On Clustering Histograms with k-Means by Using Mixed α-Divergences
Stronglysingularintegralsalongcurves on α-modulationspaces
Transcript of Stronglysingularintegralsalongcurves on α-modulationspaces
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 DOI 10.1186/s13660-017-1458-0
R E S E A R C H Open Access
Strongly singular integrals along curveson α-modulation spacesXiaomei Wu1* and Xiao Yu2
*Correspondence: [email protected] College, Zhejiang NormalUniversity, Jinhua, 321004, P.R. ChinaFull list of author information isavailable at the end of the article
AbstractIn this paper, we study the strongly singular integrals
Tn,β ,γ f (x) = p.v.∫ 1
–1f (x –�θ (t))
e–2π i|t|–β
t|t|γ dt
along homogeneous curves �θ (t). We prove that Tn,β ,γ is bounded on theα-modulation spaces, including the inhomogeneous Besov spaces and the classicalmodulation spaces.
Keywords: α-modulation spaces; strongly singular integrals; Besov spaces;homogeneous curves
1 IntroductionThe two dimension strongly singular integrals along curves Tβ ,γ are defined by
Tβ ,γ (f )(x, y) = p.v.∫
–f(x – t, y – �(t)
)e–π i|t|–β
t|t|γ dt,
where x, y ∈ R,β ,γ > . Zielinski [] showed that Tβ ,γ is bounded on L(R) along thecurve (t, t) if and only if β > γ . Later, Chandarana [] extended the result to the generalcurves (t, |t|m) or (t, sgn(t)|t|m) with m ≥ and showed that Tβ ,γ is bounded on Lp(R) for
+γ (β + )
β(β + ) + β – γ< p <
β(β + ) + (β – γ )γ (β + )
+ .
Moreover, Chandarana also studied the strongly singular integrals along curves in R
(see [] for details). In [], Chen et al. considered the operator for high dimension n. Letθ = (θ, θ, . . . , θn) ∈R
n, and
�θ (t) =(θ|t|p , θ|t|p , . . . , θn|t|pn
)
or
�θ (t) = sgn(t)(θ|t|p , θ|t|p , . . . , θn|t|pn
).
© The Author(s) 2017. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in anymedium, pro-vided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, andindicate if changes were made.
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 2 of 13
Then the operator Tn,β ,γ is defined as
Tn,β ,γ f (x) = p.v.∫
–f(x – �θ (t)
)e–π i|t|–β
t|t|γ dt(β > γ > , x ∈R
n).
Suppose p, p, . . . , pn,α and β are positive numbers. In [], the authors proved that Tn,β ,γ
is bounded on Lp(Rn) whenever β > (n + )γ and
β
β – (n + )γ< p <
β
(n + )γ.
Later, Cheng-Zhang [] and Cheng [] extended the results to the modulation space. Theyshowed that the strongly singular integral Tn,β ,γ is bounded on the modulation spaces Ms
p,qfor all p > . It is worth to point out that the modulation space is a better substitution tostudy the strongly singular integrals because there is no restriction on the index p.
Here we will consider the strongly singular integrals along homogeneous curves Tn,β ,γ
on the α-modulation spaces. The α-modulation spaces Ms,αp,q were first introduced by
Gröbner in []. They contain the inhomogeneous Besov spaces Bsp,q in the limit case α =
and the classical modulation spaces Msp,q in the case α = , respectively. It is proposed as
an intermediate function space; see [, ] for more details.In recent years, there were numerous papers on these spaces and its applications, such as
[–] and the references therein. Motivated by the work of Cheng-Zhang [] on the mod-ulation spaces, one naturally expects that the strongly singular integral operators Tn,β ,γ
have the boundedness property on the α-modulation spaces for all ≤ α ≤ . In this pa-per, we will affirm this.
This paper is organized as follows. In Section , we will recall the definition of the α-modulation spaces and the Besov spaces. Some lemmas will also be presented in this sec-tion. In Section , we will give the main results and prove the theorems. In addition, wewill consider the strongly singular integrals along a well-curved �(t) in R
n. Throughoutthis paper, we use the notation A � B meaning that there is a positive constant C indepen-dent of all essential variables such that A ≤ CB. We denote A ∼ B to stand for A � B andB � A.
2 Preliminaries and lemmasBefore giving the definition of the α-modulation spaces, we introduce some notations fre-quently used in this paper. Let S = S(Rn) be the subspace of C∞(Rn) of Schwartz rapidlydecreasing functions and S ′ = S ′(Rn) be the space of all tempered distribution on R
n. Fork = (k, k, . . . , kn) ∈ Z
n, we denote
|k| =(k
+ k + · · · + k
n)
, 〈k〉 =( + |k|)
.
We define the ball
Brk :=
{ξ ∈R
n :∣∣ξ – 〈k〉 α
–α k∣∣ < r〈k〉 α
–α}
and Brk denotes
{ξ ∈R
n :∣∣ξ – 〈k〉 α
–α k∣∣ < r〈k〉 α
–α}
.
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 3 of 13
The Fourier transform F (f ) and the inverse Fourier transform F–(f ) are defined by
F (f )(ξ ) = f̂ (ξ ) =∫Rn
f (x)e–π ix·ξ dx,
F–(f )(ξ ) = f̆ (ξ ) =∫Rn
f (x)eπ ix·ξ dx.
To define the α-modulation spaces, we introduce the α-decomposition. Let ρ be a non-negative smooth radial bump function supported in B(, ), satisfying ρ(ξ ) = for |ξ | < and ρ(ξ ) = for |ξ | ≥ . For any k = (k, k, . . . , kn) ∈ Z
n, we set
ραk (ξ ) = ρ
(ξ – 〈k〉 α
–α kr〈k〉 α
–α
)
and denote
ηαk (ξ ) = ρα
k (ξ )(∑
l∈Zn
ραl (ξ )
)–
.
It is easy to check that {ηαk }k∈Zn satisfy
suppηαk ⊂ Br
k ; (.)
ηαk (ξ ) = c, ∀ξ ∈ Br
k ; (.)∑k∈Zn
ηαk (ξ ) ≡ , ∀ξ ∈R
n (.)
and
∥∥F–ηαk∥∥
L � . (.)
Corresponding to the above sequence {ηαk }k∈Zn , we can construct an operator sequence
{�αk }k∈Zn by
�αk = F–ηα
k F .
For ≤ α < , < p, q ≤ ∞, s ∈ R, using this decomposition, we define the α-modulationspaces as
Ms,αp,q
(R
n) :={
f ∈ S ′ : ‖f ‖Ms,αp,q =
(∑k∈Zn
〈k〉 sq–α
∥∥�αk f
∥∥qLp
) q
< ∞}
.
We have the usual modification when p, q = ∞. We denote Ms,p,q = Ms
p,q. It is the classicalmodulation space. Its related decomposition is called uniform decomposition; see [, ]and [] for details. In order to define the Besov spaces, we introduce the dyadic decom-position. Let ψ be a smooth bump function supported in the ball {ξ : |ξ | ≤
}. We may
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 4 of 13
assume ψ(ξ ) = if |ξ | ≤ . Denote φ(ξ ) = ψ(ξ ) – ψ(ξ ) and a function sequence {φj}∞j=:
⎧⎨⎩
φj(ξ ) = φ(–jξ ), j ∈N,
φ(ξ ) = –∑∞
j= φj(ξ ).
Define the Littlewood-Paley (or dyadic) decomposition operators as
j = F–φjF , j ∈N∪ {}.
Let ≤ p, q ≤ ∞, s ∈ R. For a tempered distribution f , we define the (inhomogeneous)Besov space Bs
p,q as
Bsp,q
(R
n) :=
{f ∈ S ′ : ‖f ‖Bs
p,q(Rn) =
( ∞∑j=
sjq‖ jf ‖qLp
) q
< ∞}
.
With the usual modification when p, q = ∞. Obviously, the α-decomposition is bigger thanthe uniform decomposition and thinner than the dyadic decomposition. This decompo-sition on frequency extends the dyadic and the uniform decomposition.
In order to prove the theorems, we also need some lemmas.
Lemma . Van der Corput lemma ([], p.). Let ϕ and φ be real valued smooth func-tions on the interval (a, b) and k ∈ N. If |ϕ(k)(t)| ≥ for all t ∈ (a, b) and () k = , ϕ′(t) ismonotonic on (a, b), or () k ≥ , then we have
∣∣∣∣∫ b
aeiλϕ(t)φ(t) dt
∣∣∣∣ ≤ ckλ–
k
(∣∣φ(b)∣∣ +
∫ b
a
∣∣φ′(t)∣∣dt
).
Lemma . () If ≤ α < , ≤ p ≤ ∞, s, s ∈R, k ∈ Zn and
∥∥�αk (Tβ ,γ f )
∥∥Lp � 〈k〉 s
–α ‖f ‖Lp ,
then, for any < q ≤ ∞, Tβ ,γ is bounded from Ms+s,αp,q to Ms,α
p,q.() If ≤ p ≤ ∞, s, s ∈ R, j ∈ Z and
∥∥ αj (Tβ ,γ f )
∥∥Lp � js‖f ‖Lp ,
then, for any < q ≤ ∞, Tβ ,γ is bounded from Bs+s,αp,q to Bs,α
p,q.
Proof () It suffices to show that
∥∥�αk (Tβ ,γ f )
∥∥Lp � 〈k〉 s
–α
∥∥�αk f
∥∥Lp .
Denote �(k) = {j ∈ Zn : ηα
k · ηαj �= }. Then by the support condition (.), we have
〈j〉 ∼ 〈k〉, if j ∈ �(k)
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 5 of 13
and
��(k) � .
Recall that the choice of ηαj satisfies
∑j∈Zn ηα
j ≡ . By Minkowski’s inequality, we obtain
∥∥�αk (Tβ ,γ f )
∥∥Lp =
∥∥∥∥∑j∈Zn
�αk[Tβ ,γ
(�α
j f)]∥∥∥∥
Lp≤
∑j∈Zn
∥∥�αj[Tβ ,γ
(�α
k f)]∥∥
Lp
�∑
j∈�(k)
〈j〉 s–α
∥∥�αk f
∥∥Lp
� 〈k〉 s–α
∥∥�αk f
∥∥Lp .
Thus, the first part of the lemma holds. The second part of this lemma is similar to theproof of the first part. Here we omit the details. �
3 Main results and proofsTheorem . Let �(t) = |t|m or �(t) = |t|m sgn(t). If ≤ α < ,β > γ > , ≤ p ≤ ∞, <q ≤ ∞ and s ∈R, then Tβ ,γ is bounded from Ms+α,α
p,q to Ms,αp,q.
Proof By checking the following proof, we can only consider the operator
Tβ ,γ (f )(x, y) =∫
f(x – t, y – tm)e–π it–β
tγ + dt.
It is similar for – ≤ t < . First, let us choose a C∞ function θ (t) with support in [ , ] on
the real line satisfying
∞∑j=
θ(jt
) ≡ .
Then we can decompose Tβ ,γ as
Tβ ,γ (f )(x, y) =∞∑j=
∫
θ(jt
)f(x – t, y – tm)e–π it–β
tγ + dt :=∞∑j=
Tj(f )(x, y). (.)
Using the Fourier transformation, the operator Tj can be written as
T̂jf (ξ, ξ) = mj(ξ, ξ)f̂ (ξ, ξ),
where
mj(ξ, ξ) =∫
θ(jt
)e–π i(ξt+ξtm) e–π it–β
tγ + dt.
Let �αk,j be the kernel of �α
k Tj, so we have
�αk Tj(f )(x) = F–(ηα
k (ξ )mj(ξ )) ∗ f (x) = �α
k,j ∗ f (x).
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 6 of 13
By the Young inequality, we obtain
∥∥�αk Tj(f )
∥∥Lp ≤ ∥∥�α
k,j∥∥
L‖f ‖Lp .
Let η̃αk (ζ ) = ηα
k (〈k〉 α–α ζ + k〈k〉 α
–α ). Then supp η̃αk (ζ ) ⊂ B(, r). By a simple substitution and
the Fubini theorem, we get
�αk,j(x) =
∫R
eπ ix·ξ ηαk (ξ )
∫
θ(jt
)e–π i(ξt+ξtm) e–π it–β
tγ + dt dξ dξ
=∫R
ηαk (ξ )
∫
e–π i[ξ(t–x)+ξ(tm–x)]e–π it–β
θ(jt
)t––γ dt dξ dξ
= 〈k〉 α–α
∫R
η̃αk (ζ )
∫
e–π i[(ζ〈k〉 α
–α +k〈k〉 α–α )(t–x)+(ζ〈k〉 α
–α +k〈k〉 α–α )(tm–x)]
× e–π it–βθ(jt
)t––γ dt dζ dζ
= 〈k〉 α–α eπ i〈k〉 α
–α k·x∫
e–π i(t–β +k〈k〉 α
–α t+k〈k〉 α–α tm)θ
(jt
)t––γ
× ˆ̃ηαk(〈k〉 α
–α (t – x), 〈k〉 α–α
(tm – x
))dt.
Let ϕ(t) = –π [t–β + k〈k〉 α–α t + k〈k〉 α
–α tm], then
�αk,j(x) = 〈k〉 α
–α eπ i〈k〉 α–α k·x
∫ –j+
–j–eiϕ(t)θ
(jt
)t––γ ˆ̃ηα
k(〈k〉 α
–α (t – x), 〈k〉 α–α
(tm – x
))dt
:= 〈k〉 α–α eπ i〈k〉 α
–α k·xIj.
First, let us estimate Ij. We divide it into two cases.Case . If k ≥ , then
∣∣ϕ′′(t)∣∣ = π
[β(β + )t–β– + m(m – )k〈k〉 α
–α tm–] � t–β– � j(β+).
By the Van der Corput lemma, we have
Ij � – j(β+) j(+γ )∣∣ ˆ̃ηα
k(〈k〉 α
–α(–j+ – x
), 〈k〉 α
–α(–jm+m – x
))∣∣
+ – j(β+)
∫ –j+
–j–
∣∣∣∣ ddt
[θ(jt
)t––γ ˆ̃ηα
k(〈k〉 α
–α (t – x), 〈k〉 α–α
(tm – x
))]∣∣∣∣dt
� – j(β–γ )
∣∣ ˆ̃ηαk(〈k〉 α
–α(–j+ – x
), 〈k〉 α
–α(–jm+m – x
))∣∣
+ – j(β+)
∫ –j+
–j–jθ ′(jt
)t––γ
∣∣ ˆ̃ηαk(〈k〉 α
–α (t – x), 〈k〉 α–α
(tm – x
))∣∣dt
+ – j(β+)
∫ –j+
–jt––γ
∣∣ ˆ̃ηαk(〈k〉 α
–α (t – x), 〈k〉 α–α
(tm – x
))∣∣dt
+ – j(β+) 〈k〉 α
–α
∫ –j+
–jt––γ
∣∣∣∣∂ˆ̃ηα
k∂x
(〈k〉 α–α (t – x), 〈k〉 α
–α(tm – x
))∣∣∣∣dt
+ – j(β+) 〈k〉 α
–α
∫ –j+
–jt––γ +m
∣∣∣∣∂ˆ̃ηα
k∂x
(〈k〉 α–α (t – x), 〈k〉 α
–α(tm – x
))∣∣∣∣dt.
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 7 of 13
Then, by a substitution and the Fubini theorem, we obtain
∥∥�αk,j
∥∥L � – j(β–γ )
〈k〉 α–α
∫R
∣∣ ˆ̃ηαk(〈k〉 α
–α(–j+ – x
), 〈k〉 α
–α(–jm+m – x
))∣∣dx
+ – jβ 〈k〉 α
–α
∫R
∫ –j+
–j–t––γ
∣∣ ˆ̃ηαk(〈k〉 α
–α (t – x), 〈k〉 α–α
(tm – x
))∣∣dt dx
+ – j(β+) 〈k〉 α
–α
∫R
∫ –j+
–j–t––γ
∣∣ ˆ̃ηαk(〈k〉 α
–α (t – x), 〈k〉 α–α
(tm – x
))∣∣dt dx
+ – j(β+) 〈k〉 α
–α
×∫R
∫ –j+
–j–t––γ
∣∣∣∣∂ˆ̃ηα
k∂x
(〈k〉 α–α (t – x), 〈k〉 α
–α(tm – x
))∣∣∣∣dt dx
+ – j(β+) 〈k〉 α
–α
×∫R
∫ –j+
–j–t––γ +m
∣∣∣∣∂ˆ̃ηα
k∂x
(〈k〉 α–α (t – x), 〈k〉 α
–α(tm – x
))∣∣∣∣dt dx
� – j(β–γ )
∥∥ ˆ̃ηαk∥∥
L + – jβ∥∥ ˆ̃ηα
k∥∥
L
∫ –j+
–j–t––γ dt
+ – j(β+)
∥∥ ˆ̃ηαk∥∥
L
∫ –j+
–j–t––γ dt + – j(β+)
〈k〉 α–α
∥∥∥∥∂ ˆ̃ηαk
∂x
∥∥∥∥L
∫ –j+
–jt––γ dt
+ – j(β+) 〈k〉 α
–α
∥∥∥∥∂ ˆ̃ηαk
∂x
∥∥∥∥L
∫ –j+
–jt––γ +m dt
� – j(β–γ )
∥∥ ˆ̃ηαk∥∥
L + – j(β+–γ ) 〈k〉 α
–α
∥∥∥∥∂ ˆ̃ηαk
∂x
∥∥∥∥L
+ – j(β+m–γ ) 〈k〉 α
–α
∥∥∥∥∂ ˆ̃ηαk
∂x
∥∥∥∥L
� – j(β–γ ) 〈k〉 α
–α . (.)
Case . If k < , then
∣∣ϕ′′′(t)∣∣ =
∣∣–π[–β(β + )(β + )t–β– + m(m – )(m – )k〈k〉 α
–α tm–]∣∣� t–β– � j(β+).
By the Van der Corput lemma, we have
Ij � – j(β+) j(+γ )∣∣ ˆ̃ηα
k(〈k〉 α
–α(–j+ – x
), 〈k〉 α
–α(–jm+m – x
))∣∣
+ – j(β+)
∫ –j+
–j–
∣∣∣∣ ddt
[θ(jt
)t––γ ˆ̃ηα
k(〈k〉 α
–α (t – x), 〈k〉 α–α
(tm – x
))]∣∣∣∣dt
� – j(β–γ )
∣∣ ˆ̃ηαk(〈k〉 α
–α(–j+ – x
), 〈k〉 α
–α(–jm+m – x
))∣∣
+ – j(β+)
∫ –j+
–j–t––γ jθ ′(jt
)∣∣ ˆ̃ηαk(〈k〉 α
–α (t – x), 〈k〉 α–α
(tm – x
))∣∣dt
+ – j(β+)
∫ –j+
–j–t––γ
∣∣ ˆ̃ηαk(〈k〉 α
–α (t – x), 〈k〉 α–α
(tm – x
))∣∣dt
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 8 of 13
+ – j(β+) 〈k〉 α
–α
∫ –j+
–j–t––γ
∣∣∣∣∂ˆ̃ηα
k∂x
(〈k〉 α–α (t – x), 〈k〉 α
–α(tm – x
))∣∣∣∣dt
+ – j(β+) 〈k〉 α
–α
∫ –j+
–j–t––γ +m
∣∣∣∣∂ˆ̃ηα
k∂x
(〈k〉 α–α (t – x), 〈k〉 α
–α(tm – x
))∣∣∣∣dt.
Thus, similar to (.), we get
∥∥�αk,j
∥∥L � – j(β–γ )
∥∥ ˆ̃ηα
k∥∥
L + – j(β+–γ ) 〈k〉 α
–α
∥∥∥∥∂ ˆ̃ηαk
∂x
∥∥∥∥L
+ – j(β+m–γ ) 〈k〉 α
–α
∥∥∥∥∂ ˆ̃ηαk
∂x
∥∥∥∥L
� – j(β–γ ) 〈k〉 α
–α . (.)
Combining (.) and (.), we have
∥∥�αk,j
∥∥L � – j(β–γ )
〈k〉 α–α .
Therefore, noticing β > γ and combining with (.), we get
∥∥�αk Tβ ,γ (f )
∥∥Lp ≤
∞∑j=
∥∥�αk Tj(f )
∥∥Lp �
∞∑j=
– j(β–γ ) 〈k〉 α
–α ‖f ‖Lp � 〈k〉 α–α ‖f ‖Lp .
So, by Lemma .(), we have
∥∥Tβ ,γ (f )∥∥
Ms,αp,q
� ‖f ‖Ms+α,αp,q .
We finished the proof of Theorem .. �
Theorem . Let �(t) = |t|m or �(t) = |t|m sgn(t). If β > γ > , ≤ p ≤ ∞, < q ≤ ∞ ands ∈R, then Tβ ,γ is bounded from Bs+
p,q to Bsp,q.
Proof As the proof of Theorem ., using the Fourier transformation, the operator Tβ ,γ
can be written as
T̂β ,γ f (ξ, ξ) = mβ ,γ (ξ, ξ)f̂ (ξ, ξ),
where
mβ ,γ (ξ, ξ) =∫
e–π i(ξt+ξtm) e–π it–β
tγ + dt.
Let �j,β ,γ be the kernel of jTβ ,γ , so we have
jTβ ,γ (f )(x) = F–(φj(ξ )mβ ,γ (ξ )) ∗ f (x) = �j,β ,γ ∗ f (x).
Using the Young inequality, we obtain
∥∥ jTβ ,γ (f )∥∥
Lp ≤ ‖�j,β ,γ ‖L‖f ‖Lp .
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 9 of 13
First, let us estimate �j,β ,γ . By a substitution and integrating by parts, we have
�j,β ,γ (x) =∫R
eπ ix·ξφj(ξ )∫
e–π i(ξt+ξtm) e–π it–β
tγ + dt dξ
=∫R
φj(ξ )∫
e–π i[ξ(t–x)+ξ(tm–x)]e–π it–β
t––γ dt dξ dξ
= j∫R
φ(ζ )∫
e–π i[jζ(t–x)+jζ(tm–x)]e–π it–β
t––γ dt dζ dζ
=j
πβi
∫R
φ(ζ )e–π i[jζ(t–x)+jζ(tm–x)]e–π it–βtβ–γ | dζ
–j
πβi
∫R
φ(ζ )∫
e–π it–β
d[tβ–γ e–π i[jζ(t–x)+jζ(tm–x)]]dζ
=j
πβiφ̂(j( – x), j( – x)
)e–π i
–j
πβi(β – γ )
∫R
φ(ζ )∫
e–π it–β
tβ–γ –e–π i[jζ(t–x)+jζ(tm–x)] dt dζ
+j
β
∫R
φ(ζ )∫
e–π it–β
tβ–γ e–π i[jζ(t–x)+jζ(tm–x)](ζ + ζmtm–)dt dζ .
Then, using the Fubini theorem, we get
∣∣�j,β ,γ (x)∣∣ � j∣∣φ̂(
j( – x), j( – x))∣∣
+ j∣∣∣∣∫
e–π it–β
tβ–γ –φ̂(j(t – x), j(tm – x
))dt
∣∣∣∣
+ j∣∣∣∣∫
e–π it–β
tβ–γ ∂φ̂
∂x
(j(t – x), j(tm – x
))dt
∣∣∣∣
+ j∣∣∣∣∫
e–π it–β
tβ–γ +m– ∂φ̂
∂x
(j(t – x), j(tm – x
))dt
∣∣∣∣.
Thus, using a substitution and the Minkowski inequality, it follows that
‖�j,β ,γ ‖L � j∫R
∣∣φ̂(j( – x), j( – x)
)∣∣dx
+ j∫R
∣∣∣∣∫
e–π it–β
tβ–γ –φ̂(j(t – x), j(tm – x
))dt
∣∣∣∣dx
+ j∫R
∣∣∣∣∫
e–π it–β
tβ–γ ∂φ̂
∂x
(j(t – x), j(tm – x
))dt
∣∣∣∣dx
+ j∫R
∣∣∣∣∫
e–π it–β
tβ–γ +m– ∂φ̂
∂x
(j(t – x), j(tm – x
))dt
∣∣∣∣dx
� ‖φ̂‖L + j∥∥∥∥ ∂φ̂
∂x
∥∥∥∥L
+ j∥∥∥∥ ∂φ̂
∂x
∥∥∥∥L
� j.
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 10 of 13
Therefore, we get
∥∥ jTβ ,γ (f )∥∥
Lp � j‖f ‖Lp .
Thus, following Lemma .(), we have
∥∥Tβ ,γ (f )∥∥
Bsp,q
� ‖f ‖Bs+p,q
.
We finished the proof of Theorem .. �
Theorem . If ≤ α ≤ , β > (n + )γ , ≤ p ≤ ∞, < q ≤ ∞ and s ∈ R, then Tn,β ,γ isbounded from Ms+α,α
p,q to Ms,αp,q.
Proof When α = , it is similar to the proof of Theorem .. Thus, we only prove the caseof ≤ α < . Checking the proof of this theorem, we only need to show the case, < t ≤ .When – ≤ t < , the proof is similar. For convenience, we rewrite the operator as follows:
Tn,β ,γ f (x) =∫
f(x – �θ (t)
)e–π it–β
t+γdt.
As in Theorem ., choose a C∞ function �(t) with support in [ , ] on the real line sat-
isfying
∞∑j=
�(jt
) ≡ .
Then we can decompose Tn,β ,γ as
Tn,β ,γ (f )(x) =∞∑j=
∫
�
(jt
)f(x – �θ (t)
)e–π it–β
tγ + dt :=∞∑j=
Tn,j(f )(x).
Using the Fourier transformation, the operator Tn,j can be written as
T̂n,jf (ξ ) = mn,j(ξ )f̂ (ξ ),
where
mn,j(ξ ) =∫
�
(jt
)e–π i(�θ (t)·ξ+t–β )
tγ + dt.
Therefore, we have
�αk Tn,j(f )(x) = F–(ηα
k mn,j) ∗ f (x).
By the Young inequality, we get
∥∥�αk Tn,j(f )
∥∥Lp ≤ ∥∥F–(ηα
k (·)mn,j(·))∥∥
L‖f ‖Lp .
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 11 of 13
Now, we estimate ‖F–(ηαk (·)mn,j(·))‖L . By scaling, we can assume θ = (, , . . . , ) in the
definition of �θ (t). Let η̃αk (ζ ) = ηα
k (〈k〉 α–α ζ + k〈k〉 α
–α ). By a simple substitution and the Fu-bini theorem, we obtain
F–(ηαk mn,j
)(x)
=∫Rn
eπ ix·ξ ηαk (ξ )
∫
�
(jt
)e–π i(�θ (t)·ξ+t–β )t––γ dt dξ
=∫
e–π it–β
t––γ �(jt
)∫Rn
ηαk (ξ )e–π i[(tp –x)ξ+(tp –x)ξ+···+(tpn –xn)ξn] dξ dt
= e∑n
l= π ixlkl〈k〉 α–α 〈k〉 nα
–α
∫
e–π i[t–β +
∑nl= kl〈k〉 α
–α tpl ]t––γ �(jt
)
×∫Rn
η̃αk (ζ )e–π i
∑nl=(tpl –xl)ζl〈k〉 α
–α dζ dt
= e∑n
l= π ixlkl〈k〉 α–α 〈k〉 nα
–α
∫ –j+
–j–eiφ(t)t––γ �
(jt
)
× ˆ̃ηαk(〈k〉 α
–α(tp – x
), 〈k〉 α
–α(tp – x
), . . . , 〈k〉 α
–α(tpn – xn
))dt.
Denote φ(t) = –π [t–β +∑n
l= kl〈k〉 α–α tpl ]. By a normal computation, we get
φ(m)(t) = –π (–)m[β(β + ) · · · (β + m – )t–β–m + (–)mA
m(t)t–m],
where
Am(t) =
n∑l=
kl〈k〉 α–α pl(pl – ) · · · (pl – m + )tpl .
Set Km = {t : (–)mAm(t) ≥ }. Checking the proof in [], when t ∈ Km, we get
∣∣φ(m)(t)∣∣ ≥ Cβ ,mt–β–m.
Then, using the ideas in [] and following the proofs in Theorem ., we obtain Theo-rem .. Here we omit the details. �
Remark . In Theorems . and ., if we take α = , we will get the sufficient resultsin []. Unlike [], we use the discrete definition of modulation spaces. So, our method isdifferent from []. Unfortunately, if < α ≤ , for the scaling property of the decompo-sitions(see Section for details), we will lose the regularity of the space by our method.Maybe we need some new ideas to overcome this limitation.
Remark . In [], the authors mention the fact that the α-modulation space cannot beobtained by interpolation between modulation spaces α = and Besov spaces α = . Thus,it shows that our proofs for < α < in Theorems . and . are meaningful.
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 12 of 13
Remark . We can extend this result to the well-curved γ in Rn. Let γ (t) be a smooth
mapping such that γ () = and
dkγ (t)dtk
∣∣∣t=
, k = , , . . .
spanRn (smooth mappings of finite type in a small neighborhood of the origin). Thenwe call γ (t) well curved. See [, ] for details. According to [] (Proposition .), toevery smooth well-curved γ (t) there exists a constant nonsingular matrix M such thatγ̃ (t) = Mγ (t) is of standard type; that is, approximately homogeneous, taking the form
γk(t) =tak
ak !+ higher order terms
for k = , , . . . , n with ≤ a < a < · · · < an, γ̃ (t) = (γ(t),γ(t), . . . ,γn(t)). Following theideas in [], combining with Theorem ., we can get
Theorem . Let �θ (t) be well-curved. If ≤ α ≤ , β > (n + )γ , ≤ p ≤ ∞, < q ≤ ∞and s ∈R, then Tn,β ,γ is bounded from Ms+α,α
p,q to Ms,αp,q.
4 ConclusionsIn this paper, using the equivalent discrete definition of α-modulation spaces, combin-ing the Fourier transform and Van der Corput lemma, we obtained the strongly singu-lar integrals along homogeneous curves are bounded on the α-modulation spaces for all ≤ α ≤ . Our results extend the main results in []. Our method is also different from [].
AcknowledgementsThis work was completed with the support of National Natural Science Foundation of China (Grant No. 11501516 and11561057), Natural Science Foundation of Zhejiang Province (Grant No. LQ15A010003) and Natural Science Foundationof Jiangxi Province (Grant No. 20151BAB211002).
Competing interestsThe authors declare that they have no competing interests.
Authors’ contributionsAll authors read and approved the final manuscript.
Author details1Xingzhi College, Zhejiang Normal University, Jinhua, 321004, P.R. China. 2Department of Mathematics, Shangrao NormalUniversity, Shangrao, 334001, P.R. China.
Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Received: 28 April 2017 Accepted: 19 July 2017
References1. Zielinski, M: Highly oscillatory singular integrals along curves. PhD thesis, University of Wisconsin, Madison (1995)2. Chandarana, S: Lp-bounds for hypersingular integral operators along curves. Pac. J. Math. 175(2), 389-416 (1996)3. Chen, JC, Fan, DS, Wang, M, Zhu, XR: Lp bounds for oscillatory hyper-Hilbert transform along curves. Proc. Am. Math.
Soc. 136(9), 3145-3153 (2008)4. Cheng, MF, Zhang, ZQ: Hypersingular integrals along homogeneous curves on modulation spaces. Acta Math. Sin.,
Chin. Ser. 53(3), 531-540 (2010)5. Cheng, MF: Hypersingular integral operators on modulation spaces for 0 < p < 1. J. Inequal. Appl. 2012, 165 (2012)6. Gröbner, P: Banachräume glatter funktionen and zerlegungsmethoden. PhD thesis, University of Vienna (1992)7. Han, JS, Wang, BX: α modulation spaces (I). J. Math. Soc. Jpn. 66(4), 1315-1373 (2014)8. Borup, L, Nielsen, M: Boundedness for pseudodifferential operators on multivariate α-modulation spaces. Ark. Mat.
44, 241-259 (2006)
Wu and Yu Journal of Inequalities and Applications (2017) 2017:185 Page 13 of 13
9. Huang, Q, Chen, JC: Cauchy problem for dispersive equations in α–modulation spaces. Electron. J. Differ. Equ. 2014,158 (2014)
10. Huang, Q, Fan, DS, Chen, JC: Critical exponent for evolution equations in modulation spaces. J. Math. Anal. Appl. 443,230-242 (2016)
11. Song, CW: Unimodular Fourier multipliers with a time parameter on modulation spaces. J. Inequal. Appl. 2014, 43(2014)
12. Wu, XM, Chen, JC: Boundedness of fractional integral operators on α-modulation spaces. Appl. Math. J. Chin. Univ.29(3), 339-351 (2014)
13. Zhao, GP, Chen, JC, Fan, DS, Guo, WC: Unimodular Fourier multipliers on homogeneous Besov spaces. J. Math. Anal.Appl. 425, 536-547 (2015)
14. Wang, B, Hao, C, Huo, C: Harmonic Analysis Method for Nonlinear Evolution Equations. I. World Scientfic, Hackensack(2011)
15. Stein, EM: Harmonic Analysis: Real-Variable Methods, Orthogonality, and Oscillatory Integrals. Princeton UniversityPress, Princeton (1993)
16. Guo, WC, Fan, DS, Wu, HX, Zhao, GP: Sharpness of complex interpolation on α-modulation spaces. J. Fourier Anal.Appl. 22(2), 427-461 (2016)
17. Laghi, N, Lyall, N: Strongly singular integrals along curves. Pac. J. Math. 233, 403-415 (2007)18. Stein, EM, Wainger, S: Prblems in harmonic analysis related to curvature. Bull. Am. Math. Soc. 84, 1239-1295 (1978)