In this paper, p-biharmonic equations involving Hardy potential and negative exponents with a parameter λ are considered. By means of the structure and properties of Nehari manifold, we give uniform lower bounds for \(\varLambda >0\), which is the supremum of the set of λ. When \(\lambda \in (0, \varLambda )\), the above problems admit at least two positive solutions.
Hinweise
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
1 Introduction and preliminaries
In this paper, we consider a p-biharmonic equation with Hardy potential and negative exponents:
where \(0\in \varOmega \subset \mathbb{R}^{N}\) is a bounded smooth domain with \(1< p<\frac{N}{2}\), \(\Delta ^{2}_{p}u=\Delta (\vert \Delta u\vert ^{p-2} \Delta u)\) is the p-biharmonic operator. \(\lambda >0\) is a parameter, \(0<\mu <\mu _{N,p}=(\frac{(p-1)N(N-2p)}{p^{2}})^{p}\), \(0< q<1\) and \(p-1<\gamma <p^{*}-1\), where \(p^{*}=\frac{Np}{N-2p}\) is called the critical Sobolev exponent. \(f(x)\geq 0\), \(f(x)\not \equiv 0\), \(g(x)\) satisfies the requirement that the set \(\{x\in \varOmega : g(x)>0 \}\) has positive measures, \(\operatorname{supp}f \cap \{x\in \varOmega : g(x)>0 \} \neq \emptyset \) and \(f, g\in C(\overline{\varOmega })\). Biharmonic equations describe the sport of a rigid body and the deformations of an elastic beam. For example, this type of equation provides a model for considering traveling wave in suspension bridges [5, 16, 27, 30, 36]. Various methods and tools have been adopted to deal with singular problems, such that fixed point theorems [14], topological methods [37], Fourier and Laurent transformation [18, 19], monotone iterative methods [21], global bifurcation theory [12], and degree theory [22, 31].
In recent years, there was much attention focused on the existence, multiplicity and qualitative properties of solutions for p-biharmonic equations under Dirichlet boundary conditions or Navier boundary conditions with Hardy terms [4, 15, 17, 32, 34]. Xie and Wang [32] studied the following p-biharmonic equation with Dirichlet boundary conditions:
where \(\frac{\partial }{\partial n}\) is the outer normal derivative. By using the variational method, the existence of infinitely many solutions with positive energy levels for (1.2) was established. Huang and Liu [15] considered the following p-biharmonic equation with Navier boundary conditions:
where \(1< p<\frac{N}{2}\). By using invariant sets of gradient flows, the authors proved that (1.3) possesses a sign-changing solution. Furthermore, Yang, Zhang and Liu [34] showed that (1.3) has a positive solution, a negative solution and a sequence of sign-changing solutions when f satisfies appropriate conditions. Bhakta [4] established the qualitative properties of entire solutions for a noncompact problem related to p-biharmonic type equations with Hardy terms.
Anzeige
On the other hand, nonlinear biharmonic equations with negative exponents have been studied expensively [1, 6, 8, 13, 20]. Guerra [13] gave a complete description of entire radially symmetric solutions for the following biharmonic equation:
where \(\varOmega \subset \mathbb{R}^{N}\) (\(N\geq 1\)) is a bounded \(C^{4}\)-domain, λ and μ are positive parameters and \(0<\alpha <1\), \(0<\gamma <1\) are constants. Here M, h and k are given continuous functions satisfying suitable hypotheses. By using the Galerkin method and the sharp angle lemma, the authors proved that problem (1.4) has a positive solution for \(0<\mu < (\frac{N(N-4)}{4} ) ^{2}\).
We say that \(u\in W:=W^{2,p}(\varOmega )\cap W_{0}^{1,p}(\varOmega )\) is a weak solution of (1.1), if for every \(\varphi \in W\), there holds
$$ \int _{\varOmega } \vert \Delta u \vert ^{p-2}\Delta u \Delta \varphi \,dx- \int _{\varOmega }\frac{\mu }{ \vert x \vert ^{2p}} \vert u \vert ^{p-2}u \varphi \,dx= \int _{\varOmega }f(x)u^{-q}\varphi \,dx+\lambda \int _{\varOmega } g(x)u^{\gamma }\varphi \,dx. $$
(1.5)
The following Rellich inequality will be used in this paper:
$$ \int _{\varOmega } \vert \Delta u \vert ^{p}\,dx\geq \mu _{N,p} \int _{\varOmega }\frac{ \vert u \vert ^{p}}{ \vert x \vert ^{2p}}\,dx, \quad \forall u\in W, $$
and it is not achieved [9, 24]. For any \(u\in W\), and \(0<\mu <\mu _{N,p}\). The energy functional corresponding to (1.1) is defined by
$$ \begin{aligned}[b] I_{\lambda ,\mu }(u)={}&\frac{1}{p} \int _{\varOmega } \biggl( \vert \Delta u \vert ^{p}- \frac{\mu }{ \vert x \vert ^{2p}} \vert u \vert ^{p} \biggr)\,dx- \frac{1}{1-q} \int _{\varOmega }f(x) \vert u \vert ^{1-q}\,dx \\ &{}- \frac{\lambda }{\gamma +1} \int _{\varOmega }g(x) \vert u \vert ^{\gamma +1}\,dx. \end{aligned} $$
(1.6)
For \(\mu \in [0,\mu _{N,p})\), W is equipped with the following norm:
$$ \Vert u \Vert ^{p}_{\mu }= \int _{\varOmega } \biggl( \vert \Delta u \vert ^{p}- \frac{\mu }{ \vert x \vert ^{2p}} \vert u \vert ^{p} \biggr)\,dx. $$
Negative exponent term \(u^{-q}\) implies that \(I_{\lambda ,\mu }\) is not differential on W, therefore, critical point theory cannot be applied to the problem (1.1) directly. We consider the following manifold:
$$ \mathcal{M}= \biggl\{ u\in W: \Vert u \Vert ^{p}_{\mu } = \int _{\varOmega }f(x) \vert u \vert ^{1-q}\,dx+\lambda \int _{\varOmega } g(x) \vert u \vert ^{\gamma +1}\,dx \biggr\} , $$
and make the following splitting for \(\mathcal{M}\):
In this paper, we will study the dependence of problem (1.1) on q, γ, f, g and Ω and evaluate the extremal value of λ related to multiplicity of positive solutions for problem (1.1). Our idea comes from [7, 28, 29]. Our results improve and complement previous ones obtained in [23, 25]. Denote \(\Vert u\Vert _{t}^{t}= \int _{\varOmega }\vert u\vert ^{t}\,dx\) and \(D^{2, p}(\mathbb{R}^{N})\) be the closure of \(C_{0}^{\infty }(\mathbb{R}^{N})\) with respect to the norm \((\int _{\mathbb{R}^{N}}\vert \Delta u\vert ^{p} \,dx )^{\frac{1}{p}}\).
Anzeige
\(\lambda _{1}\) denotes the smallest eigenvalue for
$$ \Delta _{p}^{2}u-\frac{\mu }{ \vert x \vert ^{2p}} \vert u \vert ^{p-2}u=\lambda _{1} \vert u \vert ^{p-2}u, \quad x\in \varOmega \setminus \{0\}, u\in W, $$
(1.10)
and \(\varphi _{1}\) denotes the corresponding eigenfunction with \(\varphi _{1}>0\) in Ω [3, 10, 26, 33, 35]. The following minimization problem will be useful in the following discussions:
$$ S_{\mu }=\inf \biggl\{ \int _{\mathbb{R}^{N}} \biggl( \vert \Delta u \vert ^{p}- \frac{\mu }{ \vert x \vert ^{2p}} \vert u \vert ^{p} \biggr)\,dx, u\in D^{2,p} \bigl(\mathbb{R}^{N} \bigr), \int _{\mathbb{R}^{N}} \vert u \vert ^{p^{*}}\,dx=1 \biggr\} >0, $$
(1.11)
and \(S_{\mu }\) is achieved by a family of functions [4, 11]. Thus, for every \(u\in W\setminus \{0\}\), \(\Vert u\Vert _{p^{*}}\leq \frac{ \Vert u\Vert _{\mu }}{\sqrt[p]{S_{\mu }}}\). Therefore, combining with the Hölder inequality, we deduce that
Then problem (1.1) admits at least two solutions\(u_{0}\in \mathcal{M}^{+}\), \(U_{0}\in \mathcal{M}^{-}\), with\(\Vert U_{0}\Vert _{\mu }> \Vert u_{0}\Vert _{\mu }\).
Corollary 1.2
Let\(U_{\lambda , \mu ,\varepsilon } \in \mathcal{M}^{-}\)be the solution of problem (1.1) with\(\gamma = \varepsilon +p-1\), where\(\lambda \in (0,T_{\mu })\). Then
There exists\(\lambda ^{*} =\lambda ^{*} (N, \varOmega , \mu , q, \gamma )>0\)such that problem (1.1) with\(f=g=1\)admits at least a positive solution for every\(0<\lambda <\lambda ^{*}\)and has no solution for every\(\lambda >\lambda ^{*}\).
2 Some lemmas
Lemma 2.1
Assume that\(\lambda \in (0,T_{\mu })\), where\(T_{\mu }\)is defined in (1.15). Then\(\mathcal{M}^{\pm }\neq \emptyset \)and\(\mathcal{M}^{0}=\{0\}\).
Proof
(i) We can choose \(u^{*}\in \mathcal{M}\setminus \{0 \}\) such that \(\int _{\varOmega }f(x)\vert u^{*}\vert ^{1-q}\,dx>0\) and \(\int _{\varOmega }g(x) \vert u^{*}\vert ^{\gamma +1}\,dx>0\) from the conditions imposed on f and g. Denote
It is easy to check that \(\varphi _{\mu }(t)\rightarrow -\infty \) as \(t\rightarrow 0^{+}\) and \(\varphi _{\mu }(t)\rightarrow -\lambda \int _{\varOmega } g(x)\vert u^{*}\vert ^{\gamma +1}\,dx<0\) as \(t\rightarrow \infty \). Furthermore, \(\varphi _{\mu }(t)\) attains its maximum at \(t_{\max }\). By (1.12) and (1.13), we obtain
In turn, this is also true. Hence \(A(\mu ,\lambda )=0\) if and only if \(\lambda =T_{\mu }\). Thus for \(\lambda \in (0,T_{\mu })\), we have \(A(\mu ,\lambda )>0\). Moreover, by (2.2), we derive that \(\varphi _{ \mu }(t_{\max })>0\). Consequently, there exist two numbers \(t_{\mu } ^{-}\) and \(t_{\mu }^{+}\) such that \(0< t_{\mu }^{-}< t_{\max }< t_{ \mu }^{+}\), and
$$ (p-1-\gamma ) \Vert tu \Vert _{\mu }^{p}+(q+\gamma ) \int _{\varOmega }f(x) \vert tu \vert ^{1-q}\,dx>0, $$
i.e.,
$$ (p-1-\gamma ) \Vert tu \Vert _{\mu }^{p}+(q+\gamma ) \biggl[ \Vert tu \Vert _{\mu }^{p} - \lambda \int _{\varOmega }g(x) \vert tu \vert ^{\gamma +1}\,dx \biggr]>0. $$
Note that \(tu\in \mathcal{M}\), we have
$$ (p+q-1) \Vert tu \Vert _{\mu }^{p}-\lambda (q+\gamma ) \int _{\varOmega }g(x) \vert tu \vert ^{\gamma +1}\,dx>0. $$
Thus \(tu\in \mathcal{M}^{+}\). By a similar argument, if \(\varphi _{ \mu }(t)=0\) and \(\varphi '_{\mu }(t)<0\), then \(tu\in \mathcal{M}^{-}\). Therefore, both \(\mathcal{M}^{+}\) and \(\mathcal{M}^{-}\) are non-empty sets for every \(\lambda \in (0,T_{\mu })\).
(ii) We claim that \(\mathcal{M}^{0}=\{0\}\). Otherwise, we suppose that there exists \(u_{*}\in \mathcal{M}^{0}\) and \(u_{*}\neq 0\). Since \(u_{*}\in \mathcal{M}^{0}\), we have
Consequently, \(M_{\mu }(\lambda )=M_{\mu ,0}\) if and only if \(\lambda =T_{\mu }\) and \(N_{\mu }(\lambda )=N_{\mu ,0}\) if and only if \(\lambda =T_{\mu }\) respectively. This completes the proof of Lemma 2.2. □
Lemma 2.3
Assume that\(\lambda \in (0,T_{\mu })\). Then\(\mathcal{M}^{-}\)is a closed set inW-topology.
Proof
We choose a sequence \(\{U_{n}\}\) such that \(\{U_{n}\} \subset \mathcal{M}^{-}\) and \(U_{n}\rightarrow U_{0}\) with \(U_{0} \in W\). Then
that is, \(U_{0}\neq 0\). Combining with Lemma 2.1, we obtain \(U_{0}\notin \mathcal{M}^{0}\). Thus \(U_{0}\in \mathcal{M}^{-}\). Therefore \(\mathcal{M}^{-}\) is a closed set in W-topology for every \(\lambda \in (0,T_{\mu })\). □
Lemma 2.4
For\(u\in \mathcal{M}^{\pm }\), there exist a number\(\varepsilon >0\)and a continuous function\(\widetilde{g}(h)>0\)with\(h\in W\)and\(\Vert h\Vert <\varepsilon \)such that
$$\begin{aligned}& \widetilde{F}(0,1)= \Vert u \Vert _{\mu }^{p}- \int _{\varOmega }f(x) \vert u \vert ^{1-q}\,dx -\lambda \int _{\varOmega }g(x) \vert u \vert ^{\gamma +1}\,dx=0, \end{aligned}$$
and
$$ \widetilde{F}_{s}(0,1)=(p-1+q) \Vert u \Vert _{\mu }^{p}-(q+\gamma )\lambda \int _{\varOmega }g(x) \vert u \vert ^{\gamma +1}\,dx>0. $$
(2.3)
At \((0,1)\), using the implicit function theorem, we know that there exists \(\overline{\varepsilon }>0\) such that for \(h\in W\) and \(\Vert h\Vert <\overline{\varepsilon }\), the equation \(\widetilde{F}(h,s)=0\) has a unique continuous solution \(s=\widetilde{g}(h)>0\). Hence \(\widetilde{g}(0)=1\) and
together with (2.3), these imply that we can choose \(\varepsilon >0\) small enough (\(\varepsilon <\overline{\varepsilon }\)) such that for every \(h\in W\) and \(\Vert h\Vert <\varepsilon \)
Since \(\mathcal{K}''(\Vert u\Vert _{\mu })>0\) for all \(\Vert u\Vert _{\mu }>0\) with \(\mathcal{K}(\Vert u\Vert _{\mu })\rightarrow 0\) as \(\Vert u\Vert _{\mu }\rightarrow 0\) and \(\mathcal{K}(\Vert u\Vert _{\mu })\rightarrow \infty \) as \(\Vert u\Vert _{ \mu }\rightarrow \infty \). Therefore \(\mathcal{K}(u)\) attains its minimum at \((\Vert u\Vert _{\mu })_{\min }\), and
Thus \(I_{\lambda ,\mu }(u)\) is bounded below on \(\mathcal{M}\). According to Lemma 2.3, if \(\lambda \in (0,T_{\mu })\), then \(\mathcal{M}^{+} \cup \mathcal{M}^{0}\) and \(\mathcal{M}^{-}\) are two closed sets in \(\mathcal{M}\). Therefore, we apply the Ekeland variational principle [2] to derive a minimizing sequence \(\{u_{n}\}\subset \mathcal{M}^{+}\cup \mathcal{M}^{0}\) satisfying:
which implies that \(\inf_{\mathcal{M}^{+}}I_{\lambda ,\mu }(u)<0\). For \(\lambda \in (0,T_{\mu })\), it follows from Lemma 2.1 that \(\mathcal{M}^{0}=\{0 \}\). Thus \(u_{n}\in \mathcal{M}^{+}\) for n large enough and \(\inf_{\mathcal{M}^{+}\cup \mathcal{M}^{0}}I_{\lambda ,\mu }(u) =\inf_{\mathcal{M}^{+}}I_{\lambda ,\mu }(u)<0\). Therefore
Set \(\phi \in \mathcal{M}\) with \(\phi \geq 0\). Using Lemma 2.4, there exists \(\widetilde{g}_{n}(t)\) such that \(\widetilde{g}_{n}(0)=1\) and \(\widetilde{g}_{n}(t)(u_{n}+t\phi )\in \mathcal{M}^{+}\). Thus
where \(\widetilde{g}'_{n}(0)\) denotes the right derivative of \(\widetilde{g}_{n}(t)\) at zero. If it does not exist, \(\widetilde{g}'_{n}(0)\) should be replaced by \(\lim_{k\rightarrow \infty }\frac{\widetilde{g}_{n}(t_{k})- \widetilde{g}_{n}(0)}{t_{k}}\) for some sequence \(\{t_{k}\}_{k=1}^{ \infty }\) with \(\lim_{k\rightarrow \infty }t_{k} =0\) and \(t_{k}>0\).
Combining with (3.7) and (3.8), we have \(\widetilde{g}'_{n}(0)\neq - \infty \). Now we prove that \(\widetilde{g}'_{n}(0)\neq +\infty \). Otherwise, we suppose that \(\widetilde{g}'_{n}(0)=+\infty \). Note that \(\widetilde{g}_{n}(t)>\widetilde{g}_{n}(0)=1\) for n large enough, and
Since \(\int _{\varOmega }u_{0}^{-q}\varphi _{1}\,dx<\infty \), we have \(u_{0}>0\) a.e. in Ω. For every \(\phi \in \mathcal{M}\) and \(\phi \geq 0\), we have
where \(\varOmega _{1}=\{x\vert u_{0}(x)+\varepsilon \phi (x)>0, x\in \varOmega \}\) and \(\varOmega _{2}=\{x\vert u_{0}(x)+\varepsilon \phi (x)\leq 0, x\in \varOmega \}\). Since the measure of \(\varOmega _{2}\) tends to zero as \(\varepsilon \rightarrow 0\), we have \(\int _{\varOmega _{2}} \vert \Delta u_{0}\vert ^{p-2}\Delta u_{0} \Delta \phi \,dx \rightarrow 0\) as \(\varepsilon \rightarrow 0\). By the same arguments, we have \(\lambda \varepsilon ^{\gamma }\Vert g\Vert _{\infty } \int _{\varOmega _{2}}\vert \phi \vert ^{\gamma +1}\,dx \longrightarrow 0\) and \(\lambda \int _{\varOmega _{2}}g(x)u_{0}^{\gamma }\phi \,dx\longrightarrow 0\) as \(\varepsilon \rightarrow 0\). Dividing by ε and taking the limit for \(\varepsilon \rightarrow 0\), we deduce that
Therefore \(u_{0}\) is a positive weak solution of problem (1.1).
We adopt the Ekeland variational principle again to derive a minimizing sequence \(U_{n}\subset \mathcal{M}^{-}\) for the minimization problem \(\inf_{\mathcal{M}^{-}} I_{\lambda ,\mu }\) such that for \(U_{n}\in \mathcal{M}\), \(U_{n} \rightharpoonup U_{0}\) weakly in \(\mathcal{M}\) and pointwise a.e. in Ω. By similar arguments to those in (3.4) and (3.6), for \(\lambda \in (0,T_{\mu })\), we have
for n large enough and a positive constant \(C_{5}\). Therefore \(U_{0}>0\) is the positive weak solution of problem (1.1). Furthermore \(U_{0}\in \mathcal{M}\). By (3.14), we obtain
i.e., \(U_{0}\in \mathcal{M}^{-}\). According to Lemma 2.2, we know that problem (1.1) has at least two positive weak solutions \(u_{0}\in \mathcal{M}^{+}\) and \(U_{0}\in \mathcal{M}^{-}\) with \(\Vert U_{0}\Vert _{ \mu }>\Vert u_{0}\Vert _{\mu }\) for every \(\lambda \in (0,T_{\mu })\). This completes the proof of Theorem 1.1.
Let \(U_{\lambda , \mu ,\varepsilon }\in \mathcal{M}^{-}\) be the solution of problem (1.1) with \(\gamma =\varepsilon +p-1\), where \(\lambda \in (0,T_{\mu })\). Then
$$\begin{aligned}& \lambda ^{*}=\lambda ^{*}(N,\varOmega ,\mu ,q,\gamma )=\sup \bigl\{ \lambda >0: \text{problem (5.1) has a positive solution} \bigr\} . \end{aligned}$$
Using Theorem 1.1, we provide uniform estimates for \(\lambda ^{*}(N, \varOmega ,\mu ,q,\gamma )\).
Lemma 5.1
For\(1< p<\frac{N}{2}\), \(0<\mu <\mu _{N,p}\), \(0< q<1<\gamma <p^{*}-1\)and\(\varOmega \in \mathbb{U}\), where\(\mathbb{U}=\{\varOmega \in \mathbb{R}^{N}: \varOmega \textit{ is an open and bounded domain}\}\), we have
(1) Assume that \(\lambda \in (0,\lambda ^{-})\), then problem (5.1) has at least two solutions. By the definition of \(\lambda ^{*}\), we have \(\lambda ^{*}\geq \lambda ^{-}>0\).
(2) Assume that (5.1) has a positive solution u. Integrating over Ω by multiplying (5.1) by \(\varphi _{1}\), we obtain
We may choose \(\lambda =\lambda _{1}^{\frac{\gamma +q}{q-1+p}} (\frac{ \gamma -p+1}{\gamma +q} )^{\frac{\gamma -p+1}{q+p-1}}\frac{-1+p+q}{ \gamma +q}+\frac{1}{2}=\lambda ^{+} >0\) such that
We only prove the case that \(0<\lambda <\lambda ^{*}\). By the definition of \(\lambda ^{*}\), there exists \(\overline{\lambda }\in (\lambda , \lambda ^{*})\) such that the problem
$$\begin{aligned}& \Delta ^{2}_{p}u-\mu \frac{ \vert u \vert ^{p-2}u}{ \vert x \vert ^{2p}}=u^{-q}+ \overline{ \lambda } u^{\gamma } \end{aligned}$$
has a positive solution, denoted by \(u_{\overline{\lambda }}\). It follows that
Hence \(u_{\overline{\lambda }}\) is an upper solution of (5.1). Note that \(\lim_{t\rightarrow 0^{+}}G_{\lambda }(t)=\infty \), we can take \(\varepsilon >0\) small enough with \(\varepsilon \varphi _{1}< u_{\overline{ \lambda }}\) and \(G_{\lambda }(\varepsilon \varphi _{1})\geq 0\). Thus
namely, \(\varepsilon \varphi _{1}\) is a lower solution of (5.1). Note that \(\Delta _{p}^{2}-\frac{\mu }{\vert x\vert ^{2p}}\) is monotone, then problem (5.1) has a positive solution \(u_{\lambda }\) with \(\varepsilon \varphi _{1}\leq u_{\lambda }\leq u_{\overline{\lambda }}\). □
6 Conclusions
In this paper, we study a class of p-biharmonic equations with Hardy potential and negative exponents. We establish the dependence of the above problem on q, γ, f, g and Ω and evaluate the extremal value of λ related to the multiplicity of positive solutions for this problem.
Availability of data and materials
No data were used to support this study.
Competing interests
The authors declare that they have no competing interests.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.