\documentstyle{pspum-l} \newtheorem{theorem}{Theorem}[section] \newtheorem{lemma}[theorem]{Lemma} \newtheorem{prop}[theorem]{Proposition} \theoremstyle{definition} \newtheorem{definition}{Definition} \renewcommand{\thedefinition}{} \newtheorem{example}[theorem]{Example} \theoremstyle{remark} \newtheorem{remark}[theorem]{Remark} \numberwithin{equation}{section} \newcommand{\bt}{\begin{theorem}} \newcommand{\et}{\end{theorem}} \newcommand{\bl}{\begin{lemma}} \newcommand{\el}{\end{lemma}} \newcommand{\bc}{\begin{corollary}} \newcommand{\ec}{\end{corollary}} \newcommand{\bp}{\begin{prop}} \newcommand{\ep}{\end{prop}} \newcommand{\bd}{\begin{definition}} \newcommand{\ed}{\end{definition}} \newcommand{\be}{$$} \newcommand{\ee}{$$} \newcommand{\beq}{\begin{eqnarray}} \newcommand{\eeq}{\end{eqnarray}} \newcommand{\bpf}{\begin{pf}} \newcommand{\epf}{\end{pf}} \newcommand{\la}{\label} \newcommand{\ci}{\cite} \newfont{\msbm}{msbm10 }%blackboardbold \newfont{\msbms}{msbm7 } %blackboardbold subscript \newfont{\msbml}{msbm10 scaled\magstep 2} \newcommand{\bbr}{\mbox{$\mbox{\msbm R}$}} \newcommand{\bbn}{\mbox{$\mbox{\msbm N}$}} \newcommand{\bbi}{\mbox{$\mbox{\msbm I}$}} \newcommand{\bbc}{\mbox{$\mbox{\msbm C}$}} \newcommand{\bbk}{\mbox{$\mbox{\msbm K}$}} \newcommand{\bbe}{\mbox{$\mbox{\msbm E}$}} \newcommand{\bbz}{\mbox{$\mbox{\msbm Z}$}} \newcommand{\bbp}{\mbox{$\mbox{\msbm P}$}} \newcommand{\bbt}{\mbox{$\mbox{\msbm T}$}} \newcommand{\bbrs}{\mbox{$\mbox{\msbms R}$}} %for subscripts \newcommand{\bbns}{\mbox{$\mbox{\msbms N}$}} \newcommand{\bbis}{\mbox{$\mbox{\msbms I}$}} \newcommand{\bbcs}{\mbox{$\mbox{\msbms C}$}} \newcommand{\bbks}{\mbox{$\mbox{\msbms K}$}} \newcommand{\bbes}{\mbox{$\mbox{\msbms E}$}} \newcommand{\bbzs}{\mbox{$\mbox{\msbms Z}$}} \newcommand{\bbps}{\mbox{$\mbox{\msbms P}$}} \newcommand{\B}{\mbox{$\mbox{\msbm B}$}} \newcommand{\Bs}{\mbox{$\mbox{\msbms B}$}} %for subscripts \newcommand{\E}{\mbox{$\mbox{\msbm E}$}} \newcommand{\de}{\delta} \newcommand{\De}{\Delta} \newcommand{\ba}{\beta} \newcommand{\ga}{\gamma} \newcommand{\Ga}{\Gamma} \newcommand{\ch}{\chi} \newcommand{\si}{\sigma} \newcommand{\om}{\omega} \newcommand{\ka}{\kappa} \newcommand{\lb}{\lambda} \newcommand{\ze}{\zeta} \newcommand{\th}{\theta} \newcommand{\vp}{\varphi} \newcommand{\ph}{\phi} \newcommand{\ps}{\psi} \newcommand{\Ph}{\Phi} \newcommand{\La}{\Lambda} \newcommand{\ve}{\varepsilon } \newcommand{\e}{\rm e} \newcommand{\R}{{\cal R}} \newcommand{\I}{{\cal I}} \newcommand{\cH}{\cal H} \newcommand{\F}{{\cal F}} \newcommand{\K}{{\cal K}} \newcommand{\T}{{\cal T}} \newcommand{\cB}{{\cal B}} \newcommand{\A}{{\cal A}} \newcommand{\M}{{\cal M}} \newcommand{\X}{{\cal X}} \newcommand{\cV}{{\cal V}} \newcommand{\cW}{{\cal W}} \newcommand{\cL}{{\cal L}} \begin{document} \title[Random Schr\"odinger Operators] { Absolutely Continuous Spectrum in Random Schr\"odinger Operators} \author{ Abel Klein} \address{Department of Mathematics, University of California, Irvine, CA 92717-3875} \email{aklein@@math.uci.edu} % Don't type a period at the end; it will be supplied. \thanks{The author was supported in part by NSF Grant DMS-9208029.} \keywords{ Random Schr\"odinger operators, Anderson model, extended states, absolutely continuous spectrum, localization} % Math Subject Classifications \subjclass{Primary 82B44; Secondary 81Q10} \date{} \begin{abstract} The spectrum of the Anderson Hamiltonian $\;H_\lb=-\De +\lb V$ on the Bethe Lattice is absolutely continuous inside the spectrum of the Laplacian, if the disorder $\lb$ is sufficiently small. More precisely, given any closed interval $I$ contained in the interior of the spectrum of the (centered) Laplacian $\De$ on the Bethe lattice, for small disorder $H_\lb$ has purely absolutely continuous spectrum in $I$ with probability one (i.e., $\si_{ac}( H_\lb) \cap I = I$ and $\si_{pp}( H_\lb) \cap I =\si_{sc}( H_\lb) \cap I= \emptyset$ with probability one). The proof is discussed and regularity properties are proven for the spectral measures restricted to such intervals of absolute continuity. \end{abstract} \maketitle \section{Introduction} The Anderson Hamiltonian \ci{And} is the random Schr\"odinger operator $$H_\lb\;\;=\;\;{\textstyle\frac{1}{2}}\De\, +\,\lb V\;\;\;\mbox{on}\;\;\; \ell^2({\bbz^d})\,. \la{ha}$$ Here the (centered) Laplacian $\De$ is defined by $$(\De u)(x)\;\;=\;\;\sum_{y} {u(y)}\;,$$ where the sum runs over all nearest neighbors of $x\,$, and $V$ is a random potential, with $V(x)$, $x \in {{\bbz^d}}$, being independent, identically distributed random variables with common probability distribution $\mu$. The real parameter $\lb$ is called the {\em disorder}. The spectrum of the Hamiltonian $\, H_\lb$ is given by $$\sigma(H_\lb) = \sigma({\textstyle\frac{1}{2}}\Delta) + \lb\, \mbox{supp} \; \mu = [-d,d] +\lb\, \mbox{supp} \; \mu \la{erg}$$ with probability one, due to ergodic considerations \ci{P,CL}. For each choice of $V$ the spectrum of $H_\lb$ can be decomposed into pure point spectrum, $\sigma_{pp}(H_\lb)$, absolutely continuous spectrum, $\sigma_{ac}(H_\lb)$, and singular continuous spectrum, $\sigma_{sc}(H_\lb)$. Ergodicity also gives the existence of sets $\Sigma_{\lb,pp}\; , \; \Sigma_{\lb,ac}\; , \; \Sigma_{\lb,sc} \subset {{\bbr}}$ such that $\sigma_{pp}(H_\lb) = \Sigma_{\lb,pp} \; , \; \sigma_{ac}(H_\lb) = \Sigma_{\lb,ac}$ and $\sigma_{sc}(H_\lb) = \Sigma_{\lb,sc}$ with probability one \ci{KS1,CL}. In the physics literature (\ci{And,MT,T,AAT,AT,MS} and others) the following picture is given: In one and two dimensions, as long as the potential is random (i.e., $\lb \not= 0$), the model shows {\em exponential localization} (i.e., pure point spectrum with exponentially decaying eigenfunctions ). In three and more dimensions both localized and {\em extended states} (i.e., absolutely continuous spectrum) are expected for small disorder, with the energies of extended and localized states being separated by a {\em mobility edge}. In one dimension there are now mathematical proofs of exponential localization for any disorder (e.g., \ci{GMP,KS1,CKM,DK} and others). In the multidimensional case exponential localization has been proven for large disorder or low energy (e.g., \ci{FS,FMSS,DLS,SW,DK,AM,A,K3,Gf} and others); for small disorder there exist energies $E_{\lambda, loc}^\pm$, with $|E_{\lambda, loc}^\pm| > d$ and \mbox{$\lim_{\lambda \to 0} E_{\lambda, loc}^\pm= \pm d$}, such that $H_{\lambda}$ has pure point spectrum in \mbox{$(-\infty, E_{\lambda, loc}^- ) \cup ( E_{\lambda, loc}^+,\infty)$ \cite{Sp,FK,A,Sp2}.} Only localization in two dimensions for small disorder is still an open problem. But up to now there is no mathematical proof of the occurrence of absolutely continuous spectrum in the Anderson model on $\bbz^d$. The Bethe lattice (or Cayley tree), ${{\B}}$, is an infinite connected graph with no closed loops and a fixed degree (number of nearest neighbors) at each vertex (site or point). The degree is called the coordination number and the connectivity, $K$, is one less the coordination number ($K \ge 2$, so $\B$ is not the line $\bbr$) . The distance between two sites $x$ and $y$ will be denoted by $d(x,y)$ and is equal to the length of the shortest path connecting $x$ and $y$. The Anderson Hamiltonian on the Bethe lattice is given by (\ref{ha}) with ${{\B}}$ substituted for ${\bbz^d}$. It was first studied by Abou-Chacra, Anderson and Thouless \ci{AAT}; their self-consistent approximation for the study of localization becomes exact in the Bethe lattice. The resulting equations were further studied by Abou-Chacra and Thouless \ci{AT}, who argued that, on the Bethe lattice, for sufficiently low disorder there should be an energy at which localization breaks down, which converges to $\frac{K +1}{2}$ in the zero disorder limit. An outline of expected results was given by Kunz and Souillard \cite{KS}. Miller and Derrida \ci{MD} performed a weak disorder expansion inside the spectrum of the zero disorder Hamiltonian, and computed perturbatively the density of states and conducting properties corresponding to extended states. They also found the existence of an energy, which converges (from outside) to the edge $\sqrt{K}$ of the spectrum of ${1 \over 2}\De$ in the zero disorder limit, above which the density of states and conducting properties vanish to all orders in perturbation theory. The Bethe lattice model was also discussed by Mirlin and Fyodorov \ci{MF}. Ergodicity considerations still apply (see Acosta and Klein \cite[Appendix]{AK}), so the first equality in (\ref{erg}) and the following statements are still valid in the Bethe lattice, where we have $\sigma(\Delta) \,=\, [-2\sqrt{ K},2\sqrt{ K}]\,$. In particular, $$\sigma(H_\lb)\;\; =\;\;[-\sqrt{ K},\sqrt{ K}] +\lb\, \mbox{supp} \; \mu$$ with probability one. The nature of the spectrum for energies $E$ with $\sqrt{K} \le |E| \le \frac{K +1}{2}$, at low disorder, does not seem to have been properly discussed in the physics literature. The Abou-Chacra and Thouless calculations indicate that we do not have localization in these intervals for small disorder; Miller and Derrida's weak disorder expansions suggest that the corresponding states cannot be too extended", since they should not have conducting properties. So there should be continuous spectrum, but of which type? It could be absolutely continuous, but it would be of a different nature then the one inside the free spectrum. It is plausible that near $E=\pm \frac{K +1}{2}$ there is a transition from pure point spectum to either singular continuous spectrum or absolutely continuous spectrum without conducting properties", and that near $E=\pm \sqrt{K}$ there is another transition where the spectrum changes to absolutely continuous with conducting properties". If so, there is a {\it mobility interval} instead of a {\it mobility edge}! Localization for large disorder or low energies has only recently been proven by Aizenman and Molchanov \ci{AM}, and Aizenman \ci{A} proved localization for energies beyond $\frac{K +1}{2}$ at weak disorder, confirming half of Abou-Chacra and Thouless' prediction \ci{AT}. On the Bethe lattice we proved that the Anderson Hamiltonian has extended states'' for small disorder \cite{K1,K2}. More precisely, given any closed interval $I$ contained in the interior of the spectrum of ${1 \over 2}\De$ on the Bethe lattice, we proved that for small disorder $H_\lb$ has purely absolutely continuous spectrum in $I$ with probability one, and its integrated density of states is continuously differentiable on the interval. These results agree with Miller and Derrida's conclusions \ci{MD}. We have also studied the spreading of wave packets evolving under the Anderson Hamiltonian on the Bethe Lattice: for small disorder we showed that the averaged mean square distance travelled by a particle in a time $t$ \ grows as $t^2$ for large $t$ \cite{K4}. The precise results requires hypotheses about the single site potential probability distribution $\mu$: the localization results require $\mu$ to have a bounded density with respect to Lebesgue measure, while the extended states results call for $\mu$ to have a characteristic function $h(t)=\int {{\e}^{-itv}d\mu (v)}$ which is twice differentiable with bounded first and second derivatives on $(0,\infty)$. (True for any any probability distribution $\mu$ with a finite second moment (e.g., uniform, Gaussian or Bernoulli distributions) and for the Cauchy distribution). In this article we show that by requiring more differentiability of the characteristic function $h(t)$ we can get absolutely continuous spectrum with certain regularity properties, which will be expressed in terms of the spectral measures $\nu_{\lb,x},\; x \in \B$, given by $$d\nu_{\lb,x}(E) = \left\langle {\de_x,dP_\lb (E)\de_x} \right\rangle\;,$$ where $dP_\lb (E)$ is the spectral measure of the operator $H_\lb$. \bd Let $p \in \bbn$. A probability measure $\mu$ will be said to be $p$-admissible if its characteristic function $h(t)$ is $2p$-times differentiable on $(0,\infty)$, with all $2p$ derivatives bounded on $(0,\infty)$. (True for any any probability distribution $\mu$ with $\int |v|^{2p} d\mu (v) < \infty$). \ed \bt \la{main} Let $\;H_\lb$ be the Anderson Hamiltonian on the Bethe lattice with $\mu$ $p$-admissible, $p \in \bbn$. Then for any $E \in (0 , \sqrt K)$ there exists $\lb(E) > 0$, such that given any $\lb$ with$|\lb| < \lb(E)$, with probability one the spectrum of $\;H_\lb$ in $[-E, E]$ is purely absolutely continuous (i.e., $\Sigma_{\lb,ac} \cap [-E, E] = [-E, E]$ and $\Sigma_{\lb,pp} \cap [-E, E] = \Sigma_{\lb,sc} \cap [-E, E]=\emptyset$), the spectral measures $\nu_{\lb,x}$ being absolutely continuous in this interval with $\frac{d\nu_{\lb,x}}{dE} \in L^{2p}([-E, E])$ for all $x \in \B$. \et The $p=1$ case was proved in \cite{K1,K2}. The Green's function of $H_\lb$ is given by $$G_{\lb}\, (x,y;z)\;\;=\;\;\left\langle {\de_x,(H_{\lb} -z)^{-1}\de_y} \right\rangle$$ for $x,y \in {\B}$ and $z = E + i\eta$ with $E \in \bbr$, $\eta> 0$. For any $x \in \B$ and any potential $V$, $G_{\lb}\, (x,x; E +i\eta)$ is a continuous function of $(\lb, E,\eta) \in \bbr \times \bbr \times (0, \infty)$, so $\E (|G_{\lb}\, (x,x; E +i\eta)|^{2p})$ is also a continuous function of $(\lb, E,\eta) \in \bbr \times \bbr \times (0, \infty)$ for any $p \in \bbn$. Theorem \ref{main} will follow from the fact that we can let $\eta \downarrow 0$ inside the spectrum of $\frac{1}{2} \De$. \bt \la{cor} Let $\;H_\lb$ be the Anderson Hamiltonian on the Bethe lattice with $\mu$ $p$-admissible, $p \in \bbn$. Then for any $E \in (0 , \sqrt K)$ there exists $\lb(E) > 0$, such that for all $x \in \B$ the continuous function $$(\lb, E',\eta) \in (-\lb(E),\lb(E)) \times [-E,E] \times (0, \infty) \;\longrightarrow\; \E (|G_{\lb}\, (x,x; E' +i\eta)|^{2p} )$$ has a continuous extension to $(-\lb(E),\lb(E)) \times [-E,E] \times [0, \infty)$. In particular, $$\sup_{\lb;\, |\lb| < \lb(E)}\, \sup_{E';\,|E' | \le E} \, \sup_{\eta;\, 0 <\eta }\, \E (|G_{\lb}\, (x,x; E' +i\eta)|^{2p} )\;\;<\;\; \infty\;. \la{sup}$$ \et Theorem \ref{main} follows from (\ref{sup}). Let $E \in (0 , \sqrt K)$ and $\lb(E) > 0$ as in Theorem \ref{cor}, so (\ref{sup}) holds. For $|\lb| < \lb(E)$ and any $x \in \B$ we use Fubini's Theorem and Fatou's Lemma to obtain \beq \lefteqn{ \E \left( \liminf_{\eta \downarrow 0}\int_{-E}^E |G_{\lb}\, (x,x; E' +i\eta)|^{2p}\,dE' \right) \;\;\le \quad \quad \quad \quad \quad } \\ &&\quad \quad \liminf_{\eta \downarrow 0} \int_{-E}^E \E (|G_{\lb}\, (x,x; E' +i\eta)|^{2p} )\,dE'\;\;< \;\; \infty\;. \nonumber \eeq Thus we must have \beq \lefteqn{ \liminf_{\eta \downarrow 0}\int_{-E}^E ({\rm Im}\,G_{\lb}\, (x,x; E' +i\eta))^{2p}\,dE' \;\;\le \qquad \qquad \qquad} \\ && \quad \quad \quad \quad \liminf_{\eta \downarrow 0}\int_{-E}^E |G_{\lb}\, (x,x; E' +i\eta)|^{2p}\,dE' \;\;<\;\;\infty \nonumber \eeq with probability one. Since $G_{\lb}\, (x,x; E +i\eta)$ is the Stieltjes transform of the measure $\nu_{\lb,x}$, it follows (see \ci[Theorem 2.1]{simon}) that, with probability one, the spectral measures $\nu_{\lb,x}$ are absolutely continuous in the interval $(-E, E)$ with $\frac{d\nu_{\lb,x}}{dE} \in L^{2p}(-E, E)$ for all $x \in \B$. Theorem \ref{main} now follows. \section{Formulas} Let $\;H_\lb$ be the Anderson Hamiltonian on the Bethe lattice with $\mu$ $p$-admissible. We fix an arbitrary site in $\B$ which we will call the origin and denote by $0$. Given two nearest neighbors sites $x,y \in \B$, we will denote by ${\B}^{(x|y)}$ the lattice obtained by removing from $\B$ the branch emanating from $x$ that passes through $y$; if we do not specify which branch was removed we will simply write ${\B}^{(x)}$. Each vertex in ${\B}^{(x)}$ has degree $K +1$, with the single exception of $x$ which has degree $K$. Given $\La \subset \B$, we will use $\,H_{\lb, \La}$ to denote the operator $\,H_{\lb}$ restricted to $\ell^2 (\Lambda)$ with Dirichlet boundary conditions. The Green's function corresponding to $\,H_{\lb, \La}$ will be denoted by $$G_{\lb, \La}\, (x,y;z)\;\;=\;\;\left\langle {\de_x,(H_{\lb, \La} -z)^{-1}\de_y} \right\rangle$$ for $x,y \in {\La}$ and $z = E + i\eta$ with $E \in \bbr$, $\eta> 0$. We will write $H_{\lb}$, $H_{\lb}^{(x|y)}$ and $H_{\lb}^{(x)}$ for $H_{\lb, \Bs}$, $H_{\lb, {\Bs}^{(x|y)}}$ and $H_{\lb, {\Bs}^{(x)}}$, respectively. Similarly, we will use $G_{\lb}\, (x,y;z)$ for $G_{\lb, \Bs}\, (x,y;z)$ and $G_{\lb}\, (z)$, $G_{\lb}^{(x|y)}\, (z)$, $G_{\lb}^{(x)}\, (z)\;$ for $G_{\lb}\, (0,0;z)$, $G_{\lb,{\Bs}^{(x|y)}}\, (x,x;z)$, $G_{\lb,{\Bs}^{(x)}}\,(x,x;z)\;$, respectively. We start by stating some useful formulas (see \ci{K2} for more details). The next proposition is a consequence of the resolvent equation. \bp \la{res} For any $\lb \in \bbr$, $E \in \bbr$ and $\eta > 0$ we have \beq G_{\lb}\, (z)\;\;=\;\; -\,\left(z -\lb V(0) +{\textstyle{\frac{1}{4}}} \!\sum_{x:\,d(x,0)=1} G_{\lb}^{(x|0)}\, (z) \right)^{-1}\; \la{G3} \eeq and, for any two nearest neighbors sites $x,y \in \B$, \beq G_{\lb}^{(x|y)}\, (z)\;\;=\;\; -\,\left(z - \lb V(x) +{\textstyle{\frac{1}{4}}} \!\sum_{x':\,d(x',x)=1,\,x'\not=y} G_{\lb}^{(x'|x)}\, (z) \right)^{-1}\;. \la{G4} \eeq \ep \bp For any $\lb \in \bbr$, $E \in \bbr$ and $\eta > 0$ we have \beq G_{\lb}\, (z)\;\;=\;\; \frac{ i}{\pi} \int_{\bbrs^2} \e^{i(z -\lb V(0))\vp ^2} \exp \left\{{\textstyle{\frac{i}{4}}} \!\sum_{x:\,d(x,0)=1} G_{\lb}^{(x|0)}\, (z)\, \vp^2\right\}\, d^2\vp \la{G2}\;, \eeq and, for any two nearest neighbors sites $x,y \in \B$, \beq \lefteqn{\qquad \quad \e^{ \frac{i}{4} G_{\lb}^{(x|y)}\, (z) \vp^2}\;\;=} \la{G22} \\ &&-\frac{1}{\pi}\int_{\bbrs^2} \e^ { - i \vp\cdot\vp'}\, \partial\left\{\e^{i(z -\lb V(x)) {{\vp'}^{2}}} \exp \left\{{\textstyle{\frac{i}{4}}} \!\sum_{x':\,d(x',x)=1,\,x'\not=y}\!\! G_{\lb}^{(x'|x)}\, (z) \,{{\vp'}^{2}}\right \} \right\}\, d^2\vp'\;, \nonumber \eeq where $\vp^2=\vp\cdot \vp$ and $\partial f(\vp^2)= f'(\vp^2)$. \ep \bpf If we perform the integration in (\ref{G2}) and (\ref{G22}) we obtain (\ref{G3}) and (\ref{G4}). \epf \bd For any $\lb \in \bbr$, $E \in \bbr$ and $\eta > 0$ let \beq \xi_{\lb,z} ( t, s)\;\;&=&\;\; \E \left(\exp{ \left\{\frac{i}{4} \left ( G_{\lb}^{(0)}\, (z) \,t\, - \,\overline{ G_{\lb}^{(0)}\, (z) }\, s\right)\right\}}\right) \la{xia} \\ &=&\;\;\E \left(\exp{ \left\{\frac{1}{4} \left[i{\R}_{\lb}^{(0)} (z) \,(t-s)\, -\, {\I}_{\lb}^{(0)} (z)\, (t+s)\right]\right\}}\right) \la{xi} \eeq for $t,s >0$, where ${\R}_{\lb}^{(x|0)} (z) + i{\I}_{\lb}^{(x|0)} (z)$ is the decomposition of $G_{\lb}^{(x|0)}\, (z)$ into its real and imaginary parts. \ed If $\lb =0$ we can calculate $G_{0}^{(0)}\, (z)$ \ci{AK} obtaining \beq \xi_{0,z} (t, s)\;\;=\;\; \e^ { \frac{ i }{2K } \{(-z +\sqrt{z^2 - K})t\, - \, \overline{(-z +\sqrt{z^2 - K})}s \} } \;,\la{ze0} \eeq where we always make the choice Im $\sqrt{ \ } >0$. If $|E| <\sqrt{ K}$, we have the pointwise limit %@xi00 \beq \xi_{0,E} (t, s)\;\;\equiv \;\;\lim_{\eta \downarrow 0}\xi_{0,z} (t, s)\;\;=\;\; \e^ { \frac{ 1 }{2K }\{-iE(t-s) -\sqrt{ K- E^2}(t+s) \} } \;.\la{xi00} \eeq \bl \la{eqs} For any $\lb \in \bbr$, $E \in \bbr$ and $\eta > 0$ we have \beq \lefteqn{\qquad \qquad\E(| G_{\lb}\, (z)|^{2p}) \; \;=} \la{EGG}\\ &&{\textstyle\frac{1}{\pi^{2p}}}{\displaystyle \int_{\bbrs^{2p} \times\bbrs^{2p} } } \e^{iE (\vp_+^2-\vp_- ^2) -\eta(\vp_+^2+\vp_- ^2)} h(\lb (\vp_+^2-\vp_- ^2)) [\xi_{\lb,z} ( \vp_+^2, \vp_- ^2)]^{K+1} d^{2p}\vp_+ d^{2p}\vp_- \;, \nonumber \eeq and \beq \lefteqn{\qquad \qquad \quad \xi_{\lb,z} ( \vp_+^2, \vp_- ^2)\;\;=\;\; {\textstyle\frac{1}{\pi^{2p}}}{\displaystyle \int_{\bbrs^{2p} \times\bbrs^{2p} } } \e^ { - i (\vp_+\cdot\vp_+' - i \vp_-\cdot\vp_-')} \;\times } \la{xiK} \\ && \partial_+^p \partial_-^p \left\{\e^{iE ({\vp_+'}{^2}-{\vp_-'}{^2}) -\eta({\vp_+'}{^2}+{\vp_-'}{^2})} h(\lb ({\vp_+'}{^2}-{\vp_-'}{^2})) [\xi_{\lb,z} ( {\vp_+'}{^2}, {\vp_-'}{^2})]^K \right \} d^{2p}{\vp}_+' d^{2p}{\vp}_-' , \nonumber \eeq with \beq \partial_\pm g(\vp_+^2,\vp_-^2 )\;\;= \frac{\partial}{\partial \vp_{\pm}^2} g(\vp_+^2,\vp_-^2 )\;. \eeq \el \bpf Equations (\ref{EGG}) and (\ref{xiK}) follow from (\ref{G2}) and (\ref{G22}) by taking the $p$th power of each side, multiplying it by its complex conjugate, and taking expectations of each side, using the independence of the potential at different sites. \epf \section{Nonlinear analysis} To handle the nonlinear equations (\ref{EGG}) and (\ref{xiK}), we introduce the Banach spaces ${\K}_r$, $1 \le r \le \infty$, given by the completion of $$\{g\!:\;[0,\infty) \times [0,\infty) \to \bbc \;\;\mbox{of class C^\infty }\,;\;\; \|g\|_{{\K}_r} \equiv|\!|\!|\!|g|\!|\!|\!|_2 + |\!|\!|\!|g|\!|\!|\!|_r < \infty \}\;,$$ where \beq |\!|\!|\!|g|\!|\!|\!|_r ^2 \;\;= \;\; \sum_{a,b=0}^p \| (2\partial_+)^a (2\partial_-)^b g(\vp_+^2,\vp_-^2 )\|_{L^r(\bbrs^{2p} \times\bbrs^{2p} ,d^{2p}\vp_+ d^{2p}\vp_-)}^2\;\;. \nonumber \eeq We define linear operators \beq \lefteqn{({\T} g)( \vp_+^2, \vp_- ^2)\;\;=} \\ &&{\textstyle\frac{1}{\pi^{2p}}} {\displaystyle \int_{\bbrs^{2p} \times\bbrs^{2p} } } \e^ { - i (\vp_+\cdot\vp_+' - i \vp_-\cdot\vp_-')}\, \partial_+^p \partial_-^p \left\{g ( {\vp_+'}{^2}, {\vp_-'}{^2}) \right \} d^{2p}{\vp}_+' d^{2p}{\vp}_-' \nonumber \eeq and \beq {\cB} (\lb,z) = M(\e^{iE (\vp_+^2-\vp_- ^2) -\eta(\vp_+^2+\vp_- ^2)} h(\lb (\vp_+^2-\vp_- ^2)) )\;, \eeq where $M(g(\vp_+^2,\vp_-^2 ))$ denotes multiplication by the function $g(\vp_+^2,\vp_-^2 )$. It turns out that $\T$ is unitary on ${\K}_2$ \ci{CK} and is a bounded linear operator from ${\K}_1$ to ${\K}_\infty$, ${\cB}_{\lb,z}$ is a bounded linear operator on all ${\K}_r$, and $g \to g^n$ is a continuous map from ${\K}_\infty$ to ${\K}_1$ for any $n=2,3,\ldots$. We can prove the following lemma. \bl \la{xita} \begin{enumerate} \item[(I)] $\xi_{\lb,z} \in {\K}_\infty$ for all $\lb \in \bbr$ and $z=E+i\eta$ with $\eta >0$. The map $(\lb,E, \eta) \to \xi_{\lb,E +i\eta}$ is continuous from $\bbr \times \bbr \times (0,\infty)$ to ${\K}_\infty$. \item[(II)] If $|E| < \sqrt{K}$ we have $\xi_{0,E} \in {\K}_\infty$ and \beq \lim_{\eta \downarrow 0} \xi_{0,E+i\eta}\;\;=\;\;\xi_{0,E} \;\;\;\mbox{in}\;\; {\K}_\infty \;. \eeq \item[(III)] The integral equation(\ref{xiK}) can be rewritten as a fixed point equation in ${\K}_\infty$: %@fpx \beq \la{fpx} \xi_{\lb,z} \;\;=\;\;{\T}{\cB}_{\lb,z} \xi_{\lb,z}^K \;, \eeq valid for all $\lb \in \bbr$ and $z=E+i\eta$ with $\eta >0$, and also valid for $\lb=0$ and $z=E$ with $|E| < \sqrt{K}$ . \end{enumerate} \el The next step is a fixed point analysis. \bl \la{Q} The map $Q:\,\bbr \times \bbr \times [0,\infty) \times {\K}_\infty \to {\K}_\infty$, defined by \beq Q(\lb, E,\eta,g)\;\;=\;\;{\T} {\cB}_{\lb,E + i \eta} g^K \;-g \;\;, \eeq is continuous. $Q$ is continuously Frechet differentiable with respect to $g$, the partial derivative being \beq Q_g(\lb, E,\eta,g)\;\;=\;\;K {\T} {\cB}_{\lb,E + i \eta} M( g^{K-1}) \;-\; I \;. \eeq Moreover, for any $E$ such that $|E| < \sqrt{K}$ we have $Q(0, E,0,\xi_{0,E}) =0$ and \beq 0 \notin \si( Q_g(0, E,0,\xi_{0,E})) \;. \la{000} \eeq \el \bpf The proof of this lemma is routine except for (\ref{000}). We have $Q_g(0, E,0,\xi_{0,E}) = K{\A}_{0,E} - I$ where ${\A}_{0,E}= {\T}{\cB}_{0,E } M( \xi_{0,E}^{K-1})$. As in \ci{K2}, following Acosta and Klein \ci{AK} (see their Propositions 3.2 and 3.3 and Theorem 3.5), we can show that ${\A}_{0,E}^2$ is a compact operator on ${\K}_\infty$ and \beq \si({\A}_{0,E}) \;\;=\;\;\{ {\cal E}_{i,j} =E_i\bar{E}_j \,;\;\; i,j= 0,1,2,\ldots \} \cup \{0\}\;, \eeq with \beq E_n\;\;=\;\; \left(\frac{-E +i\sqrt{ K- E^2}}{K} \right)^{2n}\;\;\;\mbox{for}\;\; n= 0,1,2,\ldots \;. \la{E} \eeq Since ${\cal E}_{i,j} \not= {1 \over K}$ for any $i,j=0, 1,2,\ldots$, (\ref{000}) follows. \epf Lemma \ref{Q} tells us that the hypotheses of the Implicit Function Theorem (see \ci[2.7.2]{N}) are satisfied by the function $Q(\lb, E,\eta,g)$ at $(0,E,0,\xi_{0,E} )$, if $|E| < \sqrt{K}$. It follows that for each $E$ such that $|E| < \sqrt{K}$ there exist $\lb_E > 0$, $\varepsilon_E >0$, $\eta_E >0$ and $\de_E >0$, such that for each $$(\lb, E',\eta) \in (-\lb_E,\lb_E)\times (E - \ve_E,E + \ve_E) \times [0, \eta_E)$$ there is a unique $\, \omega_{\lb,E',\eta} \in {\K}_\infty$ with $\| \omega_{\lb,E',\eta}- \xi_{0,E} \|_{ {\K}_\infty}< \de_E$, such that $Q(\lb, E',\eta,\omega_{\lb,E',\eta}) =0$. Moreover, the map $$(\lb, E',\eta) \in (-\lb_E,\lb_E)\times (E - \ve_E,E + \ve_E) \times [0, \eta_E) \;\longrightarrow\; \omega_{\lb,E',\eta} \in {\K}_\infty$$ is continuous. Combining with Lemma \ref{xita}, and using the uniqueness of $\omega_{\lb,E',\eta}$ as above for each $(\lb, E',\eta)$, we get \bt \la{xize} For any $E$ such that $|E| < \sqrt{K}$ there exist $\lb_E > 0$, $\varepsilon_E >0$ and $\de_E >0$, such that the map $$(\lb, E',\eta) \in (-\lb_E,\lb_E)\times (E - \ve_E,E + \ve_E) \times (0, \infty) \;\longrightarrow\; \xi_{\lb,E'+i\eta} \in {\K}_\infty \la{xizexi}$$ has a continuous extension to $(-\lb_E,\lb_E)\times (E - \ve_E,E + \ve_E) \times [0, \infty)$ satisfying (\ref{fpx}). \et Theorem \ref{cor} now follows from (\ref{EGG}), Theorem \ref{xize}, the translation invariance of expectations, and a simple compactness argument. Full details for the $p=1$ case can be found in Klein \ci{K2}. \bibliographystyle{amsplain} \renewcommand{\bibname}{References (style RC)} \begin{thebibliography}{99} \bibitem{AAT} Abou-Chacra, R., Anderson, P., Thouless, D. J.: A selfconsistent theory of localization. J. Phys. C: Solid State Phys. {\bf 6}, 1734-1752 (1973) \bibitem{AT} Abou-Chacra, P., Thouless, D. J.: Selfconsistent theory of localization: II. Localization near the band edges. J. Phys. C: Solid State Phys. {\bf 7}, 65-75 (1974) \bibitem{AK} Acosta, V., Klein, A.: Analyticity of the density of states in the Anderson model in the Bethe lattice. J. Stat. Phys. {\bf 69}, 277-305 (1992) \bibitem{A} Aizenman, M. : Localization at weak disorder: some elementary bounds. Rev. Math. Phys., to appear. \bibitem{AM} Aizenman, M., Molchanov, S.: Localization at large disorder and extreme energies: an elementary derivation. Commun. Math. Phys. {\bf 157}, 245-278 (1993) \bibitem{And} Anderson, P.: Absence of diffusion in certain random lattices. Phys. Rev. {\bf 109}, 1492-1505 (1958) \bibitem{CK} Campanino, M., Klein, A.: A supersymmetric transfer matrix and differentiability of the density of states in the one-dimensional Anderson model. Comm. Math. Phys. {\bf 104}, 227-241 (1986) \bibitem{CKM} Carmona, R., Klein, A., Martinelli, F.: Anderson localization for Bernoulli and other singular potentials. Commun. Math. Phys. {\bf 108}, 41-66 (1987) \bibitem{CL} Carmona, R., Lacroix, J.: Spectral Theory of Random Schrodinger Operators. Boston, MA: Birkhauser, 1990 \bibitem{DLS} Delyon, F., Levy, Y., Souillard, B.: Anderson localization for multidimensional systems at large disorder or low energy. Commun. Math. Phys. {\bf 100}, 463-470 (1985) \bibitem{DK} von Dreifus, H., Klein, A.: A new proof of localization in the Anderson tight binding model. Commun. Math. Phys. {\bf 124}, 285-299 (1989) \bibitem{FK} Figotin, A., Klein, A.: Localization phenomenon in gaps of the spectrum of random lattice operators. J. Stat. Phys. {\bf 75}, 997-1021 (1994) \bibitem{FMSS} Fr\"ohlich, J., Martinelli, F., Scoppola, E., Spencer, T.: Constructive proof of localization in the Anderson tight binding model. Commun. Math. Phys. {\bf 101}, 21-46 (1985) \bibitem{FS} Fr\"ohlich, J., Spencer, T.: Absence of diffusion in the Anderson tight binding model for large disorder or low energy. Commun. Math. Phys. {\bf 88}, 151-184 (1983) \bibitem{GMP} Gol'dsheid, Ya., Molchanov, S., Pastur, L.: Pure point spectrum of stochastic one dimensional Schr\"odinger operators. Funct. Anal. Appl. {\bf 11}, 1-10 (1977) \bibitem{Gf} Graf, G. M.: Anderson localization and the space-time characteristic of continuum states. J. Stat. Phys. {\bf 75}, 337-346 (1994) \bibitem{K} Klein, A.: The supersymmetric replica trick and smoothness of the density of states for random Schrodinger operators. Proc. Symposia in Pure Mathematics {\bf 51}, 315-331 (1990) \bibitem{K3} Klein, A.: Localization in the Anderson model with long range hopping. Braz. J. Phys. {\bf 23}, 363-371 (1993) \bibitem{K1} Klein, A.: Absolutely continuous spectrum in the Anderson model on the Bethe lattice. Mathematical Research Letters {\bf 1}, 399-407 (1994) \bibitem{K2} Klein, A.: Extended states in the Anderson model on the Bethe lattice. Advances in Math., to appear. \bibitem{K4} Klein, A.: Spreading of wave packets in the Anderson model on the Bethe lattice. Commun. Math. Phys., to appear. \bibitem{KMP} A. Klein, F. Martinelli and J. F. Perez, A rigorous replica trick approach to Anderson localization in one dimension. Commun. Math. Phys. {\bf 106}, 623-633 (1986) \bibitem{KS1} Kunz, H., Souillard, B.: Sur le spectre des operateurs aux differences finies aleatoires. Commun. Math. Phys. {\bf 78}, 201-246 (1980) \bibitem{KS} Kunz, H., Souillard, B.: The localization transition on the Bethe lattice. J. Phys. (Paris) Lett. {\bf 44} , 411-414 (1983) \bibitem{MS} McKane, A. J., Stone, M.: Localization as an alternative to Goldstone's theorem. Ann. Phys. {\bf 131}, 36-55 (1981). \bibitem{MD} Miller, J., Derrida, B.: Weak disorder expansion for the Anderson model on a tree. J. Stat. Phys. {\bf 75}, 357-389 (1994) \bibitem{MF} Mirlin, A. D., Fyodorov, Y. V.: Localization transition in the Anderson model on the Bethe lattice: spontaneous symmetry breaking and correlation functions. Nucl. Phys. B {\bf 366}, 507-532 (1991) \bibitem{MT} Mott, N., Twose, W.: The theory of impurity conduction. Adv. Phys. {\bf 10}, 107-163 (1961) \bibitem{N} Nirenberg, L.: Topics in Nonlinear Functional Analysis. New York: Courant Institute of Mathematical Sciences, 1974 \bibitem{P} Pastur, L.: Spectra of random selfadjoint operators. Russ. Math. Surv. {\bf 28}, 1-67 (1973) \bibitem{simon} Simon, B.: $L^p$ norms of the Borel transform and the decomposition of measures. Proc. Amer. Math. Soc., to appear \bibitem{SW} Simon, B., Wolff, T.: Singular continuum spectrum under rank one perturbations and localization for random Hamiltonians. Commun. Pure. Appl. Math. {\bf 39}, 75-90 (1986) \bibitem{Sp} Spencer, T.: Localization for random and quasiperiodic potentials. J. Stat. Phys. {\bf 51}, 1009-1019 (1988) \bibitem{Sp2} Spencer, T.: Lifshitz tails and localization. Preprint \bibitem{T} Thouless, D.: Electrons in disordered systems and the theory of localization. Phys. Rev. {\bf 13}, 93-106 (1974) \end{thebibliography} \end{document}