Fan e Zhao - de Giorgi e Continuidade 1997 PDF

Nonlinear Analysis 36 (1999) 295 – 318
A class of De Giorgi type and Holder continuity1

Xianling Fan∗ , Dun Zhao
Department of Mathematics, Lanzhou University, Lanzhou 730000, China
Received 30 August 1995; received in revised form 30 October 1996; accepted 12 March 1997
Keywords: Sobolev spaces; Minimizers; Weak solutions; Holder continuity
1. Introduction
In order to investigate the regularity of generalized solutions of quasi-linear elliptic

equations and variational problems Ladyzhenskaya and Ural’tseva have developed the
method pioneered by De Giorgi [3] and introduced the class Bm (
; M; ; 1 ; ; 1=q) and
proved Holder continuity of the functions belonging to this class (see [13]). It is well
known that the class Bm (· · ·) is very available in the case that so-called standard m-
growth conditions are satised (see e.g. [9, 11, 13, 16]). Following an example given
by Giaquinta [10] a lot of papers concerning the regularity of the solutions under
nonstandard growth conditions have appeared (see e.g. [1, 2, 8, 12, 14, 15]).
We are here interested in the case that so-called m(x)-growth conditions are satised.
For example, let us consider the variational functional
Z
J (u) = F(x; Du) dx

in which
⊂ Rn and F :
× Rn → R satises the following condition:
c1 |z|m(x) − c0 ≤ f(x; z) ≤ c2 |z|m(x) + c0 ; (1.1)

R
where 1¡p ≤ m(x) ≤ q for x ∈
. The simplest model is the functional
|Du|m(x) dx
which was investigated by Zhikov [19, 20] in other questions. Acerbi and Fusco [2]
have studied Holder regularity of minimizers of the functional J when m(x) takes only
∗Corresponding author.
1Research supported by the National Science Foundation of China (19671040) and the State Education
Commission Doctoral Foundation of China.
0362-546X/98/$19.00 ? 1998 Elsevier Science Ltd. All rights reserved.

PII: S 0 3 6 2 - 5 4 6 X ( 9 7 ) 0 0 6 2 8 - 7
296 X. Fan, D. Zhao / Nonlinear Analysis 36 (1999) 295 – 318
two values in
. Note that the m(x)-growth condition (1.1) is a special case of the
so-called (p; q)-growth conditions.
In the present paper, we shall consider the case that m(x) is a continuous function.
We shall introduce the class Bm(x) which is a generalization of the class Bm , and prove
the Holder continuity of elements in the class Bm(x) if the function m(x) satises some
additional condition (in particular if m(x) is Holder continuous). As applications of
this result we can prove the Holder continuity of minimizers of variational functionals
and of solutions of quasilinear equations with principal part in divergence form both
with m(x)-growth conditions.
In Section 2 we will give the statement and proof of our main result Theorem 2.1.
Sections 3 and 4 are applications of Theorem 2.1. In Section 3, we will discuss the
Holder continuity of minimizers of variational functional and in Section 4 we will
discuss the Holder continuity of solutions for quasilinear elliptic equations, our main
results are Theorems 3.1 and 4.4.
2. Notations and main results
The following notations will be used throughout this paper. Rn is the n-dimensional
euclidean space.
is a bounded domain in Rn . W 1; p (
) (p ≥ 1) is the Sobolev space
with norm
Z 1=p Z 1=p
p p
kukW 1; p (
) = |u| dx + |Du| dx :

n
For a measurable set E in R we denote by mes E or |E| the n-Lebesgue measure
of E. For a measurable function u dened in
and a measurable set E ⊂
denote
max u(x) = ess sup u(x); min u(x) = ess inf u(x);
E x∈E E x∈E
osc(u; E) = max u(x) − min u(x):

E E
For x0 ∈ Rn and ¿0; B (x0 ) = {x ∈ Rn : |x − x0 |¡}. B (x0 ) is also written simply

by B . B1 and B2 are the balls with common center. !n = |B1 |.
= {u: u is Holder continuous on
C 0; (
) with Holder exponent }
0 ); ∀
0 ⊂ ⊂
};
C 0; (
) = {u: u ∈C 0; (
where 0¡ ≤ 1.
L∞ ∞
+ (
) = {m∈L (
): 1¡ min m(x)}:

In the following discussion m ∈L∞

+ (
) is a given function. We write
p = min m(x); q = max m(x);

1; m(x)
R
W (
) = {u ∈W (
):
|Du|m(x) dx¡∞};
1; 1
X. Fan, D. Zhao / Nonlinear Analysis 36 (1999) 295 – 318 297
W01; m(x) (
) = W 1; m(x) (
) ∩ W01; 1 (
);
1; m(x)
Wloc (
) = {u: u ∈ W 1; m(x) (
0 ); ∀
0 ⊂ ⊂
}:
Note that p¿1 since m ∈L∞ + (

). Obviously, we have W
1; q
(
) ⊂ W 1; m(x)
1; p
(
) ⊂ W (
).
Now we list several well known lemmas which will be used below.
Lemma 2.1 (see Lemma 4.7 of Chapter II of [13]). Suppose a sequence yh ; h = 0; 1;

2; : : : ; of nonnegative numbers satises the recursion relation
yh+1 ≤ cbh yh1+ ; h = 0; 1; 2; : : : (2.1)

2
with some positive constants c; and b¿1. If y0 ≤ = c−(1=) b−(1= ) ; then yh ≤ b−(h=)
and consequently yh → 0 for h → ∞.
Lemma 2.2 (see Lemma 3.9 of Chapter 2 of [13]). For any u ∈W 1; 1 (B ) and
arbitrary numbers k and l with l¿k; the following inequality holds
Z
1
1− n n
(l − k)|Al; | ≤ |Du| dx (2.2)
|B − Ak; | Ak; −Al;
where Ak; = {x ∈B : u(x)¿k}; = (n)¿1 is a constant depends only on n.
Lemma 2.3 (see Lemma 4.8 of Chapter II of [13]). Suppose a function u(x) is mea-
surable and bounded in some ball BR0 . Consider balls BR and BbR which have a
common center with BR0 ; where b¿1 is a xed constant, and suppose that for any
0¡R ≤ b−1 R0 at least one of the following two inequalities is valid:
osc{u; BR } ≤ c1 R ; osc{u; BR } ≤ osc{u; BbR };
where c1 ; ≤ 1 and ¡1 are positive constants. Then u ∈C 0; (BR0 ); where = min{;
−logb }.
Lemma 2.4 (see Lemma 2.4 of [8]). Let u ∈ W 1; p (

)(p¿1): If for any B ⊂ ⊂
with ¡R0 and for any ∈(0; 1) and any k ≥ k0 ¿0

Z "Z ∗ #
u − k p
p
|Du| dx ≤ c r
dx + (k + 1)|Ak; | (2.3)
Ak ; − Ak;
where 0¡r ≤ p∗ ; p∗ is the Sobolev embedding exponent of p; c is a positive constant,

then u is locally bounded above on
.
In this paper, our main result is Theorem 2.1. Here we give
Denition 2.1 Let M; ; 1 ; are positive constants with ≤ 2. We will say that a
function u(x) belongs to class Bm(x) (
; M; ; 1 ; ) if u ∈W 1; m(x) (
); max
|u(x)| ≤ M;
and the functions w(x) = ±u(x) satisfy the inequalities

Z Z
w(x) − k m(x)
|Dw|m(x) dx ≤ dx + 1 |Ak ; | (2.4)

Ak ;− Ak ;
for arbitrary B ⊂
; ∈ (0; 1) and such k
k ≥ max w(x) − M; (2.5)

B
where Ak; = {x ∈ B : w(x)¿k}.
Remark 2.1. In [13] the class Bm (

; M; ; 1 ; ; 1=q) with a constant m¿1 is dened.
We may dene the general class Bm(x) (
; M; ; 1 ; ; 1=q). For brevity in Denition 2.1
we consider only the case of 1=q = 0.
Denition 2.2. We will say that a function m :

→ R satises the condition (H) in
if there exists a constant L¿0 such that
R−osc{m; BR } ≤ L for all B R ⊂

: (2.6)
Remark 2.2. It is easy to see that

(i) If m satises condition (H) in
; then m is continuous on
.
then m satises condition (H).
(ii) If m ∈C 0; (
);
(iii) If m satises the condition
|x − y|−|m(x)−m(y)| ≤ L1 for all x; y ∈

;
or
|m(x) − m(y)| · ln|x − y|−1 ≤ ln L1 for all x; y ∈

:
then m satises condition (H).
Theorem 2.1. If m satises condition (H); then Bm(x) (

; M; ; 1 ; ) ⊂ C 0; (
); where
the constant ∈(0; 1] depends only on the parameters n; p; q; L; and ; but it is
independent of M and 1 .
Let u ∈ Bm(x) (
; M; ; 1 ; ) in which m(x) satises the condition (H). Without loss
of generality we may assume that M ≥ 1, ≥ 1 and L ≥ 1.
To prove Theorem 2.1 it suces to prove that for each x0 ∈
there is a ball
BR0 (x0 ) ⊂
such that u ∈ C 0; (BR0 (x0 )) where is the constant as in Theorem 2.1.
Now let x0 ∈
be given arbitrarily. By the continuity of m(x) we can nd a positive
number R0 ¡1 such that B R0 (x0 ) ⊂
and
M osc{m; BR0 } ≤ 2: (2.7)

Take arbitrarily R ∈ (0; R0 ]. It is easy to see that at least one of the two functions
w = ± u satises the following condition

1 1
mes x ∈ BR=2 : w(x)¿ max w(x) − osc{u; BR } ≤ mes BR=2 : (2.8)
BR 2 2
From now on we denote by w the function identied to u or −u that satises

Eq. (2.8). Set
= max{2; 2=}; = −1 osc{u; BR } (2.9)
and
k 0 = max w(x) − : (2.10)

BR
Then
k 0 ≥ max w(x) − 12 osc{u; BR } (2.11)

BR
and
k 0 ≥ max w(x) − M: (2.12)

BR
Note that Eq. (2.12) implies that Eqs. (2.4) and (2.5) hold for k ≥ k 0 and ≤ R.
Now let
m− = m− (BR ) = min m(x); m+ = m+ (BR ) = max m(x):

BR BR
Under the above assumption and notations we can give the proof of Theorem 2.1
through the following Lemmas 2.5–2.8.
Lemma 2.5. There is a positive constant = (n; q; L; ) such that if
|Ak 0 ; R=2 | ≤ R n ; (2.13)
then at least one of the following two inequalities holds:

1=m+
+ 1 + 1
H≤ R; (2.14)

H
max w(x) ≤ k 0 + ; (2.15)
BR=4 2
where
0¡H ¡ ; k 0 = max w(x) − H: (2.16)

BR
Proof. For h = 0; 1; 2; : : : ; set

R R H H
h = + h+2 ; kh = k 0 + − h+1 ;
4 2 2 2
yh = R−n |Akh ; h |; D h+1 = Akh ; h+1\Akh+1 ; h+1 :
Applying Eq. (2.4) to k = kh , = h and − = h+1 for h = 0; 1; 2; : : : we obtain
Z Z
|Dw|m(x) dx ≤ |Dw|m(x) dx
Dh+1 Akh ; h+1
Z
w(x) − kh m(x)
≤ dx + 1 |Akh ; h |: (2.17)
h − h+1
Akh ; h
From Eq. (2.17) it follows that

Z Z
w(x) − kh m+
|Dw|m− dx ≤ dx + ( + 1 + 1)|Ak ; |

Akh ; h h − h+1
h h
Dh+1

+ 1 + 1 n
≤ 2(h+3)m+ R−m+ H m+ + R yh : (2.18)

Assume that Eq. (2.14) does not hold, then we have

+ 1 + 1
¡R−m+ H m+

and from this and Eq. (2.18) it follows that
Z
|Dw|m− dx ≤ [2(h+3)m+ R−m+ H m+ + R−m+ H m+ ]Rn yh
Dh+1
≤ 2(h+4)m+ H m+ Rn−m+ yh : (2.19)

Using Holder inequality and Eq. (2.19) we obtain
Z Z 1=m−
m−
|Dw| dx ≤ |Dw| dx |Dh+1 |1−(1=m− )
Dh+1 Dh+1
(h+4)m+
≤ ( 2 H m+ Rn−m+ yh )1=m− (R n yh )1−(1=m− )
≤ 2(h+4)q H (m+ =m− ) Rn−(m+ =m− ) yh : (2.20)
Applying inequality (2.2) to k = kh , l = kh+1 and = h+1 we get
Z
|Dw| dx ≥ −1 (kh+1 − kh )(R n yh+1 )1−(1=n) −n
h+1 |Bh+1\Akh ; h+1 |
Dh+1
1−(1=n) n −n
≥ −1 2−(h+2) HRn−1 yh+1 2 R (|BR=4 | − |A R |): (2.21)
k0; 2
If Eq. (2.13) holds and

1 −n
≤ · 4 !n ; (2.22)
2
then
1 −n
|BR=4 | − |A R| ≥ 4−n R n !n − · 4 !n R n = 2−2n−1 !n R n (2.23)
k0; 2 2
and consequently from Eq. (2.21) it follows that
Z
1−(1=n)
|Dw| dx ≥ −1 2−(h+3+n) HRn−1 !n yh+1 : (2.24)
Dh+1
From Eqs. (2.24), (2.20), (2.7) and (2.6) and
0¡H ¡ ≤ M; M ≥ 1; ≥ 1; L ≥ 1; R¡1
it follows that
Z
1−(1=n) h+3+n −1 1−n
yh+1 ≤ 2 H R !n−1 |Dw| dx
Dh+1
≤ 2h+3+n H −1 R1−n !n−1 2(h+4)q H m+ =m− Rn−(m+ =m− ) yh

≤ 2(q+1)h 23+n+4q !n−1 M (m+ −m− )=m− R1−(m+ =m− ) yh
≤ 2(q+1)h 24+n+4q !n−1 Rm− −m+ yh
≤ c1 bh1 yh ; (2.25)
where b1 = 2q+1 ¿1, c1 = 24+n+4q !n−1 L¿1 and = (n) as in Lemma 2.2. Hence
yh+1 ≤ cbh yh1+[1=(n−1)] = cbh yh1+ (2.26)
where b = bn=(n−1)
1 = b(n; q)¿1, c = c1n=(n−1) = c(n; q; ; L) and = 1=(n − 1).
Now we choose

1 −n −(1=) −(1=2 )
= min · 4 !n ; c b : (2.27)
2
Then = (n; q; L; ) and when
1=m+
n + 1 + 1
|Ak 0 ; R=2 | ≤ R and H¿ R

we have
2
y0 = R−n |Ak 0 ; R=2 | ≤ ≤ c−1= b−1= (2.28)
and consequently, by Lemma 2.1, yh → 0 as h → ∞. This shows

H
max w(x) ≤ lim kh = k 0 + :
BR=4 h→∞ 2
Lemma 2.5 is proved.
Lemma 2.6. For any given ¿0 there is a natural number s = s(; n; q; L; )¿2 such
that either
1=m+
+ 1 + 1
≤ 2s R (2.29)

or
|Ak 0 ; R=2 | ≤ R n (2.30)
holds; where
k 0 = max w(x) − 2−s+1 : (2.31)

BR
Proof. If Eq. (2.29) does not hold, i.e. supposing

1=m+
+ 1 + 1
¿2s R; (2.32)

where the natural number s will be chosen later.
For t = 0; 1; 2; : : : ; s − 1, set
kt = max w(x) − 2−t ; Dt = Akt ; R=2 \Akt+1 ; R=2 :

BR
Applying inequality (2.4) to = R, − = R=2 and k = kt for t = 0; 1; : : : ; s − 2, and

with (2.32) we can obtain
Z " Z #
+ 1 + 1
|Dw|m− dx ≤ 2m+ R−m+ (w(x) − kt )m+ dx + |Akt ; R |
Dt Akt ; R
≤ [2m+ R−m+ (2−t )m+ + (2−s )m+ R−m+ ]!n R n
≤ 2q+1 !n (2−t )m+ Rn−m+ (t = 0; 1; : : : ; s − 2): (2.33)
Applying inequality (2.2) to w with k = kt , l = kt+1 , = R=2 for t = 0; 1; 2; : : : ; s − 2,

we obtain
|Aks−1 ; R=2 |1−(1=n) ≤ |Akt+1 ; R=2 |1−(1=n)

Z
≤ 2−n Rn (kt+1 − kt )−1 |BR=2 \Akt ; R=2 |−1 |Dw| dx
Dt
Z 1=m−
−n n −(t+1) −1 −1 −1 n −1 m−
≤ 2 R (2 ) (2 !n (2 R) ) |Dw| dx |Dt |1−(1=m− )
Dt
Z 1=m−
≤ 4!n−1 2t −1
|Dw|m− dx |Dt |1−(1=m− ) : (2.34)
Dt
From Eqs. (2.34) and (2.33) it follows that

Z
|Aks−1 ; R=2 |(n−1)m− =n ≤ (4!n−1 2t −1 )m− |Dw|m− dx · |Dt |m− −1
Dt
≤ (2n !n−1 )q 2q+1 !n (2−t )m+ −m− Rn−m+ |Dt |m− −1 : (2.35)
As 2−t ≤ ≤ M , from Eq. (2.7) we have
(2−t )m+ −m− ≤ M m+ −m− ≤ 2
and consequently by Eq. (2.35)
|Aks−1 ; R=2 |(n−1)m− =n ≤ c1 Rn−m+ |Dt |m− −1 (t = 0; 1; : : : ; s − 2); (2.36)
where c1 = (2n !n−1 )q 2q+2 !n + 1 = c(n; q; )¿1.

From Eq. (2.36) we have
|Aks−1 ; R=2 |(n−1)m− =n(m− −1) ≤ cRn−m+ =m− −1 |Dt | (t = 0; 1; : : : ; s − 2); (2.37)
where c = c11=(p−1) = c(n; p; q; ).

Summing Eq. (2.37) with respect to t from 0 to s − 2 and noting that
s−2
X n
R
|Dt | ≤ |BR=2 | = !n
2
t=0
we obtain
c2−n !n (nm− −m+ )=m− −1
|Aks−1 ; R=2 |((n−1)m− )=n(m− −1) ≤ R
s−1
and therefore
(n(m− −1))=(n−1)m−
c2−n !n
|Aks−1 ; R=2 | ≤ R−(n(m+ −m− ))=(n−1)m− Rn
s−1
(n(m− −1))=(n−1)m−
c2−n !n
≤ Ln=n−1 Rn ; (2.38)
s−1
here we have made use of the condition (H).
Now we choose a natural number s such that s − 1¿c2−n !n and
−n (n(p−1))=(n−1)p
c2 !n
Ln=n−1 ≤ : (2.39)
s−1
Note that s = s(; c; n; p; L) = s(; n; p; q; L; ). Then for such s, when Eq. (2.29) does
not hold, from Eq. (2.38) we have
|Aks−1 ; R=2 | ≤ Rn

which means |Ak 0 ; R=2 | ≤ Rn , i.e., Eq. (2.20) holds, since
k 0 = max w(x) − 2−s+1 = ks−1 :

BR
Lemma 2.6 is proved.
Lemma 2.7. There is a number s = s(n; p; q; L; )¿2 such that

( 1=m+ )
s + 1 + 1
≤ 2 max max w(x) − max w(x); R (2.40)
BR BR=4
for any R ∈ (0; R0 ].
Proof. Let R ∈ (0; R0 ] and let = (n; p; q; L; ) be the constant as in Lemma 2.5. Ap-
plying Lemma 2.6 to this we can nd a constant s = s(; n; p; q; L; ) = s(n; p; q; L; )
such that at least one of Eqs. (2.29) and (2.30) holds. If Eq. (2.29) holds, then
Eq. (2.40) is obviously true. Now assume that Eq. (2.29) does not hold. Then by
Lemma 2.6, Eq. (2.30) holds, i.e.
|Ak 0 ; R=2 | ≤ Rn ;
where k 0 = maxBR w(x) − 2−s+1 .

Set H ¿2−s+1 . Then, since Eq. (2.29) does not hold,
1=m+ 1=m+
+ 1 + 1 + 1 + 1
H ¿2−s+1 · 2s R¿ R:

By Lemma 2.5
H
max w(x) ≤ k 0 + = max w(x) − 2−s
BR=4 2 BR
i.e.

≤ 2s max w(x) − max w(x) :
BR BR=4
This shows that Eq. (2.40) holds. Lemma 2.7 is proved.
Lemma 2.8. For any R ∈ (0; R0 ] at least one of the following two inequalities holds
+ 1 + 1
osc{u; BR } ≤ 2s R; (2.41)

osc{u; BR=4 } ≤ (1 − −1 2−s )osc{u; BR }; (2.42)
where = max{2; 2=}; s = s(n; p; q; L; ) is the constant as in Lemma 2.7.

Proof. By Lemma 2.7, Eq. (2.40) holds, this implies that at least one of the following
two inequalities holds
1=m+
+ 1 + 1
≤ 2s R; (2.43)

s
≤ 2 max w(x) − max w(x) : (2.44)
BR BR=4
When Eq. (2.43) holds we have

1=m+
s + 1 + 1 + 1 + 1
osc{u; BR } = ≤ 2 R ≤ 2s R;

i.e. Eq. (2.41) holds. When Eq. (2.44) holds we have

s
osc{u; BR } = ≤ 2 max w(x) − max w(x)
BR BR=4
≤ 2s [osc{w; BR } − osc{w; BR=4 }]
= 2s [osc{u; BR } − osc{u; BR=4 }]

this implies Eq. (2.42). Lemma 2.8 is proved.
By Lemma 2.3 and it follows from Lemma 2.8 that u ∈ C 0; (BR0 (x0 )) where
= min{1; − log4 (1 − −1 2−s )} = (n; p; q; L; ; ):
By the arbitrarity of x0 ∈
we have u ∈ C 0; (
). The proof of Theorem 2.1 is com-
pleted.
Now we give
Denition 2.3. We will say that u ∈ m(x) (

; M; ; 1 ; ) if u ∈ m(x) (
; M; ; 1 ; ) and
the inequalities (2.4) hold also for arbitrary ball B with B ∩ @
6= ; ∈ (0; 1) and

k ≥ max max w(x) − M; max w(x) ;

@

where
= B ∩
; @
= @
∩ B ; Ak; = {x ∈
: w(x)¿k}.
Denition 2.4 (see Section 1, Chapter 1 of [13]). We will say that the boundary @
of
satises the condition (A), if there exist two positive constants a0 and 0 such
that for any ball B with center on @
and radius ≤ a0 and for any connected branch
i of B ∩
, the following inequality holds:
|
i | ≤ (1 − 0 )|B |:
Similar to Theorem 2.1, we can prove

M; ; 1 ; ). If m satises condition (H), @

satises the
Theorem 2.2. Let u ∈ m(x) (
;
condition (A) and u|@
∈ C (@
); then u ∈ C 0; (
)
0; 1 where = (n; p; q; L; ; ; 1 ).
Remark 2.3. It is similar to Section 7 of Chapter II of [13] that we can dene the

class Bm(x) (
∪ S1 ; M; ; 1 ; ) with S1 ⊂ @
and prove the corresponding results.
3. Application to minimizers
Let F :
× R × Rn → R be a Caratheodory function satisfying
|z|m(x) − b(|u|r(x) + 1) ≤ F(x; u; z) ≤ a|z|m(x) + b(|u|r(x) + 1); (3.1)
where a and b are positive constants, m ∈ L∞ ∞

+ (
); r ∈ L (
) and
m(x) ≤ r(x)¡m∗ (x) for x ∈

; (3.2)
here m∗ (x) is the Sobolev embedding exponent of m(x).

Put
Z
1; m(x)
J (u; E) = F(x; u; Du) dx for E ⊂ ⊂
and u ∈ Wloc (
):
E
1; m(x)
Denition 3.1. A function u ∈ Wloc (
) is said to be a local minimizer of J if
J (u; supp ’) ≤ J (u + ’; supp ’) for ’ ∈ W 1; m(x) (

) with supp ’ ⊂ ⊂
: (3.3)
Theorem 3.1. Let Eqs. (3.1) and (3.2) hold and let the functions m(x) and r(x) be
1; m(x)
continuous on
. Suppose that u ∈ Wloc (
) is a local minimizer of J . Then
(i) u is locally bounded in
.
(ii) u ∈ C 0; (
) if m(x) satises the condition (H ); where the constant ∈ (0; 1] only
depends on the constants n; p; q; L and a.
We now let the assumptions of Theorem 3.1 hold. Let us rst prove the assertion
(i) of Theorem 3.1, i.e. prove the local boundedness of u in
.
Take arbitrarily x0 ∈
. Let B R0 (x0 ) ⊂
where R0 will be chosen below. We will
prove that u is bounded on BR0 =2 (x0 ).
Let
m− = min m(x); m+ = max m(x); r+ = max r(x):

BR0 BR0 BR0
We will restrict ourselves to the case 1¡m− ≤ n, since when m− ¿n the Holder
continuity of u on B R0 is trivial. Let
 nm
 −
if m− ¡n;
m∗− = n − m−

r+ + 1 if m− = n:
By Eq. (3.2) and the continuity of m(x) and r(x) we may choose R0 ¿0 suciently
small such that
r+ ¡m∗− : (3.4)
Note that Eq. (3.4) implies that
m+ ¡m∗− (3.5)
since by Eq. (3.2) m(x) ≤ r(x).

By the Sobolev inequality there is a constant c0 = c0 (m− ; n)¿0 such that for all
w ∈ W 1; m− (Bs ) that satisfy the condition
1
mes{x ∈ Bs : w(x) 6= 0} ≤ mes Bs (3.6)
2
the inequality
Z m− =m∗− Z
m∗
|w(x)| − dx ≤ c0 |Dw|m− dx (3.7)
Bs Bs
holds.
Note that here c0 = c0 (m− ; n)¿0 is independent of s.
We can choose R0 ¿0 suciently small such that
Z !(r+ −m− )=m∗−
r+ m∗ ∗ 1
b2 c0 |u| − dx |BR0 |1−(r+ =m− ) ≤ : (3.8)
BR0 2
For k ∈ (−∞; +∞) and s ∈ (0; R0 ] set
Ak = {x ∈ BR0 : u(x)¿k}; Ak; s = Ak ∩ Bs :
It is easy to see that there is a k0 ¿0 such that

1
|Ak | ≤ |BR0 =2 | for k ≥ k0 : (3.9)
2
Let R0 =2 ≤ t¡s ≤ R0 , k ≥ k0 , and
w(x) = max{u(x) − k; 0}: (3.10)
Let ∈ C ∞ (
) with supp ⊂ Bs ; 0 ≤ ≤ 1; ≡ 1 on Bt ; |D| ≤ 2(s − t)−1 .
Set v = u − w, by the minimality of u and Eq. (3.1), we have
Z
|Du|m(x) dx
Ak; s
Z Z Z
m(x) r(x)
≤a |Dv| dx + b (|v| + 1) dx + b (|u|r(x) + 1) dx
Ak; s Ak; s Ak; s
Z Z
u − k m(x)
≤ a · 2q−1 |Du|m(x) dx + a · 22q−1 dx

Ak; s \Ak; t Ak; s s − t
Z
+ b2r+ + |w|r+ dx + (b2r+ + k r+ + 2b2r+ + 2b)|Ak; s |: (3.11)
Ak; s
By Eqs. (3.9) and (3.10) for k ≥ k0 and R0 =2 ≤ s ≤ R0 we have

1 1
mes{x ∈ Bs : w(x) 6= 0} ≤ |Ak | ≤ |BR0 =2 | ≤ |Bs |
2 2
and therefore by Eq. (3.8)
Z Z Z r+ =m∗−
r+ r+ m∗ ∗
|w| dx = |w| dx ≤ |w|− dx |Bs |1−(r+ =m− )
Ak; s Bs Bs
Z (r+ −m− )=m∗− Z m− =m∗−

m∗ m∗ ∗
= |w| − dx |w| − dx |Bs |1−(r+ =m− )
Bs Bs
Z (r+ −m− )=m∗− Z

m∗ 1−(r+ =m∗
−)
≤ c0 |w| − dx |Bs | |Dw|m− dx: (3.12)
Bs Bs
Noting that |Bs | ≤ |BR0 | from Eqs. (3.12) and (3.8) we obtain
Z Z Z
1 1 1
b2r+ |w|r+ dx ≤ |Dw|m− dx ≤ |Du|m(x) dx + |Ak; s |: (3.13)
Ak; s 2 Bs 2 Ak; s 2

Z Z
m(x)
|Du| dx ≤ |Du|m(x) dx
Ak; t Ak; s
Z Z
u − k m(x)
≤ a · 2q |Du|m(x) dx + a · 22q dx

Ak; s \Ak; t Ak; s s − t
+ c0 (k r+ + 1)|Ak; s |; (3.14)
where c0 = c0 (b; Rr+ ).

Adding a · 2q Ak; t |Du|m(x) dx to both sides of Eq. (3.14) we get
Z Z Z
u − k m(x)
|Du|m(x) dx ≤ |Du|m(x) dx + c1 dx + c2 (k r+ + 1)|Ak; s |;
s−t
Ak; t Ak; s Ak; s
(3.15)
where = a2q =(a2q + 1) = (a; q)¡1; c1 = a22q =(a2q + 1) = c1 (a; q); c2 = c2 (a; b; q; r+ )
are positive constants.
Using the similar method that was used in the proof of Lemma 5.1 of Chapter V
of [9], from Eq. (3.15) we can deduce that
Z Z
u − k m(x)
m(x)
|Du| dx ≤ dx + c(k r+ + 1)|Ak; | (3.16)

Ak; − Ak;
for R0 =2 ≤ − ¡ ≤ R0 ; ∈ (0; 1) and k ≥ k0 , where = (; q) = (a; q); c = c(a; b;

q; r+ ) are positive constants.
Z Z ∗
u − k m−
m−
|Du| dx ≤ r+ + 1)|Ak; |;
dx + c(k (3.17)
Ak; − Ak;
b; q; r+ ) is a positive constant.
where c = c(a;
By Lemma 2.4 of [8], from Eq. (3.17) it follows that
max u(x)¡+∞:
BR0 =2
This shows that u is locally bounded above on

.
Since −u minimizes the functional
Z
J (v;
) = F(x;
v; Dv) dx

v; z) = F(x; −v; −z) satisfying the same growth condition (3.1), hence, by the
with F(x;
same reason, −u is locally bounded above on
and so u is locally bounded in
.
The assertion (i) of Theorem 3.1 is proved.
The assertion (ii) of Theorem 3.1 follows from Theorem 2.1 and the following
Lemma 3.1.
Lemma 3.1. Under the assumptions of Theorem 3.1 for any open set
1 in
with
1 ⊂
there are positive constants M; and 1 such that u ∈ Bm(x) (
1 ; M; ; 1 ; 1) and
in which the constant = (a; q) is independent of
1 .
1 ⊂
. By assertion (i) of Theorem 3.1 there
Proof. Let
1 be an open set in
with
is a constant M ¿0 such that
max |u(x)| ≤ M: (3.18)

1
From Eq. (3.1) it follows that there is a positive constant c = c(b; r+ ; M ) such that
|z|m(x) − c ≤ F(x; u; z) ≤ a|z|m(x) + c 1;

for x ∈
|u| ≤ 4M; z ∈ Rn ; (3.19)
where r+ = max
1 r(x).
Let Bs ⊂
1 ; 0¡t¡s; k ≥ −2M and let As; k ; w(x); ; v be the same as in the proof
of assertion (i) of Theorem 3.1.
1
Noting that for x ∈
|v(x)| ≤ |u(x)| + |w(x)| ≤ M + 3M = 4M

and using the minimality of u and Eq. (3.19) we get
Z Z
|Du|m(x) dx − c|Ak; s | ≤ a |Dv|m(x) dx + c|Ak; s |: (3.20)
Ak; s Ak; s
From Eq. (3.20), because of the same reason as in the proof of assertion (i) of
Theorem 3.1 we get
Z Z
u − k m(x)
|Du|m(x) dx ≤ dx − 1 |Ak; | (3.21)

Ak; − Ak;
for B ⊂
1 ; ∈ (0; 1) and k ≥ −2M , where = (a; q) and 1 = 1 (M; a; b; q; r+ ) are
positive constants.
Set = 1. Then for every B ⊂
1
max u(x) − M ≥ −2M
B
and consequently Eq. (3.21) holds for k ≥ maxB u(x) − M with = 1.

It is easy to see that Eq. (3.21) holds when u is replaced by −u. So by
Denition 2.1 u ∈ Bm(x) (
1 ; M; ; 1 ; 1) and Lemma 3.1 is proved.
In the case when m(x) satises condition (H) by Lemma 3.1 and Theorem 2.1
1 ) for every
u ∈ C 0; (
1 ⊂
and the constant ∈ (0; 1) is independent of
1 . Hence
0;
u ∈ C (
). The proof of Theorem 3.1 is completed.
R
Remark 3.1. In the proof of assertion (i) of Theorem 3.1 for b2r+ Ak; s |w|r+ dx we ob-
tain the estimation (3.13) which guarantees that the constant = (a; q) in
Eq. (3.17) is independent of b. This is not necessary for the proof itself of the
assertion (i). In fact since 0¡s − t¡1; r+ ¡m∗− ; we have
Z Z ∗
u − k m−
r+
|w| dx ≤ dx:

Ak; s Ak; s s − t
From this inequality we can obtain Eq. (3.17) with = (a; b; q; r+ ) and then prove
assertion (i).
4. Application to quasilinear elliptic equations
In this section we will consider the problem of the regularity of weak solutions of
equation
div A(x; u; Du) + B(x; u; Du) = 0; x∈
(4.1)
n n n
where
is an open bounded subset of R (n ≥ 2); A :
× R × R → R ; B :
× R ×
R n → R; u :
→ R. It is well known that under so-called standard m-growth condi-
tions (where m¿1 is a constant) the answer to this problem is positive, see e.g.
[3, 5, 9, 11, 13, 17, 21], but under the so-called (p; q)-nonstandard growth conditions the
answer is in general negative, see e.g. [10, 12]. Recently, some results concerning the
regularity of weak solutions of Eq. (4.1) or of minimizers of variational functionals
with nonstandard growth conditions have been obtained, see e.g. [1, 2, 4, 6–8, 14, 15, 18].
We are here mainly interested in the regularity of weak solutions of Eq. (4.1) with
m(x)-growth conditions. The simplest model is the Euler equation for the integral
functional
Z
|Du|m(x) dx (4.2)

which was investigated by Zhikov [19, 20] in other contents. Acerbi and Fusco [2]
proved the Holder continuity of minimizers of the functional (4.2) when m(x) takes
only two values. The main aim of this section is to obtain the Holder continuity of
weak solutions of (4.1) with m(x)-growth conditions, where m(x) satises condition
(H). For p ≥ 1; W 1; p (
) = {u: u ∈ Lp (
); |Du|p ∈ Lp (
)} is the well-known Sobolev
space, W01; p (
) is the closure of C0∞ (
) in W 1; p (
).
Let p∗ be the Sobolev embedding exponent of p, i.e.
np
∗ if p¡n;
p = n−p
any q¿p if p ≥ n:
Now we consider the boundedness of weak solutions of Eq. (4.1) satisfying the
following growth conditions:
A(x; u; z)z ≥ a0 |z|m(x) − b|u|r (x) − c; (4.3)
|A(x; u; z)| ≤ a1 |z|m(x)−1 + b|u|(x) + c; (4.4)
|B(x; u; z)| ≤ a2 |z|(x) + b|u|r (x)−1 + c; (4.5)

and
where a0 ; a1 ; a2 ; b; c are positive constants, m; r ∈ C 0 (
)
m(x) ≤ r(x)¡m∗ (x); (4.6)
m(x) − 1
(x) = r(x); (4.7)
m(x)
r(x) − 1
(x) = m(x): (4.8)
r(x)
Denition 4.1. u ∈ W 1; m(x) (

) is said to be a weak solution of Eq. (4.1) if
Z Z
A(x; u; Du) D dx − B(x; u; Du) dx = 0 (4.9)

for every ∈ W01; m(x) (

).
Theorem 4.1. Let Eqs. (4.3)–(4.8) hold. If u ∈ W 1; m(x) (

) is a weak solution of
Eq. (4.1), then u ∈ L∞ ∞
loc (
). If in addition u|@
is bounded, then u ∈ L (
).
Proof. Let u be a weak solution of Eq. (4.1). Let x0 ∈

. We will prove that u is
locally bounded at x0 . Take a ball BR0 (x0 ) ⊂
. Set
m− = min m(x); m+ = max m(x); r+ = max r(x): (4.10)

BR0 BR0 BR0
Without loss of generality we may assume m− ≤ n. By Eq. (4.6) and the continuity of
m and r we can choose R0 ∈ (0; 1) suciently small such that
m+ ≤ r+ ¡m∗− ; (4.11)
here we set
 nm
 −
if m− ¡n;
m∗− = n − m−

r+ + 1 if m− = n:
For arbitrary balls B t (x0 ) ⊂ Bs (x0 ) ⊂ BR0 (x0 ), let be a C ∞ function such that
0 ≤ ≤ 1; supp ⊂ Bs ; ≡ 1 on Bt ; |D| ≤ 2=s − t:
For k ≥ 1 set = m+ max{u − k; 0}. Then ∈ W01; m(x) (

). Using Eq. (4.9) we get
Z
A(x; u; Du)Du m+ dx
Ak; s
Z
+ m+ A(x; u; Du) D m+ −1 (u − k) dx
Ak; s
Z
− B(x; u; Du) m+ (u − k) dx = 0; (4.12)
Ak; s
where Ak; s = {x ∈ Bs : u(x)¿k}.

For brevity we now write
Z Z ∗ Z Z
u − k m−
J= |Du| m(x) m+
dx; Q=
dx; · = ·:
Ak; s Ak; s s − t Ak; s
From Eqs. (4.3)–(4.5) and (4.12) it follows that

Z Z
a0 J ≤ b |u|r (x) m+ dx + c m+ dx
Z
+ a1 m+ |Du|m(x)−1 m+ −1 |D|(u − k) dx
Z
+ bm+ |u|(x) m+ −1 |D|(u − k) dx
Z
+ cm+ m+ −1 |D|(u − k) dx
Z
+ a2 |Du|(x) m+ (u − k) dx
Z
+ b |u|r (x)−1 m+ (u − k) dx
Z
+ c m+ (u − k) dx: (4.13)
Now let us estimate the terms of the right-hand side of Eq. (4.13). As 0¡s − t¡1,
from Eqs. (4.10) and (4.11) we get
Z Z
|u|r (x) m+ dx ≤ |u|r+ dx + |Ak; s |
Z
≤ 2r+ (|u − k|r+ + k r+ ) dx + |Ak; s |
≤ 2r+ Q + 2r+ k r+ |Ak; s | + |Ak; s |: (4.14)

Obviously
Z
m+ dx ≤ |Ak; s |: (4.15)
Using Young inequality and taking 1 ∈ (0; 1) such that

a0
a1 (m+ − 1)1m+ =(m+ −1) =
4
we get
Z
a1 m+ |Du|m(x)−1 m+ −1 |D|(u − k) dx
Z
m(x) − 1 m(x)=(m(x)−1)
≤ a1 m+ |Du|m(x) (m+ −1)m(x)=(m(x)−1) dx
m(x) 1
Z
1 −m(x)
+ 1 |D|m(x) (u − k)m(x) dx
m(x)
Z
m+ − 1 m+ =(m+ −1)
≤ a1 m+ 1 |Du|m(x) m+ dx
m+
Z m(x)
m+ −m+ m+ u − k
+ a1 2 s−t dx
m− 1
a0
≤ J + c1 Q + c1 |Ak; s |: (4.16)
4
Using Young’s inequality and taking 2 ∈ (0; 1) such that
r+ − 1 r+ =(r+ −1) a0
a2 =
r+ 2 4
we get
Z
a2 |Du|(x) (u − k)m+ dx
Z
m+ r(x) − 1 r (x)=(r (x)−1)
≤ a2 |Du|(x)r (x)=(r (x)−1)
r(x) 2

1 −r (x)
+ 2 (u − k)r (x) dx
r(x)
Z Z
r+ − 1 r+ =(r+ −1)
≤ a2 2 |Du|m(x) m+ dx + c2 (u − k)r (x) dx
r+
a0
≤ J + c20 Q + c20 |Ak; s |: (4.17)
4
Similarly we have
Z
|u|(x) m+ −1 |D|(u − k) dx
Z
m(x) − 1 (x) · (m(x)=m(x)−1) (m+ −1) · (m(x)=m(x)−1)
≤ |u| dx
m(x)
Z
1
+ |D|m(x) (u − k) m(x) dx
m(x)
Z
≤ c3 |u| r+ dx + c4 Q + c4 |Ak; s |
≤ c5 Q + c6 (k r+ + 1)|Ak; s |: (4.18)
Z Z
u−k
m+ −1 |D|(u − k) dx ≤ 2 dx ≤ 2Q + 2|Ak; s |: (4.19)
s−t
Z Z
r(x)−1 m+
|u| (u − k) dx ≤ |u| r(x)−1 (u − k) dx
Z Z
1 r(x) − 1 r(x)
≤ (u − k) r(x) dx + |u| dx
r(x) r(x)
≤ c7 Q + c8 (k r+ + 1)|Ak; s |: (4.20)
Z Z
m+
(u − k) dx ≤ (u − k) dx ≤ Q + |Ak; s |: (4.21)
From Eqs. (4.13)–(4.21) it follows that
J ≤ c9 [Q + (k r+ + 1)|Ak; s |] (4.22)
and therefore
Z "Z ∗ #
u − k m−
|Du|m− dx ≤ c0 r+
s − t dx + (k + 1)|Ak; s | : (4.23)
Ak; t Ak; s
By Lemma 2.4, Eq. (4.23) implies that u is bounded above on BR0 (x0 ) and hence u
is locally bounded above on
.
Similarly we can prove that −u is also locally bounded above on
. So u ∈ L∞
loc (
).
If in addition max@
|u(x)| = M ¡∞, then for every x0 ∈ @
, by the similar argument
as above we can prove that Eq. (4.23) holds for k ≥ M and therefore u ∈ L∞ (
). The
Theorem 4.1 is proved.
Now we discuss the Holder continuity of weak solutions of Eq. (4.1).
Theorem 4.2. Suppose that for x ∈

; u ∈ R and z ∈ R n hold
A(x; u; z)z ≥ a0 (|u|)|z| m(x) − b(|u|); (4.24)
|A(x; u; z)| ≤ a1 (|u|)|z| m(x)−1 + b(|u|); (4.25)
|B(x; u; z)| ≤ a2 (|u|)|z| m(x) + b(|u|); (4.26)
where a0 : [0; +∞) → (0; +∞) is a nonincreasing continuous function; a1 ; a2 and

b : [0; +∞) → (0; +∞) are nondecreasing continuous functions. If u ∈ W 1; m(x) (
) is
a weak solution of Eq. (4.1) and
max |u(x)| ≤ M; (4.27)

M; ; 1 ; ) where = (a0 (M ); a1 (M ); p; q);

then u ∈ m(x) (
;

a0 (M )
1 = 1 (a0 (M ); a1 (M ); a2 (M ); b(M ); p; q); = min ;2 :
4a2 (M )M
Proof. Let B t ⊂ Bs ⊂
and let is a C ∞ -function such that
2
0 ≤ ≤ 1; supp ⊂ Bs ; ≡ 1 on Bt ; |D| ≤ : (4.28)
s−t
R R
For brevity we write a0 = a0 (M ); a1 = a1 (M ); a2 = a2 (M ); b = b(M ) and · = Ak; s ·.
Let

a0
= min ;2 (4.29)
4a2 M
and = q max{u − k; 0} where
k ≥ max u(x) − M: (4.30)

Bs
Then obviously ∈ W01; m(x) (

). By Eqs. (4.9), (4.24) – (4.26) we get
Z Z
a0 |Du| m(x) q dx ≤ a1 q |Du| m(x)−1 q−1 |Du|(u − k) dx
Z
+ bq |D| q−1 (u − k) dx
Z
+ a2 |Du| m(x) q (u − k) dx
Z Z
q
+b (u − k) dx + b q dx: (4.31)

a0
a1 (q − 1)1q=(q−1) ≤
4
we get
Z
a1 q |Du| m(x)−1 q−1 |D|(u − k) dx
Z
m(x) − 1 m(x)=(m(x)−1) (q−1)m(x)=(m(x)−1)
≤ a1 q |Du| m(x) dx
m(x) 1
Z
1 −m(x)
+ a1 q |D| m(x) (u − k) m(x) dx
m(x) 1
Z Z
a0 u − k m(x)
≤ m(x) q
|Du| dx + d dx; (4.32)
4 s−t
where d = d(a0 ; a1 ; p; q) is a positive constant.
bqp−1 2 q 2p = 1
we get
Z Z
bq |D|(u − k) q−1 dx ≤ bq |D|(u − k) dx
Z
1 m(x)
≤ bq |D| m(x) (u − k) m(x) dx
m(x) 2
Z
m(x) − 1 −m(x)=m(x)−1
+ bq dx
m(x) 2
Z
u − k m(x)
≤ dx + c1 |Ak; s |; (4.33)
s−t
where c1 = c1 (b; p; q)¿0.
From Eqs. (4.29) and (4.30) we have

Z Z
a2 |Du| m(x) q (u − k) dx ≤ a2 max u(x) − k |Du| m(x) q dx
Bs
Z Z
a0
≤ a2 M |Du| m(x) q dx ≤ |Du| m(x) q dx; (4.34)
4
Z Z
q (u − k) dx ≤ (u − k) dx
≤ (max u(x) − k)|Ak; s |

Bs
a0
≤ |Ak; s |: (4.35)
4a2
From Eqs. (4.31)–(4.35) it follows that
Z Z
u − k m(x)
|Du|m(x) q dx ≤ dx + 1 |Ak; s |

Ak; s Ak; s s − t
and therefore
Z Z
u − k m(x)
|Du|m(x) dx ≤ dx + 1 |Ak; s | (4.36)

Ak; t Ak; s s − t
for B t ⊂ Bs ⊂
and k satisfying Eq. (4.30), where = (a0 ; a1 ; p; q) and 1 (a0 ; a1 ; a2 ;
b; p; q) = 1 are positive constants. Eq. (4.36) shows that u ∈ m(x) (
; M; ; 1 ; ). By a
similar argument as above we can prove that u ∈ m(x) (
; M; ; 1 ; ). Theorem 4.2 is
proved.
From Theorems 4.2 and 2.1 we obtain
Theorem 4.3. Suppose that the assumptions of Theorem 4.2 hold. If m satises the
condition (H); then u ∈ C 0; (
) where = (a0 (M ); a1 (M ); a2 (M ); M; n; p; q) ∈ (0; 1].
From Theorems 4.1–4.3 we obtain
Theorem 4.4. Let Eqs. (4.4)–(4.8) hold and let m satisfy the condition (H). Suppose
that u ∈ W 1; m(x) (
) is a weak solution of Eq. (4.1). Then u is locally Holder continu-
ous in
. If in addition @
satises the condition (A) and u|@
is Holder continuous;
then u is Holder continuous on
.
References
[1] E. Acerbi, N. Fusco, Partial regularity under anisotropic (p; q) growth conditions, J. Dierential
Equations 107 (1994) 46 – 67.
[2] E. Acerbi, N. Fusco, A transmission problem in the calculus of variations, Calc. Var. 2 (1994) 1–16.
[3] E. De Giorgi, Sulla dierenziabilita e l’analiticita delle estremali degli integrali multipli regolari, Mem.
Accad. Sci. Torino 3 (3) (1957) 25 – 43.
[4] E. Di Benedetto, C 1+ local regularity of weak solutions of degenerate elliptic equations, Nonlinear,
Anal. TMA 7 (1983) 827– 850.
[5] E. Di Benedetto, N.S. Trudinger, Harnack inequality for quasiminima of variational integrals, Ann. Inst.
H. Poincare: Anal. Nonlineaire 1 (1984) 295 –308.
[6] T. Donaldson, Nonlinear elliptic boundary value problems in Orlicz–Sobolev space, J. Dierential
Equations 10 (1971) 507–528.
[7] L.C. Evans, A new proof of local C 1; regularity for solutions of certain degenerate elliptic p.d.e.,
J. Dierential Equations 45 (1982) 356 –373.
[8] N. Fusco, C. Sbordone, Some remarks on the regularity of minima of anisotropic integrals, Comm.
Partial Dierential Equations 18 (1993) 153 –167.
[9] M. Giaquinta, Multiple Integrals in the Calculus of Variations and Nonlinear Elliptic Systems, Princeton
University Press, Princeton, 1983.
[10] M. Giaquinta, Growth conditions and regularity, a counterexample, Manuscr. Math. 59 (1987) 245 – 248.
[11] M. Giaquinta, E. Giusti, On the regularity of the minima of variational integrals, Acta Math. 148 (1982)
31– 46.
[12] M.-C. Hong, Some remarks on the minimizers of variational integrals with nonstandard growth
conditions, Boll. Unione Mat. Ital. Ser. A 6 (7) (1992) 91–102.
[13] O.A. Ladyzhenskaya, Ural’tseva, N.N., Linear and Quasilinear Elliptic Equations, 2nd Russian ed.,
Nauka, Moscow, 1973.
[14] P. Marcellini, Regularity of minimizers of integrals of the calculus of variations with nonstandard growth
conditions, Arch. Rat. Mech. Anal. 105 (1989) 267–284.
[15] P. Marcellini, Regularity and existence of solutions of elliptic equations with p; q-growth conditions,
J. Dierential Equations 90 (1991) 1–30.
[16] C.B. Morrey, Multiple Integrals in the Calculus of Variations, Springer, Berlin, 1968.
[17] J.M. Rakotoson, R. Temam, Relative rearrangement in quasilinear variational elliptic inequalities, Indiana
Univ. Math. J. 36 (1987).
[18] K. Uhlenbeck, Regularity for a class of non-linear elliptic systems, Acta Math. 138 (1977) 219 – 240.
[19] V.V. Zhikov, Averaging of functionals of the calculus of variations and elasticity theory, Math. USSR
Izv. 29 (1987) 33 –36.
[20] V.V. Zhikov, On passing to the limit in nonlinear variational problem, Mat. Sbornik 183 (8) (1992)
47– 84.
[21] W.P. Ziemer, Boundary regularity for quasiminima, Arch. Rat. Mech. Anal. 92 (4) (1986) 371–382.

Fan e Zhao - de Giorgi e Continuidade 1997 PDF

Uploaded by

Copyright:

Available Formats

Fan e Zhao - de Giorgi e Continuidade 1997 PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fan e Zhao - de Giorgi e Continuidade 1997 PDF

Uploaded by

Copyright:

Available Formats

Nonlinear Analysis 36 (1999) 295 – 318

A class of De Giorgi type and Holder continuity1

Keywords: Sobolev spaces; Minimizers; Weak solutions; Holder continuity

In order to investigate the regularity of generalized solutions of quasi-linear elliptic

c1 |z|m(x) − c0 ≤ f(x; z) ≤ c2 |z|m(x) + c0 ; (1.1)

0362-546X/98/$19.00 ? 1998 Elsevier Science Ltd. All rights reserved.

2. Notations and main results

osc(u; E) = max u(x) − min u(x):

For x0 ∈ Rn and ¿0; B (x0 ) = {x ∈ Rn : |x − x0 |¡}. B (x0 ) is also written simply

In the following discussion m ∈L∞

p = min m(x); q = max m(x);

Note that p¿1 since m ∈L∞ + (

Lemma 2.1 (see Lemma 4.7 of Chapter II of [13]). Suppose a sequence yh ; h = 0; 1;

yh+1 ≤ cbh yh1+ ; h = 0; 1; 2; : : : (2.1)

where Ak;  = {x ∈B : u(x)¿k}; = (n)¿1 is a constant depends only on n.

osc{u; BR } ≤ c1 R ; osc{u; BR } ≤ osc{u; BbR };

Lemma 2.4 (see Lemma 2.4 of [8]). Let u ∈ W 1; p (

with ¡R0 and for any  ∈(0; 1) and any k ≥ k0 ¿0

where 0¡r ≤ p∗ ; p∗ is the Sobolev embedding exponent of p; c is a positive constant,

In this paper, our main result is Theorem 2.1. Here we give

and the functions w(x) = ±u(x) satisfy the inequalities

k ≥ max w(x) − M; (2.5)

where Ak;  = {x ∈ B : w(x)¿k}.

Remark 2.1. In [13] the class Bm (

De nition 2.2. We will say that a function m :

if there exists a constant L¿0 such that

R−osc{m; BR } ≤ L for all B R ⊂

Remark 2.2. It is easy to see that

|x − y|−|m(x)−m(y)| ≤ L1 for all x; y ∈

|m(x) − m(y)| · ln|x − y|−1 ≤ ln L1 for all x; y ∈

then m satis es condition (H).

Theorem 2.1. If m satis es condition (H); then Bm(x) (

M osc{m; BR0 } ≤ 2: (2.7)

From now on we denote by w the function identi ed to u or −u that satis es

 = max{2; 2=}; = −1 osc{u; BR } (2.9)

k 0 = max w(x) − : (2.10)

k 0 ≥ max w(x) − 12 osc{u; BR } (2.11)

k 0 ≥ max w(x) − M: (2.12)

m− = m− (BR ) = min m(x); m+ = m+ (BR ) = max m(x):

Lemma 2.5. There is a positive constant  = (n; q; L; ) such that if

|Ak 0 ; R=2 | ≤ R n ; (2.13)

then at least one of the following two inequalities holds:

0¡H ¡ ; k 0 = max w(x) − H: (2.16)

Proof. For h = 0; 1; 2; : : : ; set

From Eq. (2.17) it follows that

≤ 2(h+4)m+ H m+ Rn−m+ yh : (2.19)

If Eq. (2.13) holds and

From Eqs. (2.24), (2.20), (2.7) and (2.6) and

≤ 2h+3+n H −1 R1−n !n−1 2(h+4)q H m+ =m− Rn−(m+ =m− ) yh

yh+1 ≤ cbh yh1+[1=(n−1)] = cbh yh1+ (2.26)

and consequently, by Lemma 2.1, yh → 0 as h → ∞. This shows

|Ak 0 ; R=2 | ≤ R n (2.30)

k 0 = max w(x) − 2−s+1 : (2.31)

Proof. If Eq. (2.29) does not hold, i.e. supposing

kt = max w(x) − 2−t ; Dt = Akt ; R=2 \Akt+1 ; R=2 :

Applying inequality (2.4) to  = R,  −  = R=2 and k = kt for t = 0; 1; : : : ; s − 2, and

Applying inequality (2.2) to w with k = kt , l = kt+1 ,  = R=2 for t = 0; 1; 2; : : : ; s − 2,

|Aks−1 ; R=2 |1−(1=n) ≤ |Akt+1 ; R=2 |1−(1=n)

A class of De Giorgi type and Holder continuity1

Keywords: Sobolev spaces; Minimizers; Weak solutions; Holder continuity

For x0 ∈ Rn and ¿0; B (x0 ) = {x ∈ Rn : |x − x0 |¡}. B (x0 ) is also written simply

yh+1 ≤ cbh yh1+ ; h = 0; 1; 2; : : : (2.1)

where Ak; = {x ∈B : u(x)¿k}; = (n)¿1 is a constant depends only on n.

osc{u; BR } ≤ c1 R ; osc{u; BR } ≤ osc{u; BbR };

with ¡R0 and for any ∈(0; 1) and any k ≥ k0 ¿0

k ≥ max w(x) − M; (2.5)

where Ak; = {x ∈ B : w(x)¿k}.

Denition 2.2. We will say that a function m :

R−osc{m; BR } ≤ L for all B R ⊂

then m satises condition (H).

Theorem 2.1. If m satises condition (H); then Bm(x) (

From now on we denote by w the function identied to u or −u that satises

= max{2; 2=}; = −1 osc{u; BR } (2.9)

k 0 ≥ max w(x) − M: (2.12)

Lemma 2.5. There is a positive constant = (n; q; L; ) such that if

|Ak 0 ; R=2 | ≤ R n ; (2.13)

yh+1 ≤ cbh yh1+[1=(n−1)] = cbh yh1+ (2.26)

|Ak 0 ; R=2 | ≤ R n (2.30)

Applying inequality (2.4) to = R, − = R=2 and k = kt for t = 0; 1; : : : ; s − 2, and

Applying inequality (2.2) to w with k = kt , l = kt+1 , = R=2 for t = 0; 1; 2; : : : ; s − 2,

|Aks−1 ; R=2 | ≤ Rn

|Ak 0 ; R=2 | ≤ Rn ;

osc{u; BR=4 } ≤ (1 − −1 2−s )osc{u; BR }; (2.42)

where = max{2; 2=}; s = s(n; p; q; L; ) is the constant as in Lemma 2.7.

≤ 2s [osc{w; BR } − osc{w; BR=4 }]

= 2s [osc{u; BR } − osc{u; BR=4 }]

Denition 2.3. We will say that u ∈ m(x) (

M; ; 1 ; ). If m satises condition (H), @

Remark 2.3. It is similar to Section 7 of Chapter II of [13] that we can dene the

Z (r+ −m− )=m∗− Z m− =m∗−

Z (r+ −m− )=m∗− Z

for R0 =2 ≤ − ¡ ≤ R0 ; ∈ (0; 1) and k ≥ k0 , where = (; q) = (a; q); c = c(a; b;

|z|m(x) − c ≤ F(x; u; z) ≤ a|z|m(x) + c 1;

and consequently Eq. (3.21) holds for k ≥ maxB u(x) − M with = 1.

|A(x; u; z)| ≤ a1 |z|m(x)−1 + b|u|(x) + c; (4.4)

|B(x; u; z)| ≤ a2 |z|(x) + b|u|r (x)−1 + c; (4.5)

Denition 4.1. u ∈ W 1; m(x) (

for every ∈ W01; m(x) (

0 ≤ ≤ 1; supp ⊂ Bs ; ≡ 1 on Bt ; |D| ≤ 2=s − t:

For k ≥ 1 set = m+ max{u − k; 0}. Then ∈ W01; m(x) (

Using Young inequality and taking 1 ∈ (0; 1) such that