Content-Type: multipart/mixed; boundary="-------------0001140413516" This is a multi-part message in MIME format. ---------------0001140413516 Content-Type: text/plain; name="00-20.keywords" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="00-20.keywords" Black-Box, long-range, resolvent estimate, semi-classique, trapping energy level ---------------0001140413516 Content-Type: application/x-tex; name="ResEst.tex" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="ResEst.tex" \documentstyle[epsf]{amsart} \setlength{\topmargin}{0cm} \setlength{\textheight}{21cm} \setlength{\oddsidemargin}{0in} \setlength{\evensidemargin}{0in} \setlength{\textwidth}{6.5in} \setlength{\parindent}{.25in} \pagestyle{headings} %\pagestyle{plain} \def\Bbbone{{\mathchoice {1\mskip-4mu \text{l}} {1\mskip-4mu \text{l}} { 1\mskip-4.5mu \text{l}} { 1\mskip-5mu \text{l}}}} \def\squarebox#1{\hbox to #1{\hfill\vbox to #1{\vfill}}} \newcommand{\stopthm}{\hfill\hfill\vbox{\hrule\hbox{\vrule\squarebox {.667em}\vrule}\hrule}\smallskip} \pagestyle{headings} %\pagestyle{plain} \newcommand{\1}{{\bold 1}} \newcommand{\F}{{\cal F}} \newcommand{\CC}{{\cal C}} \newcommand{\CI}{{\cal C}^\infty } \newcommand{\Oo}{{\cal O}} \newcommand{\K}{{\cal K}} \newcommand{\D}{{\cal D}} \newcommand{\G}{{\cal G}} \newcommand{\Hh}{{\cal H}} \newcommand{\pic}{{\mbox{Pic}}} \newcommand{\Z}{{\Bbb Z}} \newcommand{\Q}{{\Bbb Q}} \newcommand{\RR}{{\Bbb R}} \newcommand{\HH}{{\Bbb H}} \newcommand{\U}{{\cal U}} \newcommand{\A}{{\Bbb A}} \newcommand{\C}{{\Bbb C}} \newcommand{\N}{{\Bbb N}} \newcommand{\Tr}{\operatorname{Tr}} \newcommand{\vol}{\operatorname{vol}} \newcommand{\rank}{\operatorname{rank}} \newcommand{\half}{\frac{1}{2}} \newcommand{\itt}{\operatorname{it}} \newcommand{\supp}{\operatorname{supp}} \newcommand{\Ran}{\operatorname{Ran}} \newcommand{\itA}{\operatorname{it}} \newcommand{\im}{\operatorname{Im}} \newcommand{\point}{\operatorname{point}} \newcommand{\comp}{\operatorname{comp}} \newcommand{\loc}{\operatorname{loc}} \newcommand{\Diff}{\operatorname{Diff}} \newcommand{\tr}{\operatorname{tr}} \newcommand{\rarrow}{\operatornamewithlimits{\longrightarrow }} \newsymbol\circlearrowleft 1309 \newsymbol\restriction 1316 \newcommand{\rest}{\!\!\restriction} \newcommand{\ttt}{|\hspace{-0.25mm}|\hspace{-0.25mm}|} \renewcommand{\Re}{\mathop{\rm Re}\nolimits} \renewcommand{\Im}{\mathop{\rm Im}\nolimits} \theoremstyle{plain} \def\Rm#1{{\rm#1}} \newtheorem{thm}{Theorem} \renewcommand{\thethm}{\arabic{thm}} \newtheorem{cor}{Corollary} \renewcommand{\thecor}{\arabic{cor}} \newtheorem{lem}{Lemma} \renewcommand{\thelem}{\arabic{lem}} %\numberwithin{lem}{section} \newtheorem{prop}{Proposition} \renewcommand{\theprop}{\arabic{prop}} %\numberwithin{prop}{section} \newtheorem{rem}{Remark} \theoremstyle{definition} \newtheorem{ex}{EXAMPLE}[section] \numberwithin{equation}{section} \newcommand{\thmref}[1]{Theorem~\ref{#1}} \newcommand{\secref}[1]{Section~\ref{#1}} \newcommand{\lemref}[1]{Lemma~\ref{#1}} \newcommand{\exref}[1]{Example~\ref{#1}} \newcommand{\corref}[1]{Corollary~\ref{#1}} \newcommand{\propref}[1]{Proposition~\ref{#1}} \title[Semiclassical resolvent estimates ] {Semiclassical resolvent estimates for trapping perturbations} \author[V. Bruneau, V. Petkov]{Vincent Bruneau and Vesselin Petkov} %------------------------lettres grecs ----------------------- \def\e{\varepsilon} \def\phi {\varphi} \def \la {{\lambda}} \def \a {{\alpha}} \def\t{\theta} %-------------------------domaine--------------------- \newcommand{\omd}{\Omega_{\delta}} %analyse \def\CC{{\cal C}} \def\lap{\bigtriangleup} %-----------------------------tilde------------------- \newcommand{\tL}{\tilde L} \newcommand{\tl}{\tilde l} \newcommand{\tP}{\tilde P} \newcommand{\tR}{\tilde R} \newcommand{\DS}[1]{{\displaystyle #1}} \def \rn{{{\RR}^n}} \begin{document} \maketitle $\:\:\:\:\:$ {\bf Abstract.} {\footnotesize We study the semiclassical estimates of the resolvent $R(\lambda + i\tau),\:\:\lambda \in J \subset\subset{\RR}^{+},\: \tau \in ]0,1]$ of a self-adjoint operator $L(h)$ in the space of bounded operators ${\cal L}({\cal H}^{0,s},{\cal H}^{0,-s}),\:s > 1/2$. In the general case of long-range trapping "black-box" perturbations we prove that the estimate of the cut-off resolvent $\|\chi(x)R(\lambda + i0)\chi(x)\|_{{\cal H} \to {\cal H}} \leq C\exp(Ch^{-p}),\:\chi(x) \in C^{\infty}_0({\RR^n}),\:p \geq 1$ implies the estimate $\|R(\lambda + i\tau)\|_{s,-s} \leq C_1\exp(C_1 h^{-p})$.} \section{Introduction} The purpose of this paper is to obtain semiclassical estimates for the resolvent $$ R(z)=(L(h)-z)^{-1},\:z \in {\C} \setminus \RR$$ of a self-adjoint operator $L = L(h)$ depending on $h \in ]0,h_0]$ for $$z \in {\cal B}_{\pm} = \{z = \lambda + i\tau \in {\C}: \lambda \in J,\:\pm\tau \in ]0,1]\},$$ $J = ]\mu_0, \mu_1[ \subset \subset {\RR}^{+}.$ The operator $L$ is defined in a domain ${\cal D} \subset {\cal H}$ of a complex Hilbert space ${\cal H}$ with an orthogonal decomposition $${\cal H} = {\cal H}_{R_0} \oplus L^2({\RR}^n \setminus B(0,R_0)),\:B(0,R_0) = \{x \in {\RR}^n: |x| \leq R_0 \},\:\:n \geq 2 \: $$ and $L$ satisfies the long-range "black box" assumptions (\ref{eq:1.7})-(\ref{eq:1.13}) described below. Introduce the spaces $${\cal H}^{0,s} = {\cal H}_{R_0} \oplus L^2(\RR^n\setminus B(0,R_0), \langle x {\rangle}^s dx), \:\langle x \rangle = (1 + |x|^2)^{1/2}.$$ First we show that for $s > 1/2$ and fixed $h > 0$ the limits \begin{equation} \lim_{\epsilon \to 0,\: \epsilon > 0} R(\lambda \pm i\epsilon) = R(\lambda \pm i0),\:\:\lambda \in J \label{eq:1.1} \end{equation} exist in the space of bounded operators ${\cal L}({\cal H}^{0,s},{\cal H}^{0,-s})$, where here and below we omit in $R(z)$ the dependence of $h$. Next we give semiclassical resolvent estimates \begin{equation} \|R(\lambda \pm i\tau)\|_{s,-s} \leq Cr(h),\: \lambda \in J,\:\:\tau \in ]0,1],\:\: h \in ]0, h_0], \label {eq:1.2} \end{equation} where $\|.\|_{s,-s}$ denotes the norm in ${\cal L}({\cal H}^{0,s},{\cal H}^{0,-s})$. Such estimates have been obtained by many authors in great generality in the case when every $\lambda \in J$ is a {\em non-trapping energy level} for the principal symbol $l_0(x,\xi)$ of a differential operator $L$ (see \cite{kn:RT1}, \cite{kn:GM1}, \cite{kn:RT2}, \cite{kn:HN}, \cite{kn:W}, \cite{kn:R2}, \cite{kn:R3}, \cite{kn:VZ} and the survey article \cite{kn:R4} for other references). The proofs of these estimates are essentially based on Mourre method \cite{kn:M}, \cite{kn:CFKS}, \cite{kn:PSS} and the main point is to construct a conjugate operator $A(h)$ for which the Mourre inequality (\ref{eq:mourreesti}) holds (see Section 5.1). Roughly speaking we call these operators non-trapping perturbations and in this case we have $r(h) = h^{-1}$ in (\ref{eq:1.2}). Moreover, for such perturbations there are no resonances converging sufficiently fast to the real axis as $h \to 0.$\\ The case of {\em trapping energy levels} is more complicated. First, without the non-trapping assumption, in general, a Mourre type inequality is not true. Secondly, resonances converging exponentially fast to the real axis could exist and, consequently, we must have $r(h) = e^{Ch^{-p}}$ with some $p \geq 1.$ In the special case of a Schr\"{o}dinger operator $-h^2\Delta + V(x)$, with potential $V(x)$ having the form of a "well in an island", the results in \cite{kn:GM2}, \cite{kn:GMR} imply the estimate (\ref{eq:1.2}) with $r(h) = e^{Ch^{-1}}$. Recently, for long-range trapping perturbations in the exterior of an obstacle, Burq \cite{kn:Bu} established the estimate \begin{equation} \label{eq:1.3} \|\chi(x)R(\lambda +i0)\chi(x)\|_{{\cal H} \to {\cal H}} \leq C\exp(Ch^{-p}),\:\lambda \in J,\:0 < h \leq h_0, \end{equation} with $p=1$ and some constants $C > 0,\: h_0 > 0$, provided $\chi(x) \in C^{\infty}_0({\RR}^n)$ is equal to 1 on $\overline{B(0,R_0)}.$ In fact, Burq \cite{kn:Bu} obtained the estimate \begin{equation} \|(L_{\theta}(h) - \lambda)^{-1} \|_{L^2({\cal H}_{\theta})} \leq Ce^{C_1 h^{-p}},\:\lambda \in J, \label{eq:1.3*} \end{equation} with $p =1,\:L_{\theta}(h)$ being the operator given by complex scaling (see \cite{kn:SZ}, \cite{kn:Sj1}). On the other hand, the estimate (\ref{eq:1.3*}) implies immediately (\ref{eq:1.3}) since we have $$\chi(x) R(\lambda + i0)\chi(x) = \chi(x)(L_{\theta}(h) - \lambda)^{-1}\chi(x)$$ for suitably chosen $\chi(x)$ (see Lemma 3.5 in \cite{kn:SZ}). Next, denote by Res $L(h)$ the set of resonances of $L(h)$ and assume that there exist $\epsilon > 0, \:c > 0$ and $q \geq 1$ so that $${\rm dist}\: \{ {\rm Res}\:L(h),\: [\mu_0, \mu_1] \} \geq \epsilon \exp(-c h^{-q}),\: 0 < h \leq h_0.$$ Then an application of Lemma 1 in \cite{kn:TZ} with $g(h) = \exp(-\frac{c}{2}h^{-q})$ yields the estimate (\ref{eq:1.3*}) with $p \geq q.$ Thus we have several sufficient conditions leading to the estimate (\ref{eq:1.3}).\\ In this work we show that an estimate of the cut-off resolvent $\|\chi(x)R(\lambda + i0)\chi(x)\|_{{\cal H}},\: \lambda \in J,$ implies an estimate of $\|R(\lambda + i\tau)\|_{s, -s}$ with the same order $p$ in the bound $\exp(Ch^{-p}).$ Consequently, we extend the result of \cite{kn:Bu} replacing the cut-off factor $\chi(x)$ by a function $\psi(x) \in C^{\infty}({\RR}^n)$ which is equal to $\langle x \rangle^{-s},\: s > 1/2$ for $|x| \geq R_1 > R_0$. Moreover, for long-range perturbations in ${\Bbb R}^n$ studied in \cite{kn:Bu}, we prove the conjecture of Robert \cite{kn:R3} who claimed that (\ref{eq:1.2}) with $r(h) = C\exp(Ch^{-p})$ holds for general trapping perturbations. Under this assumption, Robert \cite{kn:R3} obtained a Weyl type asymptotic for the scattering phase related to two $h$-admissible pseudodifferential operators $L_1(h),\:\:L_2(h)$. Our main result is the following. \begin{thm}\label{theo} Let $L(h)$ be a self-adjoint operator satisfying the "black box" assumptions $(\ref{eq:1.7})- (\ref{eq:1.13})$. Then for fixed $s > 1/2$ and for sufficiently small fixed $h$ the limits $$R(\la \pm i0) = \lim_{\Im z \to 0, \: \: \pm \Im z > 0} R(z), \quad \Re z=\la \in J,$$ exist in ${\cal L}({\cal H}^{0,s},{\cal H}^{0,-s})$ uniformly with respect to $\lambda \in J$. Moreover, if there exist $\delta > 0$ and $p \geq 1$ such that \begin{equation} \|\chi(x)R(\lambda \pm i0)\chi(x)\|_{{\cal H} \to {\cal H}} \leq C\exp(Ch^{-p}),\:\lambda \in ]\mu_0 -\delta, \mu_1 + \delta[,\:0 < h \leq h_0, \label{eq:1.5} \end{equation} holds, then for each $s > \frac12$ there exist $C_s > 0,\:\: h_1 > 0$ such that for $(\lambda, \tau) \in J \times ]0, 1]$ we have \begin{equation} \label{eq:1.6} \| R(\lambda \pm i \tau) \|_{s,-s} \leq C_s \exp (C_sh^{-p}),\:\: 0 < h \leq h_1. \end{equation} \end{thm} The idea of our proof is to construct a self-adjoint operator $\tilde{L} = \tilde{L}(h)$ on $L^2({\Bbb R}^n)$ such that each $\lambda \in J$ is a non-trapping energy level for $\tilde{L}$ and $L(h)\psi = \tilde{L}(h) \psi$ for any $\psi \in C^{\infty}_0({\Bbb R}^n)$ supported away from a ball $B(0, R)$. For this purpose we show that the bounded trajectories related to the symbol of $L(h)$ and to the energy levels $\lambda \in J$ are included in a compact set. Next we represent $R(z)$ by a sum of terms involving the resolvent $\tilde{R}(z)$ of $\tilde{L}$ (see Section 3) and an application of the semiclassical estimates for non-trapping energy levels reduces the problem to the estimation of $\|\chi(x)R(z)\chi(x)\|_{{\cal H}}$ (see Proposition 3). In Section 2 our approach is motivated by the proof of Lemma 1.2 in \cite{kn:GMR} concerning the operator $L(h) = -h^2\Delta + V(x)$, where the potential $V(x)$ involves a "well in an island". Having an "island" around the trapped trajectories, G\'erard, Martinez and Robert have introduced a perturbation $W_0(x) \in C^{\infty}_0({\RR}^n)$ such that the perturbed symbol $|\xi |^2 + V(x) + W_0(x)$ becomes a non-trapping Hamiltonian for energy levels $\lambda \in J.$ We prove a similar result in the general case by constructing a perturbation $\tilde l(x,\xi)$ of $l_0(x,\xi)$ (see Proposition \ref{prop2}) so that all $\lambda \in J$ are non-trapping energy levels for $\tilde l(x,\xi)$. The existence of $\tilde l(x, \xi)$ implies immediately the existence of an operator $\tilde{L}(h)$ with the properties mentioned above. Finally, let us mention that the estimate (\ref{eq:1.6}) combined with the argument of Robert \cite{kn:R3} lead to a Weyl asymptotic of the scattering phase for general long-range perturbations in ${\Bbb R}^n$ when we have not a "black box".\\ Now we will recall the abstract "black box" assumptions given in \cite{kn:SZ}, \cite{kn:Sj1} and \cite{kn:Sj2}. Suppose that ${\cal D}$ satisfies \begin{equation}\label{ass1} {\Bbbone }_{{\RR}^n \setminus B(0,R_0)}{\cal D} = H^2({\RR}^n \setminus B(0,R_0)), \label{eq:1.7} \end{equation} uniformly with respect to $h$ in the sense of \cite{kn:Sj1}. More precisely, equip $H^2({\RR}^n \setminus B(0,R_0))$ with the norm $\|^2u\|_{L^2},\:^2 = 1 + (hD)^2$ and equip ${\cal D}$ with the norm $\|(L+i)u\|_{{\cal H}}.$ Then we require that ${\Bbbone }_{{\RR}^n \setminus B(0,R_0)}: {\cal D} \longrightarrow H^2({\RR}^n \setminus B(0,R_0))$ is uniformly bounded with respect to $h$ and this map has a uniformly bounded right inverse. Assume that %\begin{equation} %{\rm if}\: u \in H^2({\RR}^n \setminus B(0,R_0))\:\: {\rm vanishes}\:{\rm near}\:\partial B(0,R%_0),\:{\rm then}\: u \in {\cal D}, %\end{equation} \begin{equation} \label{ass2} {\Bbbone}_{B(0,R_0)}(L+i)^{-1} \hbox{is compact} \end{equation} and \begin{equation}\label{ass3} (Lu)\vert_{{\RR}^n \setminus \overline{B(0,R_0)}} = Q\Bigl( u\vert_{{\RR}^n \setminus \overline{B(0,R_0)}}\Bigr), \end{equation} where $Q$ is a formally self-adjoint differential operator \begin{equation}\label{ass4} Q u = \sum_{| \alpha | \leq 2} a_\alpha (x;h) (hD_x)^\alpha u, \end{equation} with $ a_\alpha (x;h)= a_\alpha (x)$ independent of $h$ for $| \alpha | = 2$ and $a_\alpha \in C_b^\infty(\RR^n)$ uniformly bounded with respect to $h$. Next we assume the following properties: There exists $C>0$ such that \begin{equation} \label{ass5} l_0(x,\xi) =\sum_{| \alpha | = 2} a_\alpha (x) \xi^\alpha \geq C {\langle \xi \rangle}^2,\:\:\langle \xi \rangle = (1 + |\xi|^2)^{1/2}. \label{eq:1.9} \end{equation} There exists $\gamma > 0$ such that for every $(\alpha,\beta) \in {\N}^n \times \N^n$, we have \begin{equation}\label{ass6} | {\partial}_{x}^{\alpha}{\partial}_{\xi}^{\beta} \Big(\sum_{| \alpha | \leq 2} a_\alpha (x;h) \xi^\alpha - |\xi|^2 \Big) | \leq C_{\alpha,\beta} {\langle x \rangle}^{-\gamma- \mid \alpha \mid}{\langle \xi \rangle}^2 \label{eq:1.10} \end{equation} uniformly with respect to $h$. There exist $\theta_0 \in ]0,\pi[,\:\epsilon > 0$ and $R_1 > R_0$ so that the coefficients $a_{\alpha}(x;h)$ of $Q$ can be extended holomorphically in $x$ to \begin{equation} \label{ass7} \{r\omega;\:\omega \in {\C}^n,\: {\rm dist}\:(\omega, S^{n-1}) < \epsilon, \: r \in {\C},\: |r| > R_1,\:{\rm arg} \: r \in [-\epsilon, \theta_0 + \epsilon)\} \label{eq:1.11} \end{equation} and (\ref{eq:1.10}) extend to this larger set. Let $R > R_0,\:T = ({\RR}/\tilde{R}{\Z})^n,\: \tilde{R} > 2R.$ Set $${\cal H}^{\#} = {\cal H}_{R_0} \oplus L^2(T \setminus B(0, R_0))$$ and consider a self-adjoint operator $L^{\#}: {\cal H}^{\#} \longrightarrow {\cal H}^{\#}$ with domain $${\cal D}^{\#} = \{u \in {\cal H}^{\#}: \: \chi u \in {\cal D}, \: (1-\chi)u \in H^2 \},$$ where $\chi \in C^{\infty}_0(B(0,R); [0,1])$ is equal to 1 near $\overline{B(0,R_0)}.$ Take a differential operator $$Q^{\#} = \sum_{|\alpha| \leq 2} a_{\alpha}^{\#}(x;h)(hD)^{\alpha}$$ with $a_{\alpha}^{\#}(x;h) = a_{\alpha}(x;h)$ for $|x| < R$, put $$L^{\#}u = L(\chi u) + Q^{\#}((1-\chi)u),\: u \in {\cal D}^{\#}$$ and denote by $N(L^{\#}, [-\lambda^2, \lambda^2])$ the number of eigenvalues of $L^{\#}$ in the interval $[-\lambda^2, \lambda^2]$. Then we assume that \begin{equation}\label{ass8} N(L^{\#}, [-\lambda^2, \lambda^2]) = {\cal O}(\Bigl(\frac{\lambda}{h}\Bigr)^{n^{\#}}),\: n^{\#} \geq n,\: \lambda \geq 1. \label{eq:1.12} \end{equation} Finally, we suppose that \begin{equation}\label{ass9} \sigma_{pp}(L(h)) \cap [\mu_0 - \delta, \mu_1+\delta] = \emptyset,\: h \in ]0,h_0], \label{eq:1.13} \end{equation} where $\delta > 0$ is the constant in (\ref{eq:1.5}). A typical example for an operator satisfying the conditions (\ref{eq:1.7})-(\ref{eq:1.13}) is an elliptic self-adjoint Schr\"{o}dinger operator $$P = \sum_{|\alpha| \leq 2} a_{\alpha}(x;h) (hD_x)^{\alpha}$$ with $C^{\infty}$ coefficients in the connex exterior $\Omega = {\RR}^n \setminus {\cal O}$ of a bounded obstacle ${\cal O}$ with $C^{\infty}$ boundary $\partial \Omega$ with Dirichlet or Neumann boundary condition on $\partial \Omega$ (see \cite{kn:Bu}). In this case ${\cal D} = H^2 ({\RR}^n \setminus {\cal O}) \cap H_0^1({\RR}^n \setminus {\cal O})$ and $P$ is a positive operator in $L^2({\RR}^n \setminus {\cal O}).$\\ Under the above assumptions the resonances close to the real axis can be defined by the method of complex scaling \cite{kn:SZ}, \cite{kn:Sj1} and they coincide with the poles of the meromorphic continuation of the resolvent $(L(h) - z)^{-1}: {\cal H}_{{\rm comp}} \longrightarrow {\cal D}_{{\rm loc}}$ from $\Im z > 0$ to a conic neighborhood of the positive real axis in the lower half plane. The set of resonances will be denoted by Res $L(h)$.\\ The paper is organized as follows. Section 2 is devoted to the analysis of the non-trapping energy levels and to the construction of the symbol $\tilde l(x,\xi).$ In Section 3 we prove Theorem 1 introducing the operator $\tilde{L}(h)$ and exploiting the semiclassical estimates for non-trapping energy levels. In Section 4 we obtain some microlocal estimates closely related to the bounds of $\|R(z)\|_{s, -s}$. In a further work we will apply the microlocal estimates given in Proposition 5 for the analysis of the semiclassical asymptotic of the scattering phase.\\ {\bf Acknowledgments.} The authors are grateful to Didier Robert for helpful discussions and valuable comments concerning the non-trapping perturbations. We would to thank Maciej Zworski for helpful comments and discussions as well as for his suggestion to study the long-range "black box" perturbations.\\ \section{Classical trajectories and non-trapping energy levels} Let $l(x,\xi) \in C^\infty(\rn \times \rn :\RR)$ be a classical Hamiltonian, and let $H_{l}= (\partial_\xi l, - \partial_x l)$ be the associated Hamiltonian vector field: $$H_{l}(f)(x,\xi) = \{l,f\}(x,\xi)=(\partial_\xi l \partial_x f- \partial_x l \partial_\xi f)(x,\xi),\:\: f \in C^\infty(\rn \times \rn). $$ Denote by $\Phi_l^t$ the Hamiltonian flow of $l$ given by $$\Phi_l^t = \hbox{\rm exp}(tH_{l}) \; : \; (x_0, \xi_0) \mapsto \Big( x(t, x_0, \xi_0), \xi(t, x_0, \xi_0)\Big).$$ Therefore for each $\la \in \RR$, the energy surface $\Sigma_\la=l^{-1}(\la)$ is stable by $\Phi_l^t$. In the exposition below we denote $J = ]\mu_0, \mu_1[ \subset \subset \RR^{+}$. We will say that $(x,\xi)$ is a non-critical point for $l(x,\xi)$ if $\nabla_{x,\xi} l(x,\xi) \neq 0$ and that $\lambda$ is a non-critical energy level for $l(x,\xi)$, if any $(x,\xi) \in \Sigma_\la=l^{-1}(\la)$ is a non-critical point for $l(x,\xi)$. We start by an examination of the bounded trajectories. \begin{lem}\label{lem1} Let $\mu_1 > \mu_0 > 0$ be fixed. Assume there exist $\mu \in ]0,\mu_0 [$, $R>0$ such that \begin{equation}\label{eq:2.1} | x | \geq R, \; (x,\xi) \in l^{-1}(]\mu_0,\mu_1[) \; \Rightarrow \; | \frac14 \{ l, x \cdot \partial_\xi l \} (x,\xi) - l(x,\xi) | \leq \mu. \end{equation} Then for each $\la \in J$ and each $(x_0,\xi_0) \in \Sigma_\la$ with $|x_0 | > R$ the function $x(\cdot,x_0,\xi_0)$ is non-bounded. \end{lem} \begin{pf} By definition, $|x(t)|^2$ satisfies \begin{equation}\label{eq:2.2} \left\{ \begin{array}{ccl} (|x|^2)' = \left\{l,x^2 \right\}(x,\xi) = 2 x\cdot \partial_\xi l(x,\xi), \\ (x\cdot \partial_\xi l(x,\xi))' = \left\{l,x\cdot \partial_\xi l \right\}(x,\xi). \end{array} \right. \end{equation} For $(x_0,\xi_0) \in \Sigma_\la$ we have $(x(t),\xi(t)) \in \Sigma_\la$ and this yields \begin{equation}\label{eq:2.3} ( x\cdot \partial_\xi l(x,\xi))' = 4\la + \left\{l,x\cdot \partial_\xi l \right\}(x,\xi) - 4 l(x,\xi). \end{equation} Let $|x_0| > R$ and $(x_0,\xi_0) \in l^{-1}(]\mu_0,\mu_1[)$. Denote by $[T_m,T_M]$ the maximal interval containing $t_0=0$ so that \begin{equation}\label{eq:2:4} |\Big( \{ l, x \cdot \partial_\xi l \} - 4 l \Big)(x(t),\xi(t)) | \leq 4 \mu, \:\: t \in [T_m, T_M]. \end{equation} By the continuity of $x(t)$ and $|\{ l, x \cdot \partial_\xi l \} -4 l |(x(t),\xi(t))$ a such interval exists and it is not reduced to $\{0\}$. From the relation (\ref{eq:2.3}) we deduce $$( x\cdot \partial_\xi l(x,\xi))'(t) \geq 4 (\la - \mu) \geq 4 (\mu_0 - \mu),\:\:\forall t \in [T_m,T_M],$$ that is $$ \left\{\begin{array}{cccc} (x\cdot \partial_\xi l(x,\xi))(t) \geq x_0\cdot \partial_\xi l(x_0,\xi_0) + 4(\mu_0 - \mu) t,\:\:\forall t \in [0,T_M], \\ ( x\cdot \partial_\xi l(x,\xi))(t) \leq x_0\cdot \partial_\xi l(x_0,\xi_0) + 4(\mu_0 - \mu) t,\:\:\forall t \in [T_m,0].\:\: \\ \end{array} \right. $$ Then applying (\ref{eq:2.2}), we get $$ |x(t)|^2 \geq |x_0|^2 + 2 x_0\cdot \partial_\xi l(x_0,\xi_0) t + 4 ( \mu_0-\mu) t^2,\:\:\forall t \in [T_m,T_M]. $$ Consequently, if $\pm x_0\cdot \partial_\xi l(x_0,\xi_0) \geq 0$, for $\pm t \geq 0$, we have \begin{equation}\label{eq:esti2} |x(t)|^2 \geq |x_0|^2 + 4 ( \mu_0-\mu) t^2 \end{equation} which implies immediately that $T_M = + \infty$ (if $ x_0\cdot \partial_\xi l(x_0,\xi_0) \geq 0$) or $T_m = - \infty$ (if $ x_0\cdot \partial_\xi l(x_0,\xi_0) \leq 0$) and (\ref{eq:esti2}) holds for any $t \geq 0$ (or $t \leq 0$). Thus, as soon as $|x_0| > R$, and $l(x_0,\xi_0) \in ]\mu_0,\mu_1[$, the function $x(\cdot,x_0,\xi_0)$ is non-bounded. \end{pf} \begin{prop}\label{prop.1} Let $\mu_1 > \mu_0 > 0$ and let $\displaystyle{\lim_{| \xi | \rightarrow \infty} l(x,\xi) = +\infty}$. Assume there exists $\gamma > 0$ such that for every $(\alpha,\beta) \in \N^n \times \N^n$, $|\alpha| \leq 1$, $|\beta | \leq 2$ we have $$| {\partial}_{x}^{\alpha}{\partial}_{\xi}^{\beta} \Big( l(x,\xi) - | \xi |^2 \Big) | \leq C_{\alpha,\beta} {\langle x \rangle}^{-\gamma- \mid \alpha \mid}{\langle \xi \rangle}^2.$$ Then there exists a compact set $K(J)\subset \RR^{2n}$ such that for each $\la \in J$, the set $K(J) \cap \Sigma_\la$ contains all bounded trajectories of the Hamiltonian flow $\Phi_l^t$ on $\Sigma_\la$ and all critical points of $l(x,\xi)$ in $\Sigma_\la.$ \end{prop} \begin{rem} In a more restrictive setting, this result is stated and exploited by the first author $\cite{kn:Br}$ to study the scattering phase for short-range trapping perturbations of the Laplacian. \end{rem} \begin{pf} First, as $\displaystyle{\lim_{| \xi | \rightarrow \infty} l(x,\xi) = + \infty}$, it is easy to see that there exists $M_{\mu_1} > 0$ such that \begin{equation}\label{eq:estixi} (x,\xi) \in \Sigma_\la, \; \forall \la \leq \mu_1 \; \Rightarrow \; |\xi| \leq M_{\mu_1}. \end{equation} Now we will prove that the $x$-component of the bounded trajectories and the critical points are included in a compact set. This means that outside a fixed compact set the $x$-component of the Hamiltonian trajectories are non-bounded and $\nabla_{x,\xi} l(x,\xi) \neq 0$. Using Lemma \ref{lem1} it is sufficient to establish the existence of $R>0$ such that (\ref{eq:2.1}) holds with $\mu <\mu_0$. We write \begin{equation}\label{calcul1} \begin{array}{ccl} \{ l, x \cdot \partial_\xi l \} = \partial_\xi l \cdot \partial_x (x \cdot \partial_\xi l) - \partial_x l \cdot \partial_\xi (x \cdot \partial_\xi l) \\ = | \partial_\xi l |^2 + \sum_{j=1}^{n} \partial_{\xi_j} l \; x \cdot \partial_\xi \partial_{x_j} l - \partial_x l \cdot \partial_\xi (x \cdot \partial_\xi l). \end{array} \end{equation} Set $r(x,\xi)= l(x,\xi) - | \xi |^2$ and observe that $$ | \partial_\xi l |^2 = 4 | \xi |^2 + 4 \xi \cdot \partial_\xi r +| \partial_\xi r |^2,$$ which implies \begin{equation}\label{eq:2.8} \frac14 | \partial_\xi l |^2 - l(x,\xi) = \xi \cdot \partial_\xi r + \frac14 | \partial_\xi r |^2 - r(x,\xi). \end{equation} Moreover, since $\partial_{x_j} l = \partial_{x_j} r$, we deduce the following representation $$\frac14 \{ l, x \cdot \partial_\xi l \} - l(x,\xi)$$ \[ =\xi \cdot \partial_\xi r + \frac14 | \partial_\xi r |^2 - r(x,\xi) + \frac14 \sum_{j=1}^{n} \partial_{\xi_j} l \; x \cdot \partial_\xi \partial_{x_j} r - \partial_x r \cdot \partial_\xi (x \cdot \partial_\xi l). \] From the assumption on $r$ we obtain $$| \frac14 \{ l, x \cdot \partial_\xi l \} - l(x,\xi) | \leq C {\langle x \rangle}^{-\gamma}{\langle \xi \rangle}^4. $$ Thus we can find $R=R(\mu_0,\mu_1,\gamma)$ such that (\ref{eq:2.1}) holds. Moreover taking $R$ such that the RHS of (\ref{eq:2.8}) is bounded by $\mu<\mu_0$, we obtain $| \partial_\xi \l | \neq 0$ on $\l^{-1}(J) \cap \{ (x,\xi); |x| >R\}$. Consequently, each $(x,\xi) \in \l^{-1}(J) \cap \{ (x,\xi); |x| >R\}$ is non-critical for $l$ and for each $\la \in J$ every trajectory with initial data $\mid x_0 \mid > R$ is non-bounded. \end{pf} Now we pass to the analysis of the non-trapping energy levels in the sense given in \cite{kn:RT1}. Recall that $\la \in \RR$ is a {\em non-trapping energy level} for $l(x,\xi)$ if for every $R>0$ there exists $T_R > 0$ such that for $(x_0,\xi_0) \in \Sigma_\la$, $|x_0| < R$, we have $$ \quad |x(t, x_0,\xi_0)| > R,\:\:\forall | t | > T_R. $$ \begin{lem}\label{lem2} Let $\mu_1 > \mu_0 > 0$ and let $\displaystyle{\lim_{| \xi | \rightarrow \infty} l(x,\xi) = +\infty}$. If there exists $\mu \in ]0,\mu_0 [$ such that \begin{equation}\label{hyp-lem2} (x,\xi) \in l^{-1}(]\mu_0,\mu_1[) \; \Rightarrow \; | \frac14 \{ l, x \cdot \partial_\xi l \} (x,\xi) - l(x,\xi) | \leq \mu, \end{equation} then each $\la \in J$ is a non-trapping energy level for $l(x,\xi)$. \end{lem} \begin{pf} Our aim is to prove that $x(\cdot,x_0,\xi_0)$ is non-bounded in both directions $t \rightarrow \pm \infty$ for any $(x_0,\xi_0) \in \Sigma_\la$. Following the proof of Lemma \ref{lem1}, we have \begin{equation} %\label{esti1} | \Big(\{ l, x \cdot \partial_\xi l \} -4 l \Big)(x(t),\xi(t)) | \leq 4\mu,\:\:\forall t \in \RR, \quad \end{equation} (there is no restriction on $t$ by the assumption on $l$). Then we deduce $$ |x(t)|^2 \geq |x_0|^2 + 2 x_0\cdot \partial_\xi l(x_0,\xi_0)t + 4 (\mu_0-\mu) t^2,\:\:\forall t \in {\RR}.$$ On the other hand the set $\{\xi_0:\:(x_0,\xi_0) \in \Sigma_\la \;, \; \la \leq \mu_1\}$ is bounded. Consequently, for $R>0$ and $ |x_0 | < R$, we conclude that $ x_0\cdot \partial_\xi l(x_0,\xi_0)$ is bounded, and there exists $C_R >0$ such that $$ |x(t)|^2 \geq 4 |t| \Big( ( \mu_0-\mu) |t| - C_R \Big).$$ Thus, as soon as $| t | >2 C_R + 2 \sqrt{C_R^2 + (\mu_0-\mu)R^2}$, we will have $| x(t) | > R$, hence $\la$ is a non-trapping energy level. \end{pf} Introduce a function $\chi \in C_0^\infty(\{x \in \rn ; \; |x| \leq 2 \})$, $0 \leq \chi \leq 1$ such that $\chi(x) = 1$ for $|x| \leq 1$ and set $\chi_R(x)=\chi(\frac{x}{R})$. We have the following \begin{prop}\label{prop2} Let $\mu_1 > \mu_0 > 0$. %and let $\displaystyle{\lim_{| \xi | \rightarrow \infty} l(x,\xi) = %+\infty}$. Assume there exists $\gamma > 0$ such that for every $(\alpha,\beta) \in \N^n \times \N^n,\:\:|\alpha| \leq 1,\:\: |\beta| \leq 2$ we have $$| {\partial}_{x}^{\alpha}{\partial}_{\xi}^{\beta} \Big( l(x,\xi) - | \xi |^2 \Big) | \leq C_{\alpha,\beta} {\langle x \rangle}^{-\gamma- \mid \alpha \mid}{\langle \xi \rangle}^2.$$ Then there exists $R_c>0$ such that for $R \geq R_c$, each $\la \in J$ is a non-trapping and non-critical energy level for the Hamiltonian $$\tilde{l}(x,\xi) ={l}(x,\xi) -\chi_R(x) \Big( l(x,\xi) - | \xi |^2 \Big).$$ \end{prop} \begin{pf} We will take $R>0$ large enough in order to satisfy the conditions of Lemma \ref{lem2} for $$\tilde{l}(x,\xi) =l(x,\xi) -\chi(\frac{x}{R})\Big( l(x,\xi) - | \xi |^2 \Big).$$ First, we have $$\tilde{l}(x,\xi) =| \xi |^2 + \Big(1- \chi(\frac{x}{R}) \Big) \Big( l(x,\xi) - | \xi |^2 \Big),$$ and by the assumption on $l$, we obtain $$\tilde {l}(x,\xi) \geq {\langle \xi \rangle}^2 \Big(1-C_{0,0} {\langle x \rangle}^{-\gamma} \Big) -1.$$ Thus there exists $R_1$ such that for any $R \geq R_1$, $\displaystyle{\lim_{| \xi | \rightarrow \infty} \tl(x,\xi) = +\infty}$ for $|x| \geq R$ and this holds also for $|x| \leq R$, where $\tl(x,\xi) = |\xi|^2.$ Now, as in the proof of the Proposition \ref{prop.1}, we have \begin{eqnarray}\label{tcalcul1} \{ \tl, x \cdot \partial_\xi \tl \} = | \partial_\xi \tl |^2 + \sum_{j=1}^{n} \partial_{\xi_j} \tl \; x \cdot \partial_\xi \partial_{x_j} \tl - \partial_x \tl \cdot \partial_\xi (x \cdot \partial_\xi \tl). \end{eqnarray} Next notice that for $r(x,\xi)= \tl(x,\xi) - | \xi |^2$, the derivative of $\tl= | \xi |^2 + (1- \chi_R)r$ satisfies $$ | \partial_\xi \tl |^2 = 4 | \xi |^2 + 4 (1- \chi_R) \xi \cdot \partial_\xi r + (1- \chi_R)^2| \partial_\xi r |^2$$ and we deduce \begin{equation}\label{tterm1} \frac14 | \partial_\xi \tl |^2 - \tl(x,\xi) = (1- \chi_R) \xi \cdot \partial_\xi r + \frac14(1- \chi_R)^2 | \partial_\xi r |^2 - (1- \chi_R)r(x,\xi). \end{equation} On the other hand, it is clear that $$\partial_{x_j} \tl = \partial_{x_j}(1- \chi_R) r = (1- \chi_R)\partial_{x_j} r - \frac{1}{R} \partial_{x_j} \chi(\frac{x}{R})r, $$ and, consequently, \begin{eqnarray} \sum_{j=1}^{n} \partial_{\xi_j} \tl \; x \cdot \partial_\xi \partial_{x_j} \tl - \partial_x \tl \cdot \partial_\xi (x \cdot \partial_\xi \tl) \nonumber \\ = (1- \chi_R)\left( \sum_{j=1}^{n} \partial_{\xi_j} \tl \; x \cdot \partial_\xi \partial_{x_j} r - \partial_x r \cdot \partial_\xi (x \cdot \partial_\xi \tl)\right) \nonumber\\ - \sum_{j=1}^{n} \frac{1}{R} \partial_{x_j} \chi(\frac{x}{R}) \left( \partial_{\xi_j} \tl \; x \cdot \partial_\xi r - r \cdot \partial_{\xi_j} (x \cdot \partial_\xi \tl)\right).\label{tterm2} \end{eqnarray} The function $(1- \chi_R)$ (resp. $\partial_{x} \chi_R$) is supported in $\{ |x | \geq R \}$, (resp. $\{2R \geq |x | \geq R \}$) and the assumptions on $r(x,\xi)$ combined with the relations (\ref{tcalcul1}), (\ref{tterm1}), (\ref{tterm2}) lead to $$ | \frac14 \{ \tl, x \cdot \partial_\xi \tl \} - \tl (x,\xi) | \leq C {\langle R \rangle}^{-\gamma}{\langle \xi \rangle}^4. $$ Finally, $\xi$ being bounded on the the energy level $\Sigma_{\lambda}$, we can find $R_c=R_c(\mu_0,\mu_1,\gamma) \geq R_1$ such that for any $R \geq R_c$ the function $\tl$ satisfies (\ref{hyp-lem2}). Thus each $\la \in J$ is a non-trapping energy level for $\tl (x,\xi)$. Moreover, taking $R_c$ such that the right hand side of (\ref{tterm1}) is bounded by $\mu< \mu_0$, we obtain $| \partial_\xi \tl | \neq 0$ on $\tl^{-1}(J)$ and, consequently, each $\la \in J$ is a non-critical energy level for $\tl (x,\xi)$. This completes the proof. \end{pf} \begin{rem} For a more complete analysis of the qualitative aspects of classical trajectories, we refer to the recent work of A. Knauf $\cite{kn:K}$. For the Hamiltonian $l(x,\xi)=\frac12 | \xi |^2 + V(x)$, he obtained a topological condition for the existence of trapped trajectories. \end{rem} \section{Resolvent estimates} We start with the following proposition reducing the proof of the resolvent estimates to that for the cut-off resolvent. \begin{prop}\label{prop:bb} Let $L(h)$ satisfy the "black box" assumptions $(\ref{ass1}) - (\ref{ass6})$ and let $J = ]\mu_0,\:\mu_1[ \subset \subset {\RR}^{+}$. Then there exists a self-adjoint differential operator $\tL = \tL(h)$ on $L^2(\RR^n)$, satisfying the assumptions $(\ref{ass1}) - (\ref{ass6})$ such that each $\lambda \in J$ is a non-trapping and non-critical energy level for $\tL(h)$ and $$ L(h) \psi = \tL(h) \psi$$ for any $\psi \in C^\infty({\RR}^n)$ supported away from $B(0,R_0)$. % and such that $\psi(x)=1$ for $x$ sufficiently large. Moreover, for every $s > \frac12$ there exists $C > 0,\: h_0 > 0$ and $\rho > R_0$ such that for $\chi \in C_0^\infty(\RR^n)$ with $\chi(x) = 1$ for $|x| \leq \rho$, we have $$\| R(z) \|_{s,-s} \leq C h^{-2} \Big( 1 + \| \chi R(z) \chi \|_{{\cal H} \to {\cal H}} \Big),$$ uniformly with respect to $z \in {\cal B}_{\pm} = \{z \in \C;$ $ (\Re z,\pm \Im z) \in J\times ]0, 1]\}$ and $h \in (0,h_0]$. \end{prop} \begin{pf} Choose a function $\Phi \in C_0^\infty(\{x \in \rn ; \; |x| \leq 2 \}; \RR)$, $0 \leq \Phi \leq 1$ such that $\Phi(x) = 1$ for $|x| \leq 1$ and set $\chi_R(x)=\Phi(\frac{x}{R})$ for $R > R_0$, where $R_0$ is the constant in the "black box" assumptions. Introduce the self-adjoint operator on $L^2(\RR^n)$ \begin{eqnarray}\label{deftL} \tL & = & (1 - \chi_R) L \chi_R + \chi_R L (1- \chi_R) + (1 - \chi_R) L (1- \chi_R) + \chi_R (-h^2 \Delta ) \chi_R \\ & =& Q + \chi_R (-h^2 \Delta - Q ) \chi_R, \end{eqnarray} where $Q$ is the formally self-adjoint differential operator given in the Introduction. According to Proposition \ref{prop2}, there exists $R_c >R_0$ such that if $R \geq R_c$, every $\lambda \in J$ is a non-trapping and non-critical energy level in the sense mentioned in Section 2 for the principal symbol $\tilde{l}_0(x,\xi)$ of $\tL(h)$ having the form $$\tilde{l}_0(x,\xi) ={l}_0(x,\xi) -\chi^2_R(x) \Big( l_0(x,\xi) - | \xi |^2 \Big)$$ with ${\DS{l_0(x,\xi) = \sum_{| \a | =2} a_\a (x) \xi^\a}}$. Clearly, for $\psi=0$ on the support of $\chi_R$, we have $ \tL \psi = Q \psi = L \psi $. Let $\chi_1 \in C_0^\infty(\RR^n)$ be equal to $1$ on the support of $\chi_R$ and let $\chi_2 \in C_0^\infty(\RR^n)$ be equal to $1$ on the support of $\chi_1$. Using $\tL (1-\chi_1) = L (1-\chi_1)$, we have \begin{equation}\label{reseq} R(z) = R(z) \chi_2 + (1-\chi_1) \tR (z) (1-\chi_2) + R (z) [\tL,\chi_1]\tR(z)(1-\chi_2), \end{equation} where $$ \tR(z) = (\tL(h) - z )^{-1},\:\: z \in {\C},\:\Im z \neq 0.$$ Thus we obtain \begin{equation}\label{reseq1} \| R(z) \|_{s,-s} \leq C \Big( \| R(z) \chi_2 \|_{s,-s} + \| \tR(z) \|_{s,-s} + \| R(z) \chi_2 \|_{s,-s} \| [\tL,\chi_1]\tR(z) \|_{s,s} \Big), \end{equation} where here and below we denote by $C > 0$ different constants independent on $h$ and $z$. Next, $[\tL,\chi_1]$ is a first order $h$-admissible differential operator with compactly supported coefficients, so there exists $C > 0$ such that $$ \| [\tL,\chi_1]\tR(z) \|_{s,s} \leq C \| [\tL,\chi_1]\tR(z) \|_{s,-s}.$$ Moreover, the interval $J$ contains non-trapping energy levels for $\tL(h)$, and we can estimate $\|\tR(z)\|_{s,-s}$ and $\| [\tL,\chi_1]\tR(z) \|_{s,-s}$ by Lemma \ref{lem4} and Corollary \ref{corlem4} (see Section 5.1). Consequently, there exist $C>0$ and $h_0>0$ such that for any $h \in (0,h_0]$ we have \begin{equation}\label{reseq2} \| R(z) \|_{s,-s} \leq C h^{-1} \Big( 1 + \| R(z) \chi_2 \|_{s,-s} \Big) . \end{equation} On the other hand, using the equality $$ \| R(z) \chi_2 \|_{s,-s} = \| \chi_2 R({\overline z}) \|_{s,-s}$$ and introducing $\chi_2$ in the estimates (\ref{reseq1}), (\ref{reseq2}), we obtain \begin{equation}\label{reseq3} \| R(z) \chi_2 \|_{s,-s} \leq C h^{-1}\Big( 1 + \| \chi_2 R(z) \chi_2 \|_{{\cal H} \to {\cal H}} \Big). \end{equation} Finally, (\ref{reseq2}) and (\ref{reseq3}) yield $$ \| R(z) \|_{s,-s} \leq C h^{-2}\Big( 1 + \| \chi_2 R(z) \chi_2 \|_{{\cal H} \to {\cal H}} \Big). $$ \end{pf} \begin{prop}\label{prop:lapbb} Assume that $L(h)$ satisfies the assumptions $(\ref{eq:1.7})-(\ref{eq:1.13})$. Then for small $h_0 > 0$ and fixed $s > 1/2,\: h \in (0, h_0]$ the limits $$R(\la \pm i0) = \lim_{\Im z \to 0, \: \: \pm \Im z > 0} R(z), \quad \Re z=\la \in J,$$ exist in ${\cal L}({\cal H}^{0,s},{\cal H}^{0,-s})$ uniformly with respect to $\lambda \in J.$ \end{prop} \begin{pf} Take a cut-off function $\chi \in C_0^{\infty}({\RR}^n)$ equal to 1 on $B(0,R_0)$. We shall prove that there exists $h_0 > 0$ such that for fixed $h \in (0,h_0]$ and $s > 1/2$, we have $$\| R(z) - R(z') \|_{s,-s} \leq C_{h,s} \Big( | z- z'|^{\delta_1} + | z- z'| + \|\chi \Big( R(z) - R(z') \Big) \chi \|_{{\cal H} \to {\cal H}}\Big),$$ with $\delta_1 \in (0,1/2)$ uniformly with respect to $z,z' \in {\cal B}_{\pm}$. As in the proof of the Proposition \ref{prop:bb}, we choose $R > R_c$ and introduce $\chi_R$ and an operator $\tL = \tL(h)$ with non-trapping energy levels in $J$. Let $\chi_1 \in C_0^\infty(\RR^n)$ be equal to $1$ on the support of $\chi_R$ and let $\chi_2 \in C_0^\infty(\RR^n)$ be equal to $1$ on the support of $\chi_1$. Set $ R(z,z') = R(z) - R(z')$ and $ \tR(z,z') = \tR(z)-\tR(z')$. According to (\ref{reseq}), we have $$\begin{array}{ccl} R(z,z') = R(z,z') \chi_2 + (1-\chi_1) \tR(z,z') (1-\chi_2)\\ + R(z,z') [\tL,\chi_1]\tR(z)(1-\chi_2) + R(z') [\tL,\chi_1]\tR(z,z') (1-\chi_2). \end{array}$$ and we deduce $$\begin{array}{ccl} \| R(z,z') \|_{s,-s} \leq C \| R(z,z') \chi_2 \|_{s,-s} \Big( 1+ \| [\tL,\chi_1]\tR(z) \|_{s,s}\Big) \\ + C \| \tR(z,z') \|_{s,-s} + C\| R(z') \chi_2 \|_{s,-s}\; \| [\tL,\chi_1]\tR(z,z') \|_{s,s}. \end{array}$$ For fixed $h$ Lemma \ref{lem6} in Section 5.2 implies that the cut-off resolvent $\chi(x)R(z)\chi(x)$ is analytic in $\overline{{\cal B}_{+}}$ and the norm $\|\chi_2 R(z) \chi_2\|_{{\cal H} \to {\cal H}}$ will be uniformly bounded with respect to $z \in {\cal B}_{\pm}$. Applying Proposition \ref{prop:bb} for $s>1/2$, we conclude that $\|R(z') \chi_2 \|_{s,-s}$ is also uniformly bounded with respect to $z \in {\cal B}_{\pm}.$ On the other hand, since $J$ is non-trapping for $\tL$, we can apply Corollary \ref{corlem4}. Thus, as in the proof of Proposition \ref{prop:bb}, for any $s > 1/2$ there exists $C_s$ such that $$\| [\tL,\chi_1] \tR(z) \|_{s,s} \leq C_s h^{-1}$$ uniformly with respect to $z \in {\cal B}_{\pm}$. Moreover, for $s$ and $h$ fixed, there exists $C_{h,s}$ such that $$\| \tR(z,z') \|_{s,-s}+ \|[\tL,\chi_1] \tR(z,z') \|_{s,-s} \leq C_{h,s}\Big(| z - z' |+ | z - z' |^{\delta_1}\Big),$$ uniformly with respect to $z,z' \in {\cal B}_{\pm}$. Consequently, there exists $C_{h,s} > 0$, such that $$\begin{array}{ccl} \| R(z,z') \|_{s,-s} & \leq & C_{h,s} \Big( |z-z'|^{\delta_1}+ |z-z'| + \| R(z,z') \chi_2 \|_{s,-s}\Big). \end{array}$$ Next, using the equality $ \| R(z,z') \chi_2 \|_{s,-s} = \| \chi_2 R({\overline z},{\overline {z'}}) \|_{s,-s}$, and introducing the cut-off function $\chi_2$ in the above estimates, we obtain $$\| R(z,z') \chi_2\|_{s,-s} \leq C_{h,s} \Big( |z-z'|^{\delta_1}+ |z-z'| + \|\chi_2 R(z,z') \chi_2 \|_{s,-s}\Big).$$ This implies immediately $$\| R(z,z')\|_{s,-s} \leq C_{h,s} \Big( |z-z'|^{\delta_1}+ |z-z'| + \| \chi_2 R(z,z') \chi_2\|_{s,-s} \Big).$$ From Lemma 6 we deduce the analyticity of $\chi_2R(z,z')\chi_2$ so for fixed $h$ and $s > 1/2$, the limits $$R(\la \pm i0)= \lim_{\Im z \to 0, \: \: \pm \Im z > 0} R(z), \quad \Re z=\la \in J$$ exist in ${\cal L}({\cal H}^{0,s},{\cal H}^{0,-s})$. \end{pf} For the proof of our principal result we need the following \begin{lem} \label{lem3} Let $L(h)$ satisfy the assumptions $(\ref{eq:1.7})-(\ref{eq:1.13})$ and assume the estimate $(\ref{eq:1.5})$ fulfilled. Then for every $\chi \in C_0^\infty(\RR^n)$ equal to 1 on $B(0,R_0)$ there exist $C>0$ such that for $(\lambda, \tau) \in J\times ]0,1]$ we have \begin{equation} \|\chi(x) R(\lambda \pm i\tau)\chi(x) \|_{{\cal H} \to {\cal H}} \leq Ce^{Ch^{-p}},\:0 < h \leq h_0. \label{eq:3.2} \end{equation} \end{lem} \begin{pf} Set $S_{\theta} = \{ z \in {\C}: -2\theta < \arg z \leq \pi/4 \},$ where $0 < \theta \leq \theta_0$ is fixed and $\theta_0 \in ]0, \pi[$ was introduced in the assumption (\ref{eq:1.11}) making possible a complex scaling of $L(h).$ Choose $\mu/2 > \delta > 0$ so that (\ref{eq:1.5}) holds for $\lambda \in [\mu_0 - \delta, \mu_1 + \delta]$ and consider the compact set \[ \omd = \{ z \in S_{\theta}: \mu_0 - \delta \leq \Re z \leq \mu_1 + \delta \}\,. \] Then for every $m \in {\N}$, Lemma 1 in \cite{kn:TZ} yields \begin{equation} \label{eq:estre} \|\chi(L(h) - z)^{-1}\chi \| \leq C(\omd, m) \exp\left(C(\omd,m)h^{-k}\right),\: 0 < h \leq h_0(\omd,m) \end{equation} for $$z \in \Omega_\delta \setminus \bigcup_{z_j \in {\rm Res}\: L(h) \cap \omd} D(z_j : h^m)$$ with $k \geq n$ and $D(z_j : h^m) = \{ z \in \C: |z - z_j| \leq h^m \}.$ For the number of the scattering resonances we have the estimate (see \cite{kn:SZ}, \cite{kn:Sj1}) \[ \sharp \{ z_j: z_j \in {\rm Res}\: L(h) \cap \omd \} \leq C_1(\omd) h^{-n^{\sharp}} \] with $n^{\sharp} \geq n.$ Next we fix an integer $m$ satisfying $2m \geq n^{\sharp} + 1$ and take $0 < h_1(\omd,m) \leq h_0(\omd,m)$ small enough to arrange \[4 C_1(\omd) h^{2m - n^{\sharp}} < \delta \, \] for $0 < h \leq h_1(\omd, m).$ This implies immediately the existence of $\gamma_1(h),\:\gamma_2(h)$ so that \[ \mu_0/2 \leq \mu_0 - \delta < \gamma_1(h) < \mu_0 - \frac{\delta}{2},\:\: \mu_1 +\frac{\delta}{2} < \gamma_2(h) < \mu_1 + \delta\,, \] and \[ \gamma_j (h) + i \tau \notin \bigcup_{z_j \in {\rm Res}\: L(h) \cap \omd} D(z_j : h^m),\:\:j =1,2, \: \tau \geq 0\,.\] Thus for $k > n$ fixed the estimate (\ref{eq:estre}) holds for $z = \gamma_j(h) +i\tau,\:\:j =1,2,\:\: \tau \geq 0.$ Set $\epsilon = \frac{\delta}{10}$ and introduce the domain \[ \Pi_{\delta} = \{ z = \lambda + i \tau: \mu_0- \delta \leq \lambda \leq \mu_1 + \delta,\:\: 0 \leq \tau \leq h^M \} \subset \omd \,, \] where $2M > k.$ By a modification of the argument in \cite{kn:TZ}, Lemma 2, we may construct a function $f(z,h)$ holomorphic in $\Pi_{\epsilon}$ with the properties:\\ (i) $|f(z,h)| \leq e$ in $\Pi_{\delta}$,\\ (ii) $|f(z,h)| \geq 1/2$ in $[\mu_0 - \epsilon, \mu_1 + \epsilon] + i[0, h^M],$\\ (iii) for $z \in \Pi_{\delta} \cap \{z: \Re z \leq \mu_0 -4\epsilon,\: {\rm or}\: \Re z \geq \mu_1 + 4\epsilon\}$ we have $$|f(z,h)| \leq C_{\epsilon}\exp\Bigl(-\frac{\epsilon^2}{2h^{2M}}\Bigr), \: 0 \leq h \leq h(\epsilon).$$ To construct $f(z,h)$, take a function $\psi(x) \in C^{\infty}_0({\Bbb R}^n),\: 0 \leq \psi(x) \leq 1,$ such that $$\psi(x) = \left \{\aligned 0,\:\: x \notin [\mu_0 - 3\epsilon, \mu_1 + 3\epsilon],\\ 1, \:\: x \in [\mu_0 -2\epsilon, \mu_1 + 2\epsilon] \endaligned \right.$$ and put $$f(z,h) = (\pi \alpha^2)^{-1/2} \int \exp \Bigl( -\frac{(x-z)^2}{\alpha^2}\Bigr)\psi(x) dx,\:\alpha = h^M.$$ The check of (i) is trivial. For (ii) write $z = u +iv$ and take $u \in [\mu_0 -\epsilon, \mu_1 + \epsilon]$. Then $$|f(z,h) - 1| \leq (\pi \alpha^2)^{-1/2} \exp\Bigl(\frac{v^2}{\alpha^2}\Bigr) \int \exp\Bigl(- \frac{(x-u)^2}{\alpha^2}\Bigr) |\psi(x) -1 |dx$$ $$\leq \pi^{-1/2}e \int e^{-y^2} |\psi(\alpha y + u) - 1| dy \leq 2\pi^{-1/2}e \int_{{\Bbb R} \setminus B(0, \frac{\epsilon}{\alpha})} e^{-y^2} dy \leq \frac{1}{2},$$ for $0 < h \leq h(\epsilon)$ since $\frac{\epsilon}{\alpha} \longrightarrow \infty$ as $h \to 0.$ Finally, to deal with (iii), assume for example that $u \leq \mu_0 - 4\epsilon..$ We have $$|f(z,h)| \leq (\pi \alpha^2)^{-1/2} \exp\Bigl(\frac{v^2}{\alpha^2}\Bigr) \int \exp\Bigl(-\frac{(x-u)^2}{\alpha^2}\Bigr) \psi(x)dx$$ $$\leq A\pi^{-1/2} h^{-M}\exp \Bigl(-\frac{\epsilon^2}{\alpha^2}\Bigr) \leq C_{\epsilon}\exp \Bigl(-\frac{\epsilon^2}{2\alpha^2}\Bigr), \: 0 \leq h \leq h(\epsilon).$$ Our assumptions imply that the cut-off resolvent $$F(z,h) = \chi(x) R(z) \chi(x)$$ has no poles on the real axis (see Appendix 5.2). Thus the function $G(z,h) = f(z,h)F(z,h)$ is holomorphic in $\Pi_{\epsilon}$. For $\Im z = 0$ we have $\|G(z,h)\| \leq eC e^{\frac{C}{h^p}}$, while for $\Im z = h^M$ we get $$\|G(z,h)\| \leq \frac{Ce}{h^M} \leq e^{\frac{C_1}{h}}.$$ For $z \in \Pi_{\delta},\: \Re z = \gamma_j(h),\:\:j =1,2$, combining the estimates for $|f(z,h)|$ and $\|F(z,h)\|,$ we obtain $$\|G(z,h)\| \leq C_{\epsilon}e^{\frac{C}{h^k}} e^{-\frac{\epsilon^2}{h^{2M}}} \leq C_2,\:\: 0 R, \; d^{-1} < |\xi | < d, \; \frac{\langle x.\xi \rangle}{|x||\xi|} \; ^>_< \; \sigma_\pm \}, \end{equation} where $d>1$, $-1 < \sigma_\pm < 1$ and $R \gg 1$ is large enough. Consider a symbol $ \omega_\pm(x,\xi) \in C^\infty (\RR^{2n})$ such that supp $\omega_\pm \subset \Gamma^\pm (R,d,\sigma_\pm)$ and \begin{equation}\label{defomegapm} | \partial_x^\alpha \partial_\xi^\beta \omega_\pm (x,\xi) | \leq C_{\alpha,\beta,L} \langle x \rangle^{-|\alpha |} \langle \xi \rangle^{- L } \end{equation} for any $L \gg 1$. Then we have the following \begin{prop}\label{prop5} Let $J \subset \subset \RR^+$ and $\psi \in C_b^\infty({\RR}^n)$ be supported away from $B(0,R_0)$. Then there exists $\chi \in C_0^\infty({\RR}^n),\: \chi = 1$ on $B(0,R_0)$ such that for any $(\la,\tau) \in J \times [0,1]$ and $h \in (0,h_0] $ the following assertions hold: i) For any $s > 1/2$, $\delta >1$ there exist $C > 0$ and $h_0 > 0$ such that for $h \in (0,h_0]$ we have \begin{equation}\label{eq:mltrap.es1} \|\psi R(\la\pm i \tau)\psi \omega_\pm(x,hD_x) \|_{-s+\delta,-s} \leq C h^{-2} ( 1 + \| \chi R(\la\pm i \tau) \chi \|_{{\cal H} \to {\cal H}}). \end{equation} ii) If $\sigma_+ > \sigma_-$, then for any $s \gg 1 $, there exist $C > 0$ and $h_0 > 0$ such that for $h \in (0,h_0]$ we have \begin{equation}\label{eq:mltrap.es2} \| \omega_\mp(x,hD_x) \psi R(\la\pm i \tau)\psi \omega_\pm(x,hD_x) \|_{-s,s} \leq C h^{-2} ( 1 + \| \chi R(\la\pm i \tau) \chi \|_{{\cal H} \to {\cal H}}). \end{equation} \end{prop} \begin {pf} As in the proof of the Proposition \ref{prop:bb}, there exists a self-adjoint differential operator $\tL = \tL(h)$ on $L^2(\RR^n)$, such that each $\lambda \in J$ is a non-trapping energy level for $\tL$ and the equality (3.3) holds. Then $i)$ is a consequence of Proposition \ref{prop:bb} and $i)$ of Lemma \ref{lem5}. We use also the resolvent estimate (as in Corollary \ref{corlem4}) and the fact that $ \langle x \rangle^{-s} \omega_\pm(x,hD_x)\langle x \rangle^{s}$ is a bounded pseudo-differential operator for any $s \in \RR$ (see \cite{kn:DS}, \cite{kn:R1}). According to (\ref{reseq}) we have $$\begin{array}{ccl} \omega_\mp(x,hD_x) \psi R(z) \psi \omega_\pm(x,hD_x) = \omega_\mp(x,hD_x) \psi R(z) \chi_2 \Big(1 +[\tL,\chi_1]\tR(z)(1-\chi_2)\Big) \psi \omega_\pm(x,hD_x)\\ + \omega_\mp(x,hD_x) (1-\chi_1) \tR(z) (1-\chi_2) \omega_\pm(x,hD_x). \end{array}$$ However, $\omega_\mp(x,hD_x) \psi R(z) \chi_2 = \Big( \chi_2 R(\overline{z}) \psi \omega_\mp(x,hD_x)^* \Big)^*$, then using (\ref{reseq}), the Proposition \ref{prop:bb} and Lemma \ref{lem5} once more we obtain that the operator $\omega_\mp(x,hD_x) \psi R(\la\pm i \tau) \chi_2 $ exists in ${\cal L}(L^2_{-s},L^2_{s})$ for $(\la,\tau) \in J \times [0,1]$ and following estimates hold $$ \| \omega_\mp(x,hD_x) \psi R(\la\pm i \tau) \chi_2 \|_{-s,s} \leq C h^{-1} ( 1 + \| \chi_2 R(\la\pm i \tau) \chi_2 \|_{{\cal H} \to {\cal H}}).$$ Consequently, according to Lemma \ref{lem5}, $\omega_\mp(x,hD_x) \psi R(\la\pm i \tau)\psi \omega_\pm(x,hD_x)$ exists in ${\cal L}(L^2_{-s},L^2_{s})$ for $(\la,\tau) \in J \times [0,1]$ and these operators satisfy the claimed estimate (\ref{eq:mltrap.es2}). \end{pf} \section{Appendix }\label{sec.appendix} \subsection{Resolvent estimates in the non-trapping case} In this section we recall some resolvent estimates in a non-trapping energy interval for a self-adjoint differential operator $\tL = \tL(h)$ defined in a dense domain in $L^2({\RR}^n).$ We say that $J \subset \subset \RR^+$ is non-trapping for $\tL(h)$ if every $\la \in J$ is a {\em non-trapping energy level} for the principal symbol of $\tL(h)$ in the sense of Section 2. Denote by $\| . \|_{s,s'}$ the norm of ${\cal L}(L^{2,s}, L^{2,s'})$, where $L^{2,s}$ is the weight space $L^2(\RR^n, \langle x \rangle^s dx)$. \begin{lem}\label{lem4} Let $\tL = \tL(h)$ be a self-adjoint differential operator having the form $(\ref{ass4})$ and satisfying $(\ref{ass5})-(\ref{ass6})$. Let $J \subset \subset \RR^+$ be an open non-trapping interval for $ \tL(h)$ and let $s > 1/2.$ Then i) for $h\in (0,h_0]$ fixed, there exists $C_{h,s} > 0$ such that for any $z,z' \in {\cal B}_{\pm}$ we have $$\| (\tL-z)^{-1}-(\tL-z')^{-1} \|_{s,-s} \leq C_{h,s} | z - z' |^{\delta_1}, \quad \delta_1 = (s-1/2)/(2s-1/2).$$ ii) for $\la \in J$, the limits $$(\tL-\la \mp i0)^{-1}:= \lim_{\Im z \to 0, \: \: \pm \Im z > 0} (\tL-z)^{-1}, \quad \Re z=\la,$$ exist in ${\cal L}(L^{2,s}, L^{2,-s})$, and there exists $C > 0$ and $h_0 > 0$ such that for $z \in {\cal B}_{\pm}$ and $h \in (0,h_0]$ we have \begin{equation}\label{eq:trap.es} \|(\tL-z)^{-1} \|_{s,-s} \leq C h^{-1}. \end{equation} \end{lem} \begin{pf} This result can be obtained following the Mourre theory. First, there exists a {\em conjugate operator} $A = A(h)$ to $\tL$ satisfying the following Mourre inequality: $\forall \Phi \in C_0^\infty(J)$, $\exists \gamma_0>0$ such that for all $h \in (0,h_0]$, we have \begin{equation}\label{eq:mourreesti} \Phi(\tL) i^{-1} [\tL,A ] \Phi(\tL) \geq \gamma_0 h \Phi(\tL)^2. \end{equation} For the construction of $A(h)$ we refer to \cite{kn:R3}, where a more general setup has been considered. The operator $A(h)$ has the form of a $h$-pseudodifferential operator with symbol $$F(x,\xi) = K(x,\xi) + \frac{2 x.\xi}{\langle \xi \rangle^2},$$ where $K \in C^\infty(\RR^n \times \RR^n)$ is compactly supported with respect to $x$ and $K$ is uniformly bounded with respect to $(x,\xi)$. According to Mourre type results, we deduce that $\tL$ has no eigenvalues in $J$ for $h$ small enough (see \cite{kn:M}, \cite{kn:CFKS}) and for $s>1/2$ the following assertions hold (see \cite{kn:PSS}, \cite{kn:JMP} and \cite{kn:HN},\cite{kn:R2}, \cite{kn:R3} for operators depending on $h$): i') for $h\in (0,h_0]$ fixed, there exists $C_h>0$ such that for $z,z' \in {\cal B}_{\pm} = \{z \in \C;$ $ (\Re z,\pm \Im z) \in J\times ]0, 1]\}$ we have $$\| \langle A \rangle^{-s}\Big( (\tL-z)^{-1}-(\tL-z')^{-1}\Big) \langle A \rangle^{-s} \|_{L^2 \to L^2} \leq C_h | z - z' |^{\delta_1} % , \quad \delta_1 = (s-1/2)/(2s-1/2) .$$ ii') there exist $C>0$ and $h_0>0$ such that for $z \in {\cal B}_{\pm}$ and $h \in (0,h_0]$ we have \begin{equation}\label{eq:trap.es'} \|\langle A \rangle^{-s} (\tL-z)^{-1} \langle A \rangle^{-s} \|_{L^2 \to L^2} \leq C h^{-1}, \: = (1 + A^2)^{1/2} \end{equation} Let us note that the operators $ \langle x \rangle^{-s} \langle A \rangle^{s},\: \langle A \rangle^{s} \langle x \rangle^{-s} $, $s \geq 0$, are uniformly bounded with respect to $h$ in ${\cal L}(L^2)$. This follows from the functional calculus developed by Dimassi-Sj\"ostrand (see Sections 7 and 8 in \cite{kn:DS}) or by using a complexe interpolation as in Lemma 8.2 of \cite{kn:PSS}. Then, writing $$\langle x \rangle^{-s} (\tL-z)^{-1} \langle x \rangle^{-s}= \langle x \rangle^{-s} \langle A \rangle^{s}\; \langle A \rangle^{-s} (\tL-z)^{-1} \langle A \rangle^{-s} \; \langle A \rangle^{s} \langle x \rangle^{-s},$$ we complete the proof. \end{pf} \begin{cor}\label{corlem4} Let $P$ be a $p$-order $h$-admissible differential operator, $p\leq 2$ and let $s > 1/2.$ Then, under the assumptions of the lemma \ref{lem4} the following assertions hold: i'') for $h\in (0,h_0]$ fixed, there exists $C_{h,s} > 0$ such that for $z,z' \in {\cal B}_{\pm}$ we have $$\| P \Big( (\tL-z)^{-1}-(\tL-z')^{-1} \Big) \|_{s,-s} \leq C_{h,s}\Big(| z - z' |+ | z - z' |^{\delta_1}\Big) % \quad \delta_1 = (s-1/2)/(2s-1/2) .$$ ii'') there exists $C > 0$ and $h_0 > 0$ such that for $z \in {\cal B}_{\pm}$ and $h \in (0,h_0]$ we have \begin{equation}\label{eq:trap.es} \| P (\tL-z)^{-1} \|_{s,-s} \leq C h^{-1}. \end{equation} \end{cor} \begin{pf} The ellipticity of $\tL$ and the functional calculus developed by Dimassi-Sj\"ostrand (see Sections 7 and 8 in \cite{kn:DS}) imply that the operator $\langle x \rangle^{-s}P(\tL-i)^{-1}\langle x \rangle^{s}$ is uniformly bounded in ${\cal L}(L^2)$ with respect to $h$, for $s \geq 0$. Then ii'') is a direct consequence of ii) and the resolvent equation $$(\tL-z)^{-1} = (\tL-i)^{-1} + (z-i)(\tL-i)^{-1}(\tL-z)^{-1}.$$ In the same way i'') is a direct consequence of i) and the relation $$(\tL-z)^{-1}-(\tL-z')^{-1} = (z-z') (\tL-i)^{-1}(\tL-z)^{-1} + (z'-i)(\tL-i)^{-1}\Big((\tL-z)^{-1}-(\tL-z')^{-1}\Big).$$ \end{pf} Finally, we formulate some microlocal resolvent estimates which generalize Lemma 2.3 of \cite{kn:RT2}. \begin{lem}\label{lem5} Let $\tL = \tL(h)$ be a self-adjoint differential operator having the form $(\ref{ass4})$ and satisfying $(\ref{ass5})-(\ref{ass6})$. Let $J \subset \subset \RR^+$ be an open non-trapping interval for $ \tL(h)$ and let $ \omega_\pm(x,\xi)$ be a symbol supported in an outgoing (incoming) region as in Section 4. Then for any $(\la,\tau) \in J \times [0,1]$ the following assertions hold: i) For any $s > 1/2$, $ \delta >1$ there exist $C > 0$ and $h_0 > 0$ such that for $h \in (0,h_0]$ we have \begin{equation}\label{eq:mlntrap.es1} \|(\tL-\la\mp i\tau)^{-1} \omega_\pm(x,hD_x) \|_{-s+\delta,-s} \leq C h^{-1}. \end{equation} ii) If $\sigma_+ > \sigma_-$ then for any $s \gg 1 $, there exist $C > 0$ and $h_0 > 0$ such that for $h \in (0,h_0]$ we have \begin{equation}\label{eq:mlntrap.es2} \| \omega_\mp(x,hD_x) (\tL-\la\mp i\tau)^{-1} \omega_\pm(x,hD_x) \|_{-s,s} \leq C h^{N}. \end{equation} for any $N \in \N$. \end{lem} The proof of this lemma is analogous to the proof of Lemma 2.3 of \cite{kn:RT2} exploiting the constructions of long time approximations in section 4 of \cite{kn:R3} for a large class of long range perturbations of the Laplacian. The essential point is the construction of the phase $\varphi_\pm$ (see Proposition 4.1 of \cite{kn:R3}). \subsection{Boundary values of the cut-off resolvent $\chi R(z) \chi$} In this section we will show that for fixed $h \in (0, h_0]$ the limits $$\chi(x) R(\lambda \pm i0) \chi(x) = \lim _{\Im z \to 0, \: \: \pm \Im z > 0} \chi(x) R(z) \chi(x),\: \lambda \in J$$ exist in ${\cal L}({\cal H},\: {\cal H}),$ provided $\chi \in C_0^{\infty}({\RR}^n)$ is equal to 1 on $B(0, R_0)$ and $\chi(x) = 0$ for $|x| \geq R_0 + a_0,\: a_0 > 0.$ To do this it is sufficient to show that there are no resonances $z_j$ in the interval $]\mu_0 -\delta, \mu_1 + \delta[.$ \begin{lem}\label{lem6} Assume $(\ref{eq:1.7})-(\ref{eq:1.13})$ fulfilled. Then for $0 < h \leq h_0$ we have $${\rm Res} \: L(h) \cap ]\mu_0 -\delta, \mu_1 + \delta[ = \emptyset.$$ \end{lem} \begin{pf} The proof can be obtained from the results in \cite{kn:SZ}, \cite{kn:Sj1}. Here we will only give indications of how to extract it from there. We assume $h$ fixed and let $z_0 \in J$ be a resonance for $L(h)$. Denote by ${\cal H}_{R_0 +a_0}$ the space of elements of ${\cal H}$ that vanish outside $B(0, R_0 + a_0).$ Then by using a complex scaling argument and Lemma 3.5 in \cite{kn:SZ}, we conclude that for $v \in {\cal H}_{R_0 + a_0}$ and $z$ in a neighborhood of $z_0$ we have \begin{equation} (L - z)^{-1}v = \sum_{j =1}^N (z_0 - z)^{-j}v_{-j} + G(z) \label{eq:4.1} \end{equation} with $G(z)$ holomorphic in a neighborhood of $z_0$ and $N$ independent of $v$. Consider the spectral projection $$\pi_{\theta, z_0} = \frac{1}{2\pi i} \int_{\gamma(z_0)} (z - L_{\theta})^{-1}dz,$$ $L_{\theta}$ being the operator obtained by complex dilation defined in a dense set ${\cal D}_{\theta}$ of the space ${\cal H}_{\theta} = {\cal H} \oplus L^2(\Gamma_{\theta} \setminus B(0, R_0))$ with $0 <\theta \leq \theta_0$ (see \cite{kn:Sj1}), while $\gamma(z_0):\: [0,2\pi] \ni s \longrightarrow z_0 + \epsilon e^{is}$ with $ \epsilon > 0$ small enough. Let $F_{\theta, z_0}$ be the image $\pi_{\theta, z_0}({\cal H}_{\theta})$ and let $$\pi_{0,z_0} = \frac{1}{2\pi i} \int_{\gamma(z_0)} (z - L)^{-1}dz: \: {\cal H}_{{\rm comp}} \longrightarrow {\cal D}_{{\rm loc}}.$$ Then exploiting the equality (\ref{reseq}) and the fact that the resolvent $\tilde R(z)$ is holomorphic in a small neighborhood of $z_0$, as in Proposition 3.6 in \cite{kn:SZ}, we conclude that there exists a bijection between the spaces $F_{\theta, z_0}$ and $ \pi_{0,z_0}({\cal H}_{R_0 + a_0}).$ Since $z_0$ is a resonance, the space $F_{\theta, z_0}$ is not trivial and we can find $v \in {\cal H}_{R_0 + a_0}$ so that (\ref{eq:4.1}) holds with $\sum_{j =1}^N (z_0 - z)^{-j}v_{-j} \neq 0$. For simplicity assume that $v_{-j} = 0$ for $j = 2,..., N$ and let $v_{-1} \neq 0.$ Take a cut-off function $\Phi(x) \in C_0^{\infty}({\RR}^n)$ equal to 1 on $B(0, R_0)$ so that $(\Phi, v_{-1})_{{\cal H}} \neq 0.$ Therefore, $$\lim_{\epsilon \to 0} (\Phi, i\epsilon(L - z_0 + i\epsilon)^{-1}v_{-1})_{\cal H} = (\Phi, P_{\{z_0\}}v_{-1})_{\cal H} \neq 0,$$ $P_{\{z_0\}}$ being the spectral projector of $L$ related to $z_0$, and we get $z_0 \in \sigma_{pp}(L)$ which contradicts (\ref{eq:1.13}). The case of a multiple pole can be treated in the same way. \end{pf} {\footnotesize \begin{thebibliography}{99} \bibitem{kn:Br} V. Bruneau, {\em Semi-classical behavior of the scattering phase for trapping perturbations of the Laplacian}, Comm. in P.D.E., {\bf 24}, 5 \& 6, (1999),1095-1125. \bibitem{kn:Bu} N. Burq, {\em Absence de r\'esonances pr\`es du r\'eel pour l'op\'erateur de Schr\"odinger}, Expos\'e XVII, S\'eminaire EDP, Ecole Polytechnique, 1997/1998. \bibitem{kn:CFKS} H. L. Cycon, R. G. Froese, W. Kirsch, B. Simon, {\em Schr\"{o}dinger operators with application to quantum mechanics and global geometry}, Texts and Monographs in Physics, Springer-Verlag, Berlin, Heidelberg, 1987. \bibitem{kn:DS} M. Dimassi, J. Sj\"ostrand, {\em Spectral asymptotics in the semi-classical limit}, Lecture Notes, (to appear). \bibitem{kn:GM1} C. G\'erard, A. Martinez, {\em Principe d'absorption limite pour des op\'erateurs de Schr\"odinger \`a longue port\'ee}, C. R. Acad. Sci. Paris, {\bf 306} (1988), 121-123. \bibitem{kn:GM2} C. G\'erard, A. Martinez, {\em Semi-classical asymptotics of the spectral function of long range Schr\"{o}dinger operators}, J. Funct. Anal. {\bf 84} (1989), 226-254. \bibitem{kn:GMR} C. G\'erard, A. Martinez and D. Robert, {\em Breit-Wigner formulas for the scattering poles and total scattering cross-section in the semi-classical limit}, Comm. Math. Phys. {\bf 121} (1989), 323-336. \bibitem{kn:HN} P. Hislop, S. Nakamura, {\em Semiclassical resolvents estimates}, Ann. Inst. H. Poincar\'e (Physique th\'eorique), {\bf 51} (1989), 187-198. \bibitem{kn:IK} H. Isozaki, H. Kitada, {\em Modified wave operators with time independent modifiers}, J. Fac. Sc. Univ. Tokyo Sect IA. {\bf 32} (1985), 77-104. \bibitem{kn:JMP} A. Jensen, E. Mourre, P. Perry, {\em Multiple Commutator Estimates and Resolvent Smoothness in Quantum Scattering Theory}, Ann. Inst. H. Poincar\'e (Physique th\'eorique) {\bf 41} (1984), 207-225. \bibitem{kn:K} A. Knauf, {\em Qualitative Aspects of Classical Potential Scattering}, Preprint, 1998. \bibitem{kn:M} E. Mourre, {\em Absence of singular continuous spectrum for certain selfadjoint operators}, Comm. Math. Phys., {\bf 78} (1981), 391-408. \bibitem{kn:PSS} P. Perry, I. M. Sigal, B. Simon, {\em Spectral analysis of N-body Schr\"{o}dinger operators}, Ann. Math. {\bf 114} (1981), 519-567. \bibitem{kn:R1} D. Robert, {\em Autour de l'approximation semi-classique} PM, {\bf 68}, Birkh\"auser, Basel, 1987. \bibitem{kn:R2} D. Robert, {\em Asymptotique de la phase de diffusion \`a haute \'energie pour des perturbations du second ordre du Laplacien}, Ann. Sci. Ecole Norm. Sup., {\bf 25} (1992), 107-134. \bibitem{kn:R3} D. Robert, {\em Relative time-delay for perturbations of elliptic operators and semiclassical asymptotics}, J. Funct. Anal. {\bf 126} (1994), 36-82. \bibitem{kn:R4} D. Robert, {\em Semiclassical approximation in quantum mechanics. A survey of old and recent mathematical results}, Helv. Phys. Acta {\bf 71} (1998), 44-116. \bibitem{kn:RT1} D. Robert and H. Tamura, {\em Semiclassical estimates for resolvents and asymptotics for total scattering cross-sections}, Ann. Inst. H. Poincar\'e (Physique th\'eorique), {\bf 46} (1987), 415-442. \bibitem{kn:RT2} D. Robert and H. Tamura, {\em Asymptotic behavior of the scattering amplitudes in semi-classical and low energy limits}, Ann. Inst. Fourier, {\bf 39} (1989), 155-192. \bibitem{kn:Sj1} J. Sj\"ostrand, {\em A trace formula and review of some estimates for resonances}, in {\em Microlocal analysis and spectral theory (Lucca, 1996),} 377--437, NATO Adv. Sci. Inst. Ser. C Math. Phys. Sci., 490, Kluwer Acad. Publ., Dordrecht, 1997. \bibitem{kn:Sj2} J. Sj\"ostrand, {\em Resonances for bottles and trace formulae}, Preprint, Ecole Polytechnique, 1998. \bibitem{kn:SZ} J. Sj\"{o}strand and M. Zworski, {\em Complex scaling and the distribution of scattering poles}, J. Amer. Math. Soc. {\bf 4} (1991), 729-769. \bibitem{kn:TZ} S. Tang and M. Zworski, {\em From quasimodes to resonances}, Math. Res. Lett., {\bf 5} (1998), 261-272. \bibitem{kn:VZ} A. Vasy and M. Zworski, {\em Semiclassical estimates in asymptotically Euclidean scattering}, Preprint, 1999. \bibitem{kn:W} X.P. Wang, {\em Time-Decay of Scattering Solutions and Resolvents Estimates for Semi-classical Schr\"odinger Operators}, J. Differential Equations, {\bf 71} (1988), 348-395. \end{thebibliography}} \bigskip \noindent {\sc D\'epartement de Math\'ematiques Appliqu\'ees, Universit\'e de Bordeaux I, \\ 351, Cours de la Lib\'eration, 33405 Talence, FRANCE}\\ \email{vbruneau@@math.u-bordeaux.fr}\\ \email{petkov@@math.u-bordeaux.fr} \medskip \end{document} ---------------0001140413516--