On Some Symmetric Systems of Difference Equations
Abstract
Here we show that the main results in the papers by Yalcinkaya (2008), Yalcinkaya and Cinar (2010), and Yalcinkaya, Cinar, and Simsek (2008), as well as a conjecture from the last mentioned paper, follow from a slight modification of a result by G. Papaschinopoulos and C. J. Schinas. We also give some generalizations of these results.
1. Introduction
Studying difference equations and systems which possess some kind of symmetry attracted some attention recently (see, e.g., [1–25] and the related references therein).
Based on this observation we may, and will, assume that a = 1 in systems of difference equations (1)–(3).
In the main results in [23–25] it is proved that when a = 1, the positive equilibrium point of systems (1)–(3) is globally asymptotically stable.
Here, among others, we show that all the results and conjectures mentioned above follow from a slight modification of a result in the literature published before papers [23–25]. For related systems see also [2, 5–10, 12, 17–20].
2. Main Results
Let ℝ+ = (0, +∞) and be the set of all positive n-dimensional vectors. The following theorem was proved in [4].
Theorem A. Let (M, d) be a complete metric space, where d denotes a metric and M is an open subset of ℝn, and let T : M → M be a continuous mapping with the unique equilibrium x* ∈ M. Suppose that for the discrete dynamic system
It is known that the part-metric p is a continuous metric on , is a complete metric space, and that the distances induced by the part-metric and by the Euclidean norm are equivalent on (see, e.g., [4]).
Based on these properties and Theorem A, the following corollary follows.
Corollary 1. Let be a continuous mapping with a unique equilibrium . Suppose that for the discrete dynamic system (7), there is some k ∈ ℕ such that for the part-metric p inequality
Some applications of various part-metric-related inequalities and some asymptotic methods in studying difference equations related to symmetric ones can be found, for example, in [1, 3–5, 10, 11, 13–16, 22] (see also the related references therein).
In Lemma 2.3 in [10], Papaschinopoulos and Schinas formulated a variant of the following result, without giving a proof. However, the part concerning the equality in inequality (12) below, is not mentioned, but it is crucial in applying Corollary 1 (see inequality (10)). For this reason, the completeness and the benefit of the reader we will give a complete proof of it.
Proposition 2. Let , be continuous functions. We suppose that the system of two difference equations,
Proof. First, we prove that for every n ∈ ℕ0
To prove (13), it is enough to prove that
The proofs of inequalities (14) and (15) are the same (up to the interchanging letters u and v) so it is enough to prove (14).
Now note that if the equality holds in the first inequality in (12), then we have that
Now suppose that the first inequality in (12), is strict. Then, if un > un+r, directly follows that w/un+r > w/un, while from the first inequality in (12) it follows that un+r/w > w/un. Hence
If un < un+r, then un+r/w > un/w, while from the first inequality in (12), it follows that w/un+r > un/w. From these two inequalities, we have that
If (14) and (15) hold then if un ≠ w and vn ≠ w, inequality (13) immediately follows by using the following elementary implication: if a > b and c > d, then min {a, c} > min {b, d}.
If un ≠ w and vn = w, then from the second inequality in (12), we have that vn+r = vn = w. Hence
Finally, note that if un = vn = w, then from (12), we have that un+r = un = w and vn+r = vn = w, so that the first equality in (20) holds and
Now we define the map as follows:
Then we get
By using inequality (13) and the fact that the inequalities 1 ≥ ai > bi, i ∈ I⊆{1, …, m}, I ≠ ∅, along with equalities ai = bi = 1, i ∈ {1, …, m}∖I, imply the inequality min 1≤i≤m ai > min 1≤i≤m bi, we have that for each vector such that ,
It is not difficult to see that the following extension of Proposition 2 can be proved by slight modifications of the proof of Proposition 2.
Proposition 3. Let , i = 1, …, l, be continuous functions. Suppose that the system of difference equations
Now we use Proposition 2 in proving the results in papers [23–25].
Corollary 4. Let k, l ∈ ℕ0, k ≠ l. Consider the system
Proof. We may assume that m = k. From system (28), we have that
Clearly if
By symmetry (see (30)), we have that if equality holds in the second inequality in (31), then yn−k = 1. Therefore, equalities in (31) hold if and only if (xn−k, yn−k) = (1,1). Hence all the conditions of Proposition 2 are fulfilled from which it follows that the positive equilibrium (1,1) is globally asymptotically stable with respect to the set .
Remark 5. Corollary 4 extends and gives a very short proof of the main result in [23], which is obtained for k = 1 and l = 0. Further, it also extends and gives a very short proof of the main result in [25], which is obtained for k = 0 and l = 1. Moreover, it confirms the conjecture in [25], which is obtained for k = 0 and l ∈ ℕ∖{1}.
Corollary 6. Let k, l ∈ ℕ0, k ≠ l. Consider the system
Proof. We may assume that m = k. From system (34), we have that
Remark 7. Corollary 6 extends and gives a very short proof of the main result in [24], which is obtained for k = 1 and l = 0.
Remark 8. Corollary 6 is also a consequence of Corollary 4. Namely, by using the change of variables (xn, yn) = (1/un, 1/vn), system (34) is transformed into the system
Remark 9. Similar type of issues appear in some literature on scalar difference equations (see, e.g., related results in papers [1, 5, 11, 13]).
It is of some interest to extend results in Corollaries 4 and 6 by using Proposition 2. The next result is of this kind and it extends a result in [5].
Corollary 10. Let and with k, l ∈ ℕ, 0 ≤ r1 < ⋯<rk and 0 ≤ m1 < ⋯<ml ≤ rk and satisfy the following two conditions:
- (H1)
,
- (H2)
,
Then is the unique positive equilibrium of the system of difference equations
Proof. Let
From (40) and (41), we see if we show that and have the same sign for n ∈ ℕ, then PnQn will be nonpositive.
We consider four cases.
Case 1. , fn ≥ 1. Clearly in this case . By (H1) and (H2), we have that
Hence and consequently
Case 2. , fn ≤ 1. Since 1/fn ≥ 1, we obtain . On the other hand, by (H1) and (H2), we have
Case 3. Case , fn ≥ 1. Then we have that 1/fn ≤ 1 and consequently . On the other hand, we have
Case 4. Case , fn ≤ 1. Then . On the other hand, we have
Assume that PnQn = 0, then Pn = 0 or Qn = 0. Using (40) or (41) along with (43) in any of these two cases, we have that
Let
Finally, let (x*, y*) be a solution of the system
Then we have that
From all above mentioned and by Proposition 2, we get the result.
Acknowledgments
S. Stević would like to express his sincere thanks to Professors G. Papaschinopoulos and C. J. Schinas for useful conversations and their help during writing this paper. The first author is supported by the Grant P201/10/1032 of the Czech Grant Agency (Prague). The fourth author is supported by the grant FEKT-S-11-2-921 of Faculty of Electrical Engineering and Communication, Brno University of Technology. This paper is also supported by the Serbian Ministry of Science Projects III 41025, III 44006, and OI 171007.