## Are you sure?

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.

K84 2001 515′.R.W. Reasonable efforts have been made to publish reliable data and information. Kulenovi´ . or for resale. A wide variety of references are listed.com © 2002 by Chapman & Hall/CRC No claim to original U. including photocopying. Includes bibliographical references and index.Library of Congress Cataloging-in-Publication Data Kulenovi´. electronic or mechanical. QA431 . G.. for creating new works.S.R. Florida 33431. (Mustafa R. I.) c Dynamics of second order rational difference equations : with open problems and conjectures / M.S. and recording. Corporate Blvd. or by any information storage or retrieval system. Difference equations—Numerical solutions. microﬁlming. G. Trademark Notice: Product or corporate names may be trademarks or registered trademarks.E. 2000 N. Ladas. ISBN 1-58488-275-1 1. and sources are indicated.crcpress. Direct all inquiries to CRC Press LLC. 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 .625—dc21 20010370 This book contains information obtained from authentic and highly regarded sources. without intent to infringe. c p. Ladas. for promotion. 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. The consent of CRC Press LLC does not extend to copying for general distribution. and are used only for identiﬁcation and explanation. cm. Boca Raton. II.S.S. without prior permission in writing from the publisher. Speciﬁc permission must be obtained in writing from CRC Press LLC for such copying. Title. Visit the CRC Press Web site at www. M. Reprinted material is quoted with permission.

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

.6. . .Riccati Equation . .1 Semicycle Analysis .4 Global Behavior of Solutions . . . . . 2)-Type Equations 77 6. . . . The Case γ = A = 0 : xn+1 = Bxα+βxn n +Cxn−1 6. 6. . . . . . .7. . .3 The Case β = A = C = 0 : xn+1 = α+γxn . .7 The Case α = β = A = 0 : xn+1 = 4. . . . . 6. . . . . . . . . . . . .3 The Case q > 1 . . . . . . . . . . . . .2 Semicycle Analysis . . . . . . . . . . . .5 The Case β = C = 0 : xn+1 = α+γxn−1 . .6 The Case α = β = C = 0 : xn+1 = 4.4 . . . . . . 6. .7 The Case α = C = 0 : xn+1 = βxA+Bxn−1 . . 4. 1)-Type Equations 5. . . . . . . .5. . . .1 The Case p < 1 . . . . . . . . . . . . . .3 The Case β = γ = A = 0 : xn+1 = Bxn +Cxn−1 . .7. . . .1 Introduction . .2 Invariant Intervals . . .3 The Case γ = B = 0 : xn+1 = A+Cxn−1 . . . . . 5. . . . . . .4 Global Attractivity When q < 1 . . . .4 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 4. 4. . . . . . . A+Bxn 6. 6. . . . . . . .1 Existence and Local Stability of Period-Two Cycles . . . .7. . . . . . . . . . . . . 78 n α+βxn 6. . . . . . . . . 53 α 4. .4. . . . . . . . . . . . . . . . . . . . . 6.7. . .7. . . . . . . . . . .5. . . .6. . . .4 The Case α = γ = B = 0 : xn+1 = 4. . . . . . . . . . .1 Introduction . . . . . . . . . . . . . . . . . . . . 77 α+βx 6. . . . . . . . . . . . . . 6. . . . . . . . . . 57 58 59 60 61 62 62 64 69 69 70 72 72 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Semicycle Analysis . . . .6 The Case β = A = 0 : xn+1 = Bxn +Cxn−1 . . . . . . . . . . . .Lyness’ Equation Cxn−1 n−1 5. . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Introduction . . Bx 5. . . . . . . . . . . . . . .3 Global Stability and Boundedness . . . .5 The Case α = γ = A = 0 : xn+1 = 4. . . . . . . . . . . . . . . . .7. .8 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6. . . . . . . . .2 The Case γ = A = B = 0 : xn+1 = α+βxn . . . . . .2 Semicycle Analysis of Solutions When q ≤ 1 . . . . . . . . . . . . . . . 6. . . . . . . . . . . A+Bxn γxn−1 Bxn +Cxn−1 Pielou’s Equation . .1 Invariant Intervals . . . 2)-Type Equations 53 4. . . . . . . . . .3 Semicycle Analysis and Global Attractivity When q = 1 . . . . 79 6. . n 6. . . . . 4. . . . . . . . . . . . 6. . . . . .2 The Case q < 1 . . . . . . . . . . . . . . . . . . . . . .5. . . .6. . . . . . . 6 (2. . . . 54 α 4. 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4. . . . . . . . . . . . 6. . . . . . . . . . . . . α+γxn−1 6. . . . . . .4. βxn A+Cxn−1 βxn Bxn +Cxn−1 γxn−1 . . . . . .vi Contents 4 (1. . . . . . . . . . . . . . . . .2 The Case β = γ = C = 0 : xn+1 = A+Bxn . . . .3 The Case p > 1 . . . . . . . . . . 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 The Case p = 1 . . . . . . . . . .1 Invariant Intervals . . . 6. . . .2 The Case γ = C = 0 : xn+1 = A+Bxn . . . . . . . . . . . . 82 83 84 87 89 90 90 91 92 93 96 97 100 101 102 103 105 107 . . . 5 (2. . . . . . . . . . . . . . . . . n +γx 6. . . .7. . . . . . . . . . . . .

. . . . .2 The Case γ = 0 : xn+1 = A+Bxn +Cxn−1 .2 Invariant Intervals . . . . . . . . .1 Introduction . .2 Convergence of Solutions . . . .4 Global Stability Analysis When p > q . . . . . . . . . . . . 141 141 141 142 143 144 147 149 150 154 156 157 157 158 160 162 163 . . .9. . . . . . . . . .2. . . . . vii 107 109 110 112 113 114 117 117 123 123 7 (1. . . . . . .5 Semicycle Analysis When pr = q + qr . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9. . . . . . . . . 9. . . .5 Open Problems and Conjectures .1 Existence and Local Stability of Period-Two Cycles 9. 131 +Cx 7. . . . . . . . . . . . . . . . 9. . . . . . . . . . . . . . . . . .4 The Case α = 0 : xn+1 = A+Bx+γxn−1 . . . 3)-Type Equations 9. . . . . . . .3 The Case α = γ = 0 : xn+1 = A+Bxβxn n−1 . . . .8. . . . 9. .4 Global Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9. . . 131 7. 6.1 Existence of a Two Cycle . . . . . . . n +Cxn−1 9. .2 Open Problems and Conjectures . . .3 The Case β = 0 : xn+1 = A+Bxn +Cxn−1 . . . . . . . 2 9.9. . . . . 6. . . 9. . . . . . . . .1 Invariant Intervals and Semicycle Analysis . . . . . . βxn 9. . . . . . . . . . . . . . . . .8. . . . . . . . . . . . . . .1 Boundedness of Solutions . . . . . .10 Open Problems and Conjectures . . . . . . . . . . . . 9. . .2 Global Stability of the Positive Equilibrium . . . . . . . . . . . . . . . . . . . 6. . . . . . . . . . . 137 8. . . . .2. . . . . . . . . . . . . . . . . . . . . . . . .3. . . . . . . . .4. . .4 Global Asymptotic Stability . . . . .3 Convergence of Solutions . . . . . . . . . . . . . . .2 Semicycle Analysis . 138 9 (2. . . . . . . . . . . . . . . . . . 6. . . .2. . . . . . . . . . . . . . . . 132 n +Cx γx 7. . . . . . . . . . . . . . . . . . .4 The Case α = β = 0 : xn+1 = A+Bxnn−1 n−1 . . . . . .3. . . . . . . .3 Global Stability Analysis When p < q . . .1 Introduction . . . . 6. . . . . . . . . . . . . . . . . . . . . . .1 Introduction . . . 9. . .3. . . . .9 The Case α = A = 0 : xn+1 = Bxn +Cxn−1 . . . . . . . . .3. . . .Contents 6. . . A+Cxn−1 6.9. . . . . . . . . . . . . . . . . . . . . . . . . . . α+βxn 9. . . . . . 3)-Type Equations 131 7. .1 Invariant Intervals . . . . . . . . . . . 134 8 (3. . . . . . . . . . . 9. . . .2 Invariant Intervals . . . . . . . . . 133 +Cx 7. . n−1 6. . . . . . . . . . .8 The Case α = B = 0 : xn+1 = βxn +γxn−1 . . . . . 1)-Type Equations 137 8. 9. . .3 Semicycle Analysis . . . . . . . . . . . . . . . . . . . . . . α+γxn−1 9. . . . . . . . . . . . . . . . . .7. . . βxn +γx 6. . . . . . . . . . . . .5 Open Problems and Conjectures . . . . . . . . . . . . . . . . . . . .5 Long-term Behavior of Solutions When q > 1 6. . . . . . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . .2 The Case β = γ = 0 : xn+1 = A+Bxnα n−1 . . . . .4. . . . . . .

.1 Introduction . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . .2. . . . . . .4 The Case A = 0 : xn+1 = α+βx+Cxn−1 . . . Convergence Results . . . . . . . . .2. . . . . .3. . . . . . . . . . . . . . . . . .4. . . .1 Proof of Part (b) of Theorem 10. . . Invariant Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 11. . . . . . . . . . . . . . . . . . . . . . .2 Proof of Part (c) of Theorem 10. . . . . . . . . .1 Invariant Intervals . . . .viii 10 (3.2 Convergence of Solutions . . . . . . . . . . . . . . . .1 Global Stability . . . . . . . . . . . . .3 Convergence of Solutions . . . . . . . .1 Invariant Intervals . . . . . n +γxn−1 10. . .2. . . . . . .3. . . . . 11. 10. . . Bxn 10. . . . . . . . . . 11 The 11. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 The Case B = 0 : xn+1 = α+βxn +γxn−1 . . . . 10. . . . . . 10. . . . . . . . . . . . . . . 10. . . . . . . . . . . . . . . . . . . A+Cxn−1 10. . . . . . α+βx (3. .1 10. . . . . . A Global Attractivity for Higher Order Equations Bibliography Index . . . . . . . . . . . . . . . . . . . . . A+Bxn 10. . . . . . . . . . . . 3)-Type Equation xn+1 = A+Bxn +γxn−1 n +Cxn−1 Linearized Stability Analysis . . . . . . . . . . . .1 11. . 2)-Type Equations 10.4 Open Problems and Conjectures . . . .5 Open Problems and Conjectures . . . . . . .3 Contents 167 167 167 169 172 172 173 173 174 175 177 179 180 183 183 184 189 191 192 201 211 217 . . . . .1 10.2. . . . . . . . . . . . . .2 The Case C = 0 : xn+1 = α+βxn +γxn−1 . 10. . . . . . . . . . . .4. . . . . . . . . . . . . . . . . .3. . . .2 Global Stability When p ≥ r . . . . . . . . 11.

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

.

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

.

To Senada and Theodora .

.

and the periodicity of solutions of all rational diﬀerence equations of the form xn+1 = α + βxn + γxn−1 . we present the known results and derive several new ones on the boundedness. B. . Eq(1) includes among others the following well known classes of equations: 1. such as [1]. γ. 1. C are nonnegative real numbers and the initial conditions x−1 and x0 are arbitrary nonnegative real numbers such that A + Bxn + Cxn−1 > 0 for all n ≥ 0. [41].Introduction and Classiﬁcation of Equation Types In this monograph. 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. the global stability. [42]. A. . 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.6 where. a large number of equations whose dynamics have not been thoroughly established yet. The Riccati diﬀerence equation when γ = C = 0. [21]. etc. A + Bxn + Cxn−1 n = 0. . This equation is studied in many textbooks on diﬀerence equations. It is an amazing fact that Eq(1) contains as special cases. β. See Section 1. 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. (1) where the parameters α. in addition to the basic description of the solutions 1 .

See [24] and Sections 2. For example. .2. 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. 1 + xn n = 0. 75). . See [42]. [55].2 Introduction and Classiﬁcation of the Riccati equation. . [32]. 3. 1. [44]. [67]. A characteristic feature of this equation is that it possesses an invariant which can be used to understand the character of its . and the rational equation xn+1 = α + αxn + βxn−1 . See [13].2. The problem of existence of solutions for diﬀerence equations is of paramount importance but so far has been systematically neglected. 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 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. [66]. we introduce the forbidden set of the equation.7 and 10. 1. n = 0. See ([42]. / if and only if x0 ∈ F.4. and Section 4. Lyness’ equation when γ = A = B = 0. . An example of such a problem is the following: Open Problem Find all initial points (x−1 . 2. and Section 5. This is a fascinating equation with many interesting extensions and generalizations. . See [13]. [18]. [57]. Hence the solution {xn } of Eq(2) exists for all n ≥ 0. . and Section 1. [49]. [69]. it is interesting to note the similarities of the dynamics of the population model xn+1 = α + βxn−1 e−xn .6 for some results in this regard. p. . Pielou’s discrete delay logistic model when α = γ = B = 0. x0 ) ∈ R × R through which the equation xn−1 xn+1 = −1 + xn is well deﬁned for all n ≥ 0.

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

.

1). xn = x for n ≥ 0 is a solution of Eq(1. [33]. The reader may just glance at the results in this chapter and return for the details when they are needed in the sequel. or equivalently. [22]. See also [1]. 5 . Then for every set of initial conditions x0 . that is. n = 0.1.1). . [63]. xn−1 ). x−1 ∈ I. x is a ﬁxed point of f . x). [21]. [4]. and [68]. [41]. Deﬁnition 1. [34]. (1. 1.1 Deﬁnitions of Stability and Linearized Stability Analysis f :I ×I →I Let I be some interval of real numbers and let be a continuously diﬀerentiable function.Chapter 1 Preliminary Results In this chapter we state some known results and prove some new ones about diﬀerence equations which will be useful in our investigation of Eq(1).1) has a unique solution {xn }∞ . 1. . n=−1 A point x ∈ I is called an equilibrium point of Eq(1.1) if x = f (x. . the diﬀerence equation xn+1 = f (xn . [16].1 Let x be an equilibrium point of Eq(1.

(vi) The equilibrium x of Eq(1. we have n→∞ lim xn = x.1).2) p= is called the linearized equation associated with Eq(1. x−1 ∈ I we have lim xn = x. there exists δ > 0 such that for all x0 .3) Let lie in the open unit disk |λ| < 1.1. if there exists r > 0 such that for all x0 .1) is called a source.1) is called unstable if it is not stable. x) ∂u ∂v denote the partial derivatives of f (u.1) is locally asymptotically stable. (iii) The equilibrium x of Eq(1. . ∂f ∂f (x. (1. or a repeller. and if there exists γ > 0 such that for all x0 . Clearly a source is an unstable equilibrium. (v) The equilibrium x of Eq(1. there exists N ≥ 1 such that |xN − x| ≥ r. then the equilibrium x of Eq(1. . (ii) The equilibrium x of Eq(1. Then the equation yn+1 = pyn + qyn−1 .1) is called locally asymptotically stable if it is locally stable.1) about the equilibrium point x.1 (Linearized Stability) (a) If both roots of the quadratic equation λ2 − pλ − q = 0 (1. Theorem 1. x−1 ∈ I with 0 < |x0 − x| + |x−1 − x| < r. v) evaluated at the equilibrium x of Eq(1. x) and q = (x. x−1 ∈ I with |x0 − x| + |x−1 − x| < δ. x−1 ∈ I with |x0 − x| + |x−1 − x| < γ. 1. . Preliminary Results (i) The equilibrium x of Eq(1. n→∞ (iv) The equilibrium x of Eq(1. we have |xn − x| < ε for all n ≥ −1. n = 0. .1) is called a global attractor if for every x0 .1) is called locally stable if for every ε > 0.6 Chapter 1.1) is called globally asymptotically stable if it is locally stable and a global attractor.

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).

. ≤ M1 ≤ M0 and Set mi ≤ xk ≤ Mi m = lim mi i→∞ for k ≥ 2i + 1.7 below are new and deal with the remaining two cases relative to the monotonic character of f (x. The next two theorems are slight modiﬁcations of results which were obtained in [54] and were developed while we were working on a special case of the equation that we are investigating in this monograph. Global Attractivity of the Positive Equilibrium 11 Very often the best strategy for obtaining global attractivity results for Eq(1.1. set Mi = f (Mi−1 . . at the end of this monograph. . 2. y) is non-decreasing in x ∈ [a. b]. Proof.4. Theorem 1. b] × [a. . .9. y). (b) If (m.4 and 6.8 and 1. and f (x. See Sections 6.4. In this regard there are four possible scenarios depending on whether f (x. b] for each y ∈ [a. b] → [a. i→∞ and M = lim Mi . y) is nonincreasing in y ∈ [a. y) is non-decreasing in x and non-increasing in y and the second theorem deals with the case where the function f (x. See also the Appendix. b].1) converges to x. y) is monotonic in its arguments.1) has a unique equilibrium x ∈ [a.8 where these theorems are utilized. y) is non-increasing in x and nondecreasing in y. b] be an interval of real numbers and assume that f : [a. Set m0 = a and M0 = b and for i = 1. m0 ≤ m1 ≤ . Now observe that for each i ≥ 0. . b] for each x ∈ [a.4. b] is a solution of the system f (m. b] and every solution of Eq(1.5 ([54]) Let [a. or nonincreasing in both arguments. M ) ∈ [a. ≤ Mi ≤ . then m = M .) The ﬁrst theorem deals with the case where the function f (x.1) is to work in the regions where the function f (x. ≤ mi ≤ . b] is a continuous function satisfying the following properties: (a) f (x. . b] × [a. Theorems 1. which describes the extension of these results to higher order diﬀerence equations. mi−1 ) and mi = f (mi−1 . y) is nondecreasing in both arguments. or nonincreasing in one and nondecreasing in the other. m) = M. . (See Section 6.4. . . M ) = m and f (M. Then Eq(1. Mi−1 ).

. m0 ≤ m1 ≤ .6 ([54]) Let [a. b] and every solution of Eq(1. . Now observe that for each i ≥ 0. b] for each x ∈ [a. b] → [a. m) and M = f (m. Mi−1 ) and mi = f (Mi−1 . 2 . b].4. and Set mi ≤ xk ≤ Mi m = lim mi i→∞ for k ≥ 2i + 1. y) is nondecreasing in y ∈ [a.12) holds and by the continuity of f . ≤ Mi ≤ . b].1) has a unique equilibrium x ∈ [a. m = f (M. Preliminary Results M ≥ lim sup xi ≥ lim inf xi ≥ m i→∞ i→∞ (1. y) is non-increasing in x ∈ [a. b] be an interval of real numbers and assume that f : [a. In view of (b). m = f (m. . . b] is a continuous function satisfying the following properties: (a) f (x. . i→∞ and M = lim Mi . Set m0 = a and M0 = b and for i = 1.1) converges to x. set Mi = f (mi−1 .12 Then Chapter 1. M ) and M = f (M. Proof. m=M from which the result follows. 2 Theorem 1. m=M from which the result follows. Therefore in view of (b). (b) The diﬀerence equation Eq(1.1) has no solutions of prime period two in [a. ≤ M1 ≤ M0 . Then clearly (1. m).12) and by the continuity of f . Then Eq(1. mi−1 ). . b] for each y ∈ [a. . 2. M ). b]. . ≤ mi ≤ . b] × [a. . and f (x.

4. b] is a continuous function satisfying the following properties: (a) f (x. Proof. has a unique positive solution. b] × [a. b] is a solution of the system f (m. . . b] → [a. b] be an interval of real numbers and assume that f : [a. 2 Theorem 1. b] × [a. m). b] × [a. 13 then m = M . y) is non-increasing in each of its arguments. ≤ Mi ≤ . Now observe that for each i ≥ 0. 2. In view of (b). .8 Let [a. (b) The equation f (x. b] is a continuous function satisfying the following properties: (a) f (x. mi−1 ) and mi = f (Mi−1 . m = f (M. y) is non-decreasing in each of its arguments. . .1) converges to x. m=M =x from which the result follows.4. Then clearly (1. ≤ mi ≤ . b] → [a. . Then Eq(1.1) has a unique equilibrium x ∈ [a. (b) If (m. . i→∞ and M = lim Mi .7 Let [a. . . and Set mi ≤ xk ≤ Mi m = lim mi i→∞ for k ≥ 2i + 1. . set Mi = f (mi−1 . b] and every solution of Eq(1.12) holds and by the continuity of f . Set m0 = a and M0 = b and for i = 1. Global Attractivity of the Positive Equilibrium Theorem 1.1) converges to x. Then Eq(1. Mi−1 ). m) = M and f (M. M ) ∈ [a.1. b] be an interval of real numbers and assume that f : [a. ≤ M1 ≤ M0 .1) has a unique equilibrium x ∈ [a. M ) and M = f (m. M ) = m. x) = x. m0 ≤ m1 ≤ . b] and every solution of Eq(1.4.

M ). B ∈ J. . n ≥ −k + 1. . mi−1 ). m) and M = f (M. In this section we ﬁrst give a self-contained version of limiting solutions the way we will use them in this monograph. Now observe that for each i ≥ 0. . Assume that {xn }∞ n=−1 is a solution of Eq(1. and let k ≥ 1 be a positive integer. and mi ≤ xk ≤ Mi Set m = lim mi i→∞ for k ≥ 2i + 1.12) is satisﬁed and by the continuity of f . 2. m0 ≤ m1 ≤ . set Mi = f (Mi−1 . m = M = x. . . Set Chapter 1. Let J be some interval of real numbers. Now with the above hypotheses.13) .14 Proof. . . (1. J]. . from which the result follows. . Mi−1 ) and mi = f (mi−1 . In view of (b). . . ≤ Mi ≤ . Then clearly the inequality (1. ≤ mi ≤ . Let L be a limit point of the solution {xn }∞ n=−1 (For example L is the limit superior S or the limit inferior I of the solution). ∞ we will show that there exists a solution {Ln }n=−k of the diﬀerence equation xn+1 = f (xn . where A. ≤ M1 ≤ M0 . m = f (m. f ∈ C[J × J. Preliminary Results m0 = a and M0 = b and for i = 1.1) such that A ≤ xn ≤ B for all n ≥ −1. . xn−1 ). 2 1. i→∞ and M = lim Mi .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.

It follows by induction that the sequence L−k . lim xnij = L and lim xnij −1 = L−1 . . L1 = 1.13) with Then clearly.5. xn−1 ). (See Theorem 1. −k ) = f (L−k+1 . L−k+1 .14) associated with the solution {xn }∞ n=−1 of Eq(1.1). satisﬁes Eq(1. L−1 . . i→∞ i→∞ Continuing in this way and by considering further and further subsequences we obtain ∞ numbers L−2 . = f( −k+1 . n=−1 The following result. . j→∞ lim xnj −k = L−k . .. (1. the term LN is a limit point of the solution {xn }∞ . xnj −k ) = lim xnj −k+2 = L−k+2 . . . L0 = L = 0 .5.) ∞ We now proceed to show the existence of a limiting solution {Ln }n=−k .. Limiting Solutions such that L0 = L 15 and for every N ≥ −k. Therefore.13) and every term of this sequence is a limit point of the solution {xn }∞ . Karakostas takes k = ∞ and calls the solution {Ln }∞ n=−∞ a full limiting sequence of the diﬀerence equation xn+1 = f (xn .1).13) associated with the The solution {Ln }∞ n=−k solution {xn }∞ n=−1 of Eq(1. n=−1 is called a limiting solution of Eq(1. .2 below. L−3 . . Clearly there exists a subsequence {xni }∞ of {xn }∞ i=1 n=−1 such that i→∞ lim xni = L. which is a consequence of the above discussion.1. . −k Let { n }∞ n=−k be the solution of Eq(1. L−k ) j→∞ = lim f (xnj −k+1 . L−k and a subsequence {xnj }n=−1 such that j→∞ lim xnj = L j→∞ lim xnj −1 = L−1 . . . is stated in the way we will use it in this monograph: . −k+2 j→∞ = L−k and −k+1 = L−k+1 . B] and so it has a ∞ further subsequence {xnij }j=1 which converges to some limit which we call L−1 . ∞ Next the subsequence {xni −1 }i=1 of {xn }∞ n=−1 also lies in interval [A.

. S ∈ J. .2 to diﬀerence equations of the form xn+1 = Ai . S] for all n ∈ Z and such that for every N ∈ Z. . and let {xn }∞ n=−ν be a bounded solution of the diﬀerence equation xn+1 = f (xn . J]. f ∈ C[J ν+1 . . . J]. and let k ≥ 1 be a positive integer. with I = lim inf xn . IN and SN are limit points of {xn }∞ . S] for n ≥ −k. . 1.1) with limit inferior I and limit superior S. which by the change of variables xn = y1 reduce to rational diﬀerence equations of the n form 1 . . . .16) associated with the solution {xn }∞ n=−k of Eq(1. . 0. Let Z denote the set of all integers {. . n→∞ S = lim sup xn n→∞ n = 0.16) lim xrn +N = IN and n→∞ lim xln +N = SN for every N ≥ m. Sn ∈ [I. . Then there exist two solutions ∞ {In }∞ n=−∞ and {Sn }n=−∞ of the diﬀerence equation xn+1 = f (xn . n = 0. as developed by Karakostas. 1. Preliminary Results Theorem 1. . and In . (1. with I0 = I.5. . . . 1. .5.1 Let J be some interval of real numbers.5. n=−ν Furthermore for every m ≤ −ν. In . . −1. i=0 xn−i k n = 0. with I. . ∞ The solutions {In }∞ n=−∞ and {Sn }n=−∞ of the diﬀerence equation (1. xn−ν ). f ∈ C[J × J. .2 ([37]) Let J be some interval of real numbers.15). . .16 Chapter 1. See [23] and [25] for an application of Theorem 1. S ∈ J.}. . 1. yn+1 = k i=0 Ai yn−i .16) are called full limiting solutions of Eq(1. S0 = S. Then Eq(1. xn−1 . . S0 = S We now state the complete version of full limiting sequences. Suppose that {xn }∞ n=−1 is a bounded solution of Eq(1. xn−ν ) which satisfy the equation for all n ∈ Z. Theorem 1. there exist two subsequences {xrn } and {xln } of the solution {xn }∞ n=−ν such that the following are true: n→∞ (1.15) and with I.13) has two solutions {In }∞ n=−k with the following properties: and {Sn }∞ n=−k I0 = I. xn−1 . Sn ∈ [I.

6 The Riccati Equation α + βxn . .17) when x0 ∈ F.17) will become zero for some value of n ≥ 0 is called the forbidden set F of Eq(1.17) reduces to βA + βxn β xn+1 = B = for all n ≥ 0.18) Indeed when B = 0. Theorem 1. In the sequel.) The other goal we have is to describe in detail the long. One of our goals in this section is to determine the forbidden set F of Eq(1. β.21) β B (1. in addition to (1.6. B.and short-term behavior of / solutions of Eq(1. Then (1. The proof is straightforward and will be omitted. (1. The ﬁrst result in this section addresses a very special case of Eq(1. Eq(1. γ. .17) regarding period-two solutions.17) possesses a prime periodtwo solution. x0 are all real numbers) other than the material presented here.18). A. β. To avoid degenerate cases.17) is a linear equation and when αB − βA = 0. (See also [13] and [69].17).17) where the parameters α. (1.1. we will assume that B = 0 and αB − βA = 0.6. we will assume that β + A = 0.18) holds and that Eq(1. which is extracted from [32]. It should be mentioned here that there is practically nothing known about the forbidden set of Eq(1) (when the parameters α.20) . C and the initial conditions x−1 . (1. A + Bxn B The set of initial conditions x0 ∈ R through which the denominator A+Bxn in Eq(1.19) β + A = 0. . The Riccati Equation 17 1.17). A.1 Assume that (1. 1. A + Bxn A diﬀerence equation of the form xn+1 = n = 0. Furthermore when (1. Eq(1.17) with x0 = is periodic with period two. B and the initial condition x0 are real numbers is called a Riccati diﬀerence equation.19) holds every solution of Eq(1.

1.17).22). 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 . w Also note that the signum of R is equal to the signum of the derivative y =1− f (x) = βA − αB .23) βA − αB (β + A)2 is a nonzero real number. Similar results can be obtained for Eq(1. (A + Bx)2 of the function For the remainder of this section we focus on Eq(1. Preliminary Results for n ≥ 0 (1.22) transforms Eq(1. wn n = 0.18 Now observe that the change of variable xn = β+A A wn − B B Chapter 1.23) it follows that when R< 1 4 .17) into the diﬀerence equation wn+1 = 1 − where R= R . . (1. . called the Riccati number of Eq(1. .23).23) and present its forbidden set F and the character of its solutions. From Eq(1.17) through the change of variables (1. It is interesting to note that the four parameters of Eq(1.17) have been reduced to the single parameter R of Eq(1.

1.24) are the equilibrium points of Eq(1. 4 1− √ √ 1 − 4R 1 + 1 − 4R < = w+ 2 2 19 When Eq(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. . n = 0.23). 2 Finally. Note that in these two cases the characteristic roots of Eq(1.6. The forbidden set of Eq(1. which can be solved explicitly. 4 1 R= .24) yn+1 yn for n = 0.23) has exactly one equilibrium point namely.23) is clearly the set of points where the sequence {yn }∞ n=0 vanishes. respectively.1.24) are real numbers. 4 Eq(1. where 4 the characteristic roots of Eq(1. 1 w= . ). . when 1 R> .23) has the two equilibrium points w− and w+ with w− = and |w− | < w+ . . . Now observe that the change of variables wn = with y0 = 1 and y1 = w0 reduces Eq(1.23) has no equilibrium points. (1. 1 The next two theorems deal with the cases R < 1 and R = 4 . The Riccati Equation Eq(1. . .

27) . 2. 4 1 2 is the only equilibrium point of Eq(1.23) is the sequence of points fn = n−1 n−1 w+ − w− w+ w− n n w+ − w− for n = 1.23) is given by / n+1 n+1 (w0 − w− )w+ − (w+ − w0 )w− . . if R > 0 and fn < w− while (fn − w− )(fn+1 − w− ) < 0 if R < 0. . .3 Assume Then w= 1 R= .6. wn = n n (w0 − w− )w+ − (w+ − w0 )w− n = 1. the solution of Eq(1.2 Assume Then 1 R< . .26) It is now clear from Theorem 1. (1.6.6. (1. . . The forbidden set F of Eq(1. 2. Preliminary Results √ √ 1 − 4R 1 + 1 − 4R w− = and w+ = 2 2 are the only equilibrium points of Eq(1. 4 1− Chapter 1.23). 2. . . Theorem 1. .23) is the sequence of points fn = n−1 2n for n = 1. The forbidden set F of Eq(1. .25) When w0 ∈ F. .23). . (1.2 that n→∞ lim fn = w− .20 Theorem 1. The equilibrium w+ is a global attractor as long as w0 ∈ F ∪ {w− } / and the equilibrium w− is a repeller.

The Riccati Equation 21 which converges to the equilibrium from the left. . 1.3 it follows that the equilibrium w = 1 of Eq(1. Hence wn = √ √ 4R − 1 cos(n + 1)φ + (2w0 − 1) sin(n + 1)φ √ . .1.23) is given by / wn = 1 + (2w0 − 1)(n + 1) 2 + 2(2w0 − 1)n for n = 0. more precisely.23) is a global 2 / attractor for all solutions with w0 ∈ F but is locally unstable. 1. .6. π φ ∈ (0.24) 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 .28) From Theorem 1. . 2 R n = 0. . . . (1. . 1.29) n Finally consider the case √ R. 4R − 1 n = 0. yn = R 2 (C1 cos(nφ) + C2 sin(nφ)).6. the solution of Eq(1. ) 2 and √ 4R − 1 √ sin φ = . R 4R − 1 cos(nφ) + (2w0 − 1) sin(nφ) . . (1. 1 R> . When w0 ∈ F. 4 The characteristic roots of Eq(1. . it is a sink from the right and a source from the left. 2w0 − 1 C1 = 1 and C2 = √ .

2 Let θ ∈ (− π . 2. wn = (4R − 1)2 + (2x0 − 1)2 .6. the solution of Eq(1. . when R > 1 . .22 Chapter 1. . φ= π k π ∈ (0. For example.4 Assume that R> and let φ ∈ (0. that is.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. .23) is given by (1. . that is. assume φ is a rational multiple of π.32) . (1. π ) be such that 2 1 cos φ = √ 2 R and 1 4 √ 4R − 1 √ sin φ = . .23).31) from which the character of the solutions of Eq(1. . 2. 2 n = 0.29). . . can now be easily 4 revealed. 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). 2 2 When w0 ∈ F. w0 = − 2 2 The following theorem summarizes the above discussion. π ) is a rational multiple of π. . / (1. Theorem 1. π ) be such that 2 2 √ 4R − 1 2x0 − 1 and sin θ = cos θ = r r where r= Then we obtain.23) is the sequence of points √ 1 4R − 1 fn = − cot(nφ) for n = 1. √ 4R − 1 1 cot(nφ) for n = 1. ) N 2 (1. 1. Preliminary Results as long as the denominator is diﬀerent from zero. . 2 R Then the forbidden set F of Eq(1.

. then the following result is true: Theorem 1.5 is not a rational multiple of π.23) is periodic with period N . p Then every solution of Eq(1.6. . √ √ 1 1 4R − 1 4R − 1 wN = − tan(kπ − θ) = − tan(−θ) = w0 2 2 2 2 and so every solution of Eq(1.23) is periodic. The Riccati Equation 23 for some k. . 2.6. p − 1 . ) p 2 1 4 cos2 φ with φ = k π N Then where k and p are positive comprimes and suppose that √ 4R − 1 1 √ cos φ = √ and sin φ = .6.23) with w0 ∈ F is dense in R.5 is not a rational multiple of π. N ∈ {1.} and 2k < N. 2 R 2 R or equivalently k 4R cos2 ( π) = 1. . Equivalently assume that A= where k.1.23) 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. 2. then every solution of Eq(1.23) is periodic with period N . .6. . / for n = 1.6.6 Assume that the number φ in Theorem 1. .} with 2k < N.5 Assume that φ= k π π ∈ (0. When the number φ in Theorem 1. . Theorem 1. (ii) The set of limit points of a solution of Eq(1. . . 2. The following theorem summarizes the above discussion. N ∈ {1. Then the following statements are true: (i) No solution of Eq(1.

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

.33) consists of a “string” of terms {xl . ∞)] is such that: f (x. xm }. Proof. (See Section 6. xN −1 ) < f (x. Then by using the monotonic character of f (x. (b) A sequence {xn } is said to oscillate about x if the sequence xn − x oscillates. xl+1 .33). ∞). ∞) × (0. The sequence {xn } is called strictly oscillatory about x if the sequence xn − x is strictly oscillatory.7. . . all less than the equilibrium x.5. Thus xN +1 < x < xN +2 or m < ∞ and xm+1 ≥ x. x) = x.7. with l ≥ −1 and m ≤ ∞ and such that either l = −1.33) with at least two semicycles. Semicycle Analysis 25 A negative semicycle of a solution {xn } of Eq(1. Let {xn } be a solution of Eq(1. Let x be a positive equilibrium of Eq(1.1. x) = x and xN +2 = f (xN +1 . (0. there exist n1 . Then except possibly for the ﬁrst semicycle. A sequence {xn } is called strictly oscillatory if for every n0 ≥ 0. xN ) > f (x. 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. y) is decreasing in x for each ﬁxed y. Deﬁnition 1. every solution of Eq(1.7. y) is increasing in y for each ﬁxed x. or l > −1 and xl−1 ≥ x and either m = ∞.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. n2 ≥ n0 such that xn1 xn2 < 0.1 ([28]) Assume that f ∈ C[(0. y) we have xN +1 = f (xN .33) has semicycles of length one. Then there exists N ≥ 0 such that either xN −1 < x ≤ xN or xN −1 ≥ x > xN . We will assume that xN −1 < x ≤ xN . Otherwise the sequence is called nonoscillatory. and f (x. .) Theorem 1.

Chapter 1. which is a contradiction. (0. Assume that {xn } is an oscillatory solution with two consecutive terms xN −1 and xN in a positive semicycle xN −1 ≥ x and xN ≥ x with at least one of the inequalities being strict. Proof.33) has semicycles of length at most two. ∞).7. xN −1 ) > 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. y) is increasing in both arguments. y) is increasing in x and decreasing in y. Then by using the decreasing character of f we obtain: xN +1 = f (xN . Let x be a positive equilibrium of Eq(1. Then by using the increasing character of f we obtain: xN +1 = f (xN . xN −1 ) < f (x. every oscillatory solution of Eq(1. The next result applies when the function f is increasing in both arguments. 2 . Then except possibly for the ﬁrst semicycle. ∞). y) is decreasing in both arguments. ∞) × (0. Let x be a positive equilibrium of Eq(1. ∞)] and that f (x.33) has semicycles of length one.33). The proof in the case of negative semicycle is similar and is omitted. Proof.7.3 Assume that f ∈ C[(0.2 Assume that f ∈ C[(0.33). It follows by induction that all future terms of this solution belong to this positive semicycle. x) = x which completes the proof. Preliminary Results 2 The next result applies when the function f is decreasing in both arguments. ∞)] and that f (x. Theorem 1. with at least one of the inequalities being strict. (0. ∞) × (0. x) = x which shows that the next term xN +1 also belongs to the positive semicycle. Theorem 1. Then every oscillatory solution of Eq(1. The proof in the case of negative semicycle is similar and is omitted. 2 The next result applies when the function f (x.26 and the proof follows by induction.

and in the second term if the semicycle has three terms. Then by using the increasing character of f we obtain xN +1 = f (xN . xN −1 ) > f (xN . Proof. 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. every oscillatory solution of Eq(1. y) is decreasing in y for each ﬁxed x. ∞)] is such that: f (x.7. if we assume that f (u. (0. Assume that {xn } is an oscillatory solution with two consecutive terms xN −1 and xN such that xN −1 < x ≤ xN .33).34) we obtain: xN +1 = f (xN . xN −1 ) < f (xN . and f (x. The proof in the case xN −1 ≥ x > xN . Semicycle Analysis 27 Theorem 1. If xN > xN −1 > x then by using the increasing character of f and Condition (1. except possibly for the ﬁrst semicycle.34) we obtain xN +1 = f (xN .7. y) < x for every x > y > x (1. u) = x for every u and f (x. x) = x which shows that the next term xN +1 also belongs to the positive semicycle. is similar and is omitted. xN −1 ) > f (x. ∞) × (0. xN ) = x (1. Then by using the increasing character of f and Condition (1.1. ∞). xN ) = x which shows that the positive semicycle has length two.4 Assume that f ∈ C[(0. 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.34) .33) has semicycles of length at least two. Then. Let x be a positive equilibrium of Eq(1. Furthermore.35) then {xn } oscillates about the equilibrium x with semicycles of length two or three. y) is increasing in x for each ﬁxed y.

36) is satisﬁed and that x0 > x−1 > x. we obtain x1 = f (x0 . y) < x for every x < y < x then if x0 < x−1 < x. If.33) has monotonic solutions. y) is increasing in x for each ﬁxed y. and f (x. and the proof follows by induction. f satisﬁes Condition (1. y) is decreasing in y for each ﬁxed x.34) is not enough to guarantee the above result.7. y) > x for every x > y > x. Assume that f satisﬁes Conditions (1. x−1 ) > x0 . Then if x0 > x−1 > x. assume that Condition (1. xN −1 ) < xN and the proof is complete. Preliminary Results xN +1 = f (xN . First. Theorem 1.33).35) we ﬁnd Chapter 1. The proof of the remaining case is similar and will be omitted. x−1 ) = x. then Eq(1.37) .5 Assume that f ∈ C[(0. on the other hand. the corresponding solution is decreasing. ∞)] is such that: f (x. Proof.36) f (x. the corresponding solution is increasing.34) and (1. 2 (1. (0. x−1 ) > f (x−1 . If. and f (x.28 and by using Condition (1. instead of Condition (1. 2 Condition (1. Let x be a positive equilibrium of Eq(1.34). we assume (1. Then by the monotonic character of f and (1.36) we get x1 = f (x0 . ∞).36).34). ∞) × (0.35). In view of Condition (1.

B. . Zero is an equilibrium point of Eq(2. In view of the above restriction on the initial conditions of Eq(2. the equilibrium points of Eq(2.1) are the nonnegative solutions of the equation x= or equivalently (B + C)x2 − (β + γ − A)x − α = 0. in addition to the zero equilibrium . (2. . C ∈ [0.2) and where the initial conditions x−1 and x0 are arbitrary nonnegative real numbers such that the right hand side of Eq(2. Semicycles.1) When (2. ∞) (2.5) holds. Periodicity. B + C ∈ (0. ∞) with α + β + γ. γ.1) if and only if α = 0 and A > 0.1) is well deﬁned for all n ≥ 0. A.3) n = 0.4) α + (β + γ)x A + (B + C)x (2.1). (2. . β.5) (2. A + Bxn + Cxn−1 Here we investigate the equilibrium points of the general equation xn+1 = where α.1 Equilibrium Points α + βxn + γxn−1 . 1. and Invariant Intervals 2.1) has a positive equilibrium if and only if β + γ > A.Chapter 2 Local Stability. 29 . Eq(2.

v) = (A + Bu + Cv)2 (2.4).8). . .) In summary.1) has a positive equilibrium x. v) = A + Bu + Cv and observe that (βA − αB) + (βC − γB)v fu (u. then x is unique.1). Finally when α>0 the only equilibrium point of Eq(2.2).7) β+γ−A . . and [69] for some results in this regard. A + Bxn there is very little known.4) would also be an equilibrium worth investigating. [18].6 about the Riccati equation α + βxn xn+1 = . [32]. Set α + βu + γv f (u. Periodicity.3) and (2. except for our presentation in Section 1.30 Chapter 2. This observation simpliﬁes the investigation of the local stability of the positive equilibrium of Eq(2.1). and Invariant Intervals In fact in this case the positive equilibrium of Eq(2. it is interesting to observe that when Eq(2. when α = 0. it satisﬁes Eqs(2.1) is positive and is given by x= β+γ . the quantity β + γ is positive. Local Stability. . If we had allowed negative initial conditions then the negative solution of Eq(2. and it is given by (2.9) . B+C (2. Semicycles.2 Stability of the Zero Equilibrium Here we investigate the stability of the zero equilibrium of Eq(2.1) when the initial conditions and the coeﬃcients of the equation are real numbers is of great mathematical importance and. 2. The problem of investigating Eq(2.1) is unique and is given by x= When α = 0 and A = 0 the only equilibrium point of Eq(2. B+C (2.8) of the quadratic equation (2. 1.4).6) Note that in view of Condition (2. (See [13].1) is the positive solution x= β+γ−A+ (β + γ − A)2 + 4α(B + C) 2(B + C) (2. n = 0.

x). . 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. x) and q = fv (x.1.1. n = 0.1) we have p= β A and q = γ A and so the linearized equation associated with Eq(2.10) If x denotes an equilibrium point of Eq(2.1).3.1) about the zero equilibrium point is β γ zn+1 − zn − zn−1 = 0. A A For the stability of the zero equilibrium of Eq(2. .1).1) about the equilibrium point x is zn+1 − pzn − qzn−1 = 0 where p = fu (x.1. (A + Bu + Cv)2 31 (2. . in addition to the local stability Theorem 1.2. we have the luxury of the global stability Theorem 1.1 Assume that A > 0. For the zero equilibrium of Eq(2. is globally asymptotically stable when β+γ ≤A and is unstable when β + γ > A. 1. . As a consequence of these two theorems we obtain the following global result: Theorem 2.2. 1. . .2. then the linearized equation associated with Eq(2. Then the zero equilibrium point of the equation xn+1 = βxn + γxn−1 .1. A + Bxn + Cxn−1 n = 0. The local stability character of x is now described by the linearized stability Theorem 1. .1. v) = (γA − αC) − (βC − γB)u . Stability of the Zero Equilibrium and fv (u. .

In these cases the positive equilibrium x is unique.1) has a positive equilibrium when either or α > 0.15).1.14) (2.11) and q > −1. By using the identity (B + C)x2 = (β + γ − A)x + α.8). Inequalities (2.1. Eq(2. x) = and q = fv (x. it satisﬁes Eqs(2.3) and (2. and it is given by (2.4).12) (A + β + γ)x > −α − (2.13) (2.3 Local Stability of the Positive Equilibrium α = 0 and β + γ > A As we mentioned in Section 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 (2.12) 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. 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. Semicycles. Local Stability.1) the following: |p| < 1 − q (2.13)-(2.32 Chapter 2.11) and (2. are equivalent to some inequalities of the form x>ρ . Periodicity. x) = (γA − αC) − (βC − γB)x (γA − αC) − (βC − γB)x = 2 .15) A2 + (γA − αC) βC − γB x > −α − . and Invariant Intervals 2. we see that p = fu (x.

4 When is Every Solution Periodic with the Same Period? 1 .) Theorem 2. n = 0.1) with the property that every solution of the equation is periodic with the same period. 33 2. 5. . . ∞).18) (2.16) (2. = xn−1 are remarkable in the sense that all positive solutions of each of these four nontrivial equations are periodic with periods 2.1) xn+1 = xn+1 = xn+1 = xn+1 n = 0. . Then A = B = γ = 0. 1. Then the following statements are true: (i) Assume C > 0.4. . . . . . (2. 1. n = 0. (See also [75].1) is periodic with period p. The following four special examples of Eq(2. and 6 respectively. 1. .1 Let p ≥ 2 be a positive integer and assume that every positive solution of Eq(2. and [74]. xn−1 xn . 4. .2. The following result characterizes all possible special cases of equations of the form of Eq(2. σ ∈ [0. .4. Then γ(α + β) = 0. it is clear that F (x) = 0 and that x > ρ if and only if F (ρ) < 0 while x<σ if and only if F (σ) > 0. n = 0.17) (2. . Now if we set F (u) = (B + C)u2 − (β + γ − A)u − α.19) 1 + xn . . When is Every Solution Periodic with the Same Period? and/or x<σ for some ρ. (ii) Assume C = 0. [73]. xn 1 xn−1 . 1.

1 Let p ∈ {2. A+B C (2. Thus (2.1) xp = Hence (A + B)x0 + (Cx0 − γ)xp−2 = α + β. Consider the solution with x−1 = 1 and x0 ∈ (0. (i) Assume that C > 0.16)-(2. Periodicity. The proof is complete. assume that γ(α + β) > 0. Local Stability. A+B C we see that (2. Hence (2.20) is impossible.1) is periodic with period p. 5. A + B + Cxp−2 and xp−1 = x−1 = 1 we see that (2. .21) we now claim that also γ = 0.20) α + β + γxp−2 = x0 .21) holds.20) is impossible. Semicycles.20) is impossible. Then we claim that A = B = 0.22) holds . and Invariant Intervals Proof. Then by choosing x0 suﬃciently small. for the sake of contradiction.22) Otherwise γ > 0 and by choosing x0 < min α+β γ . 3. Assume that every positive solution of Eq(2.1) reduces to one of the equations (2.34 Chapter 2. In addition to (2. Then up to a change of variables of the form xn = λyn Eq(2. ∞). Then clearly x0 = xp and so from Eq(2. we see that (2. 6}.4. 2 Corollary 2. (ii) Assume C = 0 and. Otherwise A + B > 0 and by choosing x0 > max α+β γ .19) . 4. (2.21) (2.

23) α + βψ + γφ A + Bψ + Cφ α + βφ + γψ A + Bφ + Cψ . φ. . .1) to have a prime period-two solution and we exhibit all prime period-two solutions of the equation. ψ. φ. ψ.26) αC + β(γ − β − A) .27) hold . 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. (See [50]. Furthermore when (2. is a prime period-two solution of Eq(2.27) (2. .2.1) if and only if α + β(φ + ψ) = Bφψ with φ. . Existence of Prime Period-Two Solutions 35 2.1).5 Existence of Prime Period-Two Solutions In this section we give necessary and suﬃcient conditions for Eq(2. solution with period p and furthermore the set of solutions of the equation with prime period p is nonempty.24) (2. . φ.1) converges to a period-two solution in a nontrivial way according to the following convention.1) has a prime period-two solution if and only if γ = β + A. . . .5.26) and (2. φ. Assume that . ∞) and φ = ψ (2. Convention Throughout this book when we say that “every solution of a certain diﬀerence equation converges to a periodic solution with period p” we mean that every solution converges to a. ψ ∈ [0. C (2.25) (2. . . . is a prime period-two solution of Eq(2. not necessarily prime. ψ.) Furthermore we are interested in conditions under which every solution of Eq(2.

. C C with 0 ≤ φ < γ−A . α = β = 0. .6 Let Local Asymptotic Stability of a Two Cycle . the two cycle is . Then it follows from (2.1) possesses a prime period-two solution .24) holds and that Eq(2. . .23) and (2. . ψ. . 0. . φ. and Invariant Intervals or equivalently β φ> . . 1. φ. C C Finally when (2. .. .30) (2..31) holds. be a two cycle of the diﬀerence equation (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. ψ. ψ. . the two cycle is . φ.29) (2. . B φ= β+ √ β 2 + αB . . and φψ > 0. Periodicity. and φψ ≥ 0. φ. . and φψ = 0.. B = C. . .1). C γ−A γ−A . α = β = 0.25) that one of the following three conditions must be satisﬁed: B = C. . Semicycles. . . (2. − φ.28) Next assume that (2. . φ. B and ψ = α + βφ .30) holds. γ−A γ−A − φ. 0.31) γ > β + A and When (2. . When (2. . −β + Bφ (2. (γ − β − A)[B(γ − β − A) − C(γ + 3β − A)] α< . Set un = xn−1 and vn = xn for n = 0. Local Stability.29) holds. . . B > C.32) 4C 2 2. . . ψ. . α + β > 0. That is.36 Chapter 2. (2. . φ.

ψ) = . ∂v (A + Bψ + Cφ)2 [(βA − αB) − (γB − βC)ψ][(γA − αC) + (γB − βC)ψ] ∂h (φ. n = 0. . v) h(u. ψ ∂h ∂h (φ. ∞) deﬁned by: T Then u v = v βv+γu+α Bv+Cu+A 37 . Furthermore T2 where g(u. v) = u v = g(u. φ ψ is a ﬁxed point of T 2 . ψ) = . 1. ∂u (A + Bφ + Cψ)2 (A + Bψ + Cφ)2 . v) = α + βv + γu A + Bv + Cu and h(u. ψ) ∂v (φ.2. n +Cun Let T be the function on [0. ψ) = . ∂u (A + Bψ + Cφ)2 ∂g (βA − αB) − (γB − βC)φ (φ. evaluated at ψ By deﬁnition ∂g ∂g (φ. . . ∞) × [0. ψ) ∂u φ = JT 2 . v) α+βv+γu α + β A+Bv+Cu + γv α+βv+γu A + B A+Bv+Cu + Cv . . 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. JT 2 . ψ) ∂u we ﬁnd the following identities: (γA − αC) + (γB − βC)ψ ∂g (φ. Local Asymptotic Stability of a Two Cycle and write Eq(2.6.1) in the equivalent form: un+1 = vn α+βv vn+1 = A+Bvn +γun . the second iterate of T . ψ) ∂v (φ.

φ.1).33) converges to a period-two solution. 2. be a two cycle of Eq(2. and 10. + (A + Bφ + Cψ)2 Set S= and D= ∂h ∂g (φ. .2) to obtain conditions on α. β.1 (c) that both eigenvalues of JT 2 unit disk |λ| < 1. φ.1. . That is. ψ. . 6.9. ψ. B ∈ [0.6.5. 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 Convergence to Period-Two Solutions When C=0 α + βxn + γxn−1 . 6. . ∞) with α + β + γ. Local Stability. β. ψ) ∂u ∂v ∂h ∂g ∂h ∂g (φ. 1. Periodicity. γ. ψ) + (φ. Semicycles.7. A and B so that every solution of Eq(2.1) converges to the two cycle. See Sections 6. ψ) (φ. A. ψ) − (φ.33) . 6. ψ) = ∂v (A + Bφ + Cψ)2 (A + Bψ + Cφ)2 (γA − αC) + (γB − βC)φ . . and Invariant Intervals [(βA − αB) − (γB − βC)ψ][(βA − αB) − (γB − βC)φ] ∂h (φ.7. ∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1. x0 ) through which the solution of Eq(2. (2. . Deﬁnition 2.1 (Basin of Attraction of a Two Cycle) Let . .4 for some applications. . A + Bxn In this section we consider the equation xn+1 = where α. The basin of attraction of this two cycle is the set B of all initial conditions (x−1 . γ. .6. every solution converges to a (not necessarily prime) period-two solution and n = 0. if and only if |S| < 1 + D and D < 1.38 Chapter 2. and 7. ψ) (φ. ψ).

2 where we present the complete character of solutions of Eq(2.33).33). In this case.7. ∞ Lemma 2. In particular. the solution would converge to a period-two solution.2. if and only if J0 > 0 and β + A > 0. (β + A)J0 = 0. Statement (ii) holds if and only if J0 < 0 and β +A > 0. (See also Section 10.27) holds.34) Jn = α + β(xn−1 + xn ) − Bxn−1 xn for n ≥ 0. (ii) xn+1 − xn−1 < 0 for all n ≥ 0. In this case the subsequences {x2n }∞ and {x2n−1 }∞ of the solution {xn }∞ are both decreasing to ﬁnite limits and n=0 n=0 n=0 so in this case the solution converges to a period-two solution. one and only one of the following three statements is true for each solution of Eq(2. if and only if. the sign of the quantity Jn is constant along each solution. Jn+1 = (b) xn+1 − xn−1 = Jn A + Bxn for n ≥ 0. the solution {xn } is periodic with period-two. Clearly (i) holds. Set (2. 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.27) holds. and let {xn }n=−1 be a solution of Eq(2. Now observe that xn+1 − xn−1 = β+A (xn − xn−2 ) for n ≥ 1 A + Bxn . (iii) xn+1 − xn−1 > 0 for all n ≥ 0.33): (i) xn+1 − xn−1 = 0 for all n ≥ 0. Convergence to Period-Two Solutions When C=0 39 there exist prime period-two solutions. Finally (iii) holds. Then the following statements are true: (a) β+A Jn for n ≥ 0. the proof of which follows by straightforward computations.33) converges to a periodtwo solution if and only if (2.) Now before we see that when B > 0 every solution of Eq(2. A + Bxn In particular.1 Assume that (2.7. we need the following result.

Then there exists N ≥ 0 such that ρ= Clearly for all n > N . . Periodicity. + |x2N +2 − x2N | + x2N m for m = 1. (b) When (2. To this end assume. . ≤K ρi + x2N < i=1 which contradicts the hypothesis that the subsequence of even terms is unbounded. . that the subsequence of even terms increases to ∞. A + Bx2n+1 A + Bx2n A + Bx2N +1 A + Bx2N Set Then K = |x2N − x2N −2 |. .33) converges to a period-two solution. Kρ + x2N 1−ρ . ψ.1 Assume B > 0. |x2N +2 − x2N | = A+β A+β < 1.7.40 and so Chapter 2. 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 | + . . . Local Stability. . 2. for the sake of contradiction. Theorem 2.35) We will now employ (2. Then the following statements hold: (a) Eq(2. of Eq(2.28). . The proof for the subsequence of odd terms is similar and will be omitted.33) has a prime period-two solution if and only if (2. A+β A+β A+β A+β < = ρ. . In summary we have established the following result. We will give the details for the subsequence of even terms. . .33) are given by (2. Semicycles.27) holds every solution of Eq(2. and Invariant Intervals n xn+1 − xn−1 = (x1 − x−1 ) β+A . ψ. φ.27) holds.35) to show that the subsequences of even and odd terms of the solution are bounded. (c) When (2. . A + Bxk k=1 (2. φ.27) holds all prime period-two solutions .

1). In other words I is an invariant interval for the diﬀerence equation xn+1 = f (xn . we obtain the following table of invariant intervals for Eq(2. (2. The following table gives the signs of fx and fy in all possible nondegenerate cases. then xn ∈ I for every n > N . (A + Bx + Cy)2 − + − − − − − − Using this table. 1. y) = N − Mx .8. Invariant intervals for Eq(2. xn−1 ). . .10) and are fx (x. 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 . Signs of derivatives fx and fy fx > 0 N N fy > 0 if x < M .1) the partial derivatives are given by (2. Invariant Intervals 41 2. y) = where L = βA − αB. 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.8 Invariant Intervals An invariant interval for Eq(2. . and N = γA − αC. xN ∈ I for some N ≥ 0. . n = 0.9) and (2. fx < 0 if y < − M fy < 0 fx < 0 and fy = 0 fx < 0 and fy < 0 fx < 0 and fy > 0 fx < 0 and fy > 0 fx < 0 N N fy > 0 if x > M . For Eq(2. y) is monotonic in its arguments. 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 . fy < 0 if x < M L + My (A + Bx + Cy)2 and fy (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.2.36) if xN −1 .36) are determined from the intervals where the function f (x.

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

(2. . Is it true that the linearized equation about a positive equilibrium point of Eq(2.9. = 1 + xn . . 1. 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. 4. ∞)] and let p ≥ 2 be an integer.1) with α.37) converges to a periodic solution with period p. ∞) × (0.7. . 6.9. . Is it true that p ∈ {2.9. and 6.2. . ∞) converges to a solution with period p .37) is periodic with period p ≥ 2.40) 1 + xn−1 . 6}? Open Problem 2. 4. C ∈ [0.1 Assume that f ∈ C 1 [(0.5. .7. (2. . B. . (0. . n = 0. Open Problems and Conjectures 43 2. 1. Find necessary and suﬃcient conditions so that every nonnegative solution of the diﬀerence equation (2.3 Let p ≥ 2 be an integer and assume that every positive solution of Eq(2.9. Open Problem 2. β. 5. . . ∞).9.7) that every positive solution of each of the following three equations xn+1 = 1 + xn+1 = xn+1 xn−1 .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.38) (2.39) (2. 1 + xn xn + 2xn−1 . every solution is periodic with period p ≥ 2. γ. 1. xn−1 ). if and only if every root of the characteristic equation is a pth root of unity. Open Problem 2. ∞)] is such that every positive solution of the equation xn+1 = f (xn .4 It is known (see Sections 2. In particular address the case p = 2. [0. n = 0.2 Let f ∈ C 1 [[0. ∞) × [0. . A. ∞). n = 0. xn n = 0.

9. Show that every positive solution of Eq(2.7.3 Assume α. and Invariant Intervals converges to a solution with (not necessarily prime) period-two: . B. γ. A. A. . .9. C ∈ (0. β.38) or Eq(2. C ∈ (0. β. β. C ∈ (0.41) is a period-two solution of Eq(2. and 10. Conjecture 2. . ∞) for which the solution {xn }∞ n=−1 converges to the period-two solution (2. γ. Determine the set of initial conditions x−1 and x0 for which the solution {xn }∞ n=−1 of Eq(2. . Show that the positive equilibrium is globally asymptotically stable. Periodicity. ∞).2 Assume α. A. φ. determine all initial conditions (x−1 .1) is bounded. if (2. (See Sections 2.5. 4.7.5 Assume α.1) has a positive prime period-two solution.) Conjecture 2. . ∞) × (0.2. ∞) and that Eq(2. . ψ. ψ. Conjecture 2.9.33) is bounded.39) or Eq(2. φ. B. γ. Conversely. A. B. Semicycles. B ∈ (0.1) is a saddle point.44 Chapter 2. determine φ and ψ in terms of the initial conditions x−1 and x0 . Open Problem 2.1 Assume α. (2. x0 ) ∈ (0. . γ. 6. ∞) and that Eq(2. . Show that the positive equilibrium of Eq(2. ∞) and γ > β + A. Local Stability. 6.40). β.9.41).1) has no period-two solution.41) In each case.7.

.7.8 Assume f ∈ C[(0. .2).9. Show that Eq(2. B. (0.9.9 Extend Theorem 2.1) cannot have a positive prime two cycle which attracts all positive non-equilibrium solutions.4 Assume that α. . f (xn ) . γ. Open Problem 2. Conjecture 2. 45 Obtain necessary and suﬃcient conditions in terms of the coeﬃcients so that every positive solution of Eq(2. .1 to third order diﬀerence equations xn+1 = α + βxn + γxn−1 + δxn−2 .6 Assume α. ∞). with nonnegative parameters and nonnegative initial conditions such that γ = β + δ + A. What additional conditions should the parameters satisfy? . xn n = 0.9.7 Assume f ∈ C[(0. (0. Obtain necessary and suﬃcient conditions on f for every solution of the diﬀerence equation xn+1 = to be bounded. A. ∞). A.2. . B. 1.9. (See Section 4. f (xn−1 ) . Open Problem 2. xn−1 n = 0. . .7). ∞)]. γ.9. . β. β. C ∈ [0. Obtain necessary and suﬃcient conditions on f for every positive solution of the equation xn+1 = 1 + to converge to a period-two solution. C ∈ [0. ∞). Open Problems and Conjectures Open Problem 2. ∞). . 1. Open Problem 2. A + Bxn−1 + Dxn−2 n = 0.1) is bounded.9. ∞)]. (See Section 5. 1.

xn−2 ). and Invariant Intervals Open Problem 2.1. . . But how about necessary and suﬃcient conditions for the equilibrium of Eq(2.10 Extend the linearized stability Theorem 1. . xn−2 . n = 0. or a saddle point. 1.9. Semicycles.1 to third order difference equations xn+1 = f (xn . Local Stability. . . . or nonhyperbolic? Extend these results to fourth order diﬀerence equation xn+1 = f (xn . xn−1 . .46 Chapter 2.42) For the equation λ3 + aλ2 + bλ + c = 0 one can see that all roots lie inside the unit disk |λ| < 1 if and only if |a + c| < 1 + b |a − 3c| < 3 − b b + c2 < 1 + ac . xn−1 . . n = 0. Periodicity.42) to be a repeller. . 1. (2. xn−3 ).

. 1. . . (3. (3. . .5) xn+1 = βxn . . . 1. 1)-Type Equations 3. 1. 1.3) xn+1 = βxn . 1)-type: xn+1 = n = 0. 1. (3. . B n = 0. . (3. . (3.2) xn+1 = α . .6) . . .1 Introduction α . . .Chapter 3 (1. . 1.4) xn+1 = β . (3. Cxn−1 n = 0. A Eq(1) contains the following nine equations of the (1. Bxn n = 0. A n = 0. . Cxn−1 47 n = 0. .1) xn+1 = α .

. the initial conditions are nonnegative. C n = 0. namely. and (3. .7) are linear diﬀerence equation. Eqs (3.7) xn+1 = and γxn−1 . Indeed the solution with initial conditions y−1 and y0 is the six-cycle: y−1 . (3.10) is periodic with period six. 1)-type Eq(3. . 1. y−1 y−1 y0 y0 It is interesting to note that except for the equilibrium solution y = 1. 1.10) has prime period six. 1. Eqs(3. . Of these nine equations. . 1)-Type Equations xn+1 = γxn−1 . . . (3. (3. .10) It is easy to see that every positive solution of Eq(3.2) is periodic with period two and every solution of Eq(3. 3. .1). . y0 1 1 y−1 . . 1.48 Chapter 3.5)..9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. Eqs(3.2 The Case α = γ = A = B = 0 : xn+1 = β yn C βxn Cxn−1 This is the (1. .6) and (3. y0 . A n = 0. yn−1 n = 0. . .4) and (3. . (3.8) xn+1 = γ . . .4) is periodic with period four. . Bxn n = 0. and the denominators are always positive.. . every other solution of Eq(3.8).6) which by the change of variables xn = reduces to the equation yn+1 = yn .. (3.9) are trivial. (1. The remaining two equations. are treated in the next two sections. Every solution of Eq(3.

Eq(3.13) n = 0. .12) . . n = 0. The general solution of Eq(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. while Eq(3.3.11) + C2 −1 − 2 √ 5 n . (0.2). from which the behavior of solutions is easily derived.3 The Case α = β = A = C = 0 : xn+1 = γ yn e B This is the (1. . . The Case α = β = A = C = 0 : xn+1 = γxn−1 Bxn 49 γxn−1 Bxn 3. . . . where f ∈ C[(0. Eq(3. n = 0. (3. ∞)]. 1)-type Eq(3. 3. 1. 1. 1.8) which by the change of variables xn = reduces to the linear equation yn+1 + yn − yn−1 = 0. Also every solution of Eq(3. namely two and six. . ∞)].11) is yn = C1 −1 + 2 √ 5 n n = 0. which possess the property that every positive nontrivial solution of each of these two equations is periodic with the same prime period. . .2) and (3. 1. (3. respectively.14) (3. xn−1 ).6) is of the form xn+1 = f (xn . .6). .3) also has the property that every nontrivial positive solution is periodic with prime period four but this equation is really a variant of Eq(3. (3. ∞). (3. (3. ∞) × (0.2) is of the form xn+1 = f (xn ).7) is periodic with period two but this is a linear equation. ∞).3.4 Open Problems and Conjectures Of the nine equations in this chapter we would like to single out the two nonlinear equations. . where f ∈ C[(0. (0.

5. In particular address the cases k = 4. with real initial conditions is periodic with period k.4. 6.13) holds. k = 5. 8.1 Let k ≥ 2 be a positive integer and assume that (3. xn−1 n = 0.4. Obtain necessary and suﬃcient conditions on f so that every positive solution of Eq(3. γ so that every solution of the equation xn+1 = α |xn | + βxn−1 + γ. .1 Let f ∈ C 1 [[0. . γ ∈ (−∞. [0. Show that f (x) = 1 + x. In particular address the cases: (See [10]. ∞)] be such that every positive nontrivial solution of the diﬀerence equation xn+1 = f (xn ) . (3. n = 0. 1)-Type Equations Open Problem 3.12) is periodic with period k. Open Problem 3. β.4. .4.50 Chapter 3. ∞).2 Let f ∈ C 1 [[0.15) holds. 6. 7. ∞). [0. 1.14) is periodic with period k.2 Let k ≥ 2 be a positive integer and assume that (3. Conjecture 3. and [59]. (1.16) is periodic with prime period six.4. 9. 1. ∞)] be such that every positive nontrivial solution of the diﬀerence equation (3. [17]. Show that f (x) = x.3 Let k be a positive integer and assume that α. . Open Problem 3.16) is periodic with prime period ﬁve. Obtain necessary and suﬃcient conditions on f so that every positive solution of Eq(3. β. ∞). . Obtain necessary and suﬃcient conditions on α.) Conjecture 3. . .

[0. . ∞). . (ii) {pn }∞ converges to ±∞.4.4 Obtain necessary and suﬃcient conditions on f ∈ C[[0.5 Let {pn }∞ be a sequence of nonzero real numbers.18) xn−1 ∞ (i) {pn }n=0 converges to a ﬁnite limit. n = 0. f (xn ) . (3. . ∞)] such that every positive nontrivial solution of the equation xn+1 = is periodic with prime period three. .3. n=0 (iii) {pn }∞ is periodic with prime period k ≥ 2. 1.4. .4. Open Problems and Conjectures 51 Open Problem 3. xn−1 n = 0. . (3.17) Open Problem 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 = . 1. .

.

. 1. (4.5) xn+1 = βxn . .2) xn+1 = α . . 2)-type: xn+1 = n = 0. (4. 1.1) xn+1 = α . .4) xn+1 = βxn . . (4. . . . . .1 Introduction α . . A + Cxn−1 n = 0. 1. . . A + Cxn−1 n = 0. . A + Bxn Eq(1) contains the following nine equations of the (1. Bxn + Cxn−1 53 n = 0.3) xn+1 = βxn . (4.6) .Chapter 4 (1. . . 1. (4. Bxn + Cxn−1 n = 0. 1. (4. A + Bxn n = 0. 1. . . 2)-Type Equations 4.

. and the denominators are always positive. Bxn + Cxn−1 n = 0. and (4. .2). from which the global behavior of solutions is easily derived. 1. 2)-Type Equations xn+1 = γxn−1 .4) are Riccati-type diﬀerence equations.1). (4. (4. .2). 1.2 The Case β = γ = C = 0 : xn+1 = A yn B α A+Bxn This is the (1. then the subsequences {x2n−1 } and {x2n } satisfy the Riccati equation of the form of Eq(4.8) γxn−1 . . A + Cxn−1 n = 0. 1. Also Eqs(4. (4. (1.8) will not be discussed any further in this chapter. (4.4) to the linear equation yn+1 = A B yn + . . . . Eqs(4. In view of the above.1) and (4.1) which by the change of variables xn = . .54 Chapter 4. . β β n = 0.7) xn+1 = and xn+1 = γxn−1 . . (See Section 1.9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive.6 in the Preliminary Results).8) are essentially Riccati equations. . . . . Similarly. 4. 2)-type Eq(4. yn+1 = A + Cyn It is also interesting to note that the change of variables xn = 1 yn reduces the Riccati equation (4. . the initial conditions are nonnegative. namely Eqs(4. 1. Indeed if {xn } is a solution of Eq(4. A + Bxn n = 0.2) and (4. then the subsequences {x2n−1 } and {x2n } satisfy the Riccati equation γyn n = 0. if {xn } is a solution of Eq(4. .4). 1. Two of these equations. .8).

2. Theorem 4.3) which by the change of variables √ α xn = yn is transformed to the diﬀerence equation yn+1 = C B + .2. . (See also [12].3. . . . 4.3 The Case β = γ = A = 0 : xn+1 = α Bxn +Cxn−1 This is the (1. It is easy to see that if a solution of Eq(4. 1. Proof.) We will ﬁrst show that every solution of Eq(4. (4. lim sup yn = ∞ and n→∞ lim inf yn = 0.11) (4.12) converges to √ its equilibrium B + C.6. The Case β = γ = A = 0 : xn+1 = reduces to the Riccati equation yn+1 = where α Bxn +Cxn−1 55 p . n→∞ . A2 The Riccati number associated with this equation (see Section 1. (4. 2)-type Eq(4.4. .12) is bounded and then we will use the method of “limiting solutions” to show that every solution of Eq(4. .6) is R = −p < 0 and so the following result is a consequence of Theorem 1. That is.10) is globally asymptotically stable. .12) is bounded and persists.3.10) αB . 1 + yn p= n = 0. Theorem 4.1 Every solution of Eq(4. 1. yn yn−1 n = 0.12) is bounded from above then it is also bounded from below and vice-versa.12) was investigated in [65].1 The positive equilibrium y= −1 + √ 2 1 + 4p of Eq(4.12) Eq(4. So if the theorem is false there should exist a ∞ solution {yn }n=−1 which is neither bounded from above nor from below.

2 √ B + C of Eq(4. B C B+C + > yj−1 yj−2 yi C B+C B + ≤ . 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. . The linearized equation of Eq(4. Theorem 4.3. 1. y is locally asymptotically stable for all positive values of B and C.1 .5) there exist two solutions {In }∞ n=−3 and {Sn }∞ n=−3 of the diﬀerence equation yn+1 = with the following properties: I0 = I. −2. . . {yn }n=−1 is bounded and persists. To ∞ this end. It remains to be shown that y is a global attractor of all solutions of Eq(4.1. .1. Then by Theorem 4.12). S0 = S B C + . B + C < yi yj ≤ B + C which is impossible. (1.12) is globally asymptotically √ B + C is (4. B+C B+C n = 0. . .12) about the equilibrium y = zn+1 = − B C zn − zn−1 . and S = lim sup yn . let {yn }∞ n=−1 be a solution of Eq(4. n→∞ for n ≥ −1. and by Theorem 1.3.56 Chapter 4. 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 . yn yn−1 n = −3. yi−1 yi−2 yj Proof. . . .12). . . j − 1}. By the theory of limiting solutions (see Section 1. .13) for all n ∈ {−1.2 The equilibrium y = stable.

C βxn A+Cxn−1 .14) from which it follows that (4. . 1.16) .Pielou’s Equation 57 In . 1 + yn−1 β . .14) and (4. and I= Hence Thus B C B+C B C + =S= = + S−1 S−2 I I I from which it follows that S−1 = S−2 = I.4. (4. Sn ∈ [I.4. S] for n ≥ −3. I−1 I−2 S B + C = SI.15) we conclude that S=I and the proof is complete.5) which by the change of variables xn = reduces to Pielou’s diﬀerence equation yn+1 = where p= pyn .4 The Case α = γ = B = 0 : xn+1 = Equation A yn .Pielou’s This is the (1. Therefore B C B C + = I = S−1 = + S S S−2 S−3 S = S−2 .15) 2 4. A n = 0. 2)-type Eq(4. The Case α = γ = B = 0 : xn+1 = and Then βxn A+Cxn−1 . From (4. . S= B C B+C + ≤ S−1 S−2 I C B+C B + ≥ . (4.

Then the zero equilibrium of Eq(4.58 Chapter 4.16) is globally asymptotically stable.16) that every positive solution converges to 0. 0 is locally asymptotically stable when p ≤ 1 and unstable when p > 1. Theorem 4. In fact by employing Theorem 1. it follows that when (4. (4. x is globally asymptotically stable. So the interesting case for Eq(4. which is locally asymptotically stable.79) as a discrete analogue of the delay logistic equation dN N (t − τ ) = rN (t) 1 − .17) In this case the zero equilibrium of Eq(4.1 (a) Assume p ≤ 1. p. (4. Then the positive equilibrium y = p − 1 of Eq(4. 2)-type Eq(4. 4.5 The Case α = γ = A = 0 : xn+1 = β . .16) is unstable and Eq(4. 1. Furthermore. . We summarize the above discussion in the following theorem.6) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = p + yn .) When p ≤ 1.4.18) .16) was investigated in [55]. p + yn−1 n = 0.16) possesses the unique positive equilibrium y = p − 1. (See also ([42].16) is when p > 1.16) is globally asymptotically stable.22) and ([67].4. (b) Assume y0 ∈ (0. t≥0 dt P which is a prototype of modelling the dynamics of single-species. p.75). (1. . it follows from Eq(4. Eq(4. ∞) and p > 1.2. 2)-Type Equations This equation was proposed by Pielou in her books ([66]. p.17) holds. B + Cyn βxn Bxn +Cxn−1 This is the (1.

The following local result is a straightforward application of the linearized stability Theorem 1. The Case α = β = C = 0 : xn+1 = where γxn−1 A+Bxn 59 B . Then 0 and y = 1 − p are the only equilibrium points of Eq(4.4.19) was investigated in [28]. C Eq(4. p + yn n = 0. 2)-type Eq(4.7.6.18) was investigated in ([42] Corollary 3. It follows directly from Eq(4.6 The Case α = β = C = 0 : xn+1 = γ yn . . That is. p. ∞). . B γxn−1 A+Bxn This is the (1. γ Eq(4.1 (e). (4. Lemma 4.19) has semicycles of length one. In fact.19) and it is locally asymptotically stable. The oscillatory character of the solutions of Eq(4.1. 73) where it was shown that its equilibrium one is globally asymptotically stable . .7) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= yn−1 .19) and they are both unstable.1.19) is a consequence of Theorem 1. (b) Assume p < 1. every solution of Eq(4. 0 is a repeller and y = 1 − p is a saddle point equilibrium.4. Then 0 is the only equilibrium point of Eq(4.1 (a) Assume p > 1.1. except possibly for the ﬁrst semicycle. 1.5) that 1 yn+1 < yn−1 p and so when p>1 for n ≥ 0 . p= 4.19) (see also Section 2.19) A ∈ (0.6.

B This equation was investigated in [7] where it was shown that the following statements are true: p= . . . . 2)-Type Equations the zero equilibrium of Eq(4.9) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = p + where yn−1 . . . . φ. 4.19) converges to a period-two solution . 1. 1.7 The Case α = β = A = 0 : xn+1 = γ Byn γxn−1 Bxn +Cxn−1 This is the (1. It follows from the identity (1 − p) − yn yn−1 . .19) is globally asymptotically stable. . (1.60 Chapter 4. (4. On the other hand when p = 1. . 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. ψ. yn+1 < yn−1 for n ≥ 0 and so every positive solution of Eq(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. . ψ. . φ.20) C ∈ (0. n = 0. Furthermore we can see that φψ = 0 but whether there exists solutions with φ=ψ=0 remains an open question. . yn n = 0. . 2)-type Eq(4.

(ii) When p = 1.21) As a simple consequence of the linearized stability Theorem 1. n→∞ lim 2.20) such that 0 < 1 y−1 ≤ 1 and y0 ≥ 1−p . In fact the following result is true. p+1 p+1 n = 0. Then {yn }∞ n=−1 converges monotonically to y = p + 1. Theorem 4. Theorem 4. every solution of Eq(4. 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. and so y0 > p + 1.1 we obtain the following result: Theorem 4.1 The Case p < 1 Here we show that there exist solutions of Eq(4. Clearly the only equilibrium point of Eq(4. ∞ (b) Let {yn }n=−1 be a solution of Eq(4. Then {yn }∞ n=−1 is oscillatory. (iii) When p > 1. The next result about semicycles is an immediate consequence of some straightforward arguments and Theorem 1.20) is y = p + 1. .2 (a) Let {yn }∞ n=−1 be a solution of Eq(4. . the equilibrium y = p + 1 of Eq(4. every semicycle has length one. with the possible exception of the ﬁrst semicycle. The linearized equation of Eq(4.20) converges to a period-two solution.7.1. and let {yn }∞ n=−1 be a solution of Eq(4.7. It suﬃces to show that y1 ∈ (p. . Moreover. 1.20) which are unbounded.20) is globally asymptotically stable.20) is locally asymptotically stable when p > 1 and is an unstable saddle point when p < 1.7. . lim y2n = ∞. n→∞ y2n+1 = p. . (4. We now proceed to establish the above results. 4.7.7. 1] and y2 ≥ p + y0 and use induction to complete the proof.20) which consists of at least two semicycles. Note that 1 1−p > p + 1.7.1.1 The equilibrium point y = p + 1 of Eq(4.3 Let p < 1.20) about the equilibrium point y = p + 1 is zn+1 + 1 1 zn − zn−1 = 0.20) which consists of a single semicycle. Proof. Then the following statements are true: 1.4.20) to be bounded.

and that p < y0 < p + 1 < y−1 . that p < yn for all n ≥ −1.7. Then the following statements are true.2 that we may assume that every semicycle of {yn }∞ n=−1 has length one. The following result will be useful. n→∞ p p−1 n→∞ Proof. ∞ (b) Suppose {yn }∞ n=−1 consists of at least two semicycles. 4. It follows from Theorem 4.3 The Case p > 1 Here we show that the equilibrium point y = p + 1 of Eq(4. ∞ (a) Suppose {yn }n=−1 consists of a single semicycle. Then p+ p2 p−1 ≤ lim inf yn ≤ lim sup yn ≤ . and let {yn }∞ n=−1 be a solution of Eq(4.5. . Hence y2 = p + y0 y1 ≥ p + y0 .7. Theorem 4. 2)-Type Equations > p. Then {yn }n=−1 converges to a prime period-two solution of Eq(4. Lemma 4. y1 = p + y−1 1 ≤p+ ≤ 1.4 Let p = 1. and let {yn }∞ n=−1 be a solution of Eq(4. y2n+1 < p + So as every solution of the diﬀerence equation 1 ym+1 = p + ym .2 The Case p = 1 The following result follows from Lemma 2. . Note that for n ≥ 0.20). Also. 1.7. p m = 0. See also Section 2. 4. y2n−1 .20) is globally asymptotically stable. Then {yn }∞ n=−1 converges monotonically to y = 2. . p .7. 1].1 Let p > 1.7. (1.20) converges to a period-two solution if and only if p = 1.20). (c) Every solution of Eq(4. y0 y0 2 and so y1 ∈ (p.1 and some straightforward arguments.20).62 Indeed y1 = p + y1 y0 Chapter 4.7. p2 We shall ﬁrst show that lim supn→∞ yn ≤ p−1 .

7. The Case α = β = A = 0 : xn+1 = 2 γxn−1 Bxn +Cxn−1 2 63 p p converges to p−1 . and set a = p and b = Note that f and p2 + .5 Let p > 1. ∞).7.7. it follows that lim supn→∞ yn ≤ p−1 . we have p−1 p3 + p(p − 1) =p+ . there exist no solutions of Eq(4.7. Then > 0. set p−1 p2 + >p . p−1 For x. Let p N ≥ 0 such that for all n ≥ N . ∞) × (0. Theorem 4. ∞). y ∈ p. So let {yn }∞ n=−1 be a solution of Eq(4. Then the equilibrium y = p + 1 is globally asymptotically stable. Proof.20) with prime period two. y2n−1 p + p2 + is arbitrary. (0. There clearly exists y2n = p + So as p−1 p3 + p + p(p − 1) y2n−2 >p+p 2 = .1 that y = p + 1 is a locally asymptotically stable equilibrium point of Eq(4. Hence p < f (x. and f is increasing in y ∈ (0.p = p + p 2 p−1 p + p2 + p−1 p2 + p−1 = p2 + p3 + p = . . y) = p + . ∞) for each x ∈ (0. n→∞ y f (x. p2 + y2n−1 < . ∞). Let > 0. It suﬃces to show that lim yn = p + 1.20). Recall that by Theorem 4. p2 − p p−1 p2 + .4. x Then f ∈ C[(0. ∞)] . lim inf yn ≥ 2 n→∞ p p 2 We are now ready for the following result.4. ∞). y) < for all x. f is decreasing in x ∈ (0. We shall next show that p + p−1 ≤ lim inf n→∞ yn . We know by Theorem 4. p−1 f p. ∞) for each y ∈ (0. p−1 Let n ≥ N . y ∈ (0.20).

. φ1 .8 Open Problems and Conjectures xn+1 = A xn−k + B . A+B A+B A. . .22). . .23) is well deﬁned for all n ≥ 0. [60]. + yn yn−1 (4. . (a) Find all initial conditions y−1 .1. xn−l n = 0.1 Let .6. B ∈ (0. . . . φ2 . ∞) and k. .1 Consider the diﬀerence equation where and where the initial conditions x−l . .} with k < l.2 (see [18]) Assume that A. n=−l Open Problem 4. p<p+ Chapter 4.8. . (1. B ∈ (0. either to the equilibrium or to a periodic solution with period p ≥ 2 and that what happens and the exact value of p are uniquely determined from the characteristic roots of the linearized equation λl+1 + (See [19]. φp . l ∈ {0. . [23]. y0 with y−1 y0 < 0 for which the equation yn+1 = B A . (b) Assume that the initial conditions y−1 . Is it bounded? Does limn→∞ yn exist? When does {yn } converge to a two cycle? . ∞).4. Show that every solution of Eq(4. . . .8. and [61].8. 2 4. x0 such that {xn }∞ converges to this periodic solution. [25]. Investigate the global behavior of the solution {yn }. 2)-Type Equations and so by Theorem 1.22) converges. . . (4. y0 are such that y−1 y0 < 0 and such that Eq(4. .) Open Problem 4. p−1 p2 + p2 ≤ lim inf yn ≤ lim sup yn ≤ < n→∞ p p−1 p−1 n→∞ n→∞ lim yn = p + 1. . .23) A B λl−k + = 0. . . x0 are arbitrary positive real numbers.64 Finally note that by Lemma 4.22) Conjecture 4. Determine the set of all positive initial conditions x−l . . 1. be a given prime period p solution of Eq(4. is well deﬁned for all n ≥ 0. 1.7.

20) possesses a solution {yn }∞ n=−1 which remains above the equilibrium for all n ≥ −1. Determine the set of initial conditions y−1 . x0 ) ∈ R × R through which the equation xn−1 xn+1 = −1 + (4.8. Open Problem 4.8.8. y0 ∈ R be such that Eq(4. Show that Eq(4. for these initial points.8. the asymptotic behavior.20).20) is well deﬁned for all n ≥ 0.20) is well deﬁned for all n ≥ 0. Open Problems and Conjectures 65 Conjecture 4.5 (a) Assume p ∈ (0.19) has a solution which converges to zero. investigate the global character ∞ of {yn }n=−1 . Conjecture 4.24) is well deﬁned for all n ≥ 0 and bounded.24) xn is well deﬁned for all n ≥ 0. . ∞).8.2 Assume that p = 1.6 (See [13]) Find the set G of all initial points (x−1 .3 Assume that p < 1.4. Open Problem 4. (b) Let y−1 .4 Determine the set G of all initial conditions (y−1 .4 Assume that (x−1 . y0 ) ∈ R × R through which the equation yn−1 yn+1 = 1 + yn is well deﬁned for all n ≥ 0 and.3 Show that Eq(4.19) is bounded. and the periodic ∞ nature of the solution {yn }n=−1 of Eq(4. Open Problem 4. Show that {xn }∞ n=−1 is a three cycle. y0 ∈ (0.8. Open Problem 4.8.8. x0 ) ∈ R × R is such that the equation (4. Conjecture 4. For such an initial point. Determine all initial conditions y−1 . investigate the boundedness. y0 ∈ R such that the equation (4. ∞) ∞ for which the solution {yn }n=−1 of Eq(4. (c) Discuss (a) and (b) when p < 0.

f (xn ) .8. 1. ∞) does the sequence deﬁned by xn−1 . ∞). [66]. . 1 + xn−k n = 0. . Conjecture 4. . Obtain necessary and suﬃcient conditions on f for every positive solution of the equation xn+1 = to be bounded. . ai ∈ (0. . . with positive initial conditions. 1.) Assume α ∈ (1. 1. 1. + ak xn−k n = 0.}. . 2)-Type Equations Open Problem 4. 1. . Open Problem 4. see [55]. . a + a0 xn + . 1.5 (See [27]) Assume r ∈ (0. ∞)]. . . (1. . and [67]. Show that the positive equilibrium of Pielou’s discrete logistic model xn+1 = αxn . . < 2 cos α 2k + 1 Open Problem 4. . . ∞) and k ∈ {0. n = 0. ∞) for i = 0. . . . . xn + xn−1 n = 0. n = 0.6 (Pielou’s Discrete Logistic Model. Show that every positive solution of the population model Jn+1 = Jn−1 er−(Jn +Jn−1 ) .66 Chapter 4. . k. converges to a period-two solution. . ∞).8 Let a.9 (See [5]) For which initial values x−1 .8. xn+1 = 1 + xn converges to two? . . x0 ∈ (0.8. .8.7 Let f ∈ C[(0.8. Conjecture 4. with positive initial conditions is globally asymptotically stable if and only if kπ α−1 . Investigate the global asymptotic stability of the equilibrium points of the diﬀerence equation xn+1 = xn . . (0.

20) are bounded. . y0 ∈ (0. 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 + . y0 ∈ [0.7) Open Problem 4. n = 0.10 (See [31]) Consider the diﬀerence equation yn+1 = yn−1 e−yn .26) (4. lim (a) Find all initial points y−1 . . n = 0. ∞) through which the solutions of Eq(4. In other words. lim p = n→∞ pn . 1. . 1. (b) Determine the limits of the subsequences of even and odd terms of every solution of Eq(4. (4. ∞ One can easily see that every solution {yn }n=−1 of Eq(4.25) with nonnegative initial conditions. . . Investigate the asymptotic behavior of {yn }n=−1 . . . in terms of the initial conditions y−1 and y0 of the solution.8. each of the subsequences {y2n }∞ and {y2n+1 }n=−1 is decreasing and n=0 lim [n→∞ y2n ][n→∞ y2n+1 ] = 0.25) converges to a period-two solution . 0. Open Problem 4. 67 (4. 1. Open Problems and Conjectures (See also Section 4.25). ﬁnd the basin of attraction of the equilibrium of Eq(4. Open Problem 4.4. .8. 1). . yn yn−1 pn yn . . . ∞) such that the solutions {yn }∞ n=−1 of Eq(4. . ∞ (b) Let y−1 . y0 ∈ B.11 Assume p ∈ (0. . .8. . 1 + yn−1 n = 0. .8. ∞ More precisely.12 Let {pn }∞ be a convergent sequence of nonnegative real numn=0 bers with a ﬁnite limit.25). .25) converge to zero.27) . (a) Find the set B of all initial conditions y−1 . 0. .

. . 2)-Type Equations yn+1 = n = 0. investigate the long term behavior of the solution {yn }∞ n=−1 : yn+1 = yn .14 For each of the following diﬀerence equations determine the “good” set G ⊂ R × R of all initial conditions (y−1 . (1. Open Problem 4.30) yn+1 = yn+1 = pn + Open Problem 4.26) . . 1. . 1 − yn−1 yn−1 . pn + yn yn−1 .29) (4. 1 − yn (4. (4. y0 ) ∈ R × R through which the equation is well deﬁned for all n ≥ 0. . . 1 − yn−1 1 + yn . y0 ) ∈ G. . yn For those equations from the above list for which you were successful. 1.13 Let {pn }∞ be a periodic sequence of nonnegative real numbers n=0 with period k ≥ 2. 1.30).32) (4.(4.68 pn + yn . Then for every (y−1 . n = 0.8.34) yn+1 = yn+1 = −1 + yn−1 . n = 0. yn Chapter 4. . Investigate the global character of all positive solutions of each of Eqs(4.8. . . pn + yn−1 yn−1 .28) (4.33) (4.31) yn+1 = (4. extend your result by introducing arbitrary real parameters in the equation. .

. (5. . . (5. Bxn n = 0. . .3) xn+1 = α + γxn−1 . (5. . (5. 1. 1)-Type Equations 5. . 1. . . (5. 1.5) xn+1 = α + γxn−1 . .Chapter 5 (2. 1)-type: xn+1 = n = 0.6) . A n = 0. . Bxn n = 0. . . . 1. . (5.1 Introduction α + βxn .4) xn+1 = α + γxn−1 .2) xn+1 = α + βxn . . Cxn−1 n = 0. Cxn−1 69 n = 0. . . 1. 1. A Eq(1) contains the following nine equations of the (2.1) xn+1 = α + βxn .

3) which by the change of variables reduces to the equation yn+1 = n = 0.20) with p = β . 1.9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive.7) xn+1 = and xn+1 = βxn + γxn−1 . 1)-type Eq(5. yn−1 α+βxn Cxn−1 . Therefore there remain Eqs(5.2). (5. (2. .7) are linear. 1. (5. . 1)-Type Equations xn+1 = βxn + γxn−1 . A n = 0. (5.70 Chapter 5. The change of variables γ xn = yn . γ Finally the change of variables xn = γ β + yn .2 The Case γ = A = B = 0 : xn+1 = Equation p + yn .3) and (5. Eq(5. (5. which was investigated in Section 4. 5. . . .18) with p = β . which will be investigated in the next two sections.1). Eqs(5. . . B reduces Eq(5. Bxn n = 0. which was investigated in Section 4. .7.8) βxn + γxn−1 . . 1. (5.4) and (5. .Lyness’ This is the (2.6) is essentially like Eq(5. 1.8) to Eq(4. . . .9) to Eq(4.5).2) is a Riccati equation and Eq(5. the initial conditions are nonnegative. Cxn−1 n = 0. Of these nine equations. and the denominators are always positive. C C γ reduces Eq(5.10) .5.

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

13) was investigated in [28]. γ2 Eq(5. By the linearized stability Theorem 1.3 The Case β = A = C = 0 : xn+1 = γ yn B n = 0.1.13) converges monotonically to the equilibrium y.13) is a consequence of the identity yn+1 − yn−1 = yn−1 − yn−2 yn for n ≥ 1.1 Show that every solution of Lyness’ 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 ∞. 1)-Type Equations 5.13) one can easily see that every nonoscillatory solution of Eq(5. Open Problem 5. every oscillatory solution of Eq(5.10) is bounded from above and from below by positive numbers.13) αB .12).1 one can see that except possibly for the ﬁrst semicycle. α+γxn−1 Bxn This is the (2. 1)-type Eq(5. 5. From this it follows that a solution of Eq(5. To this day we have not found a proof for the boundedness of solutions of Eq(5.12).13) is a saddle point. yn (5. Concerning the nonoscillatory solutions of Eq(5.7.1 one can see that the unique equilibrium point √ 1 + 1 + 4p y= 2 of Eq(5.72 Chapter 5.4.5) which by the change of variables xn = reduces to the equation yn+1 = where p= p + yn−1 . . 1.4 Open Problems and Conjectures By using the invariant (5. . . (2. The global character of solutions of Eq(5. one can easily see that every solution of Lyness’ Eq(5. .13) has semicycles of length one.10) without using essentially the invariant. By Theorem 1.10) is bounded without using the invariant (5.

∞).5. When we allow the initial conditions to be real numbers the equation yn+1 = p + yn yn−1 (5.4.4. Conjecture 5. f (xn ) . (0.4. Show that the subset of G through which the solutions of Eq(5. Open Problems and Conjectures 73 Open Problem 5.2 Let G be the set of initial conditions deﬁned in the Open problem 5. we oﬀer the following open problems and conjectures. ∞).4. Determine the set G of all initial conditions y−1 . ∞)]. . xn−1 n = 0. . Open Problem 5. (5. (0.3 Assume that f ∈ C[(0. ∞)] and that the diﬀerence equation (5. y0 ∈ (−∞. 1. Obtain necessary and suﬃcient conditions in terms of f so that every positive solution of the diﬀerence equation xn+1 = is bounded.14) possesses a (nontrivial) invariant.4. (0.4 (See [26]. ∞). ∞) through which Eq(5.14) about x does not have both eigenvalues in |λ| < 1 and does not have both eigenvalues in |λ| > 1. .1 Assume that f ∈ C[(0. To illustrate. ∞)]. Obtain necessary and suﬃcient conditions in terms of f so that the diﬀerence equation (5.) Assume p ∈ (−∞.15) is well deﬁned for all n ≥ 0.15) are unbounded is a set of measure zero.4. ∞).14) possesses a unique positive equilibrium point x and a nontrivial invariant.4.14) Open Problem 5. . What is it that makes Lyness’ equation possess an invariant? What type of diﬀerence equations possess an invariant? Along these lines.4. Conjecture 5. we oﬀer the following open problems and conjectures.2 Assume that f ∈ C[(0. Show that the linearized equation of Eq(5.15) may not even be well deﬁned for all n ≥ 0.

1. Conjecture 5.4. . Show that the set of points (y−1 . Determine the periodic character and the asymptotic behavior of all solutions with (y−1 . (2.4. + yn−k .4. . yn n = 0. . 1.6 Show that the equation yn+1 = 1 + yn−1 . .74 Chapter 5. Open Problem 5. y0 ) ∈ R × R through which the equation 1 + yn−1 yn+1 = yn is well deﬁned for all n ≥ 0. yn−k−1 n = 0. . . .15) is not well deﬁned for all n ≥ 0.3 Assume p ∈ (−∞. converges to a period-three solution. Conjecture 5.6 Find the set G of all initial points (y−1 . . is a set of points without interior. y0 ) ∈ (−∞.4. y0 ) ∈ G. . .5 (See [20]) Show that every positive solution of the equation yn+1 = 1 1 . + yn yn−1 yn−3 yn−4 n = 0. Open Problem 5. Conjecture 5. ∞) through which Eq(5.4. . 1. . yn yn−4 + . has a nontrivial positive solution which decreases monotonically to the equilibrium of the equation. . 1 + . 1.4 Show that every positive solution of the equation yn+1 = converges to a period-six solution. ∞). . 1)-Type Equations Conjecture 5. ∞) × (−∞. .4.5 Determine all positive integers k with the property that every positive solution of the equation yn+1 = is periodic. yn−1 yn−3 n = 0.

Open Problems and Conjectures 75 Conjecture 5.16) to where xn = −1+yn 2 if A > 1 if A = 1 if A < 1 n = 0. xn xn−1 n = 0. 1.8 (The Gingerbreadman Map. . 2 αxn . . . . .4. 1. Eq(5.16) A A 1+yn 2 reduces Eq(5. . ∞).4. Show that every positive solution of May’s Host Parasitoid Model xn+1 = is bounded. 1 if A < 1 if A = 1 δ= 0 −1 if A > 1. Show that every positive solution of the equation xn+1 = is bounded. .17) is called the Gingerbreadman Map and was investigated by Devaney [17]. (5. Note that the change of variables n e y2 max{x2 . When δ = 1.17) yn+1 = |yn | − yn−1 + δ. 1. .5. (1 + xn )xn−1 n = 0. see [59]) Assume α > 1. . Conjecture 5.7 (May’s Host Parasitoid Model. see [59]) Let A ∈ (0.4. . A} n . (5.

.

A + Cxn−1 n = 0. . 1.2) xn+1 = α + βxn . 1. . (6. 1. . . . 2)-type: xn+1 = n = 0.4) xn+1 = α + γxn−1 . . A + Bxn Eq(1) contains the following nine equations of the (2. (6. . (6. .Chapter 6 (2. (6.3) xn+1 = α + γxn−1 . 1. A + Cxn−1 n = 0.1 Introduction α + βxn . 2)-Type Equations 6. (6. 1.5) xn+1 = α + γxn−1 . Bxn + Cxn−1 n = 0. .1) xn+1 = α + βxn . . . . . . . A + Bxn n = 0. 1. (6.6) . Bxn + Cxn−1 77 n = 0. . . .

. Eq(6. . A + Cxn−1 n = 0. .1) which is in fact a Riccati equation. In fact if {xn }∞ n=−1 is a positive solution of ∞ ∞ Eq(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. .9) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive.78 Chapter 6. α + γzn . . 6. . 2)-Type Equations xn+1 = βxn + γxn−1 . A + Czn n = 0. 0. .8) xn+1 = βxn + γxn−1 . See Section 1. Bxn + Cxn−1 n = 0. . Therefore we will omit any further discussion of Eq(6. (6. 2)-type Eq(6.2 The Case γ = C = 0 : Equation xn+1 = α+βxn A+Bxn . . A + Czn for n ≥ 0 n = −1. for n ≥ −1.5). . . 1. 1. . α + γzn . . 1.Riccati This is the (2.6.7) xn+1 = βxn + γxn−1 . the initial conditions are nonnegative. . (2. . (6. . .5) is essentially similar to Eq(6. A + Bxn n = 0. (6. 1. and the denominators are always positive.1). To avoid a degenerate situation we will assume that αB − βA = 0.

3.1 we obtain the following result. 1 + yn−1 n = 0.2: Theorem 6. To this day. The main results known about Eq(6.3 The Case γ = B = 0 : xn+1 = A yn C α+βxn A+Cxn−1 This is the (2. .1 The equilibrium y of Eq(6. 2)-type Eq(6. the conjecture has not been proven or refuted.2. .10) αC β and q = .1) is globally asymptotically stable.10) is locally asymptotically stable for all values of the parameters p and q. By applying the linearized stability Theorem 1.) Eq(6. Eq(6. Lemma 6.10) are the following: Theorem 6. 1.10) is bounded from above and from below by positive constants.1 The positive equilibrium of Eq(6.3. . See also [36].6.10) was investigated in [42] and [43]. (6. The Case γ = B = 0 : xn+1 = α+βxn A+Cxn−1 79 The Riccati number associated with this equation is R= and clearly βA − αB (β + A)2 1 R< . .2) which by the change of variables xn = reduces to the equation yn+1 = where p + qyn . 2 A A (Eq(6.1.10) has the unique positive equilibrium y given by p= y= q−1+ (q − 1)2 + 4p 2 . 6. 4 Now the following result is a consequence of Theorem 1.1 Every solution of Eq(6.10) is a very simple looking equation for which it has long been conjectured that its equilibrium is globally asymptotically stable.6.3.

10) we see that yn+1 < qyn + p for n ≥ 0 and so k→∞ lim ynk = lim ynk −1 = ∞.10) is described by the following result: ∞ Theorem 6. has at least two terms.2 Let {yn }n=−1 be a nontrivial solution of Eq(6.10). Let {yn } be a solution of Eq(6. k→∞ Hence. then p yn+1 ≥ 1+M and so it is also bounded from below.80 Chapter 6. (2. 0 ≤ y1+nk − ynk = p + [(q − 1) − ynk −1 ]ynk <0 1 + ynk −1 2 which is a contradiction and the proof is complete. (ii) The extreme in each semicycle occurs at either the ﬁrst term or the second. for suﬃciently large k. Then. there is at most one term which follows the extreme. Furthermore after the ﬁrst. their extrema cannot be consecutive terms. if the solution is bounded from above by a constant M . (iv) Assume q ≤ p. and y1+nk = max{yn : n ≤ nk } for k ≥ 0. We present the proofs for positive semicycles only. Proof. From (6.10). every semicycle contains two or three terms. except perhaps for the ﬁrst one. 2)-Type Equations Proof. Furthermore.3. except possibly for the ﬁrst semicycle. in every semicycle. the remaining terms in a positive semicycle are strictly decreasing and in a negative semicycle are strictly increasing. Then the following statements are true: (i) Every semicycle. k→∞ lim y1+nk = ∞. 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. Then there exists a subsequence {y1+nk }∞ such k=0 that k→∞ lim nk = ∞. Clearly. The oscillatory character of solutions for Eq(6. (iii) In any two consecutive semicycles. .

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 = 1.10) 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). and yN +2 are all in a positive semicycle. α+βxn A+Cxn−1 81 yN −1 < y Then yN +1 = and yN ≥ y. Then yN ≥ y.6. Then yN +3 = p + qyN +2 p + qyN +1 p + qy = y. 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. 1 + yN −1 1 + yN (iv) Assume that for some N ≥ 0. 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 ﬁrst two terms in a positive semicycle are yN and yN +1 . Then yN +1 = p + qyN p + qyN +1 < = yN +2 .3. the terms yN . The Case γ = B = 0 : xn+1 = (i) Assume that for some N ≥ 0. yN +1 . .3 The positive equilibrium of Eq(6.3. p + qyN p + qy = y. > 1 + yN −1 1+y (ii) Assume that for some N ≥ 0. < < 1 + yN +1 1 + yN +1 1+y 2 Theorem 6.

10) yields i≥ Hence and so because q < 1. B (6.4.3) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= yn + p . . 1+i from which the result follows. For the proof when see ([42].3. it remains to show that every solution {yn }∞ of Eq(6.3.2.1. .11) and q = Eq(6. . 71). 2)-Type Equations (i) In view of Lemma 6. p + (q − 1)i ≤ is ≤ p + (q − 1)s i≥s p + qi 1+s and s ≤ p + qs . yn + qyn−1 αB β2 n = 0. .1. The following result is a consequence of the linearized stability Theorem 1. This result was ﬁrst established in [43]. 1. Then Eq(6. Chapter 6. To this end. p. 2)-type Eq(6.3 (f).11) was investigated in [15].1. (2.82 Proof. Theorem 3.10) n=−1 tends to y as n → ∞. let i = lim inf yn n→∞ and s = lim sup yn n→∞ which by Theorem 6.1 exist and are positive numbers.4 The Case γ = A = 0 : xn+1 = β yn B α+βxn Bxn +Cxn−1 This is the (2. 1 ≤ q ≤ p ≤ 2(q + 1) 2 6. C . (ii) The proof when p<q follows by applying Theorem 1.4.

yk−1 . q for all n > k. two consecutive values yk−1 and yk of a solution lie in I. the q function u+p f (u. Then yn ∈ 1. Lemma 6.11) is locally asymptotically stable for all values of the parameters p and q. v ∈ [ p . (a) The basic ingredient behind the proof is the fact that when u.11).4. ∞). Then the following statements are true: (a) Assume p < q and suppose that for some k ≥ 0. p q p .4.6.1 Let {yn }∞ n=−1 be a solution of Eq(6.1 The unique positive equilibrium point y= 1+ 1 + 4p(q + 1) 2(q + 1) of Eq(6. for some k ≥ 0. (b) The proof is similar and will be omitted.1 Invariant Intervals The following result establishes that the interval I with end points 1 and p is an invariq ant interval for Eq(6.4. . yk ∈ Then yn ∈ p . yk ∈ 1.4. 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. (b) Assume p > q and suppose that for some k ≥ 0. v) = u + qv is increasing in u and decreasing in v. 6.11) in the sense that if. Proof. then ym ∈ I for all m > k. yk + p yk + p yk+1 = ≤ = 1. yk−1 .1 .1 q p . Indeed. The Case γ = A = 0 : xn+1 = α+βxn Bxn +Cxn−1 83 Theorem 6. q for all n > k.

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)

.

. . 1].5 applies and so every solution of Eq(6. we can see that if a solution is not eventually in [1.14) to this period four solution we obtain a contradiction and so every solution of Eq(6. Now. p ] converges to y. 2.11) converges to y. v) = u + qv is decreasing in both variables and so it follows by applying Theorem 1. By applying (6. L2 . L4 . First. L1 .22) is a periodic solution of Eq(6. q Hence for every solution {yn }∞ n=−1 of Eq(6.22) we see that Li = 1 for i = 1.11) must eventually lie in the interval [ p . (2. .2 that [1. 2)-Type Equations is either identically equal to one or else it is strictly monotonically convergent. .7 that every solution of Eq(6. By applying (6. we assume that p > q. it converges to a periodic solution with period-four which q is not the equilibrium. Set Li = n→∞ y4n+i lim Thus clearly . 3. .11) must lie eventually in [1.4. lim yn = 1. (6.11) n→∞ lim yn = y and the proof is complete. 2. Under the assumption that q ≤ 1 + 4p. 4. .19) to the solution (6. Recall from Lemma 6. 4 and so n→∞ for i = 1. Case 2 Assume p = q. p ]. Theorem 1. p ]. v) = u+p u + qv is increasing in u and decreasing in v. 2 .4. 3. As in Case 2 we can see that every solution of Eq(6. By q using an argument similar to that in the case p = q.88 Chapter 6.4. L3 . In this interval the q function u+p f (u. (b) The proof follows from the discussion in part (a). p ] is an invariant interval. assume that p < q.11) with period four.11) with two consecutive values in [1. In this interval the function q f (u.

) The only equilibrium point of Eq(6.25) (b) Assume (6. φ.1 we obtain the following result: Theorem 6.23) is locally asymptotically stable when q < 1. if and only if q = 1.25) holds.24) are given by {φ. . ψ.1 The equilibrium y of Eq(6. . The Case β = C = 0 : xn+1 = α+γxn−1 A+Bxn 89 α+γxn−1 A+Bxn 6.23) converges to a period two solution. the following result is a consequence of the results in Section 2.25) holds. (6. (c) Assume that (6. .1.23) has a prime period-two solution .24) . φ. γ .23) was investigated in [28]. . A (6.23) and q = (Eq(6. 1 + yn αB A2 n = 0. ψ ∈ (0.6. ∞) : φψ = p}.5. Concerning prime period-two solutions. . (6. Theorem 6.2 (a) Eq(6. Then every solution of Eq(6.5.5 The Case β = C = 0 : xn+1 = A yn B This is the (2.5. . Then the values φ and ψ of all prime period-two solutions (6. ψ. 1. .7. 2)-type Eq(6. and is an unstable saddle point when q > 1.4) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + qyn−1 . . .23) is y= q−1+ (q − 1)2 + 4p 2 and by applying the linearized stability Theorem 1.

5. yn0 > yn0 −1 .23) is globally asymptotically stable. (6.2 The Case q < 1 The main result in this section is the following: Theorem 6. 1−q n ≥ 0. yn0 ) < f (y. It is suﬃcient to show that {yn } is a decreasing sequence for n ≥ K. We will assume that there exists a positive integer K such that yn−1 ≥ y for every n ≥ K.23) are as follows: Theorem 6. (2. Proof.4 Assume that q < 1. (b) Assume p ≥ q.23).5. yn0 −1 ) < f (yn0 . Then yn0 +1 = which is impossible.1 (b) Let {yn } be a nonoscillatory solution of Eq(6. 2)-Type Equations 6. Proof.23) oscillates about the equilibrium y with semicycles of length one.90 Chapter 6. Clearly yn+1 = Set M= p + qyn−1 < p + qyn−1 . Then every nonoscillatory solution of Eq(6.23) and let y denote the unique positive equilibrium of Eq(6.3 Let {yn } be a nontrivial solution of Eq(6.7.26) .5.23). Then the positive equilibrium of Eq(6. Then the following statements are true: (a) After the ﬁrst semicycle.1 Semicycle Analysis Our results on the semicycles of solutions for Eq(6. Clearly. an oscillatory solution {yn } of Eq(6. y) = y. 1 + yn p + . p + qyn0 −1 = f (yn0 .5. (a) This follows immediately from Theorem 1. So assume for the sake of contradiction that for some n0 ≥ K.23) converges monotonically to the equilibrium y. the condition p ≥ q implies that the function f (x) = is decreasing. The case where yn−1 < y for n ≥ K is similar and will be omitted. 1 + yn0 2 p + qx 1+x 6.

b]. From Eq(6.23).23) lies eventually in the interval [0. By applying Theorem 1. 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. one of which is monotonically approaching 0 and the other is monotonically approaching ∞.4. M ]. 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. 2 6.6 it follows that y is a global attractor of all solutions of Eq(6. Set p + qy p f (x.5. a ≤ f (x.5. y2n−1 < q − 1 and y2n > y0 + n Hence n→∞ p . y) = .26) and Eq(6. The Case β = C = 0 : xn+1 = α+γxn−1 A+Bxn 91 where is some positive number. and b = + . y) ∈ [a. . y) ≤ b. The proof is complete.3 The Case q > 1 In this case we will show that there exist solutions that have two subsequences.6. lim y2n = ∞ p + qy2n−1 =0 1 + y2n and lim y2n+1 = lim n→∞ n→∞ and the proof is complete.5. b] × [a. The local stability of y was established in Theorem 6. a = 0.23). 1+x 1−q Then clearly. for all (x. it now follows that every solution of Eq(6.1.

27) and q = (Eq(6.92 Chapter 6. C (6. 1. B . 1+q (y + p)(1 + q) n = 0. . .29) .27) about y is zn+1 + q qy − p zn − zn−1 = 0.30) (6. The linearized equation associated with Eq(6.27) has the unique positive equilibrium y= 1+ 1 + 4p(1 + q) 2(1 + q) .1. F (u) = (1 + q)u2 − u − p.6) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + yn−1 . . 2)-type Eq(6. .28) By employing Theorem 1.6 The Case β = A = 0 : xn+1 = γ yn C α+γxn−1 Bxn +Cxn−1 This is the (2. qyn + yn−1 αC γ2 n = 0. If we now set it is easy to see that (6. . Clearly the equilibrium y is the positive solution of the quadratic equation (1 + q)y 2 − y − p = 0.1 we see that y is locally asymptotically stable when (q − 1)y < 2p and unstable (a saddle point) when (q − 1)y > 2p.27) was investigated in [45]. . 2)-Type Equations 6. (6. (6. 1. (2.) Eq(6.29) is satisﬁed if and only if either q≤1 or q > 1 and F 2p q−1 > 0. .

6. . (6.1 Let Existence and Local Stability of Period-Two Cycles .27) has a prime period-two solution .6. ψ. . . ψ. qφ + ψ It now follows after some calculations that the following result is true: Theorem 6. The following result is now a consequence of the above discussion and some simple calculations: Theorem 6. φ. when q > 1 + 4p.32) holds. The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 93 Similarly (6. ψ. .2 Eq(6. ψ. Furthermore when (6. .27) is locally asymptotically stable when q < 1 + 4p and unstable. φ. 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. . . . be a period-two cycle of Eq(6.6.32) (6. and more precisely a saddle point equilibrium. if and only if q > 1 + 4p.6. q−1 . φ. . φ. .6.27). Then φ= p+φ qψ + φ and ψ = p+ψ .1 The equilibrium y of Eq(6. .30) is satisﬁed if and only if q > 1 and F 2p q−1 < 0.31) 6. .

∂v (qψ + φ)2 . . ψ. the second iterate of T . φ ψ is a ﬁxed point of T 2 .94 To investigate the local stability of the two cycle . ∂g qψ − p (φ. . . qg(u. 1. v) = Clearly the two cycle is locally asymptotically stable when the eigenvalues of the φ lie inside the unit disk. v) + v and h(u. ψ) = . Jacobian matrix JT 2 . ψ) ∂h (φ. . By a simple calculation we ﬁnd that T2 where g(u. ∞) deﬁned by: T Then u v = v p+u qv+u . . 2)-Type Equations for n = 0.27) in the equivalent form: un+1 = vn vn+1 = p+un qvn +un . . ψ) = − . ψ. ∞) × (0. and write Eq(6. Let T be the function on (0. . v) h(u. ψ) ∂v (φ. φ. φ. . 1. ψ) ∂u φ JT 2 = ψ ∂h (φ. . Chapter 6. evaluated at ψ We have ∂g ∂g (φ. n = 0. ∂u (qψ + φ)2 ∂g (p + φ)q (φ. . . v) p+v . . v) = p+u qv + u u v = g(u. (2. . we set un = yn−1 and vn = yn . ψ) ∂u ∂v where.

33) is equivalent to the following three inequalities: S <1+D −1 − D < S D < 1. The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 95 q(p + ψ)(qψ − p) ∂h (φ.34) is equivalent to (6.33) (6.34) (6. p )[q − (1 + 4p)] < 0 p+1 . Inequality (6. ψ) = − . First we will establish Inequality (6.35) (6.34).36) (qψ − p)(qφ + ψ)2 + (qφ − p)(qψ + φ)2 + q 2 (p + ψ)(p + φ) − (qψ − p)(qφ − p) < (qψ + φ)2 (qφ + ψ)2 which is true if and only if p (q 3 + 2q 2 − 4qp + 2q 2 p + 2p − 3q) + q − p + q 2 p2 + qp − p2 < (qψ + φ)2 (qφ + ψ)2 q−1 which is true if and only if −q 3 p + 7q 2 p − 8qp2 + 4q 2 p2 + 4p2 − 7qp + 2q 2 − q + p − q 3 < 0 which is true if and only if −(p + 1)(q − 1)(q − which is true because q > 1 + 4p. ψ) = . ∂u (qφ + ψ)2 (qψ + φ)2 ∂h q 2 (p + φ)(p + ψ) qφ − p (φ.6.6. ψ) + (φ. ψ) − (φ.1 that both eigenvalues of JT 2 unit disk |λ| < 1. if and only if |S| < 1 + D < 2.1. ψ) (φ. ∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1. + 2 (qψ + φ)2 ∂v (qφ + ψ) (qφ + ψ)2 Set S= and D= ∂h ∂g (φ. ψ) (φ. ψ). ψ) ∂u ∂v ∂g ∂h ∂h ∂g (φ. To this end 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 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. we establish Inequality (6. Theorem 6. 6. Finally.6. p ≤ yN ≤ 1.35) 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.27) either remain forever outside the interval I or the solution is eventually trapped into I.35). Then the following statements are true. (a) Assume that p ≤ q and that for some N ≥ 0.27). 2)-Type Equations Next we will establish Inequality (6.96 Chapter 6. Observe that (6.6. q .36). (2.2 Invariant Intervals Here we show that the interval I with end points one and p is an invariant interval for q Eq(6.27). More precisely we show that the values of any solution {yn }∞ n=−1 of Eq(6. Observe that (6.3 Let {yn }∞ n=−1 be a solution of Eq(6.

3 Semicycle Analysis Here we present a thorough semicycle analysis of the solutions of Eq(6. 6. The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 97 (b) Assume that p ≥ q and that for some N ≥ 0.38) (6. Let {yn }∞ n=−1 be a solution of Eq(6. (b) Clearly.6.27) relative to the equilibrium y and relative to the end points of the invariant interval of Eq(6.6.39) 2 qyn−1 (yn − p ) + (yn − 1)(yn−1 yn + qyn ) q . p 1 ≤ yN ≤ . Then observe that the following identities are true: p − yn for n ≥ 0.27).6.27). (6. (a) Indeed we have yN +1 = and yN +1 = p + yN −1 p + yN −1 ≤ =1 qyN + yN −1 p + yN −1 p q p + yN −1 p = p + yN −1 q p q for all n ≥ N. p + yN −1 p + yN −1 ≥ = q qyN + yN −1 (pyN + p yN −1 ) p q and the proof follows by induction.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. for n ≥ 0. q Then 1 ≤ yn ≤ Proof. 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.

p ] for n ≥ N . yN ≥ p . yN < 1. then yN +1 > 1.27). then yN +1 < p . pyn + yn−1 Chapter 6. In particular [1. yN ≤ 1.98 When p=q that is for the diﬀerence equation yn+1 = p + yn−1 . yN ≥ 1. then yn ∈ [1.42) The following three lemmas are now direct consequences of (6.1 Assume that p>q and let {yn }∞ n=−1 be a solution of Eq(6. (6. (6. q (vi) If for some N ≥ 0. yN < p .40) the above identities reduce to the following: yn+1 − 1 = p and yn − yn+2 = (yn − 1) 1 − yn pyn + yn−1 for n ≥ 0 for n ≥ 0. .41) 2 pyn−1 + yn−1 yn + pyn p(p + yn−1 ) + yn (pyn + yn−1 ) (6. . Then the following statements are true: (i) If for some N ≥ 0.27). p ] is an q q q invariant interval for Eq(6. yN = p . (viii) p 1<y< . q (ii) If for some N ≥ 0. then yN +1 < 1.2 Assume that p=q ∞ and let {yn }n=−1 be a solution of Eq(6. Lemma 6.27).37)-(6. yN > p . yN ≤ p .6.42). . q (v) If for some N ≥ 0. . q (iii) If for some N ≥ 0. (2. then yN +1 > 1. 2)-Type Equations n = 0. then yN +2 > yN . 1. (vii) If for some N ≥ 0. then yN +1 = 1.6. then yN +2 < yN . q Lemma 6. Then the following statements are true: (i) If for some N ≥ 0. q (iv) If for some N ≥ 0.

q (iv) If for some N ≥ 0. (v) If for some N ≥ 0. q (iii) If for some N ≥ 0.27).1 and 1. (vii) If for some N ≥ 0.6.3: . yN ≤ 1. 1] for n ≥ N .7.1-6.6. then yN +1 ≥ 1. then yN +1 > p . yN = 1. yN > 1.6. (ix) p < y < 1. q (viii) If for some N ≥ 0. then yN +1 = 1. The Case β = A = 0 : xn+1 = α+γxn−1 Bxn +Cxn−1 99 (ii) If for some N ≥ 0. yN > p .2 and Lemmas 6. then yN +1 < 1. then yn ∈ [ p .3 Assume that p<q ∞ and let {yn }n=−1 be a solution of Eq(6. q Note that the function f (u. q (vi) If for some N ≥ 0. yN < 1. (iv) If for some N ≥ 0. q (ii) If for some N ≥ 0.6.27). Then the following statements are true: (i) If for some N ≥ 0. q (v) If for some N ≥ 0.7. yN = p . yN ≥ 1. q q The following result is now a consequence of Theorems 1. then yN +2 > yN . Lemma 6. then yN +1 < 1. In particular [ p . then yN +2 < yN . then yN > yN +2 > 1. then yN +1 > 1.6. yN ≤ p . v) = p+v qu + v is always decreasing in u but in v it is decreasing when u < p and increasing when u > p . yN ≤ p . 1] is an q q invariant interval for Eq(6. then yN +1 = 1. then yN < yN +2 < 1. yN < p . (iii) If for some N ≥ 0. yN ≤ 1. yN > 1.

must converge to y.6 and 1. .32) holds. (6.6.4 Let {yn }∞ n=−1 be a solution of Eq(6.27) has a “unique” prime period-two solution . On the other hand when (6. Hence. . 2)-Type Equations Then either all the even terms of the solution lie in J and all odd terms lie in K.27) has a unique equilibrium y which is locally stable when (6. or for some N ≥ 0. ψ. But this two cycle lies q inside the interval I. if Theorem 6. When (6. Eq(6.4.32) holds. and only then. The character of these solutions remains an open question. any solution which remains forever outside the interval I = [ p .43) 6. . . (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 = (0.27) eventually enters and remains in the interval I.44) (6.6.45). ∞). in view of Theorems 1.45) there are no period-two solutions and so the solutions must eventually enter and remain in I and. except for the length of the ﬁrst semicycle of the solution.27) is global attractor. The above observations are summarized in the following theorem: Theorem 6. q−1 (6. yn ∈ I for n ≥ N. q Now when q ≤ 1 + 4p Then the equilibrium y of Eq(6.4.7. φ. while if p>q the length is at most two.6. or stays forever outside I. .31) holds.100 Chapter 6. (2. every solution of Eq(6. with φ + ψ = 1 and φψ = p . ψ.46) Also recall that a solution {yn }∞ n=−1 of Eq(6.5 (a) Assume q ≤ 1 + 4p. or vice-versa. p<q the length is one.27). . . when (6.32) holds.44)-(6.4 Global Behavior of Solutions Recall that Eq(6. 1] must converge to the two cycle (6. When (6.27) either eventually enters the interval I with end points 1 and p and remains there. φ.43) holds.

6. . Then the zero equilibrium of Eq(6. (b) Assume p + q > 1.47) is globally asymptotically stable. 2)-type equation Eq(6. 1].7. 1 + yn n = 0. Then every solution of Eq(6.) The equilibrium points of Eq(6. The following result follows from Theorem 1.1: Theorem 6.7) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where pyn + qyn−1 .47) was investigated in [52]. (6.27) eventually enters and remains in the interval [ p . . A A (Eq(6. .47) are the solutions of the equation p= y= py + qy . y = p + q − 1 is the only positive equilibrium point of Eq(6.1 and the linearized stability Theorem 1. Then the zero equilibrium of Eq(6. .47) is unstable.3. q 6. More precisely it is a saddle point when 1−p<q <1+p and a repeller when q > 1 + p.47). 1+y So y = 0 is always an equilibrium point. The Case α = C = 0 : xn+1 = (b) Assume βxn +γxn−1 A+Bxn 101 q > 1 + 4p.7.47) β γ and q = .1.1 (a) Assume p + q ≤ 1.7 The Case α = C = 0 : xn+1 = A yn B βxn +γxn−1 A+Bxn This is the (2. 1. and when p + q > 1.

Then the positive equilibrium y = p + q − 1 of Eq(6.47) is locally asymptotically stable. (2.3 (a) Assume that q < 1.7. Furthermore. φ−p 6. φ. every solution of Eq(6.47) with q = 1. 2)-Type Equations 1 − p < q < 1 + p. Then the positive equilibrium y = p + q − 1 of Eq(6. (6.1 Invariant Intervals q=1 (6.2 Eq(6. ∞) : ψ = = 2p}. [ p . (6. . q (b) Assume that q > 1. (d) Assume q > 1 + p.3.7. remains in this interval.7.51) Then every solution of Eq(6. .48) holds. q . Concerning prime period-two solutions it follows from Sections 2. That is.47) q with initial conditions in (0.47) eventually enters the interval [ p . . The case is treated in Section 6.47). ∞). ∞) is an invariant interval for Eq(6.47). p ].47) is an unstable saddle point. . if and only if q = 1 + p. ψ ∈ (p. p ] is an invariant interval for Eq(6.102 (c) Assume Chapter 6. . ψ.49) The main result here is the following theorem about Eq(6.5 that the following result is true: Theorem 6.47) has a prime period-two solution .48) Furthermore when (6. φ.50) Then every solution of Eq(6. q (0. (6. Theorem 6.7. . p ]. the values of φ and ψ of all prime period-two solutions are given by: pφ {φ. ψ. Furtherq more.47) eventually enters the interval (0.

6. yk ≥ and yk+2 ≤ p p =⇒ yk+2 ≥ p > q q p p =⇒ yk > m+1 > p. Then the following statements are true for every n ≥ 0: p + q > 1 and q ≤ 1. Consider the function g : [0. qm q 6. m ∈ N. 1+x Observe that g(y) = y and that g increases on [0. y) = px + qy .50) holds. Let f : [0.2 Here we discuss the behavior of the semicycles of solutions of Eq(6.7. Then for any k. m q q (b) Assume that (6. ∞) × [0.4 Suppose Let {yn }∞ n=−1 be a positive solution of Eq(6. whose proof is straightforward and will be omitted.47). Hereafter. .1 (a) Assume that (6.7. m ∈ N.47) when p + q > 1 and q ≤ 1.7.7. ∞) deﬁned by g(x) = (p + q)x . ∞) denote the function f (x.51) holds. yk ≤ and yk+2 ≥ p p =⇒ yk+2 ≤ p < q q p p =⇒ yk > m+1 > p.47). ∞) − {x}. Then for any k. ∞) → [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. ∞) → [0. (x − x)(f (x. ∞). y refers to the positive equilibrium p + q − 1 of Eq(6. 1+x Semicycle Analysis of Solutions When q ≤ 1 Then f satisﬁes the negative feedback condition. that is. x) − x) < 0 for x ∈ (0. Lemma 6. Theorem 6.

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

52) holds. More precisely the following statements are true. The Case α = C = 0 : xn+1 = (d) Suppose yn < y ≤ yn−1 . . Suppose p − qyn−1 ≥ 0. yn−1 ≥ qyn−1 > y + (q − 1)yn > yn 2 pyn + qyn−1 qyn−1 (yn + 1) < ≤ yn−1 . (a) Assume y−1 ≥ p and y0 ≥ p.47) is globally asymptotically stable.7. Theorem 6. Then yn ≤ p for all n > 0 and (6. Case 2. Suppose p − qyn−1 < 0. we must consider two cases: Case 1.5 Suppose that q = 1 and let y−1 + y0 > 0. 1 + yn 1 + yn−1 The last inequality follows from the negative feedback property.3 Semicycle Analysis and Global Attractivity When q = 1 q = 1. Here we discuss the behavior of the solutions of Eq(6. Then yn+1 = (e) Observe that yn+1 = and so from which the result follows. Then {yn }∞ n=−1 oscillates about the equilibrium p with semicycles of length one and (6. (c) Assume either y−1 < p < y0 or y−1 > p > y0 . Then yn ≥ p for all n > 0 and n→∞ lim yn = p.6. Then the equilibrium y of Eq(6.52) holds. 1 + yn 1 + yn pyn + qyn−1 >y =p+q−1 1 + yn 6.7.7. Then px+qyn−1 increases in x and so 1+x yn+1 = pyn−1 + qyn−1 pyn + qyn−1 ≤ < yn−1 .47) when Here the positive equilibrium of Eq(6. To see that yn+1 < yn−1 .52) (b) Assume y−1 ≤ p and y0 ≤ p. 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. (6.47) is y = p.

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

7. y0 ∈ [p. Here we consider the case where and we show that the equilibrium point y = p + q − 1 of Eq(6.3 (a) we may assume that the initial conditions y−1 and y0 lie in the interval I = (0. 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). By Theorem 6.4. 2 6.47) we may assume.7. y) = px + qy 1+x satisﬁes the hypotheses of Theorem 1. ∞). The Case α = C = 0 : xn+1 = βxn +γxn−1 A+Bxn 107 and the claim follows by induction. .7. ∞).47). Observe that the consistency condition y ≤ p is equivalent to the condition q q ≤ 1.6. and so convergent.7. Then every positive solution of Eq(6. Now clearly the function q f (x. In a similar way we can show that the subsequence {y2k }∞ is decreasing.47) when which is equivalent to p y> .7. that y−1 .4 Global Attractivity When q < 1 1 − p < q < 1. Concerning the long-term behavior of solutions of Eq(6. every positive solution of Eq(6. without loss of generality. Thus.6 Assume 1 − p < q < 1. Theorem 6.47) lies eventually in the interval [p.47) converges to the positive equilibrium y.5 in the interval I from which the result follows. Proof. q In this case. m = M . p ]. 2 6.47) is global attractor of all positive solutions of Eq(6.5 Long-term Behavior of Solutions When q > 1 q>1 Here we discuss the behavior of the solutions of Eq(6.

. By applying the results of Section 6.47) converges to the positive equilibrium of Eq(6.7. transforms Eq(6. 2)-Type Equations p(q − 1) + qun−1 1 + p + un (6.47) has unbounded solutions. Then every positive solution of Eq(6.47). Then Eq(6. Theorem 6. The character of solutions of Eq(6.47) oscillates about the positive equilibrium y with semicycles of length one. Then every solution of Eq(6.47) when q > 1: Theorem 6.53) was investigated in Section 6.47) to the equation un+1 = Chapter 6.7. (b) Assume 1 < q < 1 + p.7 After the ﬁrst semicycle. (2.47) converges to a period-two solution. (c) Assume q > 1 + p. we obtain the following theorems about the solutions of Eq(6. an oscillatory solution of Eq(6.8 (a) Assume q = 1 + p.5 to Eq(6.53).5.53) where un ≥ 0 for n ≥ 0.108 Then the change of variable yn = un + p.

The Case α = B = 0 : xn+1 = βxn +γxn−1 A+Cxn−1 109 βxn +γxn−1 A+Cxn−1 6.55) holds and y−1 + y0 > 0.6.54) is unstable and the positive equilibrium y = p + 1 − q of Eq(6. A . Our goal is to show that when (6. γ (6. 1. Theorem 6.1.8) which by the change of variables reduces to the diﬀerence equation yn+1 = where p= pyn + yn−1 .54) and q = (Eq(6.54) is globally asymptotically stable. Furthermore the zero equilibrium is a saddle point when 1−p<q <1+p and a repeller when q < 1 − p. the positive equilibrium y = p + 1 − q of Eq(6.54) was investigated in [53]. (b) Assume p + 1 > q.54) is globally asymptotically stable.3. .8 The Case α = B = 0 : xn+1 = xn = γ yn C This is the (2.55) Then the zero equilibrium of Eq(6.8. In the remainder of this section we will investigate the character of the positive equilibrium of Eq(6.1 (a) Assume p + 1 ≤ q.1.54) are the solutions of the equation y= py + y .55) holds.54) and so we will assume without further mention that (6. . (6.54) also possesses the unique positive equilibrium y = p + 1 − q.54) is locally asymptotically stable.8.1 and 1. Eq(6. Then the zero equilibrium of Eq(6.54) and when p + 1 > q. 2)-type Eq(6. . q+y So y = 0 is always an equilibrium point of Eq(6. The following result follows from Theorems 1. .) The equilibrium points of Eq(6. q + yn−1 β γ n = 0.

54) is such that: < 1 if p < q < p + 1 y is = 1 if p = q > 1 if p > q. for n ≥ 0.58) Note that the positive equilibrium y =p+1−q of Eq(6.54). p + yn−1 n = 0. (6. (6.110 Chapter 6. (2. (6.60) (6. 2)-Type Equations 6. .56) q + yn−1 q p2 [yn − p ] + (p − q)yn−1 q yn+1 − = p p(q + yn−1 ) 2 for n ≥ 0.61) The following three lemmas are now direct consequences of the above identities. When p=q that is for the diﬀerence equation yn+1 = pyn + yn−1 .54). Then the following identities are easily established: q yn − p yn+1 − 1 = p for n ≥ 0.1 Assume that p<q (6.8.8. 1. Then the following statements are true: .1 Invariant Intervals and Semicycle Analysis Let {yn }∞ n=−1 be a positive solution of Eq(6.62) and let {yn }∞ n=−1 be a positive solution of Eq(6. .56)-(6.59) the identities (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 ≥ 0.57) and yn+1 − yn = (p − q)yn + (1 − yn )yn−1 q + yn−1 (6. (6. Lemma 6. .

(v) If for some N ≥ 0.8. yN ≤ 1. yN > p .6. q (ii) If for some N ≥ 0. q (iii) If for some N ≥ 0. Lemma 6.2 Assume that p=q (6. yn < 1 and the solution is strictly increasing. (ii) If y0 = 1. then yN +1 > yN .63) and let {yn }∞ n=−1 be a positive solution of Eq(6. then for n ≥ 0. p q then yN +1 > p . (vi) If for some N ≥ 0. . (iii) If y0 > 1. (iv) If for some N ≥ 0. yN > p .8.3 Assume that p>q (6.8. then yn = 1 for n ≥ 0. (vi) If for some N ≥ 0. yN = p .64) and let {yn }∞ n=−1 be a positive solution of Eq(6.54). then for n ≥ 0. yN < p . then yN +1 > 1. then yN +1 < 1. then yN +1 < 1. (v) If for some N ≥ 0. yN = p . The Case α = B = 0 : xn+1 = βxn +γxn−1 A+Cxn−1 111 q (i) If for some N ≥ 0. yn > 1 and the solution is strictly decreasing. Then the following statements are true: q (i) If for some N ≥ 0. yN > q 2 . (iv) If for some N ≥ 0. yN > q 2 . yN ≥ 1. then yN +1 < yN . yN ≥ 1. then yN +1 = 1. p q then yN +1 < p . then yN +1 < 1. then yN +1 > q p and yN +2 > 1. then yN +1 = 1.54). yN < p . Then the following statements are true: (i) If y0 < 1. yN ≤ 1. then yN +1 > 1. q (iii) If for some N ≥ 0. q (ii) If for some N ≥ 0. Lemma 6.

63) holds. (6. ∞). 1) the function f (u. y is a global attractor. Next.62) holds.54) eventually enters and remains in the interval (0. Now in the interval (0.64) holds. by Lemma 6. q+1 From the above observations and in view of Theorem 6. .112 Chapter 6.8. 2)-Type Equations 6. 2)-type equation was investigated in Section 6.4. (2. . Set yn = 1 + (q + 1)un for n ≥ −1.1 that if a solution {yn }∞ n=−1 is such that yn ≥ 1 for all n ≥ 0 then the solution decreases and its limit lies in the interval [1.54) converges to the equilibrium y = 1. it follows from Lemma 6. Hence.8.8.54) eventually enters and remains in the interval (1. ∞). Without loss of generality we will assume that yn > 1 for n ≥ −1.1 we have the following result: Theorem 6.8. 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 .65) with positive parameters and positive initial conditions. When (6.3 where we established in Theorem 6.2 Global Stability of the Positive Equilibrium When (6.1. . v) = pu + v q+v is increasing in both arguments and by Theorem 1.8. n = 0. it follows from Lemma 6. Here it is clear from Lemma 6.3 that every solution of Eq(6.3 that every solution converges to its positive equilibrium u= p−q .54) is globally asymptotically stable. This (2. assume that (6. . 1).8.8.2 that every solution of Eq(6.2 Assume that p + 1 > q and that y−1 + y0 > 0.8. This is impossible because y < 1. Then the positive equilibrium y = p + 1 − q of Eq(6.3. every positive solution of Eq(6.

1. . (6.1 (a) Assume that p > q. .67) .6. (b) Assume that p < q.66) is locally asymptotically stable. .66) B β and q = . . (6.) The only equilibrium point of Eq(6.1 we obtain the following result: Theorem 6. 1. . (p + 1)(q + 1) (p + 1)(q + 1) n = 0. (Eq(6.66) was investigated in [54].9 The Case α = A = 0 : xn+1 = γ yn C This is the (2. Then the positive equilibrium of Eq(6. The Case α = A = 0 : xn+1 = βxn +γxn−1 Bxn +Cxn−1 113 βxn +γxn−1 Bxn +Cxn−1 6.66) about y is zn+1 − p−q p−q zn + zn−1 = 0.68) (6. .9. 2)-type Eq(6. . Then the positive equilibrium of Eq(6. γ C To avoid a degenerate situation we will also assume that p= p = q. By applying the linearized stability Theorem 1.9) which by the change of variables xn = reduces to the equation yn+1 = where pyn + yn−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. qyn + yn−1 n = 0.66) is y= p+1 q+1 and the linearized equation of Eq(6.9.

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

The Case α = A = 0 : xn+1 = Set P = Then JT 2 βxn +γxn−1 Bxn +Cxn−1 115 p−q . ψ > 0.6.71) is equivalent to (p − q)φ (p − q)ψ + +1>0 (qψ + φ)2 (qφ + ψ)2 which is true if and only if (q − p)[(qφ + ψ)2 ψ + (qψ + φ)2 φ] < (qψ + φ)2 (qφ + ψ)2 .72) is always true. We will use the fact that φ + ψ = 1 − p. Observe that φ > 0. φ= pψ + φ qψ + φ φψ = p(1 − p) q−1 pφ + ψ .71) (6. qφ + ψ (6.72) φ ψ and ψ = Inequality (6.9. By applying Theorem 1.1 (c). P < 0 and Q < 0 and so Inequality (6.71). Next we will establish Inequality (6. (qφ + ψ)2 p−q (qψ + φ)2 φ ψ = and Q = and its characteristic equation is Pφ −P ψ −P Qψ 2 P Qφψ − Qφ λ2 + (P ψ + Qφ − P Qφψ)λ + P Qφψ = 0.1.73) (6. 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.

Inequality (6.68) holds. Inequality (6.74) = (q − p)(1 − p)2 (qp + 1 + 2p). = [q(φ + ψ)2 + (q − 1)2 φψ]2 (q − p) φψ < (1 − p)(q − p) . 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. 2)-Type Equations (q − p)[(qφ + ψ)(pφ + ψ) + (qψ + φ)(pψ + φ)] < [(qφ + ψ)(qψ + φ)]2 .71) is true if and only if (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. Now observe that the lefthand side of (6. (2.74) is p(1 − p) (p − 1)(q − 1)] q−1 (6. Finally.116 if and only if Chapter 6.73) is equivalent to (p − q)2 φψ < (qψ + φ)2 (qφ + ψ)2 which is true if and only if (q − p) φψ < (qψ + φ)(qφ + ψ) if and only if if and only if if and only if if and only if (q − p) φψ < (q 2 + 1)φψ + q(φ2 + ψ 2 ) (q − p) φψ < (q 2 + 1)φψ + q[(φ + ψ)2 − 2φψ] (q − p) φψ < (q − 1)2 φψ + q(φ + ψ)2 (q − p) φψ < (q − 1)2 if and only if p(1 − p) + q(1 − p)2 q−1 = (1 − p)2 (q − p)2 .

the function 2 f (x. Then {yn } oscillates about the equilibrium y with semicycles of length two or three.3 Global Stability Analysis When p < q The main result in this section is the following Theorem 6.3 Let {yn } be a nontrivial solution of Eq(6. we present a semicycle analysis of the solutions of Eq(6. . This function also satisﬁes Condition (1.35). where φ and ψ are the two positive and distinct roots of the quadratic Eq(6. The extreme in each semicycle occurs in the ﬁrst term if the semicycle has two terms and in the second term if the semicycle has three terms. Theorem 6.6. .9.67) holds. Proof. . (b) Assume p < q.66) posseses the prime period-two solution .9.68) holds. Then the following statements are true: (a) Assume p > q. ψ. y) = qx + y is increasing in x and decreasing in y.2 Semicycle Analysis In this section. Furthermore this prime period-two solution is locally asymptotically stable. φ.9. Then the positive equilibrium y of Eq(6.66): Theorem 6. . Then Eq(6. except possibly for the ﬁrst semicycle which may have length one. φ. y) is increasing in y and decreasing in x.66).9. after the ﬁrst semicycle. 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.69) of Eq(6. ψ. the following result is true about the local stability of the prime periodtwo solution (6. In summary.66) is globally asymptotically stable.70).9.66). (b) The proof follows from Theorem 1. .7.2 Assume that Condition (6.9.1 by observing that when p < q. or {yn } converges monotonically to y. Then either {yn } oscillates about the equilibrium y with semicycles of length one. 6.4 Assume that p<q and (6. (a) The proof follows from Theorem 1. 6.7.4 by observing that the condition p > q implies that the function px + y f (x. . .

φ.5 Assume that (6. .75) holds. Set f (x. . ψ. φ) = φ and in general yN +2k ≥ ψ and yN +2k+1 ≤ φ for k = 0.68) holds. ψ) = ψ. y > 0. .4. qx + y from which we conclude that limk→∞ yN +2k = ψ and limk→∞ yN +2k+1 = φ. 2 The method employed in the proof of Theorem 1.4. Set f (x. ψ. and lim inf yn = φ n→∞ Now as in the proof of Theorem 1. Proof. φ. ψ. We consider ﬁrst the case where eventually consecutive terms yi . . . Now the conclusion of Theorem 6.68) holds.69) when instead of (6.66) converge to the two cycle (6. Theorem 6. .67).75) yN ≥ ψ and yN +1 ≤ φ. yN ) ≥ f (φ. Then this solution converges to the two cycle φ. Assume that (6..67). denote the two cycle of Eq(6.118 Proof..6 can also be used to establish that certain solutions of Eq(6. .6 lim sup yn = ψ n→∞ px + y . y) ≤ 1 for all x. . yN +1 ) ≤ f (ψ. yN +2 = f (yN +1 .4 follows as a consequence of Theorem 1. y) = Then clearly.9. y) = px + y . Let φ. Eq(6. qx + y Chapter 6. (6. yN +3 = f (yN +2 . 1. Also clearly.9. y) is decreasing in x for each ﬁxed y.4. 2)-Type Equations and note that f (x. q Finally in view of (6. (2.66). Assume that for some solution {yn }∞ n=−1 of Eq(6.66) has no prime period-two solution. with φ < ψ. and increasing in y for each ﬁxed x. 2 Next we want to ﬁnd more cases where the conclusion of the last theorem holds.66) and for some index N ≥ −1.6 and the fact that y is locally asymptotically stable. . (6. . yi+1 lie between m and M . ψ. . p ≤ f (x.

. and the result follows by induction. M ) = M. (B) y2n+1 > M and M ≥ y2n > m for every n. φ.6. φ. ψ. . .2 In all cases (A)-(D) the corresponding solution {yn } converges to the two cycle . Proof. M ] (that is. Proof. From Eq(6. and so M ≥ lim supn→∞ y2n ≥ lim inf n→∞ y2n ≥ M. then we have one of the following four cases: (A) y2n > M and M ≥ y2n+1 > m for every n. Since the proofs are similar for all cases we will give the details in case (A).9. (C) y2n < m and m ≤ y2n+1 < M for every n. . By our earlier observation we have that M ≥ lim supi→∞ yi . for every i. Using the monotonic character of the function f we have m = f (M. (D) y2n+1 < m and m ≤ y2n < M for every n. Then yk ∈ [m. M ] for every k > i. 2 If we never get two successive terms in the interval [m. qy2n+2 − p M − M2 lim y2n+1 = . either yi > M and yi+1 < m or yi < m and yi+1 > M ). m ≤ yi .9. M ] and if we never get two successive terms outside of the interval [m. qM − p = m and the proof is complete. yi+1 ≤ M. . ψ.9. m) ≤ f (yi+1 . . m= 2 Since the case in which the solution lies outside of the interval [m. M ) = pm + M qm + M for m we ﬁnd Thus.9. Hence limn→∞ y2n = M . The Case α = A = 0 : xn+1 = βxn +γxn−1 Bxn +Cxn−1 119 Lemma 6. .66) we have y2n+1 = and so y2n − y2n y2n+2 . yi ) = yi+2 ≤ f (m. Lemma 6. we are left with the case in which the solution lies eventually in [m. limn→∞ y2n+1 M − M2 . n→∞ qM − p On the other hand by solving the equation M = f (m. M ] is answered by Theorem 6.5. M ].1 Suppose that for some i ≥ 0.

3 Suppose M ≥ yi+2 ≥ yi ≥ y ≥ yi+1 ≥ yi+3 ≥ m and not both yi+2 and yi+3 are equal to y. ψ. φ. Induction on i then establishes the monotonicity of the two sequences. We denote this value by y . Proof. . and for m < y < y. ψ. (2) x > y if and only if x < y . y) = y. . In view of the uniqueness of the two cycle and the fact that at least one of yi . yi+3 ) ≤ f (yi+2 . 2)-Type Equations Lemma 6. (2. Therefore. y ) < y and y > y . y ) > y and y > y. . . Thus. but this situation cannot occur. yi+2 ) ≥ f (yi+1 . in which case Eq(6. yi ) = yi+2 and yi+5 = f (yi+4 . so we are led to the general equation f (x.76) has a q unique positive solution for x.The case M ≥ yi+3 ≥ yi+1 ≥ y ≥ yi ≥ yi+2 ≥ m is handled in a similar way. yi ) ≤ yi . Let us consider the ﬁrst case. the proof is complete. yi+1 is diﬀerent from y.9. . which of course depends on y. Then {yn } converges to the two cycle . Proof. f (y. . The correspondence between y and y has the following properties: Lemma 6. f (y. both subsequences converge and their limits form a two cycle. y) > y if and only if x < y .9.120 Chapter 6. The second case can be handled in a similar way. Here we have yi+4 = f (yi+3 . . In this case we need to know more about the condition yi+2 = f (yi+1 . (3) For y < y < M. (6. as the following results show. y) = y. 2 Using a similar technique one can prove that if M ≥ yi ≥ yi+2 ≥ y ≥ yi+3 ≥ yi+1 ≥ m or M ≥ yi+1 ≥ yi+3 ≥ y ≥ yi+2 ≥ yi ≥ m we would get convergence to y . Thus f (y . and the sequence {yi+2k+1 } is a non increasing sequence and bounded below by m.76) It makes sense to consider this equation only for y ∈ ( p .4 (1) f (x. yi+1 ) = yi+3 . the sequence {yi+2k } is non decreasing and bounded above by M . φ. 1).

so the roots of the numerator are not real. for m < y < y.9. M . y) > f (y . Then yi+3 ≤ yi+1 with equality only in the case of a two cycle. By (1) then. 1). y) and y = f (y. y>y . (p − qy)2 y − y2 . y < y . 2 Now we will use this result to prove the following: Lemma 6. so y is decreasing function q of y. Hence q m < y < y ⇒ f (y. x < y if and only if f (x. y ) < 1.5 Assume that for some i. g (y) < 0. qy − p The discriminant of the numerator is 4p2 − 4pq = 4p(p − q) < 0.9. y ) < y . Simplifying this equation we get a fourth degree polynomial equation with roots 0. y ) < y . Thus for y < m.6. y ) − y changes sign as y passes each of these values. For y ∈ ( p . Thus each of these roots is simple and so f (y. y. Clearly as y approaches p y approaches ∞. m. and for y < y < M. f (y. y) = y for y gives y = g(y) = Computing the derivative of g we get g (y) = − p − 2y + qy 2 . The Case α = A = 0 : xn+1 = βxn +γxn−1 Bxn +Cxn−1 121 (1) Since f is decreasing in the ﬁrst variable and f (y . y ) > y . However q p < f (y. M ≥ yi ≥ yi+2 ≥ y ≥ yi+1 ≥ m. . y ) > y . y < y < M ⇒ f (y. M < y <⇒ f (y. y) = y. y) = y. y ) we get px + x − x2 = qx − p qx + x−x2 qx−p x−x2 qx−p . (2) Solving f (y . (3) Eliminating y from the equations y = f (y .

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

77) and observe that when p > q. x0 ) ∈ F. d ∈ R. Proof. We will employ Theorem 1. y) is increasing in x for each ﬁxed y.9. c. the function f (x. Open Problems and Conjectures 123 6.77) holds. Open Problem 6.4. qx + y (6. Also 1 ≤ f (x. the only solution of the system M= is m = M. 1. To this end. cxn + dxn−1 n = 0. set f (x. Finally observe that when (6. b. .9. .5. Then the positive equilibrium y of Eq(6.10.10 Open Problems and Conjectures a. y) = px + y . .10.66) is globally asymptotically stable.4.4 Global Stability Analysis When p > q Here we have the following result: Theorem 6. y > 0. and is decreasing in y for each ﬁxed x.1 (See [64]) Assume that (a) Investigate the forbidden set F of the diﬀerence equation xn+1 = axn + bxn−1 xn . y) ≤ p q for all x.6. Now the result is a consequence of Theorem 1. investigate the asymptotic behavior and the periodic ∞ nature of the solution {xn }n=−1 . / (b) For every point (x−1 .7 Assume that p>q and p ≤ pq + 1 + 3q. . qm + M 6. .5. qM + m m= pm + M . 2 pM + m .

1.6 for the autonomous case. not necessarily prime. (See Section 1. Bn ∈ R for n = 0. Show that every positive solution of Eq(6.124 (See Section 1. . . . .2.10. .) Chapter 6. Show that the equation yn+1 = p + yn−1 .11) has a ﬁnite limit. . .10. An . 2)-Type Equations Open Problem 6.3 Assume p ∈ (0. every positive solution converges to a. ∞). (Note that by Theorem 6.) .5. An + Bn xn n = 0. (Note that by Theorem 6. this conjecture has been conﬁrmed for q ≤ 1 + 4p. has a positive and monotonically decreasing solution.) Conjecture 6.10) has a ﬁnite limit.4. 1 + yn n = 0.2 Assume that αn . (2. . ∞). . . 1. we need only to conﬁrm this conjecture for p > 2(q +1) and q ≥ 1.1 Assume p. q ∈ (0.10. 1. Investigate the forbidden set and the asymptotic character of solutions of the nonautonomous Riccati equation xn+1 = αn + βn xn .3. βn .4.2 Assume that p. Show that every positive solution of Eq(6. See also [36].10. q ∈ (0. (Note that by Theorem 6.) Conjecture 6.6.) Conjecture 6. ∞).3. are convergent sequences of real numbers with ﬁnite limits. period-two solution.

27). (See Section 6. Open Problems and Conjectures Conjecture 6. . q ∈ (0. ∞) and q > pq + 1 + 3p.9. (Note that by Theorem 6.) Open Problem 6.4 Assume that p. Open Problem 6.10.5 Find the set of all initial conditions (y−1 .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. Find the basin of attraction of the period-two solution of Eq(6.10.4 Assume that p.10. as n → ∞.27) either converges to a ﬁnite limit or to a two cycle. ∞) and q > 1 + 4p. q ∈ (0. Conjecture 6.10.10. 125 Show that every positive solution of Eq(6.10.5 Assume p > q. q ∈ (0.7.66).3 Assume that p. Find the basin of attraction of the period-two solution of Eq(6. ∞). this conjecture has been conﬁrmed for p ≤ pq + 1 + 3q. Show that every positive solution of Eq(6.6.66) has a ﬁnite limit.) Open Problem 6.

C ∈ (0. 1. ∞).80) (6. . y0 ) ∈ R × R through which the equation is well deﬁned for all n ≥ 0. Investigate the asymptotic behavior and the periodic nature of solutions of the diﬀerence equation αxn + βxn−1 + γxn−2 .78) (6.10. .7 For each of the following diﬀerence equations determine the “good” set G ⊂ R × R of initial conditions (y−1 . (6. 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. β.83) yn − yn−1 yn − 1 yn − yn−1 1 − yn−1 1 − yn−1 yn − yn−1 yn + yn−1 . B. 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 (6.126 Chapter 6. extend your result by introducing arbitrary real parameters in the equation. γ.10.82) (6. y0 ) ∈ G. . n = 0. (2.79) (6.81) (6. xn+1 = Axn + Bxn−1 + Cxn−2 Extend and generalize. A. Open Problem 6. .84) yn − yn−1 For those equations from the above list for which you were successful. .8 Assume α.Then for every (y−1 . 2)-Type Equations Open Problem 6.6 Find the set of all initial conditions (y−1 .10. as n → ∞. Open Problem 6.

1.10.86) (6. . ∞).91) (6. (a) Find the set B of all initial conditions (y−1 . 1.10 Assume that {pn }∞ and {qn }∞ are period-two sequences n=0 n=0 of nonnegative real numbers. y0 ) ∈ (0. . qn yn + yn−1 pn yn + qn yn−1 . Open Problems and Conjectures 127 ∞ Open Problem 6. ∞) such that the solutions {yn }∞ n=−1 of Eq(6. . .85) (6.92).92) pn + qn yn . . . qn yn + yn−1 yn+1 = yn+1 = Open Problem 6. n = 0. p = n→∞ pn lim lim and q = n→∞ qn . . n = 0. yn + qn yn−1 pn + qn yn−1 . 1. . y0 ) ∈ B. 1. Investigate the asymptotic behavior and the periodic nature of all positive solutions of each of the following eight diﬀerence equations: yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = pn + yn . 1.87) (6.9 Assume that {pn }n=0 and {qn }∞ are convergent sequences of n=0 nonnegative real numbers with ﬁnite limits. .10.89) (6. n = 0. . .23) are bounded. n=−1 . (6. . ∞) × (0. 1 + yn−1 yn + pn .85)-(6. Investigate the global character of all positive solutions of Eqs(6. . qn + yn−1 pn yn + yn−1 . 1. . . . n = 0. . Open Problem 6. (b) Let (y−1 . 1 + yn pn yn + yn−1 . .6. . .11 Assume q ∈ (1. .88) (6. . n = 0. 1 + yn pn + yn−1 .10. .10. Extend and generalize. n = 0.90) (6. 1. n = 0. qn + yn n = 0. 1. . . Investigate the asymptotic behavior of {yn }∞ .

∞). 1) with α + β > 1 + β . 2)-Type Equations (a) Find the set B of all initial conditions (y−1 . y0 ∈ (0. Open Problem 6. ∞). and the asymptotic behavior of the solution {xn }∞ n=−1 of Eq(6. y0 ) ∈ B. ∞) × (0. . Investigate the boundedness character. (b) Let (x−1 . (a) Determine the set G of initial conditions (y−1 . 2 Show that the positive equilibrium of System (6. . ∞). ∞) and that α ∈ (2. n = 0. y0 ) ∈ (0.12 Assume q ∈ (1. and γ ∈ (0. (2. See [14].128 Open Problem 6. Chapter 6. 1. ∞) such that the solutions {yn }∞ n=−1 of Eq(6. (6. and [2]) Assume α ∈ (1. 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 . x0 ) ∈ G. 1. ∞) through which the solutions {xn }∞ n=−1 of the diﬀerence equation xn+1 = (1 − xn − xn−1 )(1 − e−Axn ). Investigate the asymptotic behavior of {yn }n=−1 . 1 β = 1 and γ = .14 (Discrete Epidemic Models.6 Assume x0 .10. . the periodic nature. . .) Let A ∈ (0. . n = 0.10.94) . ∞).94). ∞) × (0. y0 ) ∈ (0.10. [3]. β ∈ (0. (6.47) are bounded.13 (A Plant-Herbivore System. See [9]. 3).93) is globally asymptotically stable.93) Conjecture 6. ∞ (b) Let (y−1 . . Open Problem 6. are nonnegative.10. (c) Extend and generalize.

∞) with c1 + c2 > 0. . n = 0. . . . . Investigate the global behavior of all positive solutions of each of the following diﬀerence equations: p + qyn yn+1 = . . 1.101) .}. b ∈ (0. .98) (6.97) (6. 3. Investigate the global character of all positive solutions of the system . Open Problem 6. n = 0.17 Assume that p. n = 0. 1.10. 1. ∞) and k ∈ {2. . 1. n = 0. . qyn + yn−k yn + pyn−k . 1. . . 1). n = 0. . 1.96) (6. ∞). Open Problems and Conjectures Open Problem 6. (6. . .10. 1.6. n = 0. 1) and xn+1 = αxn e−yn + β yn+1 = αxn (1 − e −yn β ∈ (1. 1. q + yn−k pyn + yn−k . n = 0. .99) (6. (6. See [48]). . .100) (6.95) 1 + yn−k yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn+1 = yn + p . yn + qyn−k p + qyn−k . and c1 .10. Assume a ∈ (0.15 (The Flour Beetle Model. . with positive initial conditions. . 1 + yn p + yn−k . c2 ∈ [0. qyn + yn−k n = 0. n = 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 . .16 (A Population Model) Assume α ∈ (0. . ) which may be viewed as a population model. .10. . . . . 1. . q ∈ [0. ∞). Open Problem 6. yn + q pyn + yn−k . . . . .

.10. 2)-Type Equations Open Problem 6.20 Assume that Eq(6.11) and (6. (6. . . which extends and uniﬁes the global asymptotic stability results for Eqs(6. Obtain necessary and suﬃcient conditions on f so that the same two cycle is a locally asymptotically stable solution of Eq(6. . n = 0. (2. . 1.103).102) has a two cycle which is locally asymptotically stable. . n = 0.130 Chapter 6. . . . yn−1 ).19 Assume that every positive solution of an equation of the form xn+1 = f (xn . xn−1 ). . also converges to a period-two solution. (6.18 Obtain a general result for an equation of the form yn+1 = f (yn . Extend and generalize.103) Open Problem 6. n = 0. 1. 1. xn−1 ).10.27). Obtain necessary and suﬃcient conditions on f so that every positive solution of the equation xn+1 = f (xn−2 . Open Problem 6.10.102) converges to a period-two solution.

3)-Type Equations 7. A + Bxn + Cxn−1 n = 0. A2 .4) αC .2) Eq(1) contains the following three equations of the (1. (7. . A + Bxn + Cxn−1 n = 0.Chapter 7 (1. A + Bxn + Cxn−1 βxn . (7. 1. . 1 + pyn + qyn−1 αB A2 and q = 131 n = 0.1) (7. . .2 The Case β = γ = 0 : xn+1 = α yn A α A+Bxn +Cxn−1 This is the (1. . . 1. 1. . . 1. 7. .3) α .1) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= 1 .(7.1 Introduction xn+1 = xn+1 = and xn+1 = γxn−1 . . 3)-type: Please recall our classiﬁcation convention in which all the parameters in the above (1. . n = 0. 3)-type Eq. . 3)-type equations are positive and that the initial conditions are nonnegative. . (7.

5) is globally asymptotically stable. It also follows by applying Theorem 1. . A C Eq(7. or Theorem 1.5) always has zero as an equilibrium point and when p>1 it also has the unique positive equilibrium point y= p−1 . Theorem 7.4) is globally asymptotically stable.3.4.7 in the interval [0. 7.2. The following result is now a straightforward consequence of either the stability trichotomy Theorem 1. .2 Assume that y−1 . 1 + pyn + qyn−1 p= n = 0.2) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where yn .1 Assume p ≤ 1. . 3)-type Eq(7. (1. . 1]. ∞) and p > 1.4. p ]. Then the zero equilibrium of Eq(7. Then the positive equilibrium of Eq(7. 3)-Type Equations One can easily see that the positive equilibrium of Eq(7.5 in the interval [0. y0 ∈ (0.1. Theorem 7.4.4) has no prime period two solutions.3. q+1 The following result is a straightforward consequence of Theorem 1.132 Chapter 7. q Theorem 7.5) β B and q = .5) is globally asymptotically stable.2.4) is locally asymptotically stable for all values of the parameters and that Eq(7.4.4. 1.3 The Case α = γ = 0 : xn+1 = A yn C βxn A+Bxn +Cxn−1 This is the (1.3.1 The positive equilibrium of Eq(7. (7. The following result is a straightforward consequence of Theorem 1.

1. .6) always has the zero equilibrium and when p<1 it also has the unique positive equilibrium y= 1−p . p + qyn + yn−1 A γ n = 0. . ψ. Then the positive equilibrium of Eq(7.4. . Theorem 7. φ. ψ.3.3) which by the change of variables xn = reduces to the equation yn+1 = where p= yn−1 .6) and q = Eq(7.6) is locally asymptotically stable when q<1 and is unstable (a saddle point) when q > 1.1.1 and 1. . 3)-type Eq(7.4. C (7. if and only if p < 1.7. . . .6) has a prime periodtwo solution . .6) is globally asymptotically stable when p≥1 and is unstable (a repeller) when p < 1.1. . The Case α = β = 0 : xn+1 = γxn−1 A+Bxn +Cxn−1 133 γxn−1 A+Bxn +Cxn−1 7. . 1+q The subsequent result is a straightforward application of Theorems 1.4 The Case α = β = 0 : xn+1 = γ yn C This is the (1. (b) Assume that p < 1. B .1 (a) The zero equilibrium of Eq(7. Concerning prime period-two solutions one can see that Eq(7. φ.

1). 0.2 Assume that p<1 Then the period-two solution . φ. (7. 1 − p − φ. . . φ.1 Assume Show that every positive solution of Eq(7. 1 − p. 1 − p.6) are given by . 2 and q > 1. 0. .6 that here JT 2 0 1−p = 1 p+q(1−p) q(1−p) − p+q(1−p) 1−p . On the other hand.6) is locally asymptotically stable. with 0 ≤ φ ≤ 1 − p and φ = Theorem 7. . of Eq(7.4. . 2 and in view of (7. Proof. (1.8) 0 p 0 lie in the disk |λ| < 1.134 Furthermore when Chapter 7.6) has a ﬁnite limit. . 3)-Type Equations p < 1 and q = 1 the only prime period-two solution is . 0.5.7) (7. Conjecture 7. . when p < 1 and q = 1 all prime period-two solutions of Eq(7. . .7) both eigenvalues of JT 2 7. 0. . . q ∈ (0. 1 − p − φ. . It follows from Section 2. . . .5 Open Problems and Conjectures p. . . 1 − p. 1 − p.8) is locally asymptotically stable. . . . Therefore 1−p the prime period-two solution (7.

5. Extend and generalize. . . 1 − p.6) converges to a.7.1 Assume p ∈ (0.2 Assume p < 1 and q ≥ 1. 1 + pxn + qxn−1 + xn−2 n = 0.11) Now for each of the equations (7.10) (7.5.2 Investigate the asymptotic character and the periodic nature of all positive solutions of the diﬀerence equation xn+1 = where p. ∞). . Find the basin of attraction of the two cycle . 0. let (x−1 .5. Open Problem 7.11). 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 .6).3 For each of the equations given below. Open Problem 7. x0 ) ∈ G be an initial point through which the corresponding equation is well deﬁned for all n ≥ 0. 1) and q ∈ (1. 1 − p.9) (7. not necessarily prime.5. 1. .5. . . Open Problems and Conjectures Conjecture 7. and investigate ∞ the long term-behavior of the solution {xn }n=−1 . . Open Problem 7. 1 + xn + xn−1 (7. q ∈ [0. of Eq(7. ﬁnd the set G of all points (x−1 . periodtwo solution. ∞). .9)-(7. . 135 Show that every positive solution of Eq(7. 0. . xn−2 .

.5 Assume that {pn }∞ and {qn }∞ are period-two sequences of n=0 n=0 nonnegative real numbers. (1. 1 + pn yn + qn yn−1 yn .12)-(7. 3)-Type Equations Open Problem 7. . n = 0.136 Chapter 7.14). 1 + pn yn + qn yn−1 yn−1 . p = lim pn n→∞ and q = n→∞ qn . .14) Open Problem 7.5. . 1. . (7. pn + qn yn + yn−1 n = 0. . . . n = 0. Investigate the global character of all positive solutions of Eqs(7.4 Assume that {pn }∞ and {qn }∞ are convergent sequences of n=0 n=0 nonnegative real numbers with ﬁnite limits. . 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 . Extend and generalize.12) (7. . .5. 1.13) (7. 1.

Eq(8.2) Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. β + Byn n = 0. Bxn Eq(1) contains the following three equations of the (1. 1. B is reduced to a (2. . C .1) is a linear diﬀerence equation and can be solved explicitly. Eq(8. . .5). . (8. 1. 1)-Type Equations 8. Eq(8. yn+1 = (α + βγ ) + γyn−1 B . . . 1.Chapter 8 (3.4) (see Section 6. . n = 0. Cxn−1 n = 0. the initial conditions are nonnegative. (8. 2)-type equation of the form of Eq(6.2) by the change of the variables xn = yn + β . . A α + βxn + γxn−1 . 3)-type: xn+1 = xn+1 = and xn+1 = α + βxn + γxn−1 . namely. .1) (8.1 Introduction α + βxn + γxn−1 . and the denominators are always positive. . 1.3) n = 0. .3) by the change of the variables xn = yn + 137 γ . . . .

. .3) namely.1 Assume γ > β. (8. φ. Hence there is nothing else remaining to do about these equations other than to pose some Open Problems and Conjectures.4). 2)-type equation of the form of Eq(6. .1 Show that every positive solution of Eq(8.2. x0 ) ∈ R × R through which the equation 1 + xn + xn−1 xn+1 = xn is well deﬁned for all n and for these initial points determine the long-term behavior of the solutions {xn }∞ . .2. Determine the basin of attraction of (8. Determine the values of φ and ψ in terms of the initial conditions x−1 and x0 . . Open Problem 8. x0 ) ∈ R × R through which the equation 1 + xn + xn−1 xn+1 = xn−1 is well deﬁned for all n and for these initial points determine the long-term behavior of the solutions {xn }∞ . .2 (a) Assume γ = β and let .2. (b) Let {xn }∞ n=−1 be a positive solution of Eq(8. ∞) × (0.2. Conjecture 8. x0 ) ∈ (0. n=−1 Open Problem 8. . 8. n=−1 .4) be a period-two solution of Eq(8. (3. yn+1 (α + βγ ) + βyn C .2). .4).3) converges to the positive equilibrium of the equation. 1. .2) (see Section 6. ∞) through which the solutions of Eq(8.2) which converges to (8.2 Open Problems and Conjectures Open Problem 8. Open Problem 8.138 Chapter 8.3 Determine the set G of all initial points (x−1 . 1)-Type Equations is reduced to a (2.4 Determine the set G of all initial points (x−1 .2) are bounded. ψ. φ.2. . Determine the set of initial conditions (x−1 . ψ. = γ + Cyn−1 n = 0.

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.

y4k+1 > q. y4k ≥ 1. . y4k+1 < q.4 Assume p = q(qr + 1) and let {yn }∞ n=−1 be a solution of Eq(9. (2. 1. (iv) y−1 < q and y0 ≤ q =⇒ y4k−1 < q. Then yn+1 − q = Furthermore when qr < 1.10) is a consequence of the fact that in this case qr ∈ (0. y4k ≥ 1. Theorem 9.3 Suppose that p = q + rq 2 and let {yn }∞ n=−1 be a nontrivial solution of Eq(9.2.5). excluding the ﬁrst. ∞) and Eq(9. and y4k+2 ≥ q. y4k ≤ 1. (ii) y−1 < q and y0 ≥ q =⇒ y4k−1 < q. y4k+1 < q.4) when p = q(1 + rq). Identity (9. then n→∞ Chapter 9. and the only equilibrium point is y = q. and y4k+2 ≤ q. (iii) y−1 > q and y0 ≥ q =⇒ y4k−1 > q. 2 . 3)-Type Equations qr (q − yn−1 ). y4k+1 > q.9) follows by straightforward computation. The proof follows from identity (9.2.10) Proof. and y4k+2 ≤ q. In this case Eq(9. . . (9.4). 1 + yn + ryn−1 n = 0.4). 2 Here we present a semicycle analysis of the solutions of Eq(9. is equal to four. The limit (9.4) becomes yn+1 = q + rq 2 + qyn . Proof. Then this solution is oscillatory and the sum of the lengths of two consecutive semicycles. the following statements are true for all k ≥ 0: (i) y−1 > q and y0 ≤ q =⇒ y4k−1 > q.9) (9. and y4k+2 ≥ q. 1 + yn + ryn−1 lim yn = y.146 Lemma 9. y4k ≤ 1. More precisely.4) has no period-two solution.

if for some N . y ∈ I. yN +1 < p−q . (1 + yn + ryn−1 )(1 + b) n ≥ 0. q−1+ (q − 1)2 + 4p 2 Hence. First assume that p ≤ q. the following is true: Lemma 9. qr and yN +2 < . q−1+ (q−1) +4p ] if p ≤ q. yN ≤ b. qr p−q . every solution of Eq(9. Then every solution of Eq(9.2.4) lies eventually in I. (ii) yN > q. Next assume that q < p < q(qr + 1) and. p > q(qr + 1). p−q ] qr if if q < p < q(qr + 1). also assume that the solution {yn }∞ n=−1 is not eventually in the interval I. there exists N > 0 such that one of the following three cases holds: (i) yN > q.2.4) described in Lemma 9.2. yN +1 > q. Then clearly. Set b= and observe that yn+1 − b = (q − p)(yn − b) − r(p + qb)yn−1 . More precisely.1. Now assume for the sake of contradiction that yn > b for n > 0.2.4). for the sake of contradiction. q] qr [q.9.2. Then yn > y for n ≥ 0 and so which is a contradiction. The Case γ = 0 : xn+1 = α+βxn A+Bxn +Cxn−1 147 9. ∞ Proof.3.5 Let I denote the interval which is deﬁned as follows: √ 2 [0. Let {yn }n=−1 be a solution of Eq(9. then yN +1 < b. qr n→∞ lim yn = y ∈ I and yN +2 < p−q .4) is eventually trapped into one of the three invariant intervals of Eq(9. Then by Lemma 9.4 Global Asymptotic Stability The following lemma establishes that when p = q(qr + 1). 2 I= [ p−q .

the following results are true for Eq(9. Proof.5. we can obtain some global asymptotic stability results for the solutions of Eq(9.2. (2. 3)-Type Equations ≤ yN +1 ≤ q. p−q qr Chapter 9. Then the equilibrium y of Eq(9. 1.4) is globally asymptotically stable. 1+r Then the equilibrium y of Eq(9. p< p≤q or q < p < q + q2r and that one of the following conditions is also satisﬁed: (i) q ≤ 1. together with the appropriate convergence Theorems 1.4.2. For example. then yN + ryN − p < 0. then yN + ryN − p > 0 and 2 if yN ≤ p−q . (b) Assume that either (iii) r > 1 (ii) r ≤ 1. 3}.7 and 1. 2.4): Theorem 9.4) is globally asymptotically stable. and yN +2 < p−q . or q .7) from which it follows that for j ∈ {0.4. The proof when p > q(qr + 1) is similar and will be omitted.4). qr Also observe that 2 if yN ≥ q. each subsequence {yn+4k+j }∞ with all its terms outside k=0 the interval I converges monotonically and enters in the interval I.1. . qr The desired contradiction is now obtained by using the Identity (9.148 (iii) yN > q. and (q − 1)2 (r − 1) ≤ 4p.4 (a) Assume that either p ≥ q + q 2 r. y) = p + qx 1 + x + ry 2 in each of the intervals in Lemma 9. By using the monotonic character of the function f (x.

11) A yn B αB . Clearly now if Condition (ii) or (iii) is satisﬁed. .2.4.1 we see that when y−1 . √ q−1+ (q−1)2 +4p (b) In view of Lemma 9.2. 3)-type Eq(9. 2 decreases in y.9. this is the case if Condition (i) is satisﬁed. y0 ∈ 0. 2 2 √ 2 +4p q−1+ (q−1) .1 and 9.2.1 and Theorem 1. then M = m. 1 + M + rm Hence (M − m)(1 − q + M + m) = 0. A2 and r = .3 The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 This is the (2.4. 2 9. then yn ∈ 2 √ √ q−1+ (q−1)2 +4p q−1+ (q−1)2 +4p for all n ≥ 0. 1. M ) and M = f (M. The proof when q < p < q + q 2 r holds follows from Lemma 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 . C . We will employ Theorem 1.2.7 and 1. Now if m + M = q − 1.2. This system has the form m= p + qm 1 + m + rM and M = p + qM . m) then M = m. It is easy to check that y ∈ 0. . A n = 0.4 and Theorems 1. B (9. If m + M = q − 1 then m and M satisfy the equation (r − 1)m2 + (r − 1)(1 − q)m + p = 0. . The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 149 (a) The proof follows from Lemmas 9. 0. m = M from which the result follows.3. For instance.4. .4.5.5 and so it remains to be shown that if m = f (m. 1 + yn + ryn−1 q= γ .

ψ. .1 Eq(9. Furthermore when (9. if and only if (9. 1 + φ + rψ It now follows after some calculation that the following result is true.12) 9.1 we obtain the following result. Here we make some simple observations about linearized stability. . By employing the linearized stability Theorem 1. Lemma 9.12) holds. 3)-Type Equations This equation has not been investigated yet.3. . .150 Chapter 9. .3. Eq(9. t+ r 1−r . . φ. . φ.12) holds. be a period-two cycle of Eq(9. and more precisely a saddle point. .1 Let Existence and Local Stability of Period-Two Cycles . when q > 1 and (r − 1)(q − 1)2 + 4pr2 < 0. . φ. (9.11) is unstable. (b) The equilibrium y of Eq(9. Then φ= p + qφ 1 + ψ + rφ and ψ = p + qψ . invariant intervals.1.11) 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 (a) The equilibrium y of Eq(9.3. . 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 = 0.11) has a prime period-two solution . . ψ. and the convergence of solutions in certain regions of initial conditions. φ. ψ.11).11) is locally asymptotically stable when either q≤1 or q > 1 and (r − 1)(q − 1)2 + 4pr2 > 0. (2. Theorem 9. ψ.

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

(2. ψ) = + . 3)-Type Equations ∂h (p + qψ)(pr − q − qψ) (φ. ψ) + (φ.14).16) First we will establish Inequality (9. Inequality (9.1 that both eigenvalues of JT 2 unit disk |λ| < 1. ψ) (φ. This inequality is equivalent to (q − pr + qφ) (q − pr + qψ) − (1 + ψ + rφ)2 (1 + φ + rψ)2 < 0.1. ψ) = .152 Chapter 9. (9. ∂u ∂v ∂v ∂u φ ψ lie inside the Then it follows from Theorem 1. 2 (1 + φ + rψ)2 ∂v (1 + ψ + rφ) (1 + φ + rψ)2 Set S= and D= ∂h ∂g (φ.13) is equivalent to the following three inequalities: D<1 S <1+D −1 − D < S. ψ) ∂u ∂v ∂h ∂g ∂h ∂g (φ.13) (9. ψ) − (φ. if and only if |S| < 1 + D < 2. ψ). ψ) (φ. Observe that.15) (9. Thus .14) (9. ∂u (1 + ψ + rφ)2 (1 + φ + rψ)2 ∂h (p + qψ)(p + qφ) q − pr + qφ (φ. (q − pr + qφ) (q − pr + qψ) = (q − pr)2 + q(q − pr)(φ + ψ) + q 2 φψ = = −2q 2 r + q 2 r2 + 3qpr2 − 2qpr3 − p2 r3 + p2 r4 − q 3 + rq 3 + q 2 − pr2 q 2 − qpr r (−1 + r) (1 + ψ + rφ)2 (1 + φ + rψ)2 = 1 + φ + rψ + ψ + ψφ + rψ 2 + rφ + φ2 r + r2 φψ = −pr2 + qr + pr − q + q 2 r 2 2 and that .

3.12).12). Finally Inequality (9. we have established the following result: . In summary.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. 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 (9.15). r2 (1 − r) −r2 p + rp − q + q 2 + qr = rp(1 − r) + q(q − 1) + qr > 0. 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) = . After some lengthy calculation one can see that this inequality is also a consequence of Condition (9. Next we turn to Inequality (9.9. Also by using Condition (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.

ψ. r q when pr ≥ q + q2 . 3)-Type Equations Theorem 9. r when 1 < q < pr < q + q2 q − . (b) pr − q q .3.11) has a unique prime period-two solution . φ.11) possesses the following invariant intervals: (a) 0. . ψ. this period-two solution is locally asymptotically stable. . 1 + yk + qyk−1 1 + ryk−1 The proof follows by induction. q−1+ (q − 1)2 + 4pr 2r when pr ≤ q.11). yk ∈ [0.2 Invariant Intervals The following result. (2. Lemma 9. if and only if (9. r r Proof. Using Eq(9. 9.2 Eq(9.12) holds.3. Furthermore when (9. .11) we see that when yk−1 . then yk+1 = p + qyk−1 p + qyk−1 ≤ = g(yk−1 ) ≤ g(b) ≤ b. which can be established by direct calculation. b]. gives a list of invariant intervals for Eq(9.2 Eq(9. .3. (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. q r (c) q pr − q . . . .12) holds.154 Chapter 9. φ.

q p + qy 1 + x + ry . (c) It is clear that the function f (x. y) = p + qy 1 + x + ry pr−q q . r The proof follows by induction. r By using the fact that f (x. q r q = . < r + (r + 1)q q The proof follows by induction. q r pr−q .r q is increasing in y for x > pr−q . yk ∈ q then q pr − q p + qyk−1 pr − q = f (yk .9. yk−1 ) ≤ f 1 + yk + ryk−1 pr − q q . Now by using the monotonic character of the function f (x. yk ∈ p + qyk−1 = f (yk . ≥ yk+1 = 1 + yk + ryk−1 r q q . Using Eq(9. . 2 . yk−1 ) ≥ f 1 + yk + qyk−1 pr − q pr − q . y) and the condition 2 q < pr < q + qr we obtain yk+1 = p + qyk−1 = f (yk . yk−1 ) ≥ f . y) = is decreasing in y for x < q pr−q . then yk+1 = Using Eq(9. The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 155 (b) It is clear that the function f (x. q q q = .11) we see that when yk−1 . y) is decreasing in both arguments and the condition pr > q + rq2 we obtain yk+1 = q q p + qyk−1 = f (yk . 1 + yk + qyk−1 r r = pr + q 2 pr − q . yk−1 ) ≤ f .3.11) we see that when yk−1 .

3.3.4.3 Assume that pr ≤ q and that one of the following conditions is also satisﬁed: (i) q ≤ 1. .11): Theorem 9. the function f decreases in x and increases 2r in y.3. for all n ≥ 0.8) we can obtain some convergence results for the solutions with initial conditions in the invariant intervals.6. q−1+ (q − 1)2 + 4pr 2r converges to the equilibrium y. y0 ∈ 0. . It is easy to check that y ∈ 0.4 (a) Assume that 1 < q < pr < q+ qr − q and that one of the following r conditions is also satisﬁed: (i) r ≥ 1.4. 3)-Type Equations 9. Then every solution of Eq(9.3. For example.3 Convergence of Solutions p + qy 1 + x + ry By using the monotonic character of the function f (x.2. The result now follows by employing Theorem 1. 2r 2r √ q−1+ (q−1)2 +4pr and that in the interval 0. (ii) r ≥ 1. 2 q−1+ √ (q−1)2 +4pr (ii) r < 1 and (r − 1)(q − 1)2 + 4pr2 ≥ 0. (iii) r < 1 and (r − 1)(q − 1)2 + 4pr2 ≥ 0. 2r √ √ q−1+ (q−1)2 +4pr q−1+ (q−1)2 +4pr yn ∈ 0.3. y) = in each of the intervals in Lemma 9. together with the appropriate convergence theorem (from among Theorems 1.156 Chapter 9. 2 Theorem 9. (2.4.11) with initial conditions in the invariant interval 0. the following results are true for Eq(9. In view of Lemma 9.2 we see that when y−1 . then Proof.5-1.

q r converges to the equilibrium y.2 we conclude that yn ∈ for all n ≥ 0. r q converges to the equilibrium y. The Case β = 0 : xn+1 = α+γxn−1 A+Bxn +Cxn−1 157 Then every solution of Eq(9. we obtain the following global asymptotic stability result. 1.4 Global Stability By applying Theorem 1. q r interval q .r q 2 check that y ∈ and that the function f (x.9. q2 r q2 . It is easy to 9. The result now follows by employing Theorem 1. . r 9. .11) with initial conditions in the invariant interval q pr − q . Then every solution of Eq(9. (b) Assume that pr > q + qr .11) with initial conditions in the invariant interval pr − q q .3. pr−q and that in the r q in both x and in y. Proof. r2 + r2 yn + r3 yn−1 n = 0. . Theorem 9. the function f decreases check that y ∈ q .4.4. In and the only positive equilibrium is y = q .3. 2 q pr−q .3. pr−q r q for all n ≥ 0.11) when pr = q + this case Eq(9.7.11) is globally asymptotically stable.r q pr−q q . The result now follows by employing Theorem 1.5 Assume r ≥ 1 and pr ≤ q.5 Semicycle Analysis When pr = q + Here we present a semicycle analysis of the solutions of Eq(9.2 we conclude that yn ∈ pr−q q .3. y) decreases in x and increases in y. (a) In view of Lemma 9.3.3.11).4. r .6.11) becomes yn+1 = qr + q 2 + qr2 yn−1 . It is easy to (b) In view of Lemma 9. Then the positive equilibrium of Eq(9.3 to Eq(9.

γ q= B . (9. C and r = A . r + qyn + yn−1 β .158 2 Chapter 9. Then after the ﬁrst semicycle.3.11). Eq(9. 3)-Type Equations Theorem 9. q+1 (9. invariant intervals.3. and the convergence of solutions in certain regions of initial conditions.6 Suppose that pr = q + qr and let {yn }∞ n=−1 be a nontrivial solution of Eq(9. γ This equation has not been investigated yet.1. .17) are the nonnegative solutions of the equation y= (p + 1)y . . C n = 0. r r r (1 + yn + ryn−1 ) 2 9.1 and Theorem 1. .17) also possesses the unique positive equilibrium y= p+1−r .18) The following theorem is a consequence of Theorem 1.3) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= pyn + yn−1 . (2. Here we make some simple observations about linearized stability. .4 The Case α = 0 : xn+1 = βxn +γxn−1 A+Bxn +Cxn−1 This is the (2. 3)-type Eq(9. The proof follows from the relation yn+1 − q q qr2 = ( − yn ) 3 . The equilibrium points of Eq(9. r + (1 + q)y Hence zero is always an equilibrium point and when p + 1 > r. Proof.17) γ yn . 1. the solution oscillates about the equilibrium y with semicycles of length one.1.

ψ. φ. .4.2 Assume that (9. Theorem 9. (b) Assume that p + 1 > r. The Case α = 0 : xn+1 = Theorem 9.18) holds.5 that the following result is true. and unstable (a saddle point) when q + r > 3p + 1 + qr + pq.17) is locally asymptotically stable when q + r < 3p + 1 + qr + pq.1 βxn +γxn−1 A+Bxn +Cxn−1 159 (a) Assume that p + 1 ≤ r. Then the zero equilibrium of Eq(9. 1.17). . (9.3 Eq(9. Furthermore the zero equilibrium is a saddle point when 1−p<r <1+p and a repeller when r < 1 − p.9.4. .17) is globally asymptotically stable. . Then the zero equilibrium of Eq(9.17) is unstable. . φ.1.20) 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) = 0.19) Concerning prime period-two solutions for Eq(9. Then the positive equilibrium of Eq(9.4. (p + 1)(q + 1) (p + 1)(q + 1) n = 0. .20) (9. q−1 . it follows from Section 2. if and only if (9. Furthermore when (9.1. . ψ. .17) has a prime period-two solution . . Theorem 9.17) about its positive equilibrium y is zn+1 − p − q + qr q−p+r zn − zn−1 = 0.4. The linearized equation associated with Eq(9.20) holds. The following result is a consequence of Theorem 1.

K2 −p [0.1 Invariant Intervals The following result. ((q − r)(q 2 − p))2 + 4p3 qr2 2pqr (q − r)(q 2 − p) + . (b) 0.4. gives a list of invariant intervals for Eq(9.4. when q 2 − qr ≤ p < q 2 and p + q > r. where K > 0 is an arbitrary number and pr .1 Eq(9. and [0.17) possesses the following invariant intervals: (a) 0. which can be established by direct calculation. p+q−r q+1 when p = q 2 and q 2 + q > r where K > 0 is an arbitrary number and qr . K] when p = q 2 where K > 0 is an arbitrary number. when p > q 2 + r where ((r − q)(p − q 2 ) + qr)2 + 4q 3 r(p − q 2 ) 2q(p − q 2 ) .160 Chapter 9. K1 p − q2 K1 = (c) 0. q2 where K2 = when q 2 > p + qr.17). K] when q 2 < p ≤ q 2 + r and p + q ≤ r. (2. Lemma 9. q2 pr −p (q − r)(p − q 2 ) − qr + [0. K] when q 2 − qr ≤ p < q 2 and p + q ≤ r. qr p − q2 when q 2 < p ≤ q 2 + r and p + q > r. and q 2 + q ≤ r. 3)-Type Equations 9. .

K2 q 2 −p Second assume that p < q 2 − qr. yk+1 = f (yk . qr Hence for yk−1 . −p Then one can see that for yk−1 . p−q 2 0 ≤ f (x. while when q 2 + q > r the inequality is true when p+q−r K≥ . K) ≤ K is equivalent to q 2 +q−r ≤ (q+1)K. 1 ≤ K. q+1 (b) It is clear that the function f (x. First assume that q 2 < p ≤ q 2 + r . Now observe that the inequality f (K. K2 ≤ yk+1 = f (yk . yk ∈ f one can see that ≤ K2 . K) = (p + q)K Second assume that p > q 2 + r. 2 p−q p − q2 pr . q 2 −p First assume that q 2 − qr ≤ p < q 2 . K1 ≤ yk+1 = f (yk . yk−1 ) ≤ f (K. yk−1 ) ≤ f (K1 . Then for yk−1 . r + (q + 1)K pr . y) is increasing in both arguments for y < and it is increasing in y and decreasing in x for y ≥ q2pr . y) is increasing in both arguments for x < qr and it is increasing in x and decreasing in y for x ≥ p−q2 . K) = (p + q)K with K = pr . q 2 −p (c) It is clear that the function f (x. K). K1 p−q 2 we can see that qr qr .9. 1 ≤ K. y) ≤ f (K. 2 pr pr . yk ∈ f qr . 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. q2pr we obtain −p yk+1 = f (yk . ) ≤ K1 . this inequality is satisﬁed for any K > 0. r + (q + 1)K qr where we set K = p−q2 . y) = r + qx + y is increasing in both arguments. p−q2 we obtain qr . Therefore for any K > 0. yk ∈ 0. The last inequality is always satisﬁed when p + q ≤ r and when p + q > r. yk ∈ 0.4. When q 2 + q ≤ r. yk−1 ) ≤ f (K. 2 q2 − p q −p . Then for yk−1 . The Case α = 0 : xn+1 = Proof. yk−1 ) ≤ f K2 .

1 and Theorem 1.1. For example. Assume that p = q 2 and q 2 + q ≤ r. Then every solution of Eq(9. converges to the equilibrium y. the following results are true for Eq(9.5-1. Assume that q 2 + r ≥ p > q 2 and p + q ≤ r. Then every solution of Eq(9.4. converges to the equilibrium y.17) with initial conditions in the invariant interval 0. Proof.17): Theorem 9.8. K]. Then every solution of Eq(9.2 Convergence of Solutions px + qy r + qx + y By using the monotonic character of the function f (x. converges to the equilibrium y.162 Chapter 9. y) = in each of the intervals in Lemma 9.4. we can obtain some convergence results for the solutions with initial conditions in the invariant intervals.4.17) with initial conditions in the invariant interval 0.17) with initial conditions in the interval [0. Then every solution of Eq(9. Then every solution of Eq(9. for any K > 0. for any K > 0.4 (a) Assume that p = q 2 and q 2 + q > r. K].17) with initial conditions in the interval [0. for any K > 0. K].4. 2 q2 pr −p qr p − q2 q2 + q − r q+1 .4. (c) Assume that q 2 − qr ≥ p < q 2 and p + q > r.8).17) with initial conditions in the interval [0. 3)-Type Equations 9. converges to the equilibrium y.17) with initial conditions in the invariant interval 0.4.4. (b) Assume that q 2 + r ≥ p > q 2 and p + q > r. Then every solution of Eq(9. converges to the equilibrium y. The proof is an immediate consequence of Lemma 9. converges to the equilibrium y. (2. together with the appropriate convergence theorem (from among Theorems 1. Assume that q 2 − qr ≤ p < q 2 and p + q ≤ r.

p−q 2 Proof. C ∈ (0. 1−q 163 and p ≤ Then every solution of Eq(9.6 Assume that q 2 > p + qr and that Condition (9.6. Open Problems and Conjectures Theorem 9. The proof is a consequence of Lemma 9. ∞). A.1 and Theorem 1. Conjecture 9. Show that every positive solution of Eq(9.1 and Theorem 1.5. C ∈ (0.9.2) is bounded.5. Show that every positive solution of Eq(9.5 Assume that p > q2 + r and that one of the following conditions is also satisﬁed: (i) p ≤ q + r.11) with initial conditions in the invariant interval qr .4.5.5.2 Assume α. γ.3 Assume β. Proof. 2 9. The proof is an immediate consequence of Lemma 9. ∞). C ∈ (0. B. (iii) q < 1 qr . (ii) q ≥ 1. Conjecture 9. K2 converges to the equilibrium y. ∞). 2 Theorem 9.4. A. Conjecture 9. Then every solution of Eq(9.4. K1 p−q 2 3q 2 +q−qr+r .4.1) has a ﬁnite limit. A. .20) is not satisﬁed.3) is bounded. B.5.4.11) with initial conditions in the invariant interval converges to the equilibrium y.4.1 Assume Show that every positive solution of Eq(9.5 Open Problems and Conjectures α. β. B. γ.

Open Problem 9. B.5.3 Determine the set G of all initial points (x−1 . and for these initial points investigate the long-term behavior of the solutions {xn }∞ .2) has no prime period-two solutions. Suppose that Eq(9.164 Conjecture 9. Show that the positive equilibrium of Eq(9. ∞) and suppose that Eq(9.5.5 Assume that x−1 . A. x0 ) ∈ R × R through which the equation 1 + xn xn+1 = 1 + xn + xn−1 is well deﬁned for all n ≥ 0. Show that the positive equilibrium of Eq(9. γ. ∞) with x−1 + x0 > 0 and that β.5.1 Determine the set G of all initial points (x−1 . x0 ∈ [0.5. A. B.2 Determine the set G of all initial points (x−1 . Extend and generalize. C ∈ (0. Extend and generalize.3) is globally asymptotically stable.4 Assume Chapter 9. and for these initial points investigate the long-term behavior of the solutions {xn }∞ . ∞).3) has no prime period-two solutions. n=−1 Open Problem 9. and for these initial points investigate the long-term behavior of the solutions {xn }∞ . x0 ) ∈ R × R through which the equation 1 + xn−1 xn+1 = 1 + xn + xn−1 is well deﬁned for all n ≥ 0. (2.2) is globally asymptotically stable. n=−1 Open Problem 9. 3)-Type Equations α. n=−1 . γ. Extend and generalize.5. C ∈ (0. x0 ) ∈ R × R through which the equation xn + xn−1 xn+1 = 1 + xn + xn−1 is well deﬁned for all n ≥ 0. Conjecture 9.

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 . 1. . . 1. . . ψ.5. . . p = lim pn . φ.6 Assume that (9. .5 Assume that {pn }∞ . and {rn }∞ are period-two sen=0 n=0 quences of nonnegative real numbers. (9. Open Problem 9. . . .4 Assume that {pn }∞ . . ψ. φ. Extend and generalize.6 Assume that (9. ψ. .23). . .5.5. n→∞ q = n→∞ qn .5. {qn }n=0 . Open Problem 9. . . n = 0. 1. Open Problems and Conjectures 165 Open Problem 9.17). .23) pn yn + yn−1 . φ. 1 + yn + rn yn−1 pn + qn yn−1 . ψ. ψ.22) (9. Investigate the basin of attraction of the two cycle . . . Conjecture 9. n = 0.11). and {rn }∞ are convergent sen=0 n=0 n=0 quences of nonnegative real numbers with ﬁnite limits. . ψ. of Eq(9. {qn }∞ . of Eq(9. φ.12) holds. .5. . φ.20) holds.5. Show that the period-two solution . of Eq(9. φ. . . . . . Investigate the basin of attraction of the two cycle .21)-(9.17) is locally asymptotically stable. rn + qn yn + yn−1 ∞ Open Problem 9. lim and r = n→∞ rn .9.7 Assume that (9. 1 + yn + rn yn−1 n = 0. . Investigate the global character of all positive solutions of Eqs(9.21) (9.20) holds. .

8 Assume that p. . . q. 1 + yn + ryn−k pyn + yn−k .}.5. 1. . r + qyn + yn−k n = 0. 1 + yn + ryn−k yn+1 = yn+1 = p + qyn−k . n = 0.166 Open Problem 9.24) yn+1 = . . . . 1. . .26) . . (9. 1. . (2. r ∈ [0. Investigate the global behavior of all positive solutions of each of the following diﬀerence equations: p + qyn (9. . 3. . . ∞) and Chapter 9.25) (9. n = 0. 3)-Type Equations k ∈ {2.

1. and the denominators are always positive. 1 + yn αB . 2)-type Eq(10. (10. . . 2)-Type Equations 10. n = 0. . . . . . 1.1 Introduction α + βxn + γxn−1 . A + Cxn−1 Eq(1) contains the following three equations of the (3. .2 The Case C = 0 : xn+1 = A yn B α+βxn +γxn−1 A+Bxn This is the (3. A + Bxn α + βxn + γxn−1 . (10. γ .4) and r = 167 .3) Bxn + Cxn−1 Please recall our classiﬁcation convention in which all parameters that appear in these equations are positive. . A n = 0.2) 10.1) (10. n = 0. 2)-type: xn+1 = xn+1 = and α + βxn + γxn−1 . A (10.Chapter 10 (3.1) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + qyn + ryn−1 . the initial conditions are nonnegative. . . xn+1 = n = 0. . 1. . A2 q= β . 1.

7.4) possesses prime period-two solutions. 2)-Type Equations (Eq(10.4) exhibit the following trichotomy character. See Sections 2.2.2.5) . Then the equilibrium of Eq(10. Then Eq(10. Then every solution of Eq(10.7. for a more general equation. (3.4) converges to a period-two solution. .7) (10.) The linearized equation of Eq(10. .4) is globally asymptotically stable.4) possesses unbounded solutions.4) about its unique equilibrium y= is zn+1 − q+r−1+ (q + r − 1)2 + 4p 2 q − p − ry r zn − zn−1 = 0.4) was investigated in [29].1 and 10. The main result in this section is that the solutions of Eq(10. (c) Assume that r > q + 1. When r =q+1 Eq(10.168 Chapter 10. Theorem 10.1 we see that y is locally asymptotically stable when r <1+q and is an unstable (saddle point) equilibrium when r > 1 + q.2. The proofs of parts (b) and (c) of Theorem 10.1 was given.1.2 below. 2 (1 + y) 1+y n = 0.6) (10. (10. The proof of part (a) of Theorem 10.2. 1. in Section 2. (b) Assume that r < q + 1.1 (a) Assume that r = q + 1.2.5 and 2. By applying the linearized stability Theorem 1. .1 are given in Sections 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. .2.2. 2)-type Eq(6. (10.5 and therefore the proof is complete when (10.12) (10.8) holds. .10.23) which was investigated in Section 6.8) which is of the form of the (2. So in the remaining part of the proof we assume that p < q. Observe that (p − q) + ryn−1 yn+1 = q + .4) is globally asymptotically stable.6) holds and we prove that the equilibrium of Eq(10.6) holds the equilibrium is locally asymptotically stable so it remains to be shown that every solution of Eq(10. (10.9) for n ≥ 1.4) satisfy the following identity yn+1 − q = Note that when q= that is when p + qr − q = 0 the Identity (10.10) .4) is attracted to the equilibrium. 1. The Case C = 0 : xn+1 = α+βxn +γxn−1 A+Bxn 169 10.11) q−p r yn−1 − 1 + yn r q−p r for n ≥ 0. . 1 + yn and so when p≥q we have yn ≥ q Therefore the change of variables yn = q + zn transforms Eq(10. Now observe that the solutions of Eq(10. . (10.1 Here we assume that Eq(10.2.1 Proof of Part (b) of Theorem 10.4) to the diﬀerence equation zn+1 = (p + rq − q) + rzn−1 . (1 + q) + zn n = 1. n = 0. We know already that when Eq(10. .

10) that yN +1+2K > q for all K ≥ 0. To complete the proof in this case it is suﬃcient to show that eventually (10. Then we can see from (10. (10.15) yn ≥ q and then use the known result for Eq(10.13) (10.14) First assume that (10. It follows from (10.17) holds is similar and will be omitted. To this end.9). (10.17) q−p r for some N ≥ 0. We will assume that (10. note that now q−p <q r and employ (10. 2)-Type Equations |yn+1 − q| < r|yn+1 − q| and so n→∞ lim yn = q which completes the proof when (10.10). q−p r q−p r for n ≥ 0 for n ≥ 0.13) holds. Still to be considered are cases where p + qr − q > 0 and p + qr − q < 0. (3.4) is equal to q.16) (10.170 and the equilibrium of Eq(10.11) holds.18) q−p q−p r y2n − > r y2n − 1 + yn+1 r r . If (10.12) that Chapter 10.15) is not eventually true then either y2n < or y2n−1 < This is because when yN −1 > then it follows from (10. (10.10) that y2n+2 − q = and so y2n+2 > ry2n + p for n ≥ 0.16) holds. The case where (10.

13) and completes the proof when (10.2.10.2. + rp + p r and so p q−p ≥ r 1−r which contradicts (10. The Case C = 0 : xn+1 = α+βxn +γxn−1 A+Bxn 171 This is impossible when r ≥ 1. . it follows from (10.8 and the fact that in this case the function f (x.14) and completes the proof of part (b) of Theorem 10. To this end assume for the sake of contradiction that the solution is not eventually in the interval 0.1. Note that in this case r < 1 and so q−p < y2n+2 < rn+1 y0 + rn p + . Next assume that (10. q−p r q−p < r y2n − y2n − 1 + y2n+1 r r . Then from (10. q−p . Once we establish this result.20) r We will assume that (10. because by (10.10) and r the fact that now q−p q< .4) lies eventually in the interval 0. Then one can see from (10. + rp + p r which implies that p q−p ≤ . y2n−1 > (10. Now we claim that every solution of Eq(10.4.14) holds.20) holds is similar and will be omitted. r 1−r This contradicts (10. the proof that the r equilibrium is a global attractor of all solutions would be a consequence of Theorem 1. r that either q−p y2n > for n ≥ 0 (10.16) and (10.19) r or q−p for n ≥ 0. y) = p + qx + ry 1+x is increasing in both arguments.18) that q−p > y2n+2 > rn+1 y0 + rn p + .13) holds.16) the solution is bounded.19) holds. On the other hand. . q−p . . when r < 1.10) we see that y2n+2 − q = and so y2n+2 < ry2n + p for n ≥ 0. . The case where (10.

A2 and r = .172 Chapter 10.4) possesses unbounded solutions. To this end observe that yn+1 = q + q p + r(yn−1 − r ) 1 + yn for n ≥ 0 and so when (10.4) is a saddle point and so by the stable manifold theorem. 1. That is when y−1 . It should be mentioned that when (10. .9) and the conclusion of part (c) follows from Section 6. A (10. .5 for appropriate initial conditions y−1 and y0 .2) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= p + qyn + ryn−1 . in addition to unbounded solutions. Eq(10. 2)-type Eq(10. γ .7) holds and show that Eq(10. y0 ∈ [q.7) holds the interval [q. ∞). ∞) is invariant.2 Proof of Part (c) of Theorem 10. have the property that the subsequences of even terms converge to ∞ while the subsequences of odd terms converge to q.4) with initial conditions such that q ≤ y−1 < r − 1 and y0 > r−1 + p + q(r − 1) r−1−q for n ≥ 0.2. 1 + yn−1 q= β .4).1 Here we assume that (10.4) possesses bounded solutions which in fact converge to the equilibrium of Eq(10.7) holds. (3. . A n = 0.3 The Case B = 0 : xn+1 = A yn C α+βxn +γxn−1 A+Cxn−1 This is the (3. the equilibrium of Eq(10.2. 10.4) to Eq(10.21) αC . 2)-Type Equations 10. then yn ≥ q Now the change of variables yn = q + zn transforms Eq(10. More speciﬁcally the solutions of Eq(10.

zn + 1+y (1 + y)2 n = 0. p 1−q when q < 1 and p ≥ r.1 we see that y is locally asymptotically stable for all values of the parameters p.1 Eq(10.1 Invariant Intervals The following result. which can be established by direct calculation. q. for some integer N ≥ 0.10. . . yN +1 ∈ I then yn ∈ I for n ≥ N.2 Here we discuss the behavior of the solutions of Eq(10. 1. and r. . r−p q when q < 1 and p ≥ r(1 − q). 10.3.3.3. Recall that I is an invariant interval.1.21) possesses the following invariant intervals: (a) 0. Lemma 10. (b) 0. invariant intervals. The linearized equation of Eq(10.21) when p ≥ r. . . gives a list of invariant intervals for Eq(10. The Case B = 0 : xn+1 = α+βxn +γxn−1 A+Cxn−1 173 This equation has not been investigated yet.3. if whenever. 1. yn+1 = r + 1 + yn−1 and so when p≥r Global Stability When p ≥ r . yN . .21) about its positive equilibrium y= is zn+1 − q+r−1+ (q + r − 1)2 + 4p 2 q p − r + qy zn = 0.21). By applying Theorem 1. 10. Observe that (p − r) + qyn . . n = 0. and convergence of solutions in certain regions of initial conditions. Here we make some simple observations about linearized stability.

.3. 2. converges to the equilibrium y. 1−q Then every solution of Eq(10. Theorem 10.2 (a) Assume that q < 1 and r ≥ p . 2)-Type Equations for n ≥ 1. Then the following statements are true. n = 1.174 yn ≥ r Therefore the change of variables yn = r + zn transforms Eq(10. 2)-type Eq(6. (1 + r) + zn−1 Chapter 10.3. (b) The positive equilibrium of Eq(10.21) to the diﬀerence equation zn+1 = p + qr − r + qzn . (a) Every positive solution of Eq(10. 10.3. which is of the form of the (2.21) with initial conditions in the invariant interval 0. . Theorem 10.21) is globally asymptotically stable if one of the following conditions holds: (i) q < 1.1 Assume p ≥ r.21) is bounded. (3. r−p q . The following result is now a consequence of the above observations and the results in Section 6.3. (ii) q≥1 and either p + qr − r ≤ q or q < p + qr − r ≤ 2(q + 1). .3.21) when p < r.2) which was investigated in Section 6.3 Convergence of Solutions Here we discuss the behavior of the solutions of Eq(10.

4. It is easy to check that y ∈ r−p . αC . (a) In view of Lemma 10.10.1 we see that when y−1 .4. y0 ∈ for all n ≥ 0.4 The Case A = 0 : xn+1 = γ yn C α+βxn +γxn−1 Bxn +Cxn−1 This is the (3. C n = 0. γ2 (10.21) with initial conditions in the invariant interval r−p . y) increases in both x and y.4. The result is now a consequence of Theorem 1.22) β .K q r−p . then yn ∈ 0. The result is now a consequence of Theorem 1. .8. r−p and that in the interval 0. .5. r−p .3) which by the change of variables xn = reduces to the diﬀerence equation yn+1 = where p= r + pyn + yn−1 .3. γ and r = . It is easy to check that y ∈ 0. 2)-type Eq(10. . then yn ∈ r−p . y0 ∈ 0.K q converges to the equilibrium y.K q the function f (x. y) increases in x and decreases in y. r−p q q for all n ≥ 0.3. r−p q q the function f (x. qyn + yn−1 q= B . Proof. 2 10.K q and that in the interval . 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 . r−p .K q pq + r2 − pr . 1−q and p < r < Then every solution of Eq(10. K= q + r − p − q2 (b) In view of Lemma 10.1 we see that when y−1 .

φ.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. When r = 0. The second condition reduces to q > pq + 1 + 3p. the inequality in (10.9. as in Section 6. Here we make some simple observations about linearized stability. .23) is strict and Condition (10. invariant intervals.24) 2r q − pq − 1 − 3p >0 (10.24) is void. q > 1 and 4r < (1 − p)(q − pq − 3p − 1).176 Chapter 10. 1. (3. . ψ.25) Furthermore in this case the prime period-two solution . (10. q−1 . The linearized equation of Eq(10. . φ. (10. ψ. . (1 + p) + (1 + p)2 + 4r(1 + q) 2(1 + q) By applying Theorem 1. . it follows from Section 2.23) It is interesting to note that the above condition for the local asymptotic stability of y requires the assumption that r > 0. . and the convergence of solutions in certain regions of initial conditions.1. (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. . 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. . 2)-Type Equations This equation has not been investigated yet. . Concerning prime period-two solution.5 that a period-two solution exists if and only if p < 1. .

and p < q. b) ≤ f (x. Hence a≤ r + (p + 1)b r + (p + 1)a = f (b.4.4. (q + 1)b (q + 1)a qr . (a) It is easy to see that when p = q the function f (x.4. p−q q (c) 1. y) ≤ f (a.26) when p > q and p2 − pq < r. q p−q qr p .1 q−p Proof.1 Invariant Intervals Here we present some results about invariant intervals for Eq(10. r q−p r . y) is decreasing in both arguments when y < qr is increasing in x and decreasing in y for y ≥ p−q . Lemma 10. p−q and (10. First assume that p > q and p2 −pq q2 and it < r. a) = ≤ b. p > q. y) = is decreasing in both arguments.21) possesses the following invariant intervals: (a) [a. The Case A = 0 : xn+1 = α+βxn +γxn−1 Bxn +Cxn−1 177 10.21). r + px + y qx + y (b) Clearly the function f (x. when q > p + r.1 Eq(10. q2 when p > q and when p < q < p + r. q2 p2 − pq > r. b] when p = q a and b are positive numbers such that r + (p + 1)a ≤ (q + 1)ab ≤ r + (p + 1)b.10. We consider the cases where p = q. (b) p qr . .

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

2 Theorem 10.2 Assume that p > q. (c) Assume that p < q < p+r. r < p2 −pq .4. Then every solution of Eq(10.4.4.22) with initial conditions in the invariant interval r . Theorem 10.3 Assume that p < q. p + r < q.4.22) with initial conditions r in the invariant interval 1.10.2 Convergence of Solutions Here we obtain some convergence results for Eq(10.4.4.4. The proof is an immediate consequence of Lemma 10. 2 Theorem 10. 2 . (b) Assume that p > q and r > p2 −pq . Then every solution of Eq(10.22) with initial conditions in the invariant interval [a. Proof. q p−q conditions in the invariant interval converges to the equilibrium y. The proof is an immediate consequence of Lemma 10.5.25) is not satisﬁed. q−p converges to the equilibrium y.4. q2 and that either p≤1 or p > 1 and (q − 1)[(p − 1)2 (q − 1) − 4q(1 − p − qr)] ≤ 0.22) with initial conditions in the invariant interval converges to the equilibrium y. b].22).7. p−q q Proof. Then every solution of Eq(10. The proof is an immediate consequence of Lemma 10. and that Condition (10.1 and Theorem 1.1 and Theorem 1.4.22) with initial p qr .26) converges to the equilibrium y. qr p .1 (a) Assume that p = q. q−p Proof.6. Then every solution of Eq(10.1 and Theorem 1.4. and r <1 q−p 2 10.4. where 0 < a < b satisfy (10. q2 Then every solution of Eq(10. The Case A = 0 : xn+1 = The inequalities α+βxn +γxn−1 Bxn +Cxn−1 179 (p + r)(q − p) + r r ≥ q(q − p) + r q−p follow from the inequality q > p + r. 1 converges to the equilibrium y.

. . .4) converges to a not necessarily prime period-two solution . (10. (a) Determine the set of initial conditions y−1 and y0 for which φ = ψ. 2)-Type Equations 10.180 Chapter 10.30) xn+1 = xn + xn−1 Extend and generalize.28) 1 + xn 1 + xn + xn−1 xn+1 = (10. ∞) × (0. Conjecture 10.29) 1 + xn−1 1 + xn + xn−1 . .21) converges to the positive equilibrium of the equation.5. y0 ) ∈ (0.5. (3.27).3 For each of the equations listed below. y0 ) ∈ B. . ∞) such that the ∞ solutions {yn }n=−1 of Eq(10. . Investigate the asymptotic behavior of {yn }n=−1 .5. if and only if r = 1 + q. Open Problem 10. (10. ψ. (a) Find the set B of all initial conditions (y−1 .27) (b) Assume (10.5 Open Problems and Conjectures ∞ Open Problem 10.1 Every positive solution of Eq(10. determine the set G of all initial conditions (x−1 . and for these initial points investigate the long-term behavior of the corresponding solution: 1 + xn + xn−1 xn+1 = (10. Open Problem 10.5. x0 ) ∈ R × R through which the equation is well deﬁned for all n ≥ 0. ψ.1 We know that every solution {yn }n=−1 of Eq(10. Determine the set of initial conditions y−1 and y0 for which {yn }∞ n=−1 converges to (10. φ.2 Assume r > q + 1.4) are bounded. . φ.27) is a prime period-two solution of Eq(10. ∞ (b) Let (y−1 .4).

5. and {rn }∞ are period-two sen=0 n=0 n=0 quences of nonnegative real numbers.2 Assume that p. is locally asymptotically stable. when it exists. . 1 + yn pn + qn yn + rn yn−1 . . but not globally. Open Problems and Conjectures Conjecture 10. 1. (10. Investigate the global character of all positive solutions of Eqs(10. . . . qn yn + yn−1 n = 0. . p = lim pn . . 181 (b) When Eq(10. .5.22). and {rn }n=0 are convergent sen=0 n=0 quences of nonnegative real numbers with ﬁnite limits. {qn }∞ . q.22) (a) Local stability of the positive equilibrium implies global stability.33).31) yn+1 = yn+1 = n = 0. ∞ Open Problem 10.5 Assume that {pn }∞ . 1 + yn−1 rn + pn yn + yn−1 .4 Assume that {pn }∞ . ∞).22) has no prime period-two solutions the equilibrium y is globally asymptotically stable.33) Open Problem 10.5. 1. 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 . r ∈ (0. 1. (d) The period-two solution of Eq(10.10. n→∞ q = n→∞ qn . Extend and generalize.5.22) possesses a period-two solution. lim lim and r = n→∞ rn .31)-(10. (c) When Eq(10. . Then the following statements are true for Eq(10. n = 0. . the equilibrium y is a saddle point. (10. {qn }∞ .32) (10. .

. . q. 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 . . . 3. 1. . 1 + yn−k p + qyn + ryn−1 . .5. ∞) and k ∈ {2. 1. .6 Assume that Chapter 10.182 Open Problem 10. . 2)-Type Equations p. (3. 1. .}. 1 + yn p + qyn + ryn−k . .34) (10. .36) . . n = 0. . n = 0.35) (10. (10. qyn + yn−k n = 0. r ∈ [0.

The linearized equation about x is (see Section 2.3) zn+1 − (βA − Bα) + (βC − Bγ)x zn + α(B + C) + (B + C)(A + β + γ)x (γA − Cα) − (βC − γB)x = 0. 1. This equation has not been investigated yet.1) Please recall our classiﬁcation convention in which all coeﬃcients of this equation are now assumed to be positive and the initial conditions nonnegative.1 Linearized Stability Analysis (B + C)x2 − (β + γ − A)x − α = 0. 11.Chapter 11 The (3. A + Bxn + Cxn−1 n = 0. Eq(11. x= β+γ−A+ (A − β − γ)2 + 4α(B + C) 2(B + C) . 3)-type equation xn+1 = α + βxn + γxn−1 . (11. invariant intervals. . . − z 2 + α(B + C) + (B + C)(A + β + γ)x n−1 A A2 183 n = 0.2) . . Here we make some simple observations about linearized stability. . . 1. . 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. . (11.1) has a unique equilibrium which is the positive solution of the quadratic equation That is. .

.1) by testing whether the hypotheses of one of the four Theorems 1. when (11. .13).15) are satisﬁed. . φ.1). Conditions (2. then clearly fx = M = βC − γB. C C(B − C) .5. φ.184 Chapter 11. 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. ψ. 2 (A + Bx + Cy) (A + Bx + Cy)2 Recall that an interval I is an invariant interval for Eq(11. . yN +1 ∈ I then for some integer N ≥ 0. if. ψ. is discussed in Section 2. B > C. see Section 2. Eq(11.2 Invariant Intervals Here we obtain several invariant intervals for Eq(11. that is.32) are satisﬁed. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 The equilibrium x is locally asymptotically stable if Conditions (2. (11. γ > β+A. The (3.1) then we may be able to obtain a convergence result for the solutions of Eq(11.4) The local asymptotic stability of the two cycle 11.1) if whenever. . ψ. and α < (γ − β − A)[B(γ − β − A) − C(γ + 3β − A)] .4. . for n ≥ N.8 are satisﬁed.1) has a prime period-two solution . (11. .3) 4C 2 Furthermore. .5-1.31) and (2. φ. . φ.3). If we set L = βA − αB. These intervals are derived by analyzing the intervals where the function f (x. . yn ∈ I . ψ. y) = is increasing or decreasing in x and y. and N = γA − αC α + βx + γy A + Bx + Cy L + My N − My and fy = . . (2. Once we have an invariant interval for Eq(11. and (2.14).6.4. yN . .

(v) Assume that α β γ γA − αC > . (ii) Assume that β γ α > > B C A and γ N ≤ . A is an invariant interval for Eq(11. y) is decreasing in both variables. y) is increasing in both variables. y) is increasing in both variables. y) is decreasing in both variables. ≥ > βC − γB A B C γA−αC Then 0.1) and in this interval the function f (x.1) and in this interval the function f (x. A is an invariant interval for Eq(11.1) and in this interval the function f (x. (iii) Assume that β α γ > > C B A and β+γ αB − βA ≤ . C B A 185 α β Then A .1) and in this interval the function f (x. y) is increasing in x and decreasing in y. C M α γ Then A . A B C α Then 0. βC−γB is an invariant interval for Eq(11. B+C is an invariant interval for Eq(11. A is an invariant interval for Eq(11.1 (i) Assume that β α γ = > . (vii) Assume that α β γ = > .2. C B A β α Then B . Invariant Intervals Lemma 11. C is an invariant interval for Eq(11. . (iv) Assume that β α γ = < . y) is increasing in both variables. B+C βC − γB α β+γ Then A .11.1) and in this interval the function f (x. y) is decreasing in both variables. B is an invariant interval for Eq(11.2.1) and in this interval the function f (x. (vi) Assume that β γ α γA − αC . < < < B C A βC − γB α Then 0.1) and in this interval the function f (x.

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

α = f (0. A B+C β+γ β+γ . y) ≤ f (0.11. Thus. Thus assuming A ≤ M . which imply that N < 0. Thus assuming B+C ≤ − M . y) ≤ f (0. y) ≤ f A for all x. L < 0. in this case the function f is decreasing in y. this function α N N is decreasing in y for x < M . B+C . this function is increasing in x for y < − M . and N < 0. 0) = α for all x. Thus. ≤ f (x. 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. in this case the function f is increasing in y. we obtain. B A A A Here we used the fact that under the condition β α γ = < C B A the function f (u. y) ≤ f (0. B+C B+C ≤ β+γ . for every y. Thus assuming A ≤ − M . which in turn imply that N < 0. α . in this case the function f is decreasing in both variables for all values of x and y and α α β α ≤f . α L ≤− . in this case the function f is decreasing in x. Thus. we obtain 0 ≤ f (x.2. for every x. γ β α (iv) The condition C = B < A is equivalent to M = 0 and L < 0. − M . In addition. In β+γ L L addition. y ∈ 0. α L L this function is decreasing in x for y < − M . Thus. y ∈ 0. In addition. B+C B β γ α (v) The condition A > B > C is equivalent to M > 0 and L < 0. which in turn imply that N > 0. y ∈ α β+γ . A . u) ≥ β+γ β = . we obtain 0 ≤ f (x. for every y. 0) ≤ f (x. A M γ β α (vi) The condition A > C > B is equivalent to M < 0. A . u) is decreasing and consequently f (u. 0) = .

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

M ] if AM +(B+C)N < αM +βK+γN N N [ M . − M ] if AM −L(B+C) < − M αM −βL+γM L L [− M . . n→∞ . K1 ] β−A+ Signs of derivatives N M [0.11.4. . 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. M ] if AM +(B+C)N < N [ M . Theorem 11.2.1 Let {xn } be a solution of Eq(11. [0. fx < 0 if y < − M fy < 0 fx < 0 fy < 0 fx < 0 N N fy > 0 if x > M .1).3. 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 . fx < 0 if y > − M fy > 0 fx > 0 fy < 0 fx < 0 fy > 0 L L fx > 0 if y > − M .5-1. Then in each of the following cases lim xn = x. B ] γ [0. A ] √ L L fx > 0 if y < − M .4.8. ((γ − A)M − BN )2 + 4BM (αM + βN ) 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 . K2 ] if BM K+AM −γL > − M −CL β+γ−A+ β [0. .3 Convergence Results The following result is a consequence of Lemma 11.3. C ] αM −L(β+γ) αM −L(β+γ) L L [ AM −L(B+C) . − M ] if AM −(B+C)L < − M βM K+αM L L [− M . Convergence Results 189 Case 1 2 3 4 5 6 7 8 9 Invariant intervals αM +(β+γ)N N [0. fy < 0 if x < M αM +(β+γ)N αM +(β+γ)N N [ AM +(B+C)N . K3 ] if AM −BL+CMK > − M K α [0.1 and Theorems 1.

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

C and x−1 . . The proofs of statements (vii) and (viii) follow by applying Lemma 11. Theorem 1.11.4. The proofs of statements (iv-vi) follow by applying Lemma 11. C B A Then the equilibrium x of Eq(11.2.2. γ . 2 11.6. By applying Theorem 1.1) we obtain the following global asymptotic result.1 Global Stability Theorem 1. x0 ∈ 0.8.1 (i-iii) and Theorem 1.3.5. A + Bm + CM M= α + βM + γm A + BM + Cm is m = M .1 (ix) and (x) respectively.2 Assume that γ β α < min .4.2. x0 ∈ 0. m). .3.1) is globally asymptotically stable.4. γ .4. The proofs of statements (i-iii) follow by applying Lemma 11.2 and linearized stability imply the following global stability result: Theorem 11. and by observing that Eq(11. M = f (M. and by observing that the only solution of the system of equations m = f (m. Convergence Results (ix) Condition (11.2.1 (iv-vi) and Theorem 1. Theorem 1.7.1) does not possess a prime period-two solution. β γ α = < A B C (x) Condition (11. α γ β < ≤ B A C and x−1 .3) is not satisﬁed. that is.4.1 (vii) and (viii) respectively.3) is not satisﬁed.3 to Eq(11.4. M ). The proofs of statements (ix) and (x) follow by applying Lemma 11. C 191 Proof. m= α + βm + γM .3.

3 Assume that B ≤ C.2 Assume α. γB > βC.1) is a saddle point and the period-two solution is locally asymptotically stable.3. B. γ. C ∈ (0.4. Then the equilibrium x of Eq(11.4 Open Problems and Conjectures α. .1 Assume Show that every positive solution of Eq(11. β.1) is globally asymptotically stable. β.4 (a) Assume that αC ≥ γA.1) has no positive prime period-two solution. Show that every positive solution of Eq(11. Show that the positive equilibrium of Eq(11. Conjecture 11. Then the equilibrium x of Eq(11.4.1) is globally asymptotically stable. C ∈ (0.4. B.1) is bounded. C ∈ (0.1) converges to the positive equilibrium. Conjecture 11. Theorem 11. γ. and 3γB ≥ βC ≥ γB. γ.1) is globally asymptotically stable. γA ≥ αC. The (3. and 3βC ≥ γB ≥ βC. and γA > αC. The following global stability result is a consequence of linearized stability and the Trichotomy Theorem 1. A. 11. A. αB > βA. β. (b) Assume that αB ≥ βA.3. ∞) and that Eq(11.1) has a positive prime period-two solution. βA ≥ αB.3 Assume α.4.192 Chapter 11.4. Then the equilibrium x of Eq(11. A. ∞) and that Eq(11. B. ∞). 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Theorem 11. Conjecture 11.

∞). Ai ∈ [0. C ∈ (0. In particular.5) An + Bn xn + Cn xn−1 Open Problem 11. {Bn }∞ .4 Assume that (11.7) is periodic with period-k. Show that Eq(11. Obtain necessary and suﬃcient conditions so that every positive solution of Eq(11. .4.4. . Open Problems and Conjectures Conjecture 11. n = 0. A.6) holds. Investigate the asymptotic behavior and the periodic nature of all positive solutions of Eq(11.4.4. 1.11. ai . .1 Assume {αn }n=0 . ∞) for i = 0. . B. . Open Problem 11. 193 ∞ ∞ Open Problem 11. k ≥ 2. {βn }∞ .2 Assume that the parameters in Eq(11.3 Assume that a. {An }n=0 . investigate the cases where 2 ≤ k ≤ 8. . . {Cn }∞ n=0 n=0 n=0 n=0 are convergent sequences of nonnegative real numbers with ﬁnite limits. .1) cannot have a prime period-k solution for any k ≥ 3. . a + a0 xn + .4 Assume α. + Ak xn−k (11. . . .5). .5 Assume that (11. xn+1 = (11. Obtain necessary and suﬃcient conditions so that every positive solution of Eq(11. .4.6) holds. k.4. 1.5) are period-two sequences of positive real numbers. β. Investigate the asymptotic behavior and the periodic nature of all positive solutions of the diﬀerence equation αn + βn xn + γn xn−1 .6) Obtain necessary and suﬃcient conditions so that every positive solution of the equation xn+1 = has a ﬁnite limit. .4. {γn }∞ . . γ.7) converges to a period-two solution.7) Open Problem 11. . 1. A + A0 xn + . + ak xn−k . (11. Open Problem 11. A. n = 0.

Show that every positive solution of the diﬀerence equation xn+1 = xn + . . . δ ∈ [0. Extend and generalize. j ∈ {1. .4.) αxn + βxn−1 xn−2 . 1. 1.6 Let α. 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 . .5 Let k ≥ 2 be a positive integer. . 1.6 Let aij ∈ (0. + xn−k−2 converges to 1. . .4. the periodic nature.7 Assume that p.194 Chapter 11. Investigate the boundedness character. and the asymptotic behavior of all positive solutions of the diﬀerence equation xn+1 = (See [8] and [61]. Extend to equations with real parameters.4. and [60]. ∞). 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Open Problem 11. . (11. [47]. 2. . xn xn−1 + xn−2 + . ∞) with γ + δ > 0. .4.8) converges to a period-two solution. . n = 0. n = 0. (See [32] for the two dimensional case. q ∈ [0. β.) Conjecture 11.) Conjecture 11. 3}. . . γ. The (3. γxn xn−1 + δxn−2 Conjecture 11. ∞) for i. + xn−k + xn−k−1 xn−k−2 . . (See [8]. n = 0. .

Open Problems and Conjectures (a) Show that every positive solution of the equation xn+1 = pxn−1 + xn−2 . xn−1 + pxn−2 n = 0.4. .11. ∞).8 Assume that p ∈ (0. .) Conjecture 11. (c) Show that when p>1+q the equation possesses positive unbounded solutions. . . 195 converges to a period-two solution if and only if p = 1 + q.4. (a) Show that every positive solution of the equation xn+1 = xn−1 . .4. . (See [62]. q ∈ [0. 1. (b) Show that when p<1 the positive equilibrium x = 1 1+p is globally asymptotically stable. Conjecture 11. . ∞). (b) Show that when p<1+q every positive solution has a ﬁnite limit. converges to a period-two solution if p ≥ 1. 1. q + xn−2 n = 0.9 Assume that p.

(See [62]. converges to a period-two solution if and only if q = 1. .4. . xn−2 + q n = 0. xn n = 0.10 Show that every positive solution of the equation xn+1 = xn + xn−2 . 1. (iii) When p < 1 there exist positive unbounded solutions. (See [62].196 Chapter 11. . (c) Show that when q<1 the equation possesses positive unbounded solutions. . xn−1 n = 0. (ii) When p = 1 every positive solution converges to a period-ﬁve solution.) Conjecture 11. . has the following trichotomy character: (i) When p > 1 every positive solution converges to the positive equilibrium. . . (See [62]. . .) Conjecture 11.) . 1. 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 .11 Show that the diﬀerence equation xn+1 = p + xn−2 . The (3. (b) Show that when q>1 the positive equilibrium of the equation is globally asymptotically stable. 1. converges to a period-four solution.4.

. .7 Consider the diﬀerence equation xn+1 = α + βxn + γxn−1 + δxn−2 . (11.) . (See [62]. Then one can show that Eq(11. 1. .9) has prime period-two solutions if and only if γ = β + δ + A. 197 converges to a period-six solution of the form . .9) converges to the positive equilibrium? (c) Assume γ > β + δ + A. xn−1 + xn−2 n = 0. .10). . . . . ψ.11.) ψ 1 1 φ .12 Show that every positive solution of the equation xn+1 = 1 + xn .9) possesses unbounded solutions? (See [6]. ψ ∈ (0. what other conditions are needed so that every positive solution of Eq(11. φ...9) converges to a period-two solution? (b) Assume γ < β + δ + A. 1. . Open Problems and Conjectures Conjecture 11.4. n = 0. φ φ ψ ψ Open Problem 11. What other conditions are needed so that every positive solution of Eq(11.4.10) (a) In addition to (11. . where φ. . ∞).. A + Bxn + Dxn−2 (11.4.9) with nonnegative parameters and nonnegative initial conditions such that B + D > 0 and A + Bxn + Dxn−2 > 0 for n ≥ 0. What other conditions are needed so that Eq(11.

. converges to a period-two solution. xn 1 + xn−1 + xn−2 . converges to a period-two solution. 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Conjecture 11. 1.4. . ∞).15 Show that every positive solution of the equation xn+1 = xn−2 . 1. Conjecture 11. xn−1 + xn−2 + xn−3 n = 0. (See [6].13 Show that every positive solution of each of the two equations xn+1 = and xn+1 = 1 + xn−1 + xn−2 .4. (a) Show that when q=r every positive solution converges to a period-four solution.) Conjecture 11.198 Chapter 11. r ∈ (0. 1. . . . . . xn−1 n = 0.14 Show that every positive solution of the equation xn+1 = 1 + xn−1 . converges to a period-three solution. . 1. . (See [6]. 1 + xn + xn−2 n = 0. . n = 0. . . . .16 Consider the diﬀerence equation xn+1 = where p ∈ [0.) Conjecture 11. .4. The (3. p + qxn + rxn−2 . xn−2 n = 0. . (b) Show that when q>r every positive solution converges to the positive equilibrium. ∞) and q.4. 1.

Bk ∈ [0. Open Problems and Conjectures (c) Show that when q<r the equation possesses positive unbounded solutions.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 . . Obtain necessary and suﬃcient conditions on the parameters A and B0 . A + B0 xn + .11. 1. ∞) with i=0 Bi > 0. xn−1 . . . .9 Assume that k is a positive integer and that k A. xn−2 n = 0. . .4. B1 . . 1 + xn + xn−3 n = 0. . . . xn + xn−2 + xn−3 n = 0.4. . 1. xn−2 + xn−3 . . . . . 1.4. . B0 . . .) 199 Open Problem 11. xn n = 0. . converges to a period-(k + 1) solution of this equation. . 1. B1 . . + Bk xn−k n = 0. (See [62]. Open Problem 11. 1 + xn−3 . Bk so that every positive solution of the diﬀerence equation xn+1 = xn−k . .) . . 1. (See [6]. .

. x0 .11) converges to a period-(k + 1) solution of the form . . (a) Does Eq(11. . 3)-Type Equation xn+1 = α+βxn +γxn−1 A+Bxn +Cxn−1 Open Problem 11. xn−k+1 n = 0. . 0. φ. .11) possess a positive solution which converges to zero? (b) Determine.) Conjecture 11. . 1 + xn + ..11). + xn−k+1 n = 0. . . . . . (c) Determine the set of all nonnegative initial conditions x−k . x0 for which the ∞ corresponding solution {xn }n=−k of Eq(11. . . . . the value of φ in (11. The (3. .. converges to a periodic solution of period 2k.12) with φ ≥ 0. . k−terms (11.10 Assume that k is a positive integer. 1. . 0. . .4. Show that every positive solution of xn+1 = 1 + xn + xn−k .12) which corresponds to the nonnegative solution {xn }∞ n=−k of Eq(11. . .200 Chapter 11. 0. . . Then we can easily see that every nonnegative solution of the equation xn+1 = xn−k . .17 Let k be a positive integer. 1.4. in terms of the nonnegative initial conditions x−k .11) converges to a period-(k+1) solution of the form (11.12) with a given φ ≥ 0 ? (See [6]. (11.

b] and every solution of Eq(A. or non increasing in one and non decreasing in the other two.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. xn−2 ). As we have seen in previous chapters. Let I be some interval of real numbers and let f :I ×I ×I →I be a continuously diﬀerentiable function. n = 0. (b) The equation f (x. . b]3 → [a. y.0. y. .1) converges to x. (A.1) is to work in the regions where the function f (x.8. the diﬀerence equation xn+1 = f (xn .1) has a unique equilibrium x ∈ [a. We state the results for third order equations from which higher order extensions will become clear. or non decreasing in all three.1 Let [a. z) is non-decreasing in each of its arguments. y. z) is non decreasing in all three arguments.4. Theorem A. z) is monotonic in its arguments. very often a good strategy for obtaining global attractivity results for Eq(A. b] is a continuous function satisfying the following properties: (a) f (x. x−2 ∈ I. or non increasing in two arguments and non decreasing in the third. x−1 .1) ∞ has a unique solution {xn }n=−2 . Then for every set of initial conditions x0 . b]. b] be an interval of real numbers and assume that f : [a. xn−1 . 201 . . x) = x. x.5-1. has a unique solution in the interval [a. Then Eq(A. 1.4. In this regard there are eight possible scenarios depending on whether f (x.

m = M = x. . b] for each z ∈ [a. i→∞ and M = lim Mi .2 Let [a. ≤ M1 ≤ M0 . and is nonincreasing in z ∈ [a. Mi−1 ) and mi = f (mi−1 . y. ≤ Mi ≤ . Then Eq(A. (A. M ) ∈ [a. . M ) and M = f (M. 2 Theorem A. . from which the result follows. m) then m = M.202 Proof. Set m0 = a and M0 = b and for i = 1. m.1) has a unique equilibrium x ∈ [a. M. mi−1 . b]. . . . M. mi−1 ). b] for each x and y ∈ [a. m) and M = f (M. Now observe that for each i ≥ 0. M ). set Mi = f (Mi−1 . z) is non-decreasing in x and y ∈ [a. In view of (b). b] and every solution of Eq(A. .0. b] × [a. Mi−1 . ≤ mi ≤ . b]. b] be an interval of real numbers and assume that f : [a. m = f (m.1) converges to x. m0 ≤ m1 ≤ . (b) If (m. 2. . b] is a solution of the system m = f (m. m. . b]3 → [a. b] is a continuous function satisfying the following properties: (a) f (x. and mi ≤ xk ≤ Mi Set m = lim mi i→∞ Appendix for k ≥ 3i + 1.2) Then M ≥ lim sup xi ≥ lim inf xi ≥ m i→∞ i→∞ and by the continuity of f . .

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

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

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

b] be an interval of real numbers and assume that f : [a. ≤ mi ≤ .1) converges to x. . . . b] is a solution of the system M = f (m. mi−1 . M = f (M. . i→∞ and M = lim Mi . b] and every solution of Eq(A. b]. M.2) holds and by the continuity of f . Mi ≤ . m. (b) If (m. Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . b] is a continuous function satisfying the following properties: (a) f (x. m) and m = f (m. b] and is non-increasing in x and z for each y ∈ [a. Mi−1 .206 Proof. M ) ∈ [a. . M ) then m = M.0.6 Let [a. . Then Eq(A. set Mi = f (Mi−1 . M. Mi−1 ). Then (A. . y.1) has a unique equilibrium x ∈ [a. m) and m = f (M. z) is non-decreasing in y for each x and z ∈ [a. mi−1 ) and mi = f (mi−1 . b]3 → [a. . m. Therefore in view of (b) m=M from which the results follows. b] × [a. . 2. 2 Theorem A. M ). M1 ≤ M0 = b and mi ≤ xk ≤ Mi Set m = lim mi i→∞ Appendix for k ≥ 3i + 1. Set m0 = a and M0 = b and for each i = 1. .

M ). m. Set m0 = a and M0 = b and for each i = 1. z) is non-decreasing in z for each x and y in [a.1) has a unique equilibrium x ∈ [a. Mi−1 . M. b] is a solution of the system M = f (m. b] × [a. 2 Theorem A. Mi−1 ). .Appendix Proof.1) converges to x. Then Eq(A. m) and m = f (M. 2. . Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . mi−1 .2) holds and by the continuity of f . Mi ≤ . m. b] is a continuous function satisfying the following properties: (a) f (x. b]3 → [a. m) then m = M. . . M ) and m = f (M. b] be an interval of real numbers and assume that f : [a. b] and is non-increasing in x and y for each z ∈ [a. b] and every solution of Eq(A. ≤ mi ≤ . . . Then (A. M. i→∞ and M = lim Mi . . y. . mi−1 ) and mi = f (Mi−1 . M = f (m. . . b].0.7 Let [a. (b) If (m. M ) ∈ [a. set Mi = f (mi−1 . Therefore in view of (b) m=M from which the results follows. M1 ≤ M0 = b and mi ≤ xk ≤ Mi Set m = lim mi i→∞ 207 for k ≥ 3i + 1.

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

. Then (A. . . M ).1-A. Mi−1 ). . . . . ≤ mi ≤ . mi−1 . (A. M. .Appendix Proof. . Theorem A. This gives 2p results of the type found in Theorems A. .0. M1 ≤ M0 = b and m i ≤ x k ≤ Mi Set m = lim mi i→∞ 209 for k ≥ 3i + 1. . . Let us formulate one of these results. there are 2p possibilities for the function f to be non decreasing in some arguments and non increasing in the remaining arguments. . p xn+1 = f (xn . m) and m = f (M. xn−1 .0. f : Ip → I n = 0. .3) 2 The above theorems can be easily extended to the general diﬀerence equation of order . M = f (m. u2 . b] is a continuous function satisfying the following properties: (a) f (u1 . . up ) is non decreasing in the second and the third variables u2 and u3 and is non increasing in all other variables. b]p → [a. . 1. Set m0 = a and M0 = b and for each i = 1.2) holds and by the continuity of f . Now observe that for each i ≥ 0 m0 ≤ m1 ≤ . . mi−1 ) and mi = f (Mi−1 . . . i→∞ and M = lim Mi . Mi ≤ . set Mi = f (mi−1 . Therefore in view of (b) m=M from which the results follows. xn−p+1 ).9 Let [a.0. b] be an interval of real numbers and assume that f : [a. where is a continuous function. 2.8. . Mi−1 . m. In this case.

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

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

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

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

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

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

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

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

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

- Rosas - A Robust State Observer for Ndof Lagrangian Systems
- Industry
- Stability Revieweer
- 1000013976687.pdf
- Titu Andreescu, Cristinel Mortici, Marian Tetiva - Mathematical Bridges
- XUlaa1
- A+SURVEY+ON+SEMI-TENSOR+PRODUCT+OF+MATRICES
- Article for Review
- Design and Real-Time Control of a 4-DOF Biped Robot
- STABLE INDIRECT ADAPTIVE TYPE-2 FUZZY SLIDING MODE
- Michel Miller Ordinary Differential Equations
- Solution of homework
- Construction of Global Lyapunov Functions Using Radial Basis Functions
- Math312S12FinalPrac1(2).pdf
- 00186309
- EPW - Indexing the Tax Enforcement
- ch05
- A Graphical Introduction to Survey Adjustment777
- ENAE 641
- MScMCF2014_AdmissionExercise.pdf
- Managing Diversification SSRN-id1358533
- ENCIT2012-0318_4546
- zj_paper_39
- 04282107
- A g 4301181184
- 10.1.1.26.5159
- Linear Algebra Linear Systems Differential Equations
- Matrices 2
- CompQM
- Elements

Sign up to vote on this title

UsefulNot usefulClose Dialog## Are you sure?

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

Close Dialog## This title now requires a credit

Use one of your book credits to continue reading from where you left off, or restart the preview.

Loading