doi:10.1155/2009/938706
Research Article
Oscillation Criteria for Second-Order Forced Dynamic Equations with Mixed Nonlinearities
Ravi P. Agarwal
1and A. Zafer
21Department of Mathematical Sciences, Florida Institute of Technology, Melbourne, FL 32901, USA
2Department of Mathematics, Middle East Technical University, 06531 Ankara, Turkey
Correspondence should be addressed to A. Zafer,zafer@metu.edu.tr Received 21 January 2009; Accepted 19 April 2009
Recommended by Mariella Cecchi
We obtain new oscillation criteria for second-order forced dynamic equations on time scales containing mixed nonlinearities of the formrtΦαxΔΔ ft, xσ et,t ∈ t0,∞T with ft, x qtΦαx n
i1qitΦβix,Φ∗u |u|∗−1u, where t0,∞Tis a time scale interval with t0∈T, the functionsr, q, qi, e: t0,∞T → Rare right-dense continuous withr >0,σis the forward jump operator,xσt:xσt, andβ1>· · ·> βm> α > βm1>· · ·βn >0. All results obtained are new even forTRandTZ. In the special case whenTRandα1 our theorems reduce toY. G. Sun and J. S. W. Wong, Journal of Mathematical Analysis and Applications. 3372007, 549–560. Therefore, our results in particular extend most of the related existing literature from the continuous case to arbitrary time scale.
Copyrightq2009 R. P. Agarwal and A. Zafer. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
1. Introduction
LetTbe a time scale which is unbounded above andt0 ∈Ta fixed point. For some basic facts on time scale calculus and dynamic equations on time scales, one may consult the excellent texts by Bohner and Peterson 1,2.
We consider the second-order forced nonlinear dynamic equations containing mixed nonlinearities of the form
rtΦαxΔΔ
ft, xσ et, t∈ t0,∞T, 1.1
with
ft, x qtΦαx n
i1
qitΦβix, Φ∗u |u|∗−1u, 1.2
where t0,∞Tdenotes a time scale interval, the functionsr, q, qi, e: t0,∞T → Rare right- dense continuous withr >0,σis the forward jump operator,xσt:xσt, and
β1>· · ·> βm> α > βm1>· · ·βn>0. 1.3
By a proper solution of1.1on t0,∞T we mean a functionx ∈ C1rd t0,∞T which is defined and nontrivial in any neighborhood of infinity and which satisfies 1.1 for all t ∈ t0,∞T, where C1rd t0,∞T denotes the set of right-dense continuously differentiable functions from t0,∞TtoR. As usual, such a solutionxtof1.1is said to be oscillatory if it is neither eventually positive nor eventually negative. The equation is called oscillatory if every proper solution is oscillatory.
In a special case,1.1becomes rtΦαxΔΔ
ctΦβxσ et, 1.4
which is called half-linear forβ α, super-half-linear forβ > α, and sub-half-linear for 0 <
β < α. IfTR,1.4takes the form
rtΦαx
ctΦβx et. 1.5
The oscillation of1.5has been studied by many authors, the interested reader is referred to the seminal books by Doˇsl ´y and ˇReh´ak 3and Agarwal et al. 4,5, where in addition to mainly oscillation theory, the existence, uniqueness, and continuation of solutions are also discussed. In 3, one may also find several results related to the oscillation of 1.4 when TZ, that is, for
ΔrkΦαΔxk ckΦβxk1 ek, 1.6
whereΔis the forward difference operator.
There are several methods in the literature for finding sufficient condition for oscillation of solutions in terms of the functions appearing in the corresponding equation, and almost all such conditions involve integrals or sums on infinite intervals 3–19. The interval oscillation method is different in a sense that the conditions make use of the information of the functions on a union of intervals rather than on an infinite interval. Following El-Sayed 20, many authors have employed this technique in various works 20–30. For instance, Sun et al. 26, Wong 28, and Nasr 25have studied1.5whenα1 andβ≥1, while the case α 1 and 0< β < 1 is taken into account by Sun and Wong in 16. The results in 25,28 have been extended by Sun 27to superlinear delay differential equations of the form
xt ct|xτt|β−1xτt et. 1.7
Further extensions of these results can be found in 30,31, where the authors have studied some related super-half-linear differential equations with delay and advance arguments.
Recently, there have been also numerous papers on second-order forced dynamic equations on time scales, unifying particularly the discrete and continuous cases and
handling many other possibilities. For a sampling of the work done we refer in particular to 6,8,9,12,13,22,32,33and the references cited therein. In 22Anderson and Zafer have extended the above mentioned interval oscillation criteria to second-order forced super-half- linear dynamic equations with delay and advance arguments including
rtΦαxΔtΔ
ctΦβxτt et. 1.8
Our motivation in this study stems from the work contained in 34, where the authors have derived interval criteria for oscillation of second-order differential equations with mixed nonlinearities of the form
xft, x et, t≥t0, 1.9
with
ft, x qtxm
i1
qitΦβix 1.10
by using a Riccati substitution and an inequality of geometric-arithmetic mean type. As it is indicated in 34, further research on the oscillation of equations of mixed type is necessary as such equations arise in mathematical modeling, for example, in the growth of bacteria population with competitive species. We aim to make a contribution in this direction for a class of more general equations on time scales of the form1.1by combining the techniques used in 22,34. Notice that whenα1,rt≡1, andT R,1.1coincides with1.9, and therefore our results provide new interval oscillation criteria even for T R when α /1.
Moreover, for the special case T Z we obtain interval oscillation criteria for difference equations with mixed nonlinearities of the form
ΔrkΦαΔxk qkΦαxk1 n
i1
qikΦβixk1 ek, 1.11
for which almost nothing is available in the literature.
2. Lemmas
We need the following preparatory lemmas. The first two lemmas are given by Wong and Sun as a single lemma 34, Lemma 1forα 1. The proof for the caseα /1 is exactly the same, in fact one only needs to replace the exponentsαibyβi/αin their proof.Lemma 2.3is the well-known Young inequality.
Lemma 2.1. For any givenn-tuple{β1, β2, . . . , βn}satisfying
β1>· · ·> βm> α > βm1>· · ·> βn>0, 2.1
there corresponds ann-tuple{η1, η2, . . . , ηn}such that n
i1
βiηiα, n i1
ηi<1, 0< ηi <1. 2.2
Ifn2 andm1cf. 34for the caseα1one may take
η1 α−β2 1−η0
β1−β2 , η2 β1 1−η0
−α
β1−β2 , 2.3
whereη0is any positive number withβ1η0 < β1−α.
Lemma 2.2. For any givenn-tuple{β1, β2, . . . , βn}satisfying
β1>· · ·> βm> α > βm1>· · ·> βn>0, 2.4
there corresponds ann-tuple{η1, η2, . . . , ηn}such that n
i1
βiηiα, n
i1
ηi1, 0< ηi<1. 2.5
Ifn2 andm1, it turns out that
η1 α−β2
β1−β2, η2 β1−α
β1−β2. 2.6
Lemma 2.3Young’s Inequality. Ifp >1 andq >1 are conjugate numbers1/p1/q1, then
|u|p p |v|q
q ≥ |uv|, ∀u, v∈R, 2.7
and equality holds if and only ifu|v|q−2v.
Let γ > δ. Put u Aδ/γ, p γ/δ, andv Bα1−δ/γγ −δδ/γ−1. It follows from Lemma 2.3that
AxγB≥γδ−δ/γγ−δδ/γ−1Aδ/γB1−δ/γxδ 2.8
for allA, B, x≥0. Rewriting the above inequality we also have
Cxδ−D≤δ−γ/δδγ−δγ/δ−1Cγ/δD1−γ/δxγ 2.9
for allC, x≥0 andD >0.
3. The Main Results
Following 21,22,30, denote fora, b∈ t0,∞Twitha < bthe admissible set Aa, b:
u∈C1rd a, bT:ua 0ub, u /≡0 . 3.1
The main results of this paper are contained in the following three theorems. The arguments used in the proofs have common features with the ones developed in 22,30,34.
Theorem 3.1. Suppose that for any given T ∈ t0,∞T there exist subintervals a1, b1T and a2, b2Tof T,∞Tsuch that
qit≥0 fort∈ a1, b1T∪ a2, b2T,i1,2, . . . , n,
−1ket≥0/≡0 fort∈ ak, bkT,k1,2. 3.2
Let {η1, η2, . . . , ηn} be an n-tuple satisfying 2.2 in Lemma 2.1. If there exists a function u ∈ Aak, bk,k1,2, such that
bk
ak
|uσt|α1
qt η|et|η0n
i1
qηiit
− |uΔt|α1rt
Δt≥0 3.3
fork1,2, where
η01−n
i1
ηi, ηn
i0
η−ηi i, 3.4
then1.1is oscillatory.
Proof. To arrive at a contradiction, let us suppose thatxis a nonoscillatory solution of1.1.
First, we assume thatxtis positive for allt∈ t1,∞T, for somet1 ∈ t0,∞T. Lett∈ a1, b1T, wherea1∈ t1,∞Tis sufficiently large. Define
wt −rtΦα
xΔt
Φαxt . 3.5
It follows that
wΔt ft, xσ
Φαxσt− et
ΦαxσtrtΦ xΔt
ΦαxtΔ
ΦαxtΦαxσt , 3.6
and hence
wΔt qt n
i1
qitΦβi−αxσt |et|
ΦαxσtrtΦ xΔt
ΦαxtΔ
ΦαxtΦαxσt . 3.7
By our assumptions3.2we haveqit≥0 andet≤0 fort∈ a1, b1T. Set
ui 1
ηiqitΦβi−αxσt, u0 1 η0
|et|
Φαxσt. 3.8
Then3.7becomes
wΔt qt n
i0
ηiuirtΦ xΔt
ΦαxtΔ
ΦαxtΦαxσt . 3.9
In view of the arithmetic-geometric mean inequality, see 35, n
i0
ηiui≥n
i0
uηii, 3.10
and equality3.9we obtain
wΔt≥qt η|et|η0n
i1
qiηit rtΦ xΔt
ΦαxtΔ
ΦαxtΦαxσt . 3.11
Multiplying both sides of inequality3.11by|uσ|α1and then using the identity
uΦαuwΔuσΦαuσwΔ |u|α1Δw 3.12
result in
uΦαuwΔ≥ |uσ|α1Q− |uΔ|α1rGu, w, 3.13
where
Qt qt η|et|η0n
i1
qηiit,
Gu, w |uΔ|α1r |u|α1Δ w|uσ|α1r Φα
xΔ
ΦαxΔ ΦαxΦαxσ .
3.14
As demonstrated in 7,12, we know thatGu, w≥0, and thatGu, w 0 if and only if
uΔ Φ−1α
−w r
u, 3.15
whereΦ−1α stands for the inverse function. In our case, since 1μΦ−1α −w/r xσ/x > 0, dynamic equation 3.15 has a unique solution satisfying ua1 0. Clearly, the unique solution isu≡0. Therefore,Gu, w>0 on a1, b1T.
For the benefit of the reader we sketch a proof of the fact thatGu, w≥0. Note that if tis a right-dense point, then we may write
Gu, w α1
Φ−1α r
|Φ−1α ruΔ|α1
α1 wΦαuΦ−1α ruΔ |wΦαu|α1/α α1/α
. 3.16
Applying Young’s inequalityLemma 2.3with
pα1, u Φ−1α ruΔ, vwΦαu, 3.17
we easily see thatGu, w≥0 holds. Iftis a right-scattered point, thenGcan be written as a function ofuμtuΔandvuas
Gu, v 1 μ
r
μα|u|α1 wr
Φα
Φ−1α r μΦ−1α w|uv|α1−w |v|α1
. 3.18
Using differential calculus, see 7, the result follows.
Now integrating the inequality3.13froma1tob1and usingGu, w>0 on a1, b1T we obtain
b1
a1
|uσt|α1Qt− |uΔt|α1rt Δt <0, 3.19
which of course contradicts3.3. This completes the proof whenxtis eventually positive.
The proof whenxtis eventually negative is analogous by repeating the arguments on the interval a2, b2Tinstead of a1, b1T.
A close look at the proof of Theorem 3.1reveals that one cannot takeet ≡ 0. The following theorem is a substitute in that case.
Theorem 3.2. Suppose that for any givenT ∈ t0,∞Tthere exists a subinterval a1, b1Tof T,∞T such that
qit≥0 fort∈ a1, b1T,i1,2, . . . , n. 3.20
Let {η1, η2, . . . , ηn} be an n-tuple satisfying 2.5 in Lemma 2.2. If there exists a function u ∈ Aa1, b1such that
b1
a1
|uσt|α1
qt η
n i1
qηiit
− |uΔt|α1rt
Δt≥0, 3.21
where
ηn
i1
η−ηi i, 3.22
then1.1withet≡0 is oscillatory.
Proof. We proceed as in the proof ofTheorem 3.1to arrive at3.7withet≡0, that is,
wΔt qt n
i1
qitΦβi−αxσt rtΦ xΔt
ΦαxtΔ
ΦαxtΦαxσt . 3.23
Setting
ui 1
ηiqitΦβi−αxσt, 3.24
and using again the arithmetic-geometric mean inequality n
i1
ηiui≥n
i1
uηii, 3.25
we have
wΔt≥qt η n
i1
qηiit rtΦ xΔt
ΦαxtΔ
ΦαxtΦαxσt . 3.26
The remainder of the proof is the same as that ofTheorem 3.1.
As it is shown in 34 for the sublinear terms case, we can also remove the sign condition imposed on the coefficients of the sub-half-linear terms to obtain interval criterion which is applicable for the case when some or all of the functionsqit,i m1, . . . , n, are nonpositive. We should note that the sign condition on the coefficients of super-half-linear terms cannot be removed alternatively by the same approach. Furthermore, the functionet cannot take the value zero on intervals of interest in this case. We have the following theorem.
Theorem 3.3. Suppose that for any given T ∈ t0,∞T there exist subintervals a1, b1T and a2, b2Tof T,∞Tsuch that
qit≥0 fort∈ a1, b1T∪ a2, b2T,i1,2, . . . , m,
−1ket>0 fort∈ ak, bkT,k1,2. 3.27
If there exist a functionu∈ Aak, bk,k1,2, and positive numbersλiandμiwith m
i1
λi n
im1
μi1, 3.28
such that
bk
ak
|uσt|α1
qt m
i1
Pit− n
im1
Rit
− |uΔt|α1rt
Δt≥0 3.29
fork1,2, where
Pit βiβi−αα/βi−1α−α/βiλ1−α/βi iqiα/βit|et|1−α/βi,
Rit βiα−βiα/βi−1α−α/βiμ1−α/βi i−qiα/βit|et|1−α/βi, 3.30
with
−qi
t max
−qit,0
, 3.31
then1.1is oscillatory.
Proof. Suppose that1.1has a nonoscillatory solution. We may assume thatxtis eventually positive on a1, b1Twhena1is sufficiently large. Ifxtis eventually negative, then one can repeat the proof on the interval a2, b2T. Rewrite1.1as follows:
rtΦαxΔΔ
qtΦαxσ gt, xσ 0, t∈ a1, b1T, 3.32
with
gt, x m
i1
qitxβiλi|et|
− n
im1
−qitxβix−μi|et|
. 3.33
Clearly,
gt, x≥m
i1
qitxβiλi|et|
− n
im1
−qitxβi−μi|et|
, 3.34
where
−qit max
−qit,0
. 3.35
Applying2.8and2.9to each summation on the right side with
Aqit, Bλi|et|, γβi, δα,
C −qit, Dμi|et|, δβi, γα, 3.36
we see that
gt, x≥ m
i1
Pit− n
im1
Rit
xα, 3.37
where
Pit βiβi−αα/βi−1α−α/βiλ1−α/βi iqiα/βit|et|1−α/βi, Rit
βi α
1−βi α
α/βi−1
μ1−α/βi i−qiα/βit|et|1−α/βi. 3.38
From3.32and inequality3.37we obtain rtΦαxΔΔ
QtΦαxσ≤0, t∈ a1, b1T, 3.39
where
Qt qt m
i1
Pit− n
im1
Rit. 3.40
Set
wt −rtΦα
xΔt
Φαxt . 3.41
In view of inequality3.39it follows that
wΔt≥Qt rtΦ
xΔt
ΦαxtΔ
ΦαxtΦαxσt . 3.42
The remainder of the proof is the same as that ofTheorem 3.1, hence it is omitted.
4. Applications
To illustrate the usefulness of the results we state the corresponding theorems for the special casesT R,T Z, andT qN,q > 1. One can easily provide similar results for other specific time scales of interest.
4.1. Differential Equations
LetTR, then we havefΔf,σt t, and rtΦαx
qtΦαx n
i1
qitΦβix et, t∈ t0,∞, 4.1
wherer, q, qi, e: t0,∞ → Rare continuous functions withr > 0, andβ1 >· · · > βm > α >
βm1>· · ·βn>0. LetA1a, b:{u∈C1 a, b:ua 0ub, u /≡0}.
Theorem 4.1. Suppose that for any givenT∈ t0,∞there exist subintervals a1, b1and a2, b2of T,∞such that
qit≥0 fort∈ a1, b1∪ a2, b2,i1,2, . . . , n,
−1ket≥0/≡0 fort∈ ak, bk,k1,2. 4.2
Let {η1, η2, . . . , ηn} be an n-tuple satisfying 2.2 in Lemma 2.1. If there exists a function u ∈ A1ak, bk,k1,2, such that
bk
ak
|ut|α1
qt η|et|η0n
i1
qηiit
− |ut|α1rt
dt≥0 4.3
fork1,2, where
η0 1−n
i1
ηi, ηn
i0
η−ηi i, 4.4
then4.1is oscillatory.
Theorem 4.2. Suppose that for any givenT ∈ t0,∞there exists a subinterval a1, b1of T,∞ such that
qit≥0 fort∈ a1, b1,i1,2, . . . , n. 4.5 Let {η1, η2, . . . , ηn} be an n-tuple satisfying 2.5 in Lemma 2.2. If there exists a function u ∈ A1a1, b1such that
b1
a1
|ut|α1
qt η
n i1
qηiit
− |ut|α1rt
dt≥0, 4.6
where
ηn
i1
η−ηi i, 4.7
then4.1withet≡0 is oscillatory.
Theorem 4.3. Suppose that for any givenT∈ t0,∞there exist subintervals a1, b1and a2, b2of T,∞such that
qit≥0 fort∈ a1, b1∪ a2, b2,i1,2, . . . , m,
−1ket>0 fort∈ ak, bk,k1,2. 4.8
If there exist a functionu∈ Aak, bk,k1,2, and positive numbersλiandμiwith m
i1
λi n
im1
μi1, 4.9
such that
bk
ak
|uσt|α1
qt m
i1
Pit− n
im1
Rit
− |ut|α1rt
dt≥0 4.10
fork1,2, where
Pit βiβi−αα/βi−1α−α/βiλ1−α/βi iqiα/βit|et|1−α/βi,
Rit βiα−βiα/βi−1α−α/βiμ1−α/βi i−qiα/βit|et|1−α/βi, 4.11 with
−qi
t max
−qit,0
, 4.12
then4.1is oscillatory.
4.2. Difference Equations
LetTZ, then we havefΔk Δfk fk1−fk,σk k1, and
ΔrkΦαΔxk qkΦαxk1 n
i1
qikΦβixk1 ek, k∈ k0,∞N, 4.13
where k0,∞N{k0, k01, k02, . . .},r, q, qi, e: k0,∞N → Rwithrk>0, andβ1>· · ·>
βm> α > βm1>· · ·βn>0. Let a, bN{a, a1, a2, . . . , b}, andA2a, b:{u: a, bN → R, ua 0ub, u /≡0}.
Theorem 4.4. Suppose that for any given K ∈ k0,∞N there exist subintervals a1, b1N and a2, b2Nof K,∞Nsuch that
qi j
≥0 forj ∈ a1, b1N∪ a2, b2N,i1,2, . . . , n,
−1ke j
≥0/≡0 forj∈ ak, bkN,k1,2. 4.14
Let {η1, η2, . . . , ηn} be an n-tuple satisfying 2.2 in Lemma 2.1. If there exists a function u ∈ A2ak, bk,k1,2, such that
bk−1 jak
|uj1|α1
q j
η|ej|η0n
i1
qiηi j
− |Δuj|α1r j
≥0 4.15
fork1,2, where
η0 1−n
i1
ηi, ηn
i0
η−ηi i, 4.16
then4.13is oscillatory.
Theorem 4.5. Suppose that for any given K ∈ k0,∞N there exists a subinterval a1, b1N of K,∞Nsuch that
qi
j
≥0 forj∈ a1, b1N,i1,2, . . . , n. 4.17
Let {η1, η2, . . . , ηn} be an n-tuple satisfying 2.5 in Lemma 2.2. If there exists a function u ∈ A2a1, b1such that
b1−1 ja1
|uj1|α1
q j
η n
i1
qηii j
− |Δuj|α1r j
≥0, 4.18
where
ηn
i0
η−ηi i, 4.19
then4.13withek≡0 is oscillatory.
Theorem 4.6. Suppose that for any given K ∈ k0,∞N there exist subintervals a1, b1N and a2, b2Nof K,∞Nsuch that
qi
j
≥0 forj∈ a1, b1N∪ a2, b2N,i1,2, . . . , m,
−1ke j
>0 forj∈ ak, bkN,k1,2. 4.20
If there exist a functionu∈ A2ak, bk,k1,2, and positive numbersλiandμiwith m
i1
λi n
im1
μi1, 4.21
such that
bk−1 jak
|uj1|α1
q j
m
i1
Pi j
− n
im1
Ri j
− |Δuj|α1r j
≥0 4.22
fork1,2, where
Pit βiβi−αα/βi−1α−α/βiλ1−α/βi iqiα/βit|et|1−α/βi,
Rit βiα−βiα/βi−1α−α/βiμ1−α/βi i−qiα/βit|et|1−α/βi, 4.23 with
−qi
t max
−qit,0
, 4.24
then4.13is oscillatory.
4.3.q-Difference Equations
LetTqNwithq >1, then we haveσt qt,fΔt Δqft fqt−ft/qt−t,and
Δq
rtΦα
Δqxt
ptΦα
x qt
n
i1
pitΦβi
x qt
et, t∈ t0,∞q, 4.25
where t0,∞q :{qt0, qt01, qt02, . . .}witht0 ∈N,r, p, pi, e: t0,∞q → Rwithrt> 0, and β1 >· · · > βm > α > βm1 > · · ·βn > 0. Let a, bq {qa, qa1, qa2, . . . , qb}witha, b ∈N, and A3a, b:{u: a, bq → R, uqa 0uqb, u /≡0}.
Theorem 4.7. Suppose that for any givenT ∈ t0,∞qthere exist subintervals a1, b1qand a2, b2q of T,∞qsuch that
pit≥0 fort∈ a1, b1q∪ a2, b2q,i1,2, . . . , n,
−1ket≥0/≡0 fort∈ ak, bkq,k1,2. 4.26
Let {η1, η2, . . . , ηn} be an n-tuple satisfying 2.2 in Lemma 2.1. If there exists a function u ∈ A3ak, bk,k1,2, such that
bk−1 jak
qj
|uqj1|α1
p qj
η|eqj|η0n
i1
pηii qj
− |Δquqj|α1r qj
≥0 4.27
fork1,2, where
η01−n
i1
ηi, ηn
i0
η−ηi i, 4.28
then4.25is oscillatory.
Theorem 4.8. Suppose that for any givenT ∈ t0,∞qthere exists a subinterval a1, b1qof T,∞q such that
pit≥0 fort∈ a1, b1q,i1,2, . . . , n. 4.29
Let {η1, η2, . . . , ηn} be an n-tuple satisfying 2.5 in Lemma 2.2. If there exists a function u ∈ A3a1, b1such that
b1−1 ja1
qj
|uqj1|α1
p qj
η n
i1
pηii qj
− |Δquqj|α1r qj
≥0, 4.30
where
ηn
i0
η−ηi i, 4.31
then4.25withet≡0 is oscillatory.
Theorem 4.9. Suppose that for any givenT ∈ t0,∞qthere exist subintervals a1, b1qand a2, b2q of T,∞qsuch that
pit≥0 fort∈ a1, b1q∪ a2, b2q,i1,2, . . . , m,
−1ket>0 fort∈ ak, bkq,k1,2. 4.32
If there exist a functionu∈ A3ak, bk,k1,2, and positive numbersλiandμiwith m
i1
λi n
im1
μi1 4.33
such that
bk−1 jak
qj
|uqj1|α1
p qj
m
i1
Pi qj
− n
im1
Ri qj
− |Δquqj|α1r qj
≥0 4.34
fork1,2, where
Pit βiβi−αα/βi−1α−α/βiλ1−α/βi iqiα/βit|et|1−α/βi,
Rit βiα−βiα/βi−1α−α/βiμ1−α/βi i−qiα/βit|et|1−α/βi 4.35
with
−pi
t max
−pit,0
, 4.36
then4.25is oscillatory.
5. Examples
We give three simple examples to illustrate the importance of our results. For clarity, we have takenn2 andet≡0. Then,
η1 α−β2
β1−β2, η2 β1−α
β1−β2, β1> α > β2>0. 5.1
Example 5.1. Consider the constant coefficient differential equation |x|α−1x
a|x|α−1xb|x|β1−1xc|x|β2−1x0, t≥0, 5.2
whereb, c >0 andaare real numbers.
Letut sint−a1,a1 mand b1 mπ,m ∈ Nis arbitrarily large. Applying Theorem 4.2we see that every solution of5.2is oscillatory if
a b
η1
η1 c η2
η2
≥1. 5.3
Example 5.2. Consider the constant coefficient difference equation
Δ
|Δxk|α−1Δxk
a|xk1|α−1xk1 b|xk1|β1−1xk1 c|xk1|β2−1xk1 0, k≥1,
5.4
whereb, c >0 andaare real numbers.
Letuj 1−−1j, anda12mandb1 2m2,m∈Nis arbitrarily large. It follows fromTheorem 4.5that every solution of5.4is oscillatory if
a b
η1 η1
c η2
η2
≥2. 5.5
Example 5.3. Consider the constant coefficientq-difference equation Δq
|Δqxt|α−1Δqxt
a|xqt|α−1x qt
b|xqt|β1−1x qt c|xqt|β2−1x
qt
0, t≥1, 5.6
whereq >1,b, c >0 andaare real numbers.
Letut qb1 −tt−qa1, anda1 mandb1 m2,m∈ Nis arbitrarily large. In view ofTheorem 4.8, we see that every solution of5.6is oscillatory if
a b
η1 η1
c η2
η2
>0. 5.7
6. Remarks
(1) LiteratureEquation 1.1 has been studied by Sun and Wong 34 for the case T R and α 1.
Our results in Section 4.1 coincide with theirs when α 1, and therefore the results can be considered as an extension fromα 1 to α > 0. Since the results in 34 are linked to many well-known oscillation criteria in the literature, the interval oscillation criteria we have obtained provide further extensions of these to time scales.
The results in Sections4.2and4.3are all new for all values of the parameters. Although there are some results for difference equations in the special case n 1, there is hardly any interval oscillation criteria for theq-difference equations case.
Moreover, since our main results inSection 4are valid for arbitrary time scales, similar interval oscillation criteria can be obtained by considering other particular time scales.
(2) Generalization
The results obtained in this paper remain valid for more general equations of the form rtΦαxΔtΔ
qtgxσ n
i1
qitfixσ et, t∈ t0,∞T, 6.1
provided thatg, fi :R → Rare continuous and satisfy the growth conditions
xgx≥ |x|α1, xfix≥ |x|βi1 ∀x∈R. 6.2
To see this, we note that ifxtis eventually positive, then taking into account the intervals whereqandqiare nonnegative, the above inequalities result in
rtΦαxΔtΔ
qtΦαxσ n
i1
qitqixσ≤et, t∈ t0,∞T. 6.3
The arguments afterward follow analogously.
(3) Forms Related to1.1
Related to1.1are the dynamic equations with mixed delta and nabla derivatives rtΦαxΔ∇
ft, x et, t∈ t0,∞T, 6.4
rtΦαx∇Δ
ft, x et, t∈ t0,∞T, 6.5
rtΦαx∇∇
ft, xρ et, t∈ t0,∞T, 6.6
whereρdenotes the backward jump operator and
ft, x qtΦαx n
i1
qitΦβix. 6.7
It is not difficult to see that time scale modifications of the previous arguments give rise to completely parallel results for the above dynamic equations. For an illustrative example we provide below the version ofTheorem 3.1for6.4. The other theorems for6.4,6.5, and 6.6can be easily obtained by employing arguments developed for1.1in this paper.
Theorem 6.1. Suppose that for any given T ∈ t0,∞T there exist subintervals a1, b1T and a2, b2Tof T,∞Tsuch that
qit≥0 fort∈ a1, b1T∪ a2, b2T,i1,2, . . . , n,
−1ket≥0/≡0 fort∈ ak, bkT,k1,2. 6.8
Let {η1, η2, . . . , ηn} be an n-tuple satisfying 2.2 in Lemma 2.1. If there exists a function u ∈ Bak, bk:{u∈C1ld a, bT:ua 0ub, u /≡0},k1,2, such that
bk
ak
|ut|α1
qt η|et|η0n
i1
qηiit
− |u∇t|α1rρt
Δt≥0 6.9
fork1,2, where
η0 1−n
i1
ηi, ηn
i0
η−ηi i, 6.10
then6.4is oscillatory.
(4) An Open Problem
It is of theoretical and practical interest to obtain interval oscillation criteria when there are only sub-half-linear terms in1.1, that is, whenβi < αholds for alli 1,2, . . . , n. Also, the open problems stated in 34for the special caseT Rwithα 1 naturally carry over for 1.1.