Iterated Commutators for Multilinear Singular Integral Operators on Morrey Space with Non-Doubling Measures

Let μ be a non-negative Radon measure on R ^d which only satisfies the following growth condition that there exists a positive constant C such that μ( B( x, r)) ≤ C r ⁿfor all x∈ R ^d, r > 0 and some fixed n ∈ (0, d]. This paper is interested in the properties of the iterated commutators of multilinear singular integral operators on Morrey spaces . Precisely speaking, we show that the iterated commutators generated by multilinear singular integrals operators are bounded from to where (Regular Bounded Mean Oscillation space) and 1 < q_j ≤ p_j <∞ with 1/ p = 1/ p ₁ + ... + 1/ p _m and 1/ q = 1/ q ₁+ ... + 1/ q_m.

• Non-Doubling Measures
• Morrey Space
• Multilinear Singular Integral Operators
• RBMO
• Commutator

1. Introduction

Let μ be a positive Radon measures on ℝ d satisfying only the growth condition, that is, there exists a constant C > 0 and n ∈ ( 0, d ] such that

μ ( Q ) ≤ C l ( Q ) n (1)

for any cube Q ⊂ ℝ d with sides parallel to the coordinate axes. Q ( x , l ( Q ) ) will be the cube centered at x with side length l ( Q ) . For r > 0 , rQ will denote the cube with the same center as Q and with l ( r Q ) = r l ( Q ) . The set of all cubes Q ⊂ ℝ d , satisfying μ ( Q ) > 0 is denoted by L ( μ ) . In this note, we do not assume that μ is doubling.

Nazarov, Treil and Volberg developed the theory of the singular integrals for the measures with growth condition to investigate the analytic capacity on the complex plane [1] [2]. Tolsa showed that the analytic capacity is subadditive and that it is bi-Lipschitz invariant [3] [4] and defined for the growth measures RBMO (regular bounded mean oscillation) space, the Hardy space H 1 ( μ ) and the Littlewood-Paley decomposition [5] [6]. He also gave his H 1 ( μ ) space in terms of the grand maximal operator [7]. Recently many people paid attention to the measure with growth condition because of recovering the Calderón-Zygmund theory and solving the long-standing open Painlevé problem.

The boundedness of fractional integral operators on the classical Morrey spaces was studied by Adams [8], Chiarenza et al. [9]. In [9], by establishing a pointwise estimate of fractional integrals in terms of the Hardy-Littlewood maximal function, they showed the boundedness of fractional integral operators on the Morrey spaces. In 2005, Sawano and Tanaka [10] gave a natural definition of Morrey spaces for Radon measures which might be non-doubling but satisfied the growth condition, and they investigated the boundedness in these spaces of some classical operators in harmonic analysis. Later on, Sawano [11] defined the generalized Morrey spaces on ℝ n for non-doubling measure and showed the properties of maximal operators, fractional integral operators and singular operators in this space.

A classical result of commutator is due to Coifman, Rochberg and Weiss [12], if b ∈ B M O and T is a Calderón-Zygmund operator, then the commutator [ b , T ] is bounded on L p spaces for 1 < p < ∞ . The same result for the multilinear commutator was obtained by Pérez and Trujillo-Gonzalez [13]. Tolsa [5] developed the theory of Calderón-Zygmund operators and their commutators with RBMO functions in the setting of non-doubling measures. Hu, Meng and Yang [14] considered the multilinear commutator on Lebesgue spaces with non-doubling measures. Chen and Sawyer [15] modified the definition of RBMO to investigate the commutators of the potential operators and RBMO functions.

In the last decade, multilinear singular integrals of Calderón-Zygmund type have attracted great attentions. Some interesting results refer to [16] [17] [18] [19] [20] in the text of Lebesgue measures. It points out that Perez and Pradolini [21] introduced a said iterated commutators generated by the multilinear singular integral operators with Calderón-Zygmund type and vector function b → ∈ R B M O m and obtained the boundedness from L p 1 × ⋯ × L p m to L p with 1 / p = 1 / p 1 + ⋯ + 1 / p m for 1 < p 1 , ⋯ , p m < ∞ (in fact, they considered the weighted case). Xu [22] extended the result to the case of the non-doubling measures. Very recently, Tao and He [23] obtained the boundedness of the multilinear Calderón-Zygmund operators on the generalized Morrey spaces over the quasi-metric space of non-homogeneoustype. The aim of this paper is to study the iterated commutators of multilinear singular integral operators on Morrey spaces with non-doubling measures.

Before stating our result, we recall some definitions and notation. Given β d ≥ 2 d + 1 large enough but depending only on the dimension d, we say that a cube Q ⊂ ℝ d is doubling if μ ( 2 Q ) ≤ β d μ ( Q ) . For any fixed cube Q ⊂ ℝ d , let N be the smallest nonnegative integer such that 2 N Q is doubling. We denote this cube by Q ˜ .

For two cubes Q ⊂ R in ℝ d , we suppose

K Q , R = 1 + ∑ k = 1 N Q , R μ ( 2 k Q ) l ( 2 k Q ) n , (2)

where N Q , R is the first positive integer k such that l ( 2 k Q ) ≥ l ( R ) . This was introduced by Tolsa in [5].

Let m Q f be the mean value of f on Q, namely, m Q f = 1 μ ( Q ) ∫ Q     f ( x ) d μ . The regularity bounded mean oscillations function spaces were introduced by Tolsa [5].

Definition 1.1. Let η > 1 be a fixed constant. We say that f ∈ L l o c 1 ( μ ) is in RBMO if there exists a constant A such that

1 μ ( η Q ) ∫ Q | f ( y ) − m Q ˜ f | d μ ( y ) ≤ A , (3)

for any cube Q, and

| m Q f − m R f | ≤ A K Q , R , (4)

for any two doubling cubes Q ⊂ R . The mininal constant A is the R B M O ( μ ) norm of f, and it will be denoted by ‖ f ‖ ∗ .

The definition of the Morrey space with non-doubling measure is given in the following [10].

Definition 1.2. Let k > 1 and 1 ≤ q ≤ p < ∞ , the Morrey space M q p ( k , μ ) is defined as

M q p ( k , μ ) : = { f ∈ L l o c q ( μ ) : ‖ f ‖ M q p ( k , μ ) < ∞ } ,

where

‖ f ‖ M q p ( k , μ ) : = sup Q ∈ L ( μ ) μ ( k Q ) 1 p − 1 q ( ∫ Q | f | q d μ ) 1 q . (5)

It is easy to observe that L p ( μ ) = M p p ( k , μ ) , and Hölder’s inequality tells us ‖ f ‖ M q 2 p ( k , μ ) ≤ ‖ f ‖ M q 1 p ( k , μ ) for all 1 ≤ q 2 ≤ q 1 ≤ p , then we have L p ( μ ) = M p p ( k , μ ) ⊂ M q 1 p ( k , μ ) ⊂ M q 2 p ( k , μ ) . The space M q p ( k , μ ) is a Banach space with its norm ‖ f ‖ M q p ( k , μ ) and the parameter k > 1 appearing in the definition does not affect it. The Morrey space norm reflects local regularity of f more precisely than the Lebesgue space norm. See [10] [11] [24] for details. We will denote M q p ( 2, μ ) by M q p ( μ ) .

Denoting by f → = ( f 1 , f 2 , ⋯ , f m ) , we consider the multilinear singular integral operator T m as follows,

T m ( f → ) ( x ) = ∫ ( ℝ d ) m     K ( x , y 1 , ⋯ , y m ) f 1 ( y 1 ) ⋯ f m ( y m ) d μ ( y 1 ) ⋯ d μ ( y m )     for   x ∈ ℝ d , (6)

whenever f 1 , ⋯ , f m are C ∞ -functions with compact support and x ∉ ∩ j = 1 m   supp   f j . Moreover,

| K ( y 0 , y 1 , ⋯ , y m ) | ≤ A ( ∑ k , l = 0 m | y k − y l | ) − m n (7)

and, for some ϵ > 0 ,

| K ( y 0 , y 1 , ⋯ , y j , ⋯ , y m ) − K ( y 0 , y 1 , ⋯ , y ′ j , ⋯ , y m ) | ≤ A | y j − y ′ j | ϵ ( ∑ k , l = 0 m | y k − y l | ) m n + ϵ (8)

provided that 0 ≤ j ≤ m and | y j − y ′ j | ≤ 1 2 max 0 ≤ k ≤ m | y j − y k | .

Let b i ∈ R B M O ( μ ) for i = 1 , 2 , ⋯ , m and let b → = ( b 1 , b 2 , ⋯ , b m ) , then the iterated commutator ( T m ) ∏ b → is formally defined as

( T m ) ∏ b → ( f → ) ( x ) = ∫ ( ℝ d ) m ∏ j = 1 m ( b j ( x ) − b j ( y j ) ) K ( x , y 1 , ⋯ , y m )     × f 1 ( y 1 ) ⋯ f m ( y m ) d μ ( y 1 ) ⋯ d μ ( y m ) . (9)

Suppose ‖ μ ‖ = ∞ the main result in this paper can be stated as follow.

Theorem 1.1. Let ( T m ) ∏ b → as in (9) and satisfying conditions (7) and (8). Let 1 < q j ≤ p j < ∞ with 1 p = 1 p 1 + ⋯ + 1 p m and 1 q = 1 q 1 + ⋯ + 1 q m . Suppose b i ∈ R B M O ( μ ) for i = 1 , 2 , ⋯ , m . If T m maps L 1 ( μ ) × ⋯ × L 1 ( μ ) to L 1 / m , ∞ ( μ ) , then the commutators ( T m ) ∏ b → are bounded from M q 1 p 1 ( μ ) × ⋯ × M q m p m ( μ ) to M q p ( μ ) , that is,

‖ ( T m ) ∏ b → ( f → ) ( x ) ‖ M q p ( μ ) ≤ C ∏ i = 1 m ‖ b i ‖ ∗ ‖ f i ‖ M q i p i ( μ ) . (10)

More generally, denote by C i m the family of all subsets σ = { σ 1 , σ 2 , ⋯ , σ i } of i different elements of { 1,2, ⋯ , m } , and let σ ′ = { 1 , 2 , ⋯ , m } \ σ and b → σ = { b σ 1 , b σ 2 , ⋯ , b σ i } . For any σ ∈ C i m , we define

( T m ) ∏ b → σ ( f → ) ( x ) = ∫ ( ℝ d ) m ∏ σ j ∈ σ ( b σ j ( x ) − b σ j ( y σ j ) ) K ( x , y 1 , ⋯ , y m )     × f 1 ( y 1 ) ⋯ f m ( y m ) d μ ( y 1 ) ⋯ d μ ( y m ) . (11)

In case σ = { 1 , 2 , ⋯ , m } , one sees that ( T m ) ∏ b → σ is just the commutator ( T m ) ∏ b → . So we have a more generalization version of the theorem as following.

Theorem 1.2. Let ( T m ) ∏ b → σ as in (11) and satisfying conditions (7) and (8). Let 1 < q j ≤ p j < ∞ with 1 p = 1 p 1 + ⋯ + 1 p m and 1 q = 1 q 1 + ⋯ + 1 q m . Suppose b i ∈ R B M O ( μ ) for i = 1 , 2 , ⋯ , m . If T m maps L 1 ( μ ) × ⋯ × L 1 ( μ ) to L 1 / m , ∞ ( μ ) , then for all σ ⊂ { 1,2, ⋯ , m } , the commutators ( T m ) ∏ b → σ are bounded from M q 1 p 1 ( μ ) × ⋯ × M q m p m ( μ ) to M q p ( μ ) , that is,

‖ ( T m ) ∏ b → σ ( f → ) ( x ) ‖ M q p ( μ ) ≤ C ∏ j ∈ σ ‖ b j ‖ ∗ ∏ i = 1 m ‖ f i ‖ M q i p i ( μ ) . (12)

2. Proof of Main Results

Before proving our theorem, we recall the following maximal operator,

M k f ( x ) = sup x ∈ Q ⊂ L ( μ ) 1 μ ( k Q ) ∫ Q | f ( y ) | d μ ( y ) . (13)

we will use the sharp maximal estimates. Let f be a function in L l o c 1 ( μ ) , the sharp maximal function of f is defined by

M # f ( x ) = sup Q ∋ x 1 μ ( 3 2 Q ) ∫ Q | f ( y ) − m Q ˜ f | d μ ( y ) + sup R ⊃ Q ∋ x Q , R   doubling | m Q f − m R f | K Q , R (14)

The non-centered doubling maximal operator is defined by

N f ( x ) = sup Q ∋ x ,   Q   doubling 1 μ ( Q ) ∫ Q | f ( y ) | d μ ( y ) . (15)

By the Lebesgue differential theorem, it is easy to see that | f ( x ) | ≤ N f ( x ) for any f ∈ L l o c 1 ( μ ) and μ − a . e . x ∈ ℝ d . Define the non-centered maximal operator,

M τ , η f ( x ) = sup Q ∋ x { 1 μ ( η Q ) ∫ Q | f ( y ) | τ d μ ( y ) } 1 τ . (16)

for η > 1 and τ > 1 , where the supremum is taking over all the cubes Q containing the point x.

To prove Theorem 1.2 is reduced to the following lemmas.

Lemma 2.1. Let τ > 1 , s i > 1 , b i ∈ R B M O ( μ ) and f i ∈ L q i ( μ ) , i = 1,2, ⋯ , m and σ ⊂ { 1,2, ⋯ , m } . If T m maps L 1 ( μ ) × ⋯ × L 1 ( μ ) to L 1 / m , ∞ ( μ ) and satisfying conditions (7) and (8). Then we have

M # ( ( T m ) ∏ b → σ ( f → ) ) ( x ) ≤ C [ ∏ j ∈ σ ‖ b j ‖ ∗ M τ , 3 2 ( T m ( f → ) ) ( x )     + ∑ σ 1 ∪ σ 2 = σ σ 1 ≠ ∅ ,   σ 2 ≠ ∅ ∏ j ∉ σ 1 ‖ b j ‖ ∗ M τ , 3 2 ( T m ) ∏ b → σ 2 ( f → ) ( x )     + ∏ j ∈ σ ‖ b j ‖ ∗ ∏ j = 1 m M s j , 9 8 f j ( x ) ] . (17)

We postpone the proof of Lemma 2.1 after of Theorem 1.2.

Lemma 2.2. [10] Let q > τ > 1 , η > 1 , and 1 < q ≤ p < ∞ , then the operator M τ , η is bounded on M q p ( μ ) and

‖ M τ , η ( f ) ‖ M q p ( μ ) ≤ C ‖ f ‖ M q p ( μ ) .

with the constant C independent of f.

Lemma 2.3. [24] Suppose that 1 < q ≤ p < ∞ , and there exists an increasing sequence of concentric doubling cubes, I 0 ⊂ I 1 ⊂ ⋯ ⊂ I k ⊂ ⋯ , such that

lim k → ∞ m I k ( f ) = 0     and     ∪ k = 0 ∞ I k = ℝ d .

Then there exist a constant C > 0 independent on f such that

‖ N f ‖ M q p ( μ ) ≤ C ‖ M # ‖ M q p ( μ ) .

Lemma 2.4. [25] Let 1 < p i < ∞ and 1 p = 1 p 1 + ⋯ + 1 p m . If T m maps L 1 ( μ ) × ⋯ × L 1 ( μ ) to L 1 / m , ∞ ( μ ) and satisfying conditions (7) and (8). Then there exists a constant C independent of f i such that

‖ T m ( f → ) ‖ L p ( μ ) ≤ C ∏ i = 1 m ‖ f i ‖ L p i ( μ ) .

Remark 2.1. If T m maps L 1 ( μ ) × ⋯ × L 1 ( μ ) to L 1 / m , ∞ ( μ ) and satisfying conditions (7) and (8), with 1 < q i ≤ p i < ∞ , 1 p = 1 p 1 + ⋯ + 1 p m and 1 q = 1 q 1 + ⋯ + 1 q m . From Corollary 1.8 in [23], we can easily get that T m is bounded from M q 1 p 1 ( μ ) × ⋯ × M q m p m ( μ ) to M q p ( μ ) .

Proof of Theorem 1.2. Using Lemma 2.1, Lemma 2.2, Lemma 2.3 and Lemma 2.4, we get that

‖ ( T m ) b i ( f → ) ‖ M q p ( μ ) ≤ C ‖ N ( ( T m ) b i ( f → ) ) ‖ M q p ( μ ) ≤ C ‖ M # ( ( T m ) b i ( f → ) ) ‖ M q p ( μ ) ≤ C ‖ b i ‖ ∗ ‖ M τ , 3 2 ( T m ( f → ) ) ( x ) + ∏ i = 1 m M s i , 9 8 f i ‖ M q p ( μ ) ≤ C ‖ b i ‖ ∗ ‖ M τ , 3 2 ( T m ( f → ) ) ‖ M q p ( μ ) + C ‖ b i ‖ ∗ ‖ ∏ i = 1 m M s i , 9 8 f i ‖ M q p ( μ ) ≤ C ‖ b i ‖ ∗ ‖ T m ( f → ) ‖ M q p ( μ ) + C ‖ b i ‖ ∗ ∏ i = 1 m ‖ f i ‖ M q j p j ( μ ) ≤ C ‖ b i ‖ ∗ ∏ j = 1 m ‖ f i ‖ M q j p j ( μ ) .

Applying the inequality (17) in Lemma 2.1, for σ ⊂ { 1,2, ⋯ , m } , we have

‖ ( T m ) ∏ b → σ ( f → ) ‖ M q p ( μ ) ≤ C ‖ N ( ( T m ) ∏ b → σ ( f → ) ) ‖ M q p ( μ ) ≤ C ‖ M # ( ( T m ) ∏ b → σ ( f → ) ) ‖ M q p ( μ ) ≤ C ‖ ∏ j ∈ σ ‖ b j ‖ ∗ M τ , 3 2 ( T m ( f → ) ) + ∑ σ 1 ∪ σ 2 = σ σ 1 ≠ ∅   σ 2 ≠ ∅ ∏ j ∈ σ 1 ‖ b j ‖ ∗ M τ , 3 2 ( ( T m ) ∏ b → σ 2 ( f → ) )   + ∏ j ∈ σ ‖ b j ‖ ∗ ∏ j = 1 m M s j , 9 8 f j ‖ M q p ( μ ) ≤ C ∏ j ∈ σ ‖ b j ‖ ∗ ∏ j = 1 m ‖ f j ‖ M q j p j ( μ ) + C ∑ σ 1 ∪ σ 2 = σ σ 1 ≠ ∅   σ 2 ≠ ∅ ∏ j ∈ σ 1 ‖ b j ‖ ∗ ‖ M τ , 3 2 ( ( T m ) ∏ b → σ 2 ( f → ) ) ‖ M q p ( μ ) ≤ C ∏ j ∈ σ ‖ b j ‖ ∗ ∏ j = 1 m ‖ f j ‖ M q j p j ( μ ) + C ∑ σ 1 ∪ σ 2 = σ σ 1 ≠ ∅   σ 2 ≠ ∅ ∏ j ∈ σ 1 ‖ b j ‖ ∗ ‖ ( T m ) ∏ b → σ 2 ( f → ) ‖ M q p (μ)

where σ 1 and σ 2 are two nonempty subsets of σ and σ 1 ∩ σ 2 = ∅ . Hence, we can make use of induction on σ ⊂ { 1,2, ⋯ , m } to get that

‖ ( T m ) ∏ b → σ ( f → ) ‖ M q p ( μ ) ≤ C ∏ j ∈ σ ‖ b j ‖ ∗ ∏ i = 1 m ‖ f i ‖ M q i p i ( μ ) . (18)

This completes the proof of Theorem 1.2.☐

For simplicity of the notation, we only show the special case σ = { 1 , ⋯ , m } of Lemma 2.1. The similar process with minor modification will be to able prove Lemma 2.1 for the general case.

Lemma 2.5. Let τ > 1 , s i > 1 , b i ∈ R B M O ( μ ) and f i ∈ L q i ( μ ) , i = 1 , 2 , ⋯ , m . If T m maps L 1 ( μ ) × ⋯ × L 1 ( μ ) to L 1 / m , ∞ ( μ ) and satisfying conditions (7) and (8), then there exists a constant C > 0 independent of b i and f i such that

M # ( ( T m ) ∏ b → ( f → ) ) ( x ) ≤ C [ ∏ i = 1 m ‖ b i ‖ ∗ M τ , 3 2 ( T m ( f → ) ) ( x )         + ∑ i = 1 m − 1 ∑ σ ∈ C i m ∏ j ∈ σ ‖ b j ‖ ∗ M η , 3 2 ( ( T m ) ∏ b → σ ′ ( f → ) ) ( x )         + ∏ i = 1 m ‖ b i ‖ ∗ M s i , 9 8 f i ( x ) ] (19)

In order to prove Lemma 2.5, we have the following decomposition for the commutators ( T m ) ∏ b → . For any λ → = ( λ 1 , λ 2 , ⋯ , λ m ) ∈ ℝ m , writing b → σ ( x ) − b → σ ( y ) = ( b → σ ( x ) − λ → σ ) − ( b → σ ( y ) − λ → σ ) , thus it is clear that ( T m ) ∏ b → σ ( f → ) = ( T m ) ∏ ( b → σ − λ → σ ) ( f → ) . Moreover,

( T m ) ∏ b → ( f → ) ( x ) = ∑ i = 0 m ∑ σ ∈ C i m ( − 1 ) m − i ∏ σ j ∈ σ ( b σ j ( x ) − λ σ j ) × ∫ ( ℝ d ) m ∏ σ k ∈ σ ′ ( b σ k ( y ) − λ σ k )   × K ( x , y 1 , ⋯ , y m ) f 1 ( y 1 ) ⋯ f m ( y m ) d μ ( y 1 ) ⋯ d μ ( y m ) . (20)

By expanding b σ k ( y ) − λ σ k = ( b σ k ( y ) − b σ k ( x ) ) − ( b σ k ( x ) − λ σ k ) and

∏ σ k ∈ σ ′ ( b σ k ( y ) − λ σ k ) = ∑ σ ( 1 ) ∪ σ ( 2 ) = σ ′ σ ( 1 ) ∩ σ ( 2 ) = ∅   ∏ σ u ∈ σ ( 1 ) ( b σ u ( y ) − b σ u ( x ) ) ∏ σ v ∈ σ ( 2 ) ( b σ v ( x ) − λ σ v ) ,

hence we can obtain from the equality (20) that

( T m ) ∏ b → ( f → ) ( x ) = ∏ j = 1 m ( b j ( x ) − λ j ) T m ( f → ) ( x ) + ( − 1 ) m T m ( ( b → − λ → ) f → ) ( x )         + ∑ i = 1 m − 1 ∑ σ ∈ C i m C m , i ∏ σ j ∈ σ ( b σ j ( x ) − λ σ j ) × ( T m ) ∏ b → σ ′ ( f → ) ( x ) (21)

where C m , i are constants depending only on m and i.

Proof of Lemma 2.5. For simplicity, we denote by R ( b → , f → ) ( x ) the quantities on the right hand side of the inequality (19). Recall the definition of the sharp maximal operator M # , and use the standard technique, see [15] for example, we only need to prove that

1 μ ( 3 2 Q ) ∫ Q | ( T m ) ∏ b → ( f → ) ( z ) − h Q | d μ ( z ) ≤ C R ( b → , f → ) ( x ) (22)

and

| h Q − h R | ≤ C ( K Q , R ) 2 m R ( b → , f → ) ( x ) (23)

with the absolute constant C independent of b → , f → , Q and R, where R is any doubling cube with Q ⊂ R . In fact, we take

h Q = ( − 1 ) m m Q ( T m ( ( m Q ˜ ( b 1 ) − b 1 ) f 1 χ ℝ d \ 4 3 Q , ⋯ , ( m Q ˜ ( b m ) − b m ) f m χ ℝ d \ 4 3 Q ) ) (24)

and clearly

h R = ( − 1 ) m m R ( T m ( ( m R ˜ ( b 1 ) − b 1 ) f 1 χ ℝ d \ 4 3 R , ⋯ , ( m R ˜ ( b m ) − b m ) f m χ ℝ d \ 4 3 R ) ) (25)

Recall the equality (21), for any z ∈ Q , we have that

| ( T m ) ∏ b → ( f → ) ( z ) − h Q | ≤ | ∏ j = 1 m ( b j ( x ) − m Q ˜ ( b j ) ) T m ( f → ) ( z ) |         + | ∑ i = 1 m − 1 ∑ σ ∈ C i m C i , m ∏ σ j ∈ σ ( b σ j ( z ) − m Q ˜ ( b σ j ) ) ( T m ) ∏ b → σ ′ ( f → ) ( z ) |         + | ( − 1 ) m T m ( ( b → − m Q ˜ ( b → ) ) f → ) ( z ) − h Q | = : I ( z ) + I I ( z ) + I I I ( z ) . (26)

In order to show the inequality (22), we will calculate the integrals for the three functions above, respectively. Firstly, for τ > 1 , by the Hölder inequality one sees that

1 μ ( 3 2 Q ) ∫ Q | I ( z ) | d μ ( z ) ≤ ∏ i = 1 m ( 1 μ ( 3 2 Q ) ∫ Q | b i ( z ) − m Q ˜ ( b i ) | τ i d μ ( z ) ) 1 τ i ( 1 μ ( 3 2 Q ) ∫ Q | T m ( f → ) | τ d μ ( z ) ) 1 τ ≤ C ‖ b 1 ‖ ∗ ‖ b 2 ‖ ∗ ⋯ ‖ b m ‖ ∗ M τ , 3 2 ( T m ( f → ) ) ( x ) (27)

where we have choose τ i > 1 such that 1 τ 1 + 1 τ 2 + ⋯ + 1 τ m + 1 τ = 1 .

Similarly, for τ > 1 , by the Hölder inequality, we also deduce that

1 μ ( 3 2 Q ) ∫ Q | I I ( z ) | d μ ( z ) ≤ C ∑ i = 1 m − 1 ∑ σ ∈ C i m ∏ σ j ∈ σ ‖ b σ j ‖ ∗ M τ , 3 2 ( ( T m ) b σ ′ ( f → ) ) ( x ) . (28)

To estimate the integral related to the function I I I ( z ) , we split f i as f i = f i 0 + f i ∞ , where f i 0 = f χ 4 3 Q and f i ∞ = f i − f i 0 , this yields

∏ j = 1 m f j ( y j ) = ∑ α 1 , α 2 , ⋯ , α m ∈ { 0 , ∞ } f 1 α 1 ( y 1 ) ⋯ f m α m ( y m ) = ∏ j = 1 m f j 0 ( y j ) + ∑ ∗ f 1 α 1 ( y 1 ) ⋯ f m α m ( y m ) + ∏ j = 1 m f j ∞ ( y j )     , (29)

where each term in ∑ ∗ satisfies that α j 1 = α j 2 = ⋯ = α j λ = 0 , for some 1 ≤ λ < m and some { j 1 , j 2 , ⋯ , j λ } ⊂ { 1,2, ⋯ , m } . So we can decompose the function I I I ( z ) further into three parts as follows

I I I ( z ) ≤ | T m ( ( b 1 − m Q ˜ ( b 1 ) ) f 1 0 , ⋯ , ( b m − m Q ˜ ( b m ) ) f m 0 ) ( z ) |     + ∑ ∗ | T m ( ( b 1 − m Q ˜ ( b 1 ) ) f 1 α 1 , ⋯ , ( b m − m Q ˜ ( b m ) ) f 1 α m ) ( z ) |     + | ( − 1 ) m T m ( ( b 1 − m Q ˜ ( b 1 ) ) f 1 ∞ , ⋯ , ( b m − m Q ˜ ( b m ) ) f 1 ∞ ) ( z ) − h Q | = : I I I 1 ( z ) + I I I 2 ( z ) + I I I 3 ( z ) . (30)

For s i > 1 ,we can take 1 < μ i < s i such that 1 v = 1 u 1 + 1 u 2 + ⋯ + 1 u m and v > 1 . Let 1 u i = 1 s i + 1 t i for each i = 1 , 2 , ⋯ , m , then 1 < t i < ∞ . Using Lemma 2.4, we know that the T m is bounded from L u 1 ( μ ) × L u 2 ( μ ) × ⋯ × L u m ( μ ) to L v ( μ ) . Hence, by this boundedness and Hölder inequality, we have

1 μ ( 3 2 Q ) ∫ Q | I I I 1 ( z ) | d μ ( z ) ≤ μ ( Q ) 1 − 1 v μ ( 3 2 Q ) ‖ T m ( ( b 1 − m Q ˜ ( b 1 ) ) f 1 0 , ⋯ , ( b m − m Q ˜ ( b m ) ) f m 0 ) ( z ) ‖ L v ( μ ) ≤ C μ ( 3 2 Q ) − 1 v ∏ i = 1 m ‖ ( b i − m Q ˜ ( b i ) ) f i 0 ‖ L u i ( μ ) ≤ C ∏ i = 1 m ‖ b i ‖ ∗ M s i , 9 8 f i ( x ) . (31)

In order to estimate the integral of terms I I I 2 ( z ) and I I I 3 ( z ) over Q, we will give their point-wise estimates. In fact, for z ∈ Q , since 1 ≤ λ ≤ m − 1 we observe that

I I I 2 ( z ) = ∑ ∗ | T m ( ( b 1 − m Q ˜ ( b 1 ) ) f 1 α 1 , ⋯ , ( b m − m Q ˜ ( b m ) ) f 1 α m ) ( z ) | ≤ C ∑ ∗ ∏ j ∈ { j 1 , ⋯ , j λ } ∫ 4 3 Q | b j ( y j ) − m Q ˜ ( b j ) | | f j ( y j ) | d μ ( y j )     × ∫ ( ℝ d \ 4 3 Q ) m − λ ∑ j ∉ { j 1 , ⋯ , j λ } | b j ( y j ) − m Q ˜ ( b j ) | | f j ( y j ) | d μ ( y j ) ( ∑ j ∉ { j 1 , ⋯ , j λ } | z − y j | ) m n ≤ C ∑ ∗ ∏ j ∈ { j 1 , j 2 , ⋯ , j λ } ‖ b j ‖ ∗ M s j , 9 8 f j ( x )     × ∑ k = 1 ∞     2 − k λ n ∏ j ∉ { j 1 , j 2 , ⋯ , j λ } ‖ b j ‖ ∗ ( k + 1 ) M s j , 9 8 f j ( x ) ≤ C ∏ j = 1 ∞ ‖ b j ‖ ∗ M s j , 9 8 f j ( x ) (32)

where we have used the fact (see [5] ) that, there is an absolute constant C such that, for any b ∈ R B M O , integer k ≥ 0 and cubes Q,

| m 2 k 4 3 Q ˜ ( b ) − m Q ˜ ( b ) | ≤ C ‖ b ‖ ∗ K Q ˜ , 2 k 4 3 Q ˜ ≤ C ‖ b ‖ ∗ K Q ,2 k 4 3 Q ≤ C k ‖ b ‖ ∗ . (33)

On the other hand, for I I I 3 ( z ) , we note for any z , y ∈ Q that

| T m ( ( b 1 − m Q ˜ ( b 1 ) ) f 1 ∞ , ⋯ , ( b m − m Q ˜ ( b m ) ) f m ∞ ) ( z )   − T m ( ( b 1 − m Q ˜ ( b 1 ) ) f 1 ∞ , ⋯ , ( b m − m Q ˜ ( b m ) ) f m ∞ ) ( y ) | ≤ ∫ ( ℝ d \ 4 3 Q ) m | K ( z , y 1 , ⋯ , y m ) − K ( y , y 1 , ⋯ , y m ) |     × | ∏ i = 1 m ( b i ( y i ) − m Q ˜ ( b i ) ) f i ∞ | d μ ( y 1 ) d μ ( y 2 ) ⋯ d μ (ym)

≤ ∫ ( ℝ d \ 4 3 Q ) m | z − y | ϵ ∏ i = 1 m | ( b i ( y i ) − m Q ˜ ( b i ) ) f i ∞ | ( ∑ k . l = 0 m | y k − y l | ) m n + ϵ d μ ( y 1 ) ⋯ d μ ( y m ) ≤ C ∏ i = 1 m ∑ k = 1 ∞ ∫ 2 k 4 3 Q \ 2 k − 1 4 3 Q l ( Q ) ϵ m l ( 2 k 3 2 Q ) n + ϵ m | ( b i ( y i ) − m Q ˜ ( b i ) ) f i | d μ ( y i ) ≤ C ∏ i = 1 m ∑ k = 1 ∞     2 − k m ( k + 1 ) ‖ b i ‖ ∗ M s i , 9 8 f i ( x ) ≤ C ∏ i = 1 m ‖ b i ‖ ∗ M s i , 9 8 f i ( x ) , (34)

where we have use the inequation (33) again.

Taking the mean over y ∈ Q , we can obtain that

1 μ ( 3 2 Q ) ∫ Q ( | I I I 2 ( z ) | + | I I I 3 ( z ) | ) d μ ( z ) ≤ C ∏ i = 1 m ‖ b i ‖ ∗ M s i , 9 8 f i ( x ) . (35)

Combing the inequalities (26) (27), (28), (30), (31) and (33), we see from the estimates of I , I I , I I I 1 , I I I 2 and I I I 3 that the desired inequality (22) holds.

Next we turn to estimate the inequality (23). For any cubes Q ⊂ R with x ∈ Q , where R is doubling. We denote N Q , R + 2 by N, then 2 N Q ⊃ 2 Q and 2 N Q ⊃ 2 R . We recall the equality (29) and let f i 0 = f i χ 2 N Q \ 4 3 Q and f i R = f i χ 2 N Q \ 4 3 R , and let f i ∞ = f i χ ℝ d \ 2 N Q . Then we can write

| h Q − h R | = | m Q [ T m ( ( m Q ˜ ( b 1 ) − b 1 ) f 1 χ ℝ d \ 4 3 Q , ⋯ , ( m Q ˜ ( b m ) − b m ) f m χ ℝ d \ 4 3 Q ) ]     − m R [ T m ( ( m R ˜ ( b 1 ) − b 1 ) f 1 χ ℝ d \ 4 3 R , ⋯ , ( m R ˜ ( b m ) − b m ) f m χ ℝ d \ 4 3 R ) ] | ≤ | m Q [ T m ( ( m Q ˜ ( b 1 ) − b 1 ) f 1 ∞ , ⋯ , ( m Q ˜ ( b m ) − b m ) f m ∞ ) ]     − m R [ T m ( ( m Q ˜ ( b 1 ) − b 1 ) f 1 ∞ , ⋯ , ( m Q ˜ ( b m ) − b m ) f m ∞ ) ] |

    + | m R [ T m ( ( m Q ˜ ( b 1 ) − b 1 ) f 1 ∞ , ⋯ , ( m Q ˜ ( b m ) − b m ) f m ∞ ) ]     − m R [ T m ( ( m R ( b 1 ) − b 1 ) f 1 ∞ , ⋯ , ( m R ( b m ) − b m ) f m ∞ ) ] |     + ∑ α 1 , α 2 , ⋯ , α m ∈ { 0 , ∞ } at   least   one   α i ≠ ∞ | m Q [ T m ( ( m Q ˜ ( b 1 ) − b 1 ) f 1 α 1 , ⋯ , ( m Q ˜ ( b m ) − b m ) f m α m ) ] |     + ∑ α 1 , α 2 , ⋯ , α m ∈ { R , ∞ } at   least   one   α i ≠ ∞ | m R [ T m ( ( m R ( b 1 ) − b 1 ) f 1 α 1 , ⋯ , ( m R ( b m ) − b m ) f m α m ) ] | = : A 1 + A 2 + A 3 + A 4 . (36)

For the term A 1 , noting

| m R ( b i ) − m Q ˜ ( b i ) | ≤ C ‖ b i ‖ ∗ K Q , R (37)

and the similar argument as that for the estimate of I I I 3 , we can obtain that

A 1 ≤ C ∏ i = 1 m K Q , R ‖ b i ‖ ∗ M s i , 9 8 f i ( x ) ≤ C ( K Q , R ) m ∏ i = 1 m ‖ b i ‖ ∗ M s i , 9 8 f i ( x ) . (38)

To estimate A 2 , we recall the notations and note that, for any sequences ξ j and ζ j ,

∏ j = 1 m ( ξ j + ζ j ) = ∑ i = 0 m ∑ σ ∈ C i m ∏ j ∈ σ ξ j ∏ j ′ ∈ σ ′ ζ j ′ . (39)

Using this equality and expanding m Q ˜ ( b j ) − b j ( y ) = [ m Q ˜ ( b j ) − b j ( z ) ] + [ b j ( z ) − b j ( y ) ] , we observe that

T m ( ( m Q ˜ ( b 1 ) − b 1 ) f 1 ∞ , ⋯ , ( m Q ˜ ( b m ) − b m ) f m ∞ ) ( z ) = ∑ i = 0 m ∑ σ ∈ C i m ∏ j ∈ σ ( m Q ˜ ( b j ) − b j ( z ) ) ( T m ) ∏ b → σ ′ ( f → χ ℝ d \ 2 N Q ) ( z ) (40)

Similarly,

T m ( ( m R ( b 1 ) − b 1 ) f 1 ∞ , ⋯ , ( m R ( b m ) − b m ) f m ∞ ) ( z ) = ∑ i = 0 m ∑ σ ∈ C i m ∏ j ∈ σ ( m R ( b j ) − b j ( z ) ) ( T m ) ∏ b → σ ′ ( f → χ ℝ d \ 2 N Q ) ( z ) . (41)

Thus

| T m ( ( m Q ˜ ( b 1 ) − b 1 ) f 1 ∞ , ⋯ , ( m Q ˜ ( b m ) − b m ) f m ∞ ) ( z )   − T m ( ( m R ( b 1 ) − b 1 ) f 1 ∞ , ⋯ , ( m R ( b m ) − b m ) f m ∞ ) ( z ) | ≤ | ∑ i = 1 m ∑ σ ∈ C i m ∏ j ∈ σ ( m Q ˜ ( b j ) − b j ( z ) ) ( T m ) ∏ b → σ ′ ( f → χ ℝ d \ 2 N Q ) ( z ) |     + | ∑ i = 1 m ∑ σ ∈ C i m ∏ j ∈ σ ( m R ( b j ) − b j ( z ) ) ( T m ) ∏ b → σ ′ ( f → χ ℝ d \ 2 N Q ) ( z ) | = : B 1 ( z ) + B 2 ( z ) . (42)

To estimate the integrals above, we recall that f i ∞ = f i χ ℝ d \ 2 N Q and let f j Q = f j χ 2 N Q , then we can write that f j ∞ = f j − f j Q and f j = f j ∞ + f j Q , and thus we have

∏ j = 1 m f j ∞ ( y j ) = ∏ j = 1 m f j ( y j ) + ∑ i = 1 m ( − 1 ) i ∑ ρ ∈ C i m ∏ j ∈ ρ f j Q ( y j ) ∏ j ′ ∈ ρ ′ f j ′ ( y j ′ ) = ∏ j = 1 m f j ( y j ) + ∑ λ = 1 m   ∑ { j 1 , j 2 , ⋯ , j λ } ⊂ { 1 , 2 , ⋯ , m }     × C j 1 , j 2 , ⋯ , j m f j 1 Q ( y j 1 ) , ⋯ , f j λ Q ( y j λ ) f j λ + 1 ∞ ( y j λ + 1 ) , ⋯ , f j m ∞ ( y j m ) (43)

where C j 1 , j 2 , ⋯ , j m are constant independent of f → and Q. From the equality (43), we can deduce that

( T m ) ∏ b → σ ′ ( f → χ ℝ d \ 2 N Q ) ( z ) = ( T m ) ∏ b → σ ′ f → ( z ) + ∑ λ = 1 m   ∑ { j 1 , j 2 , ⋯ , j m } ⊂ { 1 , 2 , ⋯ , m } C j 1 , j 2 , ⋯ , j m T m ( g 1 F 1 , g 2 F 2 , ⋯ , g m F m ) ( z ) (44)

where

g j ( y ) = { b j ( z ) − b j ( y ) ,   if   j ∈ σ ′ 1 ,   if   j ∈ σ

and

F j ( y ) = { f j Q ( y ) ,   if   j ∈ { j 1 , ⋯ , j λ } f j ∞ ( y ) ,   if   j ∉ { j 1 , ⋯ , j λ }

Along the same lines as that of the pointwise estimates of I I I 2 ( z ) ,we can obtain that, for x , z ∈ R ⊂ 2 N − 1 Q and if 1 ≤ λ ≤ m ,

| T m ( g 1 F 1 , g 2 F 2 , ⋯ , g m F m ) ( z ) | ≤ C ∏ j ∈ { j 1 , j 2 , ⋯ , j λ } 1 l ( 2 N Q ) n ∫ 2 N Q | g j ( y j ) f j ( y j ) | d μ ( y j )     × ∑ k = 1 ∞     2 − k λ n ∏ j ∉ { j 1 , j 2 , ⋯ , j λ } 1 ≤ j ≤ m 1 l ( 2 k 2 N Q ) n ∫ 2 k 2 N Q | g j ( y j ) f j ( y j ) | d μ ( y j ) . (45)

Let ( g j ) k , s ′ j * = 1 if j ∈ σ ; and

( g j ) k , s ′ j * = { ( 1 l ( 2 N Q ) n ∫ 2 N Q | b j ( z ) − b j ( y ) | s ′ j d μ ( y ) ) 1 s ′ j ,   if   j ∈ σ ′ ∩ { j 1 , j 2 , ⋯ , j λ } ( 1 l ( 2 k + N Q ) n ∫ 2 k + N Q | b j ( z ) − b j ( y ) | s ′ j d μ ( y ) ) 1 s ′ j ,   if   j ∈ σ ′ \ { j 1 , j 2 , ⋯ , j λ }

thus for j ∈ σ ′ ,

( g j ) k , s ′ j * ≤ C ‖ b j ‖ ∗ + C | m 2 N Q ( b j ) − b j ( z ) | + C | m 2 k + N Q ( b j ) − b j ( z ) | .

Hence we get from (45) that, for τ > 1 ,

( 1 μ ( R ) ∫ R | T m ( g 1 F 1 , g 2 F 2 , ⋯ , g m F m ) ( z ) | τ d μ ( z ) ) 1 τ ≤ C ( ∏ j = 1 m M s j , 9 8 f j ( x ) ) ∑ j = 1 ∞ 2 − k λ n × ( 1 μ ( R ) ∫ R | ∏ j ∈ σ ′ ( g j ) k , s ′ j * | τ d μ ( z ) ) 1 τ ≤ C ( ∏ j = 1 m M s j , 9 8 f j ( x ) ) ∑ j = 1 ∞     2 − k λ n ∏ j ∈ σ ′ ‖ b j ‖ ∗ ( 1 + K R , 2 N Q + K R , 2 k + N Q ) ≤ C ( ∏ j = 1 m M s j , 9 8 f j ( x ) ) ∏ j ∈ σ ′ ‖ b j ‖ ∗ , (46)

where we have used the fact that the cubes R and 2 N Q are comparable, which implies K R ,2 N Q ≤ C and K R ,2 k + N Q ≤ C ( 1 + k ) . Using the inequality (46) above and the identity (44), we obtain that, for τ > 1 ,

A 2 ≤ 1 μ ( R ) ∫ R ( | B 1 ( z ) | + | B 2 ( z ) | ) d μ ( z ) ≤ C ∑ i = 1 m ∑ σ ∈ C i m ( ∏ j ∈ σ K Q , R ‖ b j ‖ ∗ )     × ( 1 μ ( R ) ∫ R | ( T m ) ∏ b → σ ′ ( f → χ ℝ d \ 2 N Q ) | τ d μ ( z ) ) 1 τ ≤ C ∑ i = 1 m ∑ σ ∈ C i m ( ∏ j ∈ σ K Q , R ‖ b j ‖ ∗ ) M τ , 3 2 ( ( T m ) ∏ b → σ ′ ( f → ) ) ( x )     + C ∏ j = 1 m   K Q , R ‖ b j ‖ ∗ M s j , 9 8 f j ( x ) .

The estimates of A 3 and A 4 is very similar to the one used in the estimate of A 2 . In fact, repeating the similar procedures used in (45) and (46) for τ > 1 , and noting that K Q ,2 k + N Q ≤ K Q , R + K R ,2 k + N Q ≤ C ( 1 + k ) + K Q , R since 2 N − 3 Q ⊂ 2 R ⊂ 2 N Q by the definition of N, we can deduce that

∑ α 1 , α 2 , ⋯ , α m ∈ { 0 , ∞ } at   least   one   α i ≠ ∞ and   one   α j ≠ 0 | m Q [ T m ( ( m Q ˜ ( b 1 ) − b 1 ) f 1 α 1 , ⋯ , ( m Q ˜ ( b m ) − b m ) f m α m ) ] | + ∑ α 1 , α 2 , ⋯ , α m ∈ { R , ∞ } at   least   one   α i ≠ ∞ | m R [ T m ( ( m R ( b 1 ) − b 1 ) f 1 α 1 , ⋯ , ( m R ( b m ) − b m ) f m α m ) ] | ≤ C ∏ j = 1 m   K Q , R ‖ b j ‖ ∗ M s j , 9 8 f j ( x ) . (47)

It is left to estimate the term in A 3 of the case α 1 = α 2 = ⋯ = α m = 0 . A small modification is needed to estimate this term. For z ∈ Q and x ∈ Q , one sees

| T m ( ( m Q ˜ ( b 1 ) − b 1 ) f 1 0 , ⋯ , ( m Q ˜ ( b j ) − b j ) f j 0 , ⋯ , ( m Q ˜ ( b m ) − b m ) f m 0 ) ( z ) | ≤ C ∑ k = 1 N ∏ j = 1 m 1 l ( 2 k 4 3 Q ) n ∫ 2 k 4 3 Q | ( m Q ˜ ( b j ) − b j ( y j ) ) f j ( y j ) | d μ ( y j ) ≤ C ∑ k = 1 N ∏ j = 1 m μ ( 2 k 3 2 Q ) l ( 2 k 4 3 Q ) n ( 1 + K Q , 2 k 4 3 Q ) ‖ b j ‖ ∗ M s j , 9 8 f j ( x ) ≤ C ( K Q , R ) 2 m ∏ j = 1 m ‖ b j ‖ ∗ M s j , 9 8 f j ( x ) .

This and the inequality (47) follows

A 3 + A 4 ≤ C ( K Q , R ) 2 m ∏ j = 1 m ‖ b j ‖ ∗ M s j , 9 8 f j ( x ) .

Moreover, combing the estimates of A 1 , A 2 , A 3 and A 4 , we obtain the desired inequality (23).

Finally, let us show how to acquire the inequality (19) from the two inequalities (22) and (23). Fix the point x and let Q be any cube that x ∈ Q . notice K Q , Q ˜ ≤ C , hence we see from the inequalities (22) and (23) that

1 μ ( 3 2 Q ) ∫ Q | ( T m ) ∏ b → ( f → ) ( z ) − m Q ˜ ( ( T m ) ∏ b → ( f → ) ) | d μ ( z ) ≤ 1 μ ( 3 2 Q ) ∫ Q | ( T m ) ∏ b → ( f → ) ( z ) − h Q | d μ ( z ) + 1 μ ( 3 2 Q ) ∫ Q | h Q − h Q ˜ | d μ ( z ) + 1 μ ( 3 2 Q ) ∫ Q | m Q ˜ ( ( T m ) ∏ b → ( f → ) ) − h Q ˜ | d μ ( z ) ≤ C R ( b → , f → ) ( x ) . (48)

On the other hand for all doubling cubes Q ⊂ R with x ∈ Q such that K Q , R ≤ P 0 , where P 0 is the constant in Lemma 6 in [15], using (23), we have

| h Q − h R | ≤ C K Q , R P 0 2 m − 1 R ( b → , f → ) ( x ) (49)

and moreover the inequality (49) holds for any doubling cubes Q, R with Q ⊂ R . Therefore,

m Q ( ( T m ) ∏ b → ( f → ) ) − m R ( ( T m ) ∏ b → ( f → ) ) | ≤ | m Q ( ( T m ) ∏ b → ( f → ) ) − h Q | + | h R − m R ( ( T m ) ∏ b → ( f → ) ) | + | h Q − h R | ≤ C K Q , R R ( b → , f → ) ( x ) . (50)

According to the estimates (48) (50) and the definition of the sharp maximal function, we deduce the inequality (19) and so finish the proof of the Lemma 2.5.☐

3. Conclusions

The proof of Lemma 2.5 can be slightly modified to prove the conclusion of Lemma 2.1. Therefore we show that the iterated commutators generated by multilinear singular integrals operators ( T m ) ∏ b → are bounded from

M q 1 p 1 ( μ ) × ⋯ × M q m p m ( μ ) to M q p ( μ ) . Suppose ‖ μ ‖ = ∞ , the detailed conclusion can be described as follows: Let ( T m ) ∏ b → as in (9) and satisfying conditions (7)

and (8). Let 1 < q j ≤ p j < ∞ with 1 p = 1 p 1 + ⋯ + 1 p m and 1 q = 1 q 1 + ⋯ + 1 q m . Suppose b i ∈ R B M O ( μ ) for i = 1 , 2 , ⋯ , m . If T m maps L 1 ( μ ) × ⋯ × L 1 ( μ ) to L 1 / m , ∞ ( μ ) , then the commutators ( T m ) ∏ b → are bounded from M q 1 p 1 ( μ ) × ⋯ × M q m p m ( μ ) to M q p ( μ ) , that is,

‖ ( T m ) ∏ b → ( f → ) ( x ) ‖ M q p ( μ ) ≤ C ∏ i = 1 m ‖ b i ‖ ∗ ‖ f i ‖ M q i p i ( μ ) .

More generally, let ( T m ) ∏ b → σ as in (11) and satisfying conditions (7) and (8). Let 1 < q j ≤ p j < ∞ with 1 p = 1 p 1 + ⋯ + 1 p m and 1 q = 1 q 1 + ⋯ + 1 q m . Suppose b i ∈ R B M O ( μ ) for i = 1 , 2 , ⋯ , m . If T m maps L 1 ( μ ) × ⋯ × L 1 ( μ ) to L 1 / m , ∞ ( μ ) , then for all σ ⊂ { 1,2, ⋯ , m } , the commutators ( T m ) ∏ b → σ are bounded from M q 1 p 1 ( μ ) × ⋯ × M q m p m ( μ ) to M q p ( μ ) , that is,

‖ ( T m ) ∏ b → σ ( f → ) ( x ) ‖ M q p ( μ ) ≤ C ∏ j ∈ σ ‖ b j ‖ ∗ ∏ i = 1 m ‖ f i ‖ M q i p i ( μ ) .

Funding

Supported by the National Natural Science Foundation of China (Grant No. 11161044 and Grant No. 11661061) and Natural science foundation of Inner Mongolia (No. 2019MS01003).

Conflicts of Interest

The authors declare no conflicts of interest regarding the publication of this paper.

References

1. Nazarov, F., Treil, S. and Volberg, A. (1997) Cauchy Integral and Calderón-Zygmund Operators on Nonhomogeneous Spaces. International Mathematics Research Notices, 15, 703-726. https://doi.org/10.1155/S1073792897000469

2. Nazarov, F., Treil, S. and Volberg, A. (1998) Weak Type Estimates and Cotlar Inequalities for Calderón-Zygmund Operators on Nonhomogeneous Spaces. International Mathematics Research Notices, 1998, 463-487. https://doi.org/10.1155/S1073792898000312

3. Tolsa, X. (2003) Painlevé’s Problem and the Semiadditivity of Analytic Capacity. Acta Mathematica, 190, 105-149. https://doi.org/10.1007/BF02393237

4. Tolsa, X. (2003) Bilipschitz Maps, Analytic Capacity, and the Cauchy Integral. Annals of Mathematics, 190, 1243-1304. https://doi.org/10.4007/annals.2005.162.1243

5. Tolsa, X. (2001) BMO, H1, and Calderón-Zygmund Operators for Non Doubling Measures. Mathematische Annalen, 319, 89-149. https://doi.org/10.1007/PL00004432

6. Tolsa, X. (2003) The Space H1 for Nondoubling Measures in Terms of a Grand Maximal Operator. Transactions of the American Mathematical Society, 355, 315-348. https://doi.org/10.1090/S0002-9947-02-03131-8

7. Tolsa, X. (2001) Littlewood-Paley Theory and the T(1) Theorem with Non-Doubling Measures. Advances in Mathematics, 164, 57-116.https://doi.org/10.1006/aima.2001.2011

8. Adams, D.R. (1975) A Note on Riesz Potentials. Duke Mathematical Journal, 42, 765-778. https://doi.org/10.1215/S0012-7094-75-04265-9

9. Chiarenza, F. and Frasca, M. (1987) Morrey Spaces and Hardy-Littlewood Maximal Function. Rendiconti del Seminario Matematico della Università di Padova, 7, 273-279.

10. Sawano, Y. and Tanaka, H. (2005) Morrey Spaces for Non-Doubling Measures. Acta Mathematica Sinica, 21, 1535-1544. https://doi.org/10.1007/s10114-005-0660-z

11. Sawano, Y. (2008) Generalized Morrey Spaces for Non-Doubling Measures. Nonlinear Differential Equations and Applications NoDEA, 15, 413-425. https://doi.org/10.1007/s00030-008-6032-5

12. Coifman, R., Rochberg, R. and Weiss, G. (1976) Factorization Theorems for Hardy Spaces in Several Variables. Annals of Mathematics, 103, 611-635.https://doi.org/10.2307/1970954

13. Pérez, C. and Trujillo-Gonzalez, R. (2002) Sharp Weighted Estimates for Multilinear Commutators. Journal of the London Mathematical Society, 65, 672-692.https://doi.org/10.1112/S0024610702003174

14. Hu, G., Meng, Y. and Yang, D. (2005) Multilinear Commutators of Singular Integrals with Non-Doubling Measures. Integral Equations and Operator Theory, 51, 235-255. https://doi.org/10.1007/s00020-003-1251-y

15. Chen, W. and Sawyer, E. (2002) A Note on Commutators of Fractional Integral with Function. Illinois Journal of Mathematics, 46, 1287-1298.https://doi.org/10.1215/ijm/1258138480

16. Grafakos, L. and Torres, R. (2002) Multilinear Calderón-Zygmund Theory. Advances in Mathematics, 165, 124-164. https://doi.org/10.1006/aima.2001.2028

17. Grafakos, L. and Torres, R. (2000) On Multilinear Singular Integrals of Calderoń-Zygmund Type. Proceedings of the 6th International Conference on Harmonic Analysis and Partial Differential Equations, 57-91.https://doi.org/10.5565/PUBLMAT_Esco02_04

18. Grafakos, L. and Torres, R. (2002) Maximal Operator and Weighted Norm Inequalities for Multilinear Singular Integrals. Indiana University Mathematics Journal, 51, 1261-1276. https://doi.org/10.1512/iumj.2002.51.2114

19. Kenig, C. and Stein, E. (1999) Multilinear Estimates and Fractional Integration. Mathematical Research Letters, 6, 1-15. https://doi.org/10.4310/MRL.1999.v6.n1.a1

20. Pérez, C. and Torres, R. (2003) Sharp Maximal Function Estimates for Multilinear Singular Integrals. In: Harmonic Analysis at Mount Holyoke, Contemporary Mathematics, AMS, Providence, RI, 323-331. https://doi.org/10.1090/conm/320/05615

21. Pérez, C. and Pradolini, G. (2011) End-Point Estimates for Iterated Commutators of Multilinears of Multilinear Singular Integrals. Classical Analysis and ODEs, arXiv: 1004.4976v2.

22. Xu, J. (2007) Boundedness in Lebesgue Spaces for Commutators of Multilinear Singular Integrals and RBMO Functions with Non-Doubling Measures. Science in China Series A: Mathematics, 50, 361-376. https://doi.org/10.1007/s11425-007-2073-1

23. Tao, X. and He, S. (2013) The Boundedness of Multilinear Operators on Generalized Morrey Spaces over the Quasi-Metric Space of Non-Homogeneous Type. Journal of Inequalities and Applications, 2013, Article No. 330.https://doi.org/10.1186/1029-242X-2013-330

24. Sawano, Y. and Tanaka, H. (2007) Sharp Maximal Inequality and Commutators on Morrey Spaces with Non-Doubling Measures. Taiwanese Journal of Mathematics, 11, 1091-1112. https://doi.org/10.11650/twjm/1500404805

25. Xu, J. (2007) Boundedness of Multilinear Singular Integrals for Non-Doubling Measures. Journal of Mathematical Analysis and Applications, 327, 471-480.https://doi.org/10.1016/j.jmaa.2006.04.049