Directory UMM :Journals:Journal_of_mathematics:GMJ:

Georgian Mathematical Journal
1(94), No. 4, 367-376

TWO-WEIGHTED Lp -INEQUALITIES FOR SINGULAR
INTEGRAL OPERATORS ON HEISENBERG GROUPS
V. S. GULIEV

Abstract. Some sufficient conditions are found for a pair of weight
functions, providing the validity of two-weighted inequalities for singular integrals defined on Heisenberg groups.

Estimates for singular integrals of the Calderon–Zygmund type in various spaces (including weighted spaces and the anisotropic case) have attracted a great deal of attention on the part of researchers. In this paper
we will deal with singular integral operators T on the Heisenberg group H n
which have an essentially different character as compared with operators
of the Calderon–Zygmund type. We have obtained the two-weighted Lp inequality with monotone weights for singular integral operators T on H n .
Applications are given.
Let H n be the Heisenberg group (see [1], [2]) realized as a set of points
x = (x0 , x1 , . . . , x2n ) = (x0 , x′ ) ∈ R2n+1 with the multiplication
n

€
1X

(xi yn+i − xn+i yi ), x′ + y ′ .
xy = x0 + y0 +
2 i=1

The corresponding Lie algebra is generated by the left-invariant vector
fields


, Xi =
+
∂x0
∂xi

1

=
− xi
,
∂xn+i
2 ∂x0


X0 =
Xn+i


1
,
xn+i
2
∂x0
i = 1, . . . , n,

which satisfy the commutation relation
1
X0 ,
4
[X0 , Xi ] = [X0 , Xn+i ] = [Xi , Xj ] = [Xn+i , Xn+j ] = [Xi , Xn+j ] = 0,
[Xi , Xn+i ] =

1991 Mathematics Subject Classification. 42B20, 42B25, 42B30.


367
c 1994 Plenum Publishing Corporation
1072-947X/94/0700-0367$12.50/0 

368

V. S. GULIEV

i, j = 1, . . . , n i =
6 j.
The dilation δt : δt x = (t2 x0 , tx′ ), t > 0, is defined on H n . The Haar measure on this group coincides with the Lebesgue measure dx = dx0 dx1 · · · dx2n .
The identity element in H n is e = 0 ∈ R2n+1 , while the element x−1 inverse
to x is (−x).
The function f defined in H n is said to be H-homogeneous of degree m,
on H n , if f (δt x) = tm f (x), t > 0. We also define the norm on H n
2n
€X
2 ƒ1/4
‚

|x|H = x20 +
xi2
i=1

which is H-homogeneous of degree one. This also yields the distance function, namely, the distance
d(x, y) = d(y −1 x, e) = |y −1 x|H ,
|y −1 x|H =

‚€

x0 − y0 −

n
2
1X
(xi yn+i − xn+i yi ) +
2 i=1

2n
2 ƒ1/4

€X
+
(xi − yi )2
.
i=1

d is left-invariant in the sense that d(x, y) remains unchanged when x and y
are both left-translated by some fixed vector in H n . Furthermore, d satisfies
the triangle inequality d(x, z) ≤ d(x, y) + d(y, z), x, y, z ∈ H n . For r > 0
and x ∈ H n let
B(x, r) = {y ∈ H n ; |y −1 x|H < r} (S(x, r) = {y ∈ H n ; |y −1 x|H = r})
be the H-ball (H-sphere) with center x and radius r.
The number Q = 2n + 2 is called the homogeneous dimension of H n .
Clearly, d(δt x) = tQ dx.
Given functions f (x) and g(x) defined in H n , the Heisenberg convolution
(H-convolution) is obtained by
Z
Z
f (y)g(y −1 x)dy =
(f ∗ g)(x) =

f (xy −1 )g(y)dy,
Hn

Hn

n

where dy is the Haar measure on H .
is summable
The kernel K(x) admitting the estimate |K(x)| ≤ C|x|α−Q
H
in the neighborhood of e for α > 0 and in that case K ∗ g is defined for
the function g with bounded support. If however the kernel K(x) has a
singularity of order Q at zero, i.e., |K(x)| ∼ |x|−Q
H near e, then there arises
a singular integral on H n .

TWO-WEIGHTED Lp -INEQUALITIES

369


Let ω(x) be a positive measurable function on H n . Denote by Lp (H n , ω)
a set of measurable functions f (x), x ∈ H n , with the finite norm
‘1/p
Z
, 1 ≤ p < ∞.
|f (x)|p ω(x)dx
kf kLP (H n ,ω) =
Hn

We say that a locally integrable function ω : H n → (0, ∞) satisfies
Muckenhoupt’s condition Ap = Ap (H n ) (briefly, ω ∈ Ap ), 1 < p < ∞, if
there is a constant C = C(ω, p) such that for any H-ball B ⊂ H n
Z
Z
‘

‘

1

1
+ ′ = 1,
ω 1−p (x)dx ≤ C,
ω(x)dx |B|−1
|B|−1
p
p
B
B
where the second factor on the left is replaced by ess sup{ω −1 (x) : x ∈ B}
if p = 1.
Let K(x) be a singular kernel defined on H n \{e} and satisfying the confunction of degree −Q, i.e., K(δt x) =
ditions: K(x) is an H-homogeneous
R
t−Q K(x) for any t > 0 and SH K(x)dσ(x) = 0, where dσ(x) is a measure
element on SH = S(e, 1).
Denote by ωK (δ) the modulus of continuity of the kernel on SH :
ωK (δ) = sup{|K(x) − K(y)| : x, y ∈ SH ,
It is assumed that


Z

1

ωK (t)
0

dt
< ∞.
t

We consider the singular integral operator T :
Z
Z
−1
K(xy )f (y)dy =: lim
T f (x) =
Hn

|y −1 x|H ≤ δ}.


ε→0+
|xy −1 |H >ε

K(xy −1 )f (y)dy.

As is known, T acts boundedly in Lp (H n ), 1 < p < ∞ (see [3], [4]).
For singular integrals with Cauchy–Szeg¨
o kernels the weighted estimates
were established in the norms of Lp (H n , ω) with weights ω satisfying the
condition Ap [5]. These results extend to the more general kernels considered
above [4].
Theorem 1 [4]. Let 1 < p < ∞ and ω ∈ Ap ; then T is bounded in
Lp (H n , ω).
In the sequel we will use
Theorem 2. Let 1 ≤ p ≤ q < ∞ and U (t), V (t) be positive functions
on (0, ∞).

370


V. S. GULIEV

1) The inequality
Œq ‘1/q
ŒZ t
Z
Z ∞
Œ
Œ
≤ K1
ϕ(τ )dτ Œ dt
U (t)Œ

|ϕ(t)|p v(t)dt

0

0

0



‘1/p

with the constant K1 not depending on ϕ holds iff the condition
‘p−1
‘p/q  Z t
Z ∞

0

is fulfilled;
2) The inequality
Z
Z ∞
U (t)|

t

0

0

t



ϕ(τ )dτ |q dt

‘1/q

≤ K2

Z

0



‘1/p
|ϕ(t)|p V (t)dt

with the constant K2 not depending on ϕ holds iff the condition
Z t
‘p−1
‘p/q  Z ∞

V (τ )1−p dτ
0

t

0

is fulfilled.
Note that Theorem 2 was proved by G.Talenti, G.Tomaselli, B.Muckenhoupt [7] for 1 ≤ p = q < ∞, and by J.S.Bradley [8], V.M.Kokilashvili [9],
V.G.Maz’ya [10] for p < q.
epq (γ), γ > 0, if
We say that the weight pair (ω, ω1 ) belongs to the class A
either of the following conditions is fulfilled: a) ω(t) and ω1 (t) are increasing
functions on (0, ∞) and
‘p/q  Z t/2
‘p−1
Z ∞

−1−γq/p′
ω(τ )1−p τ γ−1 dτ

< ∞;
ω(τ )τ
sup
t>0

0

t

b) ω(t) and ω1 (t) are decreasing functions on (0, ∞) and
‘p−1
‘p/q  Z ∞
 Z t/2


γ−1
< ∞.
ω1 (τ )τ
ω(τ )1−p τ −1−γp /q dτ

sup
t>0

0

t

ep (Q) ≡
Theorem 3. Let 1 < p < ∞ and the weight pair (ω, ω1 ) ∈ A
n
epp (Q). Then for f ∈ Lp (H , ω(|x|H )) there exists T f (x) for almost all
A
x ∈ H n and
Z
Z
|T f (x)|p ω1 (|x|H )dx ≤ C
|f (x)|p ω(|x|)H )dx,
(1)
Hn

Hn

where the constant C does not depend on f .

TWO-WEIGHTED Lp -INEQUALITIES

371

Corollary. If ω(t), t > 0 is increasing (decreasing) and the function
ω(t)t−β is decreasing (increasing) for some β ∈ (0, Q(p − 1)) (β ∈ (−Q, 0)),
then T is bounded on Lp (H n , ω(|x|H )).
Proof of Theorem 3. Let f ∈ Lp (H n , ω(|x|H )) and ω, ω1 be positive increasing functions on (0, ∞). We will prove that T f (x) exists for almost all
x ∈ H n . We take any fixed τ > 0 and represent the function f in the norm
of the sum f1 + f2 , where
(
f (x), if |x|H > τ /2
f1 (x) =
, f2 (x) = f (x) − f1 (x).
0,
if |x|H ≤ τ /2
Let ω(t) be a positive increasing function on (0, ∞) and f ∈ Lp (H n , ω(|x|H )).
Then f1 ∈ Lp (H n ) and therefore T f1 (x) exists for almost all x ∈ H n . Now
we will show that T f2 converges absolutely for all x : |x|H ≥ τ . Note that
C(K) = supx∈SH |K(x)| < ∞. Hence
Z

|T f2 (x)| ≤ C(K)

|f (y)|

|y|H ≤τ /2

€ 2  Qp

τ

Z

|xy−1 |Q
H

dy ≤
(2)

1

|f (y)|ω(|y|H ) p
1

ω(|y|H ) p

|y|H ≤τ /2

dy,

since |xy −1 |H ≥ |x|H − |y|H ≥ τ /2. Thus, by the H¨older inequality we can
estimate (2) as
|T f2 (x)| ≤ Cτ

−Q/p

kf kLp (H n ,ω(|x|H ))

Z

τ /2



ω(t)1−p tQ−1 dt

0

‘1/p′

.

Therefore T f2 (x) converges absolutely for all x : |x|H ≥ τ and thus T f (x)
exists for almost all x ∈ H n . Assume ω
¯ 1 (t) to be an arbitrary continuous
¯ 1 (t) ≤ ω1 (t), ω
increasing function on (0, ∞) such that ω
¯ 1 (0) = ω1 (0+) and
Rt
ω
¯ 1 (t) = 0 ϕ(τ )dτ + ω
¯ 1 (t) exists;
¯ 1 (0), t ∈ (0, ∞) (it is obvious that such ω
Rt
for example, ω
¯ 1 (t) = 0 ω1′ (τ )dτ + ω1 (t)).
We observe that the condition a) implies
∃C1 > 0, ∀t > 0, ω1 (t) ≤ C1 ω(t/2).

(3)

Indeed, from
Z

∃C2 > 0, ∀t > 0,
‘p−1
‘ Z t/2

ω(τ )1−p τ Q−1 dτ
≤ C2
ϕ(τ )τ −Q(p−1) dτ


t

0

(4)

372

V. S. GULIEV

we obtain (3), since
Z ∞
Z

ω1 (τ )τ −1−Q(p−1) dτ ≥ Cω1 (t)t−Q(p−1) ,

t
t/2



ω(τ )1−p τ Q−1 dτ

0

and, besides,

‘p−1

≤ Cω(t/2)−1 tQ(p−1)

Z ∞
Z ∞
Z ∞
1
−Q(p−1)
dτ =
ϕ(τ )dτ
λ−1−Q(p−1) dλ =
ϕ(τ )τ
Q(p − 1) t
t
τ
Z λ
Z ∞
Z ∞
λ−1−Q(p−1) dλ
ϕ(τ )dτ ≤
ω1 (τ )τ −1−Q(p−1) dτ.
=
t

t

t

We have
kT f kLp ,¯ω1 (H n ) ≤


¯ 1 (0)
+ ω

Z

Z

|T f (x)|p dx

Hn

|T f (x)|p dx

Hn

Z

‘1/p

|x|H

ϕ(t)dt
0

‘1/p

+

= A1 + A2 .

If ω(0+) > 0, then Lp (H n , ω(|x|H )) ⊂ Lp (H n ), and if ω(0+) = 0, then
¯ (t) ≤ ω1 (t) ≤ Cω(t/2) implies ω
ω
¯ 1 (0) = 0. Therefore in the case ω(0+) = 0
we have A2 = 0.
If ω(0) > 0, then f ∈Lp (Rn ) and we have
Z

‘1/p
Z
‘1/p
A2 ≤ C ω

¯ 1 (0)
|f (x)|p ω1 (|x|H )dx
≤C
|f (x)|p dx
Hn

Hn

≤ Ckf kLp (H n ,ω(|x|H )) .
Now we can write
Z
A1 ≤

Z



ϕ(t)dt

0

|x|H >t

‘1/p
≤ A11 + A12 .
|T f (x)p dx

where
p
A11
=

Ap12 =

Z

Z



ϕ(t)dt
0

Z

Z

|x|H >t |y|H >t/2


ϕ(t)dt
0

Œ
Œ
Œ

Z

Œ
Œ
Œ

Z

|x|H >t |y|H t/2

|y|H >t/2

implies f ∈ Lp ({y ∈ H n : |y|H > t}) for any t > 0.
Hence, on account of (3), we have
Z
‘1/p
Z ∞
|f (x)|p dx
ϕ(t)dt
=
A11 ≤ C
0

=C
≤C

Z

Z

|x|H >t/2

|f (x)|p dx

Hn

Z

2|x|H

ϕ(t)dt
0

|f (x)|p ω1 (2|x|H )dx

Hn

‘1/p

‘1/p



≤ Ckf kLp,ω(|x|H ) (H n ).

Obviously, if |x|H > t, |y|H < t/2, then 21 |x|H ≤ |y −1 x|H ≤
Therefore
Z Œ Z
Œp
Œ
Œ
K(xy −1 )f (y)dy Œ dx ≤
Œ

3
2 |x|H .

|x|H >t |y|H t

≤2

Qp

C(K)



Z

|xy −1 |−Q
H |f (y)|dy

|y|H Q(1 +
Z

0

SH

|x|H >t

1
p′ ),

by virtue of the H¨older inequality, we have

|f (y)|dy = α

Z

dσ(¯
y)

SH

|y|H