• 検索結果がありません。

A Remark on Polynomial Norms and

N/A
N/A
Protected

Academic year: 2022

シェア "A Remark on Polynomial Norms and"

Copied!
9
0
0

読み込み中.... (全文を見る)

全文

(1)

Photocopying permittedbylicenseonly the Gordon andBreachScience Publishersimprint.

Printed in Singapore.

A Remark on Polynomial Norms and

Their Coefficients

SUNG GUENKIM*

Topology andGeometryResearchCenter,KyungpookNational University,Taegu,Korea(702-701)

(Received20 December1998; Revised 29January1999)

Thispaper presents new lowerbounds for thenorms of2-homogeneous real-valued polynomialsonlpspaces for0<p< which aresharper than those recently given by the author.

Keywords: Coefficients;lpspaces; Polynomialnorms 1991 MathematicsSubjectClassification: 46B20;46E15

This note isconcernedwiththe generalproblemof the relation between the norm ofa polynomial and its coefficients. This type ofproblem has been studied in manycontexts[1-6,8-11]overtheyears,because of both its relevancetonon-trivialproblemsinmathematics andbecause ofits ourinherentinterest.Inthisnotewefocusourattentiononlower bounds forthenormsof2-homogeneousreal-valuedpolynomialson

lp

spaces. Recentlythe author

[9]

gave lower bounds for thenormsof 2- homogeneous real-valued polynomialson

lp

spaces for 0

<

p

< .

We

hereimprovethem.

Let Ebe arealBanach spacewiththeunit sphere

SE

and rn

>

2, a

naturalnumber.

79(mE)

denotesthe Banach spaceofm-homogeneous real-valued polynomials on

E,

endowed with the polynomial norm

*E-mail:[email protected].

(2)

IIPII

supllxll

IP(x)]. See

Dineen

[7]

for more details about thetheory ofpolynomialsonBanachspaces.

LEMMA

LetO7ScS’ cN

andx,

aijERfori,

jES’with i<j. Then

max

Ix+ Z

aijeiej

ek=d:l,kS’

i,jS’,i<j

Proof

Letm

>

2 beapositive integer. Itisenoughtoshow that

max

Ix+ Z

ek=-t-1, <k<_m

l<i<j<m ek=’+-l,l<k<m+l

l<i<j<m+l a#’eij

Let

max

[x

nt-

Z

aijeif-j

=:1:1, <k<m

<i<j<m

--x+

<_i<j<_m

Let forsomesignchoices

f-(,...,

%.

f-m+ aim+

ifx

+ 21<i<j<_m aijf-[f-j >

0 and

aim+ i

otherwise.Thenwehave

max

Ix +

et=+l, l<_k<_m+l

<_i<j<_m+

>x+

<i<j<m+

x+

<i<j<m

Z aim+lf-

l<i<m

(3)

LEMMA

2 Then

Letrn

>_

2 bea positive integer.Letx,aij

(1 <

<j

< m)

ER.

max

ek=4-1, l<_k<_m

x+ 2_

aijeiej

l<i<j<_m

Ixl+

<i<j<mmax

lail.

The equality holds

if

andonly

if

the following conditionsare

satisfied.

Withoutloss

of

generality,assumethatmax1_<i<j<_m

[a01- la21.

(a)

aij O

for

3

< <

j

<

m.

(b)

xal2alia2i

<

0 and

[ale[- la2glfor

each 3

< <

m.

(C) E3

_<i_< m

[ali[ _< min{lx[, la121}.

Proof

Useinduction onm. Ifm 2, then the lemma is truebecause

max{Ix + al, Ix a12[} Ixl + lal. Suppose

that the lemma is true

for 2,3,...,m-1. Without loss of generality, we may assume that

max

<_i<j<_m

[a/[- [a34[.

Put

M=e=-t-maxl<k<m

x+ Z

aijeiej

l<i<j<m

Substitutingel-4-1,weget

max

e=+l, 2<_k<_m

3<i<j<m

<M

aijeiej)

(1)

and

max

ek=-t-1,2<k<_m

3<i<j<m

(2)

(4)

By

adding

(1)

and

(2)

and thetriangleinequality,weget

ek=4-1, 2<k<m

3<j<m 3<_i<j<m

aijeiej

_<. (3)

Again, by substituting

2--+1

into

(3)

and adding each other and the triangleinequality,weget

max

Ix

-’t- ao’e.iej

ek=+l 3<k<m

3<i<j<m

_<

M.

(4)

By

induction hypothesis and

(4),

wehave

Ixl +

l<i<j<mmax

la+l- Ixl + 1a341- Ixl +

3<i<j<mmax

[aijl

<

Ck--q-1,max<_k<_m

Ixq-aije.ie.j

3<i<j<m

Suppose

that conditions

(a)-(c)

are satisfied. From now on, we will assumethat

l<i<j<m

Thenby

(a),

--al2l--l<_i<_m(ali--a2i)i3 Ix--al21nt-l<i<m(ali--a2i)i3

Withoutlossof generality,assumethatxa12

>

O.Then by

(b),

Ix + a21 + (ali + a2i)i

3<i<m

(5)

for any sign choices3,’’

m

and, by

(b)

and

(c),

Ix a121 + (ali- a2i)i

3<i<m

<_ Ix a21 +

2

Z lai[

3<i<m

<_ IX al2l

q- 2

min{Ix i, lal2l}

--Ixl-t-la12[ (6)

foranysign choices 3,’’’,

m.

Thus M=

Ixl + la121.

Letus provethe

necessary condition. Firstwewill proveitwhenm 4.Somecomputa- tionshows that

max

Ix+

aijeiej

ek=+l,

l<k<4[

<i<j<4

max{Ix +

a12 -4-

a34[-+-I(al3 + a23)

-4-

(a14 -+- a24)1,

IX

a124-

a341 + [(a13 a23)

4-

(a14 a24)]}.

By

somecalculation,weget

(a)

a34 0.

(b)

xa12alia2i

<

0and

lal,I--la2/I

fori=3, 4.

(C) E3

_<i_<4

_< min{lxl, lal2l}.

Letm

>

4.

Suppose

that M

Ixl

/

laa21.

Let 3

< i0 <Jo <_

mbe fixed.

Letcrbethepermutationon

{

1,2,...,

m}

suchthat

cr(3)=i0,

tr(4)=j0,

cr(i0)=3, or(j0)=4.

Define

bij

ar(i)r(j)for each <j_<m.

By

Lemma 1,

max x

+ Z b0"i

ek=+l, l_<k<4

l<i<j<4

<

max

ek=-+-1,l<k<m

x+

<i<j<rn

bijeij

=M

andbythe firstclaimofLemma 2,

max

k--Zt=1, l_<k_<4X

+ b0.i

j

l<i<j<4

Ixl +

max

Ibijl

l_<i<j_<4

(6)

SO

max

ek=+l,l<k<4

x-+- Z

bijeiej

l<i<j<4

Ixl +

l_<i<j<4max

Ib l.

By

the above argument for m--4 case, we have

0--b34--aiojo

and

xb12b13b24

xal2alioaljo

<_

0 and

[ali0[ Ib131 Ib24[--lau0l,

showing

(a)

and

(b). Suppose

that

(c)

isnot true.

By (a), (b), (5)

and the triangle inequality,

M

>

ek=+l, 3<k<mmax x

a12[ + 21

aliei

3<i<m

IX- a121 +

2

lail > Ix a=l +

2

min{Ix I, la9.l}

3<i<m

acontradiction. Thereforewecompletetheproof.

Remark Lemmain

[9]

canbe improvedasfollows. Let Ebeanormed space over a field

(C

or

R)

and m

>

2, a natural number. Let x,aij

(1 <

<j

< m)

EE.Then

x+

l<i<j<m

eiejaij max

{llxll Ila/ll}.

<i<j<m

UsingLemma 2,weobtainthemainresult of thispaper.

THEOREM 3 wehave

Let

P(x) 2i<_j

bijxixj

79(2/p), b/ R,

0

<

p

<_

c.Then

Ilpll

sup

f {

mN, (Wl,w2 wm,O )ESlp

Y biiw2i

<i<m

+

< <maxj<_m

Ibijwiwj[).

(7)

A REMARKONPOLYNOMIAL NORMS 7

Proof

Itfollows fromLemma2because

Ilell Ie(wa,...,

f.mwm,

O,O,..

- biiw"- bijwiwjiJ]

l<i<m l<i<j<m

for any

(Wl,

w2, Wm,

0,...)

E

Sll,

X

l<i<m biiw

and aij=

bijeiey

for anysignchoicesel,...,

em.

COROLLARY4 Thenwehave

(a)LetP(x) Yi<jbijxixj p(2/p),

bij R,0<p

< .

(b)

Let

P(x) -]i<jbijxixj p(2/), bij

R.Thenwehave

IIPII mENSUp{ <i<m bii +

l<i<j<mmax

Ibil}.

Proof (a)

followsbytakingwk

1/m

lipfork 1,2,...,m.

(b)

followsby takingwk 1 fork 1,2,...,m.

PROPOSITION 5

(a)

Thenwehave

Let

P(x)= Eil inair"inXi

Xin

(n/2)

air-i, K.

IIPII lai,...il

(b) If

then

(8)

Proof Use

inductionon n.

Casen 2.Forx

(Xk)

E

St2,

k

(by the H61der inequality)

k k,j

lajle

(by the Hblder inequality)

Suppose

that forn

_<

k,thepropositionis true.Forx

(Xk) Sty,

aiv..ik+Xil Xik+

il ik+l

(bythe inductionhypothesis)

(bythe H61der inequality)

References

[1] R.M. Aron, M. Lacruz, R.A. RyanandA.M. Tonge,The generalized Rademacher functions.NotediMate.(lecee)12(1992),15-25.

[2] R.M. Aronand I.Zalduendo, Polynomialsnormsand coefficients, ExtractaMath.

(1992),13-20.

(9)

[10]

[11]

[3] B. Beauzamy, Jensen’sinequality for polynomials withconcentrations atlow degrees, Numer.Math. 49(1986),221-225.

[4] B. Beauzamy, E.Bombieri,P.Enflo andH.Montgomery,Products ofpolynomialsin manyvariables, J. Number Th. 36(1990),219-245.

[5] B. BeauzamyandP.Enflo,Estimationsde produits depolynomes, J. NumberTh.21 (1985),390-420.

[6] B. Beauzamy,J.L.Frotand C. Millour, Representingamany-variable polynomialon ahypercube, preprint.

[7] S.Dineen, Complex analysisinlocally convex spaces,MathematicsStudies, Vol.57, North Holland(1981).

[8] P.Enflo,Ontheinvariantsubspace probleminBanach spaces,ActaMath.158(1987), 213-313.

[9] S.G.Kim,Aninequality concerning polynomialnormsand theircoefficients,Indian J.PureAppl. Math. 29(1998),277-283.

A.K.Rigler,S.Y.Trimble and R.S.Varga, Sharp lower bounds forageneralized Jensen inequality, Rocky Mtn. J. Math.19(1989),353-373.

I.Zalduendo,Anestimate formultilinearformson pspaces,Pro.R.IrishAcad.93A (1993),137-142.

参照

関連したドキュメント

Key words: Dunkl operators, Dunkl transform, Dunkl translation operators, Dunkl convolu- tion, Besov-Dunkl spaces.. Abstract: In this paper, we define subspaces of L p by

Theorem 1 For every graph G, the modular and integral flow and tension polyno- mials of G can be realized as Ehrhart polynomials of compressed, integral inside-out polytopes.... To

Finally, as an application of our second method, we give a lower bound for the number of certain partitions of n which are interesting in the modular representation theory of

It follows from Theorem 1 that when the momentum map is properly normalized (see Lemma 2) this stratification does not depend on the choice of ω in its cohomology class.. When Z is

Aluthge, “On p-hyponormal operators for 0 &lt; p &lt; 1,” Integral Equations and Operator Theory, vol.. Ito, “Some classes of operators associated with generalized

They proved that if Y is a (real or complex) rearrangement-invariant nonatomic function space on [0, 1] isometric to L p [0, 1] for some 1 ≤ p &lt; ∞ then the isometric isomorphism

This is reminiscent of the cµ rule, where priority is given according to the index c i µ i , known to be optimal under linear queue-length cost with weights c i : note in

Once we have applied the transformation from Lemma 2.2 we will obtain an elliptic operator in a domain Ω 0 with a matrix of the form (2.1), but no guarantee that the Lipschitz