Abstract
In this paper we study the multiplicity of weak solutions to (possibly resonant) nonlocal equations involving the fractional p-Laplacian operator. More precisely, we consider a Dirichlet problem driven by the fractional p-Laplacian operator and involving a subcritical nonlinear term which does not satisfy the technical Ambrosetti–Rabinowitz condition. By framing this problem in an appropriate variational setting, we prove a multiplicity theorem.
Similar content being viewed by others
1 Introduction
In the standard framework of the p-Laplacian operator, there are a lot of interesting problems widely studied in the literature. A natural question is whether or not the results got in this classical context can be extended to the nonlocal case in the presence of fractional-type operators. Nonlocal fractional equations appear in many fields, and after the seminal works by Caffarelli and Silvestre [8–10], an increasing interest has been devoted to these topics; we refer, for instance, the recent papers [7, 12], the monograph [21] for several results on fractional variational problems and related topics, and [25–27] as well as references therein for some existence and multiplicity results involving the fractional p-Laplacian operator.
In most of the papers concerning with fractional Laplacian equation, it is assumed that the right-hand side has a superlinear, but subcritical growth (cf., e.g., [5, 22, 28, 29] and references therein). In the recent papers [3, 14] it has been firstly studied in the nonlocal setting the so-called asymptotically linear case: as it is well known, the lack of the classical Ambrosetti–Rabinowitz assumption requires more efforts in order to obtain a compactness Palais–Smale-type condition.
Further difficulties arise in the so-called resonant case (cf., e.g., [1] and references therein): indeed, the resonance affects both the compactness property and the geometry of the Euler–Lagrange functional arising in a suitable variational approach. For the local case we recall the contributions [2, 18, 24].
Motivated by this interest in the current literature, we would like to focus here our attention on p-fractional “asymptotically linear” problems, also in the presence of resonance. Indeed, we look for solutions of the nonlocal elliptic problem
where \(1<p<+\infty , s\in ]0,1[, (-\varDelta )^s_p\) denotes the fractional p-Laplacian which (up to normalization factors) may be defined for any \(x\in {\mathbb {R}}^{N}\) as
along any \(\varphi \in C^\infty _0(\mathbb {R} ^N)\), where \(B_\varepsilon (x):=\left\{ y\in \mathbb {R} ^N\,:\,\left| x-y\right| <\varepsilon \right\} \), \(\Omega \) is an open bounded domain of \(\mathbb {R} ^N\) (\(N>sp\)) with Lipschitz boundary \(\partial \Omega \) and \(f:\Omega \times \mathbb {R} \rightarrow \mathbb {R} \) is a Carathéodory function such that:
- \(({h}_1)\) :
-
\( \displaystyle \sup _{|t|\le a}|f(\cdot ,t)|\in L^\infty (\Omega ) \quad \forall \, a>0; \)
- \(({h}_2)\) :
-
there exist
$$\begin{aligned} \lim _{|t|\rightarrow + \infty }\frac{f(x,t)}{|t|^{p-2}t}\ =\ 0 \end{aligned}$$(1.2)and
$$\begin{aligned} \lim _{t\rightarrow 0}\frac{f(x,t)}{|t|^{p-2}t} \ = \alpha \in \mathbb {R}\ \end{aligned}$$(1.3)uniformly with respect to a.e. \(x \in \Omega \).
Here we also deal with the resonant case, under the following further assumption (cf., e.g., [18]):
- \((h_3)\) :
-
there exists
$$\begin{aligned} \lim _{|t|\rightarrow + \infty }\ \Big (f(x,t)t-pF(x,t)\Big )=\ +\infty \end{aligned}$$uniformly with respect to a.e. \(x \in \Omega \), where, as usual, we set
$$\begin{aligned} F(x,t):=\displaystyle \int _0^t f(x,\tau ) \; \mathrm{d}\tau . \end{aligned}$$
In order to state our multiplicity result we introduce some notations. Recalling that the Gagliardo seminorm of a measurable function \(u:\mathbb {R} ^N\rightarrow \mathbb {R} \) is defined by
the fractional Sobolev space is defined by
and is equipped with the norm
where \(|\cdot |_p\) denotes the norm on \(L^p(\mathbb {R} ^N)\).
Our problem is set in the closed linear space
endowed with the norm \(\Vert \cdot \Vert = [\cdot ]_{s,p}\). The space \((X(\Omega ), \Vert \cdot \Vert )\) is uniformly convex and setting
it results:
and
Let us recall that \(\lambda \in \mathbb {R} \) is an eigenvalue for \((-\varDelta )^s_p\) if there exists a nontrivial weak solution on \(X(\Omega )\) of the problem
that is \(u\in X(\Omega ){\setminus }\{0\}\) such that
for every \(\varphi \in X(\Omega )\).
As usual, the set of the eigenvalues is named spectrum and it is denoted by \(\sigma ((-\varDelta )^s_p))\).
It is well known that if \(p=2\) the spectrum of \((-\varDelta )^s\) in \(X(\Omega )\) consists of a diverging sequence \(\{\lambda _n\}_{n\in \mathbb {N}}\) of eigenvalues, repeated according to their multiplicity, satisfying \(0<\lambda _1<\lambda _2\le \cdots \le \lambda _k\le \cdots \) (see [29]). Moreover, for \(p=2\) the structure of \(\sigma ((-\varDelta )^s)\) is very closed to that of the local operator \(-\varDelta \) and also provides a decomposition by means of the eigenfunctions.
On the other hand, when \(p\not =2\) the full spectrum of \((-\varDelta )^s_p\) is still almost unknown, even if some important properties of the first eigenvalue and of the higher-order (variational) eigenvalues have been established in [15, 19]. We also point out that a sequence of eigenvalues has been introduced in [17] by means of the cohomological index.
As we shall see in the proof of our main result, the definition of the quasi-eigenvalues proposed here and inspired to that in [11] fits in with our purposes, as a suitable decomposition of \(X(\Omega )\) can be introduced and it turns out to be the known one for \(p=2\) (cf. Sect. 2 for the details).
Thus, the interaction of the nonlinearity f with the spectrum \(\sigma ((-\varDelta )^s_p)\) of the fractional p-Laplacian operator must be taken into account. To our knowledge the only contribution in this direction is by [17], where the authors obtain existence results via Morse theory.
Now we state our main result.
Theorem 1.1
Let \(s\in ]0,1[, N>sp\) and \(\Omega \) be an open bounded subset of \(\mathbb {R} ^N\) with continuous boundary. Assume that \((h_1)\)–\((h_2)\) hold, \(f(x,\cdot )\) is odd for a.e. \(x\in \Omega \) and that
- \((h_4)\) :
-
there exist \(h, k \in \mathbb {N} \) with \(k\ge h\) such that
$$\begin{aligned} \alpha + \lambda<\ \beta _h\le \ \gamma _k\ <\lambda , \end{aligned}$$where \(\{\beta _n\}_{n\in \mathbb {N}}, \{\gamma _n\}_{n\in \mathbb {N}}\) are, respectively, as in (2.11) and (2.13) below.
Then, problem (1.1) has at least \(k-h+1\) distinct pairs of nontrivial weak solutions, provided either
- (a):
-
\(\lambda \not \in \sigma ((-\varDelta )^s_p)\) or
- (b):
-
\(\lambda \in \sigma ((-\varDelta )^s_p)\) and \((h_3)\) holds true.
For the sake of completeness we recall here that a function \(u\in X(\Omega )\) is a weak solution of problem (1.1) if
for every \(\varphi \in X(\Omega )\).
This paper is organized as follows. In Sect. 2 we introduce the sequences of quasi-eigenvalues and depict some properties; then, in Sect. 3 we prove Theorem 1.1 by making use of a pseudo-index result recalled in Appendix as well as other classical tools.
2 Splitting of the fractional space \(X(\Omega )\)
In this section we adapt the arguments in [11, Section 5] to the nonlocal setting, thus constructing a first sequence of quasi-eigenvalues for \((-\varDelta )_p^s\) on \(X(\Omega )\). Let us define the subset
the functional \(\Phi :X(\Omega ) \rightarrow \mathbb {R} \) by
and
(\(\beta _1\) is indeed the first eigenvalue of the p-fractional operator). Let us remark that
-
\(X(\Omega )\) is a reflexive Banach space;
-
\({\mathcal S}\) is weakly closed in \(X(\Omega )\) (by (1.4));
-
\(\Phi \) is coercive on \(X(\Omega )\);
-
\(\Phi \) is weakly lower semicontinuous on \(X(\Omega )\).
Then, by the generalized Weierstrass theorem, there exists a function \(\psi _1 \in X(\Omega )\) such that
therefore,
Let us consider the linear operator \({\mathcal L}_1: L^p(\mathbb {R} ^N) \rightarrow \mathbb {R} \) related to \(\psi _1\) (\(\psi _1\in W^{s,p}(\mathbb {R} ^N)\Rightarrow \psi _1\in L^p(\mathbb {R} ^N)\) thus \(|\psi _1|^{p-1}\in L^{p'}(\mathbb {R} ^N)\), where \(p'\) is the conjugate of p) defined by
By definition, \({\mathcal L}_1 \in L^{p'}(\mathbb {R} ^N)\), while (2.3) implies \({\mathcal L}_1\psi _1 = 1\), then \(\Vert {\mathcal L}_1\Vert _{L^{p'}} = 1\). Denoting again by \({\mathcal L}_1\) the restriction to the subspace \(X(\Omega )\), it is also \({\mathcal L}_1 \in (X(\Omega ))'\).
Now, we define the new constraint
and the corresponding constrained infimum
We have that \(\beta _2 > \beta _1\) (the first eigenvalue is isolated). We claim that also \({\mathcal S}_1\) is weakly closed in \(X(\Omega )\). In fact, taking a sequence \(\{v_m\}_{m\in \mathbb {N}} \subset {\mathcal S}_1\) and \(v\in X(\Omega )\) such that
by (1.4) it follows that
and, since \({\mathcal L}_1 \in L^{p'}\) and \(v_m \in {\mathcal S}_1\) for each \(m \in \mathbb {N} \), we get that \({\displaystyle \int _{\mathbb {R} ^N} |v(x)|^p \mathrm {d}x = 1}\) and \({\mathcal L}_1 v_m \rightarrow {\mathcal L}_1 v\); therefore, \(v \in {\mathcal S}_1\).
Thus, the generalized Weierstrass theorem applies again and there exists \(\psi _2 \in {\mathcal S}_1\) such that \(\Phi (\psi _2)=\beta _2\), i.e.,
The procedure can be repeated, so fixing any \(n\in \mathbb {N} \) we can define some positive numbers
and some functions
such that, for each \(i \in \{1,\ldots ,n\}\), related to \(\psi _i\) we can consider the linear operator \({\mathcal L}_{i}\in L^{p'}\) defined by
such that
and
hence
while \({\mathcal L}_j\psi _i=0\) for all \(j\in \{1,\ldots , i-1\}\), thus
Therefore, we can define \({\mathcal S}_0:={\mathcal S}\),
and the corresponding constrained infimum
We claim that there exists \(\psi _{n+1} \in {\mathcal S}_n\) such that
To this aim, arguing as above, it is enough proving that each \({\mathcal S}_n\) is weakly closed. Indeed, if \(\{v_m\}_{m\in \mathbb {N}}\subset {\mathcal S}_n\) weakly converges to \(v\in X(\Omega )\), by (1.4) and \({\mathcal L}_i \in L^{p'}\), it follows that \(|v|_p=1\) and \({\mathcal L}_i v_m \rightarrow {\mathcal L}_i v\) for all \(i \in \{1,\ldots ,n\},\) hence \(v \in {\mathcal S}_n\). Thus, \(\Phi \) attains its infimum on \({\mathcal S}_n\) and (2.8) holds.
Summing up, by induction, we construct a sequence of positive numbers \(\{\beta _n\}_{n\in \mathbb {N}}\), of functions \(\{\psi _n\}_{n\in \mathbb {N}} \subset X(\Omega )\) and of linear operators \(\{{\mathcal L}_n\}_{n\in \mathbb {N}} \subset L^{p'}\) such that (2.4)–(2.6) hold for all \(n\in \mathbb {N} \); furthermore, it is
and \(\psi _n \ne \psi _m\) if \(n \ne m\).
Now we recall that if \(V\subseteq X\) is a closed subspace of a Banach space X, a subspace \(W\subseteq X\) is a topological complement of V, briefly \(X=V\oplus W\), if W is closed and every \(x\in X\) can be uniquely written as \(v+w\), with \(v\in V\) and \(w\in W\); furthermore, the projection operators onto V and W are linear and continuous; hence, there exists \(L:=L(V,W)>0\) such that
When \(X=V\oplus W\) and V has finite dimension, we say that W has finite codimension, with \(\mathrm{codim}\, W=\dim V\).
By using again (1.4), the proofs of [11, Lemmas 5.2 and 5.3 and Proposition 5.4] can be adapted with minor changes to our setting; thus, the following properties can be stated:
-
the increasing sequence \(\{\beta _n\}_{n\in \mathbb {N}}\) diverges positively;
-
fixing any \(n \ge 1\) and setting
$$\begin{aligned} X_n := & {} \mathrm{span}\{\psi _1,\ldots ,\psi _n\} = \left\{ v \in X(\Omega ): \; \exists \ b_1,\ldots ,b_n \in \mathbb {R}\ \hbox {s.t.}\ v = \sum _{i=1}^n b_i\psi _i\right\} , \\ Y_n := & {} \bigcap _{i=1}^{n}\ker ({\mathcal L}_i) = \{w \in X(\Omega ): {\mathcal L}_1 w =\dots = {\mathcal L}_{n}w = 0\}, \end{aligned}$$we have
$$\begin{aligned} X(\Omega ) = X_n \oplus Y_n; \end{aligned}$$(2.10) -
taking \(n \ge 1\), for all \(w\in Y_n\) we get
$$\begin{aligned} \beta _{n+1}\ \int _{\mathbb {R} ^N} |w(x)|^p \mathrm {d}x \le \ [w]_{s,p}^p; \end{aligned}$$(2.11) -
the sequence \(\{\psi _n\}_{n\in \mathbb {N}}\) generates the whole space \(X(\Omega )\).
Now, following [18] we introduce another sequence of positive numbers. For all \(n\in {\mathbb {N}}\), taking \(\psi _1\) as in (2.3), we set
and
with \({\mathcal S}\) as in (2.1). By the previous definitions, it follows that \(\beta _1=\gamma _1\) and \({\mathbb {W} _{n+1}}\subseteq {\mathbb {W} _n}\); hence, \(\{\gamma _n\}_{n\in \mathbb {N}}\) is an increasing sequence of quasi-eigenvalues.
Remark 2.1
For \(p=2\) the sequences \(\{\beta _n\}_{n\in \mathbb {N}}\) and \(\{\gamma _n\}_{n\in \mathbb {N}}\) reduce to the known sequence of eigenvalues \(\{\lambda _n\}_{n\in \mathbb {N}}\) of \((-\varDelta )^s\) (see, for instance, the paper [29]).
By using the genus we can construct a sequence of eigenvalues \(\{\mu _n\}_{n\in \mathbb {N}}\) for the nonlinear operator \((-\varDelta )_s^p\) on \(X(\Omega )\), alike in the case of local p-Laplacian as in [16, 20].
Let us consider the nonlinear eigenvalue problem (1.5) and set
where (cf. Sect. 1) \(\displaystyle \Sigma _n:=\{A\in \Sigma : \gamma (A)\ge n\}\) with \(\Phi \) as in (2.2),
and consider the even functional
The critical values and the critical points of \(\Psi \) restricted to the manifold \({\mathcal {S}}\) defined in (2.1) are eigenvalues and eigenfunctions of \((-\varDelta )^s_p\) on \(X(\Omega )\), respectively. We can state the following proposition.
Proposition 2.2
For every \(n\in \mathbb {N} \) the numbers \(\mu _n\) in (2.14) are eigenvalues for the nonlinear operator \((-\varDelta )^s_p\) on \(X(\Omega )\).
Proof
By [13, Lemma 4] and (1.4) the functional \(\Psi \big |_{{\mathcal {S}}}\) satisfies the Palais–Smale condition. Then, by using a suitable version of the deformation lemma (cf., e.g., [6]), standard mini–max arguments give the result. \(\square \)
Furthermore, we point out that \(\mu _1=\beta _1=\gamma _1\). Slight changes in the proof of [2, Proposition 2.9] and in [18, Remark 1.1(4)] provide the following proposition stating a comparison among the sequence \(\{\mu _n\}_{n\in \mathbb {N}}\) and the sequences of quasi-eigenvalues \(\{\beta _n\}_{n\in \mathbb {N}}\) and \(\{\gamma _n\}_{n\in \mathbb {N}}\) of \((-\varDelta )^s_p\).
Proposition 2.3
For all \(n\in \mathbb {N} \) we have that \(\beta _n\le \mu _n\le \gamma _n\).
Remark 2.4
The properties of \(\{\beta _n\}_{n\in \mathbb {N}}\) and Proposition 2.3 imply that \(\{\mu _n\}_{n\in \mathbb {N}}\) and \(\{\gamma _n\}_{n\in \mathbb {N}}\) are diverging sequences. Moreover, as \(\{\gamma _n\}_{n\in \mathbb {N}}\) is increasing, we have also \(\beta _h\le \gamma _k\) for \(k\ge h\ge 1\); therefore, this inequality is not an assumption in \((h_4)\).
3 Proof of Theorem 1.1
From \((h_1)\) and (1.2) for all \(\varepsilon >0\) there exists \(K_\varepsilon >0\) such that
The weak solutions of problem (1.1) are the critical points of the \(C^1\)-functional
on \(X(\Omega )\) whose derivative is given by
for any \(\varphi \in X(\Omega )\).
For the sake of simplicity we introduce the operator \(A:X(\Omega )\rightarrow (X(\Omega ))^*\), defined for all \(u,\varphi \in X(\Omega )\) by
In next proposition we prove that the functional J satisfies the \((\mathrm C)\) condition (cf. Appendix) both in the nonresonant case and in the resonant one, up to assume also assumption \((h_3)\).
Proposition 3.1
Assume that \((h_1)\)–\((h_2)\) hold. Then
-
(i)
if \(\lambda \not \in \sigma ((-\varDelta )^s_p)\), the functional \(J_\lambda \) in (3.2) satisfies \((\mathrm C)\) in \(\mathbb {R};\)
-
(ii)
if \(\lambda \in \sigma ((-\varDelta )^s_p)\) and \((h_3)\) holds, the functional \(J_\lambda \) in (3.2) satisfies \((\mathrm C)\) in \(\mathbb {R} \).
Proof
(i) Let \(c\in \mathbb {R} \) and \(\{u_m\}_{m\in \mathbb {N}}\) be a sequence in \(X(\Omega )\) such that (3.30) holds; then in particular
for every \(\varphi \in X(\Omega )\), where o(1) denotes an infinitesimal sequence.
In order to prove the statement, it is enough to show that \(\{\Vert u_m\Vert \}_{m\in \mathbb {N}}\) is bounded (cf. [23, Proposition 1.3]). Then, arguing by contradiction, let us assume that
Setting \(w_m:=\displaystyle \frac{u_m}{\Vert u_m\Vert }\), \(\{w_m\}_{m\in \mathbb {N}}\) is bounded in \(X(\Omega )\) and there exists \(w\in X(\Omega )\) such that, up to subsequences, we have
and
Evaluating (3.3) in \(w_m-w\) and dividing by \(\Vert u_m\Vert ^{p-1}\), we get
Let us analyze this last equation. Firstly, by (3.6) it follows that
Furthermore, (3.1), (3.4) and (3.6) imply that
Hence, by (3.7)
and by [23, Proposition 1.3]
Thus, by the definition of \(w_m\) it follows \(w\not =0\).
Now, dividing (3.3) by \(\Vert u_m\Vert ^{p-1}\), for all \(\varphi \in X(\Omega )\) we have that
Again (3.1), (3.4) and (3.6) give
Therefore, by (3.9) and (3.11), passing to the limit in (3.10), we get
But this means that \(\lambda \in \sigma ((-\varDelta )^s_p)\), against our assumption; thus, the proof is complete.
(ii) Let \(c\in \mathbb {R} \) and \(\{u_m\}_{m\in \mathbb {N}}\) be a sequence in \(X(\Omega )\) such that (3.30) holds. Set
By using (3.12) we have that
and
with \(\displaystyle G(x,t):=\int _0^t g(x,\tau )\mathrm{d}\tau .\)
By assumption \(({h}_3)\) there exists \(\eta _1>0\) such that
On the other hand, by using condition \(({h}_1)\) there exists \(C_1=C_1(\eta _1)>0\) such that
for every \(m\in \mathbb {N} \). Fixing \(\varepsilon >0\), by (3.12) in addition to (1.2) of \(({h}_2)\), there exists \(\eta _\varepsilon >0\) such that
Now, taking \(q\in ]p,p_s^*[\), there exists \(C>0\) such that
(cf. (1.4)). Hence, let us set
with c as in (3.13) and \(C_1\) as in (3.16).
Again by \(({h}_3)\) we get the existence of \(\eta _2 :=\eta _2(\kappa )>\max \{\eta _1, \eta _\varepsilon \}\) such that
Then, for \(\eta _2\) as above, we define
and
for every \(m\in \mathbb {N} \).
By (3.13)–(3.16) and (3.20) it follows that
Hence, from the above inequality, one has
Taking \(r>p\), by (3.13) and (3.14) we have that
Moreover, by \(({h}_1)\) there exists \(C_2:=C_2(\Omega ,g,\eta _2, r)>0\) such that
Hence, by (3.22) and (3.23) we infer that
Further, by (3.15) and (3.17) it follows that
Now, by the Hölder inequality, (3.18), (3.19) and (3.21) we have that
Thus, the sequence \(\{\Vert u_m\Vert \}_{m\in \mathbb {N}}\) is bounded in \(X(\Omega )\). \(\square \)
Lemma 3.2
Assume that \((h_1)\)–\((h_2)\) hold. Let \(\beta _h\) be as in \((h_4)\) and \(Y_{h-1}\) as in (2.10). Then, there exist \(\rho >0\) and \(c_0>0\) such that, setting \(S_\rho :=\{u\in X(\Omega ): \Vert u\Vert = \rho \},\) the functional \(J_\lambda \) in (3.2) verifies
Proof
By \((h_2)\) it follows that, uniformly with respect to almost every \(x \in \Omega \), there exist
and
Therefore, for every \(\varepsilon >0\) there exist \(R_\varepsilon , \delta _\varepsilon >0\) such that, for almost every \(x\in \Omega \),
and
without loss of generality with \(R_\varepsilon \ge 1\). On the other hand, by \((h_1)\), taking any \(l\in [0, p_s^*-p[\), there exists \(k_{R_\varepsilon }>0\) such that, for almost every \(x\in \Omega \),
The inequalities (3.25)–(3.27) imply that for any \(\varepsilon >0\) there exists \(k_\varepsilon >0\) such that
We infer that
For a suitable \(k'_\varepsilon >0\) we have
Let us recall that by the decomposition (2.10) it is \(X(\Omega )=X_{h-1}\oplus Y_{h-1}\), where \(X_{h-1}:=\mathrm{span}\{\psi _1,\ldots ,\psi _{h-1}\}\) and \(Y_{h-1}\) is its complement. Thus by (2.11) and (3.28) it follows that
and by \((h_4)\), for a suitable \(\varepsilon \), there exists \(k''_\varepsilon >0\) such that
Thus we conclude that if \(\rho \) is small enough there exists \(c_0>0\) such that (3.24) holds. \(\square \)
Lemma 3.3
Assume that \((h_1)\) and (1.2) hold. Let \(\gamma _k\) as in \((h_4)\), \(\mathbb {W} _{k}\) as in (2.12) and \(c_0\) as in Lemma 3.2. Then, there exist a k-dimensional space \(V\in \mathbb {W} _k\) and \(c_\infty >c_0\) such that the functional \(J_\lambda \) in (3.2) verifies
Proof
By (3.1), fixing any \(\varepsilon >0\) there exists \(C_\varepsilon >0\) such that
Let \(\gamma _k\) be as in \((h_4)\) and take \(\varepsilon >0\) such that \(\gamma _k+\varepsilon <\lambda \). From definition (2.13), for such a fixed \(\varepsilon >0\) there exists a subspace \(V^\varepsilon _k\) in \(\mathbb {W} _k\), with \(\dim V^\varepsilon _k\ge k\), such that
Thus it results that
and, as without loss of generality we can assume that \(V^\varepsilon _k\) is a k-dimensional subspace, the functional \(J_\lambda \) tends to \(-\infty \) as \(\Vert u\Vert \) diverges in \(V^\varepsilon _k\), so there exists \(c_\infty =c_\infty (\varepsilon )\) (with \(c_\infty >c_0\)), such that (3.29) holds. \(\square \)
Proof of Theorem 1.1
(a) Firstly, by Proposition 3.1—part (i) the functional \(J_\lambda \) in (3.2) satisfies \((\mathrm C)\) in \(\mathbb {R} \), and by assumption, it is even.
Let us consider \(\beta _h\), \(Y_{h-1}\), \(\rho , c_0\) as in Lemma 3.2 and \(\gamma _k, \mathbb {W} _{k}\), \(V_k^\varepsilon \), \(c_\infty \) as in Lemma 3.3.
Then, we consider the pseudo-index theory \((S_\rho \cap Y_{h-1}, {\mathcal {H}}^*, \gamma ^*)\) related to the genus and \(S_\rho \cap Y_{h-1}\). By Remark 3.7 applied to \(V:=V^\varepsilon _k\), \(\partial B:= S_\rho \) and \(W:=Y_{h-1}\), we get
which implies
The proof is then complete: in fact Theorem 3.6 applies with \({\tilde{A}}:= V^\varepsilon _k\) and \(S:=S_\rho \cap Y_{h-1}\) and J has at least \(k-h+1\) distinct pairs of critical points corresponding to at most \(k-h+1\) distinct critical values \(c_i\), where \(c_i\) is as in (3.32).
(b) In the resonant case, by Proposition 3.1—part (ii) the functional \(J_\lambda \) satisfies \((\mathrm C)\), and we can proceed as above. \(\square \)
Remark 3.4
We point out that Theorem 1.1 holds also with slight changes in the proof when \((h_4)\) is replaced by
References
Bartolo, P., Benci, V., Fortunato, D.: Abstract critical point theorems and applications to some nonlinear problems with “strong” resonance at infinity. Nonlinear Anal. 7, 981–1012 (1983)
Bartolo, R., Candela, A.M., Salvatore, A.: \(p\)-Laplacian problems with nonlinearities interacting with the spectrum. Nonlinear Differ. Equ. Appl. 20, 1701–1721 (2013)
Bartolo, R., Molica Bisci, G.: A pseudo-index approach to fractional equations. Expo. Math. 33, 502–516 (2015)
Benci, V.: On the critical point theory for indefinite functionals in the presence of symmetries. Trans. Am. Math. Soc. 274, 533–572 (1982)
Binlin, Z., Molica Bisci, G., Servadei, R.: Superlinear nonlocal fractional problems with infinitely many solutions. Nonlinearity 28, 2247–2264 (2015)
Bonnet, A.: A deformation lemma on a \(C^1\) manifold. Manuscr. Math. 81, 339–359 (1993)
Cabré, X., Tan, J.: Positive solutions of nonlinear problems involving the square root of the Laplacian. Adv. Math. 224, 2052–2093 (2010)
Caffarelli, L.A., Silvestre, L.: An extension problem related to the fractional Laplacian. Commun. Partial Differ. Equ. 32, 1245–1260 (2007)
Caffarelli, L.A., Silvestre, L.: Regularity theory for fully nonlinear integro-differential equations. Commun. Pure Appl. Math. 62, 597–638 (2009)
Caffarelli, L.A., Silvestre, L.: Regularity results for nonlocal equations by approximation. Arch. Ration. Mech. Anal. 200, 59–88 (2011)
Candela, A.M., Palmieri, G.: Infinitely many solutions of some nonlinear variational equations. Calc. Var. Partial Differ. Equ. 34, 495–530 (2009)
Di Nezza, E., Palatucci, G., Valdinoci, E.: Hitchhiker’s guide to the fractional Sobolev spaces. Bull. Sci. Math. 136, 521–573 (2012)
Drábek, P., Robinson, S.: Resonance problems for the \(p\)-Laplacian. J. Funct. Anal. 169, 189–200 (1999)
Fiscella, A., Servadei, R., Valdinoci, E.: A resonance problem for non-local elliptic operators. Z. Anal. Anwend. 32, 411–431 (2013)
Franzina, G., Palatucci, G.: Fractional \(p\)-eigenvalues. Riv. Mat. Univ. Parma 5, 315–328 (2014)
García Azorero, J., Peral Alonso, I.: Existence and nonuniqueness for the \(p\)-Laplacian: nonlinear eigenvalues. Commun. Partial Differ. Equ. 12, 1389–1430 (1987)
Iannizzotto, A., Liu, S., Perera, K., Squassina, M.: Existence results for fractional \(p\)-Laplacian problems via Morse theory. Adv. Calc. Var. (2014). doi:10.1515/acv-2014-0024
Li, G., Zhou, H.S.: Multiple solutions to \(p\)-Laplacian problems with asymptotic nonlinearity as \(u^{p-1}\) at infinity. J. Lond. Math. Soc. 65, 123–138 (2002)
Lindgren, E., Lindqvist, P.: Fractional eigenvalues. Calc. Var. 49, 795–826 (2014)
Lindqvist, P.: On a nonlinear eigenvalue problem. Berichte Univ. Jyväskylä Math. Inst. 68, 33–54 (1995)
Molica Bisci, G., Rădulescu, V., Servadei, R.: Variational Methods for Nonlocal Fractional Problems. With a Foreword by Jean Mawhin, Encyclopedia of Mathematics and its Applications, vol. 162. Cambridge University Press, Cambridge (2016)
Molica Bisci, G., Repovš, D.: Existence and localization of solutions for nonlocal fractional equations. Asymptot. Anal. 90, 367–378 (2014)
Perera, K., Agarwal, R.P., O’Regan, D.: Morse Theoretic Aspects of \(p\)–Laplacian Type Operators Math. Surveys Monogr. vol. 161. Am. Math. Soc., Providence, RI (2010)
Perera, K., Szulkin, A.: \(p\)-Laplacian problems where the nonlinearity crosses an eigenvalue. Discrete Contin. Dyn. Syst. 13, 743–753 (2005)
Piersanti, P., Pucci, P.: Existence theorems for fractional \(p\)–Laplacian problems. Anal. Appl. (2015) (in press)
Pucci, P., Saldi, S.: Multiple solutions for an eigenvalue problem involving non-local elliptic \(p\)-Laplacian operators, In: Citti, G., Manfredini, M., Morbidelli, D., Polidoro, S., Uguzzoni, F. (eds.) Geometric Methods in PDE’s, vol. 13. Springer INdAM Series, pp. 159–176 (2015)
Pucci, P., Xiang, M., Zhang, B.: Multiple solutions for nonhomogeneous Schrödinger-Kirchhoff type equations involving the fractional \(p\)-Laplacian in \({\mathbb{R}}^N\). Calc. Var. Partial Differ. Equ. 54, 2785–2806 (2015)
Servadei, R., Valdinoci, E.: Mountain pass solutions for non-local elliptic operators. J. Math. Anal. Appl. 389, 887–898 (2012)
Servadei, R., Valdinoci, E.: Variational methods for non-local operators of elliptic type. Discrete Contin. Dyn. Syst. 33, 2105–2137 (2013)
Struwe, M.: Variational Methods. Applications to Nonlinear Partial Differential Equations and Hamiltonian Systems, 4th edn, vol. 34, no. 4. Ergeb. Math. Grenzgeb. Springer, Berlin (2008)
Acknowledgments
The authors warmly thank the anonymous referee for her/his useful and nice comments on the paper. The manuscript was realized within the auspices of the INdAM—GNAMPA Projects 2016 titled: Problemi variazionali su varietà Riemanniane e gruppi di Carnot and Fenomeni non-locali: teoria, metodi e applicazioni.
Author information
Authors and Affiliations
Corresponding author
Appendix: Abstract framework
Appendix: Abstract framework
Throughout this paper \((X,\Vert \cdot \Vert _X)\) is a Banach space, \((X',\Vert \cdot \Vert _{X'})\) its dual, I a \(C^1\) functional on X, \(I^b := \{u\in X : I(u)\le b\}\) the sublevel of I corresponding to \(b\in \mathbb {R} \) and
the set of the critical points of I in X at the critical level \(c \in \mathbb {R} \).
In Sect. 3 we have seen that problem (1.1) has a variational structure; thus, next we recall some abstract tools used before.
Firstly, we recall the so-called Cerami’s variant of the Palais–Smale condition; even if it is a condition weaker than the classical one, it is enough in order to state a deformation theorem and some critical point theorems (cf. [1]).
Definition 3.5
The functional I satisfies the Cerami’s variant of the Palais–Smale condition at level c \(\mathopen {(}c\in \mathbb {R} \mathclose {)}\), if any sequence \(\{u_m\}_{m\in \mathbb {N}} \subseteq X\) such that
converges in X, up to subsequences. In general, if \(-\infty \le a<b\le +\infty \), I satisfies \((\mathrm C)\) in ]a, b[ if so is at each level \(c\in ]a,b[\).
In the proof of our main theorem we use [1, Theorem 2.9] rewritten on Banach spaces where the index theory related to the genus acts. The proof is based on the use of a pseudo-index theory, and before introducing such a definition, we recall some notions of the index theory on Banach spaces X for an even functional with symmetry group \(\mathbb {Z} _2 := \{\mathrm{id}, -\mathrm{id}\}\).
Define
and
Taking \(A \in \Sigma \), \(A\ne \emptyset \), the genus of A is
if such an infimum exists, otherwise \(\gamma (A) = +\infty \). Assume \(\gamma (\emptyset ) = 0\).
The index theory \((\Sigma ,{\mathcal {H}}, \gamma )\) related to \(\mathbb {Z} _2\) is also called genus (we refer for more details, e.g., to [30, Section II.5]).
According to [4], the pseudo-index related to the genus, an even functional \(I:X\rightarrow \mathbb {R} \) and \(S\in \Sigma \) is the triplet \((S, {\mathcal {H}}^*,\gamma ^*)\) such that \({\mathcal {H}}^*\) is a group of odd homeomorphisms and \(\gamma ^*: \Sigma \longrightarrow \mathbb {N} \cup \{+\infty \}\) is the map defined by
Since
then
The following mini–max theorem was proved in [1, Theorem 2.9] in the setting of Hilbert spaces; the same proof holds on Banach spaces.
Theorem 3.6
Consider \(a,b, c_0, c_\infty \in \bar{\mathbb {R}}\), \(-\infty \le a<c_0<c_\infty <b\le +\infty \). Let I be an even functional, \((\Sigma ,{\mathcal {H}},\gamma )\) the genus theory on X, \(S\in \Sigma \), \((S, {\mathcal {H}}^*,\gamma ^*)\) the pseudo-index theory related to the genus, I and S, with
Assume that:
-
(i)
the functional I satisfies \((\mathrm C)\) in ]a, b[;
-
(ii)
\(S\subseteq I^{-1}([c_0,+\infty [);\)
-
(iii)
there exist \({\tilde{k}}\in \mathbb {N} \) and \({\tilde{A}}\in \Sigma \) such that \({\tilde{A}}\subseteq I^{c_\infty }\) and \(\gamma ^*({\tilde{A}})\ge {\tilde{k}}\).
Then the numbers
with \(\Sigma _i^*:= \{A\in \Sigma : \gamma ^*(A)\ge i\}\), are critical values for I and
Furthermore, if \(c=c_i=\cdots =c_{i+r}\), with \(i\ge 1\) and \(i+r\le {\tilde{k}}\), then \(\gamma (K_c)\ge r+1\).
Remark 3.7
In order to apply the theorem above, we need a lower bound for the pseudo-index of a suitable \({\tilde{A}}\) as in (iii) of Theorem 3.6. Thus, let us consider the genus theory \((\Sigma ,{\mathcal {H}},\gamma )\) on X and V, W two closed subspaces of X. If
then, for every odd bounded homeomorphism h on X and every open bounded symmetric neighborhood B of 0 in X, it results
Rights and permissions
About this article
Cite this article
Bartolo, R., Molica Bisci, G. Asymptotically linear fractional p-Laplacian equations. Annali di Matematica 196, 427–442 (2017). https://doi.org/10.1007/s10231-016-0579-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10231-016-0579-2
Keywords
- Fractional p-Laplacian
- Integro-differential operator
- Variational methods
- Asymptotically linear problem
- Resonant problem
- Pseudo-genus