stability in m -dimensional linear delay difference system

19
This article was downloaded by: [University of Saskatchewan Library] On: 02 May 2013, At: 03:51 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Journal of Difference Equations and Applications Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/gdea20 Stability in m-dimensional linear delay difference system X. H. Tang a & Zhiyuan Jiang a a School of Mathematical Science and Computing Technology, Central South University, Changsha, 410083, People's Republic of China Published online: 19 Sep 2007. To cite this article: X. H. Tang & Zhiyuan Jiang (2007): Stability in m-dimensional linear delay difference system, Journal of Difference Equations and Applications, 13:10, 927-944 To link to this article: http://dx.doi.org/10.1080/10236190701388419 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Upload: zhiyuan

Post on 09-Dec-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

This article was downloaded by: [University of Saskatchewan Library]On: 02 May 2013, At: 03:51Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Journal of Difference Equations and ApplicationsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/gdea20

Stability in m-dimensional linear delay differencesystemX. H. Tang a & Zhiyuan Jiang aa School of Mathematical Science and Computing Technology, Central South University,Changsha, 410083, People's Republic of ChinaPublished online: 19 Sep 2007.

To cite this article: X. H. Tang & Zhiyuan Jiang (2007): Stability in m-dimensional linear delay difference system, Journal ofDifference Equations and Applications, 13:10, 927-944

To link to this article: http://dx.doi.org/10.1080/10236190701388419

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form toanyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contentswill be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses shouldbe independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly inconnection with or arising out of the use of this material.

Stability in m-dimensional linear delaydifference system†

X. H. TANG* and ZHIYUAN JIANG

School of Mathematical Science and Computing Technology, Central South University,Changsha 410083, People’s Republic of China

(Received 30 October 2006; revised 12 February 2007; in final form 9 April 2007)

In this paper, we give some sufficient conditions for the zero solution of an m-dimensional delaydifference equation of the form

xðnþ 1Þ2 xðnÞ ¼ AðnÞxðn2 kÞ; n ¼ 0; 1; 2; . . .

to be uniformly stable and asymptotically stable.

Keywords: Stability; m-Dimensional; Delay difference equation; Uniformly stable

2000 Mathematics Subject Classification: 39 A

1. Introduction

Consider the m-dimensional linear delay difference system

xðnþ 1Þ2 xðnÞ ¼ AðnÞxðn2 kÞ; n ¼ 0; 1; 2; . . . ; ð1Þ

where {x(n)} is an m-dimensional vector sequence, {A(n)} is an m £ m matrix sequence, k is

a nonnegative integer.

In the one-dimensional case of (1), the stability of the zero solution of equation (1) has

been much studied [1,2,4–6,8,10]. In the multi-dimensional case, the results dealt with the

stability of the zero solution of equation (1) are relatively scarce. We only find a paper [7]

which gives a necessary and sufficient condition for the asymptotic stability of equation (1)

in terms of the eigenvalues of the coefficient matrix A when AðnÞ ; A is a constant matrix.

However, when A(n) is not a constant matrix, the theory of characteristic equation fails to

apply equation (1). In this case, it seems a common method to construct a working

“Liapunov” function. But this method is also difficult in general. Motivated by the work on

stability in higher dimensional delay differential equations in [3] and [9], we will give

sufficient conditions for the uniform stability and asymptotic stability of the zero solution of

equation (1) by a “geometric method” instead of “Liapunov method”.

For x; y [ Rm, jxj is the Euclidean norm, and x·y is the inner product of x and y. For

A [ Rm£m, jAj denotes the matrix norm induced by the above vector norm. Let f :

{ 2 k;2k þ 1; . . . ; 0} ! Rm and Ekm be the set of m-dimensional vector system

Journal of Difference Equations and Applications

ISSN 1023-6198 print/ISSN 1563-5120 online q 2007 Taylor & Francis

http://www.tandf.co.uk/journals

DOI: 10.1080/10236190701388419

†This work was supported by NNSF (No: 10471153) of China.*Corresponding author. Email: [email protected]

Journal of Difference Equations and Applications,

Vol. 13, No. 10, October 2007, 927–944

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

ðfð2kÞ;fð2k þ 1Þ; . . . ;fð0ÞÞ, and let

kfk ¼ max{jfðsÞj : s ¼ 2k;2k þ 1; . . . ; 0}:

For any m-dimensional vector sequence {y(n)} and any integer n . 0, we define ynðsÞ ¼

yðnþ sÞ for s [ { 2 k;2k þ 1; . . . ; 0}. For any initial value f [ Ekm and integer n0 $ 0,

we denote by xðn; n0;fÞ the solution of equation (1) with initial condition xn0¼ f.

Definition 1. We say the zero solution of equation (1) is uniformly stable, if for any 1 . 0

there exists d(1) . 0 such that if n0 $ 0 and f [ Ekm, then kfk , dð1Þ implies

jxðn; n0;fÞj , 1 for all n $ n0 2 k.

2. Main results

First, we state a lemma which is taken from [9].

Lemma 1[9]. Let a, b and c be points in Rm with jaj ¼ jbj ¼ jcj ¼ 1. Then

2ða·bÞða·cÞðb·cÞ $ ða·bÞ2 þ ða·cÞ2 2 1:

Lemma 2. Let K [ (0,1] and

qKðxÞ ¼ð1 2 xÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ x2 2 2x

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

:

Then

(i) there exists a unique a [ ð0; 1Þ such that

1 2 a ¼a2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

p

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ a2 2 2a

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pp ; ð2Þ

(ii) min0#x#1qKðxÞ ¼ qKðaÞ.

Proof. Set a ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

p. Then qKðxÞ ¼ ð1=2Þð1 2 xÞ2 þ

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ x2 2 2ax

p,

q 0KðxÞ ¼ 21 þ xþ

x2 affiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ x2 2 2ax

p ; q 0Kð0Þ ¼ 2a , 0; q 0

Kð1Þ ¼

ffiffiffiffiffiffiffiffiffiffiffi1 2 a

2

r. 0;

and

q 00KðxÞ ¼ 1 þ

1 2 a2

ð1 þ x 2 2 2axÞffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ x2 2 2ax

p . 0; for x [ ½0; 1�:

Therefore, the equation q 0KðxÞ ¼ 0 has a unique foot a [ (0,1), i.e. there exists a unique

a [ (0,1) such that (2) holds. It follows that (ii) holds also. The proof is complete. A

Theorem 1. Assume that there exists a K [ ð0; 1� such that

2AðnÞx

jAðnÞxj·x

jxj$ K ð3Þ

X. H. Tang et al.928

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

for all ðn; xÞ [ {0; 1; 2; . . . ; } £ Rm with AðnÞx – 0, and that

Xni¼n2k

jAðiÞj #ð1 2 aÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ a2 2 2a

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

; n $ 0; ð4Þ

where a [ (0,1) is defined by (2). Then the zero solution of equation (1) is uniformly stable.

Proof. By (2), we have

M U qKðaÞ ¼ð1 2 aÞ2

a2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

p

1 2 a¼

ð1 2 aÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ a2 2 2a

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

:

Then from (1), we have

jxðnÞj # jxðn0Þj þXn21

i¼n0

jAðiÞj jxði2 kÞj # jxðn0Þj þXn21

i¼n0

jAðiÞj kxik:

It follows that

kxnk # kxn0k þ

Xn21

i¼n0

jAðiÞj kxik; n $ n0:

By Discrete Grownwall’s inequality, we have

kxnk # kxn0kYn21

i¼n0

ð1 þ jAðiÞjÞ # kfkexpXn21

i¼n0

jAðiÞj

!; n $ n0;

which implies that

jxðnÞj # kfke2M ; n0 # n # n0 þ 2k: ð5Þ

Let

r ¼ max{jxðiÞj : n0 2 k # i # n0 þ 2k}:

Assume that r . 0. In what follows, we will show that

jxðnÞj # r for n $ n0 þ 2k: ð6Þ

Suppose (6) is false. Let N þ 1 ¼ min{n : jxðnÞj . r}. Then jxðN þ 1Þj . r;

N þ 1 . n0 þ 2k, and jxðnÞj # r for n0 # n # N. Hence, from (1), we have

jxðnþ 1Þ2 xðnÞj # jAðnÞj jxðn2 kÞj # jAðnÞj jxðN þ 1Þj; n0 # n # N þ 1: ð7Þ

For N 2 k # n # N, from (7), we have

jxðn2 kÞj # jxðn2 kÞ2 xðN 2 kÞj þ jxðN 2 kÞj

¼XN2k21

i¼n2k

½xðiþ 1Þ2 xðiÞ�

����������þ jxðN 2 kÞj

# jxðN þ 1ÞjXN2k21

i¼n2k

jAðiÞj þ jxðN 2 kÞj:

Linear delay difference system 929

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

Substituting this into the first inequality in (7), then we have

jxðnþ 1Þ2 xðnÞj # jAðnÞj jxðN þ 1ÞjXN2k21

i¼n2k

jAðiÞj þ jxðN 2 kÞj

" #; N 2 k # n # N:

On combining this and (7), we obtain

jxðnþ 1Þ2 xðnÞj # jxðN þ 1Þj jAðnÞjmin 1;XN2k21

i¼n2k

jAðiÞj þjxðN 2 kÞj

jxðN þ 1Þj

( );

N 2 k # n # N:

ð8Þ

Note that

jxðN þ 1Þj2 jxðN 2 kÞj # jxðN þ 1Þ2 xðN 2 kÞj # jxðN þ 1ÞjXN

i¼N2k

jAðiÞj # MjxðN þ 1Þj:

It follows that 1 2 jxðN 2 kÞj=jxðN þ 1Þj # M. Then there are two possible cases to

consider:

Case 1. M $PN

i¼N2kjAðiÞj . 1 2 jxðN 2 kÞj=jxðN þ 1Þj. In this case, there exist an integer

l with 0 # l # k2 1 and j [ ½0; 1Þ such that

XNi¼N2l

jAðiÞj þ jjAðN 2 l2 1Þj ¼ 1 2jxðN 2 kÞj

jxðN þ 1Þj:

Hence, from (4) and (8), we have

jxðN þ 1Þ2 xðN 2 kÞj

#XN2l22

n¼N2k

jxðnþ 1Þ2 xðnÞj þ ð1 2 jÞjxðN 2 lÞ2 xðN 2 l2 1Þj

þ jjxðN 2 lÞ2 xðN 2 l2 1Þj þXN

n¼N2l

jxðnþ 1Þ2 xðnÞj

# jxðN þ 1ÞjXN2l22

i¼N2k

jAðiÞj þ ð1 2 jÞjAðN 2 l2 1Þj

"

þjjAðN 2 l2 1ÞjXN2k21

i¼N2l2k21

jAðiÞj þjxðN 2 kÞj

jxðN þ 1Þj

!

þXN

n¼N2l

jAðnÞjXN2k21

i¼n2k

jAðiÞj þjxðN 2 kÞj

jxðN þ 1Þj

!#

¼ jxðN þ 1ÞjXN2l21

i¼N2k

jAðiÞj2 jjAðN 2 l2 1Þj

!"

X. H. Tang et al.930

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

£XN

n¼N2l

jAðnÞj þ jjAðN 2 l2 1Þj þjxðN 2 kÞj

jxðN þ 1Þj

!

þjjAðN 2 l2 1ÞjXN2k21

i¼N2l2k21

jAðiÞj þjxðN 2 kÞj

jxðN þ 1Þj

!

þXN

n¼N2l

jAðnÞjXN2k21

i¼n2k

jAðiÞj þjxðN 2 kÞj

jxðN þ 1Þj

!#

¼ jxðN þ 1ÞjXN

n¼N2l

jAðnÞjXN2l21

i¼n2k

jAðiÞj2 jjAðN 2 l2 1Þj

!" #

þjjAðN 2 l2 1ÞjXN2l21

i¼N2l2k21

jAðiÞj2 jjAðN 2 l2 1Þj

!

þjxðN 2 kÞj

jxðN þ 1Þj

XNi¼N2k

jAðiÞj

#

# jxðN þ 1Þj MXNi¼N2l

jAðiÞj2XN

n¼N2l

jAðnÞjXni¼N2l

jAðiÞj

"

þjjAðN 2 l2 1Þj M 2XNi¼N2l

jAðiÞj2 jjAðN 2 l2 1Þj

jxðN 2 kÞj

jxðN þ 1ÞjM

#

¼ jxðN þ 1Þj MXNi¼N2l

jAðiÞj21

2

XNi¼N2l

jAðiÞj

!2

21

2

XNi¼N2l

jAðiÞj2

24

þjjAðN 2 l2 1Þj M 2XNi¼N2l

jAðiÞj2 jjAðN 2 l2 1Þj

jxðN 2 kÞj

jxðN þ 1ÞjM

#

¼ jxðN þ 1Þj MXNi¼N2l

jAðiÞj þ jjAðN 2 l2 1Þj

jxðN 2 kÞj

jxðN þ 1ÞjM

"

21

2

XNi¼N2l

jAðiÞj þ jjAðN 2 l2 1Þj

!2

21

2

XNi¼N2l

jAðiÞj2þ j2jAðN 2 l2 1Þj

2

!#

Linear delay difference system 931

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

# jxðN þ 1Þj MXNi¼N2l

jAðiÞj þ jjAðN 2 l2 1Þj

jxðN 2 kÞj

jxðN þ 1ÞjM

"

21

2

XNi¼N2l

jAðiÞj þ jjAðN 2 l2 1Þj

!235

¼ jxðN þ 1Þj M 21

21 2

jxðN 2 kÞj

jxðN þ 1Þj

� �2" #

:

Case 2.PN

i¼N2kjAðiÞj # 1 2 jxðN 2 kÞj=jxðN þ 1Þj # M. Then from (4) and (8), we have

jxðN þ 1Þ2 xðN 2 kÞj

#XN

n¼N2k

jxðnþ 1Þ2 xðnÞj

# jxðN 2 kÞjXN

n¼N2k

jAðnÞj þ jxðN þ 1ÞjXN

n¼N2k

jAðnÞjXN2k21

i¼n2k

jAðiÞj

¼ jxðN 2 kÞjXN

n¼N2k

jAðnÞj þ jxðN þ 1ÞjXN

n¼N2k

jAðnÞjXni¼n2k

jAðiÞj2Xn

i¼N2k

jAðiÞj

!

# jxðN 2 kÞjM þ jxðN þ 1Þj MXN

n¼N2k

jAðnÞj21

2

XNi¼N2k

jAðiÞj

!2

21

2

XNi¼N2k

jAðiÞj2

24

35

# jxðN 2 kÞjM þ jxðN þ 1Þj MXN

i¼N2k

jAðiÞj21

2

XNi¼N2k

jAðiÞj

!224

35

# jxðN þ 1Þj MjxðN 2 kÞj

jxðN þ 1ÞjþM 1 2

jxðN 2 kÞj

jxðN þ 1Þj

� �2

1

21 2

jxðN 2 kÞj

jxðN þ 1Þj

� �2" #

¼ jxðN þ 1Þj M 21

21 2

jxðN 2 kÞj

jxðN þ 1Þj

� �2" #

:

Combining Case 1 and Case 2, we have

jxðN þ 1Þ2 xðN 2 kÞj # jxðN þ 1Þj M 21

21 2

jxðN 2 kÞj

jxðN þ 1Þj

� �2" #

: ð9Þ

From (1), we have

jxðN þ 1Þj2 jxðNÞj ¼2xðN þ 1Þ · ½xðN þ 1Þ2 xðNÞ�2 jxðN þ 1Þ2 xðNÞj

2

jxðN þ 1Þj þ jxðNÞj

¼2xðN þ 1Þ ·AðNÞxðN 2 kÞ2 jAðNÞxðN 2 kÞj

2

jxðN þ 1Þj þ jxðNÞj

#2xðN þ 1Þ ·AðNÞxðN 2 kÞ

jxðN þ 1Þj þ jxðNÞj:

X. H. Tang et al.932

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

It follows from the fact that jxðN þ 1Þj2 jxðNÞj . 0 that

AðNÞxðN 2 kÞ

jAðNÞxðN 2 kÞj

xðN þ 1Þ

jxðN þ 1Þj. 0: ð10Þ

From (3), we have

2AðNÞxðN 2 kÞ

jAðNÞxðN 2 kÞj

xðN 2 kÞ

jxðN 2 kÞj$ K: ð11Þ

Hence, by (10), (11) and Lemma 1, we have

xðN þ 1Þ

jxðN þ 1Þj

xðN 2 kÞ

jxðN 2 kÞj,

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

p: ð12Þ

Set u ¼ jxðN 2 kÞj=jxðN þ 1Þj and use the fact that

2xðN þ 1ÞxðN 2 kÞ ¼ jxðN þ 1Þj2þ jxðN 2 kÞj

22 jxðN þ 1Þ2 xðN 2 kÞj

2:

It follows from (9) that

xðN þ 1Þ

jxðN þ 1Þj

xðN 2 kÞ

jxðN 2 kÞj$

1

2u1 þ u2 2 M 2

ð1 2 uÞ2

2

� �2( )

: ð13Þ

Combining (12) and (13), we have

2uffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

p. 1 þ u2 2 M 2

ð1 2 uÞ2

2

� �2

:

It follows Lemma 2 (ii) that

M .ð1 2 uÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ u2 2 2u

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

$ min0#u#1

ð1 2 uÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ u2 2 2u

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq� �

¼ð1 2 aÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ a2 2 2a

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

¼ M:

This is a contradiction, hence, (6) holds. Combining (5) and (6), we have

jxðnÞj # kfke2M for n $ n0 2 k:

Let 1 . 0 be any given and choose d ¼ ð1=2Þ1e22M . Then f [ Ekm and kfk , d implies

jxðnÞj ¼ jxðn; n0;fÞj , 1. The proof is complete. A

Theorem 2. Assume that there exist a K [ ð0; 1�, N0 . 0 and a nonnegative sequence

{mðnÞ} with S1mðnÞ ¼ 1 such that (4) holds and

jAðnÞxj $ mðnÞjxj ð14Þ

for all ðn; xÞ [ {N0;N0 þ 1; . . . } £ Rm with AðnÞx – 0, and that

lim supn!1

Xni¼n2k

jAðiÞj ,ð1 2 aÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ a2 2 2a

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

; ð15Þ

where a [ ½0; 1Þ is defined by (2). Then every solution of equation (1) tends zero as n!1.

Linear delay difference system 933

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

Proof. Set M ¼ ð1 2 aÞ2=2 þffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ a2 2 2a

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pp. By (15), there exist M1 , M and

N1 . N0 such that Xni¼n2k

jAðiÞj # M1; n $ N1: ð16Þ

From (1), we have

jxðnþ 1Þj2 jxðnÞj ¼2xðnþ 1ÞAðnÞxðn2 kÞ2 jAðnÞxðn2 kÞj

2

jxðnþ 1Þj þ jxðnÞj: ð17Þ

In view of Theorem 1, jxðnÞj is bounded. In what follows, we will show

limn!1

jxðnÞj ¼ 0 ð18Þ

in two possible cases.

Case 1. {jxðnþ 1Þj2 jxðnÞj} is eventually non-negative or eventually non-positive. In this

case, {jxðnÞj} is eventually monotone and the limit limn!1jxðnÞj ¼ c exists. If c . 0, then

limn!1

jxðn2 kÞj

jxðnÞj¼ lim

n!1

jxðn2 2kÞj

jxðnÞj¼ 1: ð19Þ

From (1) and (16), we have

jxðnþ1Þ2 xðn2 kÞj#Xni¼n2k

jAðiÞj jxði2 kÞj#M1max{jxðnþ1Þj; jxðn22kÞj}; n$N1;

which, together with (19), implies that there exists N2 .N1 such that

xðnþ1Þ

jxðnþ1Þj

xðn2 kÞ

jxðn2 kÞj

¼jxðnþ1Þj

2þjxðn2 kÞj

22 jxðnþ1Þ2 xðn2 kÞj

2

2jxðnþ1Þj jxðn2 kÞj

$jxðnþ1Þj

2þjxðn2 kÞj

22M2

1max{jxðnþ1Þj2; jxðn22kÞj

2}

2jxðnþ1Þj jxðn2 kÞj

$ 121

2M 2; n$N2:

On the other hand, by (2), we have

2AðnÞxðn2 kÞ

jAðnÞxðn2 kÞj

xðn2 kÞ

jxðn2 kÞj$K; jAðnÞxðn2 kÞj– 0; n$N0:

Hence, in view of Lemma 1, we have

xðnþ1Þ

jxðnþ1Þj

AðnÞxðn2 kÞ

jAðnÞxðn2 kÞj#

xðnþ1Þjxðnþ1Þj

xðn2kÞjxðn2kÞj

2

þ AðnÞxðn2kÞjAðnÞxðn2kÞj

xðn2kÞjxðn2kÞj

2

21

2 xðnþ1Þjxðnþ1Þj

xðn2kÞjxðn2kÞj

AðnÞxðn2kÞjAðnÞxðn2kÞj

xðn2kÞjxðn2kÞj

#12 12M 2

2

2

2K 2

22 xðnþ1Þjxðnþ1Þj

xðn2kÞjxðn2kÞj

AðnÞxðn2kÞjAðnÞxðn2kÞj

xðn2kÞjxðn2kÞj

;jAðnÞxðn2 kÞj– 0; n$N2:

X. H. Tang et al.934

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

Let

b¼ 12M 2

2

� �2

þK 2 21:

Note that

M¼ð12aÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þa2 22a

ffiffiffiffiffiffiffiffiffiffiffiffiffiffi12K 2

pq

¼ min0#u#1

ð12uÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þu2 22u

ffiffiffiffiffiffiffiffiffiffiffiffiffiffi12K 2

pq� �

,

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi222

ffiffiffiffiffiffiffiffiffiffiffiffiffiffi12K 2

pq

:

It follows that

1

4M 4 2M 2 þK 2 . 0;

and so b. 0. Hence, for n$N2 and jAðnÞxðn2 kÞj– 0,

xðnþ1Þ

jxðnþ1Þj

AðnÞxðn2 kÞ

jAðnÞxðn2 kÞj#2

b

22 xðnþ1Þjxðnþ1Þj

xðn2kÞjxðn2kÞj

AðnÞxðn2kÞjAðnÞxðn2kÞj

xðn2kÞjxðn2kÞj

#2b

2; n$N2:

ð20Þ

Combining (14), (17) and (20), we obtain

jxðnþ1Þj2 jxðnÞj ¼2xðnþ1ÞAðnÞxðn2 kÞ2 jAðnÞxðn2 kÞj

2

jxðnþ1Þjþ jxðnÞj

#2xðnþ1ÞAðnÞxðn2 kÞ

jxðnþ1Þj

#2bjAðnÞxðn2 kÞj

#2bmðnÞjxðn2 kÞj; n$N2:

The above shows jxðnÞj nonincreasing when n$N2. Summing the above from N2 to 1,

we get

c2 jxðN2Þj#2cbX1n¼N2

mðnÞ ¼21:

This is a contradiction. Hence, c ¼ 0, and so (18) holds.

Case 2. {jxðnþ 1Þj2 jxðnÞj} is oscillatory and jxðnþ 1Þj2 jxðnÞj � 0 eventually.

Let v ¼ lim supn!1jxðnÞj. It suffices to show v ¼ 0. Note that M1 , M. If v . 0, then we

can choose 1 . 0 is sufficient small so that

M1 þ21

vþ 1¼ M2 , M 2

21

vþ 1: ð21Þ

Choose a sequence {nj} of integers with n1 . N0 þ 2k such that

jxðnj þ 1Þj2 jxðnjÞj . 0; jxðnj þ 1Þj . v2 1;

limj!1

jxðnj þ 1Þj ¼ v; and jxðnÞj , vþ 1; n $ n1 þ k:

8<: ð22Þ

Linear delay difference system 935

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

Then from (1) and (22), we have

jxðnþ 1Þ2 xðnÞj # jAðnÞj jxðn2 kÞj # jAðnÞjðvþ 1Þ; n $ n1: ð23Þ

Hence, for large j

jxðn2 kÞj # jxðn2 kÞ2 xðnj 2 kÞj þ jxðnj 2 kÞj

¼Xnj2k21

i¼n2k

½xðiþ 1Þ2 xðiÞ�

����������þ jxðnj 2 kÞj

# ðvþ 1ÞXnj2k21

i¼n2k

jAðiÞj þ jxðnj 2 kÞj; nj 2 k # n # nj:

Substituting this into the first inequality in (23), then for nj 2 k # n # nj we have

jxðnþ 1Þ2 xðnÞj # jAðnÞj ðvþ 1ÞXnj2k21

i¼n2k

jAðiÞj þ jxðnj 2 kÞj

" #:

On combining this and (23), we obtain

jxðnþ 1Þ2 xðnÞj # ðvþ 1ÞjAðnÞjmin 1;Xnj2k21

i¼n2k

jAðiÞj þjxðnj 2 kÞj

vþ 1

( );

nj 2 k # n # nj:

ð24Þ

Note that

jxðnj þ 1Þj2 jxðnj 2 kÞj # jxðnj þ 1Þ2 xðnj 2 kÞj # ðvþ 1ÞXnj

i¼nj2k

jAðiÞj # M1ðvþ 1Þ:

It follows that 1 2 jxðnj 2 kÞj=ðvþ 1Þ # M2. There are two possible cases to consider:

Case 2.1. M2 $Pnj

i¼nj2kjAðiÞj . 1 2 jxðnj 2 kÞj=ðvþ 1Þ. Then there exist an integer l with

0 # l # k2 1 and a j [ ½0; 1Þ such that

Xnji¼nj2l

jAðiÞj þ jjAðnj 2 l2 1Þj ¼ 1 2jxðnj 2 kÞj

vþ 1:

Hence, from (24), we have

jxðnjþ1Þ2xðnj2kÞj

#Xnj2l22

n¼nj2k

jxðnþ1Þ2xðnÞjþð12jÞjxðnj2 lÞ2xðnj2 l21Þj

þjjxðnj2 lÞ2xðnj2 l21ÞjþXnj

n¼nj2l

jxðnþ1Þ2xðnÞj

X. H. Tang et al.936

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

# ðvþ1ÞXnj2l22

i¼nj2k

jAðiÞjþð12jÞjAðnj2 l21Þj

24

þjjAðnj2 l21ÞjXnj2k21

i¼nj2l2k21

jAðiÞjþjxðnj2kÞj

vþ1

0@

1A

þXnj

n¼nj2l

jAðnÞjXnj2k21

i¼n2k

jAðiÞjþjxðnj2kÞj

vþ1

!35

¼ðvþ1ÞXnj2l21

i¼nj2k

jAðiÞj2jjAðnj2 l21Þj

0@

1A Xnj

n¼nj2l

jAðnÞjþjjAðnj2 l21Þjþjxðnj2kÞj

vþ1

0@

1A

24

þjjAðnj2 l21ÞjXnj2k21

i¼nj2l2k21

jAðiÞjþjxðnj2kÞj

vþ1

0@

1A

þXnj

n¼nj2l

jAðnÞjXnj2k21

i¼n2k

jAðiÞjþjxðnj2kÞj

vþ1

!35

¼ðvþ1ÞXnj

n¼nj2l

jAðnÞjXnj2l21

i¼n2k

jAðiÞj2jjAðnj2 l21Þj

!24

þjjAðnj2 l21ÞjXnj2l21

i¼nj2l2k21

jAðiÞj2jjAðnj2 l21Þj

0@

1Aþ

jxðnj2kÞj

vþ1

Xnji¼nj2k

jAðiÞj

35

# ðvþ1Þ M2

Xnji¼nj2l

jAðiÞj2Xnj

n¼nj2l

jAðnÞjXni¼nj2l

jAðiÞjþjjAðnj2 l21Þj

24

M22Xnji¼nj2l

jAðiÞj2jjAðnj2 l21Þj

0@

1Aþ

jxðnj2kÞj

vþ1M2

35

¼ðvþ1Þ M2

Xnji¼nj2l

jAðiÞj21

2

Xnji¼nj2l

jAðiÞj

0@

1A

2

21

2

Xnji¼nj2l

jAðiÞj2þjjAðnj2 l21Þj

24

M22Xnji¼nj2l

jAðiÞj2jjAðnj2 l21Þj

0@

1Aþ

jxðnj2kÞj

vþ1M2

35

Linear delay difference system 937

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

# ðvþ1Þ M2

Xnji¼nj2l

jAðiÞjþjjAðnj2 l21Þj

0@

1Aþ

jxðnj2kÞj

vþ1M2

24

21

2

Xnji¼nj2l

jAðiÞjþjjAðnj2 l21Þj

0@

1A

235

¼ðvþ1Þ M221

212

jxðnj2kÞj

vþ1

� �2" #

:

Case 2.2.Pnj

i¼nj2kjAðiÞj # 1 2 jxðnj 2 kÞj=ðvþ 1Þ # M2. Then from (24), we have

jxðnj þ 1Þ2 xðnj 2 kÞj

#Xnj

n¼nj2k

jxðnþ 1Þ2 xðnÞj

# jxðnj 2 kÞjXnj

n¼nj2k

jAðnÞj þ ðvþ 1ÞXnj

n¼nj2k

jAðnÞjXnj2k21

i¼n2k

jAðiÞj

¼ jxðnj 2 kÞjXnj

n¼nj2k

jAðnÞj þ ðvþ 1ÞXnj

n¼nj2k

jAðnÞjXni¼n2k

jAðiÞj2Xn

i¼nj2k

jAðiÞj

0@

1A

# jxðnj 2 kÞjM2 þ ðvþ 1Þ M2

Xnjn¼nj2k

jAðnÞj21

2

Xnji¼nj2k

jAðiÞj

0@

1A

2

21

2

Xnji¼nj2k

jAðiÞj2

24

35

# jxðnj 2 kÞjM2 þ ðvþ 1Þ M2

Xnji¼nj2k

jAðiÞj21

2

Xnji¼nj2k

jAðiÞj

0@

1A

224

35

# ðvþ 1Þ M2

jxðnj 2 kÞj

vþ 1þM2 1 2

jxðnj 2 kÞj

vþ 1

� �2

1

21 2

jxðnj 2 kÞj

vþ 1

� �2" #

¼ ðvþ 1Þ M2 21

21 2

jxðnj 2 kÞj

vþ 1

� �2" #

:

Combining Case 2.1 and Case 2.2, we have

jxðnj þ 1Þ2 xðnj 2 kÞj # ðvþ 1Þ M2 21

21 2

jxðnj 2 kÞj

vþ 1

� �2" #

: ð25Þ

X. H. Tang et al.938

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

Similar the proof of Theorem 1, we have

xðnj þ 1Þ

jxðnj þ 1Þj·xðnj 2 kÞ

jxðnj 2 kÞj,

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

p: ð26Þ

Set u ¼ jxðnj 2 kÞj=ðvþ 1Þ. Then from (25),

xðnj þ 1Þ

jxðnj þ 1Þj

xðnj 2 kÞ

jxðnj 2 kÞj

¼jxðnj þ 1Þj

2þ jxðnj 2 kÞj

22 jxðnj þ 1Þ2 xðnj 2 kÞj

2

2jxðnj þ 1Þj jxðnj 2 kÞj

$jxðnj þ 1Þj

2þ jxðnj 2 kÞj

22 ðvþ 1Þ2 M2 2

12ð1 2 uÞ2

� �2jxðnj þ 1Þj jxðnj 2 kÞj

a

¼

jxðnjþ1Þj

vþ1

2

þu2 2 M2 212ð1 2 uÞ2

� �22u

jxðnjþ1Þj

vþ1

$vþ 1

2uðv2 1Þ

v2 1

vþ 1

� �2

þu2 2 M2 2ð1 2 uÞ2

2

� �2( )

¼1

2aua2 þ u2 2 M2 2

ð1 2 uÞ2

2

� �2( )

;

where a ¼ ðv2 1Þ=ðvþ 1Þ. Combining this and (26), we have

2auffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

p. a2 þ u2 2 M2 2

ð1 2 uÞ2

2

� �2

:

It follows that

M2 .ð1 2 uÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia2 þ u2 2 2au

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

$ min0#u#1

ð1 2 uÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia2 þ u2 2 2au

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq� �

¼ð1 2 aðaÞÞ2

aðaÞ2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

p

1 2 aðaÞ;

where a(a) [ (0,1) satisfy that

1 2 aðaÞ ¼aðaÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

p

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia2 þ ½aðaÞ�2 2 2aaðaÞ

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq : ð27Þ

Let 1! 0þ in (27), then lim1!0þaðaÞ ¼ lima!1aðaÞ ¼ a and so

M1 $ð1 2 aÞ2

a2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

p

1 2 a¼ M:

This is a contradiction, hence, v ¼ 0 and so (18) holds. The proof is complete. A

Linear delay difference system 939

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

3. Remarks and examples

In the last section, we give some remarks and examples to illustrate our results. Note that

qKðaÞ ¼ð1 2 aÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ a2 2 2a

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

¼ min0#u#1

ð1 2 uÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ u2 2 2u

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq� �

. min0#u#1

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ u2 2 2u

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

¼ K;

and

qKðaÞ ¼ð1 2 aÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ a2 2 2a

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

¼ min0#u#1

ð1 2 uÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ u2 2 2u

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq� �

,

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 2 2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

:

Hence,

K , qKðaÞ ,

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 2 2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 2 K 2

pq

; 0 , K # 1: ð28Þ

For comparing K with qKðaÞ, we give some values of K and qKðaÞ as follows:

Example 1. Consider the 2-dimensional delay difference equation

xðnþ 1Þ2 xðnÞ ¼2aðnÞ bðnÞ

2bðnÞ 2aðnÞ

!xðn2 kÞ; ð29Þ

where {aðnÞ}; {bðnÞ} are two real sequences and aðnÞ . 0. Let

AðnÞ ¼2aðnÞ bðnÞ

2bðnÞ 2aðnÞ

!:

K qK ðaÞ

0.1 0.1000114110.2 0.2001700850.3 0.3008159410.4 0.4024885550.5 0.5059790810.6 0.6124816530.7 0.7239634970.8 0.8442805660.9 0.9828663891 1.242217666

X. H. Tang et al.940

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

Then for any x [ R2,

jAðnÞxj ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia2ðnÞ þ b2ðnÞ

pjxj;

and for x – 0,

2AðnÞx

jAðnÞxj

x

jxj¼

aðnÞffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia2ðnÞ þ b2ðnÞ

p :

Let

N ¼ supn$0

Xns¼n2k

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia2ðsÞ þ b2ðsÞ

p; and K ¼ inf

n$0

aðnÞffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia2ðnÞ þ b2ðnÞ

p :

By Theorem 1, if N # qKðaÞ, then the zero solution of (29) is uniformly stable.

Furthermore, if

lim supn!1

Xns¼n2k

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia2ðsÞ þ b2ðsÞ

p, qKðaÞ and

X1s¼0

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia2ðsÞ þ b2ðsÞ

p¼ 1:

Then by Theorem 2, every solution of (29) tends zero as n!1.

In particular, let aðnÞ ¼ rðnÞ cos u and bðnÞ ¼ rðnÞ sin u, where juj , ðp=2Þ and

rðnÞ $ 0; n ¼ 0; 1; 2; . . . . Then if

lim supn!1

Xns¼n2k

rðsÞ , f ðuÞ andX1s¼0

rðsÞ ¼ 1; ð30Þ

then every solution of the equation

xðnþ 1Þ2 xðnÞ ¼ rðnÞ2cos u sin u

2sin u 2cos u

!xðn2 kÞ; ð31Þ

tends to zero as n!1, where

f ðuÞ ¼ð1 2 aÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ a2 2 2ajsin uj

p;

and a [ ½0; 1Þ satisfies the equation

1 2 a ¼a2 jsin ujffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

1 þ a2 2 2ajsin ujp : ð32Þ

Remark 1. When rðnÞ ; r . 0, it is shown in [7] that every solution of the equation of

equation (31) tends to zero as n!1 if and only if

r , gðuÞ ¼ 2 coskpþ juj

2k þ 1: ð33Þ

In this case, condition (30) reduces to

r ,f ðuÞ

k þ 1¼

1

k þ 1

ð1 2 aÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ a2 2 2ajsin uj

p� �: ð34Þ

Linear delay difference system 941

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

To compare our result (34) with (33) in obtained in [7], we let k ¼ 2, and give some values of

f ðuÞ=3 and gðuÞ as follows:

Remark 2. When rðnÞ � constant, for example

rðnÞ ¼½1 þ ð21Þn� f ðuÞ

k þ 3þ

1

nþ 1; n ¼ 0; 1; 2; . . . :

It is easy to see that rðnÞ satisfies condition (30), therefore, every solution of the equation of

equation (31) tends to zero as n!1. However, the results in [7] can not be used for the

above rðnÞ.

Example 2. Consider more general 2-dimensional delay difference equation

xðnþ 1Þ2 xðnÞ ¼2aðnÞ 2bðnÞ

2cðnÞ 2dðnÞ

!xðn2 kÞ; ð35Þ

where {aðnÞ}; {bðnÞ}; {cðnÞ} and {dðnÞ} are four sequences of real numbers and

aðnÞ; dðnÞ . 0. Let

AðnÞ ¼2aðnÞ 2bðnÞ

2cðnÞ 2dðnÞ

!:

Then

jAðnÞj ¼

ffiffiffi2

p

2

�a2ðnÞ þ b2ðnÞ þ c2ðnÞ þ d 2ðnÞ

þ

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi½a2ðnÞ þ b2ðnÞ þ c 2ðnÞ þ d 2ðnÞ�2 2 4½aðnÞdðnÞ2 bðnÞcðnÞ�2

q �1=2

;

and for any x [ R2,

2AðnÞx · x ¼ aðnÞx21 þ dðnÞx2

2 þ ½bðnÞ þ cðnÞ�x1x2

$1

2aðnÞ þ dðnÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi½aðnÞ2 dðnÞ�2 þ ½bðnÞ þ cðnÞ�2

q� �jxj

2;

u f(u)/3 g(u)0 0.4141 0.6174p6

0.3108 0.4157p3

0.1687 0.2086p2

0 0

X. H. Tang et al.942

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

and

jAðnÞxj2¼ ½a2ðnÞ þ c2ðnÞ�x2

1 þ ½b2ðnÞ þ d 2ðnÞ�x22 þ 2½aðnÞbðnÞ þ cðnÞdðnÞ�x1x2

$1

2a2ðnÞ þ b2ðnÞ þ c2ðnÞ þ d 2ðnÞ

2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi½a2ðnÞ þ b2ðnÞ þ c 2ðnÞ þ d 2ðnÞ�2 2 4½aðnÞdðnÞ2 bðnÞcðnÞ�2

q �jxj

2

; m2ðnÞjxj2:

It follows that for x – 0,

2AðnÞx

jAðnÞxj·x

jxj$

2AðnÞxx

jAðnÞj jxj2

$1

2jAðnÞjaðnÞ þ dðnÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi½aðnÞ2 dðnÞ�2 þ ½bðnÞ þ cðnÞ�2

q� �:

Set

K ¼ infn$0

1

2jAðnÞjaðnÞ þ dðnÞ2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi½aðnÞ2 dðnÞ�2 þ ½bðnÞ þ cðnÞ�2

q� �� �:

Then by Theorem 2, if

2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiaðnÞdðnÞ

p$ bðnÞ þ cðnÞ;

X1s¼0

mðsÞ ¼ 1;

and

lim supn!1

Xns¼n2k

jAðsÞj , qKðaÞ;

then every solution of (35) tends to zero as n!1, where a [ ð0; 1Þ is defined by (2) and

mðnÞ ¼

ffiffiffi2

p

2a2ðnÞ þ b2ðnÞ þ c2ðnÞ þ d 2ðnÞ

2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi½a2ðnÞ þ b2ðnÞ þ c 2ðnÞ þ d 2ðnÞ�2 2 4½aðnÞdðnÞ2 bðnÞcðnÞ�2

q �1=2

:

Acknowledgements

The authors are grateful to the referees for their valuable comments and suggestions.

References

[1] Elaydi, S.N., 1996, An Introduction to Difference Equations (New York: Springer-Verlag).[2] Erbe, L.H., Xia, H. and Yu, J.S., 1995, Global stability of a linear nonautonomous delay difference equation.

Journal of Difference Equations and Applications, 1, 151–161.

Linear delay difference system 943

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3

[3] Grossman, S.E., 1972, Stability in n-dimensional differential-delay equations. Journal of MathematicalAnalysis and Applications, 40, 541–546.

[4] Kocic, V.L. and Ladas, G., 1993, Global Behavior of Nonlinear Difference Equations of Higher Order withApplications (Dordrecht: Kluwer Academic Publishers).

[5] Kuruklis, S.A., 1994, The asymptotic stability of xnþ1 2 axn þ bxn2k ¼ 0. Journal of Mathematical Analysisand Applications, 188, 719–731.

[6] Levin, S.A. and May, R.M., 1976, A note on difference-delay equations. Theoretical Population Biology,9, 178–187.

[7] Matsunaga, H. and Hara, T., 1999, The asymptotic stability of a two-dimesional linear delay differenceequation. Dynamics of Continuous, Discrete and Impulsive Systems, 6, 465–473.

[8] Papanicolaou, V.G., 1996, On the asymptotic stability of a class of linear difference equations. MathematicsMagazine, 69, 34–43.

[9] Tang, X.H., 2004, Stability in n-dimensional delay differential equations. Journal of Mathematical Analysisand Applications, 295, 485–501.

[10] Yu, J.S., 1998, Asymptotic stability for a linear difference equation with variable delay. Computers andMathematics with Applications, 36, 203–210.

X. H. Tang et al.944

Dow

nloa

ded

by [

Uni

vers

ity o

f Sa

skat

chew

an L

ibra

ry]

at 0

3:51

02

May

201

3