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

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.

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

.1 Introduction . . . . . . . . . . . . .1 Deﬁnitions of Stability and Linearized Stability Analysis 1. . . . . . . . . . .1 Equilibrium Points . . . . . . . . . . . . .6 Local Asymptotic Stability of a Two Cycle . . . . . . . 1. . . βxn 3. . . . . . . . v . . . . . . . 2. Periodicity. . . .2 The Stable Manifold Theorem in the Plane . . . . . . . . . . . . . . . . . . 2. .5 Existence of Prime Period-Two Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . 2 Local Stability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . and Invariant Intervals 2. 2. . . . . . . . . . . .3 The Case α = β = A = C = 0 : xn+1 = γxn−1 . . . . . . . . . . . . . 1.4 Global Attractivity of the Positive Equilibrium . . .7 Convergence to Period-Two Solutions When C=0 . . . Bxn 3.4 When is Every Solution Periodic with the Same Period? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. . . .8 Invariant Intervals . . . . . . . . . . . . . . . . . . .7 Semicycle Analysis . . . . . . . . . . . . . . . . .6 The Riccati Equation . . . . . . . . . .2 The Case α = γ = A = B = 0 : xn+1 = Cxn−1 3. . . . . . . . . . . 2. . 1)-Type Equations 3. . . . . . 1. . . . . . . . . . . . . . .9 Open Problems and Conjectures . . . . . . . . . . . . . . . . 2. . . . . . . . . . 1. . . . . . . . . . . . . .Contents Preface Acknowledgements Introduction and Classiﬁcation of Equation Types 1 Preliminary Results 1. . . .2 Stability of the Zero Equilibrium . . . . . . . .5 Limiting Solutions . . . . . . . . . . . . .3 Global Asymptotic Stability of the Zero Equilibrium . . . 2. . 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 (1. . . . . . Semicycles. 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 . . . . . . . . . . . . 2. .3 Local Stability of the Positive Equilibrium . . .

. . . 4. . . . . . .6. . . . . . .3 Global Stability and Boundedness . . . .7. . . . . . . .2 The Case q < 1 . . . . . . . . . . . . . . . . . . . . . . . . . . .vi Contents 4 (1. . . 5. . . . . . . . . . . . . 4. . . . . . . .7. . α+γxn−1 6. . .2 The Case p = 1 . . . . . . . . .2 The Case γ = A = B = 0 : xn+1 = α+βxn . . . .2 The Case β = γ = C = 0 : xn+1 = A+Bxn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 The Case α = β = C = 0 : xn+1 = 4. . .2 The Case γ = C = 0 : xn+1 = A+Bxn . . . . . . . . . .1 Invariant Intervals . . . .Riccati Equation .1 Semicycle Analysis . . . . . . . . . . . . . . . . . . . . . . . . . n +γx 6. . . . . . . . . . . . . . . .3 The Case q > 1 . . . . . . . . . 53 α 4. . . . 6. 6. . . . . . . . . . . 2)-Type Equations 77 6. . . . 5 (2. . . . . . . . . .3 The Case β = A = C = 0 : xn+1 = α+γxn . . . . . . . . . .2 Invariant Intervals . . . . .2 Semicycle Analysis . . . . . . . . . .5 The Case β = C = 0 : xn+1 = α+γxn−1 . . . 6. . 79 6. . . . . . . . . . . . . . .3 The Case p > 1 . . . . . . . . . . .6. . 57 58 59 60 61 62 62 64 69 69 70 72 72 . .7. . . . . . . . . . . . . . . . . .3 The Case γ = B = 0 : xn+1 = A+Cxn−1 . . . . . . 6. . . 54 α 4. . . . . . . . . . . .3 Semicycle Analysis . . . . . . . .4. . . . . . .1 Existence and Local Stability of Period-Two Cycles . . . . . . . . . . . . . . . . . . . . . . . .4. . . . The Case γ = A = 0 : xn+1 = Bxα+βxn n +Cxn−1 6. . . . .4 . . . . . . . . . . . .4 Open Problems and Conjectures . . . . . . .8 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7. . . . . . . . . . .7. . . . .1 Introduction . . . . . . . . . . . . . . . .2 Semicycle Analysis of Solutions When q ≤ 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 n α+βxn 6. . . βxn A+Cxn−1 βxn Bxn +Cxn−1 γxn−1 . . . . . . . . . . . . . . . .7 The Case α = C = 0 : xn+1 = βxA+Bxn−1 . . . . . . . . .4 Global Attractivity When q < 1 . . . . . . . . . . . . . . A+Bxn 6. . . 6. . . . .6 The Case β = A = 0 : xn+1 = Bxn +Cxn−1 . . . . . . . . . . . . 4. . . . .1 Invariant Intervals . . . . . . 6 (2. . . . .7 The Case α = β = A = 0 : xn+1 = 4. . . 6. . . . . . . . . . . . 6. . . Bx 5. . . . . 2)-Type Equations 53 4. . 6. . . . . . . . . . .5 The Case α = γ = A = 0 : xn+1 = 4. . . . . . . . . . . . . . . . .6. . . .5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . n 6. 55 4. . . . . . . . . . . . . . . . . . . A+Bxn γxn−1 Bxn +Cxn−1 Pielou’s Equation . . . . . . . . . . . . . . . . . . .Lyness’ Equation Cxn−1 n−1 5. . . . . . . 6. . . . . . 77 α+βx 6.1 Introduction . . . . . . . . . 6. . . . . . .3 The Case β = γ = A = 0 : xn+1 = Bxn +Cxn−1 . . 6. . .4. . . .6.1 The Case p < 1 .4 Global Behavior of Solutions . . . . . . . . . . . . . . . .1 Introduction . . . . . . . .5. . . . . . .4 The Case α = γ = B = 0 : xn+1 = 4. .5. . . .3 Semicycle Analysis and Global Attractivity When q = 1 . . . . .7. . . . 82 83 84 87 89 90 90 91 92 93 96 97 100 101 102 103 105 107 . . . . . . .7. . . . . . . 6. . . . . . . 1)-Type Equations 5.

. . . . . . . . . . . . . . . .3. . . .3. . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . .1 Introduction .3. . . . . . .5 Open Problems and Conjectures . . . . . . . 9. . . . .5 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 The Case α = β = 0 : xn+1 = A+Bxnn−1 n−1 . . . . n +Cxn−1 9. . . . . .3 The Case α = γ = 0 : xn+1 = A+Bxβxn n−1 . . . . . . . . . . . . . . .5 Long-term Behavior of Solutions When q > 1 6. . . 6. . . . . . 9. . . . . . .4 Global Stability Analysis When p > q . . . . .3 Global Stability Analysis When p < q . . . . . . . . α+γxn−1 9. . .2 The Case γ = 0 : xn+1 = A+Bxn +Cxn−1 . . . . βxn +γx 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3)-Type Equations 131 7. . . .2 The Case β = γ = 0 : xn+1 = A+Bxnα n−1 . .1 Invariant Intervals . .9. . . . .4. . . . . . . . . . . . . . n−1 6. . . . . .2 Invariant Intervals . . . .1 Existence and Local Stability of Period-Two Cycles 9. . . .7. . . . . . . . . . . . . . . . . . 6. . . . . 6.1 Introduction . . . . . . . . . . . . . . . . .3. . . . . .8. . . . . . . . . . . . A+Cxn−1 6. . . . . 132 n +Cx γx 7. . . . . .3 The Case β = 0 : xn+1 = A+Bxn +Cxn−1 . . . . . . . 137 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 +Cx 7.4 Global Stability . . . . . . . . . . . . . . . . . . . . . . . . . . 9. . . . . . . 9. . . . .Contents 6. . .9. . . . . . . . . . . . . 131 +Cx 7. .4 The Case α = 0 : xn+1 = A+Bx+γxn−1 . . . . . . . . . . . . 2 9. . . . . . . . . . . . . . . . .3 Convergence of Solutions . . . vii 107 109 110 112 113 114 117 117 123 123 7 (1. . 9. . . . . . . . .5 Semicycle Analysis When pr = q + qr . . . . . . . . . . . 131 7. . .1 Boundedness of Solutions . 3)-Type Equations 9. . . . .1 Invariant Intervals and Semicycle Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Introduction . . . . . . . . . . . . . . . . . . . .8 The Case α = B = 0 : xn+1 = βxn +γxn−1 . . . . . . . . . . . . .2 Invariant Intervals . . . . . . . . . . 1)-Type Equations 137 8. . . . . . . . . .4 Global Asymptotic Stability . . . . . . . . . . . . . . . . . . α+βxn 9. . . . . . . .1 Existence of a Two Cycle . . . . . . . . . . . . . . . . . .2. .10 Open Problems and Conjectures . . . . . . . . . . . . .2 Convergence of Solutions . . . . 9. . . . . . . .3 Semicycle Analysis . . . . 138 9 (2. βxn 9. . .8. . . . . . . . . . . . . . .9. . . .9. . . . . . .2 Open Problems and Conjectures .3. . . . . . . . . . . . . . . . . 6. . . . . . . . . . .2 Global Stability of the Positive Equilibrium . . 141 141 141 142 143 144 147 149 150 154 156 157 157 158 160 162 163 . . . . . . . . . . . . . . . . . .4. . . . . .2 Semicycle Analysis . . . . .2.2. . . . 6.9 The Case α = A = 0 : xn+1 = Bxn +Cxn−1 . . . . . . . . . . . . . 134 8 (3. . . . . . . . . . . 9. 9. 9.

. . .1 Global Stability . . . .1 10. . . . . . . . .5 Open Problems and Conjectures . . . . . 3)-Type Equation xn+1 = A+Bxn +γxn−1 n +Cxn−1 Linearized Stability Analysis . . . . . . . . . . . .2. . . . 2)-Type Equations 10. . . . . . . . . . . . . . .2 11. . . . . 10. . . . . 10. . . A+Cxn−1 10. . . .4 The Case A = 0 : xn+1 = α+βx+Cxn−1 . . Invariant Intervals . . . . . . . . . . . . . . . . .3. 10. . . . . . . .3.2. . .2 Global Stability When p ≥ r . .2 Proof of Part (c) of Theorem 10. . . . . . . . . . . . . . . . . . . . . . . . . . . . .4. . . . . . . . . . . . . . .viii 10 (3. . . . . . . . α+βx (3. . . . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . .2 Convergence of Solutions . . . . . . . . . . . . . . . . . . .2 The Case C = 0 : xn+1 = α+βxn +γxn−1 . . . . . . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . .4 Open Problems and Conjectures . . . . . . 11. . . . . . . . . . . 10. . . . . . . . .1 Proof of Part (b) of Theorem 10. . . . . . .1 10. . . . . . . . . . Bxn 10. . . . . . .1 Introduction . . . . . . . .1 Invariant Intervals . . .1 11. . . . . . . . . . . . 11. . . . . . . . . . . .4. . . . . . . . . . .2. . . . . . . . . n +γxn−1 10. . . . . . . . . . . A+Bxn 10. 11 The 11. . . . A Global Attractivity for Higher Order Equations Bibliography Index .3 Convergence of Solutions . . . . . . . . . . 10. . . . . . Convergence Results . . . . .3 The Case B = 0 : xn+1 = α+βxn +γxn−1 . . . . . . . . . . . . . . . . . .3 Contents 167 167 167 169 172 172 173 173 174 175 177 179 180 183 183 184 189 191 192 201 211 217 . . . .3. . . . . . . .1 Invariant Intervals . . . . . . . . . . . . .

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

.

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

.

To Senada and Theodora .

.

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

4. and Section 5.2. and Section 4. n = 0. Lyness’ equation when γ = A = B = 0. [57]. See [13]. A characteristic feature of this equation is that it possesses an invariant which can be used to understand the character of its . See [42]. [69]. . [44]. and the rational equation xn+1 = α + αxn + βxn−1 . p. See ([42]. [66]. 1. . 3. [32]. [18]. 1.2. it is interesting to note the similarities of the dynamics of the population model xn+1 = α + βxn−1 e−xn . / if and only if x0 ∈ F. 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. we introduce the forbidden set of the equation.7 and 10. .6 for some results in this regard. 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. Hence the solution {xn } of Eq(2) exists for all n ≥ 0. x0 ) ∈ R × R through which the equation xn−1 xn+1 = −1 + xn is well deﬁned for all n ≥ 0. The problem of existence of solutions for diﬀerence equations is of paramount importance but so far has been systematically neglected. This is a fascinating equation with many interesting extensions and generalizations. For example. [55]. [67]. . . [49]. See [13].2 Introduction and Classiﬁcation of the Riccati equation. 1 + xn n = 0. . See [24] and Sections 2. 75). and Section 1. Pielou’s discrete delay logistic model when α = γ = B = 0. An example of such a problem is the following: Open Problem Find all initial points (x−1 . We believe that the techniques which we develop in this monograph to understand the dynamics of Eq(1) will also be of paramount importance in analyzing the equations in the mathematical models of various biological systems and other applications. 2. .

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

.

Then for every set of initial conditions x0 . [16]. Deﬁnition 1. that is. [34]. The reader may just glance at the results in this chapter and return for the details when they are needed in the sequel. . [63].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). [41]. n=−1 A point x ∈ I is called an equilibrium point of Eq(1.1. . . n = 0. [22]. 1. x). 1. xn = x for n ≥ 0 is a solution of Eq(1. [21]. the diﬀerence equation xn+1 = f (xn .1 Let x be an equilibrium point of Eq(1. [33]. [4].1) if x = f (x.1).1) has a unique solution {xn }∞ .1). and [68]. 5 . x−1 ∈ I. (1.1 Deﬁnitions of Stability and Linearized Stability Analysis f :I ×I →I Let I be some interval of real numbers and let be a continuously diﬀerentiable function. or equivalently. See also [1]. x is a ﬁxed point of f . xn−1 ).

1. Preliminary Results (i) The equilibrium x of Eq(1. we have |xn − x| < ε for all n ≥ −1. (vi) The equilibrium x of Eq(1. Clearly a source is an unstable equilibrium. if there exists r > 0 such that for all x0 . (iii) The equilibrium x of Eq(1.1). ∂f ∂f (x. Theorem 1.1) is called a global attractor if for every x0 . (1. x−1 ∈ I we have lim xn = x. x−1 ∈ I with 0 < |x0 − x| + |x−1 − x| < r. and if there exists γ > 0 such that for all x0 . v) evaluated at the equilibrium x of Eq(1. n→∞ (iv) The equilibrium x of Eq(1. .1) is called locally stable if for every ε > 0.2) p= is called the linearized equation associated with Eq(1.1 (Linearized Stability) (a) If both roots of the quadratic equation λ2 − pλ − q = 0 (1. . x) ∂u ∂v denote the partial derivatives of f (u. x) and q = (x.1) is called locally asymptotically stable if it is locally stable.6 Chapter 1.1) is called unstable if it is not stable.1) is called globally asymptotically stable if it is locally stable and a global attractor. or a repeller.1) is locally asymptotically stable. (v) The equilibrium x of Eq(1. we have n→∞ lim xn = x. Then the equation yn+1 = pyn + qyn−1 . x−1 ∈ I with |x0 − x| + |x−1 − x| < γ. there exists N ≥ 1 such that |xN − x| ≥ r. 1. .1) about the equilibrium point x. then the equilibrium x of Eq(1. x−1 ∈ I with |x0 − x| + |x−1 − x| < δ. there exists δ > 0 such that for all x0 .1) is called a source. (ii) The equilibrium x of Eq(1.3) Let lie in the open unit disk |λ| < 1. . n = 0.

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).

Proof. i→∞ and M = lim Mi .4. or nonincreasing in both arguments. See also the Appendix. mi−1 ) and mi = f (mi−1 . (b) If (m.4. ≤ Mi ≤ . m0 ≤ m1 ≤ . b] is a solution of the system f (m. which describes the extension of these results to higher order diﬀerence equations. Theorems 1. . y) is non-decreasing in x and non-increasing in y and the second theorem deals with the case where the function f (x. b] for each x ∈ [a. y). b] for each y ∈ [a. . (See Section 6. Then Eq(1. Set m0 = a and M0 = b and for i = 1.5 ([54]) Let [a. Mi−1 ). b]. ≤ mi ≤ .7 below are new and deal with the remaining two cases relative to the monotonic character of f (x.) The ﬁrst theorem deals with the case where the function f (x. .1) converges to x.8 and 1. at the end of this monograph. then m = M . m) = M. b] × [a. b] → [a. . M ) = m and f (M. y) is non-increasing in x and nondecreasing in y. Theorem 1. M ) ∈ [a. set Mi = f (Mi−1 .4 and 6. See Sections 6. b] and every solution of Eq(1. b] × [a. y) is monotonic in its arguments. y) is nondecreasing in both arguments. b] be an interval of real numbers and assume that f : [a. 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]. . . .8 where these theorems are utilized.9. In this regard there are four possible scenarios depending on whether f (x. and f (x. .4.1) is to work in the regions where the function f (x.1) has a unique equilibrium x ∈ [a. ≤ M1 ≤ M0 and Set mi ≤ xk ≤ Mi m = lim mi i→∞ for k ≥ 2i + 1.4. b] is a continuous function satisfying the following properties: (a) f (x.1. . or nonincreasing in one and nondecreasing in the other. 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 ∈ [a. y) is nonincreasing in y ∈ [a. 2. Now observe that for each i ≥ 0. .

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

4. b] is a continuous function satisfying the following properties: (a) f (x. . b] → [a. set Mi = f (mi−1 . b] be an interval of real numbers and assume that f : [a.4. m=M =x from which the result follows.12) holds and by the continuity of f . M ) and M = f (m.7 Let [a.8 Let [a. Global Attractivity of the Positive Equilibrium Theorem 1. and Set mi ≤ xk ≤ Mi m = lim mi i→∞ for k ≥ 2i + 1. . . Then Eq(1. m) = M and f (M. Mi−1 ). . b] and every solution of Eq(1. i→∞ and M = lim Mi . . ≤ M1 ≤ M0 .1) has a unique equilibrium x ∈ [a. mi−1 ) and mi = f (Mi−1 . 2 Theorem 1. b] × [a. M ) ∈ [a. y) is non-increasing in each of its arguments.1) converges to x. b] be an interval of real numbers and assume that f : [a. m = f (M. Set m0 = a and M0 = b and for i = 1. M ) = m.4. Then clearly (1. b] × [a. Proof. . m).1) converges to x. ≤ mi ≤ . Now observe that for each i ≥ 0.1) has a unique equilibrium x ∈ [a. 2. b] and every solution of Eq(1. b] is a solution of the system f (m. . .1. (b) The equation f (x. Then Eq(1. ≤ Mi ≤ . (b) If (m. 13 then m = M . b] is a continuous function satisfying the following properties: (a) f (x. has a unique positive solution. In view of (b). m0 ≤ m1 ≤ . . . x) = x. b] × [a. b] → [a. y) is non-decreasing in each of its arguments.

. . and mi ≤ xk ≤ Mi Set m = lim mi i→∞ for k ≥ 2i + 1. ∞ we will show that there exists a solution {Ln }n=−k of the diﬀerence equation xn+1 = f (xn . In view of (b). ≤ mi ≤ .1) such that A ≤ xn ≤ B for all n ≥ −1. Now with the above hypotheses. . 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). i→∞ and M = lim Mi . Assume that {xn }∞ n=−1 is a solution of Eq(1. mi−1 ). In this section we ﬁrst give a self-contained version of limiting solutions the way we will use them in this monograph. (1. m = f (m. where A. Now observe that for each i ≥ 0. . xn−1 ). set Mi = f (Mi−1 . m) and M = f (M.12) is satisﬁed and by the continuity of f . . J]. . . . B ∈ J. ≤ M1 ≤ M0 . . Let J be some interval of real numbers. ≤ Mi ≤ . Set Chapter 1. Then clearly the inequality (1. 2 1. . m = M = x. M ). and let k ≥ 1 be a positive integer. Preliminary Results m0 = a and M0 = b and for i = 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. n ≥ −k + 1. 2.13) . from which the result follows. f ∈ C[J × J. m0 ≤ m1 ≤ .14 Proof. Mi−1 ) and mi = f (mi−1 . . .

. . . n=−1 The following result. . is stated in the way we will use it in this monograph: . xnj −k ) = lim xnj −k+2 = L−k+2 .13) with Then clearly. which is a consequence of the above discussion. L−k ) j→∞ = lim f (xnj −k+1 .14) associated with the solution {xn }∞ n=−1 of Eq(1. L1 = 1. −k+2 j→∞ = L−k and −k+1 = L−k+1 . −k Let { n }∞ n=−k be the solution of Eq(1. Therefore. L0 = L = 0 . Karakostas takes k = ∞ and calls the solution {Ln }∞ n=−∞ a full limiting sequence of the diﬀerence equation xn+1 = f (xn . (1. (See Theorem 1.5.1). . lim xnij = L and lim xnij −1 = L−1 . .5.13) associated with the The solution {Ln }∞ n=−k solution {xn }∞ n=−1 of Eq(1.1)..13) and every term of this sequence is a limit point of the solution {xn }∞ . .) ∞ We now proceed to show the existence of a limiting solution {Ln }n=−k .. L−k and a subsequence {xnj }n=−1 such that j→∞ lim xnj = L j→∞ lim xnj −1 = L−1 . −k ) = f (L−k+1 . Clearly there exists a subsequence {xni }∞ of {xn }∞ i=1 n=−1 such that i→∞ lim xni = L. the term LN is a limit point of the solution {xn }∞ . satisﬁes Eq(1. L−k+1 . i→∞ i→∞ Continuing in this way and by considering further and further subsequences we obtain ∞ numbers L−2 . L−3 .1. L−1 . B] and so it has a ∞ further subsequence {xnij }j=1 which converges to some limit which we call L−1 . Limiting Solutions such that L0 = L 15 and for every N ≥ −k. . xn−1 ). .2 below. It follows by induction that the sequence L−k . = f( −k+1 . . . ∞ Next the subsequence {xni −1 }i=1 of {xn }∞ n=−1 also lies in interval [A. n=−1 is called a limiting solution of Eq(1. j→∞ lim xnj −k = L−k .

1 Let J be some interval of real numbers. S ∈ J. . with I = lim inf xn . . i=0 xn−i k n = 0. n = 0. . . there exist two subsequences {xrn } and {xln } of the solution {xn }∞ n=−ν such that the following are true: n→∞ (1. S0 = S. 0. J]. . .5. Theorem 1. with I0 = I. as developed by Karakostas. S] for all n ∈ Z and such that for every N ∈ Z. . Sn ∈ [I. In . . .2 to diﬀerence equations of the form xn+1 = Ai . Then Eq(1. and let k ≥ 1 be a positive integer. Let Z denote the set of all integers {. xn−1 . .16) lim xrn +N = IN and n→∞ lim xln +N = SN for every N ≥ m. . and In . n→∞ S = lim sup xn n→∞ n = 0.16 Chapter 1.15). which by the change of variables xn = y1 reduce to rational diﬀerence equations of the n form 1 . . S ∈ J. xn−1 . 1. yn+1 = k i=0 Ai yn−i . . .13) has two solutions {In }∞ n=−k with the following properties: and {Sn }∞ n=−k I0 = I.}. 1. xn−ν ). n=−ν Furthermore for every m ≤ −ν. 1. Suppose that {xn }∞ n=−1 is a bounded solution of Eq(1. .5. . Preliminary Results Theorem 1. (1. See [23] and [25] for an application of Theorem 1. . J]. . Then there exist two solutions ∞ {In }∞ n=−∞ and {Sn }n=−∞ of the diﬀerence equation xn+1 = f (xn . . with I. f ∈ C[J × J.15) and with I. S] for n ≥ −k.5. .1) with limit inferior I and limit superior S. 1.16) associated with the solution {xn }∞ n=−k of Eq(1. ∞ The solutions {In }∞ n=−∞ and {Sn }n=−∞ of the diﬀerence equation (1.2 ([37]) Let J be some interval of real numbers. . f ∈ C[J ν+1 . Sn ∈ [I. .16) are called full limiting solutions of Eq(1. . IN and SN are limit points of {xn }∞ . xn−ν ) which satisfy the equation for all n ∈ Z. S0 = S We now state the complete version of full limiting sequences. −1. . and let {xn }∞ n=−ν be a bounded solution of the diﬀerence equation xn+1 = f (xn .

18). (1.21) β B (1. x0 are all real numbers) other than the material presented here. C and the initial conditions x−1 .17).18) Indeed when B = 0. It should be mentioned here that there is practically nothing known about the forbidden set of Eq(1) (when the parameters α.17). Theorem 1. which is extracted from [32]. A. A + Bxn A diﬀerence equation of the form xn+1 = n = 0.6. A + Bxn B The set of initial conditions x0 ∈ R through which the denominator A+Bxn in Eq(1.17) with x0 = is periodic with period two. The proof is straightforward and will be omitted.1.17) where the parameters α. Then (1. .17) possesses a prime periodtwo solution.1 Assume that (1. To avoid degenerate cases. (See also [13] and [69]. . γ.17) regarding period-two solutions. In the sequel. A.18) holds and that Eq(1.6. we will assume that β + A = 0. The Riccati Equation 17 1. B. Eq(1.17) when x0 ∈ F. . we will assume that B = 0 and αB − βA = 0.and short-term behavior of / solutions of Eq(1. One of our goals in this section is to determine the forbidden set F of Eq(1. Furthermore when (1.19) β + A = 0.17) is a linear equation and when αB − βA = 0.17) reduces to βA + βxn β xn+1 = B = for all n ≥ 0. (1. in addition to (1.19) holds every solution of Eq(1. 1.) The other goal we have is to describe in detail the long. Eq(1. β.20) . (1.17) will become zero for some value of n ≥ 0 is called the forbidden set F of Eq(1. β.6 The Riccati Equation α + βxn . The ﬁrst result in this section addresses a very special case of Eq(1. B and the initial condition x0 are real numbers is called a Riccati diﬀerence equation.

17) into the diﬀerence equation wn+1 = 1 − where R= R . 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 .18 Now observe that the change of variable xn = β+A A wn − B B Chapter 1. wn n = 0.22) transforms Eq(1. Preliminary Results for n ≥ 0 (1. 1.23) and present its forbidden set F and the character of its solutions.22).23) βA − αB (β + A)2 is a nonzero real number. It is interesting to note that the four parameters of Eq(1.23). called the Riccati number of Eq(1. w Also note that the signum of R is equal to the signum of the derivative y =1− f (x) = βA − αB . . . (A + Bx)2 of the function For the remainder of this section we focus on Eq(1. Similar results can be obtained for Eq(1.17). . From Eq(1.23) it follows that when R< 1 4 . (1.17) through the change of variables (1.17) have been reduced to the single parameter R of Eq(1.

23) has exactly one equilibrium point namely. The forbidden set of Eq(1.23) has the two equilibrium points w− and w+ with w− = and |w− | < w+ .24) are the equilibrium points of Eq(1. where 4 the characteristic roots of Eq(1. .1. which can be solved explicitly. 2 Finally. when 1 R> . . Note that in these two cases the characteristic roots of Eq(1.23) has no equilibrium points. Now observe that the change of variables wn = with y0 = 1 and y1 = w0 reduces Eq(1. ).6. 1.23). 1.23) to the second order linear diﬀerence equation yn+2 − yn+1 + Ryn = 0. (1. Also w− < 0 if R < 0 and 1 w− > 0 if R ∈ (0. . .23) is clearly the set of points where the sequence {yn }∞ n=0 vanishes.24) yn+1 yn for n = 0. 1 w= . 4 1− √ √ 1 − 4R 1 + 1 − 4R < = w+ 2 2 19 When Eq(1. . n = 0.24) are real numbers. respectively. The Riccati Equation Eq(1. . . 4 1 R= . 1 The next two theorems deal with the cases R < 1 and R = 4 . 4 Eq(1.

23) is the sequence of points fn = n−1 2n for n = 1. . (1. .23) is given by / n+1 n+1 (w0 − w− )w+ − (w+ − w0 )w− .23).23) is the sequence of points fn = n−1 n−1 w+ − w− w+ w− n n w+ − w− for n = 1.27) .6.3 Assume Then w= 1 R= .23). Theorem 1. . . (1. .2 Assume Then 1 R< . 2. 4 1− Chapter 1. 4 1 2 is the only equilibrium point of Eq(1.6. The equilibrium w+ is a global attractor as long as w0 ∈ F ∪ {w− } / and the equilibrium w− is a repeller. 2. . . .6. if R > 0 and fn < w− while (fn − w− )(fn+1 − w− ) < 0 if R < 0. The forbidden set F of Eq(1.20 Theorem 1. The forbidden set F of Eq(1.2 that n→∞ lim fn = w− .26) It is now clear from Theorem 1. (1. wn = n n (w0 − w− )w+ − (w+ − w0 )w− n = 1. Preliminary Results √ √ 1 − 4R 1 + 1 − 4R w− = and w+ = 2 2 are the only equilibrium points of Eq(1. . .25) When w0 ∈ F. the solution of Eq(1. 2. . .

23) is a global 2 / attractor for all solutions with w0 ∈ F but is locally unstable. . . . . 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 .6. 4 The characteristic roots of Eq(1. R 4R − 1 cos(nφ) + (2w0 − 1) sin(nφ) . . it is a sink from the right and a source from the left. 1 R> .1.3 it follows that the equilibrium w = 1 of Eq(1.29) n Finally consider the case √ R. the solution of Eq(1. 4R − 1 n = 0. 2 R n = 0. (1. 2w0 − 1 C1 = 1 and C2 = √ . more precisely. .28) From Theorem 1. ) 2 and √ 4R − 1 √ sin φ = .6. . π φ ∈ (0. . (1. yn = R 2 (C1 cos(nφ) + C2 sin(nφ)). The Riccati Equation 21 which converges to the equilibrium from the left. 1. 1. .23) is given by / wn = 1 + (2w0 − 1)(n + 1) 2 + 2(2w0 − 1)n for n = 0. When w0 ∈ F. Hence wn = √ √ 4R − 1 cos(n + 1)φ + (2w0 − 1) sin(n + 1)φ √ .

30) 1 Clearly when R > 4 the character of the solution {wn } and the forbidden set F depends on whether or not the number φ ∈ (0. 2 R Then the forbidden set F of Eq(1. . 2. . that is. and so √ cos(nφ + φ − θ) √ cos(nφ − θ) cos φ − sin(nφ − θ) sin φ R = R cos(nφ − θ) cos(nφ − θ) √ = R (cos φ − sin φ tan(nφ − θ)) 1 wn = − 2 √ 4R − 1 tan(nφ − θ). 1.31) from which the character of the solutions of Eq(1.4 Assume that R> and let φ ∈ (0.23) is given by (1. . π ) be such that 2 1 cos φ = √ 2 R and 1 4 √ 4R − 1 √ sin φ = .23) is the sequence of points √ 1 4R − 1 fn = − cot(nφ) for n = 1.29). w0 = − 2 2 The following theorem summarizes the above discussion. . . . Preliminary Results as long as the denominator is diﬀerent from zero. that is. For example. 2 Let θ ∈ (− π . . π ) is a rational multiple of π. can now be easily 4 revealed. / (1. 2 2 When w0 ∈ F.22 Chapter 1. .23). the solution of Eq(1. 2. ) N 2 (1. Theorem 1. . . . wn = (4R − 1)2 + (2x0 − 1)2 .32) . assume φ is a rational multiple of π.6. φ= π k π ∈ (0. (1. √ 4R − 1 1 cot(nφ) for n = 1. when R > 1 . π ) be such that 2 2 √ 4R − 1 2x0 − 1 and sin θ = cos θ = r r where r= Then we obtain. 2 n = 0.

} and 2k < N.23) is periodic with period N .5 is not a rational multiple of π. p − 1 . 2. ) 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 = . .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.5 Assume that φ= k π π ∈ (0. . then every solution of Eq(1.6. then the following result is true: Theorem 1.23) is periodic. Equivalently assume that A= where k.5 is not a rational multiple of π. Then the following statements are true: (i) No solution of Eq(1. / for n = 1.6.6. . . .6.23) is periodic with period N .23) with w0 ∈ F is dense in R.1. . 2 R 2 R or equivalently k 4R cos2 ( π) = 1. 2. . . . (ii) The set of limit points of a solution of Eq(1. Theorem 1.6. N ∈ {1. N ∈ {1. 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. . The following theorem summarizes the above discussion. When the number φ in Theorem 1.} with 2k < N. The Riccati Equation 23 for some k.6 Assume that the number φ in Theorem 1. 2.

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. . . . xn−1 ).23) and the fact that when φ is not a rational multiple of π.33) relative to an equilibrium point x. then the values (nφ − θ)(mod π) for n = 0.33) under appropriate hypotheses on the function f . 2 2 2 1. . . π ). xm }. .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. or l > −1 and xl−1 < x and either m = ∞. or m < ∞ and xm+1 < x. The proof of statement (ii) is a consequence of the form of the solution of Eq(1. are dense in the interval (− π . . all greater than or equal to the equilibrium x. . 1. φ is a rational multiple of π. A positive semicycle of a solution {xn } of Eq(1. Statement (i) is true because. wk = wl if and only if Chapter 1. with l ≥ −1 and m ≤ ∞ and such that either l = −1. (1. xl+1 . .24 Proof. 1. we present some results about the semicycle character of solutions of some general diﬀerence equations of the form xn+1 = f (xn . .1) consists of a “string” of terms {xl . First we give the deﬁnitions for the positive and negative semicycle of a solution of Eq(1. n = 0. . In this section.

Then there exists N ≥ 0 such that either xN −1 < x ≤ xN or xN −1 ≥ x > xN .5. The ﬁrst result was established in [28] while we were working on a special case of the equation we are investigating in this monograph. All other cases are similar and will be omitted. ∞)] is such that: f (x. y) we have xN +1 = f (xN .1 ([28]) Assume that f ∈ C[(0.7. n2 ≥ n0 such that xn1 xn2 < 0.33) consists of a “string” of terms {xl . A sequence {xn } is called strictly oscillatory if for every n0 ≥ 0.1. Thus xN +1 < x < xN +2 or m < ∞ and xm+1 ≥ x. y) is increasing in y for each ﬁxed x. Semicycle Analysis 25 A negative semicycle of a solution {xn } of Eq(1. xN ) > f (x. Then except possibly for the ﬁrst semicycle. . y) is decreasing in x for each ﬁxed y. with l ≥ −1 and m ≤ ∞ and such that either l = −1. ∞) × (0. xN −1 ) < f (x. (See Section 6. (0.7.1 (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. .7. . xl+1 . The sequence {xn } is called strictly oscillatory about x if the sequence xn − x is strictly oscillatory. Let {xn } be a solution of Eq(1.33) with at least two semicycles.33) has semicycles of length one. x) = x. . Proof. ∞). and f (x.) Theorem 1. xm }. Then by using the monotonic character of f (x. every solution of Eq(1. (b) A sequence {xn } is said to oscillate about x if the sequence xn − x oscillates.33). all less than the equilibrium x. x) = x and xN +2 = f (xN +1 . Otherwise the sequence is called nonoscillatory. We will assume that xN −1 < x ≤ xN . Let x be a positive equilibrium of Eq(1. Deﬁnition 1. or l > −1 and xl−1 ≥ x and either m = ∞. there exist n1 . .

2 Assume that f ∈ C[(0. The proof in the case of negative semicycle is similar and is omitted. The proof in the case of negative semicycle is similar and is omitted.3 Assume that f ∈ C[(0. with at least one of the inequalities being strict.7. Then except possibly for the ﬁrst semicycle. Chapter 1. y) is increasing in both arguments.33) has semicycles of length at most two.33) has semicycles of length one. Let x be a positive equilibrium of Eq(1. (0. Proof. xN −1 ) < f (x. Theorem 1. Then every oscillatory solution of Eq(1. Then by using the decreasing character of f we obtain: xN +1 = f (xN . 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. (0. Proof.7. x) = x which shows that the next term xN +1 also belongs to the positive semicycle. Preliminary Results 2 The next result applies when the function f is decreasing in both arguments. ∞) × (0. xN −1 ) > f (x. Let x be a positive equilibrium of Eq(1. ∞)] and that f (x. 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. y) is increasing in x and decreasing in y. It follows by induction that all future terms of this solution belong to this positive semicycle. ∞) × (0. ∞).33). Theorem 1.26 and the proof follows by induction. x) = x which completes the proof.33). Then by using the increasing character of f we obtain: xN +1 = f (xN . The next result applies when the function f is increasing in both arguments. ∞)] and that f (x. 2 The next result applies when the function f (x. y) is decreasing in both arguments. which is a contradiction. every oscillatory solution of Eq(1. ∞). 2 .

xN −1 ) < f (xN . is similar and is omitted. Semicycle Analysis 27 Theorem 1. Let x be a positive equilibrium of Eq(1. (0. if we assume that f (u. If xN > xN −1 > x then by using the increasing character of f and Condition (1.33) has semicycles of length at least two. and in the second term if the semicycle has three terms. and f (x. y) < x for every x > y > x (1.7.34) . Furthermore. except possibly for the ﬁrst semicycle. ∞). Then. except possibly for the ﬁrst semicycle which may have length one. The proof in the case xN −1 ≥ x > xN . xN ) = x which shows that the positive semicycle has length two. xN ) = x (1. Assume that {xn } is an oscillatory solution with two consecutive terms xN −1 and xN such that xN −1 < x ≤ xN .4 Assume that f ∈ C[(0. u) = x for every u and f (x.35) then {xn } oscillates about the equilibrium x with semicycles of length two or three.34) we obtain: xN +1 = f (xN . xN −1 ) > f (x. ∞)] is such that: f (x. y) is decreasing in y for each ﬁxed x. x) = x which shows that the next term xN +1 also belongs to the positive semicycle. ∞) × (0. Proof.1. xN −1 ) > f (xN . y) is increasing in x for each ﬁxed y.34) we obtain xN +1 = f (xN . Then by using the increasing character of f we obtain xN +1 = f (xN .33). The extreme in each semicycle occurs in the ﬁrst term if the semicycle has two terms. every oscillatory solution of Eq(1. Then by using the increasing character of f and Condition (1.7. 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.

Assume that f satisﬁes Conditions (1. Preliminary Results xN +1 = f (xN .33). ∞). y) is increasing in x for each ﬁxed y. Theorem 1. In view of Condition (1.35) we ﬁnd Chapter 1. 2 Condition (1.37) .36) is satisﬁed and that x0 > x−1 > x. y) > x for every x > y > x.7.33) has monotonic solutions.34) and (1. assume that Condition (1. on the other hand. y) is decreasing in y for each ﬁxed x.28 and by using Condition (1. 2 (1.36). then Eq(1. If. x−1 ) > x0 . x−1 ) > f (x−1 . ∞) × (0. instead of Condition (1.36) we get x1 = f (x0 . the corresponding solution is increasing.36) f (x. First. The proof of the remaining case is similar and will be omitted. we obtain x1 = f (x0 . and f (x. and the proof follows by induction. y) < x for every x < y < x then if x0 < x−1 < x.34). ∞)] is such that: f (x.34) is not enough to guarantee the above result. If. x−1 ) = x. (0. f satisﬁes Condition (1.5 Assume that f ∈ C[(0.34). and f (x. Then by the monotonic character of f and (1. we assume (1. the corresponding solution is decreasing.35). Let x be a positive equilibrium of Eq(1. Proof. xN −1 ) < xN and the proof is complete. Then if x0 > x−1 > x.

1) has a positive equilibrium if and only if β + γ > A. B + C ∈ (0.4) α + (β + γ)x A + (B + C)x (2. and Invariant Intervals 2. β. A + Bxn + Cxn−1 Here we investigate the equilibrium points of the general equation xn+1 = where α. (2.2) and where the initial conditions x−1 and x0 are arbitrary nonnegative real numbers such that the right hand side of Eq(2.1) When (2. Eq(2. Zero is an equilibrium point of Eq(2. 29 . In view of the above restriction on the initial conditions of Eq(2. . γ. Periodicity.1) if and only if α = 0 and A > 0.3) n = 0.Chapter 2 Local Stability. B.1) is well deﬁned for all n ≥ 0. ∞) (2.1 Equilibrium Points α + βxn + γxn−1 . 1. the equilibrium points of Eq(2.5) holds. . Semicycles.1) are the nonnegative solutions of the equation x= or equivalently (B + C)x2 − (β + γ − A)x − α = 0. . A. ∞) with α + β + γ. in addition to the zero equilibrium .5) (2.1). C ∈ [0. (2.

2). when α = 0.6) Note that in view of Condition (2.4) would also be an equilibrium worth investigating. Semicycles.1) is unique and is given by x= When α = 0 and A = 0 the only equilibrium point of Eq(2. n = 0.30 Chapter 2.7) β+γ−A . except for our presentation in Section 1.4). [32].1) is positive and is given by x= β+γ . 1. v) = (A + Bu + Cv)2 (2. and Invariant Intervals In fact in this case the positive equilibrium of Eq(2.1) is the positive solution x= β+γ−A+ (β + γ − A)2 + 4α(B + C) 2(B + C) (2.6 about the Riccati equation α + βxn xn+1 = . . and [69] for some results in this regard. [18]. Periodicity. v) = A + Bu + Cv and observe that (βA − αB) + (βC − γB)v fu (u.1). it satisﬁes Eqs(2.2 Stability of the Zero Equilibrium Here we investigate the stability of the zero equilibrium of Eq(2.4). Set α + βu + γv f (u. The problem of investigating Eq(2. (See [13]. B+C (2.3) and (2.8). . If we had allowed negative initial conditions then the negative solution of Eq(2. . Finally when α>0 the only equilibrium point of Eq(2. .1) when the initial conditions and the coeﬃcients of the equation are real numbers is of great mathematical importance and.8) of the quadratic equation (2.) In summary.1) has a positive equilibrium x. B+C (2. 2.9) . then x is unique.1). the quantity β + γ is positive. Local Stability. A + Bxn there is very little known. This observation simpliﬁes the investigation of the local stability of the positive equilibrium of Eq(2. it is interesting to observe that when Eq(2. and it is given by (2.

1. 1. v) = (γA − αC) − (βC − γB)u . .2. . A + Bxn + Cxn−1 n = 0.1).2. . The local stability character of x is now described by the linearized stability Theorem 1. Then the zero equilibrium point of the equation xn+1 = βxn + γxn−1 .10) If x denotes an equilibrium point of Eq(2. Furthermore the zero equilibrium point is a sink when a saddle point when a repeller when A>β+γ A < β + γ < A + 2β β + γ > A + 2β.1) about the equilibrium point x is zn+1 − pzn − qzn−1 = 0 where p = fu (x. .1.1 Assume that A > 0. 1. in addition to the local stability Theorem 1. A A For the stability of the zero equilibrium of Eq(2. . x). For the zero equilibrium of Eq(2. then the linearized equation associated with Eq(2. (A + Bu + Cv)2 31 (2. .2. x) and q = fv (x.1).1) we have p= β A and q = γ A and so the linearized equation associated with Eq(2.1. is globally asymptotically stable when β+γ ≤A and is unstable when β + γ > A. .1. n = 0. 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. .3. Stability of the Zero Equilibrium and fv (u. As a consequence of these two theorems we obtain the following global result: Theorem 2.1.

11) and q > −1. and it is given by (2. By using the identity (B + C)x2 = (β + γ − A)x + α.1) has a positive equilibrium when either or α > 0.32 Chapter 2. are equivalent to some inequalities of the form x>ρ .11) and (2. Semicycles. In these cases the positive equilibrium x is unique. x) = (γA − αC) − (βC − γB)x (γA − αC) − (βC − γB)x = 2 .4). Periodicity.8).1.3 Local Stability of the Positive Equilibrium α = 0 and β + γ > A As we mentioned in Section 2.1.3) and (2.1) the following: |p| < 1 − q (2.12) (A + β + γ)x > −α − (2. and Invariant Intervals 2. Local Stability. Inequalities (2.13)-(2. 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. we see that p = fu (x. it satisﬁes Eqs(2.13) (2. 2 (A + (B + C)x) A + α(B + C) + (B + C)(A + β + γ)x (βA − αB) + (βC − γB)x (βA − αB) + (βC − γB)x = 2 2 (A + (B + C)x) A + α(B + C) + (B + C)(A + β + γ)x Hence the conditions for the local asymptotic stability of x are (see Theorem 1.14) (2.12) are equivalent to the following three inequalities: A+β+γ+2 βC − γB A2 + (βA − αB) − (γA − αC) x > −α − B+C B+C A2 − (βA − αB) − (γA − αC) B+C (2.15). Eq(2. x) = and q = fv (x.15) A2 + (γA − αC) βC − γB x > −α − .

and [74]. n = 0.1 Let p ≥ 2 be a positive integer and assume that every positive solution of Eq(2. 1. . . . When is Every Solution Periodic with the Same Period? and/or x<σ for some ρ. Now if we set F (u) = (B + C)u2 − (β + γ − A)u − α. 4.1) with the property that every solution of the equation is periodic with the same period. (ii) Assume C = 0.4 When is Every Solution Periodic with the Same Period? 1 . 1. The following four special examples of Eq(2. . Then the following statements are true: (i) Assume C > 0. . The following result characterizes all possible special cases of equations of the form of Eq(2.19) 1 + xn . ∞). 33 2.2. 1. . . 5. xn 1 xn−1 . 1. (See also [75]. .4. . n = 0. . and 6 respectively.4. Then A = B = γ = 0. .18) (2. [73]. 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) > 0.17) (2. n = 0. . (2. σ ∈ [0.1) xn+1 = xn+1 = xn+1 = xn+1 n = 0. xn−1 xn .) Theorem 2.16) (2.1) is periodic with period p. Then γ(α + β) = 0. = xn−1 are remarkable in the sense that all positive solutions of each of these four nontrivial equations are periodic with periods 2. .

and Invariant Intervals Proof. A+B C (2. assume that γ(α + β) > 0.1) is periodic with period p. Then by choosing x0 suﬃciently small. In addition to (2.4. 4. A + B + Cxp−2 and xp−1 = x−1 = 1 we see that (2. 2 Corollary 2. (ii) Assume C = 0 and. Otherwise A + B > 0 and by choosing x0 > max α+β γ . Then clearly x0 = xp and so from Eq(2.21) holds. Local Stability.1 Let p ∈ {2. Semicycles. we see that (2. The proof is complete. Consider the solution with x−1 = 1 and x0 ∈ (0.20) is impossible. Then up to a change of variables of the form xn = λyn Eq(2. 6}. 3.21) (2. Hence (2. Periodicity.22) Otherwise γ > 0 and by choosing x0 < min α+β γ . .16)-(2. Assume that every positive solution of Eq(2.20) is impossible. Then we claim that A = B = 0.1) reduces to one of the equations (2.19) . ∞).22) holds . (i) Assume that C > 0. for the sake of contradiction.21) we now claim that also γ = 0.20) is impossible. Thus (2.34 Chapter 2. (2. 5.1) xp = Hence (A + B)x0 + (Cx0 − γ)xp−2 = α + β. A+B C we see that (2.20) α + β + γxp−2 = x0 .

5. . Assume that . φ.27) hold . φ. .1) converges to a period-two solution in a nontrivial way according to the following convention.26) and (2.1) has a prime period-two solution if and only if γ = β + A.24) (2.25) (2.1) to have a prime period-two solution and we exhibit all prime period-two solutions of the equation. (See [50]. 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. φ. is a prime period-two solution of Eq(2.5 Existence of Prime Period-Two Solutions In this section we give necessary and suﬃcient conditions for Eq(2. is a prime period-two solution of Eq(2. ψ. 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. ψ ∈ [0. . . .1) if and only if α + β(φ + ψ) = Bφψ with φ.2. C (2.23) α + βψ + γφ A + Bψ + Cφ α + βφ + γψ A + Bφ + Cψ . ψ. Furthermore when (2. .1). .27) (2. Existence of Prime Period-Two Solutions 35 2.26) αC + β(γ − β − A) . ∞) and φ = ψ (2. . not necessarily prime. . . . . ψ. ψ. solution with period p and furthermore the set of solutions of the equation with prime period p is nonempty. φ.) Furthermore we are interested in conditions under which every solution of Eq(2.

α + β > 0. C C with 0 ≤ φ < γ−A . .1). 0. . . and Invariant Intervals or equivalently β φ> . C C Finally when (2. γ−A γ−A − φ. B and ψ = α + βφ . . . . . . ψ. − φ. φ. (2. Local Stability. B > C. . Periodicity. . C γ−A γ−A . the two cycle is .23) and (2.31) holds.. . B = C.32) 4C 2 2. .28) Next assume that (2.31) γ > β + A and When (2.. .24) holds and that Eq(2. 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. . α = β = 0. .. and φψ > 0. φ.6 Let Local Asymptotic Stability of a Two Cycle . φ. . Semicycles.1) possesses a prime period-two solution .25) that one of the following three conditions must be satisﬁed: B = C. ψ. Set un = xn−1 and vn = xn for n = 0.30) holds. and φψ = 0. . ψ. and φψ ≥ 0. φ.29) (2. .29) holds. the two cycle is . . When (2. φ.30) (2. α = β = 0. . 1. . That is. .36 Chapter 2. . . B φ= β+ √ β 2 + αB . be a two cycle of the diﬀerence equation (2. . . 0. −β + Bφ (2. (2. ψ. φ. . (γ − β − A)[B(γ − β − A) − C(γ + 3β − A)] α< . Then it follows from (2. .

∂u (A + Bψ + Cφ)2 ∂g (βA − αB) − (γB − βC)φ (φ. ψ) = . 1. ψ ∂h ∂h (φ. ψ) = . n = 0.6. φ ψ is a ﬁxed point of T 2 . . v) = α + βv + γu A + Bv + Cu and h(u. Local Asymptotic Stability of a Two Cycle and write Eq(2.2. v) α+βv+γu α + β A+Bv+Cu + γv α+βv+γu A + B A+Bv+Cu + Cv . Furthermore T2 where g(u. ψ) ∂v (φ. ∂u (A + Bφ + Cψ)2 (A + Bψ + Cφ)2 . ∞) deﬁned by: T Then u v = v βv+γu+α Bv+Cu+A 37 . evaluated at ψ By deﬁnition ∂g ∂g (φ. . v) h(u. ψ) ∂u we ﬁnd the following identities: (γA − αC) + (γB − βC)ψ ∂g (φ. 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.1) in the equivalent form: un+1 = vn α+βv vn+1 = A+Bvn +γun . n +Cun Let T be the function on [0. JT 2 . the second iterate of T . ψ) = . . ∂v (A + Bψ + Cφ)2 [(βA − αB) − (γB − βC)ψ][(γA − αC) + (γB − βC)ψ] ∂h (φ. ψ) ∂v (φ. . v) = u v = g(u. ψ) ∂u φ = JT 2 . ∞) × [0.

Semicycles. A.6. . .33) . ∞) with α + β + γ. . A + B ∈ (0. ψ) = ∂v (A + Bφ + Cψ)2 (A + Bψ + Cφ)2 (γA − αC) + (γB − βC)φ .7. . Periodicity. ψ) + (φ. be a two cycle of Eq(2.1). . and Invariant Intervals [(βA − αB) − (γB − βC)ψ][(βA − αB) − (γB − βC)φ] ∂h (φ. x0 ) through which the solution of Eq(2. 1. A + Bxn In this section we consider the equation xn+1 = where α. See Sections 6. ψ) − (φ.1) converges to the two cycle. . .38 Chapter 2. Deﬁnition 2. That is. 6. The basin of attraction of this two cycle is the set B of all initial conditions (x−1 . ∞) and nonnegative initial conditions and obtain a signum invariant for its solutions which will be used in the sequel (See Sections 4. 6. if and only if |S| < 1 + D and D < 1.1 (Basin of Attraction of a Two Cycle) Let . ψ. A and B so that every solution of Eq(2. φ.9.7. γ. ψ. β.1. β.6.7 Convergence to Period-Two Solutions When C=0 α + βxn + γxn−1 .5. Local Stability. γ.4 for some applications. ψ). .33) converges to a period-two solution.1 (c) that both eigenvalues of JT 2 unit disk |λ| < 1. (2. every solution converges to a (not necessarily prime) period-two solution and n = 0. ψ) ∂u ∂v ∂h ∂g ∂h ∂g (φ. 6. and 7. and 10. B ∈ [0. ∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1. . 2.2) to obtain conditions on α. ψ) (φ. φ. + (A + Bφ + Cψ)2 Set S= and D= ∂h ∂g (φ. ψ) (φ.

the solution would converge to a period-two solution. Set (2. Clearly (i) holds. 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. the sign of the quantity Jn is constant along each solution. In particular. (iii) xn+1 − xn−1 > 0 for all n ≥ 0.34) Jn = α + β(xn−1 + xn ) − Bxn−1 xn for n ≥ 0. Statement (ii) holds if and only if J0 < 0 and β +A > 0. In this case the subsequences {x2n }∞ and {x2n−1 }∞ of the solution {xn }∞ are both decreasing to ﬁnite limits and n=0 n=0 n=0 so in this case the solution converges to a period-two solution.2. Then the following statements are true: (a) β+A Jn for n ≥ 0.2 where we present the complete character of solutions of Eq(2. if and only if J0 > 0 and β + A > 0. (β + A)J0 = 0. the solution {xn } is periodic with period-two.1 Assume that (2.27) holds. one and only one of the following three statements is true for each solution of Eq(2. ∞ Lemma 2. and let {xn }n=−1 be a solution of Eq(2. Jn+1 = (b) xn+1 − xn−1 = Jn A + Bxn for n ≥ 0.) Now before we see that when B > 0 every solution of Eq(2. we need the following result.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 . In this case.7.33).33) converges to a periodtwo solution if and only if (2. Finally (iii) holds. Convergence to Period-Two Solutions When C=0 39 there exist prime period-two solutions. (ii) xn+1 − xn−1 < 0 for all n ≥ 0. A + Bxn In particular. (See also Section 10. the proof of which follows by straightforward computations. if and only if.7.33).27) holds.

+ |x2N +2 − x2N | + x2N m for m = 1. A + Bx2n+1 A + Bx2n A + Bx2N +1 A + Bx2N Set Then K = |x2N − x2N −2 |. Then the following statements hold: (a) Eq(2. Then there exists N ≥ 0 such that ρ= Clearly for all n > N . 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. . .27) holds. φ. ψ. To this end assume. A + Bxk k=1 (2. A+β A+β A+β A+β < = ρ.33) are given by (2.1 Assume B > 0. Local Stability. 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 | + .40 and so Chapter 2.35) We will now employ (2.28).33) converges to a period-two solution. of Eq(2. that the subsequence of even terms increases to ∞. |x2N +2 − x2N | = A+β A+β < 1.7. Periodicity. . Semicycles. Theorem 2. . for the sake of contradiction. . and Invariant Intervals n xn+1 − xn−1 = (x1 − x−1 ) β+A . In summary we have established the following result.33) has a prime period-two solution if and only if (2. . . φ. . (b) When (2. ψ. . (c) When (2.27) holds all prime period-two solutions . . 2.35) to show that the subsequences of even and odd terms of the solution are bounded.27) holds every solution of Eq(2. We will give the details for the subsequence of even terms. Kρ + x2N 1−ρ . . .

In other words I is an invariant interval for the diﬀerence equation xn+1 = f (xn . then xn ∈ I for every n > N . 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 . (2. fx < 0 if y > − M fy > 0 fx > 0 and fy < 0 fx > 0 and fy < 0 fx < 0 and fy > 0 fx < 0 and fy > 0 L L fx > 0 if y > − M . we obtain the following table of invariant intervals for Eq(2. The following table gives the signs of fx and fy in all possible nondegenerate cases. xN ∈ I for some N ≥ 0.8 Invariant Intervals An invariant interval for Eq(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. n = 0.10) and are fx (x. 1.2. Invariant intervals for Eq(2.36) if xN −1 . For Eq(2. . xn−1 ). (A + Bx + Cy)2 − + − − − − − − Using this table. . Invariant Intervals 41 2. y) = where L = βA − αB. Signs of derivatives fx and fy fx > 0 N N fy > 0 if x < M . y) = N − Mx .1).36) are determined from the intervals where the function f (x.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 . and N = γA − αC.1) the partial derivatives are given by (2. fy < 0 if x < M L + My (A + Bx + Cy)2 and fy (x.9) and (2.8. . y) is monotonic in its arguments.

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

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. 1. In particular address the case p = 2. . . if and only if every root of the characteristic equation is a pth root of unity. ∞)] and let p ≥ 2 be an integer. (2.37) is periodic with period p ≥ 2. .3 Let p ≥ 2 be an integer and assume that every positive solution of Eq(2.9. 1.37) 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. Open Problem 2. 4. xn−1 ). ∞)] is such that every positive solution of the equation xn+1 = f (xn . β. A.7) that every positive solution of each of the following three equations xn+1 = 1 + xn+1 = xn+1 xn−1 . B. . every solution is periodic with period p ≥ 2.2.9. (2. 4. γ.1) with α.38) (2. ∞). xn n = 0.9. ∞) × (0. n = 0. 6}? Open Problem 2.7.39) (2. . . Is it true that the linearized equation about a positive equilibrium point of Eq(2. (0.1 Assume that f ∈ C 1 [(0. and 6. = 1 + xn .2 Let f ∈ C 1 [[0. Find necessary and suﬃcient conditions so that every nonnegative solution of the diﬀerence equation (2. ∞) converges to a solution with period p . 1.37) converges to a periodic solution with period p.40) 1 + xn−1 . C ∈ [0. Is it true that p ∈ {2. . 6. 1 + xn xn + 2xn−1 .9.4 It is known (see Sections 2. . . . 5. n = 0. . [0. ∞) × [0.7.9. ∞). n = 0. .5. Open Problems and Conjectures 43 2. 1.

A. β. γ. Open Problem 2. Semicycles. .7. determine all initial conditions (x−1 . ψ. C ∈ (0. . .2 Assume α.38) or Eq(2. ∞). ∞) and that Eq(2. ∞) for which the solution {xn }∞ n=−1 converges to the period-two solution (2. β. γ. B. Periodicity.5. Local Stability.1) is a saddle point.7. and Invariant Intervals converges to a solution with (not necessarily prime) period-two: . determine φ and ψ in terms of the initial conditions x−1 and x0 .41).41) In each case.44 Chapter 2. A. β. if (2.9. and 10. . γ.1) has a positive prime period-two solution. A.5 Assume α. 6. ∞) × (0. ψ. B. .9. .33) is bounded.2. φ.39) or Eq(2. B ∈ (0. (2. γ. φ. 6. B.41) is a period-two solution of Eq(2.9. A.1) has no period-two solution. Show that the positive equilibrium of Eq(2. Show that the positive equilibrium is globally asymptotically stable.40). Conversely. Show that every positive solution of Eq(2.7. x0 ) ∈ (0. . Conjecture 2. β. C ∈ (0. 4. . Determine the set of initial conditions x−1 and x0 for which the solution {xn }∞ n=−1 of Eq(2.) Conjecture 2. C ∈ (0.9. ∞) and that Eq(2.3 Assume α. Conjecture 2. (See Sections 2.1 Assume α.1) is bounded. ∞) and γ > β + A.

7 Assume f ∈ C[(0. β. Open Problem 2. Conjecture 2. ∞)].9 Extend Theorem 2.6 Assume α.9. (See Section 5. f (xn ) . A.1) is bounded. 1. 1. β. Open Problem 2. C ∈ [0. . .2). . ∞). Obtain necessary and suﬃcient conditions on f for every positive solution of the equation xn+1 = 1 + to converge to a period-two solution. .9.7). C ∈ [0.1) cannot have a positive prime two cycle which attracts all positive non-equilibrium solutions. (0. . xn−1 n = 0. γ. . . (See Section 4.4 Assume that α. .2.9. . f (xn−1 ) . Open Problems and Conjectures Open Problem 2. 1. A. γ.1 to third order diﬀerence equations xn+1 = α + βxn + γxn−1 + δxn−2 .9. B. 45 Obtain necessary and suﬃcient conditions in terms of the coeﬃcients so that every positive solution of Eq(2.9.7. B. A + Bxn−1 + Dxn−2 n = 0. xn n = 0. ∞).9.8 Assume f ∈ C[(0. with nonnegative parameters and nonnegative initial conditions such that γ = β + δ + A. ∞)]. Show that Eq(2. (0. Obtain necessary and suﬃcient conditions on f for every solution of the diﬀerence equation xn+1 = to be bounded. ∞). ∞). What additional conditions should the parameters satisfy? . Open Problem 2.

. . 1. n = 0. .42) to be a repeller.1. xn−1 . and Invariant Intervals Open Problem 2. xn−1 . .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 . or a saddle point. xn−3 ).10 Extend the linearized stability Theorem 1. Semicycles. .1 to third order difference equations xn+1 = f (xn . But how about necessary and suﬃcient conditions for the equilibrium of Eq(2.46 Chapter 2.9. . 1. . xn−2 . . Local Stability. . Periodicity. xn−2 ). (2. n = 0. or nonhyperbolic? Extend these results to fourth order diﬀerence equation xn+1 = f (xn .

.1 Introduction α . (3. . 1)-Type Equations 3.3) xn+1 = βxn .6) . . 1. . 1. . 1)-type: xn+1 = n = 0.4) xn+1 = β . . .2) xn+1 = α . A n = 0. 1. . 1. . . . (3.1) xn+1 = α . . A Eq(1) contains the following nine equations of the (1. . (3. B n = 0. (3. Cxn−1 n = 0. . . . Bxn n = 0. 1.Chapter 3 (1. 1. . (3. . (3.5) xn+1 = βxn . Cxn−1 47 n = 0.

. and (3.8).8) xn+1 = γ .4) and (3. y0 . . The remaining two equations. . 1. .. (3.7) xn+1 = and γxn−1 .10) has prime period six. 1.9) are trivial. 3.6) and (3. . . . 1)-type Eq(3.48 Chapter 3. . Eqs(3. y−1 y−1 y0 y0 It is interesting to note that except for the equilibrium solution y = 1. Indeed the solution with initial conditions y−1 and y0 is the six-cycle: y−1 .4) is periodic with period four. yn−1 n = 0. .. . (3. . (1. (3.2) is periodic with period two and every solution of Eq(3.10) It is easy to see that every positive solution of Eq(3. 1)-Type Equations xn+1 = γxn−1 . .9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive.10) is periodic with period six. Every solution of Eq(3. . . 1.2 The Case α = γ = A = B = 0 : xn+1 = β yn C βxn Cxn−1 This is the (1. and the denominators are always positive. the initial conditions are nonnegative. . Eqs(3. . Bxn n = 0. (3. A n = 0. are treated in the next two sections.1).. . Of these nine equations.6) which by the change of variables xn = reduces to the equation yn+1 = yn . .7) are linear diﬀerence equation. every other solution of Eq(3. 1. namely. Eqs (3. C n = 0. y0 1 1 y−1 .5). . (3.

.2) and (3.7) is periodic with period two but this is a linear equation. n = 0. . (3. ∞)]. . n = 0.3. . The Case α = β = A = C = 0 : xn+1 = γxn−1 Bxn 49 γxn−1 Bxn 3. xn−1 ). .12) . 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.4 Open Problems and Conjectures Of the nine equations in this chapter we would like to single out the two nonlinear equations. 1. ∞) × (0.3 The Case α = β = A = C = 0 : xn+1 = γ yn e B This is the (1. Eq(3. .3. (3. namely two and six. where f ∈ C[(0.11) is yn = C1 −1 + 2 √ 5 n n = 0. (3. ∞). 1)-type Eq(3. respectively. ∞)]. (3.2).6) is of the form xn+1 = f (xn . where f ∈ C[(0. Eq(3. 1. from which the behavior of solutions is easily derived. . 1. Also every solution of Eq(3.6).2) is of the form xn+1 = f (xn ). .8) which by the change of variables xn = reduces to the linear equation yn+1 + yn − yn−1 = 0. . (0.3) also has the property that every nontrivial positive solution is periodic with prime period four but this equation is really a variant of Eq(3. (3.14) (3. . . which possess the property that every positive nontrivial solution of each of these two equations is periodic with the same prime period. ∞). (0. while Eq(3.11) + C2 −1 − 2 √ 5 n . .13) n = 0. 1. The general solution of Eq(3.

2 Let k ≥ 2 be a positive integer and assume that (3. [17]. In particular address the cases k = 4. xn−1 n = 0. ∞)] be such that every positive nontrivial solution of the diﬀerence equation (3. (3. β. ∞).4. 1. with real initial conditions is periodic with period k.4. Obtain necessary and suﬃcient conditions on α. (1. Obtain necessary and suﬃcient conditions on f so that every positive solution of Eq(3. . 1.14) is periodic with period k. γ so that every solution of the equation xn+1 = α |xn | + βxn−1 + γ.50 Chapter 3. n = 0.4.2 Let f ∈ C 1 [[0.16) is periodic with prime period six.) Conjecture 3. Conjecture 3. ∞).4. Obtain necessary and suﬃcient conditions on f so that every positive solution of Eq(3. Show that f (x) = x.3 Let k be a positive integer and assume that α. 9. . . . . 6. ∞). k = 5. 5.15) holds. ∞)] be such that every positive nontrivial solution of the diﬀerence equation xn+1 = f (xn ) . Open Problem 3. 6.16) is periodic with prime period ﬁve. β.1 Let f ∈ C 1 [[0. 8. In particular address the cases: (See [10]. Show that f (x) = 1 + x. and [59].4. 1)-Type Equations Open Problem 3. . γ ∈ (−∞. [0. .13) holds.12) is periodic with period k. 7. [0. Open Problem 3.1 Let k ≥ 2 be a positive integer and assume that (3.

n=0 (iii) {pn }∞ is periodic with prime period k ≥ 2. .4.3.4 Obtain necessary and suﬃcient conditions on f ∈ C[[0.4.18) xn−1 ∞ (i) {pn }n=0 converges to a ﬁnite limit. f (xn ) .5 Let {pn }∞ be a sequence of nonzero real numbers. (3. 1. . (ii) {pn }∞ converges to ±∞. 1. ∞). . [0. . . xn−1 n = 0.17) Open Problem 3.4. (3. n=0 . 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 = . ∞)] such that every positive nontrivial solution of the equation xn+1 = is periodic with prime period three. . Open Problems and Conjectures 51 Open Problem 3. . n = 0.

.

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

then the subsequences {x2n−1 } and {x2n } satisfy the Riccati equation γyn n = 0. Similarly. . 1.8) are essentially Riccati equations. . . 2)-Type Equations xn+1 = γxn−1 . (4. namely Eqs(4. . if {xn } is a solution of Eq(4. 1.1). Indeed if {xn } is a solution of Eq(4.8). . . (See Section 1. (4. 1. (4. from which the global behavior of solutions is easily derived. .8) γxn−1 .2).1) and (4. then the subsequences {x2n−1 } and {x2n } satisfy the Riccati equation of the form of Eq(4. Two of these equations. β β n = 0.7) xn+1 = and xn+1 = γxn−1 . .4) are Riccati-type diﬀerence equations. (4. Eqs(4.9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. . In view of the above.1) which by the change of variables xn = . A + Bxn n = 0. Bxn + Cxn−1 n = 0. the initial conditions are nonnegative. . . Also Eqs(4. . 4. yn+1 = A + Cyn It is also interesting to note that the change of variables xn = 1 yn reduces the Riccati equation (4. . .54 Chapter 4.2) and (4. A + Cxn−1 n = 0.2). .4). 1. and (4.2 The Case β = γ = C = 0 : xn+1 = A yn B α A+Bxn This is the (1. 2)-type Eq(4. . . and the denominators are always positive. (1.8) will not be discussed any further in this chapter.4) to the linear equation yn+1 = A B yn + .6 in the Preliminary Results). 1.

2.10) is globally asymptotically stable.1 Every solution of Eq(4. 4.12) is bounded and persists.12) is bounded and then we will use the method of “limiting solutions” to show that every solution of Eq(4. . . That is.3.6) is R = −p < 0 and so the following result is a consequence of Theorem 1. A2 The Riccati number associated with this equation (see Section 1.3 The Case β = γ = A = 0 : xn+1 = α Bxn +Cxn−1 This is the (1. .12) Eq(4. Theorem 4.12) converges to √ its equilibrium B + C. .6. 1.10) αB .11) (4.2. The Case β = γ = A = 0 : xn+1 = reduces to the Riccati equation yn+1 = where α Bxn +Cxn−1 55 p . So if the theorem is false there should exist a ∞ solution {yn }n=−1 which is neither bounded from above nor from below.3) which by the change of variables √ α xn = yn is transformed to the diﬀerence equation yn+1 = C B + . (4.3. . It is easy to see that if a solution of Eq(4.1 The positive equilibrium y= −1 + √ 2 1 + 4p of Eq(4. .12) was investigated in [65].4. lim sup yn = ∞ and n→∞ lim inf yn = 0. Proof. 2)-type Eq(4. (4. . n→∞ . Theorem 4. (See also [12].) We will ﬁrst show that every solution of Eq(4. 1 + yn p= n = 0. 1.12) is bounded from above then it is also bounded from below and vice-versa. yn yn−1 n = 0.

j − 1}.12). . let {yn }∞ n=−1 be a solution of Eq(4. . Then by Theorem 4.3. yn yn−1 n = −3.1 .1. B C B+C + > yj−1 yj−2 yi C B+C B + ≤ . Theorem 4. and S = lim sup yn . yi−1 yi−2 yj Proof. . 2 √ B + C of Eq(4. . S0 = S B C + . The linearized equation of Eq(4.1.3. n→∞ for n ≥ −1. . .12) about the equilibrium y = zn+1 = − B C zn − zn−1 . . {yn }n=−1 is bounded and persists. . B + C < yi yj ≤ B + C which is impossible. B+C B+C n = 0. . .12) is globally asymptotically √ B + C is (4. 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.13) for all n ∈ {−1.56 Chapter 4. and by Theorem 1. 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 .12). 1. −2.2 The equilibrium y = stable. .5) there exist two solutions {In }∞ n=−3 and {Sn }∞ n=−3 of the diﬀerence equation yn+1 = with the following properties: I0 = I. y is locally asymptotically stable for all positive values of B and C. (1. To ∞ this end. By the theory of limiting solutions (see Section 1. . It remains to be shown that y is a global attractor of all solutions of Eq(4.

. Sn ∈ [I. S] for n ≥ −3.4 The Case α = γ = B = 0 : xn+1 = Equation A yn .14) from which it follows that (4. (4. A n = 0. 2)-type Eq(4. 1.5) which by the change of variables xn = reduces to Pielou’s diﬀerence equation yn+1 = where p= pyn .15) we conclude that S=I and the proof is complete. . (4. S= B C B+C + ≤ S−1 S−2 I C B+C B + ≥ .14) and (4. I−1 I−2 S B + C = SI. From (4. Therefore B C B C + = I = S−1 = + S S S−2 S−3 S = S−2 .Pielou’s This is the (1. 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.16) . .Pielou’s Equation 57 In . The Case α = γ = B = 0 : xn+1 = and Then βxn A+Cxn−1 . 1 + yn−1 β .15) 2 4. C βxn A+Cxn−1 .4.4.

5 The Case α = γ = A = 0 : xn+1 = β . So the interesting case for Eq(4. (4. B + Cyn βxn Bxn +Cxn−1 This is the (1.1 (a) Assume p ≤ 1.16) that every positive solution converges to 0.17) In this case the zero equilibrium of Eq(4. x is globally asymptotically stable. Then the positive equilibrium y = p − 1 of Eq(4. Theorem 4. Furthermore. (1.16) possesses the unique positive equilibrium y = p − 1. We summarize the above discussion in the following theorem. . 1. which is locally asymptotically stable. In fact by employing Theorem 1.2. t≥0 dt P which is a prototype of modelling the dynamics of single-species.) When p ≤ 1. Eq(4.17) holds.79) as a discrete analogue of the delay logistic equation dN N (t − τ ) = rN (t) 1 − . .16) was investigated in [55].18) . p. Then the zero equilibrium of Eq(4. it follows from Eq(4.16) is globally asymptotically stable.58 Chapter 4. ∞) and p > 1. p. (See also ([42]. 4.16) is globally asymptotically stable.6) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = p + yn . (b) Assume y0 ∈ (0. (4.22) and ([67]. .75).16) is when p > 1. 2)-Type Equations This equation was proposed by Pielou in her books ([66].4. 2)-type Eq(4. 0 is locally asymptotically stable when p ≤ 1 and unstable when p > 1. p.4. p + yn−1 n = 0.16) is unstable and Eq(4. it follows that when (4.

19) has semicycles of length one. ∞). p.1. (b) Assume p < 1. (4. In fact.1 (e).19) and it is locally asymptotically stable. Then 0 and y = 1 − p are the only equilibrium points of Eq(4. except possibly for the ﬁrst semicycle. 1. The oscillatory character of the solutions of Eq(4.19) and they are both unstable.4.1. 2)-type Eq(4. That is.6 The Case α = β = C = 0 : xn+1 = γ yn .6.7) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= yn−1 .4. γ Eq(4. Lemma 4. 0 is a repeller and y = 1 − p is a saddle point equilibrium. 73) where it was shown that its equilibrium one is globally asymptotically stable .19) is a consequence of Theorem 1.6. B γxn−1 A+Bxn This is the (1. . . p + yn n = 0.5) that 1 yn+1 < yn−1 p and so when p>1 for n ≥ 0 .19) was investigated in [28]. Then 0 is the only equilibrium point of Eq(4. p= 4. .1.1 (a) Assume p > 1. The following local result is a straightforward application of the linearized stability Theorem 1.18) was investigated in ([42] Corollary 3. The Case α = β = C = 0 : xn+1 = where γxn−1 A+Bxn 59 B . It follows directly from Eq(4.7.19) A ∈ (0.19) (see also Section 2. every solution of Eq(4. C Eq(4.

4. φ. 2)-type Eq(4. 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. . . . . .19) is globally asymptotically stable. n = 0. yn n = 0. yn+1 < yn−1 for n ≥ 0 and so every positive solution of Eq(4. . . B This equation was investigated in [7] where it was shown that the following statements are true: p= . . (1. It follows from the identity (1 − p) − yn yn−1 .19) converges to a period-two solution . . . φ. (4. 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 0.60 Chapter 4. ψ.7 The Case α = β = A = 0 : xn+1 = γ Byn γxn−1 Bxn +Cxn−1 This is the (1. . . On the other hand when p = 1. 1. ψ.20) C ∈ (0.9) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = p + where yn−1 . Furthermore we can see that φψ = 0 but whether there exists solutions with φ=ψ=0 remains an open question. 2)-Type Equations the zero equilibrium of Eq(4. 1. . ∞).

7. ∞ (b) Let {yn }n=−1 be a solution of Eq(4. Moreover. n→∞ lim 2. The next result about semicycles is an immediate consequence of some straightforward arguments and Theorem 1. In fact the following result is true.2 (a) Let {yn }∞ n=−1 be a solution of Eq(4. 1] and y2 ≥ p + y0 and use induction to complete the proof. Then {yn }∞ n=−1 converges monotonically to y = p + 1.20) is y = p + 1.20) is globally asymptotically stable. (ii) When p = 1. . . . p+1 p+1 n = 0. Then the following statements are true: 1. n→∞ y2n+1 = p. every semicycle has length one. and so y0 > p + 1. every solution of Eq(4.20) which are unbounded. 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(4.20) converges to a period-two solution. We now proceed to establish the above results.20) is locally asymptotically stable when p > 1 and is an unstable saddle point when p < 1. and let {yn }∞ n=−1 be a solution of Eq(4.1.20) to be bounded.20) which consists of a single semicycle.1. Proof.7.1 The Case p < 1 Here we show that there exist solutions of Eq(4.4. .7. lim y2n = ∞.20) about the equilibrium point y = p + 1 is zn+1 + 1 1 zn − zn−1 = 0.7. (4.7. Clearly the only equilibrium point of Eq(4.21) As a simple consequence of the linearized stability Theorem 1.7. . It suﬃces to show that y1 ∈ (p.20) such that 0 < 1 y−1 ≤ 1 and y0 ≥ 1−p . the equilibrium y = p + 1 of Eq(4. with the possible exception of the ﬁrst semicycle. Note that 1 1−p > p + 1.3 Let p < 1.20) which consists of at least two semicycles.1 we obtain the following result: Theorem 4. Theorem 4. The linearized equation of Eq(4. Theorem 4. 1. 4. Then {yn }∞ n=−1 is oscillatory.1 The equilibrium point y = p + 1 of Eq(4. (iii) When p > 1.

Then p+ p2 p−1 ≤ lim inf yn ≤ lim sup yn ≤ .4 Let p = 1. Note that for n ≥ 0. p2 We shall ﬁrst show that lim supn→∞ yn ≤ p−1 .2 The Case p = 1 The following result follows from Lemma 2. See also Section 2. . 2)-Type Equations > p. It follows from Theorem 4. Then {yn }n=−1 converges to a prime period-two solution of Eq(4.7. Then {yn }∞ n=−1 converges monotonically to y = 2.5.20) is globally asymptotically stable.20).62 Indeed y1 = p + y1 y0 Chapter 4.7. n→∞ p p−1 n→∞ Proof. and that p < y0 < p + 1 < y−1 . y1 = p + y−1 1 ≤p+ ≤ 1. (c) Every solution of Eq(4.7.20) converges to a period-two solution if and only if p = 1. .20).7. p m = 0. (1.1 and some straightforward arguments. Then the following statements are true.20). Also. ∞ (b) Suppose {yn }∞ n=−1 consists of at least two semicycles.1 Let p > 1.2 that we may assume that every semicycle of {yn }∞ n=−1 has length one. 4. and let {yn }∞ n=−1 be a solution of Eq(4. and let {yn }∞ n=−1 be a solution of Eq(4. p . 4. Theorem 4. y0 y0 2 and so y1 ∈ (p. Lemma 4. .7.7. that p < yn for all n ≥ −1. 1. 1]. y2n+1 < p + So as every solution of the diﬀerence equation 1 ym+1 = p + ym .3 The Case p > 1 Here we show that the equilibrium point y = p + 1 of Eq(4. The following result will be useful. ∞ (a) Suppose {yn }n=−1 consists of a single semicycle. y2n−1 . Hence y2 = p + y0 y1 ≥ p + y0 .

5 Let p > 1. So let {yn }∞ n=−1 be a solution of Eq(4. Hence p < f (x. y2n−1 p + p2 + is arbitrary. (0. ∞). y) = p + . and set a = p and b = Note that f and p2 + . ∞) for each y ∈ (0. we have p−1 p3 + p(p − 1) =p+ .1 that y = p + 1 is a locally asymptotically stable equilibrium point of Eq(4. y ∈ (0.20). Proof. Let p N ≥ 0 such that for all n ≥ N . it follows that lim supn→∞ yn ≤ p−1 . Then the equilibrium y = p + 1 is globally asymptotically stable.7. The Case α = β = A = 0 : xn+1 = 2 γxn−1 Bxn +Cxn−1 2 63 p p converges to p−1 . ∞)] . y ∈ p. It suﬃces to show that lim yn = p + 1. y) < for all x. We know by Theorem 4.4. and f is increasing in y ∈ (0. p−1 Let n ≥ N .20). p2 − p p−1 p2 + . ∞) × (0. lim inf yn ≥ 2 n→∞ p p 2 We are now ready for the following result. p−1 For x. ∞). We shall next show that p + p−1 ≤ lim inf n→∞ yn .7.20) with prime period two. Let > 0.4. p2 + y2n−1 < . There clearly exists y2n = p + So as p−1 p3 + p + p(p − 1) y2n−2 >p+p 2 = . ∞). f is decreasing in x ∈ (0. ∞) for each x ∈ (0. x Then f ∈ C[(0. Recall that by Theorem 4. Theorem 4. n→∞ y f (x. set p−1 p2 + >p . . ∞).p = p + p 2 p−1 p + p2 + p−1 p2 + p−1 = p2 + p3 + p = . p−1 f p.7. Then > 0. there exist no solutions of Eq(4.7.

B ∈ (0.2 (see [18]) Assume that A. . B ∈ (0. (b) Assume that the initial conditions y−1 .22) converges. . . p<p+ Chapter 4. .4. .23) is well deﬁned for all n ≥ 0. .} with k < l. y0 with y−1 y0 < 0 for which the equation yn+1 = B A . x0 such that {xn }∞ converges to this periodic solution. ∞). p−1 p2 + p2 ≤ lim inf yn ≤ lim sup yn ≤ < n→∞ p p−1 p−1 n→∞ n→∞ lim yn = p + 1.8. [60]. . . φ1 . . . . Determine the set of all positive initial conditions x−l .8. .64 Finally note that by Lemma 4. Show that every solution of Eq(4. Investigate the global behavior of the solution {yn }. . . A+B A+B A. y0 are such that y−1 y0 < 0 and such that Eq(4. φp . .22) Conjecture 4. . + yn yn−1 (4. xn−l n = 0. 2 4. 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]. is well deﬁned for all n ≥ 0.22). (4. .1 Consider the diﬀerence equation where and where the initial conditions x−l . (1. [23].1 Let . . φ2 . .) Open Problem 4.7. . . 2)-Type Equations and so by Theorem 1.23) A B λl−k + = 0. [25]. (a) Find all initial conditions y−1 . 1. . and [61]. . be a given prime period p solution of Eq(4. ∞) and k.6. Is it bounded? Does limn→∞ yn exist? When does {yn } converge to a two cycle? . n=−l Open Problem 4.8 Open Problems and Conjectures xn+1 = A xn−k + B . .8. x0 are arbitrary positive real numbers. 1. l ∈ {0.1.

24) is well deﬁned for all n ≥ 0 and bounded.20) possesses a solution {yn }∞ n=−1 which remains above the equilibrium for all n ≥ −1. y0 ∈ (0. the asymptotic behavior. Open Problem 4.19) has a solution which converges to zero.8. (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. y0 ∈ R be such that Eq(4. ∞) ∞ for which the solution {yn }n=−1 of Eq(4.4 Determine the set G of all initial conditions (y−1 .3 Assume that p < 1. For such an initial point. Determine all initial conditions y−1 .19) is bounded. Open Problem 4. Conjecture 4. Show that Eq(4.2 Assume that p = 1.8.8. (b) Let y−1 . Open Problem 4. Show that {xn }∞ n=−1 is a three cycle.8. for these initial points. and the periodic ∞ nature of the solution {yn }n=−1 of Eq(4.8.20) is well deﬁned for all n ≥ 0.8. y0 ∈ R such that the equation (4.3 Show that Eq(4. .8.4 Assume that (x−1 . ∞).20) is well deﬁned for all n ≥ 0. Open Problem 4. x0 ) ∈ R × R through which the equation xn−1 xn+1 = −1 + (4. investigate the global character ∞ of {yn }n=−1 .5 (a) Assume p ∈ (0.6 (See [13]) Find the set G of all initial points (x−1 .8. Open Problems and Conjectures 65 Conjecture 4.4. Conjecture 4. x0 ) ∈ R × R is such that the equation (4. Determine the set of initial conditions y−1 .20).24) xn is well deﬁned for all n ≥ 0. investigate the boundedness.

.9 (See [5]) For which initial values x−1 . .8.7 Let f ∈ C[(0. . 1. 1. 1. ∞) does the sequence deﬁned by xn−1 . . . and [67]. ∞) for i = 0. .8. < 2 cos α 2k + 1 Open Problem 4.5 (See [27]) Assume r ∈ (0. .) Assume α ∈ (1. 2)-Type Equations Open Problem 4. Conjecture 4. . 1 + xn−k n = 0. x0 ∈ (0. Investigate the global asymptotic stability of the equilibrium points of the diﬀerence equation xn+1 = xn . see [55]. . Show that every positive solution of the population model Jn+1 = Jn−1 er−(Jn +Jn−1 ) . . . . . ∞)]. converges to a period-two solution. n = 0. with positive initial conditions. Obtain necessary and suﬃcient conditions on f for every positive solution of the equation xn+1 = to be bounded. a + a0 xn + . f (xn ) . 1. ai ∈ (0. k. . .6 (Pielou’s Discrete Logistic Model. with positive initial conditions is globally asymptotically stable if and only if kπ α−1 . ∞) and k ∈ {0. xn+1 = 1 + xn converges to two? . ∞). n = 0. ∞).8 Let a. . Show that the positive equilibrium of Pielou’s discrete logistic model xn+1 = αxn .8. . . (1. (0.}. Open Problem 4. .66 Chapter 4. 1.8. xn + xn−1 n = 0. + ak xn−k n = 0. [66]. . 1.8. . . . Conjecture 4. .

25) with nonnegative initial conditions.8. lim p = n→∞ pn . . . In other words. Open Problem 4. n = 0. ∞ (b) Let y−1 .4. 1. . 1. Open Problems and Conjectures (See also Section 4.11 Assume p ∈ (0. ∞ More precisely. y0 ∈ B. Investigate the asymptotic behavior and the periodic nature of all positive solutions of each of the following diﬀerence equations: yn+1 = yn+1 = 1 pn + . (a) Find the set B of all initial conditions y−1 . . . 1). ∞) such that the solutions {yn }∞ n=−1 of Eq(4. y0 ∈ [0. .25). ∞) through which the solutions of Eq(4.20) are bounded. . 67 (4. 1 + yn−1 n = 0.25) converge to zero.25). each of the subsequences {y2n }∞ and {y2n+1 }n=−1 is decreasing and n=0 lim [n→∞ y2n ][n→∞ y2n+1 ] = 0.10 (See [31]) Consider the diﬀerence equation yn+1 = yn−1 e−yn .8.27) . . . .8. Investigate the asymptotic behavior of {yn }n=−1 . .26) (4. . .8. in terms of the initial conditions y−1 and y0 of the solution. Open Problem 4.7) Open Problem 4. 1. ∞ One can easily see that every solution {yn }n=−1 of Eq(4.25) converges to a period-two solution . 0. . . (4. ﬁnd the basin of attraction of the equilibrium of Eq(4. . (b) Determine the limits of the subsequences of even and odd terms of every solution of Eq(4. y0 ∈ (0. lim (a) Find all initial points y−1 . yn yn−1 pn yn . 0.12 Let {pn }∞ be a convergent sequence of nonnegative real numn=0 bers with a ﬁnite limit. . n = 0. .

(4.34) yn+1 = yn+1 = −1 + yn−1 . .14 For each of the following diﬀerence equations determine the “good” set G ⊂ R × R of all initial conditions (y−1 . . . . 1. . . yn For those equations from the above list for which you were successful.33) (4. 2)-Type Equations yn+1 = n = 0.(4. extend your result by introducing arbitrary real parameters in the equation.30) yn+1 = yn+1 = pn + Open Problem 4. Then for every (y−1 . n = 0. 1 − yn (4.30). 1. Investigate the global character of all positive solutions of each of Eqs(4. 1. . .26) . Open Problem 4.31) yn+1 = (4. 1 − yn−1 1 + yn .8. .8.29) (4.28) (4. pn + yn−1 yn−1 . 1 − yn−1 yn−1 . yn Chapter 4.68 pn + yn .13 Let {pn }∞ be a periodic sequence of nonnegative real numbers n=0 with period k ≥ 2. y0 ) ∈ R × R through which the equation is well deﬁned for all n ≥ 0. y0 ) ∈ G. investigate the long term behavior of the solution {yn }∞ n=−1 : yn+1 = yn . . (1. . n = 0.32) (4. pn + yn yn−1 .

Cxn−1 69 n = 0. . (5. 1. . . (5. (5.3) xn+1 = α + γxn−1 . . A Eq(1) contains the following nine equations of the (2. A n = 0. 1. 1)-Type Equations 5. . . . . . Cxn−1 n = 0.2) xn+1 = α + βxn . 1.6) . 1. . . .5) xn+1 = α + γxn−1 . 1. (5. Bxn n = 0. 1)-type: xn+1 = n = 0.1 Introduction α + βxn . .1) xn+1 = α + βxn . . (5. . .Chapter 5 (2. Bxn n = 0. . (5. . 1.4) xn+1 = α + γxn−1 .

Of these nine equations. γ Finally the change of variables xn = γ β + yn . . yn−1 α+βxn Cxn−1 . Eq(5. (5.7) xn+1 = and xn+1 = βxn + γxn−1 .9) to Eq(4. (5. which was investigated in Section 4. Therefore there remain Eqs(5. 1. 1. (5. . . B reduces Eq(5. The change of variables γ xn = yn .6) is essentially like Eq(5. A n = 0.20) with p = β . Bxn n = 0. . 1. which was investigated in Section 4.18) with p = β .5.2) is a Riccati equation and Eq(5. . (5.8) βxn + γxn−1 . 1)-type Eq(5.1). . .3) and (5. C C γ reduces Eq(5. . Cxn−1 n = 0.4) and (5.2 The Case γ = A = B = 0 : xn+1 = Equation p + yn .9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. 5. . 1.Lyness’ This is the (2. .70 Chapter 5. . . (2.5).3) which by the change of variables reduces to the equation yn+1 = n = 0.8) to Eq(4. Eqs(5. . the initial conditions are nonnegative. 1)-Type Equations xn+1 = βxn + γxn−1 . and the denominators are always positive.7) are linear.7.2).10) . (5. which will be investigated in the next two sections.

p.10) is stable but not asymptotically stable. . . It was also shown in [44] by using KAM theory that the positive equilibrium y of Eq(5.10) where αC .10) are true: (a) The absolute extreme in a semicycle occurs in the ﬁrst or in the second term. See [11]. after the ﬁrst one.2. 1 + y0 1 + y−1 + y0 1 + y−1 . and [72] and the references cited therein. . y0 . It was shown in [30] that no nontrivial solution of Eq(5.. .Lyness’ Equation 71 p= The special case of Eq(5. the equation becomes 1 + yn yn+1 = .10) has a limit.12) from which it follows that every solution of Eq(5. . Then the following statements about the positive solutions of Eq(5.5.11) with initial conditions y−1 and y0 is the ﬁve-cycle: y−1 . .11) yn−1 every solution of which is periodic with period ﬁve. (b) Every nontrivial semicycle. contains at least two and at most three terms. Concerning the semicycles of solutions of Eq(5. 131) for the history of the problem. 1. (5.1 Assume that p > 0. [71]. There is substantial literature on Lyness’ Equation. [43]. .. This result was also established in [49] by using a Lyapunov function.10) the following result was established in [43]: Theorem 5. Indeed the solution of Eq(5.) In this special case. (See ([42]..10) is bounded from above and from below by positive constants. [57]-[59]. n = 0. β2 p=1 was discovered by Lyness in 1942 while he was working on a problem in Number Theory. The Case γ = A = B = 0 : xn+1 = where α+βxn Cxn−1 . y−1 y−1 y0 y0 Eq(5.2. [42].10) possesses the invariant In = (p + yn−1 + yn ) 1 + 1 yn−1 1+ 1 yn = constant (5.

13) αB . By Theorem 1. 1)-Type Equations 5.13) is a consequence of the identity yn+1 − yn−1 = yn−1 − yn−2 yn for n ≥ 1.13) converges monotonically to the equilibrium y.72 Chapter 5. α+γxn−1 Bxn This is the (2. The global character of solutions of Eq(5.7. 1)-type Eq(5. Open Problem 5.3 The Case β = A = C = 0 : xn+1 = γ yn B n = 0. Concerning the nonoscillatory solutions 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 {y2n } and {y2n+1 } converge monotonically one to zero and other to ∞. one can easily see that every solution of Lyness’ Eq(5.12). 5. (2.4.4 Open Problems and Conjectures By using the invariant (5. By the linearized stability Theorem 1.10) is bounded without using the invariant (5. .13) is a saddle point. γ2 Eq(5. 1.12).13) one can easily see that every nonoscillatory solution of Eq(5.5) which by the change of variables xn = reduces to the equation yn+1 = where p= p + yn−1 . From this it follows that a solution of Eq(5.1 one can see that except possibly for the ﬁrst semicycle.10) without using essentially the invariant. To this day we have not found a proof for the boundedness of solutions of Eq(5.10) is bounded from above and from below by positive numbers.1 Show that every solution of Lyness’ Eq(5.1 one can see that the unique equilibrium point √ 1 + 1 + 4p y= 2 of Eq(5.13) has semicycles of length one. .13) was investigated in [28]. . yn (5.1. . every oscillatory solution of Eq(5.

(0. . ∞)] and that the diﬀerence equation (5. (0. (5.14) possesses a (nontrivial) invariant. . we oﬀer the following open problems and conjectures.4. Obtain necessary and suﬃcient conditions in terms of f so that the diﬀerence equation (5. To illustrate.14) about x does not have both eigenvalues in |λ| < 1 and does not have both eigenvalues in |λ| > 1. xn−1 n = 0.5. ∞). ∞) through which Eq(5.15) are unbounded is a set of measure zero.4.2 Let G be the set of initial conditions deﬁned in the Open problem 5.4. ∞)]. 1.3 Assume that f ∈ C[(0.15) is well deﬁned for all n ≥ 0. Obtain necessary and suﬃcient conditions in terms of f so that every positive solution of the diﬀerence equation xn+1 = is bounded. Open Problems and Conjectures 73 Open Problem 5. . Show that the linearized equation of Eq(5.4. f (xn ) .2 Assume that f ∈ C[(0. we oﬀer the following open problems and conjectures.14) Open Problem 5. When we allow the initial conditions to be real numbers the equation yn+1 = p + yn yn−1 (5. ∞).14) possesses a unique positive equilibrium point x and a nontrivial invariant.4 (See [26]. Determine the set G of all initial conditions y−1 .15) may not even be well deﬁned for all n ≥ 0. y0 ∈ (−∞. Conjecture 5.4.4.) Assume p ∈ (−∞.4. Open Problem 5. ∞). . ∞)]. (0. ∞).1 Assume that f ∈ C[(0.4. Show that the subset of G through which the solutions of Eq(5. What is it that makes Lyness’ equation possess an invariant? What type of diﬀerence equations possess an invariant? Along these lines. Conjecture 5.

1. converges to a period-three solution. yn−1 yn−3 n = 0.5 Determine all positive integers k with the property that every positive solution of the equation yn+1 = is periodic.4. . has a nontrivial positive solution which decreases monotonically to the equilibrium of the equation. 1. . + yn yn−1 yn−3 yn−4 n = 0.4. . ∞) × (−∞.4. Conjecture 5. . . 1. yn−k−1 n = 0. (2. Show that the set of points (y−1 . . yn yn−4 + . 1. .4. .3 Assume p ∈ (−∞. 1)-Type Equations Conjecture 5. ∞) through which Eq(5. yn n = 0.4 Show that every positive solution of the equation yn+1 = converges to a period-six solution.6 Show that the equation yn+1 = 1 + yn−1 . y0 ) ∈ R × R through which the equation 1 + yn−1 yn+1 = yn is well deﬁned for all n ≥ 0.4.4. . Open Problem 5.74 Chapter 5. is a set of points without interior. 1 + . ∞). Determine the periodic character and the asymptotic behavior of all solutions with (y−1 . y0 ) ∈ (−∞. + yn−k . .6 Find the set G of all initial points (y−1 . . . . . Open Problem 5. y0 ) ∈ G.5 (See [20]) Show that every positive solution of the equation yn+1 = 1 1 . .15) is not well deﬁned for all n ≥ 0. Conjecture 5. Conjecture 5.

1. . . (1 + xn )xn−1 n = 0. see [59]) Let A ∈ (0. 1. .5.16) to where xn = −1+yn 2 if A > 1 if A = 1 if A < 1 n = 0.17) is called the Gingerbreadman Map and was investigated by Devaney [17]. see [59]) Assume α > 1. .8 (The Gingerbreadman Map. .4. Open Problems and Conjectures 75 Conjecture 5. Show that every positive solution of the equation xn+1 = is bounded. When δ = 1. . ∞). .4. xn xn−1 n = 0. (5.4. . . A} n . 1. Note that the change of variables n e y2 max{x2 .7 (May’s Host Parasitoid Model. Conjecture 5.17) yn+1 = |yn | − yn−1 + δ. 2 αxn . Show that every positive solution of May’s Host Parasitoid Model xn+1 = is bounded. . (5.16) A A 1+yn 2 reduces Eq(5. Eq(5. 1 if A < 1 if A = 1 δ= 0 −1 if A > 1.

.

.Chapter 6 (2. . .5) xn+1 = α + γxn−1 . (6. . 1. . . . 2)-Type Equations 6. 1. 1.3) xn+1 = α + γxn−1 .4) xn+1 = α + γxn−1 . (6. (6. A + Bxn n = 0. 1. .1) xn+1 = α + βxn . 1. . (6.1 Introduction α + βxn . 1. . . . Bxn + Cxn−1 77 n = 0. . . Bxn + Cxn−1 n = 0. 2)-type: xn+1 = n = 0. A + Cxn−1 n = 0. A + Bxn Eq(1) contains the following nine equations of the (2. . (6. . A + Cxn−1 n = 0.6) .2) xn+1 = α + βxn . . (6. .

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.7) xn+1 = βxn + γxn−1 .1) which is in fact a Riccati equation.5). 2)-type Eq(6. (6. . 6. .5) is essentially similar to Eq(6.8) xn+1 = βxn + γxn−1 . . A + Bxn n = 0. the initial conditions are nonnegative. 2)-Type Equations xn+1 = βxn + γxn−1 . and the denominators are always positive. . 1. . (6. for n ≥ −1. Bxn + Cxn−1 n = 0.Riccati This is the (2. A + Czn for n ≥ 0 n = −1.6. In fact if {xn }∞ n=−1 is a positive solution of ∞ ∞ Eq(6. A + Cxn−1 n = 0. . . . α + γzn . Eq(6. To avoid a degenerate situation we will assume that αB − βA = 0. .2 The Case γ = C = 0 : Equation xn+1 = α+βxn A+Bxn .9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. See Section 1. (2. 0.78 Chapter 6. α + γzn . . (6. 1. . . . . . .1). . 1. 1. A + Czn n = 0. Therefore we will omit any further discussion of Eq(6.

.10) are the following: Theorem 6.10) was investigated in [42] and [43].1 The equilibrium y of Eq(6.3 The Case γ = B = 0 : xn+1 = A yn C α+βxn A+Cxn−1 This is the (2.10) has the unique positive equilibrium y given by p= y= q−1+ (q − 1)2 + 4p 2 . 4 Now the following result is a consequence of Theorem 1.6.3.10) is locally asymptotically stable for all values of the parameters p and q. the conjecture has not been proven or refuted.10) αC β and q = .3. 6. . To this day. 1. . Lemma 6. By applying the linearized stability Theorem 1.1) is globally asymptotically stable. See also [36]. 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< . 1 + yn−1 n = 0. 2)-type Eq(6.2.1 The positive equilibrium of Eq(6.) Eq(6.1. The main results known about Eq(6. Eq(6.2) which by the change of variables xn = reduces to the equation yn+1 = where p + qyn .10) is a very simple looking equation for which it has long been conjectured that its equilibrium is globally asymptotically stable.1 Every solution of Eq(6. (6.6. 2 A A (Eq(6.3.1 we obtain the following result.10) is bounded from above and from below by positive constants.2: Theorem 6. .

in every semicycle. The oscillatory character of solutions for Eq(6. We present the proofs for positive semicycles only. then p yn+1 ≥ 1+M and so it is also bounded from below. The proofs for negative semicycles are similar and will be omitted. Now assume for the sake of contradiction that the solution is not bounded from above. and y1+nk = max{yn : n ≤ nk } for k ≥ 0. Then.10) is described by the following result: ∞ Theorem 6.80 Chapter 6. (2. k→∞ lim y1+nk = ∞. Let {yn } be a solution of Eq(6. has at least two terms.3.10). the remaining terms in a positive semicycle are strictly decreasing and in a negative semicycle are strictly increasing. k→∞ Hence. Furthermore.10) we see that yn+1 < qyn + p for n ≥ 0 and so k→∞ lim ynk = lim ynk −1 = ∞. for suﬃciently large k. 0 ≤ y1+nk − ynk = p + [(q − 1) − ynk −1 ]ynk <0 1 + ynk −1 2 which is a contradiction and the proof is complete. Furthermore after the ﬁrst. if the solution is bounded from above by a constant M . Proof.10). From (6.2 Let {yn }n=−1 be a nontrivial solution of Eq(6. (iii) In any two consecutive semicycles. except possibly for the ﬁrst semicycle. their extrema cannot be consecutive terms. except perhaps for the ﬁrst one. there is at most one term which follows the extreme. Then there exists a subsequence {y1+nk }∞ such k=0 that k→∞ lim nk = ∞. every semicycle contains two or three terms. 2)-Type Equations Proof. Then the following statements are true: (i) Every semicycle. (iv) Assume q ≤ p. Clearly. (ii) The extreme in each semicycle occurs at either the ﬁrst term or the second. .

6. the terms yN . The Case γ = B = 0 : xn+1 = (i) Assume that for some N ≥ 0.3 The positive equilibrium of Eq(6. yN +1 . 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.3. α+βxn A+Cxn−1 81 yN −1 < y Then yN +1 = and yN ≥ y. (ii) q≥1 and either p ≤ q or q < p ≤ 2(q + 1). < < 1 + yN +1 1 + yN +1 1+y 2 Theorem 6.10) is globally asymptotically stable if one of the following two conditions holds: (i) q < 1. . > 1 + yN −1 1+y (ii) Assume that for some N ≥ 0. Then yN +1 = p + qyN p + qyN +1 < = yN +2 . 1 + yN −1 1 + yN (iv) Assume that for some N ≥ 0. Then yN ≥ 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 . The opposite case is similar and will be omitted. yN +1 > y and p p +q +q yN +2 1 p + qyN +1 yN +1 = = < y = 1.3. the ﬁrst two terms in a positive semicycle are yN and yN +1 . Then yN +3 = p + qyN +2 p + qyN +1 p + qy = y. p + qyN p + qy = y. and yN +2 are all in a positive semicycle.

1 exist and are positive numbers.1. The following result is a consequence of the linearized stability Theorem 1. it remains to show that every solution {yn }∞ of Eq(6. Theorem 3. 1 ≤ q ≤ p ≤ 2(q + 1) 2 6. 1+i from which the result follows. (ii) The proof when p<q follows by applying Theorem 1. . yn + qyn−1 αB β2 n = 0.11) and q = Eq(6.3. For the proof when see ([42].10) yields i≥ Hence and so because q < 1.11) was investigated in [15].4.3 (f).1. let i = lim inf yn n→∞ and s = lim sup yn n→∞ which by Theorem 6. Then Eq(6. This result was ﬁrst established in [43]. (2. B (6.2.3.10) n=−1 tends to y as n → ∞. p + (q − 1)i ≤ is ≤ p + (q − 1)s i≥s p + qi 1+s and s ≤ p + qs . . 71).3) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= yn + p .4. p. .82 Proof.1. Chapter 6. 2)-type Eq(6. 1. .4 The Case γ = A = 0 : xn+1 = β yn B α+βxn Bxn +Cxn−1 This is the (2. To this end. C . 2)-Type Equations (i) In view of Lemma 6.

the q function u+p f (u. v) = u + qv is increasing in u and decreasing in v. ∞). . Lemma 6. Indeed. (b) The proof is similar and will be omitted. q for all n > k. 6. two consecutive values yk−1 and yk of a solution lie in I. (b) Assume p > q and suppose that for some k ≥ 0.4.1 q p . yk−1 .11) in the sense that if. q for all n > k. for some k ≥ 0.4. The Case γ = A = 0 : xn+1 = α+βxn Bxn +Cxn−1 83 Theorem 6.4. p q p .1 The unique positive equilibrium point y= 1+ 1 + 4p(q + 1) 2(q + 1) of Eq(6.1 . Proof. yk−1 . yk + qyk−1 yk + q p q yk+1 yk + p = ≥ yk + qyk−1 p q p q +p +q > p q 2 and the proof follows by induction. Then the following statements are true: (a) Assume p < q and suppose that for some k ≥ 0.1 Let {yn }∞ n=−1 be a solution of Eq(6. v ∈ [ p . yk ∈ 1. (a) The basic ingredient behind the proof is the fact that when u.11).11) is locally asymptotically stable for all values of the parameters p and q. yk ∈ Then yn ∈ p . Then yn ∈ 1. yk + p yk + p yk+1 = ≤ = 1.1 Invariant Intervals The following result establishes that the interval I with end points 1 and p is an invariq ant interval for Eq(6. then ym ∈ I for all m > k.4.6.

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)

.

L2 . lim yn = 1.11) must lie eventually in [1.22) we see that Li = 1 for i = 1. Now. (b) The proof follows from the discussion in part (a). 4 and so n→∞ for i = 1.11) converges to y. .5 applies and so every solution of Eq(6. p ] is an invariant interval. In this interval the function q f (u. Theorem 1. .7 that every solution of Eq(6.11) with period four. v) = u+p u + qv is increasing in u and decreasing in v. By applying (6. By applying (6. In this interval the q function u+p f (u. 2.11) with two consecutive values in [1. L1 .22) is a periodic solution of Eq(6.4. 2. 3. p ] converges to y. Under the assumption that q ≤ 1 + 4p. 4.19) to the solution (6. .14) to this period four solution we obtain a contradiction and so every solution of Eq(6. 2)-Type Equations is either identically equal to one or else it is strictly monotonically convergent. 1].4. L3 .11) n→∞ lim yn = y and the proof is complete. 2 . (2. p ].2 that [1. p ]. (6. v) = u + qv is decreasing in both variables and so it follows by applying Theorem 1. Case 2 Assume p = q. . First. it converges to a periodic solution with period-four which q is not the equilibrium. 3. q Hence for every solution {yn }∞ n=−1 of Eq(6.88 Chapter 6. L4 . . Recall from Lemma 6. By q using an argument similar to that in the case p = q. As in Case 2 we can see that every solution of Eq(6. assume that p < q.11) must eventually lie in the interval [ p . Set Li = n→∞ y4n+i lim Thus clearly .4. we can see that if a solution is not eventually in [1. . we assume that p > q.

5 The Case β = C = 0 : xn+1 = A yn B This is the (2. . if and only if q = 1. φ. 2)-type Eq(6. Concerning prime period-two solutions.23) was investigated in [28]. and is an unstable saddle point when q > 1.) The only equilibrium point of Eq(6.23) and q = (Eq(6. Theorem 6.23) has a prime period-two solution .2 (a) Eq(6. ∞) : φψ = p}. . γ .5. . .23) is locally asymptotically stable when q < 1.24) . the following result is a consequence of the results in Section 2.1 we obtain the following result: Theorem 6.25) (b) Assume (6. .6.23) is y= q−1+ (q − 1)2 + 4p 2 and by applying the linearized stability Theorem 1. ψ.7. ψ ∈ (0. The Case β = C = 0 : xn+1 = α+γxn−1 A+Bxn 89 α+γxn−1 A+Bxn 6. Then the values φ and ψ of all prime period-two solutions (6. . 1 + yn αB A2 n = 0.25) holds.25) holds.23) converges to a period two solution. φ.1 The equilibrium y of Eq(6. . Then every solution of Eq(6. 1.5.4) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + qyn−1 . (c) Assume that (6. . (6.5.24) are given by {φ. (6. .1. A (6. ψ.

(6. an oscillatory solution {yn } of Eq(6. It is suﬃcient to show that {yn } is a decreasing sequence for n ≥ K.23). yn0 > yn0 −1 . Then the following statements are true: (a) After the ﬁrst semicycle.23) is globally asymptotically stable.26) .1 Semicycle Analysis Our results on the semicycles of solutions for Eq(6.90 Chapter 6. Clearly yn+1 = Set M= p + qyn−1 < p + qyn−1 . yn0 ) < f (y.5.23) oscillates about the equilibrium y with semicycles of length one. y) = y. Proof.1 (b) Let {yn } be a nonoscillatory solution of Eq(6.5. The case where yn−1 < y for n ≥ K is similar and will be omitted. Clearly.23) are as follows: Theorem 6. (b) Assume p ≥ q.4 Assume that q < 1. Then yn0 +1 = which is impossible. Then every nonoscillatory solution of Eq(6.23) converges monotonically to the equilibrium y.7.5.2 The Case q < 1 The main result in this section is the following: Theorem 6. p + qyn0 −1 = f (yn0 . 1 + yn p + .5. 1−q n ≥ 0. (a) This follows immediately from Theorem 1. Proof. Then the positive equilibrium of Eq(6. So assume for the sake of contradiction that for some n0 ≥ K.23). yn0 −1 ) < f (yn0 . We will assume that there exists a positive integer K such that yn−1 ≥ y for every n ≥ K. the condition p ≥ q implies that the function f (x) = is decreasing. (2.23) and let y denote the unique positive equilibrium of Eq(6.3 Let {yn } be a nontrivial solution of Eq(6. 1 + yn0 2 p + qx 1+x 6. 2)-Type Equations 6.

6. b].5. By applying Theorem 1. it now follows that every solution of Eq(6.26) and Eq(6. y) ≤ b. a ≤ f (x. The Case β = C = 0 : xn+1 = α+γxn−1 A+Bxn 91 where is some positive number.3 The Case q > 1 In this case we will show that there exist solutions that have two subsequences.6 it follows that y is a global attractor of all solutions of Eq(6. for all (x.5. .5. y2n−1 < q − 1 and y2n > y0 + n Hence n→∞ p .23).23). 2 6. b] × [a. From Eq(6. M ]. Set p + qy p f (x.1.23) lies eventually in the interval [0. one of which is monotonically approaching 0 and the other is monotonically approaching ∞. The local stability of y was established in Theorem 6. y) ∈ [a. lim y2n = ∞ p + qy2n−1 =0 1 + y2n and lim y2n+1 = lim n→∞ n→∞ and the proof is complete. In fact this is true for every solution {yn }∞ n=−1 whose initial conditions are such that the following inequalities are true: y−1 < q − 1 and y0 > q − 1 + To this end observe that y1 = and y2 = and by induction. 1+x 1−q Then clearly. and b = + . 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 proof is complete. y) = . a = 0.4.

B .27) has the unique positive equilibrium y= 1+ 1 + 4p(1 + q) 2(1 + q) .27) was investigated in [45].29) .30) (6. . 2)-Type Equations 6. .92 Chapter 6. If we now set it is easy to see that (6.1.28) By employing Theorem 1.27) about y is zn+1 + q qy − p zn − zn−1 = 0.6 The Case β = A = 0 : xn+1 = γ yn C α+γxn−1 Bxn +Cxn−1 This is the (2. (2. The linearized equation associated with Eq(6. . . .) Eq(6. 2)-type Eq(6. (6. 1. .29) is satisﬁed if and only if either q≤1 or q > 1 and F 2p q−1 > 0. (6.1 we see that y is locally asymptotically stable when (q − 1)y < 2p and unstable (a saddle point) when (q − 1)y > 2p.27) and q = (Eq(6. Clearly the equilibrium y is the positive solution of the quadratic equation (1 + q)y 2 − y − p = 0. 1+q (y + p)(1 + q) n = 0. . C (6.6) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + yn−1 . 1. F (u) = (1 + q)u2 − u − p. qyn + yn−1 αC γ2 n = 0.

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. ψ. .2 Eq(6. (6. q−1 . . and more precisely a saddle point equilibrium.27).1 The equilibrium y of Eq(6.6. φ. . be a period-two cycle of Eq(6.30) is satisﬁed if and only if q > 1 and F 2p q−1 < 0.27) is locally asymptotically stable when q < 1 + 4p and unstable. .31) 6. φ. qφ + ψ It now follows after some calculations that the following result is true: Theorem 6. . when q > 1 + 4p. φ. . Then φ= p+φ qψ + φ and ψ = p+ψ .6. Furthermore when (6.6. .6. . . ψ. ψ.6. The following result is now a consequence of the above discussion and some simple calculations: Theorem 6.27) has a prime period-two solution . ψ.32) holds. . The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 93 Similarly (6. .1 Let Existence and Local Stability of Period-Two Cycles .32) (6. φ. if and only if q > 1 + 4p. .

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

First we will establish Inequality (6. + 2 (qψ + φ)2 ∂v (qφ + ψ) (qφ + ψ)2 Set S= and D= ∂h ∂g (φ. Inequality (6. ψ) + (φ. ψ) (φ. The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 95 q(p + ψ)(qψ − p) ∂h (φ. ψ) ∂u ∂v ∂g ∂h ∂h ∂g (φ. ψ) = . if and only if |S| < 1 + D < 2. ψ) (φ.34) (6.1.36) (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.6. To this end observe that (6. ψ) − (φ.33) (6.34) is equivalent to (6. ψ).34).35) (6. ∂u (qφ + ψ)2 (qψ + φ)2 ∂h q 2 (p + φ)(p + ψ) qφ − p (φ.6.33) is equivalent to the following three inequalities: S <1+D −1 − D < S D < 1. ψ) = − . ∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1.1 that both eigenvalues of JT 2 unit disk |λ| < 1. p )[q − (1 + 4p)] < 0 p+1 .

27) either remain forever outside the interval I or the solution is eventually trapped into I.35). we establish Inequality (6.2 Invariant Intervals Here we show that the interval I with end points one and p is an invariant interval for q Eq(6.36). q .3 Let {yn }∞ n=−1 be a solution of Eq(6. More precisely we show that the values of any solution {yn }∞ n=−1 of Eq(6.27). Theorem 6. p ≤ yN ≤ 1. Finally.6. Observe that (6. (a) Assume that p ≤ q and that for some N ≥ 0.27).6. Observe that (6. 6. (2.36) 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. 2)-Type Equations Next we will establish Inequality (6.96 Chapter 6. Then the following statements are true.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. q Then p ≤ yn ≤ 1 for all n ≥ N.

The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 97 (b) Assume that p ≥ q and that for some N ≥ 0.3 Semicycle Analysis Here we present a thorough semicycle analysis of the solutions of Eq(6. p 1 ≤ yN ≤ . Let {yn }∞ n=−1 be a solution of Eq(6.27) relative to the equilibrium y and relative to the end points of the invariant interval of Eq(6. for n ≥ 0. q Then 1 ≤ yn ≤ Proof.27). (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. (6.38) (6.27).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 ≥ 0.6. Then observe that the following identities are true: p − yn for n ≥ 0.39) 2 qyn−1 (yn − p ) + (yn − 1)(yn−1 yn + qyn ) q . p + yN −1 p + yN −1 ≥ = q qyN + yN −1 (pyN + p yN −1 ) p q and the proof follows by induction. (6. 6. (b) Clearly.6. yN +1 = and yN +1 = p + yN −1 p + yN −1 =1 ≥ qyN + yN −1 p + yN −1 p q p + yN −1 ≤ pyN + p yN −1 q p q p p + yN −1 = p + yN −1 q 2 p + yN −1 = qyN + yN −1 and the proof follows by induction.

q (ii) If for some N ≥ 0. (2.1 Assume that p>q and let {yn }∞ n=−1 be a solution of Eq(6. yN ≥ 1. then yN +1 < 1. then yN +1 < p . then yN +1 > 1.41) 2 pyn−1 + yn−1 yn + pyn p(p + yn−1 ) + yn (pyn + yn−1 ) (6. yN = p . (6. In particular [1. (vii) If for some N ≥ 0. yN < p . yN ≥ p . yN ≤ p .37)-(6. yN < 1. Then the following statements are true: (i) If for some N ≥ 0.42). q (iv) If for some N ≥ 0. pyn + yn−1 Chapter 6. 1. then yN +1 > 1. . yN ≤ 1. Lemma 6. q (iii) If for some N ≥ 0.27). q Lemma 6. . . Then the following statements are true: (i) If for some N ≥ 0.27). .42) The following three lemmas are now direct consequences of (6. then yN +2 > yN . then yN +2 < yN . (6. q (v) If for some N ≥ 0. q (vi) If for some N ≥ 0.27). 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 .6. yN > p . then yn ∈ [1. then yN +1 = 1. p ] for n ≥ N .6.2 Assume that p=q ∞ and let {yn }n=−1 be a solution of Eq(6.40) 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. 2)-Type Equations n = 0. (viii) p 1<y< .

yN < p . then yN < yN +2 < 1. Then the following statements are true: (i) If for some N ≥ 0.3 Assume that p<q ∞ and let {yn }n=−1 be a solution of Eq(6. then yN +1 < 1. q (iii) If for some N ≥ 0. then yN +1 ≥ 1. q (vi) If for some N ≥ 0.27). (vii) If for some N ≥ 0. then yN +1 = 1. yN > p .6. q (viii) If for some N ≥ 0. yN < 1. q (ii) If for some N ≥ 0. then yN +1 < 1. q (v) If for some N ≥ 0. (iii) If for some N ≥ 0.3: .2 and Lemmas 6.7.6. yN ≥ 1.7. q (iv) If for some N ≥ 0. (iv) If for some N ≥ 0. q q The following result is now a consequence of Theorems 1. then yN +1 > p . yN ≤ 1. In particular [ p . (ix) p < y < 1. then yN +1 = 1. yN = p . then yN +2 > yN . then yN +1 > 1. yN > 1. The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 99 (ii) If for some N ≥ 0. yN ≤ 1. yN ≤ p . 1] is an q q invariant interval for Eq(6. q Note that the function f (u. (v) 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 . 1] for n ≥ N .6. then yN +2 < yN . yN = 1.1-6. then yn ∈ [ p .1 and 1.6.27). then yN > yN +2 > 1.6. Lemma 6. yN ≤ p .

100 Chapter 6.44) (6. Eq(6. (6. or vice-versa.27) has a “unique” prime period-two solution . φ. The above observations are summarized in the following theorem: Theorem 6. yn ∈ I for n ≥ N. and only then.7.32) holds. must converge to y. 1] must converge to the two cycle (6.43) 6. with φ + ψ = 1 and φψ = p .27) either eventually enters the interval I with end points 1 and p and remains there.4.6.27). when (6. . .4 Global Behavior of Solutions Recall that Eq(6. any solution which remains forever outside the interval I = [ p .4.32) holds. ∞).45). .31) holds.43) holds. p<q the length is one.6. But this two cycle lies q inside the interval I. or for some N ≥ 0.44)-(6.6 and 1. On the other hand when (6. Hence. (2. q Now when q ≤ 1 + 4p Then the equilibrium y of Eq(6. . . if Theorem 6. while if p>q the length is at most two.45) there are no period-two solutions and so the solutions must eventually enter and remain in I and.46) Also recall that a solution {yn }∞ n=−1 of Eq(6. in view of Theorems 1. or stays forever outside I.27) eventually enters and remains in the interval I. ψ.4 Let {yn }∞ n=−1 be a solution of Eq(6. except for the length of the ﬁrst semicycle of the solution.32) holds. The character of these solutions remains an open question. When (6. 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. ψ. . .6. q−1 (6. When (6.27) is global attractor. 2)-Type Equations Then either all the even terms of the solution lie in J and all odd terms lie in K. φ. every solution of Eq(6.5 (a) Assume q ≤ 1 + 4p.27) has a unique equilibrium y which is locally stable when (6. (6.

. 1+y So y = 0 is always an equilibrium point. Then the zero equilibrium of Eq(6. 1 + yn n = 0. More precisely it is a saddle point when 1−p<q <1+p and a repeller when q > 1 + p.47) are the solutions of the equation p= y= py + qy .47) β γ and q = .1: Theorem 6.6.3. 1]. 2)-type equation Eq(6. The Case α = C = 0 : xn+1 = (b) Assume βxn +γxn−1 A+Bxn 101 q > 1 + 4p. (b) Assume p + q > 1.1 (a) Assume p + q ≤ 1. Then every solution of Eq(6. A A (Eq(6.27) eventually enters and remains in the interval [ p .1 and the linearized stability Theorem 1.7 The Case α = C = 0 : xn+1 = A yn B βxn +γxn−1 A+Bxn This is the (2. The following result follows from Theorem 1. q 6. Then the zero equilibrium of Eq(6. .47) is globally asymptotically stable. 1.7.47) is unstable. .1. and when p + q > 1.7. .7) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where pyn + qyn−1 .47).) The equilibrium points of Eq(6.47) was investigated in [52]. y = p + q − 1 is the only positive equilibrium point of Eq(6. (6.

∞). (6. p ]. q (0.5 that the following result is true: Theorem 6. the values of φ and ψ of all prime period-two solutions are given by: pφ {φ. . ψ. Theorem 6.47) eventually enters the interval [ p . q (b) Assume that q > 1.7. Then the positive equilibrium y = p + q − 1 of Eq(6. [ p . if and only if q = 1 + p.3. ψ. .102 (c) Assume Chapter 6. φ.47) q with initial conditions in (0. φ. . p ].49) The main result here is the following theorem about Eq(6. The case is treated in Section 6. Furthermore.47). (6. .47).1 Invariant Intervals q=1 (6. Concerning prime period-two solutions it follows from Sections 2.47) with q = 1. ∞) is an invariant interval for Eq(6. p ] is an invariant interval for Eq(6.50) Then every solution of Eq(6. 2)-Type Equations 1 − p < q < 1 + p.47) is an unstable saddle point.7.3 (a) Assume that q < 1. every solution of Eq(6. That is.48) Furthermore when (6.2 Eq(6. Furtherq more. ∞) : ψ = = 2p}. φ−p 6.48) holds. .7. remains in this interval. (6. (d) Assume q > 1 + p. (2. ψ ∈ (p.7. q . Then the positive equilibrium y = p + q − 1 of Eq(6. .47) eventually enters the interval (0.51) Then every solution of Eq(6.47) is locally asymptotically stable.47) has a prime period-two solution .

47).6.51) holds. (x − x)(f (x. ∞) → [0.7. m ∈ N.50) holds. y refers to the positive equilibrium p + q − 1 of Eq(6. that is. Lemma 6.7. 1+x Observe that g(y) = y and that g increases on [0.7. m ∈ N. ∞) denote the function f (x. m q q (b) Assume that (6. whose proof is straightforward and will be omitted. Hereafter. yk ≤ and yk+2 ≥ p p =⇒ yk+2 ≤ p < q q p p =⇒ yk > m+1 > p. x) − x) < 0 for x ∈ (0.47). ∞) → [0. Then the following statements are true for every n ≥ 0: p + q > 1 and q ≤ 1.4 Suppose Let {yn }∞ n=−1 be a positive solution of Eq(6. ∞). y) = px + qy . qm q 6. Let f : [0. . 1+x Semicycle Analysis of Solutions When q ≤ 1 Then f satisﬁes the negative feedback condition. ∞) − {x}. ∞) × [0.7.2 Here we discuss the behavior of the semicycles of solutions of Eq(6. Theorem 6.1 (a) Assume that (6. Then for any k. Then for any k. Consider the function g : [0. 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. ∞) deﬁned by g(x) = (p + q)x .47) when p + q > 1 and q ≤ 1. yk ≥ and yk+2 ≤ p p =⇒ yk+2 ≥ p > q q p p =⇒ yk > m+1 > p.

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

yn−1 ≥ qyn−1 > y + (q − 1)yn > yn 2 pyn + qyn−1 qyn−1 (yn + 1) < ≤ yn−1 . The Case α = C = 0 : xn+1 = (d) Suppose yn < y ≤ yn−1 . (6. Suppose p − qyn−1 ≥ 0. To see that yn+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. . we must consider two cases: Case 1.47) is globally asymptotically stable.5 Suppose that q = 1 and let y−1 + y0 > 0.7.3 Semicycle Analysis and Global Attractivity When q = 1 q = 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 (6.7. Then yn+1 = (e) Observe that yn+1 = and so from which the result follows. Here we discuss the behavior of the solutions of Eq(6.47) is y = p. Case 2. Then the equilibrium y of Eq(6.47) when Here the positive equilibrium of Eq(6. 1 + yn 1 + yn pyn + qyn−1 >y =p+q−1 1 + yn 6.52) (b) Assume y−1 ≤ p and y0 ≤ p.6. Suppose p − qyn−1 < 0. More precisely the following statements are true.52) holds.7. (c) Assume either y−1 < p < y0 or y−1 > p > y0 . Then yn ≤ p for all n > 0 and (6.52) holds. (a) Assume y−1 ≥ p and y0 ≥ p. Then yn ≥ p for all n > 0 and n→∞ lim yn = p. Theorem 6. Then px+qyn−1 increases in x and so 1+x yn+1 = pyn−1 + qyn−1 pyn + qyn−1 ≤ < yn−1 .

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

y) = px + qy 1+x satisﬁes the hypotheses of Theorem 1. By Theorem 6. Concerning the long-term behavior of solutions of Eq(6.47).7. p ]. 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).6.5 in the interval I from which the result follows. y0 ∈ [p. and so convergent. Then every positive solution of Eq(6.47) lies eventually in the interval [p. 2 6.6 Assume 1 − p < q < 1.47) converges to the positive equilibrium y.47) is global attractor of all positive solutions of Eq(6.3 (a) we may assume that the initial conditions y−1 and y0 lie in the interval I = (0. ∞). that y−1 .4 Global Attractivity When q < 1 1 − p < q < 1.7. Theorem 6. . Observe that the consistency condition y ≤ p is equivalent to the condition q q ≤ 1.47) we may assume. Thus. Now clearly the function q f (x. ∞).7. m = M . every positive solution of Eq(6. q In this case.47) when which is equivalent to p y> .7.5 Long-term Behavior of Solutions When q > 1 q>1 Here we discuss the behavior of the solutions of Eq(6.4. In a similar way we can show that the subsequence {y2k }∞ is decreasing.7. The Case α = C = 0 : xn+1 = βxn +γxn−1 A+Bxn 107 and the claim follows by induction. Proof. without loss of generality. Here we consider the case where and we show that the equilibrium point y = p + q − 1 of Eq(6. 2 6.

47) converges to a period-two solution. transforms Eq(6.47). (b) Assume 1 < q < 1 + p. we obtain the following theorems about the solutions of Eq(6. Then Eq(6.108 Then the change of variable yn = un + p.7. The character of solutions of Eq(6.53) was investigated in Section 6.53). an oscillatory solution of Eq(6.7 After the ﬁrst semicycle.5. (2.53) where un ≥ 0 for n ≥ 0. Theorem 6.47) converges to the positive equilibrium of Eq(6.7. Then every solution of Eq(6.5 to Eq(6. By applying the results of Section 6. 2)-Type Equations p(q − 1) + qun−1 1 + p + un (6.8 (a) Assume q = 1 + p. .47) oscillates about the positive equilibrium y with semicycles of length one. Then every positive solution of Eq(6.47) to the equation un+1 = Chapter 6.47) when q > 1: Theorem 6. (c) Assume q > 1 + p.47) has unbounded solutions.

54) is locally asymptotically stable. .55) Then the zero equilibrium of Eq(6.8. γ (6.1 and 1. The following result follows from Theorems 1.54) is unstable and the positive equilibrium y = p + 1 − q of Eq(6.1.54) is globally asymptotically stable.8 The Case α = B = 0 : xn+1 = xn = γ yn C This is the (2.3. q + yn−1 β γ n = 0. The Case α = B = 0 : xn+1 = βxn +γxn−1 A+Cxn−1 109 βxn +γxn−1 A+Cxn−1 6. .54) and so we will assume without further mention that (6.) The equilibrium points of Eq(6.8. Theorem 6. In the remainder of this section we will investigate the character of the positive equilibrium of Eq(6.54) was investigated in [53].54) also possesses the unique positive equilibrium y = p + 1 − q. A .55) holds and y−1 + y0 > 0. Eq(6.54) and when p + 1 > q.6. .54) and q = (Eq(6.8) which by the change of variables reduces to the diﬀerence equation yn+1 = where p= pyn + yn−1 . the positive equilibrium y = p + 1 − q of Eq(6.1. (b) Assume p + 1 > q. 2)-type Eq(6.54) are the solutions of the equation y= py + y .54) is globally asymptotically stable. q+y So y = 0 is always an equilibrium point of Eq(6. (6. 1.1 (a) Assume p + 1 ≤ q. .55) holds. Furthermore the zero equilibrium is a saddle point when 1−p<q <1+p and a repeller when q < 1 − p. Then the zero equilibrium of Eq(6. Our goal is to show that when (6.

p + yn−1 n = 0.1 Assume that p<q (6.60) (6. .54) is such that: < 1 if p < q < p + 1 y is = 1 if p = q > 1 if p > q.1 Invariant Intervals and Semicycle Analysis Let {yn }∞ n=−1 be a positive solution of Eq(6. . Lemma 6. When p=q that is for the diﬀerence equation yn+1 = pyn + yn−1 .110 Chapter 6. 2)-Type Equations 6.57) and yn+1 − yn = (p − q)yn + (1 − yn )yn−1 q + yn−1 (6.58) 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 ≥ 0.62) and let {yn }∞ n=−1 be a positive solution of Eq(6. (6.54).61) The following three lemmas are now direct consequences of the above identities. Then the following identities are easily established: q yn − p yn+1 − 1 = p for n ≥ 0.56) q + yn−1 q p2 [yn − p ] + (p − q)yn−1 q yn+1 − = p p(q + yn−1 ) 2 for n ≥ 0. for n ≥ 0.58) Note that the positive equilibrium y =p+1−q of Eq(6. (6.8.59) the identities (6. . (6. (2.56)-(6. Then the following statements are true: .8.54). (6. 1.

The Case α = B = 0 : xn+1 = βxn +γxn−1 A+Cxn−1 111 q (i) If for some N ≥ 0. yN ≥ 1. q (ii) If for some N ≥ 0.3 Assume that p>q (6. yN ≥ 1. yN ≤ 1. . then for n ≥ 0. (iv) If for some N ≥ 0.64) and let {yn }∞ n=−1 be a positive solution of Eq(6.63) and let {yn }∞ n=−1 be a positive solution of Eq(6. yN < p . then yN +1 < 1. (v) If for some N ≥ 0. (vi) If for some N ≥ 0. Then the following statements are true: (i) If y0 < 1. q (iii) If for some N ≥ 0. then yN +1 > 1. then yN +1 = 1. then yN +1 < 1. then yn = 1 for n ≥ 0. p q then yN +1 < p . yN = p . (iii) If y0 > 1. then yN +1 < yN . q (iii) If for some N ≥ 0. then yN +1 > 1. (v) If for some N ≥ 0.54). Lemma 6. then yN +1 < 1. yN = p . then yN +1 > q p and yN +2 > 1. yN < p . Lemma 6. Then the following statements are true: q (i) If for some N ≥ 0. yn < 1 and the solution is strictly increasing. (ii) If y0 = 1.6. yN ≤ 1. p q then yN +1 > p . (iv) If for some N ≥ 0.8. yN > q 2 . q (ii) If for some N ≥ 0. yn > 1 and the solution is strictly decreasing. yN > p . (vi) If for some N ≥ 0.54). yN > p .8.2 Assume that p=q (6. then for n ≥ 0. then yN +1 = 1. then yN +1 > yN . yN > q 2 .8.

1 that if a solution {yn }∞ n=−1 is such that yn ≥ 1 for all n ≥ 0 then the solution decreases and its limit lies in the interval [1. When (6.8.64) holds.54) is globally asymptotically stable. q+1 From the above observations and in view of Theorem 6. .4.54) eventually enters and remains in the interval (0.3 where we established in Theorem 6. 1) the function f (u. y is a global attractor.1.8.8. by Lemma 6. ∞).8. .2 Global Stability of the Positive Equilibrium When (6.3 that every solution of Eq(6. Hence.112 Chapter 6.3. This (2. Now in the interval (0.8. (2. (6. 1.8. ∞).54) converges to the equilibrium y = 1. Set yn = 1 + (q + 1)un for n ≥ −1.8.3 that every solution converges to its positive equilibrium u= p−q . Without loss of generality we will assume that yn > 1 for n ≥ −1. 2)-type equation was investigated in Section 6.65) with positive parameters and positive initial conditions. This is impossible because y < 1. 2)-Type Equations 6.63) holds. Then the positive equilibrium y = p + 1 − q of Eq(6. assume that (6.54) eventually enters and remains in the interval (1. Here it is clear from Lemma 6.8.2 that every solution of Eq(6. 1). . 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 . it follows from Lemma 6. n = 0. v) = pu + v q+v is increasing in both arguments and by Theorem 1.62) holds.2 Assume that p + 1 > q and that y−1 + y0 > 0. it follows from Lemma 6. every positive solution of Eq(6. .1 we have the following result: Theorem 6. Next.

66) about y is zn+1 − p−q p−q zn + zn−1 = 0. qyn + yn−1 n = 0. .9.9.1. (b) Assume that p < q. 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. Then the positive equilibrium of Eq(6.66) is locally asymptotically stable. By applying the linearized stability Theorem 1. The Case α = A = 0 : xn+1 = βxn +γxn−1 Bxn +Cxn−1 113 βxn +γxn−1 Bxn +Cxn−1 6. (6.66) is y= p+1 q+1 and the linearized equation of Eq(6. Then the positive equilibrium of Eq(6. .66) was investigated in [54]. .1 (a) Assume that p > q.67) . .68) (6. (Eq(6.1 we obtain the following result: Theorem 6. (6.6.) The only equilibrium point of Eq(6. . 1. 1.66) is locally asymptotically stable when q < pq + 1 + 3p and is unstable (and more precisely a saddle point equilibrium) when q > pq + 1 + 3p. .66) B β and q = . . (p + 1)(q + 1) (p + 1)(q + 1) n = 0.9) which by the change of variables xn = reduces to the equation yn+1 = where pyn + yn−1 . 2)-type Eq(6.

66) possesses a unique prime period-two solution: . On the other hand when p<q and q > pq + 1 + 3p.70) t2 − (1 − p)t + = 0. v) p pv+u + v qv+u q pv+u + v qv+u v pv+u qv+u . 1. we set un = yn−1 and vn = yn for n = 0. φ.9. 2)-Type Equations 6. Eq(6. .69) is asymptotically stable if the eigenvalues of the φ Jacobian matrix JT 2 . evaluated at lie inside the unit disk. . . ∞) deﬁned by: T Then u v = φ ψ u v g(u. v) = = . the second iterate of T . The prime period-two solution (6. . ψ. n = 0. . One can see that T2 where pv + u g(u. . and write Eq(6. φ. .69) where the values of φ and ψ are the (positive and distinct) solutions of the quadratic equation p(1 − p) (6. .66) has no prime period-two solutions. . (2.66) in the equivalent form: un+1 = vn n +un vn+1 = pvn +un . . (6. . . .1 Existence of a Two Cycle It follows from Section 2. q−1 In order to investigate the stability nature of this prime period-two solution. Eq(6. ∞) × (0.5 that when p > q. qv Let T be the function on (0. 1. 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 . .114 Chapter 6. v) h(u. v) = qv + u and h(u. is a ﬁxed point of T 2 . ψ.

73) (6. Next we will establish Inequality (6. ψ > 0.9.72) is always true.72) φ ψ and ψ = Inequality (6. We will use the fact that φ + ψ = 1 − p. P < 0 and Q < 0 and so Inequality (6. By applying Theorem 1.1.71) 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 . (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. qφ + ψ (6. Observe that φ > 0.6.71). 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 < 1.1 (c). The Case α = A = 0 : xn+1 = Set P = Then JT 2 βxn +γxn−1 Bxn +Cxn−1 115 p−q .71) (6. φ= pψ + φ qψ + φ φψ = p(1 − p) q−1 pφ + ψ .

74) = (q − p)(1 − p)2 (qp + 1 + 2p). Finally. Now observe that the lefthand side of (6. Inequality (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.74) is p(1 − p) (p − 1)(q − 1)] q−1 (6. Inequality (6.74) 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 (6.68) holds. = [q(φ + ψ)2 + (q − 1)2 φψ]2 (q − p) φψ < (1 − p)(q − p) .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 .71) is true if and only if (6. (2.116 if and only if Chapter 6. 2)-Type Equations (q − p)[(qφ + ψ)(pφ + ψ) + (qψ + φ)(pψ + φ)] < [(qφ + ψ)(qψ + φ)]2 .

35). . Proof.67) holds. .7. (b) Assume p < q.66): Theorem 6.2 Assume that Condition (6. .70). φ. y) is increasing in y and decreasing in x.3 Let {yn } be a nontrivial solution of Eq(6. φ. . Then the following statements are true: (a) Assume p > q. In summary. .66). ψ.9.68) holds. (a) The proof follows from Theorem 1.69) of Eq(6.66) posseses the prime period-two solution . . ψ. after the ﬁrst semicycle.3 Global Stability Analysis When p < q The main result in this section is the following Theorem 6. Furthermore this prime period-two solution is locally asymptotically stable. Then {yn } oscillates about the equilibrium y with semicycles of length two or three. the function 2 f (x.4 by observing that the condition p > q implies that the function px + y f (x.9.66) is globally asymptotically stable. Then either {yn } oscillates about the equilibrium y with semicycles of length one. or {yn } converges monotonically to y.9.4 Assume that p<q and (6. Then the positive equilibrium y of Eq(6. 6.2 Semicycle Analysis In this section.7.66).9. (b) The proof follows from Theorem 1. except possibly for the ﬁrst semicycle which may have length one.9. where φ and ψ are the two positive and distinct roots of the quadratic Eq(6. the following result is true about the local stability of the prime periodtwo solution (6. Theorem 6. Then Eq(6. This function also satisﬁes Condition (1.9. 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. y) = qx + y is increasing in x and decreasing in y.1 by observing that when p < q. . 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. 6. we present a semicycle analysis of the solutions of Eq(6.6.

69) when instead of (6. . Theorem 6. . Assume that for some solution {yn }∞ n=−1 of Eq(6. yN +1 ) ≤ f (ψ. (6. φ) = φ and in general yN +2k ≥ ψ and yN +2k+1 ≤ φ for k = 0. 2 Next we want to ﬁnd more cases where the conclusion of the last theorem holds.6 lim sup yn = ψ n→∞ px + y . 1. y) ≤ 1 for all x. q Finally in view of (6.6 can also be used to establish that certain solutions of Eq(6. Then this solution converges to the two cycle φ.. ψ.68) holds. . denote the two cycle of Eq(6.4 follows as a consequence of Theorem 1. . Also clearly.67).118 Proof.4. We consider ﬁrst the case where eventually consecutive terms yi .66) and for some index N ≥ −1. y) = Then clearly.6 and the fact that y is locally asymptotically stable.75) yN ≥ ψ and yN +1 ≤ φ. .68) holds. (6. Eq(6. Set f (x. Set f (x. p ≤ f (x.66). . qx + y from which we conclude that limk→∞ yN +2k = ψ and limk→∞ yN +2k+1 = φ. yN +2 = f (yN +1 .4. Proof. with φ < ψ. and increasing in y for each ﬁxed x. φ. yi+1 lie between m and M .66) converge to the two cycle (6. ψ. yN +3 = f (yN +2 . y) = px + y . Assume that (6..75) holds. and lim inf yn = φ n→∞ Now as in the proof of Theorem 1.4.66) has no prime period-two solution.9. . yN ) ≥ f (φ. Now the conclusion of Theorem 6. y > 0. ψ. φ. .5 Assume that (6.67). Let φ. .9. qx + y Chapter 6. 2)-Type Equations and note that f (x. y) is decreasing in x for each ﬁxed y. ψ) = ψ. . . ψ. (2. 2 The method employed in the proof of Theorem 1.

(B) y2n+1 > M and M ≥ y2n > m for every n.9.9. M ] and if we never get two successive terms outside of the interval [m. limn→∞ y2n+1 M − M2 . Hence limn→∞ y2n = M . yi+1 ≤ M. ψ. Proof. φ. qy2n+2 − p M − M2 lim y2n+1 = . m= 2 Since the case in which the solution lies outside of the interval [m. From Eq(6. Since the proofs are similar for all cases we will give the details in case (A). m ≤ yi . The Case α = A = 0 : xn+1 = βxn +γxn−1 Bxn +Cxn−1 119 Lemma 6. m) ≤ f (yi+1 .6. n→∞ qM − p On the other hand by solving the equation M = f (m. M ] is answered by Theorem 6. By our earlier observation we have that M ≥ lim supi→∞ yi .5. Lemma 6. .66) we have y2n+1 = and so y2n − y2n y2n+2 . . M ] for every k > i. M ]. M ) = M. Then yk ∈ [m. and so M ≥ lim supn→∞ y2n ≥ lim inf n→∞ y2n ≥ M. yi ) = yi+2 ≤ f (m. then we have one of the following four cases: (A) y2n > M and M ≥ y2n+1 > m for every n.2 In all cases (A)-(D) the corresponding solution {yn } converges to the two cycle . ψ. . . . Proof.9. and the result follows by induction. M ) = pm + M qm + M for m we ﬁnd Thus. for every i. Using the monotonic character of the function f we have m = f (M. M ] (that is. . (D) y2n+1 < m and m ≤ y2n < M for every n. we are left with the case in which the solution lies eventually in [m. (C) y2n < m and m ≤ y2n+1 < M for every n. either yi > M and yi+1 < m or yi < m and yi+1 > M ). . qM − p = m and the proof is complete. φ. 2 If we never get two successive terms in the interval [m.9.1 Suppose that for some i ≥ 0.

the proof is complete. yi+1 ) = yi+3 . y) = y. 1). . in which case Eq(6. the sequence {yi+2k } is non decreasing and bounded above by M .The case M ≥ yi+3 ≥ yi+1 ≥ y ≥ yi ≥ yi+2 ≥ m is handled in a similar way.4 (1) f (x. Proof. Thus. y) > y if and only if x < y . f (y. We denote this value by y . but this situation cannot occur.76) It makes sense to consider this equation only for y ∈ ( p . (2. yi+3 ) ≤ f (yi+2 . yi+2 ) ≥ f (yi+1 . . In view of the uniqueness of the two cycle and the fact that at least one of yi . Let us consider the ﬁrst case.76) has a q unique positive solution for x. y ) > y and y > y. which of course depends on y. .120 Chapter 6. 2)-Type Equations Lemma 6. φ. and the sequence {yi+2k+1 } is a non increasing sequence and bounded below by m. ψ. . The second case can be handled in a similar way. (2) x > y if and only if x < y . y ) < y and y > y . The correspondence between y and y has the following properties: Lemma 6.9. . 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 . f (y.9. yi ) = yi+2 and yi+5 = f (yi+4 . φ. y) = y. so we are led to the general equation f (x. Induction on i then establishes the monotonicity of the two sequences. (3) For y < y < M. In this case we need to know more about the condition yi+2 = f (yi+1 . Proof. . as the following results show. yi ) ≤ yi . (6. Therefore. Thus f (y . ψ. Here we have yi+4 = f (yi+3 . Then {yn } converges to the two cycle . and for m < y < y. both subsequences converge and their limits form a two cycle. yi+1 is diﬀerent from y.3 Suppose M ≥ yi+2 ≥ yi ≥ y ≥ yi+1 ≥ yi+3 ≥ m and not both yi+2 and yi+3 are equal to y. .

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

y0 } that are on opposite sides of the equilibrium y. Remark. Since f is increasing in the second argument and yi ≥ yi+1 we get yi+2 = f (yi+1 . f (yi+1 . . solving for limn→∞ yi+2n gives limn→∞ yi+2n = M . then yi+2n+1 = y for all n. To show yi+3 = f (yi+2 . ≥ yi+2 ≥ yi ≥ y ≥ . yi+1 ). But then yi = yi+1 = y.66) converges to this two cycle. The above sequence of lemmas leads to the following result: Theorem 6. since m ≤ yi+1 ≤ y. the second case is similar. as well as 2 yi+3 = yi+1 . as required. Since yi+2 = f (yi+1 ..122 Chapter 6.66) possesses the two-cycle solution (6. . . . Proof. Then every oscillatory solution of Eq(6. ≥ y ≥ yi+1 ≥ yi+3 ≥ . If c = m. ≥ m. . .9. and our solution is a two cycle. . . To get equality we must have yi+1 = yi or equivalently yi+2 = yi . yi+1 ) ≥ yi+1 . We will consider the ﬁrst case. yi ) ≤ yi . we have yi+1 ≥ yi . We also obtain the following lemma: Lemma 6. 2)-Type Equations Proof. so either c = y or c = m. we must show yi+2 ≥ yi+1 . Thus yi ≥ yi+1 . The case m ≤ yi ≤ yi+2 ≤ y ≤ yi+1 ≤ M is similar. .9.6 that all such pairs belong to the stable manifold of the two cycle. or M ≥ . . Thus by (2) of the previous lemma yi ≥ yi+1 .66) are generated by initial values {y−1 . . there is no i such that either M ≥ yi ≥ yi+2 ≥ . in which case yi+2n+1 = m for all n.69). and that Eq(6. By (3) of the previous lemma. The value c will be one value in a two cycle. ≥ yi+3 ≥ yi+1 ≥ . yi+1 ) ≤ yi+1 . Since {yi+2k+1 } is a bounded monotonic sequence it has a limit c. Thus equality occurs only in the case of a two cycle.6 Assume that p < q.6 Unless {yn } is a period-two solution. and by (3) yi ≥ yi . This can only happen if yi+2n = M 2 for all n. yi ) ≥ f (yi+1 . (2. Since some of the oscillatory solutions of Eq(6. If c = y. it follows by Theorem 6. ≥ m.9. which also gives a two cycle. Thus yi+2 ≥ yi+1 . and solving for yi+2n we see yi+2n = y also.

5.4 Global Stability Analysis When p > q Here we have the following result: Theorem 6.77) and observe that when p > q. set f (x. Also 1 ≤ f (x. . . the only solution of the system M= is m = M. . y > 0. the function f (x. y) is increasing in x for each ﬁxed y. We will employ Theorem 1. b. .10 Open Problems and Conjectures a.10.9.6. Then the positive equilibrium y of Eq(6. cxn + dxn−1 n = 0. Open Problems and Conjectures 123 6.1 (See [64]) Assume that (a) Investigate the forbidden set F of the diﬀerence equation xn+1 = axn + bxn−1 xn .66) is globally asymptotically stable. x0 ) ∈ F. investigate the asymptotic behavior and the periodic ∞ nature of the solution {xn }n=−1 .4. and is decreasing in y for each ﬁxed x.7 Assume that p>q and p ≤ pq + 1 + 3q. c. Now the result is a consequence of Theorem 1. 1. 2 pM + m . Finally observe that when (6. d ∈ R.9. / (b) For every point (x−1 .77) holds. qx + y (6. y) = px + y .5. To this end.4. . qm + M 6. Proof.10. y) ≤ p q for all x. qM + m m= pm + M . Open Problem 6.

4. every positive solution converges to a. Show that the equation yn+1 = p + yn−1 . .6 for the autonomous case. not necessarily prime. (Note that by Theorem 6.) Chapter 6. .5. q ∈ (0. this conjecture has been conﬁrmed for q ≤ 1 + 4p. (2.) Conjecture 6. Show that every positive solution of Eq(6. . . period-two solution.2.1 Assume p. are convergent sequences of real numbers with ﬁnite limits.10.6. Bn ∈ R for n = 0. βn . An + Bn xn n = 0. See also [36]. Show that every positive solution of Eq(6. 1 + yn n = 0. ∞). ∞).124 (See Section 1. (See Section 1. we need only to conﬁrm this conjecture for p > 2(q +1) and q ≥ 1.11) has a ﬁnite limit. An . 1.) . q ∈ (0. . ∞).3 Assume p ∈ (0. .3. 1. . .3. (Note that by Theorem 6.) Conjecture 6.10) has a ﬁnite limit.10. has a positive and monotonically decreasing solution. 2)-Type Equations Open Problem 6.2 Assume that αn . 1. (Note that by Theorem 6.) Conjecture 6.4.10. . Investigate the forbidden set and the asymptotic character of solutions of the nonautonomous Riccati equation xn+1 = αn + βn xn .2 Assume that p.10. .

7.10.) Open Problem 6.3 Assume that p. 125 Show that every positive solution of Eq(6. q ∈ (0.5 Find the set of all initial conditions (y−1 .9. q ∈ (0. ∞) and q > 1 + 4p. (See Section 6.10.66).) Open Problem 6.4 Assume that p.10. (Note that by Theorem 6.66) has a ﬁnite limit. q ∈ (0.10.6. this conjecture has been conﬁrmed for p ≤ pq + 1 + 3q.27).10. as n → ∞. Open Problem 6. Conjecture 6. Open Problems and Conjectures Conjecture 6.4 Assume that p.6.5 Assume p > q. ∞). Find the basin of attraction of the period-two solution of Eq(6.10. . Show that every positive solution of Eq(6. ∞) and q > pq + 1 + 3p. Find the basin of attraction of the period-two solution of Eq(6. y0 ) ∈ R × R through which the solution of the equation yn+1 = yn + yn−1 1 + yn is well deﬁned and converges to 1.27) either converges to a ﬁnite limit or to a two cycle.

extend your result by introducing arbitrary real parameters in the equation.80) (6.8 Assume α.10. Open Problem 6. .78) (6. . Investigate the asymptotic behavior and the periodic nature of solutions of the diﬀerence equation αxn + βxn−1 + γxn−2 . 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. ∞). .83) yn − yn−1 yn − 1 yn − yn−1 1 − yn−1 1 − yn−1 yn − yn−1 yn + yn−1 .10.79) (6. as n → ∞.10. Open Problem 6. A. n = 0. . xn+1 = Axn + Bxn−1 + Cxn−2 Extend and generalize. 1. .7 For each of the following diﬀerence equations determine the “good” set G ⊂ R × R of initial conditions (y−1 . (6.Then for every (y−1 . C ∈ (0. (2. β.82) (6.84) yn − yn−1 For those equations from the above list for which you were successful. y0 ) ∈ G.126 Chapter 6. y0 ) ∈ R × R through which the equation is well deﬁned for all n ≥ 0. 2)-Type Equations Open Problem 6.81) (6. γ. B.6 Find the set of all initial conditions (y−1 . y0 ) ∈ R × R through which the solution of the equation yn+1 = yn + yn−1 1 + yn−1 is well deﬁned and converges to 1.

. n = 0. (a) Find the set B of all initial conditions (y−1 .10. 1. .87) (6.92). Investigate the global character of all positive solutions of Eqs(6.23) are bounded. y0 ) ∈ B. y0 ) ∈ (0.10. 1. qn + yn−1 pn yn + yn−1 .89) (6. n = 0. . 1. n = 0. 1. n=−1 . .85)-(6.88) (6.92) pn + qn yn . . ∞) × (0. n = 0. . 1. qn yn + yn−1 pn yn + qn yn−1 . Open Problems and Conjectures 127 ∞ Open Problem 6. (b) Let (y−1 . . . . . . 1. . 1. n = 0. n = 0.10.86) (6. 1 + yn pn yn + yn−1 . Extend and generalize. . 1 + yn pn + yn−1 . . . 1. . p = n→∞ pn lim lim and q = n→∞ qn .11 Assume q ∈ (1.91) (6. . . . . . Investigate the asymptotic behavior of {yn }∞ . 1 + yn−1 yn + pn . qn + yn n = 0. .10 Assume that {pn }∞ and {qn }∞ are period-two sequences n=0 n=0 of nonnegative real numbers. 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 . ∞). ∞) such that the solutions {yn }∞ n=−1 of Eq(6.9 Assume that {pn }n=0 and {qn }∞ are convergent sequences of n=0 nonnegative real numbers with ﬁnite limits.10. (6. Open Problem 6.90) (6.85) (6. n = 0. .6. yn + qn yn−1 pn + qn yn−1 . .

See [14]. . (6. and the asymptotic behavior of the solution {xn }∞ n=−1 of Eq(6. Investigate the boundedness character. 1 β = 1 and γ = . y0 ) ∈ B. ∞) × (0. Open Problem 6. are nonnegative. n = 0. and γ ∈ (0. and [2]) Assume α ∈ (1.) Let A ∈ (0. 3). 2)-Type Equations (a) Find the set B of all initial conditions (y−1 . (b) Let (x−1 . [3]. y0 ∈ (0. .10. . ∞) such that the solutions {yn }∞ n=−1 of Eq(6. 1) with α + β > 1 + β . y0 ) ∈ (0. the periodic nature. Chapter 6. .6 Assume x0 .10. ∞ (b) Let (y−1 . 1. ∞) through which the solutions {xn }∞ n=−1 of the diﬀerence equation xn+1 = (1 − xn − xn−1 )(1 − e−Axn ). ∞). ∞) and that α ∈ (2. n = 0.12 Assume q ∈ (1. (c) Extend and generalize.13 (A Plant-Herbivore System.93) Conjecture 6.10. (6.93) is globally asymptotically stable.10. (a) Determine the set G of initial conditions (y−1 . y0 ) ∈ (0. ∞). (2.94). .47) are bounded.128 Open Problem 6.94) . Open Problem 6. γ Obtain conditions for the global asymptotic stability of the positive equilibrium of the system xn+1 = αxn βxn +eyn yn+1 = γ(xn + 1)yn . ∞).14 (Discrete Epidemic Models. ∞). 2 Show that the positive equilibrium of System (6. See [9]. β ∈ (0. . ∞) × (0. x0 ) ∈ G. Investigate the asymptotic behavior of {yn }n=−1 . 1. .

n = 0. Open Problems and Conjectures Open Problem 6. Investigate the global character of all positive solutions of the system .6. b ∈ (0. . .99) (6.97) (6. . . qyn + yn−k n = 0.101) . 1. . .17 Assume that p. qyn + yn−k yn + pyn−k . 1) and xn+1 = αxn e−yn + β yn+1 = αxn (1 − e −yn β ∈ (1. . ∞) with c1 + c2 > 0. q ∈ [0. 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 . ∞). . 1 + yn p + yn−k .16 (A Population Model) Assume α ∈ (0. with positive initial conditions.10. and c1 . .15 (The Flour Beetle Model. .100) (6. . . .98) (6. 1. ∞) and k ∈ {2. n = 0. yn + qyn−k p + qyn−k . . Investigate the global behavior of all positive solutions of each of the following diﬀerence equations: p + qyn yn+1 = . 1). 1. yn + q pyn + yn−k .10. . ∞). . . . n = 0. Open Problem 6. Assume a ∈ (0. n = 0.10. .95) 1 + yn−k yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn + p . n = 0. n = 0. . . . . 1. 1. . Open Problem 6. . . (6. 1. .}. ) which may be viewed as a population model. . . n = 0. c2 ∈ [0.96) (6. 1. 1. . (6. n = 0. 1. . .10. q + yn−k pyn + yn−k . 3. See [48]).

.10. n = 0. .19 Assume that every positive solution of an equation of the form xn+1 = f (xn . 1. . .10. Open Problem 6. 1. . 2)-Type Equations Open Problem 6.27). (6. . n = 0.20 Assume that Eq(6. Obtain necessary and suﬃcient conditions on f so that the same two cycle is a locally asymptotically stable solution of Eq(6. Obtain necessary and suﬃcient conditions on f so that every positive solution of the equation xn+1 = f (xn−2 . also converges to a period-two solution. .130 Chapter 6. yn−1 ).103). . (2. xn−1 ). Extend and generalize. . which extends and uniﬁes the global asymptotic stability results for Eqs(6.102) has a two cycle which is locally asymptotically stable.103) Open Problem 6.18 Obtain a general result for an equation of the form yn+1 = f (yn .10.11) and (6. (6.102) converges to a period-two solution. . xn−1 ). 1. n = 0.

.1) (7. .(7. . . 1. (7. .2) Eq(1) contains the following three equations of the (1. . .2 The Case β = γ = 0 : xn+1 = α yn A α A+Bxn +Cxn−1 This is the (1. 1 + pyn + qyn−1 αB A2 and q = 131 n = 0.4) αC . (7. A2 . . .1 Introduction xn+1 = xn+1 = and xn+1 = γxn−1 . (7. 3)-type equations are positive and that the initial conditions are nonnegative. n = 0.3) α . A + Bxn + Cxn−1 n = 0. 1. .Chapter 7 (1. . A + Bxn + Cxn−1 βxn . A + Bxn + Cxn−1 n = 0. . 3)-type: Please recall our classiﬁcation convention in which all the parameters in the above (1. . 7. 3)-type Eq. 1. 3)-Type Equations 7. 1.1) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= 1 .

Then the zero equilibrium of Eq(7.4) has no prime period two solutions. The following result is now a straightforward consequence of either the stability trichotomy Theorem 1.5) is globally asymptotically stable.5) is globally asymptotically stable.4) is locally asymptotically stable for all values of the parameters and that Eq(7. 1].4. 1.3 The Case α = γ = 0 : xn+1 = A yn C βxn A+Bxn +Cxn−1 This is the (1. .2) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where yn .4.1 Assume p ≤ 1. 3)-Type Equations One can easily see that the positive equilibrium of Eq(7. (1.2 Assume that y−1 . Then the positive equilibrium of Eq(7. The following result is a straightforward consequence of Theorem 1. (7.4) is globally asymptotically stable. 3)-type 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 .1 The positive equilibrium of Eq(7.1. .4.2. q Theorem 7. y0 ∈ (0.4. . ∞) and p > 1.132 Chapter 7.3.3. .7 in the interval [0. 1 + pyn + qyn−1 p= n = 0.5 in the interval [0. 7. It also follows by applying Theorem 1. A C Eq(7.2. or Theorem 1. p ]. Theorem 7.5) β B and q = . Theorem 7.4.3. q+1 The following result is a straightforward consequence of Theorem 1.

The Case α = β = 0 : xn+1 = γxn−1 A+Bxn +Cxn−1 133 γxn−1 A+Bxn +Cxn−1 7. 1. . C (7. .6) is globally asymptotically stable when p≥1 and is unstable (a repeller) when p < 1.6) has a prime periodtwo solution .6) and q = Eq(7. ψ. . 1+q The subsequent result is a straightforward application of Theorems 1. . . .3) which by the change of variables xn = reduces to the equation yn+1 = where p= yn−1 . . φ. Theorem 7.1 and 1. if and only if p < 1. . Then the positive equilibrium of Eq(7.1. p + qyn + yn−1 A γ n = 0. .3. φ.4 The Case α = β = 0 : xn+1 = γ yn C This is the (1. (b) Assume that p < 1. . ψ.6) is locally asymptotically stable when q<1 and is unstable (a saddle point) when q > 1. 3)-type Eq(7. Concerning prime period-two solutions one can see that Eq(7.6) always has the zero equilibrium and when p<1 it also has the unique positive equilibrium y= 1−p . B .4.1 (a) The zero equilibrium of Eq(7.7.1.4.

On the other hand. φ. .1 Assume Show that every positive solution of Eq(7.8) 0 p 0 lie in the disk |λ| < 1. 1 − p − φ.7) both eigenvalues of JT 2 7. .4. .5. . .6) has a ﬁnite limit. . 0. . Conjecture 7. 1 − p. . Therefore 1−p the prime period-two solution (7. 1 − p − φ.2 Assume that p<1 Then the period-two solution . 1).7) (7. q ∈ (0. 0. .5 Open Problems and Conjectures p. (7. It follows from Section 2.134 Furthermore when Chapter 7.8) is locally asymptotically stable.6 that here JT 2 0 1−p = 1 p+q(1−p) q(1−p) − p+q(1−p) 1−p . . 2 and q > 1. φ. 2 and in view of (7. 0. . . . (1. . when p < 1 and q = 1 all prime period-two solutions of Eq(7. 1 − p.6) is locally asymptotically stable. 0. Proof. 1 − p. with 0 ≤ φ ≤ 1 − p and φ = Theorem 7. . 1 − p. . . of Eq(7. 3)-Type Equations p < 1 and q = 1 the only prime period-two solution is . . . .6) are given by .

. ﬁnd the set G of all points (x−1 . . 1.2 Assume p < 1 and q ≥ 1.9)-(7. x0 ) ∈ G be an initial point through which the corresponding equation is well deﬁned for all n ≥ 0. periodtwo solution. Find the basin of attraction of the two cycle .6) converges to a. .9) (7.7. ∞). 1 − p.6). Open Problem 7. not necessarily prime. . Open Problems and Conjectures Conjecture 7. 1 − p. Open Problem 7. .5. 135 Show that every positive solution of Eq(7.2 Investigate the asymptotic character and the periodic nature of all positive solutions of the diﬀerence equation xn+1 = where p.5.5. ∞). .10) (7. let (x−1 . 1) and q ∈ (1.11) Now for each of the equations (7.5.11). . q ∈ [0. 1 + pxn + qxn−1 + xn−2 n = 0. Open Problem 7. .1 Assume p ∈ (0.5. of Eq(7. . x0 ) ∈ R × R through which the equation is well deﬁned for all n ≥ 0: xn+1 = xn+1 = xn+1 = 1 1 + xn + xn−1 xn 1 + xn + xn−1 xn−1 . 0. and investigate ∞ the long term-behavior of the solution {xn }n=−1 . xn−2 . . 1 + xn + xn−1 (7. 0. Extend and generalize.3 For each of the equations given below.

14).136 Chapter 7. . . .4 Assume that {pn }∞ and {qn }∞ are convergent sequences of n=0 n=0 nonnegative real numbers with ﬁnite limits. 1 + pn yn + qn yn−1 yn . . 1 + pn yn + qn yn−1 yn−1 . pn + qn yn + yn−1 n = 0.5. (7. Extend and generalize. (1. Investigate the global character of all positive solutions of Eqs(7.5 Assume that {pn }∞ and {qn }∞ are period-two sequences of n=0 n=0 nonnegative real numbers.12)-(7.5. n = 0. .14) Open Problem 7. 1. n = 0. . . lim Investigate the asymptotic behavior and the periodic nature of all positive solutions of each of the following three diﬀerence equations: yn+1 = yn+1 = yn+1 = 1 . 3)-Type Equations Open Problem 7. .13) (7. 1.12) (7. . 1. . p = lim pn n→∞ and q = n→∞ qn . .

(8. Eq(8. and the denominators are always positive. Eq(8. 3)-type: xn+1 = xn+1 = and xn+1 = α + βxn + γxn−1 . . . B is reduced to a (2. . . (8.5). namely. 1)-Type Equations 8. .1) (8. .4) (see Section 6. the initial conditions are nonnegative. C . .1) is a linear diﬀerence equation and can be solved explicitly. A α + βxn + γxn−1 . . 1.3) n = 0. .Chapter 8 (3. Cxn−1 n = 0. yn+1 = (α + βγ ) + γyn−1 B . . 2)-type equation of the form of Eq(6.2) by the change of the variables xn = yn + β .2) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. 1. 1. 1. n = 0. Bxn Eq(1) contains the following three equations of the (1. . . β + Byn n = 0.1 Introduction α + βxn + γxn−1 .3) by the change of the variables xn = yn + 137 γ . . Eq(8. .

φ.1 Assume γ > β.4) be a period-two solution of Eq(8. x0 ) ∈ R × R through which the equation 1 + xn + xn−1 xn+1 = xn is well deﬁned for all n and for these initial points determine the long-term behavior of the solutions {xn }∞ . 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 }∞ . (8. φ. ∞) through which the solutions of Eq(8.138 Chapter 8. (3.3) namely.4). . . Determine the basin of attraction of (8. n=−1 Open Problem 8. yn+1 (α + βγ ) + βyn C . .2.2). 1. . . n=−1 . Conjecture 8. ∞) × (0. 1)-Type Equations is reduced to a (2. 2)-type equation of the form of Eq(6. Open Problem 8.2.2. Hence there is nothing else remaining to do about these equations other than to pose some Open Problems and Conjectures. Determine the set of initial conditions (x−1 . ψ.2) which converges to (8. . ψ.1 Show that every positive solution of Eq(8. = γ + Cyn−1 n = 0.2) are bounded.2. .4 Determine the set G of all initial points (x−1 . .3) converges to the positive equilibrium of the equation.4). .2 (a) Assume γ = β and let . x0 ) ∈ (0. Open Problem 8. (b) Let {xn }∞ n=−1 be a positive solution of Eq(8.2.3 Determine the set G of all initial points (x−1 .2) (see Section 6. .2 Open Problems and Conjectures Open Problem 8. 8. Determine the values of φ and ψ in terms of the initial conditions x−1 and x0 .

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.

Then this solution is oscillatory and the sum of the lengths of two consecutive semicycles. y4k ≤ 1.4 Assume p = q(qr + 1) and let {yn }∞ n=−1 be a solution of Eq(9. In this case Eq(9. More precisely. and the only equilibrium point is y = q.10) Proof. 1 + yn + ryn−1 n = 0.5). (iv) y−1 < q and y0 ≤ q =⇒ y4k−1 < q. (ii) y−1 < q and y0 ≥ q =⇒ y4k−1 < q.9) follows by straightforward computation.2. Then yn+1 − q = Furthermore when qr < 1. y4k ≥ 1. 1 + yn + ryn−1 lim yn = y. and y4k+2 ≥ q.4) becomes yn+1 = q + rq 2 + qyn . y4k ≤ 1. . and y4k+2 ≥ q.2. 2 Here we present a semicycle analysis of the solutions of Eq(9.4).4). is equal to four. (9. ∞) and Eq(9. (iii) y−1 > q and y0 ≥ q =⇒ y4k−1 > q. y4k+1 < q.4) when p = q(1 + rq). 3)-Type Equations qr (q − yn−1 ). 1. (2.10) is a consequence of the fact that in this case qr ∈ (0.4) has no period-two solution. . 2 . y4k ≥ 1. y4k+1 < q. and y4k+2 ≤ q. y4k+1 > q. y4k+1 > q. The limit (9.9) (9. The proof follows from identity (9. Proof. Identity (9.3 Suppose that p = q + rq 2 and let {yn }∞ n=−1 be a nontrivial solution of Eq(9. . and y4k+2 ≤ q. the following statements are true for all k ≥ 0: (i) y−1 > q and y0 ≤ q =⇒ y4k−1 > q. then n→∞ Chapter 9. excluding the ﬁrst. Theorem 9.146 Lemma 9.

there exists N > 0 such that one of the following three cases holds: (i) yN > q.2.3. First assume that p ≤ q. q−1+ (q−1) +4p ] if p ≤ q. yN +1 > q. qr p−q . q] qr [q. p > q(qr + 1). The Case γ = 0 : xn+1 = α+βxn A+Bxn +Cxn−1 147 9.4) described in Lemma 9. Let {yn }n=−1 be a solution of Eq(9. Now assume for the sake of contradiction that yn > b for n > 0.2. also assume that the solution {yn }∞ n=−1 is not eventually in the interval I.4) is eventually trapped into one of the three invariant intervals of Eq(9.4). q−1+ (q − 1)2 + 4p 2 Hence. the following is true: Lemma 9. yN +1 < p−q . Then every solution of Eq(9. every solution of Eq(9. More precisely. (ii) yN > q. then yN +1 < b.2. ∞ Proof. qr and yN +2 < .4 Global Asymptotic Stability The following lemma establishes that when p = q(qr + 1). Then yn > y for n ≥ 0 and so which is a contradiction. yN ≤ b. qr n→∞ lim yn = y ∈ I and yN +2 < p−q .1. (1 + yn + ryn−1 )(1 + b) n ≥ 0. for the sake of contradiction. p−q ] qr if if q < p < q(qr + 1). Set b= and observe that yn+1 − b = (q − p)(yn − b) − r(p + qb)yn−1 . Then clearly. y ∈ I. Next assume that q < p < q(qr + 1) and.4) lies eventually in I.9.2. 2 I= [ p−q . Then by Lemma 9.2.5 Let I denote the interval which is deﬁned as follows: √ 2 [0. if for some N .

7) from which it follows that for j ∈ {0.4.1.4) is globally asymptotically stable. 3)-Type Equations ≤ yN +1 ≤ q. 2. By using the monotonic character of the function f (x. For example. then yN + ryN − p > 0 and 2 if yN ≤ p−q . and (q − 1)2 (r − 1) ≤ 4p. The proof when p > q(qr + 1) is similar and will be omitted. we can obtain some global asymptotic stability results for the solutions of Eq(9. 1.2. 3}. Then the equilibrium y of Eq(9. .2. the following results are true for Eq(9. qr Also observe that 2 if yN ≥ q. Proof. p−q qr Chapter 9.4): Theorem 9. then yN + ryN − p < 0. (b) Assume that either (iii) r > 1 (ii) r ≤ 1.5. y) = p + qx 1 + x + ry 2 in each of the intervals in Lemma 9.7 and 1. together with the appropriate convergence Theorems 1.4) is globally asymptotically stable.4. each subsequence {yn+4k+j }∞ with all its terms outside k=0 the interval I converges monotonically and enters in the interval I. and yN +2 < p−q .4).148 (iii) yN > q. 1+r Then the equilibrium y of Eq(9. qr The desired contradiction is now obtained by using the Identity (9.4 (a) Assume that either p ≥ q + q 2 r. or q . p< p≤q or q < p < q + q2r and that one of the following conditions is also satisﬁed: (i) q ≤ 1. (2.

It is easy to check that y ∈ 0. This system has the form m= p + qm 1 + m + rM and M = p + qM . .2. A2 and r = . . 0.5. . this is the case if Condition (i) is satisﬁed.7 and 1.4. 1 + yn + ryn−1 q= γ . 1 + M + rm Hence (M − m)(1 − q + M + m) = 0.2. The proof when q < p < q + q 2 r holds follows from Lemma 9. √ q−1+ (q−1)2 +4p (b) In view of Lemma 9.4. 2 2 √ 2 +4p q−1+ (q−1) . For instance. 2 decreases in y. We will employ Theorem 1. y0 ∈ 0.2.11) A yn B αB . 3)-type Eq(9.5 and so it remains to be shown that if m = f (m. m = M from which the result follows. A n = 0. B (9. the function f increases in x and and that in the interval 0.2) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + qyn−1 . M ) and M = f (M. then M = m. 2 9. m) then M = m.1 and 9. The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 149 (a) The proof follows from Lemmas 9. C .2. then yn ∈ 2 √ √ q−1+ (q−1)2 +4p q−1+ (q−1)2 +4p for all n ≥ 0.3 The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 This is the (2.3. Now if m + M = q − 1.2. 1.1 and Theorem 1.9.4. Clearly now if Condition (ii) or (iii) is satisﬁed.1 we see that when y−1 .4. If m + M = q − 1 then m and M satisfy the equation (r − 1)m2 + (r − 1)(1 − q)m + p = 0. .4 and Theorems 1.

φ. . be a period-two cycle of Eq(9. Theorem 9. 1 + φ + rψ It now follows after some calculation that the following result is true. . .12) 9. ψ. and the convergence of solutions in certain regions of initial conditions. .11) is locally asymptotically stable when either q≤1 or q > 1 and (r − 1)(q − 1)2 + 4pr2 > 0.11) is unstable.150 Chapter 9. By employing the linearized stability Theorem 1. when q > 1 and (r − 1)(q − 1)2 + 4pr2 < 0. .11). Eq(9. Here we make some simple observations about linearized stability. Then φ= p + qφ 1 + ψ + rφ and ψ = p + qψ .1 (a) The equilibrium y of Eq(9.12) holds. φ. φ.12) holds.3.11) has a unique equilibrium y which is positive and is given by y= q−1+ (q − 1)2 + 4p(r + 1) 2(r + 1) . . .1 Eq(9.1 we obtain the following result. ψ. (2. invariant intervals. φ. .11) has a prime period-two solution .3. . 3)-Type Equations This equation has not been investigated yet. . ψ.1. (b) The equilibrium y of Eq(9. and more precisely a saddle point. (9. t+ r 1−r . 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. Lemma 9. Furthermore when (9.3.1 Let Existence and Local Stability of Period-Two Cycles . ψ. if and only if (9. . .

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

∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1. This inequality is equivalent to (q − pr + qφ) (q − pr + qψ) − (1 + ψ + rφ)2 (1 + φ + rψ)2 < 0. (9.1 that both eigenvalues of JT 2 unit disk |λ| < 1. (2. if and only if |S| < 1 + D < 2. (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 . ψ) (φ. ψ) (φ. Inequality (9. ψ) − (φ. ψ) + (φ.14). Observe that.16) First we will establish Inequality (9.15) (9.14) (9.1. 3)-Type Equations ∂h (p + qψ)(pr − q − qψ) (φ.13) (9. ψ) = . ∂u (1 + ψ + rφ)2 (1 + φ + rψ)2 ∂h (p + qψ)(p + qφ) q − pr + qφ (φ. ψ).152 Chapter 9. Thus . ψ) ∂u ∂v ∂h ∂g ∂h ∂g (φ.13) is equivalent to the following three inequalities: D<1 S <1+D −1 − D < S. 2 (1 + φ + rψ)2 ∂v (1 + ψ + rφ) (1 + φ + rψ)2 Set S= and D= ∂h ∂g (φ. ψ) = + .

In summary. Finally Inequality (9. The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 153 (q − pr + qφ) (q − pr + qψ) − (1 + ψ + rφ)2 (1 + φ + rψ)2 = −pr3 q 2 + 2q 2 r − q 2 r2 − q 2 + 3p2 r3 − 2p2 r4 − 3rq 3 + 3qpr3 − 5qpr2 r2 (1 − r) + 2qpr + 4pr2 q 2 + 2q 3 + q 4 r − p2 r2 − 2prq 2 − q 4 + r2 q 3 r2 (1 − r) Note that (2r2 p − rp + q − q 2 − qr + q 2 r) (−r2 p + rp − q + q 2 + qr) = . Also by using Condition (9. Next we turn to Inequality (9. After some lengthy calculation one can see that this inequality is also a consequence of Condition (9. we have established the following result: .3.12).12). r2 (1 − r) −r2 p + rp − q + q 2 + qr = rp(1 − r) + q(q − 1) + qr > 0. 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 < 0.16) 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 a consequence of Condition (9.9.15).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.

Furthermore when (9.2 Eq(9.3. r q when pr ≥ q + q2 . b].12) holds.3. . yk ∈ [0. φ.2 Eq(9. .11) we see that when yk−1 . 9. this period-two solution is locally asymptotically stable. φ.11) possesses the following invariant intervals: (a) 0.12) holds. (2.3. gives a list of invariant intervals for Eq(9. which can be established by direct calculation. if and only if (9. ψ. r when 1 < q < pr < q + q2 q − .154 Chapter 9.2 Invariant Intervals The following result. (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. . 1 + yk + qyk−1 1 + ryk−1 The proof follows by induction. 3)-Type Equations Theorem 9. . (b) pr − q q . Using Eq(9. . q−1+ (q − 1)2 + 4pr 2r when pr ≤ q. Lemma 9. q r (c) q pr − q .11).11) has a unique prime period-two solution . r r Proof. ψ. . then yk+1 = p + qyk−1 p + qyk−1 ≤ = g(yk−1 ) ≤ g(b) ≤ b. .

r The proof follows by induction. q q q = . yk ∈ p + qyk−1 = f (yk .r q is increasing in y for x > pr−q . yk ∈ q then q pr − q p + qyk−1 pr − q = f (yk . y) = is decreasing in y for x < q pr−q . y) = p + qy 1 + x + ry pr−q q . . Now by using the monotonic character of the function f (x. q p + qy 1 + x + ry . < r + (r + 1)q q The proof follows by induction.9. yk−1 ) ≤ f .3. The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 155 (b) It is clear that the function f (x. then yk+1 = Using Eq(9. y) is decreasing in both arguments and the condition pr > q + rq2 we obtain yk+1 = q q p + qyk−1 = f (yk . r By using the fact that f (x.11) we see that when yk−1 . q r q = . yk−1 ) ≥ f . y) and the condition 2 q < pr < q + qr we obtain yk+1 = p + qyk−1 = f (yk . 1 + yk + qyk−1 r r = pr + q 2 pr − q . yk−1 ) ≥ f 1 + yk + qyk−1 pr − q pr − q . (c) It is clear that the function f (x. q r pr−q . ≥ yk+1 = 1 + yk + ryk−1 r q q . Using Eq(9.11) we see that when yk−1 . yk−1 ) ≤ f 1 + yk + ryk−1 pr − q q . 2 .

11): Theorem 9. the function f decreases in x and increases 2r in y. Then every solution of Eq(9. The result now follows by employing Theorem 1. 2r 2r √ q−1+ (q−1)2 +4pr and that in the interval 0.11) with initial conditions in the invariant interval 0.4. 2 q−1+ √ (q−1)2 +4pr (ii) r < 1 and (r − 1)(q − 1)2 + 4pr2 ≥ 0.4 (a) Assume that 1 < q < pr < q+ qr − q and that one of the following r conditions is also satisﬁed: (i) r ≥ 1.3.2 we see that when y−1 . .8) we can obtain some convergence results for the solutions with initial conditions in the invariant intervals. then Proof. y) = in each of the intervals in Lemma 9. for all n ≥ 0.4. (iii) r < 1 and (r − 1)(q − 1)2 + 4pr2 ≥ 0.156 Chapter 9. y0 ∈ 0.6.4. . the following results are true for Eq(9.3. It is easy to check that y ∈ 0. (2.3.3. q−1+ (q − 1)2 + 4pr 2r converges to the equilibrium y. (ii) r ≥ 1.3. 3)-Type Equations 9. 2 Theorem 9.5-1. 2r √ √ q−1+ (q−1)2 +4pr q−1+ (q−1)2 +4pr yn ∈ 0.3 Assume that pr ≤ q and that one of the following conditions is also satisﬁed: (i) q ≤ 1.3 Convergence of Solutions p + qy 1 + x + ry By using the monotonic character of the function f (x. In view of Lemma 9. together with the appropriate convergence theorem (from among Theorems 1. For example.2.

3.3.6.2 we conclude that yn ∈ for all n ≥ 0.11) with initial conditions in the invariant interval q pr − q .5 Assume r ≥ 1 and pr ≤ q.11) becomes yn+1 = qr + q 2 + qr2 yn−1 .3. 2 q pr−q . Then the positive equilibrium of Eq(9. The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 157 Then every solution of Eq(9. .3. q2 r q2 . . Then every solution of Eq(9.4 Global Stability By applying Theorem 1. (b) Assume that pr > q + qr .3 to Eq(9. q r converges to the equilibrium y.7.11).11) with initial conditions in the invariant interval pr − q q . r 9. The result now follows by employing Theorem 1. pr−q r q for all n ≥ 0. It is easy to (b) In view of Lemma 9.2 we conclude that yn ∈ pr−q q .3.r q pr−q q . pr−q and that in the r q in both x and in y. q r interval q .5 Semicycle Analysis When pr = q + Here we present a semicycle analysis of the solutions of Eq(9. Theorem 9. In and the only positive equilibrium is y = q .4. . r .4.11) is globally asymptotically stable. (a) In view of Lemma 9. Proof.11) when pr = q + this case Eq(9. r2 + r2 yn + r3 yn−1 n = 0. 1. we obtain the following global asymptotic stability result. It is easy to 9.4. The result now follows by employing Theorem 1.r q 2 check that y ∈ and that the function f (x.9.3. y) decreases in x and increases in y. r q converges to the equilibrium y. the function f decreases check that y ∈ q .

3. Eq(9. The proof follows from the relation yn+1 − q q qr2 = ( − yn ) 3 . γ q= B .1 and Theorem 1. The equilibrium points of Eq(9. invariant intervals.6 Suppose that pr = q + qr and let {yn }∞ n=−1 be a nontrivial solution of Eq(9. 3)-type Eq(9. . 1. the solution oscillates about the equilibrium y with semicycles of length one. . r + qyn + yn−1 β .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. (2. Proof. (9.1. . 3)-Type Equations Theorem 9.1. C and r = A . Here we make some simple observations about linearized stability. .3) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= pyn + yn−1 . C n = 0.18) The following theorem is a consequence of Theorem 1. r r r (1 + yn + ryn−1 ) 2 9.17) also possesses the unique positive equilibrium y= p+1−r .3. Then after the ﬁrst semicycle. and the convergence of solutions in certain regions of initial conditions. q+1 (9.158 2 Chapter 9.11).17) γ yn .4 The Case α = 0 : xn+1 = βxn +γxn−1 A+Bxn +Cxn−1 This is the (2. γ This equation has not been investigated yet.

1.18) holds. . The following result is a consequence of Theorem 1. .4. . φ. Theorem 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) = 0.17) about its positive equilibrium y is zn+1 − p − q + qr q−p+r zn − zn−1 = 0. if and only if (9. q−1 . Furthermore when (9.4.1. it follows from Section 2. Then the zero equilibrium of Eq(9.5 that the following result is true. φ.17) has a prime period-two solution .3 Eq(9.4.20) holds.1 βxn +γxn−1 A+Bxn +Cxn−1 159 (a) Assume that p + 1 ≤ r. Then the zero equilibrium of Eq(9. The linearized equation associated with Eq(9.9. The Case α = 0 : xn+1 = Theorem 9.17). . .17) is locally asymptotically stable when q + r < 3p + 1 + qr + pq.17) is unstable.2 Assume that (9.1. ψ. . (9. (b) Assume that p + 1 > r. ψ. and unstable (a saddle point) when q + r > 3p + 1 + qr + pq. Furthermore the zero equilibrium is a saddle point when 1−p<r <1+p and a repeller when r < 1 − p.4.17) is globally asymptotically stable. . . Theorem 9.19) Concerning prime period-two solutions for Eq(9. (p + 1)(q + 1) (p + 1)(q + 1) n = 0. . Then the positive equilibrium of Eq(9.20) (9.

17).4.1 Eq(9. Lemma 9. K2 −p [0.160 Chapter 9.17) possesses the following invariant intervals: (a) 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 . 3)-Type Equations 9. when p > q 2 + r where ((r − q)(p − q 2 ) + qr)2 + 4q 3 r(p − q 2 ) 2q(p − q 2 ) . when q 2 − qr ≤ p < q 2 and p + q > r.1 Invariant Intervals The following result. qr p − q2 when q 2 < p ≤ q 2 + r and p + q > r. where K > 0 is an arbitrary number and pr . q2 where K2 = when q 2 > p + qr. ((q − r)(q 2 − p))2 + 4p3 qr2 2pqr (q − r)(q 2 − p) + . (b) 0. K] when q 2 − qr ≤ p < q 2 and p + q ≤ r. . K] when p = q 2 where K > 0 is an arbitrary number. and [0. q2 pr −p (q − r)(p − q 2 ) − qr + [0. which can be established by direct calculation.4. (2. K] when q 2 < p ≤ q 2 + r and p + q ≤ r. K1 p − q2 K1 = (c) 0. gives a list of invariant intervals for Eq(9.

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

K].8. (b) Assume that q 2 + r ≥ p > q 2 and p + q > r.4.4. Then every solution of Eq(9. converges to the equilibrium y. together with the appropriate convergence theorem (from among Theorems 1. converges to the equilibrium y. for any K > 0. we can obtain some convergence results for the solutions with initial conditions in the invariant intervals. Then every solution of Eq(9.5-1.4. for any K > 0. For example. (2.4. converges to the equilibrium y. y) = in each of the intervals in Lemma 9.4. Then every solution of Eq(9. converges to the equilibrium y. Then every solution of Eq(9.1 and Theorem 1.1. the following results are true for Eq(9.4.2 Convergence of Solutions px + qy r + qx + y By using the monotonic character of the function f (x. Then every solution of Eq(9.17) with initial conditions in the interval [0. Then every solution of Eq(9. Assume that q 2 + r ≥ p > q 2 and p + q ≤ r. for any K > 0.4. 2 q2 pr −p qr p − q2 q2 + q − r q+1 . (c) Assume that q 2 − qr ≥ p < q 2 and p + q > r.8).162 Chapter 9.17) with initial conditions in the invariant interval 0. The proof is an immediate consequence of Lemma 9. Assume that p = q 2 and q 2 + q ≤ r.4 (a) Assume that p = q 2 and q 2 + q > r. Proof.17) with initial conditions in the invariant interval 0.17) with initial conditions in the interval [0.17) with initial conditions in the interval [0. converges to the equilibrium y. K].17): Theorem 9. Assume that q 2 − qr ≤ p < q 2 and p + q ≤ r.17) with initial conditions in the invariant interval 0. converges to the equilibrium y. 3)-Type Equations 9. K].

5. K1 p−q 2 3q 2 +q−qr+r . B. ∞).1 Assume Show that every positive solution of Eq(9.20) is not satisﬁed.5. p−q 2 Proof. β.5 Open Problems and Conjectures α.6.4. A.3) is bounded. 2 9.5. K2 converges to the equilibrium y. .5. γ. 1−q 163 and p ≤ Then every solution of Eq(9. 2 Theorem 9. (ii) q ≥ 1. B. A.1 and Theorem 1.4. Show that every positive solution of Eq(9. Then every solution of Eq(9. ∞).5.4. Conjecture 9.4. ∞).1 and Theorem 1.1) has a ﬁnite limit.9. The proof is a consequence of Lemma 9. C ∈ (0.11) with initial conditions in the invariant interval qr . Conjecture 9. γ.11) with initial conditions in the invariant interval converges to the equilibrium y. Conjecture 9.4. A. Show that every positive solution of Eq(9. Proof.4. Open Problems and Conjectures Theorem 9. (iii) q < 1 qr .2 Assume α. C ∈ (0.5 Assume that p > q2 + r and that one of the following conditions is also satisﬁed: (i) p ≤ q + r. The proof is an immediate consequence of Lemma 9.6 Assume that q 2 > p + qr and that Condition (9. C ∈ (0.3 Assume β.2) is bounded. B.

4 Assume Chapter 9. x0 ) ∈ R × R through which the equation 1 + xn−1 xn+1 = 1 + xn + xn−1 is well deﬁned for all n ≥ 0. B.5. Show that the positive equilibrium of Eq(9.3 Determine the set G of all initial points (x−1 .5. Show that the positive equilibrium of Eq(9.1 Determine the set G of all initial points (x−1 . A. and for these initial points investigate the long-term behavior of the solutions {xn }∞ . n=−1 . Conjecture 9. x0 ) ∈ R × R through which the equation xn + xn−1 xn+1 = 1 + xn + xn−1 is well deﬁned for all n ≥ 0. Open Problem 9.5. n=−1 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. ∞) with x−1 + x0 > 0 and that β. and for these initial points investigate the long-term behavior of the solutions {xn }∞ . (2.2 Determine the set G of all initial points (x−1 .5 Assume that x−1 .3) has no prime period-two solutions. Extend and generalize. ∞). A.2) is globally asymptotically stable. C ∈ (0. x0 ∈ [0. 3)-Type Equations α.2) has no prime period-two solutions. γ. Suppose that Eq(9. Extend and generalize.164 Conjecture 9. n=−1 Open Problem 9. C ∈ (0.3) is globally asymptotically stable. Extend and generalize. ∞) and suppose that Eq(9. B.5.5. and for these initial points investigate the long-term behavior of the solutions {xn }∞ . γ.

. . ψ.9.6 Assume that (9.5 Assume that {pn }∞ . Extend and generalize. Investigate the basin of attraction of the two cycle . φ. 1 + yn + rn yn−1 pn + qn yn−1 . .17). . φ. . ψ. . n = 0. lim and r = n→∞ rn .6 Assume that (9. φ. ψ. . p = lim pn . Open Problem 9. of Eq(9. φ. .7 Assume that (9. . ψ. and {rn }∞ are convergent sen=0 n=0 n=0 quences of nonnegative real numbers with ﬁnite limits. Investigate the basin of attraction of the two cycle . and {rn }∞ are period-two sen=0 n=0 quences of nonnegative real numbers.20) holds. {qn }∞ .20) holds. .23). {qn }n=0 . of Eq(9.4 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 . Investigate the global character of all positive solutions of Eqs(9. ψ.5. φ. . Open Problems and Conjectures 165 Open Problem 9. .21)-(9. of Eq(9.5.5.5.22) (9.17) is locally asymptotically stable.23) pn yn + yn−1 . . n→∞ q = n→∞ qn . . . Show that the period-two solution . 1 + yn + rn yn−1 n = 0. n = 0.12) holds. 1. 1. . .5. .11).21) (9. . . . . ψ. . Conjecture 9.5. φ. . . rn + qn yn + yn−1 ∞ Open Problem 9. . 1. Open Problem 9. . (9.

. . . q. 1. 1. (2.26) . .166 Open Problem 9. (9.8 Assume that p. . . .}. . n = 0. 3. 1 + yn + ryn−k pyn + yn−k . r ∈ [0.25) (9. Investigate the global behavior of all positive solutions of each of the following diﬀerence equations: p + qyn (9.5. r + qyn + yn−k n = 0. . 1 + yn + ryn−k yn+1 = yn+1 = p + qyn−k . 3)-Type Equations k ∈ {2. n = 0. . . 1. ∞) and Chapter 9. . .24) yn+1 = .

. the initial conditions are nonnegative.1) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + qyn + ryn−1 . n = 0.4) and r = 167 . 1 + yn αB . A + Cxn−1 Eq(1) contains the following three equations of the (3. . 1.1) (10. and the denominators are always positive. 1. A (10.Chapter 10 (3. 2)-type: xn+1 = xn+1 = and α + βxn + γxn−1 . 2)-type Eq(10. (10. (10. .2) 10. . γ . . 2)-Type Equations 10. . A n = 0.2 The Case C = 0 : xn+1 = A yn B α+βxn +γxn−1 A+Bxn This is the (3. A2 q= β . .3) Bxn + Cxn−1 Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. 1. n = 0.1 Introduction α + βxn + γxn−1 . . . . . xn+1 = n = 0. . A + Bxn α + βxn + γxn−1 . 1. .

6) (10. The main result in this section is that the solutions of Eq(10.7. (c) Assume that r > q + 1.4) converges to a period-two solution.1.5 and 2. in Section 2.4) possesses prime period-two solutions.2.2. Then the equilibrium of Eq(10.1 are given in Sections 10. .5) . The proofs of parts (b) and (c) of Theorem 10.2.7. (3. See Sections 2.4) is globally asymptotically stable. Theorem 10. The proof of part (a) of Theorem 10. for a more general equation. 1.4) 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. 2 (1 + y) 1+y n = 0. By applying the linearized stability Theorem 1.1 (a) Assume that r = q + 1. Then every solution of Eq(10. .2.1 and 10. When r =q+1 Eq(10. Then Eq(10.7) (10.2 below.1 was given. 2)-Type Equations (Eq(10.4) exhibit the following trichotomy character. (b) Assume that r < q + 1. (10.2.4) was investigated in [29].) The linearized equation of Eq(10.168 Chapter 10.4) possesses unbounded solutions.1 we see that y is locally asymptotically stable when r <1+q and is an unstable (saddle point) equilibrium when r > 1 + q. . .

1 Proof of Part (b) of Theorem 10. The Case C = 0 : xn+1 = α+βxn +γxn−1 A+Bxn 169 10. .4) satisfy the following identity yn+1 − q = Note that when q= that is when p + qr − q = 0 the Identity (10.11) q−p r yn−1 − 1 + yn r q−p r for n ≥ 0.2. .5 and therefore the proof is complete when (10.8) which is of the form of the (2. n = 0.4) to the diﬀerence equation zn+1 = (p + rq − q) + rzn−1 . .2.8) holds. .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. .23) which was investigated in Section 6. 1 + yn and so when p≥q we have yn ≥ q Therefore the change of variables yn = q + zn transforms Eq(10. (10.10) reduces to yn+1 − q = Also in this case 0<r =1− r (yn − q) for n ≥ 0. 1 + yn p <1 q (10. Observe that (p − q) + ryn−1 yn+1 = q + .4) is attracted to the equilibrium. (10.10. 2)-type Eq(6. . (10.10) .4) is globally asymptotically stable. (1 + q) + zn n = 1.9) for n ≥ 1.6) holds and we prove that the equilibrium of Eq(10. Now observe that the solutions of Eq(10.2. We know already that when Eq(10. 1. So in the remaining part of the proof we assume that p < q.12) (10.

18) q−p q−p r y2n − > r y2n − 1 + yn+1 r r . It follows from (10. note that now q−p <q r and employ (10. The case where (10. Then we can see from (10.15) yn ≥ q and then use the known result for Eq(10. If (10.13) holds.10) that y2n+2 − q = and so y2n+2 > ry2n + p for n ≥ 0.9). To this end.10) that yN +1+2K > q for all K ≥ 0.15) is not eventually true then either y2n < or y2n−1 < This is because when yN −1 > then it follows from (10.17) holds is similar and will be omitted. (10.12) that Chapter 10. (10. We will assume that (10.4) is equal to q.14) First assume that (10.11) holds.17) q−p r for some N ≥ 0.13) (10.10).16) holds. To complete the proof in this case it is suﬃcient to show that eventually (10. Still to be considered are cases where p + qr − q > 0 and p + qr − q < 0. (3.16) (10.170 and the equilibrium of Eq(10. 2)-Type Equations |yn+1 − q| < r|yn+1 − q| and so n→∞ lim yn = q which completes the proof when (10. q−p r q−p r for n ≥ 0 for n ≥ 0. (10.

14) and completes the proof of part (b) of Theorem 10. Next assume that (10. Note that in this case r < 1 and so q−p < y2n+2 < rn+1 y0 + rn p + . q−p .20) r We will assume that (10.1. y2n−1 > (10. when r < 1.18) that q−p > y2n+2 > rn+1 y0 + rn p + . q−p .4) lies eventually in the interval 0.4. r that either q−p y2n > for n ≥ 0 (10. To this end assume for the sake of contradiction that the solution is not eventually in the interval 0.10) and r the fact that now q−p q< .16) and (10.19) holds. y) = p + qx + ry 1+x is increasing in both arguments. q−p r q−p < r y2n − y2n − 1 + y2n+1 r r . The case where (10.13) holds. Now we claim that every solution of Eq(10.20) holds is similar and will be omitted. it follows from (10. r 1−r This contradicts (10. . + rp + p r which implies that p q−p ≤ . Then one can see from (10.2.14) holds. 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. On the other hand. Then from (10.16) the solution is bounded.19) r or q−p for n ≥ 0. . . . the proof that the r equilibrium is a global attractor of all solutions would be a consequence of Theorem 1.10. + rp + p r and so p q−p ≥ r 1−r which contradicts (10.8 and the fact that in this case the function f (x. Once we establish this result. because by (10.10) we see that y2n+2 − q = and so y2n+2 < ry2n + p for n ≥ 0.2.

3 The Case B = 0 : xn+1 = A yn C α+βxn +γxn−1 A+Cxn−1 This is the (3.4) is a saddle point and so by the stable manifold theorem. Eq(10.21) αC .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. . γ .9) and the conclusion of part (c) follows from Section 6.2.7) holds. It should be mentioned that when (10. A (10.172 Chapter 10.2 Proof of Part (c) of Theorem 10. 1 + yn−1 q= β . That is when y−1 . More speciﬁcally the solutions of Eq(10.4). have the property that the subsequences of even terms converge to ∞ while the subsequences of odd terms converge to q.1 Here we assume that (10. . To this end observe that yn+1 = q + q p + r(yn−1 − r ) 1 + yn for n ≥ 0 and so when (10.7) holds and show that Eq(10. the equilibrium of Eq(10. ∞). A n = 0.4) possesses bounded solutions which in fact converge to the equilibrium of Eq(10. then yn ≥ q Now the change of variables yn = q + zn transforms Eq(10. y0 ∈ [q. 2)-type Eq(10.7) holds the interval [q. ∞) is invariant. (3.2) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + qyn + ryn−1 . 10. 2)-Type Equations 10.5 for appropriate initial conditions y−1 and y0 . in addition to unbounded solutions.2. 1. A2 and r = .4) to Eq(10. .4) possesses unbounded solutions.

Lemma 10. Observe that (p − r) + qyn . n = 0. p 1−q when q < 1 and p ≥ r.1.10. yn+1 = r + 1 + yn−1 and so when p≥r Global Stability When p ≥ r .21) possesses the following invariant intervals: (a) 0.3. Here we make some simple observations about linearized stability. gives a list of invariant intervals for Eq(10. which can be established by direct calculation. and convergence of solutions in certain regions of initial conditions. if whenever. . r−p q when q < 1 and p ≥ r(1 − q). for some integer N ≥ 0.21) when p ≥ r. 1. yN +1 ∈ I then yn ∈ I for n ≥ N. 10.3. . yN . The linearized equation of Eq(10. 1.21) about its positive equilibrium y= is zn+1 − q+r−1+ (q + r − 1)2 + 4p 2 q p − r + qy zn = 0. zn + 1+y (1 + y)2 n = 0. and r.1 Eq(10.1 Invariant Intervals The following result. q. (b) 0. . . Recall that I is an invariant interval.3.2 Here we discuss the behavior of the solutions of Eq(10.3. 10.1 we see that y is locally asymptotically stable for all values of the parameters p. . . The Case B = 0 : xn+1 = α+βxn +γxn−1 A+Cxn−1 173 This equation has not been investigated yet. . invariant intervals. By applying Theorem 1.21).

Then the following statements are true. (1 + r) + zn−1 Chapter 10. The following result is now a consequence of the above observations and the results in Section 6. r−p q .3.21) is bounded. (3.2 (a) Assume that q < 1 and r ≥ p . 2)-type Eq(6. Theorem 10. (ii) q≥1 and either p + qr − r ≤ q or q < p + qr − r ≤ 2(q + 1). which is of the form of the (2.3.2) which was investigated in Section 6.21) with initial conditions in the invariant interval 0.21) when p < r. 2.21) is globally asymptotically stable if one of the following conditions holds: (i) q < 1. Theorem 10. (a) Every positive solution of Eq(10.3. converges to the equilibrium y.3 Convergence of Solutions Here we discuss the behavior of the solutions of Eq(10. 1−q Then every solution of Eq(10. .21) to the diﬀerence equation zn+1 = p + qr − r + qzn . (b) The positive equilibrium of Eq(10. .1 Assume p ≥ r. .3.174 yn ≥ r Therefore the change of variables yn = r + zn transforms Eq(10. 2)-Type Equations for n ≥ 1.3. 10. n = 1.

γ2 (10.K q pq + r2 − pr . qyn + yn−1 q= B . y) increases in x and decreases in y. 1−q and p < r < Then every solution of Eq(10.4 The Case A = 0 : xn+1 = γ yn C α+βxn +γxn−1 Bxn +Cxn−1 This is the (3.K q the function f (x. r−p and that in the interval 0. 2)-type Eq(10.3) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= r + pyn + yn−1 .8. It is easy to check that y ∈ r−p . r−p . The result is now a consequence of Theorem 1.K q converges to the equilibrium y. The result is now a consequence of Theorem 1. r−p q q the function f (x. Proof.22) β . y0 ∈ 0. . 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 . . γ and r = . C n = 0.5. It is easy to check that y ∈ 0. K= q + r − p − q2 (b) In view of Lemma 10. y) increases in both x and y.10.1 we see that when y−1 .4.1 we see that when y−1 .3.3. r−p . then yn ∈ 0. 2 10.4.K q and that in the interval . αC . y0 ∈ for all n ≥ 0. .4. then yn ∈ r−p . (a) In view of Lemma 10. r−p q q for all n ≥ 0.21) with initial conditions in the invariant interval r−p .K q r−p .

The linearized equation of Eq(10. the inequality in (10. (q + 1)(r + (p + 1)y) (q + 1)(r + (p + 1)y) n = 0. . .22) about its positive equilibrium y= is zn+1 − (p − q)y − qr (p − q)y + r zn + zn−1 = 0.9. . 1.1. (10. q−1 .24) 2r q − pq − 1 − 3p >0 (10. invariant intervals.23) is strict and Condition (10. .24) is void. . φ. ψ. . φ. 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 second condition reduces to q > pq + 1 + 3p.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. Concerning prime period-two solution. and the convergence of solutions in certain regions of initial conditions. . (10. as in Section 6.5 that a period-two solution exists if and only if p < 1.23) It is interesting to note that the above condition for the local asymptotic stability of y requires the assumption that r > 0. q > 1 and 4r < (1 − p)(q − pq − 3p − 1). . 2)-Type Equations This equation has not been investigated yet. ψ.25) Furthermore in this case the prime period-two solution . . (3. When r = 0. (1 + p) + (1 + p)2 + 4r(1 + q) 2(1 + q) By applying Theorem 1. . it follows from Section 2. Here we make some simple observations about linearized stability.176 Chapter 10.

(q + 1)b (q + 1)a qr . when q > p + r.4. p−q and (10. q p−q qr p . b) ≤ f (x. b] when p = q a and b are positive numbers such that r + (p + 1)a ≤ (q + 1)ab ≤ r + (p + 1)b.1 Invariant Intervals Here we present some results about invariant intervals for Eq(10.21).21) possesses the following invariant intervals: (a) [a. p−q q (c) 1. . p > q. (b) p qr . We consider the cases where p = q. y) = is decreasing in both arguments. r + px + y qx + y (b) Clearly the function f (x.4.1 q−p Proof. Hence a≤ r + (p + 1)b r + (p + 1)a = f (b.10. y) ≤ f (a. a) = ≤ b. (a) It is easy to see that when p = q the function f (x.26) when p > q and p2 − pq < r. r q−p r . and p < q.1 Eq(10. The Case A = 0 : xn+1 = α+βxn +γxn−1 Bxn +Cxn−1 177 10. q2 p2 − pq > r. First assume that p > q and p2 −pq q2 and it < r.4. Lemma 10. y) is decreasing in both arguments when y < qr is increasing in x and decreasing in y for y ≥ p−q . q2 when p > q and when p < q < p + r.

q p−q Chapter 10. y) is decreasing in both arguments for x < r and it is increasing in y and decreasing in x for x ≥ q−p . assume that p < q < p + r. we obtain qr p . r . y) ≤ f (1.178 Then for x. 2)-Type Equations we obtain ≤ f (x. p−q q p2 −pq q2 and p qr < q p−q < r. y) ≤ f p qr . y ∈ qr p . q p−q p = . p−q q ≤ f (x. Using the decreasing character of f . Using the decreasing character of f in x and the increasing character of f in y. y) ≤ f p p . y ∈ p =f q p qr . 1) = r+p+1 r ≤ . q−p q−p 1< ≤ f (x. we obtain 1=f The inequalities r r . 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. p(q + 1) p−q qr qr . q q = qr qr + p(p + 1) ≤ . Next assume that q > p + r. 1 = 1. p−q . (3. we obtain r (p + r)(q − p) + r = f 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. First. q(q − p) + r q−p ≤ f (x. p2 −pq q2 > r. . y) ≤ f r . 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. q−p (c) It is clear that the function f (x.

1 converges to the equilibrium y. q2 Then every solution of Eq(10.26) converges to the equilibrium y. q p−q conditions in the invariant interval converges to the equilibrium y. p−q q Proof. qr p .7.4. q−p converges to the equilibrium y.22) with initial conditions r in the invariant interval 1.4. and that Condition (10.4. The proof is an immediate consequence of Lemma 10. Proof. p + r < q. The proof is an immediate consequence of Lemma 10.4. q2 and that either p≤1 or p > 1 and (q − 1)[(p − 1)2 (q − 1) − 4q(1 − p − qr)] ≤ 0.1 and Theorem 1. 2 Theorem 10.2 Assume that p > q.10. q−p Proof.22). Then every solution of Eq(10. where 0 < a < b satisfy (10. Then every solution of Eq(10.4.4.22) with initial p qr . Then every solution of Eq(10. 2 .2 Convergence of Solutions Here we obtain some convergence results for Eq(10.5.22) with initial conditions in the invariant interval [a. and r <1 q−p 2 10.4.4.3 Assume that p < q.4. 2 Theorem 10.4. 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.22) with initial conditions in the invariant interval r . (b) Assume that p > q and r > p2 −pq .1 (a) Assume that p = q.25) is not satisﬁed.1 and Theorem 1.6. r < p2 −pq . Theorem 10.4.1 and Theorem 1. b]. Then every solution of Eq(10. The proof is an immediate consequence of Lemma 10. (c) Assume that p < q < p+r.22) with initial conditions in the invariant interval converges to the equilibrium y.

.28) 1 + xn 1 + xn + xn−1 xn+1 = (10.29) 1 + xn−1 1 + xn + xn−1 .27).5.1 Every positive solution of Eq(10.1 We know that every solution {yn }n=−1 of Eq(10. y0 ) ∈ B. x0 ) ∈ R × R through which the equation is well deﬁned for all n ≥ 0. . ψ. .5. Open Problem 10. if and only if r = 1 + q. y0 ) ∈ (0. .4) are bounded.180 Chapter 10. .5 Open Problems and Conjectures ∞ Open Problem 10. . Open Problem 10.5. (10. Conjecture 10. φ. ψ. (a) Determine the set of initial conditions y−1 and y0 for which φ = ψ. φ. . 2)-Type Equations 10.27) (b) Assume (10. (3. 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 . Determine the set of initial conditions y−1 and y0 for which {yn }∞ n=−1 converges to (10.27) is a prime period-two solution of Eq(10. determine the set G of all initial conditions (x−1 . ∞) such that the ∞ solutions {yn }n=−1 of Eq(10. ∞ (b) Let (y−1 .5.30) xn+1 = xn + xn−1 Extend and generalize.3 For each of the equations listed below.21) converges to the positive equilibrium of the equation.4). ∞) × (0. (a) Find the set B of all initial conditions (y−1 . Investigate the asymptotic behavior of {yn }n=−1 .2 Assume r > q + 1. (10.

22) possesses a period-two solution. {qn }∞ . 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 . and {rn }n=0 are convergent sen=0 n=0 quences of nonnegative real numbers with ﬁnite limits.4 Assume that {pn }∞ .33).22) has no prime period-two solutions the equilibrium y is globally asymptotically stable. Extend and generalize.5. . .22) (a) Local stability of the positive equilibrium implies global stability. n = 0. Then the following statements are true for Eq(10. is locally asymptotically stable.5.5 Assume that {pn }∞ . 181 (b) When Eq(10.31)-(10.22).10. Investigate the global character of all positive solutions of Eqs(10. . 1 + yn−1 rn + pn yn + yn−1 .2 Assume that p. . p = lim pn . 1.33) Open Problem 10. lim lim and r = n→∞ rn . qn yn + yn−1 n = 0. and {rn }∞ are period-two sen=0 n=0 n=0 quences of nonnegative real numbers. when it exists.5. but not globally. . 1. . (d) The period-two solution of Eq(10. (10. .5. q. {qn }∞ . (c) When Eq(10.32) (10. ∞ Open Problem 10. r ∈ (0. . . the equilibrium y is a saddle point.31) yn+1 = yn+1 = n = 0. 1 + yn pn + qn yn + rn yn−1 . ∞). . n→∞ q = n→∞ qn . . 1. Open Problems and Conjectures Conjecture 10. (10.

r ∈ [0.182 Open Problem 10. 1 + yn−k p + qyn + ryn−1 . 1. 1 + yn p + qyn + ryn−k . . n = 0.34) (10. . ∞) and k ∈ {2. . .}. . . .6 Assume that Chapter 10. . 1. (3. 2)-Type Equations p. . . (10. q. n = 0. . .36) . 3. qyn + yn−k n = 0. .5. 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 .35) (10. 1.

2) . (11. . invariant intervals.1) Please recall our classiﬁcation convention in which all coeﬃcients of this equation are now assumed to be positive and the initial conditions nonnegative. x= β+γ−A+ (A − β − γ)2 + 4α(B + C) 2(B + C) .1 Linearized Stability Analysis (B + C)x2 − (β + γ − A)x − α = 0. . A + Bxn + Cxn−1 n = 0. . . 1. This equation has not been investigated yet. Here we make some simple observations about linearized stability. 3)-type equation xn+1 = α + βxn + γxn−1 . . .1) has a unique equilibrium which is the positive solution of the quadratic equation That is. . (11. and the convergence of solutions in certain regions of initial conditions. 3)-Type Equation α+βx xn+1 = A+Bxn+γxn−1 n+Cx n−1 In this chapter we discuss the (3. 1. 11. The linearized equation about x is (see Section 2. Eq(11.3) zn+1 − (βA − Bα) + (βC − Bγ)x zn + α(B + C) + (B + C)(A + β + γ)x (γA − Cα) − (βC − γB)x = 0. .Chapter 11 The (3. − z 2 + α(B + C) + (B + C)(A + β + γ)x n−1 A A2 183 n = 0.

then clearly fx = M = βC − γB.13). These intervals are derived by analyzing the intervals where the function f (x. γ > β+A.1) by testing whether the hypotheses of one of the four Theorems 1. 2 (A + Bx + Cy) (A + Bx + Cy)2 Recall that an interval I is an invariant interval for Eq(11. is discussed in Section 2. (11. Eq(11. φ. φ. .14). 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. and α < (γ − β − A)[B(γ − β − A) − C(γ + 3β − A)] . Conditions (2.1) if whenever. if. (2. . 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. when (11. .3) 4C 2 Furthermore. If we set L = βA − αB. .2 Invariant Intervals Here we obtain several invariant intervals for Eq(11. yN +1 ∈ I then for some integer N ≥ 0. . . . ψ.6.5. ψ. and N = γA − αC α + βx + γy A + Bx + Cy L + My N − My and fy = .1). Once we have an invariant interval for Eq(11. . for n ≥ N. y) = is increasing or decreasing in x and y. . .8 are satisﬁed. . (11.1) then we may be able to obtain a convergence result for the solutions of Eq(11. and (2.4.4.32) are satisﬁed. ψ. yN . The (3. see Section 2. .4) The local asymptotic stability of the two cycle 11.1) has a prime period-two solution .184 Chapter 11. C C(B − C) .15) are satisﬁed. ψ. φ. that is.5-1.31) and (2.3). φ. yn ∈ I .

C B A 185 α β Then A .1) and in this interval the function f (x. (iv) Assume that β α γ = < .1) and in this interval the function f (x. Invariant Intervals Lemma 11.1) and in this interval the function f (x. βC−γB is an invariant interval for Eq(11.1) and in this interval the function f (x. B+C βC − γB α β+γ Then A . (iii) Assume that β α γ > > C B A and β+γ αB − βA ≤ . C M α γ Then A .2. A is an invariant interval for Eq(11.1 (i) Assume that β α γ = > . y) is increasing in x and decreasing in y. . (v) Assume that α β γ γA − αC > . ≥ > βC − γB A B C γA−αC Then 0. (ii) Assume that β γ α > > B C A and γ N ≤ . y) is decreasing in both variables. C is an invariant interval for Eq(11. y) is increasing in both variables. A is an invariant interval for Eq(11. y) is increasing in both variables. y) is increasing in both variables.1) and in this interval the function f (x. y) is decreasing in both variables.2.1) and in this interval the function f (x. < < < B C A βC − γB α Then 0. A B C α Then 0. (vii) Assume that α β γ = > . A is an invariant interval for Eq(11.11. C B A β α Then B . (vi) Assume that β γ α γA − αC .1) and in this interval the function f (x. B is an invariant interval for Eq(11. y) is decreasing in both variables. B+C is an invariant interval for Eq(11.

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

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

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

n→∞ . K1 ] β−A+ Signs of derivatives N M [0. .3. M ] if AM +(B+C)N < αM +βK+γN N N [ M .1). − M ] if AM −L(B+C) < − M αM −βL+γM L L [− M .1 Let {xn } be a solution of Eq(11. Theorem 11. A ] √ L L fx > 0 if y < − M . fx < 0 if y < − M fy < 0 fx < 0 fy < 0 fx < 0 N N fy > 0 if x > M .4. K2 ] if BM K+AM −γL > − M −CL β+γ−A+ β [0. fx < 0 if y > − M fy > 0 fx > 0 fy < 0 fx < 0 fy > 0 L L fx > 0 if y > − M . Then in each of the following cases lim xn = x. ((γ − A)M − BN )2 + 4BM (αM + βN ) 11.4.11. 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 . [0.5-1. fy < 0 if x < M αM +(β+γ)N αM +(β+γ)N N [ AM +(B+C)N . Convergence Results 189 Case 1 2 3 4 5 6 7 8 9 Invariant intervals αM +(β+γ)N N [0. C ] αM −L(β+γ) αM −L(β+γ) L L [ AM −L(B+C) .2. . − M ] if AM −(B+C)L < − M βM K+αM L L [− M . K3 ] if AM −BL+CMK > − M K α [0. B ] γ [0.8. 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 .3. 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 [0.3 Convergence Results The following result is a consequence of Lemma 11. .1 and Theorems 1.

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

m= α + βm + γM .3 to Eq(11.1 (ix) and (x) respectively. m).3. The proofs of statements (iv-vi) follow by applying Lemma 11.11. and by observing that the only solution of the system of equations m = f (m. . The proofs of statements (ix) and (x) follow by applying Lemma 11.2.1 Global Stability Theorem 1. By applying Theorem 1. x0 ∈ 0.1) is globally asymptotically stable. .3. Theorem 1.1) does not possess a prime period-two solution.2. 2 11.1 (vii) and (viii) respectively. The proofs of statements (vii) and (viii) follow by applying Lemma 11.2 Assume that γ β α < min . M ).3) is not satisﬁed. that is.1 (iv-vi) and Theorem 1. α γ β < ≤ B A C and x−1 . C and x−1 .2.2. A + Bm + CM M= α + βM + γm A + BM + Cm is m = M .4.2 and linearized stability imply the following global stability result: Theorem 11. C 191 Proof.3.7. M = f (M. x0 ∈ 0.4. The proofs of statements (i-iii) follow by applying Lemma 11.4.1 (i-iii) and Theorem 1.1) we obtain the following global asymptotic result. Convergence Results (ix) Condition (11.4.6. β γ α = < A B C (x) Condition (11.5. γ .8. Theorem 1.4. γ . and by observing that Eq(11.3) is not satisﬁed. C B A Then the equilibrium x of Eq(11.4.

αB > βA.1) converges to the positive equilibrium. γ. The following global stability result is a consequence of linearized stability and the Trichotomy Theorem 1.3. Conjecture 11. B.4. C ∈ (0.4. . 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Theorem 11. B.4. ∞) and that Eq(11. Conjecture 11. C ∈ (0.4. ∞) and that Eq(11. Theorem 11. A. Then the equilibrium x of Eq(11. and γA > αC. γ.4 Open Problems and Conjectures α. β.4. ∞). γB > βC. B. Then the equilibrium x of Eq(11.1) has a positive prime period-two solution.192 Chapter 11.1) has no positive prime period-two solution. C ∈ (0.1) is bounded. β. γA ≥ αC.1) is globally asymptotically stable. and 3βC ≥ γB ≥ βC.2 Assume α. The (3.4 (a) Assume that αC ≥ γA. and 3γB ≥ βC ≥ γB. Then the equilibrium x of Eq(11. Conjecture 11.1) is a saddle point and the period-two solution is locally asymptotically stable. A. γ.3 Assume α. Show that every positive solution of Eq(11.3. 11. Show that the positive equilibrium of Eq(11. β. A. (b) Assume that αB ≥ βA.3 Assume that B ≤ C.1 Assume Show that every positive solution of Eq(11. βA ≥ αB.1) is globally asymptotically stable.1) is globally asymptotically stable.

2 Assume that the parameters in Eq(11.4 Assume that (11.5 Assume that (11.3 Assume that a.4.4. A.4.4. Open Problem 11. .4 Assume α.1) cannot have a prime period-k solution for any k ≥ 3. k ≥ 2. {An }n=0 . . B.7) Open Problem 11. ∞) for i = 0. k.7) converges to a period-two solution. . . . Investigate the asymptotic behavior and the periodic nature of all positive solutions of the diﬀerence equation αn + βn xn + γn xn−1 . . A.6) Obtain necessary and suﬃcient conditions so that every positive solution of the equation xn+1 = has a ﬁnite limit. 193 ∞ ∞ Open Problem 11.11. {Cn }∞ n=0 n=0 n=0 n=0 are convergent sequences of nonnegative real numbers with ﬁnite limits. {γn }∞ . . ai . A + A0 xn + .6) holds. Open Problems and Conjectures Conjecture 11.6) holds. . β. . . .5). + Ak xn−k (11.1 Assume {αn }n=0 . . n = 0. C ∈ (0. γ. . investigate the cases where 2 ≤ k ≤ 8. ∞).5) An + Bn xn + Cn xn−1 Open Problem 11. n = 0. a + a0 xn + . xn+1 = (11. + ak xn−k . .4. Show that Eq(11. Ai ∈ [0. Open Problem 11. (11.7) is periodic with period-k. Investigate the asymptotic behavior and the periodic nature of all positive solutions of Eq(11. 1. In particular.5) are period-two sequences of positive real numbers. . Obtain necessary and suﬃcient conditions so that every positive solution of Eq(11. {Bn }∞ . Obtain necessary and suﬃcient conditions so that every positive solution of Eq(11. 1.4. . . 1.4. {βn }∞ .

and the asymptotic behavior of all positive solutions of the diﬀerence equation xn+1 = (See [8] and [61].4.194 Chapter 11. (11.) Conjecture 11. β. the periodic nature. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Open Problem 11. . The (3.4. . 1. . ∞) with γ + δ > 0. ∞) for i. Show that every positive solution of the diﬀerence equation xn+1 = xn + . xn xn−1 + xn−2 + . j ∈ {1. Extend to equations with real parameters. 2. . . + xn−k−2 converges to 1.4.) αxn + βxn−1 xn−2 . γ.8) converges to a period-two solution. δ ∈ [0. . . . Show that every positive solution of the system xn+1 = yn+1 = zn+1 = a11 xn a21 xn a31 xn + + + a12 yn a22 yn a32 yn + + + a13 zn a23 zn a33 zn . and [60]. . γxn xn−1 + δxn−2 Conjecture 11.6 Let α. Investigate the boundedness character. q ∈ [0.5 Let k ≥ 2 be a positive integer.4. [47]. 3}. .) Conjecture 11. . 1. .6 Let aij ∈ (0. (See [8]. . 1. Extend and generalize. ∞). . (See [32] for the two dimensional case. + xn−k + xn−k−1 xn−k−2 . n = 0. n = 0. n = 0. .7 Assume that p.

.4. . q + xn−2 n = 0. 195 converges to a period-two solution if and only if p = 1 + q. xn−1 + pxn−2 n = 0. . ∞). (c) Show that when p>1+q the equation possesses positive unbounded solutions. .9 Assume that p. q ∈ [0. 1.11.4. (b) Show that when p<1+q every positive solution has a ﬁnite limit.) Conjecture 11. Open Problems and Conjectures (a) Show that every positive solution of the equation xn+1 = pxn−1 + xn−2 . Conjecture 11.4. 1.8 Assume that p ∈ (0. . (See [62]. (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 . ∞). converges to a period-two solution if p ≥ 1.

1. (See [62]. converges to a period-four solution.10 Show that every positive solution of the equation xn+1 = xn + xn−2 . (See [62]. converges to a period-two solution if and only if q = 1. . has the following trichotomy character: (i) When p > 1 every positive solution converges to the positive equilibrium. 1. .4.4.196 Chapter 11. . . 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 .) Conjecture 11. (iii) When p < 1 there exist positive unbounded solutions.) Conjecture 11. . xn−2 + q n = 0. 1. (See [62]. xn n = 0. (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. (ii) When p = 1 every positive solution converges to a period-ﬁve solution. . .) . The (3. .11 Show that the diﬀerence equation xn+1 = p + xn−2 . xn−1 n = 0. .

(See [62].12 Show that every positive solution of the equation xn+1 = 1 + xn .9) with nonnegative parameters and nonnegative initial conditions such that B + D > 0 and A + Bxn + Dxn−2 > 0 for n ≥ 0. . .4. A + Bxn + Dxn−2 (11.9) has prime period-two solutions if and only if γ = β + δ + A.. ψ. . where φ..11. ∞). φ φ ψ ψ Open Problem 11.4. Then one can show that Eq(11. what other conditions are needed so that every positive solution of Eq(11. .) . . xn−1 + xn−2 n = 0. . . .9) converges to the positive equilibrium? (c) Assume γ > β + δ + A. What other conditions are needed so that Eq(11. ψ ∈ (0. .9) possesses unbounded solutions? (See [6]. Open Problems and Conjectures Conjecture 11.9) converges to a period-two solution? (b) Assume γ < β + δ + A..7 Consider the diﬀerence equation xn+1 = α + βxn + γxn−1 + δxn−2 . 197 converges to a period-six solution of the form . 1. . n = 0.) ψ 1 1 φ . .10) (a) In addition to (11.4.10). φ. . What other conditions are needed so that every positive solution of Eq(11. (11. 1.

. The (3. converges to a period-three solution. 1 + xn + xn−2 n = 0. . r ∈ (0. . 1. .4.4.198 Chapter 11. Conjecture 11. converges to a period-two solution. (See [6]. . .) Conjecture 11. 1. . (a) Show that when q=r every positive solution converges to a period-four solution. xn 1 + xn−1 + xn−2 . 1. .14 Show that every positive solution of the equation xn+1 = 1 + xn−1 . 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Conjecture 11.15 Show that every positive solution of the equation xn+1 = xn−2 . 1. xn−1 n = 0.13 Show that every positive solution of each of the two equations xn+1 = and xn+1 = 1 + xn−1 + xn−2 .4. 1. ∞) and q.) Conjecture 11. n = 0. converges to a period-two solution. xn−2 n = 0. . ∞). (See [6]. . xn−1 + xn−2 + xn−3 n = 0.4. . . p + qxn + rxn−2 .16 Consider the diﬀerence equation xn+1 = where p ∈ [0. . . . . (b) Show that when q>r every positive solution converges to the positive equilibrium.

(See [62].4. 1. A + B0 xn + . . . .) 199 Open Problem 11. converges to a period-(k + 1) solution of this equation.) . 1. .11. . . xn−2 n = 0. . . . . . B0 . (See [6]. . Bk so that every positive solution of the diﬀerence equation xn+1 = xn−k . . 1 + xn−3 . .4. . . . xn + xn−2 + xn−3 n = 0. 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 . . B1 . + Bk xn−k n = 0. xn−2 + xn−3 . 1. 1.8 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 . .4. Open Problem 11. . ∞) with i=0 Bi > 0. 1. . Bk ∈ [0. .9 Assume that k is a positive integer and that k A. xn−1 . . B1 . 1 + xn + xn−3 n = 0. . xn n = 0.

1 + xn + . x0 for which the ∞ corresponding solution {xn }n=−k of Eq(11. converges to a periodic solution of period 2k. . φ. xn−k+1 n = 0.17 Let k be a positive integer.12) with a given φ ≥ 0 ? (See [6]. . (c) Determine the set of all nonnegative initial conditions x−k . . Then we can easily see that every nonnegative solution of the equation xn+1 = xn−k . 1.12) which corresponds to the nonnegative solution {xn }∞ n=−k of Eq(11. . 0. . .. (a) Does Eq(11.12) with φ ≥ 0. the value of φ in (11.11) possess a positive solution which converges to zero? (b) Determine. . . 1. in terms of the nonnegative initial conditions x−k . .. .) Conjecture 11. x0 . Show that every positive solution of xn+1 = 1 + xn + xn−k . + xn−k+1 n = 0. 0.200 Chapter 11. .11) converges to a period-(k + 1) solution of the form . The (3. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Open Problem 11. (11.11) converges to a period-(k+1) solution of the form (11. . . 0. . .11). . . .4. . . . .. .4. k−terms (11.10 Assume that k is a positive integer. .

xn−1 . . very often a good strategy for obtaining global attractivity results for Eq(A. Let I be some interval of real numbers and let f :I ×I ×I →I be a continuously diﬀerentiable function. z) is monotonic in its arguments.1) has a unique equilibrium x ∈ [a. (b) The equation f (x. x−1 . .0. b]3 → [a. b] and every solution of Eq(A. has a unique solution in the interval [a. or non increasing in two arguments and non decreasing in the third. y.4.1) ∞ has a unique solution {xn }n=−2 . or non decreasing in all three.1 Let [a. Then for every set of initial conditions x0 . b].5-1. z) is non decreasing in all three arguments. n = 0. z) is non-decreasing in each of its arguments.8.1) converges to x. b] is a continuous function satisfying the following properties: (a) f (x. .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. In this regard there are eight possible scenarios depending on whether f (x. 201 . 1. x. As we have seen in previous chapters. Theorem A. x) = x. or non increasing in one and non decreasing in the other two. the diﬀerence equation xn+1 = f (xn . We state the results for third order equations from which higher order extensions will become clear.1) is to work in the regions where the function f (x. xn−2 ).4. Then Eq(A. b] be an interval of real numbers and assume that f : [a. y. x−2 ∈ I. y. (A.

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

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

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

2 Theorem A. 2. set Mi = f (mi−1 . . M ) ∈ [a. b]. .1) has a unique equilibrium x ∈ [a. . M ) and m = f (M. b] is a continuous function satisfying the following properties: (a) f (x. b] × [a. Then Eq(A. Therefore in view of (b) m=M from which the results follows. m. m).2) holds and by the continuity of f . i→∞ and M = lim Mi . M. Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . M = f (m. .1) converges to x. Set m0 = a and M0 = b and for each i = 1. .0.5 Let [a. Then (A. M. M ) then m = M. b] is a solution of the system M = f (M. ≤ mi ≤ . m) and m = f (m. b] and is non-increasing in y and z for each x ∈ [a. b] be an interval of real numbers and assume that f : [a. M1 ≤ M0 = b and mi ≤ xk ≤ Mi Set m = lim mi i→∞ 205 for k ≥ 3i + 1. . mi−1 ). . mi−1 . .Appendix Proof. b]3 → [a. y. Mi ≤ . Mi−1 . . b] and every solution of Eq(A. (b) If (m. m. z) is non-decreasing in x for each y and z ∈ [a. . Mi−1 ) and mi = f (Mi−1 .

b] and is non-increasing in x and z for each y ∈ [a. Therefore in view of (b) m=M from which the results follows. b] is a solution of the system M = f (m. Mi ≤ . M. m) and m = f (M. y. Mi−1 ). . Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . i→∞ and M = lim Mi . Mi−1 . m) and m = f (m. M ). M = f (M. M1 ≤ M0 = b and mi ≤ xk ≤ Mi Set m = lim mi i→∞ Appendix for k ≥ 3i + 1.1) has a unique equilibrium x ∈ [a.0. mi−1 ) and mi = f (mi−1 . mi−1 . 2. . b] × [a. b]3 → [a. Set m0 = a and M0 = b and for each i = 1. Then Eq(A. . . b] and every solution of Eq(A.1) converges to x. .6 Let [a. . ≤ mi ≤ . M. set Mi = f (Mi−1 . . M ) ∈ [a. b].2) holds and by the continuity of f . b] be an interval of real numbers and assume that f : [a. 2 Theorem A.206 Proof. M ) then m = M. . m. . (b) If (m. m. . z) is non-decreasing in y for each x and z ∈ [a. Then (A. b] is a continuous function satisfying the following properties: (a) f (x.

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

b]3 → [a. y.2) holds and by the continuity of f . Then (A. Set m0 = a and M0 = b and for each i = 1. . M1 ≤ M0 = b and mi ≤ xk ≤ Mi Set m = lim mi i→∞ Appendix for k ≥ 3i + 1. 2 Theorem A. Mi−1 . . z) is non-increasing in all three variables x. y. M ) and m = f (M. . m). M. mi−1 .1) converges to x.8 Let [a. b] × [a. . . z in [a. Therefore in view of (b) m=M from which the results follows. . mi−1 ). .1) has a unique equilibrium x ∈ [a. b]. . M = f (m. (b) If (m. b] is a continuous function satisfying the following properties: (a) f (x. M ) then m = M. m. 2.0. M. b] is a solution of the system M = f (m. . ≤ mi ≤ . m) and m = f (M. i→∞ and M = lim Mi . b] and every solution of Eq(A.208 Proof. b] be an interval of real numbers and assume that f : [a. Mi−1 ) and mi = f (Mi−1 . Then Eq(A. M ) ∈ [a. Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . Mi ≤ . . m. set Mi = f (mi−1 .

M. (A.0. . p xn+1 = f (xn . . . . . 2. mi−1 .Appendix Proof. . f : Ip → I n = 0. M ). . xn−1 . .0. u2 . . b] be an interval of real numbers and assume that f : [a. ≤ mi ≤ . Set m0 = a and M0 = b and for each i = 1. Mi−1 ).9 Let [a. . Mi ≤ . set Mi = f (mi−1 . m) and m = f (M. M = f (m.1-A. where is a continuous function. Then (A. In this case.8. Mi−1 . . m. . Therefore in view of (b) m=M from which the results follows.3) 2 The above theorems can be easily extended to the general diﬀerence equation of order . Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . up ) is non decreasing in the second and the third variables u2 and u3 and is non increasing in all other variables. . 1.2) holds and by the continuity of f . Let us formulate one of these results. .0. This gives 2p results of the type found in Theorems A. i→∞ and M = lim Mi . mi−1 ) and mi = f (Mi−1 . . Theorem A. . . there are 2p possibilities for the function f to be non decreasing in some arguments and non increasing in the remaining arguments. . b]p → [a. xn−p+1 ). 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. .

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

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

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

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

173 (1993). 1(4) (1995). J. Teixeira. path stability. K. Anal. J. T. Kuang and J. F. and E. S. G. and N. J. Lett. Math. 641-646. R. Kruse and T. Appl. Nesemann. Dordrecht. S. Ladas. Hot-Line 2(1998). 6(5) (2000). Kulenovi´. On the recursive c p+yn−1 sequence yn+1 = qyn +yn−1 . [49] M. and I. n +Cxn−1 [51] M. On the recursive sequence xn+1 = c αxn +βxn−1 . Ladas. Ladas. G. Global asymptotic stability for discrete dynamical systems. [43] V. Kulenovi´. Equations Appl. γxn +Cxn−1 . Math. [54] M.5. 571-586. Martins. Kluwer Academic Publishers. (to appear). Dynamics of Continuous. Invariants and related Liapunov functions for diﬀerence equac tions. Math. Appl. [46] U. 245-254. R. [45] W. and I. Diﬀer. Ladas. Comput. Academic Press. Global Behavior of Nonlinear Diﬀerence Equations of Higher Order with Applications. On a rational diﬀerence equation. Diﬀer. On the stability of Lyness’ equation. and N. 1993. Open problems and conjectures: On period two c α+βx solutions of xn+1 = A+Bxn +γxn−1 . 1+xn [53] M. Cushing. Anal. 31-37.214 BIBLIOGRAPHY [41] W. G. R. M. no. 13 (7)(2000). Math. Kulenovi´ and G. Kosmala. Global stability in a nonlinear diﬀerence-delay equation model of ﬂour beetle population growth. G. Appl. G. Math. Appl. 151-158. [48] Y. L. W. R. S. and relative stability for positive nonlinear diﬀerence equations of higher order. Rodrigues. Anal. 41(2001). Thomas. G. Math. G. L. Kulenovi´. J. 323-346. L. 563-576. Equations Appl. Stability trichotomy. Prokup. Holland. Equations Appl. New York. Appl. Ladas. R. S. On the rational recursive sequences. S. G. Prokup. 1-16. R.. Ladas. Ladas. 2(1)(1996). M. J. R. 1-8. Sizer. [42] V. Discrete and Impulsive Systems 1(1995). and W. On the recursive sequence xn+1 = c αxn +βxn−1 . Kelley and A. [44] V. S. Tzanetopoulos. 6(5) (2000). c Comput. R. 251(2000). Kulenovi´. 235 (1999). Math. Kocic. and C. Equations Appl. Sci. W. [50] M. S. Kocic. Peterson. Ladas. G. Krause. C. A. Res. [47] N. R. J. Diﬀer. 671-678. Kulenovi´. Appl. 127-157. S. J. Ladas. Diﬀer. Diﬀerence Equations. Kulenovi´. 1991. Kocic and G. Rodrigues. L. [52] M. On the dynamics c α+βxn of xn+1 = A+Bxn +Cxn−1 facts and conjectures.

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

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

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

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

Sign up to vote on this title

UsefulNot useful- Skirms Stability
- 515nts0
- Discrete Competition Models
- STRESS CONCENTRATIONS IN FILAMENTARY STRUCTURES
- Robotic Application
- 01318710
- Che386HW1
- A Note on Stability of Motion of a Projectile
- Shone - An Introduction to Economic Dynamic
- Ch-6 RH Stability Webpage
- 291-IJC03
- Sum Constructions
- Ra Material
- Network-Based H Inf Output Tracking Control
- A Model to Predict Stall Inception of Transonic Axial Flow
- 1-s2.0-0301679X78900993-main
- Periodic Functions, The Unit Step Function and the Second Shifting Theorem
- Y.wang, P.L.yung - A Subelliptic Bourgain-Brezis Inequality
- CE
- GATE Exam Syllabus Mechanical Engineering - ME PDF Download
- Strum Ecellenca2
- A Theoretical Foundation for the Gravity Equation
- Infinity
- IRJET- Onset Of Convection Of Maxwellian Fluid In Porous Medium With Variable Gravity Using Positive Operator Method.
- shenshengping-7
- s 2010194512005399
- me471_lec02
- Articulo Networked control systems with communication constraints.pdf
- Sess 6 Tuenbaeva Nazarov
- Course 1 Introduction
- Dynamics of Second Order Rational Diff[1]

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.