On stability and instability of standing waves for 2d-nonlinear Schrödinger equations with point interaction

We consider the following nonlinear Schrödinger equation (NLS) with a focusing power nonlinearity in two spatial dimension:

where $p>1$ and $-\Delta_{\alpha}$ is the Laplacian with a point interaction with strength $\alpha\in\mathbb{R}$ at the origin (see (1.2)–(1.5) for the precise definition).

The study of the Laplace operator with point interactions in $\mathbb{R}^{N}$ ($N=1$, $2$, or $3$) seems to become intensive research area in the last decades. The first rigorous attempt to define and study the spectral properties of these operators was done in [14] by Berezin and Faddeev in 1961. Their study was extended in the work [10] by Albeverio and Høegh-Krohn in 1981.

A simplest way to introduce a singular perturbation at a point is to consider the potential perturbation $-\Delta+V_{\varepsilon}$ of the Laplace operator with regular potentials $V_{\varepsilon}$ spiking up and shrinking around a point in the limit $\varepsilon\to 0$. This approach is well-known for dimension $N=1$ [9], $N=2$ [7], and $N=3$ [6] (we also refer to [8] for a comprehensive overview). In the case of one singular point, this approximation gives a self-adjoint operator $-\Delta_{\alpha}$ depending on the parameter $\alpha\in\mathbb{R}$. To be more precise, starting with the symmetric operator of $-\Delta|_{C_{c}^{\infty}(\mathbb{R}^{N}\setminus\{0\})}$, one can characterize all its nontrivial self-adjoint extension on $L^{2}(\mathbb{R}^{N})$ by means of a parameter $\alpha\in\mathbb{R}$. A detailed overview of the construction and the main properties of $-\Delta_{\alpha}$ can be found in [18]. Fractional powers of the Laplace operator $-\Delta_{\alpha}$ with singular perturbations are studied in [35, 49].

The dispersive properties of the Schrödinger group $(e^{it\Delta_{\alpha}})_{t\in\mathbb{R}}$ have been studied intensively in the last years. In one-dimensional case, the dispersive and Strichartz estimates and boundedness wave operators have been studied, for example, in [4, 27, 39]. In higher dimensional case, such properties have been studied recently by [22, 59] in $N=2$, [25] in $N=3$. See also [24] for the weighted dispersive estimate in three dimension.

By using these dispersive properties, one can establish existence of nonlinear evolution flow. The local well-posedness for (1.1) (in two or three-dimensional cases) in the domain $D(-\Delta_{\alpha})$ has been studied in [18]. See, e.g., [4, 27] for the one-dimensional case.

Recently, the nonlinear dynamics around standing waves in one dimension are intensively studied in various contexts depending on whether the potential is attractive or repulsive. In the attractive regime, the orbital stability and instability of standing waves have been studied by [34, 36, 41], the asymptotic stability has been done in [23, 46, 47], and the strong instability has been done in [51, 29]. In the repulsive regime, the orbital stability and instability have been studied by [31, 45], and the global dynamics below ground states were studied in [40].

As related topics, we mention the NLS with the concentrated nonlinearity. See [3, 17, 42] for $N=1$, [1, 2] for $N=2$, and [5] for $N=3$.

However, much less is known about the nonlinear dynamics and the existence and stability/instability properties of the ground states for (1.1). Up to our knowledge, there are no results treating such properties in the two-dimensional case. The main goal of the work is to cover this case and establish fundamental properties of ground states for (1.1) such as existence, uniqueness, nondegeneracy, and its orbital stability/instability.

To state our main results, we shortly give the definition of the operator $-\Delta_{\alpha}$ in two dimension (see [8, Chapter I.5] for more details). The class of self-adjoint extensions in $L^{2}(\mathbb{R}^{2})$ of the positive and densely defined symmetric operator $-\Delta|_{C^{\infty}_{0}(\mathbb{R}^{2}\setminus\{0\})}$ is a one-parameter family of operators $(-\Delta_{\alpha})_{\alpha\in(-\infty,+\infty]}$. The extension $-\Delta_{\alpha=\infty}$ is the Friedrichs extension and is precisely the free Laplacian on $L^{2}(\mathbb{R}^{2})$ with domain $H^{2}(\mathbb{R}^{2})$. All other extensions for $\alpha\in\mathbb{R}$ represent nontrivial operators with a point interaction at the origin, and are characterized explicitly by

Here $\lambda>0$ is a fixed constant with $\lambda\neq-e_{\alpha}$ (see (1.11) for the definition of $e_{\alpha}$),

$\gamma>0$ denoting Euler–Mascheroni constant, and $G_{\lambda}$ is the Green function of $-\Delta+\lambda$ on $\mathbb{R}^{2}$, defined by the distributional relation $(-\Delta+\lambda)G_{\lambda}=\delta$, that is

where $\mathcal{F}^{-1}$ is the inverse Fourier transform. Note that $f(0)$ makes sense for $f\in H^{2}(\mathbb{R}^{2})$ from the embedding $H^{2}(\mathbb{R}^{2})\hookrightarrow C(\mathbb{R}^{2})$.

From the the expression (1.5), we can easily check that

for all $\varepsilon>0$ and $\lambda,\mu>0$. The function $G_{\lambda}$ is also represented as

where $K_{0}$ is the modified Bessel function of the second kind (or the Macdonald function) of order zero. From the expression (1.8), $G_{\lambda}$ is positive, radial, and strictly decreasing function with

(see [26, Chapter 10] for more properties of $K_{\nu}$).

From the property (1.7) we can check that the definition of $-\Delta_{\alpha}$ is independent of $\lambda$. Indeed, if $g=f+f(0)\beta_{\alpha}(\lambda)^{-1}G_{\lambda}$ for some $f\in H^{2}(\mathbb{R}^{2})$ and $\lambda\neq-e_{\alpha}$, then for $\mu\neq-e_{\alpha}$, we have the decomposition

and the relation

This means the independence.

The spectral properties of $-\Delta_{\alpha}$ are also known (see [8, Theorem 5.4]). The essential spectrum of $-\Delta_{\alpha}$ are given by

where $\sigma_{\mathrm{ac}}$ and $\sigma_{\mathrm{sc}}$ denote the set of the absolutely and singularly continuous spectrum, respectively. The operator $-\Delta_{\alpha}$ has a simple negative eigenvalue

In this sense, $-\Delta_{\alpha}$ can be regarded as an Schrödinger operator with an attractive potential. The normalized eigenfunction corresponding to the eigenvalue $e_{\alpha}$ is

Note that $\beta_{\alpha}(\lambda)$ defined in (1.4) is expressed as

so one has

The energy space (or the form domain) $H_{\alpha}^{1}(\mathbb{R}^{2})$ associated with $-\Delta_{\alpha}$ can be characterized by general results on the Kreĭn–Višik–Birman extension theory (see e.g. [48]), and it is given explicitly by

Note that $H_{\alpha}^{1}(\mathbb{R}^{2})$ is independent of the choice of $\alpha\in\mathbb{R}$ and $\lambda>0$ from (1.7). For $\lambda>-e_{\alpha}$, one can define the maximal extension of the form

with $g=f+f(0)\beta_{\alpha}(\lambda)^{-1}G_{\lambda}\in D(-\Delta_{\alpha})$. This extension defines a positive quadratic form well-defined on $g\in H_{\alpha}^{1}(\mathbb{R}^{2})$ explicitly given by

for $g=f+cG_{\lambda}\in H_{\alpha}^{1}(\mathbb{R}^{2})$. By using the relation

we can rewrite (1.13) as

We denote the $H_{\alpha}^{1}$-norm depending on the parameter $\lambda>-e_{\alpha}$ by

for $g\in H_{\alpha}^{1}(\mathbb{R}^{2})$. The following equivalence holds:

As a special case, we choose $\lambda=1-e_{\alpha}$ and also use the notation

To study the stability properties of the standing waves, we need the local well-posedness in the energy spaces $H^{1}_{\alpha}(\mathbb{R}^{2})$. We have the following statement (see Appendix B for the proof).

Now let us consider standing wave solutions with the form

where $\phi\in H_{\alpha}^{1}(\mathbb{R}^{2})$ is a nontrivial solution of the stationary equation

Equation (1.16) can be rewritten as $S_{\omega}^{\prime}(\phi)=0$, where $S_{\omega}$ is the action functional defined by

We denote the set of all nontrivial solution of (1.16) by

and the set of all ground states (minimal action solution) by

Now we state our main results. First, we state the results about the existence, symmetry, and uniqueness of ground states.

Next, we state the results about orbital stability and instability of standing waves. The definition of orbital stability is as follows.

The following two statements are main results of this paper. The first one concerns the stability for $\omega$ close to $-e_{\alpha}$.

The second one concerns the stability/instability for large frequency $\omega$. We denote the unique ground state given in Theorem 1.4 by $\phi_{\omega}$.

One can observe the similarity between the results in [28, 30, 32, 33, 34] and Theorems 1.6 and 1.7. The paper [34] treats NLS with attractive $\delta$-potential in one dimension, and the papers [28, 30, 32, 33] concern NLS with general attractive potential $V(x)$. Since $-\Delta_{\alpha}$ has a unique simple negative eigenvalue, we regard it as a Schrödinger operator with an attractive potential, so it is natural to choose to follow the approach in these papers.

Let us give a short outline of the proofs. The local well-posedness (Proposition 1.1) follows from the energy methods in [52] (see also [20, Chapter 3]) and the Strichartz estimates for the operator $-\Delta_{\alpha}$ obtained by [22]. The existence of ground states (Theorem 1.2) follows from a standard variational method by using the Nehari manifold. The positivity and symmetry of ground states (Theorem 1.3) follow from the maximal principle and the symmetric rearrangement. In particular, we use the result of Brothers and Ziemer [16] to obtain the radial symmetry and decrease of ground states.

To investigate the stability properties, we consider rescaled ground states and use a perturbation argument as in [30, 32, 33]. Let $(\phi_{\omega})_{\omega>-e_{\alpha}}$ be a family of positive ground states with $\phi_{\omega}=\phi_{\alpha,\omega}\in\mathcal{G}_{\alpha,\omega}$. For $\omega$ close to $-e_{\alpha}$, we normalize the ground states as

Then $\widehat{\phi}_{\omega}$ is a positive solution of

Since we can verify $\lVert\phi_{\omega}\rVert_{L^{2}}\to 0$ as $\omega\to-e_{\alpha}$, $\widehat{\phi}_{\omega}$ can be regarded as a perturbation of the solution for the linear equation

that is, $\widehat{\phi}_{\omega}$ is close to the eigenfunction $\chi_{\alpha}$. On the other hand, for large $\omega$, we rescale $\phi_{\omega}$ as

Then $\widetilde{\phi}_{\omega}$ is a positive solution of

where

Since $-\Delta_{\alpha=\infty}$ is the free Laplacian $-\Delta$, $\widetilde{\phi}_{\omega}$ can be regarded as a perturbation of the solution for the equation

Therefore, we can investigate the stability properties by using the limiting equation (1.19) for small $\omega$ and (1.20) for large $\omega$.

The stability for small frequency (Theorem 1.6) follows from the argument of [33]. If $\omega$ is sufficiently close to $-e_{\alpha}$, we can obtain the following coercivity property for the linearlized operator around the ground state.

It is known that this coercivity implies the stability (see, e.g., [37, 44]). Because we can prove Proposition 1.8 exactly in the same way as [33, Section 4], we omit the proof.

To investigate the properties of the ground states for large frequency $\omega$, we use the limiting equation (1.20). It is well known that (1.20) has the unique positive radial ground state $\widetilde{\phi}_{\infty}\in H^{1}(\mathbb{R}^{2})$ (see, e.g., [12, 43]), and it is nondegenerate in the radial space, that is, the kernel of the linearized operator $\widetilde{L}_{\infty}\mathrel{\mathop{:}}=-\Delta+1-p\widetilde{\phi}_{\infty}^{p-1}$ is trivial: $\ker\widetilde{L}_{\infty}|_{H_{\mathrm{rad}}^{1}}=\{0\}$. By using these properties, we establish the uniqueness (Theorem 1.18) and nondegeneracy (Lemma 6.1) for large $\omega$ following the argument of [30, Proposition 2 (v)]. Moreover, we can obtain the regularity of the map $\omega\mapsto\phi_{\omega}$ (Corollary 8.2). To obtain the stability and instability, we use the following criteria.

From Proposition 1.10, the stability/instability problems can be reduced to the investigation of the sign of the derivative $\frac{d}{d\omega}\|\phi_{\omega}\|_{L^{2}}^{2}$. When $\alpha=\infty$, i.e., without interaction, one can show by the scaling invariance for the equation that the ground states $\phi_{\infty,\omega}$ of (1.16) satisfies $\frac{d}{d\omega}\|\phi_{\infty,\omega}\|_{L^{2}}^{2}>0$ if $1<p<3$ and $\frac{d}{d\omega}\|\phi_{\infty,\omega}\|_{L^{2}}^{2}<0$ if $p>3$ for all $\omega>0$. This means that when $\alpha=\infty$, the ground-state standing wave $e^{i\omega t}\phi_{\infty,\omega}$ of (1.1) is stable if $1<p<3$ [19] and unstable if $p\geq 3$ [13] (see [57] for $p=3$).

To investigate the sign of $\frac{d}{d\omega}\|\phi_{\omega}\|_{L^{2}}^{2}$ for large $\omega$, we apply the argument of [30, 28]. Instead of $\frac{d}{d\omega}\|\phi_{\omega}\|_{L^{2}}^{2}$, we calculate the rescaled version $\frac{d}{d\omega}\|\widetilde{\phi}_{\omega}\|_{L^{2}}^{2}$ and use the convergence $\widetilde{\phi}_{\omega}\to\widetilde{\phi}_{\infty}$. To estimate some error terms, we establish and use a boundedness of the inverse linearized operator of $\widetilde{\phi}_{\omega}$. After that, we can determine the sign of $\frac{d}{d\omega}\|\phi_{\omega}\|_{L^{2}}^{2}$, and combining Proposition 1.10 we obtain Theorem 1.7.

The difficulty of the proofs of our results mainly comes from the treatment of functions in the energy space $H_{\alpha}^{1}(\mathbb{R}^{2})$ and the domain $D(-\Delta_{\alpha})$. Of special importance in one-dimensional case is the fact that the fundamental solution of $(1-\Delta)$ is in $H^{1}(\mathbb{R})$, so one can use $H^{1}(\mathbb{R})$ as a natural space of the nonlinear flow associated with the corresponding NLS. The situation changes essentially in two dimension since we are forced to work with the perturbed $H_{\alpha}^{1}(\mathbb{R}^{2})$ space, so there are nontrivial difficulties to apply of the variational technique from [34] and the cases of slowly decaying potentials [28, 30, 32, 33]. For a function in the spaces $H_{\alpha}^{1}(\mathbb{R}^{2})$ or $D(-\Delta_{\alpha})$, we need to decompose it into the regular and singular parts and to treat these separately, and we have to avoid several difficult points requiring appropriate new treatments.

• •

  The local well-posedness for the standard 2d NLS with or without potential requires the use of Strichartz estimates in Sobolev spaces

  (1.21)

  $$H^{s,p}(\mathbb{R}^{2})=\{g=(1-\Delta)^{-s}\psi\colon\,\psi\in L^{p}(\mathbb{R}^{2})\},\quad p\in(1,2),\ s\in(0,1],$$

  if a contraction argument is applied. On the other hand, the case of singular perturbed Laplacian $-\Delta_{\alpha}$ requires the replacement of the classical Sobolev space $H^{1}(\mathbb{R}^{2})$ by the perturbed space $H^{1}_{\alpha}(\mathbb{R}^{2})$, and we need to decompose functions $g\in H^{1}_{\alpha}(\mathbb{R}^{2})$ as

  (1.22)

  $$g=f+cG_{\lambda}.$$

  There is no flexible treatment (up to our knowledge) of appropriate generalization of generic spaces $H^{s,p}$ for the Laplacian of type $-\Delta_{\alpha}$. For this we have chosen another approach based on compactness argument and the results in [52].

• •

  The existence of ground states seems to be closely connected with the inclusion

  $$H^{1}(\mathbb{R}^{2})\subset H^{1}_{\alpha}(\mathbb{R}^{2}).$$

  However, the ground states $\phi=f+f(0)\beta_{\alpha}(\omega)^{-1}G_{\omega}$ from Theorem 1.3 have nontrivial singular part, since $e^{i\theta}f(0)$ is positive. This fact shows that the ground states associated with $\alpha\in\mathbb{R}$ are different from ground states with $\alpha=\infty$. Moreover the ground state $\phi$ from Theorem 1.3 is not in $H^{1}(\mathbb{R}^{2})$.

• •

  The symmetry of the ground state for the classical NLS can be obtained by Schwartz symmetrization. Since we have the decomposition $\phi=f+f(0)\beta_{\alpha}(\omega)^{-1}G_{\omega}$ for any $\phi\in D(-\Delta_{\alpha})$ into regular and singular parts, a formal symmetrization

  $$\phi^{*}=\left(f+\frac{f(0)}{\beta_{\alpha}(\lambda)}G_{\lambda}\right)^{*}$$

  cannot work. We have chosen the following symmetrization

  $$f+\frac{f(0)}{\beta_{\alpha}(\lambda)}G_{\lambda}\to f^{*}+\frac{|f(0)|}{\beta_{\alpha}(\lambda)}G_{\lambda}.$$

  The technical difficulties associated with this symmetrization can be overcome by using the results in [11].

• •

  The uniqueness and nondegeneracy of ground states require careful use of the decomposition (1.22) and modify accordingly the approach in [32, 33, 30].

• •

  To determine the sign of $\frac{d}{d\omega}\|\widetilde{\phi}_{\omega}\|_{L^{2}}^{2}$, we need to estimate the error term $\partial_{\omega}\widetilde{f}_{\omega}(0)$ coming from the interaction of $-\Delta_{\alpha}$, where $\widetilde{f}_{\omega}$ is the regular part of $\widetilde{\phi}_{\omega}$. As in the previous work [28, 30], we use the boundedness of the inverse of the linearized operator $\widetilde{L}_{\omega}$. However, it is not trivial how to express and estimate the term $\partial_{\omega}\widetilde{f}_{\omega}(0)$ by using the operator $\widetilde{L}_{\omega}^{-1}$. To overcome this difficulty we make a good use the expression (1.3) of the operator and the expression of the bilinear form as

  $$f(0)=\left\langle(\Delta_{\alpha}+\lambda)G_{\lambda},f+\frac{f(0)}{\beta_{\alpha}(\lambda)}G_{\lambda}\right\rangle$$

  for $f\in H^{2}(\mathbb{R}^{2};\mathbb{R})$. For more details, see Lemma 9.4.

The rest of organization of this paper is as follows. In Section 2 we correct the properties of $G_{\lambda}$ used in this paper. In Section 3 we prove Theorem 1.2 through the characterization with the Nehari functional. Section 4 is devoted to the proof of Theorem 1.3. In Section 5 we show that a family of rescaled ground states converges to the ground state of NLS without interaction (i.e. $\alpha=\infty$) as $\omega\to\infty$. In Section 6 we show lower boundedness and nondegeneracy of the linearized operator around the ground state for large $\omega$. This lower boundedness will be used in Section 9 as the boundedness of the inverse operator. In Section 7 we prove Theorem 1.18. In Section 8 we discuss the regularity of the map $\omega\mapsto\phi_{\omega}$ for large $\omega$. Finally we prove Theorem 1.7 in Section 9. In Appendix A we review the properties of wave operators and Strichartz estimates for the operator $-\Delta_{\alpha}$. In Appendix B we prove Proposition 1.1.

The aim of this section is to recall the main properties of the singular-perturbed Laplace operator $-\Delta_{\alpha}$ and the Green function $G_{\lambda}$.

Note that (1.6) and (1.9) imply

This fact leads easily to the following Sobolev inequality

In this section, we prove existence of ground states for (1.16) by using a standard variational method and properties of the operator $-\Delta_{\alpha}$. Throughout this section, we fix $\omega>-e_{\alpha}$. We define the Nehari functional by

for $v\in H_{\alpha}^{1}(\mathbb{R}^{2})$. We denote

For simplicity of notations, we shall often omit the subscript $\alpha$ like $S_{\omega}$, $K_{\omega}$, and so on. We note that $\mathcal{G}_{\omega}\subset\mathcal{A}_{\omega}\subset\mathcal{K}_{\omega}$.

We will prove the following.

By using the functional $K_{\omega}$, we can rewrite the action as

and $d_{\alpha}(\omega)$ as