq(s)ldS)
for
tEIT.
(3.6.10)
for
T::;t<b
(3.6.12)
Ix/(t, T) -
11 ::;
(3.6.13)
Let Tk E (a, b), Tk+l < Tk for k = 1,2,, limk-4oo Tk = a. Suppose
Uk(t) = x(t, n) for Tk ::; t < b for k = 1,2,. From (3.6.11) and
(3.6.12) the sequences {udk=l and {u~}k=l are uniformly bounded
and equicontinuous in (a, b), (i.e., on each segment contained in (a, b)).
Hence, without loss of generality, by the Arzela-Ascoli theorem, we can
assumethat limk-4oou~i)(t)=x~i)(t), i=O,1 uniformlyin (a,b).
It is easy to see that the function XI(t) is a solution of (3.6.1). Further,
by (3.6.11) and (3.6.13) the function Xl (t) satisfies condition (3.6.4) also.
Thus, we have proved that the set of solutions n of the problem (3.6.1),
(3.6.4) is non-empty.
Chapter 3
116
Now, we shall show that all functions in n are defined on the entire
segment [a, b]. For this, let Xl E n be a noncontinuable solution. Then
analogous to what has been done above, we find
for tEl,
(3.6.14)
where
I = {t
Ix~(s)1
00,
<
00
for
(b-t)lx~(t)1
<:: (b-a)+
IXl(t)1 <:: 1 +
t-a
~b
1
-a
It
a
It
(s-a)C>(b-s)CXlq(s)
s)x~ (s)11-ads.
Adding these inequalities, using (3.6.9) and applying Lemma 1.1.1, we obtain
IXl(t)1 +(b-t)lx'(t)1
t- a
<
(1
+b-
Ib
a) exp (
l+b-a
b_ a
(s - a)"(b - s)a1q(s)ldS)
for tEl.
)I
(3.6.15)
a
[(s-a)(b-sW q ( S )IXl(s)l
-b-s
s-a
x I(b - S)X~ (8)1 1 - " sgn xl(s)ds
Xl ()
t -_ t - a - (b - t
[(s--a)(b-s)]oq(S)
xl(s)ds
from which in view of (3.6.15) and Proposition 3.6.1, we conclude that there
exists a finite limit Xl (b-).
117
lim (t - a)lx'(t) I =
t--+a+
(3.6.16)
(3.6.17)
t--+b-
liminf (t-a)lx'(t)1
~~
(liminf (b-t)lx'(t)1
t~-
0).
For this, assume that (3.6.17) does not hold. Then there exist c
and E > 0 such that
t~a
Ix'(t)l>
for
a<t<c
(lx'(t)l>
b~t
for
(a, b)
C<t<b).
c-a
Ix(t) - x(c)1 > d n - t-a
(Ix(t)-X(C)I >
a<t<c
for
dn~=~
for
c<t<b),
(t - a)x'(t) = (T - a)x'(T)
it
-iT
+
(
+ x(t) - X(T)
(b - t)x'(t)
(b - T)X'(T)
+ X(T)
(3.6.18)
- x(t)
a<t<r<b)
(t - a)lx'(t)1 S (T - a)lx'(T)1
it
(s - a)"'lq(s)I(lx(s)1
+ Ix(t) -
+ (s -
x(T)1
Chapter 3
118
+ Ix(t)
- x(7)1
Suppose
c(T)
2max{lx(t)-x(a+)I:
for a:::;t:::;T}
+ iT(S - a)'"'lq(s)llx(s)lds
( crT)
2 ma:{lx(t) - x(b-)I:
a<T <b
for
1r (b - sr'lq(s)llx(s)lds
) ,
a<T <b
for
and
it
I
(8 - a)"'lq(s)I((s - a)lx'(s)l)ds
for
(3.6.20)
<
(b-S)''',q(S)'((b-S)'X'(S)')ds).
for
<T <t<7<b
(t-a)lx'(t)1 :::; M(7, T) exp (iT (s - a)"'lq(s)ldS) for a < T < t < T < b
((b-t)IX'(t)1 :::; M(7, T) exp (J:(b-S)"'lq(S)ldS) for a < T < t < 7 < b)
Now, in view of (3.6.17) and (3.6.20), we find
((b-t)IX'(t)l:'S C(T)ex p
(1r (b-S)"'lq(S)ldS)
b
for
for
a<t<T <b
a<T<t<b).
=0
(limT--+b~
119
Lemma 3.6.4. Let al E (a, b), bl E (al' b) and let x(t) and y(t) be
solutions of equation (3.6.1) satisfying the conditions
(3.6.21 )
u(t) > 0, u'(t) > 0 for al :::: t :::: bl (u(t) > 0, u'(t) < 0 for al :::: t :::: bl )
'ul/(t) :::: q(t)lu(t)I"lx'(t)ll-a for al < t < bl .
Then the following hold
(3.6.22)
and
y'(t)
u'(t)
y(t) < u(t) for al:::: t :::: h
u'(t)
)
( Y'(t)
y(t) > u(t) for al:::: t :::: bl .
(3.6.23)
Proof. Suppose to the contrary that (3.6.22) does not hold. Then by
(3.6.21) there exists c E (al,b l ) such that x'(t) > 0 for a] < t <
c (x'(t) < 0 for c < t :::: bl ) and x'(c) = O. Define
p(t)
for
al
< t < bl
and
o-(t) = Ix'(t)
I x(t)
IQ
sgn x'(t)
for
a<
t:::: c
for
al
(for c:::: t
< bd.
for
< t < bl
(3.6.24)
(3.6.26)
From (3.6.26) there exist Cl E (al'c) and C2 E (CI'C) (Cl E (cl,b l ) and
C2 E (c, cd) such that
p(t) > o-(t) > 0 for Cl < t < C2, P(Cl) = o-(cd
(p(t) < o-(t) < 0 for C2 < t < Cl, P(Cl) = o-(Cl))'
(3.6.27)
120
Chapter 3
aCt) =
a(Cl)
+ Q:
it
q(s)ds -
Q:
c}
a(t)
Q:
Q:
1~ Ip(s)l(a+l)/ads
= a(cl) -
< P(Cl) -
la(s)l(a+l)/ads
Cl
it
Q:
Cl
< t < C2
)
Q:
Q:
for
C2
< t < Cl
'
x(a+) = 0
(3.6.28)
(x(b-) = 0)
t < b),
C
1=
for
a<t ~ c
(for
~ t
< b).
(3.6.29)
x~(cd =
o.
Cl E
x~ (t)
-- <
Xl(t)
Define
u'(t)
u(t)
for
(3.6.31 )
Cl,
g(t) = q(t)h(t)(Xl(t)U(t))Q
for
121
(3.6.32)
Ig(t)1
_x:j~~ (t)) a :
a<t< a
Cl } .
y(t)
(_1(a+c
+ cd/2
:2:
y (a
~ Cl )
exp
:2:
y (a
~ Cl )
+ cd/2].
it follows that
Il/ 2 9(S)dS)
a + Cl
for a < t ::; - 2 - '
(3.6.33)
t->a+
However, from the proof of Lemma 3.6.3 and condition (3.6.4), we have
liminf y(t) =
t ..... a+
-x~(t)u(t))
::; 0,
(3.6.35)
If false, then there exists C2 E (a, co) such that X'(C2) = O. By Lemma
3.6.4, we have x'(t)/x(t) < x~(t)/Xl(t) for a < t < C2. Thus, as above it
follows that yi(t) = gl(t)Yl(t) for a < t < C2, where
Yl(t)
X'(t)Xl(t) - x~(t)x(t)
for
a<t <
C2,
122
Chapter 3
gl(t)
h t
1 ()
_
-
liminf (t - a) Ix'((t
x(a+) = 0,
t---+a+
(X(b-) = 0,
t---+b-
00
< 00).
<
(3.6.36)
x'(t)
X t
.
( hm
(b - t) -x'(t)
Hbx(t)
lim (t - a)-(-) = 1
t---+a+
1) .
(3.6.37)
a < t < Co
(3.6.38)
(t _ a) x'(t)
x(t)
:::;
()X'(T)
1 + T - a X(T)
(b _ t) Ix'(t)1
x(t)
<
[1
It( S T
+ (s - a) X'x((s)]
s)
ds
)"'1 ( )1
for
qs
l+(b_T)lx'(T)1 + r(b-s)"'lq(s)1
X(T)
it
x
[l+(b-s)l~g}l]dS
for
co<t<T<b
(3.6.39)
I
I(t-a )-x'(t)
--1
x(t)
123
< (T-a)x'(T)+X(T)
rt
x(t)
x' (s) ]
+ JT (s - a)"'lq(s)1 1 + (s - a) x(s) ds for a < T < t <
(b - t)-x'(t)
-+l I
I
x(t)
(
Co
)
Co
<t <T<b
.
(3.6.40)
(t-a)
~g}
<::; M(T)+
(b-t)l~g}1
<::; M(T)
~(~}) ds
+ I (b-S)'''lq(S)1
((b-s)I~(~}I)dS),
for Co < t <
where
M(T)
X'(T)
1 + (T - a) X(T)
(M(T) = 1 + (b - T)
jeo(s - a)Qlq(s)lds
I~(~}I + l>b -
s)"lq(s)lds
<b
for
for
(t-a)~g}
<::; M(T)ex p
(l (s-a)"'lq(s)lds)
((b-t)I~(~}1
<::; M(T)ex p
(l
for a<T<t<co
x'(t)
(lx'(t)1
)
(t-a) x(t) <::; Mo for a < t < Co
(b-t) x(t) <::; Mo for Co < t < b .
From (3.6.40), we find
(t-a)x'(t)
x(t)
-11
<::; (T-a)x'(T)+X(T)
x(t)
+ (I+Mo)jt(s-a)Qlq(s)ldS
a
for a
Co
124
Chapter 3
S)"'lq(S)ldS)
<b
l(t-a)~g?-11
)l
)l
:.: : : (I+Mo
(l(b-t)~g?+11 : ;
(I+Mo
(s-a)"'lq(s)ldS
for
a<t<co
(b-S)"'lq(S)ldS
for
Co<t<b).
(t - a)x'(t) = x(t)
(b - t)lx'(t)1 = x(t)
it
(s - a)q(s)lx(s)I"'lx'(sW-"'ds
+ Ib(b -
for
a < t < Co
(b - t) Ix'((t)) I >
( liminf
t-+bX t
> 0 and
Cl
Cl
(a, co)
(Cl
0).
Cl
f'"
b
(b-S)"',q(S)'[(b-S)'X'(S) '1 dS,
Cl
Cl
125
)x'(T)
a x(T) < exp
((b-T)I:(~?I:::;
(f" I
(s-a)'',q(s)'ds)
eXP(f"lb(b-S)"lq(S)ldS)
for a
for
< T < Cl
Cl<T<b).
x(a+) = 0,
(X(b-)
liminf Ix'(t)1 = 0
t-+a+
= 0,
liminf Ix'(t)1 = 0) .
t-+b-
(3.6.43)
x'(t)
X'(T)
it
and
x(t)
X(T)
+ (t -
T)X'(T)
it
t-a
T-a
it
(s - a)"lq(s)1
+ IX'(T)I +
it
T
(s _ a)"lq(s)1 [lx(s)1
s-a
+ IX'(s)l] ds.
Hence, by Lemma 3.6.7 there exist f > 0 and Cl E (a, co) so that
(3.6.42) holds. Combining the last two inequalities, using Lemma 1.1.1 and
condition (3.6.42), we obtain for a < T < t < b, T < CI,
Chapter 3
126
Now, from (3.6.43) and the above inequality it follows that Ix(t)1 + (ta)lx'(t)1 = 0 for a < t < b, which is a contradiction. This completes the
proof.
= O} \ {t
i- 0.
First we shall show that if Xo E B([a}, b}j), where b} < b then there
exist b2 E (b},b) and a nontrivial solution Xo of (3.6.1) such that
(3.6.44)
Indeed, by Remark 3.6.1, either
xo(b)
or
xo(b) = 0,
i-
(3.6.45)
0,
x~(b)
i- o.
(3.6.46)
p(b}) =
1x~(b})
I
xo(b}) I'" sgn xo(b}).
x(t) _
-
{XO(t)
for
xo(b l ) exp
(la\::; t ::;
b
b}
for
bl
::;
t ::; b2 .
The function x(t) clearly satisfies equation (3.6.1) for t E (a, ~). By
Lemma 3.6.2 there exists a solution xo(t) of (3.6.1) such that xo(t) = x(t)
for t E [a}, b2 J and satisfies (3.6.44).
Suppose now that (3.6.46) is satisfied. Denote by Yo(t) a solution of
the Cauchy problem
x(t) = { xo(t)
yo(t) for
b}::; t ::; b2 .
127
rf.
(3.6.47)
I.
Yk
(3.6.48)
k = 1,2,.
Define
x(t)
Yl ( t)
a:<::; t < iI
for tk-l:<::;t<tk,
for
Yk(t)
k=2,3,.
(3.6.49)
{t
mes Ak = 0,
{t
{t
k = 1,2,
(a,bl ): (x(t' = O} c {t
Uk=l A k
Since mes A '=' I:~l mes Ak = 0, this implies that x E B([a, bl]), and
consequently bl E I, which contradicts (3.6.47).
Remark 3.6.3. In the above lemma, we proved that the problem (3.6.1),
(3.6.4) has at least one proper solution. Similarly, we can show that the
problem (3.6.1), (3.6.5) has at least one proper solution, and for any to E
( a,b) and Cl, C2 E lR, ICII + 1c21 i= 0 the Cauchy problem
xl/(t)
x(to)
Cl,
x'(to) =
C2
Chapter 3
128
has at least one proper solution.
Lemma 3.6.10. Let there exist c E (a, b)
y E ACloJ(a, c) U (c, b)) having finite limit
y(a+)
~ 0,
y(b-)
~ 0,
y'(c-) > 0,
y'(c+) < 0
y"(t) :S q(t)ly(t)IOly'(tW-a
for
a < t < b,
a < t < c,
Then, q E Uo((a,b)).
Assume to the contrary. Let q rf. Ua((a,b)). Then there exist
and a solution xo(t) of (3.6.1) satisfying the
conditions
Proof.
al E [a,b), bl E (al,b]
(3.6.50)
By Lemma 3.6.8 and Remark 3.6.1 there exist
h, bd such that
x~(t)
x~(t)
> 0 for
< 0 for
'I
(al,b l ) and
'2
X~(Tl) = 0
X~(T2) = O.
x~ (t d = 0
X~(t2) = O.
Lemma 3.6.11.
Let there exist c E (a, b) and a function 0" E
ACLoc((a, c) U (c,b)) having finite limits O"(c-) ~ dc+) and satisfying
the conditions
Then, q
for
a<t <b
(3.6.51)
(3.6.52)
129
Proof. Assume to the contrary that q E Oo((a, b)). Then there exist
al E [a, b), b1 E (al,bJ and a proper solution xo(t) of (3.6.1) satisfying
the condition (3.6.50). Suppose al = a and b1 = b (the lemma for the
other cases can be proved analogously). First, we introduce the function
p by
p(t)
~~~!~
0
1
sgn
x~(t)
for
a < t < b.
Clearly, we have
for
a<t<b
(3.6.53)
t-+a+
lim (b - t)Op(t) =
t-+b-
- l.
(3.6.54)
From (3.6.52) and (3.6.54) we conclude that there exist tl E (a, b) and
a E AC([tl' tl + ED, and
for
tl < t < tl
+ E, a(td = p(td.
::::; t ::::; tl + E. Then
(3.6.55)
there exists an
It follows from this fact, (3.6.51) and (3.6.53) that u'(t)::::; -h(t)u(t) for
tl < t < tl + E, and hence we conclude that u(t) < 0 for tl < t < tl + E,
which contradicts (3.6.55).
_
Q(t, to, a) -
{ a(t - a)O
ito
q(s)ds
for
a < t < to
a(b - t)O
q(s)ds
for
to < t < b,
to
Q*(to,a)
Q*(to, a)
Theorem 3.6.1. Let there exist ,x E (a, b) and I-l E (,x, b) such that
-a
11-'q(s)ds ~
(3.6.56)
Chapter 3
130
a < t < b.
for
(3.6.57)
Let
pet)
xo(t)
xo(t)
sgn xo(t)
a<t <b
(3.6.58)
so that
for
a < t < b.
(3.6.59)
Al : (t -
a)O pet)
-I}
i= (a, b).
(3.6.60)
+ A _1 a
Jar>- (s -
A _0: a
Jar>-
(A - a)Op(A)
a)o+lq(s)ds
(3.6.61)
Multiplying both sides of (3.6.59) by (b - t)<>+l, integrating from tt to
b and taking into account Lemmas 3.6.3 and 3.6.7, we get
Clearly, we have
(3.6.64)
131
(A - a)a peA)
1+ A
~ a 1>' (s -
a)a+lq(s)ds,
(3.6.65)
(b_p,)O<p(p,) > _1 _ _
a_ (b(b_s)o<+lq(s)ds
(3.6.66)
b - p, il-'
-a
11-' q(s)ds
peA) - pep,)
(3.6.67)
(3.6.68)
Then, q E O",((a,b)).
Proof.
[(a
p, E
+ b)/2, b)
and
-a(a + 1) [(A _
+(b_:)o<+l
+ b)/2]
Q(s, a;b,a)ds]
2::
(A_~)O<+l + (b_:)o<+l
(3.6.69)
1
a
>.
a +b )
Q ( s, -2-,a
ds
--
+1
1>' (s - a)a+lq(s)ds]
a
and
{b
il-'
a +b
Q s, -2-,a ds
=
_1_ [(b_P,)o<+l
q(s)ds+ {b(b_S)O<+lq(S)dS].
a +1
i(a+b)12
il-'
132
Chapter 3
In view of these facts, we find from (3.6.69) that the inequality (3.6.56) is
satisfied. Hence, by Theorem 3.6.1, q E Oa((a, b)).
a<t <b
q(t) :S 0 for
a)
to -Q*(to,a) :S - 1 - max { ( b - to
'
b- to )
to - a
(3.6.70)
a }
(3.6.71)
Then, q E Oa((a,b)).
Proof. By (3.6.71) there exists c E (a, b) \ {to} such that
-Q(c,to,a) 2
l+max{(~o_-t:)a, (:o-_t~r'}.
(3.6.72)
Suppose
). =
{ to if c < to
c if c > to.
c if c < to
to if c > to,
This and inequality (3.6.70) show that (3.6.56) holds, and hence by Theorem 3.6.1, q E Oa((a,b)).
In the case when condition (3.6.70) holds Theorem 3.6.1 can be improved. In fact, the following result holds.
Theorem 3.6.2. Let condition (3.6.70) hold and there exist ). E (a, b),
j.l E ()., b) and n E IN such that
(3.6.73)
where
al(t)
t - a, ak+1(t) = t - a +
it
133
X~ (t)
or there exist
X~(t)
CI
(a, b) and
x~(t)
CI,
C2
< 0,
(3.6.74)
< t < b,
x~(t) = 0,
~ t ~
CI
C2.
(3.6.75)
Suppose
pet) =
IX~(t)la sgnx~(t)
Xl (t)
for
a<t<b
pet)
ak"'(t)
for
(3.6.76)
CI,
p-l/"'(t)
t -
a+
lot a~+I(s)lq(s)lds
for
a<t <
Cl'
From (3.6.75), we find that pet) ~ ak_;\ (t) for a < t < b. Thus, it is
proved by induction that for any k E 1N the inequality (3.6.76) holds.
Analogously, we can show that for any k E 1N,
pet)
- bk"'(t)
for
a < t < b.
(3.6.77)
Chapter 3
134
Integrating (3.6.59) from A to 11. and using (3.6.76) and (3.6.77) (for
Iq(s)lds a;:;-a(A) +b;:;-a(p), which contradicts
(3.6.73).
k = n), we obtain - a
s:
1:
Theorem 3.6.3.
Let there exist c E (a, b)
AC([a, c]) and 9 E AC([c, b]) such that
f(a) = 0,
g(b) = 0,
a < t < c,
c<t<b,
Ig'(tW+ 1
g<>(t)
and
1
c
f(s)q(s)ds
+ f(c)
1
C
lb
g(S)Q(S)dS]
11'(s)la+1 ds
f<>(s)
+ f(c)
Ib
c
Ig'(s)I<>+l dS].
g"(s)
(3.6.78)
Then, q E O,,((a,b)).
1t
+ 1)f1/(a+1l(t) <
-
llt
a
11'(s)I"+l dS]
f"(s)
Q+1
0+'
t--+a+
lim g(t)p(t) = O.
t--+b-
(3.6.79)
-a
1
c
f(s)q(s)ds
+ f(c)p(c)
- f(a
+ E)p(a + c)
a+
1
c
[f'(s)p(s) - a f (s)lp(s)l(a+1)/a] ds
a+
<
a+
(3.6.80)
135
b-ry
: :; l
( _A )"'+1
0+1
for
-og(c)
and
-of (c)
f(s)q(s)ds
+ g(c)f(c)p(c) <
g(c)
(0 + 1)"'+1
f(c)
(0
+ 1)"'+1
T E JR.
Ib
c
If'(s)I",+l
ds
f"(s)
19'(s)I"'+1
ds.
go(S)
-0
_0
(s-a)o+l(b-s)o+1 q(s)ds>
[1(a+b)/2 (s _ a)Aq(s)ds
a
+ {b
(b
a)0+2
,
0+2
(3.6.82)
(b - S)Aq(S)dS]
J(a+b)/2
2A o +1
(b_a)A-O
> (A _ 0)(0 + 1)0+1 -2,
(3.6.83)
where
A> 0
or
_lbsino+1(7r~s~:))q(S)dS >
2(0+2)'" ( 7r
(0+1)0+1 b-a
)0
(3.6.84)
Then, q E O",a,b)).
Proof. Condition (3.6.82) follows from (3.6.81) in the case when f(t) =
[(t - a)(b - t)]o+1 = g(t) for a < t < b; condition (3.6.83) is obtained in
the case when f(t) = (t - a)A and g(t) = (b - t)A for a < t < b, and
condition (3.6.84) in the case when f(t) = sin (7r(t - a)/b - a) = g(t) for
a < t < b.
Chapter 3
136
Finally, we shall present results which ensure that the function q(t) E
A(t)
B(t) =
It
it
(b - t)l-A
(s _ a)l+o.A(b _ s)C>A[q(s)rds
(t-a)A a
a)l-A
(t
+ (;_ t)A t (s - a)O>'(b - s)l+o.>'[q(s)rds,
Ib
(s-a)l+o.>'(b-S)C>A[q(S)l-dS-l (s-a)cx>'(b-S)l+o.>'[q(S)l-ds
Yo(t)
+ A(t)
for
a < t < b.
(3.6.85)
(3.6.86)
Let Xl (t) and X2(t) be proper solutions of the problems (3.6.1), (3.6.4)
and (3.6.1), (3.6.5) respectively. By Lemma 3.6.5 (in this case u(t) = yo(t)
for a < t < b) there exist tl E [co, b) and t2 E (a, co] such that
X~(t)
x~(t)
Suppose tl
>
<
for
for
a<t<tl,
t2 < t < b,
x~(tr)=O
X~(t2) =
0.
(3.6.87)
Xl (t)
for
Xl (co)
X2(t)-(-)
X2 Co
a < t < Co
Co < t < b
for
Define
l7(t)
~~~~~ 10.
Oo((a, b.
sgn yb(t)
for
(3.6.88)
a <t < b
137
then by (3.6.85), (3.6.86) and Proposition 3.6.1, we find that aCt) satisfies the conditions of Lemma 3.6.11, and hence q rJ. Ooa, b)), which
contradicts (3.6.88).
Consequently, iI > t2. Then from (3.6.87) there exists c E (tl' t2)
such that x~ (c) > 0 > x~ (c). It can be easily verified that the function
= {
u(t)
Xl(t)
for
Xl(C)
X2(t) X2(C)
a<tS;c
for
<t <b
where
C = max
{l
(s-a)HO(b-S)Ct[q(S)]-dS'
(s-a)O(b-s)O+l[q(S)]-ds}.
then q E Uoa,b)).
~(A),
(_0_)0<
where A> _ _
1_
0+1 0+1
root of the equation T - ITlo/(o<+l) - A = O.
Denote by
0+1
( -0 -
0+1
'
)0
and
(3.6.89)
Then, q E Uoa, b)).
Proof. Let k = ~(Q*(to,o)),
aCt)
Q(t, to, 0) - k
for a < t < to
(t - a)o<
{
Q(t, to, 0) - k
(b-t)Ct
for to<t<b.
(3.6.90)
Chapter 3
138
a<t <b
for
(3.6.91 )
from which it follows that a(t) sgn (to - t) > 0 for a < t < b, t =I- to and
la(t)l(a+l)/a :::;; k h(t)
where
h(t) =
( _
t
)a+1
y(t)
(3.6.92)
a < t :::;; to
for
(3.6.93)
(b _ t)a+l
Define
a < t < b,
for
1:
to < t < b.
for
la(s)11/a d s)
for
a < t < b.
y"(t)
q(t)ly(tWly'(t)1 1 -
+ y(t)la(t)l(l-o)/a
(la(t)l(o+l)/a - kh(t)
- -1a+1
( -a-
a+1
)0
_ a )a+1
a+1
Then, q E Ua((a, b.
Theorem 3.6.6. Let
- -1-
a+1
( -a-
a+1
)a
and either
or
Q*(t o, a)
and
Q*(to, a) < 2.
(3.6.95)
139
AC!oc((a, to)
u(to+)
(to, b)),
-(b-to)a'
lim (t - a)au(t)
t ..... a+
k<1
for
a < t < b.
a+1
a+1
a ) a
a+1
Q*(to, 1) =f. 0,
or
Q*(to, 1) = 0,
Then, q 'f- Ol((a,b)) (i.e"
Q*(to, 1) < 2.
q E U1((a, b)).
140
Chapter 3
w'(t)
[T,oo).
::::: 0 on
(3.7.1)
for all
t 2:': T.
(3.7.2)
u(t)
is a solution of
Isl(1-</os.
exp ( [ 1j;-1
(a(t)1j;(x'(t)'
(:f:j) ) ds
+ qdt)1j;(x(t =
0,
where
1j;-l(s) =
10
00
a-1/O(s)ds
00
and
Joo q(s)ds
< 00.
(3.7.3)
It:
141
(3.7.5)
Proof. (i) => (ii). Let x(t) be a nonoscillatory solution of (3.1.1).
Without loss of generality, we can assume that x(t) > 0 on [T,oo) for
some T ~ to. Define
1j;(x'(t))
w(t) = a(t) 1j;(x(t))
on
(3.7.6)
[T,oo).
w'(t)
Let t
+ aa~l/<>(t)lw(t)I(Q+l)/<> + q(t)
on
[T,oo).
w(~)-w(t)+a
If. a~l/Q(s)lw(s)I(<>+l)/<>ds+If.
[t,~],
(3.7.7)
we get
q(s)ds = 0 for
~ ~ t ~ T.
(3.7.8)
We claim that
(3.7.9)
Suppose to the contrary that ftOO a~l/Q(s)lw(s)I(<>+l)/Qds = 00. Then
in view of (3.7.8) and (3.7.3) there is a To ~ t large enough so that for
~ ~ To,
w(O
=
rf. a~l/<>(s)lw(s)I(<>+l)/<>ds
w(t) -If. q(s)ds - a ITO a~l/Q(s)lw(s)I(Q+l)/Qds
+a
lTo
< - 1,
or equivalently,
-w(~) ~ 1 + a
rf. a~l/<>(s)lw(s)I(Q+l)/Qds
lTo
for
~ To.
(3.7.10)
(3.7.11)
142
Chapter 3
~,
we obtain
w(t)
00
q(s)ds + a
00
a-1/a(s)lw(s)l(a+l)/ads.
Then, (3.7.1) holds. Now, it follows from Theorem 3.7.1 that equation
(3.1.1) is nonoscillatory.
h'(x)
= {
Theorem 3.7.3. Let condition (3.7.3) hold. Then the following three
statements are equivalent:
(i)
143
y(t)
00
[IQ(s)
+ O:Y(S)I(o+l)/o -
a-1/o(s)J.l[S, t]ds
(3.7.13)
Z(t) =
00
+ o:z(s)I(O+l)/a -
[IQ(s)
C([T,oo),lR)
x a-1/a(s)J.l[S, t]ds.
(3.7.14)
y(t)
+
~
J.l[T, t]y(T)
I
l
[IQ(S)+O:Y(sW-::! -
0+1
IQ(s)l---;:;-a-"(s)J.l[s,t]ds.
(3.7.15)
The left hand side of (3.7.15) is independent of T, and hence as T -t- 00,
(3.7.13) follows.
(ii) =} (iii). Suppose there exists a function y(t) E C([T, 00), lR) which
satisfies (3.7.13). Define
Zo(t)
y(t)
Chapter 3
144
(iii) =} (i). Assume that (3.7.14) has a solution z(t), then an easy
computation shows that w(t) = -[Q(t) + az(t)]ja(t) defines a solution of
(3.7.1). Now, as a consequence of Theorem 3.7.1, we conclude that equation
(3.1.1) is nonoscillatory.
q(t)
7T(t)
00
a-l/<>(s)ds
and
7T(to) < 00
(3.7.16)
Lemma 3.7.2.
alxl(n+l)/Ct
+ AX
g(x)
Proof. Since
g'(x)
(a
+ l)lxl"
sgn x
+ A 2"
0 for
x 2" -(Aj(a
+ I))",
Lemma 3.7.3. Assume that q(t) E C([to, 00), JR+) and 7T(to) < 00. Let
x(t) be a solution of equation (3.1.1) such that x(t) # 0 for t 2" T 2" to.
Then, x(t) as well as the :unction
<>( ) ()1jJ(X'(t))
7T tat 1jJ(X(t)).
(3.7.17)
1jJ x t
t-'>oo
o.
(3.7.18)
(3.7.19)
Proof. Without loss of generality, we assume that x(t) > 0 for t 2" T 2"
to. Since a(t)1/J(x'(t)) is nonincreasing on [T,oo), we find that x'(t)
is eventually of constant sign, i.e., x' (t) > 0 for t 2" T, or there is a
TI > T such that x'(t) < 0 for t 2" TI and
for
s 2" t 2" T.
(3.7.20)
Dividing both sides of (3.7.20) by al/<> (s) and integrating it over [t,~]
gives
x(~)
<::: x(t)
+ al/<>(t)x'(t)
if,
a-l/a(s)ds
for
~ 2" t 2" T.
(3.7.21)
145
00,
t> T.
for
we conclude that
a
1j;(x'(t))
t~~ n (t)a(t) 1j;(x(t)) ~ O.
0:
+1
(3.7.22)
and there exists a function y(t) E C([T, (0), JR), T 2 to such that
nQ(t)y(t)
2l
oo
+0:
n),(s)q(s)ds +..\
loo a~I/Q(s)n)'~I(s)y(s)ds
loo a~I/Q(s)n),(s)ly(s)I(Q+I)/Qds
(3.7.24)
for t 2 T.
Proof. (The 'only if' part). Let x(t) be a solution of equation (3.1.1)
such that x(t) lOon [T, (0) for some T 2 to. Define w(t) =
a(t)1j;(x'(t))N(x(t)) on [T, (0). Then, we have
w'(t)
+ o:a~I/Q(t)lw(t)I(Q+I)/a + q(t)
on
[T, (0).
(3.7.25)
Chapter 3
146
we obtain
1(.
nA(~)w(~) -
nA(t)w(t) + A
a-1/0(s)nA-l(s)w(s)ds
y(t)
n-A(t)
+0:
[1
00
00
00
-1 for t
(3.7.26)
~ T is
a- l / O(s)nA-l(s)w(s)ds
a-1/Q(s)nA(s)lw(s)I(O+l)/QdS] .
nQ(t)w(t)
nA(s)q(s)ds + A
y(t) for t
nQ(t)y(t)
00
a-l/o(s)nA-Q-l(s)ds
0:+1
for t
1 ( -A- ) Q+l
A-O: 0:+1
=
~
- --
Since
y'(t)
for t
y'(t)
T, we have
+ q(t) + o:a-l/O(t)ly(t)I(Q+l)/Q
An- l (t)a- l / o (t)y(t) _An-l(t)a- l / o (t)w(t) - o:a-l/a (t) Iw(t) I(o+l)/a
+ o:a-l/O(t)ly(t)l(o+l)/Q
a-l/o(t)n-o-l(t) {[o:lnO(t)y(t)l(o+l)/O
- [o:lnO(t)w(t)l(a+l)/o
+ AnO(t)y(t)]
+ AnO(t)w(t)]}
147
(al(t)'lj;(y'(t)' +ql(t)'lj;(y(t
(3.7.27)
0,
11= q(s)dsl
1=
ql(s)ds
00,
and
t 2: to
for
w'(t)
+ ql(t) + aa~l/a(t)lw(t)l(a+l)/a
~ 0
for
t 2: T.
Using the fact that aCt) 2: al (t) in the above inequality, we obtain
w'(t)
+ ql(t) + aa-I/O(t)lw(t)l(o+I)/o
~ 0
for
t 2: T.
(a(t)'lj;(x'(t)))'
+ ql(t)'lj;(X(t =
(3.7.28)
ly(t)1
>
11=
>
1=
1=
ql(s)ds + a
a-I/O(s)ly(s)l(o+I)/O!dsl
a-I/O!(s)ly(s)l(o+I)/O!dsl
Once again, we can apply Theorem 3.7.2 to conclude that equation (3.1.1)
is nonoscillatory. This completes the proof.
Next, we shall consider equations (3.1.1) and (3.7.27) where q(t), ql(t)
EC([to,oo),JR+), and
1=
t
--ds
ales) .
(3.7.29)
Chapter 3
148
:s:
:s:
q(t)
on
[T, (0).
(3.7.31)
1.
and
:s: -q(t),
2: to
(3.7.33)
(3.7.34)
cf>'(t)
If
t-+oo
149
w(t)
where A = ((a
+ l)/a)O;.
Aa(t)
_ A,.'()
'+'
- Ak)
+ 1hO;(t)
,
(_a )0;+1
1
< 1,
a+1
a+ 1
A(t)
to
Then, we have
= (~)O;.!.
(a(t)w(t))'
(3.7.35)
and
q(t)
+ aa(t)lw(t)I(O;+l)/O; - (a(t}w(t))'
:S q(t)
1-Ak) 1(0;+1)/0;
a(l-Ak) a- 1 / o (t)
-q(t)A
h a+1(t)
= aa-1/O(t)h-(o+1)(t) [I(t)ha(t) +
:Saa-1/<>(t)h-(<>+l)(t)
ra+1)/a _
~ Ak]
= aa- 1/<>(t)h-(o+1)(t) (
=
1 -A Ak
A(o+l)/o
.!.A + k)
h'(t) :S - a- 1/0;(t)
for
t ~ to
(3.7.36)
and
'(t)
q(t)
for
t ~ to,
(3.7.37)
respectively.
The proof of Theorem 3.7.9 is the same as that of Theorem 3.7.8 except
that now, we let
1
w(t) = Aa(t)
1- Ak)
A(t) + h<>(t) .
150
Chapter 3
Theorem 3.7.10. Let the functions h(t) and (t) be as in Theorem 3.7.8
such that conditions (3.7.32) and (3.7.33) hold. If there exists a number
k > 0 such that
_ko)(o+l) -- k
:s
hOo(t)I(t)1
:s
(3.7.38)
Let
w(t)
1 ( (t)
a(t)
k )
+ hOo(t)
Then, we have
(a(t)w(t))'
,h'(t)
- (t) + ak h a+1(t) > q(t)
a- 1/"'(t)
+ ak h a+1(t)
and
q(t)
+ aa(t)lw(t)I("'+l)/c>
1
- (a(t)w(t))'
(t)
+ h"(t)
) 1("+1)/00
a- 1 /a(t)
- q(t) - ak hO:+ 1(t)
+ kl(a+1)/a
- k]
o.
+ hQk(t))
(3.7.39)
where f3 > a
(t)
h(t)
~t(o:-(3)/""
f3-a
(t)
= _()_t(3-a
f3-a
f3-a
151
lim <f>(t)
exists and
t-'too
'(t) ~ - h"'(t)q(t),
(3.7.40)
(a(t)w(t))'
h'(t)
'(t)
a- 1 / Q (t)
o.(m+(t)) hQ+1(t) - hQ(t) > q(t) +o.(m+(t)) hc>+l(t) ,
which implies
+ o.a(t)lw(t)I(Q+1)/c> - (a(t)w(t))'
m + (t) 1(<>+1)/C>
a- 1 /<>(t)
~ q(t) + o.a(t) 1a(t)hc>(t)
- q(t) - o.(m + (t)) hc>+l(t)
< o.a- 1/ Q(t)h-(c>+1) (t)[(m + <f>(t)) - (m + (t))] = O.
q(t)
= exists and
or
1
a
h'(t) - a- 1 / (t) 2: -
t-'too
2: 1/0.,
(3.7.41)
(3.7.42)
152
Chapter 3
Then, we have
q(t)
::; h(}+l(t)
+a
ah-(<>+l)(t)
a- 1/<>(t)
h'(t)
h<>+1(t) - a h<>+l(t)
(~ + a- 1/<>(t) -
h'(t))
::; 0
for
00
t 2: T.
a- 1/<>(s)</>(cx+1)/CX(s)ds <
--
) (<>+1)/<>
a+1
</>(t) ,
w(t)
__1_
a(t)
[</>(t)
+ a(a + 1)(0+1)/<>
roo a- 1/
it
(s)</>(a+1)/<> (S)dS] .
Then, we have
which implies
q(t)
+ aa(t)lw(t)I(<>+1)/<> - (a(t)w(t))'
~+1
::; q(t)+aa(t)
= o.
153
Then, we have
+ l)l/o. a-l/o(t)
aa-l/o.(t) [r/>(t)
x
{[r/>(t)
+ r/>(o.+l)/o.(t)]
+ a(a + l)l/Or/>l(t)]
[a(a + l)l/o.r/>l/o.(t)r/>l(t)
- (a
+ 1)1/0r/>1/0(t)}
+ a(a + l)l/r/>l(t)]
-(a + l)l/o.r/>l/o.(t)] =
o.
11
00
r/>o.(s)q(s)dsl < 00 on
[T,oo),
(3.7.43)
154
Chapter 3
Then, we have
o<
1 + 21
00
t::;> T1 .
Thus, we find
2-(0+1)/0cjJ-(0+1)(t) [1 + 21 00 cjJO(S)q(S)dS]
ly(t)l(a+1)/0
<
~cjJ-(Q+1)(t)
[1+21
00
cjJ<>(s)q(S)dS].
Consequently, (3.7.44) implies that (3.7.1) holds on [T1' 00). Now, Theorem
3.7.1 guarantees that equation (3.1.1) is nonoscillatory.
(i)
RQ(t)
00
t,
155
(ii)
RO(t)
00
have
00
!..O! (_0!_)0+1
0!+1
t.
1 ( O! )0+1
-R-O(t).
O! 0!+1
q1(s)ds = -
00
a- 1/ O(s)lc(s) + Q(S)I(0+1)/Odsi
for t 2': T.
(3.7.46)
Corollary 3.7.1.
If Q(t) 2': 0 and there exists a function
C ([T, 00), JR +) for some T 2': to such that
c(t) 2': O!
00
a- 1/ O(s)lc(s) + Q(s)I(0+1)/ods
for
2': T,
c(t)
(3.7.47)
Chapter 3
156
Corollary 3.7.3. If
00
a-l/<>(s)IQ(s)I{<>+l)/<>ds <
a+1
) (a+l)/<>
IQ(t)l,
a_)
<>+1 ,
_k<>/{<>+l) - k :::; R<>(t)Q(t) :::; k<>/{<>+l) - k :::; _1 ( __
a a+1
then equation (3.1.1) is nonoscillatory.
Now, taking yet) = Ql(t)+b(t) in (3.7.13) for some bet) E C([T, oo),lR)
where T ~ to, we have the following nonoscillatory characterization result.
Theorem 3.7.20. Let I,ds, t], Q(t) and Ql(t) be as in Theorem 3.7.3.
Then equation (3.1.1) is nonoscillatory if and only ifthere exists a function
bet) E C([T, 00), lRo) for some T ~ to such that
bet)
00
(,X - l)Ql(t)
00
{IQ(s)
(a
+ l),XIQ(sW/<>
157
1=
Cl!~l)
(3.7.50)
ft"o
Thus, we have
<
1=
Q(OI+!)/<>(s)a- l / OI (S)J.ll[S,tjds
1
r= Q(<>+l)j<>(s)ds <
l-'Y(o+l)it
o
~ 1 (( 0:
1)01 - 1) Q(t).
Chapter 3
158
where ). is a constant. If one of the following conditions holds
(i)
(ii)
)'~1
a).
).0 _
1 ;3(t),
a).
IAI
A;3(t) :::: 4>(t) :::: -aA;3(t),
-1 +
0 sgn
14>(t) +aA;3(t)IC a + I)/a - A{14>(t)I Co+I)/a + aAI4>(t)I I /;3(t) sgn 4>(t)} :::: O.
Let wet)
q(t)
+ aa(t)lw(t)I,t' - (a(t)w(t)'
-A [14>(t)1 0;;'
aa-I/a(t) {14>(t)
+ aAI4>(tW/;3(t)
+ aA;3(t) I0;;'
00.
t ~ to
(3.7.51)
159
(i)
1)
_(
7rt
(ii)
00
7r0+1(s)q(s)ds:::;
()<>+1
_0._
0:+1
t,
-1()
7rt
00
(3.7.52)
+ 1))<>+1 ,
(i)
(ii)
,-a
_o._t(<>--Y)/O
and
(_a
,-a
)0+1
(fYJx'(t)J
O -
0,
t:::: to
(3.7.53)
160
Chapter 3
where
(i)
'I > a
is a constant. Then,
00
< (
s-b-o:)(o:+l)/u q (s)ds
a ) ('I~
~ a) 0:
1=
> A* > (
s-b-o:)(e>+l)/D q (s)ds
a )
0'+1
('I ~ 0')0:
0'+1
+ F(t, x(t))
(3.7.54)
+ q(t)f(x(t))
0,
(3.7.55)
(lx'(t)l a - 1x'(t))'
and
(lx'(tW-1X'(t))'
where
a > 0 is a constant,
(i)
(ii) FE C([to, 00) x IR, IR), sgn F(t, x) = sgn x for each t E [to, 00),
(iii) f E C(IR,IR), sgn f(x) = sgn x, and
(iv) q(t) E C([to,oo),IR+).
Theorem 3.7.23. Suppose F(t,x) is nondecreasing in x for each fixed
t and has the partial derivative with respect to t such that
of
sgn Tt(t, x)
~ sgn
for
t 2': to
(3.7.56)
j= IF
and
<
001
00
1m sup
\--+0
F(S,AS)d
\
AO:
<
00,
(3.7.57)
(3.7.58)
Proof.
{td~l
161
for k = 1,2, .. '. Let Tk be the first zero of x'(t) which is located to the
right of tk' Since x'(t) > 0 on (t k , Tk) and x'(t) is decreasing there,
we have
x(t) = X(tk)
for
Jtk
t E
(tk' Tk)'
(3.7.59)
Define
a
V(t) = --lx'(t)I"+l
a+1
x (t)
F(t,u)du
for
t 2:: tl.
(3.7.60)
V'(t)
(lx'(t)I",-IX'(t))' x'(t)
rCt)
+ F(t, x(t))x'(t) + J o
rct) of
Jo
7it(t,u)du:::; 0 for
of
7it(t, u)du
t 2:: tl'
This implies that V(t) is bounded for t 2:: tl' Thus, in view of (3.7.60),
x'(t) remains bounded as t ---+ 00. It follows that there exists a constant
M > 0 such that
(3.7.61)
Ix'(t)1 :::; M for t 2:: tl
Now, we integrate (3.7.54) from tk to Tk, and use (3.7.59) and (3.7.61),
to obtain
<
roo
F(s, Ms)ds,
rkF(s,Ms)ds:::;
Jt k
Jtk
r k F(s, SX'(tk))
(X'(tk))" ds
1 :::; Jtk
and
00,
for
= 1,2,.
(3.7.62)
<
-
roo
00,
Chapter 3
162
Theorem 3.7.24.
lR+) and satisfies
00
[q'(s)]+
ds <
q()
s
00,
(3.7.63)
(3.7.64)
Joo q(s)lf(cs)lds
and
<
00
<
(3.7.65)
00.
Vet) = --lx'(t)ICt+1
a
+1
+ q(t)
(t)
f(u)du,
t 2: to
V'(t)
r(t)
q'(t) io
r(t)
f(u)du::; ([q'(t)]+) io
f(u)du,
which implies that V'(t) satisfies the first order differential inequality
for
t 2: to.
Hence, we have
(it [q'(s)]+
dS),
s)
to
q(
t 2: to.
Thus, Vet) is bounded. The definition of Vet) now shows that x'(t)
remains bounded as claimed. Proceeding as in Theorem 3.7.23, we get the
required contradiction.
t 2: 1
(3.7.66)
0:,
{3, ,
163
(3
are oscillatory.
+ q(t)x(t)
0:
(3.8.1)
1, (3.8.1) reduces to
o.
(3.8.2)
In the oscillatory theory of (3.8.1) there are several results which are
similar to those known for the equation (3.8.2). We state some of these in
the following theorems.
Theorem 3.8.1. Equation (3.8.1) is oscillatory if
0:
(_
0:
+1
and is nonoscillatory if
0:
limsup t"+lq(t) < ( _
t->oo
0:
+1
),,+1
),,+1
164
Chapter 3
ex: 1)
t<>+lq(t) :::::
<>+1
+ to
and is nonoscillatory if
for all sufficiently large t.
Theorem 3.8.2. Suppose q(t) > 0 for t::::: to ::::: O. Then equation
(3.8.1) is oscillatory if either one of the following conditions holds
(i)
(ii)
00
q(s)ds >
(0:
oP
+ 1)<>+1
00
q( s )ds <
0:0/
(0:+1)<>+1'
S(t
where 11"0/
+ 11"0/) = S(t)
and
IS(tW+1
+ IS'(tW+1 =
1,
211"
.
(0: + 1) sin(1I"/(0: + 1)
x(t)
p(t)S((t,
x'(t)
p(t)S'((t).
p'(t)
p(t)
and
'(t)
IS'((t))10/+ 1
+ ~q(t)IS((t)W+1.
0:
(3.8.3)
165
(3.8.4)
mf
x#o
(i)
IF(t,x)
x I 1x
C>~
F(t,x)
~l
x a X
II
()
2': q t for all sufficiently large t.
sup
:<::;
q(t)
(3.8.5)
(3.8.6)
(3.8.7)
for t 2': 1, where a, b, (3, A and J-t are positive constants. Since
and
sup
b+ti3 lxI A
< -
x#o a+tl-'lxlA -
a'
(3
< J-t
for all large t, it follows from Theorems 3.8.1' and 3.8.5 that
(i)
+ 1)t+1 ,
+ l))a+1 .
Finally, we shall prove two results for the more general equation, namely,
0,
(3.8.8)
166
Chapter 3
Theorem 3.8.6.
00.
If
lim
t-+oo
it
to
q(s)ds =
JO
a- 1/ n (s)ds =
(3.8.9)
00,
Proof.
w(t)
a(t)
Ix'(t)IO- 1 x'(t)
()
xQ t
for
(3.8.10)
t:::: to.
Then, we have
O.
t:::: to.
(3.8.11)
Thus, from (3.8.11) it follows that w'(t) +q(t) ::; 0 for t:::: to. Integrating
this inequality from to to t, we obtain 1L(t)-w(to)+ft:q(s)ds::;O, t::::
to. Hence, w(t) ---c> -00 as t ---c> 00, and therefore x'(t) < 0 for t:::: t1
for some sufficiently large t1 :::: to Choose t2 :::: t1 so that
q( s )ds > 0
for t > t 2 . Integrating (3.8.8) from t2 to t, we get
-
fL
(3.8.12)
Now, from the relation
fL
we find that
q(s)(xO(s))ds :::: 0 for t:::: t2' Thus, from (3.8.12) we
can conclude tha:-t a(t)lx'(t)10-1x'(t)::; a(t2)lx'(t2)10- 1x'(t2), t:::: t2 and
hence
(3.8.13)
An integration of (3.8.13) shows that x(t) ---c> -00 as t ---c> 00, which contradicts the fact that x(t) > 0 eventually. This completes the proof.
Theorem 3.8.7.
t-+oo
i (i8 a(u)du)-1/0 ds =
t
to
to
00.
(3.8.14)
t-lim
'>oo -litis
to q(u)duds
t
to
00,
to ~ 0
167
(3.8.15)
to
to
(3.8.16)
V(t) <
It: Iw(s)lds.
(L
a(s)ds
:<::: co,
) 1/(a+1) (
10 a- 1/a(s)lw(s)ICa+1)/a ds)
t
a/(a+1)
or
(
:<:::
lto
a- 1/a(s)lw(s)l(a+1)/ads for t
(3.8.17)
to such that
< 0 for t
--
~ to.
~ t1.
(3.8.18)
V(t)
it
to
---+-
r/
v(a+1)/a(s)
(It: a(u)du
ds < 0,
v(a+1)/a(t)
Let
W(t)
Chapter 3
168
Then, we have
W'(t)
;::: 0:(<>+1)/<>
W(<>+l)/<>(t)
(lot
a(s)ds) -1/<>
t;::: t2'
for
(3.8.19)
i (1S
t
t,
to
W (t)
Wet,)
a(u)du
-1/<>
ds:S
it W(<>+l)~<>(s)
W'()
t,
ds
dv
[
]
= 0: W- 1/<>(t1) - W- 1/<>(t)
v(<>+l)/<>
:S o:W- 1/<>(td
it (1
lim
t-+oo
t,
to
a(U)du)-l/<> ds :S 0:-aw- 1/ a (t 1)
<
00,
(3.9.1)
(3.9.2)
H(t, t) = 0
for
t;::: to,
H(t, s) > 0
for
and
h(t, s)
If
.
hmsup H(
t-+oo
t, to
1t [
to
H(t, s)q(s) -
a(s)ha+l(t, S)]
Ha()
ds
0:0
t, s
00,
(3.9.3)
169
Proof.
for t
w'(t) =
and consequently,
H(t, s)w'(s)ds
to
to, we have
q(t) - a
Iw(t)j(<>+l)/<>
a1/<>(s)
it
to
Since
H(t, s)w'(s)ds
- w(to)H(t, to) -
H(t, s)q(s)ds
S;
H(t, to)w(to)
to
-a
to
. to
H(t, s)
Iw(s)I(<>+1)/<>
1/<>() ds.
a
s
(3.9.4)
f)
W(S)-f) H(t, s)ds,
(3.9.5)
to
to
l.t
H(t, s)q(s)ds - a
it
h(t, s)lw(s)lds
to
(3.9.6)
Iw(s)I(<>+1)/a
H(t, s)
1/<>() ds.
a
s
Letting
A
and
(aH(t, s0/(0+1)
B
Iw(s)1
Iw(s)lh(t,s) - aH(t,s)
for all t
1
it
H(t, to)
to
+1
to,
Iw(s)I(<>+l)/Ct
a1/Ct(s)
t, to
A= a
H(
a 1/(0+1)(s)
<
it
to
a(s)hCt+l(t, s)
aoH<> (
t,s) .
a(s)h<>+l(t s)
HCt( ') ds (3.9.7)
ao
t, s
it
to
for
~ to.
Chapter 3
170
Taking the upper limit as t ---+ 00, we get a contradiction, which completes
the proof.
- osH(t,s) = h(t,s)Ha/(o+l)(t,S)
If
limsup H( 1
t, to
t->oo
it
to
for
(t,s) E V.
(3.9.8)
(3.9.9)
00,
t, to
t->oo
r [H(t'S)P(S)q(S)
lto
_ a(s)p(s) (h(t,s)
aoH"'(t, s)
+ p'es) H(t,S))"'+l] ds =
pes)
00,
(3.9.10)
then equation (3.1.1) is oscillatory.
Proof. Let x(t) be a nonoscillatory solution of equation (3.1.1). Without
loss of generality we assume that x(t) #- 0 for t ~ to. Now, we define
wet) = p(t)a(t)lxl(t)I"'-lx'(t)/(lx(t)I"'x(t)) for t ~ to. Then for every
s ~ to, we have
I
w (s) = - p(S)q(S)
p'es)
Iw(s)I(",+l)/a
peS) w(s) - a (a(s)p(s))l/o'
r H(t, s)w'(s)ds
lto
it
H(t, s)p(s)q(s)ds +
to
(3.9.11)
it -(-)
p'es)
to
Iw(s)l(a+l)/a
w(s)ds
l:
~ H(t, to)w(to)+
H (t, s)p(s)q(s)ds
1:
(h(t, s)+
171
Iw(s)l(o+l)/o
-a lto H(t, s) (a(s)p(s))1/0ds.
(3.9.12)
(H(
= a
t, s
))<>/(0+1)
Iw(s)1
(a(s)p(s))1/(<>+l) '
and
p'(s)
+ p(s) H(t, s)
)<>
to, we obtain
Iw(s)1 ( h(t, s)
p'(s))
Iw(s)I(<>+l)/<>
p(s) H(t, s) - aH(t, s) p1/<>(S)
a(s)p(s)
(
~ (a + 1)<>+lHo(t, s) h(t, s)
(3.9.13)
p'(s)
) <>+1
p(s) H(t, s)
t r
1
a(s)p(s) (
H(t, to) lto lH(t, s)p(s)q(s) - aoH<>(t, s) h(t, s)
p'(s)
\ <>+1]
p(s) H(t, s))
ds
~
00,
w(to).
(3.9.14)
we obtain a contradiction.
it
to
a(s)p(s) (
p'(s)
)<>+1
H<>(t, s) h(t, s) + p(s) H(t, s)
ds <
and
limsup H( 1
it
)
H(t, s)p(s)q(s)ds
t, to to
then the conclusion of Theorem 3.9.3 holds.
t-J.oo
00
00,
(3.9.15)
(3.9.16)
(t"lx'(t)1 0
1 x'(t))'
[-yfY- 3 (2-cost)
= 0
(3.9.17)
172
Chapter 3
1/
are
Here, we choose p(t) = t 2 and H(t, s) = (t - S)2 for t 2:: s 2:: to.
Then since p(t)q(t) = [fY(2 - cost)]', we find
p(s)q(s)ds 2:: fY - ko
for some constant ko > 0, and therefore,
It:
1 {t(
)2 ( ) ( )d
2fY
kl
k2
k
t2lt o t-s psqs s 2:: (!+I)(!+2)+t 2 +T- 0,
where
and
2t"+1
k2 = 2koto _ _0_.
,+1
t2
it
to
s 1/+2
(t - s) 20
= 2 +1t o - 1
0
2
)0+1 ds
2(t-s)+_(t-S)2
S
t ) 1-0 tl/- a +2
tl/- o +2
0
l/-a+2
~ ltor
[(t - s)np(s)q(s)
~o a(s)p(s)(t -
s)n-a-l
(~g? (t -
s)
n)
0+1] ds = 00,
(3.9.18)
then equation (3.1.1) is oscillatory.
Corollary 3.9.3. If
limsup
t-400
r [(t - stq(s) -
lto
n o+ 1 a(s)(t - s)n-o-l] ds
ao
00, (3.9.19)
o<
inf
s?.to
{liminf
t-400
:?'
s))} :::; 00
t, to
(3.9.20)
173
lto
a(s)p(s) (
H"'(t, s) h(t, s)
nE
li~~p
H(:, T)
)"'+1
ds < 00.
(3.9.21)
C([to, 00), lR) such that for every T 2:: to,
[H(t, s)p(s)q(s)
a(s)p(s) (
- ooH"'(t, s) h(t, s)
and
oo
to
where n+(t)
p'es)
+ pes) H(t, s)
(3.9.22)
p'es)
)"'+1]
pes) H(t, s)
ds >
= max{O, n(t)},
00
neT)
'
(3.9.23)
Proof.
li~~p
H(:, T)
[H(t, s)p(s)q(s)
_ a(s)p(s) (h(t,s)
00
pes)
neT) : : ;
and
lim sup H( 1
t, to
t ..... oo
it
to
weT)
for every
T 2:: to
U(t)
H(t,to)
and
Wet) = H( 1 )
t, to
where
(J(t, s) = h(t, s)
Iw(s) 1(<>+1)/'"
it
to
(J(t, s)lw(s)lds,
p'es)
pes) H(t, s),
t 2:: s 2:: to
(3.9.24)
(3.9.25)
Chapter 3
174
liminf[U(t) - W(t)]
t-+oo
t, to
t-+oo
:s:
1t
to
H(t, s)p(s)q(s)ds
00.
(3.9.26)
Next, we claim that
1=
to
Iw(s)l(a+l)/a
~'--C-'-....,...-;-:-:;-;-ds
(a( s )p(s) )I/a
<
00.
(3.9.27)
00.
(3.9.28)
= Iw(s)l(a+l)/a
-:--:'--:--c-:-:-:;-;-
(a(s)p(s))l/a
to
ds
s2>:to
{ lim inf
t-+=
II(t,s) }
(
)
H t, to
>
such that
> O.
(3.9.29)
Let /-l be an arbitrary positive number. Then it follows from (3.9.28) that
there exists a tl > to such that
1
t
to
Iw(s)I(a+l)/a
/-l
( ) ( 8 )) l/a ds > -~
(asp
U(t)
= H(t,Q to)
1t
1t (a
(1
H(t, s)d
for all
IW(T)I(a+l)/a
(a(T)p(T))l/a dT
) (1
H(t, to)
t (-.!!-.H(t, 05)) (1
H(t, to) it,
as
Q
to
to
--H(t, s)
as
/-lQ
~H(t, to)
1t
t,
to
t:::: tl
to
S
to
IW(T)I(a+l)/a dT ) ds
(a(T)p(T))l/a
IW(T)I(a+l)/a dT) ds
(a(T)p(T))l/a
From (3.9.29), we find that there is a t2 > tl such that H(t, td/ H(t, to) ::::
for all t:::: t 2, which implies that U(t):::: /-lQ for all t:::: t2' Since jJ
is arbitrary
lim U(t) = 00.
(3.9.30)
~
t-+oo
00)
= 00
n--+oo
= liminf
[U(t) - W(t)] <
t---+oo
00.
175
lim U(Tn)
n-+oo
(3.9.31) implies that
lim W(Tn)
n-+oo
(3.9.32)
00,
(3.9.33)
00.
Therefore, we have
W(Tn)
1
U(Tn) > 2"
(3.9.34)
n.
.
W<>+l(Tn)
hm
=
n-+oo U<>(Tn)
(3.9.35)
00.
<
and hence
1
<>
H(T)
n, to
Tn
to
f3<>+l(T s)
a(s)p(s) H<>(T n, ) ds.
n, S
Chapter 3
176
n--+oo
H(T
to
n,
which gives
.
11m
t--+oo
H t, to
Tn
to
(3<>+I(T s)
a(s)p(s) HO(T n, ) ds
00,
n, S
a(05)p(o5)
to
(30+1(t,S)
( ) ds =
HO: t,05
00.
(Q+(S))(u+l)/o:
(a(s)p(s))l/O ds <
OO
to
!w(s)!(a+l)/o
(a(s)p(s))l/ods <
00,
it
liminf H( 1 )
t--+oo
t, to
to
liminf
t--+oo
(1
H t, T
.IT
00),
00.
(3.9.36)
[H(t, s)p(s)q(s)
a(s)p(s) (
- aoHo(t, s) h(t, s)
(3.9.37)
p'(s)
)"'+1]
p(s) H(t, s)
ds :2: Q(T)
t--+oo
t, to
it
to
H(t,05)p(o5)q(s)do5 <
00,
where U(t) and W(t) are as in the proof of Theorem 3.9.4. Now, it
follows from condition (3.9.37) that
..
-11mlllf
t--+oo
t, to
it
it
1
()
H t, to
to
H(t, s)p(s)q(s)ds
to
a(s)p(s) (
p'(s)
)<>+1
H ( ) h(t,s)+-(-)H(t,s)
ds,
t, s
Ps
ao
177
t--'too
ao
H( t, to )
1t a(s)p(s) (
H a (t, S ) h(t, s)
to
p'(s)
+ -(-)
P s H(t, s)
)a+1
ds < 00.
Now, define H(t, s) = (1r(t) -1r(s))n for t 2:: s 2:: to, where n> a is
a constant and 1r(t) E C 1([to,oo),lRo) with 1r'(t) = 1/a 1/ a (t) for t 2:: to
and limt--'too 1r(t) = 00. Then, H(t, s) is continuous on V = {(t, s) ; t 2::
s 2:: to} and satisfies H(t, t) = 0 for t 2:: to, H(t, s) > 0 for t> s 2:: to
Moreover, H(t, s) has a continuous and nonpositive partial derivative on
V with respect to the second variable. Furthermore, the function
t > s 2:: to
- :s H(t, s)
.
H(t, s)
11m
t--'too H(t, to)
If p(t)
1m
Hoc
(1r(t) _1r(s))n
(1r(t) -1r(tO))n
1.
= 1, then since
--1
1
limsup
a(s) _n_(1r(t)_1r(s))(n-<>-1 l /(<>+1 l
t--'tCXl 1rn(t) to
a1/"(s)
1t
<>+1
ds
1
lim -nn<>+la- 1/a(s)(1r(t) -1r(s)t- a - 1ds
t--'too 1r (t) to
n a +1 (1r(t) - 1r(to))n-<>
lim - - 0
t--'tCXl 1rn(t)
n_a
-,
1
t
to
(n+(s))(<>+1l/a
l/a()
s
ds - 00
(3.9.38)
Chapter 3
178
and
li~~p
7r n1(t)
[7r(t) - 7r(s)tq(s)ds
n(T),
for all T
~ to
(3.9.39)
to and some n
> a,
rt (t _
JT
ao
n(T)
(3.9.40)
:s:
where v, .x, a are constants such that - 1 < .x 1, v < n, a -=1= 2 and
a 2.x ~ (a + l)(v - a). Taking H(t, s) = (t - s)2 for t ~ s ~ to, we find
<
tV (t - to)2-o
v>O
t 2 2-a
to (t - to)2-0
v<O
t 2 2-a
tV-a (
to ) 2-0
- - 1-v>O
2-a
t
'
{
{
~~
2 - a to
(1- to )2-0
t
'
< O.
it [
T
(t - s)2 SA cos S
f.
Now, set n(t) = _TA cosT - f. Then there exists an integer N such
that (2N + 1)7r - (7r/4) > tl and if n 2': N, (2n + 1)7r - (7r/4) ::; T ::;
(2n+ 1)7r+ (7r/4), n(T) ~ oTA, where 0 is a small constant. Now, since
>.
~ (a
00
to
+ 1)(1/ -
a), we obtain
(n+(s))(o+1)/o
1/0( )
179
ds > ~
5(0+1)/0 j(2n+1)7r+(7r/4)
L..n=N
(2n+l)7r-(7r/4)
1(2n+l)7r+(7r/4)
00
5(0+1)/0
n=N
(2n+l)7r-(7r/4)
sA(o+I)/o-(v/o)ds
ds
= 00.
Hence, all conditions of Corollary 3.9.5 are satisfied and therefore equation
(3.9.41) is oscillatory.
8
8tH(t,s) = hl(t,S)
(3.9.42)
and
+ 1)0+1.
180
Chapter 3
w(t) = p(t)
where p(t)
HEr,
(3.9.43)
b
Ib a(s)p(s)
I c H(b, s)p(s)q(s)ds :::; H(b, e)w(e) + c ooHa(b, s)
p'(s)
x [H(b,s) p(s)
+ h2(b,s)
]a+1
(3.9.44)
ds.
Proof. From equation (3.1.1) and (3.9.43), we have for s E [e, b),
p(s)q(s) =
,p'(s)
Iw(s)IC<>+1)/a
w (s) + p(s) w(s) - a (a(s)p(s)l/a
lbH(b, s)p(s)q(s)ds
(3.9.45)
-0
b
Iw(s)ICa+1)/a
I c H(b, c) (a(s)p(s)l/a ds
lw(s)lds
b
Iw(s)ICa+1)!a
-a I c H(b, s) (a(s)p(s)l/a ds.
Thus, from Lemma 3.2.1 with
_ (
A and
B
oH(b, s
(_0_)
0+1
a (
a!Ca+1)
Iw(s)1
(a(s)p(s)l/Ca+1)'
A-
+1
p'(s)
Iw(s)I Ca+1)/a
Iw(s)1 ( h2(b, s) + p(s) H(b, s) - aH(b, s) (a(s)p(s)l/a
<
-
a(s)p(s) (h (b s)
ooHa(b,s)
2,
+ p'(s) H(b
p(S)
S)a+1
181
lb
H(b, s)p(s)q(s)ds
H(b ) ,()
,c u c
S;
( h2(b, s)
+ 00
Jb HI>
a(s)p(s)
(b )
,s
)00+1
p'(s)
p(s) H(b, s)
ds.
1
C
1
c
a(s)p(s)
00
DoH (s, a)
p'(s)
Q+1
H(s,a)p(s)+h 1(s,a)
ds.
(3.9.46)
Proof.
Following Lemma 3.9.1, we multiply (3.9.45) by H(s, a)
integrate from a to c, to obtain
1H
(s,a)p(s)q(s)ds
=-1"
H(s,a)'lL"(s)ds+
-0
S;
Ja
1H(s,a):(~}'lL'(S)dS
c
1'lL'(s) 1(00+1)/",
H(s, a) (a(s)p(s))l/Q ds
r[
p'(s)
1
-H(c, a)'lL'(c) + Ja H(s,a) p(s) +h1(s,a)j l'lL'(s)lds
-0
r H(s, a) Iw(s)I(",+l)/oo
(a(s)p(s))l/OO ds.
Ja
A = (H(
))"'/(Q+1)
Iw(s)1
o
s,a
(a(s)p(s))l/(ct+ 1) '
and
1'lL'(s) 1( h1(s,a)
S;
p'(s))
Iw(s)I(<>+l)/<>
p(s) H(s,a) - oH(s,a) (a(8)p(s))1/<>
a(s)p(s) (
p'(s)
)<>+1
H (s, a ) h1(s,a) +-(-)
H(s,a)
p s
00
and
182
Chapter 3
1
C
H(s, a)p(s)q(s)ds
<::::
-H(c, a)w(c)
x ( hds,a)
+ -ao
1
c
a(s)p(s)
HQ(
)
S,
p'(s)
)0+1
p(s) H(s,a)
ds.
H(~,a)
<::::
1
c
H(s,a)p(s)q(s)ds+
1
c
-(-)
H c, a
+ H(b, c)
H(~,c)
lb
H(b,s)p(s)q(s)ds
a(s)p(s) (
p'(s)
)0+1
Q() H(s,a)-(-) +h1(s,a)
ds
aoH s, a
ps
lb
c
a(s)p(s) (
p'(s)
aoHCb, s) H(b, s) p(s)
+ h2(b, s)
)0+1
ds.
(3.9.47)
Proof. Divide (3.9.44) and (3.9.46) by H(b, c) and H(c, a), respectively
and add them together.
Corollary 3.9.7. Assume that p(t) E C 1([to, 00), lR+), p'(t) 2:: 0 for
t 2:: to and for some c E (a, b) and HEr,
1
H(c, a)
Jar
> H(c,a)
Ib
c
H(b, s)p(s)q(s)ds
r aoH<>(s,
a(s)p(s) (
p'(s)
)Q+1
a) H(s,a) p(s) +h1(s,a)
ds
Ja
lb
a(s)p(s) (
p'(s)
)<>+1
aoHo(b,s) H(b,s) p(s) +h2(b,s)
ds.
(3.9.48)
Then every solution of equation (3.1.1) has at least one zero in (a, b).
1
+ H(b, c)
Now, we state and prove the main interval oscillation criterion for
(3.1.1).
183
J: H(s, r)p(s)q(s)ds
lim sup
00
(3.9.49)
00,
(3.9.50)
t-+oo
and
lim sup
t-+oo
1: H(t,s)p(s)q(s)ds
Ir
a s)p(s (
W. t,s)
H(t, s)
p'(S)
p(s)
+ h2(t, s)
)<>+1
ds
H(s,a)p(s)q(s)ds > ao
In (3.9.50), we choose r
JbH(b, s)p(s)q(s)ds
C
jC HO:(s,a)
a(s)p(s) (
p'(s)
)<>+1
H(8,a) p(s) +h1(s,a)
ds.
= c.
> ao
(3.9.51 )
Then there exists a b> c such that
Jb HO:(b,
a(s)p(s) (
p'(s)
)0:+1
s) H(b, s) p(s) + h2(b, s)
ds.
C
(3.9.52)
Combining (3.9.51) and (3.9.52), we obtain (3.9.48). Now, it follows from
Theorem 3.9.6 that equation (3.1.1) is oscillatory.
184
Chapter 3
p(t)
lim sup
t-400
jt LfH(s, r)p(s)q(s) _
r
a~~(s))
Qo
(H(S, r) p'((s))
s, r
p s
+ hI (s, r)\,,'+I] ds
)
> 0
(3.9.53)
and
limsuPjt fH(t, s)p(s)q(s)r
t-400
a~~(s))
s, r
Qo
(H(t, s)p'((s))
p s
+ h2(t, S)\,,'+1] ds
)
> 0,
(3.9.54)
1
c
a(s)p(s) (
p'(s)
H(s,a)p(s)q(S) - QoH"'(s,a) H(s,a) p(S)
In (3.9.54), we choose r
+ h1(s,a) )"'+1]
ds > o.
(3.9.55)
a(s)p(s) (
p'(s)
H(b, s)p(s)q(s) - QoH"'(b, s) H(b, s) p(s)
+ h2(b, s) )"'+1]
ds >
o.
(3.9.56)
Combining (3.9.55) and (3.9.56), we obtain (3.9.48). Now, the conclusion
follows from Theorem 3.9.6.
p(t)
>
1
_1_1c
c
1
H(s,a)q(s)ds+ H(b,c}
H(c, a)
a(s)hf+1(s,a) ds
QoH"'(s, a)
lb
+ _1_1b
c
H(b,s)q(s)ds
H(b, c)
a(s)h~+I(b,s) ds
QoH"'(b, s)
,
(3.9.57)
:tH(t, s)
= h 1(t, s)H",/(a.+1)(t, s)
and :s (t, s)
185
1 a)
H(c,
>
H(s, a)q(s)ds
:0 [H(~,a) l
+ H(b,1 c)
lb
c
H(b, s)q(s)ds
a(s)hr+l(s,a)ds+
H(~,C) Ib a(S)h~+I(b'S)dSl'
(3.9.59)
J a H(s,0,)q(s)d8
or
lb
H(b, s)q(S)d8
> _1 JCa(S)hr+l(s,a)dS
aD
Hc>(s, a)
lb
> _1
ao
a(s)h~+I(b,s)d8
HO'(b, 8)
,
(3.9.60)
(3.9.61 )
or
Ib
aD
lb a(s)h~+I(b,
aD
a(s)hr+l(s, a)ds,
(3.9.62)
s)ds,
(3.9.63)
(a(t)lx'(t)IO'-IX'(t))'
+ q(t)lx(t)Ic>-lx(t)
0,
a(t) =
e-2n(0'+1)
for
for
2n < t < 2n + 1
2n + 1 < t < 2n + 2
(3.9.64)
186
Chapter 3
and
q(t) =
2n < t < 2n + 1
for
et-2n(a+2)
2n + 1 < t < 2n + 2,
for
f;
(e E _I)Q+2
a+2
q1 (eE(<>+l) _ 1) > a + 1
and condition (3.9.63) holds for each
q1
= 2n + 1 - c E (0,1) if
1':1
a+2
(e_e E1 )<>+2
(e -
(3.9.65)
> a + l'
eE1 (<>+1)
(3.9.66)
For an appropriate choice of q1, we note that both the conditions (3.9.65)
and (3.9.66) hold, and hence equation (3.9.64) is oscillatory.
For the case
denote them by
is denoted by f
f:
I:
Ib
a(s)h<>+1(b - s)ds
Ib
and
Thus, (3.9.67) implies that (3.9.59) holds for each HE fo, and therefore,
equation (3.1.1) is oscillatory.
187
(lx'(t)IO-1x'(t))'
+ q(t)lx(t)IO-1X(t)
0,
(3.9.68)
q(t) =
IoDO q(s)ds
-00.
(I)
li~~p
1t
[(s - r)nq(s) -
limsup
1t
and
t--+oo
aD
or
(II) the following inequality holds
lim sup
t--+DO
1~
t
n o + 1 (s_r)n-o-l (a(s)+a(2t-s))] ds
(s-r)n(q(s)+q(2t-s)) - __
aD
> O.
The following corollary is an immediate consequence of Theorem 3.9.10.
Corollary 3.9.12. Let a(t) = 1. Then equation (3.1.1) is oscillatory if
for any r?: to and some n > a either
(I)
t--+DO
It
r
(s - r)nq(s)ds >
ao(n - a)
Chapter 3
188
and
or
(II) the following inequality holds
f:
and
h 2 (t, s)
= na-I/"'(s)(R(t) - R(s))(n-"'-I)/(a+I).
and
li~~p Rn-"'(t)
It
r
n",+1
(R(t) - R(s)tq(s)ds > oo(n - 0)'
+c'"I Ix(t)I"'-IX(t)
tnt
0,
(3.9.69)
a- I /", (s)ds =
where 0 > 0 and C > 0 are constants. Let R(t) =
Int, t > 1. Then, we have H(t, s) = (R(t) - R(s))n = (In(t/s))n for
t 2: s > 1, where n > 0 is a constant. It is easy to check that all the
conditions of Corollary 3.9.13 are satisfied and hence equation (3.9.69) is
oscillatory for all c > 0 and 0 > O.
189
li~~p (
(l/Ct)-k
g(s)ds
where
Gk(t)
ft g(s)
(f g(r)dr)k
(f a(r)g<>+1(r)dr)
ds.
(3.9.71)
/Ct
. t
a(s)gCt+1(s)ds
t
J g( s )ds
hm
t---+oo
(3.9.72)
O.
f= g(s)ds
00.
(3.9.73)
w'(t)
t:2: T.
(3.9.74)
w(t) = w(s) - a
it
a- 1 / Ct (r)lw(r)I(<>+1)/ Ct dr
-it
q(r)dr
(3.9.75)
()
A s t
9
g(r)
r q(u)dudr
J: g(r)dr
;:.;S,,---,-,-~s---'---'---
Lemma 3.9.3. Assume that w(t) satisfies equation (3.9.74) on [T, 00)
for some T:2: to. If there exists g E n such that
liminf A g (., t)
t--+oo
> - 00,
(3.9.76)
Chapter 3
190
then
00.
Joo a-l/O(s)lw(s)l(o+l)/ods
(3.9.77)
00.
it
=
g(s)w(s)ds
= w(~)ltg(S)dS =
where t
al tg (s)
1
8
~ ~ ~
w(~)
Since 9 E
w(T) -
n,
A(~,t)
if.
I~g(s)dsA(T,t)_
If. g(s)ds
(f.q(s)ds_I';g(s)/;q(T)dTds
If. g(s)ds
iT
if.
T
q(s)ds + 0(1)
as
t~
00.
Thus, we get
w(~)
A(~,
w(T) _
t)
I~ g( s )ds A(T, t)
If. g(s)ds
-a hE a- l / O(s)lw(s)l(o+1)/Ods
Since 9 E
>.-1/0
n
<
(1)
- a
limsup
>.
(Jt g(s)ds )
+ 0(1)
as t
(l/o)-k
~ 00.
(3.9.79)
[Gk(oo) - Gk(t)],
t-400
(3.9.80)
191
Jtt, g(s)w(s)ds.
([,8 a(T)gO+l(T)dT)
I/O.
it,
it,
it,
(3.9.82)
Hence,
P'(t)
Ag(t)
+ ag(t)ly(t)l(o+l)/<>
and
o :s;
t g(s)ds
it,
1,
a(s)go+1(s)ds
)-1/0
(3.9.83)
(3.9.84)
> aAkg(t)
(I:
9(S)dS) k
(I:
a(s)g<>+1(S)dS) -1/<>.
t2 to u and
A-1/<>
-a- ~
; -
) ( t
kit! g(s)ds
) (l/<-k
[Gk(OO) - Gk(t)],
Joo
192
Chapter 3
-c'--_=---_.,.--_ _ _ _ __
I; g(s)ds
I; g(s)ds
(3.9.85)
no,
rt
lim
Hoo
fS
0+1
J g(s)J a-"Q(r)lw(r)l--;>drds
I g(s )ds
00
n+1
a-"Q(s)lw(s)l--;>ds <
00.
o<
lim
- Hoo
<
II; g(s)w(s)dsl
-'-----------:;:-t- - - - - - ' -
11m
I g(s)ds
1, t a(s)g<>+l(s)ds)
1/(<>+1) (
1, t a- 1/<>(s)lw(s)li<>+lJ/<>ds)
<>/(<>+1)
I g(s)ds
- Hoo
From Lemmas 3.9.3 and 3.9.4 the following oscillation criterion for
(3.1.1) is immediate.
Theorem 3.9.11. If there exists 9 E
lim Gk(t) =
t-HX)
00
for some
no
such that
[0,1/0:)
and
lim A g (., t) =
t~oo
00,
(3.9.86)
such that
t--+oo
00,
(3.9.87)
t--+oo
t--+oo
(3.9.88)
193
Proof. Let a and b be two numbers satisfying limt--+oo Ag(T, t) < a <
b<limt--+ooAh(T,t). Let f(t)=h(t) for T:St<tl, where tl is
such that Ah(T, tl) ~ band J; h(s)ds ~ 1. Further, let f(t) = g(t) for
tl :S t < t2, where t2 is such that Af(T, t2) :S a and J;' f(s)ds ~ 2.
This is possible because
J;2 f(s)ds
J;' g(s)ds
J;2 g(s)ds
Ag(T, t2)[1
+ o(l)J + 0(1)
as
t2 ---+
00.
Continuing in this manner, we obtain nonnegative, bounded and nonintegrable function f(t) defined on [T,oo) such that limsuPt--+oo Af(T, t) ~
b > a ~ liminft--+oo Af(T, t) and
This implies that f(t) "= no and lim Af(T, t) does not exist. By Theorem
3.9.11 equation (3.1.1) is then oscillatory.
k(t) =
it
a-1/Q(s)ds
-t
00
as
t ---+
(3.9.89)
00.
to
If
-00
lItIs
It IS
a-1/Q(u)q(u)duds
(3.9.90)
a-1/Q(u)q(u)duds :::;
00,
Lemma 3.9.5. Assume that c(s) and P(s, t) are nonnegative continuous
functions for T:S s, t < 00. If
00
for
~T
(3.9.91)
194
Chapter 3
y(t)
c(t)
+0
for
t:::: T
(3.9.92)
S Y2(t)
S
c(t)
+0
c(t)
+ 0(0 + 1)-(a+l)/oc(t) S
+ l)c(t)
(0
and Yl(t) S Y2(t) S (o+l)c(t) for t:::: T. Suppose Y1(t) S Y2(t) s ... ::;
Yn(t) ::; (0 + l)c(t), t:::: T where n is a positive integer. Then, we have
Yn(t)
c(t)
+0
c(t) +01
c(t)
00
P(s, t)IYn_1(S)I(0+1)/Ods
00
P(s,t)IYn(s)l(o+l) jo ds
+ 0(0 + 1)(Q+1)/0
00
Yn+1(t)
P(s, t)c(Q+1)jQ(s)ds
for
t::::T.
Lemma 3.9.6. Assume that c(t) and P(s, t) are nonnegative continuous
functions for T S s, t < 00. If there exists E > 0 such that
00
t=
for
t:::: T,
(3.9.93)
+0
for
t:::: T
(3.9.94)
195
Proof.
Suppose to the contrary that y(t) is a continuous function
satisfying (3.9.94) for t::>: T. Then, y(t)::>: c(t) ::>: 0 for t::>: T, which
implies y(c<+l)/Ct)::>: C(c<+l)/Ct) ::>: 0 for t::>: T. Thus, we have
y(t)
>
c(t)
>
[1
+a
c(t)
!3n+1 = 1 + a!3~n+1)/<>(a
+ 1)-(0+1)/"(1 + E),
t::>: T.
for
where !31
I, !3n <
n = 1,2,.
(3.9.95)
We claim that
lim
n->oo
!3n =
(3.9.96)
00.
Assume to the contrary that liIll n ->oo!3n = A < 00, where A is a constant.
Clearly, A::>: l. Now, it follows from (3.9.95) that
(3.9.97)
But, it is easy to verify that (3.9.97) is impossible for A::>: l. This contradiction proves (3.9.96). Then, c(t) == 0 for t::>: T, which contradicts
(3.9.93). This completes the proof.
c(t) =
and
1=
t
loo q(s)ds
a- 1 / n (s)c(n+1l/n(s)ds ~
> 0
(3.9.98)
1 ) (0+1)/0
00+1
c(t),
for
t::>: T.
Chapter 3
196
2 T,
(3.9.99)
then equation (3.1.1) is oscillatory.
Proof. Suppose to the contrary that equation (3.1.1) is nonoscillatory.
Then following the proof of Theorem 3.7.2 there exist a number T1 and
a function 11'(t) E C 1([T1,00),lR) satisfying (3.7.4) on [T],oo). Without
loss of generality let T = T 1 , then it follows from (3.9.98) that for 9 E n,
liminfH= Ag(T, t) =
q(s)ds > -00. Now, from Lemma 3.9.3, we have
a- 1/"(s)lw(s)l(a+l)/ads < 00. Thus, in view of (3.9.98) and (3.9.75),
w(t) satisfies w(t) = c(t) +0' ft= a- 1 /a(s)lw(s)l(a+1)/ads. But, by (3.9.99)
and Lemma 3.9.6, the equation y( t) = c( t) + a ft= a- 1/a (s) Iy( s) I(0+ 1)/ads
does not have a continuous solution, which is a contradiction.
f;
f;
(lx'(t)IO-]x'(t))'
(a: 1)
-(0+1)
(1
+ t)-(<>+I)lx(t)l a - 1x(t)
O.
(3.9.100)
IX) (1
00
q(T)dT) ds =
(1)
a ) (a+1)/a
= (--
(a: 1)
1=
O'-(1+2a)/a(1
+ t)-a
( 1)
(a+1)/a 00
q(s)ds = - q(s)ds.
0'+1
a
t
0'+1
t
This means that the number (1/(0' + l))(<>+I)/a in Theorems 3.9.12 and
3.9.13 is the best possible.
(a+l)/o
R(t) =
1t
to
a- 1/a(s)ds --+
00
as t --+
00
a.nd Q(t)
00
q(s)ds <
00.
(3.10.1)
197
To prove the main result of this section, we shall need the following lemma.
Lemma 3.10.1. If
00,
(3.10.2)
then
(3.10.3)
for any nonoscillatory solution x(t) of (3.1.1) and for any constant A>
o.
Joo RCt-l(s)a-1/Ct(s)y(s)ds
Joo R Ct (s)a- 1/"(s)lw(s)I(,,+1)/ads < Now, since
<
00,
for t 2: T.
and hence
00.
ln (x(t)
t a-1/Ct(s)lw(s)I(1-Ct)/Ctw(s)ds,
iT
x(T)
1
n
( ~)
x(T)
<
Ra(s)a- 1/Ct(s)lw(s)l(a+1)/ads
(iTt
1/(Ct+1)
a- 1/ Ct (s) ) Ct/("+1)
R(s) ds
,t2:T.
for
t2:T.
Therefore, we have
00.
198
Chapter 3
+ 1)(a+l)/a
such
00,
(3.10.4)
then equation (3.1.1) is oscillatory.
Let x(t) be a nonoscillatory solution of equation (3.1.1), say,
x(t) > 0 for t::;> T ::;> to. From (3.10.4) we find that (3.10.2) holds, and
this in view of Lemma 3.10.1 implies (3.10.3). Since x(t) is nonoscillatory,
it follows from Theorem 3.7.2 that (3.7.4) holds for t::;> T where w(t) is
as in (3.7.6). Hence, for t::;> T,
Proof.
which implies
t::;> T.
Joo
00,
Joo
Corollary 3.lD.1. If
exp (-2
Q(E)dE) ds < 00, then the linear
equation x//(t) + q(t)x(t) = 0, where q(t) is as in (3.1.1) is oscillatory.
Proof. Take 0:
+ 1(<>+l)/a
R(t) = t,
Q(t)
_C a ,
0:
(0:
0:
and
ds <
00.
199
lif~~f R"'(t)Q(t)
>
~a (_a_)Q+1,
0'+1
(3.10.6)
Proof.
(0'/(0'
+ 1))"'+1
Remark 3.10.1.
From Example 3.10.1 it is clear that the inequality
(3.10.6) is sharp, i.e., the constant (1/0')(0'/(0'+1))",-1 is the best possible.
Corollary 3.10.3.
such that
Hoo
1- a
R(t)
+ ~a(1-,,)/a(t)a'(t)] >
a
1,
(3.10.7)
1- a
R(t)
+ 'a(1-n)/a(t)a'(t)l > I
~
for t
Then, we have
J'xo Ix(s)ln+lds =
00.
~ T.
Chapter 3
200
Proof.
d
-lnx(t)
dt
a-1/O(t)lw(t)I(l-o)/ow(t) ~ a-1/O(t)IQ(t)I(1-O)/OQ(t),
which implies
x(t)
x(T) exp
(t a-1/a(s)IQ(s)I(1-a)/aQ(S)dS) .
Joo Ix(s)lo+lds =
00.
This completes
1
1 a-l/"'(s)(h~_l(S))("'+l)/"'ds+ho(t),
00
ho(t)
hn(t)
q(s)ds
Q(t)
00
n=1,2,.
(3.10.9)
Clearly, h1(t) ~ ho(t) and this implies that hi(t) ~ ht(t) for t ~ to.
Also, by induction
hn+l (t)
hn(t),
n = 1,2, . . .
for
to,
(3.10.10)
lim hn(t)
t ..... oo
= h(t) <
00
for
T.
(3.10.11)
w(t)
00
a-1/a(s)lw(s)l(a+l)/ods
00
q(s)ds
for
~ T ~ to.
Thus, w(t) ~ ho(t), and hence w+(t) ~ ht(t) for t ~ T. This implies
that
w(t)
> a
> a
00
1
1
00
00
a-1/"'(s)lw(s)I(O+l)/ads +
00
q(s)ds
a-1/"'(s)[w+(s)](,,+1)/"'ds + ho(t)
a-1/"'(s)[ht(s)] (<>+l)/"'ds
+ ho(t) = h1(t)
for t
~T
201
and, by induction
w(t)
hn(t),
= 0, 1,""
T.
(3.10.12)
Now, from (3.10.10) and (3.10.12), we find that the sequence {hn(t)} is
bounded above on [T,oo). Hence condition (3.10.11) holds.
such that
1 =
0' 1 a~l/a(s)[g~_l
00
go(t)
ql(s)ds
Ql(t)
00
gn(t)
(s)](a+l)/ads
+ Igo(t)l,
n = 1,2,.
(3.10.13)
gn+1(t) 2 gn(t),
n = 1,2 ..
and
t ~ to,
(3.10.14)
Rl(t) =
t a~l/a(s)ds
ltD
Theorem 3.10.4.
-+
00
as t -+
00
and Ql(t)
<
00.
(3.10.15)
T 2: to such that
n-too
<
00
for
t ~ T.
(3.10.16)
Proof. If (3.10.16) holds, then from (3.10.14) and (3.10.16) it follows that
gn(t) ::; g(t), n = 0,1, . .. for t ~ T. Applying the monotone convergence
theorem, we have
g(t)
0'
1 a~l/a(s)[g+(s)](a+l)/Qds +
00
Thus, we get
g+(t)
1 a~l/a(s)[g+(s)]("'+1)/Qds +
10' 1 a~l/Q(s)[g+(s)](a+l)/ads +
0'
= g(t)
>
Igo(t)1 2: O.
00
00
Igo(t)1
go(t)l
Chapter 3
202
Clearly, g+(t) E C([T, CXJ), m+), and now it follows from Theorem 3.10.2
that equation (3.2.15) is nonoscillatory. This completes the proof.
(i)
such that
(ii) hn(t) is defined for n = 1,2" . " but for arbitrarily large T* 2: to
there is a t* 2: T* such that limn--+= hn(t*) = 00.
Now, using Theorems 3.10.3 and 3.10.4, we shall prove the following
result.
Theorem 3.10.5.
o<
Assume that
If (3.2.15) is nonoscillatory, then equation (3.1.1) is nonoscillatory, or equivalently, if (3.1.1) is oscillatory, then equation (3.2.15) is oscillatory.
Proof. Suppose equation (3.2.15) is nonoscillatory. It follows from Theorem 3.10.3 that there exists a T 2: to such that
n--+oo
00
for
t 2: T.
(3.10.18)
hl(t)
Cl'
<
Cl'
1= a~l/a(s)[hri(s)J(O+1)/Ods
1= a~l/Q(s)[gt(s)J(Q+l)/Qds
and hence
+ Iho(t)1
+ go(t)
gl(t).
203
Also, by induction
n = 0, 1,2,
and
(3.10.19)
T.
h(t) =
lim hn(t)::;
n~CX)
lim gn(t)
n-too
= g(t) <
00
for
~ T.
Hence, by Theorem 3.10.4, equation (3.1.1) is nOlloscillatory. This completes the proof.
li~~p 7T<>-A(t)
00
A )<>+1
+1
' (3.10.20)
and
00
00,
in (3.7.26), we find
00
I7TO(S)w(s)I(<>+!)!o
a- 1!<>(S)7T A(s)lw(s)I(<>+1)!Ods
+ -7T<>(S)W(S) + _1
a
(_A
a+1
for
)0+1
~ T.
> 0,
Chapter 3
204
which implies
00
),,+1 7r -"(t).
A
Hence, we get
which
limsup
t-+oo
1t
[7r"(S)q(S) _
to
1/"(S)7r-1(S)] ds = 00,
(_a_)"+l
+
a
a-
(3.10.21)
then equation (3.1.1) is oscillatory.
Proof. Assume to the contrary that equation (3.1.1) is nonoscillatory. We
define the function w(t) as in Theorem 3.7.4. Then there exists a T ~ to
such that (3.7.26) holds for t ~ T with A = a. By Lemma 3.7.2, we have
~
0,
which implies
-00,
t-+oo
to
(3.10.22)
205
A
In"(s)w(s)l(a+1)/a + -nG(s)'w(S)
+ -1 ( -A- )
a
0+1
a+ 1
2: 0,
which implies
Then, we find
Hence, we obtain
limsup n"-"(O
[,-+1'
l'
nA(s)q(s)ds < 1 + - -
a - A
-A- )0+1
a +1
'
= sgn x.
e(t),
(3.1l.2)
Chapter 3
206
dw
a F(l, w)
If
00
q(s)ds =
to
00,
E IR.
(3.11.3)
~0
(3.11.4)
to
Ix'(t)I"-l X'(t)
x"(t)
w(t)
for
t ~ to.
w'(t) : : : - q(t)F(1,w(t
for
t ~ to.
(3.11.5)
(t)
w(to)
- v() : : : _
it
F 1, v
to
q(s)ds -+ _
00
as
t -+
00,
(3.11.6)
Remark 3.11.1. When condition (3.11.3) does not hold the argument
presented above shows that every solution of (3.11.1) is either oscillatory
or tends to zero monotonically as t -+ 00. In fact, in this case inequality
(3.11.6) yields
lim w(t) =
t-+oo
lim
t-+oo
00,
and
(_1)f3+ 1 = 1
(3.11.7)
L= (1=
207
is a constant. If
q(u)du r/<> ds
00,
(3.11.8)
Proof.
lim x'(t) =
Hoo
x(t)
o.
(3.11.9)
and consequently, from the continuity of F(x, y), given a fixed positive
constant E, E < F(l, 0), there exists a t2 ~ t1 such that
+E
(3.11.10)
x'(t)
>
or
x'(t)
xf3iot)
E)I/o<
k1/0< (
I/o< ,
it
(3.11.11)
is
(t)
v-~/O<dv
<
00,
X(t2)
t-->=
{t e(s)ds = _ 00,
iT
t-->=
{Tt e( s )ds =
i7
00,
(3.11.12)
208
Chapter 3
and
liminf
t~oo
lim sup
t~oo
1
1
t
a-1/Ot(s)
(1S e(u)du)
(1S e(u)du)
l/Ot ds
a-1/Q(s)
00,
(3.11.13)
l/Ot ds
00,
s:
a(t)lx'(t)IOt-1X'(t) - a(t1)lx'(t1)l',,-lx'(t1)
s:
it
e(s)ds.
(3.11.14)
t,
x'(t)
s:
a-1/Ot(t)
and
x(t)
s:
x(T)
+[
Theorem 3.11.4.
82 < t2 such that
a- 1/ Ot (8)
-00,
e(s)ds
<0
)l/Ot
(lS e(U)du)
l/Ot ds.
s: Sl < t1 s:
(3.11.15)
Denote by
(3.11.16)
for i
= 1,2
and (_l) Ot +l
209
+ q(t) -
e(t)
xo(t)
t::::: to.
for
(3.11.17)
By assumption, we can choose 81, tl ::::: to so that e(t) ::; 0 on the interval
SI < tl. On the interval I, w(t) in view of (3.11.15)
satisfies the differential inequality
1= [SI' hJ where
on
I.
(3.11.18)
(3.11.19)
Integrating (3.11.19) by parts and using the fact that U(Sl) = u(h) = 0,
we obtain
-(a+1)
or
>
+ l)uO(s)lu'(s)llw(s)l]
ds.
+~
((aa(s))I/(O+llu'(s)r+ 1
+ l)u<>(s)lu'(s)llw(s)1
:::::
o.
When x(t) is eventually negative, we use u E D(S2, t2) and e(t) ::::: 0
on [S2, t2J to reach a similar contradiction. This completes the proof.
Remark 3.11.2. If a
can be replaced by
Qi(U)
l~i
1,2.
Chapter 3
210
Example 3.11.1. Consider the forced half-linear equation
(3.11.20)
Here, the zeros of the forcing term sin0 are (mT)2. Let u(t) = sin 0.
For any T 2': 0 choose n sufficiently large so that (mT? 2': T, and set
81 = (mT)2 and tl = (n + 1)21T2 in (3.11.16). It is easy to verify that
(n+l)271'2 [
(n7l')2
1
]
sin4 .jS - ---;=;. cos 4 .jS d8
2y 8
168y 8
r;.
r(nH)7I'
in
[~
4
71'
~1T _
4
8n(n + 1)1T
> 0
(lx'(t)la:-lX'(t))'
+ q(t)lx(t)la:-lX(t) =
0,
>0
(3.12.1)
(3.12.2)
and
<P2(t)
= P2(t) + k
it
1<P2(tW'd8,
(3.12.3)
then
211
Proof. First, consider the case PI (t) < P2(t) for all t E [a, b]. We
shall show that cPl(t) < cP2(t) on [a, b]. Suppose to the contrary that
cPl(t*) = cP2(t*) for some t* E [a, b]. Since cP2(a) = p2(a) > Pl(a) = cPl(a),
there must be a smallest t, say, tl > a, tl :'::: t* such that cP2 (tl) = 4>1 (tl)
and <!>2(t) > cPl(t) on [a, td.
cPl(t 1 )
Pl(t 1 )
+k
< P2(t 1 ) + k
ltl
ltl
o.
G(cPl(s))ds
G(cP2(s))ds = cP2(tt},
which contradicts the fact that <!>l(td = <!>2(td. Hence, <!>l(t) < cP2(t) for
all t E [a, b]. Now, a continuity argument completes the proof.
Q(t)
J:
J: q(s)ds
:'::: 0 for t
(3.12.4)
Then, we find
q(t)
w'(t)
which implies
w(t)
Q(t)
+a
+ alw(t)l(o+l)/a, t E [c,b)
it
Iw(s)l(a+l)/ads,
(3.12.5)
[c, b)
(3.12.6)
lim w(t) =
t ..... b-
00
and
w(t) 2: 0,
(c, b).
(3.12.7)
Wl(t)
it
IWl(S)I(a+l)/ads,
t E [c,b)
it follows from Lemma 3.12.1 that w(t):,::: Wl(t) = 0 for all t E [c,b).
This contradicts the fact that limt ..... b- w(t) = 00.
Chapter 3
212
(b - e)a sup
c:::;t~b
It
. c
q(s)dsl > l.
(3.12.8)
(b - e)a sup
c~t~b
It
q(s)ds > l.
(3.12.9)
. c
Proof. Without loss of generality, we assume that x(t) > 0 for all
t E [e, b). Define w(t) as in (3.12.4), and let
cjJ(t)
it
Iw(s)l(a+l)/ads
for
t E [e, b).
(3.12.10)
for
t E [e,b),
(3.12.11)
w(t) = cjJ(t)
where Q(t) =
limHb- cjJ(t) =
I: q(s)ds.
limHb- w(t) =
00.
Now set Q*
cjJ'(t)
+ Q(t)
= sUPc~t9IQ(t)l,
so that
+ cjJ(t))(a+l)/a
for
t E [e, b)
and
a(Q*
cjJ'(t)
+ cjJ(t))(a+l)/a
::; 1 for
tE[c,b).
(3.12.12)
(Q*+~(t))l/a[:
::; b-e,
II: q(s)dsl
I:
213
Lemma 3.12.2 that Q* > O. Hence, by (3.12.11), 0::::; w(t) ::::; Q* + 4>(t).
The rest of the proof is similar to that presented above, and hence we omit
the details.
I.Itr
(T - a),' sup
a::;t::;T
q(s)dsl > l.
(3.12.14)
iT
(T - a)" sup
a::;t::;T
q(s)ds > l.
(3.12.15)
(b - c)'" sup
a::;t::;b
lib
c
q(S)dSI ::::; 1,
a::;t::;c
II
a
q(s)dSI ::::; 1,
= min{ 4, 4"'}.
(b-a)"j
[,U!2
and
J[a,b]\(Il uJ,)
q(s)ds > TJ
q(s)ds ::::; O.
(3.12.16)
(3.12.17)
x(t) on [a, b], respectively. (If there is only one zero of x'(t) in (a, b),
Chapter 3
214
then c and T coincide). Thus, x'(c) = 0, x(b) = 0 and x'(t) =f. 0 for all
t E (c, b]. By Theorem 3.12.1 inequality (3.12.9) holds. Thus, there exists
bI E (c, b] such that
(b - c)<>
b,
q(s)ds > 1
f b, q(s)ds
and
fb
(3.12.18)
:2: . c q(s)ds.
Similarly, it follows from Theorem 3.12.2 that there exists al E [a, T) such
that
11' q(s)ds
and
a,
0:
11' q(s)ds.
:2:
a,
> 1, then
(b-a)o:j
I , Uh
+ (T - a)]'"
(fbI
c
q(s)ds +
11'a, q(S)dS)
(b-a) ( j
(3.12.19)
I , UI 2
0:
~ a)" ]
1
0/2
1
]
~(b-c) 0/2 (b-C)<>/2
+ (T-a) (T-a)<>/2
- 4.
< 1, then
q(S)dS)
I/o
= (b - a)
l'
b
q(s)ds +
?: (b-a)
b
)
[ (l'q(s)ds
?: [(b-c)+(T-a)]
11'a, q(s)ds
1/<>
1/"
U~q(S)dS)
[(b~C) + (T~a)]
J:,
JI,UI 2
q(s)ds >
J:'
'T].
1/<>]
?:4.
Next,
o=
w(c) - W(T) =
J: q(s)ds
q(s)ds + 0:
J:
Iw(s)l(a+l)/ods.
h of [a,b],
215
(b - a)"
, I1UI,
q(s)ds:S
(3.12.20)
'I].
Then each solution of equation (3.12.1) has at most one zero on [a, b].
Proof. Suppose to the contrary, there exists a nontrivial solution x(t) of
equation (3.12.1) with X(7) = x(c) = 0 for a:S 7 < c :S b. Without loss
of generality, we assume that x'(t) i- 0 for t E (7, c). By Theorem 3.12.3
there exist two disjoint subintervals hand 12 on [7, c) C [a, b] such
that (C-7)a hUI, q(s)ds > '1]. Thus, we have (b-a)" fhuI, q(s)ds > '1],
which is a contradiction.
N <
[(b-'I]a)c< jq(S)dsf/ +1
(3.12.21)
(3.12.22)
and
[a,b]\!
q(s)ds < 0,
2
I
h
I = UiN-l
were
=1 Uj =1 ij.
Proof. Let ti, i = 1,2, .. , N be the zeros of x(t) in [a, b]. By Theorem
3.12.4 for each i = 1,2, ... , N - 1 there are two disjoint subintervals, say,
IiI and Ii2 of [ti, ti+l] such that
i I il UI i2
and
q(s)ds >
(3.12.23)
'I]
(ti+l - ti)"
q(s)ds :S O.
(3.12.24)
i[ti,ti+d\(Iil UIi,)
iI q(s)ds >
N-l
'I]
> 'I](N -
1
(ti+l - ti)a
1]
1
[
1)
iI)<> ...
1]
[ 1
(t2 -
1/(N-l)
(tN - tN_l)a
l/a(N-l)
>
'I]
(N
1)<>+1
(tN - tl)et'
Chapter 3
216
which implies (N - 1),,+1 < ((b - a)")/rJ II q(s)ds. This implies that
(3.12.21) holds. From (3.12.24) it is easy to deduce (3.12.22). This com
pletes the proof.
Example 3.12.1. Consider the differential equation
0,
(3.12.25)
(kTJ/4)1/" .
(i)
Since q(t)
sup
099
I (t q(S)dSI
io
sup
099
II;
It
io
sinksdsl <
k'
Thus, bSUPO:St:Sb
q(s)dsl :::; 1. Now, by Corollary 3.12.1 equation
(3.12.25) is right disfocal on [0, b). Similarly, (3.12.25) is also left disfocal
on (0, bj.
(ii) It follows from 0 < b :::; (k'17/4)1/" that for any two disjoint
subintervals hand 12 of [0, bj, II , uI2 q(s)ds = II , uI2 sinksds :::; 4/k,
which implies b" f h uI2 q(s)ds :::; TJ. Now, by Corollary 3.12.2 each solution
of equation (3.12.25) has at most one zero on [0, bj.
In what follows for any t and 0 > 0, we shall denote by (h u12 )(t, 0)
the union of two disjoint subintervals hand h of [t, t + oj.
Theorem 3.12.4. Let x(t) be an oscillatory solution of equation (3.12.1).
If
lim sup
t~oo
(0" 1
(l,UI2 )(t,8)
q( s )dS)
<
'17
(3.12.26)
for all 0 > 0 and for every two disjoint subintervals hand h of
[t, t+oj, then the distance between consecutive zeros of x(t) is unbounded
as t -+ 00.
Proof. Suppose to the contrary that equation (3.12.1) has an oscillatory
solution x(t) whose zeros {tn}~=l contain a subsequence {tnk}~l such
that 0 < t nk + 1 - tnk < 0 for some 0 > 0 and all k. Then by Theorem
3.12.3 there are two disjoint subintervals h (tnk' 0) and h(tnk' 0) of
[t nk , t nk +,] satisfying
217
bo.l
q(s)ds > TJ
k,
for all
(h U!Z)(t"k ,6)
lim
t-+oo
(IluI~)(t,.50)
q(s)ds =
(3.12.27)
for every two disjoint subintervals hand h of [t, t + bo], then the
distance between consecutive zeros of x(t) is unbounded as t --+ 00.
q(s)ds = 0.
(3.12.28)
Proof.
We will first show that for all
subintervals hand h of [t, t + bo],
lim
t-+oo
(hU!z)(t,.5)
0 >
L1
+ ioo,
i =
k-l
(IIUlz)(t,.5)
q(s)ds =
i=O
(IiI U1.,)(ti,.50 )
q(s)ds,
(3.12.29)
q( s )ds = 0,
i = 0, 1, ... , k - 1.
(3.12.30)
t-+oo
(00.1
(II UI2)(t,.5)
q(S)dS)
for all b > 0 and any two disjoint subintervals hand 12 of [t, t
The result now follows from Theorem 3.12.4.
+ 0].
It:
(3.12.31)
Chapter 3
218
limsup (<>e>
t---+oo
(I, UI2)(t,li)
q(S)dS) <
(3.12.32)
1]
[t, t+<>], then the distance between consecutive zeros of x(t) is unbounded
as t -+ 00.
lim
t---+oo
(I,Uh)(t,li o )
q(s)ds = 0
(3.12.33)
for every two disjoint subintervals hand h of [t, t + <>0], then the
distance between consecutive zeros of x(t) is unbounded as t -+ 00.
(3.13.1)
(a(t)lx'(t)lo<-lX'(t))' - q(t)lx[g(t)]lo<-lX[g(t)]
(3.13.2)
and
0,
where
(i)
(ii)
IX
Joo
a-1/0s)ds
00,
(lx'(t)lo<-lX'(t))'
q(t)lx(t)IO<-lX(t)
(3.13.3)
219
(lx'(t)1 0
1 x'(t))'
+ alx(t)la- 1 x(t) =
0,
a> 0
satisfying the initial conditions x(O) = 0, x' (0) = 1. Elbert [15] has shown
that Sa(t) exists uniquely on 1R and it is periodic of period ?T a ,
and
J:
x[g(t)] 2: k
for
a'(t):s: 0,
t 2: Tk 2: T,
(3.13.4)
or
x[g(t)] 2: k
(g~t)) x(t)
for
a'(t) 2: 0,
t 2: Tk 2: T,
(3.13.5)
t 2: T.
(3.13.6)
220
Chapter 3
(I) It suffices to consider only those t for which get) < t. Since
al/e>(t)x'(t) is decreasing, it follows from the mean-value theorem that for
each t> get) ~ T, there exists a tl E (g(t), t) such that
Proof.
x(t) - x[g(t)]
:s
x(t)
x[g(t)]
for t > g( t)
a(td
x[g(t)]
[t - get)]
(3.13.7)
T. Similarly,
x[g(t)]- x(T)
~ (a~~;:?])
1/0
x[g(t)]
~ (a~~t~?])
1/0
x[g(t)]
X'[g(t)]
x[g(t)]
(a[g(t)]) 1/0
X'[g(t)] ~
a(t2)
klg(t)
for
x(t)
x[g(t)]
(3.13.8)
t ~ T kl
~
Tkl
T,
<
x[g(t)]
a(tl)
klg(t)
a(tl)
klg(t)
'
or
(3.13.9)
221
we find
(3.13.10)
x(t)
x(t) - x(T)
(a 1 /O!(t)X'(t)).i a- 1 /<>(s)ds,
Lemma 3.13.2. Let x(t) E C 1 ([to, 00), JR+) be such that a1 / Q(t)x'(t) E
C1([to,00),JR), where a and a(t) satisfy conditions (i) and (ii) respectively, and (a(t)lx'(t)IQ-l x'(t))' ~ 0 for t ~ T ~ to. Then,
X(T) ~ al /<>(O')R[T,O']x'(O')
T ~ a ~ T.
(3.13.12)
x(O')
(12) Since
X(T)
(a(t)lz'(t)IQ-1z'(t))'
+ q(t)lz[g(t)]I!3- 1z[g(t)]
::; 0,
(3.13.13)
where conditions (i) - (iv) hold and (3 > 0 is a constant. Let z(t)
be an eventually positive solution of (3.13.13). Then for some to and
0< Xo ::; z(to) there exists a solution x(t) of
(3.13.14)
Chapter 3
222
with x(to)
= Xo
= 0, 1
t::::: tl.
for
(3.13.15)
which contradicts the fact that z(t) > 0 eventually. Thus, z'(t) > 0
eventually. Now, let Xo be such that 0 < Xo < z(T) for some T::::: to.
Then integrating (3.13.13) from t to u with u::::: t ::::: T and letting
u -+ 00, we obtain for t::::: T,
(3.13.16)
t::::: T.
(3.13.17)
t::::: T, we obtain
<I>(t,z).
<I>(s,z)ds
z(T) + \[J(t,z),
Now, define
z(t)
Xo + \[J(t, x n ),
1,2" ...
(3.13.18)
Then from (3.13.17), we get by induction that 0 < xn(t) ~ z(t) for
= 0,1,2""
and Xn+l(t) ~ xn(t) for t::::: T, n = 0,1,2,.
Consequently, letting limn-+ex> xn(t) = x(t) and applying Lebesgue's
monotone convergence theorem, we find x(t) = Xo + \[J(t, x). Now, it
follows easily that x(t) has the desired properties. It is also clear that a
corresponding result holds for the negative solution z(t) of (2.13.13).
t::::: T, n
Lemma 3.13.4. Let q(t) E C([to, (0), JR+), T(t) E Cl([to, (0), JR), T(t) ~
t and T'(t)::::: 0 for t::::: to and limHex> T(t) = 00. If
liminf
t-+ex>
q(s)ds > -,
T(t)
e
223
+ q(t)X[T(t)]
+ q(t)X[T(t)]
(a3)
+ q(t)X[T(t)] = 0
Lemma 3.13.5. Let q(t) E C([to, 00), lR+), T(t) E C1([to, 00), lR), T(t) ~
~ 0 for t ~ to. If
t and T'(t)
liminf
t-+oo
(t)
q(s)ds > -,
e
(b l ) then the inequality x'(t) - q(t)X[T(t)] ~ 0 eventually, has no eventually positive solution,
(b 2) then the inequality x'(t) - q(t)X[T(t)] ::; 0 eventually, has no eventually negative solution,
(b3)
= 0 is oscillatory.
It
y'(t)
sgn y[g(t)]
= 0,
t~T
(3.13.19)
that
x'(t) > 0
for
t ~ tl
(3.13.20)
for
t ~ t2.
(3.13.21)
Chapter 3
224
Setting u:(t)
= a(t)(x'(t))c>,
t:::: t2
t:::: t2
to
v,
and letting
v,
-+
00,
we find
Clearly, the function w(t) is strictly decreasing for t:::: t2' Hence, by
Lemma 3.13.6, there exists a positive solution y(t) of equation (3.13.19)
with limHoo y(t) = O. But this contradicts the assumption that (3.13.19)
is oscillatory. This completes the proof.
t q(s)RC>[g(s), To]ds
} get)
liminf
t-+oo
or
> -
00
when
when
t:::: T,
0'= (3,
0'< {3,
such that
li~~p
where,),
t [
to p(s)q(s)
1/(0'
+ l)(a+I),
(p'(S))c>+I ]
-')'a[g(s)] (p(s)g'(s))a ds
00,
(3.13.23)
Proof.
that for
t:::: t I ,
x'(t) > 0
and
a(t)(x'(t))a:s; a[g(t)](x'[g(t)])a.
(3.13.24)
225
p'(t)
,
( a(t)
w'(t) < -p(t)q(t)+ p(t) w(t)-ap(t)g (t)a(t) a[g(t)]
t1,
)1/'" (x[g(t)]
x'(t) )"'+1
,
or
ag'(t)(a[g(t)]p(t))-l/"'W(",+1)/", (t).
(3.13.26)
Set
'( ))"'/("'+1)
ag t
w(t)
(a[g(t)]p(t))1/("'+l) '
>.=a+1>1
a
and
+1
for
t~tl'
,
W (t) ::; - p(t)q(t)
(p'(t))"'+l
+ ,a[g(t)] (p(t)g'(t))'"
for
t ~ tl
Taking lim sup on both sides of (3.13.27) as t --+ 00, we find a contradiction to condition (3.13.23). This completes the proof.
When a
result.
226
Chapter 3
t [
to
1
(p'(S))2 ]
p(s)q(s) - -a[g(s)] ( ) '() ds
4
psg s
00,
and
lim
t ...... oo
to
r' p(s)q(s)ds
(3.13.28)
00
lto
(p'(S))Q+l
a[g(s)] (psg
( ) '( S )) ds <
(3.13.29)
00,
w(t)
It
+ It
p(s)q(s)ds
t,
-It
t,
we find
p'((s)) w(s)do5
P
05
(3.13.30)
t,
00
t,
It
p'(s)
-(-) w(05)do5 =
Ps
t,
p(s)q(s)ds::; c+
t,
p'(s)
-(-) w(s)ds-a
PS
It
t,
(3.13.31)
00.
u+1
g'(s)(a[g(s)]p(s))-;;;w~(s)ds.
(3.13.32)
227
t,
p'(S) w(s)ds =
p(s)
It
t,
((a[g(s)]p(s))-1/(a+l)(ag' (s))a/(a+1)w(s)) ds
(I,
I, (:((;;)
a/(o+l)
<
agl(s)(a[g(S)jp(s))-1/O'u/a+1)/ads)
I
<>+1
) 1/(0+1)
a[g(s)jp(s)(ag'(s))-O'ds
(3.13.33)
(p'(s)w(s)jp(s))ds --+ 00 as t --+ 00, we
Since (3.13.31) implies that
find from (3.13.29) and (3.13.33) that there exists a t2 2': tl such that for
itt,
t 2': t2,
It
t,
It
g'(s)(a[g(s)jp(s))-1/0w(o+1)/a(s)ds.
t,
(a(t)lx'(t)la- X'(t))
k
+ ta+1Ix[Atjla-1x[Atj
=
0,
t>0
(3.13.34)
where a, A, k are positive constants, 0 < A :s; 1 and aCt) E C([to, 00),
lR +). It is easy to verify the following:
t"'+l and
k >
1 (
A'"
+1
",+1
'
then all conditions of Theorem 3.13.2 are satisfied, and hence equation
(3.13.34) is oscillatory.
(A 2 ) If aCt) = 1jtf3 where {J > 0 is a constant, then for all (J, A and
k the hypotheses of Corollary 3.13.2 are satisfied with pet) = t a +!, and
therefore, equation (3.13.34) is oscillatory.
Now, for each t 2': to, we define ,et)
,et) 2': t and go ,et) = t.
Theorem 3.13.3. Let conditions (i) - (iv) hold, and a = (J, get) :s; t for
t 2': to. Then equation (3.13.1) is oscillatory if either one of the following
conditions holds
228
Chapter 3
(3.3.35)
(3.13.36)
t---+oo
(3.13.37)
Proof.
that
(3.13.38)
Integrating (3.13.1) from t 2 t2 to u and letting u --+ 00, we find
a(t)(x'(t<> 2
00
q(s)xc<[g(s)]ds.
(3.13.39)
xQ[g(t)] 2 kQ (:s:)
or
x<>[g(t) 2 k<>
(g~t) QxQ(t)
(g~t)
x<>(t)
(3.13.41)
229
0<
q(s)ds.
it
or
it
= b2 <
00.
{sn}~=1
(3.13.42)
For
b2 ; 1
(b 2
+ 1))1/'" <
~ kfR"'[sn,t2Jl~(g~S))a q(s)ds
k1
>
<
for n
~ N.
(3.13.43)
(b2~1)
C2;1)
for all sufficiently large n. This contradiction shows that condition (3.13.36)
does not hold. The other case can be considered similarly and hence omitted.
Next, by 'Y(t}
xa(t)
~ t2'
Corollary 3.13.4.
t ~ to and
a(t) == 1. If either
it
230
Chapter 3
or
limsup to.
t--+CXl
CXl
q(s)ds > 1,
let)
ha.lf~linear
differential equation
(3.13.44)
where a and c are positive constants. Let R[t,T] = J;a~l/<>(s)ds =
1 (g(
CXl
1 (1) '"
00
(~)a
2
"2
c )0.+1 ds
s(ln es
2~ac (~)a
c
_
a(lnet)<> -
Inet
p'(t) ~ 0,
k(t)
a1 /<>[g(t)]p'(t)
g'(t)
and k'(t):s;
If
JCXl
p(s)q(s)ds
(3.13.46)
00,
for
t;:::: t2
~ t1.
(3.13.47)
a1/<> (t)x'(t)
w'(t) < - p(t)q(t) + b xP[g(t)]
for
t;:::: t3.
(3.13.48)
231
t ~ T and hence
t
iT
for
t ~ T.
(al/c<[g(s)]p'(s)) (x'[g(s)]g'(S)) d
g'(s)
x!3[g(s)]
,s
( a 1/C<9[(T)]P'(T)) (f. x' [o9( s )]09' (s) ds
g'(T)
iT x!3[g(s)]
[g(f.)]
k(T)
k(T)
u-!3du = - - [x-!3+ l [g(T)]_
(3 - 1
x[g(T)]
== K <
x-!3+1[g(~)ll
(Xl,
iT
g'(s)
x!3[g(s)]
t ~ T.
for
(3.13.49)
J;
Theorem 3.13.5. Let conditions (i) - (iv) hold, (3 > a, g(t) ::; t and
~ 0 for t ~ to. If
g'(t)
00
a-l/c<[g(s)]g'(s)
(1
00
q(U)dU) l/c< ds
(3.13.50)
(Xl,
OO
a-1/C<[g(t)]g'(t)
(1
00
x~~/~1~~~~~)
for
~T ~t
l .
[g(t)]
v-!3/udv
x[g(T)]
232
Chapter 3
/t
a -1/'" [g(s)]g'(s)
(100)
s
q(u)du
Jt
s (
I/n
1 ) ds
-~
4ets 2
ds ~
/t ~ (/00
Js
= -1 In t
Set
t >0
-.-t 00 as
(3.13.51)
1/",
ds
t -+ 00
(a(t)lx'(t)I"-l X '(t)'
+ q(t)F(x[g(t)]) =
(3.13.52)
0,
(v)
IRto
(-00,0) U (0,00)
C(IR)
{j
to > 0
if to = 0,
[to, (0)
if
i= O}
and
233
for
(3.13.53)
x:f. 0,
conditions (i) - (v) hold, get) :S; t for t 2 to. If for every constant k > 0
and all large T 2 To 2 to, get) 2 To for t 2 T, the equation
(a(t)lx'(t)IQ-1x'(t)'
(3.13.54)
Proof.
Let x(t) be a nonoscillatory solution of equation (3.13.52).
Without loss of generality assume that x(t) > 0 for t 2 to > o. As in
the proofs given earlier it is easy to see that x'(t) > 0 and a(t)(x'(tQ
is nonincreasing for t 2 h 2 to. Thus, there exists a t2 2 tl and a
constant b1 > 0 such that x'(t) :S; b1a-1jct) for t 2 t 2. Integrating this
inequality from t2 to t, we conclude that there exist a constant b > 0
and a t3 2 t2 such that
for
t 2 t 2
(3.13.55)
(a(t)(x'(tc<)'
:S;
(3.13.56)
(a(t) I(x(t)
0,
(3.13.57)
Chapter 3
234
where a, (3, a, 9 and q satisfy conditions (i) - (iv), and
p(t) E C([to,oo),lRo),
(vii) T(t) E C([to, 00), lR), T'(t) > 0 and limt--+oo T(t) = 00.
(vi)
(a(t)lz'(t)la-IZ'(t)'
(3.13.58)
(a(t)lw'(t)la-lW'(t)'
+ q(t)(P[g(t)])!3lw[a(t)]I!3- l w [a(t)]
0 (3.13.59)
is oscillatory, where
and
T- I
T,
(a(t)ly'(t)I"-ly'(t)'
+ q(t)lx[g(t)]I!3- l x [g(t)]
O.
(3.13.60)
It is easy to verify that y(t) > 0 and y'(t) > 0 for t::::: tl ::::: to. Now
using the hypotheses of Theorem 3.13.7, we find
x(t)
>
y(t) - p(t)X[T(t)]
y(t) - p(t)[Y[T(t)]- p[T(t)]X[T 0 T(t)]]
y(t) - p(t)Y[T(t)] > (1- p(ty(t) for t::::: T::::: tl.
(3.13.61)
(a(t)(Y'(t")'
for
t::::: T.
235
x(t)
(3.13.62)
Using (3.13.62) in equation (3.13.60), we find
(a(t)(y'(t"')'
S 0 for
2: T*.
(3.13.63)
and
( to.l(x(t)
0,
(3.13.64)
where o!, /3, c, c*, Cl and ci are positive constants such that Cl <
1, ci > 1, ci S c* and ql (t), q2(t) E C([to, 00), lR+) are oscillatory if the
associated differential equations
=
(3.13.65)
and
=
(3.13.66)
(a(t)lx'(t)lo.-1x'(t))'
+ f(t, x[t -
0,
(3.13.67)
Chapter 3
236
where Q and a satisfy conditions (i) and (ii) respectively,
rea.l constants, and FE C([to, oo) x m?,lR).
We shall assume that there exist positive constants
function q(t) E C([to,oo),lR+) such that
for
xy
-=1=
and rr are
f3 and
0, t ~ to.
{L
and a
(3.13.68)
(3.13.69)
x,6[t - T] ~ c
for
t ~ t2.
(3.13.70)
Theorem 3.13.10.
Let condition (3.13.68) hold, f3 + {L :5.
rr ~ o. If for all large T, t ~ T + T, the delay equation
Q,
and
T ~
x[t - T]
~ T
+ T.
(3.13.72)
y'(t)
+ q(t)~[t -
+ T,
we find
for
t ~T
+ T.
237
The rest of the proof is similar to that of Theorem 3.13.9 and hence omitted.
(t2lx'(t)lx'(t))'
+ q(t)lx'[t -
Tllx[t - T] = 0,
t >
(3.13.73)
y'(t)
+ In(t t-
T) q(t)y[t - T] = 0,
t> 1 + T
(3.13.74)
t-T
(In(S-T)) q(s)ds
.
S -
1
e
> -
(3.13.75)
: lR
(3.13.77)
(3.13.78)
Chapter 3
238
Proof.
Let x(t) be a nonoscillatory solution of equation (3.13.76).
Without loss of generality, we can suppose that x(t) > 0 for t 2: to > O.
From equation (3.13.76), it follows that
Now, we shall present criteria for the existence of nonoscillatory solutions of equation (3.13.76) of the types (I), (II) and (III).
Theorem 3.13.11. Suppose for each fixed k -=I 0 and T 2: to 2: 0,
w,T(a;g(t)
k>O Wk,T(a;g(t))
lim
40,
(3.13.79)
uniformly on any interval of the form [T* , 00) for some T* > T. Then
equation (3.13.76) has a nonoscillatory solution of the type (I), if and only
if,
roo W(s,cwk(a;g(s))lds
ito
<
00
(3.13.80)
=f. 0
and
239
> 0,
Proof (The 'only if' part). Let x(t) be a nonoscillatory solution of the
type (I) of equation (3.13.76). Without loss of generality, we assume that
x(t) > 0 eventually. There exist positive constants CI, kl and tl such
that x[g(t)] ~ cI'llkJ(a;g(t)) for t ~ t l . An integration of(3.13.76) yields
ItOO F(s,x[g(s)])ds < 00, which when combined with the earlier inequality
leads to
F(s, cl'llk J (a;g(s)))ds < 00.
It":
(The 'if' part). Suppose (3.13.80) holds for some constants c > 0 and
k > O. Because of (3.13.79) we can choose constants C> 0 and T ~ to > 0
so that C < k/2, T* = inft>T g(t) and
{x
X =
C([T*, (0))
of
x(t)
C.
(3.13.81)
W: X --+
x(t)
'lIu,T(a; t),
0 for T* ~ t ~ T}
~
(3.13.82)
and
Wx(t)
0,
T*
T.
(3.13.83)
(i)
o~
lS
F(u,x[g(u)])du
loo
X, then since
'lj;-l (afs)) ds
Wx(t) <
'lj;-l
(a~!)) ds,
~T
~T
implying that W x E X.
(ii) W is continuous. Let {xn}~=l be a sequence in X converging to
x E X as n --+ 00 in the topology of C([T, (0)). Fix m E {l, 2, .. }
with m ~ T. The Lebesgue dominated convergence theorem implies that
F(S,Xn[g(s)])ds --+
F(s,x[g(s)])ds as n --+ 00, and so Wxn(t)--+
Wx(t) uniformly on [T,m]. Thus, WXn --+ Wx in C([T, (0)).
I;
I;
Chapter 3
240
:<:::
(Wx)'(t) =
<
t"2T
00
F(S,X[9(S)])dsJ
for
t"2
tl'
t "2
tl
Cl
for
(The 'if' part). Suppose (3.13.84) holds. Let C> 0 be fixed arbitrarily
and take T"2 to > 0 so large that T* = inft2:T g(t), and
241
and define
{x
and
Vx(t) = {
-1
t:2': T*}
00
00
into a compact
which gives the
(3.13.85)
for every nonzero constant b such that kb
> O.
Proof. It suffices to consider the case where k > 0 and b > O. Let
d > 0 be an arbitrary fixed constant and choose t > 0 small enough and
T :2': to > 0 large enough so that d + W(ajt) ::::; cWk(ajt) for t:2': T,
and
F(s,d+w(ajs))ds::::; t. This is possible because of (3.13.79) and
the fact that limt ..... oo Wk(ajt) = 00. Now, consider the set Y and the
mapping U defined by
J;
and
Ux(t)
d+
It ~-l
rats)
(t - lS
F(u, X(U))dU) ] ds
for
t:2': T.
Chapter 3
242
From (3.13.85), we have
x(t)
2:
1>r
+
d+
> d
[ats)
'ljJ-I [ats)
l
l
co
co
F(u, X(U))dU] ds
F(u,d)dU] ds
00
0,
as
00.
(3.13.86)
where a, (3, A are positive constants, 0 < A :s: 1 and q(t) E C(JR+, JR+).
Equation (3.13.86) is a special case of (3.13.76) in which a(t) = 1, 'ljJ(x) =
x"', g(t) = At, F(t,x) = q(t)x i3 , 'ljJ-l(x) = xli"', and 'l1 k ,T(a;t) =
kl/i3[t-T], t 2: T. Clearly, conditions (3.13.77) and (3.13.79) are satisfied
for (3.13.86).
The possible types of asymptotic behavior at infinity of nonoscillatory
solutions of (3.13.86) are as follows:
(I) limt ..... co x(t)/t = constant -=I 0,
(II) limt-+co x(t)/t = 0, limt-+oo Ix(t)1 =
(III) limt-+oo x(t) = constant -=I O.
00,
00
si3q(s)ds <
(3.13.87)
00
r/'"
and that (3.13.86) has a solution of the type (III) if and only if
100 (100
q(u)du
ds <
00.
II",
ds =
00
(3.13.88)
243
(a(t)lx'(t)I,,-lX'(t))'
+ F(t, x[g(t)]) =
0,
a> 0
(3.13.89)
= sgn x for
t 2: to.
Suppose for positive constants , and L with < L there exist positive
constants m and M depending on and L such that for :S Ixl :S L,
t2:to.
/00 a-l/,,(s) (1
00
(3.13.90)
x(t)
such that
-I- O.
(3.13.91)
Proof (The 'only if' part). Let x(t) be a nonoscillatory solution of
equation (3.13.89) such that limHoo x(t) = constant -I- O. Without loss of
generality, we assume that limt-too x(t) > 0, so that there exist positive
constants , Land tl 2: to such that :S x(t) :S Land :S x[g(t)] :S L
for t 2: tl. By (3.13.90), we have
(3.13.92)
t 2: tb
it
F(u, x[g(u)])du
which gives
a-l/,,(s)
1/0:
< x'(s),
Chapter 3
244
Thus, it follows that
1:
a-I/<>(s)
(100
t 2 tl
(The 'if' part). Suppose condition (3.13.91) holds for some constant c =f. O.
We can assume that c> O. By (3.13.90) there exists a constant M such
that c/2::::: x(t)::::: c implies F(t,x[g(t)])::::: MF(t,c) for t 2 t1 2 to
Choose T 2 t1 so large that T* = inft::::T get) and
M1/<>
00 a-1/<>(s)
(100
F(u, c)du
r/
<> ds <
and define the set Xc C([T*, (0)) and the mapping V: X --+ C([T*, (0))
by X = {x E C([T, (0)) : c/2 ::::: x(t) ::::: c, t 2 T*} and
Vx(t) =
-1
00
Vx(T)
a- 1/ <>(s)
for
(100
for
t2T
T*::::: t ::::: T.
The rest of the proof is similar to that of Theorem 3.13.12 and hence omitted.
Theorem 3.13.15. Let conditions (i), (iii) and (ivy hold, and suppose
that for positive constants f and L there exist positive constants m
and M (depending on f and L respectively) such that Izl::::: L implies
Izl::::: f
:::
2T
(3.13.93a)
2T
(3.13.93b)
implies
mF(t, fR[g(t) , hD
:::::
f;
00
00
c =f. O.
(3.13.94)
245
Proof (The 'only if' part). Suppose equation (3.13.89) has a solution
x(t) such that limt~oo x(t)/ R(t) = constant i= O. We can assume that
limt~oo x(t)/R(t) > O. Then there exist positive constants i, tl and
T (T > t l ) such that x(t) ~ iR[t, tlJ and x[g(t)J ~ iR[g(t), tlJ for
t ~ T. The rest of the proof is similar to that of Theorems 3.13.11 and
3.13.13 and hence omitted.
(The 'if' part). Let condition (3.13.94) hold with c = 2k which can be assumed to be positive. By (3.13.93) there exists a constant !vI > 0 such that
o ~ z[g(t)J ~ 2k implies F(t, z[g(t)JR[g(t), tID ~ MF(t, 2kR[g(t) , tID,
get) ~ tl ~ to. Let T ~ tl be large enough and T* = inft>T get) ~ II
so that M
F(s,2kR[g(s),td)ds ~ (2'" -l)k"'. Consider the set Xc
C ([T* , ()()) and the mapping W: X --+ C ([T* ,
defined by
J;
(0
X = {x EC([T*,oo,
and
and
Wx(t)
a-l/",(s) [(2k)Q _
for
T*
for
kR[t,TJ~x(t)~2kR[t,tIJ
~ t ~
x(t) = 0 for
T*
t~T
T}
F(u,X[g(U)])dUf/ ds for t
~T
T.
The rest of the proof is similar to that of Theorem 3.13.11 and hence omitted.
(3.13.95)
0,
(f)m F(t,i)
F(t,x)
~ (~)m F(t,L)
for
m>n.
It can easily be verified that (3.13.91) holds if and only if J1, > O! + II + 1
and (3.13.94) holds if and only if J1, + m > II + n + 1. Therefore, in view
of Theorems 3.13.14 and 1.13.15 necessary and sufficient conditions for
(3.13.95) to have nonoscillatory solutions x(t) satisfying limt~oo x(t) =
constant i= 0 and limHoo x( t) / R( t) = constant i= 0 are respectively,
J1, > O! + II + 1 and J1, + m > II + n + 1.
246
Chapter 3
:s
(3.13.96)
Proof. We can assume that x(t) > 0 for t 2:' to. In fact, a similar
argument holds if x(t) < 0 for t 2:' to. Since (a(t)lx'(t)IQ-Ix'(t))' =
-F(t, x[g(t)]) < 0 for t 2:' to, a(t)lx'(tlIQ-Ix'(t) is decreasing on [to, (0),
so that x' (t) is eventually of constant sign, i.e., either x' (t) > 0 for t 2:' to,
or there is some tl 2:' to such that x' (t) < 0 for t 2:' tl.
Suppose first that x'(t) > 0 for t 2:' to. Then, we have a(t)(x'(t))Q :S
a(to)(x'(to)a, t 2:' to which implies that x'(t):S al/a(to)x'(to)a-l/Q(t),
t 2:' to. Integrating this inequality from to to t, we find
x(t)
lto
t 2:' to.
Suppose next that x'(t) < 0 for t 2:' tl. Then, since a(t)lx'(t)la-lx'(t)
= -a(t)(-x'(t)G, we have a(s)(-x'(sa 2:' a(t)(-x'(t)"', s 2:' t 2:' t l ,
or equivalently,
(3.13.97)
Dividing (3.13.97) by al/a(s) and integrating over [t, T],
--t
00
iT
we obtain
a-1/CX(s)ds,
T 2:' t
gives
2:' tl
(3.13.98)
247
From Lemma 3.13.9 and its proof we can conclude that one and only
one of the following three possibilities occurs for the asymptotic behavior
of any nonoscillatory solution x(t) of (3.13.89).
= constant> 0,
(II) limt-+= x(t) = 0, liml.-->= l:r(t)I/7r(t) =
(III) limt-+= Ix(t)I/7r(t) = constant > o.
liml._,= Ix(t)1
(I)
00,
= (
1.0
a(s)
1
s
to
IF(u, c)ldu
) l/a
ds <
00
(3.13.99)
# O.
1
t
to
which implies that Ito;;' F(s,x[g(s)])ds < 00. Combining this inequality
with the fact that x[g(t)];::: Cl, t ;::: tl ;::: to for some positive constant
Cl, we obtain Ito;;' F(s, cl)ds < 00. Hence, it follows that
1=(118
to
a(s)
1.0
F(u, cl)du
)1/0 ds:S;
7r(to)
I.,
tl.
<
00.
Again an integration of
248
Chapter 3
or
1
( a(t)
F( s, X[g( s )])ds
) 1/<>
:S - x' (t)
Integrating the above inequality over [tl' t2J and noting that C1 :S x[g( t) J :S
C2: t 2- iI for some positive constants C1 and C2, we find that
for any
t2
(=
iT
(3.13.100)
and define the set X C C ([T* , (0)) and the mapping V : X ---+ C ([T* , (0))
by X = {x E C([T*, (0)) : c/2 :S :r(t) :S c, t 2- T*} and
c-l
Vx(t) = {
(a/s)
Vx(T),
lS
t 2- T
T*:S t:S T.
x(t)
c-
iTt
l/a
ds,
t 2- T.
Theorem 3.13.17.
A necessary and sufficient condition for equation
(3.13.89) to have a nonoscillatory solution of the type (III) is that
00
00
(3.13.101 )
to
o.
249
Itt,
It
It
F(.s, Cl7r[g(s)])d.s
F(s, x [g(s)])d.s
~ (;~~D
x[g(t)] ~ C27r[g(t)]
h. From equation
t ~ tl. Combining
for t ~ to, which
<
c~
for t
~ t2
00.
(ii). (Sufficiency).
We only need to consider the case when the constant C in (3.13.101) is positive. Let k > 0 be a constant such that
2k < cQ and take T > to so large that T* = inft2:T get) ~ to, and
F(s, (2k)1/Q 7r [g(.s)])ds ~ k. Consider the set
I;
Z = {z E C([T*, (0)) :
Wz(t) = { lex)
Wz(T),
If z E Z,
then for t
kl/o lex)a-l/Q(.s)d.s
T*
~ t ~
t ~T
T.
(3.13.102)
T, we have
~ Wz(t)
(t
\
k + iT F(u, (2k/ IQ 7r[g(u)]du)
it(ex)a-l/Q(s)
l/Q
ds
= (2k)1/Q 7r (t),
From this integral equation it follows that z(t) satisfies equation (3.13.89)
on [T, (0), and by L'Hospital's rule limHex) z(t)/7r(t) = constant E
Chapter 3
250
[kI/e> , (2k)I/e>]. Thus, z(t) is a solution of the type (III) of (3.13.89). This
completes the proof.
Theorem 3.13.18. The equation (3.13.89) has a solution of the type (II)
if
8
)1/e>
(3.13.103)
100(11
a(s)
to
to
ds <
IF(u, c)ldu
00
100
IF(s,d7r(s))lds
(3.13.104)
00
to
1 r
iTroo ( a(s)
iT F(u, c)du) 1/e> ds:::;
k 1/e>.
Then it can be verified by the Schauder fixed point theorem that the mapping W defined by (3.13.102) has a fixed element in the set
Y
= {y
E C([T*,
00)) :
k 1/ e>7r(t):::; y(t):::; c, t ~ T
k 1/ e>7r(T) :::; yet) :::; c, T*:::; t :::; T}.
In fact, if y E Y, then
Wy(t)
<
100
a-1/n(s) [(2k)1/e>
100
(2 s F(U,Y[9(U)])dU) lie>] ds
(a(s)
t~T
'et)
lim '1L-() =
7r ' t
lim
t ...... oo
t ...... oo
>
lim (k
t ...... oo
lim
t ...... oo
iTrt F(s, k
(jt
)1/0:
k+
F(s,y[g(s)])ds
T
1/0:7r(S)) ds
00.
251
(3.13.105)
where a, /3, A are positive constants and J.l is any constant. Applying
Theorems 3.13.16 and 3.13.17 to (3.13.105), respectively, we find
(i) equation (3.13.105) possesses a nonoscillatory solution of the type (I)
(which is bounded from below and above by positive or negative constants)
if and only if J.l < A, and
(ii) equation (3.13.105) possesses a nonoscillatory solution of the type (III)
(which behaves like a constant and multiple of e-(>./a)t as t -+ 00) if and
only if J.lIA < /3/a.
Theorem 3.13.18 applies only to the case where a > /3. In fact, in this
case the condition /31 a :':::: J.lI A < 1 ensures the existence of a nonoscillatory
solution x(t) of equation (3.13.105) satisfying limHoo x(t) = 0 and
limHoo e(>.}<tlx(t) I = 00.
(3.13.106)
has a positive solution on [tl' 00) for some tl ?: to, so does the equation
(3.13.107)
Chapter 3
252
y(T)
+ J; F(u., y[g(u)])du.
zl(T)
zn(t)
Let
y(t) for t 2 t l ,
y(t) for t E [tl,T]
ht
y(T)
and
be as in Lemma
3.13.10 and suppose g(t) E CI([to, 00), JR.) and g'(t) > 0 for t 2 to 20,
a(t) E C([to, oo),JR.+) and Joo a-l/O(s)ds = 00, and a> 0 is a constant.
If (3.13.89) has a nonoscillatory solution, then the equation
(a(s)ly'(s)IO-Iy'(s))'
+ g'[g2 1 (S)]F(9-I(s),y(s)) =
0,
('
= :s)
(3.13.109)
a(T)(x'(T))o. - a(s)(x'(s))O
iT
F(u, x [g(u)J)du = 0
x'(s) 2 a-l/o(s) (
9 (g
(v))
s 2 tl'
x'(s) 2 a-1/a(s) (
Js
g'(g-l(v))
1/0
Tx(s),
s 2 iI
253
(lx'(t)I"-lX'(t))' + q(t)lx[g(t)]I"-lx[g(t)]
0,
(3.13.110)
where q(t) E C([t o, (0), lR+), get) E C([to, (0), lR), get) :S t, g'(t) > 0 for
t 2': to 2': 0, limt-+oo g( t) = 00, and a > 0 is a constant, is oscillatory if
the ordinary half-linear equation
(ly'(s)IO-ly'(s))' +
(:}~~ll~~]]) ly(s)IO-ly(s)
= 0,
(' =
:s)
(3.13.111)
is oscillatory. Further, applying Corollary 3.13.3, we find that if
oo
,\ q[g-l(S)]
s g'[g_l(s)]ds =
00,
(3.13.112)
0 <). < a
JOO g'\(t)q(t)dt
(3.13.113)
00.
x(t)x'(t) < 0
or
2': T
(3.13.114)
(C l
limsup
q(s)R"'[g(t),g(s)]ds> 1,
t--+oo } g(t)
(3.13.115)
Chapter 3
254
(C 2 )
lim sup
t-->oo
(C 3 )
it
a-l/",(s)
g( t)
(it
(3.13.116)
y'(t)
(3.13.117)
(3.13.118)
is oscillatory, or
(C 4 )
liminf
t-->oo
(t
i(t+g(t/2
t:2: s:2: T.
= - (a(s)(-x'(s))"')'
it
q(s)R<>[g(t),g(s)]ds, t:2: T
get)
and hence
a[g(t)](-x'[g(t)])O
[it
q(s)RO[g(t),g(s)]ds
get)
-1] : :
t:2: T.
for
[CY, t] gives
a(t)( -x'(t))O +
a(cy)( -x'(cy))"
>
it
it
q(U)x'" [g(u)]du
~ T,
which implies
t:2: cy:2: T.
(3.13.119)
255
x(s) = x(t)
we find
.r
x(s) 2:
Putting s
we get
x[g(t)]
it
a-l/"'(u)
(-x'(u))du
for
t 2: s 2: T,
(1 q(U)x'" [9(U)]dU)
t
1/",
(3.13.120)
t 2: s 2: T.
du,
)
l(t) a-1/",(u) (1t
q(u)d'U
t
l/n
1<
du - 1
for
t 2: T,
x[g(t)] 2: a-lin
w(t)
+ q(t)R'"
[g(t), t
+ ~(t)]
CXJ
11
t 2: T.
a(t)( -x'(t))'"
[t +~(t)] ::;
w(t) 2:
for u and
t 2: T, we obtain
w'(t)
+ g(t))/2
for
0 for t 2: to.
+~(s)]
ds,
00,
we
t 2: T.
(e-tlx'(t)lx'(t))'
Ix[t/2]lx[t/2]
(3.13.121)
256
Chapter 3
(AI)
(A 2 )
(A3)
i 9 q(s)R"[g(s),g(t)]ds> 1,
i 9(t) (a(s)1 i q(u)du 1/" ds> 1, or
lim sup
i(t+9(t))/2 q(s)R" [g(s),
s + g(s)]
1
liminf
ds> -.
lim sup
(t)
t ..... oo
t ..... oo
t ..... oo
(3.13.122)
(3.13.123)
(3.13.124)
[i
9 (t)
1<
q(s)R"[g(s),g(t)]ds - 1
0,
x'({}") 2 a- 1 /"({}")
with the relation xes)
xes) 2
(i
t
er
q(u)x"[g(u)]du
)1/" ,
= x(t) + J/ x'({}")d{}", s 2 t 2 T
1
a({}")
er
we obtain
s 2 t 2 T.
x[g(t)]
[itret)
(1 itr
a({}")
q(u)du
0 for
2 T,
257
x[g(t)]
aI/a [t
= g(t) and
= (t + g(t))/2 in
y'(t)
q(t)RO [g(t), t
= a(t)(x'(t))O, t
~ T.
~
T we
+:(t)] y [t +:(t)] , t ~ T.
But, in view of Lemma 3.13.5 and condition (3.13.124), the above inequality has no eventually positive solutions, which is a contradiction. This
completes the proof.
(e-tlx'(t)lx'(t))'
3et lx[2t]lx[2t],
t~0
(3.13.125)
o.
(3.3.126)
Remark 3.13.5. Theorems 3.13.20 and 3.13.21 fail to apply to equations
of the type (3.13.2) and/or inequalities of the form (3.13.126) when g(t) =
t. Also, oscillation of all solutions of (3.13.2) does not follow from Theorem
3.13.20 or 3.13.21, because the deviating argument under consideration is
only either retarded or advanced. However, for equations of mixed type, i.e.,
involving both retarded and advanced arguments, it is possible to establish
oscillation of all solutions. We shall show this for the mixed type equation
n
L qi(t) IX[gi(t)]10-
X[gi( t)],
(3.13.127)
i=l
where
(i)
(ii)
>0
is a constant,
qi(t) E C([to, 00), JR.+) such that
to, i = 1,2, ... ,n,
a
Cha.pter 3
258
(iii) gi(t) E CI([to, 00), JR), g~(t) 2: 0 for t 2: to and
00, i = 1,2"", n,
(iv)
a(t)
limt-+oo gi(t) =
(3.13.129)
(3.13.130)
(Ix' (t)
1 x'
(t))'
where a, k, C, u and
259
(1)
from conditions (C I ) or (C 2 ) of (M I ) of Theorem 3.13.22 that all
bounded solutions of (3.13.129) are oscillatory if
a+1
a
> --,
(3.13.132)
(2)
from conditions (Ad or (A2) of (M2) of Theorem 3.13.21 that all
unbounded solutions of (3.13.130) are oscillatory if
70+1
> a + 1 or
j71/"T(o+I)/0
> a + 1.
a
'
(3.13.133)
(3)
from the last statement of Theorem 3.13.22 tha.t all solutions of
(3.13.131) are oscillatory provided both (3.13.132) and (3.13.133) hold.
x(t)x'(t) > 0
for
t:::., tl,
(3.13.134)
or
x(t)
-1= ~q,(
= 1,2"", n.
Similarly, if x(t) is an eventually positive solution of (3.13.127) satisfying (3.13.135), then integration of (3.13.127) twice from t to v and the
letting v 4 00 gives
x(t)
(3.13.137)
Chapter 3
260
< 00,
i = 1,2,, n.
(3.13.138)
./00
i=l
c, i
implies
(The 'if' part). Suppose (3.13.138) holds. Let k > 0 be arbitrarily fixed,
and let T > to so large that
(3.13.139)
and
x(t)
0, T*:S t :S T}
and
Wx(t)
h
t
a- 1 /"(s) k"
-1
00
t;qi(U)X"[9i(U)]dU
]1/0.
ds, t 2:: T
0, T*:S t :S T.
It is clear that X is a closed convex subset of the Fnchet space C ([T, 00))
of continuous functions on [T, 00) with the usual metric topology, and
261
W is well-defined and continuous on X. It can be shown without difficulty that W maps X into itself and W(X) is relatively compact in
C( [T*, (0)). Therefore, by the Schauder-Tychonov fixed point theorem,
W has a fixed element x in X which satisfies
x(t)
h
t
a- 1/ c>(s) kQ
-.i 8
00
q;(u)xc>[g;(u)]du
ll/Q
ds,
t '2 T.
J [1
DO
a(s)
00
q;(u)du ]
1/0
ds <
00,
i=1,2,,n.
(3.13.140)
Proof. The 'only if' part follows from (3.13.137). To prove the 'if' part,
snppose (3.13.140) is satisfied. Choose T> to so that (3.13.39) holds and
]lie> ds::S:
iT(exo [a-1/e>(s) 100
s
q;(u)du
and define X
X
c C ([T* , 00 ))
= {x
2"
and V: X -+ C ([T* , 00 ) ) by
E C([T*,
Vx(t)
kl
1=
Vx(T),
a-'I"(.,)
t ';> T
T*::S: t ::s: T.
It remains to discuss the existence of an unbounded nonoscillatory solution x(t) of (3.13.127) which has the property that limHexo x(t)/ R(t) =
00,
and of bounded solution x(t) of (3.13.127) having the property that
Chapter 3
262
limt-+oo x(t) = O. However, this is a difficult problem and there is no general criteria available for the existence of such solutions. Therefore, we
confine ourself to the case where at least one of the gi(t) is retarded and
present sufficient conditions which guarantee that (3.13.127) has a nonoscillatory solution which tends to zero as t -+ 00. Such a solution is often
referred to as a decaying nonoscillatory solution. Our derivation is based
on the following theorem.
00
a- 1 /<>(s)ds
00,
yet) ~
1
t
00
a- 1 /<>(s)
t ~T
(3.13.142)
where T is chosen so that inft>T gi(t) ~ to, i = 1,2, ... ,n. Then equation
(3.13.127) has a nonoscillatory -solution which tends to zero as t -+ 00.
Proof. Let Z = {z E C([T,oo: 0 < z S; yet), t ~ T}. With each
E C([to, 00 defined by
z E Z we associate the function
z(t)
= {
z(t)
z(T)
for
>T
+ [yet) -
yeT)]
for
(3.13.143)
to S; t S; T.
as follows
t ~ T.
~T.
(3.13.144)
Twice differentiation of (3.13.144) gives
n
(-a(t)(-z'(t<'
Lqi(t)(Z[gi(t)j)",
i=l
t ~T
263
[to, T*), and hence L:'=l qi (T*)(z[g; (T)J)<' > 0, since to :; .% (T*) < T*
for some io E {I, 2"", n}. Hence, (-a(t)( -z'(t))c'lt=To f O. On
the other hand, z is identically zero on [T* ,(0), which means that
(-a(t)( -z'(t))Q)'lt=T* = 0, a contradiction. Therefore, the solution z(t)
is positive. This completes the proof.
To apply Theorem 3.13.25 in practice, we need to distinguish the following three cases:
JOO
q;(s)ds <
00
<
and
i=l
JOO
tlJ
;=1
and
00,
(3.13.145)
i (8)ds
<
00
and
oc
(3.13.146)
J=L
n
q;(s)ds =
00.
(3.13.147)
;=1
(3.13.148)
Chapter 3
264
Now, we shall present results which guarantee the existence of decaying nonoscillation solutions and are applicable for the cases (3.13.146) and
(3.13.147).
Theorem 3.13.27.
n}, and
lim sup
t-+oo
00
[1 t
aCt)
~,
00
a-I/,,(s)
i=I
(3.13.149)
Q(t)
a~t)
1 ~Qi(S)dS
00
1/"
QT = sup
t2::T
Define y(t)
= exp ( -(l/QT)
y[gi(t)]
exp (Q1
It
Q(s)ds <
aCt)
It: Q(s)dS).
t
ThiW
< e exp ( - dT
(3.13.150)
Since for i
Q(S)dS) exp
1:
= 1,2,, n,
(_QlThort Q(S)dS)
Q(S)dS) = ey(t),
t 2:: T
265
r( t t,
at,)
< e
1=
q, (<<)(ylq, (u)J)"
Q(s)y(s)ds :.:; e
< eQTexp (- dT
1:
00
Q(s) exp ( - dT
1:
Q(u)dU) ds
Q(S)dS)
t:::: T.
Now, from Theorem 3.13.25 it follows that equation (3.13.127) has a decaying nonoscillatory solution.
Theorem 3.13.28. Suppose (3.13.141) holds for someio E {I, 2"", n}.
Further, suppose there exists aT> to such that inft2'T a(t) :::: to,
r=
O'(t)
(Q* )<>/(0+1)
+1
--.I..
(3.13.151)
where the functions Q(t) and a(t) are as in Theorem 3.13.27. Then
equation (3.13.127) has a nonoscillatory solution which tends to zero as
t ---+
00.
Proof. Let
PT = sup
t2'T
it
and
y(t) =
exp (-
(~)
aPT
1 ~qi(S)(Y[9i(S)WdS
1
00
e<>+l
00
<
a
1t
tq;(S)dS) .
t Oi =l
+ l)ja) y(t),
1= (~q;(S))
+ 1
---p:-
:::; e<>+l
Ln q;(s) exp
i=l
PT e<>+l exp (- a p+
+1
q;(t)ds
<T(t) ;=1
11t
to ;=1
(y(s))<>ds
qi(u)du ) ds
to
q;(S)dS) '
t:::: T.
266
Chapter 3
r ( rt,
Consequently, we obtain
at,)
_P_
T_ ) l/n
~ (
+1
QT
~
~
e(a+1)/<>
__
T_
a 1
l/n
e(o+l)/a
for
exp
100
d,
_a__1
aPT
qi(S) exp
1/0. e(a+l)/a
L qi(u)du
n
ds
(+ 118 L
+ I1t t
to i=l
aPT
(PT )
(a+l)QT a+l
yet)
100
t
1(P,+
~ -*
r
(+ 18
__
a__
to i=l
q;(S)dS)
exp (_ a
aPT
qi(u)du ) ds
aPT
tO i =l
T.
>0
(lx'(t)lx'(t))' = c'Ylx[Atllx[At],
,>
(3.13.152)
t> 0
where
1, 0 < A < 1 are constants. This is a special case of (3.13.127)
with a = 2, aCt) = 1, n = 1, ql(t) = r'Y and gl(t) = At.
(11) Let, > 3. Then both (3.13.138) and (3.13.140) hold for (3.13.152)
and so by Theorems 3.13.23 and 3.13.24, equation (3.13.152) has nonoscillatory solutions X1(t) and X2(t) such that limHoo Xl(t)/t = constant of. 0
and limt-+oo X2(t) = constant of. 0 regardless of the values of A, 0 < A < 1.
(12) Let, = 3. An easy computation shows that (3.13.149) is satisfied
for (3.13.152) if 1 < A < exp(vI2/e), since
1
t
g,(t)
it (1
At
00
u- 3 du )
1/2
ds
>:.
1In In 1
v2
From Theorem 3.13.27 it follows that for each A equation (3.13.152) has
a nonoscillatory solution which tends to zero as t ---+ 00.
(13) Let 1 < , < 3. Then, (3.13.151) is satisfied for (3.13.152) since
QT = 1, and
ft
19l(t)
ql(s)ds
r s-'Yds
lAt
Therefore, by Theorem 3.13.28 there exists a decaying nonoscillatory solution of equation (3.13.152).
267
(~) ",/(",+1)
or (J <
-,-(O'_+_l:...~l_/~(a~+~l_)
(a: 1)
",/(",+1) k- 1/(a+1).
(3.13.153)
It would be of interest to compare (3.13.153) with (3.13.133), which can be
rewritten as
2. The results of Section 3.3 are extracted from Agarwal et. a1. [8] and
generalize a well-known theorem of Levin [38].
3. Theorem 3.4.1 is due to Hartman [25], also see Swanson [58]. All
other results in Section 3.4 are borrowed from Li and Yeh [45]. We note
that Theorem 3.4.5 is an extension of theorem 1 of Singh [57J.
4. Theorem 3.5.1 is taken from Duzurnak and Millgarilli [14], whereas
Lemmas 3.5.1 and 3.5.2 and Theorem 3.5.2 are adapted from Wong and
Yell [62J.
5. Section 3.6 contains the work of Chantladze et. a1. [12].
268
Chapter 3
Subsection 3.9.2 generalize the work of Kong [29]. The results in Subsection 3.9.3 are taken from Li and Yeh [42,43]. These results extend and
improve the work of Willett [59].
9. Section 3.10 contains the work of Li and Yeh [41-44,48]. Corollary
3.10.1 is a result due to Wintner [61] (see also theorem 2.17 in Swanson
[58]), whereas Theorems 3.10.3 and 3.10.4 extend and improve some results
ofYan [68].
3.15. References
1. R.P. Agarwal and S.R. Grace, Oscillations of forced functional dif-
269
5. R.P. Agarwal and S.R. Grace, Second order nonlinear forced oscillations, Dyn. Sys. Appl., to appear.
6. R.P. Agarwal, S.R. Grace and D. O'Regan, Oscillation criteria for
certain nth order differential equations with deviating arguments, J. Math.
Anal. Appl., to appear.
7. R.P. Agarwal, S.R. Grace and D. O'Regan, On the oscillation of
second order functional differential equations, Advances Mathl. Sci. Appl.,
to appear.
8. R.P. Agarwal, W.-C. Lian and C.C. Yeh, Levin's comparison theorems for nonlinear second order differential equations, Applied Math. Letters 9(1996), 29-35.
9. A. Besicovitch, Almost Periodic Functions, Dover, New York, 1954.
10. I. Bihari, An oscillation theorem concerning the half-linear differential
equations of second order, Magyar Tud. Akad, lVlat. Kutato Int. Kozl.
8(1963), 275-280.
11. I. Bihari, Oscillation and monotonicity theorems concerning nonlinear differential equations of the second order, Acta Math. Sci. Hungar. 9(1968),
83-104.
12. T.A. Chanturia, N. Kandelaki and A. Lomtatidze, On zeros of solutions of a second order singular half-linear equation, Mem. Diff. Eqns.
Math. Phy. 17(1999), 127-154.
13. O. Dosly, Oscillation criteria for half-linear second order differential equations, Hiroshima Math. J. 28(1998), 507-521.
14. A. Dzurnak and A.B. Mingarelli, Sturm-Liouville equations with Besicovitch almost periodicity, Proc. Amer. Math. Soc. 106(1989), 647-653.
15. A. Elbert, A half-linear second order differential equation, in Qualitative Theory of Differential equations, Szeged-Societas Janos Bolyai, Colloq.
Math. Soc. Janos Bolyai 30, 1979, 153-180.
16. A. Elbert, Oscillation and nonoscillation theorems for some nonlinear ordinary differential equations, Lecture Notes in Math., 964, Springer-Verlag,
New York, 1982, 187-212.
17. A. Elbert, Asymptotic behavior of autonomous half-linear differential
systems on the plane, Studia Sci. Math. Hungar. 19(1984), 447-464.
18. A. Elbert, On the half-linear second order differential equations, Acta.
Math, Hungar. 49(1987), 487-508.
19. A. Elbert and T. Kusano, Oscillation and nonoscillation theorems for a
class of second order quasilinear differential equations, Acta. Math. Hungar, 56(1990), 325-336.
20. L.H. Erbe, Oscillation criteria for second order nonlinear delay equations,
Ganad. Math. Bull. 16(1973), 49-56.
21. S.R. Grace, Oscillation theorems for nonlinear differential equations of
second order, J. Math. Anal. Appl. 171(1992), 220-241.
22. S.R. Grace and B.S. Lalli, Oscillatory behavior for nonlinear second
order functional differential equations with deviating arguments, Bull. Jnst.
Math. Aad. Sinica 14(1986), 187-196.
23. G.H. Hardy, J .E. Littlewood and G. Polya, Inequalities, 2nd ed.
Cambridge Univ. Press, 1988.
270
Chapter 3
271
42. H.J. Li and C.C. Yeh, Nonoscillation theorems for second order quasilinear differential equations, PubJ. l\1ath. (Debrecen) 47(1995), 271-279.
43. H.J. Li and C.C. Yeh, Oscillation of half linear second order differential
equations, Hiroshima Math. J. 25(1995), 585-594.
44. H.J. Li and C.C. Yeh, Oscillation criteria for nonlinear differential equations, Houston J. Math. 21(1995), 801811.
45. H.J. Li and C.C. Yeh, An oscillation criterion of almost periodic SturmLiouville equations, Rocky Mount. J. Math. 25(1995),1417-1429.
46. H.J. Li and C.C. Yeh, On the nonoscillatory behavior of solutions of a
second order linear differential equation, Math. Nachr. 182(1996), 295315.
47. H.J. Li and C.C. Yeh, Oscillation of nonlinear functional differential
equations of the second order, AppJ. Math. Letters 11(1)(1998), 71-77.
48. H.J. Li and C.C. Yeh, Oscillation and nonoscillation criteria for second
order linear differential equations, Math. Nachr. 194(1998),171-184.
49. W.C. Lian, C.C. Yeh and H.J. Li, The distance between zeros of an
oscillatory solution to a half linear differential equation, Comput. Math.
Appl. 29(1995), 39-43.
50. W.E. Mahfoud, Oscillation and asymptotic behavior of solutions of nth
order nonlinear delay differential equations, J. Differential Equations 24
(1977), 75-98.
51. W.E. Mahfoud, Comparison theorems for delay differential equations,
Pacific J. Math. 83(1979), 187-197.
52. W.E. Mahfoud and S.M. Rankin, Some properties of solutions of
(r(t)lJ!(x)x')' + a(t)f(x) = 0, SIAM J. Math. Anal. 10(1979),49-54.
53. J.V. Manojlovic, Oscillation criteria for a second order half-linear differential equation, Math. Comput. Modelling 30(5-6)(1999),109--119.
54. Ch.G. Philos, On the existence of nonoscillatory solutions tending to
zero at 00 for differential equations with positive delays, Arch. Ivlath.
36(1981),168-178.
55. Ch.G. Philos, Oscillation theorems for linear differential equations of second order, Arch. Math. 53(1989), 482-492.
56. M. Del Pino and R. Manasevich, Oscillation and nonoscillation for
(lu'I P - 2 u')' + a(t)luI P - 2 u = 0, p> 1, Houston J. Math. 14(1988), 173177.
57. B. Singh, Comparative study of asymptotic nonoscillation and quickly
oscillation of second order linear differential equations, J. Ivlathl. Phyl.
Sci. 4(1974), 363-376.
58. C.A. Swanson, Comparison and Oscillation Theory of Linear Differential
Equations, Academic Press, New York, 1968.
59. D. Willett, On the oscillatory behavior of the solutions of second order
linear differential equations, Ann. Polon. Math. 21(1969), 175-194.
60. A. Wintner, A criterion for oscillatory stability, Quart. Appl. Math.
7(1949), 115-117.
61. A. Wintner, On the nonexistence of conjugate points, Amer. J. Math.
73(1951), 368-380.
272
Chapter 3
62. F.R. Wong and C.C. Yeh, An oscillation criterion for Sturm-Liouville
equations with Besicovitch almost periodic coefficients, Hiroshima Math.
J. 21(1991), 521-528.
63. J .S.W. Wong, Oscillation criteria for a forced second order linear differential equation, J. Math. Ana1. App1. 231(1999), 235-240.
64. P.J.Y. Wong and R.P. Agarwal, Oscillation theorems and existence
criteria of asymptotically monotone solutions for second order differential
equations, Dynam. Systems App1. 4(1995), 1177-496.
65. P.J.Y. Wong and R.P. Agarwal, On the oscillation and asymptotically
monotone solutions of second order quasilinear differential equations, Appl.
Math. Comput. 79(1996), 207-237.
66. P.J.Y. Wong and R.P. Agarwal, Oscillatory behavior of solutions of
certain second order nonlinear differential equations, J. Ivlath. Ana1. Appl.
198(1996), 813-829.
67. P.J.Y. Wong and R.P. Agarwal, Oscillation criteria for half-linear differential equations, Advances Math. Sci. Appl. 9(1999), 649-663.
68. J. Yan, Oscillation property for second order differential equations with
an 'integral small' coefficient, Acta Math. Sinica 30(1987), 206--215.
Chapter 4
Oscillation Theory for
Superlinear Differential
Equations
4.0. Introduction
This chapter presents oscillation and nonoscillation theory for solutions
of second order nonlinear differential equations of superlinear type. Section
4.1 deals with the oscillation of superlinear equations with sign changing
coefficients. Here, first we shall discuss some results which involve integrals
and weighted integrals of the alternating coefficients, and then provide several criteria which use average behaviors of these integrals. More general
averages such as 'weighted averages' and 'iterated averages' are also employed. In Section 4.2, we impose some additional conditions on the superlinear terms which allow us to proceed further and extend and improve
some of the results established in Section 4.1. In fact, an asymptotic study
has been made and interesting oscillation criteria have been proved. In
Section 4.3, first we shall provide sufficient conditions which guarantee the
existence of nonoscillatory solutions, and then present necessary and sufficient conditions for the oscillation of superlinear equations. Oscillation
results via comparison of nonlinear equations of the same form as well as
with linear ones of the same order are also established. In Section 4.4,
we shall extend some of the results of the previous sections and establish
several new oscillation criteria for more general superlinear equations. Necessary and sufficient conditions for such equations to be oscillatory are also
given. Section 4.5 deals with the oscillation of forced-super linear differential equations with alternating coefficients. Finally, Section 4.6 presents the
oscillation and nonoscillation criteria for second order superlinear equations
with nonlinear damping terms.
274
Chapter 4
J;"(t)
+ q(t)f(x(t))
0,
+ q(t)f(x(t))
(a(t)x'(t))'
(4.1.1 )
(4.1.2)
(a(t)x'(t))'
+ p(t)x'(t) + q(t)f(x(t))
0,
(4.1.3)
where
(i)
(ii)
(iii) f E C(JR, JR),
and satisfies
f'(x) 2
for
x =I 0
( 4.1.4)
<
(4.1.5)
-= du
f(u)
x"(t)
+ q(t)lx(t)17
(a(x)x'(t))'
and
(a(t)x'(t))'
sgn x(t)
+ q(t)lx(t)17
1) is of particular
= 0,
sgn x(t)
+ p(t)x'(t) + q(t)lx(t)17
b >
00.
(4.1.6)
= 0
sgn x(t)
(4.1.7)
(4.1.8)
arc prototype of (4.1.1), (4.1.2) and (4.1.3) respectively, and will be discussed extensively.
We begin with the following oscillation criteria for equation (4.1.2).
Theorem 4.1.1.
1=
to
L
OO
lim inf
t--+CXl
-ds =
a( s)
00,
(4.1.9)
q(s)ds <
00,
(4.1.10)
J7fTt q( s )ds
(4.1.11)
i 1
and
00
lim t -1()
q( u )duds
t--->oo to a s s
275
(4.1.12)
00.
w'(t)
T:::: tl
For
q(t) -
f'~~g)) w 2(t) ~
- q(t)
a(t)x'(t)
a(T)x'(T)
f(x(t)) ~ f(x(T)) -
t:::: t 1 .
for
t ( )d
iT q
(4.1.13)
s.
o ~
t:::: T,
a(T)x'(T)
f(x(T)) -
roo
iT
q(s)ds.
The same type of reasoning as above guarantees for t:::: T that we have
ftOO q(s)ds ~ a(t)x'(t)/ f(x(t)). Dividing both sides of this inequality by
aCt) and integrating from T to t, we get
it 1
1
00
a(s)
q(u)duds <
-
it
T
l (t)
x'(s)
ds =
f(x(s))
x(T)
de,
.
f(e,)
it
TN
and hence
a(t)x'(t)
h~~p f(x(t))
a(TN)x'(TN)
f(x(TN))
+ li~s~p
()d s,
qs
( tq(s)ds
)
-
iTN
< 0,
Chapter 4
276
TI to t, we find
a{t)x'(t)
= a(Tdx'(Td - t q(s)f(x(sds
iT!
iT,
iT!
iT,
q(u)duds
S a(T1)x'(TI).
Thus, x(t) S X(TI) + a(TI)x'(TI) J;! dsja(s) ---+ -00 as t ---+ 00, which
contradicts the fact that x(t) > 0 for t ~ i}. A similar proof holds when
x(t) <0 for t~tl'
lim l t q(s)ds =
t--+oo to
(4.1.11h
00,
00
q(s)ds
(4.1.11h
For this, let (4.1.11) hold. Define Q*(t) = Jt: q(s)ds. We first claim that
liminft--+oo Q*(t) i= -00. Suppose this is not true. Then there exists a
sequence {tn}~=l' tn ---+ 00 as n ---+ 00 such that Q*(tn ) S -no For
large T, wehave J;nq(s)ds=Q*(tn)-Q*(T)S-n-Q*(T)---+-oo as
t ---+ 00, which contradicts (4.1.11). If liminft--+oo Q*(t) = 00, we have
(4.1.11h. Thus, we assume that liminfQ*(t) = a i= 00. Suppose that
t--+oo
limsuPt--+oo Q*(t) = f3 > a. Then there exist two sequences {tn}~=l and
{Tn}~=l' tn ---+ 00 and Tn ---+ 00 as n ---+ 00 such that Q*(tn ) ---+ a
and Q*(Tn) ---+ f3 as n ---+ 00. Then, J;~ q(s)ds = Q*(t m ) - Q*(Tn) <
(f3-a)j2 (with appropriate modification if f3 = 00) for m, n sufficiently
large. This contradicts (4.1.11). Thus, limt--+oo Q*(t) exists, and Q(T) =
limt--+oo J; q(s)ds = liminft--+oo J; q(s)ds ~ 0 as assumed in (4.1.11). The
sufficiency part is obvious.
Theorem 4.1.2. Let a(t) Sal, t ~ to where al is a positive constant,
and condition (4.1.5) holds. If for every large T there exists a TI ~ T
such that
lim inf
t--+oo
and
t r q(u )duds
iT! iT
00
to
sq(s)ds
> 00,
00
(4.1.14)
(4.1.15)
277
Proof.
t a(s)x'(s)
iTt f(x(s)) ds
:S;
t r
iTt iT q(u)duds
~ T1
ta(t)x'(t)
T1a(Tdx'(Tt)
f(x(t)) :S;
f(x(T1))
+ -
{t
()d
iTt sq S
s,
ta(t)x'(t)
f(x(t))
tla(tl)x'(td
f( X (tl ))
t, f(x(s))
+ al
t,
(t) du
it
f( ) sq(s)ds,
x(t,)
t,
ta(t)x'(t)
(t f'(x(s))x/(s) i
t3
t3
+ t3 a(t3)X'(t3) In!...
al
t3
--+ -
00
as
t --+
00
Chapter 4
278
to.
1
1
00
and
00
p(s)q(s)ds =
00
(4.1.16)
00.
(4.1.17)
--:--:--:---:- ds
a(s)p(s)
Joo
du
-- <
+0 f(u)
00
and
du
Lo f(u)
<
(4.1.18)
00
hold, and
IfJ'(s)lds < 00, then all bounded solutions of (4.1.3) are
oscillatory. If in addition condition (4.1.5) holds, then (4.1.3) is oscillatory.
Let x(t) be a nonoscillatory solution of equation (4.1.3), say,
~ to. Define w(t) = p(t)a(t)x'(t)/ f(x(t)), t ~ t} for some
to. Then for t ~ t}, we have
Proof.
t} ~
w'(t)
-p(t)q(t)
x'(t)
+ f3(t) f(x(t))
1
2('( ( ))
- a(t)p(t) w t)f x t
x'(t)
w(t) ~ w(t}) -
x'(s)
it}r p(s)q(s)ds + itr1f3(s) f(x(s))
ds.
Now, we shall show that each of the conditions (h) and (i2) ensures
that
f3(s)x'(s)/f(x(s))ds is bounded above. Consider the following
two cases:
Jtt}
I}. Let (i}) hold. By the Bonnet theorem, for a fixed t ~ tl and ~ E [t}, t),
we have
x'(s)
it 1f3(s) f(x(s)) ds
($. x'(s)
= fJ(tI) i h f(x(s)) ds
r($.)
f3(tI)
iX(tl)
du
f(u)
279
we find
[
where KI
t,
f(u) <
x(ttl
du
()
00
x(t,)
du
-f()
u
x'(s)
(3(s) f(x(s ds < KI
(4.1.19)
t,
t?: tl,
for all
x(~)?: x(td,
if
+ w(t l ) + KI
[t
t,
( ) ( ) wZ(s)f'(x(sds.
asp s
(4.1.20)
t,
x'(s)
(3(s) f(x(s ds
ret)
= (3(t) io
-
t,
du
ret,) du
f(u) - (3(td io
f(u)
(r(S) f(u)
du )
(3'(s) io
ds.
(4.1.21)
I;(tJ
(I;(S)
t,
x'(s)
(3(s) f(x(s ds :s: MI
where MI = (3(td
t,
+ MIN + M
I;(t,) du/f(u).
p(s)q(s)ds :; - w(t)
[00 1(3'(s)lds
t,
Thus, we obtain
+ W(tl) + Kz -
it
t,
( ) ( ) wZ(s)f'(x(sds.
asp s
(4.1.22)
Set K = max{KI, K z }. Then, (4.1.20) and (4.1.22) can be written as
t
t,
+ w(td + K
-it
t,
()1 ( ) wZ(s)f'(x(sds.
asp s
(4.1.23)
280
Chapter 4
Ittl
-w(t) ::::: 1 +
1 () ()
1
tl
asp s
w 2(s)f'(:r(s))ds,
(4.1.24)
or
1l.,2(t)f'(:r(t))/a(t)p(t)
>
x'(t) ,
f(x(t)/
(x(t))
1 -+ 1/1 (w 2(S)f'(x(s))/a(s)p(s)) ds
and consequently for all t::::: t3,
In{~ (1+
where )..
1+
It
(w 2(s)f'(x(s))/a(s)p(s))dS)} :::::
In ~(~(;:/;,
1( 2 '
t
tl
)
f(X(t3))
W (s)f (x(s))/a(s)p(s) ds ::::: ).. f(x(t))
Thus, (4.1.24) yields x'(t) <::: -)..f(X(t3))/a(t)p(t) for t::::: t3. Therefore,
we have
t
1
1
t3
( ) ( ) ds
asp s
-00.
Finally, we remark that in the above proof the oscillation of all bounded
solutions of (4.1.3) is trivially included.
IX! sq(s)ds
00,
(4.1.25)
ta )' >0
( -a(t)
-,
281
(4.l.26)
= 0, t 2" to = n/2
l. Let p(t) = t. Then, j3(t) = a(t)p'(t) -- p(t)p(t) = 2/t and
x (lx(t)I"Y) sgn :r(t)
where 'Y>
for every t 2" to, we have
t
to
sq(s)ds
j
j
~s-1/2(2+COss)J
t
n/2
[-sl/2sins+
t
7r/2
d[sl/2(2+coss)]
> t 1/ 2
2( n /2) 1/2
-7 00
ds
t 1/ 2(2+cost)-2(n/2)1/2
as t
-7 00.
X(t)
du
y'U(I
+ u)
= 2 tan -1 Vx and so
(4.1.27)
and - f( -x) = f(x).
OO du
f (u)
<
00.
fX)
It:
00,
for
=1=
0.
(4.1.28)
282
Cha.pter 4
(3(t)
t ~ to (4.1.29)
and
1 it a(s)p(s)ds <
lim sup "2
If
t ....= t
to
li~~f
p(s)q(s)ds > -
1:
00.
(4.1.30)
00
(4.1.31)
and
lim sup -l i t i s p(u)q(u)duds
t ....
to
(4.1.32)
00,
to
w'(t) =
Hence, for all t
- p(t)q(t)
~
+ (3(t) f~~~ti))
- a(t)lp(t) w 2 (t)f'(x(t)).
i to p(s)q(s)ds = - w(t)
+ w(to) +
it (3(s)x'(s)
it w 2 (s)f'(x(s))
to f(x(s)) ds to
a(s)p(s) ds.
to there exists a
t
x'(s)
i x'(s)
i to (3(s) f(x(s)) ds = (3(to) to f(x(s)) ds
~ E
(3(to)
< (3(to)
Thus, for every t
du
x(to} f(u)
x (}
00
x(to}
du ).
-f(
U
to, we find
t ( ) ( )d
() M - it w 2 (s)f'(X(S))d s
i to psqs s:::; -wt+
to
a(s)p(s)
,
(4.1.33)
it
~
it
w 2 (s)
~asps
283
Next, we need to consider the following three cases for the behavior of x'(t).
Case 1. x'(t) is oscillatory. Then there exists a sequence {tm}~=l in
[to, 00) with limm-too tm = 00 such that x'(tm) = 0, m = 1,2,.
Thus, (4.1.34) gives
t",
to
11
W2(S)
M
()()ds:::;-k--k
asp s
,
tm
to
p(s)q(s)ds,
m=1,2,
00
to
w2(s)
-:-:-'---'--,-ds <
a( s )p( s)
(4.1.35)
00.
It:
I-l
W(S)dSI'
: :; [1:
a(s)p(s)dsJ
1: a~~~~~)dS
Ct 2 for t ~ to,
to
-It
p(u)q(u)duds :::;
to
to
It: w(s)ds
+M
a(s)p(s)ds.
:::; VNCt.
to
litis
i.e.,
Itto
to.
0 such that
I 15
c>
1:
:::; N
to
It:
!
t
rt 15 p(u)q(u)duds
iT
to
:::; M
(1 -~)t
for
I -
00
to
to. If
w2(S)!'(X(S))d
s <
a(s)p(s)
00
~ T,
Chapter 4
284
then condition (4.1.28) ensures that (4.1.35) holds. But, then as Case 1 we
get a contradiction. Thus, we assume that I = 00. Now, exactly as in
the last part of the proof of Theorem 4.1.3, we arrive at the contradiction
limHoo x(t) = -00.
In the following result we shall employ the averaging technique and the
weighted integral functions to discuss the oscillatory behavior of equation
(4.1.3).
Theorem 4.1.5. Let conditions (4.1.5) and (4.1.28) hold, and assume
that there exist p(t) E C1([to, 00), m+) and cf; E <I>(t, to), where <I>(t, to)
denotes the class of all positive and locally integrable functions on [to, 00)
such that (4.1.29) is satisfied,
(It:
cf;(s)
cf;(u)du) >.
rS ()
)
2(
)
ds
lto au p(u cf; u) du
00
/
is
and
lim
t---+oo fto
it cf;(s)
cf;( s )ds
to
00
(4.1.36)
p(u)q(u)duds
00.
(4.1.37)
to
Proof.
to
is
cf;(S)W(S)ds+klt cf;(s)
to
to
~~(~)
t
::; Mit cf;(S)dS-l cf;(s)
to
to
M -
fto
o ::;
cf;( s )ds
F(t)
it is
cf;(s)
to
)
is
a u Pu
duds
(4.1.38)
p(u)q(u)duds.
to
to such that
= k
it is
to
cf;(s)
tl
to
to
w 2 (u)
( ) ( ) duds < a u Pu
it
to
cf;(s)w(s)ds
and
(4.1.39)
285
[ltDt
[4>(s)Ja(s)p(s)] (
w (s)
)
(ltDt a(s)p(s)l(s)ds) (rltDt a(s)p(s)
ds
4>tt)
P(t) :2: k
(4.1.40)
<
We also have
w(s) ) dS]2
Ja(s)p(s)
(1:
a(s)p(s)4>2(S)dS) P'(t)
i (itl
t
t,
4>(s)
for
t;::: tl
it
2
W
( )(U))
( ) du ds = c
4>(s)ds,
a u pUt,
(4.1.41)
fto a(s)p(s)4>2(s)ds
::;
p A - 2 (t)P'(t)
for all
t;::: t l ,
(4.1.42)
t,
4>(s)
(It: 4>(u)du) A l l
ds < - A
<
Ito a(u)p(u)4>2 (u)du
1 - A pl- (tI)
s
00,
Remark 4.1.3.
Condition (4.1.28) fails in the special case f(x) =
lxi' sgn x, x E JR b > 1). However, it is satisfied for the following
functions:
(I)
(II)
f(x) =
f(x) =
(III) f(x) =
(IV) f(x) =
mx + lxi' sgn x, x E 1R. for 'Y > 1 and any constant m> 0,
xln2(p + Ix!), x E JR for any constant p> 1,
xe AlxI , x E lR for any constant A> 0,
sinh x, x E JR.
it it [
.
1
1 [32 (u) ]
lIm t
4>(s)
p(u)q(u) - k ( ) ( ) duds
t ..... oo fto 4>(s)ds to
to
4 au p u
00,
(4.1.43)
Chapter 4
286
then equation (4.1.3) is oscillatory.
Proof.
x(t) >
t ?: to,
for
(3(t)
1
2
,
-p(t)q(t) + a(t)p(t) w(t) - a(t)p(t) w (t)f (x(t
w'(t)
k
[
(3(t)] 2
1 (32(t)]
< - [p(t)q(t) - 4k a(t)p(t) - a(t)p(t) w(t) - 2k
Hence, we have
w(t)
+k
t
1
[
(3(s)] 2
to a(s)p(s) w(s) - 2k ds
-s: w(t o) -
t [
1 (32(s) ]
to p(s)q(s) - 4k a(s)p(s) ds.
oo
(32(S)
a(s)p(s) ds <
(4.1.44)
00.
C~X/(t)'+S~~txl(t)+~(~-sint)f(x(t
t>to=l (4.1.45)
0,
l:
t (32(s) d
to a(s)p(s) s
p(S)q(S)dS
{t (1- sins)2 d
= Jl
it(~-sins)ds
s3
<
00,
and
1-2
t -1
jt
s[lns - 2Jds
-t 00
as t -t 00. Thus, all the conditions of Corollary 4.1.5 are satisfied with
A E (0,1), where A is the constant in (4.1.36), and hence (4.1.45) is
oscillatory.
287
(t 2 x'(t))' - 2tx'(t)
20 + cost + -2-(2
2tsint)f(x(t))
(4.1.46)
where f(x) is as in Remark 4.l.2. We take pet) = 1/t 2 and (t) = t, t 2::
to> 7r/2. Then, (3(t) = a(t)p'(t) - p(t)p(t) = 0 for t > 7r/2. Further,
1
t
p(s)q(s)(1.s
to
([-~Sin8+ 2ys
\:;(2+COSS)]dS
./7r/2
./7r /2
Vi(2
and
t 1
1 t ( s)
Ito ( s )ds to
1
8
to
d [vIS(2
+ coss)]
+ cost)
- 2J(7r/2) >
t - 7r
/2)21
Vi -
7r /2
-+
2J(7r/2)
(vis - Vq)
ds
"2
2
00
as
t -+
00.
Thus, all the hypotheses of Corollary 4.l.5 are satisfied with A E [1/2,1) C
(0,1), and hence (4.l.46) is oscillatory.
Now, we shall employ some integral inequalities to study the oscillation of some of the above equatiolls. For this, we shall lleed the following
lemmas.
Lemma 4.1.1. Let k(t, s, y) be a real-valued function of t, s in [b, c)
and y in [bl, CI) such that for fixed t = to and s = So, keto, So, y) is a
nondecreasing function of y. Let G(t) be a given function on [b, c); let
v. and v be functions on [b, c) such that 11.( s) and v( s) are in [b l , CI)
for all s in [b,c); let k(t,s,v(s)) and k(t,s,u(s)) be locally integrable
in s for fixed t; and for all t E [b, c). Let
vet) = G(t)
it
k(t,s,v(s))ds
and
it
k(t,s,11.(s))ds.
-w(b l
1
t
b,
q(s)ds
1b -(
b,
1
)w 2(s)f'(x(s))ds 2:: C,
a s
(4.1.47)
Chapter 4
288
where
then
a(t)x'(t) :::;
(4.1.48)
for b:::; t :::; c. Now, the integral in (4.1.48) is nonnegative, so from the
condition f'(x) ~ 0 for x#-O and the definition of w, we find that
x(t)x'(t) < 0 on [b, c). If x(t) > 0, we let u(t) = -a(t)x'(t), then
(4 .1.48) becomes
u(t)
Cf(x(t))
It
b
Define
k(t,s,y)
f(x(t))
f'(x(s))( -x'(s))
j2(x(s))
y
for t and s in [b, c) and y E [0,00) and observe that in this domain
k(t, s, y) is nondecreasing in 11 for fixed t and s. Therefore, Lemma
4.1.1 applies with G(t) = Cf(x(t)), and we have u(t) ~ v(t), where
v(t) satisfies
G(t) = -Cf(x(t)).
iXi-400
(4.1.49)
oo
t-400
(4.1.50)
(4.1.51)
289
CXl 1
[00
= [ t a(s) w 2(s)f'(x(s))ds + t q(s)ds
w(t)
(4.1.52)
t2
- ()ds ---+ a s
00
as
t ---+ 00,
which contradicts the fact that x(t) > 0 for t ::;> tl' The prooffor the case
when x(t) < 0 for t::;> tl is similar and hence omitted. This completes
the proof of (4.1.50).
Now, since
- q(t),
we have
w(T) +
[T ats) w2(s)J'(x(sds
= w(t) +
[T q(s)ds,
(4.1.53)
which together with (4.1.10) and (4.1.50) implies that limT---+oo w(T) exists, say, w(T) ---+ C 2 as T ---+ 00, where C 2 is a constant. Then from
(4.1.53), we find
w(t) = C 2 + [ t
CXl
q(s)ds +
[CXl
t
a(s) w 2 (s)J'(x(s)ds,
t ::;> tl'
(4.1.54)
Il
q(s)ds
CXl 1 2 '
::; -C-2 for t::;> Tl and
- ()w (s)f (x(s))ds < - C
-2.
4
n as
4
290
Chapter 4
U'(t)
tl.
a(t)x'(t) > C 2
f(x(t)) - 2
Therefore, we have
()f'(x(s))(X'(S))2 d
j2(x(s))
s
iTo a s
> C2
2
f'(x(s))x'(s) ds = C 2 In f(x(t))
j2(x(s))
2
f(x(T2))
iTo
The last inequality together with (4.1.49) and (4.1.50) implies that x(t)
is bounded from above. Since x' (t) > 0 for t 2: T2 and f' (x) 2: 0 for
x -I 0, it follows that f(x(t)) 2: f(x(T2 )) for t 2: T 2. Therefore, we
obtain
x'(t) > C2 f(x(t)) > C2 f(x(T2)) for t 2: T2 .
- 2 a(t)
- 2
a(t)
Integrating the above inequality from T2 to t, we get
C f(X(t2))
x(t) 2: x(T2) + ~
2
it
T2
-1()ds,
a s
which contradicts the boundedness of x(t). This completes the proof that
C2 = 0 for the case x(t) > 0, t 2: tl. A similar proof holds for the case
when x(t) < 0, t 2: t l .
x'(t)
1
f(x(t)) 2: a(t)
['XJ
it
q(s)ds,
t
itl
x'(s) d
f(x(s)) S
and hence
G(X(tl)) - G(x(t)) 2:
r(t)
du
ix(t,)
f(u)
i 1
t
t,
1
-()
ass
00
to t
G(x(tIl) - G(x(t)),
q(u)duds --+
00
as
t--+oo,
291
Theorem 4.1.7.
and
11 f~~)
= 00 = [1
f~~) .
(4.1.56)
- 00,
(4.1.57)
00,
(4.1.58)
i
1
lim inf
t--+oo
then either
(4.1.55)
00
to
. to
1
-(-) Q( s )ds >
a s
1 ( + ) 2
_
-(-) Q (s) ds a s
or
lim sup
t--+=
lto
_(1) (Q(s)
a s
roo
ls
_(1) (Q+(u2 dU ) ds
a U
00,
(4.1.59)
x'(t)
f(x(t
t:::: to
(4.1.61)
This together with (4.1.57) and (4.1.56) implies that x(t) is bounded below
away from zero. Hence, by (4.1.55), J'(x(t:::: c > 0 for some constant
c. From (4.1.61), we find
(
X' ( t ) )
f(x(t
1
+
:::: a2(t) (Q (t
Chapter 4
292
Thus, it follows that
00
(f~~~;)
a(s)
f'(x(sds 2 c
00
ats) (Q+(s2ds.
00
as
. lt
(I) hmsup
t-4oo
to
Q( s) ds -_
-(-)
a s
or (II)
00,
1 1
00
to
-1()
00
ass
-1() (Q + (u
aU
duds -_
00.
lim sup
t-400
lt
to
_(1) (Q(s)
x'(t)
f(x(t
cl
ass
a(t) Q(t)
+ a(t)
00
a(s)
f'(X(UdU) ds
aU
CXl
(4.1.62)
(X'(S)2,
f(x(s)
f (x(sds.
(4.1.63)
roo
1
(+
2
ito a(s)6(s) Q (s ds
where 6(t) = .ftto dsja(s),
can be very large),
ito a s s
00,
to 2 1
(4.1.64)
293
Thus, there exist a large t1 2' to and positive constants 61, 62 such that
- 61~1(t) ::; -x1-'Y(t) ::; 62~dt) for t 2' t 1, or "(X'Y-1(t) 2' 616(t) for
t 2' t 1, where 6 = "(161, Now, since
(x'(s)?
a(sh x'Y +1( s )ds 2'
t.
00
00
(s)) ds,
a ()
s 6 ()(Q
s
(4.1.66)
liminf
t--+oo
to
q(s)ds >
( 4.1.67)
00
and either
(4.1.68)
or
lim sup
t-+oo
It
to
[Q(s)
1=
s
(Q+(U)?dU] ds
00,
(4.1.69)
It
to
[Q(s)
+ s(Q+(S))2] ds
00.
This follows by changing the order of integration in (4.1.69) and then using
(4.1.67).
2. Butler's theorem [4] is in fact Corollary 4.1.6 without (4.1.56). In the
proof [4, pp. 79] he attempted to show that either (4.1.56) or (4.1.28) must
Chapter 4
294
hold in general, but the proof only works when f' is continuous in a neighborhood of zero, a condition which is not assumed in the hypotheses. In
other words, Corollary 4.1.6 holds if (4.1.56) is replaced by the assumption
tha.t f is continuously differentiable.
3. Under further restrictions on the negative part of Q(t), Butler [4]
showed that (4.1.69) is also necessary for the oscillation of (4.1.1). More
deta.ils of this will be given in Section 4.3.
ft
OO
1 ~(Q+(s))2ds
00
to
00,
to:::: 1
(4.1.70)
it
to
[A(S)
+0
100 ~(Q+(u))2dU]
s
ds
(4.1.71)
00,
Remark 4.1.5.
it
to
[Q(s)
+ 8(Q+(s))2] ds
00.
it
to
[Q(s)
+ 82100 ~Q~(U)dU] ds =
00,
x' ((t))
X7 t
Q(t)
+ 81 ~(Q+(s))2ds
00
295
into (4.1.52).
For the formulation of our next theorem, we note that if (4.1.10) holds,
then the function ho(t) = (1/ ya(t)) j;'OO q(s)ds is well defined on [to, (0).
Further, as long as the improper integrals involved converge, we can define
00
hdt)
hn+,(t)
([ho(s)]+)2 ds
r ([h
..
O(')
Ik
~:~:;Jr d"
"
1,2,
+ k hN(t) >
ho(t)
y'a(t)
(4.1.72)
oo
_1_ [ho(S)
y'a(t)
+ k hN(S) ds
Va(S)
(4.1.73)
00,
00
tJ
- ()w2(s)ds <
a s
(4.1.75)
00.
(4.1. 76)
x'(t)
1
[h ()
f(x(t)) 2: Ja(t)
0 t
k hN(t)
+ Ja(t)
for
2: t2'
296
Chapter 4
'X(t)
X(t2)
1t
v, >
-df(v,) -
t2
xl/(t)
Here a(t)
h ( )
~'
NS
h()
k d
--o
s+
-s
O.
(4.l.77)
t>O.
00
ho(t) =
q(s)ds =
2 + cost
1
t3/4
> t3/4'
which implies that (4.l.72) and (4.1.73) are satisfied with N = l. Observe
that f'(x) = 3x 2 + 1 > 1 for x> 0, so (4.l.28) holds with k = l. Thus,
all the hypotheses of Theorem 4.l.9 are satisfied and hence all solutions of
(4.l. 77) are oscillatory.
is such that
Joo
v;:;:
min
~d ]2
{ [1 1
and
dv, <
00
feu)
inf
x>o
00
00
J- oo
and
v;:;:
dv, <
(4.2.1)
00
v,
du
feu)
'
~~t Jf'(x) ix
oo
.fF(U5
f(v,) dv,
> O.
297
{~~t f'(x)
1 f~:)'
inf f'(x)
00
x<o
-00
f(U)
> o.
ix
r(sgn x)oo du
(f1(u2du ::::: c ix
f(u)
f'(x)
1
l(s gn x)oo du
lim - - ::::: c x
f(u)
lul ..... oo f'(u)
for
x -=f.
o.
(4.2.2)
f (x) ix
f(u) > ~
for every
x -=f. 0
J f'(X)
ix
f(u) du > k
for all
x -=f. O.
r(sgn x)oo
ix
J7'[U5 ) -IF(X5
f(u) du
and hence
lx
J7'[U5
f(x)
k
sgn x > f(x) sgn x
Jf'(z)
r(sgn
f(u) du 7(z)dz > k ix
X)CXl
du
f(u)
298
Chapter 4
[ r(sgn
ix
J f'(u) dU]2
r(sgn
f(u)
2: k ix
x)oo
x)oo
du
f(u)
for every
0,
f'(x) =
and consequently
f(x) sgn x
Jf'(x)
>
("f [A + (1 + ~) ]) 1/2
Ixlh+1)/2
for all
(s g n
x)oo
v'P(U5
f(u) du
_
-
>
O.
h+
1J
00
Ixl
f'(u)
f(u)
--=---..:.~-"-du
2("f [A - (1 + 1)])1/2
Ixl
--''-----''----'---,---'--'-..:.~-I x I(1-,) /2
(A + l)h - 1)
299
Jf'(x)
(s gn
x)oo
Jf'(u)
21' [,\ - (1 + ~)]
f(u) du >
('\+1)h-1)
> 0,
f(x) sgn x
v'f'(x)
>
where
h + 1)/2 > 1.
Ixl"! + X
hlxl"!-l + 1)1/2
Ixl"!
h + 1)1/2
IxlC'Y+1)/2
'
f(x)f"(x)
(f'(X))2
Moreover,
(s gn
x)oo
roo
du =
du
f(u)
ilxl f(u)
roo
-=I- 0,
du
ilxl u 2 ,,!
roo du
+ ilxl
u"!
IxI 1- 2"!
IxI 1-"!
+
-.
21' - 1
I' - 1
Chapter 4
300
Furthermore, we find
yfF(x)
/,1/2Ixl(2,-1)/2 (2
f(x)
sgn x
Ix1
<
[J
(sgn
x)oo
Jx
(sgn
for x
i= O.
+ Ixl,)1/2
2,
:s;
(2/,)1/2
Ixl(r+l)/2
h + 1)/2 >
1, ensures the
E"Ei du ] 2
f(,,)
x)oo
f(1t)
t:(t)
."
to
a(s)p(s)
ds
and
lim c.(t)
t-+oo
00 .
( 4.2.3)
2: to. (4.2.4)
Moreover, let
{3(t)
a(t)p'(t) - p(t)p(t) 2: 0
If
liminf
t-+oo lto
and
1
lim sup t:( )
t-+oo
." t
it
to
and
p(s)q(s)ds > -
( )1 ()
asp s
is
to
(4.2.5)
00
p(u)q(u)duds
00,
(4.2.6)
(a(t)y'(t))'
+ p(t)y'(t) + q(t)f*(y(t))
0,
= p(t)a.(t)x'(t)/f(x(t)),
I
[(a.(t)XI(t))'
( x'(t)
(t) = p(t)
f(x(t)) - a.(t) f(x(t))
(t) = - p(t)q(t)
t ~ T. Then for
T,
Xl
(t)
+ (3(t) f(x(t))
T, we find
301
t (3(s)x
(s)
rt w 2a.(s)p(s)
(s)f'(x(s))
f(x(s)) ds - .IT
ds.
l
XI(S)
JT (3(s) f(x(s)) ds
(4.2.7)
(3(T)
XI(S)
f(x(s)) ds = (3(T)
JT
r(ry)
Jx(T)
7) E
[T, t],
du
f(u)
du
{
f(u) <
r(ry)
Jx(T)
we obtain
t p(s)q(s)ds
JT
l(T)
(Xl
t (3(s) f(x(s))
x/(s)
ds
JT
where K = (3(T)
for t ~ T that
if
:s; K
for all
~ x(T),
t ~ T,
(4.2.8)
+ w(T) + K
t w 2a.(s)p(s)
(s)f'(x(s))
ds.
iT
(4.2.9)
I(T, t) =
JTt
a.(s)p(s) w2(s)j'(x(s))ds
302
Chapter 4
t
I
x'(s)Jf'(x(s)) dsl2
iT
f(x(s))
Ja(s)p(s)
[It a(s)~(s)dsJ
<
Thus, in view of (4.2.10), we find
IiTt
T we obtain
2
w(s)Jf'(x(s)) ] d 1
Ja(s)p(s)
s
J(T,t).
1
x'es)
12 < N t
ds
f(x(s) J f'(x(s))ds
T a(s)p(s)
N~(t)
for all t
T.
(4.2.11)
Next, using (FI)' we get
l(t)
CXl
du
< !vI
feu) -
l(t)
CXl
Jf'(u) ]
du
feu)
(4.2.12)
t ~T
KI =
d
feU) > 0 and
x(T) U
CXl
w(s)
l i t x'es) 1
IJTt a(s)p(s)
ds = iT f(x(s ds =
IKI
-1:) f~:)
< KI + M
[l
CXl..jF(Uj
x(t)
Kl
feu) du
]2
00
x(T)
J f'(U)
f( ) du > 0
U
we obtain
T,
K2 =
(xU) du
ix(T) feu)
1:) f~:)
[ l (t)..jF(Uj
X
Kl + M K2 -
x(T)
feu) du
]2
r*)..jF(Uj 1]2
< Kl + M [K2 + 1iX(T) feu) du
KI
+M
[K2
+ Ilt
f~~~s;)
I i a(s)lp(s) W(S)dSI
f'(X(S)dsIJ
T, we have
rTO
ito a(s)p(s)
303
> to so that
ds ~ l.
(4.2.13)
Ilt
a(s)1p(s) w(s)dsl :S
C~(t)
for every
~ T*,
lt p(s)q(s)ds :S - w(t)
+ w(T) + K
for
(4.2.14)
~T
~
T*,
iT a(s)p(s) iT p(u)q(u)duds
rt
( ) ()
toasps
=
+ Jw(T) + KJ.
is
to
C*~(t),
p(u)q(u)duds
r
+ [ rT P(U)q(U)dU] t _(
ito
iT asp
t
r
+ iT a(s)p(s) iT p(u)q(u)duds
r
T
()1 ( )
p(s)q(u)duds
ito asp s ito
)1 ( )ds
s
:S
11:
a(s{p(s)
1:
P(U)q(U)dUdSI +
11:
P(U)q(U)dUI
+ C*~(t)
a(s)1p(s) ds
to
( ) ()
asp s
is
to
p(u)q(u)duds:S Cl~(t)
for all
t ~ T*,
304
Chapter 4
where
-w(t) ?: m
+ I(T, t)
for every
t?: T.
(4.2.15)
>
m+I(T,t)
x'(t)f'(x(t))
f(x(t))
+ I(T, t) >
1 f(x(TI))
n f(x(t)) .
x(t) :S x(Td -
MI
()ds
asp s
( )
T,
for all
t?: T I ,
-00.
lim x(t)
t ..... oo
The following corollaries, which are well known basic results in the literature, are immediate from Theorem 4.2.1.
Corollary 4.2.1. If
liminf
t ..... oo
and
11 l
t q(s)ds
lto
to
to
> -
q(r)drds
(4.2.16)
00
00,
(4.2.17)
305
Corollary 4.2.3. Suppose condition (FI) holds and let p(t) E Cl([to, (0),
JR+) be such that p' (t) is nonnegative and decreasing on [to, (0). If
condition (4.2.5) holds, and
lim sup
t-+oo
[I
to
-1()ds]
p
-ll l
t
to
1
-()
PS
to
p(u)q(u)duds
00,
(4.2.18)
C~
A-
~) } .
t-+= Jto
s>'q(s)ds >
00
Equation (4.1.1) is
Chapter 4
306
and
limsuptA-l
t-tCXJ
lim sup - 1
t-too In t
Example 4.2.4.
(0x'(t)),+
t S-A ltot
lto
it I1
-
to S
uq(u)duds
to
U A q(1L)duds
for 0 <:::;)..
00
<1
for A = 1.
00
l~x'(t)+ l-sint+~(2+cost)jf(x(t))
2t
= 0,
2y t
(4.2.20)
t ::;> to = 7r /2
where the function f is as in Examples 4.2.1 - 4.2.3. Let pet) = 0
t::;> to. Then, we have f3(t) = a(t)p'(t) - p(t)p(t) = 0 and
a(t)p(t) = t,
~(t)
(p(s)q(s)ds
1t o
and
11r/2
11r/2
1
t
to
1 1/
to
[-.JSSins+
d[y'S(2+coss)]
I ( / ) 1t n 2t 7r
to S
( ) ( ) ds = In -to -+
asp s
as
00
t -+
for
00,
1;-;,(2+COSS)j ds
2ys
y'uq( u)duds
([1v's--;Y"2
::;>
2Fj
In(2t/7r)11r/2
ds
20
2,Ji/2
In(2t/7r) - In(2t/7r) - 2y
"2
and consequently,
liminf1t y1Sq(s)ds >
t-too
and
lim sup I (21 / )
t-too
n t 7r
to
it I1
-
to S
to
00
y'uq(u)duds
00.
" [1
x (t)+ -
~tsint+2t ~t(2+cost)
y~
y~
j [x(t)['Ysgnx(t)
0,
t ::;> to = 7r/2
(4.2.21)
and
1lt 11"
-1-
to S
nt
to
>
uq(u)duds
307
to, we find
In t
7r
F[In2"-l.
Vi
2
2 1nt -2 Y 2"+lntY2"
7r
Thus, all conditions of Corollary 4.3.4 are satisfied and hence (4.2.21) is
oscillatory. On the other hand, Corollary 4.2.1 is not applicable for (4.2.21).
Indeed, for t ~ to, we have
it [- ~sins+
tq (s)ds =
it
to
<
-
7r/2
7r /2
- Vi
it
7r/2
11tl"
to
1r;;(2+COSS)]dS
2sy S
(- 1
- sin s + -3 -1-) ds = -1 cos t
JS
2 sJS
Vi
< - 1 cos t + 2
Y S
to
1
--ds
sJS
= -1
Vi
it (- -
7r/2
cos t - 4 [1
.- -
Vi
1 cos s + -3-) ds
2sJS
2sJS
~j
7r
y2
< - -3 + 4 -.
Vi
7r
to,
q(u)duds :::;
-~+~
fi+4 II
Vi tY2" y-;
~it
[-~+4
IIjds
t
JS
y-;
7r/2
and hence
11tl"
to
to
q(u)duds <
00,
limsup :
t-+oo
lt
to
(t - s)n-2
[(t _ S)2p(S)q(s) _
(k )] ds
4 v s
00
(4.2.22)
then
Chapter 4
308
T. Define w(t)
~ du
J f'(X(t)) (')0
.ix(t)
2 A for t 2 T,
f'(x(t))
[1
A2
00
x(t)
A2 [B
y7'(Uj duj-2 = A2 [B
f(u)
A2 [B-lt
rt
(4.2.23)
_l
X
(t)
x(T)
y7'(Uj duj-2
f(u)
f~~~~))Jf'(X(S))dSr2
+ 1.iT f~x(:)) J
I(
f'(x(s))ds
I] -2
f'(x(t) 2
for all
c/~(t)
P( t ) q( t ) < - w t
+ (3
2 T*
(4.2.24)
ma.x{To, T},
()
x' (t)
t f(x(t
2( )
- a(t)p(t)~(t) w t
for all
2T ,
or
p(t)q(t) ~ - w'(t)
Thus, for t 2 T*,
rt
x'(t)
+ (3(t) f(x(t
rt
- cv(t)w 2(t)
for all
t 2 T*. (4.2.25)
rt
X'(S)
.irY - stp(s)q(s)ds ~ - .iT 5t - s)nw'(s)ds + .iT,(t - s)n{3(s) f(x(s ds
(4.2.26)
309
t Y - s)n-l [fT,(3(u)
t f(x(u))
x'(u)]
du ds
n
x'(s)
fTY - s) (3(s) f(x(s)) ds
n fT
:::;
Kn {t (t _ s)n-1ds
fT'
K(t _ T*)n.
fT'
(t - s)np(s)q(s)ds
h-
+ KJ
(t - T*)n[w(T*)
- t
iT-
iT-
fT-
c(t - stv(s)w2(s)ds
+ KJ + -n
(t - T*)n[w(T*)
_t
- n
4c
[JC(t _ s)nv(s)w(s)
h-
(t - s)n-1w(s)ds
!,t (t - s)n-2_(-)ds
1
T'
n(t - s)n-l
2Jc(t - s)nv(s)
]2 ds
1
< (t - T*)n[w(T*) + KJ + -n21t (t - s)n-2_(
)ds.
4c
T'
Now, since
n1
t
it
to
<
<
(t - s)np(s)q(s)ds
( t)~
1-
+~
tn
(n2)
4c
fto
fT"
(t _ s)n-2_1_ds
v(s)
for all
2': T*
(4.2.27)
it follows that
limsup
{t
t-+oo t fto
[(t _ s)np(s)q(s) _
<
T-
fto
n 2 (t _ s)n-2_(1)] ds
4c
v S
00,
Chapter 4
310
which contradicts condition (4.2.22).
.ltot a(s)p(s)~(s)ds
it
to
<
(4.2.28)
00
(t - s)np(s)q(s)ds
(4.2.29)
00
it
1
lim sup -;;(t - s)nq(s)ds
t-+oo t to
(4.2.30)
00
(rix'(t)), +
~rx'(t) +
[-r! sint +
~t-%(2 + cost)]
for
f(x(t)) = 0
to
= 7r/2,
(4.2.31)
where f(x) is as in Examples 4.2.1- 4.2.3. Let p(t) = t 5 / 6 . Then, we have
j3(t) = a(t)p'(t) - p(t)p(t) = 1/(6t), t ~ to and l/v(t) = a(t)p(t)~(t) =
t - to, t ~ to also for all t ~ to,
rtp(s)q(s)ds =
to
1,,/2
t s5/6[-s-1/3sins+~s-4/3(2+COSS)]dS
J"/2
[-v'sSins+ 2 \;;(2+COSS)]dS
Y s
rt d[v's(2
1,,/2
+ coss)]
vt - 2f%.
> -
00.
311
it [
to
(t - s) 2 s 5/6 q( s) - - C ] ds
v(s)
t~
>
(1:(t-s)
(it
[1:
(t-s)
1f/2
[,;8-2 VEl'2 J
~Vt _ 2 E + ~ (~)3/2 ~
V'2
15
t2
dS_C(t-7r/2)2)
2
_.: (1 _ ~)2
2t
and consequently,
lim sup
t400
~
t
ito
00.
V S
Thus, condition (4.2.22) holds for every constant c> 0 and n = 2. Now,
Theorem 4.2.2 ensures that (4.2.31) is oscillatory.
We note that Corollary 4.2.5 is also applicable to equation (4.2.31).
x"(t)
[_C 1/ 3 sin t
+ ~C4/3(2 + cost)]
t;::: to
7r/2,
(4.2.32)
ito
q(s)ds
i
i
[_S-1/3 sin s
n/2
<
(_s-1/3 sins +
1f/2
< C 1/ 3 cos t
C 1 / 3 cost
~s-4/3) ds
2
it [~S-4/3
+ -11it
t- 1/ 3 cos t +
n /2
11
-"2
ds
cos S
+ ~S-4/3]
2
ds
S-4/3ds
7[/2
[C
1/ 3 -
(7r)-1/3]
'2
<
121
(7r'2)-1/3
Chapter 4
312
~ i t (t -
to
7r
00,
(1 _ ~)
.
2t
fails,
Theorem 4.2.3. Suppose (F2) holds and let the functions p(t), (3(t), f,(t)
and v(t) be as in Theorems 4.2.1 and 4.2.2 such that condition (4.2.3) is
satisfied. If condition (4.2.5) holds and for every constant c> 0,
limsup
t-+oo
to
n)2] ds
=
00
(4.2.33)
Proof.
t (t iT'
s)np(s)q(s)ds
< (t - T*)nw(T*) +
- iTo
t c(t ft
(t - s)n-l [(t - s)
arsi;~s) -
n] w(s)ds
s)nv(s)w 2 (s)ds
(t - T*tw(T*)
- iT'
iT'
Jc(t - s)nv(s)w(s) -
_c_ ds
4v(s)
(t-s)n-l(t-S)~)l2
a(s)p(s)
2Jc(t _ s)nv(s)
[(t-s)aCi;~s)
ds
-nf 4v~s)ds.
313
it
.
1
lnllsup-;;
t--+oo
to
(t-s)
n-2 [
(3( s)
]2 1
(t-s) () () -n - ()ds < 00
asp s
v s
(4.2.34)
(t 1 / 6 X'(t))' +
6~x'(t)+ [-C1/3sint+~C4/3(2+Cost)]
f(x(t)) = 0
to = 1f/2,
(4.2.35)
where f is as in Example 4.2.6. Let p(t) = t 5 / 6 , t ~ 7r /2. Then, we
have (3(t) = a(t)p'(t) - p(t)p(t) = (5 - C 1 / 6 ) /6 > 0 for t ~ 1f/2 and
for
(3'(t) = C
7/ 6
J:/2ds/s
-1
t2
it [
to
t-s
1
="2
t
Tr/2
(3(s) -2
a(s)p(s)
[
]2 -1d s
1
6s
v(s)
-(t-s)(5-s
2s
/)-2 ] 2 sln-ds
1f
> 6ln 2t ,
-1 6'
1f
(4.2.36)
~
t2
it
S5/6 q (s)ds
Tr /2
>
-
!i.,;t _ 2 E + ~ (~)3/2 ~.
15
V"2
t2
(4.2.37)
Now, from (4.2.36) and (4.2.37) it follows that the conditions of Theorem
4.2.3 are satisfied, and hence equation (4.2.35) is oscillatory.
Next, we present the following oscillation criteria for (4.1.3) when condition (4.2.22) fails for some integer n > 1.
Theorem 4.2.4.
p(t), (3(t) ,
~(t)
iT
for every T
O(T) (4.2.38)
to, and
00,
(4.2.39)
Chapter 4
314
where n+(t)
= max{n(t), O},
Proof.
Case 1. l(T, 00) < 00. Proceeding as in Theorem 4.2.2 - Case 1, we obtain
(4.2.26). The rest of the proof is exactly the same as that of Theorem 2.5.12
for H(t, s) = (t - s)n, n> 1 and hence omitted.
Case 2. l(T,oo) = 00. The proof is similar to that of Theorem 4.2.1 Case 2.
Theorem 4.2.5.
pet), (3(t),
~(t)
liminf -1
t .... oo
tn
it
to
(t - s)np(s)q(s)ds < 00
T::::: to,
for every
(4.2.40)
Proof. The proof is similar to that of Theorem 4.2.4 except that now we
employ Theorem 2.5.13 instead of Theorem 2.5.12.
In Theorems 4.2.4 and 4.2.5 if condition (4.2.4) fails, then we have the
following results.
Theorem 4.2.6.
pet), (3(t),
li~~p t n
~(t)
c (
(3( s)
iTrt(t_s)n-2 [(t-s)2 p(s)q(s)- 4v(s)
(t-s) a(s)p(s) -
2] ds
::::: neT)
for every
T::::: to,
Theorem 4.2.7.
pet), (3(t) ,
~(t)
315
(4.2.3) and (4.2.5) are satisfied. If there exist an integer n > 1, and a
function O(t) E C([to, 00), JR) such that conditions (4.2.39) and (4.2.40)
hold, and for every constant c > 0,
1ft(t-st- 2 f(t-S)2p(S)q(s) -
liminf - n
t-+oo t
c ( (t-s) (j3(s)
-(-)
) ( ) - n
4v s
asp s
)2] ds
;::: O(T)
for every T;::: to, then equation (4.1.3) is oscillatory.
Example 4.2.9. Consider the differential equatioll
+ (cost)f(x(t))
(e-tx'(t))'
t;::: to>
0,
(4.2.41 )
where
)]-l
[1- e-(t-t o
1
liminf2"
t
t-+oo
and
to
t > to> 0,
(t-s)2cos s ds
liminf
t-+oo
> - 00,
to
1ft [(t-s)2coss-c(1-e-(t-to))]ds;:::
liminf2"
t-+oo
1t
for
l'
-sinT-K
for every T;::: to, where c and K are positive constants and K is
sufficiently small. Let O(T) = - sin T - K. Next, we consider an integer
N such that (2N + 1)71" + (71"/4) > to. Then for all integers n;::: Nand
(2n + 1)71" + (71"/4) ::; T::; 2(n + 1)71" - (71"/4), O(T) = - sinT - K ;::: 5T,
where 5 is a small constant. Thus, we have
lim
t-+oo
1
t
v(s)(O+(s))2ds;:::
L
00
52
n=N
to
2 (n+l)-(n/4)
s2ds
00.
(2n+1)n+(n/4)
Hence, all conditions of Theorem 4.2.5 are satisfied and therefore equation
(4.2.41) is oscillatory.
Theorem 4.2.8. Assume that condition (4.1.5) holds,
min{~~~f'(X)
> 1
( 4.2.42)
and suppose there exists p(t) E C1([to, 00), JR+) such that condition (4.2.3)
holds and a(t)p'(t)
- 00
(4.2.43)
Chapter 4
316
and
lim F(t)
t-'>oo
where
F(t)
1
-()
~ t
it
to
1
8
( )1 ( )
asp s
to
(4.2.44)
p(u)q(u)duds,
(4.2.45)
t-'>oo
t:::- to.
(4.2.46)
w' (t)
Cl
( )t+a ()
a (t )p
t p(t )
where
J(to, t) =
and
Cl
= a(to)p(to)w'(to),
w(t)
~(t)
w(t o )
1:
~ t
it
to
to
p(s)q(s)ds + asp
() (s /(t o, t),
a(s)p(s)(w'(s))2 j'(x(s))ds
and
1
~(t) + Cl + ~(t)
1
+-()
it
it
to
1
8
a(s)p(s)
to
1
-()-() J(to, s)ds.
p(u)q(u)duds
(4.2.47)
asp s
J(to, t)
lim w(t) =
Hoo ~(t)
Let
o.
(4.2.48)
4.
(4.2.49)
317
where
G(tJ, t)
itl
a(s)p(s)f'(x(s)) ds
(4.2.50)
If G(t1' (0) <
is bounded on
that G(tJ, (0)
[t1' (0)
(4.2.51 )
But condition (4.2.42) guarantees that
f'(x(t))w(t) = f'(x(i))
r= jd(U)
ix(t)
>
for
2: il'
w(t)
<
y'E
a(t)p(t) - a(t)p(t)
or
[i
[titl
w(s) d]
a(s)p(s) s
]/2
w(s)
] -1/2 w(t)
a(s)p(s) ds
a(t)p(t)::;
t,
VE
for all
t::::: t2,
for all
(4.2.52)
t 2: t 2
1,
t
w(s) d] 1/2 2
a(s)p(s) s
-
<
VE [~(t)
[lt za(s)p(s)
w(s) d]
s
1/2
t,
-l:z a(s)~(s)dS]
<
VE ~(t).
Therefore, on setting
t3 =
max {t2'
~
y'E
[ltz a(s)p(s)
w(s) dS]
t,
1/2}
Chapter 4
318
for all t
t3, we obtain
[i
()
WS
a(s)p(s)
t,
] 1/2
it
lim 1:(1)
t-+oo <,
to
- C1 -
()1 ( )
asp s
1
8
1
lim - ()
t-+oo ~
p(u)q(u)duds
it
to
to
( ) ( /(ta,s)ds,
asp
s
It:
00.
>
(4.2.53)
t-+oo
where B(t) =
J(ta, s)/(a(s)p(sds. Assume that (4.2.53) does not
hold, then there exists a T1 ~ ta such that
B(t)
{Lw(t)
for
t ~ T 1.
(4.2.54)
(W'(t2f'(X(t)1/2 >
w'(t)
B(t)
- {L1 wet)
(4.2.55)
= J(c/{L) > 1.
a(t)p(t)B'(t) )) 1/2
( In ( a(T
(lnB(t
2 )p(T2 )B'(T2)
wet)
1/2
(4.2.56)
wet)
..,fiilln w(T2)
..,fiil[lnB(t) -lnw(T2)]
..,fiillnB(t) for t
T 2.
B-ILI (t)B'(t)
>!
-
319
(4.2.57)
k a(t)p(t)
lIlt
> - - [B l -IL1(T2 )
/11 - 1
Bl-~'l(t)l ?: -k
,
T2
( )1 ( )ds
asp s
---t
00
as t ---t
00,
which is a contradiction. Thus, the assertion (4.2.53) holds, and there exists
a sequence {tn}~=l with limn--+oo tn = 00 and such that
(4.2.58)
Using (4.2.58) in (4.2.47), we find for sufficiently large tn,
( -1
1)
+ P, ~(tn) B(tn)
tn a(s)p(s)
1
r
lto p(u)q(u)duds.
B(t)/~(t) ---t 00
l.t
liminf -1()
~
Cl
(4.2.59)
t--+=
w(to)
?:
( )1 ()
asp s
to
as t ---t
1
8
to
(4.1.59) implies
00,
p(u)q(u)duds
- 00,
6(t)
It:
and
1t 1
1t 1
to
1
lim -(-)
t--+=
6 t
1
-()
a s
to
to
1
-()
a s
q(u)duds > -
to
00
(4.2.60)
q(u)duds
< liminf!
t--+=
1t 1
8
to
to
1t 1
8
to
to
q(u)duds, (4.2.61)
320
Chapter 4
.;
it
to
(is p(u)q(u)du )2 ds =
( )1 ( )
asp
to
00.
(4.2.62)
Case 1.
l~
p(s)q(s)ds = Kl
+ a(t)p(t)w'(t) -
ho
a(s)p(s)(w'(s2 f'(x(sds,
K2
3K~ + 3
(L
OO
a(s)p(s)(w'(s))2 f'(X(S)dS) 2
then, we have
(1:
p(s)q(S)dS) 2 :s;
3K~ + 3(a(t)p(t)w'(t2
< K2 + ';(t)
< K2 + .;tt)
1:
lto
ds
a(s)p(s)(w'(s2ds
a(s)p(s)(w'(s2 f'(x(sw(s)ds
i.e.,
to~sst
lto
321
t::::
max w(s) S;
max w(s)
to:S,:ST
to:S'$!
for all
t:::: T.
t:::: T,
Thus, for
+ ~(t)
(4.2.63) gives
ds
S;
[to~;~Tw(s) + ~(t)] .
K2 + ;:)
Therefore, we have
lim sup _(1)
t-+oo
~ t
it ()\) (1'
to
asp s
to
P(U)q(U)dU)2 ds
+ K3 <
K2
S;
Corollary 4.2.11.
by
it
lim sup (: 1( )
t-+oo
00,
<,,1
_(1)
a s
to
(1' q(U)dU)
to
ds =
00.
Corollary 4.2.12. If
11 l'
! it (f"
t
liminf t-+oo
and
lim sup
t-+oo
to
to
to
lto
q(u)duds > -
q(U)dU) ds
00
00,
(4.2.64)
Theorem 4.2.10.
by
liminf -1
t-+oo
tn
it
to
(t - stp(s)q(s)ds
>
-00,
(4.2.65)
Proof.
t::::
322
Cha.pter 4
c = infx>o f'(x) ixOO du/ feu). In view of condition (4.2.42), we have c> l.
Hence, we can choose an integer k;:: max{n,2} such that
k
- - < c.
(4.2.66)
k-1
rt (t - s)kp(s)q(s)ds
liminf 1k
t--too
.lto
> -
00.
(4.2.67)
+ it(t -
i t (t - s)kp(s)q(s)ds
to
s)ka(s)p(s)(w'(s))2 f'(x(s))ds
to
i t (t - s)k(a(s)p(s)w'(s))'ds
to
- (t - to)ka(tO)p(to)u-'(to)
+ kit (t -
s)k-1a(s)p(s)w'(s)ds.
to
to
-1:
(t - s)ka(s)p(s)(w'(s))2 f'(X(S))dS] .
(4.2.68)
~
c
(_k_)
k-l
<
J.L
< l.
(4.2.69)
We claim that for every T;:: to, there exists at;:: T such that
Otherwise, there is a T
t (t - s)ka(s)p(s)(w'(s))2 f'(x(s))ds
.lto
:s; '5..
T,
t (t - s)k-1a(s)p(s)w'(s)ds .
J.L .lto
(4.2.70)
323
0::::
t (t -
ltD
s)k- 1a(s)p(s)w'(s)ds
k
]1/2 [1
]1/2
: : [ "ML
(t-s)k- 1a(s)p(s)w'(s)ds
~lo (t-s)k- 2a(s)p(s)w(s)ds
t
t (t - s)k- 1a(s)p(s)w'(s)ds
ho
t::::: T,
1.
::::
~~
we have
1,
+ -_1k
1
Using the fact that (a(t)p(t))':::: 0 for t::::: to in the above inequality, we
get
rt (t _ s)k- 2a(s)p(s)w(s)ds
::::
+-k
_1
1
[ftC (k ~
1) - 1] tk~l 1:
[ftC (k ~
k (
:::: "M 1 - t;
But, J(to, 00)
= 00
1
lim k-1
t~oo
)k-1
(t - s)ka(s)p(s)(w'(s))2 f'(x(s))ds
(4.2.72)
a(to)p(to)w(to)
for
t::::: T.
implies
00.
Chapter 4
324
On the other hand, (4.2.69) implies that ILC((k - l)/k) - 1 > O. Thus,
(4.2.72) leads to a contradiction. Hence, the claim follows and so we can
consider a sequence {tm}~=l with limm-too tm = 00 such that
tm
to
>
tm
IL to
(tm - sl-la(s)p(s)w'(s)ds
(m
= 1,2", .).
tm
tm
(t m
s)kp(s)q(s)ds < -
t)k a(to)p(to)w'(to)
1 - -.2..
tm
tm
l)T
(tm - s)ka(s)p(s)(w'(s)f j'(x(s))ds,
1
tm to
m= 1,2,.
(4.2.73)
It
lim ~
(t - s)ka(s)p(s)(w'(s))2 j'(x(s))ds
t-too t to
= J(to, 00) =
00
and by (4.2.69) the constant IL < 1, the inequality (4.2.73) ensures that
=
It
to
(t - s)kp(s)q(s)ds
= -
- 00.
00,
liminf -1
t-too t n
It
to
(t - s)nq(s)ds > -
00
n> 1, (4.2.74)
lItiS
lim q(u)duds
t-too t to to
325
lim sup
t-->oo
~n rt (t - s)n-2a(s)p(s)~(s)ds
t lto
and
1
lim sup --;-
t-->=
it
(t - s
to
<
p( s )q( s )ds
00
(4.2.75)
00,
(4.2.76)
Proof.
w(t) -::
C1~(t)
for
t 2': T.
(4.2.77)
y'(t) = p(t)q(t)
+ a(t)lp(t) y2(t)j'(X(t)),
t 2': T.
(4.2.78)
+ c1a(t)~(t)~(t) y2(t)
for
t 2': T.
(4.2.79)
Chapter 4
326
it
T
(t - s)np(s)q(s)ds
- n
it
or
t [(
(t-s)n
C1a(s)p(s)~(s)
+ -C11
it
T
(t - s)n-1y(s)ds
)1/2 yes) -
f\t-s)np(s)q(s)ds - n 2C1
iT
(t - st ( ) ( ) ( ) y2(s)ds
asp s ~ s
:s: - (t -
T)ny(T),
n(t_s)(n-2)/2]2
2(C1a(s)p(s)~(s-1/2
ds
iT
tn
iT
4t n
iT
(4.2.80)
Taking limsup on both sides of (4.2.80) as t -+ 00 and using (4.2.75), we
obtain a contradiction to condition (4.2.76). This completes the proof.
Corollary 4.2.16.
(4.2.74) holds and
lim sup -1
t .... oo t n
it
to
(t - stq(s)ds =
00
n > 1,
(4.2.81)
1. We note that (4.2.16) implies the first part of (4.2.61) and inductively
condition (4.2.74).
2. The condition limt-too
(4.2.17) implies (4.2.61).
It: q(s)ds =
00
327
l' (x),
i.e., the
Theorem 4.2.13. Suppose conditions (F2) and (4.1.9) hold and Q(t) =
oo q(s)ds exists, and liminfHoo
Q(s)/a(s)ds > -00, and either
(Q+(s))2/(a(s)6(s))ds = 00, or for every constant c> 0,
It
It";
It:
lim sup
t ..... oo
where 6(t)
00
to
ass
Itto ds/a(s),
(~(U)(Q+(U))2 du jdS
_(1) [Q(s)+cjoo
al 1 ~l
00,
Proof.
1'(x(t)) 2
~;tt)
for
00
Cl
t2
(4.2.82)
T.
a(s) ( x'es)
6(s) f(x(s))
)2 ds for t 2 T.
ho(t)
00
y'a(t)6(t)
1
= 1
00
q(s)ds,
6(t)
to
ds
a(s)'
([ho(s)]+)2 ds
00
([hO(S)
y!a(s)~l(S)
hn(S)]
+)
ds
for n
= 1,2,
328
Chapter 4
0,1, ... ,N. If for ever constant c > 0, and for all sufficiently large t,
ho(t) + chN(t)/ a(t)6(t);::: 0 and
/ 00
y'a(s)6(s)
[ho(S)
y'a(s)6(s)
hN(S)] ds
=,
Proof.
It
It
where Q-(t)
= Q+(t) - Q(t).
ds <
=,
(4.3.3)
Then,
(4.3.4)
329
sup f(u),
bC:;uC:;c
Al
sup f'(u)
(4.3.5)
bC:;uC:;c
(4.3.6)
Now, we choose T > 0 so large that
(4.3.7)
for all s, t:2: T,
and
lXo P(s)ds
:::;
(4.3.8)
E.
Ilx'(t 2 )
x'(tl)11 :::; A
11:
q(s)dSI
+ it 2
tIl
(Tnx)(t) =
0'
-1 1
0' -
00
DO
00
q(u)f(x(u))duds
JT+n s
for
T:::;
DO
q(u)f(x(u))duds
for
t:::;
+n
+ n:::; t.
330
Chapter 4
(Tnx)'(t) =
00
q(s)f(x(s))ds = f(x(t))Q(t)+
00
I(Tnx)'(t)1
< AIQ(t)1 +
< AIQ(t)1 +
~B
(1
00
00
f'(x(s))Q(s)x'(s)ds.
1
(1
00
AlIQ(s)1 [AIQ(s)1 +B
00
Q2(s)ds + AlB
00
Q2(S)dS)
-s; P(t).
IQ(s)ldS)
I(Tnx)'(t2) - (Tnx)'(tl)1
jI(x(t d )
11:
< A
1:
11:
1:
+ n,
we have
q(s)f(x(s))dSI
q(s)dsl
Q2(U)dU] ds
AlP(s)
11
00
(l t2 q(U)dU) dsl
q(U)dul ds.
ITnx(t) - 01
-s;
+ n,
n
+ [AIQ(s)1
we find
11 1
00
+B
00
q(U)f(x(u))dUdSI
1 Q2(U)d~
00
ds
+ l:/(x[T+n])
11
00
q(u)dul ds.
Thus, by (4.3.8), we get ITnx(t) - 01 -s; ftOO P(s)ds -s; E. Since Tnx(t) is
constant for t 2 T + n, this shows that Tn(Xn) C X n .
331
Ilx - yll
ao(t5)
=
b
<
sup
u,
'tI
<
Tiu - viT:'O .s
<
sup
u, u
<
a.
+1':
1f'(u)-f'(v)l
iTu - vil:'O .s
Then, we have
I(Tnx)' - (Tny)'I(t)
00
11 [J(x(s)) - J(y(s))]q(s)dsl
IJ(x(t)) - J(y(t))IIQ(t)1
+ 11[J'(x(s))x'(s)- J'(y(s))Y'(s)]Q(s)dsl
IJ(x(t)) - J(y(t))IIQ(t)i
ao(t5)IQ(t)1
ao(t5)IQ(t)1
+ (B
00
IJ'(x(s)) - f'(y(s))lx'(s)IIQ(s)lds
00
+A 1 t5 ( AIQ(s)1
00
[a 1 (8) ( AIQ(s)1
+B
00
(1=
00
Q2(U)dU)
Q2(U)dU)] IQ(s)lds
+ (al (8) + 8A 1 )
Q2(S)dS)
+B
[A
00
Q2(s)ds
IQ(s)ldS) ]
00
Q2(s)ds
b(8)Q(t),
ITnx - TnYI
b(t5)
00
00
P(s)ds
+ b1 (8)
00
P(s)ds
+ ITnx -
TnYI(T + n)
+ b1 (8)),
E(b(8)
say,
where b1 (8) has the same property as b(8), and so IITnx - TnYl1 ~
(b(t5) + b1 (8))(1 + E). Hence, it follows that Tn is continuous. Now, we
can apply the Schauder fixed point theorem to obtain a function Xn E Xn
such that Tnxn = x n . Thus, for T ~ t < T+n, x~(t)+q(t)J(xn(t)) = O.
The sequence {x n } is equicontinuous on [T, 00 ). It follows by a standard
diagonal argument that there is a subsequence {x nk } which converges uniformly on compact intervals to a function x which is nontrivial, bounded
nonoscillatory solution on [T, (0) of equation (4.1.1).
Chapter 4
332
We can combine Corollary 4.1.4 and Theorem 4.3.1 to obtain the following necessary and sufficient condition for the oscillation of (4.1.1).
It
Corollary 4.3.1. Let Q(t) = OO q(s)ds exist with Q(t) 20 for t 2 to,
and let f be strongly nonlinear, i.e., f satisfies the conditions
C 1(lR,lR), xf(x)
(f1)
(f2)
(a1(t)x'(t))'
+ q1(t)X(t)
(a(t)x'(t))' + q(t)x(t)
0,
(4.3.9)
(4.3.10)
= 1,
and let
Q(t)
Q1(t)
00
ql(s)ds,
00
q(s)ds
(4.3.11)
exist with 0 ~ Ql(t) ~ Q(t) for all t in [to, (0). If equation (4.3.9) is
oscillatory, then (4.3.10) is also oscillatory.
This result has been extended to the following form:
333
Theorem 4.3.3. Let a(t) be bounded above on [to, 00). Let Q1 (t), Q(t)
in (4.3.11) exist and let 0 < a(t) ::; a1(t), 0::; IQ1(t)1 ::; Q(t) for all t in
[to, 00). If equation (4.3.9) is oscillatory, then (4.3.10) is also oscillatory.
(a(t)x'(t))'
+ q(t)f(x(t)) = o.
(4.3.13)
Theorem 4.3.4. Let a, aI, q, q1 be continuous 011 [to, 00) such that
Q(t) and Q1(t) defined in (4.3.11) exist and such that 0 < a(t) ::; a1(t),
IQ1(t)1 ::; Q(t) for all t E [to, 00). Assume that the function f satisfies
(i)
or
(iii) liminf 1xl -4oo f'(x) > 0 and condition (4.1.5) holds.
If equation (4.3.12) is oscillatory, then (4.3.13) is also oscillatory.
Proof. It will be convenient to separate the proof into the following three
cases:
(I)
(II)
If I
w(t)
w(T)
iT iT
t
q(s)ds +
a(s) w 2 (s)f'(x(s))ds.
(4.3.14)
Letting T -+ 00 in (4.3.14), the second term on the right-hand side has the
nonnegative limit Q(t), whereas the third term has either a nonnegative
limit, or tends to 00. Hence, it follows that limT-4oo w(T) = (3, where
- 00 ::; (3 < 00.
We shall show that 0::; (3 < 00. Suppose on the contrary that - 00 ::;
(3 < O. Then for sufficiently large t, say, t?: l, we have a(t)x'(t) < O.
If for some E, a(t)x'(t)::; -E < 0 for t?: t* ?: l, then it follows
that x(t) - x(t*) ::; -E
ds/a(s) -+ -00 as t -+ 00, which contradicts
Itt.
Chapter 4
334
(l
= V(t) -
tn
tn
q(s)f(x(sds >
(1\(U)dU )f(X(S
,, q(s)f(x(sds
(x'(;~~'(~~~s
)dS
,, H(s)V(s)ds,
where
V(t) =
(l tn q(S)dS) f(x(t
and
Let W(t) = Ittn H(s)V(s)ds. Then the inequality V(t) < Ittn H(s)V(s)ds
gives W'(t) = -H(t)V(t) > -H(t)W(t) from which it follows that
and hence 0 = W(tn) exp (Itt,. H(s)ds) > W(t) exp (I; H(s)ds) , which
implies that W(t) and thus V(t) is negative on [f, tn). But this in
turn implies that Ittn q( s )ds < 0 for f::::; t < tn for n sufficiently
large, contradicting that 0::::; Q(t) = Itoo q(s)ds. Hence, on taking limit in
(4.3.14), we find
w(t)
where 0::::; (3 <
have
00
(3 + Q(t)
00
a(s) w 2 (s)f'(x(sds,
(t) ~ _
f(u) -
lto
(4.3.15)
= w(t)la(t),
we
(w(s) ds
a(s)
,
where 0: = x(to) > O. Denoting I:(t) dulf(u) by n(x) for x> 0 and
the inverse function of n by r, we get
335
Conditions (i) and (ii) imply that r is monotone increasing from some
interval (-oo,c) to lR+ (c maybe +00) and that we can write (4.3.16)
as
w = M(w),
(4.3.17)
maps the set C = {u E C([to,oo),lR+) with O:'S: u(t) :'S:
[to,oo)} into itself. lVI is monotone in the sense that lVIu < lVIv
for any u, v E C with u < v under the na.tural ordering of C. Now
where M
w(t), t
z = L(z).
(4.3.19)
Since IQI(t)l:'S: Q(t) and 0 < a(t):'S: al(t) on [to, 00), we find that L
is a self map on B = {u E C([to, 00), lR+) with QI(t):'S: u(t) :'S: z(t), t E
[to,oo)}. Clearly, B is a closed convex subset of C = C([to, 00), lR)
the Fnchet space of continuous real-valued functions on [to, 00) with the
compact-open topology. Let z E B. Then, we have
< t :'S: T,
t E [to, 00)
ILz(t)l:'S: w(t),
ilt
ql(u)dui +
ilt
(4.3.20)
From (4.3.20) and (4.3.21) it follows that the functions in the image L(B)
of B are uniformly bounded a.nd equicontinuous on compact subintervals
of [to, 00). We claim that L is continuous on B. For this, for z E C
define <1>( z; s) by
<1>(z;s) =
(Z2(S))
f' (r [1 (~) dU])
al(S)
to
al(U)
8
so that
(Lz)(t)
(J+QI(t)+
Zn ---}
00
<1>(z;s)ds.
(4.3.22)
= [to, tIl,
(4.3.23)
Chapter 4
336
I(Lz)(t) - (Lzn)(t) 1
-3E
iTl
+~ <
3
E,
d2 y
ds 2
+ a(t(s))q(t(s))f(y(s)) =
= x(t(s)) trans-
(4.3.24)
O.
The oscillation properties of the equation are invariant under this transformation. Let r(s) = a(t(s))q(t(s)), and use the substitution u = t(r), to
obtain
1=
{=
l= (1=
R(s)
r(r)dr
iT
1=
(=
a(t(r))q(t(r))dr
R(s)ds =
iT
R 2 (r)dr) ds
Q(t(s))ds =
1=
1=
t(T)
q(u)du
t(s)
Q(t(s)) > 0,
(4.3.25)
(Q((u))) du,
au
(4.3.26)
It follows from Corollary 4.3.1 and (4.3.24) - (4.3.27) that equation (4.3.13)
is oscillatory if and only if
J=
1 [ Q(u)
a(u)
1= (Q
2
(7])) d7] ] du =
a(7])
00.
(4.3.28)
al(u) IQI(U)I +
1 (Qi(7]))
00
al(7])
d7] du =
00.
(4.3.29)
:s:
o:s:
337
:s:
Q1(t) imply that (4.3.12) has a nonoscillatory solution. For this, we use
the transformation s = g(t) = (ftOO dT/adT)) -1, y(s) = sx(t). Now, if
we denote g( to) by So and the inverse function of 9 by h, then this
transformation changes (4.3.12) into
d2 y2
ds
where rds)
+ r1 (s )f (y(s))
-_
s
0,
SE[to,::x:J)
(4.3.30)
o:s:
f'(x):S: K
whenever
0:S:x:S:2
(4.3.31 )
(4.3.32)
Consider the integral equation
(4.3.33)
which we write as y
= F(y). Let
(4.3,34)
Next, we denote [To, 00) by J and let X to be the Banach space
of bounded continuously differentiable functions y on J such that ty'
is bounded on J In the space X we introduce the norm as Ilyll =
IIYlloo + Ilty'lloo, where I 1100 denotes the supremum on J
Let 13 be the closed ball in X with center 1 and radius 1/2. We shall
show that F maps 13 continuously into a compact subset of 13, so that
the Schauder fixed point theorem can be used to guarantee the existence
of y* E 13 such that F(y*) = y*. The function y*(s) will then be a
Cbapter 4
338
Noting that 0::::; y(s)/s < 2/s < 2 and using (4.3.31)
and so
I[F(y)](s)
Q*
(4.3.34), we get
(4.3.37)
From (4.3.36) and (4.3.37), we find that F(y) E B and hence F maps
y, z E B with lIy - zll : : ; E. We have
<
11
I[F(y)]'(s) - [F(z)]'(s) I =
00
rl
(T)
00
(4.3.38)
where
U(T)
f'
Kf
S
(4.3.39)
Writing U(T) as
IU(T)I ::::;
(:2)'(~) +2K;2
<
:2C!(f) + Kf),
(4.3.40)
I[F(y)]'(s) - [F(z)]/(s)1
~
<
f'
339
22KQiE + 4Qi(r(E)
+ KE)
(1 d:)
00
S s T
~[Qi(2KE
+ ,(E))]
S
(4.3.41 )
and therefore
I[F(y)](s) - [F(z)](s)1
-[Qi(2KE + ,(E))].
(4.3.42)
I[F(y)]'(s) - [F(Y)]/(T)I =
lis
Tl(/L)f
(Y~)) dTI ~
2K
is
!rI(/L)ld/L.
(4.3.43)
From (4.3.43), it follows that the image F(13) of 13 under F consists
uniformly bounded functions on J, with uniformly bounded derivatives on
J that are equicontinuous on compact subsets of 1. Thus, we can apply
the Ascoli-Arzela theorem to deduce that F(13) has a compact closure
in X. Now, the Schamler fixed-point theorem completes the proof of case
(III).
(a(t)y/(t))/
+ kq(t)y(t)
(4.3.44)
Chapter 4
340
is oscillatory, then equation (4.1.2) is oscillatory.
Proof.
Riccati equation
w'(t) = q(t)
+ pet) w 2(t)
(p(t)y'(t))'
+ q(t)y(t) =
0,
t;::: O.
(4.3.45)
(4.3.46)
is oscillatory, where
B(t)
and vet)
p(t)q(t)
+k
f3(t)
)'
k ( f3(t) )2]
[( 2a(t)p(t)v(t)
- 4v(t) a(t)p(t)
l/(a(t)p(t)~(t)), t ;:::
Case 1. l(T,oo) < 00. As in Theorem 4.2.2, we obtain (4.2.25) which takes
the form
+ a~)(;~t) wet) -
ev(t)w 2 (t),
t;::: T.
Let z(t)
341
(4.3.47)
(vtt)Y'(t)), +kp(t)q(t)y(t) = 0
is oscillatory for every constant k > 0,
then equation (4.1.3) is oscillatory.
where v(t)
l/(a(t)p(t)~(t))
Proof.
f'(x(t)) ?:
~~t)
for
(4.3.48)
t?: Tl ?: to,
)'
( a(t)p(t),
f'(x(t)) Y (t) + p(t)q(t)y(t) = 0
(4.3.49)
is nonoscillatory. Now, by (4.3.48) and as an application of the PiconeSturm comparison theorem to (4.3.49), we conclude that the equation
(y'(t)/v(t))' + cp(t)q(t)y(t) = 0 is nonoscillatory, which contradicts the
given hypothesis.
Remark 4.3.2.
(a(t)lx'(t)I"-lX'(t))'
+ q(t)f(x(t))
0,
(4.4.1)
Cha.pter 4
342
where
(i) O! is a positive constant,
(ii) aCt) E C([to,oo),IR+) and q(t) E C([to, 00), IR),
(iii) f E C(IR, IR), f is non-decreasing and xf(x) > 0 for x
-I- O.
The results we shall prove extend Lemmas 4.1.2 and 4.1.3 and Theorems
4.1.7 - 4.1.9. For this, we need the following lemma.
Lemma 4.4.1. Let K(t, s, x) : IR x IR X IR+ -* IR be such that for all
fixed t, s the function K(t, s,') is nondecreasing. Further, let pet) be
a given function and u(t), vet) satisfy for t::::: to,
u(t) :::::
(~) pet) +
and
vet) = pet)
it
K(t, s, u(s))ds
(4.4.2)
to
it
K(t,s,v(s))ds.
(4.4.3)
to
_ a(to)!x'(to)i",,-lx'(t) +
f(x(to))
it
q(s)ds +
to
it'
to
(4.4.4)
for
[iI, b].
(4.4.5)
w'(t)
f(x(t))
q(t),
(4.4.6)
( wet) )'
f(x(t))
w'(t)
f(x(t))
w(t)J'(x(t))x'(t)
J2(x(t))
!x'(t)!O+l J'(x(t))
-q(t) - aCt)
J2(x(t))
.
---
(4.4.7)
343
Let t E [tl' b]. Then on integrating (4.4.7) from to to t and then applying
(4.4.4), we find
rt
w(to)
- f(x(t o))
w(t)
f(x(t) )
---
(3 +
2:
1
t
tl
rt
a(s)
Ix'(s)ICt+1 f'(x(s))
J2(x(s))
ds
IX'(s)i<>+1 f'(x(s))
J2(x(s))
ds >
o.
(4.4.8)
Case 1. Suppose x(t) is positive. Then it follows from (4.4.8) that --w(t) >
0, or x'(t) <
for t E [tl, tl. Let u(t) = -w(t) = a(t)lx'(t)Ii3. Then,
(4.4.8) provides
2:
u(t)
(3f(x(t))
1t
+ 1t
+
tl
(3f(x(t))
(4.4.9)
K(t, s, u(s))ds,
tl
K(t,s,y)
is defined by
Ix'(s)If'(x(s))
J2(x(s)) y, t,s E [tl,b] and y E JR+. (4.4.10)
f(x(t))
u(t) 2: v*(t)
for
t E [tl, b],
(4.4.11)
v(t)
(3f(x(t))
(3f(x(t))
+
+
K(t, s, v(s))ds
itl
t f(x(t)) Ix'(s;If'(x(s))
ds
it
f (x(s))
(4.4.12)
( ~)'
f(x(t))
[(3
1t
tl
Ix'(s)If'(x(s)) ( )d ]'
J2(x(s)) V S S
Ix'(t)If'(x(t)) ()
J2(x(t)) v t .
(4.4.13)
v(t) )'
f(x(t))
f(x(t))
v'(t)
f(x(t))
_ v(t)f'(x(t))v'(t)
J2(x(t))
v(t)lx'(t)If'(x(t))
+
J2(x(t))
(4.4.14)
344
Cha.pter 4
u(t)
-f3f(x(t))
it[-
t,
-f3f(x(t)
1,t K(t,s,u(s)ds,
(4.4.15)
t E [tl,b]
v(t)
-f3f(x(t
-f3f(x(t
it
i
K(t, s, v(sds
t,
t
t, [- f(x(t]
[x'(s)[f'(x(s
j2(x(s
v(s)ds.
(4.4.16)
T/(t) =
Jto
a-1/0I.(s)ds
-00
and
<
and
lim T/(t)
t---+oo
roo q(s)ds
Jto
lim [f(x)[
Ix 1---+00
<
00,
(4.4.17)
(4.4.18)
00
(4.4.19)
00.
00
to
a(s)
.
[x'(s)["+l f'(x(s
j2(x(s
ds <
t~~ a(t)
[x'(t)["-lX'(t)
f(x(t
= 0
00,
(4.4.20)
(4.4.21)
345
and
a(t)
00
a(s)
. t
Ix'(s)la+l f'(x(s))
J2(x(s))
ds
[00 q(s)ds
t
(4.4.22)
for all sufficiently lage t.
[t a-l/a(s)ds --+
as
(X)
t --+
(X),
t,
to to t gives
a(t)
Ix'(t)la-lX'(t)
f(x(t))
( ) Ix'(to)l<>-lX'(tO)
a to
f(x(to))
-
to
it (
to
)d
IX'(s)l<>+l f'(x(s))
a(s)
J2(x(s))
ds.
(4.4.24)
In view of (4.4.18) and (4.4.20) it is clear from (4.4.24) that the limit
L
exists. Now, let t --+
a(t)
(X)
t~~ a(t)
Ix'(t)la-lx'(t)
f(x(t))
(4.4.25)
Ix'(t)la-lX'(t) _
f(x(t))
- L+
(4.4.26)
Therefore, (4.4.21) and (4.4.22) hold if L = O. For this, we consider the
following two cases:
Case 1. Suppose L < O. Then, (4.4.18) and (4.4.20) lead to
L
3
for
t 2 tl
(4.4.27)
Chapter 4
346
a(s)
Next, substituting t
= to
and
00
1 t,
a to
L
3
Ix'(s)I<>+1 f'(x(s)) I
J2(x(s))
ds <
(4.4.28)
in (4.4.26), we get
Ix'(to)l"-lX'(tO)
f (x (to) )
-1
00
to
(
)d
qS S
-1
00
to
as
Ix'(s)lo+ l f'(X(S))d
J2 (x (s ))
s.
(4.4.29)
Now, using (4.4.29) and the inequalities (4.4.27) and (4.4.28), we find for
t 2: tl,
-a(to)
Ix'(to)I<>-lX'(tO)
f(x(to))
_ L
-1
00
q(s)ds
2: - L
lt
-1
to
q(s)ds +
00
t,
ltl
to
a(s)
Ix'(s)I,,+l f'(x(s))
J2(x(s))
ds
(3
> 0,
i.e., (4.4.4) holds. Hence, we can apply Lemma 4.4.2 to obtain a contradiction as in the proof of (4.4.20).
Case 2. Suppose L > O. From the definition of L in (4.4.25), we can
assume that
L
Ix'(t)I<>-lX'(t)
a(t)
> - for t > t .
(4.4.30)
f(x(t))
From (4.4.30), we find that x'(t) > 0 for t 2: tl. Now, using (4.4.30) in
(4.4.20) leads to
00
>
00
t,
a(s)
00
t,
x'(s)f'(x(s))
f(x(s)) ds
~ lim In [ f(x(t)) ]
2 Hoo
f(X(tl)) '
00,
or
00.
t-400
(4.4.31)
347
du
jI/e>(u) <
= Loo q(s)ds
(1a(s) Q(s) )
00
l/e>
( 4.4.32)
<00
and
00
ds =
( 4.4.33)
00,
or
f'(x)
)/
f (e>-l
e>(x)
(4.4.34)
and
(4.4.35)
00,
t (_1_
ito a(s)
[Q(s)
+k
00
1/0
ds =
00,
( 4.4.36)
> 0 for t
x'(t)
jI/a(x(t
it follows that
X (t)
x(tll
du
jI/e>(u) >
it
tl
for
t>t
- 1
for some
tl >_ to,
l/a
ds --+
00
as
t --+
00.
Chapter 4
348
This contradicts (4.4.32). We also find that
f'(x(s)) ds
j OOa(s)(x'(s))O+l
J2(x(s))
= jOOa(s)(x'(S))O+l (
f'(x(s))
) ds
f(O+ll/o(x(s)) f(O-ll/o(x(s))
00
a-l/O(s)Q(o+ll/<>(s)ds.
+k
00
a-l/"'(s)Q(a+ll/o(s)ds,
t ~ tl
or
~ tl .
x'(t)
> (_1_ [Q(t) + k jOOa-l/"'(s)Q(a+ll/"'(S)dS]) 1/", for t
J1/"'(x(t)) - a(t)
t
ho(t) =
00
q(s)ds
exist ?nd
ho(t)
~0
and for a positive integer n and a positive constant k define the following
functions
and
00
a-l/",(s) [ho(s)
+ khn(s)]("'+ll/'" ds,
= 1,2,.
hn(t) exists for n=0,1,,N-1 and hN(t) does not exist, (4.4.37)
then equation (4.4.1) is oscillatory.
Proof. Suppose equation (4.4.1) has an eventually positive solution x(t).
Then from (4.4.22), we have
a(t)lx'(t)I"'-lx'(t)
f(x(t))
roo a(s)lx'(s)la+l
ho(t)
+ it
ho(t)
f(a+ll/a(x(s))
for
tl
f'(x(s))
j<a-ll/a(x(s))
for some
tl
ds
to.
(4.4.38)
:> tl'
349
))<>+1
100
it/00 a(s) ('(
;2tC(S)) f'(x(s))ds ~ k t a-I/<>(s)[ho(s)]C<>+I)/<>ds =
for
khl(t).
(4.4.39)
If N = 1 in (4.4.37), then the right-hand side of (4.4.39) does not exist.
This is a contradiction to (4.4.20).
Next, it follows from (4.4.38) and (4.4.39) that
f(x(t))
~ ho(t)
+ kh1(t)
for
t ~ tl
(4.4.40)
l Xl
Ix'(s)Ic>+l
a(s) J2(x(s)) f'(x(s))ds ~.
~ tl ,
(4.4.41 )
If N = 2 in (4.4.37), then the right-hand side of (4.4.41) does not exist.
This again contradicts (4.4.20). A similar argument yields a contradiction
for any integer N > 2. This completes the proof.
Remark 4.4.1. Lemmas 4.4.2 and 4.4.3 and Theorems 4.4.1 and 4.4.2
can be obtained for forced equations of the form
(a(t)lx'(t)IO-Ix'(t))'
+ q(t)f(x(t)) =
e(t),
(4.4.42)
roo
le(s)lds < 00. In this case the conwhere e(t) E C([to, (0), JR) and
clusion of Theorems 4.4.1 and 4.4.2 need to be replaced by: every solution
x(t) of equation (4.4.42) is either oscillatory, or liminfHoo Ix(t)1 = o.
The formulation of these results are left to the reader.
To prove our next linearization result, we need the following lemma.
Let a > 0 be a constant, a(t) E C([to, (0), JR+) and
C([to,oo),JR). Then the half-linear differential equation
Lemma 4.4.4.
q(t)
(a(t)lx'(t)I,,-lX'(t))'
+ q(t)lx(t)I<>-lX(t)
w'(t)
=
~
0 on
to and a function
[T, (0).
(a(t)lx'(t)I,,-IX'(t))'
+ q(t)lx(t)Ii3- 1 X(t) =
0,
(4.4.43)
Chapter 4
350
where {3 > 0 is a constant.
liminf
t-+co
it
to
q(s)ds > -
(4.4.44)
00.
(4.4.45)
w'(t)
-q(t) - (3a(t)
Ix'(t)IQ+l
xi3+ 1 (t)
(4.4.46)
and hence
wet) = w(to) -
to
q(s)ds - /3
it
to
a(s)
Ix'(sW+ 1
X
i3+ 1 ( ) ds.
(4.4.4 7)
I(to, t)
to
a(s)
IX'(SW+l
X
i3+l() ds
S
x1-'Y(t) - x1-'Y(to)
b-
~ b-1)(I(to,t))<>~,
1)
(l:
1: :~~;)
I
Ids
a- i (S)dS) <>+1
Cl
351
= (1/M)(l3-a)/la h -l)].
W'(t) -S
~ q(t) ~
(c 1 /3)
0' (
0'
a(t)rJ""(t)
1/",
Iw(t)I(",+I)/a
for
~ T.
for every
t 2': to.
(4.4.49)
Let T 2': to be snch that A = >.. + f3I(to, T) > O. Then, (4.4.49) implies
that w(t) is negative on [T,oo). Now, (4.4.49) gives
Ix'(t)I
+1
x(t) ,
t 2': T
> In (X(T))f3
~
x(t)
0,
(4.4.50)
W'(t) =
(4.4.51)
352
Chapter 4
~(a)a-1/<>(t)lw(t)I(0+1)/0
(F).[
(f'(u)) 1/(0+1)
J2(u)
du <
-00
00
and
(f'(U)) 1/(<>+1)
J2(u)
du <
00
(4.4.52)
and
. {.
f'(x)
;~~ j(o-l)/o(x)
mm
inf (
x<o
) 0/(0+1)
ix
1-
f'(x)
)0/(0+1)
j(o-l)/o(X)
x
J2(u)
00
du,
+ c1q(t)lz(t)I<>-l z (t)
(a(t)rt(t)lz'(t)IO- 1z'(t))'
(4.4.53)
:r(to, t)
t
to
Ix'(s)lo+ 1 ,
li
to
x'es)
(f'(X(s)))
J2(x(s))
O~,
ds
(i a- 1/O(s)ds
t
to
0+1
t 2': to.
(4.4.54)
353
f'(x(t))
)<>/(0+1) (= (f'(u) )1/(0+1) du 2 5
f(o-I)/Oo(x(t))
.Jx(t)
J2(u)
for
(4.4.55)
1=
* -
t 2 to,
(f'(v.))I/(Oo+1)
~()
d11. > O.
x(to)
f'(x(t))
f(Oo-l)/Oo(X(t))
5(0+1)/0
(f'(
)) du
_11._
J2(11.)
[K* _ l (f'(_11._))
[1
1/(a+1)
]-(<>+1)/<'>
x(t)
5(00+1)/0
x(to)
L
t
5(0+1)/<>
1/(0+1)
X(t)
K* -
2 5(0+1)/0 [K*
x'(s)
lit
to
d11.
] -(00+1)/"
J2(u)
,
(;2~:~:~D
]-(0+1)/<;
1/(<>+1)
ds
f'(x(t))
C
jCa-l)/a(x(t)) 2 7)(t)
for all
(4.4.56)
t 2 to,
w'(t)
:<::::
q(t) - a
(~a (a(t)7)a(t))l/o
1
)
Iw(t)l(a+1)/Oo
for
t 2 to.
The rest of the proof is similar to that of Theorem 4.4.4 and hence omitted.
354
Chapter 4
Case 2. J (to, (0) = 00. The proof for this case is similar to that of Theorem
4.4.3 Case 2 and hence omitted.
Remark 4.4.3.
1. If
2. From Theorems 4.4.3 ~ 4.4.5 it is clear that the oscillatory results for
equations presented in Chapter 3 can be used to obtain interesting oscillation criteria for nonlinear equations of type (4.4.1). The
formulations of these results are left to the reader.
half~linear
3. Theorems 4.4.1 and 4.4.2 hold for (4.4.1) without (4.4.34). This can
be proved by using Theorem 4.4.5 and replacing tpe function a(t) with
a(t)rl"'(t) in (4.4.35) and (4.4.36) of Theorem 4.4.1 and in hl(t) and hn(t)
in Theorem 4.4.2. Further, in both theorems k > 0 is any constant.
In the following results we shall obtain oscillation criteria for (3.13.76)
and (3.13.89) when g(t) = t.
10
00
1'1/1-1
(a~s)) 1ds
00
Q.
ki= o.
(4.4.59)
i= O.
Proof. The 'only if' part follows from Theorem 3.13.12. To prove the
'if' part assume that (3.13.76) has a nonoscillatory solution x(t), say,
x(t) > 0 for t ~ to. Integrating (3.13.76) from t to T and letting
355
1p-1 (att)
1=
~ 0,
we find a(t)1jJ(x'(t)) ~
( 4.4.60)
F(s, X(S))dS) .
> 'Ij; -1
> 1j;-1
-1
a(t)
F(s,x(S))d .5 )
00
xC(s)
(_1_
roo F(s, x(s)) dS)
a(t) it
XC(S)
for
to
Since :r(t) ~ m for some constant m > 0, ill view of the strong superlinearity of (3.13.76), we have x-C(t)F(t,x(t)) ~ m-CF(t,m) for t ~ to,
so that
x'(t)
((m-
00
>
1p-1
>
att)
F(s, m)ds)
for
to.
<
00.
and
00
du
<
1j;-l og (U)
00
for any
> O.
( 4.4.62)
Chapter 4
356
Proof. Since the 'only if' part follows from Theorem 3.13.12, it suffices
to prove the 'if' part. Assume that (3.13.76) has a nonoscillatory solution
x( t) > 0 for t ~ to > O. As in Theorem 4.4.6, we obtain (4.4.60) which
implies
x'(t)
1p-1 (a(t)
00
q(s)g(X(S))dS) ,
~ to
x'(t)
. -1 (
> 1/J
>
1/J-l
;;'-(0
(a(t)
j'CXJ
t
g(x(s)))
q(s) g(x(t)) ds
[X) q(S)dS)
for
(4.4.64)
~ to,
where we have used the fact that the function g( x( s)) is increasing and
~ 1 for s ~ t.
g(x(s))/g(x(t))
l t (1 1
00
1/J-l
to
a(s)
q(u)du ) ds:S:
L=
1/J-1
(ats)
lt
x'(s)
ds =
1/J-1(g(x(s)))
to
00
q(U)dU) ds <
(t)
x(tol
1p-1
dw
,
0 g(w)
00.
ExaIllple 4.4.1. Consider (4.4.43) and assume that ICXJ a-1/n(s)ds = 00.
From Theorem 4.4.7 it is clear that a necessary and sufficient condition for
(4.4.43) to be oscillatory is that
00
to
Jt
OO
a(s)
[= q(u)du)
1/0.
ds
00
if
(3 > a.
= t. Let 7r(t)
lto
and
(00
lto
( 4.4.65)
00
00,
(4.4.66)
357
C2
> 0,
which contradicts the condition (4.4.66). Similarly, (3.13.89) does not admit
eventually negative solutions of type II. This completes the proof.
lto
00
(4.4.67)
x'(t) > 0, t
to,
or
If x'(t) > 0 for t ~ to, then x(t) ~ Cl, t ~ to for some constant
> 0. As in Theorem 3.13.16, we find that
F(s,x(s))ds < 00 and
hence
F(s, cl)ds < 00. This implies that
F{s, Cl7T(S))ds < 00,
which contradicts condition (4.4.67).
(i)
Cl
It:'
It:'
It:'
Chapter 4
358
(ii) If x'(t) < 0 for t 2': tl, then there exists a constant C2 > 0 such that
x(t} 2': C21T(t) for t 2': t l , (cf. Lemma 3.13.9), and strong superliniearity
of (3.13.89) implies that
(4.4.68)
where c >
for
t 2': t l
[_ ( -a(t)(x'(t))"*r(C-O)/"], =
C: 0(-a(t)(x'(t))"*)
-c/o
c:
Ix'l"
(4.4.69)
sgn x' that
--ciCF(t,c21T(t))
for
t2':t l .
--cic
it
t,
F(s, C21T(S))ds:S
-a(td(x'(td)O
00.
"J -(c-a)/o
Example 4.4.2. Once again consider (4.4.43) and assume that aCt) = eAt
and q(t) = eJ.Lt, t 2': 0 where ), and J-t are constants and ), > O. Let
o < (3. Then by Theorem 4.4.9 equation (4.4.43) is oscillatory if and only
if J-t/), 2': (3/0.
(a(t)x'(t))'
+ p(t)x'(t) + q(t)lx(t)I'"Y
sgn x(t)
= e(t),
(4.5.1)
where I> 0, aCt), e(t), pet) and q(t) E C([to,oo),lR) and aCt) > 0 for
t 2': to
359
Theorem 4.5.1. Suppose for every T 2 0 there exist T:'S Sl < t1 :'S
S2 < t2 such that
Qi(U) =
for 'i
E
t E
[Sl' tIl
[S2, t2].
and
u(t)
(4.5.2)
E
D(Si,ti)
{u(t)
(3()
ca.(s) (u.'(s)
+ a(s);(s) u(s)
)2j ds
2 0
(4.5.3)
,_
_ e(t)p(t)
w (t) - p(t)q(t)
()
x' t
+ IX
,-I
w 2 (t)
j3(t)
(t) I) ( ) + () ( ) lL(t).
a, t p t a t p t
I
(4.5.4)
w'(t) 2 p(t)q(t)
on
a(t)p(t)
I.
(4.5.5)
Jf 2u(s)u'(s)w(s)ds
(3(s)
Jf u 2(s)p(s)q(s)ds + . f a(s)p(s)
u 2(s)w(s)ds
(
+ Jf ,X
1
2
w 2 (s)
- (s)u (s) a(s)p(s)ds,
which is equivalent to
02 (
Jf
( [p(s)q(s)u2(s) -
+ Jf
a(s)p(s) ( ,
IX,-l(S) U (s)
u(s){3(s))
2a(s)p(s)
(a(s)p(s))~] ds
u(s){3(s)
+ 2a(s)p(s)
IX,-l(S)
)2] ds.
(4.5.6)
360
Chapter 4
Now, we consider the following two cases:
K 1- ,
_1_
'Y
'Y
-xl-,(t) <_
CIon
I,
(4.5.7)
Next, we shall present the following criterion for the oscillation of all
solutions of the superlinear equation (4.5.1), i.e., when 'Y> 1.
Theorem 4.5.2. Suppose for every T ~ 0 there exist T::; Sl < tJ ::;
S2 < t2 such that condition (4.5.2) holds and q(t) ~ 0 on [Sl, tl]U[S2' t2].
If there exist pet) E Cl([to, 00), JR+) and v,(t) E D(Si' ti), i = 1,2 where
D(Si, t i ) is as in Theorem 4.5.1 such that
Qi(V,)
l~i [E(S)V,2(S) -
a(s)p(s) (v,'(s)
+ v,(s) 2afs\~(s) 2]
ds
(4.5.8)
E(t) = (p(t)q(tlh(le(t)lp(t(r-l)h,
then equation (4.5.1) with 'Y
> 1 is oscillatory.
Proof.
(t) - p(t)q(t)x
,-1
(J(t)
e(t)p(t)
(t) + a(t)p(t) wet) - x(t)
+ a(t)p(t) w
(t). (4.5.9)
w'(t)
p(t)q(t)X,-I(t) + le(t)lp(t)
x(t)
(J(t) wet)
a(t)p(t)
a(t)p(t)
w2(t).
361
+ le(t)IP(t)]
2: (p(t)q(t))lh(le(t)lp(t))h-1lh
x(t)
on I, we have
1O'(t) 2: (p(t)q(t))lh(le(t)lp(t))h-1lh
=
E(t)
(3(t)
a(t)p(t) w(t)
a(t)p(t)
10 2(t)
a(t)p(t)
(4.5.10)
Multiplying both sides of (4.5.10) by u 2 (t) and integrating over I, we
obtain
02: {E(s)u 2(s)ds
if
([(2U I (S)U(s)
if
+ (3(S)U 2(S))
a(s)p(s)
1o(s)
+ u 2(S)10 2(s)]
a(s)p(s)
ds
if
Ja(s)p(s)
2a s p s
Remark 4.5.1.
(t p(s)/a(s)ds) ,
then (3(t) =
o.
Thus,
with this choice of the weight function p(t) simpler forms of Theorems
4.5.1 and 4.5.2 can be reformulated. The details are left to the reader.
Example 4.5.1.
+ t12X'(t) + t"l
m sint Ix(t}l"l sgn x(t)
( ~X'(t))'
t
cost,
t > 0 (4.5.11)
where 'Y > 1 and m > O. For any T > 0 choose n sufficiently large so
that 2m!" 2: T, and let Sl = 2m!" and t1 = 2m!" + (7r /2) in (4.5.8), take
p(t) = t and u(t) = sin 7r(t - Sl)/(t1 - Sl) if t E [Sl, tlJ. Then, we have
Q1(U) =
where
1:'
("/2
4m 1h io
2(311"1
sin 2 +(lhl s
COS 3 -(lhl
(3 + 1.) r (4 - 1.)
"I
r(7)
11"1
m
"I
cos 2 2sds
7r,
7rr(7)
2:
("/2
s ds - 4 io
2r (3 + ~ ) r (4 _ ~ )'
(4.5.12)
Chapter 4
362
(4.0.1)
where f is a superlillear restoring force, i.e., f (x) E C 1 (lR, lR), f' (x) 2:
0, xf(x) > 0 for x =J 0, f(O) = 0 and f satisfies condition (4.1.5),
g(x) E C(lR, lR) and 9 is locally Lipschitz. In addition p(t), q(t) E
C([to, 00), lR) (locally integrable would suffice).
In what follows we shall extend the well-known result, namely, if q(t) >
0, t 2: to then all solutions of (4.1.6) oscillate if and only if ftC; sq(s)ds =
00.
!too
Proof. Let x (t) be posi tive and monotone increasing in [T, 00) for some
T 2: to with limt---+oo x(t) = c > 0, where c is a constant. All other
possibilities can be treated similarly.
Since x'(t) 2: 0 on [T,oo) and x(t) is bounded, we find
lim inf x'(t)
t---+oo
Suppose for some
EO
O.
(4.6.2)
EO.
(4.6.3)
> 0,
lim sup x' (t) 2:
t---+oo
363
such that EO = x'(a n ) > x'(t) > x'(b n ) = Eo/2, whenever t E (an, bn ).
Integration of equation (4.6.1) from an to bn gives
1
-2" EO
rbn
Jan
p(s)g(x'(s))ds -
ibn
a"
q(s)f(x(s))ds.
(4.6.4)
Now, from the integration by parts and the general mean-value theorem
for integrals, we get
1~" q(s)f(x(s))ds
Q(bn)f(x(bn)) - Q(an)f(x(an))
+ 1~" Q(s)f'(x(s))x'(s)ds
In the remainder of this section, we shall assume that there exist constants K, b with b > 1, such that
o < g(u)
for all
u,
(4.6.6)
,(u) = {
g(u) if u~O
u
o if u = O.
The next two lemmas show that certain types of nonoscillatory behavior
can be excluded by appropriate sign or integral conditions on the functions
p(t) and q(t).
Chapter 4
364
Lemma 4.6.2. (i) p(t):::": 0 for all sufficiently large t and the existence
of Q(t) :::": 0 with Q(t) rf. .c1[to, 00) imply that there is no solution x(t)
of equation (4.6.1) satisfying x(t)x'(t):::": 0 for all large t.
(ii) p(t) E .c 1/(2-i3l[to,00), q(t) :::": 0 for all sufficiently large t with
Q(t) rf. .c1[to, 00) imply that there is no bounded solution x(t) of equation
(4.6.1) satisfying x(t)x'(t):::": 0 for all large t.
Proof. Assume that x(t) is a solution of equation (4.6.1) which is positive
and increasing on [to, 00). We also suppose that to is large enough so that
all sign hypotheses hold on [to, 00). Define w(t) = x'(t)/ f(x(t)), t :::": to.
Then for t:::": to, we have w(t):::": 0 for t:::": to, and
(4.6.7)
for
t:::,,: to.
(4.6.8)
ItT
It
1to00 Q(s)ds
::;
100 w(s)ds
to
::;
100
x(tol
(ii) For this case, we write the equation (4.6.7) in the form
(w(t)ex p
U:
(1:
p(sh(x'(S))dS) .
w(t) = w(T)(T)
where (s) = exp
w(t) :::":
[q(s)
(ItS p(uh(x'(u))du).
iT
q(s) exp
(4.6.9)
+ f'(x(s))w 2 (s)](s)ds,
This yields the inequality
(100 p(uh(X'(U))dU) .
(4.6.10)
365
K (x' (a) )/3- 1. Thus, for s, t :::-: to sufficiently large and Holder's inequality,
we get
K o,
say, where Tn = 1/(2 - (3) and c = limt-hXJ x(t). Using this estimate
in (4.6.10) and the fact that q(t) :::-: 0 for sufficiently large t, we find
'W(t) :::-: e- Ko ftT q(s)ds. The rest of the proof is similar to that of Case (i)
above.
The proof of the nonexistence of negative, decreasing solutions of equation (4.6.1) is similar.
(IIl
[1 + J: Ip(s)ldS]
-1
t/. 1/(b-1)[to,00),
and Q(t)
Proof. We shall consider only the proof of nonexistence of positive, decreasing solutions. The proof for negative, increasing solutions is similar.
Assume that x(t) is a positive solution of equation (4.6.1), monotone decreasing on [to, (0) with to large enough so that sign hypotheses hold on
[to, (0).
(h) q(t):::-: 0 implies that x"(t) + p(t)g(x'(t)) <:; 0 for t:::-: to. Let
v(t) = -x' (t). Then, we have v' (t) + p( t)g( -v( t)) :::-: 0 and so by condition
(4.6.6) ,
(4.6.11)
The hypothesis q( t) l' 0 on any half-line is to exclude the possibility
that x(t) is eventually constant. Hence, we can assume without loss of
generality that v(to) > 0 and integrate (4.6.11) over an interval [to, t] on
which v(t) is nonvanishing, to obtain
- K(b - 1)
it
Ip(s)lds.
(4.6.12)
to
v(t) :::-:
[v1-b(tO)
+ K(b -1)
L
t
IP(S)ldS]
-1/{b-1)
,
t:::-: to
(4.6.13)
Chapter 4
366
Now, the hypotheses imply that
+ K(b -1)
V 1 - b (tO)
however, 0 -c::
Lt]-1
Ip(s)lds
rt
(I/o
ltD
t:;;. to.
( 4.6.14)
u'(t)
+ q(t) exp
(1:
p(s)ry(x'(S))dS) f(x(s)) =
u(t) - u(t n )
lt
q(s) exp
f(x(t))
lt
(1:
(4.6.15)
o>
o.
p(o}y(x'(O))dO) f(x(s))ds
q(s)(s)ds
lt
f'(x(s))x'(s)
(l
tn
q(O)(O)dO) ds,
(4.6.16)
where (s)
as
exp
(4.6.17)
where
F(t) = f(x(t))
(n
it
q(s)(s)ds
and
H(t)
f'(x(t))x'(t)
f(x(t))
:;;. O.
iT
iT
Q(s)'(s)ds
367
from the hypotheses it follows that C/>, c/>', Q are all nonnegative on [to, 00)
with limT---+oo Q(T) = 0, and from (4.6.14), we find that </>(T) ::; eKo for
T 2:: to Thus, limT---+CX) It q(s)c/>(s)ds 2:: O. This contradiction completes
the proof.
Now, we are prepared to state and prove the following oscillation criteria..
Theorem 4.6.1. Let condition (4.6.6) hold. Then,
(I)
[1 + Itiop(s)dsr1 (j 1/(b-1)[to,
00) and Q(t) (j l[to, 00) imply that equation (4.6.1) is oscillatory.
(II) p(t) E 1/(2-i3)[to,00), q(t) eventually positive with Q(t) (j 1 [to,oo)
imply that all bounded solutions of equation (4.6.1) oscillate.
(III) p(t) eventually positive, p(t) E 1/(2-13) [to, 00), Q(t) exists and is
eventually positive with Q(t) (j 1 [to, 00) imply that all bounded solutions
of equation (4.6.1) have oscillatory derivatives.
Proof. In (I) and (II), the positivity hypothesis on q(t) implies that
any nonoscillatory solution x(t) of equation (4.6.1) satisfies the inequality
x"(t) sgn x(t) < 0 for any sufficiently large t for which x'(t) vanishes.
This implies that any nonoscillatory solution must be monotonic. The
result now follows from the appropriate parts of Lemmas 4.6.2 and 4.6.3.
For (III) we can use the above lemmas to conclude that any bounded
nonoscillatory solution of equation (4.6.1) cannot be monotonic, which is
equivalent to the statement that its derivative oscillates.
Remark 4.6.1. The condition p(t) E 1/(2-13) [to, 00) in (II) is stronger
than the condition
[1 + Ittop(s)dsr1 (j 1/(b-1)[to,00).
vi [lulb-1 + Ivlb- 1 ]
(i2) q(t)El[to,OO), to 2::0 and there exist To 2:: to, k with k>l such
Chapter 4
368
that
Ql(t) =
100
q+(s)ds > k
100
q-(s)ds
Proof. We note that the proof presented here is similar to that of Theorem
4.3.1. We observe that condition (i2) implies that for t ~ To,
itrOO [q+(s) -
Q(t)
>
q-(s)] ds ~
(k-1) QI(t)
-k-
(k-k- 1) Q3(t),
where Q3(t) = !too Iq(s)lds. Thus, Q(t) E LI[to, 00) shows that Q3(t) E
.c1[to, 00). We also note that Q3(t) ---+ 0 monotonically as t ---+ 00. Let
c > 1 and define
A =
max
f(u).
(4.6.18)
c-I:O;u:O;c+1
Choose T
To so large that
1
loo
Q3(s)ds
min{1,1/(2A)}
(4.6.19)
(4.6.20)
1,
(4.6.22)
IIQ;lx'll <
(4.6.23)
369
and
+K
q(s)ds
t, -
[It?-lp(s)l l /(2-(3)ds J
2-(3
t,
(4.6.24)
[T,T+n].
t2 E
(Ynx)(t) =
-100
r r
JT+nJs
(X)
C -
00
q(a)f(x(a))dads, T
+ n < t.
Now, since
:s;
1 Ip(s)IQ~(s)ds
00
(1
Q.(s)ds )(3-1
00
( 4.6.26)
Combining (4.6.25) and (4.6.26) and then using the inequalities (4.6.20)
and (4.6.21), we find
2-(3
(1 Q*(S)dS)(3-l +AQ3(t)
00
Q.(t).
[l,
+ n]
( 4.6.27)
gives
2-(3
l(YnX)'(t2)-(ynxr(tl)1 :s; K
Ip(sW/(2-(3)dS]
+AQ3(t), (4.6.28)
370
Chapter 4
I(Ynx)(t) - cl
<
lOO 2AQ3(S)ds
+n 2AQ3(S)ds +
roo
JT+n
AQ3(S)ds
::; 1.
(4.6.29)
We also have
(Ynx)(t) = (Y,.,x)(T + n)
for all
t ~ T + n.
(4.6.30)
x, z
a(8) =
max
U,1I
E [c - 1, c
1]
If(u) - f(v)l
Now, by condition (il)' the definition of a(8), and the fact that Ix'(s)1 ::; 1
and Iz'(s)l::; 1 for all s, we find for t E [T, T + n],
l(Ynx)'(t) - (Ynz)'(t) I
:;1
:;l
00
p(s)llg(x'(s - g(z'(slds +
00
Iq(s)llf(x(s)) - f(z(slds
OO,q (s)'ds.
(4.6.32)
From (4.6.31) and (4.6.32) it is clear that Yn is continuous. Thus, we can
apply the Schauder fixed point theorem to obtain Xn E Bn which satisfies
Ynx n = x n . Finally, the equicontinuity of the sequence {x n } on [T,oo)
completes the proof of the theorem.
371
Theorem 4.6.3.
Let 'Y > 1, b ~ 1, f3 = min{b,2}. Assume that
p(t), q(t) E C([to, 00), lR+) and p(t) E 1/(2-/3) [to, 00). Then all solutions
of
x"(t)
sgn x(t) = 0
Remark 4.6.2.
+ -t 211nt [x(t)[/3
sgn x(t)
0,
(4.6.33)
Chapter 4
372
(a(t)'Ij;(x(t))x'(t))'
+ p(t)x'(t) + q(t)f(x(t)) =
0,
where a(t), p(t), q(t) E C([to,oo),JR), a(t) > 0 for t 2 to, 'Ij; E C(JR,JR+)
and f E C(JR, JR). It would be interesting to obtain oscillatory results
similar to those presented in Sections 4.2 and 4.5 for equations of the type
(4.4.1).
4.8. References
1. R.P. Agarwal and S.R. Grace, Second order nonlinear forced oscillations, Dyn. Sys. Appl., to appear.
2. R.P. Agarwal, S.R. Grace and D. O'Regan, Linearization of second
order superlinear oscillation theorems, to appear.
3. R.P. Agarwal, S.-H. Shieh and C.-C. Yeh, Oscillation criteria for
second order retarded differential equations, Math. Comput. Modelling
26( 4)(1997), 1-11.
4. G.J. Butler, On the oscillatory behavior of a second order nonlinear differential equation, Ann. Mat. Pura Appl. 105(1975), 73-92.
5. G.J. Butler, The oscillatory behavior of a second order nonlinear differential equation with damping, J. Math. Anal. Appl. 57(1977), 273-289.
6. G.J. Butler, Hille--Wintner type comparison theorems for second order
ordinary differential equations, Proc. Amer. Math. Soc. 76(1979), 51-59.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
373
374
Chapter 4
26. S.R. Grace and B.S. Lalli, An oscillation criterion for second order
sublinear ordinary differential equations with damping term, Bull. Polish.
Acad. Sci. Math. 35(1987), 181-184.
27. S.R. Grace and B.S. Lalli, Almost oscillation and second order functional differential equations with forcing term, Boll. Un. A1at. Ita1.
B(7)1(1987), 509-522.
28. S.R. Grace and B.S. Lalli, Oscillation theorems for second order nonlinear differential equations, J. Math. Anal. Appl. 124(1987), 213-224.
29. S.R. Grace and B.S. Lalli, Oscillation theorems for second order nonlinear differential equations with deviating arguments, Internat. J. 1I1ath.
Math. Sci. 10(1987), 35-45.
30. S.R. Grace and B.S. Lalli, On the second order nonlinear oscillations,
Bull. lnst. Math. Acad. Sinica 15(1987), 297-309.
31. S.R. Grace and B.S. Lalli, Integral averaging and the oscillation of
second order nonlinear differential equations, Ann. A1at. Pura App1.
151(1988), 149-159.
32. S.R. Grace and B.S. Lalli, Oscillation in second order differential equations with alternating coefficients, Period. Math. Hungar. 19(1988), 6978.
33. S.R. Grace and B.S. Lalli, Oscillation theorems for nonlinear second
order differential equations with a damping term, Comment. Math. Univ.
Carolin. 30(1989), 691-697.
34. S.R. Grace and B.S. Lalli, Integral averaging techniques for the oscillation of second order nonlinear differential equations, J. Math. Ana1. App1.
149(1990), 277-311.
35. S.R. Grace and B.S. Lalli, Oscillation theorems for second order superlinear differential equations with damping, J. Austra1. Math. Soc. Ser. A
53(1992), 156-165.
36. S.R. Grace and B.S. Lalli, Oscillation theorems for second order nonlinear functional differential equations with damping, Comput. Math. Applic.
25(1993), 107-113.
37. S.R. Grace and B.S. Lalli, Oscillation criteria for damped functional
differential equations, Dynam. Stability Systems 9(1994), 215-222.
38. S.R. Grace, B.S. Lalli and C.C. Yeh, Oscillation theorems for nonlinear second order differential equations with a nonlinear damping term,
SIAM J. Math. Anal. 15(1984), 1082-1093.
39. S.R. Grace, B.S. Lalli and C.C. Yeh, Addendum: Oscillation theorems
for nonlinear second order differential equations with a nonlinear damping
term, SIAM J. Math. Anal. 19(1988), 1252-1253.
40. J.R. Graef, S.M. Rankin and P.W. Spikes, Oscillation theorems for
perturbed nonlinear differential equations, J. Math. Anal. Appl. 65(1978),
375-390.
41. J .R. Graef and P. W. Spikes, Asymptotic behavior of solutions of a second order nonlinear differential equation, J. Differential Equations 17(1975),
461-476.
375
Chapter 5
Oscillation Theory for
Sublinear Differential
Equations
5.0. Introduction
In this chapter we shall present oscillation and nonoscillation criteria
for all solutions of second order nonlinear differential equations of sublinear
type with alternating coefficients. In Section 5.1, our sublinear oscillation
results involve integrals and weighted integrals of the alternating coefficients. In some results we employ integral averaging techniques. In Section
5.2, we impose some additional conditions on the sublinear term which allow us to proceed further and extend and improve several known theorems
in the literature. In fact, we make an asymptotic study which results in
new oscillation criteria. Section 5.3 provides some new linearized oscillation
results for second order sublinear differential equations. In Section 5.4, we
shall present criteria for the nonoscillation of sublinear Emden-Fowler type
equations. In Section 5.5, we compare the oscillatory behavior of certain
nonlinear equations with the related half-linear differential equations. Here
oscillation of general nonlinear differential equations is also discussed.
+ q(t)f(x(t)) =
(a(t)x'(t))' + q(t)f(x(t))
x"(t)
(a(t)x'(t))'
(5.1.1)
0,
+ p(t)x'(t) + q(t)f(x(t))
(5.1.2)
0,
(5.1.3)
377
where
(i)
and satisfies
f is strongly
du
j+0. feu)
xi- 0
(5.1.4)
du
feu) < 00.
(5.1.5)
< 00 and
-0
and
(a(t)x'(t))'
+ p(t)x'(t) + q(t)lx(t)I"Y
0,
=
(5.1.6)
(5.1.7)
sgn x(t) = 0
(5.1.8)
/)0
and
00
/
a(s)p(s)
ito
00.
p(u)q(u)duds
(5.1.10)
(it) If f3(t) = a(t)p'(t) - p(t)p(t) :S 0 and f3'(t) 2': 0 for t 2': to, then all
bounded solutions of equation (5.1.3) are oscillatory or tend monotonically
to zero as t -t 00. If in addition, condition (5.1.5) holds, then all solutions
of (5.1.3) are oscillatory.
(h) If foo 1f3'(s)lds < 00, then every bounded solution of equation (5.1.3)
is oscillatory.
Let x(t) be a nonoscillatory solution of equation (5.1.3), say,
x(t) > 0 for t 2': to. Define wet) = p(t)a(t)x'(t)/f(x(t)), t 2': tl for some
tl 2': to Then for t 2': t l , we obtain
Proof.
w'(t)
<
x'(t)
-p(t)q(t)
+ f3(t) f(x(t))
-p(t)q(t)
+ f3(t) f(x(t))
x'(t)
Chapter 5
378
it
x'(s)
t
i t1 p(s)q(s)ds + t1 (3(s) f(x(s ds.
We shall show that both conditions (id and (i2) imply that Ittl (3(s)x'(s)
/f(x(s)ds is bounded above. Since,
tl
x'(s)
(3(s) f(x(s ds
ret) du
r(t 1) du
(3(t) Jo
f(u) - (3(td J o
f(u)
-I:
(3'(s)
(foX(S)
f~:) ds
(5.1.11)
tl
x'(s)
r(tl) du
(3(s) f(x(s ::::; - (3(td J o
f(u) = kl'
(5.1.12)
where kl is some constant. Also, if (i2) holds, then 1(3(t)l::::; b for some
constant b> 0 and since x(t) is bounded, we find that Iox(t) du/ f(u)
is also bounded. This together with the fact that (3'(t) E C[to, 00) shows
that
t
x'(s)
t1 (3(s) f(x(s ds ::::; k2
Let k
= w(td + max{k 1 , k 2 }.
k2 > O.
Then, we have
(5.1.13)
t1
x'(t)
a(t)p(t) f(x(t
or
x'(t)
f(x(t
::::; -
::::; 1
1 it
2 a(t)p(t)
t1 p(s)q(s)ds,
it
p(s)q(s)ds.
T), we get
S
ret) du
(x(T) du
(t 1
Jo
f(u)::::; J o
f(u) - JT a(s)p(s) t/(u)q(u)duds -+
which is a contradiction.
such that
-00
as t -+
00,
379
Finally, we note that the proof of the criterion for oscillation of all
bounded solutions of (5.1.3) in the case (i 1 ) is obvious.
00
a(s)p(s) ds
(5.1.14)
00.
x' (t)
--- < -
for
f(x(t)) -
a(t)p(t)
Integrating this inequality from t to T, we get
t 2: T.
ret) du
reT) du
rt
1
io
f(u)::::; io
f(u) - ciT a(s)p(s) ds -t
which is a contradiction.
00
as t -t
00,
/00 s"Yq(s)ds
(5.1.15)
00,
(5.1.16)
for every constant M, then the conclusion of Theorem 5.1.1 holds.
Let x(t) be a nonoscillatory solution of equation (5.1.3), say,
x(t) > 0 for t 2: to. Proceeding as in Theorem 5.1.1, we obtain (5.1.13),
Proof.
i.e.,
f~~W))
::::; a(t)lp(t) [k
-1:
p(u)q(U)dU] .
t,
x' (s)
f(x(s))ds::::;
it
t,
[it
a(s)p(s) k -
t,
p(u)q(u)du ds.
Chapter 5
380
Now from condition (5.1.16), we have
G(t)
x'(s)
f(x(s ds -+ - 00
t1
as
-+ 00,
I:S;)
G(t) =
du/ f(u). If x(t) ~ X(tl) for all large t, then
G(t) > 0 which is a contradiction. Hence, for all large t, x(t) S x(t 1), so
where
G(t) = -
(t ll
x(t)
r(t d f(u)
du
du
f(u) >
- io
> - 00,
0, t > 0 (5.1.17)
hmsup a(t)p(t)
t-4oo
If
lim
t-4oo
r a(s)ds
J~
it 1"
a(s)
t0
a(t)
< 00.
Ito a( s )ds
(5.1.18)
p(u)q(u)duds
00,
(5.1.19)
it
t1
x'(s)
a(s)p(s)a(s) f(x(s ds
it i"
t1
a(s)
t1
p(u)q(u)duds ~ k
it
t1
a(s)ds.
(5.1.20)
381
ret)
a(t)p(t)a(t) Jo
du
f(u) -
1t 1
1 +
s
a(s)
t,
t,
~ k
t,
a(s)ds
1t
t,
res)
(a(s)p(s)a(s))' Jo
du
f(u) ds
(5.1.21)
p(u)q(u)duds
r:(t,)
a(t1)p(tl)a(t1) Jo
du
f(u)
Set G(t) = f;(t) du/ f(u) and n(t) = a(t)p(t)a(t). We consider two cases:
either there exists a sequence {Tn}~=l' limn->oo Tn = 00 such that
n(t)G(t)
-it
t,
n'(s)G(s)ds
for
t 2: t2.
(5.1.22)
Itt,
write C(t) =
n'(s)G(s)ds which is nonnegative and nondecreasing.
Let t3 2: t2 be chosen so large that C(t) > 0 for t 2: t3. From (5.1.22),
we deduce that n(t)C'(t) ~ n'(t)C(t), or (n(t)/C(t))' 2: 0 for t 2: t3.
Thus,
n(t)G(t)
C(t)
1 1
t
t,
a(s)
t,
p(u)q(u)duds
kit
t,
C(t3) ~(~t;)
for
t 2: iJ.
= kl'
a(s)ds
+ n(t)G(tl) + kln(t).
Itt, a(s)ds
(5.1.23)
(tx'(t))'
sgn x(t) = 0,
t>0
(5.1.24)
where 0 < I < 1. It is easy to verify all conditions of Theorem 5.1.4 with
p(t) = l/t and a(t) = t, and hence equation (5.1.24) is oscillatory.
Chapter 5
382
lim sup
t--+oo
If
t~n~
t ~
ho a(s)p(s) ds
It
to
f,tto
O'(t)
~d
a(s)p(s)
O'(s)
asp s
( ) ()
< 00.
S
to
00,
p(u)q(u)duds
Proof.
x'(t)
f(x(t))
+ a(t)p(t)
it p( )q( )d
t,
a(t)p(t) '
t?: tl
ret) feu)
du
it, (s) Jo(Xes) feu)
du
it a(s)p(s)
O'(S) is
ds +
p(u)q(u)duds
ret,) du
i O'(S)
O'(t) J o
t, 0'
t,
t,
a(s)p(s) ds
+ O'(tl) Jo
t,
feu)'
00
s"Y-IQ(s)ds = 00,
(5.1.25)
to
Proof.
the equation
(t 2y'(t))' + tl+"Yq(t)y"Y(t)
O.
(5.1.26)
w'(t)
tl+"Y
(5.1.27)
383
w(T) _ w(t)
T1+,
t1+,
iT(l + "(
iT iT
)w(s) d
s2+, s
Let z(t)
q(s)ds -
"(
SI-'(Y'(s))2
1+ ()
ttY's
= x'(t)/xOt(t), t
z'(t)
(5.1.28)
ds.
q(t) - ,,(X,-I(t)Z2(t).
(5.1.29)
z(T) = z(t)
-iT -iT
q(s)ds
,,(X,-I(S)z2(s}ds.
(5.1.30)
t,
x(t)
CI,
t-4oo tl/(l-,)
y(t)
lim t,/(l-,)
CI,
t-4oo
where
lim w(T) =
C2
x'(t)
00.
Cl
- -,
lim t,/(I-,)
t-4OO
1-,,(
y'(t)
cn
lim t(2,-1)/(1-,)
t-4oo
1-"(
CI
T-4OO
where
lim
T-4OO
T 2Y'(T) = C2,
T1+,y'(T)
But, since
T1+,
oo w(s)
2+"ds
s
'
C3(t),
00 ::;
00
(5.1.31)
in (5.1.28), we find
C3(t) <
00.
Chapter 5
384
where
'ljJ(t) =
T w(s)
to
- 2 ds
S
yl-,(t)
1-,
1-,
it follows that C3(t) cannot be -
CXJ
1t
yl-,(tO)
- =--~---'-
as
CXJ,
and
CXJ.
(5.1.32)
CXJ
in (5.1.28), we
T --+
w(s)
w(t)
( 1 +s2+,
, ) - d st - < -Q(t) ,
1+, -
and hence
w(t) _ _ 1_1= (1
t2
t 1- , t
s2+,-
it 1=
t
to
w(s) --ds
S2
to
-1S1-,
w(u)
(1 +,)--duds?
u 2+,
t? to.
for
it
to
(5.1.33)
s,-lQ(s)ds,
i.e.,
where
C4
lit
is some constant.
-,
to
t'l=
w(s)
ds-S2
'Y
w(s)
(1+,)-ds
--+
S2+,
CXJ
as
t --+
CXJ.
(5.1.34)
Now, from (5.1.31) for any f> 0 there is T(f)? to such that
~~) ds
iT() s '
Hence, we find
'ljJ(T)
lT
I
I
w(s)d
to
T ()W(S)d
to
<
T ()
to
<
I
+ 'iT
T (<)
to
for all
w(s)d
S
S+
T? T(f).
iT w(s)
'd
2+,S S
s
iT ,s,-118 -w(u)d
- ud
2
T)
w(s)d
T
-- S s2
T(e) s2+,
T(e)
w(s)
-2-ds + 2dY for T? T(f).
-- s
T(e)
u +,
385
Therefore, we have
lim 1f;(T)
t--+oo
o.
T'Y
(5.1.35)
Now, since
T -l- 00
in (5.1.36), we obtain
(5.1.37)
'Ij;(t) - (1
+ 'Y)t'Y
1 1~~)'Y
00
ds
as
-l- 00
(5.1.38)
-l- 00.
- L, say, we find
s~:'Y
it
lim
p'Y(s)q(s)ds
t--+oo to
00,
(5.1.39)
Chapter 5
386
then equation (5.1.6) is oscillatory.
Proof.
w(t) =
(X(t))'
p(t)
t;::: to.
for
~(
) (p(t)w<>(t)"
w t
(p(t)WO-l(t))"
( _0:_)
0:-1
+_l_ p"(t)w<>-l(t),
(5.1.40)
+ o:p(t)W<>-3(t)(W'(t2
t;::: to.
1-0:
(5.1.41)
for
t;::: to.
(5.1.42)
Since p"(t) SO for t;::: to and 0: > 1, the last two terms in (5.1.41) are
nonnegative, hence we may combine (5.1.41) and (5.1.42), to obtain
_0:_(p(t)Wo-1(t))" S
0:-1
- p(t)q(t)
t;::: to.
for
(5.1.43)
p(t)w<>-l(t) ::;
Cl
+ cot -
-1) it is p'(u)q(u)duds,
0: 0:
to
(5.1.44)
to
lit is
to
to
p'(u)q(u)duds
00,
(5.1.45)
e' sin t
387
du
F(x) = J+o f(u)
for
x> 0,
F(x) = [ :
f~:)
for
x <0
Clearly, 0
f(x)
infx<o F(x)f'(x) }
1 + minx<o F(x)f'(x) .
(3
(3
The following examples produce lower bounds for (3 which are essential
for our oscillation theory (see for example Corollary 5.2.6).
Example 5.2.1. Let f(x) = (lxi' sgn x)/(1 + Ixl'), x E 1R. We observe
that f(x) is continuous 011 1R and has the sign property xf(x) > 0
for x oF o. Moreover, f'(x) = blxl,-1 )/([1 + Ixl'12) > 0 for x oF 0 and
condition (5.1.5) holds. Furthermore, we have
F(x) = Ixl
IxI I -,
-,
+ -1-
for
x oF 0
and consequently,
inf
(3 =
(_1_ + x,)
'
x>0(I+x,)21-,
O.
F(x)
fix i ~ <
J+0
U'
+u
fix i du =
J+0
U'
IxI 1-,
1- ,
and consequently,
inf F(x)f'(x)
x<o
1-,
inf F(x)f'(x) < inf _x_
x>o
x>o 1 - ,
[1 +,X,-I]
,
1-,
388
Chapter 5
=J 0,
F(x) = 1+0
and
F(x) -
ixi
du
1/,'Y + 'U
+0
du
10
/1 U'Y
du
,+0
=J 0,
IxI1-'Y
(Ixl du
>
we find
+ 1/,
?::
(I du
Ixl::; 1
if
1
2(1-')')'
1+02u'Y
if
Ixl?::
we obtain
and
F(x)1'(x)
for
x<o
=J 0,
Ixl?:: 1.
and consequently
(')'
2 1 - ')'
).
Finally, we have
')'
Clearly,
provided
f (x)
I
')'lxl'Y- 1
=J 0,
=J 0,
F(x) =
lXI
+0
u'Y {k +
holds
du
sin(ln[l
+ u])}
>
k+
ixi du
1 +0 u'Y
1 IxI1-'Y
-----
k+11-,),
389
fixi
du
l+o u'Y{k + sin(ln[l
F(x) -
+ ul)}
,[k-(l+~)]
(1 -,)(k + 1)
,[k+(l+~)]
(1 -,)(k - 1) ,
which gives
k -1 ,k - , - I
k + 1 k + 2, .
Now, we shall prove the following results.
Theorem 5.2.1. Suppose condition (F) holds, and let p(t)
1R.+ ) satisfy
[
(3p (t) +
C 2([to, 00),
(pf3(t))' ( a(t))
p(t)]2
a(t)
pf3-1(t) - p(t) a(t)
4{32
:S 1 _ {3 [(1 - (3)(P'(t))2 - p(t)p"(t)]
If
limsup
t--+oo
~
t
(5.2.1)
for
r (pf3((u))) q(u)duds =
ltD ltD
a u
t;::: to
(5.2.2)
00,
x(t)
w/(t)
(3pf3- 1(t)p/(t)F(x(t))
+ (pf3(t)) (a(t)x/(t))
a(t)
f(x(t))
and
w"(t)
+ (3~-l(t)p"(t)F(x(t))
[(
x' (t ) )
f(x(t))
X I (t))2
{pl(t)
f(x(t))
- (3 p(t)
2]
390
Chapter 5
1
]
( p{3(t))' ( a(t)) P(t)} x'(t)
aCt)
p{3(t) - aCt) f(x(t)) f'(x(t))
+ (p{3(t))'
aCt)
(~) _
p{3(t)
pf3-2(t) [{ I
+ 4f'(x(t))
(3p (t)
pet)
aCt)
}]2
(pf3(t))' ( a(t))
p(t)}2
aCt)
p{3-1(t) - pet) aCt)
it 1
:::::
p{3u)) q(u)duds
ttotoau
8
~ltlSp{3u))duds:::::
t
to a u
to
w'(to) <
00,
(3p (t)
for t
(a{3(t))' aCt)
aCt)
p{3-1(t)
]2
<
14~2(3
(5.2.3)
to, and
pet) 2 0,
k(t)
p{3(t)p(t)/(a(t)
t ~ to.
(5.2.4)
Proof. Let x(t) be a nonoscillatory solution of equation (5.1.3) and
assume that x(t) > 0 for t ~ to. Put wet) = p{3(t)F(x(t)) , t ~ to. Then
as in Theorem 5.2.1, for every t ~ to, we have
w
"( )
t
p{3(t) (a(t)x'(t))'
< - -'---:'-;---'--:'7-'--
aCt)
f(x(t))
391
pf3(t)
"pf3(t)
x'(t)
a(t) q(t) < - w (t) - a(t) p(t) f(x(t
for all
t ~ to
and therefore,
1: (p:(~~)
f~~~;dS
for a.ll t
~ to
Now, by the Bonnet theorem, for a fixed t ~ to and for some ~ E [to, tJ,
we find
to
x'(s)
(-k(s f(x(sds
k(to)
since
x
(t )
a
1(~)
x'(s)
f(u)
(t O )
xCO
if x(O
du
-- <
x(to )
io
du
f(u)
du
f( ) S; k(to)F(x(to,
u
> x(to)
if
x(~)
S;
x(to)
1(~)
f(u)
~
t
it 1
s
to
to
du
-- <
~
[0
x(to)
du
f(u)
if
x(~) ~ x(to)
to, we obtain
(pf3((u) q(u)duds
a u
<
~ to.
Chapter 5
392
(vtx'(t)'
Since f(x)
for every t
2yt
,
[ f3p (t)
Ixll/2sgn x, we have f3
to > 0, we find
= 1/2.
0,
~ to> O.
= t.
(5.2.5)
Then
(p,6(t))' a(t)
p(t)
]2
a(t)
p,6-1(t) - a(t)p(t)
~(1- sint)2
= 1 is oscillatory.
t ~ to.
:::; 1 =
-I1t
t to
1
8
to
p.6(u)
(-) q(u)duds = -1 lt 1 u 5 / 4 sin ududs -+
a u t to to
8
as
00
t -+
00.
Thus, all conditions of Theorem 5.2.1 are satisfied and hence (5.2.5) is
oscillatory.
Example 5.2.5. Consider the differential equation
(t 2x'(t'
v1nt
t ~ to> 1.
0,
= t,
(5.2.6)
we find for
~
t
lt 1
8
to
to
(p.6((u))) q(u)duds = ~
aut
lt 1
8
to
to
U5/ 4
sin ududs -+
00
as t -+
00.
393
Therefore, all conditions of Theorem 5.2.2 are satisfied and hence (5.2.6) is
oscillatory.
We note that Theorem 5.2.2 is not applicable to (5.2.5) since condition
(5.2.4) does not hold. Also, Theorem 5.2.1 fails to apply to (5.2.6) since
condition (5.2.1) is not satisfied. We also observe that conditions of the
type Joo ds / a( s) = 00 and Joo ds / a( s) < 00 are not required.
1 - (3
4(3
p(t)
a(t)
Theorem 5.2.3. Suppose condition (F) holds, and let p(t) E C 2 ([to, (0),
1R+), satisfy (5.2.1). If
lim sup
t-+oo
t (t - s)n (p!3((s)))
q(s )ds
a s
.fto
00,
(5.2.8)
(P:(~~))q(S)dS
:::;
-l:(t-s)n w l/(s)dS
(t - to)nw'(to) - n(n - 1)
+ n(t - to)n-1w(to)
1:
(t - s)n- 2w(s)ds
if n> 1
<
(to)n,
(to)n-l w(to)
1- t
w (to)+n 1- t
-t-
394
Chapter 5
~ ltot (t -
p(t)
for
e'/i3
t::2: to if f3 = 0
for t::2: to if f3 > O.
Our next result provides sufficient conditions for the oscillaion of (5.1.3)
when condition (5.2.2) fails. For this, we need to define the following functions
g(t)
p"(t)
p(t)
and
h(t)
p'(t)
p(t) -
(1(3) a'(t)
~ a(t) ,
t::2: to (5.2.9)
(5.2.10)
t::2: to
Theorem 5.2.4. Suppose condition (F) holds, and let p(t) E C 2 ([to, =),
1R+), D(t) E C([t o,oo),1R), a(t) E C 1 ([to,=),1R+) and p(t) E C([to,=),
1Ro) so that
1
lim sup -T
T->oo
rt
T
lt lt
for every
t::2: to.
(5.2.11)
= max{D(t), O},
= S (P((S)))2 ds <
a s
=,
(5.2.13)
395
00,
t-HX)
(5.2.14)
x(t)
01 0
_
w'(t) -
p'(t)
x'(t)
!3
and
w"(t)
+(
p!3(t))'
x'(t)
aCt)
aCt) f(x(t))
(~)2
f(x(t))
p'(t) ,
+ 13 pet) w (t) + 13
(P'(t))'
pet) wet)
p!3(t)
!3 pet) x'(t)
- aCt) q(t) - p (t) aCt) f(x(t))
_f'(X(t))F(X(t))w~t)
[Wl(t)
-f3~g?W(t)r
pet)
pet)
Using the definition of the number 13 and applying the method of completing the square, we obtain for t ~ to,
~(t) (p(t)) ~
aCt)
f(x(t))
(113 ) 1 [,
{ (t) (1 - 13) (t) } ]
13 wet) w (t) - pet) - 2i3 aCt) wet)
13 [pll(t)
pet)
( 1-
a'
or
w" (t)
<
ret)
- aCt) q(t) -
~(t) aCt)
pet) x'(t)
f(x(t))
+f3g(t)w(t) -
(5.2.15)
Chapter 5
396
- liT,
t) w (t)
we obtain
iT is --q(u)duds
pf3(u)
a(u)
x'(u)
1 iT is
+ T1 iT is ~(u) (p(u))
a(u) f(x(u)) duds + {3T
-g(u)w(u)duds
-weT)
- - + -wet) + ( 1 T
+(
(3
) 1
1 - (3
IT IS
t
::::: -1
T
1
w(u)
[w'(u) - h(u)w(uWduds.
(5.2.16)
Now, taking into account condition (5.2.13) and using the Bonnet theorem,
we conclude that for any s, t with s::::: t ::::: to, there exists a number
~ E [t, s] such that
x' (u) )
is (_13 (u) p(a(u)u) ) f(x(u))
x' (u)
i s(_13 (u) p(a(u)u)) ( f(x(u))
du =
du
_13 pet) 1~ f(x(u))
x'(u)
_13 (t) pet)
1t x'(u)
= - (t) aCt)
du =
aCt) ~ f(x(u)) du
t
fJ
fJ
= ~(t)p(t)
fJ
(t)
3:JL :::;
pl3(t)p(t) F(x(t))
aCt)
since
-fJ
(t)
fey) <
x()
dy
Pa((tt))w(t),
io
x(t)
dy
fey)
if
x(~):::; x(t)
(t)
dy
x(~) fey)
_ weT)
T
>
+ wet) +
T
< {
10
x(t)
dy
fey)
if
x(~)::::: x(t)
(1-.!.)
[Wl(t) + pet) wet)]
T
aCt)
iT is w(u)
1
[w'(u) - h(u)W(U)]2duds.
t
(5.2.17)
397
w'(t)
:~!~ w(t)
{3 1
+-(3
1-
00
{31O (-g(u))w(u)du
1 [w'(u) - h(u)W(u)]2du
-(-)
w u
+ :~!~ w(t) 2:
w'(t)
D.(t)
for every
t 2: to,
. . w(T)
hl1l1I1f - T < 00,
1.
and
1.
(5.2.19)
T-+oo
00
(5.2.18)
(5.2.20)
to
00
to
1
-(-) [w'(u) - h(u)w(uWdu
W tL
< 00.
(5.2.21 )
1.
to
1.
1.
T [w'(u)J2
iT
iT
( ) du - 2
h(u)w'(u)du +
h2(u)w(u)du
~
wu
~
~
T [w'(uW
( ) du - 2h(T)w(T) + 2h(to)w(to)
w u
to
1:
+
>
(T[h2(u)
+ 2h'(u)Jw(u)du
ito
[w~~~J2 du -
- (c + 2)
1.
2Th(T)
(w~)) + 2h(to)w(to)
(-g(u))w(u)du.
to
) [w'(u) - h(u)W(u)]2du
i toT _(1
w u
2:
1.
00
to
[W'~Ur
du - 2 [lil1lS11 P Th(T)]
w u
T-+oo
+ 2h(to)w(to) -
[lil1linf WT(T)]
T-+oo
(c + 2) {oo (-g(u))w(u)du,
ito
Chapter 5
398
which because of (5.2.14) and (5.2.19) - (5.2.21) yields
00
. to
[u,t(u)F
( ) du <
W U
w(t)
:::;
:::;
to, we have
+ {[w(tW/ 2 - [W(tO)P/2}
[[u,(toW/2
:::;
(5.2.22)
00 .
2w(to)
+ 2 [[w(t)]1/2 -
2w(to)
2w(to)
+~
2w(to)
and consequently,
[w(to)p/2f
u,'(u)
(It. It
1
+ -t
11
2w(to)
( ) <t { wt
- + -2
to
-
[It.
00
to
dU)
t.o
00
t.o
t.o
]2
[w'(u)F du
'w(u)
[w'(ulF
( ) du
W U
[w'(u)F du }
w(u)
for
to.
(5.2.23)
where
= 2w(to) + ~
to
00
t.o
[w'(s)j2 ds <
w(s)
00.
t ~ to. (5.2.24)
399
p(t) = p(to)
it
to
~ to.
00.
1
lim sup -T
T-4OO
iT 1
8
pJ3(u)q(u)duds
D(t)
for every
~ to.
(5.2.25)
~ to
and
00.
Corollary 5.2.5. Let condition (F) hold, and let p(t) E C 2 ([to, (0), lR+),
and D(t) E C([to, (0), lR) be such that condition (5.2.25) hold. Then
equation (5.1.1) is oscillatory if conditions (5.2.12) and (5.2.24) are satisfied.
Corollary 5.2.6.
Let condition (F) hold, and let
DE C([to,oo),lR) be such that
liTiS
lim sup -T
T-4OO
uAq(u)duds ~ D(t)
and
t ~ to
for every
p(t)
= tA/ J3
1
liminf -T
T-4OO
iT 1
8
p/3((u)) q(u)duds
tau
D(t)
for every
~ to.
(5.2.26)
[it
to
Sh 2(S)dS]
-lit
to
[D+(S)]2 ds
S
00
(5.2.27)
400
Chapter 5
and
limsup
t-4CXJ
[I Sh2(S)dS] -ll
t
to
to
s (P((S)))2 ds <
a s
00.
Proof.
x(t)
i-
w'(t)
+ :~~~ w(t)
2:
O(t)
+--(3(3 1 -(-)
CXJ
1-
Also,
lim sup
(31
CXJ
(-g(s))w(s)ds
T-4CXJ
we::)
<
(5.2.28)
00
T-4CXJ
also (5.2.18), (5.2.20) and (5.2.21) hold. Furthermore, for T 2: to, we find
T [w'(s)JZ ds =
) [w'(s) - h(s)w(sWds + iT 2h(s)w'(s)ds
i toT ~(
w(s)
w s
to
to
_ (T h2(s)w2(s)ds
Jto
LT
w~s) [w'(s) -
h(s)w(s)J 2ds
+ 2Th(T)
< loT
-I
(w~))
h2(s)w(s)ds
to
w~S)[w'(S)-h(s)w(sWdS+2Th(T) (w~))
lT
to
~(
) [w'(s) w s
-2h(to)w(to)
to
h(s)w(s)fds + 2M(Th(T))
to
where
sup w(T)
T?to
<
00
(by (5.2.28)).
(5.2.29)
401
[w'(sW
,( ) ds ::::: N
to
lt
to
11.. 8
Sh2(S)ds
for
t 2:: to.
(5.2.30)
to 1L S
+ P((S))W(8)12 d8
a s
to
a s
to
a s
to
[I Sh2(S)dSl-llt [!1+(sW ds
t
to
to
r1> (~~:~r
1
ds <
==
If a(t)
Remark 5.2.3.
reduces to
p'(t) 2:: 0
00,
and
p"(t) ::::: 0
t 2:: to.
for
(5.2.31 )
liminf -T
T-+oo
j.T IS pf3(u)q(u)duds
t
2:: !1(t)
[I
to
(p'(S))2
-(-) ds
P
for every
t 2:: to.
j-l lt
to
[!1+(s)j2
S
ds
(5.2.33)
00
[I
to
(5.2.32)
to
00.
Chapter 5
402
Corollary 5.2.8. Let condition (F) hold, and let pet) E C 2 ([to, (0), lR+)
and net) E C([to, (0), lR) be such that condition (5.2.32) hold. Then
equation (5.1.1) is oscillatory if conditions (5.2.31) and (5.2.33) are satisfied.
Corollary 5.2.9. Let condition (F) hold, and let n E C([to, (0), lR) be
such that
liTiS
liminf -T
T->oo
for every
to.
(5.2.34)
it
hmsup t->oo In t to
[n+(s)F
S
ds =
(5.2.35)
00.
to.
t S
to
it
(p'(S))2 ds ~ b2
pes)
to
ds = b2 [lnt- Int ol
< 2b2lnt,
(II) f(x) = Ixl"! sgn x + x, x E lR (0 < 'Y < 1) with 'Y/2 ~ /3 < I,
(h) f(x) = Ixl l / 2 sgn x/(I + Ixl l / 4 ), x E lR with /3 = 1/4,
(13) f(x) = Ixl"![k+sin(ln[I+lxl])] sgn x, x E lR (0 < 'Y < I, k ~ I+(1h))
with (k -I)("(k - 'Y -I)/(k + I)(k + 2'Y) ~ /3 < 1.
Also, see Examples 5.2.1 - 5.2.3.
ExaIllple 5.2.6. Consider the differential equation
(to:x'(t))'
+ tOx'(t) + t'>'(sint)f(x(t))
t ~ to> 0
0,
(5.2.36)
get) =
a2
[ 4/3
2
1)
a - /3] (1-/3) ( t 2
and h(t)
/3 - a(I2/3
(1)
/3)] t '
403
>0
and
1jTjS
to, we obtain
+~
[t i5 sin t
+ 28t"-1 cos t -
T---+oo
jT
s"-3cos sds.
to, we find
1jTjS
lim sup -T
-0(0-1)(0-2)
Thus, for t
1jT
+ (1 - ~ )
+ 0(0 -
uf3 q (u)duds
1jT j"
lim sup -T
T---+oo
> t i5 cost -
00
s"-3cossds
Jl,
n=N
2mr +(1l'/4)
2mr-(7I'/4)
-1
LIn
2
00
n=N
-ds
S
2]
1+-8n-1
00,
Chapter 5
404
(5.2.37)
0,
~lT1S rJ(u)q(u)duds
- sinT + (1-
~)
~lTls usinududs
[tcost - sintJ +
1l
T
-=----'-~ds
to
>
>
1
1
2n7r-(7r /4)
S
2n7r
1
+(7r/4)
sds
8 2n7r-(7r/4)
and therefore,
.
1
hmsup t--+oo In t
it
to
1
n--+oo In [2n7r + ~]
7r 2 n
> lim
- n--+oo 8In [2n7r + ~J
[n+(s)j2.
ds > hmsup
S
ds
to
ds
00.
0,
2: to> 1. (5.2.38)
405
",
(t) =
h1(t) = f'(x)F(x)
gl(t)
f3P;~~~) -
f3",(t)
[ I I p ' (t)
2",(t)
~g; -
h () = 1-13 [~ ()
2 t
13
2'" t
and
(~gD
+ ]'(x)F(x)hi(t),
~f3(l+f3) pl(t)]
+2
1 - 13
p(t)
[1 + L]
(p'(t))2 + _f3_ h 2(t)].
1 - 13
p(t)
1 - 13
2
h~(t):S
- Cg2(t)
and
h~(t)
+ h~(t) 2: g2(t)
t~oo
(5.2.39)
for some constant c> 0 and t 2: to. Then equation (5.1.3) is oscillatory
if there exists a function 0 E C([to, 00), IR) satisfying (5.2.12) and
1
lim sup -;t~oo
Jt (t - s)n_(
p!3(s)
-)
T
a s
q(s)ds 2: O(T)
for every
T 2: to. (5.2.40)
Proof.
x'(t)
pf3(t) f(x(t))
p!3(t) a(t)x'(t)
a(t) f(x(t))
w'(t) -
f3~g} w(t),
406
Chapter 5
a(t)
f(x(t))
p{3(t)
f3 p(t) x'(t)
1 [,
P'(t)] 2
- a(t) q(t) - p (t) a(t) f(x(t)) - W(t) W (t) - (3 p(t) W(t)
x f'(x(t))F(x(t))
(5.2.41)
p(t)
p(t)
w(t)
-(3[1
p(t)
w(t)
+ (3f'(x(t))F(x(t))]
(32 f'(x(t))F(x(t))
(~gD 2w(t)
p(t)
p(t)
pf3(t)
1
,
2
a(t) q(t) - (-91(t))W(t) - w(t) [w (t) - hI (t)w(t)]
X
f'(x(t))F(x(t)).
t:::: to,
(5.2.42)
(5.2.42) takes
(5.2.43)
(t -
:s; -
s)nP:(~) q(s)ds +
{3
1-,u
--(.I
ht
ht
T
(t - s)n( -92(S))w(s)ds
1
(t - sr-(-)
[w'(s) - h2(S)W(sWds
WS
(t - s)nw"(s)ds = (t - T)nw'(T) - n
(t - T)w'(T) - w(t)
ht
+ w(T)
ht
(t - s)n-1w'(s)ds
if n = 1
407
~ t
tn
iT
(t - st p!3((s)) q(s)ds + ~
(t - s)n( -g2(s))w(s)ds
as
t n iT
iTt (t -
+ -.L~n
l-(3t
(1-
(1 - f)
<
sr-1-[w'(s) - h2(S)W(s)fds
w(s)
if n = 1
1) t: It (t - s)n- 2w(s)ds
w'(T) - n(n -
T)n-l w(T)
+n ( 1 - -t
t
Thus, for every T 2: to, we get
1
lim sup --;:;
t---'>oo
(3
+ --(3
1-
It
1
pf3(s)
(t - s)n_(
-) q(s)ds +
00
00
if n > l.
(-g2(s))w(s)ds
if n> 1
w'(T) 2: O(T)
(3
-g2(s)w(s)ds + -{3
00
1-
liminf w(t)
t---'>oo
00
-(-) [w'(S)-h2(S)W(S))2ds
W S
if n = 2
n(n - 1) liminf -1
t---'>oo t n
lt
T
(t - s)n-2w(s)ds if n > 2.
-g2(s)w(s)ds <
00
00,
to
ito
00
and
liminf w(t) <
{
t---'>oo
liminf
~n t
t---'>oo
ito
00
if n
=1
(5.2.44)
00
if n> l.
408
Chapter 5
. f w(t)
1Imlll
-- <
t->oo
t
(5.2.45)
00
TI : : :
-1
tn
it
to
jt
jt (
at
1
-;;;
>
(t - s)n- 2 w(s)ds
(t - s)n- 2 w(s)ds
Tl
P, 1
> --;;;
T,
t-s )n -2sds
T )
1 (
; [ n-1 1--f
Since,
n-I
lim [ -1- ( 1 - T
-.!. )
- -1 ( 1 - T
-.!. )
t->oo n - 1
t n t
we can choose a T 2
:::::
n(n - 1)
TI so that
n]
t n t
>
for
all
> a,
t::::: T2.
-1
it
t n to
which gives
lim -1
t->oo t n
since p,
follows.
it
to
for every
(t - s)n-2w(s)ds
t::::: T2 ,
00,
'YI(t)
'f
(3P'(t)
p(t) ,
h (t) =
2
p'(t)
p(t)
and
p"(t)
g2(t) = (3 p(t) .
(5.2.46)
409
(Fl
Proof.
x(t)
i=
ll l
T
lim -T
T-+oo
to
and define
Q(t)
If
l
lim -T
T-+oo
u"Yq(u)duds
[T [S u"Yq(u)duds
00
to
to
-[Q(s)
S
+ k(S)]2ds
1R
(5.2.47)
t 2: to.
(5.2.48)
exists in
for
00
(5.2.49)
limHoo k(t) = 0,
then
410
Chapter 5
and
w'(t)
'-t-
w" (t)
- 'wet)
T + ,{y-1x-'Y(t)x'(t)
+ t'Y x-, (t )x" (t)
+, [Wl(t)
,Wl(t) _,wet)
t
t2
_,wet)]
t2
__
, __1_
l-,w(t)
t'x-,(t)x"(t) _ _
, __1_ [Wl(t) _ wet)] 2
l-,w(t)
t
Thus, it follows that
-w"(t) = t'Yq(t)
This gives
_ weT)
T
u'q(u)duds
+1
=, ~ iTis w~u)
[WI(U) _
w'(t)
t 2: To.
for all
(1 _.!.)
w/(t)
T
+ wet) +
iTis
l-,w(t)
11 1
, 1 -+-lim _T
T-too
To,
duds
we obtain
weT)
+ T-too
lim T
[,
W(S)]2
w (s) - - ds
00
1-,
u'q(u)duds
w~u)]
w(s)
and consequently,
<
lim weT)
T-too
and
00
to
w(s)
(5.2.50)
00
<
00.
(5.2.51)
Hence, we have
w'(t)
where
G(t)
Q(t)
lim wT(T)
T-too
_'_1
1-,
00
+ G(t)
for every
t 2: To,
t 2: To.
(5.2.52)
411
Now, we set
w(T)
k(t) _
-
T--+oo
w(t)
+ G( t)
lim - T
w(T)
+ G(to)
lim - T
T--+oo
if
- --
w(to)
- --
> to
if
to
In view of (5.2.50) and (5.2.51), k(t) E C([to, (0), JR) and limHoo k(t) =
O. Furthermore, from (5.2.50) there exists a positive constant. K such
that
w(T)
-T :::;
T 2: to
for all
(5.2.53)
[WI(S) _ W(S)]2 ds
roo _1_
.fTo w(s)
roo _(1)
.fTo W S
00
To
[Q(s)
+ G(s) _ W(S)]2 ds
lim w(T)
T
T--+oo
1
-(-) [Q(s)
w s
+ k(s)J 2ds 2: K
00
To
-[Q(s)
S
+ k(s)fds
001
-[Q(s)
to
+ k(S)]2ds
:::;
TO
to
+K
+ k(sWds
-[Q(s)
s
lTo
<
x"(t)
00,
sgn x(t)
0,
t 2: to > 0
(5.2.54)
where 0 < 'Y < 1 and - 'Y < .\ < 1 - 'Y. We set 5 = 'Y +.\ so that
0<5 < 1. Now, for any T, t with T 2: t 2: to, we have
T1
jT jS u'q(u)duds
t
T1
jTjS
t
Osinududs
+ 5(5 -
sin T
5 (5 - 1)( 5 + 1) t li - 2 sin t]
1jT
-5(5-1)(5+1)(5-2)-
1)(5 + 1)TIi -
sli-3
sinsds
Chapter 5
412
lT
Jt8-I
and consequently,
Q(t)
T-too
urq(u)duds
t ii cos t
+ ",(t)
for all
t:::: to,
where for
",(t)
11 1
lim -T
-JtO-1sint-J(J-1)tO- 2 cost-J(J-1)(J-2)
00
t:::: to,
sO-3cos sds.
o.
00
to
-[Q(s)
s
00
to
-1
L
8
+ k(s)J 2 ds
00
n=N
2n 1l'+(1l'/4)
2n1l'-(1l'/4)
-ds
S
2n7l'+(1l'/4)
00
n=N
1L In
= -
00
n=N
2n1l'-(1l'/4)
2]
1+-8n - 1
-1 ( v21
M -
2v2
)2ds
00.
Q*(t) =
t Q(s)ds
lto
- 315(15 - 1)
ft
ho
= t 8 sin t
it~ 1s
00
2Jtg- 1 cos to
u li - 3 cos ududs
and hence the function Q*(t) is unbounded. Therefore, Theorem 5.2.8 can
be applied to equations of type (5.1.6) without any additional hypothesis
that Q*(t) is a bounded function.
413
Theorem 5.2.9. Suppose condition (F) holds, and let p(t) E C 2 ([to, (0),
JR+) be such that for some constant c> 0,
(5.2.55)
iT is
and
lim -1
T->oo
to
to
rJ3(u)
(-) q( v,)dv,ds
a u
Define
1
lim -T
T->oo
exists in
IT IStau
rJ3(u) q(u)dv,ds
- (-)
JR.
(5.2.56)
t 2: to
for
- qo >
t ..... oo
qo
00,
is a positive constant.
001
ta
-;-[Q1(S)
+ k(sWds
(5.2.57)
= 0,
(5.2.58)
00,
Proof.
x(t) =I-
To,
with T 2: t 2:
we have
(1 _i)
w'(t)
T
..!:..jT
IS p{3(u)
q(v,)dv,ds
T
a(v,)
1
+ T1 jT IS -gl(v,)w(v,)dv,ds + T1 jT jS w(u)
[w'(u) - hI (u)w(uW
_ w(T)
T
+ w(t) +
T
x f'(x(u))F(x(u))duds
w'(t)
1
lim -T
T ..... oo
00
Il
T
+I
t
OO
-
g1 (u)w(v,)dv,
. w(T)
11m
-- <
T->oo
00,
(5.2.59)
414
Chapter 5
roo -gl(u)w(u)du
<
(5.2.60)
00
JTo
and
00
To
100
t
-gl(s)w(s)ds+
100
t
(5.2.61)
00.
w(s) [w'(s)-hl(s)w(s)ff'(x(s))F(x(s))ds,
we find
w'(t) = Ql(t)
lim WT(T)+G(t)
forevery
'1'--+00
t2:To.
(5.2.62)
Next, we define
k(t)
weT)
T
{
lim weT)
'1'--+00
T
hm - -
T--+oo
wet) .
If t > To
t
+ Gl(To) _ w(To)
To
+ Gl(t)
- -
and observe from (5.2.59), (5.2.60) and (5.2.61) that k(t) E C([to, 00), JR.)
with limt--+oo k(t) = O. Moreover, in view of (5.2.59), there is a constant
K > 0 such that
we;:)
for all
T 2: To.
(5.2.63)
+ k(t) 2: -
f-t
for all
-qo
2: To.
>
-00
(5.2.64)
JTo w s
2: K(/- (3)
! ~[w'(s)
rt
- hl(S)W(sWds
(3
1 [
K(l- (3) JTo :; Ql(S)
K(/- (3)
: ~
(T)
+ )~oo w T + G1(s) -
{[Ql(S) + k(s)] +
D-
h1(s)w(s)
2
hl(S)] w(S)} ds
]2
ds
[1
>
K(:-
(t
h 1(S)]
-s- w(s)ds + iTo
s2 -
(3)
[1
h 1(s)]2 2
-s- w (s)ds
S2 -
415
2/L
h~
~[QI(S)+k(s)fdS-2/LC,( -91(S)W(S)dS]
For the special case when a(t) = 1, p(t) = 0 and p(t) = t, we find
h ()
1
= f!..
t
[1 +!'(x)F(x)
f'(X)F(X)]
91(t)
and
(32
t2
[1~ -
1]
1 - f'(x)F(x)
1 (T
lim -T
r uf3q(u)duds
Q2(t)
1
lim -T
T-+=
iTt is
exists in
uf3 q(u)duds
for
ffi..
t:;::: to
and suppose that liminft-+= Q2(t) > -00. If for every k(t) E C([to, 00),
ffi.) with limt-+= k( t) = 0,
00,
x"(t)
0,
t:;::: to> 0
(5.2.65)
Chapter 5
416
Q2(t) =
lim
T-+oo
ITI
uf3 q(u)duds =
lITIs
lim -T
T-+oo
sinududs = cost,
00
12n11"+(11"/4)
n=N
2n11"-(11"/4)
1L
"8
00
n=N
1
1]2
-1 [ds
)2 2)2
S
2]
In 1 + 8n _ 1
[
00.
Thus, all conditions of Corollary 5.2.12 are satisfied, and therefore, equation
(5.2.65) is oscillatory.
The following result not only extends Theorem 5.1.6 to more general
equations of the type (5.1.1), but also removes the condition Q(t)
!too q( s )ds :::::
for t::::: to.
it
lim
sf3- 1 Q(s)ds
t-+oo to
00,
ft q(s)ds exist
OO
(5.2.66)
lim f(x)
x-+oo
= 00.
(5.2.67)
In fact, from condition (F) and the definition of F(x), we have F'(x) =
l/f(x) and
(5.2.68)
417
and hence
( f(x)
f(xo)
)C
F(x)
F(xo)
> --
(5.2.69)
(5.2.70)
x'(t)
f(x(t))
Q(t) +
oo
x'(s)
f(x(s))
)2 f ,(x(s))ds,
(5.2.71)
In particular, we have
00
x'(s)
f(x(s))
to
)2 J'(x(s))ds
<
(5.2.72)
00
from which we shall show that x(t) has the following asymptotic behavior
lim F(x(t))
t-+oo
= o.
(5.2.73)
F(x(t)) - F(X(t1))
<
li
tl
x'(s)
f(x(s)) ds
[i t(f~X(:))
'())2
tl
]1/2
J'(X(S))dSJ
[it
tl
f'(x(s)) ds
]1/2
(5.2.74)
418
Chapter 5
for t 2': tl 2': to. In view of (5.2.72), we choose tl sufficiently large so that
for any E > 0,
00 (
x' ( s ) ) 2 I
E
f (x(sds <
f(x(s
t,
4'
F(x(t
F(X(tl
;E 1, F(x(sds )1/2
fo (
(5.2.75)
;E 1, F(x(sds )1/2
~ fo (
F(X(tl
F(x(t
fo
(1:
for
F(X(SdS) 1/2
(5.2.76)
1,
F(x(sds
) 1/2
1,t?
F(x(sds
) 1/2
~ fo (t _ t2) < fo t
22'
(5.2.77)
(t 2 F (X(sds <
itl
~Et2
t2': t3.
forall
Thus, from (5.2.76) and (5.2.77), (5.2.73) follows immediate. Now, we define wet) = t iJ - l F(x(t)), t 2': to. Then for t 2': to, we have x'(t)/ f(x(t =
tl-iJw'(t) + (1 - (3)c;jw(t) and
I
f (x(t
x'(t)
f(x(t
)2
(5.2.78)
Integration of the second term on the right-hand side of (5.2.78) from t
to T 2': t 2': to yields
w~) +{31TW}+s~dsl.
S"
(5.2.79)
419
In view of (5.2.73) and the fact that w(t) is nonnegative the limits of
the two integrals in (5.2.79) exist and can be infinite. Noting (5.2.72) and
(3 < 1, we can integrate both sides of (5.2.78) and deduce that the limit
on the left-hand side of (5.2.79) must be finite. Hence, the integral on the
right-hand side of (5.2.79) is likewise finite, and therefore, we can define an
energy function
E(t)
w(t) - (1
t 2 to.
(5.2.80)
Recall that x(t) is a solution of (5.1.1) and satisfies the integral equation
(5.2.71). Computing E'(t), we find
x'(t)
E'(t) = t f3 - 1 _
_
f(x(t))
(3({3 + 1)tf3 - 1
00
w(s)
ds
SM1
t 2 to.
for
(5.2.81)
Combining the second and the last terms in (5.2.81) and using (5.2.71), we
obtain
E'(t)
tf3-1 [Q(t)
-(3(1
00
x'(s)
f(x(s))
)2 f'(X(S))dS] + 2{3w(t)
t 2 to
(5.2.82)
where (3 = 1/(1 + c) < 1 and c = (1 - (3)/(3 > O. Furthermore, we
apply (5.2.73) and combine (5.2.78) and (5.2.79) with (5.2.82), to obtain
for t 2 to,
E'(t) 2 t f3 - 1Q(t)
1 f3 - 1
+ _t
c
00
Integrating this inequality and using condition (5.2.66), we get limHoo E(t)
= 00. From the definition of E(t) in (5.2.80), we also deduce that
limHoow(t) = 00. Define tn = sup{t : w(t) ::; n}, which satisfies
limn-+oo tn = 00 and also t 2 tn implies w(t) 2 n = w(t n ). Replacing
t in (5.2.80) by tn and using its definition, we obtain
Chapter 5
420
which contradicts the fact that limn-too E(t n )
proof.
= 00.
(5.2.83)
where 0 < 'Y < 1, t 2': to> O. Let f(x) = Ixl'Ysgn x so that c = (l-'Y)h
and fJ = 'Y. Also,
Q(t)
1=
q(s)ds =
1
-[l+vtsint],
t'Y
t 2': to
It is easy to check that all hypotheses of Theorem 5.2.10 are satisfied and
hence equation (5.2.83) is oscillatory.
We note that Theorem 5.1.6 fails to apply to equation (5.2.83) since the
function Q(t) assumes negative values for all large values of t and satisfies
condition (5.1.25). Therefore, we conclude that Theorem 5.2.10 improves
Theorem 5.1.6.
Example 5.2.12. Consider the differential equation
x"(t) +
t~
G ~vtsint
+
(5.2.84)
We let
(5.2.85)
f;
so that F(x) =
du/ f(u) = 2 tan-1..jX. Clearly, condition (F) is satisfied
with fJ = 1/2, i.e., c = (1 - fJ) / fJ = 1. All conditions of Theorem 5.2.10
are satisfied and hence equation (5.2.84) is oscillatory.
It is interesting to note that f(x) in (5.2.85) satisfies both conditions
of sublinearity, i.e., (5.1.5) as well as of superlinearity, i.e., (4.1.5).
min
{inf
f'(x) r fd(U) , inf f'(x) jX fd(U)}
x>o
J+ o
x<o
-0
> O.
(5.2.86)
(a(t)p(t))':::; 0,
~(t)=
a(t)p'(t)
= p(t)p(t)
() ()ds-too
toasps
as
for
t-too.
421
t-400
and
liminf - 1
t-400
tm
it
to
it
to
( )1 ( )
asp s
[is
to
p(u)q(u)du
(t - s)mp(s)q(s)ds > -
ds
00
00
(5.2.88)
m21,
(5.2.89)
-p(t)q(t)
+ (a(t)p'(t) -
x'(t)
p(t)p(t)) f(x(t))
It:
t 2 to
(5.2.90)
a(s)p(s)(w'(s))2 f'(x(s))ds
Case 1. I(to, 00) < 00. In view of (5.2.86), there exists a positive constant
a such that
f'(x(t))w(t)
ret)
f'(x(t)) J+o
du
f(u) > a
w(t)
hm -(-)
t-400 ~
for all
t 2 to
(5.2.91 )
(5.2.92)
O.
11: w'(s)dsl
~ (.:r(tl, t))1/2
for all
t 2 tl'
(5.2.94)
Chapter 5
422
If J(tl, (0) < 00, (5.2.94) ensures that wet) is bounded, and so (5.2.92)
holds. Thus, we restrict our attention to the case when J(tl, (0) = 00. In
this case, we can choose a
> h so that W(tl)::; (-JfJi/2) (J(tl,t1/2
for all t ~ t2. Thus from (5.2.94) it follows that wet) ::;
(J(tl,
for all t ~ t 2 , and so, by using (5.2.91), we have
t2
t1/2
..;w.
(5.2.95)
or
{
L
t
}-1/2
w(s)
a(s)p(s) ds.
vE
w(t)::;
~
for all
t2, we obtain
() d }1/2
ws
{i a(s)p(s) ds }1/2 - {it? a(s)p(s)
t
()
WS
t,
tJ
~ t 2
<
-
vE
to
a(s)p(s)
ds.
Hence, by setting
max { t2,
t3
for t
{l
-IE
t2
t,
w(s)
a(s)p(s) ds
}1/2}
t3, we obtain
Thus, from (5.2.95) it follows that wet) < E~(t) for all t ~ t3. As E > 0
is arbitrary, the proof of (5.2.92) is complete. Now, from equation (5.2.90),
p(u)q(u)du = C l - a(t)p(t)w'(t) -I(to, t), where
for t ~ to, we have
C1 = a(to)p(to)w'(to). Hence, for every t ~ to, we find
It:
[1:
P(U)q(U)dUf
[Cl
+ 3 [I(to, (0)]2.
423
to,
it
s::
3 to a(s)p(s)(w'(s))2ds
C2 + ~(t)
s::
C2
+ (3/(;))
~
a(s)p(s)(w'(s))2 f'(x(s))w(s)ds
.Jto
~(t)
to:::;s:::;t
t a(s)p(s)(w'(s))2 f'(x(s))ds
.Jto
to,
itto a(s)p(s)
1
[isto p(u)q(u)du] 2 ds s::
C2
+ C3
(5.2.96)
where C 3 = (3/a)I(to, (0) > O. But, by (5.2.92) we can choose T ~ to
so that w(t) s:: ~(t) for every t ~ T. Therefore, maxto<s<t w(s) s::
maXto:::;s:::;T w(s) + ~(t) for all t ~ to. Thus, (5.2.96) gives - -
for all t
limsup C(l)
t-.oo
."
it
to
()1 ( )
asp s
[is
to
p(u)q(U)dU] ds
s::
C2 +C3 <
00,
to, we have
-1t
+ 1t
it
to
00.
to
to
(t - s)m(a(s)p(s)w'(s))'ds
(t - to)a(to)p(to)w'(to) - a(t)p(t)w(t)
to
(a(s)p(s))'w(s)ds
(t - to)ma(to)p(to)w'(to)
+m
to
for m
+ a(to)p(to)w(to)
=1
+ m(t -
to)m-la(to)p(to)w(to)
424
::;
Chapter 5
(t - to)a(to)p(to)w'(to) + a(to)p(to)w(to) if m = 1
(t-to)ma(to)p(to)w'(to) + m(t-to)m- 1 a(to)p(to)w(t o) if m 2: 2.
() m
()m-1
(5.2.97)
But, since
lim
t-400
~
i t (t m
t
s)ma(s)p(S)(W'(S))2 f'(x(s))ds
to
I(to, (0)
00.
t-400
tm
to
= -
00,
Corollary 5.2.13. Let conditions (5.1.5) and (5.2.86) hold, and a'(t) ::; 0
for t 2: to and ~l(t) = Itto ds/a(s) ----+ 00 as t ----+ 00. If
1
lim sup -(-)
t-400
and
liminf - 1
t-400
tm
it
to
6 t
it [is
to
1
-(-)
a s
to
(t - s)mq(s)ds > -
q(u)du
00
ds
00
(5.2.98)
m 2: 1, (5.2.99)
Corollary 5.2.14. Let conditions (5.1.5), (5.2.86) and (5.2.99) hold, and
limsuP~lt
t
t-400
to
[is
to
q(U)dU] ds
00,
(5.2.100)
425
Remark 5.2.6.
1. The condition
lim sup -1
t
t--+oo
jt.jS q(u)duds
to
(5.2.101 )
00
to
t--+oo
jt IS q(u)duds >
(5.2.102)
00.
to. t.o
Now, we can state that equation (5.1.6) with 0 < --y < 1 is oscillatory
if either one of the following holds:
I1 l
to
to
I1t I
to
to
Next, we shall prove the following more general result for (5.1.3). It
extends the criterion (Id above.
Theorem 5.2.12. Let conditions (5.l.5) and (5.2.86) hold, and assume
that there exists a function p(t) E C 1 ([to,00),lR+) such that conditions
(5.2.87) and (5.2.89) are satisfied. If, for some integer n > I,
lim sup -1
t--+CXJ
tn
and
lim sup -1
t--+=
tn
It
(t - s)rlp(s)q(s)ds
(5.2.104)
00
to
It
(t - s)n-2~(s)ds
< 00,
(5.2.105)
to
t :::> to which
a(t)p(t)w'(t), t :::>
to.
(5.2.106)
,
z (t)
J'(x(t)) 2)
a(t)p(t) z (t
+ p(t
q(t) = 0,
t :::>
> 0 for
As in Theorem 5.2.11 we consider the two cases I and II. The proof of case
II is exactly the same as earlier and hence omitted. Now, for the case I
426
Chapter 5
z'(t)
for
t 2:: to
tl
w(t)
s:
~(t)
for
(5.2.107)
2:: tl.
2::
tl
we have
(5.2.108)
s:
for
t 2::
(5.2.109)
tl.
or
it
t,
(t-s)np(s)q(s)ds
s:
t)t
1 - .2
Z(tl)+ ( -
40
t,
it
to
n2 c
] ds = 00, (5.2.111)
(t - st- 2 (t - s)p(s)q(s) - -~(s)
4
427
Proof.
1~ (t -
s)n-2
[(t - s)2p(s)q(s) -
~~ ~(s)] ds
(1 _t; ) n z(t
::;
1 ).
It:
it
it
1
limsup--;t
t-->oo
and
1
lim sup --;-
t-->oo
or
limsup :
t-->oo
to
to
to
(t-s)nq(s)ds
(5.2.113)
00,
where c = (1 - (3)1(3,
(5.2.112)
00
00,
(5.2.114)
Corollary 5.2.16. Let either condition (F) hold with (3 > 0, or conditions (5.1.5) and (5.2.86) hold. If conditions (5.2.99) and (5.2.112) hold,
then equation (5.1.1) is oscillatory.
Corollary 5.2.15 follows from Theorem 5.2.11 (Theorem 5.2.12) by letting pet) = 1 and pet) = 0, while Corollary 5.2.16 follows from Corollary
5.2.15 by letting aCt) = 1 and noting that for n> 1,
lim -1
t-->oo
tn
it
t1
n n- 1
) <
00.
Remark 5.2.7.
It is known that (5.2.112) alone is sufficient for the
oscillation of sublinear equation (5.1.6). Also, from Theorem 5.2.12, we
note that the condition
lim -1
t-->oo t n
lt
to
(t - s)nq(s)ds =
00
n> 1
(5.2.115)
Chapter 5
428
alone is sufficient for the oscillation of (5.1.6) with 0 < 'Y < 1.
We also note that conditions (5.2.112) and (5.2.115) can be improved
by allowing n to be any real number > 1 and the proof remains the
same.
Remark 5.2.8. In Theorem 5.2.12 the restriction of selecting the weight
(it:
= 0, 0 < 0: < 1,
t 2: to
= 'IT /2.
(5.2.116)
Here, f(x) = Ixl"sgn x + x, f'(x) = o:lxl"-l + 1 > 0 for all x f. 0, and
F(x) = fJxl du/(u" + u). To estimate F(x) from below, we observe that
F x
()
fixi ~ >
u'" + u -
fix
10
10
and
F(x) =
Therefore, for x
1xI
f.
-du- >
u'"
+u
11
0
Ixll-"
2(1 - 0:)
-du =
1
2(1 - 0:)
2u
2u a
2u"
if
Ixl S; 1,
if Ixl 2: 1.
J'(x)F(x) >
2(1 - 0:)
2(1 - 0:)
{(t)
1t
(1f/2)
= t5/ 6 .
1
a(s)p(s)
Then for t 2: to =
ds
1t
(1f/2)
SI/6ds
'IT /2,
= -7
6
e/6 -
'IT
7/6]
C)
429
to,
s)
2"
t
s)
(11"/2)
(11"/2)
sin 1t +
d [u 1 / 2
s)
(11"/2)
1/
cosu)] ds
coss) _ 2
(11"/2)
(~) 1/2] ds
2
ds
t2
So,
Theorem 5.3.1.
Let condition (F) hold, and suppose there exists a
function p(t) E C2([to, 00), JR+) satisfying (5.2.24). If
lItIS
liminf t-400
to
to
rI(u)q(u)duds > - 00
(5.3.1)
(tv'(t' + C(t)v(t)
(5.3.2)
0,
is oscillatory, where
t,
Chapter 5
430
Proof.
5.2.5, we have
w(t)
hmsup- <
t--+oo
(5.3.3)
00.
We also have
w
jJ-I'(
(t)
(3p
x'(t)
+ PO( t) f(x(t))
'
((
or
z(t),
t;::: to
and
z'(t)
(pO(t)
f~S;)))'
'(t)
Pt
t;::: to.
w t
(5.3.4)
From condition (F) and (5.3.3), we find that there exist a tl ;::: to and a
constant C1 > 0 such that
f'(x(t)F(x(t)) ;::: 1 ~ (3
w(t)
for
t;::: tl
t;::: t]
and
S; CIt
(5.3.b)
z'(t)
where
S;
pO(t)q(t)
/(t)
( z(t) _
~tP/(t))' <
2c p(t)
-pO(t)q(t) _
-~t
[z(t) -
tl
we have
(tP'(t))/
p(t)
2c
~tP/(t)]2
2c p(t)
,
+ (32 t (p/(t))2
t>
-
p(t)
4c
tl
Now, let
y(t) = z(t) -
~tP'(t),
2c p(t)
t;::: tl
so that
y'(t)
S;
+~
2c
(tP'(t))'j_ ~y2(t),
p(t)
t
t;::: tl.
(5.3.6)
Finally, applying Lemma 2.2.1, we find that equation (5.3.2) is nonoscillatory, which is a contradiction. This completes the proof.
431
(tv'(t))'
Proof.
x(t)
-=F
Y (t) ::::::
132-1
- tf3(t)q(t) - 4c t
(t-p(t)(t) ) 2 p'
-y (t)
t
lit is
liminf t-too
(tv'(t))'
to
to
p!3(u)
(5.3.7)
a u
(5.3.8)
is oscillatory, where
P(t) =
(5.3.9)
Chapter 5
432
Proof.
pf3(t) a(t)x'(t)
aCt) f(x(t
z(t),
t;::::to
and
Z
pf3(t)
[aCt) (pf3(t) , pet)]
1 2
,
(t) = - aCt) q(t)+ pf3(t) aCt) - aCt) z(t)- wet) z (t)f (x(tF(x(t.
(5.3.10)
From condition (F) and (5.3.3) there exist a tl ;:::: to and a constant CI > 0
such that (5.3.5) is satisfied. Using (5.3.5) in equation (5.3.10), we find
where
C
(3
CI
(1 _ (3)
and
,
pet)
t;:::: to
we have
pf3(t)
1
,1
2
--q(t) - -(tP(t + -tP (t)
aCt)
2c
4c
-~
[z(t) -
;c
tP(t)f,
t;::::
tl
or
Y, (t) :::;
- [pf3(t)
-(-) q()
t
a t
1 (tP ('
+ -2
t c
1
2')]
C 2( t,
)
-tP
(t - -y
4c
t
t;::::
tl
where yet) = z(t) - (1/2c)tP(t), t ;:::: tl' The rest of the proof is similar
to that of Theorem 5.3.1 and hence omitted.
+ kp(t)q(t)v(t)
(5.3.11)
433
Proof.
z'(t) = - p(t)q(t) _
or
y'(t) = p(t)q(t)
()1 ( ) z2(t)f'(X(t)),
a t p t
+ a(t)p(t)y2(t)f'(x(t)), t
2: to
434
Chapter 5
q(t) 2: 0 for
t 2: to
and
00
to
(5.4.1)
It:
Q( t)
00
q( s )ds
t 2: to.
(5.4.2)
t,
(5.4.3)
00
(5.4.4)
to
o<
At t
=~,
~ =
E [1, {].
(5.4.5)
we have either
x~(~) =
(if ~ = {I)
or
x~(t)
x~({) =
[1,~],
-it
2m
(if ~ = 6).
(5.4.6)
we find
q(s)x'Y(s)ds.
(5.4.7)
lit q(S)X~(S)dSI
I[Q(l) -
Q(t)]x~(t) +
it
[Q(s) -
Q(l)](x~(s))'dsl
it IQ(s)l(x~(s))'ds
(5.4.8)
435
1ft Q(s)(x~(s))'dsl
s;
<
ft P(S)(X'(s))'ds
P(t)x7r,(t) + ft 1F'(s)lx7r,(s)ds.
(5.4.9)
Since Q(t) -+ 0 as t --+ 00, Q(t) is bounded on [1, .;), i.e., IQ(t)1 s;
kI, t E [1,';) where k is a constant. By assumption there exists a constant
k2 such that It' P(t)1 S; k2, t E [1,.;). For t E [1, .;), we also have from
(5.4.5),
ft 1F'(s)lx7r,(s)ds
S;
(2m)'
ft s'IF'(s)lds
k3(2m)',
S;
(5.4.10)
for some constant k3 > O. Using (5.4.9) and (5.4.10) in (5.4.8), we find
1ft q(S)X;n(S)dSI
S;
[3k1
+ k2 + k3](2m)' = K(2m)'.
(5.4.11)
+ K(2m)'
for all
<
t E [1,';].
x~(~)
xl/(t)
+ (t A sin t)lx(t)I'
sgn x(t)
= 0,
t 2: to > 0 (5.4.12)
xl/(t)
+ [t~(lnt)l-'sint]
0,
t 2: to> 0
(5.4.13)
where p, S; ~2. We find that P(t) can be taken to be a multiple of
(lnt)l-'/t. If P(t) E C1([to,00),lRo) and P'(t) 2: 0 for t 2: to such that
< 00,
(5.4.14)
Chapter 5
436
then F(t) satisfies the hypotheses of Theorem 5.4.1, i.e., F(t) = O(t-/,)
as t -+ 00, and (5.4.4) holds. To see this we apply integration by parts,
to obtain
1 F(t)
-t'
'"Y
'"Y
S/,-l
F(s)ds
1
+ -F(I).
'"Y
(5.4.15)
'"Y
i- 1
dq+(s)
00
()
q S
to
< 00.
(5.4.16)
t-+oo
[00 s/'q(s)ds =
t
t-+oo
00
0,
qh+l)h(s)ds
= 0,
lim t 2 q( t) = 0,
t-+oo
(Is)
t-+oo
The equation
x"(t)
+ t 2 (lnt)/' Ix(tW
sgn x(t) = 0,
(5.4.17)
437
(5.4.18)
where ql(t) is a positive nonincreasing function, and q2(t) is a positive
nondecreasing function. Furthermore, if q(t) satisfies (5.4.16), then 0 <
kl < q2(t) ::::: k2 < 00 for all t, where kl and k2 are fixed positive
constants.
Proof. This follows from the identity
Remark 5.4.1.
exp
q2(t) is bounded from below and above by positive constants, we find that
q(t) satisfies anyone of the conditions (Ii)' 'i = 2,3,4,5 if and only if
ql (t) satisfies the same condition. Hence, in the proofs of some of the
results of this section, we can assume without loss of generality that q(t)
itself is nonincreasing.
(5.4.19)
Proof. We shall prove that (14) <=} (15), (12) <=} (15),
(13)
(14). The case (II)
(14) is left to the reader.
'*
'*
o=
lim
t--+oo
'* (15)
(15)
'*
(13) and
l/t
t--+oo -1/t2
t--+oo
Now, suppose that (15) holds, but (14) does not hold. Then there exists a
sequence {t n }, limn--+oo tn = 00 such that
(5.4.20)
Chapter 5
438
By (15), if tn is large enough, then
(5.4.21)
On the other hand, since (5.4.20) holds and q(t) is nonincreasing q(t) 2:
for t E [t n /2, tn]. Integration of this inequality over [t n /2, t n ] yields
(tn/2) It:~/2 q( s )ds 2: J / 4, which contradicts (5.4.21).
Jt:;;2
=?
(5.4.22)
Letting u(t) =
It" s'Yq(s)ds,
t'Y :S -
f2/Cl-'Y)U-2/Cl-'Y) (t)u'(t),
tl
2:
tl.
to t 2: t l , we find
(5.4.23)
and hence
(5.4.24)
Now, an integration by parts from t to T 2: t 2: tl gives
(5.4.25)
It
00
Since
It
q(s)ds
= 0
(f
2/ C
1+'Y)C l )
as
-t 00.
(5.4.26)
439
Now suppose that (Is) holds. Then for any E> 0 there exists t1 2': to
such that f tOO q(s)ds ::; Elt for all t 2': t 1 . An integration by parts from
t to T 2': t 2': t1 yields the following identity
lT
s'Yq(s)ds
t'Y
lT
lT
q(s)ds +,
00
q(s)ds ::;
lT (ITq(~)d~)
s'Y- 1
q(s)ds =
(~)
as
t -+
ds.
(5.4.27)
00
implies that the right-hand side of (5.4.27), and hence the left-hand side,
converges as T -+ 00. Thus,
s'Yq(s)ds < 00. Furthermore, for t 2': tb
00
ftoo
et f-'Y ,
1
(5.4.28)
where E~-'Y = EI(l - ,). Now, for any t 2': t1, we need to consider the
two mutually exclusive subcases (i). q(t) 2': Edt 2 and (ii). q(t) < Edt2.
In the sub case (i), estimate (5.4.27) gives
In the sub case (ii), we set v = (Edq(t))1/2 > t, t 2': tl and observe
that q(t) is non-increasing. Then it follows from (5.4.28) that
qh- 1)/2(t)
Since
00
qh- 1)/2(t)
s'Yq(s)ds
::;
qh- 1)/2(t)
<
E1
h+1)/2
,+1
[l s'Yq(s)ds + 1 S'Yq(S)dS]
V
[1:
00
s'Yq(s)ds +
+ E~1-'Y)/2 =
(~) 1-"1]
O( y'E) as t -+
00.
Case 3. (Is) =? (h). We consider the identity for 0 < , < J.L < 1,
lT
Since
sl-'q(s)ds = tl-'
lT
q(s)ds + J.L
as t -+
00
lT lT q(~)d~ds,
SI-'-l
(5.4.29)
from (5.4.29) it follows that
00,
sl-'q(s)ds
o(tl-'-l).
(5.4.30)
Chapter 5
440
By applying Holder's inequality, we have
r=
.Jt
ql/h+l)(s)ds:::
(1
00
sf-Lq(s)ds
1/h+ 1) (
00
r=
.Jt
s-f-Lhds
)1'/h+ 1)
(5.4.31 )
00,
a (th-1)/h+ l )) .
(5.4.32)
E2
(5.4.33)
For any given t 2: to, we again consider two mutually exclusive sub cases
(i). q(t):;. C2/t2 and (ii). q(t) < Edt2. In the sub case (i), from (5.4.33),
we have the following estimate
(5.4.34)
In the subcase (ii), we set v
creasing, we have
(E2/q(t))1/2
(5.4.35)
On the other hand, using estimate (5.4.33), we find
l/h+1) h-l)/(2h+1))
E2
E2
1/2
E2 .
(5.4.36)
Combining (5.4.35) and (5.4.36), we obtain
qh- 1)/(2(J+l))(t)
1=
2E~/2.
(5.4.37)
Since E2 > 0 is arbitrary, (5.4.34) and (5.4.37) establish the desired assertion (13).
Case 4. (13)
(1+ 1)/(1- 1 )
=}
>
(14). Let
1. Choose
0:
Ea
100
ql/(r+ll(s)ds
441
<
E.
(5.4.38)
Denote f(t) = q/3/ 2 (t), t 2': t1, then fO(t) = ql/{l+'Y)(t). Suppose (14)
fails. Then there exist constants c and 6 > 0 such that c 2': 2t1 and
q(c) 2': 6/c 2 , or
(5.4.39)
where
61
= 61/(r+1) >
(5.4.40)
= 3c/4 that
(5.4.41)
f(t) >
~ 100 r(s)ds
>
[i
r(s)ds
r(S)ds]
g(y) >
~61C-!3 + ~ fY gG(u)du.
4E
Jo
(5.4.43)
G(y)
~61C-i3 + ~ fY G""(u)du.
4E
Jo
(5.4.44)
442
Chapter 5
00
_E (4E) 1/,13
a-I
bl
as y ~ (E/(a - 1(4E/b)I/13 c.
1
(5.4.46)
> 4'
G(x) will blow up on the finite interval [0, c/4]' and so will g(y). But,
this contradicts the fact that f(t) is defined on [c/2,3c/4J. Recall that
bl is determined only from the fact that (14) fails and is independent of
the choice of E. Hence, (5.4.46) must hold for all E > 0, an obvious
impossibility. This completes the proof of (13) =} (14) and also that of
Theorem 5.4.3.
E(x(t
[X'(t)J2
q(t)
= - - + --xl'+! (t),
'Y
+1
t::::: to.
dt E(x(t
'(~
(5.4.47)
since q'(t) ::; 0 for t::::: to. From (5.4.47) it follows that {X'(tn)/ql/2(t n )}
forms a nondecreasing sequence, and in particular
[r,rlJ.
J:
443
In case condition (5.1.16) does not hold, we have the following extension
of criterion (b).
Theorem 5.4.4.
Let q1 (t)
exp ( -
1=
s"Yq(s)ds
0,
(5.4.51)
E(x(t)) = [x'(tJF
q(t)
+ _2_X"Y+1(t)
,),+1
'
so that
(1
t
to
dq+(s))
(5.4.52)
Chapter 5
444
(a(t)lxl(t)I"'-lXI(t)1
+ q(t)f(x(t =
0,
(5.5.1)
where
(i) 0: is a positive constant,
(ii) aCt) E C([to ,oo),1R+), q(t) E G([to, 00), 1R),
(iii) f E G(1R, 1R), xf(x) > 0 and f'(x)::::: 0 for x
-I- O.
1](t)
ltD
a- 1/", (s)ds
lim 1](t) =
t--'>(X)
00.
(5.5.2)
(a(t)lxl(t)I"'-lXI(t)1
+ q(t)lx(t)I13-1 X(t)
0,
(5.5.3)
liminf
t--'>co
ltD
(5.5.4)
(5.5.5)
Proof.
445
w'(t)
Ix'(t)I<>+1
-q(t) - /3a(t) x,B+1(t)
(5.5.6)
w(t) = w(to) -
to
q(s)ds - /3
it
to
a(s)
Ix'(s)la+1
,B+1() ds.
X
(5.5.7)
It:
x 1--r(t) - x 1--r(to)
it I
L
= (-y - 1)
:S (-y - 1)[I(t)]1/(a+1)
x'(s) IdS
x-r (s)
] a/(a+1)
a- 1/a(s)ds
to
x(,B-a)/<>(t) > ~
- 7J(t)
where
w'(t)
C4
for
t> T
-
(5.5.8)
+ q(t) + (C~)
0:
:S 0
for
t;:::: T.
446
Chapter 5
and
min { inf
x>o f
1'(x) lX
1
inf 1'(x)
~
fl/a( ) du, x<o
fY(x)
(x) +0
u
Q
jX
-0
p/a(u)
du }
> O.
(5.5.10)
Remark 5.5.1. Conditions (5.5.9) and (5.5.10) when 0: = 1 reduce to
(5.1.5) and (5.2.86). Also, we can replace condition (5.5.10) by
l' (x ) )
( f(a-l)/a(x)
1
---,--,,------,-...,...du
> -1 > 0 for all
p/a(u)
c
x,
(5.5.11)
where c is a constant.
N ow we sha.ll prove the following theorem.
Theorem 5.5.2. Let conditions (5.5.2), (5.5.4), (5.5.9) and (5.5.10) hold.
If for every constant k > 0, the half-linear differential equation (5.5.5) is
oscillatory, then (5.5.1) is oscillatory.
Ix'(t)laH ,
w'(t) = - q(t) - aCt) j2(x(t f (x(t),
or
W '(t)
q(t) - a-1/a(t)lu'(t)l(aH)/a (
1'(x(t
)
fCa-l)/a(x(t
'
(5512)
..
and hence
rt
rt
Ix'(s)laH,
wet) = w(to) - lto q(s)ds - lto a(s) j2(x(s) f (x(sds.
Now we need to consider two cases, when
I(to,t)
Ix'(s)la+l,
lto a(s) j2(x(s)) f (x(sds
00.
00.
1'(x(t
) lxCt)
1
( JCa-l)/a(x(t
+0 p/a(u) du > c for t 2: to
(5.5.13)
447
(5.5.14)
ry(t)
Ec)a+1
( 2"
(5.5.15)
where
_
:l(t1' t) -
It -l/a
(f(a-1)/a x (s)
t, a
(s)
f'(x(s)
l/a
ds,
(5.5.16)
If :l(t1, 00) < 00, (5.5.16) ensures that v(t) is bounded and hence (5.5.14)
holds. So, we restrict our attention to the case when :l(t1'00) = 00. Let
t2 > t1 be such that V(t1):S: (cE/2)[.J(t1' t)]a/(a+1) for t::::: t2' Then
from (5.5.16) it follows that
(5.5.17)
and hence, from (5.5.13), we obtain
v(t)
:s:
c1/(a+1)E
[1.
a- 1/a(s)v1/a(s)ds
a/(a+1)
or
0:0 : 1
-
[It
t, a- 1/a(s)v1/a(s)ds ]
:s:
0+'
0:0 : 1
-
c1/(a+1)E (t a- 1/a(s)ds
it2
[l t2
t,
:s:
t::::: t2,
we get
a- 1/rY.(s)v1/a(s)ds ]
c1/(a+1)Ery(t).
0+'
Chapter 5
448
Now, there exists a t3 2': t2 such that
[
11 a-1/a(s)v1/a(s)ds
t
] a/(aH)
Now from (5.5.13) and (5.5.14) there exist a T 2': t3 and a constant
C2 > 0 such that
f'(x(t))
2': C2
f(a-l)/a(x(t))
'I}(t)
for
2': T.
(5.5.19)
w'(t)
for all
t 2': T.
Remark 5.5.2. By Theorems 5.5.1 and 5.5.2 it is clear that the results
established in Chapter 3 for half-linear differential equations can be employed to obtain some interesting oscillation criteria for nonlinear equations
of types (5.5.1) and (5.5.3). The formulation of such results are left to the
reader.
Remark 5.5.3. From the results presented in earlier chpaters it is clear
that condition (5.5.4) in Theorems 5.5.1 and 5.5.2 can be weakened.
In the following results we shall present oscillation criteria for (3.13.76)
and (3.13.89) with g(t) = t, i.e., we shall consider the equations
(a(t)~(x'(t)))'
+ F(t, x(t)) =
(5.5.20)
0,
(5.5.21)
and
(a(t)(x'(t))a*)' +F(t,x(t))
are as in equations
449
o< c <
and
00
~ o.
(5.5.23)
1UTI
-+0, k>0
\II,to(a, t)
0
=
t)
(5.5.24)
uniformly on any interval of the form ttl, (0) for some t1 > to. Suppose
moreover that
(5.5.25)
Then equation (5.5.20) is oscillatory if and only if
00
(5.5.26)
00
to
x'(t)
~ ~r1
(att))
~r1(a(t)'I/J(x'(t)))
to
to t
for
G(u) =
~ to
(5.5.27)
uo
t ~ to.
dv
['I/J-1(uW'
(5.5.28)
(5.5.29)
>
F(t,x(t))
['I/J-1 (a(t)'I/J(x'(t)) W
['l1 1,to (a, tW(x( t))-C F(t, x(t)).
(5.5.30)
Chapter 5
450
From the strong sublinearity and the inequality x(t) :::; COWk,to(a, t), t 2: to
where Co > 0 is a constant, it follows that
(5.5.31)
for t > to. Substituting (5.5.31) into (5.5.30) and using the inequality
Wl,to(a, t) >
Wk.to(a,t) -
1p-l
(.!.)
t > to
k'
O 1p-l
1~1 [1P-~~'VW'
:::;
hI
for
(t, y)
JR+
(5.5.32)
1R.
g(xy) 2: g(x)g(y)
and
du
i+o go Iw-1(u)1
<
00
and
x, y > 0
for any
-0
go
1 dU1 ( )1 <
w-
(5.5.33)
00.
(5.5.34)
00
q(S)g(Wk(a, sds
(5.5.35)
00
to
-I- O.
g(x(t
2: g(w-1(a(t)w(x'(t)g(wl,to(a, t,
t 2: to
(5.5.36)
Define
dv
10
H(u) =
451
gO'lj;-l(V)'
(5.5.37)
u> 0
where Uo = a(to)'lj;(x'(to)) > O. Then by (5.5.36) and (5.5.37) for t > to,
we have
F(t, x(t))
g( ~/r 1 (a( t )~j;( x' (t)))
q(t)g(x(t))
;::: q(t)g(\}il,to(a, t)).
>
g( ~p-l (a(t)~j;(x'(t)))
(5.5.38)
Integrating (5.5.38) from t1 > to to t2 > tI, we find
t2 q(S)g(\}il,to(a, s))ds
t,
where Ui
:s: 111
U2
= a(ti)'lj;(X'(ti)),
0:~l( )'
~
J'X)
7l'(t)
00
a- 1 /0I(s)ds
and
7l'(to) <
00.
(5.5.39)
lOO (ats)
1:
IF(u,k)ldU) 1/01 ds
00
(5.5.40)
1 (1
00
to
)1/0.
ds:S: 7l'(to)
<
00,
Chapter 5
452
Case II. If x'(t) < 0 for t ~ tl ~ to, then an integration of (5.5.21) gives
for
t,
~ tI ,
or equivalently,
-x'(t)
Since, x(t) ::;
implies
It
)
( aCt) l F(s,x(sds
~
C2,
1/",
C2
(5.5.41 )
for some constant c < a (see Definition 5.5.1). Combining (5.5.41) with
(5.5.42) and using the decreasing property of x(t), we find
-x'(t)
> c;-C/"'a-I/"'(t)
(t)'
= _
>
c;-C/"'a-I/<>(t)xC/<>(t)
(-x a;:c
(1:
(a:
(1:
~ tl.
(5.5.43)
- x(<>-c)/<>(t) and using (5.5.43), we obtain
c) x-c/<>(t)x'(t)
(a: C) c;-c/<> 1:
1:
lOO (a~) 1:
2
(ats)
F(u, c2)du
t2 -4 00
y/a
F(U,C2)duy/a ds <
00,
Example 5.5.2. Consider (5.5.3) with aCt) = eAt and q(t) = el-'t for
t > 0, where >. and JL are real constants and >. > O. Let Q: > (3. Then
by Theorem 5.5.5, equation (5.5.3) is oscillatory if and only if JL
>..
453
5.7. References
1. R.P. Agarwal, S.R. Grace and D. O'Regan, Linearization of second
454
Chapter 5
25(}-255.
4. G.J. Butler, Integral averages and the oscillation of second order ordinary
differential equations, SIAM J. Math. Anal. 11(1980), 19(}-200.
5. G.J. Butler, An integral criterion for the oscillation of a second order
sublinear ordinary differential equations, Indian J. Math. 24(1982), 1-7.
6. A Elbert and T. Kusano, Oscillation and nonoscillation theorems for a
class of second order quasilinear differential equations, Acta. lvlath. Hungar. 56(1990), 325-336.
7. H.E. Gollwitzer, Nonoscillation theorems for a nonlinear differential equ-
ation, Proc. Amer. Math. Soc. 26(1970), 78-84.
8. S.R. Grace, Oscillation theorems for second order nonlinear differential
equations with damping, Math. Nachr. 141(1989), 117-127.
9. S.R. Grace, Oscillation theorems for nonlinear differential equations of
second order, J. Math. Anal. Appl. 171(1992), 22(}-241.
10. S.R. Grace, Oscillation theorems for damped functional differential equations, Funkcial. Ekvac. 35(1992), 261-278.
11. S.R. Grace, Oscillation of nonlinear differential equations of second order,
Publ. Math. Debrecen 40(1992), 143-153.
12. S.R. Grace, On the oscillatory behavior of solutions of second order nonlinear differential equations, Publ. Math. Debrecen 43(1993), 351-357.
13. S.R. Grace and B.S. Lalli, Oscillation theorems for certain second order
perturbed nonlinear differential equations, J. Math. Anal. Appl. 77(1980),
205-214.
14. S.R. Grace and B.S. Lalli, Oscillation of solutions of damped nonlinear
second order functional differential equations, Bull. Inst. lvlath. Acad.
Sinica 12(1984), 5-9.
15. S.R. Grace and B.S. Lalli, Oscillation theorems for a second order nonlinear ordinary differential equation with damping term, Comment. Math.
Univ. Caralin. 27(1986), 449-453.
16. S.R. Grace and B.S. Lalli, Oscillatory behavior of solutions of second order differential equations with alternating coefficients, Math. Nachr.
127(1986), 165-175.
17. S.R. Grace and B.S. Lalli, An oscillation criterion for certain second order strongly sublinear differential equations, J. Math. Anal. Appl.
123(1987), 584-588.
18. S.R. Grace and B.S. Lalli, An oscillation criterion for second order
sublinear ordinary differential equations with damping term, Bull. Polish.
Acad. Sci. Math. 35(1987), 181-184.
19. S.R. Grace and B.S. Lalli, Oscillation theorems for second order nonlinear differential equations, J. Math. Anal. Appl. 124(1987), 213-224.
20. S.R. Grace and B.S. Lalli, On the second order nonlinear oscillations,
Bull. Inst. Math. Acad. Sinica 15(1987), 297-309.
21. S.R. Grace and B.S. Lalli, Integral averaging and the oscillation of
second order nonlinear differential equations, Ann. Mat. Pura Appl.
151(1988), 149--159.
455
22. S.R. Grace and B.S. Lalli, Oscillations in second order differential equations with alternating coefficients, Period. Math. Hungar. 19(1988), 6978.
23. S.R. Grace and B.S. Lalli, Oscillation theorems for nonlinear second
order differential equations with a damping term, Comment. lvIath. Univ.
Carolin. 30(1989), 691-697.
24. S.R. Grace and B.S. Lalli, Integral averaging techniques for the oscillation of second order nonlinear differential equations, J. Math. Anal. Appl.
149(1990), 277-311.
25. S.R. Grace, B.S. Lalli and C.C. Yeh, Oscillation theorems for nonlinear second order differential equations with a nonlinear damping term,
SIAM J. Math. Anal. 15(1984), 1082-1093.
26. J.R. Graef, S.M. Rankin and P.W. Spikes, Oscillation theorems for
perturbed nonlinear differential equations, J. Math. Anal. Appl. 65(1978),
375-390.
27. J.W. Heidel, A nonoscillation theorem for a nonlinear second order differential equation, Proc. Amer. Math. Soc. 22(1969), 485-488.
28. LV. Kamenev, Oscillation of solutions of second order nonlinear equations with sign variable coefficients, Differencial'nye Uravnenija 6(1970),
1718-1721.
29. T. Kusano, A. Ogata and H. Usami, Oscillation theory for a class of
second order quasilinear ordinary differential equations with application to
partial differential equations, Japan J. IvIath. 19(1993),131-147.
30. T. Kusano, H. Onose and H. Tobe, On the oscillation of second order
nonlinear ordinary differential equations, Hiroshima Math. J. 4(1974), 491499.
31. M.K. Kwong and J.S.W. Wong, On the oscillation and nonoscillation
of second order sublinear equations, Proc. Amer. Math. Soc. 85(1982),
547-551.
32. M.K. Kwong and J.S.W. Wong, Linearization of second order nonlinear oscillation theorems, 'nans. Amer. Math. Soc. 279(1983), 705-722.
33. M.K. Kwong and J.S.W. Wong, On an oscillation theorem of Belahorec, SIAM J. Math. Anal. 14(1983), 474-476.
34. M.K. Kwong and J.S.W. Wong, Nonoscillation theorems for a second
order sublinear ordinary differential equation, Proc. Amer. Math. Soc.
87(1983), 467-474.
35. H. Onose, On Butler's conjecture for oscillation of an ordinary differential
equation, Quart. J. Math. 34(1983), 235-239.
36. Ch.G. Philos, Oscillation of sublinear differential equations of second
order, Nonlinear Analysis 7(1983), 1071-1080.
37. Ch.G. Philos, On second order sublinear oscillation, Aequations Math.
27(1984), 242-254.
38. Ch.G. Philos, Integral averaging techniques for the oscillation of second
order sublinear ordinary differential equations, J. Austral. Math. Soc. Ser.
A 40(1986),111-130.
456
Chapter 5
39. Ch.G. Philos, On the oscillation of second order sublinear ordinary differential equations with alternating coefficients, Math. Nachr. 146(1990),
105-116.
40. Ch.G. Philos, Integral averages and oscillation of second order sublinear
differential equations, Differential and Integral Equations 4( 1991), 205-213.
41. Ch.G. Philos and I.K. Purnaras, On the oscillation of second order
nonlinear differential equations, Arch. Math. 59(1992), 260-271.
42. J .S.W. Wong, Oscillation theorems for second order nonlinear differential
equations, Bull. Inst. Math. Acad. Sinica 3(1975), 283-309.
43. J .S.W. Wong, Remarks on nonoscillation theorems for a second order
nonlinear differential equation, Proc. Amer. Math. Soc. 83(1981), 541546.
44. J .S.W. Wong, Oscillation theorems for second order nonlinear differential
equations, Proc. Amer. Math. Soc. 106(1989), 1069-1077.
45. J .S.W. Wong, A sublinear oscillation theorem, J. Math. Anal. Appl.
139(1989), 197-215.
46. J .S.W. Wong, An oscillation theorem for second order sublinear differential equations, Proc. Amer. Math. Soc. 110(1990), 633-637.
47. J .S.W. Wong, Oscillation criteria for second order nonlinear differential
equations with integrable coefficients, Proc. Amer. Math. Soc. 115(1992),
389-395.
48. J .S.W. Wong, An oscillation criterion for second order nonlinear differential equations with iterated integral averages, Differential and integral
Equations 6(1993), 83-91.
49. J.S.W. Wong, Nonoscillation theorems for second order nonlinear differential equations, Proc. Amer. Math. Soc. 127(1999), 1387-1395.
Chapter 6
Further Results on the
Oscillation of Differential
Equations
6.0. Introduction
In this chapter we shall discuss some techniques which are different
than those employed in the previous chapters to obtain oscillatory criteria
for differential equations. In Section 6.1 we shall present oscillation and
nonoscillation theorems for nonlinear second order differential equations by
using the method of Olech, Opial and Wazewski. Section 6.2 is concerned
with the oscillation of half-linear second order differential equations by employing the variational inequality given in Lemma 3.2.6. In Section 6.3 we
shall begin with some preliminaries of Liapunov functions, and then apply Liapunov second method to obtain criteria for the oscillation of second
order nonlinear equations.
x"(t)
+ q(t)f(x(t))
(6.1.1)
0,
(6.1.2)
i= 0
i= 0,
Chapter 6
458
(6.1.3)
in
or
lim approxinf
t -+ 00
ft q(s)ds <
limapproxsup
t -+ 00
io
(6.1.4)
fat q(s)ds,
in
t
io
q(s)ds
(6.1.5)
00,
S by
(6.1.6)
g(x)
f"(x)f(x)
(f'(x))2'
=1=
g(x)
'Y
o.
=1=
o.
(6.1. 7)
= Ixll' sgn x,
(6.1.8)
With respect to the function g(x), we will assume that the following
condition holds: There exist numbers M > 0 and m < 1 such that
Ig(x)1 ::; M
for all
=1=
o.
(6.1.9)
459
Theorem 6.1.1.
b(M, m)
00,
M2 -4m+4
(6.1.10)
(6.1.11)
and
lim
t-+CXJ, tES
q(s)ds
io(
00.
(6.1.12)
w'(t)
(6.1.13)
w(t)
w(to)
-it -it
q(s)ds
to
f'(x(s))w 2 (s)ds.
(6.1.14)
du
f(u)'
(6.1.15)
to
0(x)
ix
it
to
w(s)ds =
where Co
it
to
x'(s)
f(x(s)) ds = -
r(t
ix(t)
o)
du
f(u) = - 0(x(t))
+ co,
0(x(t)),
(6.1.16)
0(x(to)), if we set
y(t)
Co
-it
to
w(s)ds
460
Chapter 6
then x(t)
(6.1.l4),
q(s)ds
y'(t)
-it
t 2
(6.1.17)
to
to
t 2 to
W(t) =
.f:
f'(n-1(y(s))(y'(s))2ds,
2 to
(6.1.l8)
Now (6.1.12) implies that q(t) is not identically zero on any half-line
(tl, (0), and so y'(t) is not identically zero on any half-line (tIl (0)
for some t1 2 to. It follows that W (t) > 0 for all t sufficiently large.
Without loss of generality we may assume that W(t) > 0 for t 2 to. We
now let the function P(t) be defined by the equation
y'(t) = P(t)W(t)
for
> to.
(6.1.l9)
We then have
for
t > to.
and
U(t) =
V(t)~(t)'
> to,
we find that V(t) > 0 and U(t) > 0 for t > to, and
V'(t)
j"(x(t))x'(t)
j"(x(t)) :t (n-1(y(t)))
(6.1.20)
for
> to
We also have
U'(t)
V'(t)
W'(t)
V2(t)W(t)
V(t)W2(t)
1" (x( t) )f(x( t) )P( t) W (t)
p 2(t) W 2(t)f' (x(t))
V2(t)W(t)
V(t)W2(t)
g(x(t))P(t) - p 2(t) for t > to,
(6.1.21)
Sl
(6.1.22)
461
TheIl from (6.1.17) and (6.1.19) for all sufficiently large k > 0, we have
(6.1.23)
Assume now that k is fixed so that tk > to and (6.1.23) holds, and let
5 = Sl n [tk' (0). It follows that
Ji(S
.
1lmsup
n [0, t])
t
1--'>00
Now on
5,
1.
Ji(S
Ilnsup
t--'>oo
n [tk' t])
t
Ji(5n[tk,t])
.
<::: 1mlSUp
t--'>oo
(6.1.21)
(since the function rnP - p 2 for P 2: 1 is maximized at P = 1). Further
on 5c = [tb cc)\5 (= the complement of 5 in [tk' (0)), P(t) < 1 so that
the function P(a - P) is maximized at P = a/2. Thus, from (6.1.21)
and (6.1.9), we obtain
1
5=
U(t)
}s(t)
-4]1;[2
for
t E
5c .
(6.1.25)
then we have
U'(s)ds + (
}se(t)
U'(s)ds,
t 2: tk.
1 set)
U'(s)ds
+ (
}se(t)
U'(s)ds 2: - L > -
and all
(6.1.26)
00
1s(t)
t [(rn -l)Ps(t)
+ ~M2(1- Ps(t))] 2: -
L > -
00
Chapter 6
462
which implies
4L
M2]
t [PS(t) - M2 _ 4m + 4
00
t
io
q(s)ds =
(6.1.29)
00.
(2.=2)2]
+
t-+oo, tES
iot
i'
t-+oo
q(s)ds
00.
00
(6.1.30)
(6.1.31)
fot q(s)ds
io
= 00.
(6.1.32)
xf(x) > 0 and f'(x);:::: C> 0 for some constant C> 0 and all x
=1= O.
(6.1.33)
463
If we assume 0 < 'Y1 < 'Y2 < ... < 'Yn, then clearly a necessary
condition for (6.1.33) to hold is C1 > 0, Cn > 0 and 'Y1 ~ 1 ~ 'Yn.
Theorem 6.1.2. Assume f(x) satisfies condition (6.1.33) and assume
that condition (6.1.32) holds. Then equation (6.1.1) is oscillatory.
Proof. Since (6.1.32) holds, it follows that the equation xl/(t)+cq(t)x(t) =
is oscillatory by Corollary 6.1.3, where c is the same as in (6.1.33).
Hence, by Theorem 4.3.5 equation (6.1.1) is oscillatory.
limsupps(t) <
t....;oo
'Y - 1)2
(- ,
'Y + 1
'Y-=/=1.
(6.1.34)
t....;oo,
tES
10t
q(s)ds =
00
(6.1.35)
t [ps(t) + EO -
(~ ~ ~
r]
Bo,
(6.1.36)
Choose an open set U C IRo such that S C U and p,(U\S) ~ 1 and define
Q1 = IRo\U, Q2 = U\S. Thus, Q1 is closed, Q1nS = 0, IRo = SUQ1UQ2.
Hence, P,(Q2) ~ 1 and so, tpQ2(t) ~ 1 and tpQ, (t) :::: t[l - Ps(t)]- 1.
Define the function pet) by
1+E1, tES
{ 'Y - 1
~' tEQ1
pet) =
where 0 < E1 < 1 will be chosen sufficiently small. It follows that Ip(t)1 ~
2 + Ib -1)/(2'Y)1 on S U Q1, and we can extend pet) to all of IRo so
that it is of class C 1 andsatisfies Ip(t)I~Co=2+lb-l)/(2'Y)1 forall
t E IRo. We now let the function wet) be defined by
t E IRo
(6.1.37)
Chapter 6
464
and notice that
t E S
w(t)
(6.1.38)
and
Iw(t)1 :S
1'Y~IICo+c6'
tElRo.
(6.1.39)
C1
I"I ~ 11 Co + C6 and A =
E1
(~ + 1 + (1) .
(6.1.40)
lRo by
u(t) = B1
lot w(s)ds,
(6.1.41)
u(t) = B1
r w(s)ds + Jr
1s
w(s)ds +
Qlt
r w(s)ds.
lQ2t
(6.1.42)
1st w(s)ds
= -
r w(s)ds =
JQlt
(~ + A) tps(t),
(6.1.43)
('Y-l)2tPQ1(t)
(6.1.44)
2"1
and
(6.1.45)
Hence, we have from (6.1.42) - (6.1.45) and (6.1.36) after rearranging
(6.1.46)
where
465
and
r'(t) = utt)p2(t)r(t),
on IRo \Q2 = S U Ql. We have
p2 (t ) >
C~
u(t) - Bl + CIt'
r(O) = 1
(6.1.47)
{Ill} .
C 2 = min 1, '"'( 2~
(6.1.48)
lim r( t)
t-+oo
(6.1.49)
00.
v(t)
r(t)u(t) > 0,
(6.1.50)
E IRo
= v(t),
f'(x(t))
then, we have
hmt-+oo x(t)
i.e.,
v(t)) l/(-y-l)
= (-
x(t)
'"'(
{Oif'"'(>l
f 0
001
1
<'"'.
With
f(x)
(6.1.51)
Now, by (6.1.41),
w(t) =
u'(t)
=
f"(x(t))f(x(t)) ()
2( )
(f'(x(t)))2 P t - P t
f"(x(t))f(x(t)) ()
2( )
v 2 (t)
P t - pt.
Chapter 6
466
On the other hand, we also have u(t)
therefore by (6.1.47) and (6.1.51)
,
v'(t)
u (t) = - r(t)v2(t)
r'(t)
r2(t)v(t)
= l/(r(t)v(t
-
by (6.1.50), and
f"(x(tx'(t)
2
r(t)v2(t) - p (t).
x'(t)
f(x(t
(6.1.52)
p(t)r(t).
q(t)
J;
(6.1.53)
so that (6.1.49) and (6.1.53) imply that limHoo,tEs
J; q(s)ds =
00.
('ljJ(x'(t)'
+ q(t)'ljJ(x(t =
0,
(6.2.1)
where
(i)
q(t) E C(lRo,lR),
(ii) 'ljJ(8) = 181",-18 with a> 0 is a constant.
() _ 'ljJ(x'(t
t - 'ljJ(x(t
(6.2.2)
467
w'(t)
+ q(t) + alw(t)I(<>+l)/<> = o.
(6.2.3)
I(x;a,b) =
lb
[lx'(s)Ia+ 1
q(s)lx(s)I<>+l] ds
in the class of functions satisfying x(a) = 0, x(b) = 0 and the discojugacy of (6.2.1) in [a, b]. Moreover, we shall investigate (6.2.1) not as
a perturbation of (6.2.4), but as a perturbation of the generalized Euler
equation
(6.2.5)
('lj;(x'(t)))' + t7~1 'lj;(x(t)) = 0,
where 'Yo = (a/(a+l))<>+l is the so called critical constant in this equation.
As in the linear case, if we replace 'Yo by a constant 'Y > 'Yo ('Y < 'Yo),
then equation (6.2.5) becomes oscillatory (remains nonoscillatory).
(a(t)x'(t))'
+ q(t)x(t)
I(x;a,b) =
lb
(6.2.6)
i.e., any
for every nontrivial, piecewise GI[a, b] function for which x(a) = 0 = x(b).
Another important concept in oscillation theory of linear equations is the
principal solution. A solution Xo of (6.2.6) is said to be principal if
limHoo xo(t)/x(t) = 0 for any nonzero solution x of equation (6.2.6)
which is linearly independent with Xo (solution x is said to be nonprincipa0. A principal solution of equation (6.2.6) exists (uniquely up to
multiplication by a nonzero real constant) if and only if (6.2.6) is nonoscillatory.
Now consider (6.2.6) as a perturbation of the nonoscillatory equation
(a(t)x'(t))'
+ qo(t)x(t)
= 0,
(6.2.7)
Chapter 6
468
t-H>a
Xo t
1=
t
[q(s) -
qo(s)]x~(s)ds
> 1,
(6.2.8)
XI(t)
t~n~ xo(t)
00
+ ?
1
[q(s) - qo(s)] xo(s)ds < 4'
(6.2.9)
lim(lnt)l[q(s)-~lsds
t
4s
t--+oo
> 1
(6.2.10)
x"(t)
+ q(t)x(t)
(6.2.11)
t--+=
1=
t
[q(s) -
~]
4s
sds <
(6.2.12)
(a(t)v;(x'(t)))'
on [a, b] satisfying x(t)
+ q(t)~(x(t)) =
0,
(6.2.13)
y(b)
and
y(t) f 0 on (a,b)}.
J(y;a,b) =
lb
[a(s)ly'(s)I"+l - q(s)ly(s)I"+l] ds 2 0,
(6.2.14)
are proportional.
A closer examination of the proof of this lemma shows that the condition
[a, b] in definition of Y can be replaced by a wea.ker condition:
y E Cl
469
tE
I(y; a, b) <:: 0, then (6.2.13) is conjugate in [a, b], i.e., there exists a nontrivial solution with at least two zeros in [a, bj. Conversely, if I(y; a, b) > 0
for every nontrivial y E Y1 , then (6.2.13) is disconjugate in [a,bj. Indeed,
if y is a nontrivial solution with consecutive zeros t 1, t2 E [a, b], then for
we have
to
[a(s)ly'(s)lo+ 1
tl
a(t)y(t)?fJ(y'(t))l
q(s)ly(s)la+lj ds
o.
tl
t~~,ont)
Suppose that
where 10 = (0:/(0:
00
[q(s) -
+ 1))<>+1,
S2~1] s"'ds
>
2C:t: 1)
a,
(6.2.15)
Proof. In view of Lemma 6.2.1 it suffices to find for any d > 0 a piecewise
differentiable function x with compact support in (d,oo), say, [to, t 3 ]
such that
x(t) =
~(t{f
i:
<::t:o<:: t <:: tl
if
t 2: t 3 ,
470
Chapter 6
H(t2' t3)
Vo
t~/(Q+I) [Vot;-a/(a+l)
eX: 1)
a,
xo(t)
P(9'(t 2))] ,
-1
ta/(a+I).
Next, we claim that the functions f /xo and g/xo are strictly monotonic on
(to, tl) and (t2' t3) respectively. If (f /xo)'(f) = 0 for some t E (to, tr),
thenwehave (f'/f)(f) = (xb/xo) (f). Denote by WI = '1j;(f')N(f) , W2=
'ljJ(x~)/'ljJ(xo). Clearly, the functions WI and W2 satisfy the generalized
Riccati equation
W' + ~ + alwl(a+l)/a = 0
(6.2.16)
t a +1
471
But this is an obvious contradiction (since f(to) = 0), i.e., (f /xo)' > 0
on this interval. The same argument applies to prove the monotonicity of
g/xo over (t2J t3).
Note that the second mean value theorem of the integral calculus implies
the existence of a constant (1 E (to, td such that
Similarly,
j ~2
t2
q(s) -
I]
oS
"'~1 s"'ds
(4;(z'))' - mj;(z')
and
+ 104;(z) =
(6.2.17)
Vi = av - 10 - alvl(o+l)/", = F(v),
(6.2.18)
V (U)
veT)
d(
F(()
(6.2.19)
= v - T.
left~hand
The function F(v) can be expressed in the form (since F'(vo) = 0),
F(v) = (1/2)F"(vo)(v - VO)2 + O((v - von as v -+ Vo, and hence
_1_
_ 2 + 0(_1_)
F(v) - F"(VO)(V - VO)2
vo
V -
as
v -+ vo,
Chapter 6
472
this implies
2
as
u- T
F"()(
va va-v ) + O(ln Iv - vol)
u -+
00.
Thus, we have
2
u -+
as
F"(vo)
00.
u->oo
v(u))(u - T) =
(1'0 -
v - va
!F"(vo)u + O(1n u)
_
F"(vo)u
~F"(vo)u (1 + 0
(1_0(lnu))
__
F"(vo)u
e:
U ))
+o(lnu)
u2
t w(t) -
1'0
2
(In(lnt))
= - FI/(vo)
In t + 0 ~
as
t -+
00.
Let w(t, to) = 'ljJ(f'(t))j'lj;(f(t)). Then the function w solves (6.2.16), and
hence
o
[va tH(t 2, t)
3 -- to/(a+l)
2
2 /(a+1)
as t3 -+
00,
we
a
t 2a2 /(a+1) W (t 2, t)]
3
-- Vo - t 2 W (t 2, t)
3 ,
473
I(x;to,t3)
G(to,tl) +H(t2,t3)
G(to, tl)
:::; G(d,t1) [ G(d,tl)
-1~2
[q(s) -
1~2
H(t2' t3)
+ G(d,t 1 )
G(d,6) ~,
s~~l] s"'ds
1'0 ]
q(s) - S,,+l s ds .
(6.2,20)
Next, we let E > 0 be a sufficiently small number such that the limit
in (6.2.15) is greater than 2(0:/(0: + lCY(l + 60) = 21'0(1 + 6E). We have
limt--+=(lnt)G(d,t) = 21'0, i.e., (Inx)G(d,t) < 2Vo(1+E) if t is sufficiently
large, and in view of (6.2.15), to > d can be chosen in such a way that
G(;6)
. G(to,t)
11m
-=,:-:-,--,-
t--+=
G(d,t)
1
)
(In6)G(d,6)
>
21'0 (1 + 6f)
> 1 + 4f,
2vo(1 + f)
(6.2.21)
. vo-t"w(t,to)
111n
--'---....:.,-c.....::e-
t-+oo
11m
t--+oo
1'0 - tCYw( t, d)
F"(:::o2)lnt
-2
F"(vo) Int
+ 0 (~)
+0
(In(lnt)
JnTt
G(to, t 1 )
< 1 + f.
G(d,tl)
Further, (6.2.21) implies the existence of t2 > tl such that
G(d1 ~ )
,1
1'] s"'ds
6
[q(s) - <>~l
E,
s
> 1 + 3f whenever
1,
474
Chapter 6
Finally, since H(t2' t3) -40 as t3 -4 00, we find that H(t2' t3)C- 1(d, tl) <
if t3 is sufficiently large. Combining the above calculations, it follows
from (6.2.20) that I(x; to, t 3) < -E, and hence I(x; to, t3) < 0, and now
by Lemma 6.2.1 equation (6.2.1) is oscillatory.
Remark 6.2.1.
1. As in Chapter 3, we observe that the transformation of the independent
variable
lot a-1/a(s)ds
(6.2.22)
d~ (~ (:u x) ) + a1/a[t(u)]q(t[u]N1(x)
0,
Joo a-1/a(s)ds
and
(6.2.23)
00
00
q(S)dS) > 1,
t:::; to
0,
x(t)
ito
tl:::; t :::; t2
1,
t3
a-1/a(s)ds /
1:
a-1/O(s)ds,
to:::; t :::; tl
475
Now, if to < t1 < t2 < t3 are sufficiently large, then as before we have
I(x; to, t3) < O.
(6.2.24)
for any nontrivial C 1 function x with compact support in (N, 00) (see
Lemma 6.2.1). For this, first we shall establish the following inequality:
Let M be a positive differentiable function for which M'(t) =1= 0 in [a, bJ
and let z E Y (Y is defined as in Lemma 6.2.1). Then,
l/
and let N E R be such that the expression in (6.2.24) is less than v for
t > N. Using (6.2.25) and (6.2.24), we find for any differentiable x with
476
Chapter 6
LXJ q(s)Jx(s)J"+lds
(a
+ 1)
:::; (a
+ 1)
:::;
oo
q+(S)JX(s)J"+lds
L (J: X'(U)~j;(X(U))dU)
LJx'(s)J~j;(x'(s))M(s) (J8 !;(~~)dU)
L
< (a + l)v
oo
q+(s)
ds
oo
oo
oo
M(s)Jx'(s)J1f;(x(s))ds
<>
:::; (a+1)v(LOOJM'(S)JJx(s)J"+ldS)
< (a + l)"+lv
: :; L=
-
ds
00
<>+1
(LOOJ~~l;~J:IJx'(s)J"+ldS)
1
<>+1
JM(S)J"+l
Jx'(s)J"+lds
JM'(s)J"
a(s)Jx'(s)J"+lds
L
oo
(a(t)x'(t))'
(6.3.1)
x'(t)
y(t)
a(t) ,
y'(t)
y(t))
-F ( t,x(t), a(t) .
(6.3.2)
477
6.3.1. Preliminaries
Consider the system of first order differential equations
x'(t) = F(t, x(t)),
(6.3.3)
~~.3.3)(t, x) =
(6.3.4)
Let x(t) be a solution of(6.3.3) which stays in S and denote by V'(t, x(t))
the upper right-hand derivative of Vet, x(t)), i.e.,
V'(t,x(t)) =
(6.3.5)
where
Vet, x),
(6.3.6)
.
1
hmsup -h {Vet
h-.O+
+ h, x(t + h)) -
Vet, x(t))}
+ h,x + hF(t,x)) -
(6.3.7)
V(t,x)}.
~ Vet
(6.3.8)
Chapter 6
478
Thus, from (6.3.8) and (6.3.7), we obtain
(6.3.9)
1
liminf -h {Vet + h, x + hF(t, x)) - Vet, x(t))}.
(6.3.10)
h--+O+
V(6.33)(t,X)
av av
at
+ ax . F(t,x),
(6.3.11)
V'(t,X(t))
.
1
hmsuP -h {V(t+h,x(t+h)) - V(t,x(t))}
h--+O+
and
V'et, yet)) = lim sup -hI {Vet + h, yet + h)) - Vet, y(t))}.
h--+O+
V'et, yet))
:::::
h--+O+
479
<
00
y <
00
t, where {x(t),y(t)}
0 for all large t and
x(t), y(t))
.
1
bmsup -h {V(t
h .... a+
.
1
bmsup -h {W(t
h .... a+
(6.3.12)
for all x(t) > 0, ly(t)1 < 00 and t ~ t i . However, by condition (ii), we
have V(t,x(t),y(t)) :S V(ta,x(ta),y(ta)) for all t ~ ta, which contradicts (6.3.12). When x(t) < 0 for t ~ ta, by employing the function
W(t,x(t),y(t)) we arrive again at a contradiction. This completes the
proof.
Now consider a Liapunov function v(t, x, y), and define V(6.3.2) (t, x, y)
by
V(6.32)(t, x, y)
480
Chapter 6
(1 1 )
y
# 0, t
.t
T,
Tl
T.
(6.3.14)
Moreover, assume that there exist a T2 and a w(t, x, y) for all large
such that T2 2: T and w(t, x, y) is a Liapunov function defined
for t ~ T2 , X > 0, y < 0 (x < 0, y > 0) which satisfies the following
conditions:
(13) y:::; w(t,x,y) (-y:::; w(t,x,y)) and w(T2 ,x,y):::; m(y) where m(y)
= 0 and m(y) < 0 (y # 0),
(I4) w(6.32)(t, x, y) :::; -c(t)w(t, x, y) where c(t) E C([T2' (0), IRa) and
is continuous, m(O)
l~ ats) exp (
-l:
C(U)dU) ds =
00.
(6.3.15)
Then if {x( t), y( t)} is a solution of the system (6.3.2) such that x( t) > 0
for all large t, then y(t) 2: 0 (y(t) :::; 0) for all large t.
Proof. Let x(t) > 0 for t 2: to. The case when x(t) < 0 for t ~ to can
be treated similarly. Now assume that there exists a sequence {tn}~=l
such that tn -t 00 as n -t 00. Let tn ~ T and tn be sufficiently large
so that
liminf
t-+CXJ
it
tTl,
k(s)ds
for
~
~ tn.
(6.3.16)
-l~ k(s)ds.
From (6.3.16) it follows that there is To > to such that for all t 2: To,
k(s)ds ~ (l/2)v(t n ,X(tn),y(tn )), because v(tn,x(tn),y(t n )) < o.
Tl~erefore, for t 2: To, we obtain v(t, x(t), y(t)) :::; (1/2)v(tn' x(t n ), y(t n ))
< 0, which implies that y(t) < 0 for all t 2: To.
Itt
481
J;
ht C(S)dS)
C(S)dS)
for
t 2': T.
ht C(S)dS) ,
(6.3.17)
and hence
+ m(y(T))
ht ats)
exp (-
Since x(t) > 0 for t 2': T and m(y(T)) < 0, we find by condition (6.3.15)
that x(t) ---+ -00 as t ---+ 00, which is a contradiction. This completes
the proof.
Remark 6.3.1.
replaced by:
When a(t)
00,
Chapter 6
482
0 eventually. A similar
(i I )
(6.3.18)
ito
0,
0 for x
qI(t)JI(X)
(i3)
f2(x)
t-too
iTt q2(s)ds ~
> 0,
f~(x) ~
~ to
(6.3.19)
F(t,x,u).
0 for x
f=
(6.3.20)
For t ~ to and x ~ 0,
E C(lR, lR) such that
liminf
f=
~ to
and
(6.3.21)
(6.3.22)
V(6.3.2)(t,x,y)
n~x)
-qI(t).
Hence, this v(t, x, y) satisfies the hypotheses of Lemma 6.3.1 with k(t) =
qI(t).
w(t,x,y)
483
Lemma 6.3.1 it follows that if {x(t), y(t)} is a solution of (6.3.1) such that
x(t) < 0 for all large t, then y(t):::::: 0 for all large t.
Now by employing Theorem 6.3.1 we shall prove the following results.
Corollary 6.3.1. Let conditions (id - (i3) hold. If
00
qi(s)ds =
i = 1,2
00,
(6.3.23)
to
V(t, x, y)
~ {I
00
h~X) +
t
<
1:
<
set
for
ql(s)ds
for
ql(s)ds
00
y 2: 0
< o.
to
00
00
<y<
00
R~x)
V(~.3.2)(t, x, y)
ql
+ ql (t)
o.
W(t, x, y)
{it
q2(s)ds
to
h~x) +
for
y> 0
for
q2(s)ds
y:::::: 0
I 1
t
to
00
-1()
qi(u)duds ---+
ass
00
as t ---+
00,
i = 1,2
(6.3.24)
and
00
du
h(u)
<
00
and
du
-00
;
-E
h(u)
<
00
for some
> 0, (6.3.25)
484
Chapter 6
Proof. For
t:::: T*,
V(t, x, y)
00
du
JI(u)
set
00
rt
+ ito a(s)
00
(6.3.26)
ql(u)duds.
For a solution x(t) which satisfies x( t) > 0 for all large t, we can
assume that x(t) > 0, y(t) :::: 0 for t:::: To :::: to, To is sufficiently large,
and hence
1
JI~~~~ + 1
1
y(t)
- JI(x(t a(t)
att) [-
00
+ a(t)
00
ql(s)ds
ql(S)dS] .
V'(6_3_2)(t, x, y) = -
R~x) [-F (t, x, art) JI(x) - y!~ (x) art)] -ql(t) :::: O.
t:::: To.
r-
W(t,x,y) = ix
OO
du
h(u)
+ ito
00
a(s)
q2(u)duds.
i 1
t
-1()
to a s s
00
q2(u)duds -+
00
as t -+
00
and
ft"; ql(s)ds =
roo
J-f
du
h(u) <
00,
00,
=I 0, i = 1,2
and
485
> 0,
V(t,x,y) = h(X)Pl(t)
+ .fto Pl(S)
W(t,x,y)
Y
= h(X)P2(t)
+
[ql(S) -
1t [
to
p2(S) q2(S) -
1 a( s) (p~ (s) ) 2]
4-bPl(S)
ds
1 a(s) (p~(S)2]
4-bP2(S)
ds
Theorem 6.3.3. Let the hypotheses of Lemma 6.3.2 hold and assume
that for each 6 > 0 and m > 0 there exist a T(6, m) > 0 and two
Liapunov functions V(t,x,y) and W(t,x,y) such that V(t,x,y) is
defined for t ~ T(6, m), x> 6, 0:<:::; y < m and W(t, x, y) is defined for
t ~ T(6, m), x < -6, - m < y < 0, moreover, assume that V(t, x, y)
and W(t, x, y) satisfy the following conditions:
00
uniformly for x, y as t -t
00,
486
Chapter 6
By Lemma 6.3.2 there is a constant m > 0 such that 0:::; y(t) < m for all
t 2: t l . Since x'(t) = y(t)/a(t) 2: 0 for t 2: tI, we have x(t) 2: x(td > 0
for t 2: tI
Consider the Liapunov function V(t, x, y) defined for t 2: T(tS, m), x >
8, 0 :::; y < m where 8 = x(t I )/2, where T 2: t l . The rest of the
proof is similar to that of Theorem 6.3.1. When x(t) < 0 eventually, by
employing the Liapunov function W(t, x, y) and Lemma 6.3.2 we arrive
at a contradiction. Thus, equation (6.3.1) is oscillatory.
x"(t)
(6.3.27)
x'(t)
y'(t)
=
=
y(t)
- a(t, x(t), y(ty(t) - F(t, x(t), y(t
(6.3.28)
We assume that
(ad F(t, x, y) E C([to, 00) x m?, lR) and xF(t, x, y) > 0 for x i= 0,
(a2) a(t,x,y) E C([to,oo) x lR2 ,lR) and there exist two functions g(t),
p(t) E C([to, 00), lR o) such that - g(t) :::; a(t, x, y) :::; p(t) for t 2: to, x E
lR, y E lR,
(a3) for any 8 > 0 and m > 0 there exist a T(8, m) and a function
q(t; 8, m) 2: 0 defined for t 2: T(8, m) such that limt-+oo J;(cl,m) q(s; 8, m)ds
= 00, and for Ixl 2: 8, Iyl :::; m and xy > 0, IF(t, x, y)1 2: q(t; 8, m),
(a4)
roo
ito
lim
t~oo ito
exp (-
itor P(U)dU) ds =
00.
For t 2: to 2: 0, x > 0,
v(t, x, y)
Iyl < 00
the function
487
satisfies the conditions of Lemma 6.3.1 with k(t) = 0. For any T 2: to,
the function w(t, x, y) = y defined for t 2: T, x > 0, y <
satisfies the
conditions of Lemma 6.3.1. For this, we have
V(6.3.28)(t, x, y) =
00
00.
{ (-y)exp (l>(S)ds)
v(t,x,y) -
t 2: to, x < 0, y 2: 0
for
(-y)exp (-
9(S)dS)
for
w(t,x,y) = -y for t 2: T, x < 0, y > 0 and u(t,x,y) = exp (-2 It: g(s)ds)
for t 2: to, x < 0, y <
satisfy the conditions of Lemma 6.3.2. Next, for
and m > 0 define V(t, x, y) for t 2: T(8, m), x > 8, 0 :::;
each 8 >
y < m by
t 9(S)dS) + L t
lto
IT(o,m)
q(s; 8, m)ds,
-1:
-1:
V(~.3.28)(t, x, y)
exp (
< exp (
+ Lq(t; 8, m)
+ Lq(t; 8, m)
+ Lq(t; 8, m) :::; 0.
+L
rt
q(s; 8, m)ds
IT(o,m)
488
Chapter 6
Corollary 6.3.6.
repla.ced by
(a5) for any fJ> 0 there exist a T(6) > 0 and a q(t;6) > 0 defined for
T(6) such that
t ~
( exp (
and for
Ixl
(a6) lim
t->oo
-1:
~
00
t --+
as
(6.3.29)
00
(6.3.30)
Proof.
From conditions (al), (a2) and (a5) it follows that there are
Liapunov functions which satisfy the conditions of Lemma 6.3.1 as was
seen in Theorem 6.3.2. For t ~ T(6), x> 6, y ~ 0 define
V(t,x,y) = yexp
(-it q(S)dS) +
to
Then, we have
V(~.3.28)(t,x,y)
exp
(-1:
< exp (
W(t, x, y)
q(s;6)ds.
+ q(t; fJ)
+ (ex p
~
IT(8)
-g(t) (exp
For t
to
-1:
-1:
(-1:
-1:
+ exp (
(expit -9(S)dS)
O.
0 if we define W(t, x, y) by
489
(a7) given 8> 0 there exists a T(8) > to and a q(t; 8) ;::: 0 for t;::: T(8)
with
-1
t
lt
T(Ii)
q(s; 8)ds -+
as
00
t -+
00
and such that for Ixl;::: <5, xy > 0, IF(t, x, y)1 ;::: q(t; <5),
(as) for some To;::: to and a constant lvI,
~ exp (l>(S)dS)
< M
for
t;::: To
(a(t)x'(t))'
and
(a(t)x'(t))'
+ q(t)f(x(t)) =
+ q(t)lx(t)I'"Y
sgn x(t) = 0,
where a(t) E C([to,oo),lR+). For more details see Butler and Erbe [3].
2. The results of Section 6.2 are taken from Dosly [5]. The method used
in Theorems 6.2.1 and 6.2.2 as well to establish (6.2.8) and (6.2.9) suggests
the following general approach to the investigation of oscillatory properties
of equation (6.2.13). In this method, this equation would be viewed as a
perturbation of the general nonoscillatory equation
(a(t)1f;(x'(t)))' + qo(t)1f;(x(t)) = O.
(6.4.1)
However, except for the special case treated in Theorem 6.2.1, no 'halflinear' analogy of (6.2.8) and (6.2.9) is known. The main reason is the
absence of equivalents of principal and non principal solutions, and also one
does not have a transformation theory similar to that for linear equation.
Using the linear transformation theory, for example, criterion (6.2.8) may
be proved as follows:
If Xo is the principal solution of equation (6.2.7), then the transformation x = xou transforms this equation into the one term equation
(a(t)x5(t)u'(t))'
0,
(6.4.2)
(a(t)x6(t)u'(t))'
+ [q(t) -
qo(t)]x6(t)u(t)
O.
490
Chapter 6
6.5. References
1. R.P. Agarwal and P.J.Y. Wong, Advanced Topics in Difference Equa-
Chapter 7
Oscillation Results for
Differential Systems
7.0. Introduction
In this chapter we are concerned with the oscillation of nonlinear twodimensional differential systems and second order vector-matrix differential
equations. In Section 7.1 we shall present criteria for the oscillation of nonlinear two-dimensional differential systems. This includes the superlinear,
linear, and sublinear cases. Section 7.2 deals with the oscillation of linear
second order differential systems. Here, first the system considered will be
reduced to a certain scalar Riccati inequality, so that the known results
from the literature can be applied to obtain oscillation criteria. Then, we
shall employ the notation and definitions of Section 6.1 to present some general results. Finally, we shall use Riccati and variational techniques which
involve assumptions on the behavior of the eigenvalues of the coefficient matrix (or of its integral) to present a number of sufficient conditions which
guarantee the oscillation of linear second order systems. In Section 7.3 we
shall discuss the oscillation of nonlinear second order differential systems
with functionally commutative matrix coefficients. Here, we shall show
that the oscillation theory of such systems can be effectively reduced to
the study of diagonal systems of scalar second order differential equations.
In Section 7.4 we shall prove some comparison theorems of Hille-Wintner
type for second order operator-valued linear differential equations. In Section 7.5 some oscillation results for second order differential systems with
a forcing term are given.
Chapter 7
492
x'(t)
ql(t)h(y(t))
-q2(t)!2(X(t)),
v'et)
(7.1.1)
where
(i) qi(t) E C([to,oo),JR), i = 1,2 and ql(t) 2:: 0 for t 2:: to,
(ii) fi(X)EC 1 (JR,JR) and satisfy xfi(x) >0 for x~O, i=1,2, hex)
is increasing on JR and f~(x) 2:: 0 for every x ~ O.
In what follows, we shall restrict our attention only to the solutions of
the differential system (7.1.1) which exist on some ray [To, 00), where
To 2:: to may depend on the particular solutions. Of course, under quite
general conditions existence of such solutions can always be guaranteed. As
usual, a continuous real-valued function defined on an interval [To,oo) is
said to be oscillatory if it has arbitrarily large zeros, and otherwise, it is
said to be nonoscillatory. A solution (x, y) of the system (7.1.1) is called
oscillatory if both x and yare oscillatory, and otherwise, it is called
nonoscillatory. The differential system (7.1.1) is said to be oscillatory if an
its solutions are oscillatory. We remark that if the coefficient ql(t) is not
identically zero on any interval of the form [TO, 00) where TO 2:: to, then
from the first equation of (7.1.1) for any solution (x, y) the oscillation of
x implies the same for y. So, if (x, y) is a nonoscillatory solution of
(7.1.1), then x is always nonoscillatory.
The special case where fi(X) = Ixl>'i sgn x, x E JR (Ai> 0), i = 1,2 is
of particular interest. In this case the differential system (7.1.1) becomes
x'(t)
v'et)
where Ai, i
of (7.1.1).
= 1,2
(7.1.2)
In the particular case where Ql(t) > 0 for t 2:: to and hex)
JR, (7.1.1) reduces to the equation
== 1 for
xl/(t)
O.
x, x
(7.1.3)
+ Q2(t)f2(X(t))
O.
(7.1.4)
493
xl/(t)
+ q2(t)lx(tW
= 0,
sgn x(t)
(7.1.5)
(7.1.6)
Moreover, it is well-known that if q2(t) ::::: 0 for t::::: to, then condition
(7.1.6) is also necessary for t.he oscillation of equation (7.1.5). Our first
t.heorem generalizes this result to the system (7.1.1). For t.his, we define
R(t) = ftto q1(s)ds and assume that
lim R(t) =
t -+co
(7.1. 7)
00,
CO
du
h(u) <
00
and
du
-CO
12(u) <
00.
(7.1.8)
TheoreIll 7.1.1. Suppose conditions (7.1.7) and (7.1.8) hold. If q2(t) ::::: 0
for t::::: to and
00
R(S)q2(S)ds
00,
(7.1.9)
to
A(v( t))
Z' (t)
q1 (t)
v'(t)
-Q2(t)i2(z(t)),
where ii(u) = - ii( -u), u E lR, i = 1,2. The functions ii, i = 1,2 are
subject to the conditions posed on ii, i = 1,2. Thus, we can restrict our
494
Chapter 7
or
00 ),
and assume
(h) Let y(t) ::; 0 for t 2:: to. The second equation of (7.1.1) implies
that y(t) is decreasing on [to, (0) and y(t) -+ either - 00, or a finite
negative number as t -+ 00. Similarly, it follows that h (y(t)) -+ either
- 00,
or finite negative number as t -+ 00. This in view of condition
(7.1.7) implies that
q}(s)h(y(s))ds = -00. Now, integration of the
first equation of (7.1.1) gives
It':
x(t)
= x(to) +
t qds)h(y(s))ds
-+ -
1to
as t -+
00
00,
which contradicts the assumption that x(t) > 0 for t 2:: to.
(1 2 ) Let y(t) 2:: 0 for t 2:: to. Define w(t) = h(y(t))/h(x(t)), t 2:: to.
Then as a consequence of the system (7.1.1) the following equation holds
w'(t)
+ q}(t)f~(X(t))W2(t) + q2(t)ff(y(t)) =
(7.1.10)
O.
Clearly, q}(t)f~(x(t))w2(x(t)) 2:: 0 for t 2:: to, and since x(t) > 0 and
y(t) 2:: 0 for t 2:: to, y(t) decreases to a nonnegative constant. Also there
for
We thus have
w' (t) ::; - kq2 (t)
[O,y(to)].
(7.1.11)
for t 2:: to
(7.1.12)
R(t)w(t) Since,
to
t q}(s)w(s)ds
1to
q}(s)w(s)ds =
it
to
::; - k
x'(s)
f ( ( ))ds
2 X S
t R(S)q2(S)ds.
(7.1.13)
lto
=
(t)
x(to)
du
-(f) <
2 U
kIt:
00,
00,
which
lto
(7.1.14)
= ft q2(s)ds,
OO
495
y(t o)
- h(x(to))
ft
fT,
f~(x(s))
ito
y'(s)
ds
f2(x(s))
y(t)
y(to)
- h(x(t)) + h(x(to)) -
t
' f~(x(s))
ito y(s)x (s) fi(x(s)) ds
y(t)
f2(x(t))
y(to)
[ - f(x(to))
'
ft
f~(x(s))
fT,
for every
2: T.
t [- h(x(s))
y(s)] [ , f~(X(S))]
-x (s) h(x(s)) ds.
y(t)
h(x(t)) 2: c + iT
w(t)
h(x(t))
ft
w(s)
f~(X(S))]
for t 2: T.
du
JI 0 12 (u) <
h(u)
-00
00
and
JI(h(u)) satisfy
du
JI 0 h(u)
<00
(7.1.15)
and suppose that conditions (7.1.7) and (7.1.14) hold, and for all sufficiently
small u and every v> 0,
(7.1.16)
496
Chapter 7
If
00 ql (s)h (Q(s))ds
(7.1.17)
00,
to
00
to
f'(x(s))
y(s)x/(s) f~(x(s)) ds <
(7.1.18)
00.
For this, assume that (7.1.18) is not satisfied. By condition (7.1.14) there
exists a constant k such that
y(to)
- Jz(x(to))
lt
to
>
tl
to
f'(x(s))
y(s)x/(s) f~(x(s))ds > 1- k.
y(to)
- h(x(to))
lt
to
q2(s)ds+
itl /
to
fHx(s))
y(s)x (s) n(x(s))ds 2': 1
lt
x(t) - x(td =
ql(s)h(y(s))ds
t,
h(b)
lt
ql(s)ds,
t,
-00,
a contradic-
Q(to) - Q(t)
lt
to
q2(s)ds = -
lt
to
y/(s)
Jz(x(s)) ds
y(t)
y(to)
- Jz(x(t)) + h(x(to)) -
00
it /
to
f' (x(s))
y(s)x/(s) fhx(s)) ds,
f~(x(s))
y(t)
h(x(t)) = 8 + Q(t)
497
rOO
f~(x(s))
(7.1.19)
100
y(to)
8 = f2(x(to)) - Q(to) -
to
f~(x(s))
rOO
iT
f~(x(s))
roo
it
and
q2(s)ds <
y(to)
- h(x(to))
lt
to
q2(s)ds +
IT
I
f~(x(s))
y(s)x (s) fi(x(s)) ds
to
00
f~(x(s))
100
t
Q2(s)ds -
f2(x(S))
and so Lemma 4.3.1 ensures that y(t):s; b1 for all t 2': T, where b1 =
(8/2)h(x(T)) < O. Hence, exactly as in establishing (7.1.18) we arrive at
the contradiction limt-.oo x(t) = -00, which proves the claim.
Finally, (7.1.19) guarantees that y(t) 2': Q(t)f2(x(t)) for every t 2': to.
Hence, in view of limHoo Q(t) = 0 and the condition (7.1.16), from the
first equation of (7.1.1), we obtain for t 2': to,
and consequently
du
(t)
x(to)
h(u)
2':
it
to
Ql(s)h(Q(s))ds
for all
2': to
to
Ql(s)h(Q(s))ds :s;
100
x(to)
f ; ( ) <00
1
2 U
for
2': to,
498
Chapter 7
yl/3(t)
y'(t)
sin t
- ( --+
t
roo
= it
q2(s)ds
roo
= it
(7.1.20)
2 + cos t) 5()
x t
t2
= 1, q2(t) = (sint/t)+(2+cost)/t2 ,
-d
(2 +
2+cost
coss)
s
and hence Q(to) = Q(1f/2) = 4/1f exists as a real number. Also, we find
that F(u) = h 0 h(u) = u 5 / 3 which is a superlinear function and satisfies
Joo u- 5 / 3du < 00. Finally,
= i t (2+COSS)1/3 dS -+
tq1(s)h(Q(s))ds
ito
7r/2
00
as
t-+oo.
Thus all conditions of Theorem 7.1.2 are satisfied, and hence the system
(7.1.20) is oscillatory.
1
du
io hoh(u)
If
<
00
00
(7.1.21)
<
and
q2(s)h(R(sds =
00.
00,
to
(7.1.22)
(7.1.23)
499
Proof. First we shall show that for any fixed tl > to,
(7.1.24)
Since limHoo R(t) = 00, there is a t2
t > t2.
for
(7.1.25)
Thus, we have
(7.1.26)
and hence by condition (7.1.21) for t > t2,
(7.1.27)
(b) Assume that y(t);::: 0 for t;::: to. From the system (7.1.1) we note
that x(t) is an increasing function on [to, 00) while y(t) is a decreasing
function on [to, 00). Integrating the first equation of (7.1.1) from tl to
t, we obtain
x(t) = x(to)
lto
ql(s)h(y(s))ds;:::
lto
R(t)fI(y(t)).
ql(s)h(y(s))ds
(7.1.28)
fI(y(t)).
y'(t)
fI(y(t)) > q2(t)h(R(t))
for
t;::: to
(7.1.29)
y'(t) S; -q2(t)
(7.1.30)
Y (t O)
y(t)
du
2
f ( );:::
1 U
it
to
q2(s)h(R(s))ds.
(7.1.31)
Chapter 7
500
The right-hand side of (7.1.31) tends to 00 as t --+ 00, while the left-hand
side of (7.1.31) remains bounded by (7.1.22) as y(t) --+ 0 as t --+ 00. This
contradiction completes the proof.
Example 7.1.2.
x'(t)
y3/5(t)
y'(t)
r14/9x5/9(t),
2 to
1.
(7.1.32)
= x 5/ 9
Thus all conditions of Theorem 7.1.3 are satisfied, and hence the system
(7.1.32) is oscillatory.
satisfies
(7.1.33)
'U---*(X)
If
lim
t .....
=lto
fJi(s)ds =
00,
= 1,2
(7.1.34)
_ltot h(x(s))
y'(s) ds
t q2(s)ds
lto
= Q*(t).
(7.1.35)
y(t)
- h(x(t))
Q (t)
+ -
y(to))
h(x(to))
+ lto
,f~(x(s))
y(s)x (s) fi(x(s)) ds. (7.1.36)
501
+ (-y(to)/h(x(to))) 2': 1
t ( -x (s) /1(x(s))
fHX(S)))
(-y(s))ds
y(t)
- h(x(t)) 2': 1 + it1
for all
2': t 1
(7.1.37)
Hence, in particular y(t) <
for t 2': tl. Now, Lemmas 4.1.1 and 4.4.1
allow us to compare -y(t) with the solution u(t) of the integral equation
u(t)
h(x(t))
1+
to conclude that
f~(X(S)))
it1rt (-x,(8) n(x(s))
(u(s))ds
for all
for all
2': h (7.1.38)
t 2': t 1 .
(7.1.39)
It
Ql(8)JI(y(s))ds :::;
t,
kIt
ql(s)ds -> -
00
as
t ->
00.
t,
x(t) = X(tl) -
t ql(s)JI(y(s))ds.
it l
(7.1.40)
Remark 7.1.1. In all the results proved above some positivity condition
has to be imposed on one of the coefficients qi(t). To see this, we will
construct an example of a nonoscillatory system (7.1.1) for which condition
(7.1.34) is satisfied.
(7.1.41)
-
where Q(t) = w(to) + fo q2(s)ds. The question whether the system (7.1.1)
has a solution x(t) of fixed sign (say positive) is equivalent to the question
whether (7.1.41) has a continuous solution on [0, (0).
Chapter 7
502
In what follows we shall construct step functions Q(t) and q2(t) so that
(7.1.41) has a solution on [0,00). We shall allow qi(t) to change sign, even
though condition (7.1.34) is satisfied. Although our smoothness condition
on qi(t) precludes such step functions, and the solution has jumps at each
of the points 3k, 3k + 1, modification of our example by smoothing out
the abrupt jumps easily leads to an acceptable counterexample.
_{-1,tE[3k,3k+1)
.
rt
_
1, t E [3k + 1, 3k + 3). Then, hmHoo Jo ql(s)ds - 00.
Let qi(t) -
a(k) 2 k -+ 00
as
k-+oo.
(7.1.42)
k
- - - a < O.
k+1
(7.1.43)
(7.1.44)
We assumed that the above inequality is true when the upper limit of
integration is 3k and concluded that the same is true for 3(k + 1).
This induction step allows us to infer that (7.1.44) holds for all k > O.
Continuing this process over each interval, we obtain a solution w(t)
defined on the whole half-line [0,00), and so we have an example of a
nonoscillatory system (7.1.1) for which condition (5.3.34) is satisfied.
It is natural to ask whether some relaxation on the positivity requirement is possible. As shown by the above example, it is not sufficient to
assume that at each t one of the q/s is nonnegative.
Remark 7.1.2. We note that Theorem 7.1.4 holds without the assumption
that II is increasing on JR. Clearly, the increasing character of II on
JR implies condition (7.1.33).
503
- tanh y(t)
x'(t)
y' (t)
- [-v'tsint+
2~(2+cost)]
(7.l.45)
for
j q2(s)ds =
t
to
} 1f /2
t ---+
as
00
} 1f /2
d[VS(2+cos s)]
00.
All conditions of Theorem 7.1.4 are satisfied, and hence the system (7.1.45)
is oscillatory. However, Theorems 7.1.1 - 7.1.3 are not applicable.
x"(t)
+ Q(t)x(t) =
(7.2.1 )
0,
where x: [to, (0) ---+ JRn and Q(t) is a continuous real symmetric n x n
matrix function for t 2: to 2: O. In 1980, Hinton and Lewis conjectured
that (7.2.1) is oscillatory whenever
lim Al
t~oo
(jt Q(S)dS)
to
(7.2.2)
00,
~t
tr
t~oo
or
liminf
t~oo
where tr()
oscillatory.
tr
(jt Q(S)dS)
to
> -
(jt j8 Q(U)dUdS)
to
to
(7.2.3)
00,
> _
00,
(7.2.4)
504
Chapter 7
We recall some pertinent definitions and notation which will be subsequently used. For any n x n matrix A, the transpose will be denoted
by A*, similarly x* denotes the transpose of the column vector x. If
h, t2 E [to, 00), tl =1= t2 and if there exists a nontrivial solution of (7.2.1)
which vanishes at hand t2, then tl and t2 are said to be (mutually)
conjugate relative to (7.2.1). System (7.2.1) is said to be disconjugate on an
interval J c [to,oo) if every nontrivial solution of (7.2.1) vanishes at most
once in J and system (7.2.1) is said to be oscillatory if for each tl > to
there exists a t2 > tl such that (7.2.1) is not disconjugate on [tl, t2].
The matrix differential system associated to (7.2.1) is
X"(t)
+ Q(t)X(t)
0,
[to,oo)
(7.2.5)
X*(t)X'(t) - (X*(t))'X(t)
0,
[to, 00)
(7.2.6)
X*(t)X'(t) - (X*(t))'X(t)
C,
(7.2.7)
(7.2.8)
and
(7.2.9)
k=l
505
W(t)
It:
W(to)
where Ql(t) =
Q(s)ds.
obtain the scalar equation
+ Ql(t) +
W2(s)ds,
(7.2.10)
= tr W(to) + tr Ql(t) +
tr W(t)
ltD
ltD
tr W 2(s)ds.
(7.2.11)
Now suppose condition (7.2.2) holds, i.e., limt-;oo Al(Ql(t)) = 00. Then,
Al(W(tO)+Ql(t))--+oo as t--+oo, and given 1'>0 we can find a large
T 2:: to such that
(7.2.12)
Hence, for all t 2:: T,
tr W 2(t) =
We use this estimate in (7.2.11) to derive an inequality for the scalar quantity tr W(t). For this, we split the integral term in (7.2.11) as follows
ltD
tr W2(s)ds =
I'
{t tr W2(s)ds + (1 - E)
ltD
ltD
tr W2(s)ds.
(7.2.14)
Assuming t 2:: T and ignoring the contribution from the interval [to, T] to
the last integral, we estimate the remainder by means of (7.2.13), to obtain
(7.2.15)
In the first integral we use the simple estimate tr W 2(t) 2:: 2:~=1 w~k(t) 2::
(l/n)(tr W(t))2, where Wkk(t) is the kth diagonal entry of W, to get
the inequality
(7.2.16)
Thus, combining equations (7.2.11), (7.2.14) with inequalities (7.2.15) and
(7.2.16), we find
tr W(t) 2:: tr Ql(t)+tr W(to)+(1-E)3
lT
(7.2.17)
Cbapter 7
506
This inequality holds for all t 2': to. Using condition (7.2.2) we get for all
sufficiently large t 2': TI 2': T,
q(t)
tr QI(t)
replace
81
by
>'i(QI(S))ds,
nE
t 2': to
(7.2.18)
to
Elt w
2(s)ds
for
t 2': T I .
(7.2.19)
to
Next, we state some known sufficient conditions which ensure that the
inequality (7.2.19) is incompatible with the existence of a continuous function w(t) on the entire half-line [to, 00), to 2': O. For t 2': to, we let
Q1(t) = ft~ Q(s)ds and Q2(t) = ftto QI(s)ds.
00)
with mes J =
(7.2.20)
tE J, t-->oo
q'Y(s)ds
00
00)
00
(7.2.21 )
00
. . f
I1m
1Il
ftto
Hoo
p(s)q(s)ds
t
fto
p(s)ds
> -
00,
>
t _ to
(7.2.22)
t-->oo
It
to
[(q(s) - >.)+fds
= max{4>(t),O}.
00
(7.2.23)
507
. I:
Hoc
Ito p( s )ds
Now we shall apply these criteria to present oscillation results for (7.2.1).
IJ
Proof. Suppose condition (7.2.2) holds. Then equation (7.2.1) is oscillatory or there exists a function w(t) which is defined and continuous
on the half-line [to, (0) such that inequality (7.2.19) holds for all sufficiently large t. The later possibility is excluded by each of the criteria (Ii)'
i = 1,2,3,4.
q* (t)
~ tr Q2(t) > -
00.
(7.2.25)
lit
>0
1
(trQ1(s)+cs)ds 2': -ct.
508
Cha.pter 7
tm).
In our next result we shall use the notation and definitions of Section
6.1, and will need the following lemma.
1t
to
V 2(S)dS)]
L <
00,
(7.2.26)
gt
1t
to
>'l(V2(s))ds <
(7.2.27)
00.
Proof.
limapproxinf -1()
t --+ 00
9 t
then
1
limapproxinf -()
t --+ 00
9 t
1t
to
We define H(t) = tr
1t
tr(V2(s))ds
(Itt
to
Al(V2(s))ds
00,
(1 V2 (s )dS)
t
to
o V2(s)ds)
00.
{t:
509
then by (7.2.26) we find that P(E1) < 00; since M > L and so the
complement of E1 relative to [to, 00) which we denote by Ef, satisfies
p(ED = 00, and H(t) < -tr V(t) on E'( Since p(E n ED = 00 we
see that the inequality g(t) < H(t) < -tr V(t) holds on En E'( Thus,
:s:
g2(t)
:s:
H2(t)
Ef.
holds (a.e.) on E n
.!.
1t
+ Mg'(t))
H2(t) -
H2(t)
n tr (V2(T)) = n(H'(t)
n -
:s:
(tr V(t))2
H2(t)
g2(t)
(7.2.28)
1
t
to
H'(S)
H 2( S ) ds
+M
to
g'(S)
~() ds
9
<
(XJ u- 2 du + M
roo
u- 2 du
19(t o )
1 H(to)
- + g(to)
-- <
H(to)
00
00.
limapproxinf _(1)tr
t---+oo
gt
(1 Q(S)dS) -l >
t
to
-00,
l is a constant, (7.2.29)
00
and that
lim .A1
t-+oo
(1 Q(S)dS)
t
00.
to
(7.2.30)
(7.2.31)
_V'(t) = Q(t)
+ V2(t),
t?,: h.
(7.2.32)
-V(t)
+ V(t1) =
1t
t1
Q(s)ds +
1t
t1
V2(s)ds
(7.2.33)
Chapter 7
510
and hence
-V(t)
+ V(t 1 ) 2: it Q(s)ds.
(7.2.34)
t,
Al (- V(t))
+ Al (V(tl))'
~p~r~ sup
lim~p~r~sup
00.
we find
Now, by conditions
[t V (S)dS) ) ]
2
00,
limapproxinf - ()
t --+
00
(J/, Q(s)d.s)
<
00.
(7.2.36)
for all large t, and now taking lim approx inf on both sides of (7.2.36)
as t -+ 00 we get a contradiction to condition (7.2.30). This completes
the proof.
Remark 7.2.1.
implies that
If g(t)
liminf
replaced by
511
-ktO"
where q(t)
-00
= 8t b - I -17t'1-I, t 2:: to =
as t --+
g(t) =
Then, gCt)
-q(t)
00
and
)11
[AI
too
Q(U)dU)
00
dS) -y
as t --+
where
= t'1 -
t b --+
Let
00.
1 it tr (Q(s))ds
g(t)
1. Then, I: tr(Q(s))ds
(it (is
.
(
t'1 - t b > - to--Y
2'1+1 )
,
t(2'1+ Ih -
r ~ [l (~, (l
rv
00
d,
Q{U)dU)) '
d'l
->
00
as t --+ 00. By Theorem 7.2.3 oscillation is guaranteed if 0 < 17 < 8 < 217+ 1
(choose 'Y = (8/217) + 1). If 8> 1, we have (l/t) I: tr(Q(s))ds --+ -00,
so we cannot deduce oscillation by Theorem 7.2.1.
In our next result we will relax condition (7.2.2), and replace hypotheses
(7.2.29) and (7.2.30) of Theorem 7.2.3 by a condition involving the relative
rates of the growth of the largest and smallest eigenvalues of It: Q(s)ds
as t --+ 00. For this, we denote and order the eigenvalues of It: Q(s)ds
as AI(t) 2:: A2(t) 2:: ... 2:: An(t).
We shall need the following lemma.
Lemma 7.2.2. Let p(t) be locally bounded, nonnegative and measurable
on [to, 00) and nonzero a.e. Let q(t) be nonnegative and locally integrable
such that p(t) 2:: q(t) It: p2(s)ds for almost all t 2:: to. Then for all
sufficiently large T 2:: to, q E .c 2 [T, 00).
Proof. Let P(t) = It: p2(s)ds. Then, P(t) is absolutely continuous
and by hypothesis, P(t) > 0 for t 2:: tI, say. We have P'tt) = p2(t) 2::
q2(t)p 2(t), and hence PI(t)/ p 2(t) 2:: q2(t). Integrating this inequality
from T>tl to t, weobtain (1/P(T))-(1/P(t))=I;q2(s)ds, andso
I; q2(s)ds::; l/P(T), proving the lemma.
Chapter 7
512
(I) (i)
limt~r~inf >'l(t) =
(II) (i)
lim~pJ::r:sup
A1(t)
00,
and (ii)
and (ii)
= 00,
<I>(t) = W(t)
-It
(t Q(s)ds + C,
W2(s)ds =
t 2> t1 2> to
ltD
t,
(7.2.37)
(1:
y*(t)
Q(S)dS) y(t)
(7.2.38)
Adt).
Let the eigenvalues of W(t) be J.l1(t) 2> J.l2(t) 2> ... > J.ln(t). By
the preceding remark, we may select a system of (orthonormal) locally
integrable eigenvectors ei (t), i = 1, 2, ... ,n such that
W(t)ei(t) = J.li(t)ei(t),
1 if i
0 if i
ei (t)ej(t) = lS ij =
= 1,, n, to
s, t <
<::::
00
I'= j.j
(7.2.39)
by
y(t)
L Ci(S, t)ei(s).
(7.2.40)
i=l
W(s)y(t)
(7.2.41)
i=l
y*(t)<I>(t)y(t) = t;J.li(t)c7(t,t) -
t (
t;J.l7(s)c7(s,t))
ds.
(7.2.42)
513
cos ()(t),
C2(tO, t) = sinB(t).
(7.2.43)
(7.2.44)
From (7.2.43) and (7.2.44), we find that Cl(S, t) = cos[B(t) - (s)], C2(S, t)
= sin[()(t) - (s)]. If we put o:(t) = ()(t) - (s), we may write (7.2.42) as
-1:
p,~(s) sin2[o:(t)+(t)-(s)]ds.
(7.2.45)
Now y*(t)<P(t)y(t) = Al(t) + y*(t)Cy(t). Since C is a constant matrix
and Ily(t)11 = 1, we have
limapproxinf
t ---+ 00
or
y*(t)if>(t)y(t)
limapproxsup
t ---+ 00
00,
(7.2.46)
00
(7.2.47)
y*{t)if>(t)y{t)
51,
(7.2.48)
if
(I)
holds,
(7.2.49)
if
(II)
holds.
(7.2.50)
p,{5'1) <
00
:s: A1(t)//j.
514
Chapter 7
For any x
[y(t) + x]*iI>(t)[y(t) + x]
=
>
)'1 (t)
>
Adt)-~A1(t)-1166Adt)-(1+~6rIICII
~O)
IICII
(7.2.51 )
if tE51
x(t)
={
0 if la(t)1 :::: 6/8 (mod 7f) and la(t) - (7f/2)1 :::: 0/8 (mod 7f),
x(t) otherwise,
and let y(t) = y(t) + x(t). If 5(t) is defined by y(t) = cos B(t)edto)
sinB(t)e2(t), 5(t) = B(t) - (t), then we have
A
la(t)1 ::::
8'
(mod 7f),
15(t) -
~I
::
(mod 7f)
(7.2.52)
y*(t)iI>(t)j)(t)
-it J-L~(S)
-it J-L~(S)
to
(7.2.53)
to
::::
1
3A1(t) - 211ctl,
t E 51'
Part (i) of the hypotheses of the theorem yields the existence of a subset
52 of [to, (0) such that
1
t E 52
(7.2.54)
J-L(5:jJ <
00
if (I) holds,
(7.2.55)
J-L(52) =
00
if (II) holds,
(7.2.56)
515
where 8 2 = [to, 00) - 8 2. Now (7.2.50), (7.2.52) - (7.2.56) show that there
exists a subset 8 = 8 1 n 8 2 of [to, 00) with M(8) = 00 such that
ltD
where
O! (
> 0 for
t E 8,
ltD
+ >(t)
>(s)]ds
- >(s)]ds
(7.2.57)
(7.2.58)
and let J 1 = {t: U1(t) > O}. By (7.2.52) there exist 01 > 0 and a positive
integer m such that
1
Iwl:S; -
(mod 1f).
(7.2.59)
ltD
= { IM1 (t) I, t
Ai
0, otherwIse.
(fi,i(S))2ds.
(7.2.60)
It:
U2(t)
(t M~(S) sin2[&(t)
ltD
+ >(t) -
>(s)]ds
and h = {t: U2(t) > O}, then M(82 ) < 00. But this contradicts (7.2.56),
and this completes the proof of the theorem for the case n = 2.
For the general case n > 2, the basic idea of the proof is the same.
We introduce the orthogonal matrix U(s) whose rows are the eigenvectors
ei(s) of W(s) (see (7.2.39)). Denote the vector (C1(S, t),, cn(s, t))*
by c(s, t). Assuming without loss of generality that U(to) = I, where I
is the identity matrix, then (7.2.40) gives
c(s, t)
= U- 1(s)c(to, t).
(7.2.61)
Chapter 7
516
1:
tJ-Li(t)V;(t) =
tJ-L;(S)[Vi(t)
+ 'Wi(S, t)fds.
(7.2.62)
y*(t)cp(t)y(t)
(7.2.63)
where IVi (t) I 2 60 > 0 on Sand S is a subset of [to, 00) of infinite
measure, Wi(S, t) is the ith component of [V(s) - V(t)]c(t o, t) and
Ilc(to, t)11 = 1. Note that (7.2.63) is the analog of (7.2.53) and (7.2.56).
Since the n x n orthogonal matrices (identified with zn-l) form a
compact set, we may find a finite decomposition of zn-l, {Edk=l' say,
such that G, H E Ek imply IIG - HII < 60 /2, k = 1,2"", m. Now
define Ak = {t 2 to: Vet) E Ed. Then we shall have
[Vj(t) + Wj(s, tW 2
whenever s, tEAk, k
~65,
(7.2.64)
With (7.2.63), (7.2.64) and Lemma 7.2.2, we may complete the proof as
in the case n = 2.
lot Q(s)ds
=
0]
~tl/2sin2t[1 + t n -
+ cost)]
(1
+ cos t)
, A,l(t) , > 1 if
A,3(t) -
C3
517
lim
T-+oo
IT
t
W2(s)ds
[to, 00).
(7.2.65)
L(Q)
litiS
liminf -
t-+oo
to
tr Q(u)duds >
to
(7.2.66)
00.
Proof. Suppose first that the limit in (7.2.65) exists for some solution
W(t) = -X'(t)X-1(t) of (7.2.10) where det X(t) =/= 0 for t ~ to. Then
from (7.2.10), we have
tr W(t)
where C
+ tr
00
[tr W(s)
: :; ~ [1:
-1
00
Ito;' W 2(s)ds.
tr W 2(U) dU
[(tr W(S2
(1
t -+
1: (1
00
tr Q(s)ds - C,
(7.2.67)
to
= -tr W(to) - tr
1:
1t
W2(s)ds =
00
Since
ds
tr W 2(U)dU)
we find (l/t)
tr W2(u)du
2] dS]
(7.2.68)
It: tr W2(s)ds -+ 0
ds -+ 0 as t -+
as
00
~
t
it [1
to
to
tr Q(u)du -
c]
ds -+ 0
as
t -+
00.
(7.2.69)
Chapter 7
518
Now since in view of Cauchy-Schwarz inequality
it follows that
lim ~
t-+oo t
Jtot [Jtot
tr Q(U)dU] ds
exists
(7.2.70)
lt
1
--
to
W(s)ds+-1
t
lt 1
5
to to
l I l t1
5
W2(s)duds--(t-to)W(to)-t
to to
Q(u)duds
(7.2.71)
11t
-t
to
tr W(s)ds
11t 1
+-
to
/'0
tr W2(u)duds ~ M
Itto tr
11 l
t
to to
tr W2 (u )duds --+
00
as
It:
Itto l
to
tr W2(u)duds ~ -1
t
Itto
tr W(s)ds
t --+
00
(7.2.72)
Itto tr W2(s)ds
2 (s )ds
--+
00
as
00.
as t --+
21tto
+M < t
00
and so
tr W(s)ds.
(7.2.73)
I~
10
t
tr W(s)ds
[1 10
10
so (7.2.73) gives
< 4n
t
Jto
tr W 2(s)ds.
(7.2.74)
h: h: tr W2(u)duds,
t::::: tl
519
tl::::: to
(7.2.75)
1 < H'(t)
4nt - H2(t) ,
(7.2.76)
h:
exists as a finite limit. This implies the existence of limt-+oo ftto W2(s)ds,
as the following argument shows: Let the (operator) norm of a matrix A be
denoted by IAI. For to s:: s s:: t, define A[s, tl by A[s, tl =
W 2(u)du.
Then, A[s, tl is a nonnegative definite matrix and
f:
s::
tr Als, tl = 1 \ r W2(u)du -+ 0 as s, t -+
00.
f:
(Id
lim sup
(12)
lim sup
t-+oo
(13)
(14)
t-+oo
~ 1.t Al (1.
t to
Q( u )du ) ds
limapproxsup Al
t -+
to
00
limapproxinf Al
t -+ 00
(1. Q(S)dS)
(1. Q(S)dS)
t
to
t
to
(7.2.77)
00,
00,
(7.2.78)
00,
(7.2.79)
- 00.
(7.2.80)
Proof. (II) Assume conditions (7.2.66) and (7.2.77) hold. Suppose there
exists a prepared solution X(t) of equation (7.2.5) which is not oscillatory.
Without loss of generality, we may suppose that det X(t) i- 0, t:::::: to so
that from equation (7.2.10), we have
(7.2.81 )
Chapter 7
520
By the convexity of ),1 and the fact that ftto W2(s)ds ~ 0, t ~ to, (i.e.,
nonnegative definite for t ~ to), we find from (7.2.81) that ),l(W(t)) +
),l(-W(tO))
lit
to
),l(W(s))ds
t - to
+ --),l(-W(tO))
~t
1t (1
to
A1
to
T1
iT
n to
A1(W(s))ds ---+
Q(U)dU) ds
{Tn}~=l'
Tn ---+
as
00
(7.2.82)
Tn ---+ 00
(7.2.83)
00.
[L t
(),,(W(')))'d'l 'I'
as Tn ---+
00.
But since L(Q) > -00, Lemma 6.2.3 implies that ft'; A1(W 2 (S))ds S
ft'; tr W2(s)ds < 00. This contradiction proves part (It).
roo W2(s)ds + 0
it
Q(s)ds,
(7.2.84)
<
),1
(W(t)
< tr (W(t)
1
+1
1
+
00
00
W2(s)ds
W2(s)ds
00
0)
or
W 2(S)dS) 2 + 2tr 0 2
(1=
W 2 (S) dSr
+2tr 0 2,
+ B)2
~ 2(tr
521
A2 + tr B2. Therefore,
ds
~ ~ l>r
41t
+t
to
(1
00
W2(u)du
tr W (s)ds
ds
+ 2(t - to)
t
tr C .
(7.2.85)
If we set A(t) = ftOO W2(s)ds, then tr A(t) -+ 0 as t -+ 00 and A(t) ;::: 0
so that Al(A(t -+ 0, and hence Al(A2(t -+ 0 as t -+ 00. Therefore,
tr A2 (t) -+ 0 as t -+ 00. Thus, the integrals on the right-hand side of
(7.2.85) tend to zero as t -+ 00, and the last term is bounded. However,
condition (7.2.78) implies that the left-hand side of (7.2.85) is not bounded.
This contradiction completes the proof of part (12).
(13) Assume conditions (7.2.66) and (7.2.79) hold. As in part (1 2 ), we
obtain equation (7.2.84) so that
Al(W(t +
00
Al(W2(sds ;::: Al
(1:
Q(S)dS) + An(-C).
(7.2.86)
!too
00
+ 1,
(7.2.88)
i.e., fEJAl(W(S)2ds = 00, where Ek
contradiction proves part (13).
(14) Assume conditions (7.2.66) and (7.2.80) hold. The proof is similar to
part (13). Since for any M > 0,
(7.2.89)
and since
Chapter 7
522
it follows that if M> l+IAl(-C)I, we have p,{t: Al(W(t)) :::; -I} = CXJ
so that
(AI (W(s)))2ds = CXJ, which is a contradiction. This completes
the proof of the theorem.
It':
An (
limapproxsup
t --+ 00
-CXJ,
limapproxsup
t --+ 00
An (
lto
Q(S)dS) > -
(7.2.90)
CXJ.
and
lto
Q(S)dS) = m > -
CXJ.
as t -+
Al ( -
CXJ,
and hence
Al
(Itto W
It:
2 ( S ) dS)
Al (W(t)
-1:
Q(S)dS)
-+
CXJ
Al(W(t)) - An
(-1:
(1:
{t : (It: Q(s)ds) ~ m -
1
-tr
n
it
to
W2(s)ds
-+
-1:
-1:
-1:
-1:
Q(S)dS)
(7.2.91)
Q(S)dS) .
E}
CXJ.
~tr
(W(t)
Q(s)ds - W(to))
~tr
(W(t)
Q(S)dS) -
< Al (W(t)
Since
CXJ.
< Al(W(t)) + Al
as
Q(S)dS) -
< Al (W(t)) + Al (
Al (W(t)) - An
(1:
Now a.gain
~tr W(to)
~tr W(to)
Q(S)dS) -
Q(S)dS) -
~tr W(to)
~tr W(to)
(7.2.92)
523
and since (l/n)tr It: W2(s)ds ;::: (l/n) It: A1(W 2(Sds
follows that
~tr W(to) + ~
1:
(1:
from (7.2.92) it
Q(S)dS) , t;::: to
(7.2.93)
{ l I l+ -t
Jl t: -tr W(to)
n
to
00
(7.2.94)
00
as t ---+
00,
we find
then Jl(E) =
>'1 (W2(t
r P'(s)
lE p2(s)ds
;::: 4n 2Jl(E)
00,
r P'(s)
lE P2(S) ds
This completes the proof.
::::;
P(to
+ 1)'
lto
lto
ls
p2(u)duds
= It: It:p2(u)duds.
.c2[T, (0).
P'(t) ;:::
(7.2.95)
to
(_1_)
(it P(S)dS)
t - to
to
1/2
;:::
= It:p2(s)ds
(_1_)
q2(t)p2(t).
t - to
and
(7.2.96)
Chapter 7
524
Now, since I;'(P'(s)/p 2 (sds <
00
Al
(1:
Q(S)dS)
(7.2.97)
and
liminf
Hoo
Al (Itt Q(s)ds)
0
> O.
(7.2.98)
An (It: Q(s)ds)
t->oo
(7.2.99)
00.
cI>(t) = Wet) -
t W2(s)ds
ito
it
Q(s)ds + C,
i 2: to
(7.2.100)
to
Al
(1:
=
Q(S)dS)
+ y*(t)Cy(t)
tJ.li(t)vt(t)
i=l
-it
be
the
be
i =
y*(t)cI>(t)y(t)
J.lT(s) [Vi(t)
+ Wi(S, t)]2ds.
to i=l
Again as in the proof of Theorem 7.2.4 the given hypotheses of the theorem
allow us to find a unit vector yet), functions Pij(t), i = 1,"', n, j =
1, ,m and ~ > 0 such that
~Al
(1:
Q(S)dS) -
2: [Pi j (t) - ~
-,1
1:
(7.2.101)
Pij (s )dS] .
525
Hypothesis (7.2.99) and (7.2.101) imply that there exists a tI ;::: to such
that
(7.2.102)
If jlij == 0 a.e. for t;::: to, let ijij(t) = 0, otherwise define ijij by
(7.2.103)
By Lemma 7.2.4 each ijij has the property that ijij(t)IJi E 2[T, (0) if
> tI is sufficiently large, and so ij( t) = 2::i,j ijij (t) E 2 [T, (0) if T IS
sufficiently large. However, (7.2.102) implies that
[ij(t) -
l2:.1: 1:
t >T
',J
for
t > T,
ij( t) / Vi E
(It:
(It:
Q(s)ds) , or [>'1
Q(s)ds)
by these
place the average of >'1
expressions alone, and that one cannot replace 'lim approx sup' by 'lim
sup' in parts (13) and (14). For this, it suffices to consider a scalar counterexample. We choose v(t) E CI([to,oo),IR) n 2[to, (0) and define
q(t) = v'(t) - v 2(t). Then, x(t) = exp ( -
It: v(s)ds)
is a nonoscillatory
It:
Example 7.2.4.
The condition (7.2.99) is not by itself an oscillation
criterion for (7.2.5) when n> 1. To see this consider the case when Q(t)
is a 2 x 2 diagonal matrix for all t, say Q(t) = diag {q(t), q2(t)}. Then
(7.2.5) is uncoupled as two scalar equations
x~'(t)
+ qi(t)Xi(t) = 0,
= 1,2.
(7.2.104)i
Chapter 7
526
(7.2.104h are both nonoscillatory with
P2(t) ;:::
and
lim -1
t--->oo
it
0
pi(s)ds
(7.2.105)
pi(t) = max{0,P1(t)}.
where
00,
t~
for
(7.2.106)
C1(t)
1, tEsupportpi(t)
Al
0, otherwise,
= (C1(t),C2(t))*,
(lot Q(S)dS)
0, t E support pi(t)
1, otherwise
we have
;:::
y*(t)
Ci(t)P1(t)
+ C~(t)P2(t)
;::: pi(t).
{ ~: !~ ~~ ~ t ::::;
1,
10
am+l ;::: am
+I
m+2E~
w~(s)ds = am
+ b~fm >
S7n+m
( ~)m
2
= 1/(8bm+l),
~ (~) m
then in view of
16
2'am.
527
(3)-m
1
<
- 2(m+2)"
Now we extend WI (t) to [Sm+l, Sm+2] so that WI (t) = bm+1 on [Sm+l +
E
m+
1
<
-16
-2
-
1L'dt) -
iot
'IL,!
(s)ds ::0: bm
am -
Define ql(t) by ql(t) = w~(t) - wi(t). Then for t E [Sk + Ek,Sk + 2Ek],
we find Pl(t) =
ql(s)ds::O: (1/2)b k . Thus, if Sm :s; t:S; Sm+l, we obtain
Jd
I1t
t
Since
Sm rv
t1 L
m
pt(s)ds::O:
In m for sm:S; t
11t
k=1
p+(s)ds
1
(7) .
"2 bkEk = 16
rv
-1
16
t)
( ~
rv
et so that
lR n : 7](0) = 7](;3) = 0,
7](t) E AC[o,;3], 7]'(t) E L 2(0,;3)}.
(7.2.107)
The basic result we need here is that system (7.2.5) is oscillatory if and
only if there is a sequence of intervals {[an, bn ]} with limn-+oo an = 00
and a sequence of functions {<Pn}, <Pn E Al (an, bn) such that
A 1 (oc,;3)
{T,: [0,;3]
(7.2.108)
The idea now is to utilize conditions on the behavior of Q(t), or its integral
to construct sequences fulfilling (7.2.108).
Our first result in this direction is the following.
Theorem 7.2.8. Suppose that
limSUp.Al
t-+oo
(it
to
Q(8)dS)
00.
(7.2.109)
Chapter 7
528
Then equation (7.2.5) is oscillatory if either
(id AI(Q(t)) is bounded above on [to, 00), or
(h) An(Q(t)) is bounded below on [to, 00).
Proof. Choose any sequence ak
Yk
(l:
limsuPHoo Al
Q(S)dS) Yk 2: k,
lR n
(I:k Q(s)ds) =
with Ilxkll
k = 1,2, .. .
00.
= 1, so
(7.2.110)
Suppose that condition (h) holds. Then there exists M such that Al(Q(t))
~ M for t 2: to. Define (/>k E A(ak' bk) as follows:
(t - ak)Yk, ak ~ t ~ ak + 1
(/>k(t) = { Yk, ak + 1 ~ t ~ bk - 1
(-t + bk)Yk, bk - 1 ~ t ~ bk.
Clearly,
J:: 1<t{(s)1 ds = 2,
2
(7.2.111)
and
(t + 1 - ak)Yk, ak - 1 ~ t ~ ak
{ Yk, ak ~ t ~ bk
(-t + bk + l)Yk, bk ~ t ~ bk + 1.
(7.2.113)
= 2,
and we have
bk+l
l ak- 1>'k(s)Q(s)1>k(s)ds
1
1:~1 1>'k(s)Q(s)1>k(s)ds +
+
1:
529
1:d1
1>'k(s)Q(s)1>k(s)ds
1>'k(s)Q(s)1>ds)ds
k
k
[s + 1 - ak]2Y'kQ(S)Ykds
l akb Y'kQ(s)Ykds + la
ak-1
bk + 1
+
[b k + 1 - t]2Y'kQ(s)ykds
bk
k- M
[1:~1 [s + 1 -
ak]2ds
l:
+1
[bk
(7.2.114)
+ 1 - s]2ds
k--M
3
'
bk+l
l ak- [11>~(s)12
1
- 1>'k(s)Q(S)1>k(S)] ds <
2+
3M
- k < 0,
lot Al(Q(s))ds
lot q+(s)ds
lot[w~(s) -w~(s)]+ds.
j sm+<rn
w~(s)ds
Sm
Chapter 7
530
and hence,
Theorem 7.2.9. Let Q(t) = (qij(t)), i,j = 1"", n and each qij(t) is
bounded on [to,oo), and qii(t) = 0, i = 1"", n. Then system (7.2.5) is
oscillatory if there exist i, j such that
limsuplt qij(s)ds =
t-+oo
(7.2.115)
00.
to
x"(t)
+ Qij(t)x(t)
(7.2.116)
00
(7.2.117)
m E AI(am,bm ) (with n
Ia~
b~
Define 'm E Al (am, bm ) by 'm(t) = col (0, ... , m(t), 0, ... , m(t), ... ,0),
where the only nonzero entries are the ith and jth. Then,
Theorem 7.2.10. Suppose for each positive integer m::::: to there exists
a positive number Em, and for each positive integer k there exists a unit
vector Ymk E 1Rn such that the set
m-(e/2)::;t~m
IIQ(t)11 =
= km
1
am
(7.2.118)
so that
2
k > 2 + 12amEm.
(7.2.119)
531
p(Smk)
I!
Y;"k
and
Y;"k
(I:j
Q(S)dS) Ymk
k,
1::; j ::; r
(7.2.120)
(7.2.121)
k,
(7.2.122)
Let ao and bo be defined to m, <50 to be 0, and let 6.j
= 0,1,, r.
= 2:{=0 <5j ,
rjJm(t)
(7.2.123)
Pm(t)Ymk,
where
Pm(t)
Clearly,
1
2Em
and
l br
rjJ:n(s)Q(S)rjJm(s)ds
m-(1/2)f m
+~
j=O
+ 6.r
(7.2.125)
lbr
rjJ:n(s)Q(s)rjJm(s)ds
m-(1/2)f~
bj
i+
rjJ;"(s)Q(s)rjJ(s)ds
+~
j=l
bj
rjJ;"(s)Q(S)rjJm(s)ds.
aj
(7.2.126)
Chapter 7
532
By (7.2.120) and (7.2.121), it follows that
;"(s)Q(S)m(s)ds 2: -am
m-(1/2)f~
[s + (1/2)Em - m]2ds
m-(1/2)f~
-24 amEm
(7.2.127)
and by (7.2.120),
r ' ;"(s)Q(s)m(s)ds
Jb o
2Emk .
(7.2.128)
For 1::; j ::; r - 1, by (7.2.120) and (7.2.121) we also have
aH1
((1/2)Em - .0.j)
</>;"(s)Q(s)</>m(s)ds
l bj
p;,,(bj
aH1
Y;'kQ(s)Ymkds
l bj
l bj
= -Ion
O.
(7.2.129)
and
[aj, bj ],
p;"(S)Y;"kQ(S)ymkds
Y;"kQ(s)Ymkds
p~(t)
l bj
2Pm(s)
(is
Y;"kQ(U)Ymkdu) ds.
(7.2.130)
By (7.2.120) and (7.2.121), we further have
l bj
Y;"kQ(S)Ymk ds
I:
Y;"kQ(s)ymkds -
I:
y;'kQ(S)ymkds 2: 0,
(7.2.131)
(with equality if 1 ::; j ::; r - 1). Since t E [aj, bj ] implies t E Smk,
using (7.2.120) we find that for t E raj, bj ],
l bj
J
</>;"(s)Q(s)</>m(s)ds 2: 0,
= 1,, r.
(7.2.133)
533
m-/2)
[11>'(s)1 2 -1>;"(s)Q(S)1>rn(S)] ds s:
Ern
+ 2amE~ 4
2Emk < O.
+ G(X*(t))Q(t)H(X(t))
= 0
(7.3.1)
where
(i) G, HE F, and the set F will be defined later,
(ii) Q(t) is a real-valued, continuous, symmetric n x n matrix valued
function on [to,oo), to > 0,
(iii) X(t) is a twice continuously differentiable n x n matrix valued
function on [to,oo), and X* denotes the transpose of X.
A solution of the system (7.3.1) is a real-valued n x n matrix X(t)
satisfying (7.3.1) on [T,oo), T 2:: to (T may depend on X) and not
identically singular on a neighborhood of infinity. Any solution X(t)
of (7.3.1) satisfies (7.2.7), and if it satisfies (7.2.6), X(t) is said to be
prepared. X(t) oscillates if del X(t) has arbitrarily large zeros.
The continuous n x n matrix-valued function
functionally commutative on an interval J if
Q(s)Q(t)
Q(t)Q(s)
for all
Q(t)
s, t E J.
is said to be
(7.3.2)
We shall consider the general system (7.3.1) with symmetric, functionally commutative coefficients and show that the oscillation theory for such
systems can be effectively reduced to the study of a diagonal system of
scalar second order equations. In the case of equations like (7.3.1), this
system of scalar equations is uncoupled, allowing one to directly employ
the existing theory to obtain oscillation and nonoscillation criteria.
The following notation and preliminaries are required: Let Mn be the
Banach space of real n x n matrices with any of the usual norms. Let Vn
and N n be the subsets of diagonal and nonsingular matrices, respectively.
534
Chapter 7
In what follows, we shall consider the linear system (7.2.5) and the
general system (7.3.1), where G, HE F, and Q is continuous, symmetric
and functionally commutative. We shall denote the set of such Q by C.
The following lemma gives a characterization of functionally commutative matrices.
Lemma 7.3.1. Let the n x n matrix A(t) be diagonalizable, and
functionally commutative on an interval J. Then, A(t) has a constant set
of eigenvalues, and therefore there exists a constant matrix P such that
A(t)
p- 1 D(t)P
where
D(t) E V n ,
t E
J.
In the case of linear system (7.2.5), we can readily reduce our discussion
to that of an associated diagonal system.
Theorem 7.3.1. Let Q E C, and assume that there exists a constant matrix P so that Q(t) = P-lD(t)P, where D(t) = diag{ql(t),,qn(t)},
where qi(t) E C([to, CXl), lR), i = 1,2, ... , n. Then all prepared solutions
of (7.2.5) oscillate if and only if at least one of an associated system of n
scalar equations
yr(t) + qi(t)Yi(t)
0,
i = 1,2, ... , n
(7.3.3)
is oscillatory.
Proof. Since Q(t) is symmetric, the matrix P is actually orthogonal.
Let yet) = PX(t), then the system (7.2.5) is transformed into
Y"(t)
+ D(t)Y(t)
O.
(7.3.4)
535
Yo (to)
= 0,
Y~ (to)
= I,
(7.3.5)
In fact, in view of the Morse separation theorem [42] the oscillation of this
solution will guarantee oscillation of all solutions. Clearly, the solution of
(7.3.3) satisfying the initial condition (7.3.5) is diagonal for all t:::: to,
and so for this solution, system (7.3.4) may be written in the form (7.3.3),
where Yo(t) = diag{vdt),"',Vn(t)}, cmd D(t) = {lJl(t),,qn(t)}.
Now since det Yo(t) = TI~=l Yi(t), Yo(t) oscillates if and only if one of
the functions Yi(t) oscillates, i.e., if and ouly if one of the scalar equations
(7.3.3) is oscillatory. This completes the proof.
Now we shall study the general system (7.3.1). For this, first we note the
following simple consequence of a solution which is symmetric and prepared.
Theorem 7.3.2. Any symmetric and prepared solution X(t) of (7.3.1)
which is nonsingulaT on an interval J, commutes with Q(t) for all t E .J.
Proof. Since X(t) is symmetric and prepared, (7.3.1) shows that X(t)
commutes with X'(t) for all t:::: to. Differentiation of the identity
X'(t)X(t) - X(t)X'(t) = 0
leads to the fact that X(t) commutes with X"(t) for all t:::: to. If X(t)
is nonsingular on J, system (7.3.1) gives X(t)G(X(t))Q(t)H(X(t)) =
G(X(t))Q(t)H(X(t))X(t). Since H(X(t)) is nonsingular on J, we can
wri te this as
X(t)G(X(t))Q(t) = G(X(t))Q(t)H(X(t))X(t)H-l(X(t))
t E J.
(7.3.6)
Now, let X(t) = P-1(t)Y(t)P(t), where P(t) EN,., and X(t) E Dn.
Then, since Y(t) and H(Y(t)) E Dnl we have
H(X(t))X(t)
P-l(t)H(Y(t))Y(t)P(t)
for all
X(t)H(x(t)).
Similarly, X(t)G(X(t))
xX(t)Q(t)
536
Chapter 7
(I)
A commute pairwise.
i=1,2, .. ,n
0,
(7.3.7)
of system
G(X(tQ(t)
G(p-lY(t)p)p-1 D(t)P,
P-IG(Y(tpp-1 D(t)P
yet) E Vn
P-IG(Y(tD(t)P
where
=
p- D(t)G(Y(tP = Q(t)G(X(t,
l
(7.3.8)
since G(Y(t, D(t) E V n .
Therefore, equation (7.3.1) can be written as
X"(t)
+ Q(t)F(X(t
0,
t::;> to.
(7.3.9)
Thus, we have
pylI(t)p-1
+ PQ(t)F(P-IY(t)p)P- 1
which reduces to
ylI(t)
+ D(t)F(Y(t =
O.
0,
(7.3.10)
537
y"(t)
+ q(t)y(t)
O.
(7.3.12)
t-'>oo
Jtort Q( s )dS)
00.
so
Al (
(It:
t
Jt.rto Q(S)dS)
max
l:Si:Sn
Jtort qi(s)ds.
Itto
Now, limHoo Al
Q(s)ds) = 00 implies lim approx SUPHoo
qi(s)ds
for at least one index i. By Corollary 7.3.1 it follows that for such
indices i the equation y?( t) + qi (t )Yi (t) = 0 is oscillatory. The oscillation
of (7.2.5) now follows from Theorem 7.3.1.
= 00
538
Chapter 7
(1 Q(S)dS)
t
to
00
t--'>oo
1t
to
SAI(Q(s))ds
It:
x"(t) + q(t)f(x(t))
0,
where q(t) E C([to, 00), JR+) and f E C(JR, JR), xf(x) > 0,
for X =I O.
(II) Let
only if
(1 2 ) Let f(x)
(7.3.13)
f'(x) 2: 0
= xa/f3, where 0:, /3 are odd integers with 0 <, = 0://3 <
Joo siq(s)ds = 00.
OO du
f(u) <
00.
(7.3.14)
539
Proof.
limt->oo >'1
Itto sqi(s)ds =
00,
(it SQ(S)dS)
to
(I/o sQ(s)ds)
if and only if
= 00
I:7=1
00.
(7.3.15)
It o
Theorem 7.3.7. Let f(x) = xo./ f3 , where a, /3 are odd integers with
o < I = a//3 < 1. Let Q E C be as in Theorem 7.3.1 and Q(t) is
nonnegative definite for all t ~ to. Then all symmetric prepared solutions
of the system (7.3.1) oscillate if and only if
lim Al
t->oo
(it S'YQ(S)dS)
to
00.
Proof. The proof is the same as that of Theorem 7.3.6 except now we use
criterion (h).
X"(t)
+ (X*(t))k
Q(t)Xk+l(t) = 0,
(7.3.16)
where k is a positive integer, and Q and X satisfy (ii) and (iii) of the
system (7.3.1) gives the following corollary.
it
lim
s2k+lJ.L1(S)ds <
t->oo to
00.
Chapter 7
540
Then all symmetric prepared solutions of the system (7.3.1) are nonoscillatory.
X"(t)
and
+ Q(t)X(t)
X"(t) + QI(t)X(t)
(7.4.1)
(7.4.2)
541
It
(i)
+(B), t
It
[to,oo).
= X'(t)X-l(t)
[to,oo),
W(t) = W(T) +
iT
Q(s)ds +
iT
W2(s)ds,
to
~ t <T
is
an upward directed set in C(B) and is order bounded above by W(t). For
Chapter 7
542
follows
It
W(t)
+ P(T)
- if>(t, T) 2:: 0
for
to:S T :S T.
(7.4.5)
For each T 2:: T the left-hand side of (7.4.5) is in C(B), which is a normed
vector lattice and therefore has norm-closed positive cone .c+(B). Letting
T --* 00 in (7.4.5), we obtain W(t) - if>(t, T) 2:: 0,
which completes the
proof of the assertion. Since .cr(B) is order complete, it follows that
if>(t) = SUPT if>(t, T) exists. Let y E B+. Then, (7.4.3) implies that
W(T)y = W(t)y -
(iT
(7.4.6)
Since B has order continuous norm, if>(t, T)y --* if>(t)y in norm, so if>(t)
is the strong operator limit of if>(t, T) as T --* 00. Thus the limit in norm
as T --* 00 exists for the right-hand side of (7.4.6). We conclude that the
limit in norm of W(T)y as T --* 00 exists (in B+ by norm-closedness
of B+), we call it C(y). Thus,
(7.4.7)
W(t) = C + P(t)
+ if>(t),
C E .c+(B),
t E [to, 00).
(7.4.8)
Zo(t)
Zn+l(t)
C, t 2:: to
=
iT Z~(s)ds,
n = 0, 1,, t 2:: to
(7.4.9; n)
where s-lim denotes the strong operator limit. To justify this, inductively
we shall show
(7.4.10; n)
543
Clearly (7.4.10;0) is true. Now assume that (7.4.1O;n) holds. Then, 0:::;
Zn(s) :::; W(s) implies 0:::; Z;(s) :::; W 2(s). This is a consequence of
the cone + (B) being closed under multiplication. (It does not, in general, hold in the B*-algebra case). Thus,
{itT Z~(s)ds}
increasing in T
since
ItT
is monotone
1;1' Z~(s)ds :::;
Zn+l(t) - Zn+l(S)
Pl(t) - P1 (s)
-1t Z~(T)dT,
(7.4.11)
ItT
where W(t)
s-limn~= it Z~(T)dT
(7.4.13)
o :::;
lilt
Z~(u)dull
as
t-
--+ 0
I:
Z~(T)dT
Y E B+, so
544
Chapter 7
Example 7.4.1.
Let B be a Banach lattice with order continuous
norm and let Bl be its dual space. (Bl is also a Banach lattice, under
the dual norm, which is order complete). Let c E B+ and E Bt.
Let q(t) E C2([to,00),IR+) such that q'(t) > 0 and q"(t) < 0 for all
t E [to, 00). Define Q: [to, 00) ---+ .c(B) by
Q( t)y
q"(t)(y)c
1 + q(t)(c) ,
yE
Then,
Q(t) E .c+(B)
for all
t E [to, 00),
(jjQ(t)jj
jq"(t)jjjjjjjcjj )
1 + q(t)II(c)11 .
Note that
111
00
Q(s)dSII =
(1=
-1
+q~~:~(C)dS) 11llltcll ~
q'(t)IIllllcll
Wetly
q'(t)(y)c
1 + q(t)(c) '
YE B
o -<
00
q~(s)
ds <
1 + ql(S)l(Cl)
-
00
q"(s)
1 + q(s)(c)
ds
t E [to,oo)
Pl(t)
and to dispense with hypothesis (ii) (which holds automatically in the scalar
case for a nonoscillatory solution X(t) of (7.4.1) when P(t);::: 0).
As far as (ii) is concerned, we may make the following remarks:
545
Po(t) = P(t)
Pn(t) = P(t)
+ s-limT->~
n = 1,2,.
= P(t)
The equivalence of (ii) and (iii) can be established by using essentially the
same arguments as were used in the proof of Theorem 7.4.l.
When B is finite dimensional we can replace (i) by (i)'.
Theorem 7.4.2. Let B
P(t)
lim
T--+oo
iT
t
Q(s)ds,
(7.4.16)
exist such that (i)' holds. If there exists a nonoscillatory solution X(t)
of (7.4.1) such that (ii) holds, then the system (7.4.2) has a nonoscillatory
solution on [to, 00).
Proof. As in the proof of Theorem 7.4.1, we put W(t) = X'(t)X-l(t)
in (7.4.1) and obtain (7.4.3) which leads to (7.4.8). Let Z be the set of
continuous functions Z: [to, 00) --+ .c(lRn) such that IZ(t)1 ~ W(t) for all
t E [to, 00). Then, Z is closed, convex subset of M, the Frechet space of
continuous operator-valued functions on [to, 00) with the compact-open
topology.
Now if Z E Z, then for each t E [to,oo), Z(t) may be represented
by an n x n matrix, say, Z(t) = [(Z(t))ij]. Similarly, we may represent
W(t) by the n x n matrix W(t) = [(W(t))i'J, and we have I(Z(t))ijl ~
(W(t))ij, i,j = 1, .. , n. It follows that IZd(t)1 ~ W 2 (t). Furthermore,
we have IIZ2(t)11 ~ IIW 2(t)ll, where II II denotes the uniform operator
norm on .c(B). Since W 2 (t) is integrable on [to, 00), it follows that
Z2 (t) is integrable on [to, 00 ) , and
(7.4.17)
It follows from (7.4.8), (7.4.17) and (i)' that F on Z defined by
(7.4.18)
546
Chapter 7
= Z.
II(FZn)(t) - (FZ)(t)11
CXJ
CXJ
(7.4.19)
~
W2(s)dsll and
Z2(s) ds ll ~
W 2(s)dsll, we can for any given E > 0 choose T = T(E) so that
Z~(.s)dsll < c/3, n = 1,2,'"
and
Z2(s)dsll < E/3. By the
uniform convergence of Zn to Z on [to, (0), we have
Now since
III;
III;
III; Z~(s)dsil
III;
III;
III;
iT IIZ~(s)
- Z2(s)llds <
for all
n 2: N(E),
say
for
i.e.,
I: IIW2(
(7.4.20)
It follows that the system (7.4.2) has a nonoscillatory solution.
with an appropriate
(7.4.21)
fo
547
(7.4.22)
v(t) = v(T)
IT
q(y'Tis)ds + n
IT
v 2 (s)ds.
(7.4.23)
vet)
1
b + y'Tip(
vnt ) + n 1 v (s)ds.
t
00
where
(7.4.24)
Let Zo be the set of continuous functions Z(t) = [(Z(t))ij] from [to, (0)
to (JRn ) such that !(Z(t))ij! s; v(t), t E [to, (0) for each i, j. Define
the map Fo: Zo --+ Zo by
(7.4.25)
for t E [to, (0), i, j
= 1,2,n.
It follows from the hypothesis (iv) and (7.4.24) that Fo is well defined, Zo is closed, convex subset of M, the Frechet space of continuous
operator-valued functions on [to, (0) with the compact-open topology.
Arguments similar to those used in the proof of Theorem 7.4.2 show that
Fo is continuous on Zo and that Fo(Zo) is precompact subset of Zoo
Hence Tychonov's theorem yields the existence of a fixed point of F o, i.e.,
there exists Z(t) = [(Z(t))ij] such that
(Z(t))ij = (P1(t))ij
for t E [to, (0), i, j
1 (~(Z(S))ik(Z(S))kj)
00
ds
(7.4.26)
= 1,2,,n.
Example 7.4.2.
n = 2. Let
Q(t) = t21
[ac d
b1
'
548
Chapter 7
Po(t) = P(t) = -1
t
[ac
b
d
00,
where
1,
If
(7.4.27)
has a solution.
It can be shown that the solvability of (7.4.28) is equivalent to the
existence of positive solution m of the equation
(7.4.29)
and
Q,
(3, ,
[l_m_ d: a], (3 !
=~, 6= ~ [l-m+ d:a].
~
< 1/4
if d
+ a + J(d -
=a
a)2
and
+ 4be
be
=0
:::: 1/2
if d
f. a
or
bc
f. O.
(7.4.30)
549
t-'>oo
00
qll(s)ds < a,
C,
lim sup t 1
t-'>oo
lim sup t 1
t-'>oo
00
q12 (s )ds
00
<b
(7.4.31 )
(7.4.32)
XI/(t)
(7.5.1)
n ,
to 20
CJ
00.
(7.5.2)
to
550
Chapter 7
Proof. Let X(t) be a solution of (7.5.1) and set Y(t) = X(t) - V(t).
Then it follows from the system (7.5.1) that
yl/(t)
o.
(7.5.3)
Y(t) = K
(7.5.4)
has a solution on [t l , (0) for some tl > to. For this, it suffices show that
the operator
(TY)(t) = K
00
(t - s)Q(s, Y(s)
+ V(s))ds,
BE
--t
~ tl
matrix~valued
(7.5.5)
functions.
t-....}CX)
Clearly, B{ is a Banach space under the sup norm. From (7.5.2), there
exists a h > to such that for every t ~ tl,
IIKII + 100J (s -
(7.5.6)
00
00
(7.5.7)
00.
then
(s - t)q(s, A + IjV(s)ll)ds.
(7.5.8)
Since the integral on the right~hand side of (7.5.8) tends to zero independently of Y, it follows that the set T( S) is equiconvergent. Consequently,
T(S) is relatively compact in Bt,. Thus, by Tychonov's theorem, the operator T has a fixed point in the ball S. This fixed point Y (t) satisfies
equation (7.5.1). The conclusion of the theorem now follows by letting
X(t) = Y(t) + V(t).
551
Next, we employ Theorem 7.5.1 to obtain an infinity of oscillatory solutions of the system (7.5.1).
Theorem 7.5.2. Let conditions (i) and (ii) hold, and assume that for
some A > 0 condition (7.5.2) is satisfied, and
+ V(t)]
<)..
and
> 5,
+ V(t)] < -
(7.5.9)
5.
(7.5.10)
(7.5.11)
and
(7.5.12)
for every] ::::> ]0. Consequently, from (7.5.12) we deduce that det(X(tj)) >
5/2 for all ]::::> ]0. Similarly, we can show that det(X(si)) < -5/2 for
is a sequence in [t1' (0) with
all large i, where {Si}, i = 1,2""
limi->oo Si = 00. Obviously, det(X(t)) has an unbounded set of zeros on
[t1' (0) and this completes the proof.
Example 7.5.1.
Q(t, X)
(t
+ 1)311XII
and
F(t) =
sint
1/(t + 1)
1/(t+l) ]"
cost
It is easy to see that the hypotheses of Theorem 7.5.2 are satisfied, and
hence (7.5.1) has at least one oscillatory solution.
Chapter 7
552
comparison-type theorems continue to hold and oscillation criteria can easily be obtained via the traditional methods (see Chapter 3).
(W(t))
(W(to)) + (Q1(t)) +
it
(W2(s))ds.
(7.6.1)
to
It has been shown by Akiyama [3], that any continuous positive linear
functional on the cone of nonnegative real symmetric matrices is equivalent
553
with the trace functional. Hence, there exists a positive constant c, which
depends only on , such that (W2(t));c c tr(W 2(t)). Thus, instead of
(7.2.13) we have the inequality
(W2(t)) ;c c(l-tlAi(Ql(t))
forall
t;cT.
(7.6.2)
t (W2(s))ds
lto
= f
lto
(7.6.3)
Here we use (7.6.2) to establish the last integral, whereas in the first integral
we use the elementary inequality (W2(t));c ((J))-l[(W(t))j2, where
J is the n x n identity matrix. Thus, we obtain instea.d of (7.2.17),
(W(t))
+ ~I)
1:
Ai(Ql(S))ds
(7.6.4)
[(W(s)Wds.
;c c(l for t;c Tl ;c T. Thus if we let 8 = c(l define w(t) = (W(t)), t;c to then
q(t) = (Ql(t))
+8
lto
f)4,
f)4
t Ai(Ql(S))ds (7.6.5)
lto
replace
by (I)E, and
Ai(Ql(S))ds.
It:
6. Theorem 7.2.3 can be generalized by considering the principal subQ(s)ds. We recall (cf. 6], p. 113): For any n x n symmatrices of
metric matrix A, the sequence of symmetric matrices Ak = (aij), i, j =
1,2, ... ,k for k = 1,2, ... ,n satisfies A]+l (Ak+1) ::; Aj (Ak) ::; Aj (A k + 1),
where Aj(Ak) denotes the j th characteristic roots of A k .
(itto QdS)dS)
> -
00,
(7.6.6; k)
Chapter 7
554
00,
and
lim)'1
t--+oo
(fiQds)ds)
. to
00.
(7.6.7; k)
(7.6.8; k)
7.7. References
1. S. Ahmad and A.C. Lazer, A n-dimensional extension of the Sturm
separation and comparison theory to a class of nonsclf-adjoint systems,
SIAM J. Math. Anal. 9(1978), 1137-1150.
2. S. Ahmad and A.C. Lazer, On an extension of Sturm's comparison theorem to a class of nonself-adjoint second order systems, Nonlinear Analysis
4(1980), 497-501.
3. K. Akiyama, On the maximum eigenvalue conjecture for the oscillation
of second order differential systems, M.Sc. Thesis, University of Ottawa,
1983.
555
556
Chapter 7
557
Chapter 8
Asymptotic Behavior
of Solutions of Certain
Differential Equations
8.0. Introduction
The study of behavioral properties of solutions of differential equations
near infinity is of immense importance and hence it continues to attract
many researchers. Therefore, in this chapter we shall present some recent
contributions on the asymptotic behavior of solutions of second order differential equations as well as the behavioral properties of positive solutions
of singular Emden-Fowler-type equations. In Section 8.1 it is shown that
for a large class of differentia'! equations, not only can the existence of
nonoscillatory solutions be proved, but also an explicit asymptotic form of
the nonoscillatory solutions may be provided. Then, we shall impose more
restrictions on the sign of the integrable coefficient of the equation, and get
necessary and sufficient conditions so that the solutions have the specified
asymptotic behavior as t -+ 00, i.e., solutions which behave asymptotically like a nonzero constant and a.!so those which behave asymptotically
like et, e i- O. For this, various averaging techniques of the types employed in the previous chapters to study the oscillatory behavior of such
equations have been used. Section 8.2 is devoted to the study of existence, uniqueness and asymptotic behavior of positive solutions of singular
Emden-Fowler-type equations. The cases when the coefficient of the equation under consideration is of constant sign, or of an alternating sign are
systematically discussed. Then the existence as well as nonexistence results
for the positive solutions of Emden-Fowler-type systems are proved.
559
x"(t)
+ q(t)f(x(t)) =
(8.1.1)
0,
(8.1.2)
x(t) = c+o(l)
x(t) = o(t)
or, the condition
and
as
lim x(t) =
x(t) = ct + o(t)
as
t--+oo,
00
as
t --+
(8.1.3)
t --+
00,
00,
(8.1.4)
(8.1.5)
t-400
it
to
(8.1.6)
Q(t)
00
q(s)ds,
t:2: to.
(8.1.7)
If q( t) :2: 0 for t :2: to, then it is easy to show that a nonoscillatory solution
x(t) of (8.1.1) satisfies exactly one of the three asymptotic conditions (8.1.3)
- (8.1.5). In the following result, we shall prove that this fact remains valid
even when Q(t):2: 0 for t:2: to
Theorem 8.1.1.
Let
Q(t) :2: 0 for t:2: to. Then for each nonoscillatory solution x(t) of
Chapter 8
560
equation (8.1.2), exactly one of the three asymptotic conditions (8.1.3) (8.1.5) is satisfied.
Let x(t) be a nonoscillatory solution of equation (8.1.2), say,
x(t) > 0 for t 2: to > O. It is known (cf. Lemma 4.1.3) that x(t) satisfies
Proof.
the equality
x'(t)
= Q(t)x'Y(t) +ryx'Y(t)
00
x-'Y-1(s)(x'(s)fds
for
2: to,
x'(t) 2: Q(t)x'Y(t)
t 2: to.
for
(8.1.8)
x'(u)-Q(u)x'Y(u)+ry
Q(s)x'Y-l(s)x'(s)ds
x'(t)-Q(t)x'Y(t), (8.1.9)
/00 Q(s)x'Y-l(s)x'(s)ds
<
00.
(8.1.10)
x'(t)
0:
0: =
+ Q(t)x'Y(t) + ryk1(t)
0:
for
2: to.
(8.1.12)
/00 Q2(s)x2'Y-l(S)ds
<
00.
(8.1.13)
561
x(t) = x(to)
+ a(t -
to)
ito
t Q(s)x'(s)ds <
ito
(1:
X(S)dS) 1/2
the first
(8.1.16)
t;:::: to.
for
Thus, for
x(t)
(1:
t;:::: to,
(8.1.15)
:s
x(to)
+ a(t -
+ [k2(tO)(t -
to)
to)X(t)J1/2
+ "(kI (to)(t -
X1/ 2(t)
:s ~
+ D1/2(t)]
for
to).
X 1/ 2
(t).
t;:::: to,
x(t) :S mt
for
t;:::: to.
(8.1.17)
o :S -lit Q(s)x'(s)ds
t
J;
o :S
to
[mk2(T)p/2.
lit Q(s)x'(s)ds
lim -
t--400
to
t;:::: T.
(8.1.18)
lit Q(s)x'(s)ds:S
00
for
00,
O.
(8.1.19)
letting T ---+
00
in
(8.1.20)
Chapter 8
562
In view of (8.1.15), (8.1.20) and the fact that kl(t)
find limt-HXJ x(t)/t = Q:.
0 as t
~ 00,
we
Recall that x(t) is nondecreasing for t:2: to. Now, there are three
cases to consider:
(i)
(ii)
(iii)
Q:
Q:
Q:
Case (i) implies (8.1.3) with c = limHoo x(t) > 0, while case (iii) implies
(8.1.5) with c = Q: > O. It is also clear that case (ii) implies (8.1.4). This
completes the proof.
/00 IQ(s)lds
<
00
(8.1.21)
and
(8.1.22)
Then for any constant c
that
x(t)
#- 0,
c+ 0
(1
00
[IQ(s)1
and
x'(t)
as t ~
00,
O(IQ(t)1
+ pes)] dS)
+ pet)),
(8.1.23)
(8.1.24)
Proof. Condition (8.1.22) implies that pet) is nonincreasing and integrable on [to, 00). We may assume that c> O. Let
mm'
+ bm'
loo
IQ(s)lds < b
(8.1.25)
Let F
{OO
iT
IQ(s)lds
(OO
+ b iT
P(s)ds ~
2".
563
(8.1.26)
{x
F : Ix(t) - cl
~~,
Ix'(t)1
~ mQ(t) + bP(t)
(<I>x)(t) = c
-1
00
Q(s)f(x(sds
-1 (1
00
00
we
Q(u)f'(x(u))X'(U)dU) ds.
(8.1.27)
by employing the Schauder-
(It) <I> is well-defined on X and maps X into itself. For this, let x E X.
For r 2': s 2': T, we have
11T Q(u)f'(x(ux'(U)dU\
<
<
I(cpx)'(t) I <
IQ(t)f(x(t1
\100
Q(U)f'(X(Ux'(U)dU\
~ ooIQ(S)f(x(sldS+ 00
1100
Q(u)f'(x(ux'(u)dul ds
~ ~.
in X
564
Chapter 8
1(<I>xk)'(t) - (<I>x)'(t) I
-:;
IQ(t)llf(Xk(t)) - f(x(t))1
1 IQ(u)IIf'(Xk(U))X~(U)
00
- f'(x(u))x'(u)ldu
1(<I>Xk)(t) - (<I>X)(t) I
-:;
1I
00
(<I>xd(s) - (<I>x)'(s)lds
x"(t)
sgn x(t)
0,
(8.1.28)
where k, A and 'Y > 0 are constants. Applying Theorem 8.1.2 to the
case f(x) = Ixl'"Y sgn x, IQ(t)1 = too ks Asinsdsl -:; 21klt A (). < 0), we
find that if A < -1, then for any nonzero constant c, equation (8.1.28)
has a solution x(t) such that x(t) = c + O(tA+l) and x'(t) = O(tA) as
t -4 00. Note that (8.1.28) has a nonoscillatory solution if and only if
IJ
565
f
or,
1 (
[415])
see,
,
lim tQ(t)
t--too
(8.1.29)
are satisfied. Then equation (8.1.1) has a nonoscillatory solution x(t) such
that for any constant c =I 0,
x(t)
c+o(l)
and
x'(t)
o(C I )
as
t---+oo.
(8.1.30)
Proof.
It suffices to note that conditions (8.1.21) and (8.1.29) imply
OO
(8.1.22) and t OO Q2(s)ds
sQ2(s)ds ---+ 0 as t ---+ 00.
ft
:s: ft
(i)
for any constant c =I 0, there exists a solution x(t) of equation
(8.1.1) satisfying (8.1.3),
(ii) for some constant c =I 0, there exists a solution x(t) of equation
(8.1.1) satisfying (8.1.3),
(iii) the integral conditions (8.1.21) and (8.1.22) are satisfied.
Proof. (i) implies (ii) trivially, and (iii) implies (i) by Corollary 8.1.1. We
claim that (ii) implies (iii). Let x(t) be a solution of equation (8.1.1) for
which (8.1.3) holds for some constant c =I O. We may assume that c> O.
There is a number T such that c/2:S: x(t) :s: 2c for t ~ T. Recall
Lemma 4.1.3, if we remove condition (4.1.49) and let a(t) = 1, then the
equality (4.1.52) takes the form
x'(t)
f(x(t))
= a + Q(t) +
00
x'(s)
f(x(s))
)2 f ,(x(s))ds,
(8.1.31 )
Chapter 8
566
where a is a nonnegative constant. Since Q(t):::: 0,
x'(t)
f(x(t
>
- Q(t)
00
(8.1.31) gives
for
t:::: T,
where m ] = min {I' (x) : c/2 ::; x ::; 2c} > O. An integration of the above
inequality over [T, t] gives
X(t)
. x(T)
It [
- du > . T
f(u) -
Q(s)
00
m]
Q 2 (u)du J ds
for
t:::: T.
Since the left-hand side of t.he above inequality remaius bouuded as t--+
00,
we conclude that (8.1.21) and (8.1.22) are satisfied. This completes
the proof.
For the superlinear case, i.e., equation (8.1.2) with "f > 1,
result is now clear.
Theorem 8.1.4.
t::::
the next
(i)
(ii)
(i)
for any constant c
(8.1.1) satisfying (8.1.30),
0,
0,
x' (t)
f(x(t
= Q(t) +
00
x' (s ) )
f(x(s
f (x(sds,
tx'(t)
tQ(t) = f(x(t - t
00
x'es)
f(x(s
)2 f ,(x(sds
567
(8.1.32)
.1
t~! t
00
x' (s ) )
f(x(s
. ft ft C(:~:~)
OO
f (x(sds
hm
t-too
f'(x(sds
2
----'-----'-----':-'-~----
.!l..
dt
(1)
t
t~c!, j2(x(t tx t
O.
+ tQ(t)f'(x(tx'(t).
x(t) - tx'(t)
0'1 -
tQ(t)f(x(t
Q(s)[f(x(s + sf'(x(sx'(s)]ds,
it
(8.1.33)
Remark 8.1.2. From Theorems 8.1.3 and 8.1.5, we note that even for
solutions which have the same limit as t -+ 00, there is an essential
difference between restricting and not restricting the asymptotic behavior
of their derivatives.
Chapter 8
568
t1
jt If(ks)IIQ(s)lds -+ 0 as t -+
jt f'(ks)IQ(s)lds <
and
IX) If(klS)If'(k2S)Q2(s)ds
00,
(8.1.37)
00,
<
(8.1.36)
(8.1.38)
00.
Then for any nonzero constant c equation (8.1.1) has a solution x(t)
such that
x(t)
and
as t -+
+ Pc(S)]dS)
+ Pc(t)) ,
(8.1.39)
(8.1.40)
where
_ 3 /2
d
c , an
Cl -
C2 -
X = {x
F : Ix(t) - ctl
S;
(c/2)t, Ix'(t)1
S; c +
f(Clt)Q(t)
(fllx)(t) = ct + [
Q(s)f(x(s))ds + [
t:::: T
569
As in the proof of Theorem 8.1.2 it can be shown that (11) cI> is welldefined on X and maps X into itself, (h) cI> is continuous on X, and
(h) cI>X is compact. By the Schauder-Tychonov fixed point theorem the
operator cI> has a fixed point x E X. This fixed point x = x(t) is a
f(kt)Q(t) -+ 0 as t -+
and
J=
f'(ks)IQ(s)lds <
(8.1.41)
00
(8.1.42)
00.
Then for any nonzero constant e equation (8.1.1) has a solution x(t)
such that
x'(t) = e + 0(1)
as
t -+
00.
(8.1.43)
Next, we shall prove the converse of Corollaries 8.1.3 and 8.1.4 when
Q( t) is of constant sign.
Theorem 8.1. 7 . Suppose Q (t) :2: 0 for all large t. Moreover, suppose
either condition (8.1.34) or (8.1.35) holds, f'{x) > 0 for x =I 0 and for
all nonzero constants kl' k2'
(8.1.44)
Chapter 8
570
Then equation (8.1.1) has a solution
nonzero constant c provided
lim
~ jt If(ks)IQ(s)ds
t-+oo t
JOO f'(ks)Q(s)ds
JOO If(k
<
00
s)IJ'(k2s)Q2(S)ds <
x(t)
00
(8.1.45)
(8.1.46)
x'(t)
00 (
x'(s)
f(x(s
)2 f'(x(sds,
(8.1.48)
where a is a nonnegative constant. On the other hand, an integration by
parts of (8.1.1) from T to t gives
(8.1.49)
af(x(t)+ f(x(t
1 (f~:~~)
00
f'(x(sds
f3- ltQ(s)f'(x(Sx'(S)dS.
(8.1.50)
It Q(s)f'(x(sx'(s)ds <
00.
(8.1.51)
f;
lit
+~ It
x(t) - X(T)
571
Q(s)f(x(s))ds + -O!
t
(l
f(x(s))
CXJ
It
f(x(s))ds
(f~~~~))
f'(X(U))dU) ds.
(8.1.53)
Since x(t) satisfies (8.1.5), from (8.1.52) and (8.1.53) it follows that
lim -1
t--'tCXJ
It
Q(s)f(x(s))ds = c -
(8.1.54)
f).
lit
o ::; t
Q(s)f(x(s))ds
: ; ~ 1J,(~:)
(
ds
) 1/2 (
1
t
f(x(s))f'(x(s))Q2(s)ds
) 1/2
(8.1.55)
h
- - { c/2 if condition (8.1.34) holds
Note that
were c 2c if cOl1dition (8.1.35) holds.
(8.1.44) implies that there exists a positive constant m independent of T
such that
1
f(2cs)
t2
f'(cs) ds ::; m 2 for t ~ T.
for
T,
t
iT
o ::;
C -
f) ::;
00
m (
(8.1.56)
lim -
t--'tCXJ
jt Q(s)f(x(s))ds
O.
f).
(8.1.57)
(8.1.58)
572
Chapter 8
For equation (8.1.2), Corollary 8.1.3 and Theorem 8.1.7 yield the following result.
Theorem 8.1.8. Suppose Q(t)
statements are equivalent:
(i)
for any c
satisfying (8.1.5),
i-
i-
of equation (8.1.2)
/>0 s"i-lQ(s)ds
<
00
(8.1.59)
and
(8.1.60)
Now we are ready to prove the following result.
Theorem 8.1.9. Let 0 < , < 1. Suppose condition (8.1.6) holds and that
Q(t) ~ 0 for t ~ to. Then the following three statements are equivalent:
(i)
(ii)
Joo
Suppose that equation (8.1.2) has a solution x(t), say, x(t) > 0 for
to. Then the equalities and inequalities in the proof of Theorem 8.1.1
remain valid. Now, we consider the following two possibilities:
(h) 2,-1 ~ O. Then the desired condition (8.1.60) follows from (8.1.13)
and (8.1.17).
(12) 2, - 1 > O. In (8.1.15), the first three terms of the right-hand side
are nonnegative and kl (t) is nonincreasing on [to, 00). Thus, we have
x(t) ~ ,(t - to)k1(t) for t ~ to. Hence, in view of (8.1.8), (8.1.11) and
(8.1.14), we get x(t) ~ ,(t-tO)k2(t) for t ~ to, and so by the assumption
573
2,-1> 0,
Q2(t)x21'-1(t)k~-21'(t)
- 2! 2, k~-21' (t)
+ 2!
for
,21'-1
1:
(s - to)21'-lQ2(S)ds.
Note that the left-hand side of the above inequality is bounded on [to, 00).
This implies that the desired condition (8.1.60) is satisfied. This completes
the proof.
lim f(kt)Q(t)
t-'>oo
(8.1.62)
and
/00 f'(ks)IQ(s)lds
<
00
(8.1.63)
1 (f~:~~))
00
f'(x(s))ds
for
t;::: T.
(8.1.64)
Using l'Hospital's rule we find that the second term of the right-hand side
of (8.1.64) tends to c as t ---+ 00, i.e.,
d
( dt
itroo (
x' (s ) )
f(x(s))
f (x(s))ds
d
1 )
dt f(x(t))
-1
x (t) ---+
as t ---+
00.
Chapter 8
574
(i)
for any constant c =I=- 0, there exists a solution x(t) of equation
(8.1.2) satisfying (8.1.43),
(ii) for some constant c =I=- 0, there exists a solution x(t) of equation
(8.1.2) satisfying (8.1.43),
(iii) the following integral conditions are satisfied
lim {YQ(t) =
t-+oo
/00
and
s'Y-1IQ(s)lds <
00.
It is to be noted that the condition (iii) of Theorem 8.1.11 can be replaced by limt-+oo s'Y q( s )ds exists and is finite.
litiS
to to
q(u)duds > -
00.
(8.1.65)
it is
to
to
q(u)duds
(8.1.66)
Q2(t)
lim
T-+oo
lITIS
-T
q(u)duds,
to.
(8.1.67)
The role of Q(t) in the above results is thus taken by the function Q2(t).
In fact, it can be shown that if Q(t) is replaced by Q2(t) in the results
of Subsections 8.1.1 - 8.1.3, then the corresponding results remain valid.
575
x(t) = c+o(l),
x'(t)=o(C 1 )
as
t---+oo,
(8.1.68)
x'(t)
x'(T)
f(x(t)) - f(x(T))
rt (f(x(s))
x'es) )2 ,
rt
f (x(s))ds+ iT q(s)ds = 0 (8.1.69)
+ iT
and
f(x(t))
lOO (f~:~~)))
J'(x(s))ds <
(8.1.70)
00.
(8.1.71)
x'(t)
f(x(t))
-1
= oCt )
and
x'(t) )
( f(x(t))
f (x(t))
iT
f'(x(s))x'(s) ds
J2(x(s))
f(x(T)) - f(x(t))'
(8.1.72)
Since the right-hand side of (8.1.72) has a finite limit as t ---+ 00, so does
the left-hand side. Therefore, in view of the fact that x'(t) ---+ c =f. 0
576
Chapter 8
as t ---+
proof.
00,
Remark 8.1.3. Theorem 8.1.12 implies that when condition (8.1.6) does
not hold, equation (8.1.1) does not have any asymptotically linear solution x of the form (8.1.43) or (8.1.68). Therefore, the assumption of the
condition (8.1.6) in Subsections 8.1.1 - 8.1.3 is justified.
We now return to the problem of existence of solutions x satisfying
the asymptotic conditions (8.1.3) and (8.1.5). Note that the asymptotic
behavior of the derivatives of solutions x is not restricted in (8.1.3) and
(8.1.5).
is a constant.
r(t) f(u)
du
-
ix(T)
x'(T)
f(x(T))
(t--t T)
rt r (f(X(T))
X'(T) )2 ,
- +t
f (x(T))dTds
1
litis
+t
iT iT
q(T)dTds =
for
t;:::: T. (8.1.74)
The first term of the left-hand side of (8.1.74) has a finite limit as t ---+ 00.
In fact, for the case when x(t) satisfies (8.1.3) with c i- 0, it is clear
that this term tends to zero as t ---+ 00. For the case when x(t) satisfies
(8.1.5) with c i- 0, we assume that c>
(the same will be done for the
Icase c < 0), then it follows that
1r(t) [ 1 1]
t~~ t
ix(T)
f(u) - f(oo) du
00
iT
f(x(r))
(8.l.75)
f (x(r))dr,
x'(T)
a- f(x(T))
577
00
in (8.1.74),
{=( f(x(r))
x'(r) )2
t t
f (x(r))dr+h~~~p t iT iT q(r)drds =
+ iT
O.
From this equality and condition (8.1.65) it follows that the integral (8.1.75)
is finite. Then in the limit as t -+ 00 in (8.1.74), we find that (8.1.66)
holds, and moreover, we have
0: -
x'(T)
f(x(T))
{= ( x'(r)
+ iT
f(x(r))
)2 f (x(r))dr + Q2(T)
I
0,
(8.l.76)
lim
t--+oo
11
-tn -
it
to
(t - s)"-1q(s)ds
(8.l.77)
Qn(t) =
iT
1
lim --=1
T-+cxo
Tn
(T - s)n-1q(s)ds,
(8.1.78)
n~1
it
to
(t - s)n-1q(s)ds > --
00,
(8.1.79)
Chapter 8
578
x'(t)
f(x(t))
roo (
+ Qn(t) + it
x'(s)
f(x(s))
)2 ,
f (x(s))ds
(8.1.80)
From equation (8.1.1), for every T, t with T 2': t 2': To, we have
itr (T T
s)
n-l
1T
1
- (n - l)(n - 2) Tn-l
1
- Tn-l
1T(
T-s
q(s)ds
)n-l (
( t )n-l ( x'(t) )
1- T
f(x(t))
(T - s)n-3
X'(S)
f(x(s))
[1
X'(T) dT ] ds
f(X(T))
(8.1.81)
T~ Tn-l
itrT(T - s )n-3 [1
X'(T) d ] d
f(X(T)) T S
(8.1.82)
T .....
oo
1T
t
(T - s)n-3
[l
(S)
x(t)
-du
f () ds.
U
(8.1.83)
579
forevery
r2:t 1 .
(8.1.84)
1 1 1]
X
(S) [
x(t)
f(u) - f(oo) du
< 11 (t') [
x(t)
<
f(u) - f(oo) du
<
(c - C)tlJ.
T:-I
< T:-l iT (T -
<
and consequently,
(n! (1 _~ )
2)T
n-2 ,
Chapter 8
580
Since
is arbitrary, we find
)~!,
(T - s)n-3
{lX(S) [f(u)
1x(t)
T
T:_Ii (T-S)n-3
I]}
f(oo) du
ds
o.
(8.1.85)
[rx{s) dU] ds
.Jx(t)
+ T:-I
iT (T - s)n-3x (s)ds
t,
1
1 (
t ) n-2
x(t)
- (n - 2) T 1 - T
<
;2 rtl
~
T
X(S)dS+(C+f)T:_Ii (T-s)n-3 sds
t,
__1_~
(1- .!.)n-2 x(t)
2)
itl x(s)dS+(C+f) [(n-2)Ttl ( tlT )n-2
T2
(n -
I-
(tl)n-l]
1
1 (
t )n-2
1- T
- (n _ 2) T 1 - T
x(t)
- (n - 2)(n - 1)
C+f
-+ (n-2 )( n-1 ) as T -+
and
TLI iT (T - st- 3
[l
(S)
00
dU] ds
x(t)
> -1
T2
l
1 iT
1- t
~) n-3 i t x(s)ds + (c - f ) (T - s)n-3 sds
T
t
Tn-I t,
_ _1_~
(n - 2) T
-..!...
T2
(1 _.!.)n-2 x(t)
+ (n - 2)(n - 1)
T
tl
(tl)n-l]
T
1-
C-f
-+ (n-2 )( n-l )
x(s)ds + (c _ f)
as
T -+
00.
[_I_tl
(1- tlT )n-2
(n - 2) T
1
1 (
(n _ 2) T
t)n-2
1- T
x(t)
T--+oo
581
= ( _ 2)( _ )"
n
(8.1.86)
From (8.1.85) and (8.1.86) it follows that the limit (8.1.83) is equal to the
real number
J =
(n - 2)(n - 1)
(1)
f(oo)
lTD
T--+oo
x'(To)
.
1
{T(
)n-1 ( x'(s)
f(x(To)) - a - )~~ Tn-1 lTD T - s
f(x(s))
)2 f '( x ( )) ds,
s
{T(
)n-l ( x'(s)
f(x(s))
)2 f '(x (s ))ds
<
00.
T~m= Tn-1
[T(T
t
00
)2 f'( x (s ))ds
x'(s) )2 ,
f(x(s))
f (x(s))ds <
)n-1 ( x'(s)
f(x(s))
S
t
00.
T--+oo
1
n
-1
[T (T - s)n-1q(s)ds
[= (f(x(s))
x'(s) )2 ,
f (x(s))ds
t
x'(t)
f(x(t)) - a -
for
t;::: To.
So, the solution x(t) satisfies (8.1.80). This completes the proof.
582
Chapter 8
lim
-1
tn -
t-->oo
it
to
(t - s)n- 2Qn(s)ds = 0,
(8.1.87)
lim
t-->oo
--1
tn -
it
to
(t - s)n-2h(s)ds
Proof. (i)
1
Tn-l
Jtr
iT
iT
to
- (n - 1)
However, since for T
IT:-11:(T- s
T:-1
1:
lim TL1
T-->oo
lim
T-->oo
1
n-1
1T
t
(T - s)n-1q(s)ds -
T
~
t- 2
:-1 it
to
[1:
(T - s)n-2
it follows that
to, we have
it
to
(T - s)n-2
Tt)n-l1tto q(r)dr
[1
q(T)dr] ds.
to
to,
q(r)dr] dsl
[1:
(T - s)n-2
(T_s)n-lq(s)ds
i.e.,
(T - s)n-1q(s)ds
1-
Iq(r)ldr] ds <
[1
to
Qn(tO)
.!.It [1
T
q(r)dr] ds
to,
Qn(t)
it
1
(T - s)n-1q(s)ds - Tn-1
Tn-1
(8.1.88)
= h(t).
(T - s)n-1q(s)ds
1
Tn-1
0,
1
lim Tn-1
T-->oo
-it
to
to
for
iT
to
q(r)dr
to
Iq(r)ldr] ds
t ~ to.
(T-s)n-lq(s)ds-
for
~ to.
it
to
q(r)dr,
(8.1.89)
583
t n1- 1
1t
(t - st- 2 Qn(s)ds
l:(t -
to
tn~l
st- 2 [Qn(to) -
L
S
q(T)dT] ds
1
(
t )n-1
1
(n _ 1) 1 - :
Qn(tO) - tn-1
t)
rt (t -
st- 2
.flo
1
-1- ( 1-..!!. n-1 Q (to) - -1- (n - 1)
t
n
(n - 1) t n - 1
11
to
[1 q(T)dT
S
to
ds
(t - st-1q(s)ds
From h'(t)
-1t
q(s)ds
for
t:::: to.
(8.l.90)
to
t n 1- 1
t::::
(t - s)n-2h(s)ds
to
1
(
t
- - 1-..!!.
(n - 1)
t
- -(-1
- ) - -11
n - 1 tn -
)n-I h(to)
1t (
t - s )n-l q (s )ds.
to
q(t) = {
otherWIse,
for
t E In,
n = 1,2,
(8.1.91)
Chapter 8
584
Here q(t) is regarded as a function on the interval lRo
h(t)
+ cos(nA(t ~ 2n1T
t E In, n
for
1,2,
(8.1.92)
otherwise.
lit
8
n
8
n
o s: -t o
h(s)ds n
s: (2 ~ n -A)"'"'
1T
Ii ds = 2n - n- A "'"'iA .
Thus, limHoo(1/t) J~ h(s)ds = 0 and so, by Lemma 8.1.1, (8.1.88) with
n = 2 is satisfied, and Q2(t) = h(t). Noting that Q2(t) (= h(t
satisfies Q2(t) s: 2 for t E In, n = 1,2, and Q2(t) 2: 1 for
t E I n , n = 1,2, ... , where
we can easily show that integral conditions (8.1.21), (8.1.22), (8.1.59) and
(8.1.60) with Q2(t) inplace of Q(t), are satisfied if and only if ). >
1, ). > 2, ). >, and ). > 2" respectively. Therefore, we can conclude by
Theorem 8.1.2 with Q2(t) inplace of Q(t) and condition (8.1.66) instead
of (8.1.6) that equation (8.1.2) with (8.1.92) has a solution x(t) of the
form (8.1.3) (c =1= 0) if and only if ). > 2. Similarly, we can conclude by
Theorem 8.1.8 with Q2(t) inplace of Q(t) and condition (8.1.66) instead
of (8.1.6) that equation (8.1.2) with (8.1.91) has a solution x(t) of the
form (8.1.5) (c =1= 0) if and only if ). > 2,.
x"(t)
+ q(t)x-A(t)
x"(t)
= q(t)x-A(t)
0,
(8.2.1)
(8.2.2)
585
xl/(t)
p(t)y-A(t)
q(t)x-J.L(t) ,
y"(t)
(8.2.3)
In what follows we are concerned with the proper solutions of equation (8.2.1) and discuss existence, uniqueness and asymptotic behavior of
solutions satisfying (II) or (1 2 ).
xl/(t)
+ q(t)x-A(t)
Theorem 8.2.1. If
00
0,
s-Aq(s)ds =
x(c)
00,
= a, x'(c)
/3.
to;:=: 1
(8.2.4)
to;:=: 1
(8.2.5)
to
00
s-Aq(s)ds <
00,
to
/3 < /30,
/30
such that
586
Chapter 8
x(t) = x(a)
and
yet)
yea)
-it
-it
+ (t - a)x'(a)
+ (t - a)y'(a)
(t - s)q(s)x-A(s)ds
(t - s)ql(s)y-A(s)ds.
x(t) - yet)
[x(a) - yea)]
+
it
+ [x'(a) - y'(a)](t - a)
(8.2.6)
Since x'(a) > y'(a), there exists an t > 0 such that x(t) > yet) for
a < t :::; a + t. Suppose x(t) :::; yet) for some t E (a, b]. Then there exists
e E (a, b] such that x(c) = y(e) and x(t) > yet) for a < t < c. Letting
t = e in (8.2.6), then the left-hand side is zero and the right-hand side is
positive, which is a contradiction. Hence, x(t) > yet) for a < t :::; b, and
since
x'(t) = x'(a)
-it
we have x'(t)
-it
ql(s)y-A(s)ds,
587
o.
-z"(t) = q(t)[x(c)
+ x'(e)(t - c)]->-,
z(e) = :1:(c),
z'(e) = x'(e).
Since 0 <S x(t) <S :r;(c)+x'(e)(t-e) for t:::: e, we have -z"(t) <S -y"(t)
for t:::: c. Hence, z'(t) - :J;'(t) = f:[z"(,5) - ;r;"(s)]ds > 0 and since
:J:'(t) :::: 0 for t:::: c, we have z'(t):::: 0 for t:::: c. But,
Z'(t)
and therefore,
x'(e)
00
-it
q(s)[x(e)
q(s)[x(c)
+ x'(e)(s -
+ x'(c)(s -
ell-Ads <
ell-Ads
00.
Hence, flOG s-Aq(s)ds < 00. Conversely, suppose that condition (8.2.5)
holds and a> O. Choose (3 > 1 + feoo q(s)[a + s - cl-Ads. Let x(t) be
the solution of equation (8.2.7) with x'(e) = (3. If x'(t) > 1 for all t :::: e,
then clearly x(t) is positive and limHoo x'(t) > O. If :J;'(t) <S 1 for some
t> c, let a = inf{t > c: x'(t) = I}. Then, y'(t):::: 1 for c <S t <S a and
hence x(t)::::a+(t-c) for e<St<Sa. So,
(3
= x'(c) = x'(a) +
q(s)x-A(s)ds
l
+1
<
1+
<
00
q(s)[a
+ s - erAds
q(s)[a + s - cl-Ads
x"(t)
+ q(t)x-A(t) =
Furthermore, x( 00)
0,
x(e) = a,
x'(oo) = O.
(8.2.8)
00
00.
(8.2.9)
Chapter 8
588
So, y'(t) - x'(t) 2:: y'(O) - x'(O) for t 2:: 0, and hence y'(oo) - x'(oo) 2::
y'(O) - x'(O). This contradicts y'(oo) = x'(oo) = 0, and hence the
uniqueness follows.
Let
x{3 (t)
be the solution of
x"(t) + q(t)x->'(t) = 0,
x(O) =
x'(O) = {3.
0:,
Let T = {{3 : x~(t) 2:: 0 for t 2:: O} and S = {(3 : x~(t) < 0 for
some t 2:: O}. By Lemma 8.2.1 if {31 E Sand {32 E T, then {31 < {32.
By Proposition 8.2.1, T i= (/) and every negative number belongs to S.
Thus, sup S = inf T and if b = sup S, then 0::; b < 00. By continuous
dependence on initial conditions b tJ. S. So, bET and we need only to
show that x~(oo) = O.
Let x(t) = Xb(t) and suppose x'(oo) > O. Then, x(t)j(t + 1) is
continuous and positive for t 2:: 0 and limHoo x(t)j(t + 1) = x'(oo) > O.
Therefore, if M = inf {x(t)j(t + 1) : t 2:: O}, then M> O. Choose 8> 0
such that
2 ] >.+1
8 exp ( >. [ M
Jo
M
<""4
and b - 8> Mj2 (since b 2:: x'(oo) 2:: M this is possible). Let y(t) =
Xb-t5(t). If y(t)j(t + 1) > Mj2 for all t > 0, then clearly b - 8 E T
contradicting b = infT. Suppose y(t)j(t + 1) ::; M/2 for some t > O.
Since y(O) = x(O) > M/2, y(t)/(t + 1) 2:: M/2 for 0::; t ::; a, where
a = inf{t > 0: y(t)j(t + 1) ::; Mj2}. Also, y(a)j(a + 1) = Mj2. Now, for
o ::; t ::; a, we have
~ _ .!!l!l
t +1
and hence
+1
::; 8 exp
x(a)
a+1
[>. (2)
>'+1 t
M
Jo
< y(a) + M
a+1
q( s)
1<
(s + 1)>' ds
3M
4
4'
589
contradicting the choice of M. Thus, x' (00) = 0 and this proves the
existence.
Let x(t) be the solution of (8.2.8). Then integrating (8.2.8) from t
to u and letting u -+ 00, we obtain
x'(t) =
00
q(s)x-A(s)ds.
00
= a+
00
(8.2.10)
now gives
sq(s)x-A(s)ds
(8.2.11)
and since a < x(t) < x(oo) it follows from (8.2.11) that x(oo) <
and only if condition (8.2.9) holds.
00
if
Asymptotic Behavior
The following theorem deals with the asymptotic behavior of proper
solutions of equation (8.2.1) which satisfy case (Id.
Theorem 8.2.2.
Let x(t) be a proper solution of equation (8.2.1)
satisfying (II). Then there exists a constant a > 0 such that
(8.2.12)
Furthermore, if
00
s-Mlq(s)ds <
00,
to::::: 1
(8.2.13)
to
as t -+
00.
(8.2.14)
Proof. Let a = limHoo x'(t). Then, a> 0 and x(t) = at[1 + 0(1)] as
00. Integrating equation (8.2.1) from t to v and letting v -+ 00,
we obtain
t -+
x'(t) - a
00
q(s)x-A(s)ds
Chapter 8
590
by b. Then integrating (8.2.15) from t to
x(t)
= at+b-a->'[l+o(l)Jl
OO
00
100
gives
u->'q(u)duds.
(8.2.16)
x(t) = a - a->'[l
as t --+
00.
(8.2.17)
Proof.
Since condition (8.2.9) holds, we have by Theorem 8.2.1 that
limt-400 x(t) = a < 00 for some constant a > O. Integrating equation
(8.2.1) from t to u and letting u --+ 00, we obtain
x'(t) =
100 q(s)x-A(s)ds
x(t)
a - a- A[l
=
00
gives
+ o(l)J 100
00
q(u)duds.
(8.2.18)
(8.2.19)
00
sq(s)ds
00,
(8.2.20)
Lemma 8.2.3. Let y(t) and w(t) be two positive proper solutions of
equation (8.2.1) satisfying (h). If condition (8.2.20) holds, then y(t) rv
z(t) as t --+ 00. (The notation y(t) rv z(t) as t --+ 00 means that
limHOO y(t)/z(t) = 1).
591
Proof. Without any loss of generality, we can assume that yet) and z(t)
are defined for t::::: 0 and yeO) > z(O). By Proposition 8.2.1, condition
(8.2.5) holds. And, by Proposition 8.2.2, yet) cannot intersect with z(t).
Thus, yet) > z(t) for t::::: O. Also, y'(t):s; z'(t) for t::::: 0, if for
y'(a) > z'(a) for some constant a::::: 0, then for t::::: a,
y'(t)-z'(t) = y'(a)-z'(a)+
(8.2.21)
Since condition (8.2.20) holds, we have by Proposition 8.2.2 that z(t) ---+
as t ---+ 00. Thus, by (8.2.21), yet) '" z(t) as t ---+ 00.
,>
00
Lemma 8.2.4. Suppose q(t), ql(t) E C([a,oo),R o), a::::: 0, q(t):s; ql(t)
and
0 is a constant. Let x(t) and yet) be positive solutions of
equation (8.2.1) with x(a)
x'(oo) = 0 and the equation
="
(8.2.23)
respectively. Then, y(t)::::: x(t) for t::::: a.
Proof. Suppose there exists a constant b > a such that y(b) < x(b).
Let c = max{t < b : yet) ::::: x(t)}. Then, y(e) = x(e) and yet) < x(t)
for e < t :s; b. Thus, for some dE (e, b), y'(d) < x'(d) and y(d) < xed).
Hence, by Lemma 8.2.1, yet) < x(t) for t::::: d. Therefore,
x"(t) - y"(t) = ql(t)y->'(t) - q(t)x->'(t) >
for
t::::: d.
Hence, x' (t) - y' (t) ::::: x' (d) - y' (d) > 0 for t::::: d, contradicting
limHoo x'(t) = limHoo y'(t) = O. Thus, y(t)::::: x(t) for t::::: a.
y"(t)
+ ql(t)y->'(t)
0,
y'(oo) = 0,
00.
Proof. Let pet) = ql(t)/q(t), t ::::: 0 and the numbers E, 'rl E (0,1).
Choose a constant a> 0 such that (1- E)r < pet) < (1 + E)r for t::::: a.
We can assume that x(t) and yet) are defined for t::::: O. By Proposition
Chapter 8
592
8.2.1, condition (8.2.5) holds. And, by Proposition 8.2.2 there exist unique
solutions YI (t) and Y2(t) of
y~'(t)
+ (1- E)rq(t)Y1A(t)
0,
YI(a)
y~(t)
+ (1 + E)rq(t)Y2"A(t) =
0,
= yea),
y~(=) = 0
and
respectively. Since (1 - E)rq (t) ::; qI (t) ::; (1 + E)rq (t) for t 2: a, we have
by Lemma 8.2.4 that YI(t)::; yet) ::; Y2(t) for t;::: 0,. By Lemma 8.2.2,
[(1- E)rJI/(A+Ilx(t) and [(1 + E)rj1/(A+ Il x(t) are solutions of
y~(t)
+ (1- E)rq(t)Y1A(t) =
0,
y~(=)
0,
y~(=)=o,
and
y~(t)+(I+E)rq(t)Y1A(t)
=0
Since E and 'f/ are arbitrary numbers in (0,1), yet) '" rl/(A+I)x(t) as
t~=.
q(t)
00
Sql (s)ds
(8.2.23)
z"(t)
+ B(t)Z-A(t) =
0,
where
1 [
tf"(t)]
B(t) = A + 1 2 + J'(t) ql(t).
(8.2.24)
593
Since z(t) > 0 and B(t) > 0 for all sufficiently large t, we have by (8.2.24)
that limHoo z'(t) ~ O. Let 00= limHOO z'(t). If a> 0, then z(t) ~ oo(t)
as t ---c> 00, and hence f(t) ~ a as t ---c> 00, contradicting f(t) ---c> 0 as
t ---c> 00. Hence, z'(oo) = O. Also, limHoo B(t)/ql(t) = (2+r)/(1+'\) > O.
So, by Theorem 8.2.4,
z(t)
rv
(21 + r)
1/(.\+1)
+,\
x(t)
as
---c> 00.
Therefore,
x(t)
1+,\)1/(.\+1)
( -tf(t)
2+r
as
---c> 00.
Remark 8.2.1.
1. In Theorem 8.2.5, the number r can be calculated by the following
equation
tJ"(t)
(8.2.25)
I'(t)
= ,\
+ 1 and (3
-1).
ftC; sq(s)ds =
00,
594
Chapter 8
(8.2.26)
as t --+
or
00,
x(t) = a-a->-[1+0(1)lloo(S-t)so(Ins)!3ds
(8.2.27)
-m-1
(In t)n+l
-n -1
tt
if
if
< -1
(8.2.28)
=c=
-1
and
n <-1
r=
tf"(t)
hm--=
Hoc
f'(t)
Then
a +2
..
, - - 2 If (ld holds
{ ,,+1
-l'f
1 (.12 ) 110ld S.
Since r > -2, we have by Theorem 8.2.5 that if x(t) is a proper solution
of equation (8.2.1) satisfying (12)' then
x(t) ~ (
A+
r
1)
+2
1/(>-+1)
(A
tf(t)
+ 1)2
(a+2)(A-1-a)
] 1/(>.+1)
-(3 - 1
] 1/(>-+1(lnt)!3/(>-+1)t(2+c/(>.+1) if (i ) holds
x(t)
at+o(l:lO u O ->-(lnU)!3dudS)
as
t--+oo.
y(t) = (
~: ~ )
1/(>-+1)
(In t)(!3+ 1 )/(>-+l).
595
Then, - y"(t)y>'(t) '" q1(t) '" q(t) as t -+ 00. Hence, by Theorem 8.2.4,
if x( t) is a proper solution of equation (8.2.1) satisfying (i3) then
(8.2.29)
for some constants a and b, a >
o.
x(t)
lim x(t) = 0
lim x'(t) = O.
and
t ..... oo
t ..... oo
(8.2.30)
Q(t)
00
q(s)ds
(8.2.31)
ito
00
(8.2.32)
and
ito
00.
(8.2.33)
596
Chapter 8
Then for any constant P> 0 equation (8.2.2) has a positive solution x(t)
satisfying
(8.2.34)
lim x(t) = P,
lim x'(t) = 0
t--">oo
t--">oo
x(t)
00
P+
~ to.
(8.2.35)
First, we shall solve equation (8.2.35) in some neighborhood of infinity,
say, for t ~ T, T being sufficiently large. As in Theorem 8.1.1, choose
k, m > 0 and T ~ to so that
p-A
oo
Q(s)ds + >.p-A-l m
+>.P-A-1k
P- A ::::; k
and
oo
sQ2(s)ds
(h
OO
(h
OO
Q(S)dS)
>.p-A-l ( m
+k
oo
SQ2(S)dS)
< P,
Q(S)dS) ::::; k.
o ::; For t
+k
00
Q2(s)ds
for
~ T.
(Fx)(t) = P+
00
Q(s)x- A(s)ds->.lO
(1
00
Q(u)x-A-1(u)X'(U)dU) ds.
597
We claim that x'(t) < 0 for t 2': T. To this end suppose that X'(T)
for some T E [T, 00 ). Since
=0
x'(t) = - Q(t)x-A(t) + A
.(JO Q(s)x-A-I(s)x'(s)ds
for
t 2': T (8.2.36)
I;
o=
Q(T)X-A(T)
+A
LX) Q(s)x-A-I(s)x'(s)ds
< O.
= [to, (0).
x(t) -+
as
00
x(t)
= x(T) + x'(T)(t - T) + [,
t -+ tt.
(8.2.37)
(tl, T]
which follows from (8.2.2) shows that x(t) tends to a finite limit as t -+ tt.
This contradicts (8.2.37), and so, we must have J = [to, (0), i.e., x(t)
exists on the whole interval [to, (0). This completes the proof.
100
to
slq(s)1
(100 Q(u)du)
-A/(A+1)
ds <
00,
(8.2.38)
so that condition
let xn(t) be a positive solution
Chapter 8
598
of equation (8.2.2) satisfying x~(t) < 0 for t::::O: to, and xn(t)---t
lin, x~(t) ---t 0 as t ---t 00. The existence of such Xn (for each n) is
guaranteed by Lemma 8.2.5. Also, xn(t) for t::::O: to satisfies
Xn(t)
=
Since
1 (1
~ +1
~+
00
00
x~(t)
00
q(U)x;;A(U)dU) ds
Q(s)x;;A(s)ds - A
1 (1
00
00
Q(u)x;;A(u)X'(U)dU) ds.
(8.2.39)
which implies
00
Xn(t)::::O:
Q(s)ds
l/(Ml)
(8.2.40)
t::::O: to
xn(t)
1
+1
<
~+
<
(s - t)lq(s)1
(1
slq(s)1
00
00
(1
00
00
Q(U)dU) -A/(MI) ds
t::::O: to.
(8.2.41 )
Similarly,
x~(t)
Ix~(t)1 ~
00
Iq(s)1
roo
is
Q(u)du
-A/(A+l)
ds,
t::::o: to
(8.2.42)
Inequalities (8.2.41) and (8.2.42) assert that the sequence {xn(t)} is uniformly bounded and euicontinuous on each compact subset of [to, 00).
Therefore, Ascoli-Arzela's theorem implies that {xn} has a subsequence
{XnJ converging to some x E C([to,oo),lR) uniformly on each compact
subset of [to, 00). From (8.2.40) it follows that x(t) > 0 for t::::o: to
Letting ni ---t 00 in the equation
599
Hence, x(t) is a positive decaying solution of equation (8.2.2). This completes the proof.
Since q(t)
00
q(t) =
_(I~sint)/,
t +c
(C 2 -
as t --t
Q(s)ds
f )
t-l-f
1+
E>A.
and
+ 0 (C 2 -
--
00,
t;:::l,
f )
;::: CC 1 -
as
t --t
00,
for some constant C > 0, a.ll conditions of Theorem 8.2.6 are sa.tisfied and
hence there exists a positive decaying solution of this equation.
Theorem 8.2.7. Let q(t);::: 0 for t;::: to and A E (0,1). Suppose there
exists a nonincreasing function p( t) E C ([to, 00 ), lR+) such that
s:
q(t)
and
00
pet)
t;::: to
for
p1/2(s)ds <
(8.2.43)
(8.2.44)
00.
to
J::
X};'+1)/2(t) <
and
(-1)
(.H1)/2
(X~(t))2 s:
(1
1/2
00
1 2
p / (s)ds
t > to
,-
(8.2.45)
C~ A) p(t)x~->'(t),
(~[x~(t)f)'
2)
+ -+2-A) (I-A
t;::: to
(8.2.46)
= x n ), we find
q(t)x;;->'(t)x~(t)
for
t;::: to
Chapter 8
600
This establishes (8.2.46). Since (8.2.46) can be rewritten as
_(X~1+>')/2(t))'
::;
C;>') C~>.y/2p1/2(t)
for
t"2to
x = Xn together with (8.2.40) shows the uniform bounded ness and the
equicontinuity of {x~} on each compact subset of [to, (0). Hence, it
follows by Ascoli-Arzela's theorem that there exists a subsequence {xni}
of {xn} which converges to a function x E C 2 ([to,00),lR) in the C 2 topology on each compact subset of [to, (0). It is clear that x(t) > 0 on
[to, (0). Letting ni ---* 00 in the equation X~i(t) = q(t)X;;i>'(t), t"2 to
and the inequality
X~:+1)/2(t):::;
t"2 to
Remark 8.2.2.
-X'(t)X>'(t) -
>.100 x>.-l(s)(x'(s)?ds
= Q(t),
t"2 to
which implies - (x>'+l(t)j(>. + I)' "2 Q(t) for t"2 to. Integrating this
inequality from t"2 to to T"2 t, we obtain
Letting
>'+1
100
Q(s)ds,
t"2 to.
(8.2.47)
601
Joo
It:
w'(t)
A 1 q(t)
+ A+
(1
t
00
Q(s)ds
)-1 w(t) =
0,
(8.2.48)
w'(t)
+ c(t)w(t)
0,
t 2: to
(8.2.49)
where
Since each x
that
= Xi
A 1 q(t)
c(t) ::; >. +
(1
t
00
Q(s)ds
)-1
for t 2: to.
Chapter 8
602
Corollary 8.2.1.
If
A 1 t 2q(t)
A+
(1
00
Q(s)ds
)-1
::;
t,
then equation (8.2.2) has exactly one positive decaying solution. In fact,
the singular Emden-Fowler equation x"(t) = r 2-o.x-)..(t), t 2: 1, a> 0
has exactly one positive decaying solution
x(t) =
provided Aa(a
) 2 ] l/()"+l)
A+ 1
a(a + A + 1)
Co./()"+l)
for
t 2: 1
y"(t)
p(t)y-)..(t) ,
(8.2.50)
where p(t), q(t) E C1([to, 00), JR.+). In the following theorem, we assert
that if p(t) and q(t) have the same asymptotic behavior in some sense,
then so do the positive decaying solutions of these equations as t ---+ 00.
For this, we notice that when equation (8.2.50) admits a positive decaying
solution, then the integral P(t) = ftoo(s - t)p(s)ds converges for t 2: to.
To prove our result we shall need the following elementary lemma.
Lemma 8.2.6. If x(t) E Cl([to, 00), JR.) satisfies Joo x2(s)ds < 00 and
Ix'(t)l::; c, t 2: to for some constant c> 0, then limHoox(t) = O.
Theorem 8.2.9. Let x(t) and y(t) be positive decaying solutions of
equations (8.2.2) and (8.2.50), respectively. Suppose that
lim q(t) = 1,
t--+oo
00
p{t)
(8.2.51 )
(8.2.52)
o
where Y(t)
603
(8.2.53)
t-+oo
00
11 ds
< 00,
(8.2.54)
or
(8.2.55)
is satisfied. Then, x(t),..., y(t) as t ---+ 00.
C1 P1 /(>.+l l (t)
x(t),
y(t)
~ c21O (s -
to,
t)p(s)P->'/('\+ll(s)ds (8.2.56)
for some constants Ci = Ci(A) > 0, i = 1,2. In fact, the first inequality
in (8.2.56) is obtained from (8.2.47), while the second inequality in (8.2.56)
follows from y(t) = ftoo (fsoo p(u)y->'(u)du) ds, t ~ to. Define the new
function v by v(t) = x(t)/y(t) for t ~ to. Then, v(t) satisfies the
equation
vl/(t)
q(t)y->.-l(t)v->'(t) (8.2.57)
d2 v
dT2
+ p(t)y3->'(t)v =
q(t)y3->'(t)v->-,
ftto
y-2(u)du,
T ~ O.
(8.2.58)
v-
+v =
g(s)v->-,
~ 0,
(-
= :s)
(8.2.59)
where
f(s) =
and
g(s) = q(s)/p(s)
for
s ~ O.
8-+00
(8.2.60)
Chapter 8
604
o<
s--+oo
/00 Ig(s)lds
<
<
00,
(8.2.61 )
(8.2.62)
00
(8.2.63)
00,
s2:0
(8.2.64)
By rewriting equation
(8.2.67)
C3
605
C4
00,
> 0,
Joo v (s)d8
2
<
(8.2.68)
00.
From the boundedness of v(s), (8.2.68) and Lemma 8.2.6, we find that
limHooi'(8) = O. Accordingly, it follows from (8.2.67) that the limit
must exist as a finite value, i.e., lim s-+ oo V(8) = (finite). Letting 8 ---* 00
inequation(8.2.59),weget lims-+oov(8)=->--. If ->-_1-0, then
the boundedness of V(8) is violated. Hence, ->- = (I, i.e., (I = 1 as
desired.
Next, let condition (8.2.55) hold. We notice that equation (8.2.67) is
equivalent to
(8.2.69)
for some real constant C5. Condition (8.2.63) ensures the convergence of the
right-hand side of (8.2.59). Hence, we have (8.2.68) and lims-+cxo {;(8) = 0,
which in turn imply that the limit
exists in JR. The rest of the proof is similar to that of presented above and
hence omitted. This completes the proof.
limsup p-l/2(t)
t-+cxo
cxo
(P*(8))1/2ds <
00.
Chapter 8
606
Consider the case when the function pet) behaves like eta where c
and 0 are constants, c> 0 and 0 < -2, i.e., assume that
t .... oo
and
p'(t)
o<
ta
t .... oo
-p'(t)
t .... oo
ta -
t .... oo
(8.2.70)
00
(8.2.71 )
00.
Then for any positive decaying solution yet) of equation (8.2.50), (8.2.56)
shows that C 1t(a+2)/(A+l) :::; yet) :::; C 2 t(a+2)/(A+1), t ~ to for some
positive constants C 1 and C 2 . Moreover, the equation - y'(t) =
ftOO p(s)y-A(s)ds, t ~ to gives the estimates C3t(a+2)/(M1)-1 :::; -y'(t)
:::; C4 t(a+2)/(A+1)-1, t
to.
OOl- l q (S)
- - 1 ds
s pes)
rv
<
00,
or
y(t) as t --+
00.
/00
or
(8.2.73)
00
holds. Then any positive decaying solution x(t) of equation (8.2.2) satisfies
x( t)
Proof.
rv
(0
(A
1)2
+ 2)(0 -
A + 1)
11/(HA)
t(o:+2)/(Ml)
as
--+
00.
] 1/(A+1)
( A 1)2
+
t(a+2)/(M1)
(0 + 2)(0 - A + 1)
c
for
t>
-
to
607
A:::: 3
and
0 <
.
q(t)
rllllm
. f -q( t) < IllllSUP
<
t ..... oo
1. .....
to:
t'"
00
00
.
-q'(t)
rImln
. f-q'(t)
- < lUll sup - - 1 - <
1. .....
to<-1
t ..... oo
to:00
00.
It:
d2 v
d/ 2
+ q(t)y3-A(t)[V -
v-
f(s)v
+v
v-A,
V-A] = 0,
~ O.
(8.2.74)
I; (q[t(E)]y3-A[t(E)]) 1/2 dE
s
~ 0,
(- = :s)
trans-
(8.2.75)
over
[Sn, Sn+1]
yields
=1.
This
Chapter 8
608
x(t), yet)
rv
KtlT
x'(t), y'(t)
and
rv
where
K =
c().
+ 1)2
] 1/(>+1)
and
(a+2)(a-).+1)
a+2
(J= ),+1'
C- 1 / 2 K(A+l)/2[). -
2a
t--+oo
lim x(t)
t--+oo
c,
c E lR*
= lR+ U {oo}.
(8.2.77)
(8.2.78)
609
o.
>
x'(t) = x'(to)
+ it q(s)x-:>"(s)ds ~
x'(to)
00
for t:2: to. Since x'(t) is an increasing function for t:2: to it must have
a finite limit as t -+ 00, and this completes the proof of the assertion.
From this observation, we restrict our attention to the case when condition (8.2.78) fails to hold. We shall use the following notation: For
t :2: to, Ql(t) = It: s-:>"q(s)ds, Q2(t) = Itto Qi/(:>"+l) (s)ds and Q3(t) =
ItDO Q:;2(s)ds. Now, we shall prove the following lemma.
Lemma 8.2.7. Suppose
(8.2.79)
Then any positive increasing solution x(t) of equation (8.2.2) satisfies
. f x(t)
1Imln
Q ()
2
t-+oo
>
(\ + 1)1/(:>"+1).
_ /\
Proof.
x(t o) + (t - to)x'(t)
for
t:2: to
cx~t~:+l)'
:2: C>'q(t)[l
+ 0(1)]
as
t -+ 00.
and
+ 0(1)]
as
t -+ 00.
J'(t)
..
J(t)
J(t)
f'(t)
< hmmf
-() < hmsup -() < hmsup ~().
t-+oo 9 t
t-+oo 9 t
t-+oo 9 t
Chapter 8
610
y(t) =
(8.2.80)
where C is a positive constant. Hence, it is natural to expect that the
positive increasing solutions of (8.2.2) behave like y(t) when the coefficient
of (8.2.80) is asymptotic to the function q(t). This observation leads to
the following results.
Theorem 8.2.12. Let 0
lim
t-+oo tQ~/(>'+l)(t)
a E 1R+
(8.2.81 )
hold. Then any positive increasing solution x(t) of equation (8.2.2) satisfies
(8.2.82)
Theorem 8.2.13. Suppose conditions (8.2.79) and (8.2.81) hold. Further,
assume that
lim q(t)Q~->'(t)Q~(t) = bE 1R+
(8.2.83)
t->oo
and either
00
/
Q3-1( S)Q2-2()
S
Q2(S)
1/(>.+1)
sQ 1
(s)
a ds <
00,
(8.2.84)
or
hold. Then any positive increasing solution x(t) of equation (8.2.2) has
the asymptotic form (8.2.82).
Theorem 8.2.14.
Suppose in addition to (8.2.79), (8.2.81), (8.2.83),
(8.2.84) the following condition holds
Also, from Lemmas 8.2.7 and 8.2.8 the following result is immediate.
611
Theorem 8.2.15. Let 0 < A < 1, and x(t) and y(t) be positive
increasing solutions of equations (8.2.2) (I,nd (8.2.50), respectively, such that
x(t)
;::0:
my(t)
for
t;::o: to,
(8.2.85)
for some constant m > O. Suppose further that condition (8.2.79) and
. q( t)
11m -()
= 1
P t
(8.2.86)
t-+=
hold. Then,
x(t)~y(t)
as t--+oo.
],\_,\
Q2(t)
z"(t) = q(t) [ tQ~/('\+l\t) z (t)
for
t;::o: tl > to
;::0:
Lemma 8.2.7 and Theorem 8.2.15 show that the assumption 0 < A < 1
in Theorem 8.2.12 is superfluous if a = 1 in condition (8.2.81). We state
this result in the following.
Corollary 8.2.4. If conditions (8.2.79) and (8.2.81) with a = 1 hold,
then any positive increasing solution x(t) of equation (8.2.2) satisfies
x(t) ~ (A + 1)1/(A+l)Q2(t) as t --+ 00.
Theorem 8.2.16. Let x(t) and y(t) be positive increasing solutions
of equations (8.2.2) and (8.2.50) respectively, satisfying (8.2.85). Suppose
conditions (8.2.79), (8.2.86),
lim p(t)y3-'\(t)
t-+cxo
(1= y-2(S)d8)2
t
= E R+,
(8.2.87)
(8.2.88)
and either
(8.2.89)
or
(8.2.90)
y(t) as t --+
00.
612
Chapter 8
for
t 2: to
(8.2.91)
for some constant M > O. Set v(t) = x(t)/y(t) for t 2: to. Since v(t)
satisfies equation (8.2.57), the change of variable T = (JtOO y-2(u)du)-1
implies that v( T) satisfies
v + il + p(s)v
q(s)v->"
for
s 2: So = ln TO,
C :s)
=
(8.2.92)
(8.2.93)
= f,
(8.2.94)
8-+00
lim q(s)
1
1 I~~:~ -11
8-+00
00
\(q(s'\ds <
00
(8.2.95)
00
ds <
(8.2.96)
00
and
(8.2.97)
respectively. By (8.2.91), we have
m :s; v(s) :s; M,
s 2: So
(8.2.98)
Since (8.2.93) and (8.2.94) imply that lims-+oo q(s) = f, the argument used
in the proof of Theorem 8.2.9 shows that v(s) is bounded on [so, 00),
and from equation (8.2.92), we see that v(s) is also bounded. Multiplying
equation (8.2.92) by v(s), we get
p-()
S (q(s)
p(s)
-1) v
->...v,
s>
_ so
(8.2.99)
613
We shall prove that lims--+oo v(s) = 1 when condition (8.2.89) holds. The
other case when (8.2.90) holds can be proved similarly, and will not be
included. Integrating equation (8.2.99) over [so, sJ, we obtain
{)2(s)
2
1
-1
+
i,2(7])d7] + p(s)
[v 2(S) _ V1-A(S)]
2
So
1:
So
1 - .\
[:~~~
p(rl)
(8.2.100)
s
~ so,
Consider the case when q(t) behaves like a positive constant multiple
of to:, a -.\ + 1 > O. Lemma 8.2.7 shows that for every positive increasing
solution x(t) of equation (8.2.2), x(t) ~ mt(<>+2)/(A+1), t ~ to for some
constant m > O. On the other hand, it is easily seen that the equation
y"(t) = ct<>y-A(t), t ~ to where c > 0 is a constant, admits a positive
increasing solution y(t) given by
y(t)
( .\
1)2
] l/(A+1)
(a+2)(a-.\+1)
t(<>+2)/(A+l)
for
> to.
-
ct a ,
Corollary 8.2.5. Let a> .\-1. Suppose q(t) = eta [l+E(t)J, where e is
a positive constant, E(t) E C([to, (0), lR), limt--+oo E(t) = O. If 0 < .\ < 1,
then
l/(A+l)
x(t) rv
e(.\ + 1)
t(a+2)/(A+1) as t --+ 00.
2]
[(a+2)(a-.\+1)
-1
y-2(s) Ip(s)y3-A(S)
(f.oo y-2(U)dU)
2 _
RI ds <
00
Chapter 8
614
and either (8.2.89) or (8.2.90) hold. Then, x(t)
A > 0, 11
(ii)
hoc
q(s)ds
lim x(t) = 0,
t-'too
lim y(t) = 0,
t-'too
and
{
lim x(t)
= ,
lim y(t)
= 0,
t--l-oo
t-'toc
=0
lim x'(t)
t-'toc
lim y' ( t) = 0
(8.2.101)
t--l-OC
lim x'(t) = 0
t-'too
t-'too
= 0,
(8.2.102)
where f > 0 is a constant. For this, we need the following basic lemma.
Lemma 8.2.9. Suppose
1
1
1
(8.2.103)
Q(s)ds < 00
(8.2.104)
(8.2.105)
to
00
to
and
00
to
615
= = f,
lim y(t) = m,
t ..... =
lim x(t)
t .....
lim x'(t) = 0
t ..... =
lim y'(t)
t ..... oo
(8.2.106)
=0
x(t) = f
y(t) = m
.
(8.2.108)
for t::::: to. First, we solve this system in some neighborhood of infinity,
say, t::::: T ::::: to. Choose positive constants c, k and T::::: to so that
>'m-:>..-l (rl' +k
00 P(S)dS) ::; c,
JLf-l'-l (m-:>.. +c
00 Q(S)dS) ::; k,
Consider the set X of all functions (x, y) E C 1 ([T, 00), ]R+) X C 1 ([T, 00 ),
]R+) = Y, where components satisfy the inequalities f::; x(t) ::; 2f,
m ::; y(t) ::; 2m,
o< -
o ::; - y'(t)
::; rI'Q(t)
100
+ 100
+c
P(s)Q(s)ds
P(s)Q(s)ds
Xl(t)
=f+
x~(t) =
616
Chapter 8
and
Yl(t)
m+
00
1 (1
fJ
Q(s)x-IL(s)ds -
00
00
Q(u)X- IL - 1 (u)X'(U)dU) ds
00
P(u)y->--l(u)y'(u)du
:::; m->--l
00
00
P(u)Q(u)du + k
+ Am->--l
+c
00
(1
00
00
P(~)Q(~)d~) dU]
P(U)dU)
(LX) P(U)Q(U)dU)]
P(s)Q(s)ds
for
00
P(S)Q(S)dS)
2:: T
and
o :::; x(t) -
1 (1
00
1 (1
00
00
P(U)dU)
:::; m->-looP(s)ds
:::; C for
(1
00
P(U)y->--i(u)IY'(u)ldU) ds
[rlL 1 (1
00
00
00
P(U)Q(U)dU) ds
P(u)Q(U)dU)dS]
t 2:: T.
The estimates for YI (t) and y~ (t) can be obtained similarly. Thus, we find
that F X c X. Furthermore, it is easy to show (see previous arguments)
that the mapping F is continuous and F X is compact. Therefore, the
Schauder-Tychonov theorem implies that F has a fixed element (x,y) E
X. Hence, the system (8.2.107), (8.2.108) admits a positive solution (x, y)
in [T,oo).
617
x'(t)
y'(t)
1=
- P(t)y->'(t)
+A
Q(t)x-Il-(t)
+ /-t
00
P(s)y->'-l(s)y'(s)ds
(8.2.109)
Q(s):r;-Il--l(s)x'(s)ds
(8.2.110)
for t ~ T. Now, we claim that x'(t) < 0, y'(t) < 0 in [T,oo). Suppose
not, then X'(7) = 0 for some 7 ~ T. It follows from (8.2.109) that
A
P(s)y->.-l(s)y'(s)ds = 0, i.e., P(t)y'(t) == 0 in [7,00). Multiplying
(8.2.110) by P(t), we find that P(t)Q(t) == 0 in [7,00). This contradicts
condition (v). Hence, x'(t) < 0 in [T,oo). Similarly, we can show that
y'(t) < 0 in [T,oo).
I:"
Now, for the case AJL < 1 we shall prove the following results which
guarantee the existence of positive solutions of (8.2.3) satisfying (8.2.101)
or (8.2.102) with f> O.
Theorem 8.2.18. Let A/-t < 1 and conditions (8.2.103) and (8.2.104)
hold. Further, suppose that
(8.2.111)
and
(8.2.112)
Then the system (8.2.3) has a positive solution (x, y) satisfying (8.2.101).
Proof.
Clearly, the assumptions of Lemma 8.2.9 are fulfilled. Thus,
for n E IN there exists a positive solution (xn,Yn) E C 2([to,00),1R+) x
C2([to, 00), 1R+) = Y of the system (8.2.3) satisfying
x~(t)
< 0,
y~(t)
< 0 for t 2. to
(8.2.113)
618
Chapter 8
Xn(t) =
=
Yn(t) =
l (1=
~ +l
~ 1= (1=
~ +1
~+
CXl
CXl
p(U)y;:A(U)dU) ds
P(s)y;;A(s)ds - A
1=(1= P(U)y;;A-l(U)Y~(Zl)dU)
ds
(8.2.114)
q(U).T;:I'(U)dU) ds
00
Q(s)x;;!l(s)ds - ft
1 (1=Q(U)Xr/,-l(u)x~
00
(U)dU) ds.
(8.2.115)
1
1
00
P(s)ds,
t:::: to
(8.2.116)
00
Q(s)ds,
t:::: to.
(8.2.117)
Thus, it follows from (8.2.114) and (8.2.117) and the decreasing nature of
-x~(t) =
00
p(s)y;;A(s)ds :S
<
or
_ ( x1-A!l(t))' <
1 - Aft
-
00
00
Ip(s)1
x~!l(t)
ip(s)1
~
(1
1=
(100
s
00
Ip(sll
Q(U)dU) -A
(1
Q(u)du
and letting
cY0
x~l'(s)ds
Q(U)dU) -A ds,
)-A ds.
(8.2.118)
-+ 00, we get
1 (1
~i + 1= (1
~i +
Xni (t)
Yni(t) =
00
00
00
ni
--+
619
00
in the
2: to
p(u)y;,A(U)dU) ds,
q(u)X;;(U)dU) ds,
t 2: to
Theorem 8.2.19. Let All < 1, (8.2.111) and ft':' slq(s)lds < 00 hold.
Then for any constant . > 0, the system (8.2.3) admits a positive solution
(x,y) satisfying (8.2.102).
l-AI-'(t)
.l-AI-'
A
::; - - , 1 - Jl
1 - /\11
xn
100
t
slp(s)1
t 2: to
Now, it is easy to see that xn(t) 2: for t 2: to, and hence by the above
inequality, we get the following estimate for Yn (t)
Yn(t) ::; -1
n
+ -1-'
1=
t
slq(s)lds,
t 2: to
(Id If conditions (8.2.104) and (8.2.111) hold, then the system (8.2.3)
admits no positive solution (x, y) satisfying
lim x(t) = 0,
t-+oo
lim x'(t) = 0,
t-+=
lim yet) = Cl E IR o,
t-+oo
x'(t) < 0, y'(t) < 0
Cl
lim y'(t) =
t-+=
is a constant
(8.2.119)
Chapter 8
620
(h) If conditions (8.2.103) and (8.2.112) hold, then the system (8.2.3)
admits no positive solution (x, y) satisfying
{
lim y(t)
lim x(t)
= C2
t-+oo
t-too
x'(t) < 0,
0,
lim y'(t)
IRo,
is a constant
t.......:;,oo
y'(t) <
C2
lim x'(t)
0,
t-H)O
(8.2.120)
t.
Proof. We only consider (Id, because (12) can be treated similarly. Let
(x, y) be a positive solution of the system (8.2.3) with the required property
(8.2.119). As in Lemma 8.2.9, we find that it satisfies (8.2.107) and (8.2.108)
with C = 0 and m = y(oo) E IRo. Now choose T 2 to so large that
x'(t) < 0, y'(t) < 0 for t 2 T. Then as in Theorem 8.2.18, we get
x(t)
<
::;
1 (1
1 [1 (1
1
(1
00
p(u)y->-'(U)dU) ds
Ip(u)1
00
x-)'I'(t)
x 1 -),I'(t)
00,
00
00
i.e.,
Letting t -+
00
00
00
(s - t)lp(s)1
slp(s)1
(1
00
00
2 T.
Example 8.2.3.
q satisfying
Let AIL
= 1,2
> 0 and {3 -
lLa
> 0,
(8.2.122)
then the system (8.2.3) has a positive solution (x, y) satisfying (8.2.101).
In particular, the singular Emden-Fowler system
(8.2.123)
621
+ 1)
b"f2b2 + 1)
a"llb1
b->',
- "11 - 2
>""12 - 2 - 0:
a- IL ,
"12 - 2
P,'Y1 - 2 - {3
or
{3 - p,o: > 0,
(8.2.125)
then the system (8.2.123) never has a positive solution (x, y) of the form
(8.2.124). On the other hand, Theorem 8.2.20 concludes that if condition
(8.2.125) holds, then the system (8.2.3) with (8.2.121) does not have any
positive solution (x, y) satisfying (8.2.101).
Example 8.2.4.
q satisfying
for t 2 1, where ai, bi, i = 1,2 are positive constants, 0:1 > 0:2 > 0 and
{31 > (32 > 0 are parameters. Let >..p, < 1. Then it follows from Theorem
8.2.18 that there exist O:i = D:i(>",p,) > 0 and (3i = (3i(>",p,) such that
the system (8.2.3), with this D:i(>", p,) and (3i(>", p,) has a positive solution
(x, y) satisfying (8.2.101). To see this it suffices to notice the fact that for
any given >.., p, satisfying >..p, < 1 inequalities
> 0,
{32 - P,0:1 > 0,
0:2 - >"{31
> 0
{31 - {32 > 0
0:1 - 0:2
admit a positive solution O:i, {3i, i = 1,2. Here, we may adopt the following
known results from convex analysis. For each n x n matrix M exactly
one of the following two cases holds:
(M 1 ) There exists n-vector satisfying
(M 2 ) There exists n-vector ", -=1= 0 satisfying M", < 0 and ",20.
(The order relation v 2 W [v > w] for vectors v = (Vi), W = (Wi)
defined as Vi 2 Wi [Vi> Wi] for all i).
IS
622
Chapter 8
Consider the system (8.2.3) with
Example 8.2.5.
(t)
p(t)
(1
E,
+ sin t )
t 2+ < '
q( t) =
q(t) = 0 (C 2 -
E)
1=
1=
ii ) ,
_ (1 + sin
t2+ii
Let AfJ
t)
P(s)ds 2'
CIt
Q(s)ds 2'
C2 C1 - ii
1-<
as t --+ 00 for some positive constants C1, C2. Therefore, Theorems 8.2.18
and 8.2.19 assert that system (8.2.3) has a positive solution (x, y) satisfying
(8.2.101) if A(1 + 5) < E and /1(1 + E) < 5 and that the system (8.2.3)
has a positive solution (x, y) satisfying (8.2.102), > 0 if A(1 + 5) < c.
It:
I1 l
liminf t .....
to
to
q(u)duds > -
(8.3.1)
00,
f'(x) 2' 0
for
i= 0
(8.3.2)
It: It:
623
(a(t)x'(t))'
+ q(t)f(x(t))
0,
where a(t) E C([to, (0), JR+). Also, to equations with deviating arguments
of the form
xl/(t) + q(t)f(x[g(t)]) = 0,
where g(t) E C([to,oo),JR) and
to the reader.
limHoog(t) =
00.
(lx'(t)la-1x'(t))' + q(t)lx(t)I~-lX(t) = 0,
(lx'(t)la-lX'(t))' + q(t)f(x(t)) =
and
(lx'(t)la-1x'(t))'
+ q(t)f(x[g(t)]) =
0,
where g(t), q(t) E C([to, (0), JR), f E C(JR, JR) and limHOO g(t) =
00.
(a(t)x'(t))'
+ i5q(t)x- A (t) =
a(t)
>.
0,
624
Chapter 8
where 0 = 1, a and), are positive constants, and q(t) E C([to, (0), lR).
8.4. References
1. G.J. Butler, On the oscillatory behavior of a second order nonlinear differential equation, AIlIl. AJat. Pura Appl. 105(1975), 73-92.
2. G.J. Butler, Oscillation theorems for a nonlinear analogue of Hill's equation, Quart. J Math. 27(1976), 159-171.
3. G.J. Butler, Integral averages and the oscillation of second order ordinary
differential equations, SIAM J. Math. Anal. 11(1980), 190--200.
4. T. K ura, Oscillation theorems for a second order sublinear ordinary differential equation, Proc. ArneI'. Math. Soc. 84(1982), 535-538.
5. M.K. Kwong and J.S.W. Wong, On the oscillation and nonoscillation
of second order sublinear equations, Proc. Amer. IvIath. Soc. 85(1982),
547-551.
6. M. Naito, Asymptotic behavior of solutions of second order differential equations with integrable coefficients, Trans. Amer. Math. Soc.
282(1984), 577-588.
7. M. Naito, Nonoscillatory solutions of second order differential equations
with integrable coefficients, Pmc. Amer. Math. Soc. 109(1990), 769-774.
8. M. Naito, Integral averages and the asymptotic behavior of solutions
of second order ordinary differential equations, J. Math. Anal. Appl.
164(1992), 37{}-380.
9. Ch.G. Philos and I.K. Purnaras, Asymptotic behavior of solutions of
second order nonlinear ordinary differential equations, Nonlinear Analysis
24(1995), 81-90.
10. S.D. Taliaferro, On the positive solutions of y" +<f;(t)y->' = 0, Nonlinear
Analysis 2(1978), 437-446.
Chapter 9
Miscellaneous Topics
9.0. Introduction
This chapter contains some special results. In Section 9.1, we shall
extend the Sturm-Picone theorem, obtain nonoscillation theorems for perturbed second order nonlinear differential equations, and present a nonlinear Picone type identity to enable us to prove some Sturm-Picone type
comparison theorems for nonlinear equations. Section 9.2 is devoted to the
study of nonoscillatory solutions of forced differential equations of second
order. In Section 9.3 we shall present some limit cycle criteria and discuss
its related properties for nonlinear second order differential equations. Finally, in Section 9.4 we shall present some properties of solutions of very
general second order differential equations.
(a(t)x'(t))'
and
+ q(t)x(t) =
(9.1.1)
(al(t)y'(t))' + ql(t)y(t) = 0,
(9.1.2)
where a(t), al(t) E CI([to, (0), JR+), q(t), ql(t) E C([to, (0), JR). It states
that if
a(t)
:s:
al(t)
and
q(t)
ql(t)
on
to> 0
then if any solution y(t) of (9.1.2) has two (or more) zeros on [tl, t2],
every solution of (9.1.1) also has at least one zero on [t l , t2J.
We shall compare the oscillatory behavior of solutions of (9.1.1) with
the solutions of the equation
(a(t)p(t)y'(t))'
+ q(t)p(t)y(t)
0,
(9.1.3)
626
Chapter 9
where p(t) E Cl([to, oo),lR). For this, we shall need the following lemma.
Lemma 9.1.1. If there exists a function y(t) E C l ([t l , t2J, lR) with y
not identically zero, and y(tl) = y(t2) = 0 such that
t2
t,
0=
1T2 (a(s)p(s)x(s)x'(s))'ds
T2 [12"a( s )p' (s) dsd (x 2(s)) +a(s )p( s )(x' (s))2 +p(s )x(s )(a(s )x' (s))'Jds
1
T1
T1
= -
'1
Thus, we have
1~2
(9.1.4)
Miscellaneous topics
627
Example 9.1.1.
((sint)y'(t))'
+ (k 2 sint)y(t) =
0,
k > 2.
(9 .l.5)
We shall show that any solution of (9.l.5) has a 2em on the interval [0,7f].
For this, we compare (9.l.5) with the equation x"(t) + k 2:r(t) = 0, which
has solutions with two zeros on the interval [0 + E, 7f - E] for a sufficiently
small E > O. Here a(t) = 1, q(t) = k 2 and p(t) = sint. Thus,
a(t)p(t) = sint > 0 and (a(t)p'(t))' = -sint < 0 on [E,7f - E]. Now by
Theorem 9.l.2 every solution of (9.l.5) has a zero on [E,7f - Elo and hence
on [0,7f].
Next, we shall consider the equation
(a(t)x'(t))'
+ q(t)x(t) =
g(t)f(x(t)),
(9.l.6)
where the functions a(t) and q(t) are as in (9.l.1), g(t) E l[to,OO)
locally, f(x) E C(JR, JR) and xf(x) > 0 for x of O.
Theorem 9.1.3. Suppose x(t) is a solution of (9.l.6) such that X(tl) =
0= X(t2)' t2 > tl > to. If there exists a function p(t) E C 1([tl, t2], JR)
such that a(t)p(t):::: 0, (a(t)p'(t))' ::; 0, g(t)p(t) ::::
on [tl' t2] and
g(t)p(t) E l[to,oo) locally, and if a1(t)::; a(t)p(t) and q1(t):::: q(t)p(t)
on [t1' t2], then allY solution of equation (9.1.2) will have a zero on [t1, t2]'
t2 (a(s)p(s)x(s)x'(s))'ds
itl
-'2
t,
it l
t2 [a(s)p(s)(x'(s))2 _ q(s)p(s)x2(s)] ds
itl
::; O.
The rest of the proof is similar to that of Theorem 9.l.1 and hence omitted.
Chapter 9
628
Corollary 9.1.1.
Subject to the hypotheses of Theorem 9.1.3 if any
solution of (9.1.2) has no zero on the interval [t I ,t 2 ], then any solution of
equation (9.1.6) will have at most one zero on [tI' t2]'
Example 9.1.2. Consider the differential equations
y"(t)
and
(2x'(t))'
+ y(t) =
+ x(t) =
(9.1.7)
(9.1.8)
Here a(t) = 2, q(t) = 1, g(t) = sint and f(x) = x3 . We claim that any
solution of (9.1.8) has at most one zero on the interval ['if /3, 7T /2]. For this,
we choose p(t) = sint. Then, a(t)p(t) = 2sint ~ 1 = aI(t), q(t)p(t) =
sint S 1 = ql(t), (a(t)p'(t))' = -2sint < 0 and g(t)p(t) = sin 2 t > O.
Hence, Corollary 9.1.1 is applicable, since there exists a solution of (9.1.7)
having no zero on [7T/3,7T/2).
Remark 9.1.2. If the condition xf(x) > 0 for x f. 0 in Theorem 9.1.3
is replaced by xf(x) < 0 for x =f 0, then Theorem 9.1.3 remains true
provided g(t)p(t) ~ 0 is replaced by g(t)p(t) S O.
(aI(t)x'(t))'
and
+ QI(t,X(t),x'(t)) = P(t,x(t),x'(t))
(9.1.9)
(9.1.10)
1,2. We
(9.1.11)
x
and
(9.1.12)
(9.1.13)
Miscellaneous topics
629
[:~~~ (al(t)X'(t)y(t) -
a2(t)Y'(t)X(t))]'
[Q2(t'~W),y'(t))
QI(t,:~tl)'X'(t))] + [al(t)
- a2(t)](X'(t))2
P(t,x,x') :S 0 for
(t,x,x')
(9.1.14)
QI(t,X,X'):S 0 if x:SO
(9.1.15)
QI(t,x,X') 2: 0 if x2:0.
(9.1.16)
y"(t)
+ q(t)y(t)
0,
(9.1.17)
Chapter 9
630
"( )
t
x(t)
+ q(t) x 2()
t +1
is also nonoscillatory.
Now we shall consider the special case of equations (9.1.9) and (9.1.10),
namely,
(al(t)x'(t))'
+ ql(t)J'I(X(t))
and
(a2(t)y'(t))'
P(t,x(t),x'(t))
+ q2(t)12(y(t)) =
0,
(9.1.18)
(9.1.19)
where aI, a2 and P are as in equations (9.l.9) and (9.1.10), ql(t), q2(t) E
C([to, 00), lR) and J'I, 12 E C(lR, JR).
Theorem 9.1.6. Assume that condition (9.l.13) holds,
(9.l.20)
J'I(x):::;'O
if
x:::;'O
and
h(O)=O,
f~(y) >
(9.l.22)
Proof. Let x(t) be a solution of (9.1.18) with the above properties and
suppose that y(t) is a solution of (9.1.19) with y(t) ej on [tIl t2]' Since
( X ((al(t)x
t ) ) , (t)fz(y(t)) ~ a2(t)y , (t)h(x(t)))
[h
fz(y(t))
= (al(t)x'(t))'J'I(x(t))
J'
+ al(t)f{(x(t))(x'(t))2 ~
(a2(t)y'(t))'j~i~g;;
Miscellaneous topics
631
j2
a2(t)ji~~~!~n
tl
to
t2
and
h (x)
::; 0
if
x::; O.
(9.1.26)
k < I,
Example 9.1.4.
t?: 2
(9.1.27)
and
0,
t?: 2
(9.1.28)
where q(t) E C([to, (0), lRo ), and assume that (9.1.28) is nonoscillatory.
Since f{(x) = (1/2) sech2(x/2)::; 1/2 and al(t) = 1+t?: (2t+sint)/2 =
a2(t)/2 for t?: t2, in view of Theorem 9.1.7 equation (9.1.27) is nonoscillatory.
In the following result, we shall replace condition (9.1.20) by
(9.1.29)
Chapter 9
632
= _k2ql (t) /f(x(t) +k2 h (x(t) )P( t, x( t), x' (t)) +k2al (t)f{ (x( t) )(x' (t) f
+ kq2(t)il(x(t)) - 2ka2(t)y'(t) h
(xX~~~~~~(t)) x'(t)
+ ka (t)(y'(t))d?(x(t))fHy(t))
2
Ji(y(t))
;::: kfi (x( t)) [q2 (t) - kql (t)] + k 2h (x(t) )P( t, x(t), x' (t))
+ k2al (t)f{ (x(t)) (x' (t))2 +a2(t)
x x'(t)y'(t)
=
+ k 2f{(x(t))(x'(t))2[al(t) - a2(t)]
+ a2(t)f{(x(t)) [h(x(t))y'(t) - kh(y(t))X'(t)]2.
The remainder of the proof follows as before.
Remark 9.1.5.
1. It is easy to obtain many corollaries and theorems similar to those of
Corollary 9.1.2 and Theorems 9.1.5 and 9.1.7. The formulation of such
results are left to the reader.
(a(t)(b(t)x(t))')'
O.
(9.2.1)
Miscellaneous topics
633
(a(t)(b(t)x(t))')'
+ F(t, x(t))
(9.2.2)
(a(t)(b(t)x(t))')'
(9.2.3)
(a(t)(b(t)y(t))')'
0,
(9.2.4)
and
where
C([to,oo),lR)
Fi
and
rt
R[t,T]
iT
ds
a(s)'
and
7r(t) =
to
and
IX) a~:)
and
lim R[t,T]
t-+oo
00
(9.2.5)
(9.2.6)
w(t)
ht
X(t,s)
00
Y(v,s)H(v,u(v))dvds
and Y
for
satisfy
~ T,
Z(t)
has a solution
w(t)
ht
X(t,s)
00
Y(v,s)H(v,z(v))dvds
F1(t,x)
F2(t,x)
for
x> 0
(9.2.7)
634
Chapter 9
(a(t)(b(t)TJi(t))')'
+ ei(t)
t2":to,
0,
i=1,2
(9.2.8);
and
lim b(t)r/i(t)
1-+00
0,
i=1,2
0,
(9.2.9)
. b(t)
11m - (
) T];(t)
t
1-+00 If
1,2
(9.2.10)
Proof. Let x(t) be a positive solution of (9.2.3) defined on [to, (0). Set
w(t) = x(t) - T]l (t). Then for t 2": to, we find
(a(t)(b(t)w(t))')'
o.
(9.2.11)
Thus, w(t) is eventually of constant sign. If w(t) < 0 for t 2": tl for
some tl 2": to, then 0 < x(t) < T]l(t), t 2": tl which contradicts the fact
that T]l(t) is oscillatory and hence we must have w(t) > 0 for t 2": tl.
Now we need to consider two cases:
(i) Suppose condition (9.2.5) holds. It is easy to see that a(t)(b(t)lL'(t))' >
for t 2": tl and the finite limit C2 = limHoo a(t)(b(t)w(t))' 2": 0 exists.
Integrating equation (9.2.11) from t 2": tl to u 2": t and letting u -+ 00,
we obtain
a(t)(b(t)w(t))' =
C2
00
Fl(S,X(s))ds,
2": tl.
(9.2.12)
b(t)w(t)
Cl
+C2R[t,h]
l -() 1=
1
t,
ass
or
b(t)x(t) = cl+b(t)T]1(t)+c2R[t,tl]+
for t 2": tl, where
t2 2": h such that
Cl
Fl(U,X(u))duds,
l 1=
t
t,
1
-()
ass
Fl(U,X(u))duds
(9.2.13)
(9.2.9) there is a
(9.2.14)
Miscellaneous topics
635
it 1
to
1
-()
00
a s
F2(u, x(u))duds,
(9.2.15)
Now by Lemma 9.2.1 it follows that there exists a continuous
solution y( t) of the integral equation
t :::: t2'
1 +b(t)ry2(t)+c2R[t, tll+
b(t)y(t) = -Cl
it 1
00
1
-()
has
(9.2.16)
satisfying
for
(9.2.18)
O.
(9.2.19)
and
(a(t)7r 2 (t)
Since
to
-~-,--,-ds
a(s)7r 2(s)
- - - -+
7r(t) 7r(t o)
00
as
t -+
00,
the above arguments in (i) apply to equations (9.2.18) and (9.2.19), and
the desired conclusion follows immediately. This completes the proof.
t::::
(It) If condition (9.2.5) holds, a solution x(t) of (9.2.3) with the property
limHOOb(t)x(t)jR[t,toJ = constant> 0 (respectively, limHoob(t)x(t) =
Chapter 9
636
(i) Let condition (9.2.5) hold. If equation (9.2.3) has a solution x(t) such
that limHoo b(t)x(t) = constant> 0, then (9.2.4) has a solution y(t)
such that y(t) :s; x(t) for all large t and limHoo b(t)y(t) = constant> O.
(ii) Let condition (9.2.6) hold. If equation (9.2.3) has a solution x(t) such
that limHoo b(t)x(t)j7f(t) = constant> 0, then (9.2.4) has a solution
y(t) such that y(t):s; x(t) for all large t and limHoo b(t)y(t)j7f(t) =
constant> O.
Proof. It suffices to prove the statement (i). Let x(t) be a positive
solution on [to,oo) of (9.2.3) such that limHoo b(t)x(t) = c > 0, where
c is a constant. Put w(t) = x(t) - 'T71 (t). As in the proof of Theorem
9.2.1, we see that w(t) is eventually of constant sign, say, for t 2: t1 2: to.
If w(t) < 0, t 2: tl then x(t) < 7/1(t), t 2: t1 which implies that
liminfHoob(t)'T71(t) 2: c> 0, a contradiction to condition (9.2.9). Thus,
we have w(t) > 0 for t 2: tl' Proceeding as in the proof of Theorem 9.2.1(i)
and noting that limHoo b(t)w(t) = C, C2 = limHoo a(t)(b(t)w(t))' = 0,
we conclude that there exists a continuous solution y(t) of the integral
equation
1
b(t)y(t) = -Cl
2
+ b(t)'T72(t) +
it -(-) 1
t2
1
ass
00
F2(u, y(u))duds
b(t)
Inll -R[j'T71(t)
t--+oo t, to
0,
b(t)
11m -R[j'T72(t)
t--+oo t, to
(9.2.20)
637
Miscellaneous topics
when condition (9.2.5) holds, and
(9.2.21)
XJ
e(t).
Corollary 9.2.1. Suppose there exists a function 'r/(t) E C([t o, 00), JR)
with the properties
(a(t)(b(t)'r/(t))')'
lim b(t)'r/(t)
t400
+ e(t)
0,
t:::: to
(9.2.22)
(9.2.23)
Chapter 9
638
b(t)
11m -(-)TJ(t)
t-)OO
n t
..
(9.2.24)
If equation (9.2.2) has an eventually positive solution x(t), then (9.2.1) has
an eventually positive solution y(t) such that y(t) ::; x(t) for all large t.
Corollary 9.2.2. Suppose there exists a functionf)(t) E C([to, 00), JR)
satisfying condition (9.2.22) and either condition (9.2.23) or (9.2.24).
(i) Let condition (9.2.5) hold. If equation (9.2.2) has a solution x(t) such
that limt-400 b(t)x(t) = constant> 0, then (9.2.1) has a solution y(t)
such that y(t) ::; x(t) for all large t and limHoo b(t)y(t) = constant> O.
(ii) Let condition (9.2.6) hold. If equation (9.2.2) has a solution x(t) such
that limHoo b(t)x(t)/n(t) = constant> 0, then (9.2.1) has a solution
y(t) such that y(t)::; x(t) for all large t and limHoo b(t)y(t)(rr(t) =
constant> O.
Corollary 9.2.3. Suppose there exists a function TJ(t) E C([to, 00), JR)
satisfying condition (9.2.22) and
t-400
lim b(t)T](t)
t~oo
(i) Let condition (9.2.5) hold. If equation (9.2.2) has a solution x(t) such
that limHoo b(t)x(t)/ R[t, to] = constant> 0, then (9.2.1) has a solution
y(t) such that y(t)::; x(t) for all large t and lill1Hoo b(t)y(t)/ R[t, to] =
constant> O.
(ii) Let condition (9.2.6) hold. If equation (9.2.2) has a solution x(t) such
that limHCXl b(t)x(t) = constant> 0, then (9.2.1) has a solution y(t)
such that y(t) ::; x(t) for all large t and limHoo b(t)y(t) = constant> O.
Remark 9.2.2. In the above results we were concerned only with eventually positive solutions of the equations under consideration. It is clear that
these results have counterparts for eventually negative solutions provided
F i , F E C ( [to, 00) x JR - , JR - ), i = 1, 2.
(a(t)x'(t))'
+ q(t)f(x(t))
e(t).
(9.3.1)
Miscellaneous topics
639
We will say that equation (9.3.1) is of limit circle type if all its solutions
satisfy
I:)Q x(u)f(x(udu
<
(9.3.2)
00.
J;
2(n+1),
(9.3.3)
(9.3.4)
and
(9.3.5)
= 1/(2(n + 1
1.
s =
it
a-(3(u)qC> (u)du
and
and
= x(t)
(9.3.6)
(. = d/ds)
(9.3.7)
y(s)
to
ii + ap(t)y + Q(t)f(y)
E(t),
y=
z -
-p(t)y
1=
to
[(a(u)q(u']- d
a(u)q(u)
u <
00,
(9.3.9)
640
Chapter 9
rOO I [
lto
(a(u)q(u' ]'
aa(u)q<>+l(u) -
00
. to
le(u)1
d
U <
(a( v,)q( u))a
~..,-:.--:.--'-:-:--
00,
(9.3.10)
(9.3.11)
00
and
(9.3.12)
Then any solution x(t) of (9.3.1) satisfies (9.3.2) and
Proof.
(~ -
V(s)
+ Q(t)F(y),
-Q(t)p(t)
(~ -
then
(~Yf(Y) + Q(t)F(y)
1 ] [(a(t)q(t)'] (a!3(t))
+Q(t) [ (f3 - o:)F(y) - kyf(y)
a(t)q(t)
qa(t)
V(s)
[B - 2
k:
+E(t)z+ [p(t)-
qa(t)
(~-0:)p2(t)] (~yz).
Iyzl
111
~ 2,y2 + 2,Z2 ~ 2,z2
+ AF(y) + K,
so
V(s) < [B - 2
-
+IE(t)1
Since pet)
(~- 0:)]
k
([(a(t)q(t),l-) (a!3(t).) V
a(t)q(t)
qa(t)
= p(t)a!3(t)q-a(t),
let)
Ip(t) Ip'(t) -
we have
0:) p2(t) I
(~ - 0:) p2(t)q<>(t)a- i3 (t)i a!3(t)q-<>(t).
(~ -
Miscellaneous topics
641
Let T(S) denote the inverse function of s(t), we obtain by (9.3.10) that
converges. Similarly,
So
IE[T(~)lld~ =
It
cy,
IE(u)lq (u)a
to
-(3
It
(u)du -
to
le(u)1
(a(u)q(un du
= (a(t)q(t(3-a F(y(s
(a(t)q(t(3-a F(x(t
~ Kl
for some constant Kl 2: 0, and now the conclusion of the theorem follows
from conditions (9.3.4) and (9.3.12).
x"(t)
+ q(t)x'Y(t)
0,
(9.3.13)
1I
00
to
and
1I
00
where ()
(2n
q"(U)
[q'(u)j2 Idu <
- - (}--qO ( u)
q>' (u)
(u
+ 3)/(2(n + 1
and A = (4n
roo q-n/(n+l)(u)du
ito
(9.3.14)
00
+ 5)/(2(n + 1.
<
00,
(9.3.15)
00,
If
(9.3.16)
642
Chapter 9
+ 1)
o.
and B =
Clearly,
00.
> 0, N
x 2 < Mxf(x)
1=
and
0, m
+ N,
Iq(t)
e(t) I
l=le(u)ldU
q(u)
< c,
-
to
<
>0
(9.3.17)
(9.3.18)
00
(9.3.19)
00.
to
If x(t) is a limit circle solution of equation (9.3.1), i.e., (9.3.2) holds, then
= a(u),
to
- () [x (u)] du <
q
(9.3.20)
00.
a(t),(
a(t1),()
()
() x t)x ()
t - -(-)
X t1 x t1
q t
q t1
it
t,
x(u)f(x(u))du
-it
it
a(u) '()'() ()
~() q u x u X u du
t, q
t, q U
it
e((u)) x(u)du.
t, q
(9.3.21)
643
Miscellaneous topics
By the Schwarz inequality, we have
<
::; m 2 M x(t)f(x(t))
+ N ;~tl) [q'(t)f,
roo q~(u)
[q'(uWx2(u)du
(u)
itl
::; Kl <
00,
Kl is a constant.
Since
<
Ix (t )I e(t)
() _
qt
e(t) 2()
e(t) < ~'1' ()f( ())
2q (t ) x t + 2q ()
t _ 2 lVj X t x t +
(N+1) qe(t)t
2
( )'
(tx'(t))'
+ t 2 x(t)
t2
+t
for
2: 1
IS
(9.3.22)
644
Chapter 9
and equation (9.3.1) is limit circle. Then every solution of equation (9.3.1)
either oscillates or converges monotonically to zero as t --+ 00.
Let x(t) be a nonoscillatory solution of equation (9.3.1), say,
x(t) > 0 for t:;:, tl :;:, to. Clearly, liminft._HXl x(t) = 0 for otherwise
condition (9.3.2) will be violated. If x(t) is eventually monotonic, we are
done, so we assume that x(t) is Hot eventuR.lly monotonic. If x(t) does
Proof.
L= (1:' atv)
a t
It
t3
~K1
[e(u) ~ q(u)f(x(u))]du.
(9.3.23)
x(t)
X(t3)
+ (ft __
(1 )d1L) ft[e(u)
13
~ ft
t3
au
/3
~ q(u)f(x(u))]du
1~
If t4 > h is any zero of x'(t), then (9.3.23) shows that the first integral
above vanishes at t = t 4 , and then we have
i.e.,
t :;:, t3,
Next, we state two results for the oscillation of unforced and forced
equations, respectively.
Theorem 9.3.5.
If in addition to the hypotheses of Theorem 9.3.4,
e(t) = 0 and f= ds/a(s) = 00, then all solutions of equation (9.3.1)
are oscillatory.
Theorem 9.3.6. Suppose the hypotheses of Theorem 9.3.4 hold except
possibly condition (9.3.22). If for every constant c > 0 and all large
T:;:, to,
Miscellaneous topics
and
645
00,
h(t) = [(a(t)q(t))']a(t)q(t)
+ le(t)1
and H(t) =
[(a(t)q(t))']+
a(t)q(t)
le(t)1
+ (a(t)q(t))l/2
(9.3.24)
and either
L= !~:~
1~ le(u)1 (exp
or
(h)
1~ exp (-
L
1:'
exp ( -
L
U
h(V)dV) du =
(9.3.25)
h(V)dV) du <
H(V)dV) du =
00,
00,
00,
r= le(u)1 ( r
)
ltD (a(u)q(u))I/2
ltD H(v)dv du <
exp
(9.3.26)
00.
Then equation (9.3.1) is oflimit point type, i.e., there is a solution of (9.3.1)
which does not satisfy condition (9.3.2).
{ :; :
(9.3.27)
Then,
646
Chapter 9
Now, we have
so it follows that
V'es)
a(t)q(t)
(9.3.28)
(1:
00.
Let (x(t), yet)) be a solution of system (9.3.27) such that (x(to), y(to)) =
(xo, yo) and V(xo, Yo, to) = Veto) > KI + L Then from (9.3.28), we get
(V(t)ex p
(1:
h(U)dU)), ;:: -
~le(t)lexp
(1:
h(U)dU).
Vet) exp
(1:
h(U)dU)
Hence,
In view of Theorem 9.3.3 this shows that x(t) cannot be a limit circle
solution of equation (9.3.1).
If (1 2 ) holds, we define
y2
a(t)-()
2q t
+ F(x).
Then, we have
e(t) I [
le(t)1
]
Iq(t)Y
(a(t)q(t))l/2
we find
V{(t)
+ H(t)Vi(t)
;::
(
1
VI
1)
+"2 '
le(t)1
-"2 (a(t)q(t))l/2
Miscellaneous topics
647
The rest of the proof is similar to the case (II) and hence omitted.
In our next limit point theorem we shall need the following lemma.
Lemma 9.3.1.
Suppose there exist constants
> 0 such that
Tnl
x 2 :s; Mxf(x)+N,
(a(t)q(t))' I
Ia 1/2(t)q3/2(t)
:s; ml and
(9.3.29)
roo [(a(u)q(u))']2 du
lto
a(u)q3(u)
<
00.
(9.3.30)
lto
a(u)q3(u)
<
00.
lto
a(u)q3(u)
< miM
-
+
Theorem 9.3.8.
Bl ::::: 0 such that
kl
00
to
roo x(u)f(x(u))du
lto
[(a(u)q(u))']2d
u <
a(u)q3(u)
00.
(9.3.33)
648
Chapter 9
n) aQ(t)qQ+l(t)
[(a(t)q(t))'](v + ~)
-Ip(t) - (~ - n) p2(t)1 ~ [v + K+ A~]
kl
kl
Q(t)
V(s) ?: -
(~ -
-Q(t)
> _
-
-IE(t)1
z2
kl
[(f3 -
n)F(y) _
a(t)q(t)
kl
qQ(t)
-Ip(t) -
G(t)
+ le(t)1
{ [Bl + 2 (~kl _ n)] [(a(t)q(t))']a(t)q(t)
(a(t)q(t))Q
'
+ Ip (t) -
( kl1
- n
qQ (t) I [
1 ] 1 } ai3 (t)
(t) af3(t) 1 + AQ(t) kl qQ(t)
and
K I'
g(t) = kl P (t) so that V(s)
1
kl -
n) p
+ G(t)V(s) ?: -g(t),
qQ(t) I
le(t)1
af3(t)
(t) af3(t) + 2(a(t)q(t))Q qQ(t) ,
and hence
?: - g(t) exp
(1: G[T(~)]d~)
?: - Klg(t)
(fa
(9.3.34)
G[T(~)]d~) :S
since conditions (9.3.9) - (9.3.11) guarantee that exp
Kl < 00 for some constant Kl > O. In view of (9.3.10) and (9.3.11), we
have Kl
g[T(~)]d~ :S K2 < 00 for some constant K2 > O. Now let
x(t) be any solution of (9.3.1) such that V(y(so), z(so), so) > K2 + l.
Integrating (9.3.34) from So to s, we find
Is";
V(s) exp
(1: G[T(~)]d~)
?: V(So) - K2 > 1
and thus V(s)?: 1/ Kl for s;::: so. Now dividing this inequality by
(a(t)q(t))i3- Q and rewriting the left-hand side in terms of t, we get
a(t) (x'(t))2
2q(t)
2k 2a(t)q3(t)
>
;1
(a(t)q(t))Q-i3.
(9.3.35)
649
Miscellaneous topics
If x(t) was a limit circle solution of equation (9.3.1), then since conditions (9.3.5) and (9.3.24) imply (9.3.29), in view of Theorem 9.3.3, we have
ftC;: (a(u)jq(u)) (x'(u))2du < 00. Also, by Lemma 8.3.1
(X)
:::;
(a(;;r~~))' x(u)x'(u)dul
[r=
lto
(X)
and by condition (9.3.24) and the assumption that x(t) was a limit circle
solution ItC;: F(x(u))du < 00. Now integrating (9.3.35) we get the desired
contradiction.
(a(t)x'(t))'
+ q(t)X 2n - 1 (t) =
e(t),
(9.3.36)
.c
00.
< 00.
x"(t) + q(t)x2n-l(t)
(9.3.37)
r= [q'(u)]- du <
ltD
q( u)
00,
OX) [q'(u)F
~(
) du <
to
00,
650
Chapter 9
Jto
Example 9.3.1.
00.
(9.3.38)
0,
(9.3.39)
xl/(t)
+ t":C 2n -
1 (t)
> 1 + (lin).
(9.3.40)
It is easy to check that all the hypotheses of Corollary 9.3.2 are satisfied,
and hence condition (9.3.40) is a necessary and sufficient condition for all
solutions of equation (9.3.39) to belong to 2n[to,00).
(a(t)7/;(X(t))X'(t))'
+ q(t)f(x(t))
e(t),
(9.4.1)
x//(t)
+ q(t)f(x(t))
e(t).
(9.4.2)
We define h: JR ---+ JR by
h(x)
fox 7j;(u)du.
(9.4.3)
Miscellaneous topics
651
Corollary 9.4.1. Every oscillatory solution of (9.4.1) generates an oscillatory solution of equation (9.4.5).
Corollary 9.4.2. There is one-to-one correspondence between the nonoscillatory solutions of equations (9.4.1) and (9.4.5).
Corollary 9.4.3. Suppose e(t) = 0 and 1jJ(0) = O. If the solution
z(t) of (9.4.5) with z(to) = z'(t o) = 0 for every to?> 0 is unique, then
equation (9.4.1) has no nontrivial oscillatory solutions. If in addition (9.4.5)
is oscillatory, then equation (9.4.1) has no nontrivial continuable solutions.
Proof. Suppose x(t) is a nontrivial solution of (9.4.1), then by Theorem
9.4.1 and (9.4.3), z(t) = h(x(t)) is a nontrivial solution of equation (9.4.5)
such that
z'(t) = ljJ(X(t))x'(t).
(9.4.6)
Example 9.4.1.
(9.4.7)
and let x(t) be its solution, then z(t) = x 3 (t)/3 is a solution of the linear
equation z"(t) + z(t) = O. Thus, z(t) = Asin(t + 0) for some constants
A and 0, and hence x(t) = Csin 1 / 3 (t + 0) where C = (3A)1/3. As
x'(t) does not exist for t = br - 0, k = 1,2"" equation (9.4.7) has no
nontrivial continuable solution.
Chapter 9
652
Example 9.4.2.
where
X 2n + 1
/(2n
+ 1)
z"(t)
+ (2n + 1)
C:)
z(t)
h(x)
Here,
(9.4.9)
0.
k:::
(a(t)1j;(X(t))X'(t))'
for x
+ q(t)g(z(t))
0,
(a(t)z'(t))'
(9.4.10)
Ie 0.
The transforma-
+ q(t)f(x(t)) =
(9.4.11)
= R(t) -
it
to
du
a(u)
(9.4.12)
y + R*(s)Q(s)g(y) =
where R*(s)
0,
(-
= djds)
(9.4.13)
Jo+
where F(x) =
1j;(u) du
JF(u)
00
foX 1j;(u)f(u)du.
and
Jo-
1f;(u) du = _
JF(u)
00
Miscellaneous topics
653
Theorem 9.4.3. Suppose 'ljJ(x) > 0 for x -=J- 0, ft~o 'ljJ(u)du = oo,
and q(t) < 0 on [h, t 2 ], tl ;:::: O. Then equation (9.4.10) has a solution
x(t) such that limt-tT Ix(t)1 = 00 for some T E (tl, t2) if and only if
00
1j;(U)
d
'u <
Jl+F(u)
---r===~;:==:=
where F(x)
00
or
io
,'/1,
> -
00,
= fox'ljJ(u)f(u)du.
f'(x);:::: 0 for
for
x-=J-O,
(9.4.14)
00.
(9.4.15)
The next theorem describes the behavior of oscillatory solutions of equation (9.4.10) when 1/;(0) = O.
Theorem 9.4.6.
Suppose x(t) is a solution of equation (9.4.10) on
[tl' t2J such that x(t l ) = X(t2) = O. If 'ljJ(0) = 0 and q(t) does not
change sign on [tl: t2], then x(t) = 0 on [tl: t2J.
Proof. Suppose there exists t3 E (tl' t2) such that x( t 3) -=J- 0, then
there exist T I , T2 E [tl' t2J such that x(TIl = x(T2) = 0 and x(t) 1= 0
on (TI ,T2). Integrating equation (9.4.10) from TI to T 2, we obtain
T2
iT!
(a(s)'ljJ(x(s))x'(s))'ds
T2
iT!
q(s)f(x(s))ds = o.
As x(Td = x(T2) = 0 and 'ljJ(0) = 0, the first integral is zero, and hence
q(s)f(x(s))ds = o. As the integrand is of one sign and a continuous
function of t, it follows that q(t)f(x(t)) = 0 for all t E [TI ,T2], and
hence x(t) == 0 on [TI , T2J, which is a contradiction.
{i,'
Corollary 9.4.4. Suppose 'ljJ(0) = 0 and q(t) does not change sign.
Then the only oscillatory solution of equation (9.4.10) is the eventually
identically zero solution.
654
Chapter 9
9.6. References
1. F.V. Atkinson, Nonlinear extensions of limit point criteria, Math. Z.
130(1973), 297-312.
2. F.V. Atkinson, On second order differential inequalities, Proc. Roy. Soc.
Edinburgh Sec. A 72(1974), 109-127.
3. R. Bellman, Stability Theory of Differential Equations, McGraw Hill,
New York, 1953.
4. J. Burlak, On the nonexistence of Lz-solutions of nonlinear differential
equations, Proc. Edinburgh Math. Soc. 14(1965), 257-268.
5. T.A. Burton and R. Grimmer, On the asymptotic behavior of solutions
of x" + a(t)f(x) = e(t), Pacific J. Math. 41(1972), 43-55.
6. T.A. Burton and W.T. Patula, Limit circle results for second order
equations, Monatsh. Math. 81(1976), 185-194.
7. T.A. Chanturija, Some comparison theorems for higher order ordinary
differential equations, Bull. Acad. Polan. Sci. Ser. Sci. Math. Astronom.
Phys. 25(1977), 749-756 (in Russian).
8. S.R. Grace, Oscillation theorems for nonlinear differential equations of
second order, J. Math. Anal. Appl. 171(1992), 220-241.
9. S.R. Grace and B.S. Lalli, Integral averaging techniques for the oscillation of second order nonlinear differential equations, J. Math. Anal. Appl.
149(1990), 277-311.
10. S.R. Grace, B.S. Lalli and C.C. Yeh, Oscillation theorems for nonlinear second order differential equations with a nonlinear damping term,
SIAM J. Math. Anal. 15(1984), 1082-1093.
11. J.R. Graef, Limit circle criteria and related properties for nonlinear equations, J. Differential Equations 35(1980), 319-338.
Miscellaneous topics
655
12. J.R. Graef and P.W. Spikes, Sufficient conditions for nonoscillation of
a second order nonlinear differential equation, Proc. Amer. l\lath. Soc.
50(1975), 289-292.
13. J.R. Graef and P.W. Spikes, Comparison and nonoscillation results
for perturbed nonlinear differential equations, Ann. Mat. Pura Appl.
116(1978), 135-142.
14. D. Hinton, Limit point - Limit circle criteria for (py')' + qy = >"ky, in
Or'dina'ry and Partial Differential Equations, Lecture Notes in Math. 415,
Springer-Verlag, New York, 1974, 173-183.
15. A.G. Kartsatos, On the maintenance of oscillation of nth order equations
under the effect of small forcing term, J. Differential Equations 10(1971),
355-363.
16. R.M. Kauffman, T.T Read and A. Zettl, The Deficiency Index Problem for Powers of Ordinary Differential Expressions, Lecture Notes in
Math. 621, Springer-Verlag, New York, 1977.
17. N. Kawano, T. Kusano and M. Naito, Nonoscillatory solutions of
forced differential equations of the second order, J. Math. Anal. Appl.
90(1982), 323-342.
18. I. Knowles, On a limit-circle criterion for second order differential operators, Quart. J. Math. 24(1973), 451-455.
19. V. Komkov, A note on a generalization of the Sturm-Picone theorem,
Colloq. Math. XXXIX(1978), 173-176.
20. K. Kreith, Oscillation Theory, Lecture Notes in Math. 324, SpringerVerlag, New York, 1973.
21. W. Leighton, Comparison theorems for linear differential equations of
second order, Proc. Amer. Math. Soc. 13(1962), 603-610.
22. W.E. Mahfoud and S.M. Rankin, Some properties of solutions of
(r(t)'lI(x)x')' + a(t)f(x) = 0, SIAM J. Math. Anal. 10(1979), 49--54.
23. W.T. Patula and P. Waltman, Limit point classification of second order
linear differential equations, J. London Math. Soc. 8(1974), 209--216.
24. W.T. Patula and J.S.W. Wong, An LP analog of the Weyl alternative,
Math. Ann. 197(1972), 9-28.
25. P.W. Spikes, On the integrability of solutions of perturbed nonlinear
differential equations, Proc. Roy. Soc. Edinburgh Sec. A 77(1977), 309318.
26. C.A. Swanson, Comparison and Oscillation Theory of Linear Differential
Equations, Academic Press, New York, 1968.
27. C.A. Swanson, Picone's identity, Rend. Mat. 8(1975), 373-397.
28. H. Weyl, Uber gew6hnliche Differentialgleichungen mit Singularitaten
und die augeh6rige Entwicklung willkiirlicher Funktionen, Math. Ann.
68(1910), 220--269.
29. J.S.W. Wong, Remarks on the limit--circle classification of second order
differential operators, Quart. J. Math. 24(1973), 423-425.
30. J.S.W. Wong and A. Zettl, On the limit point classification of second
order differential equations, Math. Z. 132(1973), 297-304.
Chapter 10
Nonoscillation Theory
for Multivalued Differential
Equations
10.0. Introduction
In our previous chapters, we have presented several nonoscillation criteria for second order differential equations. In the present chapter, we shall
introduce nonoscillatory theory for second order differential and neutral inclusions. Our results rely on fixed point theorems for multivalued maps,
and on a compactness criterion.
10.1. Preliminaries
In this chapter, we shall provide nonoscillatory results for the second
order differential inclusions
(a(t)x'(t))' E e(t)
+ F(t, x(t)),
2: to 2: 0
(10.1.1)
:t (a(t)
:t
(x(t)
+ px[t -
T]))
F(t, x(t)),
2: to 2: o.
(10.1.2)
657
We shall also need the followillg compactness criterion in B [T, (0) (the
Banach space of all continuous, bounded real valued functions 011 [T, (0)
endowed with the usual supremum norm, i.e., Ilull oo = snPtE[T,oo) lu(t)1
for u E B[T, (0).)
Theorem 10.1.3 [7]. Let E be all equicontinuous and uniformly bounded
subset of the Banach space B [T, (0). If E is equiconvergent at 00, then
it is relatively compact.
this we mean
(a). for each measurable u: [to, (0) ---+ lR the map t
measurable single valued selections,
(b). for a.e. t E [to, (0) the map u
(c). for each r > 0 there exists hr E L1 [to, (0) with IF(t, u)1 ::; hr(t)
for a.e. t E [to, (0) and all u E lR with lui::; r; here IF(t, u)1 =
sup{lzl: z E F(t, u)},
(v)
-00.
Remark 10.2.1. In (iv), part (a) could be replaced by: the map t
F(t,u) is measurable for all u E JR..
(3 > -
inf
tE [to,oo)
TJ(t),
(10.2.1)
Chapter 10
658
and let d> 0 be such that
f3 + inf
'T)(t) 2 d.
tE[to,oo)
(10.2.2)
J 100
and
OO
M> d with
there exists
-1()
ass
(10.2.3)
(10.2.4)
t~to
sup
wE[d,Mj
(10.2.5)
t---+oo
f3 and
t 2 T
t---+oo
rOO _(1) 00
iT
ass
sup
wE[d,Mj
!F(t,w)ldtds::; M -
Q =
{xEB[T,oo):
[f3+ sup
'T/(t)].
t~T
d::;x(t)::;M
for
(10.2.6)
and
t2T}.
N x(t)
1
= f3 + 'T/(t) + itrOO a(s)
1
s
00
F(v, x( v))dvds
for t 2 T.
The Proposition 1.1 in [6, pp. 777] guarantees that N: Q -+ C(E); here
C(E) denotes the family of nonempty, convex subsets of E. We shall first
show that
(10.2.7)
N: Q -+ C(Q).
For notational purposes for any x E Q let
F(x)
= {UELl[T,oo): u(t)EF(t,x(t))
fora.e. tE[T,oo)}.
659
w(t) = f3 + ry(t)
1
a(s)
s
+ it
rOO
00
T(v)dvds
for t 2 T.
w(t)
1 1
00
-1
00
a( s)
sup
wE [d,MJ
!FCt', w)ldvds
t?T
w(t) 2 f3
+ ry(t) 2
f3 + inf ry(t) 2 d.
t?to
C( Q)
(10.2.8)
is a compact map.
For this, we will use Theorem 10.1.3. For x E Q let G(x) = N x - ry.
Take any x E Q and wE G (x). Then there exists T E F(x) such that
w(t) = f3
+ it(=
a(s)
00
T(v)dvds
for t 2 T.
Ilwll=
sup
tE[T,=)
for each x E Q.
Iw(t)-f31 :S
1 1
00
1
-()
00
ass
sup
wE[d,MJ
IF(v,w)ldvds
(10.2.9)
IW(t2) - w(tl)1 :S
t2
t,
00
1
sup !F(v,w)ldvds
a s s wE[d,MJ
- ()
Chapter 10
660
N : Q -+ CK(Q)
(10.2.10)
From (10.2.10) and [4, pp. 465] it is sufficient to show that the graph of
E giN) with (J.Ln, :r,,) -+ (J.L, :r);
here n E lNl = {1, 2, .}. We must show J.L E N x. Fix t E [T,oo).
Note that IlxilCXl S NI, Ilxnlloo S NI for n E lN 1 , since x, Xn E Q for
n E lN l . Also there exists Zn E F(xn) with
J.Ln(t)
(3
tOO
1
a(s)
+ "l(t) + Jt
tOO
Js
zn(v)dvds
J.L(t) = (3 + "l(t)
(10.2.11)
F(x) with
00
1 (=
a(s)
u(v)dvds.
Js
Notice (iv) guarantees that there exists a hM ELI [T, (0) with IZn(s)1 S
hl\1(s) for a.e. s E [T,oo). Consider {zn}nElN, Take k E lNl and
k > t. From (10.2.11), we have
IJ.Ln(t) -
I lv
(3 - '(I(t) - .
I"
Zn(V)
a(s) dsdv
I S
Jk{= hM(V)
lv
t
a(s) dsdv.
(10.2.12)
Now, a standard result from the literature [14, Proposition 1.4] guarantees that Fk : B[T, k] = CrT, k] -+ Ll [T, k] is upper semicontinuous
with respect to the weak topology (w u.s.c.) and also weakly completely
continuous; here Fk is given by
Fk(W) = {u
00
lv
- ()dsdv.
k
t a s
(10.2.14)
Similarly, we can show that there exists Uk+1 E Ll [T, k + 1] and a subsequence of S, say SI, with Zn --' Uk+l in Ll [T, k + 1] as n..--t 00 in
SI and with Uk+l E Fk+l (x). Of course this implies Zn --' Uk+l in
Ll[T,k] as n..--too in SI so Uk+l(V)=udv) fora.e. VE[T,k]. In
addition note that IUk+l(v)l:::; hM(V) for a.e. v E [T, k + 1]. Continue
and construct Uk+2, Uk+3,. For 1 E {k, k + I,} = 1Nk let uHv) be
any extension to [T,oo) of Uz with /ui(v)1 ~ hM(v) for a.e. v E (l,oo).
Also, let
~
Ft(w)
{z
L l [T,oo):
hM(V)
661
a~) dSdVI ~
00
hM(V)
a(s) dsdv.
(10.2.15)
F (x)
n Ft
lEK
(x)
(here K
(10.2.16)
(and :F (w) is nonempty, closed and convex). Note first that Ilxll oo :::; M
so IF(v, x(v))1 :::; hM(V) for a.e. v E [T,oo). Let Xk be the restriction
to the interval [T, k], k E 1N l , of x. Note [10,11 or 12] that
h (Xk) = {z
:Fk (Xk) = {z
L1[T, 00) :
It is immediate that :Fk (Xk) is a closed set in L1 [T, 00) for each k E K.
Let
Rk
{z E 1[T,oo):
Cbapter 10
662
and notice it is clear that :Fio (x) = :Fio (Xk) EEl Rk. It is clear that
:Fio (x) is a closed set in L1 [T, 00 ). Also, for each k E K, we have
:F (x) <;;; :Fio (x), and so :F (x) <;;; nCEK :Ft (x). On the other hand if
Z E :Fe' (x) for each IE K, then z(v) E F(v, x(v)) for a.e. v E [T, (0),
and so nCEK Ft (x) <;;; F (x). Thus, (10.2.16) holds and a.lso F (x) is a
closed subset of LI[T, (0). Now, since u belongs to nCEK Fc* (x) (note
for each I E K that u E Ft (x)), we have u E F (x). Let k -+ 00 in
(10.2.15), to obtain
J1(t)-/3-7](t)-
1
+1
t
and hence
J1(t) = /3 + 7](t)
00
u(v)
IV a(s)dsdv
1
00
a(s)
roo
is
u(l')dvds.
Consequently, Q(N) is closed, so N: Q -+ C(Q) is a upper semicontinuous map (see [4, pp. 465]); in fact (10.2.8) guarantees that N: Q -+
CK(Q).
Theorem 10.1.1 now guarantees that there exists x E Q with x E N x,
i.e., for every t 2: T, we have
x(t) E /3 + 7](t)
and so
(a(t)x'(t))'
In addition
<
e(t)
00
roo
F(v, x(v))dvds,
+ F(t,x(t))
for a.e. t 2: T.
a(s)
1 1
00
-1
a(s)
-+0
and
is
00
sup
wE[d,MJ
00
sup
wE[d,MJ
IF(v, w)ldvds
ast-+oo
IF(v, w)ldv -+
as
t -+
00.
We shall next show that an existence result for (10.1.1) can be obtained
if (10.2.3) is replaced by
(10.2.17)
Theorem 10.2.2.
Suppose (i) - (v) and (10.2.17) hold, and let /3
(respectively d) be chosen as in (10.2.1) (respectively (10.2.2)). Also,
assume that the following two conditions are satisfied
there exist
K >1
and
d
M > K
with
663
and
OG
o.(s)
(XJ
sup
Js
wEld/K,MI
(10.2.19)
oo
0.(8)
[00
Js
W(t,w)ldtd8::;
sup
{xEB[T,oo):
(K;;
wEld/K,MI
=
(E[T, 00),
1) d.
II . 1100)
(10.2.20)
and
N: Q
-7
(10.2.21 )
C(Q).
Let x E Q and take wEN x. Then there exists T E F(x) such that
1 1
J= 1=
w(t) = (3 + ry(t) +
If t::;> T,
w(t)
we also have
>
(3
+ ry(t) -
00
00
ass
-1
a(s)
-1(
.)
for
sup
W(t, w)ldtd8
::;> d-
(K-1)
- - d = Kd
WEld/K,MI
- d
(-K-1)
K
T(1,)dl,d8
As a result w(t)::;> dJK for t::;> T for each wEN x. On the other
hand for t 2.'- T, we have w(t)::; (3 + ry(t) ::; (3 + SUPt>T ry(t)::; M. Thus,
w(t) ::; M for t ::;> T for each '11' E N x. Hence, (10.2.21) holds. Essentially
the same reasoning as in Theorem 10.2.1 guarantees that N: Q -7 CK(Q)
is a upper semicontinuous, compact map. Now, Theorem 10.1.1 proves the
result.
In our next result, we will show that the assumption (10.2.17) (respectively, (10.2.3)) in Theorem 10.2.2 (respectively, Theorem 10.2.1) can be
removed if we combine the analysis of both the theorems. In this case
(10.2.18) has to be adjusted slightly.
Theorem 10.2.3. Suppose (i) ~ (v) hold, and let (3 (respectively d)
be chosen as in (10.2.1) (respectively, (10.2.2)). Also, assume that the
following two conditions are satisfied
there exist
K > 1 and
M>K
with
664
Chapter 10
and
oo
a( s)
/00
S
sup
wE Id/ K,Mj
(10.2.23)
roo
iT
00
a(s)
sup
WEld/ K,Mj
IF(t, 'W)ldtds
(10.2.24)
(a(t)(b(t)x(t))')'
e(t)
+ F(t, x(t)),
t 2: to 2: o.
F:[to,oo) x (0,00) -,
Ipl =I
and
there exists K
CK((~oo,O])
1
~
a(s)
00
(10.3.1)
(10.3.2)
J 1
oo
sup
wE[K/2,Kj
665
Ipi <
Case (I).
(= 1
JT
Let TI
I . 11(0)
a(s)
=T -
Ipi <
1. Choose T 2: to so thFlt t -
1=
Ipl >
1 separately.
( 10.3.4)
= (B[TI' (0),
T.
1 and
sup
wE[K/2,K]
and
_
-
4'(1
and
N2 x(t)
4'(1
+ p)K -
px[T - T],
+ p)K -
px[t - T],
r 1=
{0 } ,
TI
- JT
a(s)
-s: t -s: T
F(v,x(v))dl'ds,
t 2: T.
F(x)
{u
E [TI' oo)}.
Let x E Q so that K/2 -s: x(t) -s: K for t E [TI' (0), and take wE N2 x.
Then there exists TO E F( x) with
w(t) =
r a(s) 1= To(v)dV ds ,
{O},
TI
- JT
-s: t -s: T
s
t;:: T
666
Chapter 10
- 1<
< O.
Subcase (1).
Nl x(t)
+ wet)
4(1
+ p)K -
px[T - T]
~ ~(l+P)K-PK
(~-~P)K
>
and
+ wet)
+ p)K -
+ 4(1 -
:s;
4(1
<
3
K
1
4(1+p)K- P2+4(1-p)K
px[t - T]
p)K
= K.
Tl for each w E N2 x.
+ wet)
4(1
>
+ p)K -
px[T - T]
~(1 + p)K _ pK
4
and
If t ~ T,
Nl x(t)
wet)
~ ~(1 + p)K -
px[t - T]
(~+ ~p) K ~ ~,
>
pK +
4(1 + p)K
K.
667
N2 : Q -+ C(E)
(10.3.6)
whereas if t
Q -+ E
Xl,
X2
is a contractive map.
(10.3.7)
T, we have
N: Q -+ CK(Q)
+ p)K -
px[t - T] -
it 1
T
1
-()
ass
00
Q with x E Nl
F(v, x(v))dvds.
roo
iT
> 1. Choose T
1
a(s)
00
~ to so that t -
sup
wE[K/2,KJ
> to for t
~ T
and
3(1 + p)
{4
-p-
Nl x(t) =
K - px[T + T],
P) K - -X[t+T],
1
-3(1+
-4
668
Chapter 10
and
N2 x(t) =
{O},
1
-~
Tl '5:. t '5:. T
jt+r T+T
o.(s)
00
F(v, x(vdvds,
T.
10.5. References
1. R.P. Agarwal, S.R. Grace and D. O'Regan, On nonoscillatory solutions of differential inclusions, Proc. Amer. Math. Soc., to appear.
2. R.P.Agarwal, S.R. Grace and D. O'Regan, Oscillation Theory for
Second Order Dynamic Equations, to appear.
3. R.P. Agarwal, M. Meehan and D. O'Regan, Fixed Point Theory and
Applications, Cambridge University Press, Cambridge, 2001.
4. C.D. Aliprantis and K.C. Border, Infinite Dimensional Analysis, Springer, Berlin, 1994.
5. J.P. Aubin and A. Cellina, Differential Inclusions, Springer, Berlin,
1984.
5. M. Cecchi, M. Marini and P. Zecca, Existence of bounded solutions
for multivalued differential systems, Nonlinear Analysis 9(1985), 775~785.
7. N. Dunford and J. Schwartz, Linear Operators, Interscience, New York,
1958.
8. L.H. Erbe, Q.K. Kong and B.G. Zhang, Oscillation Theory for Functional Differential Equations, Marcel Dekker, New York, 1995.
9. P.M. Fitzpatrick and W.V. Petryshyn, Fixed point theorems for multivalued noncompact acyclic mappings, Pacific Jour. Math. 54(1974),
17~23.
669
Subject Index
Arzela-Ascoli theorem
asymptotic behavior
averaging technique
Banach lattice
Besicovitch almost periodic functions
Besicovitch seminorm
bounded solution
compactness criterion
comparison theorems
conjugate
continuability
converges to zero
damped equation
decaying nonoscillatory solution
differential inclusion
differential system
differential system with forcing term
disconjugate
distance of zeros
Emden-Fowler equation
Euler differential equation
Euler-Lagrange equation
Fitzpatrick-Petryshyn fixed point theorem
forced equation
functional differential equations
functionally commutative
general damped equation
generalized Emden-Fowler equation
generalized Euler equation
generalized Prufer transform
generalized Riccati equation
generalized sine function
Gronwall inequality
half-linear differential equation
11
549,559,562,567,589,608
40
540
107
107
4,5,6,8,10,562
11,657
15
469
2
6,7,8,10
205,376
262
656,657
503
549
467,504
210
36,433,457,493,641
17,23,38,69
94
657
82,205,632
218,237
533
274
559
467
164
209,467,470
164
2
94,444,466,552
Subject Index
Hartman's oscillation criterion
Hille theorem
Hille-Wintner comparison theorem
integrably small
integral averaging technique
interval criteria
interval oscillation
Kummer transformation
Ky-Fan fixed point theorem
Lebesgue dominated convergence theorem
left dis focal
Leighton's oscillation criterion
Levin-type comparison theorem
Liapunov function
Liapunov inequality
Liapunov second method
limit circle type
limit point type
linearization result
maximal solution
minimal solution
mixed type equation
neutral equation
neutral inclusions
nonlinear damping term
nonoscillatory
nonoscillatory characterization
nonprincipal solution
Olech-Opial-Wazewski method
operator-valued differential equation
oscillatory
Picone identity
positive decaying solution
positive increasing solution
positive solution
prepared solution
principal solution
proper solution
quickly oscillatory
Riccati differential equation
Riccati inequality
671
90
601
332,540
70
168
179
64,66
90
656
11
213
90
100
477
105
476
639
639
339,429
342,636,637
342,636,637
257
233
656,664
362
1,492
15
467
457
540
1,492,504
14
595
608
585,614
504
467
113,585
107
501
506
672
Riccati integral equation
Riccati transformation
Riccati variable
right disfoca.l
right-hand derivative
Schauder fixed point theorem
Schaudcr-Tychonov fixed point theorem
self-conjugate
singular Emden-Fowler equation
singular Emden-Fowler system
solution
strongly sublinear
strongly superlinear
Sturm theorem
Sturmian identity
Sturm-Picone theorem
sublinear
superlinear
superlinear forced equation
two-dimensional differential system
unforced equation
variational principle
weighted function
weighted-average oscillation criteria
weighted-integral condition
Willett's oscillation criterion
Wirtinger type inequality
Wronskian
subject Index
22,501,505
466
544
213
477
11
11
504
584,602,606
614,620
1,4
448
274
13,94,99,626
14
15,625,628
376,377,433,498,569
274,296,493,566
358
492
82
467
46
188
45
90
96
14