Australian Journal of Basic and Applied Sciences, 5(11): 72-77, 2011

ISSN 1991-8178

Stability and Convergence of A Higher Order Rational Difference Equation

Hossein Shojaei, Saeed Parvandeh, Tayebeh Mohammadi, Zeinab Mohammadi and

Narjes Mohammadi

Department of Mathematics, Mobarakeh Branch, Islamic Azad University, Mobarakeh, Isfahan, Iran.

Abstract: In this paper the asymptotic stability of equilibria and periodic points of the following higher
order rational difference equation

, 1, 0,1, …
…

is studied where the parameters , , are positive real numbers, and the initial conditions ,…,
are given arbitrary real numbers. The forbidden set of this equation is found and then, the order reduction
method is used to facilitate the analysis of its asymptotic dynamics.

Key words: Difference equation; Equilibrium point; Periodic point; Convergence; semiconjugacy.

INTRODUCTION

Difference equations may appear as solutions of various phenomena or as a discretized system of delay or
non-delay differential equations. They are very important in both theory and applications; for applications in
biology see (Brauer et al., 2001), in economics see (Ishiyama et al., 2005) and (Matti, 1996), in medical sciences
see (Moye, et al., 2000), and in military sciences see (Epstein, 1985). The main goal in the study of difference
equations is to understand the asymptotic behavior of solutions rather than trying to find an explicit formula for
solutions. This basically is not only because the explicit solutions are hard to find but also the explicit solutions
may still represent a complex dynamics and they yet may require a qualitative analysis to understand their
dynamics.
The ratio of any two polynomials of a recursive sequence is called a rational difference equation. Rational
difference equations are important as practical classes of difference equations. Most of the works in the literature
of rational difference equations have treated the first and second order difference equations. For second order
rational difference equations we refer the reader to the monograph of Kullenovic and Ladas, 2002. In this paper
we study the dynamics of the following 1 -order rational difference equation

, 0,1, … (1)
…

where 1 and the parameters , , are positive. We allow the initial conditions , ,… to take
any arbitrary value out of the forbidden set of equation (1). Let be some interval of positive real numbers and
: be a continuously differentiable function. Then for every initial conditions , ,…,
, the difference equation

, ,…, , 0,1, … (2)

∞
has a unique solution . The point is called an equilibrium of equation (2) (or simply an
equilibrium of ) if , ,…, , i.e., for all 0 (such a solution is also called a trivial
solution). The point ,…, is called a 1 -cycle if for all 0,1, … , . In this
case we say that is periodic with period 1 . The linearized equation associated with equation (2) about
the equilibrium is

,… ,

and its corresponding characteristic equation is defined by

Corresponding Author: Hossein Shojaei, Department of Mathematics, Mobarakeh Branch, Islamic Azad University,
Mobarakeh, Isfahan, Iran.
E-mail: [email protected]
72
 Aust. J. Basic & Appl. Sci., 5(11): 72-77, 2011

∑ ,… 0. (3)

An equilibrium of equation(2) is called locally stable if for every 0, there exists 0 such that for
∞
the solution of equation (2) with | | | | | | we have | |
∞
for all 1. Furthermore, if there exists 0 such that for the solution of equation (2) with
| | | | | | we have lim ∞ , then is called locally asymptotically
stable. Linearized stability theorem indicates that if all roots of equation (3) lie inside the open unit disk | | 1,
then the equilibrium of equation (2) is locally asymptotically stable. If at least one of the roots of equation (3)
has modulus greater than one, then the equilibrium of equation (2) is unstable, see e.g., (Kocic, et al., 1993).
∞
is called a global attractor on an interval if for every solution of equation (2) with
, ,…, , we have lim ∞ . When is locally stable and a global attractor, we call it
globally asymptotically stable. The forbidden set of equation (2) is the set of all (k+1)-tuples , ,… ,
where they can not be taken as the initial conditions for an infinite well-defined sequence on its domain. In
other words, the forbidden set of equation (2) is the set of all initial conditions such that arbitrary iterations of the
right hand side of equation (2) are not well-defined.
The following definitions, lemma, corollary, and theorems are needed to study the global behavior of
solutions of equation (1).

Definition 1:
(for original ideas see (Elaydi, 2000) or (Sedaghat, 2003)) Consider equation (2) and assume that
is nonempty.
1. Suppose that ,… . Then max ,…, denotes the sup-norm of . Also, if
is the equilibrium of then ,…, is an equilibrium of its vectorization that is

,… ,… , ,…, .

2. If there is a non-constant function , such that on for some ,

then is also called a , , -semiconjugate of . The mapping is called a link map and is
called the factor map. For each , the level set , abbreviated as is called a fiber of
in .
3. A continuous mapping : is said to be bending at a point if is not an isolated point in
and , where denotes the interior of .
4. An equilibrium of equation (2) is exponentially stable under relative to some nontrivial interval
containing if there is 0,1 such that for every solution of equation (2) with initial values
,…, we have for all 1 that | | where ,… is independent of .
5. An equilibrium of a map : is semistable (from the right) if for any 0 there exists 0 such
that if 0 then | | for all 1. If in addition, lim ∞ whenever
0 for some 0 , then is said to be semiasymptotically stable (from the right).
Semistability (semiasymptotic stability) from the left is defined analogously.

Theorem 1.1:
(see (Sedaghat, 2003)). Let is a , , -semiconjugate map and is an equilibrium of .
1. is an equilibrium of .
2. (Boundedness). Assume that | | ∞ as ∞. If the sequence is bounded for some
, then the sequence with is bounded.
3. (Stability and instability). Assume that is bending at . If we set then
(a) If is stable (asymptotically stable) under , then is a stable (asymptotically stable) equilibrium of
.
(b) If is unstable under , then is unstable under .
4. (Attractivity of invariant fibers). Let be an isolated equilibrium of which attracts all points in . If
and is bounded, then converges to the invariant fiber .

2. The Forbidden Set:

Consider equation (1). If 0, the solution is trivial. If 0 then equation (1) reduces to a linear
equation. If 0 then the solution is periodic with period 1. Thus, we assume that all the parameters are
nonzero. A change of variables followed by the change reduces equation (1) to

73
 Aust. J. Basic & Appl. Sci., 5(11): 72-77, 2011

, 0,1, … (4)
…

where . Hence, , we consider equation (4) instead of equation (1), hereafter. Now we are ready to
obtain the forbidden set of equation (4).

Theorem 2.1:
Consider equation (4). Assume that is the forbidden set for this equation. Then
∞
1
, ,…, : …
∑

Proof. Assume the initial conditions , ,… satisfy

1
…
∑

for some 0 . Then equation (4) implies that

1
… ∑ 1
… .
1 … 1 ∑ ∑

Therefore, we obtain from equation (4) that

1
… ∑ 1
… .
1 … 1 ∑ ∑

Continuing in this fashion by induction we obtain that

1
… 1.
∑

As a result the iteration process stops at . Conversely, assume iteration process stops at some point
, i.e., … 1. Then equation (4) implies that

1
… .
1

Again using this fact and equation (4) we obtain that

1
… .
1

Continuing in this manner we get

1
…
∑

or equivalently , ,… . This completes our inductive proof.

3. Linearized Stability:
The first step to understand of the dynamics is to find the equilibria and find their stability type. In this section
we investigate the local asymptotic stability of the equilibria of equation (4). Simple calculations show that origin
is always an equilibrium for equation (4) and if 1 then it has a second equilibrium √ 1. The
following theorem deals with the local asymptotic stability of the equilibria. Note that throughout the rest of this
paper, the initial conditions are assumed to be taken out of the forbidden set .

74
 Aust. J. Basic & Appl. Sci., 5(11): 72-77, 2011

Theorem 3.1:
Consider equation (4). Then,
(a) origin is locally asymptotically stable for 1, while it is unstable when 1.
(b) For the case of 1, the positive equilibrium √ 1 is stable.

Proof. For part (a), the linearized equation associated with equation (4) about origin is
0, 0. Therefore, the corresponding characteristic equation is 0, i.e., √ . Since the origin is
locally asymptotically stable if 1 and is unstable if 1.
(b) Simple calculations show that the linearized equation associated with equation (4) about the positive
equilibrium √ 1 is

1 1 1 1
0, 0,1, …

Therefore the corresponding characteristic equation is

0, (5)

Some algebra show that equation (5) is equivalent to

/
0, 1, (6)

Therefore, 1/ is one of the roots of equation (6). Since 1, by local asymptotic stability Theorem
we conclude that the positive equilibrium √ 1 is unstable. The proof is complete.

Remark 1:
The above theorem is sufficient to completely describe the local dynamics of the equilibrium for the case
1. However, for the case 1, more investigation are needed. For 1 the characteristic equation about
origin has modulus equal to one, the linearization fails to analyze the asymptotic stability of origin. Thus, it may
have a complex dynamics near this point. On the other hand consider equation (6) and assume that 1.
Equation (6) has a real root 1/ with modulus less than one but this equation has 1 roots in the
following form

2 2
cos sin , 0,1, … , ,
1 1

all with modulus equal to one. So, if 1 then linearization tells us nothing about the stability of the positive
equilibrium. In the next section we discuss these cases in details.

3. Semiconjugate Factorization:
The main purpose of this section is to analyze the local dynamics near equilibria for the cases 1 as well
as of the global asymptotic dynamics of the equation (4). Although the previous section, using linearization,
showed that the origin is locally asymptotically stable for 1. Yet the global nature of asymptotic stability can
not be inferred from linearization. Also, we saw that for 1 the linearized equation about the positive
equilibrium has several roots; all with modulus equal to unity. As a result, more powerful methods are necessary
in order to analyze the dynamics of equation (4). In this section, we apply semiconjugacy analysis to examine the
global nature of equation (4). The idea is to reduce the order of a higher order difference equation such that the
analysis of the reduced system is feasible. Then, this is helpful when this analysis can facilitate the understanding
of the dynamics for the original difference equation.

Let , ,…, . Consider the vectorization of i.e.,

…

, ,…, , ,…, .
1 …

Set , ,…, … and note that

75
 Aust. J. Basic & Appl. Sci., 5(11): 72-77, 2011

… , ,…,
, ,…, ,
1 … 1 , ,…,

So the map makes a , , -semiconjugate map with being the nonnegative orthant of ,
i.e., 0, ∞ , and the map

,
1

serving as the factor map on 0, ∞ . Therefore, equation (4) is a semiconjugae factorization of the well-known
Ricatti difference equation . In the sequence, we consider three cases as follow:

Case I:
0 1; In this case we claim that

/
0 max , ,…, , ,

where is the greatest integer which is less than or equal to . To prove the above claim note that it is true for
all 1. Assume that it is true for all integers less than or equal to some integer 1 . Then

1 …

max , ,…,
max , ,…, ,

and this completes our inductive proof. Therefore, in this case every positive solution of equation (4) converges
exponentially to origin.

Case II:
1; In this case the factor map has a positive equilibrium 1. Recall that in a Ricatti equation
the positive equilibrium is globally asymptotically stable (see (Kulenovic et al., 2002 )). Hence, is globally
∞
asymptotically stable equilibrium of . Also, since the sequence , 0, ∞ is bounded (this is
∞
evident by the global asymptotic stability of ), Theorem 1.1(ii) implies that the sequence ,
∞
0, ∞ is bounded. Thereby, we conclude from Theorem 1.1(iv) that the sequence converges to
∞
the invariant fiber since the sequence converges to . Now, consider the fiber . If
, ,… then some simple algebra show that

, ,…, , ,…, ,

Therefore, every member of is a 1 -cycle of ; in other words, it is a periodic orbit of

equation (4) of period 1 . Then, in this case every solution of equation (4) converges to a 1 -cycle.

Case III:
1; In this case origin is the unique equilibrium of equation (4) and . Some calculations show that
′ ′′
0 1, 0 2,

Therefore by Theorem 1.2 origin is semiasymptotic stable equilibrium of form the right, i.e., origin attracts
the sequence for all 0, ∞ . So, by an analysis precisely similar to that of used in the previous case
we conclude, by Theorem 1.1(iv), that the the fiber attracts every solution of equation (4) since origin
attracts every solution of the sequence . Now, assume that , ,…, . Then

… 0,

and therefore, every solution of equation (4) either converges to origin or to a point in one of 1 coordinate
planes in , i.e., the following set

76
 Aust. J. Basic & Appl. Sci., 5(11): 72-77, 2011

, ,…, | 1 1 0 . (7)

elements of the attracting set (note that , where are the coordinate planes) are 1 -cycles of
equation (4). To prove this consider , … , , 0, ,…, . Then, it is easy to verify that

, … , , 0, ,…, , … , , 0, ,…, . (8)

Consequently, every solution of equation (4) either converges to the origin or to a periodic orbit of period
1 , that is in a coordinate plane of . We now summarize the above arguments into the following
theorem that is one of the main results of this paper.

Theorem 4.1:
(Basic convergence Theorem) Assume Equation (4) is given. Then,
(i) For 0 1, every positive orbit of equation (4) converges exponentially to origin. This implies that the
equation has no other periodic orbit except the trivial fixed point (i.e., origin).
(ii) When 1, any cycle of the equation (4) is a 1 -cycle and the set of all 1 -cycles is given by
equation (7). Furthermore
1. For 1, every positive orbit of equation (4) either converges to origin or to a 1 -cycle.
2. When 1, every positive solution of equation (4) converges to a 1 -cycle.

REFERENCES

Braverman, E., 2005. On a discrete model of population dynamics with impolsive harvisting or recruitment,
Nonlinear Analysis, 63: 751-759.
Brauer, F., C. Castillo-Chavez, 2001. Mathematical Models in Population Biology and Epidemiology,
Springer, New York.
Elaydi, S.N., 2000. Discrete Chaos, Chapman & Hall/CRC.
Epstein, J.M., 1985. The Calculus of Conventional War, Brookings inst. Studies in defence policy,
Washington, DC.
Ishiyama, K., Y. Saiki, 2005. Unstable periodic orbits and chaotic economic growth, Chaos, Solitins and
Fractals, 26: 33-42.
Kocic, V.L., G. Ladas, 1993. Global Behavior of Nonlinear Difference Equations of Higher Order with
Applications, Kluwer Accademic Publishers.
Kulenovic, M.R.S., G. Ladas, 2002. Dynamics of Second Order Rational Difference Equations with Open
Problems and Conjectures, Chapman and Hall/CRC.
Matti, L., 1996. Oscillations in some nonlinear economic relationships, Chaos, Solitons and Fractals, 7:
2235- 2245.
Moye, L.A., A.S. Kapadia, 2000. Differenc equations with public health applications, Marcel Dekker, Inc.
Sedaghat, H., 2002. Converges, oscillations and chaos in a discrete model of combat, SIAM. Review., 44:
74-92.
Sedaghat, H., 2003. Nonlinear Difference Equations: Theory with Applications to Social Science Models,
Kluwer, Dordrecht, The Netherlands.
Sedaghat, H., M. Shojaei, 2007. A class of second order difference equations inspired by Euler’s
discretization method, International Journal of Applied Mathematics and Statistics, 9: 110-123.
Shojaei, M., R. Saadati, H. Adibi, 2009. Stability and periodic character of a rational third order difference
equation, Chaos, Solutons and Fractals, 39: 1203-1209.

77