Sharp bound on the largest positive eigenvalue for one-dimensional Schr\"odinger operators
Extracted source text ยท arXiv 1709.05611
\documentclass[10pt]{amsart}
\usepackage{amssymb,amsthm,amsmath}
\usepackage[numbers,sort&compress]{natbib}
\usepackage{color}
\usepackage{graphicx}
\usepackage{tikz}
%\linespread{2}
\hoffset -3.5pc
\renewcommand{\textwidth}{34pc}
\title[ ]{Sharp bound on the largest positive eigenvalue for one-dimensional Schr\"odinger operators}
%\author{ Svetlana Jitomirskaya}
%\address[ Svetlana Jitomirskaya]{ Department of Mathematics, University of California, Irvine, California 92697-3875, USA}
%\address{Current address: Department of Mathematics, University of California, Irvine, California 92697-3875, USA}
%\email{szhitomi@math.uci.edu}
\author{Wencai Liu}
\address[Wencai Liu]{Department of Mathematics, University of California, Irvine, California 92697-3875, USA} \email{liuwencai1226@gmail.com}
%\address{Current address: Department of Mathematics, University of California, Irvine, California 92697-3875, USA}
%\subjclass{Primary ; Secondary , }
%\keywords{Klein-Gordon-Maxwell system, variational methods, critical point theorem.\\
%$^\dag$ Corresponding author. Tel:+862368253135. Email:tangcl@swu.edu.cn.}
%\thanks{{\em 2010 Mathematics Subject Classification.} Primary: 35J61. Secondary:35C06, 35J20.}
%\thanks{This work is supported by National Natural Science Foundation of China (No. 11071198), Fundamental Funds for the Central Universities (No. XDJK2013D007) and Fund of China Scholarship Council.}
\newcommand{\dint}[4]{\displaystyle\int_{#1}^{#2} \! #3 \, \operatorname{d} \! #4}
\newcommand{\dd}{\,\mathrm{d}}
\newcommand{\norm}[1]{\left\| #1 \right\|}
\newcommand{\abs}[1]{\left\lvert #1 \right\rvert}
%\newcommand{\RR}{\mathbb{R}}
\DeclareMathOperator{\essinf}{essinf}
\DeclareMathOperator{\meas}{meas}
\DeclareMathOperator{\divv}{div}
\DeclareMathOperator{\spann}{span}
\newcommand{\R}{\mathbb{R}}
\newcommand{\Z}{\mathbb{Z}}
\newcommand{\Q}{\mathbb{Q}}
\newcommand{\T}{\mathbb{T}}
\newcommand{\C}{\mathbb{C}}
\theoremstyle{plain}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem{lemmaa}[theorem]{Lemma A.}
\theoremstyle{definition}
\newtheorem{definition}[theorem]{Definition}
\newtheorem{conjecture}[theorem]{Conjecture}
\newtheorem{example}[theorem]{Example}
\newtheorem{remark}[theorem]{Remark}
\newtheorem*{note}{Note}
\newtheorem{case}{Case}[section]
\begin{document}
%%% ----------------------------------------------------------------------
\begin{abstract}
Let $H=-D^2+V$ be a Schr\"odinger operator on $ L^2(\mathbb{R})$, or on $ L^2(0,\infty)$.
Suppose the potential satisfies $\limsup_{x\to \infty}|xV(x)|=a<\infty$. We prove that $H$ admits no eigenvalue larger than
$ \frac{4a^2}{\pi^2}$. For any positive $a$ and $\lambda$ with $0<\lambda< \frac{4a^2}{\pi^2}$, we construct potentials $V$
such that $\limsup_{x\to \infty}|xV(x)|=a $ and the associated Sch\"rodinger operator $H=-D^2+V$ has eigenvalue $\lambda$.
\end{abstract}
%%% ----------------------------------------------------------------------
\maketitle
%%% ----------------------------------------------------------------------
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%INTRODUCTION%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Introduction and main results}
Let $H=-D^2+V$ be a one-dimensional
Sch\"rodinger operator on $ L^2(\mathbb{R})$, or on $ L^2(0,\infty)$, where the potential $V$ satisfies
\begin{equation*}
\limsup_{x}|xV(x)|=a<\infty.
\end{equation*}
By the classical Weyl theorem, the essential spectrum $\sigma_{\rm ess}(H)=[0,\infty)$.
In this paper, we are interested in quantitative relation of the largest positive eigenvalue and $a$.
It is well known that there is no eigenvalue larger than $a^2$, see paper \cite{kato} for example.
This implies there is no eigenvalue embedded into the essential spectrum if $V(x)=o(x^{-1})$ as $x$ goes to infinity.
As a standard example with Wigner-von Neumann type potentials shows, for any $0<\lambda<\frac{a^2}{4}$, there exists $V$ with $\limsup_{x}|xV(x)|=a$ such that $\lambda$ is an eigenvalue.
Thus $V(x)=o(x^{-1})$ is the spectral transition for existence of eigenvalue embedded into the essential spectrum.
Naboko \cite{naboko} and Simon \cite{simon} constructed examples to show that dense point spectra can be embedded into essential spectrum if
the potential decreases slightly slower than $V(x)=O(x^{-1})$.
The fact that $V(x)=\frac{O(1)}{1+x}$ is the spectral transition for singular continuous embedding into the essential spectrum was established by Kiselev \cite{ki1}.
Kiselev-Last-Simon \cite{ki2} proved that
if $\limsup_{x}|xV(x)|=a$, then the sum of eigenvalues $\lambda_n$ of $-D^2+V$ is finite, that is
$ \sum \lambda_n\leq \frac{a^2}{2}$.
In this paper, we focus on the largest eigenvalue and obtain the following sharp result.
\begin{theorem}\label{thm1}
%Let $H=-D^2+V$ be a Schr\"odinger operator.
Suppose potential $V$ satisfies
\begin{equation*}
\limsup_{x\to \infty}|xV(x)|=a<\infty.
\end{equation*}
Then $H$ admits no eigenvalue larger than
$ \frac{4 a^2}{\pi^2}$.
\end{theorem}
\begin{theorem}\label{thm2}
For any positive $a$ and $\lambda$ with $0<\lambda< \frac{4a^2}{\pi^2}$, there exist potentials $V$
such that $\limsup_{x\to \infty}|xV(x)|=a $ and the associated Schr\"odinger operator $H=-D^2+V$ has eigenvalue $\lambda$.
\end{theorem}
Our proof is based on the modified Pr\"ufer transformation and some basic analysis. See \cite{ki2} for more details about modified Pr\"ufer transformation.
\begin{remark}\footnote{I would like to thank Barry Simon for telling me the full history of the problem.}
After we finished this paper, we noticed that Theorem 1.2 has been proved by Halvorsen \cite{Hal}. See \cite{atk,eas,ani} for more generalization.
Later, Remling also addressed the problem \cite{remling1998absolutely} and showed that $H$ has purely absolute continuous spectrum in $ (\frac{4 a^2}{\pi^2},\infty)$.
We refer readers to Simon's article for full history \cite{simon2017tosio}.
\end{remark}
\section{Proof of Theorems \ref{thm1} and \ref{thm2} }
Without loss of generality, we only consider the half line case $L^2(0,\infty)$.
Let $\lambda=k^2$ with $k>0$ and
suppose $ u $ is a solution of
\begin{equation}\label{Gei}
-u^{\prime\prime}(x) +V(x)u(x)= k^2 u(x).
\end{equation}
Change variables to
\begin{eqnarray*}
% \nonumber to remove numbering (before each equation)
u^{\prime}(x) &=& kR(x)\cos(\theta(x)) \\
u(x) &=& R(x)\sin(\theta(x)).
\end{eqnarray*}
We get a pair of equations
\begin{eqnarray}
% \nonumber to remove numbering (before each equation)
\frac{d\theta}{dx} &=& k-\frac{V(x)}{k}\sin^2\theta \label{Gtheta}\\
\frac{d\log R}{dx}&=& \frac{1}{2k} V(x)\sin2\theta.\label{GR}
\end{eqnarray}
If $V=0$, then $\theta(x)=\theta_0+kx$ and $R(x)=R_0$ is a solution.
The following lemma is well known. The proof is basic, the readers can see Lemma 4.2 in \cite{ki2} for the details.
\begin{lemma}[Lemma 4.2, \cite{ki2}]\label{Lem1}
If $u\in L^2(0,\infty)$ is a solution of \eqref{Gei}, then
\begin{equation*}
R(\cdot)\in L^2(0,\infty).
\end{equation*}
\end{lemma}
{\textbf{Proof of Theorem \ref{thm1}}}
\begin{proof}
Under the assumption of Theorem \ref{thm1}, for any $\epsilon>0$,
there exists $x_0>0$ such that for all $x>x_0$
\begin{equation*}
| V(x)|\leq \frac{a+\epsilon}{1+x}.
\end{equation*}
In the following, we always assume $\epsilon>0$ is sufficiently small and may change even in the same formula.
We also assume $x>0$ is large enough.
By \eqref{GR}, one has
\begin{equation}\label{e4}
\log R(x)\geq \log R(x_0)- \frac{a+\epsilon}{2k}\int_{x_0}^x\frac{|\sin2\theta(y)|}{y+1}dy.
\end{equation}
Now we will estimate $\int_{x_0}^x\frac{|\sin2\theta(y)|}{y+1}dy$.
Let $i_0$ be the largest positive integer such that $2\pi i_0<\theta(x_0)$.
By \eqref{Gtheta}, there exist
$x_0<x_1<x_2<\cdots<x_n<x<x_{n+1}$ such that
\begin{equation*}
\theta(x_i)= 2\pi i_0+i\frac{\pi}{2}
\end{equation*}
for $i=1,2,\cdots,n,n+1$.
By \eqref{Gtheta}, one has
\begin{equation*}
|x_{i+1}-x_{i}|=\frac{\pi}{2k}+ \frac{O(1)}{x_i+1}.
\end{equation*}
Then, one has
\begin{equation}\label{e2}
n\leq (\frac{2k}{\pi}+\epsilon)x
\end{equation}
and
\begin{equation}\label{e3}
(\frac{\pi}{2k}-\epsilon)i<x_i-x_0<(\frac{\pi}{2k}+\epsilon)i.
\end{equation}
For $y\in[x_i,x_{i+1})$, we have
\begin{eqnarray*}
% \nonumber to remove numbering (before each equation)
\theta(y) &=&2\pi i_0+i\frac{\pi}{2}+k(y-x_i)+\frac{O(1)}{1+x_i} \\
&=& 2\pi i_0+i\frac{\pi}{2}+k(y-x_i)+\frac{O(1)}{1+i}.
\end{eqnarray*}
Which implies
\begin{eqnarray}
% \nonumber to remove numbering (before each equation)
\nonumber \int_{x_i}^{x_{i+1}}|\sin2\theta(y)|dy &=& \int_{0}^{\frac{\pi}{2k}}\sin( 2ky )dy+\frac{O(1)}{1+i} \\
&=&\frac{1}{k}+\frac{O(1)}{1+i}. \label{e1}
\end{eqnarray}
By \eqref{e2}, \eqref{e3} and \eqref{e1}, we obtain
\begin{eqnarray*}
% \nonumber to remove numbering (before each equation)
\int_{x_0}^x\frac{|\sin2\theta(y)|}{y+1}dy&\leq& \sum_{i=1}^{n+1}(\frac{1}{k}+\frac{O(1)}{1+i})\frac{2k}{(\pi-\epsilon)i}+O(1)\\
&\leq&( \frac{2}{\pi}+\epsilon)\ln x+O(1).
\end{eqnarray*}
Combining with \eqref{e4}, we have
\begin{equation*}
\log R(x)\geq -(\frac{a}{k\pi}+\epsilon)\ln x +O(1).
\end{equation*}
Assume $k>\frac{2a}{\pi}$, one has
\begin{equation*}
\frac{a}{k\pi}+\epsilon<\frac{1}{2}.
\end{equation*}
Thus
\begin{equation*}
R(x)\geq \frac{1}{\sqrt{x}}
\end{equation*}
for large $x$. By Lemma \ref{Lem1}, $u\notin L^2 (0,\infty)$. Thus $\lambda=k^2$ is not an eigenvalue.
%We also assume there is no $x_n<x_{n+1}
\end{proof}
{\textbf{Proof of Theorem \ref{thm2} }}
\begin{proof}
We define
\begin{equation}\label{Def}
V(x)=-\frac{a}{1+x}{\rm sgn}(\sin2\theta(x)),
\end{equation}
where ${\rm sgn}(\cdot) $ is the sign function.
Substitute \eqref{Def} into \eqref{Gtheta}, and
solve the nonlinear system for $\theta$ with the initial condition $\theta(0)=\theta_0$.
It is not difficult to see that \eqref{Gtheta} has a unique piecewise
smooth global solution by a standard ODE existence and uniqueness theorem.
Thus $V(x)$ is well defined and
\begin{equation}\label{e6}
\frac{d\log R}{dx}= - \frac{a}{2k} \frac{\sin2\theta}{1+x}.
\end{equation}
Under the notations in the proof of Theorem \ref{thm1},
we have
\begin{eqnarray*}
% \nonumber to remove numbering (before each equation)
\int_{1}^x\frac{|\sin2\theta(y)|}{y+1}dy&\geq& \sum_{i=1}^{n+1}(\frac{1}{k}+\frac{O(1)}{1+i})\frac{2k}{(\pi+\epsilon)i}+O(1)\\
&\geq&( \frac{2}{\pi}-\epsilon)\ln x+O(1).
\end{eqnarray*}
Combining with \eqref{e6}, it is easy to see
\begin{equation*}
\log R(x)\leq -(\frac{a}{k\pi}-\epsilon)\ln x +O(1).
\end{equation*}
Assume $k<\frac{2a}{\pi}$, then
\begin{equation*}
\frac{a}{k\pi}-\epsilon>\frac{1}{2}+\epsilon.
\end{equation*}
Thus
\begin{equation*}
R(x)^2\leq \frac{1}{x ^{1+\epsilon}},
\end{equation*}
for large $x$. This yields $u\in L^2(0,\infty)$, so $\lambda=k^2$ is an eigenvalue.
\end{proof}
\section*{Acknowledgments}
The paper is inspired by Fan Yang's talk at ergodic Schr\"odinger operators weekly graduate seminar at UCI, where she presented paper \cite{ki2}. I am appreciated her fascinating presentation.
I would like to thank Svetlana Jitomirskaya for comments on earlier versions of the manuscript. I also
thank Barry Simon for telling me the full history of the problem (see footnote 1).
The author was supported by the AMS-Simons Travel Grant 2016-2018 and NSF DMS-1700314. This research was also
partially supported by NSF DMS-1401204.
\footnotesize
\begin{thebibliography}{10}
\bibitem{ani}
L.~Anikeeva.
\newblock The absence of solutions in ${L}^2 [x_0,\infty]$ of a second order
linear differential equation.
\newblock {\em Differential Equations}, 6:1721--1724, 1970.
\bibitem{atk}
F.~Atkinson and W.~Everitt.
\newblock Bounds for the point spectrum for a {S}turm-{L}iouville equation.
\newblock {\em Proceedings of the Royal Society of Edinburgh Section A:
Mathematics}, 80(1-2):57--66, 1978.
\bibitem{eas}
M.~S.~P. Eastham and H.~Kalf.
\newblock {\em Schr{\"o}dinger-type operators with continuous spectra},
volume~65.
\newblock Pitman Publishing, 1982.
\bibitem{Hal}
S.~Halvorsen.
\newblock Sharp bounds and ${L}_p-$stability for solutions of secondorder
linear ordinary differential equations.
\newblock {\em Report No. 9/75, Matematisk Institutt, Trondheim}, 1975.
\bibitem{kato}
T.~Kato.
\newblock Growth properties of solutions of the reduced wave equation with a
variable coefficient.
\newblock {\em Communications on Pure and Applied Mathematics}, 12(3):403--425,
1959.
\bibitem{ki1}
A.~Kiselev.
\newblock Imbedded singular continuous spectrum for {S}chr{\"o}dinger
operators.
\newblock {\em Journal of the American Mathematical Society}, 18(3):571--603,
2005.
\bibitem{ki2}
A.~Kiselev, Y.~Last, and B.~Simon.
\newblock Modified {P}r{\"u}fer and {E}{F}{G}{P} transforms and the spectral
analysis of one-dimensional {S}chr{\"o}dinger operators.
\newblock {\em Communications in mathematical physics}, 194(1):1--45, 1998.
\bibitem{naboko}
S.~N. Naboko.
\newblock Dense point spectra of {S}chr{\"o}dinger and {D}irac operators.
\newblock {\em Theoretical and Mathematical Physics}, 68(1):646--653, 1986.
\bibitem{remling1998absolutely}
C.~Remling.
\newblock The absolutely continuous spectrum of one-dimensional
{S}chr{\"o}dinger operators with decaying potentials.
\newblock {\em Communications in mathematical physics}, 193(1):151--170, 1998.
\bibitem{simon}
B.~Simon.
\newblock Some {S}chr{\"o}dinger operators with dense point spectrum.
\newblock {\em Proceedings of the American Mathematical Society},
125(1):203--208, 1997.
\bibitem{simon2017tosio}
B.~Simon.
\newblock Tosio {K}ato's {W}ork on {N}on--{R}elativistic {Q}uantum {M}echanics.
\newblock {\em arXiv preprint arXiv:1711.00528}, 2017.
\end{thebibliography}
\end{document}