global asymptotic stability of a second order rational difference equation
TRANSCRIPT
Applied Mathematics and Computation 233 (2014) 377–382
Contents lists available at ScienceDirect
Applied Mathematics and Computation
journal homepage: www.elsevier .com/ locate/amc
Global asymptotic stability of a second order rational differenceequation
http://dx.doi.org/10.1016/j.amc.2014.01.1770096-3003/� 2014 Elsevier Inc. All rights reserved.
⇑ Corresponding author.E-mail address: [email protected] (L.-X. Hu).
Lin-Xia Hu ⇑, Hong-Ming XiaDepartment of Mathematics, Tianshui Normal University, Tianshui, Gansu 741001, People’s Republic of China
a r t i c l e i n f o
Keywords:Difference equationBoundednessPeriod-two solutionStableGlobal attractorGlobally asymptotically stable
a b s t r a c t
The main goal of this paper is to investigate the global asymptotic stability of the differenceequation
ynþ1 ¼pn þ yn
pn þ yn�1; n ¼ 0;1;2; . . . ;
where
pn ¼a; if n is evenb; if n is odd
�and a > 0; b > 0; a–b
and the initial conditions y�1; y0 2 ½0;1Þ . We show that the unique equilibrium �y ¼ 1 isglobally asymptotically stable under the certain conditions.
� 2014 Elsevier Inc. All rights reserved.
1. Introduction and preliminaries
Our aim in this paper is to investigate the global stability of the following difference equation
ynþ1 ¼pn þ yn
pn þ yn�1; n ¼ 0;1;2; . . . ; ð1:1Þ
where
pn ¼a; if n is evenb; if n is odd
�and a > 0; b > 0; a – b
and the initial conditions y0; y�1 2 ½0;1Þ. In [10, Open Problem 4.8.12], Kulenovic and Ladas proposed the following openproblem. See also [3, Open Problem 5.26.1].
Open Problem 1.1. Let fpng1n¼0 be a periodic-two sequence of nonnegative real numbers. Investigate the global character of
all positive solutions of Eq. (1.1).Recently, systematic analysis of difference equations with periodic coefficients was considered by several authors, see, for
example Refs. [6,7,11]. Inspired by the above open problem, we investigate the global stability of Eq. (1.1) with a periodic-two coefficients, which is the non-autonomous equation that corresponds to the following autonomous equation
378 L.-X. Hu, H.-M. Xia / Applied Mathematics and Computation 233 (2014) 377–382
ynþ1 ¼pþ yn
pþ yn�1; n ¼ 0;1;2; . . . ; p > 0; y�1; y0 2 ½0;1Þ: ð1:2Þ
In [9], it is shown that the unique equilibrium �y ¼ 1 of Eq. (1.2) is globally asymptotically stable. See also [8, p.73] and [3,p.168]. Our main goal in this paper is to extend the result to the non-autonomous difference Eq. (1.1) under the certainconditions.
Before our discussion, we present some definitions and the known results which will be useful in the sequel. For the gen-eral theory of difference equations, one can refer to the monographes of Kocic and Ladas [8] and Kulenovic and Ladas [10].For other related results on nonlinear difference equations, see, for example, [1–16] and the references therein.
Consider the system:
unþ1 ¼ f ðun; vnÞ;vnþ1 ¼ gðun;vnÞ:
�n ¼ 0;1; . . . : ð1:3Þ
Let k � k be the norm of vector ðu;vÞ 2 R2. Then, we present some definitions and some useful lemmas.
Definition 1.2. The equilibrium point ð�u; �vÞ is said to be:
(i) stable if given � > 0 and N > 0 there exists d > 0 such that kðu0;v0Þ � ð�u; �vÞk < d implies that kðun;vnÞ � ð�u; �vÞk < � forall n > N, and unstable if it is not stable;
(ii) attracting if there exists g > 0 such that kðu0;v0Þ � ð�u; �vÞk < g implies that limn!1ðun;vnÞ ¼ ð�u; �vÞ;(iii) asymptotically stable if it is stable and attracting.
The main result in linearized stability analysis is following lemma which is extracted from [4], see also [10].
Lemma 1.3. Let F ¼ ðf ; gÞ be a continuously differentiable function defined on an open set D � R2.
(a) If the eigenvalues of the Jacobian matrix JFðð�u; �vÞÞ, that is, both roots of its characteristic equation
k2 � TrJFðð�u; �vÞÞkþ DetJFðð�u; �vÞÞ ¼ 0; ð1:4Þ
lie inside the unit disk, then the equilibrium ð�u; �vÞ of Eq. (1.3) is locally asymptotically stable.(b) A necessary and sufficient condition for both roots of Eq. (1.4) to lie inside the unit disk is
jTrJFðð�u; �vÞÞj < 1þ DetJFðð�u; �vÞÞ < 2:
In this case, the locally asymptotically stable equilibrium ð�u; �vÞ is also called a sink.
The following Lemma can be founded in [12].
Lemma 1.4. Let B ¼ ½a; b� � ½c; d� 2 R2. Assume that F ¼ ðf ; gÞ : B! B is a continuous function satisfying the following properties:
(i) f ðu;vÞ is non-increasing in the first variable and non-decreasing in the second variable and g is non-increasing in each of itsvariables for each ðu;vÞ 2 B;
(ii) If ðm;M; l; LÞ 2 B� B is a solution of the system
M ¼ f ðm; LÞ; m ¼ f ðM; lÞ;L ¼ gðm; lÞ; l ¼ gðM; LÞ:
�
then m ¼ M and l ¼ L.
Then the system (1.3) has a unique equilibrium ð�u; �vÞ and every solution of Eq. (1.3) with ðu0;v0Þ 2 B converges to the equilibriumð�u; �vÞ.
2. Period-two character and linearized stability
Theorem 2.1. Eq. (1.1) has no positive prime period-two solution.
Proof. Assume for the sake of contradiction that
. . . ;/;u;/;u; . . .
is a positive prime period-two solution of Eq. (1.1), then
L.-X. Hu, H.-M. Xia / Applied Mathematics and Computation 233 (2014) 377–382 379
/ ¼ aþuaþ /
; u ¼ bþ /bþu
;
from which it follows that
a/þ /2 ¼ aþu; buþu2 ¼ bþ /:
So yields
ða/þ /2Þðbþ /Þ ¼ ðaþuÞðbuþu2Þ
and
ð/�uÞ½/2 þ /uþu2 þ ðaþ bÞð/þuÞ þ ab� ¼ 0:
Thus / ¼ u, a contradiction.
The proof is complete. h
Clearly, Eq. (1.1) has the unique equilibrium �y ¼ 1. To determine the stability of the equilibrium �y, we set
un ¼ y2n�1; vn ¼ y2n for n ¼ 0;1;2; . . .
and write Eq. (1.1) in the following equivalent form
unþ1 ¼ aþvnaþun
;
vnþ1 ¼ aþabþbunþvnðaþunÞðbþvnÞ ;
(n ¼ 0;1;2; . . . : ð2:1Þ
Applying Theorem 2.1, we are sure that Eq. (2.1) has a unique equilibrium E� ¼ ð1;1Þ in the first quadrant.Consider the map T : ½0;1Þ2 ! ½0;1Þ2 associated to Eq. (2.1), i.e.
Tðu; vÞ ¼T1ðu;vÞT2ðu;vÞ
� �¼
aþvaþu
aþabþbuþvðaþuÞðbþvÞ
" #:
Calculating the partial derivatives of the function T1ðu;vÞ and T2ðu;vÞ shows that
@T1
@u¼ � aþ vðaþ uÞ2
;@T1
@v ¼1
aþ u;
@T2
@u¼ � aþ vðaþ uÞ2ðbþ vÞ
;@T2
@v ¼b� a� ab� bu
ðaþ uÞðbþ vÞ2:
Clearly, @T1@u ;
@T2@u < 0; @T1
@v > 0 for u;v > 0, and @T2@v P 0 for u 6 1� a� a=b; @T2
@v 6 0 for u P 1� a� a=b.
The Jacobian matrix of T evaluated at E� is
JTðE�Þ ¼
� 1aþ1
1aþ1
� 1ðaþ1Þðbþ1Þ � a
ðaþ1Þðbþ1Þ
" #
and its characteristic equation associated with E� is
k2 þ aþ bþ 1ðaþ 1Þðbþ 1Þ kþ
1ðaþ 1Þðbþ 1Þ ¼ 0:
Noticing that inequality j aþbþ1ðaþ1Þðbþ1Þ j < 1þ 1
ðaþ1Þðbþ1Þ < 2 holds and applying Lemma 1.3, the following statement is true.
Theorem 2.2. The equilibrium E� of Eq. (2.1) is locally asymptotically stable for all values a; b 2 ð0;1Þ.
Theorem 2.3. Eq. (2.1) has no positive prime period-two solution.
Proof. Assume for the sake of contradiction that
. . . ; ð/1;u1Þ; ð/2;u2Þ; ð/1;u1Þ; ð/2;u2Þ . . .
is a positive prime period-two solution of Eq. (2.1), then they satisfy the following systems
/1 ¼aþu2
aþ /2; /2 ¼
aþu1
aþ /1
380 L.-X. Hu, H.-M. Xia / Applied Mathematics and Computation 233 (2014) 377–382
and
u1 ¼bþ /1
bþu2; u2 ¼
bþ /2
bþu1:
Then
a/1 þ /1/2 ¼ aþu2; a/2 þ /1/2 ¼ aþu1
bu1 þu1u2 ¼ bþ /1; bu2 þu1u2 ¼ bþ /2:
So
að/1 � /2Þ ¼ u2 �u1; and bðu1 �u2Þ ¼ /1 � /2;
from which it follows that
ð/1 � /2Þðu1 �u2Þðabþ 1Þ ¼ 0:
Clearly, u1 ¼ u2 () /1 ¼ /2 and this will lead to a contradiction.The proof is complete. h
3. Global asymptotic stability
Let fðun;vnÞg be a solution of Eq. (2.1). Clearly, the second equation of system (2.1) implies that for n P 0,
vnþ1 ¼aþ abþ bun þ vn
ðaþ unÞðbþ vnÞ6
maxfaþ ab;b;1gminfab;a;bg
and
vnþ2 ¼bþ unþ2
bþ vnþ1P
bbþ vnþ1
:
Set
Sv ¼maxfaþ ab; b;1g
minfab;a;bg ; Iv ¼b
bþ Sv: ð3:1Þ
Then Iv 6 vn 6 Sv for n P 2.On the other hand, for n P 2, the first equation of system (2.1) implies that
unþ1 ¼aþ vn
aþ un6
aþ vn
a6
aþ Sv
a¼ 1þ 1
aSv
and
unþ2 ¼aþ vnþ1
aþ unþ1P
aþ b=ðbþ SvÞaþ 1þ ð1=aÞSv
¼ abþ a2bþ a2Sv
ða2 þ aþ SvÞðbþ SvÞ:
Set
Iu ¼abþ a2bþ a2Sv
ða2 þ aþ SvÞðbþ SvÞ; Su ¼ 1þ 1
aSv : ð3:2Þ
Then Iu 6 un 6 Su for n P 4.Especially, if a < b 6 aþ ab, then 1� a� a=b 6 0 and un P 1� a� a=b hold for n P 0. Using the monotonic character
that the function T2 is decreasing in u, we can get that for n P 0,
vnþ1 ¼aþ abþ bun þ vn
ðaþ unÞðbþ vnÞ6
aþ abþ bð1� a� a=bÞ þ vn
ðaþ 1� a� a=bÞðbþ vnÞ¼ b
b� a:
If b > aþ ab, then 1� a� a=b > 0 holds. Similarly, applying the monotonic character of the function T2, we get, for someN, if uN P 1� a� a=b, then vNþ1 6
bb�a, and if uN 6 1� a� a=b, then vNþ1 6
bb�a.
So in the case where b > a yields
vn 6b
b� a
and
vnþ1 ¼bþ unþ1
bþ vnP
bbþ b=ðb� aÞ ¼
b� ab� aþ 1
;
L.-X. Hu, H.-M. Xia / Applied Mathematics and Computation 233 (2014) 377–382 381
unþ1 ¼aþ vn
aþ un6
aþ vn
a6
aþ b=ðb� aÞa
¼ aðb� aÞ þ baðb� aÞ
for n P 1. Moreover, we get
unþ2 ¼aþ vnþ1
aþ unþ1P
aþ ðb� aÞ=ðb� aþ 1Þaþ ðabþ b� a2Þ=ðab� a2Þ ¼
aðab� a2 þ bÞðb� aÞða2 þ aÞðb� aÞ2 þ ða2 þ aþ bÞðb� aÞ þ b
>aðb� aÞ
ðaþ 1Þðb� aÞ þ 1:
So we have the following conclusion.
Theorem 3.1. Every solution fðun;vnÞg of Eq. (2.1) eventually enters an invariable bounded rectangle B� ¼ ½Iu; Su� � ½Iv ; Sv �, whereIv ; Sv and Iu; Su are defined by (3.1) and (3.2). Further, if b > a, then
B ¼ aðb� aÞðaþ 1Þðb� aÞ þ 1
;aðb� aÞ þ b
aðb� aÞ
� �� b� a
b� aþ 1;
bb� a
� �ð3:3Þ
is also an invariable rectangle of Eq. (2.1).Now we turn to the investigation of the global asymptotic stability of the equilibrium point E�.
Theorem 3.2. Assume that ab P 1 and b < aðbþ 1Þ þ 12 a2ðbþ 1Þ. Then the unique positive equilibrium E� of Eq. (2.1) is globally
asymptotically stable. The global stable manifold is given by WsðE�Þ ¼ fðu;vÞ : u P 0;v P 0g.
Proof. Let fðun;vnÞg be a solution of Eq. (2.1). In view of Theorems (2.2), it is sufficient to show that every nonnegative solu-tion converges to E�. In order to achieve our aim, we divide the proof of the theorem into the following two cases.
Case (i) b 6 aðbþ 1Þ.The condition that b 6 aðbþ 1Þ implies that 1� a� a=b 6 0, and thus un P 1� a� a=b for n P 0. From Theorem 3.1,
every solution fðun;vnÞgwith initial point ðu0;v0Þ 2 fðu;vÞ : u P 0;v P 0g eventually enters an invariant bounded rectangleB� ¼ ½Iu; Su� � ½Iv ; Sv �. In this rectangle B�, the function f is non-increasing in the first argument and non-decreasing in thesecond, and the function g is non-increasing in each of its arguments. Let ðm;M; l; LÞ be a solution of the following system.
M ¼ aþ Laþm
; m ¼ aþ laþM
ð3:4Þ
and
L ¼ aþ abþ bmþ lðaþmÞðbþ lÞ ; l ¼ aþ abþ bM þ L
ðaþMÞðbþ LÞ : ð3:5Þ
Then in view of Lemma 1.4, to establish attractivity, we must show that m ¼ M and l ¼ L.Eq. (3.4) implies that
L ¼ ðaþmÞM � a; and l ¼ ðaþMÞm� a ð3:6Þ
and
L� l ¼ aðM �mÞ: ð3:7Þ
Eq. (3.5) implies that
ðbþ lÞL ¼ bþ aþ laþm
; and ðbþ LÞl ¼ bþ aþ LaþM
; ð3:8Þ
from which it follows that
bðL� lÞ ¼ aþ laþm
� aþ LaþM
: ð3:9Þ
Substituting expressions (3.6) and (3.7) into Eq. (3.9) leads to the following equation
ðM �mÞ½a2 þ a3bþ a2bM þ a2bmþ ðab� 1ÞmM� ¼ 0:
Notice that ab P 1, thus m ¼ M and l ¼ L. In view of Lemma 1.4, we get that the equilibrium E� of Eq. (2.1) is a global attrac-tor. Further,
WsðE�Þ ¼ fðu; vÞ : u P 0; v P 0g:
Case (ii) aðbþ 1Þ < b < aðbþ 1Þ þ 12 a2ðbþ 1Þ.
From Theorem 3.1, we get that, under the condition that b > aðbþ 1Þ, every nonnegative solution fðun;vnÞg of Eq. (2.1)eventually enters an invariant bounded rectangle B which is defined by (3.3). Thus there exists a positive integer N such thatun P aðb�aÞ
ðaþ1Þðb�aÞþ1 for n > N.
382 L.-X. Hu, H.-M. Xia / Applied Mathematics and Computation 233 (2014) 377–382
Notice that b > aþ ab implies that b� a > ab P 1 , and b < aðbþ 1Þ þ 12 a2ðbþ 1Þ implies that aþ a
b ¼ ab ðbþ 1Þ > 2
aþ2,thus
un Paðb� aÞ
ðaþ 1Þðb� aÞ þ 1P
aaþ 2
¼ 1� 2aþ 2
> 1� a� ab; for n > N:
Therefore, in the invariant rectangle B, the function f is non-increasing in the first argument and non-decreasing in thesecond, and the function g is non-increasing in each of its arguments. Applying Lemma 1.4, then as the same argument inthe case (i), one can show that the result follows when ab P 1.
The proof is complete. h
As far as Eq. (1.1) is concerned, the statement in Theorem 3.2 can be described by the following Theorem.
Theorem 3.3. Assume that ab P 1 and b < aðbþ 1Þ þ 12 a2ðbþ 1Þ. Then the unique positive equilibrium �y ¼ 1 of Eq. (1.1) is
globally asymptotically stable.
References
[1] R.P. Agarwal, W.T. Li, P.Y.H. Pang, Asymptotic behavior of a class of nonlinear delay difference equations, J. Differ. Equ. Appl. 8 (2002) 719–728.[2] D. Clark, M.R.S. Kulenovic, On a coupled system of rational difference equations, Comput. Math. Appl. 43 (2002) 849–867.[3] E. Camouzis, G. Ladas, Dynamics of Third-Order Rational Difference Equations with Open Problems and Conjectures, Chapman Hall/CRC, Boca Raton,
2007.[4] S. Elaydi, An Introduction to Difference Equations, 3rd ed., Science+Business Media Inc., 2005.[5] C.H. Gibbons, M.R. Kulenovic, G. Ladas, On the recursive sequence xnþ1 ¼ ðaþ bxn�1Þ=ðcþ xnÞ, Math. Sci. Res. Hot-Line 4 (2) (2000) 1–11.[6] C.H. Gibbons, C.B. Overdeep, On the trichotomy character of xnþ1 ¼ ðan þ cnxn�1Þ=ðAn þ BnxnÞ with period-two coefficients, J. Differ. Equ. Appl. 11 (7)
(2005) 655–667.[7] L.X. Hu, W.S. He, H.M. Xia, Global asymptotic behavior of a rational difference equation, Appl. Math. Comput. 218 (2010) 7818–7828.[8] V.L. Kocic, G. Ladas, Global Behavior of Nonlinear Difference Equations of Higher Order with Application, Kluwer Academic Publishers, Dordrecht, 1993.[9] V.L. Kocic, G. Ladas, I.W. Rodrigues, On rational recursive sequences, J. Math. Anal. Appl. 173 (1993) 127–157.
[10] M.R. Kulenovic, G. Ladas, Dynamics of Second Order Rational Difference Equations with Open Problems and Conjectures, Chapman Hall/CRC, BocaRaton, 2002.
[11] M.R. Kulenovic, G. Ladas, C.B. Overdeep, On the dynamics of xnþ1 ¼ pn þ xn�1xn
with a preiod-two coefficient, J. Differ. Equ. Appl. 10 (2004) 905–914.[12] M.R.S. Kulenovic, O. Merino, Global attractivity result for maps with invariant boxes, Discrete Continuous Dyn. Syst. Ser. B 6 (2006) 97–110.[13] W.T. Li, H.R. Sun, Global attractiveness in a rational recursive equation, Dyn. Syst. Appl. 11 (2002) 339–345.[14] W.T. Li, H.R. Sun, Dynamics of a rational difference equation, Appl. Math. Comput. 163 (2005) 577–591.[15] Y.H. Su, W.T. Li, Global attractivity of a higher order nonlinear difference equation, J. Differ. Equ. Appl. 11 (2005) 947–958.[16] Y.H. Su, W.T. Li, S. Stevic, Dynamics of a high order nonlinear rational difference equation, J. Differ. Equ. Appl. 11 (2005) 133–150.