New York Journal of Mathematics
New York J. Math. 5 (1999) 83{90.
Quasilinear Elliptic Systems in Divergence Form
with Weak Monotonicity
Norbert Hungerbuhler
Abstract. We consider the Dirichlet problem for the quasilinear elliptic sys-
tem
; div (x u(x) Du(x)) = f
on on @ m
for a function u : ! R , where is a bounded open domain in Rn . For
arbitrary right hand side f 2 W ;1p () we prove existence of a weak solution
under classical regularity, growth and coercivity conditions, but with only very
mild monotonicity assumptions.
u(x) = 0
0
Contents
1. Introduction
2. Galerkin approximation
3. The Young measure generated by the Galerkin approximation
4. Passage to the limit
References
1.
83
85
86
86
90
Introduction
On a bounded open domain Rn we consider the Dirichlet problem for the
quasilinear elliptic system
(1)
; div (x u(x) Du(x)) = f on (2)
u(x) = 0 on @ for a function u : ! Rm . Here, f 2 W ;1p () for some p 2 (1 1), and
satises the conditions (H0){(H2) below. A feature of this paper is that we treat
a class of problems for which the classical monotone operator methods developed
by Visik 11], Minty 10], Browder 2], Brezis 1], Lions 9] and others do not apply. The reason for this is that does not need to satisfy the strict monotonicity
condition of a typical Leray-Lions operator. The tool we use in order to prove the
0
Received May 18, 1999.
Mathematics Subject Classication. 35J65, 47H15.
Key words and phrases. Quasilinear elliptic systems, monotone operators, Young measures.
c 1999 State University of New York
83
ISSN 1076-9803/99
Norbert Hungerbuhler
84
needed compactness of approximating solutions is Young measures. The methods
are inspired by 3].
To x some notation, let M mn denote the real vector space of m n matrices
equipped with the inner product M : N = Mij Nij (with the usual summation
convention).
The following notion of monotonicity will play a r^ole in part of the exposition:
Instead of assuming the usual pointwise monotonicity condition for , we will also
use a weaker, integrated version of monotonicity which is called quasimonotonicity
(see 3]). The denition is phrased in terms of gradient Young measures. Note,
however, that although quasimonotonicity is \monotonicity in integrated form",
the gradient D of a quasiconvex function is not necessarily quasimonotone.
Denition. A function : M mn ! M mn is said to be strictly p-quasimonotone,
if
Z
(() ; ( )) : ( ; )d () > 0
Mm n
for all homogeneous W 1p -gradient Young measures with center of mass = h idi
which are not a single Dirac mass.
A simple example is the following: Assume that satises the growth condition
j(F )j 6 C jF jp;1
with p > 1 and the structure condition
Z
((F + r') ; (F )) : r' dx > c
Z
jr'jp dx
for all ' 2 C01 () and all F 2 M mn . Then is strictly p-quasimonotone. This
follows easily from the denition if one uses that for every W 1p -gradient Young
measure there exists a sequence fDvk g generating for which fjDvk jp g is equiintegrable (see 4], 6]).
Now, we state our main assumptions.
(H0) (Continuity) : Rm M mn ! M mn is a Caratheodory function, i.e.,
x 7! (x u F ) is measurable for every (u F ) 2 Rm M mn and (u F ) 7!
(x u F ) is continuous for for almost every x 2 .
(H1) (Growth and coercivity) There exist c1 > 0, c2 > 0, 1 2 Lp (), 2 2 L1 (),
3 2 L(p=) (), 0 < < p and 0 < q 6 n np;;1p such that
j (x u F )j 6 1 (x) + c1 (jujq + jF jp;1 )
(x u F ) : F > ;2 (x) ; 3 (x)juj + c2 jF jp
(H2) (Monotonicity) satises one of the following conditions:
(a) For all x 2 and all u 2 Rm , the map F 7! (x u F ) is a C 1 -function
and is monotone, i.e.,
( (x u F ) ; (x u G)) : (F ; G) > 0
for all x 2 , u 2 Rm and F G 2 M mn .
(b) There exists a function W : Rm M mn ! R such that (x u F ) =
@W (x u F ), and F 7! W (x u F ) is convex and C 1 .
@F
(c) is strictly monotone, i.e., is monotone and
( (x u F ) ; (x u G)) : (F ; G) = 0 implies F = G:
0
0
Quasilinear Elliptic Systems
85
(d) (x u F ) is strictly p-quasimonotone in F .
The condition (H0) ensures that (x u(x) U (x)) is measurable on for measurable
functions u : ! Rm and U : ! M mn . (H1) are standard growth and
coercivity conditions. The main point is that we do not require strict monotonicity
or monotonicity in the variables (u F ) in (H2) as it is usually assumed in previous
work (see, e.g., 7] or 8]). For example, take a potential W (x u F ), which is
only convex but not strictly convex in F , and consider the corresponding elliptic
problem (1){(2) with (x u F ) = @W
@F (x u F ). Even such a very simple situation
cannot be treated by conventional methods: The problem is that the gradients of
approximating solutions do not converge pointwise where W is not strictly convex.
The idea is now, that in a point where W is not strictly convex, it is locally ane,
and therefore, passage to the limit should locally still be possible. Technically, this
can indeed be achieved by a suitable blow-up process, or (and this seems to be
much more ecient) by considering the Young measure generated by the sequence
of gradients.
We prove the following result:
Theorem. If satises the conditions
(H0){(H2), then the Dirichlet problem (1),
(2) has a weak solution u 2 W01p () for every f 2 W ;1p ().
2.
Galerkin approximation
Let V1 V2 : : : W01p () be a sequence of nite dimensional subspaces with
the property that i2N Vi is dense in W01p (). We dene the operator
F : W01p () ! W ;1pZ()
;
u 7! w 7!
(x u(x) Du(x)) : Dw dx ; hf wi 0
where h i denotes the dual pairing of W ;1p () and W01p (). Observe that for
arbitrary u 2 W01p (), the functional F (u) is well dened by the growth condition
in (H1), linear, and bounded (again by the growth condition in (H1)).
By the continuity assumption (H0) and the growth condition in (H1), it is easy to
check, that the restriction of F to a nite linear subspace of W01p () is continuous.
Let us x some k and assume that Vk has dimension r and that '1 : : : 'r is a
basis of Vk . Then we dene the map
0a11 0hF (ai ' ) ' i1
i 1
2C
i 'i ) '2 iC
B
B
a
h
F
(
a
CC :
G : Rr ! Rr B
B@ ... CCA 7! BB@
..
A
.
r
i
a
hF (a 'i ) 'r i
G is continuous, since F is continuous on nite dimensional subspaces. Moreover,
for a = (a1 : : : ar )t and u = ai 'i 2 Vk , we have by the coercivity assumption in
(H1) that
G(a) a = (F (u) u) ! 1
as kakRr ! 1. Hence, there exists R > 0 such that for all a 2 @BR (0) Rr we
have G(a) a > 0 and the usual topological argument (see, e.g., 10] or 9]) gives
0
Norbert Hungerbuhler
86
that G(x) = 0 has a solution in BR (0). Hence, for all k there exists uk 2 Vk such
that
hF (uk ) vi = 0 for all v 2 Vk .
3.
The Young measure generated by the Galerkin
approximation
From the coercivity assumption in (H1) it follows that there exists R > 0 with
the property, that hF (u) ui > 1 whenever kukW01p () > R. Thus, for the sequence
of Galerkin approximations uk 2 Vk constructed above, there is a uniform bound
(3)
kuk kW01p () 6 R for all k.
Thus, we may extract a subsequence (again denoted by uk ) such that
uk * u in W01p ()
and such that
uk * u in measure and in Ls ()
for all s < p . The sequence of gradients Duk generates a Young measure x , and
since uk converges in measure to u, the sequence (uk Duk ) generates the Young
measure u(x) x (see, e.g., 5]). Moreover, for almost all x 2 , x
(i) is a probability measure,
(ii) is a homogeneous W 1p -gradient Young measure, and
(iii) satises hx idi = Du(x).
The proofs are standard (see, e.g., 3]).
4.
Passage to the limit
Let us consider the sequence
;
;
Ik := (x uk Duk ) ; (x u Du) : Duk ; Du
and prove, that its negative part I;k is equiintegrable: To do this, we write I;k in
the form
Ik = (x uk Duk ) : Duk ; (x uk Duk ) : Du
; (x u Du) : Duk + (x u Du) : Du =: IIk + IIIk + IVk + V:
The sequences II;k and V; are easily seen to be equiintegrable by the coercivity
condition in (H1). Then, to see equiintegrability of the sequence IIIk we take a
measurable
subset 0 and write
Z
j (x uk Duk ) : Dujdx 6
0
Z
Z
;
1=p ;
p
6 j (x uk Duk )j dx
jDujp dx 1=p
Z
;
;Z jDujpdx1=p :
p
qp
6 C (j1 (x) + juk j + jDuk jp )dx 1=p
0
0
0
0
0
0
0
0
0
The rst integral is uniformly bounded in k by (3). The second integral is arbitrarily
small if the measure of 0 is chosen small enough. A similar argument gives the
equiintegrability of the sequence IVk .
Quasilinear Elliptic Systems
87
Having established the equiintegrability of I;k , we may use 3, Lemma 6] which
gives that
X := lim
inf
k!1
(4)
Z
Ik >
Z Z
(x u ) : ( ; Du)dx ()dx:
M mn
On the other hand, we will now see that X 6 0. To prove this, we x " > 0. Then,
there exists k0 2 N such that dist(u Vk ) < " for all k > k0 , or equivalently, that
dist(uk ; u Vk ) < " for all k > k0 . Then, for vk 2 Vk , we may estimate X as follows
X = lim
inf
k!1
= lim
inf
k!1
6 lim
inf
k!1
Z
(x uk Duk ) : (Duk ; Du)dx
Z
;Z
R
(x uk Duk ) : D(uk ; u ; vk )dx +
j
Z
;
(x uk Duk )jp dx 1=p
0
0
1=p
Z
(x uk Duk ) : Dvk dx
jD(uk ; u ; vk )jp dx
1=p + hf v i :
k
0
The term j (x uk Duk )jp dx
is bounded uniformly in k by the growth
condition in (H1) and (3). On the other hand, by choosing
vk 2 Vk in such a way
;R
that kuk ; u ; vk kW01p () < 2" for all k > k0 , the term jD(uk ; u ; vk )jp dx 1=p
is bounded by 2". Moreover, we have
jhf vk ij 6 jhf vk ; (uk ; u)ij + jhf uk ; uij 6 2"kf kW 1p() + o(k):
Since " > 0 was arbitrary, this proves X 6 0. We conclude from (4), that
0
;
Z Z
(5)
M mn
(x u ) : dx ()dx 6
Z Z
M mn
(x u ) : Dudx ()dx:
Now, we have to prove the theorem separately in the cases (a), (b), (c) and (d)
of (H2). We start with the easiest case:
Case (d): Suppose that x is not a Dirac mass on a set x 2 M of positive
Lebesgue measure jM j > 0. Then, by the strict p-quasimonotonicity of (x u ),
we have for a.e. x 2 M
Z
(x u ) : dx () >
M mn
Z
M mn
(x u )dx () :
Z Z
M mn
(x u )dx () : Du(x)dx >
M mn
(x u ) : dx ()dx >
dx () :
| M m n {z
}
= Du(x)
Hence, by integrating over , we get
Z Z
Z
Z Z
M mn
(x u )dx () : Du(x)dx
which is a contradiction. Hence, we have x = Du(x) for almost every x 2
. From this, it follows that Duk ! Du in measure for k ! 1, and thus,
(x uk Duk ) ! (x u Du) almost everywhere. Since, by the growth condition
in (H1), (x uk Duk ) is equiintegrable, it follows that (x uk Duk ) ! (x u Du)
in L1 () by the Vitali convergence theorem. This implies that hF (u) vi = 0 for all
v 2 k2N Vk and hence F (u) = 0, which proves the theorem in this case.
Norbert Hungerbuhler
88
To prepare the proof in the remaining cases (a){(c), we proceed as follows:
From (5), we infer that
(6)
Z Z
; (x u ) ; (x u Du) : ; ; Dud ()dx 6 0:
x
n
M m
On the other hand, the integrand in (6) is nonnegative by monotonicity. It follows
that the integrand must vanish almost everywhere with respect to the product
measure dx dx. Hence, we have that for almost all x 2 (7)
( (x u ) ; (x u Du)) : ( ; Du) = 0 on spt x
and thus
(8)
spt x f j ( (x u ) ; (x u Du)) : ( ; Du) = 0g:
Now, we proceed with the proof in the single cases.
Case (c): By strict monotonicity, it follows from (7) that x = Du(x) for almost
all x 2 , and hence Duk ! Du in measure. The reminder of the proof in this case
is exactly as in case (d).
Case (b): We start by showing that for almost all x 2 , the support of
x is contained in the set where W agrees with the supporting hyper-plane L :=
f( W (x u Du) + (x u Du)( ; Du))g in Du(x), i.e., we want to show that
spt x Kx = f 2 M mn : W (x u ) = W (x u Du) + (x u Du) : ( ; Du)g:
If 2 spt x then by (8)
(9)
(1 ; t)( (x u Du) ; (x u )) : (Du ; ) = 0 for all t 2 0 1].
On the other hand, by monotonicity, we have for t 2 0 1] that
(10)
0 6 (1 ; t)( (x u Du + t( ; Du)) ; (x u )) : (Du ; ):
Subtracting (9) from (10), we get
(11)
0 6 (1 ; t)( (x u Du + t( ; Du)) ; (x u Du)) : (Du ; )
for all t 2 0 1]. But by monotonicity, in (11) also the reverse inequality holds and
we may conclude, that
(12)
( (x u Du + t( ; Du)) ; (x u Du)) : ( ; Du) = 0
for all t 2 0 1], whenever 2 spt x . Now, it follows from (12) that
Z1
W (x u ) = W (x u Du) +
(x u Du + t( ; Du)) : ( ; Du)dt
0
= W (x u Du) + (x u Du) : ( ; Du)
as claimed.
By the convexity of W we have W (x u ) W (x u Du)+ (x u Du) : ( ; Du)
for all 2 M mn and thus L is a supporting hyper-plane for all 2 Kx. Since the
mapping 7! W (x u ) is by assumption continuously dierentiable we obtain
(13)
(x u ) = (x u Du) for all 2 Kx spt x
and thus
Z
(14)
:=
(x u ) dx () = (x u Du) :
M mn
Quasilinear Elliptic Systems
89
Now consider the Caratheodory function
g(x u p) = j (x u p) ; (x)j :
The sequence gk (x) = g(x uk (x) Duk (x)) is equiintegrable and thus
gk * g weakly in L1 ()
and the weak limit g is given by
g(x) =
=
Z
Z
RmM mn
spt x
j (x ) ; (x)j du(x) () dx ()
j (x u(x) ) ; (x)j dx () = 0
by (13) and (14). Since gk > 0 it follows that
gk ! 0 strongly in L1 ().
This again suces to pass to the limit in the equation and the proof of the case (b)
is nished.
Case (a): We claim that in this case for almost all x 2 the following identity
holds for all 2 M mn on the support of x :
(x u ) : = (x u Du) : + (r (x u Du)) : (Du ; )
(15)
where r is the derivative with respect to the third variable of . Indeed, by the
monotonicity of we have for all t 2 R
( (x u ) ; (x u Du + t)) : ( ; Du ; t) 0
whence, by (7),
; (x u ) : (t) > ; (x u Du) : ( ; Du) + (x u Du + t) : ( ; Du ; t)
;
= t (r (x u Du))( ; Du) ; (x u Du) : + o(t):
The claim follows from this inequality since the sign of t is arbitrary. Since the
sequence (x uk Duk ) is equiintegrable, its weak L1-limit is given by
=
=
Z
Z
spt x
spt x
(x u )dx ()
(x u Du)dx () + (r (x u Du))t
= (x u Du)
Z
spt x
(Du ; )dx ()
where we used (15) in this calculation. This nishes the proof of the case (a) and
hence of the theorem.
Remark. Notice, that in case (b) we have (x uk Duk ) ! (x u Du) in L1().
In the cases (c) and (d), we even have Duk ! Du in measure as k ! 1.
Norbert Hungerbuhler
90
References
1] H. Brezis, Operateurs Maximaux Monotones et Semi-groupes de Contractions dans les Espaces de Hilbert , North-Holland Mathematics Studies, No. 5. Notas de Matematica (50).
North-Holland Publishing Co., Amsterdam-London American Elsevier Publishing Co., Inc.,
New York, 1973, MR 50 #1060 Zbl 252.47055.
2] F. E. Browder, Existence theorems for nonlinear partial dierential equations , Global Analysis (Berkeley, 1968), Proc. Sympos. Pure Math., no. XVI, AMS, Providence, 1970, pp. 1{60,
MR 42 #4855.
3] G. Dolzmann, N. Hungerbuhler, S. Muller, Non-linear elliptic systems with measure-valued
right hand side , Math. Z. 226 (1997), 545{574, Zbl 895.35029.
4] I. Fonseca, S. Muller, P. Pedregal, Analysis of concentration and oscillation eects generated
by gradients , SIAM J. Math. Anal. 29 (1998), 736{756, MR 99e:49013 Zbl 980.20907.
5] N. Hungerbuhler, A renement of Ball's Theorem on Young measures, New York J. Math.
3 (1997), 48{53, Zbl 887.46010.
6] D. Kinderlehrer, P. Pedregal, Gradient Young measures generated by sequences in Sobolev
spaces, J. Geom. Anal. 4 (1994), 59{90, MR 95f:49059, Zbl 808.46046.
7] R. Landes, On Galerkin's method in the existence theory of quasilinear elliptic equations, J.
Funct. Anal. 39 (1980), 123{148, MR 83j:35064, Zbl 452.35037.
8] R. Landes, V. Mustonen, On pseudomonotone operators and nonlinear noncoercive variational problems on unbounded domains, Math. Ann. 248 (1980), 241{246, MR 81i:35057,
Zbl 424.35086.
9] J.-L. Lions, Quelques Methodes de Resolution des Problemes aux Limites non Lineaires,
Dunod Gauthier-Villars, Paris 1969, MR 41 #4326, Zbl 189.40603.
10] G. J. Minty, Monotone (nonlinear) operators in Hilbert space, Duke Math. J. 29 (1962),
341{346, MR 29 #6319, Zbl 111.31202.
11] M. I. Visik, Quasi-linear strongly elliptic systems of dierential equations of divergence form,
(Russian) Trudy Moskov. Mat. Obsc. 12 (1963), 125{184, MR 27 #6017.
Max-Planck Institute for Mathematics in the Sciences, Inselstrasse 22{26, 04103
Leipzig (Germany)
buhler@mis.mpg.de http://personal-homepages.mis.mpg.de/buhler
This paper is available via http://nyjm.albany.edu:8000/j/1999/5-5.html.