Principale 10.0000/cms.math.ca/ca7e3530b54982da75d0

# 10.0000/cms.math.ca/ca7e3530b54982da75d0

Quanto ti piace questo libro?
Qual è la qualità del file?
Scarica il libro per la valutazione della qualità
Qual è la qualità dei file scaricati?
Lingua:
english
DOI:
10.0000/cms.math.ca/ca7e3530b54982da75d0
File:
PDF, 137 KB

Puoi scrivere una recensione del libro e condividere le tue esperienze. Gli altri lettori saranno sempre interessati alla tua opinione sui libri che hai letto. Che tu abbia amato o meno il libro, se dai i tuoi pensieri onesti e dettagliati, le persone troveranno nuovi libri adatti a loro.
1

Anno:
2011
Lingua:
english
File:
PDF, 1,10 MB
2

### A Critical Evaluation of the Ontogeny of Human Sexual Behavior

Anno:
1965
Lingua:
english
File:
PDF, 802 KB
```Canad. Math. Bull. Vol. 52 (4), 2009 pp. 521–534

The Parabolic Littlewood–Paley Operator
with Hardy Space Kernels
Yanping Chen and Yong Ding
Abstract. In this paper, we give the L p boundedness for a class of parabolic Littlewood–Paley g-function with its kernel function Ω is in the Hardy space H 1 (Sn−1 ).

1 Introduction
Let Rn be the Euclidean space with the routine norm |x| for each x ∈ Rn . Denote
by Sn−1 = {x ∈ Rn : |x| = 1} the unit sphere on Rn equipped with the Lebesgue
measure σ(x ′ ). Let α1 , . . . , αn be fixed real numbers with αi ≥ 1. It is easy to see
that for fixed x ∈ Rn , the function
F(x, ρ) =

n
X
xi 2
ρ2αi
i=1

is a strictly decreasing function of ρ > 0. Therefore, there exists a unique ρ(x) such
that F(x, ρ) = 1. It was proved in  that ρ(x) is a metric on Rn . For x ∈ Rn , set
x1 = ρα1 cos ϕ1 · · · cos ϕn−2 cos ϕn−1
x2 = ρα2 cos ϕ1 · · · cos ϕn−2 sin ϕn−1
..
.
xn−1 = ραn−1 cos ϕ1 sin ϕ2
xn = ραn sin ϕ1 .
′
′
α−1
Then dx = ρα−1 J(x
J(x ′ ) is the Jacobian of the above transPn)dρdσ(x ), and′ ρ
form, where α = i=1 αi and J(x ) = α1 x1′2 + · · · + αn xn′2 . It is easy to see that
J(x ′ ) ∈ C ∞ (Sn−1 ) with 1 ≤ J(x ′ ) ≤ M for some M ≥ 1. Without loss of generality,
we may assume αn ≥ αn−1 ≥ · · · ≥ α1 ≥ 1.
For t > 0, let At = diag[t α1 , . . . , t αn ]. Suppose that Ω(x) is a real valued and
measurable function defined on Rn . We say Ω(x) is homogeneous of degree zero
with respect to At , if for any t > 0 and x ∈ Rn

(1.1)

Ω(At x) = Ω(x).

Received by the editors October 5, 2006; revised February 19, 2009.
The research was supported by NSF of China (Grants: 10571015, 10826046) and RFDP of China
(Grant: 20050027025).
AMS subject classification: Primary: 42B20; secondary: 42B25.
Keywords: parabolic Littlewood-Paley operator, Hardy space, rough kernel.

521

522

Y. Chen and Y. Ding
Moreover, we also assume that Ω(x) satisfies the following cancellation condition:
(1.2)

Z

Ω(x ′ ) J(x ′ )dσ(x ′ ) = 0.

Sn−1

In 1966, Fabes and Rivière [7; ] proved that if Ω ∈ C 1 (Sn−1 ) satisfies (1.1) and (1.2),
then the parabolic singular integral operator TΩ is bounded on L p (Rn ) for 1 <
p < ∞, where TΩ is defined by
TΩ f (x) = p.v.

Z

Rn

Ω(y)
f (x − y) dy.
ρ(y)α

In 1976, Nagel, Rivière and Wainger  improved the above result. They showed TΩ
is still bounded on L p (Rn ) for 1 < p < ∞ if replacing Ω ∈ C 1 (Sn−1 ) by a weaker
condition Ω ∈ L log+L(Sn−1 ).
On the other hand, in 1974, Madych considered the L p boundedness with respect
to the transform At of the Littlewood–Paley operator. Let ψ ∈ S(Rn ) satisfy ψ̂(0) =
0, where and below, ψ̂ denotes the Fourier transform of ψ. Let ψt (x) = t −α ψ(At −1 x)
for t > 0. Then the Littlewood–Paley operator related to At is defined by
gψ ( f )(x) =

Z

0

Theorem A

∞

|ψt ∗ f (x)|2

dt  1/2
.
t

 The Littlewood–Paley operator gψ is of type (p, p) for 1 < p < ∞.

Inspired by the works in [7–9], recently Ding, Xue and Yabuta  improved the
above result. More precisely, the authors in  proved that the parabolic Littlewood–
Paley operator is still bounded on L p if ψ(x)is replaced by a kernel function φ(x) =
Ω(x)ρ(x)−α+1 χ{ρ(x)≤1} (x) with Ω ∈ Lq (Sn−1 ) (q > 1) satisfying (1.1) and (1.2).
Theorem B  If Ω ∈ Lq (Sn−1 )(q > 1) satisfies (1.1) and (1.2), then gφ is of type
(p, p) for 1 < p < ∞.
Notice that on the unit sphere Sn−1 , there are the following containing relationships:
C ∞ \$ Lq (q > 1) \$ L log+ L \$ H 1 \$ L1 ,
where H 1 denotes the Hardy space on Sn−1 (see §2 for its definition). Hence, a natural
question is whether the size condition assumed on Ω can be weakened further. The
purpose of this paper is to give a positive answer to this question.
Theorem 1.1
1 < p < ∞.

If Ω ∈ H 1 (Sn−1 ) satisfies (1.1) and (1.2), then gφ is of type (p, p) for

Remark. If α1 = · · · = αn = 1, then ρ(x) = |x| and α = n. In this case, gφ = µΩ
and the latter is just the classical Marcinkiewicz integral, which was studied by many
authors. (See [1, 4, 10], for example.) Moreover, note also that the Ω in Theorem 1.1
(also Theorem B) has no any smoothness on Sn−1 .

The Parabolic Littlewood–Paley Operator with Hardy Space Kernels

523

2 Definitions and Lemmas
Let us begin with the definition of Hardy space H 1 (Sn−1 ). For f ∈ L1 (Sn−1 ) and
x ′ ∈ Sn−1 , we denote
Z
+
′
P f (x ) = sup
f (y ′ )Ptx ′ (y ′ ) dσ(y ′ ) ,
0<t<1

where Ptx ′ (y ′ ) =

1−t 2
|y ′ −tx ′ |n

Sn−1

for y ′ ∈ Sn−1 . Then

H 1 (Sn−1 ) = { f ∈ L1 (Sn−1 ) : kP+ f kL1 (Sn−1 ) < ∞},
and we define k f kH1 (Sn−1 ) = kP+ f kL1 (Sn−1 ) if f ∈ H 1 (Sn−1 ).
A very useful characterization of the space H 1 (Sn−1 ) is its atomic decomposition.
Let us first recall the definition of atoms. A regular H 1 (Sn−1 ) atom is a function a(x ′ )
on L∞ (Sn−1 ) satisfying the following conditions:
(2.1)

supp(a) ⊂ Sn−1
∩ {y ∈ Rn : |y − ξ ′ | < r for some ξ ′ ∈ Sn−1 and r ∈ (0, 2]};

(2.2)

Z

a(x ′ )Y m (x ′ ) dσ(x ′ ) = 0

Sn−1

for any spherical harmonic polynomial Y m with degree m ≤ N, where N is any fixed
integer;
kakL∞ (Sn−1 ) ≤ r1−n .

(2.3)

An exceptional H 1 (Sn−1 ) atom u(x ′ ) is an L∞ (Sn−1 ) function bounded by 1.
From , we find that any Ω ∈ H 1 (Sn−1 ) has an atomic decomposition
Ω=

∞
X
j=1

λjaj +

∞
X

δi u i ,

i=1

where each a j is a regular H 1 (Sn−1 ) atom and each ui is an exceptional atom. Moreover,
∞
∞
X
X
|δi | ≤ CkΩkH1 (Sn−1 ) .
|λ j | +
i=1

j=1

We note that for any x ′ ∈ Sn−1 ,
∞
X
i=1

δi ui (x ′ ) ≤

∞
X
i=1

|δi |.

524

Y. Chen and Y. Ding
Without loss of generality, we can assume
∞
X

δi ui (x ′ ) ≤ kΩkH1 (Sn−1 ) .

i=1

Thus we write

′

with ω(x ) =
(2.4)

P∞

∞
X
i=1

i=1 δi ui (x

Ω(x ′ ) =

δi ui (x ′ ) = kΩkH1 (Sn−1 ) ω(x ′ ),

∞
X

′

)/kΩkH1 (Sn−1 ) . In this new definition, for x ′ ∈ Sn−1 ,

λ j a j (x ′ ) + kΩkH1 (Sn−1 ) ω(x ′ )

and

kωkL∞ (Sn−1 ) ≤ 1.

j=1

The following Lemmas 2.1 and 2.2 can be found in .
Lemma 2.1
(2.5)

 Suppose that n ≥ 3 and b satisfies (2.1), (2.3), and
Z
b(y ′ ) dσ(y ′ ) = 0.
Sn−1

Let
2 (n−3)/2

Fb (s) = (1 − s )

χ(−1,1) (s)

Z

Sn−2

Gb (s) = (1 − s2 )(n−3)/2 χ(−1,1) (s)

Z

Sn−2

b(s, (1 − s2 )1/2 ye)dσ(e
y ),

|b(s, (1 − s2 )1/2 ye)|dσ(e
y ).

Then there exists a constant C, independent of b, such that
(2.6)

supp(Fb ) ⊂ (ξ1′ − 2r(ξ ′ ), ξ1′ + 2r(ξ ′ )),

(2.7)

supp(Gb ) ⊂ (ξ1′ − 2r(ξ ′ ), ξ1′ + 2r(ξ ′ )),

(2.8)

kFb k∞ ≤ C/r(ξ ′ ), kGb k∞ ≤ C/r(ξ ′ ),
Z
Fb (s) ds = 0,

(2.9)

R

where r(ξ ′ ) = |ξ|−1 |Lr ξ| and Lr ξ = (r2 ξ1 , rξ2 , . . . , rξn ).
 Suppose that n = 2 and b satisfies (2.1), (2.3) and (2.5). Let

Fb (s) = (1 − s2 )−1/2 χ(−1,1) (s) b(s, (1 − s2 )1/2 ) + b(s, −(1 − s2 )1/2 ) ,

Gb (s) = (1 − s2 )−1/2 χ(−1,1) (s) |b(s, (1 − s2 )1/2 )| + |b(s, −(1 − s2 )1/2 )| .

Lemma 2.2

Then Fb (s) satisfies (2.6) and (2.9), and kFb kq ≤ C|Lr (ξ ′ )|−1+1/q . And Gb (s) satisfies
(2.7) and kGb kq ≤ C|Lr (ξ ′ )|−1+1/q for some q ∈ (1, 2).

The Parabolic Littlewood–Paley Operator with Hardy Space Kernels
Lemma 2.3

525

 For Ω ∈ L1 (Sn−1 ), denote
σ2t (x) = 2−t Ω(x)ρ(x)−α+1 χ{ρ(x)≤2t } (x),

and σ ∗ ( f )(x) = supt∈R ||σ2t | ∗ f (x)|. Then kσ2t k1 ≤ C and kσ ∗ ( f )k p ≤ Ck f k p for
1 < p < ∞, where the constant C is independent of f and t.
Lemma 2.4  Suppose that m denotes the distinct numbers of {α j }. Then for any
x, y ∈ Rn , 0 ≤ β ≤ 1
Z 2
β
dλ
e−iAλ x·y
≤ C|x · y|− m ,
λ
1
where C > 0 is independent of x and y.

3 Proof of Theorem 1.1
Since Ω ∈ H 1 (Sn−1 ) satisfies the cancellation condition (1.2), by (2.4) we can
write
∞
X
λ j a j (x ′ ) + kΩkH1 (Sn−1 ) ω(x ′ ),
Ω(x ′ ) =
j=1

1

where each a j is a regular H (Sn−1 ) atom and kωkL∞ (Sn−1 ) ≤ 1. Moreover,
∞
X

|λ j | ≤ CkΩkH1 (Sn−1 ) .

j=1

For y ∈ Rn (y 6= 0), we write
Ω(y) =

∞
X

λ j ã j (y) + kΩkH1 (Sn−1 ) ω̃(y),

j=1

where ã j (y) = a j (Aρ(y)−1 y) and ω̃(y) = ω(Aρ(y)−1 y). It is easy to check that ω̃(y ′ ) =
ω(y ′ ), ã j (y ′ ) = a j (y ′ ) for y ′ ∈ Sn−1 and ω̃ and ã j satisfy (1.1) for any t > 0 and
y ∈ Rn .
x
Noticing that J( |x|
)|x|2 is a homogeneous polynomial of degree 2 on Rn by [11,
Theorem 2.1], we can write
 x 
J
|x|2 = P2 (x) + |x|2 P0 (x),
|x|
where Pk (x) is a harmonic polynomial of degree k (k = 0, 2). Then J(x ′ ) = P2 (x ′ ) +
P0 (x ′ ), where Pk (x ′ ) is a spherical harmonic polynomial of degree k (k = 0, 2). So by
(2.2), we have
(3.1)

Z

a j (y ′ ) J(y ′ ) dσ(y ′ )

Sn−1

=

Z

Sn−1

′

′

′

a j (y )P2 (y ) dσ(y ) +

Z

Sn−1

a j (y ′ )P0 (y ′ ) dσ(y ′ ) = 0,

526

Y. Chen and Y. Ding
for all j = 1, 2, . . . . Thus by (2.4) and (3.1), we know
Z

(3.2)

ω(y ′ ) J(y ′ ) dσ(y ′ ) = 0.
Sn−1

Therefore,
(3.3)

kgφ ( f )k p ≤

∞
X

|λ j |kga j ( f )k p + kΩkH1 (Sn−1 ) kgω ( f )k p ,

j=1

where
ga j ( f )(x) =

Z

∞

ρ(y)≤t

0

gω ( f )(x) =

Z

Z

∞

0

Z

ρ(y)≤t

ã j (y)
f (x − y) dy
ρ(y)α−1
ω̃(y)
f (x − y) dy
ρ(y)α−1

2 dt  1/2

t3

2 dt  1/2

t3

,
.

Since ω(x ′ ) ∈ L∞ (Sn−1 ) and satisfies the cancellation condition (3.2), by Theorem B
we get
kgω ( f )k p ≤ Ck f k p ,

(3.4)

where C is independent of ω and f . Thus, to prove Theorem 1.1, by (3.3) and (3.4)
it suffices to show that there exists C > 0, independent of the atoms a j and f , such
that for j = 1, 2, . . . ,
kga j ( f )k p ≤ Ck f k p .

(3.5)

We only prove (3.5) for the case n > 2. The case for n = 2 can be dealt with using
the same method and Lemma 2.2. From now we denote simply a j , ã j and ga j by a, ã,
and ga , respectively. Without loss of generality, we may also assume that supp(a) is
contained in B(1, r) ∩ Sn−1 , where B(1, r) = {y : |y − 1| < r} and 1 = (1, 0, . . . , 0).
Choose a C 0∞ (Rn ) function ϕ such that Rϕ(x) = ϕ(ρ(x)), 0 ≤ ϕ ≤ 1 satisfying
∞
supp(ϕ) ⊂ {y : 1/2 ≤ ρ(y) ≤ 2} and 0 ϕ(t)/t dt = 1. Define functions Φ
b
b
and ∆ by Φ(ξ)
= ϕ(ρ(Lr ξ)) and ∆(ξ)
= ϕ(ρ(ξ)), respectively, where Lr ξ is defined in Lemma 2.1. If we denote Φt (x) = t −α Φ(At −1 x) and ∆t (x) = t −α ∆(At −1 x),
ct (ξ) = ϕ(tρ(Lr ξ)), ∆
ct (ξ) = ϕ(tρ(ξ)), and Φt (x) =
then it is easy to check that Φ
1 −α
−1
−1
t
∆(L
A
x),
where
r
t
r n+1
Lr−1 At −1 x = (r−2t −α1 x1 , r−1t −α2 x2 , . . . , r−1t −αn xn ).
For any f ∈ S(Rn ), by taking Fourier transform we have
(3.6)

f (x) =

Z

∞

−∞

Φ2t ∗ f (x)dt ∼

Z

0

∞

Φt ∗ f (x)

dt
.
t

The Parabolic Littlewood–Paley Operator with Hardy Space Kernels

527

Define
gΦ ( f )(x) =

Z

∞

|Φt ∗ f (x)|2

0

dt  1/2 
∼
t

Z

∞

|Φ2t ∗ f (x)|2 dt

−∞

 1/2

.

Now we claim that
kgΦ ( f )k p ≤ Ck f k p ,

(3.7)

with C independent of r > 0. In fact, by the definition of Φt , we have
Z
1
∆(Lr−1 At −1 y) f (x − y) dy
Φt ∗ f (x) = n+1 t −α
r
Rn
Z
= t −α
∆(At −1 y) f (Lr (Lr−1 x − y)) dy
Rn

= ∆t ∗ h(Lr−1 x),
R
b
= ϕ(0) = 0, by Theorem A we get
where h(x) = f (Lr x). Since Rn ∆(x) dx = ∆(0)
Z
 ∞
dt  1/2
kgΦ ( f )k p =
|Φt ∗ f (·)|2
t
p
0
nZ Z ∞
dt  p/2 o 1/p
=
|∆t ∗ h(Lr−1 x)|2
dx
t
Rn
0
Z Z ∞
n
dt  p/2 o 1/p
dx
= rn+1
|∆t ∗ h(x)|2
t
Rn
0
n+1

≤ Cr p khk p
Z

 1/p
n+1
=C r
= Ck f k p .
| f (Lr x)| p dx
Rn

This is (3.7). Now we denote σ2t (y) = 2−t ã(y)ρ(y)−α+1 χ{ρ(y)≤2t } (y). Then
Z ∞ Z
2 dt  1/2
ã(y)
f
(x
−
y)dy
ga ( f )(x) =
α−1
t3
0
ρ(y)≤t ρ(y)
Z ∞
 1/2
∼
|σ2t ∗ f (x)|2 dt
.
−∞

By (3.6) and the Minkowski inequality, we obtain
Z ∞ Z ∞
2  1/2
ga ( f )(x) ∼
Φ2s+t ∗ σ2t ∗ f (x)ds dt
≤

Z

−∞

∞

−∞

=:

Z

Z

−∞

∞

|σ2t ∗ Φ2s+t ∗ f (x)|2 dt

−∞

∞

−∞

Qs ( f )(x) ds.

 1/2

ds

528

Y. Chen and Y. Ding
Using Minkowski’s inequality again yields
(3.8)

kga ( f )k p ≤ C

Z

∞

kQs ( f )k p ds +

Z

0

−∞

0


kQs ( f )k p ds .

By (3.8), it is easy to see that the proof of (3.5) can be reduced to show the following
estimates
(
C2−sγ k f k p
kQs ( f )k p ≤
C2sτ k f k p

(3.9)

for s > 0,
for s < 0,

where τ and γ are some positive constants, and C is independent s and f .
The proof of (3.9) will be completed in two steps.
Step 1: There exists C > 0, independent of s and f , such that
kQs ( f )k p ≤ Ck f k p

(3.10)

for 1 < p < ∞.

First we consider the case 1 < p < 2. Denote Gs+t (x) = Φ2s+t ∗ f (x). Since a(x ′ ) ∈
L1 (Sn−1 ), by Lemma 2.3, we know kσ2t k1 ≤ C, then
Z

(3.11)

∞

σ ∗ Gs+t (·)dt
2t

−∞

1

≤C

Z

∞

Gt (·)dt

−∞

1

.

On the other hand, for 1 < q < ∞, also by Lemma 2.3, we get
(3.12)

k sup |σ2t ∗ Gs+t |kq ≤ kσ ∗ (sup |Gt |)kq ≤ Ck sup |Gt |kq .
t∈R

t∈R

t∈R

If we define TGs+t (x) = σ2t ∗Gs+t (x), then (3.11) and (3.12) show that T is a bounded
operator on L1 (L1 (R), Rn ) and Lq (L∞ (R), Rn ), respectively. Since 1 < p < 2, we can
take q > 1 such that 1/q = 2/p − 1. Then by using the operator interpolation
theorem between (3.11) and (3.12), we know that the operator T is also bounded on
L p (L2 (R), Rn). That is
Z

∞

|σ2t ∗ Gs+t (·)|2 dt

−∞

 1/2

p

≤C

Z

∞

−∞

|Gt (·)|2 dt

 1/2

p

.

From this and (3.7), we prove (3.10) for 1 < p < 2. Moreover, by (3.7) and the L2
boundedness of σ ∗ , (3.10) holds for the case p = 2. Now let us deal with the case
p > 2. Let q = (p/2) ′ . Then
kQs f k2p

= sup
ν

Z Z
Rn

∞

|σ2t ∗ Φ2s+t ∗ f (x)|2 ν(x) dt dx ,
−∞

The Parabolic Littlewood–Paley Operator with Hardy Space Kernels

529

where the supremum is taken over all ν(x) ∈ Lq (Rn ) with kνkq ≤ 1. Applying
Hölder’s inequality and noting the fact kσ2t k1 ≤ C,
Z Z ∞
|σ2t ∗ Φ2s+t ∗ f (x)|2 ν(x) dt dx
Rn

−∞

≤

Z

∞

−∞

Z Z
Rn

|Φ2s+t ∗ f (y)|2 |σ2t (x − y)| dy

Rn

×

Z

|σ2t (x − y)| dy

Rn

≤ kσ2t k1
≤C

Z

Z

−∞

∞

−∞

=C

∞

Z Z
Rn

Z

Z Z
Rn

 1/2  2

 1/2

|ν(x)| dx dt

|Φ2s+t ∗ f (y)|2 |σ2t (x − y)||ν(x)| dy dx dt

Rn

|Φ2t ∗ f (y)|2 σ ∗ (|ν|)(y) dy dt

Rn

∞

|Φ2t ∗ f (y)|2 dt σ ∗ (|ν|)(y) dy,

−∞

where C is independent of s, f and ν. Using Hölder’s inequality again and (3.7),
Lemma 2.3, we obtain
kQs f k2p ≤ C sup kgΦ ( f )k2p kσ ∗ (|ν|)kq ≤ Ck f k2p .
ν

Thus we have (3.10) for p > 2. From the proof of (3.10) above, it is easy to check
that the constant C is independent of s and f .
Step 2: There exists C > 0, independent of f and s, such that
(
C2−s k f k2
for s > 0,
(3.13)
kQs ( f )k2 ≤
βs/m
C2
k f k2 for s < 0,
where 0 < β < 2α1 n and m denotes the distinct numbers of {α j }.
By Plancherel’s theorem,
Z ∞Z
(3.14)
kQs f k22 ≤
| fb(ξ)|2 |ϕ(2s+t ρ(Lr ξ))|2 |c
σ2t (ξ)|2 dξdt,
Rn

−∞

where

−t
σc
2t (ξ) = 2
1

n−1

Z

0

2t

Z

′

a(y ′ ) J(y ′ )e−2πiξ·Aρ y dσ(y ′ )dρ

Sn−1

and a is a regular H (S ) atom supported in B(1, r) ∩ Sn−1 ,where 1 = (1, 0, . . . , 0).
We first give the estimate of |c
σ2t (ξ)|. Let η(y ′ ) = a(y ′ ) J(y ′ )/k JkL∞ (Sn−1 ) . By (3.1)
′
∞ n−1
and J(y ) ∈ C 0 (S ), we know η(y ′ ) satisfies (2.3) and (2.5), and supp(η) ⊂
B(1, r) ∩ Sn−1 . Then
Z tZ
′
k JkL∞ (Sn−1 ) 2
η(y ′ )e−2πiξ·Aρ y dσ(y ′ )dρ.
(3.15)
σc
2t (ξ) =
t
2
0
Sn−1

530

Y. Chen and Y. Ding
In the following, we want to prove |c
σ2t (ξ)| ≤ C min{|Lr A2t ξ|, |Lr A2t ξ|−β/m }, where
0 < β < 2α1 n and m denotes the distinct numbers of {α j }. For any ξ 6= 0, denote
Aρ ξ
|Aρ ξ|

=: ζ := (ζ1′ , ζ∗ ) ∈ Sn−1 , where ζ∗ ∈ Rn−1 . We choose a rotation O in Rn such
that O(ζ) = 1. Since O−1 = Ot , where O−1 and Ot denote the inverse and transpose
of O, respectively, it is easy to check that ζ is the first row vector of O. Thus, we have
O2 (ζ) = (ζ1′ , γ∗ ), where γ∗ ∈ Rn−1 . Now, we take a rotation Qn−1 in Rn−1 such
0
that Qn−1 (ζ∗ ) = γ∗ . Set R = 10 Qn−1
; then R is a rotation in Rn , such that for any
′
′
′
n−1
′
y := (ℓ, y 2 , . . . , y n ) in S , h1, Ry i = ℓ. Thus
k JkL∞ (Sn−1 )
σc
2t (ξ) =
2t

t
Z 2Z

0

′

η(O−1 (Ry ′ ))e−2πi|Aρ ξ|h1,Ry i dσ(y ′ )dρ.

Sn−1

Now η(O−1 (Ry ′ )) also satisfies (2.3) and (2.5), and is supported in B(ζ, r) ∩ Sn−1 .
Thus we have
t
Z 2Z

k JkL∞ (Sn−1 )
σc
2t (ξ) =
2t

0

Fη (ℓ)e−2πi|Aρ ξ|ℓ dℓdρ,

R

where Fη (ℓ) is the function defined in Lemma 2.1. By Lemma 2.1, we know that Fη
α1
|L A ξ|
ξ1
. Thus
is supported in (−2r(ζ) + δ1 , 2r(ζ) + δ1 ), where r(ζ) = |Ar ρ ρξ| and δ1 = ρ|Aρ ξ|
δ1
δ1
N(ℓ) = r(ζ)Fη (r(ζ)ℓ) is a function with support in the interval (−2 + r(ζ)
, 2 + r(ζ)
),
R
and kNk∞ < C (C is independent of η and ρ) and R N(ℓ) dℓ = 0. After changing a
variable we have

σc
2t (ξ) =

k JkL∞ (Sn−1 )
2t

t
Z 2Z

0

N(ℓ)e−2πiℓ|Lr Aρ ξ| dℓdρ.

R

So by the cancellation property of N, we obtain that

k JkL∞ (Sn−1 )
(3.16) |c
σ2t (ξ)| =
2t
t
Z 2Z
≤ C2−t
0

t
Z 2Z

0

ξ1

] dℓdρ

R

ζ

1 |
|ℓ− r(ζ)

≤ 2|N(ℓ)||Lr Aρ ξ| ℓ −
≤ C|Lr A2t ξ|.

α1

N(ℓ)[e−2πi|Lr Aρ ξ|ℓ − e−2πiρ

ζ1
dℓ dρ ≤ C
r(ζ)

Z

0

1

|Lr A2t ρ ξ| dρ

The Parabolic Littlewood–Paley Operator with Hardy Space Kernels

531

On the other hand, using Hölder’s inequality and (3.15), we have
t
Z Z
2
k JkL∞ (Sn−1 ) 2
′ −2πiξ·Aρ y ′
′
σ2t (ξ)| =
η(y
)e
dσ(y
)
dρ
(3.17)|c
2t
0 Sn−1
t
Z
Z
2
′
1 2
≤C t
η(y ′ )e−2πiξ·Aρ y dσ(y ′ ) dρ
2 0
Sn−1
Z 2t+ j Z
0
X
′
1
2 j−1 t+ j−1
=C
η(y ′ )e−2πiξ·Aρ y dσ(y ′ )
2
n−1
t+
j−1
S
2

2

2

dρ

j=−∞
0
X

≤C

2 j−1

0
X

2t+ j

2t+ j−1

j=−∞

=C

Z

Z

′

η(y ′ )e−2πiξ·Aρ y dσ(y ′ )

2 dρ

ρ

Sn−1

2 j−1 Bt, j (ξ),

j=−∞

where

Z

Bt, j (ξ) =

2t+ j

2t+ j−1

Then we get
Bt, j (ξ) =

Z

2t+ j

2t+ j−1

ZZ

≤C

ZZ

Z

′

η(y ′ )e−2πiξ·Aρ y dσ(y ′ )

2 dρ

ρ

Sn−1

η(y ′ )η(x ′ )e−2πiAρ (y

′

−x ′ )·ξ

.

dσ(y ′ )dσ(x ′ )

Sn−1 ×Sn−1

|η(y ′ )||η(x ′ )|

Sn−1 ×Sn−1

Z

2t+ j

e−2πiAρ (y

′

−x ′ )·ξ dρ

ρ

2t+ j−1

dρ
ρ

dσ(y ′ )dσ(x ′ ).

By Lemma 2.4, we know
Z

2t+ j

e−2πiAρ (y

′

−x ′ )·ξ dρ

2t+ j−1

ρ

=

Z

2

e−2πiA2t+ j−1 ρ (y

′

−x ′ )·ξ dρ

ρ

−2β/m
≤ C |(y ′ − x ′ ) · A2t+ j−1 ξ|
,
1

where 0 < β < 2α1 n and m denotes the distinct numbers of {α j }. Then by the above
inequality we get
(3.18) Bt, j (ξ) ≤ C

ZZ

|η(y ′ )||η(x ′ )|

Sn−1 ×Sn−1

where
I1 (ξ) =

ZZ

Sn−1 ×Sn−1

 −2β/m
× |(y ′ − x ′ ) · A2t+ j−1 ξ|
dσ(y ′ )dσ(x ′ ) = CI1 (ξ),
 −2β/m
|η(y ′ )||η(x ′ )| |(y ′ − x ′ ) · A2t+ j−1 ξ|
dσ(y ′ )dσ(x ′ ).

532

Y. Chen and Y. Ding
As was done above, for any ξ 6= 0, we choose a rotation O in Rn such that
O(A2t+ j−1 ξ) = |A2t+ j−1 ξ|1 = |A2t+ j−1 ξ|(1, 0, . . . , 0).
Thus, we may take another rotation R in Rn such that for any
y ′ = (y 1′ , y 2′ , . . . , y n′ ) ∈ Sn−1 ,
h1, Ry ′ i = y 1′ = h1, y ′ i. Now, let y ′ = (s, y 2′ , y 3′ , . . . , y n′ ), x ′ = (δ, x2′ , x3′ , . . . , xn′ ).
Then it is easy to see that
ZZ
|η(O−1 (Ry ′ ))||η(O−1 (Rx ′ ))|
I1 (ξ) =
Sn−1 ×Sn−1

 −2β/m
× |(y ′ − x ′ ) · |A2t+ j−1 ξ|1|
dσ(y ′ )dσ(x ′ ),

where O−1 is the inverse of O. Now η(O−1 (Ry ′ )) satisfies (2.3) and (2.5), and is
ξ
A
supported in B(ϑ, r) ∩ Sn−1 where ϑ = |A2t+t+ j−1
ξ| . Thus we have
2 j−1
ZZ
 −2β/m
Gη (s)Gη (δ) |A2t+ j−1 ξ||s − δ|
dsdδ,
I1 (ξ) =
R ×R

where Gη (s) is the function defined in Lemma 2.1. By Lemma 2.1, we know
|L A j−1 ξ|
2(t+ j−1)α1 ξ1
supp(Gη ) ⊂ (−2r(ϑ) + ϑ1 , 2r(ϑ) + ϑ1 ), where r(ϑ) = |Ar t+2t+j−1
ξ| and ϑ1 = |A2t+ j−1 ξ| .
2

ϑ1
) is a function supported in the interval (−2, 2),
Thus ϕ(s) = r(ϑ)Gη r(ϑ)(s − r(ϑ)
and kϕk∞ < C (C is independent of r, t, j and ϑ ). Since 2β/m < 1, we get
Z 2Z 2
 −2β/m
I1 (ξ) =
ϕ(s)ϕ(δ) |Lr A2t+ j−1 ξ||s − δ|
dsdδ
−2

−2

≤ C|Lr A2t+ j−1 ξ|−2β/m

Z

2

−2

Z

2

|s − δ|−2β/m dsdδ

−2

≤ C|Lr A2t+ j ξ|−2β/m .
This together with (3.18) gives
Bt, j (ξ) ≤ C|Lr A2t+ j ξ|−2β/m .

(3.19)
Since 0 < β <
(3.20)

1
2αn

and m ≥ 1, then by (3.17) and (3.19), we get
|c
σ2t (ξ)|2 ≤ C

0
X

2 j−1 |Lr A2t+ j ξ|−2β/m

j=−∞

≤C

0
X

2 j(1−2βαn /m) |Lr A2t ξ|−2β/m

j=−∞

≤C

0
X

2 j(1−2βαn /m) |Lr A2t ξ|−2β/m

j=−∞

≤ C|Lr A2t ξ|−2β/m .

The Parabolic Littlewood–Paley Operator with Hardy Space Kernels

533

By (3.16) and (3.20), we have
|c
σ2t (ξ)| ≤ C min{|Lr A2t ξ|, |Lr A2t ξ|−β/m }.
Now we give the estimates kQs ( f )k2 . For s > 0, by (3.14) and the properties of ϕ,
using the estimate |c
σ2t (ξ)| ≤ C|Lr A2t ξ| and the Plancherel theorem, we get

kQs ( f )k22 ≤ C

=C

≤C

×

1
rn+1
1
rn+1
Z

≤ C2

Z

Z

−∞

∞

−∞

Z

∞

∞

−∞

2−s−1 ≤2t ρ≤2−s+1

Z

Sn−1

ρ
−s−1− log
log 2

−∞

≤ C2

1
rn+1

Z

Rn

Z

Sn−1

| fb(ξ)|2 |Lr A2t ξ|2 dξdt

J(ξ ′ )| fb(Lr−1 Aρ ξ ′ )|2 |A2t Aρ ξ ′ |2 ρα−1 dσ(ξ ′ )dρdt

J(ξ ′ )| fb(Lr−1 Aρ ξ ′ )|2 (2−s+1 )2α1 + · · · + (2−s+1 )2αn




dt ρα−1 dσ(ξ ′ )dρ

Z∞ Z

1
rn+1

−2s

2−s−1 ≤ρ(Lr A2t ξ)≤2−s+1

Z

ρ
−s+1− log
log 2

−2sα1

Z

Sn−1

J(ξ ′ )| fb(Lr−1 Aρ ξ ′ )|2 ρα−1 dσ(ξ ′ )dρ

| fb(Lr−1 ξ)|2 dξ

≤ C2−2s k f k22 .
So we have kQs ( f )k2 ≤ C2−s k f k2 for s > 0. Using the estimate
|c
σ2t (ξ)| ≤ C|Lr A2t ξ|−β/m
and the same idea, we have kQs ( f )k2 ≤ C2βs/m k f k2 for s < 0. Thus we get (3.13),
and obviously, the constant C is independent of s and f .
Applying the Riesz–Thorin interpolation theorem of sub-linear operators  between (3.10) and (3.13), we know that there exist two constants γ, τ > 0 such that
kQs ( f )k p ≤ C2−γs k f k p

for s > 0, 1 < p < ∞,

kQs ( f )k p ≤ C2τ s k f k p

for s < 0, 1 < p < ∞.

Thus, we obtain (3.9) and (3.5) follows.

534

Y. Chen and Y. Ding
Acknowledgement The authors would like to express their gratitude to the referee

References


A. Benedek, A. Calderón and R. Panzone, Convolution operators on Banach space valued functions.
Proc. Natl. Acad. Sci. U.S.A. 48(1962), 356–365.
 A. Calderón and A. Zygmund, A note on the interpolation of sublinear operator. Amer. J. Math,
78(1956), 282–288.
 L. Colzani, Hardy Spaces on Sphere. Ph.D. Thesis, Washington University, St. Louis, MO, 1982.
 Y. Ding, D. Fan, and Y. Pan, L p boundedness of Marcinkiewicz integrals with Hardy space function
kernels. Acta Math. Sin. Ser. (Engl. Ser.), 16(2000), 593–600.
 Y. Ding, Q. Xue, and K. Yabuta, Parabolic Littlewood-Paley g-function with rough kernels. Acta
Math. Sin. (Engl. Ser.) 24(2008), 2049–2060.
 D. Fan and Y. Pan, A singular integral operator with rough kernel. Proc. Amer. Math. Soc.
125(1997), 3695–3703.
 E. Fabes and N. Rivière, Singular integrals with mixed homogeneity. Studia Math. 27(1966), 19–38.
 W. R. Madych, On Littlewood-Paley functions. Studia Math. 50(1974), 43–63.
 A. Nagel, N. Rivière, and S. Wainger, On Hilbert transforms along curves. II. Amer. J. Math.
98(1976), 395–403.
 E. M. Stein, On the functions of Littlewood-Paley, Lusin, and Marcinkiewicz. Trans. Amer. Math. Soc.
88(1958), 430–466.
 E. M. Stein and G. Weiss, Introduction to Fourier Analysis on Euclidean Spaces. Princeton University
Press, Princeton, NJ, 1971.
Department of Mathematics and Mechanics Applied Science School, University of Science and Technology
Beijing, Beijing 100083, The People’s Republic of China
e-mail: yanpingch@126.com
School of Mathematical Sciences, Beijing Normal University, Laboratory of Mathematics and Complex Systems (BNU), Ministry of Education, Beijing 100875, The People’s Republic of China
e-mail: dingy@bnu.edu.cn

```