BLAISE PASCAL
Bechir Amri & Mohamed Sifi
Riesz transforms for Dunkl transform Volume 19, no1 (2012), p. 247-262.
<http://ambp.cedram.org/item?id=AMBP_2012__19_1_247_0>
© Annales mathématiques Blaise Pascal, 2012, tous droits réservés.
L’accès aux articles de la revue « Annales mathématiques Blaise Pas- cal » (http://ambp.cedram.org/), implique l’accord avec les condi- tions générales d’utilisation (http://ambp.cedram.org/legal/). Toute utilisation commerciale ou impression systématique est constitutive d’une infraction pénale. Toute copie ou impression de ce fichier doit contenir la présente mention de copyright.
Publication éditée par le laboratoire de mathématiques de l’université Blaise-Pascal, UMR 6620 du CNRS
Clermont-Ferrand — France
cedram
Article mis en ligne dans le cadre du
Centre de diffusion des revues académiques de mathématiques
Riesz transforms for Dunkl transform
Bechir Amri Mohamed Sifi
Abstract
In this paper we obtain theLp-boundedness of Riesz transforms for the Dunkl transform for all 1< p <∞.
Transformées de Riesz associés à la transformée de Dunkl
Résumé
Dans cet article, nous étudions la bornitude des transformées de Riesz associées à la transformée de Dunkl sur les espacesLp, 1< p <∞.
1. Introduction
On the Euclidean space RN, N ≥ 1, the ordinary Riesz transform Rj, j = 1, ..., N is defined as the multiplier operator
R\j(f)(ξ) =−i ξj
kξkf(ξ).b (1.1)
It can also be defined by the principal value of the singular integral Rj(f)(x) =d0lim
ε→0
Z
kx−yk>ε
xj−yj
kx−ykf(y)dy
where d0 = 2N2 Γ(N+12 )
√π . It follows from the general theory of singular integrals that Riesz transforms are bounded onLp(RN, dx) for all 1< p <
∞. What is done in this paper is to extend this result to the context of Dunkl theory where a similar operator is already defined.
Dunkl theory generalizes classical Fourier analysis on RN. It started twenty years ago with Dunkl’s seminal work [3] and was further developed by several mathematicians. See for instance the surveys [5, 6, 7, 9] and the
Keywords:Dunkl transforms, Riesz Transforms, Singular integrals.
Math. classification:17B22, 32A55, 43A32, 42A45.
references cited therein. The study of theLp-boundedness of Riesz trans- forms for Dunkl transform onRN goes back to the work of S. Thangavelu and Y. Xu [10] where they established boundedness result only in a very special case of N = 1. It has been noted in [10] that the difficulty arises in the application of the classical Lp- theory of Caldéron-Zygmund, since Riesz transforms are singular integral operators. In this paper we describe how this theory can be adapted in Dunkl setting and gives anLp-result for Riesz transforms for all 1< p <∞. More precisely, through the fundamen- tal result of M. Rösler [6] for the Dunkl translation of radial functions, we reformulate a Hörmander type condition for singular integral operators.
The Riesz kernel is given by acting Dunkl operator on Dunkl translation of radial function.
This paper is organized as follows. In Section 2 we present some defi- nitions and fundamental results from Dunkl’s analysis. The Section 3 is devoted to proving Lp-boundedness of Riesz transforms. As applications, we will prove a generalized Riesz and Sobolev inequalities. Throughout this paper C denotes a constant which can vary from line to line.
Acknowledgments. The authors are partially supported by the project DGRST 04/UR/15-02 and the cooperation programs PHC Utique / CMCU 07G 1501 and 10G 1503.
2. Preliminaries
In this section we collect notations and definitions and recall some basic facts. We refer to [5, 3, 6, 7, 9].
Let G⊂ O(RN) be a finite reflection group associated to a reduced root system R and k : R → [0,+∞) be a G-invariant function (called multiplicity function). Let R+ be a positive root subsystem. We shall assume that R is normalized in the sense that kαk2 = hα, αi = 2 for all α ∈R, whereh, iis the standard Euclidean scalar product on RN.
The Dunkl operators Tξ, ξ ∈ RN are the following k–deformations of directional derivatives ∂ξ by difference operators :
Tξf(x) =∂ξf(x) + X
α∈R+
k(α)hα, ξif(x)−f(σα. x)
hα, xi , x∈RN
whereσα denotes the reflection with respect to the hyperplane orthogonal toα. For the standard basis vectors ofRN, we simply write Tj =Tej.
The operators ∂ξ and Tξ are intertwined by a Laplace–type operator Vkf(x) =
Z
RN
f(y)dµx(y),
associated to a family of compactly supported probability measures {µx|x∈RN}.
Specifically, µx is supported in the the convex hull co(G.x).
For every λ∈CN, the simultaneous eigenfunction problem, Tξf =hλ, ξif, ξ∈RN
has a unique solutionf(x) =Ek(λ, x) such thatEk(λ,0) = 1, which is given by
Ek(λ, x) = Vk(ehλ, .i)(x) = Z
RN
ehλ,yidµx(y), x∈RN. Furthermore λ7→Ek(λ, x) extends to a holomorphic function on CN.
Let mk be the measure onRN, given by dmk(x) = Y
α∈R+
|hα, xi|2k(α)dx.
Forf ∈L1(mk) (the Lebesgue space with respect to the measuremk) the Dunkl transform is defined by
Fk(f)(ξ) = 1 ck
Z
RN
f(x)Ek(−i ξ, x)dmk(x), ck = Z
RN
e−|x|
2
2 dmk(x).
This new transform shares many analogous properties of the Fourier trans- form.
(i) The Dunkl transform is a topological automorphism of S(RN) (Schwartz space).
(ii) (Plancherel Theorem) The Dunkl transform extends to an isomet- ric automorphism of L2(mk).
(iii) (Inversion formula) For everyf∈L1(mk) such thatFkf∈L1(mk), we have
f(x) =Fk2f(−x), x∈RN.
(iv) For allξ ∈RN and f ∈ S(RN)
Fk(Tξ(f))(x) =< iξ, x >Fk(f)(x), x∈RN. (2.1) Letx∈RN, the Dunkl translation operatorτxis defined onL2(mk) by, Fk(τx(f))(y) =Ek(ix, y)Fkf(y), y∈RN. (2.2) If f is a continuous radial function in L2(mk) withf(y) =fe(kyk), then
τx(f)(y) = Z
RN
fe qkxk2+kyk2+ 2< y, η >dµx(η). (2.3) This formula is first proved by M. Rösler [6] for f ∈ S(RN) and recently is extended to continuous functions by F. and H. Dai Wang [2].
We collect below some useful facts : (i) For all x, y∈RN,
τx(f)(y) =τy(f)(x). (2.4) (ii) For allx, ξ∈RN and f ∈ S(RN),
Tξτx(f) =τxTξ(f). (2.5) (iii) For all x∈RN and f, g∈L2(mk),
Z
RN
τx(f)(−y)g(y)dmk(y) = Z
RN
f(y)τxg(−y)dmk(y). (2.6) (iv) For allx∈RN and 1≤p≤2, the operator τx can be extended to
all radial functions f inLp(mk) and the following holds
kτx(f)kp,k≤ kfkp,k. (2.7) k.kp,k is the usual norm of Lp(mk).
3. Riesz transforms for the Dunkl transform.
In Dunkl setting the Riesz transforms (see [10]) are the operators Rj, j = 1...N defined onL2(mk) by
Rj(f)(x) =dklim
ε→0
Z
|y|>ε
τx(f)(−y) yj
kykpkdmk(y), x∈RN
where
dk = 2pk
−1 2
Γ(p2k)
√π ; pk = 2γk+N + 1 and γk= X
α∈R+
k(α).
It has been proved by S. Thangavelu and Y. Xu [10], thatRjis a multiplier operator given by
Fk(Rj(f))(ξ) =−i ξj
kξkFk(f)(ξ), f ∈ S(RN), ξ∈RN, (3.1) The authors state that if N = 1 and 2γk∈Nthe operatorRj is bounded onLp(mk), 1< p <∞. In [1] this result is improved by removing 2γk ∈N, where Riesz transform is called Hilbert transform. If γk= 0 (k= 0), this operator coincides with the usual Riesz transform Rj given by (1.1). Our interest is to prove the boundedness of this operator forN ≥2 andk≥0.
To do this, we invoke the theory of singular integrals. Our basic is the following,
Theorem 3.1. LetKbe a measurable function onRN×RN\{(x, g.x); x∈ RN, g∈G} and S be a bounded operator from L2(mk) into itself, associ- ated with a kernel K in the sense that
S(f)(x) = Z
RN
K(x, y)f(y)dmk(y), (3.2) for all compactly supported function f in L2(mk) and for a.e x ∈ RN satisfying g.x /∈supp(f), for allg∈G. If K satisfies
Z
ming∈Gkg.x−yk>2ky−y0k
|K(x, y)− K(x, y0)|dmk(x)≤C, y, y0 ∈RN, (3.3) then S extends to a bounded operator from Lp(mk) into itself for all1 <
p≤2.
Proof. We first note that (RN, mk) is a space of homogenous type, that is, there is a fixed constant C >0 such that
mk(B(x,2r))≤Cmk(B(x, r)), ∀x∈RN, r >0 (3.4) where B(x, r) is the closed ball of radius r centered at x (see [8], Ch 1).
Then we can adapt to our context the classical technic which consist to show that S is weak type (1,1) and conclude by Marcinkiewicz interpola- tion theorem.
In fact, the Calderón-Zygmund decomposition says that for all f ∈ L1(mk)∩L2(mk) and λ >0, there exist a decomposition of f, f =h+b with b =Pjbj and a sequence of balls (B(yj, rj))j =(Bj)j such that for some constantC, depending only on the multiplicity function k
(i) khk∞≤Cλ;
(ii) supp(bj)⊂Bj; (iii)
Z
Bj
bj(x)dmk(x) = 0;
(iv) kbjk1,k ≤C λ mk(Bj);
(v) X
j
mk(Bj)≤C kfk1,k
λ .
The proof consists in showing the following inequality hold for w=hand w=b :
ρλ(S(w)) =mk
{x∈RN; |S(w)(x)|> λ
2}≤C kfk1,k
λ . (3.5)
By using the L2-boundedness of S we get ρλ(S(h))≤ 4
λ2 Z
RN
|S(h)(x)|2dmk(x)≤ C λ2
Z
RN
|h(x)|2dmk(x). (3.6) From (i) and (v),
Z
∪Bj
|h(x)|2dmk(x)≤Cλ2µk(∪Bj)≤Cλkfk1,k. (3.7) Since on (∪Bj)c,f(x) =h(x), then
Z
(∪Bj)c
|h(x)|2dmk(x)≤Cλkf|1,k. (3.8) From (3.6), (3.7) and (3.8), the inequality (3.5) is satisfied for h.
Next we turn to the inequality (3.5) for the functionb. Consider Bj∗ =B(yj,2rj ); and Q∗j = [
g∈G
g.B∗j.
Then
ρλ(S(b))≤mk
[
j
Q∗j+mk{x∈ [
j
Q∗jc;|S(b)(x)|> λ 2}.
Now by (3.4) and (v) mk [
j
Q∗j≤ |G|X
j
mk(Bj∗)≤CX
j
mk(Bj)≤C kfk1,k λ .
Furthermore if x /∈Q∗j, we have
ming∈Gkg.x−yjk>2ky−yjk, y∈Bj. Thus, from (3.2),(iii) ,(ii), (3.3), (iv)and (v)
Z
(∪Q∗j)c
|S(b)(x)|dmk(x)
≤X
j
Z
(Q∗j)c
|S(bj)(x)|dmk(x)
=X
j
Z
(Q∗j)c
Z
RN
K(x, y)bj(y)dmk(y)dmk(x)
=X
j
Z
(Q∗j)c
Z
RN
bj(y)K(x, y)− K(x, yj)dmk(y)dmk(x)
≤X
j
Z
RN
|bj(y)|
Z
(Q∗j)c
|K(x, y)− K(x, yj)|dmk(x)dmk(y)
≤X
j
Z
RN
|bj(y)|
Z
g∈Gminkg.x−yjk>2ky−yjk|K(x, y)− K(x, yj)|dmk(x)dmk(y)
≤CX
j
kbjk1,k
≤C kfk1,k. Therefore,
mk{x∈ [
j
Q∗jc;|S(b)(x)|> λ 2}
≤ 2 λ
Z
(∪Q∗j)c
|S(b)(x)|dmk(x)≤C kfk1,k λ .
This achieves the proof of (3.5) for b.
Now, we will give an integral representation for the Riesz transformRj. For this end, we put for x, y∈RN and η∈co(G.x)
A(x, y, η) = q
kxk2+kyk2−2< y, η >= q
ky−ηk2+kxk2− kηk2. It is easy to check that
ming∈Gkg.x−yk ≤A(x, y, η)≤max
g∈G kg.x−yk. (3.9) The following inequalities are clear
∂A`
∂yr
(x, y, η)≤CA`−1(x, y, η),
∂2A`
∂yr∂ys
(x, y, η)≤CA`−2(x, y, η) (3.10) and for α∈R+,
∂A`
∂yr(x, σα.y, η)≤CA`−1(x, σα.y, η),
∂2A`
∂yr∂ys(x, σα.y, η)≤CA`−2(x, σα.y, η), (3.11) where r, s= 1, ...N and `∈R.
Let us set K(1)j (x, y) =
Z
RN
ηj−yj
Apk(x, y, η)dµx(η) Kj(α)(x, y) = 1
< y, α >
Z
RN
h 1
Apk−2(x, y, η) − 1
Apk−2(x, σα.y, η)
idµx(η), Kj(x, y) = dknK(1)j (x, y) + X
α∈R+
k(α)αj
pk−2K(α)j (x, y)o, where α∈R+.
Proposition 3.2. If f ∈L2(mk) with compact support, then for all x ∈ RN satisfying g.x /∈supp(f) for all g∈G, we have
Rj(f)(x) = Z
RN
Kj(x, y)f(y)dmk(y).
Proof. Let f ∈ L2(mk) be a compact supported function and x ∈ RN, such that g.x /∈supp(f) for all g ∈G. For 0< ε <min
g∈G min
y∈supp(f)
|g.x−y|
and n∈N, we consider ϕen,ε aC∞-function onR, such that:
• ϕen,ε is odd .
• ϕen,ε is supported in{t∈R; ε≤ |t| ≤n+ 1}.
• ϕen,ε= 1 in{t∈R; ε+n1 ≤t≤n}.
• |ϕen,ε| ≤1.
Let
φen,ε(t) = Z t
−∞
ϕen,ε(u)
|u|pk−1 du and φn,ε(y) =φen,ε(kyk),
for t∈R and y ∈RN. Clearly,φn,ε is a C∞ radial function supported in the ballB(0, n+ 1) and
n→+∞lim ϕen,ε(kyk) = 1, ∀y∈RN, kyk> ε.
The dominated convergence theorem, (2.5) and (2.6) yield Z
kyk>ετx(f)(−y) yj
kykpkdmk(y)
= lim
n→∞
Z
RN
τx(f)(−y) yj
kykpkϕen,ε(kyk)dmk(y)
= lim
n→∞
Z
RN
τx(f)(−y)Tj(φn,ε)(y)dmk(y)
= lim
n→∞
Z
RN
f(y)Tjτx(φn,ε)(−y)dmk(y).
Now we have Tjτx(φn,ε)(−y)
= Z
RN
(ηj−yj)ϕen,ε(A(x, y, η)) Apk(x, y, η) dµx(η)
+ X
α∈R+
k(α)αj
Z
RN
φen,ε(A(x, σα.y, η))−φen,ε(A(x, y, η))
< y, α > dµx(η),
where from (3.9)
ε < A(x, y, η) ; ε < A(x, σα.y, η), y∈supp(f), η ∈co(G.x).
Then with the aid of dominated convergence theorem
n→∞lim Tjτx(φn,ε)(−y) = 1
dkKj(x, y), and
dk
Z
kyk≥ετx(f)(−y) yj
kykpkdmk(y) = Z
RN
Kj(x, y)f(y)dmk(y).
Letting ε→0, it follows that Rj(f)(x) =
Z
RN
Kj(x, y)f(y)dmk(y),
which proves the result.
Now, we are able to state our main result.
Theorem 3.3. The Riesz transformRj,j = 1...N, is a bounded operator from Lp(mk) into itself, for all 1< p <∞.
Proof. Clearly, from (3.1) and Plancherel’s theorem Rj is bounded from L2(mk) into itself, with adjoint operatorR∗j =−Rj. Thus, via duality it’s enough to consider the range 1< p≤2 and apply Theorem 3.1. In view of Proposition 3.2 it only remains to show that Kj satisfies condition (3.3).
Let y, y0∈RN,y6=y0 andx∈RN, such that
ming∈Gkg.x−yk>2ky−y0k. (3.12) By mean value theorem,
|K(1)j (x, y)− K(1)j (x, y0)| =
N
X
i=0
(yi−(y0)i) Z 1
0
∂K(1)j
∂yi (x, yt)dt
=
N
X
i=0
(yi−(y0)i) Z 1
0
Z
RN
δi,j
Apk(x, yt, η) +pk((yt)i−ηi)(ηj−(yt)j)
Apk+2(x, yt, η)
dµx(η)
≤ Cky−y0k Z 1
0
Z
RN
1
Apk(x, yt, η) dµx(η)dt.
where yt=y0+t(y−y0) and δi,j is the Kronecker symbol.
In view of (3.9) and (3.12), we obtain
ky−y0k< A(x, yt, η), η ∈co(G.x).
Therefore,
K(1)j (x, y)− K(1)j (x, y0)
≤ Cky−y0k Z 1
0
Z
RN
1
ky−y0k2+A2(x, yt, η)
pk 2
dµx(η)dt.
≤ Cky−y0k Z 1
0
τx(ψ)(yt)dt where ψ is the function defined by
ψ(z) = 1
(ky−y0k2+kzk2)pk2 , z∈RN. Using Fubini’s theorem, (2.4) and (2.7), we get
Z
ming∈Gkg.x−yk>2ky−y0k
|K(1)j (x, y)− K(1)j (x, y0)|dmk(x)
≤ Cky−y0k Z 1
0
Z
RN
τ−yt(ψ)(x)dmk(x)dt
≤ C|y−y0| Z
RN
ψ(z)dmk(z) =C Z
RN
du (1 +u2)pk2
=C0.
This established the condition (3.3) for Kj(1).
To deal with Kj(α), α ∈ R+, we put for x, y ∈ RN , η ∈ co(G.x) and t∈[0,1]
U(x, y, η) = A2pk−4(x, y, η),
Vα(x, y, η) = Apk−2Apαk−2(Apk−2+Apαk−2), hα,t(y) = y+t(σα.y−y) =y−t < y, α > α, By mean value theorem we have
K(α)j (x, y) = Z
RN
1
< y, α >
U(x, σα.y, η)−U(x, y, η) Vα(x, y, η) dµx(η)
= −
Z
RN
Z 1 0
∂αU(x, hα,t(y), η)
Vα(x, y, η) dt dµx(η) and
K(α)j (x, y)− Kj(α)(x, y0)
= Z
RN
Z 1
0
Z 1
0
∂y−y0
∂αU(x, hα,t(.), η) Vα(x, ., η)
(yθ)dθ dt dµx(η). (3.13) Here the derivations are taken with respect to the variabley.
To simplify, let us denote by
A=A(x, yθ, η); Aα =A(x, σα.yθ, η) Then using (3.10) and the fact
kη−hα,t(yθ)k ≤max(kη−yθk,kη−σα(yθ)k), we obtain
∂U
∂yr(x, hα,t(yθ), η) ≤ CA2pk−5+A2pαk−5
∂2U
∂yr∂ys
(x, hα,t(yθ), η) ≤ CA2pk−6+A2pαk−6, r, s= 1, ..., N.
This gives us the following estimates
∂αU(x, hα,t(yθ), η)≤CA2pk−5+A2pαk−5, (3.14)
∂y−y0
∂αU(x, hα,t(.), η)(yθ)≤Cky−y0kA2pk−6+A2pαk−6. (3.15) By (3.10) and (3.11), we also have
∂Vα
∂yr((x, yθ, η))≤CApk−3Apαk−3(Apk−2+Apαk−2)(A+Aα).
The elementary inequality u+v
u`+v` ≤ 3
u`−1+v`−1, u, v >0, `≥1, leads to
∂y−y0Vα(x, yθ, η) Vα2(x, yθ, η)
≤ Cky−y0k Aα+A
Apk−1Apαk−1(Apk−2+Apαk−2)
≤ Cky−y0k 1
Apk−1Apαk−1(Apk−3+Apαk−3). (3.16) Now (3.14), (3.15) and (3.16) yield
∂y−y0
∂αU(x, hα,t(.), η) Vα(x, ., η)
(yθ)
≤ Cky−y0k A2pk−6+A2pαk−6 Apk−2Apαk−2(Apk−2+Apαk−2) + Cky−y0k A2pk−5+A2pαk−5
Apk−1Apαk−1(Apk−3+Apαk−3)
≤ Cky−y0k 1 A2Apαk−2
+ 1
Apk−2A2α
+ Cky−y0k 1 AApαk−1
+ 1
Apk−1Aα
≤ Cky−y0k 1 Apk + 1
Apαk
where in the last equality we have used the fact that 1
uv`−1 ≤ 1 u` + 1
v`, u, v >0 and `≥1.
Thus, in view of (3.13),
K(α)j (x, y)− Kj(α)(x, y0)
≤Cky−y0k Z 1
0
Z
RN
h 1
Apk(x, yθ, η) + 1 Apk(x, σαyθ, η)
i
dµx(η)dθ.
Then by same argument as for K(1)j we obtain Z
ming∈Gkg.x−yk>2ky−y0k
|K(2)j (x, y)− K(2)j (x, y0)|dmk(x)≤C, which established the condition (3.3) for the kernelK(α)j and furnishes the
proof.
As applications, we will prove a generalized Riesz and Sobolev inequal- ities
Corollary 3.4 (Generalized Riesz inequalities). For all 1< p <∞ there exists a constant Cp such that
||TrTs(f)||k,p≤Cp||∆kf||k,p, for all f ∈ S(RN), (3.17) where ∆k is the Dunkl laplacian: ∆kf =
N
X
r=1
Tr2(f) Proof. From (2.1) and (3.1) one can see that
TrTs(f) =RrRs(−∆k)(f), r, s= 1...N, f ∈ S(RN).
Then (3.17) is concluded by Theorem 3.3.
Corollary 3.5 (Generalized Sobolev inequality). For all 1 < p ≤ q <
2γ(k) +N with 1 q = 1
p− 1
2γ(k) +N, we have
kfkq,k ≤Cp,qk∇kfkp,k (3.18) for all f ∈ S(RN). Here ∇kf = (T1f, ..., TNf) and|∇kf|= (
N
X
r=1
|Trf|2)12.
Proof. For allf ∈ S(RN), we write Fk(f)(ξ) = 1
kξk
N
X
r=1
−iξr kξk
iξrFk(f)(ξ)
= 1
kξk
d
X
r=1
−iξr kξk
Fk(Trf)(ξ). This yields to the following identity
f =Ik1
N
X
j=1
Rj(Tjf), where
Ikβ(f)(x) = (dβk)−1 Z
RN
τyf(x)
kyk2γ(k)+N−βdmk(y), here
dβk = 2−γ(k)−N/2+β Γ(β2) Γ(γ(k) +N−β2 ).
Theorem 1.1 of [4] asserts that Ikβ a bounded operator from Lp(mk) to Lq(mk). Then (3.18) follows from Theorem 3.3.
References
[1] B. Amri, A. Gasmi & M. Sifi – “Linear and bilinear multiplier operators for the dunkl transform”, Mediterranean Journal of Math- ematics 7 (2010), p. 503–521.
[2] F. Dai & H. Wang – “A transference theorem for the dunkl transform and its applications”, Journal of Functional Analysis 258 (2010), p. 4052–4074.
[3] C. F. Dunkl – “Differential–difference operators associated to re- flection groups”, Trans. Amer. Math.311(1989), p. 167–183.
[4] S. Hassani, S. Mustapha & M. Sifi – “Riesz potentials and frac- tional maximal function for the dunkl transform”, J. Lie Theory 19 (2009, no. 4), p. 725–734.
[5] M. de Jeu – “The dunkl transform”, Invent. Math. 113 (1993), p. 147–162.
[6] M. Rosler– “Dunkl operators: theory and applications, in orthogo- nal polynomials and special functions (leuven, 2002),Rn”,Lect. Notes Math.1817 (2003), p. 93–135.
[7] , “A positive radial product formula for the dunkl kernel”, Trans. Amer. Math. Soc.355 (2003), p. 2413–2438.
[8] E. M. Stein –Harmonic analysis: Reals-variable methods, orthogo- nality and oscillatory integrals, PrincetonS, New Jersey, 1993.
[9] S. Thangavelu & Y. Xu – “Convolution operator and maximal function for dunkl transform”, J. Anal. Math.97 (2005), p. 25–55.
[10] , “Riesz transforms and riesz potentials for the dunkl trans- form”, J. Comp. and Appl. Math. 199(2007), p. 181–195.
Bechir Amri
Department of Mathematics University of Tunis
Preparatory Institute of Engineer Studies of Tunis
1089 Montfleury, Tunis, Tunisia bechir.amri@ipeit.rnu.tn
Mohamed Sifi
Department of Mathematics University of Tunis El Manar Faculty of Sciences of Tunis
2092 Tunis El Manar, Tunis, Tunisia mohamed.sifi@fst.run.tn