REGULARITY OF HlnLn~ WEAK SOLUTIONS FOR NONLINEAR ELLIPTIC SYSTEMS· He X 'lJikJrrbg (1PJ Jtg ~ ) Dept. ot Math.,
Abstract Let us consider the following elliptic systems of second order -Da(.A?(x,
and suppose i) j.A?($, tt, Du) 1 ~L(1+ I Dul); ii) (1+lpl)-1A?(x,u,p)are Holder-continuous functions with some exponent uniformly with respect to p, I.e,
lim sup .A~($+O', u+-r, 1') -.A..?(aJ, u, 1') ~K(u) < +00; (1+ 11'1) (Ia l + l-rj)~
A?(aJ, u, 1') are differentiable functiona in p with bounded and continuous derivatives \Afp.i (aJ,
Remarks. Only bounded open set LP,).,«(J)=={UELP«(J):
\7'8>0.
which is called a Morrey Space. Let assumptions i)-iT) hold, Giaquinta and Modica t1,2 ] have proved the regularity of both 1+.! and the H 1 n L oo weak solutions of (1) n
n+2
IBI ~a(!pl"+ lul~ +b, O<'Y~ under natural growth condition IBI ~ajpI2+b with a
the Hl weak solutions of (1) under controllable growth condition
which implys that the H 1 n Loo weak solutions have the same
regularity in the case of 1+~ <'Y<2. In the case of 'Y=2, many counterexamples (see [2J) showed n ')'-1 that u must be in H 1nL-, while in the case of 1+.!<'Y<2, we consider the H 1 n Ln 2=Y weak n ,,-1 solutions of (1), weaken the integrability conditions upon them (from L" to L-M) and obtain the
* Received May 16, 1988. Revised Dec. 26, 1988. This work is supported in part by the Foundatio.n of Zhongshan University, Advanced Research Center.
174
ACTA MATHEMATICA SCIENTIA
~::::~
same regularity-results. Finally we show that the exponent n
Vol 10
can not be decreased anymore
for the sake of the regularity results. Definition 1.
f
.jQ
We call uEH1 nLn 2=Y(Q, BN) be a weak solution of (1), provided that ~-t
f
-y-l
Al(x, u, Du)Da¢>'lk=- oBi-ere, u, DU)¢>'dre,
V¢E Han Ln-::Y(Q, BN),
(2)
a==l, •••, n } wheres==(!lJb ••."xn)E·Q, u=(u1, .•••, uN),Du- { DciU ' , .•. N • 1==1, •••, We use the convention that repeated indices are summed- i, j go from 1 to N and a, f3 from 1 to n.
§ ·1. Introduction Under natural growth conditions for the regularity of elliptio equations or systems of second order, authors, in general, assumed that ruE H1 nU", But 0_ A. Ladyzhenskaya and Ural'tseva[3J showed that for single equations, if we want to prove tho Holder continuity of weak solution u, it is suffioient to restrain u in
H1 n L
2
n 1'-1 2 - 1'
.1-1--<1'<2. n In this paper, we oonslder the completely non-linear elliptio systems (1) under natural growth condition v), generalize furthermore the technique in [2J, which was 'employed to treat the case B=O, and obtain two more sophisticated LP-
n
n 1'-1
estimates for the derivatives of the Hl L 2-')' weak solutions of (1). Using these est.imates and the method of perturbation we prove the Holder continuity for the 2
derivatives. Finally, with an example, we show that in the case of 1+-<1'<2 if , n we consider the H?
nY
weak solution of (1), then it follows that r>» 1'2 -1
other-
-I'
wise the local uniqueness principle .for the Dirichlet problem would be Violated. Hence the integrability condition for the weak solutton can not be decreased from 1'-1
L"~. Still more, as what concerns with the growth condition of
we have made our best effort.
Remarks.
9'b+2 1'-1 1) 7:~max( n-2' 2- . I'
n+2
) = m-2' 8 1'-1 { <2-1"
2) Suppose b(x) ELU(D), CT>n, then b(;r) EL W e use the following notations:
f
Q
U
if
&= TQT
Q
U
ax, UR='U"".R=
r +s
2 n
f
•
b(a;) and B on u,
1+.!<'Y~.£.+l:-; 3
n
1
2
n
2+;-<,,<2. (D).
2n 2' 2.= n+2' 2m n; uda; 2* = n-
BR=BB(WO)CR: stands for the ball with radius R and centre a;o, 0 represents the constant at various oocaslons and s,a,() the sufficiently small positive constants.
§ 2. Two Lp-Estimates for the Gradient Let QcRn be an ~-oube, QR=QR(XO) ={xERn:
Iwa-zoal
0;=1,
n},
!,
9 be nonnegative and gE Lfl(Q), Lemma 1. Suppose
fbr
175
WEAK SOLUTIONS FOR NONLIN.ELLIPT. SYS.
No.2
IE Lr(Q)
with tr>g>l, then we.have
( if OO~Ol( ( 11 oo)q + ( jq eJ,a;+() ( if drt JOB JQIl~ JQIB JQIB each aioEQ and each B< ~ dist(a;o, aQ) ARo, where u; 0 1 , and
() are.eenstants
with 0 1>1, R o >0,0,9<1. Then gELi1oc(Q) for pE [q, q+s), and
(t. g"eJ,;rriP ~02 {(to..
Ylq + (t./peJ,a;Y/P}
ifOO
for QJBcQ, R
,
Proposition 1. Let i), ill), Iv), v) be sablsfied w.lbh 'li'o;;;;;max
~(a;)E L 2••';-2.
(),
(n+22' 7-1) 2-1' . '1l.-
(D) and uE H 1 n~, 1+1..<1'<2, be a weak solution of (1), then n
there exist some Ro>O andp>2 s.t, uEHf~(D, RN) andforeaohaioED, R
. {t./t (1+ IDuI)'OO riP t., (1+ IDuIJl)OO, <;,,0
-
'Y-l
where O=O~n, 'A, a, Ilbll L2., -j-2. (D)), R o is a constant depending on the L"2-'Y -norm of u and L2- n or m of Du, Proof. From i), iii) and iv) it follows
..4.~(IlJ, u, p)p~";;;!;~ IplJl-k, L 1>O. Let aioED. R
~
LIDu/
JI-rI'OO<;"OB"+2L{t.
f + t. Ib.(a;) /·IU-UB/-rI'da;
IDuI·lu-UBI711.D7]1da;
+ B. IU-UB /71 ID71 Ida;} +afa. (I Du I'Y+ IU1'1") IU-UBI71 2OO and by the Young inequality, the Sobolev-Potnoare inequality and the Holder inequality, we get
2L
JBB IDul·lu-UBI77I D77ldai
<;" : 2L
1..
jDu 1 -r1'OO + 0 B2
2
(1..
It»
1+ 2
la.eJ,a;)
lr
t. lu-uBI71ID71IOO<;"B"+OB-JI (JB.1DulJl.OO )
GfB. 1.Du17·lu-UBI~daJ
1+ 2 If.
176
r (t.
ACTA MATI-IEMATICA SCIENTIA
..;;;;a(t. IDuro! th;Y/2[(JB. Iu-u.l 2·th; «o
(J
luI nl=~ dz)-n 2-'Y
B.
J IDu/ B.
2
/2 '
Vol. 10
22 Iu-uBI"~da:Y- / ' r-"l'/ll
dz = h1 ( /I u IILn ~~~
(B s
»)
f IDuI
2
. BR
dZ,
where h1(t) and h,(t) , t=2, 3, 4, afterwards are strictly increasing functions with h,(O) =0, 1,=1, 2, 3, 4. 2 3 1 . n+2 In the case of 1+<"Y '-2 +-, it IS sufficient to assnme 'r=--2-. n n n-
a
t. /u ,... ·lu-uBI7J' d
r· 1
da:
By the Sobolev embedding theorem (see [5J, Lemma 5.10), there exists constant
o independent of'U and R, s.t,
(t. lul
";;;;O/GR/G
y+Olc(t. IDuI y, 1
2
&
1
2
k:» n(n-2) ~ 2(n+2) •
&
Hence by the boundedness of Q we have OJ
J Iu \ BR
'U •
IU -
U,R 11]2
r
2 ";;;;0R" 2(t . 1U & 1
I
da: ·1
~O~+ [h.2(IIU~LI(B.»)
17+?
n:7
(t. IDul:J& y+0 (t. 1
+ha(I/DuI/LI(B.»)]
3 1 ')'-1 For -2 +-<"Y<2, 'r<-:;--,we have n ~-I' a
J'B
R
2
IDul 2dx )n=2t., IDuI 2d4l
IDuj2 dai,
t. IU, .... /U-uBI7J'&..;;;;a(t. lu/ 20",&Y/2·(fa.lu-U t! l " th;Y/2' ";;;;OR"/lI(t.
OB. IDullIth;r
lu/ .....dxr/..
~O~+h4(lluIlLn ~~: (B
s»)
1
J
Bit
IDul.2dz.
For b(z)EL2·,i 2* ({J) , we have
t.
Ib
(x) 1·1 u-u. I rio! dx..;;;;(JB.
,,;;;;OCllbh··.;-.·,o)).R"/2
«otr-vo
J IDu B.
1
r (t. r
I b 1 • do:
(t. IDul
2
2
da:
I
u- UB 12* dxrlS"
1
\2dx, 0>0 sufficiently small.
(3)
177
WEAK SOLUTIONS FOR NONLIN. ELLIPT. SYS.
No.2
By the absolute continuity theorem of Lebesgue we get
!UIL"*
~D'UIILI(B.)~O, when R~O, hence hr-+O, ~==1, 2, 3,
and
4, thus there exist someRo s.t, for all R<:Ro
t."
IDuI2dai";;;;0.R"+OR-ll(t.IDttI2.ck)
1+ 2
Or rewritten as
t.,O
1r
+1/2"+1l
1+ 2
(1+ IDuI)2 dai ";;;; 0 (fB. (1+ IDuI)ll.tko)
1r
+
By Lemma 1 there exist some p>2, s.t.
(tR"
~
t.
t.
(1+ IDttlll)tko
(1+ IDul)lltko.
(1+ IDul)flck)ll/fI<;;0ts. (1+ IDuF')da;.
0
Set ¢(R) = ( iDu- (Du)Blllda;, we have
JBR
Proposition 2. Let uEL"E} be a weak solution of (1) and assumptionS.i)v) hold. Then there exist some GliE (2, p), 8>0 and constant 0 s.t; for every zoE Q, R
(t."
IDu- (Duh/41"da;
r/
« <;.0
where h (t) is an increasing function. Proof. Define
t.
IDu- (Du)Bllldai 1
+h(luBI + I (Du)BI-1-¢(R)~)Jl8
Gr(Qj, y) = Af(y, u(y), Du(x)) -Af(Qj, u(Qj), DU(Qj)) I
J:
G(ai.
lI)={~i1[Gf(ai, lI)]ll}i,
A~t= .A~.j(lI,
U(lI), l'+t(Du(ai) -I'»dt,
1'=
(v~).
From iii) and i v) we get
.
11~f! ,L, A:fe~~~AI~12, VeER"N. Z=!.
Hence, for every cf>E Htn L" ~-'Y (D, RN) and eyery yE D we have
tlf!(DsUi-v~)D..¢ida;= Let p
p= :
t
p and
7]1
be a standard test funotion on
EOo(B p) , O~'71~1, ID'711 ~O/p, 7}1=1 on B p / 2. Inserting Du(y) (a;-a;o))7)~ and ')'==Du(y) in (4) we get
to,. I
Du(ai) -Du(lI) Illda;
.,;;;;0 {llpll
JB3
(4)
Gf(ai, lI)D..¢'da;+tB'(ai, u, Du)¢'ck.
IU(ai) -~-Dtt(lI) (ai- :To)
1
2
dai +
t:o
Gll(ai, 1I)da;+
Bep) ,
7]1
ep=eplC= (u(~)-~-
t.. IBI·I¢lldai}
178
ACTA MATHEMATICA SCIENTIA
«o {llpaJ .
lu(w)-uz-
~
Vol. 10
J GfJ(w, Y)dX+f'
(Du)2I(w-wo)ada;+
~
IBI··:I'4>~ldx
~.
+pnIDu(y)-(Du)~llI }. By the Sobolev-TotnoareJnequalfty and taking nheaverage in yon: B:a we have
tP/I dytP/I 1.Du(ai) -.Du(y) 11Ida;
~
IDu(y) - (Du),.llIdy+f tlyf' ~
G2(i, Y)dy+f
~.
~
dyf
~
;/Bl oI4hlda; }'.
(5)
.
In order to estimate the second term on the right hand side of (5), inserting ep
-=¢a = (u(ro) - u p - (Du) p(a;:--. xri)7Ji and' ')'= (D'll) p in (4) ,where TJaE 0 0 (B p) , O~7]:a'
:1., IDTJal "alp,
tp
and 'YJa=l on B~ we obtain
jDu(ai) - (Du)plllda;
,0 {.llpaJ ',I Bp
U-U p-
(Du) p(a;-mo) 12aa,
+J
Bp
(P(w,
y)dQi+J
Bp
IBI ·1
By the Bobolev-Potnoare inequality and taking the average in 'Y on B; we get (
JB 2S
IDu(ai)-(Du),llIdai
<:0 {( ( IDU(ai) - (Du)p IlIodai) J~
1+* + J~ ( dyf· ~(x, ~
Y)dai+f
~
dye J~
\BlolcPlildaf. (6)
Let's estimate the last term of (5) and (6),
of
BZ
IBllcP1Idx
Bz
IB/·I (Du(y) -
(Du)p) (W-aJo) 100
+fB3IBloIU-'l1,,- (Du)~(ai-aio) Idai}.
(7)
Estimating the first term on the right hand side of (7), we get
fB3IBloIDu(y) - (Duh! ·lai- aiolcl4; "OpIDu(y)-(Du)pl·[f
f
Bz
B:a
1.DuI"dQi~f
IDuI"dai+fBz IUI'Vdx+fBz l(bx)lox]
B~
(1+
IDul)Pd~==F(p).
Obviously F (p) ,OF (p) •
For 1+~ <1' ~ 3 2 n
+1.., 'Ti= n
n+ 2 , by the choice of s; and (3) we have
n- 2
r/l!"
fB3 luiTdai< OlP-nT/lI*GB" Iu /1I*da;
y 1
<;'OllP-
n 7i
/
2
*
[pn/lI* (t~ luillda; +(fB~ IDu12da;
rr 1
No.2
WEAK SOLUTIONS. FOB NONLIN. ELLIPT. SYS.
~O [1+p'fG1I» IDuljda;
179
r/J 1
r
l 0[( (1+ IDul)'da; ' .,.;; J1I» { o[ 1+pl»o!1I» IDul'da; ]
for
71~P
<:OF(p). 3 1
For
2+-;-<"1<2, we have
t1l» lul'fda;"';;Op-a{S1l]lIul nr=;da;y,/n "';;Op-·', where 0:1 =
2-'>'
2 <1. Hence, for 1+-<')'<2, we obtain 1 "1n
'11-
-
t1l]lIul'fda;"';;0(J~
(8)
lulJ.'fda;r/J*.o;;;O(F(p)+p-.').
For b(a;) E y*,n2*/2+s(D)
f
B3
ibIda;~( ( Ib IJ.da;1 /a..o;;;Op-<1+n/J)pn/J+8/J* = Op-"
(9)
JEAS
where (12=1-8/2.<1. Thus we have (
JB:a
IBI·IDu(y) - (Du)pI·Ia;-a;o]da;
~OIDu(y)
- (Du)pl (pF(p)
+ pl-tJl+pl-al)
<; IDu(y) - (Du)pI2+0 p2aF2(p)
where o:=min(l-tIl, 1-0::a) >0. Now let's estimate the second term on the right hand side of (7). By (8) and (9) we have
o JB2S ( (lul-'-+lb(a;)I)/u-u,,-(Du)p(a;-a:o)/da; ",0
[(fa" Iula.-'-da;ria. +(t"IbIJ·aa;r/J. ](t,,1 u-u
p-
(Du)P(ai-a:o)
1
.,.;;0 [F(p) +p-a'+p-a']p(t"
",OF
J(p)p:l4+
and O(
JB2S
",0
~(:r'
t"
Io«: (Du)pIJdr.r)~
IDu-(Du)pIJdw
IDul'Ylu-u p - (Du);;(a;-a;o)
Ida;
(fa" IDu I' da;'YI'[ pi (Du);; I + (f1l]l lu- u"I,I<'-'Y)da;y-" I'].
Note that for 2" l+'Y-p < 1..=.!.. P-'Y . 2-')"
80
we have
rIP
IJ*d,a;
180
Vol. 10
ACTA MATHEMATICA SCIENTIA
fB'1l I
(fB]j IU- tl]jI,.da;)9/91'/<'-'1') d'l:'''Op-"",
u
U-
JB 2
(J
B21
1+')'-1' )"" Iu In P=::ydo:
IDuI" dai,
_prp-2) . ( ) >0. Thus we have 1'-1) P-I'
where aa- (
o fB'1lIDul'l'ju-u
- (Du) ('l:'-(2)o) Id'l:' )(")'+1)/1
(
"Op ( h'1l (1+ IDui)'d'l:'
( (
+Opo, h'1l (1+ IDui)'d'l:'
)
~/2), ~= (P-2)/(1'-1) >0.
"Op!J"F!J(jJ) for some 0<6"min (;,
Hence we obtain (
JB2
IBI·lu-u.-(Du)p«(li-(lio)ldz
~ (:
"Op!J"F!JCjJ) +
fB]J IBII
Y" t'1l
IDu- (Du)pl!Jd(li for some 8>0,
+OIDu(y) - (Du)pI!J+
~
(10)
(:Y" t:a I
Du((li) - (Du) , l!Jdz.
(11)
Since we have
( IDU(Qi) -
JB 2 setting it
(Du)
12dai~2f IDU(Qi) -
Du(y)
R2
IJ dQi + 21Du(y) -
(Du) p 1 2 ,
in (11) and taking the average In y, we get
of
R2
dye
J R2
IBII
+ 41 JB~ ( dye IDu«(li)-Du(y) l!Jdx. JB A Analogous to (1.1) wo have (substitute p with p)
o JB(
p
IBIIc/>2/
dZ<:OpfJ8F2(p)
+
Since now for every p>l
fs; IDU(Qi) -
f
s;
(Du)pl'dQi~fB
1 /Du(ai)- (DU)p/2dai= 2
p
f
1 2
fB p
/Du-(Du)plfJdz.
(13)
JDu(Qi) -Du(y) l'dQi,
(14)
B3
dyf
Bp
(12)
dye
JB
p
(15)
IDu«(li)-Du(y) l:ldx,
by (5), set in (12), (6) and (13), finally by (14) withp=2. and (15) we obtain
t""
I
dyt"" DU«(li) -Du(y)
+ JB (
p
dye
JB
1!Jda;<;;o{(t" dyfB.o IDU«(li) -Du(y) I2-d!/' 1
p
(GfJ(z. y) +p!JaF!J(p»da;} + 2
f
Bp
y+i
B IDu(Qi) -Du(y) 1
dyf
VQioE D, p
p
2 d ai ,
(1.6)
181
WEAK SOLUTIONS FOR NONLIN. ELLIPT. SYS.
:No.2
Therefore, applying Lemma 1 in DxQ with g= IDu(a;) -Du(y) )2*, f= [a!" (G 1
+p·F)]ll., q=1+
u;/\disb (wo, ea,
~
and ()= ; , we get for some
~E (2,
p), p<1+')', VXoE D, R<
(t.,. (Zyt." IDu(x) -Du(y)/a
IDu(x) -Du(y) jllth (17)
By Proposition 1 we have F(R/2)
=1../. (1+ 1.DuI)'th~o(t. (1+ IDul)llth )'/
1
~O(l+
1
I CDU)BI +ef> (R)2)'.
It remains only to estimate the second term on the right hand side of (17). Since we have established Proposition 1, by the standard Giaqutnta technique ([1, 2]) we have
(t..
dyt.,. e"'(x, y)dx
y/'" ~h (luRI + I (Du)BI +~(B)~)B-.
Then from (14) and (15) Proposition 2 holds
§ 3. Holder Continuity of the Gradient In this section we'll use the above V-estimates and the method of per burbatton to prove the Holder continuity for the gradient of weak sol.ublon of (1). At first, let's establish the following Lemma. 1'-1
Lemma 2. Let uEL"2=Y(D, RN) be a weak solution of (1), and assumptions
i)-V) hold. Then there exist positive constants 0 and 8 s.t, for every woE D and O
t.o
IDu-
t.
(DU)p/2dx~O[(p/R)"+1l+X(B)] .
1
IDu- (Du).12th
+HoCJUBI + I CDU)BI +cPCR)2)R"+8, 1
(1.8)
where x(R) =cu( IUB f + I CDu)Bl +cPCR) 2", c/>CR)). wet, s) is inoreasing in t for fixed s, and w-+O uniformly with respect to tE [-M, M] when 8~O; HCt) is inoreasing in t, Proof. It is sufficient to prove (18) for p
Wo
E Q and set
J:
A~= A~.i(",o, 'Y-l
UB, (Du)B/4+ t (.Du(",) - (Du)B/4) )dt.
Let 'lJEHlnL"~(BB/4' BN) be the solution of the Diriohlet problem:
182
ACTA MATHEMATICA SCIENTIA
Vol. 10
Da,{Att.Dsvi) =0 on B~/4 { v-uE H&(BB/4' RN) then from the theory of linear elliptic systems (see [2J, Chap. III), we have for R/4 IDv- (Dv)pI2dx~O(p/R)"+JJ JDv- (DV)R/41 2dx.
J
~
p<
B~
/,-·1
Set w=v-u, then wEH5,1'nLn2-'Y (BB/4' RN) and
t"
IDu- (DU)pjlldx
(19)
Note that w is a weak solution of the following elliptio system
J
Aft.Ds~Da,¢'dili=JB [A~-Aff] [Ds1i -
BB"
Il"
+Jo
BBl,
.
(DsUi)R/4]Da,¢'dili
[Af(ilio, UR, DU(ili»)-At(x. U(ili), Du(ili»)]Da,¢'dili
V¢EH~nL";=~ (B R/4,
+ tSI' B,(x, U, DU)¢'dili,
RN).
Putting ep=w into the above equations we deduce
J
BB'
IDwI2dx~OJB ~IA~-AffI2IDu- (DU)R/4f 2dx
+0 +0
J
R •
IAf(xo,
UR, DU(aJ») - Af(x, u(z), Du(x)) 12dx
tSI' jB(ili, U, Du) Ilwldili. RRI'
(20)
Now let's estimate the right hand side of (20). Without loss of generality we assume pE (2, 2,,), then O<~= (2,,-p) /(p-'Y) <2(1'-1)/(2-'Y),we get
tRl< IDu I
'Y
~(J ~ (J
BBl.
Iw Idili<
IDu I' dai)" "1/'[(" J IDu I
fJ
B RI •
(tSI. IDu III da:r II(t.,. Iw /1I1
\J BBl,
B RI•
1'
dx)"I/ [ ( BR1I.
IDw IJ dx)(J
~ORn+(1'-2)/(1'-1)F(R/4) +1. J '±
B ll t
BRII.
Iw I"PIll dili )2/"J1-"I/1'
I win ;=~ ax)
8(2-'Y) n()'-l)
R
2-8 •
!=.!. '}'-1 ] 1-')",
l-Vw1 2 ax.
Similar to (8) and (9) we have
o
t.,.
(!u! ... + I b(ili) /) /1LJ!dili
~ORn(f
JBBI.
«ou-» (
Iwlll·dili)1/2* [(( 1
BRII.
JBBII.
lulll.... dili)1/2*
+(f
IDWI2dx)2' [F(RI4)+R~a,'+R-a,.]
+1.J
~ORn+2aF2(R/4)
4
BRII.
IDwI 2 da1.
fb.J2*d.~)1/2*]
BBl.'
WEAK SOLUTIONS FOR NONLIN. ELLIPT. SYS.
No.2
Thus for some 8>0 we have
o
· JB./t
.
IB(x,
U,
Du) 1·lwlttxoe;;;;O.R"+8pll(R/4) +
1 · "2J
B."
183
/DwI 2da:.
By Propositton 1, Proposition 2 and the standard Giaquinta technique (see [2], Chap. VI) we have the following estimates for the first and second terms on the right hand side of (20):
fB ~ IA~(xo,
'UB,
Du(x)) - Af(x, u, Du(x))
R"
~H1(luBI"+'1 (DU)BI
f
B R"
1
+ep(R)I)R-+.,
~IAft: -A~tl~·IDu- (DU)altI 2dai .
oe;;;;o {X(R) where H1(t),
t.,.
I~ dz
H~(t)
t.
IDu- (Du)B/2dx+.R'*eHJl(luBI
+ I (Du)BI +ep(R)i)},
are increasing functions. So from (20) we have
/DwI 2dxoe;;;;Ox(R)
t.
IDu- (Du)B/Jldai+H(luB/
Hence the results follows from
Definition 2. Let p~1,
linear space of functions
(1~).
A~O, Q(zo,
uE Y(D) s.t,
sup
X.ED
p-Af
O
+ I (Duhl +ep(R)J).R"+e.
p) =zQn Bp(xo), by y,A(Q) we denote the
IU(ai)-Ueo,pl'aai<+oo.
D(".,p)
Lemma 3. Let D be of type (A), Le, for any aioEQ, p
where
inff
2 1=={QioED: lim R-+O+ 22
={iVo E .o:
S
B.
IDu(x) -
(Du)BI 2dw> 0} ,
~ p( /uBI + I (DU)BI) =+oo}.
Proof. By Lemma 2 and the standard Giaquinta technique we have: suppose,
Qio
Et 2 1 U2 2 ,
"then
eav,
(21) Since ep(R) and IUBI + I (DU)BI are continuous functions of xo, if (21) holds for a point xoE D, then (21) holds for every x on a ball B-y(xo), I.e, DuE L2,n+8(B,,!(xo) , BrlN). Thus from Lemma 3, DuE OO,a(B"!(xo) , Rn.), so there exists an open Qos.t. Du EOO,a(Qo, RrtM) with.Q\Doc21U2~, and meas(D\Qo) =0 follows from the Lebesgue ep(p)~O(p/Ro)·,
'I'heorem,
Vp
184
ACTA MATHEMATICA SCIENTIA
Vol. 10
§ 4. Arguments on the Exponent n-')'-1
-')' 2--
Let us oonsider the fol low.lng Dirichlet problem
Llu=O!Du r { u=O
where
O=(~=~
t
1+ :
in B.R(O)
<,,<2,
n~3
on oBR(O)
(~=I +2-n), then both U= lail ~=r
-R
~-i
and U=O are
the weak solutions of the above Diriohlet problem whioh Violates the local unique-
ness prinoiple for the Diriohlet problem. Note that uE R 1 nY, 'v'p
H1
nirr», 1'-1
2-')'
so under natural growth condition 1+-<,,<2, the admissible class
n
n
')'-1
rb
for the weak solutions must at least be H1 L 2-')' to except the case of the above example. In this sense, the exponent can not be improve anymore.
Note that 1(,,/) =n~=~ (1+
:
<,,/<2) is a strictly Increasing and continuous
function with inf f (,,) = 2*, which is the oritical exponent in the embedding Jll~ Y, while limi(,,) =00; under controllable growth condition, it is only required ')'~2
that uE H1 while under quadratic growth condition uE H 1 just as a link between Hi and H1 L",
n
nL", bhus L n~ appears ~-1
References [ 1]
Giaquinta, M., G. Modica, Almost everywhere regularity results for solutions of nonlinear elliptic systems, Manu. Math., 28 (1979), 109--158. [ 2 ] Giaquinta,M.,M'llltiple Integrals in the Calculus of Variations and Nonlinea'J· Elliptic Systems,Princeton University Press, .1983. [ 3] Ladyzhenskaya, O. A., N. N. Ural'tseva, Linea'!" and Quarilinear Elliptic Equations, Nauka, Noscow, 196~; English Translation,Academic Press, New York, 1968; Second Russian Edition, Nauka, Moscow, 1973. [ 4 ] Giaquinta, M., G. Modica, Regularity results for some classes of higher order nonlinear elliptic systems, J. fur Reine una, Angew. Math., 311, 312 (1979), 145-109. [5] Adams, R. A., Sobole» Sf!