Oscillation results for difference equations with oscillating coefficients

14
OSCILLATION RESULTS FOR DIFFERENCE EQUATIONS WITH SEVERAL OSCILLATING COEFFICIENTS G. E. CHATZARAKIS, M. LAFCI, AND I. P. STAVROULAKIS H Abstract. This paper presents a new su¢ cient condition for the oscillation of all solutions of di/erence equations with several deviating arguments and oscillating coe¢ cients. Corresponding di/erence equations of both retarded and advanced type are studied. Examples illustrating the results are also given Keywords: Oscillating coe¢ cients, retarded argument, advanced argu- ment, oscillatory solutions, nonoscillatory solutions. 2010 Mathematics Subject Classication: 39A10, 39A21. 1. INTRODUCTION In the present paper, we study the oscillatory behavior of the solutions of the retarded di/erence equation x(n)+ m X i=1 p i (n)x( i (n)) = 0, n 2 N 0 , (E R ) where N 3 m 2, p i , 1 i m, are sequences of real numbers and f i (n)g n2N0 , 1 i m, are sequences of integers such that i (n) n 1, n 2 N 0 , and lim n!1 i (n)= 1, 1 i m (1.1) and the (dual) advanced di/erence equation rx(n) m X i=1 p i (n)x( i (n)) = 0, n 2 N (E A ) where N 3 m 2, p i , 1 i m, are sequences of real numbers and f i (n)g n2N , 1 i m, are sequences of integers such that i (n) n +1, n 2 N, 1 i m. (1.2) Here, N 0 = f0; 1; 2;:::g and N = f1; 2;:::g. Also, as usual, denotes the forward di/erence operator x(n)= x(n +1) x(n) and r denotes the backward di/erence operator rx(n)= x(n) x(n 1). By a solution of (E R ), we mean a sequence of real numbers fx(n)g nw which satises (E R ) for all n 2 N 0 : Here, w := min n2N0 1im i (n) 2 N 0 . H Corresponding author : I. P. Stavroulakis; email address: [email protected]; tel. +30-26510- 08283; Greece . 1

Transcript of Oscillation results for difference equations with oscillating coefficients

OSCILLATION RESULTS FOR DIFFERENCE EQUATIONSWITH SEVERAL OSCILLATING COEFFICIENTS

G. E. CHATZARAKIS, M. LAFCI, AND I. P. STAVROULAKISH

Abstract. This paper presents a new su¢ cient condition for the oscillationof all solutions of di¤erence equations with several deviating arguments andoscillating coe¢ cients. Corresponding di¤erence equations of both retardedand advanced type are studied. Examples illustrating the results are alsogiven

Keywords: Oscillating coe¢ cients, retarded argument, advanced argu-ment, oscillatory solutions, nonoscillatory solutions.

2010 Mathematics Subject Classi�cation: 39A10, 39A21.

1. INTRODUCTION

In the present paper, we study the oscillatory behavior of the solutions of theretarded di¤erence equation

�x(n) +

mXi=1

pi(n)x(� i(n)) = 0, n 2 N0, (ER)

where N 3 m � 2, pi, 1 � i � m, are sequences of real numbers and f� i(n)gn2N0 ,1 � i � m, are sequences of integers such that

� i(n) � n� 1, n 2 N0, and limn!1

� i(n) =1, 1 � i � m (1.1)

and the (dual) advanced di¤erence equation

rx(n)�mXi=1

pi(n)x(�i(n)) = 0, n 2 N (EA)

where N 3 m � 2, pi, 1 � i � m, are sequences of real numbers and f�i(n)gn2N,1 � i � m, are sequences of integers such that

�i(n) � n+ 1, n 2 N, 1 � i � m. (1.2)

Here, N0 = f0; 1; 2; : : :g and N = f1; 2; : : :g. Also, as usual, � denotes the forwarddi¤erence operator �x(n) = x(n+1)�x(n) and r denotes the backward di¤erenceoperator rx(n) = x(n)� x(n� 1).By a solution of (ER), we mean a sequence of real numbers fx(n)gn��w which

satis�es (ER) for all n 2 N0: Here,w := � min

n2N01�i�m

� i(n) 2 N0.

HCorresponding author : I. P. Stavroulakis; email address: [email protected]; tel. +30-26510-08283; Greece .

1

2 G. E. CHATZARAKIS, M. LAFCI, AND I. P. STAVROULAKISH

It is clear that, for each choice of real numbers c�w; c�w+1; : : : ; c�1; c0, thereexists a unique solution fx(n)gn��w of (ER) which satis�es the initial conditionsx(�w) = c�w; x(�w + 1) = c�w+1; : : : ; x(�1) = c�1; x(0) = c0.By a solution of the advanced di¤erence equation (EA), we mean a sequence of

real numbers fx(n)gn2N0 which satis�es (EA) for all n 2 N.A solution fx(n)gn��w

�fx(n)gn2N0

�of (ER) [(EA)] is called oscillatory (around

zero), if for any positive integer n0 � �w [n0 � 0] there exist n1; n2 � n0 such thatx(n1)x(n2) � 0. Otherwise, the solution is said to be nonoscillatory.In the last few decades, the oscillatory behavior of all solutions of di¤erence

equations has been extensively studied when the coe¢ cients pi(n) are nonnegative.However, for the general case when pi(n) are allowed to oscillate, it is di¢ cultto study the oscillation of (ER) [(EA)], since the di¤erence �x(n) [rx(n)] of anynonoscillatory solution of (ER) [(EA)] is in general oscillatory. Therefore, the resultson oscillation of di¤erence and di¤erential equations with oscillating coe¢ cients arerelatively scarce. Thus, a small number of paper are dealing with this case. See,for example, [1�16] and the references cited therein.In 1992 Qian, Ladas and Yan [12], in 2000 Yu and Tang [16] and in 2001 Tang

and Cheng [13] derived oscillation conditions for a special case of equation (ER) thefollowing equation with one oscillating coe¢ cient and constant delay of the form

xn+1 � xn + pnxn�k = 0, n 2 N0,while in 1996 Yan and Yan [15] studied the di¤erence equation with several oscil-lating coe¢ cients of the form

xn+1 � xn +mXi=1

pi(n)xn�ki(n) = 0, n 2 N0,

under some additional conditions on the oscillating coe¢ cients.

For equations (ER) and (EA) with several oscillating coe¢ cients, very recently,Bohner, Chatzarakis and Stavroulakis [2,3] and Berezansky, Chatzarakis, Domosh-nitsky and Stavroulakis [1] established the following theorems.

Theorem 1.1 (See [2, Theorem 2.4]). Assume (1.1) and that the sequences� i are increasing for all i 2 f1; : : : ;mg. Suppose also that for each i 2 f1; : : : ;mgthere exists a sequence fni(j)gj2N such that limj!1 ni(j) =1 and

pk(n) � 0 for all n 2m\i=1

8<:[j2N

[�(�(ni(j))); ni(j)] \ N

9=; 6= ;, 1 � k � m (1.3)

where�(n) = max

1�i�m� i(n), n 2 N0. (1.4)

If, moreover

lim supj!1

mXi=1

n(j)Xq=�(n(j))

pi(q) > 1, (1.5)

where n(j) = min fni(j) : 1 � i � mg, then all solutions of (ER) oscillate.

Theorem 1.2 (See [2, Theorem 3.4]). Assume (1.2) and that the sequences�i are increasing for all i 2 f1; : : : ;mg. Suppose also that for each i 2 f1; : : : ;mg

OSCILLATING COEFFICIENTS 3

there exists a sequence fni(j)gj2N such that limj!1 ni(j) =1 and

pk(n) � 0 for all n 2m\i=1

8<:[j2N

[ni(j); �(�(ni(j)))] \ N

9=; 6= ;, 1 � k � m, (1.6)

where�(n) = min

1�i�m�i(n), n 2 N. (1.7)

If, moreover

lim supj!1

mXi=1

�(n(j))Xq=n(j)

pi(q) > 1, (1.8)

where n(j) = max fni(j) : 1 � i � mg, then all solutions of (EA) oscillate.

Theorem 1.3 (See [3, Theorem 2.1]). Assume (1.1) and that the sequences� i are increasing for all i 2 f1; : : : ;mg. Suppose also that for each i 2 f1; : : : ;mgthere exists a sequence fni(j)gj2N such that limj!1 ni(j) =1,

pk(n) � 0 for all n 2m\i=1

8<:[j2N

[� i(� i(ni(j))); ni(j)] \ N

9=; 6= ;, 1 � k � m (1.9)

and

lim supn!1

mXi=1

pi(n) > 0 for all n 2m\i=1

8<:[j2N

[� i(� i(ni(j))); ni(j)] \ N

9=; . (1.10)

If, moreover

lim infj!1

mXi=1

ni(j)�1Xq=� i(ni(j))

pi(q) >1

e, (1.11)

then all solutions of (ER) oscillate.

Theorem 1.4 (See [3, Theorem 3.1]). Assume (1.2) and that the sequences�i are increasing for all i 2 f1; : : : ;mg. Suppose also that for each i 2 f1; : : : ;mgthere exists a sequence fni(j)gj2N such that limj!1 ni(j) =1,

pk(n) � 0 for all n 2m\i=1

8<:[j2N

[ni(j); �i(�i(ni(j)))] \ N

9=; 6= ;, 1 � k � m (1.12)

and

lim supn!1

mXi=1

pi(n) > 0 for all n 2m\i=1

8<:[j2N

[ni(j), �i(�i(ni(j)))] \ N

9=; . (1.13)

If, moreover

lim infj!1

mXi=1

�i(ni(j))Xq=ni(j)+1

pi(q) >1

e, (1.14)

then all solutions of (EA) oscillate.

4 G. E. CHATZARAKIS, M. LAFCI, AND I. P. STAVROULAKISH

Theorem 1.5 (See [1, Theorem 2.1]). Assume that (1.1) holds, the sequences � iare increasing for all i 2 f1; : : : ;mg and � is de�ned by (1.4). Suppose also that foreach i 2 f1; : : : ;mg there exists a sequence fni(j)gj2N such that limj!1 ni(j) =1,

pk(n) � 0 for all n 2m\i=1

8<:[j2N

[�(�(ni(j))); ni(j)] \ N

9=; 6= ;, 1 � k � m (1.3)

and

� := lim infj!1

mXi=1

n(j)�1Xq=�(n(j))

pi(q), (1.15)

where n(j) = min fni(j) : 1 � i � mg.If 0 < � < 1, and

lim supj!1

mXi=1

n(j)Xq=�(n(j))

pi(q) > 1��2

4 (1� �) , (1.16)

then all solutions of (ER) oscillate.

Theorem 1.6 (See [1, Theorem 3.1]). Assume (1.2) holds, the sequences �iare increasing for all i 2 f1; : : : ;mg and � is de�ned by (1.7). Suppose also that foreach i 2 f1; : : : ;mg there exists a sequence fni(j)gj2N such that limj!1 ni(j) =1,

pk(n) � 0 for all n 2m\i=1

8<:[j2N

[ni(j), �(�(ni(j)))] \ N

9=; 6= ;, 1 � k � m (1.6)

and

� := lim infj!1

mXi=1

�(n(j))Xq=n(j)+1

pi(q), (1.17)

where n(j) = max fni(j) : 1 � i � mg.If 0 < � < 1, and

lim supj!1

mXi=1

�(n(j))Xq=n(j)

pi(q) > 1��2

4 (1� �) , (1.18)

then all solutions of (EA) oscillate.

In the present paper, the authors study further the equation (ER) [(EA)] and

essentially improve the upper bound of the ratio x(�(n))=x(n+1)hx(n�1)x(�(n))

ifor pos-

sible nonoscillatory solutions x of (ER) [(EA)], when neither (1.5) [(1.8)] nor (1.11)[(1.14)] is satis�ed, and derive a new su¢ cient oscillation condition. Examplesillustrating the results are also given.

2. RETARDED EQUATIONS

In this section, we present a new su¢ cient condition for the oscillation of allsolutions of (ER) when the conditions (1.5) and (1.11) are not satis�ed, under theassumption that the sequences � i are increasing for all i 2 f1; : : : ;mg. To that end,the following lemma provides a useful tool.

OSCILLATING COEFFICIENTS 5

Lemma 2.1. Assume that (1.1) holds, the sequences � i are increasing for alli 2 f1; : : : ;mg, � is de�ned by (1.4), � is de�ned by (1.15) and (x(n))n��w is anonoscillatory solution of (ER). Suppose also that for each i 2 f1; : : : ;mg thereexists a sequence fni(j)gj2N such that limj!1 ni(j) =1, and

pk(n) � 0 for all n 2m\i=1

8<:[j2N

[�(�(ni(j))); ni(j)] \ N

9=; 6= ;, 1 � k � m. (1.3)

If 0 < � � 1=2, then

lim infj!1

x(n(j) + 1)

x(�(n(j)))� 1

2

�1� ��

p1� 2�

�. (2.1)

Proof. Since the solution fx(n)gn��w of (ER) is nonoscillatory, it is eithereventually positive or eventually negative. As f�x(n)gn��w is also a solution of(ER), we may restrict ourselves only to the case where x(n) > 0 eventually.By (1.3), it is obvious that there exists j0 2 N such that

pk(n) � 0 for all n 2m\i=1

[�(�(ni(j0))); ni(j0)] \ N, 1 � k � m (2.2)

and

x(�k(n)) > 0 for all n 2m\i=1

[�(�(ni(j0))); ni(j0)] \ N, 1 � k � m. (2.3)

In view of (2.2) and (2.3), (ER) gives

x(n+ 1)� x(n) = �mXi=1

pi(n)x(� i(n)) � 0,

for every n 2mTi=1

[�(�(ni(j0))); ni(j0)] \ N. This guarantees that the sequence x is

decreasing onmTi=1

[�(�(ni(j0))); ni(j0)] \ N.

Also, by (1.15) we have

mXi=1

n(j0)�1Xq=�(n(j0))

pi(q) � �� ", (2.4)

where " is an arbitrary real number with 0 < " < �. It is clear that there exists a�xed natural number n�(j0) � n(j0) such that

mXi=1

n�(j0)�1Xq=n(j0)

pi (q) <�� "2

andmXi=1

n�(j0)Xq=n(j0)

pi (q) ��� "2

. (2.5)

This is because in the case where pi (q) < ��"2 , there exists n�(j0) > n(j0) such

that (2.5) is satis�ed, while in the case where pi(q) � ��"2 , then n�(j0) = n(j0),

and thereforemXi=1

n�(j0)�1Xq=n(j0)

pi (q) =mXi=1

n(j0)�1Xq=n(j0)

pi (q) (by which we mean) = 0 <�� "2

6 G. E. CHATZARAKIS, M. LAFCI, AND I. P. STAVROULAKISH

andmXi=1

n�(j0)Xq=n(j0)

pi (q) =mXi=1

n(j0)Xq=n(j0)

pi (q) � pi(n(j0)) ��� "2

.

That is, in both cases (2.5) is satis�ed.Now, we will show that

�(n�(j0)) � n(j0)� 1:

Indeed, in the case where pi(n(j0)) � ��"2 , since n�(j0) = n(j0), it is obvious

that �(n�(j0)) = �(n(j0)) � n(j0) � 1. In the case where pi(n(j0)) < ��"2 , then

n�(j0) > n(j0). Assume, for the sake of contradiction, that �(n�(j0)) > n(j0) � 1.Hence, n(j0) � �(n�(j0)) � n�(j0)� 1 and then

mXi=1

n�(j0)�1Xq=�(n�(j0))

pi (q) �mXi=1

n�(j0)�1Xq=n(j0)

pi (q) <�� "2

,

which contradicts (2.4). Thus, in both cases, we have �(n�(j0)) � n(j0)� 1.Therefore

mXi=1

n(j0)�1Xq=�(n�(j0))

pi (q) =mXi=1

n�(j0)�1Xq=�(n�(j0))

pi (q)�mXi=1

n�(j0)�1Xq=n(j0)

pi (q)

> (�� ")� �� "2

=�� "2

. (2.6)

Summing up (ER) from n(j0) to n�(j0), and using the fact that the function x isdecreasing and the function � (as de�ned by (1.4)) is increasing, we have

x(n(j0)) = x(n�(j0) + 1) +

mXi=1

n�(j0)Xq=n(j0)

pi (q)x(� i(q))

� x(n�(j0) + 1) +mXi=1

n�(j0)Xq=n(j0)

pi (q)x(�(q)),

or

x(n(j0)) � x(n�(j0) + 1) + x(�(n�(j0)))mXi=1

n�(j0)Xq=n(j0)

pi (q) ,

which, in view of (2.5), gives

x(n(j0)) � x(n�(j0) + 1) +�� "2

x(�(n�(j0))). (2.7)

Summing up (ER) from �(n�(j0)) to n(j0)� 1, and using the same arguments, wehave

x(�(n�(j0))) = x(n(j0)) +mXi=1

n(j0)�1Xq=�(n�(j0))

pi (q)x(� i(q))

� x(n(j0)) +mXi=1

n(j0)�1Xq=�(n�(j0))

pi (q)x(�(q)),

OSCILLATING COEFFICIENTS 7

or

x(�(n�(j0))) � x(n(j0)) + x(�(n(j0)� 1))mXi=1

n(j0)�1Xq=�(n�(j0))

pi (q) ,

which, in view of (2.6), gives

x(�(n�(j0))) > x(n(j0)) +�� "2

x(�(n(j0)� 1)). (2.8)

Combining inequalities (2.7) and (2.8), it is obvious that

x(n(j0)) >

��� "2

�2x(�(n(j0)� 1)) = b1x(�(n(j0)� 1)), (2.9)

where

b1 =

��� "2

�2.

Combining inequalities (2.7),(2.8) and (2.9), we have

x(n(j0)) >

�b1 +

�� "2

�x(�(n�(j0)))

>

�b1 +

�� "2

��x(n(j0)) +

�� "2

x(�(n(j0)� 1))�

>

�b1 +

�� "2

�2x(�(n(j0)� 1)) = b2x(�(n(j0)� 1)), (2.10)

where

b2 =

�b1 +

�� "2

�2.

Following the above procedure, we can inductively construct a sequence of positivereal numbers fb�g��1 with

b�+1 =

�b� +

�� "2

�2,

such thatx(n(j0)) > b�x(�(n(j0)� 1)) (� = 1; 2; ...). (2.11)

It is easy to prove that fb�g��1 is a strictly increasing and bounded sequence ofpositive real numbers. Therefore, it follows that lim�!1 b� exists as a positive realnumber. Set

` = lim�!1

b� .

Because of the de�nition of fb�g��1, it holds

` =

�`+

�� "2

�2,

i.e.,

` =1

2

h1� (�� ")�

p1� 2(�� ")

i,

or

` =1

2

h1� (�� ") +

p1� 2(�� ")

i.

8 G. E. CHATZARAKIS, M. LAFCI, AND I. P. STAVROULAKISH

In both cases, we have

` � 1

2

h1� (�� ")�

p1� 2(�� ")

iand consequently (2.11) yields

x(n(j0))

x(�(n(j0)� 1))� ` � 1

2

h1� (�� ")�

p1� 2(�� ")

i,

orx(n(j0) + 1)

x(�(n(j0)))� 1

2

h1� (�� ")�

p1� 2(�� ")

i.

Thus,

lim infj0!1

x(n(j0) + 1)

x(�(n(j0)))� 1

2

h1� (�� ")�

p1� 2(�� ")

i.

The last inequality holds true for all real numbers " with 0 < " < �. Hence, we canobtain (2.1).The proof of the lemma is complete.

Theorem 2.1. Assume that (1.1) holds, the sequences � i are increasing for alli 2 f1; : : : ;mg, � is de�ned by (1.4) and de�ne � by (1.15). Suppose also that foreach i 2 f1; : : : ;mg there exists a sequence fni(j)gj2N such that limj!1 ni(j) =1,and

pk(n) � 0 for all n 2m\i=1

8<:[j2N

[�(�(ni(j))); ni(j)] \ N

9=; 6= ;, 1 � k � m. (1.3)

If 0 < � � 1=2, and

lim supj!1

mXi=1

n(j)Xq=�(n(j))

pi(q) >1 + �+

p1� 2�

2, (2.12)

then all solutions of (ER) oscillate.

Proof. Assume, for the sake of contradiction, that fx(n)gn��w is an eventuallypositive solution of (ER). Then there exists j0 2 N such that (2.2) and (2.3) hold.Also, as it has already been shown in the proof of Lemma 2.1, the sequence x is

decreasing onmTk=1

[�(�(nk(j0))); nk(j0)] \ N.

Summing up (ER) from �(n(j0)) to n(j0), and using the fact that the functionx is decreasing and the function � (as de�ned by (1.4)) is increasing, we obtain

x(�(n(j0))) = x(n(j0) + 1) +mXi=1

n(j0)Xq=�(n(j0))

pi (q)x(� i(q))

� x(n(j0) + 1) + x(�(n(j0)))mXi=1

n(j0)Xq=�(n(j0))

pi (q) .

Consequently,mXi=1

n(j0)Xq=�(n(j0))

pi (q) � 1�x(n(j0) + 1)

x(�(n(j0))),

OSCILLATING COEFFICIENTS 9

which gives

lim supj0!1

mXi=1

n(j0)Xq=�(n(j0))

pi (q) � 1� lim infj0!1

x(n(j0) + 1)

x(�(n(j0))). (2.13)

Assume that 0 < � � 1=2 and (2.12) holds. Then by Lemma 2.1, inequality (2.1)is ful�lled, and so (2.13) leads to

lim supj0!1

mXi=1

n(j0)Xq=�(n(j0))

pi (q) �1 + �+

p1� 2�

2,

which contradicts condition (2.12). The proof of the theorem is complete.

3. ADVANCED EQUATIONS

In the case of the advanced di¤erence equation (EA) the following Lemma 3.1(analogue to Lemma 2.1) can be easily derived and an easy modi�cation of theproof of Theorem 2.1 leads to the following Theorem 3.1.

Lemma 3.1. Assume that (1.2) holds, the sequences �i are increasing for alli 2 f1; : : : ;mg, � is de�ned by (1.7), � is de�ned by (1.17) and (x(n))n�0 is anonoscillatory solution of (EA). Suppose also that for each i 2 f1; : : : ;mg thereexists a sequence fni(j)gj2N such that limj!1 ni(j) =1, and

pk(n) � 0 for all n 2m\i=1

8<:[j2N

[ni(j); �(�(ni(j)))] \ N

9=; 6= ;, 1 � k � m.

If 0 < � � 1=2, then

lim infj!1

x(n(j)� 1)x(�(n(j)))

� 1

2

�1� ��

p1� 2�

�.

Theorem 3.1. Assume (1.2) holds, the sequences �i are increasing for alli 2 f1; : : : ;mg, � is de�ned by (1.7) and de�ne � by (1.17). Suppose also that foreach i 2 f1; : : : ;mg there exists a sequence fni(j)gj2N such that limj!1 ni(j) =1,and

pk(n) � 0 for all n 2m\i=1

8<:[j2N

[ni(j), �(�(ni(j)))] \ N

9=; 6= ;, 1 � k � m. (1.6)

If 0 < � � 1=2, and

lim supj!1

mXi=1

�(n(j))Xq=n(j)

pi(q) >1 + �+

p1� 2�

2, (3.1)

then all solutions of (EA) oscillate.

Remark 3.1. When � ! 0, then the conditions (2.12) and (3.1) reduce tothe condition (1.5) and (1.8) respectively. However the improvement is clear when0 < � � 1=2. It is easy to see that

1

2

�1� ��

p1� 2�

�>

�2

4 (1� �) ;

10 G. E. CHATZARAKIS, M. LAFCI, AND I. P. STAVROULAKISH

which means that the condition (2.12) is weaker than the condition (1.16) and thecondition (3.1) is weaker than the condition (1.18).

4. EXAMPLES

The signi�cance of the results is illustrated in the following examples.

Example 4.1. Consider the retarded di¤erence equation

�x(n) + p1(n)x(n� 3) + p2(n)x(n� 4) = 0; n 2 N0, (4.1)

where p1(n) and p2(n) are oscillating coe¢ cients, as shown in Figure 1.

n0 11

1216

23

0n

)n(p1

)n(p2

23

4.0

02.0

3.0

12

Figure 1.

11

07.0

635.0

16

In view of (1.4), it is obvious that �(n) = n� 3: Observe that forn1(j) = 24j + 11; j 2 N,

we have p1(n) � 0 for every n 2 A, where

A =[j2N

[�(�(n1(j))); n1(j)] \ N =[j2N

[24j + 5; 24j + 11] \ N.

Also, forn2(j) = 24j + 9; j 2 N,

we have p2(n) � 0 for every n 2 B, where

B =[j2N

[�(�(n2(j))); n2(j)] \ N =[j2N

[24j + 3; 24j + 9] \ N.

Therefore

p1(n) � 0 and p2(n) > 0 for all n 2 A \B =[j2N

[24j + 5; 24j + 9] \ N 6= ;.

OSCILLATING COEFFICIENTS 11

Observe that

n(j) = min fni(j) : 1 � i � 2g = 24j + 9, j 2 N.

Now,

� = lim infj!1

2Xi=1

n(j)�1Xq=�(n(j))

pi(q)

= lim infj!1

24 24j+8Xq=24j+6

p1(q) +

24j+8Xq=24j+6

p2(q)

35 = 3 � 7100

+ 3 � 2100

= 0:27

lim supj!1

2Xi=1

n(j)Xq=�(n(j))

pi(q)

= lim supj!1

24 24j+9Xq=24j+6

p1(q) +

24j+9Xq=24j+6

p2(q)

35 = 4 � 7100

+ 3 � 2100

+635

1000= 0:975

Observe that

0:975 >1 + �+

p1� 2�

2' 0:974116499;

that is, condition (2.12) of Theorem 2.1 is satis�ed and therefore all solutions ofequation (4.1) oscillate.On the other hand,

0:975 < 1,

0:975 < 1� �2

4 (1� �) ' 0:975034246.

Observe that p1(n) � 0 for every n 2 A0 = A and p2(n) � 0 for every n 2 B0, where

B0 =[j2N

[�2(�2(n2(j))); n2(j)] \ N =[j2N

[24j + 1; 24j + 9] \ N.

Therefore

p1(n) � 0 and p2(n) > 0 for all n 2 A0 \B0 =[j2N

[24j + 5; 24j + 9] \ N 6= ;.

Also,

lim infj!1

2Xi=1

ni(j)�1Xq=� i(ni(j))

pi(q)

= lim infj!1

24 24j+10Xq=24j+8

p1(q) +

24j+8Xq=24j+5

p2(q)

35 = 3 � 7100

+ 4 � 2100

= 0:29 <1

e.

Therefore none of the conditions (1.5), (1.16) and (1.11) are satis�ed.

Example 4.2. Consider the advanced di¤erence equation

rx(n)� p1(n)x(n+ 2)� p2(n)x(n+ 3) = 0; n 2 N, (4.2)

12 G. E. CHATZARAKIS, M. LAFCI, AND I. P. STAVROULAKISH

where p1(n) and p2(n) are oscillating coe¢ cients, as shown in Figure 2.

n0 11

12 23

0n

)n(p1

)n(p2

23

5.0

06.012

Figure 2.

11

08.0

612.0

In view of (1.7), it is obvious that �(n) = n+ 2: Observe that for

n1(j) = 24j + 3; j 2 N,

we have p1(n) � 0 for every n 2 A, where

A =[j2N

[n1(j); �(�(n1(j)))] \ N =[j2N

[24j + 3; 24j + 7] \ N.

Also, for

n2(j) = 24j + 4; j 2 N,

we have p2(n) � 0 for every n 2 B, where

B =[j2N

[n2(j); �(�(n2(j)))] \ N =[j2N

[24j + 4; 24j + 8] \ N.

Therefore

p1(n) � 0 and p2(n) > 0 for all n 2 A \B =[j2N

[24j + 4; 24j + 7] \ N 6= ;.

Observe that

n(j) = max fni(j) : 1 � i � 2g = 24j + 4, j 2 N.

OSCILLATING COEFFICIENTS 13

Now

� = lim infj!1

2Xi=1

�(n(j))Xq=n(j)+1

pi(q)

= lim infj!1

24 24j+6Xq=24j+5

p1(q) +

24j+6Xq=24j+5

p2(q)

35 = 2 � 8100

+ 2 � 6100

= 0:28.

Also

lim supj!1

2Xi=1

�(n(j))Xq=n(j)

pi(q)

= lim supj!1

24 24j+6Xq=24j+4

p1(q) +

24j+6Xq=24j+4

p2(q)

35 = 3 � 8100

+ 2 � 6100

+612

1000= 0:972

Observe that

0:972 >1 + �+

p1� 2�

2' 0:971662479;

that is, condition (3.1) of Theorem 3.1 is satis�ed and therefore all solutions ofequation (4.2) oscillate.On the other hand,

0:972 < 1,

0:972 < 1� �2

4 (1� �) ' 0:972777777.

Observe that p1(n) � 0 for every n 2 A0 = A and p2(n) � 0 for every n 2 B0, where

B0 =[j2N

[n2(j); �2(�2(n2(j)))] \ N =[j2N

[24j + 4; 24j + 10] \ N.

Therefore

p1(n) � 0 and p2(n) > 0 for all n 2 A0 \B0 =[j2N

[24j + 4; 24j + 7] \ N 6= ;.

Also,

lim infj!1

2Xi=1

�i(ni(j))Xq=ni(j)+1

pi(q)

= lim infj!1

24 24j+5Xq=24j+4

p1(q) +

24j+7Xq=24j+5

p2(q)

35 = 2 � 8100

+ 3 � 6100

= 0:34 <1

e.

Therefore none of the conditions (1.8), (1.18) and (1.14) are satis�ed.

Acknowledgement The authors would like to thank both referees for the con-structive remarks which improved the presentation of the paper.

14 G. E. CHATZARAKIS, M. LAFCI, AND I. P. STAVROULAKISH

References

[1] L. Berezansky, G. E. Chatzarakis, A. Domoshnitsky and I. P. Stavroulakis, Oscilla-tions of di¤erence equations with several oscillating coe¢ cients, Abstract Appl. Anal.,(2014), article ID 392097, 9 pages, doi: 10.1155/2014/392097.

[2] M. Bohner, G. E. Chatzarakis and I. P. Stavroulakis, Qualitative behavior of solutionsof di¤erence equations with several oscillating coe¢ cients, Arab. J. Math.. 3 (2014),1�13.

[3] M. Bohner, G. E. Chatzarakis and I. P. Stavroulakis, Oscillation criteria for di¤erenceequations with several oscillating coe¢ cients, Bull. Korean Math. Soc., 51 (2014), (inpress).

[4] G. E. Chatzarakis, T. Kusano and I. P. Stavroulakis, "Oscillation conditions for di¤er-ence equations with several variable arguments, Math. Bohemica , (2014), (to appear).

[5] G. E. Chatzarakis, J. Manojlovic, S. Pinelas and I. P. Stavroulakis, Oscillation criteriaof di¤erence equations with several deviating arguments, Yokohama Math. J., 60(2014), (in press).

[6] N. Fukagai and T. Kusano, Oscillation theory of �rst order functional-di¤erentialequations with deviating arguments, Annl. Mat. Pura Appl., 136 (1984), 95�117.

[7] M. K. Grammatikopoulos, R. Koplatadze and I. P. Stavroulakis, On the oscillation ofsolutions of �rst-order di¤erential equations with retarded arguments, Georgian Math.J., 10 (2003), No. 1, 63�76.

[8] H. Khatibzadeh, An oscillation criterion for a delay di¤erence equation, Comp. Math.Appl. 57 (2009), 37�41.

[9] M. R. Kulenovic and M. K. Grammatikopoulos, First order functional di¤erentialinequalities with oscillating coe¢ cients, Nonlinear Anal., 8 (1984), 1043�1054.

[10] G. Ladas, G. S�cas and I. P. Stavroulakis, Functional-di¤erential inequalities andequations with oscillating coe¢ cients. Trends in theory and practice of nonlinear dif-ferential equations (Arlington, Tex., 1982), 277�284, Lecture Notes in Pure and Appl.Math., 90, Dekker, New York, 1984.

[11] X. Li, D. Zhu and H. Wang, Oscillation for advanced di¤erential equations withoscillating coe¢ cients, Internat. J. Math. Math. Sci., 33 (2003), 2109�2118.

[12] C. Qian, G. Ladas and J. Yan, Oscillation of di¤erence equations with oscillatingcoe¢ cients, Radovi Mathematicki, 8 (1992), 55�65.

[13] X. H. Tang and S. S. Cheng, An oscillation criterion for linear di¤erence equationswith oscillating coe¢ cients, J. Comput. Appl. Math., 132 (2001), 319�329.

[14] T. Xianhua, Oscillation of �rst order delay di¤erential equations with oscillating co-e¢ cients, Appl. Math. J. Chinese Univ. Ser. B, 15 (2000), 252�258.

[15] W. Yan and J. Yan, Comparison and oscillation results for delay di¤erence equationswith oscillating coe¢ cients, Internat. J. Math. Math. Sci., 19 (1996), 171�176.

[16] J. S. Yu and X. H. Tang, Su¢ cient conditions for the oscillation of linear delay di¤er-ence equations with oscillating coe¢ cients, J. Math. Anal. Appl. 250 (2000), 735�742.

Department of Electrical and Electronic Engineering Educators, School of Peda-gogical and Technological Education (ASPETE), 14121, N. Heraklio, Athens, Greece

E-mail address : [email protected], [email protected]

Department of Mathematics, Faculty of Sciences, Ankara University, Tando¼Gan,Ankara, 06100, Turkey

E-mail address : [email protected]

Department of Mathematics, University of Ioannina, 451 10 Ioannina, GreeceE-mail address : [email protected]