Applied Mathematics and Computation 217 (2010) 661–676
Contents lists available at ScienceDirect
Applied Mathematics and Computation journal homepage: www.elsevier.com/locate/amc
Positive periodic solutions for impulsive predator–prey model with dispersion and time delays Ruixi Liang a,*, Jianhua Shen b a b
School of Mathematical Sciences and Computing Technologies, Central South University, Changsha, Hunan 410075, China Department of Mathematics, Hangzhou Normal University, Hangzhou, Zhejiang 310036, China
a r t i c l e
i n f o
a b s t r a c t In this paper, we study the existence and global attractivity of positive periodic solutions for impulsive predator–prey systems with dispersion and time delays. By using coincidence degree theorem, a set of easily verifiable sufficient conditions are obtained for the existence of at least one strictly positive periodic solutions, and by means of a suitable Lyapunov functional, the uniqueness and global attractivity of positive periodic solution is presented. Some known results subject to the underlying systems without impulses are improved and generalized. Ó 2010 Elsevier Inc. All rights reserved.
Keywords: Periodic solution Global attractivity Time delay Dispersion Impulses
1. Introduction The effect of environment change in the growth and diffusion of a species in a heterogenous habitat is a very interesting subject in the ecological literature (see [2–8,16,19]). Because of human activities, the location of manufacturing industries, the pollution of the atmosphere, of soil, of rivers, etc., more and more habitats are broken into patches and some of them are polluted. In some of these patches, without the contribution from other patches, the species will go to extinction. As a consequence, the species will occupy a week patchy environment. Since the pioneering theoretical work by Skellam [19], many works have focused on the effect of spatial factors which play a crucial rule in the stability and persistence of a population ([3,4,6–9,12,13] and reference cited therein). Xu et al. [4] investigated the following two-species non-autonomous path-system with time delays:
8 0 > < x1 ðtÞ ¼ x1 ðtÞ½r1 ðtÞ a11 ðtÞx1 ðtÞ a13 ðtÞyðtÞ þ D1 ðtÞ½x2 ðtÞ x1 ðtÞ; x02 ðtÞ ¼ x2 ðtÞ½r2 ðtÞ a22 ðtÞx2 ðtÞ a23 ðtÞyðtÞ þ D2 ðtÞ½x1 ðtÞ x2 ðtÞ; > : 0 y ðtÞ ¼ yðtÞ½r3 ðtÞ þ a31 ðtÞx1 ðt s1 Þ þ a32 ðtÞx2 ðt s1 Þ a33 ðtÞyðt s2 Þ;
ð1:1Þ
with initial conditions
xi ðhÞ ¼ /i ðhÞ; /i ð0Þ > 0;
yðhÞ ¼ wðhÞ;
wð0Þ > 0;
/i ;
h 2 ½s; 0; w 2 Cð½s; 0Þ; Rþ Þ;
i ¼ 1; 2;
where xi(t) denote the density of species x in patch i, i = 1, 2, at time t, and y(t) denotes the total predator population for both patches. ri(t) is the intrinsic growth rate of the prey at patch i, i = 1, 2; aii(t) (i = 1, 2) are the density-dependent coefficients of the prey at patch i; a13(t) and a23(t) are the capturing rates of the predator in patch 1 and patch 2, respectively, a31(t)/a13(t) and a32(t)/a23(t) are the conversion rates of nutrients into the reproduction of the predator, r3(t) is the death rate of the * Corresponding author. E-mail address:
[email protected] (R. Liang). 0096-3003/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.amc.2010.06.003
662
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
predator; Di(t) is dispersion rate of prey species x, i = 1, 2. s = max{s1, s2}. s1 is the delay due to gestation, that is, mature adult predators can only contribute to the production of predator biomass. In addition, the model have included the term a33(t)y(t s2) in the dynamics of the predator y to incorporate the negative feedback of predator crowding. In [4], the authors determined sufficient conditions on parameters of the model that ensure the existence, uniqueness and global stability of positive periodic solution of the system. However, in population dynamics, many evolutionary processes experience short-time rapid changes after undergoing relatively long smooth variation. For examples, the harvesting and stocking occur at fixed moments, and some species usually immigrate at the same time every year, etc. If we still thought of the population dynamical systems with these phenomena as continuous systems, it would be unreasonable or incorrect. We should establish systems with impulsive effects. Recently, theories for impulsive differential equations have been introduced into population dynamics [1,5,10,11]. To the authors knowledge, the population dynamical systems with diffusion and impulsive effects are seldom discussed. In this paper, we mainly study the following impulsive system:
8 0 x ðtÞ ¼ x1 ðtÞ½r 1 ðtÞ a11 ðtÞx1 ðtÞ a13 ðtÞyðtÞ þ D1 ðtÞ½x2 ðtÞ x1 ðtÞ; > > > 10 > > > < x2 ðtÞ ¼ x2 ðtÞ½r 2 ðtÞ a22 ðtÞx2 ðtÞ a23 ðtÞyðtÞ þ D2 ðtÞ½x1 ðtÞ x2 ðtÞ; y0 ðtÞ ¼ yðtÞ½r 3 ðtÞ þ a31 ðtÞx1 ðt s1 Þ þ a32 ðtÞx2 ðt s1 Þ a33 ðtÞyðt s2 Þ; > > > > Dxi ðt k Þ ¼ bik xi ðt k Þ; > > : Dyðt k Þ ¼ b3k yðtk Þ;
t – tk ; t – tk ; t – tk ;
ð1:2Þ
i ¼ 1; 2; k ¼ 1; 2; . . . ; k ¼ 1; 2; . . . ;
where bikxi(tk) (i = 1, 2) and b3ky(tk) represent the population xi(t) and y(t) at tk regular harvest pulse. Throughout this paper, for system (1.2) the following conditions are assumed. (C1) ri(t), aij(t)(i, j = 1, 2, 3), D1(t) and D2(t) are continuously positive periodic functions with periodic x, s1 and s2 are nonnegative constants. (C2) 1 < bik 6 0, i = 1, 2, 3 for all k 2 N and there exists a positive integer q such that tk+q = tk + x, bi(k+q) = bik, i = 1, 2, 3 and tk s1, tk s2 – tm. In the following, we shall use the notation
f ¼ 1
x
Z 0
x
f ðsÞds;
f L ¼ min f ðtÞ; t2½0;x
f M ¼ max f ðtÞ: t2½0;x
Without loss of generality, we shall assume tk – 0, x and [0, x] \ {tk} = {t1,t2, . . . , tq}. It is clear that without the impulses system (1.2) reduces to system (1.1). The existence of positive periodic solutions of (1.1) is investigated in [4], and the following result was obtained. Theorem 1.1. In addition to (C1), assume further that the following hold: M M (H2) aL33 ðr1 D1 Þ aM 13 Aða31 þ a32 Þ > 0; L M M (H3) a33 ðr2 D2 Þ a23 Aða31 þ aM 32 Þ > 0; L M M M (H4) aL31 aM 22 ðr 1 D1 Þ þ a32 a11 ðr 2 D2 Þ r 3 a11 a22 > 0;where
( ) M M ðr 1 D1 ÞM þ DM 1 ðr 2 D2 Þ þ D2 A ¼ max ; : aL11 aL22
Then system (1.1) has at least one positive x-periodic solution. The Example 2 in [4] shows that Theorem 1.1 has room for improvement. The organization of this paper is as follows. In the next section, we establish some simple criteria for the existence of positive periodic solutions of system (1.2). We also note that our results improve Theorem 1.1 as bik 0, because our results do not need the condition (H2) or (H3). In Section 3, the uniqueness and global attactivity of periodic solutions are presented. Finally, we give an example to show our results.
2. Existence of periodic solutions In this section, by using continuation theorem which was proposed in [14] by Gaines and Mawhin, we will establish the existence conditions of at least one positive periodic solution to system (1.2). To do so, we need to make some preparations. Let X, Z be real Banach spaces, L: Dom L X ? Z be a Fredholm mapping of index zero (index L = dim Ker L codim Im L), and let P: X ? X, Q: Z ? Z be continuous projectors such that Im P = Ker L, Ker Q = Im L and X = Ker L Ker P, Z = Im L Im Q. Denote by LP the restriction of L to Dom L \ Ker P, KP: Im L ? Ker P \ Dom L the inverse (to LP), and J: Im Q ? Ker L an isomorphism of Im Q onto Ker L. For convenience, we first introduce Mawhin’s continuation theorem [14] as follows:
663
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
Lemma 2.1. Let X X be an open bounded set. Let L be a Fredholm mapping of index zero and N be L-compact on X. Assume (a) Lx – kNx for each k 2 (0,1), x 2 oX \ Dom L, (b) for each x 2 Ker L \ oX, QNx – 0, \ DomL. (c) deg {JQN,X \ Ker L,0} – 0. Then Lx = Nx has at least one solution in X To prove the main conclusion by means of the continuation theorem, we need introduce some function spaces. Let PC(R, R3) = {x:R ? R3jx be continuous at t – tk ; xðt þ k Þ; xðt k Þ exist and xðt k Þ ¼ xðt k Þ; k ¼ 1; 2; . . . ; g, let X = {(u1(t), u2(t), u3(t))T 2 PC(R, R3):ui(t + x) = ui(t), i = 1, 2, 3} with the norm
kðu1 ðtÞ; u2 ðtÞ; u3 ðtÞÞkT ¼
3 X
sup jui ðtÞj;
i¼1 t2½0;x
and
Y ¼ X R3q with the norm kukY ¼ kxk þ kyk; for u 2 Y; x 2 X; y 2 R3q ; here j j denotes the Euclidean norm. Then X and Y are Banach spaces. Theorem 2.1. In addition to (C1) and (C2), assume further that the following hold: M M L (C3) aL33 ðr1 D1 Þx aM 13 Aða31 þ a32 Þx þ a33 B1 > 0;
(C4)
aL31 aM ½ðr 1 D1 ÞxþB1 þaL32 aM ½ðr 2 D2 ÞxþB2 22 11 r 3 xB3
Bi ¼
q X
lnð1 þ bik Þ;
M > aM 11 a22 ; where A is defined in Theorem 1.1, and
i ¼ 1; 2; 3:
k¼1
Then system (1.2) has at least one positive x-periodic solution. Proof. Let
u1 ðtÞ ¼ ln½x1 ðtÞ;
u2 ðtÞ ¼ ln½x2 ðtÞ;
u3 ðtÞ ¼ ln½yðtÞ;
ð2:1Þ
then system (1.2) can be translated to
u01 ðtÞ ¼ r 1 ðtÞ D1 ðtÞ a11 ðtÞeu1 ðtÞ a13 ðtÞeu3 ðtÞ þ D1 ðtÞeu2 ðtÞu1 ðtÞ ; u02 ðtÞ ¼ r 2 ðtÞ D2 ðtÞ a22 ðtÞeu2 ðtÞ a23 ðtÞeu3 ðtÞ þ D2 ðtÞeu1 ðtÞu2 ðtÞ ; u03 ðtÞ ¼ r 3 ðtÞ þ a31 ðtÞeu1 ðts1 Þ þ a32 ðtÞeu2 ðts1 Þ a33 ðtÞeu3 ðts2 Þ ;
Dui ðtk Þ ¼ lnð1 þ bik Þ;
i ¼ 1; 2; 3; k ¼ 1; 2; . . .
ð2:2Þ
It is easy to see that if system (2.2) has one x-periodic solution ðu 1 ðtÞ; u 2 ðtÞ; u 3 ðtÞÞT , then ðx 1 ðtÞ; x 2 ðtÞ; y ðtÞÞT ¼ ðexp½u 1 ðtÞ; exp½u 2 ðtÞ; exp½u 3 ðtÞÞT is a positive x-periodic solution of (1.2). Therefore, to complete the proof, we need only to prove that (2.2) has one x-periodic solution. Let L: Dom L X ? Y, u ? (u0 , Du(t1), . . . , Du(tq)),
32 3 2 31 3 2 lnð1 þ b1q Þ lnð1 þ b12 Þ lnð1 þ b11 Þ r 1 ðtÞ D1 ðtÞ a11 ðtÞeu1 ðtÞ a13 ðtÞeu3 ðtÞ þ D1 ðtÞeu2 ðtÞu1 ðtÞ 6 7 6 7 6 7C B6 7 Nu ¼ @4 r 2 ðtÞ D2 ðtÞ a22 ðtÞeu2 ðtÞ a23 ðtÞeu3 ðtÞ þ D2 ðtÞeu1 ðtÞu2 ðtÞ 5; 4 lnð1 þ b21 Þ 5; 4 lnð1 þ b22 Þ 5;. .. ; 4 lnð1 þ b2q Þ 5A: lnð1 þ b3q Þ lnð1 þ b31 Þ lnð1 þ b32 Þ r3 ðtÞ þ a31 ðtÞeu1 ðts1 Þ þ a32 ðtÞeu2 ðts1 Þ a33 ðtÞeu3 ðts2 Þ 02
Evidently
Ker L ¼ fu : uðtÞ ¼ c 2 R3 ; t 2 ½0; xg; ( Z x
Im L ¼
z ¼ ðf ; a1 ; . . . ; aq Þ 2 Y :
f ðsÞds þ
0
q X
) ak ¼ 0 ;
k¼1
and
dimKerL ¼ 3 ¼ codim Im L: So Im L is closed in Y, L is a Fredholm mapping of index zero. Define
Px ¼
1
x
Z
x
xðtÞdt;
0
Qz ¼ Qðf ; a1 ; a2 ; . . . ; aq Þ ¼
1
x
"Z 0
x
f ðsÞds þ
q X k¼1
#
!
ak ; 0; . . . ; 0 :
664
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
It is easy to show that P and Q are continuous projectors satisfying
Im P ¼ Ker L; Im L ¼ Ker Q ¼ ImðI QÞ: Furthermore, by an easy computation, we can find that the inverse KP: Im L ? Ker P \ Dom L has the form
K P ðzÞ ¼
Z
t
f ðsÞds þ
0
X
ak
tk
1
x
Z 0
x
Z
t
f ðsÞdsdt
0
q X
ak :
k¼1
Thus
00 1 R x
1 1 u1 ðtÞ a13 ðtÞeu3 ðtÞ x 0 ½r 1 ðtÞ D1 ðtÞ a11 ðtÞe C BB C C BB C q P C BB C u2 ðtÞu1 ðtÞ 1 þD1 ðtÞe dt þ x lnð1 þ b1k Þ; C C BB C BB C k¼1 C BB C C BB 1 R x C u ðtÞ u ðtÞ C BB x 0 ½r 2 ðtÞ D2 ðtÞ a22 ðtÞe 2 a23 ðtÞe 3 C C BB C C BB C q QNu ¼ BB C; 0; . . . ; 0C; P u1 ðtÞu2 ðtÞ 1 C BB C þD2 ðtÞe dt þ x lnð1 þ b2k Þ; C C BB C BB C k¼1 C BB R C C BB 1 x u1 ðts1 Þ u2 ðts1 Þ C C BB x 0 ½r 3 ðtÞ þ a31 ðtÞe C þ a32 ðtÞe C BB C C BB C q A @@ A P a33 ðtÞeu3 ðts2 Þ dt þ x1 lnð1 þ b3k Þ; k¼1
and
2Rt 0
½r1 ðsÞ D1 ðsÞ a11 ðsÞeu1 ðsÞ a13 ðsÞeu3 ðsÞ þ D1 ðsÞeu2 ðsÞu1 ðsÞ ds þ
P
lnð1 þ b1k Þ
3
7 6 t>t k 7 6Rt P 7 6 ½r2 ðsÞ D2 ðsÞ a22 ðsÞeu2 ðsÞ a23 ðsÞeu3 ðsÞ þ D2 ðsÞeu1 ðsÞu2 ðsÞ ds þ lnð1 þ b Þ 2k K P ðI Q ÞNu ¼ 6 0 7 t>t k 7 6R P 5 4 t u1 ðss1 Þ u2 ðss1 Þ u3 ðss2 Þ ½r 3 ðsÞ þ a31 ðsÞe þ a32 ðsÞe a33 ðsÞe ds þ lnð1 þ b3k Þ 0 t>t k
3 lnð1 þ b Þ 1k 7 6x 0 7 6 k¼1 7 6 R R q P x t 61 u2 ðsÞ u3 ðsÞ u1 ðsÞu2 ðsÞ a23 ðsÞe þ D2 ðsÞe ds þ lnð1 þ b2k Þ 7 6 x 0 0 ½r2 ðsÞ D2 ðsÞ a22 ðsÞe 7 7 6 k¼1 7 6 q R R P 5 4 1 x t u1 ðss1 Þ u2 ðss1 Þ u3 ðss2 Þ ½r ðsÞ þ a ðsÞe þ a ðsÞe a ðsÞe ds þ lnð1 þ b Þ 3 31 32 33 3k 0 x 0 2
1
Rx Rt
½r ðsÞ D1 ðsÞ a11 ðsÞeu1 ðsÞ a13 ðsÞeu3 ðsÞ þ D1 ðsÞeu2 ðsÞu1 ðsÞ ds þ 0 1
q P
k¼1
3 q P u1 ðsÞ u3 ðsÞ u2 ðsÞu1 ðsÞ ½r ðsÞ D ðsÞ a ðsÞe a ðsÞe þ D ðsÞe ds þ lnð1 þ b Þ 1 11 13 1 1k 7 6 x 2 0 1 7 6 k¼1 7 6 q R P 6 t 1 t ½r ðsÞ D ðsÞ a ðsÞeu2 ðsÞ a ðsÞeu3 ðsÞ þ D ðsÞeu1 ðsÞu2 ðsÞ ds þ lnð1 þ b2k Þ 7 6 x 2 0 2 7: 2 22 23 2 7 6 k¼1 7 6 q P 5 4 t 1 R t u1 ðss1 Þ u2 ðss1 Þ u3 ðss2 Þ þ a32 ðsÞe a33 ðsÞe ds þ lnð1 þ b3k Þ x 2 0 ½r 3 ðsÞ þ a31 ðsÞe 2
t
Rt 1
k¼1
Clearly, QN and KP(I Q)N are continuous. Using Lemma 2.4 in [1], it is not difficult to show that QNðXÞ; K p ðI Q ÞNðXÞ are relatively compact for any open bounded set X X. Hence N is L-compact on X for any open bounded set X X. Now we reach the position to search for an appropriate open, bounded subset X for the application of the continuation theorem. Corresponding to equation Lu = kNu, k 2 (0,1), we have
u01 ðtÞ ¼ k½r 1 ðtÞ D1 ðtÞ a11 ðtÞeu1 ðtÞ a13 ðtÞeu3 ðtÞ þ D1 ðtÞeu2 ðtÞu1 ðtÞ ; u02 ðtÞ ¼ k½r 2 ðtÞ D2 ðtÞ a22 ðtÞeu2 ðtÞ a23 ðtÞeu3 ðtÞ þ D2 ðtÞeu1 ðtÞu2 ðtÞ ;
ð2:3Þ
u03 ðtÞ ¼ k½r 3 ðtÞ þ a31 ðtÞeu1 ðts1 Þ þ a32 ðtÞeu2 ðts1 Þ a33 ðtÞeu3 ðts2 Þ ;
Dui ðtk Þ ¼ k lnð1 þ bik Þ;
i ¼ 1; 2; 3; k ¼ 1; 2; . . . :
Since ui(t) (i = 1, 2, 3) are x-periodic functions, we need only to prove the result in the interval [0, x]. Integrating (2.3) over the interval [0, x] leads to
Z
x
a11 ðtÞeu1 ðtÞ dt þ
Z
0
Z 0
x
a13 ðtÞeu3 ðtÞ dt ¼ 0
x
a22 ðtÞeu2 ðtÞ dt þ
Z
x
ðr 1 ðtÞ D1 ðtÞÞdt þ
0
x
a23 ðtÞeu3 ðtÞ dt ¼ 0
Z
Z 0
Z
x
D1 ðtÞeu2 ðtÞu1 ðtÞ dt þ
0
x
ðr 2 ðtÞ D2 ðtÞÞdt þ
Z 0
q X
lnð1 þ b1k Þ;
ð2:4Þ
lnð1 þ b2k Þ;
ð2:5Þ
k¼1
x
D2 ðtÞeu1 ðtÞu2 ðtÞ dt þ
q X k¼1
665
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
Z
x
a31 ðtÞeu1 ðts1 Þ dt þ
Z
0
x
a32 ðtÞeu2 ðts1 Þ dt ¼
Z
0
x
r 3 ðtÞdt þ
0
Z
x
a33 ðtÞeu3 ðts2 Þ dt
0
q X
lnð1 þ b3k Þ:
ð2:6Þ
k¼1
It follows from (2.3)–(2.6) that
Z
x
ju1 ðtÞjdt <
0
¼2
Z
a11 ðtÞeu1 ðtÞ dt þ 2
ju02 ðtÞjdt < ¼2
Z
ju03 ðtÞjdt <
0
¼2
Z
x
x
D1 ðtÞeu2 ðtÞu1 ðtÞ dt þ q X
a13 ðtÞeu3 ðtÞ dt Z
k¼1
x
D2 ðtÞeu1 ðtÞu2 ðtÞ dt þ
x
r 3 ðtÞdt þ
q X
x
a23 ðtÞeu3 ðtÞ dt
k¼1
Z
0
x
a31 ðtÞeu1 ðts1 Þ dt þ
a22 ðtÞeu2 ðtÞ dt þ
x
x
a32 ðtÞeu2 ðts1 Þ dt þ
0
x
a23 ðtÞeu3 ðtÞ dt
x
D2 ðtÞdt;
a32 ðtÞeu2 ðts1 Þ dt þ
q X
Z 0
0
Z
a13 ðtÞeu3 ðtÞ dt
D1 ðtÞdt;
x
Z
x
x
0
Z
0
Z
Z 0
0
Z
lnð1 þ b2k Þ þ 2
0
a31 ðtÞeu1 ðts1 Þ dt þ 2
a11 ðtÞeu1 ðtÞ dt þ
0
0
Z
x
lnð1 þ b1k Þ þ 2
0
Z
Z 0
x
ðr 2 ðtÞ þ D2 ðtÞÞdt þ
a22 ðtÞeu2 ðtÞ dt þ 2
x
Z
x
0
Z 0
0
Z
x
0
Z
ðr 1 ðtÞ þ D1 ðtÞÞdt þ
x
x
0
x
0
0
Z
Z
lnð1 þ b3k Þ 6 2
Z
Z
x
a33 ðtÞeu3 ðts2 Þ dt
0
x
a31 ðtÞeu1 ðts1 Þ dt þ 2
0
k¼1
Z
x
a32 ðtÞeu2 ðts1 Þ dt:
0
ð2:7Þ Multiplying the first equation of (2.3) by eu1 ðtÞ and integrating over [0, x], we obtain
p X
b1k eu1 ðtk Þ þ
Z
x
a11 e2u1 ðtÞ dt 6 ðr 1 D1 ÞM
Z
0
k¼1
x
0
eu1 ðtÞ dt þ DM 1
Z
x
eu2 ðtÞ dt:
0
Since 1 < b1k 6 0, so we have
Z
x
a11 ðtÞe2u1 ðtÞ dt 6 ðr 1 D1 ÞM
Z
0
x
eu1 ðtÞ dt þ DM 1
0
Z
x
eu2 ðtÞ dt; 0
which yields
aL11
Z
x
e2u1 ðtÞ dt 6 ðr 1 D1 ÞM
Z
0
x
0
eu1 ðtÞ dt þ DM 1
Z
x
eu2 ðtÞ dt:
ð2:8Þ
0
Similarly, multiplying the second equation in (2.3) by eu2 ðtÞ and integrating over [0, x] gives
aL22
Z
x
e2u2 ðtÞ dt < ðr 2 D2 ÞM
0
Z
x 0
eu2 ðtÞ dt þ DM 2
Z
x
eu1 ðtÞ dt:
ð2:9Þ
0
By using the inequalities
Z
x
eui ðtÞ dt
2
6x
0
Z
x
e2ui ðtÞ dt;
i ¼ 1; 2;
0
and (2.8) and (2.9), we have
aL11 aL22 If
Rx 0
Z
x
eu1 ðtÞ dt
0
Z
x
eu2 ðtÞ dt
2
< xðr 1 D1 ÞM
Z
Z
x
0
< xðr 2 D2 ÞM
0
Z
x
0
eu2 ðtÞ dt 6
aL11
2
Rx 0
eu1 ðtÞ dt þ DM 1 x eu2 ðtÞ dt þ DM 2 x
Z
x
eu2 ðtÞ dt;
ð2:10Þ
eu1 ðtÞ dt:
ð2:11Þ
0
Z
x
0
eu1 ðtÞ dt, then we derive from (2.10) that
x
eu1 ðtÞ dt
2
< xðr 1 D1 ÞM
0
Z 0
x
eu1 ðtÞ dt þ DM 1 x
Z
x
eu1 ðtÞ dt;
0
which implies
Z
x
eu2 ðtÞ dt 6
0
If
Rx 0
0
x
eu1 ðtÞ dt <
0
eu1 ðtÞ dt 6
Z
Z
Rx 0
xðr1 D1 ÞM þ xDM1 aL11
:
ð2:12Þ
:
ð2:13Þ
eu2 ðtÞ dt, then we can conclude
x
eu1 ðtÞ dt 6
Z 0
x
eu2 ðtÞ dt <
xðr2 D2 ÞM þ xDM2 aL22
666
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
Set
A ¼ max
( ) M M ðr 1 D1 ÞM þ DM 1 ðr 2 D2 Þ þ D2 ; ; aL11 aL22
ð2:14Þ
so, from (2.12)–(2.14) we get
Z
x
eui ðtÞ dt < xA;
i ¼ 1; 2:
ð2:15Þ
0
Since u(t) 2 X, there exist ni, gi 2 [0, x] such that
ui ðgi Þ ¼ max ui ðtÞ;
ui ðni Þ ¼ min ui ðtÞ; t2½0;x
i ¼ 1; 2; 3:
t2½0;x
ð2:16Þ
From (2.15) and (2.16), we see that
ui ðni Þ < ln A; Noticing that
aL33
Z
Rx 0
i ¼ 1; 2:
eu1 ðts1 Þ dt ¼
x 0
eu3 ðtÞ dt ¼ aL33 ¼
Z
Z
x
ð2:17Þ
Rx 0
x
eu1 ðtÞ dt and
eu3 ðts2 Þ dt 6
Rx 0
Z
0
x
Rx 0
eu3 ðtÞ dt, from (2.6) and (2.15), one obtains
a33 ðtÞeu3 ðts2 Þ dt
0
a31 ðtÞeu1 ðts1 Þ dt þ
Z
0
Z
eu3 ðts2 Þ dt ¼
x
a32 ðtÞeu2 ðts1 Þ dt
Z
0
x
6
a31 ðtÞeu1 ðts1 Þ dt þ
Z
0
x
r3 ðtÞdt þ
0
x
a32 ðtÞeu2 ðts1 Þ dt ¼
Z
0
q X
lnð1 þ b3k Þ
k¼1
x
a31 ðtÞeu1 ðtÞ dt þ
0
Z 0
x M a32 ðtÞeu2 ðtÞ dt 6 ðaM 31 þ a32 ÞxA;
which implies
Z
x
eu3 ðtÞ dt 6
0
M ðaM 31 þ a32 ÞxA aL33
ð2:18Þ
and
u3 ðn3 Þ 6 ln
M ðaM 31 þ a32 ÞA : aL33
ð2:19Þ
It follows from (2.7), (2.15) and (2.18) that
Z 0
Z
0
M X ðaM 31 þ a32 ÞxA þ 2xD1 lnð1 þ b1k Þ :¼ c1 aL33 k¼1
M ju02 ðtÞjdt 6 2aM 22 xA þ 2a23
M X ðaM 31 þ a32 ÞxA þ 2xD2 lnð1 þ b2k Þ :¼ c2 L a33 k¼1
q
q
x
0
Z
M ju01 ðtÞjdt 6 2aM 11 xA þ 2a13
x
ð2:20Þ
x M ju03 ðtÞjdt 6 2ðaM 31 þ a32 ÞxA :¼ c 3 :
Thus, from (2.17), (2.19) and (2.20), we have
8 Rt 0 P > > < u1 ðn1 Þ þ n1 u1 ðsÞds þ n
lnð1 þ b1k Þ; t 2 ½0; n1 > : u1 ðn1 Þ þ n1 u1 ðsÞds 6 u1 ðn1 Þ þ
t6t k 6n 1
Z
x
ju01 ðtÞjdt
0
< ln A þ c1
q X
lnð1 þ b1k Þ
k¼1 q X
lnð1 þ b1k Þ
ð2:21Þ
k¼1
u2 ðtÞ 6 u2 ðn2 Þ þ
Z
x 0
ju02 ðtÞjdt
q X k¼1
lnð1 þ b2k Þ < ln A þ c2
q X k¼1
lnð1 þ b2k Þ
ð2:22Þ
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
u3 ðtÞ 6 u3 ðn3 Þ þ
Z
x
ju03 ðtÞjdt
0
X
M X ðaM 31 þ a32 ÞA þ c3 lnð1 þ b3k Þ: L a33 k¼1
667
q
lnð1 þ b3k Þ < ln
k¼1
ð2:23Þ
From (2.4) and (2.18) we obtain
Z
x
a11 ðtÞeu1 ðtÞ dt >
Z
0
x
ðr 1 ðtÞ D1 ðtÞÞdt
Z
0
q X
x
a13 ðtÞeu3 ðtÞ dt þ
0
P xðr 1 D1 Þ aM 13
lnð1 þ b1k Þ
k¼1
xAðaM31 þ aM32 Þ aL33
þ
q X
lnð1 þ b1k Þ > 0
k¼1
which implies
u1 ðg1 Þ > ln
xðr1 D1 Þ aM13
xAðaM þaM Þ 31 32 aL33
þ
Pq
k¼1
lnð1 þ b1k Þ :¼ d1 :
31 a
This, together with (2.20), leads to
u1 ðtÞ P u1 ðg1 Þ
Z
x
ju01 ðtÞjdt þ
0
q X
lnð1 þ b1k Þ > d1 c1 þ
q X
k¼1
lnð1 þ b1k Þ:
ð2:24Þ
k¼1
Let
( R1 ¼ max j ln Aj þ c1
q X
lnð1 þ b1k Þ; jd1 j þ c1
k¼1
q X
) lnð1 þ b1k Þ ;
k¼1
it follows from (2.21) and (2.24) that
max ju1 ðtÞj < R1 :
ð2:25Þ
t2½0;x
From (2.5) we have
22 eu2 ðg2 Þ P ðr2 D2 Þ þ D2 eu1 ðn1 Þu2 ðg2 Þ a
Z
x
a23 ðtÞeu3 ðtÞ dt þ
0
lnð1 þ b2k Þ
k¼1
> ðr 2 D2 Þ þ D2 eu1 ðn1 Þ eu2 ðg2 Þ aM 23 P ðr2 D2 Þ þ D2 eR1 eu2 ðg2 Þ aM 23
q X
xAðaM31 þ aM32 Þ aL33
q X
þ
lnð1 þ b2k Þ
k¼1
xAðaM31 þ aM32 Þ aL33
þ
q X
lnð1 þ b2k Þ;
k¼1
which implies
u2 ðg2 Þ
e
P
dþ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 22 D2 eR1 d þ 4a ; 2a22
where d ¼ ðr2 D2 Þ aM 23
xAðaM þaM Þ 31 32 aL33
þ
Pq
k¼1
lnð1 þ b2k Þ, so
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 22 D2 eR1 d þ d þ 4a u2 ðg2 Þ P ln :¼ d2 : 22 2a
ð2:26Þ
It follows from (2.20) and (2.26) that
u2 ðtÞ P u2 ðg2 Þ
Z
x 0
ju02 ðtÞjdt þ
q X
lnð1 þ b2k Þ P d2 c2 þ
k¼1
q X
lnð1 þ b2k Þ:
k¼1
This, together with (2.22), leads to
max ju2 ðtÞj < maxfj ln Aj þ c2
t2½0;x
From (2.4) and (2.5), we have
q X k¼1
lnð1 þ b2k Þ; jd2 j þ c2
q X k¼1
lnð1 þ b2k Þg :¼ R2 :
ð2:27Þ
668
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
aM 11 aM 22
Z
x
eu1 ðtÞ dt þ aM 13
0
Z 0
x
eu2 ðtÞ dt þ aM 23
Z
x
eu3 ðtÞ dt P ðr1 D1 Þx þ 0
Z
x
eu3 ðtÞ dt P ðr2 D2 Þx þ 0
q X k¼1 q X
lnð1 þ b1k Þ; ð2:28Þ lnð1 þ b2k Þ;
k¼1
We derive from (2.6) that
Z
x
a33 ðtÞeu3 ðts2 Þ dt ¼
Z
0
x
Z
a31 ðtÞeu1 ðts1 Þ dt þ
0
a32 ðtÞeu2 ðts1 Þ dt
0
P aL31 ¼ aL31
x
Z
x 0
Z
Z
x
0
eu1 ðtÞ dt þ aL32
x
eu2 ðts1 Þ dt r 3 x þ
0
r 3 ðtÞdt þ
x
eu2 ðtÞ dt r 3 x þ
Rx 0
q X
q X
0
lnð1 þ b3k Þ
lnð1 þ b3k Þ
eu3 ðtÞ dt þ
q P
lnð1 þ b1k Þ
k¼1
þ aL32
aM 11 Rx
lnð1 þ b3k Þ
k¼1
k¼1
P aL31 ðr2 D2 Þx aM 23
q X
k¼1
0
ðr 1 D1 Þx aM 13
x
0
Z
eu1 ðts1 Þ dt þ aL32
Z
eu3 ðtÞ dt þ
q P
lnð1 þ b2k Þ
k¼1
r3 x þ
aM 22
q X
lnð1 þ b3k Þ
ð2:29Þ
k¼1
From (2.28), (2.29) and condition (C4), we have
u3 ðg3 Þ P ln
aL31 aM 22 ½ðr 1 D1 Þx þ
Pq
L M k¼1 lnð1 þ b1k Þ þ a32 a11 ½ðr 2 D2 Þ L M M M L M L M a11 a22 a33 þ a11 a32 a23 þ aM 22 a31 a13
xþ
Pq
k¼1
Pq M aM 11 a22 ½r 3 x k¼1 lnð1 þ b3k Þ :¼ d3 : M M L M M L M aL11 aM 22 a33 þ a11 a32 a23 þ a22 a31 a13
lnð1 þ b2k Þ
ð2:30Þ
So, from (2.20) and (2.30), we obtain
u3 ðtÞ P u3 ðg3 Þ
Z 0
x
ju03 ðtÞjdt þ
q X
lnð1 þ b3k Þ P d3 c3 þ
k¼1
q X
lnð1 þ b3k Þ:
ð2:31Þ
k¼1
This, together with (2.23), leads to
max ju3 ðtÞj < R3 ;
t2½0;x
here
( ) q q M X X ðaM 31 þ a32 ÞA R3 ¼ max ln lnð1 þ b3k Þ; jd3 j þ c3 lnð1 þ b3k Þ : þ c3 aL 33
k¼1
k¼1
Clearly, R1, R2 and R3 are independent of k. Similar to the proof of Theorem 2.1 of [4], we can find a sufficiently large M > 0, denote the set
X ¼ fuðtÞ ¼ ðu1 ðtÞ; u2 ðtÞ; u3 ðtÞÞT 2 X : kuk < M; uðtþk Þ 2 X;
k ¼ 1; 2; . . . ; qg;
it follows that for each u 2 Ker L \ oX, QNu – 0 and
degfJQNu; X \ KerL; 0g ¼ 1 – 0: By now we have proved that X verifies all the requirements in Lemma 2.1. Hence (2.2) has at least one x-periodic solution. Accordingly, system (1.2) has at least one positive x-periodic solution. The proof is complete. h Remark 1. If bik 0, (i = 1, 2, 3), k = 1, 2 . . . , then (1.2) is translated to (1.1). In this case, the conditions (C3) and (C4) are the same as (H2) and (H4) of Theorem 1.1, but we see that (H3) is not need in here. Hence our result improve and generalized the corresponding result of [4]. Similarly, we can easily get the following theorem. Theorem 2.2. The conclusion of Theorem (2.1) is still valid if the condition C3 in Theorem 2.1 is replaced by the following M M L (C 3 ) aL33 ðr2 D2 Þx aM 23 Aða31 þ a32 Þx þ a33 B2 > 0.
669
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
3. Uniqueness and global stability of periodic solutions Under the hypotheses (C1) (C2), we consider the nonimpulsive delay differential equation
Y
z01 ðtÞ ¼ z1 ðtÞ½r1 ðtÞ A11 ðtÞz1 ðtÞ A13 ðtÞz3 ðtÞ þ D1 ðtÞ½z2 ðtÞ
ð1 þ b2k Þð1 þ b1k Þ1 z1 ðtÞ
0
Y
z02 ðtÞ ¼ z2 ðtÞ½r2 ðtÞ A22 ðtÞz2 ðtÞ A23 ðtÞz3 ðtÞ þ D2 ðtÞ½z1 ðtÞ
ð1 þ b1k Þð1 þ b2k Þ1 z2 ðtÞ
ð3:1Þ
0
z03 ðtÞ ¼ z3 ðtÞ½r 3 ðtÞ þ A31 ðtÞz1 ðt s1 Þ þ A32 ðtÞz2 ðt s1 Þ A33 ðtÞz3 ðt s2 Þ with initial conditions
zi ðhÞ ¼ /i ðhÞ; z3 ðhÞ ¼ wðhÞ; h 2 ½s; 0 /i ð0Þ > 0; wð0Þ > 0; /i ; w 2 Cð½s; 0Þ; Rþ Þ; where
Q
Aii ðtÞ ¼ aii ðtÞ
ð1 þ bik Þ;
ð3:2Þ
i ¼ 1; 2;
ði ¼ 1; 2; 3Þ;
0
Q
A13 ðtÞ ¼ a13 ðtÞ
ð1 þ b3k Þ; A23 ðtÞ ¼ a23 ðtÞ
0
Q
A31 ðtÞ ¼ a31 ðtÞ
Q
ð1 þ b3k Þ
0
ð1 þ b1k Þ;
0
Q
A32 ðtÞ ¼ a32 ðtÞ
ð1 þ b2k Þ;
t P 0:
0
The following lemmas will be used in the proofs of our results. The proof of the first lemma is similar to that of Theorem 1 in [15], and it will be omitted. Lemma 3.1. Assume that (C1) (C2) hold. Then Q Q (i) if z(t) = (z1(t),z2(t),z3(t))T is a solution of (3.1) on [ s, + 1), then xi ðtÞ ¼ 0 0 such that
0 < zi ðtÞ 6 M i ;
ði ¼ 1; 2; 3Þ;
where
M1 ¼ M2 >
M 1 ;
e 1 ðtÞ ¼ D1 ðtÞ D
M 1
Y
for t P T 2 ;
( ) eM M eM rM 1 þ D 1 r2 þ D 2 ¼ max ; ; AL11 AL22
ð3:3Þ
M3 ¼
M AM 31 M 1 þ A32 M 2
AL33
M
M
eðA31 M1 þA32 M2 Þs2 ;
ð1 þ b2k Þð1 þ b1k Þ1 ;
ð3:4Þ
0
e 2 ðtÞ ¼ D2 ðtÞ D
Y
ð1 þ b1k Þð1 þ b2k Þ1 :
0
Proof. We define
VðtÞ ¼ maxfz1 ðtÞ; z2 ðtÞg: Calculating the upper-right derivative of V(t) along the positive solution of system (3.1), we have the following (P1) If z1(t) > z2(t) or z1(t) = z2(t), then
"
Dþ VðtÞ ¼ z01 ðtÞ ¼ z1 ðtÞ½r1 ðtÞ A11 ðtÞz1 ðtÞ A13 ðtÞz3 ðtÞ þ D1 ðtÞ z2 ðtÞ
Y 0
L L M eM eM 6 z1 ðtÞ½r M 1 A11 z1 ðtÞ þ D 1 z2 ðtÞ 6 z1 ðtÞ½r 1 þ D 1 A11 z1 ðtÞ:
(P2) If z1(t) < z2(t) or z1(t) = z2(t), then L eM Dþ VðtÞ ¼ z02 ðtÞ 6 z2 ðtÞ½r M 2 þ D 2 A22 z2 ðtÞ:
# ð1 þ b2k Þð1 þ b1k Þ1 z1 ðtÞ
670
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
It follows from (P1) and (P2) that L eM Dþ VðtÞ 6 zi ðtÞ½r M i þ D i Aii zi ðtÞ; ði ¼ 1 or 2Þ:
ð3:5Þ
By (3.5) we can derive (A) If max{z1(0), z2(0)} 6 M1, then max{z1(t), z2(t)} 6 M1, t P 0. L eM (B) If max{z1(0), z2(0)} > M1. Let a ¼ maxfM 1 ðr M i þ D i Aii M 1 Þ; i ¼ 1; 2g, (a > 0). We consider the following two possibilities: (a) V(0) = z1(0) > M1. (z1(0) P z2(0)). (b) V(0) = z2(0) > M1. (z1(0) < z2(0)). If (a) holds, then there exists e > 0 such that if t 2 [0,e), then V(t) = z1(t) > M1, and we have
Dþ VðtÞ ¼ z01 ðtÞ < a < 0: If (b) holds, then there exists e > 0 such that if t 2 [0,e), then V(t) = z2(t) > M1, and also we have
Dþ VðtÞ ¼ z02 ðtÞ < a < 0: From what has been discussed above, we can conclude that if V(0) > M1, then V(t) is strictly monotone decreasing with speed at least a. Therefore there exists a T1 > 0 such that if t P T1, then
VðtÞ ¼ maxfz1 ðt; z2 ðtÞg 6 M1 :
ð3:6Þ
In addition, from the third equation of system (3.1) and (3.6) we derive that for t > T1 + s1, M L z03 ðtÞ 6 z3 ðtÞ½AM 31 M 1 þ A32 M 2 A33 z3 ðt s2 Þ:
A similar argument in proof of Lemma 2.1 in [16] shows that there exists a T2 P T1 + s1 such that
z3 ðtÞ 6
M AM 31 M 1 þ A32 M 2
AL33
M
M
eðA31 M1 þA32 M2 Þs2 :¼ M3 ; for t P T 2 ;
which completes the proof.
h
Lemma 3.3. Let (C1) (C2) hold. Assume further that ðC 5 ÞðAL31 þ AL32 Þ min Then there exist positive constants T > T2 and mi (i = 1, 2, 3) such that
nrL DM AM M 1
1
AM 11
13
3
;
M r L2 DM 2 A23 M 3
AM 22
mi < zi ðtÞ; ði ¼ 1; 2; 3Þ for t P T;
o
> rM 3 .
ð3:7Þ
in which
( ) M M M L r L1 DM 1 A13 M 3 r 2 D2 A23 M 3 ; m1 ¼ m2 < min ; AM AM 11 22 m3 <
m 3
¼
AL31 m1 þ AL32 m2 r M 3 AM 33
ð3:8Þ
;
and M3 are defined in (3.4). Proof. Define
VðtÞ ¼ minfz1 ðtÞ; z2 ðtÞg: Then calculating to lower-right derivative of V(t) along the positive solution of system (3.1). Similar to the discussion for inequality (3.5), it is easy to obtain by (3.3)
( Dþ VðtÞ ¼ z0i ðtÞði ¼ 1 or 2Þ P
M M z1 ðtÞ½rL1 AM 11 z1 ðtÞ A13 M 3 D1 ði ¼ 1Þ M M for t P T 2 ; z2 ðtÞ½rL2 AM 22 z1 ðtÞ A23 M 3 D2 ði ¼ 2Þ
where T2 be defined in Lemma 3.2. (I) If V(T2) = min{z1(T2),z2(T2)} P m1, then min{z1(t),z2(t)} P m1, t P T2. (II) If V(T2) = min{z1(T2),z2(T2)} < m1, and let
l ¼ minfz1 ðT 2 ÞðrL1 AM11 m1 AM13 M3 DM1 Þ; z2 ðT 2 ÞðrL2 AM22 m1 DM2 Þg:
671
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
There are three cases: (i) V(T2) = z1(T2) < m1,(z1(T2) < z2(T2)). (ii) V(T2) = z2(T2) < m1,(z2(T2) < z1(T2)). (iii) V(T2) = z1(T2) = z2(T2) < m1. If (i) holds, then there exists e > 0 such that if t 2 [T2,T2 + e) we have V(t) = z1(t) and
Dþ VðtÞ ¼ z01 ðtÞ > l > 0: If (ii) holds, similar to (i), there exists [T2, T2 + e) such that if t 2 [T2,T2 + e), we have V(t) = z2(t) and
Dþ VðtÞ ¼ z02 ðtÞ > l > 0: If (iii) holds, in the same way also there exists [T2, T2 + e) such that if t 2 [T2, T2 + e), we have V(t) = zi(t) (i = 1 or 2), and
Dþ VðtÞ ¼ z_ i ðtÞ > l > 0 ði ¼ 1 or 2Þ: From (i)-(iii), we know that if V(T2) < m1, V(t) will strictly monotonically in crease with speed l. So there exists T3 > T2 such that if t P T3, we have
VðtÞ ¼ minfz1 ðtÞ; z2 ðtÞg P m1 : From the third equation of system (3.1), we know that L L M z03 ðtÞ P z3 ðtÞ½r M 3 þ A31 m1 þ A32 m2 A33 z3 ðt s2 Þ; for t P T 3 þ s1 :
AL31 m1
AL32 m2
AL31 m1
AL32 m2
ð3:9Þ
AM 33 m3 ;
By (C5), þ rM þ rM ðk > 0Þ, then if z3(t) 6 m3 for all t P T3 + s1 + s2, (3.9) 3 > 0. Let k ¼ 3 implies that z03 ðtÞ > kz3 ðtÞ. This will lead to a contradiction. Hence there must exist a T4 P T3 + s1 + s2 such that z3(T4) > m3. If z3(t) > m3 for all t P T4, then the conclusion holds. If not, suppose z3 ðtÞ 6 m3 , where t > T 4 . Then from the above discussion there exists t* and t** such that
z3 ðt Þ ¼ z3 ðt Þ ¼ m3 and z3 ðtÞ < m3 for all t < t < t ; where T 4 6 t < t < t . Now suppose z3(t) with t* 6 t 6 t** attains its maximum at tˇ, t* < tˇ < t**. Then since z03 ðtÞ ¼ 0, (3.9) implies L L M r M 3 þ A31 m1 þ A32 m2 A33 z3 ðt s2 Þ 6 0;
this leads to
z3 ðt s2 Þ P
AL31 m1 þ AL32 m2 r M 3 AM 33
> m3 :
From (3.9) we have L L z03 ðtÞ=z3 ðtÞ P rM 3 þ A31 m1 þ A32 m2 :
Integrating the above inequality from t s2 to tˇ, we have L L L L M M z3 ðtÞ > z3 ðt s2 ÞeðA31 m1 þA32 m2 r3 Þs2 > m3 eðA31 m1 þA32 m2 r3 Þs2 > m3 ;
this contradicts to z3(t) < m3 for all t* < t < t**. So we have
z3 ðtÞ > m3 for all t P T 4 : Let T > T4, then for t P T, (3.7) holds. The proof is complete. h Now we formulate the uniqueness and global stability of the x-periodic solution x*(t) in Theorem 2.1. It is immediate that if x*(t) is globally asymptotically stable then x*(t) is in fact unique. Theorem 3.1. In addition to (C15), assume further that (C6) limt?1 inf Ai(t) > 0, where
Z tþs1 þs2 eM D 2 A31 ðt þ s1 ÞM3 A33 ðsÞds; m2 tþs1 Z tþs1 þs2 eM D A2 ðtÞ ¼ A22 ðtÞ A32 ðt þ s1 Þ 1 A32 ðt þ s1 ÞM3 A33 ðsÞds; ; m2 tþs1
A1 ðtÞ ¼ A11 ðtÞ A31 ðt þ s1 Þ
A3 ðtÞ ¼ A33 ðtÞ A13 ðtÞ A23 ðtÞ ðr3 ðtÞ þ A31 ðtÞM 1 þ A32 ðtÞM 2 þ A33 ðtÞM 3 Þ A33 ðt þ s2 ÞM 3
Z
Z
tþs2
A33 ðsÞds
t tþ2s2
A33 ðsÞds
ð3:10Þ
tþs2
Then system (1.2) has a unique positive x-periodic solution x ðtÞ ¼ ðx 1 ðtÞ; x 2 ðtÞ; y ðtÞÞT which is globally asymptotically stable.
672
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
Proof. Due to the conclusion of Theorem 2.1, we only need to show the global asymptotic stability of the positive periodic solution of (1.2). Let x ðtÞ ¼ ðx 1 ðtÞ; x 2 ðtÞ; y ðtÞÞT be a positive x-periodic solution of system (1.2), x(t) = (x1(t), x2(t), y(t))T be Q Q any positive solution of (1.2). Thus z ðtÞ ¼ ðz 1 ðtÞ; z 2 ðtÞ; z 3 ðtÞÞT ; ðz i ðtÞ ¼ 0 0, M1 and mi (defined by (3.4) and (3.8), respectively) such that for all t P T,
mi < z i ðtÞ 6 M i ;
mi < zi ðtÞ 6 Mi ; i ¼ 1; 2; 3:
ð3:11Þ
We define
V 1 ðtÞ ¼ j ln z 1 ðtÞ ln z1 ðtÞj þ j ln z 2 ðtÞ ln z2 ðtÞj:
ð3:12Þ
Calculating the upper-right derivative of V1(t) along the solution of (3.1), it follows for t P T that
Dþ V 1 ðtÞ ¼
2 0 X z ðtÞ i¼1
i z i ðtÞ
z i ðtÞ sgnðz i ðtÞ zi ðtÞÞ zi ðtÞ
e 1 ðtÞ z2 ðtÞ z2 ðtÞ þ sgnðz 2 ðtÞ ¼ sgnðz 1 ðtÞ z1 ðtÞÞ A11 ðtÞðz 1 ðtÞ z1 ðtÞÞ A13 ðtÞðz 3 ðtÞ z3 ðtÞÞþ D z1 ðtÞ z1 ðtÞ
e 2 ðtÞ z1 ðtÞ z1 ðtÞ z2 ðtÞÞ A22 ðtÞðz 2 ðtÞ z2 ðtÞÞ A23 ðtÞðz 3 ðtÞ z3 ðtÞÞ þ D z 2 ðtÞ z2 ðtÞ 6 A11 ðtÞjz 1 ðtÞ z1 ðtÞj þ ðA13 ðtÞ þ A23 ðtÞÞjz 3 ðtÞ z3 ðtÞj A22 ðtÞjz 2 ðtÞ z2 ðtÞj þ D11 ðtÞ þ D22 ðtÞ;
ð3:13Þ
where
D11 ðtÞ ¼
8 z2 ðtÞ z2 ðtÞ e > > < D 1 ðtÞ z 1 ðtÞ z1 ðtÞ ; z1 ðtÞ > z1 ðtÞ; > > :D e 1 ðtÞ z2 ðtÞ z2 ðtÞ ; z ðtÞ < z1 ðtÞ: 1 z1 ðtÞ z ðtÞ 1
8 e 2 ðtÞ z1 ðtÞ z1 ðtÞ ; z ðtÞ > z2 ðtÞ; > D > 2 ðtÞ ðtÞ z z < 2 2
D22 ðtÞ ¼
> > :D e 2 ðtÞ z1 ðtÞ z1 ðtÞ ; z ðtÞ < z2 ðtÞ: 2 z ðtÞ z2 ðtÞ 2
We estimate D11(t) under the following two cases: (i) If z 1 ðtÞ P z1 ðtÞ, then
D11 ðtÞ 6
e 1 ðtÞ eM D D jz ðtÞ z2 ðtÞj: ðz ðtÞ z2 ðtÞÞ 6 z 1 ðtÞ 2 m1 2
(ii) If z 1 ðtÞ < z1 ðtÞ, then
D11 ðtÞ 6
e 1 ðtÞ eM D D ðz2 ðtÞ z 2 ðtÞÞ 6 1 jz 2 ðtÞ z2 ðtÞj: z1 ðtÞ m1
Combining the conclusions in (i)–(ii), we obtain
D11 ðtÞ 6
eM D 1 jz ðtÞ z2 ðtÞj: m1 2
ð3:14Þ
A similar argument in the discussion above shows that
D22 ðtÞ 6
eM D 2 jz ðtÞ z1 ðtÞj: m2 1
ð3:15Þ
It follows from (3.14) and (3.15) that
Dþ V 1 ðtÞ 6 A11 ðtÞjz 1 ðtÞ z1 ðtÞj A22 ðtÞjz 2 ðtÞ z2 ðtÞj þ ðA13 þ A23 ðtÞÞðtÞjz 3 ðtÞ z3 ðtÞj þ þ
eM D 2 jz ðtÞ z1 ðtÞj: m2 1
eM D 1 jz ðtÞ z2 ðtÞj m1 2 ð3:16Þ
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
673
Define
V 21 ðtÞ ¼ j ln z 3 ðtÞ ln z3 ðtÞj:
ð3:17Þ
Calculating the upper-right derivative of V21(t) along the solution of (3.1), we derive for t P T that
z 03 ðtÞ z03 ðtÞ sgnðz 3 ðtÞ z3 ðtÞÞ ¼ sgnðz 3 ðtÞ z3 ðtÞÞ A33 ðtÞðz 3 ðt s2 Þ z3 ðt s2 ÞÞþA31 ðtÞðz 1 ðt s1 Þ z3 ðtÞ z3 ðtÞ
z1 ðt s1 ÞÞ þ A32 ðtÞðz 2 ðt s1 Þ z2 ðt s1 ÞÞ ¼ sgnðz 3 ðtÞ z3 ðtÞÞ A33 ðtÞðz 3 ðtÞ z3 ðtÞÞ þ A31 ðtÞðz 1 ðt s1 Þ Z t z1 ðt s1 ÞÞþA32 ðtÞðz 2 ðt s1 Þ z2 ðt s1 ÞÞ þ A33 ðtÞ ðz 03 ðuÞ z03 ðuÞÞdu ð3:18Þ
Dþ V 21 ðtÞ ¼
ts2
By substituting (3.1) into (3.18), we obtain Dþ V 21 ðtÞ ¼ sgnðz 3 ðtÞ z3 ðtÞÞ A33 ðtÞðz 3 ðtÞ z3 ðtÞÞ þ A31 ðz 1 ðt s1 Þ z1 ðt s1 ÞÞ þ A32 ðz 2 ðt s1 Þ Z t ½ðr3 ðuÞ þ A31 ðuÞz1 ðu s1 ÞþA32 ðuÞz2 ðu s1 Þ A33 ðuÞz3 ðu s2 ÞÞðz 3 ðuÞ z3 ðuÞÞ z2 ðt s1 ÞÞþA33 ðtÞ ts
2
þA31 ðuÞz 3 ðuÞðz 1 ðu s1 Þ z1 ðu s1 ÞÞ þ A32 ðuÞz 3 ðuÞðz 2 ðu s1 Þ z2 ðu s1 ÞÞA33 ðuÞz 3 ðuÞðz 3 ðu s2 Þ z3 ðu s2 ÞÞdu :
ð3:19Þ
It follows from (3.7) and (3.19) that for t P T + s
Dþ V 21 ðtÞ 6 A33 jz 3 ðtÞ z3 ðtÞj þ A31 ðtÞjz 1 ðt s1 Þ z1 ðt s1 Þj þ A32 ðtÞjz 2 ðt s1 Þ z2 ðt s1 Þj Z t ½ðr 3 ðuÞ þ A31 ðuÞM 1 þ A32 ðuÞM 2 þ A33 ðuÞM3 Þjz 3 ðuÞ z3 ðuÞj þ A31 ðuÞM3 jz 1 ðu s1 Þ þ A33 ðtÞ ts2
z1 ðu s1 Þj þ A32 ðuÞM 3 jz 2 ðu s1 Þ z2 ðu s1 Þj þ A33 ðuÞM 3 jz 3 ðu s2 Þ z3 ðu s2 Þjdu
ð3:20Þ
Define
V 22 ðtÞ ¼
Z
t
A31 ðs þ s1 Þjz 1 ðsÞ z1 ðsÞjds þ
ts1
Z
tþs2
Z
Z
t
ts1
A32 ðs þ s1 Þjz 2 ðsÞ z2 ðsÞjds
t
A33 fðr 3 ðuÞ þ A31 ðuÞM 1 þ A32 ðuÞM 2 þA33 ðuÞM 3 Þjz 3 ðuÞ z3 ðuÞj þ A31 ðuÞM 3 jz 1 ðu s1 Þ z1 ðu s1 Þj
ð3:21Þ þ A32 ðuÞM3 jz 2 ðu s1 Þ z2 ðu s1 ÞjþA33 ðuÞM3 jz 3 ðu s2 Þ z3 ðu s2 Þj du ds:
þ
ss2
t
It follows from (3.20) and (3.21) that for t P T + s
Dþ V 21 ðtÞ þ V 022 ðtÞ 6 A33 ðtÞjz 3 ðtÞ z3 ðtÞj þ A31 ðt þ s1 Þjz 1 ðtÞ z1 ðtÞj þ A32 ðt þ s1 Þjz 2 ðtÞ z2 ðtÞj Z tþs2 þ A33 ðsÞds ðr 3 ðtÞ þ A31 ðtÞM1 þ A32 ðtÞM2 þ A33 ðtÞM 3 Þjz 3 ðtÞ z3 ðtÞjþA31 ðtÞM 3 jz 1 ðt s1 Þ z1 ðt s1 Þj t
þA32 ðtÞM 3 jz 2 ðt s1 Þ z2 ðt s1 ÞjþA33 ðtÞM 3 jz 3 ðt s2 Þ z3 ðt s2 Þj
ð3:22Þ
We now define
V 2 ðtÞ ¼ V 21 ðtÞ þ V 22 ðtÞ þ V 23 ðtÞ;
ð3:23Þ
in which
V 23 ðtÞ ¼ M 3
Z
Z
t
ts1
uþs1 þs2
A33 ðsÞA31 ðu þ s1 Þjz 1 ðuÞ z1 ðuÞjds du þ M3
uþs1
Z
z2 ðuÞjds du þ M3
t
ts2
Z
uþ2s2
uþs2
Z
t
Z
ts1
uþs1 þs2
uþs1
A33 ðsÞA32 ðu þ s1 Þjz 2 ðuÞ
A33 ðsÞA33 ðu þ s2 Þjz 3 ðuÞ z3 ðuÞjds du:
ð3:24Þ
It follows from (3.22)–(3.24) that for t P T + s
Dþ V 2 ðtÞ 6 A33 ðtÞjz 3 ðtÞ z3 ðtÞj þ A31 ðt þ s1 Þjz 1 ðtÞ z1 ðtÞj þ A32 ðt þ s1 Þjz 2 ðtÞ z2 ðtÞj þ
Z t
tþs2
A33 ðsÞds ðr 3 ðtÞ þ A31 ðtÞM1 þ A32 ðtÞM2 þ A33 ðtÞM 3 Þjz 3 ðtÞ z3 ðtÞj
þ A31 ðt þ s1 ÞM 3
Z
tþs1 þs2
tþs1
þ A33 ðt þ s2 ÞM3
Z
tþ2s2
tþs2
A33 ðsÞdsjz 1 ðtÞ z1 ðtÞj þ A32 ðt þ s1 ÞM 3
A33 ðsÞdsjz 3 ðtÞ z3 ðtÞj :
Z
tþs1 þs2
tþs1
A33 ðsÞdsjz 2 ðtÞ z2 ðtÞj ð3:25Þ
674
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
Now, define a Lyapunove functional V(t) as
VðtÞ ¼ V 1 ðtÞ þ V 2 ðtÞ:
ð3:26Þ
Then it follows from (3.16), (3.25) and (3.26) that for t P T + s
Dþ VðtÞ 6
3 X
Ai ðtÞjz i ðtÞ zi ðtÞj;
ð3:27Þ
i¼1
where Ai(t) (i = 1, 2, 3) are defined in (3.10). By hypothesis (C6), there exist constants ai > 0 (i = 1, 2, 3) and T* P T + s such that
Ai ðtÞ P ai > 0 for t P T :
ð3:28Þ
Integrating both sides of (3.27) on interval [T*,t]
VðtÞ þ
3 Z X i¼1
T
t
Ai ðsÞjz i ðsÞ zi ðsÞjds 6 VðT Þ:
ð3:29Þ
It follows from (3.28) and (3.29) that 3 X
ai
i¼1
Z T
t
jz i ðsÞ zi ðsÞjds 6 VðT Þ < 1; for t P T :
ð3:30Þ
* * 0 Since z 0 i ðtÞ and zi ðtÞ ði ¼ 1; 2; 3Þ are bounded for t P T , so jzi ðtÞ zi ðtÞj ði ¼ 1; 2; 3Þ are uniformly continuous on [T , 1). By Barbalat’s lemma [17], we have
"
Y
lim jz i ðtÞ zi ðtÞj ¼ lim
t!1
t!1
0
" lim jz ðtÞ t!1 3
z3 ðtÞj ¼ lim
t!1
# ð1 þ bik Þ1 ðx i ðtÞ xi ðtÞÞ ¼ 0;
Y
ði ¼ 1; 2Þ
# ð1 þ b3k Þ
1
ðx 3 ðtÞ
x3 ðtÞÞ ¼ 0:
0
Hence
lim jx i ðtÞ xi ðtÞj ¼ 0; ði ¼ 1; 2Þ;
t!1
lim jy ðtÞ yðtÞj ¼ 0:
t!1
By Theorems 7.4 and 8.2 in [18], we know that the positive periodic solution x ðtÞ ¼ ðx 1 ðtÞ; x 2 ðtÞ; y ðtÞÞ of Eq. (1.2) is uniformly asymptotically stable. The proof of Theorem 3.1 is complete. h
4. An example In this section, we give an example to illustrate the feasibility of our main results. Example 4.1. We consider the following impulsive Lotka–Volterra predator–prey model with prey dispersal and time delays:
x01 ðtÞ ¼ x1 ðtÞ½8 þ sin t 5x1 ðtÞ 6yðtÞ þ 0:5½x2 ðtÞ x1 ðtÞ;
t – tk ;
x02 ðtÞ
¼ x2 ðtÞ½2 sin t 5x2 ðtÞ 5yðtÞ þ 3½x1 ðtÞ x2 ðtÞ; t – t k ; 1 1 1 þ x2 t 8y t ; y ðtÞ ¼ yðtÞ 1 þ ð3 þ sin tÞx1 t 10 10 1000 0
t – tk ;
ð4:1Þ
Dxi ðt k Þ ¼ bik xi ðt k Þ; i ¼ 1; 2; k ¼ 1; 2; . . . ; Dyðt k Þ ¼ b3k yðt k Þ; k ¼ 1; 2; . . . ; We fix the parameters bik ¼ 16 ; i ¼ 1; 2; b3k ¼ 18 ; t kþ2 ¼ t k þ 2p; ½0; 2p \ ftk g ¼ ft 1 ; t 2 g. Obviously, ðr2 D2 Þ ¼ 1. Thus, the result in Xu et al. [4] cannot be applied to this case. However, it is easy to verified that our result is feasible. In fact, simple computation shows that
A ¼ max and
( ) M M ðr 1 D1 ÞM þ DM 9 1 ðr 2 D2 Þ þ D2 ¼ ; ; 5 aL11 aL22
675
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
x1 x2 y
1.2 1 0.8 0.6 0.4 0.2 0
0
10
20
30
40
50
60
Fig. 1. The existence of positive 2p-periodic solution of system (4.1).
2 X
lnð1 þ bik Þ ¼ 2 ln 5 2 ln 6; i ¼ 1; 2;
k¼1
2 X
lnð1 þ b3k Þ ¼ 2 ln 7 2 ln 8:
k¼1
Therefore we have M M L aL33 ðr1 D1 Þx aM 13 Aða31 þ a32 Þx þ a33 B1 ¼ 12p 16ðln 6 ln 5Þ > 0;
and L M aL31 aM 140p 30ðln 6 ln 5Þ M 22 ½ðr 1 D1 Þx þ B1 þ a32 a11 ½ðr 2 D2 Þx þ B2 ¼ > 25 ¼ aM 11 a22 : r 3 x B3 2p þ 2 ln 8 2 ln 7
The above inequalities show that system (4.1) satisfies all the assumptions in Theorem 2.1. Thus, by Theorem 2.1, system (4.1) has at least one positive 2p-periodic solution.Numerical simulation shows that system (4.1) has at least one positive 2p-periodic solution (see Fig. 1). Therefore, in some sense, we generalize the results in [4]. 5. Conclusion In this paper, a model which describes the non-autonomous periodic predator–prey system with prey dispersal and impulse is proposed. By using coincidence degree theorem, a set of easily sufficient conditions are obtained for the existence of at least one positive periodic solution. We also present the uniqueness and global attractivity of positive periodic solution by means of Lyapunov functional method. From the view point of biology, the mathematical results are full of biological meanings and can be used to provide reliable foundation for making control strategy. The conditions of Theorem 2.1 show that human activities may save the extinct species living in a week environment (r2 D2 < 0, see Example 4.1). Hence, the results obtained imply impulsive perturbation is one of important control factors in determining the existence of periodic solutions of system (2.1). Numerical simulation is given to illustrate the feasibility of our main result. Acknowledgements The authors would like to thank the anonymous referees for their careful reading of the manuscript and excellent comments, which improved the presentation of this paper. Also, this work is supported by the National Science Foundation of China (10871062). References [1] D.D. Bainov, P.S. Simeonov, Impulsive Differential Equations: Periodic Solutions and Applications, Longman Scientific, Technical, New Yoke, 1993. [2] L.L. Wang, W.T. Li, Periodic solutions and permanence for a delayed nonautonomous ratio-dependent predator–prey model with Holling type functional response, J. Comput. Appl. Math. 162 (2004) 341–357.
676
R. Liang, J. Shen / Applied Mathematics and Computation 217 (2010) 661–676
[3] S. Chen, J. Zhang, T. Yong, Existence of positive periodic solution for nonautonomous predator–prey system with diffusion and time delay, J. Comput. Appl. Math. 159 (2003) 375–386. [4] R. Xu, M.A.J. Chaplain, F.A. Davidson, Periodic solution for a delayed predator–prey model of prey dispersal in two-patch environments, Nonlinear Anal. RWA 5 (2004) 183–206. [5] J. Zhen, Zh. Ma, M. Han, The existence of periodic solutions of the n-species Lotka–Volterra competition systems with impulsive, Chaos, Solitons & Fractals 22 (2004) 181–188. [6] J. Zhang, L. Chen, X. Chen, Persistence and global stability for two-species nonautonomous competition Lotka–Volterra patch-system with time delay, Nonlinear Anal. TMA 37 (1999) 1019–1028. [7] R. Xu, L. Chen, Persistence and stability for a two-species ratio-dependent predator–prey system with time delay in a two-patch environment, Comput. Math. Appl. 40 (2000) 577–588. [8] X.Y. Song, L.S. Chen, Persistence and global stability for nonautonomous predator–prey systems with diffusion and time delay, Comput. Math. Appl. 35 (1998) 33–40. [9] M. Kouche, N. Tatar, Existence and global attractivity of a periodic solution to a nonautonomous dispersal system with delay, Appl. Math. Model. 31 (2007) 780–793. [10] S. Lenci, G. Rega, Periodic solutions and bifurcations in an impact inverted pendulum under impulsive excitation, Chaos, Solitons & Fractals 11 (2000) 2453–2472. [11] X. Liu, L. Chen, Complex dynamics of Holling type II Lotka–Volterra predator–prey system with impulsive perturbations on the predator, Chaos, Solitons & Fractals 16 (2003) 311–320. [12] W. Sun et al, On positive periodic solution of periodic competition Lotka–Valterra system with time delay and diffusion, Chaos, Solitons & Fractals 33 (2007) 971–978. [13] Z. Teng, Z. Lu, The effect of dispersal on single-species nonautonomous dispersal models with delays, J. Math. Biol. 42 (2001) 439–454. [14] R.E. Gaines, J.L. Mawhin, Coincidence Degree and Nonlinear Differential Equations, Springer-Verlag, Berlin, 1977. [15] J. Yan, A. Zhao, oscillation and stability of linear impulsive delay differential equation, J. Math. Anal. Appl. 227 (1998) 187–194. [16] W. Wang, Z. Ma, Harmless delays for uniform persistence, J. Math. Anal. Appl. 158 (1991) 256–268. [17] L. Barbalat, Systemes d’equations differentielles d’oscillations nonlineaires, Rev. Roumaine Math. Pures Appl. 4 (1959) 267–270. [18] T. Yoshizawa, Stability Theory by Liapunov’s Second Method, Math. Soc. Japan, Tokyo, 1966. [19] J.G. Skellam, Random dispersal in theoretical populations, Biometrika 38 (1951) 16–218.