DOI 10.1007/s11868-015-0138-4
Some new estimates for the Cherednik–Opdam transform in the space L2α,β(R)
S. El Ouadih1 · R. Daher1
Received: 3 October 2015 / Accepted: 11 October 2015
© Springer Basel 2015
Abstract New estimates are proved for the Cherednik–Opdam transform in the space L2α,β(R)on certain classes of functions characterized by the generalized continuity modulus.
Keywords Cherednik–Opdam operator·Cherednik–Opdam transform· Generalized translation
Mathematics Subject Classification 42B37
1 Introduction
Abilov et al. [1], proved two useful estimates for the Fourier transform in the space of square integrable functions on certain classes of functions characterized by the generalized continuity modulus, using a translation operator.
In this paper, we prove new estimates in certain classes of functions characterized by a generalized continuity modulus and connected with the Cherednik–Opdam transform associated to T(α,β) in L2α,β(R) analogs of the statements proved in [1]. For this purpose, we use a generalized translation operator.
In Sect.2, we give some definitions and preliminaries concerning the Cherednik–
Opdam transform. The new estimates are proved in Sect.3.
B S. El Ouadih
salahwadih@gmail.com R. Daher
rjdaher024@gmail.com
1 Department of Mathematics, Faculty of Sciences Aïn Chock, University Hassan II, Casablanca, Morocco
2 Preliminaries
In this section, we develop some results from harmonic analysis related to the differential-difference operatorT(α,β). Further details can be found in [2,3]. In the following we fix parametersα,βsubject to the constraintsα≥β ≥ −12andα > −21. Letρ = α+β +1 and λ ∈ C. The Opdam hypergeometric functions G(α,β)λ onRare eigenfunctionsT(α,β)G(α,β)λ (x)=iλG(α,β)λ (x)of the differential-difference operator
T(α,β)f(x)=f(x)+ [(2α+1)cothx+(2β+1)tanhx]f(x)− f(−x)
2 −ρf(−x)
that are normalized such thatG(α,β)λ (0)=1. In the notation of Cherednik one would writeT(α,β)as
T(k1+k2)f(x)= f(x)+
2k1
1+e−2x + 4k2
1−e−4x
(f(x)− f(−x))
−(k1+2k2)f(x),
withα=k1+k2−12andβ=k2−12. Herek1is the multiplicity of a simply positive root andk2the (possibly vanishing) multiplicity of a multiple of this root. By [2] or [3], the eigenfunctionG(α,β)λ is given by
G(α,β)λ (x)=ϕα,βλ (x)− 1 ρ−iλ
∂
∂xϕλα,β(x)=ϕλα,β(x)+ ρ
4(α+1)sinh(2x)ϕλα+1,β+1(x), whereϕα,βλ (x)=2 F1(ρ+2iλ;ρ−2iλ;α+1; −sinh2x)is the classical Jacobi function.
Lemma 2.1 [4]The following inequalities are valids for Jacobi functionsϕα,βλ (x) (i) |ϕλα,β(x)| ≤1.
(ii) 1−ϕα,βλ (x)≤x2(λ2+ρ2).
DenoteL2α,β(R), the space of measurable functions f onRsuch that
f2,α,β =
R|f(x)|2Aα,β(x)d x 1/2
<+∞,
where
Aα,β(x)=(sinh|x|)2α+1(cosh|x|)2β+1. The Cherednik–Opdam transform of f ∈Cc(R)is defined by
Hf(λ)=
R f(x)G(α,β)λ (−x)Aα,β(x)d x for all λ∈C.
The inverse transform is given as H−1g(x)=
Rg(λ)G(α,β)λ (x) 1− ρ
iλ
dλ 8π|cα,β(λ)|2 here
cα,β(λ)= 2ρ−iλ (α+1) (iλ)
1
2(ρ+iλ) 1
2(α−β+1+iλ).
The corresponding Plancherel formula was established in [2], to the effect that
R|f(x)|2Aα,β(x)d x = +∞
0
|Hf(λ)|2+ |Hfˇ(λ)|2 dσ(λ),
where fˇ(x):= f(−x)anddσ is the measure given by dσ(λ)= dλ
16π|cα,β(λ)|2.
According to [3] there exists a family of signed measures μ(α,β)x,y such that the product formula
G(α,β)λ (x)G(α,β)λ (y)=
RG(α,β)λ (z)dμ(α,β)x,y (z), holds for allx,y∈Randλ∈C, where
dμ(α,β)x,y (z)=
⎧⎨
⎩
Kα,β(x,y,z)Aα,β(z)d z if x y=0
dδx(z) if y=0
dδy(z) if x=0
and
Kα,β(x,y,z)=Mα,β|sinhx.sinhy.sinhz|−2α π
0
g(x,y,z, χ)α−β−+ 1
×
1−σxχ,y,z+σxχ,z,y+σzχ,y,x+ ρ
β+12 cothx.cothy.cothz(sinχ)2
×(sinχ)2βdχ
ifx,y,z ∈ R\{0}satisfy the triangular inequality||x| −y||<|z|<|x| + |y|, and Kα,β(x,y,z)=0 otherwise. Here
∀x,y,z∈R, χ ∈ [0,1], σxχ,y,z =
⎧⎨
⎩
coshx+coshy−coshzcosχ
sinhxsinhy if x y=0
0 if x y=0
andg(x,y,z, χ)=1−cosh2x−cosh2y.cosh2z+2 coshx.coshy.coshz.cosχ. Lemma 2.2 [3]For all x,y∈R, we have
(i) Kα,β(x,y,z)=Kα,β(y,x,z).
(ii) Kα,β(x,y,z)=Kα,β(−x,z,y).
(iii) Kα,β(x,y,z)=Kα,β(−z,y,−x).
The product formula is used to obtain explicit estimates for the generalized translation operators
τx(α,β)f(y)=
R f(z)dμ(α,β)x,y (z).
It is known from [3] that
Hτx(α,β)f(λ)=G(α,β)λ (x)Hf(λ) (1)
H(T(α,β)f)(λ)=iλHf(λ) (2)
for f ∈Cc(R).
Forα > −21, we introduce the Bessel normalized function of the first kind jα defined by
jα(x)= (α+1)∞
n=0
(−1)n(x2)2n
n! (n+α+1), x∈R.
In the terms of jα(x), we have (see [5])
√hx Jα(hx)=O(1),hx ≥0, (3)
where Jα(x)is Bessel function of the first kind, which is related to jα(x)by the formula
jα(x)=2α (α+1)
xα Jα(x). (4)
Lemma 2.3 [6]Letα≥β ≥ −21, α = −21. Then for|ν| ≤ρ, there exists a positive constant c0such that
|1−ϕλ+α,βiν(x)| ≥c0|1−jα(λx)|.
The generalized modulus of continuity of a function f ∈ L2α,β(R)is defined by ω(f, δ)2,α,β = sup
0<h≤δNhf2,α,β, δ >0,
whereNh:=τh(α,β)+τ−(α,β)h −2Iand I is the identity operator inL2α,β(R).
Denote byW2r(T(α,β)),r =0,1,2..., the class of functions f ∈L2α,β(R)that have generalized derivatives satisfying(T(α,β))r f ∈L2α,β(R).
i.e.,
W2r(T(α,β))=
f ∈L2α,β(R):(T(α,β))rf ∈L2α,β(R) ,
where(T(α,β))0f = f,(T(α,β))rf =T(α,β)((T(α,β))r−1f);r=1,2, ...
3 Main result
The goal of this work is to prove several new estimates for the integral
JN2(f)= +∞
N
|Hf(λ)|2+ |Hfˇ(λ)|2 dσ(λ),
in certain classes of functions inL2α,β(R). Lemma 3.1 If f ∈Cc(R), then
Hˇτx(α,β)f(λ)=G(α,β)λ (−x)Hfˇ(λ) (5)
Proof For f ∈Cc(R), we have Hτˇx(α,β)f(λ)=
Rτx(α,β)f(−y)G(α,β)λ (−y)Aα,β(y)d y
=
Rτx(α,β)f(y)G(α,β)λ (y)Aα,β(y)d y
=
R
R f(z)Kα,β(x,y,z)Aα,β(z)d z
G(α,β)λ (y)Aα,β(y)d y
=
R f(z)
RG(α,β)λ (y)Kα,β(x,y,z)Aα,β(y)d y
Aα,β(z)d z.
SinceKα,β(x,y,z)=Kα,β(−x,z,y), it follows from the product formula that
Hτˇx(α,β)f(λ)=G(α,β)λ (−x)
R f(z)G(α,β)λ (z)Aα,β(z)d z
=G(α,β)λ (−x)
R f(−z)G(α,β)λ (−z)Aα,β(z)d z
=G(α,β)λ (−x)Hfˇ(λ).
Lemma 3.2 For f ∈W2r(T(α,β)), then
Nh(T(α,β))r f(x)22,α,β=4 +∞
0
λ2r|ϕλα,β(h)−1|2
|Hf(λ)|2+ |Hfˇ(λ)|2 dσ(λ).
Proof From formulas (1) and (5), we have H(Nhf)(λ)=
G(α,β)λ (h)+G(α,β)λ (−h)−2
H(f)(λ),
and
H(Nˇhf)(λ)=
G(α,β)λ (−h)+G(α,β)λ (h)−2
H(fˇ)(λ).
Since
G(α,β)λ (h)=ϕλα,β(h)+ ρ
4(α+1)sinh(2h)ϕλα+1,β+1(h), andϕλα,βis even, then
H(Nhf)(λ)=2
ϕα,βλ (h)−1
H(f)(λ)
and
H(Nˇhf)(λ)=2
ϕα,βλ (h)−1
H(fˇ)(λ)
Now by formula (2) and Plancherel Theorem, we have the result.
Theorem 3.3 Given r and f ∈W2r(T(α,β)). Then there exist a constant c>0such that, for all N >0,
JN(f)=O
N−rω((T(α,β))r f,cN−1)2,α,β
.
Proof Firstly, we have JN2(f)≤
+∞
N |jα(λh)|dμ(λ)+ +∞
N |1−jα(λh)|dμ(λ), (6)
withdμ(λ)=
|Hf(λ)|2+ |Hfˇ(λ)|2
dσ(λ). The parameterh >0 will be chosen in an instant.
In view of formulas (3) and (4), there exist a constantc1>0 such that
|jα(λh)| ≤c1(λh)−α−12. Then
+∞
N
|jα(λh)|dμ(λ)≤c1(h N)−α−12JN2(f).
Choose a constantc2such that the numberc3=1−c1c−α−
1 2
2 is positif.
Settingh =c2/Nin the inequality (6), we have c3JN2(f)≤
+∞
N
|1−jα(λh)|dμ(λ). (7)
By Hölder inequality and Lemma2.3the second term in (7) satisfies +∞
N
|1−jα(λh)|dμ(λ)= +∞
N
|1− jα(λh)|.1.dμ(λ)
≤ +∞
N
|1− jα(λh)|2dμ(λ)
1/2 +∞
N
dμ(λ) 1/2
≤ +∞
N
λ−2r|1− jα(λh)|2λ2rdμ(λ) 1/2
JN(f)
≤ N−r +∞
N
|1− jα(λh)|2λ2rdμ(λ) 1/2
JN(f)
≤ N−r c0
+∞
N
λ2r|1−ϕλα,β(h)|2dμ(λ) 1/2
JN(f).
From Lemma3.2, we conclude that +∞
N
λ2r|1−ϕα,βλ (h)|2dμ(λ)≤ Nh(T(α,β))rf(x)22,α,β. Therefore
+∞
N
|1− jα(λh)|dμ(λ)≤ N−r
c0 Nh(T(α,β))rf(x)2,α,βJN(f).
Forh =c2/N, we obtain c3JN2(f)≤ N−r
c0
ω
(T(α,β))r f,c2/N
2,α,β JN(f).
Consequently
c0c3JN(f)≤N−rω
(T(α,β))r f,c2/N
2,α,β.
for allN >0. The theorem is proved withc=c2. Theorem 3.4 Let f ∈ L2α,β(R). Then, for all N >0,
ω(f,N−1)2,α,β =O
⎛
⎜⎝N−2 N−1
l=0
(l+1)3Jl2(f) 12⎞
⎟⎠.
Proof From Lemma3.2, we have Nhf(x)22,α,β =4
+∞
0
|1−ϕλα,β(h)|2
|Hf(λ)|2+ |Hfˇ(λ)|2 dσ (λ).
This integral is divided into two +∞
0
= N
0
+ +∞
N
=4(I1+I2),
whereN = [h−1]. We estimate them separately.
From (i) of Lemma2.1, we have the estimate I2≤c4
+∞
N
|Hf(λ)|2+ |Hfˇ(λ)|2
dσ(λ)=c4JN2(f).
Now, we estimateI1. From formula (ii) of Lemma2.1, we have I1≤h4
N
0
(λ+ρ)4
|Hf(λ)|2+ |Hfˇ(λ)|2 dσ(λ)
=h4
N−1 l=0
l+1
l
(λ+ρ)4
|Hf(λ)|2+ |Hfˇ(λ)|2 dσ(λ)
≤h4
N−1 l=0
(l+ρ+1)4
Jl2(f)−Jl2+1(f) ,
From the inequalityl+ρ+1≤(ρ+1)(l+1)we conclude
I1≤(ρ+1)4h4
N−1 l=0
al
Jl2(f)−Jl2+1(f)
withal =(l+1)4. For all integersm≥1, the Abel transformation shows m
l=0
al
Jl2(f)−Jl2+1(f)
=a0J02(f)+ m l=1
(al−al−1)Jl2(f)−amJm2+1(f)
≤a0J02(f)+ m l=1
(al−al−1)Jl2(f),
becauseamJm2+1(f)≥0. Hence
I1≤(ρ+1)4h4
J02(f)+
N−1 l=1
(l+1)4−l4
Jl2(f)−N4JN2(f)
.
Moreover by the finite increments theorem, we have (l+1)4−l4≤4(l+1)3. Then
I1≤(ρ+1)4N−4
J02(f)+4
N−1 l=1
(l+1)3Jl2(f)−N4JN2(f)
,
sinceN≤ 1h. Combining the estimates forI1andI2gives
Nhf(x)22,α,β =O
N−4
N−1 l=0
(l+1)3Jl2(f)
,
which implies
ω(f,N−1)2,α,β =O
⎛
⎜⎝N−2 N−1
l=0
(l+1)3Jl2(f) 12⎞
⎟⎠,
and this ends the proof.
Theorem 3.5 Let f ∈ L2α,β(R). If the series ∞
l=1
lr−1Jl(f), r =1,2, ...
converges, then f ∈W2r(T(α,β))and, for all N >0,
ω
(T(α,β))r f,N−1
2,α,β =O
N−4
N−1 l=0
(l+1)2r+3Jl2(f) 12
+O
⎛
⎜⎝ ∞ l=[N2]
lr−1Jl(f)
⎞
⎟⎠.
Proof Let f ∈ L2α,β(R). By formula (2) and Plancherel Theorem, we have
(T(α,β))r f22,α,β = +∞
0
λ2r
|Hf(λ)|2+ |Hfˇ(λ)|2 dσ(λ)
= ∞ l=0
l+1
l λ2r
|Hf(λ)|2+ |Hfˇ(λ)|2 dσ(λ).
Using an Abel transformation we obtain
(T(α,β))r f22,α,β ≤J02(f)+2r ∞
l=1
(l+1)2r−1Jl2(f).
From the inequalityl+1≤2lwe conclude
(T(α,β))rf22,α,β ≤c5
J02(f)+∞
l=1
l2r−1Jl2(f)
.
Hence
(T(α,β))rf2,α,β =O ∞
l=1
lr−1Jl(f)
.
Since the series
∞ l=1
lr−1Jl(f), r=1,2, ...
converges, we see that f ∈W2r(T(α,β)).
From Lemma3.2, we have Nh(T(α,β))r f(x)22,α,β=4
+∞
0
λ2r|1−ϕλα,β(h)|2
|Hf(λ)|2+s|Hfˇ(λ)|2 dσ (λ).