Nonlinear differ. equ. appl.

15 (2008), 495–509
c 2008 Birkhäuser Verlag Basel/Switzerland
1021-9722/040495-15
published online 26 November 2008 Nonlinear Differential Equations
DOI 10.1007/s00030-008-7031-2 and Applications NoDEA

Flows and Critical Points

Kanishka Perera and Martin Schechter

Abstract. We use flows and the cohomological index to adapt the method of
sandwich pairs to better suit quasilinear elliptic boundary value problems.
Mathematics Subject Classification (2000). Primary 35J65, Secondary 47J10,
47J30.
Keywords. p-Laplacian, boundary value problems, nonlinear eigenvalues, vari-
ational methods, sandwich pairs.

1. Introduction
The notion of sandwich pairs was introduced by Schechter [12] based upon the
sandwich theorem for complementing subspaces by Silva [13] and Schechter [9,10].
Definition 1. We say that a pair of subsets A, B of a Banach space E forms a
sandwich pair if for any G ∈ C 1 (E, R),
−∞ < b0 := inf G ≤ sup G =: a0 < +∞ (1)
B A
implies that there is a sequence {uj } ⊂ E and a c ∈ [b0 , a0 ] such that
G(uj ) → c , G (uj ) → 0 . (2)
The sandwich pairs used in the literature so far have been formed using the
eigenspaces of a semilinear operator and are therefore unsuitable for dealing with
quasilinear problems where there are no eigenspaces. The purpose of the present
paper is to show how this method can be modified to apply to p-Laplacian problems
of the form 
−Δp u = f (x, u) in Ω
(3)
u=0 on ∂Ω
 
where Ω is a bounded domain in Rn , n ≥ 1, Δp u = div |∇u|p−2 ∇u is the
p-Laplacian of u, p ∈ (1, ∞), and f is a Carathéodory function on Ω × R with
subcritical growth, i.e.,
 
|f (x, t)| ≤ C |t|r−1 + 1 ∀(x, t) ∈ Ω × R (4)
496 K. Perera and M. Schechter NoDEA

for some r ∈ [1, p∗ ), where


np
, p<n
p∗ = n−p
(5)
∞, p≥n
is the critical Sobolev exponent, and C > 0.
Solutions of (3) coincide with the critical points of the C 1 functional

G(u) = |∇u|p − p F (x, u) , (6)
Ω
where  t
F (x, t) = f (x, s) ds , (7)
0
defined on the Sobolev space W01, p (Ω). Sandwich pairs that produce Palais–Smale
sequences for G were constructed in [7]. Here we modify the method to produce
Cerami sequences, which give better results. The distinguishing feature of our
current method is the use of flows to obtain critical values. This allows us more
flexibility in dealing with minimax situations.

2. Flows
Let E be a Banach space and let Σ be the set of all continuous maps σ = σ(t)
from E × [0, 1] to E such that
1. σ(0) is the identity map,
2. for each t ∈ [0, 1], σ(t) is a homeomorphism of E onto E.
Because of the special geometry of our problem, we shall require special types of
flows. Let A1 , B1 be a pair of disjoint nonempty closed symmetric subsets of the
unit sphere S in E and let
A = π −1 (A1 ) ∪ {0} , B = π −1 (B1 ) ∪ {0} (8)
where π : E \{0} → S, u → u/ u is the radial projection onto S. For each R > 1,
let K = K(R) = e3 + R, and define
 
K = λv : v ∈ B1 , 0 ≤ λ ≤ K .
B
Let

 
R =
Σ K > 0, max σ(t)(λu) ≤ K,
σ ∈ Σ : min d σ(t)(Ru), B
t∈[0,1] t∈[0,1]

u ∈ A1 , 0 ≤ λ ≤ R . (9)

 R for each R > 1. This follows from the fact that

Note that σ(t) u ≡ u is in Σ
K > R and consequently that
d(Ru, B) = d(Ru, B K ) , u ∈ A1 .
Vol. 15 (2008) Flows and Critical Points 497

We shall prove

Theorem 2. Assume that d(RA1 , B) > e3 ,

K = ∅ ,
σ(1) A ∩ B R ,
σ∈Σ (10)
and
−∞ < b0 := inf G ≤ sup G =: a0 < +∞ . (11)
B A

c satisfying b0 ≤ 
Then there are a number  c ≤ a0 and a sequence {uj } ⊂ E
satisfying
 
G(uj ) → 
c, 1 + uj  G (uj ) → 0 . (12)

For our applications, we shall need conditions which will imply (10). This
will be done in the next section.

3. Cohomological index
We recall the construction and some properties of the cohomological index of
Fadell and Rabinowitz [5]. Writing the group Z2 multiplicatively as {1, −1}, a
paracompact Z2 -space is a paracompact space X together with a mapping μ :
Z2 × X → X, called a Z2 -action on X, such that
μ(1, x) = x , −(−x) = x ∀x ∈ X (13)
where −x := μ(−1, x). The action is fixed-point free if
−x = x ∀x ∈ X . (14)
A subset A of X is invariant if
 
−A := − x : x ∈ A = A, (15)
and a map f : X → X  between two paracompact Z2 -spaces is equivariant if
f (−x) = −f (x) ∀x ∈ X . (16)
Two spaces X and X  are equivalent if there is an equivariant homeomorphism
f : X → X  . We denote by F the set of all paracompact free Z2 -spaces, identifying
equivalent ones.
A principal Z2 -bundle with paracompact base is a triple ξ = (E, p, B) con-
sisting of an E ∈ F, called the total space, a paracompact space B, called the base
space, and a map p : E → B, called the bundle projection, such that there are
1. an open covering {Uλ }λ∈Λ of B,
2. for each λ ∈ Λ, a homeomorphism ϕλ : Uλ × Z2 → p−1 (Uλ ) satisfying
ϕλ (b, −1) = −ϕλ (b, 1) , p ϕλ (b, ±1) = b ∀b ∈ B . (17)
498 K. Perera and M. Schechter NoDEA

Then each p−1 (b), called a fiber, is some pair {e, −e} , e ∈ E. A bundle map
f : ξ → ξ  consists of an equivariant map f : E → E  and a map f : B → B  such
that p f = f p, i.e., the diagram
f
E −−−−→ E 
⏐ ⏐
⏐
p
⏐
p 

f
B −−−−→ B 
commutes. Two bundles ξ and ξ  are equivalent if there are bundle maps f : ξ → ξ 
and f  : ξ  → ξ such that f  f and f f  are the identity bundle maps on ξ and ξ  ,
respectively. We denote by PrinZ2 B the set of principal Z2 -bundles over B and
Prin Z2 the set of all principal Z2 -bundles with paracompact base, identifying
equivalent ones.
Each X ∈ F can be identified with a ξ ∈ Prin Z2 as follows. Let X = X/Z2
be the quotient space of X ∈ F with each x and −x identified, called the orbit
space of X, and π : X → X the quotient map. Then
P : F → Prin Z2 , X → ξ := (X, π, X) (18)
is a one-to-one correspondence.
A map f : B → B  induces a bundle f ∗ ξ  = (f ∗ (E  ), p, B) ∈ Prin Z2 , called
the pullback, where
 
f ∗ (E  ) = (b, e ) ∈ B × E  : f (b) = p (e ) , −(b, e ) = (b, −e ) (19)
and
p(b, e ) = b . (20)
Homotopic maps induce equivalent bundles, so for each B  ∈ Prin Z2 , we have the
mapping
T : [B, B  ] → PrinZ2 B , [f ] → f ∗ ξ  (21)
where [B, B ] is the set of homotopy classes of maps from B to B  . For the bundle


ξ  = (S ∞ , π, RP∞ ), called the universal principal Z2 -bundle, where S ∞ is the

unit sphere in R∞ , RP∞ is the infinite dimensional real projective space, and π
identifies antipodal points ±x, T is a one-to-one correspondence (see Dold [3]).
Thus for each X ∈ F, there is a map f : X → RP∞ , unique up to homotopy and
called the classifying map, such that
T ([f ]) = P(X) . (22)
∗ ∗ ∞ ∗
Let f : H (RP ) → H (X) be the induced homomorphism of the Alexander–
Spanier cohomology rings. The cohomological index of X is defined by
  
sup k ≥ 1 : f ∗ (ω k−1 ) = 0 , X = ∅
i(X) = (23)
0, X=∅
where ω ∈ H 1 (RP∞ ) is the generator of the polynomial ring H ∗ (RP∞ ) = Z2 [ω].
The index i : F → N ∪ {0, ∞} has the usual properties of an index theory:
Vol. 15 (2008) Flows and Critical Points 499

1. Definiteness: i(X) = 0 if and only if X = ∅.

2. Monotonicity: If f : X → Y is an equivariant map, in particular, if X ⊂ Y ,
then
i(X) ≤ i(Y ) .
Thus, equality holds when f is an equivariant homeomorphism.
3. Subadditivity: If X ∈ F and A, B are closed invariant subsets of X such that
X = A ∪ B, then
i(A ∪ B) ≤ i(A) + i(B) .
4. Continuity: If X ∈ F and A is a closed invariant subset of X, then there is
a closed invariant neighborhood N of A in X such that
i(N ) = i(A) .
5. Neighborhood of zero: If U is a bounded symmetric neighborhood of 0 in a
Banach space E, then
i(∂U ) = dim E .
Now let E be a Banach space and let A ⊂ F denote the class of symmetric
subsets of E \ {0}. The suspension SA of a nonempty subset A of E is the quotient
space of A × [−1, 1] with A × {1} and A × {−1} collapsed to different points, which
can be realized in E ⊕ R as the union of all line segments joining the two points
(0, ±1) ∈ E ⊕R to points of A ⊂ E. The cohomological index also has the following
important stability property: If A ∈ A is closed, then
i(SA) = i(A) + 1 (24)
(see Fadell and Rabinowitz [5]). Let S ∈ A be the unit sphere in E and let π be
the radial projection onto S. Our main theorem for this section is
Theorem 3. Let A1 , B1 be a pair of disjoint nonempty closed symmetric subsets
of S such that
i(A1 ) = i(S \ B1 ) < ∞ (25)
and let
A = π −1 (A1 ) ∪ {0} , B = π −1 (B1 ) ∪ {0} . (26)
Assume that
−∞ < b0 := inf G ≤ sup G =: a0 < +∞ . (27)
B A
Then there are a number 
c satisfying b0 ≤  c ≤ a0 and a sequence {uj } ⊂ E
satisfying
 
G(uj ) → 
c, 1 + uj  G (uj ) → 0 . (28)
In proving Theorem 3, we shall make use of the following considerations.
Theorem 4. If (25) holds, then for each R > 1,
K = ∅ ,
σ(1) A ∩ B R .
σ∈Σ (29)
500 K. Perera and M. Schechter NoDEA

 R such that
Proof. Suppose there is a σ ∈ Σ
K = ∅ .
σ(1) A ∩ B (30)
 R , we have
By the definition of Σ
 
K = ∅ .
σ(t)(RA1 ) : t ∈ [0, 1] ∩ B (31)
Let ⎧
⎪
⎨(1 − 3t + 3Rt) u , u ∈ A1 , 0 ≤ t ≤ 1/3
Γ(t) u = σ(3t − 1)(Ru) , u ∈ A1 , 1/3 < t ≤ 2/3 (32)
⎪
⎩
σ(1) (3(1 − t) Ru) , u ∈ A1 , 2/3 < t ≤ 1 .
By (30) and (31),
 
K = ∅ .
Γ(t) A1 : t ∈ [0, 1] ∩ B (33)
Moreover, since
max σ(t)(λu) ≤ K , u ∈ A1 , 0≤λ≤R (34)
t∈[0,1]

and
u > K , K ,
u∈B\B (35)
we also have    
K = ∅ .
Γ(t) A1 : t ∈ [0, 1] ∩ B \ B (36)
So Γ = Γ(t) is a continuous map from A1 × [0, 1] to E \ B such that Γ(0) is the
identity map on A1 and Γ(1) A1 is the single point σ(1)(0) ∈ E. Then

π(Γ(t) u) , u ∈ A1 , t ∈ [0, 1]
SA1 → S \ B1 , (u, t) → (37)
−π(Γ(−t)(−u)) , u ∈ A1 , t ∈ [−1, 0)
is an odd map and hence
i(S \ B1 ) ≥ i(SA1 ) = i(A1 ) + 1 (38)
by the monotonicity of the index and (24), contradicting (25). 
Theorem 5. Assume that d(RA1 , B) > e3 and
 

cR := inf sup G σ(1) u (39)
 R u∈A
σ∈Σ

is finite. Then there is a sequence {uj } ⊂ E satisfying

 
G(uj ) → cR , K ) G (uj ) → 0 .
1 + d(uj , B (40)

Theorem 5 will be proved in Section 5.

Proof of Theorem 2. Assumption (10) guarantees that 
cR is finite. In fact, we have
b0 ≤ 
cR ≤ a0 .
K ),
Then by Theorem 5, there is a sequence satisfying (40). Since u ≤ K +d(u, B
the result follows. 
Vol. 15 (2008) Flows and Critical Points 501

Proof of Theorem 3. Take R so large that d(RA1 , B) > e3 . By (25) and Theo-
rem 4, (10) holds, so the result follows from Theorem 2. 

4. Ordinary differential equations

In proving Theorem 5 we shall make use of various extensions of Picard’s theorem
in a Banach space. Some are well known.
Lemma 6. Let γ(t) and ρ(t) be continuous functions on [0, ∞), with γ(t) nonneg-
ative and ρ(t) positive. Assume that
 ∞  T
dτ
> γ(s) ds , (41)
u0 ρ(τ ) t0

where t0 < T and u0 are given positive numbers. Then there is a unique solution
of
 
u (t) = γ(t)ρ u(t) , t ∈ [t0 , T ) , u(t0 ) = u0 (42)
which is positive in [t0 , T ) and depends continuously on u0 .
Proof. One can separate variables to obtain
 u  t
dτ
W (u) = = γ(s) ds .
u0 ρ(τ ) t0

The function W (u) is differentiable and increasing in R, positive in [u0 , ∞), de-
pends continuously on u0 and satisfies
 ∞  T
dτ
W (u) → L = > γ(s) ds , as u → ∞ .
u0 ρ(τ ) t0

Thus, for each t ∈ [t0 , T ) there is a unique u ∈ [u0 , ∞) such that

 t 
−1
u=W γ(s) ds
t0

is the unique solution of (42), and it depends continuously on u0 . 

Lemma 7. Let γ(t) and ρ(t) be continuous functions on [0, ∞), with γ(t) nonneg-
ative and ρ(t) positive. Assume that
 u0  T
dτ
> γ(s) ds , (43)
m ρ(τ ) t0

where t0 < T and m < u0 are given positive numbers. Then there is a unique
solution of
 
u (t) = −γ(t)ρ u(t) , t ∈ [t0 , T ) , u(t0 ) = u0 (44)
which is ≥ m in [t0 , T ) and depends continuously on u0 .
502 K. Perera and M. Schechter NoDEA

Proof. One can separate variables to obtain

 u0  t
dτ
W (u) = = γ(s) ds .
u ρ(τ ) t0

The function W (u) is differentiable and decreasing in R, positive in [m, u0 ], de-

pends continuously on u0 and satisfies
 u0  T
dτ
W (u) → L = > γ(s) ds , as u → m .
m ρ(τ ) t0

Thus, for each t ∈ [t0 , T ) there is a unique u ∈ [m, u0 ] such that

 t 
u = W −1 γ(s) ds
t0

is the unique solution of (44), and it depends continuously on u0 . 

Theorem 8. Let g(t, x) be a continuous map from R × H to H, where H is a

Banach space. Assume that for each point (t0 , x0 ) ∈ R × H, there are constants
K, b > 0 such that
g(t, x)−g(t, y) ≤ Kx−y , |t−t0 | < b , x−x0  < b , y −x0  < b . (45)
Assume also that
g(t, x) ≤ γ(t)ρ(x) , x∈H, t ∈ [t0 , ∞) , (46)
where γ(t), ρ(t) satisfy the hypotheses of Lemma 6 with ρ(t) nondecreasing. Then
for each x0 ∈ H and t0 > 0 there is a unique solution x(t) of the equation
dx(t)  
= g t, x(t) , t ∈ [t0 , ∞) , x(t0 ) = x0 . (47)
dt
Moreover, x(t) depends continuously on x0 and satisfies
x(t) ≤ u(t) , t ∈ [t0 , ∞) , (48)
where u(t) is the solution of (42) in that interval satisfying u(t0 ) = u0 ≥ x0 .

We also have the following.

Theorem 9. Let ρ, γ satisfy the hypotheses of Lemma 7, with ρ locally Lipschitz

continuous. Let u(t) be the solution of (44), and let h(t) be a continuous function
satisfying
 t
 
h(t) ≥ h(s) − γ(r)ρ h(r) dr , t0 ≤ s < t < T , h(t0 ) ≥ u0 . (49)
s

Then
u(t) ≤ h(t) , t ∈ [t0 , T ) . (50)
Vol. 15 (2008) Flows and Critical Points 503

Proof. Assume that there is a point t1 in the interval such that

h(t1 ) < u(t1 ) .
Let
y(t) = u(t) − h(t) , t ∈ [t0 , T ) .
Then, y(t0 ) ≤ 0 and y(t1 ) > 0. Let τ be the largest point < t1 such that y(τ ) = 0.
Then
y(t) > 0 , t ∈ (τ, t1 ] . (51)
Moreover, by (44) and (49) we have
 t    t
  
y(t) ≤ − γ(s) ρ u(s) − ρ h(s) ds ≤ L y(s) ds , (52)
τ τ
where L is the Lipschitz constant for ρ at u(τ ) times the maximum of γ in the
interval. Let  t
w(t) = y(s) ds .
τ
Then    
e−Lt w(t) = e−Lt y(t) − Lw(t) ≤ 0 , t ∈ [τ, t1 ] .
Consequently,
e−Lt w(t) ≤ e−Lτ w(τ ) = 0 , t ∈ [τ, t1 ] .
Hence,
y(t) ≤ Lw(t) ≤ 0 , t ∈ [τ, t1 ] ,
contradicting (51). This completes the proof. 

5. Proof of Theorem 5
Let ρ(r) = 1 + r. If the theorem were not true, there would be a δ > 0 such that
 
K ) G (u) ≥ 1
ρ d(u, B (53)
would hold for all u in the set
 
Q= u∈E: cR − 3δ ≤ G(u) ≤ 
cR + 3δ . (54)
Let
 
Q0 = u ∈ Q : 
cR − 2δ ≤ G(u) ≤ 
cR + 2δ (55)
 
Q1 = u ∈ Q : 
cR − δ ≤ G(u) ≤ 
cR + δ (56)
and  
Q2 = E \ Q0 , η(u) = d(u, Q2 )/ d(u, Q1 ) + d(u, Q2 ) . (57)
It is easily checked that η(u) is locally Lipschitz continuous on E and satisfies
⎧
⎪
⎨η(u) = 1 , u ∈ Q1 ,
η(u) = 0 , u ∈ Q2 , (58)
⎪
⎩
η(u) ∈ (0, 1) , otherwise .
504 K. Perera and M. Schechter NoDEA

For any θ < 1 there is a locally Lipschitz continuous map Y (u) of Ê = {u ∈ E :

G (u) = 0} into itself such that
 
Y (u) ≤ 1 , θG (u) ≤ G (u), Y (u) , u ∈ Ê (59)
(cf., e.g., [11]). Let σ(t) be the flow generated by
 
K ) .
W (u) = −η(u)Y (u)ρ d(u, B (60)
 and W (u) is locally Lipschitz continuous, σ(t) exists
Since W (u) ≤ ρ(d(u, B))
for all t ∈ R+ in view of Theorem 8. We also have
   
dG σ(t)u /dt = G (σ), σ  (61)
    
= −η(σ) G (σ), Y (σ) ρ d(σ, BK )
 
≤ −θη(σ)G (σ)ρ d(σ, BK )
= −θη(σ)
in view of (53) and (59). Let T satisfy 2δ/θ < T < 3, and suppose u ∈ Q1 is such
that there is a t1 ∈ [0, T ] for which σ(t1 )u ∈
/ Q1 . Then
 
G σ(t1 )u <  cR − δ ,
since we cannot have G(σ(t1 )u) > 
cR + δ for u ∈ Q1 by (61). But this implies
 
G σ(T )u <  cR − δ . (62)
On the other hand, if σ(t)u ∈ Q1 for all t ∈ [0, T ], then
 T
 
G σ(T )u ≤ G(u) − θ dt ≤ cR + δ − θT < 
cR − δ
0
by (61). Thus, (62) holds for u ∈ Q1 .
We claim that σ1 (t) = σ(tT ) ∈ Σ R . To see this, note that since
 t
 
σ(t)u − u = W σ(τ )u dτ , (63)
0
we have  t   
σ(t)u − σ(s)u ≤ K dr .
ρ d σ(r)u, B
s
K , we have
If v ∈ B
  
  t 
K ≤ σ(s)u − v ≤ σ(t)u − v +
h(s) = d σ(s)u, B  dr .
ρ d σ(r)u, B
s
This implies,
 t  
h(s) ≤ h(t) + ρ h(r) dr . (64)
s
Moreover, if we replace u by Ru, we see by Lemma 7 and Theorem 9, that h(s)
satisfies
m(s) ≤ h(s) , 0 ≤ s ≤ T ,
Vol. 15 (2008) Flows and Critical Points 505

where m(s) is given by

 m(0)
dτ
= s.
m(s) ρ(τ )
Note that for u ∈ A1 , we have e3 m(T ) ≥ 1. Thus
 
min d σ1 (t)(Ru), BK > 0
t∈[0,1]

when u ∈ A1 . Moreover, by Theorem 8, we have

x(t) = σ(t)(λu) − λu ≤ e3 , u ∈ A1 , 0 ≤ λ ≤ R.
This follows from the fact that x(0) = 0 and
x(t) ≤ m(t)
 , t ∈ [0, T ] ,

where m(t) satisfies
 
m(t)
ds
= t, t ∈ [0, T ] .
0 1+s

Thus, m(t) ≤ eT ≤ e3 . Hence,
σ(t)(λu) ≤ K , t ∈ [0, T ] , u ∈ A1 , 0 ≤ λ ≤ R.
 R and
This tells us that σ1 ∈ Σ
 
G σ1 (1)u < 
cR − δ , u ∈ A. (65)
But this contradicts (11). Hence (53) cannot hold for u satisfying (54). The proof
is complete.

6. Applications to the p-Laplacian

We shall now study problem (3). Consider the nonlinear eigenvalue problem

− Δp u = λ |u|p−2 u in Ω
(66)
u=0 on ∂Ω .
Its eigenvalues coincide with the critical values of the C 1 functional
1
I(u) =  (67)
Ω
|u|p
defined on the unit sphere S in W01, p (Ω). Let F denote the class of symmetric
subsets of S and set
λl = inf sup I(u) (68)
M ∈F u∈M
i(M )≥l
where i is the cohomological index. Then
0 < λ 1 < λ2 ≤ · · · → ∞ (69)
are eigenvalues of (66) (cf. Theorem 4.2.1 of Perera et al. [8]).
506 K. Perera and M. Schechter NoDEA

Setting H(x, t) = p F (x, t) − tf (x, t), we shall prove

Theorem 10. If λl < λl+1 and
λl |t|p − W (x) ≤ p F (x, t) ≤ λl+1 |t|p + W (x) , t∈R (70)
for some l > 0 and W ∈ L (Ω), and
1

H(x, t)
H(x, t) ≤ C (|t|α + 1) , H(x) := lim < 0 a.e. (71)
|t|→∞ |t|α
for some α satisfying 0 < α ≤ p, then (3) has a solution.
Theorem 11. If λl < λl+1 and (70) holds for some l > 0 and W ∈ L1 (Ω), and
H(x, t)
H(x, t) ≥ −C (|t|α + 1) , H(x) := lim > 0 a.e. (72)
|t|→∞ |t|α
for some α satisfying 0 < α ≤ p, then (3) has a solution.
Theorem 12. If λl < λl+1 and (70) holds for some l > 0 and W ∈ L1 (Ω), and
H(x, t) ≤ W1 (x) ∈ L1 (Ω) , H(x, t) → −∞ as |t| → ∞ , (73)
then (3) has a solution.
Theorem 13. If λl < λl+1 and (70) holds for some l > 0 and W ∈ L1 (Ω), and
H(x, t) ≥ −W1 (x) ∈ L1 (Ω) , H(x, t) → +∞ as |t| → ∞ , (74)
then (3) has a solution.
Similar resonance problems have been studied by Perera [6] when
f (x, t)/|t|p−2 t → α± (x) ∈ L∞ (Ω) as t → ±∞ and by Arcoya and Orsina [1],
Bouchala and Drábek [2], and Drábek and Robinson [4] for the special case α± =
const.
Proof of Theorem 10. Let
   
A1 = u ∈ S : I(u) ≤ λl , B1 = u ∈ S : I(u) ≥ λl+1 . (75)
Then
i(A1 ) = i(S \ B1 ) = l (76)
by Theorem 4.2.2 of Perera et al. [8]. Let A, B be as in Theorem 3 and let G be
given by (6). Since  
|∇u| ≥ λl+1
p
|u|p , u ∈ B (77)
Ω Ω
and  
|∇u|p ≤ λl |u|p , u ∈ A, (78)
Ω Ω
(70) implies  
− W ≤ inf G ≤ sup G ≤ W, (79)
Ω B A Ω
so there is a sequence {uj } ⊂ W01, p (Ω) satisfying (28) by Theorem 3.
Vol. 15 (2008) Flows and Critical Points 507

We claim that {uj } is bounded and hence has a convergent subsequence by

a standard argument. If ρj = uj  → ∞, a subsequence of u j = uj /ρj converges
1, p
 weakly in W0 (Ω), strongly in L (Ω), and a.e. in Ω. Then
to some u p


H(x, uj ) G (uj ) uj /p − G(uj )
= →0 (80)
Ω ρα
j ραj

by (12) and
  
H(x, uj ) H(x, uj )
lim ≤ lim |
uj |α ≤ H(x) |
u|α ≤ 0 (81)
Ω ρα
j Ω |uj |α Ω

by (71). Since H < 0 a.e., it follows that u = 0 a.e. Now passing to the limit in
 
G(uj ) p F (x, uj ) W
1− = ≤ λl+1 |
uj |p + p (82)
ρpj Ω ρ p
j Ω ρj
gives

1≤ λl+1 |
u|p ,
Ω
 = 0 a.e.
contradicting the fact that u 

Proof of Theorem 11. As in the proof of Theorem 10, there is a sequence {uj } ⊂
W01, p (Ω) satisfying (28). We claim that {uj } is bounded and hence a subsequence
converges to a critical point of G. If ρj = uj  → ∞, a subsequence of u
j = uj /ρj
1, p
converges to some u  weakly in W0 (Ω), strongly in L (Ω), and a.e. in Ω. Then
p


H(x, uj ) G (uj ) uj /p − G(uj )
= →0 (83)
Ω ρα
j ραj

by (12) and
  
H(x, uj ) H(x, uj )
lim ≥ lim |
uj |α ≥ H(x) |
u|α ≥ 0 (84)
Ω ρα
j Ω |uj |α Ω

by (72). Since H > 0 a.e., it follows that u = 0 a.e. Now passing to the limit in
 
G(uj ) p F (x, uj ) W
1− p = p ≤ λl+1 |
uj |p + p (85)
ρj Ω ρj Ω ρj
again gives

1≤ λl+1 |
u|p ,
Ω
 = 0 a.e.
contradicting the fact that u 

Proof of Theorem 12. We follow the proof of Theorem 10. We can conclude that
there is a sequence {uj } ⊂ W01, p (Ω) satisfying (28). Again we claim that {uj }
is bounded, and hence a subsequence converges to a critical point of G. If ρj =
508 K. Perera and M. Schechter NoDEA

uj  → ∞, a subsequence of u j = uj /ρj converges to some u weakly in W01, p (Ω),

p
strongly in L (Ω), and a.e. in Ω. Now
 
  
H(x, uj ) dx = G (uj ) uj /p − G(uj ) dx → −c. (86)
Ω Ω
This implies  
 
 H(x, uj ) dx ≤ K . (87)
 
Ω
(x) ≡ 0. Let Ω0 be the subset of Ω on which ũ = 0. Then
As before, we show that u
|uj (x)| = ρj |ũj (x)| → ∞, x ∈ Ω0 . (88)
If Ω1 = Ω \ Ω0 , then we have
    
H(x, uj ) dx = + ≤ H(x, uj ) dx + W1 (x) dx → −∞ . (89)
Ω Ω0 Ω1 Ω0 Ω1

This contradicts (87), and we see that ρj = uj  is bounded. 

Proof of Theorem 13. In this case we have
    
H(x, uj ) dx = + ≥ H(x, uj ) dx − W1 (x) dx → ∞ , (90)
Ω Ω0 Ω1 Ω0 Ω1

which contradicts (87) as well. 

References
[1] D. Arcoya and L. Orsina. Landesman–Lazer conditions and quasilinear elliptic equa-
tions. Nonlinear Anal., 28(10):1623–1632, 1997.
[2] J. Bouchala and P. Drábek. Strong resonance for some quasilinear elliptic equations.
J. Math. Anal. Appl., 245(1):7–19, 2000.
[3] A. Dold. Partitions of unity in the theory of fibrations. Ann. of Math. (2), 78:223–255,
1963.
[4] P. Drábek and S. B. Robinson. Resonance problems for the p-Laplacian. J. Funct.
Anal., 169(1):189–200, 1999.
[5] E. R. Fadell and P. H. Rabinowitz. Generalized cohomological index theories for Lie
group actions with an application to bifurcation questions for Hamiltonian systems.
Invent. Math., 45(2):139–174, 1978.
[6] K. Perera. One-sided resonance for quasilinear problems with asymmetric nonlinear-
ities. Abstr. Appl. Anal., 7(1):53–60, 2002.
[7] K. Perera and M. Schechter. Sandwich pairs in p-Laplacian problems. Topol. Methods
Nonlinear Anal., 29(1):29–34, 2007.
[8] K. Perera, R. P. Agarwal, and D. O’Regan. Morse-theoretic aspects of p-Laplacian
type operators. Progress in Nonlinear Differential Equations and their Applications,
Birkhäuser Boston Inc., Boston, MA, to appear.
[9] M. Schechter. A generalization of the saddle point method with applications. Ann.
Polon. Math., 57(3):269–281, 1992.
Vol. 15 (2008) Flows and Critical Points 509

[10] M. Schechter. New saddle point theorems. In Generalized Functions and Their Ap-
plications (Varanasi, 1991), pages 213–219. Plenum, New York, 1993.
[11] M. Schechter. Linking Methods in Critical Point Theory, Birkhäuser Boston, 1999.
[12] M. Schechter. Sandwich pairs in critical point theory. Trans. Amer. Math. Soc.,
360(6):2811–2823, 2008.
[13] Silva, Elves A. B. Linking theorems and applications to semilinear elliptic problems
at resonance. Nonlinear Anal., 16(5):455–477, 1991.

Kanishka Perera
Department of Mathematical Sciences
Florida Institute of Technology
Melbourne, FL 32901-6975
USA
e-mail: [email protected]
Martin Schechter
Department of Mathematics
University of California
Irvine, CA 92697-3875
USA
e-mail: [email protected]

Received: 17 July 2007.

Accepted: 12 February 2008.