• Aucun résultat trouvé

An uncertainty principle on Sturm-Liouville hypergroups

N/A
N/A
Protected

Academic year: 2021

Partager "An uncertainty principle on Sturm-Liouville hypergroups"

Copied!
3
0
0

Texte intégral

(1)

An uncertainty principle on Sturm-Liouville hypergroups

By Radouan DAHERÞand Takeshi KAWAZOEÞ (Communicated by Heisuke HIRONAKA,M.J.A., Nov. 12, 2007)

Abstract: As an analogue of the classical uncertainty inequality on the Euclidean space, we shall obtain a generalization on the Sturm-Liouville hypergroupsðRþ;ðAÞÞ. Especially, we shall obtain a condition on Aunder which the discrete part of the Plancherel formula vanishes.

Key words: uncertainty principle; Sturm-Liouville; hypergroup.

1. Sturm-Liouville hypergroups. Sturm- Liouville hypergroups are a class of one-dimensional hypergroups on Rþ¼ ½0;1Þ with the convolution structure related to the second order differential operators

L¼ d2

dx2þA0ðxÞ AðxÞ

d dx;

where A satisfies the following conditions (see [1,2]):

A >0 onRþ¼ ð0;1Þ, and is inC2ðRþÞ; ð1Þ

on a neighborhood of 0;

ð2Þ

A0ðxÞ

AðxÞ ¼2þ1

x þBðxÞ;

where 12 and (a) if >0, B and B0 are integrable, (b) if ¼0, logxB and xlogxB0 are integrable, (c) if12 < <0,x2Bandx2þ1B0are integrable, (d) if¼ 12,B0is integrable,

A0

A 0 onRþ and limx!1A0ðxÞ AðxÞ ¼2;

ð3Þ

1 2

A0 A

0

þ1 4

A0 A

2

2 is integrable at1: ð4Þ

SinceA0=A¼ ðlogAÞ0, (3) implies thatAis increas- ing, and thus, Að0Þ<1. Under the conditions (1) to (3), the second order differential equation:

Luþ ð2þ2Þ ¼0, 2C, has a unique solution satisfying uð0Þ ¼1, u0ð0Þ ¼0, which we denote by . Furthermore, under (4), if =0, then there exists another solution ðxÞ, which behaves as

ffiffiffiffiffiffiffiffi p=2 ffiffiffiffiffiffi

px

Hð1Þ at 1, where Hð1Þ is the Hankel function. Similarly, we have ðxÞfor =0, and for2Rþ, there existsCðÞ 2Csuch thatðxÞ ¼ CðÞ ðxÞ þCðÞ ðxÞ.

LetC1c;eðRÞdenote the set ofC1even functions f onR. Forf2Cc;e1ðRÞthe Fourier transformff^is defined by

f^ fðÞ ¼

Z 1 0

fðxÞðxÞAðxÞdx:

Then the inverse transform is given as fðxÞ ¼X

2D

ffðÞ^ ðxÞ

þ 1 2

Z 1 0

f^

fðÞðxÞ d jCðÞj2; where D is a finite set in the interval ið0; Þ and ¼ kk2L2ðRþ;AdxÞ. We denote this decomposition as

fþfP

and we call fP and f the principal part and the discrete part of f respectively. We denote by FðÞ ¼ ðFðÞ;fagÞ a function on Rþ[D defined by

FðÞ ¼ FðÞ if¼2Rþ a if¼2D.

We putFðÞ ¼ ðFðÞ;fagÞand define the product ofFðÞ ¼ ðFðÞ;fagÞandGðÞ ¼ ðGðÞ;fbgÞas

ðFGÞðÞ ¼ ðFðÞGðÞ;fabgÞ:

Letd denote the measure onRþ[Ddefined by Z

Rþ[D

FðÞd¼X

2D

a

þ 1 2

Z 1 0

FðÞjCðÞj2d:

#2007 The Japan Academy

2000 Mathematics Subject Classification. Primary 43A62, 43A32, 42A38.

Þ De´partement de Mathe´matiques et d’Informatique, Faculte´ des Siences, Universite´ Hassan II, BP. 5366, Maarif, Casablanca, Maroc.

Þ Department of Mathematics, Keio University, 5322 Endo, Fujisawa, Kanagawa 252-8520, Japan.

Nos. 9–10] Proc. Japan Acad.,83, Ser. A (2007) 167

(2)

Forf 2Cc;e1ðRÞ, we put

ff^ðÞ ¼ ðff^ðÞ;fffðÞgÞ:^

Then the Parseval formula onCc;e1ðRÞcan be stated as follows: Forf; g2C1c;eðRÞ

Z 1 0

fðxÞgðxÞAðxÞdx¼ Z

Rþ[D

ff^ðÞggðÞd:^ ð5Þ

The map f !ff^, f2C1c;eðRÞ, is extended to an isometry between L2ðAÞ ¼L2ðRþ; AðxÞdxÞ and L2ðÞ ¼L2ðRþ[D; dÞ. Actually, each function f inL2ðAÞis of the form

fðxÞ ¼X

2D

ff^ðxÞ

þ 1 2

Z 1 0

f^

fðÞðxÞjCðÞj2d

¼fþfP

and theirL2-norms are given as Z 1

0

jfðxÞj2AðxÞdx¼X

2D

jff^j2;

Z 1 0

jfPðxÞj2AðxÞ ¼ 1 2

Z 1 0

jff^PðÞj2jCðÞj2d:

Therefore, if we define ff^ðÞ ¼ ðffðÞ;^ fff^gÞ, then kfkL2ðAÞ¼ kff^kL2ðÞ holds. In particular, if f 2 Cc;e1ðRÞ, thenff^¼ff^ðÞfor all2D.

2. Uncertainty inequality. We retain the notations in the previous sections. We put for x2Rþ,

aðxÞ ¼ Z x

0

AðtÞdtand vðxÞ ¼ aðxÞ ð6Þ AðxÞ

and for2C,

wðÞ ¼ ð2þ2Þ1=2:

Theorem 2.1. For allf 2L1ðAÞ \L2ðAÞ, kfvk2L2ðAÞ

Z

Rþ[D

jff^ðÞj2wðÞ2d1

4kfk4L2ðAÞ; ð7Þ

where the equality holds if and only iffis of the form

fðxÞ ¼ce

Z x 0

vðtÞdt

for somec; 2Cand < <0.

Proof. Without loss of generality we may suppose that f 2Cc;e1ðRÞ. Since ðLfÞ^ðÞ ¼ f^

fðÞð2þ2Þ ¼ffðÞwðÞ^ 2 and wðÞ is positive onRþ[D, the Parseval formula (5) yields that

Z

Rþ[D

jff^ðÞj2wðÞ2d¼ Z 1

0

ðLfÞðxÞfðxÞAðxÞdx

¼ Z 1

0

jf0ðxÞj2AðxÞdx:

Hence it follows that Z 1

0

jfðxÞj2vðxÞ2AðxÞdx Z

Rþ[D

jff^ðÞj2wðÞ2d

¼ Z 1

0

jfðxÞj2vðxÞ2AðxÞdx Z 1

0

jf0ðxÞj2AðxÞdx

Z 1 0

<ðfðxÞf0ðxÞÞvðxÞAðxÞdx

2

¼1 4

Z 1 0

ðjfðxÞj2Þ0aðxÞdx

2

¼1 4

Z 1 0

jfðxÞj2AðxÞdx

2

:

Here we used the fact thata0¼A(see (6)). Clearly, the equality holds if and only if fv¼cf0 for some c2C, that is, f0=f¼c1v. This means that logðfÞ ¼c1

Z x 0

vðtÞdtþC and thus, the desired

result follows.

Remark 2.2. WhenðRþ;ðAÞÞis the Bessel- Kingman hypergroup, the equality holds for ex2,

< <0. However, when it is the Jacobi hypergroup, each function satisfying the equality has an expo- nential decayex.

Since w2ðÞ ¼2þ2, (7) can be rewritten as follows:

Corollary 2.3. Let f be the same as in Theorem 2.1.

kfvk2L2ðAÞ

Z

Rþ[D

jff^ðÞj22d ð8Þ

1

4kfk2L2ðAÞ

Z 1 0

jfðxÞj2ð142vðxÞ2ÞAðxÞdx:

3. Vanishing condition of the discrete part. We shall prove that under the assumption:

0vðxÞ 1 2; ð9Þ

it follows thatD¼ ;. We suppose thatD6¼ ;and we takef ¼,2D. Then, sinceff^ðÞ ¼1if¼ and 0 otherwise, it follows from (8) that

kfvk2L2ðAÞ2 1

4kfk2L2ðAÞ

Z 1 0

jfðxÞj2ð142vðxÞ2ÞAðxÞdx:

168 R. DAHERand T. KAWAZOE [Vol. 83(A),

(3)

Here we recall that2<0, becauseDið0; Þand 142vðxÞ20 by (9). This is contradiction.

Therefore, we obtain the following Theorem 3.1. If0v 1

2, then D¼ ;. For example, ifAsatisfies the inequality:

aðxÞA0ðxÞ ¼ Z x

0

AðxÞdxA0ðxÞ A2ðxÞ;

ð10Þ

thenAsatisfies (9). Actually, (10) implies

v0ðxÞ ¼A2ðxÞ aðxÞA0ðxÞ A2ðxÞ 0:

Hence v is increasing on Rþ and vðxÞ ¼aðxÞ=

AðxÞ AðxÞ=A0ðxÞ because A=A0>0 by (3). Then it follows from (3) thatAsatisfies (9).

Corollary 3.2. If A satisfies the inequality (10), thenD¼ ;.

Remark 3.3. It is well-known that D¼ ; for Che´bli-Trime´che hypergroups where A0=A is decreasing and (4) is not required (cf. [1]). This fact easily follows from our argument. Since A=A0 is increasing and 0A=A01=2 by (3), we see that aA=2 by integration and thus, (9) holds.

HneceD¼ ;by Theorem 3.1.

4. Uncertainty principle. We suppose thatD¼ ;. Then (8) is of the form:

kfvk2L2ðAÞ

1 2

Z 1 0

jff^ðÞj22 d jCðÞj2 1

4kfk2L2ðAÞ

Z 1 0

jfðxÞj2ð142vðxÞ2ÞAðxÞdx:

Sincevis increasing,vð0Þ ¼0, and142vðxÞ20 by (9), it follows that f and ff^ both cannot be concentrated around the origin.

In general, if D6¼ ;, then we must pay atten- tion to the discrete part off to consider uncertainty principles. We refer to [3] for the Jacobi hyper- groups.

References

[ 1 ] W. R. Bloom and H. Heyer,Harmonic analysis of probability measures on hypergroups, de Gruyt- er, Berlin, 1995.

[ 2 ] O. Bracco, Fonction maximale associe´e a` des ope´rateurs de Sturm-Liouville singuliers, The`se, Univ. Louis Pasteur, Strasbourg, 1999.

[ 3 ] T. Kawazoe, Uncertainty principles for the Jacobi transforms, Tokyo J. Math. (to appear).

Nos. 9–10] Uncertainty principle on hypergroups 169

Références

Documents relatifs

He proved that if the boundary conditions are fixed and the reduced spectrum is slightly perturbed (in P), then there exists a symmetric potential which gives

Finally, we note the additional corollary below which, in essence, assures us t h a t Theorem 5.1 has non-vacuous content in certain specific circumstances... This

zeros of two linearly independent solutions of a second order linear differential. equation separate each other,

The class of operators considered in our study is a non-linear gene- ralization of a bounded self-adjoint operator, namely the class of abstract variational

Slepian, Prolate spheroidal wave functions, Fourier analysis and uncertainty IV: Extensions to many dimensions; generalized prolate spheroidal functions, Bell System Tech. Widom,

We consider a Cauchy problem for a Sturm-Liouville type differential inclusion in non separable Banach spaces under Filippov type assumptions and prove the existence of solutions..

Montrer que toute solution de (2) non identiquement nulle admet un nombre fini de z´ eros sur tout intervalle compact3. Montrer que toute solution non identiquement nulle de (2) a

[34] Yannick Privat, Emmanuel Tr´ elat and Enrique Zuazua, Complexity and regularity of optimal observation domains for the wave equation with fixed initial data, Discrete Cont.