Open Access

Boundedness of Marcinkiewicz integrals with mixed homogeneity along compound surfaces

Journal of Inequalities and Applications20142014:265

https://doi.org/10.1186/1029-242X-2014-265

Received: 23 February 2014

Accepted: 30 June 2014

Published: 22 July 2014

Abstract

In this note we establish the L p boundedness of Marcinkiewicz integrals with mixed homogeneity along compound surfaces, which improve and extend some previous results. The main ingredient is to present a systematic treatment with several singular integral operators.

MSC:42B20, 42B15, 42B25.

Keywords

Marcinkiewicz integralsrough kernelsmixed homogeneityLittlewood-Paley theory

1 Introduction

Let R n , n 2 , be the n-dimensional Euclidean space and S n 1 denote the unit sphere in R n equipped with the induced Lebesgue measure . Let α j 1 ( j = 1 , , n ) be fixed real numbers. Define the function F : R n × ( 0 , ) R by F ( x , ρ ) = j = 1 n x j 2 ρ 2 α j , x = ( x 1 , x 2 , , x n ) . It is clear that, for each fixed x R n , the function F ( x , ρ ) is a decreasing function in ρ > 0 . We let ρ ( x ) denote the unique solution of the equation F ( x , ρ ) = 1 . Fabes and Rivière [1] showed that ( R n , ρ ) is a metric space, which is often called the mixed homogeneity space related to { α j } j = 1 n . For λ > 0 , we let A λ be the diagonal n × n matrix A λ = diag { λ α 1 , , λ α n } . Let R + : = ( 0 , ) and φ : R + R + , we denote A φ ( ρ ( y ) ) y by A φ ( y ) for y R n , where y = A ρ ( y ) 1 y S n 1 .

The change of variables related to the spaces ( R n , ρ ) is given by the transformation
x 1 = ρ α 1 cos θ 1 cos θ n 2 cos θ n 1 , x 2 = ρ α 2 cos θ 1 cos θ n 2 sin θ n 1 , , x n 1 = ρ α n 1 cos θ 1 sin θ 2 , x n = ρ α n sin θ 1 .
Thus d x = ρ α 1 J ( x ) d ρ d σ ( x ) , where ρ α 1 J ( x ) is the Jacobian of the above transform and α = j = 1 n α j , J ( x ) = j = 1 n α j ( x j ) 2 . Obviously, J ( x ) C ( S n 1 ) and there exists M > 0 such that
1 J ( x ) M , x S n 1 .
It is easy to see that
ρ ( x ) = | x | , if  α 1 = α 2 = = α n = 1 .
Let Ω be integrable on S n 1 and satisfy
S n 1 Ω ( u ) J ( u ) d σ ( u ) = 0 ,
(1.1)
Ω ( A s x ) = Ω ( x ) , s > 0  and  x R n .
(1.2)
For d 2 and a suitable function Φ : R n R d , we define the parabolic Marcinkiewicz integral operator M Ω Φ on R d by
M Ω Φ ( f ) ( x ) = ( 0 | 1 t ρ ( y ) t Ω ( y ) ρ ( y ) α 1 f ( x Φ ( y ) ) d y | 2 d t t ) 1 / 2 .
(1.3)
When α 1 = = α n = 1 , we denote M Ω Φ by μ Ω Φ . Clearly, if n = d and Φ ( y ) = y , the operator μ Ω Φ reduces to the classical Marcinkiewicz integral operator denoted by μ Ω , which was introduced by Stein [2] and investigated by many authors (see [39] for example). In particular, Ding et al. [5] proved that if Ω H 1 ( S n 1 ) , then μ Ω is bounded on L p ( R n ) for 1 < p < . Subsequently, Chen et al. [4] showed that μ Ω is bounded on L p ( R n ) for 2 β / ( 2 β 1 ) < p < 2 β if Ω F β ( S n 1 ) for some β > 1 . Here
F β ( S n 1 ) : = { Ω L 1 ( S n 1 ) : sup ξ S n 1 S n 1 | Ω ( y ) | ( log 1 | ξ y | ) β d σ ( y ) < } , β > 0 .
(1.4)
The functions class F β ( S n 1 ) was introduced by Grafakos and Stefanov [10] in the study of L p boundedness of singular integral operator with rough kernels. It follows from [10] that F β 1 ( S n 1 ) ⊆̷ F β 2 ( S n 1 ) for 0 < β 2 < β 1 , and q > 1 L q ( S n 1 ) ⊆̷ F β ( S n 1 ) for any β > 0 . Moreover,
β > 1 F β ( S n 1 ) H 1 ( S n 1 ) β > 1 F β ( S n 1 )
and
β > 1 F β ( S n 1 ) L log + L ( S n 1 ) .

Later on, Al-Salman et al. [11] proved that μ Ω is bounded on L p ( R n ) for 1 < p < provided that Ω L ( log + L ) 1 / 2 ( S n 1 ) . It is well known that L ( log + L ) 1 / 2 ( S n 1 ) and H 1 ( S n 1 ) do not contain each other. When n = d and Φ ( y ) = P ( | y | ) y with P ( y ) being a real polynomial on satisfying P ( 0 ) = 0 , Wu [12] proved that μ Ω Φ is bounded on L p ( R n ) for 1 + 1 / ( 2 β ) < p < 1 + 2 β provided that Ω F β ( S n 1 ) for some β > 1 / 2 . The L p boundedness for the Marcinkiewicz integral operator associated to polynomial mappings has also been obtained (see [6, 13]).

When α j 1 ( j = 1 , , n ), n = d and Φ ( y ) = y , we denote M Ω Φ by M Ω . In 2008, Ding et al. [14] proved that M Ω is bounded on L p ( R n ) for 1 < p < , provided that Ω L q ( S n 1 ) for fixed q > 1 . Chen and Ding [15] extended the above result to the case Ω L ( log L ) 1 / 2 ( S n 1 ) . Later on, Chen and Lu [16] proved that M Ω is bounded on L p ( R n ) for 2 β / ( 2 β 1 ) < p < 2 β , provided that Ω F β ( S n 1 ) for some β > 1 . This result was recently refined by Liu and Wu [17], who extended the range of β to the case β > 1 / 2 and the range of p to the case 1 + 1 / ( 2 β ) < p < 1 + 2 β . When n = d and Φ ( y ) = A φ ( y ) , Al-Salman [18] obtained the following result.

Theorem A Let n = d and Φ ( y ) = A φ ( y ) . Suppose that Ω F β ( S n 1 ) for some β > 1 with satisfying (1.1)-(1.2).
  1. (i)

    If φ ( t ) = P ( t ) with P being a real polynomial on , then M Ω Φ are bounded on L p ( R n ) for 2 β / ( 2 β 1 ) < p < 2 β . The bounds are independent of the coefficients of P.

     
  2. (ii)
    If φ F , then M Ω Φ are bounded on L p ( R n ) for 2 β / ( 2 β 1 ) < p < 2 β . Here F is the set of all functions ϕ which satisfy:
    1. (a)

      ϕ : R + R + is continuous increasing C 1 function satisfying that ϕ is monotonous;

       
    2. (b)

      there exist constant C ϕ and c ϕ such that t ϕ ( t ) C ϕ ϕ ( t ) and ϕ ( 2 t ) c ϕ ϕ ( t ) for all t > 0 .

       
     

Remark 1.1 There are some model examples in the class F , such as t α ( α > 0 ), t α ( ln ( 1 + t ) ) β ( α , β > 0 ), t ln ln ( e + t ) , real-valued polynomials P on with positive coefficients and P ( 0 ) = 0 and so on. For φ F , there exists a constant B φ > 1 such that φ ( 2 t ) B φ φ ( t ) (see [19]).

It is natural to ask whether Theorem A also holds if the range of β is relaxed to β > 1 / 2 and the range of p is relaxed to 1 + 1 / ( 2 β ) < p < 1 + 2 β . In this paper, we will give an affirmative answer to this question. Our main results can be stated as follows.

Theorem 1.1 Let n = d and Φ ( y ) = ( P 1 ( φ ( ρ ( y ) ) ) y 1 , , P n ( φ ( ρ ( y ) ) ) y n ) with P i ( t ) being real valued polynomials on satisfying P i ( 0 ) = 0 and φ F . Suppose that Ω F β ( S n 1 ) for some β > 1 / 2 satisfying (1.1)-(1.2). Then M Ω Φ are bounded on L p ( R n ) for 1 + 1 / ( 2 β ) < p < 1 + 2 β . The bounds are independent of the coefficients of P j for all 1 j n but depend on max 1 j d deg ( P j ) and φ.

Theorem 1.2 Let n = d and Φ ( y ) = A P N ( φ ) ( y ) with φ F and P N ( t ) = i = 1 N a i t i and P N ( t ) > 0 if t 0 . Suppose that Ω F β ( S n 1 ) for some β > 1 / 2 satisfying (1.1)-(1.2). Then M Ω Φ are bounded on L p ( R n ) for 1 + 1 / ( 2 β ) < p < 1 + 2 β . The bounds are independent of the coefficients of P N but depend on N and φ.

Remark 1.2 It is clear that Theorem 1.1 implies Theorem 1.2. When α 1 = = α n = 1 , φ ( t ) = t and P 1 ( t ) = = P n ( t ) = i = 1 N a i t i , Theorem 1.1 implies the result of [12]. In fact, Theorem 1.2 with φ ( t ) = t extends the result of [12] to the mixed case. Comparing Theorem A with Theorem 1.2, the range of β is extended to the case β > 1 / 2 and the range of p is enlarged to the case 1 + 1 / ( 2 β ) < p < 1 + 2 β . Thus Theorem 1.2 essentially improves and generalizes the corresponding results in Theorem A. In addition, Theorem 1.2 implies the result [[17], Theorem 1.3] when P N ( t ) = φ ( t ) = t .

When n = 2 , we have the following result.

Theorem 1.3 Let ϕ = ( ϕ 1 , , ϕ d ) be real analytic on S 1 . Let Φ ( y ) = P N ( φ ( ρ ( y ) ) ) ϕ ( y ) = ( P N ( φ ( ρ ( y ) ) ) ϕ 1 ( y ) , , P N ( φ ( ρ ( y ) ) ) ϕ d ( y ) ) with P N ( t ) = i = 1 N a i t i and φ F . Suppose that Ω F β ( S 1 ) for some β > 1 satisfying (1.1)-(1.2). Then M Ω Φ are bounded on L p ( R d ) for 1 + 1 / ( 2 β ) < p < 1 + 2 β . The bounds are independent of the coefficients of P N but depend on φ and N.

We remark that when α 1 = = α n = 1 and φ ( t ) = t , the surface { Φ ( y ) : y R n } given as in Theorem 1.3 recovers { ( P N ( | y | ) ϕ 1 ( y ) , , P N ( | y | ) ϕ d ( y ) ) ; y R n } , which was originally introduced by Al-Balushi and Al-Salman [20] in the study of L p bounds of singular integrals associated to certain surfaces.

The third type of surfaces we consider are polynomial compound subvarieties. To state the rest of our result, we need to recall some notations. Let A ( n , m ) be the set of polynomials on R n which have real coefficients and degrees not exceeding m, and let V ( n , m ) be the collection of polynomials in A ( n , m ) which are homogeneous of degree m. For P A ( n , m ) , we set
P = | λ | m a λ y λ = ( | λ | m | a λ | 2 ) 1 / 2 .

Definition 1.1 ([21])

Let n 2 , m N and β > 0 . An integrable function Ω on S n 1 is said to be in the space F ( n , m , β ) if
sup P V ( n , m ) , P = 1 S n 1 | Ω ( y ) | ( log + 1 | P ( y ) | ) β d σ ( y ) < .
(1.5)
It should be pointed out that the condition (1.5) was introduced by Al-Salman and Pan [21] (also see [22]) in a study of the L p boundedness of singular integrals with rough kernels. It is easy to check that F ( n , 1 , β ) = F β ( S n 1 ) . Moreover, it was shown in [21] that
F β ( S 1 ) = m = 1 F ( 2 , m , β ) .
(1.6)

The rest of the results can be stated as follows.

Theorem 1.4 Let P = ( P 1 , , P d ) with P j : R n R being a polynomial for 1 j d . Let Φ ( y ) = P ( φ ( ρ ( y ) ) y ) and φ F . Suppose that Ω satisfies (1.1)-(1.2) and Ω s = 1 F ( n , s , β ) for some β > 1 / 2 . Then M Ω Φ are bounded on L p ( R d ) for 1 + 1 / ( 2 β ) < p < 1 + 2 β . The bounds are independent of the coefficients of P j for all 1 j d but depend on max 1 j d deg ( P j ) and φ.

Theorem 1.5 Let P = ( P 1 , , P d ) with P j : R 2 R being a polynomial for 1 j d . Let Φ ( y ) = P ( φ ( ρ ( y ) ) y ) and φ F . Suppose that Ω satisfies (1.1)-(1.2) and Ω F β ( S 1 ) for some β > 1 / 2 . Then M Ω Φ are bounded on L p ( R d ) for 1 + 1 / ( 2 β ) < p < 1 + 2 β . The bounds are independent of the coefficients of P j for all 1 j d but depend on max 1 j d deg ( P j ) and φ.

Remark 1.3 When α j = 1 ( j = 1 , , n ), Theorem 1.4 implies Theorem 1.2 with ρ = 1 in [13]. Obviously, Theorem 1.5 follows from Theorem 1.4 because of (1.6).

The rest of this paper is organized as follows. After recalling some preliminary notations and lemmas in Section 2, we will prove our results in Section 3. We would like to remark that the main methods employed in this paper is a combination of ideas and arguments from [12, 21, 23]. The main ingredient in our proofs is to give a systematic treatment with these operators mentioned above.

Throughout this paper, we let p satisfy 1 / p + 1 / p = 1 . The letter C, sometimes with additional parameters, will stand for positive constants, not necessarily the same one at each occurrence, but independent of the essential variables.

2 Preliminaries

Lemma 2.1 Let { σ j , t } be a family of measures. Suppose that
sup j Z sup t > 0 | | σ j , t | g | p C g p
holds for some p > 1 and g L p ( R n ) . Then there exists a constant C > 0 such that
( 1 2 j Z | σ j , t g j | 2 d t ) 1 / 2 p C ( j Z | g j | 2 ) 1 / 2 p

for arbitrary functions { g j } j Z L p ( 2 , R n ) .

Proof By the assumption, we have
sup j Z sup t [ 1 , 2 ] | σ j , t g j | p sup j Z sup t > 0 | σ j , t | sup j Z | g j | p C sup j Z | g j | p .
On the other hand, by the dual argument, there exists a function h L p ( R n ) satisfying h p = 1 such that
1 2 j Z | σ j , t g j | d t p = R n j Z 1 2 | σ j , t g j ( x ) | d t h ( x ) d x R n j Z 1 2 | σ j , t | | g j ( x ) | d t | h ( x ) | d x R n j Z | g j ( x ) | sup j Z sup t [ 1 , 2 ] | σ j , t | | h ¯ | ( x ) d x j Z | g j | p sup j Z sup t > 0 | σ j , t | | h ¯ | ( x ) p C j Z | g j | p ,

where h ¯ ( x ) = h ( x ) . Thus, Lemma 2.1 follows from the standard interpolation arguments. □

Let { a k } k Z be a sequence of real positive numbers with satisfying inf k Z a k + 1 / a k = a > 1 . Let { λ k } k Z be a collection of C ( 0 , ) functions satisfying the following conditions:
supp ( λ k ) [ a k + 1 1 , a k 1 1 ] ; 0 λ k 1 ; k Z λ k 2 ( t ) = 1 ; | d λ k ( t ) / d t | C / t ,
where C is independent of t and k. Let M N { 0 } and L : R n R M be a linear transformation. For each k Z , we define the multiplier operators S k in R n by
S k f ˆ ( ξ ) = λ k ( | L ( ξ ) | ) f ˆ ( ξ ) .
(2.1)

By an argument which is similar to those used in [[8], Proposition 3.1], one can easily get the following lemma. The details are omitted here.

Lemma 2.2 Let S k be as in (2.1) and { g j , k , t } arbitrary functions on R n . Then
  1. (i)
    for each fixed 1 < p < 2 and 1 < q < p ,
    ( j Z 1 2 | k Z S j + k g j , k , t | 2 d t ) 1 / 2 p q C k Z ( j Z 1 2 | g j , k , t | 2 d t ) 1 / 2 p q ;
    (2.2)
     
  2. (ii)
    for each fixed 2 < p < and 1 < q < p ,
    ( j Z 1 2 | k Z S j + k g j , k , t | 2 d t ) 1 / 2 p q C k Z ( 1 2 ( j Z | g j , k , t | 2 ) 1 / 2 p 2 d t ) q / 2 .
    (2.3)
     

The following lemma is our main ingredient in the proof of our main results.

Lemma 2.3 Let { τ k , t : k Z , t R + } be a family of uniformly bounded Borel measures on R n . Let { a k } k Z be a sequence of real numbers with satisfying inf k Z a k + 1 / a k = a > 1 . Let M N { 0 } and L : R n R M be a linear transformation. Suppose that
(i) | τ k , t ˆ ( ξ ) | C min { 1 , a k | L ( ξ ) | } ;
(2.4)
(ii) | τ k , t ˆ ( ξ ) | C ( log | a k L ( ξ ) | ) β for some  β > 0 ,  if  a k | L ( ξ ) | > 1 ;
(2.5)
(iii) sup k Z sup t > 0 | | τ k , t | f | q C f q
(2.6)
for all 1 < q < . Then for p ( 1 + 1 / ( 2 β ) , 1 + 2 β ) and β > 1 / 2 , there exists a constant C ( a ) > 0 such that
( 1 2 k Z | τ k , t f | 2 d t ) 1 / 2 p C ( a ) f p .
Proof Let S k be as in (2.1). Then we can write
G ( f ) ( x ) : = ( 1 2 k Z | τ k , t f ( x ) | 2 d t ) 1 / 2 = ( 1 2 k Z | τ k , t ( j Z S j + k S j + k f ) ( x ) | 2 d t ) 1 / 2 = ( k Z 1 2 | j Z S j + k ( τ k , t S j + k f ) ( x ) | 2 d t ) 1 / 2 .
(2.7)
Case 1. 1 + 1 / ( 2 β ) < p < 2 . It follows from (2.2) and (2.7) that
G ( f ) p q C j Z ( k Z 1 2 | τ k , t S j + k f | 2 d t ) 1 / 2 p q , 1 < q < p .
(2.8)
For each fixed k Z , we set
I j f ( x ) : = ( k Z 1 2 | τ k , t S j + k f ( x ) | 2 d t ) 1 / 2 .
Invoking Lemma 2.1 and the Littlewood-Paley theory imply
I j f p C ( k Z | S j + k f | 2 ) 1 / 2 p C f p , 1 < p < .
(2.9)
On the other hand, by Plancherel’s theorem and (2.4)-(2.5), we have
I j f 2 2 = 1 2 k Z R n | f ˆ ( ξ ) | 2 | λ j + k ( | L ( ξ ) | ) | 2 | τ k , t ˆ ( ξ ) | 2 d ξ d t C 1 2 k Z { a j + k + 1 1 | L ( ξ ) | a j + k 1 1 } | f ˆ ( ξ ) | 2 | τ k , t ˆ ( ξ ) | 2 d ξ d t C B j 2 f 2 2 ,
where B j = a j + 1 χ { j 1 } + ( | j + 1 | log a ) β χ { j < 1 } . That is,
I j f 2 C B j f 2 .
(2.10)
Interpolating between (2.9) and (2.10), there exists ϵ ( 2 / ( 2 β + 1 ) , 1 ) such that
I j f p C B j ϵ f p , 1 + 1 / ( 2 β ) < p < 2 .
For fixed p ( 1 + 1 / ( 2 β ) , 2 ) and β > 1 / 2 , we can choose q ( 1 , p ) such that q ϵ β > 1 . Thus
j Z I j f p q C ( j 1 a q ϵ ( j 1 ) + j < 1 ( | j + 1 | log a ) q ϵ β ) f p q C ( a ) f p q ,
which, together with (2.8), implies
G ( f ) p C ( a ) f p , for  1 + 1 / ( 2 β ) < p < 2 .
(2.11)
Case 2. 2 < p < 1 + 2 β . By (2.3) and (2.7), we have for 2 < p < and 1 < q < p ,
G ( f ) p q C j Z ( 1 2 ( k Z | τ k , t S j + k f | 2 ) 1 / 2 p 2 d t ) q / 2 .
(2.12)
Let
J j , t f ( x ) : = ( k Z | τ k , t S j + k f ( x ) | 2 ) 1 / 2 .
By (2.6), [[23], p.544, Lemma] and the Littlewood-Paley theory, we have, for k Z and t [ 1 , 2 ] ,
J j , t f p 0 C ( k Z | S j + k f | 2 ) 1 / 2 p 0 C f p 0 , 1 < p 0 < .
(2.13)
On the other hand, by the same arguments as in (2.10), we have
J j , t f 2 C B j f 2 ,
(2.14)
where B j is as in (2.10). On interpolation between (2.13) and (2.14), for fixed p ( 2 , 1 + 2 β ) and β > 1 / 2 , we can choose q ( 1 , p ) and δ ( 2 / ( 2 β + 1 ) , 1 ) such that q δ β > 1 and
J j , t f p C B j δ f p , for  2 < p < 1 + 2 β .
This, combined with (2.12), implies
G ( f ) p q C ( j 1 a q δ ( j 1 ) + j < 1 ( | j + 1 | log a ) q δ β ) f p q C ( a ) f p q ,

which, together with (2.11), completes the proof of Lemma 2.3. □

Lemma 2.4 ([[13], Lemma 2.2])

Suppose Φ ( t ) = t α 1 + μ 2 t α 2 + + μ n t α n and φ F , where μ 2 , , μ n are real parameters, and α 1 , , α n are distinct positive (not necessarily integer) exponents. Then for any r > 0 and λ R { 0 } ,
| r / 2 r exp ( i λ Φ ( φ ( t ) ) ) d t t | C ( φ ) | λ φ ( r ) α 1 | ϵ ,

where ϵ = min { 1 / α 1 , 1 / n } and C ( φ ) does not depend on μ 2 , , μ n .

Lemma 2.5 ([[24], Lemma 2.2])

Let P ( t ) = ( P 1 ( t ) , , P d ( t ) ) with P j being real polynomials defined on R + . Suppose that φ F . Then the operator M P , φ defined by
M P , φ ( f ) ( x ) = sup r > 0 r 2 r | f ( x P ( φ ( t ) ) ) | d t t

is bounded on L p ( R d ) for 1 < p < . The bound is independent of the coefficients of P j for all 1 j d and f but depends on φ.

Lemma 2.6 ([25])

Let Φ : S 1 R d , Φ = ( Φ 1 , , Φ d ) be real analytic on S 1 . Suppose that { Φ 1 , , Φ d } is linearly independent set. If Ω F β ( S 1 ) for some β > 1 , then
sup ξ S d 1 S 1 | Ω ( y ) | ( log + 1 | ξ Φ ( y ) | ) β d σ ( y ) < .

3 Proofs of main theorems

Proof of Theorem 1.1 Let N = max 1 j n deg ( P j ) . For 1 l n , let P l ( t ) = i = 1 N a i , l t i . For 1 s N , and 1 l n , let P l ( s ) ( t ) = i = 1 s a i , l t i and P ( s ) ( t ) = ( P 1 ( s ) ( t ) , , P n ( s ) ( t ) ) . Set P ( 0 ) ( t ) = 0 and
Φ s ( y ) = ( P 1 ( s ) ( φ ( ρ ( y ) ) ) y 1 , , P n ( s ) ( φ ( ρ ( y ) ) ) y n ) .
Then we can write
Φ s ( y ) ξ = l = 1 n ξ l y l P l ( s ) ( φ ( ρ ( y ) ) ) = l = 1 n i = 1 s ξ l y l a i , l φ ( ρ ( y ) ) i = i = 1 s ( L i ( ξ ) y ) φ ( ρ ( y ) ) i ,
(3.1)
where L i : R n R n is the linear transformation given by
L i ( ξ ) = ( a i , 1 ξ 1 , , a i , n ξ n ) .
For each j Z , t R + and 1 s N , we define the measures { σ j , t s } and { | σ j , t s | } by
σ j , t s ˆ ( ξ ) = 1 2 j t 2 j 1 t < ρ ( y ) 2 j t exp ( 2 π i Φ s ( y ) ξ ) Ω ( y ) ρ ( y ) α 1 d y , | σ j , t s | ˆ ( ξ ) = 1 2 j t 2 j 1 t < ρ ( y ) 2 j t exp ( 2 π i Φ s ( y ) ξ ) | Ω ( y ) | ρ ( y ) α 1 d y .
We get from (3.1)
| σ j , t s ˆ ( ξ ) σ j , t s 1 ˆ ( ξ ) | 1 2 j t 2 j 1 t < ρ ( y ) 2 j t | exp ( 2 π i Φ s ( y ) ξ ) exp ( 2 π i Φ s 1 ( y ) ξ ) | | Ω ( y ) | ρ ( y ) α 1 d y C ( φ ( 2 j t ) s | L s ( ξ ) | ) .
(3.2)
On the other hand, by a change of variable, we have
| σ j , t s ˆ ( ξ ) | = 1 2 j t | 2 j 1 t 2 j t S n 1 exp ( 2 π i k = 1 s L k ( ξ ) θ φ ( r ) k ) Ω ( θ ) J ( θ ) d σ ( θ ) d r | C S n 1 | Ω ( θ ) | | I j , t , s , ξ ( θ ) | d σ ( θ ) ,
(3.3)
where
I j , t , s , ξ ( θ ) : = 1 2 j t 2 j 1 t 2 j t exp ( 2 π i k = 1 s L k ( ξ ) θ φ ( r ) k ) d r .
By Lemma 2.4, we have
| I j , t , s , ξ ( θ ) | C | φ ( 2 j t ) s L s ( ξ ) θ | 1 / s .
Combining the trivial inequality | I j , t , s , ξ ( θ ) | C with the fact that t / ( log t ) β is increasing in ( e β , ) , we have
| I j , t , s , ξ ( θ ) | C ( log e β s | η θ | 1 ) β ( log | φ ( 2 j t ) s L s ( ξ ) | ) β , if  | φ ( 2 j t ) s L s ( ξ ) | > 1 ,
(3.4)
where η = L s ( ξ ) / | L s ( ξ ) | . This, together (3.3) with the fact that Ω F β ( S n 1 ) , implies
| σ j , t s ˆ ( ξ ) | C ( log | φ ( 2 j t ) s L s ( ξ ) | ) β , if  | φ ( 2 j t ) s L s ( ξ ) | > 1 .
(3.5)
Now we can choose a function ψ C 0 ( R ) such that ψ ( t ) 1 for | t | 1 / 2 and ψ ( t ) 0 for | t | > 1 . For 1 s N , j Z and t R + , we define the measures { τ j , t s } by
τ j , t s ˆ ( ξ ) = σ j , t s ˆ ( ξ ) k = s + 1 N ψ ( | φ ( 2 j t ) k L k ( ξ ) | ) σ j , t s 1 ˆ ( ξ ) k = s N ψ ( | φ ( 2 j t ) k L k ( ξ ) | ) .
Here we use the convention j a j = 1 . It is easy to see that
σ j , t N = s = 1 N τ j , t s .
(3.6)
It follows from (3.2), (3.5), and the trivial estimate | σ j , t s ˆ ( ξ ) | C that, for 1 s N ,
| τ j , t s ˆ ( ξ ) | C ( φ ) min { 1 , φ ( 2 j t ) s | L s ( ξ ) | } ,
(3.7)
| τ j , t s ˆ ( ξ ) | C ( φ ) ( log | φ ( 2 j t ) s L s ( ξ ) | ) β , if  | φ ( 2 j t ) s L s ( ξ ) | > 1 .
(3.8)
By the definition of σ j , t s and (3.6), we can write
M Ω Φ ( f ) ( x ) = ( 0 | j = 0 2 j σ j , t N f ( x ) | 2 d t t ) 1 / 2 j = 0 2 j ( 0 | σ j , t N f ( x ) | 2 d t t ) 1 / 2 2 ( j Z 2 j 2 j + 1 | σ 0 , t N f ( x ) | 2 d t t ) 1 / 2 2 ( j Z 1 2 | σ j , t N f ( x ) | 2 d t t ) 1 / 2 2 s = 1 N ( 1 2 j Z | τ j , t s f ( x ) | 2 d t t ) 1 / 2 .
(3.9)
On the other hand, by a change of variable we have
| | σ j , t s | f ( x ) | 1 2 j t 2 j 1 t < ρ ( y ) 2 j t | f ( x Φ s ( y ) ) | | Ω ( y ) | ρ ( y ) α 1 d y C S n 1 | Ω ( y ) | ( 1 2 j t 2 j 1 t 2 j t | f ( x Φ s ( A r y ) ) | d r ) d σ ( y ) C S n 1 | Ω ( y ) | M P , φ ( f ) ( x ) d σ ( y ) ,
where M P , φ is as in Lemma 2.5 and P ( t ) = ( P 1 ( s ) ( t ) y 1 , , P n ( s ) ( t ) y n ) . By Lemma 2.5 and Minkowski’s inequality, we have
sup j Z sup t > 0 | | σ j , t s | f | p C f p .
This inequality, together with the definition of τ j , t s , yields
sup j Z sup t > 0 | | τ j , t s | f | p C f p .
(3.10)

Then Theorem 1.1 follows from (3.7)-(3.10) and Lemma 2.3. □

Proof of Theorem 1.3 Let Φ, P N , φ, ϕ be as in Theorem 1.3. For 1 s N , we set P s ( t ) = m = 1 s a m t m . Define the measures { σ j , t s } and { | σ j , t s | } by
σ j , t s ˆ ( ξ ) = 1 2 j t 2 j 1 t < ρ ( y ) 2 j t exp ( 2 π i P s ( φ ( ρ ( y ) ) ) ϕ ( y ) ξ ) Ω ( y ) ρ ( y ) α 1 d y ;
(3.11)
| σ j , t s | ˆ ( ξ ) = 1 2 j t 2 j 1 t < ρ ( y ) 2 j t exp ( 2 π i P s ( φ ( ρ ( y ) ) ) ϕ ( y ) ξ ) | Ω ( y ) | ρ ( y ) α 1 d y .
(3.12)
Following the notation in [20], let { ϕ i 1 , , ϕ i l } be a maximal linearly independent subset of { ϕ 1 , , ϕ d } , where 1 l d , 1 i r d and r = 1 , , l . Thus, for j { i 1 , , i l } , there exist a ( j ) = ( a j , 1 , , a j , l ) R l such that
ϕ j ( y ) = a ( j ) ( ϕ i 1 ( y ) , , ϕ i l ( y ) ) = k = 1 l a j , k ϕ i k ( y ) .
This implies that there exists a linear transformation L : R d R l such that
ϕ ( y ) ξ = L ( ξ ) ϕ ˜ ( y ) , ξ R d ,
where ϕ ˜ ( y ) = ( ϕ i 1 ( y ) , , ϕ i l ( y ) ) . Thus
| σ j , t s ˆ ( ξ ) | C S 1 | Ω ( z ) | | I j , t , s , ξ ( z ) | d σ ( z ) ,
where
I j , t , s , ξ ( z ) : = 1 2 j t 2 j 1 t 2 j t exp [ 2 π i P s ( φ ( u ) ) ( L ( ξ ) ϕ ˜ ( z ) ) ] d u .
By Lemma 2.4, we have
| I j , t , s , ξ ( z ) | C | φ ( 2 j t ) s a s L ( ξ ) ϕ ˜ ( z ) | 1 / s .
(3.13)
Since t / ( log t ) β is increasing in ( e β , ) for any β > 0 , and | ϕ ˜ ( z ) | B with B > 1 for any z S n 1 . We can deduce from (3.13) and the trivial estimate | I j , t , s , ξ ( z ) | C that
| I j , t , s , ξ ( z ) | C ( log ( B e β s | ζ ϕ ˜ ( z ) | 1 ) ) β ( log | φ ( 2 j t ) s a s L ( ξ ) | ) β , if  | φ ( 2 j t ) a s L ( ξ ) | > 1 ,
(3.14)
where ζ = L ( ξ ) / | L ( ξ ) | . Invoking Lemma 2.6 and (3.14), we obtain, for β > 1 ,
| σ j , t s ˆ ( ξ ) | C ( log | φ ( 2 j t ) s a s L ( ξ ) | ) β , if  | φ ( 2 j t ) s a s L ( ξ ) | > 1 .
(3.15)
On the other hand, we have
| σ j , t s ˆ ( ξ ) σ j , t s 1 ˆ ( ξ ) | 1 2 j t 2 j 1 t < ρ ( y ) 2 j t | exp ( 2 π i ( a s φ ( ρ ( y ) ) s | L ( ξ ) | ϕ ˜ ( y ) ) ) 1 | | Ω ( y ) | ρ ( y ) α 1 d y C | φ ( 2 j t ) s a s L ( ξ ) | S 1 | Ω ( z ) | | ϕ ˜ ( z ) | d σ ( z ) C | φ ( 2 j t ) s a s L ( ξ ) | sup z S 1 | ϕ ˜ ( z ) | C | φ ( 2 j t ) s a s L ( ξ ) | .
(3.16)
Notice that
| | σ j , t s | f ( x ) | 1 2 j t 2 j 1 t < ρ ( y ) 2 j t | f ( x P s ( φ ( ρ ( y ) ) ) ϕ ( y ) ) | | Ω ( y ) | ρ ( y ) α 1 d y C S 1 | Ω ( y ) | ( 0 2 j t | f ( x P s ( φ ( u ) ) ϕ ( y ) ) | d u u ) d σ ( y ) .
This combining with Lemma 2.5 and Minkowski’s inequality, implies
sup j Z sup t > 0 | | σ j , t s | f | p C f p .
(3.17)

Then the rest of the proof of Theorem 1.3 follows from an argument which is similar to those in the proof of Theorem 1.1 and (3.15)-(3.17). We omit the details. □

Proof of Theorem 1.4 Let n 2 , P = ( P 1 , , P d ) , where P j : R n R is a polynomial for 1 j d . Let
M = max { deg ( P 1 ) , , deg ( P d ) } ,
and
P i ( y ) = | γ | M a i γ y γ for  i = 1 , , d .
For 1 s M , we let
P ( s ) = ( P 1 , s , , P d , s ) ,
where
P i , s ( y ) = | γ | s a i γ y γ .

Set P ( 0 ) = 0 and Φ s ( y ) = P ( s ) ( φ ( ρ ( y ) ) y ) .

For each j Z , t R + and 0 s M , we define the measures { σ j , t s } and { | σ j , t s | } by
σ j , t s ˆ ( ξ ) = 1 2 j t 2 j 1 t < ρ ( y ) 2 j t exp ( 2 π i Φ s ( y ) ξ ) Ω ( y ) ρ ( y ) α 1 d y ; | σ j , t s | ˆ ( ξ ) = 1 2 j t 2 j 1 t < ρ ( y ) 2 j t exp ( 2 π i Φ s ( y ) ξ ) | Ω ( y ) | ρ ( y ) α 1 d y .
For 1 s M , let l s denote the number of multi-indices γ = ( γ 1 , , γ n ) satisfying | γ | = s , and define the linear transformation L s : R d R l s by
L s ( ξ ) = ( ( L s ( ξ ) ) γ ) | γ | = s = ( i = 1 d a i γ ξ i ) | γ | = s .
By the change of variables, we have
| σ j , t s ˆ ( ξ ) | = | S n 1 1 2 j t 2 j 1 t 2 j t exp ( 2 π i ξ P ( s ) ( φ ( u ) θ ) ) d u Ω ( θ ) J ( θ ) d σ ( θ ) | C S n 1 | Ω ( θ ) | | 1 2 j t 2 j 1 t 2 j t exp ( 2 π i ξ P ( s ) ( φ ( u ) θ ) ) d u | d σ ( θ ) C S n 1 | Ω ( θ ) | | 1 2 j t 2 j 1 t 2 j t exp ( 2 π i | γ | s i = 1 d ξ j a j γ θ γ φ ( u ) | γ | ) d u | d σ ( θ ) C S n 1 | Ω ( θ ) | | J j , t , s , ξ ( θ ) | d σ ( θ ) ,
where
J j , t , s , ξ ( θ ) : = 1 2 j t 2 j 1 t 2 j t exp ( 2 π i | γ | = s ( L s ( ξ ) ) γ θ γ φ ( u ) s + lower powers in  u ) d u .
Let
Q s , ξ ( θ ) : = | L s ( ξ ) | 1 | γ | = s ( L s ( ξ ) ) γ θ γ .
By Lemma 2.4, we have
| J j , t , s , ξ ( θ ) | C ( φ ( 2 j t ) s | L s ( ξ ) | | Q s , ξ ( θ ) | ) 1 / s .
By this inequality, together with the trivial estimate | J j , t , s , ξ ( θ ) | C , we get
| J j , t , s , θ ( ξ ) | C ( log ( e β s l s 1 / 2 | Q s , ξ ( θ ) | 1 ) ) β ( log | φ ( 2 j t ) s L s ( ξ ) | ) β , if  | φ ( 2 j t ) s L s ( ξ ) | > 1 .
Since Ω s = 1 F ( n , s , β ) , Q s , ξ V ( n , s ) and Q s , ξ = 1 , we immediately obtain
| σ j , t s ˆ ( ξ ) | C ( log | φ ( 2 j t ) s L s ( ξ ) | ) β , if  | φ ( 2 j t ) s L s ( ξ ) | > 1 .
(3.18)
On the other hand, we have
| σ j , t s ˆ ( ξ ) σ j , t s 1 ˆ ( ξ ) | 1 2 j t 2 j 1 t < ρ ( y ) 2 j t | exp ( 2 π i j = 1 d | ι | = s ξ j a j ι φ ( ρ ( y ) ) s ( y ) ι ) 1 | | Ω ( y ) | ρ ( y ) α 1 d y C | φ ( 2 j t ) s L s ( ξ ) | .
(3.19)
In addition, using Lemma 2.5, one can easily check that
sup j Z sup t > 0 | | σ j , t s | f | p C f p .
(3.20)

Then the rest proof of Theorem 1.4 follows from similar arguments to the proof of Theorem 1.1 and (3.18)-(3.20). Details will be omitted. □

Declarations

Acknowledgements

The authors would like to thank the referees for their carefully reading and invaluable comments. This work was supported by the NNSF of China (Nos. 11101339, 11371295).

Authors’ Affiliations

(1)
School of Mathematical Sciences, Xiamen University
(2)
College of Mathematics and Systems Science, Shandong University of Science and Technology

References

  1. Fabes E, Rivière N: Singular integrals with mixed homogeneity. Stud. Math. 1966, 27: 19-38.MATHGoogle Scholar
  2. Stein EM: On the function of Littlewood-Paley, Lusin and Marcinkiewicz. Trans. Am. Math. Soc. 1958, 88: 430-466. 10.1090/S0002-9947-1958-0112932-2View ArticleGoogle Scholar
  3. Benedek A, Calderón A, Panzone R: Convolution operators on Banach space valued functions. Proc. Natl. Acad. Sci. USA 1962, 48: 356-365. 10.1073/pnas.48.3.356View ArticleMATHGoogle Scholar
  4. Chen J, Fan D, Pan Y: A note on a Marcinkiewicz integral operator. Math. Nachr. 2001, 227: 33-42. 10.1002/1522-2616(200107)227:1<33::AID-MANA33>3.0.CO;2-0MathSciNetView ArticleMATHGoogle Scholar
  5. Ding Y, Fan D, Pan Y: L p -boundedness of Marcinkiewicz integrals with Hardy space function kernel. Acta Math. Sin. Engl. Ser. 2000, 16: 593-600. 10.1007/s101140000015MathSciNetView ArticleMATHGoogle Scholar
  6. Ding Y, Pan Y: L p bounds for Marcinkiewicz integrals. Proc. Edinb. Math. Soc. 2003, 46: 669-677. 10.1017/S0013091501000682MathSciNetView ArticleMATHGoogle Scholar
  7. Ma B, Wu H, Zhao X: Rough Marcinkiewicz integrals along certain smooth curves. Front. Math. China 2012,7(5):857-872. 10.1007/s11464-012-0237-yMathSciNetView ArticleMATHGoogle Scholar
  8. Wu H: L p bounds for Marcinkiewicz integrals associates to surfaces of revolution. J. Math. Anal. Appl. 2006, 321: 811-827. 10.1016/j.jmaa.2005.08.087MathSciNetView ArticleMATHGoogle Scholar
  9. Xu H, Chen J, Ying Y:A note on Marcinkiewicz integrals with H 1 kernels. Acta Math. Sci. 2003,23(1):133-138.MathSciNetMATHGoogle Scholar
  10. Grafakos L, Stefanov A: L p bounds for singular integrals and maximal singular integrals with rough kernels. Indiana Univ. Math. J. 1998, 47: 455-469.MathSciNetView ArticleMATHGoogle Scholar
  11. Al-Salman A, Al-Qassem H, Cheng LC, Pan Y: L p bounds for the function of Marcinkiewicz. Math. Res. Lett. 2002, 9: 697-700. 10.4310/MRL.2002.v9.n5.a11MathSciNetView ArticleMATHGoogle Scholar
  12. Wu H: On Marcinkiewicz integral operators with rough kernels. Integral Equ. Oper. Theory 2005, 52: 285-298. 10.1007/s00020-004-1339-zView ArticleMATHGoogle Scholar
  13. Liu, F, Wu, H, Zhang, D: Parametric Marcinkiewicz integrals with rough kernels supported by compound subvariaties. Acta. Math. Sci. (Chin. Ser.) (2014, in press)Google Scholar
  14. Ding Y, Xue Q, Yabuta K: Parabolic Littlewood-Paley g-function with rough kernel. Acta Math. Sin. Engl. Ser. 2008,24(10):2049-2060.MathSciNetMATHGoogle Scholar
  15. Chen Y, Ding Y: L p bounds for parabolic Marcinkiewicz integral with rough kernels. J. Korean Math. Soc. 2007, 44: 733-745. 10.4134/JKMS.2007.44.3.733MathSciNetView ArticleMATHGoogle Scholar
  16. Chen D, Lu S: L p boundedness of parabolic Littlewood-Paley operator with rough kernel belonging to F α ( S n 1 ) . Acta Math. Sci. 2011,31(2):343-350.MathSciNetMATHGoogle Scholar
  17. Liu F, Wu H: Rough Marcinkiewicz integrals with mixed homogeneity on product spaces. Acta Math. Sin. Engl. Ser. 2013,29(7):1231-1244. 10.1007/s10114-013-1675-5MathSciNetView ArticleMATHGoogle Scholar
  18. Al-Salman A: A note on parabolic Marcinkiewicz integrals along surfaces. Proc. A. Razmadze Math. Inst. 2010, 154: 21-36.MathSciNetMATHGoogle Scholar
  19. Al-Salman A: Parabolic Marcinkiewicz integrals along surfaces on product domains. Acta Math. Sin. Engl. Ser. 2011,27(1):1-18. 10.1007/s10114-010-9653-7MathSciNetView ArticleMATHGoogle Scholar
  20. Al-Balushi, K, Al-Salman, A: Certain L p bounds for rough singular integrals. Preprint (2013)Google Scholar
  21. Al-Salman A, Pan Y: Singular integrals with rough kernels. Can. Math. Bull. 2004,47(1):3-11. 10.4153/CMB-2004-001-8MathSciNetView ArticleMATHGoogle Scholar
  22. Al-Salman A, Al-Qassem H: Rough Marcinkiewicz integral operators. Int. J. Math. Math. Sci. 2001,27(8):495-503. 10.1155/S0161171201006548MathSciNetView ArticleMATHGoogle Scholar
  23. Duoandikoetxea J, Rubio de Francia JL: Maximal and singular integral operators via Fourier transform estimates. Invent. Math. 1986, 84: 541-561. 10.1007/BF01388746MathSciNetView ArticleMATHGoogle Scholar
  24. Liu F, Wu H: Multiple singular integrals and Marcinkiewicz integrals with mixed homogeneity along surfaces. J. Inequal. Appl. 2012,2012(189):1-23.Google Scholar
  25. Al-Salman, A: Estimates of singular integral operators of convolution type with rough kernels. PhD Dissertation, University of Pittsburgh (1999)Google Scholar

Copyright

© Zhang and Liu; licensee Springer. 2014

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.