See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/51892645

On the Eigenvalue of $p(x)$-Laplace Equation

ARTICLE · MAY 2011

Source: arXiv

READS

47

2 AUTHORS:

Yushan Jiang Yongqiang Fu

Northeastern University (Shenyang, China) Harbin Institute of Technology
8 PUBLICATIONS 0 CITATIONS 70 PUBLICATIONS 389 CITATIONS

SEE PROFILE SEE PROFILE

Available from: Yushan Jiang

Retrieved on: 19 February 2016
arXiv:1105.4225v1 [math.AP] 21 May 2011

ON THE EIGENVALUE OF p(x)-LAPLACE

EQUATION

Yushan Jiang, Yongqiang Fu

May 24, 2011

 Abstract

The main purpose of this paper is to show that there exists a positive
number λ1 , the first eigenvalue, such that some p(x)-Laplace equation ad-
mits a solution if λ = λ1 and that λ1 is simple, i.e., with respect to the
first eigenvalue solutions, which are not equal to zero a. e., of the p(x)-
Laplace equation forms an one dimensional subset. Furthermore, by devel-
oping Moser method we obtained some results concerning Hölder continuity
and bounded properties of the solutions. Our works are done in the setting
of the Generalized-Sobolev Space. There are many perfect results about
p-Laplace equations, but about p(x)-Laplace equation there are few results.
The main reason is that a lot of methods which are very useful in dealing
with p-Laplace equations are no longer valid for p(x)-Laplace equations. In
this paper, many results are obtained by imposing some conditions on p(x).
Stimulated by the development of the study of elastic mechanics, inter-
est in variational problems and differential equations has grown in recent
decades, while Laplace equations with nonstandard growth conditions share
a part. The equation discussed in this paper is derived from the elastic me-
chanics.
Keyword:p(x)-Laplace equation; eigenvalue; Hölder continuity
Contents

1 Introduction 2
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Overview of Differential Operator with Nonstandard Expo-
nent Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Origin of Problem . . . . . . . . . . . . . . . . . . . . . . . . 6

2 Generalized Sobolev Space 8

2.1 Conceptions and Properties . . . . . . . . . . . . . . . . . . . 8
2.2 Sobolev Embedding Inequalities . . . . . . . . . . . . . . . . . 9
2.3 Notations and Preliminary Results . . . . . . . . . . . . . . . 10

3 On the First Eigenvalue Problem of p(x)-Laplace Equation 12

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 On the Local Boundary of Eigenvalue Problem . . . . . . . . 13
3.2.1 Gradient Estimate of Solution . . . . . . . . . . . . . . 13
3.2.2 Local Bounded of Solution . . . . . . . . . . . . . . . 15
3.3 Hölder Continuity of Solution . . . . . . . . . . . . . . . . . . 17
3.3.1 Sobolev-Poincaré Inequality on Solution . . . . . . . . 17
3.3.2 Harnack Inequality on Solution . . . . . . . . . . . . . 17
3.4 On the First Eigenvalue . . . . . . . . . . . . . . . . . . . . . 21
3.4.1 Comparison Principle . . . . . . . . . . . . . . . . . . 21
3.4.2 The Eigenvalue Problem of Solution . . . . . . . . . . 23

4 Conclusion 29

1
Chapter 1

Introduction

1.1 Background
In recent years, there has been increasing interest toward variable exponent
Lebesgue and Sobolev spaces.It is clear that we cannot simply replace p
by p(x) in the usual definition of the norm in Lp . However, the Lebesgue
spaces can be considered as particular cases of the Orlicz spaces belong to
a larger family of so called modula spaces. This approach enables to define
corresponding counterparts of the Luxemburg and Orlicz norms in Lp(x) .
The present line of investigation toward variable exponent Lebesgue and
Sobolev spaces goes back to a paper by O.Kovác̆ik and J.Rákosnı́k [1] from
1991.After this paper not much happened till the late 1990’s. At this point
the subject seems to have been rediscovered by several researchers indepen-
dently:S.Samko [3, 4], working based on earlier Russian work(I.Sharapudinov
and V.Zhikov [6]), X.Fan and collaborators drawing inspiration from the
study of differential equations [40, 41, 42, 43]. The last couple of years have
seen the integration of the separate lines of investigation, but much still
remains to be done.
The main incentive for many of the investigators of variable exponent
spaces is relaxing the coercivity conditions assumed for the solutions of a
differential equation or the corresponding variational integral.while Laplace
equations with nonstandard growth conditions share a part.One such ap-
plication has been investigated in greater detail, electro-rheological fluids.
These fluids have the interesting property that their viscosity depends on
the electric field in the fluid. For some technical applications the mathemat-
ical theory was presented by some investigators:M.Rǔžička, E.Acerbi and
G.Mingione [8, 9, 10, 23].
The remain part of this part strive to give a little more detailed an
account of the mathematics of variable exponent spaces and in the next
section we present the theory of Generalized Lebesgue spaces and that of
Generalized Sobolev spaces. In the final section or the main section we

2
present a generalization of the eigenvalue problem on some p(x)-Laplace
equation by Mitsuharu Ôtani and Toshiakı́ Teshima [34].

1.2 Overview of Differential Operator with Non-

standard Exponent Growth
The Harnack Inequality
In a bounded domain D of Euclidean space Rn , n ≥ 2, Yu.A.Alkhutov[30]
proved the Harnack inequality and an interior a priori estimate for the Hölder
norm of solutions about the equation as following:
n  
X ∂ p(x)−2 ∂u
|∇u| =0 (1.1)
∂xi ∂xi
i=1

where p(x) is a measurable function in D and 1 < p1 ≤ p(x) ≤ p2 < ∞, the

domain D is divided by a part of a Lipschitz surface into two subdomains,
in each of which p(x) is constant.

The Hardy-Littlewood Maximal Operator.

Assume that 1 < p− ≤ p+ < ∞ and there exists a constant C > 0 such that
C
|p(x) − p(y)| ≤
− log |x − y|
1
for every x, y ∈ Rn , |x − y| ≤ 2 and
C
|p(x) − p(y)| ≤
− log(e + |y|)
for every x, y ∈ Rn , |y| ≥ |x|. Under these assumption on p(x), Cruz-Uribe,
Fiorenza and Neugebauer [13] proved that the Hardy-Littlewood maximal
operator is bounded from Lp(·) (Rn ) to itself. This was an improvement of
earlier work by Diening [19, 20, 21, 22] and Nekvinda [36]. Maximal operator
have also been studied in weighted Lp(·) spaces by Kokilashvili and Samko
[37, 4, 3].

Strong Maximum Principle of p(x)-Laplace Equation

If Ω ⊂ RN (N ≥ 2) be an open set, p(x) ∈ C 1 (Ω), and p(x) > 1(x ∈
Ω), q(x) ∈ C 0 (Ω), and p(x) ≤ q(x) ≤ p∗ (x) (p∗ = NN−p(x)
p(x)
for p(x) < N ;
p∗ = ∞ for p(x) ≥ N ), d(x) ∈ L∞ (Ω), d(x) ≥ 0 a.e.. Fan.X and Zhao.Y [45]
given a strong maximum principle for super-solutions of the p(x)-Laplace
equations

−div(|∇u|p(x)−2 ∇u) + d(x)|u|q(x)−2 u = 0 (1.2)

3
Fan.X and Zhao.Y proved that the nonnegative weak upper solution of (1.2)
u satisfies u ≥ c, x ∈ K a.e., for any given nonempty compact subset K ⊂ Ω,
where c > 0 is a constant. Furthermore if u ∈ C 1 (Ω∪x1 ), u(x1 ) = 0, x1 ∈ ∂Ω
and u satisfies the inner sphere conditions, then ∂u(x
∂γ
1)
> 0, where γ is the
unit inner normal vector of ∂Ω at x1 .

Existence of Solutions for Elliptic Systems with Nonuniform Growth

Let Ω ⊂ Rn be a bounded Lipschitz domain. For the following systems:

∂Aiα
(x, u(x), Du(x)) = B i (x, u(x), Du(x)), x ∈ Ω, i = 1, . . . , N (1.3)
∂Aα

ui (x) = 0, x ∈ ∂Ω, i = 1, . . . , N (1.4)

Fu.Yongqiang proved that if the coeifficients of (1.3) satisfy Aiα : Ω × RN ×
M N ×n → R, B i : Ω × RN × M N ×n → R, i = 1, . . . , N, α = 1, . . . , n, are
Carathéodory functions.
′
|A(x, s, ξ)| ≤ C1 |ξ|p(x)−1 + C2 |s|p(x)−1 + G(x), where G ∈ Lp (x) (Ω),
C1 , C2 ≥ 0 and C2 small.
′
|B(x, s, ξ)| ≤ C1′ |ξ|p(x)−1 + C2′ |s|p(x)−1 + G(x), where G ∈ Lp (x) (Ω),
C1′ , C2′ ≥ 0 and C2′ small.
Aiα (x, s, ξ)ξαi ≥ λ0 |ξ|p(x) − C|s|p(x) + h(x), where λ0 > 0, C ≥ 0 small and
h ∈ L1 (Ω).
For almost every x0 ∈ Ω, s0 ∈ RN , the mapping ξ 7→ A(x0 , s0 , ξ) satisfies
Z Z
Aiα (x0 , s0 , ξ0 + Dz(x))zαi (x)dx ≥ ν |Dz(x)|p(x) dx
G G

for each ξ0 ∈ M N ×n , G ⊂ Rn , z ∈ C01 (G, RN ) where ν > 0 and (Du(x))iα =

∂ui (x)/∂xα = uiα (x). p : Ω → [1, ∞] is a measurable function and p′ is
its conjugate function. Then the Dirichlet problem (1.3),(1.4) has at least
1,p(·)
one weak solution in W0 (Ω, RN ), that is to say, there exists at least one
1,p(·)
u ∈ W0 (Ω, RN ) satisfying
Z
[Aiα (x, u, Du)zαi (x) + B i (x, u, Du)z i (x)]dx = 0 (1.5)
Ω

1,p(·)
for all x ∈ W0 (Ω, RN ). This generalizes the result of Acerbi and Fusco
[11].

Hölder Continuity of Minimizers of Functionals with Variable Growth

Exponent
Let a ∈ W 1,s (Ω)(s > n), r be two nonnegative measurable functions such
that 1 < p0 ≤ a(x) ≤ q0 ≤ p∗0 , 0 ≤ r(x) ≤ r ≤ p∗0 and let f : Ω × R × Rn

4
be a Carathéodory function satisfying the growth assumptions c1 (|ξ|a(x) −
|u|r(x) −1) ≤ f (x, u, ξ) ≤ c2 (|ξ|a(x) +|u|r(x) +1) and let u be a quasiminimizer
of the functional as following
Z
F(u) = f (x, u, Du)dx (1.6)
Ω

Valeria ChiadòPiat and Alessandra Coscia[31]proved the the locally Hölder

continuous of u in Ω.

Hölder continuity of p(x)-Laplace equation

Let Ω be a open set in RN , for the following equation:

p(x)−2
−div λ + |∇u|2 2
∇u = F (x, u), x ∈ Ω ⊂ RN (1.7)

where λ ≥ 0, F ∈ C 0 (Ω×R) satisfies |F (x, u)| ≤ c1 +c2 |u|q(x) , ∀(x, u) ∈ Ω×R

where 1 < q(x) < p∗ (x)(p∗ = NN−p(x)
p(x)
f or p(x) < N, p∗ = +∞ f orp(x) ≥
N ), p ∈ C 1 (Ω), p(x) > 1(∀x ∈ Ω). X.Fan and Zhao Dun [38] proved the
local C 1,α regularity u, that is to say, the weak solution of (1.7) satisfies
1,α
u ∈ CLoc (Ω).

On the Positive Solution of p(x)-Laplace Equation

Let Ω be a bounded domain in RN (N > 1), for the following equation:

−div(|∇u|p(x)−2 ∇u) = λ|u|α(x)−2 u + |u|β(x)−2 u x ∈ Ω


(1.8)
u(x) = 0 x ∈ ∂Ω

where p, α, β are the continuous functions on Ω, and p(x) < N, λ > 0.

X.Fan[39] proved that if 1) p(x) : Ω → R is Lipschitz continuous and p− > 1;
2) 1 < α− ≤ α+ < p− ≤ p+ < β− , β(x) ≤ p∗ (x). then (1.8) has at least two
positive solutions for small λ.

Dirichlet Boundary Value Problem

Consider a differential operator A of order 2k in the divergence form
X
Au(x) = (−1)|α| D α aα (x, δk u(x)) (1.9)
|α|≤k

where the functions aα (x, δk u(x)) ∈ CAR(Ω, m), m P = ♯{α ∈ NN 0 : |α| ≤

k}, fulfill the growth condition |aα (x, ξ)| ≤ g(x) + c |α|≤k |ξα |p(x)−1 with
g ∈ C ps(x) (Ω) and c > 0. Let Q be a Banach space of functions on Ω
equipped with a norm k · kQ and such that C0∞ (Ω) is dense in Q and more-
k,p(x)
over, W0 (Ω) Q A function u ∈ W k,p(x)(Ω) is a weak solution to the

5
Dirichlet boundary value problem (A, u0 , f ) for the equation Au = f with
k,p(x)
the boundary condition given by u0 , if u−u0 ∈ W0 (Ω) and if the identify
X Z
aα (x, δk u(x))D α υ(x)dx = hf, υi (1.10)
|α|≤k Ω

k,p(x)
holds for every υ ∈ W0 (Ω). O.Kovác̆ik and J.Rákosnı́k [1] proved that
if p(x) ∈ P (Ω) satisfy

1 < ess inf p(x) ≤ ess sup p(x) < ∞

Ω Ω

the functions aα satisfy

X
[aα (x, ξ) − aα (x, η)] (ξα − ηα ) ≥ 0, (1.11)
|α|≤k

X X
aα (x, ξ)ξα ≥ c1 |ξα |p(x) − c2
|α|≤k |α|≤k

for every ξ, η ∈ Rm
and for a.e. x ∈ Ω with some constants c1 , c2 > 0.
Then the boundary value problem (A, u0 , f ) has at least one weak solution
u ∈ W k,p(x)(Ω). If moreover, the inequality (1.11) is strict for ξ 6= η then
the solution is unique.

Existence of Positive Solution on p(x)-Laplace Equation

For the equations as following:

−div(|∇u|p(x)−2 ∇u) = a(x)|u|β(x) u


x∈Ω
(1.12)
u(x) = 0 x ∈ ∂Ω

X.Fan [44] proved the existence of positive solution for (1.12) with β + < p−
or β− > p+ .

1.3 Origin of Problem

Our study comes from the article [34] written by M.ôtani and T.Teshima.
in which they study the eigenvalue of the equation as following:

−∆p u(x) + a(x)|u(x)|p−2 u(x) = λb(x)|u(x)|p−2 u(x) x ∈ Ω



u(x) = 0 x ∈ ∂Ω

where ∆p u(x) = Div(|∇u(x)|p−2 ∇u(x)), λ > 0.

M.ôtani and T.Teshima. proved that eigenvalue problem above has a
nontrivial nonnegative solution u if and only if λ = λ1 and Jλ1 := A(u) −

6
λ1 B(u) = 0. Furthermore, the set of all solutions consists of tu1 ; t ∈ R1
where u1 is a solution and u1 ∈ C 1,θ (Ω) for some θ ∈ (0, 1). But if p is a
function of x ∈ Ω. it’s a more difficult situation. It’s clear that we can not
simply replace p by p(x) in the equation about [43]. However, we can extend
the definition of p-Laplace operator by

∆p(x) u(x) = Div(p(x)|∇u(x)|p(x)−2 ∇u(x)) (1.13)

which is named p(x)-Laplace Operator. Our work tries to give some results
about the eigenvalue problem of p(x)-Laplace equation as following:

−∆p(x) u(x) + a(x)|u(x)|p(x)−2 u(x) = λb(x)|u(x)|p(x)−2 u(x) (1.14)

7
Chapter 2

Generalized Sobolev Space

2.1 Conceptions and Properties

For a set Ω ∈ RN with |Ω| > 0, we define the family of all measurable
functions p : Ω → [1, ∞] by P(Ω). we put Ω1 = {x ∈ Ω : p(x) = 1},
Ω∞ = {x ∈ Ω : p(x) = ∞}, Ω0 = Ω\(Ω0 ∪ Ω∞ ); also, we define p+ =
ess supx∈Ω0 p(x) and p− = ess inf x∈Ω0 p(x). We define the variable exponent
Lebesgue space Lp(·) (Ω) to consist of all measurable functions u : Rn → R
such that
Z
̺p(·) (λu) = |λu(x)|p(x) dx + ess sup |u(x)| < ∞ (2.1)
Rn Ω∞

for some λ > 0. The function ̺p(·) : Lp(·) (Ω) → [0, ∞] is called the modular
of the space Lp(·) (Ω). we define a norm, the so-called Luxemburg norm, on
this space by the formula

kukp(·) = inf{λ > 0 : ̺p(·) (u/λ) ≤ 1} (2.2)

If p is a constant function, then the variable exponent Lebesgue spaces

coincides with the classical Lebesgue spaces and so the notation can give
rise to no confusion. The varialbe exponent Sobolev space W k,p(x)(Ω) is the
subspace of functions Lp(·) (Ω) whose distributional gradient exists almost
everywhere and lies in Lp(·) (Ω). The norm of W k,p(x)(Ω) defined by
X
kukk,p = kD α ukp(·) (2.3)
|α|≤k

k,p(x)
By W0 (Ω) we denote the subspace of W k,p(x)(Ω) which is the closure of
∞
C0 with respect to the norm (2.3).
Basic properties. Variable exponent Lebesgue spaces resemble classical
Lebesgue spaces in many respects [1] –they are Banach spaces,the Hölder
inequality holds,they are reflexive if and only if 1 < p− ≤ p+ < ∞ and

8
continuous functions are dense if p+ < ∞; The inclusion between Lebesgue
spaces also generalizes naturally:if 0 < |Ω| < ∞ and p, q are variable ex-
ponent so that p(x) ≤ q(x) almost everywhere in Ω then there exists an
imbedding Lq(·) (Ω) ֒→ Lp(·) (Ω) whose norm does not exceed |Ω| + 1; If
p+ < ∞ and (fi ) is a sequence of functions in Lq(·) (Ω), then kfi kp(·) → 0 if
k,p(x)
and only if ̺p(·) (fi ) → 0; The spaces W k,p(x)(Ω) and W0 (Ω) are Banach
spaces,which are separable and reflexive if 1 < p− ≤ p+ < ∞; If q(x) ≤ p(x)
for a.e. x ∈ Ω then W k,p(x)(Ω) W k,q(x) (Ω).

2.2 Sobolev Embedding Inequalities

As we know,in dealing with some partial differential equation problems
Sobolev embedding inequality is useful for us. Many good results are derived
from these inequalities.

Theorem 2.1 ([40]). Assum Ω be a open domain in RN , with cone property

let p(x) ∈ P(Ω) be a Lipschitz continuous function, if q(x) ∈ P(Ω) satisfy
p(x) ≤ q(x) ≤ P ∗ (x) := NN−kp(x)
p(x)
a.e.x ∈ Ω then there exists a continuous
embedding W k,p(x)(Ω) ֒→ Lq(x) (Ω).

Theorem 2.2 ([40]). Assum Ω be a open domain in RN , with cone property;

If p(x) : Ω → R is uniform continuous and satisfy 1 < p− ≤ p+ < Nk then for
any measurable function q(x) defined in Ω with p(x) ≤ q(x), a.e.x ∈ Ω and
ess inf x∈Ω (p∗ (x) − q(x)) > 0 there is a continuous embedding W k,p(x)(Ω) ֒→
Lq(x) (Ω).

Theorem 2.3 ([40]). Assume that Ω be a open domain in RN with cone

property,If Ω is bounded, p(x) ∈ C(Ω) and q(x) is the same as in theorem
2.2, then there is a continuous compact embedding W k,p(x)(Ω) Lq(x) (Ω).

Theorem 2.4 ([43]). Suppose that p : RN → R is a uniformly continuous

and radically symmetric function satisfying 1 < p− ≤ p+ < N then, for any
measurable function α : RN → R with p(x) ≪ α(x) ≪ p∗ (x), ∀x ∈ RN , there
1,p(x) 1,p(x)
be a compact imbedding Wr (RN ) Lα(x) (RN ). where Wr (RN ) :=
{u ∈ W 1,p(x) N
(R ) : uis radially symmetric.}.

Theorem 2.5 ([43]). If p : RN → R is a uniformly continous and satisfies

1 < p− ≤ p+ < N then for any measurable function α with p ≪ α ≪
1,p(x)
p∗ , x ∈ RRN we have the compact imbedding Wr (RN ) Lα(x) (RN ). where
N
p(x) = G p(g(x))dµ(g), ∀x ∈ R , G = O(N ) be the orthogonal group on
RN , µ be a Haar measure on the compact group G, and µ(G) = 1.

Theorem 2.6 ([43]). Let G be a subgroup of O(N ) and Ω be a invariant

open subset in RN compatible with G, p : RN → R be G-invariant and
uniformly continuous such that 1 < p− ≤ p+ < N holds. Then ,for any

9
measurable function α with p ≪ α ≪ p∗ , x ∈ Ω, we have the compact
1,p(x) 1,p(x) 1,p(x)
imbedding W0,G (Ω) Lα(x) (Ω). where W0,G (Ω) := {u ∈ W0 (Ω) :
u is G-invariant}.

2.3 Notations and Preliminary Results

Let Ω be a bounded open subset of Rn , (n ≥ 2). we use the standard no-
tation for the Generalize Lebesgue and Generalize Sobolev spaces Lp(x) (Ω)
and W k,p(x)(Ω); in particular we will denote by k · kp and k · kk,p the corre-
sponding norms. The Lebesgue measure of a measurable set A ⊂ Rn will
be denoted by |A|,whereas a ball of radius R will be denoted by BR and all
balls mentioned in a single proposition will always be assumed to be con-
centric;moreover if u : Ω → R, k ∈ R and BR is a ball strictly contianed in
Ω, we set

A(k, R) = {x ∈ BR : u(x) > k}

M (u, R) = sup u
BR
m(u, R) = inf u
BR
osc(u, R) = M (u, R) − m(u, R)

Finally ,for all k ∈ R and R > 0 we set

1
I Z
udx = udx
|BR | BR
I BR
1
Z
udx = udx
A(k,R) |BR | A(k,R)

we will denote by the same letter C (or C(· · · ) to stress the dependence on
some arguments) several constants, whose value may change from line to
line.
In the proof of the article we shall use the following Lemma which can
be found in [48],[33],[31].

Lemma 2.7 (Moser iteration inequality). Let {xi }i be a sequence of positive

−1/β 2
real numbers such that x0 ≤ C −1/βB , xi+1 ≤ CB i x1+β
i where β >
0, C > 0, B > 1. then xi → 0 as i → +∞.

Lemma 2.8 (Poicaré inequality[48]). Let u be a function in W01,p (Rn ) then

∗
Z n
(Z  1 ) pn
∗
Y p
|u|p dx ≤ |u|p dx (2.4)
Rn i=1 Rn

np
holds for 1 ≤ p ≤ n,where p∗ = n−p .

10
Lemma 2.9 (Sobolev-Poicaré inequality[33]). for any given bounded do-
1,p(x)
main Ω,if p(x) ∈ L∞ (Ω), u(x) ∈ W0 (Ω) then
Z Z
|u(x)|p(x) dx ≤ C |∇u(x)|p(x) dx (2.5)
Ω Ω

where C is a constant depended on Ω.

Lemma 2.10 (Sobolev-Poicaré inequality with variable exponent[31]). As-

sume that a ∈ W 1,s (Ω), s > n satisfies 1 < p0 < a(x) < q0 ≤ p∗0 ; then for
every M > 0 there exists a positive radius R1 = R1 (M, s, n, kak1,s ) such that
for every γ > n1 − 1s there exist two positive constants χ = χ(n, p0 , s, γ, kak1,s )
and C = C(n, p0 , q0 ) for which
I  u a(x) n dx  n−1
n
I
n−1
≤ c |Du|a(x) dx+χ|{x ∈ BR : |u| > 0}|γ (2.6)
 
R
 
BR BR

holdsR for every BR ⊂ Ω with 0 < R < R1 ,and every u ∈ W 1,p0 (BR ) such
that BR |Du|a(x) dx < +∞, supBR ≤ M u = 0 on ∂BR .

11
Chapter 3

On the First Eigenvalue

Problem of p(x)-Laplace
Equation

3.1 Introduction
Suppose Ω is a bounded domain in Rn with a smooth boundary ∂Ω. For
np
certain given p(x) ∈ P (Ω), where 1 < p ≤ p(x) ≤ p∗ < +∞(p∗ = (n−p) ) and
∂p
p(x) is continuous in Ω and the partial differential of p(x), ∂x i
is bounded
a.e. in Ω. Thinking the eigenvalue problem of p(x)-Laplace equation (E)λ
as following:

−∆p(x) u(x) + a(x)|u(x)|p(x)−2 u(x) = λb(x)|u(x)|p(x)−2 u(x) x ∈ Ω



u(x) = 0 x ∈ ∂Ω

where

−∆p(x) u(x) = −Div(p(x)|∇u(x)|p(x)−2 ∇u(x)) (3.1)

a(x) ∈ L∞
+ (Ω) = {f
∞
∈ L (Ω)|f (x) ≥ 0a.e.x ∈ Ω} (3.2)
∞ ∞ +
b(x) ∈ L0 (Ω) = {f ∈ L (Ω)|f (x) =: max{f (x), 0} =
6 0} (3.3)

We say that u(x) is the solution of the eigenvalue problem (E)λ if u(x) ∈
1,p(x)
W0 (Ω) satisfies the equation in the general sense, that is to say, for any
given function h(x) ∈ C0∞ (Ω) there stands
Z
−Div(p(x)|∇u(x)|p(x)−2 ∇u(x))h(x)dx
ΩZ

+ a(x)|u(x)|p(x)−2 u(x)h(x)dx
ΩZ

= λ b(x)|u(x)|p(x)−2 u(x)h(x)dx (3.4)

Ω

12
In the generalized sobolev space, we study the eigenvalue of the equation.By
Moser iteration, we obtain some properties about the solution of eigenvalue
problem (E)λ :boundary,Hölder continouty,and so on.

3.2 On the Local Boundary of Eigenvalue Problem

3.2.1 Gradient Estimate of Solution
Lemma 3.1 (Caccioppoli Inequality on Solution). Suppose u(x) is the so-
lution of (E)λ , p(x) satisfies 1 < p ≤ p(x) ≤ p < p∗ < +∞ in a certain open
set ω included in Ω, then for any given spherical neighborhood BR in ω and
every 0 < s < t < R < 1 , k > 0 there stands
!
 u(x) − k q
Z Z  
p q
|∇u(x)| dx ≤ C  t − s  dx + (1 + k )|A(k, R)| (3.5)
 
A(k,s) A(k,t)

where C = C(p, q, λ)
Proof : Let ζ(x) be a cut-off function between Bs and Bt with |∇ζ(x)| ≤
2
t−s , by using a test function ϕ = −ζ(u − k)+ in equation (??), one arrives
at
Z Z
p(x) ∇(ζ(x)(u(x) − k)+ )p(x) dx
 
|∇u(x)| dx ≤
A(k,s) A(k,t)

Z
= |∇(ζ(x)(u(x) − k))|p(x) dx
A(k,t)
Z
= |ζ(x)∇(u(x) − k) + (u(x) − k)∇ζ(x)|p(x) dx
A(k,t)
(Z
≤C |ζ(x)|p(x) |∇(u(x) − k)|p(x) dx
A(k,t)
Z )
+ |(u(x) − k)∇ζ(x)|p(x) dx
A(k,t)
(Z Z )
p(x) p(x)
≤C |∇u(x)| dx + |(u(x) − k)∇ζ(x)| dx
A(k,t) A(k,t)
(Z Z )
p(x) p(x)
≤ C(p, q, λ) |u(x)| dx + |(u(x) − k)∇ζ(x)| dx
A(k,t) A(k,t)

the last inequality can be deduced from (3.4) in which we take h(x) = u(x)
such that
Z  
p(x)−2
−Div p(x) |∇u(x)| ∇u(x) u(x)dx
Ω

13
 Z
=λ b(x) |u(x)|p(x)−2 u(x)u(x)dx
Ω
or
Z n o Z
p(x) |∇u(x)|p(x) + a(x) |u(x)|p(x) u(x)dx = λ b(x) |u(x)|p(x) dx
Ω Ω
(3.6)
since a(x), b(x) ∈ L∞ (Ω), 1 < p ≤ p(x) ≤ q ≤ p∗ < +∞, we get the result
above.
2
Moreover,by |∇u(x)| ≤ t−s we obtain
Z
|∇u(x)|p(x) dx
A(k,s)
 u(x) − k p(x)
Z Z  
p(x)
≤ C(λ) |u(x)| dx + C 
 t−s 
 dx
A(k,t) A(k,t)
 u(x) − k p(x)
Z Z  
p(x)
≤ C(λ) |u(x) − k + k| dx + C 
 t−s 
 dx
A(k,t) A(k,t)
 u(x) − k p(x)
Z   Z  
p(x) p(x)
≤ C(λ) |u(x) − k| + |k| dx + C 
 t−s 
 dx
A(k,t) A(k,t)
(Z  p(x) )
u(x) − k
Z
|k|p(x) dx + C
 
≤ C(λ) 
 t−s 
 dx
A(k,t) A(k,t)

applying for Younger inequality and sobolev embedding theorem, we have

Z Z  
|∇u(x)|p dx ≤ 1 + |∇u(x)|p(x) dx
A(k,s) A(k,s)
(Z )
 u(x) − k p(x)
Z  
p(x)
≤ C(λ) |k| dx + C 
 t−s 
 dx + |A(k, s)|
A(k,t) A(k,t)
(Z )
 u(x) − k p(x)
 
q
≤ C(λ) 
 t−s 
 dx + (1 + k )|A(k, t)|
A(k,t)
(Z )
 u(x) − k q
   
q
≤ C(λ) 1 +   dx + (1 + k )|A(k, t)|
A(k,t) t−s 
(Z )
 u(x) − k q
 
q
≤ C(λ)  t − s  dx + (1 + k )|A(k, t)|
 
A(k,t)

Corollary 3.2. Suppose u(x) is the solution of (E)λ , p(x) satisfies 1 < p ≤
p(x) ≤ p < p∗ < +∞ in a certain open set ω included in Ω, then for any
given spherical neighborhood BR in ω and every 0 < ρ < R < 1, k > 0 there
stands
!
 u(x) − k p(x)
Z Z  
p(x)
|∇u(x)| dx ≤ C 
 R−ρ 
 dx + |A(k, R)| (3.7)
A(k,ρ) A(k,R)

14
and
!
 u(x) − k q
Z Z  
p(x)
|∇u(x)| dx ≤ C  R − ρ  dx + |A(k, R)|
  (3.8)
A(k,ρ) A(k,R)

where C = C(p, q, λ), 1 < p◦ ≤ p ≤ p(x) ≤ p < p∗ < +∞

3.2.2 Local Bounded of Solution

Theorem 3.3 (Local Bounded on solution). the solution of (E)λ is local
bounded, that is, for every spherical neighborhood BR in ω ⊂ Ω and 0 <
R < 1, there exist a certain given positive number k > 0 such that u(x) ≤ k
where p(x) satisfies 1 < p ≤ p(x) ≤ p < p∗ < +∞.
R R
Proof : For fixed BR ⊂⊂ Ω and R ≤ 1, k ≥ k0 > 0 let ϑh = 2 + 2h+1
,
ϑ +ϑ 1
ϑh = h 2 h+1 , kh = k(1 − 2h+1 ),h = 0, 1, 2, · · · . Obviously ϑh monotonously
decreases to R2 and kh monotonously increases to k as h → +∞. Define a
functional Jh as
Z
∗
Jh = |u(x) − kh |p dx
A(kh ,ϑh )

Take a function ξ(t) ∈ C 1 ([0, +∞]) such that 0 ≤ ξ(t) ≤ 1, |ξ(t)| ≤ C(constant)
and ξ(t) = 1 when 0 ≤ t ≤ 12 , ξ(t) = 0 when t ≥ 43 . from above making the
h+1
cutting function ξh (x) as ξh (x) = ξ( 2 R (|x| − R2 )) by the Poincaré inequal-
ity(2.4) and Caccioppoli inequality(3.5) we have
Z
∗
Jh = |u(x) − kh |p dx
A(k ,ϑ )
Z h h
∗
≤ |(u(x) − kh )ξh (x)|p dx
A(k ,ϑ )
Z h h
∗
= |(u(x) − kh )+ ξh (x)|p dx
A(kh ,ϑh )
 p∗
p ! p1  n

n  +
 ∂((u(x) − kh+1 ) ξh (x)) 
Y Z
≤   dx
 A(kh ,ϑh )
 ∂xi  
i=1
∗

n p ! 1  pn
p

 ∂u(x) 
Y Z Z
 dx + 2hp p
≤ C 
 ∂xi  |u(x) − kh+1 | dx

i=1 A(kh ,ϑh ) A(kh ,ϑh ) 

15
 ( n Z  ∗
 u(x) − kh+1 p
Y 
≤ C(λ)  ϑ − ϑ  dx
 
i=1 A(kh ,ϑh ) h h
∗
! 1  pn
Z p
p ∗
+2hp |u(x) − kh+1 |p dx + (1 + kh+1 ) |A(kh+1 , ϑh )|
A(kh ,ϑh ) 
( n Z
∗ p ∗
|u(x) − kh+1 |p dx + (1 + kh+1
Y
hp∗
≤ C(λ) 2 ) |A(kh+1 , ϑh )|
i=1 A(kh ,ϑh )
∗
! 1  pn
Z p

+2hp |u(x) − kh+1 |p dx

A(kh ,ϑh ) 
or
∗
( n 
p
 1 ) pn
∗ p∗ −p p
p
Y ∗
Jh+1 ≤ C(λ) 2hp Jh + (1 + kh+1 ) |A(kh+1 , ϑh )| + 2hp |A(kh+1 , ϑh )| p Jh p∗

i=1
(3.9)
but
∗
(kh+1 − kh )p |A(kh+1 , ϑh )|
Z Z
∗ ast
= |kh+1 − kh |p dx ≤ |u(x) − kh |p dx ≤ Jh
A(kh ,ϑh ) A(kh ,ϑh )
or
2h+2 p∗ ∗ ∗
|A(kh+1 , ϑh )| ≤ ( ) Jh , kp |A(kh+1 , ϑh )| ≤ (2h+2 )p Jh (3.10)
k
combining(3.9) and (3.10) we get
( n 
Y 1
Jh+1 ≤ C(λ) 2hp∗ Jh + (1 + p∗ )(2h+2 )p∗ Jh
k
i=1
∗
 1  pn
 h+2  p∗p−p
∗ ∗
p −p p
p
2 p∗ p∗ 

+ ( ) Jh Jh
k 

∗
(  p∗ −p ) pn
1 ∗ 1
= C(λ) 2hp∗ Jh + (1 + p∗ )2hp∗+2p∗ Jh + 2(h+2)(p −p) Jh
k k
n o p∗
hp∗ hp∗+2p∗ h(p∗−p) n
≤ C(λ, k) 2 Jh + 2 Jh + 2 Jh
hp∗2 p∗−p
≤ C(λ, k)2 p Jhp+1
 2 h p∗−p
p∗
= C(λ, k) 2 p Jhp+1

16
by choosing k > 1 such that
∗
k p
Z 
−1 −1
dx ≤ C(λ) η (2I ) η2

J0 = u(x) − (3.11)
A( k2 ,R)
 2 

∗ 2 ∗ 2
where I = (pp) , η = (p )p −p = n−p
p
. Then with the help of Moser iterative
inequality Lemma 2.7, we have

lim Jh = 0 (3.12)
h→+∞

or
Z
∗
|u(x) − k|p dx = 0 (3.13)
A(k, R
2
)

This shows that u(x) ≤ k, x ∈ B R . this completes the proof of theorem 3.3
2
Consequently, by the similar argument to −u(x) we can prove that u(x)
is local bounded. then with the compact property of u(x) we get theorem
3.4 as following:

Theorem 3.4 (Bounded on solution). the solution of (E)λ is bounded.

3.3 Hölder Continuity of Solution

3.3.1 Sobolev-Poincaré Inequality on Solution
Lemma 3.5. if p(x) ∈ P (Ω) satisfies ∂p/∂xi is bounded almost everywhere
in Ω, sup u(x) ≤ M, x ∈ BR , then there exist R1 = R1 (M, n, p(x)) such
that for every spherical neighborhood BR ⊂⊂ Ω, 0 < R < R1 and 1 < p ≤
p(x) ≤ p < p∗ < +∞, (x ∈ BR ) the solution of eigenvalue problem (E)λ
u(x) satisfies
  n−1
  p(x)n n
 u(x)  n−1
I I
 
 R 
 dx  ≤ C(n, p, q) |∇u(x)|p(x) dx+C(n)|BR | (3.14)
BR BR

Proof: In lemma 2.10 let γ = 1, notice that |{x ∈ BR : |u(x)| > 0}| ≤
|BR | we can get the inequality easily.

3.3.2 Harnack Inequality on Solution

Theorem 3.6. if p(x) ∈ P (Ω) satisfies ∂p/∂xi is bounded almost every-
where in Ω, 1 < p ≤ p(x) ≤ p < p∗ < +∞, in ω ⊂ Ω, sup u(x) ≤ M, x ∈ BR ,

17
then there exist R1 = R1 (M, n, p(x)) such that for every spherical neighbor-
hood BR ⊂ ω, 0 < R < R1 the solution of eigenvalue problem (E)λ u(x)
satisfies
 β I !1/q
|A(0, R)| p(x)
sup u(x) ≤ CRp/q |u/R| dx + Rn (3.15)
BR Rn A(0,R)
2

where β > 0, β(β + 1) = n1 , C = C(n, p, q, p(x), M )

Proof: Let h < k, R2 ≤ ρ < σ ≤ R < R1 < 1 where R1 is the same as

4
Lemma3.5. making the cutting function on Bρ such that |∇ζ(x)| ≤ σ−ρ .
By Hölder inequality, we have
 u − k p(x)
I  

 ρ 
 dx
A(k,ρ)
  n−1
I  +
 p(x)n n  1
 (u − k) ζ(x)  n−1 A(k, ρ) n
≤ C(n, q)    dx 
BR
 R  Rn

 u − k p(x)
I  

 ρ 
 dx
A(k,ρ)
I     1
 +

p(x)  σ + ρ  A(k, ρ) n
≤ C(n, p, q)  ∇ (u − k) ζ(x)  dx + A k,

BR 2  Rn
(3.16)

However,by assumed conditions for x ∈ A k, σ+ρ

2 we have

 
+
 u(x) − k 
|∇(u(x) − k) ζ(x)| ≤ |∇u(x)| + 4 
  (3.17)
σ−ρ 

then it follows from (3.17) and (3.16) that

 u − k p(x)
I  

 ρ 
 dx
A(k,ρ)
  1 (I  p(x) !
|A(k, ρ)| n u − k
|∇u(x)|p(x) + 
 
≤ C(n, p, q)  dx
Rn A(k, σ+ρ ) σ − ρ
2
  
 σ + ρ 
+ A k, (3.18)
2 

18
combining (3.18) and Lemma 3.2 we have

 u − k p(x)
I  

 ρ 
 dx
A(k,ρ)
 1 (I )
 u − k p(x)
  
|A(k, ρ)| n
≤ C(n, p, q)   dx + |A (k, σ)|
Rn A(k,σ) σ − ρ
 

(3.19)

moreover when h < k

σp  u − h p(x)
Z  
|A(k, ρ)| ≤   dx (3.20)
(k − h)q A(h,σ)  σ 
 u − k p(x)  u − k p(x)
I   I  

 σ 
 dx ≤ 
 σ 
 dx (3.21)
A(k,σ) A(h,σ)

therefore, from (3.19) and (3.21)

 u − k p(x)
I  

 ρ 
 dx
A(k,ρ)
 1 (I
 u − h p(x)
  
|A(h, ρ)| n
≤ C(n, p, q)   dx
Rn A(h,σ) σ − ρ
 
)
σp  u − h p(x)
I  
+   dx (3.22)
(k − h)h A(h,σ)  σ − ρ 

noticing that
β !β
σp  u − h p(x)
  
|A(k, ρ)|
I
≤ C(n)   dx (3.23)
Rn (k − h)q A(h,σ)
 σ 

if we multiply the two sides of (3.23) by

 u − k p(x)
I  

 ρ 
 dx
A(k,ρ)

and
 1 (I )
 u − h p(x) p  u − h p(x)
    
|A(h, ρ)| σ
n
I
C(n, p, q)   dx +   dx
Rn A(h,σ) σ − ρ
  (k − h)h A(h,σ)  σ − ρ 

19
we get

|A(k, ρ)| β  u − k p(x)

  I  
  dx
Rn A(k,ρ)
 ρ 
1 !β
σp  u − h p(x)
  
|A(h, ρ)| n
I
≤ C(n, p, q)   dx
Rn (k − h)q A(h,σ)  σ 
(I )
 u − h p(x) p  u − h p(x)
   
σ
I

σ − ρ
 dx + h
  dx (3.24)
A(h,σ) (k − h) A(h,σ)
σ − ρ

p
R
Now we take σ = Ri = R2 + 2i+1 , ρ = Ri+1 , h = ki = dR q (1 − 21i ), k = ki+1
for every i ∈ N and some d ∈ R to be chosen later. Taking into account that
p
dR q R
ki+1 − ki = 2I+1
, Ri − Ri+1 = 2i+1
from (3.24) we have

|A(ki+1 , Ri+1 )| β  u − ki+1 p(x)

  I  
  dx
Rn A(ki+1 ,Ri+1 )
 Ri+1 
1 I !1+β
2(1+β)qi Rn  u − ki p(x)
   
|A(ki , Ri )| n
≤ C 1+ q   dx
dqβ d Rn A(ki ,Ri )
 Ri 

where C = C(n, p, q). let

|A(k, ρ)| β  u − k p(x)

  I  
ϕ(k, ρ) =   dx
Rn A(k,ρ)
 ρ 

we have
2(1+β)qi Rn
 
ϕ(ki+1 , Ri+1 ) ≤ C(n, p, q) 1+ q ϕ1+β (ki , Ri ) (3.25)
dqβ d
(1+β)q
−
choosing d such that dq ≤ Rn and ϕ(k0 , R0 ) = ϕ(0, R) ≤ C(n, p, q)2 β2 dq
from (3.25) and Lemma2.7, we get
p R
lim ϕ(ki , Ri ) = ϕ(dR q , )=0 (3.26)
i→+∞ 2
taking d = Rn + C(n, p, q)ϕ(0, R) we deduce the desired result.

Theorem 3.7. the weak solution of (E)λ is local Hölder continuous.

Proof: by the sane proof with −u(x) and notice that BR |u(x)|p(x) dx is
H

bounded with p < p(x) < q we can get the estimate of u(x) on BR

R p
osc(u(x), ) ≤ C(n, p, q, M )R q (3.27)
2
where M = supBR u(x).

20
3.4 On the First Eigenvalue
3.4.1 Comparison Principle
Lemma 3.8 (Comparison Principle). Let F (x, u) : Ω×R1 → R1 be measur-
able in x and monotone nondecreasing in u, let u1 , u2 ∈ W 1,p(x)(Ω) satisfies

−∆p(x) u1 + F (x, u1 ) ≤ −∆p(x) u2 + F (x, u2 ) (3.28)

′
in W −1,p (x) (Ω), p′ (x) = p(x)/p(x) − 1 Then u1 ≤ u2 on ∂Ω implies u1 ≤ u2
in Ω.

Proof :Put ω(x) = max(u1 − u2 , 0) by u1 , u2 ∈ W 1,p(x) (Ω), ω(x) ∈

W 1,p(x) (Ω).
Multiplying(3.28) by ω and using the monotonicity of F (x, u),
we have
Z


−∆p(x) u1 (x) + F (x, u1 (x)) ω(x)dx
ZΩ


≤ −∆p(x) u2 (x) + F (x, u2 (x)) ω(x)dx
ZΩ Z
−∆p(x)u1 (x)ω(x)dx + F (x, u1 (x))ω(x)dx
Ω Ω
Z Z
≤ −∆p(x)u2 (x)ω(x)dx + F (x, u2 (x))ω(x)dx
Ω Ω
or
Z Z
−∆p(x) u1 (x)ω(x)dx + F (x, u1 (x))ω(x)dx
{x∈Ω:u1 (x)≥u2 (x)} {x∈Ω:u1 (x)≥u2 (x)}
Z Z
≤ −∆p(x) u2 (x)ω(x)dx + F (x, u2 (x))ω(x)dx
{x∈Ω:u1 (x)≥u2 (x)} {x∈Ω:u1 (x)≥u2 (x)}

Z
−∆p(x)u1 (x)ω(x)dx
{x∈Ω:u1 (x)≥u2 (x)}
Z
≤ −∆p(x)u2 (x)ω(x)dx
{x∈Ω:u1 (x)≥u2 (x)}

according to the definition of −∆p(x) u(x) it follows from above that

Z
p(x)(|∇u1 (x)|p(x)−2 ∇u1 (x)−|∇u2 (x)|p(x)−2 ∇u2 (x))(∇u1 (x)−∇u2 (x))dx ≤ 0
D
(3.29)

where D = {x ∈ Ω : u1 (x) ≥ u2 (x)} but

(|∇u1 (x)|p(x)−2 ∇u1 (x) − |∇u2 (x)|p(x)−2 ∇u2 (x))(∇u1 (x) − ∇u2 (x)) ≥ 0

21
hence,

∇u1 (x) = ∇u1 (x), x ∈ {x ∈ Ω : u1 (x) ≥ u2 (x)}

it’s means ∇ω(x) = 0 or u1 (x) = u2 (x) when x ∈ {x ∈ Ω : u1 (x) ≥ u2 (x)}

which implies u1 (x) ≤ u2 (x), x ∈ Ω.
1,p(x)
Lemma 3.9 (Extremum Principle). If u(x) ∈ W0 (Ω) ∩ C 1 (Ω) satisfies

 −∆p(x) u(x) + M up(x)−1 (x) ≥ 0, in W −1,p(x) (Ω), M ≥ 0



u(x) > 0, x∈Ω

u(x) = 0, x ∈ ∂Ω


∂u
Then the outer normal derivative ∂n of u is strictly negative on ∂Ω.
Proof: for any given x0 ∈ ∂Ω and a sufficiently small R > 0, There
exists y ∈ Ω such that B2R (y) ⊂ Ω and x0 ∈ ∂B2R (y) ∩ ∂Ω where Bρ (z) =:
{x ∈ Rn : |z − x| < ρ}. Set

υ(x) = α(3R − r)δ − αRδ , r = |x − y|

for fixed δ taking sufficiently small α, R such that

−∆p(x) υ(x) + M υ p(x)−1 (x) ≤ 0, x ∈ ΩR (3.30)
υ(x) ≤ u(x), x ∈ ∂ΩR (3.31)
where ΩR = B2R (y) \ BR (y). Now we proof that (3.30) and (3.31) is
valid. when x ∈ ∂B2R (y), υ(x) ≤ u(x) is trivial. noticing that υ(x) =
(2δ − 1)αRδ , x ∈ ∂BR (y), u(x) > 0, x ∈ ∂BR . we can get (3.31) by taking
sufficiently small α, R.
For (3.30) according to the chain rule of differential,
−Div(p(x)|∇(α(3R − r)δ − αRδ )|p(x)−2 ∇(α(3R − r)δ − αRδ ))
n
X ∂p(x) ∂r
= (αδ(3R − r)δ−1 )p(x)−1 ·
∂r ∂xi
i=1
n
δ−1 p(x)−1 δ−1
X ∂p(x) ∂r
+(αδ(3R − r) ) ln αδ(3R − r) ·
∂r ∂xi
i=1
p(x) − 1
−p(x)(αδ(3R − r)δ−1 )p(x)−1 αδ(δ − 1)(3R − r)δ−2
αδ(3R − r)δ − 1
n 1
+p(x)(αδ(3R − r)δ−1 )p(x)−1 ( − 3 )
r r
≡ H1 + H2 + H3 + H4
By assumed conditions, ∂p(x) ∂r
∂xi and ∂xi are both bounded for every i =
1, . . . , n. hence H1 ≤ C1 (αδ(3R − r)δ−1 )p(x)−1 ≤ C2 (αδRδ−1 )p(x)−1 Tak-
ing sufficient small R such that ln αδ(3R − r)δ−1 ≤ 0, nr − r13 ≤ 0 then

22
we get H2 ≤ 0, H4 ≤ 0. because p(x) is bounded in Ω, we have H3 ≤
δ−1
−C3 (αδRδ−1 )p(x)−1 3R−r . And noticing that M υ(x)p(x)−1 = M (αδ(3R −
rξ )δ−1 (2R − r))p(x)−1 where r ≤ rξ ≤ 2R we also have M υ(x)p(x)−1 ≤
M1 (αδRδ−1 )p(x)−1 stands for sufficiently small 2R − r All the conditions
shown above imply that

−∆p(x) υ(x) + M υ p(x)−1 (x)

δ−1
≤ C2 (αδRδ−1 )p(x)−1 − C3 (αδRδ−1 )p(x)−1 + M1 (αδRδ−1 )p(x)−1
3R − r
Let
δ−1
C2 (αδRδ−1 )p(x)−1 − C3 (αδRδ−1 )p(x)−1 + M1 (αδRδ−1 )p(x)−1 ≤ 0
3R − r
then we get
δ−1
C4 ≥ C2 + M1
3R − r
it stands for sufficiently small R. For fixed δ we prove that (3.30) and (3.31)
are valid. according to Lemma 3.8 we get the desired result.

3.4.2 The Eigenvalue Problem of Solution

We give some definitions about (E)λ as following

Defination 3.10. the first eigenvalue of (E)λ1 is

 
1 B(υ)  1,p(x)
o
= sup R(υ) := υ ∈ W =: W 0 (Ω)\{0} . (3.32)
λ1 A(υ) 

where
Z  
p(x) a(x) p(x)
A(υ) = |∇υ(x)| + |υ(x)| dx
Ω p(x)
b(x)
Z
B(υ) = |υ(x)|p(x) dx
Ω p(x)
Theorem 3.11 (Boundedness of the First Eigenvalue ). For the first eigen-
value,there exists C1 , C2 > 0 such that C1 < λ1 < C2 .
1,p(x)
proof: suppose B(u) is no positive for all u ∈ W0 (Ω), then there
1,p(x)
exists a function sequence fn in W0 (Ω) such that fn (x) ≤ 0 and fn (x) →
b (x) =: max{b(x), 0}, (n → ∞) which implies B(b+ (x)) ≤ 0 so b+ ≡ 0 it
+

contradict with the definition of b(x) therefore there must be exist a function
u0 (x) such that

B(u0 (x)) > 0 (3.33)

23
or
1
0 < λ1 < (3.34)
R(u0 (x))
On the other hand, by Lemma 2.9 and the definition of the first eigenvalue
we have
1
≤ C(Ω)kb(x)kL∞ (3.35)
λ1
So we have
1 1
≤ λ1 ≤ (3.36)
C(Ω)kb(x)kL∞ R(u0 (x))
1,p(x)
Theorem 3.12. There exists u ∈ W0 so that Jλ1 (u) = 0 implies u is
the solution of eigenvalue problem (E)λ . where
Z  
a(x) b(x)
Z
p(x) p(x)
Jλ1 (u) = |∇u(x)| + |u(x)| dx − λ1 |u(x)|p(x) dx
Ω p(x) Ω p(x)
Proof: make the Fréchet derivation of Jλ1 (u) we have
nR   o
p(x) + a(x) |u(x)|p(x) dx − λ
R b(x) p(x) dx
′
d Ω |∇u(x)| p(x) 1 Ω p(x) |u(x)|
Jλ1 (u) =
du(x)
p(x) dx − Ω |∇u(x)|p(x) dx
R R
Ω |∇(u(x) + th(x))|
= lim
t→0 th(x)
R a(x) p(x) dx −
R a(x) p(x) dx
Ω p(x) |u(x) + th(x)| Ω p(x) |u(x)|
+ lim
t→0 th(x)
R b(x) p(x) dx −
R b(x) p(x) dx
Ω p(x) |u(x) + th(x)| Ω p(x) |u(x)|
−λ1 lim
t→0 th(x)
= G1 + G2 + G3 = 0
where h(x) ∈ C0∞ (Ω) and
p(x) dx − p(x) dx
R R
|∇(u(x) + th(x))| Ω |∇u(x)|
G1 = lim Ω
t→∞ th(x)
1 p(x) − |∇u(x)|p(x) dx
R

lim t→∞ t |∇(u(x) + th(x))|
= Ω
h(x)

1 d|∇(u(x) + th(x))|p(x) 
Z
=  dx
h(x) Ω dt 
t=0
1
Z
= p(x)|∇u(x)|p(x)−2 ∇u(x)∇h(x)dx
h(x) Ω
1
Z  
= −Div p(x)|∇u(x)|p(x)−2 ∇u(x) h(x)dx (3.37)
h(x) Ω

24
 a(x) a(x)
+ th(x)|p(x) dx − p(x) dx
R R
Ω p(x) |u(x) Ω p(x) |u(x)|
G2 = lim
t→0 th(x)
1
Z
= a(x)|u(x)|p(x)−1 h(x)dx (3.38)
h(x) Ω

b(x) b(x)
+ th(x)|p(x) dx − p(x) dx
R R
Ω p(x) |u(x) Ω p(x) |u(x)|
G3 = lim
t→0 th(x)
1
Z
= b(x)|u(x)|p(x)−1 h(x)dx (3.39)
h(x) Ω

combining (3.37), (3.38) and (3.39) we get

Z   Z
p(x)−2
−Div p(x)|∇u(x)| ∇u(x) h(x)dx + a(x)|u(x)|p(x)−1 h(x)dx
Ω Z Ω

= λ1 b(x)|u(x)|p(x)−1 h(x)dx
Ω

therefore, u(x) is the solution of eigenvalue problem (E)λ . Noticing that

1,p(x)
Jλ1 (u(x)) = Jλ1 (|u(x)|) and W0 (Ω) ֒→ Lp(x) (Ω) is a continuously com-
pact imbedding. we obtain the existence of nonegative solution about the
eigenvalue problem (E)λ1 .

Theorem 3.13. The first eigenvalue λ1 is simple, that is to say, the set of
solutions is {tu(x) : t ∈R}

Proof: the sufficient is obvious. Let u1 , u2 be two solutions of (E)λ1 and

M (t, x) = max(u1 , tu2 ), m(t, x) = min(u1 , tu2 ). because

1 B(υ) 1,p(x) B(M )

= sup{R(υ) := ; υ ∈ W := W0 (Ω)\{0}} ≥
λ1 A(υ) A(M )

therefore

Jλ1 (M ) = A(M ) − λ1 B(M ) ≥ 0 (3.40)

In the same like,

Jλ1 (m) = A(m) − λ1 B(m) ≥ 0 (3.41)

Now we show that M, m are the solutions of (E)λ1 or Jλ1 (M ) = Jλ1 (m) = 0.

25
By the definition of Jλ1 it’s easy to see that

Jλ1 (M ) + Jλ1 (m)

= A(M ) − λ1 B(M ) + A(m) − λ1 B(m)
Z   Z
p(x) p(x)
= |∇M | + a|M | dx − λ1 b|M |p(x) dx
ΩZ Ω
Z
 
p(x) p(x)
+ |∇m| + a|m| dx − λ1 b|m|p(x) dx
Z Ω  ZΩ
p(x) p(x)
= |∇M | + a|M | dx − λ1 b|M |p(x) dx
I u1 I u1
Z   Z
+ |∇m|p(x) + a|m|p(x) dx − λ1 b|m|p(x) dx
I u1 I u1
Z   Z
+ |∇M |p(x) + a|M |p(x) dx − λ1 b|M |p(x) dx
Itu2 Itu2
Z   Z
+ |∇m|p(x) + a|m|p(x) dx − λ1 b|m|p(x) dx
Itu Itu
Z  2  Z 2

= |∇u1 |p(x) + a|u1 |p(x) dx − λ1 b|u1 |p(x) dx

ΩZ ΩZ
 
+ |∇u2 |p(x) + a|u2 |p(x) dx − λ1 b|u2 |p(x) dx
Ω Ω
= Jλ1 (u1 ) + Jλ1 (u2 ) = 0

where Iu1 = {x ∈ Ω : u1 (x) ≥ u2 (x)}, Itu2 = {x ∈ Ω : u1 (x) < u2 (x)} ac-

cording to (3.40) and (3.41) u1 , tu2 be two solutions of (E)λ1 . From theorem
3.7 we have M ∈ C01,θ (Ω) for all t ≥ 0.
For certain given x0 ∈ Ω, we take t0 = uu21 (x(x0 )
0)
. As we know, for every
vector e there stands

u1 (x0 + he) − u1 (x0 ) ≤ max(u1 (x0 + he), t0 u2 (x0 + he)) − u1 (x0 )

= M (t0 , x0 + he) − M (t0 , x0 )

therefore, the partial derivative of u1 (x) and M (t, x) at x0 satisfies

∂u1 (x0 ) u1 (x0 + hei ) − u1 (x0 )
= lim
∂xi h→0+ h
M (x0 + hei ) − u1 (x0 ) ∂M (x0 )
≥ lim = ,
h→0+ h ∂xi
∂u1 (x0 ) u1 (x0 + hei ) − u1 (x0 )
= lim
∂xi h→0− h
M (x0 + hei ) − u1 (x0 ) ∂M (x0 )
≤ lim =
h→0+ h ∂xi

26
or
∂u1 (x0 ) ∂M (x0 )
= (3.42)
∂xi ∂xi
where ei (i = 1, 2, · · · , n) are the unit normal
 vectors. Moreover,
 we have
∂ ∂ ∂
∇x u1 (x0 ) = ∇x M (t0 , x0 ), where ∇x = ∂x1 , ∂x2 , · · · , ∂xn By the same
way, we obtain ∇x t0 u2 (x0 ) = ∇x M (t0 , x0 ). Hence, the gradient of t0uu12 at
x0 is
 
u1 (x0 ) u2 (x0 )∇x u1 (x0 ) − u1 (x0 )∇x u2 (x0 )
∇x = =0
t0 u2 (x0 ) u1 (x0 ) − u2 (x0 )
By the arbitrary of x0 , we get
u1 (x)
≡ const, x ∈ Ω (3.43)
u2 (x)

Theorem 3.14. (E)λ has no solution for λ > λ1 .

Proof: By theorem 3.12, we need only verify it for the positive solution.
Let u, υ be positive solutions of (E)λ1 and (E)λ respectively. Assume b(x) ≥
0, from above we can select some solutions u, υ such that u ≤ υ for all x ∈ Ω.
then we deduce that there must exist 0 < η < 1 so that

−∆p(x) u + aup(x)−1 ≤ −∆p(x) (ηυ) + a(ηυ)p(x)−1 (3.44)

By the definition of solution and u ≤ υ ,because for all h ∈ C0∞ (Ω) and
h>0
Z Z
−Div(p(x)|∇u|p(x)−2 ∇uhdx + a|u|p(x)−2 uhdx
Ω Z Ω

= λ1 b|u|p(x)−2 uhdx
ZΩ

≤ λ1 b|υ|p(x)−2 υhdx
Ω
λ1 λ1
Z Z
p(x)−2
= −Div(p(x)|∇υ| ∇υhdx + a|υ|p(x)−2 υhdx
λ Ω λ Ω
We need only to prove that
λ1 λ1 a
Z Z
|∇υ|p(x)−2 ∇υ∇hdx + |υ|p(x)−2 hdx
λ Ω λ Ω p(x)
a
Z Z
p(x)−2
≤ |∇(ηυ)| ∇(ηυ)∇hdx + |ηυ|p(x)−2 hdx
Ω Ω p(x)

for certain 0 < η < 1. We obtain the desired result by taking inf η p(x)−1 ≥
λ1
λ . therefore applying for Lemma 3.8 u ≤ ηυ in Ω. Repeating this prcedure,

27
we deduce that u ≤ η n υ in Ω for all n ∈ N, which follows u ≡ 0. This is a
contradiction.
For general case, let B + (x) =: max(b(x), 0), b− (x) =: max(−b(x), 0).
Then above result implies the equation

−∆p(x) ω(x) + a(x) + λb− (x) ω(x)p(x)−1 = µb+ (x)ω p(x)−1



has a nontrivial positive solution ω if

µ ≤ µ1 = λ1 (a(x) + λb− (x), b+ (x))

and
b− (x) b+ (x)
Z Z
Iµ1 (ω(x)) = A(ω(x)) + λ |ω(x)|p(x) dx − µ1 |ω(x)|p(x) dx
Ω p(x) Ω p(x)
= min {Iµ1 (z(x)); z(x) ∈ W }
= 0

Since υ is a positive solution of the above equation with µ1 = λ we deduce

that λ ≤ µ1 and

Jλ (υ) = Iλ (υ) ≥ Iµ1 (υ) = min {Iµ1 (z(x)); z(x) ∈ W } = 0

However,

Jλ (u) = Jλ1 (u) − (λ − λ1 )B(u) < 0

1,p(x)
for all u ∈ W0 . This is a contradiction.

Remark
In this chapter ,we give many properties of the solutions of (E)λ in the
sense of weak. The most important part we discussed are the bounded-
ness and Hölder continuity of the weak solutions, which are also important
to weak solution. Therefore,the eigenvalue study in the partial differen-
tial operators, in essence, includes the solution research. Finally, we show
that the conditions restrict to p(x) are not necessary.it can be replaced
by1 < p ≤ p(x) ≤ q < +∞.

28
Chapter 4

Conclusion

In this paper we introduced certain p(x)-Laplace operator under the gener-

alized Sobolev Space and proved some conclusions as following:
We first proved the existence,boundary and Hölder continuity of the
solutions about the p(x)-Laplace equation which generalizes the result of
M.ôtani and T.Teshima and in the same time we obtained two inequalities:
the Caccioppoli inequality and the Harnack inequality about the equation
with p(x) exponent which are two basic conclusions in the research of equa-
tions with nonstandard exponent condition.
On the other hand, about the eigenvalue of p(x)-Laplace equation we
showed some properties:All eigenvalues are bounded. Solutions about the
first eigenvalue λ1 is simple i.e. all the nontrivial solutions form an one
1,p(x)
dimension subset of the space W0 (Ω). The equation has no solution with
λ > λ1 . Furthermore,we show that the outer norm vector of the positive
solutions is strictly negative at the boundary ∂Ω and got the comparison
principle. Consequently, by the similar argument we generalized the result
of M.ôtani and T.Teshima.
From the view of this point, how about the second eigenvale, the third,· · ·
and the solution about them? this is our aftertimes works.

29
Acknowlege

In this article, my supervisor, YongQiang Fu, has given me so many advices

that I can finish it completely.I give my best grateful thanks to him. Some
relative problem about the generalized Sobolev Space has been studied in
several articles written by him.

30
Bibliography

[1] O.Kováik and J.Rákosnik. On Spaces Lp(x) (Ω) and W k,p(x)(Ω).

Czechoslovak Math. J. 1991, 41(116): 592-618

[2] P.Maracellini. Regularity and Existence of Solutions of Elliptic Equa-

tions with p,q-growth Conditions. J. Defferential Equations. 1991,
1(50): 1-30

[3] S. Samko. Convolution and Potential Typeoperators in Lp(x) (Rn ). In-

tegral Transform. Spec. Funct. 1998, 3-4(7):261-284

[4] S. Samko. Convolution Type Operators in Lp(·) . Integral Transform.

Spec. Funct. 1998, 7(1-2): 123-144

[5] I.Isharapudinov. On the Topology of the Space Lp(x) [0, 1]. Matem. Za-
metki. 1978, 4(26): 613-632

[6] V.V.Zhikov. Averaging of Functionals of the Calculus of Variations an-

dElasticity theory. Math. USSR Izvestiya. 1987, 1(29): 33-66

[7] Ondrej Kovacik,Zilina and Jiri Rakosnik,Praha. on Spaces Lp(x) (Ω) and
W k,p(x)(Ω)

[8] E. Acerbi and G. Mingione. Regularity Results for a class of Functionals

with Non-standard Growth. Arch. Ration. Mech. Anal. 2001, (156):
121-140

[9] E. Acerbi and G. Mingione. Regularity Results for Stationary Electro-

rheological Fluids. Arch. Ration. Mech. Anal. 2002, (164): 213-259

[10] E. Acerbi and G. Mingione. Regularity Results for Electro-rheological

Fluids. the Stationary case C. R. Acad. Sci. ParisSer. 2002, (334): 817-
822

[11] E.Acerbi and N.Fusco. Semicontinuity problems in the calculus of vari-

ations. Arch. Rational Mech. Anal. 86(1984), 125-135

[12] A. Coscia and G. Mingione. Holder Continuity of the Gradient of p(x)-

harmonic Mappings. C. R. Acad. Sci. Paris Ser. 1999, I(328): 363-368

31
[13] D.Cruz-Uribe, A.Fiorenza and C.J.Neugebauer. the Maximal Function
on Variable Lp Spaces. Ann.Acad.Sci.Fenn.Ser.AIMath. 2003,(28):223-
238

[14] D. E. Edmunds and J. Lang and A. Nekvinda. On Lp(·) Norms R. Soc.

Lond. Proc. Ser. Amath. Phys. Eng. Sci. 1999, 1981(455): 219-225

[15] D. E. Emunds and A. Meskhi. Potential-type Operators in Lp(·) Spaces.

Z. Anal. Anwendungen. 2002, 3(21): 681-690

[16] D. E. Emunds and J.Rakosnik. Density of Smooth Functions in W k,p(·).

Proc.Roy.Soc.London.Ser. 1992,A(437):229-236

[17] D. E. Emunds and J.Rakosnik. Sobolev Embending with Variable Ex-

ponent. Studia Math. 2000,(143):267-293

[18] D. E. Emunds and J.Rakosnik. S obolev Embending with Variable Ex-

ponent ,II. Math.Nachr. 2002,(246-247):53-67

[19] L.Diening. Maximal Function on Generalized Lebesgue Spaces Lp .

Math.Inequal.Appl. to appear

[20] L.Diening. Riesz Potential and Sobolev Embendings of Generalized

Lebesgue Spaces Lp(·) and Sobolev Spaces W k,p(·) and Math. Nachr.
to appear

[21] L.Diening and M.Ruzicka. Calderon-Zygmund Operators on General-

ized Lebesgue Spaces Lp(·) and Problems related to Fluid Dynamics.
preprint

[22] L.Diening and M.Ruzicka. Integral Operators on the Halfspace in Gen-

eralized Lebesgue Spaces Lp(·) . priprint

[23] M.Ruzicka. Electro-rheological Fluids.Modeling and Mathematical

Theory. Springer.Verlag.Berlin. 2000

[24] P.Harjulehto and P.Hasto. A Capacity Approach to the Poincar In-

equality and Sobolev Imbending in Variable Exponent Sobolev Space.
Rev.Mat.Comput. to appear

[25] P.Harjulehto and P.Hasto. Lebesgue Points in Variable Exponent

Spaces. preprint

[26] P.Harjulehto, P.Hasto and M.Koskenoja. The Dirichlet Energy Integral

on Intervals in Variable Exponent Sobolev Spaces, preprint

[27] J.Heinonen, T.Kilpelainen and O.Martio. Nonlinear Potentials The-

ory ofDegenerate Elliptic Equations, Oxford Mathematical Mono-
graphs,Oxford University Press,Oxford, 1993

32
[28] A. Fiorenza. A Mean Continuity Type Results for Certain Sobolev
Spaces with Variable Exponent. Commun. Contemp. Math. 2 002, 3(4):
587-605

[29] T.C.Halsey. Electroheological fluids. Science. 1992,(258):761-766

[30] Yu.A.Alkhutov. the Harnack Inequality and the Hölder Property of

Solutions of Nonlinear Elliptic Equations with a Nonstandard Growth
Conditions. Differential Equations. 1997,33(12)1653-1662

[31] Valeria ChiadòPiat and Alessandra Coscia. Hölder Continuity of Min-

imizers of Functionals with Variable Growth Exponent. Manuscripta
Math. 1997,(93):283-299

[32] J. Rakosnik. Sobolev Inequality with Variable Exponent in Function

Spaces. Differential Operators and Nonlinear Analysis (Syote, Finland).
1999: 220-228

[33] Yongqiang Fu. the Existence of Solutions for Elliptic Systems with
Nonuniform Growth. Studia Mathematica. 2002,151(3):227-246

[34] M.ôtani and T.Teshima. on the First Eigenvalue of Some Quasilinear

Elliptic Equations. Proc.Japan.Acad. 1998,Ser.A(64):8-10

[35] Carlo Sbordone and Nicola Fusco. some Remarks on the Regularity of
Minima of Anisotropic Integrals. Commun.in Partial Differential Equa-
tions. 1993,(18):153-161

[36] A. Nekvinda: Hardy-Littlewood maximal operator on Lp(·) (Rn ). Math.

Inequal. Appl. to appear.

[37] V. Kokilashvili and S. Samko. Maximal and fractional operators in

weighted spaces Lp(·) (Rn ). Rev.Mat.Iberoamericana, to appear.

[38] X.Fan and Zhao Dun. the local C 1,α regularity of solution for p(x)-
Laplace equation. Journal of Gansu Education College. 2001,15(2):1-5

[39] X.Fan. On the positive solutions of p(x)-Laplace equation. Journal of

Gansu Education College. 2001,15(1):1-3

[40] X. Fan, J. Shen and D.Zhao. Sobolev Embedding Theorems for Spaces
W k,p(x)(Ω). J. Math. Anal. Appl. 2001, (262): 749-760

[41] X. Fan and D. Zhao. The Quasi-minimizer of Integal Functionals with

m(x) Growth Conditions. Nonlinear Anal. 2000, 39: 807-816

[42] X. Fan and D. Zhao. On the Spaces Lp(x) (Ω) and W k,p(x)(Ω). J. Math.
Anal. Appl. 2002, (263): 424-446

33
[43] X. Fan, Y. Zhao and D. Zhao. Compact Imbedding Theorems with
Symmetry of Strass-Lions Type for the Space W k,p(x)(Ω). J. Math.
Anal. Appl. 2001, (255): 333-348

[44] X. Fan. Existence of Positive Solutions for p(x)-Laplace Equa-

tions. Journal of Northwest Minorities University(Natural Science).
2000,(21):1-4

[45] X.Fan. A Strong Maximun Principle for p(x)-Laplace equation. Chinese

Journal of Contemporary Mathematics, 2003,(24)3:495-500

[46] X.L.Fan, Q.H.Zhang Existence of solutions for p(x)-Laplacian Dirichlet

problem Nonlinear Analysis 2003,(52):1843C1852

[47] A.Adams. “Sobolev Space”(Qixiao Ye et al.,trans). People’s Education

Publishing House,Beijing,1983[in Chinese].

[48] Chen.Ya.Zhe. “second order elliptic equation and equation system” sci-
ence Press, Beijing, 1991[in Chinese]

34