2. The A(α, β, γ; E)-class
In this section, we first introduce the A(α, β, γ; E)-class which is an extension of the A
r
(E)-weight. Then, we study the properties of this class. We will use the following Hölder inequality repeatedly in this article.
Lemma 2.1. Let 0 < α < ∞, 0 < β < ∞ and s-1 = α-1 + β-1. If f and g are measurable functions on ℝ
n
, then || fg ||
s,E
≤|| f ||α,E⋅|| g ||
β,E
for any E ⊂ ℝ
n
.
We introduce the following class of functions which is an extension of the several existing classes of weights, such as -weights, A
r
(λ, E)-weights, and A
r
(E)-weights.
Definition 2.2. We say that a measurable function
g(
x) defined on a subset
E ⊂ ℝ
n
satisfies the
A(
α,
β,
γ;
E)-condition for some positive constants
α,
β,
γ, write
g(
x) ∈
A(
α,
β,
γ;
E) if
g(
x) > 0 a.e., and
(2.1)
where the supremum is over all balls B ⊂ E.
We should notice that there are three parameters in the definition of the A(α, β, γ; E)-class. If we choose some special values for these parameters, we may obtain the existing weights. For example, if α = λ, β = 1/(r - 1) and γ = 1 in above definition, the A(α, β, γ; E) -class becomes A
r
(λ, E)-weight, that is A
r
(λ, E) = A(λ, 1/(r - 1),1;E). Similarly, . Also, it is easy to see that the A(α, β, γ; E)-class reduces to the usual A
r
(E)-weight if α = γ = 1 and β = 1/(r - 1). Moreover, we have the following theorem which establishes the relationship between the A
r
(E)-weight and the A(α, β, γ; E)-class.
Theorem 2.3. Let r > 1 be any constant and E ⊂ ℝ
n
. Then, (i) There exists a constant α0 > 1 such that A
r
(E) ⊂ A(α0,1/(r- 1),α0; E). (ii) For any α with 0 < α < 1, A
r
(E) ⊂ A(α,1/(r-1), α; E).
Proof. For
w(
x) ∈
A
r
(
E), by the reverse Hölder inequality for the
A
r
(
E)-weight, there are constants
α0 > 1 and
C1 > 0 such that
(2.2)
for all balls
B ⊂
E, i.e.,
(2.3)
From (2.3) and (1.1), we obtain
(2.4)
where the supremum is over all balls
B ⊂
E. Thus,
w ∈
A(
α0, 1/(
r - 1),
α0;
E). Hence,
A
r
(
E) ⊂
A(
α0, 1
/(
r -1),
α0;
E). We have completed the proof of the first part of Theorem 2.3. Next, we prove the second part of the theorem. Let
α ∈ (0,1) be any real number. Using the Hölder inequality with 1/
α = 1 + (1 -
α)/
α, we have
(2.5)
that is
which can be written as
(2.6)
Similar to inequality (2.4), using (2.6) and the definitions of the A
r
(E)-weight and the A(α, β,γ; E)-class, we obtain that A
r
(E) ⊂ A(α, 1/(r-1), α; E) for any α with 0 < α < 1. The proof of Theorem 2.3 has been completed.
Example 2.4. Let Ω ⊂ ℝ
n
be a bounded domain containing the origin and g(x) = |x|
p
, x ∈ Ω. We all know that g(x) = |x|
p
∈ A
r
(Ω) for some r > 1 if and only if -n < p < n(r - 1). Now, we consider an example in ℝ2, that is n = 2. Assume that D ⊂ ℝ2 is a bounded domain containing the origin and g(x) = |x|-3 is a function in D. Since p = -3 < -2 = -n, then g(x) = |x|-3 ∉ A
r
(D) for any r > 1. However, it is easy to check that g(x) = |x|-3 ∈ A(α, β, γ; D) for any positive numbers α, β, γ with 0 < α < 2/3.
Combining Theorem 2.3 and Example 2.4, we find that A
r
(E) is a proper subset of A(α, β, γ; E) for any positive constants α, β, γ and r with 0 < α < 2/ 3 and r > 1.
Theorem 2.5. If g1(x), g2(x) ∈ A(α, β, γ; E) for some α ≥ 1, β, γ > 0 and a subset E ⊂ ℝ
n
, then g1(x) + g2(x) ∈ A(α, β, γ; E).
Proof. Let
g1(
x),
g2(
x) ∈
A(
α,
β,
γ;
E). By Minkowski inequality, we find that
(2.7)
Since |
a +
b|
s
≤ 2
s
(|
a|
s
+ |
b|
s
) for any constants
a,
b,
s with
s > 0, from (2.7) , we have
(2.8)
Note that
g1(
x),
g2(
x) ∈
A(
α,
β,
γ;
E). Using (2.8) , we obtain
Thus, g1(x) + g2(x) ∈ A(α, β, γ; E). The proof of Theorem 2.5 has been completed.
Theorem 2.6. Let g1(x) ∈ A(α1, β1, α1γ; E) and g2(x) ∈ A(α2, β2, α2γ; E) for some γ > 0 and any subset E ⊂ ℝ
n
, where α
i
, β
i
> 0, i = 1,2, and . Then, g1(x)g2(x) ∈ A(α, β, αγ; E).
Proof. Using Lemma 2.1 with
and
, respectively, we have
(2.9)
(2.10)
Combining (2.9) and (2.10) yields
(2.11)
which is equivalent to
(2.12)
Noticing that
g1(
x) ∈
A(
α1,
β1,
α1 γ;
E) and
g2(
x) ∈
A(
α2,
β2,
α2γ;
E), we obtain
(2.13)
Thus, g1(x)g2(x) ∈ A(α, β, αγ; E). The proof of Theorem 2.6 has been completed.
Proposition 2.7. Let 0 < p < 1 and g(x) ∈ A(α, βp, γ; E). Then, g
p
(x) ∈ A(α, β, γ; E).
Proof. Using Lemma 2.1 with
yields
that is
(2.14)
Since
g(
x) ∈
A(
α,
βp,
γ;
E), using (2.14) , we find that
(2.15)
Therefore, g
p
(x) ∈ A(α, β, γ; E). The proof of Proposition 2.7 has been completed.
Let α, β, γ > 0 be any constants. It is easy to prove that (i) if and only if g(x) ∈ A(β, α, αβ/γ; E). (ii) g
p
(x) ∈ A(α, β, γ; E) if and only if g(x) ∈ A(αp, βp, γp; E) for any constant p > 0. Also, using the Hölder inequality and the definition of the A(α, β, γ; E)-class, we can prove the following monotone properties of the A(α, β, γ; E)-class.
Proposition 2.8. If α1 < α2, then A(α2, β, γ; E) ⊂ A(α1, β, γ; E) for any β,γ > 0. If β1 < β2, then A(α, β2, γ; E) ⊂ A(α, β1, γ; E) for any α, γ > 0.
From Theorem 2.3 and Proposition 2.8, we know that for every r > 1, there exists a constant α0 > 1 such that A
r
(E) ⊂ A(α, 1/(r - 1),α; E) for any α with 0 < α < α0.
3. Local Poincaré inequalities
As applications of the
A(
α,
β,
γ;
E)-class, we prove the local Poincaré inequalities with the Radon measure for the differential forms satisfying the non-homogeneous
A-harmonic equation. Differential forms are extensions of functions in ℝ
n
. For example, the function
u(
x1,
x2,...,
x
n
) is called a 0-form. The 1-form
u(
x) in ℝ
n
can be written as
. If the coefficient functions
u
i
(
x1,
x2,...,
x
n
),
i = 1,2,...,
n, are differentiable, then
u(
x) is called a differential 1-form. Similarly, a differential
k-form
u(
x) is generated by
, that is,
, where
I = (
i1,
i2,...,
i
k
), 1 ≤
i1 <
i2 < ... <
i
k
≤
n. Let ∧
l
= ∧
l
(ℝ
n
) be the set of all
l-forms in ℝ
n
and
L
p
(Ω, Λ
l
) be the
l-forms
u(
x) = Σ
I
u
I
(
x)
dx
I
in Ω satisfying ∫
Ω |
u
I
|
p
< ∞ for all ordered
l-tuples
I,
l = 1,2,...,
n. We denote the exterior derivative by
d and the Hodge star operator by *. The Hodge codifferential operator
d* is given by
d* = (-1)
nl+1*
d*,
l = 0,1,...,
n - 1. We consider here the solutions to the nonlinear partial differential equation
(3.1)
which is called non-homogeneous
A -harmonic equation, where
A : Ω × ∧
l
(ℝ
n
) → ∧
l
(ℝ
n
) and
B : Ω × ∧
l
(ℝ
n
) → ∧
l-1(ℝ
n
) satisfy the conditions: |
A(
x,
ξ)| ≤
a|
ξ|
p-1,
A(
x,
ξ) ⋅
ξ ≥ |
ξ|
p
and |
B(
x,
ξ)| ≤
b|
ξ|
p-1for almost every
x ∈ Ω and all
ξ ∈ ∧
l
(ℝ
n
). Here
a,
b > 0 are constants and 1 <
p < ∞ is a fixed exponent associated with (3.1). A solution to (3.1) is an element of the Sobolev space
such that ∫
Ω A(
x,
du) ⋅
dφ +
B(
x,
du) ⋅
φ = 0 for all
with compact support. If
u is a function (0-form) in ℝ
n
, the equation (
3.1) reduces to
(3.2)
If the operator
B = 0, Equation (
3.1) becomes
d*A(
x,
du) = 0, which is called the (homogeneous)
A -harmonic equation. Let
A : Ω × ∧
l
(ℝ
n
) → ∧
l
(ℝ
n
) be defined by
A(
x,
ξ) =
ξ|
ξ|
p- 2with
p > 1. Then,
A satisfies the required conditions and
d*
A(
x,
du) = 0 becomes the
p-harmonic equation
d*(
du|
du|
p-2) = 0 for differential forms. See [
5,
6,
9‐
16] for recent results on the solutions to the different versions of the
A-harmonic equation. The operator
K
y
with the case
y = 0 was first introduced by Cartan [
17]. Then, it was extended to the following version in [
18]. Let
D be a convex and bounded domain. To each
y ∈
D there corresponds a linear operator
K
y
:
C∞(
D, ∧
l
) →
C∞(
D, ∧
l-1) defined by
. A homotopy operator
T :
C∞(
D, ∧
l
) →
C∞(
D, ∧
l- 1) is defined by averaging
K
y
over all points
y ∈
D:
Tu = ∫
D
φ(
y)
K
y
udy, where
is normalized so that ∫
D
φ(
y)
dy = 1. The
l-form is defined by
ω
D
= |
D|
-1 ∫
D
ω(
y)
dy, l = 0, and
ω
D
=
d(
T ω),
l = 1,2,...,
n for all
ω ∈
L
p
(
D, ∧
l
), 1 ≤
p ≤ ∞. For any differential form
, we have
(3.3)
Lemma 3.1. [
14]
Let u be a differential form satisfying the non-homogeneous A-harmonic equation (3.1)
in Ω,
σ > 1
and 0 <
s,
t < ∞.
Then, there exists a constant C,
independent of u,
such that ||
du||
s, B≤
C|
B|
(t-s)/st||
du||
t,σB
for all balls or cubes B with σB ⊂ Ω.
Theorem 3.2.
Let be a solution of the non-homogeneous A-
harmonic equation (3.1)
in a domain and 1 <
s < ∞.
Then, there exists a constant C, independent of u, such that (3.4)
for all balls B with σB ⊂ Ω, where the Radon measure μ is defined by dμ = g(x)dx and g ∈ A(α, β, α; Ω), α > 1, β > 0.
Proof. By the decomposition theorem of differential forms, we have u = d(Tu) + T(du) = u
B
+ T(du), where d is the exterior differential operator and T is the homotopy operator.
From (3.3), we obtain
(3.5)
for any
t > 1. Now, choose
t =
αs/(
α - 1), then,
t >
s. Using the Hölder inequality and (3.5), we obtain
(3.6)
Let
m =
βs/(1 +
β), then 0 <
m <
s. From Lemma 3.1, we have
(3.7)
where
σ1 > 1 is a constant. Using the Hölder inequality again, we find that
(3.8)
Since
g ∈
A(
α,
β,
α; Ω), it follows that
(3.9)
Combining (3.6), (3.7), and (3.8) and using (3.9), we have
that is
We have completed the proof of Theorem 3.2.
Let , where x
B
be the center of the ball B ⊂ Ω and . Then, g(x) ∈ A (α, β, α; Ω). From Theorem 3.2, we have the following corollary.
Corollary 3.3.
Let be a solution of the non-homogeneous A-
harmonic equation (3.1)
in a domain and 1 <
s < ∞.
Then, there exists a constant C, independent of u, such that (3.10)
for all balls B with σB ⊂ Ω, where the Radon measure μ is defined by , x
B
is the center of the ball and α > 1 is a constant.
4. Global Poincaré inequalities
In this section, we will prove the global Poincaré inequalities with the Radon measure for solutions of the nonhomogeneous
A-harmonic equation in
L
s
(
μ)-averaging domains. In 1989, Staples [
19] introduced the following
L
s
-averaging domains.
Definition 4.1. A proper subdomain Ω ⊂ ℝ
n
is called an
L
s
-averaging domain,
s ≥ 1, if there exists a constant C such that
for all .
Also, in [
19], the
L
s
-averaging domain is characterized in terms of the quasi-hyperbolic metric. Particularly, Staples proved that any John domain is
L
s
-averaging domain, see [
20] for more results on the averaging domains. In [
15], the
L
s
-averaging domains were extended to the following
L
s
(
μ)-averaging domains.
Definition 4.2. We call a proper subdomain Ω ⊂ ℝ
n
an
L
s
(
μ)-averaging domain,
s ≥ 1, if there exists a constant
C such that
for some ball B0 ⊂ Ω and all , where the Radon measure μ(x) is defined by dμ = w(x)dx and w(x) is a weight. Here, the supremum is over all balls B with B ⊂ Ω.
Theorem 4.3.
Let u ∈
L
s
(Ω, ∧
0)
be a solution of the non-homogeneous A -harmonic equation (3.2)
in a domain Ω,
du ∈
L
s
(Ω, ∧
1), 1 <
s < ∞
. Then, there exists a constant C, independent of u,
such that (4.1)
for any L
s
(μ)-averaging domain Ω ⊂ ℝ
n
with μ (Ω) < ∞, where B0 is some ball appearing in Definition 4.2 and the Radon measure μ is defined by dμ = g(x)dx, g(x) ∈ A(α, β, α; Ω), α >1, β > 0.
Proof. We may assume
g(
x) ≥ 1 a.e. in Ω. Otherwise, let Ω
1 = Ω ⋂ {
x ∈ Ω : 0 <
g(
x) < 1} and Ω
2 = Ω ⋂ {
x ∈ Ω :
g(
x) ≥ 1}. Then, Ω = Ω
1 ∪ Ω
2. We define
G(
x) by
Then, G(x) ≥ g(x) and it is easy to check that g(x) ∈ A(α, β, α; Ω) if and only if G(x) ∈ A(α, β, α; Ω).
Thus,
(4.2)
with
G(
x) ≥ 1. Hence, we may suppose that
g(
x) ≥ 1 a.e. in Ω. Thus, for any
D ⊂ Ω, we have
(4.3)
Note that
diam(
B) =
C1|
B|
1/n. From Theorem 3.2, we obtain
(4.4)
By definition of the
L
s
(
μ) -averaging domain, (4.3) , (4.4) and noticing that 1 + 1/
n - 1/
s > 0, we find that
that is
The proof of Theorem 4.3 has been completed.
In [
15], it has been proved that any John domain is an
L
s
(
μ)-averaging domain. Hence, we have the following corollary.
Corollary 4.4.
Let u ∈
L
s
(Ω, ∧
0)
be a solution of the non-homogeneous A-harmonic equation (3.2)
in a John domain Ω
with μ(Ω) < ∞,
du ∈
L
s
(Ω, ∧
1), 1 <
s < ∞.
Then, there exists a constant C, independent of u, such that (4.5)
where B0 is some ball appearing in Definition 4.2 and the Radon measure μ is defined by dμ = g(x)dx and g(x) ∈ A(α, β, α; Ω), α > 1, β > 0.
Example 4.5. Since the usual p-harmonic equation div (∇u|∇u|p-2) = 0 and the A-harmonic equation div A (x, ∇u) = 0 for functions are the special cases of the non-homogeneous A- harmonic equation, all results proved in Sections 3 and 4 are still true for p-harmonic functions and A-harmonic functions.
Remark. (i) Since an
L
s
-averaging domain is a special
L
s
(
μ)-averaging domain, then the inequality (4.1) still holds in any
L
s
-averaging domain. (ii) Since
μ(Ω) < ∞, the inequality (4.1) can be written as
where Ω is an L
s
(μ)-averaging domain Ω ⊂ ℝ
n
with μ(Ω) < ∞ and B0 is some ball appearing in Definition 4.2, and the Radon measure μ is defined by dμ = g(x)dx and g(x) ∈ A(α, β, α; Ω), α > 1, β > 0. (iii) The inequalities obtained in this article are extensions of the usual A
r
(E)-weighted inequalities since the A
r
(E) is a proper subset of the A(α, β, α; E)-class which can be used to extend many results with the A
r
(E)-weight into the A(α, β, α; E)-weight.