• No results found

Existence of positive solutions of elliptic mixed boundary value problem

N/A
N/A
Protected

Academic year: 2020

Share "Existence of positive solutions of elliptic mixed boundary value problem"

Copied!
11
0
0

Loading.... (view fulltext now)

Full text

(1)

R E S E A R C H

Open Access

Existence of positive solutions of elliptic

mixed boundary value problem

Guofa Li

*

*Correspondence:

[email protected] Department of Mathematics and Information Science, Qujing Normal University, Qujing, 655011, P.R. China

Abstract

In this paper, we use variational methods to prove two existence of positive solutions of the following mixed boundary value problem:

⎧ ⎪ ⎨ ⎪ ⎩

u=f(x,u), x

,

u= 0, x

σ

,

∂u

∂ν=g(x,u), x

.

One deals with the asymptotic behaviors off(x,u) near zero and infinity and the other deals with superlinear off(x,u) at infinity.

MSC: 35M12; 35D30

Keywords: elliptic mixed boundary value problem; positive solutions; mountain pass theorem; Sobolev embedding theorem

1 Introduction and preliminaries

This paper is concerned with the existence of positive solutions of the following elliptic mixed boundary value problem:

⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩

u=f(x,u), x,

u= , xσ,

∂u

∂ν =g(x,u), x,

()

whereis a bounded domain inRnwith Lipschitz boundary,σ=,σ=Ø,

is a sufficiently smooth (n– )-dimensional manifold, andνis the outward normal vector on. We assumef :×R→R,g:×R→Rare continuous and satisfy

(S) f(x,t)≥,∀t≥,x,f(x, ) = .f(x,t)≡,∀t< ,x. (S) For almost everyx, f(xt,t)is nondecreasing with respect tot> . (S) limt→f(xt,t)=p(x),limt→+∞f(xt,t)=q(x)≡uniformly in a.e.x, where

p(x)∞<λ,λis the first eigenvalue of (),≤p(x),q(x)∈L∞().

(S) There existsc,c> such that|f(x,t)| ≤c+c|t|p–for somep∈(,n–n)asn≥

andp∈(, +∞)asn= , .

(2)

The eigenvalue problem of () is studied by Liu and Su in []

⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩

u=λu in,

u=  onσ,

∂u

∂ν =λu on.

()

There exists a set of eigenvalues {λk} and corresponding eigenfunctions {uk} which

solve problem (), where  ≤λ ≤λ ≤ · · · ≤λk ≤ · · ·, λk → ∞ as k → ∞, λ = inf=uV

|∇u|dx

|u|dx+

|u|ds.

There have been many papers concerned with similar problems at resonance under the boundary condition; see [–]. Moreover, some multiplicity theorems are obtained by the topological degree technique and variational methods; interested readers can see [–]. Problem () is different from the classical ones, such as those with Dirichlet, Neuman, Robin, No-flux, or Steklov boundary conditions.

In this paper, we assumeV:={vH() :v|

σ = }is a closed subspace ofH(). We

define the norm in V asu=|∇u|dx+|γu|ds, · Lp() is the Lp() norm,

· Lp() is theLp() norm,γ :VL() is the trace operator with γu=u for all

uH(), that is continuous and compact (see []). Furthermore, we define g=γf, ≤g(x,t)≤ |γf(x,t)|fort>  (see []). Then, by (S), we obtain

lim

t→+∞ g(x,t)

tt→lim+∞

|γf(x,t)|

t =q(x)≡, a.e.x. ()

Letbe a bounded domain with a Lipschitz boundary; there is a continuous embedding

V Ly() fory[,n

n–] whenn≥, andy∈[, +∞) whenn= , . Then there exists γy> , such that

uLy()≤γyu, ∀uV. ()

Moreover, there is a continuous boundary trace embeddingVLz() forz∈[,(nn––)] whenn≥, andz∈[, +∞) whenn= , . Then there existskz> , such that

uLz()kzu, ∀u∈V. ()

It is well known that to seek a nontrivial weak solution of problem () is equivalent to finding a nonzero critical value of theCfunctional

J(u) =  

|∇u|dx

F(x,u)dx

G(s,u)ds, ()

whereuV,F(x,u) =uf(x,t)dt,G(x,u) =ug(x,t)dt. Moreover, by (S) and the Strong maximum principle, a nonzero critical point ofJis in fact a positive solution of (). In order to find critical points of the functional (), one often requires the technique condition, that is, for someμ> ,∀|u| ≥M> ,x,

 <μF(x,u)≤uf(x,u), F(x,u) =

u

(3)

It is easy to see that the condition (AR) implies thatlimu→+∞F(ux,u)= +∞, that is,f(x,u)

must be superlinear with respect touat infinity. In the present paper, motivated by [] and [], we study the existence and nonexistence of positive solutions for problem () with the asymptotic behavior assumptions (S) offat zero and infinity. Moreover, we also study superlinear off at infinity withq(x)≡+∞in (S), which is weaker than the (AR) condition, that is the (AR) condition does not hold.

In order to get our conclusion, we define the minimization problem

=inf

|∇u|dx:uV,

q(x)udx+

q(s)uds=  , ()

then> , which is achieved by someϕV withϕ(x) >  a.e. in; see Lemma .

We denote byc,c,cuniversal constants unless specified otherwise. Our main results are as follows.

Theorem  Let conditions (S) to (S) hold, then:

(i) If> , then the problem () has no any positive solution inV. (ii) If< , then the problem () has at least one positive solution inV.

(iii) If= , then the problem () has one positive solutionu(x)∈Vif and only if there exists a constantc> such thatu(x) =(x)andf(x,u) =q(x)u(x),

g(x,u) =q(x)u(x)a.e.x, whereϕ(x) > is the function which achieves.

Corollary  Let conditions (S) to (S) with q(x)≡l> hold, then:

(i) Ifl<λ, then the problem () has no any positive solution inV.

(ii) Ifλ<l< +∞, then the problem () has at least one positive solution inV. (iii) Ifl=λ, then the problem () has one positive solutionu(x)∈Vif and only if there

exists a constantc> such thatu(x) =(x)andf(x,u) =λu(x),g(x,u) =λu(x) a.e.x, whereϕ(x) > is the eigenfunction of theλ.

Theorem  Let conditions (S) to (S) with q(x)≡+∞hold, then the problem () has at least one positive solution in V .

2 Some lemmas

We need the following lemmas.

Lemma  If q(x)∈L∞(), q(x)≥, q(x)≡, then> and there existsϕ(x)∈V such

that=|∇ϕ|dx and

q(x)ϕ

dx+

q(s)ϕ

ds= . Moreover,ϕ(x) > a.e. in V .

Proof By the Sobolev embedding functionV L() and Fatou’s lemma, it is easy to know that >  and there existsϕ(x)∈V, which satisfies, that is,

q(x)ϕ

dx+

q(s)ϕds= . Furthermore, we assumeϕ(x)≥, thenϕ(x) could replace by|ϕ(x)|.

By the Strong maximum principle, we knowϕ(x) >  a.e. inV.

Lemma  If conditions (S) to (S) hold, then there existsβ,ρ> such that J|∂Bρ()≥β,uV ,u=ρ.

Proof By condition (S), there existsδ> ,ε>  such that f(xu,u)≤λ–ε, g(xu,u)≤ γf(ux,u) ≤ λ–εas  <|u| ≤δ. Which implies thatF(x,u)≤(λ–ε)u+c|u|y,G(x,u)≤(λ–ε)u+

(4)

By () and (), we obtain

J(u) = 

∇u

L()

F(x,u)dx

G(s,u)ds

≥ 

∇u

L()+

 γu

L()

 γu

L()

(λ–ε)u 

L()

cuyLy()– 

(λ–ε)u 

L()cuzLz()

≥ 

u

(λε)  λ

u–yyuy–

(λε+ )  λ+ 

u–ckzzuz

=

ε(λ+ ) λ(λ+ )

– 

u–yyuyckz zuz.

Hence,y,z> ; we takeεwhich satisfies ε(λ+) λ(λ+)–

 > , that is,ε> λ(λ+)

λ+ . Then we

take a positive constantβsuch thatJ|∂Bρ()≥βasu=ρ, and is small enough.

Lemma  If conditions (S) to (S) hold,< (x) > is defined by Lemma , then

J((x))→–∞as t→+∞.

Proof If< ,ϕ(x) >  is defined by Lemma , by Fatou’s lemma, and (S), we have

lim

t→+∞

J((x))

t

= 

ϕ(x) 

dx– lim

t→+∞

F(x,(x))dx

t –tlim+

G(s,(s))ds

t

≤

ϕ(x) 

dx

lim

t→+∞

F(x,(x))

tϕ(x)

ϕ(x)dx

lim

t→+∞

G(s,(s))

tϕ(s)

ϕ(s)ds

= 

ϕ(x)dx

 

f(x,(x))

(x)

ϕ(x)dx– 

g(s,(s))

(s)

ϕ(s)ds

= 

ϕ(x) 

dx– 

q(x)ϕ(x)dx+

q(s)ϕ(s)ds

= 

(– )

ϕ(x) 

dx

< .

So,J((x))→–∞ast→+∞.

Lemma  Let conditions (S) and (S) hold. If a sequence{un} ⊂V satisfiesJ(un),un

as n→+∞, then there exists a subsequence of{un}, still denoted by{un}such that J(tun)≤

+t

n +J(un)for all t> , n≥.

Proof SinceJ(un),un → asn→+∞, for a subsequence, we may assume that

–

n<

J(un),un

=∇unL()

f(x,un)undx

g(s,un)unds<

(5)

For any fixedxandn≥, set

ψ(t) =

t

f(x,un)unF(x,tun), ψ(t) =

t

g(s,un)unG(s,tun).

Then (S) implies that

ψ(t) =tf(x,un)unf(x,tun)un

=tun

f(x,un) –

f(x,tun)

t = ⎧ ⎨ ⎩

≥,  <t≤;

≤, t> .

It implies thatψ(t)≤ψ(),∀t> . Following the same procedures, we obtainψ(t)≤ ψ(),∀t> .

For allt>  and positive integern, by (), we have

J(tun) =

t

∇un

L()

F(x,tun)dx

G(s,tun)ds

t

  n+

f(x,un)undx+

g(s,un)unds

F(x,tun)dx

G(s,tun)ds

t

n+

f(x,un)unF(x,un)

dx+

g(s,un)unG(s,un)

ds. ()

On the other hand, by (), one has

J(un) =

 ∇un

L()

F(x,un)dx

G(s,un)ds

≥   – n+

f(x,un)undx+

g(s,un)unds

F(x,un)dx

G(s,un)ds

= –  n+

f(x,un)unF(x,un)

dx+

g(s,un)unG(s,un)

ds. One has

f(x,un)unF(x,un)

dx+

g(s,un)unG(s,un)

dsJ(un) +

n. ()

Combining () and (), we haveJ(tun)≤+t

n +J(un).

Lemma (see []) Suppose E is a real Banach space, JC(E,R)satisfies the following

geometrical conditions:

(6)

(ii) There existseE\Bρ()such thatJ(e)≤. Letbe the set of all continuous paths joiningande:

=

hC[, ],E|h() = ,h() =e,

and

c= inf

h max

t∈[,]J

h(t).

Then there exists a sequence {un} ⊂ E such that J(un)→cβ and ( +un

J(u

n)E∗→.

3 Proofs of main results

Proof of Theorem  (i) IfuVis one positive solution of problem (), by (), one has

 =J(u),u=

|∇u|dx

f(x,u)u dx

g(s,u)u ds.

That is,

|∇u|dx=

f(x,u)u dx+

g(s,u)u ds

q(x)udx+

q(s)uds= .

It implies that≤. This completes the proof of Theorem (i).

(ii) By Lemma , there existsβ,ρ>  such thatJ|∂Bρ()≥βwithu=ρ. By Lemma ,

we obtainJ(tϕ(x)) <  ast→+∞. Define

=hC[, ],V|h() = ,h() =tϕ(x)

, ()

c= inf

htmax∈[,]J

h(t), ()

whereϕ(x) >  is given by Lemma . Thencβ>  and by Lemma , there exists{un} ⊂

V such that

J(un) =

 ∇un

L()

F(x,un)dx

G(s,un)ds=c+o(), ()

 +unJ(un)V∗→. ()

() implies that

J(un),un

=∇unL()

f(x,un)undx

g(s,un)unds=o(). ()

(7)

If{un}is bounded inV, whenis bounded andf(x,u),g(x,u) are subcritical, we can get

{un}has a subsequence strong convergence to a critical value ofJ, and our proof is

com-plete. So, to prove the theorem, we only need show that{un}is bounded inV. Supposing

that{un}is unbounded, that is,un →+∞asn→+∞. We order

tn=

√c un

, wn=tnun=

√cun

un

. ()

Then{wn}is bounded inV. By extracting a subsequence, we supposewnwis a strong

convergence inL(),w

nwis a convergence a.e.x,wnwis a weak convergence

inV.

We claim thatw= . In fact, by (S) and (S), we know∀x∈,un≥, and there exists

M,M>  such that|f(xu,unn)| ≤M,|g(xu,unn)| ≤M. Ifw= ,wn→ is a strong convergence

inL(), and by () and () we know

c = tnun=tn

∇unL()+γunL()

= tn

f(x,un)undx+tn

g(s,un)unds+tnγunL()+o()

=

f(x,un)

un

wndx+

g(s,un)

un

wnds+tnunL()+o()

M

wndx+M

wnds+wnL()+o()

→.

It is contradiction withc> , sow= . As follows, we provew=  satisfies

ϕ(x)∇w(x)dx

q(x)ϕ(x)w(x)dx

q(s)ϕ(s)w(s)ds= .

We order

pn(x) =

⎧ ⎨ ⎩

f(x,un)/un, un≥,x,

, un< ,x,

qn(x) =

⎧ ⎨ ⎩

g(x,un)/un, un≥,x,

, un< ,x.

By (S) and (S), there existsM>  such that ≤pn(x)≤M, ≤qn(x)≤M,∀x∈. We select a suitable subsequence and there existsh(x)∈L(),h(x)∈L() such that

pn(x)→h(x) is a strong convergence inL(),qn(x)→h(x) is a strong convergence in

L(), and h

(x)≤M, ≤h(x)≤M,∀x∈.

It follows fromwnwis a strong convergence inL() that

pn(x)wn(x)ϕ(x)dx=

pn(x)w+n(x)ϕ(x)dx

h(x)w+(x)ϕ(x)dx,

qn(s)wn(s)ϕ(s)ds=

qn(s)w+n(s)ϕ(s)ds

(8)

Hence,{pn(x)wn(x)}is bounded inL(),pn(x)wn(x)h(x)w+(x) inL();{qn(x)wn(x)}

is bounded inL(),qn(x)wn(x)h(x)w+(x) inL(). By (), we have

wn(x)∇ϕ(x)dx

pn(x)wn(x)ϕ(x)dx

qn(s)wn(s)ϕ(s)ds

=

tnun(x)

ϕ(x)dx

pn(x)tnun(x)ϕ(x)dx

qn(s)tnun(s)ϕ(s)ds

= 

c un

un(x)∇ϕ(x)dx

pn(x)un(x)ϕ(x)dx

qn(s)un(s)ϕ(s)ds

→.

Sincewnwis a weak convergence inV, we obtain

ϕ(x)∇w(x)dx

h(x)ϕ(x)w+(x)dx

h(s)ϕ(s)w+(s)ds= , ϕV.

We orderϕ=w; this yieldsw= , sow=w+. By the Strong maximum principle, we knoww>  a.e. in, soun→ ∞a.e. in. Combining (S) and (), we obtain

ϕ(x)∇w(x)dx

q(x)ϕ(x)w(x)dx

q(s)ϕ(s)w(s)ds= , ∀ϕV.

This is a contradiction with< . This completes the proof of Theorem (ii). (iii) If= , by Lemma , there exists someϕ(x) > , such that

∇v(x)∇ϕ(x)dx=

q(x)v(x)ϕ(x)dx+

q(s)v(s)ϕ(s)ds. ()

Ifuis a positive solution of (), for the aboveϕ(x), we have

∇u(x)∇ϕ(x)dx=

fx,u(x)ϕ(x)dx+

gs,u(s)ϕ(s)ds. ()

We orderv=uin (), and it follows from () that

∇u(x)∇ϕ(x)dx=

q(x)u(x)ϕ(x)dx+

q(s)u(s)ϕ(s)ds

=

fx,u(x)ϕ(x)dx+

gs,u(s)ϕ(s)ds

q(x)u(x)ϕ(x)dx+

q(s)u(s)ϕ(s)ds,

which implies that(f(x,u) –q(x)u(x))ϕ(x)dx+

(g(s,u) –q(s)u(s))ϕ(s)ds= .

Whenϕ(x) >  a.e. in, combining (S), (S), and (), we obtain

f(x,u)≤q(x)u(x), g(x,u)≤q(x)u(x).

Then we must havef(x,u) =q(x)u(x),g(x,u) =q(x)u(x) a.e. in,u(x) >  also achieves (= ). Whenu=,c> , we have

|∇ϕ|dx=

q(x)ϕdx+

q(s)ϕds, which

(9)

On the other hand, if for somec> ,u(x) =(x) andf(x,(x)) =cq(x)ϕ(x),g(x,u) =

cq(x)ϕ(x) a.e.x, since(x) also achieves. This meansu(x) =(x) is a solution

of problem () as= . This completes the proof of Theorem (iii).

Proof of Corollary  Note that whenq(x)≡l, then=λ

l. The conclusion follows from

Theorem .

Proof of Theorem  Whenq(x)≡+∞, we can replaceϕbyϕin () and definecas in (), then following the same procedures as in the proof of Theorem (ii), we need to show only that{un}is bounded inV. For this purpose, let{wn}be defined as in (). If{wn}is

bounded inV, we knowwnwis a strong convergence inL(),wnwis convergence

a.e.x,wnwis a weak convergence inV, andwV.

Ifun →+∞, thentn→ andw(x)≡. We set={x∈:w(x) = },={x∈:

w(x)= }. Obviously, by (),|un| →+∞a.e. in. Whenq(x)≡+∞in (S), there exists

K,K>  andnlarge enough we have|f(xu,unn)| ≥K,|g(xu,nun)| ≥K uniformly in x. Hence, by () and (), we obtain

c= lim

n→+∞t

nun

= lim

n→+∞t

n

∇unL()+γunL()

= lim

n→+∞t

n

f(x,un)undx+

g(s,un)unds+γunL()

= lim

n→+∞

f(x,un)

un

wndx+

g(s,un)

un

wnds+tnγunL()

K

wdx+K

wds+wL().

Noticing thatw(x)=  inandK,K can be chosen large enough, som≡ and thenw(x)≡ in.

Then we knowlimn→+∞

F(x,wn)dx+limn→+∞

G(s,wn)ds= , and consequently,

J(wn) =

 ∇wn

L()+o()

=  wn

 wn

L()+o()

≥ 

 –  λ+ 

wn+o()

= c

 –  λ+ 

+o(). ()

Byun →+∞,tn→ asn→+∞, then it follows Lemma  and (), we obtain

J(wn) =J(tnun)≤

 +tn

nc. ()

Obviously, () and () are contradictory. So{un}is bounded inV. This completes the

(10)

4 Example

In this section, we give two examples onf(x,u): One satisfies (S) to (S) withq(x)≡+∞, but does not satisfy the (AR) condition; the other illustrates how the assumptions on the boundary are not trivial and compatible with the inner assumptions in.

Example  Set:

f(x,t) =

⎧ ⎨ ⎩

, t≤;

tln( +t), t> .

Then it is easy to verify thatf(x,t) satisfies (S) to (S) withp(x) =  ast→ andq(x) = +∞ast→+∞. In addition,

F(x,t) =  t

ln( +t) –  t

+ t

ln( +t).

So, for someμ> ,μF(x,t) =tln( +t)(μlnμ(+t)+tlnμ(+t)μt) >tln( +t), for allt

large.

This meansf(x,t) does not satisfy the (AR) condition.

Example  Consider the following problem:

⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩

u(x) =αu(x),  <x<l,

u() = ,

u(l) =αu(l),

()

whereα>  is a constant. It is obvious thatg=γf asf(x,u) =αu(x). Problem () is a case of (); we can obtain the nontrivial solution:u(x) =Csin√αx,C= .

Competing interests

The author declares that he has no competing interests.

Author’s contributions

Li G carried out all studies in this article.

Acknowledgements

The author would like to thank the referees for carefully reading this article and making valuable comments and suggestions.

Received: 19 January 2012 Accepted: 6 August 2012 Published: 16 August 2012 References

1. Liu, H, Su, N: Well-posedness for a class of mixed problem of wave equations. Nonlinear Anal.71, 17-27 (2009). doi:10.1016/j.na.2008.10.027

2. Landesman, EM, Lazer, AC: Nonlinear perturbations of linear elliptic boundary value problems at resonance. J. Math. Mech.19, 609-623 (1970)

3. Ahmad, S, Lazer, AC, Paul, JL: Elementary critical point theory and perturbations of elliptic boundary value problems at resonance. Indiana Univ. Math. J.25, 933-944 (1976)

4. Berestycki, H, De Figueiredo, DG: Double resonance in semilinear elliptic problems. Commun. Partial Differ. Equ.6, 91-120 (1981)

5. Costa, DG, Oliveira, AS: Existence of solution for a class of semilinear elliptic problems at double resonance. Bol. Soc. Bras. Mat.19, 21-37 (1988)

6. Omari, P, Zanolin, F: Resonance at two consecutive eigenvalues for semilinear elliptic equations. Ann. Mat. Pura Appl.

(11)

7. Binding, PA, Drábek, P, Huang, YX: Existence of multiple solutions of critical quasilinear elliptic Neumann problems. Nonlinear Anal.42, 613-629 (2000). doi:10.1016/S0362-546X(99)00118-2

8. Li, G, Zhou, HS: Asymptotically linear Dirichlet problem for thep-Laplacian. Nonlinear Anal.43, 1043-1055 (2001). doi:10.1016/S0362-546X(99)00243-6

9. Escobar, JF: A comparison theorem for the first non-zero Steklov eigenvalue. J. Funct. Anal.178, 143-155 (2000). doi:10.1006/jfan.2000.3662

10. Kaur, BS, Sreenadh, K: Multiple positive solutions for a singular elliptic equation with Neumann boundary condition in two dimensions. Electron. J. Differ. Equ.43, 1-13 (2009)

11. Ahmad, S: Multiple nontrivial solutions of resonant and nonresonant asymptoticaly problems. Proc. Am. Math. Soc.

96, 405-409 (1986)

12. Chang, KC: Infinite Dimensional Morse Theory and Multiple Solutions Problems. Birkhäuser, Boston (1993) 13. Hirano, N, Nishimura, T: Multiplicity results for semilinear elliptic problems at resonance and with jumping

nonlinearities. J. Math. Anal. Appl.180, 566-586 (1993). doi:10.1006/jmaa.1993.1417

14. Landesman, E, Robinson, S, Rumbos, A: Multiple solution of semilinear elliptic problem at resonance. Nonlinear Anal.

24, 1049-1059 (1995). doi:10.1016/0362-546X(94)00107-S

15. Liu, SQ, Tang, CL, Wu, XP: Multiplicity of nontrivial solutions of semilinear elliptic equations. J. Math. Anal. Appl.249, 289-299 (2000). doi:10.1006/jmaa.2000.6704

16. Li, S, Willem, M: Multiple solution for asymptotically linear boundary value problems in which the nonlinearity crosses at least one eigenvalue. Nonlinear Differ. Equ. Appl.5, 479-490 (1998)

17. Mawhin, J, Willem, M: Critical Point Theory and Hamiltonien Systems. Springer, New York (1989)

18. Lions, JL, Magenes, E: Non-Homogeneous Boundary Value Problems and Applications. Springer, Berlin (1972) 19. Zhou, H: An application of a mountain pass theorem. Acta Math. Sin. Engl. Ser.18(1), 27-36 (2002)

20. Cheng, BT, Wu, X: Existence results of positive solutions of Kirchhoff type problems. Nonlinear Anal.71, 4883-4892 (2009). doi:10.1016/j.na.2009.03.065

21. Ambrosetti, A, Rabinowitz, PH: Dual variational methods in critical points theory and applications. J. Funct. Anal.14, 349-381 (1973). doi:10.1016/0022-1236(73)90051-7

doi:10.1186/1687-2770-2012-91

References

Related documents