1: %%Last version sent after the third referee's report (17/01/02)
2:
3: \documentclass{ajour}
4: \usepackage{amsmath,amssymb}
5: %\usepackage[dvips]{graphicx}
6: %\usepackage{showkeys}
7:
8: %\vsize=27.9cm
9: %\hsize=21cm
10: %\setlength{\topmargin}{16pt} \setlength{\headheight}{20pt}
11: %\setlength{\headsep}{30pt} \setlength{\textwidth}{15cm}
12: %\setlength{\textheight}{18cm}
13: %\setlength{\oddsidemargin}{1cm} %odd pages
14: %\setlength{\evensidemargin}{1cm} %even pages
15:
16:
17: %\renewcommand{\baselinestretch}{1.5} %%2 means double spaced; 1.5 means ...
18: %\textwidth=150mm
19: %\textheight=185mm
20: %\parindent=8mm
21: %\frenchspacing
22:
23:
24: %\baselineskip=22pt
25:
26: \newcommand{\cat}{\operatorname{cat}}
27: \renewcommand{\tfrac}{\textstyle\frac}
28: \newcommand{\R}{{\mathbb R}} %%reali
29: \newcommand{\Rn}{{{\mathbb R}^n}} %% R^n
30: \newcommand{\C}{{\mathbb{C}}}%%C
31: \newcommand{\N}{{\mathbb N}} %% naturali
32: \newcommand{\Z}{{\mathbb Z}} %% interi
33: \newcommand{\weakly}{\rightharpoonup}
34: \newcommand{\bra}{\langle} %% apre prod. scal. in Rn
35: \newcommand{\ket}{\rangle} %%chiude prod. scal.
36: \newcommand{\limn}{\lim_{n \to +\infty}}
37: \newcommand{\0}{\R\setminus \{0\}}
38: \newcommand{\iy}{\infty}
39: \newcommand{\p}{\partial}
40: \newcommand{\irn}{\int\limits_{\Rn}}
41: \newcommand{\map}{\longrightarrow }
42: \newcommand{\imp}{\Longrightarrow }
43: \newcommand{\re}{\operatorname{Re}}
44: \newcommand{\Wn}{W^{1,2}(\Rn)}
45: \newcommand{\dob}{{\mathcal D}^{1,2}(\Rn)}
46: \newcommand{\ulu}{\underline u}
47: \newcommand{\ulw}{\underline w}
48: \newcommand{\ulz}{\underline z}
49: \newcommand{\ulv}{\underline v}
50: \newcommand{\uls}{\underline s}
51: \newcommand{\olu}{\overline u}
52: \newcommand{\olv}{\overline v}
53: \newcommand{\ols}{\overline s}
54: \newcommand{\olb}{\overline\b}
55: \newcommand{\olx}{\overline\xi}
56: \newcommand{\ovar}{\overline\var}
57: \newcommand{\dw}{\dot{w}}
58: \newcommand{\dz}{\dot{z}}
59: \newcommand{\wv}{\widetilde v} \newcommand{\wu}{\widetilde u}
60: \newcommand{\wz}{\widetilde z}
61: \newcommand{\wZ}{\widetilde Z} \newcommand{\ws}{\widetilde s}
62: \newcommand{\ball}{{\mathcal B}} \newcommand{\qd}{\qquad\qquad}
63: \newcommand{\qdd}{\qd\qd}
64:
65: %%%%%%%%%%%%%%%%%%%% %definici\'on de letras griegas
66: %%%%%%%%%%%%%%%%%%%%
67: \renewcommand{\a }{\alpha }
68: \renewcommand{\b }{\beta }
69: \renewcommand{\d }{\delta }
70: \newcommand{\D }{\Delta }
71: \newcommand{\e }{\varepsilon }
72: \newcommand{\g }{\gamma}
73: \newcommand{\G }{\Gamma }
74: \renewcommand{\l }{\lambda }
75: \renewcommand{\L }{\Lambda }
76: \newcommand{\m }{\mu }
77: \newcommand{\n }{\nabla }
78: \newcommand{\var }{\varphi }
79: \newcommand{\s }{\sigma }
80: \newcommand{\Sig }{\Sigma}
81: \renewcommand{\t }{\tau }
82: \renewcommand{\o }{\omega }
83: \renewcommand{\O }{\Omega }
84: \newcommand{\z }{\zeta }
85: \newcommand{\de}{\partial}
86: %%%%%%%%%%%%%%%%%%%%
87: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
88: %\newenvironment{pf}{\noindent{\sc Proof}.\enspace}{\rule{2mm}{2mm}\medskip}
89: %\newtheorem{Theorem}{Theorem}[section]
90: %\newtheorem{Corollary}[Theorem]{Corollary}
91: %\newtheorem{Lemma}[Theorem]{Lemma}
92: %\newtheorem{Proposition}[Theorem]{Proposition}
93: %\newtheorem{definition}[Theorem]{Definition}
94: %\newtheorem{remark}[Theorem]{Remark}
95: %\newtheorem{remarks}[Theorem]{Remarks}
96: \newenvironment{Remark}{\begin{remark} \rm}{\end{remark}}
97: %% scrive i remarks non in corsivo
98: %\newenvironment{Remarks}{\begin{remarks}}
99:
100:
101: \begin{document}
102:
103: \authorrunninghead{Cingolani, Secchi}
104: \titlerunninghead{Schr\"odinger equations with electromagnetic fields}
105:
106: \title{\bf Semiclassical limit for
107: nonlinear Schr\"odinger equations with
108: electromagnetic fields}
109:
110:
111: %\author{Silvia CINGOLANI \thanks{Supported by MURST}\\
112: %Dip. Inter. Matematica, Politecnico di Bari, Via E.Orabona 4,
113: %70125 Bari, Italy \\
114: % \and
115: %\\
116: %Simone SECCHI \thanks{Supported by MURST, national project
117: %\emph{Variational methods and nonlinear differential equations}.}\\ SISSA,
118: %Via Beirut 2-4, 34013 Trieste, Italy}
119:
120:
121: \author{Silvia CINGOLANI \thanks{Supported by MURST}}
122: \affil{Dip. Inter. Matematica, Politecnico di Bari, Via E.Orabona 4,
123: 70125 Bari, Italy}
124: \email{cingolan@dm.uniba.it}
125:
126: \and
127:
128: \author{Simone SECCHI \thanks{Supported by MURST, national project
129: \emph{Variational methods and nonlinear differential equations}.}}
130: \affil{SISSA, via Beirut 2/4, 34014 Trieste, Italy}
131: \email{secchi@ma.sissa.it}
132:
133: %\date{}
134:
135: \begin{article}
136:
137: %\maketitle
138:
139: %Adress for manuscript correspondence:
140: %
141: %Silvia Cingolani,
142: %
143: %Dip. Inter. Matematica, Politecnico di Bari, Via E.Orabona 4,
144: %70125 Bari, Italy
145: %
146: %e-mail: cingolan@pascal.dm.uniba.it
147: %
148: %phone number: +390805963699, \quad fax number: +390805963612
149: %
150: %
151: %Running title: Schr\"odinger equations with electromagnetic fields
152:
153: \section{Introduction}\label{sec:int}
154:
155: Let us consider the nonlinear Schr\"{o}dinger equation
156: \begin{equation}\label{eq:1.1}
157: ih \frac{\de \psi}{\de t} = \left( \frac{h}{i} \nabla -A(x)\right)^2
158: \psi +U(x)\psi - f(x,\psi), \quad x\in\Rn
159: \end{equation}
160: where $t \in \R$, $x \in \Rn$ $(n \geq 2)$.
161: The function $\psi(x,t)$ takes on complex values,
162: $h$ is the Planck constant, $i$ is the imaginary unit.
163: Here $A\colon \Rn \to \Rn$ denotes a magnetic potential and
164: the Schr\"odinger operator is defined by
165: $$
166: \left( \frac{h}{i} \nabla -A(x)\right)^2 \psi
167: := -h^2 \varDelta \psi - \frac{2h}{i} A \cdot \nabla \psi + |A|^2 \psi -
168: \frac{h}{i} \psi\, \operatorname{div}A\, .
169: $$
170: Actually, in general dimension $n \geq 2$,
171: the magnetic field $B$ is a $2$-form where
172: $B_{i,j}= \partial_j A_k - \partial_k A_j$ ;
173: in the case $n=3$, $B= \operatorname{curl}A$.
174: The function $U\colon \Rn \to \R$ represents an electric potential.
175: In the sequel, for the sake of simplicity, we limit ourselves
176: to the particular case in which $f(x,t)=K(x)|t|^{p-1}t$, with $p>1$ if
177: $n=2$ and $1<p<\frac{n+2}{n-2}$ if $n\geq 3$.
178:
179: It is now well known that the
180: nonlinear Schr\"{o}dinger equation \eqref{eq:1.1}
181: arises from a perturbation approximation for strongly nonlinear
182: dispersive wave systems. Many papers are devoted to the nonlinear
183: Schr\"{o}dinger equation and its solitary wave solutions.
184:
185: In this paper we seek for standing wave solutions to
186: \eqref{eq:1.1}, namely waves of the form
187: $\psi (x,t)=e^{-iEth^{-1}}u(x)$ for some function
188: $u\colon \Rn \longrightarrow \C$.
189: Substituting this
190: {\it ansatz} into \eqref{eq:1.1}, and denoting for
191: convenience $\e=h$, one is led to solve the complex equation in $\Rn$
192: \begin{equation}
193: \left( \tfrac{\e}{i} \nabla -A(x)\right)^2 u +
194: (U(x)-E) u = K(x)|u|^{p-1}u \,.
195: \tag{NLS}
196: \end{equation}
197: Renaming $V(x)+1=U(x)-E$, we assume from now on that $1+V$ is
198: strictly positive on the whole $\Rn$.
199: Moreover, by an obvious change of variables,
200: the problem becomes that of finding some function
201: $u\colon\Rn\to\C$ such that
202: \begin{equation}\label{S}
203: \left( \tfrac{\nabla}{i} -A(\e x)\right)^2 u + u +
204: V(\e x)u = K(\e x)|u|^{p-1}u, \quad x\in\Rn \,.
205: \tag{$S_\e$}
206: \end{equation}
207: Concerning nonlinear Schr\"odinger equation with external magnetic field,
208: we firstly quote a paper by Esteban and Lions \cite{EL}, where
209: concentrations and compactness arguments are applied to solve some
210: minimization problems associated to $(S_\e)$ under
211: suitable assumptions on the magnetic field.
212:
213: The purpose of this paper is
214: to study the time--independent nonlinear
215: Schr\"{o}dinger equation $(S_\e)$ in the semiclassical limit.
216: This seems a very interesting problem since
217: the Correspondence's Principle establishes that
218: Classical Mechanics is, roughly speaking, contained in
219: Quantum Mechanics.
220: The mathematical transition is obtained letting to zero
221: the Planck constant ($\e \to 0$)
222: and solutions $u(x)$ of $(S_\e)$
223: which exist for small value of $\e$ are
224: usually referred as semi-classical ones (see \cite{RS}).
225:
226: We remark that in the
227: linear case, Helffer {\it et al.} in
228: \cite{hel1,helS} have studied the asymptotic
229: behavior of the eigenfunctions
230: of the Schr\"{o}dinger operators with magnetic fields
231: in the semiclassical limit.
232: Note that in these papers
233: the {\sl wells} of the Schr\"{o}dinger operators with magnetic fields
234: are the same as those without magnetic field,
235: so that one doesn' t `see' the magnetic field in the
236: definition of the {\sl well}. See also \cite{brum}
237: for generalization of the results by \cite{helS}
238: for potentials which degenerate at infinity.
239:
240: In the case $A=0$, (no magnetic field),
241: a recent extensive literature is devoted to
242: study the time--independent nonlinear
243: Schr\"{o}dinger equation $(S_\e)$ in the semi-classical limit.
244: We shortly recall the main results in literature.
245: The first paper is due to Floer and Weinstein which
246: investigated the one-dimensional nonlinear Schr\"{o}dinger equation
247: (with $K(x)=1$) and gave a description of the limit behavior of $u(x)$
248: as $\e \to 0$.
249: Really they proved that if the potential $V$ has a non-degenerate
250: critical point, then $u(x)$ concentrates near this
251: critical point, as $\e \to 0$.
252:
253: Later, other authors proved that this problem is really local in nature and
254: the presence of an isolated critical point of the potential $V$
255: (in the case $K(x)=1$)
256: produces a semi-classical solution $u(x)$ of $(S_\e)$ which
257: concentrates near this point.
258: Different approaches are used to cover different cases
259: (see \cite{ABC,dpf,Li,oh,xw}).
260: Moreover when $V$ oscillates, the existence of multibumps solutions has also
261: been studied in \cite{amberti,CN,dpf1,gui}.
262: Furthermore multiplicity results are obtained in
263: \cite{CL1,CL2} for potentials $V$ having a set of degenerate
264: global minima and recently in \cite{ams},
265: for potentials $V$ having a set of
266: critical points, not necessarily global minima.
267:
268:
269: \smallskip
270: A natural answer arises:
271: {\sl how does the presence of an external magnetic field
272: influence the existence and the
273: concentration behavior of
274: standing wave solutions to \eqref{eq:1.1} in the semi-classical limit}?
275: \smallskip
276:
277: A first result in this direction is contained in \cite{ku}
278: where Kurata has proved the existence of
279: least energy solutions to $(S_\e)$ for any $\e >0$,
280: under some assumptions linking the magnetic field
281: $B= (B_{i,j})$ and the electric potential $V(x)$.
282: The author also investigated the semi-classical limit of the found
283: least energy solutions and
284: showed a concentration phenomenon near global
285: minima of the electric potential in the case $K(x) = 1$
286: and $|A|$ is small enough.
287:
288: Recently in \cite{cingolani}, Cingolani
289: obtained a multiplicity result of semi-classical standing waves
290: solutions to $(S_\e)$,
291: relating the number of solutions to $(S_\e)$ to
292: the {\sl richness} of a set $M$ of global minima
293: of an auxiliary function $\Lambda$ defined by setting
294: \[
295: \Lambda (x)=\frac{(1+V(x))^\theta}{K(x)^{-2/(p-1)}},\qquad \theta =
296: \frac{p+1}{p-1}-\frac{n}{2},
297: \]
298: (see $(\ref{def:auxiliary})$ in section 4 for details) depending on $V(x)$ and $K(x)$.
299: We remark that, if $K(x)=1$ for any $x \in \Rn$, global
300: minima of $\Lambda$ coincides with global minima of $V$.
301: The variational approach, used in \cite{cingolani},
302: allows to deal with unbounded potential $V$
303: and does not require assumptions on the magnetic field.
304: However this approach works only near global minima of
305: $\Lambda$.
306:
307: In the present paper we deal with the more
308: general case in which the auxiliary function
309: $\Lambda$ has a manifold $M$ of stationary points,
310: not necessarily global minima.
311: For bounded magnetic potentials $A$,
312: we are able to prove a multiplicity result of semi-classical
313: standing waves of $(S_\e)$,
314: following the new perturbation approach
315: contained in the recent paper \cite{ams} due to
316: Ambrosetti, Malchiodi, Secchi.
317:
318: Now we briefly describe the proof of the result.
319: First of all, we highlight that solutions of $(S_\e)$ naturally
320: appear as {\em orbits}: in fact, equation $(S_\e)$ is invariant under the
321: multiplicative action of $S^1$. Since there is no danger of confusion, we
322: simply speak about solutions.
323: The complex--valued solutions to $(S_\e)$ are
324: found near least energy solutions of the equation \begin{equation}\label{risc}
325: \bigg(\frac{\nabla}{i} - A(\e \xi)\bigg)^2u + u + V(\e \xi)u=K(\e\xi)|u|^{p-1}u.
326: \end{equation}
327: where $\e \xi$ is in a neighborhood of $M$.
328: The least energy of (\ref{risc}) have the form
329: \begin{equation}\label{def:zed}
330: z^{\e\xi ,\sigma} \colon x\in\Rn \mapsto e^{i\sigma +
331: i A(\e\xi)\cdot x}
332: \left(\frac{1+V(\e\xi)}{K(\e\xi)}\right)^\frac{1}{p-1}
333: U((1+V(\e\xi))^{1/2} (x-\xi))
334: \end{equation}
335: where $\e \xi$ belongs to $M$ and $\sigma \in [0, 2 \pi]$.
336: As in \cite{ams}, the proof relies on a suitable
337: finite dimensional reduction,
338: and critical points of the Euler functional $f_\e$ associated to problem $(S_\e)$
339: are found near critical point of a finite dimensional
340: functional $\Phi_\e$ which is defined on a suitable neighborhood of $M$.
341: This allows to use Lusternik-Schnirelman category in the case $M$ is a
342: set of local maxima or minima of $\Lambda$.
343: We remark that the case of maxima cannot be handled by using
344: direct variational arguments as in \cite{cingolani}.
345:
346: Moreover in the case $M$ is a set of critical points non-degenerate
347: in the sense of Bott (see \cite{bott}) we are able to prove the existence of (at least) cup long of
348: $M$ solutions
349: concentrating near points of $M$. For the definition of the cup long, refer to section 5.
350:
351: Firstly we present a special case of our results.
352:
353: \begin{theorem}\label{th:nond}
354: Assume that
355: \begin{itemize}
356: \item[$(K1)$] \, $K\in L^\infty (\Rn)\cap C^2(\Rn)$ is strictly positive and $K''$ is
357: bounded;
358: \item[$(V1)$] \, $V\in L^\infty (\Rn)\cap C^2(\Rn)$ satisfies $\inf_{x\in\Rn}
359: (1+V(x)) > 0$, and $V''$ is bounded;
360: \item[$(A1)$] \, $A\in L^\infty (\Rn,\Rn)\cap C^1(\Rn,\Rn)$, and
361: the jacobian $J_A$ of $A$ is globally bounded in $\Rn$.
362: \end{itemize}
363:
364: \noindent
365: If the auxiliary function $\Lambda$ has
366: a non-degenerate critical point $x_0\in \Rn$, then
367: for $\e > 0$ small, the problem $(S_\e)$ has at least a
368: (orbit of) solution concentrating near $x_0$.
369: \end{theorem}
370:
371: Actually, we are able to prove the following generalization.
372:
373: \begin{theorem}\label{th:main}
374: As in Theorem 1.1, assume again (K1), (V1) and (A1) hold.
375:
376: \noindent
377: If the auxiliary function $\Lambda$ has
378: a smooth, compact, non-degenerate manifold of critical points $M$,
379: then for $\e > 0$ small, the problem $(S_\e)$ has at least
380: $\ell (M)$ (orbits of) solutions concentrating near points of $M$.
381: \end{theorem}
382:
383: Finally we point out that the presence of
384: an external magnetic field
385: produces a phase in the complex wave which depends on
386: the value of $A$ near $M$.
387: Conversely the presence of $A$ does not seem to influence the location of
388: the peaks of the modulus of the complex wave. Although we will not deal with this problem,
389: we believe that in order to have a
390: local $C^2$ convergence of the solutions, some assumption about the smallness of the
391: magnetic potential $A$ should be added, as done in \cite{ku} for minima of $V$.
392:
393: \bigskip
394: Finally we point out that
395: Theorem~\ref{th:nond} and Theorem~\ref{th:main} hold for problems involving
396: more general nonlinearities. See Remark \ref{nonl} in the last section.
397:
398: \vspace{10pt}
399:
400:
401: \begin{center}\textbf{Notation}\end{center}
402: \begin{enumerate}
403: \item The complex conjugate of any number $z\in\C$ will be denoted by $\bar z$.
404: \item The real part of a number $z\in\C$ will be denoted by $\re z$.
405: \item The ordinary inner product between two vectors $a,b\in\Rn$ will be denoted by $a \cdot
406: b$.
407: \item From time to time, when no confusion can arise, we omit the symbol $dx$ in integrals over
408: $\Rn$.
409: \item $C$ denotes a generic positive constant, which may vary inside a chain of inequalities.
410: \item
411: We use the Landau symbols. For example $O(\e)$ is a generic function such that
412: $\limsup\limits_{\e\to 0} \frac{O(\e)}{\e} <
413: \infty$, and $o(\e)$ is a function such that $\lim\limits_{\e\to 0} \frac{o(\e)}{\e}=0$.
414: \end{enumerate}
415:
416:
417: \section{The variational framework}
418: We work in the real Hilbert space $E$ obtained as the completion of
419: $C_0^\infty (\Rn,\C)$ with respect to the norm associated to the inner product
420: \[
421: \langle u\mid v\rangle = \operatorname{Re} \int_\Rn \nabla u \cdot \overline{\nabla v} + u \bar v
422: ..
423: \]
424: Solutions to ($S_\e$) are, under some conditions we are going to point out, critical points of the
425: functional formally defined on $E$ as
426: \begin{multline}\label{feps}
427: f_\e (u)=\frac{1}{2}\int_\Rn \bigg(\bigg| \bigg(\frac{1}{i}\nabla - A(\e x)\bigg)u\bigg|^2
428: +|u|^2+V(\e x)|u|^2\bigg)\, dx \\- \frac{1}{p+1}\int_\Rn K(\e x)|u|^{p+1}\, dx.
429: \end{multline}
430: In what follows, we shall assume that the functions $V$, $K$ and $A$
431: satisfy the following assumptions:
432: \begin{itemize}
433: \item[(K1)] $K\in L^\infty (\Rn)\cap C^2(\Rn)$ is strictly positive and $K''$ is
434: bounded;
435: \item[(V1)] $V\in L^\infty (\Rn)\cap C^2(\Rn)$ satisfies $\inf_{x\in\Rn}
436: (1+V(x)) > 0$, and $V''$ is bounded;
437: \item[(A1)] $A\in L^\infty (\Rn,\Rn)\cap C^1(\Rn,\Rn)$, and
438: the jacobian $J_A$ of $A$ is globally bounded in $\Rn$.
439: \end{itemize}
440: Indeed,
441: \begin{multline*}
442: \int_\Rn \bigg(\bigg| \bigg(\frac{1}{i}\nabla - A(\e x)\bigg)u\bigg|^2 \bigg)\, dx
443: = \\ = \int_\Rn \bigg(|\nabla u|^2 + |A(\e x)u|^2 -2
444: \operatorname{Re} ( \textstyle\frac{\nabla u}{i} \cdot A(\e x)\overline{u}) \bigg)\, dx,
445: \end{multline*}
446: and the last integral is finite thanks to the Cauchy--Schwartz
447: inequality and the boundedness of $A$.
448:
449: \noindent It follows that $f_\e$ is actually well-defined on $E$.
450:
451: In order to find possibly multiple critical points of \eqref{feps},
452: we follow the approach of \cite{ams}. In our context, we need to find
453: complex--valued solutions, and so some further remarks are due.
454:
455: Let $\xi\in\Rn$, which will be fixed suitably later on: we look for solutions to \eqref{S} ``close'' to
456: a particular solution of the equation
457: \begin{equation}\label{eqlimite}
458: \bigg(\frac{\nabla}{i} - A(\e \xi)\bigg)^2u + u + V(\e\xi)u=K(\e\xi)|u|^{p-1}u.
459: \end{equation}
460: More precisely, we denote by $U_c \colon \Rn \to \C$ a least--energy solution
461: to the scalar problem
462: \begin{equation}\label{eq:4}
463: -\varDelta U_c + U_c + V(\e\xi)U_c = K(\e\xi) |U_c|^{p-1}U_c
464: \mbox{\quad in }\Rn.
465: \end{equation}
466: By energy comparison (see \cite{ku}), one has that
467: \[
468: U_c (x)=e^{i\sigma} U^\xi (x-y_0)
469: \]
470: for some choice of $\sigma \in [0,2\pi]$ and $y_0\in\Rn$, where
471: $U^\xi\colon\Rn\to\R$ is the unique solution of the problem
472: \begin{equation}
473: \begin{cases}
474: -\varDelta U^\xi + U^\xi + V(\e\xi) U^\xi = K(\e\xi)| U^\xi|^{p-1} U^\xi \\
475: U^\xi (0) = \max_\Rn U^\xi\\
476: U^\xi > 0.
477: \end{cases}
478: \end{equation}
479: If $U$ denotes the unique solution of
480: \begin{equation}
481: \begin{cases}
482: -\varDelta U + U = U^p \mbox{\quad in }\Rn\\
483: U(0) = \max_\Rn U \\
484: U > 0,
485: \end{cases}
486: \end{equation}
487: then some elementary computations prove that
488: $U^\xi (x)=\alpha(\e\xi)U(\beta(\e\xi)x)$,
489: where
490: \begin{eqnarray*}
491: \alpha (\e\xi) &= \left(\frac{1+V(\e\xi)}{K(\e\xi)}\right)^\frac{1}{p-1} \\
492: \beta(\e\xi)&=(1+V(\e\xi))^{1/2}.
493: \end{eqnarray*}
494: It is easy to show, by direct computation, that the function
495: $u(x)= e^{i A(\e\xi)\cdot x} U_c (x)$ actually solves \eqref{eqlimite}.
496:
497: For $\xi \in \Rn$ and $\sigma\in [0,2\pi]$, we set
498: \begin{equation}\label{def:zeta}
499: z^{\e\xi ,\sigma} \colon x\in\Rn \mapsto e^{i\sigma + i A(\e\xi)\cdot
500: x}\alpha(\e\xi)U(\beta(\e\xi)(x-\xi)).
501: \end{equation}
502:
503: Sometimes, for convenience, we shall identify $[0,2\pi]$ and $S^1\subset \C$, through $\eta =
504: e^{i\sigma}$.
505:
506: Introduce now the functional $F^{\e\xi,\sigma}\colon E \to \R$ defined by
507: \begin{multline*}
508: F^{\e\xi,\sigma}(u)=\frac{1}{2}\int_\Rn \bigg(\bigg|\bigg(\frac{\nabla u}{i} -
509: A(\e\xi)u \bigg)\bigg|^2 + |u|^2 + V(\e\xi)|u|^2 \bigg) \, dx \\ - \frac{1}{p+1}
510: \int_\Rn K(\e\xi)|u|^{p+1}\, dx,
511: \end{multline*}
512: whose critical point correspond to solutions of \eqref{eqlimite}.
513:
514:
515: The set
516: \[
517: Z^\e =\{{z^{\e\xi,\sigma}}\mid \xi \in \Rn \, \land \,
518: \sigma \in [0,2\pi]\} \simeq S^1 \times \Rn
519: \]
520: is a regular manifolds of critical points for the functional $F^{\e\xi,\sigma}$.
521:
522: It follows from elementary differential geometry that
523: \begin{multline*}
524: T_{z^{\e\xi,\eta}}Z^\e = \operatorname{span}_{\R}
525: \{\tfrac{\de}{\de \sigma}
526: {{{{{z^{\e\xi,\sigma}}}}}},\tfrac{\de}{\de \xi_1}
527: {{z^{\e\xi,\sigma}}},\dots,\tfrac{\de}{\de \xi_n}
528: {{{z^{\e\xi,\sigma}}}}\}=\\ =\operatorname{span}_{\R}
529: \{i{{{z^{\e\xi,\sigma}}}},\tfrac{\de}{\de \xi_1}
530: {{{z^{\e\xi,\sigma}}}},\dots,\tfrac{\de}{\de \xi_n}{{z^{\e\xi,\sigma}}}\},
531: \end{multline*}
532: where we mean by the symbol $\operatorname{span}_{\R}$ that all
533: the linear combinations must have real coefficients.
534:
535: We remark that, for $j=1,\dots,n$,
536: \begin{multline*}
537: \frac{\de}{\de \xi_j}{{z^{\e\xi,\sigma}}}=-\frac{\de}{\de x_j}{{z^{\e\xi,\sigma}}} + O(\e|\nabla
538: V(\e\xi)|)+\\
539: +i\alpha (\e\xi)e^{i A(\e\xi)\cdot x + i\sigma} U(\beta(\e\xi)(x-\xi))\left(\frac{\de}{\de
540: \xi_j}\left(A(\e\xi)\cdot x\right)
541: +A_j(\e\xi)\right)=\\
542: =-\frac{\de}{\de x_j} {{z^{\e\xi,\sigma}}} + O(\e|\nabla V(\e\xi)|+\e|J_A
543: (\e\xi)|)+i{{{z^{\e\xi,\sigma}}}}
544: A_j(\e\xi),
545: \end{multline*}
546: so that
547: \[
548: \frac{\de}{\de \xi_j}{{z^{\e\xi,\sigma}}} =
549: - \frac{\de}{\de x_j}{{z^{\e\xi,\sigma}}} +
550: i {z^{\e\xi,\sigma}}A_j(\e\xi)+ O(\e).
551: \]
552: Collecting these remarks,
553: we get that any
554: $\zeta \in T_{{{{{z^{\e\xi,\sigma}}}}}}Z^\e$ can be written as
555: $$
556: \zeta = i\ell_1 {{{z^{\e\xi,\sigma}}}}+ \sum_{j=2}^{n+1} \ell_j
557: \frac{\de}{\de x_{j-1}}{{{{z^{\e\xi,\sigma}}}}}+O(\e)$$
558: for some real coefficients $\ell_1,\ell_2,\dots,\ell_{n+1}$.
559:
560:
561:
562: The next lemma shows that $\nabla f_\e ({{{z^{\e\xi,\sigma}}}})$ gets small when
563: $\e\to 0$.
564:
565: \begin{lemma}\label{lem:1}
566: For all $\xi\in \Rn$, all $\eta \in S^1$ and all $\e>0$ small, one has that
567: $$
568: \|\nabla f_\e ({z^{\e\xi,\sigma}})\|\leq C\left(\e |\nabla
569: V(\e\xi)|+\e |\nabla K(\e\xi)|+\e |J_A (\e\xi)|+\e
570: |\operatorname{div} A(\e\xi)|+\e^{2}\right), $$
571: for some constant $C>0$.
572: \end{lemma}
573: \begin{proof}
574: From
575: \begin{multline}\label{eq:7}
576: f_{\e}(u) = F^{\e\xi,\eta}(u)+\frac{1}{2}\int_\Rn \bigg( \bigg| \frac{\nabla
577: u}{i} - A(\e x)u \bigg|^2 - \bigg| \frac{\nabla u}{i}-A(\e\xi)u
578: \bigg|^2 \biggr) +\\
579: + \frac{1}{2}\int_{\Rn}\left[V(\e x)-V(\e\xi)\right]u^{2}
580: -\frac{1}{p+1}\int_\Rn [K(\e x)-K(\e\xi)]|u|^{p+1}
581: \end{multline}
582: and since $z^{\e\xi,\eta}$ is a
583: critical point of $F^{\e\xi,\eta}$, one has (with $z=z^{\e\xi,\eta}$)
584: \begin{multline*}
585: \langle \nabla f_\e (z)\mid v \rangle = \re \int_\Rn \bigg(
586: \frac{1}{i}\nabla - A(\e\xi)\bigg)z \cdot (A(\e\xi)-A(\e x))\bar v
587: \\+ \re\int_\Rn (A(\e\xi)-A(\e x))z \cdot \overline{\bigg( \frac{1}{i}\nabla -
588: A(\e\xi)\bigg)v} + \\ \re\int_\Rn (A(\e\xi)-A(\e x))z \cdot
589: (A(\e\xi)-A(\e x))\bar v \\
590: + \re\int_\Rn (V(\e x)-V(\e\xi))z\bar v - \re\int_\Rn (K(\e
591: x)-K(\e\xi))|z|^{p-2}z\bar v \\ = \e \operatorname{Re} \int_\Rn \frac{1}{i}
592: (\operatorname{div}\, A(\e x)) z \bar v + 2 \operatorname{Re} \int_\Rn
593: (A(\e\xi)-A(\e x))z \cdot \overline{\bigg( \frac{1}{i} \nabla -
594: A(\e\xi)\bigg)v} \\+\re\int_\Rn (V(\e x)-V(\e\xi))z\bar v -
595: \re\int_\Rn (K(\e x)-K(\e\xi))|z|^{p-2}z\bar v
596: \end{multline*}
597:
598:
599: \noindent
600: From the assumption that $|D^{2}V(x)|\leq {\rm const.}$ one infers
601: \[
602: |V(\e x)-V(\e\xi)|\leq \e |\nabla V(\e\xi)|\cdot
603: |x-\xi|+c_{1}\e^{2}|x-\xi|^{2}.
604: \]
605: This implies
606: \begin{eqnarray}
607: \int_{\Rn}|V(\e x)-V(\e\xi)|^{2}{{{z^{\e\xi,\sigma}}}}^{2}
608: &\leq& c_1\e^{2}|\nabla V(\e\xi)|^{2}\int_{\Rn}|x-\xi|^{2}z^{2}(x-\xi) + \nonumber \\
609: & & c_{2}\e^{4}\int_{\Rn}|x-\xi|^{4}z^{2}(x-\xi).
610: \label{eq:1.3}
611: \end{eqnarray}
612: A direct calculation yields
613: \begin{eqnarray*}
614: \int_{\Rn}|x-\xi|^{2}z^{2}(x-\xi) & = &
615: \a^{2}(\e\xi)\int_{\Rn}|y|^{2}U^{2}(\b(\e\xi) y)dy \\
616: & = & \alpha (\e\xi)^{2} \beta (\e\xi)^{-n-2}\int_{\Rn}|y'|^{2}U^{2}(y')dy'\leq c_{3}.
617: \end{eqnarray*}
618: {From} this (and a similar calculation for the last integral in the
619: above formula) one infers
620: \begin{equation}
621: \int_{\Rn}|V(\e x)-V(\e\xi)|^{2}|z^{\e\xi,\sigma}|^{2}
622: \leq c_{4}\e^{2}|\nabla
623: V(\e\xi)|^{2} + c_{5}\e^{4}.
624: \label{eq:1.4}
625: \end{equation}
626: Of course, similar estimates hold for the terms involving $K$.
627: It then follows that
628: \[
629: \|\nabla f_\e (z^{\e\xi,\eta})\| \leq C(\e | \operatorname{div}A(\e\xi)| + \e
630: | \nabla V (\e\xi)| + \e | J_A (\e\xi)| + \e^2),
631: \]
632: and the lemma is proved.
633: \end{proof}
634:
635: \section{The invertibility of $D^2f_\e$ on $(TZ^\e)^\bot$}
636:
637: To apply the perturbative method, we need to exploit some non--degeneracy properties of
638: the solution ${z^{\e\xi,\sigma}}$ as a critical point of $F^{\e\xi,\sigma}$.
639:
640: Let $L_{\e,\sigma,\xi}\colon (T_{{{z^{\e\xi,\sigma}}}}Z^\e)^\bot
641: \longrightarrow
642: (T_{{z^{\e\xi,\sigma}}}Z)^\bot$ be the operator defined by \[\langle
643: L_{\e,\sigma,\xi} v \mid w \rangle =
644: D^2 f_\e ({{{z^{\e\xi,\sigma}}}})(v,w)\] for all
645: $v,w \in (T_{{z^{\e\xi,\sigma}}}Z^\e)^\bot$.
646:
647: The following elementary result will play a fundamental role in the present
648: section.
649:
650: \begin{lemma}
651: Let $M \subset \Rn$ be a bounded set.
652: Then there exists a constant $C>0$ such that for all $\xi\in M$ one has
653: \begin{equation}\label{norme:equiv}
654: \int_\Rn \bigg| \bigg( \frac{\nabla}{i} - A(\xi) \bigg) u \bigg|^2 + |u|^2 \geq C \int_\Rn (|\nabla
655: u|^2 + |u|^2) \qquad\forall u \in E.
656: \end{equation}
657: \end{lemma}
658:
659: \begin{proof}
660: To get a contradiction, we assume on the contrary the existence of a sequence
661: $\{ \xi_n\}$ in $M$ and a sequence $\{u_n\}$ in $E$ such that $\|u_n\|_E = 1$
662: for all $n\in\mathbb{N}$ and
663: \begin{equation}\label{eq:contradiction}
664: \lim_{n\to + \infty} \bigg[ \int_\Rn \bigg| \bigg( \frac{\n}{i} - A(\xi)
665: \bigg)u_n \bigg|^2 + \int_\Rn |u_n|^2 \bigg] = 0.
666: \end{equation}
667: In particular, $u_n \to 0$ strongly in $L^2 (\Rn,\C)$. Moreover, since $M$
668: is bounded, we can assume also $\xi_n \to \xi^* \in \overline{M}$ as
669: $n\to\infty$. From
670:
671: \begin{multline*}
672: \int_\Rn \bigg| \bigg( \frac{\n}{i} - A(\xi_n)
673: \bigg)u_n \bigg|^2 = \\
674: \int_\Rn \left( |\n u_n|^2 + |A(\xi_n)|^2 |u_n|^2 - 2 \re
675: \frac{1}{i} \n u_n \cdot A(\xi_n) \overline{u_n} \right)
676: \end{multline*}
677: we get
678: \begin{eqnarray*}
679: \lim_{n \to + \infty} \int_\Rn | \n u_n |^2 &=& 1 \\
680: \lim_{n \to + \infty} \re \int_\Rn \tfrac{1}{i} \n u_n \cdot A(\xi_n) \overline{u_n} &=& \frac{1}{2} \,.
681: \end{eqnarray*}
682: Therefore,
683: \begin{eqnarray*}
684: \limsup_{n\to\infty} \int_\Rn |\n u_n| \, |A(\xi_n)| \, |u_n| &\geq&
685: \limsup_{n\to\infty} \left| \int_\Rn \textstyle\frac{1}{i} \n u_n \cdot
686: A(\xi_n) \overline{u_n} \right| \\
687: &\geq& \limsup_{n\to\infty} \re \int_\Rn \tfrac{1}{i} \n u_n \cdot A(\xi_n)
688: \overline{u_n} = \frac{1}{2} \,.
689: \end{eqnarray*}
690: From this we conclude that
691: \begin{eqnarray*}
692: \frac{1}{2} &\leq& |A(\xi^*)| \limsup_{n\to\infty} \| \n u_n \|_{L^2} \|u_n
693: \|_{L^2} \leq \\
694: &\leq& |A(\xi^*)| \limsup_{n\to\infty} \|u_n \|_{L^2} = 0,
695: \end{eqnarray*}
696: which is clearly absurd. This completes the proof of the lemma.
697: \end{proof}
698:
699:
700: At this point we shall prove the following result:
701:
702: \begin{lemma}\label{lemma:3.1}
703: Given $\bar\xi >0$, there exists $C > 0$ such that for $\e$ small enough one
704: has
705: \begin{equation}\label{eq:posdef}
706: |\langle L_{\e,\sigma,\xi}v\mid v \rangle | \geq C \|v\|^2, \quad \forall |\xi|
707: \leq \bar\xi, \;\forall\sigma\in [0,2\pi],\;\forall v\in
708: (T_{{z^{\e\xi,\sigma}}}Z^\e)^\bot.
709: \end{equation}
710: \end{lemma}
711:
712: \begin{proof}
713: We follow the arguments in \cite{ams}, with some minor modifications due to the presence of
714: $A$. Recall that $$T_{{z^{\e\xi,\sigma}}}Z=\operatorname{span}_{\R} \{ \frac{\de}{\de
715: \xi_1}{z^{\e\xi,\sigma}},\dots ,\frac{\de}{\de
716: \xi_n}{z^{\e\xi,\sigma}},{i z^{\e\xi,\sigma}} \}.$$
717: Define
718: $$
719: \mathcal{V}=\operatorname{span}_{\R} \{ \tfrac{\de}{\de
720: x_1}{z^{\e\xi,\sigma}},\dots ,\tfrac{\de}{\de
721: x_n}{z^{\e\xi,\sigma}},{z^{\e\xi,\sigma}} ,i{z^{\e\xi,\sigma}}\}.
722: $$
723: As in \cite{ams}, it suffices to prove \eqref{eq:posdef} for all
724: $v\in\operatorname{span}_\R \{{{z^{\e\xi,\sigma}}},\phi\}$,
725: where $\phi \perp \mathcal{V}$.
726: More precisely, we shall prove that for some constants $C_1 >0$, $C_2 > 0$,
727: for all $\e$ small enough and all $|\xi|\leq \bar\xi$ the following hold:
728: \begin{equation}\label{eq:13}
729: \langle L_{\e,\sigma,\xi}{{z^{\e\xi,\sigma}}}\mid {z^{\e\xi,\sigma}} \rangle \leq
730: -C_1 < 0,
731: \end{equation}
732: \begin{equation}\label{eq:14}
733: \langle L_{\e,\sigma,\xi}\phi\mid \phi \rangle
734: \geq C_2 \|\phi\|^2 \quad\forall\phi\perp\mathcal{V}.
735: \end{equation}
736: For the reader's convenience, we reproduce here the expression for the second
737: derivative of $F^{\e\xi,\sigma}$:
738: \begin{multline*}
739: D^2 F^{\e\xi,\sigma}(u)(v,v)=\int_\Rn \bigg| \bigg(
740: \frac{\nabla}{i}-A(\e\xi)\bigg)v\bigg|^2 + |v|^2 + V(\e\xi)|v|^2 \\
741: - K(\e\xi)\bigg[(p-1) \re \int_\Rn |u|^{p-3} \re (u \bar v) u \bar v +
742: \int_\Rn |u|^{p-1}|v|^2 \bigg].
743: \end{multline*}
744: Moreover, since ${{z^{\e\xi,\sigma}}}$ is a solution of \eqref{eqlimite}, we
745: immediately get
746: \[
747: \int_\Rn
748: \bigg( \bigg| \bigg( \frac{\nabla}{i}-A(\e\xi)\bigg){{{{z^{\e\xi,\sigma}}}}}\bigg|^2
749: + V(\e\xi)|{z^{\e\xi,\sigma}}|^2 +|{{z^{\e\xi,\sigma}}}|^2
750: \bigg) = K(\e\xi)\int_\Rn
751: |{z^{\e\xi,\sigma}}|^{p+1}.
752: \]
753: From this it follows readily that we can find some $c_0 > 0$ such that for all
754: $\e > 0$ small, all $|\xi| \leq \bar \xi$ and all $\sigma\in [0,2\pi]$ it
755: results
756: \begin{equation} D^2
757: F^{\e\xi,\sigma}({{z^{\e\xi,\sigma}}}) ({z^{\e\xi,\sigma}},{z^{\e\xi,\sigma}})<c_0 <0.
758: \end{equation}
759: Recalling \eqref{eq:7}, we find
760: \begin{multline*}
761: \langle L_{\e,\sigma,\xi}{{z^{\e\xi,\sigma}}}\mid {z^{\e\xi,\sigma}} \rangle =D^2
762: F^{\e\xi,\sigma}({{z^{\e\xi,\sigma}}})({z^{\e\xi,\sigma}},{z^{\e\xi,\sigma}})+ \\ +\int_\Rn [V(\e
763: x)-V(\e\xi)]|{{z^{\e\xi,\sigma}}}|^2
764: -\int_\Rn [K(\e x)-K(\e\xi)]|{{z^{\e\xi,\sigma}}}|^{p+1}
765: +\\ \int_\Rn
766: \bigg( \bigg| \bigg(\frac{\nabla}{i}-A(\e x)\bigg){{{z^{\e\xi,\sigma}}}}\bigg|^2 - \bigg|
767: \bigg(\frac{\nabla}{i}-A(\e \xi)\bigg){{{z^{\e\xi,\sigma}}}}\bigg|^2
768: \bigg) \,.
769: \end{multline*}
770: It follows that
771: \begin{multline}
772: \langle L_{\e,\sigma,\xi}{{z^{\e\xi,\sigma}}}\mid {z^{\e\xi,\sigma}} \rangle \leq D^2
773: F^{\e\xi,\sigma}({{z^{\e\xi,\sigma}}})({z^{\e\xi,\sigma}},{z^{\e\xi,\sigma}}) + \\ +c_1
774: \e |\nabla V(\e\xi)|+c_2\e |\nabla K (\e\xi)| + c_3 \e |J_A (\e\xi)| + c_4
775: \e^2. \end{multline}
776: Hence \eqref{eq:13} follows. The proof
777: of \eqref{eq:14} is more involved. We first prove the following
778: claim.
779:
780: \vspace{5pt}
781:
782: {\em Claim.} There results \begin{equation}\label{eq:17}
783: D^2 F^{\e\xi}({z^{\e\xi,\sigma}})(\phi,\phi) \geq c_1 \|\phi\|^2 \quad
784: \forall\phi\perp\mathcal{V}.
785: \end{equation}
786:
787: \vspace{3pt}
788:
789: Recall that the complex ground state $U_c$ introduced in \eqref{eq:4} is a
790: critical point of mountain-pass type for the corresponding energy functional
791: $J\colon E \longrightarrow \R$ defined by
792: \begin{equation}
793: J(u)=\frac{1}{2}\int_\Rn ( |\nabla u|^2 + |u|^2 + V(\e\xi) |u|^2 ) -
794: \frac{1}{p+1}\int_\Rn K(\e\xi)|u|^{p+1} .
795: \end{equation}
796: Let
797: \[
798: \mathcal{M} = \left\{u\in E \colon
799: \textstyle\int_\Rn (|\n u|^2 + |u|^2 + V(\e\xi)|u|^2 )
800: =\textstyle \int_\Rn |u|^{p+1} \right\}
801: \]
802: be the Nehary manifold of $J$, which has codimension one. Let
803: \[
804: \mathcal{N}=\left\{u\in E \colon \textstyle\int_\Rn \left( \left| \left( \frac{\n}{i} - A(\e\xi)
805: \right) u\right|^2 + |u|^2 + V(\e\xi)|u|^2 \right)
806: = \textstyle\int_\Rn |u|^{p+1} \ \right\}
807: \]
808: be the Nehari manifold of $F^{\e\xi,\sigma}$.
809: One checks readily that $\operatorname{codim} \mathcal{N}=1$.
810: Recall (\cite{ku}) that $U_c$ is, up to multiplication by a constant phase,
811: the {\em unique} minimum of $J$ restricted to $\mathcal{M}$. Now, for every
812: $u\in\mathcal{M}$, the function $x \mapsto e^{i A(\e\xi)\cdot x} u(x)$ lies in
813: $\mathcal{N}$, and viceversa. Moreover
814: \[
815: J(u)=F^{\e\xi,\sigma} (e^{i A(\e\xi)\cdot x} u).
816: \]
817: This immediately implies that $\min_{\mathcal{N}} F^{\e\xi,\sigma}$ is
818: achieved at a point which differs from $e^{i A(\e\xi)\cdot x} U_c (x)$ at most
819: for a constant phase. In other words, $z^{\e\xi,\sigma}$ is a critical point
820: for $F^{\e\xi,\sigma}$ of mountain-pass type, and the claim follows by standard
821: results (see \cite{chang}).
822:
823:
824:
825: \medskip
826:
827: Let $R\gg 1$ and consider a radial smooth function
828: $\chi_{1}:\R^{n}\longrightarrow \R$ such that
829: \begin{equation}\label{eq:c1} \chi_{1}(x) = 1, \quad \hbox{ for
830: } |x| \leq R; \qquad \chi_{1}(x) = 0, \quad \hbox{ for } |x| \geq 2 R;
831: \end{equation}
832: \begin{equation}\label{eq:c2}
833: |\n \chi_{1}(x)| \leq \frac{2}{R}, \quad \hbox{ for } R \leq |x| \leq 2 R.
834: \end{equation}
835: We also set $ \chi_{2}(x)=1-\chi_{1}(x)$.
836: Given $\phi$ let us consider the functions
837: $$
838: \phi_{i}(x)=\chi_{i}(x-\xi)\phi(x),\quad i=1,2.
839: $$
840: A straightforward computation yields:
841: $$
842: \irn |\phi|^2 = \irn |\phi_1|^2 + \irn |\phi_2|^2 + 2\re\irn \phi_{1} \, \bar
843: \phi_{2},
844: $$
845: $$
846: \irn |\n \phi|^2 = \irn |\n \phi_1|^2 + \irn |\n \phi_2|^2 + 2\re\irn
847: \n\phi_{1} \cdot \overline{\n \phi_{2}},
848: $$
849: and hence
850: $$
851: \| \phi \|^2 = \| \phi_1 \|^2 + \| \phi_2 \|^2+
852: 2 \re\irn\left[ \phi_{1} \, \overline \phi_{2}
853: + \n\phi_{1} \cdot \overline{\n \phi_{2}}\right].
854: $$
855: Letting $I$ denote the last integral,
856: one immediately finds:
857: $$
858: I=\underbrace{\irn \chi_{1}\chi_{2}(\phi^{2}+|\n \phi|^{2}) }_{I_{\phi}} +
859: \underbrace{\irn\phi^{2}\n\chi_{1}\cdot \n\chi_{2} }_{I'}+
860: \underbrace{\irn (\phi_{2}\n\chi_{1}\cdot\overline{\n\phi}+\overline{\phi_{1}}\,\n
861: \phi\cdot\n\chi_{2})}_{I''}.
862: $$
863: Due to the definition of $\chi$, the two integrals $I'$ and $I''$
864: reduce to integrals from $R$ and $2R$, and thus they are $o_{R}(1)\|\phi\|^{2}$,
865: where $o_{R}(1)$ is a function which tends to $0$, as $R \to +\infty$.
866: As a consequence we have that
867: \begin{equation}\label{eq:d}
868: \| \phi \|^2 = \| \phi_1 \|^2 + \| \phi_2 \|^2 + 2I_\phi +
869: o_R(1)\| \phi \|^2 \,.
870: \end{equation}
871: After these preliminaries, let us evaluate the three terms in the
872: equation below:
873: $$
874: (L_{\e,\sigma,\xi}\phi|\phi)=
875: \underbrace{(L_{\e,\sigma,\xi}\phi_{1}|\phi_{1})}_{\a_{1}}+
876: \underbrace{(L_{\e,\sigma,\xi}\phi_{2}|\phi_{2})}_{\a_{2}}+
877: 2\underbrace{(L_{\e,\sigma,\xi}\phi_{1}|\phi_{2})}_{\a_{3}}.
878: $$
879: One has:
880: \begin{multline*}
881: \alpha_{1}=\langle L_{\e,\sigma,\xi}\phi_{1}\mid \phi_{1}
882: \rangle =D^2
883: F^{\e\xi,\sigma}({{z^{\e\xi,\sigma}}})(\phi_{1},\phi_{1}) +
884: \int_\Rn [V(\e x)-V(\e\xi)]|\phi_{1}|^2 \\ -\int_\Rn [K(\e
885: x)-K(\e\xi)]|\phi_{1}|^{p+1} + \int_\Rn \bigg|
886: \bigg( \bigg(\frac{\nabla}{i}-A(\e x)\bigg)\phi_{1}\bigg|^2 - \bigg|
887: \bigg(\frac{\nabla}{i}-A(\e \xi)\bigg)\phi_{1}\bigg|^2 \bigg) \, .
888: \end{multline*}
889:
890: In order to use \eqref{eq:17}, we introduce the function
891: $\phi^*_{1}=\phi_{1}-\psi$, where
892: $\psi$ is the projection of $\phi_{1}$ onto $\cal V$:
893: \begin{multline*}
894: \psi=(\phi_{1}|z^{\e\xi,\sigma})z^{\e\xi,\sigma} \| z^{\e\xi,\sigma} \|^{-2} +
895: (\phi_{1}|iz^{\e\xi,\sigma})iz^{\e\xi,\sigma}\|z^{\e\xi,\sigma}\|^{-2}\\
896: +\sum (\phi_{1}|\partial_{x_{i}}{{z^{\e\xi,\sigma}}})
897: \partial_{x_{i}}{z^{\e\xi,\sigma}}\| \partial_{x_{i}}{{{z^{\e\xi,\sigma}}}} \|^{-2}.
898: \end{multline*}
899: Then we have:
900: \begin{equation}
901: D^{2}F^{\e\xi}[\phi_{1},\phi_{1}]=
902: D^{2}F^{\e\xi}[\phi_{1}^*,\phi_{1}^*]+
903: D^{2}F^{\e\xi}[\psi,\psi]+2\re D^{2}F^{\e\xi}[\phi^*_{1},\psi] \,.
904: \label{eq:alfa2}
905: \end{equation}
906: Since ${z^{\e\xi,\sigma}}$ is orthogonal to $\partial_{x_{i}}{{{z^{\e\xi,\sigma}}}}$,
907: $i=1,\ldots,n$, then one
908: readily
909: checks that $\phi^*_{1}\perp {\cal V}$ and hence \eqref{eq:17}
910: implies
911: \begin{equation}
912: D^{2}F^{\e\xi}[\phi^*_{1},\phi^*_{1}]\geq c_{1}
913: \|\phi^*_{1}\|^{2}.
914: \label{eq:alfa3}
915: \end{equation}
916: On the other side, since $(\phi|{{z^{\e\xi,\sigma}}})=0$ it follows:
917: \begin{eqnarray*}
918: (\phi_{1}|{{z^{\e\xi,\sigma}}}) & = & (\phi|{z^{\e\xi,\sigma}})-(\phi_{2}|{{z^{\e\xi,\sigma}}})=
919: -(\phi_{2}|{z^{\e\xi,\sigma}}) \\
920: & = & -\re\irn\phi_{2}{{z^{\e\xi,\sigma}}}-\re\irn \n {{z^{\e\xi,\sigma}}}\cdot \n \phi_{2} \\
921: &=& -\re\irn\chi_{2}(y)z(y)\phi(y+\xi)dy-\re\irn \n z(y)\cdot \n
922: \chi_{2}(y)\phi(y+\xi)dy .
923: \end{eqnarray*}
924: Since $\chi_{2}(x)=0$ for all $|x|<R$, and since $z(x)\to 0$ as
925: $|x|=R\to\infty$, we infer
926: $(\phi_{1}|{{z^{\e\xi,\sigma}}})=o_{R}(1)\|\phi\|$. Similarly one shows that
927: $(\phi_{1}|\partial_{x}{{{{z^{\e\xi,\sigma}}}}})=o_{R}(1)\|\phi\|$ and it follows that
928: \begin{equation}\label{eq:alfa4}
929: \|\psi\|=o_{R}(1)\|\phi\|.
930: \end{equation}
931: We are now in position to estimate the last two terms in equation (\ref{eq:alfa2}).
932: Actually, using Lemma 3.1 we get
933:
934: \begin{eqnarray}\label{eq:alfa5}
935: & & D^{2}F^{\e\xi}[\psi,\psi] \geq C\|\psi\|^{2}
936: + V(\e\xi)\irn\psi^{2} \\ &- & \notag K(\e\xi)\bigg[ \re (p-1)
937: \int_\Rn |z^{\e\xi,\sigma}|^{p-3}
938: \re (z^{\e\xi,\sigma} \bar \psi) z^{\e\xi,\sigma} \bar \psi
939: \\ & & \notag
940: +\int_\Rn |z^{\e\xi,\sigma}|^{p-1}|\psi|^2 \bigg] =
941: o_{R}(1)\|\phi\|^{2}.
942: \end{eqnarray}
943: The same arguments readily imply
944: \begin{equation}\label{eq:alfa6}
945: \re D^{2}F^{\e\xi}[\phi^*_{1},\psi]=o_{R}(1)\|\phi\|^{2}.
946: \end{equation}
947: Putting together (\ref{eq:alfa3}), (\ref{eq:alfa5}) and (\ref{eq:alfa6})
948: we infer
949: \begin{equation} \label{eq:alfa7}
950: D^{2}F^{\e\xi}[\phi_{1},\phi_{1}]\geq C\|\phi_{1}\|^{2}+o_{R}(1)\|\phi\|^{2}.
951: \end{equation}
952: Using arguments already carried out before, one has
953: \begin{eqnarray*}
954: \irn|V(\e x)-V(\e\xi)|\phi_{1}^{2} & \leq
955: &\e c_{2}\irn|x-\xi|\chi_1^{2}(x-\xi)\phi^{2}(x) \\
956: & \leq & \e c_{3}\irn |y|\chi_1^{2}(y)\phi^{2}(y+\xi)dy \\
957: & \leq & \e c_{4} R \|\phi\|^{2},
958: \end{eqnarray*}
959: and similarly for the terms containing $K$.
960: This and (\ref{eq:alfa7}) yield
961: \begin{equation}
962: \a_{1}=(L_{\e,\sigma,\xi}\phi_{1}|\phi_{1})\geq c_{5}
963: \|\phi_{1}\|^{2}-
964: \e c_{4}R \|\phi\|^{2}+o_{R}(1)\|\phi\|^{2}.
965: \label{eq:alfa8}
966: \end{equation}
967: Let us now estimate $\a_{2}$. One finds
968: \begin{equation}
969: \a_{2} = \langle L_{\e,\sigma,\xi} \phi_{2} \mid \phi_{2} \rangle \geq
970: c_{6} \|\phi_{2}\|^{2}+o_{R}(1)\|\phi\|^{2}. \label{eq:alfa9}
971: \end{equation}
972: In a quite similar way one shows that
973: \begin{equation}
974: \a_{3} \geq c_{7}I_{\phi}+o_{R}(1)\|\phi\|^{2}.
975: \label{eq:alfa10}
976: \end{equation}
977: Finally, (\ref{eq:alfa8}), (\ref{eq:alfa9}), (\ref{eq:alfa10}) and the fact
978: that $I_{\phi}\geq 0$, yield
979: \begin{eqnarray*}
980: (L_{\e,\sigma,\xi}\phi|\phi) & = & \a_{1}+\a_{2}+2\a_{3}\\
981: & \geq & c_{8}\left[\|\phi_{1}\|^{2}+\|\phi_{2}\|^{2}
982: +2 I_{\phi}\right]-c_{9} R \e\|\phi\|^{2}+ o_{R}(1)\|\phi\|^{2}.
983: \end{eqnarray*}
984: Recalling (\ref{eq:d}) we infer that
985: $$
986: (L_{\e,\sigma,\xi}\phi|\phi)\geq c_{10}\|\phi\|^{2}-c_{9} R \e\|\phi\|^{2}+
987: o_{R}(1)\|\phi\|^{2}.
988: $$
989: Taking $R = \e^{-1/2}$, and choosing $\e$ small,
990: equation \eqref{eq:14} follows. This completes the proof.
991: \end{proof}
992:
993: \section{The finite dimensional reduction}
994: \label{sec:fdr}
995: In this Section we will show that the existence of critical points of $f_{\e}$
996: can be reduced to the search of critical points of an auxiliary finite
997: dimensional functional. The proof will be carried out in two
998: subsections dealing, respectively, with a Liapunov-Schmidt reduction, and
999: with the behaviour of the auxiliary finite dimensional functional.
1000:
1001: \smallskip
1002:
1003: \subsection{A Liapunov-Schmidt type reduction}\label{subsec:LS} The main
1004: result of this section is the following lemma.
1005:
1006: \begin{lemma}\label{lem:w}
1007: For $\e>0$ small, $|\xi|\leq \overline{\xi}$ and $\sigma\in [0,2\pi]$, there
1008: exists a unique $w=w(\e,\sigma,\xi)\in
1009: (T_{z^{\e\xi,\sigma}} Z^\e)^{\perp}$ such that
1010: $\nabla f_\e (z^{\e\xi,\sigma} + w)\in T_{z^{\e\xi,\sigma}} Z^\e$.
1011: Such a $w(\e,\sigma,\xi)$
1012: is of class $C^{2}$, resp. $C^{1,p-1}$, with respect to $\xi$, provided that
1013: $p\geq 2$, resp. $1<p<2$.
1014: Moreover, the functional $\Phi_\e (\sigma,\xi)=f_\e
1015: (z^{\e\xi,\sigma}+w(\e,\sigma,\xi))$ has the same regularity as $w$ and
1016: satisfies: $$
1017: \n \Phi_\e(\sigma_0,\xi_0)=0\quad \Longleftrightarrow\quad \n
1018: f_\e\left(z_{\xi_0}+w(\e,\sigma_0,\xi_0)\right)=0.
1019: $$
1020: \end{lemma}
1021: \begin{proof}
1022: Let $P=P_{\e\xi,\sigma}$ denote the projection onto $(T_{z^{\e\xi,\sigma}} Z^\e)^\perp$. We
1023: want
1024: to find a
1025: solution $w\in (T_{z^{\e\xi,\sigma}} Z)^{\perp}$ of the equation
1026: $P\nabla f_\e(z^{\e\xi,\sigma} +w)=0$. One has that $\n f_\e(z+w)=
1027: \n f_\e (z)+D^2 f_\e(z)[w]+R(z,w)$ with $\|R(z,w)\|=o(\|w\|)$, uniformly
1028: with respect to
1029: $z=z^{\e\xi,\sigma}$, for $|\xi|\leq \overline{\xi}$. Using the notation introduced
1030: in the previous section, we are led to the equation:
1031: $$
1032: L_{\e,\sigma,\xi}w + P\n f_\e (z)+PR(z,w)=0.
1033: $$
1034: According to Lemma \ref{lemma:3.1}, this is equivalent to
1035: $$
1036: w = N_{\e,\xi,\sigma}(w), \quad \mbox{where}\quad
1037: N_{\e,\xi,\sigma}(w)=-L_{\e,\sigma,\xi}^{-1}\left( P\n f_\e (z)+PR(z,w)\right).
1038: $$
1039: {From} Lemma \ref{lem:1} it follows that
1040:
1041: \begin{equation}\label{eq:N}
1042: \|N_{\e,\xi,\sigma}(w)\|\leq c_1 (\e|\n V(\e\xi)|++\e|\nabla
1043: K(\e\xi)|+\e |J_A (\e\xi)|+\e^2)+ o(\|w\|). \end{equation}
1044: Then one readily checks that $N_{\e,\xi,\sigma}$ is a contraction on some ball in
1045: $(T_{z^{\e\xi,\sigma}} Z^\e)^{\perp}$
1046: provided that $\e>0$ is small enough and $|\xi|\leq \overline{\xi}$.
1047: Then there exists a unique $w$ such that $w=N_{\e,\xi,\sigma}(w)$. Let us
1048: point out that we cannot use the Implicit Function Theorem to find
1049: $w(\e,\xi,\sigma)$, because the map
1050: $(\e,u)\mapsto P\n f_\e (u)$ fails to be
1051: $C^2$. However, fixed $\e>0$ small, we can apply the Implicit
1052: Function Theorem to the map $(\xi,\sigma,w)\mapsto P\n f_\e (z^{\e\xi,\sigma} +
1053: w)$. Then, in particular, the function $w(\e,\xi,\sigma)$ turns out to be of
1054: class $C^1$ with respect to $\xi$ and $\sigma$. Finally, it is a standard
1055: argument, see \cite{ambad, ABC}, to check that the critical points of
1056: $\Phi_\e (\xi,\sigma)=f_\e (z+w)$ give rise to critical points of $f_\e$.
1057: \end{proof}
1058:
1059: \begin{Remark}\label{remark:psi}
1060: Since $f_\e (z^{\e\xi,\sigma})$ is actually independent of $\sigma$, the implicit function $w$ is
1061: constant with respect to that variable. As a result, there exists a functional $\Psi_\e \colon \Rn
1062: \to \R$ such that
1063: \[
1064: \Phi_\e (\sigma,\xi)=\Psi_\e (\xi), \qquad \forall\sigma\in [0,2\pi], \quad \forall \xi\in\Rn.
1065: \]
1066: In the sequel, we will omit the dependence of $w$ on $\sigma$, even it is defined over
1067: $S^1\times\Rn$.
1068: \end{Remark}
1069:
1070: \begin{Remark}\label{rem:w}
1071: From (\ref{eq:N}) it immediately follows that:
1072: \begin{equation} \label{eq:w}
1073: \|w\|\leq C \left(\e |\n V(\e\xi)|+\e|\n K(\e\xi)| +\e
1074: |J_A (\e\xi)|+\e^2\right), \end{equation}
1075: where $C>0$.
1076: \end{Remark}
1077:
1078: \noindent The following result can be proved by adapting the same argument as
1079: in \cite{ams}.
1080:
1081:
1082: \begin{lemma}\label{lem:Dw}
1083: One has that:
1084: \begin{equation} \label{eq:Dw}
1085: \|\nabla_\xi w\|\leq c \left(\e |\n
1086: V(\e\xi)|+\e|\n K(\e\xi)|+\e |J_A(\e\xi)|+O(\e^2)\right)^\gamma,
1087: \end{equation}
1088: where $\gamma=\min\{1,p-1\}$ and $c > 0$ is some constant.
1089: \end{lemma}
1090:
1091: \subsection{The finite dimensional functional}\label{subs:3}
1092:
1093: The purpose of this subsection is to give an explicit form to the
1094: finite--dimensional functional $\Phi_\e (\sigma,\xi)=\Psi_\e (\xi)=f_\e
1095: (z^{\e\xi,\sigma}+w(\e,\xi))$.
1096:
1097: Recall the precise definition of $z^{\e\xi,\sigma}$ given in \eqref{def:zeta}.
1098: For brevity, we set in the sequel $z=z^{\e\xi,\sigma}$ and $w =
1099: w(\e,\xi)$.
1100:
1101: Since $z$ satisfies \eqref{eqlimite}, we easily find the following relations:
1102: \begin{equation}
1103: \int_\Rn \bigg|\bigg( \frac{\n}{i} - A(\e\xi)\bigg)z \bigg|^2 + |z|^2 +
1104: V(\e\xi)|z|^2=\int_\Rn K(\e\xi)|z|^{p+1}
1105: \end{equation}
1106: \begin{multline}
1107: \re\int_\Rn \bigg( \frac{\n}{i} - A(\e\xi)\bigg)z \cdot \overline{\bigg(
1108: \frac{\n}{i} - A(\e\xi)\bigg)w} + \re\int_\Rn z\bar w \\
1109: +\re\int_\Rn V(\e\xi)z \bar w = \re\int_\Rn K(\e\xi) |z|^{p-1}z\bar w \,.
1110: \end{multline}
1111:
1112: Hence we get
1113: \begin{multline}\label{eq:Phi}
1114: \Phi_\e (\sigma,\xi)=f_\e (z^{\e\xi,\sigma}+w(\e,\sigma,\xi))
1115: =\\ =K(\e \xi) \left( \frac{1}{2}-\frac{1}{p+1}\right) \int_\Rn
1116: |z|^{p+1}
1117: + \frac{1}{2} \int_\Rn |A(\e\xi)-A(\e x)|^2 z^2\,
1118: +\\ \re\int_\Rn (A(\e\xi) -A(\e x))z \cdot (A(\e\xi)-A(\e x))\bar w
1119: +\e\re\int_\Rn \frac{1}{i} z \bar w \, \operatorname{div}A(\e x)
1120: \\ + \frac{1}{2}\int_\Rn \bigg| \bigg(\frac{\n}{i} - A(\e x)\bigg)w \bigg|^2
1121: +\re\int_\Rn [V(\e x)-V(\e\xi)]z \bar w
1122: \\ + \frac{1}{2}\int_\Rn [V(\e x)-V(\e\xi)] |w|^2
1123: + \frac{1}{2} \int_\Rn [V(\e x)-V(\e\xi)] z^2
1124: \\ +\frac{1}{2} V(\e\xi) \int_\Rn |w|^2\,
1125: - \frac{1}{p+1} \re \int_\Rn K(\e x)(|z+w|^{p+1} -
1126: |z|^{p+1} - (p+1) |z|^{p-1} z \bar w )\,
1127: \\ + \re K(\e\xi)\int_\Rn |z|^{p-1}z\bar w
1128: +O(\e^2) \,.
1129: \end{multline}
1130: Here we have used the estimate
1131: \[
1132: \int_\Rn \bigg(\frac{1}{2} K(\e x)-\frac{1}{p+1} K(\e\xi) \bigg) |z|^{p+1} \,
1133: = \bigg( \frac{1}{2} - \frac{1}{p+1} \bigg) \int_\Rn K(\e\xi) |z|^{p+1} \,
1134: + O(\e^2),
1135: \]
1136: which follows from the boundedness of $K''$. Since we know that
1137: \begin{eqnarray*}
1138: \alpha (\e\xi)&=&\left( \frac{1+V(\e\xi)}{K(\e\xi)}\right)^{\frac{1}{p-1}}\\
1139: \beta (\e\xi)&=&\left( 1+V(\e\xi)\right)^\frac{1}{2},
1140: \end{eqnarray*}
1141: we get immediately
1142: \begin{equation}
1143: \int_\Rn |z^{\e\xi,\sigma}|^{p+1} = C_0 \, \Lambda (\e\xi)
1144: [K(\e\xi)]^{-1},
1145: \end{equation}
1146: where we define the auxiliary function
1147: \begin{equation}\label{def:auxiliary}
1148: \Lambda (x)=\frac{(1+V(x))^\theta}{K(x)^{-2/(p-1)}},\qquad \theta =
1149: \frac{p+1}{p-1}-\frac{n}{2},
1150: \end{equation}
1151: and $C_0 = \|U\|_{L^2}$.
1152: Now one can estimate the various terms in \eqref{eq:Phi} by means of
1153: \eqref{eq:w} and \eqref{eq:Dw}, to prove that
1154: \begin{equation}\label{eq:expansion}
1155: \Phi_\e (\sigma,\xi) = \Psi_\e (\xi)= C_1 \, \Lambda (\e\xi) + O(\e).
1156: \end{equation}
1157: Similarly,
1158: \begin{equation}\label{eq:expansionD}
1159: \nabla \Psi_\e (\xi )=C_1 \n \Lambda (\e\xi) + \e^{1+\gamma} O(1),
1160: \end{equation}
1161: where $C_1 = \left(
1162: \frac{1}{2}-\frac{1}{p+1} \right)C_0$. We omit the details, which can be
1163: deduced without effort from \cite{ams}.
1164:
1165: \section{Statement and proof of the main results}
1166:
1167: \bigskip
1168:
1169: In this section we exploit the finite-dimensional reduction
1170: performed in the previous section to find existence and multiple
1171: solutions of (NLS).
1172: Recalling Lemma~\ref{lem:w}, we have to look for critical points
1173: of $\Phi_\e$ as a function of the variables
1174: $(\sigma,\xi)\in [0,2\pi ]\times \Rn$
1175: (or, equivalently, $(\eta,\xi)\in S^1 \times \Rn$).
1176:
1177: In what follows, we use the
1178: following notation: given a set $\Omega \subset \Rn$ and a number $\rho > 0$,
1179: \[
1180: \Omega_\rho \overset{\rm def}{=} \{ x\in\Rn \mid \operatorname{dist}(x,\Omega) < \rho \}.
1181: \]
1182:
1183: We start with the following result, which deals with local extrema.
1184:
1185: \begin{theorem}\label{pippo}
1186: Suppose that (K1), (V1) and (A1) hold. Assume moreover that there is
1187: a compact set $M \subset \Rn$ over which $\Lambda$ achieves an isolated strict local
1188: minimum with value $a$. By this we mean that for some $\delta > 0$,
1189: \begin{equation}\label{eq:slm}
1190: b\; \overset{\rm def}{=} \inf_{x\in \de M_\delta} \Lambda (x) > a.
1191: \end{equation}
1192: Then there exists $\e_\delta > 0$ such that $(S_\e)$ has at least $\operatorname{cat}
1193: (M,M_\delta)$ (orbits of) solutions concentrating near $M_\delta$, for all $0 < \e < \e_\delta$.
1194:
1195: Conversely, assume that $K$ is a compact set $\Rn$ over which $\Lambda$
1196: achieves an isolated strict local maximum with value $b$, namely
1197: for some $\gamma > 0$,
1198: \begin{equation}\label{eq:max}
1199: a \; \overset{\rm def}{=} \inf_{x\in \de K_\gamma} \Lambda (x) < b.
1200: \end{equation}
1201: Then there exists $\e_\gamma > 0$ such that $(S_\e)$ has at
1202: least $\operatorname{cat} (K,K_\gamma)$ (orbits of)
1203: solutions concentrating near $K_\gamma$, for all $0 < \e < \e_\gamma$.
1204:
1205: \end{theorem}
1206:
1207: \begin{proof}
1208: As in the previous theorem, one has $\Phi_\e (\eta,\xi)=\Psi_\e (\xi)$.
1209: Now choose $\bar\xi > 0$ in such a way that $M_\delta \subset \{x\in\Rn \mid \ |x| < \bar\xi\}$.
1210: Define again $\overline{\Lambda}$ as in the proof of Theorem \ref{th:5.3}. Let
1211: \begin{eqnarray*}
1212: N^\e &=& \{\xi\in \Rn \mid \e\xi\in M\} \\
1213: N_\delta^\e &=& \{ \xi\in \Rn \mid \e\xi\in M_\delta\}\\
1214: \Theta^\e &=& \{ \xi\in \Rn \mid \Psi_\e (\xi) \leq C_1 \tfrac{a+b}{2}\}.
1215: \end{eqnarray*}
1216: From \eqref{eq:expansion} we get some $\e_\delta > 0$ such that
1217: \begin{equation}\label{eq:inclusioni}
1218: N^\e \subset \Theta^\e \subset N_\delta^\e ,
1219: \end{equation}
1220: for all $0 < \e < \e_\delta$. To apply standard category theory, we need to prove that
1221: $\Theta^\e$ is compact. To this end, as can be
1222: readily checked, it suffices to prove that $\Theta^\e$ cannot touch $\de N_\delta^\e$. But if
1223: $\e\xi\in\de M$, one has $\overline{\Lambda} (\e\xi) \geq b$ by the very definition of $\delta$,
1224: and so
1225: \[
1226: \Psi_\e (\xi) \geq C_1 \overline{\Lambda} (\e\xi) + o_\e (1) \geq C_1 b + o_\e (1).
1227: \]
1228: On the other hand, for all $\xi\in \Theta^\e$ one has also $\Psi_\e (\xi) \leq C_1 \frac{a+b}{2}$.
1229: We can conclude from \eqref{eq:inclusioni} and elementary properties of the
1230: Lusternik--Schnirel'man category that $\Psi_e$ has at least
1231: \[
1232: \operatorname{cat} (\Theta^\e,\Theta^\e) \geq \operatorname{cat} (N^\e , N_\delta^\e) =
1233: \operatorname{cat} (N,N_\delta)
1234: \]
1235: critical points in $\Theta^\e$, which correspond to at least $\operatorname{cat}(M,M_\delta)$
1236: orbits of solutions to $(S_\e)$.
1237: Now, let $(\eta^*,\xi^*)\in S^1 \times M_\delta$ a critical point of $\Phi_\e$. Hence this point
1238: $(\eta^*,\xi^*)$ localizes a solution $u_{\e,\eta^*,\xi^*} (x)=z^{\e\xi^*,\eta^*} (x) +
1239: w(\e,\eta^*,\xi^*)$ of $(S_\e)$. Recalling the change of variable which allowed us to pass
1240: from (NLS) to $(S_\e)$, we find that
1241: \[
1242: u_{\e,\eta^*,\xi^*} (x) \approx z^{\e\xi^*,\eta^*} (\tfrac{x-\xi^*}{\e}).
1243: \]
1244: solves (NLS). The concentration statement follows from standard arguments (\cite{ABC,ams}).
1245: The Proof of the second part of Theorem \ref{th:5.3} follows with analogous
1246: arguments.
1247: \end{proof}
1248:
1249: Theorem 1.1 in the Introduction is an immediate corollary of the previous one when $x_0$ is
1250: either a nondegenerate local maximum or minimum for $\Lambda$.
1251: We remark that the case in which $\Lambda$ has a maximum cannot be handled
1252: using a direct variational approach and the arguments in \cite{cingolani}
1253: cannot be applied.
1254:
1255: To treat the general case, we need some more work.
1256: In order to present our main result, we need to introduce
1257: some topological concepts.
1258:
1259: \smallskip
1260: Given a set $M \subset \Rn$, the {\em cup long} of $M$ is by definition
1261: \[
1262: \ell (M)=1+\sup \{k\in\mathbb{N} \mid (\exists \alpha_1,\dots,\alpha_n \in \check{H}^{*}
1263: (M)\setminus \{1\})(\alpha_1 \cup \dots \cup \alpha_k \neq 0)\}.
1264: \]
1265: If no such classes exists, we set $\ell (M)=1$. Here $\check{H}^{*}(M)$ is the Alexander
1266: cohomology of $M$ with real coefficients, and $\cup$ denotes the cup product. It is well known
1267: that $\ell (S^{n-1})=\cat (S^{n-1})=2$, and $\ell (T^n)=\cat (T^n) = n+1$, where $T^n$ is the
1268: standard $n$--dimensional torus. But in general, one has $\ell (M) \leq \cat (M)$.
1269:
1270: The following definition dates back to Bott (\cite{bott}).
1271:
1272: \begin{definition}
1273: We say that $M$ is non-degenerate for a $C^2$ function
1274: $I\colon \R^N \to \R$ if $M$ consists of Morse theoretically
1275: non-degenerate critical points for the restriction $I_{|M^\bot}$.
1276: \end{definition}
1277:
1278: To prove our existence result, we need the next theorem, which is a slightly modified statement
1279: of
1280: Theorem 6.4 in chapter II of \cite{chang}.
1281:
1282: \begin{theorem}\label{th:chang}
1283: Let $I\in C^1 (V)$ and $J\in C^2 (V)$ be two functionals defined on
1284: the Riemannian manifold $V$,
1285: and let $\Sigma \subset V$ be a smooth, compact, non-degenerate manifold of critical
1286: points of $J$. Denote by $\mathcal{U}$ a neighborhood of $\Sigma$.
1287:
1288: If $\| I-J\|_{C^1 (U)}$ is small enough, then the functional $I$ has at least $\ell (\Sigma)$ critical
1289: points contained in $\mathcal{U}$.
1290: \end{theorem}
1291:
1292: We only remark that Theorem \ref{th:chang} can also be proved in the framework of Conley
1293: theory (\cite{conley}).
1294:
1295: We are now ready to prove an existence and multiplicity result for (NLS).
1296:
1297: \begin{theorem}\label{th:5.3}
1298: Let (V1), (K1) and (A1) hold. If the auxiliary function $\Lambda$ has
1299: a smooth, compact, non-degenerate manifold of critical points $M$,
1300: then for $\e > 0$ small, the problem $(S_\e)$ has at least $\ell (M)$ (orbits of) solutions
1301: concentrating near points of $M$.
1302: \end{theorem}
1303: \begin{proof}
1304: By Remark \ref{remark:psi}, we have to find critical points of
1305: $\Psi_\e=\Psi_\e (\xi)$. Since $M$ is compact, we can choose $\bar\xi > 0$ so
1306: that $|x| < \bar\xi$ for all points $x\in M$. From this moment, $\bar\xi$ is
1307: kept fixed. %Moreover, by the non-degeneracy assumption on %$M$, any set of the
1308: form $\{\eta^*\} \times M$ is obviously a non-degenerate critical manifold
1309: We set now $V=\Rn$, $J=\Lambda$, $\Sigma = M$, and $I(\xi)=\Psi_\e (\eta,\xi /
1310: \e)$. Select $\delta > 0$ so that $M_\delta \subset \{x \colon |x| < \bar\xi
1311: \}$, and no critical points of $\Lambda$ are in $M_\delta$, except fot those
1312: of $M$. Set $\mathcal{U}=M_\delta$. From \eqref{eq:expansion} and
1313: \eqref{eq:expansionD} it follows that $I$ is close to $J$ in $C^1
1314: (\overline{\mathcal{U}})$ when $\e$ is very small. We can apply Theorem
1315: \ref{th:chang} to find at least $\ell (M)$ critical points
1316: $\{\xi_1,\dots,\xi_{\ell (M)}\}$ for $\Psi_\e$, provided $\e$ is small enough.
1317: Hence the orbits $S^1\times\{\xi_1\}, \dots, S^1\times\{\xi_{\ell (M)}\}$
1318: consist of critical points for $\Phi_\e$ which produce solutions of $(S_\e)$.
1319: The concentration statement follows as in \cite{ams}. \end{proof}
1320:
1321:
1322: \noindent
1323: \begin{Remark}\label{nonl}
1324: We point out that
1325: Theorem 1.1, Theorem \ref{pippo}
1326: and Theorem \ref{th:5.3} hold for problems involving
1327: more general nonlinearities $g(x,u)$
1328: satisfying the same assumptions in \cite{grossi} (see also Remark 5.4
1329: in \cite{ams}).
1330: For our approach, we need the uniqueness of the radial solution
1331: $z$ of the corresponding scalar equation
1332: \begin{equation}\label{eta}
1333: - \varDelta u + u + V(\e \xi)u=g(\e\xi,u), \quad u>0, \ \ u \in W^{1,2}(\Rn) \,.
1334: \end{equation}
1335: Let us also remark that in \cite{cingolani} the class
1336: of nonlinearities handled does not require that
1337: equation \eqref{eta} has a unique solution.
1338: \end{Remark}
1339:
1340: \section*{Acknowledgements}
1341: The authors would like to thank Prof.~Ambrosetti for
1342: several comments and suggestions.
1343:
1344:
1345:
1346: \begin{thebibliography}{99}
1347:
1348: \bibitem{ambad} A.~Ambrosetti, M.~Badiale, {\it Variational perturbative
1349: methods and bifurcation of bound states from the essential spectrum}, Proc.
1350: Royal Soc. Edinburgh, {\bf 128 A}, (1998), 1131--1161.
1351:
1352: \bibitem{ABC} A.~Ambrosetti, M.~Badiale, S.~Cingolani, {\it Semiclassical
1353: states of nonlinear {S}chr\"{o}dinger equations}, Arch. Rational Mech. Anal.
1354: {\bf 140}, (1997), 285--300.
1355:
1356:
1357: \bibitem{ams} A.~Ambrosetti, A.~Malchiodi, S.~Secchi, {\it Multiplicity
1358: results for some nonlinear {S}chr\"{o}dinger equations with potentials},
1359: Arch. Rat. Mech. Anal. {\bf 159}, (2001), 253--271.
1360:
1361: \bibitem{amberti} A.~Ambrosetti, M.~Berti,
1362: {\it Homoclinics and complex dynamics in slowly oscillating systems},
1363: Discr.Cont. Dyn. Systems, {\bf 4-3}, (1998), 285--300.
1364:
1365:
1366: \bibitem{bott} R.~Bott, {\it Nondegenerate critical manifolds}, Annals of
1367: Math. {\bf 60-2}, (1954), 248--261.
1368:
1369: \bibitem{brum} R.~Brummelhuis, {\it Exponential decay in the
1370: semiclassical limit for eigenfunctions of {S}chr\"{o}dinger operators
1371: with magnetic fields and potentials which degenerate at infinity},
1372: Comm. Part. Diff. Eqs. {\bf 16}, (1991), 1489--1502.
1373:
1374: \bibitem{chang} K.~C.~Chang, {\it Infinite dimensional Morse theory and
1375: multiple solution problems}, Birkh\"{a}user, 1993.
1376:
1377: \bibitem{cingolani} S.~Cingolani, {\it Semiclassical stationary states of
1378: nonlinear {S}chr\"{o}dinger equations with an external magnetic field},
1379: preprint, Rapporto del Dip. Inter. di Matematica di Bari, n. 51/00.
1380:
1381: \bibitem{CL1} S.~Cingolani, M.~Lazzo, {\it Multiple semiclassical standing
1382: waves for a class of nonlinear {S}chr\"{o}dinger equations}, Top. Meth.
1383: Nonlin. Anal. {\bf 10} (1997), 1--13.
1384:
1385: \bibitem{CL2} S.~Cingolani, M.~Lazzo, {\it Multiple positive solutions to
1386: nonlinear {S}chr\"{o}dinger equations with competing potential functions}, J.
1387: Diff. Equations, {\bf 160}, (2000), 118--138.
1388:
1389: \bibitem{CN} S.~Cingolani, M.~Nolasco, {\it Multi-peaks periodic semiclassical
1390: states for a class of nonlinear {S}chr\"{o}dinger equations}, Proc. Royal Soc.
1391: Edinburgh, {\bf 128}, (1998), 1249--1260.
1392:
1393: \bibitem{conley} C.~C.~Conley, {\it Isolated invariant sets and the Morse index}, CBMS
1394: Regional Conf. Series in Mathematics, {\bf 38}, AMS (1978).
1395:
1396: \bibitem{dpf} M.~Del~Pino, P.~Felmer, {\it Local mountain passes for semilinear elliptic problems
1397: in unbounded domains}, Calc. Var. PDE, {\bf 4}, (1996), 121--137.
1398:
1399: \bibitem{dpf1} M.~Del~Pino, P.~Felmer, {\it
1400: Multi-peak bound states for nonlinear {S}chr\"{o}dinger equations},
1401: Ann. Inst. Henri Poincar\'e {\bf 15}, (1998), 127--149.
1402:
1403:
1404: \bibitem{EL} M.Esteban, P.~L.~Lions, {\it Stationary solutions of nonlinear
1405: {S}chr\"{o}dinger equations with an external magnetic field}, in PDE and
1406: Calculus of Variations, in honor of E.~De~Giorgi, Birkh\"{a}user, 1990.
1407:
1408: \bibitem{grossi} M.~Grossi, {\it
1409: Some results on a class of nonlinear {S}chr\"{o}dinger equations},
1410: Math. Zeit. (to appear).
1411:
1412: \bibitem{gui} C.~Gui, {\it Existence of multi-bump solutions for nonlinear
1413: {S}chr\"{o}dinger equations}, Comm. Partial Diff. Eq. {\bf 21}, (1996),
1414: 787--820.
1415:
1416: \bibitem{hel1} B.~Helffer, {\it On Spectral Theory for Schr\"odinger
1417: Operators with Magnetic Potentials},
1418: Advanced Studies in Pure Mathematics vol. 23, 113--141 (1994).
1419:
1420: \bibitem{RS} B.~Helffer,
1421: {\it Semiclassical analysis for Schr\"odinger operator with magnetic wells},
1422: in Quasiclassical methods (J.~Rauch, B.~Simon Eds.).
1423: The IMA Volumes in Mathematics and its applications vol. 95, Springer--Verlag
1424: New--York 1997.
1425:
1426: \bibitem{helS} B.~Helffer, J.~Sj\"ostrand, {\it Effet tunnel pour l'\'equation de Schr\"odinger
1427: avec champ magn\'etique}, Annali Scuola Normale Superiore di Pisa, {\bf 14} (1987),
1428: 625--657.
1429:
1430:
1431: \bibitem{ku} K.~Kurata, {\it Existence and semi-classical limit of the least
1432: energy solution to a nonlinear {S}chr\"{o}dinger equation with electromagnetic
1433: fields}, Nonlinear Anal. {\bf 41}, (2000), 763--778.
1434:
1435:
1436: \bibitem{Li} Y.~Y.~Li, {\it On a singularly perturbed elliptic equation}, Adv.
1437: Diff. Equat. {\bf 2}, (1997), 955--980.
1438:
1439: \bibitem{oh} Y.~G.~Oh, {\it Existence of semiclassical bound states of
1440: nonlinear {S}chr\"{o}dinger equations}, Comm. Partial Diff. Eq. {\bf 13},
1441: (1988), 1499--1519.
1442:
1443: \bibitem{xw} X.~Wang, B.~Zeng, {\it On concentration of
1444: positive bound states of nonlinear {S}chr\"{o}dinger
1445: equation with competing potential functions}, SIAM J. Math. Anal. {\bf 28},
1446: (1997), 633--655.
1447:
1448: \end{thebibliography}
1449:
1450: \end{article}
1451: \end{document}
1452:
1453:
1454:
1455:
1456: