Volume 2010, Article ID 727954,27pages doi:10.1155/2010/727954
Research Article
Global Caccioppoli-Type and Poincar ´e Inequalities with Orlicz Norms
Ravi P. Agarwal
1and Shusen Ding
21Department of Mathematical Sciences, Florida Institute of Technology, Melbourne, FL 32901, USA
2Department of Mathematics, Seattle University, Seattle, WA 98122, USA
Correspondence should be addressed to Shusen Ding,sding@seattleu.edu Received 16 July 2009; Revised 1 December 2009; Accepted 14 March 2010 Academic Editor: Jozsef Szabados
Copyrightq2010 R. P. Agarwal and S. Ding. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
We obtain global weighted Caccioppoli-type and Poincar´e inequalities in terms of Orlicz norms for solutions to the nonhomogeneousA-harmonic equationdAx, dω Bx, dω.
1. Introduction
The Lp-theory of solutions of the homogeneousA-harmonic equation dAx, dω 0 for differential forms has been very well developed in recent years. Many Lp-norm estimates and inequalities, including the Hardy-Littlewood inequalities, Poincar´e inequalities, Caccioppoli-type estimates, and Sobolev imbedding inequalities, for solutions of the homogeneousA-harmonic equation have been established; see1–11. Among these results, the Caccioppoli-type inequalities and the Poincar´e inequalities for differential forms have become more and more important tools in analysis and related fields, including partial differential equations and potential theory. However, the study of the nonhomogeneous A-harmonic equationdAx, dω Bx, dωjust began4,6. Roughly, the Caccioppoli-type inequalities or estimates provide upper bounds for the norms of∇uordu in terms of the corresponding norm ofuoru−c, whereuis a differential form or a function satisfying certain conditions. For example,umay be a solution of anA-harmonic equation or a minimizer of a functional, andcis some constant ifuis a function or a closed form ifuis a differential form.
Different versions of the Caccioppoli-type inequalities and the Poincar´e inequalities have been established during the past several decades. For instance, Sbordone proved in12the following version of the Caccioppoli-type inequality:
−
BR/2
A|du|dx≤C−
BR
A
|u−uR| R
dx 1.1
for a quasiminimizeruof the functionalFΩ;v
ΩA|dv|dx, whereAis a continuous, convex, and strictly increasing function satisfying the so-calledΔ2-condition,BRis a ball with radiusR >0, anduR−
BRu dx; see12. Using the above Caccioppoli-type inequality, Fusco and Sbordone obtained in13the higher integrability result
−
BR/2
Ar|du|dx≤C
−
BR
A|du|dx r
1.2
for the gradient of minimizers of the functional IΩ;v, where r > 1 is some constant. In 14, Greco et al. studied the variational integrals whose integrand grows almost linearly with respect to the gradient and the related equation divAx, f∇u 0, whereAis slowly increasing to∞. For instance,At logα1t, α >0, orAt log loget. They proved that the minimizerusubject to the Dirichlet datavsatisfies the estimate
Ω|∇u|A1±ε|∇u|dx≤C
Ω|∇v|A1±ε|∇v|dx 1.3
at least for some smallε >0. In15, Cianchi and Fusco investigated the higher integrability properties of the gradient of local minimizers of an integral functional of the formJu,Ω
Ωfx, u, dudx, whereΩ is an open subset ofRn,n ≥ 2, andf is a Carathodory function defined in Ω × RN × RnNsatisfying some growth conditions. Using a new form of the Caccioppoli inequality and some other tools, such as the Sobolev inequality and a generalized version of the Gehring lemma, they proved that if u is a local minimizer of Ju,Ω, for Ω0⊂⊂Ωthere existsδ >0 such that
Ω0
A|du|
A|du|
|du|
δ
dx <∞, 1.4
whereAsatisfies the so-calledΔ2-condition. However, all versions of the Caccioppoli-type inequality developed or used in12–15are about the minimizeruof some functional. In this paper, we will prove the Caccioppoli-type inequalities and the Poincar´e inequalities with the LslogLα-norm for differential forms satisfying the nonhomogeneousA-harmonic equation.
The method developed in this paper could be used to establish other LslogLα-norm inequalities for solutions of the homogeneousA-harmonic equation or the nonhomogeneous A-harmonic equation.
Throughout this paper, we always assume thatΩis an open subset ofRn,n≥2. Then- dimensional Lebesgue measure of a setE⊆Rnis denoted by|E|. We say thatwis a weight if w∈L1locRnandw >0 a.e. For 0< p <∞, we denote the weightedLp-norm of a measurable functionfoverEby f p,E,wα
E|fx|pwαxdx1/p,whereαis a real number. We write f p,E f p,E,wα ifw 1. A continuously increasing functionϕ : 0,∞ → 0,∞with ϕ0 0 andϕ∞ ∞is called an Orlicz function. The Orlicz spaceLϕΩconsists of all measurable functionsfonΩsuch that
Ωϕf k
dx <∞ 1.5
for somekkf>0.LϕΩis equipped with the nonlinear Luxemburg functional f
ϕ inf
k >0 : 1
|Ω|
Ωϕf k
dx≤1
. 1.6
A convex Orlicz functionϕis often called a Young function. Ifϕis a Young function, then · ϕ defines a norm inLϕΩ, which is called the Luxemburg norm or Orlicz norm. The Orlicz spaceLψΩ with ψt tplogαet/c will be denoted byLplogLαΩ and the corresponding norm will be denoted by f LplogLαΩ, where 1 ≤ p < ∞,α ≥ 0, andc > 0 are constants. The spacesLplogL0ΩandL1logL1Ωare usually referred asLpΩand LlogLΩ, respectively. From16, we have the equivalence
f
LplogLαΩ≈
Ω
fplogα
e f f
p,Ω
dx
1/p
. 1.7
Similarly, we have f
LplogLαΩ,μ≈
Ω
fplogα
e f f
p,Ω
dμ
1/p
, 1.8
whereμis a measure defined bydμwxdxandwxis a weight. In this paper, we simply write
f
LplogLαE,wα
E
|f|plogα
e f f
p,E
wαdx
1/p
, 1.9
and f LplogLαE f LplogLαE,1, wherewis a weight.
We keep using the traditional notations related to differential forms in this paper. Let Λ Λ Rnbe the linear space of the -covectors onRn, 1,2, . . . , n. It is a normed space of dimensionn
. A differential -formωonΩis a Schwartz distribution onΩwith values inΛ Rn. We writeDΩ,Λ for the space of all differential -forms andLpΩ,Λ for all -formsωx
IωIxdxI
ωi1i2···i xdxi1∧dxi2∧ · · · ∧dxi withωI ∈LpΩ,Rfor all ordered -tuplesI. Thus,LpΩ,Λ is a Banach space with norm
ω p,Ω
Ω|ωx|pdx 1/p
⎛
⎝
Ω
I
|ωIx|2 p/2
dx
⎞
⎠
1/p
. 1.10
We useLplogLαΩ,Λ to denote the space of all differential -formsuonΩwith
u LplogLαΩ
Ω|u|plogα
e |u|
u p,Ω
dx 1/p
<∞. 1.11
We use d : DΩ,Λ → DΩ,Λ 1 to denote the differential operator and d : DΩ,Λ 1 → DΩ,Λ to denote the Hodge codifferential operator given byd −1nl1 donDΩ,∧l1, 0,1, . . . , n. Hereis the well-known Hodge star operator. We useBto denote a ball andσB,σ >0, is the ball with the same center asBand with diameterσdiamB.
A differential formuis called closed ifdu 0 and a differential formvis called coclosed if dv0.
Definition 1.1. LetA:Ω×Λ Rn → Λ RnandB:Ω×Λ Rn → Λ −1Rnbe two operators satisfying the conditions:
|Ax, ξ| ≤a|ξ|p−1, Ax, ξ·ξ≥ |ξ|p, |Bx, ξ| ≤b|ξ|p−1 1.12
for almost everyx∈Ωand allξ∈Λ Rn. Then the nonlinear elliptic equation
dAx, dω Bx, dω 1.13
is called the nonhomogeneousA-harmonic equation for differential forms. Herea, b >0 are constants and 1< p <∞is a fixed exponent associated with1.13.
We should notice that if the operatorBequals 0 in1.13, then1.13reduces to the following homogeneousA-harmonic equation, or theA-harmonic equation:
dAx, dω 0, 1.14
which has received much investigation during the recent years; see3,5,7–11. A solution to 1.13is an element of the Sobolev spaceWloc1,pΩ,Λ −1such that
ΩAx, dω·dϕBx, dω·ϕ0 1.15
for allϕ ∈ Wloc1,pΩ,Λ −1with compact support. The solutions of theA-harmonic equation are calledA-harmonic tensors. For any differential formωdefined in a bounded and convex domainD, there is a decomposition
ωdTω Tdω. 1.16
Using the operatorT, we can define thel-formωD∈DD,Λ by
ωD|D|−1
D
ω y
dy, 0, andωDdTω, 1,2, . . . , n, 1.17
for allω∈LpD,Λ , 1≤p <∞. It is known thatuDis a closed form. Hence,u−uDis still a solution of1.13wheneveruis a solution of1.13.
2. Preliminaries
The purpose of this section is to establish some preliminary results that will be used in the proof of our main theorems. In 6, the weighted Poincar´e inequality for solutions of the nonhomogeneousA-harmonic equation was established. From 7, we have the following local Poincar´e inequality.
Lemma 2.1. Letu ∈DΩ,∧ be a differential form in a domainΩ ⊂ Rnanddu ∈ LsΩ,∧ 1, 0,1, . . . , n. Assume that 1< s <∞. Then
u−uB s,B≤C|B|1/n du s,σB 2.1
for all ballsBwithσB⊂Ω. HereCis a constant independent ofuandσ >1 is some constant.
From7, we have the following local Caccioppoli-type inequality.
Lemma 2.2. Letu ∈ DΩ,Λl,l 0,1, . . . , n, be a solution of the nonhomogeneousA-harmonic equation1.13in a domainΩ⊂Rnand letρ >1 be some constant. Assume that 1< s <∞is a fixed exponent associated with theA-harmonic equation1.13. Then there exists a constantC, independent ofu, such that
du s,B ≤C|B|−1/n u−c s,ρB 2.2
for all ballsBwithρB⊂Ωand all closed formsc.
The following weak reverse H ¨older inequality appears in7.
Lemma 2.3. Letube a solution of 1.13inΩand 0< s,t < ∞. Then there exists a constantC, independent ofu, such that
u s,B≤C|B|t−s/st u t,σB 2.3
for all balls or cubesBwithσB⊂Ωfor someσ >1.
Now, we prove the following local Orlicz norm estimates.
Proposition 2.4. Letube a solution of 1.13inΩ,α >0, σ >1, and 1< p <∞. Then there exists a constantC, independent ofu, such that
uB LplogLαB≤C u LplogLασB, 2.4 u−uB LplogLαB ≤C u−c LplogLασB 2.5 for all ballsBwithσB⊂Ωand diamB≥d0. Hered0 >0 is a constant andcis any closed form.
Proof. LetB⊂Ωbe a ball with diamB≥d0 >0. Chooseε >0 small enough and a constant Mlarge enough such that|B|−ε/p2 ≤M. FromLemma 2.3, we have
uB pε,B≤C1|B|p−pε/ppε uB p,σB 2.6 for some σ > 1. Similar to 3.4 in the proof of Theorem 3.1, we may assume that
|uB|/ uB p,B≥1 onB. For aboveε >0, there existsC2>0 such that
logα
e |uB| uB p,B
≤C2
|uB| uB p,σB
ε
. 2.7
From2.6and2.7, it follows that
uB LplogLαB
B
|uB|plogα
e |uB| uB p,B
dx
1/p
≤C3
1 uB ε
p,σB
B
|uB|pεdx 1/p
≤ C3
uB ε/p p,σB
B
|uB|pεdx
1/pεpε/p
≤ C4 uB ε/p
p,σB
|B|p−pε/ppε uB p,σB
pε/p
≤C5|B|−ε/p2 uB p,σB.
2.8
From17, we know that
uB p,σB≤C6 u p,σB. 2.9
Putting2.9into2.8and noting that
logα
e |u|
u p,σB
≥1 2.10
forα >0, we obtain
uB LplogLαB≤C7 u p,σB≤C8 u LplogLασB. 2.11 This ends the proof of inequality2.4. Ifcis a closed differential form, from1.16and1.17, we find that
cdTc Tdc dTc cB. 2.12
Applying triangle inequality and2.4, we conclude that
u−uB LplogLαB u−c−uB−cB LplogLαB
u−c−u−cB LplogLαB
≤ u−c LplogLαB u−cB LplogLαB
≤ u−c LplogLαBC9 u−c LplogLασB
≤C10 u−c LplogLασB
2.13
for any closed formc. The proof ofProposition 2.4has been completed.
Next, extend the weak reverse H ¨older inequality above to the case of Orlicz norms.
Lemma 2.5. Letube a solution of 1.13inΩ,σ >1, and 0< s, t <∞. Then there exists a constant C, independent ofu, such that
u LslogLαB≤C|B|t−s/st u LtlogLβσB 2.14
for any constantsα >0 andβ >0, and all ballsBwithσB⊂Ωand diamB≥d0>0, whered0is a fixed constant.
The proof of Lemma 2.5is similar to that of Proposition 2.4. For completeness, we proveLemma 2.5as follows.
Proof. For any ballB ⊂Ωwith diamB≥d0 >0, we may chooseε >0 small enough and a constantC1such that
|B|−ε/st≤C1. 2.15
FromLemma 2.3, we have
u sε,B≤C2|B|t−sε/tsε u t,σB 2.16
for someσ >1. Similar to3.5in the proof ofTheorem 3.1, we may assume that|u|/ u t,B≥1 onB. For aboveε >0, there existsC3>0 such that
logα
e |u|
u s,B
≤C3
|u|
u t,σB ε
. 2.17
From2.16and2.17, we have
u LslogLαB
B
|u|slogα
e |u|
u s,B
dx 1/s
≤C4
1 u εt,σB
B
|u|sεdx 1/s
≤ C4 u ε/st,σB
B
|u|sεdx
1/sεsε/s
≤ C5
u ε/st,σB
|B|t−sε/tsε u t,σBsε/s
≤C6|B|t−s−ε/st u t,σB.
2.18
From2.15and2.18and using logβe|u|/ u t,σB≥1,β >0, we obtain
u LslogLαB≤C6|B|t−s−ε/st u t,σB
≤C7|B|t−s−ε/st
σB
|u|tlogβ
e |u|
u t,σB
dx 1/t
≤C7|B|t−s−ε/st u LtlogLβσB
≤C8|B|t−s/st u LtlogLβσB.
2.19
This ends the proof ofLemma 2.5.
Using a similar method developed in the proof ofLemma 2.5and from Lemma 2.9 in 6, we can prove the following version of the weak reverse H ¨older inequality with Orlicz norms. Note that the following version of the weak reverse H ¨older inequality cannot be obtained by replacingubyduinLemma 2.5sincedumay not be a solution of1.13.
Lemma 2.6. Letube a solution of1.13inΩ,σ >1, and 0< s,t <∞. Then there exists a constant C, independent ofu, such that
du LslogLαB≤C|B|t−s/st du LtlogLβσB 2.20
for all ballsBwithσB⊂Ωand diamB≥d0>0. Hered0is a fixed constant, andα >0 andβ >0 are any constants.
It is easy to see that for any constantk, there exist constantsm >0 andM >0, such that
mloget≤log
e t k
≤Mloget, t >0. 2.21
From the weak reverse H ¨older inequalityLemma 2.3, we know that the norms u s,Band u t,B are comparable when 0 < d1 ≤ diamB ≤ d2 < ∞. Hence, we may assume that 0 <
m1≤ u s,B ≤M1<∞and 0< m2≤ u t,B≤M2 <∞for some constantsmiandMi, i1,2.
Thus, we have
C1loge|u|≤log
e |u|
u s,B
≤C2loge|u|,
C3loge|u|≤log
e |u|
u t,B
≤C4loge|u|
2.22
for anys >0 andt >0, whereCiis a constant,i1,2,3,4. Using2.22, we obtain
C5
B
|u|slogα
e |u|
u t,B
dx 1/s
≤ u LslogLαB≤C6
B
|u|slogα
e |u|
u t,B
dx 1/s
, 2.23 C7 u LtlogLαB ≤
B
|u|tlogα
e |u|
u s,B
dx 1/t
≤C8 u LtlogLαB 2.24
for any ballBand anys >0, t >0,andα >0. Consequently, we see that u LslogLαB<∞if and only if
B
|u|slogα
e |u|
u t,B
dx 1/s
<∞. 2.25
We recall the Muckenhoupt weights as follows. More properties and applications of Muckenhoupt weights can be found in1.
Definition 2.7. A weight wxis called an ArEweight in a set E ⊂ Rn for r > 1, write w∈ArE, if
sup
B
1
|B|
B
w dx 1
|B|
B
1 w
1/r−1 dx
r−1
<∞ 2.26
for any ballB⊂E.
We will need the following reverse H ¨older inequality forArE-weights.
Lemma 2.8. Ifw ∈ArE,r >1, then there exist constantsk > 1 andC, independent ofw, such that
w k,Q≤Cq1−k/k w 1,Q 2.27
for all balls or cubesQ⊂E.
3. Caccioppoli-Type Estimates
In recent years different versions of Caccioppoli-type estimates have been established; see 1,2,4,12–15,17–19. The Caccioppoli-type estimates have become powerful tools in analysis and related fields. The purpose of this section is to prove the following Caccioppoli-type estimates withLplogLα-norms for solutions to the nonhomogeneousA-harmonic equation.
Theorem 3.1. Letu∈LplogLαΩ,Λ , 0,1, . . . , n−1, be a solution to the nonhomogeneous A-harmonic equation1.13inΩ⊂Rn. Then, there exists a constantC, independent ofu, such that
du LplogLαB≤C|B|−1/n u−c LplogLασB 3.1
for some constantσ >1 and all ballsBwithσB ⊂ Ωand diamB≥ d0 >0. Hered0, 1< p < ∞ andα >0 are constants, andc∈LplogLαΩ,Λ is any closed form.
Proof. LetB⊂Ωbe a ball with diamB≥d0 >0. Letε >0 be small enough and a constant C1large enough such that
|B|−ε/p2≤C1. 3.2
Applying Lemma 2.9 in6, we have
du pε,B≤C2|B|p−pε/ppε du p,σB 3.3
for someσ > 1. We may assume that|du|/ du p,B ≥ 1 onB. Otherwise, settingB1 {x ∈ B:|du|/ du p,B ≥1},B2 {x∈B :|du|/ du p,B <1}, and using the elementary inequality
|ab|s≤2s|a|s|b|s, wheres >0 is any constant, we have
du LplogLαB
B
|du|plogα
e |du|
du p,B
dx 1/p
B1|du|plogα
e |du|
du p,B
dx
B2|du|plogα
e |du|
du p,B
dx 1/p
≤21/p
⎛
⎝
B1
|du|plogα
e |du|
du p,B
dx 1/p
B2
|du|plogα
e |du|
du p,B
dx 1/p⎞
⎠. 3.4
First, we estimate the first term on the right. Since |du|/ du p,B > 1 onB1, then forε > 0 appeared in3.2, there existsC3>0 such that
logα
e |du|
du p,B
≤C3
|du|
du p,σ1B ε
. 3.5
Combining3.2,3.3, and3.5, we obtain
B1
|du|plogα
e |du|
du p,B
dx 1/p
≤C4 1
du εp,σ1B
B1
|du|pεdx 1/p
≤C4 1
du εp,σ1B
B
|du|pεdx 1/p
C4 du ε/pp,σ1B
B
|du|pεdx
1/pεpε/p
≤ C5
du ε/pp,σ1B
|B|p−pε/ppε du p,σ1Bpε/p
≤C6 du p,σ1B,
3.6 whereσ1 >1 is a constant. Since
logα
e |du|
du p,B
≤M1logαe1≤M2, x∈B2, 3.7
we can estimate the second term similarly
B2
|du|plogα
e |du|
du p,B
dx 1/p
≤C7 du p,σ2B, 3.6
whereσ2 >1 is a constant. From3.4,3.6, and3.6, we have
du LplogLαB≤C8 du p,σ3B, 3.8
whereσ3 max{σ1, σ2}. ByLemma 2.2, we obtain
du p,σ3B≤C9|B|−1/n u−c p,σ4B 3.9 for someσ4> σ3and all closed formsc. Note that
logα
e |u−c|
u−c p,σ2B
≥1, α >0. 3.5
Combining last three inequalities, we obtain
du LplogLαB≤C10|B|−1/n u−c p,σ4B≤C10|B|−1/n u−c LplogLασ4B. 3.10
The proof ofTheorem 3.1has been completed.
If we revise3.5and3.5in the proof ofTheorem 3.1, we obtain the following version of Caccioppoli-type estimate.
Corollary 3.2. Letu∈LplogLαΩ,Λ , 0,1, . . . , n−1, be a solution to the nonhomogeneous A-harmonic equation1.13inΩ⊂Rn. Then, there exists a constantC, independent ofu, such that
B
|du|plogα
e |du|
du p,Ω
dx 1/p
≤ C
diamB
σB
|u−c|plogα
e |u−c|
u−c p,Ω
dx 1/p
3.11 for some constantσ >1 and all ballsBwithσB ⊂ Ωand diamB≥ d0 >0. Hered0, 1< p < ∞ andα >0 are constants, andc∈LplogLαΩ,Λ is any closed form.
Theorem 3.3. Letu∈LplogLαΩ,Λ , 0,1, . . . , n−1, be a solution to the nonhomogeneous A-harmonic equation1.13in a bounded domainΩ⊂Rnandwx∈ArΩfor somer >1. Then, there exists a constantC, independent ofu, such that
du LplogLαB,w≤C|B|−1/n u−c LplogLασB,w 3.12 for any closed formc, some constantσ >1 and all ballsBwithσB⊂Ωand diamB≥d0>0. Here d0, 1< p <∞andα >0 are constants, andc∈LplogLαΩ,Λ is any closed form.
Proof. LetBbe a ball withσB ⊂ Ωand diamB ≥ d0 > 0. SinceΩis bounded, thend0 ≤ diamB ≤ diamΩ < ∞. Thus, 0 < v1 ≤ |B| ≤ v2 < ∞for some constants v1 and v2. By Lemma 2.3, we have
m1 u s,ρ1B≤ u t,B ≤m2 u s,ρ2B 3.13
for any solutionuof1.13and any constantss, t >0, where 0 < ρ1 <1, ρ2 >1, 0< m1 <1, andm2 >1 are some constants. ByLemma 2.8, there exist constantsk >1 andC0 >0, such that
w k,B≤C0|B|1−k/k w 1,B. 3.14
Choosespk/k−1, then 1< p < sandks/s−p. We know thatu∈LplogLαΩ,Λ impliesu∈LpΩ,Λ . Then, for any closed formc∈LplogLαΩ,Λ , it follows thatu−c∈ LplogLαΩ,Λ . Thus, u−c ∈ LpΩ,Λ . By Caccioppoli inequality with Lp-norms, we know thatdu ∈ LpΩ,Λ which gives du p,Ω N < ∞. If du p,B 0, thendu 0 a.e.
on B and Theorem 3.3follows. Thus, we may assume that 0 < m1 ≤ du s,B < M1 and 0 < m2 ≤ du p,B < M2 by3.13. Since 1/p 1/s s−p/ps, by the H ¨older inequality, 3.14and2.23, we have
du LplogLαB,w
B
|du|plogα
e |du|
du p,B
w dx 1/p
B
|du|logα/p
e |du|
du p,B
w1/p p
dx 1/p
≤
B
|du|slogαs/p
e |du|
du p,B
dx 1/s
B
ws/s−pdx
s−p/sp
≤C1
B
|du|slogαs/p
e |du|
du s,B
dx
1/s
B
wkdx
1/k1/p
≤C2|B|1−k/kp w 1/p1,B
B
|du|slogαs/p
e |du|
du s,B
dx 1/s
.
3.15
ApplyingTheorem 3.1yields
B
|du|slogαs/p
e |du|
du s,B
dx 1/s
≤C3|B|−1/n u−c LslogLαs/pσ1B. 3.16
Herecis any closed form. Next, chooset p/r. Using3.16andLemma 2.5withβ α/r, we obtain
B
|du|slogαs/p
e |du|
du p,B
dx 1/s
≤C4|B|−1/n|B|t−s/st u−c LtlogLβσ2B 3.17
for someσ2> σ1. Using the H ¨older inequality again with 1/t1/p p−t/pt, we obtain
u−c LtlogLβσ2B
σ2B
|u−c|tlogβ
e |u−c|
u−c t,σ2B
dx 1/t
⎛
⎝
σ2B
|u−c|logβ/t
e |u−c|
u−c t,σ2B
w1/pw−1/p t
dx
⎞
⎠
1/t
≤
σ2B
|u−c|plogβp/t
e |u−c|
u−c t,σ2B
w dx
1/p
σ2B
1 w
t/p−t dx
p−t/pt
≤ u−c LplogLασ2B,w
σ2B
1 w
1/r−1 dx
r−1/p .
3.18
Combining3.15,3.17, and3.18, we conclude that
du LplogLαB,w
≤C5|B|−r/p−1/n u−c LplogLασ2B,w
⎛
⎝
B
w dx
σ2B
1 w
1/r−1 dx
r−1⎞
⎠
1/p
≤C5|B|−r/p−1/n u−c LplogLασ2B,w
w 1,σ2B· 1
w
1/r−1,σ2B
1/p
.
3.19
Sincew∈ArΩ, then
w 1,σ2B· 1
w
1/r−1,σ2B
1/p
⎛
⎝
σ2B
w dx
σ2B
1 w
1/r−1 dx
r−1⎞
⎠
1/p
⎛
⎝|σ2B|r 1
|σ2B|
σ2B
w dx 1
|σ2B|
σ2B
1 w
1/r−1 dx
r−1⎞
⎠
1/p
≤C6|B|r/p.
3.20
Substituting the last inequality into3.19it follows obviously that
du LplogLαB,w≤C7|B|−1/n u−c LplogLασ2B,w. 3.21
This ends the proof ofTheorem 3.3.
Letα1 inTheorem 3.3; we obtain the following corollary.
Corollary 3.4. Letu∈LplogLΩ,Λ , 0,1, . . . , n−1, be a solution to the nonhomogeneous A-harmonic equation1.13in a bounded domainΩ⊂Rnandwx∈ArΩfor somer >1. Then, there exists a constantC, independent ofu, such that
du LplogLB,w≤C|B|−1/n u−c LplogLσB,w 3.22 for any closed formc, some constantσ >1 and all ballsBwithσB⊂Ωand diamB≥d0>0. Here d0and 1< p <∞are constants, andc∈LplogLΩ,Λ is any closed form.
We know that ifw ∈ArEand 0 < λ≤ 1, thenwλ ∈ ArE. Thus, under the same conditions ofTheorem 3.3, we also have the following estimate:
du LplogLαB,wλ ≤C|B|−1/n u−c LplogLασB,wλ, 3.23 wherec is any closed form, and 0 < λ ≤ 1 andα > 0 are any constants. Chooseλ 1/p, 1< p <∞, in3.23. Then, for closed formcand any constantα >0, we have
du LplogLαB,w1/p≤C|M|−1/n u−c LplogLασB,w1/p. 3.24 We have proved Caccioppoli-type inequalities withLplogLα-norms for solutions to the nonhomogeneousA-harmonic equation. Using the same method developed in12, we can obtain the more general version of the Caccioppoli-type inequality for differential forms satisfying certain conditions. A special useful Young functionψ : 0,∞ → 0,∞, termed
anN-function, is a continuous Young function such thatψx 0 if and only ifx 0 and limx→0ψx/x0, limx→ ∞ψx/x ∞.We say that a differential formu∈Wloc1,1Ω,Λ is a k-quasiminimizer for the functional
IΩ;v
Ωψ|dv|dx 3.25
if and only if, for everyφ∈Wloc1,1Ω,Λ with compact support, I
suppφ;u
≤k·I
suppφ;uφ
, 3.26
wherek >1 is a constant. We say thatψsatisfies the so-calledΔ2-condition if there exists a constantp >1 such thatψ2t≤pψtfor allt >0, from which it follows that
ψλt≤λpψt 3.27
for anyt >0 andλ≥1; see12.
We will need the following lemma which can be found in19or12.
Lemma 3.5. Letftbe a nonnegative function defined on the intervala, bwitha≥ 0. Suppose that fors, t∈a, bwitht < s,
ft≤ A
s−tα Bθfs 3.28 holds, where A, B, α, and θ are nonnegative constants with θ < 1. Then, there exists a constant CCα, θsuch that
f ρ
≤C A
R−ρα B
3.29
for anyρ, R∈a, bwithρ < R.
Theorem 3.6. Letube ak-quasiminimizer for the functional3.25and letψ be a Young function satisfying theΔ2-condition. Then, for any ball BR ⊂ Ω with radius R, there exists a constant C, independent ofu, such that
BR/2
ψ|du|dx≤C
BR
ψ
|u−c|
R
dx, 3.30
wherecis any closed form.
The proof ofTheorem 3.6is the same as that of Theorem 6.1 developed in12. For the complete purpose, we include the proof ofTheorem 3.6as follows.
Proof. LetBR Bx0, R ⊂ be a ball with radiusR and center x0, R/2 < t < s < R. Set ηx g|x−x0|, where
gτ
⎧⎪
⎪⎪
⎨
⎪⎪
⎪⎩
1, 0≤τ≤t, affine, τ < t < s, 0, τ≥s.
3.31
Then,η∈W01,∞Bs,ηx 1 onBt, and dηx
⎧⎨
⎩
s−t−1, t≤ |x−x0| ≤s,
0, otherwise. 3.32
Letvx ux ηxpc−ux. We find that
dv 1−ηp
duηppdη
η c−ux. 3.33
Sinceψis an increasing convex function satisfying theΔ2-condition, we obtain
ψ|dv|≤ 1−ηp
ψ|du| ηpψ
pdη
η |c−ux|
. 3.34
Using the definition of thek-quasiminimizer and3.27, it follows that
Bs
ψ|du|dx≤k
Bs
ψ|dv|dx
≤k
Bs\Bt
1−ηp
ψ|du|dx
Bs
ηpψ
pdη
η |c−ux|
dx
≤k
Bs\Bt
ψ|du|dxpp
Bs
ψdη|u−c|
dx
.
3.35
Applying3.35,3.32, and3.27, we have
Bt
ψ|du|dx≤
Bs
ψ|du|dx
≤k
Bs\Bt
ψ|du|dxpp
Bs
ψ
2R|u−c|
s−tR
dx
≤k
Bs\Bt
ψ|du|dx
2pRp s−tp
Bs
ψ
|u−c|
R
dx
.
3.36
Addingk
Btψ|du|dxto both sides of inequality3.36yields
Bt
ψ|du|dx≤ k k1
Bs
ψ|du|dx
2pRp
s−tp
Bs
ψ
|u−c|
R
dx
. 3.37
Next, write ft
Btψ|du|dx,fs
Bsψ|du|dx,A 2pRp
Bsψ|u−c|/Rdx, and B 0. From 3.37, we find that the conditions ofLemma 3.5 are satisfied. Hence, using Lemma 3.5withρR/2 andαp, we obtain3.30immediately. The proof ofTheorem 3.6 has been completed.
It should be noticed thatc ∈ LplogLαΩ,Λ is any closed form on the right side of each version of the Caccioppoli-type inequality. Hence, we may choosec 0 in each of the above Caccioppoli-type inequalities. For example, if we choose c 0 in Theorem 3.1 andTheorem 3.6, we obtain the following Corollaries3.7and3.8, respectively, which can be considered as the special version of the Caccioppoli-type inequality.
Corollary 3.7. Letu∈LplogLαΩ,Λ , 0,1, . . . , n−1, be a solution to the nonhomogeneous A-harmonic equation1.13inΩ⊂Rn. Then, there exists a constantC, independent ofu, such that
du LplogLαB≤C|B|−1/n u LplogLασB 3.38 for some constantσ >1 and all ballsBwithσB ⊂ Ωand diamB≥ d0 >0. Hered0, 1< p < ∞ andα >0 are constants.
Corollary 3.8. Letu be ak-quasiminimizer for the functional 3.25 and ψ be a Young function satisfying theΔ2-condition. Then, for any ball BR ⊂ Ω with radius R, there exists a constant C, independent ofu, such that
BR/2
ψ|du|dx≤C
BR
ψ |u|
R
dx. 3.39
4. Poincar ´e Inequalities
In this section, we focus our attention on the local and global Poincar´e inequalities with LplogLα-norms. The main result for this section is Theorem 4.2, the global Poincar´e inequality for solutions of the nonhomogeneous A-harmonic equation. The following definition ofLϕμ-domains can be found in1.
Definition 4.1. Letϕbe a Young function on0,∞withϕ0 0. We call a proper subdomain Ω⊂RnanLϕμ-domain, if there exists a constantCsuch that
Ωϕσ|u−uΩ|dμ≤Csup
B⊂Ω
B
ϕστ|u−uB|dμ 4.1
for allusuch thatϕ|u|∈L1locΩ;μ, where the measureμis defined bydμwxdx,wx is a weight andτ, σare constants with 0< τ ≤1, 0< σ≤1, and the supremum is over all balls B⊂Ω.
Theorem 4.2. Assume thatΩ⊂Rnis a boundedLϕμ-domain withϕt tplogαet/k, where k u−uB0 p,Ω, 1 < p < ∞, andB0 ⊂ Ωis a fixed ball. Letu ∈ DΩ,Λ0be a solution of the nonhomogeneousA-harmonic equation inΩand du ∈ LpΩ,Λ1as well asw ∈ ArΩfor some r >1. Then, there is a constantC, independent ofu, such that
u−uΩ LplogLαΩ,w≤C|Ω|1/n du LplogLαΩ,w 4.2
for any constantα >0.
To proveTheorem 4.2, we need the following local Poincar´e inequalities, Theorems4.3 and4.4, with Orlicz norms.
Theorem 4.3. Letu ∈ DΩ,∧ be a solution of the nonhomogeneousA-harmonic equation in a domainΩ ⊂ Rnand du ∈LpΩ,∧ 1, 0,1, . . . , n. Assume that 1< p < ∞. Then, there is a constantC, independent ofu, such that
u−uB LplogLαB ≤C|B|1/n du LplogLαρB 4.3
for all ballsBwithρB⊂Ωand diamB≥d0. Hereα >0 is any constant andρ >1 andd0 >0 are some constants.
Proof. LetB⊂Ωbe a ball with diamB≥d0 >0. Chooseε >0 small enough and a constant C1such that
|B|−ε/p2≤C1. 4.4
FromLemma 2.3, we have
u−uB pε,B ≤C2|B|p−pε/ppε u−uB p,ρ1B 4.5
for someρ1>1. Similar to the proof ofTheorem 3.1, we may assume that|u−uB|/ u−uB p,B ≥ 1. Hence, for aboveε >0, there existsC3>0 such that
logα
e |u−uB| u−uB p,B
≤C3
|u−uB| u−uB p,ρ1B
ε
. 4.6