DYNAMICS of

SECOND ORDER
RATIONAL DIFFERENCE
EQUATIONS
With Open Problems and Conjectures

CHAPMAN & HALL/CRC
DYNAMICS of
SECOND ORDER
RATIONAL DIFFERENCE
EQUATIONS
M. R. S. Kulenovi´c
G. Ladas
Boca Raton London New York Washington, D.C.
With Open Problems and Conjectures

This book contains information obtained from authentic and highly regarded sources. Reprinted material
is quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable
efforts have been made to publish reliable data and information, but the author and the publisher cannot
assume responsibility for the validity of all materials or for the consequences of their use.
Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic
or mechanical, including photocopying, microfilming, and recording, or by any information storage or
retrieval system, without prior permission in writing from the publisher.
The consent of CRC Press LLC does not extend to copying for general distribution, for promotion, for
creating new works, or for resale. Specific permission must be obtained in writing from CRC Press LLC
for such copying.
Direct all inquiries to CRC Press LLC, 2000 N.W. Corporate Blvd., Boca Raton, Florida 33431.

Trademark Notice:

Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation, without intent to infringe.

Visit the CRC Press Web site at www.crcpress.com

© 2002 by Chapman & Hall/CRC
No claim to original U.S. Government works
International Standard Book Number 1-58488-275-1
Library of Congress Card Number 20010370
Printed in the United States of America 1 2 3 4 5 6 7 8 9 0
Printed on acid-free paper

Library of Congress Cataloging-in-Publication Data

Kulenovi´ c, M.R.S. (Mustafa R.S.)
Dynamics of second order rational difference equations : with
open problems and conjectures / M.R.S. Kulenovi´ c, G. Ladas.
p. cm.
Includes bibliographical references and index.
ISBN 1-58488-275-1
1. Difference equations—Numerical solutions. I. Ladas, G.E.
II. Title.
QA431 .K84 2001
515



.625—dc21 20010370


disclaimer Page 1 Monday, June 18, 2001 9:53 AM
Contents
Preface ix
Acknowledgements xi
Introduction and Classification of Equation Types 1
1 Preliminary Results 5
1.1 Definitions of Stability and Linearized Stability Analysis . . . . . . . . . 5
1.2 The Stable Manifold Theorem in the Plane . . . . . . . . . . . . . . . . . 7
1.3 Global Asymptotic Stability of the Zero Equilibrium . . . . . . . . . . . 9
1.4 Global Attractivity of the Positive Equilibrium . . . . . . . . . . . . . . . 9
1.5 Limiting Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.6 The Riccati Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.7 Semicycle Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2 Local Stability, Semicycles, Periodicity, and Invariant Intervals 29
2.1 Equilibrium Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2 Stability of the Zero Equilibrium . . . . . . . . . . . . . . . . . . . . . . 30
2.3 Local Stability of the Positive Equilibrium . . . . . . . . . . . . . . . . . 32
2.4 When is Every Solution Periodic with the Same Period? . . . . . . . . . 33
2.5 Existence of Prime Period-Two Solutions . . . . . . . . . . . . . . . . . . 35
2.6 Local Asymptotic Stability of a Two Cycle . . . . . . . . . . . . . . . . . 36
2.7 Convergence to Period-Two Solutions When
C = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.8 Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.9 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . 43
3 (1, 1)-Type Equations 47
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 The Case α = γ = A = B = 0 : x
n+1
=
βx
n
Cx
n−1
. . . . . . . . . . . . . . . 48
3.3 The Case α = β = A = C = 0 : x
n+1
=
γx
n−1
Bx
n
. . . . . . . . . . . . . . . . 49
3.4 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . 49
v
vi Contents
4 (1, 2)-Type Equations 53
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2 The Case β = γ = C = 0 : x
n+1
=
α
A+Bx
n
. . . . . . . . . . . . . . . . . . 54
4.3 The Case β = γ = A = 0 : x
n+1
=
α
Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . 55
4.4 The Case α = γ = B = 0 : x
n+1
=
βx
n
A+Cx
n−1
- Pielou’s Equation . . . . . . 57
4.5 The Case α = γ = A = 0 : x
n+1
=
βx
n
Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . 58
4.6 The Case α = β = C = 0 : x
n+1
=
γx
n−1
A+Bx
n
. . . . . . . . . . . . . . . . . . 59
4.7 The Case α = β = A = 0 : x
n+1
=
γx
n−1
Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . 60
4.7.1 The Case p < 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.7.2 The Case p = 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.7.3 The Case p > 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.8 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . 64
5 (2, 1)-Type Equations 69
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.2 The Case γ = A = B = 0 : x
n+1
=
α+βx
n
Cx
n−1
- Lyness’ Equation . . . . . . . 70
5.3 The Case β = A = C = 0 : x
n+1
=
α+γx
n−1
Bx
n
. . . . . . . . . . . . . . . . . 72
5.4 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . 72
6 (2, 2)-Type Equations 77
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.2 The Case γ = C = 0 : x
n+1
=
α+βx
n
A+Bx
n
- Riccati Equation . . . . . . . . . . 78
6.3 The Case γ = B = 0 : x
n+1
=
α+βx
n
A+Cx
n−1
. . . . . . . . . . . . . . . . . . . 79
6.4 The Case γ = A = 0 : x
n+1
=
α+βx
n
Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . . . 82
6.4.1 Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.4.2 Semicycle Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.4.3 Global Stability and Boundedness . . . . . . . . . . . . . . . . . . 87
6.5 The Case β = C = 0 : x
n+1
=
α+γx
n−1
A+Bx
n
. . . . . . . . . . . . . . . . . . . 89
6.5.1 Semicycle Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.5.2 The Case q < 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.5.3 The Case q > 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.6 The Case β = A = 0 : x
n+1
=
α+γx
n−1
Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . . . 92
6.6.1 Existence and Local Stability of Period-Two Cycles . . . . . . . . 93
6.6.2 Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.6.3 Semicycle Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.6.4 Global Behavior of Solutions . . . . . . . . . . . . . . . . . . . . . 100
6.7 The Case α = C = 0 : x
n+1
=
βx
n
+γx
n−1
A+Bx
n
. . . . . . . . . . . . . . . . . . 101
6.7.1 Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.7.2 Semicycle Analysis of Solutions When q ≤ 1 . . . . . . . . . . . . 103
6.7.3 Semicycle Analysis and Global Attractivity When q = 1 . . . . . . 105
6.7.4 Global Attractivity When q < 1 . . . . . . . . . . . . . . . . . . . 107
Contents vii
6.7.5 Long-term Behavior of Solutions When q > 1 . . . . . . . . . . . 107
6.8 The Case α = B = 0 : x
n+1
=
βx
n
+γx
n−1
A+Cx
n−1
. . . . . . . . . . . . . . . . . . 109
6.8.1 Invariant Intervals and Semicycle Analysis . . . . . . . . . . . . . 110
6.8.2 Global Stability of the Positive Equilibrium . . . . . . . . . . . . 112
6.9 The Case α = A = 0 : x
n+1
=
βx
n
+γx
n−1
Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . . . 113
6.9.1 Existence of a Two Cycle . . . . . . . . . . . . . . . . . . . . . . . 114
6.9.2 Semicycle Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.9.3 Global Stability Analysis When p < q . . . . . . . . . . . . . . . 117
6.9.4 Global Stability Analysis When p > q . . . . . . . . . . . . . . . 123
6.10 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . 123
7 (1, 3)-Type Equations 131
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7.2 The Case β = γ = 0 : x
n+1
=
α
A+Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . . 131
7.3 The Case α = γ = 0 : x
n+1
=
βx
n
A+Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . . 132
7.4 The Case α = β = 0 : x
n+1
=
γx
n−1
A+Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . . 133
7.5 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . 134
8 (3, 1)-Type Equations 137
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
8.2 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . 138
9 (2, 3)-Type Equations 141
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
9.2 The Case γ = 0 : x
n+1
=
α+βx
n
A+Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . . . . 141
9.2.1 Boundedness of Solutions . . . . . . . . . . . . . . . . . . . . . . 142
9.2.2 Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . 143
9.2.3 Semicycle Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 144
9.2.4 Global Asymptotic Stability . . . . . . . . . . . . . . . . . . . . . 147
9.3 The Case β = 0 : x
n+1
=
α+γx
n−1
A+Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . . . . 149
9.3.1 Existence and Local Stability of Period-Two Cycles . . . . . . . . 150
9.3.2 Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . 154
9.3.3 Convergence of Solutions . . . . . . . . . . . . . . . . . . . . . . . 156
9.3.4 Global Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
9.3.5 Semicycle Analysis When pr = q +
q
2
r
. . . . . . . . . . . . . . . . 157
9.4 The Case α = 0 : x
n+1
=
βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . . . . 158
9.4.1 Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . 160
9.4.2 Convergence of Solutions . . . . . . . . . . . . . . . . . . . . . . 162
9.5 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . 163
viii Contents
10 (3, 2)-Type Equations 167
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
10.2 The Case C = 0 : x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
. . . . . . . . . . . . . . . . . . . . 167
10.2.1 Proof of Part (b) of Theorem 10.2.1 . . . . . . . . . . . . . . . . . 169
10.2.2 Proof of Part (c) of Theorem 10.2.1 . . . . . . . . . . . . . . . . . 172
10.3 The Case B = 0 : x
n+1
=
α+βx
n
+γx
n−1
A+Cx
n−1
. . . . . . . . . . . . . . . . . . . . 172
10.3.1 Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . 173
10.3.2 Global Stability When p ≥ r . . . . . . . . . . . . . . . . . . . . . 173
10.3.3 Convergence of Solutions . . . . . . . . . . . . . . . . . . . . . . . 174
10.4 The Case A = 0 : x
n+1
=
α+βx
n
+γx
n−1
Bx
n
+Cx
n−1
. . . . . . . . . . . . . . . . . . . . 175
10.4.1 Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . 177
10.4.2 Convergence of Solutions . . . . . . . . . . . . . . . . . . . . . . . 179
10.5 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . 180
11 The (3, 3)-Type Equation x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
183
11.1 Linearized Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 183
11.2 Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
11.3 Convergence Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
11.3.1 Global Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
11.4 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . 192
A Global Attractivity for Higher Order Equations 201
Bibliography 211
Index 217
Preface
In this monograph, we present the known results and derive several new ones on the
boundedness, the global stability, and the periodicity of solutions of all rational difference
equations of the form
x
n+1
=
α + βx
n
+ γx
n−1
A +Bx
n
+Cx
n−1
, n = 0, 1, . . . (1)
where the parameters α, β, γ, A, B, C and the initial conditions x
−1
and x
0
are nonneg-
ative real numbers.
We believe that the results about Eq(1) are of paramount importance in their own
right and furthermore we believe that these results offer prototypes towards the devel-
opment of the basic theory of the global behavior of solutions of nonlinear difference
equations of order greater than one. The techniques and results which we develop in
this monograph to understand the dynamics of Eq(1) are also extremely useful in ana-
lyzing the equations in the mathematical models of various biological systems and other
applications.
It is an amazing fact that Eq(1) contains, as special cases, a large number of equations
whose dynamics have not been thoroughly understood yet and remain a great challenge
for further investigation. To this end we pose several Open Problems and Conjec-
tures which we believe will stimulate further interest towards a complete understanding
of the dynamics of Eq(1) and its functional generalizations.
Chapter 1 contains some basic definitions and some general results which are used
throughout the book. In this sense, this is a self-contained monograph and the main
prerequisite that the reader needs to understand the material presented here and to be
able to attack the open problems and conjectures is a good foundation in analysis. In an
appendix, at the end of this monograph, we present some global attractivity results for
higher order difference equations which may be useful for extensions and generalizations.
In Chapter 2 we present some general results about periodic solutions and invariant
intervals of Eq(1). In particular we present necessary and sufficient conditions for Eq(1)
to have period-two solutions and we also present a table of invariant intervals.
In Chapters 3 to 11 we present the known results and derive several new ones on the
various types of equations which comprise Eq(1).
Every chapter in this monograph contains several open problems and conjectures
related to the particular equation which is investigated and its generalizations.
ix
Acknowledgements
This monograph is the outgrowth of lecture notes and seminars which were given at the
University of Rhode Island during the last five years. We are thankful to Professors R.
D. Driver, E. A. Grove, J. Hoag, W. Kosmala, L. F. Martins, O. Merino, S. Schultz,
and W. S. Sizer and to our graduate students A. M. Amleh, W. Briden, E. Camouzis,
C. A. Clark, K. Cunningham, R. C. DeVault, H. El-Metwally, J. Feuer, C. Gibbons, E.
Janowski, C. Kent, L. McGrath, M. Predescu, N. Prokup, M. Radin, I. W. Rodrigues,
C. T. Teixeira, S. Valicenti, and K. P. Wilkinson for their enthusiastic participation and
useful suggestions which helped to improve the exposition.
xi
To
Senada and Theodora
Introduction and Classification of
Equation Types
In this monograph, we present the known results and derive several new ones on the
boundedness, the global stability, and the periodicity of solutions of all rational difference
equations of the form
x
n+1
=
α + βx
n
+ γx
n−1
A + Bx
n
+ Cx
n−1
, n = 0, 1, . . . (1)
where the parameters α, β, γ, A, B, C are nonnegative real numbers and the initial con-
ditions x
−1
and x
0
are arbitrary nonnegative real numbers such that
A + Bx
n
+ Cx
n−1
> 0 for all n ≥ 0.
We believe that the results about Eq(1) are of paramount importance in their own
right and furthermore we believe that these results offer prototypes towards the devel-
opment of the basic theory of the global behavior of solutions of nonlinear difference
equations of order greater than one. The techniques and results which we develop in
this monograph to understand the dynamics of Eq(1) are also extremely useful in ana-
lyzing the equations in the mathematical models of various biological systems and other
applications.
It is an amazing fact that Eq(1) contains as special cases, a large number of equa-
tions whose dynamics have not been thoroughly established yet. To this end we pose
several Open Problems and Conjectures which we believe will stimulate further
interest towards a complete understanding of the dynamics of Eq(1) and its functional
generalizations.
Eq(1) includes among others the following well known classes of equations:
1. The Riccati difference equation when
γ = C = 0.
This equation is studied in many textbooks on difference equations, such as [1], [21],
[41], [42], etc. See Section 1.6 where, in addition to the basic description of the solutions
1
2 Introduction and Classification
of the Riccati equation, we introduce the forbidden set of the equation. This is the
set F of all initial conditions x
0
∈ R through which the equation
x
n+1
=
α + βx
n
A + Bx
n
(2)
is not well defined for all n ≥ 0. Hence the solution {x
n
} of Eq(2) exists for all n ≥ 0,
if and only if x
0
/ ∈ F.
The problem of existence of solutions for difference equations is of paramount
importance but so far has been systematically neglected. See [13], [18], [32], [69], and
Section 1.6 for some results in this regard.
Throughout this monograph we pose several open problems related to the existence
of solutions of some simple equations with the hope that their investigation may throw
some light into this very difficult problem.
An example of such a problem is the following:
Open Problem Find all initial points (x
−1
, x
0
) ∈ R×R through which the equa-
tion
x
n+1
= −1 +
x
n−1
x
n
is well defined for all n ≥ 0. See [13].
2. Pielou’s discrete delay logistic model when
α = γ = B = 0.
See ([42], p. 75), [55], [66], [67], and Section 4.4.
We believe that the techniques which we develop in this monograph to understand
the dynamics of Eq(1) will also be of paramount importance in analyzing the equations
in the mathematical models of various biological systems and other applications. For
example, it is interesting to note the similarities of the dynamics of the population model
x
n+1
= α + βx
n−1
e
−x
n
, n = 0, 1, . . .
and the rational equation
x
n+1
=
α + αx
n
+ βx
n−1
1 + x
n
, n = 0, 1, . . . .
See [24] and Sections 2.7 and 10.2.
3. Lyness’ equation when
γ = A = B = 0.
See [42], [44], [49], [57], and Section 5.2. This is a fascinating equation with many
interesting extensions and generalizations. A characteristic feature of this equation is
that it possesses an invariant which can be used to understand the character of its
Introduction and Classification 3
solutions. Without the use of this invariant we are still unable to show, for example,
that every positive solution of the difference equation
x
n+1
=
α + x
n
x
n−1
, n = 0, 1, . . .
where α ∈ (0, ∞), is bounded.
It is an amazing fact that Eq(1) contains, as special cases, a large number of equations
whose dynamics have not been thoroughly established yet.
For convenience we classify all the special cases of Eq(1) as follows.
Let k and l be positive integers from the set {1, 2, 3}. We will say that a special case
of Eq(1) is of (k, l)-type if the equation is of the form of Eq(1) with k positive parameters
in the numerator and l positive parameters in the denominator.
For example the following equations
x
n+1
=
x
n
x
n−1
, n = 0, 1, . . .
x
n+1
=
α + βx
n
γ + x
n−1
, n = 0, 1, . . .
x
n+1
=
1 + x
n
x
n−1
, n = 0, 1, . . .
x
n+1
=
βx
n
+ γx
n−1
A + Bx
n
+ Cx
n−1
, n = 0, 1, . . .
are of types (1, 1), (2, 2), (2, 1), and (2, 3) respectively.
In this sense, Eq(1) contains 49 equations as follows:
9 equations of type (1, 1)
9 equations of type (1, 2)
9 equations of type (2, 1)
9 equations of type (2, 2)
3 equations of type (1, 3)
3 equations of type (3, 1)
3 equations of type (2, 3)
3 equations of type (3, 2)
1 equation of type (3, 3).
Of these 49 equations, 21 are either trivial, or linear, or of the Riccati type. The
remaining 28 are quite interesting nonlinear difference equations. Some of them have
been recently investigated and have led to the development of some general theory about
difference equations. See [7], [15], [28], [29], [42], [43], [45] and [52]-[54]. However to this
day, many special cases of Eq(1) are not thoroughly understood yet and remain a great
challenge for further investigation.
Our goal in this monograph is to familiarize readers with the recent techniques and
results related to Eq(1) and its extensions, and to bring to their attention several Open
Research Problems and Conjectures related to this equation and its functional
generalization.
Chapter 1
Preliminary Results
In this chapter we state some known results and prove some new ones about difference
equations which will be useful in our investigation of Eq(1).
The reader may just glance at the results in this chapter and return for the details
when they are needed in the sequel. See also [1], [4], [16], [21], [22], [33], [34], [41], [63],
and [68].
1.1 Definitions of Stability and Linearized Stability
Analysis
Let I be some interval of real numbers and let
f : I ×I →I
be a continuously differentiable function.
Then for every set of initial conditions x
0
, x
−1
∈ I, the difference equation
x
n+1
= f(x
n
, x
n−1
), n = 0, 1, . . . (1.1)
has a unique solution {x
n
}

n=−1
.
A point x ∈ I is called an equilibrium point of Eq(1.1) if
x = f(x, x);
that is,
x
n
= x for n ≥ 0
is a solution of Eq(1.1), or equivalently, x is a fixed point of f.
Definition 1.1.1 Let x be an equilibrium point of Eq(1.1).
5
6 Chapter 1. Preliminary Results
(i) The equilibrium x of Eq(1.1) is called locally stable if for every ε > 0, there
exists δ > 0 such that for all x
0
, x
−1
∈ I with |x
0
−x| +|x
−1
−x| < δ,
we have
|x
n
−x| < ε for all n ≥ −1.
(ii) The equilibrium x of Eq(1.1) is called locally asymptotically stable if it is
locally stable, and if there exists γ > 0 such that for all x
0
, x
−1
∈ I with |x
0
−x| +
|x
−1
−x| < γ,
we have
lim
n→∞
x
n
= x.
(iii) The equilibrium x of Eq(1.1) is called a global attractor if for every x
0
, x
−1
∈ I
we have
lim
n→∞
x
n
= x.
(iv) The equilibrium x of Eq(1.1) is called globally asymptotically stable if it is
locally stable and a global attractor.
(v) The equilibrium x of Eq(1.1) is called unstable if it is not stable.
(vi) The equilibrium x of Eq(1.1) is called a source, or a repeller, if there exists r > 0
such that for all x
0
, x
−1
∈ I with 0 < |x
0
−x| +|x
−1
−x| < r, there exists N ≥ 1
such that
|x
N
−x| ≥ r.
Clearly a source is an unstable equilibrium.
Let
p =
∂f
∂u
(x, x) and q =
∂f
∂v
(x, x)
denote the partial derivatives of f(u, v) evaluated at the equilibrium x of Eq(1.1).
Then the equation
y
n+1
= py
n
+ qy
n−1
, n = 0, 1, . . . (1.2)
is called the linearized equation associated with Eq(1.1) about the equilibrium point x.
Theorem 1.1.1 (Linearized Stability)
(a) If both roots of the quadratic equation
λ
2
−pλ −q = 0 (1.3)
lie in the open unit disk |λ| < 1, then the equilibrium x of Eq(1.1) is locally
asymptotically stable.
1.2. The Stable Manifold Theorem in the Plane 7
(b) If at least one of the roots of Eq(1.3) has absolute value greater than one, then the
equilibrium x of Eq(1.1) is unstable.
(c) A necessary and sufficient condition for both roots of Eq(1.3) to lie in the open
unit disk |λ| < 1, is
|p| < 1 −q < 2. (1.4)
In this case the locally asymptotically stable equilibrium x is also called a sink.
(d) A necessary and sufficient condition for both roots of Eq(1.3) to have absolute value
greater than one is
|q| > 1 and |p| < |1 −q| .
In this case x is a repeller.
(e) A necessary and sufficient condition for one root of Eq(1.3) to have absolute value
greater than one and for the other to have absolute value less than one is
p
2
+ 4q > 0 and |p| > |1 −q| .
In this case the unstable equilibrium x is called a saddle point.
(f ) A necessary and sufficient condition for a root of Eq(1.3) to have absolute value
equal to one is
|p| = |1 −q|
or
q = −1 and |p| ≤ 2.
In this case the equilibrium x is called a nonhyperbolic point.
Definition 1.1.2 (a) A solution {x
n
} of Eq(1.1) is said to be periodic with period p
if
x
n+p
= x
n
for all n ≥ −1. (1.5)
(b) A solution {x
n
} of Eq(1.1) is said to be periodic with prime period p, or a
p-cycle if it is periodic with period p and p is the least positive integer for which
(1.5) holds.
1.2 The Stable Manifold Theorem in the Plane
Let T be a diffeomorphism in I × I, i.e., one-to-one, smooth mapping, with smooth
inverse.
Assume that p ∈ I ×I is a saddle fixed point of T, i.e., T(p) = p and the Jacobian
J
T
(p) has one eigenvalue s with |s| < 1 and one eigenvalue u with |u| > 1.
8 Chapter 1. Preliminary Results
Let v
s
be an eigenvector corresponding to s and let v
u
be an eigenvector correspond-
ing to u.
Let S be the stable manifold of p, i.e., the set of initial points whose forward orbits
(under iteration by T)
p, T(p), T
2
(p), . . .
converge to p.
Let U be the unstable manifold of p, i.e., the set of initial points whose backward
orbits (under iteration by the inverse of T)
p, T
−1
(p), T
−2
(p), . . .
converge to p. Then, both S and U are one dimensional manifolds (curves) that contain
p. Furthermore, the vectors v
s
and v
u
are tangent to S and U at p, respectively.
The map T for Eq(1.1) is found as follows:
Set
u
n
= x
n−1
and v
n
= x
n
for n ≥ 0.
Then
u
n+1
= x
n
and v
n+1
= f(v
n
, u
n
) for n ≥ 0
and so
T

u
v

=

v
f(v, u)

.
Therefore the eigenvalues of the Jacobian J
T

x
x

at the equilibrium point x of Eq(1.1)
are the roots of Eq(1.3).
In this monograph, the way we will make use of the Stable Manifold Theorem in our
investigation of the rational difference equation
x
n+1
=
α + βx
n
+ γx
n−1
A + Bx
n
+ Cx
n−1
, n = 0, 1, . . . (1.6)
is as follows:
Under certain conditions the positive equilibrium of Eq(1.6) will be a saddle point
and under the same conditions Eq(1.6) will have a two cycle
. . . φ, ψ, φ, ψ, . . . (1.7)
which is locally asymptotically stable. Under these conditions, the two cycle (1.7) cannot
be a global attractor of all non-equilibrium solutions. See, for example, Sections 6.6 and
6.9.
1.3. Global Asymptotic Stability of the Zero Equilibrium 9
1.3 Global Asymptotic Stability of the Zero Equi-
librium
When zero is an equilibrium point of the Eq(1.6), which we are investigating in this
monograph (with β > 0), then the following theorem will be employed to establish
conditions for its global asymptotic stability. Amazingly the same theorem is a powerful
tool for establishing the global asymptotic stability of the zero equilibrium of several
biological models. See, for example, [31] and [48].
Theorem 1.3.1 ([31]. See also [48], and [70]) Consider the difference equation
x
n+1
= f
0
(x
n
, x
n−1
)x
n
+ f
1
(x
n
, x
n−1
)x
n−1
, n = 0, 1, . . . (1.8)
with nonnegative initial conditions and
f
0
, f
1
∈ C[[0, ∞) ×[0, ∞), [0, 1)].
Assume that the following hypotheses hold:
(i) f
0
and f
1
are non-increasing in each of their arguments;
(ii) f
0
(x, x) > 0 for all x ≥ 0;
(iii) f
0
(x, y) + f
1
(x, y) < 1 for all x, y ∈ (0, ∞).
Then the zero equilibrium of Eq(1.8) is globally asymptotically stable.
1.4 Global Attractivity of the Positive Equilibrium
Unfortunately when it comes to establishing the global attractivity of the positive equi-
librium of the Eq(1.6) which we are investigating in this monograph, there are not
enough results in the literature to cover all various cases. We strongly believe that the
investigation of the various special cases of our equation has already played and will
continue to play an important role in the development of the stability theory of general
difference equations of orders greater than one.
The first theorem, which has also been very useful in applications to mathematical
biology, see [31] and [48], was really motivated by a problem in [35].
Theorem 1.4.1 ([31]. See also [35], [48], and ([42], p. 53). Let I ⊆ [0, ∞) be some
interval and assume that f ∈ C[I ×I, (0, ∞)] satisfies the following conditions:
(i) f(x, y) is non-decreasing in each of its arguments;
(ii) Eq(1.1) has a unique positive equilibrium point x ∈ I and the function f(x, x)
satisfies the negative feedback condition:
(x −x)(f(x, x) −x) < 0 for every x ∈ I −{x}.
Then every positive solution of Eq(1.1) with initial conditions in I converges to x.
10 Chapter 1. Preliminary Results
The following theorem was inspired by the results in [55] that deal with the global
asymptotic stability of Pielou’s equation (see Section 4.4), which is a special case of
the equation that we are investigating in this monograph.
Theorem 1.4.2 ([42], p. 27) Assume
(i) f ∈ C[(0, ∞) ×(0, ∞), (0, ∞)];
(ii) f(x, y) is nonincreasing in x and decreasing in y;
(iii) xf(x, x) is increasing in x;
(iv) The equation
x
n+1
= x
n
f(x
n
, x
n−1
), n = 0, 1, . . . (1.9)
has a unique positive equilibrium x.
Then x is globally asymptotically stable.
Theorem 1.4.3 ([27]) Assume
(i) f ∈ C[[0, ∞) ×[0, ∞), (0, ∞)];
(ii) f(x, y) is nonincreasing in each argument;
(iii) xf(x, y) is nondecreasing in x;
(iv) f(x, y) < f(y, x) ⇐⇒x > y;
(v) The equation
x
n+1
= x
n−1
f(x
n−1
, x
n
), n = 0, 1, . . . (1.10)
has a unique positive equilibrium x.
Then x is a global attractor of all positive solutions of Eq(1.10).
The next result is known as the stability trichotomy result:
Theorem 1.4.4 ([46]) Assume
f ∈ C
1
[[0, ∞) ×[0, ∞), [0, ∞)]
is such that
x

∂f
∂x

+ y

∂f
∂y

< f(x, y) for all x, y ∈ (0, ∞). (1.11)
Then Eq(1.1) has stability trichotomy, that is exactly one of the following three cases
holds for all solutions of Eq(1.1):
(i) lim
n→∞
x
n
= ∞ for all (x
−1
, x
0
) = (0, 0) .
(ii) lim
n→∞
x
n
= 0 for all initial points and 0 is the only equilibrium point of Eq(1.1).
(iii) lim
n→∞
x
n
= x ∈ (0, ∞) for all (x
−1
, x
0
) = (0, 0) and x is the only positive equi-
librium of Eq(1.1).
1.4. Global Attractivity of the Positive Equilibrium 11
Very often the best strategy for obtaining global attractivity results for Eq(1.1) is
to work in the regions where the function f(x, y) is monotonic in its arguments. In this
regard there are four possible scenarios depending on whether f(x, y) is nondecreasing
in both arguments, or nonincreasing in both arguments, or nonincreasing in one and
nondecreasing in the other.
The next two theorems are slight modifications of results which were obtained in [54]
and were developed while we were working on a special case of the equation that we are
investigating in this monograph. (See Section 6.9.) The first theorem deals with the case
where the function f(x, y) is non-decreasing in x and non-increasing in y and the second
theorem deals with the case where the function f(x, y) is non-increasing in x and non-
decreasing in y. Theorems 1.4.8 and 1.4.7 below are new and deal with the remaining
two cases relative to the monotonic character of f(x, y). See Sections 6.4 and 6.8 where
these theorems are utilized. See also the Appendix, at the end of this monograph, which
describes the extension of these results to higher order difference equations.
Theorem 1.4.5 ([54]) Let [a, b] be an interval of real numbers and assume that
f : [a, b] ×[a, b] →[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y) is non-decreasing in x ∈ [a, b] for each y ∈ [a, b], and f(x, y) is non-
increasing in y ∈ [a, b] for each x ∈ [a, b];
(b) If (m, M) ∈ [a, b] ×[a, b] is a solution of the system
f(m, M) = m and f(M, m) = M,
then m = M.
Then Eq(1.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(1.1) con-
verges to x.
Proof. Set
m
0
= a and M
0
= b
and for i = 1, 2, . . . set
M
i
= f(M
i−1
, m
i−1
) and m
i
= f(m
i−1
, M
i−1
).
Now observe that for each i ≥ 0,
m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . ≤ M
i
≤ . . . ≤ M
1
≤ M
0
and
m
i
≤ x
k
≤ M
i
for k ≥ 2i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
12 Chapter 1. Preliminary Results
Then
M ≥ limsup
i→∞
x
i
≥ liminf
i→∞
x
i
≥ m (1.12)
and by the continuity of f,
m = f(m, M) and M = f(M, m).
Therefore in view of (b),
m = M
from which the result follows. 2
Theorem 1.4.6 ([54]) Let [a, b] be an interval of real numbers and assume that
f : [a, b] ×[a, b] →[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y) is non-increasing in x ∈ [a, b] for each y ∈ [a, b], and f(x, y) is non-
decreasing in y ∈ [a, b] for each x ∈ [a, b];
(b) The difference equation Eq(1.1) has no solutions of prime period two in [a, b].
Then Eq(1.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(1.1) converges
to x.
Proof. Set
m
0
= a and M
0
= b
and for i = 1, 2, . . . set
M
i
= f(m
i−1
, M
i−1
) and m
i
= f(M
i−1
, m
i−1
).
Now observe that for each i ≥ 0,
m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . ≤ M
i
≤ . . . ≤ M
1
≤ M
0
,
and
m
i
≤ x
k
≤ M
i
for k ≥ 2i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
Then clearly (1.12) holds and by the continuity of f,
m = f(M, m) and M = f(m, M).
In view of (b),
m = M
from which the result follows. 2
1.4. Global Attractivity of the Positive Equilibrium 13
Theorem 1.4.7 Let [a, b] be an interval of real numbers and assume that
f : [a, b] ×[a, b] →[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y) is non-increasing in each of its arguments;
(b) If (m, M) ∈ [a, b] ×[a, b] is a solution of the system
f(m, m) = M and f(M, M) = m,
then m = M.
Then Eq(1.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(1.1) con-
verges to x.
Proof. Set
m
0
= a and M
0
= b
and for i = 1, 2, . . . set
M
i
= f(m
i−1
, m
i−1
) and m
i
= f(M
i−1
, M
i−1
).
Now observe that for each i ≥ 0,
m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . ≤ M
i
≤ . . . ≤ M
1
≤ M
0
,
and
m
i
≤ x
k
≤ M
i
for k ≥ 2i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
Then clearly (1.12) holds and by the continuity of f,
m = f(M, M) and M = f(m, m).
In view of (b),
m = M = x
from which the result follows. 2
Theorem 1.4.8 Let [a, b] be an interval of real numbers and assume that
f : [a, b] ×[a, b] →[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y) is non-decreasing in each of its arguments;
(b) The equation
f(x, x) = x,
has a unique positive solution.
Then Eq(1.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(1.1) con-
verges to x.
14 Chapter 1. Preliminary Results
Proof. Set
m
0
= a and M
0
= b
and for i = 1, 2, . . . set
M
i
= f(M
i−1
, M
i−1
) and m
i
= f(m
i−1
, m
i−1
).
Now observe that for each i ≥ 0,
m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . ≤ M
i
≤ . . . ≤ M
1
≤ M
0
,
and
m
i
≤ x
k
≤ M
i
for k ≥ 2i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
Then clearly the inequality (1.12) is satisfied and by the continuity of f,
m = f(m, m) and M = f(M, M).
In view of (b),
m = M = x,
from which the result follows. 2
1.5 Limiting Solutions
The concept of limiting solutions was introduced by Karakostas [37] (see also [40]) and
is a useful tool in establishing that under certain conditions a solution of a difference
equation which is bounded from above and from below has a limit.
In this section we first give a self-contained version of limiting solutions the way we
will use them in this monograph.
Let J be some interval of real numbers, f ∈ C[J ×J, J], and let k ≥ 1 be a positive
integer. Assume that {x
n
}

n=−1
is a solution of Eq(1.1) such that
A ≤ x
n
≤ B for all n ≥ −1,
where A, B ∈ J. Let L be a limit point of the solution {x
n
}

n=−1
(For example L is the
limit superior S or the limit inferior I of the solution). Now with the above hypotheses,
we will show that there exists a solution {L
n
}

n=−k
of the difference equation
x
n+1
= f(x
n
, x
n−1
), n ≥ −k + 1, . . . (1.13)
1.5. Limiting Solutions 15
such that
L
0
= L
and for every N ≥ −k, the term L
N
is a limit point of the solution {x
n
}

n=−1
.
The solution {L
n
}

n=−k
is called a limiting solution of Eq(1.13) associated with the
solution {x
n
}

n=−1
of Eq(1.1).
Karakostas takes k = ∞and calls the solution {L
n
}

n=−∞
a full limiting sequence
of the difference equation
x
n+1
= f(x
n
, x
n−1
), (1.14)
associated with the solution {x
n
}

n=−1
of Eq(1.1). (See Theorem 1.5.2 below.)
We now proceed to show the existence of a limiting solution {L
n
}

n=−k
.
Clearly there exists a subsequence {x
n
i
}

i=1
of {x
n
}

n=−1
such that
lim
i→∞
x
n
i
= L.
Next the subsequence {x
n
i
−1
}

i=1
of {x
n
}

n=−1
also lies in interval [A, B] and so it has a
further subsequence {x
n
i
j
}

j=1
which converges to some limit which we call L
−1
. There-
fore,
lim
i→∞
x
n
i
j
= L and lim
i→∞
x
n
i
j
−1
= L
−1
.
Continuing in this way and by considering further and further subsequences we obtain
numbers L
−2
, L
−3
, . . . , L
−k
and a subsequence {x
n
j
}

n=−1
such that
lim
j→∞
x
n
j
= L
lim
j→∞
x
n
j
−1
= L
−1
. . .
lim
j→∞
x
n
j
−k
= L
−k
.
Let {
n
}

n=−k
be the solution of Eq(1.13) with
−k
= L
−k
and
−k+1
= L
−k+1
.
Then clearly,

−k+2
= f(
−k+1
,
−k
) = f(L
−k+1
, L
−k
)
= lim
j→∞
f(x
n
j
−k+1
, x
n
j
−k
) = lim
j→∞
x
n
j
−k+2
= L
−k+2
.
It follows by induction that the sequence
L
−k
, L
−k+1
, . . . , L
−1
, L
0
= L =
0
, L
1
=
1
, . . .
satisfies Eq(1.13) and every term of this sequence is a limit point of the solution {x
n
}

n=−1
.
The following result, which is a consequence of the above discussion, is stated in the
way we will use it in this monograph:
16 Chapter 1. Preliminary Results
Theorem 1.5.1 Let J be some interval of real numbers, f ∈ C[J ×J, J], and let k ≥ 1
be a positive integer. Suppose that {x
n
}

n=−1
is a bounded solution of Eq(1.1) with limit
inferior I and limit superior S, with I, S ∈ J. Then Eq(1.13) has two solutions {I
n
}

n=−k
and {S
n
}

n=−k
with the following properties:
I
0
= I, S
0
= S
and
I
n
, S
n
∈ [I, S] for n ≥ −k.
We now state the complete version of full limiting sequences, as developed by Karakostas.
Theorem 1.5.2 ([37]) Let J be some interval of real numbers, f ∈ C[J
ν+1
, J], and let
{x
n
}

n=−ν
be a bounded solution of the difference equation
x
n+1
= f(x
n
, x
n−1
, . . . , x
n−ν
), n = 0, 1, . . . (1.15)
with
I = liminf
n→∞
x
n
, S = limsup
n→∞
x
n
and with I, S ∈ J.
Let Z denote the set of all integers {. . . , −1, 0, 1, . . .}. Then there exist two solutions
{I
n
}

n=−∞
and {S
n
}

n=−∞
of the difference equation
x
n+1
= f(x
n
, x
n−1
, . . . , x
n−ν
) (1.16)
which satisfy the equation for all n ∈ Z, with
I
0
= I, S
0
= S, I
n
, S
n
∈ [I, S] for all n ∈ Z
and such that for every N ∈ Z, I
N
and S
N
are limit points of {x
n
}

n=−ν
.
Furthermore for every m ≤ −ν, there exist two subsequences {x
r
n
} and {x
l
n
} of the
solution {x
n
}

n=−ν
such that the following are true:
lim
n→∞
x
r
n
+N
= I
N
and lim
n→∞
x
l
n
+N
= S
N
for every N ≥ m.
The solutions {I
n
}

n=−∞
and {S
n
}

n=−∞
of the difference equation (1.16) are called
full limiting solutions of Eq(1.16) associated with the solution {x
n
}

n=−k
of Eq(1.15).
See [23] and [25] for an application of Theorem 1.5.2 to difference equations of the form
x
n+1
=
k

i=0
A
i
x
n−i
, n = 0, 1, . . .
which by the change of variables x
n
=
1
y
n
reduce to rational difference equations of the
form
y
n+1
=
1

k
i=0
A
i
y
n−i
, n = 0, 1, . . . .
1.6. The Riccati Equation 17
1.6 The Riccati Equation
A difference equation of the form
x
n+1
=
α + βx
n
A + Bx
n
, n = 0, 1, . . . (1.17)
where the parameters α, β, A, B and the initial condition x
0
are real numbers is called a
Riccati difference equation.
To avoid degenerate cases, we will assume that
B = 0 and αB −βA = 0. (1.18)
Indeed when B = 0, Eq(1.17) is a linear equation and when αB − βA = 0, Eq(1.17)
reduces to
x
n+1
=
βA
B
+ βx
n
A + Bx
n
=
β
B
for all n ≥ 0.
The set of initial conditions x
0
∈ R through which the denominator A+Bx
n
in Eq(1.17)
will become zero for some value of n ≥ 0 is called the forbidden set F of Eq(1.17).
One of our goals in this section is to determine the forbidden set F of Eq(1.17). It
should be mentioned here that there is practically nothing known about the forbidden
set of Eq(1) (when the parameters α, β, γ, A, B, C and the initial conditions x
−1
, x
0
are
all real numbers) other than the material presented here, which is extracted from [32].
(See also [13] and [69].)
The other goal we have is to describe in detail the long- and short-term behavior of
solutions of Eq(1.17) when x
0
/ ∈ F.
The first result in this section addresses a very special case of Eq(1.17) regarding
period-two solutions. The proof is straightforward and will be omitted.
Theorem 1.6.1 Assume that (1.18) holds and that Eq(1.17) possesses a prime period-
two solution. Then
β + A = 0. (1.19)
Furthermore when (1.19) holds every solution of Eq(1.17) with
x
0
=
β
B
(1.20)
is periodic with period two.
In the sequel, in addition to (1.18), we will assume that
β + A = 0. (1.21)
18 Chapter 1. Preliminary Results
Now observe that the change of variable
x
n
=
β + A
B
w
n

A
B
for n ≥ 0 (1.22)
transforms Eq(1.17) into the difference equation
w
n+1
= 1 −
R
w
n
, n = 0, 1, . . . (1.23)
where
R =
βA −αB
(β + A)
2
is a nonzero real number, called the Riccati number of Eq(1.17).
It is interesting to note that the four parameters of Eq(1.17) have been reduced to
the single parameter R of Eq(1.23). If we set
y = g(w) = 1 −
R
w
then we note that the two asymptotes
x = −
A
B
and y =
β
B
of the function
y = f(x) =
α + βx
A + Bx
have now been reduced to the two asymptotes
w = 0 and y = 1
of the function
y = 1 −
R
w
.
Also note that the signum of R is equal to the signum of the derivative
f

(x) =
βA −αB
(A + Bx)
2
.
For the remainder of this section we focus on Eq(1.23) and present its forbidden set F
and the character of its solutions. Similar results can be obtained for Eq(1.17) through
the change of variables (1.22).
From Eq(1.23) it follows that when
R <
1
4
1.6. The Riccati Equation 19
Eq(1.23) has the two equilibrium points w

and w
+
with
w

=
1 −

1 −4R
2
<
1 +

1 −4R
2
= w
+
and
|w

| < w
+
.
Also
w

< 0 if R < 0
and
w

> 0 if R ∈ (0,
1
4
).
When
R =
1
4
,
Eq(1.23) has exactly one equilibrium point namely,
w =
1
2
.
Finally, when
R >
1
4
,
Eq(1.23) has no equilibrium points.
Now observe that the change of variables
w
n
=
y
n+1
y
n
for n = 0, 1, . . .
with
y
0
= 1 and y
1
= w
0
reduces Eq(1.23) to the second order linear difference equation
y
n+2
−y
n+1
+ Ry
n
= 0, n = 0, 1, . . . (1.24)
which can be solved explicitly.
The forbidden set of Eq(1.23) is clearly the set of points where the sequence {y
n
}

n=0
vanishes.
The next two theorems deal with the cases R <
1
4
and R =
1
4
, respectively, where
the characteristic roots of Eq(1.24) are real numbers. Note that in these two cases the
characteristic roots of Eq(1.24) are the equilibrium points of Eq(1.23).
20 Chapter 1. Preliminary Results
Theorem 1.6.2 Assume
R <
1
4
.
Then
w

=
1 −

1 −4R
2
and w
+
=
1 +

1 −4R
2
are the only equilibrium points of Eq(1.23).
The forbidden set F of Eq(1.23) is the sequence of points
f
n
=

w
n−1
+
−w
n−1

w
n
+
−w
n

w
+
w

for n = 1, 2, . . . . (1.25)
When w
0
/ ∈ F, the solution of Eq(1.23) is given by
w
n
=
(w
0
−w

)w
n+1
+
−(w
+
−w
0
)w
n+1

(w
0
−w

)w
n
+
−(w
+
−w
0
)w
n

, n = 1, 2, . . . . (1.26)
It is now clear from Theorem 1.6.2 that
lim
n→∞
f
n
= w

,
and
f
n
< w

if R > 0
while
(f
n
−w

)(f
n+1
−w

) < 0 if R < 0.
The equilibrium w
+
is a global attractor as long as
w
0
/ ∈ F ∪ {w

}
and the equilibrium w

is a repeller.
Theorem 1.6.3 Assume
R =
1
4
.
Then
w =
1
2
is the only equilibrium point of Eq(1.23).
The forbidden set F of Eq(1.23) is the sequence of points
f
n
=
n −1
2n
for n = 1, 2, . . . , (1.27)
1.6. The Riccati Equation 21
which converges to the equilibrium from the left.
When w
0
/ ∈ F, the solution of Eq(1.23) is given by
w
n
=
1 + (2w
0
−1)(n + 1)
2 + 2(2w
0
−1)n
for n = 0, 1, . . . . (1.28)
From Theorem 1.6.3 it follows that the equilibrium w =
1
2
of Eq(1.23) is a global
attractor for all solutions with w
0
/ ∈ F but is locally unstable, more precisely, it is a sink
from the right and a source from the left.
Finally consider the case
R >
1
4
.
The characteristic roots of Eq(1.24) in this case are
λ
±
=
1
2
±i

4R −1
2
with

±
| =

R.
Let
φ ∈ (0,
π
2
)
be such that
cos φ =
1
2

R
and sin φ =

4R −1
2

R
.
Then
y
n
= R
n
2
(C
1
cos(nφ) + C
2
sin(nφ)), n = 0, 1, . . .
with
C
1
= y
0
= 1
and

R(C
1
cos φ + C
2
sin φ) = y
1
= w
0
.
It follows that
C
1
= 1 and C
2
=
2w
0
−1

4R −1
.
Hence
w
n
=

R

4R −1 cos(n + 1)φ + (2w
0
−1) sin(n + 1)φ

4R −1 cos(nφ) + (2w
0
−1) sin(nφ)
, n = 0, 1, . . . (1.29)
22 Chapter 1. Preliminary Results
as long as the denominator is different from zero, that is,
w
0
=
1
2


4R −1
2
cot(nφ) for n = 1, 2, . . . .
The following theorem summarizes the above discussion.
Theorem 1.6.4 Assume that
R >
1
4
and let φ ∈ (0,
π
2
) be such that
cos φ =
1
2

R
and sin φ =

4R −1
2

R
.
Then the forbidden set F of Eq(1.23) is the sequence of points
f
n
=
1
2


4R −1
2
cot(nφ) for n = 1, 2, . . . . (1.30)
When w
0
/ ∈ F, the solution of Eq(1.23) is given by (1.29).
Clearly when R >
1
4
the character of the solution {w
n
} and the forbidden set F
depends on whether or not the number φ ∈ (0,
π
2
) is a rational multiple of π.
Let θ ∈ (−
π
2
,
π
2
) be such that
cos θ =

4R −1
r
and sin θ =
2x
0
−1
r
where
r =

(4R −1)
2
+ (2x
0
−1)
2
.
Then we obtain,
w
n
=

R
cos(nφ + φ −θ)
cos(nφ −θ)
=

R
cos(nφ −θ) cos φ −sin(nφ −θ) sin φ
cos(nφ −θ)
=

R(cos φ −sin φtan(nφ −θ))
and so
w
n
=
1
2


4R −1
2
tan(nφ −θ), n = 0, 1, . . . (1.31)
from which the character of the solutions of Eq(1.23), when R >
1
4
, can now be easily
revealed.
For example, assume φ is a rational multiple of π, that is,
φ =
k
N
π ∈ (0,
π
2
) (1.32)
1.6. The Riccati Equation 23
for some
k, N ∈ {1, 2, . . .} with 2k < N.
Then
w
N
=
1
2


4R −1
2
tan(kπ −θ) =
1
2


4R −1
2
tan(−θ) = w
0
and so every solution of Eq(1.23) is periodic with period N.
Equivalently assume that
A =
1
4 cos
2
φ
with φ =
k
N
π
where
k, N ∈ {1, 2, . . .} and 2k < N,
then every solution of Eq(1.23) is periodic with period N.
The following theorem summarizes the above discussion.
Theorem 1.6.5 Assume that
φ =
k
p
π ∈ (0,
π
2
)
where k and p are positive comprimes and suppose that
cos φ =
1
2

R
and sin φ =

4R −1
2

R
,
or equivalently
4Rcos
2
(
k
p
π) = 1.
Then every solution of Eq(1.23) with
w
0
=
sin(n
k
p
π) −

4R −1 cos(n
k
p
π)
2 sin(n
k
p
π)
for n = 1, 2, . . . , p −1
is periodic with period p.
When the number φ in Theorem 1.6.5 is not a rational multiple of π, then the
following result is true:
Theorem 1.6.6 Assume that the number φ in Theorem 1.6.5 is not a rational multiple
of π. Then the following statements are true:
(i) No solution of Eq(1.23) is periodic.
(ii) The set of limit points of a solution of Eq(1.23) with w
0
/ ∈ F is dense in R.
24 Chapter 1. Preliminary Results
Proof. Statement (i) is true because,
w
k
= w
l
if and only if
tan(kφ −θ) = tan(lφ −θ)
if and only if for some integer N
kφ −θ −(lφ −θ) = Nπ
if and only if
(k −l)φ = Nπ
that is,
φ is a rational multiple of π.
The proof of statement (ii) is a consequence of the form of the solution of Eq(1.23) and
the fact that when φ is not a rational multiple of π, then the values
(nφ −θ)(mod π) for n = 0, 1, . . .
are dense in the interval (−
π
2
,
π
2
). 2
1.7 Semicycle Analysis
We strongly believe that a semicycle analysis of the solutions of a scalar difference equa-
tion is a powerful tool for a detailed understanding of the entire character of solutions
and often leads to straightforward proofs of their long term behavior.
In this section, we present some results about the semicycle character of solutions of
some general difference equations of the form
x
n+1
= f(x
n
, x
n−1
), n = 0, 1, . . . (1.33)
under appropriate hypotheses on the function f.
First we give the definitions for the positive and negative semicycle of a solution of
Eq(1.33) relative to an equilibrium point x.
A positive semicycle of a solution {x
n
} of Eq(1.1) consists of a “string” of terms
{x
l
, x
l+1
, . . . , x
m
}, all greater than or equal to the equilibrium x, with l ≥ −1 and m ≤ ∞
and such that
either l = −1, or l > −1 and x
l−1
< x
and
either m = ∞, or m < ∞ and x
m+1
< x.
1.7. Semicycle Analysis 25
A negative semicycle of a solution {x
n
} of Eq(1.33) consists of a “string” of terms
{x
l
, x
l+1
, . . . , x
m
}, all less than the equilibrium x, with l ≥ −1 and m ≤ ∞ and such
that
either l = −1, or l > −1 and x
l−1
≥ x
and
either m = ∞, or m < ∞ and x
m+1
≥ x.
Definition 1.7.1 (Oscillation)
(a) A sequence {x
n
} is said to oscillate about zero or simply to oscillate if the terms
x
n
are neither eventually all positive nor eventually all negative. Otherwise the
sequence is called nonoscillatory. A sequence {x
n
} is called strictly oscillatory
if for every n
0
≥ 0, there exist n
1
, n
2
≥ n
0
such that x
n
1
x
n
2
< 0.
(b) A sequence {x
n
} is said to oscillate about x if the sequence x
n
− x oscillates.
The sequence {x
n
} is called strictly oscillatory about x if the sequence x
n
−x
is strictly oscillatory.
The first result was established in [28] while we were working on a special case of the
equation we are investigating in this monograph. (See Section 6.5.)
Theorem 1.7.1 ([28]) Assume that f ∈ C[(0, ∞) ×(0, ∞), (0, ∞)] is such that:
f(x, y) is decreasing in x for each fixed y, and f(x, y) is increasing in y for each fixed
x.
Let x be a positive equilibrium of Eq(1.33). Then except possibly for the first semi-
cycle, every solution of Eq(1.33) has semicycles of length one.
Proof. Let {x
n
} be a solution of Eq(1.33) with at least two semicycles. Then there
exists N ≥ 0 such that either
x
N−1
< x ≤ x
N
or
x
N−1
≥ x > x
N
.
We will assume that
x
N−1
< x ≤ x
N
.
All other cases are similar and will be omitted. Then by using the monotonic character
of f(x, y) we have
x
N+1
= f(x
N
, x
N−1
) < f(x, x) = x
and
x
N+2
= f(x
N+1
, x
N
) > f(x, x) = x.
Thus
x
N+1
< x < x
N+2
26 Chapter 1. Preliminary Results
and the proof follows by induction. 2
The next result applies when the function f is decreasing in both arguments.
Theorem 1.7.2 Assume that f ∈ C[(0, ∞)×(0, ∞), (0, ∞)] and that f(x, y) is decreas-
ing in both arguments.
Let x be a positive equilibrium of Eq(1.33). Then every oscillatory solution of
Eq(1.33) has semicycles of length at most two.
Proof. Assume that {x
n
} is an oscillatory solution with two consecutive terms x
N−1
and x
N
in a positive semicycle
x
N−1
≥ x and x
N
≥ x
with at least one of the inequalities being strict. The proof in the case of negative
semicycle is similar and is omitted. Then by using the decreasing character of f we
obtain:
x
N+1
= f(x
N
, x
N−1
) < f(x, x) = x
which completes the proof. 2
The next result applies when the function f is increasing in both arguments.
Theorem 1.7.3 Assume that f ∈ C[(0, ∞)×(0, ∞), (0, ∞)] and that f(x, y) is increas-
ing in both arguments.
Let x be a positive equilibrium of Eq(1.33). Then except possibly for the first semi-
cycle, every oscillatory solution of Eq(1.33) has semicycles of length one.
Proof. Assume that {x
n
} is an oscillatory solution with two consecutive terms x
N−1
and x
N
in a positive semicycle
x
N−1
≥ x and x
N
≥ x,
with at least one of the inequalities being strict. The proof in the case of negative
semicycle is similar and is omitted. Then by using the increasing character of f we
obtain:
x
N+1
= f(x
N
, x
N−1
) > f(x, x) = x
which shows that the next term x
N+1
also belongs to the positive semicycle. It follows by
induction that all future terms of this solution belong to this positive semicycle, which
is a contradiction. 2
The next result applies when the function f(x, y) is increasing in x and decreasing
in y.
1.7. Semicycle Analysis 27
Theorem 1.7.4 Assume that f ∈ C[(0, ∞) ×(0, ∞), (0, ∞)] is such that:
f(x, y) is increasing in x for each fixed y, and f(x, y) is decreasing in y for each fixed
x.
Let x be a positive equilibrium of Eq(1.33).
Then, except possibly for the first semicycle, every oscillatory solution of Eq(1.33)
has semicycles of length at least two.
Furthermore, if we assume that
f(u, u) = x for every u (1.34)
and
f(x, y) < x for every x > y > x (1.35)
then {x
n
} oscillates about the equilibrium x with semicycles of length two or three, except
possibly for the first semicycle which may have length one. The extreme in each semicycle
occurs in the first term if the semicycle has two terms, and in the second term if the
semicycle has three terms.
Proof. Assume that {x
n
} is an oscillatory solution with two consecutive terms x
N−1
and x
N
such that
x
N−1
< x ≤ x
N
.
Then by using the increasing character of f we obtain
x
N+1
= f(x
N
, x
N−1
) > f(x, x) = x
which shows that the next term x
N+1
also belongs to the positive semicycle. The proof
in the case
x
N−1
≥ x > x
N
,
is similar and is omitted.
Now also assume that {x
n
} is an oscillatory solution with two consecutive terms
x
N−1
and x
N
such that
x
N−1
> x
N
≥ x and x
N−2
< x.
Then by using the increasing character of f and Condition (1.34) we obtain
x
N+1
= f(x
N
, x
N−1
) < f(x
N
, x
N
) = x
which shows that the positive semicycle has length two. If
x
N
> x
N−1
> x
then by using the increasing character of f and Condition (1.34) we obtain:
x
N+1
= f(x
N
, x
N−1
) > f(x
N
, x
N
) = x
28 Chapter 1. Preliminary Results
and by using Condition (1.35) we find
x
N+1
= f(x
N
, x
N−1
) < x
N
and the proof is complete.
2
Condition (1.34) is not enough to guarantee the above result. If, instead of Condition
(1.35), we assume
f(x, y) > x for every x > y > x. (1.36)
then Eq(1.33) has monotonic solutions.
Theorem 1.7.5 Assume that f ∈ C[(0, ∞) ×(0, ∞), (0, ∞)] is such that:
f(x, y) is increasing in x for each fixed y, and f(x, y) is decreasing in y for each fixed
x.
Let x be a positive equilibrium of Eq(1.33). Assume that f satisfies Conditions (1.34)
and (1.36).
Then if x
0
> x
−1
> x, the corresponding solution is increasing.
If, on the other hand, f satisfies Condition (1.34), and
f(x, y) < x for every x < y < x (1.37)
then if x
0
< x
−1
< x, the corresponding solution is decreasing.
Proof. First, assume that Condition (1.36) is satisfied and that x
0
> x
−1
> x. Then
by the monotonic character of f and (1.34), we obtain
x
1
= f(x
0
, x
−1
) > f(x
−1
, x
−1
) = x.
In view of Condition (1.36) we get
x
1
= f(x
0
, x
−1
) > x
0
.
and the proof follows by induction.
The proof of the remaining case is similar and will be omitted. 2
Chapter 2
Local Stability, Semicycles,
Periodicity, and Invariant Intervals
2.1 Equilibrium Points
Here we investigate the equilibrium points of the general equation
x
n+1
=
α +βx
n
+γx
n−1
A +Bx
n
+Cx
n−1
, n = 0, 1, . . . (2.1)
where
α, β, γ, A, B, C ∈ [0, ∞) with α +β + γ, B +C ∈ (0, ∞) (2.2)
and where the initial conditions x
−1
and x
0
are arbitrary nonnegative real numbers such
that the right hand side of Eq(2.1) is well defined for all n ≥ 0.
In view of the above restriction on the initial conditions of Eq(2.1), the equilibrium
points of Eq(2.1) are the nonnegative solutions of the equation
x =
α + (β +γ)x
A + (B +C)x
(2.3)
or equivalently
(B +C)x
2
−(β + γ −A)x −α = 0. (2.4)
Zero is an equilibrium point of Eq(2.1) if and only if
α = 0 and A > 0. (2.5)
When (2.5) holds, in addition to the zero equilibrium , Eq(2.1) has a positive equilibrium
if and only if
β +γ > A.
29
30 Chapter 2. Local Stability, Semicycles, Periodicity, and Invariant Intervals
In fact in this case the positive equilibrium of Eq(2.1) is unique and is given by
x =
β +γ −A
B +C
. (2.6)
When
α = 0 and A = 0
the only equilibrium point of Eq(2.1) is positive and is given by
x =
β +γ
B +C
. (2.7)
Note that in view of Condition (2.2), when α = 0, the quantity β +γ is positive.
Finally when
α > 0
the only equilibrium point of Eq(2.1) is the positive solution
x =
β +γ −A +

(β + γ −A)
2
+ 4α(B + C)
2(B +C)
(2.8)
of the quadratic equation (2.4).
If we had allowed negative initial conditions then the negative solution of Eq(2.4)
would also be an equilibrium worth investigating. The problem of investigating Eq(2.1)
when the initial conditions and the coefficients of the equation are real numbers is of
great mathematical importance and, except for our presentation in Section 1.6 about
the Riccati equation
x
n+1
=
α +βx
n
A +Bx
n
, n = 0, 1, . . . ,
there is very little known. (See [13], [18], [32], and [69] for some results in this regard.)
In summary, it is interesting to observe that when Eq(2.1) has a positive equilibrium
x, then x is unique, it satisfies Eqs(2.3) and (2.4), and it is given by (2.8). This ob-
servation simplifies the investigation of the local stability of the positive equilibrium of
Eq(2.1).
2.2 Stability of the Zero Equilibrium
Here we investigate the stability of the zero equilibrium of Eq(2.1).
Set
f(u, v) =
α +βu + γv
A +Bu + Cv
and observe that
f
u
(u, v) =
(βA −αB) + (βC −γB)v
(A +Bu +Cv)
2
(2.9)
2.2. Stability of the Zero Equilibrium 31
and
f
v
(u, v) =
(γA −αC) −(βC −γB)u
(A +Bu +Cv)
2
. (2.10)
If x denotes an equilibrium point of Eq(2.1), then the linearized equation associated
with Eq(2.1) about the equilibrium point x is
z
n+1
−pz
n
−qz
n−1
= 0
where
p = f
u
(x, x) and q = f
v
(x, x).
The local stability character of x is now described by the linearized stability Theorem
1.1.1.
For the zero equilibrium of Eq(2.1) we have
p =
β
A
and q =
γ
A
and so the linearized equation associated with Eq(2.1) about the zero equilibrium point
is
z
n+1

β
A
z
n

γ
A
z
n−1
= 0, n = 0, 1, . . . .
For the stability of the zero equilibrium of Eq(2.1), in addition to the local stability The-
orem 1.1.1, we have the luxury of the global stability Theorem 1.3.1. As a consequence
of these two theorems we obtain the following global result:
Theorem 2.2.1 Assume that A > 0. Then the zero equilibrium point of the equation
x
n+1
=
βx
n
+γx
n−1
A +Bx
n
+Cx
n−1
, n = 0, 1, . . .
is globally asymptotically stable when
β + γ ≤ A
and is unstable when
β +γ > A.
Furthermore the zero equilibrium point is
a sink when A > β +γ
a saddle point when A < β +γ < A + 2β
a repeller when β +γ > A + 2β.
32 Chapter 2. Local Stability, Semicycles, Periodicity, and Invariant Intervals
2.3 Local Stability of the Positive Equilibrium
As we mentioned in Section 2.1, Eq(2.1) has a positive equilibrium when either
α = 0 and β +γ > A
or
α > 0.
In these cases the positive equilibrium x is unique, it satisfies Eqs(2.3) and (2.4), and it
is given by (2.8). By using the identity
(B +C)x
2
= (β + γ −A)x +α,
we see that
p = f
u
(x, x) =
(βA −αB) + (βC −γB)x
(A + (B +C)x)
2
=
(βA −αB) + (βC −γB)x
A
2
+ α(B +C) + (B +C)(A +β + γ)x
and
q = f
v
(x, x) =
(γA −αC) −(βC −γB)x
(A + (B +C)x)
2
=
(γA −αC) −(βC −γB)x
A
2
+α(B +C) + (B +C)(A +β +γ)x
.
Hence the conditions for the local asymptotic stability of x are (see Theorem 1.1.1) the
following:
|p| < 1 −q (2.11)
and
q > −1. (2.12)
Inequalities (2.11) and (2.12) are equivalent to the following three inequalities:
¸
A +β + γ + 2
βC −γB
B + C
¸
x > −α −
A
2
+ (βA −αB) −(γA −αC)
B +C
(2.13)
(A +β +γ)x > −α −
A
2
−(βA −αB) −(γA −αC)
B + C
(2.14)
¸
A +β + γ −
βC −γB
B +C
¸
x > −α −
A
2
+ (γA −αC)
B +C
. (2.15)
Out of these inequalities we will obtain explicit stability conditions by using the following
procedure:
First we observe that Inequalities (2.13)-(2.15), are equivalent to some inequalities
of the form
x > ρ
2.4. When is Every Solution Periodic with the Same Period? 33
and/or
x < σ
for some ρ, σ ∈ [0, ∞).
Now if we set
F(u) = (B +C)u
2
−(β +γ −A)u −α,
it is clear that
F(x) = 0
and that
x > ρ if and only if F(ρ) < 0
while
x < σ if and only if F(σ) > 0.
2.4 When is Every Solution Periodic with the Same
Period?
The following four special examples of Eq(2.1)
x
n+1
=
1
x
n
, n = 0, 1, . . . (2.16)
x
n+1
=
1
x
n−1
, n = 0, 1, . . . (2.17)
x
n+1
=
1 +x
n
x
n−1
, n = 0, 1, . . . (2.18)
x
n+1
=
x
n
x
n−1
, n = 0, 1, . . . (2.19)
are remarkable in the sense that all positive solutions of each of these four nontrivial
equations are periodic with periods 2, 4, 5, and 6 respectively.
The following result characterizes all possible special cases of equations of the form
of Eq(2.1) with the property that every solution of the equation is periodic with the
same period. (See also [75], [73], and [74].)
Theorem 2.4.1 Let p ≥ 2 be a positive integer and assume that every positive solution
of Eq(2.1) is periodic with period p. Then the following statements are true:
(i) Assume C > 0. Then A = B = γ = 0.
(ii) Assume C = 0. Then γ(α +β) = 0.
34 Chapter 2. Local Stability, Semicycles, Periodicity, and Invariant Intervals
Proof. Consider the solution with
x
−1
= 1 and x
0
∈ (0, ∞).
Then clearly
x
0
= x
p
and x
p−1
= x
−1
= 1
and so from Eq(2.1)
x
p
=
α +β + γx
p−2
A + B +Cx
p−2
= x
0
.
Hence
(A +B)x
0
+ (Cx
0
−γ)x
p−2
= α +β. (2.20)
(i) Assume that C > 0. Then we claim that
A = B = 0. (2.21)
Otherwise A +B > 0 and by choosing
x
0
> max

α +β
A +B
,
γ
C
¸
we see that (2.20) is impossible. Hence (2.21) holds. In addition to (2.21) we now
claim that also
γ = 0. (2.22)
Otherwise γ > 0 and by choosing
x
0
< min

α +β
A +B
,
γ
C
¸
we see that (2.20) is impossible. Thus (2.22) holds .
(ii) Assume C = 0 and, for the sake of contradiction, assume that
γ(α +β) > 0.
Then by choosing x
0
sufficiently small, we see that (2.20) is impossible.
The proof is complete. 2
Corollary 2.4.1 Let p ∈ {2, 3, 4, 5, 6}. Assume that every positive solution of Eq(2.1)
is periodic with period p. Then up to a change of variables of the form
x
n
= λy
n
Eq(2.1) reduces to one of the equations (2.16)-(2.19) .
2.5. Existence of Prime Period-Two Solutions 35
2.5 Existence of Prime Period-Two Solutions
In this section we give necessary and sufficient conditions for Eq(2.1) to have a prime
period-two solution and we exhibit all prime period-two solutions of the equation. (See
[50].)
Furthermore we are interested in conditions under which every solution of Eq(2.1)
converges to a period-two solution in a nontrivial way according to the following con-
vention.
Convention Throughout this book when we say that “every solution of a certain
difference equation converges to a periodic solution with period p” we mean that every
solution converges to a, not necessarily prime, solution with period p and furthermore
the set of solutions of the equation with prime period p is nonempty.
Assume that
. . . , φ, ψ, φ, ψ, . . .
is a prime period-two solution of Eq(2.1). Then
φ =
α +βψ +γφ
A +Bψ + Cφ
and
ψ =
α + βφ +γψ
A +Bφ +Cψ
from which we find that
C(φ +ψ) = γ −β −A (2.23)
and when
C > 0 (2.24)
we also find that
(B −C)φψ =
αC +β(γ −β −A)
C
. (2.25)
One can show that when
C = 0 and B > 0 (2.26)
Eq(2.1) has a prime period-two solution if and only if
γ = β + A. (2.27)
Furthermore when (2.26) and (2.27) hold
. . . , φ, ψ, φ, ψ, . . .
is a prime period-two solution of Eq(2.1) if and only if
α + β(φ +ψ) = Bφψ with φ, ψ ∈ [0, ∞) and φ = ψ
36 Chapter 2. Local Stability, Semicycles, Periodicity, and Invariant Intervals
or equivalently
φ >
β
B
, φ =
β +

β
2
+ αB
B
, and ψ =
α + βφ
−β +Bφ
. (2.28)
Next assume that (2.24) holds and that Eq(2.1) possesses a prime period-two solution
. . . , φ, ψ, φ, ψ, . . . .
Then it follows from (2.23) and (2.25) that one of the following three conditions must
be satisfied:
B = C, α = β = 0, and φψ = 0; (2.29)
γ > β + A and B = C, α = β = 0, and φψ ≥ 0; (2.30)
B > C, α + β > 0, and φψ > 0. (2.31)
When (2.29) holds, the two cycle is
. . . , 0,
γ −A
C
, 0,
γ −A
C
, . . . .
When (2.30) holds, the two cycle is
. . . , φ,
γ −A
C
−φ, φ,
γ −A
C
−φ, . . . with 0 ≤ φ <
γ −A
C
.
Finally when (2.31) holds, the values φ and ψ of the two cycle are given by the two roots
of the quadratic equation
t
2

γ −β −A
C
t +
αC +β(γ −β −A)
C(B −C)
= 0
and we should also require that the discriminant of this quadratic equation be positive.
That is,
α <
(γ −β −A)[B(γ −β −A) −C(γ + 3β −A)]
4C
2
. (2.32)
2.6 Local Asymptotic Stability of a Two Cycle
Let
. . . , φ, ψ, φ, ψ, . . .
be a two cycle of the difference equation (2.1).
Set
u
n
= x
n−1
and v
n
= x
n
for n = 0, 1, . . .
2.6. Local Asymptotic Stability of a Two Cycle 37
and write Eq(2.1) in the equivalent form:
u
n+1
= v
n
v
n+1
=
α+βv
n
+γu
n
A+Bv
n
+Cu
n
, n = 0, 1, . . . .
Let T be the function on [0, ∞) ×[0, ∞) defined by:
T

u
v

=

v
βv+γu+α
Bv+Cu+A

.
Then

φ
ψ

is a fixed point of T
2
, the second iterate of T. Furthermore
T
2

u
v

=

g(u, v)
h(u, v)

where
g(u, v) =
α +βv + γu
A + Bv +Cu
and h(u, v) =
α +β
α+βv+γu
A+Bv+Cu
+γv
A +B
α+βv+γu
A+Bv+Cu
+ Cv
.
The two cycle is locally asymptotically stable if the eigenvalues of the Jacobian matrix
J
T
2, evaluated at

φ
ψ

lie inside the unit disk.
By definition
J
T
2

φ
ψ

=

¸
¸
∂g
∂u
(φ, ψ)
∂g
∂v
(φ, ψ)
∂h
∂u
(φ, ψ)
∂h
∂v
(φ, ψ)
¸

.
By computing the partial derivatives of g and h and by using the fact that
φ =
α +βψ + γφ
A +Bψ +Cφ
and ψ =
α +βφ + γψ
A + Bφ +Cψ
we find the following identities:
∂g
∂u
(φ, ψ) =
(γA −αC) + (γB −βC)ψ
(A +Bψ +Cφ)
2
,
∂g
∂v
(φ, ψ) =
(βA −αB) −(γB −βC)φ
(A +Bψ +Cφ)
2
,
∂h
∂u
(φ, ψ) =
[(βA −αB) −(γB −βC)ψ][(γA −αC) + (γB −βC)ψ]
(A +Bφ +Cψ)
2
(A + Bψ +Cφ)
2
,
38 Chapter 2. Local Stability, Semicycles, Periodicity, and Invariant Intervals
∂h
∂v
(φ, ψ) =
[(βA −αB) −(γB −βC)ψ][(βA −αB) −(γB −βC)φ]
(A + Bφ +Cψ)
2
(A +Bψ + Cφ)
2
+
(γA −αC) + (γB −βC)φ
(A +Bφ + Cψ)
2
.
Set
S =
∂g
∂u
(φ, ψ) +
∂h
∂v
(φ, ψ)
and
D =
∂g
∂u
(φ, ψ)
∂h
∂v
(φ, ψ) −
∂g
∂v
(φ, ψ)
∂h
∂u
(φ, ψ).
Then it follows from Theorem 1.1.1 (c) that both eigenvalues of J
T
2

φ
ψ

lie inside the
unit disk |λ| < 1, if and only if
|S| < 1 +D and D < 1.
See Sections 6.6, 6.9, and 7.4 for some applications.
Definition 2.6.1 (Basin of Attraction of a Two Cycle)
Let
. . . , φ, ψ, φ, ψ, . . .
be a two cycle of Eq(2.1). The basin of attraction of this two cycle is the set B of all
initial conditions (x
−1
, x
0
) through which the solution of Eq(2.1) converges to the two
cycle.
2.7 Convergence to Period-Two Solutions When
C = 0
In this section we consider the equation
x
n+1
=
α +βx
n
+γx
n−1
A +Bx
n
, n = 0, 1, . . . (2.33)
where
α, β, γ, A, B ∈ [0, ∞) with α + β +γ, A +B ∈ (0, ∞)
and nonnegative initial conditions and obtain a signum invariant for its solutions which
will be used in the sequel (See Sections 4.7, 6.5, 6.7, and 10.2) to obtain conditions on
α, β, γ, A and B so that every solution of Eq(2.33) converges to a period-two solution.
That is, every solution converges to a (not necessarily prime) period-two solution and
2.7. Convergence to Period-Two Solutions When C = 0 39
there exist prime period-two solutions. (See also Section 10.2 where we present the
complete character of solutions of Eq(2.33).)
Now before we see that when B > 0 every solution of Eq(2.33) converges to a period-
two solution if and only if (2.27) holds, we need the following result, the proof of which
follows by straightforward computations.
Lemma 2.7.1 Assume that (2.27) holds, and let {x
n
}

n=−1
be a solution of Eq(2.33).
Set
J
n
= α +β(x
n−1
+x
n
) −Bx
n−1
x
n
for n ≥ 0. (2.34)
Then the following statements are true:
(a)
J
n+1
=
β +A
A +Bx
n
J
n
for n ≥ 0.
In particular, the sign of the quantity J
n
is constant along each solution.
(b)
x
n+1
−x
n−1
=
J
n
A +Bx
n
for n ≥ 0.
In particular, one and only one of the following three statements is true for each solution
of Eq(2.33):
(i)
x
n+1
−x
n−1
= 0 for all n ≥ 0;
(ii)
x
n+1
−x
n−1
< 0 for all n ≥ 0;
(iii)
x
n+1
−x
n−1
> 0 for all n ≥ 0.
Clearly (i) holds, if and only if, (β + A)J
0
= 0. In this case, the solution {x
n
} is
periodic with period-two.
Statement (ii) holds if and only if J
0
< 0 and β+A > 0. In this case the subsequences
{x
2n
}

n=0
and {x
2n−1
}

n=0
of the solution {x
n
}

n=0
are both decreasing to finite limits and
so in this case the solution converges to a period-two solution.
Finally (iii) holds, if and only if J
0
> 0 and β +A > 0. In this case the subsequences
{x
2n
}

n=0
and {x
2n−1
}

n=0
of the solution {x
n
}

n=0
are both increasing and so if we can
show that they are bounded, the solution would converge to a period-two solution.
Now observe that
x
n+1
−x
n−1
=
β +A
A +Bx
n
(x
n
−x
n−2
) for n ≥ 1
40 Chapter 2. Local Stability, Semicycles, Periodicity, and Invariant Intervals
and so
x
n+1
−x
n−1
= (x
1
−x
−1
)
n
¸
k=1
β +A
A + Bx
k
. (2.35)
We will now employ (2.35) to show that the subsequences of even and odd terms of
the solution are bounded. We will give the details for the subsequence of even terms.
The proof for the subsequence of odd terms is similar and will be omitted. To this end
assume, for the sake of contradiction, that the subsequence of even terms increases to
∞. Then there exists N ≥ 0 such that
ρ =
A +β
A + Bx
2N+1
A +β
A +Bx
2N
< 1.
Clearly for all n > N,
A +β
A +Bx
2n+1
A + β
A +Bx
2n
<
A + β
A +Bx
2N+1
A +β
A +Bx
2N
= ρ.
Set
K = |x
2N
−x
2N−2
|.
Then
|x
2N+2
−x
2N
| =
A + β
A +Bx
2N+1
A +β
A +Bx
2N
|x
2N
−x
2N−2
| < Kρ
|x
2N+4
−x
2N+2
| =
A + β
A + Bx
2N+3
A + β
A + Bx
2N+2
|x
2N+2
−x
2N
| < Kρ
2
and by induction
|x
2N+2m
−x
2N+2(m−1)
| < Kρ
m
for m = 1, 2, . . . .
But then
x
2N+2m
≤ |x
2N+2m
−x
2N+2m−2
| +. . . +|x
2N+2
−x
2N
| +x
2N
≤ K
m
¸
i=1
ρ
i
+x
2N
<

1 −ρ
+x
2N
which contradicts the hypothesis that the subsequence of even terms is unbounded.
In summary we have established the following result.
Theorem 2.7.1 Assume B > 0. Then the following statements hold:
(a) Eq(2.33) has a prime period-two solution if and only if (2.27) holds.
(b) When (2.27) holds all prime period-two solutions
. . . , φ, ψ, φ, ψ, . . .
of Eq(2.33) are given by (2.28).
(c) When (2.27) holds every solution of Eq(2.33) converges to a period-two solution.
2.8. Invariant Intervals 41
2.8 Invariant Intervals
An invariant interval for Eq(2.1) is an interval I with the property that if two con-
secutive terms of the solution fall in I then all the subsequent terms of the solution also
belong to I. In other words I is an invariant interval for the difference equation
x
n+1
= f(x
n
, x
n−1
), n = 0, 1, . . . (2.36)
if x
N−1
, x
N
∈ I for some N ≥ 0, then x
n
∈ I for every n > N.
Invariant intervals for Eq(2.36) are determined from the intervals where the function
f(x, y) is monotonic in its arguments.
For Eq(2.1) the partial derivatives are given by (2.9) and (2.10) and are
f
x
(x, y) =
L +My
(A +Bx + Cy)
2
and f
y
(x, y) =
N −Mx
(A +Bx + Cy)
2
,
where
L = βA −αB, M = βC −γB, and N = γA −αC.
The following table gives the signs of f
x
and f
y
in all possible nondegenerate cases.
Case L M N Signs of derivatives f
x
and f
y
1 + + +
f
x
> 0
f
y
> 0 if x <
N
M
; f
y
< 0 if x >
N
M
2 + + 0 f
x
> 0 and f
y
< 0
3 + + − f
x
> 0 and f
y
< 0
4 + 0 + f
x
> 0 and f
y
> 0
5 + 0 0 f
x
> 0 and f
y
= 0
6 + − +
f
x
> 0 if y < −
L
M
; f
x
< 0 if y > −
L
M
f
y
> 0
7 0 + 0 f
x
> 0 and f
y
< 0
8 0 + − f
x
> 0 and f
y
< 0
9 0 − + f
x
< 0 and f
y
> 0
10 0 − 0 f
x
< 0 and f
y
> 0
11 − + −
f
x
> 0 if y > −
L
M
; f
x
< 0 if y < −
L
M
f
y
< 0
12 − 0 0 f
x
< 0 and f
y
= 0
13 − 0 − f
x
< 0 and f
y
< 0
14 − − + f
x
< 0 and f
y
> 0
15 − − 0 f
x
< 0 and f
y
> 0
16 − − −
f
x
< 0
f
y
> 0 if x >
N
M
; f
y
< 0 if x <
N
M
Using this table, we obtain the following table of invariant intervals for Eq(2.1).
42 Chapter 2. Local Stability, Semicycles, Periodicity, and Invariant Intervals
Case Invariant intervals
1
[0,
N
M
] if
αM+(β+γ)N
AM+(B+C)N
<
N
M
[
N
M
,
(β−A)M−CN+

((β−A)M−NC)
2
+4BM(αM+γN)
2BM
] if
αM+(β+γ)N
AM+(B+C)N
>
N
M
and B > 0
[
N
M
,
αM+γN
(A−β)M+CN
] if B = 0 and (A−β)M +CN > 0
2
[0,
β−A+

(β−A)
2
+4Bα
2B
] if B > 0
[0,
α
A−β
] if B = 0, α > 0 and A > β
3
[0,
β−A+

(β−A)
2
+4Bα
2B
] if B > 0
[0,
α
A−β
] if B = 0, α > 0 and A > β
4 [0,
β+γ−A+

(β+γ−A)
2
+4α(B+C)
2(B+C)
] if B +C > 0
5
[
α
A
,
β−A+

(β−A)
2
+4αB
2B
] if B > 0
[
α
A
,
α
A−β
] if B = 0 and A > β
6
[0, −
L
M
] if
αM−(β+γ)L
AM−(B+C)L
< −
L
M
[−
L
M
,
(A−γ)M−BL+

((A−γ)M−BL)
2
+4CM(αM−βL)
−2CM
] if C = 0 and
βMK+αM−γL
BMK+AM−CL
> −
L
M
[−
L
M
,
βL−αM
(γ−A)M+BL
] if C = 0, (γ −A)M +BL > 0 and −(βM +BL)K ≥ αM +AL
7
[
γ
C
,
β
B
] if B > 0
[
γ
C
,
γ
2
C(γ−β)
] if B = 0 and β < γ
8
[0,
β
B
] if B = 0
[
γ
C
,
αC+γ
2
γ−βC
] if B = 0 and γ > βC
9
[0,
γ
C
] if C > 0
[k, K] if C = 0 where k and K satisfy α +βk + (γ −A)K ≤ BkK ≤ α +βK −Ak
10
[
β
B
,
γ
C
] if C > 0
[
β
B
,
β
2
B(β−γ)
] if C = 0 and β > γ
11
[
αM−L(β+γ)
AM−L(B+C)
, −
L
M
] if
αM−L(β+γ)
AM−L(B+C)
< −
L
M
[−
L
M
,
(β−A)M+LC+

((β−A)M+LC)
2
+4BM(αM−γL)
2BM
] if
αM−βL+γMK
AM−BL+CMK
> −
L
M
12
[
α(A+β)
A
2
+Bα
,
α
A
] if A > 0
[
β
B
,
β
B
+
α
β
] if A = 0 and β > 0
13
[0,
α
A
] if A > 0
[k, K] if A = 0 where k and K satisfy
α+(β+γ)k
B+C
≤ kK ≤
α+(β+γ)K
B+C
14
[0,
γ
C
] if C > 0
[k, K] if C = 0 where k and K satisfy α +βk + (γ −A)K ≤ BkK ≤ α +βK −Ak
15
[0,
γ
C
] if C > 0
[
β
B
,
αB+β
2
B(β−γ)
] if A = 0 and β > γ
16
[
αM+(β+γ)N
AM+(B+C)N
,
N
M
] if
αM+(β+γ)N
AM+(B+C)N
<
N
M
[
N
M
,
(γ−A)M−BN+

((γ−A)M−BN)
2
+4BM(αM+βN)
2CM
] if
αM+βK+γNM
AM+BK+CNM
>
N
M
2.9. Open Problems and Conjectures 43
2.9 Open Problems and Conjectures
What is it that makes all solutions of a nonlinear difference equation periodic with the
same period?
For a linear equation, every solution is periodic with period p ≥ 2, if and only if
every root of the characteristic equation is a pth root of unity.
Open Problem 2.9.1 Assume that f ∈ C
1
[(0, ∞) × (0, ∞), (0, ∞)] is such that every
positive solution of the equation
x
n+1
= f(x
n
, x
n−1
), n = 0, 1, . . . (2.37)
is periodic with period p ≥ 2.
Is it true that the linearized equation about a positive equilibrium point of Eq(2.37)
has the property that every one of its solutions is also periodic with the same period p?
What is it that makes every positive solution of a difference equation converge to a
periodic solution with period p ≥ 2 ?
Open Problem 2.9.2 Let f ∈ C
1
[[0, ∞) × [0, ∞), [0, ∞)] and let p ≥ 2 be an integer.
Find necessary and sufficient conditions so that every nonnegative solution of the differ-
ence equation (2.37) converges to a periodic solution with period p. In particular address
the case p = 2.
Open Problem 2.9.3 Let p ≥ 2 be an integer and assume that every positive solution
of Eq(2.1) with
α, β, γ, A, B, C ∈ [0, ∞)
converges to a solution with period p . Is it true that
p ∈ {2, 4, 5, 6}?
Open Problem 2.9.4 It is known (see Sections 2.7, 4.7, 6.5, and 6.7) that every pos-
itive solution of each of the following three equations
x
n+1
= 1 +
x
n−1
x
n
, n = 0, 1, . . . (2.38)
x
n+1
=
1 +x
n−1
1 + x
n
, n = 0, 1, . . . (2.39)
x
n+1
=
x
n
+ 2x
n−1
1 + x
n
, n = 0, 1, . . . (2.40)
44 Chapter 2. Local Stability, Semicycles, Periodicity, and Invariant Intervals
converges to a solution with (not necessarily prime) period-two:
. . . , φ, ψ, φ, ψ, . . . . (2.41)
In each case, determine φ and ψ in terms of the initial conditions x
−1
and x
0
.
Conversely, if (2.41) is a period-two solution of Eq(2.38) or Eq(2.39) or Eq(2.40), de-
termine all initial conditions (x
−1
, x
0
) ∈ (0, ∞)×(0, ∞) for which the solution {x
n
}

n=−1
converges to the period-two solution (2.41).
Conjecture 2.9.1 Assume
α, β, γ, A, B, C ∈ (0, ∞).
Show that every positive solution of Eq(2.1) is bounded.
Open Problem 2.9.5 Assume
α, β, γ, A, B ∈ (0, ∞) and γ > β +A.
Determine the set of initial conditions x
−1
and x
0
for which the solution {x
n
}

n=−1
of
Eq(2.33) is bounded.
(See Sections 2.7, 4.7, 6.5, 6.7, and 10.2.)
Conjecture 2.9.2 Assume
α, β, γ, A, B, C ∈ (0, ∞)
and that Eq(2.1) has no period-two solution. Show that the positive equilibrium is globally
asymptotically stable.
Conjecture 2.9.3 Assume
α, β, γ, A, B, C ∈ (0, ∞)
and that Eq(2.1) has a positive prime period-two solution. Show that the positive equi-
librium of Eq(2.1) is a saddle point.
2.9. Open Problems and Conjectures 45
Open Problem 2.9.6 Assume
α, β, γ, A, B, C ∈ [0, ∞).
Obtain necessary and sufficient conditions in terms of the coefficients so that every
positive solution of Eq(2.1) is bounded.
Conjecture 2.9.4 Assume that
α, β, γ, A, B, C ∈ [0, ∞).
Show that Eq(2.1) cannot have a positive prime two cycle which attracts all positive
non-equilibrium solutions.
Open Problem 2.9.7 Assume f ∈ C[(0, ∞), (0, ∞)]. Obtain necessary and sufficient
conditions on f for every solution of the difference equation
x
n+1
=
f(x
n
)
x
n−1
, n = 0, 1, . . .
to be bounded.
(See Section 5.2).
Open Problem 2.9.8 Assume f ∈ C[(0, ∞), (0, ∞)]. Obtain necessary and sufficient
conditions on f for every positive solution of the equation
x
n+1
= 1 +
f(x
n−1
)
x
n
, n = 0, 1, . . .
to converge to a period-two solution.
(See Section 4.7).
Open Problem 2.9.9 Extend Theorem 2.7.1 to third order difference equations
x
n+1
=
α + βx
n
+ γx
n−1
+δx
n−2
A +Bx
n−1
+ Dx
n−2
, n = 0, 1, . . .
with nonnegative parameters and nonnegative initial conditions such that
γ = β +δ +A.
What additional conditions should the parameters satisfy?
46 Chapter 2. Local Stability, Semicycles, Periodicity, and Invariant Intervals
Open Problem 2.9.10 Extend the linearized stability Theorem 1.1.1 to third order dif-
ference equations
x
n+1
= f(x
n
, x
n−1
, x
n−2
), n = 0, 1, . . . . (2.42)
For the equation
λ
3
+aλ
2
+ bλ +c = 0
one can see that all roots lie inside the unit disk |λ| < 1 if and only if
|a +c| < 1 +b
|a −3c| < 3 −b
b +c
2
< 1 +ac

.
But how about necessary and sufficient conditions for the equilibrium of Eq(2.42) to
be a repeller, or a saddle point, or nonhyperbolic?
Extend these results to fourth order difference equation
x
n+1
= f(x
n
, x
n−1
, x
n−2
, x
n−3
), n = 0, 1, . . . .
Chapter 3
(1, 1)-Type Equations
3.1 Introduction
Eq(1) contains the following nine equations of the (1, 1)-type:
x
n+1
=
α
A
, n = 0, 1, . . . (3.1)
x
n+1
=
α
Bx
n
, n = 0, 1, . . . (3.2)
x
n+1
=
α
Cx
n−1
, n = 0, 1, . . . (3.3)
x
n+1
=
βx
n
A
, n = 0, 1, . . . (3.4)
x
n+1
=
β
B
, n = 0, 1, . . . (3.5)
x
n+1
=
βx
n
Cx
n−1
, n = 0, 1, . . . (3.6)
47
48 Chapter 3. (1, 1)-Type Equations
x
n+1
=
γx
n−1
A
, n = 0, 1, . . . (3.7)
x
n+1
=
γx
n−1
Bx
n
, n = 0, 1, . . . (3.8)
and
x
n+1
=
γ
C
, n = 0, 1, . . . . (3.9)
Please recall our classification convention in which all parameters that appear in these
equations are positive, the initial conditions are nonnegative, and the denominators are
always positive.
Of these nine equations, Eqs(3.1), (3.5), and (3.9) are trivial. Eqs (3.4) and (3.7)
are linear difference equation. Every solution of Eq(3.2) is periodic with period two and
every solution of Eq(3.4) is periodic with period four. The remaining two equations,
namely, Eqs(3.6) and (3.8), are treated in the next two sections.
3.2 The Case α = γ = A = B = 0 : x
n+1
=
βx
n
Cx
n−1
This is the (1, 1)-type Eq(3.6) which by the change of variables
x
n
=
β
C
y
n
reduces to the equation
y
n+1
=
y
n
y
n−1
, n = 0, 1, . . . . (3.10)
It is easy to see that every positive solution of Eq(3.10) is periodic with period six.
Indeed the solution with initial conditions y
−1
and y
0
is the six-cycle:
y
−1
, y
0
,
y
0
y
−1
,
1
y
−1
,
1
y
0
,
y
−1
y
0
, . . . .
It is interesting to note that except for the equilibrium solution y = 1, every other
solution of Eq(3.10) has prime period six.
3.3. The Case α = β = A = C = 0 : x
n+1
=
γx
n−1
Bx
n
49
3.3 The Case α = β = A = C = 0 : x
n+1
=
γx
n−1
Bx
n
This is the (1, 1)-type Eq(3.8) which by the change of variables
x
n
=
γ
B
e
y
n
reduces to the linear equation
y
n+1
+ y
n
−y
n−1
= 0, n = 0, 1, . . . . (3.11)
The general solution of Eq(3.11) is
y
n
= C
1

−1 +

5
2

n
+ C
2

−1 −

5
2

n
, n = 0, 1, . . .
from which the behavior of solutions is easily derived.
3.4 Open Problems and Conjectures
Of the nine equations in this chapter we would like to single out the two nonlinear
equations, (3.2) and (3.6), which possess the property that every positive nontrivial
solution of each of these two equations is periodic with the same prime period, namely
two and six, respectively. Eq(3.3) also has the property that every nontrivial positive
solution is periodic with prime period four but this equation is really a variant of Eq(3.2).
Also every solution of Eq(3.7) is periodic with period two but this is a linear equation.
Eq(3.2) is of the form
x
n+1
= f(x
n
), n = 0, 1, . . . (3.12)
where
f ∈ C[(0, ∞), (0, ∞)], (3.13)
while Eq(3.6) is of the form
x
n+1
= f(x
n
, x
n−1
), n = 0, 1, . . . (3.14)
where
f ∈ C[(0, ∞) ×(0, ∞), (0, ∞)]. (3.15)
What is it that makes every solution of a nonlinear difference equation periodic with the
same period? We believe this is a question of paramount importance and so we pose the
following open problems.
50 Chapter 3. (1, 1)-Type Equations
Open Problem 3.4.1 Let k ≥ 2 be a positive integer and assume that (3.13) holds.
Obtain necessary and sufficient conditions on f so that every positive solution of Eq(3.12)
is periodic with period k.
Open Problem 3.4.2 Let k ≥ 2 be a positive integer and assume that (3.15) holds.
Obtain necessary and sufficient conditions on f so that every positive solution of Eq(3.14)
is periodic with period k. In particular address the cases
k = 4, 5, 6.
Open Problem 3.4.3 Let k be a positive integer and assume that
α, β, γ ∈ (−∞, ∞).
Obtain necessary and sufficient conditions on α, β, γ so that every solution of the equa-
tion
x
n+1
= α|x
n
| + βx
n−1
+ γ, n = 0, 1, . . .
with real initial conditions is periodic with period k. In particular address the cases:
k = 5, 6, 7, 8, 9.
(See [10], [17], and [59].)
Conjecture 3.4.1 Let f ∈ C
1
[[0, ∞), [0, ∞)] be such that every positive nontrivial so-
lution of the difference equation
x
n+1
=
f(x
n
)
x
n−1
, n = 0, 1, . . . (3.16)
is periodic with prime period six. Show that
f(x) = x.
Conjecture 3.4.2 Let f ∈ C
1
[[0, ∞), [0, ∞)] be such that every positive nontrivial so-
lution of the difference equation (3.16) is periodic with prime period five. Show that
f(x) = 1 + x.
3.4. Open Problems and Conjectures 51
Open Problem 3.4.4 Obtain necessary and sufficient conditions on f ∈ C[[0, ∞), [0, ∞)]
such that every positive nontrivial solution of the equation
x
n+1
=
f(x
n
)
x
n−1
, n = 0, 1, . . . (3.17)
is periodic with prime period three.
Open Problem 3.4.5 Let {p
n
}

n=0
be a sequence of nonzero real numbers. In each of
the following cases investigate the asymptotic behavior of all positive solutions of the
difference equation
x
n+1
=
p
n
x
n
x
n−1
, n = 0, 1, . . . . (3.18)
(i) {p
n
}

n=0
converges to a finite limit;
(ii) {p
n
}

n=0
converges to ±∞;
(iii) {p
n
}

n=0
is periodic with prime period k ≥ 2.
Chapter 4
(1, 2)-Type Equations
4.1 Introduction
Eq(1) contains the following nine equations of the (1, 2)-type:
x
n+1
=
α
A + Bx
n
, n = 0, 1, . . . (4.1)
x
n+1
=
α
A + Cx
n−1
, n = 0, 1, . . . (4.2)
x
n+1
=
α
Bx
n
+ Cx
n−1
, n = 0, 1, . . . (4.3)
x
n+1
=
βx
n
A + Bx
n
, n = 0, 1, . . . (4.4)
x
n+1
=
βx
n
A + Cx
n−1
, n = 0, 1, . . . (4.5)
x
n+1
=
βx
n
Bx
n
+ Cx
n−1
, n = 0, 1, . . . (4.6)
53
54 Chapter 4. (1, 2)-Type Equations
x
n+1
=
γx
n−1
A + Bx
n
, n = 0, 1, . . . (4.7)
x
n+1
=
γx
n−1
A + Cx
n−1
, n = 0, 1, . . . (4.8)
and
x
n+1
=
γx
n−1
Bx
n
+ Cx
n−1
, n = 0, 1, . . . . (4.9)
Please recall our classification convention in which all parameters that appear in these
equations are positive, the initial conditions are nonnegative, and the denominators are
always positive.
Two of these equations, namely Eqs(4.1) and (4.4) are Riccati-type difference equa-
tions. (See Section 1.6 in the Preliminary Results). Also Eqs(4.2) and (4.8) are essen-
tially Riccati equations. Indeed if {x
n
} is a solution of Eq(4.2), then the subsequences
{x
2n−1
} and {x
2n
} satisfy the Riccati equation of the form of Eq(4.1). Similarly, if {x
n
}
is a solution of Eq(4.8), then the subsequences {x
2n−1
} and {x
2n
} satisfy the Riccati
equation
y
n+1
=
γy
n
A + Cy
n
n = 0, 1, . . . .
It is also interesting to note that the change of variables
x
n
=
1
y
n
reduces the Riccati equation (4.4) to the linear equation
y
n+1
=
A
β
y
n
+
B
β
, n = 0, 1, . . .
from which the global behavior of solutions is easily derived.
In view of the above, Eqs(4.2), (4.4), and (4.8) will not be discussed any further in
this chapter.
4.2 The Case β = γ = C = 0 : x
n+1
=
α
A+Bx
n
This is the (1, 2)-type Eq(4.1) which by the change of variables
x
n
=
A
B
y
n
4.3. The Case β = γ = A = 0 : x
n+1
=
α
Bx
n
+Cx
n−1
55
reduces to the Riccati equation
y
n+1
=
p
1 + y
n
, n = 0, 1, . . . (4.10)
where
p =
αB
A
2
.
The Riccati number associated with this equation (see Section 1.6) is
R = −p < 0
and so the following result is a consequence of Theorem 1.6.2.
Theorem 4.2.1 The positive equilibrium
y =
−1 +

1 + 4p
2
of Eq(4.10) is globally asymptotically stable.
4.3 The Case β = γ = A = 0 : x
n+1
=
α
Bx
n
+Cx
n−1
This is the (1, 2)-type Eq(4.3) which by the change of variables
x
n
=

α
y
n
(4.11)
is transformed to the difference equation
y
n+1
=
B
y
n
+
C
y
n−1
, n = 0, 1, . . . . (4.12)
Eq(4.12) was investigated in [65]. (See also [12].)
We will first show that every solution of Eq(4.12) is bounded and then we will use
the method of “limiting solutions” to show that every solution of Eq(4.12) converges to
its equilibrium

B + C.
Theorem 4.3.1 Every solution of Eq(4.12) is bounded and persists.
Proof. It is easy to see that if a solution of Eq(4.12) is bounded from above then it is
also bounded from below and vice-versa. So if the theorem is false there should exist a
solution {y
n
}

n=−1
which is neither bounded from above nor from below. That is,
limsup
n→∞
y
n
= ∞ and liminf
n→∞
y
n
= 0.
56 Chapter 4. (1, 2)-Type Equations
Then clearly we can find indices i and j with 1 ≤ i < j such that
y
i
> y
n
> y
j
for all n ∈ {−1, . . . , j −1}.
Hence
y
j
=
B
y
j−1
+
C
y
j−2
>
B + C
y
i
and
y
i
=
B
y
i−1
+
C
y
i−2

B + C
y
j
.
That is ,
B + C < y
i
y
j
≤ B + C
which is impossible. 2
Theorem 4.3.2 The equilibrium y =

B + C of Eq(4.12) is globally asymptotically
stable.
Proof. The linearized equation of Eq(4.12) about the equilibrium y =

B + C is
z
n+1
= −
B
B + C
z
n

C
B + C
z
n−1
, n = 0, 1, . . . . (4.13)
and by Theorem 1.1.1, y is locally asymptotically stable for all positive values of B and
C.
It remains to be shown that y is a global attractor of all solutions of Eq(4.12). To
this end, let {y
n
}

n=−1
be a solution of Eq(4.12). Then by Theorem 4.3.1 , {y
n
}

n=−1
is
bounded and persists. Therefore there exist positive numbers m and M such that
m ≤ y
n
≤ M for n ≥ −1.
Let
I = liminf
n→∞
y
n
and S = limsup
n→∞
y
n
.
We will show that
I = S.
By the theory of limiting solutions (see Section 1.5) there exist two solutions {I
n
}

n=−3
and {S
n
}

n=−3
of the difference equation
y
n+1
=
B
y
n
+
C
y
n−1
, n = −3, −2, . . .
with the following properties:
I
0
= I, S
0
= S
4.4. The Case α = γ = B = 0 : x
n+1
=
βx
n
A+Cx
n−1
- Pielou’s Equation 57
and
I
n
, S
n
∈ [I, S] for n ≥ −3.
Then
S =
B
S
−1
+
C
S
−2

B + C
I
and
I =
B
I
−1
+
C
I
−2

B + C
S
.
Hence
B + C = SI.
Thus
B
S
−1
+
C
S
−2
= S =
B + C
I
=
B
I
+
C
I
from which it follows that
S
−1
= S
−2
= I. (4.14)
Therefore
B
S
+
C
S
= I = S
−1
=
B
S
−2
+
C
S
−3
from which it follows that
S = S
−2
. (4.15)
From (4.14) and (4.15) we conclude that
S = I
and the proof is complete. 2
4.4 The Case α = γ = B = 0 : x
n+1
=
βx
n
A+Cx
n−1
- Pielou’s
Equation
This is the (1, 2)-type Eq(4.5) which by the change of variables
x
n
=
A
C
y
n
,
reduces to Pielou’s difference equation
y
n+1
=
py
n
1 + y
n−1
, n = 0, 1, . . . (4.16)
where
p =
β
A
.
58 Chapter 4. (1, 2)-Type Equations
This equation was proposed by Pielou in her books ([66], p.22) and ([67], p.79) as a
discrete analogue of the delay logistic equation
dN
dt
= rN(t)

1 −
N(t −τ)
P

, t ≥ 0
which is a prototype of modelling the dynamics of single-species.
Eq(4.16) was investigated in [55]. (See also ([42], p.75).)
When
p ≤ 1,
it follows from Eq(4.16) that every positive solution converges to 0.
Furthermore, 0 is locally asymptotically stable when p ≤ 1 and unstable when p > 1.
So the interesting case for Eq(4.16) is when
p > 1. (4.17)
In this case the zero equilibrium of Eq(4.16) is unstable and Eq(4.16) possesses the
unique positive equilibrium
y = p −1,
which is locally asymptotically stable.
In fact by employing Theorem 1.4.2, it follows that when (4.17) holds, x is globally
asymptotically stable.
We summarize the above discussion in the following theorem.
Theorem 4.4.1 (a) Assume
p ≤ 1.
Then the zero equilibrium of Eq(4.16) is globally asymptotically stable.
(b) Assume y
0
∈ (0, ∞) and
p > 1.
Then the positive equilibrium y = p −1 of Eq(4.16) is globally asymptotically stable.
4.5 The Case α = γ = A = 0 : x
n+1
=
βx
n
Bx
n
+Cx
n−1
This is the (1, 2)-type Eq(4.6) which by the change of variables
x
n
=
β
B + Cy
n
,
reduces to the difference equation
y
n+1
=
p + y
n
p + y
n−1
, n = 0, 1, . . . (4.18)
4.6. The Case α = β = C = 0 : x
n+1
=
γx
n−1
A+Bx
n
59
where
p =
B
C
.
Eq(4.18) was investigated in ([42] Corollary 3.4.1 (e), p. 73) where it was shown that
its equilibrium one is globally asymptotically stable .
4.6 The Case α = β = C = 0 : x
n+1
=
γx
n−1
A+Bx
n
This is the (1, 2)-type Eq(4.7) which by the change of variables
x
n
=
γ
B
y
n
,
reduces to the difference equation
y
n+1
=
y
n−1
p + y
n
, n = 0, 1, . . . (4.19)
where
p =
A
γ
∈ (0, ∞).
Eq(4.19) was investigated in [28].
The following local result is a straightforward application of the linearized stability
Theorem 1.1.1.
Lemma 4.6.1 (a) Assume
p > 1.
Then 0 is the only equilibrium point of Eq(4.19) and it is locally asymptotically stable.
(b) Assume
p < 1.
Then 0 and y = 1 − p are the only equilibrium points of Eq(4.19) and they are both
unstable. In fact, 0 is a repeller and y = 1 −p is a saddle point equilibrium.
The oscillatory character of the solutions of Eq(4.19) is a consequence of Theorem
1.7.1. That is, except possibly for the first semicycle, every solution of Eq(4.19) has
semicycles of length one.
It follows directly from Eq(4.19) (see also Section 2.5) that
y
n+1
<
1
p
y
n−1
for n ≥ 0
and so when
p > 1
60 Chapter 4. (1, 2)-Type Equations
the zero equilibrium of Eq(4.19) is globally asymptotically stable. On the other hand
when
p = 1,
y
n+1
< y
n−1
for n ≥ 0
and so every positive solution of Eq(4.19) converges to a period-two solution
. . . , φ, ψ, φ, ψ, . . . .
Furthermore we can see that
φψ = 0
but whether there exists solutions with
φ = ψ = 0
remains an open question.
Finally when
p < 1
by invoking the stable manifold theory one can see that there exist nonoscillatory solu-
tions which converge monotonically to the positive equilibrium 1−p. It follows from the
identity
y
n+1
−y
n−1
=
(1 −p) −y
n
p + y
n
y
n−1
, n = 0, 1, . . .
that every oscillatory solution is such that the subsequences of even and odd terms
converge monotonically one to ∞ and the other to 0.
4.7 The Case α = β = A = 0 : x
n+1
=
γx
n−1
Bx
n
+Cx
n−1
This is the (1, 2)-type Eq(4.9) which by the change of variables
x
n
=
γ
By
n
reduces to the difference equation
y
n+1
= p +
y
n−1
y
n
, n = 0, 1, . . . (4.20)
where
p =
C
B
∈ (0, ∞).
This equation was investigated in [7] where it was shown that the following statements
are true:
4.7. The Case α = β = A = 0 : x
n+1
=
γx
n−1
Bx
n
+Cx
n−1
61
(i) p ≥ 1 is a necessary and sufficient condition for every solution of Eq(4.20) to be
bounded;
(ii) When p = 1, every solution of Eq(4.20) converges to a period-two solution;
(iii) When p > 1, the equilibrium y = p + 1 of Eq(4.20) is globally asymptotically
stable.
We now proceed to establish the above results.
Clearly the only equilibrium point of Eq(4.20) is y = p + 1.
The linearized equation of Eq(4.20) about the equilibrium point y = p + 1 is
z
n+1
+
1
p + 1
z
n

1
p + 1
z
n−1
= 0, n = 0, 1, . . . . (4.21)
As a simple consequence of the linearized stability Theorem 1.1.1 we obtain the following
result:
Theorem 4.7.1 The equilibrium point y = p + 1 of Eq(4.20) is locally asymptotically
stable when p > 1 and is an unstable saddle point when p < 1.
The next result about semicycles is an immediate consequence of some straightfor-
ward arguments and Theorem 1.7.1.
Theorem 4.7.2 (a) Let {y
n
}

n=−1
be a solution of Eq(4.20) which consists of at least
two semicycles. Then {y
n
}

n=−1
is oscillatory. Moreover, with the possible excep-
tion of the first semicycle, every semicycle has length one.
(b) Let {y
n
}

n=−1
be a solution of Eq(4.20) which consists of a single semicycle. Then
{y
n
}

n=−1
converges monotonically to y = p + 1.
4.7.1 The Case p < 1
Here we show that there exist solutions of Eq(4.20) which are unbounded. In fact the
following result is true.
Theorem 4.7.3 Let p < 1, and let {y
n
}

n=−1
be a solution of Eq(4.20) such that 0 <
y
−1
≤ 1 and y
0

1
1−p
. Then the following statements are true:
1. lim
n→∞
y
2n
= ∞.
2. lim
n→∞
y
2n+1
= p.
Proof. Note that
1
1−p
> p + 1, and so y
0
> p + 1. It suffices to show that
y
1
∈ (p, 1] and y
2
≥ p + y
0
and use induction to complete the proof.
62 Chapter 4. (1, 2)-Type Equations
Indeed y
1
= p +
y
1
y
0
> p. Also,
y
1
= p +
y
−1
y
0
≤ p +
1
y
0
≤ 1,
and so y
1
∈ (p, 1]. Hence y
2
= p +
y
0
y
1
≥ p + y
0
. 2
4.7.2 The Case p = 1
The following result follows from Lemma 2.7.1 and some straightforward arguments. See
also Section 2.5.
Theorem 4.7.4 Let p = 1, and let {y
n
}

n=−1
be a solution of Eq(4.20). Then the
following statements are true.
(a) Suppose {y
n
}

n=−1
consists of a single semicycle. Then {y
n
}

n=−1
converges mono-
tonically to y = 2.
(b) Suppose {y
n
}

n=−1
consists of at least two semicycles. Then {y
n
}

n=−1
converges to
a prime period-two solution of Eq(4.20).
(c) Every solution of Eq(4.20) converges to a period-two solution if and only if p = 1.
4.7.3 The Case p > 1
Here we show that the equilibrium point y = p+1 of Eq(4.20) is globally asymptotically
stable. The following result will be useful.
Lemma 4.7.1 Let p > 1, and let {y
n
}

n=−1
be a solution of Eq(4.20). Then
p +
p −1
p
≤ liminf
n→∞
y
n
≤ limsup
n→∞
y
n

p
2
p −1
.
Proof. It follows from Theorem 4.7.2 that we may assume that every semicycle of
{y
n
}

n=−1
has length one, that p < y
n
for all n ≥ −1, and that p < y
0
< p + 1 < y
−1
.
We shall first show that limsup
n→∞
y
n

p
2
p−1
. Note that for n ≥ 0,
y
2n+1
< p +
y
2n−1
p
.
So as every solution of the difference equation
y
m+1
= p +
1
p
y
m
, m = 0, 1, . . .
4.7. The Case α = β = A = 0 : x
n+1
=
γx
n−1
Bx
n
+Cx
n−1
63
converges to
p
2
p−1
, it follows that limsup
n→∞
y
n

p
2
p−1
.
We shall next show that p +
p−1
p
≤ liminf
n→∞
y
n
. Let > 0. There clearly exists
N ≥ 0 such that for all n ≥ N,
y
2n−1
<
p
2
+
p −1
.
Let n ≥ N. Then
y
2n
= p +
y
2n−2
y
2n−1
> p + p
p −1
p
2
+
=
p
3
+ p + p(p −1)
p
2
+
.
So as is arbitrary, we have
liminf
n→∞
y
n

p
3
+ p(p −1)
p
2
= p +
p −1
p
.
2
We are now ready for the following result.
Theorem 4.7.5 Let p > 1. Then the equilibrium y = p + 1 is globally asymptotically
stable.
Proof. We know by Theorem 4.7.1 that y = p + 1 is a locally asymptotically stable
equilibrium point of Eq(4.20). So let {y
n
}

n=−1
be a solution of Eq(4.20). It suffices to
show that
lim
n→∞
y
n
= p + 1.
For x, y ∈ (0, ∞), set
f(x, y) = p +
y
x
.
Then f ∈ C[(0, ∞) ×(0, ∞), (0, ∞)] , f is decreasing in x ∈ (0, ∞) for each y ∈ (0, ∞),
and f is increasing in y ∈ (0, ∞) for each x ∈ (0, ∞). Recall that by Theorem 4.7.4,
there exist no solutions of Eq(4.20) with prime period two. Let > 0, and set
a = p and b =
p
2
+
p −1
.
Note that
f

p
2
+
p −1
, p

= p + p
p −1
p
2
+
> p
and
f

p,
p
2
+
p −1

=
p
3
+ p
p
2
−p
=
p
2
+
p −1
.
Hence
p < f(x, y) <
p
2
+
p −1
for all x, y ∈

p,
p
2
+
p −1

.
64 Chapter 4. (1, 2)-Type Equations
Finally note that by Lemma 4.7.1,
p < p +
p −1
p
≤ liminf
n→∞
y
n
≤ limsup
n→∞
y
n

p
2
p −1
<
p
2
+
p −1
and so by Theorem 1.4.6,
lim
n→∞
y
n
= p + 1.
2
4.8 Open Problems and Conjectures
Conjecture 4.8.1 Consider the difference equation
x
n+1
=
A
x
n−k
+
B
x
n−l
, n = 0, 1, . . . (4.22)
where
A, B ∈ (0, ∞) and k, l ∈ {0, 1, . . .} with k < l.
and where the initial conditions x
−l
, . . . , x
0
are arbitrary positive real numbers.
Show that every solution of Eq(4.22) converges, either to the equilibrium or to a
periodic solution with period p ≥ 2 and that what happens and the exact value of p are
uniquely determined from the characteristic roots of the linearized equation
λ
l+1
+
A
A + B
λ
l−k
+
B
A + B
= 0.
(See [19], [23], [25], [60], and [61].)
Open Problem 4.8.1 Let
. . . , φ
1
, φ
2
, . . . , φ
p
, . . .
be a given prime period p solution of Eq(4.22). Determine the set of all positive initial
conditions x
−l
, . . . , x
0
such that {x
n
}

n=−l
converges to this periodic solution.
Open Problem 4.8.2 (see [18]) Assume that A, B ∈ (0, ∞).
(a) Find all initial conditions y
−1
, y
0
with
y
−1
y
0
< 0
for which the equation
y
n+1
=
A
y
n
+
B
y
n−1
, (4.23)
is well defined for all n ≥ 0.
(b) Assume that the initial conditions y
−1
, y
0
are such that y
−1
y
0
< 0 and such that
Eq(4.23) is well defined for all n ≥ 0. Investigate the global behavior of the solution
{y
n
}. Is it bounded? Does lim
n→∞
y
n
exist? When does {y
n
} converge to a two cycle?
4.8. Open Problems and Conjectures 65
Conjecture 4.8.2 Assume that p = 1. Show that Eq(4.19) has a solution which con-
verges to zero.
Open Problem 4.8.3 Assume that p < 1. Determine the set of initial conditions
y
−1
, y
0
∈ (0, ∞)
for which the solution {y
n
}

n=−1
of Eq(4.19) is bounded.
Open Problem 4.8.4 Determine the set G of all initial conditions (y
−1
, y
0
) ∈ R × R
through which the equation
y
n+1
=
y
n−1
1 + y
n
is well defined for all n ≥ 0 and, for these initial points, investigate the global character
of {y
n
}

n=−1
.
Conjecture 4.8.3 Show that Eq(4.20) possesses a solution {y
n
}

n=−1
which remains
above the equilibrium for all n ≥ −1.
Open Problem 4.8.5 (a) Assume p ∈ (0, ∞). Determine all initial conditions y
−1
, y
0

R such that the equation (4.20) is well defined for all n ≥ 0.
(b) Let y
−1
, y
0
∈ R be such that Eq(4.20) is well defined for all n ≥ 0. For such an
initial point, investigate the boundedness, the asymptotic behavior, and the periodic
nature of the solution {y
n
}

n=−1
of Eq(4.20).
(c) Discuss (a) and (b) when p < 0.
Open Problem 4.8.6 (See [13]) Find the set G of all initial points (x
−1
, x
0
) ∈ R ×R
through which the equation
x
n+1
= −1 +
x
n−1
x
n
(4.24)
is well defined for all n ≥ 0.
Conjecture 4.8.4 Assume that (x
−1
, x
0
) ∈ R × R is such that the equation (4.24) is
well defined for all n ≥ 0 and bounded. Show that {x
n
}

n=−1
is a three cycle.
66 Chapter 4. (1, 2)-Type Equations
Open Problem 4.8.7 Let f ∈ C[(0, ∞), (0, ∞)]. Obtain necessary and sufficient con-
ditions on f for every positive solution of the equation
x
n+1
=
f(x
n
)
x
n
+ x
n−1
, n = 0, 1, . . .
to be bounded.
Open Problem 4.8.8 Let a, a
i
∈ (0, ∞) for i = 0, . . . , k. Investigate the global asymp-
totic stability of the equilibrium points of the difference equation
x
n+1
=
x
n
a + a
0
x
n
+ . . . + a
k
x
n−k
, n = 0, 1, . . .
with positive initial conditions.
Conjecture 4.8.5 (See [27]) Assume r ∈ (0, ∞). Show that every positive solution of
the population model
J
n+1
= J
n−1
e
r−(J
n
+J
n−1
)
, n = 0, 1, . . .
converges to a period-two solution.
Conjecture 4.8.6 (Pielou’s Discrete Logistic Model; see [55], [66], and [67].) As-
sume
α ∈ (1, ∞) and k ∈ {0, 1, . . .}.
Show that the positive equilibrium of Pielou’s discrete logistic model
x
n+1
=
αx
n
1 + x
n−k
, n = 0, 1, . . .
with positive initial conditions is globally asymptotically stable if and only if
α −1
α
< 2 cos

2k + 1
.
Open Problem 4.8.9 (See [5]) For which initial values x
−1
, x
0
∈ (0, ∞) does the se-
quence defined by
x
n+1
= 1 +
x
n−1
x
n
, n = 0, 1, . . .
converges to two?
4.8. Open Problems and Conjectures 67
(See also Section 4.7)
Open Problem 4.8.10 (See [31]) Consider the difference equation
y
n+1
= y
n−1
e
−y
n
, n = 0, 1, . . . (4.25)
with nonnegative initial conditions.
One can easily see that every solution {y
n
}

n=−1
of Eq(4.25) converges to a period-two
solution
. . . , , 0, , 0, . . . .
More precisely, each of the subsequences {y
2n
}

n=0
and {y
2n+1
}

n=−1
is decreasing and
[ lim
n→∞
y
2n
][ lim
n→∞
y
2n+1
] = 0.
(a) Find all initial points y
−1
, y
0
∈ [0, ∞) through which the solutions of Eq(4.25)
converge to zero. In other words, find the basin of attraction of the equilibrium of
Eq(4.25).
(b) Determine the limits of the subsequences of even and odd terms of every solution
of Eq(4.25), in terms of the initial conditions y
−1
and y
0
of the solution.
Open Problem 4.8.11 Assume p ∈ (0, 1).
(a) Find the set B of all initial conditions y
−1
, y
0
∈ (0, ∞) such that the solutions
{y
n
}

n=−1
of Eq(4.20) are bounded.
(b) Let y
−1
, y
0
∈ B. Investigate the asymptotic behavior of {y
n
}

n=−1
.
Open Problem 4.8.12 Let {p
n
}

n=0
be a convergent sequence of nonnegative real num-
bers with a finite limit,
p = lim
n→∞
p
n
.
Investigate the asymptotic behavior and the periodic nature of all positive solutions of
each of the following difference equations:
y
n+1
=
p
n
y
n
+
1
y
n−1
, n = 0, 1, . . . (4.26)
y
n+1
=
p
n
y
n
1 + y
n−1
, n = 0, 1, . . . (4.27)
68 Chapter 4. (1, 2)-Type Equations
y
n+1
=
p
n
+ y
n
p
n
+ y
n−1
, n = 0, 1, . . . (4.28)
y
n+1
=
y
n−1
p
n
+ y
n
, n = 0, 1, . . . (4.29)
y
n+1
= p
n
+
y
n−1
y
n
, n = 0, 1, . . . . (4.30)
Open Problem 4.8.13 Let {p
n
}

n=0
be a periodic sequence of nonnegative real numbers
with period k ≥ 2. Investigate the global character of all positive solutions of each of
Eqs(4.26) - (4.30).
Open Problem 4.8.14 For each of the following difference equations determine the
“good” set G ⊂ R × R of all initial conditions (y
−1
, y
0
) ∈ R × R through which the
equation is well defined for all n ≥ 0. Then for every (y
−1
, y
0
) ∈ G, investigate the long
term behavior of the solution {y
n
}

n=−1
:
y
n+1
=
y
n
1 −y
n−1
, (4.31)
y
n+1
=
1 + y
n
1 −y
n−1
, (4.32)
y
n+1
=
y
n−1
1 −y
n
, (4.33)
y
n+1
= −1 +
y
n−1
y
n
. (4.34)
For those equations from the above list for which you were successful, extend your result
by introducing arbitrary real parameters in the equation.
Chapter 5
(2, 1)-Type Equations
5.1 Introduction
Eq(1) contains the following nine equations of the (2, 1)-type:
x
n+1
=
α + βx
n
A
, n = 0, 1, . . . (5.1)
x
n+1
=
α + βx
n
Bx
n
, n = 0, 1, . . . (5.2)
x
n+1
=
α + βx
n
Cx
n−1
, n = 0, 1, . . . (5.3)
x
n+1
=
α + γx
n−1
A
, n = 0, 1, . . . (5.4)
x
n+1
=
α + γx
n−1
Bx
n
, n = 0, 1, . . . (5.5)
x
n+1
=
α + γx
n−1
Cx
n−1
, n = 0, 1, . . . (5.6)
69
70 Chapter 5. (2, 1)-Type Equations
x
n+1
=
βx
n
+ γx
n−1
A
, n = 0, 1, . . . (5.7)
x
n+1
=
βx
n
+ γx
n−1
Bx
n
, n = 0, 1, . . . (5.8)
and
x
n+1
=
βx
n
+ γx
n−1
Cx
n−1
, n = 0, 1, . . . . (5.9)
Please recall our classification convention in which all parameters that appear in these
equations are positive, the initial conditions are nonnegative, and the denominators are
always positive.
Of these nine equations, Eqs(5.1), (5.4) and (5.7) are linear. Eq(5.2) is a Riccati
equation and Eq(5.6) is essentially like Eq(5.2).
The change of variables
x
n
=
γ
B
y
n
,
reduces Eq(5.8) to Eq(4.20) with p =
β
γ
, which was investigated in Section 4.7.
Finally the change of variables
x
n
=
γ
C
+
β
C
y
n
,
reduces Eq(5.9) to Eq(4.18) with p =
γ
β
, which was investigated in Section 4.5.
Therefore there remain Eqs(5.3) and (5.5), which will be investigated in the next two
sections.
5.2 The Case γ = A = B = 0 : x
n+1
=
α+βx
n
Cx
n−1
- Lyness’
Equation
This is the (2, 1)-type Eq(5.3) which by the change of variables reduces to the equation
y
n+1
=
p + y
n
y
n−1
, n = 0, 1, . . . (5.10)
5.2. The Case γ = A = B = 0 : x
n+1
=
α+βx
n
Cx
n−1
- Lyness’ Equation 71
where
p =
αC
β
2
.
The special case of Eq(5.10) where
p = 1
was discovered by Lyness in 1942 while he was working on a problem in Number Theory.
(See ([42], p. 131) for the history of the problem.) In this special case, the equation
becomes
y
n+1
=
1 + y
n
y
n−1
, n = 0, 1, . . . (5.11)
every solution of which is periodic with period five. Indeed the solution of Eq(5.11) with
initial conditions y
−1
and y
0
is the five-cycle:
y
−1
, y
0
,
1 + y
0
y
−1
,
1 + y
−1
+ y
0
y
−1
y
0
,
1 + y
−1
y
0
, . . . .
Eq(5.10) possesses the invariant
I
n
= (p + y
n−1
+ y
n
)

1 +
1
y
n−1

1 +
1
y
n

= constant (5.12)
from which it follows that every solution of Eq(5.10) is bounded from above and from
below by positive constants.
It was shown in [30] that no nontrivial solution of Eq(5.10) has a limit.
It was also shown in [44] by using KAM theory that the positive equilibrium y of
Eq(5.10) is stable but not asymptotically stable. This result was also established in [49]
by using a Lyapunov function.
Concerning the semicycles of solutions of Eq(5.10) the following result was established
in [43]:
Theorem 5.2.1 Assume that p > 0. Then the following statements about the positive
solutions of Eq(5.10) are true:
(a) The absolute extreme in a semicycle occurs in the first or in the second term.
(b) Every nontrivial semicycle, after the first one, contains at least two and at most
three terms.
There is substantial literature on Lyness’ Equation. See [11], [42], [43], [57]-[59], [71],
and [72] and the references cited therein.
72 Chapter 5. (2, 1)-Type Equations
5.3 The Case β = A = C = 0 : x
n+1
=
α+γx
n−1
Bx
n
This is the (2, 1)-type Eq(5.5) which by the change of variables
x
n
=
γ
B
y
n
reduces to the equation
y
n+1
=
p + y
n−1
y
n
, n = 0, 1, . . . (5.13)
where
p =
αB
γ
2
.
Eq(5.13) was investigated in [28].
By the linearized stability Theorem 1.1.1 one can see that the unique equilibrium
point
y =
1 +

1 + 4p
2
of Eq(5.13) is a saddle point.
By Theorem 1.7.1 one can see that except possibly for the first semicycle, every
oscillatory solution of Eq(5.13) has semicycles of length one.
Concerning the nonoscillatory solutions of Eq(5.13) one can easily see that every
nonoscillatory solution of Eq(5.13) converges monotonically to the equilibrium y.
The global character of solutions of Eq(5.13) is a consequence of the identity
y
n+1
−y
n−1
=
y
n−1
−y
n−2
y
n
for n ≥ 1.
From this it follows that a solution of Eq(5.13) either approaches the positive equilibrium
y monotonically or it oscillates about y with semicycles of length one in such a way that
{y
2n
} and {y
2n+1
} converge monotonically one to zero and other to ∞.
5.4 Open Problems and Conjectures
By using the invariant (5.12), one can easily see that every solution of Lyness’ Eq(5.10)
is bounded from above and from below by positive numbers. To this day we have not
found a proof for the boundedness of solutions of Eq(5.10) without using essentially the
invariant.
Open Problem 5.4.1 Show that every solution of Lyness’ Eq(5.10) is bounded without
using the invariant (5.12).
5.4. Open Problems and Conjectures 73
Open Problem 5.4.2 Assume that f ∈ C[(0, ∞), (0, ∞)]. Obtain necessary and suffi-
cient conditions in terms of f so that every positive solution of the difference equation
x
n+1
=
f(x
n
)
x
n−1
, n = 0, 1, . . . (5.14)
is bounded.
What is it that makes Lyness’ equation possess an invariant? What type of difference
equations possess an invariant? Along these lines, we offer the following open problems
and conjectures.
Open Problem 5.4.3 Assume that f ∈ C[(0, ∞), (0, ∞)]. Obtain necessary and suffi-
cient conditions in terms of f so that the difference equation (5.14) possesses a (non-
trivial) invariant.
Conjecture 5.4.1 Assume that f ∈ C[(0, ∞), (0, ∞)] and that the difference equation
(5.14) possesses a unique positive equilibrium point x and a nontrivial invariant. Show
that the linearized equation of Eq(5.14) about x does not have both eigenvalues in |λ| < 1
and does not have both eigenvalues in |λ| > 1.
When we allow the initial conditions to be real numbers the equation
y
n+1
=
p + y
n
y
n−1
(5.15)
may not even be well defined for all n ≥ 0. To illustrate, we offer the following open
problems and conjectures.
Open Problem 5.4.4 (See [26].) Assume p ∈ (−∞, ∞). Determine the set G of all
initial conditions y
−1
, y
0
∈ (−∞, ∞) through which Eq(5.15) is well defined for all n ≥ 0.
Conjecture 5.4.2 Let G be the set of initial conditions defined in the Open problem
5.4.4. Show that the subset of G through which the solutions of Eq(5.15) are unbounded
is a set of measure zero.
74 Chapter 5. (2, 1)-Type Equations
Conjecture 5.4.3 Assume p ∈ (−∞, ∞). Show that the set of points (y
−1
, y
0
) ∈
(−∞, ∞) × (−∞, ∞) through which Eq(5.15) is not well defined for all n ≥ 0, is a
set of points without interior.
Open Problem 5.4.5 Determine all positive integers k with the property that every
positive solution of the equation
y
n+1
=
1 + . . . + y
n−k
y
n−k−1
, n = 0, 1, . . .
is periodic.
Conjecture 5.4.4 Show that every positive solution of the equation
y
n+1
=
y
n
y
n−1
+
y
n−4
y
n−3
, n = 0, 1, . . .
converges to a period-six solution.
Conjecture 5.4.5 (See [20]) Show that every positive solution of the equation
y
n+1
=
1
y
n
y
n−1
+
1
y
n−3
y
n−4
, n = 0, 1, . . .
converges to a period-three solution.
Conjecture 5.4.6 Show that the equation
y
n+1
=
1 + y
n−1
y
n
, n = 0, 1, . . .
has a nontrivial positive solution which decreases monotonically to the equilibrium of the
equation.
Open Problem 5.4.6 Find the set G of all initial points (y
−1
, y
0
) ∈ R × R through
which the equation
y
n+1
=
1 + y
n−1
y
n
is well defined for all n ≥ 0. Determine the periodic character and the asymptotic
behavior of all solutions with (y
−1
, y
0
) ∈ G.
5.4. Open Problems and Conjectures 75
Conjecture 5.4.7 (May’s Host Parasitoid Model; see [59]) Assume α > 1. Show
that every positive solution of May’s Host Parasitoid Model
x
n+1
=
αx
2
n
(1 + x
n
)x
n−1
, n = 0, 1, . . .
is bounded.
Conjecture 5.4.8 (The Gingerbreadman Map; see [59]) Let A ∈ (0, ∞). Show that
every positive solution of the equation
x
n+1
=
max{x
2
n
, A}
x
n
x
n−1
, n = 0, 1, . . . (5.16)
is bounded.
Note that the change of variables
x
n
=







A
1+y
n
2
if A > 1
e
y
n
2
if A = 1
A
−1+y
n
2
if A < 1
reduces Eq(5.16) to
y
n+1
= |y
n
| −y
n−1
+ δ, n = 0, 1, . . . (5.17)
where
δ =





1 if A < 1
0 if A = 1
−1 if A > 1.
When δ = 1, Eq(5.17) is called the Gingerbreadman Map and was investigated
by Devaney [17].
Chapter 6
(2, 2)-Type Equations
6.1 Introduction
Eq(1) contains the following nine equations of the (2, 2)-type:
x
n+1
=
α + βx
n
A + Bx
n
, n = 0, 1, . . . (6.1)
x
n+1
=
α + βx
n
A + Cx
n−1
, n = 0, 1, . . . (6.2)
x
n+1
=
α + βx
n
Bx
n
+ Cx
n−1
, n = 0, 1, . . . (6.3)
x
n+1
=
α + γx
n−1
A + Bx
n
, n = 0, 1, . . . (6.4)
x
n+1
=
α + γx
n−1
A + Cx
n−1
, n = 0, 1, . . . (6.5)
x
n+1
=
α + γx
n−1
Bx
n
+ Cx
n−1
, n = 0, 1, . . . (6.6)
77
78 Chapter 6. (2, 2)-Type Equations
x
n+1
=
βx
n
+ γx
n−1
A + Bx
n
, n = 0, 1, . . . (6.7)
x
n+1
=
βx
n
+ γx
n−1
A + Cx
n−1
, n = 0, 1, . . . (6.8)
x
n+1
=
βx
n
+ γx
n−1
Bx
n
+ Cx
n−1
, n = 0, 1, . . . . (6.9)
Please recall our classification convention in which all parameters that appear in these
equations are positive, the initial conditions are nonnegative, and the denominators are
always positive.
Eq(6.5) is essentially similar to Eq(6.1). In fact if {x
n
}

n=−1
is a positive solution of
Eq(6.5) then {x
2n
}

n=0
and {x
2n−1
}

n=0
are both solutions of the Riccati equations
z
n+1
=
α + γz
n
A + Cz
n
, n = 0, 1, . . .
with
z
n
= x
2n
for n ≥ 0
and
z
n+1
=
α + γz
n
A + Cz
n
, n = −1, 0, . . .
with
z
n
= x
2n+1
for n ≥ −1,
respectively.
Therefore we will omit any further discussion of Eq(6.5).
6.2 The Case γ = C = 0 : x
n+1
=
α+βx
n
A+Bx
n
- Riccati
Equation
This is the (2, 2)-type Eq(6.1) which is in fact a Riccati equation. See Section 1.6.
To avoid a degenerate situation we will assume that
αB −βA = 0.
6.3. The Case γ = B = 0 : x
n+1
=
α+βx
n
A+Cx
n−1
79
The Riccati number associated with this equation is
R =
βA −αB
(β + A)
2
and clearly
R <
1
4
.
Now the following result is a consequence of Theorem 1.6.2:
Theorem 6.2.1 The positive equilibrium of Eq(6.1) is globally asymptotically stable.
6.3 The Case γ = B = 0 : x
n+1
=
α+βx
n
A+Cx
n−1
This is the (2, 2)-type Eq(6.2) which by the change of variables
x
n
=
A
C
y
n
reduces to the equation
y
n+1
=
p + qy
n
1 + y
n−1
, n = 0, 1, . . . (6.10)
where
p =
αC
A
2
and q =
β
A
.
(Eq(6.10) was investigated in [42] and [43]. See also [36].)
Eq(6.10) has the unique positive equilibrium y given by
y =
q −1 +

(q −1)
2
+ 4p
2
.
By applying the linearized stability Theorem 1.1.1 we obtain the following result.
Lemma 6.3.1 The equilibrium y of Eq(6.10) is locally asymptotically stable for all val-
ues of the parameters p and q.
Eq(6.10) is a very simple looking equation for which it has long been conjectured
that its equilibrium is globally asymptotically stable. To this day, the conjecture has
not been proven or refuted.
The main results known about Eq(6.10) are the following:
Theorem 6.3.1 Every solution of Eq(6.10) is bounded from above and from below by
positive constants.
80 Chapter 6. (2, 2)-Type Equations
Proof. Let {y
n
} be a solution of Eq(6.10). Clearly, if the solution is bounded from
above by a constant M, then
y
n+1

p
1 + M
and so it is also bounded from below. Now assume for the sake of contradiction that the
solution is not bounded from above. Then there exists a subsequence {y
1+n
k
}

k=0
such
that
lim
k→∞
n
k
= ∞, lim
k→∞
y
1+n
k
= ∞, and y
1+n
k
= max{y
n
: n ≤ n
k
} for k ≥ 0.
From (6.10) we see that
y
n+1
< qy
n
+ p for n ≥ 0
and so
lim
k→∞
y
n
k
= lim
k→∞
y
n
k
−1
= ∞.
Hence, for sufficiently large k,
0 ≤ y
1+n
k
−y
n
k
=
p + [(q −1) −y
n
k
−1
]y
n
k
1 + y
n
k
−1
< 0
which is a contradiction and the proof is complete. 2
The oscillatory character of solutions for Eq(6.10) is described by the following result:
Theorem 6.3.2 Let {y
n
}

n=−1
be a nontrivial solution of Eq(6.10). Then the following
statements are true:
(i) Every semicycle, except perhaps for the first one, has at least two terms.
(ii) The extreme in each semicycle occurs at either the first term or the second. Fur-
thermore after the first, the remaining terms in a positive semicycle are strictly
decreasing and in a negative semicycle are strictly increasing.
(iii) In any two consecutive semicycles, their extrema cannot be consecutive terms.
(iv) Assume
q ≤ p.
Then, except possibly for the first semicycle, every semicycle contains two or three
terms. Furthermore, in every semicycle, there is at most one term which follows
the extreme.
Proof. We present the proofs for positive semicycles only. The proofs for negative
semicycles are similar and will be omitted.
6.3. The Case γ = B = 0 : x
n+1
=
α+βx
n
A+Cx
n−1
81
(i) Assume that for some N ≥ 0,
y
N−1
< y and y
N
≥ y.
Then
y
N+1
=
p + qy
N
1 + y
N−1
>
p + qy
1 + y
= y.
(ii) Assume that for some N ≥ 0, the first two terms in a positive semicycle are y
N
and y
N+1
. Then
y
N
≥ y, y
N+1
> y
and
y
N+2
y
N+1
=
1
y
N+1
p + qy
N+1
1 + y
N
=
p
y
N+1
+ q
1 + y
N
<
p
y
+ q
1 + y
= 1.
(iii) We consider the case where a negative semicycle is followed by a positive one. The
opposite case is similar and will be omitted. So assume that for some N ≥ 0 the
last two terms in a negative semicycle are y
N−1
and y
N
with
y
N−1
≥ y
N
.
Then
y
N+1
=
p + qy
N
1 + y
N−1
<
p + qy
N+1
1 + y
N
= y
N+2
.
(iv) Assume that for some N ≥ 0, the terms y
N
, y
N+1
, and y
N+2
are all in a positive
semicycle. Then
y
N+3
=
p + qy
N+2
1 + y
N+1
<
p + qy
N+1
1 + y
N+1
<
p + qy
1 + y
= y.
2
Theorem 6.3.3 The positive equilibrium of Eq(6.10) is globally asymptotically stable if
one of the following two conditions holds:
(i)
q < 1;
(ii)
q ≥ 1
and
either p ≤ q or q < p ≤ 2(q + 1).
82 Chapter 6. (2, 2)-Type Equations
Proof.
(i) In view of Lemma 6.3.1, it remains to show that every solution {y
n
}

n=−1
of Eq(6.10)
tends to y as n →∞. To this end, let
i = liminf
n→∞
y
n
and s = limsup
n→∞
y
n
which by Theorem 6.3.1 exist and are positive numbers. Then Eq(6.10) yields
i ≥
p + qi
1 + s
and s ≤
p + qs
1 + i
.
Hence
p + (q −1)i ≤ is ≤ p + (q −1)s
and so because q < 1,
i ≥ s
from which the result follows.
(ii) The proof when
p < q
follows by applying Theorem 1.4.2. This result was first established in [43].
For the proof when
1 ≤ q ≤ p ≤ 2(q + 1)
see ([42], Theorem 3.4.3 (f), p. 71).
2
6.4 The Case γ = A = 0 : x
n+1
=
α+βx
n
Bx
n
+Cx
n−1
This is the (2, 2)-type Eq(6.3) which by the change of variables
x
n
=
β
B
y
n
reduces to the difference equation
y
n+1
=
y
n
+ p
y
n
+ qy
n−1
, n = 0, 1, . . . (6.11)
where
p =
αB
β
2
and q =
C
B
.
Eq(6.11) was investigated in [15].
The following result is a consequence of the linearized stability Theorem 1.1.1.
6.4. The Case γ = A = 0 : x
n+1
=
α+βx
n
Bx
n
+Cx
n−1
83
Theorem 6.4.1 The unique positive equilibrium point
y =
1 +

1 + 4p(q + 1)
2(q + 1)
of Eq(6.11) is locally asymptotically stable for all values of the parameters p and q.
6.4.1 Invariant Intervals
The following result establishes that the interval I with end points 1 and
p
q
is an invari-
ant interval for Eq(6.11) in the sense that if, for some k ≥ 0, two consecutive values
y
k−1
and y
k
of a solution lie in I, then y
m
∈ I for all m > k.
Lemma 6.4.1 Let {y
n
}

n=−1
be a solution of Eq(6.11). Then the following statements
are true:
(a) Assume p < q and suppose that for some k ≥ 0,
y
k−1
, y
k

¸
p
q
, 1
¸
.
Then
y
n

¸
p
q
, 1
¸
for all n > k.
(b) Assume p > q and suppose that for some k ≥ 0,
y
k−1
, y
k

¸
1,
p
q
¸
.
Then
y
n

¸
1,
p
q
¸
for all n > k.
Proof.
(a) The basic ingredient behind the proof is the fact that when u, v ∈ [
p
q
, ∞), the
function
f(u, v) =
u + p
u + qv
is increasing in u and decreasing in v. Indeed,
y
k+1
=
y
k
+ p
y
k
+ qy
k−1

y
k
+ p
y
k
+ q
p
q
= 1,
y
k+1
=
y
k
+ p
y
k
+ qy
k−1

p
q
+ p
p
q
+ q
>
p
q
and the proof follows by induction.
(b) The proof is similar and will be omitted. 2
84 Chapter 6. (2, 2)-Type Equations
6.4.2 Semicycle Analysis
Let {y
n
}

n=−1
be a solution of Eq(6.11). Then observe that the following identities are
true:
y
n+1
−1 = q
p
q
−y
n−1
y
n
+ qy
n−1
for n ≥ 0, (6.12)
y
n+1

p
q
=
(q −p) + pq(1 −y
n−1
)
q(y
n
+ qy
n−1
)
for n ≥ 0, (6.13)
and
y
n
−y
n+4
=
(y
n
−1)[qy
n
y
n+3
+ y
n+1
y
n+3
] + qy
n+1
(y
n

p
q
)
q(p + y
n+1
) + y
n+3
(y
n+1
+ qy
n
)
for n ≥ 0. (6.14)
First we will analyze the semicycles of the solution {y
n
}

n=−1
under the assumption
that
p > q. (6.15)
The following result is a direct consequence of (6.12)-(6.15):
Lemma 6.4.2 Assume that (6.15) holds and let {y
n
}

n=−1
be a solution of Eq(6.11).
Then the following statements are true:
(i) If for some N ≥ 0, y
N−1
<
p
q
, then y
N+1
> 1;
(ii) If for some N ≥ 0, y
N−1
=
p
q
, then y
N+1
= 1;
(iii) If for some N ≥ 0, y
N−1
>
p
q
, then y
N+1
< 1;
(iv) If for some N ≥ 0, y
N−1
≥ 1, then y
N+1
<
p
q
;
(v) If for some N ≥ 0, y
N−1
≤ 1, then y
N+1
> 1;
(vi) If for some N ≥ 0, y
N

p
q
, then y
N+4
< y
N
;
(vii) If for some N ≥ 0, y
N
≤ 1, then y
N+4
> y
N
;
(viii) If for some N ≥ 0, 1 ≤ y
N−1

p
q
, then 1 ≤ y
N+1

p
q
;
(ix) If for some N ≥ 0, 1 ≤ y
N−1
, y
N

p
q
, then y
n
∈ [1,
p
q
] for n ≥ N. That is [1,
p
q
] is
an invariant interval for Eq(6.11);
(x)
1 < y <
p
q
.
6.4. The Case γ = A = 0 : x
n+1
=
α+βx
n
Bx
n
+Cx
n−1
85
The next result, which is a consequence of Theorem 1.7.2, states that when (6.15)
holds, every nontrivial and oscillatory solution of Eq(6.11), which lies in the interval
[1,
p
q
], oscillates about the equilibrium y with semicycles of length one or two.
Theorem 6.4.2 Assume that Eq(6.11) holds. Then every nontrivial and oscillatory
solution of Eq(6.11) which lies in the interval [1,
p
q
], oscillates about y with semicycles
of length one or two.
Next we will analyze the semicycles of the solutions {y
n
}

n=−1
under the assumption
that
q = p. (6.16)
In this case, Eq(6.11) reduces to
y
n+1
=
y
n
+ p
y
n
+ py
n−1
, n = 0, 1, . . . (6.17)
with the unique equilibrium point
y = 1.
Also identities (6.12)-(6.14) reduce to
y
n+1
−1 = p
1 −y
n−1
y
n
+ py
n−1
for n ≥ 0 (6.18)
and
y
n
−y
n+4
= (y
n
−1)
py
n+1
+ py
n
y
n+3
+ y
n+1
y
n+3
p(p + y
n+1
) + y
n+3
(y
n+1
+ py
n
)
for n ≥ 0. (6.19)
and so the following results follow immediately:
Lemma 6.4.3 Let {y
n
}

n=−1
be a solution of Eq(6.17). Then the following statements
are true:
(i) If for some N ≥ 0, y
N−1
< 1, then y
N+1
> 1;
(ii) If for some N ≥ 0, y
N−1
= 1, then y
N+1
= 1;
(iii) If for some N ≥ 0, y
N−1
> 1, then y
N+1
< 1;
(iv) If for some N ≥ 0, y
N
< 1, then y
N
< y
N+4
< 1;
(v) If for some N ≥ 0, y
N
> 1, then y
N
> y
N+4
> 1.
Corollary 6.4.1 Let {y
n
}

n=−1
be a nontrivial solution of Eq(6.17). Then {y
n
}

n=−1
oscillates about the equilibrium 1 and, except possibly for the first semicycle, the following
are true:
86 Chapter 6. (2, 2)-Type Equations
(i) If y
−1
= 1 or y
0
= 1, then the positive semicycle has length three and the negative
semicycle has length one.
(ii) If (1 −y
−1
)(1 −y
0
) = 0, then every semicycle has length two.
Finally we will analyze the semicycles of the solutions {y
n
}

n=−1
under the assumption
that
p < q. (6.20)
The following result is a direct consequence of (6.12)-(6.14) and (6.20).
Lemma 6.4.4 Assume that (6.20) holds and let {y
n
}

n=−1
be a solution of Eq(6.11).
Then the following statements are true:
(i) If for some N ≥ 0, y
N−1
<
p
q
, then y
N+1
> 1;
(ii) If for some N ≥ 0, y
N−1
=
p
q
, then y
N+1
= 1;
(iii) If for some N ≥ 0, y
N−1
>
p
q
, then y
N+1
< 1;
(iv) If for some N ≥ 0, y
N−1
≤ 1, then y
N+1
>
p
q
;
(v) If for some N ≥ 0, y
N−1

p
q
, then y
N+1
>
p
q
;
(vi) If for some N ≥ 0, y
N
≥ 1, then y
N+4
< y
N
;
(vii) If for some N ≥ 0, y
N

p
q
, then y
N+4
> y
N
;
(viii) If for some N ≥ 0,
p
q
≤ y
N−1
≤ 1, then
p
q
≤ y
N+1
≤ 1;
(ix) If for some N ≥ 0,
p
q
≤ y
N−1
, y
N
≤ 1, then y
n
∈ [
p
q
, 1] for n ≥ N. That is, [
p
q
, 1] is
an invariant interval for Eq(6.11);
(x)
p
q
< y < 1.
The next result, which is a consequence of Theorem 1.7.4, states that when (6.20)
holds, every nontrivial and oscillatory solution of Eq(6.11) which lies in the interval
[
p
q
, 1], after the first semicycle, oscillates with semicycles of length at least two.
Theorem 6.4.3 Assume that (6.20) holds. Then every nontrivial and oscillatory solu-
tion of Eq(6.11) which lies in the interval [
p
q
, 1], after the first semicycle, oscillates with
semicycles of length at least two.
6.4. The Case γ = A = 0 : x
n+1
=
α+βx
n
Bx
n
+Cx
n−1
87
How do solutions which do not eventually lie in the invariant interval behave?
First assume that
p > q
and let {y
n
}

n=−1
be a solution which does not eventually lie in the interval I = [1,
p
q
].
Then one can see that the solution oscillates relative to the interval [1,
p
q
], essentially in
one of the following two ways:
(i) Two consecutive terms in (
p
q
, ∞) are followed by two consecutive terms in (0, 1),
are followed by two consecutive terms in (
p
q
, ∞), and so on. The solution never visits
the interval (1,
p
q
).
(ii) There exists exactly one term in (
p
q
, ∞), which is followed by exactly one term in
(1,
p
q
), which is followed by exactly one term in (0, 1), which is followed by exactly one
term in (1,
p
q
), which is followed by exactly one term in (
p
q
, ∞), and so on. The solution
visits consecutively the intervals
. . . ,

p
q
, ∞

,

1,
p
q

, (0, 1),

1,
p
q

,

p
q
, ∞

, . . .
in this order with one term per interval.
The situation is essentially the same relative to the interval [
p
q
, 1] when
p < q.
6.4.3 Global Stability and Boundedness
Our goal in this section is to establish the following result:
Theorem 6.4.4 (a) The equilibrium y of Eq(6.11) is globally asymptotically stable
when
q ≤ 1 + 4p. (6.21)
(b) Assume that
q > 1 + 4p.
Then every solution of Eq(6.11) lies eventually in the interval

p
q
, 1

.
Proof.
(a) We have already shown that the equilibrium y is locally asymptotically stable so
it remains to be shown that y is a global attractor of every solution {y
n
}

n=−1
of
Eq(6.11).
Case 1 Assume p = q. It follows from (6.19) that each of the four subsequences
{y
4n+i
}

n=0
for i = 1, 2, 3, 4
88 Chapter 6. (2, 2)-Type Equations
is either identically equal to one or else it is strictly monotonically convergent. Set
L
i
= lim
n→∞
y
4n+i
for i = 1, 2, 3, 4.
Thus clearly
. . . , L
1
, L
2
, L
3
, L
4
, . . . (6.22)
is a periodic solution of Eq(6.11) with period four. By applying (6.19) to the
solution (6.22) we see that
L
i
= 1 for i = 1, 2, 3, 4
and so
lim
n→∞
y
n
= 1.
Case 2 Assume p = q. First, we assume that p > q.
Recall from Lemma 6.4.2 that [1,
p
q
] is an invariant interval. In this interval the
function
f(u, v) =
u + p
u + qv
is decreasing in both variables and so it follows by applying Theorem 1.4.7 that
every solution of Eq(6.11) with two consecutive values in [1,
p
q
] converges to y. By
using an argument similar to that in the case p = q, we can see that if a solution is
not eventually in [1,
p
q
], it converges to a periodic solution with period-four which
is not the equilibrium. By applying (6.14) to this period four solution we obtain a
contradiction and so every solution of Eq(6.11) must lie eventually in [1,
p
q
].
Hence for every solution {y
n
}

n=−1
of Eq(6.11)
lim
n→∞
y
n
= y
and the proof is complete.
Now, assume that p < q.
As in Case 2 we can see that every solution of Eq(6.11) must eventually lie in the
interval [
p
q
, 1]. In this interval the function
f(u, v) =
u + p
u + qv
is increasing in u and decreasing in v. Under the assumption that q ≤ 1 + 4p,
Theorem 1.4.5 applies and so every solution of Eq(6.11) converges to y.
(b) The proof follows from the discussion in part (a).
2
6.5. The Case β = C = 0 : x
n+1
=
α+γx
n−1
A+Bx
n
89
6.5 The Case β = C = 0 : x
n+1
=
α+γx
n−1
A+Bx
n
This is the (2, 2)-type Eq(6.4) which by the change of variables
x
n
=
A
B
y
n
reduces to the difference equation
y
n+1
=
p + qy
n−1
1 + y
n
, n = 0, 1, . . . (6.23)
where
p =
αB
A
2
and q =
γ
A
.
(Eq(6.23) was investigated in [28].)
The only equilibrium point of Eq(6.23) is
y =
q −1 +

(q −1)
2
+ 4p
2
and by applying the linearized stability Theorem 1.1.1 we obtain the following result:
Theorem 6.5.1 The equilibrium y of Eq(6.23) is locally asymptotically stable when
q < 1,
and is an unstable saddle point when
q > 1.
Concerning prime period-two solutions, the following result is a consequence of the
results in Section 2.7.
Theorem 6.5.2 (a) Eq(6.23) has a prime period-two solution
. . . , φ, ψ, φ, ψ, . . . (6.24)
if and only if
q = 1. (6.25)
(b) Assume (6.25) holds. Then the values φ and ψ of all prime period-two solutions
(6.24) are given by
{φ, ψ ∈ (0, ∞) : φψ = p}.
(c) Assume that (6.25) holds. Then every solution of Eq(6.23) converges to a period
two solution.
90 Chapter 6. (2, 2)-Type Equations
6.5.1 Semicycle Analysis
Our results on the semicycles of solutions for Eq(6.23) are as follows:
Theorem 6.5.3 Let {y
n
} be a nontrivial solution of Eq(6.23) and let y denote the
unique positive equilibrium of Eq(6.23). Then the following statements are true:
(a) After the first semicycle, an oscillatory solution {y
n
} of Eq(6.23) oscillates about
the equilibrium y with semicycles of length one.
(b) Assume p ≥ q. Then every nonoscillatory solution of Eq(6.23) converges mono-
tonically to the equilibrium y.
Proof. (a) This follows immediately from Theorem 1.7.1
(b) Let {y
n
} be a nonoscillatory solution of Eq(6.23). We will assume that there exists
a positive integer K such that y
n−1
≥ y for every n ≥ K. The case where y
n−1
< y for
n ≥ K is similar and will be omitted. It is sufficient to show that {y
n
} is a decreasing
sequence for n ≥ K. So assume for the sake of contradiction that for some n
0
≥ K,
y
n
0
> y
n
0
−1
.
Clearly, the condition p ≥ q implies that the function
f(x) =
p + qx
1 + x
is decreasing. Then
y
n
0
+1
=
p + qy
n
0
−1
1 + y
n
0
= f(y
n
0
, y
n
0
−1
) < f(y
n
0
, y
n
0
) < f(y, y) = y,
which is impossible. 2
6.5.2 The Case q < 1
The main result in this section is the following:
Theorem 6.5.4 Assume that
q < 1.
Then the positive equilibrium of Eq(6.23) is globally asymptotically stable.
Proof. Clearly
y
n+1
=
p + qy
n−1
1 + y
n
< p + qy
n−1
, n ≥ 0. (6.26)
Set
M =
p
1 −q
+ ,
6.5. The Case β = C = 0 : x
n+1
=
α+γx
n−1
A+Bx
n
91
where is some positive number.
From Eq(6.26) and Eq(6.23), it now follows that every solution of Eq(6.23) lies
eventually in the interval [0, M].
Set
f(x, y) =
p + qy
1 + x
, a = 0, and b =
p
1 −q
+ .
Then clearly,
a ≤ f(x, y) ≤ b,
for all (x, y) ∈ [a, b] ×[a, b]. By applying Theorem 1.4.6 it follows that y is a global at-
tractor of all solutions of Eq(6.23). The local stability of y was established in Theorem
6.5.1. The proof is complete. 2
6.5.3 The Case q > 1
In this case we will show that there exist solutions that have two subsequences, one of
which is monotonically approaching 0 and the other is monotonically approaching ∞.
In fact this is true for every solution {y
n
}

n=−1
whose initial conditions are such that the
following inequalities are true:
y
−1
< q −1 and y
0
> q −1 +
p
q −1
.
To this end observe that
y
1
=
p + qy
−1
1 + y
0
<
p + q(q −1)
1 + y
0
< q −1
and
y
2
=
p + qy
0
1 + y
1
>
p + qy
0
q
= y
0
+
p
q
and by induction,
y
2n−1
< q −1 and y
2n
> y
0
+ n
p
q
for n ≥ 1.
Hence
lim
n→∞
y
2n
= ∞
and
lim
n→∞
y
2n+1
= lim
n→∞
p + qy
2n−1
1 + y
2n
= 0
and the proof is complete.
92 Chapter 6. (2, 2)-Type Equations
6.6 The Case β = A = 0 : x
n+1
=
α+γx
n−1
Bx
n
+Cx
n−1
This is the (2, 2)-type Eq(6.6) which by the change of variables
x
n
=
γ
C
y
n
reduces to the difference equation
y
n+1
=
p + y
n−1
qy
n
+ y
n−1
, n = 0, 1, . . . (6.27)
where
p =
αC
γ
2
and q =
B
C
.
(Eq(6.27) was investigated in [45].)
Eq(6.27) has the unique positive equilibrium
y =
1 +

1 + 4p(1 + q)
2(1 + q)
.
The linearized equation associated with Eq(6.27) about y is
z
n+1
+
q
1 + q
z
n

qy −p
(y + p)(1 + q)
z
n−1
= 0, n = 0, 1, . . . . (6.28)
By employing Theorem 1.1.1 we see that y is locally asymptotically stable when
(q −1)y < 2p (6.29)
and unstable (a saddle point) when
(q −1)y > 2p. (6.30)
Clearly the equilibrium y is the positive solution of the quadratic equation
(1 + q)y
2
−y −p = 0.
If we now set
F(u) = (1 + q)u
2
−u −p,
it is easy to see that (6.29) is satisfied if and only if either
q ≤ 1
or
q > 1 and F

2p
q −1

> 0.
6.6. The Case β = A = 0 : x
n+1
=
α+γx
n−1
Bx
n
+Cx
n−1
93
Similarly (6.30) is satisfied if and only if
q > 1 and F

2p
q −1

< 0.
The following result is now a consequence of the above discussion and some simple
calculations:
Theorem 6.6.1 The equilibrium y of Eq(6.27) is locally asymptotically stable when
q < 1 + 4p (6.31)
and unstable, and more precisely a saddle point equilibrium, when
q > 1 + 4p. (6.32)
6.6.1 Existence and Local Stability of Period-Two Cycles
Let
. . . , φ, ψ, φ, ψ, . . .
be a period-two cycle of Eq(6.27). Then
φ =
p + φ
qψ + φ
and ψ =
p + ψ
qφ + ψ
.
It now follows after some calculations that the following result is true:
Theorem 6.6.2 Eq(6.27) has a prime period-two solution
. . . , φ, ψ, φ, ψ, . . .
if and only if
q > 1 + 4p.
Furthermore when (6.32) holds, the period-two solution is “unique” and the values of φ
and ψ are the positive roots of the quadratic equation
t
2
−t +
p
q −1
= 0.
94 Chapter 6. (2, 2)-Type Equations
To investigate the local stability of the two cycle
. . . , φ, ψ, φ, ψ, . . .
we set
u
n
= y
n−1
and v
n
= y
n
, for n = 0, 1, . . .
and write Eq(6.27) in the equivalent form:
u
n+1
= v
n
v
n+1
=
p+u
n
qv
n
+u
n
, n = 0, 1, . . . .
Let T be the function on (0, ∞) ×(0, ∞) defined by:
T

u
v

=

v
p+u
qv+u

.
Then

φ
ψ

is a fixed point of T
2
, the second iterate of T. By a simple calculation we find that
T
2

u
v

=

g(u, v)
h(u, v)

where
g(u, v) =
p + u
qv + u
and h(u, v) =
p + v
qg(u, v) + v
.
Clearly the two cycle is locally asymptotically stable when the eigenvalues of the
Jacobian matrix J
T
2, evaluated at

φ
ψ

lie inside the unit disk.
We have
J
T
2

φ
ψ

=

¸
¸
∂g
∂u
(φ, ψ)
∂g
∂v
(φ, ψ)
∂h
∂u
(φ, ψ)
∂h
∂v
(φ, ψ)
¸

where,
∂g
∂u
(φ, ψ) =
qψ −p
(qψ + φ)
2
,
∂g
∂v
(φ, ψ) = −
(p + φ)q
(qψ + φ)
2
,
6.6. The Case β = A = 0 : x
n+1
=
α+γx
n−1
Bx
n
+Cx
n−1
95
∂h
∂u
(φ, ψ) = −
q(p + ψ)(qψ −p)
(qφ + ψ)
2
(qψ + φ)
2
,
∂h
∂v
(φ, ψ) =
q
2
(p + φ)(p + ψ)
(qφ + ψ)
2
(qψ + φ)
2
+
qφ −p
(qφ + ψ)
2
.
Set
S =
∂g
∂u
(φ, ψ) +
∂h
∂v
(φ, ψ)
and
D =
∂g
∂u
(φ, ψ)
∂h
∂v
(φ, ψ) −
∂g
∂v
(φ, ψ)
∂h
∂u
(φ, ψ).
Then it follows from Theorem 1.1.1 that both eigenvalues of J
T
2

φ
ψ

lie inside the
unit disk |λ| < 1, if and only if
|S| < 1 +D < 2. (6.33)
Inequality (6.33) is equivalent to the following three inequalities:
S < 1 +D (6.34)
−1 −D < S (6.35)
D < 1. (6.36)
First we will establish Inequality (6.34).
To this end observe that (6.34) is equivalent to
(qψ −p)(qφ + ψ)
2
+ (qφ −p)(qψ + φ)
2
+ q
2
(p + ψ)(p + φ) −(qψ −p)(qφ −p)
< (qψ + φ)
2
(qφ + ψ)
2
which is true if and only if
p
q −1
(q
3
+ 2q
2
−4qp + 2q
2
p + 2p −3q) + q −p + q
2
p
2
+ qp −p
2
< (qψ + φ)
2
(qφ + ψ)
2
which is true if and only if
−q
3
p + 7q
2
p −8qp
2
+ 4q
2
p
2
+ 4p
2
−7qp + 2q
2
−q + p −q
3
< 0
which is true if and only if
−(p + 1)(q −1)(q −
p
p + 1
)[q −(1 + 4p)] < 0
which is true because q > 1 + 4p.
96 Chapter 6. (2, 2)-Type Equations
Next we will establish Inequality (6.35). Observe that (6.35) is equivalent to
−(qψ + φ)
2
(qφ + ψ)
2
< (qψ −p)(qφ + ψ)
2
+ (qφ −p)
2
(qψ + φ)
2
+ q
2
(p + ψ)(p + φ) + (qψ −p)(qφ −p)
which is true if and only if
−(qψ + φ)
2
(qφ + ψ)
2
<
p
q −1
(q
3
+ 4q
2
−4qp + 2q
2
p + 2p −3q) −p + q
2
p
2
which is true if and only if
q
2
+ 2pq + 2p
2
q
2
−p
2
+ qp
2
−q
3
−3q
3
p −2q
3
p
2
−p < 0
which is true if and only if
p
2
(−2q
3
+ 2q
2
+ q −1) + p(−3q
3
+ 2q −1) −q
3
+ q
2
< 0
which is true because q > 1.
Finally, we establish Inequality (6.36). Observe that (6.36) is true if and only if
(qψ −p)(qφ −p) < (qψ + φ)
2
(qφ + ψ)
2
which is true if and only if
p
q −1
[q
2
−(q −1)(q −p)] < (pq −p + q)
2
which is true if and only if
q(pq −p + q)(pq −2p + q −1) > 0
which is true because q > 1 + 4p.
6.6.2 Invariant Intervals
Here we show that the interval I with end points one and
p
q
is an invariant interval for
Eq(6.27). More precisely we show that the values of any solution {y
n
}

n=−1
of Eq(6.27)
either remain forever outside the interval I or the solution is eventually trapped into I.
Theorem 6.6.3 Let {y
n
}

n=−1
be a solution of Eq(6.27). Then the following statements
are true.
(a) Assume that p ≤ q and that for some N ≥ 0,
p
q
≤ y
N
≤ 1.
Then
p
q
≤ y
n
≤ 1 for all n ≥ N.
6.6. The Case β = A = 0 : x
n+1
=
α+γx
n−1
Bx
n
+Cx
n−1
97
(b) Assume that p ≥ q and that for some N ≥ 0,
1 ≤ y
N

p
q
.
Then
1 ≤ y
n

p
q
for all n ≥ N.
Proof.
(a) Indeed we have
y
N+1
=
p + y
N−1
qy
N
+ y
N−1

p + y
N−1
p + y
N−1
= 1
and
y
N+1
=
p + y
N−1
qy
N
+ y
N−1
=
p + y
N−1
q
p
(py
N
+
p
q
y
N−1
)

p
q

p + y
N−1
p + y
N−1
=
p
q
and the proof follows by induction.
(b) Clearly,
y
N+1
=
p + y
N−1
qy
N
+ y
N−1

p + y
N−1
p + y
N−1
= 1
and
y
N+1
=
p + y
N−1
qy
N
+ y
N−1
=

p
q

p + y
N−1
py
N
+
p
q
y
N−1

p
q

p + y
N−1
p + y
N−1
=
p
q
and the proof follows by induction.
2
6.6.3 Semicycle Analysis
Here we present a thorough semicycle analysis of the solutions of Eq(6.27) relative to
the equilibrium y and relative to the end points of the invariant interval of Eq(6.27).
Let {y
n
}

n=−1
be a solution of Eq(6.27). Then observe that the following identities
are true:
y
n+1
−1 = q
p
q
−y
n
qy
n
+ y
n−1
for n ≥ 0, (6.37)
y
n+1

p
q
=
qy
n−1
(1 −
p
q
) + pq(1 −y
n
)
q(qy
n
+ y
n−1
)
for n ≥ 0, (6.38)
y
n
−y
n+2
=
qy
n−1
(y
n

p
q
) + (y
n
−1)(y
n−1
y
n
+ qy
2
n
)
q(p + y
n−1
) + y
n
(qy
n
+ y
n−1
)
for n ≥ 0. (6.39)
98 Chapter 6. (2, 2)-Type Equations
When
p = q
that is for the difference equation
y
n+1
=
p + y
n−1
py
n
+ y
n−1
, n = 0, 1, . . . (6.40)
the above identities reduce to the following:
y
n+1
−1 = p
1 −y
n
py
n
+ y
n−1
for n ≥ 0 (6.41)
and
y
n
−y
n+2
= (y
n
−1)
py
n−1
+ y
n−1
y
n
+ py
2
n
p(p + y
n−1
) + y
n
(py
n
+ y
n−1
)
for n ≥ 0. (6.42)
The following three lemmas are now direct consequences of (6.37)-(6.42).
Lemma 6.6.1 Assume that
p > q
and let {y
n
}

n=−1
be a solution of Eq(6.27). Then the following statements are true:
(i) If for some N ≥ 0, y
N
<
p
q
, then y
N+1
> 1;
(ii) If for some N ≥ 0, y
N
=
p
q
, then y
N+1
= 1;
(iii) If for some N ≥ 0, y
N
>
p
q
, then y
N+1
< 1;
(iv) If for some N ≥ 0, y
N
≥ 1, then y
N+1
<
p
q
;
(v) If for some N ≥ 0, y
N

p
q
, then y
N+2
< y
N
;
(vi) If for some N ≥ 0, y
N
≤ 1, then y
N+2
> y
N
;
(vii) If for some N ≥ 0, y
N

p
q
, then y
n
∈ [1,
p
q
] for n ≥ N. In particular [1,
p
q
] is an
invariant interval for Eq(6.27);
(viii)
1 < y <
p
q
.
Lemma 6.6.2 Assume that
p = q
and let {y
n
}

n=−1
be a solution of Eq(6.27). Then the following statements are true:
(i) If for some N ≥ 0, y
N
< 1, then y
N+1
> 1;
6.6. The Case β = A = 0 : x
n+1
=
α+γx
n−1
Bx
n
+Cx
n−1
99
(ii) If for some N ≥ 0, y
N
= 1, then y
N+1
= 1;
(iii) If for some N ≥ 0, y
N
> 1, then y
N+1
< 1;
(iv) If for some N ≥ 0, y
N
< 1, then y
N
< y
N+2
< 1;
(v) If for some N ≥ 0, y
N
> 1, then y
N
> y
N+2
> 1.
Lemma 6.6.3 Assume that
p < q
and let {y
n
}

n=−1
be a solution of Eq(6.27). Then the following statements are true:
(i) If for some N ≥ 0, y
N
<
p
q
, then y
N+1
> 1;
(ii) If for some N ≥ 0, y
N
=
p
q
, then y
N+1
= 1;
(iii) If for some N ≥ 0, y
N
>
p
q
, then y
N+1
< 1;
(iv) If for some N ≥ 0, y
N
≤ 1, then y
N+1
>
p
q
;
(v) If for some N ≥ 0, y
N

p
q
, then y
N+1
≥ 1;
(vi) If for some N ≥ 0, y
N
≥ 1, then y
N+2
< y
N
;
(vii) If for some N ≥ 0, y
N

p
q
, then y
N+2
> y
N
;
(viii) If for some N ≥ 0, y
N
≤ 1, then y
n
∈ [
p
q
, 1] for n ≥ N. In particular [
p
q
, 1] is an
invariant interval for Eq(6.27);
(ix)
p
q
< y < 1.
Note that the function
f(u, v) =
p + v
qu + v
is always decreasing in u but in v it is decreasing when u <
p
q
and increasing when u >
p
q
.
The following result is now a consequence of Theorems 1.7.1 and 1.7.2 and Lemmas
6.6.1-6.6.3:
100 Chapter 6. (2, 2)-Type Equations
Theorem 6.6.4 Let {y
n
}

n=−1
be a solution of Eq(6.27). Let I be the closed interval
with end points 1 and
p
q
and let J and K be the intervals which are disjoint from I and
such that
I ∪ J ∪ K = (0, ∞).
Then either all the even terms of the solution lie in J and all odd terms lie in K, or
vice-versa, or for some N ≥ 0,
y
n
∈ I for n ≥ N. (6.43)
When (6.43) holds, except for the length of the first semicycle of the solution, if
p < q
the length is one, while if
p > q
the length is at most two.
6.6.4 Global Behavior of Solutions
Recall that Eq(6.27) has a unique equilibrium y which is locally stable when (6.31) holds.
When (6.32) holds, and only then, Eq(6.27) has a “unique” prime period-two solution
. . . , φ, ψ, φ, ψ, . . . (6.44)
with
φ + ψ = 1 and φψ =
p
q −1
. (6.45)
Also recall that a solution {y
n
}

n=−1
of Eq(6.27) either eventually enters the interval I
with end points 1 and
p
q
and remains there, or stays forever outside I.
Now when
q ≤ 1 + 4p
there are no period-two solutions and so the solutions must eventually enter and remain
in I and, in view of Theorems 1.4.6 and 1.4.7, must converge to y.
On the other hand when (6.32) holds, any solution which remains forever outside the
interval I = [
p
q
, 1] must converge to the two cycle (6.44)-(6.45). But this two cycle lies
inside the interval I. Hence, when (6.32) holds, every solution of Eq(6.27) eventually
enters and remains in the interval I.
The character of these solutions remains an open question.
The above observations are summarized in the following theorem:
Theorem 6.6.5 (a) Assume
q ≤ 1 + 4p. (6.46)
Then the equilibrium y of Eq(6.27) is global attractor.
6.7. The Case α = C = 0 : x
n+1
=
βx
n
+γx
n−1
A+Bx
n
101
(b) Assume
q > 1 + 4p.
Then every solution of Eq(6.27) eventually enters and remains in the interval [
p
q
, 1].
6.7 The Case α = C = 0 : x
n+1
=
βx
n
+γx
n−1
A+Bx
n
This is the (2, 2)-type equation Eq(6.7) which by the change of variables
x
n
=
A
B
y
n
reduces to the difference equation
y
n+1
=
py
n
+ qy
n−1
1 + y
n
, n = 0, 1, . . . (6.47)
where
p =
β
A
and q =
γ
A
.
(Eq(6.47) was investigated in [52].)
The equilibrium points of Eq(6.47) are the solutions of the equation
y =
py + qy
1 + y
.
So y = 0 is always an equilibrium point, and when
p + q > 1,
y = p + q −1 is the only positive equilibrium point of Eq(6.47).
The following result follows from Theorem 1.3.1 and the linearized stability Theorem
1.1.1:
Theorem 6.7.1 (a) Assume
p + q ≤ 1.
Then the zero equilibrium of Eq(6.47) is globally asymptotically stable.
(b) Assume
p + q > 1.
Then the zero equilibrium of Eq(6.47) is unstable. More precisely it is a saddle
point when
1 −p < q < 1 + p
and a repeller when
q > 1 + p.
102 Chapter 6. (2, 2)-Type Equations
(c) Assume
1 −p < q < 1 + p.
Then the positive equilibrium y = p + q − 1 of Eq(6.47) is locally asymptotically
stable.
(d) Assume
q > 1 + p.
Then the positive equilibrium y = p+q −1 of Eq(6.47) is an unstable saddle point.
Concerning prime period-two solutions it follows from Sections 2.5 that the following
result is true:
Theorem 6.7.2 Eq(6.47) has a prime period-two solution
. . . , φ, ψ, φ, ψ, . . .
if and only if
q = 1 + p. (6.48)
Furthermore when (6.48) holds, the values of φ and ψ of all prime period-two solutions
are given by:
{φ, ψ ∈ (p, ∞) : ψ =

φ −p
= 2p}.
6.7.1 Invariant Intervals
The main result here is the following theorem about Eq(6.47) with q = 1. The case
q = 1 (6.49)
is treated in Section 6.7.3.
Theorem 6.7.3 (a) Assume that
q < 1. (6.50)
Then every solution of Eq(6.47) eventually enters the interval (0,
p
q
]. Furthermore,
(0,
p
q
] is an invariant interval for Eq(6.47). That is, every solution of Eq(6.47)
with initial conditions in (0,
p
q
], remains in this interval.
(b) Assume that
q > 1. (6.51)
Then every solution of Eq(6.47) eventually enters the interval [
p
q
, ∞). Further-
more, [
p
q
, ∞) is an invariant interval for Eq(6.47).
6.7. The Case α = C = 0 : x
n+1
=
βx
n
+γx
n−1
A+Bx
n
103
The proof of the above theorem is an elementary consequence of the following lemma,
whose proof is straightforward and will be omitted.
Lemma 6.7.1 (a) Assume that (6.50) holds. Then for any k, m ∈ N,
y
k

p
q
=⇒y
k+2
≤ p <
p
q
and
y
k+2

p
q
m
=⇒y
k
>
p
q
m+1
> p.
(b) Assume that (6.51) holds. Then for any k, m ∈ N,
y
k

p
q
=⇒y
k+2
≥ p >
p
q
and
y
k+2

p
q
m
=⇒y
k
>
p
q
m+1
> p.
6.7.2 Semicycle Analysis of Solutions When q ≤ 1
Here we discuss the behavior of the semicycles of solutions of Eq(6.47) when
p + q > 1 and q ≤ 1.
Hereafter, y refers to the positive equilibrium p + q −1 of Eq(6.47).
Let f : [0, ∞) ×[0, ∞) →[0, ∞) denote the function
f(x, y) =
px + qy
1 + x
.
Then f satisfies the negative feedback condition; that is,
(x −x)(f(x, x) −x) < 0 for x ∈ (0, ∞) −{x}.
Consider the function g : [0, ∞) →[0, ∞) defined by
g(x) =
(p + q)x
1 + x
.
Observe that g(y) = y and that g increases on [0, ∞).
Theorem 6.7.4 Suppose
p + q > 1 and q ≤ 1.
Let {y
n
}

n=−1
be a positive solution of Eq(6.47). Then the following statements are true
for every n ≥ 0:
104 Chapter 6. (2, 2)-Type Equations
(a) If y ≤ y
n−1
, y
n
, then y ≤ y
n+1
≤ max{y
n−1
, y
n
}.
(b) If y
n−1
, y
n
≤ y, then min{y
n−1
, y
n
} ≤ y
n+1
< y.
(c) If y
n−1
< y ≤ y
n
, then y
n−1
< y
n+1
< y
n
.
(d) If y
n
< y ≤ y
n−1
, then y
n
< y
n+1
< y
n−1
.
(e) If y < y
k
<
y
2−q
, for every k ≥ −1, then {y
n
} decreases to the equilibrium y.
Proof.
(a) Suppose y ≤ y
n−1
, y
n
. Then
y
n+1
=
py
n
+ qy
n−1
1 + y
n

(p + q) max{y
n−1
, y
n
}
1 + y
= max{y
n−1
, y
n
}.
Since the function
px + qx
1 + x
,
is increasing for y <
p
q
, which is equivalent to q ≤ 1, we obtain
y
n+1
=
py
n
+ qy
n−1
1 + y
n

py
n
+ qy
1 + y
n
≥ g(y) = y.
Observe that all inequalities are strict if we assume that y
n−1
> y or y
n
> y.
(b) Suppose y
n−1
, y
n
≤ y. Then in view of the previous case
y
n+1
=
py
n
+ qy
n−1
1 + y
n

py
n
+ qy
1 + y
n
< g(y) = y
and
y
n+1
=
py
n
+ qy
n−1
1 + y
n

(p + q) min{y
n−1
, y
n
}
1 + y
= min{y
n−1
, y
n
}.
Observe that all inequalities are strict if we assume that y
n−1
< y or y
n
< y.
(c) Suppose y
n−1
< y ≤ y
n
. Then
y
n+1
=
py
n
+ qy
n−1
1 + y
n
<
(p + q)y
n
1 + y
n
=
(1 + y)y
n
1 + y
n
< y
n
.
Since q ≤ 1 and y
n−1
< y,
p −qy
n−1
> p −qy = (1 −q)(p + q) ≥ 0.
Thus
px+qy
n−1
1+x
increases in x and so
y
n+1
=
py
n
+ qy
n−1
1 + y
n
>
py
n−1
+ qy
n−1
1 + y
n−1
= g(y
n−1
) > y
n−1
.
The last inequality follows from the negative feedback property.
6.7. The Case α = C = 0 : x
n+1
=
βx
n
+γx
n−1
A+Bx
n
105
(d) Suppose y
n
< y ≤ y
n−1
. Then
y
n+1
=
py
n
+ qy
n−1
1 + y
n
>
py
n
+ qy
n
1 + y
n
= g(y
n
) > y
n
by the negative feedback property.
To see that y
n+1
< y
n−1
, we must consider two cases:
Case 1. Suppose p −qy
n−1
≥ 0. Then
px+qy
n−1
1+x
increases in x and so
y
n+1
=
py
n
+ qy
n−1
1 + y
n

py
n−1
+ qy
n−1
1 + y
n−1
< y
n−1
.
The last inequality follows from the negative feedback property.
Case 2. Suppose p −qy
n−1
< 0. Then
y
n+1
=
py
n
+ qy
n−1
1 + y
n
<
qy
n−1
(y
n
+ 1)
1 + y
n
≤ y
n−1
.
(e) Observe that
y
n+1
=
py
n
+ qy
n−1
1 + y
n
> y = p + q −1
and so
y
n−1
≥ qy
n−1
> y + (q −1)y
n
> y
n
from which the result follows.
2
6.7.3 Semicycle Analysis and Global Attractivity When q = 1
Here we discuss the behavior of the solutions of Eq(6.47) when
q = 1.
Here the positive equilibrium of Eq(6.47) is
y = p.
Theorem 6.7.5 Suppose that q = 1 and let y
−1
+ y
0
> 0. Then the equilibrium y of
Eq(6.47) is globally asymptotically stable. More precisely the following statements are
true.
(a) Assume y
−1
≥ p and y
0
≥ p. Then y
n
≥ p for all n > 0 and
lim
n→∞
y
n
= p. (6.52)
(b) Assume y
−1
≤ p and y
0
≤ p. Then y
n
≤ p for all n > 0 and (6.52) holds.
(c) Assume either y
−1
< p < y
0
or y
−1
> p > y
0
. Then {y
n
}

n=−1
oscillates about the
equilibrium p with semicycles of length one and (6.52) holds.
106 Chapter 6. (2, 2)-Type Equations
Proof. (a) The case where y
−1
= y
0
= p is trivial. We will assume that y
−1
≥ p and
y
0
> p. The case y
−1
> p and y
0
≥ p is similar and will be omitted. Then
y
1
=
py
0
+ y
−1
1 + y
0

py
0
+ p
1 + y
0
= p,
and
y
2
=
py
1
+ y
0
1 + y
1
>
py
1
+ p
1 + y
1
= p.
It follows by induction that y
n
≥ p for all n > 0. Furthermore, y
−1
≥ p implies that
py
0
+ y
−1
≤ y
−1
y
0
+ y
−1
and so
y
1
=
py
0
+ y
−1
1 + y
0

y
−1
y
0
+ y
−1
1 + y
0
= y
−1
.
Similarly, we can show that y
2
< y
0
and by induction
y
−1
≥ y
1
≥ y
3
≥ . . . ≥ p
and
y
0
> y
2
> y
4
> . . . > p.
Thus, there exist m ≥ p and M ≥ p such that
lim
k→∞
y
2k+1
= m and lim
k→∞
y
2k
= M.
As Eq(6.47) has no period-two solutions when q = 1, it follows that m = M = p and
the proof of part (a) is complete.
It is interesting to note that
y
−1
= p ⇒y
2k+1
= p for k ≥ 0 and y
0
= p ⇒y
2k
= p for k ≥ 0.
(b) The proof of (b) is similar to the proof of (a) and will be omitted.
(c) Assume y
−1
< p < y
0
. The proof when y
0
< p < y
−1
is similar and will be
omitted.
y
1
=
py
0
+ y
−1
1 + y
0
<
py
0
+ p
1 + y
0
= p,
and
y
2
=
py
1
+ y
0
1 + y
1
>
py
1
+ p
1 + y
1
= p.
It follows by induction that the solution oscillates about p with semicycles of length one.
Next, we claim that the subsequence {y
2k+1
}

k=−1
is increasing, and so convergent. In
fact, since y
−1
+ py
0
> y
−1
+ y
−1
y
0
then
y
1
=
py
0
+ y
−1
1 + y
0
> y
−1
6.7. The Case α = C = 0 : x
n+1
=
βx
n
+γx
n−1
A+Bx
n
107
and the claim follows by induction. In a similar way we can show that the subsequence
{y
2k
}

k=0
is decreasing, and so convergent. Thus, there exist m ≤ p and M ≥ p such
that
lim
k→∞
y
2k+1
= m and lim
k→∞
y
2k
= M
and as in (a), m = M. 2
6.7.4 Global Attractivity When q < 1
Here we consider the case where
1 −p < q < 1,
and we show that the equilibrium point y = p + q −1 of Eq(6.47) is global attractor of
all positive solutions of Eq(6.47).
Theorem 6.7.6 Assume
1 −p < q < 1.
Then every positive solution of Eq(6.47) converges to the positive equilibrium y.
Proof. Observe that the consistency condition y ≤
p
q
is equivalent to the condition
q ≤ 1. By Theorem 6.7.3 (a) we may assume that the initial conditions y
−1
and y
0
lie
in the interval I = (0,
p
q
]. Now clearly the function
f(x, y) =
px + qy
1 + x
satisfies the hypotheses of Theorem 1.4.5 in the interval I from which the result follows.
2
6.7.5 Long-term Behavior of Solutions When q > 1
Here we discuss the behavior of the solutions of Eq(6.47) when
q > 1
which is equivalent to
y >
p
q
.
In this case, every positive solution of Eq(6.47) lies eventually in the interval [p, ∞).
Concerning the long-term behavior of solutions of Eq(6.47) we may assume, without
loss of generality, that
y
−1
, y
0
∈ [p, ∞).
108 Chapter 6. (2, 2)-Type Equations
Then the change of variable
y
n
= u
n
+ p,
transforms Eq(6.47) to the equation
u
n+1
=
p(q −1) + qu
n−1
1 + p + u
n
(6.53)
where u
n
≥ 0 for n ≥ 0.
The character of solutions of Eq(6.53) was investigated in Section 6.5. By apply-
ing the results of Section 6.5 to Eq(6.53), we obtain the following theorems about the
solutions of Eq(6.47) when q > 1:
Theorem 6.7.7 After the first semicycle, an oscillatory solution of Eq(6.47) oscillates
about the positive equilibrium y with semicycles of length one.
Theorem 6.7.8 (a) Assume
q = 1 + p.
Then every solution of Eq(6.47) converges to a period-two solution.
(b) Assume
1 < q < 1 + p.
Then every positive solution of Eq(6.47) converges to the positive equilibrium of
Eq(6.47).
(c) Assume
q > 1 + p.
Then Eq(6.47) has unbounded solutions.
6.8. The Case α = B = 0 : x
n+1
=
βx
n
+γx
n−1
A+Cx
n−1
109
6.8 The Case α = B = 0 : x
n+1
=
βx
n
+γx
n−1
A+Cx
n−1
This is the (2, 2)-type Eq(6.8) which by the change of variables
x
n
=
γ
C
y
n
reduces to the difference equation
y
n+1
=
py
n
+ y
n−1
q + y
n−1
, n = 0, 1, . . . (6.54)
where
p =
β
γ
and q =
A
γ
.
(Eq(6.54) was investigated in [53].)
The equilibrium points of Eq(6.54) are the solutions of the equation
y =
py + y
q + y
.
So y = 0 is always an equilibrium point of Eq(6.54) and when
p + 1 > q,
Eq(6.54) also possesses the unique positive equilibrium y = p + 1 −q.
The following result follows from Theorems 1.1.1 and 1.3.1.
Theorem 6.8.1 (a) Assume
p + 1 ≤ q.
Then the zero equilibrium of Eq(6.54) is globally asymptotically stable.
(b) Assume
p + 1 > q. (6.55)
Then the zero equilibrium of Eq(6.54) is unstable and the positive equilibrium
y = p + 1 − q of Eq(6.54) is locally asymptotically stable. Furthermore the zero
equilibrium is a saddle point when
1 −p < q < 1 + p
and a repeller when
q < 1 −p.
In the remainder of this section we will investigate the character of the positive
equilibrium of Eq(6.54) and so we will assume without further mention that (6.55) holds.
Our goal is to show that when (6.55) holds and y
−1
+ y
0
> 0, the positive equilibrium
y = p + 1 −q of Eq(6.54) is globally asymptotically stable.
110 Chapter 6. (2, 2)-Type Equations
6.8.1 Invariant Intervals and Semicycle Analysis
Let {y
n
}

n=−1
be a positive solution of Eq(6.54). Then the following identities are easily
established:
y
n+1
−1 = p
y
n

q
p
q + y
n−1
for n ≥ 0, (6.56)
y
n+1

q
p
=
p
2
[y
n

q
p

2
] + (p −q)y
n−1
p(q + y
n−1
)
for n ≥ 0, (6.57)
and
y
n+1
−y
n
=
(p −q)y
n
+ (1 −y
n
)y
n−1
q + y
n−1
for n ≥ 0. (6.58)
Note that the positive equilibrium
y = p + 1 −q
of Eq(6.54) is such that:
y is

< 1 if p < q < p + 1
= 1 if p = q
> 1 if p > q.
When
p = q
that is for the difference equation
y
n+1
=
py
n
+ y
n−1
p + y
n−1
, n = 0, 1, . . . (6.59)
the identities (6.56)-(6.58) reduce to the following:
y
n+1
−1 = (y
n
−1)
p
p + y
n−1
for n ≥ 0 (6.60)
and
y
n+1
−y
n
= (1 −y
n
)
y
n−1
p + y
n−1
for n ≥ 0. (6.61)
The following three lemmas are now direct consequences of the above identities.
Lemma 6.8.1 Assume that
p < q (6.62)
and let {y
n
}

n=−1
be a positive solution of Eq(6.54). Then the following statements are
true:
6.8. The Case α = B = 0 : x
n+1
=
βx
n
+γx
n−1
A+Cx
n−1
111
(i) If for some N ≥ 0, y
N
<
q
p
, then y
N+1
< 1;
(ii) If for some N ≥ 0, y
N
=
q
p
, then y
N+1
= 1;
(iii) If for some N ≥ 0, y
N
>
q
p
, then y
N+1
> 1;
(iv) If for some N ≥ 0, y
N
>

q
p

2
, then y
N+1
<
q
p
;
(v) If for some N ≥ 0, y
N
≤ 1, then y
N+1
< 1;
(vi) If for some N ≥ 0, y
N
≥ 1, then y
N+1
< y
N
.
Lemma 6.8.2 Assume that
p = q (6.63)
and let {y
n
}

n=−1
be a positive solution of Eq(6.54). Then the following statements are
true:
(i) If y
0
< 1, then for n ≥ 0, y
n
< 1 and the solution is strictly increasing;
(ii) If y
0
= 1, then y
n
= 1 for n ≥ 0;
(iii) If y
0
> 1, then for n ≥ 0, y
n
> 1 and the solution is strictly decreasing.
Lemma 6.8.3 Assume that
p > q (6.64)
and let {y
n
}

n=−1
be a positive solution of Eq(6.54). Then the following statements are
true:
(i) If for some N ≥ 0, y
N
<
q
p
, then y
N+1
< 1;
(ii) If for some N ≥ 0, y
N
=
q
p
, then y
N+1
= 1;
(iii) If for some N ≥ 0, y
N
>
q
p
, then y
N+1
> 1;
(iv) If for some N ≥ 0, y
N
>

q
p

2
, then y
N+1
>
q
p
;
(v) If for some N ≥ 0, y
N
≤ 1, then y
N+1
> y
N
;
(vi) If for some N ≥ 0, y
N
≥ 1, then y
N+1
>
q
p
and y
N+2
> 1.
112 Chapter 6. (2, 2)-Type Equations
6.8.2 Global Stability of the Positive Equilibrium
When (6.62) holds, it follows from Lemma 6.8.1 that if a solution {y
n
}

n=−1
is such that
y
n
≥ 1 for all n ≥ 0
then the solution decreases and its limit lies in the interval [1, ∞). This is impossible
because y < 1. Hence, by Lemma 6.8.1, every positive solution of Eq(6.54) eventually
enters and remains in the interval (0, 1). Now in the interval (0, 1) the function
f(u, v) =
pu + v
q + v
is increasing in both arguments and by Theorem 1.4.8, y is a global attractor.
When (6.63) holds, it follows from Lemma 6.8.2 that every solution of Eq(6.54)
converges to the equilibrium y = 1.
Next, assume that (6.64) holds. Here it is clear from Lemma 6.8.3 that every solu-
tion of Eq(6.54) eventually enters and remains in the interval (1, ∞). Without loss of
generality we will assume that
y
n
> 1 for n ≥ −1.
Set
y
n
= 1 + (q + 1)u
n
for n ≥ −1.
Then we can see that {u
n
}

n=−1
satisfies the difference equation
u
n+1
=
p−q
(q+1)
2
+
p
q+1
u
n
1 + u
n−1
, n = 0, 1, . . . (6.65)
with positive parameters and positive initial conditions. This (2, 2)-type equation was
investigated in Section 6.3 where we established in Theorem 6.3.3 that every solution
converges to its positive equilibrium
u =
p −q
q + 1
.
From the above observations and in view of Theorem 6.8.1 we have the following
result:
Theorem 6.8.2 Assume that p + 1 > q and that y
−1
+ y
0
> 0. Then the positive
equilibrium y = p + 1 −q of Eq(6.54) is globally asymptotically stable.
6.9. The Case α = A = 0 : x
n+1
=
βx
n
+γx
n−1
Bx
n
+Cx
n−1
113
6.9 The Case α = A = 0 : x
n+1
=
βx
n
+γx
n−1
Bx
n
+Cx
n−1
This is the (2, 2)-type Eq(6.9) which by the change of variables
x
n
=
γ
C
y
n
reduces to the equation
y
n+1
=
py
n
+ y
n−1
qy
n
+ y
n−1
, n = 0, 1, . . . (6.66)
where
p =
β
γ
and q =
B
C
.
To avoid a degenerate situation we will also assume that
p = q.
(Eq(6.66) was investigated in [54].)
The only equilibrium point of Eq(6.66) is
y =
p + 1
q + 1
and the linearized equation of Eq(6.66) about y is
z
n+1

p −q
(p + 1)(q + 1)
z
n
+
p −q
(p + 1)(q + 1)
z
n−1
= 0, n = 0, 1, . . . .
By applying the linearized stability Theorem 1.1.1 we obtain the following result:
Theorem 6.9.1 (a) Assume that
p > q.
Then the positive equilibrium of Eq(6.66) is locally asymptotically stable.
(b) Assume that
p < q.
Then the positive equilibrium of Eq(6.66) is locally asymptotically stable when
q < pq + 1 + 3p (6.67)
and is unstable (and more precisely a saddle point equilibrium) when
q > pq + 1 + 3p. (6.68)
114 Chapter 6. (2, 2)-Type Equations
6.9.1 Existence of a Two Cycle
It follows from Section 2.5 that when p > q, Eq(6.66) has no prime period-two solutions.
On the other hand when
p < q
and
q > pq + 1 + 3p,
Eq(6.66) possesses a unique prime period-two solution:
. . . , φ, ψ, φ, ψ, . . . (6.69)
where the values of φ and ψ are the (positive and distinct) solutions of the quadratic
equation
t
2
−(1 −p)t +
p(1 −p)
q −1
= 0. (6.70)
In order to investigate the stability nature of this prime period-two solution, we set
u
n
= y
n−1
and v
n
= y
n
for n = 0, 1, . . .
and write Eq(6.66) in the equivalent form:
u
n+1
= v
n
v
n+1
=
pv
n
+u
n
qv
n
+u
n
, n = 0, 1, . . . .
Let T be the function on (0, ∞) ×(0, ∞) defined by:
T

u
v

=

v
pv+u
qv+u

.
Then

φ
ψ

is a fixed point of T
2
, the second iterate of T. One can see that
T
2

u
v

=

g(u, v)
h(u, v)

where
g(u, v) =
pv + u
qv + u
and h(u, v) =
p
pv+u
qv+u
+ v
q
pv+u
qv+u
+ v
.
The prime period-two solution (6.69) is asymptotically stable if the eigenvalues of the
Jacobian matrix J
T
2, evaluated at

φ
ψ

lie inside the unit disk. One can see that
J
T
2

φ
ψ

=

¸
¸
¸

(p−q)ψ
(qψ+φ)
2
(p−q)φ
(qψ+φ)
2

(p−q)
2
ψ
2
(qψ+φ)
2
(qφ+ψ)
2
(p−q)φ
(qφ+ψ)
2

(p−q)ψ
(qψ+φ)
2
−1

¸

.
6.9. The Case α = A = 0 : x
n+1
=
βx
n
+γx
n−1
Bx
n
+Cx
n−1
115
Set
P =
p −q
(qψ + φ)
2
and Q =
p −q
(qφ + ψ)
2
.
Then
J
T
2

φ
ψ

=

−Pψ Pφ
−PQψ
2
PQφψ −Qφ

and its characteristic equation is
λ
2
+ (Pψ + Qφ −PQφψ)λ + PQφψ = 0.
By applying Theorem 1.1.1 (c), it follows that both eigenvalues of
J
T
2

φ
ψ

lie inside the unit disk if and only if
|Pψ + Qφ −PQφψ| < 1 + PQφψ < 2
or equivalently if and only if the following three inequalities hold:
Pψ + Qφ + 1 > 0 (6.71)
Pψ + Qφ < 1 + 2PQφψ (6.72)
and
PQφψ < 1. (6.73)
Observe that
φ > 0, ψ > 0, P < 0 and Q < 0
and so Inequality (6.72) is always true.
Next we will establish Inequality (6.71). We will use the fact that
φ + ψ = 1 −p, φψ =
p(1 −p)
q −1
φ =
pψ + φ
qψ + φ
and ψ =
pφ + ψ
qφ + ψ
.
Inequality (6.71) is equivalent to
(p −q)ψ
(qψ + φ)
2
+
(p −q)φ
(qφ + ψ)
2
+ 1 > 0
which is true if and only if
(q −p)[(qφ + ψ)
2
ψ + (qψ + φ)
2
φ] < (qψ + φ)
2
(qφ + ψ)
2
116 Chapter 6. (2, 2)-Type Equations
if and only if
(q −p)[(qφ + ψ)(pφ + ψ) + (qψ + φ)(pψ + φ)] < [(qφ + ψ)(qψ + φ)]
2
. (6.74)
Now observe that the lefthand side of (6.74) is
I = (q −p)[(qp + 1)(φ
2
+ ψ
2
) + 2(p + q)φψ]
= (q −p)[(qp + 1)(φ + ψ)
2
−2φψ(qp + 1 −p −q)]
= (q −p)[(qp + 1)(1 −p)
2
−2
p(1 −p)
q −1
(p −1)(q −1)]
= (q −p)(1 −p)
2
(qp + 1 + 2p).
The righthand side of (6.74) is
II = [(qψ + φ)(qφ + ψ)]
2
= [(q
2
+ 1)ψφ + q(φ
2
+ ψ
2
)]
2
= [q(φ + ψ)
2
+ (q −1)
2
φψ]
2
= [q(1 −p)
2
+ (q −1)p(1 −p)]
2
= (1 −p)
2
(q −p)
2
.
Hence, Inequality (6.71) is true if and only if (6.68) holds.
Finally, Inequality (6.73) is equivalent to
(p −q)
2
φψ < (qψ + φ)
2
(qφ + ψ)
2
which is true if and only if
(q −p)

φψ < (qψ + φ)(qφ + ψ)
if and only if
(q −p)

φψ < (q
2
+ 1)φψ + q(φ
2
+ ψ
2
)
if and only if
(q −p)

φψ < (q
2
+ 1)φψ + q[(φ + ψ)
2
−2φψ]
if and only if
(q −p)

φψ < (q −1)
2
φψ + q(φ + ψ)
2
if and only if
(q −p)

φψ < (q −1)
2
p(1 −p)
q −1
+ q(1 −p)
2
if and only if
(q −p)

φψ < (1 −p)(q −p)
6.9. The Case α = A = 0 : x
n+1
=
βx
n
+γx
n−1
Bx
n
+Cx
n−1
117
if and only if
p(1 −p)
q −1
< (1 −p)
2
if and only if q > pq + 1 which is clearly true.
In summary, the following result is true about the local stability of the prime period-
two solution (6.69) of Eq(6.66):
Theorem 6.9.2 Assume that Condition (6.68) holds. Then Eq(6.66) posseses the prime
period-two solution
. . . , φ, ψ, φ, ψ, . . .
where φ and ψ are the two positive and distinct roots of the quadratic Eq(6.70). Fur-
thermore this prime period-two solution is locally asymptotically stable.
6.9.2 Semicycle Analysis
In this section, we present a semicycle analysis of the solutions of Eq(6.66).
Theorem 6.9.3 Let {y
n
} be a nontrivial solution of Eq(6.66). Then the following state-
ments are true:
(a) Assume p > q. Then {y
n
} oscillates about the equilibrium y with semicycles of
length two or three, except possibly for the first semicycle which may have length one.
The extreme in each semicycle occurs in the first term if the semicycle has two terms
and in the second term if the semicycle has three terms.
(b) Assume p < q. Then either {y
n
} oscillates about the equilibrium y with semicycles
of length one, after the first semicycle, or {y
n
} converges monotonically to y.
Proof. (a) The proof follows from Theorem 1.7.4 by observing that the condition p > q
implies that the function
f(x, y) =
px + y
qx + y
is increasing in x and decreasing in y. This function also satisfies Condition (1.35).
(b) The proof follows from Theorem 1.7.1 by observing that when p < q, the function
f(x, y) is increasing in y and decreasing in x. 2
6.9.3 Global Stability Analysis When p < q
The main result in this section is the following
Theorem 6.9.4 Assume that
p < q
and (6.67) holds. Then the positive equilibrium y of Eq(6.66) is globally asymptotically
stable.
118 Chapter 6. (2, 2)-Type Equations
Proof. Set
f(x, y) =
px + y
qx + y
,
and note that f(x, y) is decreasing in x for each fixed y, and increasing in y for each
fixed x. Also clearly,
p
q
≤ f(x, y) ≤ 1 for all x, y > 0.
Finally in view of (6.67), Eq(6.66) has no prime period-two solution. Now the conclusion
of Theorem 6.9.4 follows as a consequence of Theorem 1.4.6 and the fact that y is locally
asymptotically stable. 2
The method employed in the proof of Theorem 1.4.6 can also be used to establish
that certain solutions of Eq(6.66) converge to the two cycle (6.69) when instead of (6.67),
(6.68) holds.
Theorem 6.9.5 Assume that (6.68) holds. Let φ, ψ, φ, ψ, . . ., with φ < ψ, denote the
two cycle of Eq(6.66). Assume that for some solution {y
n
}

n=−1
of Eq(6.66) and for
some index N ≥ −1,
y
N
≥ ψ and y
N+1
≤ φ. (6.75)
Then this solution converges to the two cycle φ, ψ, φ, ψ, . . ..
Proof. Assume that (6.75) holds. Set
f(x, y) =
px + y
qx + y
.
Then clearly,
y
N+2
= f(y
N+1
, y
N
) ≥ f(φ, ψ) = ψ,
y
N+3
= f(y
N+2
, y
N+1
) ≤ f(ψ, φ) = φ
and in general
y
N+2k
≥ ψ and y
N+2k+1
≤ φ for k = 0, 1, . . . .
Now as in the proof of Theorem 1.4.6
limsup
n→∞
y
n
= ψ and liminf
n→∞
y
n
= φ
from which we conclude that
lim
k→∞
y
N+2k
= ψ and lim
k→∞
y
N+2k+1
= φ.
2
Next we want to find more cases where the conclusion of the last theorem holds. We
consider first the case where eventually consecutive terms y
i
, y
i+1
lie between m and M.
6.9. The Case α = A = 0 : x
n+1
=
βx
n
+γx
n−1
Bx
n
+Cx
n−1
119
Lemma 6.9.1 Suppose that for some i ≥ 0,
m ≤ y
i
, y
i+1
≤ M.
Then y
k
∈ [m, M] for every k > i.
Proof. Using the monotonic character of the function f we have
m = f(M, m) ≤ f(y
i+1
, y
i
) = y
i+2
≤ f(m, M) = M,
and the result follows by induction. 2
If we never get two successive terms in the interval [m, M] and if we never get two
successive terms outside of the interval [m, M] (that is, for every i, either y
i
> M and
y
i+1
< m or y
i
< m and y
i+1
> M), then we have one of the following four cases:
(A) y
2n
> M and M ≥ y
2n+1
> m for every n;
(B) y
2n+1
> M and M ≥ y
2n
> m for every n;
(C) y
2n
< m and m ≤ y
2n+1
< M for every n;
(D) y
2n+1
< m and m ≤ y
2n
< M for every n.
Lemma 6.9.2 In all cases (A)-(D) the corresponding solution {y
n
} converges to the
two cycle
. . . φ, ψ, φ, ψ, . . . .
Proof. Since the proofs are similar for all cases we will give the details in case (A). By
our earlier observation we have that M ≥ limsup
i→∞
y
i
, and so
M ≥ limsup
n→∞
y
2n
≥ liminf
n→∞
y
2n
≥ M.
Hence lim
n→∞
y
2n
= M. From Eq(6.66) we have
y
2n+1
=
y
2n
−y
2n
y
2n+2
qy
2n+2
−p
.
and so
lim
n→∞
y
2n+1
=
M −M
2
qM −p
.
On the other hand by solving the equation
M = f(m, M) =
pm + M
qm + M
for m we find
m =
M −M
2
qM −p
.
Thus, lim
n→∞
y
2n+1
= m and the proof is complete. 2
Since the case in which the solution lies outside of the interval [m, M] is answered by
Theorem 6.9.5, we are left with the case in which the solution lies eventually in [m, M].
120 Chapter 6. (2, 2)-Type Equations
Lemma 6.9.3 Suppose M ≥ y
i+2
≥ y
i
≥ y ≥ y
i+1
≥ y
i+3
≥ m and not both y
i+2
and
y
i+3
are equal to y. Then {y
n
} converges to the two cycle
. . . φ, ψ, φ, ψ, . . . .
Proof. Here we have
y
i+4
= f(y
i+3
, y
i+2
) ≥ f(y
i+1
, y
i
) = y
i+2
and
y
i+5
= f(y
i+4
, y
i+3
) ≤ f(y
i+2
, y
i+1
) = y
i+3
.
Induction on i then establishes the monotonicity of the two sequences. Thus, the se-
quence {y
i+2k
} is non decreasing and bounded above by M, and the sequence {y
i+2k+1
}
is a non increasing sequence and bounded below by m. Therefore, both subsequences
converge and their limits form a two cycle. In view of the uniqueness of the two cycle
and the fact that at least one of y
i
, y
i+1
is different from y, the proof is complete.The
case M ≥ y
i+3
≥ y
i+1
≥ y ≥ y
i
≥ y
i+2
≥ m is handled in a similar way. 2
Using a similar technique one can prove that if
M ≥ y
i
≥ y
i+2
≥ y ≥ y
i+3
≥ y
i+1
≥ m
or
M ≥ y
i+1
≥ y
i+3
≥ y ≥ y
i+2
≥ y
i
≥ m
we would get convergence to y , but this situation cannot occur, as the following results
show. Let us consider the first case. The second case can be handled in a similar way.
In this case we need to know more about the condition y
i+2
= f(y
i+1
, y
i
) ≤ y
i
, so we are
led to the general equation
f(x, y) = y. (6.76)
It makes sense to consider this equation only for y ∈ (
p
q
, 1), in which case Eq(6.76) has a
unique positive solution for x, which of course depends on y. We denote this value by y

.
Thus f(y

, y) = y. The correspondence between y and y

has the following properties:
Lemma 6.9.4 (1) f(x, y) > y if and only if x < y

.
(2) x > y if and only if x

< y

.
(3) For y < y < M, f(y, y

) < y

and y > y

, and for m < y < y, f(y, y

) > y

and y

> y.
Proof.
6.9. The Case α = A = 0 : x
n+1
=
βx
n
+γx
n−1
Bx
n
+Cx
n−1
121
(1) Since f is decreasing in the first variable and f(y

, y) = y, x < y

if and only if
f(x, y) > f(y

, y) = y.
(2) Solving f(y

, y) = y for y

gives
y

= g(y) =
y −y
2
qy −p
.
Computing the derivative of g we get
g

(y) = −
p −2y + qy
2
(p −qy)
2
.
The discriminant of the numerator is 4p
2
− 4pq = 4p(p − q) < 0, so the roots of
the numerator are not real. For y ∈ (
p
q
, 1), g

(y) < 0, so y

is decreasing function
of y.
(3) Eliminating y

from the equations y = f(y

, y) and y

= f(y, y

) we get
x −x
2
qx −p
=
px +
x−x
2
qx−p
qx +
x−x
2
qx−p
.
Simplifying this equation we get a fourth degree polynomial equation with roots
0, y, m, M. Thus each of these roots is simple and so f(y, y

)−y

changes sign as y
passes each of these values. Clearly as y approaches
p
q
y

approaches ∞. However
p
q
< f(y, y

) < 1. Thus for y < m, f(y, y

) < y

. Hence
m < y < y ⇒f(y, y

) > y

,
y < y < M ⇒f(y, y

) < y

,
M < y <⇒f(y, y

) > y

.
By (1) then, for m < y < y, y < y

, and for y < y < M, y > y

.
2
Now we will use this result to prove the following:
Lemma 6.9.5 Assume that for some i,
M ≥ y
i
≥ y
i+2
≥ y ≥ y
i+1
≥ m.
Then y
i+3
≤ y
i+1
with equality only in the case of a two cycle.
122 Chapter 6. (2, 2)-Type Equations
Proof. To show y
i+3
= f(y
i+2
, y
i+1
) ≤ y
i+1
, we must show y
i+2
≥ y

i+1
.
Since y
i+2
= f(y
i+1
, y
i
) ≤ y
i
, we have y
i+1
≥ y

i
. Thus by (2) of the previous lemma
y

i
≥ y

i+1
, and by (3) y
i
≥ y

i
. Thus y
i
≥ y

i+1
. Since f is increasing in the second
argument and y
i
≥ y

i+1
we get
y
i+2
= f(y
i+1
, y
i
) ≥ f(y
i+1
, y

i+1
).
By (3) of the previous lemma, since m ≤ y
i+1
≤ y, f(y
i+1
, y

i+1
) ≥ y

i+1
. Thus y
i+2
≥ y

i+1
,
as required. To get equality we must have y
i+1
= y

i
or equivalently y
i+2
= y
i
, as well as
y
i+3
= y
i+1
. Thus equality occurs only in the case of a two cycle. 2
The case
m ≤ y
i
≤ y
i+2
≤ y ≤ y
i+1
≤ M
is similar.
We also obtain the following lemma:
Lemma 6.9.6 Unless {y
n
} is a period-two solution, there is no i such that either
M ≥ y
i
≥ y
i+2
≥ . . . ≥ y ≥ y
i+1
≥ y
i+3
≥ . . . ≥ m,
or
M ≥ . . . ≥ y
i+2
≥ y
i
≥ y ≥ . . . ≥ y
i+3
≥ y
i+1
≥ . . . ≥ m.
Proof. We will consider the first case; the second case is similar. Since {y
i+2k+1
} is a
bounded monotonic sequence it has a limit c. The value c will be one value in a two
cycle, so either c = y or c = m. If c = y, then y
i+2n+1
= y for all n, and solving for y
i+2n
we see y
i+2n
= y also. But then y
i
= y
i+1
= y, and our solution is a two cycle.. If c = m,
solving for lim
n→∞
y
i+2n
gives lim
n→∞
y
i+2n
= M. This can only happen if y
i+2n
= M
for all n, in which case y
i+2n+1
= m for all n, which also gives a two cycle. 2
The above sequence of lemmas leads to the following result:
Theorem 6.9.6 Assume that
p < q,
and that Eq(6.66) possesses the two-cycle solution (6.69). Then every oscillatory solution
of Eq(6.66) converges to this two cycle.
Remark. Since some of the oscillatory solutions of Eq(6.66) are generated by initial
values {y
−1
, y
0
} that are on opposite sides of the equilibrium y, it follows by Theorem
6.9.6 that all such pairs belong to the stable manifold of the two cycle.
6.10. Open Problems and Conjectures 123
6.9.4 Global Stability Analysis When p > q
Here we have the following result:
Theorem 6.9.7 Assume that
p > q
and
p ≤ pq + 1 + 3q. (6.77)
Then the positive equilibrium y of Eq(6.66) is globally asymptotically stable.
Proof. We will employ Theorem 1.4.5. To this end, set
f(x, y) =
px + y
qx + y
,
and observe that when p > q, the function f(x, y) is increasing in x for each fixed y, and
is decreasing in y for each fixed x. Also
1 ≤ f(x, y) ≤
p
q
for all x, y > 0.
Finally observe that when (6.77) holds, the only solution of the system
M =
pM + m
qM + m
, m =
pm + M
qm + M
,
is
m = M.
Now the result is a consequence of Theorem 1.4.5. 2
6.10 Open Problems and Conjectures
Open Problem 6.10.1 (See [64]) Assume that
a, b, c, d ∈ R.
(a) Investigate the forbidden set F of the difference equation
x
n+1
=
ax
n
+ bx
n−1
cx
n
+ dx
n−1
x
n
, n = 0, 1, . . . .
(b) For every point (x
−1
, x
0
) / ∈ F, investigate the asymptotic behavior and the periodic
nature of the solution {x
n
}

n=−1
.
124 Chapter 6. (2, 2)-Type Equations
(See Section 1.6.)
Open Problem 6.10.2 Assume that
α
n
, β
n
, A
n
, B
n
∈ R for n = 0, 1, . . .
are convergent sequences of real numbers with finite limits. Investigate the forbidden set
and the asymptotic character of solutions of the nonautonomous Riccati equation
x
n+1
=
α
n
+ β
n
x
n
A
n
+ B
n
x
n
, n = 0, 1, . . . .
(See Section 1.6 for the autonomous case.)
Conjecture 6.10.1 Assume
p, q ∈ (0, ∞).
Show that every positive solution of Eq(6.10) has a finite limit.
(Note that by Theorem 6.3.3, we need only to confirm this conjecture for p > 2(q +1)
and q ≥ 1. See also [36].)
Conjecture 6.10.2 Assume that
p, q ∈ (0, ∞).
Show that every positive solution of Eq(6.11) has a finite limit.
(Note that by Theorem 6.4.4, this conjecture has been confirmed for q ≤ 1 + 4p.)
Conjecture 6.10.3 Assume p ∈ (0, ∞). Show that the equation
y
n+1
=
p + y
n−1
1 + y
n
, n = 0, 1, . . .
has a positive and monotonically decreasing solution.
(Note that by Theorem 6.5.2, every positive solution converges to a, not necessarily
prime, period-two solution.)
6.10. Open Problems and Conjectures 125
Conjecture 6.10.4 Assume that
p, q ∈ (0, ∞).
Show that every positive solution of Eq(6.27) either converges to a finite limit or to a
two cycle.
(See Section 6.6.)
Open Problem 6.10.3 Assume that
p, q ∈ (0, ∞)
and
q > 1 + 4p.
Find the basin of attraction of the period-two solution of Eq(6.27).
Open Problem 6.10.4 Assume that
p, q ∈ (0, ∞)
and
q > pq + 1 + 3p.
Find the basin of attraction of the period-two solution of Eq(6.66).
Conjecture 6.10.5 Assume
p > q.
Show that every positive solution of Eq(6.66) has a finite limit.
(Note that by Theorem 6.9.7, this conjecture has been confirmed for p ≤ pq +1+3q.)
Open Problem 6.10.5 Find the set of all initial conditions (y
−1
, y
0
) ∈ R×R through
which the solution of the equation
y
n+1
=
y
n
+ y
n−1
1 + y
n
is well defined and converges to 1, as n →∞.
126 Chapter 6. (2, 2)-Type Equations
Open Problem 6.10.6 Find the set of all initial conditions (y
−1
, y
0
) ∈ R×R through
which the solution of the equation
y
n+1
=
y
n
+ y
n−1
1 + y
n−1
is well defined and converges to 1, as n →∞.
Open Problem 6.10.7 For each of the following difference equations determine the
“good” set G ⊂ R×R of initial conditions (y
−1
, y
0
) ∈ R×R through which the equation
is well defined for all n ≥ 0.Then for every (y
−1
, y
0
) ∈ G, investigate the long-term
behavior of the solution {y
n
}

n=−1
:
y
n+1
=
1 −y
n
1 −y
n−1
(6.78)
y
n+1
=
y
n
−1
y
n
−y
n−1
(6.79)
y
n+1
=
1 −y
n−1
1 −y
n
(6.80)
y
n+1
=
1 −y
n−1
y
n
−y
n−1
(6.81)
y
n+1
=
y
n
−y
n−1
y
n
−1
(6.82)
y
n+1
=
y
n
−y
n−1
1 −y
n−1
(6.83)
y
n+1
=
y
n
+ y
n−1
y
n
−y
n−1
. (6.84)
For those equations from the above list for which you were successful, extend your
result by introducing arbitrary real parameters in the equation.
Open Problem 6.10.8 Assume
α, β, γ, A, B, C ∈ (0, ∞).
Investigate the asymptotic behavior and the periodic nature of solutions of the difference
equation
x
n+1
=
αx
n
+ βx
n−1
+ γx
n−2
Ax
n
+ Bx
n−1
+ Cx
n−2
, n = 0, 1, . . . .
Extend and generalize.
6.10. Open Problems and Conjectures 127
Open Problem 6.10.9 Assume that {p
n
}

n=0
and {q
n
}

n=0
are convergent sequences of
nonnegative real numbers with finite limits,
p = lim
n→∞
p
n
and q = lim
n→∞
q
n
.
Investigate the asymptotic behavior and the periodic nature of all positive solutions of
each of the following eight difference equations:
y
n+1
=
p
n
+ y
n
q
n
+ y
n
, n = 0, 1, . . . (6.85)
y
n+1
=
p
n
+ q
n
y
n
1 + y
n−1
, n = 0, 1, . . . (6.86)
y
n+1
=
y
n
+ p
n
y
n
+ q
n
y
n−1
, n = 0, 1, . . . (6.87)
y
n+1
=
p
n
+ q
n
y
n−1
1 + y
n
, n = 0, 1, . . . (6.88)
y
n+1
=
p
n
+ y
n−1
q
n
y
n
+ y
n−1
, n = 0, 1, . . . (6.89)
y
n+1
=
p
n
y
n
+ q
n
y
n−1
1 + y
n
, n = 0, 1, . . . (6.90)
y
n+1
=
p
n
y
n
+ y
n−1
q
n
+ y
n−1
, n = 0, 1, . . . (6.91)
y
n+1
=
p
n
y
n
+ y
n−1
q
n
y
n
+ y
n−1
, n = 0, 1, . . . . (6.92)
Open Problem 6.10.10 Assume that {p
n
}

n=0
and {q
n
}

n=0
are period-two sequences
of nonnegative real numbers. Investigate the global character of all positive solutions of
Eqs(6.85)-(6.92). Extend and generalize.
Open Problem 6.10.11 Assume q ∈ (1, ∞).
(a) Find the set B of all initial conditions (y
−1
, y
0
) ∈ (0, ∞) × (0, ∞) such that the
solutions {y
n
}

n=−1
of Eq(6.23) are bounded.
(b) Let (y
−1
, y
0
) ∈ B. Investigate the asymptotic behavior of {y
n
}

n=−1
.
128 Chapter 6. (2, 2)-Type Equations
Open Problem 6.10.12 Assume q ∈ (1, ∞).
(a) Find the set B of all initial conditions (y
−1
, y
0
) ∈ (0, ∞) × (0, ∞) such that the
solutions {y
n
}

n=−1
of Eq(6.47) are bounded.
(b) Let (y
−1
, y
0
) ∈ B. Investigate the asymptotic behavior of {y
n
}

n=−1
.
Open Problem 6.10.13 (A Plant-Herbivore System; See [9], [3], and [2]) Assume
α ∈ (1, ∞), β ∈ (0, ∞), and γ ∈ (0, 1) with α + β > 1 +
β
γ
.
Obtain conditions for the global asymptotic stability of the positive equilibrium of the
system
x
n+1
=
αx
n
βx
n
+e
y
n
y
n+1
= γ(x
n
+ 1)y
n

, n = 0, 1, . . . . (6.93)
Conjecture 6.10.6 Assume x
0
, y
0
∈ (0, ∞) and that
α ∈ (2, 3), β = 1 and γ =
1
2
.
Show that the positive equilibrium of System (6.93) is globally asymptotically stable.
Open Problem 6.10.14 (Discrete Epidemic Models; See [14].) Let A ∈ (0, ∞).
(a) Determine the set G of initial conditions (y
−1
, y
0
) ∈ (0, ∞) ×(0, ∞) through which
the solutions {x
n
}

n=−1
of the difference equation
x
n+1
= (1 −x
n
−x
n−1
)(1 −e
−Ax
n
), n = 0, 1, . . . (6.94)
are nonnegative.
(b) Let (x
−1
, x
0
) ∈ G. Investigate the boundedness character, the periodic nature, and
the asymptotic behavior of the solution {x
n
}

n=−1
of Eq(6.94).
(c) Extend and generalize.
6.10. Open Problems and Conjectures 129
Open Problem 6.10.15 (The Flour Beetle Model; See [48]). Assume
a ∈ (0, 1), b ∈ (0, ∞), and c
1
, c
2
∈ [0, ∞) with c
1
+ c
2
> 0.
Obtain necessary and sufficient conditions for the global asymptotic stability of the Flour
Beetle Model:
x
n+1
= ax
n
+ bx
n−2
e
−c
1
x
n
−c
2
x
n−2
, n = 0, 1, . . .
with positive initial conditions.
Open Problem 6.10.16 (A Population Model) Assume
α ∈ (0, 1) and β ∈ (1, ∞).
Investigate the global character of all positive solutions of the system
x
n+1
= αx
n
e
−y
n
+ β
y
n+1
= αx
n
(1 −e
−y
n
)

, n = 0, 1, . . . .
which may be viewed as a population model.
Open Problem 6.10.17 Assume that
p, q ∈ [0, ∞) and k ∈ {2, 3, . . .}.
Investigate the global behavior of all positive solutions of each of the following difference
equations:
y
n+1
=
p + qy
n
1 + y
n−k
, n = 0, 1, . . . (6.95)
y
n+1
=
y
n
+ p
y
n
+ qy
n−k
, n = 0, 1, . . . (6.96)
y
n+1
=
p + qy
n−k
1 + y
n
, n = 0, 1, . . . (6.97)
y
n+1
=
p + y
n−k
qy
n
+ y
n−k
, n = 0, 1, . . . (6.98)
y
n+1
=
y
n
+ py
n−k
y
n
+ q
, n = 0, 1, . . . (6.99)
y
n+1
=
py
n
+ y
n−k
q + y
n−k
, n = 0, 1, . . . (6.100)
y
n+1
=
py
n
+ y
n−k
qy
n
+ y
n−k
, n = 0, 1, . . . . (6.101)
130 Chapter 6. (2, 2)-Type Equations
Open Problem 6.10.18 Obtain a general result for an equation of the form
y
n+1
= f(y
n
, y
n−1
), n = 0, 1, . . .
which extends and unifies the global asymptotic stability results for Eqs(6.11) and (6.27).
Open Problem 6.10.19 Assume that every positive solution of an equation of the form
x
n+1
= f(x
n
, x
n−1
), n = 0, 1, . . . (6.102)
converges to a period-two solution. Obtain necessary and sufficient conditions on f so
that every positive solution of the equation
x
n+1
= f(x
n−2
, x
n−1
), n = 0, 1, . . . (6.103)
also converges to a period-two solution.
Extend and generalize.
Open Problem 6.10.20 Assume that Eq(6.102) has a two cycle which is locally asymp-
totically stable. Obtain necessary and sufficient conditions on f so that the same two
cycle is a locally asymptotically stable solution of Eq(6.103).
Chapter 7
(1, 3)-Type Equations
7.1 Introduction
Eq(1) contains the following three equations of the (1, 3)-type:
x
n+1
=
α
A +Bx
n
+Cx
n−1
, n = 0, 1, . . . (7.1)
x
n+1
=
βx
n
A +Bx
n
+Cx
n−1
, n = 0, 1, . . . (7.2)
and
x
n+1
=
γx
n−1
A +Bx
n
+Cx
n−1
, n = 0, 1, . . . . (7.3)
Please recall our classification convention in which all the parameters in the above
(1, 3)-type equations are positive and that the initial conditions are nonnegative.
7.2 The Case β = γ = 0 : x
n+1
=
α
A+Bx
n
+Cx
n−1
This is the (1, 3)-type Eq.(7.1) which by the change of variables
x
n
=
α
A
y
n
reduces to the difference equation
y
n+1
=
1
1 +py
n
+qy
n−1
, n = 0, 1, . . . (7.4)
where
p =
αB
A
2
and q =
αC
A
2
.
131
132 Chapter 7. (1, 3)-Type Equations
One can easily see that the positive equilibrium of Eq(7.4) is locally asymptotically stable
for all values of the parameters and that Eq(7.4) has no prime period two solutions.
The following result is now a straightforward consequence of either the stability
trichotomy Theorem 1.4.4, or Theorem 1.4.7 in the interval [0, 1].
Theorem 7.2.1 The positive equilibrium of Eq(7.4) is globally asymptotically stable.
7.3 The Case α = γ = 0 : x
n+1
=
βx
n
A+Bx
n
+Cx
n−1
This is the (1, 3)-type Eq(7.2) which by the change of variables
x
n
=
A
C
y
n
reduces to the difference equation
y
n+1
=
y
n
1 +py
n
+qy
n−1
, n = 0, 1, . . . (7.5)
where
p =
β
A
and q =
B
C
.
Eq(7.5) always has zero as an equilibrium point and when
p > 1
it also has the unique positive equilibrium point
y =
p −1
q + 1
.
The following result is a straightforward consequence of Theorem 1.3.1.
Theorem 7.3.1 Assume
p ≤ 1.
Then the zero equilibrium of Eq(7.5) is globally asymptotically stable.
The following result is a straightforward consequence of Theorem 1.4.2. It also follows
by applying Theorem 1.4.5 in the interval [0,
p
q
].
Theorem 7.3.2 Assume that y
−1
, y
0
∈ (0, ∞) and
p > 1.
Then the positive equilibrium of Eq(7.5) is globally asymptotically stable.
7.4. The Case α = β = 0 : x
n+1
=
γx
n−1
A+Bx
n
+Cx
n−1
133
7.4 The Case α = β = 0 : x
n+1
=
γx
n−1
A+Bx
n
+Cx
n−1
This is the (1, 3)-type Eq(7.3) which by the change of variables
x
n
=
γ
C
y
n
reduces to the equation
y
n+1
=
y
n−1
p +qy
n
+y
n−1
, n = 0, 1, . . . (7.6)
where
p =
A
γ
and q =
B
C
.
Eq(7.6) always has the zero equilibrium and when
p < 1
it also has the unique positive equilibrium
y =
1 −p
1 +q
.
The subsequent result is a straightforward application of Theorems 1.1.1 and 1.3.1.
Theorem 7.4.1 (a) The zero equilibrium of Eq(7.6) is globally asymptotically stable
when
p ≥ 1
and is unstable (a repeller) when
p < 1.
(b) Assume that p < 1. Then the positive equilibrium of Eq(7.6) is locally asymptot-
ically stable when
q < 1
and is unstable (a saddle point) when
q > 1.
Concerning prime period-two solutions one can see that Eq(7.6) has a prime period-
two solution
. . . , φ, ψ, φ, ψ, . . .
if and only if
p < 1.
134 Chapter 7. (1, 3)-Type Equations
Furthermore when
p < 1 and q = 1
the only prime period-two solution is
. . . , 0, 1 −p, 0, 1 −p, . . . .
On the other hand, when
p < 1 and q = 1
all prime period-two solutions of Eq(7.6) are given by
. . . , φ, 1 −p −φ, φ, 1 −p −φ, . . .
with
0 ≤ φ ≤ 1 −p and φ =
1 −p
2
.
Theorem 7.4.2 Assume that
p < 1 and q > 1. (7.7)
Then the period-two solution
. . . , 0, 1 −p, 0, 1 −p, . . . (7.8)
of Eq(7.6) is locally asymptotically stable.
Proof. It follows from Section 2.6 that here
J
T
2

0
1 −p

=


1
p+q(1−p)
0

q(1−p)
p+q(1−p)
p


and in view of (7.7) both eigenvalues of J
T
2

0
1 −p

lie in the disk |λ| < 1. Therefore
the prime period-two solution (7.8) is locally asymptotically stable. 2
7.5 Open Problems and Conjectures
Conjecture 7.5.1 Assume
p, q ∈ (0, 1).
Show that every positive solution of Eq(7.6) has a finite limit.
7.5. Open Problems and Conjectures 135
Conjecture 7.5.2 Assume
p < 1 and q ≥ 1.
Show that every positive solution of Eq(7.6) converges to a, not necessarily prime, period-
two solution.
Open Problem 7.5.1 Assume
p ∈ (0, 1) and q ∈ (1, ∞).
Find the basin of attraction of the two cycle
. . . , 0, 1 −p, 0, 1 −p, . . .
of Eq(7.6).
Open Problem 7.5.2 Investigate the asymptotic character and the periodic nature of
all positive solutions of the difference equation
x
n+1
=
x
n−2
1 +px
n
+qx
n−1
+x
n−2
, n = 0, 1, . . .
where
p, q ∈ [0, ∞).
Extend and generalize.
Open Problem 7.5.3 For each of the equations given below, find the set G of all points
(x
−1
, x
0
) ∈ R ×R through which the equation is well defined for all n ≥ 0:
x
n+1
=
1
1 +x
n
+x
n−1
(7.9)
x
n+1
=
x
n
1 +x
n
+x
n−1
(7.10)
x
n+1
=
x
n−1
1 +x
n
+x
n−1
. (7.11)
Now for each of the equations (7.9)-(7.11), let (x
−1
, x
0
) ∈ G be an initial point
through which the corresponding equation is well defined for all n ≥ 0, and investigate
the long term-behavior of the solution {x
n
}

n=−1
.
136 Chapter 7. (1, 3)-Type Equations
Open Problem 7.5.4 Assume that {p
n
}

n=0
and {q
n
}

n=0
are convergent sequences of
nonnegative real numbers with finite limits,
p = lim
n→∞
p
n
and q = lim
n→∞
q
n
.
Investigate the asymptotic behavior and the periodic nature of all positive solutions of
each of the following three difference equations:
y
n+1
=
1
1 +p
n
y
n
+q
n
y
n−1
, n = 0, 1, . . . (7.12)
y
n+1
=
y
n
1 +p
n
y
n
+q
n
y
n−1
, n = 0, 1, . . . (7.13)
y
n+1
=
y
n−1
p
n
+q
n
y
n
+y
n−1
, n = 0, 1, . . . . (7.14)
Open Problem 7.5.5 Assume that {p
n
}

n=0
and {q
n
}

n=0
are period-two sequences of
nonnegative real numbers. Investigate the global character of all positive solutions of
Eqs(7.12)-(7.14). Extend and generalize.
Chapter 8
(3, 1)-Type Equations
8.1 Introduction
Eq(1) contains the following three equations of the (1, 3)-type:
x
n+1
=
α +βx
n
+γx
n−1
A
, n = 0, 1, . . . (8.1)
x
n+1
=
α +βx
n
+γx
n−1
Bx
n
, n = 0, 1, . . . (8.2)
and
x
n+1
=
α +βx
n
+γx
n−1
Cx
n−1
, n = 0, 1, . . . . (8.3)
Please recall our classification convention in which all parameters that appear in these
equations are positive, the initial conditions are nonnegative, and the denominators are
always positive.
Eq(8.1) is a linear difference equation and can be solved explicitly.
Eq(8.2) by the change of the variables
x
n
= y
n
+
β
B
,
is reduced to a (2, 2)-type equation of the form of Eq(6.4) (see Section 6.5), namely,
y
n+1
=
(α +
βγ
B
) +γy
n−1
β +By
n
, n = 0, 1, . . . .
Eq(8.3) by the change of the variables
x
n
= y
n
+
γ
C
,
137
138 Chapter 8. (3, 1)-Type Equations
is reduced to a (2, 2)-type equation of the form of Eq(6.2) (see Section 6.3) namely,
y
n+1
=
(α +
βγ
C
) +βy
n
γ +Cy
n−1
, n = 0, 1, . . . .
Hence there is nothing else remaining to do about these equations other than to pose
some Open Problems and Conjectures.
8.2 Open Problems and Conjectures
Open Problem 8.2.1 Assume γ > β. Determine the set of initial conditions (x
−1
, x
0
) ∈
(0, ∞) ×(0, ∞) through which the solutions of Eq(8.2) are bounded.
Open Problem 8.2.2 (a) Assume γ = β and let
. . . , φ, ψ, φ, ψ, . . . (8.4)
be a period-two solution of Eq(8.2). Determine the basin of attraction of (8.4).
(b) Let {x
n
}

n=−1
be a positive solution of Eq(8.2) which converges to (8.4). Determine
the values of φ and ψ in terms of the initial conditions x
−1
and x
0
.
Conjecture 8.2.1 Show that every positive solution of Eq(8.3) converges to the positive
equilibrium of the equation.
Open Problem 8.2.3 Determine the set G of all initial points (x
−1
, x
0
) ∈ R × R
through which the equation
x
n+1
=
1 +x
n
+x
n−1
x
n
is well defined for all n and for these initial points determine the long-term behavior of
the solutions {x
n
}

n=−1
.
Open Problem 8.2.4 Determine the set G of all initial points (x
−1
, x
0
) ∈ R × R
through which the equation
x
n+1
=
1 +x
n
+x
n−1
x
n−1
is well defined for all n and for these initial points determine the long-term behavior of
the solutions {x
n
}

n=−1
.
8.2. Open Problems and Conjectures 139
Open Problem 8.2.5 Assume that {p
n
}

n=0
and {q
n
}

n=0
are convergent sequences of
nonnegative real numbers with finite limits,
p = lim
n→∞
p
n
and q = lim
n→∞
q
n
.
Investigate the asymptotic behavior and the periodic nature of all positive solutions of
each of the following two difference equations:
y
n+1
=
p
n
+x
n
+x
n−1
q
n
x
n
, n = 0, 1, . . . (8.5)
y
n+1
=
p
n
+x
n
+x
n−1
q
n
x
n−1
, n = 0, 1, . . . . (8.6)
Open Problem 8.2.6 Assume that {p
n
}

n=0
and {q
n
}

n=0
are period-two sequences of
nonnegative real numbers. Investigate the global character of all positive solutions of
Eqs(8.5) and (8.6). Extend and generalize.
Chapter 9
(2, 3)-Type Equations
9.1 Introduction
Eq(1) contains the following three equations of the (2, 3)-type:
x
n+1
=
α + βx
n
A + Bx
n
+ Cx
n−1
, n = 0, 1, . . . (9.1)
x
n+1
=
α + γx
n−1
A + Bx
n
+ Cx
n−1
, n = 0, 1, . . . (9.2)
and
x
n+1
=
βx
n
+ γx
n−1
A + Bx
n
+ Cx
n−1
, n = 0, 1, . . . . (9.3)
Please recall our classification convention in which all the parameters in the above
(2, 3)-type equations are positive and the initial conditions are nonnegative.
9.2 The Case γ = 0 : x
n+1
=
α+βx
n
A+Bx
n
+Cx
n−1
This is the (2, 3)-type Eq(9.1) which by the change of variables
x
n
=
A
B
y
n
reduces to the difference equation
y
n+1
=
p + qy
n
1 + y
n
+ ry
n−1
, n = 0, 1, . . . (9.4)
where
p =
αB
A
2
, q =
β
A
, and r =
C
B
.
141
142 Chapter 9. (2, 3)-Type Equations
(Eq(9.4) was investigated in [51].)
Here we make some simple observations about linearized stability, invariant intervals,
and the convergence of solutions in certain regions of initial conditions.
Eq(9.4) has a unique equilibrium y which is positive and is given by
y =
q −1 +

(q −1)
2
+ 4p(r + 1)
2(r + 1)
.
By employing the linearized stability Theorem 1.1.1, we obtain the following result:
Theorem 9.2.1 The equilibrium y of Eq(9.4) is locally asymptotically stable for all
values of the parameters p, q and r.
9.2.1 Boundedness of Solutions
We will prove that all solutions of Eq(9.4) are bounded.
Theorem 9.2.2 Every solution of Eq(9.4) is bounded from above and from below by
positive constants.
Proof. Let {y
n
} be a solution of Eq(9.4). Clearly, if the solution is bounded from above
by a constant M, then
y
n+1

p
1 + (1 + r)M
and so it is also bounded from below. Now assume for the sake of contradiction that the
solution is not bounded from above. Then there exists a subsequence {y
1+n
k
}

k=0
such
that
lim
k→∞
n
k
= ∞, lim
k→∞
y
1+n
k
= ∞, and y
1+n
k
= max{y
n
: n ≤ n
k
} for k ≥ 0.
From (9.4) we see that
y
n+1
< qy
n
+ p for n ≥ 0
and so
lim
k→∞
y
n
k
= lim
k→∞
y
n
k
−1
= ∞.
Hence for sufficiently large k,
0 < y
1+n
k
−y
n
k
=
p + [(q −1) −y
n
k
−ry
n
k
−1
]y
n
k
1 + y
n
k
+ ry
n
k
−1
< 0
which is a contradiction and the proof is complete. 2
The above proof has an advantage that extends to several equations of the form
of Eq(9.1) with nonnegative parameters. The boundedness of solutions of the special
9.2. The Case γ = 0 : x
n+1
=
α+βx
n
A+Bx
n
+Cx
n−1
143
equation (9.4) with positive parameters immediately follows from the observation that
if
M = max{1, p, q}
then
y
n+1

M + My
n
1 + y
n
= M for n ≥ 0.
9.2.2 Invariant Intervals
The following result, which can be established by direct calculation, gives a list of in-
variant intervals for Eq(9.4). Recall that I is an invariant interval, if whenever,
y
N
, y
N+1
∈ I for some integer N ≥ 0,
then
y
n
∈ I for n ≥ N.
Lemma 9.2.1 Eq(9.4) possesses the following invariant intervals:
(a)

0,
q −1 +

(q −1)
2
+ 4p
2
¸
¸
, when p ≤ q;
(b)
¸
p −q
qr
, q
¸
, when q < p < q(rq + 1);
(c)
¸
q,
p −q
qr
¸
, when p > q(rq + 1).
Proof.
(a) Set
g(x) =
p + qx
1 + x
and b =
q −1 +

(q −1)
2
+ 4p
2
and observe that g is an increasing function and g(b) ≤ b. Using Eq(9.4) we see
that when y
k−1
, y
k
∈ [0, b], then
y
k+1
=
p + qy
k
1 + y
k
+ ry
k−1

p + qy
k
1 + y
k
= g(y
k
) ≤ g(b) ≤ b.
The proof follows by induction.
144 Chapter 9. (2, 3)-Type Equations
(b) It is clear that the function
f(x, y) =
p + qx
1 + x + ry
is increasing in x for y >
p−q
qr
. Using Eq(9.4) we see that when y
k−1
, y
k

p−q
qr
, q

,
then
y
k+1
=
p + qy
k
1 + y
k
+ ry
k−1
= f(y
k
, y
k−1
) ≤ f

q,
p −q
qr

= q.
Also by using the condition p < q(rq + 1) we obtain
y
k+1
=
p + qy
k
1 + y
k
+ ry
k−1
= f(y
k
, y
k−1
) ≥ f

p −q
qr
, q

=
q(pr + p −q)
(rq)
2
+ rq + p −q
>
p −q
qr
.
The proof follows by the induction.
(c) It is clear that the function
f(x, y) =
p + qx
1 + x + ry
is decreasing in x for y <
p−q
qr
. Using Eq(9.4) we see that when y
k−1
, y
k

q,
p−q
qr

,
then
y
k+1
=
p + qy
k
1 + y
k
+ ry
k−1
= f(y
k
, y
k−1
) ≥ f

p −q
qr
,
p −q
qr

= q.
Also by using the condition p > q(rq + 1) we obtain
y
k+1
=
p + qy
k
1 + y
k
+ ry
k−1
= f(y
k
, y
k−1
) ≤ f(q, q) =
p + q
2
1 + (r + 1)q
<
p −q
qr
.
The proof follows by induction.
2
9.2.3 Semicycle Analysis
Let {y
n
}

n=−1
be a solution of Eq(9.4). Then the following identitites are true for n ≥ 0:
y
n+1
−q =
qr(
p−q
qr
−y
n−1
)
1 + y
n
+ ry
n−1
, (9.5)
y
n+1

p −q
qr
=
qr

q −
p−q
qr

y
n
+ qr

y
n−1

p−q
qr

+ pr(q −y
n−1
)
qr(1 + y
n
+ ry
n−1
)
, (9.6)
y
n
−y
n+4
=
9.2. The Case γ = 0 : x
n+1
=
α+βx
n
A+Bx
n
+Cx
n−1
145
qr

y
n

p−q
qr

y
n+1
+ (y
n
−q)(y
n+1
y
n+3
+ y
n+3
+ y
n+1
+ ry
n
y
n+3
) + y
n
+ ry
2
n
−p
(1 + y
n+3
)(1 + y
n+1
+ ry
n
) + r(p + qy
n+1
)
,
(9.7)
y
n+1
−y =
(y −q)(y −y
n
) + yr(y −y
n−1
)
1 + y
n
+ ry
n−1
. (9.8)
The proofs of the following two lemmas are straightforward consequences of the Identities
(9.5)- (9.8) and will be omitted.
Lemma 9.2.2 Assume
p > q(qr + 1)
and let {y
n
}

n=−1
be a solution of Eq(9.4). Then the following statements are true:
(i) If for some N ≥ 0, y
N
>
p−q
qr
, then y
N+2
< q;
(ii) If for some N ≥ 0, y
N
=
p−q
qr
, then y
N+2
= q;
(iii) If for some N ≥ 0, y
N
<
p−q
qr
, then y
N+2
> q;
(iv) If for some N ≥ 0, q < y
N
<
p−q
qr
, then q < y
N+2
<
p−q
qr
;
(v) If for some N ≥ 0, y ≥ y
N−1
and y ≥ y
N
, then y
N+1
≥ y;
(vi) If for some N ≥ 0, y < y
N−1
and y < y
N
, then y
N+1
< y;
(vii) q < y <
p−q
qr
.
Lemma 9.2.3 Assume
q < p < q(qr + 1)
and let {y
n
}

n=−1
be a solution of Eq(9.4). Then the following statements are true:
(i) If for some N ≥ 0, y
N
<
p−q
qr
, then y
N+2
> q;
(ii) If for some N ≥ 0, y
N
=
p−q
qr
, then y
N+2
= q;
(iii) If for some N ≥ 0, y
N
>
p−q
qr
, then y
N+2
< q;
(iv) If for some N ≥ 0, y
N
>
p−q
qr
and y
N
< q then q > y
N+2
>
p−q
qr
;
(v)
p−q
qr
< y < q.
146 Chapter 9. (2, 3)-Type Equations
Lemma 9.2.4 Assume
p = q(qr + 1)
and let {y
n
}

n=−1
be a solution of Eq(9.4). Then
y
n+1
−q =
qr
1 + y
n
+ ry
n−1
(q −y
n−1
). (9.9)
Furthermore when qr < 1, then
lim
n→∞
y
n
= y. (9.10)
Proof. Identity (9.9) follows by straightforward computation. The limit (9.10) is a
consequence of the fact that in this case qr ∈ (0, ∞) and Eq(9.4) has no period-two
solution. 2
Here we present a semicycle analysis of the solutions of Eq(9.4) when p = q(1 +rq).
In this case Eq(9.4) becomes
y
n+1
=
q + rq
2
+ qy
n
1 + y
n
+ ry
n−1
, n = 0, 1, . . .
and the only equilibrium point is y = q.
Theorem 9.2.3 Suppose that p = q + rq
2
and let {y
n
}

n=−1
be a nontrivial solution of
Eq(9.4). Then this solution is oscillatory and the sum of the lengths of two consecutive
semicycles, excluding the first, is equal to four. More precisely, the following statements
are true for all k ≥ 0:
(i)
y
−1
> q and y
0
≤ q =⇒y
4k−1
> q, y
4k
≤ 1, y
4k+1
< q, and y
4k+2
≥ q;
(ii)
y
−1
< q and y
0
≥ q =⇒y
4k−1
< q, y
4k
≥ 1, y
4k+1
> q, and y
4k+2
≤ q;
(iii)
y
−1
> q and y
0
≥ q =⇒y
4k−1
> q, y
4k
≥ 1, y
4k+1
< q, and y
4k+2
≤ q;
(iv)
y
−1
< q and y
0
≤ q =⇒y
4k−1
< q, y
4k
≤ 1, y
4k+1
> q, and y
4k+2
≥ q.
Proof. The proof follows from identity (9.5). 2
9.2. The Case γ = 0 : x
n+1
=
α+βx
n
A+Bx
n
+Cx
n−1
147
9.2.4 Global Asymptotic Stability
The following lemma establishes that when p = q(qr + 1), every solution of Eq(9.4)
is eventually trapped into one of the three invariant intervals of Eq(9.4) described in
Lemma 9.2.1. More precisely, the following is true:
Lemma 9.2.5 Let I denote the interval which is defined as follows:
I =

[0,
q−1+

(q−1)
2
+4p
2
] if p ≤ q;
[
p−q
qr
, q] if q < p < q(qr + 1);
[q,
p−q
qr
] if p > q(qr + 1).
Then every solution of Eq(9.4) lies eventually in I.
Proof. Let {y
n
}

n=−1
be a solution of Eq(9.4). First assume that
p ≤ q.
Then clearly, y ∈ I. Set
b =
q −1 +

(q −1)
2
+ 4p
2
and observe that
y
n+1
−b =
(q −p)(y
n
−b) −r(p + qb)y
n−1
(1 + y
n
+ ry
n−1
)(1 + b)
, n ≥ 0.
Hence, if for some N, y
N
≤ b, then y
N+1
< b. Now assume for the sake of contradiction
that
y
n
> b for n > 0.
Then y
n
> y for n ≥ 0 and so
lim
n→∞
y
n
= y ∈ I
which is a contradiction.
Next assume that
q < p < q(qr + 1)
and, for the sake of contradiction, also assume that the solution {y
n
}

n=−1
is not even-
tually in the interval I. Then by Lemma 9.2.3, there exists N > 0 such that one of the
following three cases holds:
(i) y
N
> q, y
N+1
> q, and y
N+2
<
p−q
qr
;
(ii) y
N
> q, y
N+1
<
p−q
qr
, and y
N+2
<
p−q
qr
;
148 Chapter 9. (2, 3)-Type Equations
(iii) y
N
> q,
p−q
qr
≤ y
N+1
≤ q, and y
N+2
<
p−q
qr
.
Also observe that
if y
N
≥ q, then y
N
+ ry
2
N
−p > 0
and
if y
N

p−q
qr
, then y
N
+ ry
2
N
−p < 0.
The desired contradiction is now obtained by using the Identity (9.7) from which it
follows that for j ∈ {0, 1, 2, 3}, each subsequence {y
n+4k+j
}

k=0
with all its terms outside
the interval I converges monotonically and enters in the interval I.
The proof when
p > q(qr + 1)
is similar and will be omitted. 2
By using the monotonic character of the function
f(x, y) =
p + qx
1 + x + ry
in each of the intervals in Lemma 9.2.1, together with the appropriate convergence
Theorems 1.4.7 and 1.4.5, we can obtain some global asymptotic stability results for the
solutions of Eq(9.4). For example, the following results are true for Eq(9.4):
Theorem 9.2.4 (a) Assume that either
p ≥ q + q
2
r,
or
p <
q
1 + r
.
Then the equilibrium y of Eq(9.4) is globally asymptotically stable.
(b) Assume that either
p ≤ q
or
q < p < q + q
2
r
and that one of the following conditions is also satisfied:
(i) q ≤ 1;
(ii) r ≤ 1;
(iii) r > 1 and (q −1)
2
(r −1) ≤ 4p.
Then the equilibrium y of Eq(9.4) is globally asymptotically stable.
Proof.
9.3. The Case β = 0 : x
n+1
=
α+γx
n−1
A+Bx
n
+Cx
n−1
149
(a) The proof follows from Lemmas 9.2.1 and 9.2.4 and Theorems 1.4.7 and 1.4.2.
(b) In view of Lemma 9.2.1 we see that when y
−1
, y
0

¸
0,
q−1+

(q−1)
2
+4p
2

, then y
n

¸
0,
q−1+

(q−1)
2
+4p
2

for all n ≥ 0. It is easy to check that y ∈
¸
0,
q−1+

(q−1)
2
+4p
2

and that in the interval
¸
0,
q−1+

(q−1)
2
+4p
2

, the function f increases in x and
decreases in y.
We will employ Theorem 1.4.5 and so it remains to be shown that if
m = f(m, M) and M = f(M, m)
then M = m. This system has the form
m =
p + qm
1 + m + rM
and M =
p + qM
1 + M + rm
.
Hence (M −m)(1 −q + M + m) = 0. Now if m + M = q −1, then M = m. For
instance, this is the case if Condition (i) is satisfied. If m + M = q − 1 then m
and M satisfy the equation
(r −1)m
2
+ (r −1)(1 −q)m + p = 0.
Clearly now if Condition (ii) or (iii) is satisfied, m = M from which the result
follows.
The proof when q < p < q + q
2
r holds follows from Lemma 9.2.1 and Theorem
1.4.5.
2
9.3 The Case β = 0 : x
n+1
=
α+γx
n−1
A+Bx
n
+Cx
n−1
This is the (2, 3)-type Eq(9.2) which by the change of variables
x
n
=
A
B
y
n
reduces to the difference equation
y
n+1
=
p + qy
n−1
1 + y
n
+ ry
n−1
, n = 0, 1, . . . (9.11)
where
p =
αB
A
2
, q =
γ
A
, and r =
C
B
.
150 Chapter 9. (2, 3)-Type Equations
This equation has not been investigated yet. Here we make some simple observations
about linearized stability, invariant intervals, and the convergence of solutions in certain
regions of initial conditions.
Eq(9.11) has a unique equilibrium y which is positive and is given by
y =
q −1 +

(q −1)
2
+ 4p(r + 1)
2(r + 1)
.
By employing the linearized stability Theorem 1.1.1 we obtain the following result.
Theorem 9.3.1 (a) The equilibrium y of Eq(9.11) is locally asymptotically stable when
either
q ≤ 1
or
q > 1 and (r −1)(q −1)
2
+ 4pr
2
> 0.
(b) The equilibrium y of Eq(9.11) is unstable, and more precisely a saddle point,
when
q > 1 and (r −1)(q −1)
2
+ 4pr
2
< 0. (9.12)
9.3.1 Existence and Local Stability of Period-Two Cycles
Let
. . . , φ, ψ, φ, ψ, . . .
be a period-two cycle of Eq(9.11). Then
φ =
p + qφ
1 + ψ + rφ
and ψ =
p + qψ
1 + φ + rψ
.
It now follows after some calculation that the following result is true.
Lemma 9.3.1 Eq(9.11) has a prime period-two solution
. . . , φ, ψ, φ, ψ, . . .
if and only if (9.12) holds.
Furthermore when (9.12) holds, the period-two solution is “unique” and the values
of φ and ψ are the positive roots of the quadratic equation
t
2

q −1
r
t +
p
1 −r
= 0.
9.3. The Case β = 0 : x
n+1
=
α+γx
n−1
A+Bx
n
+Cx
n−1
151
To investigate the local stability of the two cycle
. . . , φ, ψ, φ, ψ, . . .
we set
u
n
= y
n−1
and v
n
= y
n
, for n = 0, 1, . . .
and write Eq(9.11) in the equivalent form:
u
n+1
= v
n
v
n+1
=
p+qu
n
1+v
n
+ru
n
, n = 0, 1, . . . .
Let T be the function on (0, ∞) ×(0, ∞) defined by:
T

u
v

=

v
p+qu
1+v+ru

.
Then

φ
ψ

is a fixed point of T
2
, the second iterate of T. By a simple calculation we find that
T
2

u
v

=

g(u, v)
h(u, v)

where
g(u, v) =
p + qu
1 + v + ru
and h(u, v) =
p + qv
1 + g(u, v) + rv
.
Clearly the two cycle is locally asymptotically stable when the eigenvalues of the
Jacobian matrix J
T
2, evaluated at

φ
ψ

, lie inside the unit disk.
We have
J
T
2

φ
ψ

=

¸
¸
∂g
∂u
(φ, ψ)
∂g
∂v
(φ, ψ)
∂h
∂u
(φ, ψ)
∂h
∂v
(φ, ψ)
¸

,
where
∂g
∂u
(φ, ψ) =
q −pr + qψ
(1 + ψ + rφ)
2
,
∂g
∂v
(φ, ψ) =
−p −qφ
(1 + ψ + rφ)
2
,
152 Chapter 9. (2, 3)-Type Equations
∂h
∂u
(φ, ψ) =
(p + qψ)(pr −q −qψ)
(1 + ψ + rφ)
2
(1 + φ + rψ)
2
,
∂h
∂v
(φ, ψ) =
(p + qψ)(p + qφ)
(1 + ψ + rφ)
2
(1 + φ + rψ)
2
+
q −pr + qφ
(1 + φ + rψ)
2
.
Set
S =
∂g
∂u
(φ, ψ) +
∂h
∂v
(φ, ψ)
and
D =
∂g
∂u
(φ, ψ)
∂h
∂v
(φ, ψ) −
∂g
∂v
(φ, ψ)
∂h
∂u
(φ, ψ).
Then it follows from Theorem 1.1.1 that both eigenvalues of J
T
2

φ
ψ

lie inside the
unit disk |λ| < 1, if and only if
|S| < 1 +D < 2. (9.13)
Inequality (9.13) is equivalent to the following three inequalities:
D < 1 (9.14)
S < 1 +D (9.15)
−1 −D < S. (9.16)
First we will establish Inequality (9.14). This inequality is equivalent to
(q −pr + qφ) (q −pr + qψ) −(1 + ψ + rφ)
2
(1 + φ + rψ)
2
< 0.
Observe that,
(q −pr + qφ) (q −pr + qψ) = (q −pr)
2
+ q(q −pr)(φ + ψ) + q
2
φψ =
=
−2q
2
r + q
2
r
2
+ 3qpr
2
−2qpr
3
−p
2
r
3
+ p
2
r
4
−q
3
+ rq
3
+ q
2
−pr
2
q
2
−qpr
r (−1 + r)
and that
(1 + ψ + rφ)
2
(1 + φ + rψ)
2
=

1 + φ + rψ + ψ + ψφ + rψ
2
+ rφ + φ
2
r + r
2
φψ

2
=

−pr
2
+ qr + pr −q + q
2
r

2
.
Thus
9.3. The Case β = 0 : x
n+1
=
α+γx
n−1
A+Bx
n
+Cx
n−1
153
(q −pr + qφ) (q −pr + qψ) −(1 + ψ + rφ)
2
(1 + φ + rψ)
2
=
−pr
3
q
2
+ 2q
2
r −q
2
r
2
−q
2
+ 3p
2
r
3
−2p
2
r
4
−3rq
3
+ 3qpr
3
−5qpr
2
r
2
(1 −r)
+
2qpr + 4pr
2
q
2
+ 2q
3
+ q
4
r −p
2
r
2
−2prq
2
−q
4
+ r
2
q
3
r
2
(1 −r)
=
(2r
2
p −rp + q −q
2
−qr + q
2
r) (−r
2
p + rp −q + q
2
+ qr)
r
2
(1 −r)
.
Note that
−r
2
p + rp −q + q
2
+ qr = rp(1 −r) + q(q −1) + qr > 0.
Also by using Condition (9.12) we see that
(r −1)(q −1)
2
+ 4pr
2
= q
2
r −2qr + r −q
2
+ 2q −1 + 4r
2
p < 0
and so
2r
2
p −rp + q −q
2
−qr + q
2
r < 2r
2
p −rp + q −qr + 2qr −r −2q + 1 −4r
2
p
= −2r
2
p −rp −q + qr −r + 1 = −2r
2
p −rp + (r −1) (q −1) < 0
from which the result follows.
Next we turn to Inequality (9.15). This inequality is equivalent to
(q −pr + qψ)(1 + φ + rψ)
2
+ (p + qψ)(p + qφ) + (q −pr + qφ)(1 + ψ + rφ)
2
−(q −pr + qφ) (q −pr + qψ) −(1 + ψ + rφ)
2
(1 + φ + rψ)
2
< 0.
After some lengthy calculation one can see that this inequality is a consequence of
Condition (9.12).
Finally Inequality (9.16) is equivalent to
(q −pr + qψ)(1 + φ + rψ)
2
+ (p + qψ)(p + qφ) + (q −pr + qφ)(1 + ψ + rφ)
2
+(q −pr + qφ) (q −pr + qψ) + (1 + ψ + rφ)
2
(1 + φ + rψ)
2
> 0.
After some lengthy calculation one can see that this inequality is also a consequence
of Condition (9.12).
In summary, we have established the following result:
154 Chapter 9. (2, 3)-Type Equations
Theorem 9.3.2 Eq(9.11) has a unique prime period-two solution
. . . , φ, ψ, φ, ψ, . . .
if and only if (9.12) holds.
Furthermore when (9.12) holds, this period-two solution is locally asymptotically sta-
ble.
9.3.2 Invariant Intervals
The following result, which can be established by direct calculation, gives a list of in-
variant intervals for Eq(9.11).
Lemma 9.3.2 Eq(9.11) possesses the following invariant intervals:
(a)

0,
q −1 +

(q −1)
2
+ 4pr
2r
¸
¸
when pr ≤ q;
(b)
¸
pr −q
q
,
q
r
¸
when 1 < q < pr < q +
q
2
r

q
r
;
(c)
¸
q
r
,
pr −q
q
¸
when pr ≥ q +
q
2
r
.
Proof.
(a) Set
g(x) =
p + qx
1 + rx
and b =
q −1 +

(q −1)
2
+ 4pr
2r
and observe that g is an increasing function and g(b) ≤ b. Using Eq(9.11) we see
that when y
k−1
, y
k
∈ [0, b], then
y
k+1
=
p + qy
k−1
1 + y
k
+ qy
k−1

p + qy
k−1
1 + ry
k−1
= g(y
k−1
) ≤ g(b) ≤ b.
The proof follows by induction.
9.3. The Case β = 0 : x
n+1
=
α+γx
n−1
A+Bx
n
+Cx
n−1
155
(b) It is clear that the function
f(x, y) =
p + qy
1 + x + ry
is increasing in y for x >
pr−q
q
. Using Eq(9.11) we see that when y
k−1
, y
k

pr−q
q
,
q
r

,
then
y
k+1
=
p + qy
k−1
1 + y
k
+ ry
k−1
= f(y
k
, y
k−1
) ≥ f

q
r
,
pr −q
q


pr −q
q
.
Now by using the monotonic character of the function f(x, y) and the condition
q < pr < q +
q
2
r
we obtain
y
k+1
=
p + qy
k−1
1 + y
k
+ ry
k−1
= f(y
k
, y
k−1
) ≤ f

pr −q
q
,
q
r

=
q
r
.
The proof follows by induction.
(c) It is clear that the function
f(x, y) =
p + qy
1 + x + ry
is decreasing in y for x <
pr−q
q
. Using Eq(9.11) we see that when y
k−1
, y
k

q
r
,
pr−q
q

, then
y
k+1
=
p + qy
k−1
1 + y
k
+ qy
k−1
= f(y
k
, y
k−1
) ≥ f

pr −q
q
,
pr −q
q

=
q
r
.
By using the fact that f(x, y) is decreasing in both arguments and the condition
pr > q +
q
r
2
we obtain
y
k+1
=
p + qy
k−1
1 + y
k
+ qy
k−1
= f(y
k
, y
k−1
) ≤ f

q
r
,
q
r

=
pr + q
2
r + (r + 1)q
<
pr −q
q
.
The proof follows by induction.
2
156 Chapter 9. (2, 3)-Type Equations
9.3.3 Convergence of Solutions
By using the monotonic character of the function
f(x, y) =
p + qy
1 + x + ry
in each of the intervals in Lemma 9.3.2, together with the appropriate convergence
theorem (from among Theorems 1.4.5-1.4.8) we can obtain some convergence results for
the solutions with initial conditions in the invariant intervals. For example, the following
results are true for Eq(9.11):
Theorem 9.3.3 Assume that
pr ≤ q
and that one of the following conditions is also satisfied:
(i) q ≤ 1;
(ii) r ≥ 1;
(iii) r < 1 and (r −1)(q −1)
2
+ 4pr
2
≥ 0.
Then every solution of Eq(9.11) with initial conditions in the invariant interval

0,
q −1 +

(q −1)
2
+ 4pr
2r
¸
¸
converges to the equilibrium y.
Proof. In view of Lemma 9.3.2 we see that when y
−1
, y
0

¸
0,
q−1+

(q−1)
2
+4pr
2r

, then
y
n

¸
0,
q−1+

(q−1)
2
+4pr
2r

for all n ≥ 0. It is easy to check that y ∈
¸
0,
q−1+

(q−1)
2
+4pr
2r

and that in the interval
¸
0,
q−1+

(q−1)
2
+4pr
2r

the function f decreases in x and increases
in y. The result now follows by employing Theorem 1.4.6.
2
Theorem 9.3.4 (a) Assume that 1 < q < pr < q+
q
2
r

q
r
and that one of the following
conditions is also satisfied:
(i) r ≥ 1;
(ii) r < 1 and (r −1)(q −1)
2
+ 4pr
2
≥ 0.
9.3. The Case β = 0 : x
n+1
=
α+γx
n−1
A+Bx
n
+Cx
n−1
157
Then every solution of Eq(9.11) with initial conditions in the invariant interval
¸
pr −q
q
,
q
r
¸
converges to the equilibrium y.
(b) Assume that pr > q +
q
2
r
. Then every solution of Eq(9.11) with initial conditions
in the invariant interval
¸
q
r
,
pr −q
q
¸
converges to the equilibrium y.
Proof.
(a) In view of Lemma 9.3.2 we conclude that y
n

pr−q
q
,
q
r

for all n ≥ 0. It is easy to
check that y ∈

pr−q
q
,
q
r

and that the function f(x, y) decreases in x and increases
in y. The result now follows by employing Theorem 1.4.6.
(b) In view of Lemma 9.3.2 we conclude that y
n

q
r
,
pr−q
q

for all n ≥ 0. It is easy to
check that y ∈

q
r
,
pr−q
q

and that in the interval

q
r
,
pr−q
q

the function f decreases
in both x and in y. The result now follows by employing Theorem 1.4.7.
2
9.3.4 Global Stability
By applying Theorem 1.4.3 to Eq(9.11), we obtain the following global asymptotic sta-
bility result.
Theorem 9.3.5 Assume
r ≥ 1 and pr ≤ q.
Then the positive equilibrium of Eq(9.11) is globally asymptotically stable.
9.3.5 Semicycle Analysis When pr = q +
q
2
r
Here we present a semicycle analysis of the solutions of Eq(9.11) when pr = q +
q
2
r
. In
this case Eq(9.11) becomes
y
n+1
=
qr + q
2
+ qr
2
y
n−1
r
2
+ r
2
y
n
+ r
3
y
n−1
, n = 0, 1, . . .
and the only positive equilibrium is y =
q
r
.
158 Chapter 9. (2, 3)-Type Equations
Theorem 9.3.6 Suppose that pr = q +
q
2
r
and let {y
n
}

n=−1
be a nontrivial solution of
Eq(9.11). Then after the first semicycle, the solution oscillates about the equilibrium y
with semicycles of length one.
Proof. The proof follows from the relation
y
n+1

q
r
= (
q
r
−y
n
)
qr
2
r
3
(1 + y
n
+ ry
n−1
)
.
2
9.4 The Case α = 0 : x
n+1
=
βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
This is the (2, 3)-type Eq(9.3) which by the change of variables
x
n
=
γ
C
y
n
,
reduces to the difference equation
y
n+1
=
py
n
+ y
n−1
r + qy
n
+ y
n−1
, n = 0, 1, . . . (9.17)
where
p =
β
γ
, q =
B
C
, and r =
A
γ
.
This equation has not been investigated yet. Here we make some simple observations
about linearized stability, invariant intervals, and the convergence of solutions in certain
regions of initial conditions.
The equilibrium points of Eq(9.17) are the nonnegative solutions of the equation
y =
(p + 1)y
r + (1 + q)y
.
Hence zero is always an equilibrium point and when
p + 1 > r, (9.18)
Eq(9.17) also possesses the unique positive equilibrium
y =
p + 1 −r
q + 1
.
The following theorem is a consequence of Theorem 1.1.1 and Theorem 1.3.1.
9.4. The Case α = 0 : x
n+1
=
βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
159
Theorem 9.4.1 (a) Assume that
p + 1 ≤ r.
Then the zero equilibrium of Eq(9.17) is globally asymptotically stable.
(b) Assume that
p + 1 > r.
Then the zero equilibrium of Eq(9.17) is unstable. Furthermore the zero equilibrium
is a saddle point when
1 −p < r < 1 + p
and a repeller when
r < 1 −p.
The linearized equation associated with Eq(9.17) about its positive equilibrium y is
z
n+1

p −q + qr
(p + 1)(q + 1)
z
n

q −p + r
(p + 1)(q + 1)
z
n−1
= 0, n = 0, 1, . . . .
The following result is a consequence of Theorem 1.1.1.
Theorem 9.4.2 Assume that (9.18) holds. Then the positive equilibrium of Eq(9.17)
is locally asymptotically stable when
q + r < 3p + 1 + qr + pq, (9.19)
and unstable (a saddle point) when
q + r > 3p + 1 + qr + pq. (9.20)
Concerning prime period-two solutions for Eq(9.17), it follows from Section 2.5 that the
following result is true.
Theorem 9.4.3 Eq(9.17) has a prime period-two solution
. . . φ, ψ, φ, ψ, . . .
if and only if (9.20) holds. Furthermore when (9.20) holds there is a unique period-two
solution and the values of φ and ψ are the positive roots of the quadratic equation
t
2
−(1 −p −r)t +
p(1 −p −r)
q −1
= 0.
160 Chapter 9. (2, 3)-Type Equations
9.4.1 Invariant Intervals
The following result, which can be established by direct calculation, gives a list of in-
variant intervals for Eq(9.17).
Lemma 9.4.1 Eq(9.17) possesses the following invariant intervals:
(a)
¸
0,
p + q −r
q + 1
¸
when p = q
2
and q
2
+ q > r
and
[0, K] when p = q
2
and q
2
+ q ≤ r,
where K > 0 is an arbitrary number;
(b)
¸
0,
qr
p −q
2
¸
when q
2
< p ≤ q
2
+ r and p + q > r,
[0, K] when q
2
< p ≤ q
2
+ r and p + q ≤ r,
where K > 0 is an arbitrary number
and
¸
qr
p −q
2
, K
1
¸
when p > q
2
+ r
where
K
1
=
(q −r)(p −q
2
) −qr +

((r −q)(p −q
2
) + qr)
2
+ 4q
3
r(p −q
2
)
2q(p −q
2
)
;
(c)
¸
0,
pr
q
2
−p
¸
when q
2
−qr ≤ p < q
2
and p + q > r,
[0, K] when q
2
−qr ≤ p < q
2
and p + q ≤ r,
where K > 0 is an arbitrary number
and
¸
pr
q
2
−p
, K
2
¸
when q
2
> p + qr,
where
K
2
=
(q −r)(q
2
−p) +

((q −r)(q
2
−p))
2
+ 4p
3
qr
2
2pqr
.
9.4. The Case α = 0 : x
n+1
=
βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
161
Proof.
(a) It is easy to see that when p = q
2
the function
f(x, y) =
px + qy
r + qx + y
is increasing in both arguments. Therefore for any K > 0,
0 ≤ f(x, y) ≤ f(K, K).
Now observe that the inequality f(K, K) ≤ K is equivalent to q
2
+q−r ≤ (q+1)K.
When q
2
+q ≤ r, this inequality is satisfied for any K > 0, while when q
2
+q > r
the inequality is true when
K ≥
p + q −r
q + 1
.
(b) It is clear that the function f(x, y) is increasing in both arguments for x <
qr
p−q
2
,
and it is increasing in x and decreasing in y for x ≥
qr
p−q
2
.
First assume that q
2
< p ≤ q
2
+ r .
Hence for y
k−1
, y
k

0,
qr
p−q
2

we obtain
y
k+1
= f(y
k
, y
k−1
) ≤ f(K, K) = (p + q)K
1
r + (q + 1)K
≤ K,
where we set K =
qr
p−q
2
. The last inequality is always satisfied when p +q ≤ r and
when p + q > r, it is satisfied under the condition r + q
2
> p.
Second assume that p > q
2
+ r.
Then for y
k−1
, y
k

qr
p−q
2
, K
1

we can see that
f

qr
p −q
2
, K
1

≤ y
k+1
= f(y
k
, y
k−1
) ≤ f(K
1
,
qr
p −q
2
) ≤ K
1
.
(c) It is clear that the function f(x, y) is increasing in both arguments for y <
pr
q
2
−p
,
and it is increasing in y and decreasing in x for y ≥
pr
q
2
−p
.
First assume that q
2
−qr ≤ p < q
2
.
Then one can see that for y
k−1
, y
k

0,
pr
q
2
−p

we obtain
y
k+1
= f(y
k
, y
k−1
) ≤ f(K, K) = (p + q)K
1
r + (q + 1)K
≤ K,
with K =
pr
q
2
−p
.
Second assume that p < q
2
−qr. Then for y
k−1
, y
k

pr
q
2
−p
, K
2

one can see that
f

pr
q
2
−p
, K
2

≤ y
k+1
= f(y
k
, y
k−1
) ≤ f

K
2
,
pr
q
2
−p

≤ K
2
.
2
162 Chapter 9. (2, 3)-Type Equations
9.4.2 Convergence of Solutions
By using the monotonic character of the function
f(x, y) =
px + qy
r + qx + y
in each of the intervals in Lemma 9.4.1, together with the appropriate convergence
theorem (from among Theorems 1.4.5-1.4.8), we can obtain some convergence results
for the solutions with initial conditions in the invariant intervals. For example, the
following results are true for Eq(9.17):
Theorem 9.4.4 (a) Assume that p = q
2
and q
2
+ q > r. Then every solution of
Eq(9.17) with initial conditions in the invariant interval
¸
0,
q
2
+ q −r
q + 1
¸
converges to the equilibrium y.
Assume that p = q
2
and q
2
+ q ≤ r. Then every solution of Eq(9.17) with initial
conditions in the interval [0, K], for any K > 0, converges to the equilibrium y.
(b) Assume that q
2
+r ≥ p > q
2
and p +q > r. Then every solution of Eq(9.17) with
initial conditions in the invariant interval
¸
0,
qr
p −q
2
¸
converges to the equilibrium y.
Assume that q
2
+r ≥ p > q
2
and p +q ≤ r. Then every solution of Eq(9.17) with
initial conditions in the interval [0, K], for any K > 0, converges to the equilibrium
y.
(c) Assume that q
2
−qr ≥ p < q
2
and p +q > r. Then every solution of Eq(9.17) with
initial conditions in the invariant interval
¸
0,
pr
q
2
−p
¸
converges to the equilibrium y.
Assume that q
2
−qr ≤ p < q
2
and p +q ≤ r. Then every solution of Eq(9.17) with
initial conditions in the interval [0, K], for any K > 0, converges to the equilibrium
y.
Proof. The proof is an immediate consequence of Lemma 9.4.1 and Theorem 1.4.8.
2
9.5. Open Problems and Conjectures 163
Theorem 9.4.5 Assume that
p > q
2
+ r
and that one of the following conditions is also satisfied:
(i) p ≤ q + r;
(ii) q ≥ 1;
(iii) q < 1 and p ≤
3q
2
+q−qr+r
1−q
.
Then every solution of Eq(9.11) with initial conditions in the invariant interval

qr
p−q
2
, K
1

converges to the equilibrium y.
Proof. The proof is a consequence of Lemma 9.4.1 and Theorem 1.4.5.
2
Theorem 9.4.6 Assume that q
2
> p + qr and that Condition (9.20) is not satis-
fied. Then every solution of Eq(9.11) with initial conditions in the invariant interval

qr
p−q
2
, K
2

converges to the equilibrium y.
Proof. The proof is an immediate consequence of Lemma 9.4.1 and Theorem 1.4.6.
2
9.5 Open Problems and Conjectures
Conjecture 9.5.1 Assume
α, β, A, B, C ∈ (0, ∞).
Show that every positive solution of Eq(9.1) has a finite limit.
Conjecture 9.5.2 Assume
α, γ, A, B, C ∈ (0, ∞).
Show that every positive solution of Eq(9.2) is bounded.
Conjecture 9.5.3 Assume
β, γ, A, B, C ∈ (0, ∞).
Show that every positive solution of Eq(9.3) is bounded.
164 Chapter 9. (2, 3)-Type Equations
Conjecture 9.5.4 Assume
α, γ, A, B, C ∈ (0, ∞)
and suppose that Eq(9.2) has no prime period-two solutions. Show that the positive
equilibrium of Eq(9.2) is globally asymptotically stable.
Conjecture 9.5.5 Assume that x
−1
, x
0
∈ [0, ∞) with x
−1
+ x
0
> 0 and that
β, γ, A, B, C ∈ (0, ∞).
Suppose that Eq(9.3) has no prime period-two solutions. Show that the positive equilib-
rium of Eq(9.3) is globally asymptotically stable.
Open Problem 9.5.1 Determine the set G of all initial points (x
−1
, x
0
) ∈ R × R
through which the equation
x
n+1
=
x
n
+ x
n−1
1 + x
n
+ x
n−1
is well defined for all n ≥ 0, and for these initial points investigate the long-term behavior
of the solutions {x
n
}

n=−1
. Extend and generalize.
Open Problem 9.5.2 Determine the set G of all initial points (x
−1
, x
0
) ∈ R × R
through which the equation
x
n+1
=
1 + x
n
1 + x
n
+ x
n−1
is well defined for all n ≥ 0, and for these initial points investigate the long-term behavior
of the solutions {x
n
}

n=−1
. Extend and generalize.
Open Problem 9.5.3 Determine the set G of all initial points (x
−1
, x
0
) ∈ R × R
through which the equation
x
n+1
=
1 + x
n−1
1 + x
n
+ x
n−1
is well defined for all n ≥ 0, and for these initial points investigate the long-term behavior
of the solutions {x
n
}

n=−1
. Extend and generalize.
9.5. Open Problems and Conjectures 165
Open Problem 9.5.4 Assume that {p
n
}

n=0
, {q
n
}

n=0
, and {r
n
}

n=0
are convergent se-
quences of nonnegative real numbers with finite limits,
p = lim
n→∞
p
n
, q = lim
n→∞
q
n
, and r = lim
n→∞
r
n
.
Investigate the asymptotic behavior and the periodic nature of all positive solutions of
each of the following three difference equations:
y
n+1
=
p
n
+ q
n
y
n
1 + y
n
+ r
n
y
n−1
, n = 0, 1, . . . (9.21)
y
n+1
=
p
n
+ q
n
y
n−1
1 + y
n
+ r
n
y
n−1
, n = 0, 1, . . . (9.22)
y
n+1
=
p
n
y
n
+ y
n−1
r
n
+ q
n
y
n
+ y
n−1
, n = 0, 1, . . . . (9.23)
Open Problem 9.5.5 Assume that {p
n
}

n=0
, {q
n
}

n=0
, and {r
n
}

n=0
are period-two se-
quences of nonnegative real numbers. Investigate the global character of all positive
solutions of Eqs(9.21)-(9.23). Extend and generalize.
Open Problem 9.5.6 Assume that (9.12) holds. Investigate the basin of attraction of
the two cycle
. . . , φ, ψ, φ, ψ, . . .
of Eq(9.11).
Open Problem 9.5.7 Assume that (9.20) holds. Investigate the basin of attraction of
the two cycle
. . . , φ, ψ, φ, ψ, . . .
of Eq(9.17).
Conjecture 9.5.6 Assume that (9.20) holds. Show that the period-two solution
. . . , φ, ψ, φ, ψ, . . .
of Eq(9.17) is locally asymptotically stable.
166 Chapter 9. (2, 3)-Type Equations
Open Problem 9.5.8 Assume that
p, q, r ∈ [0, ∞) and k ∈ {2, 3, . . .}.
Investigate the global behavior of all positive solutions of each of the following difference
equations:
y
n+1
=
p + qy
n
1 + y
n
+ ry
n−k
, n = 0, 1, . . . (9.24)
y
n+1
=
p + qy
n−k
1 + y
n
+ ry
n−k
, n = 0, 1, . . . (9.25)
y
n+1
=
py
n
+ y
n−k
r + qy
n
+ y
n−k
, n = 0, 1, . . . . (9.26)
Chapter 10
(3, 2)-Type Equations
10.1 Introduction
Eq(1) contains the following three equations of the (3, 2)-type:
x
n+1
=
α + βx
n
+ γx
n−1
A + Bx
n
, n = 0, 1, . . . (10.1)
x
n+1
=
α + βx
n
+ γx
n−1
A + Cx
n−1
, n = 0, 1, . . . (10.2)
and
x
n+1
=
α + βx
n
+ γx
n−1
Bx
n
+ Cx
n−1
, n = 0, 1, . . . . (10.3)
Please recall our classification convention in which all parameters that appear in these
equations are positive, the initial conditions are nonnegative, and the denominators are
always positive.
10.2 The Case C = 0 : x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
This is the (3, 2)-type Eq(10.1) which by the change of variables
x
n
=
A
B
y
n
reduces to the difference equation
y
n+1
=
p + qy
n
+ ry
n−1
1 + y
n
, n = 0, 1, . . . (10.4)
where
p =
αB
A
2
, q =
β
A
, and r =
γ
A
.
167
168 Chapter 10. (3, 2)-Type Equations
(Eq(10.4) was investigated in [29].)
The linearized equation of Eq(10.4) about its unique equilibrium
y =
q + r −1 +

(q + r −1)
2
+ 4p
2
is
z
n+1

q −p −ry
(1 + y)
2
z
n

r
1 + y
z
n−1
= 0, n = 0, 1, . . . .
By applying the linearized stability Theorem 1.1.1 we see that y is locally asymptot-
ically stable when
r < 1 + q
and is an unstable (saddle point) equilibrium when
r > 1 + q.
When
r = q + 1
Eq(10.4) possesses prime period-two solutions. See Sections 2.5 and 2.7. The main
result in this section is that the solutions of Eq(10.4) exhibit the following trichotomy
character.
Theorem 10.2.1 (a) Assume that
r = q + 1. (10.5)
Then every solution of Eq(10.4) converges to a period-two solution.
(b) Assume that
r < q + 1. (10.6)
Then the equilibrium of Eq(10.4) is globally asymptotically stable.
(c) Assume that
r > q + 1. (10.7)
Then Eq(10.4) possesses unbounded solutions.
The proof of part (a) of Theorem 10.2.1 was given, for a more general equation, in
Section 2.7. The proofs of parts (b) and (c) of Theorem 10.2.1 are given in Sections
10.2.1 and 10.2.2 below.
10.2. The Case C = 0 : x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
169
10.2.1 Proof of Part (b) of Theorem 10.2.1
Here we assume that Eq(10.6) holds and we prove that the equilibrium of Eq(10.4) is
globally asymptotically stable. We know already that when Eq(10.6) holds the equilib-
rium is locally asymptotically stable so it remains to be shown that every solution of
Eq(10.4) is attracted to the equilibrium.
Observe that
y
n+1
= q +
(p −q) + ry
n−1
1 + y
n
, n = 0, 1, . . .
and so when
p ≥ q (10.8)
we have
y
n
≥ q for n ≥ 1.
Therefore the change of variables
y
n
= q + z
n
transforms Eq(10.4) to the difference equation
z
n+1
=
(p + rq −q) + rz
n−1
(1 + q) + z
n
, n = 1, . . . (10.9)
which is of the form of the (2, 2)-type Eq(6.23) which was investigated in Section 6.5
and therefore the proof is complete when (10.8) holds.
So in the remaining part of the proof we assume that
p < q.
Now observe that the solutions of Eq(10.4) satisfy the following identity
y
n+1
−q =
r
1 + y
n

y
n−1

q −p
r

for n ≥ 0. (10.10)
Note that when
q =
q −p
r
that is when
p + qr −q = 0 (10.11)
the Identity (10.10) reduces to
y
n+1
−q =
r
1 + y
n
(y
n
−q) for n ≥ 0. (10.12)
Also in this case
0 < r = 1 −
p
q
< 1
170 Chapter 10. (3, 2)-Type Equations
and the equilibrium of Eq(10.4) is equal to q.
It follows from (10.12) that
|y
n+1
−q| < r|y
n+1
−q|
and so
lim
n→∞
y
n
= q
which completes the proof when (10.11) holds.
Still to be considered are cases where
p + qr −q > 0 (10.13)
and
p + qr −q < 0. (10.14)
First assume that (10.13) holds. To complete the proof in this case it is sufficient to
show that eventually
y
n
≥ q (10.15)
and then use the known result for Eq(10.9). To this end, note that now
q −p
r
< q
and employ (10.10). If (10.15) is not eventually true then either
y
2n
<
q −p
r
for n ≥ 0 (10.16)
or
y
2n−1
<
q −p
r
for n ≥ 0. (10.17)
This is because when
y
N−1
>
q −p
r
for some N ≥ 0,
then it follows from (10.10) that
y
N+1+2K
> q for all K ≥ 0.
We will assume that (10.16) holds. The case where (10.17) holds is similar and will be
omitted. Then we can see from (10.10) that
y
2n+2
−q =
r
1 + y
n+1

y
2n

q −p
r

> r

y
2n

q −p
r

and so
y
2n+2
> ry
2n
+ p for n ≥ 0. (10.18)
10.2. The Case C = 0 : x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
171
This is impossible when r ≥ 1, because by (10.16) the solution is bounded. On the other
hand, when r < 1, it follows from (10.16) and (10.18) that
q −p
r
> y
2n+2
> r
n+1
y
0
+ r
n
p + . . . + rp + p
and so
q −p
r

p
1 −r
which contradicts (10.13) and completes the proof when (10.13) holds.
Next assume that (10.14) holds. Now we claim that every solution of Eq(10.4) lies
eventually in the interval

0,
q−p
r

. Once we establish this result, the proof that the
equilibrium is a global attractor of all solutions would be a consequence of Theorem
1.4.8 and the fact that in this case the function
f(x, y) =
p + qx + ry
1 + x
is increasing in both arguments. To this end assume for the sake of contradiction that
the solution is not eventually in the interval

0,
q−p
r

. Then one can see from (10.10) and
the fact that now
q <
q −p
r
,
that either
y
2n
>
q −p
r
for n ≥ 0 (10.19)
or
y
2n−1
>
q −p
r
for n ≥ 0. (10.20)
We will assume that (10.19) holds. The case where (10.20) holds is similar and will be
omitted. Then from (10.10) we see that
y
2n+2
−q =
r
1 + y
2n+1

y
2n

q −p
r

< r

y
2n

q −p
r

and so
y
2n+2
< ry
2n
+ p for n ≥ 0.
Note that in this case r < 1 and so
q −p
r
< y
2n+2
< r
n+1
y
0
+ r
n
p + . . . + rp + p
which implies that
q −p
r

p
1 −r
.
This contradicts (10.14) and completes the proof of part (b) of Theorem 10.2.1.
172 Chapter 10. (3, 2)-Type Equations
10.2.2 Proof of Part (c) of Theorem 10.2.1
Here we assume that (10.7) holds and show that Eq(10.4) possesses unbounded solutions.
To this end observe that
y
n+1
= q +
p + r(y
n−1

q
r
)
1 + y
n
for n ≥ 0
and so when (10.7) holds the interval [q, ∞) is invariant. That is when
y
−1
, y
0
∈ [q, ∞),
then
y
n
≥ q for n ≥ 0.
Now the change of variables
y
n
= q + z
n
transforms Eq(10.4) to Eq(10.9) and the conclusion of part (c) follows from Section 6.5
for appropriate initial conditions y
−1
and y
0
. More specifically the solutions of Eq(10.4)
with initial conditions such that
q ≤ y
−1
< r −1 and y
0
> r
−1
+
p + q(r −1)
r −1 −q
have the property that the subsequences of even terms converge to ∞ while the subse-
quences of odd terms converge to q.
It should be mentioned that when (10.7) holds, the equilibrium of Eq(10.4) is a
saddle point and so by the stable manifold theorem, in addition to unbounded solu-
tions, Eq(10.4) possesses bounded solutions which in fact converge to the equilibrium of
Eq(10.4).
10.3 The Case B = 0 : x
n+1
=
α+βx
n
+γx
n−1
A+Cx
n−1
This is the (3, 2)-type Eq(10.2) which by the change of variables
x
n
=
A
C
y
n
reduces to the difference equation
y
n+1
=
p + qy
n
+ ry
n−1
1 + y
n−1
, n = 0, 1, . . . (10.21)
where
p =
αC
A
2
, q =
β
A
, and r =
γ
A
.
10.3. The Case B = 0 : x
n+1
=
α+βx
n
+γx
n−1
A+Cx
n−1
173
This equation has not been investigated yet. Here we make some simple observations
about linearized stability, invariant intervals, and convergence of solutions in certain
regions of initial conditions.
The linearized equation of Eq(10.21) about its positive equilibrium
y =
q + r −1 +

(q + r −1)
2
+ 4p
2
is
z
n+1

q
1 + y
z
n
+
p −r + qy
(1 + y)
2
z
n
= 0, n = 0, 1, . . . .
By applying Theorem 1.1.1 we see that y is locally asymptotically stable for all values
of the parameters p, q, and r.
10.3.1 Invariant Intervals
The following result, which can be established by direct calculation, gives a list of in-
variant intervals for Eq(10.21).
Recall that I is an invariant interval, if whenever,
y
N
, y
N+1
∈ I for some integer N ≥ 0,
then
y
n
∈ I for n ≥ N.
Lemma 10.3.1 Eq(10.21) possesses the following invariant intervals:
(a)
¸
0,
p
1 −q
¸
when q < 1 and p ≥ r;
(b)
¸
0,
r −p
q
¸
when q < 1 and p ≥ r(1 −q).
10.3.2 Global Stability When p ≥ r
Here we discuss the behavior of the solutions of Eq(10.21) when p ≥ r.
Observe that
y
n+1
= r +
(p −r) + qy
n
1 + y
n−1
, n = 0, 1, . . .
and so when
p ≥ r
174 Chapter 10. (3, 2)-Type Equations
y
n
≥ r for n ≥ 1.
Therefore the change of variables
y
n
= r + z
n
transforms Eq(10.21) to the difference equation
z
n+1
=
p + qr −r + qz
n
(1 + r) + z
n−1
, n = 1, 2, . . .
which is of the form of the (2, 2)-type Eq(6.2) which was investigated in Section 6.3.
The following result is now a consequence of the above observations and the results
in Section 6.3.
Theorem 10.3.1 Assume
p ≥ r.
Then the following statements are true.
(a) Every positive solution of Eq(10.21) is bounded.
(b) The positive equilibrium of Eq(10.21) is globally asymptotically stable if one of
the following conditions holds:
(i) q < 1;
(ii)
q ≥ 1
and
either p + qr −r ≤ q or q < p + qr −r ≤ 2(q + 1).
10.3.3 Convergence of Solutions
Here we discuss the behavior of the solutions of Eq(10.21) when p < r.
Theorem 10.3.2 (a) Assume that
q < 1 and r ≥
p
1 −q
.
Then every solution of Eq(10.21) with initial conditions in the invariant interval
¸
0,
r −p
q
¸
converges to the equilibrium y.
10.4. The Case A = 0 : x
n+1
=
α+βx
n
+γx
n−1
Bx
n
+Cx
n−1
175
(b) Suppose that either
0 ≤ q −1 < r and p < r
or
q < 1 and p < r <
p
1 −q
.
Set
K =
pq + r
2
−pr
q + r −p −q
2
.
Then every solution of Eq(10.21) with initial conditions in the invariant interval
¸
r −p
q
, K
¸
converges to the equilibrium y.
Proof.
(a) In view of Lemma 10.3.1 we see that when y
−1
, y
0

0,
r−p
q

, then y
n

0,
r−p
q

for all n ≥ 0. It is easy to check that y ∈

0,
r−p
q

and that in the interval

0,
r−p
q

the function f(x, y) increases in both x and y. The result is now a consequence of
Theorem 1.4.8.
(b) In view of Lemma 10.3.1 we see that when y
−1
, y
0

r−p
q
, K

, then y
n

r−p
q
, K

for all n ≥ 0. It is easy to check that y ∈

r−p
q
, K

and that in the interval

r−p
q
, K

the function f(x, y) increases in x and decreases in y. The result is now
a consequence of Theorem 1.4.5.
2
10.4 The Case A = 0 : x
n+1
=
α+βx
n
+γx
n−1
Bx
n
+Cx
n−1
This is the (3, 2)-type Eq(10.3) which by the change of variables
x
n
=
γ
C
y
n
reduces to the difference equation
y
n+1
=
r + py
n
+ y
n−1
qy
n
+ y
n−1
, n = 0, 1, . . . (10.22)
where
p =
β
γ
, q =
B
C
, and r =
αC
γ
2
.
176 Chapter 10. (3, 2)-Type Equations
This equation has not been investigated yet. Here we make some simple observations
about linearized stability, invariant intervals, and the convergence of solutions in certain
regions of initial conditions.
The linearized equation of Eq(10.22) about its positive equilibrium
y =
(1 + p) +

(1 + p)
2
+ 4r(1 + q)
2(1 + q)
is
z
n+1

(p −q)y −qr
(q + 1)(r + (p + 1)y)
z
n
+
(p −q)y + r
(q + 1)(r + (p + 1)y)
z
n−1
= 0, n = 0, 1, . . . .
By applying Theorem 1.1.1 we see that y is locally asymptotically stable when either
q ≤ pq + 1 + 3p (10.23)
or
q > pq + 1 + 3p and F

2r
q −pq −1 −3p

> 0
where
F(u) = (q + 1)u
2
−(p + 1)u −r.
The second condition reduces to
q > pq + 1 + 3p. (10.24)
It is interesting to note that the above condition for the local asymptotic stability of y
requires the assumption that r > 0. When r = 0, as in Section 6.9, the inequality in
(10.23) is strict and Condition (10.24) is void.
Concerning prime period-two solution, it follows from Section 2.5 that a period-two
solution exists if and only if
p < 1, q > 1 and 4r < (1 −p)(q −pq −3p −1). (10.25)
Furthermore in this case the prime period-two solution
. . . , φ, ψ, φ, ψ, . . .
is unique and the values φ and ψ are the positive roots of the quadratic equation
t
2
−(1 −p)t +
r + p(1 −p)
q −1
= 0.
10.4. The Case A = 0 : x
n+1
=
α+βx
n
+γx
n−1
Bx
n
+Cx
n−1
177
10.4.1 Invariant Intervals
Here we present some results about invariant intervals for Eq(10.21). We consider the
cases where p = q, p > q, and p < q.
Lemma 10.4.1 Eq(10.21) possesses the following invariant intervals:
(a)
[a, b] when p = q and
a and b are positive numbers such that
r + (p + 1)a ≤ (q + 1)ab ≤ r + (p + 1)b. (10.26)
(b)
¸
p
q
,
qr
p −q
¸
when p > q and
p
2
−pq
q
2
< r;
¸
qr
p −q
,
p
q
¸
when p > q and
p
2
−pq
q
2
> r.
(c)
¸
1,
r
q −p
¸
when p < q < p + r;
¸
r
q −p
, 1
¸
when q > p + r.
Proof.
(a) It is easy to see that when p = q the function
f(x, y) =
r + px + y
qx + y
is decreasing in both arguments. Hence
a ≤
r + (p + 1)b
(q + 1)b
= f(b, b) ≤ f(x, y) ≤ f(a, a) =
r + (p + 1)a
(q + 1)a
≤ b.
(b) Clearly the function f(x, y) is decreasing in both arguments when y <
qr
p−q
, and it
is increasing in x and decreasing in y for y ≥
qr
p−q
.
First assume that p > q and
p
2
−pq
q
2
< r.
178 Chapter 10. (3, 2)-Type Equations
Then for x, y ∈

p
q
,
qr
p−q

we obtain
p
q
= f

qr
p −q
,
qr
p −q

≤ f(x, y) ≤ f

p
q
,
p
q

=
qr + p(p + 1)
p(q + 1)

qr
p −q
.
The inequalities
qr + p(p + 1)
p(q + 1)

qr
p −q
and
p
q
<
qr
p −q
are equivalent to the inequality
p
2
−pq
q
2
< r.
Next assume that p > q and
p
2
−pq
q
2
> r.
For x, y ∈

qr
p−q
,
p
q

, we obtain
(qr + p)(p −q) + pq
2
r
q
3
r + p(p −q)
= f

qr
p −q
,
p
q

≤ f(x, y) ≤ f

p
q
,
qr
p −q

=
p
q
.
The inequalities
(qr + p)(p −q) + pq
2
r
q
3
r + p(p −q)

qr
p −q
and
p
q
>
qr
p −q
follow from the inequality
p
2
−pq
q
2
> r.
(c) It is clear that the function f(x, y) is decreasing in both arguments for x <
r
q−p
,
and it is increasing in y and decreasing in x for x ≥
r
q−p
.
First, assume that p < q < p + r.
Using the decreasing character of f, we obtain
1 = f

r
q −p
,
r
q −p

≤ f(x, y) ≤ f(1, 1) =
r + p + 1
q + 1

r
q −p
,
The inequalities
1 <
r
q −p
and
r + p + 1
q + 1
<
r
q −p
are equivalent to the inequality q < p + r.
Next assume that q > p + r.
Using the decreasing character of f in x and the increasing character of f in y, we
obtain
(p + r)(q −p) + r
q(q −p) + r
= f

1,
r
q −p

≤ f(x, y) ≤ f

r
p −q
, 1

= 1.
10.4. The Case A = 0 : x
n+1
=
α+βx
n
+γx
n−1
Bx
n
+Cx
n−1
179
The inequalities
(p + r)(q −p) + r
q(q −p) + r

r
q −p
and
r
q −p
< 1
follow from the inequality q > p + r.
2
10.4.2 Convergence of Solutions
Here we obtain some convergence results for Eq(10.22).
Theorem 10.4.1 (a) Assume that p = q. Then every solution of Eq(10.22) with
initial conditions in the invariant interval [a, b], where 0 < a < b satisfy (10.26)
converges to the equilibrium y.
(b) Assume that p > q and r >
p
2
−pq
q
2
. Then every solution of Eq(10.22) with initial
conditions in the invariant interval

p
q
,
qr
p−q

converges to the equilibrium y.
(c) Assume that p < q < p+r. Then every solution of Eq(10.22) with initial conditions
in the invariant interval

1,
r
q−p

converges to the equilibrium y.
Proof. The proof is an immediate consequence of Lemma 10.4.1 and Theorem 1.4.7.
2
Theorem 10.4.2 Assume that p > q, r <
p
2
−pq
q
2
, and that either
p ≤ 1
or
p > 1 and (q −1)[(p −1)
2
(q −1) −4q(1 −p −qr)] ≤ 0.
Then every solution of Eq(10.22) with initial conditions in the invariant interval

qr
p−q
,
p
q

converges to the equilibrium y.
Proof. The proof is an immediate consequence of Lemma 10.4.1 and Theorem 1.4.5.
2
Theorem 10.4.3 Assume that p < q, p +r < q, and that Condition (10.25) is not sat-
isfied. Then every solution of Eq(10.22) with initial conditions in the invariant interval

r
q−p
, 1

converges to the equilibrium y.
Proof. The proof is an immediate consequence of Lemma 10.4.1 and Theorem 1.4.6.
2
180 Chapter 10. (3, 2)-Type Equations
10.5 Open Problems and Conjectures
Open Problem 10.5.1 We know that every solution {y
n
}

n=−1
of Eq(10.4) converges
to a not necessarily prime period-two solution
. . . , φ, ψ, φ, ψ, . . . (10.27)
if and only if
r = 1 + q.
(a) Determine the set of initial conditions y
−1
and y
0
for which φ = ψ.
(b) Assume (10.27) is a prime period-two solution of Eq(10.4). Determine the set of
initial conditions y
−1
and y
0
for which {y
n
}

n=−1
converges to (10.27).
Open Problem 10.5.2 Assume
r > q + 1.
(a) Find the set B of all initial conditions (y
−1
, y
0
) ∈ (0, ∞) × (0, ∞) such that the
solutions {y
n
}

n=−1
of Eq(10.4) are bounded.
(b) Let (y
−1
, y
0
) ∈ B. Investigate the asymptotic behavior of {y
n
}

n=−1
.
Conjecture 10.5.1 Every positive solution of Eq(10.21) converges to the positive equi-
librium of the equation.
Open Problem 10.5.3 For each of the equations listed below, determine the set G of
all initial conditions (x
−1
, x
0
) ∈ R×R through which the equation is well defined for all
n ≥ 0, and for these initial points investigate the long-term behavior of the corresponding
solution:
x
n+1
=
1 + x
n
+ x
n−1
1 + x
n
(10.28)
x
n+1
=
1 + x
n
+ x
n−1
1 + x
n−1
(10.29)
x
n+1
=
1 + x
n
+ x
n−1
x
n
+ x
n−1
. (10.30)
Extend and generalize.
10.5. Open Problems and Conjectures 181
Conjecture 10.5.2 Assume that
p, q, r ∈ (0, ∞).
Then the following statements are true for Eq(10.22)
(a) Local stability of the positive equilibrium implies global stability.
(b) When Eq(10.22) has no prime period-two solutions the equilibrium y is globally
asymptotically stable.
(c) When Eq(10.22) possesses a period-two solution, the equilibrium y is a saddle
point.
(d) The period-two solution of Eq(10.22), when it exists, is locally asymptotically sta-
ble, but not globally.
Open Problem 10.5.4 Assume that {p
n
}

n=0
, {q
n
}

n=0
, and {r
n
}

n=0
are convergent se-
quences of nonnegative real numbers with finite limits,
p = lim
n→∞
p
n
, q = lim
n→∞
q
n
, and r = lim
n→∞
r
n
.
Investigate the asymptotic behavior and the periodic nature of all positive solutions of
each of the following three difference equations:
y
n+1
=
p
n
+ q
n
y
n
+ r
n
y
n−1
1 + y
n
, n = 0, 1, . . . (10.31)
y
n+1
=
p
n
+ q
n
y
n
+ r
n
y
n−1
1 + y
n−1
, n = 0, 1, . . . (10.32)
y
n+1
=
r
n
+ p
n
y
n
+ y
n−1
q
n
y
n
+ y
n−1
, n = 0, 1, . . . . (10.33)
Open Problem 10.5.5 Assume that {p
n
}

n=0
, {q
n
}

n=0
, and {r
n
}

n=0
are period-two se-
quences of nonnegative real numbers. Investigate the global character of all positive
solutions of Eqs(10.31)-(10.33). Extend and generalize.
182 Chapter 10. (3, 2)-Type Equations
Open Problem 10.5.6 Assume that
p, q, r ∈ [0, ∞) and k ∈ {2, 3, . . .}.
Investigate the global behavior of all positive solutions of each of the following difference
equations:
y
n+1
=
p + qy
n
+ ry
n−k
1 + y
n
, n = 0, 1, . . . (10.34)
y
n+1
=
p + qy
n
+ ry
n−k
1 + y
n−k
, n = 0, 1, . . . (10.35)
y
n+1
=
p + qy
n
+ ry
n−1
qy
n
+ y
n−k
, n = 0, 1, . . . . (10.36)
Chapter 11
The (3, 3)-Type Equation
x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
In this chapter we discuss the (3, 3)-type equation
x
n+1
=
α + βx
n
+ γx
n−1
A + Bx
n
+ Cx
n−1
, n = 0, 1, . . . . (11.1)
Please recall our classification convention in which all coefficients of this equation are
now assumed to be positive and the initial conditions nonnegative.
This equation has not been investigated yet. Here we make some simple observations
about linearized stability, invariant intervals, and the convergence of solutions in certain
regions of initial conditions.
11.1 Linearized Stability Analysis
Eq(11.1) has a unique equilibrium which is the positive solution of the quadratic equation
(B + C)x
2
−(β + γ −A)x −α = 0.
That is,
x =
β + γ −A +

(A −β −γ)
2
+ 4α(B + C)
2(B + C)
.
The linearized equation about x is (see Section 2.3)
z
n+1

(βA −Bα) + (βC −Bγ)x
A
2
+ α(B + C) + (B + C)(A + β + γ)x
z
n

(γA −Cα) −(βC −γB)x
A
2
+ α(B + C) + (B + C)(A + β + γ)x
z
n−1
= 0, n = 0, 1, . . . . (11.2)
183
184 Chapter 11. The (3, 3)-Type Equation x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
The equilibrium x is locally asymptotically stable if Conditions (2.13), (2.14), and (2.15)
are satisfied.
Eq(11.1) has a prime period-two solution
. . . , φ, ψ, φ, ψ, . . .
if, see Section 2.5, Conditions (2.31) and (2.32) are satisfied; that is,
γ > β+A, B > C, and α <
(γ −β −A)[B(γ −β −A) −C(γ + 3β −A)]
4C
2
. (11.3)
Furthermore, when (11.3), holds the values of φ and ψ are the positive roots of the
quadratic equation
t
2
+
γ −β −A
C
t +
αC + β(γ −β −A)
C(B −C)
= 0. (11.4)
The local asymptotic stability of the two cycle
. . . , φ, ψ, φ, ψ, . . .
is discussed in Section 2.6.
11.2 Invariant Intervals
Here we obtain several invariant intervals for Eq(11.1). These intervals are derived by
analyzing the intervals where the function
f(x, y) =
α + βx + γy
A + Bx + Cy
is increasing or decreasing in x and y.
If we set
L = βA −αB, M = βC −γB, and N = γA −αC
then clearly
f
x
=
L + My
(A + Bx + Cy)
2
and f
y
=
N −My
(A + Bx + Cy)
2
.
Recall that an interval I is an invariant interval for Eq(11.1) if whenever,
y
N
, y
N+1
∈ I for some integer N ≥ 0,
then
y
n
∈ I for n ≥ N.
Once we have an invariant interval for Eq(11.1) then we may be able to obtain a con-
vergence result for the solutions of Eq(11.1) by testing whether the hypotheses of one of
the four Theorems 1.4.5-1.4.8 are satisfied.
11.2. Invariant Intervals 185
Lemma 11.2.1 (i) Assume that
γ
C
=
β
B
>
α
A
.
Then

α
A
,
β
B

is an invariant interval for Eq(11.1) and in this interval the function
f(x, y) is increasing in both variables.
(ii) Assume that
β
B
>
γ
C
>
α
A
and
γ
C

N
M
.
Then

α
A
,
γ
C

is an invariant interval for Eq(11.1) and in this interval the function
f(x, y) is increasing in both variables.
(iii) Assume that
γ
C
>
β
B
>
α
A
and
β + γ
B + C

αB −βA
βC −γB
.
Then

α
A
,
β+γ
B+C

is an invariant interval for Eq(11.1) and in this interval the func-
tion f(x, y) is increasing in both variables.
(iv) Assume that
γ
C
=
β
B
<
α
A
.
Then

β
B
,
α
A

is an invariant interval for Eq(11.1) and in this interval the function
f(x, y) is decreasing in both variables.
(v) Assume that
γA −αC
βC −γB

α
A
>
β
B
>
γ
C
.
Then

0,
γA−αC
βC−γB

is an invariant interval for Eq(11.1) and in this interval the func-
tion f(x, y) is decreasing in both variables.
(vi) Assume that
β
B
<
γ
C
<
α
A
<
γA −αC
βC −γB
.
Then

0,
α
A

is an invariant interval for Eq(11.1) and in this interval the function
f(x, y) is decreasing in both variables.
(vii) Assume that
α
A
=
β
B
>
γ
C
.
Then

0,
α
A

is an invariant interval for Eq(11.1) and in this interval the function
f(x, y) is increasing in x and decreasing in y.
186 Chapter 11. The (3, 3)-Type Equation x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
(viii) Assume that
β
B
>
α
A

γ
C
.
Then

0,
β
B

is an invariant interval for Eq(11.1) and in this interval the function
f(x, y) is increasing in x and decreasing in y.
(ix) Assume that
α
A
=
β
B
<
γ
C
.
Then

0,
γ
C

is an invariant interval for Eq(11.1) and in this interval the function
f(x, y) is increasing in y and decreasing in x.
(x) Assume that
β
B
<
α
A

γ
C
.
Then

0,
γ
C

is an invariant interval for Eq(11.1) and in this interval the function
f(x, y) is increasing in y and decreasing in x.
Proof.
(i) The condition
γ
C
=
β
B
>
α
A
is equivalent to M = 0 and L > 0, which in turn imply
that N > 0. Thus, in this case the function f is increasing in both variables for
all values of x and y and
α
A
= f(0, 0) ≤ f(x, y) ≤ f(
β
B
,
β
B
) ≤
β + γ
B + C
=
β
B
.
Here we used the fact that under the condition
γ
C
=
β
B
>
α
A
the function f(u, u) is increasing and consequently
f(u, u) ≤
β + γ
B + C
=
β
B
.
(ii) The condition
β
B
>
γ
C
>
α
A
is equivalent to M > 0, L > 0, and N > 0. Thus, in
this case the function f is increasing in x, for every x. In addition, this function
is increasing in y for x <
N
M
. Thus, assuming
γ
C

N
M
we obtain
α
A
= f(0, 0) ≤ f(x, y) ≤ f(
N
M
,
N
M
) =
γ
C
,
for all x, y ∈

α
A
,
γ
C

.
11.2. Invariant Intervals 187
(iii) The condition
γ
C
>
β
B
>
α
A
is equivalent to M < 0 and L > 0, which in turn imply
that N > 0. Thus, in this case the function f is increasing in y, for every y. In
addition, this function is increasing in x for y < −
L
M
. Thus assuming
β+γ
B+C
≤ −
L
M
,
we obtain,
α
A
= f(0, 0) ≤ f(x, y) ≤ f

β + γ
B + C
,
β + γ
B + C


β + γ
B + C
,
for all x, y ∈

α
A
,
β+γ
B+C

.
(iv) The condition
γ
C
=
β
B
<
α
A
is equivalent to M = 0 and L < 0, which in turn imply
that N < 0. Thus, in this case the function f is decreasing in both variables for
all values of x and y and
β
B
≤ f

α
A
,
α
A

≤ f(x, y) ≤ f(0, 0) =
α
A
.
Here we used the fact that under the condition
γ
C
=
β
B
<
α
A
the function f(u, u) is decreasing and consequently
f(u, u) ≥
β + γ
B + C
=
β
B
.
(v) The condition
α
A
>
β
B
>
γ
C
is equivalent to M > 0 and L < 0, which imply that
N < 0. Thus, in this case the function f is decreasing in y, for every y. In addition,
this function is decreasing in x for y < −
L
M
. Thus assuming
α
A
≤ −
L
M
, we obtain
0 ≤ f(x, y) ≤ f(0, 0) =
α
A
≤ −
L
M
,
for all x, y ∈

0, −
L
M

.
(vi) The condition
α
A
>
γ
C
>
β
B
is equivalent to M < 0, L < 0, and N < 0. Thus, in
this case the function f is decreasing in x, for every x. In addition, this function
is decreasing in y for x <
N
M
. Thus assuming
α
A

N
M
, we obtain
0 ≤ f(x, y) ≤ f(0, 0) =
α
A
,
for all x, y ∈

0,
α
A

.
188 Chapter 11. The (3, 3)-Type Equation x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
(vii) The condition
α
A
=
β
B
>
γ
C
is equivalent to M > 0 and L = 0, which in turn imply
that N < 0. Thus, in this case the function f is increasing in x and decreasing in
y for all values of x and y. Using this we obtain,
0 ≤ f(x, y) ≤ f

α
A
, 0

=
α
A
,
for all x, y ∈

0,
α
A

.
(viii) The condition
β
B
>
α
A
>
γ
C
is equivalent to M > 0, L > 0, and N < 0. Thus, in
this case the function f is increasing in x and decreasing in y for all values of x
and y. Using this we obtain,
0 ≤ f(x, y) < f

β
B
, 0

=
β
B
,
for all x, y ∈

0,
β
B

.
(ix) The condition
α
A
=
β
B
<
γ
C
is equivalent to M < 0 and L = 0, which in turn imply
that N > 0. Thus, in this case the function f is decreasing in x and increasing in
y for all values of x and y. Using this we obtain,
0 ≤ f(x, y) ≤ f

0,
γ
C

=
γ
C
,
for all x, y ∈

0,
γ
C

.
(x) The condition
β
B
<
α
A

γ
C
is equivalent to M < 0, L < 0, and N ≥ 0. Thus, in
this case the function f is increasing in x and decreasing in y for all values of x
and y. Using this we obtain,
0 ≤ f(x, y) < f

0,
γ
C


γ
C
,
for all x, y ∈

0,
γ
C

. Here we use the fact that the function f(0, v) is increasing in
v when N > 0 and so
f(0, v) ≤
γ
C
= lim
v→∞
f(0, v),
for all v.
2
The next table summarizes more complicated invariant intervals with the signs of
the partial derivatives f
x
and f
y
.
11.3. Convergence Results 189
Case Invariant intervals Signs of derivatives
1
[0,
N
M
] if
αM+(β+γ)N
AM+(B+C)N
<
N
M
[
N
M
, K
1
] if
αM+(β+γ)N
AM+(B+C)N
>
N
M
f
x
> 0
f
y
> 0 if x <
N
M
; f
y
< 0 if x >
N
M
2
[0,
β−A+

(β−A)
2
+4Bα
2B
]
f
x
> 0
f
y
< 0
3
[0,
β+γ−A+

(β+γ−A)
2
+4α(B+C)
2(B+C)
]
f
x
> 0
f
y
> 0
4
[0, −
L
M
] if
αM−(β+γ)L
AM−(B+C)L
< −
L
M
[−
L
M
, K
2
] if
βMK+αM−γL
BMK+AM−CL
> −
L
M
f
x
> 0 if y < −
L
M
; f
x
< 0 if y > −
L
M
f
y
> 0
5 [0,
β
B
]
f
x
> 0
f
y
< 0
6 [0,
γ
C
]
f
x
< 0
f
y
> 0
7
[
αM−L(β+γ)
AM−L(B+C)
, −
L
M
] if
αM−L(β+γ)
AM−L(B+C)
< −
L
M
[−
L
M
, K
3
] if
αM−βL+γMK
AM−BL+CMK
> −
L
M
f
x
> 0 if y > −
L
M
; f
x
< 0 if y < −
L
M
f
y
< 0
8 [0,
α
A
]
f
x
< 0
f
y
< 0
9
[
αM+(β+γ)N
AM+(B+C)N
,
N
M
] if
αM+(β+γ)N
AM+(B+C)N
<
N
M
[
N
M
, K
4
] if
αM+βK+γNM
AM+BK+CNM
>
N
M
f
x
< 0
f
y
> 0 if x >
N
M
; f
y
< 0 if x <
N
M
where
K
1
=
(β −A)M −CN +

((β −A)M −NC)
2
+ 4BM(αM + γN)
2BM
,
K
2
=
(A −γ)M −BL +

((A −γ)M −BL)
2
+ 4CM(αM −βL)
−2CM
,
K
3
=
(β −A)M + LC +

((β −A)M + LC)
2
+ 4BM(αM −γL)
2BM
,
K
4
=
(γ −A)M −BN +

((γ −A)M −BN)
2
+ 4BM(αM + βN)
2CM
.
11.3 Convergence Results
The following result is a consequence of Lemma 11.2.1 and Theorems 1.4.5-1.4.8.
Theorem 11.3.1 Let {x
n
} be a solution of Eq(11.1). Then in each of the following
cases
lim
n→∞
x
n
= x.
190 Chapter 11. The (3, 3)-Type Equation x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
(i)
γ
C
=
β
B
>
α
A
and x
−1
, x
0

¸
α
A
,
β
B
¸
;
(ii)
β
B
>
γ
C
>
α
A
,
γ
C

γA −αC
βC −γB
, and x
−1
, x
0

¸
α
A
,
γ
C

;
(iii)
γ
C
>
β
B
>
α
A
,
β + γ
B + C

αB −βA
βC −γB
, and x
−1
, x
0

¸
α
A
,
β + γ
B + C
¸
;
(iv)
γ
C
=
β
B
<
α
A
and x
−1
, x
0

¸
β
B
,
α
A
¸
;
(v)
γA −αC
βC −γB

α
A
>
β
B
>
γ
C
and x
−1
, x
0

¸
0,
γA −αC
βC −γB
¸
;
(vi)
β
B
<
γ
C
<
α
A
<
γA −αC
βC −γB
and x
−1
, x
0

¸
0,
α
A

;
(vii) At least one of the following three conditions is not satisfied and x
−1
, x
0

0,
α
A

:
β > γ + A,
B < C,
and
(C −B)[(C −B)(β −γ −A)
2
−4B(αB + γ(β −γ −A))] > 0,
or
α
A
=
β
B
>
γ
C
;
(viii) At least one of the following three conditions is not satisfied and x
−1
, x
0

0,
β
B

:
β > γ + A,
B < C,
and
(C −B)[(C −B)(β −γ −A)
2
−4B(αB + γ(β −γ −A))] > 0,
or
β
B
>
α
A

γ
C
;
11.3. Convergence Results 191
(ix) Condition (11.3) is not satisfied,
α
A
=
β
B
<
γ
C
and x
−1
, x
0

¸
0,
γ
C

;
(x) Condition (11.3) is not satisfied,
β
B
<
α
A

γ
C
and x
−1
, x
0

¸
0,
γ
C

.
Proof. The proofs of statements (i-iii) follow by applying Lemma 11.2.1 (i-iii) and
Theorem 1.4.8.
The proofs of statements (iv-vi) follow by applying Lemma 11.2.1 (iv-vi) and Theo-
rem 1.4.7.
The proofs of statements (vii) and (viii) follow by applying Lemma 11.2.1 (vii) and
(viii) respectively, Theorem 1.4.5, and by observing that the only solution of the system
of equations
m = f(m, M), M = f(M, m);
that is,
m =
α + βm + γM
A + Bm + CM
, M =
α + βM + γm
A + BM + Cm
is m = M.
The proofs of statements (ix) and (x) follow by applying Lemma 11.2.1 (ix) and (x)
respectively, Theorem 1.4.6, and by observing that Eq(11.1) does not possess a prime
period-two solution.
2
11.3.1 Global Stability
Theorem 1.4.2 and linearized stability imply the following global stability result:
Theorem 11.3.2 Assume that
γ
C
< min

β
B
,
α
A
¸
.
Then the equilibrium x of Eq(11.1) is globally asymptotically stable.
By applying Theorem 1.4.3 to Eq(11.1) we obtain the following global asymptotic
result.
192 Chapter 11. The (3, 3)-Type Equation x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
Theorem 11.3.3 Assume that
B ≤ C, γB > βC, αB > βA, and γA > αC.
Then the equilibrium x of Eq(11.1) is globally asymptotically stable.
The following global stability result is a consequence of linearized stability and the
Trichotomy Theorem 1.4.4.
Theorem 11.3.4 (a) Assume that
αC ≥ γA, βA ≥ αB, and 3γB ≥ βC ≥ γB.
Then the equilibrium x of Eq(11.1) is globally asymptotically stable.
(b) Assume that
αB ≥ βA, γA ≥ αC, and 3βC ≥ γB ≥ βC.
Then the equilibrium x of Eq(11.1) is globally asymptotically stable.
11.4 Open Problems and Conjectures
Conjecture 11.4.1 Assume
α, β, γ, A, B, C ∈ (0, ∞).
Show that every positive solution of Eq(11.1) is bounded.
Conjecture 11.4.2 Assume
α, β, γ, A, B, C ∈ (0, ∞)
and that Eq(11.1) has no positive prime period-two solution. Show that every positive
solution of Eq(11.1) converges to the positive equilibrium.
Conjecture 11.4.3 Assume
α, β, γ, A, B, C ∈ (0, ∞)
and that Eq(11.1) has a positive prime period-two solution. Show that the positive equi-
librium of Eq(11.1) is a saddle point and the period-two solution is locally asymptotically
stable.
11.4. Open Problems and Conjectures 193
Conjecture 11.4.4 Assume
α, β, γ, A, B, C ∈ (0, ∞).
Show that Eq(11.1) cannot have a prime period-k solution for any k ≥ 3.
Open Problem 11.4.1 Assume {α
n
}

n=0
, {β
n
}

n=0
, {γ
n
}

n=0
, {A
n
}

n=0
, {B
n
}

n=0
, {C
n
}

n=0
are convergent sequences of nonnegative real numbers with finite limits. Investigate the
asymptotic behavior and the periodic nature of all positive solutions of the difference
equation
x
n+1
=
α
n
+ β
n
x
n
+ γ
n
x
n−1
A
n
+ B
n
x
n
+ C
n
x
n−1
, n = 0, 1, . . . . (11.5)
Open Problem 11.4.2 Assume that the parameters in Eq(11.5) are period-two se-
quences of positive real numbers. Investigate the asymptotic behavior and the periodic
nature of all positive solutions of Eq(11.5).
Open Problem 11.4.3 Assume that
a, A, a
i
, A
i
∈ [0, ∞) for i = 0, 1, . . . , k. (11.6)
Obtain necessary and sufficient conditions so that every positive solution of the equation
x
n+1
=
a + a
0
x
n
+ . . . + a
k
x
n−k
A + A
0
x
n
+ . . . + A
k
x
n−k
, n = 0, 1, . . . . (11.7)
has a finite limit.
Open Problem 11.4.4 Assume that (11.6) holds. Obtain necessary and sufficient con-
ditions so that every positive solution of Eq(11.7) converges to a period-two solution.
Open Problem 11.4.5 Assume that (11.6) holds. Obtain necessary and sufficient con-
ditions so that every positive solution of Eq(11.7) is periodic with period-k, k ≥ 2. In
particular, investigate the cases where 2 ≤ k ≤ 8.
194 Chapter 11. The (3, 3)-Type Equation x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
Open Problem 11.4.6 Let
α, β, γ, δ ∈ [0, ∞) with γ + δ > 0.
Investigate the boundedness character, the periodic nature, and the asymptotic behavior
of all positive solutions of the difference equation
x
n+1
=
αx
n
+ βx
n−1
x
n−2
γx
n
x
n−1
+ δx
n−2
, n = 0, 1, . . . .
(See [8] and [61].)
Conjecture 11.4.5 Let k ≥ 2 be a positive integer. Show that every positive solution
of the difference equation
x
n+1
=
x
n
+ . . . + x
n−k
+ x
n−k−1
x
n−k−2
x
n
x
n−1
+ x
n−2
+ . . . + x
n−k−2
, n = 0, 1, . . .
converges to 1. Extend and generalize.
(See [8], [47], and [60].)
Conjecture 11.4.6 Let
a
ij
∈ (0, ∞) for i, j ∈ {1, 2, 3}.
Show that every positive solution of the system
x
n+1
=
a
11
x
n
+
a
12
y
n
+
a
13
z
n
y
n+1
=
a
21
x
n
+
a
22
y
n
+
a
23
z
n
z
n+1
=
a
31
x
n
+
a
32
y
n
+
a
33
z
n

, n = 0, 1, . . . (11.8)
converges to a period-two solution. Extend to equations with real parameters.
(See [32] for the two dimensional case.)
Conjecture 11.4.7 Assume that
p, q ∈ [0, ∞).
11.4. Open Problems and Conjectures 195
(a) Show that every positive solution of the equation
x
n+1
=
px
n−1
+ x
n−2
q + x
n−2
, n = 0, 1, . . .
converges to a period-two solution if and only if
p = 1 + q.
(b) Show that when
p < 1 + q
every positive solution has a finite limit.
(c) Show that when
p > 1 + q
the equation possesses positive unbounded solutions.
(See [62].)
Conjecture 11.4.8 Assume that
p ∈ (0, ∞).
(a) Show that every positive solution of the equation
x
n+1
=
x
n−1
x
n−1
+ px
n−2
, n = 0, 1, . . .
converges to a period-two solution if
p ≥ 1.
(b) Show that when
p < 1
the positive equilibrium x =
1
1+p
is globally asymptotically stable.
Conjecture 11.4.9 Assume that
p, q ∈ [0, ∞).
196 Chapter 11. The (3, 3)-Type Equation x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
(a) Show that every positive solution of the equation
x
n+1
=
x
n−1
+ p
x
n−2
+ q
, n = 0, 1, . . .
converges to a period-two solution if and only if
q = 1.
(b) Show that when
q > 1
the positive equilibrium of the equation is globally asymptotically stable.
(c) Show that when
q < 1
the equation possesses positive unbounded solutions.
(See [62].)
Conjecture 11.4.10 Show that every positive solution of the equation
x
n+1
=
x
n
+ x
n−2
x
n−1
, n = 0, 1, . . .
converges to a period-four solution.
(See [62].)
Conjecture 11.4.11 Show that the difference equation
x
n+1
=
p + x
n−2
x
n
, n = 0, 1, . . .
has the following trichotomy character:
(i) When p > 1 every positive solution converges to the positive equilibrium.
(ii) When p = 1 every positive solution converges to a period-five solution.
(iii) When p < 1 there exist positive unbounded solutions.
(See [62].)
11.4. Open Problems and Conjectures 197
Conjecture 11.4.12 Show that every positive solution of the equation
x
n+1
=
1 + x
n
x
n−1
+ x
n−2
, n = 0, 1, . . .
converges to a period-six solution of the form
. . . , φ, ψ,
ψ
φ
,
1
φ
,
1
ψ
,
φ
ψ
, . . .
where
φ, ψ ∈ (0, ∞).
(See [62].)
Open Problem 11.4.7 Consider the difference equation
x
n+1
=
α + βx
n
+ γx
n−1
+ δx
n−2
A + Bx
n
+ Dx
n−2
, n = 0, 1, . . . (11.9)
with nonnegative parameters and nonnegative initial conditions such that B+D > 0 and
A + Bx
n
+ Dx
n−2
> 0 for n ≥ 0.
Then one can show that Eq(11.9) has prime period-two solutions if and only if
γ = β + δ + A. (11.10)
(a) In addition to (11.10), what other conditions are needed so that every positive
solution of Eq(11.9) converges to a period-two solution?
(b) Assume
γ < β + δ + A.
What other conditions are needed so that every positive solution of Eq(11.9) con-
verges to the positive equilibrium?
(c) Assume
γ > β + δ + A.
What other conditions are needed so that Eq(11.9) possesses unbounded solutions?
(See [6].)
198 Chapter 11. The (3, 3)-Type Equation x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
Conjecture 11.4.13 Show that every positive solution of each of the two equations
x
n+1
=
1 + x
n−1
+ x
n−2
x
n
, n = 0, 1, . . .
and
x
n+1
=
1 + x
n−1
+ x
n−2
x
n−2
, n = 0, 1, . . .
converges to a period-two solution.
(See [6].)
Conjecture 11.4.14 Show that every positive solution of the equation
x
n+1
=
1 + x
n−1
1 + x
n
+ x
n−2
, n = 0, 1, . . .
converges to a period-two solution.
(See [6].)
Conjecture 11.4.15 Show that every positive solution of the equation
x
n+1
=
x
n−2
x
n−1
+ x
n−2
+ x
n−3
, n = 0, 1, . . .
converges to a period-three solution.
Conjecture 11.4.16 Consider the difference equation
x
n+1
=
p + qx
n
+ rx
n−2
x
n−1
, n = 0, 1, . . .
where
p ∈ [0, ∞) and q, r ∈ (0, ∞).
(a) Show that when
q = r
every positive solution converges to a period-four solution.
(b) Show that when
q > r
every positive solution converges to the positive equilibrium.
11.4. Open Problems and Conjectures 199
(c) Show that when
q < r
the equation possesses positive unbounded solutions.
(See [62].)
Open Problem 11.4.8 Determine whether every positive solution of each of the fol-
lowing equations converges to a periodic solution of the corresponding equation:
(a)
x
n+1
=
x
n−2
+ x
n−3
x
n
, n = 0, 1, . . .
(b)
x
n+1
=
1 + x
n−3
1 + x
n
+ x
n−3
, n = 0, 1, . . .
(c)
x
n+1
=
1 + x
n
+ x
n−3
x
n−2
, n = 0, 1, . . .
(d)
x
n+1
=
x
n−1
x
n
+ x
n−2
+ x
n−3
, n = 0, 1, . . . .
Open Problem 11.4.9 Assume that k is a positive integer and that
A, B
0
, B
1
, . . . B
k
∈ [0, ∞) with
k
¸
i=0
B
i
> 0.
Obtain necessary and sufficient conditions on the parameters A and B
0
, B
1
, . . . B
k
so
that every positive solution of the difference equation
x
n+1
=
x
n−k
A + B
0
x
n
+ . . . + B
k
x
n−k
, n = 0, 1, . . .
converges to a period-(k + 1) solution of this equation.
(See [6].)
200 Chapter 11. The (3, 3)-Type Equation x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
Open Problem 11.4.10 Assume that k is a positive integer. Then we can easily see
that every nonnegative solution of the equation
x
n+1
=
x
n−k
1 + x
n
+ . . . + x
n−k+1
, n = 0, 1, . . . (11.11)
converges to a period-(k + 1) solution of the form
. . . , 0, 0, ..., 0
. .. .
k−terms
, φ, . . . (11.12)
with φ ≥ 0.
(a) Does Eq(11.11) possess a positive solution which converges to zero?
(b) Determine, in terms of the nonnegative initial conditions x
−k
, . . . , x
0
, the value of
φ in (11.12) which corresponds to the nonnegative solution {x
n
}

n=−k
of Eq(11.11).
(c) Determine the set of all nonnegative initial conditions x
−k
, . . . , x
0
for which the
corresponding solution {x
n
}

n=−k
of Eq(11.11) converges to a period-(k+1) solution
of the form (11.12) with a given φ ≥ 0 ?
(See [6].)
Conjecture 11.4.17 Let k be a positive integer. Show that every positive solution of
x
n+1
=
1 + x
n
+ x
n−k
x
n−k+1
, n = 0, 1, . . .
converges to a periodic solution of period 2k.
Appendix A
Global Attractivity for Higher
Order Equations
In this appendix we present some global attractivity results for higher order difference
equations which are analogous to Theorems 1.4.5-1.4.8. We state the results for third
order equations from which higher order extensions will become clear.
Let I be some interval of real numbers and let
f : I ×I ×I →I
be a continuously differentiable function.
Then for every set of initial conditions x
0
, x
−1
, x
−2
∈ I, the difference equation
x
n+1
= f(x
n
, x
n−1
, x
n−2
), n = 0, 1, . . . (A.1)
has a unique solution {x
n
}

n=−2
.
As we have seen in previous chapters, very often a good strategy for obtaining global
attractivity results for Eq(A.1) is to work in the regions where the function f(x, y, z) is
monotonic in its arguments. In this regard there are eight possible scenarios depending
on whether f(x, y, z) is non decreasing in all three arguments, or non increasing in two
arguments and non decreasing in the third, or non increasing in one and non decreasing
in the other two, or non decreasing in all three.
Theorem A.0.1 Let [a, b] be an interval of real numbers and assume that
f : [a, b]
3
→[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y, z) is non-decreasing in each of its arguments;
(b) The equation
f(x, x, x) = x,
has a unique solution in the interval [a, b].
Then Eq(A.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(A.1)
converges to x.
201
202 Appendix
Proof. Set
m
0
= a and M
0
= b
and for i = 1, 2, . . . set
M
i
= f(M
i−1
, M
i−1
, M
i−1
) and m
i
= f(m
i−1
, m
i−1
, m
i−1
).
Now observe that for each i ≥ 0,
m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . ≤ M
i
≤ . . . ≤ M
1
≤ M
0
,
and
m
i
≤ x
k
≤ M
i
for k ≥ 3i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
Then
M ≥ limsup
i→∞
x
i
≥ liminf
i→∞
x
i
≥ m (A.2)
and by the continuity of f,
m = f(m, m, m) and M = f(M, M, M).
In view of (b),
m = M = x,
from which the result follows. 2
Theorem A.0.2 Let [a, b] be an interval of real numbers and assume that
f : [a, b]
3
→[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y, z) is non-decreasing in x and y ∈ [a, b] for each z ∈ [a, b], and is non-
increasing in z ∈ [a, b] for each x and y ∈ [a, b];
(b) If (m, M) ∈ [a, b] ×[a, b] is a solution of the system
m = f(m, m, M) and M = f(M, M, m)
then
m = M.
Then Eq(A.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(A.1) converges
to x.
Appendix 203
Proof. Set
m
0
= a and M
0
= b
and for each i = 1, 2, . . . set
M
i
= f(M
i−1
, M
i−1
, m
i−1
)
and
m
i
= f(m
i−1
, m
i−1
, M
i−1
).
Now observe that for each i ≥ 0
a = m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . M
i
≤ . . . M
1
≤ M
0
= b.
and
m
i
≤ x
k
≤ M
i
for k ≥ 3i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
Then (A.2) holds and by the continuity of f,
m = f(m, m, M) and M = f(M, M, m).
Therefore in view of (b)
m = M
from which the result follows. 2
Theorem A.0.3 Let [a, b] be an interval of real numbers and assume that
f : [a, b]
3
→[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y, z) is non-decreasing in x and z in [a, b] for each z ∈ [a, b], and is non-
increasing in y ∈ [a, b] for each x and z in [a, b];
(b) If (m, M) ∈ [a, b] ×[a, b] is a solution of the system
M = f(M, m, M) and m = f(m, M, m)
then
m = M.
Then Eq(A.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(A.1) converges
to x.
204 Appendix
Proof. Set
m
0
= a and M
0
= b
and for each i = 1, 2, . . . set
M
i
= f (M
i−1
, m
i−1
, M
i−1
)
and
m
i
= f (m
i−1
, M
i−1
, m
i−1
) .
Now observe that for each i ≥ 0
a = m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . M
i
≤ . . . M
1
≤ M
0
= b.
and
m
i
≤ x
k
≤ M
i
for k ≥ 3i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
Then (A.2) holds and by the continuity of f,
m = f (m, m, M) and M = f (M, M, m) .
Therefore in view of (b)
m = M
from which the results follows. 2
Theorem A.0.4 Let [a, b] be an interval of real numbers and assume that
f : [a, b]
3
→[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y, z) is non-increasing in x for each y and z ∈ [a, b] and is non-decreasing
in y and z for each x, y ∈ [a, b];
(b) If (m, M) ∈ [a, b] ×[a, b] is a solution of the system
M = f(m, M, M) and m = f(M, m, m)
then
m = M.
Then Eq(A.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(A.1) converges
to x.
Appendix 205
Proof. Set
m
0
= a and M
0
= b
and for each i = 1, 2, . . . set
M
i
= f(m
i−1
, M
i−1
, M
i−1
)
and
m
i
= f(M
i−1
, m
i−1
, m
i−1
).
Now observe that for each i ≥ 0
m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . M
i
≤ . . . M
1
≤ M
0
= b
and
m
i
≤ x
k
≤ M
i
for k ≥ 3i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
Then (A.2) holds and by the continuity of f,
M = f(m, M, M) and m = f(M, m, m).
Therefore in view of (b)
m = M
from which the results follows. 2
Theorem A.0.5 Let [a, b] be an interval of real numbers and assume that
f : [a, b]
3
→[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y, z) is non-decreasing in x for each y and z ∈ [a, b] and is non-increasing
in y and z for each x ∈ [a, b];
(b) If (m, M) ∈ [a, b] ×[a, b] is a solution of the system
M = f(M, m, m) and m = f(m, M, M)
then
m = M.
Then Eq(A.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(A.1) converges
to x.
206 Appendix
Proof. Set
m
0
= a and M
0
= b
and for each i = 1, 2, . . . set
M
i
= f(M
i−1
, m
i−1
, m
i−1
)
and
m
i
= f(m
i−1
, M
i−1
, M
i−1
).
Now observe that for each i ≥ 0
m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . M
i
≤ . . . M
1
≤ M
0
= b
and
m
i
≤ x
k
≤ M
i
for k ≥ 3i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
Then (A.2) holds and by the continuity of f,
M = f(M, m, m) and m = f(m, M, M).
Therefore in view of (b)
m = M
from which the results follows. 2
Theorem A.0.6 Let [a, b] be an interval of real numbers and assume that
f : [a, b]
3
→[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y, z) is non-decreasing in y for each x and z ∈ [a, b] and is non-increasing
in x and z for each y ∈ [a, b];
(b) If (m, M) ∈ [a, b] ×[a, b] is a solution of the system
M = f(m, M, m) and m = f(M, m, M)
then
m = M.
Then Eq(A.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(A.1) converges
to x.
Appendix 207
Proof. Set
m
0
= a and M
0
= b
and for each i = 1, 2, . . . set
M
i
= f(m
i−1
, M
i−1
, m
i−1
)
and
m
i
= f(M
i−1
, m
i−1
, M
i−1
).
Now observe that for each i ≥ 0
m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . M
i
≤ . . . M
1
≤ M
0
= b
and
m
i
≤ x
k
≤ M
i
for k ≥ 3i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
Then (A.2) holds and by the continuity of f,
M = f(m, M, m) and m = f(M, m, M).
Therefore in view of (b)
m = M
from which the results follows. 2
Theorem A.0.7 Let [a, b] be an interval of real numbers and assume that
f : [a, b]
3
→[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y, z) is non-decreasing in z for each x and y in [a, b] and is non-increasing
in x and y for each z ∈ [a, b];
(b) If (m, M) ∈ [a, b] ×[a, b] is a solution of the system
M = f(m, m, M) and m = f(M, M, m)
then
m = M.
Then Eq(A.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(A.1) converges
to x.
208 Appendix
Proof. Set
m
0
= a and M
0
= b
and for each i = 1, 2, . . . set
M
i
= f(m
i−1
, m
i−1
, M
i−1
)
and
m
i
= f(M
i−1
, M
i−1
, m
i−1
).
Now observe that for each i ≥ 0
m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . M
i
≤ . . . M
1
≤ M
0
= b
and
m
i
≤ x
k
≤ M
i
for k ≥ 3i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
Then (A.2) holds and by the continuity of f,
M = f(m, m, M) and m = f(M, M, m).
Therefore in view of (b)
m = M
from which the results follows. 2
Theorem A.0.8 Let [a, b] be an interval of real numbers and assume that
f : [a, b]
3
→[a, b]
is a continuous function satisfying the following properties:
(a) f(x, y, z) is non-increasing in all three variables x, y, z in [a, b];
(b) If (m, M) ∈ [a, b] ×[a, b] is a solution of the system
M = f(m, m, m) and m = f(M, M, M)
then
m = M.
Then Eq(A.1) has a unique equilibrium x ∈ [a, b] and every solution of Eq(A.1) converges
to x.
Appendix 209
Proof. Set
m
0
= a and M
0
= b
and for each i = 1, 2, . . . set
M
i
= f(m
i−1
, m
i−1
, m
i−1
)
and
m
i
= f(M
i−1
, M
i−1
, M
i−1
).
Now observe that for each i ≥ 0
m
0
≤ m
1
≤ . . . ≤ m
i
≤ . . . M
i
≤ . . . M
1
≤ M
0
= b
and
m
i
≤ x
k
≤ M
i
for k ≥ 3i + 1.
Set
m = lim
i→∞
m
i
and M = lim
i→∞
M
i
.
Then (A.2) holds and by the continuity of f,
M = f(m, m, m) and m = f(M, M, M).
Therefore in view of (b)
m = M
from which the results follows. 2
The above theorems can be easily extended to the general difference equation of order
p
x
n+1
= f(x
n
, x
n−1
, . . . , x
n−p+1
), n = 0, 1, . . . (A.3)
where
f : I
p
→I
is a continuous function. In this case, there are 2
p
possibilities for the function f to
be non decreasing in some arguments and non increasing in the remaining arguments.
This gives 2
p
results of the type found in Theorems A.0.1-A.0.8. Let us formulate one
of these results.
Theorem A.0.9 Let [a, b] be an interval of real numbers and assume that
f : [a, b]
p
→[a, b]
is a continuous function satisfying the following properties:
(a) f(u
1
, u
2
, . . . , u
p
) is non decreasing in the second and the third variables u
2
and
u
3
and is non increasing in all other variables.
210 Appendix
(b) If (m, M) ∈ [a, b] ×[a, b] is a solution of the system
M = f(m, M, M, m, . . . , m) and m = f(M, m, m, M, . . . , M)
then
m = M.
Then Eq(A.3) has a unique equilibrium x ∈ [a, b] and every solution of Eq(A.3) converges
to x.
Bibliography
[1] R. Agarwal, Difference Equations and Inequalities. Theory, Methods and Applica-
tions, Marcel Dekker Inc., New York, 1992.
[2] L. S. Allen, M. K. Hannigan, and M. J. Strauss, Mathematical analysis of a model
for a plant-herbivore system, Bulletin of Mathematical Biology 55 (1993), 847-864.
[3] L. S. Allen, M. J. Strauss, M. J. Thorvilson, and W. N. Lipe, A Preliminary math-
ematical model of the Apple Twig Borer (Coleoptera: Bostrichidae) and Grapes on
the Texas High Plains Ecological Modelling 58 (1991), 369-382.
[4] K.T. Alligood, T. Sauer and J.A. Yorke , CHAOS An Introduction to Dynamical
Systems, Springer-Verlag, New York, Berlin, Heidelberg, Tokyo, 1997.
[5] American Mathemathical Monthly, Problem 10670, 106(1999), 967-968.
[6] A. M. Amleh and G. Ladas, Convergence to periodic solutions, J. Differ. Equations
Appl. 7(2001), (to appear).
[7] A. M. Amleh, D. A. Georgiou, E. A. Grove, and G. Ladas, On the recursive sequence
x
n+1
= α +
x
n−1
x
n
, J. Math. Anal. Appl. 233(1999), 790-798.
[8] A. M. Amleh, N. Kruse, and G. Ladas, On a class of difference equations with
strong negative feedback, J. Differ. Equations Appl. 5(6)(1999), 497-515.
[9] A. M. Amleh, G. Ladas, and J. T. Hoag, Boundedness of solutions of a plant-
herbivore system, Proceedings of the Third International Conference on Difference
Equations and Applications, September 1-5, 1977, Taipei, Taiwan, Gordon and
Breach Science Publishers, 1999, 1-5.
[10] R. Abu-Saris, On the periodicity of the difference equation x
n+1
= α|x
n
| + βx
n−1
,
J. Differ. Equations Appl. 5(1)(1999), 57-69.
[11] E. Barbeau, B. Gelbard, and S. Tanny, Periodicities of solutions of the generalized
Lyness recursion, J. Differ. Equations Appl. 1(3)(1995), 291-306.
[12] B. Beckermann and J. Wimp, Some dynamically trivial mappings with applications
to the improvement of simple iteration, Comput. Math. Appl. 24(1998), 89-97.
211
212 BIBLIOGRAPHY
[13] E. Camouzis, R. DeVault, and G. Ladas, On the recursive sequence x
n+1
= −1 +
x
n−1
x
n
, J. Differ. Equations Appl. 7(2001), (to appear).
[14] K. L. Cooke, D. F. Calef, and E. V. Level, Stability or chaos in discrete epidemic
models, Nonlinear Systems and Applications, V. Lakshmikantham, editor, 73-93,
Academic Press, New York, 1977.
[15] K. Cunningham, M. R. S. Kulenovi´c, G. Ladas, and S. Valicenti, On the recursive
Sequence x
n+1
=
α+βx
n
Bx
n
+Cx
n−1
, Proceedings of WCNA2000, Catania, Italy, July 19-26,
2000, Nonlin. Anal. TMA,(to appear).
[16] R. Devaney, A First Course in Chaotic Dynamical Systems: Theory and Experi-
ment, Addison-Wesley, Reading, MA, 1992.
[17] R. Devaney, A Piecewise linear model for the zones of instability of an area-
preserving map, Physica D 10(1984), 387-393.
[18] R. DeVault, L. Galminas, E. J. Janowski and G. Ladas, On the recursive sequence
x
n+1
=
A
x
n
+
B
x
n−1
, J. Differ. Equations Appl. 6(1)(2000), 121-125.
[19] R. DeVault, G. Ladas, and S. W. Schultz, On the Recursive Sequence x
n+1
=
A
x
n
+
1
x
n−2
, Proc. Amer. Math. Soc. 126 (1998), 3257-3261.
[20] R. DeVault, G. Ladas, and S. W. Schultz, On the recursive sequence x
n+1
=
A
x
n
x
n−1
+
1
x
n−3
x
n−4
, J. Differ. Equations Appl. 6(4)(2000), 481-483.
[21] S. Elaydi, An Introduction to Difference Equations, 2nd ed., Springer-Verlag, New
York, 1999.
[22] S. Elaydi, Discrete Chaos, CRC Press, Boca Raton, FL, 2000.
[23] H. El-Metwally, E. A. Grove, and G. Ladas, A global convergence result with ap-
plications to periodic solutions, J. Math. Anal. Appl., 245 (2000), 161-170.
[24] H. El-Metwally, E. A. Grove, G. Ladas, R. Levins, and M. Radin, On the difference
equation x
n+1
= α + βx
n−1
e
−x
n
, Proceedings of WCNA2000, Catania, Italy, July
19-26, 2000, (to appear).
[25] H. El-Metwally, E. A. Grove, G. Ladas, and H. D. Voulov, On the global attractivity
and the periodic character of some difference equations, J. Differ. Equations Appl.
7(2001), (to appear).
[26] J. Feuer, E. J. Janowski, and G. Ladas, Lyness-type equations in the third quadrant,
Nonlinear Anal. 30(1997), 1183-1189.
[27] J. E. Franke, J. T. Hoag, and G. Ladas, Global attractivity and convergence to a
two-cycle in a difference equation, J. Differ. Equations Appl. 5(2)(1999), 203-209.
BIBLIOGRAPHY 213
[28] C. H. Gibbons, M. R. S. Kulenovi´c, and G. Ladas, On the recursive sequence
x
n+1
=
α+βx
n−1
γ+x
n
, Math. Sci. Res. Hot-Line, 4(2)(2000), 1-11.
[29] C. H. Gibbons, M. R. S. Kulenovi´c, G. Ladas, and H. D. Voulov, On the trichotomy
character of x
n+1
=
α+βx
n
+γx
n−1
A+x
n
, J. Differ. Equations Appl. 7(2001) (to appear).
[30] E. A. Grove, E. J. Janowski, C. M. Kent, and G. Ladas, On the rational recursive
sequence x
n+1
= (αx
n
+β)/((γx
n
+C)x
n−1
), Comm. Appl. Nonlinear Anal. 1(1994),
61-72.
[31] E. A. Grove, C. M. Kent, R. Levins, G. Ladas, and S. Valicenti, Global stability
in some population models , Proceedings of the Fourth International Conference
on Difference Equations and Applications, August 27-31, 1998, Poznan, Poland,
Gordon and Breach Science Publishers, 2000, 149-176.
[32] E. A. Grove, G. Ladas, L.C. McGrath, and C. T. Teixeira, Existence and behavior
of solutions of a rational system, Comm. Appl. Nonlinear Anal. 8(2001), 1-25.
[33] J. Guckenheimer and P. Holmes, Nonlinear Oscillations, Dynamical Systems, and
Bifurcations of Vector Fields, Springer-Verlag, New York, Berlin, Heidelberg,
Tokyo, 1983.
[34] J. Hale and H. Kocak, Dynamics and Bifurcations, Springer-Verlag, New York,
Berlin, Heidelberg, Tokyo, 1991.
[35] M. L. J. Hautus and T.S. Bolis, Solution to Problem E2721, Amer. Math. Monthly,
86(1979), 865-866.
[36] J. H. Jaroma , On the global asymptotic stability of x
n+1
=
a+bx
n
A+x
n−1
, Proceedings of
the First International Conference on Difference Equations and Applications, May
25-28, 1994, San Antonio, Texas, Gordon and Breach Science Publishers, 1995,
281-294.
[37] G. Karakostas, A discrete semi-flow in the space of sequences and study of con-
vergence of sequences defined by difference equations, M. E. Greek Math. Society,
(1989), 66-74.
[38] G. Karakostas, Convergence of a difference equation via the full limiting sequences
method, Differential Equations Dynam. Systems 1(1993), 289-294.
[39] G. Karakostas, Asymptotic 2-periodic difference equations with diagonally self-
invertible responses, J. Differ. Equations Appl. 6(3)(2000), 329-335.
[40] G. Karakostas, Ch. G. Philos, and Y. G. Sficas, The dynamics of some discrete
population models, Nonlinear Anal. 17(1991), 1069-1084.
214 BIBLIOGRAPHY
[41] W. G. Kelley and A. C. Peterson, Difference Equations, Academic Press, New York,
1991.
[42] V. L. Kocic and G. Ladas, Global Behavior of Nonlinear Difference Equations of
Higher Order with Applications, Kluwer Academic Publishers, Dordrecht, Holland,
1993.
[43] V. L. Kocic, G. Ladas, and I. W. Rodrigues, On the rational recursive sequences,
J. Math. Anal. Appl. 173 (1993), 127-157.
[44] V. L. Kocic, G. Ladas, G. Tzanetopoulos, and E. Thomas, On the stability of Ly-
ness’ equation, Dynamics of Continuous, Discrete and Impulsive Systems 1(1995),
245-254.
[45] W. A. Kosmala, M. R. S. Kulenovi´c, G. Ladas, and C. T. Teixeira, On the recursive
sequence y
n+1
=
p+y
n−1
qy
n
+y
n−1
, J. Math. Anal. Appl. 251(2000), 571-586.
[46] U. Krause, Stability trichotomy, path stability, and relative stability for positive
nonlinear difference equations of higher order, J. Differ. Equations Appl. 1(4)
(1995), 323-346.
[47] N. Kruse and T. Nesemann, Global asymptotic stability for discrete dynamical
systems, J. Math. Anal. Appl. 235 (1999), 151-158.
[48] Y. K. Kuang and J. M. Cushing, Global stability in a nonlinear difference-delay
equation model of flour beetle population growth, J. Differ. Equations Appl.
2(1)(1996), 31-37.
[49] M. R. S. Kulenovi´c, Invariants and related Liapunov functions for difference equa-
tions, Appl. Math. Lett. 13 (7)(2000), 1-8.
[50] M. R. S. Kulenovi´c and G. Ladas, Open problems and conjectures: On period two
solutions of x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
, J. Differ. Equations Appl. 6(5) (2000), 641-646.
[51] M. R. S. Kulenovi´c, G. Ladas, L. F. Martins, and I. W. Rodrigues, On the dynamics
of x
n+1
=
α+βx
n
A+Bx
n
+Cx
n−1
facts and conjectures, Comput. Math. Appl., (to appear).
[52] M. R. S. Kulenovi´c, G. Ladas, and N. R. Prokup, On the recursive sequence x
n+1
=
αx
n
+βx
n−1
1+x
n
, J. Differ. Equations Appl. 6(5) (2000), 563-576.
[53] M. R. S. Kulenovi´c, G. Ladas, and N. R. Prokup, On a rational difference equation,
Comput. Math. Appl. 41(2001), 671-678.
[54] M. R. S. Kulenovi´c, G. Ladas, and W. S. Sizer, On the recursive sequence x
n+1
=
αx
n
+βx
n−1
γx
n
+Cx
n−1
, Math. Sci. Res. Hot-Line 2(1998), no.5, 1-16.
BIBLIOGRAPHY 215
[55] S. Kuruklis and G. Ladas, Oscillation and global attractivity in a discrete delay
logistic model, Quart. Appl. Math. 50 (1992), 227-233.
[56] G. Ladas, Invariants for generalized Lyness equations, J. Differ. Equations Appl.
1(2)(1995), 209-214.
[57] G. Ladas, Open problems and conjectures, J. Differ. Equations Appl. 1(1)(1995),
95-97.
[58] G. Ladas, Open problems and conjectures, J. Differ. Equations Appl. 1(3)(1995),
317-321.
[59] G. Ladas, Open Problems and Conjectures, J. Differ. Equations Appl. 1(4)(1995),
413-419.
[60] G. Ladas, On the recursive sequence x
n+1
=

k
i=0
A
i
x
n−2i
, J. Differ. Equations Appl.
4(3)(1998), 311-313.
[61] G. Ladas, Open problems and conjectures, J. Differ. Equations Appl. 4(5)(1998),
497-499.
[62] G. Ladas, Open problems and conjectures, J. Differ. Equations Appl. 7(2001), (to
appear).
[63] M. Martelli, Introduction to Discrete Dynamical Systems and Chaos, John Wiley &
Sons, New York, 1999.
[64] Mathemathics Magazine, Problem 1572, 73(2000), 160-161.
[65] Ch. G. Philos, I. K. Purnaras, and Y. G. Sficas, Global attractivity in a nonlinear
difference equation, Appl. Math. Comput. 62(1994), 249-258.
[66] E. C. Pielou, An Introduction to Mathematical Ecology, John Wiley & Sons, New
York, 1965.
[67] E. C. Pielou, Population and Community Ecology, Gordon and Breach, New York,
1974.
[68] C. Robinson, Stability, Symbolic Dynamics, and Chaos, CRC Press, Boca Raton,
FL, 1995.
[69] H. Sedaghat, Existence of solutions of certain singular difference equations, J. Dif-
fer. Equations Appl. 6(5)(2000), 535-561.
[70] H. Sedaghat, Geometric stability conditions for higher order difference equations,
J. Math. Anal. Appl. 224(1998), 255-272.
216 BIBLIOGRAPHY
[71] W. S. Sizer, Some periodic solutions of the Lyness equation , Proceedings of the
Fifth International Conference on Difference Equations and Applications, January
3-7, 2000, Temuco, Chile, Gordon and Breach Science Publishers, 2001, (to appear).
[72] E. C. Zeeman, Geometric unfolding of a difference equation,
http://www.math.utsa.edu/ecz/gu.html
[73] Y. Zheng, Existence of nonperiodic solutions of the Lyness equation x
n+1
=
α+x
n
x
n−1
,
J. Math. Anal. Appl. 209(1997), 94-102.
[74] Y. Zheng, On periodic cycles of the Lyness equation, Differential Equations Dynam.
Systems 6(1998), 319-324.
[75] Y. Zheng, Periodic solutions with the same period of the recursion x
n+1
=
α+βx
n
+γx
n−1
A+Bx
n
+Cx
n−1
, Differential Equations Dynam. Systems 5(1997), 51-58.
Index
basin of attraction of a two cycle, 38
classification of equation types, 3
convergence to period-two solutions, 38
Discrete Epidemic Models, 128
equilibrium point, 5, 29
existence of prime period-two solution,
35
existence of solutions, 2
fixed point, 5
Flour Beetle Model, 129
forbidden set, 2, 17
full limiting sequence, 15
full limiting solution, 16
gingerbreadman map, 75
global asymptotic stability of the zero
equilibrium, 9, 31
global attractivity of the positive equi-
librium, 9–14
global attractor, 6
globally asymptotically stable, 6
invariant, 71
invariant interval, 41–42, 184–189
limiting solution, 15
linearized stability analysis, 5, 183
linearized stability theorem, 6
local asymptotic stability of a two cycle,
36
local stability of the positive equilibrium
, 32
local stability of the zero equilibrium, 30
locally asymptotically stable, 6
locally stable, 6
Lyness’ equation, 2, 70
May’s host parasitoid model, 75
negative feedback condition, 9
negative semicycle, 25
nonoscillatory, 25
open problems and conjectures, 43–46,
49–51, 64–68, 72–75, 123–130, 134–
136, 138–139, 163–166, 180–182,
192–200
oscillate about x, 25
oscillate about zero, 25
periodic solution, 7
periodic solutions with the same period,
33
Pielou’s discrete delay logistic model, 2,
66
Pielou’s equation, 57
Plant-herbivore system, 128
population model, 66, 129
positive equilibrium, 9
positive semicycle, 24
prime period p solution, 7
repeller, 6, 7
Riccati difference equation, 1, 17, 78
Riccati number, 18
saddle point, 7
semicycle analysis, 24–28
signum invariant, 38
sink, 7
217
218 INDEX
source, 6
stability of equilibrium, 5
stability trichotomy, 10
stable manifold, 8
stable manifold theorem in the plane, 7
strictly oscillatory, 25
table of invariant intervals, 42, 189
trichotomy character, 168
type of equation
(1, 1)-type, 47–48
(1, 2)-type, 53–54
(1, 3)-type, 131
(2, 1)-type, 69–70
(2, 2)-type, 77–78
(2, 3)-type, 141
(3, 1)-type, 137
(3, 2)-type, 167
(3, 3)-type, 183
unstable equilibrium, 6
unstable manifold, 8

DYNAMICS of SECOND ORDER RATIONAL DIFFERENCE EQUATIONS
With Open Problems and Conjectures

DYNAMICS of SECOND ORDER RATIONAL DIFFERENCE EQUATIONS
With Open Problems and Conjectures
M. R. S. Kulenovi´ c G. Ladas

CHAPMAN & HALL/CRC
Boca Raton London New York Washington, D.C.

c p.E. The consent of CRC Press LLC does not extend to copying for general distribution. Kulenovi´ . or by any information storage or retrieval system. and are used only for identification and explanation. Specific permission must be obtained in writing from CRC Press LLC for such copying. Visit the CRC Press Web site at www. QA431 .W. Trademark Notice: Product or corporate names may be trademarks or registered trademarks. microfilming. for promotion. without prior permission in writing from the publisher.S. and recording. without intent to infringe. Florida 33431. 2000 N.com © 2002 by Chapman & Hall/CRC No claim to original U. Corporate Blvd. Ladas. including photocopying. Difference equations—Numerical solutions. II. I. Reprinted material is quoted with permission. electronic or mechanical. Government works International Standard Book Number 1-58488-275-1 Library of Congress Card Number 20010370 Printed in the United States of America 1 2 3 4 5 6 7 8 9 0 Printed on acid-free paper .S. Direct all inquiries to CRC Press LLC. but the author and the publisher cannot assume responsibility for the validity of all materials or for the consequences of their use.S.R. A wide variety of references are listed. or for resale.K84 2001 515′.Library of Congress Cataloging-in-Publication Data Kulenovi´. G. Boca Raton..S. (Mustafa R. Title. ISBN 1-58488-275-1 1. Ladas. Includes bibliographical references and index. and sources are indicated.R. Neither this book nor any part may be reproduced or transmitted in any form or by any means.crcpress.) c Dynamics of second order rational difference equations : with open problems and conjectures / M. G. cm. Reasonable efforts have been made to publish reliable data and information.625—dc21 20010370 This book contains information obtained from authentic and highly regarded sources. for creating new works. M.

. . . . . . . . . . . . . . . . . . 2. . .6 The Riccati Equation . . and Invariant Intervals 2.2 Stability of the Zero Equilibrium . . .6 Local Asymptotic Stability of a Two Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v . . . . . . . . . . . . . . . . . . . .1 Introduction . . . . 2. . . . . . .3 Local Stability of the Positive Equilibrium . . . . . . . . 2. . . . . . .1 Definitions of Stability and Linearized Stability Analysis 1. . .2 The Stable Manifold Theorem in the Plane . . .Contents Preface Acknowledgements Introduction and Classification of Equation Types 1 Preliminary Results 1. . . . . . . . . . . . . . . . . . . . . . Bxn 3. . . . .5 Limiting Solutions . . . . . . . . . . . . . . . .3 The Case α = β = A = C = 0 : xn+1 = γxn−1 . . . . βxn 3. . . .7 Semicycle Analysis . .2 The Case α = γ = A = B = 0 : xn+1 = Cxn−1 3. . . . . ix xi 1 5 5 7 9 9 14 17 24 29 29 30 32 33 35 36 38 41 43 47 47 48 49 49 . . . 1. . . . . . . . Periodicity. 2. . .7 Convergence to Period-Two Solutions When C=0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8 Invariant Intervals . . . . . . . . . . . . . . . . .5 Existence of Prime Period-Two Solutions . . . . . . . . . . . . . . . . . . . . . . . Semicycles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. . . . . . . . . . . . . . .3 Global Asymptotic Stability of the Zero Equilibrium . . . . . . . . . 3 (1. . .1 Equilibrium Points . . . 2. .9 Open Problems and Conjectures . 2. . . . . 1)-Type Equations 3. . .4 When is Every Solution Periodic with the Same Period? . . . . . 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. . 2. . . . . . . . . . . . . . . . . . . . . . . .4 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . 2. 1. . . . . . . . . . . . .4 Global Attractivity of the Positive Equilibrium . . . . 2 Local Stability. . . . . . . . . . . . . . . . .

. . . . .3 The Case β = A = C = 0 : xn+1 = α+γxn . . . . . . . . . . . .1 Introduction . . .3 The Case p > 1 . . . . . . βxn A+Cxn−1 βxn Bxn +Cxn−1 γxn−1 . . . 82 83 84 87 89 90 90 91 92 93 96 97 100 101 102 103 105 107 . . . . .7. . . . . . . . . . . . . 6. . . . . . . . . . . . . . . . . . . . A+Bxn γxn−1 Bxn +Cxn−1 Pielou’s Equation . . . . . . . .8 Open Problems and Conjectures . . . . α+γxn−1 6. . . . . 6. . . . . . . .4 Open Problems and Conjectures . . .Lyness’ Equation Cxn−1 n−1 5. . . 4. . . . . 5. . . . . . . . . . . . . . . . 77 α+βx 6. . . . . . . . . . . . 2)-Type Equations 77 6. . . . . . . . 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7. . . . . . . . . . . . . . . . 6 (2. . . . .2 Invariant Intervals . . . . . . .4 .4. . . . . . . . . . . . . .5 The Case α = γ = A = 0 : xn+1 = 4. . . 4. . . . . . . . . . . . . .4. . . . .7.4. . . . . . . . . . . 6. . .2 The Case γ = A = B = 0 : xn+1 = α+βxn . . . . . . . . . . . . . 6. 54 α 4. . . . . . . . . . . . . . .3 Semicycle Analysis . . . . . . . . . . n +γx 6. . . .7. . . . . . . . . . .6. . . . . . . .2 Semicycle Analysis . . n 6. . . . . . .3 Semicycle Analysis and Global Attractivity When q = 1 . . .7. . . .6. . . . . . .1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Invariant Intervals . . . . . . . . . . .6 The Case α = β = C = 0 : xn+1 = 4. . . . . . . . . . . . . . . . . . . . . . . . .7 The Case α = β = A = 0 : xn+1 = 4. .2 The Case p = 1 . . . . . . 79 6. . . . . .3 The Case β = γ = A = 0 : xn+1 = Bxn +Cxn−1 . . . . . . . .4 Global Behavior of Solutions . 1)-Type Equations 5. . 4. . . . . . .1 Invariant Intervals . . . . . . . . . . . . . . . . . . . . .6. . . . . . 78 n α+βxn 6. . . . Bx 5. . . . 6. . . . . . . . . . . .2 The Case γ = C = 0 : xn+1 = A+Bxn . . . . . . . . . . . . . . . .6. . . . .1 The Case p < 1 . . . .7. . . . . . . .2 Semicycle Analysis of Solutions When q ≤ 1 . . . . . .5 The Case β = C = 0 : xn+1 = α+γxn−1 . . . . . . 53 α 4. . . . 6. . . . . . . . . . . . . . . . 55 4. . 5 (2. . . . . . . . . . . . . . . . 2)-Type Equations 53 4. 57 58 59 60 61 62 62 64 69 69 70 72 72 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7. . . . The Case γ = A = 0 : xn+1 = Bxα+βxn n +Cxn−1 6. . . . . . . .1 Existence and Local Stability of Period-Two Cycles . . . .2 The Case β = γ = C = 0 : xn+1 = A+Bxn . . . . . . . . .2 The Case q < 1 . . . . . . . . . . . . . . . . . . . . . 6.3 The Case γ = B = 0 : xn+1 = A+Cxn−1 . . . . . . . . . . . . . . . . .Riccati Equation . . . . . . . . . . . . . . . . .1 Introduction . . . 6.3 Global Stability and Boundedness . 6. . . . . . . . . . . . . . . . . .5. . . . . . .7 The Case α = C = 0 : xn+1 = βxA+Bxn−1 . . 6. . .5. . . . . . . . . . . . . . . . . . . . . . A+Bxn 6. . . . . . 6. . .3 The Case q > 1 . . . . . . . . . . . . . . . . . . . . . . . . .6 The Case β = A = 0 : xn+1 = Bxn +Cxn−1 . . .vi Contents 4 (1. . . . . . . . . . .4 Global Attractivity When q < 1 . . . . . . . . .1 Semicycle Analysis . .5. . . . . .4 The Case α = γ = B = 0 : xn+1 = 4.

. . . . . . . . .5 Open Problems and Conjectures . . βxn 9. . . . .3 The Case β = 0 : xn+1 = A+Bxn +Cxn−1 . . . . . .4. .1 Introduction . . . . . . .2 Invariant Intervals . .1 Introduction . . . . . . . 9. . . . . . . . . . .2. . . . . . .1 Introduction . . . . . . . . . . . . . . . . . . .2 Global Stability of the Positive Equilibrium . . .8.2 Semicycle Analysis . A+Cxn−1 6. . . . . 9. . .7. . .5 Open Problems and Conjectures .10 Open Problems and Conjectures . . . . . . . . . .Contents 6. . . . . . . . . . . . . . .5 Semicycle Analysis When pr = q + qr . . . . . . . . . . . . . . . . . . .4 The Case α = 0 : xn+1 = A+Bx+γxn−1 . . . . . . 9. . . . . . . . α+γxn−1 9.3 Global Stability Analysis When p < q . . . . . .1 Invariant Intervals and Semicycle Analysis . . . . . . . . . .4 Global Stability . . . . . . .4. 6. . . . . . . . . . . 6. . . 9. . . . . . 3)-Type Equations 131 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Semicycle Analysis . . . . . . . . . . . . . . . . . . . . . . . .2 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9. . . 9. . . . . . . .3. . . . . . . . . 137 8. . . . . . . . . . . . .3. . . . . . . . . . . . . . . .3 The Case α = γ = 0 : xn+1 = A+Bxβxn n−1 . . . . . . 132 n +Cx γx 7.2 Invariant Intervals . . . . . .4 The Case α = β = 0 : xn+1 = A+Bxnn−1 n−1 . 6. . . . . . . . . . . .1 Boundedness of Solutions . . . . . . 138 9 (2. . 9. . . .9. . . .2 The Case γ = 0 : xn+1 = A+Bxn +Cxn−1 . . . . . . .2 The Case β = γ = 0 : xn+1 = A+Bxnα n−1 . 2 9. . . . . 141 141 141 142 143 144 147 149 150 154 156 157 157 158 160 162 163 . 6. . . . . . . . . .1 Existence of a Two Cycle . . . . . . . . . . . . . . . . . 3)-Type Equations 9. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 +Cx 7. . . vii 107 109 110 112 113 114 117 117 123 123 7 (1. . . . . . . . . . . . .2. . . . . . . . . .3. . . . . . . . . . . . . . . . . 134 8 (3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . βxn +γx 6. . . .3 Convergence of Solutions . .8. . . . . 131 7. . . . 9. .2. .9 The Case α = A = 0 : xn+1 = Bxn +Cxn−1 . . . . . . . .5 Long-term Behavior of Solutions When q > 1 6.1 Invariant Intervals . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Convergence of Solutions . . . . . . 1)-Type Equations 137 8. . . . . . . . . . . . . . . . . . n +Cxn−1 9. . . . n−1 6. . . . .9. . . . . . . 9. . . . . . . .4 Global Stability Analysis When p > q .8 The Case α = B = 0 : xn+1 = βxn +γxn−1 . . . . . 6. . . . . . . . . . . . 133 +Cx 7. . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . .9. . . . . . . .2. .1 Existence and Local Stability of Period-Two Cycles 9. . . α+βxn 9. . . .4 Global Asymptotic Stability . . . . . . .9. . . . . . . . . . . . .

. . 11. . . . 10. . . . . . . . . . . . . . . . . . . .3. . . α+βx (3. . . . . . . .1 10. 10. . . . .2 11. . . . .3 The Case B = 0 : xn+1 = α+βxn +γxn−1 . . . . . . . . . . . . . . . . . . . Bxn 10. .2.2 Proof of Part (c) of Theorem 10. .1 Invariant Intervals . . . . . . . . . . . 3)-Type Equation xn+1 = A+Bxn +γxn−1 n +Cxn−1 Linearized Stability Analysis . . . . . . 2)-Type Equations 10. . .2 Convergence of Solutions . . . . . . . . . n +γxn−1 10.3. . . . . . . . . . . . . . . . . . Convergence Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . . . . . . . . . .viii 10 (3.4. . . . . . .3 Convergence of Solutions . . . . . . . . . . . . . . . . . .1 11. .4. . . . . . . . . . . . . . . . . . . . . . 10. . . . . . . . . . . . . . . . . . . . . . .5 Open Problems and Conjectures . 11. . . . . . . . . . . . . .1 Introduction . . . . . . . . . A+Bxn 10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3. . .2 Global Stability When p ≥ r . . . . . . . . . . .1 10. . . . .2. .3. . . . . . . . . . . . Invariant Intervals . . . . .1 Proof of Part (b) of Theorem 10. . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . . .4 Open Problems and Conjectures . . . . . . . . 10. A Global Attractivity for Higher Order Equations Bibliography Index .3 Contents 167 167 167 169 172 172 173 173 174 175 177 179 180 183 183 184 189 191 192 201 211 217 . . . . . . . . . 10. . . . . . . 11 The 11. . . . . . .4 The Case A = 0 : xn+1 = α+βx+Cxn−1 . . . . . . . . . . .2 The Case C = 0 : xn+1 = α+βxn +γxn−1 . . . A+Cxn−1 10. . . . . .1 Invariant Intervals . . . . . . .1 Global Stability . . . . . .

In particular we present necessary and sufficient conditions for Eq(1) to have period-two solutions and we also present a table of invariant intervals. A + Bxn + Cxn−1 n = 0. Every chapter in this monograph contains several open problems and conjectures related to the particular equation which is investigated and its generalizations. we present some global attractivity results for higher order difference equations which may be useful for extensions and generalizations. we present the known results and derive several new ones on the boundedness. . (1) where the parameters α. A. It is an amazing fact that Eq(1) contains. Chapter 1 contains some basic definitions and some general results which are used throughout the book. In Chapters 3 to 11 we present the known results and derive several new ones on the various types of equations which comprise Eq(1). as special cases. a large number of equations whose dynamics have not been thoroughly understood yet and remain a great challenge for further investigation. and the periodicity of solutions of all rational difference equations of the form xn+1 = α + βxn + γxn−1 . The techniques and results which we develop in this monograph to understand the dynamics of Eq(1) are also extremely useful in analyzing the equations in the mathematical models of various biological systems and other applications. In this sense. To this end we pose several Open Problems and Conjectures which we believe will stimulate further interest towards a complete understanding of the dynamics of Eq(1) and its functional generalizations.Preface In this monograph. In an appendix. ix . 1. γ. β. We believe that the results about Eq(1) are of paramount importance in their own right and furthermore we believe that these results offer prototypes towards the development of the basic theory of the global behavior of solutions of nonlinear difference equations of order greater than one. In Chapter 2 we present some general results about periodic solutions and invariant intervals of Eq(1). this is a self-contained monograph and the main prerequisite that the reader needs to understand the material presented here and to be able to attack the open problems and conjectures is a good foundation in analysis. . at the end of this monograph. C and the initial conditions x−1 and x0 are nonnegative real numbers. B. the global stability. .

.

S. Merino. Hoag. Radin. Valicenti. C. C. E. L. xi . Camouzis. M. E. R. P. Rodrigues. Predescu. L. A. I. Teixeira. Prokup. Cunningham. We are thankful to Professors R. K. and K. Feuer. McGrath. F. C. Clark. T. Wilkinson for their enthusiastic participation and useful suggestions which helped to improve the exposition. Sizer and to our graduate students A. DeVault. J. W. Gibbons. El-Metwally. M.Acknowledgements This monograph is the outgrowth of lecture notes and seminars which were given at the University of Rhode Island during the last five years. W. J. C. Driver. W. Schultz. E. M. D. Kent. Janowski. Kosmala. Amleh. Briden. and W. O. A. S. S. Grove. N. H. Martins. C.

.

To Senada and Theodora .

.

A + Bxn + Cxn−1 n = 0. . . in addition to the basic description of the solutions 1 . (1) where the parameters α. etc. such as [1]. the global stability. [42]. We believe that the results about Eq(1) are of paramount importance in their own right and furthermore we believe that these results offer prototypes towards the development of the basic theory of the global behavior of solutions of nonlinear difference equations of order greater than one. To this end we pose several Open Problems and Conjectures which we believe will stimulate further interest towards a complete understanding of the dynamics of Eq(1) and its functional generalizations. [41]. See Section 1. The techniques and results which we develop in this monograph to understand the dynamics of Eq(1) are also extremely useful in analyzing the equations in the mathematical models of various biological systems and other applications. . γ.Introduction and Classification of Equation Types In this monograph. A. a large number of equations whose dynamics have not been thoroughly established yet. [21]. 1. It is an amazing fact that Eq(1) contains as special cases. This equation is studied in many textbooks on difference equations. The Riccati difference equation when γ = C = 0. and the periodicity of solutions of all rational difference equations of the form xn+1 = α + βxn + γxn−1 . C are nonnegative real numbers and the initial conditions x−1 and x0 are arbitrary nonnegative real numbers such that A + Bxn + Cxn−1 > 0 for all n ≥ 0. Eq(1) includes among others the following well known classes of equations: 1. B. β.6 where. we present the known results and derive several new ones on the boundedness.

A characteristic feature of this equation is that it possesses an invariant which can be used to understand the character of its . . Lyness’ equation when γ = A = B = 0.4. This is a fascinating equation with many interesting extensions and generalizations. . x0 ) ∈ R × R through which the equation xn−1 xn+1 = −1 + xn is well defined for all n ≥ 0. [32]. and Section 4. See [24] and Sections 2. See [13]. Hence the solution {xn } of Eq(2) exists for all n ≥ 0. 3. 75). [18]. [49]. 1. . Throughout this monograph we pose several open problems related to the existence of solutions of some simple equations with the hope that their investigation may throw some light into this very difficult problem. See ([42]. This is the set F of all initial conditions x0 ∈ R through which the equation xn+1 = α + βxn A + Bxn (2) is not well defined for all n ≥ 0. [67].2.7 and 10. p. [55]. Pielou’s discrete delay logistic model when α = γ = B = 0. 2. 1 + xn n = 0. . and Section 1. For example. it is interesting to note the similarities of the dynamics of the population model xn+1 = α + βxn−1 e−xn . See [42]. See [13]. . An example of such a problem is the following: Open Problem Find all initial points (x−1 . . We believe that the techniques which we develop in this monograph to understand the dynamics of Eq(1) will also be of paramount importance in analyzing the equations in the mathematical models of various biological systems and other applications. The problem of existence of solutions for difference equations is of paramount importance but so far has been systematically neglected.6 for some results in this regard. n = 0. [44]. [69]. and Section 5. 1. [57]. / if and only if x0 ∈ F. we introduce the forbidden set of the equation. . [66].2 Introduction and Classification of the Riccati equation. and the rational equation xn+1 = α + αxn + βxn−1 .2.

. 3) 3 equations of type (3. n = 0. Some of them have been recently investigated and have led to the development of some general theory about difference equations. We will say that a special case of Eq(1) is of (k. that every positive solution of the difference equation α + xn . 3). . or linear. . . Let k and l be positive integers from the set {1. a large number of equations whose dynamics have not been thoroughly established yet. . n = 0. n = 0. and (2. and to bring to their attention several Open Research Problems and Conjectures related to this equation and its functional generalization. However to this day. 1) 3 equations of type (2. [28]. 1. For convenience we classify all the special cases of Eq(1) as follows. . . 3) 3 equations of type (3. 1. Eq(1) contains 49 equations as follows: 9 equations of type (1. . [43]. Of these 49 equations. as special cases. Our goal in this monograph is to familiarize readers with the recent techniques and results related to Eq(1) and its extensions. 1. xn+1 = xn−1 βxn + γxn−1 . For example the following equations xn . 2. . See [7]. is bounded. for example. . [45] and [52]-[54]. [15]. (2. l)-type if the equation is of the form of Eq(1) with k positive parameters in the numerator and l positive parameters in the denominator. The remaining 28 are quite interesting nonlinear difference equations. 2) 1 equation of type (3. . 1. . . ∞). (2. 2) 9 equations of type (2. 3}. It is an amazing fact that Eq(1) contains. 1). 1) 9 equations of type (2. 1). xn+1 = A + Bxn + Cxn−1 are of types (1. [42]. 21 are either trivial. 3) respectively. γ + xn−1 1 + xn . [29]. . n = 0. 1. or of the Riccati type. .Introduction and Classification 3 solutions. 2) 3 equations of type (1. 2). xn+1 = xn−1 where α ∈ (0. many special cases of Eq(1) are not thoroughly understood yet and remain a great challenge for further investigation. . In this sense. xn+1 = xn−1 α + βxn xn+1 = . Without the use of this invariant we are still unable to show. n = 0. 1) 9 equations of type (1.

.

1. [63]. . Then for every set of initial conditions x0 . x).1). See also [1]. [22]. [21].Chapter 1 Preliminary Results In this chapter we state some known results and prove some new ones about difference equations which will be useful in our investigation of Eq(1).1). and [68].1) has a unique solution {xn }∞ . xn = x for n ≥ 0 is a solution of Eq(1.1.1 Definitions of Stability and Linearized Stability Analysis f :I ×I →I Let I be some interval of real numbers and let be a continuously differentiable function. that is. The reader may just glance at the results in this chapter and return for the details when they are needed in the sequel. [16]. [33].1) if x = f (x. .1 Let x be an equilibrium point of Eq(1. 5 . [4]. n = 0. (1. xn−1 ). Definition 1. x is a fixed point of f . or equivalently. the difference equation xn+1 = f (xn . n=−1 A point x ∈ I is called an equilibrium point of Eq(1. x−1 ∈ I. . [41]. [34]. 1.

and if there exists γ > 0 such that for all x0 . Theorem 1. x−1 ∈ I with |x0 − x| + |x−1 − x| < δ. (1.1) is called a source. 1. . x−1 ∈ I with 0 < |x0 − x| + |x−1 − x| < r. x−1 ∈ I with |x0 − x| + |x−1 − x| < γ.3) Let lie in the open unit disk |λ| < 1. (vi) The equilibrium x of Eq(1.1) is called a global attractor if for every x0 . (ii) The equilibrium x of Eq(1.1) is called globally asymptotically stable if it is locally stable and a global attractor.6 Chapter 1. ∂f ∂f (x.1. Clearly a source is an unstable equilibrium.1) is called locally stable if for every ε > 0. we have |xn − x| < ε for all n ≥ −1. Preliminary Results (i) The equilibrium x of Eq(1. there exists δ > 0 such that for all x0 . . . n→∞ (iv) The equilibrium x of Eq(1. x−1 ∈ I we have lim xn = x.1 (Linearized Stability) (a) If both roots of the quadratic equation λ2 − pλ − q = 0 (1. n = 0.1) about the equilibrium point x.1).1) is locally asymptotically stable. we have n→∞ lim xn = x. x) ∂u ∂v denote the partial derivatives of f (u. there exists N ≥ 1 such that |xN − x| ≥ r. or a repeller.1) is called locally asymptotically stable if it is locally stable.2) p= is called the linearized equation associated with Eq(1.1) is called unstable if it is not stable. x) and q = (x. (iii) The equilibrium x of Eq(1. (v) The equilibrium x of Eq(1. v) evaluated at the equilibrium x of Eq(1. if there exists r > 0 such that for all x0 . then the equilibrium x of Eq(1. . Then the equation yn+1 = pyn + qyn−1 .

1.2. The Stable Manifold Theorem in the Plane

7

(b) If at least one of the roots of Eq(1.3) has absolute value greater than one, then the equilibrium x of Eq(1.1) is unstable. (c) A necessary and sufficient condition for both roots of Eq(1.3) to lie in the open unit disk |λ| < 1, is |p| < 1 − q < 2. (1.4) In this case the locally asymptotically stable equilibrium x is also called a sink. (d) A necessary and sufficient condition for both roots of Eq(1.3) to have absolute value greater than one is |q| > 1 and |p| < |1 − q| . In this case x is a repeller. (e) A necessary and sufficient condition for one root of Eq(1.3) to have absolute value greater than one and for the other to have absolute value less than one is p2 + 4q > 0 and |p| > |1 − q| .

In this case the unstable equilibrium x is called a saddle point. (f ) A necessary and sufficient condition for a root of Eq(1.3) to have absolute value equal to one is |p| = |1 − q| or q = −1 and |p| ≤ 2. In this case the equilibrium x is called a nonhyperbolic point. Definition 1.1.2 if (a) A solution {xn } of Eq(1.1) is said to be periodic with period p xn+p = xn for all n ≥ −1. (1.5)

(b) A solution {xn } of Eq(1.1) is said to be periodic with prime period p, or a p-cycle if it is periodic with period p and p is the least positive integer for which (1.5) holds.

1.2

The Stable Manifold Theorem in the Plane

Let T be a diffeomorphism in I × I, i.e., one-to-one, smooth mapping, with smooth inverse. Assume that p ∈ I × I is a saddle fixed point of T, i.e., T(p) = p and the Jacobian JT (p) has one eigenvalue s with |s| < 1 and one eigenvalue u with |u| > 1.

8

Chapter 1. Preliminary Results

Let vs be an eigenvector corresponding to s and let vu be an eigenvector corresponding to u. Let S be the stable manifold of p, i.e., the set of initial points whose forward orbits (under iteration by T) p, T(p), T2 (p), . . . converge to p. Let U be the unstable manifold of p, i.e., the set of initial points whose backward orbits (under iteration by the inverse of T) p, T−1 (p), T−2 (p), . . . converge to p. Then, both S and U are one dimensional manifolds (curves) that contain p. Furthermore, the vectors vs and vu are tangent to S and U at p, respectively. The map T for Eq(1.1) is found as follows: Set un = xn−1 and vn = xn Then un+1 = xn and so T u v = v f (v, u) x x . and vn+1 = f (vn , un ) for n ≥ 0

for n ≥ 0.

Therefore the eigenvalues of the Jacobian JT

at the equilibrium point x of Eq(1.1)

are the roots of Eq(1.3). In this monograph, the way we will make use of the Stable Manifold Theorem in our investigation of the rational difference equation xn+1 = α + βxn + γxn−1 , A + Bxn + Cxn−1 n = 0, 1, . . . (1.6)

is as follows: Under certain conditions the positive equilibrium of Eq(1.6) will be a saddle point and under the same conditions Eq(1.6) will have a two cycle . . . φ, ψ, φ, ψ, . . . (1.7)

which is locally asymptotically stable. Under these conditions, the two cycle (1.7) cannot be a global attractor of all non-equilibrium solutions. See, for example, Sections 6.6 and 6.9.

1.3. Global Asymptotic Stability of the Zero Equilibrium

9

1.3

Global Asymptotic Stability of the Zero Equilibrium

When zero is an equilibrium point of the Eq(1.6), which we are investigating in this monograph (with β > 0), then the following theorem will be employed to establish conditions for its global asymptotic stability. Amazingly the same theorem is a powerful tool for establishing the global asymptotic stability of the zero equilibrium of several biological models. See, for example, [31] and [48]. Theorem 1.3.1 ([31]. See also [48], and [70]) Consider the difference equation xn+1 = f0 (xn , xn−1 )xn + f1 (xn , xn−1 )xn−1 , with nonnegative initial conditions and f0 , f1 ∈ C[[0, ∞) × [0, ∞), [0, 1)]. Assume that the following hypotheses hold: (i) f0 and f1 are non-increasing in each of their arguments; (ii) f0 (x, x) > 0 for all x ≥ 0; (iii) f0 (x, y) + f1 (x, y) < 1 for all x, y ∈ (0, ∞). Then the zero equilibrium of Eq(1.8) is globally asymptotically stable. n = 0, 1, . . . (1.8)

1.4

Global Attractivity of the Positive Equilibrium

Unfortunately when it comes to establishing the global attractivity of the positive equilibrium of the Eq(1.6) which we are investigating in this monograph, there are not enough results in the literature to cover all various cases. We strongly believe that the investigation of the various special cases of our equation has already played and will continue to play an important role in the development of the stability theory of general difference equations of orders greater than one. The first theorem, which has also been very useful in applications to mathematical biology, see [31] and [48], was really motivated by a problem in [35]. Theorem 1.4.1 ([31]. See also [35], [48], and ([42], p. 53). Let I ⊆ [0, ∞) be some interval and assume that f ∈ C[I × I, (0, ∞)] satisfies the following conditions: (i) f (x, y) is non-decreasing in each of its arguments; (ii) Eq(1.1) has a unique positive equilibrium point x ∈ I and the function f (x, x) satisfies the negative feedback condition: (x − x)(f (x, x) − x) < 0 for every x ∈ I − {x}.

Then every positive solution of Eq(1.1) with initial conditions in I converges to x.

10

Chapter 1. Preliminary Results

The following theorem was inspired by the results in [55] that deal with the global asymptotic stability of Pielou’s equation (see Section 4.4), which is a special case of the equation that we are investigating in this monograph. Theorem 1.4.2 ([42], p. 27) Assume (i) f ∈ C[(0, ∞) × (0, ∞), (0, ∞)]; (ii) f (x, y) is nonincreasing in x and decreasing in y; (iii) xf (x, x) is increasing in x; (iv) The equation xn+1 = xn f (xn , xn−1 ), n = 0, 1, . . . has a unique positive equilibrium x. Then x is globally asymptotically stable. Theorem 1.4.3 ([27]) Assume (i) f ∈ C[[0, ∞) × [0, ∞), (0, ∞)]; (ii) f (x, y) is nonincreasing in each argument; (iii) xf (x, y) is nondecreasing in x; (iv) f (x, y) < f (y, x) ⇐⇒ x > y; (v) The equation xn+1 = xn−1 f (xn−1 , xn ), n = 0, 1, . . . (1.10)

(1.9)

has a unique positive equilibrium x. Then x is a global attractor of all positive solutions of Eq(1.10). The next result is known as the stability trichotomy result: Theorem 1.4.4 ([46]) Assume f ∈ C 1 [[0, ∞) × [0, ∞), [0, ∞)] is such that x ∂f ∂f +y < f (x, y) for all x, y ∈ (0, ∞). ∂x ∂y (1.11)

Then Eq(1.1) has stability trichotomy, that is exactly one of the following three cases holds for all solutions of Eq(1.1): (i) limn→∞ xn = ∞ for all (x−1 , x0 ) = (0, 0) . (ii) limn→∞ xn = 0 for all initial points and 0 is the only equilibrium point of Eq(1.1). (iii) limn→∞ xn = x ∈ (0, ∞) for all (x−1 , x0 ) = (0, 0) and x is the only positive equilibrium of Eq(1.1).

at the end of this monograph. Then Eq(1. ≤ M1 ≤ M0 and Set mi ≤ xk ≤ Mi m = lim mi i→∞ for k ≥ 2i + 1. . y) is nondecreasing in both arguments. y) is monotonic in its arguments. b] × [a. . y) is non-decreasing in x ∈ [a. . (See Section 6. M ) = m and f (M. m) = M. M ) ∈ [a. set Mi = f (Mi−1 . b] × [a. b] is a continuous function satisfying the following properties: (a) f (x. 2. . b] for each x ∈ [a.1. See also the Appendix.4 and 6. b] is a solution of the system f (m.1) has a unique equilibrium x ∈ [a. b] for each y ∈ [a. b]. or nonincreasing in one and nondecreasing in the other. ≤ mi ≤ . y) is nonincreasing in y ∈ [a. ≤ Mi ≤ . i→∞ and M = lim Mi . or nonincreasing in both arguments. Proof. y) is non-increasing in x and nondecreasing in y.9. The next two theorems are slight modifications of results which were obtained in [54] and were developed while we were working on a special case of the equation that we are investigating in this monograph. . Theorems 1. y). . . b]. b] be an interval of real numbers and assume that f : [a. (b) If (m.1) is to work in the regions where the function f (x. b] → [a.4. .5 ([54]) Let [a. which describes the extension of these results to higher order difference equations. then m = M . Now observe that for each i ≥ 0.4. . mi−1 ) and mi = f (mi−1 . See Sections 6. and f (x. Global Attractivity of the Positive Equilibrium 11 Very often the best strategy for obtaining global attractivity results for Eq(1. Mi−1 ).8 and 1.7 below are new and deal with the remaining two cases relative to the monotonic character of f (x.) The first theorem deals with the case where the function f (x. m0 ≤ m1 ≤ . Theorem 1.4.8 where these theorems are utilized. In this regard there are four possible scenarios depending on whether f (x. Set m0 = a and M0 = b and for i = 1. y) is non-decreasing in x and non-increasing in y and the second theorem deals with the case where the function f (x. b] and every solution of Eq(1.4. .1) converges to x.

M ) and M = f (M.12) and by the continuity of f . b] → [a. In view of (b). b] is a continuous function satisfying the following properties: (a) f (x. y) is non-increasing in x ∈ [a.4. b]. b] × [a. . . m = f (m. 2 . Now observe that for each i ≥ 0. Mi−1 ) and mi = f (Mi−1 .12 Then Chapter 1. b] be an interval of real numbers and assume that f : [a.1) has no solutions of prime period two in [a. M ). m) and M = f (m. . m=M from which the result follows.12) holds and by the continuity of f . ≤ M1 ≤ M0 . . ≤ mi ≤ . . Then clearly (1. . and Set mi ≤ xk ≤ Mi m = lim mi i→∞ for k ≥ 2i + 1. . b] for each x ∈ [a. 2 Theorem 1. b]. 2. Preliminary Results M ≥ lim sup xi ≥ lim inf xi ≥ m i→∞ i→∞ (1. i→∞ and M = lim Mi . b]. (b) The difference equation Eq(1. m).6 ([54]) Let [a.1) has a unique equilibrium x ∈ [a. Therefore in view of (b). Proof. b] and every solution of Eq(1. and f (x. b] for each y ∈ [a. ≤ Mi ≤ . y) is nondecreasing in y ∈ [a. mi−1 ). Then Eq(1. set Mi = f (mi−1 . m=M from which the result follows. Set m0 = a and M0 = b and for i = 1. m0 ≤ m1 ≤ .1) converges to x. m = f (M. . .

. set Mi = f (mi−1 .8 Let [a.1) converges to x.4. b] → [a. Then clearly (1. ≤ M1 ≤ M0 . In view of (b). Then Eq(1. Set m0 = a and M0 = b and for i = 1. ≤ Mi ≤ . M ) ∈ [a. Now observe that for each i ≥ 0. 13 then m = M . m). b] and every solution of Eq(1. . b] × [a. x) = x. Proof. b] × [a.1. . . . mi−1 ) and mi = f (Mi−1 . y) is non-decreasing in each of its arguments. M ) and M = f (m. has a unique positive solution.4. ≤ mi ≤ . . Then Eq(1. M ) = m. . m0 ≤ m1 ≤ . b] and every solution of Eq(1. 2. b] is a continuous function satisfying the following properties: (a) f (x. b] is a continuous function satisfying the following properties: (a) f (x. Mi−1 ).7 Let [a. .1) has a unique equilibrium x ∈ [a. (b) If (m.1) converges to x. b] be an interval of real numbers and assume that f : [a. m) = M and f (M. i→∞ and M = lim Mi . b] × [a. (b) The equation f (x. b] → [a. Global Attractivity of the Positive Equilibrium Theorem 1. and Set mi ≤ xk ≤ Mi m = lim mi i→∞ for k ≥ 2i + 1. b] is a solution of the system f (m.4. y) is non-increasing in each of its arguments. .12) holds and by the continuity of f . m = f (M.1) has a unique equilibrium x ∈ [a. . 2 Theorem 1. b] be an interval of real numbers and assume that f : [a. m=M =x from which the result follows.

Now with the above hypotheses. Mi−1 ) and mi = f (mi−1 . . ≤ mi ≤ . Let L be a limit point of the solution {xn }∞ n=−1 (For example L is the limit superior S or the limit inferior I of the solution). Preliminary Results m0 = a and M0 = b and for i = 1.14 Proof. . m = f (m. Set Chapter 1. . 2 1. m = M = x. . Assume that {xn }∞ n=−1 is a solution of Eq(1. In this section we first give a self-contained version of limiting solutions the way we will use them in this monograph. In view of (b). and mi ≤ xk ≤ Mi Set m = lim mi i→∞ for k ≥ 2i + 1. 2. n ≥ −k + 1. .12) is satisfied and by the continuity of f . J]. . (1. where A. ≤ M1 ≤ M0 . m) and M = f (M. m0 ≤ m1 ≤ . Let J be some interval of real numbers. Then clearly the inequality (1. i→∞ and M = lim Mi . . and let k ≥ 1 be a positive integer.13) . set Mi = f (Mi−1 . from which the result follows. ∞ we will show that there exists a solution {Ln }n=−k of the difference equation xn+1 = f (xn .1) such that A ≤ xn ≤ B for all n ≥ −1. . . Now observe that for each i ≥ 0. ≤ Mi ≤ . . mi−1 ). B ∈ J. f ∈ C[J × J. M ). xn−1 ). .5 Limiting Solutions The concept of limiting solutions was introduced by Karakostas [37] (see also [40]) and is a useful tool in establishing that under certain conditions a solution of a difference equation which is bounded from above and from below has a limit. .

13) and every term of this sequence is a limit point of the solution {xn }∞ . L−k ) j→∞ = lim f (xnj −k+1 .5. n=−1 is called a limiting solution of Eq(1. .14) associated with the solution {xn }∞ n=−1 of Eq(1. Limiting Solutions such that L0 = L 15 and for every N ≥ −k. L1 = 1. .13) with Then clearly. the term LN is a limit point of the solution {xn }∞ . . Karakostas takes k = ∞ and calls the solution {Ln }∞ n=−∞ a full limiting sequence of the difference equation xn+1 = f (xn . j→∞ lim xnj −k = L−k .1). . satisfies Eq(1. B] and so it has a ∞ further subsequence {xnij }j=1 which converges to some limit which we call L−1 . .. .. . = f( −k+1 . L−k+1 . which is a consequence of the above discussion. . .13) associated with the The solution {Ln }∞ n=−k solution {xn }∞ n=−1 of Eq(1. −k ) = f (L−k+1 . n=−1 The following result. L−1 . . It follows by induction that the sequence L−k .1. L−k and a subsequence {xnj }n=−1 such that j→∞ lim xnj = L j→∞ lim xnj −1 = L−1 . ∞ Next the subsequence {xni −1 }i=1 of {xn }∞ n=−1 also lies in interval [A. −k Let { n }∞ n=−k be the solution of Eq(1.5. i→∞ i→∞ Continuing in this way and by considering further and further subsequences we obtain ∞ numbers L−2 . −k+2 j→∞ = L−k and −k+1 = L−k+1 . (1. is stated in the way we will use it in this monograph: . xn−1 ). L−3 . Therefore. xnj −k ) = lim xnj −k+2 = L−k+2 .) ∞ We now proceed to show the existence of a limiting solution {Ln }n=−k . L0 = L = 0 . (See Theorem 1.2 below. Clearly there exists a subsequence {xni }∞ of {xn }∞ i=1 n=−1 such that i→∞ lim xni = L. . lim xnij = L and lim xnij −1 = L−1 .1).

5. .1 Let J be some interval of real numbers. . as developed by Karakostas. and let {xn }∞ n=−ν be a bounded solution of the difference equation xn+1 = f (xn . 1. Then there exist two solutions ∞ {In }∞ n=−∞ and {Sn }n=−∞ of the difference equation xn+1 = f (xn . i=0 xn−i k n = 0. . f ∈ C[J ν+1 . Theorem 1. In . n→∞ S = lim sup xn n→∞ n = 0. . with I. S] for all n ∈ Z and such that for every N ∈ Z. and let k ≥ 1 be a positive integer. . (1. 0. . xn−1 . 1.1) with limit inferior I and limit superior S. IN and SN are limit points of {xn }∞ . . 1. S] for n ≥ −k. S ∈ J.2 to difference equations of the form xn+1 = Ai . . . .13) has two solutions {In }∞ n=−k with the following properties: and {Sn }∞ n=−k I0 = I. . Suppose that {xn }∞ n=−1 is a bounded solution of Eq(1. there exist two subsequences {xrn } and {xln } of the solution {xn }∞ n=−ν such that the following are true: n→∞ (1. J].16) lim xrn +N = IN and n→∞ lim xln +N = SN for every N ≥ m. . f ∈ C[J × J. xn−1 . with I0 = I. . Then Eq(1. . with I = lim inf xn . xn−ν ). 1. n = 0. n=−ν Furthermore for every m ≤ −ν. Sn ∈ [I. S ∈ J. . and In . .}. . S0 = S We now state the complete version of full limiting sequences. S0 = S. See [23] and [25] for an application of Theorem 1. . .2 ([37]) Let J be some interval of real numbers.16) associated with the solution {xn }∞ n=−k of Eq(1. .16) are called full limiting solutions of Eq(1. xn−ν ) which satisfy the equation for all n ∈ Z. Let Z denote the set of all integers {. . Sn ∈ [I. which by the change of variables xn = y1 reduce to rational difference equations of the n form 1 . −1.15).16 Chapter 1. ∞ The solutions {In }∞ n=−∞ and {Sn }n=−∞ of the difference equation (1.5. . . .5. J]. Preliminary Results Theorem 1. yn+1 = k i=0 Ai yn−i .15) and with I.

(1. 1.1.) The other goal we have is to describe in detail the long. which is extracted from [32]. B. γ.18).17) reduces to βA + βxn β xn+1 = B = for all n ≥ 0. To avoid degenerate cases. we will assume that β + A = 0.1 Assume that (1. we will assume that B = 0 and αB − βA = 0. C and the initial conditions x−1 . A. (1. β. A + Bxn A difference equation of the form xn+1 = n = 0. Eq(1. B and the initial condition x0 are real numbers is called a Riccati difference equation.19) β + A = 0. Then (1. in addition to (1.17). . .17) will become zero for some value of n ≥ 0 is called the forbidden set F of Eq(1.18) Indeed when B = 0. One of our goals in this section is to determine the forbidden set F of Eq(1. In the sequel. (1. β. x0 are all real numbers) other than the material presented here.17) regarding period-two solutions.17).19) holds every solution of Eq(1. Furthermore when (1.17) is a linear equation and when αB − βA = 0.17) with x0 = is periodic with period two.17) possesses a prime periodtwo solution.20) . Theorem 1. . A + Bxn B The set of initial conditions x0 ∈ R through which the denominator A+Bxn in Eq(1. (See also [13] and [69].and short-term behavior of / solutions of Eq(1. The proof is straightforward and will be omitted.18) holds and that Eq(1. Eq(1.21) β B (1. The first result in this section addresses a very special case of Eq(1. A.6 The Riccati Equation α + βxn .17) when x0 ∈ F.6. The Riccati Equation 17 1.6.17) where the parameters α. It should be mentioned here that there is practically nothing known about the forbidden set of Eq(1) (when the parameters α.

(1. It is interesting to note that the four parameters of Eq(1.17) have been reduced to the single parameter R of Eq(1. Preliminary Results for n ≥ 0 (1. From Eq(1. Similar results can be obtained for Eq(1.22) transforms Eq(1.17) through the change of variables (1. . (A + Bx)2 of the function For the remainder of this section we focus on Eq(1. . wn n = 0. If we set y = g(w) = 1 − then we note that the two asymptotes x=− of the function A B and y = β B R w α + βx A + Bx have now been reduced to the two asymptotes y = f (x) = w = 0 and y = 1 R .23) and present its forbidden set F and the character of its solutions.23) βA − αB (β + A)2 is a nonzero real number.22).18 Now observe that the change of variable xn = β+A A wn − B B Chapter 1. 1. called the Riccati number of Eq(1.23) it follows that when R< 1 4 .23). w Also note that the signum of R is equal to the signum of the derivative y =1− f (x) = βA − αB .17) into the difference equation wn+1 = 1 − where R= R .17). .

24) yn+1 yn for n = 0.23).1. (1. 4 1− √ √ 1 − 4R 1 + 1 − 4R < = w+ 2 2 19 When Eq(1. Also w− < 0 if R < 0 and 1 w− > 0 if R ∈ (0. . which can be solved explicitly. n = 0. . . 1 w= . The forbidden set of Eq(1. when 1 R> . Now observe that the change of variables wn = with y0 = 1 and y1 = w0 reduces Eq(1. respectively. ). 1. where 4 the characteristic roots of Eq(1. . 1. . .23) has the two equilibrium points w− and w+ with w− = and |w− | < w+ .23) to the second order linear difference equation yn+2 − yn+1 + Ryn = 0. 4 Eq(1.23) is clearly the set of points where the sequence {yn }∞ n=0 vanishes. The Riccati Equation Eq(1.24) are real numbers.23) has no equilibrium points.24) are the equilibrium points of Eq(1. 4 1 R= . 2 Finally.23) has exactly one equilibrium point namely. Note that in these two cases the characteristic roots of Eq(1. . 1 The next two theorems deal with the cases R < 1 and R = 4 .6.

if R > 0 and fn < w− while (fn − w− )(fn+1 − w− ) < 0 if R < 0.23).23) is the sequence of points fn = n−1 n−1 w+ − w− w+ w− n n w+ − w− for n = 1. (1. Preliminary Results √ √ 1 − 4R 1 + 1 − 4R w− = and w+ = 2 2 are the only equilibrium points of Eq(1.23). 4 1− Chapter 1. . . .6. wn = n n (w0 − w− )w+ − (w+ − w0 )w− n = 1. 2. . The equilibrium w+ is a global attractor as long as w0 ∈ F ∪ {w− } / and the equilibrium w− is a repeller. 2. the solution of Eq(1. Theorem 1.2 that n→∞ lim fn = w− . 2. 4 1 2 is the only equilibrium point of Eq(1.3 Assume Then w= 1 R= . .2 Assume Then 1 R< .6. (1.25) When w0 ∈ F.23) is the sequence of points fn = n−1 2n for n = 1. .6.27) . .23) is given by / n+1 n+1 (w0 − w− )w+ − (w+ − w0 )w− . . . .26) It is now clear from Theorem 1.20 Theorem 1. The forbidden set F of Eq(1. . The forbidden set F of Eq(1. . (1.

When w0 ∈ F. . .3 it follows that the equilibrium w = 1 of Eq(1. . (1.1. . . .6.29) n Finally consider the case √ R. 1. the solution of Eq(1. 4R − 1 n = 0. . 1 R> . Hence wn = √ √ 4R − 1 cos(n + 1)φ + (2w0 − 1) sin(n + 1)φ √ . . 4 The characteristic roots of Eq(1.6. 1. The Riccati Equation 21 which converges to the equilibrium from the left. (1. R 4R − 1 cos(nφ) + (2w0 − 1) sin(nφ) . . 1.23) is given by / wn = 1 + (2w0 − 1)(n + 1) 2 + 2(2w0 − 1)n for n = 0.28) From Theorem 1. 2w0 − 1 C1 = 1 and C2 = √ .24) in this case are √ 4R − 1 1 λ± = ± i 2 2 with |λ± | = Let be such that 1 cos φ = √ 2 R Then with C1 = y0 = 1 and It follows that √ R(C1 cos φ + C2 sin φ) = y1 = w0 . ) 2 and √ 4R − 1 √ sin φ = .23) is a global 2 / attractor for all solutions with w0 ∈ F but is locally unstable. it is a sink from the right and a source from the left. π φ ∈ (0. more precisely. . 2 R n = 0. yn = R 2 (C1 cos(nφ) + C2 sin(nφ)).

For example. 2 R Then the forbidden set F of Eq(1.31) from which the character of the solutions of Eq(1. that is. 2 n = 0. when R > 1 . . Theorem 1. . π ) be such that 2 2 √ 4R − 1 2x0 − 1 and sin θ = cos θ = r r where r= Then we obtain. φ= π k π ∈ (0. . Preliminary Results as long as the denominator is different from zero. wn = (4R − 1)2 + (2x0 − 1)2 .23) is given by (1. 2 2 When w0 ∈ F. (1. assume φ is a rational multiple of π.4 Assume that R> and let φ ∈ (0.23). ) N 2 (1. . 1. √ 4R − 1 1 cot(nφ) for n = 1. .6. π ) be such that 2 1 cos φ = √ 2 R and 1 4 √ 4R − 1 √ sin φ = . . 2. . . .23) is the sequence of points √ 1 4R − 1 fn = − cot(nφ) for n = 1. π ) is a rational multiple of π. / (1.29).32) . the solution of Eq(1. w0 = − 2 2 The following theorem summarizes the above discussion. can now be easily 4 revealed. that is. . and so √ cos(nφ + φ − θ) √ cos(nφ − θ) cos φ − sin(nφ − θ) sin φ R = R cos(nφ − θ) cos(nφ − θ) √ = R (cos φ − sin φ tan(nφ − θ)) 1 wn = − 2 √ 4R − 1 tan(nφ − θ).30) 1 Clearly when R > 4 the character of the solution {wn } and the forbidden set F depends on whether or not the number φ ∈ (0. 2. 2 Let θ ∈ (− π . .22 Chapter 1.

2. The Riccati Equation 23 for some k.5 Assume that φ= k π π ∈ (0.5 is not a rational multiple of π. 2 R 2 R or equivalently k 4R cos2 ( π) = 1. .6.23) with √ sin(n k π) − 4R − 1 cos(n k π) p p w0 = k 2 sin(n p π) is periodic with period p.23) is periodic with period N . ) p 2 1 4 cos2 φ with φ = k π N Then where k and p are positive comprimes and suppose that √ 4R − 1 1 √ cos φ = √ and sin φ = . Theorem 1. .6. √ √ 1 1 4R − 1 4R − 1 wN = − tan(kπ − θ) = − tan(−θ) = w0 2 2 2 2 and so every solution of Eq(1.1.23) is periodic.} with 2k < N. N ∈ {1.5 is not a rational multiple of π.6. . . . The following theorem summarizes the above discussion. 2.6. . .} and 2k < N. (ii) The set of limit points of a solution of Eq(1. .23) is periodic with period N . then the following result is true: Theorem 1. Equivalently assume that A= where k. / for n = 1. p Then every solution of Eq(1. Then the following statements are true: (i) No solution of Eq(1. N ∈ {1.6. . p − 1 . . 2.6 Assume that the number φ in Theorem 1. then every solution of Eq(1. When the number φ in Theorem 1.23) with w0 ∈ F is dense in R.

then the values (nφ − θ)(mod π) for n = 0. wk = wl if and only if Chapter 1. φ is a rational multiple of π.7 Semicycle Analysis We strongly believe that a semicycle analysis of the solutions of a scalar difference equation is a powerful tool for a detailed understanding of the entire character of solutions and often leads to straightforward proofs of their long term behavior. . with l ≥ −1 and m ≤ ∞ and such that either l = −1. . (1.24 Proof.33) relative to an equilibrium point x. we present some results about the semicycle character of solutions of some general difference equations of the form xn+1 = f (xn .33) under appropriate hypotheses on the function f . . . 2 2 2 1. Preliminary Results tan(kφ − θ) = tan(lφ − θ) if and only if for some integer N kφ − θ − (lφ − θ) = N π if and only if (k − l)φ = N π that is. First we give the definitions for the positive and negative semicycle of a solution of Eq(1. n = 0. xn−1 ). xm }. π ). . 1. . In this section.23) and the fact that when φ is not a rational multiple of π.1) consists of a “string” of terms {xl . The proof of statement (ii) is a consequence of the form of the solution of Eq(1. all greater than or equal to the equilibrium x. are dense in the interval (− π . . or m < ∞ and xm+1 < x. . A positive semicycle of a solution {xn } of Eq(1. xl+1 . Statement (i) is true because. . 1. . or l > −1 and xl−1 < x and either m = ∞. .

7. Proof.7. We will assume that xN −1 < x ≤ xN .33). Then except possibly for the first semicycle. n2 ≥ n0 such that xn1 xn2 < 0. with l ≥ −1 and m ≤ ∞ and such that either l = −1. Let x be a positive equilibrium of Eq(1. Definition 1. The first result was established in [28] while we were working on a special case of the equation we are investigating in this monograph. . A sequence {xn } is called strictly oscillatory if for every n0 ≥ 0. Otherwise the sequence is called nonoscillatory. ∞) × (0.33) with at least two semicycles. (See Section 6. x) = x and xN +2 = f (xN +1 . Let {xn } be a solution of Eq(1.33) has semicycles of length one.33) consists of a “string” of terms {xl . The sequence {xn } is called strictly oscillatory about x if the sequence xn − x is strictly oscillatory. . . y) we have xN +1 = f (xN . or l > −1 and xl−1 ≥ x and either m = ∞. . there exist n1 . xN −1 ) < f (x. (b) A sequence {xn } is said to oscillate about x if the sequence xn − x oscillates.5. Then there exists N ≥ 0 such that either xN −1 < x ≤ xN or xN −1 ≥ x > xN .1 (Oscillation) (a) A sequence {xn } is said to oscillate about zero or simply to oscillate if the terms xn are neither eventually all positive nor eventually all negative.1.7. every solution of Eq(1. . ∞). x) = x. y) is increasing in y for each fixed x. (0. xN ) > f (x. xl+1 . Then by using the monotonic character of f (x. ∞)] is such that: f (x.1 ([28]) Assume that f ∈ C[(0. All other cases are similar and will be omitted. y) is decreasing in x for each fixed y. and f (x. Semicycle Analysis 25 A negative semicycle of a solution {xn } of Eq(1. all less than the equilibrium x. xm }.) Theorem 1. Thus xN +1 < x < xN +2 or m < ∞ and xm+1 ≥ x.

7. Theorem 1. Then by using the increasing character of f we obtain: xN +1 = f (xN . xN −1 ) > f (x. y) is increasing in both arguments. Theorem 1. Then by using the decreasing character of f we obtain: xN +1 = f (xN . ∞). (0. 2 . which is a contradiction. ∞) × (0. The proof in the case of negative semicycle is similar and is omitted. Let x be a positive equilibrium of Eq(1. ∞). 2 The next result applies when the function f (x. Then every oscillatory solution of Eq(1. y) is decreasing in both arguments.33) has semicycles of length one. The proof in the case of negative semicycle is similar and is omitted. x) = x which shows that the next term xN +1 also belongs to the positive semicycle. with at least one of the inequalities being strict.7. Let x be a positive equilibrium of Eq(1. Proof. Assume that {xn } is an oscillatory solution with two consecutive terms xN −1 and xN in a positive semicycle xN −1 ≥ x and xN ≥ x with at least one of the inequalities being strict. every oscillatory solution of Eq(1. Then except possibly for the first semicycle. Preliminary Results 2 The next result applies when the function f is decreasing in both arguments. The next result applies when the function f is increasing in both arguments. y) is increasing in x and decreasing in y. Proof. x) = x which completes the proof.26 and the proof follows by induction.3 Assume that f ∈ C[(0.33). It follows by induction that all future terms of this solution belong to this positive semicycle.33) has semicycles of length at most two.33). (0. Assume that {xn } is an oscillatory solution with two consecutive terms xN −1 and xN in a positive semicycle xN −1 ≥ x and xN ≥ x. xN −1 ) < f (x. ∞)] and that f (x. ∞)] and that f (x. ∞) × (0.2 Assume that f ∈ C[(0. Chapter 1.

7. x) = x which shows that the next term xN +1 also belongs to the positive semicycle. xN ) = x which shows that the positive semicycle has length two. Then.7. (0. if we assume that f (u. y) is decreasing in y for each fixed x. y) is increasing in x for each fixed y. xN ) = x (1.4 Assume that f ∈ C[(0.35) then {xn } oscillates about the equilibrium x with semicycles of length two or three. The extreme in each semicycle occurs in the first term if the semicycle has two terms. and f (x. ∞). is similar and is omitted. If xN > xN −1 > x then by using the increasing character of f and Condition (1. The proof in the case xN −1 ≥ x > xN .34) we obtain: xN +1 = f (xN . Now also assume that {xn } is an oscillatory solution with two consecutive terms xN −1 and xN such that xN −1 > xN ≥ x and xN −2 < x. Furthermore.1. ∞)] is such that: f (x. Proof. u) = x for every u and f (x.33) has semicycles of length at least two. Assume that {xn } is an oscillatory solution with two consecutive terms xN −1 and xN such that xN −1 < x ≤ xN . Then by using the increasing character of f and Condition (1. every oscillatory solution of Eq(1.33). except possibly for the first semicycle. except possibly for the first semicycle which may have length one. Then by using the increasing character of f we obtain xN +1 = f (xN . ∞) × (0. xN −1 ) > f (xN .34) . xN −1 ) > f (x.34) we obtain xN +1 = f (xN . xN −1 ) < f (xN . and in the second term if the semicycle has three terms. Semicycle Analysis 27 Theorem 1. y) < x for every x > y > x (1. Let x be a positive equilibrium of Eq(1.

34). The proof of the remaining case is similar and will be omitted. we obtain x1 = f (x0 . x−1 ) = x. and the proof follows by induction.36) f (x. instead of Condition (1. y) is decreasing in y for each fixed x. f satisfies Condition (1.36) is satisfied and that x0 > x−1 > x.36). If. ∞). Preliminary Results xN +1 = f (xN . Assume that f satisfies Conditions (1.5 Assume that f ∈ C[(0. Proof. Theorem 1.35). x−1 ) > f (x−1 . Let x be a positive equilibrium of Eq(1. ∞) × (0.33) has monotonic solutions. then Eq(1. If. Then if x0 > x−1 > x. xN −1 ) < xN and the proof is complete.36) we get x1 = f (x0 . In view of Condition (1. Then by the monotonic character of f and (1. the corresponding solution is increasing. y) is increasing in x for each fixed y.35) we find Chapter 1. (0.34) is not enough to guarantee the above result. we assume (1. and f (x.37) . ∞)] is such that: f (x. y) < x for every x < y < x then if x0 < x−1 < x. y) > x for every x > y > x. and f (x. on the other hand.7.34) and (1. x−1 ) > x0 . 2 Condition (1. First.33). the corresponding solution is decreasing. assume that Condition (1. 2 (1.34).28 and by using Condition (1.

A + Bxn + Cxn−1 Here we investigate the equilibrium points of the general equation xn+1 = where α. B + C ∈ (0. 1. . γ. Zero is an equilibrium point of Eq(2. Periodicity.1) are the nonnegative solutions of the equation x= or equivalently (B + C)x2 − (β + γ − A)x − α = 0.Chapter 2 Local Stability.1) When (2. B. β.1) has a positive equilibrium if and only if β + γ > A.2) and where the initial conditions x−1 and x0 are arbitrary nonnegative real numbers such that the right hand side of Eq(2.1) if and only if α = 0 and A > 0. 29 . (2. and Invariant Intervals 2.3) n = 0. Semicycles.1). ∞) with α + β + γ. C ∈ [0. the equilibrium points of Eq(2. Eq(2. In view of the above restriction on the initial conditions of Eq(2.4) α + (β + γ)x A + (B + C)x (2.5) holds. .1 Equilibrium Points α + βxn + γxn−1 . A. ∞) (2. . (2. in addition to the zero equilibrium .5) (2.1) is well defined for all n ≥ 0.

B+C (2. Semicycles.9) . then x is unique. and it is given by (2. . A + Bxn there is very little known.8). v) = (A + Bu + Cv)2 (2. [32].1). it satisfies Eqs(2.1) when the initial conditions and the coefficients of the equation are real numbers is of great mathematical importance and. . 2. 1. and Invariant Intervals In fact in this case the positive equilibrium of Eq(2.2).8) of the quadratic equation (2.4).1) is positive and is given by x= β+γ . v) = A + Bu + Cv and observe that (βA − αB) + (βC − γB)v fu (u.) In summary. . B+C (2.1) is unique and is given by x= When α = 0 and A = 0 the only equilibrium point of Eq(2.1) is the positive solution x= β+γ−A+ (β + γ − A)2 + 4α(B + C) 2(B + C) (2. and [69] for some results in this regard. The problem of investigating Eq(2.7) β+γ−A .4).30 Chapter 2.4) would also be an equilibrium worth investigating. Periodicity. Set α + βu + γv f (u.3) and (2. . [18].1).6) Note that in view of Condition (2.2 Stability of the Zero Equilibrium Here we investigate the stability of the zero equilibrium of Eq(2. If we had allowed negative initial conditions then the negative solution of Eq(2. the quantity β + γ is positive. Local Stability. n = 0. Finally when α>0 the only equilibrium point of Eq(2. when α = 0.1) has a positive equilibrium x. it is interesting to observe that when Eq(2. except for our presentation in Section 1.6 about the Riccati equation α + βxn xn+1 = . (See [13]. This observation simplifies the investigation of the local stability of the positive equilibrium of Eq(2.

As a consequence of these two theorems we obtain the following global result: Theorem 2.1) we have p= β A and q = γ A and so the linearized equation associated with Eq(2.1 Assume that A > 0. 1. . .2.1.1) about the zero equilibrium point is β γ zn+1 − zn − zn−1 = 0.1). (A + Bu + Cv)2 31 (2. 1. is globally asymptotically stable when β+γ ≤A and is unstable when β + γ > A. For the zero equilibrium of Eq(2. Stability of the Zero Equilibrium and fv (u.2. then the linearized equation associated with Eq(2. . v) = (γA − αC) − (βC − γB)u . .1. in addition to the local stability Theorem 1.10) If x denotes an equilibrium point of Eq(2. .3. x). A A For the stability of the zero equilibrium of Eq(2. A + Bxn + Cxn−1 n = 0. Furthermore the zero equilibrium point is a sink when a saddle point when a repeller when A>β+γ A < β + γ < A + 2β β + γ > A + 2β.1.2. x) and q = fv (x. .1. . Then the zero equilibrium point of the equation xn+1 = βxn + γxn−1 .1) about the equilibrium point x is zn+1 − pzn − qzn−1 = 0 where p = fu (x.1. we have the luxury of the global stability Theorem 1.1). The local stability character of x is now described by the linearized stability Theorem 1. . n = 0.

Periodicity.12) (A + β + γ)x > −α − (2. x) = and q = fv (x.1.32 Chapter 2. and it is given by (2. x) = (γA − αC) − (βC − γB)x (γA − αC) − (βC − γB)x = 2 . 2 (A + (B + C)x) A + α(B + C) + (B + C)(A + β + γ)x (βA − αB) + (βC − γB)x (βA − αB) + (βC − γB)x = 2 2 (A + (B + C)x) A + α(B + C) + (B + C)(A + β + γ)x Hence the conditions for the local asymptotic stability of x are (see Theorem 1.13)-(2. Inequalities (2. are equivalent to some inequalities of the form x>ρ .1.4).15) A2 + (γA − αC) βC − γB x > −α − . A+β+γ− B+C B+C Out of these inequalities we will obtain explicit stability conditions by using the following procedure: First we observe that Inequalities (2. Semicycles.14) (2. In these cases the positive equilibrium x is unique. Local Stability.3 Local Stability of the Positive Equilibrium α = 0 and β + γ > A As we mentioned in Section 2. and Invariant Intervals 2.3) and (2.15). it satisfies Eqs(2.8).13) (2. Eq(2.12) are equivalent to the following three inequalities: A+β+γ+2 βC − γB A2 + (βA − αB) − (γA − αC) x > −α − B+C B+C A2 − (βA − αB) − (γA − αC) B+C (2.1) has a positive equilibrium when either or α > 0.11) and q > −1.11) and (2. By using the identity (B + C)x2 = (β + γ − A)x + α. we see that p = fu (x.1) the following: |p| < 1 − q (2.

Then γ(α + β) = 0. 4. Then A = B = γ = 0. The following result characterizes all possible special cases of equations of the form of Eq(2. . When is Every Solution Periodic with the Same Period? and/or x<σ for some ρ.18) (2. 1. . .1 Let p ≥ 2 be a positive integer and assume that every positive solution of Eq(2.1) is periodic with period p.16) (2. n = 0. . 33 2.4 When is Every Solution Periodic with the Same Period? 1 . (ii) Assume C = 0. 1. Now if we set F (u) = (B + C)u2 − (β + γ − A)u − α. . Then the following statements are true: (i) Assume C > 0. n = 0. = xn−1 are remarkable in the sense that all positive solutions of each of these four nontrivial equations are periodic with periods 2.17) (2. 5. . and 6 respectively. 1.4.4. 1. . . n = 0. xn−1 xn .) Theorem 2. ∞). The following four special examples of Eq(2. and [74]. . . σ ∈ [0. . .2. xn 1 xn−1 . (See also [75].1) with the property that every solution of the equation is periodic with the same period. [73]. (2.19) 1 + xn .1) xn+1 = xn+1 = xn+1 = xn+1 n = 0. it is clear that F (x) = 0 and that x > ρ if and only if F (ρ) < 0 while x<σ if and only if F (σ) > 0. .

Local Stability. Assume that every positive solution of Eq(2.19) . and Invariant Intervals Proof. . Consider the solution with x−1 = 1 and x0 ∈ (0. Semicycles. Then up to a change of variables of the form xn = λyn Eq(2.16)-(2. 6}.1) is periodic with period p.21) holds.20) is impossible. Then we claim that A = B = 0.4. Then clearly x0 = xp and so from Eq(2. 2 Corollary 2. Periodicity. The proof is complete. 5. 4. A + B + Cxp−2 and xp−1 = x−1 = 1 we see that (2. In addition to (2.1) xp = Hence (A + B)x0 + (Cx0 − γ)xp−2 = α + β. Hence (2. we see that (2. (ii) Assume C = 0 and. assume that γ(α + β) > 0.20) is impossible. (i) Assume that C > 0. ∞).21) we now claim that also γ = 0. Then by choosing x0 sufficiently small. 3.21) (2.1) reduces to one of the equations (2. for the sake of contradiction. A+B C we see that (2.1 Let p ∈ {2. (2.20) is impossible.34 Chapter 2.20) α + β + γxp−2 = x0 . A+B C (2. Thus (2.22) holds .22) Otherwise γ > 0 and by choosing x0 < min α+β γ . Otherwise A + B > 0 and by choosing x0 > max α+β γ .

φ. . Assume that .1). φ. . ψ ∈ [0. solution with period p and furthermore the set of solutions of the equation with prime period p is nonempty. is a prime period-two solution of Eq(2. Convention Throughout this book when we say that “every solution of a certain difference equation converges to a periodic solution with period p” we mean that every solution converges to a. . ψ. . .5 Existence of Prime Period-Two Solutions In this section we give necessary and sufficient conditions for Eq(2. Furthermore when (2. . ψ. .1) to have a prime period-two solution and we exhibit all prime period-two solutions of the equation.23) α + βψ + γφ A + Bψ + Cφ α + βφ + γψ A + Bφ + Cψ .26) and (2.26) αC + β(γ − β − A) . ψ.1) if and only if α + β(φ + ψ) = Bφψ with φ. ∞) and φ = ψ (2. ψ. φ.24) (2. . not necessarily prime.2.) Furthermore we are interested in conditions under which every solution of Eq(2.5.27) hold . Existence of Prime Period-Two Solutions 35 2. .1) converges to a period-two solution in a nontrivial way according to the following convention. (See [50]. .27) (2. φ. .25) (2. is a prime period-two solution of Eq(2. C (2. . Then φ= and ψ= from which we find that C(φ + ψ) = γ − β − A and when C>0 we also find that (B − C)φψ = One can show that when C = 0 and B > 0 Eq(2.1) has a prime period-two solution if and only if γ = β + A.

and Invariant Intervals or equivalently β φ> . α = β = 0.. . . φ. . . α + β > 0. B > C. α = β = 0. . 0. . φ. ψ. When (2.23) and (2. . − φ.. C C Finally when (2.30) (2. (2. ψ.24) holds and that Eq(2. . . γ−A γ−A − φ. . . .25) that one of the following three conditions must be satisfied: B = C. and φψ = 0. φ. φ. . ψ.1) possesses a prime period-two solution . (γ − β − A)[B(γ − β − A) − C(γ + 3β − A)] α< . ψ. . Local Stability.29) (2. .29) holds. .. . (2.30) holds. . Periodicity. and φψ > 0. B and ψ = α + βφ .31) holds. . 1. . the two cycle is . . .6 Let Local Asymptotic Stability of a Two Cycle . . the values φ and ψ of the two cycle are given by the two roots of the quadratic equation t2 − γ−β−A αC + β(γ − β − A) t+ =0 C C(B − C) and we should also require that the discriminant of this quadratic equation be positive. . and φψ ≥ 0. Set un = xn−1 and vn = xn for n = 0.32) 4C 2 2. B φ= β+ √ β 2 + αB .31) γ > β + A and When (2. φ. That is. . be a two cycle of the difference equation (2. .1). C C with 0 ≤ φ < γ−A . 0. B = C. the two cycle is . φ. −β + Bφ (2.28) Next assume that (2.36 Chapter 2. Then it follows from (2. Semicycles. . C γ−A γ−A . .

By computing the partial derivatives of g and h and by using the fact that φ= α + βψ + γφ A + Bψ + Cφ and ψ = α + βφ + γψ A + Bφ + Cψ The two cycle is locally asymptotically stable if the eigenvalues of the Jacobian matrix φ lie inside the unit disk. ψ) = . Local Asymptotic Stability of a Two Cycle and write Eq(2. ψ) ∂u φ  = JT 2  .2. ∞) × [0. ψ) = . ψ) = . evaluated at ψ By definition  ∂g  ∂g (φ. n = 0. ψ) ∂v (φ. 1. ψ ∂h ∂h (φ. v) = α + βv + γu A + Bv + Cu and h(u. JT 2 . .1) in the equivalent form: un+1 = vn α+βv vn+1 = A+Bvn +γun . n +Cun Let T be the function on [0. ∂u (A + Bψ + Cφ)2 ∂g (βA − αB) − (γB − βC)φ (φ. the second iterate of T . v) h(u. ∂v (A + Bψ + Cφ)2 [(βA − αB) − (γB − βC)ψ][(γA − αC) + (γB − βC)ψ] ∂h (φ. ∂u (A + Bφ + Cψ)2 (A + Bψ + Cφ)2 . . ∞) defined by: T Then u v = v βv+γu+α Bv+Cu+A 37 .6. ψ) ∂v (φ. ψ) ∂u we find the following identities: (γA − αC) + (γB − βC)ψ ∂g (φ. v) = u v = g(u. . v) α+βv+γu α + β A+Bv+Cu + γv α+βv+γu A + B A+Bv+Cu + Cv . Furthermore T2 where g(u. . φ ψ is a fixed point of T 2 .

β. be a two cycle of Eq(2. A. .1) converges to the two cycle. . (2. 6.6. ψ) + (φ.7. if and only if |S| < 1 + D and D < 1. and Invariant Intervals [(βA − αB) − (γB − βC)ψ][(βA − αB) − (γB − βC)φ] ∂h (φ. That is. .9. ψ) ∂u ∂v ∂h ∂g ∂h ∂g (φ.1. γ.33) . . ψ).4 for some applications. Periodicity. φ. and 10. every solution converges to a (not necessarily prime) period-two solution and n = 0.7. . .38 Chapter 2. ψ) = ∂v (A + Bφ + Cψ)2 (A + Bψ + Cφ)2 (γA − αC) + (γB − βC)φ . 1. A and B so that every solution of Eq(2. Semicycles. ψ.5. ψ) (φ.1). A + B ∈ (0. ψ. γ. ∞) with α + β + γ. .6. 6. A + Bxn In this section we consider the equation xn+1 = where α. φ. The basin of attraction of this two cycle is the set B of all initial conditions (x−1 . . See Sections 6.1 (Basin of Attraction of a Two Cycle) Let . ∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1.7 Convergence to Period-Two Solutions When C=0 α + βxn + γxn−1 .1 (c) that both eigenvalues of JT 2 unit disk |λ| < 1.33) converges to a period-two solution. β. x0 ) through which the solution of Eq(2. and 7. . B ∈ [0. + (A + Bφ + Cψ)2 Set S= and D= ∂h ∂g (φ. ∞) and nonnegative initial conditions and obtain a signum invariant for its solutions which will be used in the sequel (See Sections 4. Local Stability. Definition 2. 6. 2. ψ) (φ.2) to obtain conditions on α. ψ) − (φ.

Set (2. Now observe that xn+1 − xn−1 = β+A (xn − xn−2 ) for n ≥ 1 A + Bxn .33): (i) xn+1 − xn−1 = 0 for all n ≥ 0.2 where we present the complete character of solutions of Eq(2. if and only if J0 > 0 and β + A > 0. ∞ Lemma 2. (iii) xn+1 − xn−1 > 0 for all n ≥ 0. the sign of the quantity Jn is constant along each solution. Statement (ii) holds if and only if J0 < 0 and β +A > 0. and let {xn }n=−1 be a solution of Eq(2. A + Bxn In particular. In particular.33) converges to a periodtwo solution if and only if (2.27) holds.33). Convergence to Period-Two Solutions When C=0 39 there exist prime period-two solutions. the solution {xn } is periodic with period-two. we need the following result. Clearly (i) holds. one and only one of the following three statements is true for each solution of Eq(2. Finally (iii) holds.2.33). Jn+1 = (b) xn+1 − xn−1 = Jn A + Bxn for n ≥ 0.27) holds.7. Then the following statements are true: (a) β+A Jn for n ≥ 0.1 Assume that (2. (β + A)J0 = 0. In this case the subsequences ∞ {x2n }∞ and {x2n−1 }∞ of the solution {xn }n=0 are both increasing and so if we can n=0 n=0 show that they are bounded.) Now before we see that when B > 0 every solution of Eq(2. (See also Section 10. (ii) xn+1 − xn−1 < 0 for all n ≥ 0.7. the proof of which follows by straightforward computations. In this case. the solution would converge to a period-two solution. if and only if.34) Jn = α + β(xn−1 + xn ) − Bxn−1 xn for n ≥ 0. In this case the subsequences {x2n }∞ and {x2n−1 }∞ of the solution {xn }∞ are both decreasing to finite limits and n=0 n=0 n=0 so in this case the solution converges to a period-two solution.

. . A+β A+β A+β A+β < = ρ. . Kρ + x2N 1−ρ . .40 and so Chapter 2. The proof for the subsequence of odd terms is similar and will be omitted. (b) When (2. ψ. Then the following statements hold: (a) Eq(2. and Invariant Intervals n xn+1 − xn−1 = (x1 − x−1 ) β+A . Theorem 2. . .1 Assume B > 0. . .27) holds. Semicycles. . φ. + |x2N +2 − x2N | + x2N m for m = 1. A + Bx2n+1 A + Bx2n A + Bx2N +1 A + Bx2N Set Then K = |x2N − x2N −2 |. for the sake of contradiction.33) has a prime period-two solution if and only if (2. 2.27) holds every solution of Eq(2. ≤K ρi + x2N < i=1 which contradicts the hypothesis that the subsequence of even terms is unbounded. (c) When (2.27) holds all prime period-two solutions . φ.33) are given by (2.35) We will now employ (2. . Periodicity. ψ.7. A + Bx2N +1 A + Bx2N A+β A+β |x2N − x2N −2 | < Kρ A + Bx2N +1 A + Bx2N A+β A+β |x2N +2 − x2N | < Kρ2 |x2N +4 − x2N +2 | = A + Bx2N +3 A + Bx2N +2 and by induction |x2N +2m − x2N +2(m−1) | < Kρm But then x2N +2m ≤ |x2N +2m − x2N +2m−2 | + . A + Bxk k=1 (2. |x2N +2 − x2N | = A+β A+β < 1.33) converges to a period-two solution. Then there exists N ≥ 0 such that ρ= Clearly for all n > N . To this end assume. of Eq(2.35) to show that the subsequences of even and odd terms of the solution are bounded.28). that the subsequence of even terms increases to ∞. In summary we have established the following result. . Local Stability. . We will give the details for the subsequence of even terms.

For Eq(2. .36) if xN −1 . . y) = N − Mx . fy < 0 if x > M fx > 0 and fy < 0 fx > 0 and fy < 0 fx > 0 and fy > 0 fx > 0 and fy = 0 L L fx > 0 if y < − M . fx < 0 if y > − M fy > 0 fx > 0 and fy < 0 fx > 0 and fy < 0 fx < 0 and fy > 0 fx < 0 and fy > 0 L L fx > 0 if y > − M .36) are determined from the intervals where the function f (x. 1.10) and are fx (x. Case L 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 + + + + + + 0 0 0 0 M + + + 0 0 − + + − − 0 0 − − N + 0 − + 0 + 0 − + 0 − 0 − + 0 − M = βC − γB. then xn ∈ I for every n > N . (2.1). Signs of derivatives fx and fy fx > 0 N N fy > 0 if x < M . xn−1 ). Invariant intervals for Eq(2. fy < 0 if x < M L + My (A + Bx + Cy)2 and fy (x. In other words I is an invariant interval for the difference equation xn+1 = f (xn . n = 0. we obtain the following table of invariant intervals for Eq(2. fx < 0 if y < − M fy < 0 fx < 0 and fy = 0 fx < 0 and fy < 0 fx < 0 and fy > 0 fx < 0 and fy > 0 fx < 0 N N fy > 0 if x > M . Invariant Intervals 41 2. . xN ∈ I for some N ≥ 0. y) is monotonic in its arguments. .1) is an interval I with the property that if two consecutive terms of the solution fall in I then all the subsequent terms of the solution also belong to I.8 Invariant Intervals An invariant interval for Eq(2.1) the partial derivatives are given by (2.9) and (2.2. y) = where L = βA − αB.8. and N = γA − αC. (A + Bx + Cy)2 − + − − − − − − Using this table. The following table gives the signs of fx and fy in all possible nondegenerate cases.

M ] if AM +(B+C)N < M √ 2 αM +(β+γ)N N (β−A)M −CN + ((β−A)M −N C) +4BM (αM +γN ) [M .42 Chapter 2. C(γ−β) ] if B = 0 and β < γ βM K+αM L if C = 0 and BM K+AM −γL > − M −CL and −(βM + BL)K ≥ αM + AL β [0. B(β−γ) ] if C = 0 and β > γ αM −L(β+γ) αM −L(β+γ) L L [ AM −L(B+C) . ] if AM −BL+CMK > − M 2BM K α [ α(A+β) . B ] if B = 0 2 γ [ C . αM −(β+γ)L L AM −(B+C)L < − M √ (A−γ)M −BL+ ((A−γ)M −BL)2 +4CM (αM −βL) ] −2CM βL−αM (γ−A)M +BL ] if C = 0. ] if 2CM αM +βK+γN M AM +BK+CN M > N M . B(β−γ) ] if A = 0 and β > γ αM +(β+γ)N N αM +(β+γ)N N [ AM +(B+C)N . A−β ] if B = 0. ] if AM +(B+C)N > 2BM αM +γN N [ M . Periodicity. ] if B > 0 2B α [0. Local Stability. C ] if C > 0 β β2 [ B . C ] if C > 0 [k. A−β ] if B = 0. − M ] if L [− M . M ] if AM +(B+C)N < M √ 2 N (γ−A)M −BN + ((γ−A)M −BN ) +4BM (αM +βN ) [M . (A−β)M +CN ] if B = 0 and (A − β)M + CN > 0 √ β−A+ (β−A)2 +4Bα [0. L [− M . B ] if B > 0 γ γ2 [ C . C ] if C > 0 [k. B + α ] if A = 0 and β > 0 β α [0. K] if C = 0 where k and K satisfy α + βk + (γ − A)K ≤ BkK ≤ α + βK − Ak γ [0. and Invariant Intervals Case 1 2 3 4 5 Invariant intervals αM +(β+γ)N N N [0. (γ − A)M + BL > 0 N M and B > 0 6 7 8 9 10 11 12 13 14 15 16 γ β [ C . C ] if C > 0 β αB+β 2 [ B . − M ] if AM −L(B+C) < − M √ 2 αM −βL+γM L (β−A)M +LC+ ((β−A)M +LC) +4BM (αM −γL) L [− M . αC+γ ] if B = 0 and γ > βC γ−βC γ [0. K] if A = 0 where k and K satisfy α+(β+γ)k ≤ kK ≤ α+(β+γ)K B+C B+C γ [0. ] if B > 0 2B α [0. α > 0 and A > β √ β+γ−A+ (β+γ−A)2 +4α(B+C) [0. ] if B + C > 0 √ 2(B+C) 2 +4αB α β−A+ (β−A) [A. A ] if A > 0 A2 +Bα β β [ B . Semicycles. α > 0 and A > β √ β−A+ (β−A)2 +4Bα [0. ] if B > 0 2B α α [ A . K] if C = 0 where k and K satisfy α + βk + (γ − A)K ≤ BkK ≤ α + βK − Ak β γ [ B . A−β ] if B = 0 and A > β L [0. A ] if A > 0 [k.

9. 6}? Open Problem 2. 4. Is it true that the linearized equation about a positive equilibrium point of Eq(2. Open Problem 2.1) with α.37) converges to a periodic solution with period p.9. .4 It is known (see Sections 2. ∞) × [0. B. = 1 + xn . 4. A.7) that every positive solution of each of the following three equations xn+1 = 1 + xn+1 = xn+1 xn−1 . Is it true that p ∈ {2. . (0. n = 0. if and only if every root of the characteristic equation is a pth root of unity.37) has the property that every one of its solutions is also periodic with the same period p? What is it that makes every positive solution of a difference equation converge to a periodic solution with period p ≥ 2 ? Open Problem 2. . . .9. xn−1 ). γ. 1. (2. . every solution is periodic with period p ≥ 2. 1. ∞)] is such that every positive solution of the equation xn+1 = f (xn . .2 Let f ∈ C 1 [[0.9. .9.7. n = 0. 6.9 Open Problems and Conjectures What is it that makes all solutions of a nonlinear difference equation periodic with the same period? For a linear equation.5.2. and 6. . 5.37) is periodic with period p ≥ 2.7. . 1. ∞) × (0. β. 1 + xn xn + 2xn−1 . 1. Find necessary and sufficient conditions so that every nonnegative solution of the difference equation (2.40) 1 + xn−1 .3 Let p ≥ 2 be an integer and assume that every positive solution of Eq(2. (2. . xn n = 0. . ∞) converges to a solution with period p . Open Problem 2.38) (2. ∞).39) (2. ∞). C ∈ [0. ∞)] and let p ≥ 2 be an integer. n = 0. [0. In particular address the case p = 2.1 Assume that f ∈ C 1 [(0. Open Problems and Conjectures 43 2.

9. ∞) × (0. A. A.38) or Eq(2.7. C ∈ (0. (2.3 Assume α. 6. if (2. γ. φ. B ∈ (0. . x0 ) ∈ (0.) Conjecture 2. Conjecture 2. Periodicity. . γ. Conversely. 6.41). . ψ. (See Sections 2. ∞) for which the solution {xn }∞ n=−1 converges to the period-two solution (2. . β.41) is a period-two solution of Eq(2. determine φ and ψ in terms of the initial conditions x−1 and x0 .1) has no period-two solution. . Semicycles. and Invariant Intervals converges to a solution with (not necessarily prime) period-two: . C ∈ (0. and 10. Determine the set of initial conditions x−1 and x0 for which the solution {xn }∞ n=−1 of Eq(2.9. φ.7. C ∈ (0.2 Assume α.41) In each case. ∞) and that Eq(2. γ.39) or Eq(2.5 Assume α.7. Show that the positive equilibrium is globally asymptotically stable. A.9. B. Open Problem 2. 4.44 Chapter 2. ∞). β. A. Local Stability. γ. .40). β.1) is a saddle point. Show that every positive solution of Eq(2. ∞) and γ > β + A. ψ. B. Conjecture 2.1) is bounded. Show that the positive equilibrium of Eq(2.33) is bounded. B.5. ∞) and that Eq(2.1 Assume α. . determine all initial conditions (x−1 . β.2.9. .1) has a positive prime period-two solution.

9.9 Extend Theorem 2. . xn n = 0.8 Assume f ∈ C[(0. C ∈ [0. ∞)]. (See Section 4. Obtain necessary and sufficient conditions on f for every solution of the difference equation xn+1 = to be bounded.1 to third order difference equations xn+1 = α + βxn + γxn−1 + δxn−2 . Open Problem 2. Open Problems and Conjectures Open Problem 2.7. 1.9. 45 Obtain necessary and sufficient conditions in terms of the coefficients so that every positive solution of Eq(2. C ∈ [0. . ∞).1) cannot have a positive prime two cycle which attracts all positive non-equilibrium solutions.1) is bounded. What additional conditions should the parameters satisfy? . f (xn−1 ) . Open Problem 2. β. . A. . B. ∞). 1.9. f (xn ) . (0. Show that Eq(2.7 Assume f ∈ C[(0. xn−1 n = 0.4 Assume that α. (0. .6 Assume α. .7). .2). B. with nonnegative parameters and nonnegative initial conditions such that γ = β + δ + A. γ.2. β. Obtain necessary and sufficient conditions on f for every positive solution of the equation xn+1 = 1 + to converge to a period-two solution. Conjecture 2. . ∞). 1.9. Open Problem 2. ∞)]. A. ∞). γ. .9. A + Bxn−1 + Dxn−2 n = 0.9. (See Section 5.

9. or nonhyperbolic? Extend these results to fourth order difference equation xn+1 = f (xn .42) to be a repeller. . xn−2 ).1 to third order difference equations xn+1 = f (xn .46 Chapter 2.1. n = 0. . . 1. Semicycles. . xn−3 ). . and Invariant Intervals Open Problem 2. . xn−1 . (2. xn−1 . But how about necessary and sufficient conditions for the equilibrium of Eq(2.42) For the equation λ3 + aλ2 + bλ + c = 0 one can see that all roots lie inside the unit disk |λ| < 1 if and only if |a + c| < 1 + b |a − 3c| < 3 − b b + c2 < 1 + ac      . . n = 0. Local Stability. Periodicity. or a saddle point. xn−2 . 1.10 Extend the linearized stability Theorem 1. . .

(3. . . .2) xn+1 = α . . . A Eq(1) contains the following nine equations of the (1. (3. . 1. (3. 1. . .1 Introduction α .Chapter 3 (1.1) xn+1 = α . 1. 1)-Type Equations 3.4) xn+1 = β . (3. A n = 0. . . Cxn−1 47 n = 0. . Bxn n = 0.5) xn+1 = βxn . 1. Cxn−1 n = 0. B n = 0. .6) . . 1)-type: xn+1 = n = 0. 1. (3.3) xn+1 = βxn . 1. . . . (3. . .

Every solution of Eq(3. C n = 0. Bxn n = 0. Eqs(3. 3. (1.6) which by the change of variables xn = reduces to the equation yn+1 = yn .. .8). . . y−1 y−1 y0 y0 It is interesting to note that except for the equilibrium solution y = 1. .7) are linear difference equation. (3.6) and (3. y0 1 1 y−1 .9) are trivial. Eqs(3. . A n = 0. every other solution of Eq(3. the initial conditions are nonnegative. .. (3.5). . (3. (3. and (3. and the denominators are always positive. . 1. Eqs (3. 1. yn−1 n = 0. (3. . 1)-type Eq(3. Of these nine equations. 1.8) xn+1 = γ .10) has prime period six. are treated in the next two sections. y0 .10) is periodic with period six.4) and (3.10) It is easy to see that every positive solution of Eq(3. Indeed the solution with initial conditions y−1 and y0 is the six-cycle: y−1 .. .7) xn+1 = and γxn−1 . . .48 Chapter 3.9) Please recall our classification convention in which all parameters that appear in these equations are positive. . . .4) is periodic with period four. The remaining two equations. 1)-Type Equations xn+1 = γxn−1 .2) is periodic with period two and every solution of Eq(3.2 The Case α = γ = A = B = 0 : xn+1 = β yn C βxn Cxn−1 This is the (1. . 1. .1). namely. . .

. . ∞)]. . .4 Open Problems and Conjectures Of the nine equations in this chapter we would like to single out the two nonlinear equations. ∞)]. n = 0.11) is yn = C1 −1 + 2 √ 5 n n = 0. where f ∈ C[(0.15) What is it that makes every solution of a nonlinear difference equation periodic with the same period? We believe this is a question of paramount importance and so we pose the following open problems. Eq(3.3. xn−1 ). (0. respectively. ∞). The general solution of Eq(3.8) which by the change of variables xn = reduces to the linear equation yn+1 + yn − yn−1 = 0. (3. . . ∞) × (0. ∞).2) is of the form xn+1 = f (xn ).3 The Case α = β = A = C = 0 : xn+1 = γ yn e B This is the (1.12) .3) also has the property that every nontrivial positive solution is periodic with prime period four but this equation is really a variant of Eq(3. (3.3. . 1)-type Eq(3. 3. (3. . from which the behavior of solutions is easily derived. which possess the property that every positive nontrivial solution of each of these two equations is periodic with the same prime period. 1. namely two and six.2). Eq(3.2) and (3. (3. n = 0. (0. . . 1. where f ∈ C[(0. 1. while Eq(3. The Case α = β = A = C = 0 : xn+1 = γxn−1 Bxn 49 γxn−1 Bxn 3. .14) (3.11) + C2 −1 − 2 √ 5 n .7) is periodic with period two but this is a linear equation.13) n = 0. .6) is of the form xn+1 = f (xn . Also every solution of Eq(3. . (3. 1.6).

. 9.4. [0. xn−1 n = 0. ∞). γ so that every solution of the equation xn+1 = α |xn | + βxn−1 + γ. 1. Show that f (x) = 1 + x.50 Chapter 3. Conjecture 3. Obtain necessary and sufficient conditions on α.1 Let k ≥ 2 be a positive integer and assume that (3. (3. 6.14) is periodic with period k. . 1. Open Problem 3.15) holds.16) is periodic with prime period six. 1)-Type Equations Open Problem 3. [17]. Obtain necessary and sufficient conditions on f so that every positive solution of Eq(3. . 7. Obtain necessary and sufficient conditions on f so that every positive solution of Eq(3. .4. n = 0.4. ∞).12) is periodic with period k. . (1. and [59]. ∞)] be such that every positive nontrivial solution of the difference equation (3.1 Let f ∈ C 1 [[0. Show that f (x) = x. In particular address the cases k = 4.2 Let f ∈ C 1 [[0. . γ ∈ (−∞.4. β.2 Let k ≥ 2 be a positive integer and assume that (3. In particular address the cases: (See [10].) Conjecture 3. β. . ∞)] be such that every positive nontrivial solution of the difference equation xn+1 = f (xn ) . 5. k = 5. with real initial conditions is periodic with period k. ∞). [0. Open Problem 3.4.16) is periodic with prime period five.13) holds. 8. 6.3 Let k be a positive integer and assume that α.

(3. n=0 (iii) {pn }∞ is periodic with prime period k ≥ 2. 1. [0. 1. xn−1 n = 0. . . ∞).5 Let {pn }∞ be a sequence of nonzero real numbers. .4. . . .17) Open Problem 3. f (xn ) .4. ∞)] such that every positive nontrivial solution of the equation xn+1 = is periodic with prime period three. In each of n=0 the following cases investigate the asymptotic behavior of all positive solutions of the difference equation pn x n xn+1 = . . n=0 . n = 0. (3.4 Obtain necessary and sufficient conditions on f ∈ C[[0.18) xn−1 ∞ (i) {pn }n=0 converges to a finite limit. (ii) {pn }∞ converges to ±∞. Open Problems and Conjectures 51 Open Problem 3.4.3.

.

1 Introduction α . (4. A + Cxn−1 n = 0. . 1. Bxn + Cxn−1 n = 0. .4) xn+1 = βxn . .Chapter 4 (1. . . .6) . 2)-Type Equations 4. . 1. . . (4.3) xn+1 = βxn . A + Bxn Eq(1) contains the following nine equations of the (1. 1. .1) xn+1 = α . Bxn + Cxn−1 53 n = 0. 1. (4. A + Bxn n = 0. . 2)-type: xn+1 = n = 0.5) xn+1 = βxn . 1. . 1. (4. . (4. A + Cxn−1 n = 0.2) xn+1 = α . . . . (4. . .

(4.1) and (4. (1. namely Eqs(4.2 The Case β = γ = C = 0 : xn+1 = A yn B α A+Bxn This is the (1. then the subsequences {x2n−1 } and {x2n } satisfy the Riccati equation of the form of Eq(4.6 in the Preliminary Results). A + Cxn−1 n = 0. 1.4) to the linear equation yn+1 = A B yn + . .4). . 1. .8) will not be discussed any further in this chapter. β β n = 0. . . . . 1. Indeed if {xn } is a solution of Eq(4.2) and (4.8) are essentially Riccati equations. . and (4. yn+1 = A + Cyn It is also interesting to note that the change of variables xn = 1 yn reduces the Riccati equation (4. Bxn + Cxn−1 n = 0. .1) which by the change of variables xn = .2). . . Eqs(4. . (4. Also Eqs(4. 1.7) xn+1 = and xn+1 = γxn−1 . 4. 2)-Type Equations xn+1 = γxn−1 .1).54 Chapter 4. 1. .8) γxn−1 . Two of these equations. A + Bxn n = 0. (See Section 1.9) Please recall our classification convention in which all parameters that appear in these equations are positive. . and the denominators are always positive.8). In view of the above. 2)-type Eq(4. (4. the initial conditions are nonnegative. if {xn } is a solution of Eq(4. .2). . then the subsequences {x2n−1 } and {x2n } satisfy the Riccati equation γyn n = 0. Similarly.4) are Riccati-type difference equations. from which the global behavior of solutions is easily derived. . (4.

4.11) (4. Theorem 4.3. .2. 1. (4. .12) was investigated in [65].10) αB .) We will first show that every solution of Eq(4. A2 The Riccati number associated with this equation (see Section 1.10) is globally asymptotically stable.12) Eq(4. n→∞ . The Case β = γ = A = 0 : xn+1 = reduces to the Riccati equation yn+1 = where α Bxn +Cxn−1 55 p .2. . 1 + yn p= n = 0.12) is bounded and then we will use the method of “limiting solutions” to show that every solution of Eq(4.12) converges to √ its equilibrium B + C.3) which by the change of variables √ α xn = yn is transformed to the difference equation yn+1 = C B + . . Proof.3 The Case β = γ = A = 0 : xn+1 = α Bxn +Cxn−1 This is the (1. 1.12) is bounded from above then it is also bounded from below and vice-versa. 4.1 Every solution of Eq(4.1 The positive equilibrium y= −1 + √ 2 1 + 4p of Eq(4. lim sup yn = ∞ and n→∞ lim inf yn = 0. It is easy to see that if a solution of Eq(4. So if the theorem is false there should exist a ∞ solution {yn }n=−1 which is neither bounded from above nor from below. Theorem 4. (See also [12].3. That is. .6) is R = −p < 0 and so the following result is a consequence of Theorem 1.6. . 2)-type Eq(4. yn yn−1 n = 0.12) is bounded and persists. (4. .

12) is globally asymptotically √ B + C is (4. It remains to be shown that y is a global attractor of all solutions of Eq(4.1. yi−1 yi−2 yj Proof. 2)-Type Equations Then clearly we can find indices i and j with 1 ≤ i < j such that yi > yn > yj Hence yj = and yi = That is . y is locally asymptotically stable for all positive values of B and C. Therefore there exist positive numbers m and M such that m ≤ yn ≤ M Let I = lim inf yn n→∞ We will show that I = S.1 . The linearized equation of Eq(4.56 Chapter 4. j − 1}.12). . −2.1. 2 √ B + C of Eq(4. let {yn }∞ n=−1 be a solution of Eq(4. (1. . . Then by Theorem 4. . and S = lim sup yn .3.2 The equilibrium y = stable. S0 = S B C + .5) there exist two solutions {In }∞ n=−3 and {Sn }∞ n=−3 of the difference equation yn+1 = with the following properties: I0 = I. yn yn−1 n = −3. .13) for all n ∈ {−1.3. . n→∞ for n ≥ −1. B C B+C + > yj−1 yj−2 yi C B+C B + ≤ .12) about the equilibrium y = zn+1 = − B C zn − zn−1 . {yn }n=−1 is bounded and persists. Theorem 4.12). 1. . By the theory of limiting solutions (see Section 1. . . To ∞ this end. . and by Theorem 1. B + C < yi yj ≤ B + C which is impossible. . . B+C B+C n = 0.

1 + yn−1 β . . S] for n ≥ −3. From (4. and I= Hence Thus B C B+C B C + =S= = + S−1 S−2 I I I from which it follows that S−1 = S−2 = I.14) and (4. Sn ∈ [I.5) which by the change of variables xn = reduces to Pielou’s difference equation yn+1 = where p= pyn .4 The Case α = γ = B = 0 : xn+1 = Equation A yn . . C βxn A+Cxn−1 .4.16) . (4. .4. I−1 I−2 S B + C = SI. Therefore B C B C + = I = S−1 = + S S S−2 S−3 S = S−2 . A n = 0.15) we conclude that S=I and the proof is complete. 1.Pielou’s This is the (1. (4.Pielou’s Equation 57 In . The Case α = γ = B = 0 : xn+1 = and Then βxn A+Cxn−1 .15) 2 4. S= B C B+C + ≤ S−1 S−2 I C B+C B + ≥ .14) from which it follows that (4. 2)-type Eq(4.

(b) Assume y0 ∈ (0. Then the zero equilibrium of Eq(4. So the interesting case for Eq(4.5 The Case α = γ = A = 0 : xn+1 = β . p.75).17) holds. (1.16) was investigated in [55]. 4.16) possesses the unique positive equilibrium y = p − 1. 2)-Type Equations This equation was proposed by Pielou in her books ([66]. Theorem 4.16) is unstable and Eq(4.) When p ≤ 1. p + yn−1 n = 0.4.16) is when p > 1.16) is globally asymptotically stable. B + Cyn βxn Bxn +Cxn−1 This is the (1.1 (a) Assume p ≤ 1. ∞) and p > 1.16) is globally asymptotically stable. Eq(4. .4. 2)-type Eq(4. . Then the positive equilibrium y = p − 1 of Eq(4.6) which by the change of variables xn = reduces to the difference equation yn+1 = p + yn . t≥0 dt P which is a prototype of modelling the dynamics of single-species. 1.79) as a discrete analogue of the delay logistic equation dN N (t − τ ) = rN (t) 1 − .16) that every positive solution converges to 0. Furthermore. In fact by employing Theorem 1.17) In this case the zero equilibrium of Eq(4.58 Chapter 4. (4.2. (4. We summarize the above discussion in the following theorem. p. p.18) . (See also ([42]. x is globally asymptotically stable. it follows that when (4. . it follows from Eq(4.22) and ([67]. 0 is locally asymptotically stable when p ≤ 1 and unstable when p > 1. which is locally asymptotically stable.

γ Eq(4.19) is a consequence of Theorem 1. ∞). 1.19) and they are both unstable.6 The Case α = β = C = 0 : xn+1 = γ yn .4. p. p + yn n = 0.1. The oscillatory character of the solutions of Eq(4.1 (e). .19) was investigated in [28]. The following local result is a straightforward application of the linearized stability Theorem 1. 2)-type Eq(4. .6.19) has semicycles of length one.5) that 1 yn+1 < yn−1 p and so when p>1 for n ≥ 0 . Then 0 is the only equilibrium point of Eq(4.1 (a) Assume p > 1. B γxn−1 A+Bxn This is the (1.7) which by the change of variables xn = reduces to the difference equation yn+1 = where p= yn−1 .19) A ∈ (0. p= 4. C Eq(4.19) and it is locally asymptotically stable.18) was investigated in ([42] Corollary 3. In fact.7. The Case α = β = C = 0 : xn+1 = where γxn−1 A+Bxn 59 B . except possibly for the first semicycle. . 0 is a repeller and y = 1 − p is a saddle point equilibrium. every solution of Eq(4. Lemma 4. (b) Assume p < 1. It follows directly from Eq(4. 73) where it was shown that its equilibrium one is globally asymptotically stable . That is.1.6. (4.1. Then 0 and y = 1 − p are the only equilibrium points of Eq(4.4.19) (see also Section 2.

Finally when p<1 by invoking the stable manifold theory one can see that there exist nonoscillatory solutions which converge monotonically to the positive equilibrium 1 − p. . φ. φ.19) is globally asymptotically stable. On the other hand when p = 1. . n = 0. It follows from the identity (1 − p) − yn yn−1 .7 The Case α = β = A = 0 : xn+1 = γ Byn γxn−1 Bxn +Cxn−1 This is the (1. 2)-type Eq(4. ∞). . . (1. . B This equation was investigated in [7] where it was shown that the following statements are true: p= . . . 2)-Type Equations the zero equilibrium of Eq(4.19) converges to a period-two solution . Furthermore we can see that φψ = 0 but whether there exists solutions with φ=ψ=0 remains an open question. yn+1 − yn−1 = p + yn that every oscillatory solution is such that the subsequences of even and odd terms converge monotonically one to ∞ and the other to 0. yn+1 < yn−1 for n ≥ 0 and so every positive solution of Eq(4. . . 4.9) which by the change of variables xn = reduces to the difference equation yn+1 = p + where yn−1 . . 1.20) C ∈ (0. 1. . ψ. . ψ. (4. yn n = 0. .60 Chapter 4.

1 The Case p < 1 Here we show that there exist solutions of Eq(4. with the possible exception of the first semicycle.20) about the equilibrium point y = p + 1 is zn+1 + 1 1 zn − zn−1 = 0. 1] and y2 ≥ p + y0 and use induction to complete the proof. .20) is globally asymptotically stable. The Case α = β = A = 0 : xn+1 = γxn−1 Bxn +Cxn−1 61 (i) p ≥ 1 is a necessary and sufficient condition for every solution of Eq(4. It suffices to show that y1 ∈ (p.4. We now proceed to establish the above results. and so y0 > p + 1. n→∞ y2n+1 = p.7.1. Then the following statements are true: 1.20) which consists of at least two semicycles.21) As a simple consequence of the linearized stability Theorem 1. Theorem 4. Theorem 4.1.7. Clearly the only equilibrium point of Eq(4.20) which are unbounded. (iii) When p > 1. . .2 (a) Let {yn }∞ n=−1 be a solution of Eq(4. the equilibrium y = p + 1 of Eq(4. p+1 p+1 n = 0.20) is y = p + 1.20) converges to a period-two solution. and let {yn }∞ n=−1 be a solution of Eq(4.1 we obtain the following result: Theorem 4.7. every solution of Eq(4. In fact the following result is true. The next result about semicycles is an immediate consequence of some straightforward arguments and Theorem 1. Moreover. Proof.3 Let p < 1. ∞ (b) Let {yn }n=−1 be a solution of Eq(4. every semicycle has length one. 1.7.7. (4. lim y2n = ∞. Note that 1 1−p > p + 1.20) such that 0 < 1 y−1 ≤ 1 and y0 ≥ 1−p . Then {yn }∞ n=−1 converges monotonically to y = p + 1.7.20) is locally asymptotically stable when p > 1 and is an unstable saddle point when p < 1.20) which consists of a single semicycle. (ii) When p = 1. . .1 The equilibrium point y = p + 1 of Eq(4.20) to be bounded. The linearized equation of Eq(4. 4. Then {yn }∞ n=−1 is oscillatory. n→∞ lim 2.

See also Section 2.7.7. y2n−1 . Also. 4. (1. . The following result will be useful.20) is globally asymptotically stable. Then the following statements are true.1 and some straightforward arguments.20) converges to a period-two solution if and only if p = 1. and let {yn }∞ n=−1 be a solution of Eq(4.20). that p < yn for all n ≥ −1.7. Note that for n ≥ 0. p2 We shall first show that lim supn→∞ yn ≤ p−1 .2 The Case p = 1 The following result follows from Lemma 2. ∞ (b) Suppose {yn }∞ n=−1 consists of at least two semicycles. Then p+ p2 p−1 ≤ lim inf yn ≤ lim sup yn ≤ . 2)-Type Equations > p.20).62 Indeed y1 = p + y1 y0 Chapter 4. 1]. y1 = p + y−1 1 ≤p+ ≤ 1. 4. (c) Every solution of Eq(4. ∞ (a) Suppose {yn }n=−1 consists of a single semicycle. Lemma 4. n→∞ p p−1 n→∞ Proof. . 1.7. Then {yn }n=−1 converges to a prime period-two solution of Eq(4.3 The Case p > 1 Here we show that the equilibrium point y = p + 1 of Eq(4. y0 y0 2 and so y1 ∈ (p.7. and let {yn }∞ n=−1 be a solution of Eq(4. p m = 0.20). Theorem 4.7. and that p < y0 < p + 1 < y−1 . Then {yn }∞ n=−1 converges monotonically to y = 2.1 Let p > 1. It follows from Theorem 4.4 Let p = 1. . p .2 that we may assume that every semicycle of {yn }∞ n=−1 has length one. y2n+1 < p + So as every solution of the difference equation 1 ym+1 = p + ym .5. Hence y2 = p + y0 y1 ≥ p + y0 .

there exist no solutions of Eq(4. Hence p < f (x. Recall that by Theorem 4. n→∞ y f (x.20).5 Let p > 1. y2n−1 p + p2 + is arbitrary. p2 − p p−1 p2 + . ∞). Then the equilibrium y = p + 1 is globally asymptotically stable. ∞) × (0.4. y ∈ (0. (0. x Then f ∈ C[(0. it follows that lim supn→∞ yn ≤ p−1 .7. set p−1 p2 + >p . So let {yn }∞ n=−1 be a solution of Eq(4. We know by Theorem 4. and f is increasing in y ∈ (0.7. There clearly exists y2n = p + So as p−1 p3 + p + p(p − 1) y2n−2 >p+p 2 = .20). ∞). p−1 f p. We shall next show that p + p−1 ≤ lim inf n→∞ yn .1 that y = p + 1 is a locally asymptotically stable equilibrium point of Eq(4. y) = p + . It suffices to show that lim yn = p + 1.7. y) < for all x. . ∞). ∞).20) with prime period two. lim inf yn ≥ 2 n→∞ p p 2 We are now ready for the following result. Proof. The Case α = β = A = 0 : xn+1 = 2 γxn−1 Bxn +Cxn−1 2 63 p p converges to p−1 . Let > 0.7. Then > 0. ∞) for each y ∈ (0. f is decreasing in x ∈ (0.4. y ∈ p. p2 + y2n−1 < . p−1 Let n ≥ N . Let p N ≥ 0 such that for all n ≥ N . Theorem 4. and set a = p and b = Note that f and p2 + .p = p + p 2 p−1 p + p2 + p−1 p2 + p−1 = p2 + p3 + p = . p−1 For x. ∞)] . we have p−1 p3 + p(p − 1) =p+ . ∞) for each x ∈ (0.

xn−l n = 0.8. . 2)-Type Equations and so by Theorem 1. x0 are arbitrary positive real numbers.1 Consider the difference equation where and where the initial conditions x−l . . (4. Show that every solution of Eq(4. Determine the set of all positive initial conditions x−l . y0 with y−1 y0 < 0 for which the equation yn+1 = B A . φp . . 1.8. is well defined for all n ≥ 0. . . either to the equilibrium or to a periodic solution with period p ≥ 2 and that what happens and the exact value of p are uniquely determined from the characteristic roots of the linearized equation λl+1 + (See [19].23) is well defined for all n ≥ 0. . .8.} with k < l. p<p+ Chapter 4.1 Let . . . .) Open Problem 4. . . 1. [23].22). . 2 4.4. l ∈ {0. .8 Open Problems and Conjectures xn+1 = A xn−k + B . . and [61]. x0 such that {xn }∞ converges to this periodic solution. be a given prime period p solution of Eq(4. . [25]. ∞).64 Finally note that by Lemma 4. φ2 . + yn yn−1 (4. . Is it bounded? Does limn→∞ yn exist? When does {yn } converge to a two cycle? . (1.22) Conjecture 4. . . .2 (see [18]) Assume that A. [60]. B ∈ (0.1. n=−l Open Problem 4. . φ1 . p−1 p2 + p2 ≤ lim inf yn ≤ lim sup yn ≤ < n→∞ p p−1 p−1 n→∞ n→∞ lim yn = p + 1. (a) Find all initial conditions y−1 .23) A B λl−k + = 0. (b) Assume that the initial conditions y−1 . B ∈ (0.22) converges. Investigate the global behavior of the solution {yn }. y0 are such that y−1 y0 < 0 and such that Eq(4.7. .6. . . A+B A+B A. ∞) and k.

Open Problems and Conjectures 65 Conjecture 4.2 Assume that p = 1. the asymptotic behavior. y0 ∈ (0.8.8. For such an initial point.3 Assume that p < 1. Open Problem 4. Conjecture 4. Determine all initial conditions y−1 . y0 ∈ R such that the equation (4. x0 ) ∈ R × R through which the equation xn−1 xn+1 = −1 + (4. Open Problem 4.8.4.8.4 Determine the set G of all initial conditions (y−1 .20) is well defined for all n ≥ 0.8.4 Assume that (x−1 . investigate the global character ∞ of {yn }n=−1 .19) is bounded. y0 ) ∈ R × R through which the equation yn−1 yn+1 = 1 + yn is well defined for all n ≥ 0 and.8. Open Problem 4. Conjecture 4.20) possesses a solution {yn }∞ n=−1 which remains above the equilibrium for all n ≥ −1.20). y0 ∈ R be such that Eq(4. Determine the set of initial conditions y−1 . Show that Eq(4. ∞) ∞ for which the solution {yn }n=−1 of Eq(4.24) xn is well defined for all n ≥ 0.8. ∞).20) is well defined for all n ≥ 0.6 (See [13]) Find the set G of all initial points (x−1 .8.5 (a) Assume p ∈ (0. . (c) Discuss (a) and (b) when p < 0. x0 ) ∈ R × R is such that the equation (4. Open Problem 4. Show that {xn }∞ n=−1 is a three cycle.19) has a solution which converges to zero.24) is well defined for all n ≥ 0 and bounded. and the periodic ∞ nature of the solution {yn }n=−1 of Eq(4. investigate the boundedness.3 Show that Eq(4. (b) Let y−1 . for these initial points.

ai ∈ (0.5 (See [27]) Assume r ∈ (0. k.8 Let a. . . . . with positive initial conditions. . xn + xn−1 n = 0. + ak xn−k n = 0. ∞)]. ∞) does the sequence defined by xn−1 . . Obtain necessary and sufficient conditions on f for every positive solution of the equation xn+1 = to be bounded. . with positive initial conditions is globally asymptotically stable if and only if kπ α−1 . [66]. .8. . 1. and [67]. 1. . Show that the positive equilibrium of Pielou’s discrete logistic model xn+1 = αxn .8. ∞) and k ∈ {0.7 Let f ∈ C[(0.}. 1. . see [55]. . Conjecture 4. converges to a period-two solution. . . xn+1 = 1 + xn converges to two? . Open Problem 4. ∞) for i = 0. n = 0. (0. a + a0 xn + .8. .8. Conjecture 4. 2)-Type Equations Open Problem 4. x0 ∈ (0. . 1. Show that every positive solution of the population model Jn+1 = Jn−1 er−(Jn +Jn−1 ) . . Investigate the global asymptotic stability of the equilibrium points of the difference equation xn+1 = xn . . n = 0.6 (Pielou’s Discrete Logistic Model. f (xn ) . < 2 cos α 2k + 1 Open Problem 4. (1. 1. . ∞). . . .66 Chapter 4.8. 1 + xn−k n = 0. 1. .9 (See [5]) For which initial values x−1 . .) Assume α ∈ (1. ∞).

. In other words. .8. .26) (4. 1.20) are bounded. ∞ More precisely. 1). Investigate the asymptotic behavior of {yn }n=−1 . 1 + yn−1 n = 0. .25).4.25) converge to zero.8. 1. n = 0. y0 ∈ [0. Open Problem 4.7) Open Problem 4.25) converges to a period-two solution . lim p = n→∞ pn . ∞ One can easily see that every solution {yn }n=−1 of Eq(4. (4. .8. .27) . yn yn−1 pn yn . . each of the subsequences {y2n }∞ and {y2n+1 }n=−1 is decreasing and n=0 lim [n→∞ y2n ][n→∞ y2n+1 ] = 0. 67 (4. . lim (a) Find all initial points y−1 . n = 0. ∞) such that the solutions {yn }∞ n=−1 of Eq(4. Open Problems and Conjectures (See also Section 4.25). 0. . find the basin of attraction of the equilibrium of Eq(4. ∞) through which the solutions of Eq(4. Investigate the asymptotic behavior and the periodic nature of all positive solutions of each of the following difference equations: yn+1 = yn+1 = 1 pn + . (a) Find the set B of all initial conditions y−1 . . .12 Let {pn }∞ be a convergent sequence of nonnegative real numn=0 bers with a finite limit.11 Assume p ∈ (0. 0. .10 (See [31]) Consider the difference equation yn+1 = yn−1 e−yn .25) with nonnegative initial conditions. 1.8. . . y0 ∈ (0. (b) Determine the limits of the subsequences of even and odd terms of every solution of Eq(4. in terms of the initial conditions y−1 and y0 of the solution. . y0 ∈ B. ∞ (b) Let y−1 . . Open Problem 4. . .

n = 0.28) (4. 1 − yn (4. . yn Chapter 4.29) (4. 1 − yn−1 yn−1 . y0 ) ∈ R × R through which the equation is well defined for all n ≥ 0.33) (4. yn For those equations from the above list for which you were successful. . . pn + yn−1 yn−1 .8.26) . (4. pn + yn yn−1 . .30).8. 1.34) yn+1 = yn+1 = −1 + yn−1 .68 pn + yn .32) (4. 2)-Type Equations yn+1 = n = 0.30) yn+1 = yn+1 = pn + Open Problem 4. extend your result by introducing arbitrary real parameters in the equation. Investigate the global character of all positive solutions of each of Eqs(4.(4. . 1 − yn−1 1 + yn . .14 For each of the following difference equations determine the “good” set G ⊂ R × R of all initial conditions (y−1 . n = 0. y0 ) ∈ G. 1. investigate the long term behavior of the solution {yn }∞ n=−1 : yn+1 = yn . Open Problem 4. Then for every (y−1 . . . 1. . .13 Let {pn }∞ be a periodic sequence of nonnegative real numbers n=0 with period k ≥ 2. . (1.31) yn+1 = (4.

. . (5.5) xn+1 = α + γxn−1 . 1)-type: xn+1 = n = 0. . . . Cxn−1 69 n = 0. 1. . 1)-Type Equations 5. . Bxn n = 0. . .4) xn+1 = α + γxn−1 . 1. 1.6) . . Bxn n = 0.1 Introduction α + βxn . (5. . 1. Cxn−1 n = 0. . (5. A n = 0. (5. (5. . 1. . (5. 1.Chapter 5 (2. . . A Eq(1) contains the following nine equations of the (2. . .2) xn+1 = α + βxn .1) xn+1 = α + βxn .3) xn+1 = α + γxn−1 .

70 Chapter 5. 1. .9) to Eq(4. yn−1 α+βxn Cxn−1 . γ Finally the change of variables xn = γ β + yn .2) is a Riccati equation and Eq(5. . which was investigated in Section 4. (5.8) βxn + γxn−1 . (5.20) with p = β . . . (5. (5. . A n = 0. Cxn−1 n = 0. .Lyness’ This is the (2. . 1.1).10) . which was investigated in Section 4. the initial conditions are nonnegative.7) xn+1 = and xn+1 = βxn + γxn−1 . Eqs(5. Of these nine equations. and the denominators are always positive. 1.8) to Eq(4. . . B reduces Eq(5.5.3) and (5. (2.7) are linear.9) Please recall our classification convention in which all parameters that appear in these equations are positive. .4) and (5.3) which by the change of variables reduces to the equation yn+1 = n = 0. . which will be investigated in the next two sections. Therefore there remain Eqs(5. C C γ reduces Eq(5.2). 1)-type Eq(5. 5. .6) is essentially like Eq(5.7. 1)-Type Equations xn+1 = βxn + γxn−1 . 1. Eq(5. (5.5).18) with p = β . .2 The Case γ = A = B = 0 : xn+1 = Equation p + yn . Bxn n = 0. The change of variables γ xn = yn .

See [11]. p. . . and [72] and the references cited therein.10) possesses the invariant In = (p + yn−1 + yn ) 1 + 1 yn−1 1+ 1 yn = constant (5. Then the following statements about the positive solutions of Eq(5. [43]. y0 .Lyness’ Equation 71 p= The special case of Eq(5. the equation becomes 1 + yn yn+1 = . n = 0. .11) with initial conditions y−1 and y0 is the five-cycle: y−1 .10) is bounded from above and from below by positive constants. It was also shown in [44] by using KAM theory that the positive equilibrium y of Eq(5.10) the following result was established in [43]: Theorem 5. .10) where αC . . y−1 y−1 y0 y0 Eq(5. It was shown in [30] that no nontrivial solution of Eq(5.10) is stable but not asymptotically stable. contains at least two and at most three terms. 131) for the history of the problem.. (See ([42].12) from which it follows that every solution of Eq(5. . [42].5. The Case γ = A = B = 0 : xn+1 = where α+βxn Cxn−1 . . (b) Every nontrivial semicycle. 1.2.) In this special case. Indeed the solution of Eq(5..10) has a limit.1 Assume that p > 0. [57]-[59].2. 1 + y0 1 + y−1 + y0 1 + y−1 . This result was also established in [49] by using a Lyapunov function.10) are true: (a) The absolute extreme in a semicycle occurs in the first or in the second term. [71]. There is substantial literature on Lyness’ Equation..11) yn−1 every solution of which is periodic with period five. after the first one. Concerning the semicycles of solutions of Eq(5. (5. β2 p=1 was discovered by Lyness in 1942 while he was working on a problem in Number Theory.

5) which by the change of variables xn = reduces to the equation yn+1 = where p= p + yn−1 . yn (5. Open Problem 5.13) is a saddle point. . From this it follows that a solution of Eq(5. .4.13) is a consequence of the identity yn+1 − yn−1 = yn−1 − yn−2 yn for n ≥ 1. By Theorem 1. By the linearized stability Theorem 1.3 The Case β = A = C = 0 : xn+1 = γ yn B n = 0. To this day we have not found a proof for the boundedness of solutions of Eq(5. α+γxn−1 Bxn This is the (2. .10) without using essentially the invariant.13) converges monotonically to the equilibrium y. 1)-type Eq(5.1 one can see that except possibly for the first semicycle.10) is bounded from above and from below by positive numbers.12). γ2 Eq(5. (2.13) either approaches the positive equilibrium y monotonically or it oscillates about y with semicycles of length one in such a way that {y2n } and {y2n+1 } converge monotonically one to zero and other to ∞.1 Show that every solution of Lyness’ Eq(5.13) was investigated in [28]. 5. every oscillatory solution of Eq(5.1 one can see that the unique equilibrium point √ 1 + 1 + 4p y= 2 of Eq(5. .13) has semicycles of length one.4 Open Problems and Conjectures By using the invariant (5. Concerning the nonoscillatory solutions of Eq(5.72 Chapter 5.12). The global character of solutions of Eq(5. 1.13) αB .7.13) one can easily see that every nonoscillatory solution of Eq(5. 1)-Type Equations 5.1. one can easily see that every solution of Lyness’ Eq(5.10) is bounded without using the invariant (5.

f (xn ) . ∞). . What is it that makes Lyness’ equation possess an invariant? What type of difference equations possess an invariant? Along these lines.14) Open Problem 5.4. ∞). Determine the set G of all initial conditions y−1 . (5.15) are unbounded is a set of measure zero. When we allow the initial conditions to be real numbers the equation yn+1 = p + yn yn−1 (5.4.) Assume p ∈ (−∞. (0.4. . Open Problem 5. Conjecture 5.4.15) may not even be well defined for all n ≥ 0. ∞). 1.4.14) about x does not have both eigenvalues in |λ| < 1 and does not have both eigenvalues in |λ| > 1. Conjecture 5. ∞)]. ∞).15) is well defined for all n ≥ 0. To illustrate. Show that the subset of G through which the solutions of Eq(5. Open Problems and Conjectures 73 Open Problem 5. . Obtain necessary and sufficient conditions in terms of f so that every positive solution of the difference equation xn+1 = is bounded. (0. (0. y0 ∈ (−∞.3 Assume that f ∈ C[(0. xn−1 n = 0.2 Let G be the set of initial conditions defined in the Open problem 5. Obtain necessary and sufficient conditions in terms of f so that the difference equation (5.4 (See [26]. we offer the following open problems and conjectures. . ∞)] and that the difference equation (5.14) possesses a unique positive equilibrium point x and a nontrivial invariant. Show that the linearized equation of Eq(5.4.1 Assume that f ∈ C[(0. ∞)]. we offer the following open problems and conjectures. ∞) through which Eq(5.14) possesses a (nontrivial) invariant.4.4.2 Assume that f ∈ C[(0.5.

4 Show that every positive solution of the equation yn+1 = converges to a period-six solution. y0 ) ∈ G. . Conjecture 5. (2. 1. + yn yn−1 yn−3 yn−4 n = 0.6 Show that the equation yn+1 = 1 + yn−1 . yn yn−4 + . yn−k−1 n = 0. .4.4.3 Assume p ∈ (−∞. y0 ) ∈ (−∞. + yn−k . ∞).4. .4. 1)-Type Equations Conjecture 5. 1. .6 Find the set G of all initial points (y−1 . Conjecture 5. . 1.5 Determine all positive integers k with the property that every positive solution of the equation yn+1 = is periodic. 1. Determine the periodic character and the asymptotic behavior of all solutions with (y−1 . . Show that the set of points (y−1 . . Open Problem 5.4. . yn n = 0. . .5 (See [20]) Show that every positive solution of the equation yn+1 = 1 1 . yn−1 yn−3 n = 0. . y0 ) ∈ R × R through which the equation 1 + yn−1 yn+1 = yn is well defined for all n ≥ 0. converges to a period-three solution. ∞) × (−∞. 1 + . . Conjecture 5. Open Problem 5.74 Chapter 5. . has a nontrivial positive solution which decreases monotonically to the equilibrium of the equation. is a set of points without interior. . . ∞) through which Eq(5.4.15) is not well defined for all n ≥ 0.

4.16) to where xn =   −1+yn 2 if A > 1 if A = 1 if A < 1 n = 0. see [59]) Assume α > 1. see [59]) Let A ∈ (0.5. . .7 (May’s Host Parasitoid Model. xn xn−1 n = 0. . . Show that every positive solution of the equation xn+1 = is bounded. 2 αxn . (5. A} n . 1 if A < 1 if A = 1 δ= 0  −1 if A > 1. . Show that every positive solution of May’s Host Parasitoid Model xn+1 = is bounded. Conjecture 5.    When δ = 1. (1 + xn )xn−1 n = 0.16) A A 1+yn 2 reduces Eq(5. . Open Problems and Conjectures 75 Conjecture 5. (5. Eq(5. Note that the change of variables      n e y2 max{x2 . 1. ∞).17) is called the Gingerbreadman Map and was investigated by Devaney [17]. . . 1. .17) yn+1 = |yn | − yn−1 + δ. 1.4.8 (The Gingerbreadman Map.4. .

.

2)-type: xn+1 = n = 0. 2)-Type Equations 6. 1.1) xn+1 = α + βxn . (6.4) xn+1 = α + γxn−1 . Bxn + Cxn−1 77 n = 0. . . . . (6. .1 Introduction α + βxn . A + Bxn Eq(1) contains the following nine equations of the (2. A + Cxn−1 n = 0. . .6) . . (6. 1. 1. .2) xn+1 = α + βxn . .5) xn+1 = α + γxn−1 .Chapter 6 (2. . A + Bxn n = 0. (6. 1. . .3) xn+1 = α + γxn−1 . . . . (6. . 1. A + Cxn−1 n = 0. 1. (6. Bxn + Cxn−1 n = 0. .

See Section 1. . A + Czn n = 0. To avoid a degenerate situation we will assume that αB − βA = 0.5).5) is essentially similar to Eq(6. (6. . . for n ≥ −1. A + Bxn n = 0. (2. (6. α + γzn . A + Cxn−1 n = 0. .1). .7) xn+1 = βxn + γxn−1 . In fact if {xn }∞ n=−1 is a positive solution of ∞ ∞ Eq(6. Bxn + Cxn−1 n = 0.6. .9) Please recall our classification convention in which all parameters that appear in these equations are positive. . . and the denominators are always positive.8) xn+1 = βxn + γxn−1 . 0. 2)-type Eq(6.78 Chapter 6.Riccati This is the (2. . . . (6. 1. 1. .5) then {x2n }n=0 and {x2n−1 }n=0 are both solutions of the Riccati equations zn+1 = with zn = x2n and zn+1 = with zn = x2n+1 respectively. 1. 1. . 2)-Type Equations xn+1 = βxn + γxn−1 . . A + Czn for n ≥ 0 n = −1.2 The Case γ = C = 0 : Equation xn+1 = α+βxn A+Bxn . . . the initial conditions are nonnegative. .1) which is in fact a Riccati equation. α + γzn . 6. Therefore we will omit any further discussion of Eq(6. Eq(6.

.1 Every solution of Eq(6. By applying the linearized stability Theorem 1.10) is bounded from above and from below by positive constants.2.1 The positive equilibrium of Eq(6. The main results known about Eq(6.6. See also [36].) Eq(6. . 2 A A (Eq(6. 2)-type Eq(6. Lemma 6. To this day.1. 1.3.1) is globally asymptotically stable. 1 + yn−1 n = 0. . the conjecture has not been proven or refuted.10) is a very simple looking equation for which it has long been conjectured that its equilibrium is globally asymptotically stable. Eq(6.6.2) which by the change of variables xn = reduces to the equation yn+1 = where p + qyn .2: Theorem 6.10) has the unique positive equilibrium y given by p= y= q−1+ (q − 1)2 + 4p 2 .10) was investigated in [42] and [43].1 The equilibrium y of Eq(6. 4 Now the following result is a consequence of Theorem 1.3 The Case γ = B = 0 : xn+1 = A yn C α+βxn A+Cxn−1 This is the (2. 6. The Case γ = B = 0 : xn+1 = α+βxn A+Cxn−1 79 The Riccati number associated with this equation is R= and clearly βA − αB (β + A)2 1 R< . (6. .3.1 we obtain the following result.10) αC β and q = .10) is locally asymptotically stable for all values of the parameters p and q.3.10) are the following: Theorem 6.

Furthermore after the first. Proof. if the solution is bounded from above by a constant M . Then. Now assume for the sake of contradiction that the solution is not bounded from above.80 Chapter 6.3. . 2)-Type Equations Proof. Let {yn } be a solution of Eq(6. and y1+nk = max{yn : n ≤ nk } for k ≥ 0.10).10) is described by the following result: ∞ Theorem 6. The oscillatory character of solutions for Eq(6. in every semicycle. for sufficiently large k. then p yn+1 ≥ 1+M and so it is also bounded from below. their extrema cannot be consecutive terms. 0 ≤ y1+nk − ynk = p + [(q − 1) − ynk −1 ]ynk <0 1 + ynk −1 2 which is a contradiction and the proof is complete.10). k→∞ Hence. k→∞ lim y1+nk = ∞.2 Let {yn }n=−1 be a nontrivial solution of Eq(6. The proofs for negative semicycles are similar and will be omitted. has at least two terms. (iii) In any two consecutive semicycles. every semicycle contains two or three terms. (ii) The extreme in each semicycle occurs at either the first term or the second. (iv) Assume q ≤ p. Furthermore. there is at most one term which follows the extreme. Clearly. except perhaps for the first one. the remaining terms in a positive semicycle are strictly decreasing and in a negative semicycle are strictly increasing. From (6. except possibly for the first semicycle. (2.10) we see that yn+1 < qyn + p for n ≥ 0 and so k→∞ lim ynk = lim ynk −1 = ∞. We present the proofs for positive semicycles only. Then there exists a subsequence {y1+nk }∞ such k=0 that k→∞ lim nk = ∞. Then the following statements are true: (i) Every semicycle.

Then yN +3 = p + qyN +2 p + qyN +1 p + qy = y. α+βxn A+Cxn−1 81 yN −1 < y Then yN +1 = and yN ≥ y. 1 + yN −1 1 + yN (iv) Assume that for some N ≥ 0.3. Then yN ≥ y. yN +1 . > 1 + yN −1 1+y (ii) Assume that for some N ≥ 0. p + qyN p + qy = y. Then yN +1 = p + qyN p + qyN +1 < = yN +2 . yN +1 yN +1 1 + yN 1 + yN 1+y (iii) We consider the case where a negative semicycle is followed by a positive one. the first two terms in a positive semicycle are yN and yN +1 . the terms yN . So assume that for some N ≥ 0 the last two terms in a negative semicycle are yN −1 and yN with yN −1 ≥ yN .3.3 The positive equilibrium of Eq(6. (ii) q≥1 and either p ≤ q or q < p ≤ 2(q + 1).6. The opposite case is similar and will be omitted. and yN +2 are all in a positive semicycle. The Case γ = B = 0 : xn+1 = (i) Assume that for some N ≥ 0.10) is globally asymptotically stable if one of the following two conditions holds: (i) q < 1. yN +1 > y and p p +q +q yN +2 1 p + qyN +1 yN +1 = = < y = 1. < < 1 + yN +1 1 + yN +1 1+y 2 Theorem 6. .

3. B (6. . p + (q − 1)i ≤ is ≤ p + (q − 1)s i≥s p + qi 1+s and s ≤ p + qs . For the proof when see ([42].1. yn + qyn−1 αB β2 n = 0. 71). (ii) The proof when p<q follows by applying Theorem 1.1. (2. . . 1. p. 2)-type Eq(6.4. it remains to show that every solution {yn }∞ of Eq(6.10) n=−1 tends to y as n → ∞.4 The Case γ = A = 0 : xn+1 = β yn B α+βxn Bxn +Cxn−1 This is the (2. .10) yields i≥ Hence and so because q < 1.4. The following result is a consequence of the linearized stability Theorem 1.1 exist and are positive numbers.1. 1 ≤ q ≤ p ≤ 2(q + 1) 2 6.3. Theorem 3.2. This result was first established in [43].11) was investigated in [15].82 Proof. C .3) which by the change of variables xn = reduces to the difference equation yn+1 = where p= yn + p .11) and q = Eq(6.3 (f). let i = lim inf yn n→∞ and s = lim sup yn n→∞ which by Theorem 6. To this end. 2)-Type Equations (i) In view of Lemma 6. 1+i from which the result follows. Chapter 6. Then Eq(6.

yk−1 .4.4. . q for all n > k. ∞). (a) The basic ingredient behind the proof is the fact that when u. for some k ≥ 0.11) is locally asymptotically stable for all values of the parameters p and q. yk ∈ 1. The Case γ = A = 0 : xn+1 = α+βxn Bxn +Cxn−1 83 Theorem 6. the q function u+p f (u. yk + qyk−1 yk + q p q yk+1 yk + p = ≥ yk + qyk−1 p q p q +p +q > p q 2 and the proof follows by induction. then ym ∈ I for all m > k. Then yn ∈ 1. two consecutive values yk−1 and yk of a solution lie in I. v) = u + qv is increasing in u and decreasing in v.1 q p .11). 6.11) in the sense that if. yk ∈ Then yn ∈ p .1 Invariant Intervals The following result establishes that the interval I with end points 1 and p is an invariq ant interval for Eq(6. v ∈ [ p . Indeed. yk−1 . Proof. Then the following statements are true: (a) Assume p < q and suppose that for some k ≥ 0.1 Let {yn }∞ n=−1 be a solution of Eq(6. (b) The proof is similar and will be omitted. q for all n > k.1 The unique positive equilibrium point y= 1+ 1 + 4p(q + 1) 2(q + 1) of Eq(6. (b) Assume p > q and suppose that for some k ≥ 0.1 . p q p .4.6.4. yk + p yk + p yk+1 = ≤ = 1. Lemma 6.

84

Chapter 6. (2, 2)-Type Equations

6.4.2

Semicycle Analysis

∞ Let {yn }n=−1 be a solution of Eq(6.11). Then observe that the following identities are true: p − yn−1 (6.12) yn+1 − 1 = q q for n ≥ 0, yn + qyn−1

yn+1 − and yn − yn+4 =

p (q − p) + pq(1 − yn−1 ) = q q(yn + qyn−1 )

for n ≥ 0,

(6.13)

(yn − 1)[qyn yn+3 + yn+1 yn+3 ] + qyn+1 (yn − p ) q q(p + yn+1 ) + yn+3 (yn+1 + qyn )

for n ≥ 0.

(6.14)

First we will analyze the semicycles of the solution {yn }∞ n=−1 under the assumption that p > q. (6.15) The following result is a direct consequence of (6.12)-(6.15):
∞ Lemma 6.4.2 Assume that (6.15) holds and let {yn }n=−1 be a solution of Eq(6.11). Then the following statements are true:

(i) If for some N ≥ 0, yN −1 < p , then yN +1 > 1; q (ii) If for some N ≥ 0, yN −1 = p , then yN +1 = 1; q (iii) If for some N ≥ 0, yN −1 > p , then yN +1 < 1; q (iv) If for some N ≥ 0, yN −1 ≥ 1, then yN +1 < p ; q (v) If for some N ≥ 0, yN −1 ≤ 1, then yN +1 > 1; (vi) If for some N ≥ 0, yN ≥ p , then yN +4 < yN ; q (vii) If for some N ≥ 0, yN ≤ 1, then yN +4 > yN ; (viii) If for some N ≥ 0, 1 ≤ yN −1 ≤ p , then 1 ≤ yN +1 ≤ p ; q q (ix) If for some N ≥ 0, 1 ≤ yN −1 , yN ≤ p , then yn ∈ [1, p ] for n ≥ N . That is [1, p ] is q q q an invariant interval for Eq(6.11); (x) p 1<y< . q

6.4. The Case γ = A = 0 : xn+1 =

α+βxn Bxn +Cxn−1

85

The next result, which is a consequence of Theorem 1.7.2, states that when (6.15) holds, every nontrivial and oscillatory solution of Eq(6.11), which lies in the interval [1, p ], oscillates about the equilibrium y with semicycles of length one or two. q Theorem 6.4.2 Assume that Eq(6.11) holds. Then every nontrivial and oscillatory solution of Eq(6.11) which lies in the interval [1, p ], oscillates about y with semicycles q of length one or two.
∞ Next we will analyze the semicycles of the solutions {yn }n=−1 under the assumption that q = p. (6.16)

In this case, Eq(6.11) reduces to yn+1 = with the unique equilibrium point y = 1. Also identities (6.12)-(6.14) reduce to yn+1 − 1 = p and yn − yn+4 = (yn − 1) pyn+1 + pyn yn+3 + yn+1 yn+3 p(p + yn+1 ) + yn+3 (yn+1 + pyn ) for n ≥ 0. (6.19) 1 − yn−1 yn + pyn−1 for n ≥ 0 (6.18) yn + p , yn + pyn−1 n = 0, 1, . . . (6.17)

and so the following results follow immediately:
∞ Lemma 6.4.3 Let {yn }n=−1 be a solution of Eq(6.17). Then the following statements are true:

(i) If for some N ≥ 0, yN −1 < 1, then yN +1 > 1; (ii) If for some N ≥ 0, yN −1 = 1, then yN +1 = 1; (iii) If for some N ≥ 0, yN −1 > 1, then yN +1 < 1; (iv) If for some N ≥ 0, yN < 1, then yN < yN +4 < 1; (v) If for some N ≥ 0, yN > 1, then yN > yN +4 > 1.
∞ Corollary 6.4.1 Let {yn }∞ n=−1 be a nontrivial solution of Eq(6.17). Then {yn }n=−1 oscillates about the equilibrium 1 and, except possibly for the first semicycle, the following are true:

86

Chapter 6. (2, 2)-Type Equations

(i) If y−1 = 1 or y0 = 1, then the positive semicycle has length three and the negative semicycle has length one. (ii) If (1 − y−1 )(1 − y0 ) = 0, then every semicycle has length two. Finally we will analyze the semicycles of the solutions {yn }∞ under the assumption n=−1 that p < q. (6.20) The following result is a direct consequence of (6.12)-(6.14) and (6.20). Lemma 6.4.4 Assume that (6.20) holds and let {yn }∞ n=−1 be a solution of Eq(6.11). Then the following statements are true: (i) If for some N ≥ 0, yN −1 < p , then yN +1 > 1; q (ii) If for some N ≥ 0, yN −1 = p , then yN +1 = 1; q (iii) If for some N ≥ 0, yN −1 > p , then yN +1 < 1; q (iv) If for some N ≥ 0, yN −1 ≤ 1, then yN +1 > p ; q (v) If for some N ≥ 0, yN −1 ≤ p , then yN +1 > p ; q q (vi) If for some N ≥ 0, yN ≥ 1, then yN +4 < yN ; (vii) If for some N ≥ 0, yN ≤ p , then yN +4 > yN ; q (viii) If for some N ≥ 0,
p q p q

≤ yN −1 ≤ 1, then

≤ yN +1 ≤ 1;

(ix) If for some N ≥ 0, p ≤ yN −1 , yN ≤ 1, then yn ∈ [ p , 1] for n ≥ N . That is, [ p , 1] is q q q an invariant interval for Eq(6.11); (x) p < y < 1. q

The next result, which is a consequence of Theorem 1.7.4, states that when (6.20) holds, every nontrivial and oscillatory solution of Eq(6.11) which lies in the interval [ p , 1], after the first semicycle, oscillates with semicycles of length at least two. q Theorem 6.4.3 Assume that (6.20) holds. Then every nontrivial and oscillatory solution of Eq(6.11) which lies in the interval [ p , 1], after the first semicycle, oscillates with q semicycles of length at least two.

6.4. The Case γ = A = 0 : xn+1 =

α+βxn Bxn +Cxn−1

87

How do solutions which do not eventually lie in the invariant interval behave? First assume that p>q
∞ and let {yn }n=−1 be a solution which does not eventually lie in the interval I = [1, p ]. q Then one can see that the solution oscillates relative to the interval [1, p ], essentially in q one of the following two ways: (i) Two consecutive terms in ( p , ∞) are followed by two consecutive terms in (0, 1), q are followed by two consecutive terms in ( p , ∞), and so on. The solution never visits q the interval (1, p ). q (ii) There exists exactly one term in ( p , ∞), which is followed by exactly one term in q (1, p ), which is followed by exactly one term in (0, 1), which is followed by exactly one q term in (1, p ), which is followed by exactly one term in ( p , ∞), and so on. The solution q q visits consecutively the intervals

...,

p p p p , ∞ , 1, ,∞ ,... , (0, 1), 1, , q q q q

in this order with one term per interval. The situation is essentially the same relative to the interval [ p , 1] when q p < q.

6.4.3

Global Stability and Boundedness

Our goal in this section is to establish the following result: Theorem 6.4.4 when (b) Assume that q > 1 + 4p. Then every solution of Eq(6.11) lies eventually in the interval Proof. (a) We have already shown that the equilibrium y is locally asymptotically stable so it remains to be shown that y is a global attractor of every solution {yn }∞ n=−1 of Eq(6.11). Case 1 Assume p = q. It follows from (6.19) that each of the four subsequences {y4n+i }∞ n=0 for i = 1, 2, 3, 4
p ,1 q

(a) The equilibrium y of Eq(6.11) is globally asymptotically stable q ≤ 1 + 4p. (6.21)

.

First.88 Chapter 6. As in Case 2 we can see that every solution of Eq(6. v) = u + qv is decreasing in both variables and so it follows by applying Theorem 1. In this interval the q function u+p f (u. p ] is an invariant interval. we can see that if a solution is not eventually in [1. L3 . L1 . p ] converges to y.4.11) with two consecutive values in [1. 2)-Type Equations is either identically equal to one or else it is strictly monotonically convergent. assume that p < q. . it converges to a periodic solution with period-four which q is not the equilibrium. . . Theorem 1.22) is a periodic solution of Eq(6. L4 . By applying (6. L2 . p ].11) converges to y. 4 and so n→∞ for i = 1.5 applies and so every solution of Eq(6. lim yn = 1. (b) The proof follows from the discussion in part (a). Recall from Lemma 6.11) must eventually lie in the interval [ p .4.2 that [1. . 3. By q using an argument similar to that in the case p = q. v) = u+p u + qv is increasing in u and decreasing in v.11) n→∞ lim yn = y and the proof is complete. 2. .19) to the solution (6. q Hence for every solution {yn }∞ n=−1 of Eq(6.14) to this period four solution we obtain a contradiction and so every solution of Eq(6.11) must lie eventually in [1. Set Li = n→∞ y4n+i lim Thus clearly . we assume that p > q.11) with period four. . 3. 1]. (6. (2. 2 .4. In this interval the function q f (u. Case 2 Assume p = q. 2. 4.7 that every solution of Eq(6. Now. By applying (6. p ]. Under the assumption that q ≤ 1 + 4p.22) we see that Li = 1 for i = 1.

.1. .1 The equilibrium y of Eq(6. ψ. .4) which by the change of variables xn = reduces to the difference equation yn+1 = where p= p + qyn−1 . ∞) : φψ = p}. φ.24) are given by {φ. 1.23) is y= q−1+ (q − 1)2 + 4p 2 and by applying the linearized stability Theorem 1. (6. .23) was investigated in [28].2 (a) Eq(6. the following result is a consequence of the results in Section 2.7.23) and q = (Eq(6. A (6. 1 + yn αB A2 n = 0.6. Then the values φ and ψ of all prime period-two solutions (6.25) (b) Assume (6. Concerning prime period-two solutions.23) is locally asymptotically stable when q < 1.5.24) .5 The Case β = C = 0 : xn+1 = A yn B This is the (2. . .5.) The only equilibrium point of Eq(6. φ.1 we obtain the following result: Theorem 6. . Theorem 6. γ . The Case β = C = 0 : xn+1 = α+γxn−1 A+Bxn 89 α+γxn−1 A+Bxn 6.23) has a prime period-two solution .23) converges to a period two solution. ψ ∈ (0. (c) Assume that (6.25) holds.25) holds. 2)-type Eq(6. . if and only if q = 1. ψ. Then every solution of Eq(6.5. (6. and is an unstable saddle point when q > 1. .

1−q n ≥ 0.23). yn0 −1 ) < f (yn0 . It is sufficient to show that {yn } is a decreasing sequence for n ≥ K. (6.23) is globally asymptotically stable. Then the positive equilibrium of Eq(6.7. y) = y.90 Chapter 6.5. (a) This follows immediately from Theorem 1. (b) Assume p ≥ q. So assume for the sake of contradiction that for some n0 ≥ K. an oscillatory solution {yn } of Eq(6.5. the condition p ≥ q implies that the function f (x) = is decreasing. 1 + yn p + . yn0 > yn0 −1 .5.26) . Clearly yn+1 = Set M= p + qyn−1 < p + qyn−1 .23) oscillates about the equilibrium y with semicycles of length one.5.23).1 Semicycle Analysis Our results on the semicycles of solutions for Eq(6. Then yn0 +1 = which is impossible. Then every nonoscillatory solution of Eq(6. Clearly. Then the following statements are true: (a) After the first semicycle. We will assume that there exists a positive integer K such that yn−1 ≥ y for every n ≥ K. p + qyn0 −1 = f (yn0 .4 Assume that q < 1. Proof. yn0 ) < f (y.1 (b) Let {yn } be a nonoscillatory solution of Eq(6.23) converges monotonically to the equilibrium y.23) are as follows: Theorem 6. Proof. 1 + yn0 2 p + qx 1+x 6.2 The Case q < 1 The main result in this section is the following: Theorem 6. The case where yn−1 < y for n ≥ K is similar and will be omitted. (2.23) and let y denote the unique positive equilibrium of Eq(6. 2)-Type Equations 6.3 Let {yn } be a nontrivial solution of Eq(6.

one of which is monotonically approaching 0 and the other is monotonically approaching ∞. y) ∈ [a.23). M ]. The Case β = C = 0 : xn+1 = α+γxn−1 A+Bxn 91 where is some positive number. q−1 p + q(q − 1) p + qy−1 < <q−1 1 + y0 1 + y0 p + qy0 p + qy0 p > = y0 + 1 + y1 q q p q for n ≥ 1. b]. By applying Theorem 1. y2n−1 < q − 1 and y2n > y0 + n Hence n→∞ p . 1+x 1−q Then clearly. The local stability of y was established in Theorem 6.5. for all (x.3 The Case q > 1 In this case we will show that there exist solutions that have two subsequences. From Eq(6. lim y2n = ∞ p + qy2n−1 =0 1 + y2n and lim y2n+1 = lim n→∞ n→∞ and the proof is complete. In fact this is true for every solution {yn }∞ n=−1 whose initial conditions are such that the following inequalities are true: y−1 < q − 1 and y0 > q − 1 + To this end observe that y1 = and y2 = and by induction. Set p + qy p f (x. y) = .26) and Eq(6.6.6 it follows that y is a global attractor of all solutions of Eq(6. a = 0.1. and b = + .5.23). .4. b] × [a. it now follows that every solution of Eq(6.23) lies eventually in the interval [0. a ≤ f (x. y) ≤ b. The proof is complete. 2 6.5.

. Clearly the equilibrium y is the positive solution of the quadratic equation (1 + q)y 2 − y − p = 0.27) was investigated in [45]. . .30) (6.1.27) and q = (Eq(6. 1. (2. (6. F (u) = (1 + q)u2 − u − p.27) about y is zn+1 + q qy − p zn − zn−1 = 0.6) which by the change of variables xn = reduces to the difference equation yn+1 = where p= p + yn−1 . 1.29) . . . . B .28) By employing Theorem 1. 2)-type Eq(6. .1 we see that y is locally asymptotically stable when (q − 1)y < 2p and unstable (a saddle point) when (q − 1)y > 2p.27) has the unique positive equilibrium y= 1+ 1 + 4p(1 + q) 2(1 + q) . (6. qyn + yn−1 αC γ2 n = 0. 2)-Type Equations 6. The linearized equation associated with Eq(6.92 Chapter 6.29) is satisfied if and only if either q≤1 or q > 1 and F 2p q−1 > 0.6 The Case β = A = 0 : xn+1 = γ yn C α+γxn−1 Bxn +Cxn−1 This is the (2. If we now set it is easy to see that (6. C (6. 1+q (y + p)(1 + q) n = 0.) Eq(6.

27) is locally asymptotically stable when q < 1 + 4p and unstable. ψ. Furthermore when (6.6. φ.6.1 Let Existence and Local Stability of Period-Two Cycles . .6.32) holds. The following result is now a consequence of the above discussion and some simple calculations: Theorem 6. . .6. . . the period-two solution is “unique” and the values of φ and ψ are the positive roots of the quadratic equation t2 − t + p = 0. . and more precisely a saddle point equilibrium. . (6. φ. Then φ= p+φ qψ + φ and ψ = p+ψ . . The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 93 Similarly (6. ψ. .32) (6.27).31) 6. be a period-two cycle of Eq(6.30) is satisfied if and only if q > 1 and F 2p q−1 < 0. φ. φ. q−1 . ψ.1 The equilibrium y of Eq(6.27) has a prime period-two solution .6. . . if and only if q > 1 + 4p. ψ.2 Eq(6. . qφ + ψ It now follows after some calculations that the following result is true: Theorem 6. when q > 1 + 4p.

φ. the second iterate of T . n = 0. . evaluated at ψ We have  ∂g  ∂g (φ. we set un = yn−1 and vn = yn . . . φ ψ is a fixed point of T 2 . . Jacobian matrix JT 2 . 1.94 To investigate the local stability of the two cycle . ψ. . (2. ψ) = − . ψ) = . ψ) ∂v (φ. . By a simple calculation we find that T2 where g(u. 2)-Type Equations for n = 0. ψ. . and write Eq(6. . . ∞) × (0. v) h(u. . ∞) defined by: T Then u v = v p+u qv+u . v) + v and h(u. 1. v) = p+u qv + u u v = g(u. . . Let T be the function on (0. ψ) ∂u φ  JT 2 =   ψ ∂h (φ. . v) = Clearly the two cycle is locally asymptotically stable when the eigenvalues of the φ lie inside the unit disk. ψ) ∂h (φ. φ. ψ) ∂u ∂v where. qg(u.27) in the equivalent form: un+1 = vn vn+1 = p+un qvn +un . ∂g qψ − p (φ. v) p+v . ∂u (qψ + φ)2 ∂g (p + φ)q (φ. Chapter 6. ∂v (qψ + φ)2 .

36) (qψ − p)(qφ + ψ)2 + (qφ − p)(qψ + φ)2 + q 2 (p + ψ)(p + φ) − (qψ − p)(qφ − p) < (qψ + φ)2 (qφ + ψ)2 which is true if and only if p (q 3 + 2q 2 − 4qp + 2q 2 p + 2p − 3q) + q − p + q 2 p2 + qp − p2 < (qψ + φ)2 (qφ + ψ)2 q−1 which is true if and only if −q 3 p + 7q 2 p − 8qp2 + 4q 2 p2 + 4p2 − 7qp + 2q 2 − q + p − q 3 < 0 which is true if and only if −(p + 1)(q − 1)(q − which is true because q > 1 + 4p. + 2 (qψ + φ)2 ∂v (qφ + ψ) (qφ + ψ)2 Set S= and D= ∂h ∂g (φ.33) (6.6.34) is equivalent to (6. ψ) + (φ. ψ) ∂u ∂v ∂g ∂h ∂h ∂g (φ.1 that both eigenvalues of JT 2 unit disk |λ| < 1. ψ). ψ) (φ. Inequality (6.6. ψ) = − . First we will establish Inequality (6.35) (6.34). ψ) − (φ. ψ) = . p )[q − (1 + 4p)] < 0 p+1 . To this end observe that (6.1. ∂u (qφ + ψ)2 (qψ + φ)2 ∂h q 2 (p + φ)(p + ψ) qφ − p (φ. ψ) (φ. ∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1.33) is equivalent to the following three inequalities: S <1+D −1 − D < S D < 1. if and only if |S| < 1 + D < 2. The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 95 q(p + ψ)(qψ − p) ∂h (φ.34) (6.

Observe that (6. (a) Assume that p ≤ q and that for some N ≥ 0.36). p ≤ yN ≤ 1.35). Observe that (6.27) either remain forever outside the interval I or the solution is eventually trapped into I. 6.35) is equivalent to −(qψ + φ)2 (qφ + ψ)2 which is true if and only if < (qψ − p)(qφ + ψ)2 + (qφ − p)2 (qψ + φ)2 + q 2 (p + ψ)(p + φ) + (qψ − p)(qφ − p) −(qψ + φ)2 (qφ + ψ)2 < p (q 3 + 4q 2 − 4qp + 2q 2 p + 2p − 3q) − p + q 2 p2 q−1 which is true if and only if q 2 + 2pq + 2p2 q 2 − p2 + qp2 − q 3 − 3q 3 p − 2q 3 p2 − p < 0 which is true if and only if p2 (−2q 3 + 2q 2 + q − 1) + p(−3q 3 + 2q − 1) − q 3 + q 2 < 0 which is true because q > 1.3 Let {yn }∞ n=−1 be a solution of Eq(6.6.96 Chapter 6. q Then p ≤ yn ≤ 1 for all n ≥ N. we establish Inequality (6.2 Invariant Intervals Here we show that the interval I with end points one and p is an invariant interval for q Eq(6.6. (2. Finally. 2)-Type Equations Next we will establish Inequality (6. Theorem 6. q . More precisely we show that the values of any solution {yn }∞ n=−1 of Eq(6.36) is true if and only if (qψ − p)(qφ − p) < (qψ + φ)2 (qφ + ψ)2 which is true if and only if p [q 2 − (q − 1)(q − p)] < (pq − p + q)2 q−1 which is true if and only if q(pq − p + q)(pq − 2p + q − 1) > 0 which is true because q > 1 + 4p.27).27). Then the following statements are true.

6.6. (6.27).38) (6.27) relative to the equilibrium y and relative to the end points of the invariant interval of Eq(6. for n ≥ 0.3 Semicycle Analysis Here we present a thorough semicycle analysis of the solutions of Eq(6.6. The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 97 (b) Assume that p ≥ q and that for some N ≥ 0.37) yn+1 − 1 = q q qyn + yn−1 qyn−1 (1 − p ) + pq(1 − yn ) p q yn+1 − = q q(qyn + yn−1 ) yn − yn+2 = q(p + yn−1 ) + yn (qyn + yn−1 ) for n ≥ 0. Then observe that the following identities are true: p − yn for n ≥ 0. (6. (b) Clearly. p + yN −1 p + yN −1 ≥ = q qyN + yN −1 (pyN + p yN −1 ) p q and the proof follows by induction.27). p 1 ≤ yN ≤ . yN +1 = and yN +1 = p + yN −1 p + yN −1 =1 ≥ qyN + yN −1 p + yN −1 p q p + yN −1 ≤ pyN + p yN −1 q p q p p + yN −1 = p + yN −1 q 2 p + yN −1 = qyN + yN −1 and the proof follows by induction. q Then 1 ≤ yn ≤ Proof. Let {yn }∞ n=−1 be a solution of Eq(6.39) 2 qyn−1 (yn − p ) + (yn − 1)(yn−1 yn + qyn ) q . 6. (a) Indeed we have yN +1 = and yN +1 = p + yN −1 p + yN −1 ≤ =1 qyN + yN −1 p + yN −1 p q p + yN −1 p = p + yN −1 q p q for all n ≥ N.

then yN +1 > 1. then yN +1 = 1. 1. q (ii) If for some N ≥ 0. (2. q (iii) If for some N ≥ 0. pyn + yn−1 Chapter 6.6. q (iv) If for some N ≥ 0.41) 2 pyn−1 + yn−1 yn + pyn p(p + yn−1 ) + yn (pyn + yn−1 ) (6. yN ≥ p . q Lemma 6. then yN +2 > yN . yN ≤ 1. Then the following statements are true: (i) If for some N ≥ 0. Lemma 6.98 When p=q that is for the difference equation yn+1 = p + yn−1 .6. then yN +2 < yN . (vii) If for some N ≥ 0. then yN +1 > 1. q (vi) If for some N ≥ 0. q (v) If for some N ≥ 0. In particular [1.27). (6. yN ≥ 1. (viii) p 1<y< .42). (6. yN ≤ p .42) The following three lemmas are now direct consequences of (6. yN > p .40) the above identities reduce to the following: yn+1 − 1 = p and yn − yn+2 = (yn − 1) 1 − yn pyn + yn−1 for n ≥ 0 for n ≥ 0. .27). . yN = p . 2)-Type Equations n = 0. then yn ∈ [1. yN < 1.1 Assume that p>q and let {yn }∞ n=−1 be a solution of Eq(6. then yN +1 < 1.2 Assume that p=q ∞ and let {yn }n=−1 be a solution of Eq(6. then yN +1 < p .27). yN < p .37)-(6. . p ] for n ≥ N . . p ] is an q q q invariant interval for Eq(6. Then the following statements are true: (i) If for some N ≥ 0.

In particular [ p . yN ≥ 1. yN > p . then yN +1 > p .6. (iii) If for some N ≥ 0. Then the following statements are true: (i) If for some N ≥ 0. yN ≤ 1. q (ii) If for some N ≥ 0.27). then yN +1 > 1.6. then yN +1 = 1.3 Assume that p<q ∞ and let {yn }n=−1 be a solution of Eq(6. then yN +2 > yN . yN ≤ p . then yN +2 < yN . then yN +1 = 1.1-6. q (iv) If for some N ≥ 0. q Note that the function f (u. yN < 1. then yN > yN +2 > 1. v) = p+v qu + v is always decreasing in u but in v it is decreasing when u < p and increasing when u > p . then yn ∈ [ p . (v) If for some N ≥ 0. yN ≤ p . yN < p .7. yN = p . q (viii) If for some N ≥ 0.2 and Lemmas 6. q (iii) If for some N ≥ 0. yN > 1. yN ≤ 1.6. yN > 1. q (v) If for some N ≥ 0. then yN +1 ≥ 1. 1] is an q q invariant interval for Eq(6. then yN < yN +2 < 1. Lemma 6. then yN +1 < 1. q q The following result is now a consequence of Theorems 1.6. (iv) If for some N ≥ 0. then yN +1 < 1.3: . 1] for n ≥ N .7. The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 99 (ii) If for some N ≥ 0. (ix) p < y < 1.6. q (vi) If for some N ≥ 0.1 and 1.27). yN = 1. (vii) If for some N ≥ 0.

.5 (a) Assume q ≤ 1 + 4p. .6. 1] must converge to the two cycle (6.31) holds.27) either eventually enters the interval I with end points 1 and p and remains there. q−1 (6. or stays forever outside I. .27). ψ.4.4.27) eventually enters and remains in the interval I. and only then. must converge to y.27) is global attractor. if Theorem 6.43) 6. (6. ∞). φ.4 Let {yn }∞ n=−1 be a solution of Eq(6.6 and 1.45) there are no period-two solutions and so the solutions must eventually enter and remain in I and.44)-(6.4 Global Behavior of Solutions Recall that Eq(6. with φ + ψ = 1 and φψ = p . q Now when q ≤ 1 + 4p Then the equilibrium y of Eq(6.6. Hence. while if p>q the length is at most two. in view of Theorems 1. yn ∈ I for n ≥ N.32) holds. any solution which remains forever outside the interval I = [ p .100 Chapter 6. every solution of Eq(6. when (6. When (6.43) holds. But this two cycle lies q inside the interval I. Let I be the closed interval p with end points 1 and q and let J and K be the intervals which are disjoint from I and such that I ∪ J ∪ K = (0.27) has a unique equilibrium y which is locally stable when (6. p<q the length is one.46) Also recall that a solution {yn }∞ n=−1 of Eq(6.44) (6. When (6. (6.32) holds.32) holds. . (2. On the other hand when (6. . or vice-versa. The above observations are summarized in the following theorem: Theorem 6. 2)-Type Equations Then either all the even terms of the solution lie in J and all odd terms lie in K.7.45). Eq(6.27) has a “unique” prime period-two solution . . φ.6. ψ. The character of these solutions remains an open question. or for some N ≥ 0. except for the length of the first semicycle of the solution. .

47). (6. (b) Assume p + q > 1.7. .1 (a) Assume p + q ≤ 1.47) is globally asymptotically stable.47) was investigated in [52].47) β γ and q = .7) which by the change of variables xn = reduces to the difference equation yn+1 = where pyn + qyn−1 . 1. 1 + yn n = 0.47) are the solutions of the equation p= y= py + qy . More precisely it is a saddle point when 1−p<q <1+p and a repeller when q > 1 + p. A A (Eq(6. and when p + q > 1. .6. Then the zero equilibrium of Eq(6.7 The Case α = C = 0 : xn+1 = A yn B βxn +γxn−1 A+Bxn This is the (2.3.47) is unstable. 1]. y = p + q − 1 is the only positive equilibrium point of Eq(6. The following result follows from Theorem 1. q 6. 2)-type equation Eq(6.27) eventually enters and remains in the interval [ p . . 1+y So y = 0 is always an equilibrium point.1 and the linearized stability Theorem 1. . The Case α = C = 0 : xn+1 = (b) Assume βxn +γxn−1 A+Bxn 101 q > 1 + 4p.7.1. Then the zero equilibrium of Eq(6. Then every solution of Eq(6.) The equilibrium points of Eq(6.1: Theorem 6.

The case is treated in Section 6. . ∞) is an invariant interval for Eq(6. remains in this interval. q (0. q (b) Assume that q > 1. 2)-Type Equations 1 − p < q < 1 + p. ψ.47) q with initial conditions in (0.7.47) eventually enters the interval [ p .102 (c) Assume Chapter 6.47) eventually enters the interval (0. . ψ.5 that the following result is true: Theorem 6. . (2. if and only if q = 1 + p. ∞) : ψ = = 2p}.50) Then every solution of Eq(6.47). ∞). every solution of Eq(6.47). φ.48) Furthermore when (6. . Theorem 6. [ p . Concerning prime period-two solutions it follows from Sections 2.47) has a prime period-two solution . Furtherq more. (6. q . (6. p ].47) is an unstable saddle point.2 Eq(6.7. Furthermore.48) holds. p ] is an invariant interval for Eq(6. Then the positive equilibrium y = p + q − 1 of Eq(6. ψ ∈ (p. . That is.3.47) is locally asymptotically stable. Then the positive equilibrium y = p + q − 1 of Eq(6. φ. (d) Assume q > 1 + p. p ]. φ−p 6. (6.47) with q = 1.7.7. the values of φ and ψ of all prime period-two solutions are given by: pφ {φ. .3 (a) Assume that q < 1.51) Then every solution of Eq(6.49) The main result here is the following theorem about Eq(6.1 Invariant Intervals q=1 (6.

∞) → [0.47). Then the following statements are true for every n ≥ 0: p + q > 1 and q ≤ 1.47) when p + q > 1 and q ≤ 1. Then for any k. ∞) × [0.6. Lemma 6. ∞) → [0. ∞) denote the function f (x. Hereafter. (x − x)(f (x. m ∈ N. x) − x) < 0 for x ∈ (0. Let f : [0.2 Here we discuss the behavior of the semicycles of solutions of Eq(6. m ∈ N. ∞) defined by g(x) = (p + q)x .47). Then for any k. y) = px + qy . whose proof is straightforward and will be omitted.7. 1+x Observe that g(y) = y and that g increases on [0.7. ∞).50) holds. . that is.7. 1+x Semicycle Analysis of Solutions When q ≤ 1 Then f satisfies the negative feedback condition. Theorem 6. y refers to the positive equilibrium p + q − 1 of Eq(6.7. The Case α = C = 0 : xn+1 = βxn +γxn−1 A+Bxn 103 The proof of the above theorem is an elementary consequence of the following lemma. Consider the function g : [0.1 (a) Assume that (6. qm q 6. ∞) − {x}.51) holds. m q q (b) Assume that (6. yk ≤ and yk+2 ≥ p p =⇒ yk+2 ≤ p < q q p p =⇒ yk > m+1 > p.4 Suppose Let {yn }∞ n=−1 be a positive solution of Eq(6. yk ≥ and yk+2 ≤ p p =⇒ yk+2 ≥ p > q q p p =⇒ yk > m+1 > p.

(p + q) max{yn−1 . yn . we obtain q yn+1 = pyn + qyn−1 pyn + qy ≥ ≥ g(y) = y. 1 + yn 1 + yn Since the function Observe that all inequalities are strict if we assume that yn−1 > y or yn > y. for every k ≥ −1. yn } pyn + qyn−1 ≥ = min{yn−1 . Then in view of the previous case yn+1 = and pyn + qy pyn + qyn−1 ≤ < g(y) = y 1 + yn 1 + yn (p + q) min{yn−1 . yn . 1 + yn 1 + yn 1 + yn Since q ≤ 1 and yn−1 < y. which is equivalent to q ≤ 1. yn ≤ y. yn } pyn + qyn−1 ≤ = max{yn−1 . yn } ≤ yn+1 < y. yn+1 = (c) Suppose yn−1 < y ≤ yn . yn ≤ y. (2. 1 + yn 1+y Observe that all inequalities are strict if we assume that yn−1 < y or yn < y. 1+x is increasing for y < p . 1 + yn 1+y px + qx . (c) If yn−1 < y ≤ yn . yn }. Thus px+qyn−1 1+x p − qyn−1 > p − qy = (1 − q)(p + q) ≥ 0. Then yn+1 = pyn + qyn−1 (p + q)yn (1 + y)yn < = < yn . (a) Suppose y ≤ yn−1 . yn }. 2−q Chapter 6. (e) If y < yk < Proof. min{yn−1 . (b) Suppose yn−1 . (b) If yn−1 . yn }. then {yn } decreases to the equilibrium y. 2)-Type Equations then then then then y ≤ yn+1 ≤ max{yn−1 . yn−1 < yn+1 < yn . increases in x and so pyn + qyn−1 pyn−1 + qyn−1 > = g(yn−1 ) > yn−1 .104 (a) If y ≤ yn−1 . Then yn+1 = y . yn < yn+1 < yn−1 . (d) If yn < y ≤ yn−1 . 1 + yn 1 + yn−1 yn+1 = The last inequality follows from the negative feedback property. .

Then yn+1 = βxn +γxn−1 A+Bxn 105 pyn + qyn pyn + qyn−1 > = g(yn ) > yn 1 + yn 1 + yn by the negative feedback property. 1 + yn 1 + yn pyn + qyn−1 >y =p+q−1 1 + yn 6.7. Then yn+1 = (e) Observe that yn+1 = and so from which the result follows.6. Here we discuss the behavior of the solutions of Eq(6. Then yn ≤ p for all n > 0 and (6. (6.52) holds. Then {yn }∞ n=−1 oscillates about the equilibrium p with semicycles of length one and (6. To see that yn+1 < yn−1 .47) is globally asymptotically stable. .47) is y = p. Suppose p − qyn−1 ≥ 0. More precisely the following statements are true. The Case α = C = 0 : xn+1 = (d) Suppose yn < y ≤ yn−1 . we must consider two cases: Case 1.52) (b) Assume y−1 ≤ p and y0 ≤ p. Suppose p − qyn−1 < 0. Then yn ≥ p for all n > 0 and n→∞ lim yn = p. 1 + yn 1 + yn−1 The last inequality follows from the negative feedback property.5 Suppose that q = 1 and let y−1 + y0 > 0.3 Semicycle Analysis and Global Attractivity When q = 1 q = 1. Then the equilibrium y of Eq(6. yn−1 ≥ qyn−1 > y + (q − 1)yn > yn 2 pyn + qyn−1 qyn−1 (yn + 1) < ≤ yn−1 .7. Theorem 6. Case 2. Then px+qyn−1 increases in x and so 1+x yn+1 = pyn−1 + qyn−1 pyn + qyn−1 ≤ < yn−1 .47) when Here the positive equilibrium of Eq(6.7. (a) Assume y−1 ≥ p and y0 ≥ p.52) holds. (c) Assume either y−1 < p < y0 or y−1 > p > y0 .

We will assume that y−1 ≥ p and y0 > p. . 1 + y0 1 + y0 Similarly. Furthermore. 1 + y1 1 + y1 It follows by induction that yn ≥ p for all n > 0.106 Chapter 6. we claim that the subsequence {y2k+1 }∞ k=−1 is increasing. Then y1 = and py0 + p py0 + y−1 ≥ = p. we can show that y2 < y0 and by induction y1 = y−1 ≥ y1 ≥ y3 ≥ . y−1 ≥ p implies that y2 = py0 + y−1 ≤ y−1 y0 + y−1 and so y−1 y0 + y−1 py0 + y−1 ≤ = y−1 . (b) The proof of (b) is similar to the proof of (a) and will be omitted. 1 + y0 1 + y0 and py1 + p py1 + y0 > = p. there exist m ≥ p and M ≥ p such that k→∞ lim y2k+1 = m and k→∞ lim y2k = M. since y−1 + py0 > y−1 + y−1 y0 then y1 = py0 + y−1 > y−1 1 + y0 . > p. . The proof when y0 < p < y−1 is similar and will be omitted. y2 = 1 + y1 1 + y1 It follows by induction that the solution oscillates about p with semicycles of length one. . it follows that m = M = p and the proof of part (a) is complete. (c) Assume y−1 < p < y0 . 2)-Type Equations Proof. Next. In fact. ≥ p and y0 > y2 > y4 > . and so convergent. Thus. As Eq(6. py0 + y−1 py0 + p y1 = < = p. The case y−1 > p and y0 ≥ p is similar and will be omitted. 1 + y0 1 + y0 py1 + y0 py1 + p > = p. (2. It is interesting to note that y−1 = p ⇒ y2k+1 = p for k ≥ 0 and y0 = p ⇒ y2k = p for k ≥ 0.47) has no period-two solutions when q = 1. . (a) The case where y−1 = y0 = p is trivial.

2 6.47) is global attractor of all positive solutions of Eq(6. 2 6.7.7.4 Global Attractivity When q < 1 1 − p < q < 1. y0 ∈ [p. every positive solution of Eq(6.7. Thus. Theorem 6. m = M .47) converges to the positive equilibrium y. In a similar way we can show that the subsequence {y2k }∞ is decreasing. Here we consider the case where and we show that the equilibrium point y = p + q − 1 of Eq(6. Now clearly the function q f (x. ∞). that y−1 . ∞).5 Long-term Behavior of Solutions When q > 1 q>1 Here we discuss the behavior of the solutions of Eq(6.47) when which is equivalent to p y> . The Case α = C = 0 : xn+1 = βxn +γxn−1 A+Bxn 107 and the claim follows by induction. Then every positive solution of Eq(6.3 (a) we may assume that the initial conditions y−1 and y0 lie in the interval I = (0. Concerning the long-term behavior of solutions of Eq(6. Proof.6.47) we may assume. q In this case.47).7. y) = px + qy 1+x satisfies the hypotheses of Theorem 1. there exist m ≤ p and M ≥ p such k=0 that lim y2k+1 = m and lim y2k = M k→∞ k→∞ and as in (a).7.4. .5 in the interval I from which the result follows. without loss of generality. By Theorem 6.6 Assume 1 − p < q < 1. p ]. Observe that the consistency condition y ≤ p is equivalent to the condition q q ≤ 1. and so convergent.47) lies eventually in the interval [p.

53) was investigated in Section 6. (b) Assume 1 < q < 1 + p. (2.53) where un ≥ 0 for n ≥ 0. we obtain the following theorems about the solutions of Eq(6.47) to the equation un+1 = Chapter 6. . transforms Eq(6.7 After the first semicycle. Then every solution of Eq(6. Then Eq(6. By applying the results of Section 6. The character of solutions of Eq(6.53).47) when q > 1: Theorem 6.47) oscillates about the positive equilibrium y with semicycles of length one.7. Then every positive solution of Eq(6.47). 2)-Type Equations p(q − 1) + qun−1 1 + p + un (6. an oscillatory solution of Eq(6.47) has unbounded solutions.8 (a) Assume q = 1 + p.7.5.47) converges to the positive equilibrium of Eq(6.5 to Eq(6.108 Then the change of variable yn = un + p. (c) Assume q > 1 + p.47) converges to a period-two solution. Theorem 6.

Furthermore the zero equilibrium is a saddle point when 1−p<q <1+p and a repeller when q < 1 − p.1 (a) Assume p + 1 ≤ q. . . (b) Assume p + 1 > q.54) and q = (Eq(6.1.54) also possesses the unique positive equilibrium y = p + 1 − q.8 The Case α = B = 0 : xn+1 = xn = γ yn C This is the (2. The following result follows from Theorems 1. 1.54) and so we will assume without further mention that (6. A . The Case α = B = 0 : xn+1 = βxn +γxn−1 A+Cxn−1 109 βxn +γxn−1 A+Cxn−1 6.3. the positive equilibrium y = p + 1 − q of Eq(6.6. q + yn−1 β γ n = 0.8. Our goal is to show that when (6.55) holds and y−1 + y0 > 0.8.54) is unstable and the positive equilibrium y = p + 1 − q of Eq(6. γ (6. . 2)-type Eq(6.55) holds.) The equilibrium points of Eq(6.54) is locally asymptotically stable.54) are the solutions of the equation y= py + y . .1.54) is globally asymptotically stable. Then the zero equilibrium of Eq(6. Theorem 6.54) and when p + 1 > q.54) was investigated in [53]. Eq(6.54) is globally asymptotically stable.8) which by the change of variables reduces to the difference equation yn+1 = where p= pyn + yn−1 .55) Then the zero equilibrium of Eq(6. In the remainder of this section we will investigate the character of the positive equilibrium of Eq(6. q+y So y = 0 is always an equilibrium point of Eq(6. (6.1 and 1.

(2. p + yn−1 n = 0.62) and let {yn }∞ n=−1 be a positive solution of Eq(6.1 Invariant Intervals and Semicycle Analysis Let {yn }∞ n=−1 be a positive solution of Eq(6.8.54).58) reduce to the following: yn+1 − 1 = (yn − 1) and yn+1 − yn = (1 − yn ) p p + yn−1 yn−1 p + yn−1 for n ≥ 0 for n ≥ 0.61) The following three lemmas are now direct consequences of the above identities. . 1. for n ≥ 0.58) Note that the positive equilibrium y =p+1−q of Eq(6. (6. Lemma 6. When p=q that is for the difference equation yn+1 = pyn + yn−1 .1 Assume that p<q (6. (6.8.54) is such that: < 1 if p < q < p + 1 y is  = 1 if p = q  > 1 if p > q.54). (6. 2)-Type Equations 6. .60) (6.57) and yn+1 − yn = (p − q)yn + (1 − yn )yn−1 q + yn−1 (6. Then the following identities are easily established: q yn − p yn+1 − 1 = p for n ≥ 0.56) q + yn−1 q p2 [yn − p ] + (p − q)yn−1 q yn+1 − = p p(q + yn−1 ) 2 for n ≥ 0. Then the following statements are true: .110 Chapter 6. .59)    the identities (6. (6.56)-(6.

(vi) If for some N ≥ 0. then yN +1 > q p and yN +2 > 1. (v) If for some N ≥ 0. yN ≤ 1. (v) If for some N ≥ 0. yN ≥ 1.2 Assume that p=q (6. q (ii) If for some N ≥ 0. yN > p . then yN +1 > 1.8. p q then yN +1 < p . yN ≤ 1. yN > q 2 . then yN +1 < 1. yN = p . then yN +1 = 1.54). yN ≥ 1. p q then yN +1 > p .8.63) and let {yn }∞ n=−1 be a positive solution of Eq(6.3 Assume that p>q (6. then yN +1 > 1. (ii) If y0 = 1. Lemma 6. then yn = 1 for n ≥ 0. The Case α = B = 0 : xn+1 = βxn +γxn−1 A+Cxn−1 111 q (i) If for some N ≥ 0. then yN +1 < 1. then yN +1 < yN . yn > 1 and the solution is strictly decreasing. Then the following statements are true: q (i) If for some N ≥ 0. then yN +1 < 1. yN < p . q (iii) If for some N ≥ 0. yN > p . then yN +1 = 1. q (iii) If for some N ≥ 0. (vi) If for some N ≥ 0. . yn < 1 and the solution is strictly increasing. Then the following statements are true: (i) If y0 < 1. then yN +1 > yN .64) and let {yn }∞ n=−1 be a positive solution of Eq(6. q (ii) If for some N ≥ 0. (iv) If for some N ≥ 0.54). then for n ≥ 0. (iii) If y0 > 1. yN < p . Lemma 6. then for n ≥ 0. yN > q 2 . yN = p .6. (iv) If for some N ≥ 0.8.

(2. assume that (6.8.8.1 that if a solution {yn }∞ n=−1 is such that yn ≥ 1 for all n ≥ 0 then the solution decreases and its limit lies in the interval [1.54) is globally asymptotically stable.54) eventually enters and remains in the interval (1. Then the positive equilibrium y = p + 1 − q of Eq(6. ∞).2 that every solution of Eq(6. When (6. ∞). 1) the function f (u. 2)-Type Equations 6.8. 1). q+1 From the above observations and in view of Theorem 6. This (2. it follows from Lemma 6. by Lemma 6. every positive solution of Eq(6.112 Chapter 6. . This is impossible because y < 1. (6.4. Here it is clear from Lemma 6. n = 0.54) eventually enters and remains in the interval (0.3 that every solution of Eq(6. 1. 2)-type equation was investigated in Section 6.54) converges to the equilibrium y = 1.1.65) with positive parameters and positive initial conditions.1 we have the following result: Theorem 6. . Without loss of generality we will assume that yn > 1 for n ≥ −1.2 Global Stability of the Positive Equilibrium When (6. Now in the interval (0. y is a global attractor.3 where we established in Theorem 6. it follows from Lemma 6.62) holds. Next.8.8. . v) = pu + v q+v is increasing in both arguments and by Theorem 1.8. Set yn = 1 + (q + 1)un for n ≥ −1.8.3.2 Assume that p + 1 > q and that y−1 + y0 > 0.63) holds. Hence.8.64) holds.3 that every solution converges to its positive equilibrium u= p−q . Then we can see that {un }∞ n=−1 satisfies the difference equation un+1 = p−q (q+1)2 + p u q+1 n 1 + un−1 . .

66) was investigated in [54].66) is locally asymptotically stable when q < pq + 1 + 3p and is unstable (and more precisely a saddle point equilibrium) when q > pq + 1 + 3p. (6.9 The Case α = A = 0 : xn+1 = γ yn C This is the (2. 2)-type Eq(6. . 1.1.) The only equilibrium point of Eq(6. (p + 1)(q + 1) (p + 1)(q + 1) n = 0.68) (6.66) is locally asymptotically stable. (Eq(6.66) about y is zn+1 − p−q p−q zn + zn−1 = 0.66) B β and q = .6. . . . qyn + yn−1 n = 0.9) which by the change of variables xn = reduces to the equation yn+1 = where pyn + yn−1 . Then the positive equilibrium of Eq(6. 1. . . (b) Assume that p < q.67) . Then the positive equilibrium of Eq(6.1 we obtain the following result: Theorem 6.66) is y= p+1 q+1 and the linearized equation of Eq(6.9. (6. The Case α = A = 0 : xn+1 = βxn +γxn−1 Bxn +Cxn−1 113 βxn +γxn−1 Bxn +Cxn−1 6. By applying the linearized stability Theorem 1.1 (a) Assume that p > q.9. . γ C To avoid a degenerate situation we will also assume that p= p = q.

ψ. One can see that T2 where pv + u g(u. .  . v) = = . . ∞) × (0. (6. ψ. . Eq(6.66) has no prime period-two solutions.70) t2 − (1 − p)t + = 0. evaluated at lie inside the unit disk. is a fixed point of T 2 . Eq(6. . ∞) defined by: T Then u v = φ ψ u v g(u.5 that when p > q. .9. . n = 0. 1. 2)-Type Equations 6. .69) is asymptotically stable if the eigenvalues of the φ Jacobian matrix JT 2 . v) = qv + u and h(u. 1. . and write Eq(6. φ. . v) p pv+u + v qv+u q pv+u + v qv+u v pv+u qv+u . . the second iterate of T . One can see that ψ JT 2 φ ψ =    (p−q)ψ − (qψ+φ)2 (p−q) ψ − (qψ+φ)2 (qφ+ψ)2 2 2 (p−q)φ (qψ+φ)2 (p−q)φ (qφ+ψ)2 (p−q)ψ (qψ+φ)2 −1   . . qv Let T be the function on (0.66) possesses a unique prime period-two solution: .1 Existence of a Two Cycle It follows from Section 2.114 Chapter 6. (2.69) where the values of φ and ψ are the (positive and distinct) solutions of the quadratic equation p(1 − p) (6. we set un = yn−1 and vn = yn for n = 0. φ. On the other hand when p<q and q > pq + 1 + 3p. q−1 In order to investigate the stability nature of this prime period-two solution. The prime period-two solution (6. v) h(u.66) in the equivalent form: un+1 = vn n +un vn+1 = pvn +un . . .

72) φ ψ and ψ = Inequality (6.71) (6.72) is always true. Next we will establish Inequality (6.71).6. P < 0 and Q < 0 and so Inequality (6.1. it follows that both eigenvalues of JT 2 lie inside the unit disk if and only if |P ψ + Qφ − P Qφψ| < 1 + P Qφψ < 2 or equivalently if and only if the following three inequalities hold: P ψ + Qφ + 1 > 0 P ψ + Qφ < 1 + 2P Qφψ and P Qφψ < 1. Observe that φ > 0.9. qφ + ψ (6. ψ > 0.71) is equivalent to (p − q)φ (p − q)ψ + +1>0 (qψ + φ)2 (qφ + ψ)2 which is true if and only if (q − p)[(qφ + ψ)2 ψ + (qψ + φ)2 φ] < (qψ + φ)2 (qφ + ψ)2 . φ= pψ + φ qψ + φ φψ = p(1 − p) q−1 pφ + ψ . The Case α = A = 0 : xn+1 = Set P = Then JT 2 βxn +γxn−1 Bxn +Cxn−1 115 p−q . We will use the fact that φ + ψ = 1 − p. (qφ + ψ)2 p−q (qψ + φ)2 φ ψ = and Q = and its characteristic equation is Pφ −P ψ −P Qψ 2 P Qφψ − Qφ λ2 + (P ψ + Qφ − P Qφψ)λ + P Qφψ = 0.1 (c).73) (6. By applying Theorem 1.

74) is I = (q − p)[(qp + 1)(φ2 + ψ 2 ) + 2(p + q)φψ] = (q − p)[(qp + 1)(φ + ψ)2 − 2φψ(qp + 1 − p − q)] = (q − p)[(qp + 1)(1 − p)2 − 2 The righthand side of (6.116 if and only if Chapter 6. (2. II = [(qψ + φ)(qφ + ψ)]2 = [(q 2 + 1)ψφ + q(φ2 + ψ 2 )]2 = [q(1 − p)2 + (q − 1)p(1 − p)]2 Hence. Now observe that the lefthand side of (6.73) is equivalent to (p − q)2 φψ < (qψ + φ)2 (qφ + ψ)2 which is true if and only if (q − p) φψ < (qψ + φ)(qφ + ψ) if and only if if and only if if and only if if and only if (q − p) φψ < (q 2 + 1)φψ + q(φ2 + ψ 2 ) (q − p) φψ < (q 2 + 1)φψ + q[(φ + ψ)2 − 2φψ] (q − p) φψ < (q − 1)2 φψ + q(φ + ψ)2 (q − p) φψ < (q − 1)2 if and only if p(1 − p) + q(1 − p)2 q−1 = (1 − p)2 (q − p)2 . Inequality (6. = [q(φ + ψ)2 + (q − 1)2 φψ]2 (q − p) φψ < (1 − p)(q − p) .74) = (q − p)(1 − p)2 (qp + 1 + 2p).74) is p(1 − p) (p − 1)(q − 1)] q−1 (6. Inequality (6. Finally.68) holds.71) is true if and only if (6. 2)-Type Equations (q − p)[(qφ + ψ)(pφ + ψ) + (qψ + φ)(pψ + φ)] < [(qφ + ψ)(qψ + φ)]2 .

2 Assume that Condition (6.9. .66): Theorem 6. (b) Assume p < q.7. In summary.1 by observing that when p < q. (b) The proof follows from Theorem 1. 6. Then {yn } oscillates about the equilibrium y with semicycles of length two or three.6.9. . . The Case α = A = 0 : xn+1 = if and only if βxn +γxn−1 Bxn +Cxn−1 117 p(1 − p) < (1 − p)2 q−1 if and only if q > pq + 1 which is clearly true. .66). where φ and ψ are the two positive and distinct roots of the quadratic Eq(6.66). y) is increasing in y and decreasing in x.7.3 Let {yn } be a nontrivial solution of Eq(6.9. Then either {yn } oscillates about the equilibrium y with semicycles of length one.9.68) holds. (a) The proof follows from Theorem 1. the function 2 f (x. we present a semicycle analysis of the solutions of Eq(6. Proof. Then the positive equilibrium y of Eq(6. Theorem 6.66) is globally asymptotically stable.35). 6. ψ. ψ.9. φ.4 by observing that the condition p > q implies that the function px + y f (x. . The extreme in each semicycle occurs in the first term if the semicycle has two terms and in the second term if the semicycle has three terms.70). . the following result is true about the local stability of the prime periodtwo solution (6. or {yn } converges monotonically to y. This function also satisfies Condition (1.4 Assume that p<q and (6. Furthermore this prime period-two solution is locally asymptotically stable. φ. Then Eq(6. y) = qx + y is increasing in x and decreasing in y.3 Global Stability Analysis When p < q The main result in this section is the following Theorem 6. after the first semicycle.2 Semicycle Analysis In this section.9.69) of Eq(6. except possibly for the first semicycle which may have length one. . Then the following statements are true: (a) Assume p > q.66) posseses the prime period-two solution .67) holds.

67). ψ. φ.69) when instead of (6. φ) = φ and in general yN +2k ≥ ψ and yN +2k+1 ≤ φ for k = 0. yN +1 ) ≤ f (ψ. p ≤ f (x. y > 0. . Now the conclusion of Theorem 6. 2 Next we want to find more cases where the conclusion of the last theorem holds. ψ) = ψ. Theorem 6. yN +2 = f (yN +1 . Also clearly. . Let φ.4 follows as a consequence of Theorem 1.68) holds. y) = px + y . (2. yN ) ≥ f (φ. . ψ.4.68) holds.118 Proof.4.67). φ.6 lim sup yn = ψ n→∞ px + y .6 and the fact that y is locally asymptotically stable.4. Assume that for some solution {yn }∞ n=−1 of Eq(6. .66). 2 The method employed in the proof of Theorem 1. y) = Then clearly.66) has no prime period-two solution.75) yN ≥ ψ and yN +1 ≤ φ. We consider first the case where eventually consecutive terms yi . . Proof. 2)-Type Equations and note that f (x. Assume that (6.5 Assume that (6.. . q Finally in view of (6. yN +3 = f (yN +2 .9.66) and for some index N ≥ −1. . . qx + y from which we conclude that limk→∞ yN +2k = ψ and limk→∞ yN +2k+1 = φ. ψ. .66) converge to the two cycle (6.6 can also be used to establish that certain solutions of Eq(6. ψ. y) ≤ 1 for all x.9. and increasing in y for each fixed x. yi+1 lie between m and M . denote the two cycle of Eq(6.75) holds. y) is decreasing in x for each fixed y. and lim inf yn = φ n→∞ Now as in the proof of Theorem 1. (6. Set f (x.. with φ < ψ. qx + y Chapter 6. 1. Eq(6. (6. Then this solution converges to the two cycle φ. Set f (x. . .

9. By our earlier observation we have that M ≥ lim supi→∞ yi . qM − p = m and the proof is complete. either yi > M and yi+1 < m or yi < m and yi+1 > M ). . and so M ≥ lim supn→∞ y2n ≥ lim inf n→∞ y2n ≥ M. M ] for every k > i. φ.6. Since the proofs are similar for all cases we will give the details in case (A).9. .9. .1 Suppose that for some i ≥ 0. (C) y2n < m and m ≤ y2n+1 < M for every n. Then yk ∈ [m. then we have one of the following four cases: (A) y2n > M and M ≥ y2n+1 > m for every n. (B) y2n+1 > M and M ≥ y2n > m for every n. The Case α = A = 0 : xn+1 = βxn +γxn−1 Bxn +Cxn−1 119 Lemma 6.2 In all cases (A)-(D) the corresponding solution {yn } converges to the two cycle . ψ. m ≤ yi . we are left with the case in which the solution lies eventually in [m.66) we have y2n+1 = and so y2n − y2n y2n+2 . (D) y2n+1 < m and m ≤ y2n < M for every n. limn→∞ y2n+1 M − M2 . M ] is answered by Theorem 6. yi+1 ≤ M. .5. Proof. M ) = pm + M qm + M for m we find Thus. m) ≤ f (yi+1 . qy2n+2 − p M − M2 lim y2n+1 = . Using the monotonic character of the function f we have m = f (M. and the result follows by induction. . M ] and if we never get two successive terms outside of the interval [m. 2 If we never get two successive terms in the interval [m. From Eq(6. Hence limn→∞ y2n = M . yi ) = yi+2 ≤ f (m. ψ. . M ) = M.9. M ]. M ] (that is. φ. Proof. . m= 2 Since the case in which the solution lies outside of the interval [m. n→∞ qM − p On the other hand by solving the equation M = f (m. Lemma 6. for every i.

. . . ψ. The correspondence between y and y has the following properties: Lemma 6. 2)-Type Equations Lemma 6. y) > y if and only if x < y . Thus f (y . f (y. .3 Suppose M ≥ yi+2 ≥ yi ≥ y ≥ yi+1 ≥ yi+3 ≥ m and not both yi+2 and yi+3 are equal to y. Here we have yi+4 = f (yi+3 . in which case Eq(6. as the following results show. but this situation cannot occur.9. Thus. y) = y. yi+1 ) = yi+3 .4 (1) f (x. (3) For y < y < M. ψ. so we are led to the general equation f (x. Therefore. both subsequences converge and their limits form a two cycle. In view of the uniqueness of the two cycle and the fact that at least one of yi .The case M ≥ yi+3 ≥ yi+1 ≥ y ≥ yi ≥ yi+2 ≥ m is handled in a similar way.9. which of course depends on y. 2 Using a similar technique one can prove that if M ≥ yi ≥ yi+2 ≥ y ≥ yi+3 ≥ yi+1 ≥ m or M ≥ yi+1 ≥ yi+3 ≥ y ≥ yi+2 ≥ yi ≥ m we would get convergence to y .76) has a q unique positive solution for x. φ. φ.120 Chapter 6. We denote this value by y . The second case can be handled in a similar way. and for m < y < y. . y) = y. y ) < y and y > y . yi+1 is different from y. In this case we need to know more about the condition yi+2 = f (yi+1 . (2. (2) x > y if and only if x < y . f (y. . yi ) ≤ yi .76) It makes sense to consider this equation only for y ∈ ( p . yi ) = yi+2 and yi+5 = f (yi+4 . the proof is complete. (6. Then {yn } converges to the two cycle . yi+2 ) ≥ f (yi+1 . Proof. Proof. yi+3 ) ≤ f (yi+2 . Let us consider the first case. and the sequence {yi+2k+1 } is a non increasing sequence and bounded below by m. 1). Induction on i then establishes the monotonicity of the two sequences. y ) > y and y > y. the sequence {yi+2k } is non decreasing and bounded above by M . .

2 Now we will use this result to prove the following: Lemma 6. for m < y < y.5 Assume that for some i. 1). Thus for y < m. y) and y = f (y. (2) Solving f (y . Hence q m < y < y ⇒ f (y. y ) we get px + x − x2 = qx − p qx + x−x2 qx−p x−x2 qx−p . For y ∈ ( p .9. so y is decreasing function q of y. y < y < M ⇒ f (y. m. M < y <⇒ f (y. x < y if and only if f (x. M ≥ yi ≥ yi+2 ≥ y ≥ yi+1 ≥ m. However q p < f (y. so the roots of the numerator are not real. y ) < y .9. y. y) = y for y gives y = g(y) = Computing the derivative of g we get g (y) = − p − 2y + qy 2 .6. By (1) then. Clearly as y approaches p y approaches ∞. y ) > y . y) = y. The Case α = A = 0 : xn+1 = βxn +γxn−1 Bxn +Cxn−1 121 (1) Since f is decreasing in the first variable and f (y . y < y . f (y. and for y < y < M. y ) > y . y ) < 1. y ) < y . Thus each of these roots is simple and so f (y. y>y . Simplifying this equation we get a fourth degree polynomial equation with roots 0. (p − qy)2 y − y2 . qy − p The discriminant of the numerator is 4p2 − 4pq = 4p(p − q) < 0. Then yi+3 ≤ yi+1 with equality only in the case of a two cycle. . M . y ) − y changes sign as y passes each of these values. y) = y. y) > f (y . g (y) < 0. (3) Eliminating y from the equations y = f (y .

we have yi+1 ≥ yi . so either c = y or c = m. The value c will be one value in a two cycle. . ≥ yi+3 ≥ yi+1 ≥ . . and solving for yi+2n we see yi+2n = y also. If c = y.9. there is no i such that either M ≥ yi ≥ yi+2 ≥ . then yi+2n+1 = y for all n. To show yi+3 = f (yi+2 . and our solution is a two cycle. . To get equality we must have yi+1 = yi or equivalently yi+2 = yi . or M ≥ . . Thus yi ≥ yi+1 . This can only happen if yi+2n = M 2 for all n. and that Eq(6. we must show yi+2 ≥ yi+1 . as well as 2 yi+3 = yi+1 . ≥ y ≥ yi+1 ≥ yi+3 ≥ . it follows by Theorem 6. We also obtain the following lemma: Lemma 6. yi+1 ). (2.66) are generated by initial values {y−1 . in which case yi+2n+1 = m for all n. By (3) of the previous lemma. yi+1 ) ≥ yi+1 . Since {yi+2k+1 } is a bounded monotonic sequence it has a limit c. If c = m. . . . The case m ≤ yi ≤ yi+2 ≤ y ≤ yi+1 ≤ M is similar. Proof. ≥ yi+2 ≥ yi ≥ y ≥ . yi+1 ) ≤ yi+1 .122 Chapter 6. But then yi = yi+1 = y.9. which also gives a two cycle. The above sequence of lemmas leads to the following result: Theorem 6. solving for limn→∞ yi+2n gives limn→∞ yi+2n = M . ≥ m..69).66) converges to this two cycle. Thus yi+2 ≥ yi+1 .6 that all such pairs belong to the stable manifold of the two cycle. Remark. yi ) ≥ f (yi+1 . f (yi+1 . ≥ m. Thus equality occurs only in the case of a two cycle.9. . . Since some of the oscillatory solutions of Eq(6.66) possesses the two-cycle solution (6. 2)-Type Equations Proof. yi ) ≤ yi . .6 Unless {yn } is a period-two solution. y0 } that are on opposite sides of the equilibrium y. since m ≤ yi+1 ≤ y. the second case is similar. Since yi+2 = f (yi+1 . Then every oscillatory solution of Eq(6.6 Assume that p < q. Since f is increasing in the second argument and yi ≥ yi+1 we get yi+2 = f (yi+1 . . and by (3) yi ≥ yi . We will consider the first case. Thus by (2) of the previous lemma yi ≥ yi+1 . as required.

.7 Assume that p>q and p ≤ pq + 1 + 3q. set f (x. y) is increasing in x for each fixed y. d ∈ R. y > 0.1 (See [64]) Assume that (a) Investigate the forbidden set F of the difference equation xn+1 = axn + bxn−1 xn .10. Open Problems and Conjectures 123 6.77) and observe that when p > q. . cxn + dxn−1 n = 0.10 Open Problems and Conjectures a. qm + M 6. Finally observe that when (6. qM + m m= pm + M . . We will employ Theorem 1.4. y) ≤ p q for all x.77) holds. x0 ) ∈ F.10.9.4 Global Stability Analysis When p > q Here we have the following result: Theorem 6.4. the only solution of the system M= is m = M.5.5. investigate the asymptotic behavior and the periodic ∞ nature of the solution {xn }n=−1 . Also 1 ≤ f (x. qx + y (6. and is decreasing in y for each fixed x. Now the result is a consequence of Theorem 1. . y) = px + y . 2 pM + m . Open Problem 6. 1. b. c. . the function f (x.6. To this end. Then the positive equilibrium y of Eq(6. Proof.9. / (b) For every point (x−1 .66) is globally asymptotically stable.

4.10. .2 Assume that αn . q ∈ (0.5.) Chapter 6.3.6 for the autonomous case. 1 + yn n = 0. has a positive and monotonically decreasing solution. we need only to confirm this conjecture for p > 2(q +1) and q ≥ 1.1 Assume p.4. Show that every positive solution of Eq(6. Investigate the forbidden set and the asymptotic character of solutions of the nonautonomous Riccati equation xn+1 = αn + βn xn . Show that the equation yn+1 = p + yn−1 . . are convergent sequences of real numbers with finite limits. (2. every positive solution converges to a. period-two solution. 1.) Conjecture 6.10. not necessarily prime. this conjecture has been confirmed for q ≤ 1 + 4p. (Note that by Theorem 6.) Conjecture 6.2.10. (Note that by Theorem 6. An . βn . 1. (See Section 1. .) . Bn ∈ R for n = 0. 1. An + Bn xn n = 0. Show that every positive solution of Eq(6.124 (See Section 1. .10. ∞). 2)-Type Equations Open Problem 6.2 Assume that p.3 Assume p ∈ (0.11) has a finite limit. . . See also [36]. (Note that by Theorem 6.10) has a finite limit. .6. ∞).3.) Conjecture 6. q ∈ (0. . ∞). . .

10. Show that every positive solution of Eq(6.10. (See Section 6.27). Conjecture 6.9.6. Find the basin of attraction of the period-two solution of Eq(6.) Open Problem 6. y0 ) ∈ R × R through which the solution of the equation yn+1 = yn + yn−1 1 + yn is well defined and converges to 1.66) has a finite limit.6. Find the basin of attraction of the period-two solution of Eq(6. Open Problems and Conjectures Conjecture 6.66). q ∈ (0. 125 Show that every positive solution of Eq(6.3 Assume that p.4 Assume that p. ∞) and q > pq + 1 + 3p.7.) Open Problem 6.5 Find the set of all initial conditions (y−1 . as n → ∞.4 Assume that p. this conjecture has been confirmed for p ≤ pq + 1 + 3q. q ∈ (0.27) either converges to a finite limit or to a two cycle. ∞) and q > 1 + 4p.10. .10.10.10. q ∈ (0. Open Problem 6. ∞). (Note that by Theorem 6.5 Assume p > q.

investigate the long-term behavior of the solution {yn }∞ : n=−1 yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = 1 − yn 1 − yn−1 yn − 1 yn − yn−1 1 − yn−1 1 − yn (6.84) yn − yn−1 For those equations from the above list for which you were successful. 1. (2.10. . y0 ) ∈ R × R through which the equation is well defined for all n ≥ 0.83) yn − yn−1 yn − 1 yn − yn−1 1 − yn−1 1 − yn−1 yn − yn−1 yn + yn−1 . . as n → ∞. Open Problem 6. ∞). .126 Chapter 6. A. . Open Problem 6. n = 0. β. extend your result by introducing arbitrary real parameters in the equation.81) (6. xn+1 = Axn + Bxn−1 + Cxn−2 Extend and generalize. (6.80) (6.Then for every (y−1 . y0 ) ∈ R × R through which the solution of the equation yn+1 = yn + yn−1 1 + yn−1 is well defined and converges to 1.82) (6.6 Find the set of all initial conditions (y−1 .10. γ.79) (6.8 Assume α. 2)-Type Equations Open Problem 6. B. y0 ) ∈ G.10.7 For each of the following difference equations determine the “good” set G ⊂ R × R of initial conditions (y−1 . Investigate the asymptotic behavior and the periodic nature of solutions of the difference equation αxn + βxn−1 + γxn−2 .78) (6. C ∈ (0. .

1.6. qn + yn−1 pn yn + yn−1 . p = n→∞ pn lim lim and q = n→∞ qn . .11 Assume q ∈ (1.10.9 Assume that {pn }n=0 and {qn }∞ are convergent sequences of n=0 nonnegative real numbers with finite limits. 1. . (a) Find the set B of all initial conditions (y−1 . (6. .90) (6. Open Problem 6.91) (6. . (b) Let (y−1 . . . ∞) such that the solutions {yn }∞ n=−1 of Eq(6. .10. 1 + yn−1 yn + pn . . 1 + yn pn yn + yn−1 . y0 ) ∈ (0.85)-(6. . 1. ∞). n = 0.85) (6. . .10 Assume that {pn }∞ and {qn }∞ are period-two sequences n=0 n=0 of nonnegative real numbers. qn yn + yn−1 pn yn + qn yn−1 . n = 0. . . y0 ) ∈ B. 1 + yn pn + yn−1 .86) (6. Open Problems and Conjectures 127 ∞ Open Problem 6. . 1. 1. . qn + yn n = 0. n = 0. yn + qn yn−1 pn + qn yn−1 . n=−1 . . Extend and generalize. . .87) (6. . 1. Investigate the asymptotic behavior of {yn }∞ .23) are bounded. ∞) × (0. n = 0. qn yn + yn−1 yn+1 = yn+1 = Open Problem 6. n = 0. .10.92) pn + qn yn . Investigate the asymptotic behavior and the periodic nature of all positive solutions of each of the following eight difference equations: yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = pn + yn . . 1. .88) (6.92). . 1. Investigate the global character of all positive solutions of Eqs(6.89) (6. . n = 0. n = 0. .10.

are nonnegative. 3).) Let A ∈ (0.10. Investigate the asymptotic behavior of {yn }n=−1 . ∞) through which the solutions {xn }∞ n=−1 of the difference equation xn+1 = (1 − xn − xn−1 )(1 − e−Axn ). ∞) × (0. [3].93) is globally asymptotically stable.47) are bounded.10.14 (Discrete Epidemic Models. (a) Determine the set G of initial conditions (y−1 .6 Assume x0 . . . (6. γ Obtain conditions for the global asymptotic stability of the positive equilibrium of the system xn+1 = αxn βxn +eyn yn+1 = γ(xn + 1)yn      . ∞) and that α ∈ (2. β ∈ (0. and γ ∈ (0. n = 0. (c) Extend and generalize.13 (A Plant-Herbivore System. 2)-Type Equations (a) Find the set B of all initial conditions (y−1 .128 Open Problem 6. x0 ) ∈ G. ∞) such that the solutions {yn }∞ n=−1 of Eq(6. 2 Show that the positive equilibrium of System (6. y0 ∈ (0. the periodic nature. ∞) × (0. Open Problem 6. . ∞). Investigate the boundedness character.94) . and the asymptotic behavior of the solution {xn }∞ n=−1 of Eq(6. y0 ) ∈ B. and [2]) Assume α ∈ (1.10. ∞). . ∞ (b) Let (y−1 . (b) Let (x−1 . 1 β = 1 and γ = .12 Assume q ∈ (1. Chapter 6. y0 ) ∈ (0. See [14]. ∞).94). . (6. . (2. n = 0. See [9].93) Conjecture 6.10. Open Problem 6. y0 ) ∈ (0. ∞). 1) with α + β > 1 + β . . 1. 1.

yn + q pyn + yn−k . 1. 3. Open Problem 6. n = 0. . (6. 1. n = 0. . . ∞) and k ∈ {2. (6. 1.    Investigate the global character of all positive solutions of the system . . ∞). . . ∞). 1). with positive initial conditions. . . .   Open Problem 6. q + yn−k pyn + yn−k . 1 + yn p + yn−k . . .98) (6. ∞) with c1 + c2 > 0. . . . . 1.10. .95) 1 + yn−k yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn + p . . .15 (The Flour Beetle Model. 1. c2 ∈ [0. n = 0.10. . . and c1 .96) (6. Assume a ∈ (0. .10. . . 129 Obtain necessary and sufficient conditions for the global asymptotic stability of the Flour Beetle Model: xn+1 = axn + bxn−2 e−c1 xn −c2 xn−2 . Open Problems and Conjectures Open Problem 6. . .101) . Investigate the global behavior of all positive solutions of each of the following difference equations: p + qyn yn+1 = . b ∈ (0. n = 0. See [48]). . . . n = 0. . .100) (6.}. qyn + yn−k n = 0. 1. 1. n = 0. n = 0. 1) and xn+1 = αxn e−yn + β yn+1 = αxn (1 − e −yn β ∈ (1. 1. ) which may be viewed as a population model.99) (6. 1.16 (A Population Model) Assume α ∈ (0. qyn + yn−k yn + pyn−k . n = 0.10. yn + qyn−k p + qyn−k . q ∈ [0. . .6.17 Assume that p.97) (6.

18 Obtain a general result for an equation of the form yn+1 = f (yn . . 2)-Type Equations Open Problem 6. yn−1 ). . 1. . (6. n = 0. xn−1 ). Open Problem 6. xn−1 ). .10. 1. also converges to a period-two solution. . Obtain necessary and sufficient conditions on f so that the same two cycle is a locally asymptotically stable solution of Eq(6.102) has a two cycle which is locally asymptotically stable. n = 0. which extends and unifies the global asymptotic stability results for Eqs(6.103) Open Problem 6. . . .19 Assume that every positive solution of an equation of the form xn+1 = f (xn .10.11) and (6.130 Chapter 6.102) converges to a period-two solution. (2. Extend and generalize. .20 Assume that Eq(6. Obtain necessary and sufficient conditions on f so that every positive solution of the equation xn+1 = f (xn−2 .103).27). n = 0.10. 1. (6. .

. (7. A + Bxn + Cxn−1 n = 0. . 1 + pyn + qyn−1 αB A2 and q = 131 n = 0. 3)-Type Equations 7.1) which by the change of variables xn = reduces to the difference equation yn+1 = where p= 1 . . . .1) (7. . (7. 3)-type: Please recall our classification convention in which all the parameters in the above (1.Chapter 7 (1.3) α . . (7. . 7.(7. . 1. . 3)-type equations are positive and that the initial conditions are nonnegative. . 1.4) αC . A + Bxn + Cxn−1 n = 0. n = 0. 1. 1. A + Bxn + Cxn−1 βxn . 3)-type Eq. A2 . .1 Introduction xn+1 = xn+1 = and xn+1 = γxn−1 .2) Eq(1) contains the following three equations of the (1.2 The Case β = γ = 0 : xn+1 = α yn A α A+Bxn +Cxn−1 This is the (1. .

3.3 The Case α = γ = 0 : xn+1 = A yn C βxn A+Bxn +Cxn−1 This is the (1. (7. y0 ∈ (0.5 in the interval [0. 1]. q+1 The following result is a straightforward consequence of Theorem 1. 3)-Type Equations One can easily see that the positive equilibrium of Eq(7. Then the zero equilibrium of Eq(7. .4.4. 1. .2 Assume that y−1 . A C Eq(7.1. Theorem 7. . . or Theorem 1.7 in the interval [0. ∞) and p > 1.2.1 The positive equilibrium of Eq(7.5) β B and q = .4. 3)-type Eq(7. 1 + pyn + qyn−1 p= n = 0.3. It also follows by applying Theorem 1. 7. The following result is now a straightforward consequence of either the stability trichotomy Theorem 1.2. Then the positive equilibrium of Eq(7.1 Assume p ≤ 1.4) has no prime period two solutions. Theorem 7.132 Chapter 7. p ]. The following result is a straightforward consequence of Theorem 1.4. (1.4) is locally asymptotically stable for all values of the parameters and that Eq(7.4) is globally asymptotically stable.3.5) always has zero as an equilibrium point and when p>1 it also has the unique positive equilibrium point y= p−1 .4. q Theorem 7.5) is globally asymptotically stable.2) which by the change of variables xn = reduces to the difference equation yn+1 = where yn .5) is globally asymptotically stable.

4.7. . .1. Then the positive equilibrium of Eq(7. . B . Concerning prime period-two solutions one can see that Eq(7. . ψ.6) always has the zero equilibrium and when p<1 it also has the unique positive equilibrium y= 1−p . .6) and q = Eq(7. .6) has a prime periodtwo solution . 3)-type Eq(7.3. .4. 1+q The subsequent result is a straightforward application of Theorems 1. φ. .1 (a) The zero equilibrium of Eq(7.4 The Case α = β = 0 : xn+1 = γ yn C This is the (1. p + qyn + yn−1 A γ n = 0.1.6) is locally asymptotically stable when q<1 and is unstable (a saddle point) when q > 1.3) which by the change of variables xn = reduces to the equation yn+1 = where p= yn−1 . (b) Assume that p < 1. The Case α = β = 0 : xn+1 = γxn−1 A+Bxn +Cxn−1 133 γxn−1 A+Bxn +Cxn−1 7. φ. ψ. C (7. Theorem 7. if and only if p < 1. 1. .1 and 1.6) is globally asymptotically stable when p≥1 and is unstable (a repeller) when p < 1. .

134 Furthermore when Chapter 7.8) 0  p  0 lie in the disk |λ| < 1.6) is locally asymptotically stable.5.7) (7. 1 − p − φ.6) are given by .7) both eigenvalues of JT 2 7. On the other hand. (7. .5 Open Problems and Conjectures p.8) is locally asymptotically stable.6 that here JT 2 0 1−p =  1 p+q(1−p) q(1−p) − p+q(1−p) 1−p . . 1 − p. 0. with 0 ≤ φ ≤ 1 − p and φ = Theorem 7. . . . . (1. Conjecture 7. when p < 1 and q = 1 all prime period-two solutions of Eq(7. . .2 Assume that p<1 Then the period-two solution . . 2 and in view of (7.6) has a finite limit. 0. 2 and q > 1. .1 Assume Show that every positive solution of Eq(7. 3)-Type Equations p < 1 and q = 1 the only prime period-two solution is . of Eq(7. . . 1 − p − φ. . 0. Proof. . . φ.4. q ∈ (0. . φ. It follows from Section 2. 1 − p. . 0. . 1 − p. 1 − p. . 1). Therefore 1−p the prime period-two solution (7. .

5. Open Problem 7. xn−2 .9)-(7.10) (7. . 1 − p. 1 + pxn + qxn−1 + xn−2 n = 0. .5. 1 + xn + xn−1 (7.5.5. ∞). 0.2 Investigate the asymptotic character and the periodic nature of all positive solutions of the difference equation xn+1 = where p.6) converges to a.6). .2 Assume p < 1 and q ≥ 1. . x0 ) ∈ R × R through which the equation is well defined for all n ≥ 0: xn+1 = xn+1 = xn+1 = 1 1 + xn + xn−1 xn 1 + xn + xn−1 xn−1 . 0.9) (7. 1. of Eq(7. Open Problems and Conjectures Conjecture 7. . 1) and q ∈ (1.11). Open Problem 7.5. 1 − p. and investigate ∞ the long term-behavior of the solution {xn }n=−1 . Extend and generalize. Open Problem 7. x0 ) ∈ G be an initial point through which the corresponding equation is well defined for all n ≥ 0. . Find the basin of attraction of the two cycle . . . q ∈ [0. . .11) Now for each of the equations (7. ∞). periodtwo solution.1 Assume p ∈ (0.3 For each of the equations given below. let (x−1 . 135 Show that every positive solution of Eq(7. find the set G of all points (x−1 . not necessarily prime.7.

14). . (7. n = 0. (1. 1. .5.136 Chapter 7. 1 + pn yn + qn yn−1 yn .13) (7. pn + qn yn + yn−1 n = 0. p = lim pn n→∞ and q = n→∞ qn . 1. . Investigate the global character of all positive solutions of Eqs(7.12) (7.5 Assume that {pn }∞ and {qn }∞ are period-two sequences of n=0 n=0 nonnegative real numbers. . n = 0. . . lim Investigate the asymptotic behavior and the periodic nature of all positive solutions of each of the following three difference equations: yn+1 = yn+1 = yn+1 = 1 .4 Assume that {pn }∞ and {qn }∞ are convergent sequences of n=0 n=0 nonnegative real numbers with finite limits. . . 1. 3)-Type Equations Open Problem 7.5. .12)-(7. 1 + pn yn + qn yn−1 yn−1 . . Extend and generalize.14) Open Problem 7. .

. . . 3)-type: xn+1 = xn+1 = and xn+1 = α + βxn + γxn−1 . B is reduced to a (2. . and the denominators are always positive.1 Introduction α + βxn + γxn−1 .2) by the change of the variables xn = yn + β . . the initial conditions are nonnegative. (8. Eq(8. Eq(8.1) is a linear difference equation and can be solved explicitly. Bxn Eq(1) contains the following three equations of the (1.1) (8. Eq(8. . . yn+1 = (α + βγ ) + γyn−1 B . 1. 1. (8.5). β + Byn n = 0. 1. . . .2) Please recall our classification convention in which all parameters that appear in these equations are positive. 1. C . A α + βxn + γxn−1 . Cxn−1 n = 0. 2)-type equation of the form of Eq(6.4) (see Section 6. .Chapter 8 (3. . n = 0.3) n = 0. . namely. 1)-Type Equations 8.3) by the change of the variables xn = yn + 137 γ . .

4).4) be a period-two solution of Eq(8.2) (see Section 6. (b) Let {xn }∞ n=−1 be a positive solution of Eq(8. .3) converges to the positive equilibrium of the equation. Determine the basin of attraction of (8. x0 ) ∈ R × R through which the equation 1 + xn + xn−1 xn+1 = xn is well defined for all n and for these initial points determine the long-term behavior of the solutions {xn }∞ .2. ψ. .138 Chapter 8. yn+1 (α + βγ ) + βyn C . (3.2) which converges to (8. (8.2).1 Show that every positive solution of Eq(8. φ. φ. 8. Conjecture 8.4 Determine the set G of all initial points (x−1 . Determine the set of initial conditions (x−1 . Open Problem 8. ∞) × (0. n=−1 . = γ + Cyn−1 n = 0.2 Open Problems and Conjectures Open Problem 8. . .2. .3 Determine the set G of all initial points (x−1 .2. ψ. .2 (a) Assume γ = β and let .4). Open Problem 8. . .2. . x0 ) ∈ (0. Determine the values of φ and ψ in terms of the initial conditions x−1 and x0 .3) namely.2.1 Assume γ > β. n=−1 Open Problem 8. x0 ) ∈ R × R through which the equation 1 + xn + xn−1 xn+1 = xn−1 is well defined for all n and for these initial points determine the long-term behavior of the solutions {xn }∞ . Hence there is nothing else remaining to do about these equations other than to pose some Open Problems and Conjectures. 2)-type equation of the form of Eq(6.2) are bounded. 1)-Type Equations is reduced to a (2. 1. . ∞) through which the solutions of Eq(8.

8.2. Open Problems and Conjectures

139

Open Problem 8.2.5 Assume that {pn }∞ and {qn }∞ are convergent sequences of n=0 n=0 nonnegative real numbers with finite limits, lim p = n→∞ pn and q = n→∞ qn . lim

Investigate the asymptotic behavior and the periodic nature of all positive solutions of each of the following two difference equations: yn+1 = yn+1 = pn + xn + xn−1 , q n xn pn + xn + xn−1 , qn xn−1 n = 0, 1, . . . (8.5)

n = 0, 1, . . . .

(8.6)

∞ ∞ Open Problem 8.2.6 Assume that {pn }n=0 and {qn }n=0 are period-two sequences of nonnegative real numbers. Investigate the global character of all positive solutions of Eqs(8.5) and (8.6). Extend and generalize.

Chapter 9 (2, 3)-Type Equations
9.1 Introduction
α + βxn , A + Bxn + Cxn−1 α + γxn−1 , A + Bxn + Cxn−1 βxn + γxn−1 , A + Bxn + Cxn−1

Eq(1) contains the following three equations of the (2, 3)-type: xn+1 = xn+1 = and xn+1 = n = 0, 1, . . . . (9.3) n = 0, 1, . . . n = 0, 1, . . . (9.1) (9.2)

Please recall our classification convention in which all the parameters in the above (2, 3)-type equations are positive and the initial conditions are nonnegative.

9.2

The Case γ = 0 : xn+1 =

α+βxn A+Bxn +Cxn−1

This is the (2, 3)-type Eq(9.1) which by the change of variables xn = reduces to the difference equation yn+1 = where p= p + qyn , 1 + yn + ryn−1 q= β , A n = 0, 1, . . . C . B (9.4) A yn B

αB , A2

and r =

141

142

Chapter 9. (2, 3)-Type Equations

(Eq(9.4) was investigated in [51].) Here we make some simple observations about linearized stability, invariant intervals, and the convergence of solutions in certain regions of initial conditions. Eq(9.4) has a unique equilibrium y which is positive and is given by y= q−1+ (q − 1)2 + 4p(r + 1) 2(r + 1) .

By employing the linearized stability Theorem 1.1.1, we obtain the following result: Theorem 9.2.1 The equilibrium y of Eq(9.4) is locally asymptotically stable for all values of the parameters p, q and r.

9.2.1

Boundedness of Solutions

We will prove that all solutions of Eq(9.4) are bounded. Theorem 9.2.2 Every solution of Eq(9.4) is bounded from above and from below by positive constants. Proof. Let {yn } be a solution of Eq(9.4). Clearly, if the solution is bounded from above by a constant M , then p yn+1 ≥ 1 + (1 + r)M and so it is also bounded from below. Now assume for the sake of contradiction that the solution is not bounded from above. Then there exists a subsequence {y1+nk }∞ such k=0 that
k→∞

lim nk = ∞,

k→∞

lim y1+nk = ∞,

and y1+nk = max{yn : n ≤ nk } for k ≥ 0.

From (9.4) we see that yn+1 < qyn + p for n ≥ 0 and so
k→∞

lim ynk = lim ynk −1 = ∞.
k→∞

Hence for sufficiently large k, 0 < y1+nk − ynk = p + [(q − 1) − ynk − rynk −1 ]ynk <0 1 + ynk + rynk −1 2

which is a contradiction and the proof is complete.

The above proof has an advantage that extends to several equations of the form of Eq(9.1) with nonnegative parameters. The boundedness of solutions of the special

9.2. The Case γ = 0 : xn+1 =

α+βxn A+Bxn +Cxn−1

143

equation (9.4) with positive parameters immediately follows from the observation that if M = max{1, p, q} then yn+1 ≤ M + M yn =M 1 + yn f or n ≥ 0.

9.2.2

Invariant Intervals

The following result, which can be established by direct calculation, gives a list of invariant intervals for Eq(9.4). Recall that I is an invariant interval, if whenever, yN , yN +1 ∈ I then yn ∈ I (a) for n ≥ N. Lemma 9.2.1 Eq(9.4) possesses the following invariant intervals:

for some integer N ≥ 0,

0,

q−1+

(q − 1)2 + 4p 2

,

when p ≤ q;

(b) p−q ,q , qr (c) q, p−q , qr when p > q(rq + 1). when q < p < q(rq + 1);

Proof. (a) Set q − 1 + (q − 1)2 + 4p p + qx and b = 1+x 2 and observe that g is an increasing function and g(b) ≤ b. Using Eq(9.4) we see that when yk−1 , yk ∈ [0, b], then g(x) = yk+1 = p + qyk p + qyk ≤ = g(yk ) ≤ g(b) ≤ b. 1 + yk + ryk−1 1 + yk

The proof follows by induction.

144 (b) It is clear that the function f (x, y) =

Chapter 9. (2, 3)-Type Equations

p + qx 1 + x + ry
p−q ,q qr

is increasing in x for y > p−q . Using Eq(9.4) we see that when yk−1 , yk ∈ qr then p + qyk p−q = f (yk , yk−1 ) ≤ f q, = q. yk+1 = 1 + yk + ryk−1 qr Also by using the condition p < q(rq + 1) we obtain yk+1 = p + qyk = f (yk , yk−1 ) ≥ f 1 + yk + ryk−1

,

p−q q(pr + p − q) p−q > ,q = . 2 + rq + p − q qr (rq) qr

The proof follows by the induction. (c) It is clear that the function f (x, y) = p + qx 1 + x + ry

is decreasing in x for y < p−q . Using Eq(9.4) we see that when yk−1 , yk ∈ q, p−q , qr qr then p + qyk p−q p−q yk+1 = = q. , = f (yk , yk−1 ) ≥ f 1 + yk + ryk−1 qr qr Also by using the condition p > q(rq + 1) we obtain yk+1 = p + qyk p−q p + q2 < . = f (yk , yk−1 ) ≤ f (q, q) = 1 + yk + ryk−1 1 + (r + 1)q qr

The proof follows by induction. 2

9.2.3

Semicycle Analysis
qr( p−q − yn−1 ) qr 1 + yn + ryn−1

Let {yn }∞ n=−1 be a solution of Eq(9.4). Then the following identitites are true for n ≥ 0: yn+1 − q = qr q − p−q yn+1 − = qr
p−q qr

, + pr(q − yn−1 )

(9.5)

qr(1 + yn + ryn−1 ) yn − yn+4 =

yn + qr yn−1 −

p−q qr

,

(9.6)

9.2. The Case γ = 0 : xn+1 = qr yn −
p−q qr

α+βxn A+Bxn +Cxn−1

145

2 yn+1 + (yn − q)(yn+1 yn+3 + yn+3 + yn+1 + ryn yn+3 ) + yn + ryn − p

(1 + yn+3 )(1 + yn+1 + ryn ) + r(p + qyn+1 )

,

(9.7) yn+1 − y = (y − q)(y − yn ) + yr(y − yn−1 ) . 1 + yn + ryn−1 (9.8)

The proofs of the following two lemmas are straightforward consequences of the Identities (9.5)- (9.8) and will be omitted. Lemma 9.2.2 Assume p > q(qr + 1) and let {yn }∞ n=−1 be a solution of Eq(9.4). Then the following statements are true: (i) If for some N ≥ 0, (ii) If for some N ≥ 0, (iii) If for some N ≥ 0, (iv) If for some N ≥ 0, (v) If for some N ≥ 0, (vi) If for some N ≥ 0, (vii) q < y <
p−q . qr

yN > yN = yN <

p−q , qr p−q , qr p−q , qr

then then then
p−q , qr

yN +2 < q; yN +2 = q; yN +2 > q; q < yN +2 <
p−q ; qr

q < yN <

then

y ≥ yN −1 and y ≥ yN , then y < yN −1 and y < yN , then

yN +1 ≥ y; yN +1 < y;

Lemma 9.2.3 Assume q < p < q(qr + 1) and let {yn }∞ n=−1 be a solution of Eq(9.4). Then the following statements are true: (i) If for some N ≥ 0, (ii) If for some N ≥ 0, (iii) If for some N ≥ 0, (iv) If for some N ≥ 0, (v)
p−q qr

yN < yN = yN > yN >

p−q , qr p−q , qr p−q , qr p−q qr

then then then

yN +2 > q; yN +2 = q; yN +2 < q; then q > yN +2 >
p−q ; qr

and yN < q

< y < q.

(9. Identity (9. More precisely. y4k ≥ 1. Then this solution is oscillatory and the sum of the lengths of two consecutive semicycles. 1. 3)-Type Equations qr (q − yn−1 ).2.2. 2 Here we present a semicycle analysis of the solutions of Eq(9. y4k ≥ 1. 1 + yn + ryn−1 n = 0. The proof follows from identity (9.146 Lemma 9. y4k ≤ 1. y4k+1 < q. and y4k+2 ≤ q.4) becomes yn+1 = q + rq 2 + qyn . Proof. y4k+1 > q. and y4k+2 ≥ q.4 Assume p = q(qr + 1) and let {yn }∞ n=−1 be a solution of Eq(9. . (iv) y−1 < q and y0 ≤ q =⇒ y4k−1 < q. (2. and y4k+2 ≥ q. y4k ≤ 1.4). . y4k+1 > q. Theorem 9.10) is a consequence of the fact that in this case qr ∈ (0. (iii) y−1 > q and y0 ≥ q =⇒ y4k−1 > q. then n→∞ Chapter 9. (ii) y−1 < q and y0 ≥ q =⇒ y4k−1 < q. and y4k+2 ≤ q. and the only equilibrium point is y = q.9) follows by straightforward computation. The limit (9. y4k+1 < q.5). the following statements are true for all k ≥ 0: (i) y−1 > q and y0 ≤ q =⇒ y4k−1 > q.4) when p = q(1 + rq). In this case Eq(9. 2 .4). . 1 + yn + ryn−1 lim yn = y. Then yn+1 − q = Furthermore when qr < 1.9) (9. excluding the first.4) has no period-two solution. is equal to four.10) Proof.3 Suppose that p = q + rq 2 and let {yn }∞ n=−1 be a nontrivial solution of Eq(9. ∞) and Eq(9.

also assume that the solution {yn }∞ n=−1 is not eventually in the interval I. yN ≤ b.   2 I=             [ p−q . More precisely.4) lies eventually in I. Set b= and observe that yn+1 − b = (q − p)(yn − b) − r(p + qb)yn−1 . there exists N > 0 such that one of the following three cases holds: (i) yN > q.9.2. p > q(qr + 1). p−q ] qr if if q < p < q(qr + 1). ∞ Proof. qr n→∞ lim yn = y ∈ I and yN +2 < p−q . Then every solution of Eq(9.2. the following is true: Lemma 9.4 Global Asymptotic Stability The following lemma establishes that when p = q(qr + 1). then yN +1 < b. Then clearly. y ∈ I.4) is eventually trapped into one of the three invariant intervals of Eq(9. Let {yn }n=−1 be a solution of Eq(9. qr and yN +2 < . q] qr [q. if for some N . q−1+ (q − 1)2 + 4p 2 Hence. Next assume that q < p < q(qr + 1) and. (ii) yN > q.1. yN +1 > q.2. Then by Lemma 9.4). q−1+ (q−1) +4p ] if p ≤ q.5 Let I denote the interval which is defined as follows: √  2   [0. Then yn > y for n ≥ 0 and so which is a contradiction.2. The Case γ = 0 : xn+1 = α+βxn A+Bxn +Cxn−1 147 9. Now assume for the sake of contradiction that yn > b for n > 0. yN +1 < p−q .3. (1 + yn + ryn−1 )(1 + b) n ≥ 0. qr p−q .4) described in Lemma 9.2. First assume that p ≤ q. for the sake of contradiction. every solution of Eq(9.

Proof. together with the appropriate convergence Theorems 1. and yN +2 < p−q . 1+r Then the equilibrium y of Eq(9. 3)-Type Equations ≤ yN +1 ≤ q.2. each subsequence {yn+4k+j }∞ with all its terms outside k=0 the interval I converges monotonically and enters in the interval I. . or q . p−q qr Chapter 9.4): Theorem 9. 3}.4) is globally asymptotically stable. (b) Assume that either (iii) r > 1 (ii) r ≤ 1. qr The desired contradiction is now obtained by using the Identity (9. qr Also observe that 2 if yN ≥ q. The proof when p > q(qr + 1) is similar and will be omitted. then yN + ryN − p < 0.4 (a) Assume that either p ≥ q + q 2 r.4.2.4. the following results are true for Eq(9. Then the equilibrium y of Eq(9. and (q − 1)2 (r − 1) ≤ 4p. For example. 2.4). By using the monotonic character of the function f (x. p< p≤q or q < p < q + q2r and that one of the following conditions is also satisfied: (i) q ≤ 1.5. then yN + ryN − p > 0 and 2 if yN ≤ p−q .7) from which it follows that for j ∈ {0. 1.148 (iii) yN > q. y) = p + qx 1 + x + ry 2 in each of the intervals in Lemma 9. we can obtain some global asymptotic stability results for the solutions of Eq(9. (2.1.4) is globally asymptotically stable.7 and 1.

m) then M = m. m = M from which the result follows.11) A yn B αB .4 and Theorems 1.4.2. A2 and r = . 3)-type Eq(9.5. B (9.9.4. 1. 2 9. M ) and M = f (M.4. Now if m + M = q − 1.1 we see that when y−1 . The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 149 (a) The proof follows from Lemmas 9. . .1 and 9. 1 + M + rm Hence (M − m)(1 − q + M + m) = 0. y0 ∈ 0. C . We will employ Theorem 1. . A n = 0.2. 2 2 √ 2 +4p q−1+ (q−1) .3. The proof when q < p < q + q 2 r holds follows from Lemma 9.2. then yn ∈ 2 √ √ q−1+ (q−1)2 +4p q−1+ (q−1)2 +4p for all n ≥ 0. It is easy to check that y ∈ 0. 0. . this is the case if Condition (i) is satisfied.5 and so it remains to be shown that if m = f (m.1 and Theorem 1. 2 decreases in y. Clearly now if Condition (ii) or (iii) is satisfied. If m + M = q − 1 then m and M satisfy the equation (r − 1)m2 + (r − 1)(1 − q)m + p = 0.3 The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 This is the (2. then M = m.2. the function f increases in x and and that in the interval 0.7 and 1. This system has the form m= p + qm 1 + m + rM and M = p + qM .2) which by the change of variables xn = reduces to the difference equation yn+1 = where p= p + qyn−1 . For instance.2. 1 + yn + ryn−1 q= γ .4. √ q−1+ (q−1)2 +4p (b) In view of Lemma 9.

3)-Type Equations This equation has not been investigated yet. . when q > 1 and (r − 1)(q − 1)2 + 4pr2 < 0. By employing the linearized stability Theorem 1. .1 Let Existence and Local Stability of Period-Two Cycles . ψ. .11) has a unique equilibrium y which is positive and is given by y= q−1+ (q − 1)2 + 4p(r + 1) 2(r + 1) .1 Eq(9. . φ. 1 + φ + rψ It now follows after some calculation that the following result is true.12) holds. and the convergence of solutions in certain regions of initial conditions. Furthermore when (9.11).1 we obtain the following result. . and more precisely a saddle point.150 Chapter 9. .11) is unstable. (9. the period-two solution is “unique” and the values of φ and ψ are the positive roots of the quadratic equation t2 − p q−1 = 0. invariant intervals. t+ r 1−r . Theorem 9. . Eq(9. . (2. ψ.3. ψ.12) holds. ψ. φ.11) has a prime period-two solution . Then φ= p + qφ 1 + ψ + rφ and ψ = p + qψ .11) is locally asymptotically stable when either q≤1 or q > 1 and (r − 1)(q − 1)2 + 4pr2 > 0.3. Lemma 9. Here we make some simple observations about linearized stability. . be a period-two cycle of Eq(9. . φ. (b) The equilibrium y of Eq(9.1 (a) The equilibrium y of Eq(9. φ. if and only if (9. . .3.1.12) 9.

φ. 1 + g(u. . 1. . evaluated at ψ We have   ∂g ∂g (φ. v) = p + qu 1 + v + ru u v = g(u. and write Eq(9. φ. Let T be the function on (0. ∞) × (0. ψ) = ∂u (1 + ψ + rφ)2 ∂g −p − qφ (φ. ∞) defined by: T Then u v = v p+qu 1+v+ru . v) + rv and h(u. Jacobian matrix JT 2 . . ψ) = . (φ.11) in the equivalent form: un+1 = vn vn+1 = p+qun 1+vn +run .9. ψ. ψ) ∂v (φ. ψ. . ψ) ∂v (φ. lie inside the unit disk. By a simple calculation we find that T2 where g(u. ψ) ∂u where ∂g q − pr + qψ . . The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 151 To investigate the local stability of the two cycle . for n = 0. n = 0. v) p + qv . . .3. . the second iterate of T . . . φ ψ is a fixed point of T 2 . v) = Clearly the two cycle is locally asymptotically stable when the eigenvalues of the φ . . 1. v) h(u.  ψ ∂h ∂h (φ. . ψ) ∂u φ  = JT 2 . . we set un = yn−1 and vn = yn . ∂v (1 + ψ + rφ)2 .

ψ). (9. ψ) (φ.1. This inequality is equivalent to (q − pr + qφ) (q − pr + qψ) − (1 + ψ + rφ)2 (1 + φ + rψ)2 < 0. Observe that.16) First we will establish Inequality (9. 3)-Type Equations ∂h (p + qψ)(pr − q − qψ) (φ. ∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1.14). ψ) + (φ.152 Chapter 9. ψ) − (φ. if and only if |S| < 1 + D < 2.13) is equivalent to the following three inequalities: D<1 S <1+D −1 − D < S. Thus . ψ) = + . ∂u (1 + ψ + rφ)2 (1 + φ + rψ)2 ∂h (p + qψ)(p + qφ) q − pr + qφ (φ. ψ) ∂u ∂v ∂h ∂g ∂h ∂g (φ. ψ) (φ. (q − pr + qφ) (q − pr + qψ) = (q − pr)2 + q(q − pr)(φ + ψ) + q 2 φψ = = −2q 2 r + q 2 r2 + 3qpr2 − 2qpr3 − p2 r3 + p2 r4 − q 3 + rq 3 + q 2 − pr2 q 2 − qpr r (−1 + r) (1 + ψ + rφ)2 (1 + φ + rψ)2 = 1 + φ + rψ + ψ + ψφ + rψ 2 + rφ + φ2 r + r2 φψ = −pr2 + qr + pr − q + q 2 r 2 2 and that . (2.15) (9.13) (9.14) (9. Inequality (9. ψ) = . 2 (1 + φ + rψ)2 ∂v (1 + ψ + rφ) (1 + φ + rψ)2 Set S= and D= ∂h ∂g (φ.1 that both eigenvalues of JT 2 unit disk |λ| < 1.

r2 (1 − r) −r2 p + rp − q + q 2 + qr = rp(1 − r) + q(q − 1) + qr > 0. Next we turn to Inequality (9.15).16) is equivalent to (q − pr + qψ)(1 + φ + rψ)2 + (p + qψ)(p + qφ) + (q − pr + qφ)(1 + ψ + rφ)2 + (q − pr + qφ) (q − pr + qψ) + (1 + ψ + rφ)2 (1 + φ + rψ)2 > 0.12). The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 153 (q − pr + qφ) (q − pr + qψ) − (1 + ψ + rφ)2 (1 + φ + rψ)2 = −pr3 q 2 + 2q 2 r − q 2 r2 − q 2 + 3p2 r3 − 2p2 r4 − 3rq 3 + 3qpr3 − 5qpr2 r2 (1 − r) + 2qpr + 4pr2 q 2 + 2q 3 + q 4 r − p2 r2 − 2prq 2 − q 4 + r2 q 3 r2 (1 − r) Note that (2r2 p − rp + q − q 2 − qr + q 2 r) (−r2 p + rp − q + q 2 + qr) = . Also by using Condition (9. we have established the following result: .3.12). After some lengthy calculation one can see that this inequality is also a consequence of Condition (9. Finally Inequality (9. This inequality is equivalent to (q − pr + qψ)(1 + φ + rψ)2 + (p + qψ)(p + qφ) + (q − pr + qφ)(1 + ψ + rφ)2 − (q − pr + qφ) (q − pr + qψ) − (1 + ψ + rφ)2 (1 + φ + rψ)2 < 0.12) we see that (r − 1)(q − 1)2 + 4pr2 = q 2 r − 2qr + r − q 2 + 2q − 1 + 4r2 p < 0 and so 2r2 p − rp + q − q 2 − qr + q 2 r < 2r2 p − rp + q − qr + 2qr − r − 2q + 1 − 4r2 p = −2r2 p − rp − q + qr − r + 1 = −2r2 p − rp + (r − 1) (q − 1) < 0 from which the result follows.9. In summary. After some lengthy calculation one can see that this inequality is a consequence of Condition (9.

Using Eq(9. (b) pr − q q . .2 Eq(9. φ. ψ. then yk+1 = p + qyk−1 p + qyk−1 ≤ = g(yk−1 ) ≤ g(b) ≤ b. this period-two solution is locally asymptotically stable. ψ. r r Proof. .12) holds.11) we see that when yk−1 .12) holds. . . φ. if and only if (9. Furthermore when (9. . r q when pr ≥ q + q2 . which can be established by direct calculation. (2. gives a list of invariant intervals for Eq(9. 3)-Type Equations Theorem 9. (a) Set g(x) = p + qx 1 + rx and b = q−1+ (q − 1)2 + 4pr 2r and observe that g is an increasing function and g(b) ≤ b. Lemma 9. 9. b]. r when 1 < q < pr < q + q2 q − . .3.11). .2 Invariant Intervals The following result.11) has a unique prime period-two solution .154 Chapter 9. q−1+ (q − 1)2 + 4pr 2r when pr ≤ q. q r (c) q pr − q .11) possesses the following invariant intervals: (a)    0. 1 + yk + qyk−1 1 + ryk−1 The proof follows by induction. yk ∈ [0.2 Eq(9.3.3.

q p + qy 1 + x + ry . y) = p + qy 1 + x + ry pr−q q . y) and the condition 2 q < pr < q + qr we obtain yk+1 = p + qyk−1 = f (yk .3. yk−1 ) ≥ f .r q is increasing in y for x > pr−q . r By using the fact that f (x. (c) It is clear that the function f (x.11) we see that when yk−1 . y) is decreasing in both arguments and the condition pr > q + rq2 we obtain yk+1 = q q p + qyk−1 = f (yk . Using Eq(9. 1 + yk + qyk−1 r r = pr + q 2 pr − q . ≥ yk+1 = 1 + yk + ryk−1 r q q . q r q = . yk−1 ) ≤ f . The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 155 (b) It is clear that the function f (x. then yk+1 = Using Eq(9. yk−1 ) ≥ f 1 + yk + qyk−1 pr − q pr − q . .11) we see that when yk−1 . q q q = . 2 .9. < r + (r + 1)q q The proof follows by induction. yk−1 ) ≤ f 1 + yk + ryk−1 pr − q q . y) = is decreasing in y for x < q pr−q . yk ∈ q then q pr − q p + qyk−1 pr − q = f (yk . yk ∈ p + qyk−1 = f (yk . Now by using the monotonic character of the function f (x. r The proof follows by induction. q r pr−q .

(ii) r ≥ 1. together with the appropriate convergence theorem (from among Theorems 1.3.2. 3)-Type Equations 9.4. then Proof.4 (a) Assume that 1 < q < pr < q+ qr − q and that one of the following r conditions is also satisfied: (i) r ≥ 1. the following results are true for Eq(9.2 we see that when y−1 . 2r √ √ q−1+ (q−1)2 +4pr q−1+ (q−1)2 +4pr yn ∈ 0. It is easy to check that y ∈ 0.156 Chapter 9. q−1+ (q − 1)2 + 4pr 2r  converges to the equilibrium y.   Then every solution of Eq(9. For example. In view of Lemma 9. y) = in each of the intervals in Lemma 9.3.3 Convergence of Solutions p + qy 1 + x + ry By using the monotonic character of the function f (x. for all n ≥ 0.3 Assume that pr ≤ q and that one of the following conditions is also satisfied: (i) q ≤ 1.11) with initial conditions in the invariant interval 0. the function f decreases in x and increases 2r in y.6.8) we can obtain some convergence results for the solutions with initial conditions in the invariant intervals. (iii) r < 1 and (r − 1)(q − 1)2 + 4pr2 ≥ 0.4. 2 Theorem 9.4.11): Theorem 9. y0 ∈ 0.3.5-1. . 2r 2r √ q−1+ (q−1)2 +4pr and that in the interval 0. The result now follows by employing Theorem 1. .3.3. (2. 2 q−1+ √ (q−1)2 +4pr (ii) r < 1 and (r − 1)(q − 1)2 + 4pr2 ≥ 0.

It is easy to 9. . y) decreases in x and increases in y. pr−q and that in the r q in both x and in y. The result now follows by employing Theorem 1. r2 + r2 yn + r3 yn−1 n = 0. q r converges to the equilibrium y. (a) In view of Lemma 9.5 Assume r ≥ 1 and pr ≤ q. r q converges to the equilibrium y.11) becomes yn+1 = qr + q 2 + qr2 yn−1 .3.11) when pr = q + this case Eq(9. In and the only positive equilibrium is y = q . It is easy to (b) In view of Lemma 9. Theorem 9.4. Then the positive equilibrium of Eq(9.r q pr−q q .6.4 Global Stability By applying Theorem 1. .11).5 Semicycle Analysis When pr = q + Here we present a semicycle analysis of the solutions of Eq(9.4.11) with initial conditions in the invariant interval pr − q q . 1.9. 2 q pr−q . Proof. we obtain the following global asymptotic stability result.11) is globally asymptotically stable.3.3.2 we conclude that yn ∈ for all n ≥ 0.3.3. The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 157 Then every solution of Eq(9.4. . q2 r q2 . pr−q r q for all n ≥ 0.7.r q 2 check that y ∈ and that the function f (x. r .3. Then every solution of Eq(9.11) with initial conditions in the invariant interval q pr − q . q r interval q .3 to Eq(9. (b) Assume that pr > q + qr . r 9.2 we conclude that yn ∈ pr−q q . the function f decreases check that y ∈ q . The result now follows by employing Theorem 1.

18) The following theorem is a consequence of Theorem 1. Here we make some simple observations about linearized stability.1. .17) also possesses the unique positive equilibrium y= p+1−r .3. . Eq(9.3. γ This equation has not been investigated yet. Then after the first semicycle. invariant intervals.1.1 and Theorem 1.17) are the nonnegative solutions of the equation y= (p + 1)y . r r r (1 + yn + ryn−1 ) 2 9. Proof. C and r = A . γ q= B .158 2 Chapter 9. r + qyn + yn−1 β . The equilibrium points of Eq(9. q+1 (9. (9. the solution oscillates about the equilibrium y with semicycles of length one.3) which by the change of variables xn = reduces to the difference equation yn+1 = where p= pyn + yn−1 . . . C n = 0.17) γ yn .6 Suppose that pr = q + qr and let {yn }∞ n=−1 be a nontrivial solution of Eq(9. 3)-type Eq(9. 1. (2. 3)-Type Equations Theorem 9. The proof follows from the relation yn+1 − q q qr2 = ( − yn ) 3 .4 The Case α = 0 : xn+1 = βxn +γxn−1 A+Bxn +Cxn−1 This is the (2. and the convergence of solutions in certain regions of initial conditions.11). r + (1 + q)y Hence zero is always an equilibrium point and when p + 1 > r.

4. Then the positive equilibrium of Eq(9. .3 Eq(9. φ.20) (9.17) is locally asymptotically stable when q + r < 3p + 1 + qr + pq. q−1 . Theorem 9. Then the zero equilibrium of Eq(9.17) about its positive equilibrium y is zn+1 − p − q + qr q−p+r zn − zn−1 = 0.18) holds. . The linearized equation associated with Eq(9.5 that the following result is true. .4. . The Case α = 0 : xn+1 = Theorem 9.19) Concerning prime period-two solutions for Eq(9. . The following result is a consequence of Theorem 1.17) is globally asymptotically stable. φ. ψ.4.1. Furthermore when (9. (b) Assume that p + 1 > r.17).2 Assume that (9. Furthermore the zero equilibrium is a saddle point when 1−p<r <1+p and a repeller when r < 1 − p. ψ. 1. . . . (9.1 βxn +γxn−1 A+Bxn +Cxn−1 159 (a) Assume that p + 1 ≤ r. (p + 1)(q + 1) (p + 1)(q + 1) n = 0. and unstable (a saddle point) when q + r > 3p + 1 + qr + pq.17) has a prime period-two solution . Theorem 9.17) is unstable.9. it follows from Section 2. if and only if (9. .20) holds there is a unique period-two solution and the values of φ and ψ are the positive roots of the quadratic equation t2 − (1 − p − r)t + p(1 − p − r) = 0.20) holds.1. Then the zero equilibrium of Eq(9.4.

. K2 −p [0.1 Eq(9. and q 2 + q ≤ r. K] when p = q 2 where K > 0 is an arbitrary number.4. when q 2 − qr ≤ p < q 2 and p + q > r. p+q−r q+1 when p = q 2 and q 2 + q > r where K > 0 is an arbitrary number and qr . when p > q 2 + r where ((r − q)(p − q 2 ) + qr)2 + 4q 3 r(p − q 2 ) 2q(p − q 2 ) . K] when q 2 − qr ≤ p < q 2 and p + q ≤ r.17). Lemma 9. K1 p − q2 K1 = (c) 0. where K > 0 is an arbitrary number and pr . (b) 0. which can be established by direct calculation. qr p − q2 when q 2 < p ≤ q 2 + r and p + q > r.160 Chapter 9. ((q − r)(q 2 − p))2 + 4p3 qr2 2pqr (q − r)(q 2 − p) + .1 Invariant Intervals The following result. 3)-Type Equations 9. q2 where K2 = when q 2 > p + qr.4. (2. and [0. K] when q 2 < p ≤ q 2 + r and p + q ≤ r. gives a list of invariant intervals for Eq(9.17) possesses the following invariant intervals: (a) 0. q2 pr −p (q − r)(p − q 2 ) − qr + [0.

yk−1 ) ≤ f (K. −p Then one can see that for yk−1 . y) = r + qx + y is increasing in both arguments. qr Hence for yk−1 . yk ∈ f one can see that ≤ K2 . K2 ≤ yk+1 = f (yk . q2pr we obtain −p yk+1 = f (yk . K) = (p + q)K with K = pr . βxn +γxn−1 A+Bxn +Cxn−1 161 (a) It is easy to see that when p = q 2 the function px + qy f (x. Then for yk−1 . K) ≤ K is equivalent to q 2 +q−r ≤ (q+1)K. 2 p−q p − q2 pr . K) = (p + q)K Second assume that p > q 2 + r. p−q2 we obtain qr . yk ∈ f qr . K1 p−q 2 we can see that qr qr . this inequality is satisfied for any K > 0. Now observe that the inequality f (K. 2 q2 − p q −p . y) ≤ f (K. ) ≤ K1 . q 2 −p First assume that q 2 − qr ≤ p < q 2 . 1 ≤ K. yk−1 ) ≤ f (K1 . K2 q 2 −p Second assume that p < q 2 − qr. K). y) is increasing in both arguments for x < qr and it is increasing in x and decreasing in y for x ≥ p−q2 . yk ∈ 0. 2 pr pr . y) is increasing in both arguments for y < and it is increasing in y and decreasing in x for y ≥ q2pr . K1 ≤ yk+1 = f (yk . Therefore for any K > 0.9. while when q 2 + q > r the inequality is true when p+q−r K≥ . The last inequality is always satisfied when p + q ≤ r and when p + q > r. it is satisfied under the condition r + q 2 > p.4. q+1 (b) It is clear that the function f (x. Then for yk−1 . p−q 2 0 ≤ f (x. 1 ≤ K. First assume that q 2 < p ≤ q 2 + r . The Case α = 0 : xn+1 = Proof. r + (q + 1)K qr where we set K = p−q2 . yk−1 ) ≤ f K2 . yk−1 ) ≤ f (K. When q 2 + q ≤ r. yk+1 = f (yk . yk ∈ 0. r + (q + 1)K pr . q 2 −p (c) It is clear that the function f (x.

4. converges to the equilibrium y.4 (a) Assume that p = q 2 and q 2 + q > r.5-1. together with the appropriate convergence theorem (from among Theorems 1. converges to the equilibrium y.17) with initial conditions in the invariant interval 0.17) with initial conditions in the interval [0. For example. the following results are true for Eq(9. K]. K]. Then every solution of Eq(9. 3)-Type Equations 9.4.17): Theorem 9. K]. Assume that q 2 − qr ≤ p < q 2 and p + q ≤ r. for any K > 0. for any K > 0.4.17) with initial conditions in the interval [0.162 Chapter 9. Proof. Then every solution of Eq(9. we can obtain some convergence results for the solutions with initial conditions in the invariant intervals. Then every solution of Eq(9.17) with initial conditions in the invariant interval 0.4.17) with initial conditions in the invariant interval 0. (c) Assume that q 2 − qr ≥ p < q 2 and p + q > r. Then every solution of Eq(9. converges to the equilibrium y. y) = in each of the intervals in Lemma 9.8. The proof is an immediate consequence of Lemma 9. converges to the equilibrium y.1 and Theorem 1. (b) Assume that q 2 + r ≥ p > q 2 and p + q > r.4. converges to the equilibrium y. converges to the equilibrium y. 2 q2 pr −p qr p − q2 q2 + q − r q+1 .8).17) with initial conditions in the interval [0.4. (2. for any K > 0.1.2 Convergence of Solutions px + qy r + qx + y By using the monotonic character of the function f (x. Assume that q 2 + r ≥ p > q 2 and p + q ≤ r.4. Assume that p = q 2 and q 2 + q ≤ r. Then every solution of Eq(9. Then every solution of Eq(9.

1 Assume Show that every positive solution of Eq(9.5 Assume that p > q2 + r and that one of the following conditions is also satisfied: (i) p ≤ q + r.4. γ. Conjecture 9. B. K1 p−q 2 3q 2 +q−qr+r . A.3 Assume β. γ. 2 Theorem 9. K2 converges to the equilibrium y.4. 2 9. Open Problems and Conjectures Theorem 9. ∞). (iii) q < 1 qr . A.1) has a finite limit. The proof is a consequence of Lemma 9. B. Show that every positive solution of Eq(9. β.4. ∞). Conjecture 9. B. 1−q 163 and p ≤ Then every solution of Eq(9.5 Open Problems and Conjectures α. C ∈ (0.1 and Theorem 1.5. Proof. C ∈ (0.4.11) with initial conditions in the invariant interval converges to the equilibrium y.5.2 Assume α.6 Assume that q 2 > p + qr and that Condition (9. C ∈ (0. Conjecture 9. . ∞). p−q 2 Proof.5.11) with initial conditions in the invariant interval qr .5. Show that every positive solution of Eq(9.2) is bounded.9.4.5. (ii) q ≥ 1.20) is not satisfied.4. A.1 and Theorem 1. Then every solution of Eq(9.3) is bounded. The proof is an immediate consequence of Lemma 9.6.

5.5. ∞). x0 ) ∈ R × R through which the equation 1 + xn−1 xn+1 = 1 + xn + xn−1 is well defined for all n ≥ 0. x0 ) ∈ R × R through which the equation 1 + xn xn+1 = 1 + xn + xn−1 is well defined for all n ≥ 0. A. ∞) and suppose that Eq(9. and for these initial points investigate the long-term behavior of the solutions {xn }∞ . x0 ∈ [0. n=−1 Open Problem 9. Suppose that Eq(9. (2. C ∈ (0. Conjecture 9.5. γ.3) has no prime period-two solutions.2 Determine the set G of all initial points (x−1 .3) is globally asymptotically stable. Extend and generalize.3 Determine the set G of all initial points (x−1 . 3)-Type Equations α. B. A. Extend and generalize. Extend and generalize.5. Show that the positive equilibrium of Eq(9.5 Assume that x−1 . Open Problem 9.2) is globally asymptotically stable. γ. n=−1 .164 Conjecture 9. Show that the positive equilibrium of Eq(9.5.4 Assume Chapter 9. C ∈ (0. and for these initial points investigate the long-term behavior of the solutions {xn }∞ . n=−1 Open Problem 9.2) has no prime period-two solutions. B. ∞) with x−1 + x0 > 0 and that β. x0 ) ∈ R × R through which the equation xn + xn−1 xn+1 = 1 + xn + xn−1 is well defined for all n ≥ 0.1 Determine the set G of all initial points (x−1 . and for these initial points investigate the long-term behavior of the solutions {xn }∞ .

. . . ψ. .7 Assume that (9. 1. Investigate the basin of attraction of the two cycle . and {rn }∞ are period-two sen=0 n=0 quences of nonnegative real numbers.6 Assume that (9. 1. 1 + yn + rn yn−1 pn + qn yn−1 . . . . . {qn }∞ . Open Problem 9. 1 + yn + rn yn−1 n = 0.23) pn yn + yn−1 .17).5. . . ψ. rn + qn yn + yn−1 ∞ Open Problem 9. ψ. Investigate the basin of attraction of the two cycle . ψ.11). lim Investigate the asymptotic behavior and the periodic nature of all positive solutions of each of the following three difference equations: yn+1 = yn+1 = yn+1 = pn + qn yn .5. . . .5. . Open Problems and Conjectures 165 Open Problem 9. . . p = lim pn . Investigate the global character of all positive solutions of Eqs(9. . . and {rn }∞ are convergent sen=0 n=0 n=0 quences of nonnegative real numbers with finite limits. Conjecture 9. . n→∞ q = n→∞ qn . n = 0.5.6 Assume that (9.20) holds. lim and r = n→∞ rn . {qn }n=0 . φ. of Eq(9. (9.21)-(9. .20) holds.9. ψ.5. of Eq(9. of Eq(9.23).12) holds. Show that the period-two solution .5. . φ. .22) (9. Extend and generalize. n = 0. φ. . . ψ. .21) (9. φ. Open Problem 9. 1. .17) is locally asymptotically stable. . φ. .5 Assume that {pn }∞ .4 Assume that {pn }∞ . . φ.

8 Assume that p. Investigate the global behavior of all positive solutions of each of the following difference equations: p + qyn (9. . . . . . . . 1.26) . .25) (9.}. 1. . r + qyn + yn−k n = 0. .5. 3. ∞) and Chapter 9. n = 0. 1 + yn + ryn−k yn+1 = yn+1 = p + qyn−k .24) yn+1 = . (2. . . 3)-Type Equations k ∈ {2. n = 0. q. r ∈ [0. . (9. 1 + yn + ryn−k pyn + yn−k .166 Open Problem 9. 1.

1) which by the change of variables xn = reduces to the difference equation yn+1 = where p= p + qyn + ryn−1 . n = 0.4) and r = 167 . A + Cxn−1 Eq(1) contains the following three equations of the (3.2 The Case C = 0 : xn+1 = A yn B α+βxn +γxn−1 A+Bxn This is the (3. . . 1 + yn αB . A n = 0.2) 10. 2)-Type Equations 10. . (10. A + Bxn α + βxn + γxn−1 . 1.3) Bxn + Cxn−1 Please recall our classification convention in which all parameters that appear in these equations are positive. .1) (10. 1. . 2)-type: xn+1 = xn+1 = and α + βxn + γxn−1 . A2 q= β . 1. . 2)-type Eq(10. . n = 0. xn+1 = n = 0. . . and the denominators are always positive. . 1. .Chapter 10 (3. (10. . the initial conditions are nonnegative. γ . A (10.1 Introduction α + βxn + γxn−1 . .

4) was investigated in [29]. 2)-Type Equations (Eq(10. .2. When r =q+1 Eq(10. Then the equilibrium of Eq(10.7. . By applying the linearized stability Theorem 1.2.) The linearized equation of Eq(10.4) exhibit the following trichotomy character. Then every solution of Eq(10. for a more general equation. See Sections 2. 2 (1 + y) 1+y n = 0.5 and 2.4) is globally asymptotically stable. (c) Assume that r > q + 1.4) converges to a period-two solution. (b) Assume that r < q + 1. The proofs of parts (b) and (c) of Theorem 10.1 are given in Sections 10.6) (10. .7. The main result in this section is that the solutions of Eq(10. in Section 2. 1.1 (a) Assume that r = q + 1.1 and 10.168 Chapter 10.1 was given. Theorem 10.5) .4) possesses unbounded solutions.4) possesses prime period-two solutions. Then Eq(10. (3.2.1 we see that y is locally asymptotically stable when r <1+q and is an unstable (saddle point) equilibrium when r > 1 + q.7) (10. .1.2.2. (10. The proof of part (a) of Theorem 10.2 below.4) about its unique equilibrium y= is zn+1 − q+r−1+ (q + r − 1)2 + 4p 2 q − p − ry r zn − zn−1 = 0.

4) to the difference equation zn+1 = (p + rq − q) + rzn−1 .5 and therefore the proof is complete when (10.12) (10. (10. 1. 2)-type Eq(6.2. We know already that when Eq(10. (1 + q) + zn n = 1.8) which is of the form of the (2. .10) reduces to yn+1 − q = Also in this case 0<r =1− r (yn − q) for n ≥ 0. (10. Now observe that the solutions of Eq(10.9) for n ≥ 1.4) is attracted to the equilibrium. The Case C = 0 : xn+1 = α+βxn +γxn−1 A+Bxn 169 10. (10.10) . So in the remaining part of the proof we assume that p < q.2.6) holds and we prove that the equilibrium of Eq(10. 1 + yn p <1 q (10.10. . 1 + yn and so when p≥q we have yn ≥ q Therefore the change of variables yn = q + zn transforms Eq(10. n = 0.2.4) satisfy the following identity yn+1 − q = Note that when q= that is when p + qr − q = 0 the Identity (10. Observe that (p − q) + ryn−1 yn+1 = q + . .6) holds the equilibrium is locally asymptotically stable so it remains to be shown that every solution of Eq(10.1 Here we assume that Eq(10. .11) q−p r yn−1 − 1 + yn r q−p r for n ≥ 0.23) which was investigated in Section 6. . .4) is globally asymptotically stable.1 Proof of Part (b) of Theorem 10.8) holds.

17) q−p r for some N ≥ 0.10) that y2n+2 − q = and so y2n+2 > ry2n + p for n ≥ 0. (10. q−p r q−p r for n ≥ 0 for n ≥ 0. The case where (10. (10.17) holds is similar and will be omitted.170 and the equilibrium of Eq(10. To this end. note that now q−p <q r and employ (10. To complete the proof in this case it is sufficient to show that eventually (10.15) is not eventually true then either y2n < or y2n−1 < This is because when yN −1 > then it follows from (10. (10.15) yn ≥ q and then use the known result for Eq(10.4) is equal to q.12) that Chapter 10. Still to be considered are cases where p + qr − q > 0 and p + qr − q < 0.18) q−p q−p r y2n − > r y2n − 1 + yn+1 r r .10). If (10.9). 2)-Type Equations |yn+1 − q| < r|yn+1 − q| and so n→∞ lim yn = q which completes the proof when (10.10) that yN +1+2K > q for all K ≥ 0. (3.16) holds. Then we can see from (10.13) (10.16) (10. It follows from (10.14) First assume that (10.11) holds. We will assume that (10.13) holds.

The case where (10.14) and completes the proof of part (b) of Theorem 10. it follows from (10.16) the solution is bounded.20) holds is similar and will be omitted. . Then one can see from (10.10) we see that y2n+2 − q = and so y2n+2 < ry2n + p for n ≥ 0.13) and completes the proof when (10. Once we establish this result. the proof that the r equilibrium is a global attractor of all solutions would be a consequence of Theorem 1. .20) r We will assume that (10. y2n−1 > (10. because by (10.10) and r the fact that now q−p q< . r that either q−p y2n > for n ≥ 0 (10.8 and the fact that in this case the function f (x. To this end assume for the sake of contradiction that the solution is not eventually in the interval 0.4. Next assume that (10. The Case C = 0 : xn+1 = α+βxn +γxn−1 A+Bxn 171 This is impossible when r ≥ 1. r 1−r This contradicts (10.16) and (10. . Then from (10.10.2. Now we claim that every solution of Eq(10. .18) that q−p > y2n+2 > rn+1 y0 + rn p + .2. q−p r q−p < r y2n − y2n − 1 + y2n+1 r r . + rp + p r and so p q−p ≥ r 1−r which contradicts (10.1. y) = p + qx + ry 1+x is increasing in both arguments.4) lies eventually in the interval 0. q−p . when r < 1. Note that in this case r < 1 and so q−p < y2n+2 < rn+1 y0 + rn p + .13) holds. On the other hand.19) r or q−p for n ≥ 0.19) holds.14) holds. q−p . + rp + p r which implies that p q−p ≤ .

5 for appropriate initial conditions y−1 and y0 . ∞).4) with initial conditions such that q ≤ y−1 < r − 1 and y0 > r−1 + p + q(r − 1) r−1−q for n ≥ 0.4) possesses bounded solutions which in fact converge to the equilibrium of Eq(10. 1 + yn−1 q= β .3 The Case B = 0 : xn+1 = A yn C α+βxn +γxn−1 A+Cxn−1 This is the (3.2 Proof of Part (c) of Theorem 10. A n = 0.172 Chapter 10. 1.2. ∞) is invariant. 2)-Type Equations 10.7) holds and show that Eq(10.4) to Eq(10. A (10. 10. More specifically the solutions of Eq(10. Eq(10.2. .4). y0 ∈ [q.21) αC . That is when y−1 . It should be mentioned that when (10. To this end observe that yn+1 = q + q p + r(yn−1 − r ) 1 + yn for n ≥ 0 and so when (10. .2) which by the change of variables xn = reduces to the difference equation yn+1 = where p= p + qyn + ryn−1 . 2)-type Eq(10. A2 and r = .7) holds the interval [q.9) and the conclusion of part (c) follows from Section 6. in addition to unbounded solutions.4) possesses unbounded solutions. have the property that the subsequences of even terms converge to ∞ while the subsequences of odd terms converge to q. γ .4) is a saddle point and so by the stable manifold theorem. the equilibrium of Eq(10. then yn ≥ q Now the change of variables yn = q + zn transforms Eq(10.7) holds. . (3.1 Here we assume that (10.

1 Invariant Intervals The following result. and convergence of solutions in certain regions of initial conditions.10. The linearized equation of Eq(10. .3. which can be established by direct calculation. zn + 1+y (1 + y)2 n = 0. yn+1 = r + 1 + yn−1 and so when p≥r Global Stability When p ≥ r .3. Here we make some simple observations about linearized stability. Observe that (p − r) + qyn . 10.1. n = 0. r−p q when q < 1 and p ≥ r(1 − q). . yN . .3. q. . The Case B = 0 : xn+1 = α+βxn +γxn−1 A+Cxn−1 173 This equation has not been investigated yet. invariant intervals. gives a list of invariant intervals for Eq(10. p 1−q when q < 1 and p ≥ r. Lemma 10. and r.3. Recall that I is an invariant interval.21) possesses the following invariant intervals: (a) 0. (b) 0.2 Here we discuss the behavior of the solutions of Eq(10.21) when p ≥ r. if whenever.21) about its positive equilibrium y= is zn+1 − q+r−1+ (q + r − 1)2 + 4p 2 q p − r + qy zn = 0. 1. By applying Theorem 1.1 Eq(10.21).1 we see that y is locally asymptotically stable for all values of the parameters p. . 10. yN +1 ∈ I then yn ∈ I for n ≥ N. 1. . . for some integer N ≥ 0.

n = 1. (b) The positive equilibrium of Eq(10. Theorem 10.2) which was investigated in Section 6. which is of the form of the (2. .2 (a) Assume that q < 1 and r ≥ p . 10. . converges to the equilibrium y. .174 yn ≥ r Therefore the change of variables yn = r + zn transforms Eq(10.3. 2)-type Eq(6. (1 + r) + zn−1 Chapter 10.21) is globally asymptotically stable if one of the following conditions holds: (i) q < 1.3.3. (a) Every positive solution of Eq(10. Theorem 10.21) is bounded.3. 2)-Type Equations for n ≥ 1. The following result is now a consequence of the above observations and the results in Section 6.21) when p < r. (3.3 Convergence of Solutions Here we discuss the behavior of the solutions of Eq(10.21) with initial conditions in the invariant interval 0.3. (ii) q≥1 and either p + qr − r ≤ q or q < p + qr − r ≤ 2(q + 1). Then the following statements are true. 2.21) to the difference equation zn+1 = p + qr − r + qzn . r−p q . 1−q Then every solution of Eq(10.1 Assume p ≥ r.

K q the function f (x.K q converges to the equilibrium y.4. Proof. . γ2 (10. The result is now a consequence of Theorem 1. 2 10. The result is now a consequence of Theorem 1. 1−q and p < r < Then every solution of Eq(10. K= q + r − p − q2 (b) In view of Lemma 10.10. r−p . y) increases in x and decreases in y.1 we see that when y−1 . r−p and that in the interval 0. y0 ∈ 0.1 we see that when y−1 .3. r−p q q for all n ≥ 0. 2)-type Eq(10. then yn ∈ 0.3) which by the change of variables xn = reduces to the difference equation yn+1 = where p= r + pyn + yn−1 . y) increases in both x and y.K q r−p . αC . r−p q q the function f (x. It is easy to check that y ∈ r−p .21) with initial conditions in the invariant interval r−p . It is easy to check that y ∈ 0. (a) In view of Lemma 10.4.4 The Case A = 0 : xn+1 = γ yn C α+βxn +γxn−1 Bxn +Cxn−1 This is the (3. C n = 0.5. 1. qyn + yn−1 q= B .K q pq + r2 − pr . y0 ∈ for all n ≥ 0.22) β . then yn ∈ r−p . . The Case A = 0 : xn+1 = (b) Suppose that either or Set α+βxn +γxn−1 Bxn +Cxn−1 175 0≤q−1<r q<1 and p < r p . .3.8. γ and r = .K q and that in the interval .4. r−p .

The linearized equation of Eq(10. Concerning prime period-two solution. Here we make some simple observations about linearized stability.25) Furthermore in this case the prime period-two solution . When r = 0. .1. q−1 . .5 that a period-two solution exists if and only if p < 1.23) It is interesting to note that the above condition for the local asymptotic stability of y requires the assumption that r > 0. is unique and the values φ and ψ are the positive roots of the quadratic equation t2 − (1 − p)t + r + p(1 − p) = 0. (10.9. φ. the inequality in (10. as in Section 6.1 we see that y is locally asymptotically stable when either q ≤ pq + 1 + 3p or q > pq + 1 + 3p and F where F (u) = (q + 1)u2 − (p + 1)u − r.24) 2r q − pq − 1 − 3p >0 (10. it follows from Section 2. The second condition reduces to q > pq + 1 + 3p. . and the convergence of solutions in certain regions of initial conditions.24) is void.176 Chapter 10. (3. . 2)-Type Equations This equation has not been investigated yet. . (1 + p) + (1 + p)2 + 4r(1 + q) 2(1 + q) By applying Theorem 1. . . . 1.23) is strict and Condition (10. ψ. (q + 1)(r + (p + 1)y) (q + 1)(r + (p + 1)y) n = 0. φ. q > 1 and 4r < (1 − p)(q − pq − 3p − 1). invariant intervals. . ψ.22) about its positive equilibrium y= is zn+1 − (p − q)y − qr (p − q)y + r zn + zn−1 = 0. . (10.

q2 p2 − pq > r. r + px + y qx + y (b) Clearly the function f (x. (b) p qr .4.26) when p > q and p2 − pq < r.4.1 q−p Proof. p > q. y) = is decreasing in both arguments. (q + 1)b (q + 1)a qr . Hence a≤ r + (p + 1)b r + (p + 1)a = f (b. Lemma 10. q2 when p > q and when p < q < p + r. p−q and (10. q p−q qr p . y) is decreasing in both arguments when y < qr is increasing in x and decreasing in y for y ≥ p−q . . b) ≤ f (x.21) possesses the following invariant intervals: (a) [a. p−q q (c) 1. The Case A = 0 : xn+1 = α+βxn +γxn−1 Bxn +Cxn−1 177 10. r q−p r . when q > p + r. b] when p = q a and b are positive numbers such that r + (p + 1)a ≤ (q + 1)ab ≤ r + (p + 1)b. and p < q.21).4. y) ≤ f (a.10. We consider the cases where p = q.1 Eq(10. (a) It is easy to see that when p = q the function f (x. a) = ≤ b.1 Invariant Intervals Here we present some results about invariant intervals for Eq(10. First assume that p > q and p2 −pq q2 and it < r.

q+1 q−p r+p+1 r r < and q−p q+1 q−p are equivalent to the inequality q < p + r. q q = qr qr + p(p + 1) ≤ . y ∈ p =f q p qr . Using the decreasing character of f . y) ≤ f r . p−q q p2 −pq q2 and p qr < q p−q < r. p2 −pq q2 > r.178 Then for x. we obtain 1=f The inequalities r r . Next assume that q > p + r. p−q . First. we obtain r (p + r)(q − p) + r = f 1. . assume that p < q < p + r. q p−q p = . p(q + 1) p−q qr qr . (3. Using the decreasing character of f in x and the increasing character of f in y. q−p (c) It is clear that the function f (x. 2)-Type Equations we obtain ≤ f (x. q(q − p) + r q−p ≤ f (x. p−q q ≤ f (x. 1) = r+p+1 r ≤ . q p−q Chapter 10. we obtain qr p . p−q p−q The inequalities qr + p(p + 1) qr ≤ p(q + 1) p−q are equivalent to the inequality Next assume that p > q and For x. y ∈ qr p . y) ≤ f p p . y) ≤ f p qr . r . y) ≤ f (1. q (qr + p)(p − q) + pq 2 r =f q 3 r + p(p − q) The inequalities (qr + p)(p − q) + pq 2 r qr ≥ 3 r + p(p − q) q p−q follow from the inequality p2 −pq q2 and qr p > q p−q > r. 1 = 1. q−p q−p 1< ≤ f (x. y) is decreasing in both arguments for x < r and it is increasing in y and decreasing in x for x ≥ q−p .

The proof is an immediate consequence of Lemma 10.4.3 Assume that p < q. where 0 < a < b satisfy (10. Theorem 10. q−p converges to the equilibrium y.1 and Theorem 1. q2 and that either p≤1 or p > 1 and (q − 1)[(p − 1)2 (q − 1) − 4q(1 − p − qr)] ≤ 0. q−p Proof. 2 .4.22) with initial conditions r in the invariant interval 1.1 and Theorem 1.2 Convergence of Solutions Here we obtain some convergence results for Eq(10. q2 Then every solution of Eq(10. The Case A = 0 : xn+1 = The inequalities α+βxn +γxn−1 Bxn +Cxn−1 179 (p + r)(q − p) + r r ≥ q(q − p) + r q−p follow from the inequality q > p + r.4. (c) Assume that p < q < p+r.4. 1 converges to the equilibrium y.1 and Theorem 1.4.4. 2 Theorem 10.22) with initial conditions in the invariant interval r . The proof is an immediate consequence of Lemma 10.4. q p−q conditions in the invariant interval converges to the equilibrium y.7. Then every solution of Eq(10. The proof is an immediate consequence of Lemma 10.22) with initial p qr .4. (b) Assume that p > q and r > p2 −pq .4. p + r < q.4.10.1 (a) Assume that p = q. b]. and that Condition (10. qr p .6.22). Then every solution of Eq(10. Proof. 2 Theorem 10.4.26) converges to the equilibrium y. r < p2 −pq .5. Then every solution of Eq(10. Then every solution of Eq(10. and r <1 q−p 2 10.22) with initial conditions in the invariant interval [a.25) is not satisfied.22) with initial conditions in the invariant interval converges to the equilibrium y.2 Assume that p > q. p−q q Proof.

ψ.28) 1 + xn 1 + xn + xn−1 xn+1 = (10.5.5. φ.27). . . (10. ∞) such that the ∞ solutions {yn }n=−1 of Eq(10. ψ.2 Assume r > q + 1.29) 1 + xn−1 1 + xn + xn−1 .180 Chapter 10. y0 ) ∈ (0. (a) Determine the set of initial conditions y−1 and y0 for which φ = ψ.4) are bounded.30) xn+1 = xn + xn−1 Extend and generalize. (3. Open Problem 10. .4).5.27) (b) Assume (10. Open Problem 10.3 For each of the equations listed below. determine the set G of all initial conditions (x−1 .4) converges to a not necessarily prime period-two solution . . Investigate the asymptotic behavior of {yn }n=−1 . ∞ (b) Let (y−1 . . . Conjecture 10. x0 ) ∈ R × R through which the equation is well defined for all n ≥ 0.5. Determine the set of initial conditions y−1 and y0 for which {yn }∞ n=−1 converges to (10. 2)-Type Equations 10. (10. if and only if r = 1 + q. φ. .21) converges to the positive equilibrium of the equation.5 Open Problems and Conjectures ∞ Open Problem 10. and for these initial points investigate the long-term behavior of the corresponding solution: 1 + xn + xn−1 xn+1 = (10.1 Every positive solution of Eq(10.1 We know that every solution {yn }n=−1 of Eq(10. y0 ) ∈ B.27) is a prime period-two solution of Eq(10. (a) Find the set B of all initial conditions (y−1 . ∞) × (0.

. q. 1 + yn−1 rn + pn yn + yn−1 .31) yn+1 = yn+1 = n = 0.22) (a) Local stability of the positive equilibrium implies global stability.5. (10. {qn }∞ . . n = 0. .5.5. .32) (10. lim lim and r = n→∞ rn . Investigate the asymptotic behavior and the periodic nature of all positive solutions of each of the following three difference equations: yn+1 = pn + qn yn + rn yn−1 . the equilibrium y is a saddle point. but not globally. r ∈ (0. Then the following statements are true for Eq(10.22) possesses a period-two solution. . ∞ Open Problem 10. . and {rn }n=0 are convergent sen=0 n=0 quences of nonnegative real numbers with finite limits. (10.22). 1 + yn pn + qn yn + rn yn−1 .10. . . Open Problems and Conjectures Conjecture 10. ∞). 181 (b) When Eq(10. (c) When Eq(10. (d) The period-two solution of Eq(10. 1. 1. Investigate the global character of all positive solutions of Eqs(10. and {rn }∞ are period-two sen=0 n=0 n=0 quences of nonnegative real numbers.2 Assume that p. n→∞ q = n→∞ qn .22) has no prime period-two solutions the equilibrium y is globally asymptotically stable.5. . .31)-(10.33). p = lim pn . Extend and generalize. is locally asymptotically stable. when it exists.33) Open Problem 10.5 Assume that {pn }∞ . qn yn + yn−1 n = 0. . 1.4 Assume that {pn }∞ . {qn }∞ .

3. n = 0. Investigate the global behavior of all positive solutions of each of the following difference equations: yn+1 = yn+1 = yn+1 = p + qyn + ryn−k . . 2)-Type Equations p. . .5.182 Open Problem 10. .36) . 1 + yn p + qyn + ryn−k . q. 1. . . . (3.}. ∞) and k ∈ {2. . 1.34) (10. qyn + yn−k n = 0. 1. . r ∈ [0. 1 + yn−k p + qyn + ryn−1 . . n = 0.35) (10. . (10. . .6 Assume that Chapter 10.

A + Bxn + Cxn−1 n = 0. . − z 2 + α(B + C) + (B + C)(A + β + γ)x n−1 A A2 183 n = 0. . . . 11.1) Please recall our classification convention in which all coefficients of this equation are now assumed to be positive and the initial conditions nonnegative.3) zn+1 − (βA − Bα) + (βC − Bγ)x zn + α(B + C) + (B + C)(A + β + γ)x (γA − Cα) − (βC − γB)x = 0. invariant intervals. . . and the convergence of solutions in certain regions of initial conditions. (11. . Eq(11. 3)-type equation xn+1 = α + βxn + γxn−1 . 1. The linearized equation about x is (see Section 2. .1) has a unique equilibrium which is the positive solution of the quadratic equation That is. 3)-Type Equation α+βx xn+1 = A+Bxn+γxn−1 n+Cx n−1 In this chapter we discuss the (3.1 Linearized Stability Analysis (B + C)x2 − (β + γ − A)x − α = 0. This equation has not been investigated yet. x= β+γ−A+ (A − β − γ)2 + 4α(B + C) 2(B + C) . 1.Chapter 11 The (3. (11. Here we make some simple observations about linearized stability.2) .

ψ.3).1). (11. . see Section 2. φ. and (2. (2.4) The local asymptotic stability of the two cycle 11. yN +1 ∈ I then for some integer N ≥ 0. C C(B − C) .14).184 Chapter 11. Conditions (2. ψ. φ. holds the values of φ and ψ are the positive roots of the quadratic equation t2 + γ−β−A αC + β(γ − β − A) t+ = 0.1) if whenever.4.2 Invariant Intervals Here we obtain several invariant intervals for Eq(11.1) by testing whether the hypotheses of one of the four Theorems 1.15) are satisfied.13).5-1. The (3.1) has a prime period-two solution . .4.3) 4C 2 Furthermore. . for n ≥ N. . that is. Eq(11. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 The equilibrium x is locally asymptotically stable if Conditions (2. φ. B > C. and α < (γ − β − A)[B(γ − β − A) − C(γ + 3β − A)] . γ > β+A.6.5. then clearly fx = M = βC − γB. if.31) and (2. yN . If we set L = βA − αB. φ. y) = is increasing or decreasing in x and y. . . . when (11. .1) then we may be able to obtain a convergence result for the solutions of Eq(11. . and N = γA − αC α + βx + γy A + Bx + Cy L + My N − My and fy = .32) are satisfied. These intervals are derived by analyzing the intervals where the function f (x.8 are satisfied. is discussed in Section 2. ψ. . (11. Once we have an invariant interval for Eq(11. 2 (A + Bx + Cy) (A + Bx + Cy)2 Recall that an interval I is an invariant interval for Eq(11. ψ. . . yn ∈ I .

1) and in this interval the function f (x. C B A β α Then B . C B A 185 α β Then A . y) is decreasing in both variables.2. y) is increasing in x and decreasing in y. A B C α Then 0.1) and in this interval the function f (x.1) and in this interval the function f (x.1) and in this interval the function f (x. A is an invariant interval for Eq(11. (vi) Assume that β γ α γA − αC .1) and in this interval the function f (x. y) is increasing in both variables. A is an invariant interval for Eq(11. (ii) Assume that β γ α > > B C A and γ N ≤ . B+C βC − γB α β+γ Then A . y) is decreasing in both variables. B+C is an invariant interval for Eq(11.1) and in this interval the function f (x. (v) Assume that α β γ γA − αC > . A is an invariant interval for Eq(11. (iv) Assume that β α γ = < . y) is increasing in both variables.1 (i) Assume that β α γ = > . y) is decreasing in both variables.11. (vii) Assume that α β γ = > . < < < B C A βC − γB α Then 0. βC−γB is an invariant interval for Eq(11. (iii) Assume that β α γ > > C B A and β+γ αB − βA ≤ . C is an invariant interval for Eq(11.1) and in this interval the function f (x. B is an invariant interval for Eq(11. Invariant Intervals Lemma 11. C M α γ Then A . ≥ > βC − γB A B C γA−αC Then 0. .2. y) is increasing in both variables.

(x) Assume that β α γ < ≤ .186 (viii) Assume that Chapter 11. 0) ≤ f (x. for every x. A C . u) ≤ β β+γ = . this function γ N N is increasing in y for x < M . B+C B γ β α (ii) The condition B > C > A is equivalent to M > 0. C is an invariant interval for Eq(11. y) ≤ f ( . The (3. γ β α (i) The condition C = B > A is equivalent to M = 0 and L > 0.1) and in this interval the function f (x. B is an invariant interval for Eq(11. y ∈ α γ . in this case the function f is increasing in x. B A C β Then 0. Thus. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 β α γ > ≥ . L > 0. in this case the function f is increasing in both variables for all values of x and y and α β+γ β β β = .1) and in this interval the function f (x. A B C γ Then 0. 0) ≤ f (x. (ix) Assume that α β γ = < . ) ≤ A B B B+C B Here we used the fact that under the condition γ β α = > C B A the function f (u. y) is increasing in y and decreasing in x. y) ≤ f ( . B A C γ Then 0. ) = . u) is increasing and consequently f (u. In addition. . = f (0.1) and in this interval the function f (x. A M M C for all x. assuming C ≤ M we obtain N N γ α = f (0. y) is increasing in y and decreasing in x. C is an invariant interval for Eq(11. which in turn imply that N > 0. Thus. Proof. and N > 0. y) is increasing in x and decreasing in y. Thus.

y ∈ 0. y) ≤ f A for all x. which in turn imply that N > 0. which imply that N < 0. In β+γ L L addition. y ∈ 0.11. Thus. 0) ≤ f (x. A . for every y. Thus. Thus assuming B+C ≤ − M . y) ≤ f (0. in this case the function f is increasing in y. A M γ β α (vi) The condition A > C > B is equivalent to M < 0. 0) = . α L L this function is decreasing in x for y < − M . Thus. y) ≤ f (0. 0) = α for all x. α L ≤− . B A A A Here we used the fact that under the condition β α γ = < C B A the function f (u. y) ≤ f (0. L < 0. A . B+C . 0) = L for all x. y ∈ α β+γ . γ β α (iv) The condition C = B < A is equivalent to M = 0 and L < 0. we obtain. in this case the function f is decreasing in x. Thus assuming A ≤ − M . Invariant Intervals 187 γ β α (iii) The condition C > B > A is equivalent to M < 0 and L > 0. for every y. u) is decreasing and consequently f (u. in this case the function f is decreasing in y. u) ≥ β+γ β = . this function α N N is decreasing in y for x < M . ≤ f (x. α = f (0.2. − M . for every x. In addition. in this case the function f is decreasing in both variables for all values of x and y and α α β α ≤f . we obtain 0 ≤ f (x. A B+C β+γ β+γ . B+C B+C ≤ β+γ . B+C B β γ α (v) The condition A > B > C is equivalent to M > 0 and L < 0. Thus assuming A ≤ M . Thus. In addition. which in turn imply that N < 0. α . we obtain 0 ≤ f (x. this function is increasing in x for y < − M . and N < 0.

in this case the function f is increasing in x and decreasing in y for all values of x and y. y) < f 0. y ∈ 0. v). B B β γ α (ix) The condition A = B < C is equivalent to M < 0 and L = 0. in this case the function f is decreasing in x and increasing in y for all values of x and y. α α . 0 ≤ f (x. v) is increasing in v when N > 0 and so γ f (0. which in turn imply that N > 0. C v→∞ for all v. 0 ≤ f (x. C . 0 ≤ f (x. A . γ C = γ . Using this we obtain. C . Using this we obtain. and N < 0. y ∈ 0. y ∈ 0. . y) ≤ f α for all x. Using this we obtain. 0 ≤ f (x. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 β γ α (vii) The condition A = B > C is equivalent to M > 0 and L = 0. which in turn imply that N < 0. B . β β . in this case the function f is increasing in x and decreasing in y for all values of x and y. 2 The next table summarizes more complicated invariant intervals with the signs of the partial derivatives fx and fy . y) < f β for all x. L > 0.0 = . y ∈ 0. γ C ≤ γ . C γ β α (x) The condition B < A ≤ C is equivalent to M < 0. and N ≥ 0.188 Chapter 11. Thus.0 = . The (3. Using this we obtain. Thus. Thus. Thus. γ for all x. in this case the function f is increasing in x and decreasing in y for all values of x and y. L < 0. C γ for all x. v) ≤ = lim f (0. A A β γ α (viii) The condition B > A > C is equivalent to M > 0. Here we use the fact that the function f (0. y) ≤ f 0.

K1 ] β−A+ Signs of derivatives N M [0. − M ] if AM −(B+C)L < − M βM K+αM L L [− M .3.5-1. Then in each of the following cases lim xn = x.4.3.1 and Theorems 1.11. fx < 0 if y > − M fy > 0 fx > 0 fy < 0 fx < 0 fy > 0 L L fx > 0 if y > − M . M ] if AM +(B+C)N < αM +βK+γN N N [ M . Convergence Results 189 Case 1 2 3 4 5 6 7 8 9 Invariant intervals αM +(β+γ)N N [0. K2 ] if BM K+AM −γL > − M −CL β+γ−A+ β [0. M ] if AM +(B+C)N < N [ M . n→∞ . fy < 0 if x > N M (β−A)2 +4Bα ] 2B (β+γ−A)2 +4α(B+C) ] 2(B+C) αM −(β+γ)L L L [0.4. A ] √ L L fx > 0 if y < − M . fy < 0 if x < M αM +(β+γ)N αM +(β+γ)N N [ AM +(B+C)N . Theorem 11. . . − M ] if AM −L(B+C) < − M αM −βL+γM L L [− M .2.8. . K4 ] if AM +BK+CNM > M M N M where K1 = K2 = K3 = K4 = (β − A)M − CN + (A − γ)M − BL + (β − A)M + LC + (γ − A)M − BN + ((β − A)M − N C)2 + 4BM (αM + γN ) 2BM ((A − γ)M − BL)2 + 4CM (αM − βL) ((β − A)M + LC)2 + 4BM (αM − γL) 2BM 2CM −2CM . [0.1). ((γ − A)M − BN )2 + 4BM (αM + βN ) 11. fx < 0 if y < − M fy < 0 fx < 0 fy < 0 fx < 0 N N fy > 0 if x > M . C ] αM −L(β+γ) αM −L(β+γ) L L [ AM −L(B+C) .1 Let {xn } be a solution of Eq(11.3 Convergence Results The following result is a consequence of Lemma 11. K3 ] if AM −BL+CMK > − M K α [0. B ] γ [0. if √ αM +(β+γ)N AM +(B+C)N > N M fx > 0 fy > 0 if x < fx > 0 fy < 0 fx > 0 fy > 0 N M .

A C α β+γ . A β γ α γA − αC < < < B C A βC − γB and x−1 . B+C βC − γB and x−1 . . B A and x−1 . β γ α = > . B < C. B C A (iii) γ β α > > . x0 ∈ 0. and or (C − B)[(C − B)(β − γ − A)2 − 4B(αB + γ(β − γ − A))] > 0. A B+C γ γA − αC ≤ . . C βC − γB and x−1 . B < C. γA − αC . . A B C β (viii) At least one of the following three conditions is not satisfied and x−1 . β α γ > ≥ . x0 ∈ β α . βC − γB α . B A C . C B A (iv) γ β α = < C B A (v) and x−1 . x0 ∈ α β γ γA − αC ≥ > > βC − γB A B C (vi) and x−1 . B : β > γ + A. x0 ∈ αB − βA β+γ ≤ . and or (C − B)[(C − B)(β − γ − A)2 − 4B(αB + γ(β − γ − A))] > 0. x0 ∈ α β . A B α γ . x0 ∈ 0. x0 ∈ 0. The (3. α (vii) At least one of the following three conditions is not satisfied and x−1 . . x0 ∈ 0. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 γ β α = > C B A (ii) β γ α > > .190 (i) Chapter 11. A : β > γ + A.

M = f (M.1) is globally asymptotically stable. . α γ β < ≤ B A C and x−1 . m= α + βm + γM . x0 ∈ 0.4.3.2.3 to Eq(11. The proofs of statements (ix) and (x) follow by applying Lemma 11.4.3. By applying Theorem 1. β γ α = < A B C (x) Condition (11. C 191 Proof.11.8. m).5. The proofs of statements (iv-vi) follow by applying Lemma 11. and by observing that Eq(11. M ).1 (i-iii) and Theorem 1.2.3) is not satisfied.2.7. The proofs of statements (i-iii) follow by applying Lemma 11. 2 11.6. . γ .3.4.1 Global Stability Theorem 1.1) we obtain the following global asymptotic result.1 (vii) and (viii) respectively.3) is not satisfied. Theorem 1. that is.4.2 and linearized stability imply the following global stability result: Theorem 11.4. x0 ∈ 0. Theorem 1.2 Assume that γ β α < min . C and x−1 . γ . The proofs of statements (vii) and (viii) follow by applying Lemma 11.1 (iv-vi) and Theorem 1.4.2.1) does not possess a prime period-two solution. Convergence Results (ix) Condition (11. and by observing that the only solution of the system of equations m = f (m. A + Bm + CM M= α + βM + γm A + BM + Cm is m = M .1 (ix) and (x) respectively. C B A Then the equilibrium x of Eq(11.

B. βA ≥ αB. γ. Theorem 11.3. C ∈ (0. B.1) is globally asymptotically stable. A.4 Open Problems and Conjectures α. and 3γB ≥ βC ≥ γB. A. C ∈ (0. β. Show that every positive solution of Eq(11.1) has a positive prime period-two solution.1 Assume Show that every positive solution of Eq(11. and 3βC ≥ γB ≥ βC. A. γB > βC.1) is bounded. Conjecture 11.3 Assume that B ≤ C. B.1) is globally asymptotically stable. (b) Assume that αB ≥ βA.2 Assume α.4. The (3. and γA > αC. αB > βA. ∞).4.4.4.1) is a saddle point and the period-two solution is locally asymptotically stable. Then the equilibrium x of Eq(11. C ∈ (0. ∞) and that Eq(11. ∞) and that Eq(11. γ. Conjecture 11. 11.4.192 Chapter 11. Conjecture 11. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Theorem 11.1) has no positive prime period-two solution. β.1) converges to the positive equilibrium.1) is globally asymptotically stable. γ.4 (a) Assume that αC ≥ γA. β. Show that the positive equilibrium of Eq(11.3. Then the equilibrium x of Eq(11. γA ≥ αC.3 Assume α. Then the equilibrium x of Eq(11. The following global stability result is a consequence of linearized stability and the Trichotomy Theorem 1. .

A. . In particular.11. 193 ∞ ∞ Open Problem 11. n = 0. ai .2 Assume that the parameters in Eq(11. .1) cannot have a prime period-k solution for any k ≥ 3. ∞) for i = 0. a + a0 xn + .7) is periodic with period-k. . ∞). . investigate the cases where 2 ≤ k ≤ 8. Ai ∈ [0.4 Assume that (11.1 Assume {αn }n=0 . β.4.4. . C ∈ (0.6) Obtain necessary and sufficient conditions so that every positive solution of the equation xn+1 = has a finite limit.4 Assume α. .5) An + Bn xn + Cn xn−1 Open Problem 11. k ≥ 2. A + A0 xn + . . 1.7) converges to a period-two solution. Obtain necessary and sufficient conditions so that every positive solution of Eq(11. . Obtain necessary and sufficient conditions so that every positive solution of Eq(11. 1. {γn }∞ . Open Problem 11. {Bn }∞ . B. Show that Eq(11. .7) Open Problem 11. 1. .4.5). Open Problem 11.5) are period-two sequences of positive real numbers. . n = 0. A. {An }n=0 . + Ak xn−k (11.6) holds. Open Problems and Conjectures Conjecture 11. . {Cn }∞ n=0 n=0 n=0 n=0 are convergent sequences of nonnegative real numbers with finite limits.4. .6) holds. Investigate the asymptotic behavior and the periodic nature of all positive solutions of Eq(11. . . γ. .4. .3 Assume that a. {βn }∞ . k.5 Assume that (11.4. xn+1 = (11. (11. Investigate the asymptotic behavior and the periodic nature of all positive solutions of the difference equation αn + βn xn + γn xn−1 . + ak xn−k .4.

1.) Conjecture 11. xn xn−1 + xn−2 + .4. the periodic nature. The (3. and [60]. (See [8]. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Open Problem 11.5 Let k ≥ 2 be a positive integer. . Show that every positive solution of the difference equation xn+1 = xn + .4.7 Assume that p.194 Chapter 11. . Extend and generalize. .8) converges to a period-two solution. . . n = 0. Show that every positive solution of the system xn+1 = yn+1 = zn+1 = a11 xn a21 xn a31 xn + + + a12 yn a22 yn a32 yn + + + a13 zn a23 zn a33 zn                . ∞) with γ + δ > 0. . 3}.4. . β. n = 0. + xn−k−2 converges to 1. . and the asymptotic behavior of all positive solutions of the difference equation xn+1 = (See [8] and [61]. + xn−k + xn−k−1 xn−k−2 . γxn xn−1 + δxn−2 Conjecture 11. Investigate the boundedness character. j ∈ {1. [47].6 Let α. δ ∈ [0. 1. q ∈ [0. (See [32] for the two dimensional case.4. . n = 0. . ∞) for i. (11. . ∞). 2. . Extend to equations with real parameters. γ.6 Let aij ∈ (0. . .) αxn + βxn−1 xn−2 . . 1.) Conjecture 11.

converges to a period-two solution if p ≥ 1. . .4.9 Assume that p.) Conjecture 11.4. (b) Show that when p<1 the positive equilibrium x = 1 1+p is globally asymptotically stable. (a) Show that every positive solution of the equation xn+1 = xn−1 . . q + xn−2 n = 0.8 Assume that p ∈ (0. Open Problems and Conjectures (a) Show that every positive solution of the equation xn+1 = pxn−1 + xn−2 . q ∈ [0. 195 converges to a period-two solution if and only if p = 1 + q. . (c) Show that when p>1+q the equation possesses positive unbounded solutions. Conjecture 11.11.4. xn−1 + pxn−2 n = 0. . (See [62]. ∞). . ∞). . 1. 1. (b) Show that when p<1+q every positive solution has a finite limit.

3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 (a) Show that every positive solution of the equation xn+1 = xn−1 + p . .10 Show that every positive solution of the equation xn+1 = xn + xn−2 . . xn n = 0.196 Chapter 11.4. The (3. (iii) When p < 1 there exist positive unbounded solutions. (See [62]. . .11 Show that the difference equation xn+1 = p + xn−2 . . converges to a period-four solution. has the following trichotomy character: (i) When p > 1 every positive solution converges to the positive equilibrium. (See [62].) Conjecture 11. 1. (See [62]. . (b) Show that when q>1 the positive equilibrium of the equation is globally asymptotically stable.4. converges to a period-two solution if and only if q = 1. 1. (c) Show that when q<1 the equation possesses positive unbounded solutions. xn−1 n = 0. . (ii) When p = 1 every positive solution converges to a period-five solution. 1.) . xn−2 + q n = 0. . .) Conjecture 11.

Then one can show that Eq(11..9) possesses unbounded solutions? (See [6].10) (a) In addition to (11.4. φ.) . .12 Show that every positive solution of the equation xn+1 = 1 + xn . . φ φ ψ ψ Open Problem 11. . 1. where φ. What other conditions are needed so that every positive solution of Eq(11. What other conditions are needed so that Eq(11.) ψ 1 1 φ . .7 Consider the difference equation xn+1 = α + βxn + γxn−1 + δxn−2 . . (11. (See [62].9) with nonnegative parameters and nonnegative initial conditions such that B + D > 0 and A + Bxn + Dxn−2 > 0 for n ≥ 0. . ∞).9) converges to the positive equilibrium? (c) Assume γ > β + δ + A.. ψ. 1.. 197 converges to a period-six solution of the form .4.9) has prime period-two solutions if and only if γ = β + δ + A. .4. what other conditions are needed so that every positive solution of Eq(11. . Open Problems and Conjectures Conjecture 11. . n = 0. . . ψ ∈ (0.10). .11. xn−1 + xn−2 n = 0. A + Bxn + Dxn−2 (11.9) converges to a period-two solution? (b) Assume γ < β + δ + A.

xn−2 n = 0. xn−1 + xn−2 + xn−3 n = 0. converges to a period-two solution. . (a) Show that when q=r every positive solution converges to a period-four solution. 1. (See [6]. . . 1.4.16 Consider the difference equation xn+1 = where p ∈ [0. r ∈ (0.14 Show that every positive solution of the equation xn+1 = 1 + xn−1 . n = 0. converges to a period-two solution. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Conjecture 11.) Conjecture 11. .4. The (3.198 Chapter 11. . 1 + xn + xn−2 n = 0.4.13 Show that every positive solution of each of the two equations xn+1 = and xn+1 = 1 + xn−1 + xn−2 . p + qxn + rxn−2 . (See [6]. xn−1 n = 0. .) Conjecture 11.15 Show that every positive solution of the equation xn+1 = xn−2 . xn 1 + xn−1 + xn−2 . Conjecture 11. . 1. converges to a period-three solution. . . . . 1. . .4. ∞). (b) Show that when q>r every positive solution converges to the positive equilibrium. 1. . ∞) and q. . .

. 1 + xn + xn−3 n = 0. . xn n = 0. xn + xn−2 + xn−3 n = 0. Obtain necessary and sufficient conditions on the parameters A and B0 . xn−2 + xn−3 . 1. 1. . ∞) with i=0 Bi > 0. . . A + B0 xn + . Open Problems and Conjectures (c) Show that when q<r the equation possesses positive unbounded solutions.4.4.11. . Open Problem 11. 1. 1 + xn−3 . . . + Bk xn−k n = 0. xn−2 n = 0.) 199 Open Problem 11.8 Determine whether every positive solution of each of the following equations converges to a periodic solution of the corresponding equation: (a) xn+1 = (b) xn+1 = (c) xn+1 = (d) xn+1 = 1 + xn + xn−3 . . . . . B0 . . (See [62]. . . xn−1 . .9 Assume that k is a positive integer and that k A. B1 . 1. (See [6]. Bk ∈ [0. 1. . .4. Bk so that every positive solution of the difference equation xn+1 = xn−k . . B1 . . . .) . . converges to a period-(k + 1) solution of this equation. .

. Then we can easily see that every nonnegative solution of the equation xn+1 = xn−k .11). . . 1 + xn + . . .10 Assume that k is a positive integer.12) with φ ≥ 0. . Show that every positive solution of xn+1 = 1 + xn + xn−k . (11. . . 0.) Conjecture 11.4. The (3. the value of φ in (11. 1.11) possess a positive solution which converges to zero? (b) Determine. . . . xn−k+1 n = 0. x0 .4. k−terms (11.17 Let k be a positive integer. . x0 for which the ∞ corresponding solution {xn }n=−k of Eq(11.200 Chapter 11. .11) converges to a period-(k+1) solution of the form (11. (c) Determine the set of all nonnegative initial conditions x−k . φ. 0. .. + xn−k+1 n = 0. . . . . 1. in terms of the nonnegative initial conditions x−k . .11) converges to a period-(k + 1) solution of the form . . converges to a periodic solution of period 2k. .12) with a given φ ≥ 0 ? (See [6].. . .. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Open Problem 11. (a) Does Eq(11.12) which corresponds to the nonnegative solution {xn }∞ n=−k of Eq(11. . 0.

Then Eq(A. . y. x−2 ∈ I.0. b] be an interval of real numbers and assume that f : [a.1 Let [a.1) ∞ has a unique solution {xn }n=−2 . Then for every set of initial conditions x0 .8. (A.1) is to work in the regions where the function f (x. or non decreasing in all three. b] and every solution of Eq(A. .4.1) converges to x. . xn−2 ). Theorem A. z) is monotonic in its arguments. or non increasing in two arguments and non decreasing in the third.1) has a unique equilibrium x ∈ [a. y. the difference equation xn+1 = f (xn . or non increasing in one and non decreasing in the other two. (b) The equation f (x. Let I be some interval of real numbers and let f :I ×I ×I →I be a continuously differentiable function. has a unique solution in the interval [a. b]3 → [a. x. 1.5-1. y.4. x−1 . As we have seen in previous chapters. b]. x) = x. b] is a continuous function satisfying the following properties: (a) f (x. xn−1 . z) is non-decreasing in each of its arguments.Appendix A Global Attractivity for Higher Order Equations In this appendix we present some global attractivity results for higher order difference equations which are analogous to Theorems 1. We state the results for third order equations from which higher order extensions will become clear. n = 0. z) is non decreasing in all three arguments. 201 . In this regard there are eight possible scenarios depending on whether f (x. very often a good strategy for obtaining global attractivity results for Eq(A.

. and is nonincreasing in z ∈ [a. 2.1) has a unique equilibrium x ∈ [a. b] is a continuous function satisfying the following properties: (a) f (x. . M ) and M = f (M. Now observe that for each i ≥ 0.2 Let [a. y. ≤ mi ≤ . M ) ∈ [a. ≤ M1 ≤ M0 . .0. . mi−1 ). b] is a solution of the system m = f (m. M. M ).202 Proof. . b] × [a. m. . b] for each x and y ∈ [a. . from which the result follows. M. m = M = x. . . i→∞ and M = lim Mi . In view of (b). Set m0 = a and M0 = b and for i = 1.1) converges to x. 2 Theorem A. b] for each z ∈ [a. m. ≤ Mi ≤ . m0 ≤ m1 ≤ . b]3 → [a.2) Then M ≥ lim sup xi ≥ lim inf xi ≥ m i→∞ i→∞ and by the continuity of f . set Mi = f (Mi−1 . Mi−1 ) and mi = f (mi−1 . b] and every solution of Eq(A. m) then m = M. b]. . Then Eq(A. z) is non-decreasing in x and y ∈ [a. and mi ≤ xk ≤ Mi Set m = lim mi i→∞ Appendix for k ≥ 3i + 1. (b) If (m. Mi−1 . mi−1 . b] be an interval of real numbers and assume that f : [a. (A. m) and M = f (M. m = f (m. b].

m). m) then m = M. Mi−1 . . b]. . 2. . M ) and M = f (M. 2 Theorem A.1) has a unique equilibrium x ∈ [a. M. z) is non-decreasing in x and z in [a. b] is a continuous function satisfying the following properties: (a) f (x. mi−1 .0. M1 ≤ M0 = b. i→∞ and M = lim Mi .2) holds and by the continuity of f . and is nonincreasing in y ∈ [a.1) converges to x. b]. . . (b) If (m. Set m0 = a and M0 = b and for each i = 1. M ) and m = f (m. M ) ∈ [a. . b] be an interval of real numbers and assume that f : [a. m. . and mi ≤ xk ≤ Mi Set m = lim mi i→∞ 203 for k ≥ 3i + 1. b]3 → [a. Mi ≤ . M. . b] for each z ∈ [a. m. set Mi = f (Mi−1 . . b] for each x and z in [a. . Then (A. Now observe that for each i ≥ 0 a = m0 ≤ m1 ≤ . ≤ mi ≤ . b] × [a. mi−1 ) and mi = f (mi−1 . y. Then Eq(A. m = f (m. Therefore in view of (b) m=M from which the result follows. b] and every solution of Eq(A.Appendix Proof. b] is a solution of the system M = f (M.3 Let [a. Mi−1 ).

m.0.4 Let [a. mi−1 ) . M. ≤ mi ≤ . Set m0 = a and M0 = b and for each i = 1. M1 ≤ M0 = b. 2 and M = f (M. . . .2) holds and by the continuity of f . b] and is non-decreasing in y and z for each x.204 Proof. . b] and every solution of Eq(A. m) . Theorem A. Then Eq(A. 2. b] be an interval of real numbers and assume that f : [a. . Now observe that for each i ≥ 0 a = m0 ≤ m1 ≤ . M ) and m = f (M. Mi ≤ . M. . Then (A. .1) has a unique equilibrium x ∈ [a. M ) ∈ [a. i→∞ and M = lim Mi . b]3 → [a. . . b] is a continuous function satisfying the following properties: (a) f (x. m = f (m. set Mi = f (Mi−1 . m. Mi−1 ) and mi = f (mi−1 . mi−1 . . M ) Therefore in view of (b) m=M from which the results follows. (b) If (m. Mi−1 . m) then m = M. b].1) converges to x. y. and mi ≤ xk ≤ Mi Set m = lim mi i→∞ Appendix for k ≥ 3i + 1. b] is a solution of the system M = f (m. b] × [a. y ∈ [a. z) is non-increasing in x for each y and z ∈ [a.

m. . . . b] and every solution of Eq(A. . . M ) and m = f (M. Mi−1 ) and mi = f (Mi−1 .Appendix Proof. M. b] × [a. 2 Theorem A. Therefore in view of (b) m=M from which the results follows. z) is non-decreasing in x for each y and z ∈ [a. b]3 → [a. Mi ≤ . Then (A. i→∞ and M = lim Mi .0. M1 ≤ M0 = b and mi ≤ xk ≤ Mi Set m = lim mi i→∞ 205 for k ≥ 3i + 1. . ≤ mi ≤ . M ) then m = M. Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . m). Set m0 = a and M0 = b and for each i = 1. M = f (m. y. (b) If (m. .1) has a unique equilibrium x ∈ [a. mi−1 ). b] is a solution of the system M = f (M. Mi−1 .5 Let [a. m) and m = f (m. set Mi = f (mi−1 . . .2) holds and by the continuity of f . M ) ∈ [a. M. 2.1) converges to x. b] and is non-increasing in y and z for each x ∈ [a. . Then Eq(A. m. b] is a continuous function satisfying the following properties: (a) f (x. b] be an interval of real numbers and assume that f : [a. b]. mi−1 .

M ) then m = M. set Mi = f (Mi−1 . z) is non-decreasing in y for each x and z ∈ [a. M.206 Proof. M1 ≤ M0 = b and mi ≤ xk ≤ Mi Set m = lim mi i→∞ Appendix for k ≥ 3i + 1. Therefore in view of (b) m=M from which the results follows. .6 Let [a. b] × [a. b]3 → [a. M ).1) has a unique equilibrium x ∈ [a. b]. 2 Theorem A. Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . .2) holds and by the continuity of f . . Mi−1 . . Mi ≤ . . 2. m. M = f (M. b] and every solution of Eq(A. i→∞ and M = lim Mi . mi−1 ) and mi = f (mi−1 . y. . ≤ mi ≤ . b] is a continuous function satisfying the following properties: (a) f (x. b] is a solution of the system M = f (m.1) converges to x. m) and m = f (M. Set m0 = a and M0 = b and for each i = 1. b] be an interval of real numbers and assume that f : [a. m. Mi−1 ). . .0. (b) If (m. M. m) and m = f (m. Then Eq(A. mi−1 . . M ) ∈ [a. b] and is non-increasing in x and z for each y ∈ [a. Then (A. .

mi−1 ) and mi = f (Mi−1 . Mi−1 ). (b) If (m. ≤ mi ≤ . M. . mi−1 . b] is a continuous function satisfying the following properties: (a) f (x. z) is non-decreasing in z for each x and y in [a. 2. M ) ∈ [a. Then (A. M = f (m. . M. Set m0 = a and M0 = b and for each i = 1. set Mi = f (mi−1 . b] × [a. 2 Theorem A. M1 ≤ M0 = b and mi ≤ xk ≤ Mi Set m = lim mi i→∞ 207 for k ≥ 3i + 1. Mi ≤ .7 Let [a. y. . Then Eq(A. .Appendix Proof. . M ). b] be an interval of real numbers and assume that f : [a. m.2) holds and by the continuity of f . i→∞ and M = lim Mi . . .0. m. b]3 → [a. m) then m = M. b]. Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . Mi−1 . m) and m = f (M. b] and is non-increasing in x and y for each z ∈ [a. . .1) converges to x. b] and every solution of Eq(A.1) has a unique equilibrium x ∈ [a. Therefore in view of (b) m=M from which the results follows. b] is a solution of the system M = f (m. . M ) and m = f (M.

b] is a continuous function satisfying the following properties: (a) f (x. ≤ mi ≤ .1) converges to x. M ) and m = f (M. Set m0 = a and M0 = b and for each i = 1. m). y. M ) then m = M. . y.2) holds and by the continuity of f . (b) If (m. Then Eq(A. b] × [a. M1 ≤ M0 = b and mi ≤ xk ≤ Mi Set m = lim mi i→∞ Appendix for k ≥ 3i + 1. b] be an interval of real numbers and assume that f : [a. b].8 Let [a. . . Therefore in view of (b) m=M from which the results follows. . M. 2. 2 Theorem A. M. .0. .1) has a unique equilibrium x ∈ [a. b] is a solution of the system M = f (m. Mi−1 ) and mi = f (Mi−1 . Then (A. . m. b] and every solution of Eq(A. . i→∞ and M = lim Mi . Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . m.208 Proof. mi−1 ). Mi ≤ . M = f (m. z in [a. . . set Mi = f (mi−1 . M ) ∈ [a. m) and m = f (M. mi−1 . z) is non-increasing in all three variables x. b]3 → [a. Mi−1 .

xn−1 . up ) is non decreasing in the second and the third variables u2 and u3 and is non increasing in all other variables. . . . b] is a continuous function satisfying the following properties: (a) f (u1 . Mi−1 ). m. . Set m0 = a and M0 = b and for each i = 1. . where is a continuous function. . mi−1 . ≤ mi ≤ .Appendix Proof. . . . Let us formulate one of these results.9 Let [a. xn−p+1 ).2) holds and by the continuity of f . i→∞ and M = lim Mi . . there are 2p possibilities for the function f to be non decreasing in some arguments and non increasing in the remaining arguments. M1 ≤ M0 = b and m i ≤ x k ≤ Mi Set m = lim mi i→∞ 209 for k ≥ 3i + 1. b] be an interval of real numbers and assume that f : [a. m) and m = f (M. . b]p → [a. u2 .3) 2 The above theorems can be easily extended to the general difference equation of order . (A. . Therefore in view of (b) m=M from which the results follows. p xn+1 = f (xn . . 1.1-A. Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . . . .8. Theorem A. Then (A. M.0. . . f : Ip → I n = 0. Mi−1 . This gives 2p results of the type found in Theorems A. In this case. 2. . mi−1 ) and mi = f (Mi−1 . Mi ≤ . M = f (m. .0. M ). set Mi = f (mi−1 .0.

M ) then m = M. . . . . b] × [a. m) and m = f (M. M ) ∈ [a. M.3) has a unique equilibrium x ∈ [a. M. m. . . . m. .3) converges to x. m. M. . b] is a solution of the system M = f (m. Appendix Then Eq(A.210 (b) If (m. b] and every solution of Eq(A.

Ladas. Abu-Saris. 1992. A. 967-968. Amleh. September 1-5. Equations Appl. M. and G. Gordon and Breach Science Publishers. On a class of difference equations with strong negative feedback. M. Ladas. Allen. New York.. Amleh. Tanny. Taiwan. On the periodicity of the difference equation xn+1 = α|xn | + βxn−1 . Math. 1-5. B. 24(1998). Thorvilson. 1(3)(1995). N. [7] A. Differ. Springer-Verlag. and W. New York. M. 5(1)(1999). 497-515. [5] American Mathemathical Monthly. Ladas. Georgiou.A. 847-864. Periodicities of solutions of the generalized Lyness recursion. G. 790-798. Problem 10670. [10] R. Tokyo. 5(6)(1999). [9] A. [4] K. Lipe. Convergence to periodic solutions. 1997. and G. J. Wimp. Differ. Beckermann and J. 7(2001). [2] L. Grove. Hoag. J. J. Alligood. Amleh and G. Math. A. J. [6] A. [12] B.T. 1977. M. and S. Equations Appl. S. Appl. Strauss. [8] A. Hannigan. Equations Appl. J. 57-69. Sauer and J. On the recursive sequence n−1 xn+1 = α + xxn . T. Gelbard. J. M. Amleh. Differ. Difference Equations and Inequalities. J. Strauss. A Preliminary mathematical model of the Apple Twig Borer (Coleoptera: Bostrichidae) and Grapes on the Texas High Plains Ecological Modelling 58 (1991). Heidelberg. Agarwal. Differ. Equations Appl. M. K. 89-97. D. Barbeau. S. 211 . Taipei. Theory. Mathematical analysis of a model for a plant-herbivore system. Methods and Applications. [3] L. Ladas. E. 291-306. Appl. 233(1999). Proceedings of the Third International Conference on Difference Equations and Applications. and M. CHAOS An Introduction to Dynamical Systems. N.Bibliography [1] R. J. Anal. Kruse. and J. Yorke . [11] E. Marcel Dekker Inc. 369-382. 106(1999). Bulletin of Mathematical Biology 55 (1993). Some dynamically trivial mappings with applications to the improvement of simple iteration. Boundedness of solutions of a plantherbivore system. Berlin. M. T. Comput. Allen. 1999. (to appear).

7(2001). W. W. Italy. El-Metwally. . Lyness-type equations in the third quadrant. 7(2001). R. [25] H. 245 (2000). Proceedings of WCNA2000. 2000. Nonlinear Systems and Applications. Equations Appl. R. Global attractivity and convergence to a two-cycle in a difference equation. S. editor. Boca Raton. Italy. Academic Press. J. L. (to appear). 203-209. Amer. An Introduction to Difference Equations. Stability or chaos in discrete epidemic models. 73-93. Ladas. and G. J. xn−3 xn−4 A + xn xn−1 [21] S. A global convergence result with applications to periodic solutions. 6(4)(2000). 2000.. G. 30(1997). Camouzis. Kulenovi´. J. New York. FL. On the Recursive Sequence xn+1 = A 1 + xn−2 . Differ. Differ.(to appear). 5(2)(1999). 1183-1189.. and H. Anal. DeVault. Grove. [18] R. Catania. On the recursive sequence xn+1 = −1 + xn−1 . On the global attractivity and the periodic character of some difference equations. [19] R. Franke. Ladas. Discrete Chaos. Ladas. D. E. Schultz. and G. Lakshmikantham. Level. and S. L. TMA. G. A. 126 (1998). July 19-26. Elaydi. Nonlinear Anal. Levins. G. J. and E. Ladas. Calef. Ladas. Proceedings of WCNA2000. 6(1)(2000). Valicenti. Equations Appl. Voulov. [24] H. Cunningham. [27] J. Ladas. and M. [17] R. Radin. Math. Cooke. New York. Equations Appl. [16] R. Grove. July 19-26. Differ. R. Devaney. and S.212 BIBLIOGRAPHY [13] E. [23] H. J. Ladas. El-Metwally. El-Metwally. J. Differ. (to appear). Math. Elaydi. Anal. CRC Press. On the recursive c α+βxn Sequence xn+1 = Bxn +Cxn−1 . Differ. DeVault. xn [14] K. 1992. 481-483. G. Nonlin. 2nd ed. DeVault. Devaney. Reading. Grove. [22] S. 1977. A Piecewise linear model for the zones of instability of an areapreserving map. 161-170. M. 1999. Janowski and G. 121-125. E. Soc. and G. E. MA. Hoag. and G. Feuer. J. Physica D 10(1984). Schultz. DeVault. Addison-Wesley. Equations Appl. A. V. xn [20] R. Appl. J. Galminas. 2000. Proc. A First Course in Chaotic Dynamical Systems: Theory and Experiment. Ladas. and S. J. [26] J. A. On the recursive sequence xn+1 = 1 . T. G. On the recursive sequence A B xn+1 = xn + xn−1 . V. Catania. [15] K. Equations Appl. Springer-Verlag. Ladas. E. F. 3257-3261. E. On the difference equation xn+1 = α + βxn−1 e−xn . Ladas. 387-393. Janowski. E. (to appear). D.

T. H. 1983. New York. J. and Y. Comm.BIBLIOGRAPHY 213 [28] C. 149-176. 6(3)(2000). 2000. and H. M. Bolis. Appl. R. 1(1994).S. and Bifurcations of Vector Fields. Proceedings of the First International Conference on Difference Equations and Applications. a+bxn [36] J. Karakostas. M. Global stability in some population models . G. . 8(2001). M. Equations Appl. A. Heidelberg. G. Kent. M. J. C. L. A. 1-25. Ladas. and G. 865-866. Gibbons. Grove. Systems 1(1993). Kocak. Ch. E. 1994. Greek Math. [31] E. San Antonio. Grove. Philos. Holmes. Kulenovi´. M. [33] J. On the recursive sequence c α+βxn−1 xn+1 = γ+xn . Tokyo. Nonlinear Anal. C. Math. The dynamics of some discrete population models. Gibbons. Ladas. Hautus and T. Karakostas. Valicenti. L. A+xn [30] E. On the rational recursive sequence xn+1 = (αxn +β)/((γxn +C)xn−1 ). Equations Appl. 61-72. Differential Equations Dynam. On the global asymptotic stability of xn+1 = A+xn−1 . August 27-31. Convergence of a difference equation via the full limiting sequences method. 1-11. Poznan. Differ. A. 289-294. H. Ladas. and S. Appl. G. Sci. Nonlinear Anal. Ladas. J. Nonlinear Anal. Karakostas. Sficas. Differ. [40] G. Grove. Springer-Verlag. [38] G. Poland. [29] C. May 25-28. Karakostas. J.C. 17(1991). Tokyo. Hale and H. McGrath. (1989). Proceedings of the Fourth International Conference on Difference Equations and Applications. Jaroma . Math. Society. Gordon and Breach Science Publishers. 7(2001) (to appear). Berlin. A discrete semi-flow in the space of sequences and study of convergence of sequences defined by difference equations. and G. 66-74. 281-294. S. H. Monthly. Janowski. Asymptotic 2-periodic difference equations with diagonally selfinvertible responses. Dynamics and Bifurcations. 1991. [35] M. Comm. Kent. Ladas. Berlin. 1069-1084. [39] G. Levins. Teixeira. 329-335. New York. S. Guckenheimer and P. Amer. Hot-Line. 86(1979). G. Gordon and Breach Science Publishers. and C. [34] J. 1995. D. [32] E. Res. 4(2)(2000). Nonlinear Oscillations. G. Solution to Problem E2721. R. Kulenovi´. R. Texas. Springer-Verlag. On the trichotomy c character of xn+1 = α+βxn +γxn−1 . Voulov. E. Existence and behavior of solutions of a rational system. 1998. Heidelberg. Dynamical Systems. [37] G.

n +Cxn−1 [51] M. Equations Appl. Kosmala. Discrete and Impulsive Systems 1(1995). Res. Invariants and related Liapunov functions for difference equac tions. Martins. On a rational difference equation. W. Global Behavior of Nonlinear Difference Equations of Higher Order with Applications. R. Math. [46] U. 173 (1993). [49] M. S. G. A. Kocic. Ladas. 563-576. L. J. Rodrigues. Anal. 1991. Lett. S. R. M. S. [47] N. [42] V. Appl. Kulenovi´. J. Open problems and conjectures: On period two c α+βx solutions of xn+1 = A+Bxn +γxn−1 . Kulenovi´. [43] V. S. Hot-Line 2(1998). Kulenovi´. R. 1+xn [53] M. On the recursive c p+yn−1 sequence yn+1 = qyn +yn−1 . L. Appl. J. R. Kruse and T. W. 127-157. 1-16. Ladas. and I. 6(5) (2000). Dordrecht. On the recursive sequence xn+1 = c αxn +βxn−1 . On the stability of Lyness’ equation. Ladas. 31-37. L. Appl. and relative stability for positive nonlinear difference equations of higher order. and C. Kulenovi´. G. S. S. L. Math.214 BIBLIOGRAPHY [41] W. On the recursive sequence xn+1 = c αxn +βxn−1 . Kelley and A.. Kuang and J. R. and W. J. J. [52] M. New York. 1(4) (1995). Math. Differ. Stability trichotomy. Ladas. Anal. Kulenovi´. Tzanetopoulos. S. 6(5) (2000). Kluwer Academic Publishers. Sizer. and N. [44] V. K. [48] Y. Global stability in a nonlinear difference-delay equation model of flour beetle population growth. [45] W. Rodrigues. 151-158. Appl. R. path stability. G. Kulenovi´. G. Equations Appl. T. Kocic. Math. Math. and N. Prokup. Differ. 41(2001). R. J. Cushing. Krause. Peterson. Equations Appl. and I. Kulenovi´ and G. (to appear). R. Ladas. 251(2000). 245-254. γxn +Cxn−1 . c Comput. 235 (1999). Sci. G. G. Ladas. 2(1)(1996). Appl. G. Dynamics of Continuous. [50] M. 1993. Thomas. R. Ladas. M. C. On the dynamics c α+βxn of xn+1 = A+Bxn +Cxn−1 facts and conjectures. G. Differ. Ladas. Comput. S. Teixeira. no. F. and E. Anal. Differ. 1-8. Math.5. G. [54] M. Math. Equations Appl. Holland. Ladas. Appl. Nesemann. On the rational recursive sequences. 641-646. J. Global asymptotic stability for discrete dynamical systems. Difference Equations. Kocic and G. 671-678. 13 (7)(2000). Prokup. Academic Press. 323-346. 571-586.

Equations Appl. 1965. 1(3)(1995). Open problems and conjectures. 224(1998). John Wiley & Sons. [56] G. C. J. Equations Appl. Equations Appl. J. An Introduction to Mathematical Ecology. Sedaghat. 1995. On the recursive sequence xn+1 = 4(3)(1998). J. 73(2000). G. 1(1)(1995). 160-161. (to appear). J. Differ. Equations Appl. FL. Differ. New York. Open Problems and Conjectures. 255-272. Population and Community Ecology. [59] G. Kuruklis and G. Existence of solutions of certain singular difference equations. Ladas. Open problems and conjectures. . Ladas. J. Math. Pielou. Boca Raton.BIBLIOGRAPHY 215 [55] S. Ladas. CRC Press. 1974. Math. 497-499. Open problems and conjectures. [66] E. Problem 1572. J. Quart. Math. 249-258. Equations Appl. Differ. Oscillation and global attractivity in a discrete delay logistic model. Comput. Differ. Differ. Equations Appl. Ladas. Differ. Ladas. Geometric stability conditions for higher order difference equations. Equations Appl. k Ai i=0 xn−2i . 62(1994). [67] E. Ladas. G. J. [58] G. Appl. Global attractivity in a nonlinear difference equation. Anal. 50 (1992). Invariants for generalized Lyness equations. Ladas. Introduction to Discrete Dynamical Systems and Chaos. 227-233. [57] G. [62] G. Appl. Purnaras. Sficas. 6(5)(2000). J. 317-321. 95-97. Open problems and conjectures. New York. [61] G. Symbolic Dynamics. 1(2)(1995). 535-561. Stability. [64] Mathemathics Magazine. 209-214. [65] Ch. 1999. 4(5)(1998). C. K. 311-313. [68] C. I. Philos. Differ. Appl. J. 413-419. Robinson. [60] G. Martelli. Gordon and Breach. Ladas. Differ. New York. Equations Appl. [70] H. 1(4)(1995). and Y. [69] H. Pielou. John Wiley & Sons. 7(2001). Sedaghat. [63] M. and Chaos.

Differential Equations Dynam. (to appear). January 3-7. Existence of nonperiodic solutions of the Lyness equation xn+1 = J. Systems 5(1997). [75] Y. 319-324. Gordon and Breach Science Publishers. A+Bxn +Cxn−1 . http://www. [72] E. Proceedings of the Fifth International Conference on Difference Equations and Applications. xn−1 [74] Y. Zheng. 94-102.html [73] Y.utsa. Anal. Sizer. Zeeman.edu/ecz/gu. Zheng. Chile. α+xn . Some periodic solutions of the Lyness equation . Temuco. 2001. Appl. Systems 6(1998). Differential Equations Dynam. On periodic cycles of the Lyness equation. Zheng.math. Periodic solutions with the same period of the recursion xn+1 = α+βxn +γxn−1 . Math. 51-58. Geometric unfolding of a difference equation.216 BIBLIOGRAPHY [71] W. C. 209(1997). S. 2000.

2. 7 repeller. 2. 128 equilibrium point. 15 full limiting solution. 72–75. 6 invariant. 9 positive semicycle. 183 linearized stability theorem. 78 Riccati number. 30 217 locally asymptotically stable. 75 negative feedback condition. 184–189 limiting solution. 64–68. 18 saddle point. 25 open problems and conjectures. 36 local stability of the positive equilibrium . 163–166. 2. 7 periodic solutions with the same period. 29 existence of prime period-two solution. 33 Pielou’s discrete delay logistic model. 129 forbidden set. 180–182. 41–42. 6 local asymptotic stability of a two cycle. 66. 128 population model. 75 global asymptotic stability of the zero equilibrium. 9. 7 . 1. 192–200 oscillate about x. 16 gingerbreadman map. 6 Lyness’ equation. 7 semicycle analysis. 5.Index basin of attraction of a two cycle. 38 Discrete Epidemic Models. 35 existence of solutions. 25 oscillate about zero. 2 fixed point. 9 negative semicycle. 24 prime period p solution. 71 invariant interval. 138–139. 6. 32 local stability of the zero equilibrium. 38 sink. 24–28 signum invariant. 31 global attractivity of the positive equilibrium. 17 full limiting sequence. 17. 25 periodic solution. 6 locally stable. 43–46. 5 Flour Beetle Model. 123–130. 57 Plant-herbivore system. 134– 136. 25 nonoscillatory. 70 May’s host parasitoid model. 9–14 global attractor. 15 linearized stability analysis. 38 classification of equation types. 5. 49–51. 3 convergence to period-two solutions. 129 positive equilibrium. 66 Pielou’s equation. 7 Riccati difference equation. 6 globally asymptotically stable.

69–70 (2. 8 stable manifold theorem in the plane. 42. 1)-type. 1)-type. 3)-type. 8 INDEX . 6 unstable manifold. 25 table of invariant intervals. 5 stability trichotomy. 53–54 (1. 47–48 (1. 168 type of equation (1. 3)-type. 167 (3. 3)-type. 2)-type. 7 strictly oscillatory. 2)-type. 10 stable manifold.218 source. 141 (3. 2)-type. 189 trichotomy character. 6 stability of equilibrium. 137 (3. 1)-type. 183 unstable equilibrium. 131 (2. 77–78 (2.

Sign up to vote on this title
UsefulNot useful