Abstract
We study the nonlinear Schrödinger equation in \(\mathbb {R}^n\) without making any periodicity assumptions on the potential or on the nonlinear term. This prevents us from using concentration compactness methods. Our assumptions are such that the potential does not change the essential spectrum of the linear operator. This results in \([0, \infty )\) being the absolutely continuous part of the spectrum. If there are an infinite number of negative eigenvalues, they will converge to 0. In each case we obtain nontrivial solutions. We also obtain least energy solutions.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
We consider the semilinear Schrödinger equation
where V(x) is a given potential. One wishes to find solutions and, in particular, the so called “least energy solutions.” These are solutions that minimize the corresponding energy functional. The existence of solutions depends both on the linear operator \(\mathcal {A}\) and the nonlinear term f(x, u).
Many authors have studied the problem for the Schrödinger equation (1) under various stipulations (cf., e.g., [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20, 23, 25, 26, 28,29,30,31,32] and references quoted in them). In almost all cases it was required to stipulate that the spectrum of the linear operator \(\mathcal {A}u = - \Delta u + V(x) u\) have a gap. This caused writers to make various assumptions on the potential V(x) to guarantee that this is the case. However, many of these assumptions caused the nature of the spectrum to be far different from that of \( - \Delta u .\) Thus, any theorem proved for
did not hold for \(\mathcal {A}u=- \Delta u.\)
Some authors assumed
Others assumed that there exists a constant B such that \( V(x) \le B\) for all \(x \in \mathbb {R}^n,\) \(V(x) \rightarrow B \; as\; |x| \rightarrow \infty \) and \(\sigma ( - \Delta + V(x) ) >0\) together with other assumptions. Another approach assumes that for every \(M >0\) the set \(\omega = \{x \in \mathbb {R}^n : V(x)<M \} \) has finite Lebesgue measure. Others assumed that V(x) is in some combination of \( L^p(\mathbb {R}^n)\) spaces. In each case the growth of f(x, t) is controlled by the growth of V(x). In most cases the resulting spectrum of \(\mathcal {A}= - \Delta + V(x) \) is discrete, consisting only of isolated eigenvalues of finite multiplicity tending to \( + \infty .\) All of these assumptions cause restrictions on the the nonlinear term depending on V. In most cases the hypothesis
is used, where \(\mu >2\) and
A different approach is to assume that the potential is periodic in the coordinates of \(\mathbb {R}^n\) and then apply concentration compactness methods. In this case the resulting spectrum of \(\mathcal {A}= - \Delta + V(x) \) is absolutely continuous and consists of a finite number of disjoint closed intervals. In order to apply this method, f(x, t) must be periodic in x as well. In the few publications where 0 is permitted to be in \(\sigma ( - \Delta + V(x) ),\) an interval of the form \((-\varepsilon , 0)\) is required to be free of the spectrum.
The purpose of the present paper is solve the Eq. (1) under assumptions on V(x) such that the essential spectrum of \(\mathcal {A}= - \Delta + V(x) \) is the the same as that of \(- \Delta ,\) i.e., \([0,\infty ).\) The situation is different if there are no negative eigenvalues, one negative eigenvalue or two or more negative eigenvalues. If there are no negative eigenvalues, one can solve under the same hypotheses for f(x, t) that can be used for the equation
Otherwise, the hypotheses on f(x, t) need only take the negative eigenvalues into consideration. We can even deal with the case where the negative eigenvalues converge to 0. We do not need an interval of the form \((-\varepsilon , 0)\) to be free of the spectrum. In each of these cases different methods must be employed, requiring different assumptions on the nonlinear term.
Concerning the function V(x) we make the following assumptions:
- \( (V_1)\) :
-
$$\begin{aligned} \sup _y \int _{|x-y| <\delta }|V(x)| \;\omega _{2}(x-y)\,dx \rightarrow 0 \; as\; \delta \rightarrow 0, \end{aligned}$$
and
- \((V_2)\) :
-
$$\begin{aligned} \int _{|x-y| <1}|V(x)| \;\omega _{2}(x-y)\,dx \rightarrow 0 \; as\; |y| \rightarrow \infty , \end{aligned}$$
where
$$\begin{aligned} \omega _s(x) = {\left\{ \begin{array}{ll}|x|^{s-n}, \quad 0<s<n\\ 1-\ln |x|^2, \quad s=n \\ 1, \quad s>n.\end{array}\right. }\end{aligned}$$
These assumptions imply that there is a forms extension \(\mathcal {A}\) of the operator
on the space \(H=H^{1,2} (\mathbb {R}^n)\) having essential spectrum equal to \([0,\infty )\) and a (possibly empty) discrete, countable negative spectrum consisting of isolated eigenvalues of finite multiplicity with a finite lower bound \(-L\)
For each \(l>0,\) define the subspaces \(M=M_l\) and \(N=N_l\) of H as
For the operator \(\mathcal {A}\) there are three possibilities: (a) it has no negative eigenvalues, (b) it has only one negative eigenvalue, and (c) it has two or more negative eigenvalues. What is interesting is that each of these possibilities must be dealt with differently. We shall study all of them separately.
The following notation will be used throughout the paper:
Let q be any number satisfying
and let f(x, t) be a Carathéodory function on \(\mathbb {R}^n\times {\mathbb {R}}\). This means that f(x, t) is continuous in t for a.e. \(x\in \mathbb {R}^n\) and measurable in x for every \(t\in {\mathbb {R}}\). We make the following assumptions
(A) The function f(x, t) satisfies
and
where \(S(x) > 0\) is a function in \(L^q(\mathbb {R}^n)\) satisfying
and W is a function in \(L^{\infty }(\mathbb {R}^n)\). Here
Our other hypotheses depend only on the primitive
of f(x, t).
Let
It follows that G is a continuously differentiable functional on the whole of H (cf., e.g., [25]). It is easily checked that \(u \in H\) is a (weak) solution of (1) iff it is a critical point of G(u). Our methods will make use of this fact. They involve finding linking sets A, B which separate G, i.e., are such that
If the spectrum of \(\mathcal {A}\) has no gap, i.e., consists only of the interval \([0,\infty ), \) the choice of the sets A, B is very limited. If it has one gap, the subspace in the gap is very useful. If it has two gaps, we obtain two useful subspaces. In our situation, if there are no negative eigenvalues, there are no gaps. If there is one negative eigenvalue, there is one gap. If there are two or more negative eigenvalues, there are at least two gaps. Therefore we consider three different situations.
2 The space \(N_1\)
We let \(N_1\) be the set of measurable functions h(x) on \(\mathbb {R}^n\) satisfying
It is clear that bounded functions are in \(N_1.\) It can be shown that \(L^{p}(\mathbb {R}^n) \subset N_1\) for \(p \ge n.\) For a general description of this space, cf. [21]. These functions will be used in our theorems because of the following properties:
Lemma 1
If \(g^{-1} \in N_1,\) the following statements are true:
-
1.
there is a constant C such that
$$\begin{aligned} \Vert u\Vert _H \le C \Vert u\Vert '_H, \quad u \in H, \end{aligned}$$(5)where
$$\begin{aligned} (\Vert u\Vert '_H)^2 = \Vert \nabla u\Vert ^2+\Vert gu\Vert ^2. \end{aligned}$$ -
2.
If \(\mathcal {A}\) has no negative eigenvalues, then there is an \(\varepsilon >0\) such that
$$\begin{aligned} a(u) + \Vert gu\Vert ^2 \ge \varepsilon \Vert u\Vert _H^2, \quad u \in H. \end{aligned}$$(6) -
3.
If \(\mathcal {A}\) has only one negative eigenvalue \(\lambda _0,\) then there is an \(\varepsilon >0\) such that
$$\begin{aligned} a(u) + \Vert gu\Vert ^2 \ge \varepsilon \Vert u\Vert _H^2, \quad u \in M. \end{aligned}$$(7) -
4.
If \(\mathcal {A}\) has negative eigenvalues \(\lambda _{l-1}, \lambda _l\), then there is an \(\varepsilon >0\) such that
$$\begin{aligned} a(u) -\lambda _l \Vert u\Vert ^2+ \Vert gu\Vert ^2 \ge \varepsilon \Vert u\Vert _H^2, \quad u \in M. \end{aligned}$$(8)
If \(g \in N_1,\) then there is a constant C such that
If \(g, g^{-1} \in N_1,\) then the two norms are equivalent.
3 No negative eigenvalues
In this case,
There are no gaps in the spectrum of \(\mathcal {A}.\)
We have
Theorem 2
Assume
-
1.
There is a function g(x) such that \(g, g^{-1} \in N_1\) and
$$\begin{aligned} 2F(x,u) \le -g(x)^2 |u|^2+W(x), \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \end{aligned}$$where \(W(x) \in L^1(\mathbb {R}^n).\)
Then the Eq. (1) has a solution.
Remark 3
It is clear from the equation that the solution obtained will be nontrivial if
To guarantee that a solution will be nontrivial even when \(f(x,0) = 0,\) we have
Theorem 4
Assume
-
1.
There are constants \(0<\alpha <2,\delta >0, \) and a function g(x) such that \(g, g^{-1} \in N_1\) and
$$\begin{aligned} 2F(x,u)&\le - g(x)^2|u|^2, \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \; |u| < \delta , \\&\le -g(x)^2 |u|^2+S(x)^q |u|^\alpha , \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \; |u| > \delta . \end{aligned}$$ -
2.
There is a locally bounded function h(t) such that
$$\begin{aligned} 2F(x,u) \ge -h(u), \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \end{aligned}$$and
$$\begin{aligned} c_0=\sup _\mathbb {R}h(u)/u^2 <\infty . \end{aligned}$$
Then the Eq. (1) has a nontrivial solution.
4 Only one negative eigenvalue
Let \(\lambda _0 <0\) be the eigenvalue. In this case
We can make use of the fact that there is a gap in the spectrum of \(\mathcal {A}.\) We have
Theorem 5
Assume
-
1.
$$\begin{aligned} 2F(x,u) \ge \lambda _0|u|^2, \quad u \in \mathbb {R},\; x \in \mathbb {R}^n. \end{aligned}$$
-
2.
There are constants \(0<\alpha <2,\delta >0, \) and a function g(x) such that \(g, g^{-1} \in N_1\) and
$$\begin{aligned} 2F(x,u)&\le - g(x)^2|u|^2, \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \; |u| < \delta , \\&\le -g(x)^2 |u|^2+S(x)^q |u|^\alpha , \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \; |u| > \delta . \end{aligned}$$
Then the Eq. (1) has a solution.
To obtain a nontrivial solution we have
Theorem 6
Assume
-
1.
$$\begin{aligned} 2F(x,u) \ge \lambda _0|u|^2, \quad u \in \mathbb {R},\; x \in \mathbb {R}^n. \end{aligned}$$
-
2.
There are constants \(0<\alpha <2,\delta >0, \) and a function g(x) such that \(g, g^{-1} \in N_1\) and
$$\begin{aligned} 2F(x,u)&\le - g(x)^2|u|^2, \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \; |u| < \delta , \\&\le -g(x)^2 |u|^2+S(x)^q |u|^\alpha , \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \; |u| > \delta . \end{aligned}$$ -
3.
There is a locally bounded function h(t) such that
$$\begin{aligned} 2F(x,u) \ge -h(u), \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \end{aligned}$$and
$$\begin{aligned} c_0=\sup _{\mathbb {R}} h(u)/u^2 < \infty . \end{aligned}$$
Then the Eq. (1) has a nontrivial solution.
5 Two or more negative eigenvalues
Here again we can make use of the fact that there is more than one gap in the spectrum. This allows us to use more complicated linking methods.
Let \(\lambda _{l-1}, \; \lambda _l\) be two consecutive negative eigenvalues of \(\mathcal {A}.\) We have
Theorem 7
Assume
-
1.
$$\begin{aligned} 2F(x,u) \ge \lambda _{l-1} |u|^2, \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \end{aligned}$$
-
2.
There are constants \(0<\alpha <2,\delta >0, \) and a function g(x) such that \(g, g^{-1} \in N_1\) and
$$\begin{aligned} 2F(x,u)&\le - g(x)^2|u|^2 + \lambda _l |u|^2, \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \; |u| < \delta , \\&\le -g(x)^2 |u|^2+\lambda _l|u|^2 +S(x)^q |u|^\alpha , \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \; |u| > \delta . \end{aligned}$$
Then the Eq. (1) has a solution.
To obtain a nontrivial solution we have
Theorem 8
Assume
-
1.
$$\begin{aligned} 2F(x,u) \ge \lambda _{l-1} |u|^2, \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \end{aligned}$$
-
2.
There are constants \(0<\alpha <2,\delta >0, \) and a function g(x) such that \(g, g^{-1} \in N_1\) and
$$\begin{aligned} 2F(x,u)&\le - g(x)^2|u|^2 + \lambda _l |u|^2, \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \; |u| < \delta , \\&\le -g(x)^2 |u|^2+\lambda _l|u|^2 +S(x)^q |u|^\alpha , \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \; |u| > \delta . \end{aligned}$$ -
3.
There is a locally bounded function h(t) such that
$$\begin{aligned} 2F(x,u) \ge -h(u), \quad u \in \mathbb {R},\; x \in \mathbb {R}^n, \end{aligned}$$and
$$\begin{aligned} c_0=\sup _{\mathbb {R}} h(u)/u^2 < \infty . \end{aligned}$$
Then the Eq. (1) has a nontrivial solution.
Remark 9
Note that the hypothesis of Theorem 2 is stronger than hypotheses 1 and 2 of Theorem 5. Hypothesis 1 of Theorem 5 requires
which is stronger than hypothesis 1 of Theorem 7. Consequently, the hypotheses of Theorem 4 are stronger than those of Theorem 6 which are stronger than those of Theorem 8.
6 Least energy solutions
Let \(\mathcal {M}\) be the set of all solutions of (1). A solution \(\tilde{u}\) is called a “least energy solution” if it minimizes the functional
over the set \(\mathcal {M}.\)
We have
Theorem 10
If we add the following hypothesis to Theorems 2–8, then Eq. (1) has a least energy solution: The function given by
satisfies
We shall prove Theorems 2–10 in Sect. 8. In the next section we describe the construction of the operator \(\mathcal {A}.\) We obtain the largest self-adjoint extension of \(\mathcal {A}_0\) which preserves the essential spectrum.
7 The operator \(\mathcal {A}\)
The following was proved in [21] (Theorem 10.9, ch. 6, p. 153.)
Theorem 11
Let P(D) be an elliptic constant coefficient operator of order 2 on \(\mathbb {R}^n, \) and let V(x) be a function satisfying \((V_1)\) and \((V_2).\) If \(\rho (P_0)\) is not empty, then P(D)+V has a forms extension operator \(\mathcal {A}\) such that \(\sigma _e(\mathcal {A})=\sigma _e(P_0).\)
Here \(P_0 \) is the closure of the operator P(D) restricted to \(C^\infty _0(\mathbb {R}^n), \) and \(\sigma _e\) is the essential spectrum. Any point not in the essential spectrum is either a point in the resolvent or an isolated eigenvalue of finite multiplicity. If \(P(D) = -\Delta ,\) then \(\sigma (P_0)=[0,\infty ).\) Consequently, in our case, the negative spectrum of \(\mathcal {A}\) can have at most a countable number of isolated eigenvalues of finite multiplicity having a finite lower bound. The theorem is proved by showing that the bilinear form
is bounded and closed on H. Moreover, if the coefficients of P(D) and the function V(x) are real, then the bilinear form is symmetric and \(\mathcal {A}\) is selfadjoint.
8 Proof of Lemma 1.
Proof
We have
Hence,
To prove (6), assume that there is a sequence \( u^{(k)} \in H\) such that \(\Vert u^{(k)}\Vert ' _H =1 \) and \( a(u^{(k)}) + \Vert gu^{(k)}\Vert ^2 \rightarrow 0,\) where
Since
there is a renamed subsequence such that \( u^{(k)}\) converges to a limit \( u \in H\) weakly in H, strongly in \(L^2_{loc}(\mathbb {R}^n)\) and a.e. in \(\mathbb {R}^n.\) Since \( a(u^{(k)}) \ge 0,\) we have \( a(u^{(k)}) \rightarrow 0 \) and \( \Vert gu^{(k)}\Vert \rightarrow 0.\) By our hypothesis on V(x), there is a renamed subsequence of \( u^{(k)}\) such that \(b( u^{(k)}) \rightarrow b(u),\) where \( b(u)=(Vu,u).\) Thus, \(1+b(u)=0,\) showing that \(u \ne 0.\) Since \(g u^{(k)} \rightarrow gu \) a.e., we have \(u =0,\) providing a contradiction. The same proof can be used to prove (7) if we keep in mind that (7) holds only on M. To prove (8), assume that there is a sequence \( u^{(k)} \in M\) such that \(\Vert u^{(k)}\Vert ' _H =1 \) and \( a(u^{(k)}) -\lambda _l\Vert u^{(k)}\Vert ^2+ \Vert gu^{(k)}\Vert ^2 \rightarrow 0.\) Since
there is a renamed subsequence such that \( u^{(k)}\) converges to a limit \( u \in H\) weakly in H, strongly in \(L^2_{loc}(\mathbb {R}^n)\) and a.e. in \(\mathbb {R}^n.\) Also, there is a renamed subsequence such that \(\Vert \nabla u^{(k)} \Vert \rightarrow \nu \) and \(\Vert g u^{(k)}\Vert \rightarrow \tau ,\) where \(\nu ^2+\tau ^2=1.\) Since \( a(u^{(k)}) -\lambda _l \Vert u^{(k)}\Vert ^2\ge 0\) in M, we have \( a(u^{(k)}) -\lambda _l\Vert u^{(k)}\Vert ^2\rightarrow 0 \) and \( \Vert gu^{(k)}\Vert \rightarrow 0.\) By our hypothesis on V(x), there is a renamed subsequence of \( u^{(k)}\) such that \(b( u^{(k)}) \rightarrow b(u),\) where \( b(u)=(Vu,u).\) Since \( a(u^{(k)}) -\lambda _l\Vert u^{(k)}\Vert ^2 \rightarrow 0,\) and \(a(u^{(k)}) \rightarrow \nu ^2 +b(u),\) we see that \(\nu ^2+b(u)\le 0.\) Since \( gu^{(k)} \rightarrow gu \) a.e. and \( \Vert gu^{(k)}\Vert \rightarrow 0,\) we have \(\tau =0\) and \(u =0.\) But then, \(\nu = 1\) and \(1+ b(u) \le 0,\) showing that \(u \ne 0,\) providing a contradiction. \(\square \)
9 Proofs of the Theorems
We now give the proof of Theorem 2.
Proof
We define
By Lemma 1 there is an \(\varepsilon >0\) such that
Thus
i.e., G(u) is coercive. Let
By Corollary 3.22, p. 29, of [24] there is a sequence \(\{u^{(k)}\} \subset H\) such that
and
Since G(u) is coercive,
where
Thus there is a renamed subsequence such that \( u^{(k)}\) converges to a limit \( u \in H\) weakly in H, strongly in \(L^2_{loc}(\mathbb {R}^n)\) and a.e. in \(\mathbb {R}^n.\) From (17) we see that
from which we conclude easily that u is a solution of (1). \(\square \)
Proof of Theorem 4
By Lemma 1, there is an \(\varepsilon >0\) such that
Consequently,
by Hypothesis (A). As a consequence, there are positive constants \(\eta , \rho \) such that
Since \(0 \in \sigma _e(\mathcal {A}),\) there is a \(\varphi \in H\) such that \(\Vert \varphi \Vert =1\) and \(a(\varphi )=(\mathcal {A}\varphi , \varphi ) < \eta /2.\) Consequently,
Thus,
since
This implies that there is a sequence \( u^{(k)}\) in H such that
(Theorem 2.7.1 of [22]).
If
there is a renamed subsequence such that \( u^{(k)}\) converges to a limit \( u \in H\) weakly in H, strongly in \(L^2_{loc}(\mathbb {R}^n)\) and a.e. in \(\mathbb {R}^n.\) In particular we have
From this we see that
from which we conclude easily that u is a solution of (1). Moreover, since \(\Vert u^{(k)}\Vert '_H\) is bounded, (21) implies \(G(u)=c\ge \eta /2\) (Theorem 3.4.1 of [22]). Since, \(G(0)=0,\) we see that \(u \ne 0.\)
If
let \(\tilde{u}^{(k)}=u^{(k)}/\rho _k.\) Then, \(\Vert \tilde{u}^{(k)}\Vert '_H =1.\) There is a renamed subsequence such that \( \tilde{u}^{(k)} \) converges to a function \(\tilde{u}(x) \in H\) weakly in H, strongly in \(L^2_{loc}(\mathbb {R}^n),\) a.e. in \(\mathbb {R}^n,\) and such that \(\Vert \nabla \tilde{u}^{(k)} \Vert \rightarrow \nu \) and \(\Vert g\tilde{u}^{(k)}\Vert \rightarrow \tau ,\) where \(\nu ^2+\tau ^2=1.\) Let \(b(u,v)=(Vu,v),\; b(u)=b(u,u).\) Since \( \tilde{u}^{(k)} \) converges to \(\tilde{u} \) weakly in H, we see that \(b(\tilde{u}^{(k)}) \rightarrow b(\tilde{u}).\) Hence, \( \Vert \nabla \tilde{u}^{(k)} \Vert ^2 +b(\tilde{u}^{(k)}) +\Vert g\tilde{u}^{(k)}\Vert ^2 \rightarrow \nu ^2 +b(\tilde{u}) +\tau ^2.\) Note that
Thus,
Thus, \(1+b(\tilde{u})= \nu ^2+b(\tilde{u}) + \tau ^2\le 0.\) This implies that \(b(\tilde{u}) \ne 0,\) and consequently \( \tilde{u}(x) \not \equiv 0.\) But \(g \tilde{u}^{(k)} \rightarrow g\tilde{u}\) a.e. This means that \(\tilde{u} =0,\) a contradiction. Hence, the \(\rho _k\) are bounded, and the proof is complete. \(\square \)
Proof of Theorem 5
We follow the proof of Theorem 4. By Lemma 1 there is an \(\varepsilon >0\) such that (7) holds. This implies that there are positive constants \(\eta , \rho \) such that
by the argument given there. Note that
by Hypothesis 1. Define \(A=M \cap {\mathbf{B}}_\rho , \;\; B=N,\) where
Then A links B (Example 2, p. 38 of [22]). If \(G_1=-G,\) then
By Theorem 13.7, p. 259 of [22] or Corollary 3.22, p. 29 of [24] there is a sequence \(\{u^{(k)}\} \subset H\) such that
and
If
where
then there is a renamed subsequence such that \( u^{(k)}\) converges to a limit \( u \in H\) weakly in H, strongly in \(L^2_{loc}(\mathbb {R}^n)\) and a.e. in \(\mathbb {R}^n.\) From (26) we see that
from which we conclude easily that u is a solution of (1).
If
let \(\tilde{u}^{(k)}=u^{(k)}/\rho _k.\) Then, \(\Vert \tilde{u}^{(k)}\Vert '_H =1.\) There is a renamed subsequence such that \( \tilde{u}^{(k)} \) converges to a function \(\tilde{u}(x) \in H\) weakly in H, strongly in \(L^2_{loc}(\mathbb {R}^n),\) a.e. in \(\mathbb {R}^n,\) and such that \(\Vert \nabla \tilde{u}^{(k)} \Vert \rightarrow \nu \) and \(\Vert g\tilde{u}^{(k)}\Vert \rightarrow \tau ,\) where \(\nu ^2+\tau ^2=1.\) Let \(b(u,v)=(Vu,v),\; b(u)=b(u,u).\) Since \( \tilde{u}^{(k)} \) converges to \(\tilde{u} \) weakly in H, we see that \(b(\tilde{u}^{(k)}) \rightarrow b(\tilde{u}).\) Hence, \(0 \le \Vert \nabla \tilde{u}^{(k)} \Vert ^2 +b(\tilde{u}^{(k)}) +\Vert g\tilde{u}^{(k)}\Vert ^2 \rightarrow \nu ^2 +b(\tilde{u}) +\tau ^2.\) Since \(a(\tilde{u}^{(k)}) \ge 0,\) we have \( \nu ^2 +b(\tilde{u}) \ge 0.\) Note that
Moreover,
This implies \( \nu ^2+b(\tilde{u}) =0\) and \(\tau ^2=0.\) But \(g \tilde{u}^{(k)} \rightarrow g\tilde{u}\) a.e. This means that \(\tilde{u} =0\) and \(\nu =1,\) so that
But this implies that \(b(\tilde{u}) \ne 0,\) and consequently \( \tilde{u}(x) \not \equiv 0,\) a contradiction. Hence, the \(\rho _k\) are bounded, and the proof is complete. \(\square \)
Proof of Theorem 6
We follow the proof of Theorem 4. By Lemma 1 there is an \(\varepsilon >0\) such that (7) holds. This implies that there are positive constants \(\eta , \rho \) such that
as in the proof of Theorem 4. Note that
by Hypothesis 1. Since \(\mathcal {A}\) is not invertible on M, there is a \(\varphi \in M\) such that \(\Vert \varphi \Vert =1\) and \(a(\varphi )=(\mathcal {A}\varphi , \varphi ) < \eta .\) For \(R > \rho ,\) let
where
By Example 3, p.38 of [22], \( A_R, \; B\) link each other. Now
We can now apply Theorem 2.7.3 of [22] to conclude that there is a sequence \( u^{(k)}\) in H such that
We can now follow the proofs of Theorem 5 to reach the desired conclusion. \(\square \)
Proof of Theorem 7
Define the subspaces M and N of H as before:
Let
We note that Hypothesis 1 implies
In fact, we have
In view of inequality (8), we see that there are positive constants \(\eta , \rho \) such that
Take
where
By Example 8, p. 22 of [24], A links B. Moreover,
Hence, we may apply Corollary 2.8.2 of [22] to conclude that that there is a sequence \(\{u^{(k)}\} \subset H\) such that
and
If
there is a renamed subsequence such that \( u^{(k)}\) converges to a limit \( u \in H\) weakly in H strongly in \(L^2_{loc}(\mathbb {R}^n)\) and a.e. in \(\mathbb {R}^n.\) From (35) we see that
from which we conclude easily that u is a solution of (1). We can now follow the proof of Theorem 4 until we reach (23). In our case this becomes
Since \(\lambda _l \le 0,\)
and
we again have \(1+b(\tilde{u})\le 0,\) which leads to the desired conclusion. \(\square \)
Proof of Theorem 8
We follow the proof of Theorem 7. We note that Hypothesis 1 implies
In fact, we have
In view of inequality (8), we see that there are positive constants \(\eta , \rho \) such that
Let \(\varphi \) be an eigenfunction of \(\mathcal {A}\) corresponding to \(\lambda _l \) and satisfying \(\Vert \varphi \Vert =1.\) For \(R > \rho ,\) let
where
By Example 3, p.38 of [22], \( A_R, \; B\) link each other. Now
We can now apply Theorem 2.7.3 of [22] to conclude that there is a sequence \( u^{(k)}\) in H such that
We can now follow the proof of Theorem 4 until we reach (23). In our case this becomes
Since \(\lambda _l \le 0,\)
and
we again have \(1+b(\tilde{u})\le 0,\) which leads to the desired conclusion. \(\square \)
Proof of Theorem 10. Let
There is a sequence \(\{u^{(k)}\} \in \mathcal {M}\) such that
and
Thus,
In view of assumption (11), we see that \(\alpha \ge -\int W > - \infty .\) In view of the arguments given in the proofs of Theorems 2–8, we see that
Hence, there is a renamed subsequence such that \(u^{(k)}\) converges to a limit \(u \in H,\) weakly in H, strongly in \(L^2_{loc}(\mathbb {R}^n)\) and a.e. in \(\mathbb {R}^n.\) From (43) we see that
from which we conclude easily that u is a solution of (1). Hence, \(u \in \mathcal {M}.\) Moreover,
Thus, \(G(u)=\alpha .\) If \(\alpha \ne 0,\) then \( u \ne 0.\) This completes the proof.
References
Alama, S., Li, Y.Y.: On “multibump” bound states for certain semilinear elliptic equations. Indiana Univ. Math. J. 41(4), 983–1026 (1992)
Angenent, S.: The shadowing lemma for elliptic PDE. Dynamics of infinite-dimensional systems (Lisbon, 1986), 7–22. NATO Adv. Sci. Inst. Ser. F Comput. Systems Sci., 37, Springer, Berlin (1987)
Bartsch, T., Ding, Y.: On a nonlinear Schrödinger equation with periodic potential. Math. Ann. 313, 15–37 (1999)
Bartsch, T., Wang, Z.: Existence and multiplicity results for some superlinear elliptic problems on \(\mathbb{R}^N\). Commun. PDE 20, 1725–1741 (1995)
Bartsch, T., Pankov, A., Wang, Z.-Q.: Nonlinear Schrödinger equations with steep potential well. Commun. Contemp. Math. 3(4), 549–569 (2001)
Berestycki, H., Capuzzo-Dolcetta, I., Nirenberg, L.: Variational methods for indefinite superlinear homogeneous elliptic problems. Nonlinear Differ. Equ. Appl. 2, 553572 (1995)
Berestycki, H., Lions, P.-L.: Nonlinear scalar field equations. I. Existence of a ground state. Arch. Ration. Mech. Anal. 82(4), 313–345 (1983)
Chabrowski, J., Szulkin, A.: On a semilinear Schrödinger equation with critical Sobolev exponent. Proc. Am. Math. Soc. 130(1), 85–93 (2002)
Costa, D.G.: On a class of elliptic systems in \(\mathbb{R}^N,\) electronic. J. Differ. Equ. 7, 1–14 (1994)
Ding, Y., Szulkin, A.: Bound states for semilinear Schrödinger equations with sign-changing potential. Calc. Var. Part. Differ. Equ. 29(3), 397–419 (2007)
Ding, W.-Y., Ni, W.-M.: On the existence of positive entire solutions of a semilinear elliptic equation. Arch. Ration. Mech. Anal. 91, 183308 (1986)
Furtado, M.F., Maia, L.A., Silva, E.A.B.: On a double resonant problem in \(\mathbb{R}^N\). Differ. Integral Equ. 15(11), 1335–1344 (2002)
Jeanjean, L.: Local conditions insuring bifurcation from the continuous spectrum. Math. Z. 232, 651–664 (1999)
Kryszewski, W., Szulkin, A.: Generalized linking theorems with an application to semilinear Schrödinger equation. Adv. Differ. Equ. 3, 441–472 (1998)
Kryszewski, W., Szulkin, A.: Infinite-dimensional homology and multibump solutions. J. Fixed Point Theory Appl. 5(1), 1–35 (2009)
Li, Y.: Remarks on a semilinear elliptic equation on RN. J. Differ. Equ. 74, 3449 (1988)
Li, Y.Q., Wang, Z.Q., Zeng, J.: Ground states of nonlinear Schrödinger equations with potentials. Ann. Inst. H. Poincare Anal. Non Lineaire 23(6), 829–837 (2006)
Omana, W., Willem, M.: Homoclinic orbits for a class of Hamiltonian systems. Differ. Integral Equ. 5, 1115–1120 (1992)
Pankov, A.: Periodic nonlinear Schrödinger equation with application to photonic crystals. Milan J. Math. 73, 259–287 (2005)
Rabinowitz, P.H.: On a class of nonlinear Schrödinger equations. Z. Agnew. Math. Phys. 43, 270–291 (1992)
Schechter, M.: Spectra of Partial Differential Operators. Second edition. North-Holland Series in Applied Mathematics and Mechanics, vol. 14. North-Holland Publishing Co., Amsterdam. ISBN: 0-444-87822-X (1986)
Schechter, M.: Linking Methods in Critical Point Theory. Birkhauser Boston, Cambridge (1999)
Schechter, M.: Superlinear Schrödinger operators. J. Funct. Anal. 262, 2677–2694 (2012)
Schechter, M.: Minimax Systems and Critical Point Theory. Birkhauser Boston, Cambridge (2009)
Schechter, M.: Nonlinear Schrdinger operators with zero in the spectrum. Z. Angew. Math. Phys. 66(5), 21252141 (2015)
Schechter, M., Zou, W.: Weak linking theorems and Schrödinger equations with critical Sobolev exponent. ESAIM Control Optim. Calc. Var. 9, 601–619 (2003). (electronic)
Sirakov, B.: Existence and multiplicity of solutions of semi-linear elliptic equations in RN. Calc. Var. 11, 119142 (2000)
Szulkin, A., Weth, T.: Ground state solutions for some indefinite variational problems. J. Funct. Anal. 257(12), 3802–3822 (2009)
Troestler, C., Willem, M.: Nontrivial solution of a semilinear Schrödinger equation. Commun. Part. Differ. Equ. 21(9–10), 1431–1449 (1996)
Willem, M., Zou, W.: On a Schrödinger equation with periodic potential and spectrum point zero. Indiana Univ. Math. J. 52(1), 109–132 (2003)
Yang, M.: Ground state solutions for a periodic Schrödinger equation with superlinear nonlinearities. Nonlinear Anal. 72(5), 2620–2627 (2010)
Yang, M., Chen, W., Ding, Y.: Solutions for periodic Schrödinger equation with spectrum zero and general superlinear nonlinearities. J. Math. Anal. Appl. 364(2), 404–413 (2010)
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by A. Malchiodi.
Rights and permissions
About this article
Cite this article
Schechter, M. Global solutions of nonlinear Schrödinger equations. Calc. Var. 56, 40 (2017). https://doi.org/10.1007/s00526-017-1145-5
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00526-017-1145-5