
380 
R. 
H. 
Martin 
and 
M. 
Pierre 
where 
p, 
q 
2 
1 
and 
c 
: 
R 
--* 
R 
is 
a 
regular  function 
with  variable  sign  on 
R. 
Again  uniform L’-estimate  and equivalence of  LP-norms hold, but global 
existence does not follow.  Partial “local” estimates have been  obtained 
for 
(6.4) 
in 
[ll] 
where this kind of systems are introduced. Note that numerical 
experiments show that if dl and 
d2 
are very different 
(d2 
ss 
c(z)upvq 
can become very  large 
at 
some time.  However, in  the situations considered, 
the solutions seem 
to 
become  “smoother”  after 
a 
while. 
There are actually  elliptic  versions of  the same questions  which  could 
perhaps be looked 
at 
first to provide some hints and which are 
also 
of interest 
by  themselves.  One such example is 
where  (again) 
0 
< 
X 
2 
1, 
p, 
q, 
r, 
s 
> 
1, 
F, 
G 
are nonnegative  and smooth 
given  functions on 
R, 
p 
> 
0 
and where we  denote 
z 
= 
(11,. 
. . , 
zN). 
Again, 
we  easily prove  (as in  Proposition 
4.1) 
that 
If 
p 
is 
small enough,  this allows 
us 
to 
conclude  existence of 
a 
solution  in 
(6.5). 
For 
large 
p, 
the question of  existence seems 
to 
be open in general. 
1.6.2 
“Critical” Boundary Conditions 
Although general  boundary conditions 
of 
type 
(4.5) 
can  be handled  by 
our 
basic method, some “critical” cases are not (see the extra hypothesis 
(4.6)). 
For 
instance, in  the case of the Brusselator 
(5.1), 
it 
is 
not known if solutions 
subject to the boundary  conditions 
i3V 
an 
u= 
B/A 
and 
- 
=O 
ondRx(0,m)