SINET:

©
Ethiop. J. Sci., 45(2): 224-234, 2022 ISSN: 0379-2897 (PRINT)
College of Natural and Computational Sciences, Addis Ababa University, 2022 eISSN: 2520-7997

Date received: March 11, 2022; Date revised: July 25, 2022; Date accepted: July 28, 2022

DOI: https://dx.doi.org/10.4314/sinet.v45i2.8

Spectral problem for the Laplacian and a selfadjoint nonlinear elliptic boundary
value problem
Natnael Gezahegn and Tadesse Abdi∗
Department of Mathematics, Addis Ababa University, Ethiopia.

E-mail: tadesse.abdi@aau.edu.et

Abstract. In this paper, we present some connections between the spectral problem,
(
−∆u(x) = λ1 u(x) in Ω,
u(x) = 0 on ∂Ω
and selfadjoint boundary value problem,
(
∆u(x) − λ1 u(x) + g(x, u(x)) = h(x) in Ω,
u(x) = 0 on ∂Ω,

where λ1 is the smallest eigenvalue of −∆, Ω ⊆ Rn is a bounded domain, h ∈ L2 (Ω) and the nonlinear
function g is a Caratheodory function satisfying a growth condition. We initially investigate the

existence of solutions for the spectral problem by considering the selfadjoint boundary value problem.

The selfadjoint boundary value problem is then considered for both existence and estimation results.

We use degree argument in order to show that the selfadjoint boundary value problem has a solution

instead of the Landesman-Lazer condition or the monotonocity assumption on the second argument

of the function g.

Key words/phrases: Spectral problem, Boundary value problem, Second order elliptic partial
dierential operator, Selfadjoint operator, Laplace operator.

INTRODUCTION Among these, Nica [19] assured that the eigen-

functions for the eigenvalue problem
This paper is inspired by the search of nontrivial
−∆u(x) = λu(x) in Ω,

solutions in Iannacci et al. [10] and Iannacci and (1)
Nkashama [11, 12] for the selfadjoint boundary u(x) = 0 on ∂Ω,
value problem, form an orthonormal basis of L2 (Ω) and the cor-
Lu(x) + λ1 u(x) + g(x, u(x)) = h(x) in Ω, responding eigenvalues grow without bound.


u(x) = 0 on ∂Ω,
where λ1 is the smallest eigenvalue of −L,
Ω ⊆ Rn is a bounded domain, h ∈ Lp (Ω)
with p > n and the nonlinear function g is a
De Figueiredo and Ni [6] studied the existence
of solutions for a boundary value problem,
Lu(x) − λ1 u + g(u(x)) = h(x) in Ω,


Caratheodory function that grows at most lin- u(x) = 0 on ∂Ω,

early.
The solvability of a boundary value problem in-
volving the Laplace operator was investigated by
a number of researchers (see, e.g., [1, 4, 6, 10, 11,
12, 13, 15, 16, 19, 21] and references therein).
∗
Author to whom correspondence should be addressed.
without using the Landesman-Lazer condition
under some prescribed assumptions. Iannacci
and Nkashama [12] investigated this same prob-
lem by specifying L and provided sucient con-
ditions for the solvability of the semilinear two-
point boundary value problem,
 ′′
u (x) + u(x) + g(x, u(x)) = h(x), x ∈ (0, π),
u(0) = u(π) = 0,
SINET: Ethiop. J. Sci. , 45(2), 2022 225

in which g is not required to satisfy the into consideration.

Landesman-Lazer condition, ii. examine the solvability of problem (3)
Rπ Rπ Rπ without using the Landesman-Lazer con-
g− (x)sinxdx < h(x)sinxdx < g+ (x)sinxdx, dition or monotonicity assumption with
0 0 0
respect to the second argument of the
where g± = lim g(u) or the monotonicity as-
u−→±∞ nonlinear function g , and
sumption. After a while, Iannacci et al. [10] gen- iii. nd some estimation results by consider-
eralized the main results in De Figueiredo and ing problem (3).
Ni [6], Iannacci and Nkashama[12], Gupta [7, 8] In the above boundary value problems (2) and
and Ward [20]. In their generalization, it is indi- (3):
cated that there is a solution for the selfadjoint • Ω ⊆ Rn is a bounded domain with bound-
boundary value problem, ary ∂Ω of class C 1,µ , 0 < µ < 1;
Lu(x) + λ1 u(x) + g(x, u(x)) = h(x) in Ω, • λ1 is the smallest eigenvalue of −∆;


u(x) = 0 on ∂Ω, • h(x) ∈ L2 (Ω);

where L is a selfadjoint operator. In order to ar-
rive at this existence result they primarily studied
 • the second-order partial dierential oper-
−Lu(x) = λ1 u(x) in Ω,
u(x) = 0 on ∂Ω,
under some assumptions placed on L .
Recently, Xu and Ma [21] investigated the spec-
trum structure of the eigenvalue problem
 (4)
u (x) = λu(x), x ∈ (0, 1),
u(0) = u(1) = u′ (0) = u′ (1) = 0,
and for its application, they demonstrated the ex-
istence of solutions for the fourth order boundary
value problem
−u(4) (x) + λ1 u(x) + g(x, u(x)) = h(x),

u(0) = u(1) = u′ (0) = u′ (1) = 0,
at resonance, where λ1 is the smallest eigenvalue
of the corresponding linear eigenvalue problem.
They provided sucient conditions for solving
the above problem, in which the nonlinearity of g
does not necessarily satisfy the Landesman-Lazer
type condition or the monotonicity assumption.
The main targets of this study are to:
i. investigate the solvability of the spectral
problem for the Laplacian,
−∆u(x) = λ1 u(x) in Ω,

(2)
u(x) = 0 on ∂Ω
by taking a selfadjoint boundary value
problem,
∆u(x) − λ1 u(x) + g(x, u(x)) = h(x) in Ω,

• g : Ω × R −→ R is a Caratheodory func-
tion that grows at most linearly;
ator ∆ is the usual Laplacian.
The paper is organized as follows. In Section 2,
some facts concerning the spectral problem for
the Laplacian as well as important Lemmas re-
garding the selfadjoint boundary value problem
are discussed.
In Section 3, both existence and estimation re-
sults are presented. We have shown that there
is a solution for the spectral problem and selfad-
joint boundary value problem in an appropriate
Sobolev space. For the estimation results we con-
sidered the selfadjoint boundary value problem.
In section 4, we put a precise conclusion based
on the results obtained.
PRELIMINARIES
In this section, we describe some vital facts about
a spectral problem for the Laplacian and a self-
adjoint boundary value problem involving the
Laplace operator. Some auxiliary results are also
presented.
Denition 1. Let Ω ⊆ Rn be a bounded domain.
The spectral problem for the Laplace operator is
−∆u = λf (x, u) in Ω,

(4)
u = 0 on ∂Ω,
where f : Ω × R −→ R is a given function and λ

u(x) = 0 on ∂Ω is a real number. We say that λ is an eigenvalue

(3) of (4) if there exists u ∈ H01 (Ω) such that for any
226 Natnael Gezahegn and Tadesse Abdi

ψ ∈ H01 (Ω), where Q := H 2 (Ω) ∩ H01 (Ω). Then L is a selfad-

Z Z
∇u∇ψdx − λ f (x, u)ψ = 0.
Ω Ω
Furthermore, if λ is an eigenvalue of problem
(4), then u ∈ H01 (Ω) given in the above deni-
tion is called the eigenfunction corresponding to
λ. If f (x, u) = u, then (4) reduces to (1). The
discreteness of the spectrum of ∆ allows one to
order the eigenvalues, 0 < λ1 < λ2 < λ3 < · · · <
λn < · · · monotonically and λ1 is characterized
from a variational point of view as the minimum
of the Rayleigh quotient,
|∇u|2 dx
R
joint operator and thus L2 (Ω) admits the orthog-
onal direct sum decomposition L2 (Ω) = κ ⊕ ℜ,
where κ is the one dimensional null space of L
and ℜ is the range space of L, namely,
κ = {y(x) ∈ L2 (Ω) : y(x) = su1 (x) for some s ∈ R}
ℜ = {y(x) ∈ L2 (Ω) :
R
y(x)u1 (x)dx = 0}.
Ω
Therefore, each u ∈ H01 (Ω) has a unique decom-
position:
u = su1 (x) + w := ν(x) + φ(x),
where s ∈ R and w ∈ ℜ so that, with obvious
notation

λ1 = inf Ω R 2 . H01 (Ω) = H̄01 (Ω) ⊕ H̃01 (Ω),

u∈H01 (Ω) u dx
Ω
where H̄01 (Ω) and H̃01 (Ω) are spaces associated
Moreover, it is widely known that λ1 is simple,
with ν and φ, respectively.
all the associated eigenfunctions are multiples of
each other[5, 2, 3, 9]. Lemma 1. [10] Let w(x) ∈ L∞ (Ω) be such that
for a.e. x ∈ Ω, 0 ≤ w(x) ≤ λ2 − λ1 with w(x) in
Theorem 1. [18] For the eigenvalue problem (1), a subset of Ω with positive measure, σ > 0 and
all eigenvalues are positive. δ = δ(w) > 0. Then, for all p ∈ L∞ (Ω) satisfying

Theorem 2. [5, 18, 2] Let Ω ⊆ Rn be a bounded 0 ≤ p(x) ≤ w(x) + σ

domain. There exists a sequence of eigenvalues a.e. on Ω and all u ∈ Q one has
{λj } and eigenfunctions {uj } for (1) such that Z
(∆u(x) − u(x) + p(x)u(x))
0 < λ1 < λ2 < λ3 < · · ·
Ω
with λj −→ ∞ as j −→ ∞, uj ∈ H01 (Ω) and × (ν(x) − φ(x))dx ≥ (δ − σ)∥φ∥2H 1 .
−∆uj = λj uj . The family of eigenfunctions is a
Hilbert basis of L2 (Ω) and is orthogonal and com-
plete in H01 (Ω) equipped with the inner product
MAIN RESULTS
associated with the seminorm. From now on
• (·, ·) denotes the L2 inner product in Ω.
Theorem 3. [5, 18, 2] There exists u1 in the • we let
kernel of −∆ − λ1 I such that u1 (x) > 0 for all u(x) = ν(x) + φ(x),
x ∈ Ω.
for each u(x) ∈ H01 (Ω), ν(x) =
In other words, the eigenfunctions associated ( u(x)u1 (x)dx)u1 (x) with ∥u1 (x)∥2L2 (Ω) =
R

with the rst eigenvalue do not vanish in Ω, Ω

1 and φ(x) = u(x) − ν(x). The func-
and thus we can choose one that is strictly pos-
tions ν(x) and φ(x) as dened above are
itive in Ω. The result is a consequence of the
L2 (Ω)-orthogonal.
Krein-Rutman theorem; see [14]. As a rule, the
• u : Ω −→ R is dened by
eigenfunctions associated with the other eigen-
values do vanish on nodal sets and change sign. u(x) = u+ (x) − u− (x),
where
Dene a linear operator L : Q ⊂ L2 (Ω) −→
u+ (x) = max{u(x), 0}
L2 (Ω) by
Lu := ∆u + λ1 u, u− (x) = max{−u(x), 0}.
SINET: Ethiop. J. Sci. , 45(2), 2022 227

• (−∆φ(x), φ(x)) ≥ λ2 (φ(x), φ(x)), λ2 is have

the second eigenvalue of −∆. ∆u(x) − λ1 u(x) + z(x, u(x))u(x) = 0 in Ω,

The proof of Theorem 4 will be based on the u = 0 on ∂Ω.
following assumptions.
This in turn leads to:
(H1). H+ (x), H− (x) ∈ L2 (Ω) in which for a.e.
x ∈ Ω, H± (x) is a nonnegative function [∆u(x) − λ1 u(x) + z(x, u(x))u(x)] [ν(x) − φ(x)] = 0.
and H± (x) ≤ λ1 + λ2 with This implies that
Z Z
[(λ1 + λ2 ) − H+ (x)](u2 (x))2 dx [∆u(x) − λ1 u(x) + z(x, u(x))u(x)]
u2 (x)>0 Ω
Z × [ν(x) − φ(x)] dx
+ [(λ1 + λ2 ) − H− (x)](u2 (x))2 dx > 0 Z
u2 (x)<0 = (ν(x)∆ (ν(x) + φ(x))
(5) Ω

for all u2 (x) satisfying − φ(x)∆ (ν(x) + φ(x)))dx

Z
−∆u2 (x) = λ2 u2 (x) in Ω,

− λ1 (ν(x) + φ(x)) (ν(x) − φ(x)) dx
u2 (x) = 0 on ∂Ω. Ω
Z
(H2). For all r± (x) ∈ L2 (Ω), consider + [z(x, u(x)) (ν(x) + φ(x)) (ν(x) − φ(x))] dx
0 ≤ r± (x) ≤ H± (x). Ω
Z
(H3). Dene z : Ω × R −→ R by = [−φ(x)∆φ(x) + ν(x)∆ν(x) − λ1 (ν(x))2
r+ (x) if x ∈ Ω and u ≥ 0,

Ω
z(x, u(x)) =
r− (x) if x ∈ Ω and u < 0 − (z(x, u(x)) − λ1 )(φ(x))2 ]dx
such that
Z
+ z(x, u(x))(ν(x))2 dx.
z(x, u(x))
0 ≤ z(x, u(x)) ≤ λ1 + λ2 with ≥ λ1 . Ω
2
Since
(H4). u(x) ∈ H 2 (Ω), satises
−∆ν(x) = λ1 ν(x) in Ω,
it follows that
∆u(x) − λ1 u(x) + r+ (x)u+ (x) − r− (x)u− (x) = 0 in Ω,


u(x) = 0 on ∂Ω.
Z
[−φ(x)∆φ(x) − (z(x, u(x)) − λ1 )(φ(x))2 ]dx
(6)
Ω
Z
+ (z(x, u(x)) − 2λ1 )(ν(x))2 dx = 0.
Existence result Ω
Moreover, we have the information that
Theorem 4. If (H1)-(H4) hold, then u(x) ∈
(−∆φ(x), φ(x)) ≥ λ2 (φ(x), φ(x)) and z(x, u(x)) ≤
H (Ω) is a solution of (2).
2
λ1 + λ2 with z(x,u(x))
2 ≥ λ1 and hence
Proof. From problem (3), we have
Z
−φ(x)∆φ(x) − (z(x, u(x)) − λ1 )(φ(x))2 dx
 
∆u(x) − λ1 u(x) + g(x, u(x)) − h(x) = 0, in Ω. Ω
Let z(x, u(x))u(x) = g(x, u(x)) − h(x), where
Z
+ (z(x, u(x)) − 2λ1 )(ν(x))2 dx
z(x, u(x)) is dened in (H3). Then
Ω
z(x, u(x))u(x) =z(x, u(x))(u+ (x) − u− (x))
Z
≥ [−φ(x)∆φ(x) − λ2 (φ(x))2 ]dx
+ −
=r+ (x)u (x) − r− (x)u (x).
Ω
Upon substituting z(x, u(x))u(x) for ≥ 0.
r+ (x)u+ (x) − r− (x)u− (x) in equation (6), we
228 Natnael Gezahegn and Tadesse Abdi

From the above inequality one can see that and

Z Z
[−φ(x)∆φ(x) − λ2 (φ(x))2 ]dx = 0. (z(x, u(x)) − 2λ1 )(ν(x))2 dx = 0, (9)
Ω Ω

But, this is true if and only if φ(x) = u2 (x). i.e., where

Z −∆ν(x) = λ1 ν(x) in Ω
[−φ(x)∆φ(x) − λ2 (φ(x))2 ]dx and
Ω
Z −∆u2 (x) = λ2 u2 (x) in Ω.
= [−u2 (x)∆u2 (x) − λ2 (u2 (x))2 ]dx If z = λ1 + λ2 and z = 2λ1 respectively, then (8)
Ω and (9) will be zero.
Z
= [λ2 (u2 (x))2 − λ2 (u2 (x))2 ]dx If we can show that ν(x) = u(x), then we are
Ω done.
= 0.
Now, for φ(x) = u2 (x), we get Suppose
Z Ω1 = {x ∈ Ω : ν(x) ̸= 0}.
−u2 ∆u2 − (z(x, u) − λ1 )(u2 (x))2 dx
 
Assume that Ω1 = ∅. Then u(x) = u2 (x). From
Ω
Z (8), (H2), (H4) and the fact that H± satises
+ (z(x, u) − 2λ1 )(ν(x))2 dx = 0. (7) (5), we have
Z
Ω
0 = [λ1 + λ2 − z(x, u2 (x))](u2 (x))2 dx
On the other hand,
Z Ω
−u2 (x)∆u2 (x) − (z(x, u) − λ1 )(u2 (x))2 dx
  Z
= [λ1 + λ2 − r+ (x)](u2 (x))2 dx
Ω
Z u2 (x)>0
λ2 (u2 (x))2 − (z(x, u) − λ1 )(u2 (x))2 dx
 
=
Z
+ [λ1 + λ2 − r− (x)](u2 (x))2 dx.
Ω
Z u2 (x)<0
= [λ1 + λ2 − z(x, u(x))] (u2 (x))2 dx.
Thus by (H1) and (H2), u2 (x) = 0 and hence
Ω u(x) = 0 = ν(x).
Consequently (7) becomes,
Assume again that Ω1 ̸= ∅. Then ν(x) ̸= 0
Z
0 = [λ1 + λ2 − z(x, u(x))] (u2 (x))2 dx
and hence in (9)
Ω
Z z(x, u(x)) = 2λ1
+ (z(x, u(x)) − 2λ1 )(ν(x))2 dx
for a.e x ∈ Ω.
Ω
Z
≥ [λ1 + λ2 − z(x, u(x))] (u2 (x))2 dx In (8), λ1 + λ2 − z(x, u(x)) ̸= 0 and hence
Ω u2 (x) = 0. This implies that
≥ 0. u(x) = ν(x) + u2 (x) = ν(x).
Furthermore, From the fact that
−∆ν(x) = λ1 ν(x) in Ω,

z(x, u)] u22 dx )ν 2 dx
R R
[λ1 + λ2 − = − (z(x, u) − 2λ1
Ω Ω
ν(x) = 0 on ∂Ω,
when
Z and u(x) = ν(x) (from the computation above),
[λ1 + λ2 − z(x, u(x))] (u2 (x))2 dx = 0 (8) one can conclude that
−∆u(x) = λ1 u(x) in Ω,

Ω
u(x) = 0 on ∂Ω.
SINET: Ethiop. J. Sci. , 45(2), 2022 229

□ ∥un ∥Q ≥ n for all n ∈ N such that

∆un (x) − λ1 un (x) + pun (x) = sn [−pun (x) − g(x, un (x)) + h(x)] in Ω,


un (x) = 0 on ∂Ω.
The existence of solutions for problem (3) will (13)
be guaranteed by taking the following assump- Setting vn = ∥un ∥Q , we have
un
tions into account. (
(H5). g(x, u(x))u(x) ≥ 0 for a.e. x ∈ Ω and all
g(x,un (x)) h(x)
∆vn (x) − λ1 vn (x) + pvn (x) = sn [−pvn (x) − ∥un ∥Q + ∥un ∥Q ] in Ω,
vn (x) = 0 on ∂Ω.
u ∈ R.
(14)
(H6). For any constant σ > 0, there exist a
Dene an operator E : Q −→ L2 (Ω) by
constant R = R(σ) > 0 and a function g(x,u(x)) h(x)
Ev(x) = ∆v(x) − λ1 v(x) + pv(x) = s[−pv(x) − + ∥u∥Q ],
b = b(σ) ∈ L∞ (Ω) such that ∥u∥Q
where E is invertible and its inverse is compact
|g(x, u(x))| ≤ (w(x) + σ)|u(x)| + b(x), from L2 (Ω) into Q.
for a.e. x ∈ Ω and all u ∈ R with |u| ≥ R,
where w ∈ L∞ (Ω) is such that for a.e. It follows from (H6) and (H7), there ex-
x∈Ω ists a function c ∈ L2 (Ω) depending only on
R = R(δ) > 0 such that
0 ≤ w ≤ λ2 − λ1 .
δ
|g(x, u(x))| ≤ (q(x) + )|u(x)| + b(x) + c(x)
(H7). g : Ω × R −→ R is a Caratheodory func- 4
tion, for any constant r > 0 there exists for a.e. x ∈ Ω and all u ∈ R. Therefore, g(x,un (x))
∥un ∥Q
a function qr ∈ L2 (Ω) such that
is bounded in L2 (Ω). Similarly, from (H6), we
|g(x, u(x))| ≤ qr (x), have
for a.e. x ∈ Ω and all u ∈ R with |u| ≤ r. δ
|g(x, u(x))| ≤ (q(x) + )|u(x)| + b(x) (15)
(H8). ∆u(x) − λ1 u(x) > 0, for all u ∈ Q. 4
for a.e. x ∈ Ω and all u ∈ R with |u| ≥ R, where
Theorem 5. Assume that (H5)-(H8) hold. R is chosen such that b(x)
|u| < 4 .
δ

Then (3) has at least one solution u ∈ Q for any

h ∈ L2 (Ω) with This implies that
g(x, un (x)) h(x)
Z
h(x)u1 (x)dx = 0. (10) sn [pvn (x) − + ]
∥un ∥Q ∥un ∥Q
Ω
is bounded in L2 (Ω).
Proof. Let δ > 0 be associated with function w
and p ∈ (0, λ2 − λ1 ) be a xed constant with One can rewrite the equation in (14) as
p < 2δ . According to the Leray-Scahuder con- 
g(x, un (x)) h(x)

−1
tinuation approach [17], proving that (3) has at vn = E sn [pvn (x) − + ] .
∥un ∥Q ∥un ∥Q
least one solution is all about demonstrating that
the possible solutions of the homotopy Since E −1 : L2 (Ω) −→ Q is compact, there exists
v ∈ Q such that
∆u(x) − λ1 u(x) + (1 − s)pu(x) + sg(x, u(x)) = sh(x) in Ω,

(
u(x) = 0 on ∂Ω lim vn = v in Q,
(11) n−→∞
∥v∥Q = 1.
has a priori bound in Q.
Let us dene q̃ : Ω × R −→ R by
Our claim is to show that if u ∈ Q is a solu- for |u| ≥ R
 −1
 u g(x, u)
q̃(x, u) = R−1 g(x, R)( R
u
) + (1 − Ru
)w(x) for 0 ≤ u < R
tion of the homotopy (11), then there exists a  −1
R g(x, −R)( R u
) + (1 + Ru
)w(x) for − R < u ≤ 0
constant ρ > 0 independently of s ∈ [0, 1) such (16)
that Then (16) together with (15) and (H5) gives
∥u∥Q < ρ. (12) δ
If we assume that our claim is false, then there 0 ≤ q̃(x, u) ≤ w(x) + (17)
2
exist sequences {sn } ⊂ (0, 1) and {un } ⊂ Q with
230 Natnael Gezahegn and Tadesse Abdi

for a.e. x ∈ Ω and all u ∈ R. there exists n0 ∈ N such that n ≥ n0 , vn (x) > 0
in Ω. So that, for n ≥ n0 ,
In addition, q̃(x, u)u satises the Caratheodory
un (x) > 0 for all x ∈ Ω. (23)
condition and dene f : Ω × R −→ R by
For sn ∈ (0, 1), taking the inner product in L2 (Ω)
f (x, u) = g(x, u) − q̃(x, u)u, (18)
of the equation in (14) with ν̄n and considering
for a.e. x ∈ Ω and all u ∈ R such that by (H7) (10), we get
|f (x, u)| ≤ v(x) (19)
Z
(∆ν̄n (x) − λ1 ν̄n (x)) ν̄n (x)dx + p(1 + sn )
for some v ∈ L2 (Ω) . Ω
Z Z
sn
Upon substituting (18) in (11), we have × (ν̄n (x))2 dx+ g(x, un (x))ν̄n (x)dx
∥un (x)∥Q

∆u(x) − λ1 u(x) + (1 − s)pu(x) + sq̃(x, u(x))u + sf (x, u) = sh(x) in Ω, Ω Ω
u = 0 on ∂Ω. = 0 (24)
(20)
From the fact that p ∈ (0, λ2 − λ1 ) with p < 2δ for all n suciently large. From (H8) and
and (17), we get ν̄n (x) > 0, we have
Z
δ (∆ν̄n (x) − λ1 ν̄n (x)) ν̄n (x)dx > 0.
0 ≤ (1 − s)p + sq̃(x, u) ≤ q(x) + (21)
2
Ω
for a.e. x ∈ Ω and all u ∈ R.
Moreover, p(1 + sn ) ν̄n (x)2 dx > 0.
R
Ω
If u ∈ Q is a solution of (20) for some s ∈ (0, 1)
For (24) to be true,
(for s = 0, we have a trivial solution), then by Z
Lemma 1 and the Cauchy-Schwartz inequality, sn
g(x, un (x))ν̄n (x)dx < 0
we have ∥un ∥Q
Z Ω
(ν(x) − φ(x))[∆u(x) − λ1 u(x) + (1 − s)pu(x) and hence
Z
Ω
Z g(x, un )ν̄n dx < 0.
+ sq̃(x, u)u]dx + (ν(x) − φ(x)) Ω
Ω From (23) and (H5), one can conclude a contra-
× [sf (x, u) − sh(x)]dx diction and the proof is complete. □
δ
≥ ∥φ(x)∥2H 1 − (∥φ(x)∥L2 + ∥ν(x)∥L2 )
2
× (∥f ∥L2 + ∥h∥L2 ).
This implies from the compact embedding of Some estimation results
H 1 (Ω) into L2 (Ω) and (19) that
For the next theorems, we consider problem (3)
δ
0 ≥ ∥φ(x)∥2H 1 − α(∥φ(x)∥H 1 + ∥ν(x)∥H 1 ) (22) with s(x) ∈ L∞ (Ω).
2
for some constant α > 0.
Theorem 6. Suppose j(x) ∈ L∞ (Ω) such that
0 ≤ j(x) ≤ λ1 and µ > 0. Let
From the inequality (22), one deduces imme-
diately that φ̃n −→ 0 in H 1 (Ω) for n −→ ∞, 0 ≤ s(x) ≤ j(x) − µ.
where vn = ν̄n + φ̃n . Therefore, vn = ν̄n . Since Then there exists a constant µ = µ(j) > 0 such
∥v∥Q = 1, one is required to take that for all u ∈ Q, we have
v = eu1 for some e > 0. Z
[∆u − λ1 u + s(x)u(x)] [ν(x) − φ(x)] dx
Now, using the fact that v = 0 on ∂Ω, vn −→ v
Ω
in Q for n −→ ∞ with v > 0 in Ω, we have that
≥ 2µ∥φ∥2H 1 (Ω) .
SINET: Ethiop. J. Sci. , 45(2), 2022 231

Proof. Let g(x, u) − h(x) = s(x)u(x) in (3). Proof. Rewriting the equation in (3), we have
Then, in similar fashion as in Theorem 4, we have
Z 0 = ∆u(x)−λ1 u(x)+s(x)u(x)−h(x) in Ω. (25)
[∆u(x) − λ1 u(x) + s(x)u(x)] [ν(x) − φ(x)] dx From (25), we get
Z
Ω
Z Z [u(x)∆u(x) − λ1 (u(x))2 + s(x)(u(x))2 ]dx
= s(x)(ν(x))2 dx − φ(x)∆φ(x)dx Ω
Z
Ω Ω
Z − h(x)u(x)dx
+ (λ1 − s(x))(φ(x))2 dx. Ω
Ω
Z Z
∂u(x)
=− |∇u(x)|2 dx + u(x) dx
From the facts that s(x) ≥ 0 and Green's rst ∂η
identity, we get Ω
Z
∂Ω
Z
2
s(x)(u(x))2 dx
Z
− λ1 (u(x)) dx +
[∆u(x) − λ1 u(x) + s(x)u(x)] [ν(x) − φ(x)] dx
Ω Ω
Ω Z
Z Z
− h(x)u(x)dx
≥ −φ(x)∆φ(x)dx + (λ1 − s(x))(φ(x))2 dx
Ω
Ω Ω Z Z
|∇u(x)|2 dx − λ1 (u(x))2 dx
Z Z
= −
= |∇φ(x)|2 dx + (λ1 − s(x))(φ(x))2 dx.
Ω Ω
Ω Ω Z Z
But, s(x) ≤ j(x) − µ implies λ1 − s(x) ≥ µ − + s(x)(u(x))2 dx − h(x)u(x)dx
j(x) + λ1 and hence Ω Ω
Z Z Z
[∆u(x) − λ1 u(x) + s(x)u(x)] [ν(x) − φ(x)] dx =− |∇u(x)|2 dx + (s(x) − λ1 )(u(x))2 dx
Ω Ω Ω
Z Z Z
≥ |∇φ(x)|2 dx + (λ1 − j(x) + µ)(φ(x))2 dx − h(x)u(x)dx
Ω Ω Ω
Z Z Z
= |∇φ(x)|2 dx + (λ1 − j(x))(φ(x))2 dx ≤− h(x)u(x)dx.
Ω Ω Ω

This in turn gives h(x)u(x)dx ≤ 0 if s(x) ≤

Z R
2
+ µ(φ(x)) dx. Ω
Ω λ1 . □
Here one can choose a constant λ1 − j(x) ≥ 0 as
µ and hence
Z For the next estimation result we will use the
[∆u(x) − λ1 u(x) + s(x)u(x)] [ν(x) − φ(x)] dx following assumptions.
Ω (H9) Assume that v(x) ∈ Q solves
Z Z
∆u(x) − λ1 u(x) + s(x)u(x) = 0 in Ω,

2
≥ (λ1 − j(x))(φ(x)) dx + µ(φ(x))2 dx
u(x) = 0 on ∂Ω
Ω Ω
and w(x) ∈ Q is a solution for
Z Z
= µ(φ(x))2 dx + µ(φ(x))2 dx
∆u(x) + λ1 u(x) + s(x)u(x) = 0 in Ω,

Ω Ω
u(x) = 0 on ∂Ω.
= 2µ∥φ∥2H 1 (Ω) .
(H10) w(x) − v(x) ≥ 0 in Ω and h(x) ≥ 0 in Ω
□

Theorem 7.R If s(x) ≤ λ1 and g(x, u) = s(x)u(x) Theorem 8. Let (H9)-(H10) hold. Then
in (3), then
R
h(x)u(x)dx ≤ 0. v(x)w(x) ≤ 0.
Ω Ω
232 Natnael Gezahegn and Tadesse Abdi

Proof. By (H9), we have Multiplying (31) by v(x) and w(x) give respec-
tively
∆v(x) − λ1 v(x) + s(x)v(x) = h(x) (26)
and v(x)∆v(x) + v(x)∆w(x) − λ1 (v(x) + w(x))v(x)
∆w(x) + λ1 w(x) + s(x)w(x) = h(x). (27) + s(x)(v(x) + w(x))v(x) = h(x)v(x) (32)

Multiplying (26) by w(x) and (27) by v(x), we and

get w(x)∆v(x)+w(x)∆w(x)−λ1 (v(x)+w(x))w(x)
w(x)∆v(x) − λ1 v(x)w(x) + s(x)v(x)w(x) + s(x)(v(x) + w(x))w(x) = h(x)w(x). (33)
= h(x)w(x) (28) The above equations in (32) and (33) leads to,
and Z Z
(v∆v−w∆w)dx− λ1 ((v(x))2 −(w(x))2 )dx
v(x)∆w(x) + λ1 w(x)v(x) + s(x)v(x)w(x) Ω Ω
= h(x)v(x). (29)
Z
+ s(x)((v(x))2 − (w(x))2 )dx
Consequently, we obtain Ω
Z Z Z
(w(x)∆v(x) − v(x)∆w(x))dx − 2λ1 v(x) = h(v(x) − w(x))dx.
Ω
Ω Ω
This is the same as
Z
× w(x)dx = h(x)(w(x) − v(x))dx. (30) Z Z
Ω
(v∆v − w∆w)dx = (λ1 − s(x))((v(x))2
However, we know from Green's second identity Ω Ω
Z
that − (w(x))2 )dx + h(x)(v(x) − w(x))dx
Z
Ω
(w(x)∆v(x) − v(x)∆w(x))dx Z
Ω = (v(x) − w(x))[h(x) + (λ1 − s(x))
Z Z
dv(x) dw(x) Ω
= w(x) dx − v(x) dx = 0.
dη dη (w(x) + v(x))]dx.
∂Ω ∂Ω
As a result, by (H10) equation (30) becomes After imposing the assumptions h(x) ≥ 0, λ1 ≥
R
v(x)w(x)dx = − 2λ1 1 h(x)(w(x) − v(x))dx ≤ 0.
R s(x) and 0 ≤ w(x) ≤ v(x) in Ω, we have
Ω Ω
Z
□ (v(x)∆v(x) − w(x)∆w(x))dx ≥ 0.
Ω
Theorem 9. Let (v(x) + w(x)) ∈ Q be a solu- □
tion of ∆u(x) − λ1 u(x) + s(x)u(x) = h(x) in Ω.
If h(x) ≥ 0, λ1 ≥ s(x) and 0 ≤ w(x) ≤ v(x) in
Ω, then (v(x)∆v(x) − w(x)∆w(x))dx ≥ 0.
CONCLUSIONS
R
Ω

In this study, we took a closer look on the exis-

Proof. Since v(x)+w(x) solves ∆u(x)−λ1 u(x)+
s(x)u(x) = h(x) in Ω, we have
∆(v(x) + w(x)) − λ1 (v(x) + w(x))
+ s(x)(v(x) + w(x)) = h(x).
This suggests that
∆v(x) + ∆w(x) − λ1 (v(x) + w(x))
+ s(x)(v(x) + w(x)) = h(x). (31)
tence of solutions for problem (2) and (3) and
some estimation results involving problem (3).
The existence of solutions for problem (2) was
guaranteed by taking problem (3) into account.
In our existence result for problem (2), one can
observe that φ must be equal to u2 . If φ = au2
for any a ∈ R, then the obtained existence result
fails to be true. This is from the fact that if u
satisfying (H4), then u = ν + φ, cannot be any
SINET: Ethiop. J. Sci. , 45(2), 2022
function in H 2 (Ω) with p > n, where ν is the
function dened in Section 3 and φ = au2 for
any a ∈ R. The existence of solutions for prob-
lem (3) is based on the Leray-Scahuder continu-
ation approach. Besides this, Theorem 6 is valid
whenever the newly added assumptions, namely
λ1 − j(x) > 0 and 0 ≤ s(x) ≤ j(x) − µ hold. The
remaining estimation results are also valid after
employing their respective assumptions correctly.
ACKNOWLEDGMENTS
The authors acknowledge the reviewers for their
meticulous reading and critical suggestions which
helped us to improve the paper. In addition to
this, the authors would also like to acknowledge
the Department of Mathematics, Addis Ababa
University for the pleasant environment and the
rst author express his gratitude to International
Science Program (ISP) and the Simons Founda-
tion Fellowship, Botswana International Univer-
sity of Science and Technology (BIUST) for their
nancial support.
refrences
1 Alexander, N.K.( 1977). Separate asymp-
totics of two series of eigenvalues for a
single elliptic boundary-value problem.
Mathematical notes of the Academy of
Sciences of the USSR, 22(5):882888.
2 Dret, H.L.(2018). Nonlinear Elliptic Partial
Dierential Equations. Springer.
3 Evans, L.C.(1998). Graduate studies in math-
ematics.
4 Fan, X.(2009). Remarks on eigenvalue prob-
lems involving the p (x)-laplacian. Jour-
nal of Mathematical Analysis and Appli-
cations, 352(1):8598.
5 Gilbarg, D. & Trudinger, N.S.(2015). Ellip-
tic partial dierential equations of second
order, volume 224. springer.
6 De Figueiredo, D.G. & Ni, W.M.(1979). Per-
turbations of second order linear ellip-
tic problems by nonlinearities without
landesmanlazer condition. In Djairo G.
de Figueiredo-Selected Papers, pages 85
90. Springer.
233
7 Gupta, C.P.(1982). Perturbations of sec-
ond order linear elliptic problems by un-
bounded nonlinearities. Nonlinear Anal-
ysis: Theory, Methods & Applications,
6(9):919933.
8 Gupta, C.P.(1988). Solvability of a bound-
ary value problem with the nonlinearity
satisfying a sign condition. Journal of
mathematical analysis and applications,
129(2):482492.
9 Han, Q. & Lin, F.(2011). Elliptic partial dif-
ferential equations, volume 1. American
Mathematical Soc.
10 Iannacci, R., Nkashama, M.N., & Ward, J.R.
( 1989). Nonlinear second order ellip-
tic partial dierential equations at res-
onance. Transactions of the American
Mathematical Society, 311(2):711726.
11 Iannacci, R. & Nkashama, M.N.(1987). Un-
bounded perturbations of forced second
order ordinary dierential equations at
resonance. Journal of Dierential Equa-
tions, 69(3):289309.
12 Iannacci, R. & Nkashama, M.N.(1989). Non-
linear two-point boundary value prob-
lems at resonance without landesman-
lazer condition. Proceedings of the Amer-
ican Mathematical Society, 106(4):943
952.
13 Kawohl, B. & Schuricht, F.(2007). Dirichlet
problems for the 1-laplace operator, in-
cluding the eigenvalue problem. Commu-
nications in Contemporary Mathematics,
9(04):515543.
14 Krein, M.G. & Rutman, M.A.(1948). Lin-
ear operators leaving invariant a cone in a
banach space. Uspekhi Matematicheskikh
Nauk, 3(1):395.
15 Lê, A.(2006). Eigenvalue problems for the
p-laplacian. Nonlinear Analysis: The-
ory, Methods & Applications, 64(5):1057
1099.
16 Lian, S., Gao, W., Yuan, H. & Cao, C.(2012).
Existence of solutions to an initial dirich-
let problem of evolutional p(x)-laplace
equations. In Annales de l'IHP Analyse
non linéaire, volume 29, pages 377399.
17 Lloyd, N.(1978). Degree theory. Cambridge
234
University Press.
18 McOwen, R.C.(2004). Partial dierential
equations: methods and applications. Ts-
inghua University Press.
19 Nica, M.(2011). Eigenvalues and eigenfunc-
tions of the laplacian. The Waterloo
Mathematics Review, 1(2):2334.
Natnael Gezahegn and Tadesse Abdi
20 Ward, J.R.(1986). A note on the dirichlet
problem for some semilinear elliptic equa-
tions. preprint.
21 Xu, M. & Ma, R.(2017). On a fourth-order
boundary value problem at resonance.
Journal of Function Spaces, 2017.