• 検索結果がありません。

Hardy's inequalities for Hermite and Laguerre expansions

N/A
N/A
Protected

Academic year: 2022

シェア "Hardy's inequalities for Hermite and Laguerre expansions"

Copied!
8
0
0

読み込み中.... (全文を見る)

全文

(1)

Hardy's inequalities for Hermite and Laguerre expansions

著者 Kanjin Yuichi

journal or

publication title

Bulletin of the London Mathematical Society

volume 29

number 3

page range 331‑337

year 1997‑05‑01

URL http://hdl.handle.net/2297/3509

(2)

AND LAGUERRE EXPANSIONS

Dedicated to Professor Satoru Igari on his 60th birthday

YUICHI KANJIN

A

The well-known inequality of Hardy for Fourier coefficients of functionsf(t)C3¢n=−¢bneintin the real Hardy space is3¢n=−¢rbnr}(rn1)!¢. We shall establish analogues of this inequality for the Hermite function expansions and also for the Laguerre function expansions.

1. Introduction

Hardy’s inequality says that iff(t)C3¢n=−¢bneint is in ReH", then 3¢

n=−¢

rbnr rnr­1!¢,

where ReH" is the real Hardy space consisting of the boundary values of the real parts of functions in the Hardy spaceH"on the unit disk in the plane. The aim of this paper is to establish analogues of this inequality for the Hermite function expansions and also for the Laguerre function expansions.

Let(n(x) be the Hermite function defined by

(n(x)¯ ²π"/#2nΓ(n­1)´"/#Hn(x)e−x#/#, whereHn(x) is the Hermite polynomial of degreengiven by Hn(x)¯(®1)nexp (x#) (d}dx)nexp (®x#).

Then the system²(n´¢n=!is complete orthonormal on the real line2with respect to the ordinary Lebesgue measuredx(compare [6, 5.7]). This system leads to the formal expansion of a function f(x) on2:

f(x)C3¢

n=!

cn(f)(n(x),

wherecn(f)¯!¢¢f(x)(n(x)dxis thenth Hermite–Fourier coefficient off(x).

Received 8 September 1994 ; revised 29 May 1996.

1991Mathematics Subject Classification42C10, 33C45.

Partly supported by the Grants-in-Aid for Scientific Research, The Ministry of Education, Science and Culture, Japan.

Bull.London Math.Soc. 29 (1997) 331–337

(3)

332  

Let,αn(x),α"®1, be the Laguerre function defined by

,αn(x)¯

(

Γ(nΓ(n­α­­1)1)

*

"/#Lαn(x)e−x/#xα/#,

where Lαn(x)¯(n!)"xαex(d}dx)n²xn+αe−x´ is the Laguerre polynomial of degree n and of order α. Then ²,αn´¢n=! is a complete orthonormal system on the interval (0,¢) with respect todx(compare [6, 5.7]). We have the formal expansion

g(x)C3¢

n=!

cαn(g),αn(x) of a function g(x) on (0,¢), where cαn(g)¯!¢

! g(x),αn(x)dx is the nth Laguerre–

Fourier coefficient.

LetH"(2) be the real Hardy space consisting of the boundary values of the real

parts of functions in the Hardy spaceH"(2#+) on the upper half plane2#+. InH"(2), we consider the norm induced byH"(2#+). For a functiong(x) on (0,¢), we define the extensiongh(x) ofg(x) to2bygh(x)¯g(x) if 0!x, andgh(x)¯0 ifx%0. We define

H"(0,¢)¯ ²g(x) on (0,¢) ;gh(x)`H"(2)´,

sgsH"(!,¢)¯sghsH"(2).

Our main results of this paper are as follows.

T. (i) There exists a constant C such that 3¢

n=!

rcn(f)r

(n­1)#*/$'%CsfsH"(2) (1)

for f(x)C3¢n=!cn(f)(n(x)in H"(2).

(ii) Letα&0.Then there exists a constant C such that 3¢

n=! rcαn(g)r

n­1 %CsgsH"(!,¢) (2)

for g(x)C3¢n=!cαn(g),αn(x)in H"(0,¢).

The proof of the Theorem will be given in the next section. The atomic decomposition characterization of Hardy spaces will play an essential role in the proof. Also, the transplantation theorem (see (9) below) of R. Askey [1, p. 401, line 14] for Laguerre coefficients will be of use in the proof of part (ii) of the Theorem. The deduction of Paley’s theorem from the Theorem will be discussed in Section 3 with a remark.

For a historical survey on Hardy’s inequality and Paley’s theorem, we may refer to [2, p. 398, Comments]. The exposition of the role of the atomic decomposition characterization in the classial harmonic analysis is in R. R. Coifman and G. Weiss [3]. We shall also consult J. Garcia-Cuerva and J. L. Rubio de Francia [4] for the Hardy space theory.

2. Proof of the Theorem

AnH"atom is a functiona(x),x`2, supported in an interval (b,b­h) satisfying

ra(x)r%h" and !a(x)dx¯0. The space H"(2) is characterized in terms of atoms

(compare [4, Chapter 3]) : f(x) belongs toH"(2) if and only if f(x)¯3¢j=jaj(x),

(4)

where every aj(x) is an H" atom and 3¢j=!rλjr!¢. The H"(2) norm of f(x) is equivalent to inf3¢j=!rλjr, the infimum being taken over all decompositions. The space H"(0,¢) is also characterized as follows [4, Lemma 7.40] : g(x) belongs to H"(0,¢) if and only ifg(x)¯3¢j=jaj(x), where everyaj(x) is anH"atom satisfying suppa(x)Z(0,¢) and3¢j=!rλjr!¢.

We shall first give a lemma. Let²φn´¢n=!be an orthonormal system on an interval (c,d). For a function f(x) on (c,d), we consider the formal expansion

f(x)C3¢

n=!

bn(fn(x), wherebn(f)¯!b

cf(x)φn(x)dxis thenth Fourier coefficient with respect to the system

²φn´.

L. Suppose that max

c!x!d

)

dxd φn(x)

)

%Knδ, δ"#,

with a constant K independent of n.Then there exist constants C

"and C

#satisfying the following.

(i) For an H"atom a(x)withsuppa(x)Z(c,d),

rbn(a)r%C"nδsas## (3)

and

3¢

n=!

rbn(a)r

(n­1)(δ+#)/$%C#. (4) (ii) Let f(x)be a function of the form f(x)¯3¢j=jaj(x),where eŠery aj(x)is an H"

atom withsuppa(x)Z(c,d)and 3¢j=!rλjr!¢.Then 3¢

n=!

rbn(f)r (n­1)(δ+#)/$%C

#3¢

j=!rλjr. (5)

Proof. Leta(x) be anH"atom with support contained in (b,b­h)Z(c,d) such

thatra(x)r%h". Since

bn(a)¯

&

cd

a(x)φn(x)dx®φn(b)

&

cd

a(x)dx¯

&

bb+h

a(x)²φn(x)®φn(b)´dx, it follows from Schwarz’s inequality and the mean value theorem that

rbn(a)r%sas#

(&

bb+h

(x®b)#dx

*

"/#cmax!x!d

)

dxd φn(x)

)

%C"nδsas#h$/#,

whereCK}3"/#. By the facth%sas##, we have the inequality (3).

For simplicity, we putγ¯sas'#/(#δ+")and σ¯(δ­2)}3. It follows from (3) that 3¢

n=! rbn(a)r

(n­1)σ¯

(

n3­3

n

*

(nrb­n(a)1)rσ

%C

"sas##3

n

nδ

(n­1)σ­sas#

(

n3

1

(n­1)#σ

*

"/#

%C!"(sas##γ(#δ+")/$­sas#γ−(#δ+")/')%C

#, whereC!"andC#are positive constants independent ofa(x).

(5)

334  

Sincesas"%1, the series3¢j=jaj(x) converges inL"(c,d). We see that (4) implies (5).

We now come to the proof of the Theorem. By virtue of the Lemma and the atomic decomposition characterization, it is enough to show, for part (i) of the Theorem, that

)

dxd (n(x)

)

%Cn&/"# (6)

forx`2. Here and below, Cdenotes a positive constant which may differ at each different occurrence. It follows from the equations

Hn(x)¯2xHn−

"(x)®2(n®1)Hn−

#(x) and (d}dx)Hn(x)¯2nHn−

"(x) (compare [6, (5.5.8), (5.5.10)]) that

(d}dx)(n(x)¯(n}2)"/#(n−"(x)­((n­1)}2)"/#(n+"(x).

By the inequality

r(n(x)r%Cn"/"# (7)

(compare [5, (2.23)]), we have (6), which completes the proof of part (i) of the Theorem.

We turn to the proof of part (ii) of the Theorem. Similarly, it suffices to show that

)

dxd,αn(x)

)

%Cn, x"0. (8)

But we shall be able to prove this inequality only in the case thatα&2 orα¯0. The case 0!α!2 will be treated by the transplantation theorem for Laguerre coefficients betweenαandα­2 [1, p. 401, line 14] :

rcαn(f)r%C

0

rcαn−+#"(f)r­3j=n¢ rcαj+#(f)r

0

nj

1

α/#j"

1

, n¯1, 2,. (9)

To prove (8), we use the equation (d}dx)Lαn(x)¯®Lαn−+""(x) (compare [6, (5.1.14)]).

It follows that d

dx,αn(x)¯ταn

0

®Lαn−+""(x)e−x/#xα/α2Lαn(x)e−x/#xα/#(®"#)Lαn(x)e−x/#xα/#

1

,

whereταn¯ ²Γ(n­1)}Γ(n­α­1)´"/#. By the inequality

r,αn(x)r%C, x"0, α&0 (10) (compare [5, (2.9)]), we have

)

dxd ,αn(x)

)

%C

0

ταnrLαn−+""(x)re−x/#xα/α2ταnrLαn(x)re−x/#xα/#x1

1

%C(A­B­1), say.

We divide the matter into two cases,nx&1 and 0!nx!1. Fornx&1, we have, by (10) and ταn%Cnα/#,

AαnrLαn−+""(x)re−x/#x(α+")/#x"/#%Cn(nx)"/#%Cn,

B

αnrLαn(x)re−x/#xα/#x"%Cn(nx)"%Cn.

(6)

Thus we have (8) for nx&1 when α&0. For 0!nx!1, we use the inequality rLαn(x)r%Cnα, 0!nx!1 [6, (7.6.8)]. It follows that A%Cnα/#nα+"xα/#%Cn(nx)α/#

%Cn when α&0, and that B%Cnα/#nαxα/#x"%Cn(nx)α/#"%Cn when α&2.

Note that the termBdoes not appear whenα¯0. We conclude that (8) holds when α&2 orα¯0. Therefore we have the desired inequality (2) whenα&2 orα¯0.

We shall interpolate between α¯0 and α¯2 by using (9). Let 0!α!2 and

g(x)`H"(0,¢). By (10), we note thatrcα!(g)r%CsgsL"(!,¢)%CsgsH"(!,¢). It follows

from (9) that 3¢

n=! rcαn(g)r

n­1¯rcα

!(g)r­3¢

n="

rcαn(g)r n­1

%CsgsH"(!,¢)­C

0

n=3¢"rcnαn−+­#"(g)1r­3¢

n="

1 n­13¢

j=n

rcαj+#(g)r

0

nj

1

α/#j"

1

.

We treat the last sum. Since²3jn="nα/#(n­1)jα/#%Cforα"0, it follows that 3¢

n="

1 n­13¢

j=n

rcαj+#(g)r

0

nj

1

α/#j"¯3j=¢"

(

n=3j"

nα/#

n­1

*

jα/#j"rcαj+#(g)r

%C3¢

j="j"rcαj+#(g)r. Therefore we have

3¢

n=! rcαn(g)r

n­1%C

0

sgsH"(!,¢)­3n=¢!rcnαn+­#(g)1r

1

forα"0, which completes the proof of part (ii) of the Theorem since inequality (2) withα&2 orα¯0 has been proved.

3. Paley type theorem

The following inequalities, (11) and (12), with respect to the Laguerre function system follow from Paley’s theorem (compare [7, Vol. II, Theorem (5.1)]) with respect to general systems ²φn´ of functions orthonormal and uniformly bounded, rφn(x)r%C.

Letα&0. Forg(x)C3¢n=!cαn(g),αn(x)`Lp(0,¢), 1!p%2, 3¢

n=!rcαn(g)rp(n­1)p−#%CsgspLp(!,¢). (11) For²bn´¢n=!with 3¢n=!rbnrq(n­1)q−#, 2%q!¢, there exists a function

g(x)C3¢n=!bn,αn(x)`Lq(0,¢) such that

sgsqLq(!,¢)%C3¢

n=!rbnrq(n­1)q−#. (12) Also, by applying Paley’s theorem to the Hermite function system, we have the same inequalities as (11) and (12), where we substitute cαn(g), g(x)`Lp(0,¢), by cn(f), f(x)`Lp(2), respectively. But, for the Hermite function system, we can obtain sharper inequalities by interpolating Hardy’s inequality (1) and Parseval’s equation.

(7)

336  

P. (i) Let1!p%2.Then there exists a constant C such that 3¢

n=!rcn(f)rp(n­1)#*(p−#)/$'%CsfspLp(2) (13) for f(x)C3¢n=!cn(f)(n(x)`Lp(2).

(ii) Let2%q!¢. Then there exists a constant C such that if ²bn´¢n=! satisfies 3¢n=!rbnrq(n­1)#*(q−#)/$'!¢, then

sfsqLq(2)%C3¢

n=!rbnrq(n­1)#*(q−#)/$', (14) where f(x)C3¢n=!bn(n(x)`Lq(2).

Proof. Forκ"0, we denote by lpκ,p&1, the space of sequences²bn´¢n=! such thats²bn´sp¯ ²3¢n=!rbnrp(n­1)#κ´"/p!¢. We define an operatorTκofLp(2) tolpκ byTκf¯ ²cn(f) (n­1)κ´¢n=!. We consider the caseκ¯29}36. Then by (1), we see that Tκis a bounded linear operator ofH"(2) tol"κ, which implies thatTκis of weak type

(H"(2),l"κ). Parseval’s equation says thatTκis an isometry ofL#(2) tol#κ; in particular,

Tκis of weak type (L#(2),l#κ). We follow line by line the proof [4, pp. 308–310] of the interpolation theorem between H" space and Lp space. We conclude that Tκ is a bounded operator ofLp(2) tolpκ, 1!p%2, which means (13).

The inequality (14) is obtained by a standard duality argument (compare [7, Vol. II, proof of Theorem (5.1)]).

Lastly, we give a remark. Instead of the interpolation theorem betweenH"space

andL"space, we may use the interpolation theorem between weak-Lpspaces. But it

will become clear in the sequel that the inequality (15) below, obtained by this method, is weaker than our inequality (13).

We now considerlpκ,Tκwithκ¯11}12. Parseval’s equation implies thatTκis of weak type (L#(2),l#κ). Moreover, we see thatTκ is of weak type (L"(2),l"κ). For, we put/t¯ ²n;rcn(f) (n­1)κr"t´fort"0, wheref(x)`L"(2). It follows from (7) that

t!rcn(f) (n­1)κr!Cn"/"#nκsfsL"(2)¯Cn"!/"#sfsL"(2). Thus we have

²t}sfsL"(2)´"#/"!!n.

This gives3n`/t(n­1)#κ%C²t}sfsL"(2)´"#(−#κ+")/"!¯CsfsL"(2)}t, which means that Tκis of weak type (L"(2),l"κ). By the Marcinkiewicz interpolation theorem, we see that Tκis a bounded operator of Lp(2) tolpκ, 1!p!2, that is,

3¢

n=!rcn(f)rp(n­1)""(p−#)/"#%CsfsLpp(2). (15)

I should like to thank the referee for informing me of Kwon and Littlejohn’s work (Annals of Numerical Mathematics2 (1995) 289–303) which might lead to a further study.

References

1. R. A, ‘A transplantation theorem for Jacobi coefficients ’,Pacific J.Math. 21 (1967) 393–404.

2. L. S. B et al. (eds), Collected papers of G.H.Hardy, Vol. III (Oxford University Press, London, 1969).

3. R. R. Cand G. W, ‘ Extensions of Hardy spaces and their use in analysis ’,Bull.Amer.Math.

Soc. 83 (1977) 569–645.

(8)

4. J. G-Cand J. L. R F,Weighted norm inequalities and related topics(North- Holland, Amsterdam, 1985).

5. C. M, ‘ Nikolskii-type inequalities for Laguerre and Hermite expansions ’, Functions, series, operators, Colloq. Math. Soc. Ja!nos Bolyai 35 (ed. J. Szabados, North-Holland, Amsterdam, 1982) 811–834.

6. G. S$,Orthogonal polynomials, Amer. Math. Soc. Colloq. Publ. 23 (Amer. Math. Soc., Providence, RI, 4th edn, 1975).

7. A. Z,Trigonometric series, Vols I, II (Cambridge University Press, London, 2nd edn, 1968).

Department of Mathematics College of Liberal Arts Kanazawa University Kanazawa 920-11 Japan

参照

関連したドキュメント

We obtained the condition for ergodicity of the system, steady state system size probabilities, expected length of the busy period of the system, expected inventory level,

Making use, from the preceding paper, of the affirmative solution of the Spectral Conjecture, it is shown here that the general boundaries, of the minimal Gerschgorin sets for

[9] DiBenedetto, E.; Gianazza, U.; Vespri, V.; Harnack’s inequality for degenerate and singular parabolic equations, Springer Monographs in Mathematics, Springer, New York (2012),

This article studies the existence, stability, self-similarity and sym- metries of solutions for a superdiffusive heat equation with superlinear and gradient nonlinear terms

Here we purpose, firstly, to establish analogous results for collocation with respect to Chebyshev nodes of first kind (and to compare them with the results of [7]) and, secondly,

From (3.2) and (3.3) we see that to get the bound for large deviations in the statement of Theorem 3.1 it suffices to obtain a large deviation bound for the continuous function ϕ k

In this paper, we …rst present a new de…nition of convex interval–valued functions which is called as interval–valued harmonically h–convex functions. Then, we establish some

Specializing the members of Chebyshev systems, several applications and ex- amples are presented for concrete Hermite–Hadamard-type inequalities in both the cases of