We prove a weighted Lp boundedness of Marcinkiewicz integral operators along surfaces on product domains. For various classes of surfaces, we prove the boundedness of the corresponding operators on the weighted Lebsgue space Lp(Rn×Rm,ω1(x)dx,ω2(y)dy), provided that the weights ω1 and ω2 are certain radial weights and that the kernels are rough in the optimal space L(logL)(Sn−1×Sm−1). In particular, we prove the boundedness of Marcinkiewicz integral operators along surfaces determined by mappings that are more general than polynomials and convex functions. Also, in this paper we prove the weighted Lp boundedness of the related square and maximal functions. Our weighted Lp inequalities extend as well as generalize previously known Lp boundedness results.
Citation: Badriya Al-Azri, Ahmad Al-Salman. Weighted Lp norms of Marcinkiewicz functions on product domains along surfaces[J]. AIMS Mathematics, 2024, 9(4): 8386-8405. doi: 10.3934/math.2024408
[1] | Suixin He, Shuangping Tao . Boundedness of some operators on grand generalized Morrey spaces over non-homogeneous spaces. AIMS Mathematics, 2022, 7(1): 1000-1014. doi: 10.3934/math.2022060 |
[2] | Ancheng Chang . Weighted boundedness of multilinear integral operators for the endpoint cases. AIMS Mathematics, 2022, 7(4): 5690-5711. doi: 10.3934/math.2022315 |
[3] | Stevo Stević . Norms of some operators between weighted-type spaces and weighted Lebesgue spaces. AIMS Mathematics, 2023, 8(2): 4022-4041. doi: 10.3934/math.2023201 |
[4] | Yanqi Yang, Shuangping Tao, Guanghui Lu . Weighted and endpoint estimates for commutators of bilinear pseudo-differential operators. AIMS Mathematics, 2022, 7(4): 5971-5990. doi: 10.3934/math.2022333 |
[5] | Tianyang He, Zhiwen Liu, Ting Yu . The Weighted Lp estimates for the fractional Hardy operator and a class of integral operators on the Heisenberg group. AIMS Mathematics, 2025, 10(1): 858-883. doi: 10.3934/math.2025041 |
[6] | Mohammed Ali, Hussain Al-Qassem . On rough generalized Marcinkiewicz integrals along surfaces of revolution on product spaces. AIMS Mathematics, 2024, 9(2): 4816-4829. doi: 10.3934/math.2024233 |
[7] | Pham Thi Kim Thuy, Kieu Huu Dung . Hardy–Littlewood maximal operators and Hausdorff operators on p-adic block spaces with variable exponents. AIMS Mathematics, 2024, 9(8): 23060-23087. doi: 10.3934/math.20241121 |
[8] | Nour Abed Alhaleem, Abd Ghafur Ahmad . Intuitionistic fuzzy normed prime and maximal ideals. AIMS Mathematics, 2021, 6(10): 10565-10580. doi: 10.3934/math.2021613 |
[9] | Qianjun He, Xiang Li . Necessary and sufficient conditions for boundedness of commutators of maximal function on the p-adic vector spaces. AIMS Mathematics, 2023, 8(6): 14064-14085. doi: 10.3934/math.2023719 |
[10] | Mohammed Ali, Qutaibeh Katatbeh, Oqlah Al-Refai, Basma Al-Shutnawi . Estimates for functions of generalized Marcinkiewicz operators related to surfaces of revolution. AIMS Mathematics, 2024, 9(8): 22287-22300. doi: 10.3934/math.20241085 |
We prove a weighted Lp boundedness of Marcinkiewicz integral operators along surfaces on product domains. For various classes of surfaces, we prove the boundedness of the corresponding operators on the weighted Lebsgue space Lp(Rn×Rm,ω1(x)dx,ω2(y)dy), provided that the weights ω1 and ω2 are certain radial weights and that the kernels are rough in the optimal space L(logL)(Sn−1×Sm−1). In particular, we prove the boundedness of Marcinkiewicz integral operators along surfaces determined by mappings that are more general than polynomials and convex functions. Also, in this paper we prove the weighted Lp boundedness of the related square and maximal functions. Our weighted Lp inequalities extend as well as generalize previously known Lp boundedness results.
Let Rn(n≥2) be an n-dimensional Euclidean space, Sn−1 the unit sphere in Rn equipped with normalized Lebesgue measure dσ, and set R+=(0,∞). Furthermore, we let y′=y|y|∈Sn−1 for y≠0 and let Ω∈L1(Sn−1) be a homogeneous function of degree zero on Rn that satisfies
∫Sn−1Ω(y′)dσ(y′)=0. | (1.1) |
The classical Marcinkiewicz integral operator introduced by E. M. Stein in [1] is given by
μΩ(f)(x)=(∞∫−∞|∫|y|<2tf(x−y)Ω(y′)|y|n−1dy|2dt22t)12. | (1.2) |
When Ω∈Lipα(Sn−1)(0<α≤1), Stein [1] proved that μΩ maps Lp(Rn) into Lp(Rn) for all 1<p≤2. In [2], A. Benedek, A. Calderón, and R. Panzone proved that μΩ is bounded on Lp for all 1<p<∞ provided that Ω∈C1(Sn−1). In [3], Walsh proved that μΩ is bounded on L2(Rn) under the weak condition Ω∈L(log+L)12(Sn−1). Moreover, he showed that the L2 boundedness of μΩ may fail if the condition Ω∈L(log+L)12(Sn−1) is replaced by Ω∈L(logL)12−ε(Sn−1) for some ε>0. In 2002, Al-Salman et al. [4] improved Walsh's result by showing that the condition Ω∈L(log+L)12(Sn−1) is also sufficient for the Lp boundedness of μΩ for all p∈(1,∞). For further results and background information about the operator μΩ, we refer readers to [4,5,6,7,8,9] and references therein, among others.
In 1990, Torchinsky and Wang studied the Lp boundedness of the operator μΩ on weighted spaces. In fact, they showed in [10] that μΩ is bounded on Lp(ω)(1<p<∞) if Ω∈Lipα(Sn−1)(0<α≤1) and ω∈Ap (the Muckenhoupt weight class, see[11]). Subsequently, Ding et al. [12] proved that μΩ is bounded on Lp(ω) for p∈(1,∞) provided that Ω∈Lq(Sn−1),q>1, and ωq′∈Ap(Rn). In [13], Lee et al. proved a weighted norm inequality for μΩ under the assumption that Ω is in the Hardy space H1(Sn−1) and the weight ω is in the class ˜AIp(Rn) of radial weights introduced by Duoandikoetxea in [14]. In [15], Al-Salman studied weighted inequalities of the generalized operator
μΩ,Ψ(f)(x)=(∞∫−∞|∫|y|<2tf(x−Ψ(|y|)y′)Ω(y′)|y|n−1dy|2dt22t)12, | (1.3) |
where Ψ:(0,∞)→R is a smooth function satisfying the following growth conditions
|Ψ(t)|⩽C1td,|Ψ′′(t)|⩾C2td−2, | (1.4) |
C3td−1≤|Ψ′(t)|≤C4td−1 | (1.5) |
for some d≠0 and t∈(0,∞) where C1,C2,C3, and C4 are positive constants independent of t. We shall let G be the class of all smooth mappings Ψ:(0,∞)→R that satisfy the growth conditions (1.4)–(1.5). It is clear that G contains all power functions tα(α≠0). It is shown in [15] that μΩ,Ψ is bounded on Lp(ω) for p∈(1,∞) provided that ω∈˜AIp and that Ω is in the optimal space L(log+L)12(Sn−1). Here, we remark that for any q>1 and 0<α≤1, the following inclusions hold and that they are proper
Lipα(Sn−1)⊂Lq(Sn−1)⊂L(logL)(Sn−1)⊂H1(Sn−1), |
and
L(log+L)s(Sn−1)⊂L(log+L)r(Sn−1) whenever r<s. |
In [8], Ding considered the analogy of the operator μΩ on the product domain setting. For Ω∈L1(Sn−1×Sm−1) satisfying
∫Sn−1Ω(u′,.)dσ(u′)=∫Sm−1Ω(.,v′)dσ(v′)=0, | (1.6) |
Ω(tx,sy)=Ω(x,y), for any t,s>0, | (1.7) |
consider the Marcinkiewicz integral operator on the product domains UΩ defined by
UΩf(x,y)=(∞∫−∞∞∫−∞|Ft′,s′(f)(x,y)|2dt′ds′22(t′+s′))12; | (1.8) |
where
Ft′,s′(f)(x,y)=∫∫Λ(t′,s′)f(x−u,y−v)Ω(u′,v′)|u|n−1|v|m−1dudv | (1.9) |
and
Λ(t′,s′)={(u,v)∈Rn×Rm:|u|≤2t′and|v|≤2s′}. |
Ding proved that UΩ is bounded on L2(Rn×Rm) when Ω satisfies the additional assumption of Ω∈L(log+L)2(Sn−1×Sm−1), i.e.,
∫Sn−1∫Sm−1|Ω(u′,v′)|(log(2+|Ω(u′,v′)|)2dσ(u′)dσ(v′)<∞. |
In 2002, Chen et al. [7] improved the result of Ding and showed that UΩ is bounded on Lp(Rn×Rm)(1≤p<∞) under the same condition on Ω. Later, Choi [16] proved that the L2 boundedness of UΩ still holds under the very weak condition Ω∈L(logL)(Sn−1×Sm−1). Subsequently, Al-Qassem et al. [17] substantially improved Choi's result by showing that UΩ is bounded on Lp(Rn×Rm) for all 1<p<∞ under the same condition Ω∈L(logL)(Sn−1×Sm−1). Moreover, they proved that the condition Ω∈L(logL)(Sn−1×Sm−1) is nearly optimal in the sense that the L2 boundedness may fail if the function is assumed to be in L(logL)α(Sn−1×Sm−1)∖L(logL)(Sn−1×Sm−1) for any α<1. For further results for Marcinkiewicz integral operators on product domains, we cite [16,17,18,19,20,21,22], among others.
Motivated by the work in [15] and [18], we consider the weighted Lp boundedness of the Marcinkiewicz integral operator on product domains along surfaces. For suitable mappings Φ,Ψ:[0,∞)→R, consider the UΩ,Φ,Ψ given by
UΩ,Φ,Ψf(x,y)=(∞∫−∞∞∫−∞|FΦ,Ψt′,s′(f)(x,y)|2dt′ds′22(t′+s′))12; | (1.10) |
where
FΦ,Ψt′,s′(f)(x,y)=∫∫Λ(t′,s′)f(x−Φ(|u|)u′,y−Ψ(|v|)v′)Ω(u′,v′)|u|n−1|v|m−1dudv, | (1.11) |
and Λ(t′,s′)={(u,v)∈Rn×Rm:|u|≤2t′and|v|≤2s′}. By specializing to the case Φ(t)=Ψ(t)=t, the operator reduces to the classical Marcinkiewicz integral operator UΩ on product domains. Integral operators on product domains along surfaces have been considered by several authors. For background information, we advise the readers to consult [17,18,19,20,21,22,23] and references therein.
In order to state our results in this paper, we recall the definition of radial weights ˜AIp(Rn) introduced in [14]:
Definition 1.1. Let ω(t)≥0; and ω∈L1loc(R+). For 1<p<∞, we say that ω∈Ap(R+) if there is a positive constant C such that, for any interval I⊆R+,
(|I|−1∫Iω(t)dt)(|I|−1∫Iω(t)−1p−1dt)p−1≤C<∞. |
We say that ω∈A1(R+) if there is a positive constant C such that
ω∗(t)≤Cω(t)fora.e.t∈R+, |
where ω∗ is the Hardy-Littlewood maximal function of ω on R+.
Definition 1.2. Let 1≤p≤∞. We say that ω∈˜Ap(R+) if
ω(x)=ν1(|x|)ν2(|x|)1−p, |
where either νi∈A1(R+) is decreasing or ν2i∈A1(R+),i=1,2.
Definition 1.3. For 1<p<∞, we let
ˉAp(R+)={ω(x)=ω(|x|):ω(t)>0,ω(t)∈L1loc(R+)andω2(t)∈Ap(R+)}. |
Let AIp(Rn) be the weight class defined by exchanging the cubes in the definition of Ap for all n-dimensional intervals with sides parallel to coordinate axes. It is well known that ˉAp(R+)⊆˜Ap(R+) (see [24]). Moreover, if ω(t)∈ˉAp(R+), then ω(|x|) is the Mukenhoupt weighted class Ap(Rn) whose definition can be found in [14]. We let ˜AIp=˜Ap∩AIp.
We shall need the following lemma:
Lemma 1.4. If 1<p<∞, then the weight class ˜AIp(R+) has the following properties:
(ⅰ) ˜AIp1⊂˜AIp2, if 1≤p1<p2<∞;
(ⅱ) For any ω∈˜AIp, there exists an ε>0 such that ω1+ε∈˜AIp;
(ⅲ) For any ω∈˜AIp and p>1, there exists an ε>0 such that p−ε>1 and ω∈˜AIp−ε;
(ⅳ) ω∈˜AIp if and only if ω1−p′∈˜AIp′.
For any weights ω1 and ω2, we let Lp(Rn×Rm,ω1(x)dx,ω2(y)dy) (1<p<∞) be the weighted Lp space associated with the weight ω1 and ω2, i.e., Lp(Rn×Rm,ω1(x)dx,ω2(y)dy)=Lp(ω1,ω2) consists of all measurable functions f with ‖, where
\begin{equation} \Vert f\Vert_{L^{p}(\omega_{1}, \omega_{2})} = \left( \iint_{\mathbb{R} ^{n}\times\mathbb{R}^{m}}|f(x, y)|^{p}\, \omega_{1}(x)\, \omega_{2} (y)\, dx\, dy\right) ^{\frac{1}{p}}. \end{equation} | (1.12) |
In light of the above discussion, the following natural question arises:
Question: Let \mathcal{U}_{\Omega, \Phi, \Psi} be given by (1.8) and assume that \Omega\in L(\log L)(\mathbb{S}^{n-1}\times\mathbb{S}^{m-1}) satisfying (1.6)–(1.7). Assume that \Phi\mathit{, }\Psi\in\mathcal{G} , \omega_{1}\in\tilde{A}_{p}^{I}(\mathbb{R}^{n}) and \omega_{2} \in\tilde{A}_{p}^{I}(\mathbb{R}^{m}) for some 1 < p < \infty. Is \mathcal{U}_{\Omega, \Phi, \Psi} bounded on L^{p}(\omega _{1}, \omega_{2}) ?
In the following we shall answer the above question in the affirmative. In fact, we shall prove that the weighted L^{p} boundedness holds for various classes of mappings \Phi and \Psi .
Theorem 1.5. Suppose that \Omega\in L(\log L)(\mathbb{S}^{n-1}\times\mathbb{S}^{m-1}) satisfying (1.6)–(1.7), \omega_{1}\in\tilde{A}_{p}^{I}(\mathbb{R}^{n}) , and \omega_{2}\in\tilde{A}_{p}^{I}(\mathbb{R}^{m}) . If \Phi \mathit{, }\Psi\in\mathcal{G} , then \mathcal{U}_{\Omega, \Phi, \Psi} is bounded on L^{p}(\omega_{1}, \omega_{2}) for 1 < p < \infty .
We remark here that, by specializing to the case \Phi(t) = \Psi(t) = t , we obtain that the classical operator \mathcal{U}_{\Omega} is bounded on L^{p} (\omega_{1}, \omega_{2}) for 1 < p < \infty . This result, as far as we know, is not known previously. We shall prove in this paper that the weighted boundedness in Theorem 1.5 holds for a more mappings \Phi and \Psi . In order to state our second result, we recall the following class of mappings introduced in [5]:
Definition 1.6. A function \psi:[0, \infty)\rightarrow \mathbb{R} is said to belong to the class \mathcal{PC} _{\lambda}(d) ( d > 0 ) if there exist \lambda\in\mathbb{R} , a polynomial P , and \varphi\in\mathcal{C}^{d+1} [0, \infty) such that
\begin{equation} \begin{array} [c]{ll} (i)\, \psi(t) = P(t)+\lambda\varphi(t) & \\ (ii)\, P(0) = 0\, \, \, \text{and}\mathit{\, }\, \, \varphi^{(j)}(0) = 0\, \, \, \text{for} \, \, \, 0\leqslant j\leqslant d & \\ (iii)\, \varphi^{(j)}\, \, \, { is\; positive \;nondecreasing \;on \; }\, \, \, (0, \infty)\, \, \, \text{ for}\, \, \, 0\leqslant j\leqslant d+1. & \end{array} \end{equation} | (1.13) |
say that In fact, we prove the following:
The class \mathcal{PC}_{\lambda}(d) was introduced in [5]. It is shown in [5] that the class \cup_{d\geq0}\mathcal{PC}_{\lambda}(d) properly contains the class of polynomials \mathcal{P}_{d} of degree less than or equal d as well as the class of convex increasing functions. Examples of functions in \cup_{d\geq0}\mathcal{PC}_{\lambda}(d) that are neither convex nor polynomial are widely available. A particular example is the function \theta(t) = -t^{2}+t^{2}\ln(1+t) . Our second result in this paper is the following:
Theorem 1.7. Suppose that \Omega\in L(\log L)(\mathbb{S}^{n-1}\times\mathbb{S}^{m-1}) satisfying (1.6)–(1.7), \omega_{1}\in\tilde{A}_{p}^{I}(\mathbb{R}^{n}) , and \omega_{2}\in\tilde{A}_{p}^{I}(\mathbb{R}^{m}) . If \Phi \in\mathcal{PC}_{\lambda}(d), \, \Psi\in\mathcal{PC}_{\alpha}(b) for d, \, b > 0 and \lambda, \, \alpha\in\mathbb{R} , then \mathcal{U}_{\Omega, \Phi, \Psi} is bounded on L^{p}(\omega _{1}, \omega_{2}) for 1 < p < \infty with L^{p} bounds independent of \lambda, \alpha\in\mathbb{R} and the coefficients of the particular polynomials involved in the standard representations of \Phi and \Psi .
We remark here that; Theorem 1.7 is the analogy of Theorem 1.3 [15] in the product domain setting. On the other hand, Theorem 1.7 is a generalization of the corresponding result in [18]. More specifically, if \omega_{1}(x) = \omega_{2}(x) = 1 , then Theorem 1.7 reduces to Theorem 1.3 in [18].
We point out here that the method employed in this paper is based on interpolation between good L^{2} estimates and crude L^{p} estimates. The L^{2} estimates depend heavily on the nature of the involved surface. This is clearly expressed interns of the obtained oscillatory estimates. On the other hand, the L^{p} estimates depend on proving the boundedness of the corresponding maximal functions. The the method employed can be used to study the weighted L^{p} boundedness of more general classes of Marcinkiewicz integral operators along surfaces.
Throughout this paper, the letter C will stand for a constant that may vary at each occurrence, but it is independent of the essential variables.
This section is devoted to obtaining weighted estimates of certain square functions and maximal functions. For positive real numbers a and \, b and a Schwartz function \Phi\in \mathcal{S}(\mathbb{R}^{n}\times\mathbb{R}^{m}) , we let
\begin{equation} S_{\Phi, a, b}(f)(x, y) = \left( \int\limits_{-\infty}^{\infty}\int \limits_{-\infty}^{\infty}|\Phi_{a^{t}, \, b^{s}}\ast f(x, y)|^{2}d{t} d{s}\right) ^{\frac{1}{2}} \end{equation} | (2.1) |
where
\Phi_{a^{t}, \, b^{s}}(x, y) = a^{-n{t}}\, b^{-m{s}}\Phi(a^{-{t}}x, b^{-{s}}y). |
It can be observed here that if \Phi(x, y) = \Phi^{(1)}(x)\Phi^{(2)}(y) and f(x, y) = f_{1}(x)f_{2}(y) , then
S_{\Phi, a, b}(f)(x, y) = S_{\Phi^{(1)}, a}(f_{1})(x)S_{\Phi^{(2)}, b}(f_{2})(y) |
where S_{\Phi^{(1)}, a} and S_{\Phi^{(2)}, b} are the square functions in the one parameter setting defined in [15]. Thus, by Lemma 2.1 in [15], it follows that for two Muckenhoupt weights \omega_{1}, \omega_{2}\in A_{p} , we have
\begin{equation} \Vert S_{\Phi, a, b}(f_{1}f_{2})\Vert_{L^{p}(\omega_{1}, \omega_{2})}\leq C_{p}\, \Vert f_{1}\Vert_{L^{p}(\omega_{1})}\Vert f_{2}\Vert_{L^{p}(\omega _{2})} = C_{p}\Vert f_{1}f_{2}\Vert_{L^{p}(\omega_{1}, \omega_{2})}. \end{equation} | (2.2) |
Therefore, it is natural to question if (2.2) holds for general \Phi\in \mathcal{S}(\mathbb{R}^{n}\times\mathbb{R}^{m}) and f\in L^{p}(\omega_{1}, \omega_{2}) . In the following lemma, which is analogues to Lemma 2.1 in [15], we answer this question in the affirmative:
Lemma 2.1. Given a, \, b > 2 and let \psi, \, \theta be \mathcal{C}^{\infty} functions on \mathbb{R} that satisfy the following conditions:
(ⅰ) supp(\psi)\subseteq\left[\dfrac{4}{5a}, \dfrac{5a} {4}\right] \text{ and}\, \, \, supp(\theta)\subseteq\left[\dfrac{4}{5b}, \dfrac{5b}{4}\right].
(ⅱ) \left\vert \dfrac{d^{l}\psi}{du^{l}}(u)\right\vert, \left\vert \dfrac{d^{l}\theta}{du^{l}}(u)\right\vert \leq\dfrac{C_{l}}{u^{l} } for all u and l\geq0 where C_{l} is independent of a and b .
Let \Upsilon\in\mathcal{S}(\mathbb{R}^{n}\times\mathbb{R}^{m}) be given by \widehat{\Upsilon}(\xi, \eta) = \psi(|\xi|^{2}).\, \theta(|\eta|^{2}) and let S_{\Upsilon, a, b} be the square function S_{\Upsilon, a, b} given by (2.1) with \Phi is replaced by \Upsilon . Then, for 1 < p < \infty , \omega_{1}\in A_{p}(\mathbb{R}^{n}) , and \omega_{2}\in A_{p}(\mathbb{R}^{m}) , there exists a constant C_{p} independent of a, \, b such that
\begin{equation} \Vert S_{\Upsilon, a, b}(f)\Vert_{L^{p}(\omega_{1}, \omega_{2})}\leq C_{p}\, \Vert f\Vert_{L^{p}(\omega_{1}, \omega_{2})}. \end{equation} | (2.3) |
Proof: For (\xi, \eta)\in\mathbb{R}^{n}\times\mathbb{R}^{m} , let
m_{a, b}(\xi, \eta, t^{\prime}, s^{\prime}) = \widehat{\Upsilon}(\xi, \eta ) = \psi(|a^{t^{\prime}}\, \xi|^{2})\, \theta(|b^{s^{\prime}}\, \eta|^{2}). |
By the assumption (ii) , we have
\begin{equation} \left( \int\limits_{-\infty}^{\infty}\left\vert \dfrac{\partial^{\alpha }\, m_{a, b}(\xi, \eta, t^{\prime}, s^{\prime})}{\partial\xi^{\alpha}}\right\vert ^{2}dt^{\prime}\, \right) ^{\frac{1}{2}}\leq C_{\alpha} \, |\xi|^{-\alpha} \end{equation} | (2.4) |
and
\begin{equation} \left( \int\limits_{-\infty}^{\infty}\left\vert \dfrac{\partial^{\beta }\, m_{a, b}(\xi, \eta, t^{\prime}, s^{\prime})}{\partial\eta^{\beta}}\right\vert ^{2}ds^{\prime}\, \right) ^{\frac{1}{2}}\leq C_{\beta} \, |\eta|^{-\beta} \end{equation} | (2.5) |
for every multi-index \alpha, \beta with |\alpha|, |\beta|\geq0 , where C_{\alpha}, \, C_{\beta} are constants independent of a and b . We set
K(x, y, t^{\prime}, s^{\prime}) = \Upsilon_{a^{t^{\prime}}, b^{s^{\prime}} }(x, y) = a^{-nt^{\prime}}\, b^{-ms^{\prime}}\, \Upsilon(a^{-t^{\prime} }x, b^{-s^{\prime}}y). |
Then, by (2.4)–(2.5), and a vector-valued analogy of the argument in [25, p. 245–246], we obtain
\begin{equation} \left( \int\limits_{-\infty}^{\infty}\left\vert \dfrac {\partial^{\alpha}\, K(x, y, t^{\prime}, s^{\prime})}{\partial\, x^{\alpha} }\right\vert ^{2}dt^{\prime}\right) ^{\frac{1}{2}}\leq C\, |x|^{-n-|\alpha|}, \end{equation} | (2.6) |
\begin{equation} \left( \int\limits_{-\infty}^{\infty}\left\vert \dfrac {\partial^{\beta}\, K(x, y, t^{\prime}, s^{\prime})}{\partial\, y^{\beta} }\right\vert ^{2}ds^{\prime}\right) ^{\frac{1}{2}}\leq C\, |y|^{-m-|\beta|}, \end{equation} | (2.7) |
for |\alpha|\leq1 and |\beta|\leq1 where C is a constant independent of a and b .
Now, let
\begin{array}{l} g_{\Upsilon, a, b}(f)(x, y) = |\Upsilon_{a^{t^{\prime}}, b^{s^{\prime}}}\ast f(x, y)|, \\ g_{\Upsilon, a}(f)(x, .) = |\Upsilon_{a^{t^{\prime}}}\ast f(x, .)|, \end{array} | (2.8) |
and
\begin{equation} g_{\Upsilon, b}(f)(., y) = |\Upsilon_{b^{s^{\prime}}}\ast f(., y)|. \end{equation} | (2.9) |
Then,
\begin{equation} g_{\Upsilon, a, b}(f)(x, y)\leqslant g_{\Upsilon, a}(g_{\Upsilon, b}(f))(x, y). \end{equation} | (2.10) |
By Plancherel's theorem, we obtain
\begin{equation} \Vert g_{\Upsilon, a}(f)(x, .)\Vert_{L^{2}(\mathbb{R}^{m})}\leq C\, \Vert f\Vert_{L^{2}(\mathbb{R}^{m})} \end{equation} | (2.11) |
and
\begin{equation} \Vert g_{\Upsilon, b}(f)(., y)\Vert_{L^{2}(\mathbb{R}^{n})}\leq C\, \Vert f\Vert_{L^{2}(\mathbb{R}^{n})}. \end{equation} | (2.12) |
Hence, by the Corollary on page 205 in [25], and (2.4), (2.6), and (2.11), we have
\begin{equation} \int_{\mathbb{R}^{n}}|g_{\Upsilon, a}(f)(x, .)|^{p}w_{1}{(x)}\, dx\leqslant C\, \int_{\mathbb{R}^{n}}|(f)(x, .)|^{p}\, w_{1}{(x)}\, dx \end{equation} | (2.13) |
for w_{1}{(x)}\in A_{p}(\mathbb{R}^{n}).
Thus, by (2.13) and following similar arguments as in [26], we get
\begin{equation} \int_{\mathbb{R}^{n}}|g_{\Upsilon, a}(f)(x, y)|^{p}\, w_{1}{(x)}\, w_{2} {(y)}\, dx\leqslant C\, \int_{\mathbb{R}^{n}}|(f)(x, y)|^{p}w_{1}{(x)} \, w_{2}{(y)}dx. \end{equation} | (2.14) |
for each y\in\mathbb{R}^{m} with C independent of y . Then, by integration over \mathbb{R}^{m} , we get
\begin{equation} \int_{\mathbb{R}^{m}}\int_{\mathbb{R}^{n}}|g_{\Upsilon, a}(f)(x, y)|^{p} \, w_{1}{(x)}\, w_{2}{(y)}\, dx\, dy\leqslant C\, \int_{\mathbb{R}^{m}} \int_{\mathbb{R}^{n}}|(f)(x, y)|^{p}w_{1}{(x)}\, w_{2}{(y)}dx\, dy. \end{equation} | (2.15) |
By repeating the argument between (2.13) and (2.15) for g_{\Upsilon, b}(f)(., y) , and replacing x by y , we get
\begin{equation} \int_{\mathbb{R}^{n}}\int_{\mathbb{R}^{m}}|g_{\Upsilon, b}(f)(x, y)|^{p} \, w_{1}{(x)}\, w_{2}{(y)}\, dx\, dy\leqslant C\, \int_{\mathbb{R}^{n}} \int_{\mathbb{R}^{m}}|(f)(x, y)|^{p}w_{1}{(x)}\, w_{2}{(y)}dx\, dy. \end{equation} | (2.16) |
Finally, by (2.10), inequality (2.3) follows vector-valued analogues of the argument in the proof of the Theorem 3 in [26, p. 128], and (2.15)–(2.16). This ends the proof of Lemma 2.1.
Now, for \Omega\in L^{1}(\mathbb{S}^{n-1}\times\mathbb{S}^{m-1}) and suitable mappings \Phi, \Psi:(0, \infty)\longrightarrow\mathbb{R} , we define the family of measures \{\sigma_{\Omega, \, \Phi, \Psi, a^{t^{\prime} }, b^{s^{\prime}}}:{t^{\prime}}, \, {s^{\prime}}\in\mathbb{R}\} by
\begin{equation} \int_{\mathbb{R}^{n}\times\mathbb{R}^{m}}fd\sigma_{\Omega, \, \Phi , \, \Psi, a^{t^{\prime}}, b^{s^{\prime}}} = a^{-{t^{\prime}}}\, b^{-{s^{\prime}} }\underset{{\substack{\vert u \vert < a^{{t^{\prime}}} \\ \vert v \vert < b^{{s^{\prime}}}}}}{\iint}f(x-\Phi(|u|)\, u^{\prime}, y-\Psi(|v|)\, v^{\prime })\dfrac{\Omega(u^{\prime}, v^{\prime})}{|u|^{n-1}\, |v|^{m-1}}\, dudv. \end{equation} | (2.17) |
We let \mathcal{M}_{\Omega, \, \Phi, \, \Psi, a, b} be the maximal function corresponding to the family \{\sigma_{\Omega, \, \Phi, \, \Psi, a^{t^{\prime} }, b^{s^{\prime}}}:{t^{\prime}}, \, {s^{\prime}}\in\mathbb{R}\} , i.e.,
\begin{equation} \mathcal{M}_{\Omega, \, \Phi, \, \Psi, a, b}(f)(x, y) = \sup\limits_{{t^{\prime} }, \, {s^{\prime}}\in\mathbb{R}}\left\vert \sigma_{\Omega, \, \Phi, \, \Psi , a^{t^{\prime}}, b^{s^{\prime}}}\ast f(x, y)\right\vert . \end{equation} | (2.18) |
Then, we have the following lemma:
Lemma 2.2. Suppose that \Omega\in L^{1} (\mathbb{S}^{n-1}\times\mathbb{S}^{m-1}) satisfying (1.7). For a, b > 2 and suitable \Phi, \Psi:(0, \infty)\longrightarrow\mathbb{R} , let \mathcal{M}_{\Omega, \, \Phi, \, \Psi, a, b} be the maximal function defined by (2.18). Suppose that (i) \Phi\mathit{, } \Psi\in\mathcal{G} ; or (ii) \Phi\in\mathcal{PC}_{\lambda} (d_{1}), \, \Psi\in\mathcal{PC}_{\alpha}(d_{2}) for d_{1}, \, d_{2} > 0 and \lambda, \, \alpha\in\mathbb{R} . Then, for 1 < p < \infty and \omega_{1}\in A_{p})(\mathbb{R}^{n}), \, \omega _{2}\in A_{p})(\mathbb{R}^{m}) , there exists a constant C_{p} independent of \Omega, \, a, and b such that
\begin{equation} \Vert\mathcal{M}_{\Omega, \, \Phi, \, \Psi, a, b}(f)\Vert_{L^{p}(\omega_{1} , \omega_{2})}\leq C_{p}\, \Vert\Omega\Vert_{L^{1}}\, \, \Vert f\Vert _{L^{p}(\omega_{1}, \omega_{2})}. \end{equation} | (2.19) |
Proof: We shall start by verifying (2.19) under assumption (ii) on the functions \Phi and \Psi . Notice that
\begin{equation} \mathcal{M}_{\Omega, \, \Phi, \, \Psi, a, b}(f)\leq\mathcal{M}_{\Omega, \, \Phi , \, \Psi}(f) = \sup\limits_{t^{\prime}, \, s^{\prime}\in\mathbb{R}}\left\vert \sigma_{\Omega, \, \Phi, \, \Psi, 2^{t^{\prime}}, 2^{s^{\prime}}}\ast f(x, y)\right\vert . \end{equation} | (2.20) |
Thus, it is enough to show that
\begin{equation} \Vert\mathcal{M}_{\Omega, \, \Phi, \, \Psi}(f)\Vert_{L^{p}(\omega_{1}, \omega_{2} )}\leq C_{p}\, \Vert\Omega\Vert_{L^{1}}\, \, \Vert f\Vert_{L^{p}(\omega _{1}, \omega_{2})}. \end{equation} | (2.21) |
We define the one parameter maximal functions
\mathcal{M}_{\Omega, \, \Psi}(f)(\cdot, y) = \sup\limits_{s^{\prime}\in\mathbb{R} }\left\vert {2^{-s^{\prime}}}\int_{|v| < 2^{s^{\prime}}}f(., y-\Psi (|v|)\, v^{\prime})\dfrac{\Omega(\cdot, v^{\prime})}{|v|^{m-1}}\, dv\right\vert , |
and
\mathcal{M}_{\Omega, \Phi}(f)(x, \cdot) = \sup\limits_{t^{\prime}\in\mathbb{R} }\left\vert {2^{-t^{\prime}}}\int_{|u| < 2^{t^{\prime}}}f(x-\Phi(|u|)\, u^{\prime }, .)\dfrac{\Omega(u^{\prime}, \cdot)}{|u|^{n-1}}\, du\right\vert . |
Then,
\begin{equation} \mathcal{M}_{\Omega, \Phi, \Psi}(f)(x, y)\leq\mathcal{M}_{\Omega, \Phi }(\mathcal{M}_{\Omega, \Psi}(f)(\cdot, y)(x, \cdot). \end{equation} | (2.22) |
By polar coordinates, we have
\begin{equation} \begin{array} [c]{ll} \mathcal{M}_{\Omega, \, \Psi}f(\cdot, y) & \leq\int \limits_{\mathbb{S}^{m-1}}|\Omega(\cdot, v^{\prime})|\, \, \mathcal{M} _{\Psi, v^{\prime}}f(\cdot, y)d\sigma(v^{\prime}), \end{array} \end{equation} | (2.23) |
where
\mathcal{M}_{\Psi, v^{\prime}}(f)(., y) = \sup\limits_{s^{\prime}\in\mathbb{R} }2^{-s^{\prime}}\int\limits_{0}^{2^{s^{\prime}}}|f(\cdot, y-\Psi(r^{\prime })v^{\prime})|dr^{\prime}. |
Now, we have
\begin{align} \mathcal{M}_{\Psi, v^{\prime}}(f)(., y) & \leq\sum\limits_{j = 0}^{\infty} 2^{-j}\, \left( \sup\limits_{{s^{\prime}\in\mathbb{R}}}2^{-s^{\prime}+j} \int\limits_{2^{s^{\prime}-j-1}}^{2^{s^{\prime}-j}}|f(., y-\Psi(r^{\prime })v^{\prime})|dr^{\prime}\right) \\ & \leq C\sum\limits_{j = 0}^{\infty}2^{-j}\left( \sup\limits_{z > 0}\dfrac{1} {z}\int\limits_{0}^{cz}|f(., y-r^{\prime}\, v^{\prime})|dr^{\prime}\right) \end{align} | (2.24) |
\begin{align} & = C\sup\limits_{z > 0}\dfrac{1}{z}\int\limits_{0}^{cz}|f(., y-r^{\prime }\, v^{\prime})|dr^{\prime}; \end{align} | (2.25) |
where (2.24) follows by change of variables and (1.4). By (8) in [14] and since \omega_{2}\in\tilde{A_{p}}(\mathbb{R}^{m}) , we get
\begin{equation} \Vert\mathcal{M}_{\Psi, v^{\prime}}(f)\Vert_{L^{p}(\omega_{2})}\leq C_{p}\, \Vert f\Vert_{L^{p}(\omega_{2})}; \end{equation} | (2.26) |
where C_{p} is a constant independent of v^{\prime} . By a similar argument, for \omega_{1}\in\tilde{\mathit{A}_{p}}(\mathbb{R}^{n}) , we get
\begin{equation} \Vert\mathcal{M}_{\Phi, u^{\prime}}(f)\Vert_{L^{p}(\omega_{1})}\leq C_{p}\, \Vert f\Vert_{L^{p}(\omega_{1})}, \end{equation} | (2.27) |
where
\mathcal{M}_{\varphi, u^{\prime}}(f)(x, .)\leq C\, \sup\limits_{s > 0}\dfrac{1} {s}\int\limits_{0}^{cs}|f(x-t\, u^{\prime})|dt. |
Thus, by (2.23), (2.26), and Minkowski's inequality, we get
\begin{equation} \Vert\mathcal{M}_{\Omega, \, \Psi}(f)\Vert_{L^{p}(\omega_{2})}\leq C_{p} \, \Vert\Omega\Vert_{L^{1}}\, \Vert f\Vert_{L^{p}(\omega_{2})}. \end{equation} | (2.28) |
Similarly, for \omega_{1}\in\tilde{\mathit{A}_{p}}(\mathbb{R}^{n}) , we get
\begin{equation} \Vert\mathcal{M}_{\Omega, \, \Phi}(f)\Vert_{L^{p}(\omega_{1})}\leq C_{p} \, \Vert\Omega\Vert_{L^{1}}\Vert f\Vert_{L^{p}(\omega_{1})}. \end{equation} | (2.29) |
Now, by (2.28) and following a similar argument as in [26], we have
\begin{equation} \int_{\mathbb{R}^{m}}\mathcal{M}_{\Omega, \phi}(f)(x, y)|^{p}\, w_{1}{(x)} \, w_{2}{(y)}\, dy\leqslant C\, \Vert\Omega\Vert_{L^{1}}\, \int_{\mathbb{R}^{m} }|f(x, y)|^{p}w_{1}{(x)}\, w_{2}{(y)}\, dy, \end{equation} | (2.30) |
for x\in\mathbb{R}^{n} where C is a constant independent of x . Then, by integration with respect to x , we get
\begin{equation} \int_{\mathbb{R}^{n}}\int_{\mathbb{R}^{m}}\left\vert \mathcal{M}_{\Omega, \Psi }(f)(x, y)\right\vert ^{p}w_{1}{(x)}\, w_{2}{(y)}\, dx\, dy\leqslant C\, \Vert\Omega\Vert_{L^{1}}\, \int_{\mathbb{R}^{n}}\int_{\mathbb{R}^{m} }|f(x, y)|^{p}w_{1}{(x)}\, w_{2}{(y)}dx\, dy. \end{equation} | (2.31) |
Thus, by following a similar argument as in (2.30)–(2.31) on \mathcal{M}_{\Omega, \Phi} , replacing x by y , we get
\begin{equation} \int_{\mathbb{R}^{n}}\int_{\mathbb{R}^{m}}|\mathcal{M}_{\Omega, \Phi }(f)(x, y)|^{p}\, w_{1}{(x)}\, w_{2}{(y)}\, dx\, dy\leqslant C\, \Vert\Omega \Vert_{L^{1}}\, \int_{\mathbb{R}^{n}}\int_{\mathbb{R}^{m}}|f(x, y)|^{p} w_{1}{(x)}\, w_{2}{(y)}dx\, dy. \end{equation} | (2.32) |
Thus, by (2.22), we have
\begin{equation} \Vert\mathcal{M}_{\Omega, \, \Phi, \, \Psi}(f)\Vert_{L^{p}(\omega_{1}, \omega_{2} )}\leq C_{p}\, \Vert\Omega\Vert_{L^{1}}\, \, \Vert f\Vert_{L^{p}(\omega _{1}, \omega_{2})}. \end{equation} | (2.33) |
Hence, (2.19) follows by (2.20) and (2.33). This ends the proof of (2.19) under assumption (ⅱ) on the functions \Phi and \Psi . To prove (2.19) under assumption (ⅰ) on the functions \Phi and \Psi , we follow a similar argument as above and make use of estimates developed in the proof of Lemma 2.2 in [15]. We omit the details. This ends the proof of the lemma.
Next, we prove the following weighted inequalities for square functions:
Lemma 2.3. Suppose that \Vert\Omega\Vert_{L^{1} }\leq1 . Suppose also that a , b , \Psi , \Phi , and \Upsilon are as in Lemma 2.2. Let For t, s\in\mathbb{R} , let \sigma_{\Omega, \, \Phi, \, \Psi, a, b} be given by (2.17) where t^{\prime} and s^{\prime} are replaced by t and s , respectively. Assume that (i) \Phi\mathit{, }\Psi \in\mathcal{G} ; or (ii) \Phi\in\mathcal{PC}_{\lambda}(d_{1}), \, \Psi\in\mathcal{PC}_{\alpha}(d_{2}) for d_{1}, \, d_{2} > 0 and \lambda, \, \alpha\in\mathbb{R} . Then, for 1 < p < \infty, j, \, k\in\mathbb{Z} , \omega_{1}\in\tilde{A}_{p} ^{I}(\mathbb{R}^{n}), and \, \omega_{2}\in\tilde{A}_{p} ^{I}(\mathbb{R}^{m}) , and there exists a constant C_{p} independent of a, \, b, \, j, \, k , and \Omega such that
\begin{equation} \left\Vert \left( \int\limits_{-\infty}^{\infty}\int\limits_{-\infty} ^{\infty}|\sigma_{\Omega, \, \Phi, \, \Psi, a^{t}, b^{s}}\ast\Upsilon_{a^{t+j} , \, b^{s+k}}\ast f(x, y)|^{2}dt\, ds\right) ^{\frac{1}{2}}\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}\leq C\, \Vert f\Vert_{L^{p}(\omega_{1} , \omega_{2})}. \end{equation} | (2.34) |
Proof: Notice that
\begin{array} [c]{ll} \sup\limits_{t, s\in\mathbb{R}}\left\vert \sigma_{\Omega, \, \Phi, \, \Psi , a^{t}, b^{s}}\ast\Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)\right\vert & = \mathcal{M}_{\Omega, \, \Phi, \, \Psi, a, b}(\Upsilon_{a^{t+j}, \, b^{s+k}}\ast f)(x, y)\\ & \leq\mathcal{M}_{\Omega, \, \Phi, \, \Psi, a, b}\left( \sup\limits_{t, s\in \mathbb{R}}\left\vert \Upsilon_{a^{t+j}, \, b^{s+k}}\ast f\right\vert \right) (x, y). \end{array} |
Next, by Lemma 2.2, we have
\begin{equation} \begin{array} [c]{l} \left\Vert \sup\limits_{t, s\in\mathbb{R}}\left\vert \sigma_{\Omega , \, \Phi, \, \Psi, a^{t}, b^{s}}\ast\Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)\right\vert \right\Vert _{L^{p}(\omega_{1}, \omega_{2})}\\ \leq\left\Vert \mathcal{M}_{\Omega, \, \Phi, \, \Psi, a, b}\left( \sup \limits_{t, s\in\mathbb{R}}|\Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)|\right) \right\Vert _{L^{p}(\omega_{1}, \omega_{2})}\\ \leq C\, \left\Vert \sup\limits_{t, s\in\mathbb{R}}|\Upsilon_{a^{t+j}, \, b^{s+k} }\ast f(x, y)|\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}. \end{array} \end{equation} | (2.35) |
Now, by duality, choose a non-negative function g(x, y) with \Vert g\Vert_{{L^{p^{\prime}}}(\omega_{1}^{1-p^{\prime}}, \omega_{2}^{1-p^{\prime}})}\leq1 such that
\begin{equation} \begin{array} [c]{l} \left\Vert \int\limits_{-\infty}^{\infty}\int\limits_{-\infty }^{\infty}\left\vert \sigma_{\Omega, \, \Phi, \, \Psi, a^{t}, b^{s}}\ast \Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)\right\vert dt\, ds\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}\\ \leq\iint\limits_{\mathbb{R}^{n}\times\mathbb{R}^{m}} \int\limits_{-\infty}^{\infty}\int\limits_{-\infty}^{\infty}\left\vert \sigma_{\Omega, \, \Phi, \, \Psi, a^{t}, b^{s}}\ast\Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)\right\vert \, g(x, y)dt\, ds\, dx\, dy\\ \leq C\, \iint\limits_{\mathbb{R}^{n}\times\mathbb{R}^{m}} \int\limits_{-\infty}^{\infty}\int\limits_{-\infty}^{\infty}\left\vert \Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)\right\vert \, \left( \sup \limits_{t, s\in\mathbb{R}}|\sigma_{\Omega, \, \Phi, \, \Psi, a^{t}, b^{s}}\ast g(x, y)|\right) dt\, dsdx\, dy\\ \leq C\, \iint\limits_{\mathbb{R}^{n}\times\mathbb{R}^{m}} \int\limits_{-\infty}^{\infty}\int\limits_{-\infty}^{\infty}\left\vert \Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)\right\vert \, \mathcal{M} _{\Omega, \, \Phi, \, \Psi, a, b}(\tilde{g})(-x, -y)\, dt\, ds\, dx\, dy, \end{array} \end{equation} | (2.36) |
where \tilde{g}(x, y) = g(-x, -y) . Thus, by Lemma 2.1, (2.36), and Hölder's inequality, we get
\begin{equation} \begin{array} [c]{l} \left\Vert \int\limits_{-\infty}^{\infty}\int\limits_{-\infty }^{\infty}\left\vert \sigma_{\Omega, \, \Phi, \, \Psi, a^{t}, b^{s}}\ast \Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)\right\vert \right\Vert _{L^{p} (\omega_{1}, \omega_{2})}\\ \leq \left\Vert \int\limits_{-\infty}^{\infty}\int\limits_{-\infty }^{\infty}\left\vert \Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)\right\vert dt\, dr\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}\, \left\Vert \mathcal{M} _{\Omega, \, \varphi, \, \phi, a, b}(\tilde{g})\right\Vert _{{L^{p^{\prime}}} (\omega_{1}^{1-p^{\prime}}, \omega_{2}^{1-p^{\prime}})}. \end{array} \end{equation} | (2.37) |
By an application of Lemma 2.2, we get
\begin{equation} \begin{array} [c]{l} \left\Vert \int\limits_{-\infty}^{\infty}\int\limits_{-\infty }^{\infty}\left\vert \sigma_{\Omega, \, \Phi, \, \Psi, a^{t}, b^{s}}\ast \Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)\right\vert dt\, dr\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}\\ \leq C\, \left\Vert \int\limits_{-\infty}^{\infty}\int \limits_{-\infty}^{\infty}\left\vert \Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)\right\vert dtds\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}. \end{array} \end{equation} | (2.38) |
Hence, by interpolation between (2.35) and (2.38) in a vector-valued setting, we get
\begin{equation} \begin{array} [c]{l} \left\Vert \left( \int\limits_{-\infty}^{\infty}\int \limits_{-\infty}^{\infty}\left\vert \sigma_{\Omega, \, \Phi, \, \Psi, a^{t}, b^{s} }\ast\Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)\right\vert ^{2}dt\, ds\right) ^{\frac{1}{2}}\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}\\ \leq C\, \left\Vert \left( \int\limits_{-\infty }^{\infty}\int\limits_{-\infty}^{\infty}\left\vert \Upsilon_{a^{t+j} , \, b^{s+k}}\ast f(x, y)\right\vert ^{2}dtds\right) ^{\frac{1}{2}}\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}\\ \leq C\, \Vert f\Vert_{L^{p}(\omega_{1}, \omega_{2})}, \end{array} \end{equation} | (2.39) |
where the last inequality is obtained by Lemma 2.1. This completes the proof of Lemma 2.3.
This section is establish some preliminary estimates that are needed to prove our results.
Lemma 3.1. Let \Omega\in L^{2}(\mathbb{S} ^{n-1}\times\mathbb{S}^{m-1}) satisfying (1.6)–(1.7) with \left\Vert \Omega\right\Vert _{1}\leq1 and \left\Vert \Omega\right\Vert _{2}\leq A for some A > 2 . Suppose that \Phi\mathit{, }\Psi\in\mathcal{G} with powers d_{1}, d_{1} in (1.4)–(1.5). For t, \, s\in\mathbb{R} , let \sigma_{A, {t}, {s}}^{(\Phi, \Psi)} be the measure defined via the Fourier transform by
\begin{equation} \hat{\sigma}_{A, {t}, {s}}^{(\Phi, \Psi)}(\xi, \eta) = \frac{1}{A^{t+s}} \iint\limits_{\Gamma(A^{t}, A^{s})}e^{-i\left( \Phi(|u|)\, \xi.u^{\prime }+\, \Psi(|v|)\, \eta.v^{\prime}\right) }\dfrac{\Omega_{\kappa}(u^{\prime }, v^{\prime})}{|u|^{n-1}\, |v|^{m-1}}\, du\, dv , \end{equation} | (3.1) |
where
\begin{equation} \Gamma(A^{t}, A^{s}) = \{(u, v)\in\mathbb{R}^{n}\times\mathbb{R}^{m} :A^{t-1} < |u|\leq A^{t}\, \, \;{ and }\;\, \, A^{s-1} < |v|\leq A^{s}\}. \end{equation} | (3.2) |
Then, there exists \varepsilon\in(0, \frac{1}{2}) such that
\begin{align} \left\vert \hat{\sigma}_{A, {t}, {s}}^{(\Phi, \Psi)}(\xi, \eta)\right\vert & \leq1 \end{align} | (3.3) |
\begin{align} \left\vert \hat{\sigma}_{A, {t}, {s}}^{(\Phi, \Psi)}(\xi, \eta)\right\vert & \leq C\, \, |A^{d_{1}t}\, \xi\, |^{-\frac{\varepsilon}{2\log_{2}A}}\, \, \, |A^{d_{2} s}\, \eta\, |^{-\frac{\varepsilon}{2\log_{2}A}}; \end{align} | (3.4) |
\begin{align} \left\vert \hat{\sigma}_{A, {t}, {s}}^{(\Phi, \Psi)}(\xi, \eta)\right\vert & \leq C\, \, |A^{d_{1}t}\, \xi\, |^{-\frac{\varepsilon}{2\log_{2}A}}\, \, \, |A^{d_{2} s}\, \eta\, |^{\frac{\varepsilon}{2\log_{2}A}} \end{align} | (3.5) |
\begin{align} \left\vert \hat{\sigma}_{A, {t}, {s}}^{(\Phi, \Psi)}(\xi, \eta)\right\vert & \leq C\, \, |A^{d_{1}t}\, \xi\, |^{\frac{\varepsilon}{2\log_{2}A}}\, \, \, |A^{d_{2} s}\, \eta\, |^{-\frac{\varepsilon}{2\log_{2}A}} \end{align} | (3.6) |
\begin{align} \left\vert \hat{\sigma}_{A, {t}, {s}}^{(\Phi, \Psi)}(\xi, \eta)\right\vert & \leq C\, \, |A^{d_{1}t}\, \xi\, |^{\frac{\varepsilon}{2\log_{2}A}}\, \, \, |A^{d_{2} s}\, \eta\, |^{\frac{\varepsilon}{2\log_{2}A}} \end{align} | (3.7) |
where the constant C is independent of A, s, and t .
Proof: We shall assume that d_{1}, d_{2} > 0 . The other cases follows by similar argument. The estimate (3.3) is clear. To see the estimate (3.4), notice that
\begin{equation} \left\vert \hat{\sigma}_{A, {t}, {s}}^{(\Phi, \Psi)}(\xi, \eta)\right\vert \leq\int_{\mathbb{S}^{n-1}}\int_{\mathbb{S}^{m-1}}\left\vert \Omega(u^{\prime }, v^{\prime})\right\vert g(A, \Phi, \xi)g(A, \Psi, \eta)d\sigma(u^{\prime} )d\sigma(v^{\prime}) \end{equation} | (3.8) |
where
g(A, \Phi, \xi) = \left\vert \int\limits_{\frac{1}{A}}^{1}e^{-i\Phi(A^{t} r)\, \xi.u^{\prime}}dr\right\vert, |
and g(A, \Psi, \eta) has similar definition as g(A, \Phi, \xi) . By integration by parts along with the assumptions (1.4)–(1.5), and the observations g(A, \Phi, \xi)\leq1 and g(A, \Psi, \eta)\leq1 , there exists \varepsilon\in(0, \frac{1}{2}) such that
\begin{align} g(A, \Phi, \xi) & \leq C\, |A^{d_{1}t}\, \xi\cdot u^{\prime}\, |^{-\varepsilon } \end{align} | (3.9) |
\begin{align} g(A, \Psi, \eta) & \leq C\, |A^{d_{2}s}\, \eta\cdot v^{\prime}\, |^{-\varepsilon} . \end{align} | (3.10) |
By (3.8), (3.9)–(3.10), Hölder's inequality, and assumption on \Omega , we have
\begin{equation} \left\vert \hat{\sigma}_{A, {t}, {s}}^{(\Phi, \Psi)}(\xi, \eta)\right\vert \leq A|A^{d_{1}t}\, \xi|^{-\varepsilon}|A^{d_{2}s}\, \eta\, |^{-\varepsilon}C\tilde {C}_{\varepsilon}, \end{equation} | (3.11) |
where
\tilde{C}_{\varepsilon} = \sup\limits_{\xi^{^{\prime}}\in\mathbb{S}^{n-1}}\left( \int_{\mathbb{S}^{n-1}}\, |\xi^{^{\prime}}\cdot u^{\prime}\, |^{-2\varepsilon }d\sigma(u^{\prime})\right) ^{\frac{1}{2}}\sup\limits_{\eta^{^{\prime}}\in \mathbb{S}^{m-1}}\left( \int_{\mathbb{S}^{m-1}}\, |\eta^{^{\prime}}\cdot v^{\prime}\, |^{-2\varepsilon}d\sigma(v^{\prime})\right) ^{\frac{1}{2}}. |
Since \varepsilon < 1/2 , we have \tilde{C}_{\varepsilon} < \infty . Thus, by (3.9)–(3.10), (3.8), and an interpolation, we get
\left\vert \hat{\sigma}_{A, {t}, {s}}^{(\Phi, \Psi)}(\xi, \eta)\right\vert \leq A^{\frac{1}{2\log_{2}A}}|A^{d_{1}t}\, \xi|^{-\frac{\varepsilon}{2\log_{2}A} }|A^{d_{2}s}\, \eta\, |^{-\frac{\varepsilon}{2\log_{2}A}}C, |
which implies (3.4) since A^{\frac{1}{2\log_{2}A}} < 2 . To verify (3.5), we first notice that
\begin{equation} \frac{1}{A^{s}}\left\vert \int\limits_{A^{s-1}}^{A^{s}}\left( e^{-i\Psi (A^{s}r)\, \eta.v^{\prime}}-1\right) dr\right\vert \leq\min\{1, C_{1} |A^{d_{2}s}\, \eta\, |\}, \end{equation} | (3.12) |
which by interpolation implies
\begin{equation} \frac{1}{A^{s}}\left\vert \int\limits_{A^{s-1}}^{A^{s}}\left( e^{-i\Psi (A^{s}r)\, \eta.v^{\prime}}-1\right) dr\right\vert \leq C|A^{d_{2}s} \, \eta\, |^{\varepsilon}. \end{equation} | (3.13) |
By the cancellation property (1.6), Hölder's inequality, the assumption on \Omega , (3.9), and (3.13), we have
\begin{align} \left\vert \hat{\sigma}_{A, {t}, {s}}^{(\Phi, \Psi)}(\xi, \eta)\right\vert & \leq\frac{1}{A^{t+s}}\left\vert \iint\limits_{\Gamma(A^{t}, A^{s})}\left( e^{-i\left( \Phi(|u|)\, \xi.u^{\prime}+\Psi(|v|)\, \eta.v^{\prime}\right) }-e^{-i\Phi(|u|)\, \xi.u^{\prime}}\right) \dfrac{\Omega_{\kappa}(u^{\prime }, v^{\prime})}{|u|^{n-1}\, |v|^{m-1}}\, du\, dv\right\vert \\ & \leq C|A^{d_{2}s}\, \eta\, |^{\varepsilon}\int_{\mathbb{S}^{n-1}} \int_{\mathbb{S}^{m-1}}\left\vert \Omega(u^{\prime}, v^{\prime})\right\vert g(A, \Phi, \xi)d\sigma(u^{\prime})d\sigma(v^{\prime})\\ & \leq C|A^{d_{2}s}\, \eta\, |^{\varepsilon}\left\Vert \Omega\right\Vert _{2}\left\vert \mathbb{S}^{m-1}\right\vert \left( \int_{\mathbb{S}^{n-1} }\left\vert g(A, \Phi, \xi)\right\vert ^{2}d\sigma(u^{\prime})\right) ^{\frac{1}{2}}\\ & \leq C|A^{d_{2}s}\, \eta\, |^{\varepsilon}\left\Vert \Omega\right\Vert _{2}\left\vert \mathbb{S}^{m-1}\right\vert \, |A^{d_{1}t}\, \xi|^{-\varepsilon }\sup\limits_{\xi^{^{\prime}}\in\mathbb{S}^{n-1}}\left( \int_{\mathbb{S}^{n-1} }\, |\xi^{^{\prime}}\cdot u^{\prime}\, |^{-2\varepsilon}d\sigma(u^{\prime })\right) ^{\frac{1}{2}}\\ & \leq CA|A^{d_{2}s}\, \eta\, |^{\varepsilon}|A^{d_{1}t}\, \xi|^{-\varepsilon}, \end{align} | (3.14) |
where the last inequality follows by the same reasoning for \tilde {C}_{\varepsilon} above. Thus, (3.5) follows by (3.14), (3.3), and an interpolation. The verifications of other estimates follows by a similar argument with minor modifications. We omit the details. This completes the proof of the lemma.
Now, by the same argument as in [18], we have the following lemma:
Lemma 3.2. Let \Omega\in L^{2}(\mathbb{S} ^{n-1}\times\mathbb{S}^{m-1}) satisfying (1.6)–(1.7) with \left\Vert \Omega\right\Vert _{1}\leq1 and \left\Vert \Omega\right\Vert _{2}\leq A for some A > 2 . Let \Phi\in\mathcal{PC}_{\lambda}(d_{1}), \, \Psi\in\mathcal{PC} _{\alpha}(d_{2}) for d_{1}, \, d_{2} > 0 and \lambda, \, \alpha\in\mathbb{R} . Suppose that
\Phi(w) = P(w)+\lambda\varphi_{1}(w)\quad\quad\quad\quad\;{ and }\; \quad\quad\quad\quad\Psi(z) = Q(z)+\alpha\varphi_{2}(z), |
where P, Q, \varphi_{1} , and \varphi_{2} are as in the definition of the spaces \mathcal{PC}_{\lambda}(d_{1}) and \Psi\in\mathcal{PC}_{\alpha}(d_{2}) . For t, \, s\in\mathbb{R} , let \sigma_{A, {t}, {s}}^{(\Phi, \Psi)} , \sigma_{A, {t}, {s}} ^{(\Phi, Q)}, \sigma_{A, {t}, {s}}^{(P, \Psi)}, and \sigma_{A, {t}, {s} }^{(P, Q)} be the measures defined by (3.1) with proper modifications. Then,
(ⅰ) \Vert\sigma_{A, {t}, {s}}^{(\Phi, \Psi)}\Vert\leq C;
(ⅱ) \begin{array} [c]{ll} \left\vert \widehat{\sigma}_{A, {t, s}}^{(\Phi, \Psi)}(\xi, \eta)\right\vert \leq C\, \, |\lambda\, \varphi_{1}\, (A^{t-1})\, \xi\, |^{-\frac{1}{2\, (d_{1}+1)\log _{2}A}}\, \, \, |\alpha\, \varphi_{2}\, (A^{s-1})\, \eta\, |^{-\frac{1} {2\, (d_{2}+1)\log_{2}A}}; & \end{array}
(ⅲ) \begin{array} [c]{ll} \left\vert \widehat{\sigma}_{A, {t, s}}^{(\Phi, \Psi)}(\xi, \eta)-\widehat{\sigma }_{A, {t, s}}^{(P, \Psi)}(\xi, \eta)\right\vert \leq C\, \, \left\vert \lambda\varphi_{1}(A^{t})\, \xi\right\vert ^{\frac{1}{2\log_{2}A} }\, \, \left\vert \alpha\varphi_{2}(A^{s-1})\eta\right\vert ^{-\frac{1} {2(d_{2}+1)\, \log_{2}A}}; & \end{array}
(ⅳ) \begin{array} [c]{ll} \left\vert \widehat{\sigma}_{A, {t, s}}^{(\Phi, \Psi)}(\xi, \eta)-\widehat{\sigma }_{A, {t, s}}^{(\Phi, Q)}(\xi, \eta)\right\vert \leq C\, \left\vert \, \lambda \, \varphi_{1}(A^{t-1})\, \xi\right\vert ^{-\frac{1}{2\, (d_{1}+1)\, \log_{2}A} }\, \, \, \left\vert \, \alpha\, \varphi_{2}(A^{s})\, \eta\right\vert ^{\frac {1}{2\log_{2}A}}; & \end{array}
(ⅴ) \begin{array} [c]{ll} \left\vert \widehat{\sigma}_{A, {t, s}}^{(\Phi, \Psi)}(\xi, \eta)-\widehat{\sigma }_{A, {t, s}}^{(P, \Psi)}(\xi, \eta)-\widehat{\sigma}_{A, {t, s}}^{(\Phi, Q)} (\xi, \eta)+\widehat{\sigma}_{A, {t, s}}^{(P, Q)}(\xi, \eta)\right\vert & \\ \leq C|\lambda\varphi_{1}(A^{t})\, \xi|^{\frac{1}{2\log_{2}A}}\, |\alpha \varphi_{2}(A^{s})\, \eta|^{\frac{1}{2\log_{2}A}}; & \end{array}
(ⅵ) \left\vert \widehat{\sigma}_{A, {t, s}}^{(\Phi, Q)} (\xi, \eta)-\widehat{\sigma}_{A, {t, s}}^{(P, Q)}(\xi, \eta)\right\vert \leq C\, |\lambda\varphi_{1}(A^{t})\, \xi|^{\frac{1}{2\log_{2}A}};
(ⅶ) \left\vert \widehat{\sigma}_{A, {t}, {s}}^{(P, \Psi)} (\xi, \eta)-\widehat{\sigma}_{A, {t}, {s}}^{(P, Q)}(\xi, \eta)\right\vert \leq C\, \left\vert \, \alpha\, \varphi_{2}(A^{s})\, \eta\right\vert ^{\frac{1} {2\log_{2}A}},
where C is independent of \kappa and (\xi, \eta)\in(\mathbb{R}^{n}, \mathbb{R}^{m}) .
We end this section by the following estimates contained in the argument in [18].
Lemma 3.3. Let \Omega\in L^{2}(\mathbb{S} ^{n-1}\times\mathbb{S}^{m-1}) satisfying (1.6)–(1.7) with \left\Vert \Omega\right\Vert _{1}\leq1 and \left\Vert \Omega\right\Vert _{2}\leq A for some A > 2. Suppose that P(w) = \sum_{k = 0}^{d_{1}}c_{k, 1}\, w^{k} and Q(z) = \sum_{k = 0}^{d_{2}}c_{k, 2}\, z^{k} are polynomials of degrees d_{1} and d_{2} , respectively. For 0\leq l\leq d_{1} and 0\leq s\leq d_{2} , let
P_{l}(w) = \sum\limits_{k = 0}^{l}c_{k, 1}\, w^{k}\quad\text{and }\quad Q_{o}(z) = \sum\limits_{k = 0}^{o}c_{k, 2}\, z^{k} |
with the convention that \sum\limits_{j\in\emptyset} = 0 . For t, \, s\in\mathbb{R} , 0\leq l\leq d_{1} , and 0\leq o\leq d_{2} , let {\sigma}\, _{A, {t}, {s}}^{(l, o)} be defined by (3.1) where \Phi and \Psi are replaced by P_{l} and Q_{o} , respectively. For 0\leq l\leq d_{1}, \, 0\leq o\leq d_{2} , let {\sigma}\, _{A, {t}, {s}}^{(l, o)} = {\sigma} \, _{A, {t}, {s}}^{(P_{l}, Q_{o})} . Then, for 1\leq l\leq d_{1} and \, 1\leq o\leq d_{2} , we have
(ⅰ) \Vert\sigma_{A, {t}, {s}}^{(l, o)}\Vert\leq C;
(ⅱ) \begin{array} [c]{ll} \left\vert \widehat{\sigma}_{A, {t}, {s}}^{(l, o)}(\xi, \eta)\right\vert \leq C\, \, |c_{l, 1}A^{l(t-1)}\, \, {l}!\, \xi|^{-\frac{1}{2\, l\, \log_{2}A}} |\, c_{o, 2}\, (A^{o(s-1)}\, {o}!\eta|^{-\frac{1}{2\, o\log_{2}A}}; & \end{array}
(ⅲ) \begin{array} [c]{ll} \left\vert \widehat{\sigma}_{A, {t}, {s}}^{(l, o)}(\xi, \eta)-\widehat{\sigma }_{A, {t}, {s}}^{(l-1, o)}(\xi, \eta)\right\vert \leq C\, \, |c_{l, 1}\, A^{lt} \xi|^{\frac{1}{2\log_{2}A}}\, \, |c_{o, 2}\, A^{o(s-1)}\, {o}!\, \, \eta|^{-\frac {1}{2\, o\log_{2}A}}; & \end{array}
(ⅳ) \begin{array} [c]{ll} \left\vert \widehat{\sigma}_{A, {t}, {s}}^{(l, o)}(\xi, \eta)-\widehat{\sigma }_{A, {t}, {s}}^{(l, o-1)}(\xi, \eta)\right\vert \leq C\, \left\vert c_{l, 1} \, A^{l(t-1)}\, l!\, \xi\right\vert ^{-\frac{1}{2\, l\, \log_{2}A}}\, \, \left\vert c_{o, 2}A^{os}\, \eta\right\vert ^{\frac{1}{2\log_{2}A}}; & \end{array}
(ⅴ) \begin{array} [c]{ll} \left\vert \widehat{\sigma}_{A, {t}, {s}}^{(l, o)}(\xi, \eta)-\widehat{\sigma }_{A, {t}, {s}}^{(l-1, o)}(\xi, \eta)-\widehat{\sigma}_{A, {t}, {s}}^{(l, o-1)} (\xi, \eta)+\widehat{\sigma}_{A, {t}, {s}}^{(l-1, o-1)}(\xi, \eta)\right\vert & \\ \leq C\, |c_{l, 1}A^{lt}\, \xi|^{\frac{1}{\kappa+1}}\, |c_{o, 2}\, A^{os} \, \eta|^{\frac{1}{2\log_{2}A}}; & \end{array}
(ⅵ) \left\vert \widehat{\sigma}_{A, {t}, {s}}^{(l, o-1)} (\xi, \eta)-\widehat{\sigma}_{A, {t}, {s}}^{(l-1, o-1)}(\xi, \eta)\right\vert \leq C\, \, |c_{l, 1}\, A^{lt}\, \xi|^{\frac{1}{2\log_{2}A}};
(ⅶ) \left\vert \widehat{\sigma}_{A, {t}, {s}}^{(l-1, o)} (\xi, \eta)-\widehat{\sigma}_{A, {t}, {s}}^{(l-1, o-1)}(\xi, \eta)\right\vert \leq C\, \left\vert c_{o, 2}A^{os}\, \eta\right\vert ^{\frac{1}{2\log_{2}A}},
where C is independent of A and (\xi, \eta)\in(\mathbb{R}^{n}, \mathbb{R}^{m}).
This section is devoted for the proofs of Theorems 1.5 and 1.7. To this end, we prove the following proposition:
Proposition 4.1. Suppose that \Omega\in L^{2}(\mathbb{S}^{n-1}\times\mathbb{S}^{m-1}) satisfying (1.6)–(1.7) with \left\Vert \Omega\right\Vert _{1}\leq1 and that \left\Vert \Omega\right\Vert _{2}\leq A for some A > 2. Suppose also that \omega_{1}\in\tilde{A}_{p}^{I} (\mathbb{R}^{n}) and \omega_{2}\in\tilde{A}_{p}^{I}(\mathbb{R} ^{m}), 1 < p < \infty . Assume that the mappings \Phi, \Psi satisfies (i) \Phi\mathit{, }\Psi\in\mathcal{G} ; or (ii) \Phi\in\mathcal{PC}_{\lambda}(d_{1}), \, \Psi\in\mathcal{PC}_{\alpha}(d_{2}) for d_{1}, \, d_{2} > 0 and \lambda, \, \alpha\in\mathbb{R} . Then, for 1 < p < \infty , we have
\begin{equation} \left\Vert \mathcal{U}_{\Omega, \Phi, \Psi}(f)\right\Vert _{L^{p}(\omega _{1}, \omega_{2})}\leq\left( \log_{2}A\right) C_{p}\left\Vert f\right\Vert _{L^{p}(\omega_{1}, \omega_{2})} \end{equation} | (4.1) |
with constants C_{p} independent of A .
Proof: We shall prove (4.1) under the assumption (ⅱ) on the mappings \Phi and \Psi . The proof under the assumption (ⅰ) follows by similar argument with minor modifications. We write \Phi and \Psi as
\begin{equation} \Phi(w) = P(w)+\lambda\varphi_{1}(w)\quad\text{and} \quad Psi(z) = Q(z)+\alpha\varphi_{2}(z), \end{equation} | (4.2) |
where P and Q are polynomials of degrees d_{1} and d_{2} as in the statement of Lemma 3.3. We let \{c_{k, 1}\}, \{c_{k, 2}\}, P_{l}, Q_{o}, and {\sigma}\, _{A, {t}, {s}}^{(l, o)} be as in Lemma 3.3. Let {\sigma} \, _{A, {t}, {s}}^{(d_{1}+1, d_{2}+1)} be the measure \sigma_{A, {t}, {s}} ^{(\Phi, \Psi)} in Lemma 3.1. By simple change of variables, we have
\begin{equation} \mathcal{U}_{\Omega, \Phi, \Psi}(f)(x, y) = \left( \log_{2}A\right) \mathcal{U}_{A, \Phi, \Psi}f(x, y) , \end{equation} | (4.3) |
where
\begin{array}{c} \mathcal{U}_{A, \Phi, \Psi}f(x, y) = \left( \int_{-\infty}^{\infty}\int_{-\infty }^{\infty}\left\vert F_{A, t, s}^{{(\Phi, \Psi)}}(f)(x, y)\right\vert ^{2}2^{-2(\log_{2}A)(t+s)}dt\, ds\right) ^{\frac{1}{2}},\\ F_{A, t, s}^{{(\Phi, \Psi)}}(f)(x, y) = \int\int_{\Lambda(A^{t}, A^{s})} f(x-\Phi(|u|)u^{\prime}, y-\Psi(|v|)v^{\prime})\, \, \dfrac{\Omega(u^{\prime }, v^{\prime})}{|u|^{n-1}\, |v|^{m-1}}du\, dv, \end{array} | (4.4) |
and
\Lambda(A^{t}, A^{s}) = \{(u, v)\in\mathbb{R}^{n}\times\mathbb{R}^{m}:|u|\leq A^{t}\, \, \text{and}\, \, |v|\leq A^{s}\}\text{.} |
Thus, to prove (4.1), it suffices to show that
\begin{equation} \left\Vert \mathcal{U}_{A, \Phi, \Psi}(f)\right\Vert _{L^{p}(\omega_{1} , \omega_{2})}\leq C_{p}\left\Vert f\right\Vert _{L^{p}(\omega_{1}, \omega_{2})} \end{equation} | (4.5) |
with constant C_{p} independent of A . Let \{\sigma_{A, {t}, {s}} ^{(l, o)}:0\leq l\leq d_{1}, \, 0\leq o\leq d_{2}\} be as in Lemma 3.3. Notice that
\begin{equation} \widehat{\sigma}_{A, {t}, {s}}^{(0, 0)} = \widehat{\sigma}_{A, {t}, {s}} ^{(0, d_{2}+1)} = \widehat{\sigma}_{A, {t}, {s}}^{(d_{1}+1, 0)} = 0. \end{equation} | (4.6) |
Following the same arguments in [18], for 1\leq l\leq d_{1}, \, 1\leq o\leq d_{2} , 1 < p < \infty, j, \, k\in\mathbb{Z} , \omega_{1}\in\tilde{A}_{p}^{I}(\mathbb{R}^{n}), and \, \omega_{2}\in \tilde{A}_{p}^{I}(\mathbb{R}^{m}) , we can find linear transformations L_{l}:\mathbb{R}^{n}\mathbb{\rightarrow R}^{n} and Q_{s}:\mathbb{R} ^{m}\mathbb{\rightarrow R}^{m} and measures \{\tau_{A, t, s}^{(l, o)} :t, \, s\in\mathbb{R}, \} such that
\begin{equation} \left\vert \widehat{\tau}_{A, t, s}^{(l, o)}(\xi, \eta)\right\vert \leq C\, \, |A^{lt}L_{l}(\xi)|^{-\frac{1}{2\, \beta_{l}\, \log_{2}A}}\, \, |A^{os} Q_{o}(\eta)|^{-\frac{1}{2\, \delta_{o}\, \log_{2}A}}; \end{equation} | (4.7) |
\begin{equation} \left\vert \widehat{\tau}_{A, t, s}^{(l, s)}(\xi, \eta)-\widehat{\tau} _{A, t, s}^{(l-1, o)}(\xi, \eta)\right\vert \leq C\, \, |A^{lt}\, L_{l}(\xi )|^{\frac{1}{2\log_{2}A}}\, \, |A^{os}Q_{o}(\eta)|^{-\frac{1}{2\, \delta _{o}\, \log_{2}A}}; \end{equation} | (4.8) |
\begin{equation} \left\vert \widehat{\tau}_{A, t, s}^{(l, s)}(\xi, \eta)-\widehat{\tau} _{A, t, s}^{(l, o-1)}(\xi, \eta)\right\vert \leq C\, \left\vert A^{lt}\, L_{l} (\xi)\right\vert ^{-\frac{1}{2\, \beta_{l}\, \log_{2}A}}\, \, \left\vert A^{os}Q_{o}(\eta)\right\vert ^{\frac{1}{2\log_{2}A}}; \end{equation} | (4.9) |
\begin{equation} \begin{array} [c]{ll} \left\vert \widehat{\tau}_{A, t, s}^{(l, s)}(\xi, \eta)-\widehat{\tau} _{A, t, s}^{(l-1, o)}(\xi, \eta)-\widehat{\tau}_{A, t, s}^{(l, s-1)}(\xi , \eta)+\widehat{\tau}_{A, t, s}^{(l-1, s-1)}(\xi, \eta)\right\vert & \\ \leq C\, |A^{lt}\, L_{l}(\xi)|^{\frac{1}{2\log_{2}A}}\, |A^{os}Q_{o} (\eta)|^{\frac{1}{2\log_{2}A}}; & \end{array} \end{equation} | (4.10) |
\begin{equation} \left\vert \widehat{\tau}_{A, t, s}^{(l, o-1)}(\xi, \eta)-\widehat{\tau} _{A, t, s}^{(l-1, o-1)}(\xi, \eta)\right\vert \leq C\, \, |A^{lt}\, L_{l} (\xi)|^{\frac{1}{2\log_{2}A}}; \end{equation} | (4.11) |
\begin{equation} \left\vert \widehat{\tau}_{A, t, s}^{(l-1, o)}(\xi, \eta)-\widehat{\tau} _{A, t, s}^{(l-1, o-1)}(\xi, \eta)\right\vert \leq C\, \left\vert A^{os}Q_{o} (\eta)\right\vert ^{\frac{1}{2\log_{2}A}}; \end{equation} | (4.12) |
\begin{equation} \left\Vert \left( \int\limits_{-\infty}^{\infty}\int\limits_{-\infty} ^{\infty}|\tau_{A, t, s}^{(l, o)}\ast\Upsilon_{a^{t+j}, \, b^{s+k}}\ast f(x, y)|^{2}dt\, ds\right) ^{\frac{1}{2}}\right\Vert _{L^{p}(\omega_{1} , \omega_{2})}\leq C\, _{p}\Vert f\Vert_{L^{p}(\omega_{1}, \omega_{2})}; \end{equation} | (4.13) |
and
\begin{equation} \sum\limits_{l = 1}^{d_{1}+1}\sum\limits_{o = 1}^{d_{2}+1}\tau_{A, t, s}^{(l, o)} = \sigma_{A, {t} , {s}}^{(d_{1}+1, d_{2}+1)}; \end{equation} | (4.14) |
where
\beta_{l} = \left\{ \begin{array} [c]{c} d_{1}+1, { \ \ \ }l = d_{1}+1;\\ l, { \ \ \ \ \ \ \ \ \ \ \ }l\neq d_{1}+1, \end{array} \right. |
and
\delta_{o} = \left\{ \begin{array} [c]{c} d_{2}+1, \quad o = d_{2}+1;\\ o, { \ \ \ \ \ \ \ \ \ \ \ }o\neq d_{2}+1. \end{array} \right. |
Thus, by (4.14) and Minkowski's inequality, we obtain that
\begin{equation} \left\Vert \mathcal{U}_{A, \Phi, \Psi}(f)\right\Vert _{L^{p}(\omega_{1} , \omega_{2})}\leq C_{p}\, \sum\limits_{l = 1}^{d_{1}+1}\sum\limits_{o = 1}^{d_{2}+1}\left\Vert S_{A, l, o}(f)\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}; \end{equation} | (4.15) |
where
S_{A, l, o}(f)(x, y) = \left( \int_{-\infty}^{\infty}\int_{-\infty}^{\infty }\left\vert \left( \tau_{A, t, s}^{(l, o)}\ast(f)(x, y)\right) \right\vert ^{2}dtds\right) ^{\frac{1}{2}}. |
Now, by a similar argument as in [27], choose two collections of \mathcal{C^{\infty}} functions \{\varpi _{i}^{(l)}\}_{i\in\mathbb{Z}} and \{\varpi_{i}^{(o)}\}_{i\in\mathbb{Z}} on (0, \infty) satisfying the following properties:
\begin{equation} supp(\varpi_{i}^{(l)})\subseteq\left[ A^{-l(i+1)}, A^{-l(i-1)}\right] \, \, \, \text{and}\, \, \, supp(\varpi_{i}^{(o)})\subseteq\left[ A^{-o(i+1)} , A^{-o(i-1)}\right]; \end{equation} | (4.16) |
\begin{equation} 0\leq\varpi_{i}^{(l)}, \varpi_{i}^{(o)}\leq1; \end{equation} | (4.17) |
\begin{equation} \sum\limits_{i\in\mathbb{Z}}\varpi_{i}^{(l)})(u) = \sum\limits_{i\in\mathbb{Z}}\varpi _{i}^{(o)})(u) = 1; \end{equation} | (4.18) |
\begin{equation} \left\vert \dfrac{d^{r}\varpi_{i}^{(l)}}{du^{r}}(u)\right\vert , \, \, \, \left\vert \dfrac{d^{r}\varpi_{i}^{(o)}}{du^{r}}(u)\right\vert \leq\dfrac{C_{r}}{u^{r}}, \end{equation} | (4.19) |
where C_{r} is independent of A . Define the measures \{\upsilon_{i} ^{(l)}:i\in\mathbb{Z}\} on \mathbb{R}^{n} and \{\upsilon_{i}^{(o)} :i\in\mathbb{Z}\} on \mathbb{R}^{m} by
\widehat{(}\upsilon_{i}^{(l)})(x) = \varpi_{i}^{(l)}(|x|^{2})\, \, \, \text{and} \, \, \, \widehat{(}\upsilon_{i}^{(o)})(y) = \varpi_{i}^{(o)}(|y|^{2}). |
By (4.18), we immediately obtain
\begin{equation} \begin{array} [c]{ll} \widehat{(\tau_{A, t, s}^{(l, o)}\ast f)(\xi, \eta)} & = \widehat{\tau} _{A, t, s}^{(l, o)}(\xi, \eta)\, .\, \widehat{f}(\xi, \eta)\sum\limits_{j\in \mathbb{Z}}\widehat{\upsilon}_{j}^{(l)}(\xi)\, .\, \sum\limits_{i\in\mathbb{Z} }\widehat{\upsilon}_{i}^{(o)}(\eta)\\ & = \widehat{\tau}_{A, t, s}^{(l, o)}(\xi, \eta)\, .\, \widehat{f}(\xi, \eta )\sum\limits_{j\in\mathbb{Z}}\widehat{\upsilon}_{\lfloor t\rfloor+j}^{(l)} (\xi)\, .\, \sum\limits_{i\in\mathbb{Z}}\widehat{\upsilon}_{\lfloor s\rfloor +i}^{(o)}(\eta), \end{array} \end{equation} | (4.20) |
where \lfloor t\rfloor is the greatest integer function such that t-1 < \lfloor t\rfloor < t , and similarly for \lfloor s\rfloor (see [6,20]). Hence, by taking the inverse Fourier transform for (4.20), we get
\begin{equation} (\tau_{A, t, s}^{(l, o)}\ast f)(x, y) = \sum\limits_{j\in\mathbb{Z}}\, \sum \limits_{i\in\mathbb{Z}}\left( \upsilon_{\lfloor t\rfloor+j}^{(l)} \otimes\upsilon_{\lfloor s\rfloor+i}^{(o)}\right) \ast\tau_{A, t, s} ^{(l, o)}\ast f(x, y). \end{equation} | (4.21) |
Thus, by (4.21), we obtain
\begin{equation} S_{A, l, o}(f)(x, y)\leq C\, \sum\limits_{j\in\mathbb{Z}}\sum\limits_{i\in \mathbb{Z}}I_{A, i, j}^{(l, o)}(f)(x, y) \end{equation} | (4.22) |
where
\begin{equation} I_{A, i, j}^{(l, o)}(f)(x, y) = \left( \int_{-\infty}^{\infty}\int_{-\infty }^{\infty}\left\vert \left( \upsilon_{\lfloor t\rfloor+j}^{(l)} \otimes\upsilon_{\lfloor s\rfloor+i}^{(s)}\right) \ast\tau_{A, t, s} ^{(l, o)}\ast f(x, y)\right\vert ^{2}dt\, ds\right) ^{\frac{1}{2}}. \end{equation} | (4.23) |
By (4.7)–(4.12) and the Plancherel theorem, we get
\begin{equation} \Vert I_{A, i, j}^{(l, o)}(f)\Vert_{2}\leq\Theta_{i, j}\, \Vert f\Vert_{2}, \end{equation} | (4.24) |
where
\Theta_{i, j} = \begin{cases} 2^{\frac{is+jl}{l+j}}, \quad \text{if}\, \, i, j\leq-2;\\ 2^{-i-j}, \quad \text{if}\, \, i, j\geq3;\\ 2^{\frac{i-jl}{l}}, \quad \ \text{if}\, \, i\leq-2\, \, \text{and} \, \, j\geq3;\\ 2^{\frac{-is+j}{s}}, \quad \text{if}\, \, i\geq3\, \, \text{and} \, \, j\leq-2;\\ 1, \quad \quad \ \ \text{if}\, \, i\geq-2\, \, \text{and}\, \, j\leq3. \end{cases} |
Next, by (4.13), for 1 < p < \infty and \omega_{1}\in\tilde{A} _{p}^{I}(\mathbb{R}^{n}), \, \omega_{2}\in\tilde{A}_{p}^{I}(\mathbb{R}^{m}) , there exists a positive constant C_{p} independent of i, j , and A such that
\begin{equation} \Vert I_{A, i, j}^{(l, o)}(f)\Vert_{L^{p}(\omega_{1}, \omega_{2})}\leq C_{p}\, \Vert f\Vert_{L^{p}(\omega_{1}, \omega_{2})}. \end{equation} | (4.25) |
Now, we have three cases:
Case 1. p > 2 . Choose a q > p and \varepsilon > 0 , such that \omega_{1}^{1+\varepsilon}\in\tilde{A}_{p}^{I}(\mathbb{R}^{n})\subset \tilde{A}_{q}^{I}(\mathbb{R}^{n}) and \omega_{2}^{1+\varepsilon}\in\tilde {A}_{p}^{I}(\mathbb{R}^{m})\subset\tilde{A}_{q}^{I}(\mathbb{R}^{m}) . Thus, by (4.25) we get
\begin{equation} \Vert I_{A, i, j}^{(l, o)}(f)\Vert_{L^{q}(\omega_{1}^{1+\varepsilon}, \omega _{2}^{1+\varepsilon})}\leq C_{p}\, \Vert f\Vert_{L^{q}(\omega_{1} ^{1+\varepsilon}, \omega_{2}^{1+\varepsilon})}, \end{equation} | (4.26) |
which when combined with (4.24) and the interpolation theorem with change of measures, we have
\begin{equation} \Vert I_{A, i, j}^{(l, o)}(f)\Vert_{L^{p}(\omega_{1}, \omega_{2})}\leq C^{1-\gamma}\, \Theta_{i, j}^{\gamma}\Vert f\Vert_{L^{p}(\omega_{1}, \omega_{2})} \end{equation} | (4.27) |
for 0 < \gamma < 1 and p > 2 .
Case 2. 1 < p < 2 . Choose a 1 < q < p and \varepsilon > 0 such that \omega_{1}\in\tilde{A}_{p}^{I}(\mathbb{R}^{n}), \, \omega_{2}\in\tilde{A} _{p}^{I}(\mathbb{R}^{m}) and \omega_{1}^{1+\varepsilon}\in\tilde{A}_{q} ^{I}(\mathbb{R}^{n}), \, \omega_{2}^{1+\varepsilon}\in\tilde{A}_{q} ^{I}(\mathbb{R}^{m}) . Thus, by (4.25) we get
\begin{equation} \Vert I_{A, i, j}^{(l, o)}(f)\Vert_{L^{q}(\omega_{1}^{1+\varepsilon}, \omega _{2}^{1+\varepsilon})}\leq C_{p}\, \Vert f\Vert_{L^{q}(\omega_{1} ^{1+\varepsilon}, \omega_{2}^{1+\varepsilon})} \end{equation} | (4.28) |
for some positive constant C_{p} independent of A . Then, by the same argument as in Case 1 , we obtain (4.27) for 0 < \gamma < 1 and 1 < p < 2 .
Case 3. p = 2 . We choose \varepsilon > 0 such that \omega _{1}^{1+\varepsilon}\in\tilde{A}_{2}^{I}(\mathbb{R}^{n}), \omega_{2} ^{1+\varepsilon}\in\tilde{A}_{2}^{I}(\mathbb{R}^{n}) . Then, we follow a similar argument as in the previous two cases and get
\begin{equation} \Vert I_{A, i, j}^{(l, o)}(f)\Vert_{L^{p}(\omega_{1}, \omega_{2})}\leq C^{1-\gamma}\, \Theta_{i, j}^{\gamma}\Vert f\Vert_{L^{p}(\omega_{1}, \omega_{2})} \end{equation} | (4.29) |
for 0 < \gamma < 1 and p = 2 .
Finally, by (4.15), (4.22), and (4.27)–(4.29), we get (4.5). This completes the proof of Proposition 4.1.
Proof (of Theorem 1.5): Assume that \Omega\in L(\log L)(\mathbb{S}^{n-1}\times\mathbb{S}^{m-1}) . We write \Omega as
\begin{equation} \Omega(x, y) = \sum\limits_{{k = 0}}^{\infty}\theta_{k}\Omega_{k}(x, y), \end{equation} | (4.30) |
where \Omega_{k} satisfies (1.6)–(1.7), \Vert\Omega_{\kappa }\Vert_{1}\leq4 , \, \Vert\Omega_{k}\Vert_{2}\leq2^{2(k+1)} , and the estimate
\begin{equation} \sum\limits_{{k = 0}}^{\infty}(k+1)\, \theta_{\kappa}\leq\Vert\Omega\Vert_{L(\log L)(\mathbb{S}^{n-1}\times\mathbb{S}^{m-1})}. \end{equation} | (4.31) |
By (4.30) and Minkowski's inequality, we have
\left\Vert \mathcal{U}_{A, \Phi, \Psi}(f)\right\Vert _{L^{p}(\omega_{1} , \omega_{2})}\leq\sum\limits_{{k = 0}}^{\infty}\theta_{k}\left\Vert \mathcal{U} _{2^{2(k+1)}, \Phi, \Psi}(f)\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}. |
Thus, by Proposition 4.1 with A = 2^{2(k+1)} , we have
\begin{align*} \left\Vert \mathcal{U}_{A, \Phi, \Psi}(f)\right\Vert _{L^{p}(\omega_{1} , \omega_{2})} & \leq\sum\limits_{{k = 0}}^{\infty}\log_{2}(2^{2(k+1)})\theta _{k}\left\Vert f\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}\\ & = \left( \sum\limits_{{k = 0}}^{\infty}2(k+1)\theta_{k}\right) \left\Vert f\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}\\ & \leq2\Vert\Omega\Vert_{L(\log L)(\mathbb{S}^{n-1}\times\mathbb{S}^{m-1} )}\left\Vert f\right\Vert _{L^{p}(\omega_{1}, \omega_{2})}\text{.} \end{align*} |
This completes the proof.
Proof (of Theorem 1.7): The proof follows a similar argument as in the proof of Theorem 1.5. We omit the details.
In this paper, we proved the weighted L^{p} boundedness of Marcinkiewicz integral operators along surfaces. We considered surfaces that are determined by functions satisfying some growth conditions or mappings that are more general than polynomials and convex functions. We proved the weighted L^{p} boundedness of related square functions and maximal functions. The argument in this paper can be used to treat more general integral operators. This shall be the topic of future research.
The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.
Authors cordially thank the reviewers for their useful comments on the manuscript. The authors would like to thank Sultan Qaboos University for paying the APC.
The authors declare no conflicts of interest.
[1] |
E. M. Stein, On the function of Littlewood-Paley, Lusin and Marcinkiewicz, T. Am. Math. Soc., 88 (1958), 430–466. https://doi.org/10.2307/1993226 doi: 10.2307/1993226
![]() |
[2] |
A. Benedek, A. Calderon, R. Panzone, Convolution operators on Banach space valued functions, P. Natl. Acad. Sci. U. S. A., 48 (1962), 356–365. https://doi.org/10.1073/pnas.48.3.356 doi: 10.1073/pnas.48.3.356
![]() |
[3] |
T. Walsh, On the function of Marcinkiewicz, Stud. Math., 44 (1972), 203–217. https://doi.org/10.1090/S0002-9947-1958-0112932-2 doi: 10.1090/S0002-9947-1958-0112932-2
![]() |
[4] |
H. Al-Qassem, A. Al-Salman, L. Cheng, Y. Pan, L^{p} bounds for the functions of Marcinkiewicz, Math. Res. Lett., 9 (2002), 697–700. https://doi.org/10.4310/MRL.2002.v9.n5.a11 doi: 10.4310/MRL.2002.v9.n5.a11
![]() |
[5] |
A. Al-Salman, Marcinkiewicz functions with Hardy space kernels, Math. Inequal. Appl., 21 (2018), 553–567. https://doi.org/10.7153/mia-2018-21-40 doi: 10.7153/mia-2018-21-40
![]() |
[6] | A. Al-Salman, On Marcinkiewicz integrals along flat surfaces, Turk. J. Math., 29 (2005), 111–120. |
[7] | J. Chen, D. Fan, Y. Ying, Rough Marcinkiewicz integrals with L(\log L)^{2} kernels, Adv. Math. (China), 30 (2001), 179–181. |
[8] |
Y. Ding, L^{2} boundedness of Marcinkiewicz integral with rough kernel, Hokkaido Math. J., 27 (1998), 105–115. https://doi.org/10.14492/hokmj/1351001253 doi: 10.14492/hokmj/1351001253
![]() |
[9] |
S. Sato, Estimates for Littlewood-Paley functions and extrapolation, Integr. Equ. Oper. Theory, 62 (2008), 429–440. https://doi.org/10.1007/s00020-008-1631-4 doi: 10.1007/s00020-008-1631-4
![]() |
[10] |
A. Torchinsky, S. Wang, A note on the Marcinkiewicz integral, Colloq. Math., 60 (1990), 235–243. https://doi.org/10.4064/cm-60-61-1-235-243 doi: 10.4064/cm-60-61-1-235-243
![]() |
[11] | J. Garcia-Cuerva, J. L. R. de Francia, Weighted norm inequalities and related topics, Amsterdam: Elsevier, 2011. |
[12] |
Y. Ding, D. Fan, Y. Pan, Weighted boundedness for a class of rough Marcinkiewicz integrals, Indiana Univ. Math. J., 48 (1999), 1037–1056. https://doi.org/10.1512/iumj.1999.48.1696 doi: 10.1512/iumj.1999.48.1696
![]() |
[13] |
M. Y. Lee, C. C. Lin, Weighted L^{p} boundedness of Marcinkiewicz integral, Integr. Equ. Oper. Theory, 49 (2004), 211–220. https://doi.org/10.1007/s00020-002-1204-x doi: 10.1007/s00020-002-1204-x
![]() |
[14] |
J. Duoandikoetxea, Weighted norm inequalities for homogeneous singular integrals, T. Am. Math. Soc., 336 (1993), 869–880. https://doi.org/10.2307/2154381 doi: 10.2307/2154381
![]() |
[15] | A. Al-Salman, Certain weighted L^{p} bounds for the functions of marcinkiewicz, SE Asian B. Math., 30 (2006), 609–620. |
[16] |
Y. Choi, Marcinkiewicz integrals with rough homogeneous kernels of degree zero in product domains, J. Math. Anal. Appl., 261 (2001), 53–60. https://doi.org/10.1006/jmaa.2001.7465 doi: 10.1006/jmaa.2001.7465
![]() |
[17] |
H. Al-Qassem, A. Al-Salman, L. C. Cheng, Y. Pan, Marcinkiewicz integrals on product spaces, Stud. Math., 167 (2005), 227–234. https://doi.org/10.4064/sm167-3-4 doi: 10.4064/sm167-3-4
![]() |
[18] |
B. Al-Azriyah, A. Al-Salman, Singular and marcinkiewicz integral operators on product domains, Commun. Korean Math. S., 38 (2023), 401–430. https://doi.org/10.4134/CKMS.c210421 doi: 10.4134/CKMS.c210421
![]() |
[19] |
B. Al-Azriyah, A. Al-Salman, A note on marcinkiewicz integral operators on product domains, Kyungpook Math. J., 63 (2023), 577–591. https://doi.org/10.5666/KMJ.2023.63.4.577 doi: 10.5666/KMJ.2023.63.4.577
![]() |
[20] |
A. Al-Salman, Rough Marcinkiewicz integrals on product spaces, Int. Math. Forum, 2 (2007), 1119–1128. https://doi.org/10.12988/imf.2007.07097 doi: 10.12988/imf.2007.07097
![]() |
[21] |
A. Al-Salman, Marcinkiewicz integrals along subvarieties on product domains, International Journal of Mathematics and Mathematical Sciences, 2004 (2004), 730308. https://doi.org/10.1155/S0161171204401264 doi: 10.1155/S0161171204401264
![]() |
[22] |
J. Chen, D. Fan, Y. Ying, The method of rotation and Marcinkiewicz integrals on product domains, Stud. Math., 153 (2002), 41–58. https://doi.org/10.4064/sm153-1-4 doi: 10.4064/sm153-1-4
![]() |
[23] |
H. Al-Qassem, Y. Pan, L^{p} boundedness for singular integrals with rough kernels on product domains, Hokkaido Math. J., 31 (2002), 555–613. https://doi.org/10.14492/hokmj/1350911903 doi: 10.14492/hokmj/1350911903
![]() |
[24] |
D. Fan, Y. Pan, D. Yang, A weighted norm inequality for rough singular integrals, Tohoku Math. J., 51 (1999), 141–161. https://doi.org/10.2748/tmj/1178224808 doi: 10.2748/tmj/1178224808
![]() |
[25] | E. M. Stein, Harmonic analysis: Real-variable mathods, orthogonality and oscillatory integrals, Princeton: Princeton University Press, 1993. |
[26] |
R. Fefferman, E. M. Stein, Singular integrals on product spaces, Adv. Math., 45 (1982), 117–143. https://doi.org/10.1016/S0001-8708(82)80001-7 doi: 10.1016/S0001-8708(82)80001-7
![]() |
[27] |
A. Al-Salman, Y. Pan, Singular integrals with rough kernels in Llog^{+}L(\mathbb{S}^{n-1}), J. Lond. Math. Soc., 66 (2002), 153–174. https://doi.org/10.1112/S0024610702003241 doi: 10.1112/S0024610702003241
![]() |