• No results found

Some Lp Inequalities for B Operators

N/A
N/A
Protected

Academic year: 2020

Share "Some Lp Inequalities for B Operators"

Copied!
12
0
0

Loading.... (view fulltext now)

Full text

(1)

Some L

p

Inequalities for B-Operators

Nisar Ahmad Rather, Sajad Hussain Ahangar Department of mathematics, University of Kashmir, Harzarbal, Sringar, India.

Email: [email protected], [email protected]

Received November 8, 2012; revised December 8, 2012; accepted December 15, 2012

ABSTRACT

If P z

 

is a polynomial of degree at most n having all its zeros in z 1, then it was recently claimed by Shah and Liman ([1], estimates for the family of $B$-operators, Operators and Matrices, (2011), 79-87) that for every ,

,

1

R

1

p

 

0, ,1 2

0

 

,

1

n

p p

p

R

B P z P z

z

    

  

 where B is a n-operator with parameters   0, ,1 2 in

the sense of Rahman [2], 

 

zRz and

3

2 8

n n

2

0 1 2 0 1 2

1

, , n

       

1

. Unfortunately the proof of this re-

sult is not correct. In this paper, we present certain more general sharp Lp-inequalities for -operators which not only

provide a correct proof of the above inequality as a special case but also extend them for as well.

n

 0 p

Keywords: Lp-Inequalities; -Operators; Polynomials

n

1. Introduction and Statement of Results

Let n denote the space of all complex polynomials

 

0

n j

j j

P z

a z of degree at most n. For Pn,

define

 

 

i

0 0

1

: exp log e d ,

P z   P  

 

 

i 1

0 1

: e , 0

p p

p

P z   P   p

   

 

 

1

: Max ,

z

P z P z

and denote for any complex function : the composite function of P and , defined by

, as

P

 

z :P

 

z

z

P.

A famous result known as Bernstein’s inequality (for reference, see [3, p. 531], [4, p. 508] or [5] states that if

, then

n

P

 

 

,

P zn P z (1.1)

whereas concerning the maximum modulus of P z

 

on

the circle z  R 1, we have

 

n

 

, 1

P Rz R P z R , (1.2)

(for reference, see [6, p. 442] or [3, Vol. 1, p. 137]).

Inequalities (1.1) and (1.2) can be obtained by letting in the inequalities

p 

 

 

,

p p

P z n P z p1 (1.3)

and

 

n

 

, 1, 0

p p

P RzR P z Rp , (1.4)

respectively. Inequality (1.3) was found by Zygmund [7] whereas inequality (1.4) is a simple consequence of a result of Hardy [8] (see also [9, Th. 5.5]). Since in- equality (1.3) was deduced from M. Riesz’s interpolation formula [10] by means of Minkowski’s inequality, it was not clear, whether the restriction on p was indeed essen-

tial. This question was open for a long time. Finally Arestov [11] proved that (1.3) remains true for 0 p 1

as well.

If we restrict ourselves to the class of polynomials

n

P having no zero in z 1, then Inequalities (1.1)

and (1.2) can be respectively replaced by

 

 

,

2

n

P zP z (1.5) and

 

1

 

1.

2

n

R

P Rz   P z R (1.6)

(2)

Ankey and Ravilin [13].

Both the Inequalities (1.5) and (1.6) can be obtain by letting p  in the inequalities

 

 

,

1

p p

p

P z

P z n p

z

 

 0 (1.7)

and for R1,p0,

 

1

 

.

1

n p

p p

p

R z

P Rz P z

z

 

 (1.8)

Inequality (1.7) is due to De-Bruijn [14] for . Rahman and Schmeisser [15] extended it for

1

p

0 p 1

whereas the Inequality (1.8) was proved by Boas and Rahman [16] for and later it was extended for

by Rahman and Schmeisser [15].

1

p

0 p 1

Q. I. Rahman [2] (see also Rahman and Schmeisser [4, p. 538]) introduced a class n of operators that

carries a polynomial into

B

n

P

 

 

 

 

 

0 1

2 2

:

2 1!

,

2 2!

P z nz

B P z P z

P z

nz

 

 

  

 



 

  

(1.9)

where  0, 1 and 2 are such that all the zeros of

 

 

 

2

0 1 2

: ,1 , 2

U z  C n zC n z (1.10)

where

 

!

, 0

! !

n

C n r r n

r n r

 

  , lie in half plane

2 .

z  z n

As a generalization of Inequality (1.1) and (1.5), Q. I. Rahman [2, inequality 5.2 and 5.3] proved that if

and then for

,

n

P Bn z 1,

 

 

n

0, ,1 2

  

,

B P z     P z (1.11)

and if Pn, P z

 

0 in z 1, then z 1,

 

 

0 1 2

0

 

1 , , ,

2 n

B P z        P z (1.12) where

0 1 2

0 1 2 2 3

1

, , .

2 8

n

n n

n

        (1.13)

As a corresponding generalization of Inequalities (1.2) and (1.4), Rahman and Schmeisser [4, p. 538] proved that if Pn, then z 1,

 

0, ,1 2

  

.

n n

B P zR     P z (1.14) and if Pn, P z

 

0 in z 1, then as a special case of Corollary 14.5.6 in [4, p. 539], we have

 

0 1 2 0

 

1

, , ,

2

n n

B P z

R P

     z

 

(1.15)

where 

 

z :Rz R, 1 and    n

0, ,1 2

is defined

by (1.13).

Inequality (1.15) also follows by combining the Inequalities (5.2) and (5.3) due to Rahman [2].

As an extension of Inequality (1.14) to Lp-norm, re-

cently Shah and Liman [1, Theorem 1] proved:

Theorem A. If Pn, then for every and

,

1

R

1

p

 

1, ,2 3

  

,

n

n p

p

B P zR     P z (1.16)

where Bn, 

 

zRz and    n

0, ,1 2

is de- fined by (1.13).

While seeking the analogous result of (1.15) in Lp

norm, they [1, Theorem 2] have made an incomplete attempt by claiming to have proved the following result:

Theorem B. If Pn, and does not vanish

for

 

P z

1,

zthen for each p1, R1,

 

1, ,2 3

0

 

, 1

n n

p

R

B P z P z

z

    

  

 (1.17)

p

where Bn, 

 

zRz and    n

1, ,2 3

is de- fined by (1.13).

Further, it has been claimed in [1] to have proved the Inequality (1.17) for self-inversive polynomials as well.

Unfortunately the proof of Inequality (1.17) and other related results including the key lemma [1, Lemma 4] given by Shah and Liman is not correct. The reason being that the authors in [1] deduce:

1) line 10 from line 7 on page 84,

2) line 19 on page 85 from Lemma 3 [1] and, 3) line 16 from line 14 on page 86,

by using the argument that if

 

: n

 

1

Pzz P z , then

for 

 

zRz, R1 and z 1,

 

 

,

B P  zB P   z

 

  

 

which is not true, in general, for every R1 and

1

z  . To see this, let

 

1 0

n k

n k

P za z   a z   a za

be an arbitrary polynomial of degree n, then

 

 

1

0 1

: 1

= .

n

n n n k

k n

P z z P z

a z a za za

    

Now with 1:1n 2 and 2:2n2 8, we have

 



0 1 2

0

1 ,

n

n k n k k

k

B P z

n k n k n k a z R

    

 

 

     

(3)

and in particular for z 1, we get

 



0 1 2

0

1 ,

n n

k n

k k

B P z R z

z

n k n k n k a

R

  

  

 

 

        

 

whence

 



0 1 2

0

1

k n

n

k k

B P z

z

R n k n k n k a

R

  

 

 

 

        

 

.

But

 

0 1 2

0

1 ,

k n

n

k k

B P z

z

R k k k a

R

  

 

 

 

 

     

 

 

so the asserted identity does not hold in general for every and

1

Rz 1 as e.g. the immediate counterexample of

 

: n demonstrates in view of

z

P z P

 

z 1,

 

0

B P z   and

 

2

3

0 1 2 2 1 8

BP   z    n n n

  

 

for z 1.

Authors [1] have also claimed that Inequality (1.17) and its analogue for self-inversive polynomials are sharp has remained to be verified. In fact, this claim is also wrong.

The main aim of this paper is to establish Lp-mean

extensions of the inequalities (1.14) and (1.15) for and present correct proofs of the results mentioned in [1]. In this direction, we first present the following result which is a compact generalization of the Inequalities (1.1), (1.2), (1.14) and (1.16) and also extend

Inequality (1.17) for as well.

0  p

0 p

 1

Theorem 1. If Pn then for  with  1,

and

0  p R r 1,

 

 

0, ,1 2

  

,

R r p

n n

n p

B P z B P z

R r P z

  

    

 

 

(1.18)

where Bn, t

 

ztz and    n

0, ,1 2

is given

by (1.13). The result is best possible and equality holds in (1.18) for

 

n.

P zz

0

If we choose   in (1.18), we get the following result which extends Theorem A to 0 p 1,

Corollary 1. If Pn then for 0  p and

1,

R

 

0, ,1 2

  

,

n

n p

p

B P zR     P z (1.19)

where Bn, 

 

zRz and    n

0, ,1 2

is given

by (1.13).

Remark 1. Taking 0 0 2 in (1.19) and noting that in this case all the zeros of U(z) defined in (1.10) lie

in z  z n2 , we get for R1 and 0  p

 

n 1

 

p p

P Rz nRP z ,

which includes (1.4) as a special case. Next if we choose

1 0 2

   in (1.19), we get inequality (1.4). Inequality (1.11) also follows from Theorem 1 by letting

in (1.18).

p 

Theorem 1 can be sharpened if we restrict ourselves to the class of polynomials P z

 

which does not vanish in

1

z  In this direction, we next present the following interesting compact generalization of Theorem B which yields Lp mean extension of the inequality (1.12) for

0  p which among other things includes a correct

proof of inequality (1.17) for 1 as a special

case.

p

  

Theorem 2. If Pn and does not vanish

for

 

P z

1

z then for   with  1, 0  p and R r 1,

 

 

 

 

0, ,1 2 1 0 1

R r p

n n

n

p p p

B P z B P z

R r z

P z z

  

      

  

 

(1.20)

where Bn, t

 

ztz and    n

0, ,1 2

is de-

fined by (1.13). The result is best possible and equality holds in (1.18) for P z

 

aznb, a b 1.

If we take  0 in (1.20), we get the following result which is the generalization of Theorem B for

but also extends it for

1

p 0  p

Corollary 2. If Pn and does not vanish

for

 

P z

1

z then for 0  p and R1,

 

0, ,1 2

0

 

, 1

n

n p

p p

p

R z

B P z P z

z

    

  

(1.21)

,

n

B 

 

zRz and    n

0, ,1 2

is defined by (1.13).

By triangle inequality, the following result is an immediately follows from Corollary 2.

Corollary 3. If Pn and does not vanish

for

 

P z

1

z then for 0  p and R1,

 

0, ,1 2

0

 

1

n n p

p

R

B P z P z

z

    

  

 (1.22)

p

,

n

B 

 

zRz and    n

0, ,1 2

is defined by (1.13).

Remark 2. Corollary 3 establishes a correct proof of a result due to Shah and Liman [1, Theorem 3] for and also extends it for

1

p

(4)

Remark 3. If we choose 0 0 2

 in (1.21), we

get for R1 and 0 p ,

 

1

 

1

n

p p

p

nR

P Rz P z

z

 

which, in particular, yields Inequality (1.7). Next if we take 1 0 2

 

in (1.21), we get Inequality (1.8). Ine-quality (1.12) can be obtained from corollary 2 by letting

in (1.20). p

By using triangle inequality, the following result im- mediately follows from Theorem 2.

Corollary 4. If and does not vanish

for

n

P P z

 

1,

z then for  with  1 0  p

and R r 1,

 

 

 

 

0, ,1 2 1 0 1

R r p

n n

n

p p

B P z B P z

R r

P z z

  

      

 

 

 

(1.23)

,

n

B t

 

ttz and    n

0, ,1 2

is defined by (1.13).

A polynomial is said be self-inversive if

where n

P

 

 

P z Pz  1 and is the con-

jugate polynomial of , that is,

 

Pz

 

 

P z Pz :z Pn

 

1z .

Finally in this paper, we establish the following result for self-inversive polynomials , which includes a correct proof of an another result of Shah and Liman [1, Theo- rem 2] as a special case.

Theorem 3. If and is a self-inversive

polynomial, then for

n

P P z

 

 with  1 0  p

and R r 1,

 

 

 

 

0, ,1 2 1 0

, 1

R r p

n n

n

p

p p

B P z B P z

R r z

P z z

  

      

  

 

(1.24) where Bn, t

 

ttz and    n

0, ,1 2

is given by (1.13). The result is sharp and an extremal polynomial

is

 

n

c az

0,

P z  a , ac0.

For   we get the following result.

Corollary 5. If and is a self-inversive

polynomial, then for and

n

P

0 p

 

P z

  R1,

 

 

0, ,1 2 0

, 1

p

n

n p

p p

B P z

R z

P z z

    

(1.25)

where Bn, 

 

zRz and    n

0, ,1 2

is given

by (1.13).

The following result is an immediate consequence of

Corollary 5.

Corollary 6 If Pn

0 p

and is a self-inversive polynomial, then for

 

P z

   and R1,

 

 

0, ,1 2 0

, 1

p

n n

p p

B P z

R

P z z

    

 

 (1.26)

where Bn, 

 

zRz and    n

0, ,1 2

is given

by (1.13).

Remark 4. Corollary 6 establishes a correct proof of a result due to Shah and Liman [1, Theorem 3] for and also extends it for

1

p

0 p 1 as well.

Remark 5. A variety of interesting results can be eas-ily deduced from Theorem 3 in the same way as we have deduced from Theorem 2. Here we mention a few of these. Taking 0 0 2

p

= in (1.25), we g r

1

R

et fo and 0  ,

 

1

 

1

n

p p

p

nR

P Rz P z

z

 

 ,

which, in particular, yields a result due to Dewan and Govil [17] and A. Aziz [18] for polynomials n .

Next if we choose

P

1 0 2

   in (1.25), we get for

1

R ; 0  p

 

1

 

1

n p

p p

p

R z

P Rz P z

z

 

 .

The above inequality is a special case of a result proved by Aziz and Rather [19].

Lastly letting p  in (1.25), it follows that if

 

P z , is a self-inversive polynomial then

 

0 1 2 0

 

1

, , ,

2

n n

B P z

R P

    

 

z

(1.27)

where Bn, 

 

zRz and    n

0, ,1 2

is

de-fined by (1.13). The result is sharp.

Inequality (1.27) is a special case of a result due to Rahman and Schmeisser [4, Cor. 14.5.6].

2. Lemma

For the proof of above theorems we need the following Lemmas:

The following lemma follows from Corollary 18.3 of [20, p. 86].

Lemma 1. If Pn and P z

 

has all zeros in

1,

z then all the zeros of B P

 

 

z also lie in

1.

z

Lemma 2. If Pn and P z

 

have all its zeros in

1

(5)

 

1

 

. 1

n

R

P Rz P rz

r

 

 

 

Proof. Since all the zeros of P z

 

lie in z 1, we

write

 

i

1

e j ,

n j j

P z C z r

where rj1. Now for 0< 2π, R r 1, we have

1 2

i 2 2

i

i 2

i

2 cos

e e

2 cos

e e

1

,for 1, 2, , . 1

j j

j j j

j

j j j

j

j

j

R r Rr

R r

r r rr r r

R r R

j n

r r r

 

 

   

    

 

  



  

   

 

   

 

  

 

Hence

 

i

i i

i

i i

1

1

R e R e e

e e e

1 1 ,

1 1

j j

n

j

j j

n n

j

P r

P r r r

R R

r r

 

 

 

 

   

 

 

   

for 0  2π. This implies for z 1 and R r 1,

 

1

 

,

1

n

R

P Rz P rz

r

 

 

which completes the proof of Lemma 2.

Lemma 3. If Pn and P z

 

has no zero in

1,

z then for every   with  1, and

> 1

R r

1

z  ,

 

 

 

 

,

R r

R r

B P z B P z

B P z B P z

  

  

   

 

 

  (2.1)

where

 

: n

 

1

Pzz P z and t

 

ztz.

 

P z

Proof. Since the polynomial has all its zeros in

1,

z  therefore, for every real or complex number 

with  1, the polynomial f z

 

P z

 

P

 

z ,

where

 

: n

 

1

Pzz P z has all zeros in z 1. App-

lying Lemma 2 to the polynomial f z

 

, we obtain for every R>r1 and 0  2π,

 

ei 1

 

ei 1

n

R

f R f r

r

    

  .

0

(2.2)

Since

 

ei for every

f R   R r 1, 0  2π

and R  1 r 1, it follows from (2.2) that

 

ei 1

 

ei

 

1

n

R

f R f R f r

r

   

 

i e , for every R r 1 and 0  2π. This gives

 

<

 

for 1, and 1.

f rz f Rz zRr

Using Rouche’s theorem and noting that all the zeros of f Rz

 

lie in z 1 R1, we conclude that the polynomial

 

 

 

 

 

 

 

z f rz

P Rz P rz P Rz P rz

  

     

T zf R

1

z

has all its zeros in for every real or complex  with  1 and R r 1.

Applying Lemma polynomial T z

 

and noti

tha

 

1 to ng

t B is a linear operator, it follows that all the zeros of polynomial

 

 

 

 

 

 

 

 

R r

R r

R r

f z B f z

B P z B P z

B P z B P z

  

  

   

 

   

 

 

 

 

B T zB

where t

 

ztz. 1

z  This implies

lie in

 

 

 

 

R r

B P  P z

R r

z B

B PzB Pz

   

  (2.3)

for z 1 and If Inequality (2.3) is not true, there

1.

R r z

then exits a point z0 with z0 1 such that

 

 

 

 

0 0

R r

B P zP

0 0

R r

B z

B PzB Pz

   

  (2.4)

But all the zeros of P

 

Rz lie in z 1R1,

therefore, it follows (as in case of f z

 

) that

zeros of

all the

 

 

PRz Prz lie in z 1 Hence, by

Lemma 1, we have

 

z0 B P r

z0

0.

R

B P   

We take

 

 

 

00

 

0 0

,

R r

R r

B P z B P z

B P z B P z

  

  

 

    

   

 

 

 

then  is well defined real or complex number with

1

  and with this choice of , we obtain

 

 

0 0

B T z  where z0 1. This contradicts the fact the zeros of

that all B T

 

 

z lie in z 1. Thus (2.3) holds true for  1 and  r 1.

Next we describe a result ov [1 For

R

of Arest 1].

,

     n1

0, ,1 n  and

n j

 

j 0 j n

P z

a z  , we define

 

n

0

.

j j j j

P z a z

 

 

(6)

serves one of the following properties:

1) P z

 

has all its zeros in

z: z 1 .

2) P z

 

has all its zeros in

z: z 1 .

The result of Arestov [11] may now be stated as follows.

4] Let

Lemma 4. [11, Theorem

 

x 

logx

where is a convex non decreasing function on . Then for all Pn and each admissible operator ,

 

 

i

0

i

e d

, e d ,

P

C n P

 

 

0  

where C

 

,n max

0 ,n

.

In particular, Lemma 4 applies with : p

x x

  for

every p

0,

. Therefore, we have

 

 

1

i

0 e d

1 2π i

0

, e d .

p p

p p

C n P

 

 

(2.5)

We use (2.5) to prove the following intere

Lemma 5. If and

P

 

 

 

sting result.

n

P P z

 

does not vanish in

1,

z then for every p0, R r 1 and for real, 0  2π,

 

 

 

i i

0 B P R B P r e

 

 i 

i i

i

0 1 2 0

i

0

e e

e e d

, , e 1

e d ,

R r

p

n n

p

B P B P

R r

P

 

   

      

 

    

   

   

 

 

(2.6)

where z

 

 

n

B , t

 

z :t ,

 

:

 

t t

B P  zB P   z  and    

0, ,1 2

is defined by (1.13).

f. Sinc and

Proo e Pn

 

:

 

1 n

Pzz P z , by

Lemma 3, we have for z 1,

 

 

 

 

,

R

B P z B P r z

z P z

 

 

 

Also, since

R r

B P   B

 (2.7)

 

 

1

1

n n n n

P Rz P rz

R z P Rz r z P rz

  

 

 

 

 

 

0

1 1 2

1

1 1 2

2

2 2

1 3 2 4

2 1

1 1

1 1

2

1 1

1 1

2! 2

2 1 1 1

1 1 2 1

R r

n n n n

n n n n

n n n n

n n

n n n n

n n n n

B P z B P z

R z P Rz r z P rz

nz

nR z P Rz R z P Rz

nr z P rz r z P rz

nz

n n R z P Rz

n R z P Rz R z P Rz

n n r z P rz n r z

  

 

 

  

  

   

  

    

   

 

 

 

 

 

 

 

  

   

 

 

3

2 4

1 1

n n

P rz

rzP rz

 

and therefore,

 

 

 

 

 

 

 

3 2

0 1 2

2

1 1

1 2

2

2 2 2 2

2

1

2 8

1

2 4

. 8

R r

R

B P 



r

n n

n n

n n

B P z B P z

z B P z

n n

n

R P z R r P z r

n n

n

R zP z R r zP z r

n

R z P z R r z P z r

  

 

   

  

 

 

 

    

   

   

  

  

   

 

  

 

  

 

 

 

 

 

 

 

(2.8) Also, for z 1,

 

 

 

 

= B   .

R r

R r

B P z B P z

P z B P z

  

    

   

    

   

 

 

 

 

 

Using this in (2.7), we get for z 1,

 

 

 

 

.

R r

B P z B P z

P z

  



 

As in the proof of Lemma 3, the polynomial

R r

B P   zB  

  

 

 

,

R r

P z P z

,

has all its zeros in z 1

and by Lemma 1,  R

 

z B P r

 

z ,

 

B P

also has all its zero in z 1, therefore,

 

r

 

z has all its zer

R

 

B P   z B P   os in

1.

z  Hence by the maximum modulus principle, for

1,

z

 

 

 

 

.

B P   zB P   z

      (2.9)

R r

R r

B Pz B P

   

(7)

A direct application of Rouche’s theorem shows that

with

 

0,

n n

P za z   a

 

 

 

 

 

i

e R

B PzB PzB Pz

3 2

i

0 1 2 0

3 2

i

0 1 2 0

0

1

e 1

2 8

1

e 1

2 8

,

R r

r

n n

n n

n n

P z

B P z

n n n

R r a z

n n n

R r a

 

     

     

 

    

 

        

   

 

 

    

 

      

   

 

 

 

has all its zeros in

 

     

1

z  for every real ,

0  2π. Therefore, 

ma 4,

0

is an admissible ope .5) of Lem the desired result foll

h

ce the ra

result.

rator. ows Applying (2

immediately for eac p .

From Lemma 5, we dedu following more gene l

Lemma 6. If Pn, then for every p0,

1

R r and real 0  2π,

 

 

 

 

 

i i i

B   

0

i e

R r

p

R

 i

i

0 1 2 0

i

0

e e e

e d

, , e 1

e d ,

r

p

n n

p

P B P

B P B P

R r

P

 

  

   

  

      

 

 

   

 

(2.10)

Proof. Let and let be the zeros

of . If

n

P z z1, , ,2 zn

 

P z zj 1 for , then the

resu llows b a 5. assume that

zero i

all 1, 2, , Henceforth, we

n

j n

lt fo

 

P z ha

y Lemm

s at least one z 1 so that we can

write

 

   

 

1 2

,

k n

j j

P z P z P z

a z z

1 1

0 1, 0,

j j k

z z

k n a

  

   

where the zeros z z1, , ,2 zk of P z1

 

lie in z 1 and the zeros zk1,zk2, , zn of P z2

 

lie in z 1.

First we suppose that P z1

 

has no zero on z 1 so that all the zeros of P z1

 

lie in z 1. Since all the lie in zeros of

nk

th degree polynomial P z2

 

1

z  , all the ze jugate po omial

 

roes of its con lyn

 

2 n k 2 1

PzzP z lie in z 1 and

 

 

2 2

PzP z for z 1. Now consid e poly-

nomial

er th

 

   

 

1 2

f zP z Pz

1 1

1 j ,

j j k

a z zz

  

k j

n

then all the zeroes of z

 

f z lie in z 1, and for

1,

z

 

 

 

   

 

1 2

1 2 .

P z P P z

 

f z P z P z

z

 

(2.11)

Therefore, it follows by Rouche’s Theorem that the polynomial g z

 

P z

 

f z

 

has all its zeros in

1

z  for every , with  1 so of

that all the zeros

 

 

T zgz lie in z 1 for some  1. Apply- ing (2.9) and (2.8) to the polynomial T z

 

, we get for

1

 and

R z 1,

 

 

 

 

 

 

 

3

0 1 2

2

1 2

1 1

2

2 2 2 2

2

1

2 8

1

2 4

, 8

n n

n n

n n

z

n n

n

R T z R r T z r

n n

n

R zT z R r zT z r

n

R z T z R r z T z r

  

 

 

 

 

  

    

 

 

  

  

 

 

 

 

 

that is,

2 <

R r

R r

B T z B T z

B T z B T

 

  

    

  

 

   

 

 

3 2 0

 1 2

2

1 2

1 1

2

2 2 2 2 2 2

2

1 <

2 8

1

2 4

, 8

R r

n n

n n

n n

B T z B T z

n n

n

R g z R r g z r

n n

n

R zg z R r zg z r

n

R z g z R r z g z r

  

  

  

 

    

     

 

 

  

 



 

  

  

 

 

 

 

 

 

(2.12)

for If ei ,0 < 2π

z    , then z

 

1 1

1.

z

as 1 and we get

ei

ei

R r

(8)

 

R

B g  z B g r

 

z has all zeros in

3 2

0 1 2

i i

2

1 2

1 i i 1 i i

2

2 i i 2 i i

2

1 <

2 8

e e

1

2 4

e e e e

e e e e ,

8

n n

n n

n n

n n

n

R g R r g r

n n

n

R g R r g r

n

R g R r g r

 

   

   

  

 

 

 

 

  

 

 

 

 

 

  

  

 

 

 

 

 

Equivalently, for

1

z  and thus B g R

 

z B g r

 

z does not vanish in z 1.

An application of Rouche’s theorem shows that the polynomial

 

 

 

 

 

i e

R r

R r

L z B g z B g z

B g z B g z

  

  

 

   

 

   

  (2.13)

has all zeros in z 1. Writing in

 

:=

 

 

g z P z f z and noting that B is a linear

operator, it follows that the polynomial

1,

z

 

 

 

 

 

 

 

 

 

i

i

e

e ,

R r

R r

R R

R r

L z B g z B g z

B g z B g z

B f z B f z

B f z B f z

  

  

   

  

 

   

 



   



 

 

 

 

 

 

 

 

(2.14)

 

 

 

 

,

R r

R

B g    

  r

B g z B g z

z B g z

  

  

  

 

 

 

where Since

 

.

t z tz

 

 

g z has all its zeros in z 1, it follows that g

 

z has its zeros in z 1 and hence (pro- rly as in proof of Lemma 3) the poly- nom

ceeding simila

ial gR

 

z gr

 

z also has all its zeros has all its zeros in z 1 for every  with  1.

in z 1. By Lemma 1, We claim

 

 

 

 

 

 

 

 

i

i e

e ,

R r R r

R R R r

B P z B P z B P z B P z

B f z B f z B f z B f z

 

     

     

   

 

   

  

   

   

 

 

 

 

(2.15)

for z 1. If Inequality (2.15) is not true, then there exists a point zz0 with z0 1 such that

 

 

 

 

 

 

 

 

i

0 0 0

i

0 0 0

e

e ,

R r R r

R R R r

B P z B P z B P z B P z

B f z B f z B f z B f z

 

     

     

   

 

   

  

   

   

 

 

 

 

0

0

Since f z

 

has all its zeros in z 1, proceeding similarly as in the proof of (2.13), it follows that

 

 

ei

 

 

0

R R R r

B f  z B f z   B f   z B f   z  for z 1 We take

 

 

 

 

 

 

 

 

i

0 0

i

0 0

e e

R r

B P z B P z

B f z B f z

 

  

 

 

 

  0 0

0 0

R r

R R R r

B P z B P z

B f z B f z

  

   

    

    

    

 

   

 

 

 

 

so that  is a well-defined real or complex number with  1 and with this choice of  , from (2.14), we get . This clearly is a contradiction to the fact

that

 

0 0

L z

 

L z has all its zeros in z 1. Thus (2.15) holds, which in particular gives for each p0 and  real,

 

 

 

i i i i

e e e e d

p

R r r

B PB P   B PB

 

 

 

 

i 0

i i i i i

0

e

e e e e e d .

R

p

R R R r

P

B fB f   B fB f

      

      



   

 

   

 

 

  

 

(9)

,

f gives for each p0, Lemma 4 and (2.7) applied to

 

 

 

 

 

 

i i i i i

0

i i

0 1 2 0 0

e e e e e d

, ,

, , e 1 e d .

p

R r R r

p p

p p

n n

B P B P B P B P

R r P

    

 

i i

0 1 2 e 1 0 0 e d

n n

R rf

     

 

       

   

 

    

     

(2.16)

Now if

     

    

, then applying (2.16) to the polynomial P z

 

P1

   

z P z2

 

1

P z has a zero on z 1 where

0  1, we get for each p0, R r 1 and  real,

 

 

 

 

 

i i i i i

0

0

e e e e e d

d

p

R r R r

B P   B P   B PB P

i i

0, ,1 2 e 1 0 e .

p p

n n

R rP

         

      

       

 

(2.17)

tting

       

    

Le 1 in (2.17) and using continuity, the desired result follows immediately and this proves Lemma 6. Lemma 7. If Pn, then for every p0, R r 1 and 0  2π,

 

 

 

 

 

2π 2π i

0 0

ii

0 1 2 0

0 0

e d

, , e 1 d e d ,

p

R r R r

p p

n n

B P z B P z B P z B P z

R r P

 

d

       

        

   

 

   

 

   

(2.18)

where Bn, t

 

ztz and    n

0, ,1 2

0.

is de by

tremal polynomial for any b

fined (1.13). The result is best possible and is an ex

Proof. By Lemma 6, for each

 

n

P zbz

0

p , 0  2π and R r 1, the Inequality (2.6) holds. Since

 

 

R r

B P   z B P  z is the conjugate polynomial of B P R

 

z B P r

 

z ,

 

 

 

ei

 

e

R r R r

B P   z B P  zB P    B P   i , and therefore for each p0, R r 1 and 0  2π , we have

 

 

 

 

 

 

 

 

 

 

 

 

i i i i

0 B P e B

  i

i i i i i

0

i i i i

0

e e e e d

e e e e e d

e e e e d .

p

R r R

References

Related documents