Existence and multiplicity of positive solutions for a singular system via sub-supersolution method
and Mountain Pass Theorem
Suellen Cristina Q. Arruda
B1and Rubia G. Nascimento
*21Faculdade de Ciências Exatas e Tecnologia, Universidade Federal do Pará, 68440-000 Abaetetuba, Pará, Brazil
2Instituto de Ciências Exatas e Naturais, Universidade Federal do Pará, 66075-110 Belém, Pará, Brazil
Received 3 April 2020, appeared 1 April 2021 Communicated by Gennaro Infante
Abstract: In this paper we show the existence and multiplicity of positive solutions using the sub-supersolution method and Mountain Pass Theorem in a general singular system which the operator is not homogeneous neither linear.
Keywords: p&q-Laplacian operator, sub-supersolution method, singular system, Mountain Pass theorem.
2020 Mathematics Subject Classification: Primary 35J20, 35J50; Secondary 58E05.
1 Introduction
In this paper we treat the question of the existence and multiplicity of positive solutions for the following class of singular systems of nonlinear elliptic equation
−div(a1(|∇u|p1)|∇u|p1−2∇u) =h1(x)u−γ1 +Fu(x,u,v)inΩ,
−div(a2(|∇v|p2)|∇v|p2−2∇v) =h2(x)v−γ2+Fv(x,u,v)inΩ, u,v>0 inΩ,
u =v=0 on∂Ω,
(1.1)
where Ω⊂RN is a bounded domain with smooth boundary, N≥3 and 2≤ p1,p2 < N. For i = 1, 2, γi > 0 is a fixed constant, ai : R+ → R+ is a C1-function and hi ≥ 0 is a nontrivial measurable function. More precisely, we suppose that the functions hi and ai satisfy the following assumptions:
(H) There exists 0<φ0 ∈C10(Ω)such thathiφ−0γi ∈ L∞(Ω).
BCorresponding author. Email: scqarruda@ufpa.br
*Email: rubia@ufpa.br
(A1) There exist constantsk1,k2,k3,k4 >0 and 2≤ pi ≤qi <N such that k1tpi+k2tqi ≤ ai(tpi)tpi ≤ k3tpi+k4tqi, for allt ≥0.
(A2) The functions
t7−→ai(tpi)tpi−2 are increasing.
(A3) The functions
t7−→ Ai(tpi) are strictly convex, where Ai(t) =Rt
0 ai(s)ds.
(A4) There exist constantsµi, q1∗
1 <θs< q1
1 and q1∗
2 <θt < q1
2 such that 1
µiai(t)t≤ Ai(t), for allt ≥0, with qp1
1 ≤µ1< 1
θsp1 and qp2
2 ≤µ2< 1
θtp2.
Notice that the functions ai satisfy suitable monotonicity conditions which allow to con- sider a larger class of p&q type problems. In order to illustrate the degree of generality of the kind of problems studied here, in the following we present some examples of functionsai which are interesting from the mathematical point of view and have a wide range of applica- tions in physics and related sciences.
Example 1.1. If ai ≡ 1, for eachi=1, 2, our operator is the p-Laplacian and so problem (1.1)
becomes
−∆p1u= h1(x)u−γ1+Fu(x,u,v)inΩ,
−∆p2v= h2(x)v−γ2+Fv(x,u,v)inΩ, u=v =0 on∂Ω,
withqi = pi,k1+k2 =1 andk3+k4=1.
Example 1.2. Ifai(t) =1+tqi
−pi
pi , for eachi=1, 2, we obtain
−∆p1u−∆q1u=h1(x)u−γ1+Fu(x,u,v)inΩ,
−∆p2v−∆q2v=h2(x)v−γ2 +Fv(x,u,v)inΩ, u=v=0 on ∂Ω,
withk1=k2= k3 =k4=1.
Example 1.3. Takingai(t) =1+ 1
(1+t)pi
−2 pi
, for eachi=1, 2, we get
−div
|∇u|p1−2∇u+ |∇u|p1−2∇u (1+|∇u|p1)
p1−2 p1
= h1(x)u−γ1+Fu(x,u,v)inΩ,
−div
|∇v|p2−2∇v+ |∇v|p2−2∇v (1+|∇v|p2)
p2−2 p2
= h2(x)v−γ2+Fv(x,u,v)inΩ, u=v=0 on∂Ω,
withqi = pi,k1+k2 =1 andk3+k4=2.
Example 1.4. If we considerai(t) =1+tqi
−pi
pi + 1
(1+t)pi
−2 pi
, for eachi=1, 2, we obtain
−∆p1u−∆q1u−div
|∇u|p1−2∇u (1+|∇u|p1)
p1−2 p1
=h1(x)u−γ1+Fu(x,u,v)inΩ,
−∆p2v−∆q2v−div
|∇v|p2−2∇v (1+|∇v|p)
p2−2 p2
=h2(x)v−γ2+Fv(x,u,v)inΩ, u=v=0 on∂Ω,
wherek1 =k2=k4=1 andk3=2.
Remark 1.5. Note that by hypothesis(H)we havehi ∈ L∞(Ω)because
|hi|=|hiφ
−γi
0 φ
γi
0| ≤ khiφ
−γi
0 k∞φ
γi
0. HereFis a function onΩ×R2 of classC1 satisfying (F1) There exists 0<δ< 12 such that
−h1(x)≤Fs(x,s,t)≤0 a.e. in Ω, for all 0≤ s≤δ and
−h2(x)≤Ft(x,s,t)≤0 a.e. inΩ, for all 0≤t≤δ.
It is worthwhile to point out that, since pi <qi and by the boundedness ofΩ,W01,pi(Ω)∩ W01,qi(Ω) = W01,qi(Ω). Thus, in order to show the existence and multiplicity of solutions to system (1.1), we define the Sobolev spaceX=W01,q1(Ω)×W01,q2(Ω)endowed with the norm
k(u,v)k= kuk1,q1+kvk1,q2, where
kuk1,qi =
Z
Ω
|∇u|qidx
1 qi
.
Moreover, we say that a pair (u,v) ∈ X is a positive weak solution of system (1.1) if u,v>0 inΩand it verifies
Z
Ω
a1(|∇u|p1)|∇u|p1−2∇u∇φdx=
Z
Ω
h1(x)u−γ1φdx+
Z
Ω
Fu(x,u,v)φdx
and Z
Ω
a2(|∇v|p2)|∇v|p2−2∇v∇ϕdx =
Z
Ω
h2(x)v−γ2ϕdx+
Z
Ω
Fv(x,u,v)ϕdx, for all (φ,ϕ)∈X.
In our first theorem we apply the sub-supersolution method to establish the existence of a weak solution for system (1.1).
Theorem 1.6. Suppose that(H),(F1)and(A1)–(A3)are satisfied. Then system(1.1) has a positive weak solution ifkhik∞ is sufficiently small, for i=1, 2.
Furthermore, we assume the conditions below to prove the existence of two solutions for problem (1.1).
(F2) Fori=1, 2, there exists 1<r <q∗i = (NNq−qi
i) (q∗i =∞ifqi ≥ N)such that Fs(x,s,t)≤ h1(x)(1+sr−1+tr−1) a.e. in Ω, for all s≥0 and
Ft(x,s,t)≤h2(x)(1+sr−1+tr−1) a.e. in Ω, for allt≥0.
(F3) There exists0,t0 >0 such that
0<F(x,s,t)≤θssFs(x,s,t) +θttFt(x,s,t) a.e. inΩ, for alls≥ s0 andt ≥t0, whereθs andθt appeared in(A4).
Theorem 1.7. Suppose that(H), (F1)–(F3)and(A1)–(A4)are satisfied. Then system(1.1) has two positive weak solutions ifkhik∞is sufficiently small, for i=1, 2.
Singular problems has been much studied in last years. We are going to cite some authors in last ten years. System (1.1) with Laplacian operator in both equations was studied in [9], where it was investigated the questions of existence, non-existence and uniqueness for solutions. The results in [9] were complemented in [16]. The general operator as we consider in this paper was studied in [5] using continuous unbounded of solutions. The cases with Laplacian operator involving weights were studied in [7] and [11].
In this paper we complement the results that can be found in [5], [7], [9], [11] and [16]
because we consider a general problem with singularity without restrictions in the exponents.
Moreover, we are considering the sub-supersolution method for a system that involves a non- linear and nonhomegeneous operator. The reader can see the generality of the operator in [5].
We would like to highlight that our theorems can be applied for the model nonlinearity F(x,s,t) =h1(x)
sr
r −sδr−1
+h2(x) tr
r −tδr−1
.
This paper is organized in the following way. Section 2 is devoted to some preliminary results in order to prove the main results. The first theorem is proved in the Section 3 and the second theorem in the Section 4.
2 Preliminary results
The next lemma provides the uniqueness of solution to the linear problem. The proof can be found in [5, Lemma 1]. However, for the convenience of the reader, we also prove it here.
Lemma 2.1. Assume that the conditions (A1)and(A2)hold. Then, there exists an unique solution ui ∈W01,qi(Ω)of the linear problem
(−div(ai(|∇ui|pi)|∇ui|pi−2∇ui) =hi(x)inΩ, ui =0on ∂Ω,
where hi ∈(W01,qi(Ω))0, for all i=1, 2and2≤ pi ≤qi < N.
Proof. Consider the operatorTi :W01,qi(Ω)−→(W01,qi(Ω))0 given by hTiui,φii=
Z
Ω
ai(|∇ui|pi)|∇ui|pi−2∇ui∇φi dx.
In virtue of hypothesis (A1), we can show that the operator Ti is well defined and it is con- tinuous. Furthermore, by considering the hypothesis(A2), we argument as [8, Lemma 2.4] to obtain the following inequality
Ci|ui−vi|pi ≤ hai(|ui|pi)|ui|pi−2ui−ai(|vi|pi)|vi|pi−2vi,ui−vii, for someCi >0 and for alli=1, 2. Therefore,
hTiui−Tivi,ui−vii>0, for allui,vi ∈W01,qi(Ω)withui 6=vi, which implies thatTi is monotone. Moreover, using(A1)again we get
hTiui,uii
kuik1,qi ≥k2kuikq1,qi−1
i
and hence
kuiklim1,qi→∞
hTiui,uii
kuik1,qi = +∞,
which shows that Ti is coercive. Thus, applying the Minty–Browder Theorem [2, Theo- rem 5.15] there exists an unique ui ∈W01,qi(Ω)such thatTiui = hi(x).
Our approach in the study of system (1.1) rests heavily on the following Weak Comparison Principle for the p&q-Laplacian operator. The proof of the result below for the scalar case can be found in [6, Lemma 2.1].
Lemma 2.2. LetΩa bounded domain and consider ui,vi ∈W01,qi(Ω)satisfying
(−div(ai(|∇ui|pi)|∇ui|pi−2∇ui)≤ −div(ai(|∇vi|pi)|∇vi|pi−2∇vi)inΩ, ui ≤vi on∂Ω,
then ui ≤vi a.e. inΩ, for all i=1, 2and2≤ pi ≤qi < N.
Proof. Using the test functionφi = (ui−vi)+ :=max{ui−vi, 0} ∈W01,qi(Ω), we get
Z
Ω∩{ui>vi}
hai(|∇ui|pi)|∇ui|pi−2∇ui−ai(|∇vi|pi)|∇vi|pi−2∇vi,∇ui− ∇viidx≤0.
From Lemma 2.1,k(ui−vi)+k ≤0, which implies thatui ≤ vi a.e. in Ω.
Now, using Lemma2.2, it is possible to repeat the same arguments of [13, Hopf’s Lemma]
to obtain the next result
Lemma 2.3. Let Ω ⊂ RN be a bounded domain with smooth boundary and i = 1, 2. If ui ∈ C1(Ω)∩W01,qi(Ω), with2≤ pi ≤qi < N, and
−div(ai(|∇ui|pi)|∇ui|pi−2∇ui)≥0inΩ, ui >0inΩ,
ui =0on∂Ω.
Then, ∂u∂ηi <0on∂Ω, where ηis the outwards normal to∂Ω.
We enunciate an iteration lemma due to Stampacchia that we will use to prove the L∞- regularity of the solutions for this class of p&qtype problems.
Lemma 2.4(See [14]). Assume that φ:[0,∞)→[0,∞)is a nonincreasing function such that if h>
k > k0, for some α> 0, β > 1, φ(h) ≤ C((φ(k))β
h−k)α . Then, φ(k0+d) =0, where dα = C2βαβ−1φ(k0)β−1 and C is positive constant.
Lemma 2.5. Let ui ∈W01,qi(Ω)be solution to problem
(−div(ai(|∇ui|pi)|∇ui|pi−2∇ui) = fi inΩ,
ui =0on∂Ω, (2.1)
such that fi ∈ Lri(Ω)with ri > q∗q∗i
i−qi. Then, ui ∈ L∞. In particular, ifkfikri is small, then alsokuik∞ is small, for all i=1, 2and2≤ pi ≤qi < N.
Proof. Sinceui is the weak solution to (2.1) we can write Z
Ω
ai(|∇ui|pi)|∇ui|pi−2∇ui∇φi dx =
Z
Ω
fiφi dx, ∀φi ∈W01,qi(Ω). Fork>0, we define the test function
vi =sign(ui)(|ui| −k) =
u−k, ifu>k, 0, ifu= k, u+k, ifu<k.
Then, ui = vi+k sign(ui) and ∂u∂xi
j = ∂v∂xi
j in the set A(k) = {x ∈ Ω;|u(x)| > k}, vi = 0 in Ω−A(k) and vi ∈ W01,qi(Ω). By considering the test function vi and using the Hölder inequality, we get
Z
A(k)
ai(|∇vi|pi)|∇vi|pidx=
Z
Ω
fivi dx≤
Z
A(k)
|vi|q∗idx
1 q∗
i
Z
A(k)
|fi|ridx
1 ri
|A(k)|1−
1 q∗
i
+1
ri
,
where |A(k)|denotes the Lebesgue measure of A(k). Moreover, applying (A1)and Sobolev inequality we obtain
k2S
Z
A(k)
|vi|q∗idx
qi−1 q∗
i
≤
Z
A(k)
|fi|ridx
1 ri
|A(k)|1−
1 q∗
i
+1
ri
, (2.2)
whereSis the best constant in the Sobolev inclusion.
Note that if 0<k<h, then A(h)⊂ A(k)and
|A(k)|
1 q∗
i (h−k) =
Z
A(h)
(h−k)q∗idx
1 q∗
i
≤
Z
A(h)
|vi|q∗idx
1 q∗
i
≤
Z
A(k)
|vi|q∗idx
1 q∗
i
. (2.3)
It follows from (2.2) and (2.3) that
|A(k)| ≤ 1 (h−k)q∗i
1 (k2S)
q∗ qi−i1
kfik
q∗ qi−i1
ri |A(k)|
q∗ qi−i1
1−
1 q∗ i
+1
ri
.
Sinceri > q∗qi∗
i−qi we have β:= qq∗i
i−1
1− q1∗
i +r1
i
>1. Therefore, if we define
φ(h) =|A(h)|, α= q∗i, β:= q
∗i
qi−1
1− 1
qi∗ + 1 ri
, k0=0, we obtain that φis a nonincreasing function and
φ(h)≤ C(φ(k))β
(h−k)α , for all h>k>0.
By Lemma2.4, we conclude thatφ(d) =0 ford= Ckfik
qi1−1 ri
(k2S)
qi1−1
|Ω|β−α1 and hence,
kuik∞ ≤C kfik
1 qi−1
ri
(k2S)qi1−1
|Ω|β−α1, where β,α,SandCare constants that do not depend on fi andui.
Regarding the regularity of the solution of (2.1) the next result hold and the proof can be done repeating the same arguments of [10, Theorem 1].
Lemma 2.6. Fix hi ∈ L∞(Ω), for all i=1, 2, and consider ui ∈ W01,qi(Ω)∩L∞(Ω), with2 ≤ pi ≤ qi < N, satisfying the problem
(−div(ai(|∇ui|pi)|∇ui|pi−2∇ui) =hi inΩ, ui =0on∂Ω,
Then, ui ∈C1,α(Ω),for someα∈(0, 1).
The following result can be found in [12, Lemma 2.6]. The proof is presented for the completeness of the paper.
Lemma 2.7. Let φ,ω > 0 be any functions on C01(Ω). If ∂φ∂ν > 0 in ∂Ω, where ν is the inwards normal to∂Ω, then there exists C>0such that
φ(x)
ω(x) ≥C>0, for all x∈Ω. Proof. Forδ >0 sufficiently small, we consider the following set
Ωδ = {x∈Ω; dist(x,∂Ω)<δ}.
Sinceφ,ω >0 inΩandΩ\Ωδ is compact, there existsm>0 such that φ(x)
ω(x) ≥m, for all x∈ Ω\Ωδ. (2.4)
It follows from ∂φ∂ν >0 in∂Ωthat ∂φ∂η < 0, whereηis the outwards normal to ∂Ω. Further- more, since Ω ⊂ Rn is bounded domain, then ∂Ωis a compact set and consequently, there existsC1 <0 satisfying
∂φ(x)
∂η
≤C1, for allx ∈Ωδ. Sinceω∈C01(Ω), there existsC2>0 such that
∂ω(x)
∂η
≤C2, for allx ∈Ωδ. Consider K0 = infΩ
δ
∂ω
∂η <0 and define the function H(x) = αω(x)−φ(x), for allx ∈ Ωδ andα∈Rto be chosen later. Since 0<α< CK1
0 we obtain
∂H(x)
∂η =α∂ω(x)
∂η − ∂φ(x)
∂η ≥αK0−C1>0, for allx ∈Ωδ. Now, fix x∈Ωδ and consider the function
f(x) = H(x+sη), for alls∈R.
For everyx ∈ Ωδ, we choose an uniquexe∈ Ωδ so that there existsbs > 0 such that x+bsη= xe∈ ∂Ω. Hence, since H(∂Ω)≡0 we have
f(bs) = H(x+bsη) =H(xe) =0.
Applying the Mean Value Theorem, there existsξ ∈(0,bs)such that f(bs)− f(0) = f0(ξ)(bs−0),
which implies that
−H(x) = ∂H
∂η(x+ξη)bs>0 in Ωδ. Therefore, H(x)≤0 for all x∈Ωδ and hence,
αω(x)−φ(x)≤0, for all x∈Ωδ, which result in
αω(x)≤φ(x), for all x∈Ωδ. Thus,
φ(x)
ω(x) ≥ α>0, for allx ∈Ωδ. (2.5) By virtue of (2.4) and (2.5), we conclude that there existsC>0 so that
φ(x)
ω(x) ≥C, for allx ∈Ω.
3 Proof of Theorem 1.6
In the proof of Theorem 1.6, we combine the sub-supersolution method with minimization arguments. Before this, we need of the following definition.
We say that (u,v),(u,v) ∈ X form a pair of sub and supersolution for system (1.1) if u,u,v,v∈ L∞(Ω)with
(a) u≤ u,v≤v inΩandu=0≤u,v=0≤von∂Ω, (b) Given(φ,ϕ)∈X, withφ,ϕ≥0, we have
Z
Ω
a1(|∇u|p1)|∇u|p1−2∇u∇φdx≤
Z
Ω
h1(x)u−γ1φdx+
Z
Ω
Fu(x,u,w)φdx, for allw∈ [v,v],
Z
Ω
a2(|∇v|p2)|∇v|p2−2∇v∇ϕdx≤
Z
Ω
h2(x)v−γ2ϕdx+
Z
Ω
Fv(x,w,v)ϕdx, for allw∈ [u,u]
and
Z
Ω
a1(|∇u|p1)|∇u|p1−2∇u∇φdx≥
Z
Ω
h1(x)u−γ1φdx+
Z
Ω
Fu(x,u,w)φdx, for allw∈[v,v]
Z
Ω
a2(|∇v|p2)|∇v|p2−2∇v ∇ϕdx≥
Z
Ω
h2(x)v−γ2ϕ, dx+
Z
Ω
Fv(x,w,v)ϕdx, for allw∈[u,u].
The next result is essential to provide the existence of a subsolution and a supersolution for system (1.1) whenever we fix the value ofkhik∞with i=1, 2.
Lemma 3.1. Suppose that(H),(F1)and(A1)–(A2)are satisfied. Ifkhik∞ is small, for i=1, 2, then there exist u,u,v,v∈C1,α(Ω), for someα∈ (0, 1), such that
i) h1u−γ1,h2v−γ2 ∈ L∞(Ω), kuk∞ ≤ δ andkvk∞ ≤ δ, where δ is the constant that appeared in the hypothesis(F1);
ii) kuk∞ ≤δandkvk∞ ≤δ, where δis the constant that appeared in the hypothesis(F1); iii) 0<u(x)≤ u(x)a.e. inΩand0<v(x)≤v(x)a.e. inΩ;
iv) (u,v)is a subsolution and(u,v)is a supersolution for system(1.1).
Proof. From Lemma2.1 and maximum principle, there exists an unique positive solution 0<
u∈W01,q1(Ω)satisfying the problem below
(−div(a1(|∇u|p1)|∇u|p1−2∇u) =h1(x)in Ω,
u=0 on∂Ω. (3.1)
Similary, there exists an unique positive solution 0<v∈W01,q2(Ω)satisfying (−div(a2(|∇v|p2)|∇v|p2−2∇v) =h2(x)inΩ,
v=0 on ∂Ω. (3.2)
Sinceh1,h2 ∈L∞(Ω), it follows from Lemma2.5thatu,v∈ L∞(Ω)and there existC1,C2>
0 such that
kuk∞≤ C1kh1k
p11−1
∞ and kvk∞ ≤C2kh2k
1 p2−1
∞ ,
whereC1andC2are constants that does not depend onhi,uandv. Therefore, we may choose khik∞ sufficiently small, withi=1, 2, so that
kuk∞ ≤δ< 1
2 and kvk∞ ≤δ < 1 2.
Moreover, from Lemma 2.6 we have u,v ∈ C1,α(Ω), for some α ∈ (0, 1). Thus, by virtue of Lemmas2.3and2.7, there existC3,C4 >0 such that
u(x)−γ1
φ0(x)−γ1 ≤ C3−γ1 and v(x)−γ2
φ0(x)−γ2 ≤C−4γ2, for all x∈Ω.
Therefore, by(H)we get
|h1u−γ1| ≤C3−γ1kh1φ0−γ1k∞ and |h2v−γ2| ≤C−4γ2kh2φ−0γ2k∞, (3.3) implying thath1u−γ1,h2v−γ2 ∈ L∞(Ω), which ends the proof of condition (i).
In order to prove (ii), we invoke Lemma 2.1 and maximum principle once again to claim that there exists an unique positive solution 0<u∈W01,q1(Ω)satisfying
(−div(a1(|∇u|p1)|∇u|p1−2∇u) =h1(x)u−γ1 in Ω,
u =0 on∂Ω, (3.4)
and there exists an unique positive solution 0< v∈W01,q2(Ω)satisfying (−div(a2(|∇v|p2)|∇v|p2−2∇v) =h2(x)v−γ2 in Ω,
v=0 on ∂Ω. (3.5)
Sinceh1u−γ1,h2v−γ2 ∈ L∞(Ω), we use Lemma2.5 to obtainu,v ∈ L∞(Ω)and hence, from Lemma2.6 we obtain u,v ∈ C1,α(Ω), for some α ∈ (0, 1). Furthermore, note that using (3.3) we have
kuk∞ ≤C∗1kh1u−γ1k
1 p1−1
∞ ≤C1∗kh1k
1 p1−1
∞ C−γ1
1
p1−1
3 kφ0k−γ1
1
p1−1
∞
and
kvk∞ ≤C∗2kh2v−γ2k
p21−1
∞ ≤C2∗kh2k
p21−1
∞ C−γ2
1
p2−1
4 kφ0k−γ1
1
p1−1
∞ .
So, choosingkhik∞ sufficiently small, withi=1, 2, we obtain kuk∞ ≤δ< 1
2 and kvk∞ ≤δ < 1 2.
Now, sincekuk∞ andkvk∞are small it follows from (3.1), (3.2), (3.4) and (3.5) that
−div(a1(|∇u|p1)|∇u|p1−2∇u) =h1(x)u−γ1 ≥h1(x)kuk−∞γ1 ≥h1(x)
=−div(a1(|∇u|p1)|∇u|p1−2∇u)
and
−div(a2(|∇v|p2)|∇v|p2−2∇v) =h2(x)v−γ2 ≥h2(x)kvk−∞γ2 ≥ h2(x)
=−div(a2(|∇v|p2)|∇v|p2−2∇v).
Therefore, applying the Weak Comparison Principle for the p&q-Laplacian operator we con- clude that
0<u(x)≤ u(x) a.e. in Ω and 0<v(x)≤v(x) a.e. inΩ, which proves (iii).
Our final task is to check that the condition(iv)holds. Indeed, we invoke (F1), (i), (3.1) and (3.2) to obtain
−div(a1(|∇u|p1)|∇u|p1−2∇u)−h1(x)u−γ1−Fu(x,u,v)
≤2h1(x)−h1(x)u−γ1 ≤ h1(x)(2− kuk−∞γ1)≤0 and
−div(a2(|∇v|p2)|∇v|p2−2∇v)−h2(x)v−γ2 −Fv(x,u,v),
≤2h2(x)−h2(x)v−γ2 ≤h2(x)(2− kvk−∞γ2)≤0, which implies that (u,v)is a subsolution for system (1.1). Finally, we use(F1), (ii), (iii), (3.4) and (3.5) to get
−div(a1(|∇u|p1)|∇u|p1−2∇u)−h1(x)u−γ1−Fu(x,u,v)≥h1(x)(u−γ1−u−γ1)≥0 and
−div(a2(|∇v|p2)|∇v|p2−2∇v−h2(x)v−γ2−Fv(x,u,v)≥ h2(x)(v−γ2−v−γ2)≥0, which shows that(u,v)is a supersolution for system (1.1).
Following the same idea in [4] (see also [3]), we introduce the truncation operators T : W01,q1(Ω)→ L∞(Ω)andS:W01,q2(Ω)→ L∞(Ω)given by
Tu(x) =
u(x), ifu(x)> u(x)
u(x), ifu(x)≤ u(x)≤u(x) u(x), ifu(x)< u(x)
(3.6)
and
Sv(x) =
v(x), ifv(x)>v(x)
v(x), ifv(x)≤v(x)≤ v(x) v(x), ifv(x)<v(x).
(3.7)
It is well that the truncation operatorsTandSare continuous and bounded. Now, we consider the following functions
Gu(x,u,v) =h1(x)(Tu)−γ1+Fu(x,Tu,Sv) (3.8) and
Gv(x,u,v) =h2(x)(Sv)−γ2 +Fv(x,Tu,Sv) (3.9)
and the auxiliary problem
−div(a1(|∇u|p1)|∇u|p1−2∇u) =Gu(x,u,v)in Ω,
−div(a2(|∇v|p2)|∇v|p2−2∇v) =Gv(x,u,v)inΩ, u,v>0 inΩ,
u=v=0 on∂Ω.
(3.10)
Define the energy functionalΦ:X→Rassociated with problem (3.10) by Φ(u,v) = 1
p1 Z
Ω
A1(|∇u|p1)dx+ 1 p2
Z
Ω
A2(|∇v|p2)dx−
Z
Ω
G(x,u,v)dx, ∀(u,v)∈ X,
whereG(x,s,t) =Rs
0 Gξ(x,ξ,t)dξ+Rt
0Gξ(x,s,ξ)dξ.
It follows from Lemma3.1(i)–(iii), (3.8), (3.9) and(F1)that
|Gu(x,u,v)| ≤K1 a.e. in Ω, for someK1 >0,∀(u,v)∈ X. (3.11) Similarly,
|Gv(x,u,v)| ≤K2 a.e. inΩ, for someK2>0,∀(u,v)∈X. (3.12) Consequently, we use(A1)to show that the functional Φis well defined and it is of classC1 on Sobolev spaceXwith
Φ0(u,v)(φ,ϕ) =
Z
Ω
a1(|∇u|p1)|∇u|p1−2∇u∇φ+a2(|∇v|p2)|∇v|p2−2∇v∇ϕ dx
−
Z
Ω
Gu(x,u,v)φdx−
Z
Ω
Gv(x,u,v)ϕdx,∀(u,v),(φ,ϕ)∈ X.
Next, consider
M ={(u,v)∈ X;u≤u≤ ua.e. inΩandv≤ v≤va.e. in Ω}.
We claim that Φ is bounded from below in M. Indeed, for all (u,v) ∈ X, we use (A1), (3.11),(3.12)and continuous embeddingW01,qi(Ω),→ L1,qi(Ω), fori=1, 2, to obtain thatΦis coercive inM. Moreover, since(A3)holds andGu,Gv∈ L∞(Ω)we have thatΦis weak lower semi-continuous on M. Thus, as M is closed and convex in X, we use [15, Theorem 1.2] to conclude thatΦis bounded from below in Mand attains it is infimum at a point(u,v)∈ M.
Using the same arguments as in the proof of [15, Theorem 2.4], we see that this minimum point (u,v) is a weak solution of problem (3.10). Indeed, for all φ,ϕ ∈ C∞0 (Ω)andε > 0, let the functionsuε,vε ∈ M be given by
uε(x) =
u(x), u(x) +εφ(x)>u(x)
u(x) +εφ(x), u(x)≤u(x) +εφ(x)≤u(x) u(x), u(x) +εφ(x)<u(x)
and
vε(x) =
v(x), v(x) +εϕ(x)>v(x)
v(x) +εϕ(x), v(x)≤v(x) +εϕ(x)≤ v(x) v(x), v(x) +εϕ(x)<v(x).