Volume 2011, Article ID 635767,21pages doi:10.1155/2011/635767
Research Article
Existence and Lyapunov Stability of
Periodic Solutions for Generalized Higher-Order Neutral Differential Equations
Jingli Ren,
1Wing-Sum Cheung,
2and Zhibo Cheng
11Department of Mathematics, Zhengzhou University, Zhengzhou 450001, China
2Department of Mathematics, The University of Hong Kong, Pokfulam Road, Hong Kong
Correspondence should be addressed to Wing-Sum Cheung,wscheung@hku.hk Received 17 May 2010; Accepted 23 June 2010
Academic Editor: Feliz Manuel Minh´os
Copyrightq2011 Jingli Ren et al. 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.
Existence and Lyapunov stability of periodic solutions for a generalized higher-order neutral differential equation are established.
1. Introduction
In recent years, there is a good amount of work on periodic solutions for neutral differential equations see1–11and the references cited therein. For example, the following neutral differential equations
d
duut−kut−τ g1ut g2ut−τ1 pt, xt cxt−rf
xt
gxt−τt pt, xt−cxt−σnfxtxt g
0
−rxtsdαs
pt
1.1
have been studied in1,3,8, respectively, and existence criteria of periodic solutions were established for these equations. Afterwards, along with intensive research on thep-Laplacian,
some authors4,11start to consider the followingp-Laplacian neutral functional differential equations:
φpxt−cxt−σ
gt, xt−τt pt, φp
xt−cxt−σ f
xt
gxt−τt et, 1.2 and by using topological degree theory and some analysis skills, existence results of periodic solutions for1.2have been presented.
In general, most of the existing results are concentrated on lower-order neutral functional differential equations, while studies on higher-order neutral functional differential equations are rather infrequent, especially on higher-order p-Laplacian neutral functional differential equations. In this paper, we consider the following generalized higher-order neutral functional differential equation:
ϕpxt−cxt−σln−l F
t, xt, xt, . . . , xl−1t
, 1.3
whereϕp:R → Ris given byϕps |s|p−2swithp ≥2 being a constant,Fis a continuous function defined onRland is periodic with respect totwith periodT, that is,Ft,·, . . . ,· FtT,·, . . . ,·, Ft, a,0, . . . ,0/≡0 for alla∈R, andc,σare constants.
Since the neutral operator is divided into two cases |c|/1 and |c| 1, it is natural to study the neutral differential equation separately according to these two cases. The case
|c| 1 has been studied in 5. Now we consider1.3 for the case|c|/1. So throughout this paper, we always assume that |c|/1, and the paper is organized as follows. We first transform 1.3 into a system of first-order differential equations, and then by applying Mawhin’s continuation theory and some new inequalities, we obtain sufficient conditions for the existence of periodic solutions for1.3. The Lyapunov stability of periodic solutions for the equation will then be established. Finally, an example is given to illustrate our results.
2. Preparation
First, we recall two lemmas. LetXandYbe real Banach spaces and letL:DL⊂X → Ybe a Fredholm operator with index zero; hereDLdenotes the domain ofL. This means that ImL is closed inY and dim KerLdimY/ImL<∞. Consider supplementary subspacesX1, Y1 ofX,Y, respectively, such thatX KerL⊕X1,Y ImL⊕Y1. LetP : X → KerLand Q : Y → Y1 denote the natural projections. Clearly, KerL∩DL∩X1 {0}and so the restrictionLP :L|DL∩X1is invertible. LetKdenote the inverse ofLP.
LetΩbe an open bounded subset ofXwithDL∩Ω/∅. A mapN:Ω → Y is said to beL-compact inΩifQNΩis bounded and the operatorKI−QN:Ω → Xis compact.
Lemma 2.1see12. Suppose thatX andY are two Banach spaces, and suppose thatL:DL⊂ X → Yis a Fredholm operator with index zero. LetΩ⊂Xbe an open bounded set and letN:Ω → Y beL-compact onΩ. Assume that the following conditions hold:
1Lx /λNx, for allx∈∂Ω∩DL, λ∈0,1, 2Nx /∈Im L, for allx∈∂Ω∩KerL,
3deg{JQN,Ω∩KerL,0}/0, whereJ: ImQ → KerLis an isomorphism.
Then, the equationLxNxhas a solution inΩ∩DL.
Lemma 2.2see13. Ifω∈C1R,Randω0 ωT 0, then T
0
|ωt|pdt≤ T
πp
pT
0
ωt pdt, 2.1
wherepis a fixed real number withp >1 and
πp2
p−1/p
0
ds 1−sp/
p−11/p 2π
p−11/p psin
π/p . 2.2
For the sake of convenience, throughout this paper we denote byT a positive real number, and for any continuous functionu, we write
|u|0: max
t∈0,T|ut|. 2.3
LetA:CT → CTbe the operator onCT :{x∈CR,R:xtT xtfor allt∈R}
given by
Axt:xt−cxt−σ, ∀x∈CT, t∈R. 2.4 Lemma 2.3. The operatorAhas a continuous inverseA−1onCTsatisfying the following:
1 A−1f
t
⎧⎪
⎪⎪
⎪⎪
⎨
⎪⎪
⎪⎪
⎪⎩
ft ∞
j1
cjf t−jσ
, for|c|<1, ∀f∈CT,
−ftσ
c −∞
j1
1 cj1f
t j1
σ
, for|c|>1, ∀f∈CT,
2.5
2 A−1f
t ≤ f
0
|1− |c||, ∀f∈CT, 2.6
3 T
0
A−1f
t dt≤ 1
|1− |c||
T
0
ft dt, ∀f ∈CT. 2.7
Remark 2.4. This lemma is basically proved in3,10. For the convenience of the readers, we present a detailed proof here as follows.
Proof. We split it into the following two cases.
Case 1|c|<1. Define an operatorB:CT → CTby
Bxt:cxt−σ, ∀x∈CT, t∈R. 2.8
Clearly,Bjxt cjxt−jσandA I−B. Note also thatB< |c|<1. Therefore,Ahas a continuous inverseA−1:CT → CTwithA−1 I−B−1∞
j0Bj; hereB0xt:xt. Hence, A−1f
t ∞
j0
Bjf
t ft ∞
j1
cjf t−jσ
, 2.9
and so
A−1ft
∞ j0
Bjf
t
∞ j0
cjf t−jσ
≤ f
0
1− |c| , T
0
A−1f
t dt≤∞
j0
T
0
Bjf
t dt∞
j0
T
0
cjf
t−jσ dt≤ 1 1− |c|
T
0
ft dt.
2.10
Case 2|c|>1. Define operators
E:CT−→CT, Ext:xt−1
cxtσ, B1:CT−→CT, B1xt: 1
cxtσ.
2.11
From the definition of the linear operatorB1, we have
B1jf t 1
cjf tjσ
, ∞
j0
B1jf
t ft ∞
j1
1 cjf
tjσ .
2.12
SinceB1<1, the operatorEhas a bounded inverseE−1:CT → CT with
E−1 I−B1−1I∞
j1
B1j, 2.13
and so, for anyf∈CT,
E−1f
t ft ∞
j1
Bj1f
t. 2.14
On the other hand, fromAxt xt−cxt−σ, we have Axt xt−cxt−σ −c
xt−σ−1 cxt
. 2.15
That is,
Axt −cExt−σ. 2.16
Now, for anyf∈CT, ifxtsatisfies
Axt ft, 2.17
then we have
−cExt−σ ft, 2.18
or
Ext −ftσ
c f1t. 2.19
So, we have
xt
E−1f1
t f1t ∞
j1
B1jf1t −ftσ
c −∞
j1
B1jftσ
c . 2.20
So,A−1exists and satisfies
A−1f
t −ftσ
c −∞
j1
B1jftσ
c −ftσ
c −∞
j1
1 cj1f
t j1
σ ,
A−1f t
−ftσ
c −∞
j1
1 cj1f
t j1
σ ≤ f
0
|c| −1.
2.21
This proves1and2of Lemma2.3. Finally,3is easily verified.
By Hale’s terminology14, a solution xtof 1.3is thatxt ∈ C1R,Rsuch that Ax ∈ C1R,Rand 1.3is satisfied onR. In general,xtdoes not belong toC1R,R.But we can see easily fromAxt Axtthat a solutionxtof1.3must belong toC1R,R.
Equation1.3is transformed into
ϕp
Axl
tn−l F
t, xt, xt, . . . , xl−1t
. 2.22
Lemma 2.5see4. Ifp >1, then T
0
A−1f
t pdt≤ 1
|1− |c||p T
0
ft pdt, ∀f∈CT. 2.23
Now we consider 2.22. Define the conjugate index q ∈ 1,2 by 1/p1/q 1.
Introducing new variables
y1t xt, y2t xt, y3t xt, . . . , ylt xl−1t, yl1t ϕp
Axlt
, yl2t
ϕp
Axlt
, . . . , ynt ϕp
Axltn−l−1 . 2.24
Using the fact thatϕq◦ϕp≡id and by Lemma2.3,1.3can be rewritten as
y1t y2t, y2t y3t,
... yl−1t ylt, ylt A−1ϕq
yl1t , yl1 t yl2t,
... yn−1t ynt, ynt F
t, y1t, y2t, . . . , ylt .
2.25
It is clear that, ifyt y1t, y2t, . . . , ynt is aT-periodic solution to2.25, theny1t must be aT-periodic solution to1.3. Thus, the problem of finding aT-periodic solution for 1.3reduces to finding one for2.25.
Define the linear spaces
X Y
y
y1·, y2·, . . . , yn·
∈C0R,Rn:ytT≡yt
2.26
with normymax{y1,y2, . . . ,yn}. Obviously,XandY are Banach spaces. Define
L:DL
y∈C1R,Rn:ytT yt
⊂X−→Y 2.27
by
Lyy
⎛
⎜⎜
⎜⎜
⎜⎜
⎜⎜
⎜⎝ y1 y2 ... yl
... yn
⎞
⎟⎟
⎟⎟
⎟⎟
⎟⎟
⎟⎠
. 2.28
Moreover, define
N:X−→Y 2.29
by
Ny
⎛
⎜⎜
⎜⎜
⎜⎜
⎜⎜
⎜⎜
⎜⎝
y2t y3t
... A−1ϕq
yl1t yl2t
... F
t, y1t, y2t, . . . , ylt
⎞
⎟⎟
⎟⎟
⎟⎟
⎟⎟
⎟⎟
⎟⎠
. 2.30
Then, 2.25can be rewritten as the abstract equationLy Ny. From the definition ofL, one can easily see that KerL {y ∈ C1R,Rn : yis constant} Rn and ImL {y : y ∈ X,T
0 ysds 0}. So,Lis a Fredholm operator with index zero. Let P : X → KerLand Q:Y → ImQbe defined by
P y 1 T
T
0
ysds, Qy 1
T T
0
ysds. 2.31
It is easy to see that KerL ImQRn. Moreover, for ally∈Y, if we writey∗ y−Qy, we haveT
0 y∗sds0 and soy∗∈ImL. This is to sayY ImQ⊕ImLand dimY/ImL dim ImQ dim KerL. So,L is a Fredholm operator with index zero. Let K denote the inverse ofL|Kerp∩DL, then we have
Ky t
T 0
G1t, sy1sds, T
0
G2t, sy2sds, . . . , T
0
Gnt, synsds
, 2.32
where
Git, s
⎧⎪
⎪⎪
⎨
⎪⎪
⎪⎩ s
T, 0≤s < t≤T, s−T
T , 0≤t≤s≤T,
i1,2, . . . , n. 2.33
From2.30and2.33, it is clear thatQN andKI−QN are continuous, andQNΩis bounded, and soKI−QNΩis compact for any open boundedΩ ⊂ X. Hence,N isL- compact onΩ. For the functionyt y1t, y2t, . . . , yntdefined as2.24, we have the following.
Lemma 2.6. Ifyt∈C1R,RnandytT yt, then T
0
yit pdt≤ 1
|1−|cp T
πp
pl−i T πq
qn−lT
0
ynt qdt, 2.34
where 1/p1/q1, p≥2, i1,2, . . . , l−1.
Proof. Fromy10 y1T, there is a pointt1∈0, Tsuch thaty1t1 0. Letω1t y1tt1. Then,ω10 ω1T 0. Fromy20 y2T, there is a pointt2 ∈0, Tsuch thaty2t2 0.
Letω2t y2tt2. Then,ω20 ω2T 0.Continuing this way, we get fromyl−10 yl−1Ta pointtl−1 ∈0, Tsuch thatyl−1tl−1 0. Letωl−1t yl−1 ttl−1. Then,ωl−10 ωl−1T 0.Fromylt yltT, we haveT
0Ayltdt T
0 Ayltdt Aylt|T0 0, so there is a pointtl ∈ 0, T such thatAyltl 0; hence, we haveϕpAyltl 0. Let ωlt ϕpAylttl yl1ttl. Then,ωl0 ωlT 0. Continuing this way, we get fromyn−10 yn−1T that there is a pointtn−1 ∈ 0, Tsuch that yn−1tn−1 0. Let ωn−1t yn−1ttn−1. Then,ωn−10 ωn−1T 0.By Lemma2.2, we have
T
0
y1t pdt T
0
|ω1t|pdt
≤ T
πp pT
0
ω1t pdt
T πp
pT
0
y2t pdt
T πp
pT
0
|ω2t|pdt
≤ T
πp 2pT
0
ω2t pdt ...
≤ T
πp
pl−1T
0
ωl−1t pdt
T πp
pl−1T
0
ylt pdt.
2.35
By Lemma2.5and Lemma2.2, we have
T
0
ylt pdt T
0
A−1ϕq
yl1t pdt
≤ 1
|1− |c||p T
0
ϕq
yl1t pdt 1
|1− |c||p T
0
yl1t pq−pdt 1
|1− |c||p T
0
yl1t qdt 1
|1− |c||p T
0
|ωlt|qdt
≤ 1
|1− |c||p T
πq
qT
0
ωlt qdt 1
|1− |c||p T
πq qT
0
yl2t qdt ...
≤ 1
|1− |c||p T
πq
qn−lT
0
ωn−1t qdt 1
|1− |c||p T
πq
qn−lT
0
ynt qdt.
2.36
Combining2.35and2.36, we get
T
0
y1t pdt≤ 1
|1− |c||p T
πp
pl−1 T πq
qn−lT
0
ynt qdt. 2.37
Similarly, we get T
0
yit pdt≤ 1
|1− |c||p T
πp
pl−i T πq
qn−lT
0
ynt qdt. 2.38
This completes the proof of Lemma2.6.
Remark 2.7. In particular, if we takep2, thenq2 and
πpπqπ22 2−1/2
0
ds
1−s2/2−11/2 2π2−11/2
2 sinπ/2 π. 2.39
In this case,2.34is transformed into T
0
yit 2dt≤ 1
|1−|c2 T
π
2n−iT
0
ynt 2dt. 2.40
3. Main Results
For the sake of convenience, we list the following assumptions which will be used repeatedly in the sequel.
H1There exists a constantD >0 such that
z1Ft, z1, z2, . . . , zl>0, ∀t, z1, z2, . . . , zl∈0, T×Rl, with|z1|> D. 3.1
H2There exists a constantM >0 such that
|Ft, z1, z2, . . . , zl| ≤M, ∀t, z1, z2, . . . , zl∈0, T×Rl. 3.2
H3There exist nonnegative constantsα1, α2, . . . , αl, msuch that
|Ft, z1, z2, . . . , zl| ≤α1|z1|α2|z2|· · ·αl|zl|m, ∀t, z1, z2, . . . , zl∈0, T×Rl. 3.3
H4There exist nonnegative constantsγ1, γ2, . . . , γnsuch that
|Ft, u1, u2, . . . , un−Ft, v1, v2, . . . , vn| ≤γ1|u1−v1|γ2|u2−v2|· · ·γn|un−vn| 3.4
for allt, u1, u2, . . . , un,t, v1, v2, . . . , vn∈0, T×Rn.
Theorem 3.1. IfH1andH2hold, then1.3has at least one nonconstantT-periodic solution.
Proof. Consider the equation
LyλNy, λ∈0,1. 3.5
LetΩ1{y∈DL:LyλNy, λ∈0,1}. Ifyt y1t, y2t, . . . , ynt∈Ω1, then
y1t λy2t, y2t λy3t,
... yl−1t λylt,
ylt λϕq yl1t
, yl1t λyl2t,
...
yn−1t λynt, ynt λF
t, y1t, y2t, . . . , ylt .
3.6
We first claim that there exists a constantξ∈Rsuch that
y1ξ ≤D. 3.7
Integrating the last equation of3.6over0, T, we have T
0
F
t, y1t, y2t, . . . , ylt
dt0. 3.8
By the continuity ofF, there existsξ∈0, Tsuch that
F
ξ, y1ξ, . . . , ylξ
0. 3.9
From assumptionH1, we get3.7. As a consequence, we have
y1t y1ξ
t
ξ
y1sds ≤D
T
0
y1s ds. 3.10
On the other hand, multiplying both sides of the last equation of3.6byyntand integrating over0, T, using assumptionH2,we have
T
0
ynt 2dtλ T
0
F
t, y1t, y2t, . . . , ylt yntdt
≤ T
0
F
t, y1t, y2t, . . . , ylt ynt dt
≤M T
0
ynt dt
≤MT1/2 T
0
ynt 2dt 1/2
.
3.11
It is easy to see that there exists a constantMn>0independent ofλsuch that
T
0
ynt 2dt≤Mn. 3.12
From yn−10 yn−1T, there exists a pointt1 ∈ 0, T such thatynt1 0. By H ¨older’s inequality, we have
ynt ≤ T
0
ynt dt≤T1/2 T
0
ynt 2dt 1/2
≤T1/2Mn1/2 :Mn. 3.13
Fromyn−20 yn−2T, there exists a pointt2∈0, Tsuch thatyn−1t2 0, and we have
yn−1t ≤ T
0
yn−1t dt T
0
λynt dt≤ T
0
ynt dt≤TMn:Mn−1. 3.14
Continuing this way foryn−2, . . . , yl1, we get
yl1t ≤TMl2:Ml1. 3.15
Hence,
ylt ≤ T
0
ylt dt≤ T
0
λA−1ϕq
yl1t dt≤ 1
|1− |c||
T
0
yl1t q−1dt
≤ 1
|1− |c||TMq−1l1 :Ml, yl−1t ≤TMl:Ml−1,
...
y2t ≤TM3:M2.
3.16
Meanwhile, from3.10, we get y1t ≤D
T
0
y1t dt≤DTM2:M1. 3.17
LetM0max{M1, M2, . . . , Mn}. Then, obviouslyy1 ≤M0,y2 ≤M0, . . . ,andyn ≤M0. LetΩ2 {y ∈ KerL : Ny ∈ ImL}. Ify ∈ Ω2, then y ∈ KerL, which means that yconstant andQNy0. We see that
y20, y30,
... yn0, F
t, y1,0, . . . ,0 0.
3.18
So,
y1 ≤D≤M0, y2y3· · ·yn0≤M0. 3.19
Now takeΩ {y y1, y2, . . . , yn ∈X :y1< M01, y2 < M01, . . . , yn <
M01}. By the analysis above, it is easy to see thatΩ1 ⊂Ω,Ω2⊂Ω, and conditions1and 2of Lemma2.1are satisfied.
Next we show that condition3of Lemma2.1is also satisfied. Define an isomorphism J: ImQ → KerLas follows:
J
y1, y2, . . . , yn
:
yn, y1, . . . , yn−1
. 3.20
LetHμ, y μy 1−μJQNy,μ, y∈0,1×Ω. Then, for allμ, y∈0,1×∂Ω∩KerL,
H μ, y
⎛
⎜⎜
⎜⎜
⎜⎜
⎜⎜
⎜⎜
⎜⎝
μy11−μ T
T
0
F
t, y1,0, . . . ,0 dt y2
... A−1ϕq
yl1 ... yn
⎞
⎟⎟
⎟⎟
⎟⎟
⎟⎟
⎟⎟
⎟⎠
. 3.21
FromH1, it is obvious thatyHμ, y>0 for allμ, y∈0,1×∂Ω∩KerL. Therefore,
deg{JQN,Ω∩KerL,0}deg! H
0, y
,Ω∩KerL,0"
deg! H
1, y
,Ω∩KerL,0"
deg{I,Ω∩KerL,0}/0,
3.22
which means that condition3of Lemma2.1is also satisfied. By applying Lemma2.1, we conclude that equationLyNyhas a solutionyt∗ y1∗t, y2∗t, . . . , y∗ntonΩ; that is, 1.3has aT-periodic solutiony1∗twithy1∗< M01.
Finally, observe thaty∗1tis not constant. For, ify1∗≡aconstant, then from1.3we haveFt, a,0, . . . ,0≡0, which contradicts the assumption thatFt, a,0, . . . ,0/≡0. The proof is complete.
Theorem 3.2. IfH1andH3hold, then1.3has at least one nonconstant T-periodic solution if one of the following conditions holds:
1p >2,
2p2 and 1/|1− |c||α1Tα2T/πl−1α3T/πl−2· · ·αlT/πT/πn−l<1.
Proof. LetΩ1be defined as in Theorem3.1. Ifyt y1t, y2t, . . . , ynt ∈Ω1,then from the proof of Theorem3.1we have
ynt λF
t, y1t, y2t, . . . , ylt
, 3.23
y1
0≤D T
0
y1s ds. 3.24 We claim that|yn|0is bounded.
Multiplying both sides of 3.23 byϕqynt and integrating over 0, T, by using assumptionH3,we have
T
0
ynt qdtλ T
0
F
t, y1t, y2t, . . . , ylt ϕq
ynt dt
≤ T
0
F
t, y1t, y2t, . . . , ylt ϕq
ynt dt
≤α1
T
0
y1t ϕq
ynt dtα2
T
0
y2t ϕq
ynt dt · · ·αl
T
0
ylt ϕq
ynt dtm T
0
ϕq
ynt dt
≤α1 y1
0
T
0
ϕq
ynt dtα2
T
0
y2t ϕq
ynt dt · · ·αl
T
0
ylt ϕq
ynt dtm T
0
ϕq
ynt dt
≤α1
D
T
0
y1t dt
T 0
ϕq
ynt dtα2
T
0
y2t ϕq
ynt dt · · ·αl
T
0
ylt ϕq
ynt dtm T
0
ϕq
ynt dt.
3.25
Applying H ¨older’s inequality, we have T
0
ynt qdt≤α1
⎡
⎣DT1/p T
0
y1t qdt 1/q⎤
⎦T1/q T
0
ϕq
ynt pdt 1/p
α2 T
0
y2t qdt
1/qT
0
ϕq
ynt pdt 1/p
· · ·αl T
0
ylt qdt
1/qT
0
ϕq
ynt pdt 1/p
mT1/q T
0
ϕq
ynt pdt 1/p
≤α1T·Tp−2/qp−1 T
0
y1t pdt
1/qp−1T
0
ynt qdt 1/p
α2Tp−2/qp−1 T
0
y2t pdt
1/qp−1T
0
ynt qdt 1/p
· · ·
αlTp−2/qp−1 T
0
ylt pdt
1/qp−1T
0
|ynt|qdt 1/p
α1DmT1/q T
0
|ynt|qdt 1/p
α1T·Tp−2/p T
0
y1t pdt
1/pT
0
ynt qdt 1/p
α2Tp−2/p T
0
y2t pdt
1/pT
0
ynt qdt 1/p
· · ·αlTp−2/p T
0
ylt pdt
1/pT 0
ynt qdt 1/p
α1DmT1/q T
0
ynt qdt 1/p
.
3.26
Applying Lemma2.6and3.26, we have
T
0
ynt qdt≤α1T ·Tp−2/p 1
|1− |c||
T πp
l−1 T πq
qn−l/pT
0
ynt qdt 2/p
α2Tp−2/p 1
|1− |c||
T πp
l−1 T πq
qn−l/pT
0
ynt qdt 2/p
· · ·αlTp−2/p 1
|1− |c||
T πp
T πq
qn−l/pT
0
ynt qdt 2/p
α1DmT1/q T
0
ynt qdt 1/p
≤ 1
|1− |c||
⎡
⎣α1Tα2 T
πp l−1
α3 T
πp l−2
· · ·αl T
πp
⎤
⎦
×Tp−2/p T
πq
qn−l/p
· T
0
ynt qdt 2/p
α1DmT1/q T
0
ynt qdt 1/p
.
3.27
Case 1. Ifp2 and 1/|1− |c||α1T α2T/πl−1α3T/πl−2· · ·αlT/πT/πn−l<1, then it is easy to see that there exists a constantMn>0independent ofλsuch that
T
0
ynt qdt≤Mn. 3.28
Case 2. Ifp >2, then it is easy to see that there exists a constantMn > 0independent ofλ such that
T
0
ynt qdt≤Mn. 3.29
From yn−10 yn−1T, there exists a pointt1 ∈ 0, T such thatynt1 0. By H ¨older’s inequality, we have
ynt ≤ T
0
ynt dt≤T1/p T
0
ynt qdt 1/q
≤T1/pMn1/q :Mn. 3.30
This proves the claim, and the rest of the proof of the theorem is identical to that of Theorem3.1.
Remark 3.3. If1.3takes the form
ϕpxt−cxt−σln−l F
t, xt, xt, . . . , xl−1t
et, 3.31
whereet∈ CR,R, etT etandT
0 etdt 0, then the results of Theorems3.1and 3.2still hold.
Remark 3.4. Ifp2, then1.3is transformed into xt−cxt−σnF
t, xt, xt, . . . , xn−1t
, 3.32
and the results of Theorems3.1and3.2still hold.
Next, we study the Lyapunov stability of the periodic solutions of3.32.
Theorem 3.5. Assume thatH4holds. Then everyT-periodic solution of3.32is Lyapunov stable.
Proof. Let
z1t xt, z2t xt, . . . , znt xn−1t. 3.33
Then, system3.32is transformed into
z1t z2t, z2t z3t,
...
znt A−1Ft, z1t, z2t, . . . , znt.
3.34
Suppose now thatz∗t z∗1t, z∗2t, . . . , z∗ntis aT-periodic solution of3.34. Let zt z1t, z2t, . . . , zntbe any arbitrary solution of3.34. For anyk 1, . . . , n, write wkt zkt−z∗kt. Then, it follows from3.34that
w1t w2t, w2t w3t,
...
wnt A−1
Ft, z1t, z2t, . . . , znt−F
t, z∗1t, z∗2t, . . . , z∗nt ,
3.35
and so
w1t |w2t|, w2t |w3t|,
... wnt A−1
Ft, z1t, z2t, . . . , znt−F
t, z∗1t, z∗2t, . . . , z∗nt .
3.36
Letulk t |wklt|, l0,1, k1,2, . . . , n. Then, u1t u2t, u2t u3t,
...
unt A−1
Ft, z1t, z2t, . . . , znt−F
t, z∗1t, z∗2t, . . . , z∗nt .
3.37
Takeβmax{γ1/|1− |c||, γ2/|1− |c||1, . . . , γn/|1− |c||1}1, and define a functionV·by
Vt, u1, . . . , un:e−βt n k1
ukt. 3.38
LetUu1, . . . , un n
k1ykt. It is obvious thatVt, u1, . . . , un > 0 and Vt, u1, . . . , un ≥ Uu1, . . . , un>0. FromH4and Lemma2.3, we get
V˙t, u1, . . . , un −βe−βt n
k1
ukt
e−βtu2t · · ·unt e−βt A−1
Ft, z1t, z2t, . . . , znt−F
t, z∗1t, z∗2t, . . . , z∗nt
≤ −βe−βt n
k1
ukt
e−βtu2t · · ·unt
e−βt
|1− |c|| Ft, z1t, z2t, . . . , znt−F
t, z∗1t, z∗2t, . . . , z∗nt
≤ −βe−βt n
k1
ukt
e−βtu2t · · ·unt
e−βt
|1− |c||
γ1 z1t−z∗1t · · ·γn|znt−z∗nt|
−βe−βt n
k1
ukt
e−βtu2t · · ·unt
e−βt
|1− |c||
γ1|w1t|· · ·γn|wnt|
−βe−βt n
k1
ukt
e−βtu2t · · ·unt
e−βt
|1− |c||
γ1u1t · · ·γnunt
−β γ1
|1− |c|| u1te−βtn
k2
−β1 γk
|1− |c|| ukte−βt
<0.
3.39
Hence,Vis a Lyapunov function for nonautonomous3.32 see15, page 50, and so theT-periodic solutionz∗of3.32is Lyapunov stable.
Finally, we present an example to illustrate our result.
Example 3.6. Consider then-order delay differential equation
ϕpxt−3xt−σ4n−4 1
3πxt 1
8sinxt 1
8cosxtsin 2t1
8sinxt.
3.40