HAL Id: hal-00315641
https://hal.archives-ouvertes.fr/hal-00315641
Submitted on 29 Aug 2008
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
SEMI-LINEAR SUB-ELLIPTIC EQUATIONS ON THE HEISENBERG GROUP WITH A SINGULAR
POTENTIAL
Houda Mokrani
To cite this version:
Houda Mokrani. SEMI-LINEAR SUB-ELLIPTIC EQUATIONS ON THE HEISENBERG GROUP
WITH A SINGULAR POTENTIAL. Communications on Pure and Applied Mathematics, Wiley,
2009, 8 (5), pp.1619 - 1636. �10.3934/cpaa.2009.8.1�. �hal-00315641�
GROUP WITH A SINGULAR POTENTIAL
HOUDA MOKRANI
HOUDA.MOKRANI@ETU.UNIV-ROUEN.FR
UNIVERSIT ´E DE ROUEN, UMR 6085-CNRS, MATH ´EMATIQUES AVENUE DE L’UNIVERSIT ´E, BP.12
76801 SAINT ETIENNE DU ROUVRAY, FRANCE
Abstract. In this work, we study the Dirichlet problem for a class of semi-linear sub- elliptic equations on the Heisenberg group with a singular potential. The singularity is controlled by Hardy’s inequality, and the nonlinearity is controlled by Sobolev’s in- equality. We prove the existence of a nontrivial solution for a homogenous Dirichlet problem.
Key words: Heisenberg group, Hardy’s inequality, Sobolev’s inequality, Singular potential A.M.S. Classification
1. Introduction
In this work, we study the partial differential equations on the Heisenberg group H
d. Let us recall that the Heisenberg group is the space R
2d+1of the (non commutative) law of product
(x, y, s) · (x
0, y
0, s
0) = x + x
0, y + y
0, s + s
0+ 2 (y|x
0) − (y
0|x) . The left invariant vector fields are
X
j= ∂
xj+ 2y
j∂
s, Y
j= ∂
yj− 2x
j∂
s, j = 1, · · · , d and S = ∂
s= 1
4 [Y
j, X
j].
In all that follows, we shall denote by Z
j= X
jand Z
j+d= Y
jfor j ∈ {1, · · · , d}. We fix here some notations :
z = (x, y) ∈ R
2d, w = (z, s) ∈ H
d, ρ(z, s) = |z|
4+ |s|
21/4where ρ is the Heisenberg distance. Moreover, the Laplacian-Kohn operator on H
dand Heisenberg gradient is given by
∆
Hd=
n
X
j=1
X
j2+ Y
j2; ∇
Hd= (Z
1, · · · , Z
2d).
Let Ω be an open and bounded domain of H
d, we define thus the associated Sobolev space as following
H
1(Ω, H
d) = n
f ∈ L
2(Ω) ; ∇
Hdf ∈ L
2(Ω) o and H
01(Ω, H
d) is the closure of C
0∞(Ω) in H
1(Ω, H
d).
1
We consider the following semi-linear Dirichlet problem (1.1)
( −∆
Hdu − µV u = λu+ | u |
p−2u in Ω, u
∂Ω
= 0
where V is a positive potential function which admits the singularity on Ω, λ is a real constant and 2 < p < 2+
2d; the index 2
∗= 2+
2dis the critical index of Sobolev’s inequality on the Heisenberg group [9, 15, 16, 17, 22]
(1.2) kuk
L2∗(Ω)
≤ C
Ωkuk
H1(Ω,Hd), for all u ∈ H
01(Ω, H
d).
The potential function V is controlled by the following Hardy’s inequality, (1.3)
Z
Ω
V (w)|u(w)|
2dw ≤ k∇
Hduk
2L2(Ω), for all u ∈ H
01(Ω, H
d).
We will prove in the next section the following results
• If 0 ∈ Ω, Hardy’s inequality (1.3) holds for V (z, s) = d
4(d + 1)
2ρ(z, s)
−2.
• If 0 ∈ Ω, Hardy’s inequality (1.3) also holds for a softer potential V (z, s) = d
2|z|
2ρ(z, s)
4, see also [12, 18].
• In the Lemma 2.7, we prove that Hardy’s inequality (1.3) holds for V (z, s) = ¯ µρ
c(z, s)
−2,
where ρ
cdefined in (2.8) is the distance to a sub-manifold Σ
cof codimension ≥ 2 and ¯ µ is a constant.
Theorem 1.1. Assume that the potential function V satisfies (1.3), then for any λ > 0 and any 0 ≤ µ < 1, the Dirichlet problem (1.1) admits a nontrivial solution in H
01(Ω, H
d).
The Dirichlet problem (1.1) on the Heisenberg group is a natural generalization of the classical problem on R
d, see [5, 7, 8, 10, 11, 14, 20] and their references. The subellipticity of the operator ∆
Hdimplies that any weak solution of the Dirichlet problem (1.1) belongs to C
∞(Ω \ Σ
c) (see [25]). The issue of regularity of a weak solution near ∂Ω ∪ Σ
cis a very delicate problem.
This paper is organized as follows. In section 2, we recall Poincar´ e’s inequality and prove Hardy’s inequality on the Heisenberg group; Section 3 deals with the study of the eigenvalue problem; finally, we prove the existence of a weak solution in Sections 4 and 5 by using Rabinowitz’s Theorem and the Palais-Smale Theorem.
2. Hardy’s inequalities on H
dThe following density theorem is very useful. Let us give here an other simple proof than [2] (see also [3]),
Theorem 2.1. We have that C
0∞(Ω \ {(0, 0)}) is dense in H
01(Ω, H
d) for d ≥ 1.
Proof : By definition of H
01(Ω, H
d), it suffices to show that C
0∞(Ω) ⊂ C
0∞(Ω \ {(0, 0)}, H
d)
k.kH1. Let ϕ be a cut-off function for which
(2.1) ϕ(η) =
0 if 0 < η ≤ 1, 1 if η ≥ 2.
For u ∈ C
0∞(Ω, H
d), let ε > 0 small enough, and we set u
ε(z, s) = ϕ(
1ερ(z, s)) u(z, s). So u
ε∈ C
0∞(Ω \ {(0, 0)}, H
d) and we have
k u
ε− u k
2H1(Ω)=k ∇
Hd
(u
ε− u) k
2L2(Ω)+ k u
ε− u k
2L2(Ω). Dominated convergence theorem implies that
k u
ε− u k
2L2(Ω)→ 0, and
Z
Ω
| ϕ( 1
ε ρ(z, s)) − 1 |
2| ∇
Hdu(z, s) |
2dzds → 0, when ε → 0.
On the other hand, we have Z
Ω
| ∇
Hd( 1
ε ρ(z, s)) |
2| ϕ
0( 1
ε ρ(z, s)) |
2| u(z, s) |
2dzds
= 1 ε
2Z
Ω
| z |
2ρ(z, s)
2| ϕ
0( 1
ε ρ(z, s)) |
2| u(z, s) |
2dzds
≤ 1
ε
2k u k
2L∞(Ω)k ϕ
0k
2L∞(Ω)Z
{(z,s);ε≤ρ(z,s)≤2ε}
dzds
≤ C 1
ε
2ε
2d+2→ 0, as ε → 0.
Remark that this proof also show that the density theorem is not true for classical Sobolev space in a 2 dimensional case.
Now, we state the following precise Poincar´ e inequality,
Theorem 2.2. Let Ω be a sub-domain of H
dbounded in some direction of (z
1, · · · , z
2d), that is, there exist R > 0 and 1 ≤ j
0≤ 2d such that 0 < r =| z
j0|≤ R for all (z, s) ∈ Ω.
Then for any u ∈ H
01(Ω, H
d), Z
Ω
| u |
2dzds ≤ 4R
2Z
Ω
| ∇
Hdu |
2dzds.
(2.2)
Remark : By using the inequality (2.2), we can use k∇
Hd
uk
L2(Ω)as a norm on H
01(Ω, H
d).
The Poincar´ e inequality (2.2) holds for any Ω ⊂ {(z, s) ∈ H
d; ρ(z, s) ≤ R}. We can also obtain the Poincar´ e inequality from Bony’s maximum principle for general H¨ ormander’s vector fields but with a non-precise constant (see [6, 21]). The proof given here is a modification of L. D’Ambrosio [13].
Proof : Using the density results of Theorem 2.1, take u ∈ C
0∞(Ω \ {(0, 0)}) and let T (z, s) = (T
1(z, s), · · · , T
2d(z, s)) be a C
1vector function on Ω. Denote by
div
HdT =
2d
X
j=1
Z
jT
j,
we have Z
Ω
(div
HdT ) | u |
2dzds = −2 Z
Ω
hT, ∇
Hduiu dzds
≤ 2 Z
Ω
| hT, ∇
Hd
uiu | dzds
≤ 2[
Z
Ω
| ∇
Hdu |
2dzds]
12[ Z
Ω
| T |
2| u |
2dzds]
12≤ Z
Ω
| ∇
Hdu |
2dzds + Z
Ω
| T |
2| u |
2dzds.
Thus
Z
Ω
div
HdT − | T |
2| u |
2dzds ≤ Z
Ω
| ∇
Hdu |
2dzds.
For ε > 0, let us choose T := T
ε= −
12∇Hrdrεε
where r
ε= r
2+ ε
212and r = |z
j0|. Then div
HdT
ε= − 1
2 1
r
2ε[r
ε∆
Hdr
ε− | ∇
Hdr
ε|
2] div
HdT
ε− | T
ε|
2= − 1
2
∆
Hdr
εr
ε+ 1 4
| ∇
Hdr
ε|
2r
2εsince
∇
Hdr
ε= ∇
Hdr
22r
ε= r
r
ε, | ∇
Hdr
ε|
2= r
2r
ε2∆
Hdr
ε= ∆
Hdr
22r
ε− ∇
Hd
r
22r
2ε∇
Hd
r
ε= 1 r
ε− r
2r
3ε. So,
div
HdT
ε− | T
ε|
2= − 1
2r
2ε+ 3r
24r
ε4, and for any (z, s) ∈ Ω,
ε→0
lim div
HdT
ε− | T
ε|
2= 1 4
1 r
2≥ 1
4 1 R
2. From the dominated convergence theorem, we have
ε→0
lim Z
Ω
[div
HdT
ε− | T
ε|
2] | u |
2dzds ≥ 1 4
1 R
2Z
Ω
| u |
2dzds,
thus
1 4
1 R
2Z
Ω
| u |
2dzds ≤ Z
Ω
| ∇
Hdu |
2dzds.
Let us give a very easy proof of the classical Hardy inequality by using a radial vector field. Let U be a bounded domain of R
d, d > 2, H
01(U ) is the usual Sobolev space.
Lemma 2.3. We have, for any u ∈ H
01(U ), (2.3)
d 2 − 1
2Z
U
u
2|x|
2dx ≤ k∇
xuk
2L2(U).
Proof : As C
0∞(U \ {0}) is dense H
01(U ), we have restricted ourselves to a function u in C
0∞(U \ {0}). The proof mainly consists of an integration by parts with respect to the radial vector field R,
R =
d
X
j=1
x
j∂
xj.
We notice that R(| x |
−2) = −2 | x |
−2and divR = d, so Z
U
u
2| x |
2dx = − 1 2
Z
U
R(| x |
−2) dx
= 1 2
Z
U
div(R) u
2| x |
2dx + 1 2 Z
Ω
1
| x |
2R(u
2) dx
1 − d 2
Z
U
u
2| x |
2dx = 1 2
Z
U
1
| x |
2R(u
2) dx = Z
U d
X
j=1
u
| x | x
j| x | ∂
xju dx.
By Cauchy-Schwarz, d
2 − 1 Z
U
| u |
2| x |
2dx ≤ Z
U d
X
j=1
| ∂
xju |
2dx
12Z
U
| u |
2| x |
2dx
12.
On the Heisenberg group, if we introduce the radial vector field R =
d
X
j=1
(x
j∂
xj+ y
j∂
yj) =
d
X
j=1
(x
jX
j+ y
jY
j), then we immediately obtain for d > 1,
(d − 1)
2Z
Ω
u
2ρ(z, s)
2dzds ≤ (d − 1)
2Z
Ω
u
2|z|
2dzds ≤ k∇
Hduk
2L2(Ω)for any u ∈ H
01(Ω, H
d).
By using the idea inspired from the radial vector field, we now prove the following Hardy inequality.
Lemma 2.4. For d ≥ 1, we have that (2.4)
d
2d + 1
2Z
Ω
u
2ρ
2dzds ≤ k∇
Hduk
2L2(Ω),
for any u ∈ H
01(Ω, H
d).
Proof : By using the density theorem, we prove the inequality (2.4) for the function u ∈ C
0∞(Ω \ {(0, 0)}). Then the proof mainly consists of an integration by parts with respect to the radial vector field R
Hdadapted to the structure of H
d, namely
R
Hd= 2s∂
s+
d
X
j=1
(x
j∂
xj+ y
j∂
yj) = s 2d
d
X
j=1
[Y
j, X
j] +
d
X
j=1
(x
jX
j+ y
jY
j).
We notice that R
Hdρ
−2= −2ρ
−2and div R
Hd= 2d + 2. We have Z
Ω
u
2ρ(z, s)
2dzds = − 1 2 Z
Ω
R
Hdρ(z, s)
−2u
2dzds
= 1 2 Z
Ω
ρ
−2R
Hdu
2dzds + 1 2 Z
Ω
ρ
−2u
2div R
Hddzds.
This gives
−d Z
Ω
u
2ρ
2dzds = Z
Ω d
X
j=1
u ρ
x
jρ X
j+ y
jρ Y
ju dzds − 1 2d
Z
Ω
Y
js ρ
2u(X
ju) dzds
+ 1 2d
Z
Ω
X
js ρ
2u(Y
ju) dzds
=
1 + 1 d
Z
Ω d
X
j=1
x
ju
ρ
2X
ju + y
ju ρ
2Y
ju
dzds
+ 1 d Z
Ω d
X
j=1
s ρ
6| z |
2y
j− sx
juX
ju dzds
− 1 d Z
Ω d
X
j=1
s ρ
6| z |
2x
j+ sy
juY
ju dzds,
then
−d
2Z
Ω
u
2ρ
2dzds = Z
Ω d
X
j=1
(d + 1) − s
2ρ
4x
ju
ρ
2X
ju + y
ju ρ
2Y
ju
dzds
+ Z
Ω d
X
j=1
s | z |
2ρ
4y
ju
ρ
2X
ju − x
ju ρ
2Y
ju
dzds
= Z
Ω d
X
j=1
h
((d + 1) − s
2ρ
4) x
jρ + s | z |
2ρ
4y
jρ
i u
ρ X
ju dzds
+ Z
Ω d
X
j=1
h
((d + 1) − s
2ρ
4) y
jρ − s | z |
2ρ
4x
jρ
i u
ρ Y
ju dzds.
Setting A(z, s) =
d
X
j=1
n h
( (d + 1) − s
2ρ
4) x
jρ + s | z |
2ρ
4y
jρ i
2+ h
( (d + 1) − s
2ρ
4) y
jρ − s | z |
2ρ
4x
jρ i
2o
,
Cauchy-Schwarz inequality implies d
2Z
Ω
| u |
2ρ
2dzds ≤ Z
Ω
A(z, s) | u |
2ρ
2dzds
12Z
Ω d
X
j=1
| X
ju |
2+ | Y
ju |
2dzds
12.
For A(z, s), we have
A(z, s) = (d + 1) − s
2ρ
4 2| z |
2ρ
2+ s
2| z |
4ρ
8| z |
2ρ
2= | z |
2ρ
2h
(d + 1)
2− (2d + 1) s
2ρ
4i
= | z |
2ρ
6h
(d + 1)
2| z |
4+ d
2s
2i
= | z |
2ρ
6h
(2d + 1) | z |
4+ d
2(| z |
4+s
2) i
≤ (d + 1)
2| z |
2ρ
2≤ (d + 1)
2. So, we deduce the inequality (2.4).
The Hardy inequality on the Heisenberg group H
dis first proven in [18, 12] for a softer potential.
Lemma 2.5. We have, for any u ∈ H
01(Ω, H
d), d
2Z
Ω
| z |
2| z |
4+s
2| u |
2dzds ≤ k∇
Hduk
2L2(Ω). (2.5)
The singularity of potential in the Hardy inequalities (2.3), (2.4) and (2.5) is a isolate point of domain. We consider now the general case when the singularity is on a sub- manifold. We have first the following density result:
Lemma 2.6. Let Ω be a bounded domain of R
2d+1and Σ
ca sub-manifold of Ω such that dim Σ
c≤ 2d − 1. Then C
0∞(Ω \ Σ
c) is dense in the space H
01(Ω, H
d).
Proof : As H
01(Ω, H
d) is a Hilbert space, it is enough to prove that the orthogonal of C
0∞(Ω \ Σ
c) in H
01(Ω, H
d) is {0}. Let u be in this space. For any v in C
0∞(Ω \ Σ
c), we have
(u, v)
L2+ (∇
Hdu, ∇
Hdv)
L2= 0.
By integration by part,
∀v ∈ C
0∞(Ω \ Σ
c) , hu − ∆
Hdu, vi = 0, this implies that, as a distribution,
Supp u − ∆
Hdu
⊂ Σ
c.
Since u − ∆
Hdu belong to the classical Sobolev space H
−1(Ω) and except 0, no distribution of H
−1(Ω) can be supported in a submanifold of dimension ≤ (2d+1)−2. Thus u−∆
Hd
u = 0 on Ω. Taking the L
2scalar product with u ∈ H
01(Ω, H
d) implies that u ≡ 0. This completes the proof of Lemma 2.6.
We consider now the hyper-surface Σ = {(x, y, s) ∈ Ω : g(x, y, s) = s + f(x, y) = 0}
where Ω is a neighborhood of 0 in H
d. Assume that
(2.6) Σ
c= {w ∈ Ω : g(w) = 0, ∇
Hdg(w) = 0} ,
is a sub-manifold of dimension (2d + 1) − r − 1, r ≥ 1.
Lemma 2.7. Assume that Σ
cis a sub-manifold of dimension 2d − r and r ≥ 1. Then, there exists µ > ¯ 0 such that for any u ∈ H
01(Ω, H
d),
(2.7) µ ¯
Z
Ω
u
2ρ
2cdw ≤ k∇
Hduk
2L2(Ω)with
(2.8) ρ
c(w) = g
2(w) + |∇
Hdg(w)|
41/4.
We refer to the proof of this lemma to [4] and also [2]. The constant ¯ µ depends, of course on Σ
c, but in many interesting cases, it depends only on the dimension of Σ
c.
Here we present a proof for a model case in H
1to precise the constant ¯ µ. We take g(x, y, s) = s + 2xy, then
Σ = {(x, y, s) ∈ H
1: s + 2xy = 0}, Σ
c= {(x, 0, 0), x ∈ R }.
(2.9)
Lemma 2.8. Let Σ
cas in (2.9), then, we have for any u ∈ H
1(H
1),
(2.10) 2
25 + 2
8Z
H1
u
2ρ
2cdw ≤ k∇
H1uk
2L2(H1)Proof : We rectify Σ by setting x
0= x, y
0= y, s
0= s + 2xy, so the vector fields X and Y change to X
0= ∂
x0+ 4y
0∂
s0, Y
0= ∂
y0and
ρ
c(x
0, y
0, s
0) = (4y
0)
4+ s
021/4. Let R be a radial vector field
R = X
0(s
0)Y
0+ 2
3s
0∂
s0= 4y
0Y
0+ 2
3s
0∂
s0= 4y
0Y
0+ 2s
0[Y
0, X
0],
where R(ρ
−2c) = −8ρ
−2cand divR = 12. Using the density Lemma 2.6, we have for u ∈ C
0∞( H
1\ {(0, 0)}),
Z
H1
u
2ρ
2cdz
0ds
0= − 1 8
Z
H1
u
2R(ρ
−2c)dz
0ds
0− 1 2
Z
H1
u
2ρ
2cdz
0ds
0= Z
H1
y
0ρ
2cu Y
0u dz
0ds
0− 1 2
Z
H1
Y
0( s
0ρ
2c)u X
0u dz
0ds
0+ 1
2 Z
H1
X
0( s
0ρ
2c)u Y
0u dz
0ds
0= Z
H1
y
0ρ
2cu Y
0u dz
0ds
0+ Z
H1
2
8y
03s
0ρ
6cu X
0u dz
0ds
0+ Z
H1
h 2y
0ρ
2c− 2y
0s
02ρ
6ci
u Y
0u dz
0ds
0= Z
H1
h 3y
0ρ
c− 2y
0s
02ρ
5ci u ρ
cY
0u dz
0ds
0+ Z
H1
2
8y
03s
0ρ
5cu ρ
cX
0u dz
0ds
0We then obtain 1 2
Z
H1
| u |
2ρ
2cdz
0ds
0≤ Z
H1
A(z
0, s
0) | u |
2ρ
2dz
0ds
0 12k∇
H1uk
L2(H1),
with
A(z
0, s
0) = h 3y
0ρ
c− 2 y
0s
02ρ
5ci
2+ 2
16y
06s
02ρ
10c= 3
2y
02ρ
2c− 12 y
02(ρ
4c− 2
8y
04)
ρ
6c+ 4 y
02(ρ
4c− 2
8y
04)
2ρ
10c+ 2
16y
06(ρ
4c− 2
8y
04) ρ
10c= y
02ρ
2ch
1 + 2
8(4 + 2
8) y
04ρ
4c+ 2
8(2
10− 2
16) y
08ρ
8ci
≤ 1 2
4h 1 + 1
2
82
8(4 + 2
8) i
≤ 1
2
4(5 + 2
8).
3. Variational formulation and eigenvalue problem Thanks to Hardy’s inequality (1.3) and Poincar´ e’s inequality (2.2),
k u k
µ= ( Z
Ω
[ | ∇
Hdu(z, s) |
2−µV (z, s) | u(z, s) |
2] dzds)
12(3.1)
is equivalent to the norm on H
01(Ω, H
d) for all 0 ≤ µ < 1, so that we will use k · k
µas the norm of H
01(Ω, H
d).
We will use the variational method to study the Dirichlet problem (1.1). We define the following energy functional on H
01(Ω, H
d) :
(3.2) I
µ,λ(u) = 1 2 Z
Ω
h | ∇
Hdu |
2−µV | u |
2i
dzds − 1 p Z
Ω
| u |
pdzds − λ 2 Z
Ω
| u |
2dzds.
Similar to the classical case, I
µ,λ( · ) is well-defined on H
01(Ω, H
d) and belongs to C
1(H
01(Ω, H
d); R).
We say that u ∈ H
01(Ω, H
d) is a weak solution of the Dirichlet problem (1.1), if for any v ∈ C
0∞(Ω), there holds
Z
Ω
∇
Hdu∇
Hdv − µV u v ¯
dzds − Z
Ω
| u |
p−2u ¯ v dzds − λ Z
Ω
u v dzds ¯ = 0
So a weak solution u ∈ H
01(Ω, H
d) of the Dirichlet problem (1.1) is a critical point of I
µ,λ. The Euler-Lagrange equation of the variational problem (3.2) is exactly the semilinear equation in (1.1), and we have
hI
µ,λ0(u), vi = Z
Ω
h
∇
Hdu∇
Hdv − µV u v− | ¯ u |
p−2u ¯ v − λu¯ v i
dzds = 0 for any v ∈ H
01(Ω, H
d).
Since we consider the Dirichlet problem (1.1) for any λ > 0, we cannot use the direct method to prove the existence of the critical point for I
µ,λ. We need to use the Mountain Pass Theorem and the Linking Theorem of Rabinowitz (see [23, 24, 26]).
Thusthat we study firstly the spectral decomposition of H
01(Ω, H
d) with respect to the operator −∆
Hd
−µV where the singular potential V satisfies Hardy’s inequality (1.3). This
eigenvalue problem has also its independent interest. We have the following proposition.
Proposition 3.1. Let 0 ≤ µ < 1. Then there exist 0 < λ
1< λ
2≤ λ
3≤ ... ≤ λ
k≤ ... → +∞, such that for each k ≥ 1, the following Dirichlet problem
(3.3)
−∆
Hd
φ
k− µV φ
k= λ
kφ
k, in Ω φ
k|
∂Ω= 0
admits a nontrivial solution in H
01(Ω, H
d). Moreover, {φ
k}
k≥1constitutes an orthonormal basis of Hilbert space H
01(Ω, H
d).
Remark that the first eigenvalue λ
1is characterized by the following Poincar´ e inequality (3.4) kuk
2L2(Ω)≤ 1
λ
1Z
Ω
|∇
Hdu|
2− µV |u|
2dzds
for all u ∈ H
01(Ω, H
d).
The first step of the proof is the following compact embedding result
Lemma 3.2. Let Ω ∈ H
dbe a bounded open domain. Then H
01(Ω, H
d) is compactly embedded to L
2(Ω).
We can prove this result by the continuous embedding of H
01(Ω, H
d) into usual the Sobolev space H
01/2(Ω), then the compact embedding of H
1/2(Ω) into L
2(Ω). But the first embedding requires some careful extension results. We refer to [19] for a complete and elegant proof of this compact embedding result.
Proof of Proposition 3.1 Denote by L
µ= −∆
Hd
− µV the operator defined on the Hilbert space H
01(Ω, H
d) with the norm k∇
Hduk
L2(Ω), then Hardy’s inequality (1.3) implies
(L
µu, u)
L2(Ω)≥ (1 − µ)k∇
Hduk
2L2(Ω)> 0, ∀ u ∈ H
01(Ω, H
d), and
(L
µu, v) = (u, L
µv), ∀ u, v ∈ H
01(Ω, H
d).
Hence it is positive, definite and self-adjoint on H
01(Ω, H
d). The Lax-Milgram Theorem implies that for any g ∈ H
−1(Ω; H
d), the following Dirichlet problem
L
µu = g in Ω u = 0 on ∂Ω
admits a unique solution u belonging to H
01(Ω, H
d), where H
−1(Ω; H
d) is the dual space of H
01(Ω, H
d), g ∈ H
−1(Ω; H
d) if g ∈ D
0(Ω) and there exists C > 0 such that
|hg, ϕi| ≤ Ckϕk
H1 0(Ω,Hd)for all ϕ ∈ C
0∞(Ω) with the norm
kgk
H−1(Ω,Hd)= sup
ϕ∈C0∞(Ω)
|hg, ϕi|
kϕk
H1 0(Ω,Hd).
Then
∇
Hd: L
2(Ω) → H
−1(Ω; H
d) and ∆
Hd: H
01(Ω, H
d) → H
−1(Ω; H
d)
are continuous. The inverse operator L
−1µof L
µis well defined and it is a continuous map from H
−1(Ω, H
d) into H
01(Ω, H
d).
The compact embedding i : H
01(Ω, H
d) → L
2(Ω) and the continuous embedding i
∗:
L
2(Ω) → H
−1(Ω, H
d) imply that K
µ= L
−1µ◦ i
∗◦ i : H
01(Ω, H
d) → H
01(Ω, H
d) is a compact
and self adjoint operator. So the spectrum of the compact operator K
µis {η
k} such that η
k> 0, k ≥ 1 and η
k→ 0. If {φ
k} are the associated normal eigenvectors, we have that
K
µφ
k= η
kφ
k, ∀ k ≥ 1,
and {φ
k} form a complete basis of Hilbert space H
01(Ω, H
d), which completes the proof of Proposition 3.1.
4. Existence of critical points
We prove now the following existence result of critical points for the variational func- tional I
µ,λwhich gives the weak solution for the Dirichlet problem (1.1).
Theorem 4.1. Let 0 ≤ µ < 1, λ > 0, then I
µ,λadmits at last one nontrivial critical point on H
01(Ω, H
d).
We recall now the well-known Palais-Smale condition.
Definition 4.2. Let E be a Banach space, I ∈ C
1(E, R ) and c ∈ R . We say that I satisfies the (P S)
ccondition, if for any sequence {u
n} ⊂ E with the properties :
I (u
n) → c and k I
0(u
n) k
E0(Ω)→ 0,
there exists a subsequence which is convergent, where I
0( · ) is the Frechet differentiation of I and E
0is the dual space of E. If this holds for any c ∈ R, we say that I satisfies the (P S) condition.
We will prove in the next section the following result
Theorem 4.3. Let 0 ≤ µ < 1, λ > 0, then I
µ,λsatisfies the (P S) condition on H
01(Ω, H
d).
Let 0 < λ
1< λ
2≤ λ
3≤ ... ≤ λ
k≤ ... → +∞ be the eigenvalues of −∆
Hd− µV in Proposition 3.1. We consider firstly the case 0 < λ < λ
1and we use the following Mountain Pass Theorem to prove the existence of a critical point for I
µ,λ:
Theorem 4.4. (see [1, 23])
Let E be a Banach space and I ∈ C
1(E, R ). We suppose that I (0) = 0 and satisfies that (i) there exist R > 0, a > 0 such that if k u k
E= R, then I (u) ≥ a;
(ii) there exists e ∈ E such that k e k> R and I(e) < a. If I satisfies the (P S)
ccondition with
c = inf
h∈Γ
max
t∈[0,1]
I(h(t)), where Γ = { h ∈ C([0, 1]; E); h(0) = 0 and h(1) = e}, then c is a critical value of I and c ≥ a.
We check the above conditions for I = I
µ,λon E = H
01(Ω, H
d). We have I
µ,λ(0) = 0.
For u ∈ H
01(Ω, H
d), Sobolev’s inequality (1.2) and Hardy’s inequality imply that kuk
Lp(Ω)≤ C
Ωk∇
Hduk
L2(Ω)≤ C
Ω(1 − µ)
1/2k u k
µ. Then, for 0 < λ < λ
1I
µ,λ(u) ≥ 1
2 (1 − λ
λ
1) k u k
2µ− C
1p k u k
pµ≥ C
1k u k
2µ1
2C
1(1 − λ λ
1) − 1
p k u k
p−2µ(4.1)
where
C
1=
C
Ω(1 − µ)
1/2 p> 1.
Let
R
0= p
2C
1(1 − λ λ
1)
p−21> 0.
Then for any 0 < R < R
0,
(4.2) inf
kukµ=R
I
µ,λ(u) = a(R) > 0.
So I
µ,λsatisfies condition (i) of Theorem 4.4.
For condition (ii) of Theorem 4.4, take u ∈ H
01(Ω, H
d) such that k u k
µ= R > 0, then for θ ≥ 0,
I
µ,λ(θu) = θ
22
Z
Ω
[ | ∇
Hdu |
2−µV (z, s) | u |
2] dzds (4.3)
− θ
pp
Z
Ω
| u |
pdzds − λθ
22
Z
Ω
| u |
2dzds.
(4.4)
Since p > 2, thus
θ→+∞
lim I
µ,λ(θu) = −∞.
Then, there exists θ
1> 0 large enough such that for e = θ
1u, we have k e k
µ> R and I
µ(θ
1u) < 0 < a(R). Set now
Γ = { h ∈ C([0, 1]; H
01(Ω, H
d)); h(0) = 0 and h(1) = e}, then by continuity, we have
c = inf
h∈Γ
max
t∈[0,1]
I
µ,λ(h(t)) ≥ a(R) > 0,
and c is a local minimum. Theorem 4.3 implies that the (P S)
ccondition is satisfied. So c > 0 is a critical value by using Theorem 4.4 and the critical point is u ∈ H
01(Ω, H
d), which is nontrivial. We have proved Theorem 4.4 for 0 < λ < λ
1.
We need now the following Linking theorem from Rabinowitz [23].
Theorem 4.5. Let E be a Banach space with E = Y ⊕ X, where dim Y < ∞. Suppose that I ∈ C
1(E, R ) and satisfies
(i)there exist ρ, α > 0 such that I|
∂Bρ∩X≥ α;
(ii) there exist e ∈ ∂B
1∩ X and R > ρ such that if A ≡ ( ¯ B
R∩ Y ) ⊕ {r e, 0 < r < R}, then I |
∂A≤ 0.
If I satisfies the (P S)
ccondition with c = inf
h∈Γ
max
u∈A