This action might not be possible to undo. Are you sure you want to continue?

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

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

|λ

±

| =

√

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

<

Kρ

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

kπ

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 =

pφ

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.

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

. . . .2 The Case α = γ = A = B = 0 : xn+1 = Cxn−1 3. . . . . . . 1. . . . . . . . .3 The Case α = β = A = C = 0 : xn+1 = γxn−1 . . . Periodicity. . . . . . . . . . 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 . . . . . . . .7 Semicycle Analysis . . . . . .4 When is Every Solution Periodic with the Same Period? . 1. . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . . and Invariant Intervals 2. . . . .4 Global Attractivity of the Positive Equilibrium . . . . . . . . . . . . . . . . . . . . . v . . . . . . . . . . . . .5 Limiting Solutions . . . . . . 2. . . . . . . . . . . . . . .2 Stability of the Zero Equilibrium . . . 2. . . . . . 2 Local Stability. . . . . . . . .1 Introduction . . . . 1)-Type Equations 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 (1. . . . . . . . . 1. . . . 2. . . . . . . . . . . . . . . . . . . .4 Open Problems and Conjectures . . . . . . . . . . . . 2. . . . . . . . . 2. . . . . . . . . . . . . . . . . . . . . .7 Convergence to Period-Two Solutions When C=0 . . . . . . .3 Global Asymptotic Stability of the Zero Equilibrium . . . . . . . . . . . . . . . . . . . . . βxn 3. . . . . . . . . . .1 Equilibrium Points . . . . .3 Local Stability of the Positive Equilibrium . . . . . . . . . . . . . . . Bxn 3. . . Semicycles. . . . .Contents Preface Acknowledgements Introduction and Classiﬁcation of Equation Types 1 Preliminary Results 1. 1. . 1. . .5 Existence of Prime Period-Two Solutions . . . . . . . . . . . . . . . . .8 Invariant Intervals . . . . . . . . . . .6 Local Asymptotic Stability of a Two Cycle . . . . . . . . . . .6 The Riccati Equation . . . . . 2.1 Deﬁnitions of Stability and Linearized Stability Analysis 1. . . 2. . . . . . . . . . . . . .2 The Stable Manifold Theorem in the Plane . . . . . .9 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . .1 Introduction . . . Bx 5. . . .3 Semicycle Analysis . . . . . . . . .1 The Case p < 1 . . . . βxn A+Cxn−1 βxn Bxn +Cxn−1 γxn−1 . . . . . . . . .1 Semicycle Analysis . . . . . 6. . . . . . .7 The Case α = C = 0 : xn+1 = βxA+Bxn−1 . . . . . . . . . n 6.6. . .1 Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6. . . . . . . . . . . . . . . . . . . . . . . . . . .3 The Case β = A = C = 0 : xn+1 = α+γxn . .4. . . . . 54 α 4. 6. . . . . .5 The Case β = C = 0 : xn+1 = α+γxn−1 . . . . . . . . . . . . . . . . . . . .1 Existence and Local Stability of Period-Two Cycles . 2)-Type Equations 77 6. . . . 4. 6. . .3 Semicycle Analysis and Global Attractivity When q = 1 . . . . . . . . . . . . . . . . . . . . . . 6 (2. . . . . . .2 Semicycle Analysis . . . . . . . . . . . . .7. . . . . . . . . . . . .7. . . .4 The Case α = γ = B = 0 : xn+1 = 4.6. .6. . . . . . . . . . . . . . . . . . . . . .5 The Case α = γ = A = 0 : xn+1 = 4. . . . . . . . . . .1 Introduction . . . . .5. . . . . . 6. . . . . . . . . . . . . . . . . . . . . . . .7. . . . . . . . . . . . 1)-Type Equations 5. . .4 . .2 The Case β = γ = C = 0 : xn+1 = A+Bxn . . .2 Semicycle Analysis of Solutions When q ≤ 1 . . . . . . . . .1 Introduction . . . . . . .7. . . . . . . . . . . . 6. . . . . . . . 6. . . . . . . . . . . . . . 78 n α+βxn 6. . . . . . A+Bxn γxn−1 Bxn +Cxn−1 Pielou’s Equation . . . 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 6. . . . . .3 The Case p > 1 . . . 6. . . . . . . 5 (2. . . . . .Riccati Equation . .5. . . . . . . . . . . α+γxn−1 6. . 6. . .7 The Case α = β = A = 0 : xn+1 = 4. . . . 77 α+βx 6. . . . . . . . . . .4. . . .2 The Case q < 1 . . . . . . . . 4. . . . . 53 α 4. . . . . . . . . . . . . 5. . . . . . . . . . . . . .8 Open Problems and Conjectures . . . 6. . . .4 Global Behavior of Solutions . . 57 58 59 60 61 62 62 64 69 69 70 72 72 . . . . . . . . .Lyness’ Equation Cxn−1 n−1 5. . .1 Invariant Intervals . . . . .2 The Case p = 1 . . . . . . . . . . . 55 4. . . . . . . . . . . . . . . . . . . . . . 2)-Type Equations 53 4. . . . . . . . . . . . . . . .vi Contents 4 (1. . . . . . . . . . .6. . . . . . . . . . . . The Case γ = A = 0 : xn+1 = Bxα+βxn n +Cxn−1 6. . . . . . . . .3 Global Stability and Boundedness . . . . . . . . . . . . n +γx 6. . . .3 The Case q > 1 . . . . . . . . . . . . . . . . . . . . . . . .3 The Case γ = B = 0 : xn+1 = A+Cxn−1 . A+Bxn 6. . . . . . . . . . . . . . . . . .4 Global Attractivity When q < 1 . .4.3 The Case β = γ = A = 0 : xn+1 = Bxn +Cxn−1 . . . . .2 The Case γ = C = 0 : xn+1 = A+Bxn . . . . . .2 The Case γ = A = B = 0 : xn+1 = α+βxn . . . . . . . . . . . . . 6. . . . . . . . .7. . . . . . . . . . . .2 Invariant Intervals . . . . .7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6. . .7.6 The Case α = β = C = 0 : xn+1 = 4. . . . . . . . . . 82 83 84 87 89 90 90 91 92 93 96 97 100 101 102 103 105 107 . .6 The Case β = A = 0 : xn+1 = Bxn +Cxn−1 . . . . . . . .4 Open Problems and Conjectures . . . . . . . .5.

. . . . . 3)-Type Equations 9. . . . . . .1 Invariant Intervals and Semicycle Analysis . . . . . . . . . .4 Global Asymptotic Stability . A+Cxn−1 6. . . . . . . . . . 1)-Type Equations 137 8. .2 Semicycle Analysis . . . . .1 Existence of a Two Cycle . 3)-Type Equations 131 7. . . . . . . . βxn +γx 6. . . . . . . . . . . . . .9. 9. . . . . . . . . . . . . . 137 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . α+γxn−1 9. . . . .2. . . . . .9. . . . . . . . . . . . .5 Semicycle Analysis When pr = q + qr .2 Invariant Intervals . . . . . . . . . . . .9. . . . . . . . . . . . 131 +Cx 7. . . .3. 9. .3 Semicycle Analysis . . . . . . . . . . . . . . . . . . .Contents 6. 141 141 141 142 143 144 147 149 150 154 156 157 157 158 160 162 163 . . .2. . . . . . . . . 132 n +Cx γx 7. 9. .2 Convergence of Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10 Open Problems and Conjectures . . . . . . . . . . . . . 131 7. . . . .3 Convergence of Solutions . 9. .2 Open Problems and Conjectures . . . . . . α+βxn 9. . . . . . . . . . . .1 Introduction . . . . . . . .4 The Case α = 0 : xn+1 = A+Bx+γxn−1 . . . . . . . . . . . . . . . . . . . . . . . . . .9. .3. . . . . .2. . . . . . . 9. . . . . . . . 2 9. . . . . . . . . . 138 9 (2. . . . . . 6. . . . . . . . .9 The Case α = A = 0 : xn+1 = Bxn +Cxn−1 . . . . . βxn 9.8. . . . . . . . . . . . . . . .4. . .4 Global Stability Analysis When p > q . . .4 The Case α = β = 0 : xn+1 = A+Bxnn−1 n−1 . . . . . . . . . n−1 6. . .3. . . . 6. . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Invariant Intervals . . . . . .5 Long-term Behavior of Solutions When q > 1 6.3. . . . . . . . . . . .2 Global Stability of the Positive Equilibrium .3 The Case α = γ = 0 : xn+1 = A+Bxβxn n−1 . . . . . . . . . . . . . . . . . . . 6. . . .1 Invariant Intervals . . 6. . . . . . . .2 The Case γ = 0 : xn+1 = A+Bxn +Cxn−1 . . 9. . .5 Open Problems and Conjectures . . . . . .5 Open Problems and Conjectures .3 The Case β = 0 : xn+1 = A+Bxn +Cxn−1 . . . . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . .3. . vii 107 109 110 112 113 114 117 117 123 123 7 (1. . 9. . .1 Introduction . . . . . . n +Cxn−1 9.8. . . . . . . . . . . . . 133 +Cx 7. . . . . . . . . . . . . . . . . . . . . . . . . .1 Introduction . . . . 134 8 (3. .4 Global Stability . . . . . . . . . . . . .3 Global Stability Analysis When p < q . . . . . . . . 9. . . . . . . . . . . . . . . . 6. . . . . . . . . . . . . . .7. . . . .2 The Case β = γ = 0 : xn+1 = A+Bxnα n−1 . . . .1 Boundedness of Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8 The Case α = B = 0 : xn+1 = βxn +γxn−1 . . . . . .1 Existence and Local Stability of Period-Two Cycles 9. . . . . . . . . . . . . . 9. .

. . . . . . . . . . . . . . . . . . α+βx (3. . . . . n +γxn−1 10. . . . . . . . . . . . . . . . Convergence Results . . . . . . . . . . . . . . . . . . . . . . . .1 11. Invariant Intervals .2.1 Invariant Intervals . . . . . . . . . . . . . . . .3 Contents 167 167 167 169 172 172 173 173 174 175 177 179 180 183 183 184 189 191 192 201 211 217 . . . . .2 The Case C = 0 : xn+1 = α+βxn +γxn−1 . . . . . . . . .4. . . . . . . . . 2)-Type Equations 10. 10. . . Bxn 10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 11. . . . . . . . . . . . . . .4 Open Problems and Conjectures . . . . . . . . .1 10.2 Proof of Part (c) of Theorem 10. . . . . . .2. . . . . . . .1 Introduction . . . . . . . . . . . . 11 The 11. . . . . . . . . . . .3.4. . . . . . . . . . . . . . .2 Convergence of Solutions . . . . .3. . . . . . . . . . . . . . . . . . . . 3)-Type Equation xn+1 = A+Bxn +γxn−1 n +Cxn−1 Linearized Stability Analysis . . . . . . . . . . . . . . . . .1 10. . . . A+Cxn−1 10. . . . . . . . . . . . . . . . .1 Proof of Part (b) of Theorem 10. . . 10. . . . . . 11. . . . . . . . . . . . .3 The Case B = 0 : xn+1 = α+βxn +γxn−1 . . . .5 Open Problems and Conjectures .1 Global Stability . . . . . . . . . . . . . . . 10. . A Global Attractivity for Higher Order Equations Bibliography Index . . .2 Global Stability When p ≥ r . . . . . . .3. . . . . A+Bxn 10. . . . . . . . . . . . 10. 10. .2. . . . . . . . . . . . .3 Convergence of Solutions . . . . .3. . . . . . . . . . .1 Invariant Intervals . . .4 The Case A = 0 : xn+1 = α+βx+Cxn−1 . . . . . . . . 11. . . . . . . . . . . . . . . . . . . .viii 10 (3. . . . . . . . .2. . . . . . . . . . . .

the global stability. (1) where the parameters α. 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. γ. It is an amazing fact that Eq(1) contains. as special cases. A. β. In Chapter 2 we present some general results about periodic solutions and invariant intervals of Eq(1). In an appendix. we present some global attractivity results for higher order diﬀerence equations which may be useful for extensions and generalizations. 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. . 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. 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. at the end of this monograph. A + Bxn + Cxn−1 n = 0. . we present the known results and derive several new ones on the boundedness. 1. Chapter 1 contains some basic deﬁnitions and some general results which are used throughout the book. B. In this sense. 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). a large number of equations whose dynamics have not been thoroughly understood yet and remain a great challenge for further investigation. ix .Preface In this monograph. C and the initial conditions x−1 and x0 are nonnegative real numbers. 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. Every chapter in this monograph contains several open problems and conjectures related to the particular equation which is investigated and its generalizations. .

.

D. and K. S. Predescu. C. Feuer. W. W. Sizer and to our graduate students A. C. O. Teixeira. J. R. Briden. F. Prokup. H. El-Metwally. Kent. P. xi . M. S. Cunningham. I. E. J. McGrath. E. and W. Valicenti. C. E. Merino. DeVault. W. A. Schultz. Wilkinson for their enthusiastic participation and useful suggestions which helped to improve the exposition. S. Driver. C. Camouzis. C. Clark. N. T. M. Hoag. A. Janowski. M. Gibbons. L.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. Amleh. Rodrigues. Radin. K. We are thankful to Professors R. Martins. L. Grove. Kosmala.

.

To Senada and Theodora .

.

a large number of equations whose dynamics have not been thoroughly established yet. A + Bxn + Cxn−1 n = 0. A. The Riccati diﬀerence equation when γ = C = 0. See Section 1. [21]. 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. . Eq(1) includes among others the following well known classes of equations: 1. This equation is studied in many textbooks on diﬀerence equations. .6 where. in addition to the basic description of the solutions 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.Introduction and Classiﬁcation of Equation Types In this monograph. γ. . such as [1]. and the periodicity of solutions of all rational diﬀerence equations of the form xn+1 = α + βxn + γxn−1 . etc. [42]. (1) where the parameters α. [41]. It is an amazing fact that Eq(1) contains as special cases. 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. 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. B. 1. β. we present the known results and derive several new ones on the boundedness. the global stability.

The problem of existence of solutions for diﬀerence equations is of paramount importance but so far has been systematically neglected. 3. 1. and Section 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. 75).2. [67]. we introduce the forbidden set of the equation. A characteristic feature of this equation is that it possesses an invariant which can be used to understand the character of its .2. See [42]. p. [49]. it is interesting to note the similarities of the dynamics of the population model xn+1 = α + βxn−1 e−xn . . 1. Lyness’ equation when γ = A = B = 0. An example of such a problem is the following: Open Problem Find all initial points (x−1 . . . and Section 1. See ([42]. / if and only if x0 ∈ F. Hence the solution {xn } of Eq(2) exists for all n ≥ 0. 2.2 Introduction and Classiﬁcation of the Riccati equation. [66]. 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. . See [13]. [57]. See [24] and Sections 2. and Section 5. 1 + xn n = 0.4.7 and 10. [44]. For example. . This is a fascinating equation with many interesting extensions and generalizations. and the rational equation xn+1 = α + αxn + βxn−1 . [69]. Pielou’s discrete delay logistic model when α = γ = B = 0. n = 0.6 for some results in this regard. See [13]. [18]. . [55]. 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. [32]. . x0 ) ∈ R × R through which the equation xn−1 xn+1 = −1 + xn is well deﬁned for all n ≥ 0.

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

.

and [68].1 Let x be an equilibrium point of Eq(1.1). Then for every set of initial conditions x0 . Deﬁnition 1. that is. [34]. . xn = x for n ≥ 0 is a solution of Eq(1. [63]. [41].1. [33]. n = 0.1 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.1). x is a ﬁxed point of f . x). x−1 ∈ I.1) has a unique solution {xn }∞ .1) if x = f (x. [22]. or equivalently. See also [1]. [21]. 1. . 5 . n=−1 A point x ∈ I is called an equilibrium point of Eq(1. (1. [16].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. . 1. the diﬀerence equation xn+1 = f (xn . xn−1 ). [4].

Preliminary Results (i) The equilibrium x of Eq(1. 1. then the equilibrium x of Eq(1.1) is called a global attractor if for every x0 . (iii) The equilibrium x of Eq(1. n = 0. x) and q = (x. or a repeller. there exists N ≥ 1 such that |xN − x| ≥ r. x) ∂u ∂v denote the partial derivatives of f (u. .1 (Linearized Stability) (a) If both roots of the quadratic equation λ2 − pλ − q = 0 (1. Then the equation yn+1 = pyn + qyn−1 . (1. x−1 ∈ I with |x0 − x| + |x−1 − x| < γ. n→∞ (iv) The equilibrium x of Eq(1. if there exists r > 0 such that for all x0 . x−1 ∈ I we have lim xn = x.1) is called locally stable if for every ε > 0.6 Chapter 1.3) Let lie in the open unit disk |λ| < 1. ∂f ∂f (x.1) is called globally asymptotically stable if it is locally stable and a global attractor.1) is called a source.1) about the equilibrium point x. . we have n→∞ lim xn = x.2) p= is called the linearized equation associated with Eq(1. . x−1 ∈ I with 0 < |x0 − x| + |x−1 − x| < r. (v) The equilibrium x of Eq(1.1) is called locally asymptotically stable if it is locally stable.1) is called unstable if it is not stable. (vi) The equilibrium x of Eq(1.1). v) evaluated at the equilibrium x of Eq(1. there exists δ > 0 such that for all x0 . Clearly a source is an unstable equilibrium. x−1 ∈ I with |x0 − x| + |x−1 − x| < δ. Theorem 1. . (ii) The equilibrium x of Eq(1. we have |xn − x| < ε for all n ≥ −1. and if there exists γ > 0 such that for all x0 .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 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).

7 below are new and deal with the remaining two cases relative to the monotonic character of f (x.5 ([54]) Let [a. . b] is a solution of the system f (m. m) = M.4. M ) ∈ [a. Global Attractivity of the Positive Equilibrium 11 Very often the best strategy for obtaining global attractivity results for Eq(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. then m = M .4. set Mi = f (Mi−1 . b]. y) is monotonic in its arguments. 2. (b) If (m. or nonincreasing in one and nondecreasing in the other. b] for each y ∈ [a. y) is non-increasing in x and nondecreasing in y.) The ﬁrst theorem deals with the case where the function f (x. .4. which describes the extension of these results to higher order diﬀerence equations. See Sections 6. Set m0 = a and M0 = b and for i = 1. ≤ mi ≤ . y) is nonincreasing in y ∈ [a. and f (x. In this regard there are four possible scenarios depending on whether f (x. 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. b] is a continuous function satisfying the following properties: (a) f (x.1. Mi−1 ).1) is to work in the regions where the function f (x.4. b] → [a. M ) = m and f (M. .8 and 1. b] for each x ∈ [a. Now observe that for each i ≥ 0.8 where these theorems are utilized. ≤ Mi ≤ . . . m0 ≤ m1 ≤ . b] × [a. Theorem 1. .9. or nonincreasing in both arguments. . b] and every solution of Eq(1. b] be an interval of real numbers and assume that f : [a. b] × [a. y) is nondecreasing in both arguments.1) has a unique equilibrium x ∈ [a. . Then Eq(1. ≤ M1 ≤ M0 and Set mi ≤ xk ≤ Mi m = lim mi i→∞ for k ≥ 2i + 1. y) is non-decreasing in x ∈ [a. See also the Appendix. mi−1 ) and mi = f (mi−1 . y). . Proof. Theorems 1.4 and 6. i→∞ and M = lim Mi . (See Section 6. . b]. at the end of this monograph.1) converges to x.

Preliminary Results M ≥ lim sup xi ≥ lim inf xi ≥ m i→∞ i→∞ (1. y) is nondecreasing in y ∈ [a. Mi−1 ) and mi = f (Mi−1 . 2 . Then Eq(1.6 ([54]) Let [a.4. . Proof. m = f (M. . Therefore in view of (b).1) converges to x. .12 Then Chapter 1. M ) and M = f (M. and Set mi ≤ xk ≤ Mi m = lim mi i→∞ for k ≥ 2i + 1. b]. b] for each y ∈ [a. m = f (m. . Now observe that for each i ≥ 0. . b] × [a. m=M from which the result follows.1) has a unique equilibrium x ∈ [a. ≤ mi ≤ . Set m0 = a and M0 = b and for i = 1. In view of (b). ≤ Mi ≤ . (b) The diﬀerence equation Eq(1. y) is non-increasing in x ∈ [a. . 2 Theorem 1. b] is a continuous function satisfying the following properties: (a) f (x. i→∞ and M = lim Mi . b]. .12) and by the continuity of f . b] for each x ∈ [a. b] be an interval of real numbers and assume that f : [a. m=M from which the result follows. . b]. ≤ M1 ≤ M0 . m0 ≤ m1 ≤ . m) and M = f (m. 2.1) has no solutions of prime period two in [a. and f (x. mi−1 ). set Mi = f (mi−1 . b] and every solution of Eq(1. b] → [a. Then clearly (1. m).12) holds and by the continuity of f . M ). .

. . 2. . b] and every solution of Eq(1. and Set mi ≤ xk ≤ Mi m = lim mi i→∞ for k ≥ 2i + 1. m). . y) is non-decreasing in each of its arguments. mi−1 ) and mi = f (Mi−1 . y) is non-increasing in each of its arguments.4. x) = x. Then Eq(1. Then Eq(1. M ) = m.1) has a unique equilibrium x ∈ [a. 13 then m = M . . Then clearly (1. M ) and M = f (m.1. 2 Theorem 1.1) converges to x. m0 ≤ m1 ≤ .7 Let [a. b] be an interval of real numbers and assume that f : [a. M ) ∈ [a. b] is a continuous function satisfying the following properties: (a) f (x. Set m0 = a and M0 = b and for i = 1. Global Attractivity of the Positive Equilibrium Theorem 1. In view of (b). b] and every solution of Eq(1. b] × [a. Proof.4. . has a unique positive solution. m = f (M. b] × [a. . Now observe that for each i ≥ 0. m=M =x from which the result follows. m) = M and f (M. ≤ M1 ≤ M0 .1) converges to x. b] → [a. . b] is a continuous function satisfying the following properties: (a) f (x.4. ≤ mi ≤ . b] → [a. Mi−1 ). b] be an interval of real numbers and assume that f : [a. b] × [a.8 Let [a. (b) If (m. . (b) The equation f (x. i→∞ and M = lim Mi .12) holds and by the continuity of f . b] is a solution of the system f (m. . ≤ Mi ≤ .1) has a unique equilibrium x ∈ [a. set Mi = f (mi−1 .

m = M = x. ≤ Mi ≤ .13) . J]. Then clearly the inequality (1. M ). (1. . f ∈ C[J × J. m) and M = f (M. where A. . . . from which the result follows. ≤ mi ≤ . Assume that {xn }∞ n=−1 is a solution of Eq(1.1) such that A ≤ xn ≤ B for all n ≥ −1. Now observe that for each i ≥ 0. and mi ≤ xk ≤ Mi Set m = lim mi i→∞ for k ≥ 2i + 1. . and let k ≥ 1 be a positive integer. . m = f (m. . Preliminary Results m0 = a and M0 = b and for i = 1. . ≤ M1 ≤ M0 . xn−1 ). Let J be some interval of real numbers. In this section we ﬁrst give a self-contained version of limiting solutions the way we will use them in this monograph. 2 1. n ≥ −k + 1. ∞ we will show that there exists a solution {Ln }n=−k of the diﬀerence equation xn+1 = f (xn . Set Chapter 1. . 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). set Mi = f (Mi−1 . m0 ≤ m1 ≤ . . Mi−1 ) and mi = f (mi−1 .14 Proof. Now with the above hypotheses. mi−1 ). 2.5 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 view of (b). .12) is satisﬁed and by the continuity of f . i→∞ and M = lim Mi . B ∈ J.

n=−1 The following result. −k+2 j→∞ = L−k and −k+1 = L−k+1 . . lim xnij = L and lim xnij −1 = L−1 . L−3 . It follows by induction that the sequence L−k . . .1. .5. the term LN is a limit point of the solution {xn }∞ . . . Limiting Solutions such that L0 = L 15 and for every N ≥ −k. B] and so it has a ∞ further subsequence {xnij }j=1 which converges to some limit which we call L−1 . Karakostas takes k = ∞ and calls the solution {Ln }∞ n=−∞ a full limiting sequence of the diﬀerence equation xn+1 = f (xn . −k ) = f (L−k+1 . L−1 . . L1 = 1.1). L0 = L = 0 . xn−1 ). is stated in the way we will use it in this monograph: . i→∞ i→∞ Continuing in this way and by considering further and further subsequences we obtain ∞ numbers L−2 . = f( −k+1 . xnj −k ) = lim xnj −k+2 = L−k+2 . ∞ Next the subsequence {xni −1 }i=1 of {xn }∞ n=−1 also lies in interval [A. (1.2 below..14) associated with the solution {xn }∞ n=−1 of Eq(1. j→∞ lim xnj −k = L−k . . n=−1 is called a limiting solution of Eq(1. . .13) with Then clearly.1).13) associated with the The solution {Ln }∞ n=−k solution {xn }∞ n=−1 of Eq(1.5.. L−k+1 . . (See Theorem 1. L−k and a subsequence {xnj }n=−1 such that j→∞ lim xnj = L j→∞ lim xnj −1 = L−1 . satisﬁes Eq(1. Clearly there exists a subsequence {xni }∞ of {xn }∞ i=1 n=−1 such that i→∞ lim xni = L. which is a consequence of the above discussion. L−k ) j→∞ = lim f (xnj −k+1 . −k Let { n }∞ n=−k be the solution of Eq(1.) ∞ We now proceed to show the existence of a limiting solution {Ln }n=−k .13) and every term of this sequence is a limit point of the solution {xn }∞ . Therefore.

with I0 = I. Let Z denote the set of all integers {. xn−1 . Sn ∈ [I. . .16 Chapter 1. n=−ν Furthermore for every m ≤ −ν. . S] for all n ∈ Z and such that for every N ∈ Z. n = 0. . See [23] and [25] for an application of Theorem 1. i=0 xn−i k n = 0. .5. 1. S ∈ J. . f ∈ C[J ν+1 . . ∞ The solutions {In }∞ n=−∞ and {Sn }n=−∞ of the diﬀerence equation (1. In . S ∈ J.1 Let J be some interval of real numbers. . . Suppose that {xn }∞ n=−1 is a bounded solution of Eq(1. . Theorem 1. . Then Eq(1.16) are called full limiting solutions of Eq(1. 1. xn−ν ) which satisfy the equation for all n ∈ Z. . . .1) with limit inferior I and limit superior S. S] for n ≥ −k. Preliminary Results Theorem 1. xn−1 . J]. . . . 0. and let k ≥ 1 be a positive integer. xn−ν ).15) and with I. IN and SN are limit points of {xn }∞ .2 to diﬀerence equations of the form xn+1 = Ai . . f ∈ C[J × J.}.5.5. . which by the change of variables xn = y1 reduce to rational diﬀerence equations of the n form 1 . as developed by Karakostas. . .2 ([37]) Let J be some interval of real numbers. with I = lim inf xn . 1. 1. there exist two subsequences {xrn } and {xln } of the solution {xn }∞ n=−ν such that the following are true: n→∞ (1. n→∞ S = lim sup xn n→∞ n = 0. and In . (1.15). −1. yn+1 = k i=0 Ai yn−i . Sn ∈ [I. Then there exist two solutions ∞ {In }∞ n=−∞ and {Sn }n=−∞ of the diﬀerence equation xn+1 = f (xn .16) lim xrn +N = IN and n→∞ lim xln +N = SN for every N ≥ m. and let {xn }∞ n=−ν be a bounded solution of the diﬀerence equation xn+1 = f (xn . . S0 = S. S0 = S We now state the complete version of full limiting sequences. J].13) has two solutions {In }∞ n=−k with the following properties: and {Sn }∞ n=−k I0 = I.16) associated with the solution {xn }∞ n=−k of Eq(1. with I. . .

C and the initial conditions x−1 .6 The Riccati Equation α + βxn .) The other goal we have is to describe in detail the long. Then (1.17) where the parameters α. (See also [13] and [69]. A. (1.6.6. The ﬁrst result in this section addresses a very special case of Eq(1. β.18).and short-term behavior of / solutions of Eq(1. . A + Bxn A diﬀerence equation of the form xn+1 = n = 0. Theorem 1. It should be mentioned here that there is practically nothing known about the forbidden set of Eq(1) (when the parameters α. The Riccati Equation 17 1. . (1. B and the initial condition x0 are real numbers is called a Riccati diﬀerence equation.17) when x0 ∈ F. Furthermore when (1.17) with x0 = is periodic with period two. x0 are all real numbers) other than the material presented here. in addition to (1.19) β + A = 0.1 Assume that (1.17) is a linear equation and when αB − βA = 0.17) regarding period-two solutions. .17). Eq(1. To avoid degenerate cases. we will assume that B = 0 and αB − βA = 0. (1. A + Bxn B The set of initial conditions x0 ∈ R through which the denominator A+Bxn in Eq(1. β.20) . 1.17) reduces to βA + βxn β xn+1 = B = for all n ≥ 0.18) Indeed when B = 0.18) holds and that Eq(1.17) possesses a prime periodtwo solution.17). The proof is straightforward and will be omitted. Eq(1.21) β B (1. γ. A.1. we will assume that β + A = 0.17) will become zero for some value of n ≥ 0 is called the forbidden set F of Eq(1. In the sequel.19) holds every solution of Eq(1. One of our goals in this section is to determine the forbidden set F of Eq(1. B. which is extracted from [32].

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

Also w− < 0 if R < 0 and 1 w− > 0 if R ∈ (0. . 1. 4 1− √ √ 1 − 4R 1 + 1 − 4R < = w+ 2 2 19 When Eq(1.24) are the equilibrium points of Eq(1. The forbidden set of Eq(1.23) to the second order linear diﬀerence equation yn+2 − yn+1 + Ryn = 0. respectively.23) has no equilibrium points. . 1 w= .23) has the two equilibrium points w− and w+ with w− = and |w− | < w+ .23) is clearly the set of points where the sequence {yn }∞ n=0 vanishes. which can be solved explicitly. 2 Finally.23). (1. . 4 1 R= . 4 Eq(1. . Note that in these two cases the characteristic roots of Eq(1. . 1. Now observe that the change of variables wn = with y0 = 1 and y1 = w0 reduces Eq(1. 1 The next two theorems deal with the cases R < 1 and R = 4 . n = 0.24) yn+1 yn for n = 0.1. . ).6.23) has exactly one equilibrium point namely. where 4 the characteristic roots of Eq(1. . when 1 R> .24) are real numbers. The Riccati Equation Eq(1.

.23) is the sequence of points fn = n−1 2n for n = 1.2 Assume Then 1 R< . (1.6.25) When w0 ∈ F. The equilibrium w+ is a global attractor as long as w0 ∈ F ∪ {w− } / and the equilibrium w− is a repeller. 4 1 2 is the only equilibrium point of Eq(1. .6. .26) It is now clear from Theorem 1.3 Assume Then w= 1 R= .23).23) is the sequence of points fn = n−1 n−1 w+ − w− w+ w− n n w+ − w− for n = 1. . .23) is given by / n+1 n+1 (w0 − w− )w+ − (w+ − w0 )w− . 2. 2.23). if R > 0 and fn < w− while (fn − w− )(fn+1 − w− ) < 0 if R < 0. . . . (1. 4 1− Chapter 1. The forbidden set F of Eq(1. .6. The forbidden set F of Eq(1. Preliminary Results √ √ 1 − 4R 1 + 1 − 4R w− = and w+ = 2 2 are the only equilibrium points of Eq(1. 2. . (1. Theorem 1.27) .20 Theorem 1. . wn = n n (w0 − w− )w+ − (w+ − w0 )w− n = 1. .2 that n→∞ lim fn = w− . the solution of Eq(1.

2 R n = 0.3 it follows that the equilibrium w = 1 of Eq(1. . . . π φ ∈ (0. yn = R 2 (C1 cos(nφ) + C2 sin(nφ)). R 4R − 1 cos(nφ) + (2w0 − 1) sin(nφ) .29) n Finally consider the case √ R. 1. 1 R> . ) 2 and √ 4R − 1 √ sin φ = .6. 4R − 1 n = 0. 1.24) 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 . (1. . The Riccati Equation 21 which converges to the equilibrium from the left. . 2w0 − 1 C1 = 1 and C2 = √ . more precisely. . When w0 ∈ F. . (1.23) is given by / wn = 1 + (2w0 − 1)(n + 1) 2 + 2(2w0 − 1)n for n = 0. 4 The characteristic roots of Eq(1. the solution of Eq(1. .23) is a global 2 / attractor for all solutions with w0 ∈ F but is locally unstable. . Hence wn = √ √ 4R − 1 cos(n + 1)φ + (2w0 − 1) sin(n + 1)φ √ . 1.6.28) From Theorem 1. .1. it is a sink from the right and a source from the left.

. 2. Theorem 1. 2 R Then the forbidden set F of Eq(1.32) . . . ) N 2 (1. wn = (4R − 1)2 + (2x0 − 1)2 . . 1. .22 Chapter 1. / (1. 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). . . that is. For example.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. . π ) be such that 2 1 cos φ = √ 2 R and 1 4 √ 4R − 1 √ sin φ = .4 Assume that R> and let φ ∈ (0. 2. the solution of Eq(1. (1. w0 = − 2 2 The following theorem summarizes the above discussion.23) is given by (1. . Preliminary Results as long as the denominator is diﬀerent from zero. assume φ is a rational multiple of π. when R > 1 . π ) is a rational multiple of π. . √ 4R − 1 1 cot(nφ) for n = 1.29).6. φ= π k π ∈ (0. that is.23). π ) be such that 2 2 √ 4R − 1 2x0 − 1 and sin θ = cos θ = r r where r= Then we obtain. 2 Let θ ∈ (− π .23) is the sequence of points √ 1 4R − 1 fn = − cot(nφ) for n = 1. 2 n = 0. . can now be easily 4 revealed.31) from which the character of the solutions of Eq(1. 2 2 When w0 ∈ F.

. . . 2 R 2 R or equivalently k 4R cos2 ( π) = 1.1. Then the following statements are true: (i) No solution of Eq(1.23) is periodic with period N . The following theorem summarizes the above discussion. 2.5 Assume that φ= k π π ∈ (0.6. N ∈ {1. .6 Assume that the number φ in Theorem 1. then every solution of Eq(1. N ∈ {1. Theorem 1. (ii) The set of limit points of a solution of Eq(1. ) 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 = .6. .23) is periodic. p − 1 .5 is not a rational multiple of π.6. then the following result is true: Theorem 1. .5 is not a rational multiple of π. When the number φ in Theorem 1. . . / for n = 1.6.23) with w0 ∈ F is dense in R.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.} with 2k < N.} and 2k < N. The Riccati Equation 23 for some k. . p Then every solution of Eq(1. √ √ 1 1 4R − 1 4R − 1 wN = − tan(kπ − θ) = − tan(−θ) = w0 2 2 2 2 and so every solution of Eq(1.23) is periodic with period N . . 2. Equivalently assume that A= where k. 2.6.

.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. xn−1 ). A positive semicycle of a solution {xn } of Eq(1. . (1. with l ≥ −1 and m ≤ ∞ and such that either l = −1. . wk = wl if and only if Chapter 1. . all greater than or equal to the equilibrium x. . π ). are dense in the interval (− π . xl+1 . .24 Proof. The proof of statement (ii) is a consequence of the form of the solution of Eq(1. or l > −1 and xl−1 < x and either m = ∞. then the values (nφ − θ)(mod π) for n = 0. . Statement (i) is true because.1) consists of a “string” of terms {xl .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. φ is a rational multiple of π. 1. . n = 0. xm }. 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. . 2 2 2 1. or m < ∞ and xm+1 < x. In this section. .23) and the fact that when φ is not a rational multiple of π.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 . 1.

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. xN −1 ) < f (x. and f (x.33) has semicycles of length one.7. n2 ≥ n0 such that xn1 xn2 < 0.1 ([28]) Assume that f ∈ C[(0. Let x be a positive equilibrium of Eq(1. Then except possibly for the ﬁrst semicycle.33) consists of a “string” of terms {xl . All other cases are similar and will be omitted. . y) is increasing in y for each ﬁxed x.7. The sequence {xn } is called strictly oscillatory about x if the sequence xn − x is strictly oscillatory. Proof. The ﬁrst result was established in [28] while we were working on a special case of the equation we are investigating in this monograph. . Otherwise the sequence is called nonoscillatory. (See Section 6. xm }.) Theorem 1. every solution of Eq(1. x) = x and xN +2 = f (xN +1 .1. ∞)] is such that: f (x. xl+1 . Let {xn } be a solution of Eq(1. or l > −1 and xl−1 ≥ x and either m = ∞. . . We will assume that xN −1 < x ≤ xN . Deﬁnition 1. Semicycle Analysis 25 A negative semicycle of a solution {xn } of Eq(1.33). y) we have xN +1 = f (xN .5. Then there exists N ≥ 0 such that either xN −1 < x ≤ xN or xN −1 ≥ x > xN . Then by using the monotonic character of f (x. x) = x. . xN ) > f (x. there exist n1 . all less than the equilibrium x. (b) A sequence {xn } is said to oscillate about x if the sequence xn − x oscillates. Thus xN +1 < x < xN +2 or m < ∞ and xm+1 ≥ x. ∞) × (0. ∞). (0.7. y) is decreasing in x for each ﬁxed y. A sequence {xn } is called strictly oscillatory if for every n0 ≥ 0.33) with at least two semicycles. with l ≥ −1 and m ≤ ∞ and such that either l = −1.

2 The next result applies when the function f (x. ∞). Let x be a positive equilibrium of Eq(1. ∞) × (0. ∞) × (0. (0.33). Preliminary Results 2 The next result applies when the function f is decreasing in both arguments. 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. Then by using the increasing character of f we obtain: xN +1 = f (xN . ∞). (0.7. xN −1 ) > f (x. The proof in the case of negative semicycle is similar and is omitted. Then every oscillatory solution of Eq(1. y) is decreasing in both arguments. Then by using the decreasing character of f we obtain: xN +1 = f (xN . y) is increasing in both arguments.33). which is a contradiction.33) has semicycles of length one. y) is increasing in x and decreasing in y. x) = x which shows that the next term xN +1 also belongs to the positive semicycle.7.3 Assume that f ∈ C[(0. 2 . Let x be a positive equilibrium of Eq(1. Theorem 1. Theorem 1. The proof in the case of negative semicycle is similar and is omitted. with at least one of the inequalities being strict.2 Assume that f ∈ C[(0. x) = x which completes the proof. It follows by induction that all future terms of this solution belong to this positive semicycle. Chapter 1. Proof. ∞)] and that f (x.26 and the proof follows by induction. every oscillatory solution of Eq(1. Then except possibly for the ﬁrst semicycle. The next result applies when the function f is increasing in both arguments. 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. ∞)] and that f (x. Proof.33) has semicycles of length at most two. xN −1 ) < f (x.

and f (x. y) < x for every x > y > x (1. Assume that {xn } is an oscillatory solution with two consecutive terms xN −1 and xN such that xN −1 < x ≤ xN . xN −1 ) < f (xN . xN −1 ) > f (xN .35) then {xn } oscillates about the equilibrium x with semicycles of length two or three. The proof in the case xN −1 ≥ x > xN . Semicycle Analysis 27 Theorem 1. ∞) × (0.34) we obtain xN +1 = f (xN . Let x be a positive equilibrium of Eq(1. except possibly for the ﬁrst semicycle. (0.34) . The extreme in each semicycle occurs in the ﬁrst term if the semicycle has two terms. is similar and is omitted. xN ) = x (1. if we assume that f (u. ∞). except possibly for the ﬁrst semicycle which may have length one.33) has semicycles of length at least two. and in the second term if the semicycle has three terms.1.4 Assume that f ∈ C[(0. every oscillatory solution of Eq(1. xN −1 ) > f (x.7. Then by using the increasing character of f we obtain xN +1 = f (xN . x) = x which shows that the next term xN +1 also belongs to the positive semicycle.7. xN ) = x which shows that the positive semicycle has length two. 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. ∞)] is such that: f (x. Proof. u) = x for every u and f (x. Then by using the increasing character of f and Condition (1.34) we obtain: xN +1 = f (xN . If xN > xN −1 > x then by using the increasing character of f and Condition (1. y) is increasing in x for each ﬁxed y. Then. y) is decreasing in y for each ﬁxed x. Furthermore.33).

7. assume that Condition (1. If. Theorem 1. ∞). f satisﬁes Condition (1. we assume (1. y) is increasing in x for each ﬁxed y. y) < x for every x < y < x then if x0 < x−1 < x. then Eq(1. 2 Condition (1. and the proof follows by induction.35). on the other hand. the corresponding solution is decreasing.34). instead of Condition (1.36) is satisﬁed and that x0 > x−1 > x.28 and by using Condition (1.34). In view of Condition (1.35) we ﬁnd Chapter 1. Proof. First.36). x−1 ) = x. ∞)] is such that: f (x. y) > x for every x > y > x. ∞) × (0.36) we get x1 = f (x0 . Assume that f satisﬁes Conditions (1.33). y) is decreasing in y for each ﬁxed x. we obtain x1 = f (x0 . x−1 ) > f (x−1 . Then by the monotonic character of f and (1. Preliminary Results xN +1 = f (xN .36) f (x. Then if x0 > x−1 > x. x−1 ) > x0 .5 Assume that f ∈ C[(0. the corresponding solution is increasing.34) and (1. 2 (1. and f (x. If.34) is not enough to guarantee the above result. and f (x. xN −1 ) < xN and the proof is complete.37) .33) has monotonic solutions. Let x be a positive equilibrium of Eq(1. The proof of the remaining case is similar and will be omitted. (0.

γ. Zero is an equilibrium point of Eq(2.1) if and only if α = 0 and A > 0. . ∞) (2. A. Eq(2.1) are the nonnegative solutions of the equation x= or equivalently (B + C)x2 − (β + γ − A)x − α = 0. . 29 .Chapter 2 Local Stability. Periodicity.5) (2.1) has a positive equilibrium if and only if β + γ > A.1 Equilibrium Points α + βxn + γxn−1 .3) n = 0. β. . and Invariant Intervals 2. (2. ∞) with α + β + γ. B + C ∈ (0. Semicycles. the equilibrium points of Eq(2. in addition to the zero equilibrium .4) α + (β + γ)x A + (B + C)x (2. 1. B.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). A + Bxn + Cxn−1 Here we investigate the equilibrium points of the general equation xn+1 = where α.1) is well deﬁned for all n ≥ 0. In view of the above restriction on the initial conditions of Eq(2. C ∈ [0.5) holds.1) When (2. (2.

and Invariant Intervals In fact in this case the positive equilibrium of Eq(2. B+C (2. B+C (2. and [69] for some results in this regard. (See [13].1) is unique and is given by x= When α = 0 and A = 0 the only equilibrium point of Eq(2. n = 0. v) = A + Bu + Cv and observe that (βA − αB) + (βC − γB)v fu (u. . Finally when α>0 the only equilibrium point of Eq(2. A + Bxn there is very little known. 2.1) is the positive solution x= β+γ−A+ (β + γ − A)2 + 4α(B + C) 2(B + C) (2. The problem of investigating Eq(2. . then x is unique.4) would also be an equilibrium worth investigating.8). Set α + βu + γv f (u. This observation simpliﬁes the investigation of the local stability of the positive equilibrium of Eq(2.6) Note that in view of Condition (2.3) and (2.1) has a positive equilibrium x.4). except for our presentation in Section 1. If we had allowed negative initial conditions then the negative solution of Eq(2.2).7) β+γ−A . when α = 0. 1.30 Chapter 2.1) when the initial conditions and the coeﬃcients of the equation are real numbers is of great mathematical importance and. Periodicity.8) of the quadratic equation (2. v) = (A + Bu + Cv)2 (2. [18].1).) In summary.1) is positive and is given by x= β+γ . Semicycles. . it is interesting to observe that when Eq(2.6 about the Riccati equation α + βxn xn+1 = . [32]. the quantity β + γ is positive. it satisﬁes Eqs(2. and it is given by (2.2 Stability of the Zero Equilibrium Here we investigate the stability of the zero equilibrium of Eq(2.9) . .1).4). Local Stability.

in addition to the local stability Theorem 1. 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. . Then the zero equilibrium point of the equation xn+1 = βxn + γxn−1 . . x).1 Assume that A > 0. A A For the stability of the zero 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. Stability of the Zero Equilibrium and fv (u. 1. n = 0. For the zero equilibrium of Eq(2. v) = (γA − αC) − (βC − γB)u . .1. .1). then the linearized equation associated with Eq(2. (A + Bu + Cv)2 31 (2.3.10) If x denotes an equilibrium point of Eq(2.1. x) and q = fv (x.1. The local stability character of x is now described by the linearized stability Theorem 1. .2. . 1.1). .2.1) about the equilibrium point x is zn+1 − pzn − qzn−1 = 0 where p = fu (x.1.2. we have the luxury of the global stability Theorem 1.1) about the zero equilibrium point is β γ zn+1 − zn − zn−1 = 0. A + Bxn + Cxn−1 n = 0. is globally asymptotically stable when β+γ ≤A and is unstable when β + γ > A. .1.

1) has a positive equilibrium when either or α > 0.11) and q > −1.15).11) and (2. Inequalities (2.32 Chapter 2. it satisﬁes Eqs(2.1. In these cases the positive equilibrium x is unique.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. Periodicity. 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. and Invariant Intervals 2.13)-(2.3 Local Stability of the Positive Equilibrium α = 0 and β + γ > A As we mentioned in Section 2. Semicycles.14) (2. By using the identity (B + C)x2 = (β + γ − A)x + α. we see that p = fu (x.12) (A + β + γ)x > −α − (2.15) A2 + (γA − αC) βC − γB x > −α − . x) = (γA − αC) − (βC − γB)x (γA − αC) − (βC − γB)x = 2 .4).3) and (2. Local Stability.1.8). 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. and it is given by (2.13) (2. are equivalent to some inequalities of the form x>ρ . Eq(2. x) = and q = fv (x.1) the following: |p| < 1 − q (2.

n = 0. The following four special examples of Eq(2. . Then γ(α + β) = 0.1) with the property that every solution of the equation is periodic with the same period. . 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. .19) 1 + xn . n = 0. n = 0. 33 2.1) xn+1 = xn+1 = xn+1 = xn+1 n = 0. Now if we set F (u) = (B + C)u2 − (β + γ − A)u − α.4. . 1.16) (2. . . . . and [74]. . . xn−1 xn . 1.1 Let p ≥ 2 be a positive integer and assume that every positive solution of Eq(2. 1. Then A = B = γ = 0. and 6 respectively.2. (ii) Assume C = 0. (2. .4 When is Every Solution Periodic with the Same Period? 1 . When is Every Solution Periodic with the Same Period? and/or x<σ for some ρ. The following result characterizes all possible special cases of equations of the form of Eq(2.) Theorem 2.18) (2.1) is periodic with period p. σ ∈ [0. 1. ∞).4. 4. xn 1 xn−1 . .17) (2. [73]. 5. Then the following statements are true: (i) Assume C > 0. = xn−1 are remarkable in the sense that all positive solutions of each of these four nontrivial equations are periodic with periods 2. (See also [75]. .

Then up to a change of variables of the form xn = λyn Eq(2. Local Stability. Semicycles. 3. Then by choosing x0 suﬃciently small.1) reduces to one of the equations (2.20) α + β + γxp−2 = x0 .21) (2. assume that γ(α + β) > 0.21) holds. Consider the solution with x−1 = 1 and x0 ∈ (0. The proof is complete. ∞). 5. we see that (2.19) . Periodicity.1) xp = Hence (A + B)x0 + (Cx0 − γ)xp−2 = α + β.22) holds . (ii) Assume C = 0 and.20) is impossible. . A + B + Cxp−2 and xp−1 = x−1 = 1 we see that (2. for the sake of contradiction. (2. Hence (2. Then we claim that A = B = 0.20) is impossible.1 Let p ∈ {2. and Invariant Intervals Proof. 2 Corollary 2. Thus (2. Otherwise A + B > 0 and by choosing x0 > max α+β γ .4. 4.1) is periodic with period p.16)-(2.22) Otherwise γ > 0 and by choosing x0 < min α+β γ .20) is impossible.34 Chapter 2. A+B C we see that (2.21) we now claim that also γ = 0. Then clearly x0 = xp and so from Eq(2. A+B C (2. (i) Assume that C > 0. 6}. Assume that every positive solution of Eq(2. In addition to (2.

ψ. ψ. . . Assume that . ψ ∈ [0. φ. not necessarily prime. .25) (2. C (2. φ. . .27) (2.) Furthermore we are interested in conditions under which every solution of Eq(2.5. solution with period p and furthermore the set of solutions of the equation with prime period p is nonempty.2.1) has a prime period-two solution if and only if γ = β + A. Furthermore when (2. ψ. φ. .27) hold .1) converges to a period-two solution in a nontrivial way according to the following convention.26) and (2. . ∞) and φ = ψ (2. . is a prime period-two solution of Eq(2. is a prime period-two solution of Eq(2.1) to have a prime period-two solution and we exhibit all prime period-two solutions of the equation. (See [50]. 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. . φ.26) αC + β(γ − β − A) . 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. .1). .23) α + βψ + γφ A + Bψ + Cφ α + βφ + γψ A + Bφ + Cψ .5 Existence of Prime Period-Two Solutions In this section we give necessary and suﬃcient conditions for Eq(2.24) (2.1) if and only if α + β(φ + ψ) = Bφψ with φ. Existence of Prime Period-Two Solutions 35 2.

31) γ > β + A and When (2.6 Let Local Asymptotic Stability of a Two Cycle . be a two cycle of the diﬀerence equation (2. C γ−A γ−A . . B > C. C C with 0 ≤ φ < γ−A .30) (2. the two cycle is . Periodicity. . . B = C. α + β > 0.31) holds..36 Chapter 2. φ. φ. When (2. . .. (2. 1. and φψ ≥ 0. .32) 4C 2 2. . 0. . α = β = 0. ψ. . B and ψ = α + βφ .30) holds. . . the two cycle is . . C C Finally when (2. Semicycles. φ. . . B φ= β+ √ β 2 + αB . φ. − φ. . . . .1). γ−A γ−A − φ. and φψ = 0. 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. φ. . . .29) holds. . and φψ > 0. Then it follows from (2. −β + Bφ (2. . ψ. 0. . α = β = 0. ψ. and Invariant Intervals or equivalently β φ> . φ.28) Next assume that (2.1) possesses a prime period-two solution .23) and (2.. ψ. (2. Set un = xn−1 and vn = xn for n = 0. (γ − β − A)[B(γ − β − A) − C(γ + 3β − A)] α< . . .29) (2. .25) that one of the following three conditions must be satisﬁed: B = C.24) holds and that Eq(2. Local Stability. That is. .

. JT 2 . ψ ∂h ∂h (φ. Local Asymptotic Stability of a Two Cycle and write Eq(2.6. evaluated at ψ By deﬁnition ∂g ∂g (φ. . ∞) deﬁned by: T Then u v = v βv+γu+α Bv+Cu+A 37 . φ ψ is a ﬁxed point of T 2 . n +Cun Let T be the function on [0. ψ) = . ψ) = . ψ) ∂u φ = JT 2 . ∞) × [0. v) α+βv+γu α + β A+Bv+Cu + γv α+βv+γu A + B A+Bv+Cu + Cv . ∂u (A + Bψ + Cφ)2 ∂g (βA − αB) − (γB − βC)φ (φ. 1. ψ) = . Furthermore T2 where g(u. v) = α + βv + γu A + Bv + Cu and h(u. . .1) in the equivalent form: un+1 = vn α+βv vn+1 = A+Bvn +γun . v) h(u. ∂u (A + Bφ + Cψ)2 (A + Bψ + Cφ)2 . v) = u v = g(u. ψ) ∂v (φ.2. n = 0. 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) ∂v (φ. the second iterate of T . ψ) ∂u we ﬁnd the following identities: (γA − αC) + (γB − βC)ψ ∂g (φ. ∂v (A + Bψ + Cφ)2 [(βA − αB) − (γB − βC)ψ][(γA − αC) + (γB − βC)ψ] ∂h (φ.

and 10.33) converges to a period-two solution. . . 6. and 7. ψ) (φ. β.33) .7. β.9. x0 ) through which the solution of Eq(2. ∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1. A + B ∈ (0. ψ) + (φ. every solution converges to a (not necessarily prime) period-two solution and n = 0.5. A.1. ψ). ∞) and nonnegative initial conditions and obtain a signum invariant for its solutions which will be used in the sequel (See Sections 4. Semicycles. That is. φ. γ. . . Deﬁnition 2. ψ) − (φ. . ψ) (φ. A + Bxn In this section we consider the equation xn+1 = where α. 1. . γ. The basin of attraction of this two cycle is the set B of all initial conditions (x−1 .7 Convergence to Period-Two Solutions When C=0 α + βxn + γxn−1 . See Sections 6. φ. .7. B ∈ [0.4 for some applications. be a two cycle of Eq(2.1 (c) that both eigenvalues of JT 2 unit disk |λ| < 1. 6. ψ. ∞) with α + β + γ. . ψ) ∂u ∂v ∂h ∂g ∂h ∂g (φ. + (A + Bφ + Cψ)2 Set S= and D= ∂h ∂g (φ. ψ) = ∂v (A + Bφ + Cψ)2 (A + Bψ + Cφ)2 (γA − αC) + (γB − βC)φ .2) to obtain conditions on α. (2. 2. 6.1).1) converges to the two cycle.6. ψ. A and B so that every solution of Eq(2.6. Periodicity.1 (Basin of Attraction of a Two Cycle) Let . 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. Local Stability.38 Chapter 2.

27) holds. ∞ Lemma 2. if and only if. Set (2.2 where we present the complete character of solutions of Eq(2. one and only one of the following three statements is true for each solution of Eq(2. Statement (ii) holds if and only if J0 < 0 and β +A > 0.1 Assume that (2. Jn+1 = (b) xn+1 − xn−1 = Jn A + Bxn for n ≥ 0.7. the solution would converge to a period-two solution.33): (i) xn+1 − xn−1 = 0 for all n ≥ 0. Now observe that xn+1 − xn−1 = β+A (xn − xn−2 ) for n ≥ 1 A + Bxn .2. the proof of which follows by straightforward computations. 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. Then the following statements are true: (a) β+A Jn for n ≥ 0. if and only if J0 > 0 and β + A > 0.33) converges to a periodtwo solution if and only if (2. the sign of the quantity Jn is constant along each solution. (ii) xn+1 − xn−1 < 0 for all n ≥ 0. (See also Section 10.) Now before we see that when B > 0 every solution of Eq(2. (β + A)J0 = 0. A + Bxn In particular.34) Jn = α + β(xn−1 + xn ) − Bxn−1 xn for n ≥ 0.7. Finally (iii) holds. In this case.27) holds. we need the following result. Clearly (i) holds. Convergence to Period-Two Solutions When C=0 39 there exist prime period-two solutions. (iii) xn+1 − xn−1 > 0 for all n ≥ 0. In particular. 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.33). the solution {xn } is periodic with period-two. and let {xn }n=−1 be a solution of Eq(2.33).

Then there exists N ≥ 0 such that ρ= Clearly for all n > N . In summary we have established the following result.33) has a prime period-two solution if and only if (2. Semicycles. . . Local Stability. of Eq(2. Periodicity. Theorem 2. φ. A + Bx2n+1 A + Bx2n A + Bx2N +1 A + Bx2N Set Then K = |x2N − x2N −2 |.35) We will now employ (2. . . A+β A+β A+β A+β < = ρ. |x2N +2 − x2N | = A+β A+β < 1. for the sake of contradiction. φ. that the subsequence of even terms increases to ∞. (b) When (2.27) holds all prime period-two solutions . (c) When (2. 2.33) are given by (2. The proof for the subsequence of odd terms is similar and will be omitted. ≤K ρi + x2N < i=1 which contradicts the hypothesis that the subsequence of even terms is unbounded.28).27) holds.35) to show that the subsequences of even and odd terms of the solution are bounded. . + |x2N +2 − x2N | + x2N m for m = 1. . We will give the details for the subsequence of even terms. .27) holds every solution of Eq(2. . Then the following statements hold: (a) Eq(2. . ψ. ψ. and Invariant Intervals n xn+1 − xn−1 = (x1 − x−1 ) β+A . To this end assume. .7.33) converges to a period-two solution.40 and so Chapter 2.1 Assume B > 0. . . 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. Kρ + x2N 1−ρ .

y) = N − Mx . xn−1 ). and N = γA − αC.10) and are fx (x. In other words I is an invariant interval for the diﬀerence equation xn+1 = f (xn . xN ∈ I for some N ≥ 0. then xn ∈ I for every n > N . Invariant Intervals 41 2. we obtain the following table of invariant intervals for Eq(2.1) 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. 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 fx < 0 N N fy > 0 if x > M . (A + Bx + Cy)2 − + − − − − − − Using this table. The following table gives the signs of fx and fy in all possible nondegenerate cases.1). n = 0.8 Invariant Intervals An invariant interval for Eq(2.8. 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 . 1. Invariant intervals for Eq(2. For Eq(2. fy < 0 if x < M L + My (A + Bx + Cy)2 and fy (x. Signs of derivatives fx and fy fx > 0 N N fy > 0 if x < M . . . (2.2.9) and (2. y) is monotonic in its arguments.36) if xN −1 . . . y) = where L = βA − αB.1) the partial derivatives are given by (2. 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.36) are determined from the intervals where the function f (x.

C ] if C > 0 [k. Local Stability. ] if B > 0 2B α [0. (A−β)M +CN ] if B = 0 and (A − β)M + CN > 0 √ β−A+ (β−A)2 +4Bα [0. C ] if C > 0 β β2 [ B . α > 0 and A > β √ β+γ−A+ (β+γ−A)2 +4α(B+C) [0. K] if C = 0 where k and K satisfy α + βk + (γ − A)K ≤ BkK ≤ α + βK − Ak γ [0. A−β ] if B = 0. 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. C ] if C > 0 [k. B + α ] if A = 0 and β > 0 β α [0. B(β−γ) ] if A = 0 and β > γ αM +(β+γ)N N αM +(β+γ)N N [ AM +(B+C)N . − M ] if L [− M . A−β ] if B = 0. ] if B > 0 2B α [0. L [− M . C ] if C > 0 β αB+β 2 [ B . ] if AM −BL+CMK > − M 2BM K α [ α(A+β) . ] if B > 0 2B α α [ A . 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. A ] if A > 0 A2 +Bα β β [ B . ] if B + C > 0 √ 2(B+C) 2 +4αB α β−A+ (β−A) [A. M ] if AM +(B+C)N < M √ 2 N (γ−A)M −BN + ((γ−A)M −BN ) +4BM (αM +βN ) [M . 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 . − 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 . Semicycles. A ] if A > 0 [k. α > 0 and A > β √ β−A+ (β−A)2 +4Bα [0. ] if AM +(B+C)N > 2BM αM +γN N [ M . Periodicity. 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. ] if 2CM αM +βK+γN M AM +BK+CN M > N M . (γ − A)M + BL > 0 N M and B > 0 6 7 8 9 10 11 12 13 14 15 16 γ β [ C . and Invariant Intervals Case 1 2 3 4 5 Invariant intervals αM +(β+γ)N N N [0. K] if C = 0 where k and K satisfy α + βk + (γ − A)K ≤ BkK ≤ α + βK − Ak β γ [ B . B ] if B > 0 γ γ2 [ C . B ] if B = 0 2 γ [ C . B(β−γ) ] if C = 0 and β > γ αM −L(β+γ) αM −L(β+γ) L L [ AM −L(B+C) . A−β ] if B = 0 and A > β L [0.42 Chapter 2.

∞). ∞) × (0. n = 0.5. Open Problem 2. . 1. ∞) × [0. γ. 1. . . if and only if every root of the characteristic equation is a pth root of unity. 1 + xn xn + 2xn−1 . Open Problems and Conjectures 43 2. [0. . every solution is periodic with period p ≥ 2.9. β. Find necessary and suﬃcient conditions so that every nonnegative solution of the diﬀerence equation (2. xn n = 0. 1. (2. In particular address the case p = 2.7) that every positive solution of each of the following three equations xn+1 = 1 + xn+1 = xn+1 xn−1 .38) (2.37) is periodic with period p ≥ 2.2 Let f ∈ C 1 [[0. ∞)] is such that every positive solution of the equation xn+1 = f (xn . 1. ∞).2.39) (2. B. n = 0. .9. 5.37) converges to a periodic solution with period p. Open Problem 2. . (2. 6}? Open Problem 2. Is it true that p ∈ {2. .3 Let p ≥ 2 be an integer and assume that every positive solution of Eq(2. . 4. Is it true that the linearized equation about a positive equilibrium point of Eq(2. . 4. 6. ∞) converges to a solution with period p .9. and 6.9. = 1 + xn . C ∈ [0. . ∞)] and let p ≥ 2 be an integer.9.40) 1 + xn−1 . (0.7.1) with α.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 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. n = 0.1 Assume that f ∈ C 1 [(0. xn−1 ).7.4 It is known (see Sections 2. A. .

4. .44 Chapter 2. ψ. .9. Conjecture 2.33) is bounded. Show that the positive equilibrium is globally asymptotically stable.41). (See Sections 2.40). C ∈ (0. . . Semicycles. 6.9. . B. x0 ) ∈ (0.9. if (2. β.7. and Invariant Intervals converges to a solution with (not necessarily prime) period-two: . C ∈ (0. .1) has a positive prime period-two solution.38) or Eq(2.1) is a saddle point.41) In each case. ψ.7. φ. determine φ and ψ in terms of the initial conditions x−1 and x0 . ∞) and γ > β + A. γ. (2. B.1) has no period-two solution. ∞) and that Eq(2. and 10. C ∈ (0.41) is a period-two solution of Eq(2. B ∈ (0. Determine the set of initial conditions x−1 and x0 for which the solution {xn }∞ n=−1 of Eq(2. B.2 Assume α. β. Show that the positive equilibrium of Eq(2. A.5 Assume α. . ∞) and that Eq(2. Open Problem 2.1 Assume α. β. 6. γ. determine all initial conditions (x−1 . A.39) or Eq(2. γ.1) is bounded.) Conjecture 2.2.9. β. ∞).3 Assume α. Conjecture 2. Show that every positive solution of Eq(2. A. A. Local Stability.5. ∞) × (0. ∞) for which the solution {xn }∞ n=−1 converges to the period-two solution (2. φ. Conversely. Periodicity. . γ.7.

9.9. . (0. Obtain necessary and suﬃcient conditions on f for every positive solution of the equation xn+1 = 1 + to converge to a period-two solution. . (0. What additional conditions should the parameters satisfy? . ∞)].1 to third order diﬀerence equations xn+1 = α + βxn + γxn−1 + δxn−2 . . Open Problem 2. ∞).2. (See Section 4.7. β. A + Bxn−1 + Dxn−2 n = 0. Open Problems and Conjectures Open Problem 2.9. 45 Obtain necessary and suﬃcient conditions in terms of the coeﬃcients so that every positive solution of Eq(2. 1. ∞). . γ. Obtain necessary and suﬃcient conditions on f for every solution of the diﬀerence equation xn+1 = to be bounded.9. B.9. Open Problem 2.9 Extend Theorem 2. β. Open Problem 2. γ. 1.7). 1. ∞). xn−1 n = 0. f (xn−1 ) .1) cannot have a positive prime two cycle which attracts all positive non-equilibrium solutions. A. . (See Section 5. C ∈ [0. Conjecture 2. C ∈ [0. ∞). with nonnegative parameters and nonnegative initial conditions such that γ = β + δ + A. xn n = 0.2). .9.4 Assume that α.6 Assume α. .7 Assume f ∈ C[(0. ∞)]. .1) is bounded. . Show that Eq(2.8 Assume f ∈ C[(0. B. f (xn ) . A.

. Periodicity. . n = 0. 1. xn−3 ).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 .9. . Semicycles. .10 Extend the linearized stability Theorem 1. xn−1 . xn−1 . xn−2 . .46 Chapter 2. xn−2 ).1 to third order difference equations xn+1 = f (xn . or a saddle point. Local Stability.1. But how about necessary and suﬃcient conditions for the equilibrium of Eq(2. n = 0. . 1. .42) to be a repeller. . . or nonhyperbolic? Extend these results to fourth order diﬀerence equation xn+1 = f (xn . (2. and Invariant Intervals Open Problem 2.

1)-Type Equations 3. . B n = 0. . . . Cxn−1 47 n = 0. Bxn n = 0.4) xn+1 = β . . 1. Cxn−1 n = 0. (3.3) xn+1 = βxn . 1. 1. . . . (3. 1)-type: xn+1 = n = 0.1) xn+1 = α . . (3. . .6) . . (3. .5) xn+1 = βxn . . 1. 1.2) xn+1 = α . A Eq(1) contains the following nine equations of the (1.1 Introduction α . (3. (3. . . . . A n = 0.Chapter 3 (1. 1.

y0 . Eqs(3. namely. (3. 1. . and the denominators are always positive. . . the initial conditions are nonnegative..7) are linear diﬀerence equation. .10) has prime period six.6) and (3. (1. Eqs(3. (3.9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. . (3. 1. Indeed the solution with initial conditions y−1 and y0 is the six-cycle: y−1 . A n = 0. y0 1 1 y−1 . 3. .2 The Case α = γ = A = B = 0 : xn+1 = β yn C βxn Cxn−1 This is the (1.8) xn+1 = γ . . 1.4) and (3. y−1 y−1 y0 y0 It is interesting to note that except for the equilibrium solution y = 1. .7) xn+1 = and γxn−1 . (3. The remaining two equations. Bxn n = 0. are treated in the next two sections. yn−1 n = 0. 1. every other solution of Eq(3. . . 1)-type Eq(3. .9) are trivial.8). .1).6) which by the change of variables xn = reduces to the equation yn+1 = yn .5).10) is periodic with period six. .48 Chapter 3.. 1)-Type Equations xn+1 = γxn−1 . . .10) It is easy to see that every positive solution of Eq(3. . . Eqs (3. (3. C n = 0..2) is periodic with period two and every solution of Eq(3. and (3. . Of these nine equations. Every solution of Eq(3.4) is periodic with period four. .

12) . xn−1 ). 1.7) is periodic with period two but this is a linear equation.3. . The Case α = β = A = C = 0 : xn+1 = γxn−1 Bxn 49 γxn−1 Bxn 3. . (3.11) + C2 −1 − 2 √ 5 n .3 The Case α = β = A = C = 0 : xn+1 = γ yn e B This is the (1. ∞) × (0. . 1. The general solution of Eq(3.14) (3.6). 1. Eq(3. (0. . 1. (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. .8) which by the change of variables xn = reduces to the linear equation yn+1 + yn − yn−1 = 0. . namely two and six. which possess the property that every positive nontrivial solution of each of these two equations is periodic with the same prime period.4 Open Problems and Conjectures Of the nine equations in this chapter we would like to single out the two nonlinear equations. where f ∈ C[(0. ∞)]. 3. n = 0. 1)-type Eq(3.3. ∞). ∞)]. (3.2).11) is yn = C1 −1 + 2 √ 5 n n = 0.2) is of the form xn+1 = f (xn ). where f ∈ C[(0. . (3.13) n = 0.2) and (3. n = 0. ∞). . . Also every solution of Eq(3. Eq(3. . (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. respectively. (3. . . . from which the behavior of solutions is easily derived. while Eq(3.6) is of the form xn+1 = f (xn .

∞). (1. Show that f (x) = x. γ ∈ (−∞.1 Let k ≥ 2 be a positive integer and assume that (3.4. 6.4. 1.14) is periodic with period k. . and [59]. k = 5. Open Problem 3. Show that f (x) = 1 + x. β.15) holds. . 1. 7. γ so that every solution of the equation xn+1 = α |xn | + βxn−1 + γ. (3. xn−1 n = 0.16) is periodic with prime period ﬁve. ∞)] be such that every positive nontrivial solution of the diﬀerence equation (3. . β. In particular address the cases: (See [10].12) is periodic with period k.4. with real initial conditions is periodic with period k. 1)-Type Equations Open Problem 3. [17].16) is periodic with prime period six.1 Let f ∈ C 1 [[0. Open Problem 3. . 9. Conjecture 3.4. ∞)] be such that every positive nontrivial solution of the diﬀerence equation xn+1 = f (xn ) . n = 0. Obtain necessary and suﬃcient conditions on f so that every positive solution of Eq(3.13) holds. 6.3 Let k be a positive integer and assume that α.2 Let f ∈ C 1 [[0. In particular address the cases k = 4. Obtain necessary and suﬃcient conditions on α.4. ∞). Obtain necessary and suﬃcient conditions on f so that every positive solution of Eq(3. ∞). . 5. .2 Let k ≥ 2 be a positive integer and assume that (3.50 Chapter 3. 8. [0. .) Conjecture 3. [0.

(ii) {pn }∞ converges to ±∞. . . 1. . xn−1 n = 0. n = 0.4. [0. Open Problems and Conjectures 51 Open Problem 3. . ∞)] such that every positive nontrivial solution of the equation xn+1 = is periodic with prime period three.18) xn−1 ∞ (i) {pn }n=0 converges to a ﬁnite limit. .4.4 Obtain necessary and suﬃcient conditions on f ∈ C[[0. n=0 .3.17) Open Problem 3. n=0 (iii) {pn }∞ is periodic with prime period k ≥ 2.4. 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 = . f (xn ) . (3. (3. ∞). . . 1.5 Let {pn }∞ be a sequence of nonzero real numbers.

.

(4. 1. . . 2)-type: xn+1 = n = 0. . 2)-Type Equations 4. A + Cxn−1 n = 0. A + Bxn n = 0. Bxn + Cxn−1 53 n = 0. . . 1. A + Cxn−1 n = 0. (4. . Bxn + Cxn−1 n = 0.1 Introduction α . 1. . 1. (4. (4. .2) xn+1 = α . .1) xn+1 = α . .6) . . .3) xn+1 = βxn .Chapter 4 (1. . . . 1. A + Bxn Eq(1) contains the following nine equations of the (1.5) xn+1 = βxn .4) xn+1 = βxn . . 1. (4. . . (4.

4). and the denominators are always positive. .9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. (4.1) and (4. . if {xn } is a solution of Eq(4.8). . A + Cxn−1 n = 0.1) which by the change of variables xn = . 1.2). Two of these equations. .2). (4. . 1. .4) to the linear equation yn+1 = A B yn + .8) γxn−1 . .54 Chapter 4. 4. (See Section 1.6 in the Preliminary Results). then the subsequences {x2n−1 } and {x2n } satisfy the Riccati equation of the form of Eq(4. Also Eqs(4. Eqs(4.4) are Riccati-type diﬀerence equations. 2)-type Eq(4. (4. Similarly.7) xn+1 = and xn+1 = γxn−1 . .2) and (4. Bxn + Cxn−1 n = 0. . . (1. (4. In view of the above. the initial conditions are nonnegative. A + Bxn n = 0. from which the global behavior of solutions is easily derived. .8) will not be discussed any further in this chapter. yn+1 = A + Cyn It is also interesting to note that the change of variables xn = 1 yn reduces the Riccati equation (4.8) are essentially Riccati equations.2 The Case β = γ = C = 0 : xn+1 = A yn B α A+Bxn This is the (1. namely Eqs(4. . 1. β β n = 0. then the subsequences {x2n−1 } and {x2n } satisfy the Riccati equation γyn n = 0. . . Indeed if {xn } is a solution of Eq(4. 1. . and (4. 1.1). 2)-Type Equations xn+1 = γxn−1 . . .

Theorem 4. lim sup yn = ∞ and n→∞ lim inf yn = 0. A2 The Riccati number associated with this equation (see Section 1. .2.12) is bounded and persists. The Case β = γ = A = 0 : xn+1 = reduces to the Riccati equation yn+1 = where α Bxn +Cxn−1 55 p . (4. (See also [12].6) is R = −p < 0 and so the following result is a consequence of Theorem 1. n→∞ . 2)-type Eq(4. .3. Proof. . So if the theorem is false there should exist a ∞ solution {yn }n=−1 which is neither bounded from above nor from below. 1. (4.10) is globally asymptotically stable. .10) αB . It is easy to see that if a solution of Eq(4. yn yn−1 n = 0.) We will ﬁrst show that every solution of Eq(4.11) (4.6.12) converges to √ its equilibrium B + C.12) is bounded from above then it is also bounded from below and vice-versa.1 The positive equilibrium y= −1 + √ 2 1 + 4p of Eq(4. .3 The Case β = γ = A = 0 : xn+1 = α Bxn +Cxn−1 This is the (1. 1.12) Eq(4. 1 + yn p= n = 0. That is. 4. .4.3) which by the change of variables √ α xn = yn is transformed to the diﬀerence equation yn+1 = C B + .12) was investigated in [65]. Theorem 4.2. .12) is bounded and then we will use the method of “limiting solutions” to show that every solution of Eq(4.3.1 Every solution of Eq(4.

1 . By the theory of limiting solutions (see Section 1. . B+C B+C n = 0. let {yn }∞ n=−1 be a solution of Eq(4. n→∞ for n ≥ −1. . . and by Theorem 1.12).1.12) about the equilibrium y = zn+1 = − B C zn − zn−1 . yn yn−1 n = −3. 1. . yi−1 yi−2 yj Proof.2 The equilibrium y = stable.3. S0 = S B C + . . y is locally asymptotically stable for all positive values of B and C. B + C < yi yj ≤ B + C which is impossible. 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 . −2.1. j − 1}.13) for all n ∈ {−1. . Theorem 4. 2 √ B + C of Eq(4. . To ∞ this end. The linearized equation of Eq(4. . It remains to be shown that y is a global attractor of all solutions of Eq(4. . B C B+C + > yj−1 yj−2 yi C B+C B + ≤ .3. .12) is globally asymptotically √ B + C is (4. Then by Theorem 4.5) there exist two solutions {In }∞ n=−3 and {Sn }∞ n=−3 of the diﬀerence equation yn+1 = with the following properties: I0 = I. (1. 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. {yn }n=−1 is bounded and persists. . and S = lim sup yn .12). .56 Chapter 4.

2)-type Eq(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. I−1 I−2 S B + C = SI. Sn ∈ [I. From (4.14) from which it follows that (4. S= B C B+C + ≤ S−1 S−2 I C B+C B + ≥ . . C βxn A+Cxn−1 . (4.Pielou’s Equation 57 In .4.Pielou’s This is the (1. Therefore B C B C + = I = S−1 = + S S S−2 S−3 S = S−2 . (4.15) 2 4. .15) we conclude that S=I and the proof is complete.4.4 The Case α = γ = B = 0 : xn+1 = Equation A yn . The Case α = γ = B = 0 : xn+1 = and Then βxn A+Cxn−1 . 1 + yn−1 β .14) and (4.16) . A n = 0. 1. S] for n ≥ −3.5) which by the change of variables xn = reduces to Pielou’s diﬀerence equation yn+1 = where p= pyn .

Theorem 4. Eq(4.1 (a) Assume p ≤ 1. Then the positive equilibrium y = p − 1 of Eq(4. 4.5 The Case α = γ = A = 0 : xn+1 = β . . it follows from Eq(4. p + yn−1 n = 0.16) is globally asymptotically stable. (4.16) is when p > 1.22) and ([67]. (See also ([42]. (4. x is globally asymptotically stable. 2)-Type Equations This equation was proposed by Pielou in her books ([66].79) as a discrete analogue of the delay logistic equation dN N (t − τ ) = rN (t) 1 − . Furthermore. 1.16) is globally asymptotically stable. . In fact by employing Theorem 1. ∞) and p > 1. p. p.17) In this case the zero equilibrium of Eq(4. .4.16) was investigated in [55]. 2)-type Eq(4. We summarize the above discussion in the following theorem. it follows that when (4. (1. t≥0 dt P which is a prototype of modelling the dynamics of single-species.75). p.6) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = p + yn .2.16) is unstable and Eq(4. 0 is locally asymptotically stable when p ≤ 1 and unstable when p > 1.18) . B + Cyn βxn Bxn +Cxn−1 This is the (1.16) possesses the unique positive equilibrium y = p − 1.4.58 Chapter 4. Then the zero equilibrium of Eq(4. (b) Assume y0 ∈ (0.17) holds. So the interesting case for Eq(4. which is locally asymptotically stable.) When p ≤ 1.16) that every positive solution converges to 0.

p.19) (see also Section 2.18) was investigated in ([42] Corollary 3.6. p= 4. . In fact.1. (b) Assume p < 1.19) was investigated in [28].7) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= yn−1 . The following local result is a straightforward application of the linearized stability Theorem 1.4. ∞).5) that 1 yn+1 < yn−1 p and so when p>1 for n ≥ 0 . every solution of Eq(4. Then 0 is the only equilibrium point of Eq(4. 0 is a repeller and y = 1 − p is a saddle point equilibrium.19) has semicycles of length one.1. Then 0 and y = 1 − p are the only equilibrium points of Eq(4.6. The Case α = β = C = 0 : xn+1 = where γxn−1 A+Bxn 59 B . 73) where it was shown that its equilibrium one is globally asymptotically stable . B γxn−1 A+Bxn This is the (1. .19) and they are both unstable. (4.1 (e).19) is a consequence of Theorem 1. That is. 1. p + yn n = 0. 2)-type Eq(4.19) A ∈ (0. . The oscillatory character of the solutions of Eq(4.1 (a) Assume p > 1.1. except possibly for the ﬁrst semicycle.7. C Eq(4.19) and it is locally asymptotically stable. It follows directly from Eq(4. Lemma 4. γ Eq(4.4.6 The Case α = β = C = 0 : xn+1 = γ yn .

yn n = 0. (1. . φ. . 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.20) C ∈ (0. . 2)-type Eq(4. 4. 2)-Type Equations the zero equilibrium of Eq(4. .19) converges to a period-two solution .9) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = p + where yn−1 . yn+1 < yn−1 for n ≥ 0 and so every positive solution of Eq(4. . .7 The Case α = β = A = 0 : xn+1 = γ Byn γxn−1 Bxn +Cxn−1 This is the (1. 1. ψ. 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. . . n = 0. Furthermore we can see that φψ = 0 but whether there exists solutions with φ=ψ=0 remains an open question. ψ. .19) is globally asymptotically stable. It follows from the identity (1 − p) − yn yn−1 . . . On the other hand when p = 1. ∞).60 Chapter 4. φ. . (4. 1. B This equation was investigated in [7] where it was shown that the following statements are true: p= . .

3 Let p < 1.7. 4. every solution of Eq(4. 1] and y2 ≥ p + y0 and use induction to complete the proof. the equilibrium y = p + 1 of Eq(4. (ii) When p = 1.7.4. 1. . with the possible exception of the ﬁrst semicycle. 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.20) is globally asymptotically stable.1 we obtain the following result: Theorem 4.1.7.20) is y = p + 1. .20) which consists of a single semicycle.20) converges to a period-two solution. (4. Then {yn }∞ n=−1 converges monotonically to y = p + 1. We now proceed to establish the above results. Moreover. Clearly the only equilibrium point of Eq(4. . lim y2n = ∞. The linearized equation of Eq(4. In fact the following result is true. n→∞ y2n+1 = p.20) which consists of at least two semicycles. Proof. and let {yn }∞ n=−1 be a solution of Eq(4.2 (a) Let {yn }∞ n=−1 be a solution of Eq(4.20) such that 0 < 1 y−1 ≤ 1 and y0 ≥ 1−p . Theorem 4.20) which are unbounded.7. n→∞ lim 2. . . Note that 1 1−p > p + 1.20) about the equilibrium point y = p + 1 is zn+1 + 1 1 zn − zn−1 = 0. Theorem 4.20) to be bounded.1 The Case p < 1 Here we show that there exist solutions of Eq(4. ∞ (b) Let {yn }n=−1 be a solution of Eq(4. (iii) When p > 1.20) is locally asymptotically stable when p > 1 and is an unstable saddle point when p < 1. every semicycle has length one. Then the following statements are true: 1.7. p+1 p+1 n = 0. and so y0 > p + 1.1 The equilibrium point y = p + 1 of Eq(4.7. Then {yn }∞ n=−1 is oscillatory. The next result about semicycles is an immediate consequence of some straightforward arguments and Theorem 1.1.21) As a simple consequence of the linearized stability Theorem 1.

20). p . ∞ (b) Suppose {yn }∞ n=−1 consists of at least two semicycles. ∞ (a) Suppose {yn }n=−1 consists of a single semicycle.7.2 The Case p = 1 The following result follows from Lemma 2.20). Note that for n ≥ 0.4 Let p = 1.62 Indeed y1 = p + y1 y0 Chapter 4. . 4. Also. p2 We shall ﬁrst show that lim supn→∞ yn ≤ p−1 . p m = 0. that p < yn for all n ≥ −1.7. Then p+ p2 p−1 ≤ lim inf yn ≤ lim sup yn ≤ . y2n+1 < p + So as every solution of the diﬀerence equation 1 ym+1 = p + ym .7.7. . The following result will be useful. (c) Every solution of Eq(4.3 The Case p > 1 Here we show that the equilibrium point y = p + 1 of Eq(4. (1.5. 2)-Type Equations > p. It follows from Theorem 4. and let {yn }∞ n=−1 be a solution of Eq(4. and that p < y0 < p + 1 < y−1 .20) is globally asymptotically stable. 4. Then {yn }∞ n=−1 converges monotonically to y = 2. Theorem 4. and let {yn }∞ n=−1 be a solution of Eq(4. . n→∞ p p−1 n→∞ Proof.1 and some straightforward arguments.7. Then the following statements are true. Lemma 4.1 Let p > 1. y2n−1 . 1]. See also Section 2. Hence y2 = p + y0 y1 ≥ p + y0 .20).20) converges to a period-two solution if and only if p = 1.7. 1. Then {yn }n=−1 converges to a prime period-two solution of Eq(4.2 that we may assume that every semicycle of {yn }∞ n=−1 has length one. y1 = p + y−1 1 ≤p+ ≤ 1. y0 y0 2 and so y1 ∈ (p.

p2 + y2n−1 < . ∞). Recall that by Theorem 4.7. there exist no solutions of Eq(4. ∞)] . We know by Theorem 4.20). we have p−1 p3 + p(p − 1) =p+ . Theorem 4. Then > 0. and f is increasing in y ∈ (0. p−1 f p. x Then f ∈ C[(0. . lim inf yn ≥ 2 n→∞ p p 2 We are now ready for the following result. y2n−1 p + p2 + is arbitrary. Let > 0. ∞).4. ∞). So let {yn }∞ n=−1 be a solution of Eq(4.1 that y = p + 1 is a locally asymptotically stable equilibrium point of Eq(4.7.5 Let p > 1. We shall next show that p + p−1 ≤ lim inf n→∞ yn . y) = p + . ∞) for each x ∈ (0. set p−1 p2 + >p . p2 − p p−1 p2 + . y ∈ (0.7. Proof. p−1 For x.p = p + p 2 p−1 p + p2 + p−1 p2 + p−1 = p2 + p3 + p = . f is decreasing in x ∈ (0. (0. ∞) for each y ∈ (0. Hence p < f (x.7. It suﬃces to show that lim yn = p + 1. and set a = p and b = Note that f and p2 + .20) with prime period two. y ∈ p.20). There clearly exists y2n = p + So as p−1 p3 + p + p(p − 1) y2n−2 >p+p 2 = . n→∞ y f (x. Let p N ≥ 0 such that for all n ≥ N .4. Then the equilibrium y = p + 1 is globally asymptotically stable. ∞). y) < for all x. p−1 Let n ≥ N . it follows that lim supn→∞ yn ≤ p−1 . The Case α = β = A = 0 : xn+1 = 2 γxn−1 Bxn +Cxn−1 2 63 p p converges to p−1 . ∞) × (0.

+ yn yn−1 (4.64 Finally note that by Lemma 4. .8. 2 4. l ∈ {0.) Open Problem 4. Investigate the global behavior of the solution {yn }. . xn−l n = 0. is well deﬁned for all n ≥ 0. x0 are arbitrary positive real numbers.4. p−1 p2 + p2 ≤ lim inf yn ≤ lim sup yn ≤ < n→∞ p p−1 p−1 n→∞ n→∞ lim yn = p + 1.8. . . 1. (a) Find all initial conditions y−1 . [23]. Show that every solution of Eq(4.2 (see [18]) Assume that A. (1. Is it bounded? Does limn→∞ yn exist? When does {yn } converge to a two cycle? .22) converges. [60]. ∞) and k. . . p<p+ Chapter 4. (4.23) A B λl−k + = 0. . y0 are such that y−1 y0 < 0 and such that Eq(4.22) Conjecture 4.8. . . . y0 with y−1 y0 < 0 for which the equation yn+1 = B A .22). n=−l Open Problem 4. . . 1. . .1. . . Determine the set of all positive initial conditions x−l .1 Consider the diﬀerence equation where and where the initial conditions x−l . A+B A+B A. 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]. . . .7. φp . . (b) Assume that the initial conditions y−1 .8 Open Problems and Conjectures xn+1 = A xn−k + B . x0 such that {xn }∞ converges to this periodic solution.6. . φ2 . φ1 .1 Let .23) is well deﬁned for all n ≥ 0. ∞). be a given prime period p solution of Eq(4. B ∈ (0. B ∈ (0. . and [61].} with k < l. . . [25]. 2)-Type Equations and so by Theorem 1.

20) is well deﬁned for all n ≥ 0.8.4. the asymptotic behavior. y0 ∈ R such that the equation (4.19) is bounded.4 Assume that (x−1 . y0 ∈ R be such that Eq(4. . Conjecture 4.19) has a solution which converges to zero. and the periodic ∞ nature of the solution {yn }n=−1 of Eq(4. Open Problems and Conjectures 65 Conjecture 4.4 Determine the set G of all initial conditions (y−1 . (c) Discuss (a) and (b) when p < 0. y0 ) ∈ R × R through which the equation yn−1 yn+1 = 1 + yn is well deﬁned for all n ≥ 0 and.24) is well deﬁned for all n ≥ 0 and bounded.3 Show that Eq(4. ∞) ∞ for which the solution {yn }n=−1 of Eq(4.8. Open Problem 4.2 Assume that p = 1.6 (See [13]) Find the set G of all initial points (x−1 . investigate the boundedness.8. Conjecture 4. Determine the set of initial conditions y−1 . x0 ) ∈ R × R through which the equation xn−1 xn+1 = −1 + (4. Show that Eq(4. Open Problem 4.20).24) xn is well deﬁned for all n ≥ 0. for these initial points.8.8. Determine all initial conditions y−1 . Show that {xn }∞ n=−1 is a three cycle.20) is well deﬁned for all n ≥ 0. Open Problem 4. Open Problem 4.8. investigate the global character ∞ of {yn }n=−1 . y0 ∈ (0. x0 ) ∈ R × R is such that the equation (4.8.20) possesses a solution {yn }∞ n=−1 which remains above the equilibrium for all n ≥ −1.5 (a) Assume p ∈ (0.8. (b) Let y−1 . ∞).3 Assume that p < 1. For such an initial point.

.8. 1. n = 0. with positive initial conditions is globally asymptotically stable if and only if kπ α−1 .6 (Pielou’s Discrete Logistic Model. . ai ∈ (0. . Obtain necessary and suﬃcient conditions on f for every positive solution of the equation xn+1 = to be bounded. Conjecture 4.}. 1 + xn−k n = 0.66 Chapter 4. . 1. ∞) and k ∈ {0. 1. x0 ∈ (0. . . . ∞) does the sequence deﬁned by xn−1 . . Investigate the global asymptotic stability of the equilibrium points of the diﬀerence equation xn+1 = xn . k. converges to a period-two solution. (0. + ak xn−k n = 0. .9 (See [5]) For which initial values x−1 . . ∞) for i = 0.8.8. 1. (1. Show that the positive equilibrium of Pielou’s discrete logistic model xn+1 = αxn . n = 0. 1. and [67]. . Open Problem 4. . . [66]. xn+1 = 1 + xn converges to two? .8 Let a. . . with positive initial conditions. . < 2 cos α 2k + 1 Open Problem 4. . . xn + xn−1 n = 0. ∞). ∞). see [55]. 1. . . . 2)-Type Equations Open Problem 4.8. a + a0 xn + .) Assume α ∈ (1.7 Let f ∈ C[(0.8. ∞)]. . Conjecture 4. f (xn ) . Show that every positive solution of the population model Jn+1 = Jn−1 er−(Jn +Jn−1 ) . .5 (See [27]) Assume r ∈ (0. .

4. ∞ More precisely.25) converge to zero. . 0.7) Open Problem 4. n = 0. Open Problem 4. ∞) through which the solutions of Eq(4. 1. 1. . . . .8. . . Open Problem 4. ﬁnd the basin of attraction of the equilibrium of Eq(4. . y0 ∈ B. .25). each of the subsequences {y2n }∞ and {y2n+1 }n=−1 is decreasing and n=0 lim [n→∞ y2n ][n→∞ y2n+1 ] = 0. . In other words. . n = 0. ∞ (b) Let y−1 .10 (See [31]) Consider the diﬀerence equation yn+1 = yn−1 e−yn . . .12 Let {pn }∞ be a convergent sequence of nonnegative real numn=0 bers with a ﬁnite limit. ∞) such that the solutions {yn }∞ n=−1 of Eq(4. (4.11 Assume p ∈ (0. 1 + yn−1 n = 0. Investigate the asymptotic behavior of {yn }n=−1 . lim p = n→∞ pn . . Open Problems and Conjectures (See also Section 4.20) are bounded. (b) Determine the limits of the subsequences of even and odd terms of every solution of Eq(4. ∞ One can easily see that every solution {yn }n=−1 of Eq(4. (a) Find the set B of all initial conditions y−1 .25). 67 (4.26) (4.27) . . 1). 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 + . 0. yn yn−1 pn yn . 1.8.25) with nonnegative initial conditions. . .25) converges to a period-two solution . lim (a) Find all initial points y−1 . . y0 ∈ [0. in terms of the initial conditions y−1 and y0 of the solution.8. y0 ∈ (0.8.

. n = 0.34) yn+1 = yn+1 = −1 + yn−1 . . Investigate the global character of all positive solutions of each of Eqs(4.31) yn+1 = (4. extend your result by introducing arbitrary real parameters in the equation.33) (4.68 pn + yn . 1. .13 Let {pn }∞ be a periodic sequence of nonnegative real numbers n=0 with period k ≥ 2. .8. . . pn + yn−1 yn−1 . Open Problem 4. yn For those equations from the above list for which you were successful.14 For each of the following diﬀerence equations determine the “good” set G ⊂ R × R of all initial conditions (y−1 . .30) yn+1 = yn+1 = pn + Open Problem 4. . 2)-Type Equations yn+1 = n = 0. 1 − yn (4. . 1 − yn−1 1 + yn .(4.26) .29) (4.8. y0 ) ∈ G.32) (4. . 1. investigate the long term behavior of the solution {yn }∞ n=−1 : yn+1 = yn .28) (4. (1. (4. yn Chapter 4. . y0 ) ∈ R × R through which the equation is well deﬁned for all n ≥ 0. Then for every (y−1 .30). 1 − yn−1 yn−1 . n = 0. pn + yn yn−1 . 1.

Bxn n = 0. 1.4) xn+1 = α + γxn−1 .1 Introduction α + βxn .3) xn+1 = α + γxn−1 . . . 1. . 1. (5. .6) . .2) xn+1 = α + βxn . A n = 0.5) xn+1 = α + γxn−1 . 1. Cxn−1 n = 0. (5. 1)-type: xn+1 = n = 0.1) xn+1 = α + βxn . . . . . . . . . A Eq(1) contains the following nine equations of the (2. 1)-Type Equations 5. . 1.Chapter 5 (2. . 1. Cxn−1 69 n = 0. Bxn n = 0. . (5. (5. . . (5. (5.

. Bxn n = 0. A n = 0.5. .5). yn−1 α+βxn Cxn−1 . 1. . .10) . which was investigated in Section 4.3) and (5. (5.9) to Eq(4.2 The Case γ = A = B = 0 : xn+1 = Equation p + yn .6) is essentially like Eq(5. (5. (2. 1)-Type Equations xn+1 = βxn + γxn−1 . . (5. 1.3) which by the change of variables reduces to the equation yn+1 = n = 0. . and the denominators are always positive. .9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive.4) and (5.7) xn+1 = and xn+1 = βxn + γxn−1 . .7. . .1). .70 Chapter 5.8) to Eq(4.2) is a Riccati equation and Eq(5.2). . The change of variables γ xn = yn .18) with p = β . 1. which will be investigated in the next two sections. C C γ reduces Eq(5. 1)-type Eq(5. (5. γ Finally the change of variables xn = γ β + yn . Eq(5. the initial conditions are nonnegative. 5. B reduces Eq(5.8) βxn + γxn−1 . Cxn−1 n = 0. Eqs(5. Therefore there remain Eqs(5.7) are linear.20) with p = β . Of these nine equations. (5. 1. .Lyness’ This is the (2. which was investigated in Section 4.

11) with initial conditions y−1 and y0 is the ﬁve-cycle: y−1 . . [42]. β2 p=1 was discovered by Lyness in 1942 while he was working on a problem in Number Theory. after the ﬁrst one. contains at least two and at most three terms.) In this special case. [57]-[59]. (5.. Indeed the solution of Eq(5.Lyness’ Equation 71 p= The special case of Eq(5. 1. n = 0. See [11].10) is bounded from above and from below by positive constants. It was shown in [30] that no nontrivial solution of Eq(5. and [72] and the references cited therein. . (b) Every nontrivial semicycle.11) yn−1 every solution of which is periodic with period ﬁve. (See ([42]. . 1 + y0 1 + y−1 + y0 1 + y−1 . .5.10) possesses the invariant In = (p + yn−1 + yn ) 1 + 1 yn−1 1+ 1 yn = constant (5. Concerning the semicycles of solutions of Eq(5.10) the following result was established in [43]: Theorem 5. The Case γ = A = B = 0 : xn+1 = where α+βxn Cxn−1 . [71]. . [43]. . y0 .10) is stable but not asymptotically stable.2.10) has a limit. the equation becomes 1 + yn yn+1 = .2. y−1 y−1 y0 y0 Eq(5.10) are true: (a) The absolute extreme in a semicycle occurs in the ﬁrst or in the second term. There is substantial literature on Lyness’ Equation.. This result was also established in [49] by using a Lyapunov function.. 131) for the history of the problem. Then the following statements about the positive solutions of Eq(5. p.12) from which it follows that every solution of Eq(5. . It was also shown in [44] by using KAM theory that the positive equilibrium y of Eq(5.10) where αC .1 Assume that p > 0.

1)-Type Equations 5. (2.13) is a saddle point. α+γxn−1 Bxn This is the (2.10) without using essentially the invariant.13) has semicycles of length one.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 ∞.12).12).5) which by the change of variables xn = reduces to the equation yn+1 = where p= p + yn−1 .10) is bounded from above and from below by positive numbers. γ2 Eq(5. 5. By the linearized stability Theorem 1.7. yn (5. one can easily see that every solution of Lyness’ Eq(5.13) one can easily see that every nonoscillatory solution of Eq(5. From this it follows that a solution of Eq(5.4 Open Problems and Conjectures By using the invariant (5.72 Chapter 5. Concerning the nonoscillatory solutions of Eq(5. 1)-type Eq(5. Open Problem 5.13) converges monotonically to the equilibrium y. . To this day we have not found a proof for the boundedness of solutions of Eq(5.10) is bounded without using the invariant (5. every oscillatory solution of Eq(5. 1. .13) αB . By Theorem 1.13) was investigated in [28]. The global character of solutions of Eq(5.3 The Case β = A = C = 0 : xn+1 = γ yn B n = 0.13) is a consequence of the identity yn+1 − yn−1 = yn−1 − yn−2 yn for n ≥ 1.1.4. .1 one can see that the unique equilibrium point √ 1 + 1 + 4p y= 2 of Eq(5.1 one can see that except possibly for the ﬁrst semicycle. .1 Show that every solution of Lyness’ Eq(5.

1 Assume that f ∈ C[(0.4.4 (See [26]. y0 ∈ (−∞. ∞)]. f (xn ) . (0. ∞). . ∞).4. 1. (0.14) Open Problem 5.4.4.4. .4.14) possesses a unique positive equilibrium point x and a nontrivial invariant.3 Assume that f ∈ C[(0. ∞) through which Eq(5. xn−1 n = 0. Open Problem 5.14) about x does not have both eigenvalues in |λ| < 1 and does not have both eigenvalues in |λ| > 1.4. To illustrate. Conjecture 5.5. . Obtain necessary and suﬃcient conditions in terms of f so that every positive solution of the diﬀerence equation xn+1 = is bounded.15) is well deﬁned for all n ≥ 0. Open Problems and Conjectures 73 Open Problem 5. Determine the set G of all initial conditions y−1 .15) are unbounded is a set of measure zero. Show that the subset of G through which the solutions of Eq(5.4. ∞)]. (0. Show that the linearized equation of Eq(5.14) possesses a (nontrivial) invariant. . When we allow the initial conditions to be real numbers the equation yn+1 = p + yn yn−1 (5. ∞).2 Assume that f ∈ C[(0.15) may not even be well deﬁned for all n ≥ 0.2 Let G be the set of initial conditions deﬁned in the Open problem 5. ∞).) Assume p ∈ (−∞. Obtain necessary and suﬃcient conditions in terms of f so that the diﬀerence equation (5. ∞)] and that the diﬀerence equation (5. Conjecture 5. we oﬀer the following open problems and conjectures. 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. (5.

4.5 Determine all positive integers k with the property that every positive solution of the equation yn+1 = is periodic. .6 Show that the equation yn+1 = 1 + yn−1 . 1)-Type Equations Conjecture 5. . converges to a period-three solution. Conjecture 5. . . . . . . 1.4. . has a nontrivial positive solution which decreases monotonically to the equilibrium of the equation. .15) is not well deﬁned for all n ≥ 0. . + yn−k . + yn yn−1 yn−3 yn−4 n = 0. yn−k−1 n = 0. 1. is a set of points without interior.4.6 Find the set G of all initial points (y−1 .4 Show that every positive solution of the equation yn+1 = converges to a period-six solution. Open Problem 5. yn yn−4 + . ∞) through which Eq(5. . Determine the periodic character and the asymptotic behavior of all solutions with (y−1 .5 (See [20]) Show that every positive solution of the equation yn+1 = 1 1 . yn−1 yn−3 n = 0. 1 + . yn n = 0.3 Assume p ∈ (−∞.74 Chapter 5.4. y0 ) ∈ G. . Show that the set of points (y−1 . 1. .4. (2. ∞) × (−∞. . ∞). y0 ) ∈ R × R through which the equation 1 + yn−1 yn+1 = yn is well deﬁned for all n ≥ 0. Open Problem 5.4. 1. Conjecture 5. Conjecture 5. y0 ) ∈ (−∞.

. . ∞). (5. Note that the change of variables n e y2 max{x2 . . see [59]) Assume α > 1. Conjecture 5.4. xn xn−1 n = 0. (5. . see [59]) Let A ∈ (0.16) to where xn = −1+yn 2 if A > 1 if A = 1 if A < 1 n = 0. A} n . When δ = 1. Eq(5. 1.16) A A 1+yn 2 reduces Eq(5. (1 + xn )xn−1 n = 0.4.17) is called the Gingerbreadman Map and was investigated by Devaney [17].17) yn+1 = |yn | − yn−1 + δ. . Show that every positive solution of May’s Host Parasitoid Model xn+1 = is bounded. . 1. .4. 2 αxn . 1 if A < 1 if A = 1 δ= 0 −1 if A > 1. Open Problems and Conjectures 75 Conjecture 5. . .7 (May’s Host Parasitoid Model. . 1.5. Show that every positive solution of the equation xn+1 = is bounded.8 (The Gingerbreadman Map.

.

. . . A + Cxn−1 n = 0.1 Introduction α + βxn . .1) xn+1 = α + βxn . Bxn + Cxn−1 77 n = 0. . 2)-type: xn+1 = n = 0. . 1. 1. 2)-Type Equations 6. (6. . (6. . A + Cxn−1 n = 0.6) . . .3) xn+1 = α + γxn−1 .2) xn+1 = α + βxn .5) xn+1 = α + γxn−1 . . . Bxn + Cxn−1 n = 0. 1. (6. 1.4) xn+1 = α + γxn−1 . (6. . A + Bxn n = 0. . . . (6. (6. . . 1.Chapter 6 (2. 1. A + Bxn Eq(1) contains the following nine equations of the (2.

2)-type Eq(6. . 2)-Type Equations xn+1 = βxn + γxn−1 . .8) xn+1 = βxn + γxn−1 .5). 1. .78 Chapter 6. A + Bxn n = 0.7) xn+1 = βxn + γxn−1 . for n ≥ −1. . (2. A + Cxn−1 n = 0. (6. To avoid a degenerate situation we will assume that αB − βA = 0.9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. the initial conditions are nonnegative. .6. . . Bxn + Cxn−1 n = 0. 1. 1.Riccati This is the (2. . (6. . .1). See Section 1. In fact if {xn }∞ n=−1 is a positive solution of ∞ ∞ Eq(6. 0. .5) is essentially similar to Eq(6. . Therefore we will omit any further discussion of Eq(6. (6. . . 1. α + γzn .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. Eq(6. 6. . .2 The Case γ = C = 0 : Equation xn+1 = α+βxn A+Bxn . . A + Czn n = 0. and the denominators are always positive.1) which is in fact a Riccati equation. α + γzn . A + Czn for n ≥ 0 n = −1.

3.10) is locally asymptotically stable for all values of the parameters p and q.10) is a very simple looking equation for which it has long been conjectured that its equilibrium is globally asymptotically stable. 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< . See also [36]. 1 + yn−1 n = 0.10) is bounded from above and from below by positive constants. 6. 1. 4 Now the following result is a consequence of Theorem 1. 2)-type Eq(6.10) was investigated in [42] and [43].2: Theorem 6.1. the conjecture has not been proven or refuted. 2 A A (Eq(6.10) has the unique positive equilibrium y given by p= y= q−1+ (q − 1)2 + 4p 2 .6.3.) Eq(6.1 we obtain the following result.2.1) is globally asymptotically stable.1 The positive equilibrium of Eq(6. . .10) αC β and q = . Eq(6.10) are the following: Theorem 6.3. . The main results known about Eq(6. Lemma 6.6. .1 Every solution of Eq(6.1 The equilibrium y of Eq(6.2) which by the change of variables xn = reduces to the equation yn+1 = where p + qyn .3 The Case γ = B = 0 : xn+1 = A yn C α+βxn A+Cxn−1 This is the (2. (6. By applying the linearized stability Theorem 1. To this day.

except possibly for the ﬁrst semicycle. We present the proofs for positive semicycles only. 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). their extrema cannot be consecutive terms. then p yn+1 ≥ 1+M and so it is also bounded from below. k→∞ Hence. and y1+nk = max{yn : n ≤ nk } for k ≥ 0. (iv) Assume q ≤ p. if the solution is bounded from above by a constant M . The oscillatory character of solutions for Eq(6.10). (2. Furthermore after the ﬁrst. . Clearly. Furthermore. every semicycle contains two or three terms. except perhaps for the ﬁrst one. in every semicycle. (iii) In any two consecutive semicycles. k→∞ lim y1+nk = ∞. The proofs for negative semicycles are similar and will be omitted. there is at most one term which follows the extreme. Then the following statements are true: (i) Every semicycle. (ii) The extreme in each semicycle occurs at either the ﬁrst term or the second.80 Chapter 6. From (6. Then there exists a subsequence {y1+nk }∞ such k=0 that k→∞ lim nk = ∞. Now assume for the sake of contradiction that the solution is not bounded from above. 2)-Type Equations Proof.10) we see that yn+1 < qyn + p for n ≥ 0 and so k→∞ lim ynk = lim ynk −1 = ∞. for suﬃciently large k. has at least two terms.10) is described by the following result: ∞ Theorem 6. Then. the remaining terms in a positive semicycle are strictly decreasing and in a negative semicycle are strictly increasing. Let {yn } be a solution of Eq(6. Proof.2 Let {yn }n=−1 be a nontrivial solution of Eq(6.3.

. 1 + yN −1 1 + yN (iv) Assume that for some N ≥ 0. α+βxn A+Cxn−1 81 yN −1 < y Then yN +1 = and yN ≥ y.3 The positive equilibrium of Eq(6. 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 .3. The opposite case is similar and will be omitted. the terms yN . < < 1 + yN +1 1 + yN +1 1+y 2 Theorem 6. and yN +2 are all in a positive semicycle.10) is globally asymptotically stable if one of the following two conditions holds: (i) q < 1. Then yN +3 = p + qyN +2 p + qyN +1 p + qy = y. (ii) q≥1 and either p ≤ q or q < p ≤ 2(q + 1).3. Then yN ≥ y.6. > 1 + yN −1 1+y (ii) Assume that for some N ≥ 0. The Case γ = B = 0 : xn+1 = (i) Assume that for some N ≥ 0. Then yN +1 = p + qyN p + qyN +1 < = yN +2 . p + qyN p + qy = y. So assume that for some N ≥ 0 the last two terms in a negative semicycle are yN −1 and yN with yN −1 ≥ yN . yN +1 > y and p p +q +q yN +2 1 p + qyN +1 yN +1 = = < y = 1. yN +1 .

4.11) was investigated in [15]. 2)-Type Equations (i) In view of Lemma 6. yn + qyn−1 αB β2 n = 0. 1+i from which the result follows. Then Eq(6. . Chapter 6.3. . (2.1.10) yields i≥ Hence and so because q < 1. For the proof when see ([42].3) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= yn + p . C . let i = lim inf yn n→∞ and s = lim sup yn n→∞ which by Theorem 6. p.3 (f).1. 1 ≤ q ≤ p ≤ 2(q + 1) 2 6. . B (6. The following result is a consequence of the linearized stability Theorem 1.10) n=−1 tends to y as n → ∞.1. This result was ﬁrst established in [43].4 The Case γ = A = 0 : xn+1 = β yn B α+βxn Bxn +Cxn−1 This is the (2. 71).1 exist and are positive numbers. (ii) The proof when p<q follows by applying Theorem 1.2.4. Theorem 3. To this end. . it remains to show that every solution {yn }∞ of Eq(6. 2)-type Eq(6. p + (q − 1)i ≤ is ≤ p + (q − 1)s i≥s p + qi 1+s and s ≤ p + qs .11) and q = Eq(6.3.82 Proof. 1.

.4. q for all n > k. ∞). (b) Assume p > q and suppose that for some k ≥ 0.11) is locally asymptotically stable for all values of the parameters p and q. p q p . then ym ∈ I for all m > k.1 q p .1 The unique positive equilibrium point y= 1+ 1 + 4p(q + 1) 2(q + 1) of Eq(6.4. (a) The basic ingredient behind the proof is the fact that when u. Then the following statements are true: (a) Assume p < q and suppose that for some k ≥ 0. yk−1 . v) = u + qv is increasing in u and decreasing in v. yk−1 . The Case γ = A = 0 : xn+1 = α+βxn Bxn +Cxn−1 83 Theorem 6. yk + p yk + p yk+1 = ≤ = 1. yk ∈ 1. 6. Then yn ∈ 1. Lemma 6.4. yk ∈ Then yn ∈ p .6. v ∈ [ p .4. two consecutive values yk−1 and yk of a solution lie in I. 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. Proof. the q function u+p f (u.1 Let {yn }∞ n=−1 be a solution of Eq(6.11).11) in the sense that if.1 . q for all n > k. for some k ≥ 0. (b) The proof is similar and will be omitted. Indeed.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.

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)

.

4. 2. p ]. By applying (6. 4 and so n→∞ for i = 1. First. Now. 2)-Type Equations is either identically equal to one or else it is strictly monotonically convergent.88 Chapter 6. Set Li = n→∞ y4n+i lim Thus clearly .7 that every solution of Eq(6. q Hence for every solution {yn }∞ n=−1 of Eq(6. 2. In this interval the function q f (u. .2 that [1. 2 . it converges to a periodic solution with period-four which q is not the equilibrium. As in Case 2 we can see that every solution of Eq(6.22) is a periodic solution of Eq(6. (b) The proof follows from the discussion in part (a). lim yn = 1. Recall from Lemma 6. (6.19) to the solution (6. assume that p < q. L3 . p ] converges to y. . L4 .11) converges to y.11) must eventually lie in the interval [ p .4.14) to this period four solution we obtain a contradiction and so every solution of Eq(6. v) = u + qv is decreasing in both variables and so it follows by applying Theorem 1. p ]. L1 . we assume that p > q. L2 . Under the assumption that q ≤ 1 + 4p. .11) with period four. 3. 1]. 3.11) n→∞ lim yn = y and the proof is complete.22) we see that Li = 1 for i = 1. By q using an argument similar to that in the case p = q. p ] is an invariant interval. . 4. Theorem 1. Case 2 Assume p = q. .5 applies and so every solution of Eq(6. By applying (6. In this interval the q function u+p f (u. we can see that if a solution is not eventually in [1. . v) = u+p u + qv is increasing in u and decreasing in v. (2.11) must lie eventually in [1.11) with two consecutive values in [1.4.

25) holds.5.6.1 we obtain the following result: Theorem 6. .5.2 (a) Eq(6. ψ ∈ (0.24) . γ . . Concerning prime period-two solutions.5. A (6. . and is an unstable saddle point when q > 1.5 The Case β = C = 0 : xn+1 = A yn B This is the (2. 1 + yn αB A2 n = 0. φ. (6. .23) converges to a period two solution.1.23) is locally asymptotically stable when q < 1. . ψ. . Theorem 6. (c) Assume that (6.7.23) has a prime period-two solution .4) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + qyn−1 .1 The equilibrium y of Eq(6.23) was investigated in [28].25) holds. . φ. . the following result is a consequence of the results in Section 2.24) are given by {φ. 2)-type Eq(6. 1. (6. if and only if q = 1.23) and q = (Eq(6. Then every solution of Eq(6. Then the values φ and ψ of all prime period-two solutions (6. The Case β = C = 0 : xn+1 = α+γxn−1 A+Bxn 89 α+γxn−1 A+Bxn 6.25) (b) Assume (6. .23) is y= q−1+ (q − 1)2 + 4p 2 and by applying the linearized stability Theorem 1. ψ. ∞) : φψ = p}.) The only equilibrium point of Eq(6.

2 The Case q < 1 The main result in this section is the following: Theorem 6. yn0 −1 ) < f (yn0 .26) . (2.7.1 (b) Let {yn } be a nonoscillatory solution of Eq(6.23) oscillates about the equilibrium y with semicycles of length one. yn0 > yn0 −1 .23). It is suﬃcient to show that {yn } is a decreasing sequence for n ≥ K.1 Semicycle Analysis Our results on the semicycles of solutions for Eq(6. 1 + yn0 2 p + qx 1+x 6.5.23). Proof.23) converges monotonically to the equilibrium y.5.5.23) and let y denote the unique positive equilibrium of Eq(6. Then every nonoscillatory solution of Eq(6. (b) Assume p ≥ q. (6.3 Let {yn } be a nontrivial solution of Eq(6. Then yn0 +1 = which is impossible. The case where yn−1 < y for n ≥ K is similar and will be omitted. p + qyn0 −1 = f (yn0 . the condition p ≥ q implies that the function f (x) = is decreasing. 1 + yn p + . (a) This follows immediately from Theorem 1. Then the positive equilibrium of Eq(6. yn0 ) < f (y. y) = y. Clearly.4 Assume that q < 1. We will assume that there exists a positive integer K such that yn−1 ≥ y for every n ≥ K. 2)-Type Equations 6. Proof.90 Chapter 6.23) is globally asymptotically stable. Clearly yn+1 = Set M= p + qyn−1 < p + qyn−1 . Then the following statements are true: (a) After the ﬁrst semicycle.23) are as follows: Theorem 6.5. 1−q n ≥ 0. So assume for the sake of contradiction that for some n0 ≥ K. an oscillatory solution {yn } of Eq(6.

5.5. y2n−1 < q − 1 and y2n > y0 + n Hence n→∞ p .1.5.26) and Eq(6. From Eq(6. The proof is complete. b]. b] × [a. and b = + . a = 0. 1+x 1−q Then clearly. lim y2n = ∞ p + qy2n−1 =0 1 + y2n and lim y2n+1 = lim n→∞ n→∞ and the proof is complete. a ≤ f (x. y) ∈ [a. 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. y) ≤ b.23). one of which is monotonically approaching 0 and the other is monotonically approaching ∞. for all (x. it now follows that every solution of Eq(6. M ].3 The Case q > 1 In this case we will show that there exist solutions that have two subsequences. .23) lies eventually in the interval [0. By applying Theorem 1.6 it follows that y is a global attractor of all solutions of Eq(6. The local stability of y was established in Theorem 6. y) = . 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. The Case β = C = 0 : xn+1 = α+γxn−1 A+Bxn 91 where is some positive number.6.23). Set p + qy p f (x.4. 2 6.

C (6. (6. 1.1 we see that y is locally asymptotically stable when (q − 1)y < 2p and unstable (a saddle point) when (q − 1)y > 2p. 2)-Type Equations 6. (6.28) By employing Theorem 1.27) was investigated in [45].92 Chapter 6.27) about y is zn+1 + q qy − p zn − zn−1 = 0. . qyn + yn−1 αC γ2 n = 0. 2)-type Eq(6. If we now set it is easy to see that (6. B .27) has the unique positive equilibrium y= 1+ 1 + 4p(1 + q) 2(1 + q) .) Eq(6. . .29) is satisﬁed if and only if either q≤1 or q > 1 and F 2p q−1 > 0. . 1. . Clearly the equilibrium y is the positive solution of the quadratic equation (1 + q)y 2 − y − p = 0.6 The Case β = A = 0 : xn+1 = γ yn C α+γxn−1 Bxn +Cxn−1 This is the (2.30) (6. (2.1. . The linearized equation associated with Eq(6.27) and q = (Eq(6. . F (u) = (1 + q)u2 − u − p.29) . 1+q (y + p)(1 + q) n = 0.6) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + yn−1 .

27).6. q−1 .2 Eq(6. . ψ. be a period-two cycle of Eq(6.1 The equilibrium y of Eq(6.31) 6.27) is locally asymptotically stable when q < 1 + 4p and unstable. and more precisely a saddle point equilibrium. (6.6. . . . φ. . when q > 1 + 4p. . 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. . if and only if q > 1 + 4p.6. ψ. ψ. φ. . 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. φ. φ.6.27) has a prime period-two solution . .32) holds.30) is satisﬁed if and only if q > 1 and F 2p q−1 < 0. The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 93 Similarly (6.6. . . .32) (6. The following result is now a consequence of the above discussion and some simple calculations: Theorem 6. ψ. Furthermore when (6.1 Let Existence and Local Stability of Period-Two Cycles .

we set un = yn−1 and vn = yn . ∞) × (0. Jacobian matrix JT 2 . 1. (2. 1. . ψ) = . . . . Let T be the function on (0. Chapter 6. ∂v (qψ + φ)2 . v) = p+u qv + u u v = g(u. . ψ. . φ. qg(u. and write Eq(6. φ. ψ) ∂u φ JT 2 = ψ ∂h (φ. . . ∂u (qψ + φ)2 ∂g (p + φ)q (φ. v) p+v . . ∂g qψ − p (φ.27) in the equivalent form: un+1 = vn vn+1 = p+un qvn +un . . By a simple calculation we ﬁnd that T2 where g(u. . ψ) ∂v (φ. ψ) ∂u ∂v where. the second iterate of T . ψ) ∂h (φ.94 To investigate the local stability of the two cycle . v) h(u. ∞) deﬁned by: T Then u v = v p+u qv+u . φ ψ is a ﬁxed point of T 2 . v) = Clearly the two cycle is locally asymptotically stable when the eigenvalues of the φ lie inside the unit disk. 2)-Type Equations for n = 0. ψ. evaluated at ψ We have ∂g ∂g (φ. n = 0. ψ) = − . . v) + v and h(u. .

The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 95 q(p + ψ)(qψ − p) ∂h (φ.6.33) is equivalent to the following three inequalities: S <1+D −1 − D < S D < 1. ψ) = . To this end observe that (6. ∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1. if and only if |S| < 1 + D < 2.1 that both eigenvalues of JT 2 unit disk |λ| < 1. First we will establish Inequality (6. ψ).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. ψ) − (φ.34). + 2 (qψ + φ)2 ∂v (qφ + ψ) (qφ + ψ)2 Set S= and D= ∂h ∂g (φ. Inequality (6.34) is equivalent to (6.34) (6. ψ) = − . ψ) (φ. ψ) (φ. ψ) ∂u ∂v ∂g ∂h ∂h ∂g (φ.33) (6.35) (6. p )[q − (1 + 4p)] < 0 p+1 .6. ψ) + (φ. ∂u (qφ + ψ)2 (qψ + φ)2 ∂h q 2 (p + φ)(p + ψ) qφ − p (φ.1.

Observe that (6. Theorem 6. 2)-Type Equations Next we will establish Inequality (6. p ≤ yN ≤ 1.27). More precisely we show that the values of any solution {yn }∞ n=−1 of Eq(6. 6.96 Chapter 6. Then the following statements are true. we establish Inequality (6. q Then p ≤ yn ≤ 1 for all n ≥ N.3 Let {yn }∞ n=−1 be a solution of Eq(6. q . Observe that (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 > 1.36) 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).35). Finally.6.27) either remain forever outside the interval I or the solution is eventually trapped into I.6. (2.2 Invariant Intervals Here we show that the interval I with end points one and p is an invariant interval for q Eq(6. (a) Assume that p ≤ q and that for some N ≥ 0.36).

6. Then observe that the following identities are true: p − yn for n ≥ 0. 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. for n ≥ 0.39) 2 qyn−1 (yn − p ) + (yn − 1)(yn−1 yn + qyn ) q . (6.27). The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 97 (b) Assume that p ≥ q and that for some N ≥ 0.6. p 1 ≤ yN ≤ . (6.38) (6. 6.6. q Then 1 ≤ yn ≤ Proof. (b) Clearly.3 Semicycle Analysis Here we present a thorough semicycle analysis of the solutions of Eq(6. Let {yn }∞ n=−1 be a solution of Eq(6.37) 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. (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.27).27) relative to the equilibrium y and relative to the end points of the invariant interval of Eq(6. p + yN −1 p + yN −1 ≥ = q qyN + yN −1 (pyN + p yN −1 ) p q and the proof follows by induction.

then yN +1 < p . p ] for n ≥ N .37)-(6. q (v) If for some N ≥ 0. then yN +1 > 1. yN ≤ p . Then the following statements are true: (i) If for some N ≥ 0. p ] is an q q q invariant interval for Eq(6.98 When p=q that is for the diﬀerence equation yn+1 = p + yn−1 . then yn ∈ [1. yN ≥ p . . 1. In particular [1. then yN +1 < 1.6. (6. q (iii) If for some N ≥ 0. . yN ≥ 1.2 Assume that p=q ∞ and let {yn }n=−1 be a solution of Eq(6.27). then yN +1 = 1. then yN +2 < yN . yN < p .1 Assume that p>q and let {yn }∞ n=−1 be a solution of Eq(6. q Lemma 6. q (ii) If for some N ≥ 0. . Then the following statements are true: (i) If for some N ≥ 0. yN = p . yN > p . (viii) p 1<y< .41) 2 pyn−1 + yn−1 yn + pyn p(p + yn−1 ) + yn (pyn + yn−1 ) (6. q (iv) If for some N ≥ 0.27).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. yN < 1.42). then yN +1 > 1. q (vi) If for some N ≥ 0.42) The following three lemmas are now direct consequences of (6.27). then yN +2 > yN .6. yN ≤ 1. . (2. Lemma 6. 2)-Type Equations n = 0. (vii) If for some N ≥ 0. (6. pyn + yn−1 Chapter 6.

2 and Lemmas 6. yN ≤ 1. Lemma 6. (v) If for some N ≥ 0. yN ≤ p . q q The following result is now a consequence of Theorems 1. then yN +1 < 1. then yN +2 < yN . q (iv) If for some N ≥ 0. yN > 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 .6. The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 99 (ii) If for some N ≥ 0.6.27). q (viii) If for some N ≥ 0. then yN +2 > yN . then yN +1 < 1. then yN > yN +2 > 1.6.1 and 1. then yN +1 = 1. then yn ∈ [ p . Then the following statements are true: (i) If for some N ≥ 0. q (vi) If for some N ≥ 0. then yN < yN +2 < 1.1-6. yN < 1. yN = p . (vii) If for some N ≥ 0. In particular [ p . (iv) If for some N ≥ 0. yN > p .7. q (ii) If for some N ≥ 0. yN ≤ 1. then yN +1 > p . yN = 1. yN ≤ p .6. q Note that the function f (u. q (v) If for some N ≥ 0. (iii) If for some N ≥ 0. then yN +1 = 1. yN ≥ 1.6.27). then yN +1 > 1.3 Assume that p<q ∞ and let {yn }n=−1 be a solution of Eq(6. 1] for n ≥ N . yN < p . yN > 1. 1] is an q q invariant interval for Eq(6. q (iii) If for some N ≥ 0.7.3: . (ix) p < y < 1. then yN +1 ≥ 1.

must converge to y.43) holds. any solution which remains forever outside the interval I = [ p . Eq(6.44)-(6. every solution of Eq(6. When (6. (6. But this two cycle lies q inside the interval I. . 2)-Type Equations Then either all the even terms of the solution lie in J and all odd terms lie in K. p<q the length is one.5 (a) Assume q ≤ 1 + 4p. On the other hand when (6. or for some N ≥ 0. . (2. The above observations are summarized in the following theorem: Theorem 6. φ. q Now when q ≤ 1 + 4p Then the equilibrium y of Eq(6.43) 6. yn ∈ I for n ≥ N.27) either eventually enters the interval I with end points 1 and p and remains there.6. and only then.100 Chapter 6. ψ. if Theorem 6.27) eventually enters and remains in the interval I.27) has a unique equilibrium y which is locally stable when (6. .4 Global Behavior of Solutions Recall that Eq(6. The character of these solutions remains an open question. while if p>q the length is at most two.6.27) has a “unique” prime period-two solution . or vice-versa.32) holds. or stays forever outside I. . .32) holds. ∞). when (6.31) holds.44) (6. When (6. q−1 (6. . 1] must converge to the two cycle (6. ψ.45). in view of Theorems 1.27).6. Hence.4 Let {yn }∞ n=−1 be a solution of Eq(6.46) Also recall that a solution {yn }∞ n=−1 of Eq(6. (6.4. with φ + ψ = 1 and φψ = p .6 and 1. 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 = (0.4. except for the length of the ﬁrst semicycle of the solution.32) holds. φ.7. .27) is global attractor.45) there are no period-two solutions and so the solutions must eventually enter and remain in I and.

.47) β γ and q = . The following result follows from Theorem 1.27) eventually enters and remains in the interval [ p .47) was investigated in [52].3.6. More precisely it is a saddle point when 1−p<q <1+p and a repeller when q > 1 + p. The Case α = C = 0 : xn+1 = (b) Assume βxn +γxn−1 A+Bxn 101 q > 1 + 4p.7. 1 + yn n = 0.47) is unstable.7 The Case α = C = 0 : xn+1 = A yn B βxn +γxn−1 A+Bxn This is the (2.7. Then the zero equilibrium of Eq(6. y = p + q − 1 is the only positive equilibrium point of Eq(6.7) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where pyn + qyn−1 . q 6.1 and the linearized stability Theorem 1.) The equilibrium points of Eq(6. 1+y So y = 0 is always an equilibrium point. 2)-type equation Eq(6.47) are the solutions of the equation p= y= py + qy . .1. A A (Eq(6. (b) Assume p + q > 1. . .1 (a) Assume p + q ≤ 1. Then every solution of Eq(6. Then the zero equilibrium of Eq(6.47) is globally asymptotically stable. 1.47). and when p + q > 1. (6. 1].1: Theorem 6.

47). p ].3 (a) Assume that q < 1.47) is locally asymptotically stable. the values of φ and ψ of all prime period-two solutions are given by: pφ {φ. ∞).5 that the following result is true: Theorem 6.47). p ] is an invariant interval for Eq(6. That is. Then the positive equilibrium y = p + q − 1 of Eq(6.47) eventually enters the interval (0. (2.50) Then every solution of Eq(6.7. ψ. ψ ∈ (p. . (6.47) with q = 1. φ.7.47) has a prime period-two solution . .48) holds.47) is an unstable saddle point. ∞) is an invariant interval for Eq(6.49) The main result here is the following theorem about Eq(6. The case is treated in Section 6. Concerning prime period-two solutions it follows from Sections 2. 2)-Type Equations 1 − p < q < 1 + p. q . Theorem 6. (d) Assume q > 1 + p. ψ. (6. ∞) : ψ = = 2p}.2 Eq(6. p ]. q (b) Assume that q > 1.1 Invariant Intervals q=1 (6. Furtherq more. [ p .3. φ. every solution of Eq(6. if and only if q = 1 + p. .51) Then every solution of Eq(6.47) eventually enters the interval [ p .48) Furthermore when (6.47) q with initial conditions in (0. Furthermore. . .7. q (0.102 (c) Assume Chapter 6. remains in this interval. Then the positive equilibrium y = p + q − 1 of Eq(6. φ−p 6. (6. .7.

6. y) = px + qy . y refers to the positive equilibrium p + q − 1 of Eq(6. ∞) × [0. x) − x) < 0 for x ∈ (0.47) when p + q > 1 and q ≤ 1. Theorem 6. m q q (b) Assume that (6. Consider the function g : [0.1 (a) Assume that (6. yk ≤ and yk+2 ≥ p p =⇒ yk+2 ≤ p < q q p p =⇒ yk > m+1 > p.2 Here we discuss the behavior of the semicycles of solutions of Eq(6.7. (x − x)(f (x. m ∈ N. whose proof is straightforward and will be omitted. Then the following statements are true for every n ≥ 0: p + q > 1 and q ≤ 1. Let f : [0.7. ∞) → [0.47). ∞) − {x}. ∞) → [0. Lemma 6.51) holds.50) holds. 1+x Observe that g(y) = y and that g increases on [0. ∞) denote the function f (x.7. Hereafter. Then for any k.47). ∞) deﬁned by g(x) = (p + q)x . qm q 6. 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. that is. m ∈ N. .7. Then for any k. ∞). 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. 1+x Semicycle Analysis of Solutions When q ≤ 1 Then f satisﬁes the negative feedback condition.

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

47) when Here the positive equilibrium of Eq(6. Here we discuss the behavior of the solutions of Eq(6.47) is globally asymptotically stable.52) holds.6. Then px+qyn−1 increases in x and so 1+x yn+1 = pyn−1 + qyn−1 pyn + qyn−1 ≤ < yn−1 . 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. More precisely the following statements are true. Then yn ≥ p for all n > 0 and n→∞ lim yn = p. yn−1 ≥ qyn−1 > y + (q − 1)yn > yn 2 pyn + qyn−1 qyn−1 (yn + 1) < ≤ yn−1 . .52) (b) Assume y−1 ≤ p and y0 ≤ p.3 Semicycle Analysis and Global Attractivity When q = 1 q = 1.5 Suppose that q = 1 and let y−1 + y0 > 0. 1 + yn 1 + yn pyn + qyn−1 >y =p+q−1 1 + yn 6.52) holds.47) is y = p.7. (c) Assume either y−1 < p < y0 or y−1 > p > y0 . Then the equilibrium y of Eq(6.7. Case 2. The Case α = C = 0 : xn+1 = (d) Suppose yn < y ≤ yn−1 . (a) Assume y−1 ≥ p and y0 ≥ p. To see that yn+1 < yn−1 .7. Then yn+1 = (e) Observe that yn+1 = and so from which the result follows. Suppose p − qyn−1 < 0. we must consider two cases: Case 1. 1 + yn 1 + yn−1 The last inequality follows from the negative feedback property. Then {yn }∞ n=−1 oscillates about the equilibrium p with semicycles of length one and (6. Suppose p − qyn−1 ≥ 0. Theorem 6. Then yn ≤ p for all n > 0 and (6. (6.

47) has no period-two solutions when q = 1. In fact. 1 + y0 1 + y0 and py1 + p py1 + y0 > = p. We will assume that y−1 ≥ p and y0 > p. 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 . 1 + y0 1 + y0 py1 + y0 py1 + p > = p. Furthermore. 2)-Type Equations Proof. Next. The proof when y0 < p < y−1 is similar and will be omitted. and so convergent. (c) Assume y−1 < p < y0 . we can show that y2 < y0 and by induction y1 = y−1 ≥ y1 ≥ y3 ≥ . > p. . (a) The case where y−1 = y0 = p is trivial. 1 + y0 1 + y0 Similarly. there exist m ≥ p and M ≥ p such that k→∞ lim y2k+1 = m and k→∞ lim y2k = M. . As Eq(6. The case y−1 > p and y0 ≥ p is similar and will be omitted. 1 + y1 1 + y1 It follows by induction that yn ≥ p for all n > 0. It is interesting to note that y−1 = p ⇒ y2k+1 = p for k ≥ 0 and y0 = p ⇒ y2k = p for k ≥ 0. ≥ p and y0 > y2 > y4 > .106 Chapter 6. we claim that the subsequence {y2k+1 }∞ k=−1 is increasing. Thus. (2. Then y1 = and py0 + p py0 + y−1 ≥ = p. since y−1 + py0 > y−1 + y−1 y0 then y1 = py0 + y−1 > y−1 1 + y0 . . . (b) The proof of (b) is similar to the proof of (a) and will be omitted. py0 + y−1 py0 + p y1 = < = p. 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.

By Theorem 6.7.47) is global attractor of all positive solutions of Eq(6.47) we may assume. 2 6. Thus.5 Long-term Behavior of Solutions When q > 1 q>1 Here we discuss the behavior of the solutions of Eq(6.3 (a) we may assume that the initial conditions y−1 and y0 lie in the interval I = (0.7. 2 6. Observe that the consistency condition y ≤ p is equivalent to the condition q q ≤ 1. Proof.47) when which is equivalent to p y> .4 Global Attractivity When q < 1 1 − p < q < 1. Then every positive solution of Eq(6. ∞).4. Here we consider the case where and we show that the equilibrium point y = p + q − 1 of Eq(6. . y) = px + qy 1+x satisﬁes the hypotheses of Theorem 1.7.47) lies eventually in the interval [p. without loss of generality.6. p ]. The Case α = C = 0 : xn+1 = βxn +γxn−1 A+Bxn 107 and the claim follows by induction.5 in the interval I from which the result follows.47). In a similar way we can show that the subsequence {y2k }∞ is decreasing.6 Assume 1 − p < q < 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). y0 ∈ [p. Theorem 6. Now clearly the function q f (x.47) converges to the positive equilibrium y. m = M .7. every positive solution of Eq(6. q In this case. Concerning the long-term behavior of solutions of Eq(6. ∞). and so convergent.7. that y−1 .

8 (a) Assume q = 1 + p. Then every solution of Eq(6.53). an oscillatory solution of Eq(6.47) when q > 1: Theorem 6.47) converges to the positive equilibrium of Eq(6. By applying the results of Section 6. Theorem 6.53) where un ≥ 0 for n ≥ 0.7 After the ﬁrst semicycle.47).47) oscillates about the positive equilibrium y with semicycles of length one. (c) Assume q > 1 + p.7.47) converges to a period-two solution. (2.108 Then the change of variable yn = un + p. Then Eq(6.47) has unbounded solutions. 2)-Type Equations p(q − 1) + qun−1 1 + p + un (6. The character of solutions of Eq(6.47) to the equation un+1 = Chapter 6.53) was investigated in Section 6. transforms Eq(6.5 to Eq(6. we obtain the following theorems about the solutions of Eq(6.7.5. . (b) Assume 1 < q < 1 + p. Then every positive solution of Eq(6.

54) are the solutions of the equation y= py + y . . . γ (6. . In the remainder of this section we will investigate the character of the positive equilibrium of Eq(6. (b) Assume p + 1 > q. .54) also possesses the unique positive equilibrium y = p + 1 − q.8. Our goal is to show that when (6.3.54) and when p + 1 > q.1 (a) Assume p + 1 ≤ q.1. A .54) is unstable and the positive equilibrium y = p + 1 − q of Eq(6.6. The Case α = B = 0 : xn+1 = βxn +γxn−1 A+Cxn−1 109 βxn +γxn−1 A+Cxn−1 6.1.1 and 1. Then the zero equilibrium of Eq(6.54) and q = (Eq(6. The following result follows from Theorems 1. Furthermore the zero equilibrium is a saddle point when 1−p<q <1+p and a repeller when q < 1 − p. the positive equilibrium y = p + 1 − q of Eq(6. q+y So y = 0 is always an equilibrium point of Eq(6. 1.54) and so we will assume without further mention that (6. (6.55) Then the zero equilibrium of Eq(6.54) is locally asymptotically stable. q + yn−1 β γ n = 0. 2)-type Eq(6.) The equilibrium points of Eq(6. Theorem 6.55) holds.8) which by the change of variables reduces to the diﬀerence equation yn+1 = where p= pyn + yn−1 .54) is globally asymptotically stable.54) was investigated in [53].55) holds and y−1 + y0 > 0.8.54) is globally asymptotically stable.8 The Case α = B = 0 : xn+1 = xn = γ yn C This is the (2. Eq(6.

59) the identities (6. for n ≥ 0. (6.56)-(6.58) Note that the positive equilibrium y =p+1−q of Eq(6.8. 2)-Type Equations 6. .110 Chapter 6.1 Invariant Intervals and Semicycle Analysis Let {yn }∞ n=−1 be a positive solution of Eq(6.57) and yn+1 − yn = (p − q)yn + (1 − yn )yn−1 q + yn−1 (6. (6. (6. 1.1 Assume that p<q (6.54) is such that: < 1 if p < q < p + 1 y is = 1 if p = q > 1 if p > q. Then the following identities are easily established: q yn − p yn+1 − 1 = p for n ≥ 0.54).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. Then the following statements are true: .62) and let {yn }∞ n=−1 be a positive solution of Eq(6. (2.60) (6. (6.56) q + yn−1 q p2 [yn − p ] + (p − q)yn−1 q yn+1 − = p p(q + yn−1 ) 2 for n ≥ 0. . Lemma 6.54). When p=q that is for the diﬀerence equation yn+1 = pyn + yn−1 .8. . p + yn−1 n = 0.

yn > 1 and the solution is strictly decreasing. p q then yN +1 > p . then yN +1 < 1. then yN +1 < 1.8. Then the following statements are true: (i) If y0 < 1.3 Assume that p>q (6. Lemma 6. q (ii) If for some N ≥ 0. yN > p . yn < 1 and the solution is strictly increasing. then yN +1 > q p and yN +2 > 1. yN ≤ 1. (v) If for some N ≥ 0.8. then yN +1 < yN .63) and let {yn }∞ n=−1 be a positive solution of Eq(6. then yn = 1 for n ≥ 0. then yN +1 > 1. (v) If for some N ≥ 0. yN ≤ 1. yN ≥ 1. (ii) If y0 = 1. q (iii) If for some N ≥ 0. (iv) If for some N ≥ 0. yN < p . (vi) If for some N ≥ 0. then for n ≥ 0. yN = p . then yN +1 > yN .2 Assume that p=q (6. yN > p . Then the following statements are true: q (i) If for some N ≥ 0. q (ii) If for some N ≥ 0. then yN +1 > 1. (iv) If for some N ≥ 0.8.54).6. . q (iii) If for some N ≥ 0. (vi) If for some N ≥ 0. yN = p . yN < p . yN ≥ 1. yN > q 2 . then yN +1 < 1. then yN +1 = 1. yN > q 2 . The Case α = B = 0 : xn+1 = βxn +γxn−1 A+Cxn−1 111 q (i) If for some N ≥ 0. Lemma 6. p q then yN +1 < p . (iii) If y0 > 1.54). then yN +1 = 1.64) and let {yn }∞ n=−1 be a positive solution of Eq(6. then for n ≥ 0.

3 that every solution converges to its positive equilibrium u= p−q . .1. Hence. 1) the function f (u. .8.8. 1.8.8.54) converges to the equilibrium y = 1. Next. 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 . assume that (6. y is a global attractor. it follows from Lemma 6. 1). . v) = pu + v q+v is increasing in both arguments and by Theorem 1. Without loss of generality we will assume that yn > 1 for n ≥ −1.8.8.1 we have the following result: Theorem 6.112 Chapter 6. This (2.2 Global Stability of the Positive Equilibrium When (6.2 Assume that p + 1 > q and that y−1 + y0 > 0. Then the positive equilibrium y = p + 1 − q of Eq(6. q+1 From the above observations and in view of Theorem 6.64) holds. 2)-Type Equations 6.65) with positive parameters and positive initial conditions. ∞).62) holds. Here it is clear from Lemma 6. every positive solution of Eq(6.2 that every solution of Eq(6.4.3 where we established in Theorem 6. Now in the interval (0. Set yn = 1 + (q + 1)un for n ≥ −1.54) is globally asymptotically stable. (2. This is impossible because y < 1.1 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. 2)-type equation was investigated in Section 6.63) holds.8.54) eventually enters and remains in the interval (0. ∞). it follows from Lemma 6. . When (6.3.54) eventually enters and remains in the interval (1. by Lemma 6.3 that every solution of Eq(6.8. (6. n = 0.

. 1. 1. qyn + yn−1 n = 0.67) . The Case α = A = 0 : xn+1 = βxn +γxn−1 Bxn +Cxn−1 113 βxn +γxn−1 Bxn +Cxn−1 6.66) was investigated in [54]. .66) is y= p+1 q+1 and the linearized equation of Eq(6.6.68) (6.9.66) is locally asymptotically stable. 2)-type Eq(6. (b) Assume that p < q.1.66) about y is zn+1 − p−q p−q zn + zn−1 = 0. (Eq(6. (6. Then the positive equilibrium of Eq(6.9) which by the change of variables xn = reduces to the equation yn+1 = where pyn + yn−1 . By applying the linearized stability Theorem 1.9.) The only equilibrium point of Eq(6. (6. . . .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.1 (a) Assume that p > q.1 we obtain the following result: Theorem 6. . (p + 1)(q + 1) (p + 1)(q + 1) n = 0.66) B β and q = . Then the positive equilibrium of Eq(6. γ C To avoid a degenerate situation we will also assume that p= p = q.9 The Case α = A = 0 : xn+1 = γ yn C This is the (2. .

. ∞) × (0. .69) is asymptotically stable if the eigenvalues of the φ Jacobian matrix JT 2 . . n = 0.66) has no prime period-two solutions. ∞) deﬁned by: T Then u v = φ ψ u v g(u. .70) t2 − (1 − p)t + = 0. 2)-Type Equations 6. . . The prime period-two solution (6.66) in the equivalent form: un+1 = vn n +un vn+1 = pvn +un . q−1 In order to investigate the stability nature of this prime period-two solution. . the second iterate of T . φ. . is a ﬁxed point of T 2 . v) h(u. 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. 1. v) = qv + u and h(u. . (2. ψ. v) p pv+u + v qv+u q pv+u + v qv+u v pv+u qv+u .5 that when p > q. Eq(6. evaluated at lie inside the unit disk. we set un = yn−1 and vn = yn for n = 0. (6. . ψ. v) = = . and write Eq(6. One can see that T2 where pv + u g(u. . .66) possesses a unique prime period-two solution: . . On the other hand when p<q and q > pq + 1 + 3p. Eq(6. 1. .1 Existence of a Two Cycle It follows from Section 2.9.69) where the values of φ and ψ are the (positive and distinct) solutions of the quadratic equation p(1 − p) (6. φ.114 Chapter 6.

ψ > 0. 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 .72) φ ψ and ψ = Inequality (6.71) (6.73) (6. P < 0 and Q < 0 and so Inequality (6. qφ + ψ (6.71) 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 .72) is always true.71). We will use the fact that φ + ψ = 1 − p. 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 > 0.6. Next we will establish Inequality (6. By applying Theorem 1.1 (c).1.9. (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.

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 (6.68) holds.74) is p(1 − p) (p − 1)(q − 1)] q−1 (6.71) 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 . Now observe that the lefthand side of (6. 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.116 if and only if Chapter 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 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. Inequality (6. (2. Finally.

1 by observing that when p < q.66) posseses the prime period-two solution .4 Assume that p<q and (6.2 Semicycle Analysis In this section. . φ. Then the positive equilibrium y of Eq(6. 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) The proof follows from Theorem 1. (b) Assume p < q.9.35). . . In summary.68) holds.66). Then {yn } oscillates about the equilibrium y with semicycles of length two or three.3 Global Stability Analysis When p < q The main result in this section is the following Theorem 6. 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. This function also satisﬁes Condition (1. Then either {yn } oscillates about the equilibrium y with semicycles of length one. Furthermore this prime period-two solution is locally asymptotically stable. where φ and ψ are the two positive and distinct roots of the quadratic Eq(6.9.6. Proof. φ. 6.66): Theorem 6. ψ. . or {yn } converges monotonically to y.9. . the following result is true about the local stability of the prime periodtwo solution (6.66). y) = qx + y is increasing in x and decreasing in y.2 Assume that Condition (6.9.9. after the ﬁrst semicycle.9.4 by observing that the condition p > q implies that the function px + y f (x.70). Theorem 6. (a) The proof follows from Theorem 1. except possibly for the ﬁrst semicycle which may have length one.7. . Then the following statements are true: (a) Assume p > q. the function 2 f (x. y) is increasing in y and decreasing in x.67) holds.66) is globally asymptotically stable.3 Let {yn } be a nontrivial solution of Eq(6. . 6. Then Eq(6. we present a semicycle analysis of the solutions of Eq(6.7. ψ.69) of Eq(6.

9. Set f (x. . .118 Proof.69) when instead of (6. (2. Set f (x. 1. (6. y) is decreasing in x for each ﬁxed y.68) holds. . 2)-Type Equations and note that f (x.6 lim sup yn = ψ n→∞ px + y . φ. Now the conclusion of Theorem 6. . Eq(6. and increasing in y for each ﬁxed x. qx + y Chapter 6.66).75) yN ≥ ψ and yN +1 ≤ φ. φ) = φ and in general yN +2k ≥ ψ and yN +2k+1 ≤ φ for k = 0. yN +2 = f (yN +1 . . Let φ. y > 0. 2 The method employed in the proof of Theorem 1. We consider ﬁrst the case where eventually consecutive terms yi . ψ. p ≤ f (x. y) = px + y . Assume that (6.5 Assume that (6. ψ. Proof. ψ. and lim inf yn = φ n→∞ Now as in the proof of Theorem 1. yN +1 ) ≤ f (ψ. 2 Next we want to ﬁnd more cases where the conclusion of the last theorem holds.4.68) holds. Also clearly. denote the two cycle of Eq(6.66) and for some index N ≥ −1. . with φ < ψ. yN ) ≥ f (φ. ψ) = ψ. Then this solution converges to the two cycle φ. q Finally in view of (6. φ. qx + y from which we conclude that limk→∞ yN +2k = ψ and limk→∞ yN +2k+1 = φ. Theorem 6.67).66) has no prime period-two solution.4 follows as a consequence of Theorem 1. y) ≤ 1 for all x. y) = Then clearly. (6.6 and the fact that y is locally asymptotically stable. . ψ.. Assume that for some solution {yn }∞ n=−1 of Eq(6. . .9. .66) converge to the two cycle (6.4.67)..6 can also be used to establish that certain solutions of Eq(6.4. . yN +3 = f (yN +2 .75) holds. yi+1 lie between m and M .

From Eq(6. φ. m= 2 Since the case in which the solution lies outside of the interval [m. (D) y2n+1 < m and m ≤ y2n < M for every n. m ≤ yi . By our earlier observation we have that M ≥ lim supi→∞ yi . M ] for every k > i. M ] (that is. then we have one of the following four cases: (A) y2n > M and M ≥ y2n+1 > m for every n. Since the proofs are similar for all cases we will give the details in case (A). . either yi > M and yi+1 < m or yi < m and yi+1 > M ). qy2n+2 − p M − M2 lim y2n+1 = . m) ≤ f (yi+1 . . (C) y2n < m and m ≤ y2n+1 < M for every n. Then yk ∈ [m. The Case α = A = 0 : xn+1 = βxn +γxn−1 Bxn +Cxn−1 119 Lemma 6. ψ. and the result follows by induction. qM − p = m and the proof is complete.9. . Proof. ψ.6. (B) y2n+1 > M and M ≥ y2n > m for every n. M ) = pm + M qm + M for m we ﬁnd Thus.1 Suppose that for some i ≥ 0. and so M ≥ lim supn→∞ y2n ≥ lim inf n→∞ y2n ≥ M. yi ) = yi+2 ≤ f (m. . n→∞ qM − p On the other hand by solving the equation M = f (m.9. Lemma 6.9. . Proof. M ) = M. . . M ] is answered by Theorem 6. Hence limn→∞ y2n = M . φ.2 In all cases (A)-(D) the corresponding solution {yn } converges to the two cycle .66) we have y2n+1 = and so y2n − y2n y2n+2 . Using the monotonic character of the function f we have m = f (M. for every i.9. M ]. we are left with the case in which the solution lies eventually in [m. yi+1 ≤ M. 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. limn→∞ y2n+1 M − M2 .5.

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

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

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

.10 Open Problems and Conjectures a. .66) is globally asymptotically stable.5. d ∈ R. y > 0.77) holds. qx + y (6. . To this end. / (b) For every point (x−1 . b.77) and observe that when p > q. Proof.7 Assume that p>q and p ≤ pq + 1 + 3q.6.10.4. . 1. the function f (x. qM + m m= pm + M .1 (See [64]) Assume that (a) Investigate the forbidden set F of the diﬀerence equation xn+1 = axn + bxn−1 xn .5.4 Global Stability Analysis When p > q Here we have the following result: Theorem 6. . 2 pM + m .4. Finally observe that when (6. investigate the asymptotic behavior and the periodic ∞ nature of the solution {xn }n=−1 . Open Problems and Conjectures 123 6. y) is increasing in x for each ﬁxed y. c.9. x0 ) ∈ F. Then the positive equilibrium y of Eq(6.9. y) = px + y . cxn + dxn−1 n = 0. set f (x.10. Now the result is a consequence of Theorem 1. the only solution of the system M= is m = M. Also 1 ≤ f (x. Open Problem 6. y) ≤ p q for all x. We will employ Theorem 1. qm + M 6. and is decreasing in y for each ﬁxed x.

.) Conjecture 6.6 for the autonomous case. this conjecture has been conﬁrmed for q ≤ 1 + 4p. 1.2 Assume that αn . Show that every positive solution of Eq(6. 1 + yn n = 0. .10. . are convergent sequences of real numbers with ﬁnite limits. Investigate the forbidden set and the asymptotic character of solutions of the nonautonomous Riccati equation xn+1 = αn + βn xn . ∞). ∞). .10) has a ﬁnite limit. Show that every positive solution of Eq(6. 2)-Type Equations Open Problem 6. every positive solution converges to a.10.10.5. See also [36].1 Assume p. .10.4.2. An + Bn xn n = 0.) Conjecture 6. q ∈ (0. 1. (See Section 1.) Conjecture 6. βn . 1.4.) Chapter 6. Show that the equation yn+1 = p + yn−1 . not necessarily prime. An . q ∈ (0. (2.3 Assume p ∈ (0.2 Assume that p. .3.3. . Bn ∈ R for n = 0. period-two solution.) . (Note that by Theorem 6. . (Note that by Theorem 6. has a positive and monotonically decreasing solution. (Note that by Theorem 6. .11) has a ﬁnite limit.124 (See Section 1.6. we need only to conﬁrm this conjecture for p > 2(q +1) and q ≥ 1. . ∞).

6. Open Problems and Conjectures Conjecture 6. q ∈ (0.7.10. 125 Show that every positive solution of Eq(6. q ∈ (0. Conjecture 6. .4 Assume that p.10.10.) Open Problem 6. as n → ∞.3 Assume that p. Find the basin of attraction of the period-two solution of Eq(6. ∞) and q > 1 + 4p.10. q ∈ (0. ∞) and q > pq + 1 + 3p. y0 ) ∈ R × R through which the solution of the equation yn+1 = yn + yn−1 1 + yn is well deﬁned and converges to 1.10.4 Assume that p.27). Find the basin of attraction of the period-two solution of Eq(6.) Open Problem 6.10. ∞). (Note that by Theorem 6. this conjecture has been conﬁrmed for p ≤ pq + 1 + 3q.9.66). Open Problem 6. (See Section 6.5 Assume p > q.66) has a ﬁnite limit.6.5 Find the set of all initial conditions (y−1 . Show that every positive solution of Eq(6.27) either converges to a ﬁnite limit or to a two cycle.

80) (6.82) (6. y0 ) ∈ G.6 Find the set of all initial conditions (y−1 .79) (6. ∞). (6. Open Problem 6.84) yn − yn−1 For those equations from the above list for which you were successful.10. Investigate the asymptotic behavior and the periodic nature of solutions of the diﬀerence equation αxn + βxn−1 + γxn−2 . C ∈ (0.10.81) (6. A. . (2.Then for every (y−1 . B.10.83) yn − yn−1 yn − 1 yn − yn−1 1 − yn−1 1 − yn−1 yn − yn−1 yn + yn−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. . γ. 2)-Type Equations Open Problem 6.126 Chapter 6. . 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. .7 For each of the following diﬀerence equations determine the “good” set G ⊂ R × R of initial conditions (y−1 . extend your result by introducing arbitrary real parameters in the equation.8 Assume α.78) (6. Open Problem 6. 1. as n → ∞. n = 0. xn+1 = Axn + Bxn−1 + Cxn−2 Extend and generalize. β. . y0 ) ∈ R × R through which the equation is well deﬁned for all n ≥ 0.

. .10 Assume that {pn }∞ and {qn }∞ are period-two sequences n=0 n=0 of nonnegative real numbers. qn + yn−1 pn yn + yn−1 . ∞) such that the solutions {yn }∞ n=−1 of Eq(6. (6.6. qn yn + yn−1 pn yn + qn yn−1 . 1. qn + yn n = 0. .23) are bounded. (b) Let (y−1 .92). 1. 1. . n = 0. y0 ) ∈ B. qn yn + yn−1 yn+1 = yn+1 = Open Problem 6. 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 .86) (6. 1 + yn pn yn + yn−1 . . 1.90) (6. n=−1 .85)-(6.88) (6. ∞) × (0. Investigate the global character of all positive solutions of Eqs(6.10.92) pn + qn yn . 1 + yn pn + yn−1 . 1.89) (6. 1. . Investigate the asymptotic behavior of {yn }∞ . 1 + yn−1 yn + pn . p = n→∞ pn lim lim and q = n→∞ qn . n = 0.10. . . yn + qn yn−1 pn + qn yn−1 . . 1. Open Problems and Conjectures 127 ∞ Open Problem 6. 1. n = 0. n = 0.10. .85) (6. y0 ) ∈ (0. . . . (a) Find the set B of all initial conditions (y−1 . .87) (6.91) (6. . Open Problem 6. n = 0. . Extend and generalize. .9 Assume that {pn }n=0 and {qn }∞ are convergent sequences of n=0 nonnegative real numbers with ﬁnite limits. n = 0. . .11 Assume q ∈ (1. . . . .10. . . ∞). n = 0.

y0 ∈ (0. ∞) and that α ∈ (2. ∞) through which the solutions {xn }∞ n=−1 of the diﬀerence equation xn+1 = (1 − xn − xn−1 )(1 − e−Axn ). y0 ) ∈ B. 1) with α + β > 1 + β .14 (Discrete Epidemic Models. (6. the periodic nature. and γ ∈ (0.128 Open Problem 6. .10. (6. 2)-Type Equations (a) Find the set B of all initial conditions (y−1 . and the asymptotic behavior of the solution {xn }∞ n=−1 of Eq(6. See [9]. (a) Determine the set G of initial conditions (y−1 . . . ∞). . x0 ) ∈ G. and [2]) Assume α ∈ (1. n = 0. (c) Extend and generalize. Investigate the boundedness character. 2 Show that the positive equilibrium of System (6. Open Problem 6. Chapter 6.94). 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 . n = 0. ∞) × (0.12 Assume q ∈ (1. (b) Let (x−1 . β ∈ (0.94) .10. are nonnegative. . y0 ) ∈ (0. 1 β = 1 and γ = . . 1.10.6 Assume x0 . ∞) such that the solutions {yn }∞ n=−1 of Eq(6. ∞).93) is globally asymptotically stable. .10. Investigate the asymptotic behavior of {yn }n=−1 . ∞) × (0. [3]. ∞). (2. ∞).47) are bounded. Open Problem 6. 1. ∞ (b) Let (y−1 .93) Conjecture 6. y0 ) ∈ (0.13 (A Plant-Herbivore System. See [14].) Let A ∈ (0. 3).

Open Problem 6. ∞) with c1 + c2 > 0. Open Problems and Conjectures Open Problem 6. n = 0. qyn + yn−k yn + pyn−k . 1. n = 0. .10.10. Open Problem 6.10.99) (6. 1. . ∞) and k ∈ {2. Investigate the global behavior of all positive solutions of each of the following diﬀerence equations: p + qyn yn+1 = . 1.}. n = 0. . (6.10. . ∞). n = 0. . . 1. 1.98) (6. n = 0. and c1 . c2 ∈ [0. ∞). 1. .100) (6. q ∈ [0. . 1 + yn p + yn−k . q + yn−k pyn + yn−k . 1) and xn+1 = αxn e−yn + β yn+1 = αxn (1 − e −yn β ∈ (1. (6. yn + q pyn + yn−k . 1). . . . n = 0. See [48]).17 Assume that p. ) which may be viewed as a population model. 1.6. . . n = 0. with positive initial conditions. Investigate the global character of all positive solutions of the system . 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 . .97) (6. . . . . . . . 3. yn + qyn−k p + qyn−k . Assume a ∈ (0. . .95) 1 + yn−k yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn + p .15 (The Flour Beetle Model. . 1. qyn + yn−k n = 0. .96) (6. 1.16 (A Population Model) Assume α ∈ (0. . . . . b ∈ (0.101) . n = 0. . . .

10. n = 0.27).10. (6.10. 1. Obtain necessary and suﬃcient conditions on f so that every positive solution of the equation xn+1 = f (xn−2 . 2)-Type Equations Open Problem 6. 1.102) converges to a period-two solution. also converges to a period-two solution.19 Assume that every positive solution of an equation of the form xn+1 = f (xn . Obtain necessary and suﬃcient conditions on f so that the same two cycle is a locally asymptotically stable solution of Eq(6. . xn−1 ). .18 Obtain a general result for an equation of the form yn+1 = f (yn . . (6. .11) and (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). n = 0. yn−1 ). xn−1 ). Extend and generalize.103) Open Problem 6. . . . . Open Problem 6. 1. .130 Chapter 6.20 Assume that Eq(6. (2. .

. . . .(7. . . 3)-Type Equations 7.2 The Case β = γ = 0 : xn+1 = α yn A α A+Bxn +Cxn−1 This is the (1. (7. . .1) (7.3) α .Chapter 7 (1. (7. 3)-type equations are positive and that the initial conditions are nonnegative.4) αC . 7. 3)-type Eq. 1. n = 0. (7. . A + Bxn + Cxn−1 n = 0. 1. . 3)-type: Please recall our classiﬁcation convention in which all the parameters in the above (1.1) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= 1 . A2 . 1.2) Eq(1) contains the following three equations of the (1. A + Bxn + Cxn−1 βxn . . A + Bxn + Cxn−1 n = 0. . 1 + pyn + qyn−1 αB A2 and q = 131 n = 0.1 Introduction xn+1 = xn+1 = and xn+1 = γxn−1 . 1. .

Theorem 7.4) is globally asymptotically stable. 1 + pyn + qyn−1 p= n = 0.1. The following result is now a straightforward consequence of either the stability trichotomy Theorem 1.2. q+1 The following result is a straightforward consequence of Theorem 1.5) is globally asymptotically stable. . . (7.4. . 3)-Type Equations One can easily see that the positive equilibrium of Eq(7.7 in the interval [0.4.2) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where yn . 1.4.5) is globally asymptotically stable.3. Theorem 7. y0 ∈ (0.4) has no prime period two solutions. ∞) and p > 1.1 The positive equilibrium of Eq(7. or Theorem 1.3. 7.5 in the interval [0.5) always has zero as an equilibrium point and when p>1 it also has the unique positive equilibrium point y= p−1 . p ].1 Assume p ≤ 1.4.3. 3)-type Eq(7.2 Assume that y−1 .4. (1. It also follows by applying Theorem 1. 1].3 The Case α = γ = 0 : xn+1 = A yn C βxn A+Bxn +Cxn−1 This is the (1. A C Eq(7. Then the zero equilibrium of Eq(7.132 Chapter 7.5) β B and q = .4) is locally asymptotically stable for all values of the parameters and that Eq(7. .2. q Theorem 7. Then the positive equilibrium of Eq(7. The following result is a straightforward consequence of Theorem 1.

1 and 1. C (7. . .1. .4. 3)-type Eq(7. 1+q The subsequent result is a straightforward application of Theorems 1. ψ. .6) has a prime periodtwo solution . φ. B .7. . Theorem 7. . p + qyn + yn−1 A γ n = 0. The Case α = β = 0 : xn+1 = γxn−1 A+Bxn +Cxn−1 133 γxn−1 A+Bxn +Cxn−1 7.4 The Case α = β = 0 : xn+1 = γ yn C This is the (1.3) which by the change of variables xn = reduces to the equation yn+1 = where p= yn−1 . ψ.6) is locally asymptotically stable when q<1 and is unstable (a saddle point) when q > 1. .3. (b) Assume that p < 1. φ.1 (a) The zero equilibrium of Eq(7. .6) and q = Eq(7. 1. Then the positive equilibrium of Eq(7. . .6) always has the zero equilibrium and when p<1 it also has the unique positive equilibrium y= 1−p .1. Concerning prime period-two solutions one can see that Eq(7.6) is globally asymptotically stable when p≥1 and is unstable (a repeller) when p < 1.4. if and only if p < 1.

. . 1 − p. 1 − p.6 that here JT 2 0 1−p = 1 p+q(1−p) q(1−p) − p+q(1−p) 1−p . Proof. . . . 2 and q > 1. φ. . 1 − p − φ. Conjecture 7. 0. .6) has a ﬁnite limit. On the other hand. . Therefore 1−p the prime period-two solution (7. 1 − p.6) is locally asymptotically stable. . . . . . of Eq(7. . 1). with 0 ≤ φ ≤ 1 − p and φ = Theorem 7.4.7) both eigenvalues of JT 2 7. . q ∈ (0. . .6) are given by . (7.5. It follows from Section 2. .2 Assume that p<1 Then the period-two solution . .8) 0 p 0 lie in the disk |λ| < 1.7) (7. . 2 and in view of (7.5 Open Problems and Conjectures p. 1 − p. 3)-Type Equations p < 1 and q = 1 the only prime period-two solution is . 0. (1. 1 − p − φ. when p < 1 and q = 1 all prime period-two solutions of Eq(7. 0. φ.1 Assume Show that every positive solution of Eq(7.8) is locally asymptotically stable.134 Furthermore when Chapter 7. 0.

and investigate ∞ the long term-behavior of the solution {xn }n=−1 . of Eq(7. Open Problem 7.5.2 Investigate the asymptotic character and the periodic nature of all positive solutions of the diﬀerence equation xn+1 = where p.3 For each of the equations given below. 0.11).9)-(7. 1 + pxn + qxn−1 + xn−2 n = 0. . 1.9) (7. xn−2 . ∞). . Extend and generalize. Open Problems and Conjectures Conjecture 7.5.5. . 1 + xn + xn−1 (7. 1 − p. let (x−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 . . 135 Show that every positive solution of Eq(7. 0. Open Problem 7. . .7. ﬁnd the set G of all points (x−1 . 1 − p.10) (7. ∞). x0 ) ∈ G be an initial point through which the corresponding equation is well deﬁned for all n ≥ 0. . q ∈ [0. periodtwo solution. .5.6).6) converges to a. Find the basin of attraction of the two cycle . 1) and q ∈ (1.1 Assume p ∈ (0. .5.11) Now for each of the equations (7. Open Problem 7. not necessarily prime.2 Assume p < 1 and q ≥ 1.

. . Investigate the global character of all positive solutions of Eqs(7. . 1 + pn yn + qn yn−1 yn−1 . 1. n = 0.5. 1. . 3)-Type Equations Open Problem 7. 1. . n = 0.4 Assume that {pn }∞ and {qn }∞ are convergent sequences of n=0 n=0 nonnegative real numbers with ﬁnite limits.5 Assume that {pn }∞ and {qn }∞ are period-two sequences of n=0 n=0 nonnegative real numbers. . Extend and generalize. p = lim pn n→∞ and q = n→∞ qn . (1. . (7. .12) (7. 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 .12)-(7. . 1 + pn yn + qn yn−1 yn .136 Chapter 7. pn + qn yn + yn−1 n = 0.5.14) Open Problem 7. .14).13) (7. .

4) (see Section 6. n = 0. and the denominators are always positive. (8.1) (8. 1.1) is a linear diﬀerence equation and can be solved explicitly. . .3) by the change of the variables xn = yn + 137 γ . C . 3)-type: xn+1 = xn+1 = and xn+1 = α + βxn + γxn−1 . . yn+1 = (α + βγ ) + γyn−1 B . namely. .2) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. Eq(8. (8. B is reduced to a (2. 1. .3) n = 0. 1. 1. . .5). 1)-Type Equations 8. Cxn−1 n = 0. 2)-type equation of the form of Eq(6.Chapter 8 (3. Bxn Eq(1) contains the following three equations of the (1. A α + βxn + γxn−1 . Eq(8. . . β + Byn n = 0.2) by the change of the variables xn = yn + β . .1 Introduction α + βxn + γxn−1 . . Eq(8. . . . the initial conditions are nonnegative.

yn+1 (α + βγ ) + βyn C .4) be a period-two solution of Eq(8. (3.2. . .2).3 Determine the set G of all initial points (x−1 . . . 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 Open Problems and Conjectures Open Problem 8.2.1 Show that every positive solution of Eq(8. 8.2. 2)-type equation of the form of Eq(6.2) (see Section 6. . ψ. ∞) × (0. Determine the set of initial conditions (x−1 . . = γ + Cyn−1 n = 0. . . n=−1 Open Problem 8. φ. Hence there is nothing else remaining to do about these equations other than to pose some Open Problems and Conjectures. Conjecture 8. (8. ∞) through which the solutions of Eq(8.2 (a) Assume γ = β and let . Open Problem 8.3) namely. n=−1 .4).4 Determine the set G of all initial points (x−1 .2. (b) Let {xn }∞ n=−1 be a positive solution of Eq(8. Determine the basin of attraction of (8. 1. ψ.2. φ. Open Problem 8.2) are bounded. x0 ) ∈ (0.4). 1)-Type Equations is reduced to a (2.2) which converges to (8. .3) converges to the positive equilibrium of the equation. 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 }∞ .138 Chapter 8. Determine the values of φ and ψ in terms of the initial conditions x−1 and x0 .1 Assume γ > β. .

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.

.4) becomes yn+1 = q + rq 2 + qyn . is equal to four. and y4k+2 ≤ q.4).9) (9.9) follows by straightforward computation. Theorem 9. (iv) y−1 < q and y0 ≤ q =⇒ y4k−1 < q. In this case Eq(9.3 Suppose that p = q + rq 2 and let {yn }∞ n=−1 be a nontrivial solution of Eq(9. y4k+1 < q. 1. y4k ≤ 1. Proof. (2. The limit (9. (iii) y−1 > q and y0 ≥ q =⇒ y4k−1 > q.2. Identity (9. 1 + yn + ryn−1 n = 0. y4k+1 > q. 1 + yn + ryn−1 lim yn = y. 3)-Type Equations qr (q − yn−1 ).4 Assume p = q(qr + 1) and let {yn }∞ n=−1 be a solution of Eq(9.10) is a consequence of the fact that in this case qr ∈ (0. (9. and the only equilibrium point is y = q.4) has no period-two solution. y4k ≥ 1. y4k ≤ 1. . and y4k+2 ≤ q.2. More precisely. and y4k+2 ≥ q. 2 Here we present a semicycle analysis of the solutions of Eq(9. then n→∞ Chapter 9. Then yn+1 − q = Furthermore when qr < 1.146 Lemma 9. ∞) and Eq(9. the following statements are true for all k ≥ 0: (i) y−1 > q and y0 ≤ q =⇒ y4k−1 > q. (ii) y−1 < q and y0 ≥ q =⇒ y4k−1 < q.4) when p = q(1 + rq).4). The proof follows from identity (9. . y4k ≥ 1. Then this solution is oscillatory and the sum of the lengths of two consecutive semicycles. and y4k+2 ≥ q. y4k+1 > q. y4k+1 < q. 2 .5). excluding the ﬁrst.10) Proof.

for the sake of contradiction. q−1+ (q − 1)2 + 4p 2 Hence.1. yN +1 < p−q . (ii) yN > q.3. The Case γ = 0 : xn+1 = α+βxn A+Bxn +Cxn−1 147 9.2. then yN +1 < b.2. qr p−q . q] qr [q.2. Set b= and observe that yn+1 − b = (q − p)(yn − b) − r(p + qb)yn−1 .2. Now assume for the sake of contradiction that yn > b for n > 0. Then yn > y for n ≥ 0 and so which is a contradiction. qr n→∞ lim yn = y ∈ I and yN +2 < p−q . p > q(qr + 1). the following is true: Lemma 9. Then every solution of Eq(9.2. 2 I= [ p−q . qr and yN +2 < . ∞ Proof. yN +1 > q.4 Global Asymptotic Stability The following lemma establishes that when p = q(qr + 1). Then clearly. there exists N > 0 such that one of the following three cases holds: (i) yN > q.9.4) lies eventually in I. Let {yn }n=−1 be a solution of Eq(9. q−1+ (q−1) +4p ] if p ≤ q. (1 + yn + ryn−1 )(1 + b) n ≥ 0. if for some N . First assume that p ≤ q. every solution of Eq(9.4) is eventually trapped into one of the three invariant intervals of Eq(9.4) described in Lemma 9.5 Let I denote the interval which is deﬁned as follows: √ 2 [0.4). y ∈ I. Then by Lemma 9. More precisely. also assume that the solution {yn }∞ n=−1 is not eventually in the interval I. p−q ] qr if if q < p < q(qr + 1). yN ≤ b. Next assume that q < p < q(qr + 1) and.

4. The proof when p > q(qr + 1) is similar and will be omitted. together with the appropriate convergence Theorems 1. the following results are true for Eq(9.4 (a) Assume that either p ≥ q + q 2 r. Then the equilibrium y of Eq(9. and (q − 1)2 (r − 1) ≤ 4p. then yN + ryN − p > 0 and 2 if yN ≤ p−q . qr The desired contradiction is now obtained by using the Identity (9.2. 3)-Type Equations ≤ yN +1 ≤ q.4) is globally asymptotically stable. For example.7 and 1.4.5. (b) Assume that either (iii) r > 1 (ii) r ≤ 1. qr Also observe that 2 if yN ≥ q.148 (iii) yN > q. p−q qr Chapter 9. 1+r Then the equilibrium y of Eq(9.4).2. p< p≤q or q < p < q + q2r and that one of the following conditions is also satisﬁed: (i) q ≤ 1. . and yN +2 < p−q . By using the monotonic character of the function f (x. we can obtain some global asymptotic stability results for the solutions of Eq(9.1. Proof.7) from which it follows that for j ∈ {0. 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 .4) is globally asymptotically stable. 3}. (2. 2. then yN + ryN − p < 0. y) = p + qx 1 + x + ry 2 in each of the intervals in Lemma 9. 1.4): Theorem 9.

1 + yn + ryn−1 q= γ .1 we see that when y−1 . A2 and r = . A n = 0. M ) and M = f (M. then M = m.11) A yn B αB .2.5. We will employ Theorem 1. y0 ∈ 0. . m) then M = m. . 2 decreases in y.4. This system has the form m= p + qm 1 + m + rM and M = p + qM . B (9.2.4.1 and Theorem 1. the function f increases in x and and that in the interval 0. 3)-type Eq(9. 0.3 The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 This is the (2.5 and so it remains to be shown that if m = f (m.2) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + qyn−1 . For instance. C .3.4.2.1 and 9. 1 + M + rm Hence (M − m)(1 − q + M + m) = 0. . If m + M = q − 1 then m and M satisfy the equation (r − 1)m2 + (r − 1)(1 − q)m + p = 0. It is easy to check that y ∈ 0. 1.9.2.7 and 1.4. √ q−1+ (q−1)2 +4p (b) In view of Lemma 9. . Clearly now if Condition (ii) or (iii) is satisﬁed. m = M from which the result follows. Now if m + M = q − 1.2. then yn ∈ 2 √ √ q−1+ (q−1)2 +4p q−1+ (q−1)2 +4p for all n ≥ 0. The proof when q < p < q + q 2 r holds follows from Lemma 9. 2 9. The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 149 (a) The proof follows from Lemmas 9. this is the case if Condition (i) is satisﬁed. 2 2 √ 2 +4p q−1+ (q−1) .4 and Theorems 1.

ψ.12) holds. . 1 + φ + rψ It now follows after some calculation that the following result is true.3. ψ. if and only if (9.11) has a prime period-two solution .1 we obtain the following result. φ. . . be a period-two cycle of Eq(9. Here we make some simple observations about linearized stability. (b) The equilibrium y of Eq(9. ψ. Then φ= p + qφ 1 + ψ + rφ and ψ = p + qψ .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) . .11). . φ. . Furthermore when (9. and the convergence of solutions in certain regions of initial conditions.1 Eq(9.11) is unstable.12) holds. when q > 1 and (r − 1)(q − 1)2 + 4pr2 < 0. 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. ψ. . invariant intervals.1 Let Existence and Local Stability of Period-Two Cycles . and more precisely a saddle point.1. . 3)-Type Equations This equation has not been investigated yet.12) 9. (9. Lemma 9. Theorem 9. Eq(9. t+ r 1−r . φ.150 Chapter 9. φ.1 (a) The equilibrium y of Eq(9. . (2.3. . . By employing the linearized stability Theorem 1.3.11) is locally asymptotically stable when either q≤1 or q > 1 and (r − 1)(q − 1)2 + 4pr2 > 0. .

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

Observe that. ∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1.14) (9. 2 (1 + φ + rψ)2 ∂v (1 + ψ + rφ) (1 + φ + rψ)2 Set S= and D= ∂h ∂g (φ. This inequality is equivalent to (q − pr + qφ) (q − pr + qψ) − (1 + ψ + rφ)2 (1 + φ + rψ)2 < 0. ψ) ∂u ∂v ∂h ∂g ∂h ∂g (φ. if and only if |S| < 1 + D < 2.16) First we will establish Inequality (9. (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 .152 Chapter 9. ψ) = + . ψ) − (φ.15) (9. ∂u (1 + ψ + rφ)2 (1 + φ + rψ)2 ∂h (p + qψ)(p + qφ) q − pr + qφ (φ. ψ) (φ. ψ) (φ.13) (9.1 that both eigenvalues of JT 2 unit disk |λ| < 1.14). ψ) + (φ. Thus .13) is equivalent to the following three inequalities: D<1 S <1+D −1 − D < S. ψ). (2. Inequality (9. ψ) = . 3)-Type Equations ∂h (p + qψ)(pr − q − qψ) (φ.1. (9.

In summary. 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) = . 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).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. we have established the following result: . Next we turn to Inequality (9.9. After some lengthy calculation one can see that this inequality is also a consequence of Condition (9.12) 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.12).15). Also by using Condition (9. Finally Inequality (9.3. r2 (1 − r) −r2 p + rp − q + q 2 + qr = rp(1 − r) + q(q − 1) + qr > 0. After some lengthy calculation one can see that this inequality is a consequence of Condition (9.

12) holds. φ. if and only if (9.2 Eq(9. gives a list of invariant intervals for Eq(9. yk ∈ [0.11) has a unique prime period-two solution .3.11) we see that when yk−1 . r when 1 < q < pr < q + q2 q − .3. Furthermore when (9. . which can be established by direct calculation. then yk+1 = p + qyk−1 p + qyk−1 ≤ = g(yk−1 ) ≤ g(b) ≤ b. ψ.12) holds. q r (c) q pr − q .154 Chapter 9. .2 Invariant Intervals The following result. q−1+ (q − 1)2 + 4pr 2r when pr ≤ q. 9. b]. .2 Eq(9. ψ.11) possesses the following invariant intervals: (a) 0. Using Eq(9. Lemma 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. this period-two solution is locally asymptotically stable. 3)-Type Equations Theorem 9. . . φ. (b) pr − q q . . r q when pr ≥ q + q2 . (2.11).3. . r r Proof. 1 + yk + qyk−1 1 + ryk−1 The proof follows by induction.

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

then Proof. . the function f decreases in x and increases 2r in y. for all n ≥ 0. 3)-Type Equations 9. y0 ∈ 0.6.2.3. the following results are true for Eq(9. In view of Lemma 9.8) we can obtain some convergence results for the solutions with initial conditions in the invariant intervals.2 we see that when y−1 .11) with initial conditions in the invariant interval 0. It is easy to check that y ∈ 0. 2 Theorem 9. For example. 2r 2r √ q−1+ (q−1)2 +4pr and that in the interval 0.3 Convergence of Solutions p + qy 1 + x + ry By using the monotonic character of the function f (x.3 Assume that pr ≤ q and that one of the following conditions is also satisﬁed: (i) q ≤ 1.3. Then every solution of Eq(9.3.3. 2 q−1+ √ (q−1)2 +4pr (ii) r < 1 and (r − 1)(q − 1)2 + 4pr2 ≥ 0.156 Chapter 9. y) = in each of the intervals in Lemma 9. q−1+ (q − 1)2 + 4pr 2r converges to the equilibrium y.5-1. The result now follows by employing Theorem 1.4.3.4 (a) Assume that 1 < q < pr < q+ qr − q and that one of the following r conditions is also satisﬁed: (i) r ≥ 1.4. (iii) r < 1 and (r − 1)(q − 1)2 + 4pr2 ≥ 0. (2.4. 2r √ √ q−1+ (q−1)2 +4pr q−1+ (q−1)2 +4pr yn ∈ 0. together with the appropriate convergence theorem (from among Theorems 1. (ii) r ≥ 1.11): Theorem 9. .

3.6.3.4 Global Stability By applying Theorem 1. the function f decreases check that y ∈ q . r q converges to the equilibrium y. pr−q and that in the r q in both x and in y.5 Semicycle Analysis When pr = q + Here we present a semicycle analysis of the solutions of Eq(9. .11). Then the positive equilibrium of Eq(9.11) is globally asymptotically stable.4. Theorem 9.3. 1. q2 r q2 .4.7. r 9. Then every solution of Eq(9. . It is easy to 9. . The result now follows by employing Theorem 1. In and the only positive equilibrium is y = q .3. q r converges to the equilibrium y. The result now follows by employing Theorem 1. r . q r interval q .9.11) when pr = q + this case Eq(9.11) with initial conditions in the invariant interval pr − q q .2 we conclude that yn ∈ pr−q q .2 we conclude that yn ∈ for all n ≥ 0. (a) In view of Lemma 9. r2 + r2 yn + r3 yn−1 n = 0. we obtain the following global asymptotic stability result.5 Assume r ≥ 1 and pr ≤ q. 2 q pr−q .3 to Eq(9.4. Proof.11) becomes yn+1 = qr + q 2 + qr2 yn−1 . The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 157 Then every solution of Eq(9. It is easy to (b) In view of Lemma 9. y) decreases in x and increases in y.11) with initial conditions in the invariant interval q pr − q .r q pr−q q . (b) Assume that pr > q + qr . pr−q r q for all n ≥ 0.3.3.r q 2 check that y ∈ and that the function f (x.

r r r (1 + yn + ryn−1 ) 2 9. . invariant intervals. γ This equation has not been investigated yet. and the convergence of solutions in certain regions of initial conditions. . Proof. 3)-type Eq(9.11). 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 . 1.4 The Case α = 0 : xn+1 = βxn +γxn−1 A+Bxn +Cxn−1 This is the (2. q+1 (9. C and r = A . Eq(9. The equilibrium points of Eq(9. Then after the ﬁrst semicycle.158 2 Chapter 9. . Here we make some simple observations about linearized stability. r + (1 + q)y Hence zero is always an equilibrium point and when p + 1 > r.1.1 and Theorem 1.17) γ yn .17) are the nonnegative solutions of the equation y= (p + 1)y .3. 3)-Type Equations Theorem 9.17) also possesses the unique positive equilibrium y= p+1−r . . The proof follows from the relation yn+1 − q q qr2 = ( − yn ) 3 . γ q= B . (9. r + qyn + yn−1 β .3.6 Suppose that pr = q + qr and let {yn }∞ n=−1 be a nontrivial solution of Eq(9.1. (2. C n = 0.18) The following theorem is a consequence of Theorem 1.

17) is globally asymptotically stable. . Then the zero equilibrium of Eq(9. 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 t2 − (1 − p − r)t + p(1 − p − r) = 0.5 that the following result is true. . Then the positive equilibrium of Eq(9.20) (9.2 Assume that (9.17) has a prime period-two solution .3 Eq(9. ψ. Theorem 9.1.9.1. (9.17) about its positive equilibrium y is zn+1 − p − q + qr q−p+r zn − zn−1 = 0. . . it follows from Section 2. ψ. 1. Then the zero equilibrium of Eq(9. The Case α = 0 : xn+1 = Theorem 9. The following result is a consequence of Theorem 1. (b) Assume that p + 1 > r.17).18) holds. .17) is locally asymptotically stable when q + r < 3p + 1 + qr + pq. .1 βxn +γxn−1 A+Bxn +Cxn−1 159 (a) Assume that p + 1 ≤ r.4. (p + 1)(q + 1) (p + 1)(q + 1) n = 0. The linearized equation associated with Eq(9. q−1 . . φ. and unstable (a saddle point) when q + r > 3p + 1 + qr + pq. . Theorem 9.4.19) Concerning prime period-two solutions for Eq(9.20) holds.4. φ. if and only if (9. Furthermore the zero equilibrium is a saddle point when 1−p<r <1+p and a repeller when r < 1 − p. .4.17) is unstable.

K] when p = q 2 where K > 0 is an arbitrary number. 3)-Type Equations 9. qr p − q2 when q 2 < p ≤ q 2 + r and p + q > r. (b) 0.1 Invariant Intervals The following result. K] when q 2 < p ≤ q 2 + r and p + q ≤ r. when q 2 − qr ≤ p < q 2 and p + q > r. q2 pr −p (q − r)(p − q 2 ) − qr + [0.4. which can be established by direct calculation.1 Eq(9.17). when p > q 2 + r where ((r − q)(p − q 2 ) + qr)2 + 4q 3 r(p − q 2 ) 2q(p − q 2 ) . where K > 0 is an arbitrary number and pr .4.17) possesses the following invariant intervals: (a) 0. K2 −p [0. Lemma 9. gives a list of invariant intervals for Eq(9. (2.160 Chapter 9. . and [0. q2 where K2 = when q 2 > p + qr. K] when q 2 − qr ≤ p < q 2 and p + q ≤ r. ((q − r)(q 2 − p))2 + 4p3 qr2 2pqr (q − r)(q 2 − p) + . K1 p − q2 K1 = (c) 0. and q 2 + 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 .

yk ∈ 0. 2 q2 − p q −p . ) ≤ K1 . qr Hence for yk−1 . K) = (p + q)K with K = pr . q2pr we obtain −p yk+1 = f (yk . Therefore for any K > 0. The Case α = 0 : xn+1 = Proof. it is satisﬁed under the condition r + q 2 > p. r + (q + 1)K qr where we set K = p−q2 . 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. yk−1 ) ≤ f (K1 . 1 ≤ K. 1 ≤ K. K1 p−q 2 we can see that qr qr . −p Then one can see that for yk−1 . yk−1 ) ≤ f (K. yk ∈ f qr . while when q 2 + q > r the inequality is true when p+q−r K≥ . First assume that q 2 < p ≤ q 2 + r . Then for yk−1 . 2 p−q p − q2 pr . yk ∈ f one can see that ≤ K2 . The last inequality is always satisﬁed when p + q ≤ r and when p + q > r. p−q2 we obtain qr . K1 ≤ yk+1 = f (yk . K2 ≤ yk+1 = f (yk . this inequality is satisﬁed for any K > 0. K) ≤ K is equivalent to q 2 +q−r ≤ (q+1)K. y) ≤ f (K. yk ∈ 0. r + (q + 1)K pr . K2 q 2 −p Second assume that p < q 2 − qr. yk−1 ) ≤ f (K. y) is increasing in both arguments for y < and it is increasing in y and decreasing in x for y ≥ q2pr .9. p−q 2 0 ≤ f (x.4. q 2 −p (c) It is clear that the function f (x. y) = r + qx + y is increasing in both arguments. Now observe that the inequality f (K. y) is increasing in both arguments for x < qr and it is increasing in x and decreasing in y for x ≥ p−q2 . When q 2 + q ≤ r. K) = (p + q)K Second assume that p > q 2 + r. K). q 2 −p First assume that q 2 − qr ≤ p < q 2 . q+1 (b) It is clear that the function f (x. yk−1 ) ≤ f K2 . Then for yk−1 . yk+1 = f (yk . 2 pr pr .

we can obtain some convergence results for the solutions with initial conditions in the invariant intervals. Then every solution of Eq(9. converges to the equilibrium y.1.2 Convergence of Solutions px + qy r + qx + y By using the monotonic character of the function f (x.4. Assume that q 2 + r ≥ p > q 2 and p + q ≤ r. Then every solution of Eq(9. K]. (b) Assume that q 2 + r ≥ p > q 2 and p + q > r.17) with initial conditions in the invariant interval 0. Then every solution of Eq(9. converges to the equilibrium y.17) with initial conditions in the invariant interval 0. converges to the equilibrium y.17): Theorem 9. Assume that q 2 − qr ≤ p < q 2 and p + q ≤ r. Then every solution of Eq(9.8. Assume that p = q 2 and q 2 + q ≤ r.17) with initial conditions in the invariant interval 0. 2 q2 pr −p qr p − q2 q2 + q − r q+1 .5-1. The proof is an immediate consequence of Lemma 9. 3)-Type Equations 9. for any K > 0.8). Then every solution of Eq(9.4. for any K > 0.162 Chapter 9. y) = in each of the intervals in Lemma 9. K]. (c) Assume that q 2 − qr ≥ p < q 2 and p + q > r. for any K > 0. Then every solution of Eq(9. converges to the equilibrium y. Proof. For example.17) with initial conditions in the interval [0.4. converges to the equilibrium y. (2. the following results are true for Eq(9.17) with initial conditions in the interval [0.17) with initial conditions in the interval [0. converges to the equilibrium y.4.4 (a) Assume that p = q 2 and q 2 + q > r.4. K].4.1 and Theorem 1. together with the appropriate convergence theorem (from among Theorems 1.4.

4.4. Show that every positive solution of Eq(9.3) is bounded.1) has a ﬁnite limit.1 and Theorem 1.4.5. γ. . ∞).1 and Theorem 1. p−q 2 Proof. ∞). Then every solution of Eq(9. B.11) with initial conditions in the invariant interval converges to the equilibrium y.5. γ.5.2 Assume α.1 Assume Show that every positive solution of Eq(9.5 Open Problems and Conjectures α.6. (iii) q < 1 qr . Conjecture 9. C ∈ (0. β. (ii) q ≥ 1. A. ∞). 1−q 163 and p ≤ Then every solution of Eq(9. Conjecture 9. Open Problems and Conjectures Theorem 9. Conjecture 9. A. The proof is a consequence of Lemma 9. K1 p−q 2 3q 2 +q−qr+r . B.20) is not satisﬁed.4.4.5.4.2) is bounded. C ∈ (0. Show that every positive solution of Eq(9. Proof.9. A. 2 Theorem 9. K2 converges to the equilibrium y. The proof is an immediate consequence of Lemma 9. 2 9.5 Assume that p > q2 + r and that one of the following conditions is also satisﬁed: (i) p ≤ q + r.3 Assume β.5. C ∈ (0.11) with initial conditions in the invariant interval qr .6 Assume that q 2 > p + qr and that Condition (9. B.

(2. ∞) and suppose that Eq(9.5 Assume that x−1 . A.5. x0 ) ∈ R × R through which the equation xn + xn−1 xn+1 = 1 + xn + xn−1 is well deﬁned for all n ≥ 0.2 Determine the set G of all initial points (x−1 . n=−1 Open Problem 9.1 Determine the set G of all initial points (x−1 . and for these initial points investigate the long-term behavior of the solutions {xn }∞ . Extend and generalize. B.5. Show that the positive equilibrium of Eq(9.5. γ. Suppose that Eq(9. Conjecture 9. Show that the positive equilibrium of Eq(9. B. ∞). Extend and generalize.5. γ. x0 ∈ [0. A. 3)-Type Equations α. C ∈ (0. Open Problem 9. x0 ) ∈ R × R through which the equation 1 + xn xn+1 = 1 + xn + xn−1 is well deﬁned for all n ≥ 0. x0 ) ∈ R × R through which the equation 1 + xn−1 xn+1 = 1 + xn + xn−1 is well deﬁned for all n ≥ 0. n=−1 Open Problem 9. C ∈ (0.5.3) is globally asymptotically stable.2) has no prime period-two solutions.2) is globally asymptotically stable.4 Assume Chapter 9. and for these initial points investigate the long-term behavior of the solutions {xn }∞ .3) has no prime period-two solutions.3 Determine the set G of all initial points (x−1 .164 Conjecture 9. ∞) with x−1 + x0 > 0 and that β. and for these initial points investigate the long-term behavior of the solutions {xn }∞ . Extend and generalize. n=−1 .

. .5. ψ. φ. φ. .23) pn yn + yn−1 . . .21)-(9. Show that the period-two solution . of Eq(9. . Open Problem 9. Open Problems and Conjectures 165 Open Problem 9. .6 Assume that (9. . Investigate the global character of all positive solutions of Eqs(9. {qn }∞ . .5. Investigate the basin of attraction of the two cycle .4 Assume that {pn }∞ . 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 . .11). n = 0.23). n = 0. Investigate the basin of attraction of the two cycle . n→∞ q = n→∞ qn .5. φ.5.5. φ.17) is locally asymptotically stable.12) holds. 1 + yn + rn yn−1 pn + qn yn−1 . . . (9.17). .20) holds. .6 Assume that (9. ψ. .9. and {rn }∞ are period-two sen=0 n=0 quences of nonnegative real numbers. 1. ψ. lim and r = n→∞ rn .21) (9. and {rn }∞ are convergent sen=0 n=0 n=0 quences of nonnegative real numbers with ﬁnite limits. . . φ.22) (9.20) holds. .5. . 1. p = lim pn . Extend and generalize. rn + qn yn + yn−1 ∞ Open Problem 9. Open Problem 9. .7 Assume that (9. of Eq(9. .5 Assume that {pn }∞ . {qn }n=0 . . . . Conjecture 9. ψ. . 1 + yn + rn yn−1 n = 0. 1. . . ψ. φ. . of Eq(9. ψ. .

1 + yn + ryn−k yn+1 = yn+1 = p + qyn−k . . . r ∈ [0. (2.}.8 Assume that p. . . Investigate the global behavior of all positive solutions of each of the following diﬀerence equations: p + qyn (9. ∞) and Chapter 9. . . 3. 3)-Type Equations k ∈ {2. 1.25) (9.166 Open Problem 9. . r + qyn + yn−k n = 0. 1. n = 0. . n = 0.26) . . . 1. (9.24) yn+1 = . . 1 + yn + ryn−k pyn + yn−k . .5. q. .

A2 q= β .2 The Case C = 0 : xn+1 = A yn B α+βxn +γxn−1 A+Bxn This is the (3.1) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + qyn + ryn−1 .1) (10. A + Cxn−1 Eq(1) contains the following three equations of the (3.3) Bxn + Cxn−1 Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. the initial conditions are nonnegative. 1.1 Introduction α + βxn + γxn−1 . . . . A (10. 1 + yn αB . 1. 2)-Type Equations 10. 2)-type: xn+1 = xn+1 = and α + βxn + γxn−1 . . γ . A n = 0. . xn+1 = n = 0. . . . .2) 10. n = 0. (10. . . n = 0. . (10. 1. A + Bxn α + βxn + γxn−1 . .Chapter 10 (3. 2)-type Eq(10. and the denominators are always positive. 1.4) and r = 167 .

(b) Assume that r < q + 1.2.) The linearized equation of Eq(10. Then every solution of Eq(10. in Section 2. Theorem 10.168 Chapter 10.7) (10.7.5) . for a more general equation.1 we see that y is locally asymptotically stable when r <1+q and is an unstable (saddle point) equilibrium when r > 1 + q. Then the equilibrium of Eq(10. When r =q+1 Eq(10.4) was investigated in [29]. . The proofs of parts (b) and (c) of Theorem 10.6) (10. By applying the linearized stability Theorem 1. (3. Then Eq(10.2.4) is globally asymptotically stable. 2)-Type Equations (Eq(10. .2. The proof of part (a) of Theorem 10.1 was given. See Sections 2.2.4) possesses unbounded solutions. . 2 (1 + y) 1+y n = 0.4) 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. 1.7.2 below. .4) possesses prime period-two solutions.1.1 and 10. (10.1 are given in Sections 10.2.5 and 2.4) exhibit the following trichotomy character. The main result in this section is that the solutions of Eq(10.1 (a) Assume that r = q + 1.4) converges to a period-two solution. (c) Assume that r > q + 1.

.9) for n ≥ 1.10) . We know already that when Eq(10.8) which is of the form of the (2.2.4) satisfy the following identity yn+1 − q = Note that when q= that is when p + qr − q = 0 the Identity (10. 1 + yn and so when p≥q we have yn ≥ q Therefore the change of variables yn = q + zn transforms Eq(10. So in the remaining part of the proof we assume that p < q.2. 1. n = 0.8) holds. . (10.6) holds and we prove that the equilibrium of Eq(10.1 Proof of Part (b) of Theorem 10. (1 + q) + zn n = 1. The Case C = 0 : xn+1 = α+βxn +γxn−1 A+Bxn 169 10.5 and therefore the proof is complete when (10.4) is globally asymptotically stable. 1 + yn p <1 q (10.10) reduces to yn+1 − q = Also in this case 0<r =1− r (yn − q) for n ≥ 0. (10.4) to the diﬀerence equation zn+1 = (p + rq − q) + rzn−1 . 2)-type Eq(6. .12) (10. Observe that (p − q) + ryn−1 yn+1 = q + . .10.23) which was investigated in Section 6. Now observe that the solutions of Eq(10. (10. .4) is attracted to the equilibrium. .1 Here we assume that Eq(10.6) holds the equilibrium is locally asymptotically stable so it remains to be shown that every solution of Eq(10.2.11) q−p r yn−1 − 1 + yn r q−p r for n ≥ 0.

9).12) that Chapter 10. We will assume that (10.16) (10. 2)-Type Equations |yn+1 − q| < r|yn+1 − q| and so n→∞ lim yn = q which completes the proof when (10. Then we can see from (10. note that now q−p <q r and employ (10.13) (10. (10.15) yn ≥ q and then use the known result for Eq(10.10).17) holds is similar and will be omitted. To complete the proof in this case it is suﬃcient to show that eventually (10.10) that yN +1+2K > q for all K ≥ 0.10) that y2n+2 − q = and so y2n+2 > ry2n + p for n ≥ 0. If (10.11) holds.16) holds. It follows from (10.14) First assume that (10. To this end.170 and the equilibrium of Eq(10. The case where (10.17) q−p r for some N ≥ 0.15) is not eventually true then either y2n < or y2n−1 < This is because when yN −1 > then it follows from (10. (10.13) holds.4) is equal to q.18) q−p q−p r y2n − > r y2n − 1 + yn+1 r r . Still to be considered are cases where p + qr − q > 0 and p + qr − q < 0. (10. (3. q−p r q−p r for n ≥ 0 for n ≥ 0.

20) holds is similar and will be omitted. . + rp + p r and so p q−p ≥ r 1−r which contradicts (10. Next assume that (10.1. r that either q−p y2n > for n ≥ 0 (10. r 1−r This contradicts (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. when r < 1.19) r or q−p for n ≥ 0.18) that q−p > y2n+2 > rn+1 y0 + rn p + .14) and completes the proof of part (b) of Theorem 10. q−p .4) lies eventually in the interval 0.16) the solution is bounded.14) holds. Note that in this case r < 1 and so q−p < y2n+2 < rn+1 y0 + rn p + .16) and (10.19) holds. On the other hand.10. . Then from (10. q−p r q−p < r y2n − y2n − 1 + y2n+1 r r .4. Now we claim that every solution of Eq(10. the proof that the r equilibrium is a global attractor of all solutions would be a consequence of Theorem 1. it follows from (10. y2n−1 > (10. .13) holds.2. + rp + p r which implies that p q−p ≤ . Then one can see from (10. q−p .20) r We will assume that (10.2. y) = p + qx + ry 1+x is increasing in both arguments. Once we establish this result. The case where (10. The Case C = 0 : xn+1 = α+βxn +γxn−1 A+Bxn 171 This is impossible when r ≥ 1.13) and completes the proof when (10.10) we see that y2n+2 − q = and so y2n+2 < ry2n + p for n ≥ 0.10) and r the fact that now q−p q< . because by (10.

A n = 0.7) holds.7) holds and show that Eq(10. γ .1 Here we assume that (10.4) with initial conditions such that q ≤ y−1 < r − 1 and y0 > r−1 + p + q(r − 1) r−1−q for n ≥ 0.2) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + qyn + ryn−1 .2. 2)-Type Equations 10.4) to Eq(10. in addition to unbounded solutions.172 Chapter 10. ∞). Eq(10. 10.4) possesses unbounded solutions. A (10. 1.9) and the conclusion of part (c) follows from Section 6. More speciﬁcally the solutions of Eq(10. the equilibrium of Eq(10. 1 + yn−1 q= β . have the property that the subsequences of even terms converge to ∞ while the subsequences of odd terms converge to q. . .4) is a saddle point and so by the stable manifold theorem.5 for appropriate initial conditions y−1 and y0 . To this end observe that yn+1 = q + q p + r(yn−1 − r ) 1 + yn for n ≥ 0 and so when (10. ∞) is invariant. A2 and r = .2 Proof of Part (c) of Theorem 10.4) possesses bounded solutions which in fact converge to the equilibrium of Eq(10.7) holds the interval [q.4). That is when y−1 . y0 ∈ [q. (3.21) αC . 2)-type Eq(10. then yn ≥ q Now the change of variables yn = q + zn transforms Eq(10.2.3 The Case B = 0 : xn+1 = A yn C α+βxn +γxn−1 A+Cxn−1 This is the (3. . It should be mentioned that when (10.

1. q. Observe that (p − r) + qyn . .21). p 1−q when q < 1 and p ≥ r.21) when p ≥ r.1 Invariant Intervals The following result. . which can be established by direct calculation. The Case B = 0 : xn+1 = α+βxn +γxn−1 A+Cxn−1 173 This equation has not been investigated yet. if whenever.3. and r. Here we make some simple observations about linearized stability.1 we see that y is locally asymptotically stable for all values of the parameters p. 10. The linearized equation of Eq(10. By applying Theorem 1.1 Eq(10. invariant intervals. r−p q when q < 1 and p ≥ r(1 − q).3.1. . . . and convergence of solutions in certain regions of initial conditions.21) about its positive equilibrium y= is zn+1 − q+r−1+ (q + r − 1)2 + 4p 2 q p − r + qy zn = 0.21) possesses the following invariant intervals: (a) 0. yN .3. for some integer N ≥ 0. 1. zn + 1+y (1 + y)2 n = 0.10. . (b) 0.3. Lemma 10.2 Here we discuss the behavior of the solutions of Eq(10. gives a list of invariant intervals for Eq(10. 10. yn+1 = r + 1 + yn−1 and so when p≥r Global Stability When p ≥ r . Recall that I is an invariant interval. n = 0. . yN +1 ∈ I then yn ∈ I for n ≥ N.

2)-type Eq(6. .3.2) which was investigated in Section 6.174 yn ≥ r Therefore the change of variables yn = r + zn transforms Eq(10. n = 1.21) with initial conditions in the invariant interval 0.3 Convergence of Solutions Here we discuss the behavior of the solutions of Eq(10. 2.1 Assume p ≥ r.21) to the diﬀerence equation zn+1 = p + qr − r + qzn .3. (ii) q≥1 and either p + qr − r ≤ q or q < p + qr − r ≤ 2(q + 1). Then the following statements are true.21) when p < r. (a) Every positive solution of Eq(10. Theorem 10. 1−q Then every solution of Eq(10. which is of the form of the (2.21) is globally asymptotically stable if one of the following conditions holds: (i) q < 1. converges to the equilibrium y.3.2 (a) Assume that q < 1 and r ≥ p . (3. (1 + r) + zn−1 Chapter 10. r−p q . 2)-Type Equations for n ≥ 1. (b) The positive equilibrium of Eq(10.21) is bounded. 10.3. The following result is now a consequence of the above observations and the results in Section 6. Theorem 10. . .3.

K q pq + r2 − pr .K q and that in the interval . 2 10. (a) In view of Lemma 10.1 we see that when y−1 . γ and r = .21) with initial conditions in the invariant interval r−p . C n = 0. . qyn + yn−1 q= B . . 1. 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 .K q the function f (x. Proof.1 we see that when y−1 . γ2 (10. y) increases in x and decreases in y. r−p .3.22) β . then yn ∈ 0. y0 ∈ 0.K q r−p .3) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= r + pyn + yn−1 . The result is now a consequence of Theorem 1.4.3.K q converges to the equilibrium y.10. r−p q q the function f (x. αC .4 The Case A = 0 : xn+1 = γ yn C α+βxn +γxn−1 Bxn +Cxn−1 This is the (3. y) increases in both x and y.8. then yn ∈ r−p .4. r−p q q for all n ≥ 0. . The result is now a consequence of Theorem 1. 1−q and p < r < Then every solution of Eq(10. r−p and that in the interval 0. K= q + r − p − q2 (b) In view of Lemma 10. 2)-type Eq(10. y0 ∈ for all n ≥ 0. It is easy to check that y ∈ 0. r−p . It is easy to check that y ∈ r−p .4.5.

22) about its positive equilibrium y= is zn+1 − (p − q)y − qr (p − q)y + r zn + zn−1 = 0. it follows from Section 2. . φ. .23) is strict and Condition (10. The linearized equation of Eq(10. . φ. (10. . Concerning prime period-two solution. (10.1. 2)-Type Equations This equation has not been investigated yet. . as in Section 6. 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. the inequality in (10. invariant intervals. q > 1 and 4r < (1 − p)(q − pq − 3p − 1). 1. q−1 . ψ. . .9. .23) It is interesting to note that the above condition for the local asymptotic stability of y requires the assumption that r > 0. (1 + p) + (1 + p)2 + 4r(1 + q) 2(1 + q) By applying Theorem 1.1 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. .25) Furthermore in this case the prime period-two solution . Here we make some simple observations about linearized stability.5 that a period-two solution exists if and only if p < 1.24) is void.24) 2r q − pq − 1 − 3p >0 (10. and the convergence of solutions in certain regions of initial conditions.176 Chapter 10. (q + 1)(r + (p + 1)y) (q + 1)(r + (p + 1)y) n = 0. . When r = 0. (3. The second condition reduces to q > pq + 1 + 3p. ψ.

q2 when p > q and when p < q < p + r.1 Invariant Intervals Here we present some results about invariant intervals for Eq(10. q p−q qr p . q2 p2 − pq > r. and p < q. p > q.4. (a) It is easy to see that when p = q the function f (x. y) is decreasing in both arguments when y < qr is increasing in x and decreasing in y for y ≥ p−q . y) ≤ f (a. a) = ≤ b. .1 Eq(10.26) when p > q and p2 − pq < r. y) = is decreasing in both arguments.1 q−p Proof. Lemma 10. p−q and (10. We consider the cases where p = q. r + px + y qx + y (b) Clearly the function f (x.4. p−q q (c) 1. (b) p qr . Hence a≤ r + (p + 1)b r + (p + 1)a = f (b. (q + 1)b (q + 1)a qr .4. First assume that p > q and p2 −pq q2 and it < r. when q > p + r.21). r q−p r . The Case A = 0 : xn+1 = α+βxn +γxn−1 Bxn +Cxn−1 177 10. b] when p = q a and b are positive numbers such that r + (p + 1)a ≤ (q + 1)ab ≤ r + (p + 1)b.21) possesses the following invariant intervals: (a) [a.10. b) ≤ f (x.

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

Then every solution of Eq(10.1 and Theorem 1. (c) Assume that p < q < p+r.22).4. (b) Assume that p > q and r > p2 −pq .22) with initial conditions in the invariant interval converges to the equilibrium y. 2 Theorem 10.4. q2 and that either p≤1 or p > 1 and (q − 1)[(p − 1)2 (q − 1) − 4q(1 − p − qr)] ≤ 0.22) with initial p qr . Proof. q−p converges to the equilibrium y.25) is not satisﬁed.1 (a) Assume that p = q. b]. qr p . Then every solution of Eq(10. p−q q Proof.4.3 Assume that p < q.10.22) with initial conditions in the invariant interval r .4.4. 2 Theorem 10. and r <1 q−p 2 10. The proof is an immediate consequence of Lemma 10. p + r < q. 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.1 and Theorem 1. q p−q conditions in the invariant interval converges to the equilibrium y. q−p Proof. and that Condition (10. Then every solution of Eq(10. Theorem 10. The proof is an immediate consequence of Lemma 10. Then every solution of Eq(10.6.22) with initial conditions in the invariant interval [a. The proof is an immediate consequence of Lemma 10.22) with initial conditions r in the invariant interval 1.26) converges to the equilibrium y.2 Convergence of Solutions Here we obtain some convergence results for Eq(10.4.4.4.5. q2 Then every solution of Eq(10.4.1 and Theorem 1. where 0 < a < b satisfy (10.7. r < p2 −pq .4. 1 converges to the equilibrium y.2 Assume that p > q. 2 .4.

if and only if r = 1 + q. ψ. (10. Open Problem 10. ∞ (b) Let (y−1 . ∞) × (0. φ. (a) Find the set B of all initial conditions (y−1 . (10. ψ.27).27) (b) Assume (10. . Determine the set of initial conditions y−1 and y0 for which {yn }∞ n=−1 converges to (10. . y0 ) ∈ B.5. . Investigate the asymptotic behavior of {yn }n=−1 .1 Every positive solution of Eq(10.4).5.5. (a) Determine the set of initial conditions y−1 and y0 for which φ = ψ. y0 ) ∈ (0.29) 1 + xn−1 1 + xn + xn−1 .4) are bounded. (3.5. determine the set G of all initial conditions (x−1 . 2)-Type Equations 10. ∞) such that the ∞ solutions {yn }n=−1 of Eq(10.28) 1 + xn 1 + xn + xn−1 xn+1 = (10.3 For each of the equations listed below.2 Assume r > q + 1.5 Open Problems and Conjectures ∞ Open Problem 10.1 We know that every solution {yn }n=−1 of Eq(10. .180 Chapter 10. Open Problem 10. x0 ) ∈ R × R through which the equation is well deﬁned for all n ≥ 0. φ. .27) is a prime period-two solution of Eq(10.21) converges to the positive equilibrium of the equation.30) xn+1 = xn + xn−1 Extend and generalize. . Conjecture 10. and for these initial points investigate the long-term behavior of the corresponding solution: 1 + xn + xn−1 xn+1 = (10. .4) converges to a not necessarily prime period-two solution .

and {rn }n=0 are convergent sen=0 n=0 quences of nonnegative real numbers with ﬁnite limits. .33). r ∈ (0. . .22) (a) Local stability of the positive equilibrium implies global stability. (10. {qn }∞ . Investigate the global character of all positive solutions of Eqs(10.5. n→∞ q = n→∞ qn .5.31)-(10.5. q. {qn }∞ . is locally asymptotically stable. . 1. 1. 1 + yn pn + qn yn + rn yn−1 .32) (10.5.2 Assume that p. the equilibrium y is a saddle point. qn yn + yn−1 n = 0. 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 . (10.4 Assume that {pn }∞ . Open Problems and Conjectures Conjecture 10. .33) Open Problem 10. . .10. 1 + yn−1 rn + pn yn + yn−1 .22) possesses a period-two solution. when it exists. lim lim and r = n→∞ rn .22). ∞). Then the following statements are true for Eq(10. .31) yn+1 = yn+1 = n = 0. p = lim pn . and {rn }∞ are period-two sen=0 n=0 n=0 quences of nonnegative real numbers. (c) When Eq(10. . .22) has no prime period-two solutions the equilibrium y is globally asymptotically stable. n = 0. 181 (b) When Eq(10. 1. (d) The period-two solution of Eq(10.5 Assume that {pn }∞ . ∞ Open Problem 10. Extend and generalize. but not globally. .

6 Assume that Chapter 10.35) (10. . . . . 2)-Type Equations p. . n = 0.182 Open Problem 10. 3. . (3. (10. 1 + yn p + qyn + ryn−k . . qyn + yn−k n = 0. 1. .5. . . ∞) and k ∈ {2. . . 1.36) . 1 + yn−k p + qyn + ryn−1 . 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 . q. r ∈ [0. n = 0. 1. .}.34) (10.

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

2 Invariant Intervals Here we obtain several invariant intervals for Eq(11.14). . B > C. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 The equilibrium x is locally asymptotically stable if Conditions (2. (11.4) The local asymptotic stability of the two cycle 11. yN . and α < (γ − β − A)[B(γ − β − A) − C(γ + 3β − A)] . and (2. 2 (A + Bx + Cy) (A + Bx + Cy)2 Recall that an interval I is an invariant interval for Eq(11.4. .1) by testing whether the hypotheses of one of the four Theorems 1.13). φ. that is.6. ψ. if. C C(B − C) . . . Once we have an invariant interval for Eq(11. ψ. is discussed in Section 2. ψ.1) has a prime period-two solution . y) = is increasing or decreasing in x and y. .15) are satisﬁed.1) if whenever. .3) 4C 2 Furthermore. . If we set L = βA − αB. .5-1. (2.1) then we may be able to obtain a convergence result for the solutions of Eq(11. . φ.184 Chapter 11.31) and (2. when (11. see Section 2. These intervals are derived by analyzing the intervals where the function f (x. Conditions (2. yN +1 ∈ I then for some integer N ≥ 0.3). γ > β+A. 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+ = 0.4. .8 are satisﬁed. and N = γA − αC α + βx + γy A + Bx + Cy L + My N − My and fy = . for n ≥ N. φ. then clearly fx = M = βC − γB. Eq(11. φ.5.32) are satisﬁed.1). (11. . yn ∈ I . ψ. The (3. .

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

0) ≤ f (x. B+C B γ β α (ii) The condition B > C > A is equivalent to M > 0. in this case the function f is increasing in x. L > 0. u) ≤ β β+γ = . B A C γ Then 0. y ∈ α γ . A M M C for all x. A C . (x) Assume that β α γ < ≤ . A B C γ Then 0. for every x.1) and in this interval the function f (x.186 (viii) Assume that Chapter 11.1) and in this interval the function f (x. assuming C ≤ M we obtain N N γ α = f (0. and N > 0. B is an invariant interval for Eq(11. this function γ N N is increasing in y for x < M . = f (0. γ β α (i) The condition C = B > A is equivalent to M = 0 and L > 0. which in turn imply that N > 0. B A C β Then 0. u) is increasing and consequently f (u.1) and in this interval the function f (x. C is an invariant interval for Eq(11. Thus. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 β α 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 x and decreasing in y. ) = . y) is increasing in y and decreasing in x. C is an invariant interval for Eq(11. The (3. y) is increasing in y and decreasing in x. Thus. . (ix) Assume that α β γ = < . y) ≤ f ( . Thus. 0) ≤ f (x. y) ≤ f ( . 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 = . Proof. In addition.

for every y. Thus. L < 0. A . u) is decreasing and consequently f (u. y) ≤ f A for all x. we obtain 0 ≤ f (x. B+C .11. y ∈ α β+γ . Thus. B+C B+C ≤ β+γ . Invariant Intervals 187 γ β α (iii) The condition C > B > A is equivalent to M < 0 and L > 0. which imply that N < 0. α L L this function is decreasing in x for y < − M . we obtain 0 ≤ f (x. this function is increasing in x for y < − M . B A A A Here we used the fact that under the condition β α γ = < C B A the function f (u. In addition. In β+γ L L 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 . In addition. this function α N N is decreasing in y for x < M . and N < 0. y) ≤ f (0. in this case the function f is increasing in y. 0) ≤ f (x. A . for every y. α . Thus assuming A ≤ − M . in this case the function f is decreasing in y. α = f (0. α L ≤− . in this case the function f is decreasing in x. which in turn imply that N > 0. Thus assuming A ≤ M . A M γ β α (vi) The condition A > C > B is equivalent to M < 0. Thus. y ∈ 0. for every x. which in turn imply that N < 0. 0) = . y ∈ 0. Thus assuming B+C ≤ − M . − M . 0) = L for all x. 0) = α for all x. γ β α (iv) The condition C = B < A is equivalent to M = 0 and L < 0. y) ≤ f (0.2. we obtain. Thus. A B+C β+γ β+γ . B+C B β γ α (v) The condition A > B > C is equivalent to M > 0 and L < 0. u) ≥ β+γ β = . y) ≤ f (0. ≤ f (x.

L < 0. which in turn imply that N > 0. The (3. y) ≤ f α for all x. v) ≤ = lim f (0. Thus. in this case the function f is increasing in x and decreasing in y for all values of x and y. and N ≥ 0. β β . y) < f 0. y) < f β for all x.188 Chapter 11. C γ β α (x) The condition B < A ≤ C is equivalent to M < 0.0 = . Using this we obtain. y ∈ 0. 0 ≤ f (x. in this case the function f is decreasing in x and increasing in y for all values of x and y. y ∈ 0. and N < 0. C γ for all x. v) is increasing in v when N > 0 and so γ f (0. 0 ≤ f (x. C v→∞ for all v. 0 ≤ f (x. 0 ≤ f (x. Using this we obtain. which in turn imply that N < 0. Using this we obtain. L > 0. Here we use the fact that the function f (0. Thus. C . y ∈ 0. B . γ C = γ . A . B B β γ α (ix) The condition A = B < C is equivalent to M < 0 and L = 0. . Thus. Thus. Using this we obtain. in this case the function f is increasing in x and decreasing in y for all values of x and y. α α . y) ≤ f 0. 2 The next table summarizes more complicated invariant intervals with the signs of the partial derivatives fx and fy .0 = . y ∈ 0. v). C . in this case the function f is increasing in x and decreasing in y for all values of x and y. A A β γ α (viii) The condition B > A > C is equivalent to M > 0. γ for all x. γ C ≤ γ . 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.

. Theorem 11. M ] if AM +(B+C)N < αM +βK+γN N N [ M .4. K2 ] if BM K+AM −γL > − M −CL β+γ−A+ β [0. . 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. fx < 0 if y > − M fy > 0 fx > 0 fy < 0 fx < 0 fy > 0 L L fx > 0 if y > − M .8. − M ] if AM −L(B+C) < − M αM −βL+γM L L [− M .3. K1 ] β−A+ Signs of derivatives N M [0. ((γ − A)M − BN )2 + 4BM (αM + βN ) 11. fy < 0 if x < M αM +(β+γ)N αM +(β+γ)N N [ AM +(B+C)N .3. B ] γ [0.2. 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 . M ] if AM +(B+C)N < N [ M .4. A ] √ L L fx > 0 if y < − M .11.1). [0. − M ] if AM −(B+C)L < − M βM K+αM L L [− M . . C ] αM −L(β+γ) αM −L(β+γ) L L [ AM −L(B+C) .5-1. Convergence Results 189 Case 1 2 3 4 5 6 7 8 9 Invariant intervals αM +(β+γ)N N [0.3 Convergence Results The following result is a consequence of Lemma 11. Then in each of the following cases lim xn = x.1 Let {xn } be a solution of Eq(11. n→∞ . K3 ] if AM −BL+CMK > − M K α [0. fx < 0 if y < − M fy < 0 fx < 0 fy < 0 fx < 0 N N fy > 0 if x > M .1 and Theorems 1. 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 B C β (viii) At least one of the following three conditions is not satisﬁed and x−1 . C B A (iv) γ β α = < C B A (v) and x−1 . βC − γB α . . x0 ∈ 0. α (vii) At least one of the following three conditions is not satisﬁed and x−1 . B+C βC − γB and x−1 . γA − αC . . x0 ∈ α β γ γA − αC ≥ > > βC − γB A B C (vi) and x−1 . . x0 ∈ 0. x0 ∈ α β . x0 ∈ αB − βA β+γ ≤ . and or (C − B)[(C − B)(β − γ − A)2 − 4B(αB + γ(β − γ − A))] > 0. B < C. B : β > γ + A. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 γ β α = > C B A (ii) β γ α > > . A B+C γ γA − αC ≤ . A B α γ . x0 ∈ 0. B A C . B A and x−1 . and or (C − B)[(C − B)(β − γ − A)2 − 4B(αB + γ(β − γ − A))] > 0. . A β γ α γA − αC < < < B C A βC − γB and x−1 . B < C. A C α β+γ . x0 ∈ β α . β γ α = > . A : β > γ + A. B C A (iii) γ β α > > . x0 ∈ 0.190 (i) Chapter 11. C βC − γB and x−1 . The (3. β α γ > ≥ .

1) is globally asymptotically stable.4. The proofs of statements (i-iii) follow by applying Lemma 11.1) does not possess a prime period-two solution.4. m).11.3. The proofs of statements (ix) and (x) follow by applying Lemma 11. γ . A + Bm + CM M= α + βM + γm A + BM + Cm is m = M .8.4. . γ .3.1 Global Stability Theorem 1.2 Assume that γ β α < min .3 to Eq(11. C and x−1 .1 (i-iii) and Theorem 1.2.7.2. α γ β < ≤ B A C and x−1 .3) is not satisﬁed. m= α + βm + γM . Theorem 1.4. and by observing that the only solution of the system of equations m = f (m.4.2. β γ α = < A B C (x) Condition (11. M = f (M.1 (ix) and (x) respectively. C 191 Proof.1) we obtain the following global asymptotic result. .2.5.6.1 (vii) and (viii) respectively. x0 ∈ 0.2 and linearized stability imply the following global stability result: Theorem 11.4.3) is not satisﬁed. that is. Theorem 1. M ). The proofs of statements (vii) and (viii) follow by applying Lemma 11.1 (iv-vi) and Theorem 1. x0 ∈ 0.3. Convergence Results (ix) Condition (11. C B A Then the equilibrium x of Eq(11. and by observing that Eq(11. The proofs of statements (iv-vi) follow by applying Lemma 11. By applying Theorem 1. 2 11.

β. B. Conjecture 11. B. γB > βC. A.4. γ. βA ≥ αB. αB > βA. Theorem 11. Then the equilibrium x of Eq(11. Then the equilibrium x of Eq(11.1) is bounded. β.4 (a) Assume that αC ≥ γA.1) has no positive prime period-two solution.192 Chapter 11.4.4. .1) is globally asymptotically stable. C ∈ (0. ∞).3 Assume that B ≤ C.1) converges to the positive equilibrium.1) is a saddle point and the period-two solution is locally asymptotically stable. A.4. Show that every positive solution of Eq(11. γ.4 Open Problems and Conjectures α. The following global stability result is a consequence of linearized stability and the Trichotomy Theorem 1. and 3γB ≥ βC ≥ γB.1) is globally asymptotically stable. ∞) and that Eq(11. β. γ. C ∈ (0. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Theorem 11. Then the equilibrium x of Eq(11. A.3.1 Assume Show that every positive solution of Eq(11. Conjecture 11. C ∈ (0. and 3βC ≥ γB ≥ βC. Conjecture 11. ∞) and that Eq(11. (b) Assume that αB ≥ βA.1) is globally asymptotically stable.3.3 Assume α. and γA > αC. Show that the positive equilibrium of Eq(11. γA ≥ αC. The (3.4.1) has a positive prime period-two solution. B. 11.2 Assume α.

1) cannot have a prime period-k solution for any k ≥ 3. Obtain necessary and suﬃcient conditions so that every positive solution of Eq(11.3 Assume that a.6) holds. . . Investigate the asymptotic behavior and the periodic nature of all positive solutions of Eq(11. Show that Eq(11.4. β.4 Assume α. (11. γ.5) are period-two sequences of positive real numbers. . {Bn }∞ . {An }n=0 . B.4. {γn }∞ . A + A0 xn + . investigate the cases where 2 ≤ k ≤ 8. 1. ∞) for i = 0.6) Obtain necessary and suﬃcient conditions so that every positive solution of the equation xn+1 = has a ﬁnite limit. . .7) converges to a period-two solution. .7) is periodic with period-k. Open Problems and Conjectures Conjecture 11. {βn }∞ . Investigate the asymptotic behavior and the periodic nature of all positive solutions of the diﬀerence equation αn + βn xn + γn xn−1 .11.5) An + Bn xn + Cn xn−1 Open Problem 11. . xn+1 = (11. . + Ak xn−k (11. {Cn }∞ n=0 n=0 n=0 n=0 are convergent sequences of nonnegative real numbers with ﬁnite limits. ∞). A. a + a0 xn + .4.4. Obtain necessary and suﬃcient conditions so that every positive solution of Eq(11. .7) Open Problem 11.4 Assume that (11.5). ai . . . Ai ∈ [0.1 Assume {αn }n=0 . .5 Assume that (11. 1. 1.4. . . n = 0. C ∈ (0. Open Problem 11.4. 193 ∞ ∞ Open Problem 11.2 Assume that the parameters in Eq(11.4. . A. n = 0. .6) holds. Open Problem 11. In particular. k. . k ≥ 2. + ak xn−k .

. Investigate the boundedness character. xn xn−1 + xn−2 + . (11. γ. n = 0. 1. . 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 . 3}.4. . (See [8].8) converges to a period-two solution. . γxn xn−1 + δxn−2 Conjecture 11.4.) Conjecture 11. ∞) with γ + δ > 0. (See [32] for the two dimensional case. .) αxn + βxn−1 xn−2 .4.7 Assume that p. [47]. n = 0. Extend to equations with real parameters.6 Let α.194 Chapter 11. 2. Show that every positive solution of the diﬀerence equation xn+1 = xn + .) Conjecture 11. 1. . β. + xn−k + xn−k−1 xn−k−2 . + xn−k−2 converges to 1. .6 Let aij ∈ (0. . the periodic nature. and the asymptotic behavior of all positive solutions of the diﬀerence equation xn+1 = (See [8] and [61]. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Open Problem 11. n = 0. . . . The (3. j ∈ {1.4. 1. . q ∈ [0.5 Let k ≥ 2 be a positive integer. δ ∈ [0. Extend and generalize. and [60]. . ∞). . ∞) for i. .

. q + xn−2 n = 0. Conjecture 11. . . converges to a period-two solution if p ≥ 1. 1.4. (See [62]. 195 converges to a period-two solution if and only if p = 1 + q.9 Assume that p. Open Problems and Conjectures (a) Show that every positive solution of the equation xn+1 = pxn−1 + xn−2 .4. .8 Assume that p ∈ (0. ∞). (c) Show that when p>1+q the equation possesses positive unbounded solutions. xn−1 + pxn−2 n = 0. q ∈ [0. . ∞).11. . (b) Show that when p<1+q every positive solution has a ﬁnite limit. 1.4.) Conjecture 11. . (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 .

. (b) Show that when q>1 the positive equilibrium of the equation is globally asymptotically stable.) Conjecture 11.10 Show that every positive solution of the equation xn+1 = xn + xn−2 . (c) Show that when q<1 the equation possesses positive unbounded solutions. 1. (See [62]. has the following trichotomy character: (i) When p > 1 every positive solution converges to the positive equilibrium. . 1. xn−1 n = 0. . .4. The (3. converges to a period-two solution if and only if q = 1. (See [62]. (ii) When p = 1 every positive solution converges to a period-ﬁve solution. 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 .196 Chapter 11.) .4. 1. . .) Conjecture 11. . xn n = 0. xn−2 + q n = 0. (iii) When p < 1 there exist positive unbounded solutions.11 Show that the diﬀerence equation xn+1 = p + xn−2 . converges to a period-four solution. (See [62]. . .

φ φ ψ ψ Open Problem 11. .10). xn−1 + xn−2 n = 0. φ.. What other conditions are needed so that Eq(11. . A + Bxn + Dxn−2 (11. .9) converges to a period-two solution? (b) Assume γ < β + δ + A.4.4.4. ψ ∈ (0. 1.12 Show that every positive solution of the equation xn+1 = 1 + xn . 197 converges to a period-six solution of the form . ∞). (11.11. . Open Problems and Conjectures Conjecture 11. 1.9) with nonnegative parameters and nonnegative initial conditions such that B + D > 0 and A + Bxn + Dxn−2 > 0 for n ≥ 0.9) possesses unbounded solutions? (See [6].9) converges to the positive equilibrium? (c) Assume γ > β + δ + A. .) .. .10) (a) In addition to (11.7 Consider the diﬀerence equation xn+1 = α + βxn + γxn−1 + δxn−2 . .. . . What other conditions are needed so that every positive solution of Eq(11. n = 0. . what other conditions are needed so that every positive solution of Eq(11. (See [62]. .) ψ 1 1 φ . where φ. .9) has prime period-two solutions if and only if γ = β + δ + A. Then one can show that Eq(11. ψ.

(See [6]. 1. r ∈ (0. . . ∞). . (See [6]. Conjecture 11. . . 1.16 Consider the diﬀerence equation xn+1 = where p ∈ [0.13 Show that every positive solution of each of the two equations xn+1 = and xn+1 = 1 + xn−1 + xn−2 . xn−1 + xn−2 + xn−3 n = 0. .15 Show that every positive solution of the equation xn+1 = xn−2 . . (b) Show that when q>r every positive solution converges to the positive equilibrium. n = 0.198 Chapter 11. . . xn 1 + xn−1 + xn−2 . (a) Show that when q=r every positive solution converges to a period-four solution. converges to a period-two solution. The (3. . .4. p + qxn + rxn−2 . xn−1 n = 0.14 Show that every positive solution of the equation xn+1 = 1 + xn−1 . . . 1. 1.4. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Conjecture 11. 1.) Conjecture 11. 1 + xn + xn−2 n = 0. . .) Conjecture 11. converges to a period-three solution.4. . ∞) and q. converges to a period-two solution.4. xn−2 n = 0.

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 . . B1 . B0 .4. xn n = 0.4. 1.) 199 Open Problem 11. . xn−2 + xn−3 .11. 1 + xn + xn−3 n = 0. . .) . xn−1 . . 1. . Open Problems and Conjectures (c) Show that when q<r the equation possesses positive unbounded solutions. Obtain necessary and suﬃcient conditions on the parameters A and B0 . 1. 1. . . . . . Bk so that every positive solution of the diﬀerence equation xn+1 = xn−k . Open Problem 11. (See [62]. . . Bk ∈ [0. 1 + xn−3 .4.9 Assume that k is a positive integer and that k A. 1. xn + xn−2 + xn−3 n = 0. . converges to a period-(k + 1) solution of this equation. . B1 . . . . + Bk xn−k n = 0. . (See [6]. . xn−2 n = 0. ∞) with i=0 Bi > 0. . . . A + B0 xn + . .

xn−k+1 n = 0. 1 + xn + . φ. .12) which corresponds to the nonnegative solution {xn }∞ n=−k of Eq(11.200 Chapter 11. . . 0. (11.12) with a given φ ≥ 0 ? (See [6]. .17 Let k be a positive integer. .) Conjecture 11. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Open Problem 11. Then we can easily see that every nonnegative solution of the equation xn+1 = xn−k .11) possess a positive solution which converges to zero? (b) Determine. .10 Assume that k is a positive integer.11). The (3. . 0. 0. . x0 for which the ∞ corresponding solution {xn }n=−k of Eq(11. . . . (c) Determine the set of all nonnegative initial conditions x−k . .4. Show that every positive solution of xn+1 = 1 + xn + xn−k . .11) converges to a period-(k + 1) solution of the form . . k−terms (11.. 1. x0 . in terms of the nonnegative initial conditions x−k . + xn−k+1 n = 0. . .4. .11) converges to a period-(k+1) solution of the form (11. . . .. the value of φ in (11. . . converges to a periodic solution of period 2k. (a) Does Eq(11.12) with φ ≥ 0. 1.. . .

y.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. As we have seen in previous chapters. y. b] be an interval of real numbers and assume that f : [a. (A.1) has a unique equilibrium x ∈ [a. xn−1 . b]3 → [a. x) = x. x. y. x−2 ∈ I. or non increasing in one and non decreasing in the other two.4. . . Then for every set of initial conditions x0 . Let I be some interval of real numbers and let f :I ×I ×I →I be a continuously diﬀerentiable function. x−1 . n = 0.1) converges to x. 1. Then Eq(A. z) is monotonic in its arguments. 201 . Theorem A.1) is to work in the regions where the function f (x. . z) is non decreasing in all three arguments. xn−2 ). or non decreasing in all three.8.1 Let [a.1) ∞ has a unique solution {xn }n=−2 . has a unique solution in the interval [a. very often a good strategy for obtaining global attractivity results for Eq(A.5-1. (b) The equation f (x.0. b]. We state the results for third order equations from which higher order extensions will become clear. or non increasing in two arguments and non decreasing in the third.4. b] and every solution of Eq(A. b] is a continuous function satisfying the following properties: (a) f (x. In this regard there are eight possible scenarios depending on whether f (x. the diﬀerence equation xn+1 = f (xn . z) is non-decreasing in each of its arguments.

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

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

M. . m = f (m. . mi−1 . M1 ≤ M0 = b. b] be an interval of real numbers and assume that f : [a. (b) If (m. and mi ≤ xk ≤ Mi Set m = lim mi i→∞ Appendix for k ≥ 3i + 1. b] is a continuous function satisfying the following properties: (a) f (x. .4 Let [a.204 Proof. y. . b] × [a. ≤ mi ≤ . Mi ≤ . Theorem A. b] and every solution of Eq(A. M.1) converges to x. Mi−1 . . Then (A.2) holds and by the continuity of f . . Mi−1 ) and mi = f (mi−1 . mi−1 ) . m. m) . b]3 → [a. m. 2 and M = f (M. . i→∞ and M = lim Mi . b] and is non-decreasing in y and z for each x. .1) has a unique equilibrium x ∈ [a. set Mi = f (Mi−1 . Now observe that for each i ≥ 0 a = m0 ≤ m1 ≤ . M ) ∈ [a. y ∈ [a. b] is a solution of the system M = f (m. . b]. . Then Eq(A. M ) Therefore in view of (b) m=M from which the results follows.0. 2. M ) and m = f (M. Set m0 = a and M0 = b and for each i = 1. z) is non-increasing in x for each y and z ∈ [a. m) then m = M.

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

Then (A. . 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. ≤ mi ≤ . Set m0 = a and M0 = b and for each i = 1. M = f (M. m. . set Mi = f (Mi−1 . . Then Eq(A. b]3 → [a. Mi ≤ .206 Proof. b]. . m) and m = f (M. 2. Mi−1 ). Therefore in view of (b) m=M from which the results follows. b] is a continuous function satisfying the following properties: (a) f (x. M ) ∈ [a. Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . . y. mi−1 ) and mi = f (mi−1 . M. (b) If (m. .6 Let [a. . . m) and m = f (m. m.2) holds and by the continuity of f . . M ) then m = M. M ).0.1) has a unique equilibrium x ∈ [a. mi−1 . b] be an interval of real numbers and assume that f : [a. b] and every solution of Eq(A. M. b] is a solution of the system M = f (m. M1 ≤ M0 = b and mi ≤ xk ≤ Mi Set m = lim mi i→∞ Appendix for k ≥ 3i + 1. i→∞ and M = lim Mi .1) converges to x. b] × [a. 2 Theorem A. . Mi−1 .

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

. b]3 → [a. m. . 2 Theorem A.208 Proof. b]. m. (b) If (m.0. y. mi−1 . b] is a solution of the system M = f (m. . y. M ) and m = f (M. Therefore in view of (b) m=M from which the results follows. Then (A. M1 ≤ M0 = b and mi ≤ xk ≤ Mi Set m = lim mi i→∞ Appendix for k ≥ 3i + 1. . Set m0 = a and M0 = b and for each i = 1. .2) holds and by the continuity of f . . . . Then Eq(A. Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . M = f (m. Mi−1 ) and mi = f (Mi−1 . . z) is non-increasing in all three variables x. m). b] be an interval of real numbers and assume that f : [a. Mi−1 . mi−1 ). M ) ∈ [a. b] × [a.1) has a unique equilibrium x ∈ [a. b] is a continuous function satisfying the following properties: (a) f (x.1) converges to x. m) and m = f (M. set Mi = f (mi−1 . 2.8 Let [a. M. M. b] and every solution of Eq(A. ≤ mi ≤ . . M ) then m = M. i→∞ and M = lim Mi . z in [a. Mi ≤ .

m) and m = f (M. Theorem A. M ). . ≤ mi ≤ . Set m0 = a and M0 = b and for each i = 1. . set Mi = f (mi−1 . . . . 1. b]p → [a. mi−1 . up ) is non decreasing in the second and the third variables u2 and u3 and is non increasing in all other variables.9 Let [a. . .0. . (A. Then (A. 2.8.3) 2 The above theorems can be easily extended to the general diﬀerence equation of order . . b] is a continuous function satisfying the following properties: (a) f (u1 . M1 ≤ M0 = b and m i ≤ x k ≤ Mi Set m = lim mi i→∞ 209 for k ≥ 3i + 1. . xn−1 . This gives 2p results of the type found in Theorems A. . Mi−1 .0. there are 2p possibilities for the function f to be non decreasing in some arguments and non increasing in the remaining arguments. In this case. f : Ip → I n = 0. Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . i→∞ and M = lim Mi . M = f (m. p xn+1 = f (xn .2) holds and by the continuity of f . Mi ≤ . . . . . M.0. Let us formulate one of these results. Mi−1 ).1-A.Appendix Proof. . xn−p+1 ). mi−1 ) and mi = f (Mi−1 . . b] be an interval of real numbers and assume that f : [a. u2 . . . Therefore in view of (b) m=M from which the results follows. m. . where is a continuous function.

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

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

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

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

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

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

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

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

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

- Cntrl Systems
- 427K M2 Solutions
- ex8s
- Programa
- very important file regarding vibration analysis
- 1401.7251
- Answers to Problem Set # 7
- 01318710
- 10.1.1.85
- Algebraic Methods to Compute Mathieu Functions - Frenkel
- M408L Exam 3
- JAE_4
- ODE Stability
- Makoto Umeki- A Locally Induced Homoclinic Motion of the Vortex Filament
- Nicholas J.L. Brown, Alan D. Sokal, Harris L. Friedman - The Complex Dynamics of Wishful Thinking. the Critical Positivity Ratio (ArXiv, July 2013, 29th)
- Instructions for Formatting and Editing NCSLI Measure Submissions
- Thomas Fermi Model 1
- Decomposition of Poisson Problems
- Integrating Factors and Symmetries
- Riccati Equations
- Adamson, Gregory Dale - Science & Philosophy Two Sides of the Absolute
- pde2010_a2
- i73
- [ QUA] Quantum Field Theory for Mathematicians_ Hamiltonian Mechanics & Symplectic Geometry
- Algebra
- h82_Cartesian_Equation_of_a_Line.pdf
- ANALYTICAL SOLUTIONS OF THE MODIFIED COULOMB POTENTIAL USING THE FACTORIZATION METHOD.
- Chapter 5
- Handout_2_2011
- Lecture Notes for Section 17-5

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.

scribd