%% @texfile{ %% filename="RelOsc.tex", %% version="1.0", %% date="April-2008", %% cdate="20061012", %% filetype="LaTeX2e", %% journal="Commun. Math. Phys. 287:2, 613-640 (2009)", %% doi="10.1007/s00220-008-0600-8", %% copyright="Copyright (C) H. Krueger and G.Teschl" %% } \documentclass{amsart} %\newcommand{\href}[2]{ #2 } \usepackage{hyperref} %%%%%%%%%THEOREMS%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \newtheorem{theorem}{Theorem}[section] \newtheorem{definition}[theorem]{Definition} \newtheorem{lemma}[theorem]{Lemma} \newtheorem{corollary}[theorem]{Corollary} \newtheorem{remark}[theorem]{Remark} \newtheorem{hypothesis}[theorem]{Hypothesis {\bf H.}\hspace*{-0.6ex}} %%%%%%%%%%%%%%FONTS%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \newcommand{\R}{{\mathbb R}} \newcommand{\N}{{\mathbb N}} \newcommand{\Z}{{\mathbb Z}} \newcommand{\C}{{\mathbb C}} %%%%%%%%%%%%%%%%%%ABBRS%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \newcommand{\nn}{\nonumber} \newcommand{\be}{\begin{equation}} \newcommand{\ee}{\end{equation}} \newcommand{\ul}{\underline} \newcommand{\ol}{\overline} \newcommand{\ti}{\tilde} \newcommand{\spr}[2]{\langle #1 , #2 \rangle} \newcommand{\id}{{\rm 1\hspace{-0.6ex}l}} \newcommand{\E}{\mathrm{e}} \newcommand{\I}{\mathrm{i}} \newcommand{\sgn}{\mathrm{sgn}} \newcommand{\tr}{\mathrm{tr}} \newcommand{\im}{\mathrm{Im}} \newcommand{\re}{\mathrm{Re}} \newcommand{\dom}{\mathfrak{D}} \newcommand{\fdom}{\mathfrak{Q}} \DeclareMathOperator{\Ker}{\mathrm{Ker}} \DeclareMathOperator{\Ran}{Ran} \DeclareMathOperator{\supp}{supp} \DeclareMathOperator*{\slim}{s-lim} \DeclareMathOperator*{\trlim}{tr-lim} \newcommand{\floor}[1]{\lfloor#1 \rfloor} \newcommand{\ceil}[1]{\lceil#1 \rceil} %%%%%%%%%%%%%%%GREEK%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \newcommand{\eps}{\varepsilon} \newcommand{\vphi}{\varphi} \newcommand{\sig}{\sigma} \newcommand{\lam}{\lambda} \newcommand{\gam}{\gamma} \newcommand{\om}{\omega} %%%%%%%%%%%%%%%%%%%%%%%%NUMBERING%%%%%%%%%%%%%%%%%%%%%%%% \renewcommand{\labelenumi}{(\roman{enumi})} \numberwithin{equation}{section} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{document} \title[Relative Oscillation Theory]{Relative Oscillation Theory, Weighted Zeros of the Wronskian, and the Spectral Shift Function} \author[H. Kr\"uger]{Helge Kr\"uger} \address{Faculty of Mathematics\\ Nordbergstrasse 15\\ 1090 Wien\\ Austria} \email{\href{mailto:helge.krueger@rice.edu}{helge.krueger@rice.edu}} \urladdr{\href{http://math.rice.edu/~hk7/}{http://math.rice.edu/\string~hk7/}} \curraddr{Department of Mathematics, Rice University, Houston, TX 77005, USA} \author[G. Teschl]{Gerald Teschl} \address{Faculty of Mathematics\\ University of Vienna\\ Nordbergstrasse 15\\ 1090 Wien\\ Austria\\ and International Erwin Schr\"odinger Institute for Mathematical Physics\\ Boltzmanngasse 9\\ 1090 Wien\\ Austria} \email{\href{mailto:Gerald.Teschl@univie.ac.at}{Gerald.Teschl@univie.ac.at}} \urladdr{\href{http://www.mat.univie.ac.at/~gerald/}{http://www.mat.univie.ac.at/\string~gerald/}} \thanks{{\it Commun. Math. Phys. {\bf 287:2}, 613--640 (2009).}} \thanks{{\it Research supported by the Austrian Science Fund (FWF) under Grant No.\ Y330}} \keywords{Sturm--Liouville operators, oscillation theory, spectral shift function} \subjclass[2000]{Primary 34B24, 34C10; Secondary 34L15, 34L05} \begin{abstract} We develop an analog of classical oscillation theory for Sturm--Liouville operators which, rather than measuring the spectrum of one single operator, measures the difference between the spectra of two different operators. This is done by replacing zeros of solutions of one operator by weighted zeros of Wronskians of solutions of two different operators. In particular, we show that a Sturm-type comparison theorem still holds in this situation and demonstrate how this can be used to investigate the number of eigenvalues in essential spectral gaps. Furthermore, the connection with Krein's spectral shift function is established. \end{abstract} \maketitle \section{Introduction} \label{sec:int} This paper is concerned with oscillation theory for Sturm--Liouville operators which was originated with Sturm's celebrated memoir \cite{stu} over a hundred and seventy years ago. Since then many extensions have been made, which are far too numerous to be listed here. Rather we refer to the recent survey by Simon \cite{sim} or the recent monograph by Zettl \cite{zet}. However, there are still open problems remaining: While classical oscillation theory works perfectly below the essential spectrum, its applicability in essential spectral gaps is still limited. The aim of the present paper is to help fill this gap and provide a new powerful oscillation theory which works inside essential spectral gaps. Around 1948 Hartman and others extended classical oscillation theory to singular Sturm--Liouville operators by showing the following in a series of papers (\cite{har1}, \cite{har2}, \cite{har3}): For a given Sturm--Liouville operator $H=-\frac{d^2}{dx^2}+q(x)$ on $L^2(0,\infty)$ with Dirichlet boundary condition, let $\psi(\lam,x)$ be the solution of $H \psi = \lam \psi$ satisfying the Dirichlet boundary condition. Then the dimension of the spectral projection $P_{(-\infty, \lam)}(H)$ equals the number of zeros of $\psi(\lam)$ in $(0,\infty)$. Of course a naive use of this result to count the number of eigenvalues inside an essential spectral gap would just give $\infty-\infty$. However, they also showed that this can be avoided by taking appropriate limits. More precisely, the dimension of $P_{(\lam_0, \lam_1)}(H)$ can be obtained by considering the difference of the number of zeros of the solutions $\psi(\lam_1)$ and $\psi(\lam_0)$ inside a finite interval $(0,b)$, and performing a limit $b \to \infty$. Further important extensions were later given by Weidmann in \cite{wd}, \cite{wd1} (see also his lecture notes \cite{wdln}). Only recently it was shown in \cite{gst} by Gesztesy, Simon, and one of us, that these limits can be avoided by using a renormalized version of oscillation theory, that is, by counting zeros of Wronskians of solutions instead (see again \cite{sim} for a pedagogical discussion respectively \cite{kms3} for an interesting application to minimal surfaces). That zeros of the Wronskian are related to oscillation theory is indicated by an old paper of Leighton \cite{lei}, who noted that if two solutions have a non-vanishing Wronskian, then their zeros must intertwine each other. So there are several tools to count eigenvalues in essential spectral gaps and one might ask what could be missing. To this end observe that all the previously mentioned results have one thing in common, they all work with only one operator $H$. One the other hand, a typical question inside an essential spectral gap is, what kind of perturbations $H_1$ of a given operator $H_0$ produce a finite/infinite number of eigenvalues. Such kind of problems clearly ask for an oscillation theory which compares the spectra of two given operators $H_0$ and $H_1$. To the best of our knowledge, such an oscillation theory was not available to date. We will demonstrate in this paper, that the right object to look at are zeros of Wronskians of suitable solutions of $H_1$ respectively $H_0$. In addition, since $H_1-H_0$ is not necessarily of one sign, we will weight the zeros according to the sign of $H_1-H_0$ at a zero (see Section~\ref{sec:wz} below). This will allow us to develop an analog of Sturm's oscillation theory comparing the spectra of two operators including an analog of Sturm's comparison theorem, which is applicable inside spectral gaps. To further illustrate this, let us look at Kneser's result (\cite{kn}, see also \cite{gu} for generalizations). Let \begin{equation} \tau = -\frac{d^2}{dx^2} + q(x) \end{equation} be a Sturm--Liouville operator on the interval $(0,\infty)$ with $q(x)\to 0$ as $x\to\infty$. Then the bottom of the essential spectrum is an accumulation point of eigenvalues if \begin{equation} \liminf_{x \to\infty} -4 x^2 q(x) >1 \end{equation} and it is no accumulation point if \begin{equation} \limsup_{x \to\infty} -4 x^2 q(x) <1. \end{equation} The proof consists of the observation that the Euler equation \begin{equation} \tau_0= -\frac{d^2}{dx^2} + \frac{\mu}{x^2} \end{equation} is explicitly solvable plus an application of Sturm's comparison theorem between $\tau_0$ and $\tau$. The corresponding result within spectral gaps is Rofe-Beketov's theorem for periodic operators (\cite{rb5}, see also the recent monograph \cite{rbk}). We state it in the version of \cite{kms}, where a detailed proof plus extensions are given. Suppose $\lim_{x \rightarrow \infty} x^2 (q(x) - q_0(x))$ exists. Then a boundary point $E_n$ of the essential spectrum is an accumulation point of eigenvalues if \be \lim_{x\to\infty} \kappa_n x^2 (q(x)-q_0(x)) >1 \ee and no accumulation point if \be \lim_{x\to\infty }\kappa_n x^2 (q(x)-q_0(x)) <1, \ee where the critical constant $\kappa_n$ will be defined in Section~\ref{sec:app} below. Comparing this result with Kneser's result, there is an obvious mismatch: While Kneser's result covers all potentials which are above respectively below the critical case $q(x)=\frac{-1}{4x^2}$, the above result only covers the case where $q(x)$ has a precise asymptotic $q(x)=q_0(x) + \frac{c}{x^2} +o(x^{-2})$. As a short application we will show how a comparison argument analogous to the one for Kneser's theorem, can be used to fill this gap. While the ingredients for such a comparison argument can essentially be found in the results by Weidmann \cite{wd1} (see the paragraph after Theorem~\ref{thm:rosc} for a more detailed discussion of earlier results), we still feel there is a need to advocate their use. In addition, we hope that our novel interpretation as weighted zeros of Wronskians will lead to new relative oscillation criteria and stimulate further research in this direction. Some possible extensions are listed at the end of Section~\ref{sec:rot}. Furthermore, the arguments used in \cite{wd1} typically lead to spectral estimates (compare Satz~4.1 in \cite{wd1} respectively Lemma~\ref{lem:estproj} below). On the other hand, the results in \cite{har2} and \cite{gst} give not only estimates, but precise equalities between spectral data and zeros of solutions respectively Wronskians of solutions. Hence, as our main results we will establish precise equalities between spectral shifts and the number of weighted zeros of Wronskians of certain solutions in Theorem~\ref{thm:wronskzeros} and Theorem~\ref{thmsing}. \section{Weighted zeros of Wronskians, Pr\"ufer angles, and regular operators} \label{sec:wz} The key ingredient will be roughly speaking weighted zeros of Wronskians of solutions of different Sturm--Liouville operators. However, this naive definition has a few subtle problems which we need to clarify first. This will be done by giving an alternate definition in terms of Pr\"ufer angles and establishing equivalence for those cases where both definitions are well-defined. To set the stage, we will consider Sturm--Liouville operators on $L^{2}((a,b), r\,dx)$ with $-\infty \le a0. \end{equation} We will use $\tau$ to describe the formal differentiation expression and $H$ the operator given by $\tau$ with separated boundary conditions at $a$ and/or $b$. If $a$ (resp.\ $b$) is finite and $p^{-1},q,r$ are in addition integrable near $a$ (resp.\ $b$), we will say $a$ (resp.\ $b$) is a \textit{regular} endpoint. We will say $\tau$ respectively $H$ is \textit{regular} if both $a$ and $b$ are regular. For every $z\in\C\backslash\sig_{ess}(H)$ there is a unique (up to a constant) solution $\psi_-(z,x)$ of $\tau u = z u$ which is in $L^2$ near $a$ and satisfies the boundary condition at $a$ (if any). Similarly there is such a solution $\psi_+(z,x)$ near $b$. One of our main objects will be the (modified) Wronskian \be W_x(u_0,u_1)= u_0(x)\, p(x)u_1'(x) - p(x)u_0'(x)\, u_1(x) \ee of two functions $u_0$, $u_1$ and its zeros. Here we think of $u_0$ and $u_1$ as two solutions of two different Sturm--Liouville equations \be \label{stlij} \tau_j = \frac{1}{r} \Big(- \frac{d}{dx} p \frac{d}{dx} + q_j \Big), \qquad j=0,1. \ee Note that we have chosen $p_0=p_1\equiv p$ here. The case $p_0\ne p_1$ will be given in \cite{kt2}. Classical oscillation theory counts the zeros of solutions $u_j$, which are always simple since $u$ and $pu'$ cannot both vanish except if $u\equiv 0$. This is no longer true for the zeros of the Wronskian $W_x(u_0,u_1)$. In fact, it could even happen that the Wronskian vanishes on an entire interval if $q_0=q_1$ on this interval (cf.\ (\ref{dwr})). Such a situation will be counted as just one zero, in other words, we will only count sign flips. Furthermore, classical oscillation theory involves the spectrum of only one operator, but we want to measure the difference between the spectra of two operators. That is, we need a signed quantity. Hence we will weight the sign flips according to the sign of $q_1-q_0$. Of course this {\em definition} is not good enough for the case $q_1-q_0 \in L^1_{loc}$ considered here. Hence we will give a precise definition in terms of Pr\"ufer variables next. We begin by recalling the definition of Pr\"ufer variables $\rho_u$, $\theta_u$ of an absolutely continuous function $u$: \be u(x)=\rho_u(x)\sin(\theta_u(x)), \qquad p(x) u'(x)=\rho_u(x) \cos(\theta_u(x)). \ee If $(u(x), p(x) u'(x))$ is never $(0,0)$ and $u, pu'$ are absolutely continuous, then $\rho_u$ is positive and $\theta_u$ is uniquely determined once a value of $\theta_u(x_0)$ is chosen by requiring continuity of $\theta_u$. Notice that \be \label{wpruefer} W_x(u,v)= -\rho_u(x)\rho_v(x)\sin(\Delta_{v,u}(x)), \qquad \Delta_{v,u}(x) = \theta_v(x)-\theta_u(x). \ee Hence the Wronskian vanishes if and only if the two Pr\"ufer angles differ by a multiple of $\pi$. We will call the total difference \be \#_{(c,d)}(u_0,u_1) = \ceil{\Delta_{1,0}(d) / \pi} - \floor{\Delta_{1,0}(c) / \pi} -1 \ee the number of weighted sign flips in $(c,d)$, where we have written $\Delta_{1,0}(x)= \Delta_{u_1,u_0}$ for brevity. Next, let us show that this agrees with our considerations above. We take two solutions $u_j$, $j=1,2$, of $\tau_j u_j =\lam_j u_j$ and associated Pr\"ufer variables $\rho_j$, $\theta_j$. Since we can replace $q \to q - \lam r$ it is no restriction to assume $\lam_0=\lam_1=0$. Under these assumptions $W_x(u_0,u_1)$ is absolutely continuous and satisfies \be \label{dwr} W'_x(u_0,u_1) = -(q_0(x)-q_1(x)) u_0(x) u_1(x). \ee \begin{lemma}\label{lem:delinc} Abbreviate $\Delta_{1,0}(x) = \theta_1(x) - \theta_0(x)$ and suppose $\Delta_{1,0}(x_0) \equiv 0 \mod \pi$. If $q_0(x)-q_1(x)$ is (i) negative, (ii) zero, or (iii) positive for a.e.\ $x\in(x_0,x_0+\eps)$ respectively for a.e.\ $x\in(x_0-\eps,x_0)$ for some $\eps>0$, then the same is true for $(\Delta_{1,0}(x) - \Delta_{1,0}(x_0))/(x - x_0)$. \end{lemma} \begin{proof} By (\ref{dwr}) we have \begin{align} \nn W_x(u_0,u_1) &= - \rho_0(x)\rho_1(x)\sin(\Delta_{1,0}(x))\\ &= - \int_{x_0}^x (q_0(t)-q_1(t)) u_0(t) u_1(t) dt, \end{align} and there are two cases to distinguish: Either $u_0(x_0)$, $u_1(x_0)$ are both different from zero or both equal to zero. If both are different from zero, $u_0(t) u_1(t)$ does not change sign at $t=x_0$. If both are equal to zero, they must change sign at $x_0$ and hence again $u_0(t) u_1(t)$ does not change sign at $t=x_0$. Now the claim is evident. \end{proof} \noindent Hence $\#_{(c,d)}(u_0,u_1)$ counts the weighted sign flips of the Wronskian $W_x(u_0,u_1)$, where a sign flip is counted as $+1$ if $q_0-q_1$ is positive in a neighborhood of the sign flip, it is counted as $-1$ if $q_0-q_1$ is negative in a neighborhood of the sign flip. If $q_0-q_1$ changes sign (i.e., it is positive on one side and negative on the other) the Wronskian will not change its sign. In particular, we obtain \begin{lemma}\label{lemnbzer2} Let $u_0$, $u_1$ solve $\tau_j u_j = 0$, $j=0,1$, where $q_0-q_1\ge 0$. Then $\#_{(a,b)}(u_0,u_1)$ equals the number sign flips of $W(u_0,u_1)$ inside the interval $(a,b)$. \end{lemma} \noindent In the case $q_0-q_1\le 0$ we get of course the corresponding negative number except for the fact that zeros at the boundary points are counted as well since $\floor{-x}= -\ceil{x}$. That is, if $q_0-q_1> 0$, then $\#_{(c,d)}(u_0,u_1)$ equals the number of zeros of the Wronskian in $(c,d)$ while if $q_0-q_1< 0$, it equals minus the number of zeros in $[c,d]$. In the next theorem we will see that this is quite natural. In addition, note that $\#(u,u) = -1$. Finally, we establish the connection with the spectrum of regular operators. For this let $H_0$, $H_1$ be self-adjoint extensions of $\tau_0$, $\tau_1$ with the same separated boundary conditions. \begin{theorem} \label{thm:reg} Let $H_0$, $H_1$ be regular Sturm--Liouville operators associated with (\ref{stlij}) and the same boundary conditions at $a$ and $b$. Then \be \label{eqreg} \dim\Ran\, P_{(-\infty, \lam_1)}(H_1) - \dim\Ran\, P_{(-\infty, \lam_0]}(H_0) = \#_{(a,b)}(\psi_{0,\pm}(\lam_0), \psi_{1,\mp}(\lam_1)). \ee \end{theorem} \noindent The proof will be given in Section~\ref{sec:proofreg} employing interpolation between $H_0$ and $H_1$, using $H_\eps= (1-\eps) H_0 + \eps H_1$ together with a careful analysis of Pr\"ufer angles. It is important to observe that in the special case $H_1=H_0$, the left-hand side equals $\dim\Ran\, P_{(\lam_1,\lam_0)}(H_0)$ if $\lam_1>\lam_0$ and $-\dim\Ran\, P_{[\lam_0,\lam_1]}(H_0)$ if $\lam_1<\lam_0$. This is of course in accordance with our previous observation that $\#(\psi_{0,\pm}(\lam_0), \psi_{1,\mp}(\lam_1))$ equals the number of zeros in $(a,b)$ if $\lam_1>\lam_0$ while it equals minus the numbers of zeros in $[a,b]$ if $\lam_1<\lam_0$. \section{Relative Oscillation Theory} \label{sec:rot} After these preparations we are now ready to develop relative oscillation theory. For the connections with earlier work we refer to the discussion after Theorem~\ref{thm:rosc}. \begin{definition}\label{def:wsf} For $\tau_0$, $\tau_1$ possibly singular Sturm--Liouville operators as in (\ref{stlij}) on $(a,b)$, we define \be \underline{\#}(u_0,u_1) = \liminf_{d \uparrow b,\,c \downarrow a} \#_{(c,d)}(u_0,u_1) \quad\mbox{and}\quad \overline{\#}(u_0,u_1) = \limsup_{d \uparrow b,\,c \downarrow a} \#_{(c,d)}(u_0,u_1), \ee where $\tau_j u_j = \lam_j u_j$, $j=0,1$. We say that $\#(u_0,u_1)$ exists, if $\overline{\#}(u_0,u_1)=\underline{\#}(u_0,u_1)$, and write \be \#(u_0,u_1) = \overline{\#}(u_0,u_1)=\underline{\#}(u_0,u_1) \ee in this case. \end{definition} \noindent By Lemma~\ref{lem:delinc} one infers that $\#(u_0,u_1)$ exists if $q_0-\lam_0r- q_1+\lam_1r$ has the same definite sign near the endpoints $a$ and $b$. On the other hand, note that $\#(u_0,u_1)$ might not exist even if both $a$ and $b$ are regular, since the difference of Pr\"ufer angles might oscillate around a multiple of $\pi$ near an endpoint. Furthermore, even if it exists, one has $\#(u_0,u_1) = \#_{(a,b)}(u_0,u_1)$ only if there are no zeros at the endpoints (or if $q_0-\lam_0r-q_1+\lam_1r\ge 0$ at least near the endpoints). \begin{remark} Note that cases like $\underline{\#}(u_0,u_1)=-\infty$ and $\overline{\#}(u_0,u_1) = +\infty$ can occur. To construct such a situation let $\tau_j=-\frac{d^2}{dx^2} + q_j$ on $(0,\infty)$ and $\lam_0=\lam_1=0$. Let $u_j(x)$ be the solutions satisfying a Neumann boundary condition at $u_j'(0)=0$. Choose $q_0(x)=0$, $q_1(x)=1$ on $(0,\pi)$ such that $\theta_1(\pi)-\theta_0(\pi)=\frac{3\pi}{2}-\frac{\pi}{2}=\pi$. Next, choose $q_0(x)=1$, $q_1(x)=0$ on $(\pi,3\pi)$ such that $\theta_1(3\pi)-\theta_0(3\pi)=\frac{3\pi}{2}-\frac{5\pi}{2}=-\pi$. Next, choose $q_0(x)=0$, $q_1(x)=1$ on $(3\pi,6\pi)$ such that $\theta_1(6\pi)-\theta_0(6\pi)=\frac{9\pi}{2}-\frac{5\pi}{2}=2\pi$. Etc. \end{remark} \noindent We begin with our analog of Sturm's comparison theorem for zeros of Wronskians. We will also establish a triangle-type inequality which will help us to provide streamlined proofs below. As with Sturm's comparison theorem, the proofs are elementary. \begin{theorem}[Comparison theorem for Wronskians]\label{thm:scw} Suppose $u_j$ satisfies $\tau_j u_j = \lam_j u_j$, $j=0,1,2$, where $\lam_0 r -q_0 \le \lam_1 r - q_1 \le \lam_2 r - q_2$. If $c1$. To finish our proof, let $\ti{u}_2$ be a second linearly independent solution. Then, since $W(u_2,\ti{u}_2)$ is constant, we can assume $0<\Delta_{\ti{u}_2,u_2}(x)<\pi$. This implies $\Delta_{\ti{u}_2,u_0}(c) = \Delta_{\ti{u}_2,u_2}(c)<\pi$ and $\Delta_{\ti{u}_2,u_0}(d) = \Delta_{\ti{u}_2,u_2}(d) + \Delta_{u_2,u_0}(d)> \pi$. Consequently $W_x(u_0,\ti{u}_2)$ also has at least one sign flip in $(c,d)$. The second claim is proven analogously. \end{proof} \begin{theorem}[Triangle inequality for Wronskians]\label{thm:wtriang} Suppose $u_j$, $j=0,1,2$ are given functions with $u_j$, $pu_j'$ absolutely continuous and $(u_j(x), p(x) u_j'(x)) \neq (0,0)$ for all $x$. Then \be \ul{\#}(u_0,u_1) + \ul{\#}(u_1,u_2) - 1 \leq \ul{\#}(u_0,u_2) \leq \ul{\#}(u_0,u_1) + \ul{\#}(u_1,u_2) + 1, \ee and similarly for $\ul{\#}$ replaced by $\ol{\#}$. \end{theorem} \begin{proof} Take $a < c < d < b$. By definition $$ \#_{(c,d)} (u_0, u_2) = \ceil{\Delta_{2,0} (d) / \pi} - \floor{\Delta_{2,0} (c) / \pi} - 1, $$ and using $\floor{x}+\floor{y} \le \floor{x+y} \le \floor{x}+\floor{y} +1$ respectively $\ceil{x}+\ceil{y} -1 \le \ceil{x+y} \le \ceil{x}+\ceil{y}$ and $\Delta_{2,0} = \Delta_{2,1} + \Delta_{1,0}$, we obtain $$ \#_{(c,d)} (u_0, u_2) \leq \#_{(c,d)} (u_0,u_1) + \#_{(c,d)} (u_1,u_2) + 1. $$ Thus the result follows by taking the limits $c\downarrow a$ and $d\uparrow b$. \end{proof} \noindent We recall that in classical oscillation theory $\tau$ is called oscillatory if a solution of $\tau u = 0$ has infinitely many zeros. \begin{definition}\label{def:relosc} We call $\tau_1$ relatively nonoscillatory with respect to $\tau_0$, if the quantities $\underline{\#}(u_0, u_1)$ and $\overline{\#}(u_0, u_1)$ are finite for all solutions $\tau_j u_j = 0$, $j = 0,1$. We call $\tau_1$ relatively oscillatory with respect to $\tau_0$, if one of the quantities $\underline{\#}(u_0, u_1)$ or $\overline{\#}(u_0, u_1)$ is infinite for some solutions $\tau_j u_j = 0$, $j = 0,1$. \end{definition} \noindent Note that this definition is in fact independent of the solutions chosen as a straightforward application of our triangle inequality (cf.\ Theorem~\ref{thm:wtriang}) shows. \begin{corollary}\label{cor:indisol} Let $\tau_j u_j = \tau_j v_j = 0$, $j=0,1$. Then \be |\underline{\#}(u_0, u_1) - \underline{\#}(v_0, v_1)| \le 4,\quad |\overline{\#}(u_0, u_1) - \overline{\#}(v_0, v_1)| \le 4. \ee \end{corollary} \begin{proof} By our comparison theorem we have $|\#(u_j, v_j)|\le 1$, $j =0,1$. Now use the triangle inequality, twice. \end{proof} \noindent The bounds can be improved using our comparison theorem for Wronskians to be $\leq 2$ in the case of perturbations of definite sign. If $\tau_0$ is nonoscillatory our definition reduces to the classical one. \begin{lemma} Suppose $\tau_0$ is a nonoscillatory operator, then $\tau_1$ is relatively nonoscillatory (resp.\ oscillatory) with respect to $\tau_0$, if and only if $\tau_1$ is nonoscillatory (resp.\ oscillatory). \end{lemma} \begin{proof} This follows by taking limits in $$ |\#_{(c,d)}(u,v) - \#_{(c,d)}(u) + \#_{(c,d)}(v)|\leq 2, $$ where $\#_{(c,d)}(u)= \ceil{\theta_u(d) / \pi} - \floor{\theta_u(c) / \pi} -1$ is the number of zeros of $u$ inside $(c,d)$. \end{proof} \noindent To demonstrate the usefulness of Definition~\ref{def:relosc}, we now establish its connection with the spectra of self-adjoint operators associated with $\tau_j$, $j=0,1$. \begin{theorem} \label{thm:rosc} Let $H_j$ be self-adjoint operators associated with $\tau_j$, $j=0,1$. Then \begin{enumerate} \item $\tau_0-\lam_0$ is relatively nonoscillatory with respect to $\tau_0-\lam_1$ if and only if $\dim\Ran P_{(\lam_0,\lam_1)}(H_0)<\infty$. \item Suppose $\dim\Ran P_{(\lam_0,\lam_1)}(H_0)<\infty$ and $\tau_1-\lam$ is relatively nonoscillatory with respect to $\tau_0-\lam$ for one $\lam \in [\lam_0,\lam_1]$. Then it is relatively nonoscillatory for all $\lam\in[\lam_0,\lam_1]$ if and only if $\dim\Ran P_{(\lam_0,\lam_1)}(H_1)<\infty$. \end{enumerate} \end{theorem} \begin{proof} (i) This is item (i) of \cite[Thm.7.5]{gst}. (ii) Let $\lam, \ti{\lam} \in [\lam_0,\lam_1]$, $\tau_j u_j(\lam) = \lam u_j(\lam)$, $j=0,1$ and suppose we are relatively nonoscillatory at $\lam$. Then applying our triangle inequality twice implies \begin{align*} \overline{\#}(u_0(\ti{\lam}),u_1(\ti{\lam})) &\leq \overline{\#}(u_0(\ti{\lam}), u_1(\lam)) + \#(u_1(\lam),u_1(\ti{\lam})) +1\\ &\leq \#(u_0(\ti{\lam}), u_0(\lam)) + \overline{\#}(u_0(\lam),u_1(\lam)) + \#(u_1(\lam),u_1(\ti{\lam})) + 2, \end{align*} and similar estimates with the roles of $\lam$ and $\ti{\lam}$ interchanged and $\overline{\#}$ replaced by $\underline{\#}$. Hence if $\dim\Ran P_{(\lam_0,\lam_1)}(H_1)<\infty$ we are relatively nonoscillatory by (i). The converse direction is proven analogously. \end{proof} \noindent We remark that item (i), which corresponds to the case of equal operators $H_0=H_1$ but different spectral parameters $\lam_0\ne \lam_1$, is what was called renormalized oscillation theory by Gesztesy, Simon, and Teschl in \cite{gst}. It also follows from earlier results by Hartman \cite{har2}, Weidmann \cite[Satz~4.1]{wd1} (see also \cite{wdln}), which was pointed out in the appendix of \cite{kms} and was called relative oscillation theorem there. An argument in the spirit of item (ii) was also one of the key ingredients in Rofe-Beketov's original work (cf.\ \cite{rb5}). For a practical application of this theorem one needs of course criteria when $\tau_1-\lam$ is relatively nonoscillatory with respect to $\tau_0-\lam$ for $\lam$ inside an essential spectral gap. \begin{lemma}\label{lem:nonoscingap} Let $\lim_{x\to a} r(x)^{-1} (q_0(x) - q_1(x)) = 0$ if $a$ is singular, and similarly, $\lim_{x\to b} r(x)^{-1} (q_0(x) - q_1(x)) = 0$ if $b$ is singular. Then $\sig_{ess}(H_0)=\sig_{ess}(H_1)$ and $\tau_1 - \lam$ is relatively nonoscillatory with respect to $\tau_0 - \lam$ for $\lam \in \R \backslash\sigma_{ess}(H_0)$. \end{lemma} \begin{proof} Since $\tau_1$ can be written as $\tau_1 = \tau_0 + \ti{q}_0 + \ti{q}_1$, where $\ti{q}_0$ has compact support near singular endpoints and $|\ti{q}_1|<\eps$, for arbitrarily small $\eps>0$, we infer that $R_{H_1}(z) - R_{H_0}(z)$ is the norm limit of compact operators. Thus $R_{H_1}(z) - R_{H_0}(z)$ is compact and hence $\sig_{ess}(H_0)=\sig_{ess}(H_1)$. Let $\delta>0$ be the distance of $\lam$ to the essential spectrum and choose $a < c < d < b$, such that $$ |r^{-1}(q_1(x) - q_0(x))| \le \delta/2,\qquad x\not\in(c,d). $$ Clearly $\#_{(c,d)} (u_0,u_1) < \infty$, since both operators are regular on $(c,d)$. Moreover, observe that $$ q_0 - \lam_+ r \leq q_1- \lam r \leq q_0- \lam_- r, \qquad \lam_\pm=\lam\pm\delta/2, $$ on $I = (a,c)$ or $I =(d,b)$. Then Theorem~\ref{thm:rosc}~(i) implies $\#_I(u_0(\lam_-), u_0(\lam_+)) < \infty$ and invoking Theorem~\ref{thm:scw} shows $\#_I(u_0(\lam_\pm), u_1(\lam)) < \infty$. From Theorem~\ref{thm:wtriang} and \ref{thm:rosc}~(i) we infer $$ \ol{\#}_I(u_0(\lam),u_1(\lam)) < \#_I(u_0(\lam),u_0(\lam_+)) + \#_I(u_0(\lam_+), u_1(\lam)) + 1 < \infty, $$ and similarly for $\ul{\#}_I(u_0(\lam),u_1(\lam))$. This shows that $\tau_1-\lam$ is relatively nonoscillatory with respect to $\tau_0$. \end{proof} \noindent Our next task is to reveal the precise relation between the number of weighted sign flips and the spectra of $H_1$ and $H_0$. The special case $H_0=H_1$ is covered by \cite{gst}: \begin{theorem}[\cite{gst}] \label{thm:gst} Let $H_0$ be a self-adjoint operator associated with $\tau_0$ and suppose $[\lam_0,\lam_1]\cap\sig_{ess}(H_0)=\emptyset$. Then \be \dim\Ran P_{(\lam_0, \lam_1)} (H_0) = \#(\psi_{0,\mp}(\lam_0),\psi_{0,\pm}(\lam_1)). \ee \end{theorem} \noindent We will provide an alternate proof in Section~\ref{sec:appreg}. Combining this result with our triangle inequality already gives some rough estimates in the spirit of Weidmann \cite{wd1} who treats the case $H_0=H_1$. \begin{lemma}\label{lem:estproj} For $j=0,1$ let $H_j$ be a self-adjoint operator associated with $\tau_j$ and separated boundary conditions. Suppose that $(\lam_0, \lam_1)\subseteq\R\backslash(\sig_{ess}(H_0)\cup\sig_{ess}(H_1))$, then \begin{align} \nonumber \dim\Ran P_{(\lam_0, \lam_1)} (H_1) &- \dim\Ran P_{(\lam_0, \lam_1)} (H_0)\\ & \leq \underline{\#}(\psi_{1,\mp} (\lam_1), \psi_{0,\pm} (\lam_1)) - \overline{\#}(\psi_{1,\mp} (\lam_0), \psi_{0,\pm} (\lam_0)) + 2, \end{align} respectively, \begin{align} \nonumber \dim\Ran P_{(\lam_0, \lam_1)} (H_1) &- \dim\Ran P_{(\lam_0, \lam_1)} (H_0)\\ & \geq \overline{\#}(\psi_{1,\mp} (\lam_1), \psi_{0,\pm} (\lam_1)) - \underline{\#}(\psi_{1,\mp} (\lam_0), \psi_{0,\pm} (\lam_0)) - 2. \end{align} \end{lemma} \begin{proof} By the triangle inequality (cf.\ Theorem~\ref{thm:wtriang}) we have \begin{align*} & \#_{(c,d)}(\psi_{1,-}(\lam_1),\psi_{1,+}(\lam_0)) - \#_{(c,d)}(\psi_{0,-}(\lam_1),\psi_{0,+}(\lam_0))\\ &\qquad \le \#_{(c,d)}(\psi_{1,-}(\lam_1),\psi_{0,+}(\lam_1)) + \#_{(c,d)}(\psi_{1,-}(\lam_1),\psi_{0,+}(\lam_1)) +2. \end{align*} The result now follows by taking limits using that $$ \lim_{c\downarrow a, d \uparrow b} \#_{(c,d)}(\psi_{1,-}(\lam_1),\psi_{1,+}(\lam_0)) = \dim\Ran P_{(\lam_0, \lam_1)} (H_1) $$ and $$ \lim_{c\downarrow a, d \uparrow b} \#_{(c,d)}(\psi_{0,-}(\lam_0),\psi_{0,+}(\lam_1)) = -\dim\Ran P_{(\lam_0, \lam_1)} (H_0) $$ by the previous theorem. The second claim follows similarly. \end{proof} \noindent To turn the inequalities into equalities in Lemma~\ref{lem:estproj} will be one of our remaining goals. Observe that for semibounded operators we can choose $\lam_0$ below the spectra of $H_0$ and $H_1$, causing $\#(\psi_{1,\mp}(\lam_0),\psi_{0,\pm}(\lam_0))$ to vanish: \begin{lemma} Let $r^{-1}(q_0- q_1)\leq \delta$ for some $\delta\in\R$. Furthermore, suppose that the operator $H_0$ associated with $\tau_0$ is bounded from below, $H_0 \ge E_0$, and the form domains of $H_0$ and $H_1$ are equal. Then \be \#(\psi_{1,\mp}(\lam),\psi_{0,\pm}(\lam)) = 0, \quad\lam 0$ for a lower endpoint of a spectral gap, and $\kappa_{2n} < 0$ for an upper endpoint. Before coming to our applications, we ensure that our hypotheses from the previous section are satisfied. We begin by computing the form domain of $H_0$. \begin{lemma} Abbreviate \be \fdom= \{ f \in L^2(1,\infty) \,|\, f \in AC[1,\infty), \: \sqrt{p} f' \in L^2(1,\infty)\}. \ee The form domain of $H_0$ is given by \be \fdom(H_0)= \{ f \in \fdom \,|\, f(1)=0\}. \ee Moreover, for every $\eps>0$ there is some $C>0$ such that \be \label{qdperbd} \sup_{x_0 \le x \le x_0+\alpha} |f(x)|^2 \le \eps \int_{x_0}^{x_0+\alpha} p(x) |f'(x)|^2 dx + C \int_{x_0}^{x_0+\alpha} |f(x)|^2 dx, \quad f\in \fdom. \ee In particular, $q_1-q_0$ is infinitesimally form bounded with respect to $H_0$ if \be \int_{x_0}^{x_0+\alpha} |q_1(x) - q_0(x)| dx < C_0, \qquad x_0 \in (1,\infty), \ee where $C_0$ is independent of $x_0$. \end{lemma} \begin{proof} Equation (\ref{qdperbd}) is a standard Sobolev estimate. For the case $p(x)\ne 1$ required here compare for example \cite[Lem~A.2]{gst}. Next, set \be A = \sqrt{p} \frac{d}{dx}, \qquad \dom(A) = \{ f \in \fdom \,|\, f(1)=0\} \ee and note that $A$ is then a closed operator with adjoint given by \begin{align}\nn A^* &= -\frac{d}{dx} \sqrt{p} ,\\ \dom(A^*) &= \{ f \in L^2(1,\infty) \,|\, \sqrt{p} f \in AC[1,\infty), \: (\sqrt{p} f)' \in L^2(1,\infty)\}. \end{align} Hence, \begin{align}\nn A^* A &= -\frac{d}{dx} p \frac{d}{dx},\\ \dom(A^* A) &= \{ f \in L^2(1,\infty) \,|\, f,pf' \in AC[1,\infty), \: (p f')' \in L^2(1,\infty), \: f(1)=0\} \end{align} is self-adjoint with $\fdom(A^* A)= \dom(A)$ and by (\ref{qdperbd}) $q_0$ is infinitesimally form bounded with respect to $A^* A$. Since the same is true for $q_1$ by assumption, the lemma is proven. \end{proof} \begin{lemma}\label{lem:HS} Let $H_0$ be an arbitrary Sturm--Liouville operator on $(a,b)$. Then $\sqrt{q} R_{H_0}(z)$ is Hilbert--Schmidt if and only if \be \| \sqrt{q} R_{H_0}(z) \|_{\mathcal{J}_2}^2 = \frac{1}{\im(z)} \int_a^b |q(x)| \im(G_0(z,x,x)) r(x) dx \ee is finite. Here $G_0(z,x,y) =(H_0-z)^{-1}(x,y)$ denotes the Green's function of $H_0$. \end{lemma} \begin{proof} From the first resolvent identity we have $$ G_0(z,x,y) - G_0(z',x,y) = (z-z') \int_a^b G_0(z,x,t) G_0(z',t,y) r(t) dt. $$ Setting $x=y$ and $z'=z^*$ we obtain $$ \im(G_0(z,x,x)) = \im(z) \int_a^b |G_0(z,x,t)|^2 r(t) dt. $$ Using this last formula to compute the Hilbert--Schmidt norm proves the lemma. \end{proof} We recall that in the case of periodic operators $G(z,x,x)$ is a bounded function of $x$. In fact, we have \be\label{estgfper} |G(z,x,y)| \leq const(z) \exp(-\gam(z) |x-y|), \ee where $\gam(z)>0$ denotes the Floquet exponent. Now we are ready to apply our theory: \begin{theorem}\label{thm:app} Let $p, q_0$ $\alpha$-periodic and $q_1$ a perturbed potential which is regular at $1$, such that either $q_0 - q_1 \in L^1(1,\infty)$ or $\lim_{x\to\infty} (q_0(x) - q_1(x))=0$. Define the differential expressions on $(1, \infty)$ by \be \tau_0 = - \frac d{dx} p(x) \frac d{dx} + q_0(x), \quad \tau_1 = - \frac d{dx} p(x) \frac d{dx} + q_1(x) \ee and let $H_0$, $H_1$ be the corresponding self-adjoint operators $L^{2}(1,\infty)$. Let $E_n$ be an endpoint of a gap in the essential spectrum of $H_0$ with corresponding $\kappa_n$ given by (\ref{defkap}). Then $E_n$ is an accumulation point of eigenvalues of $H_1$ if \be \liminf_{x \to\infty} \kappa_n x^2 (q_1 (x) - q_0(x)) > 1. \ee and $E_n$ is no accumulation point of eigenvalues of $H_1$ if \be \limsup_{x \to\infty} \kappa_n x^2 (q_1 (x) - q_0(x)) < 1 \ee and $\kappa_n (q_1 - q_0) \geq 0$ near infinity. \end{theorem} \begin{proof} Lemma~\ref{lem:HS} together with (\ref{estgfper}) shows that Hypothesis~\ref{hyp:h0h1} is satisfied if $q_0 - q_1 \in L^1(1,\infty)$. Thus we can either apply Theorem~\ref{thm:wronskzeros} or Theorem~\ref{thmsing} to conclude that $\tau_1-\lam$ is nonoscillatory with respect to $\tau_0-\lam$ for any $\lam\in\R\backslash\sig_{ess}(H_0)$. Hence, Theorem~\ref{thm:rosc} (ii) is applicable and it suffices to show that $\tau_1$ is relatively oscillatory (resp.\ nonoscillatory) with respect to $\tau_0$. Without restriction, we assume $\kappa_n > 0$. For the first statement, note that we can find $c$, $\eps>0$ such that $$ q_0(x) < q_0(x) + \frac{1}{(\kappa_n - \eps)x^2} < q_1(x), \qquad x>c. $$ Now since perturbations with compact support only add finitely many eigenvalues, it is no restriction to assume $c = 1$. Next, \cite[Thm.~1]{kms} shows that $\tau_0 + (\kappa_n-\eps)^{-1} x^{-2} - E_n$ is relatively oscillatory with respect to $\tau_0 - E_n$. Hence $\tau_1 - E_n$ being relatively oscillatory with respect to $\tau_0 - E_n$ now follows using our comparison theorem for Wronskians (cf.\ Theorem~\ref{thm:scw}). For the second statement, we first note that our conditions imply $$ q_0(x) \leq q_1(x) < q_0(x) + \frac{1}{(\kappa_n+\eps)x^2} $$ near infinity, and then one proceeds as before. \end{proof} \noindent We note that even the second order term was also computed in \cite{kms}. So we obtain: \begin{theorem} Assume \be \lim_{x\rightarrow \infty} \kappa_n x^2 (q_1 (x) - q_0(x)) = 1 \ee in addition to the assumptions in Theorem~\ref{thm:app}. Then $E_n$ is an accumulation point of eigenvalues if \be \liminf_{x \to\infty} \log^2(x) (\kappa_n x^2 (q_1 (x) - q_0(x)) -1) > 1, \ee and $E_n$ is not an accumulation point of eigenvalues if \be \limsup_{x \to\infty} \log^2(x) (\kappa_n x^2 (q_1 (x) - q_0(x)) -1) < 1. \ee \end{theorem} \begin{proof} Similarly to that of Theorem~\ref{thm:app}, except, one now uses \cite[Thm.~2]{kms}. \end{proof} \noindent The main argument in \cite{kms} is a perturbation argument for the difference of Pr\"ufer angles for the solution of the unperturbed and the perturbed equation. This corresponds exactly to calculating the asymptotics of the Pr\"ufer angle of the Wronskian. \section{More on Pr\"ufer angles and the case of regular operators} \label{sec:proofreg} Now let us suppose that $\tau_{0,1}$ are both regular at $a$ and $b$ with boundary conditions \be \label{bc} \cos(\alpha) f(a) - \sin(\alpha) p(a)f'(a) =0 , \quad \cos(\beta) f(b) - \sin(\beta) p(b)f'(b) =0. \ee Hence we can choose $\psi_\pm(\lam,x)$ such that $\psi_-(\lam,a)=\sin(\alpha)$, $p(a)\psi_-'(\lam,a)=\cos(\alpha)$ respectively $\psi_+(\lam,b)=\sin(\beta)$, $p(b)\psi_+'(\lam,b)=\cos(\beta)$. In particular, we may choose \be \label{normtha} \theta_-(\lam,a) =\alpha \in [0,\pi), \quad -\theta_+(\lam,b) = \pi -\beta \in [0,\pi). \ee Next we introduce \be \tau_\eps = \tau_0 + \frac{\eps}{r} (q_1 -q_0) \ee and investigate the dependence with respect to $\eps\in[0,1]$. If $u_\eps$ solves $\tau_\eps u_\eps = 0$, then the corresponding Pr\"ufer angles satisfy \be \dot{\theta}_\eps(x) = -\frac{W_x(u_\eps, \dot{u}_\eps)}{\rho_\eps^2(x)}, \ee where the dot denotes a derivative with respect to $\eps$. \begin{lemma} \label{prwpsiepsdot} We have \be W_x(\psi_{\eps,\pm}, \dot{\psi}_{\eps,\pm}) = \left\{ \begin{array}{l} \int_x^b (q_0(t)-q_1(t)) \psi_{\eps,+}(t)^2 dt \\ -\int_a^x (q_0(t)-q_1(t)) \psi_{\eps,-}(t)^2 dt \end{array} \right. , \ee where the dot denotes a derivative with respect to $\eps$ and $\psi_{\eps,\pm}(x)= \psi_{\eps,\pm}(0,x)$. \end{lemma} \begin{proof} Integrating (\ref{dwr}) we obtain \be W_x(\psi_{\eps,\pm}, \psi_{\ti{\eps},\pm}) = (\ti{\eps}-\eps) \left\{ \begin{array}{l} \int_x^b (q_0(t)-q_1(t)) \psi_{\eps,+}(t) \psi_{\ti{\eps},+}(t) dt, \\ - \int_a^x (q_0(t)-q_1(t)) \psi_{\eps,-}(t) \psi_{\ti{\eps},-}(t) dt. \end{array} \right. \ee Now use this to evaluate the limit \be \lim_{\ti{\eps} \to \eps}W_x \Big(\psi_{\eps,\pm}, \frac{\psi_{\pm,\eps} - \psi_{\ti{\eps},\pm}}{\eps-\ti{\eps}} \Big). \ee \end{proof} \noindent Denoting the Pr\"ufer angles of $\psi_{\eps,\pm}(x)= \psi_{\eps,\pm}(0,x)$ by $\theta_{\eps,+}(x)$, this result implies for $q_0-q_1\ge 0$, \begin{align} \nn \dot{\theta}_{\eps,+}(x) &= -\frac{\int_x^b (q_0(t)-q_1(t)) \psi_{\eps,+}(t)^2 dt}{\rho_{\eps,+}(x)^2} \le 0, \\ \label{thetadot} \dot{\theta}_{\eps,-}(x) &= \frac{\int_a^x (q_0(t)-q_1(t)) \psi_{\eps,-}(t)^2 dt}{\rho_{\eps,-}(x)^2} \ge 0, \end{align} with strict inequalities if $q_1 \not\equiv q_0$. Now we are ready to investigate the associated operators $H_0$ and $H_1$. In addition, we will choose the same boundary conditions for $H_\eps$ as for $H_0$ and $H_1$. \begin{lemma} \label{lemevheps} Suppose $q_0-q_1\ge 0$ (resp.\ $q_0-q_1\leq 0$). Then the eigenvalues of $H_\eps$ are analytic functions with respect to $\eps$ and they are decreasing (resp.\ increasing). \end{lemma} \begin{proof} First of all the Pr\"ufer angles $\theta_{\eps,\pm}(x)$ are analytic with respect to $\eps$ since $\tau_\eps$ is by a well-known result from ordinary differential equations (see e.g., \cite[Thm.~13.III]{wa}). Moreover, $\lam\in\sig(H_\eps)$ is equivalent to $\theta_{\eps,+}(a)\equiv \alpha \mod \pi$ (resp.\ $\theta_{\eps,-}(b) \equiv \beta \mod \pi$), where $\alpha$ (respectively $\beta$) generates the boundary condition (cf.\ (\ref{bc})). \end{proof} \noindent In particular, this implies that $\dim\Ran P_{(-\infty,\lam)} (H_\eps)$ is continuous from below (resp. above) in $\eps$ if $q_0-q_1\ge 0$ (resp.\ $q_0-q_1\leq 0$). Now we are ready for the \begin{proof}[Proof of Theorem~\ref{thm:reg}] It suffices to prove the result for $\#(\psi_{0,+}, \psi_{1,-})$. Again we can assume $\lam_0=\lam_1=0$ without restriction. We split $q_0-q_1$ according to $$ q_0 - q_1 = q_+ - q_-,\qquad q_+, q_- \geq 0, $$ and introduce the operator $\tau_- = \tau_0 - q_-/r$. Then $\tau_-$ is a negative perturbation of $\tau_0$ and $\tau_1$ is a positive perturbation of $\tau_-$. Furthermore define $\tau_\eps$ by $$ \tau_\eps = \begin{cases} \tau_0 + 2\eps (\tau_- - \tau_0), & \eps\in[0,1/2],\\ \tau_- + 2(\eps -1/2)(\tau_1 - \tau_-), & \eps\in[1/2,1]. \end{cases} $$ Let us look at $$ N(\eps)=\#(\psi_{0,+}, \psi_{\eps,-}) = \ceil{\Delta_\eps(b)/\pi} - \floor{\Delta_\eps(a)/\pi} -1, \quad \Delta_\eps(x)=\Delta_{\psi_{0,+}, \psi_{\eps,- }}(x) $$ and consider $\eps\in [0,1/2]$. At the left boundary $\Delta_\eps(a)$ remains constant whereas at the right boundary $\Delta_\eps(b)$ is increasing by Lemma~\ref{prwpsiepsdot}. Moreover, it hits a multiple of $\pi$ whenever $0\in\sig(H_\eps)$. So $N(\eps)$ is a piecewise constant function which is continuous from below and jumps by one whenever $0\in\sig(H_\eps)$. By Lemma~\ref{lemevheps} the same is true for $$ P(\eps) = \dim\Ran\, P_{(-\infty, 0)} (H_\eps) - \dim\Ran\, P_{(-\infty, 0]} (H_0) $$ and since we have $N(0)=P(0)$, we conclude $N(\eps)=P(\eps)$ for all $\eps\in[0,1/2]$. To see the remaining case $\eps=[1/2,1]$, simply replace increasing by decreasing and continuous from below by continuous from above. \end{proof} \section{Approximation by regular operators} \label{sec:appreg} Now we want to extend our results to singular operators. We will do so by approximating a singular operator by a sequence of regular ones following \cite[Chap.~14]{wdln}. Abbreviate in the following $L^2((c,d),r\,dx)$ as $L^2(c,d)$. Fix functions $u,v\in\dom(\tau)$ and pick $c_n\downarrow a$, $d_n\uparrow b$. Define $\ti{H}_n$ \be \label{deftiHm} \ti{H}_n: \begin{array}[t]{lcl} \dom(\ti{H}_n) &\to& L^2(c_n,d_n) \\ f &\mapsto& \tau f \end{array}, \ee where \be \dom(\tilde{H}_n) = \{ f \in L^2(c_n,d_n) | \begin{array}[t]{l} f, pf' \in AC(c_n,d_n), \, \tau f \in L^2(c_n,d_n),\\ W_{c_n}(u,f) = W_{d_n}(v,f) =0 \}. \end{array} \ee Take $H_n=\alpha \id \oplus \ti{H}_n \oplus\alpha \id$ on $L^2(a,b)=L^2(a,c_n)\oplus L^2(c_n,d_n)\oplus L^2(d_n, b)$, where $\alpha$ is a fixed real constant. Then we have the following result: \begin{lemma} \label{strconh} Suppose that either $H$ is limit point at $a$ or that $u = \psi_-(\lam_0)$ for some $\lam_0\in\R$ and similarly, that either $H$ is limit point at $b$ or $v=\psi_+(\lam_1)$ for some $\lam_1\in\R$. Then $H_n$ converges to $H$ in strong resolvent sense as $n\to\infty$. Furthermore, if $H$ is limit circle at $a$ (resp. $b$), we can replace $u$ (resp. $v$) with any function in $\dom(\tau)$, which generates the boundary condition. \end{lemma} \noindent However, strong resolvent convergence is not sufficient for our purpose here. In addition we will need the following result from \cite{sw} (see also \cite{wdln}). We give a slightly refined analysis which allows eigenvalues at the boundary of the spectral intervals and the possibility of infinite-dimensional projections. We remark that for a self-adjoint projector $P$ we have \be \dim\Ran(P) = \tr(P) = \|P\|_{\mathcal{J}^1}, \ee where $\|.\|_{\mathcal{J}^1}$ denotes the trace class norm. If $P$ is not finite-rank, all three numbers equal $\infty$. Then we have the following result (\cite[Lem.~2]{te}, see also \cite{sw}): \begin{lemma}\label{strecon} Let $A_n\to A$ in strong resolvent sense and suppose $\tr(P_{(\lam_0,\lam_1)}(A_n)) \le \tr(P_{(\lam_0,\lam_1)}(A))$. Then, \be \lim_{n\to\infty} \tr(P_{(\lam_0,\lam_1)}(A_n)) = \tr(P_{(\lam_0,\lam_1)}(A)), \ee and if $\tr(P_{(\lam_0,\lam_1)}(A))<\infty$, we have \be \lim_{n\to\infty} \| P_{(\lam_0,\lam_1)}(A_n) - P_{(\lam_0,\lam_1)}(A)\|_{\mathcal{J}^1} =0. \ee \end{lemma} \begin{proof} This follows from (see e.g.\ \cite[Lem.~5.2]{gst}) \be\label{liminfsrc} \tr(P_{(\lam_0,\lam_1)}(A)) \le \liminf_{n\to\infty} \tr(P_{(\lam_0,\lam_1)}(A_n)), \ee together with Gr\"umm's theorem (\cite[Thm.~2.19]{str}). \end{proof} \noindent The key result of Stolz and Weidmann is that this lemma is applicable if certain Weyl solutions $\psi_\pm(\lam)$ are used to generate the boundary conditions of $\ti{H}_n$. As already pointed out, the version below is slightly refined since it allows $\lam_0$, $\lam_1$ to be eigenvalues of $H$. \begin{lemma}[\cite{sw}]\label{lem:swres0} Suppose $[\lam_0, \lam_1]\cap\sigma_{ess}(H) = \emptyset$ and let $H_n$ be defined as in (\ref{deftiHm}) with $u = \psi_-(\lam_-)$, $v=\psi_+(\lam_+)$ and $\lam_\pm\in[\lam_0,\lam_1]$. Then, \be \tr(P_{(\lam_0, \lam_1)}(\ti{H}_n)) \le \tr(P_{(\lam_0, \lam_1)}(H)). \ee \end{lemma} \begin{proof} Abbreviate $P=P_{(\lam_0, \lam_1)}(H)$, $P_n=P_{(\lam_0, \lam_1)}(\ti{H}_n)$. For $\ti{\psi}_1, \dots, \ti{\psi}_k \in \Ran P_n$ being eigenfunctions of $\ti{H}_n$, construct $$ \psi_j (x) = \left\{\begin{array}{cl} \gam_{j,u} u(x), & x < c_n,\\ \ti{\psi}_j(x), & c_n \leq x \leq d_n,\\ \gam_{j,v} v(x), & x > d_n, \end{array} \right. $$ where $\gam_{j,u}$, $\gam_{j,v}$ are chosen such that $\psi_j$ and $p \psi_j'$ are continuous. A computation shows that $$ \|(H - \frac{\lam_1 + \lam_0}{2}) \psi\| < \frac{\lam_1 - \lam_0}{2} \|\psi\| $$ for any $\psi$ in the linear span of the $\psi_j$'s, which yields the first result. \end{proof} \noindent This version is sufficient to give an alternative proof for the main theorem in \cite{gst}. \begin{proof}[Proof of Theorem~\ref{thm:gst}] Approximate $H_0$ by regular operators $H_n$ defined as in (\ref{deftiHm}) with $u=\psi_{0,-}(\lam_0)$, $v=\psi_{0,+}(\lam_1)$. Denote by $\psi_{0,\pm}^n(\lam)$ the solutions of the approximating problems. Then, by construction $\psi_{0,-}^n(\lam_0,x)=\psi_{0,-}(\lam_0,x)$ respectively $\psi_{0,+}^n(\lam_1,x)=\psi_{0,+}(\lam_1,x)$ for $x\in(c_n,d_n)$ and Theorem~\ref{thm:reg} in the special case $H_1=H_0$ implies $$ \tr(P_{(\lam_0, \lam_1)}(\ti{H}_n)) = \#_{(c_n,d_n)}(\psi_{0,-}(\lam_0),\psi_{0,+}(\lam_1)). $$ Letting $n\to\infty$ the left-hand side converges to $\tr(P_{(\lam_0, \lam_1)}(H_0))$ by the first part of Lemma~\ref{lem:swres0}. Hence the right-hand side converges as well and, according to Definition~\ref{def:wsf}, is given by $\#(\psi_{0,\mp}(\lam_0),\psi_{0,\pm}(\lam_1))$. \end{proof} \noindent However, the proof of our Theorem~\ref{thm:wronskzeros} requires some further extensions. In fact, in \cite{sw} Stolz and Weidmann point out that the Weyl functions of a different operator $\ti{H}$ will also do, as long as $\ti{H}$ is not too far away from $H$. Again the version below is slightly improved to allow for some border line cases. \begin{lemma}[\cite{sw}]\label{lem:swres1} Suppose $[\lam_0,\lam_1]\cap\sig_{ess}(H)=\emptyset$. Let $\ti{\tau} = \tau + \ti{q}$, where $\ti{q}$ is bounded, and pick the same boundary conditions for $\ti{H}$ as for $H$ (if any). Abbreviate \be Q_a = [\liminf_{x\to a} \ti{q}(x), \limsup_{x\to a} \ti{q}(x)], \quad Q_b = [\liminf_{x\to b}\ti{q}(x), \limsup_{x\to b} \ti{q}(x)], \ee and choose $\lam_-$ such that one of following conditions holds: \begin{enumerate} \item $\lam_- - Q_a \subseteq (\lam_0,\lam_1)$, or \item $\lam_- - Q_a \subseteq [\lam_0,\lam_1)$ and $\ti{q} \le \lam_- - \lam_0$ near $a$, or \item $\lam_- - Q_a \subseteq (\lam_0,\lam_1]$ and $\ti{q} \ge \lam_- - \lam_1$ near $a$. \end{enumerate} Similarly, choose $\lam_+$ to satisfy one of these conditions with $a$ replaced by $b$. Then, $H_n$ defined as in (\ref{deftiHm}) with $u = \ti{\psi}_-(\lam_-)$, $v=\ti{\psi}_+(\lam_+)$, satisfies \be \limsup_{n\to\infty} \tr(P_{(\lam_0, \lam_1)}(\ti{H}_n)) \le \tr(P_{(\lam_0, \lam_1)}(H)). \ee Furthermore, if we just require \be \lam_- - Q_a \subseteq [\lam_0,\lam_1], \qquad \lam_+ - Q_b \subseteq [\lam_0,\lam_1], \ee we at least have \be \limsup_{n\to\infty} \tr(P_{[\lam_0, \lam_1]}(\ti{H}_n)) \le \tr(P_{[\lam_0, \lam_1]}(H)). \ee \end{lemma} \begin{proof} Since any of our three conditions implies $|\lam_- -\frac{\lam_1+\lam_0}{2} - \ti{q}(x)| \le \frac{\lam_1-\lam_0}{2}$ for $x$ sufficiently close to $a$ and similarly $|\lam_+ -\frac{\lam_1+\lam_0}{2} - \ti{q}(x)| \le \frac{\lam_1-\lam_0}{2}$ for $x$ sufficiently close to $b$, we can proceed as in the previous lemma to prove the first claim. For the second claim choose $n$ sufficiently large such that $|\lam_\pm - \ti{q}(x) - 2^{-1}(\lam_1+\lam_0)| \le (2^{-1}(\lam_1-\lam_0) + \eps)$ for $xd_n$. Then, with the same argument as in the previous lemma, we have $$ \|(H - \frac{\lam_1 + \lam_0}{2}) \psi\| \le (\frac{\lam_1 - \lam_0}{2}+ \eps) \|\psi\| $$ and hence the second claim follows. \end{proof} \noindent Since our results involve projections to half-open intervals, we need one further step. \begin{lemma}\label{lem:swres2} Suppose $[\lam_0,\lam_1]\cap\sig_{ess}(H)=\emptyset$. Let $\ti{\tau} = \tau + \ti{q}$, where $\lim_{x\to a} \ti{q}(x)=0$ if $a$ is singular and $\lim_{x\to b} \ti{q}(x)=0$ if $b$ is singular. Furthermore, pick the same boundary conditions for $\ti{H}$ as for $H$ (if any). Define $H_n$ as in (\ref{deftiHm}) with $u = \ti{\psi}_-(\lam)$, $v=\ti{\psi}_+(\lam)$, $\lam\in\{\lam_0,\lam_1\}$. If $\lam=\lam_1$ and $\ti{q}\leq 0$ (near $a$ and $b$), then \be \lim_{n\to\infty} \tr(P_{(\lam_0,\lam_1]}(\ti{H}_n)) = \tr(P_{(\lam_0,\lam_1]}(H)), \ee and if $\lam=\lam_0$ and $\ti{q}\geq 0$, then \be \lim_{n\to\infty} \tr(P_{[\lam_0,\lam_1)}(\ti{H}_n)) = \tr(P_{[\lam_0,\lam_1)}(H)). \ee \end{lemma} \begin{proof} Without restriction, we just proove the first claim. For a sufficiently small $\eps>0$ we still have $[\lam_0,\lam_1+\eps]\cap\sig_{ess}(H)=\emptyset$ and thus by the previous lemma $$ \lim_{n\to\infty} \tr(P_{(\lam_0,\lam_1+\eps)}(\ti{H}_n)) = \tr(P_{(\lam_0,\lam_1+\eps)}(H)) $$ and $$ \lim_{n\to\infty} \tr(P_{(\lam_1,\lam_1+\eps)}(\ti{H}_n)) = \tr(P_{(\lam_1,\lam_1+\eps)}(H)). $$ Hence the result follows from $P_{(\lam_0,\lam_1]} = P_{(\lam_0,\lam_1+\eps)}-P_{(\lam_1,\lam_1+\eps)}$. \end{proof} \noindent Finally, we note: \begin{lemma} \label{lem:csf} Suppose $\tau_j u_j = \lam_j u_j$, $j=0,1$, with $q_1\le q_0$ near singular endpoints and $\lam_0<\lam_1$. If $\tau_j u_{j,n} = \lam_{j,n} u_{j,n}$, where $\lam_{j,n} \to \lam_j$ and $u_{j,n} \to u_j$, uniformly on compact sets $[c,d]\subseteq(a,b)$, then \be \liminf_{n\to\infty} \#(u_{0,n},u_{1,n}) \geq \#(u_0,u_1). \ee \end{lemma} \begin{proof} Let $N\in\N_0$ be any finite number with $N \le \#(u_0,u_1)$. Choose a compact set $[c,d]$ containing $N$ sign flips of $W(u_0,u_1)$. Then, for $n$ sufficiently large, $W(u_{0,n},u_{1,n})$ has $N$ sign flips in $[c,d]$. Hence $\#(u_{0,n},u_{1,n}) \geq \#_{(c,d)}(u_{0,n},u_{1,n}) = N$ and the claim follows. \end{proof} \noindent Now, we are ready for the \begin{proof}[Proof of Theorem~\ref{thm:wronskzeros}] It suffices to show the $\#(\psi_{1,-}(\lam_j), \psi_{0,+}(\lam_j))$ case. Define $\ti{H}_{j,n}$, $j=0,1$, as in (\ref{deftiHm}) with $u=\psi_{1,-}(\lam_0)$ and $v=\psi_{0,+}(\lam_0)$. Denote by $\psi_{j,\pm}^n(\lam)$, $j=0,1$, the solutions of the approximating problems. Then, by Theorem~\ref{thm:reg}, $$ \tr(P_{(-\infty,\lam_0)}(\ti{H}_{1,n})) - \tr(P_{(-\infty,\lam_0]}(\ti{H}_{0,n})) = \#_{(c_n,d_n)}(\psi_{1,-}^n(\lam_0),\psi_{0,+}^n(\lam_0)) $$ and we need to investigate the limits as $n\to\infty$. First of all $\psi_{1,-}^n(\lam_0,x)= \psi_{1,-}(\lam_0,x)$, $\psi_{0,+}^n(\lam_0,x)= \psi_{0,+}(\lam_0,x)$ for $x\in(c_n,d_n)$ implies \begin{align*} \lim_{n\to\infty} \#_{(c_n,d_n)}(\psi_{1,-}^n(\lam_0),\psi_{0,+}^n(\lam_0)) &= \lim_{n\to\infty} \#_{(c_n,d_n)}(\psi_{1,-}(\lam_0),\psi_{0,+}(\lam_0))\\ &= \#(\psi_{1,-}(\lam_0),\psi_{0,+}(\lam_0)). \end{align*} This takes care of the number of sign flips and it remains to look at the spectral projections. Let $\lam_0<\sig_{ess}(H_0)$, that is $H_0$ and hence also $H_1$ are bounded from below. Replacing $P_{(-\infty,\lam_0)}(H_j)$ by $P_{(\lam,\lam_0)}(H_j)$ with some $\lam$ below the spectrum of both $H_0$ and $H_1$ we infer from Lemma~\ref{lem:swres2} $$ \lim_{n\to\infty} \tr(P_{(-\infty,\lam_0)}(\ti{H}_{1,n})) = \tr(P_{(-\infty,\lam_0)}(H_1)) $$ and $$ \lim_{n\to\infty} \tr(P_{(-\infty,\lam_0]}(\ti{H}_{0,n})) = \tr(P_{(-\infty,\lam_0]}(H_0)). $$ This settles the first claim (\ref{eq:wronskzeros0}), where $\lam_0<\sig_{ess}(H_0)$. For the second claim (\ref{eq:wronskzeros1}), we first note that $\tau_1-\lam_0$ is relatively nonoscillatory with respect to $\tau_0-\lam_0$ by Lemma~\ref{lem:nonoscingap}. Next note that $\psi_{0,+}^n(\lam_1,.) \rightarrow \psi_{0,+}(\lam_1,.)$ pointwise, since \be \psi_{0,+}^n(\lam_1,x) = c_0(\lam_1,x) + m_{0,+}^n(\lam_1) s_0(\lam_1,x), \ee where $c_0(\lam,x)$, $s_0(\lam,x)$ is a fundamental system of solutions for $\tau_0-\lam$, and $m_{0,+}^n(\lam)$ are the corresponding Weyl--Titchmarsh $m$-functions. Next, strong resolvent convergence implies convergence of the Weyl $m$-function and hence uniform convergence of $\psi_{0,+}^n(\lam_1,x)\to \psi_{0,+}(\lam_1,x)$ on compact sets. Clearly the same applies to $\psi_{1,-}^n(\lam_1,x) \to \psi_{1,-}(\lam_1,x)$. Thus, by Lemma~\ref{strecon}, Lemma~\ref{lem:swres2}, and Lemma~\ref{lem:csf}, \begin{align} \label{eq:wrprin1} \tr(P_{[\lam_0, \lam_1)}(H_1)) &- \tr(P_{(\lam_0, \lam_1]}(H_0))\\ \nn &\geq \#(\psi_{1,-} (\lam_1), \psi_{0,+} (\lam_1)) - \#(\psi_{1,-} (\lam_0), \psi_{0,+} (\lam_0)). \end{align} Repeating the argument with $u=\psi_{1,-}(\lambda_1)$ and $v=\psi_{0,+}(\lambda_1)$ shows that \begin{align} \label{eq:wrprin2} \tr(P_{[\lam_0, \lam_1)}(H_1)) &- \tr(P_{(\lam_0, \lam_1]}(H_0))\\ \nn &\leq \#(\psi_{1,-} (\lam_1), \psi_{0,+} (\lam_1)) - \#(\psi_{1,-} (\lam_0), \psi_{0,+} (\lam_0)). \end{align} This proves the second claim. \end{proof} \section{Approximation in trace norm} \label{sec:apptr} Now we begin with an alternative approach toward singular differential operators by proving the case where $q_1-q_0$ has compact support. The next lemma would follow from Theorem~\ref{thm:wronskzeros}, but to demonstrate that this approach is independent of the last, we will provide an alternative proof. \begin{lemma} \label{lem:compactcase} Let $H_j$, $j=0,1$, be Sturm--Liouville operators on $(a,b)$ associated with $\tau_j$, and suppose that $r^{-1}(q_1-q_0)$ has support in a bounded interval $(c,d)\subseteq (a,b)$, where $a)}{W(\psi_{\eps,-}(z),\psi_{\eps,+}(z))}, $$ where $x_< = \min(x,y)$, $y_>=\max(x,y)$, the Green's function of $H_\eps$. As pointed out in the proof of the previous lemma, $\psi_{\eps,\pm}(z,x)$ is analytic with respect to $\eps$ and hence a simple estimate shows $$ \int_a^b \int_a^b |G_\eps(z,x,y)|^2 |r(y)^{-1}(q_1(y)-q_0(y))| r(x)dx\, r(y)dy \le C(z)^2 $$ for $\eps\in[0,1]$, which establishes the first claim. Furthermore, a straightforward calculation (using (\ref{dwr})) shows $$ G_{\eps'}(z,x,y) = G_\eps(z,x,y) + (\eps-\eps') \int_a^b G_{\eps'}(z,x,t) r^{-1}(t)(q_1(t)-q_0(t)) G_\eps(z,t,y) r(t) dt. $$ (Note that this does not follow from the second resolvent identity unless $r^{-1}(q_1-q_0)$ is relatively bounded with respect to $H_0$.) Hence, $R_{H_{\eps'}}(z) - R_{H_\eps}(z)$ can be written as the product of two Hilbert--Schmidt operators whose norm can be estimated by the first claim, \be \| R_{H_{\eps'}}(z) - R_{H_\eps}(z) \|_{\mathcal{J}_1} \le |\eps'-\eps| C(z)^2. \ee Thus $\eps \mapsto \xi(H_\eps, H_0)$ is continuous by Lemma~\ref{lem:Heps}. The rest follows from (\ref{dimxi}). \end{proof} \noindent Now we come to the \begin{proof}[Proof of Theorem~\ref{thmsing}] We first assume that we have compact support near one endpoint, say $a$. Furthermore, abbreviate $V=r^{-1}(q_0-q_1)$ which satisfies Hypothesis~\ref{hyp:h0h1} (and thus also Hypothesis~\ref{hyp:h0v}). Define by $K_\eps$ the operator of multiplication by $\chi_{(a,b_\eps]}$ with $b_\eps \uparrow b$ as $\eps\uparrow 1$. Then $K_\eps$ satisfies the assumptions of Lemma~\ref{lem:resolvconv}. Introduce $H_\eps = H_0 - K_\eps V$, and denote by $\psi_{\eps,-}(\lam,x)$ the corresponding solutions satisfying the boundary condition at $a$. By Lemma~\ref{lem:resolvconv} we have $\xi(., H_\eps, H_0) \to \xi(., H_1, H_0)$ as $\eps\to 1$ in $L^1(\R,(\lam^2+1)^{-1}d\lam)$. Moreover, $H_\eps \to H_1$ in (trace) norm resolvent sense and hence $\lam\in\rho(H_1)$ implies $\lam\in\rho(H_\eps)$ for $\eps$ sufficiently close to $1$. Since $\xi(\lam, H_\eps, H_0) \in \Z$ is constant near every $\lam\in\R\cap\rho(H_0)\cap\rho(H_\eps)$, we must have $\xi(\lam, H_\eps, H_0)= \xi(\lam, H_1, H_0)$ for $\eps \ge \eps_0$ with some $\eps_0$ sufficiently close to $1$. Now let us turn to the Wronskians. We first prove the $\#(\psi_{1,-}(\lam), \psi_{0,+}(\lam))$ case. By Lemma~\ref{lem:compactxi} we know $$ \xi(\lambda, H_\eps, H_0) = \#(\psi_{\eps,-}(\lam),\psi_{0,+}(\lam) $$ for every $\eps<1$. Concerning the right-hand side observe that $$ W_x(\psi_{\eps,-}(\lam), \psi_{0,+}(\lam)) = W_x(\psi_{1,-}(\lam), \psi_{0,+}(\lam)) $$ for $x \le b_\eps$ and that $W_x(\psi_{\eps,-}(\lam), \psi_{0,+}(\lam))$ is constant for $x \ge b_\eps$. This implies that for $\eps\ge \eps_0$ we have \begin{align*} \xi(\lam, H_1, H_0) &= \xi(\lam, H_\eps, H_0) = \#(\psi_{\eps,-}(\lam), \psi_{0,+}(\lam))\\ & = \#_{(a,b_\eps)}(\psi_{\eps,-}(\lam), \psi_{0,+}(\lam)) = \#_{(a,b_\eps)}(\psi_{1,-}(\lam), \psi_{0,+}(\lam)). \end{align*} In particular, the last item $\#_{(a,b_\eps)}(\psi_{1,-}(\lam), \psi_{0,+}(\lam))$ is eventually constant and thus has a limit which, by Definition~\ref{def:wsf}, is $\#(\psi_{1,-}(\lam), \psi_{0,+}(\lam))$. For the corresponding $\#(\psi_{1,+}(\lam), \psi_{0,-}(\lam))$ case one simply exchanges the roles of $H_0$ and $H_1$. Hence the result holds if the perturbation has compact support near one endpoint. Now one repeats the argument to remove the compact support assumption near the other endpoint as well. \end{proof} \section{Appendix: Some facts on the spectral shift function} \label{sec:xi} In this appendix we collect some facts on Krein's spectral shift function which are of relevance to us. Most results are taken from \cite{yafams} (see also \cite{sro} for an easy introduction). Two operators $H_0$ and $H_1$ are called resolvent comparable, if \be\label{eq:resolvcond} R_{H_1}(z) - R_{H_0}(z) \ee is trace class for one $z\in\rho(H_1)\cap\rho(H_0)$. By the first resolvent identity \eqref{eq:resolvcond} then holds for all $z\in\rho(H_1)\cap\rho(H_0)$. \begin{theorem}[Krein \cite{krein}]\label{thm:shifttheorem} Let $H_1$ and $H_0$ be two resolvent comparable self-adjoint operators, then there exists a function \be \xi(\lam, H_1, H_0) \in L^1(\R, (\lambda^2 + 1)^{-1}d\lam) \ee such that \be\label{eq:traceformula} \tr(f(H_1) - f(H_0)) = \int_{-\infty}^\infty \xi(\lam, H_1, H_0) f'(\lambda) d\lam \ee for every smooth function $f$ with compact support. \end{theorem} Note: Equation \eqref{eq:traceformula} holds in fact for a much larger class of functions $f$. See \cite[Thm.~8.7.1]{yafams} for this and a proof of the last theorem. The function $\xi(\lam) = \xi(\lam, H_1, H_0)$ is called Krein's spectral shift function and is unique up to an additive constant. Moreover, $\xi(\lam)$ is constant on every interval $(\lam_0, \lam_1) \subset \rho(H_0)\cap\rho(H_1)$. Hence, if $\dim\Ran P_{(\lam_0, \lam_1)}(H_j) < \infty$, $j = 0,1$, then $\xi(\lam)$ is a step function and \be \label{dimxi} \dim \Ran P_{(\lam_0, \lam_1)} (H_1) - \dim \Ran P_{(\lam_0, \lam_1)} (H_0) = \lim_{\eps\downarrow0} \Big( \xi(\lam_1-\eps) - \xi(\lam_0+\eps) \Big). \ee This formula explains the name spectral shift function. Before investigating further the properties of the SSF, we will recall a few things about trace ideals (see for example \cite{str}). First, for $1\leq p <\infty$ denote by $\mathcal{J}^p$ the Schatten $p$-class, and by $\|.\|_{\mathcal{J}^p}$ its norm. We will use $\|.\|$ for the usual operator norm. Using $\|A\|_{\mathcal{J}^p} = \infty$ if $A \notin \mathcal{J}^p$, we have the following inequalities for all operators: \be \|A B\|_{\mathcal{J}^p} \leq \|A\|\|B\|_{\mathcal{J}^p}, \quad \|A B\|_{\mathcal{J}^1} \leq \|A\|_{\mathcal{J}^2} \|B\|_{\mathcal{J}^2}. \ee Furthermore, we will use the notation of $\mathcal{J}^p$-converges to denote convergence in the respective $\|.\|_{\mathcal{J}^p}$-norm. The following result from \cite[Thm~IV.11.3]{tracedet} will be needed. \begin{lemma}\label{lem:contrace} Let $p > 0$, $A \in \mathcal{J}^p$, $T_n \xrightarrow{s} T$, $S_n \xrightarrow{s} S$ be sequences of strongly convergent bounded linear operators in some separable Hilbert space, then \be \|T_n A S_n^\ast - T A S^\ast\|_{\mathcal{J}^p} \rightarrow 0. \ee \end{lemma} We will also need the following continuity result for $\xi$. It will also allow us to fix the unknown constant. \begin{lemma} \label{lem:Heps} Suppose $H_\eps$, $\eps\in[0,1]$, is a family of self-adjoint operators, which is continuous in the metric \be \rho(A,B) = \|R_A (z_0) - R_B(z_0)\|_{\mathcal{J}^1}, \ee for some fixed $z_0\in \C\backslash\R$ and abbreviate $\xi_\eps = \xi(H_\eps, H_0)$. Then there exists a unique choice of $\xi_\eps$ such that $\eps \mapsto \xi_\eps$ is a continuous map $[0,1] \rightarrow L^1(\R,(\lam^2+1)^{-1}d\lam)$ with $\xi_0 = 0$. If $H_\eps \ge \lam_0$ is bounded from below, we can also allow $z=\lam \in (-\infty,\lam_0)$. \end{lemma} \begin{proof} The first statement can be found in \cite[Lem.~8.7.5]{yafams}. To see the second statement, let $\lam < \lam_0$ and $|\lam - z| < \lam_0 - \lam$ for some $z\in \C\backslash\R$. Abbreviate $R_\eps(z)=R_{H_\eps}(z)$. Now using the first resolvent identity gives \begin{align*} \|R_\eps(z) - R_{\eps'}(z)\|_{\mathcal{J}^1} \leq & \|R_\eps(\lam) - R_{\eps'}(\lam)\|_{\mathcal{J}^1} \\ \nn & {} + |z - \lam| \|R_\eps(z)\| \|R_\eps(\lam) - R_{\eps'}(\lam)\|_{\mathcal{J}^1}\\ \nn & {} + |z-\lam| \|R_{\eps'}(\lam)\| \|R_\eps(z) - R_{\eps'}(z)\|_{\mathcal{J}^1} \end{align*} and our conditions imply $$ |z-\lam| \|R_{\eps'}(\lam)\| \leq \frac{|z-\lam|}{\lam_0 - \lam} < 1 $$ and thus $$ \|R_\eps(z) - R_{\eps'}(z)\|_{\mathcal{J}^1} \leq \frac{1 + \frac{|z - \lam|}{|\im(z)|}}{1 - \frac{|z-\lam|}{|\lam_0 - \lam|}} \|R_\eps(\lam) - R_{\eps'}(\lam)\|_{\mathcal{J}^1}, $$ from which the statement follows. \end{proof} Our final aim is to find some conditions which allow us to verify the assumptions of this lemma. To do this, we derive some properties of relatively bounded operators multiplied by strongly continuous families of operators. The key example for these operators will be multiplication operators by characteristic functions strongly converging to the identity operator. \begin{hypothesis} \label{hyp:h0v} Suppose $H_0$ and $V$ are self-adjoint such that: \begin{itemize} \item[(i)] $V$ is relatively bounded with respect to $H_0$ with $H_0$ bound less than one or \item[(i')] $H_0$ is bounded from below and $|V|$ is relatively form bounded with respect to $H_0$ with relative form bound less than one and \item[(ii)] $|V|^{1/2} R_{H_0}(z)$ is Hilbert--Schmidt for one (and hence for all) $z\in\rho(H_0)$. \end{itemize} \end{hypothesis} We recall that (i) means that $\dom(V) \supseteq \dom(H_0)$ and for some $a < 1$, $b \ge 0$, \be \|V \psi\| \leq a \|H_0 \psi\| + b \|\psi\|, \quad \forall \psi \in \dom(H_0). \ee and (i') means that $\fdom(V) \supseteq \fdom(H_0)$ and for some $a < 1$, $b \ge 0$, \be \label{defrfb} \spr{\psi}{|V| \psi} \leq a \spr{\psi}{H_0 \psi} + b \|\psi\|^2, \quad \forall \psi \in \fdom(H_0). \ee Note that in case (i') all sums of operators are meant as form sums. \begin{lemma}\label{lem:resolvconv} Let $\eps \ni[0,1] \to K_\eps$ be a strongly continuous family of bounded self-adjoint operators which commute with $V$ and $0= K_0\le K_\eps \le K_1 = \id$. Assume Hypothesis~\ref{hyp:h0v}. Then \be H_\eps = H_0 + K_\eps V \ee are self-adjoint operators such that the assumptions of Lemma~\ref{lem:Heps} hold. \end{lemma} \begin{proof} We will abbreviate $V_\eps = K_\eps V$ and $R_\eps(z)= R_{H_\eps}(z)$. We begin with the case where $V$ is relatively bounded with respect to $H_0$. By the Kato--Rellich Theorem (\cite[Thm.~X.12]{reedsimon2}) $H_\eps$ is well-defined and self-adjoint. Moreover, there is a $z$ with $\im(z) \neq 0$ such that $\| V R_0(z) \| \le a <1$. Hence $\| V_\eps R_0(z) \| \le a$ and a straightforward calculation using the second resolvent identity, $$ V R_\eps(z) = V R_0(z) (1 + V_\eps R_0(z))^{-1}, $$ shows that $$ \| V R_\eps(z) \| \le \frac{a}{1-a}. $$ Furthermore, again using the second resolvent identity, we have $$ |V|^{1/2} R_\eps(z) = |V|^{1/2} R_0(z) ( 1 - V_\eps R_\eps(z)), $$ which shows that $$ \| |V|^{1/2} R_\eps(z) \|_{\mathcal{J}_2} \le \frac{1}{1-a} \| |V|^{1/2} R_0(z) \|_{\mathcal{J}_2}. $$ To show $\mathcal{J}^1$-continuity at some fixed $\eps\in[0,1]$ observe $$ R_{\eps'}(z) - R_\eps(z) = R_{\eps'}(z) |V|^{1/2} \Big( (K_{\eps'} - K_\eps) \sgn(V) |V|^{1/2} R_\eps \Big), $$ where the first term $R_{\eps'}(z) |V|^{1/2} \subseteq (|V|^{1/2} R_{\eps'}(z^*))^*$ is uniformly $\mathcal{J}^2$-bounded in $\eps'$ by our previous argument and the second term $\mathcal{J}^2$-converges to $0$ as $\eps'$ to $\eps$ by Lemma~\ref{lem:contrace}. Now we come to the case where $V$ is relatively form bounded with respect to $H_0$. By the KLMN theorem (\cite[Thm.~X.17]{reedsimon2}), $H_\eps$ is self-adjoint since we have $$ |\spr{\psi}{V_\eps \psi}| = |\spr{|V|^{1/2}\psi}{K_\eps \sgn(V) |V|^{1/2}\psi}| \le \spr{\psi}{|V|\psi}, \quad \psi\in\fdom(V_\eps)= \fdom(V). $$ Moreover, using (\ref{defrfb}) we have $$ \| |V|^{1/2} R_\eps(-\lam)^{1/2} \|^2 \le a, \qquad \frac{b}{a}< \lam. $$ We abbreviate: $$ U = |V|^{1/2},\quad W = \sgn(V) |V|^{1/2}. $$ For $\lam>\frac{b}{a}$ we have (see \cite[Sec.~VI.3]{kato} or \cite[Thm.~II.12]{shqf}) \begin{align*} R_\eps(-\lam) &= R_0(-\lam)^{1/2} ( 1 +C_\eps)^{-1} R_0(-\lam)^{1/2}, \\ C_\eps &= (U R_0(-\lam)^{1/2})^\ast (K_\eps W R_0(-\lam)^{1/2}). \end{align*} Hence, a straightforward calculation shows \begin{align*} R_\eps(-\lam) &= R_0(-\lam) - (U R_0(-\lam))^* (1 +\ti{C}_\eps)^{-1} (K_\eps W R_0(-\lam)),\\ \ti{C}_\eps &= K_\eps W R_0(-\lam)^{1/2} (U R_0(-\lam)^{1/2})^\ast. \end{align*} By $\|\ti{C}_\eps\| \le a < 1$, one concludes that $(1 + \ti{C}_\eps)^{-1}$ exists as a bounded operator. Then \begin{align*} D_{\eps,\eps'}\psi &= (- \ti{C}_\eps(1 + \ti{C}_\eps)^{-1} - \ti{C}_{\eps'} (1 + \ti{C}_{\eps'})^{-1})\psi \\ \nn &= (C_{\eps'} - C_\eps)(1 + C_\eps)^{-1} \psi - C_{\eps'} D_{\eps,\eps'} \psi, \end{align*} where $$ D_{\eps,\eps'} = (1 + \ti{C}_\eps)^{-1} - (1 + \ti{C}_{\eps'})^{-1}. $$ Taking norms we obtain $$ \|D_{\eps,\eps'}\psi\| = \frac{1}{1 - a} \|(C_{\eps'} - C_\eps)(1 + C_\eps)^{-1} \psi\|, $$ where the last term converges to $0$ as $\eps'\to\eps$. This implies that $(1 + \ti{C}_\eps)^{-1}$ is strongly continuous. Now, we obtain for the difference of resolvents $$ R_\eps(-\lam) - R_{\eps'}(-\lam) = (U R_0(-\lam))^* ((1 +\ti{C}_\eps)^{-1} K_\eps - (1 +\ti{C}_{\eps'})^{-1} K_{\eps'})) ( W R_0(-\lam)) $$ which $\mathcal{J}^1$-converges to $0$ as $\eps\to\eps'$ by Lemma~\ref{lem:contrace} and by $U R_0(-\lam)$ and $W R_0(-\lam)$ being Hilbert--Schmidt. \end{proof} \noindent {\bf Acknowledgments.} We thank F.\ Gesztesy for several helpful discussions and hints with respect to the literature. In addition, we are indebted to the referee for constructive remarks. \begin{thebibliography}{XXXX} \bibitem{atroscj} K. Ammann and G. Teschl, {\em Relative oscillation theory for Jacobi operators}, in Proceedings of the 14th International Conference on Difference Equations and Applications, M. Bohner (ed) et al., 105--115, U\u{g}ur--Bah\c{c}e\c{s}ehir University Publishing Company, Istanbul, 2009. \bibitem{bi} M.Sh. Birman, {\em On the spectrum of singular boundary value problems}, AMS Translations (2) {\bf 53}, 23--80 (1966). \bibitem{ea} M.S.P. Eastham, {\em The spectral theory of periodic differential equations}, Scottish Academic Press, Edinburgh, 1973. \bibitem{gs} F. Gesztesy and B. Simon, {\em A short proof of Zheludev's theorem}, Trans. Am. Math. Soc. {\bf 335}, 329--340 (1993). \bibitem{gu} F. Gesztesy and M. \"Unal, {\em Perturbative oscillation criteria and Hardy-type inequalities}, Math. Nachr.\ {\bf 189}, 121--144 (1998). \bibitem{gst} F. Gesztesy, B. Simon, and G. Teschl, {\em Zeros of the Wronskian and renormalized oscillation Theory}, Am. J. Math. {\bf 118}, 571--594 (1996). \bibitem{tracedet} I. Gohberg, S. Goldberg, and N. Krupnik, {\em Traces and Determinants of Linear Operators}, Birkh\"auser, Basel, 2000. \bibitem{har1} P.~Hartman, {\em Differential equations with non-oscillatory eigenfunctions}, Duke Math.~J. {\bf 15}, 697--709 (1948). \bibitem{har2} P.~Hartman, {\em A characterization of the spectra of one-dimensional wave equations}, Am.~J.~Math. {\bf 71}, 915--920 (1949). \bibitem{har3} P.~Hartman and C.R.~Putnam, {\em The least cluster point of the spectrum of boundary value problems}, Am.~J.~Math. {\bf 70}, 849--855 (1948). \bibitem{kato} T. Kato, {\em Perturbation Theory for Linear Operators}, Springer, New York, 1966. \bibitem{kn} A. Kneser, {\em Untersuchungen \"uber die reellen Nullstellen der Integrale linearer Differentialgleichungen}, Math. Ann. {\bf 42}, 409--435 (1893). \bibitem{khr} S.V. Khryashchev, {\em Discrete spectrum for a periodic Schr\"odinger operator perturbed by a decreasing potential}, Operator Theory: Adv. and Appl. {\bf 46}, 109--114 (1990). \bibitem{krein} M.G. Krein, {\em Perturbation determinants and a formula for the traces of unitary and self-adjoint operators}, Sov. Math. Dokl. {\bf 3}, 707--710 (1962). \bibitem{kt2} H. Kr\"uger and G. Teschl, {\em Relative oscillation theory for Sturm--Liouville operators extended}, J. Funct. Anal. {\bf 254-6}, 1702--1720 (2008). \bibitem{kt3} H. Kr\"uger and G. Teschl, {\em Effective Pr\"ufer angles and relative oscillation criteria}, J. Diff. Eq. {\bf 245}, 3823--3848 (2008). \bibitem{lei} W. Leighton, {\em On self-adjoint differential equations of second order}, J.~London Math.~Soc. {\bf 27}, 37--47 (1952). \bibitem{reedsimon2} M. Reed and B. Simon, {Methods of Modern Mathematical Physics II. Fourier Analysis, Self-Adjointness}, Academic Press, New York, 1975. \bibitem{rb1} F. S. Rofe-Beketov, {\em A test for the finiteness of the number of discrete levels introduced into gaps of a continuous spectrum by perturbations of a periodic potential}, Soviet Math. Dokl. {\bf 5}, 689--692 (1964). \bibitem{rb2} F.S. Rofe-Beketov, {\em Spectral analysis of the Hill operator and its perturbations}, FunkcionalÕny\"i analiz {\bf 9}, 144--155 (1977) (Russian). \bibitem{rb3} F.S. Rofe-Beketov, {\em A generalisation of the Pr\"ufer transformation and the discrete spectrum in gaps of the continuous one}, Spectral Theory of Operators, 146--153, Baku, Elm, 1979 (Russian). \bibitem{rb4} F.S. Rofe-Beketov, {\em Spectrum perturbations, the Kneser-type constants and the effective masses of zones-type potentials}, Constructive Theory of Functions Õ84, 757--766, Sofia, 1984. \bibitem{rb5} F.S. Rofe-Beketov, {\em Kneser constants and effective masses for band potentials}, Sov. Phys. Dokl. {\bf 29}, 391--393 (1984). \bibitem{rbk} F.S. Rofe-Beketov and A.M. Kholkin, {\em Spectral analysis of differential operators. Interplay between spectral and oscillatory properties}, World Scientific, Hackensack, 2005. \bibitem{kms} K.M. Schmidt, \textit{Critical coupling constants and eigenvalue asymptotics of perturbed periodic Sturm--Liouville operators}, Commun. Math. Phys. 211, 465--485 (2000). \bibitem{kms2} K.M. Schmidt, \textit{Relative oscillation non-oscillation criteria for perturbed periodic Dirac systems}, J. Math. Anal. Appl. {\bf 246}, 591--607 (2000). \bibitem{kms3} K.M. Schmidt, \textit{An application of Gesztesy-Simon-Teschl oscillation theory to a problem in differential geometry}, J. Math. Anal. Appl. {\bf 261}, 61--71 (2001). \bibitem{shqf} B. Simon, {\em Quantum Mechanics for Hamiltonians Defined as Quadratic Forms}, Princeton University Press, Princeton, 1971. \bibitem{str} B. Simon, {\em Trace Ideals and Their Applications}, 2nd ed., Amer. Math. Soc., Providence, 2005. \bibitem{sro} B. Simon, \textit{Spectral Analysis of Rank One Perturbations and Applications}, Lecture notes from Vancouver Summer School in Mathematical Physics, August 10--14, 1993. \bibitem{sim} B. Simon, {\em Sturm oscillation and comparison theorems}, in Sturm--Liouville Theory: Past and Present (eds. W. Amrein, A. Hinz and D. Pearson), 29--43, Birkh\"auser, Basel, 2005. \bibitem{sw} G. Stolz and J. Weidmann, {\em Approximation of isolated eigenvalues of ordinary differential operators}, J. Reine und Angew. Math. {\bf 445}, 31--44 (1993). \bibitem{stu} J.C.F. Sturm, {\em M\'emoire sur les \'equations diff\'erentielles lin\'eaires du second ordre}, J.~Math.~Pures Appl., {\bf 1}, 106--186 (1836). \bibitem{tosc} G. Teschl, {\em Oscillation theory and renormalized oscillation theory for Jacobi operators}, J. Diff. Eqs. {\bf 129}, 532--558 (1996). \bibitem{toscd} G. Teschl, {\em Renormalized oscillation theory for Dirac operators}, Proc. Amer. Math. Soc. {\bf 126}, 1685--1695 (1998). \bibitem{tjac} G. Teschl, {\em Jacobi Operators and Completely Integrable Nonlinear Lattices}, Math. Surv. and Mon. {\bf 72}, Amer. Math. Soc., Rhode Island, 2000. \bibitem{te} G. Teschl, {\em On the approximation of isolated eigenvalues of ordinary differential operators}, Proc. Amer. Math. Soc. {\bf 136}, 2473--2476 (2008). \bibitem{troscd} G. Teschl, {\em Relative oscillation theory for Dirac operators}, in preparation. \bibitem{wa} W. Walter, {\em Ordinary Differential Equations}, Springer, New York, 1998. \bibitem{wd} J. Weidmann, {\em Zur Spektraltheorie von Sturm--Liouville--Operatoren}, Math. Z. {\bf 98}, 268--302 (1967). \bibitem{wd1} J. Weidmann, {\em Oszillationsmethoden f\"ur Systeme gew\"ohnlicher Differentialgleichungen}, Math. Z. {\bf 119}, 349--373 (1971). \bibitem{wdln} J. Weidmann, {\em Spectral Theory of Ordinary Differential Operators}, Lecture Notes in Mathematics, {\bf 1258}, Springer, Berlin, 1987. \bibitem{wd2} J. Weidmann, {\em Spectral theory of Sturm--Liouville operators; approximation by regular problems}, in Sturm--Liouville Theory: Past and Present (eds. W. Amrein, A. Hinz and D. Pearson), 29--43, Birkh\"auser, Basel, 2005. \bibitem{yafams} D.R. Yafaev, {\em Mathematical Scattering Theory: General Theory}, Amer. Math. Soc., Rhode Island, 1992. \bibitem{zet} A. Zettl, {\em Sturm--Liouville Theory}, Amer. Math. Soc., Rhode Island, 2005. \bibitem{zhe} V.A. Zheludev, {\em Perturbation of the spectrum of the one-dimensional self-adjoint Schr\"odinger operator with a periodic potential}, Topics in Mathematical Physics, Vol. 4: M.Sh. Birman (ed), 55--76, Consultants Bureau, New York, 1971. \end{thebibliography} \end{document}