Journal of Applied Analysis and Computation Website:http://www.jaac-online.

com
Volume 14, Number 4, August 2024, 2441–2465 DOI:10.11948/20230449

MULTIPLICITY OF WEAK SOLUTIONS FOR

A (P (X), Q(X))-KIRCHHOFF EQUATION
WITH NEUMANN BOUNDARY CONDITIONS

A. Ahmed1,† and Mohamed Saad Bouh Elemine Vall2

Abstract The aim of this study is to investigate the existence of infinitely

many weak solutions for the (p(x), q(x))-Kirchhoff Neumann problem described
by the following equation :
  Z 
1 p(x)

 − a1 + a 2 |∇u| dx ∆p(·) u
Ω p(x)




  Z 
 1
|∇u|q(x) dx ∆q(·) u

− b + b
1 2
Ω q(x)
  
 +λ(x) |u|p(x)−2 u + |u|q(x)−2 u = f1 (x, u) + f2 (x, u)

 in Ω,


 ∂u = 0


on ∂Ω.

∂ν

By employing a critical point theorem proposed by B. Ricceri, which stems

from a more comprehensive variational principle, we have successfully estab-
lished the existence of infinitely many weak solutions for the aforementioned
problem.

Keywords Nonlinear elliptic equations, weak solutions to PDEs, Ricceri’s

variational principle, double phase problems, Musielak-Orlicz-Sobolev spaces.

MSC(2010) 35J60, 35D30, 35J20.

1. Introduction
Studying differential equations with double-phase operators is a novel and fasci-
nating subject. It is caused by factors such as extremely anisotropic materials,
Lavrentiev’s phenomenon, and nonlinear elasticity theory (see [43–46]). In recent
years, there has been a surge in interest in the study of double-phase problems, with
numerous results obtained, see for example [6, 8, 10, 26, 29, 30, 33, 34].
By taking into account the fluctuations in the string’s length during vibrations,
Kirchhoff’s differential equations, as outlined by Kirchhoff [27], extend the classical

†
The corresponding author.
1
Mathematics and Computer Sciences Department, Faculty of Science and
Technology, University of Nouakchott, Nouakchott, Mauritania
2
Department of Industrial Engineering and Applied Mathematics, Professional
University Institute, University of Nouakchott, Nouakchott, Mauritania
Email: [email protected](A. Ahmed),
[email protected](M. S. B. Elemine Vall)
2442 A. Ahmed & M. S. B. Elemine Vall

D’Alembert’s wave equation.

!
L 2
∂2u ∂2u
Z
P0 E ∂u
r 2 − + dx = 0, (1.1)
∂t h 2L 0 ∂x ∂x2

where L, h, E, P0 and r are constants.

The Kirchhoff equation (1.1) is characterized by the presence of a non-local
Z L 2 Z L 2
P0 E ∂u E ∂u
component + dx which varies depending on average
h 2L 0 ∂x 2L 0 ∂x
2
1 ∂u
of the kinetic energy on the domain [0, L], and as a result, the equation
2 ∂x
ceases to be a point-wise identity, (see [3–5, 7, 20, 25, 38]) for related topics.
The aim of this research is to show the existence of infinitely many weak solutions
to the following elliptic problem involving double phase operators of Kirchhoff type
and a Neumann boundary value condition.
  Z   Z 
1 p(x) 1 q(x)
 − a1 + a2
 |∇u| dx ∆p(·) u − b1 + b2 |∇u| dx ∆q(·) u
Ω p(x) Ω q(x)


  
 +λ(x) |u|p(x)−2 u + |u|q(x)−2 u = f1 (x, u) + f2 (x, u) in Ω,

 ∂u = 0


on ∂Ω,
∂ν
(1.2)
where Ω ⊂ RN is a bounded open domain such that his boundary ∂Ω is of class C 1 ,
and denoted by ν the outward unit normal to ∂Ω, p ≡ p(x), q ≡ q(x) ∈ C+ (Ω) with

N < p− ≤ p+ < q − ≤ q + < +∞, (1.3)

and a1 , a2 , b1 , b2 > 0, λ ∈ L∞ (Ω) and there is a positive constant λ0 satisfying

λ0 ≤ λ(x).
Let f1 , f2 : Ω × R 7−→ R are two Carathéodory functions such that for all r > 0,
we have
sup | f1 (x, t) |∈ L1 (Ω) and sup | f2 (x, t) |∈ L1 (Ω). (1.4)
|t|≤r |t|≤r

It is not surprising that there have been articles dealing with questions related to
this type of operator in the classical Sobolev spaces. We refer the reader to [4,37,40]
for some examples, where the authors are interested in the Dirichlet problem.
An important generalization of the p-Laplace operator is the p(·)-Laplace opera-
tor. The p(·)-Laplace operator has more complex nonlinearities than the p-Laplace
operator.
In variable exposent Sobolev spaces, A. Crespo-Blanco et all in [11] propose
a new type of quasi-linear elliptic equations controlled by so-called double phase
operators with variable exponents. They prove some properties of the corresponding
Musielak-Orlicz Sobolev space and properties of the new double phase operator and
show the existence and uniqueness of elliptic equations corresponding to straight
sides with dependence into the slope, see also [14, 41] for related topics. In this
article, we use Kirchhoff-type operators in an elliptic Neumann problem.
To our knowledge, few papers have been studied dealing with the elliptic problem
introducing the Kirchhoff-type operator in the case of the Neumann condition (see
[2,13,15,28,42]). The hypotheses used in this paper, as well as the results, are quite
different from the previous results.
Multiplicity of weak solutions 2443

But because of its non-homogeneities and the existence of numerous nonlinear

elements, the issue (1.2) has a more complex structure if either p or q are non
constant functions.
Our earlier work inspired us to expand these conclusions within the Musielak-
Orlicz-Sobolev space, a more all-encompassing functional framework that has gar-
nered interest from academics.
The motivation for this work is provided by its physical applications, specif-
ically the issues with equations of the Kolmogorov-type that arise in the theory
of diffusion, theory of non-Newtonian fluids with strongly inhomogeneous behavior
and a high propensity to increase their viscosity in response to shear rate, electro-
rheological fluids electric or magnetic field, and (see references [18, 19, 23, 35, 39]).
We refer to [1,15,16,21,22,31] for some more discoveries on elliptic and parabolic
problems in Musielak-Orlicz-Sobolev spaces.
The main obstacle with this type of problem is the setting of Sobolev spaces with
double phase exponents and the fact that there is a Neumann boundary condition
that makes the Theorem 1.1 difficult to apply.
We present a crucial result obtained by B. Ricceri in [36], which is necessary to
prove our primary findings.
Theorem 1.1. (See [17], Theorem 2.2). Consider two Gâteaux differentiable and
sequentially weakly lower semi-continuous functionals Φ1 , Φ2 : E −→ R on a re-
flexive real Banach space E and suppose that Φ2 is continuous with respect on the
norm topology and lim Φ2 (u) = +∞. For r > inf Φ2 , we set
kukE →∞ E

Φ1 (u) − inf Φ1 (v)

v∈(Φ−1
2 (]−∞,r[))w
ϕ(r) = inf , (1.5)
u∈Φ−1
2 (]−∞,r[)
r − Φ2 (u)

where (Φ−1 −1
2 (] − ∞, r[))w denoted the adherence of Φ2 (] − ∞, r[) with regards to the
topology of weak convergence. Then, the following claims are accurate
(a) If we have r0 > inf Φ2 and u0 ∈ E such that
E

Φ2 (u0 ) < r0 , (1.6)

and
Φ1 (u0 ) − inf Φ1 (v) < r0 − Φ2 (u0 ), (1.7)
v∈(Φ−1
2 (]−∞,r0 [))w

then the restriction of Φ1 + Φ2 to Φ−1

2 (] − ∞, r0 [) admits at least global mini-
mum point.
 
(b) If we have two sequences (rn )n ⊂ inf Φ2 , +∞ with rn → ∞ and (un )n ⊂ E
E
such that for any n
Φ2 (un ) < rn , (1.8)
and
Φ1 (un ) − inf Φ1 (v) < rn − Φ2 (un ), (1.9)
v∈(Φ−1
2 (]−∞,rn [))w

and also we assume

lim inf (Φ2 (u) + Φ1 (u)) = −∞. (1.10)
kuk→+∞
2444 A. Ahmed & M. S. B. Elemine Vall

Consequently, we can find a sequence (vn )n of local minima of Φ2 + Φ1 such

that Φ2 (vn ) → +∞ as n → ∞.
(c) If we have two sequences (rn )n ⊂ (inf Φ2 , +∞) with rn → inf Φ2 and (un )n ⊂ E
E E
such that for every n, the conditions (1.8) and (1.9) are met, and also assume
that:

The global minimizers of Φ2 are not local minimizers of Φ1 + Φ2 . (1.11)

Then, we can find a sequence (vn )n of pairwise different local minimizers of

Φ1 + Φ2 such that lim Φ2 (vn ) = inf Φ2 , and (vn )n weakly converges to a
n→∞ E
global minimizer of Φ2 .

The following describes the structure of this paper : In Section 2, we provide

some important background information. We outline an improvement in Section 3
(see Theorems 3.1 and 3.2) and support its credibility with examples (see Corollary
3.1 and 3.2).

2. Preliminary results
Consider a smooth bounded open domain Ω ⊂ RN , and let we define
n o
C+ (Ω) = z ∈ M, z(·) : Ω −→ R : 1 < z − = ess inf z(x) ≤ z + = ess sup z(x) < ∞ ,
x∈Ω x∈Ω

here M represents the collection of measurable real functions.

The variable exponent Lebesgue space Lz(·) Z
(Ω) is defined as the set of measurable
functions u : Ω 7−→ R satisfying rz(·) (u) := |u|z(x) dx < ∞ endowed with the
Ω
following norm

kukLz(·) (Ω) = kukz(·) = inf σ > 0 : rz(·) (u/σ) ≤ 1 ,

known as the Luxemburg norm. Then, the space (Lz(·) (Ω), k · kz(·) ) is a separable
reflexive and uniformly convex Banach space and its dual space is isomorphic to
0 1 1
Lz (·) (Ω), where + = 1.
z(·) z 0 (·)
An crucial instrument for our findings is the following Hölder type inequality
Z  1 1 
u(x)v(x)dx ≤ + kukz(·) kvkz0 (·) ≤ 2kukz(·) kvkz0 (·) , (2.1)
Ω z− (z − )0
0
for all u ∈ Lz(·) (Ω) and v ∈ Lz (·) (Ω).
The modular function rz(·) is a fundamental element in the study of generalized
Lebesgue spaces. Furthermore, the following result is presented:
Proposition 2.1. (See [12, 24]). If u ∈ Lz(·) (Ω), we have
− +
(a) kukz(·) > 1 is true precisely when kukzz(·) < rz(·) (u) < kukzz(·) ,
+ −
(b) kukz(·) < 1 is true precisely when kukzz(·) < rz(·) (u) < kukzz(·) .
Multiplicity of weak solutions 2445

Provided that z1 , z2 ∈ C+(Ω) and z1 (x) ≤ z2 (x) for all x ∈ Ω, we can conclude
that the following continuous embedding holds :

Lz2 (·) (Ω) ,→ Lz1 (·) (Ω). (2.2)

The variable exponent Sobolev space is given by

W 1,z(·) (Ω) = u ∈ Lz(·) (Ω) : |∇u| ∈ Lz(·) (Ω) .



The norm for this space is given by the following expression

kukW 1,z(·) (Ω) = kuk1,z(·) = kukz(·) + k∇ukz(·) .

It is worth mentioning that the space (W 1,z(·) (Ω), k·k1,z(·) ) is a Banach space that is
also separable and reflexive. For additional details on this framework, refer to [12].
Remark 2.1. If z ∈ C+ (Ω) and N < z − , then the embedding W 1,z(·) (Ω) ,→,→
C 0 (Ω) is continuous and compact. Since W 1,z(·) (Ω) is continuously embedded in
−
W 1,z (Ω).
Then we can set by (1.3)

kuk∞
C1 = sup , (2.3)
u∈W 1,p(·) (Ω)\{0} kuk1,p(·)

and
kuk∞
C2 = sup . (2.4)
u∈W 1,q(·) (Ω)\{0} kuk1,q(·)

Now, we present the Musielak-Orlicz-Sobolev spaces which is employed in the

analysis of our main results.
We start by giving the definitions of the Orlicz function and Musielak function.
Definition 2.1. An Orlicz-type function, marked as A ∈ N (Ω), is a function A :
R −→ [0, +∞[ that is even, continuous, and convex, satisfies A(0) = 0 and 0 < A(t)
for all t > 0, and also satisfies:

A(t) A(t)
lim = 0, and lim = +∞.
t→0+ t t→+∞ t

A function A : Ω × R −→ [0, +∞[ is said to be a Musielak function, denoted by

A ∈ Φ(Ω), if for each t ≥ 0, A(·, t) ∈ M and for almost every x ∈ Ω, the function
A(x, ·) is an Orlicz function.
Let A ∈ Φ(Ω), the Musielak-Orlicz space LA (Ω) is defined by
 Z  |u(x)|  
LA (Ω) := u ∈ M, and there exist σ > 0 such that A x, dx < ∞ ,
Ω σ

having the following norm, recognized as the norm of Luxemburg, given by

 Z  |u(x)|  
kukLA (Ω) := inf σ > 0 : A x, dx ≤ 1 .
Ω σ
2446 A. Ahmed & M. S. B. Elemine Vall

The space W 1 LA (Ω) is given by the following definition:

n o
W 1 LA (Ω) := u ∈ LA (Ω) with |∇u| ∈ LA (Ω) ,

equipped with the following norm

kuk1,A = kukA + k∇ukA ,

where k∇ukA = k|∇u|kA .

Definition 2.2. 1. A function A ∈ Φ(Ω) is said to be fulfilling the ∆2 -condition
noted (A ∈ ∆2 ) when there exists a positive constant k > 0 and a non-negative
function b ∈ L1 (Ω) such that A(x, 2t) ≤ kA(x, t)+b(x) for all x ∈ Ω and t ∈ R.
2. A is said to be locally integrable if A(·, t0 ) ∈ L1 (Ω) for every t0 > 0.
The function A0d (x, t) denotes the right-hand derivative of A(x, ·) at t ≥ 0 and
is defined as
A(x, t + h) − A(x, t)
A0d (x, t) = lim+ .
h→0 h
Z |t|
If t < 0, we define A0d (x, t) = −A0d (x, −t). Thus, A(x, t) = A0d (x, s)ds for all
0
t ∈ R and x ∈ Ω.
Set A∗ : Ω × R −→ [0, +∞[ by
 
A∗ (x, s) = sup st − A(x, t) for each s ∈ R and x ∈ Ω.
t∈R

According to Young’s definition A∗ is known as the complementary function to A.

It is commonly understood that A∗ meets the criteria for a Musielak function and
A also acts as complementary function to A∗ .
For the fundamental properties of these spaces, we refer to [9].
We display here some facts that will be used later.
Lemma 2.1. ( See [9]). The following norms are equivalent on W 1 LA (Ω)

kuk1,A = kukA + k|∇u|kA ,

 
kuk2,A = max kukA , k|∇u|kA ,
 Z    |∇u(x)|  
|u(x)| 
kuk = inf σ > 0 : A x, + A x, dx ≤ 1 .
Ω σ σ

Lemma 2.2. (See [32]). Suppose A and A∗ are two complementary Musielak func-
tions satisfying the ∆2 -condition, then we have

tA0d (x, t)
1 < a∗ ≤ ≤ a∗ < ∞, for any x ∈ Ω, t > 0,
A(x, t)

and for some constants a∗ , a∗ .

Furthermore, we have
Z  
∗
1. If kuk ≤ 1, kuka ≤ A(x, |u(x)|) + A(x, |∇u(x)|) dx ≤ kuka∗ ,
Ω
Multiplicity of weak solutions 2447
Z  
∗
2. If kuk > 1, kuka∗ ≤ A(x, |u(x)|) + A(x, |∇u(x)|) dx ≤ kuka ,
Ω
1
3. If un 7−→ u in W LA (Ω), then
Z Z
[A(x, |un (x)|) + A(x, |∇un (x)|)] dx −→ [A(x, |u(x)|) + A(x, |∇u(x)|)] dx.
Ω Ω

Here and in the sequel, consider p, q ∈ C+ (Ω) two variables exponents satisfy
(1.3) and the Musielak function:

A(x, t) = tp(x) + tq(x) , for all (x, t) ∈ Ω × R∗+ . (2.5)

It is clear that A and its complementary function fulfill the ∆2 -condition, and set

W 1,p(·),q(·) (Ω) = W 1 LA (Ω), (2.6)

possessing the norm kuk1,p(·),q(·) = kuk1,p(·) + kuk1,q(·) .

Proposition 2.2. (See [15]). The space W 1,p(·),q(·) (Ω) embeds continuously into
W 1,m0 (Ω) and compactly into W 1,m0 (Ω) under the hypothesis (1.3), then the fol-
lowing embedding W 1,p(·),q(·) (Ω) ,→,→ C 0 (Ω) is compact and we put

kukL∞ (Ω)
C0 = sup . (2.7)
u∈W 1,p(·),q(·) (Ω)\{0} kuk1,p(·),q(·)

3. Main results
For u ∈ W 1,p(·),q(·) (Ω) we define the following functionals

|∇u|p(x) |∇u|q(x)
Z Z
Jp(x) (u) = dx, Jq(x) (u) = dx,
p(x) q(x)
ZΩ Ω
Z
λ(x) 1 λ(x) 1
Jp(x) (u) = λ(x) |u|p(x) dx, Jq(x) (u) = λ(x) |u|q(x) dx,
p(x) q(x)
Z Ω Z Ω

H(u) = F2 (x, u)dx, Φ1 (u) = − F1 (x, u)dx,

Ω Ω
a2
2 b2
2
J(u) = a1 Jp(x) (u) + Jp(x) (u) + b1 Jq(x) (u) + Jq(x) (u)
2 2
λ(x) λ(x)
+ Jp(x) (u) + Jq(x) (u), (3.1)

and
Φ2 (u) = J(u) − H(u), (3.2)
where Z t Z t
F1 (x, t) = f1 (x, ρ)dρ, and F2 (x, t) = f2 (x, ρ)dρ. (3.3)
0 0

Definition 3.1. A measurable function u ∈ W 1,p(·),q(·) (Ω) is a weak solution of the

Neumann elliptic problem (1.2) if for any v ∈ W 1,p(·),q(·) (Ω), one has
Z
|∇u|p(x)−2 ∇u∇vdx


a1 + a2 Jp(x) (u)
Ω
2448 A. Ahmed & M. S. B. Elemine Vall
Z
|∇u|q(x)−2 ∇u∇vdx


+ b1 + b2 Jq(x) (u)
Ω
Z  
+ λ(x) |u|p(x)−2 uv + |u|q(x)−2 uv dx
Z Ω Z
= f1 (x, u)vdx + f2 (x, u)vdx. (3.4)
Ω Ω

Then, it is easy to verify that the weak solutions u ∈ W 1,p(·),q(·) (Ω) of (1.2) are
exactly the critical points of Φ1 + Φ2 .
Definition 3.2. A function F1 (x, t) is said to be of type (S) if for all compact
subset of R noted E, there exists ς ∈ E such that
F1 (x, ς) = sup f1 (x, t) for almost every x ∈ Ω. (3.5)
t∈E

Lemma 3.1. Assume that (1.3) and (1.4) are satisfied. Then, Φ2 , Φ1 ∈
C 1 (W 1,p(·),q(·) (Ω), R) and their Gâteaux derivatives are given by
Z
hΦ02 (u), vi = a1 + a2 Jp(x) (u) |∇u|p(x)−2 ∇u∇vdx


Ω
Z
|∇u|q(x)−2 ∇u∇vdx


+ b1 + b2 Jq(x) (u)
Ω
Z   Z
+ λ(x) |u|p(x)−2 uv + |u|q(x)−2 uv dx − f2 (x, u)vdx,
Ω Ω

and Z
hΦ01 (u), vi = − f1 (x, u)vdx,
Ω

for any v, u ∈ W 1,p(·),q(·) (Ω).

Proof. We divided this prove into two claims, in the first we prove the Gâteaux
differetiability of J and the second focuses on the Gâteaux differentiability of H.
Claim 1. We start by proving that Jp(x) is of class C 1 (W 1,p(·),q(·) (Ω), R). Fix
|ζ|p(x)
x ∈ Ω. We define φ : RN → R by φ(ζ) = . It is clear that, φ ∈ C 1 (RN , R)
p(x)
and ∇φ(ζ) = |ζ|p(x)−2 ζ. Thus, for all ζ, ϑ ∈ RN , we have
φ(ζ + tϑ) − φ(ϑ)
lim = |ζ|p(x)−2 ζ · ϑ.
t→0 t
As a consequence, for u, v ∈ W 1,p(·),q(·) (Ω), we obtain
|∇u + t∇v|p(x) − |∇u|p(x)
 
1
lim = |∇u|p(x)−2 ∇u · ∇v. (3.6)
t→0 p(x) t
Applying the mean value theorem, there is a θ in the range 0 < |θ| < |t| such that,
for all t ∈ R with 0 < |t| < 1:
1 |∇u + t∇v|p(x) − |∇u|p(x)
p(x) t
(3.7)
=|∇u + θ∇v|p(x)−2 (∇u + θ∇v) · ∇v|
p(x)−1
≤ (|∇u| + |∇v|) |∇v|.
Multiplicity of weak solutions 2449

Since for u, v ∈ W 1,p(·),q(·) (Ω) one has (|∇u|+|∇v|)p(x)−1 |∇v| ∈ L1 (Ω). Using (3.6)
and (3.7) and applying the dominated convergence theorem, we can conclude that:

|∇u + t∇v|p(x) − |∇u|p(x)

Z   Z
1
lim dx = |∇u|p(x)−2 ∇u∇vdx.
t→0 Ω p(x) t Ω

It means that Jp(x) is Gâteaux differentiable and for u, v ∈ W 1,p(·),q(·) (Ω), we have
Z
0
hJp(x) (u), vi = |∇u|p(x)−2 ∇u∇vdx.
Ω

λ(x) λ(x)
By similar arguments, we can show that Jq(x) , Jp(x) and Jq(x) are Gâteaux differ-
entiables and for any v, u ∈ W 1,p(·),q(·) (Ω) we have
Z
0
hJq(x) (u), vi = |∇u|q(x)−2 ∇u∇vdx,
Ω
 0 Z
λ(x)
h Jp(x) (u), vi = λ(x)|u|p(x)−2 uvdx,
Ω
 0 Z
λ(x)
h Jq(x) (u), vi = λ(x)|u|q(x)−2 uvdx.
Ω

It follows that J is Gâteaux differentiable and for u, v ∈ W 1,p(·),q(·) (Ω), we obtain

Z
hJ 0 (u), vi = a1 + a2 Jp(x) (u) |∇u|p(x)−2 ∇u∇vdx


Ω
Z
|∇u|q(x)−2 ∇u∇vdx


+ b1 + b2 Jq(x) (u)
Ω
Z  
+ λ(x) |u|p(x)−2 uv + |u|q(x)−2 uv dx.
Ω

0
Next, we prove that Jp(x) : W 1,p(·),q(·) (Ω) −→ W 1,p(·),q(·) (Ω)∗ is continuous. To this
aim we take a sequence (un )n in W 1,p(·),q(·) (Ω)
Z such that un −→ u in W
1,p(·),q(·)
(Ω)
p(x)
as n −→ ∞. By Lemma 2.2, we have lim |∇un − ∇u dx = 0. Therefore,
n−→∞ Ω
after extracting a sub-sequence, we conclude that

lim ∇un = ∇u almost every where in Ω, (3.8)

n→∞

|∇un − ∇u|p(x) is dominated by h(x) in L1 (Ω). (3.9)

Since
p(x)
|∇un |p(x) ≤ (|∇u| + |∇un − ∇u|)
+
 
≤ 2p −1 |∇u|p(x) + |∇un − ∇u|p(x)
+
 
≤ 2p −1 |∇u|p(x) + h(x) . (3.10)

For any v ∈ W 1,p(·),q(·) (Ω) with kvk1,p(·),q(·) ≤ 1, the Hölder’s inequality gives
0 0
hJp(x) (un ) − Jp(x) (u), vi
2450 A. Ahmed & M. S. B. Elemine Vall
Z  
= |∇un |p(x)−2 ∇un − |∇u|p(x)−2 ∇u · ∇vdx
Ω

≤2 |∇un |p(x)−2 ∇un − |∇u|p(x)−2 ∇u| k∇vkLp(·) (Ω)

Lp0 (·) (Ω)

≤2 |∇un |p(x)−2 ∇un − |∇u|p(x)−2 ∇u| .

Lp0 (·) (Ω)

Hence,
0 0
kJp(x) (un ) − Jp(x) (u)k(W 1,p(·),q(·) (Ω))∗
(3.11)
≤2 |∇un |p(x)−2 ∇un − |∇u|p(x)−2 ∇u| .
Lp0 (·) (Ω)

It follows from (3.8) that

p0 (x)
|∇un |p(x)−2 ∇un − |∇u|p(x)−2 ∇u −→ 0 for a.e. x ∈ Ω.

Furthermore, using (3.10), we can deduce that

p0 (x) 0
 
|∇un |p(x)−2 ∇un − |∇u|p(x)−2 ∇u ≤ 2p (x)−1 |∇un |p(x) + |∇u|p(x)
0 +
 
≤ 2(p ) −1 |∇un |p(x) + |∇u|p(x)
0 + +
 
≤ 2(p ) +p −1 |∇u|p(x) + h(x) .

0 + +
 
Since 2(p ) +p −1 |∇u|p(x) + h is integrable over Ω, we can apply the dominated
convergence theorem to conclude that
p0 (x)
Z
|∇un |p(x)−2 ∇un − |∇u|p(x)−2 ∇u dx → 0 as n → ∞.
Ω

Using Lemma 2.2, we conclude that,

|∇un |p(x)−2 ∇un − |∇u|p(x)−2 ∇u| → 0 as n → ∞.

Lp0 (·) (Ω)

Combining this with (3.11), gives

0 0
kJp(x) (un ) − Jp(x) (u)k(W 1,p(·),q(·) (Ω))∗ −→ 0 as n −→ ∞.
 ∗
0
This completes the proof that Jp(x) : W 1,p(·),q(·) (Ω) −→ W 1,p(·),q(·) (Ω) is con-
1 1,p(·),q(·)
tinuous, and therefore Jp(x) ∈ C (W (Ω), R). By similar arguments we can
λ(x) λ(x)
show that Jq(x) , Jp(x)
and are of class C 1 from W 1,p(·),q(·) (Ω) into its dual.
Jq(x)
Which means that J is of class C 1 .
Claim 2. We shall prove that H ∈ C 1 (W 1,p(·),q(·) (Ω), R).
Let u, v ∈ W 1,p(·),q(·) (Ω) be arbitrary functions, then
Z
hH 0 (u), vi = f2 (x, u)vdx.
Ω
Multiplicity of weak solutions 2451

Applying the mean value theorem again, for u, v ∈ W 1,p(·),q(·) (Ω) and t ∈ R \ {0},
we obtain
F2 (x, u(x) + tv(x)) − F2 (x, u(x))
= v(x)f2 (x, u(x) + θv(x)),
t
for some θ ∈ R with 0 < |θ| < |t|. Therefore,

F2 (x, u(x) + tv(x)) − F2 (x, u(x))

−→ v(x)f2 (x, u(x))
t
as t −→ 0 for almost every x ∈ Ω. (3.12)

Using Proposition 2.2, we see that for |t| < 1 there exists ` = kukL∞ (Ω) +kvkL∞ (Ω) >
0 such that
F2 (x, u(x) + tv(x)) − F2 (x, u(x))
= |v(x)||f2 (x, u(x) + θv(x))|
t
(3.13)
≤ |v(x)| sup |f2 (x, s)|.
|s|≤`

From Hölder’s inequality and (1.4) we obtain

Z
v(x) sup |f2 (x, s)|dx ≤ 2kvkL∞ (Ω) sup |f2 (x, s)| .
Ω |s|≤` |s|≤` L1 (Ω)

Therefore, the dominated convergence theorem together with (3.12) and (3.13) im-
plies that

F2 (x, u(x) + tv(x)) − F2 (x, u(x))

Z Z
lim dx = f2 (x, u(x))v(x)dx.
t→0 Ω t Ω

That is, H admits a Gâteaux derivative and

Z
hH 0 (u), vi = f2 (x, u(x))v(x)dx.
Ω

For the proof of the continuity of H 0 in W 1,p(·),q(·) (Ω) we use Remark 2.2 for a
sub-sequence still denoted un to get un −→ u in C 0 (Ω). Consequently,

(un )n converges uniformly to u in Ω, (3.14)

k := sup kun kL∞ (Ω) < +∞. (3.15)
n∈N

We obtain that for any v ∈ W 1,p(·),q(·) (Ω) with kuk1,p(·),q(·) ≤ 1,

Z
|hH 0 (un ) − H 0 (u), vi| ≤ |f2 (x, un (x)) − f2 (x, u(x))||v(x)|dx. (3.16)
Ω

Therefore,
 
|f2 (x, un (x)) − f2 (x, u(x))| ≤ 2 |f2 (x, un )| + |f2 (x, u)|
 
≤ 2 sup |f2 (x, s)| + sup |f2 (x, s)|
|s|≤k |s|≤k
2452 A. Ahmed & M. S. B. Elemine Vall

≤ 4 sup |f2 (x, s)|.

|s|≤k

By (3.16) we have
Z
0 0
|hH (un ) − H (u), vi| ≤ 4 sup |f2 (x, s)||v(x)|dx.
Ω |s|≤k

Note that sup |f2 (x, s)| ∈ L1 (Ω). Then, the dominated convergence theorem and
|s|≤k
(3.14), conclude that
Z
lim |f2 (x, un (x)) − f2 (x, u(x))||v(x)|dx = 0.
n→∞ Ω

Hence, from (3.16) follows that

lim kH 0 (un ) − H 0 (u)kW 1,p(·),q(·) (Ω)∗ = 0.

n→∞

This completes the proof that H 0 : W 1,p(·),q(·) (Ω) −→ W 1,p(·),q(·) (Ω)∗ is continuous,
and therefore H ∈ C 1 (W 1,p(·),q(·) (Ω), R).
Similarly as above we are able to illustrate that Φ1 ∈ C 1 (W 1,p(·),q(·) (Ω), R), and
since Φ2 = J − H, the proof is complete.
Lemma 3.2. Assume that (1.3) and (1.4) hold. Then Φ1 , Φ2 are sequentially
weakly lower semi-continuous.
Proof. We divided this prove into two claims, the first concerns the functional J
and the second focuses on the functional H.
Claim 1. For any u ∈ W 1,p(·),q(·) (Ω) we have

a2
2 b2
2 λ(x) λ(x)
J(u) = a1 Jp(x) (u)+ Jp(x) (u) +b1 Jq(x) (u)+ Jq(x) (u) +Jp(x) (u)+Jq(x) (u).
2 2
Consider a sequence (un )n such that (un )n goes to u weakly in W 1,p(·),q(·) (Ω). Then,
by the convexity of Jp(x) , we have
0
Jp(x) (u) ≤ Jp(x) (un ) + hJp(x) (u), u − un i.

When n goes to infinity, the aforementioned inequality, we can find that Jp(x) is
sequentially weakly lower semi-continuous. Its follows
Z Z
1 1
a1 |∇u|p(x) dx ≤ lim inf a1 |∇un |p(x) dx. (3.17)
Ω p(x) n→+∞ Ω p(x)

Similarly, we get
Z 2 Z 2
a2 1 p(x) a2 1 p(x)
|∇u| dx ≤ lim inf |∇un | dx , (3.18)
2 p(x) n→+∞ 2 Ω p(x)
Z Ω Z
1 1
b1 |∇u|q(x) dx ≤ lim inf b1 |∇un |q(x) dx, (3.19)
Ω q(x) n→+∞ Ω q(x)
Z 2 Z 2
b1 1 b1 1
|∇u|q(x) dx ≤ lim inf |∇un |q(x) dx , (3.20)
2 Ω q(x) n→+∞ 2 Ω q(x)
Multiplicity of weak solutions 2453
Z Z
1 1
λ(x) |u|p(x) dx ≤ lim inf λ(x) |un |p(x) dx, (3.21)
Ω p(x) n→+∞ Ω p(x)
and Z Z
1 1
λ(x) |u|q(x) dx ≤ lim inf λ(x) |un |q(x) dx. (3.22)
Ω q(x) n→+∞ Ω q(x)
Which yields
lim inf J(un )
n→+∞
" Z Z 2
1 a2 1
= lim inf a1 |∇un |p(x) dx + |∇un |p(x) dx
n→+∞ Ω p(x) 2 Ω p(x)
Z Z 2
1 q(x) b2 1 q(x)
+ b1 |∇un | dx + |∇un | dx
Ω q(x) 2 Ω q(x)
Z Z 
1 p(x) 1 q(x)
+ λ(x) |un | dx + λ(x) |un | dx
p(x) q(x)
" Ω Z
Ω
Z 2
1 a2 1
≥ lim inf a1 |∇un |p(x) dx + lim inf |∇un |p(x) dx
n→+∞ Ω p(x) n→+∞ 2 Ω p(x)
Z Z 2
1 q(x) b2 1 q(x)
+ lim inf b1 |∇un | dx + lim inf |∇un | dx
n→+∞ Ω q(x) n→+∞ 2 Ω q(x)
Z Z 
1 p(x) 1 q(x)
+ lim inf λ(x) |un | dx + lim inf λ(x) |un | dx
n→+∞ Ω p(x) n→+∞ Ω q(x)
a2
2
≥ a1 Jp(x) (u) + Jp(x) (u) + b1 Jq(x) (u)
2
b2
2 λ(x) λ(x)
+ Jq(x) (u) + Jp(x) (u) + Jq(x) (u)
2
≥ J(u), (3.23)
which means J is sequentially weakly lower semi-continuous.
Claim 2. Proving that H is weakly-lower semi-continuous.
Proposition 2.2 implies that there exists a sub-sequence of (un )n converging to
u uniformly on compact subsets of Ω. Then,
(un )n converges to u uniformly in Ω,

k := sup kun kL∞ (Ω) < +∞.

n∈N

Thus, almost everywhere in Ω, we have lim F2 (x, un (x)) = F2 (x, u(x)) and
n→∞
|F2 (x, un (x))| ≤ k sup |f2 (x, s)| for all n. Since sup |f2 (x, s)| ∈ L1 (Ω) by (1.4).
|s|≤k |s|≤k
Thus, the dominated convergence theorem gives that lim H(un ) = H(u). The
n→∞
weak semi-continuity of the functional H implies that Φ2 is sequentially weak with
lower semi-continuity, which means that Φ1 is sequentially weakly continuous.
We will now demonstrate that Φ2 is coercive. For brevity, we will write ci for
some positive constant throughout.
Proposition 3.1. Assume that G satisfies exactly one of the following two condi-
tions:
2454 A. Ahmed & M. S. B. Elemine Vall

p+ min(a1 , b1 , λ0 )
1. There exist τ > 0, 0 < ε < and θ1 , θ2 , θ3 ∈ L1 (Ω) with
2p− −1 q +
θ1 6= 0 and θ2 6= 0 such that
!
θ1 (x) p− θ2 (x) q−
|F2 (x, t)| ≤ ε − |t| + − |t| + θ3 (x),
p+ C1p kθ1 kL1 (Ω) q + C2q kθ2 kL1 (Ω)
(3.24)
for almost all x ∈ Ω and all t ≥ τ .
p− λ0
2. There exist τ > 0, 0 < ε < + and θ4 ∈ L1 (Ω) such that
q kλkL∞ (Ω)
 
λ(x) p(x) λ(x) q(x)
|F2 (x, t)| ≤ ε |t| + |t| + θ4 (x), (3.25)
p(x) q(x)

for almost all x ∈ Ω and all t ≥ τ .

Then, Φ2 is coercive.
Proof. Suppose (3.24) holds. Then, without loss of generality, we have

a2
2 b2
2
Φ2 (u) = a1 Jp(x) (u) + Jp(x) (u) + b1 Jq(x) (u) + Jq(x) (u)
2 Z 2
λ(x) λ(·)
+Jp(x) (u) + Jq(x) (u) − F2 (x, u)dx
Ω
Z Z
1 1
≥ a1 |∇u|p(x) dx + b1 |∇u|q(x) dx
Ω p(x) Ω q(x)
Z  
1 1
+ λ(x) |u|p(x) + |u|q(x) dx
Ω p(x) q(x)
Z " − −
#
εθ1 (x)|u|p εθ2 (x)|u|q
− p−
+ − + θ3 (x) dx
Ω p+ C1 kθ1 kL1 (Ω) q + C2q kθ2 kL1 (Ω)
Z Z 
min(a1 , b1 ) p(x) q(x)
≥ |∇u| dx + |∇u| dx
q+ Ω Ω
Z 
λ0 
+ + |u|p(x) + |u|q(x) dx
q
 Ω − −

εkukpL∞ (Ω) εkukqL∞ (Ω)
− kθ3 kL1 (Ω) + − + −

p+ C1p q + C2q
min(a1 , b1 )  − −
 λ 
0 − −

≥ +
k∇ukpp(·) + k∇ukqq(·) − 2 + + kukpp(·) + kukqq(·) − 2
q q
p− q−
 
εkuk1,p(·) εkuk1,q(·)
− kθ3 kL1 (Ω) + +
+ 
p q+

min(a1 , b1 , λ0 ) − min(a1 , b1 , λ0 ) −
≥ p− −1 + kukp1,p(·) + q − −1 + kukq1,q(·)
2 q 2 q
p− q−
 
kuk1,p(·) kuk1,q(·)
−ε  +  − c1
p+ q+
Multiplicity of weak solutions 2455
   
min(a1 , b1 , λ0 ) ε p− min(a1 , b1 , λ0 ) ε −
≥ p − −1 + − +
kuk 1,p(·) + q − −1 + − +
kukq1,q(·) − c1
2 q p 2 q q
p− q−
 
≥ c2 kuk1,p(·) + kuk1,q(·) − c1
− −
 
≥ c2 kukp1,p(·) + kukp1,q(·) − 1 − c3
c2 −
≥ p− −1 kukp1,p(·),q(·) − c3 . (3.26)
2
Under assumption (3.25) by using similar arguments as above, we get
min(a1 , b1 )  − −
 λ 
0 − −

Φ2 (u) ≥ +
k∇ukpp(·) + k∇ukqq(·) − 2 + + kukpp(·) + kukqq(·) − 2
q q
Z  
ελ(x) p(x) ελ(x) q(x)
− |u| + |u| + θ4 (x) dx
Ω p(x) q(x)
min(a1 , b1 )  p −
q−
 λ 
0 p− q−

≥ k∇uk p(·) + k∇uk q(·) − 2 + kuk p(·) + kukq(·) − 2
q+ q+
kλkL∞ (Ω)  − −

−ε −
kukpp(·) + kukqq(·) − 2 − c4
p
min(a1 , b1 )  p− q−

≥ k∇uk p(·) + k∇uk q(·)
q+
kλkL∞ (Ω)
 
λ0 p− q−

+ − ε kuk p(·) + kuk q(·) − c5
q+ p−
kλkL∞ (Ω)
  
min(a1 , b1 ) λ0
≥ min , −ε
q + q + p−
− − − −
 
k∇ukpp(·) + kukpp(·) + k∇ukqq(·) + kukqq(·) − c5
c6  − −

≥ q− −1 kukp1,p(·) + kukq1,q(·) − c5
2
c6  − −

≥ q− −1 kukp1,p(·) + kukp1,q(·) − c7
2
c6 −
≥ q− −1 kukp1,p(·),q(·) − c7 .
2 × 2p− −1
(3.27)
Thanks to (3.26)-(3.27) and as a result of the coercivity of Φ2 , it follows that there
are constants α1 and α2 satisfy:
−
Φ2 (u) ≥ α1 kukp1,p(·),q(·) for any kuk1,p(·),q(·) ≥ α2 . (3.28)

Lemma 3.3. Assume the hypothesis (1.3) and one of the assumptions (3.24) and
(3.25). Hence, there are positives constants δ1 , δ2 and δ3 such that
 p(x)
|ς|q(x)

|ς|
Z Z
+
λ(x) + dx − f2 (x, ς)dx ≤ δ1 |ς|q + δ2 , for any ς ∈ R.
Ω p(x) q(x) Ω
(3.29)
Proof. Under the assumptions (1.3) and (3.24) implies
 p(x)
|ς|q(x)

|ς|
Z Z
λ(x) + dx − F2 (x, ς)dx
Ω p(x) q(x) Ω
2456 A. Ahmed & M. S. B. Elemine Vall
  Z
1 p+ 1 + θ1 (x) +
≤ kλkL∞ (Ω) |Ω| −
|ς| + − |ς|q +ε p − |ς|p dx
p q Ω p− C1 kθ1 kL1 (Ω)
Z Z
θ2 (x) q+
+ε − |ς| dx + θ3 (x)dx
Ωq − C2q kθ2 kL1 (Ω) Ω
  Z
1 q+ 1 q+ ε +
≤ kλkL∞ (Ω) |Ω| |ς| + − |ς| + p−
|ς|q dx
p− p −
p C1 Ω
Z
ε +
+ − |ς|q dx + kθ3 kL1 (Ω)
p− C2q Ω
!!
2kλkL∞ (Ω) |Ω| ε|Ω| 1 1 +
≤ −
+ − p − + q − |ς|q + kθ3 kL1 (Ω) ,
p p C1 C2
for sufficiently large ς.
!!
2kλkL∞ (Ω) |Ω| ε|Ω| 1 1
Then, we have (3.29) for δ1 = + − − + − and δ2 =
p− p C1p C2q
kθ3 kL1 (Ω) .
On the other hand, assumptions (1.3) and (3.25) implies
Z   Z
1 1
λ(x) |ς|p(x) + |ς|q(x) dx − F2 (x, ς)dx
Ω p(x) q(x) Ω
 
1 q+ 1 q+
≤ kλkL∞ (Ω) |Ω| |ς| + |ς|
p− p−
Z   Z
1 1
+ε λ(x) |t|p(x) + |t|q(x) dx + θ4 (x)dx
Ω p(x) q(x) Ω
2kλkL∞ (Ω) |Ω| q+ 2εkλkL∞ (Ω) |Ω| q+
≤ |ς| + |ς| + kθ4 kL1 (Ω)
p− p−
2(1 + ε)kλkL∞ (Ω) |Ω| q+
≤ |ς| + kθ4 kL1 (Ω) , for ς large enough.
p−

2(1 + ε)kλkL∞ (Ω) |Ω|

Therefore, we establish (3.29) with δ1 and δ2 being and
p−
kθ4 kL1 (Ω) , respectively, for ς sufficiently large.
The following theorem is our first main result.
Theorem 3.1. Assuming that (1.3) and (1.4) are satisfied, and either (3.24) or
(3.25) hold, and that F satisfies condition (3.5). Additionally, we assume that F1
satisfies the following condition
Z   Z  
1 p(x) 1 q(x)
λ(x) |ς| + |ς| dx − F2 (x, ς) + F1 (x, ς) dx = −∞. (3.30)
Ω p(x) q(x) Ω

Moreover, there exist positive sequences (an )n and (bn )n such that
+
aqn
lim bn = +∞, lim p−
= 0. (3.31)
n→∞ n→∞ bn

Finally, we assume the existence of a positive integrable function with khkL1 (Ω) = 1
Multiplicity of weak solutions 2457

and some positive constants δ1 , δ2 , δ3 > 0 such that

  p− !
bn +
F1 (x, an ) + h(x) α1 − δ1 |an |q − δ2 ≥ sup F1 (x, t), (3.32)
C0 t∈[an ,bn ]
  p− !
bn +
F1 (x, −an ) + h(x) α1 − δ1 |an |q − δ2 ≥ sup F1 (x, t), (3.33)
C0 t∈[−bn ,−an ]

for any n we have for almost all x in Ω, where α1 is the !!coercivity constant defined
2kλkL∞ (Ω) |Ω| ε|Ω| 1 1
in (3.28), δ1 = + − − + − and δ2 = kθ3 kL1 (Ω) if we
p− p C1p C2q
2(1 + ε)kλkL∞ (Ω) |Ω|
assume (3.24) and δ1 = and δ2 = kθ4 kL1 (Ω) if we assume
p−
(3.25). The last inequalities (3.32) and (3.33) are strict on a non-negligible subset
of Ω.
Then, we can construct a sequence (vn )n of local minima of Φ1 + Φ2 such that
lim Φ2 (vn ) = ∞. Consequently, the problem (1.2) admits an unbounded sequence
n→∞
of weak solutions.
Proof. For r > inf Φ2 (u), we define
u∈W 1,p(·),q(·) (Ω)
n o
Θ(r) = inf κ > 0 : Φ−1
1 (] − ∞, r[) ⊂ B(0, κ) , (3.34)

where B(0, κ) denoted the ball centered at 0 with radius κ in W 1,p(·),q(·) (Ω) with
respect to the norm k · k1,p(·),q(·) , and B(0, κ) denote its closure in W 1,p(·),q(·) (Ω).
Since Φ2 is coercive, we have Θ(r) ∈]0, +∞[ for all r > inf Φ2 (u). Owing
u∈W 1,p(·),q(·) (Ω)
of (3.28), we have
−
if Φ2 (u) < α1 kukp1,p(·),q(·) , then kuk1,p(·),q(·) < α2 .

With the help of (3.34)), one may observe that Φ−1 2 (] − ∞, r[) ⊂ B(0, Θ(r)) yields
(Φ−1
2 (] − ∞, r[)) w ⊂ B(0, Θ(r)). By using (2.7), we get,

B(0, Θ(r)) ⊂ u ∈ C(Ω) : kukL∞ (Ω) ≤ C0 Θ(r) ,
which yields
inf Φ1 (v) ≥ inf Φ1 (v) ≥ inf Φ1 (v). (3.35)
v∈(Φ−1 kvk1,p(·),q(·) ≤Θ(r) kvkL∞ (Ω) ≤C0 Θ(r)
2 (]−∞,r[))w

−
Suppose κ ≥ α1 α2p and let u ∈ W 1,p(·),q(·) (Ω) satisfy Φ2 (u) < κ. If kuk1,p(·),q(·) ≥
α2 , then by (3.28), we have
−
κ > Φ2 (u) ≥ α1 kukp1,p(·),q(·) ,
  1
κ p−
this shows that kuk1,p(·),q(·) ≤ . If kuk1,p(·),q(·) < α2 , it is easy to see that
α1
  1−
κ p
kuk1,p(·),q(·) ≤ . By the definition of Θ(κ), we have
α1
  1−
κ p
Θ(κ) ≤ . (3.36)
α1
2458 A. Ahmed & M. S. B. Elemine Vall

Since F1 (x, ·) satisfies condition (3.5), for each n, there exists ςn ∈ [−an , an ] such
that
F1 (x, ςn ) = sup F1 (x, t) for almost all x in Ω. (3.37)
t∈[−an ,an ]

In order to satisfy (b) of Theorem 1.1, we consider un as the constant function with
  p−
bn
value ςn and rn = α1 , which leads to lim rn → +∞. Using (3.36), we
C0 n→∞
conclude that
bn
Θ(rn ) ≤ then C0 Θ(rn ) ≤ bn . (3.38)
C0

By (3.29), one has

Z   Z
1 1
mn = λ(x) |ς|p(x) + |ς|q(x) dx − F2 (x, ς)dx
Ω p(x) q(x) Ω
+
≤ δ1 |ς|q + δ2
+
≤ δ1 |an |q + δ2 .

For n large enough (3.31) can be write

  p−
q+ bn
δ1 |an | + δ2 < α1 = rn ,
C0

consequently we find mn < rn which means (1.8) holds. Without loss of generality,
we can assume that (1.8) holds for all n.
From (3.32)-(3.33) and (3.37), we may find the following inequality

F1 (x, ςn ) + h(x)(rn − mn ) ≥ sup F1 (x, t) a.e. in Ω, (3.39)

|t|≤bn

which is strict on a non-negligible subdomain of Ω. Using (3.38) and (3.39), we

obtain (1.9) and (1.10) follows directly from (3.30).
Then, hypotheses of Theorem 1.1 (b) hold true which completes the proof of
Theorem 3.1.
Now, we present an example to illustrate the results cited in the previous theo-
rem.
5 1
Corollary 3.1. Let Ω =]0, 1[2 then N = 2 and let p(x) = + | sin(x + y)|,
2 4
1 5 11 − 7
and q(x) = 3 + | cos(x + y)|, then p− = , p+ = , q = 3 and q + = .
2 2 4 ! 2
13 p− θ1 (x) p− −1 q − θ2 (x) q− −1
Consider the functions f2 (x, t) = |t| + − |t| where
100 p+ C1p− q + C2q
θ1 , θ2 ∈ L1 (Ω) are positives functions with kθ1 kL1 (Ω) = kθ2 kL1 (Ω) = 1, and consider
f1 (x, t) ≡ α(x)g(t), with a positive function α ∈ L1 (Ω) such that kαkL1 (Ω) = 1
and a continuous function g such that g(t) = G0 (t) and G(−t) = G(t). Then, the
Multiplicity of weak solutions 2459

following nonlinear elliptic double phase Kirchhoff-type problem

 Z !
 1 ( 5
+ 1
| sin(x+y)|)
− 1+2
|∇u| 2 4 dx ∆( 5 + 1 | sin(x+y)|) u


5 1
Ω 2 + 4 | sin(x + y)|


 2 4

 !
 Z
 1

|∇u|(3+ 2 | cos(x+y)|) dx ∆(3+ 1 | cos(x+y)|) u
 1
− 1+2


1
|




 Ω 3 + 2 cos(x + y)| 2

2 2
2+x +y 
|u|( 2 + 4 | sin(x+y)|) u + |u|(1+ 2 | cos(x+y)|) u
1 1 1
+ 2 + y2


 1 + x !



 13 10 θ 1 (x) 3 6θ 2 (x) 2

 = α(x)g(t) + |t| 2 + |t| in Ω,
100 5
7C23




 11C 1
2

 ∂u = 0 on ∂Ω,



∂ν
(3.40)
has a sequence of weak solutions (un )n in W 1,p(·),q(·) (Ω) with unbounded norm.
13 p+ min(a1 , b1 , λ0 )
Proof. It is clear that 0 < ε = < = 0.27, then (3.24) is
100 2p− 1 q +
satisfied. Furthermore, we have F1 (x, t) = α(x)G(t), and we can choose two positive
5
sequences (an )n and (bn )n such that a1 ≥ 1, bn2 = 2n2 a3n and an+1 > bn for every
n. Define G(an ) = a4n and G(bn ) such that
  b  52 
n 3
G(an ) < G(bn ) < α1 − δ1 |an | − δ2 + G(an ), (3.41)
C0
!
4 13 1 1
where δ1 = |Ω| + |Ω| 5 + and δ2 = kθ3 kL1 (Ω) .
5 250 C12 C23
 b  p−
n
Put rn = α1 and ςn = an .
C0
Since
Z Z Z Z
λ(x) λ(x)
|an |p(x) dx + |an |q(x) dx − F2 (x, an )dx − F1 (x, an )dx
Ω p(x) Ω q(x) Ω Ω
+ +
≤δ1 |an |q + δ2 − kαkL1 (Ω) aqn +1
→ −∞,

as n → ∞, then the conditions (3.30)-(3.31) hold true. Taking h(x) = α(x), then
(3.53) implies (3.32)-(3.33).
As a result, all the assumptions of Theorem 3.1 are satisfied, then problem (3.40)
has a sequence of weak solutions (un )n in W 1,p(·),q(·) (Ω) with unbounded norm.
The second main result is cited in the following theorem.
Theorem 3.2. Assume the assumptions (1.3) and (1.4) and the following hypoth-
esis:

F2 (x, t) is non-positive for almost every x ∈ Ω and for all t ∈ R. (3.42)

There exists δ,  > 0 such that

−
− F2 (x, t) ≤ δ|t|p for almost every x ∈ Ω and for |t| ≤ . (3.43)
2460 A. Ahmed & M. S. B. Elemine Vall

In addition, assume that the functional F1 satisfies the condition (3.5) and
Z Z
F1 (x, ς)dx + F2 (x, ς)dx Z  
1 p(x) 1 q(x)
lim sup Ω Ω
> λ(x) |ς| + |ς| dx.
|ς|→0 |ς|p− Ω p(x) q(x)
(3.44)
Assume that (an )n and (bn )n are positive sequences satisfying
−
apn
lim bn = 0 and lim + = 0. (3.45)
n→∞ n→∞ bqn

Additionally, there exists a positive function h ∈ L1 (Ω), where khkL1 (Ω) = 1, such
that for every n and all x in Ω, we obtain
 q + !
bn −
F1 (x, an ) + h(x) δ3 − δ4 apn ≥ sup F1 (x, t), (3.46)
C0 t∈[an ,bn ]
 q + !
bn p−
F1 (x, −an ) + h(x) δ3 − δ4 an ≥ sup F1 (x, t), (3.47)
C0 t∈[−bn ,−an ]

 
min(a1 , b1 , λ0 ) 2kλkL∞ (Ω)
with δ3 = and δ4 = + δ |Ω|, the relations (3.46) and
22q+ −2 q + p−
(3.47) are strict on a non negligible subdomain of Ω. Accordingly we can con-
struct a sequence (vn )n of separate local minima of Φ1 + Φ2 where vn tends to 0
in W 1,p(·),q(·) (Ω). Hence, a sequence of non-zero weak solutions to problem (1.2)
exists and converges strongly to 0 in W 1,p(·),q(·) (Ω).
Proof. We now proceed to demonstrate that Theorem 1.1 (c) holds by verifying
all the assumptions. Taking into account the inequality (3.42), for kuk1,p(·),q(·) ≤ 1
one has
Z
Φ2 (u) = J(u) − F2 (x, u)dx
Ω

a2
2 b2
2
≥ a1 Jp(x) (u) + Jp(x) (u) + b1 Jq(x) (u) + Jq(x) (u)
2 2
λ(x) λ(x)
+ Jp(x) (u) + Jq(x) (u)

min(a1 , b1 )  p+ q+
 λ 
0 p+ q+

≥ k∇ukp(·) + k∇ukq(·) + kuk p(·) + kuk q(·)
q+ q+
min(a1 , b1 , λ0 ) + min(a1 , b1 , λ0 ) +
≥ kukq1,p(·) + kukq1,q(·)
2q+ −1 q + 2q+ −1 q +
min(a1 , b1 , λ0 ) +
≥ 2q + −2 + kukq1,p(·),q(·)
2 q
+
≥ δ3 kukq1,p(·),q(·) ,

min(a1 , b1 , λ0 )
with δ3 = . Consequently, Φ2 is coercive, inf Φ2 = Φ2 (0) =
22q+ −2 q + W 1,p(·),q(·) (Ω)
Multiplicity of weak solutions 2461

0 and 0 is the exclusive global minimizer of Φ2 . Owing to (3.44) one has


lim sup Φ2 (ς) + Φ1 (ς)
|ς|→0
(Z )
|ς|p(x) |ς|q(x)
  Z Z
= lim sup λ(x) + dx − F1 (x, ς)dx − F2 (x, ς) dx
|ς|→0 Ω p(x) q(x) Ω Ω
(Z − −
! )
|ς|p |ς|p
Z Z
≤ lim sup λ(x) + dx − F1 (x, ς)dx − F2 (x, ς) dx
|ς|→0 Ω p(x) q(x) Ω Ω

<0,
that is, at 0, Φ1 + Φ2 does not reach a local minimum, as a result (1.11) is fulfilled.
Let r be small enough so that if Φ2 (u) < r, then kuk1,p(·),q(·) is no greater than
  1+   1+
r q r q
. It follows that Θ(r) is less than or equal to .
δ3 δ3
 q +
bn
Now put rn = δ3 If we set u0 and un to be the constant functions ς0
C0
and ςn , respectively, as prescribed by Theorem 1.1, then
C0 Θ(rn ) ≤ bn . (3.48)
The inequalities (3.43)) and (1.3) affirm the existence of a sequence (ςn )n ⊂ R such
that ςn ∈ [−an , an ] such that for any an small enough,
|ςn |p(x) |ςn |q(x)
Z   Z
mn = λ(x) + dx − F2 (x, ςn )dx
Ω p(x) q(x) Ω
 
1 p− 1 p− −
≤ kλkL∞ (Ω) −
|ςn | + − |ςn | + δ|Ω||ςn |p
p p
 
2kλkL∞ (Ω) −
≤ + δ |Ω||ςn |p
p−
−
≤ δ4 |an |p , (3.49)
 
2kλkL∞ (Ω)
where δ4 = + δ |Ω|.
p−
If we choose n to be sufficiently large, then it follows from (3.45) that
 q +
p− bn
δ4 |an | < δ3 = rn .
C0
Then (1.8) is obtained.
Because F1 (x, ·) satisfies condition (3.5), for any n, we can find ςn ∈ [−an , an ]
with
F1 (x, ςn ) = sup F1 (x, t) a.e. in Ω. (3.50)
t∈[−an ,an ]

Then, thanks to (3.46) and (3.47) we find the following inequality

sup F1 (x, t) ≤ F1 (x, ςn ) + h(x)(rn − mn ) a.e. in Ω, (3.51)
|t|≤bn
2462 A. Ahmed & M. S. B. Elemine Vall

which is strict on a non negligible subset of Ω. Then, the inequality (1.9) acquires
immediately from (3.48) and (3.51). Consequently, Theorem 1.1 (c) holds, since the
necessary hypotheses have been fulfilled.
As a result, we can conclude that a sequence (vn )n of separate local minima of
Φ1 +Φ2 exists, and satisfies lim Φ2 (vn ) = 0. This means that lim kvn k1,p(·),q(·)
n7→+∞ n7→+∞
= 0, thereby completing the proof.
We introduce now an example to illustrate the results cited in the second theo-
rem.
( N
)
X
Corollary 3.2. For N ≥ 1 we set Ω = x ∈ RN : x2i < 1 and let p(x) =
i=1
N
1X 2 1
N +1+ x , and q(x) = N + 2 + ! , then p− = N + 1,
2 i=1 i N
X
2N 1+ x2i
i=1
3 1
p = N + , q − = N + 2 and q + = N + 2 +
+
. Consider the functions F2 (x, t) =
2  2N 
N
! N
!
X X
x2i x2i
 
 2 + sin 2 + sin 
N +1 
i=1 p(x) i=1

q(x) 
− |t| + |t|

2N (N + 2) + 1 
 N 1 
N +2+ N
! 
 N + 1 + 1 X x2 X 
2 i 2
2N 1 + x
 
i=1 i
i=1
and let f1 (x, t) ≡ h(x)g(t), with h ∈ L1 (Ω) be a positive function such that
khkL1 (Ω) = 1 and g a continuous function with g(t) = G0 (t) and let (an )n and
− 1 +
(bn )n are two positives sequences with apn = bqn and bn+1 < an , with G(0) = 0,
−
n
G(an ) = apn +1 and

  b q + −

n
G(an ) < G(bn ) < δ3 − δ4 |an |p + G(an ). (3.52)
C0

Then, the following nonlinear Kirchhoff problem

  Z   Z 
1 1 1
− +3 |∇u|p(x) dx ∆p(·) u − 1 + |∇u|q(x) dx ∆q(·) u


2Ω p(x) Ω q(x)




 !!
N


 X  
x2i |u|p(x)−2 u + |u|q(x)−2 u



 + 2 + sin
i=1
 N
!!
 N +1 X  
x2i |u|p(x)−2 u + |u|q(x)−2 u


 = h(x)f1 (t) − 2 + sin
2N (N + 2) + 1


i=1



 in Ω, ∂u = 0 on ∂Ω,



∂ν
(3.53)
has weak solutions formed by a sequence (un )n in W 1,p(·),q(·) (Ω) such that

lim kun k1,p(·),q(·) = ∞.

n→∞
Multiplicity of weak solutions 2463

1
 b N +2+ 2N
n
Proof. Put rn = δ3 and ςn = an , one has
C0
Z Z
F1 (x, an )dx +
F2 (x, an )dx
Ω Ω! !
Z N 
X
2 1 1 +
− 2 + sin xi + dx|an |q
Ω i=1
p(x) q(x)
p− +1 p− λ|Ω| −
> |an | − |an | − + |an |p −→ 0 as n → ∞,
q

therefore (3.42)-(3.44) hold true. Using (3.52) we obtain the conditions (3.46)-
(3.47).
Therefore, we can conclude that the assumptions required by Theorem 3.2 are
satisfied, thus completing the proof.

References
[1] G. A. Afrouzi, S. Heidarkhani and S. Shokooh, Infinitely many solutions for
Steklov problems associated to non-homogeneous differential operators through
Orlicz-Sobolev spaces, Complex Var. Elliptic Equ., 2015, 60(11), 1505–1521.
[2] A. Ahmed and M. S. B. Elemine Vall, Perturbed nonlinear elliptic Neumann
problem involving anisotropic Sobolev spaces with variable exponents, Mathe-
matiche (Catania), 2022, 77(2), 465–486.
[3] V. Ambrosio and T. Isernia, A multiplicity result for a fractional Kirchhoff
equation in RN with a general nonlinearity, Commun. Contemp. Math., 2018
20(5), 1750054, 17 pp.
[4] V. Ambrosio and T. Isernia, A Multiplicity result for a (p, q)-Schrödinger-
Kirchhoff type equation, Ann. Mat. Pura Appl. (4), 2022, 201(2), 943–984.
[5] V. Ambrosio and T. Isernia, Concentration phenomena for a fractional
Schrödinger-Kirchhoff type equation, Math. Methods Appl. Sci., 2018, 41(2),
615–645.
[6] V. Ambrosio and V. D. Rădulescu, Fractional double-phase patterns: Con-
centration and multiplicity of solutions, J. Math. Pures Appl., 2020, 142(9),
101–145.
[7] A. Arosio and S. Panizzi, On the well-posedness of the Kirchhoff string, Trans.
Amer. Math. Soc., 1996, 348(1), 305–330.
[8] M. Cencelj, V. D. Rădulescu and D. D. Repovs̆, Double phase problems with
variable growth, Nonlinear Anal., 2018, 177, 270–287.
[9] I. Chlebicka, P. Gwiazda, A. Świerczewska-Gwiazda and A. Wróblewska-
Kamińska, Partial Differential Equations in Anisotropic Musielak-Orlicz
Spaces, Springer, 2021.
[10] F. Colasuonno and M. Squassina, Eigenvalues for double phase variational
integrals, Ann. Mat. Pura Appl. (4), 2016, 195(6), 1917–1959.
[11] A. Crespo-Blanco, L. Gasiǹski, P. Harjulehto and P. Winkert, A new class of
double phase variable exponent problems: Existence and uniqueness, J. Differ-
ential Equations, 2022, 323, 182–228.
2464 A. Ahmed & M. S. B. Elemine Vall

[12] L. Diening, P. Harjulehto, P. Hästö and M. Růz̆ic̆ka, Lebesgue and Sobolev

spaces with variable exponents, Lecture Notes in Math., 2017, 2011.
[13] N. C. Eddine and Dus̆an D. Repovs̆, The Neumann problem for a class of gen-
eralized Kirchhoff-type potential systems, Bound. Value Probl., 2023, 2023(19).
DOI: 10.1186/s13661-023-01705-6.
[14] M. El Ouaarabi, C. Allalou and S. Melliani, Weak solutions for double phase
problem driven by the (p(x), q(x))-Laplacian operator under Dirichlet boundary
conditions, Bol. Soc. Parana. Mat., 2023, 41, 1–14.
[15] M. S. B. Elemine Vall and A. Ahmed, Infinitely many weak solutions for per-
turbed nonlinear elliptic Neumann problem in Musielak-Orlicz-Sobolev frame-
work, Acta Sci. Math. (Szeged), 2020, 86(3–4), 601–616.
[16] M. S. B. Elemine Vall, A. Ahmed, A. Touzani and A. Benkirane, Existence of
entropy solutions for nonlinear elliptic equations in Musielak framework with
L1 data, Bol. Soc. Parana. Mat. (3), 2018, 36(1), 125–150.
[17] X. L. Fan and C. Ji, Existence of infinitely many solutions for a Neumann
problem involving the p(x)-Laplacian, C. J. Math. Anal. Appl., 2007, 334, 248–
260.
[18] S. Gala, Q. Liu and M. A. Ragusa, A new regularity criterion for the nematic
liquid crystal flows, Applicable Analysis, 2012, 91(9), 1741–1747.
[19] S. Gala and M. A. Ragusa, Logarithmically improved regularity criterion for
the Boussinesq equations in Besov spaces with negative indices, Appl. Anal.,
2016, 95(6), 1271–1279.
[20] J. R. Graef, S. Heidarkhani and L. Kong, A variational approach to a Kirchhoff-
type problem involving two parameters, Results Math., 2013, 63(3–4), 877–889.
[21] P. Gwiazda, P. Minakowski and A. Wróblewska-Kamin, Elliptic problems in
generalized Orlicz-Musielak spaces, Cent. Eur. J. Math., 2012, 10(6), 2019–
2032.
[22] P. Gwiazda, I. Skrzypczak and A. Zatorska-Goldstein, Existence of renormal-
ized solutions to elliptic equation in Musielak-Orlicz space, J. Differential Equa-
tions, 2018, 264(1), 341–377.
[23] P. Gwiazda and A. Swierczewska-Gwiazda, On non-Newtonian fluids with a
property of rapid thickening under different stimulus, Math. Models Methods
Appl. Sci., 2008, 18(7), 1073–1092.
[24] P. Harjulehto and P. Hästö, Sobolev Inequalities for Variable Exponents At-
taining the Values 1 and n, Publ. Mat., 2008, 52(2), 347–363.
[25] T. Isernia, Sign-changing solutions for a fractional Kirchhoff equation, Nonlin-
ear Anal., 2020, 190, 111623, 20 pp.
[26] T. Isernia and D. D. Repovs̆, Nodal solutions for double phase Kirchhoff prob-
lems with vanishing potentials, Asymptot. Anal., 2021, 124(3–4), 371–396.
[27] G. Kirchhoff, Mechanik, Teubner, Leipzig, Germany, 1883.
[28] J. Lei and H. Suo, Multiple solutions of Kirchhoff type equations involving
Neumann conditions and critical growth, AIMS Mathematics, 2021, 6(4), 3821–
3837.
Multiplicity of weak solutions 2465

[29] W. Liu and G. Dai, Existence and multiplicity results for double phase problem,
J. Differential Equations, 2018, 265(9), 4311–4334.
[30] W. Liu and G. Dai, Three ground state solutions for double phase problem, J.
Math. Phys., 2018, 59(12), 121503.
[31] D. Liu and P. Zhao, Solutions for a quasilinear elliptic equation in Musielak-
Sobolev spaces, Nonlinear Anal. Real World Appl., 2015, 26, 315–329.
[32] M. Mihăilescu and V. D. Rădulescu, Neumann problems associated to nonho-
mogeneous differential operators in Orlicz-Sobolev spaces, Ann. Inst. Fourier,
2008, 58(6), 2087–2111.
[33] N. S. Papageorgiou, V. D. Rădulescu and D. D. Repovs̆, Double-phase problems
with reaction of arbitrary growth, Z. Angew. Math. Phys., 2018, 69, 108.
[34] K. Perera and M. Squassina, Existence results for double-phase problems via
Morse theory, Commun. Contemp. Math., 2018, 20(2), 1750023.
[35] S. Polidoro and M. A. Ragusa, Harnack inequality for hypoelliptic ultraparabolic
equations with a singular lower order term, Rev. Mat. Iberoam., 2008, 24(3),
1011–1046.
[36] B. Ricceri, A general variational principle and some of its applications, J.
Comput. Appl. Math., 2000, 113(1–2), 401–410.
[37] B. Ricceri, Energy functionals of Kirchhoff-type problems having multiple global
minima, Nonlinear Anal., 2015, 115, 130–136.
[38] B. Ricceri, On an elliptic Kirchhoff-type problem depending on two parameters,
J. Global Optim., 2010, 46(4), 543–549.
[39] M. Růz̆ic̆ka, Electrorheological Fluids: Modeling and Mathematical Theory,
Lecture Notes in Mathematics, Springer, Berlin, 2000, 1748.
[40] R. Stegliǹski, Infinitely many solutions for double phase problem with un-
bounded potential in RN , Nonlinear Anal., 2022, 214, 112580.
[41] S. Yacini, M. El Ouaarabi, C. Allalou and K. Hilal, Existence result for double
phase problem involving the (p(x), q(x))-Laplacian-like operators, J. Nonlinear
Anal. Appl., 2023, 14(1), 3201–3210.
[42] J. Zhang, The critical Neumann problem of Kirchhoff type, Appl. Math. Com-
put., 2016, 274, 519–530.
[43] V. V. Zhikov, Averaging of functionals of the calculus of variations and elas-
ticity theory, Izv Akad Nauk SSSR Ser Mat., 1986, 50(4), 675–710.
[44] V. V. Zhikov, On Lavrentiev’s phenomenon, Russ. J. Math. Phys., 1995, 3(2),
249–269.
[45] V. V. Zhikov, On some variational problems, Russ. J. Math. Phys., 1997, 5(1),
105–116.
[46] V. V. Zhikov, S. M. Kozlov and O. A. Oleinik, Homogenization of Diffrential
Operators and Integral Functionals, Berlin, Springer-Verlag, 1994.