Professional Documents
Culture Documents
Sistemas Hamiltonianos Elípticos-1998
Sistemas Hamiltonianos Elípticos-1998
To cite this article: Serrin James, Zou Henghui, J. Serrin & H. Eou (1998): Existence of positive entire solutions of elliptic
hamiltonian systems, Communications in Partial Differential Equations, 23:3-4, 375-398
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss, actions,
claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
COMMUN. IN PARTIAL DIFFERENTIAL EQUATIONS, 23(3&4), 577-599 (1998)
James Serrin
School of Mathematics
University of Minnesota, Minneapolis
Henghui Zou*
Department of Mathematics
Downloaded by [Cornell University] at 00:20 23 May 2012
1 Introduction
In recent papers [8, 91 we have considered (component-wise) positive solutions
of the weakly coupled system
577
T h e n there exist infinitely many pairs (J, 7 ) E R+ x R+ such that (1.2) admits
a positive radial solution (u, v) on Rn with central values
One can ask for similar existence and uniqueness results for the system
( I . ] ) ,and even for the general Hamiltonian system (see [2, 61)
H ( u , v ) E C1 for u 2 0, v 2 0,
(1.5).
Then we have the following main result.
<
Theorem 1.1 For every suficiently small positive value there exists at least
one positive radial solution u = u ( r ) , u = v ( r ) of (1.5) with u ( 0 ) = E.
Since the system (1.5) and the conditions (1.6), (1.7) are invariant under
interchange of u and v , it is evident that Theorem 1.1 also shows that for
every sufficiently small positive value 7 there exists at least one positive radial
solution of (1.5) with v ( 0 ) = 7.
Theorem 1.1 can be extended further to state that, for any continuous
curve C in Rf x R+ with endpoints respectively on the positive [-axis and the
positive 17-axis, there is a t least one positive radial solution u , u of (1.5) with
u ( 0 ) = <, v ( 0 ) = 17 and (t,17) E C. See the remark after the proof of Theorem
1.1 in Section 3.
The question whether solutions exist whatever the value of [ (or 7 ) requires
a further condition. The basic conclusion is as follows.
'In a recent paper [I], Chen and Lu study a related (special) Hamiltonian system for
which (1.6) fails.
580 SERRIN AND ZOU
Theorem 1.2 Suppose, i n addition to the previous hypotheses (1.6) and (1.7),
that for each fixed [ > 0 we have
[sup H,(zL,v ) I7 iyf H,(u, V) (1.8)
I
whenever 7 is suitably large, where
I={O<u<[, q<v<27}
Then for every positive value .$ there exists at least one positive radial sohtzon
u = u ( r ) ,v = v ( r ) of (1.5) with u ( 0 ) = [.
Remark. Note by (1.6) that infI H,(u, u ) > 0 for all [, 7 > 0 . Thus (1.8) is
equivalent to
t s u p , Hu(u, 2') < 1 ,
7infI H,(u, v ) -
Theorem 1.2 will be the main tool for the proof of our next main existence
result. We first need the following elementary proposition.
Proposition 1.1 There exist a , @ > 0 such that
N ( u ,0 ) > Pos. Constant ua+' as u -+ oo
H ( 0 , v) 2 Pos. Constant as v -+ a.
Downloaded by [Cornell University] at 00:20 23 May 2012
vP+I
Then for every positive value [ there exists exactly one corresponding value
7 = 7 ( J )such that (1.5) has a positive radial solution with central value ([, 7 ) .
Moreover, the mapping <* q ( f ) is continuous, strictly monotone increasing
and 1-1 (onto) from R+ to R+.
In this case
Huu=fl(u), Hvu=gl(v), Hu,--0.
<
(i) For each positive value (respectively 7 ) there exists at least one positive
<
radial solution of (1.1) with u ( 0 ) = (respectively v ( 0 ) = 7 ) .
(ii) If additionally f l ( s ) 2 0 , gl(s) 2 0 for s > 0 then there exists exactly one
value 77 = 7 ( [ ) such that (1.1) has a positive radial solution with central
value ([, 7 ) . Moreover
(iii) The function q ( [ ) is continuous, strictly monotone and 1-1 from R+ to
R+.
Theorem 1.5 is an immediate corollary of Theorems 1.3 and 1..4, since for
H =F + G it is evident that (1.7) is equivalent to (1.13).
The case f ( s ) = sP, g ( s ) = sq was considered in [6, 8, 91, as noted above.
Obviously (1.13) then requires
SERRIN AND ZOU
which is exactly the condition (1.4) given earlier. Moreover, for this case (see
the remark at the beginning of the introduction) the function q(E) takes the
particular form
q(<) = ($Q+'II(P+'I,
which of course is strictly monotone and 1-1 from Rf to R
'
.
In Section 3 we consider the non-separated Hamiltonian
where X > 0, for which not only the exponents p, q but also s, t must be
suitably restricted.
A final question of interest concerns the situation when the Hamiltonian is
symmetric, i.e., H ( u , V ) = H ( v , u). One expects that in this case any positive
radial solution would have the property u(r) 2 v(r), with
Au+Hu(u,u)=O, xERn.
This is in fact the case whenever one also has Huu 2 0, Hu, _< 0, see Theorem
Downloaded by [Cornell University] at 00:20 23 May 2012
4.1.
We remark that our approach also applies when the Hamiltonian depends
on the radius 1x1. Similar conclusions can then be obtained, though with some
not entirely obvious modification of the proofs.
Theorems 1.1-1.3 will be proved in Section 3. A series of preliminary
lemmas is given in Section 2. Theorem 1.4 is proved in Section 4, while Section
5 contains the proof of a technical lemma, required in Section 2 when Hu and
: xh
H, are not Lipschitz continuous on all R '.
2 Preliminary Lemmas
With the aid of a topological argument going back to [4], we use the shooting
method and the Pohozaev-Pucci-Serrin identity to show the existence of posi-
tive radial solutions of (1.5) when (1.6), (1.7) hold. Thus one is led to consider
the ordinary differential system
n-1
+
vI1 -vl
T
+ H u ( J u l IvI)
, =0
(2.1)
n-1
d' + -ul + Hv(lul, IvI) = 0
r
ELLIPTIC HAMILTONIAN SYSTEMS
Lemma 2.1 (Local existence and uniqueness,). For each (t,7 ) E Ri x R+,
there exists a positive number p = p ( J , q ) such that (2.1)-(2.2)admits a unique
positive solution for 0 _< r 5 p.
where
and a similar one for the second integral in (2.3) (with K > 0 corresponding
to J ) , the condition F : X + X becomes exactly
again with R possibly infinite. Notice that the solution is unique on the range
[0, R], so in particular R is dependent only on the values <, q .
If R is finite, then T > R by standard continuation theorems. Moreover in
this case necessarily either u(R) = 0 or u(R) = 0, since otherwise the continu-
ation past R is certainly positive a t least for some distance, contradicting the
definition of R. Moreover, it is easy to see that R > p. Otherwise R 5 p < T,
and in turn u ( R )= 0 or v(R) = 0, contradicting the fact that the solution is
positive on [0, p ] .
The following result shows further that if a solution stays positive, then it
must go to zero as r -+ co.
whence
lim u l ( r ) = -CO.
7--tm
A = { ( E , 7 ) E R+ x R
' 1 RS,,,< 03 and ~ ( R s >, ~~ (~R )s , , ,=
) O),
and
The next two lemmas show that both A and B are non-empty and open in
R+ x R+.
Lemma 2.3 Let f , 17 > 0 be fixed such that (1.8) holds. Then ( f , 27) E A.
) a solution of (2.1) with u ( 0 ) =
Proof. Let u ( r ) ,~ ( rbe < and v ( 0 ) = 2r7.
B y (2.1)-(2.2)together with Lemma 2.2, we get
r2 .
U(T) < [- - inf H,(u, u )
2n 1 (2.5)
Downloaded by [Cornell University] at 00:20 23 May 2012
for all r such that the pair u(T), u ( r ) is in I , say 0 < r < R I . (The set I is
defined in (1.8)').
We assert that
which is again a contradiction, since 0 < u ( R I ) < u ( r o ) Thus (2.7) does not
occur. and the proof for Case 1 is complete.
Case 2. RI = cm.Then ro < Rr automatically, and
1 21~77
u ( r o ) 5 [ - - inf Hz,(u,
0) <0
2n 1 SUP, H U ( %v) -
by (1.8). On the other hand, since RI = cc we have u ( r ) > 0 for all r . This
is a contradiction, so Case 2 also cannot occur.
Lemma 2.3'. For each d > 0 there ezist positi7~ increcl,aing fi~nctions
gj = gjd : ( 0 :d ] -t R+, = I+Y : (0, d ] -t R+ such that
Downloaded by [Cornell University] at 00:20 23 May 2012
Proof. It is enough to prove the result for the set A, since the conclusion
for B can be obtained by interchanging u and v.
Define
and
< <
by (1.6). Then for 0 < E d , 0 < 7 d / 2 we have
where
rl
4(17) = 2 ( 7 m ? / 2 ) . (2.8)
This completes the proof, once we observe that D(.) is an increasing function.
Lemma 2.4 Both sets A and B are open i n R+ x R+.
Proof. It suffices to show that A is open. For any fixed (to, 170) E A, let
( u o ,uO)be the solution of (2.1)-(2.2)with initial value ( t o , 7 0 ) . Then
UO(RO)>VO(RO)=O, R o = R ~ ~ , ~ ~ < ~ ~ .
For ( E , 7 ) E R+ x R
'
, denote by (u, u ) a solution of (2.1) with initial value
( J , v), extended of course to its maximal interval.
First consider the case when H,, H, are Lipschitz continuous on R$ x R :.
Since T > Ro and vL(R0) < 0 by Lemma 2.2 it is clear that we can find c > 0
such that
+ +
uo(Ro E ) > 0 , vo(Ro C ) < 0 .
But then by continuous dependence on initial data (recall that the right hand
sides of (2.1) are Lipschitz continuous for all u , 21 when H,, H, are Lipschitz
continuous on R$ x R$)we see that
Downloaded by [Cornell University] at 00:20 23 May 2012
for all ( E , 77) sufficiently near ((0,Q), say I(<, 7 )-(to,170) 1 < 6.From (2.9)2there
+
exists in ( 0 ,Ro E ) such that ~ ( 2 =) 0. At the same time, since u l ( r ) < 0
for r > 0, we get u ( R +
-) > U ( R O E ) > 0. Hence the disk /(<,v)- (<0,17~)1 < 6
is in A (with Rt,, = R). Thus A is open, proving the lemma when H,, H , are
Lipschitz continuous on R$ x R :
.
The case when either H , or H, is not Lipschitz continuous on the closed
first quadrant R : x R$ needs more attention and is delayed until Section 5 .
The following identity will be crucial in completing the proof of Theorem
1.1.
Lemma 2.5 Let u = u ( r ) , u = u ( r ) be a solution of (2.1) such that
u(r),u(r)>O f o r O L r < R , u(R)=v(R)=O
for some R E ( 0 ,c o ) . Then
and Fl and F2 as in (2.3) - with the exception that <,7 are replaced by
) that the t integrations are carried out on the subinterval from
u ( a ) ,~ ( aand
a to r . Now 5 < R so that u ( a ) , v ( a ) > 0. But also ( ~ ( c J ) , u ( a ) ) < ( 0 ~
by)
Lemma 2.2; consequently
and a similar one for the remaining integral. The condition F : X -t X then
+
becomes P2 2 0 8 < 2p2, where p is given in (2.4). Then since a > p, we can
take
/3 = p2/2a. (2.11)
Note that if either H,,
or H , is not Lipschitz then one may lose the uniqueness
of the extension beyond R due to the lack of Lipschitz continuity at u = 0 or
v = 0.
ELLIPTIC HAMILTONIAN SYSTEMS
and
On the other hand, A n B = 0 and A and B are open. Consequently for any
fixed f E ( 0 ,b) there must exist a value 7 satisfying
u ( 0 ) = and u l ( 0 ) = ~ ' ( 0=
) 0 is an everywhere positive radial solution of
(1.5). Since (C,$ is in neither A nor B, it follows from the definition of these
sets that either
R = co,
or else
and the two functions u and u simultaneously cross the r axis at the point R
(here we write R for R , F ~ )In
. the latter case, by Lemma 2.2
and
u ( R ) = v ( R ) = 0.
But from Lemma 2.5 we then get
590 SERRIN AND ZOU
The left side is positive, while the right side is non-positive by (1.7). This
being impossible, it follows that R = co,and consequently (u, v) is a positive
radial solution of (1.5) on Rnwith central value ( f ,7).
By the same argument, for any fixed ;ii E (0, vO)there must exist a corre-
sponding value f satisfying
and for which ($,$ is in neither A nor B. The corresponding solution (u, v)
is then a positive entire solution of (1.5) with central values f, ;ii.
<
Proof of Theorem 1.2. Let > 0 be fixed and choose d = [. Then by
Lemma 2.3' the point (c,~) E B if 77 5 min{d, $(E)). On the other hand,
Downloaded by [Cornell University] at 00:20 23 May 2012
must contain at least one value 7 such that (J,7) is in neither A nor B. As
in the previous proof, the pair (t,?j)
then defines a ground state of (1.5) as
required.
Consequently
H ( 0 , v) _> v l I b ~ ( O1),
provided v 2 1. Similarly, one has
for u 2 1.
Clearly, a > 0 in view of (1.7) and the fact (see (1.6)) that H, H, > 0
when u > 0, v = 0. Similarly b > 0, and therefore necessarily a , b < 1 because
+
a b = (n - 2)/n < 1. The proof is now complete by taking
1 1
a=--1>O, p=--1>0,
a b
since H(0, I ) , H(1,O) > 0.
Remark. The particular values of a and P given here may not be the
largest ones possible. The proof simply shows (by construction) that there
exist positive values a , p satisfying (1.9).
Downloaded by [Cornell University] at 00:20 23 May 2012
Proof of Theorem 1.3. It is enough to show that, for each fixed value E > 0,
when 7 is sufficiently large, where I is given by (1.8)'. Using (1.10) and the
mean value theorem, there exists some 5 E (0, u)such that
On the other hand, for 7 sufficiently large, using (1.7) and (1.10) and the
mean value theorem again, there exists some C E (0, v) such that
bH,(u, v) 2 v-'[H(u, v) - auHu(u,w)
= v-'[H(u, v) - au(H,(u, 0) + vHu,(u, C))1
2 v-'H(u, v) - auv-'H,(~,0) + o(ua+luP).
But by (1.6) and Proposition 1.1,
U-'H(U,V)2 v - ' ~ ( 0 ,V ) 2 POS. Constant vb
for v large. Therefore
bH,(u, V) 2 POS. Const. uP - a ~ v - ~ H ~0)( u- ,o(v8)
2 Pos. Const. vP
for 0 < u < (fixed) and for v sufficiently large. It follows that
inf H,(u, v) 2 POS. Constant rlB
I
Clearly H, > 0 for u > 0 and H, > 0 for v > 0. It remains to consider (1.7)
and (1.9). For (1.10) we can clearly take
+
Figure 1. Existence region for the Hamiltonim (3.2). The point (p 1, q + 1) should lie above
the hyperbola, as shown. The point ( s , t ) should lie in the shaded region.
Downloaded by [Cornell University] at 00:20 23 May 2012
Proof of Theorem 1.4. The existence part follows immediately from Theo-
rem 1.3 since (1.11) implies (1.10). Next, if the principal conclusion were false,
c2
there would exist &, E R+ and q 1 , E~ R+ ~ such that (1.5) has two different
positive radial solutions ( u l ,vl) and ( U Z , v ~ ) ,with
and
O<E15t2, 171>V2>0,
with at least one of the inequalities 5 and > being strict
SERRIN AND ZOU
Otherwise, put
Then 0 < S < co since J1 < (2. B y direct integration of the system (2.1)-(2.2),
we have
q ( r ) = 7, - /
' dt t
/
- H.(u~,vi)sn-ids, i = 1.2
0 tn-l 0
It follows that for r <S
Downloaded by [Cornell University] at 00:20 23 May 2012
since HU(u1,wZ) - Hu(u2,v 2 ) < 0 by (1.11). Using the mean value theorem,
there exists C = ( ( s ) between vl and u2 such that
b z ( ~ -) vl(r)I+ 6 c / 6 [ ~ 2 ( -~ )V I ( s ) I + d ~ .
It is standard by the Gronwall inequality that
=
[u2(r)- v l ( r ) ] + 0, e l ( r )2 v 2 ( ) r E [0,S ] . (4.3)
Case 2. (1 = [Z and 71 > q2. The proof is essentially the same by inter-
changing the roles of u and v.
The theorem now follows a t once, since any strictly monotone, 1-1 (onto)
mapping 7 = 7(() : R+ -+ R' must be continuous.
The next result shows the symmetry of the set r, mentioned in the intro-
duction.
H,,, H,, 2 0, H,, 5 0. Then the set r is the diagonal line [ = rj in the first
quadrant of the ([,v) plane. Moreover, all positive radial solutions of (1.5)
satisfy u ( r ) = v ( r ) for r 2 0.
It follows that the set r is symmetric with respect to the diagonal line ( = 7
in the first quadrant of the ((,q) plane (interchanging u and v). Therefore
r must be contained in the diagonal line since it is monotonically increasing.
Consequently, I- must be the diagonal line itself since it is continuous.
Finally, the statement that u ( r ) = v ( r ) for r 2 0 follows from the unique-
ness Theorem 1.4, and the proof is complete.
on the data exactly a t the critical point r = R, a fact which was used crucially
in the earlier proof.)
For the proof of Lemma 2.4 in the present case, we begin as before, with
(to,qO)E .4 and with uo,u respectively denoting solutions of the initial value
problem (2.1)-(2.2)with initial values ( ( 0 , qo) and (t,q ) We assume that uo
and u are extended to their maximal intervals of existence, and let Ro = Rc,,,,,
R = Rc,,. The argument then divides into three steps.
Step 1. There exist c > 0 and 6' E ( 0 , l ) such that when I((, q ) - (lo,
q O ) /<
6' the solutions uo and u are defined on [0,Ro + c].
By Lemma 2.7, one can extend uo to the interval [0,Ro + Po) where
v) (e0,
I ( ( , - vo)j < 6' the solution u is defined and positive on 10, u ] , that is
R = R c , > 0.
We now wish to apply Lemma 2.3 to the solution u. By making 6' smaller
if necessary we can assume that p is sufficiently near po so that
p2 Pi 0
,02 2
D=- >- >- = 2c for c = po/8Ro.
20 40 4Ro
Thus if we take o to satisfy also CY 5 c (which can certainly be done), then
Hence u extends to the interval [O, Ro + c]. But c < Po, so that uo also exists
+
on [0, Ro c ] , completing the proof of Step 1. (Note also c < R o / 8 , since
Po < RO.)
Step 2. We claim that there exist eo > 0 and bo E ( 0 , I ) such that
ELLIPTIC HAMILTONIAN SYSTEMS 597
when I([, 7) - (SO,%)I < 60 and Ir - Rol L EO. We only prove (5.1)1, with
(5.2)~being obtained in the same way. For E < c and for lr - R o / 5 E we have
by integration of (2.1)
It follows that
where
where 1 2 = (0 < u < 2[0,0 < v < 2770); and where finally we have used the
fact that
for I( - (0 1 < 6". Our claim now follows with the choice do =min(cT1,6").
In what follows we assume (without loss of generality) that
Step 3. Since vo(&) = 0 it is clear from continuity that there exists a value
a' 5 €0 satisfying
1
0 .( VO(&- a') < qlub(&)l~, (5.3)
where as before we put 6 = so. Then, again by continuous dependence on
initial data, there exists > 0 such that
1
u(& - a') > -UO 2 (Ro - a'), v (& - a') < 2u0(& - a ' ) (5.4)
v(RO E)
IT-Rot9
1
< 2vo(&-or)- ilv;(&)lr<O
1
> -uO(RO- aa') - 3jub(&)I~
2
1
> --UO(&)- 31ub(Ro)k > 0
2
when ((c,17) - (to,vo)1 < min(do, dl). Therefore (as earlier) we find that A is
open; namely, if toE A, then
Acknowledgement. The authors wish to thank the referee for his careful
reading of the manuscript and in particular for suggesting a new proof for
Theorem 4.1.
ELLIPTIC HAMILTONIAN SYSTEMS
References
[I] Chen, S. and G. Lu, Existence and nun-existence of positive radial solu-
tions for a class of semilinear elliptic systems, preprint.
[2] de Figueredo, D., and C.A. Magalhaes, On nonquadratic Hamiltonian
elliptic systems, Advances Differential Eqs., 1 (1996), 881-898.
[3] Hulshof, J., E. Mitidieri and R.C.A.M. van der Vorst, Strongly indefinite
systems with critical Sobolev exponents, preprint.
141 McLeod, J.B. and J . Serrin, The existence of similar solutions of some
laminar boundary layer problems, Arch. Rational Mech. Anal., 31 (1968),
288-303.
[5] Mitidieri, E., A Rellich type identity and applications, Comm. PDE, 18
(1993), 125-151.
[6] Mitidieri, E., Non-existence of positive solutions of semilinear elliptic sys-
tems in Rn,Quaderno Matematico, 285 (1992).
[7] Pucci, P. and J. Serrin, A general variational identity, Indiana J . Math.,
35 (1986), 681-703.
Downloaded by [Cornell University] at 00:20 23 May 2012
Received: May 1 9 9 7
Revised: January 1998