Directory UMM :Journals:Journal_of_mathematics:GMJ:
Georgian Mathematical Journal
Volume 8 (2001), Number 1, 33-59
WEIGHT INEQUALITIES FOR SINGULAR INTEGRALS
DEFINED ON SPACES OF HOMOGENEOUS AND
NONHOMOGENEOUS TYPE
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
Dedicated to Professor Hans Triebel on his 65th birthday
Abstract. Optimal sufficient conditions are found in weighted Lorentz spaces for weight functions which provide the boundedness of the Calder´on–
Zygmund singular integral operator defined on spaces of homogeneous and
nonhomogeneous type.
2000 Mathematics Subject Classification: 42B20, 42B25.
Key words and phrases: Singular Integral, maximal function, Hardytype operator, space of homogeneous type, space of nonhomogeneous type,
Lorentz space.
Introduction
In the present paper optimal sufficient conditions are found for the weight
functions which provide the boundedness of Calder´on–Zygmund singular integral operator defined on spaces of homogeneous as well as nonhomogeneous type
in weighted Lorentz spaces. In the nonhomogeneous case the results are new
even in Lebesgue spaces.
Two-weight strong type inequalities for singular integrals in Lebesgue spaces
on spaces of homogeneous type (SHT) were established in [5] (see also [6],
Chapter 9, and [14]), while a similar problem for singular integrals defined on
homogeneous groups was considered in [16] in the case of Lorentz spaces.
Two-weight inequalities for Hilbert transforms with monotonic weights were
studied in [17]. Analogous problems for singular integrals in Euclidean spaces
were considered in [8] and generalized in [7] for singular integrals on Heisenberg
groups. Optimal conditions for a pair of weights ensuring the validity of twoweight inequalities for Calder´on–Zygmund singular integrals were obtained in
[3] (and also in [19] for the case of Lorentz spaces). The latter result was
generalized to the setting of homogeneous groups in [13], [15].
Finally, we would like to mention that singular integrals on spaces of nonhomogeneous type (measure spaces with a metric) were studied in [18].
c Heldermann Verlag www.heldermann.de
ISSN 1072-947X / $8.00 / °
34
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
1. Preliminaries
In this section we give the notion of SHT and defined Lorentz space. The
definition of a singular integral and several well-known results concerning the
Hardy-type operators on SHT are also given.
Definition 1.1. A space of homogeneous type (SHT) (X, d, µ) is a topological space X endowed with a complete measure µ such that: (a) the space of
compactly supported continuous functions is dense in L1 (X, µ), and (b) there
exists a non-negative real-valued function (quasimetric) d : X × X → R1 satisfying:
(i) d(x, x) = 0 for arbitrary x ∈ X;
(ii) d(x, y) > 0 for arbitrary x, y ∈ X, x 6= y;
(iii) there exists a positive constant a0 such that the inequality
d(x, y) ≤ a0 d(y, x)
holds for all x, y ∈ X;
(iv) there exists a positive constant a1 such that the inequality
d(x, y) ≤ a1 (d(x, z) + d(z, y))
holds for arbitrary x, y, z ∈ X;
(v) for every neighborhood V of any point x ∈ X there exists a number r > 0
such that the ball
n
o
B(x, r) = y ∈ X : d(x, y) < r
with center in x and radius r is contained in V ;
(vi) the balls B(x, r) are measurable for all x ∈ X, r > 0 and, moreover,
0 < µB(x, r) < ∞;
(vii) there exists a positive constant b such that the inequality (doubling
condition)
µB(x, 2r) ≤ bµB(x, r)
is true for all x ∈ X and for all positive r.
It will be supposed that for some x0 ∈ X
µ{x0 } = µ{x ∈ X : d(x0 , x) = a} = 0,
where
n
o
a ≡ sup d(x0 , x) : x ∈ X .
Note that the condition a = ∞ is equivalent to the condition µ(X) = ∞, and
if a = ∞, then µ{x ∈ X : d(x0 , x) = a} = 0.
For the definitions, various examples and properties of SHT, we refer to [2],
[23] (also to [6]).
Definition 1.2. By a spaces of nonhomogeneous type we mean a measure
space with a quasimetric (X, d, µ) satisfying conditions (i)–(v) of Definition 1.1,
i.e., the doubling condition is not assumed and may fail.
WEIGHT INEQUALITIES
35
In the sequel it will be assumed that there exists a point x0 ∈ X such that
B(x0 , R)\B(x0 , r) 6= ∅
(1)
for all r and R provided 0 < r < R < a.
In the remaining part of this section (X, d, µ) is assumed to be a space of
nonhomogeneous type.
Definition 1.3. An almost everywhere positive, locally integrable function
w : X → R1 is called a weight.
For weight functions w, we denote by Lpq
w (X) the Lorentz space with weight
w which represents the class of all µ-measurable functions f : X → R1 for which
à +∞
Z µ
Z
kf kLpq
= q
w (X)
0
¶q/p
w(x) dµ
{x∈X: |f (x)|>λ}
λ
q−1
dλ
!1/q
0
Z
¶1/p
w(x) dµ
{x∈X: |f (x)|>λ}
0.
In the sequel we shall need the following results concerning the Hardy-type
operators on the measure space (X, µ)
Hf (t) =
Z
f (x) dµ(x), t ∈ (0, a),
Z
f (x) dµ(x), t ∈ (0, a).
{x: d(x0 ,x)t}
The following result is from [5].
Proposition A. Let 1 < p ≤ q < ∞, µ{x0 } = 0. Then the inequality
µ Za
q
¶1/q
v(t)|Hf (t)| dt
0
µZ
≤c
¶1/p
p
|f (x)| w(x) dµ
X
,
holds with some c > 0 independent of f , f ∈ Lpw (X), if and only if
D = sup
0t} (·)°°
°
w(·)w1 (·)
′ ′
p s
Lw
(X)
< ∞,
then Kg(x) exists a.e. on X for arbitrary g provided that kg(·)w1 (·)kLps
< ∞.
w (X)
From Lemma 2.3 the next lemma follows easily:
WEIGHT INEQUALITIES
41
Lemma 2.4. Let a = ∞, 1 < s ≤ p < ∞. Assume that u and u1 are positive
decreasing functions on (0, ∞) and the condition
°
°
°
°
°
°
µB(d(x0 , ·))−1
χ{d(x0 ,y)>t} (·)°° p′ s′
u(d(x0 , ·))u1 (d(x0 , ·))
Lu(d(x
0. Then Kg(x) exists a.e. on X for arbitrary g satisfying the
condition g(·)u1 (d(x0 , ·)) ∈ Lps
u(d(x0 ,·)) (X).
Now we pass to the weight inequalities for the operator K.
Theorem 2.1. Let 1 < s ≤ p ≤ q < ∞, let w be a weight function on
X, suppose that σ is a positive increasing continuous function on (0, 4a1 a),
ρ ∈ Ap (X) and v(x) = σ(d(x0 , x))ρ(x). Suppose the following conditions are
satisfied:
(a) there exists a positive constant b such that the inequality
σ(2a1 d(x0 , x))ρ(x) ≤ bw(x)
holds for almost every x ∈ X;
(b)
°
° 1
°
pq
Lv (X) °
°
°
°
,y)>t}
°
B ≡ sup °°µB(x0 , d(x0 , ·))−1 χ{d(x
0
0λ}
ρ(x)
µ d(x
Z 0 ,x)
¶
!q/p
ϕ(t) dt dµ
0
dλ
!1/q
!1/q
dλ
≡ I1 + I2 .
In the case, where σ(0+) = 0, we have I1 = 0; otherwise, by Theorem B and
Lemma A (part (ii)), we have
I1 = c1 σ(0+)1/p kKf (·)kLpq
≤ c2 σ 1/p (0+)kf (·)kLpq
ρ (X)
ρ (X)
≤ c2 σ 1/p (0+)kf (·)kLps
≤ c2 kf (·)kLps
.
ρ (X)
w (X)
42
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
Now, let us estimate I2 . Set
f1t (x) = f (x)χ{d(x
t
0 ,x)> 2a1 }
, f2t (x) = f (x) − f1t (x).
We have
à Z∞
I2 = c1 q
λ
à Za
q−1
0
à Z∞
≤ c3 q
λ
q−1
0
à Z∞
+c3 q
λ
q−1
0
ϕ(t)
χ{x: |Kf
λ
1t (x)|> 2 }
{x: d(x0 ,x)>t}
µ
Z
ϕ(t)
0
Z
χ{x: |Kf
λ
2t (x)|> 2 }
{x: d(x0 ,x)>t}
I21 ≤ c4
à Z∞
ϕ(t)
µ Za
ϕ(t)kf1t (·)kpLpq
ρ (X)
0
0
≤ c5
λ
q−1
µ
≤ c5
λ
s−1
0
Ã
Z
≤ c5 kf (·)k
≤ c5
µ Za
dλ
!1/q
dλ
dλ
!p/q
!1/p
dt
ϕ(t)kf1t (·)kpLps
ρ (X)
d(x0 ,x)
µ 2a1 Z
¶
!s/p
ϕ(t) dt dµ
0
¶1/p
dt
dλ
!1/s
≤ c6 kf (·)kLps
.
w (X)
Next we shall estimate I22 . Note that if d(x0 , x) > t and d(x0 , y) <
t
2a1
d(x0 , x) ≤ a1 (d(x0 , y) + d(y, x)) ≤ a1 (d(x0 , y) + a0 d(x, y))
≤ a1
Hence
µ
¶
µ
¶
d(x0 , x)
t
+ a0 d(x, y) ≤ a1
+ a0 d(x, y) .
2a1
2a1
d(x0 , x)
≤ d(x, y),
2a1 a0
and we also have
!1/q
≡ I21 + I22 .
0
ρ(x)
!1/q
≥ 1) and using Theorem B
ρ(x) dµ
¶1/p
{x∈X: |f (x)|>λ}
Lps
w (X)
!q/p
¶q/p
}
|Kf1t (x)|> λ
2
dt
0
à Z∞
Z
{x∈X:
ρ(x) dµ dt
ρ(x) dµ dt
p
s
!q/p
¶
¶
Applying the Minkowki’s inequality twice ( pq ≥ 1,
and Lemma A (part (ii)), we obtain
à Za
dλ
ρ(x) dµ dt
{x: d(x0 ,x)>t,|Kf (x)|>λ}
µ
!q/p
¶
Z
ϕ(t)
0
à Za
0
à Za
µ
µ µ
µ(B(x, d(x0 , x))) ≤ bµ B x,
d(x0 , x)
2a1 a0
¶¶
≤ bµ(B(x, d(x, y))).
Moreover, it is easy to show that
µB(x0 , d(x0 , x)) ≤ b1 µB(x, d(x0 , x)).
Hence
µB(x0 , d(x0 , x)) ≤ b2 µB(x, d(x, y)).
, then
WEIGHT INEQUALITIES
43
Taking into account this inequality and using Proposition C we see that
°
°
I22 ≤ c7 °°
1
µB(x0 , d(x0 , ·))
Z
{d(x0 ,y)0
0 ,y)>t}
(·)µ(B(x0 , d(x0 , ·)))
°
°
−1 °
°
° 1
°
pq
Lv (X) °
n
w(·)
°
°
χ{d(x0 ,y)≤t} (·)°°
′ ′
Lpw s (X)
.
By virtue of what has been proved, if B < ∞, then the following inequality
holds:
≤ ckf (·)kLps
,
kKf (·)kLpq
w (X)
vn (X)
where the constant c > 0 does not depend on n.
By passing to the limit as n → ∞, we obtain inequality (2.1).
R∞
Using the representation σ(t) = σ(+∞)+ ψ(τ ) dτ , where σ(+∞) = lim σ(t)
t→∞
t
and ψ ≥ 0 on (0, ∞) and Proposition D, we obtain the following result.
Theorem 2.2. Let a = ∞, 1 < s ≤ p ≤ q < ∞, and suppose that σ is a
positive decreasing continuous function on (0, ∞). Assume that ρ ∈ Ap (X) and
v(X) = σ(d(x0 , x))ρ(x). Suppose the following conditions are satisfied:
(a) there exists a positive constant b such that the inequality
µ
¶
d(x0 , x)
ρ(x)σ
≤ bw(x)
2a1
is true for almost every x ∈ X;
(b)
°
° (µB(x0 , d(x0 , x)))−1
°
B ′ = sup kχ{d(x0 ,y)≤t} (·)kLpq
v (X) °
t>0
Then inequality (2.1) holds.
w(·)
°
°
χ{d(x0 ,y)>t} (·)°°
′ ′
p s
Lw
(X)
< ∞.
Now let us consider the particular cases of Theorems 2.1 and 2.2.
Theorem 2.3. Let a = ∞, 1 < p ≤ q < ∞, suppose σ1 and σ2 are positive,
increasing functions on (0, ∞), let σ1 be a continuous function and suppose that
ρ ∈ Ap (X). Put v(x) = σ2 (d(x0 , x))ρ(x), w(x) = σ1 (d(x0 , x))ρ(x). If
°
sup °°(µB(x0 , d(x0 , ·)))−1 χ{d(x
t>0
0
°
° 1
°
pq
Lv (X) °
°
(·)°°
,y)>t}
w(·)
°
°
χ{d(x0 ,y)≤t} (·)°°
′
Lpw (X)
< ∞,
44
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
then there exists a positive constant c such that
kKf (·)kLpq
≤ ckf (·)kLpw (X)
v (X)
for all f ∈ Lpw (X).
Proof. By Theorem 2.1 it is sufficient to show that there exists a positive
constant b such that σ2 (2a1 t) ≤ bσ1 (t) for all t ∈ (0, ∞).
From the doubling condition (see (vii) in Definition 1.1) and the condition
(1.1) it follows that the measure µ satisfies the reverse doubling condition at
the point x0 (see, e.g., [20] and [23], Lemma 20). In other words, there exist
constants η1 > 1 and η2 > 1 such that
µ(B(x0 , η1 r)) ≥ η2 µ(B(x0 , r))
for all r > 0.
Applying the H¨older’s inequality and using Lemma C, the reverse doubling
′
condition and the fact that ρ1−p ∈ Ap′ (X), we obtain
´−p
σ2 (2a1 t) ³
≤ µ(B(x0 , 2a1 η1 t)\B(x0 , 2a1 t))
σ1 (t)
µ
Z
×
≤
µµ
1−p′
ρ
µ
¶
¶p−1
(x) dµ
¶−p µ
µ
×
Z
¶p−1
B(x0 ,t)
0 ,2a1 η1 t)\B(x0 ,2a1
°³
´−1
°
≤ c2 ° µB(x0 , d(x0 , ·)) χ{d(x
0
¶
ρ(x) dµ
B(x0 ,2a1 η1 t)\B(x0 ,2a1 t)
′
°
≤ c1 (µB(x0 , 2a1 ηt))−p °°χ(B(x
ρ(x) dµ
σ2 (2a1 t)
σ1 (t)
Z
ρ1−p (x) dµ
σ2 (2a1 t)
σ1 (t)
°p
°
°
° 1
°
χB(x0 ,t) (·)°° p′
°
Lv (X) w(·)
Lw (X)
°p
°
1
χB(x0 ,t) (·)°° p′
≤ c.
°p
°
(·)
° pq
t))
°
°
°
Lv (X) ° w(·)
°p
°
(·)
° pq
,y)>t}
¶
B(x0 ,2a1 η1 t)\B(x0 ,2a1 t)
B(x0 ,2a1 η1 t)\B(x0 ,2a1 t)
1
µB(x0 , 2a1 ηt)
1−
η2
Z
Lw (X)
An analogous theorem dealing with singular integrals on homogeneous groups
was proved in [16], and for singular integrals on SHT in the Lebesgue space in [5].
Theorem 2.4. Let a = ∞, 1 < s ≤ p ≤ q < ∞; suppose σ1 , σ2 , u1 and
u2 are weight functions defined on X. Let ρ ∈ Ap (X), and suppose v = σ2 ρ,
w = σ1 ρ. Assume that the following conditions are fulfilled:
(1) there exists a positive constant b such that for all t > 0
1/p
1/p
sup σ2 (x) sup u2 (x) ≤ b inf σ1 (x) inf u1 (x)
Ft
holds, where Ft = {x ∈ X :
Ft
t
a1
Ft
≤ d(x0 , x) < 8a1 t};
Ft
WEIGHT INEQUALITIES
45
(2)
°
°
³
´−1
sup °u2 (·) µ(B(x0 , d(x0 , ·)))
t>0
(3)
°
°
sup °u2 (·)χ{d(x
0
t>0
Then
°
°
×°°
1
u1 (·)w(·)
°
°
χ{d(x0 ,t)>t} (·)°
Lpq
v (X)
°
°
χ{d(x0 ,y)≤t} (·)°°
′ ′
p s
Lw
(X)
< ∞;
°
°
° (µB(x0 , d(x0 , ·)))−1
°
°
χ{d(x0 ,y)>t} (·)°°
pq
°
Lv (X)
°
°
°
(·)
,t)≤t}
u1 (·)w(·)
′ ′
p s
Lw
(X)
ku2 (·)Kf (·)kLpq
≤ cku1 (·)f (·)kLps
,
v (X)
w (X)
(2.2)
where the positive constant c does not depend on f .
Proof. Let
Ek ≡ B(x0 , 2k+1 )\B(x0 , 2k ), Gk,1 ≡ B(x0 , 2k−1 /a1 ),
Gk,2 ≡ B(x0 , a1 2k+2 )\B(x0 , 2k−1 /a1 ), Gk,3 ≡ X\B(x0 , a1 2k+2 ).
We obtain
°
°X
≤ c1 °°
ku2 (·)Kf (·)kpLpq
v (X)
k∈Z
Lpq
v (X)
Lv (X)
k∈Z
°
°p
°X
°
+c1 °°
u2 (·)K(f χGk,3 )(·)χEk (·)°° pq
Lv (X)
k∈Z
°p
°
u2 (·)K(f χGk,1 )(·)χEk (·)°°
°
°p
°
°X
°
u2 (·)K(f χGk,2 )(·)χEk (·)°° pq
+c1 °
≡ c1 (S1p + S2p + S3p ).
Now we estimate S1 . Note that
d(x0 , y) <
2k−1
d(x0 , x)
≤
a1
2a1
when x ∈ Ek , and y ∈ Gk,1 . From the latter inequality we have
µB(x, d(x0 , x)) ≤ b1 µB(x, d(x, y)).
Indeed,
d(x0 , x) ≤ a1 (d(x0 , y) + d(y, x)) ≤ a1
Hence
µ
¶
d(x0 , x)
+ a0 d(x, y) .
2a1
1
d(x0 , x) ≤ d(x, y).
2a1 a0
Correspondingly,
µ
¶
< ∞.
d(x0 , x)
≤ b2 µB(x, d(x, y)).
µB(x, d(x0 , x)) ≤ b2 µB x,
2a1 a0
46
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
It is easy to see that
µB(x0 , d(x0 , x)) ≤ b3 µB(x, d(x0 , x))
and, finally, we obtain
µB(x0 , d(x0 , x)) ≤ b4 µB(x, d(x, y)).
By considering the latter inequality we have
Z |f (y)|χ (y)
¯
¯
Gk1
¯
¯
¯K(f χG )(x)¯ ≤ b5
dµ
µB(x, d(x, y))
k1
X
b6
≤
µB(x0 , d(x0 , x))
Z
|f (y)| dµ,
B(x0 ,d(x0 ,x))
when x ∈ Ek , and by Proposition C we obtain
°
°
³
´−1
S1p ≤ c2 °°u2 (·) µB(x0 , d(x0 , ·))
Z
B(x0 ,d(x0 ,x))
≤ c3 ku1 (·)f (·)kpLps
.
w (X)
°p
°
|f (y)| dµ°°
Lpq
v (X)
Now we shall estimate S3p . It is easy to check that if x ∈ Ek and y ∈ Gk,3 , then
d(x0 , y) ≤ d(x, y) and
µB(x0 , d(x0 , y)) ≤ b7 µB(x, d(x, y)).
By virtue of Proposition D we obtain
S3p
°
°
≤ c4 °°u2 (·)
Z
{d(x0 ,y)>d(x0 ,x)}
°
°p
|f (y)|
≤ c5 ku1 (·)f (·)kpLpq
.
dµ°° pq
w (X)
µB(x0 , d(x0 , y))
Lv (X)
Now let us estimate S2p . By Lemma B (part (ii)),
S2p ≤
X°
°
°u2 (·)K(f χG
k∈u
°p
°
)(·)χEk (·)°
k,2
We shall use the following notation:
Lpq
v (X)
≡
X
p
.
Sk,2
k∈u
u2,k ≡ sup u2 (x), σ2,k ≡ sup σ2 (x), u1,k ≡ inf u1 (x), σ1,k ≡ inf σ1 (x).
x∈Ek
x∈Gk,2
x∈Ek
x∈Gk,2
By Theorem C and Lemma A we have
°
°
°
1/p °
Sk,2 ≤ u2,k σ2,k °K(f χGk,2 )(·)°
Lpq
ρ (X)
°
1/p °
°
°
≤ c6 u2,k σ2,k °f (·)χGk,2 (·)°
Lpq
ρ (X)
°
°p
°
°
1/p °
1/p °
°
°
≤ c7 u1,k σ1,k °f (·)χGk,2 (·)° ps
≤ c6 u2,k σ2,k °f (·)χGk,2 (·)° ps
Lρ (X)
Lρ (X)
°
°
°
°
≤ c8 °u1 (·)f (·)χG (·)° ps .
k,2
Lw (X)
By Lemma B (part (i)) we finally obtain
.
S2p ≤ c9 ku1 (·)f (·)kLps
w (X)
WEIGHT INEQUALITIES
47
Remark 2.1. It is easy to check that Theorem 2.4 is still valid if we replace
condition (1) by the condition
(1′ )
sup σ2 (y) sup up2 (y) ≤ bσ1 (x)up1 (x),
Ex
Ex
where b > 0 does not depend on x ∈ X and where
¾
½
d(x0 , x)
≤ d(x0 , y) < 4a1 d(x0 , x) .
Ex = y :
4a1
Indeed, we have
1/p
1/p
Sk,2 ≤ σ2,k u2,k kT (f χGk,2 )(·)kLpq
≤ σ2,k u2,k kT (f χGk,2 )(·)kLpρ (X)
ρ (X)
1/p
≤ b1 σ2,k u2,k kf (·)χGk,2 (·)kLpρ (X)
Z
= b1
Gk,2
³
sup
´³
σ2 (y)
2k ≤d(x0 ,y)0
°
°
× °°
1
ϕ1 (d(x0 , ·))
°
°
χ{d(x0 ,y)≤t} (·)°°
Lp
′ s′
´−1
°
χ{d(x0 ,y)>t} (·)°°
Lpq
v(d(x
0 ,·))
(X)
0. Indeed, by Lemma A
(part (i)) and by the reverse doubling condition for µ we have
°
°
°
°
c ≥ B(t) ≥ °ϕ2 (d(x0 , ·))(µB(x0 , d(x0 , ·))−1 χ{tt} (·)°°
Lp
′ s′
¶
(X)
.
On the other hand, we have
Ã
= s′
Z∞
λ
s′ −1
0
Ã
= s
′
³ n
°
°
°(µB(x0 , d(x0 , ·))−1 χ{d(x
³
µ x:
−1
(µB(x
Z0 ,t))
≤ c4
s′ −1
o
³ n
−1
à (µB(x
Z0 ,t))
0
Lp′ s′ (X)
´´s′ /p′
µB(x0 , d(x0 , x))−1 > λ ∩ {d(x0 , x) ≥ t}
λ
0
°
°
(·)°
0 ,y)>t}
µ x : µB(x0 , d(x0 , x)) < λ
λ
s′ −1 −s′ /p′
λ
dλ
!1/s′
−1
o´s′ /p′
= c5 (µB(x0 , t))−1/p .
dλ
!1/s′
dλ
!1/s′
WEIGHT INEQUALITIES
49
Here we have used the inequality
o
n
µ x : µB(x0 , d(x0 , x)) < λ−1 ≤ bλ−1 ,
where the positive constant b is from the doubling condition for µ. Thus we
obtain
µ
¶
t
1/p
−1
B1 (t) ≤ c6 ϕ2 (t)v (t)ϕ1
≤ c7
8a21
for arbitrary t > 0.
By Theorem 2.4 we conclude that inequality (2.3) holds.
Remark 2.2. All results of this section remain valid if we omit the continuity
of σ, but require that µB(x0 , r) be continuous with respect to r, where x0 is
the same fixed point as above. This follows from the fact that in this case
the sequence σn (d(x0 , x)), where σn are absolutely continuous functions (see
the proof of Theorem 2.1.), converges to σ(d(x0 , x)) a.e. on X. On the other
hand, the continuity of µB(x0 , r) with respect to r is equivalent to the condition
µ{x : d(x0 , x) = r} = 0 for all r > 0.
3. Singular Integrals on Spaces of Nonhomogeneous Type
In this section we present weighted inequalities for Calder´on–Zygmund singular integrals defined on spaces of nonhomogeneous type.
Let (X, d, µ) be a spaces of nonhomogeneous type with metric d and measure
µ (i.e., a0 = 1, a1 = 1, we have equality instead of the inequality in (iii) and
(vii) need not be valid in Definition 1.1) satisfying the condition
µB(x, r) ≤ rα ,
x ∈ X, r > 0,
for some α > 0, where B(x, r) ≡ {y : d(x, y) ≤ r}.
Let the kernel k satisfy the following conditions:
(1)
|k(x, y)| ≤ c1 d(x, y)−α ,
for all x, y ∈ X, x 6= y;
(2) there exist c2 > 0 and ε ∈ (0, 1] such that
n
o
max |k(x, y) − k(x1 , y)|, |k(y, x) − k(y, x1 )| ≤ c2
d(x, x1 )ε
d(x, y)α+ε
whenever d(x, x1 ) ≤ 2−1 d(x, y), x 6= y. Assume also that the integral operator
T f (x) = lim
Z
ε→0
X\B(x0 ,ε)
f (y)k(x, y) dµ
is bounded in L2 (X).
The following theorem is valid.
Theorem C ([18]). The operator T is bounded in Lp (X) for 1 < p < ∞
and is of weak type (1, 1).
50
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
Recall that a ≡ sup{d(x0 , x) : x ∈ X}, where x0 is a given point of X and
B(x0 , R)\B(x0 , r) 6= ∅ whenever 0 < r < R < a.
We remark that the condition µ{x} = 0 is automatically satisfied for all
x ∈ X.
Lemma 3.1. Let 1 < p < ∞, and let w be a weight function on X. Assume
that the following two conditions are satisfied:
(i) there exists a positive increasing function v on (0, 4a) such that for almost
all x ∈ X the inequality
v(2d(x0 , x)) ≤ b1 w(x)
holds, where the positive constant b1 does not depend on x;
(ii)
Z
I(t) ≡
′
w1−p (x) dµ < ∞
B(x0 ,t)
for all t > 0. Then T ϕ(x) exists µ-a.e. for any ϕ ∈ Lpw (X).
Proof. Let 0 < α < a and let us denote
n
o
Sβ = x ∈ X : d(x0 , x) ≥ β/2 .
Suppose ϕ ∈ Lpw (X) and represent ϕ as follows:
ϕ(x) = ϕ1 (x) + ϕ2 (x),
where ϕ1 = ϕχSβ and ϕ2 = ϕ − ϕ1 . Due to condition (i) we see that
Z
X
1
≤ β
v( 2 )
Z
Sβ
v( β2 ) Z
|ϕ1 (x)| dµ = β
|ϕ(x)|p dµ
v( 2 )
p
Sβ
c1 Z
|ϕ(x)| v(2d(x0 , x)) dµ ≤ β
|ϕ(x)|p w(x) dµ < ∞
v( 2 )
p
Sβ
for arbitrary β, 0 < β < a. Consequently T ϕ1 ∈ Lp (X) and, according to
Theorem C, T ϕ1 (x) exists µ-a.e. on X.
Now let x be such that d(x0 , x) > β. If y ∈ X and d(x0 , y) < β2 , then
d(x0 , x) ≤ d(x0 , y) + d(x, y) ≤
d(x0 , x)
+ d(x, y).
2
WEIGHT INEQUALITIES
Hence β/2 <
d(x0 ,x)
2
≤ d(x, y) and we obtain
¯Z
¯
¯
¯
¯
|T ϕ2 (x)| = ¯ ϕ2 (y)k(x, y) dµ¯¯ ≤ c2
X
Z
≤ c3
≤ c4 β
−α
µ
51
B(x0 , β2 )
Z
|ϕ(y)|
dµ ≤ c4 β −α
α
d(x, y)
¶1/p µ
p
|ϕ(y)| w(y) dµ
Z
B(x0 , β2 )
Z
|ϕ(y)|
dµ
d(x, y)α
|ϕ(y)| dµ
B(x0 , β2 )
Z
w
1−p′
¶1/p′
(y) dµ
B(x0 , β2 )
B(x0 , β2 )
< ∞.
Thus T ϕ(x) is absolutely convergent for arbitrary x such that d(x0 , x) > β. As
we can take β arbitrarily small and µ{x0 } = 0, we conclude that T ϕ(x) exists
µ-a.e. on X.
Theorem 3.1. Let 1 < p < ∞. Assume that v is a positive increasing
continuous function on (0, 4a). Suppose that w is a weight on X. Let the
following two conditions hold:
(i) there exists a constant b1 > 0 such that the inequality
v(2d(x0 , x)) ≤ b1 w(x)
is fulfilled for µ-almost all x ∈ X;
(ii)
sup
0 t }
2
Ã
¶
p
{x: d(x0 ,x)>t}
Ã
φ(t)
Z
!
¯p
¯
f (y)k(x, y) dµ¯¯ dµ dt ≡ I21 + I22 .
Z
{x: d(x0 ,x)>t} {y: d(x0 ,y)≤ t }
2
Using again the boundedness of T in Lp (X) we obtain
I21 ≤ c4
Za
µ
Z
φ(t)
0
≤ c4
¶
|f (y)| dµ dt = c4
d(x0 ,y)> 2t }
{y:
Z
p
|f (y)|
Z
|f (y)|p w(y) dµ.
p
¶
φ(t) dt dµ
0
X
|f (y)|p v(2a1 d(x0 , y)) dµ ≤ c5
µ 2d(x
Z 0 ,y)
Z
X
X
Now let us estimate I22 . When d(x0 , x) > t and d(x0 , y) ≤
t
2
we have
d(x0 , x) ≤ d(x0 , y) + d(y, x) = d(x0 , y) + d(x, y)
≤
t
d(x0 , x)
+ d(x, y) ≤
+ d(x, y).
2
2
Hence
d(x0 , x)
≤ d(x, y).
2
Consequently,
I22 ≤ c6
Z
µ
µ
Z
1
dµ
d(x0 , x)αp
φ(t)
Za
φ(t)
0
≤ c7
Ã
Za
{x: d(x0 ,y)>t}
0
{x: d(x0 ,x)>t}
{y:
Z
|f (y)|
dµ(y)
d(x, y)α
d(x0 ,y)≤ 2t }
¶µ
Z
¶p
!
dµ(x) dt
¶p
|f (y)| dµ
B(x0 ,t)
dt.
It is easy to see that for any s, 0 < s < a, we have
Za
≤
{x: d(x0 ,x)≥s}
Z
φ(t)
s
Z
µ
1
d(x0 , x)αp
{x: d(x0 ,x)>t}
µ d(x
Z 0 ,x)
¶
1
dµ dt
d(x0 , x)αp
¶
φ(t) dt dµ ≤
s
Z
{x: d(x0 ,x)≥s}
v(d(x0 , x))
dµ.
d(x0 , x)αp
WEIGHT INEQUALITIES
53
Finally, using Proposition A, we obtain
I22 ≤ c8
Z
|f (x)|p w(x) dµ.
X
Lemma 3.2. Let a = ∞, 1 < p < ∞, and let w be a weight function on X.
Suppose the following conditions are fulfilled:
(i) there exists a positive decreasing function v on (0, ∞) such that
¶
µ
d(x0 , x)
≤ cw(x) a.e.;
v
2
(ii) for all t > 0
Z
′
′
w1−p (x)(d(x0 , x))−αp dµ < ∞.
X\B(x0 ,t)
Then T ϕ(x) exists µ-a.e. for arbitrary ϕ ∈ Lpw (X).
Proof. Fix arbitrarily β > 0 and let
n
o
Sβ = x : d(x0 , x) ≥ β .
Represent ϕ as follows:
ϕ(x) = ϕ1 (x) + ϕ2 (x),
where ϕ1 (x) = ϕ(x)χSβ (x) and ϕ2 (x) = ϕ(x) − ϕ1 (x).
For ϕ2 we have
Z
|ϕ2 (x)|p dµ =
X
1
≤
v(β)
Z
v(β)
v(β)
Z
|ϕ(x)|p dµ
B(x0 ,β)
c1 Z
|ϕ(x)| v(d(x0 , x)) dµ ≤
v(β)
p
B(x0 ,β)
|ϕ(x)|p w(x) dµ < ∞.
B(x,β)
Consequently, ϕ2 ∈ Lp (X) and by Theorem C we have T ϕ2 ∈ Lp (X). Hence
T ϕ2 (x) exists a.e. on X.
Now let x ∈ X and let d(x0 , x) < β/2. If d(x0 , y) ≥ β, then
d(x0 , y)
≥ d(x, y).
2
Using this inequality, we obtain
|T ϕ1 (x)| ≤ c2
Z
Sβ
≤ c3
µZ
Sβ
p
Z
|ϕ(y)|
|ϕ(y)|
dµ
≤
c
dµ
3
d(x, y)α
d(x0 , y)α
Sβ
¶1/p µ Z
|ϕ(y)| w(y) dµ
Sβ
w
1−p′
−αp′
(y)d(x0 , y)
¶1/p′
dµ
< ∞.
As we may take β arbitrarily large, we conclude that T ϕ(x) exists a.e.
54
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
Theorem 3.2. Let a = ∞, 1 < p < ∞ and let v be a positive continuous
decreasing function (0, ∞). Suppose that w is a weight function on X and the
following conditions are satisfied:
(i) for almost all x
v(d(x0 , x)/2) ≤ cw(x);
(ii)
sup
t>0
µ
Z
¶1/p µ
Z
v(d(x0 , x)) dµ
B(x0 ,t)
w
1−p′
−αp′
(x)d(x0 , x)
¶1/p′
dµ
X\B(x0 ,t)
< ∞.
Then the operator T is bounded from Lpw (X) to Lpv(d(x0 ,·)) (X).
Proof. Without loss of generality we can represent v as
v(t) = v(+∞) +
Z∞
φ(τ ) dτ,
φ ≥ 0.
t
Further,
Z
|T f (x)|p v(d(x0 , x)) dx
X
= v(+∞)
Z
p
|T f (x)| dµ +
Z
p
|T f (x)|
X
X
Z∞
µ
¶
φ(t)dt dµ ≡ I1 + I2 .
d(x0 ,x)
If v(+∞) = 0, then I1 = 0. But if v(+∞) 6= 0, then by virtue of the boundedness of T in Lp (X) we have
I1 ≤ c1 v(+∞)
Z
|f (x)|p dµ ≤ c1
Z
|f (x)|p v(d(x0 , x)) dµ ≤ c2
|f (x)|p w(x) dµ.
X
X
X
Z
Now we pass to I2 :
I2 =
Z∞
Z∞
0
(1)
where ft
Z
φ(t)
0
+ c3
µ
¶
p
|T f (x)| dµ dt ≤ c3
B(x0 ,t)
µ
φ(t)
Z
(2)
0
= c4
Z
X
φ(t)
0
Z
(1)
¶
|T ft (x)|p dµ dt
B(x0 ,t)
¶
(2)
= f χB(x0 ,2t) and ft
I21 ≤ c4
µ
φ(t)
|T ft (x)|p dµ dt = I21 + I22 ,
B(x0 ,t)
Z∞
Z∞
µ
(1)
= f − ft . Again using Theorem C we have
Z
¶
p
|f (y)| dµ dt
B(x0 ,2t)
p
|f (y)|
µ
Z∞
d(x0 ,x)
2
¶
φ(t) dt dµ ≤ c5
Z
X
|f (y)|p w(y) dµ.
WEIGHT INEQUALITIES
55
It remains to estimate I22 . If x ∈ B(x0 , t) and y ∈ X\B(x0 , 2t), then
d(x0 , y)
≤ d(x, y).
2
Consequently,
I22 ≤ c5
Z∞
µ
Z
φ(t)
0
¶µ
Z
dµ
−α
|f (y)|d(x0 , y)
dµ
X\B(x0 ,2a1 t)
B(x0 ,t)
¶p
dt.
Moreover,
Zs
0
µ
φ(t)
Z
¶
dµ dt =
B(x0 ,t)
Z
B(x0 ,s)
µ
Zs
¶
φ(t) dt dµ ≤
d(x0 ,x)
Z
v(d(x0 , x)) dµ
B(x0 ,s)
and due to Proposition B we finally obtain the boundedness of T .
Now we are going to establish weighted estimates for the operator T in
Lorentz spaces defined on spaces of nonhomogeneous type. Theorem C and
the interpolation theorem imply
Proposition E. Let 1 < p, q < ∞. Then T is bounded in Lpq (X).
The following lemmas are obtained in the same way as in the homogeneous
case. Instead of Theorem A we need to use Theorem C.
Lemma 3.3. Let 1 < s ≤ p < ∞. Let the weight functions w and w1 satisfy
the conditions:
(1) there exists an increasing function v on (0, 4a) such that the inequality
v(d(x0 , x)) ≤ bw(x)w1 (x)
holds for almost all x ∈ X;
(2) for every t, 0 < t < a, the norm
°
°
°
°
°
°
1
χ{d(x0 ,y)≤t} (·)°° p′ s′
w(·)w1 (·)
Lw (X)
is finite.
Then T g(x) exists a.e. on X for any g satisfying the condition
kg(·)w1 (·)kLps
< ∞.
w (X)
Lemma 3.4. Let 1 < s ≤ p < ∞. Suppose also that u and u1 are positive
increasing functions on (0, 4a1 a) and
°
°
°
°
°
°
1
χB(x ,t) (·)°° p′ s′
0
u(d(x0 , ·))u1 (d(x0 , ·))
Lu(d(x
0
0,
°
°
° d(x0 , ·)−α
°
°
χX\B(x ,t) (·)°°
°
0
w(·)w1 (·)
< ∞.
′ ′
p s
Lw
(X)
Then T g(x) exists a.e. on X for arbitrary g satisfying kg(·)w1 (·)kLps
< ∞.
w (X)
From the previous lemmas easily follows
Lemma 3.6. Let a = ∞, 1 < s ≤ p < ∞. Suppose also that for the
decreasing functions u and u1 on (0, ∞) the following condition is satisfied
°
°
°
°
°
°
d(x0 , ·)−α
χX\B(x ,t) (·)°° p′ s′
0
u(d(x0 , ·))u1 (d(x0 , ·))
Lu(d(x
0 ,·))
< ∞,
(X)
for all t > 0. Then T g(x) exists a.e. on X for g satisfying the condition
g(·)u1 (d(x0 , ·)) ∈ Lps
u(d(x0 ,·)) (X).
Using Propositions E and C, we obtain the following result in the same way
as Theorem 2.1.
Theorem 3.3. Let 1 < s ≤ p ≤ q < ∞, and let w be a weight function
on X. Assume that v is a positive increasing continuous function on (0, 4a).
Suppose also that the following two conditions are satisfied:
(1) there exists a positive constant c such that the inequality
v(2a1 d(x0 , x)) ≤ cw(x)
holds for almost every x ∈ X;
(2)
°
°
sup °(d(x0 , ·))−α χX\B(x
0 0
sup v 1/p (x) sup u2 (x) ≤ b inf w1/p (x) inf u1 (x)
t
a1
holds, where Ft = {x ∈ X :
(2)
°
°
sup °u2 (·)(d(x0 , ·))−α χ
t>0
(3)
°
°
sup °u2 (·)χB(x
t>0
0
Ft
Ft
Ft
°
°
(·)°
,t)
≤ d(x0 , x) < 8a1 t};
X\B(x0 ,t)
Lpq
v (X)
Ft
°
°
(·)°
°
°
°
pq
Lv (X) °
°
°
1
< ∞;
χB(x ,t) (·)°° p′ s′
0
u1 (·)w(·)
Lw (X)
°
°
°(d(x0 , ·))−α (u1 (·)w(·))−1 χ
X\B(x0
Then the following inequality holds:
°
°
(·)°
,t)
′ ′
Lpw s (X)
< ∞.
ku2 (·)T f (·)kLpq
≤ cku1 (·)f (·)kLps
,
v (X)
w (X)
where the positive constant c does not depend on f .
Remark 3.1. The results of this section remain valid if we do not require the
continuity of v, but assume that µB(x0 , r) is continuous with respect to r.
4. Examples of weight functions
Let (X, dµ) be an SHT such that the condition µ(B(x, r)) ≈ r holds (if
a < ∞, then we assume that this condition is fulfilled for 0 < r ≤ 1). It is
known (see [5]) that if µ(X) < ∞, then
Z
p
|Kf (x)| (d(x0 , x))
p−1
dµ ≤ c
X
Z
|f (x)|p (d(x0 , x))p−1 logp
X
b
dµ,
d(x0 , x)
′
where x0 ∈ X, 1 < p < ∞, and b = 8a1 aep and the positive constant c does
not depend on f .
Example 4.1. Let 1 < p < q < ∞, and let a < ∞; put v(t) = tp−1 ,
γ
w(t) = tp−1 lnγ bt for t ∈ (0, 4a1 a), where b = 8a1 ae p−1 , γ = pq + p − 1, and a1
is the positive constant from Definition 1.1. Then from Theorem 2.1 it follows
that
kKf (·)kLpq
v(d(x
0 ,·))
(X)
≤ ckf (·)kLpw(d(x
where the positive constant c does not depend on f .
0 ,·))
(X) ,
(4.1)
58
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
Example 4.2. Let 1 < p < q < ∞, a = ∞, v(t) = tp−1 when 0 < t ≤ 1
γ
p−1
and v(t) = tα when t > 1, and let w(t) = tp−1 lnγ 2e t when t ≤ 1, and
γ
w(t) = tβ lnβ (2e p−1 ) when t > 1, where γ = pq + p − 1, 0 < α ≤ β < p − 1. Then
inequality (4.1) holds.
An appropriate example for the conjugate function
1
fe(x) =
π
Zπ
−π
f (x + t)
dt
2 tg 2t
is presented in [11].
References
1. H.-M. Chung, R. A. Hunt, and D. S. Kurtz, The Hardy-Littlewood maximal function
on L(p, q) spaces with weights. Indiana Univ. Math. J. 31(1982), 109–120.
2. R. R. Coifman and G. Weiss, Extensions of Hardy spaces and their use in analysis.
Bull. Amer. Soc. 83(1977), 569–645.
3. D. E. Edmunds and V. Kokilashvili, Two-weight inequalities for singular integrals.
Canadian Math. Bull. 38(1995), 119–125.
4. D. E. Edmunds, V. M. Kokilashvili, and A. Meskhi, Boundedness and compactness of Hardy-type operators on Banach function spaces defined on measure space. Proc.
A. Razmadze Math. Inst. 117(1998), 7–30.
5. D. E. Edmunds, V. Kokilashvili, and A. Meskhi, Two weight estimates for singular
integrals defined on spaces of homogeneous type. Canadian J. Math. 52(2000), No. 3,
468–502.
6. I. Genebashvili, A. Gogatishvili, V. Kokilashvili, and M. Krbec, Weight theory
for integral transform on spaces of homogeneous type. Pitman Monographs and Surveys
in Pure and Applied Mathematics, Longman, Harlow, 1998.
7. V. Guliev, Two-weight Lp inequality for singular integral operator on Heisenberg groups.
Georgian Math. J. 1(1994), No. 4, 367–376.
8. E. G. Gusseinov, Singular integrals in the space of function summable with monotone
weight. (Russian) Mat. Sb. 132(174)(1977), No. 1, 28–44.
9. S. Hoffman, Weighted norm inequalities and vector-valued inequalities for certain rough
operators. Indiana Univ. Math. J. 42(1993), 1–14.
10. R. A. Hunt, on L(p, q) spaces. L’ǫnseign. Math. 12(1966), 249–276.
11. V. Kokilashvili, On two-weight inequalities for conjugate functions. Proc. A. Razmadze
Math. Inst. 112(1997), 141–142.
12. V. Kokilashvili and M. Krbec, Weighted inequalities in Lorentz and Orlicz spaces.
World Scientific, Singapore, New Jersey, London, Hong Kong, 1991.
13. V. Kokilashvili and A. Meskhi, Two-weight inequalities for singular integrals defined
on homogeneous groups. Proc. A. Razmadze Math. Inst. 112(1997), 57–90.
14. V. Kokilashvili and A. Meskhi, Two-weight inequalities for Hardy type transforms
and singular integrals defined on homogeneous type spaces. Proc A. Razmadze Math.
Inst. 114(1997), 119–123.
WEIGHT INEQUALITIES
59
15. V. Kokilashvili and A. Meskhi, Two-weight estimates for singular integrals on homogeneous groups. (Russian) Doklady Akademii Nauk 354(1997), No. 3, 301–303. English
transl. in Doklady Mathematics 55(1997), 362–364.
16. V. Kokilashvili and A. Meskhi, Two-weight inequalities for integral operators in
Lorentz spaces defined on homogeneous groups. Georgian Math. J. 6(1999), No. 1, 65–82.
17. B. Muckenhoupt and R. L. Wheeden, Two-weight function norm inequalities for the
Hardy–Littlewood maximal function and the Hilbert transform. Studia Math. 55(1976),
No. 3, 279–294.
18. F. Nazarov, S. Treil, and A. Volberg, Weak type estimates and Cotlar inequalities
for Calder´on–Zygmund operators on nonhomogeneous spaces. International Mathematics
Research Notices 9(1998), 463–487.
19. Y. Rakotondratsimba, On boundedness of classical operators on weighted Lorentz
spaces. Georgian Math. J. 5(1998), No. 2, 177–200.
20. E. T. Sawyer and R. L. Wheeden, Weighted inequalities for fractional integrals on
Euclidean and homogeneous spaces. Amer. J. Math. 114(1992), 813–874.
21. E. M. Stein, Harmonic analysis: real variable methods, orthogonality and oscillatory
integrals. Princeton University Press, Princeton, New Jersey, 1993.
22. I. M. Stein and G. Weiss, Introduction to Fourier analysis on Euclidean spaces. Princeton, New Jersey, Princeton University Press, 1971.
¨ mberg and A. Torchinsky, Weighted Hardy spaces. Lecture Notes in Math.
23. J. O. Stro
1381, Springer Verlag, Berlin, 1989.
(Received 23.08.2000)
Authors’ Addresses:
D. E. Edmunds
Centre for Mathematical Analysis
and its Applications
University of Sussex
Brighton BN1 9QH
Sussex
United Kingdom
E-mail: [email protected]
V. Kokilashvili and A. Meskhi
A. Razmadze Mathematical Institute
Georgian Academy of Sciences
1, M. Aleksidze St., Tbilisi 380093
Georgia
E-mail: [email protected]
[email protected]
Volume 8 (2001), Number 1, 33-59
WEIGHT INEQUALITIES FOR SINGULAR INTEGRALS
DEFINED ON SPACES OF HOMOGENEOUS AND
NONHOMOGENEOUS TYPE
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
Dedicated to Professor Hans Triebel on his 65th birthday
Abstract. Optimal sufficient conditions are found in weighted Lorentz spaces for weight functions which provide the boundedness of the Calder´on–
Zygmund singular integral operator defined on spaces of homogeneous and
nonhomogeneous type.
2000 Mathematics Subject Classification: 42B20, 42B25.
Key words and phrases: Singular Integral, maximal function, Hardytype operator, space of homogeneous type, space of nonhomogeneous type,
Lorentz space.
Introduction
In the present paper optimal sufficient conditions are found for the weight
functions which provide the boundedness of Calder´on–Zygmund singular integral operator defined on spaces of homogeneous as well as nonhomogeneous type
in weighted Lorentz spaces. In the nonhomogeneous case the results are new
even in Lebesgue spaces.
Two-weight strong type inequalities for singular integrals in Lebesgue spaces
on spaces of homogeneous type (SHT) were established in [5] (see also [6],
Chapter 9, and [14]), while a similar problem for singular integrals defined on
homogeneous groups was considered in [16] in the case of Lorentz spaces.
Two-weight inequalities for Hilbert transforms with monotonic weights were
studied in [17]. Analogous problems for singular integrals in Euclidean spaces
were considered in [8] and generalized in [7] for singular integrals on Heisenberg
groups. Optimal conditions for a pair of weights ensuring the validity of twoweight inequalities for Calder´on–Zygmund singular integrals were obtained in
[3] (and also in [19] for the case of Lorentz spaces). The latter result was
generalized to the setting of homogeneous groups in [13], [15].
Finally, we would like to mention that singular integrals on spaces of nonhomogeneous type (measure spaces with a metric) were studied in [18].
c Heldermann Verlag www.heldermann.de
ISSN 1072-947X / $8.00 / °
34
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
1. Preliminaries
In this section we give the notion of SHT and defined Lorentz space. The
definition of a singular integral and several well-known results concerning the
Hardy-type operators on SHT are also given.
Definition 1.1. A space of homogeneous type (SHT) (X, d, µ) is a topological space X endowed with a complete measure µ such that: (a) the space of
compactly supported continuous functions is dense in L1 (X, µ), and (b) there
exists a non-negative real-valued function (quasimetric) d : X × X → R1 satisfying:
(i) d(x, x) = 0 for arbitrary x ∈ X;
(ii) d(x, y) > 0 for arbitrary x, y ∈ X, x 6= y;
(iii) there exists a positive constant a0 such that the inequality
d(x, y) ≤ a0 d(y, x)
holds for all x, y ∈ X;
(iv) there exists a positive constant a1 such that the inequality
d(x, y) ≤ a1 (d(x, z) + d(z, y))
holds for arbitrary x, y, z ∈ X;
(v) for every neighborhood V of any point x ∈ X there exists a number r > 0
such that the ball
n
o
B(x, r) = y ∈ X : d(x, y) < r
with center in x and radius r is contained in V ;
(vi) the balls B(x, r) are measurable for all x ∈ X, r > 0 and, moreover,
0 < µB(x, r) < ∞;
(vii) there exists a positive constant b such that the inequality (doubling
condition)
µB(x, 2r) ≤ bµB(x, r)
is true for all x ∈ X and for all positive r.
It will be supposed that for some x0 ∈ X
µ{x0 } = µ{x ∈ X : d(x0 , x) = a} = 0,
where
n
o
a ≡ sup d(x0 , x) : x ∈ X .
Note that the condition a = ∞ is equivalent to the condition µ(X) = ∞, and
if a = ∞, then µ{x ∈ X : d(x0 , x) = a} = 0.
For the definitions, various examples and properties of SHT, we refer to [2],
[23] (also to [6]).
Definition 1.2. By a spaces of nonhomogeneous type we mean a measure
space with a quasimetric (X, d, µ) satisfying conditions (i)–(v) of Definition 1.1,
i.e., the doubling condition is not assumed and may fail.
WEIGHT INEQUALITIES
35
In the sequel it will be assumed that there exists a point x0 ∈ X such that
B(x0 , R)\B(x0 , r) 6= ∅
(1)
for all r and R provided 0 < r < R < a.
In the remaining part of this section (X, d, µ) is assumed to be a space of
nonhomogeneous type.
Definition 1.3. An almost everywhere positive, locally integrable function
w : X → R1 is called a weight.
For weight functions w, we denote by Lpq
w (X) the Lorentz space with weight
w which represents the class of all µ-measurable functions f : X → R1 for which
à +∞
Z µ
Z
kf kLpq
= q
w (X)
0
¶q/p
w(x) dµ
{x∈X: |f (x)|>λ}
λ
q−1
dλ
!1/q
0
Z
¶1/p
w(x) dµ
{x∈X: |f (x)|>λ}
0.
In the sequel we shall need the following results concerning the Hardy-type
operators on the measure space (X, µ)
Hf (t) =
Z
f (x) dµ(x), t ∈ (0, a),
Z
f (x) dµ(x), t ∈ (0, a).
{x: d(x0 ,x)t}
The following result is from [5].
Proposition A. Let 1 < p ≤ q < ∞, µ{x0 } = 0. Then the inequality
µ Za
q
¶1/q
v(t)|Hf (t)| dt
0
µZ
≤c
¶1/p
p
|f (x)| w(x) dµ
X
,
holds with some c > 0 independent of f , f ∈ Lpw (X), if and only if
D = sup
0t} (·)°°
°
w(·)w1 (·)
′ ′
p s
Lw
(X)
< ∞,
then Kg(x) exists a.e. on X for arbitrary g provided that kg(·)w1 (·)kLps
< ∞.
w (X)
From Lemma 2.3 the next lemma follows easily:
WEIGHT INEQUALITIES
41
Lemma 2.4. Let a = ∞, 1 < s ≤ p < ∞. Assume that u and u1 are positive
decreasing functions on (0, ∞) and the condition
°
°
°
°
°
°
µB(d(x0 , ·))−1
χ{d(x0 ,y)>t} (·)°° p′ s′
u(d(x0 , ·))u1 (d(x0 , ·))
Lu(d(x
0. Then Kg(x) exists a.e. on X for arbitrary g satisfying the
condition g(·)u1 (d(x0 , ·)) ∈ Lps
u(d(x0 ,·)) (X).
Now we pass to the weight inequalities for the operator K.
Theorem 2.1. Let 1 < s ≤ p ≤ q < ∞, let w be a weight function on
X, suppose that σ is a positive increasing continuous function on (0, 4a1 a),
ρ ∈ Ap (X) and v(x) = σ(d(x0 , x))ρ(x). Suppose the following conditions are
satisfied:
(a) there exists a positive constant b such that the inequality
σ(2a1 d(x0 , x))ρ(x) ≤ bw(x)
holds for almost every x ∈ X;
(b)
°
° 1
°
pq
Lv (X) °
°
°
°
,y)>t}
°
B ≡ sup °°µB(x0 , d(x0 , ·))−1 χ{d(x
0
0λ}
ρ(x)
µ d(x
Z 0 ,x)
¶
!q/p
ϕ(t) dt dµ
0
dλ
!1/q
!1/q
dλ
≡ I1 + I2 .
In the case, where σ(0+) = 0, we have I1 = 0; otherwise, by Theorem B and
Lemma A (part (ii)), we have
I1 = c1 σ(0+)1/p kKf (·)kLpq
≤ c2 σ 1/p (0+)kf (·)kLpq
ρ (X)
ρ (X)
≤ c2 σ 1/p (0+)kf (·)kLps
≤ c2 kf (·)kLps
.
ρ (X)
w (X)
42
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
Now, let us estimate I2 . Set
f1t (x) = f (x)χ{d(x
t
0 ,x)> 2a1 }
, f2t (x) = f (x) − f1t (x).
We have
à Z∞
I2 = c1 q
λ
à Za
q−1
0
à Z∞
≤ c3 q
λ
q−1
0
à Z∞
+c3 q
λ
q−1
0
ϕ(t)
χ{x: |Kf
λ
1t (x)|> 2 }
{x: d(x0 ,x)>t}
µ
Z
ϕ(t)
0
Z
χ{x: |Kf
λ
2t (x)|> 2 }
{x: d(x0 ,x)>t}
I21 ≤ c4
à Z∞
ϕ(t)
µ Za
ϕ(t)kf1t (·)kpLpq
ρ (X)
0
0
≤ c5
λ
q−1
µ
≤ c5
λ
s−1
0
Ã
Z
≤ c5 kf (·)k
≤ c5
µ Za
dλ
!1/q
dλ
dλ
!p/q
!1/p
dt
ϕ(t)kf1t (·)kpLps
ρ (X)
d(x0 ,x)
µ 2a1 Z
¶
!s/p
ϕ(t) dt dµ
0
¶1/p
dt
dλ
!1/s
≤ c6 kf (·)kLps
.
w (X)
Next we shall estimate I22 . Note that if d(x0 , x) > t and d(x0 , y) <
t
2a1
d(x0 , x) ≤ a1 (d(x0 , y) + d(y, x)) ≤ a1 (d(x0 , y) + a0 d(x, y))
≤ a1
Hence
µ
¶
µ
¶
d(x0 , x)
t
+ a0 d(x, y) ≤ a1
+ a0 d(x, y) .
2a1
2a1
d(x0 , x)
≤ d(x, y),
2a1 a0
and we also have
!1/q
≡ I21 + I22 .
0
ρ(x)
!1/q
≥ 1) and using Theorem B
ρ(x) dµ
¶1/p
{x∈X: |f (x)|>λ}
Lps
w (X)
!q/p
¶q/p
}
|Kf1t (x)|> λ
2
dt
0
à Z∞
Z
{x∈X:
ρ(x) dµ dt
ρ(x) dµ dt
p
s
!q/p
¶
¶
Applying the Minkowki’s inequality twice ( pq ≥ 1,
and Lemma A (part (ii)), we obtain
à Za
dλ
ρ(x) dµ dt
{x: d(x0 ,x)>t,|Kf (x)|>λ}
µ
!q/p
¶
Z
ϕ(t)
0
à Za
0
à Za
µ
µ µ
µ(B(x, d(x0 , x))) ≤ bµ B x,
d(x0 , x)
2a1 a0
¶¶
≤ bµ(B(x, d(x, y))).
Moreover, it is easy to show that
µB(x0 , d(x0 , x)) ≤ b1 µB(x, d(x0 , x)).
Hence
µB(x0 , d(x0 , x)) ≤ b2 µB(x, d(x, y)).
, then
WEIGHT INEQUALITIES
43
Taking into account this inequality and using Proposition C we see that
°
°
I22 ≤ c7 °°
1
µB(x0 , d(x0 , ·))
Z
{d(x0 ,y)0
0 ,y)>t}
(·)µ(B(x0 , d(x0 , ·)))
°
°
−1 °
°
° 1
°
pq
Lv (X) °
n
w(·)
°
°
χ{d(x0 ,y)≤t} (·)°°
′ ′
Lpw s (X)
.
By virtue of what has been proved, if B < ∞, then the following inequality
holds:
≤ ckf (·)kLps
,
kKf (·)kLpq
w (X)
vn (X)
where the constant c > 0 does not depend on n.
By passing to the limit as n → ∞, we obtain inequality (2.1).
R∞
Using the representation σ(t) = σ(+∞)+ ψ(τ ) dτ , where σ(+∞) = lim σ(t)
t→∞
t
and ψ ≥ 0 on (0, ∞) and Proposition D, we obtain the following result.
Theorem 2.2. Let a = ∞, 1 < s ≤ p ≤ q < ∞, and suppose that σ is a
positive decreasing continuous function on (0, ∞). Assume that ρ ∈ Ap (X) and
v(X) = σ(d(x0 , x))ρ(x). Suppose the following conditions are satisfied:
(a) there exists a positive constant b such that the inequality
µ
¶
d(x0 , x)
ρ(x)σ
≤ bw(x)
2a1
is true for almost every x ∈ X;
(b)
°
° (µB(x0 , d(x0 , x)))−1
°
B ′ = sup kχ{d(x0 ,y)≤t} (·)kLpq
v (X) °
t>0
Then inequality (2.1) holds.
w(·)
°
°
χ{d(x0 ,y)>t} (·)°°
′ ′
p s
Lw
(X)
< ∞.
Now let us consider the particular cases of Theorems 2.1 and 2.2.
Theorem 2.3. Let a = ∞, 1 < p ≤ q < ∞, suppose σ1 and σ2 are positive,
increasing functions on (0, ∞), let σ1 be a continuous function and suppose that
ρ ∈ Ap (X). Put v(x) = σ2 (d(x0 , x))ρ(x), w(x) = σ1 (d(x0 , x))ρ(x). If
°
sup °°(µB(x0 , d(x0 , ·)))−1 χ{d(x
t>0
0
°
° 1
°
pq
Lv (X) °
°
(·)°°
,y)>t}
w(·)
°
°
χ{d(x0 ,y)≤t} (·)°°
′
Lpw (X)
< ∞,
44
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
then there exists a positive constant c such that
kKf (·)kLpq
≤ ckf (·)kLpw (X)
v (X)
for all f ∈ Lpw (X).
Proof. By Theorem 2.1 it is sufficient to show that there exists a positive
constant b such that σ2 (2a1 t) ≤ bσ1 (t) for all t ∈ (0, ∞).
From the doubling condition (see (vii) in Definition 1.1) and the condition
(1.1) it follows that the measure µ satisfies the reverse doubling condition at
the point x0 (see, e.g., [20] and [23], Lemma 20). In other words, there exist
constants η1 > 1 and η2 > 1 such that
µ(B(x0 , η1 r)) ≥ η2 µ(B(x0 , r))
for all r > 0.
Applying the H¨older’s inequality and using Lemma C, the reverse doubling
′
condition and the fact that ρ1−p ∈ Ap′ (X), we obtain
´−p
σ2 (2a1 t) ³
≤ µ(B(x0 , 2a1 η1 t)\B(x0 , 2a1 t))
σ1 (t)
µ
Z
×
≤
µµ
1−p′
ρ
µ
¶
¶p−1
(x) dµ
¶−p µ
µ
×
Z
¶p−1
B(x0 ,t)
0 ,2a1 η1 t)\B(x0 ,2a1
°³
´−1
°
≤ c2 ° µB(x0 , d(x0 , ·)) χ{d(x
0
¶
ρ(x) dµ
B(x0 ,2a1 η1 t)\B(x0 ,2a1 t)
′
°
≤ c1 (µB(x0 , 2a1 ηt))−p °°χ(B(x
ρ(x) dµ
σ2 (2a1 t)
σ1 (t)
Z
ρ1−p (x) dµ
σ2 (2a1 t)
σ1 (t)
°p
°
°
° 1
°
χB(x0 ,t) (·)°° p′
°
Lv (X) w(·)
Lw (X)
°p
°
1
χB(x0 ,t) (·)°° p′
≤ c.
°p
°
(·)
° pq
t))
°
°
°
Lv (X) ° w(·)
°p
°
(·)
° pq
,y)>t}
¶
B(x0 ,2a1 η1 t)\B(x0 ,2a1 t)
B(x0 ,2a1 η1 t)\B(x0 ,2a1 t)
1
µB(x0 , 2a1 ηt)
1−
η2
Z
Lw (X)
An analogous theorem dealing with singular integrals on homogeneous groups
was proved in [16], and for singular integrals on SHT in the Lebesgue space in [5].
Theorem 2.4. Let a = ∞, 1 < s ≤ p ≤ q < ∞; suppose σ1 , σ2 , u1 and
u2 are weight functions defined on X. Let ρ ∈ Ap (X), and suppose v = σ2 ρ,
w = σ1 ρ. Assume that the following conditions are fulfilled:
(1) there exists a positive constant b such that for all t > 0
1/p
1/p
sup σ2 (x) sup u2 (x) ≤ b inf σ1 (x) inf u1 (x)
Ft
holds, where Ft = {x ∈ X :
Ft
t
a1
Ft
≤ d(x0 , x) < 8a1 t};
Ft
WEIGHT INEQUALITIES
45
(2)
°
°
³
´−1
sup °u2 (·) µ(B(x0 , d(x0 , ·)))
t>0
(3)
°
°
sup °u2 (·)χ{d(x
0
t>0
Then
°
°
×°°
1
u1 (·)w(·)
°
°
χ{d(x0 ,t)>t} (·)°
Lpq
v (X)
°
°
χ{d(x0 ,y)≤t} (·)°°
′ ′
p s
Lw
(X)
< ∞;
°
°
° (µB(x0 , d(x0 , ·)))−1
°
°
χ{d(x0 ,y)>t} (·)°°
pq
°
Lv (X)
°
°
°
(·)
,t)≤t}
u1 (·)w(·)
′ ′
p s
Lw
(X)
ku2 (·)Kf (·)kLpq
≤ cku1 (·)f (·)kLps
,
v (X)
w (X)
(2.2)
where the positive constant c does not depend on f .
Proof. Let
Ek ≡ B(x0 , 2k+1 )\B(x0 , 2k ), Gk,1 ≡ B(x0 , 2k−1 /a1 ),
Gk,2 ≡ B(x0 , a1 2k+2 )\B(x0 , 2k−1 /a1 ), Gk,3 ≡ X\B(x0 , a1 2k+2 ).
We obtain
°
°X
≤ c1 °°
ku2 (·)Kf (·)kpLpq
v (X)
k∈Z
Lpq
v (X)
Lv (X)
k∈Z
°
°p
°X
°
+c1 °°
u2 (·)K(f χGk,3 )(·)χEk (·)°° pq
Lv (X)
k∈Z
°p
°
u2 (·)K(f χGk,1 )(·)χEk (·)°°
°
°p
°
°X
°
u2 (·)K(f χGk,2 )(·)χEk (·)°° pq
+c1 °
≡ c1 (S1p + S2p + S3p ).
Now we estimate S1 . Note that
d(x0 , y) <
2k−1
d(x0 , x)
≤
a1
2a1
when x ∈ Ek , and y ∈ Gk,1 . From the latter inequality we have
µB(x, d(x0 , x)) ≤ b1 µB(x, d(x, y)).
Indeed,
d(x0 , x) ≤ a1 (d(x0 , y) + d(y, x)) ≤ a1
Hence
µ
¶
d(x0 , x)
+ a0 d(x, y) .
2a1
1
d(x0 , x) ≤ d(x, y).
2a1 a0
Correspondingly,
µ
¶
< ∞.
d(x0 , x)
≤ b2 µB(x, d(x, y)).
µB(x, d(x0 , x)) ≤ b2 µB x,
2a1 a0
46
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
It is easy to see that
µB(x0 , d(x0 , x)) ≤ b3 µB(x, d(x0 , x))
and, finally, we obtain
µB(x0 , d(x0 , x)) ≤ b4 µB(x, d(x, y)).
By considering the latter inequality we have
Z |f (y)|χ (y)
¯
¯
Gk1
¯
¯
¯K(f χG )(x)¯ ≤ b5
dµ
µB(x, d(x, y))
k1
X
b6
≤
µB(x0 , d(x0 , x))
Z
|f (y)| dµ,
B(x0 ,d(x0 ,x))
when x ∈ Ek , and by Proposition C we obtain
°
°
³
´−1
S1p ≤ c2 °°u2 (·) µB(x0 , d(x0 , ·))
Z
B(x0 ,d(x0 ,x))
≤ c3 ku1 (·)f (·)kpLps
.
w (X)
°p
°
|f (y)| dµ°°
Lpq
v (X)
Now we shall estimate S3p . It is easy to check that if x ∈ Ek and y ∈ Gk,3 , then
d(x0 , y) ≤ d(x, y) and
µB(x0 , d(x0 , y)) ≤ b7 µB(x, d(x, y)).
By virtue of Proposition D we obtain
S3p
°
°
≤ c4 °°u2 (·)
Z
{d(x0 ,y)>d(x0 ,x)}
°
°p
|f (y)|
≤ c5 ku1 (·)f (·)kpLpq
.
dµ°° pq
w (X)
µB(x0 , d(x0 , y))
Lv (X)
Now let us estimate S2p . By Lemma B (part (ii)),
S2p ≤
X°
°
°u2 (·)K(f χG
k∈u
°p
°
)(·)χEk (·)°
k,2
We shall use the following notation:
Lpq
v (X)
≡
X
p
.
Sk,2
k∈u
u2,k ≡ sup u2 (x), σ2,k ≡ sup σ2 (x), u1,k ≡ inf u1 (x), σ1,k ≡ inf σ1 (x).
x∈Ek
x∈Gk,2
x∈Ek
x∈Gk,2
By Theorem C and Lemma A we have
°
°
°
1/p °
Sk,2 ≤ u2,k σ2,k °K(f χGk,2 )(·)°
Lpq
ρ (X)
°
1/p °
°
°
≤ c6 u2,k σ2,k °f (·)χGk,2 (·)°
Lpq
ρ (X)
°
°p
°
°
1/p °
1/p °
°
°
≤ c7 u1,k σ1,k °f (·)χGk,2 (·)° ps
≤ c6 u2,k σ2,k °f (·)χGk,2 (·)° ps
Lρ (X)
Lρ (X)
°
°
°
°
≤ c8 °u1 (·)f (·)χG (·)° ps .
k,2
Lw (X)
By Lemma B (part (i)) we finally obtain
.
S2p ≤ c9 ku1 (·)f (·)kLps
w (X)
WEIGHT INEQUALITIES
47
Remark 2.1. It is easy to check that Theorem 2.4 is still valid if we replace
condition (1) by the condition
(1′ )
sup σ2 (y) sup up2 (y) ≤ bσ1 (x)up1 (x),
Ex
Ex
where b > 0 does not depend on x ∈ X and where
¾
½
d(x0 , x)
≤ d(x0 , y) < 4a1 d(x0 , x) .
Ex = y :
4a1
Indeed, we have
1/p
1/p
Sk,2 ≤ σ2,k u2,k kT (f χGk,2 )(·)kLpq
≤ σ2,k u2,k kT (f χGk,2 )(·)kLpρ (X)
ρ (X)
1/p
≤ b1 σ2,k u2,k kf (·)χGk,2 (·)kLpρ (X)
Z
= b1
Gk,2
³
sup
´³
σ2 (y)
2k ≤d(x0 ,y)0
°
°
× °°
1
ϕ1 (d(x0 , ·))
°
°
χ{d(x0 ,y)≤t} (·)°°
Lp
′ s′
´−1
°
χ{d(x0 ,y)>t} (·)°°
Lpq
v(d(x
0 ,·))
(X)
0. Indeed, by Lemma A
(part (i)) and by the reverse doubling condition for µ we have
°
°
°
°
c ≥ B(t) ≥ °ϕ2 (d(x0 , ·))(µB(x0 , d(x0 , ·))−1 χ{tt} (·)°°
Lp
′ s′
¶
(X)
.
On the other hand, we have
Ã
= s′
Z∞
λ
s′ −1
0
Ã
= s
′
³ n
°
°
°(µB(x0 , d(x0 , ·))−1 χ{d(x
³
µ x:
−1
(µB(x
Z0 ,t))
≤ c4
s′ −1
o
³ n
−1
à (µB(x
Z0 ,t))
0
Lp′ s′ (X)
´´s′ /p′
µB(x0 , d(x0 , x))−1 > λ ∩ {d(x0 , x) ≥ t}
λ
0
°
°
(·)°
0 ,y)>t}
µ x : µB(x0 , d(x0 , x)) < λ
λ
s′ −1 −s′ /p′
λ
dλ
!1/s′
−1
o´s′ /p′
= c5 (µB(x0 , t))−1/p .
dλ
!1/s′
dλ
!1/s′
WEIGHT INEQUALITIES
49
Here we have used the inequality
o
n
µ x : µB(x0 , d(x0 , x)) < λ−1 ≤ bλ−1 ,
where the positive constant b is from the doubling condition for µ. Thus we
obtain
µ
¶
t
1/p
−1
B1 (t) ≤ c6 ϕ2 (t)v (t)ϕ1
≤ c7
8a21
for arbitrary t > 0.
By Theorem 2.4 we conclude that inequality (2.3) holds.
Remark 2.2. All results of this section remain valid if we omit the continuity
of σ, but require that µB(x0 , r) be continuous with respect to r, where x0 is
the same fixed point as above. This follows from the fact that in this case
the sequence σn (d(x0 , x)), where σn are absolutely continuous functions (see
the proof of Theorem 2.1.), converges to σ(d(x0 , x)) a.e. on X. On the other
hand, the continuity of µB(x0 , r) with respect to r is equivalent to the condition
µ{x : d(x0 , x) = r} = 0 for all r > 0.
3. Singular Integrals on Spaces of Nonhomogeneous Type
In this section we present weighted inequalities for Calder´on–Zygmund singular integrals defined on spaces of nonhomogeneous type.
Let (X, d, µ) be a spaces of nonhomogeneous type with metric d and measure
µ (i.e., a0 = 1, a1 = 1, we have equality instead of the inequality in (iii) and
(vii) need not be valid in Definition 1.1) satisfying the condition
µB(x, r) ≤ rα ,
x ∈ X, r > 0,
for some α > 0, where B(x, r) ≡ {y : d(x, y) ≤ r}.
Let the kernel k satisfy the following conditions:
(1)
|k(x, y)| ≤ c1 d(x, y)−α ,
for all x, y ∈ X, x 6= y;
(2) there exist c2 > 0 and ε ∈ (0, 1] such that
n
o
max |k(x, y) − k(x1 , y)|, |k(y, x) − k(y, x1 )| ≤ c2
d(x, x1 )ε
d(x, y)α+ε
whenever d(x, x1 ) ≤ 2−1 d(x, y), x 6= y. Assume also that the integral operator
T f (x) = lim
Z
ε→0
X\B(x0 ,ε)
f (y)k(x, y) dµ
is bounded in L2 (X).
The following theorem is valid.
Theorem C ([18]). The operator T is bounded in Lp (X) for 1 < p < ∞
and is of weak type (1, 1).
50
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
Recall that a ≡ sup{d(x0 , x) : x ∈ X}, where x0 is a given point of X and
B(x0 , R)\B(x0 , r) 6= ∅ whenever 0 < r < R < a.
We remark that the condition µ{x} = 0 is automatically satisfied for all
x ∈ X.
Lemma 3.1. Let 1 < p < ∞, and let w be a weight function on X. Assume
that the following two conditions are satisfied:
(i) there exists a positive increasing function v on (0, 4a) such that for almost
all x ∈ X the inequality
v(2d(x0 , x)) ≤ b1 w(x)
holds, where the positive constant b1 does not depend on x;
(ii)
Z
I(t) ≡
′
w1−p (x) dµ < ∞
B(x0 ,t)
for all t > 0. Then T ϕ(x) exists µ-a.e. for any ϕ ∈ Lpw (X).
Proof. Let 0 < α < a and let us denote
n
o
Sβ = x ∈ X : d(x0 , x) ≥ β/2 .
Suppose ϕ ∈ Lpw (X) and represent ϕ as follows:
ϕ(x) = ϕ1 (x) + ϕ2 (x),
where ϕ1 = ϕχSβ and ϕ2 = ϕ − ϕ1 . Due to condition (i) we see that
Z
X
1
≤ β
v( 2 )
Z
Sβ
v( β2 ) Z
|ϕ1 (x)| dµ = β
|ϕ(x)|p dµ
v( 2 )
p
Sβ
c1 Z
|ϕ(x)| v(2d(x0 , x)) dµ ≤ β
|ϕ(x)|p w(x) dµ < ∞
v( 2 )
p
Sβ
for arbitrary β, 0 < β < a. Consequently T ϕ1 ∈ Lp (X) and, according to
Theorem C, T ϕ1 (x) exists µ-a.e. on X.
Now let x be such that d(x0 , x) > β. If y ∈ X and d(x0 , y) < β2 , then
d(x0 , x) ≤ d(x0 , y) + d(x, y) ≤
d(x0 , x)
+ d(x, y).
2
WEIGHT INEQUALITIES
Hence β/2 <
d(x0 ,x)
2
≤ d(x, y) and we obtain
¯Z
¯
¯
¯
¯
|T ϕ2 (x)| = ¯ ϕ2 (y)k(x, y) dµ¯¯ ≤ c2
X
Z
≤ c3
≤ c4 β
−α
µ
51
B(x0 , β2 )
Z
|ϕ(y)|
dµ ≤ c4 β −α
α
d(x, y)
¶1/p µ
p
|ϕ(y)| w(y) dµ
Z
B(x0 , β2 )
Z
|ϕ(y)|
dµ
d(x, y)α
|ϕ(y)| dµ
B(x0 , β2 )
Z
w
1−p′
¶1/p′
(y) dµ
B(x0 , β2 )
B(x0 , β2 )
< ∞.
Thus T ϕ(x) is absolutely convergent for arbitrary x such that d(x0 , x) > β. As
we can take β arbitrarily small and µ{x0 } = 0, we conclude that T ϕ(x) exists
µ-a.e. on X.
Theorem 3.1. Let 1 < p < ∞. Assume that v is a positive increasing
continuous function on (0, 4a). Suppose that w is a weight on X. Let the
following two conditions hold:
(i) there exists a constant b1 > 0 such that the inequality
v(2d(x0 , x)) ≤ b1 w(x)
is fulfilled for µ-almost all x ∈ X;
(ii)
sup
0 t }
2
Ã
¶
p
{x: d(x0 ,x)>t}
Ã
φ(t)
Z
!
¯p
¯
f (y)k(x, y) dµ¯¯ dµ dt ≡ I21 + I22 .
Z
{x: d(x0 ,x)>t} {y: d(x0 ,y)≤ t }
2
Using again the boundedness of T in Lp (X) we obtain
I21 ≤ c4
Za
µ
Z
φ(t)
0
≤ c4
¶
|f (y)| dµ dt = c4
d(x0 ,y)> 2t }
{y:
Z
p
|f (y)|
Z
|f (y)|p w(y) dµ.
p
¶
φ(t) dt dµ
0
X
|f (y)|p v(2a1 d(x0 , y)) dµ ≤ c5
µ 2d(x
Z 0 ,y)
Z
X
X
Now let us estimate I22 . When d(x0 , x) > t and d(x0 , y) ≤
t
2
we have
d(x0 , x) ≤ d(x0 , y) + d(y, x) = d(x0 , y) + d(x, y)
≤
t
d(x0 , x)
+ d(x, y) ≤
+ d(x, y).
2
2
Hence
d(x0 , x)
≤ d(x, y).
2
Consequently,
I22 ≤ c6
Z
µ
µ
Z
1
dµ
d(x0 , x)αp
φ(t)
Za
φ(t)
0
≤ c7
Ã
Za
{x: d(x0 ,y)>t}
0
{x: d(x0 ,x)>t}
{y:
Z
|f (y)|
dµ(y)
d(x, y)α
d(x0 ,y)≤ 2t }
¶µ
Z
¶p
!
dµ(x) dt
¶p
|f (y)| dµ
B(x0 ,t)
dt.
It is easy to see that for any s, 0 < s < a, we have
Za
≤
{x: d(x0 ,x)≥s}
Z
φ(t)
s
Z
µ
1
d(x0 , x)αp
{x: d(x0 ,x)>t}
µ d(x
Z 0 ,x)
¶
1
dµ dt
d(x0 , x)αp
¶
φ(t) dt dµ ≤
s
Z
{x: d(x0 ,x)≥s}
v(d(x0 , x))
dµ.
d(x0 , x)αp
WEIGHT INEQUALITIES
53
Finally, using Proposition A, we obtain
I22 ≤ c8
Z
|f (x)|p w(x) dµ.
X
Lemma 3.2. Let a = ∞, 1 < p < ∞, and let w be a weight function on X.
Suppose the following conditions are fulfilled:
(i) there exists a positive decreasing function v on (0, ∞) such that
¶
µ
d(x0 , x)
≤ cw(x) a.e.;
v
2
(ii) for all t > 0
Z
′
′
w1−p (x)(d(x0 , x))−αp dµ < ∞.
X\B(x0 ,t)
Then T ϕ(x) exists µ-a.e. for arbitrary ϕ ∈ Lpw (X).
Proof. Fix arbitrarily β > 0 and let
n
o
Sβ = x : d(x0 , x) ≥ β .
Represent ϕ as follows:
ϕ(x) = ϕ1 (x) + ϕ2 (x),
where ϕ1 (x) = ϕ(x)χSβ (x) and ϕ2 (x) = ϕ(x) − ϕ1 (x).
For ϕ2 we have
Z
|ϕ2 (x)|p dµ =
X
1
≤
v(β)
Z
v(β)
v(β)
Z
|ϕ(x)|p dµ
B(x0 ,β)
c1 Z
|ϕ(x)| v(d(x0 , x)) dµ ≤
v(β)
p
B(x0 ,β)
|ϕ(x)|p w(x) dµ < ∞.
B(x,β)
Consequently, ϕ2 ∈ Lp (X) and by Theorem C we have T ϕ2 ∈ Lp (X). Hence
T ϕ2 (x) exists a.e. on X.
Now let x ∈ X and let d(x0 , x) < β/2. If d(x0 , y) ≥ β, then
d(x0 , y)
≥ d(x, y).
2
Using this inequality, we obtain
|T ϕ1 (x)| ≤ c2
Z
Sβ
≤ c3
µZ
Sβ
p
Z
|ϕ(y)|
|ϕ(y)|
dµ
≤
c
dµ
3
d(x, y)α
d(x0 , y)α
Sβ
¶1/p µ Z
|ϕ(y)| w(y) dµ
Sβ
w
1−p′
−αp′
(y)d(x0 , y)
¶1/p′
dµ
< ∞.
As we may take β arbitrarily large, we conclude that T ϕ(x) exists a.e.
54
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
Theorem 3.2. Let a = ∞, 1 < p < ∞ and let v be a positive continuous
decreasing function (0, ∞). Suppose that w is a weight function on X and the
following conditions are satisfied:
(i) for almost all x
v(d(x0 , x)/2) ≤ cw(x);
(ii)
sup
t>0
µ
Z
¶1/p µ
Z
v(d(x0 , x)) dµ
B(x0 ,t)
w
1−p′
−αp′
(x)d(x0 , x)
¶1/p′
dµ
X\B(x0 ,t)
< ∞.
Then the operator T is bounded from Lpw (X) to Lpv(d(x0 ,·)) (X).
Proof. Without loss of generality we can represent v as
v(t) = v(+∞) +
Z∞
φ(τ ) dτ,
φ ≥ 0.
t
Further,
Z
|T f (x)|p v(d(x0 , x)) dx
X
= v(+∞)
Z
p
|T f (x)| dµ +
Z
p
|T f (x)|
X
X
Z∞
µ
¶
φ(t)dt dµ ≡ I1 + I2 .
d(x0 ,x)
If v(+∞) = 0, then I1 = 0. But if v(+∞) 6= 0, then by virtue of the boundedness of T in Lp (X) we have
I1 ≤ c1 v(+∞)
Z
|f (x)|p dµ ≤ c1
Z
|f (x)|p v(d(x0 , x)) dµ ≤ c2
|f (x)|p w(x) dµ.
X
X
X
Z
Now we pass to I2 :
I2 =
Z∞
Z∞
0
(1)
where ft
Z
φ(t)
0
+ c3
µ
¶
p
|T f (x)| dµ dt ≤ c3
B(x0 ,t)
µ
φ(t)
Z
(2)
0
= c4
Z
X
φ(t)
0
Z
(1)
¶
|T ft (x)|p dµ dt
B(x0 ,t)
¶
(2)
= f χB(x0 ,2t) and ft
I21 ≤ c4
µ
φ(t)
|T ft (x)|p dµ dt = I21 + I22 ,
B(x0 ,t)
Z∞
Z∞
µ
(1)
= f − ft . Again using Theorem C we have
Z
¶
p
|f (y)| dµ dt
B(x0 ,2t)
p
|f (y)|
µ
Z∞
d(x0 ,x)
2
¶
φ(t) dt dµ ≤ c5
Z
X
|f (y)|p w(y) dµ.
WEIGHT INEQUALITIES
55
It remains to estimate I22 . If x ∈ B(x0 , t) and y ∈ X\B(x0 , 2t), then
d(x0 , y)
≤ d(x, y).
2
Consequently,
I22 ≤ c5
Z∞
µ
Z
φ(t)
0
¶µ
Z
dµ
−α
|f (y)|d(x0 , y)
dµ
X\B(x0 ,2a1 t)
B(x0 ,t)
¶p
dt.
Moreover,
Zs
0
µ
φ(t)
Z
¶
dµ dt =
B(x0 ,t)
Z
B(x0 ,s)
µ
Zs
¶
φ(t) dt dµ ≤
d(x0 ,x)
Z
v(d(x0 , x)) dµ
B(x0 ,s)
and due to Proposition B we finally obtain the boundedness of T .
Now we are going to establish weighted estimates for the operator T in
Lorentz spaces defined on spaces of nonhomogeneous type. Theorem C and
the interpolation theorem imply
Proposition E. Let 1 < p, q < ∞. Then T is bounded in Lpq (X).
The following lemmas are obtained in the same way as in the homogeneous
case. Instead of Theorem A we need to use Theorem C.
Lemma 3.3. Let 1 < s ≤ p < ∞. Let the weight functions w and w1 satisfy
the conditions:
(1) there exists an increasing function v on (0, 4a) such that the inequality
v(d(x0 , x)) ≤ bw(x)w1 (x)
holds for almost all x ∈ X;
(2) for every t, 0 < t < a, the norm
°
°
°
°
°
°
1
χ{d(x0 ,y)≤t} (·)°° p′ s′
w(·)w1 (·)
Lw (X)
is finite.
Then T g(x) exists a.e. on X for any g satisfying the condition
kg(·)w1 (·)kLps
< ∞.
w (X)
Lemma 3.4. Let 1 < s ≤ p < ∞. Suppose also that u and u1 are positive
increasing functions on (0, 4a1 a) and
°
°
°
°
°
°
1
χB(x ,t) (·)°° p′ s′
0
u(d(x0 , ·))u1 (d(x0 , ·))
Lu(d(x
0
0,
°
°
° d(x0 , ·)−α
°
°
χX\B(x ,t) (·)°°
°
0
w(·)w1 (·)
< ∞.
′ ′
p s
Lw
(X)
Then T g(x) exists a.e. on X for arbitrary g satisfying kg(·)w1 (·)kLps
< ∞.
w (X)
From the previous lemmas easily follows
Lemma 3.6. Let a = ∞, 1 < s ≤ p < ∞. Suppose also that for the
decreasing functions u and u1 on (0, ∞) the following condition is satisfied
°
°
°
°
°
°
d(x0 , ·)−α
χX\B(x ,t) (·)°° p′ s′
0
u(d(x0 , ·))u1 (d(x0 , ·))
Lu(d(x
0 ,·))
< ∞,
(X)
for all t > 0. Then T g(x) exists a.e. on X for g satisfying the condition
g(·)u1 (d(x0 , ·)) ∈ Lps
u(d(x0 ,·)) (X).
Using Propositions E and C, we obtain the following result in the same way
as Theorem 2.1.
Theorem 3.3. Let 1 < s ≤ p ≤ q < ∞, and let w be a weight function
on X. Assume that v is a positive increasing continuous function on (0, 4a).
Suppose also that the following two conditions are satisfied:
(1) there exists a positive constant c such that the inequality
v(2a1 d(x0 , x)) ≤ cw(x)
holds for almost every x ∈ X;
(2)
°
°
sup °(d(x0 , ·))−α χX\B(x
0 0
sup v 1/p (x) sup u2 (x) ≤ b inf w1/p (x) inf u1 (x)
t
a1
holds, where Ft = {x ∈ X :
(2)
°
°
sup °u2 (·)(d(x0 , ·))−α χ
t>0
(3)
°
°
sup °u2 (·)χB(x
t>0
0
Ft
Ft
Ft
°
°
(·)°
,t)
≤ d(x0 , x) < 8a1 t};
X\B(x0 ,t)
Lpq
v (X)
Ft
°
°
(·)°
°
°
°
pq
Lv (X) °
°
°
1
< ∞;
χB(x ,t) (·)°° p′ s′
0
u1 (·)w(·)
Lw (X)
°
°
°(d(x0 , ·))−α (u1 (·)w(·))−1 χ
X\B(x0
Then the following inequality holds:
°
°
(·)°
,t)
′ ′
Lpw s (X)
< ∞.
ku2 (·)T f (·)kLpq
≤ cku1 (·)f (·)kLps
,
v (X)
w (X)
where the positive constant c does not depend on f .
Remark 3.1. The results of this section remain valid if we do not require the
continuity of v, but assume that µB(x0 , r) is continuous with respect to r.
4. Examples of weight functions
Let (X, dµ) be an SHT such that the condition µ(B(x, r)) ≈ r holds (if
a < ∞, then we assume that this condition is fulfilled for 0 < r ≤ 1). It is
known (see [5]) that if µ(X) < ∞, then
Z
p
|Kf (x)| (d(x0 , x))
p−1
dµ ≤ c
X
Z
|f (x)|p (d(x0 , x))p−1 logp
X
b
dµ,
d(x0 , x)
′
where x0 ∈ X, 1 < p < ∞, and b = 8a1 aep and the positive constant c does
not depend on f .
Example 4.1. Let 1 < p < q < ∞, and let a < ∞; put v(t) = tp−1 ,
γ
w(t) = tp−1 lnγ bt for t ∈ (0, 4a1 a), where b = 8a1 ae p−1 , γ = pq + p − 1, and a1
is the positive constant from Definition 1.1. Then from Theorem 2.1 it follows
that
kKf (·)kLpq
v(d(x
0 ,·))
(X)
≤ ckf (·)kLpw(d(x
where the positive constant c does not depend on f .
0 ,·))
(X) ,
(4.1)
58
D. E. EDMUNDS, V. KOKILASHVILI, AND A. MESKHI
Example 4.2. Let 1 < p < q < ∞, a = ∞, v(t) = tp−1 when 0 < t ≤ 1
γ
p−1
and v(t) = tα when t > 1, and let w(t) = tp−1 lnγ 2e t when t ≤ 1, and
γ
w(t) = tβ lnβ (2e p−1 ) when t > 1, where γ = pq + p − 1, 0 < α ≤ β < p − 1. Then
inequality (4.1) holds.
An appropriate example for the conjugate function
1
fe(x) =
π
Zπ
−π
f (x + t)
dt
2 tg 2t
is presented in [11].
References
1. H.-M. Chung, R. A. Hunt, and D. S. Kurtz, The Hardy-Littlewood maximal function
on L(p, q) spaces with weights. Indiana Univ. Math. J. 31(1982), 109–120.
2. R. R. Coifman and G. Weiss, Extensions of Hardy spaces and their use in analysis.
Bull. Amer. Soc. 83(1977), 569–645.
3. D. E. Edmunds and V. Kokilashvili, Two-weight inequalities for singular integrals.
Canadian Math. Bull. 38(1995), 119–125.
4. D. E. Edmunds, V. M. Kokilashvili, and A. Meskhi, Boundedness and compactness of Hardy-type operators on Banach function spaces defined on measure space. Proc.
A. Razmadze Math. Inst. 117(1998), 7–30.
5. D. E. Edmunds, V. Kokilashvili, and A. Meskhi, Two weight estimates for singular
integrals defined on spaces of homogeneous type. Canadian J. Math. 52(2000), No. 3,
468–502.
6. I. Genebashvili, A. Gogatishvili, V. Kokilashvili, and M. Krbec, Weight theory
for integral transform on spaces of homogeneous type. Pitman Monographs and Surveys
in Pure and Applied Mathematics, Longman, Harlow, 1998.
7. V. Guliev, Two-weight Lp inequality for singular integral operator on Heisenberg groups.
Georgian Math. J. 1(1994), No. 4, 367–376.
8. E. G. Gusseinov, Singular integrals in the space of function summable with monotone
weight. (Russian) Mat. Sb. 132(174)(1977), No. 1, 28–44.
9. S. Hoffman, Weighted norm inequalities and vector-valued inequalities for certain rough
operators. Indiana Univ. Math. J. 42(1993), 1–14.
10. R. A. Hunt, on L(p, q) spaces. L’ǫnseign. Math. 12(1966), 249–276.
11. V. Kokilashvili, On two-weight inequalities for conjugate functions. Proc. A. Razmadze
Math. Inst. 112(1997), 141–142.
12. V. Kokilashvili and M. Krbec, Weighted inequalities in Lorentz and Orlicz spaces.
World Scientific, Singapore, New Jersey, London, Hong Kong, 1991.
13. V. Kokilashvili and A. Meskhi, Two-weight inequalities for singular integrals defined
on homogeneous groups. Proc. A. Razmadze Math. Inst. 112(1997), 57–90.
14. V. Kokilashvili and A. Meskhi, Two-weight inequalities for Hardy type transforms
and singular integrals defined on homogeneous type spaces. Proc A. Razmadze Math.
Inst. 114(1997), 119–123.
WEIGHT INEQUALITIES
59
15. V. Kokilashvili and A. Meskhi, Two-weight estimates for singular integrals on homogeneous groups. (Russian) Doklady Akademii Nauk 354(1997), No. 3, 301–303. English
transl. in Doklady Mathematics 55(1997), 362–364.
16. V. Kokilashvili and A. Meskhi, Two-weight inequalities for integral operators in
Lorentz spaces defined on homogeneous groups. Georgian Math. J. 6(1999), No. 1, 65–82.
17. B. Muckenhoupt and R. L. Wheeden, Two-weight function norm inequalities for the
Hardy–Littlewood maximal function and the Hilbert transform. Studia Math. 55(1976),
No. 3, 279–294.
18. F. Nazarov, S. Treil, and A. Volberg, Weak type estimates and Cotlar inequalities
for Calder´on–Zygmund operators on nonhomogeneous spaces. International Mathematics
Research Notices 9(1998), 463–487.
19. Y. Rakotondratsimba, On boundedness of classical operators on weighted Lorentz
spaces. Georgian Math. J. 5(1998), No. 2, 177–200.
20. E. T. Sawyer and R. L. Wheeden, Weighted inequalities for fractional integrals on
Euclidean and homogeneous spaces. Amer. J. Math. 114(1992), 813–874.
21. E. M. Stein, Harmonic analysis: real variable methods, orthogonality and oscillatory
integrals. Princeton University Press, Princeton, New Jersey, 1993.
22. I. M. Stein and G. Weiss, Introduction to Fourier analysis on Euclidean spaces. Princeton, New Jersey, Princeton University Press, 1971.
¨ mberg and A. Torchinsky, Weighted Hardy spaces. Lecture Notes in Math.
23. J. O. Stro
1381, Springer Verlag, Berlin, 1989.
(Received 23.08.2000)
Authors’ Addresses:
D. E. Edmunds
Centre for Mathematical Analysis
and its Applications
University of Sussex
Brighton BN1 9QH
Sussex
United Kingdom
E-mail: [email protected]
V. Kokilashvili and A. Meskhi
A. Razmadze Mathematical Institute
Georgian Academy of Sciences
1, M. Aleksidze St., Tbilisi 380093
Georgia
E-mail: [email protected]
[email protected]