Content-Type: multipart/mixed; boundary="-------------0702191110579" This is a multi-part message in MIME format. ---------------0702191110579 Content-Type: text/plain; name="07-37.keywords" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="07-37.keywords" Nelson model, adiabatic perturbation theory ---------------0702191110579 Content-Type: application/x-tex; name="preprint.tex" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="preprint.tex" \documentclass[a4paper, 12pt]{article} \usepackage[english]{babel} \usepackage{amsmath, amsfonts, amsthm, mathrsfs, latexsym} %\usepackage{srcltx} \usepackage[dvips]{graphicx} \newcommand{\abs}[1]{\lvert#1\rvert} \newcommand{\norm}[1]{\lVert#1\rVert} \newcommand{\field}[2]{\mathbb{#1}^{#2}} \newcommand{\deriv}[2]{\partial_{#1}^{#2}} \newcommand{\hilbert}[1]{\mathscr{#1}} \newcommand{\fock}{\hilbert{F}} \newcommand{\coinf}{C_{0}^{\infty}} \renewcommand{\vec}[1]{\boldsymbol{#1}} \newcommand{\N}{\mathrm{N}} \newcommand{\E}{{\mathrm{e}}} \newcommand{\I}{\mathrm{i}} \newcommand{\Or}{{\mathcal{O}}} \newcommand{\qn}{Q_{\mathrm{N}}} \newcommand{\qm}{Q_{M}} \newcommand{\qt}{\mathrm{Q}_{\mathrm{T}}} \newcommand{\pepsim}{P_{M}^{\epsi}} \newcommand{\qless}[1]{Q_{\leq\mathrm{#1}}} \newcommand{\dgamma}{\mathrm{d\Gamma}} \newcommand{\hfree}{H_{0}^{\epsi}} \newcommand{\heff}{H_{\mathrm{eff}}^{\epsi}} \newcommand{\heffsigma}{H_{\mathrm{eff}}^{\epsi, \sigma}} \newcommand{\hefftwochi}{H_{\mathrm{eff},\chi}^{(2)}} \newcommand{\hdiag}{H_{\mathrm{D}}^{(2)}} \newcommand{\hdiagp}{H_{\mathrm{D}, \mathrm{p}}^{(2)}} \newcommand{\hfield}{H_{\mathrm{f}}} \newcommand{\hp}{h_{\mathrm{p}}} \newcommand{\Hp}{H_{\mathrm{p}}} \newcommand{\Hepsip}{\Hp^{\epsi}} \newcommand{\id}{\mathbf{1}} \newcommand{\rfield}{R_{\mathrm{f}}^{\perp}} \newcommand{\hfib}{H_{\mathrm{fib}}} \newcommand{\hfibsigma}{H_{\mathrm{fib}, \sigma}} \newcommand{\rfib}{R_{\mathrm{fib}}^{\perp}} \newcommand{\pop}{\hat{p}} \newcommand{\popb}{\hat{\vec{p}}} \newcommand{\epsi}{\varepsilon} \newcommand{\hepsi}{H^{\epsi}} \newcommand{\hepsisigma}{H^{\epsi, \sigma}} \newcommand{\hepsisigmaepsi}{H^{\epsi, \sigma(\epsi)}} \newcommand{\vir}{V_{\mathrm{IR}}} \newcommand{\proj}[2]{\hat{\pi}^{#1}_{#2}} \newcommand{\pizn}{\hat{\pi}_{0}^{M}} \newcommand{\uone}{\hilbert{U}} \newcommand{\uepsi}{\hilbert{U}_{\epsi}} \newcommand{\uepsisigma}{\hilbert{U}_{\epsi, \sigma}} \newcommand{\uonemchi}{U^{(1)}_{\mathrm{J}, \chi}} \newcommand{\uonemchistar}{U^{(1)\,*}_{\mathrm{J}, \chi}} \newcommand{\amdelta}{A_{\delta}^{\mathrm{M}}} \newcommand{\amdeltaj}{A_{\delta, j}^{\mathrm{M}}} \newcommand{\adeltaj}{A_{\delta, j}} \newcommand{\amzeroj}{A_{0, j}^{\mathrm{M}}} \newcommand{\phim}{\Phi^{\mathrm{J}}} \newcommand{\phimdelta}{\phim_{\delta}} \newcommand{\phisigma}{\Phi_{\sigma}} \newcommand{\phisigmaj}{\Phi_{\sigma}^{\mathrm{J}}} \newcommand{\phisigmal}{\Phi_{\sigma, l}} \newcommand{\phisigmajl}{\Phi_{\sigma, l}^{\mathrm{J}}} \newcommand{\phij}{\Phi^{\mathrm{J}}} \newcommand{\phijl}{\Phi^{\mathrm{J}}_{l}} \newcommand{\phizerojl}{\Phi^{\mathrm{J}}_{0, l}} \newcommand{\eqex}{\overset{!}{=}} \newcommand{\opw}{\mathrm{Op}^{W}_{\epsi}} \newcommand{\omegap}{\omega_{\mathrm{p}}} \newcommand{\lh}{\mathcal{L}(\hilbert{H})} \newcommand{\lhh}[2]{\mathcal{L}(\hilbert{#1}, \hilbert{#2})} \newcommand{\tr}{\mathrm{tr}} \newcommand{\vsigma}{V_{\sigma}} \newcommand{\fockvac}{\Omega_{\mathrm{F}}} \newcommand{\nop}{\widehat{N}} \newcommand{\fockfin}{\fock_{\mathrm{fin}}} \newtheorem{theorem}{Theorem} \newtheorem*{theorem*}{Theorem} \newtheorem{lemma}{Lemma} \newtheorem*{prop}{Proposition} \newtheorem{proposition}{Proposition} \newtheorem{corollary}{Corollary} %\spnewtheorem*{theorem*}{Theorem}{\bf}{\it} %\spnewtheorem*{prop}{Proposition}{\bf}{\it} \theoremstyle{remark} \newtheorem{remark}{Remark} %opening \title{Effective dynamics for particles coupled to a quantized scalar field} \author{L. Tenuta \and S. Teufel} \date{} %\institute{Mathematisches Institut, Eberhard-Karls-Universit\"at, Auf der Morgenstelle 10, 72076,\\ %T\"ubingen, Germany.\\ %\email{lucattilio.tenuta@uni-tuebingen.de, stefan.teufel@uni-tuebingen.de}} \begin{document} \maketitle \begin{center} Mathematisches Institut, Eberhard-Karls-Universit\"at, Auf der Morgenstelle 10, 72076, T\"ubingen, Germany. e-mail: lucattilio.tenuta@uni-tuebingen.de, stefan.teufel@uni-tuebingen.de \end{center} \begin{abstract} We consider a system of $N$ non-relativistic spinless quantum particles (``electrons'') interacting with a quantized scalar Bose field (whose excitations we call ``photons''). We examine the case when the velocity $v$ of the electrons is small with respect to the one of the photons, denoted by $c$ ($v/c=\varepsilon\ll 1$). We show that dressed particle states exist (particles surrounded by ``virtual photons''), which, up to terms of order $(v/c)^3$, follow Hamiltonian dynamics. The effective $N$-particle Hamiltonian contains the kinetic energies of the particles and Coulomb-like pair potentials at order $(v/c)^0$ and the velocity dependent Darwin interaction and a mass renormalization at order $(v/c)^{2}$. Beyond that order the effective dynamics are expected to be dissipative. We derive a formula for the leading order piece of the radiated photon field. The main mathematical tool we use is adiabatic perturbation theory. However, in the present case there is no eigenvalue which is separated by a gap from the rest of the spectrum, but its role is taken by the bottom of the absolutely continuous spectrum, which is not an eigenvalue. Nevertheless we construct approximate $M$-photons dressed subspaces, which are adiabatically invariant for the dynamics up to order $(v/c)\sqrt{\ln (v/c)^{-1}}$. We also give an explicit expression for the non adiabatic transitions corresponding to emission of free photons. \end{abstract} \section{Introduction} In a system of classical charges the interactions are mediated through the electromagnetic field. In the case when the velocities of the particles are small with respect to the speed of light, it is possible, loosely speaking, to expand their dynamics in powers of $v/c$. The qualitative picture which emerges has three main features. Up to terms of order $(v/c)^{3}$, the dynamics of the particles are still of Hamiltonian form. Moreover, \begin{itemize} \item at leading order, $(v/c)^{0}$, the retardation effects can be neglected and the particles interact through an instantaneous pair potential; \item at order $(v/c)^{2}$, the particles acquire an effective mass, due to the contribution of the electromagnetic mass and, to take into account the retardation effects, one has to add to the potential a velocity-dependent term (the so-called Darwin term). \end{itemize} Including the terms of order $(v/c)^{3}$, the dynamics are not Hamiltonian anymore, instead \begin{itemize} \item there is dissipation of energy through radiation. In the dipole approximation, the rate of emitted energy is proportional to the acceleration of a particle squared. \end{itemize} A formal derivation of this picture, which does not consider the problem of mass renormalization, can be found in \cite{LaLi}. A mathematical analysis in the framework of the Abraham model, i.\ e., for charges which have a rigid charge distribution, is given in \cite{KuSp1} \cite{KuSp2}. The above description is expected to remain true also for nonrelativistic quantum electrodynamics, where, neglecting the possibility of pair creation, one considers a system of $N$ nonrelativistic particles interacting with the quantized Maxwell field. In physical terms, the particles carry now a cloud of virtual photons, which makes them heavier, and interact exchanging them or dissipate energy through photons travelling freely to infinity. Instead of nonrelativistic QED we consider the massless Nelson model, which describes $N$ spinless particles (which will be called ``electrons'') coupled to a scalar Bose field of zero mass (whose excitations will be called ``photons''). In spite of the simplifications introduced, this model is expected to retain the main physical features of the original one. Therefore, since its introduction by Nelson \cite{Ne}, who analyzed its ultraviolet behavior, it has been extensively studied to get information about the spectral and scattering features of QED, mostly concerning its infrared behavior (not pretending to be exhaustive, some papers related to this aspect are \cite{Ar} \cite{Fr} \cite{LMS} \cite{Pi} and references therein). The recent monograph by Spohn \cite{Sp} contains detailed descriptions of the classical and the quantum mechanical models and results. The aim of this paper is the mathematical derivation of the Schr\"odinger equation for $N$ particles which incorporates mass renormalization and an effective pair interaction (including the velocity-dependent Darwin term) starting from the massless Nelson model with an ultraviolet cutoff and \emph{no infrared cutoff}. We give moreover an explicit expression for the wavefunction of the radiated photons. The zero order, i.\ e., a Schr\"odinger equation with an effective Coulomb interaction, was already analyzed in \cite{Te2}. To explain in more detail the kind of scaling we are interested in it is convenient to look first at the classical case. The classical equations of motion for $N$ particles with positions $\vec{q}_{j}$, mass $m_{j}$ and a rigid ``charge'' distribution $\varrho_{j}$ coupled to a scalar field $\phi(\vec{x}, t)$ \footnote[1]{We use bold italic font, $\vec{x}$, to denote vectors in $\field{R}{3}$. The only exception to this, since there is no possibility of misunderstanding, is the three-dimensional momentum of the photons, denoted by $k$. The lightface font, $x$, will be used to denote vectors in $\field{R}{3N}$.} with propagation speed $c$ are \begin{eqnarray} \frac{1}{c^{2}}\deriv{t}{2}\phi(\vec{x}, t) & = & \Delta_{\vec{x}}\phi(\vec{x}, t) - \sum_{j=1}^{N}\varrho_{j}(\vec{x}-\vec{q}_{j}(t)), \label{classicaleqfield}\\ m_{j}\ddot{\vec{q}}_{j}(t) & = & -\int dx\, (\nabla_{\vec{x}}\phi)(\vec{x}, t)\varrho_{j}(\vec{x}-\vec{q}_{j}(t)),\label{classicaleqpart} \qquad 1\leq j \leq N \quad . \end{eqnarray} We assume for simplicity that \begin{equation*} \varrho_{j}(\vec{x}) = e_{j}\varphi(\vec{x}), \end{equation*} where the form factor $\varphi$ gives rise to a sharp ultraviolet cutoff, \begin{equation}\label{formfactor} \hat{\varphi}(k)=\begin{cases} (2\pi)^{-3/2} & \abs{k}<\Lambda\quad ,\\ 0 & \text{otherwise}\quad . \end{cases} \end{equation} Taking formally the limit $c\to\infty$ in \eqref{classicaleqfield}, we get the Poisson equation for the field, and so, eliminating the field from \eqref{classicaleqpart}, we obtain equations of motion describing $N$ particles interacting through smeared Coulomb potentials. Mass renormalization does not appear at the leading order. Instead of taking $c\to\infty$, we regard as more natural to explore the regime of particle properties which gives rise to effective equations. Indeed, if we look at \emph{heavy particles} for \emph{long times}, i.\ e., if we substitute $t'=\epsi t$ and $m_{j}'=\epsi^{2}m_{j}$ in \eqref{classicaleqfield} and \eqref{classicaleqpart}, we find that, up to rescaling, the limit $\epsi\to 0$ is equivalent to the limit $c\to\infty$. After quantization however, the two limiting procedures are not equivalent anymore. The case $c\to\infty$ was analyzed by Davies \cite{Da} and by Hiroshima \cite{Hi}, who at the same time removed the ultraviolet cutoff, applying methods from the theory of the weak coupling limit. In this paper we analyze the limit $\epsi\to 0$. We define now briefly the massless Nelson model (whose presentation will be completed in section \ref{nelsonmodel}), we state our main results and the principal ideas of the proof and then compare them to the above mentioned weak coupling limit. The model is obtained through canonical quantization of the classical system described by \eqref{classicaleqfield} and \eqref{classicaleqpart}. The state space for $N$ spinless particles is $L^{2}(\field{R}{3N})$ and the Hamiltonian, assuming for simplicity that all the particles have equal mass, is given by (we switch to natural units, fixing $\hbar=1$ and $c=1$) \begin{equation*} \Hp:= -\frac{1}{2m}\sum_{j=1}^{N}\Delta_{\vec{x}_{j}} . \end{equation*} As explained above, we consider the case of heavy particles, i.\ e., \begin{equation} m = \epsi^{-2}, \qquad 0<\epsi\ll 1, \end{equation} therefore the Hamiltonian becomes \begin{equation}\begin{split} & \Hepsip = -\frac{\epsi^{2}}{2}\Delta_{x} =: \frac{1}{2}\pop^{2},\\ & x\in\field{R}{3N}, \qquad \pop := -\I\epsi\nabla_{x}\quad. \end{split} \end{equation} The particles are coupled to a scalar field, whose states are elements of the bosonic Fock space over $L^{2}(\field{R}{3})$, defined by \begin{equation}\label{fockspace} \fock := \oplus_{M=0}^{\infty}\otimes_{(s)}^{M}L^{2}(\field{R}{3}), \end{equation} where $\otimes_{(s)}^{M}$ denotes the $M$-times symmetric tensor product and $\otimes_{(s)}^{0}L^{2}(\field{R}{3}):=\field{C}{}$. We denote by $\qm$ the projector on the $M$-particles subspace of $\fock$ and by $\fockvac$ the vector $(1, 0, \ldots)\in\fock$, called the Fock vacuum. The Hamiltonian for the free bosonic field is \begin{equation}\label{hfield} \hfield := \dgamma(\abs{k}), \end{equation} where $k$ is the momentum of the photons (the reader who is not familiar with the notation can find more details in section \ref{nelsonmodel}). The particle $j$ is linearly coupled to the field through the interaction Hamiltonian \begin{equation}\label{hintj} H_{\mathrm{I}, j}:= \int_{\field{R}{3}}d\vec{y}\, \phi(\vec{y})\varrho_{j}(\vec{y}-\vec{x}_{j}), \end{equation} where $\phi$ is the field operator in position representation. The state space of the combined system particles $+$ field is \begin{equation} \hilbert{H}:= L^{2}(\field{R}{3N})\otimes\fock \simeq L^{2}(\field{R}{3N}, \fock) \end{equation} and its time evolution is generated by the Hamiltonian \begin{equation}\label{hepsi}\begin{split} & \hepsi := \hfree + \sum_{j=1}^{N}H_{\mathrm{I}, j},\quad \hfree := \Hepsip\otimes\id + \id\otimes\dgamma(\abs{k})\quad, \end{split} \end{equation} with domain \begin{equation}\label{hilbertzero} \hilbert{H}_{0}:= H^2(\field{R}{3N}, \fock) \cap L^2(\field{R}{3N}, D(\hfield)) \quad , \end{equation} which is a Hilbert space with the graph norm associated to $\hfree$. Note that there are no direct forces acting between the particles, all the interactions are mediated through the field. One goal is the construction of approximate solutions to the time-dependent Schr\"odinger equation \begin{equation} \I\epsi\deriv{t}{}\Psi(t) = \hepsi\Psi(t),\quad \Psi(0)=\Psi_{0}\in\hilbert{H} \end{equation} from solutions to an effective Schr\"odinger equation \begin{equation} \I\epsi\deriv{t}{}\psi(t) = \heff\psi(t), \quad \psi(0)=\psi_{0}\in L^{2}(\field{R}{3N}) \end{equation} for the particles only. Notice the factor $\epsi$ in front of the time derivative in both equations, which means, that, as we already mentioned, we switched to a time scale of order $\epsi^{-1}$ in the original units. This is necessary in order to see nontrivial dynamics of the particles, because, since their mass is of order $\Or(\epsi^{-2})$ and we consider states of finite kinetic energy, their velocity is of order $\Or(\epsi)$. To achieve this goal, an essential step is to remark that $\hepsi$ is given by the perturbation of a fibered Hamiltonian, because, since $H_{\mathrm{I}, j}$ depends only on the configuration $\vec{x}_{j}$ of the $j$th particle, the operator \begin{equation} \hfib(x) := \dgamma(\abs{k}) + \sum_{j=1}^{N}H_{\mathrm{I}, j}(\vec{x}_{j}) \end{equation} acts on $\fock$ for every fixed $x\in\field{R}{3N}$. This means that \begin{equation*} \hepsi = -\frac{\epsi^{2}}{2}\Delta_{x}\otimes\id + \int_{\field{R}{3N}}^{\oplus}\hfib(x)\quad . \end{equation*} There is therefore a structural similarity with the Born-Oppenheimer approximation, where the Hamiltonian describes the interaction between nuclei and electrons in a molecule, where the former have a mass of order $\Or(\epsi^{-2})$ with respect to the latter (the typical spectrum of $\hfib(x)$ for a diatomic molecule is shown in figure \ref{diatomic}). In the present case, the particles take the role of the nuclei, and the bosons the one of the electrons. \begin{figure}\begin{center} \includegraphics[width=0.7\textwidth]{bornopp.eps}\caption{\footnotesize Schematic spectrum of the fibered Hamiltonian in the case of a diatomic molecule for energies below the dissociation threshold ($r=\abs{\vec{x}_{1}-\vec{x}_{2}}$). The different eigenvalues are pointwise separated by a gap.}\label{diatomic} \end{center} \end{figure} In contrast to the molecular case however, in the Nelson model $\hfib$ has a purely absolutely continuous spectrum, and does not display a structure with pointwise separated bands (\emph{absence of both eigenvalues and spectral gap}) (see figure \ref{nelson}). \begin{figure}\begin{center} \includegraphics[width=0.7\textwidth]{spectrum.eps}\caption{\footnotesize Spectrum of $\hfib(x)$ for $N=2$ ($r=\abs{\vec{x}_{1}-\vec{x}_{2}}$). The spectrum is absolutely continuous, there is \emph{no eigenvalue at the bottom}. The oscillations in $r$ caused by the sharp ultraviolet cutoff are irrelevant, and therefore we do not show them.}\label{nelson} \end{center}\end{figure} The bottom of the spectrum, $E(x)$, can be explicitly calculated, \begin{equation}\label{ex} E(x) = \frac{1}{2}\sum_{i, j=1}^{N}V_{ij}(\vec{x}_{i}-\vec{x}_{j}) + e_{0}, \end{equation} where \begin{equation} V_{ij}(\vec{z}):= -\int_{\field{R}{3}\times\field{R}{3}}d\vec{v}d\vec{w}\, \frac{\varrho_{i}(\vec{v}-\vec{z})\varrho_{j}(\vec{w})}{4\pi\abs{\vec{v}-\vec{w}}}, \end{equation} and \begin{equation} e_{0}:= -\frac{1}{2}\sum_{j=1}^{N}\int_{\field{R}{3}\times\field{R}{3}}d\vec{v}d\vec{w}\, \frac{\varrho_{j}(\vec{v})\varrho_{j}(\vec{w})}{4\pi\abs{\vec{v}-\vec{w}}}\quad. \end{equation} The effective pair-potential $V_{ij}(\vec{z})$ coincides, up to the sign, with the electrostatic interaction energy of the charge distributions $\varrho_{i}$ and $\varrho_{j}$ at distance $\vec{z}$, while $e_{0}$ is the sum of all the self-energies. $E(x)$ becomes an eigenvalue of $\hfib(x)$ if the total charge of the $N$ particles system is equal to zero, as it happens for example in presence of an infrared cutoff. In this case, it is possible to build for every $x$ a unitary operator, $V(x)$, called the \emph{dressing operator}, which diagonalizes $\hfib(x)$ in the sense that \begin{equation} \hfib(x) = V(x)\hfield V(x)^{*} + E(x)\quad. \end{equation} Exploiting this remark, we approximate the time evolution generated by $\hepsi$ in two steps. First we define an infrared regular Hamiltonian, $\hepsisigma$, where the form factor $\hat{\varphi}$ (equation \eqref{formfactor}) is replaced by \begin{equation}\label{formfactorsigma} \hat{\varphi}_{\sigma}(k):=\begin{cases} (2\pi)^{-3/2} & \sigma<\abs{k}<\Lambda\quad ,\\ 0 & \text{otherwise}\quad . \end{cases} \end{equation} Then it holds \begin{prop}(see proposition \ref{approxsigma})\begin{equation} \big\lVert\E^{-\I t\hepsi/\epsi} - \E^{-\I t \hepsisigma/\epsi}\big\rVert_{\mathcal{L}(\hilbert{H}_{0}^{1/2}, \hilbert{H})}\leq C\abs{t}\frac{\sigma^{1/2}}{\epsi}, \end{equation} where $\hilbert{H}_{0}^{1/2}:= D\big((\hfree)^{1/2}\big)$ with the corresponding graph norm. \end{prop} Choosing $\sigma$ as a power of $\epsi$ we can then replace the original dynamics with infrared regular ones. The latter contain however two parameters, $\epsi$ and $\sigma$, therefore it is necessary to control the behavior of all the quantities that appear with respect to both. The advantage is that for $\hepsisigma$ we can build an approximate dressing operator $\uepsisigma$, acting on the whole Hilbert space $\hilbert{H}$, which is defined for every positive $\sigma$, but whose limit when $\sigma\to 0^{+}$ does not exist if the system has a total charge different from zero. $\uepsisigma$ is unitary and can be expanded in a series of powers of $\epsi$, with coefficients which are \emph{at most logarithmically divergent} in $\sigma$. Moreover, the zero order coefficient is given by $\vsigma$, the dressing operator associated to the infrared regular fibered Hamiltonian, which has therefore the property that \begin{equation*} \hepsisigma = \frac{1}{2}\pop^{2}\otimes\id + \vsigma(x)\hfield \vsigma(x)^{*} + E_{\sigma}(x). \end{equation*} Using $\uepsisigma$, we define an effective Hamiltonian \begin{equation} \heffsigma := \uepsisigma\hepsisigma\uepsisigma^{*} \end{equation} which can be expanded in a series of powers of $\epsi$ in $\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})$, with coefficients which are also at most logarithmically divergent in $\sigma$. The gain in switching to the representation defined by $\uepsisigma$ is that we can expand the dynamics of the system separating the different physical contributions according to their order in $\epsi$. The first result we find is that, even though the fibered Hamiltonian has no eigenvalues and no spectral gap, there are \emph{approximate $M$-photons dressed subspaces} which are almost invariant for the dynamics. \begin{theorem*}[Adiabatic invariance of $M$-photons dressed particles subspaces] (see theorem \ref{zeroordertime} and remark \ref{adiabaticinvm})%\label{theoremcommutator} Given any $\chi\in\coinf(\field{R}{})$ and any function $\sigma(\epsi)$ such that \begin{equation}\label{conditionsigma} \sigma(\epsi)^{1/2}\epsi^{-3}\to 0, \qquad \epsi\sqrt{\log(\sigma(\epsi)^{-1})}\to 0, \qquad \epsi\to 0^{+}\quad , \end{equation} then \begin{equation}\label{almostinvsub} \big\lVert[e^{-\I t\hepsi/\epsi}, \pepsim]\chi(\hepsi)\big\rVert_{\lh}\leq C\sqrt{M+1}\abs{t}\epsi\sqrt{\log(\sigma(\epsi)^{-1})}, \end{equation} where \begin{equation} \pepsim := \hilbert{U}^{*}_{\epsi, \sigma(\epsi)}\qm\hilbert{U}_{\epsi, \sigma(\epsi)} \quad \footnote[1]{We use the symbol $\qm$ to denote also the natural extension of $\qm$ to $\hilbert{H}$. }. \end{equation} \end{theorem*} The physical mechanism which leads to the almost invariance of the subspaces is adiabatic decoupling, i.\ e., the separation of scales for the motion of the different parts of the system, which lets the fast degrees of freedom, in our case the photons, instantaneously adjust to the motion of the slow degrees of freedom, the electrons. It is a well-known fact however, that the decoupling becomes poorer and poorer when the kinetic energies and thus the velocities of the heavy particles grow. The quadratic dispersion relation for the electrons allows them to become arbitrarily fast, therefore the decoupling holds uniformly just on states of bounded kinetic energy. This is the reason why we introduce a uniform bound on the total energy of the system through the function $\chi$, which obviously implies a bound on the kinetic energy of the electrons. Pushing the expansion in $\epsi$ one order further and using time-dependent perturbation theory, we can moreover study the effective dynamics for initial states which are in an almost invariant subspace and calculate the transitions between them, which correspond physically to the radiated piece of the wavefunction. \begin{theorem*}[Radiation](see corollary \ref{radiatedpiece} and the subsequent remark) For a system starting in the approximate dressed vacuum $(M=0)$ we have, at the leading order, \begin{equation}\begin{split} & (\id - P_{0}^{\epsi})e^{-\I t\hepsi/\epsi}P_{0}^{\epsi}\chi(\hepsi)\Psi\simeq\\ &\simeq-\frac{\epsi}{\sqrt{2}}P_{1}^{\epsi}V_{\sigma(\epsi)}\E^{-\I t(\pop^{2}/2+E(x))/\epsi}\cdot\\ &\cdot\sum_{j=1}^{N}\frac{\hat{\varrho}^{\sigma(\epsi)}_{j}(k)}{\abs{k}^{3/2}}\kappa\cdot\int_{0}^{t}ds\, \E^{\I (s-t)\abs{k}/\epsi}\opw(\ddot{\vec{x}}_{j}^{c}(s; x, p))\langle V_{\sigma(\epsi)}\Omega_{\mathrm{F}}, \chi(\hepsi)\Psi\rangle_{\fock} \quad , \end{split} \end{equation} where $\vec{x}_{j}^{c}$ is the solution to the classical equations of motion \begin{equation}\begin{split} & \ddot{\vec{x}}_{j}^{c}(s; x, p) = -\nabla_{\vec{x}_{j}}E(x^{c}(s; x, p)),\\ & \vec{x}_{j}^{c}(0; x, p)=\vec{x}_{j},\qquad \dot{\vec{x}}_{j}^{c}(0; x, p)=\vec{p}_{j},\quad j=1, \ldots, N\quad, \end{split} \end{equation} $\kappa:= k/\abs{k}$ and $\opw$ denotes the Weyl quantization. \end{theorem*} Generically (for the precise meaning of this we refer to corollary \ref{radiatedpiece}) the norm of the radiated piece is bounded below by $\Or\big(\epsi\log(\epsi\sigma(\epsi)^{-1})\big)$, which means that the subspace $P_{0}^{\epsi}$ is near \emph{optimal}, in the sense that the transitions are at least of order $\Or\big(\epsi\log(\epsi\sigma(\epsi)^{-1})\big)$. \begin{theorem*}[Effective dynamics of the particles](see theorem \ref{densitymatrix}) Given a macroscopic observable for the particles, $\opw(a)$, where $a$ is a smooth function bounded with all its derivatives, and a density matrix $\omega\in \hilbert{I}_{1}(P_{M}^{\epsi}\chi(\hepsi)\hilbert{H})$ whose time evolution is defined by \begin{equation} \omega(t):= e^{-\I t\hepsi/\epsi}\omega e^{\I t\hepsi/\epsi}, \end{equation} then \begin{equation}\begin{split} & \tr_{\hilbert{H}}\bigg(\big(\opw(a)\otimes\id_{\fock}\big)\omega(t)\bigg)=\tr_{L^{2}(\field{R}{3N})}\bigg(\opw(a)e^{-\I t H^{(2)}_{\mathrm{D, p}}}\tr_{\fock}\tilde{\omega}(0)e^{\I t H^{(2)}_{\mathrm{D, p}}}\bigg) +\\ &\qquad\qquad\qquad\qquad\qquad\qquad+\Or\big(\epsi^{2}\abs{t}\log(\sigma(\epsi)^{-1})\big) + \Or\big(\epsi^{2}\abs{t}^{2}\log(\sigma(\epsi)^{-1})\big), \end{split} \end{equation} where $\hdiagp$ depends on $\epsi$ but \emph{not} on $\sigma$, \begin{equation}\begin{split} &\hdiagp:= \sum_{j=1}^{N}\frac{1}{2m_{j}^{\epsi}}\vec{\pop}_{j}^{2}+E(x) +\\ &-\frac{\epsi^{2}}{4}\underset{(l_{1}\neq j_{1})}{\sum_{l_{1}, j_{1}=1}^{N}}\int_{\field{R}{3}}dk\, \frac{\hat{\varrho}_{l_{1}}(k)^{*}\hat{\varrho}_{j_{1}}(k)}{\abs{k}^{2}}\big[\E^{\I k\cdot(\vec{x}_{j_{1}}-\vec{x}_{l_{1}})}(\kappa\cdot\vec{\pop}_{l_{1}})(\kappa\cdot\vec{\pop}_{j_{1}})+\\ &+(\kappa\cdot\vec{\pop}_{l_{1}})(\kappa\cdot\vec{\pop}_{j_{1}})\E^{\I k\cdot(\vec{x}_{j_{1}}-\vec{x}_{l_{1}})}\big], \end{split} \end{equation} with $m_{j}^{\epsi}:=1/(1+\frac{\epsi^{2}}{2}\tilde{e}_{j})$ and \begin{equation} \tilde{e}_{j}:= \frac{1}{4\pi}\int_{\field{R}{3}\times\field{R}{3}}d\vec{x}\ d\vec{y}\ \frac{\varrho_{j}(\vec{x})\varrho_{j}(\vec{y})}{\abs{\vec{x}-\vec{y}}} \quad , \end{equation} while \begin{equation} \tilde{\omega}(0):= V_{\sigma(\epsi)}^{*}\omega V_{\sigma(\epsi)}\quad . \end{equation} \end{theorem*} The Hamiltonian $\hdiagp$ is equal to the Weyl quantization of the Darwin Hamiltonian, which, as we mentioned above, appears in classical electrodynamics when one expands the dynamics for the particles including term of order $(v/c)^{2}$, and it contains a characteristic velocity dependent term, which accounts for retardation effects. The technique used for proving these theorems is based on space-adiabatic perturbation theory, a general scheme designed to expand the dynamics generated by a pseudodifferential operator with a semiclassical symbol, described in detail in \cite{Te3}. However, this method exploits the assumption that the fibered Hamiltonian has a spectral gap, of the kind showed in figure \ref{diatomic}, therefore is not directly applicable and needs some modifications. In particular, the infrared cutoff $\sigma$ plays in our context the role of an effective gap. In the more usual framework of time-dependent Hamiltonians, adiabatic theorems without gap condition were first proven by Bornemann \cite{Bo} and Avron and Elgart \cite{AvEl1} (their proof was simplified and the result somewhat strengthened in \cite{Te1}). Using similar techniques, a space-adiabatic theorem without gap for the Nelson model was proven in \cite{Te2}. In the context of quantum statistical mechanics, an adiabatic theorem without gap for the Liouvillian (the generator of the dynamics of the states in a suitable representation of the von Neumann algebra associated to the system) is discussed in \cite{AbFr1}, while the case of time-dependent resonances was studied in \cite{AbFr2}. All these results are of the first order type, i.\ e., they describe the leading order adiabatic evolution and give upper bounds for the transitions. In our case instead, we build adiabatic dynamics including terms of the second order, and give an explicit expression for the non-adiabatic transitions, which allows us to give also a lower bound for them. In the time-dependent case, lower bounds for the transitions were calculated by Avron and Elgart \cite{AvEl2} for the Friedrichs model, which describes a ``small system'', whose Hilbert space is one-dimensional, interacting with a ``reservoir'', whose Hilbert space is $L^{2}(\field{R}{}_{+}, d\mu(k))$, with a suitable spectral density $\mu$. We are now ready to compare our results with those obtained by Davies \cite{Da}. He considers the limit $c\to\infty$ for the Hamiltonian \begin{equation*} H^{c} := H_{\mathrm{p}}\otimes\id + \id\otimes\dgamma(c\abs{k}) + \sqrt{c}H_{I} \end{equation*} which is equal to $H^{\epsi=1}$ not setting $c=1$ as we did before. Davies proves then that for all $t\in\field{R}{}$, \begin{equation*} \lim_{c\to\infty}\E^{-\I H^{c}t}(\psi\otimes\Omega_{\mathrm{F}})= (\E^{-\I (-\Delta/2 + E(x))t}\psi)\otimes\Omega_{\mathrm{F}}\quad . \end{equation*} This shows that, while at the classical level the limit $\epsi\to 0$ and $c\to\infty$ are equivalent, they differ at the quantum one. Indeed, the limit $\epsi\to 0$ is singular, because no limiting dynamics for $\epsi=0$ exist. Moreover, the effective dynamics we get refer to approximate dressed states, which contain a non zero number of photons, while the $c\to\infty$ limit is taken on states which contain no photons. Finally, we summarize the structure of our paper. After some basic facts recalled in section \ref{nelsonmodel}, we explain how to construct the approximate dressing operator $\hilbert{U}$ in section \ref{constructionunitary}. In section \ref{effectivehamiltonian}, we analyze the expansion of the effective Hamiltonian and then apply these results to the study of the dynamics in section \ref{effectivedynamics}. \section{Description of the model and preliminary facts}\label{nelsonmodel} In this section we complete the presentation of the Nelson model, discussing also the spectrum of the fibered Hamiltonian, and collect some basic facts we will use in the following. \subsection{Fock space and field operator} (The proofs of the statements we claim can be found in (\cite{ReSi2}, section X.7)). We denote by $\fockfin$ the subspace of the Fock space, defined in \eqref{fockspace}, for which $\Psi^{(M)}=0$ for all but finitely many $M$. Given $f\in L^{2}(\field{R}{3})$, one defines on $\fockfin$ the annihilation operator by \begin{equation} (a(f)\Psi)^{(M)}(k_{1}, \ldots, k_{M}):= \sqrt{M+1}\int_{\field{R}{3}}dk\, \bar{f}(k)\Psi^{(M+1)}(k, k_{1}, \ldots, k_{M}) \quad . \end{equation} The adjoint of $a(f)$ is called the creation operator, and its domain contains $\fockfin$. On this subspace they satisfy the canonical commutation relations \begin{equation}\begin{split} & [a(f), a(g)^{*}]=\langle f, g\rangle_{L^{2}(\field{R}{3})},\\ & [a(f), a(g)]=0, \quad [a(f)^{*}, a(g)^{*}]=0 \quad . \end{split} \end{equation} Since the commutator between $a(f)$ and $a(f)^{*}$ is bounded, it follows that $a(f)$ can be extended to a closed operator on the same domain of $a(f)^{*}$. On this domain one defines the Segal field operator \begin{equation} \Phi(f) := \frac{1}{\sqrt{2}}(a(f) + a(f)^{*}) \end{equation} which is essentially self-adjoint on $\fockfin$. Moreover, $\fockfin$ is a set of analytic vectors for $\Phi(f)$. From the canonical commutation relations it follows that \begin{equation}\label{commphi} [\Phi(f), \Phi(g)] = \I\Im\langle f, g\rangle_{L^{2}(\field{R}{3})}\quad . \end{equation} Given a self-adjoint multiplication operator by the function $\omega$ on the domain $D(\omega)\subset L^{2}(\field{R}{3})$, we define \begin{equation} \fock_{\omega, \mathrm{fin}} := \mathcal{L}\{\fockvac, a(f_{1})^{*}\cdots a(f_{M})^{*}\fockvac: M\in\field{N}{}, f_{j}\in D(\omega), j=1, \dots, M\}, \end{equation} where $\mathcal{L}$ means ``finite linear combinations of''. On $\fock_{\omega, \mathrm{fin}}$ we define the second quantization of $\omega$, $\dgamma(\omega)$, by \begin{equation} (\dgamma(\omega)\Psi)^{(M)}(k_{1}, \ldots, k_{M}) := \sum_{j=1}^{M}\omega(k_{j})\Psi^{(M)}(k_{1}, \ldots, k_{M}), \end{equation} which is essentially self-adjoint. In particular, the free field Hamiltonian $\hfield$, equation \eqref{hfield}, acts as \begin{equation*} (\hfield\Psi)^{(M)}(k_{1}, \ldots, k_{M}) = \sum_{j=1}^{M}\abs{k_{j}}\Psi^{(M)}(k_{1}, \ldots, k_{M}) \end{equation*} and is self-adjoint on its maximal domain. >From the previous definitions, given $f\in D(\omega)$, one gets the commutation properties \begin{equation}\label{commutators}\begin{split} & [\dgamma(\omega), a(f)^{*}] = a(\omega f)^{*}, \quad [\dgamma(\omega), a(f)] = -a(\omega f),\\ & [\dgamma(\omega), \I\Phi(f)] = \Phi(\I\omega f) \quad . \end{split} \end{equation} \subsection{The Nelson model} Using the Segal field operator (which involves taking a Fourier transform) the interaction Hamiltonian, equation \eqref{hintj} and \eqref{hepsi}, can be written as \begin{equation} \sum_{j=1}^{N}H_{\mathrm{I}, j}(\vec{x}_{j}) = \Phi(\abs{k}v(x, k)), \end{equation} where \begin{equation} v(x, k) := \sum_{j=1}^{N}\E^{\I k\cdot \vec{x}_{j}}\frac{\hat{\varrho}_{j}(k)}{\abs{k}^{3/2}} = \sum_{j=1}^{N}\E^{\I k\cdot \vec{x}_{j}}e_{j}\frac{\hat{\varphi}(k)}{\abs{k}^{3/2}} \quad . \end{equation} This form is useful to prove some standard properties of $\hepsi$ and $\hfib(x)$. \begin{lemma} \begin{enumerate} \item $\hepsi$ is self-adjoint on $\hilbert{H}_{0}$ (see eq. \eqref{hilbertzero}). \item For every $x\in\field{R}{3N}$, $\hfib(x)$ is self-adjoint on $D(\hfield)$. \end{enumerate} \end{lemma} \begin{proof} The claims are based on the standard estimates (see, e.\ g., \cite{Be}, proposition 1.3.8) \begin{equation} \norm{\Phi(f)\psi}^{2}_{\fock}\leq 2\norm{f/\sqrt{\abs{k}}}_{L^{2}(\field{R}{3})}^{2}\langle\psi, \hfield\psi\rangle_{\fock}+2\norm{f}^{2}_{L^{2}(\field{R}{3})}\norm{\psi}_{\fock}^{2}, \end{equation} for $\psi\in D(\hfield)$, and \begin{equation}\label{inequalityphi} \norm{\Phi(f)\Psi}^{2}_{\hilbert{H}}\leq 2\norm{f/\sqrt{\abs{k}}}_{L^{2}(\field{R}{3})}^{2}\langle\Psi, (\mathbf{1}\otimes\hfield)\Psi\rangle_{\hilbert{H}}+2\norm{f}^{2}_{L^{2}(\field{R}{3})}\norm{\Psi}_{\hilbert{H}}^{2}, \end{equation} for $\Psi\in L^{2}(\field{R}{3N}, D(\hfield))$, which imply respectively that, for fixed $x$, $\Phi(\abs{k}v(x, k))$ is infinitesimally small with respect to $\hfield$ and, as an operator on $\hilbert{H}$, is infinitesimally small with respect to $\hfree$, uniformly in $\epsi$. \end{proof} The fibered Hamiltonian has the form of a quadratic part, which corresponds to the free field, plus a term linear in the annihilation and creation operators. Hamiltonians of this form are usually called in the literature ``van Hove Hamiltonians'' (a review on the subject is given in \cite{De}). Their simple form allows to analyze in detail their spectral and dynamical features. In the finite dimensional case, if one considers an harmonic oscillator with a linear perturbation, like an external electric field for example, the natural way to recover the spectrum of the Hamiltonian is to translate the $x$ variable, transforming the initial Hamiltonian into a purely quadratic one. In quantum field theory, the analogous strategy would be to find a unitary operator which translates the annihilation and creation operators. Such operator would implement what is called a Bogoliubov transformation. While for a finite number of degrees of freedom the Bogoliubov transformation is always implementable, this may not be the case if the phase space is infinite dimensional. In particular this is not possible if the van Hove Hamiltonian is not sufficiently regular in the infrared region. After this preliminary remarks, we can state \begin{lemma} Given the fibered Hamiltonian \begin{equation*} K_{\mathrm{fib}}(x) := \hfield + \Phi(z(x, k)) \end{equation*} where $z(x, k)$ satisfies \begin{equation*} \alpha(x):=-\frac{1}{2}\int_{\field{R}{3}}dk\, \frac{\abs{z(x, k)}^{2}}{\abs{k}} < \infty\qquad \forall x\in\field{R}{3N} \end{equation*} one has: \begin{enumerate} \item The spectrum of $K_{\mathrm{fib}}(x)$ is given by $[\alpha(x), +\infty)$, and \begin{equation} \sigma_{\mathrm{ac}}(K_{\mathrm{fib}}(x))= (\alpha(x), +\infty) \quad . \end{equation} \item The infimum of the spectrum, $\alpha(x)$, is an eigenvalue if and only if \begin{equation*} \int_{\field{R}{3}}dk\, \frac{\abs{z(x, k)}^{2}}{\abs{k}^{2}} <\infty \end{equation*} In this case, moreover, the unitary operator \begin{equation} V(x) := \E^{\I\Phi(\I z(x, k)\abs{k}^{-1})} \end{equation} is well-defined on the Fock space and \begin{equation} K_{\mathrm{fib}}(x) = V(x)\hfield V(x)^{*} + \alpha(x). \end{equation} If $z(x, k)\abs{k}^{-1}\notin L^{2}(\field{R}{3}, dk)$, then the spectrum is absolutely continuous. \end{enumerate} \end{lemma} \begin{proof} Point $1$ is proposition $3.10$ of \cite{De} and point $2$ is proposition $3.13$ of the same paper. \end{proof} \begin{corollary} $\sigma(\hfib(x)) = \sigma_{\mathrm{ac}}(\hfib(x)) = [E(x), +\infty)$, where \begin{equation} E(x) = -\frac{1}{2}\int_{\field{R}{3}}dk\, \abs{k}\abs{v(x, k)}^{2}\quad . \end{equation} A more explicit expression for $E(x)$ is given in \eqref{ex}. \end{corollary} \begin{corollary} The infrared regularized fibered Hamiltonian $H_{\mathrm{fib}, \sigma}(x)$ (see equation \eqref{formfactorsigma}) can be written as \begin{equation*} H_{\mathrm{fib}, \sigma}(x) = \vsigma(x)\hfield \vsigma(x)^{*} + E_{\sigma}(x), \end{equation*} where \begin{equation}\label{vsigma} \vsigma = \E^{\I\Phi(\I v_{\sigma}(x, k))} \quad . \end{equation} $E_{\sigma}(x)$ is the only eigenvalue, with eigenvector $\Omega_{\sigma}(x):= \vsigma(x)\fockvac$. Moreover, for every $\alpha\in\field{N}{3N}$, \begin{equation}\label{estimateex} \deriv{x}{\alpha}E_{\sigma} = \deriv{x}{\alpha}E + \Or(\sigma^{\abs{\alpha}+1})_{\mathcal{L}(\hilbert{H})}\quad . \end{equation} \end{corollary} \begin{proof} The only thing left to prove is equation \eqref{estimateex}, which follows immediately from the fact that \begin{equation*} \deriv{x}{\alpha}E(x) - \deriv{x}{\alpha}E_{\sigma}(x) = -\frac{1}{2}\int_{\abs{k}<\sigma}dk\, \abs{k}\,\deriv{x}{\alpha}\abs{v(x, k)}^{2} \end{equation*} and \begin{equation*} \abs{k}\,\deriv{x}{\alpha}\abs{v(x, k)}^{2}\leq C\abs{k}^{\abs{\alpha}-2} \quad . \end{equation*} \end{proof} As last preliminaries, we show that we can approximate the true dynamics with infrared regular ones and that the same holds for the cutoffs in the energy. \begin{proposition}\label{approxsigma}\begin{equation} \big\lVert\E^{-\I t\hepsi/\epsi} - \E^{-\I t \hepsisigma/\epsi}\big\rVert_{\mathcal{L}(\hilbert{H}_{0}^{1/2}, \hilbert{H})}\leq C\abs{t}\frac{\sigma^{1/2}}{\epsi}, \end{equation} where $\hilbert{H}_{0}^{1/2} = D\big((\hfree)^{1/2}\big)$ with the corresponding graph norm. \end{proposition} \begin{proof} Both Hamiltonians are self-adjoint on $\hilbert{H}_{0}$, therefore, given $\Psi\in\hilbert{H}_{0}$ we can apply Duhamel formula to get \begin{equation*}\begin{split} & (\E^{-\I t\hepsi/\epsi} - \E^{-\I t\hepsisigma/\epsi})\Psi = \frac{\I}{\epsi}\int_{0}^{t}ds\, \E^{\I (s-t)\hepsisigma/\epsi}(\hepsi-\hepsisigma)\E^{-\I s\hepsi/\epsi}\Psi = \\ &= \frac{\I}{\epsi}\int_{0}^{t}ds\, \E^{\I (s-t)\hepsisigma/\epsi}\Phi\big(\abs{k}v(x, k)\id_{(0, \sigma)}(k)\big)\E^{-\I s\hepsi/\epsi}\Psi\\ &\Rightarrow \norm{(\E^{-\I t\hepsi/\epsi} - \E^{-\I t\hepsisigma/\epsi})\Psi}_{\hilbert{H}}\leq \frac{1}{\epsi}\int_{0}^{t}ds\, \norm{\Phi\big(\abs{k}v(x, k)\id_{(0, \sigma)}(k)\big)\E^{-\I s\hepsi/\epsi}\Psi}_{\hilbert{H}} \end{split} \end{equation*} where $\id_{(0, \sigma)}(k)$ is the characteristic function of the interval indicated. Using equation \eqref{inequalityphi} we get \begin{equation*}\begin{split} & \norm{\Phi\big(\abs{k}v(x, k)\id_{(0, \sigma)}(k)\big)\E^{-\I s\hepsi/\epsi}\Psi}_{\hilbert{H}} \leq\\ & \leq \sqrt{2}\norm{\abs{k}^{1/2}v(x, k)\id_{(0, \sigma)}(k)}_{L^{2}(\field{R}{3})}\norm{(\hfree)^{1/2}\E^{-\I s\hepsi/\epsi}\Psi}_{\hilbert{H}} +\\ &+ \sqrt{2}\norm{\abs{k}v(x, k)\id_{(0, \sigma)}(k)}_{L^{2}(\field{R}{3})}\norm{\Psi}_{\hilbert{H}}. \end{split} \end{equation*} Moreover, for $\beta>0$, \begin{equation}\label{normsigma}\begin{split} &\norm{\abs{k}^{\beta}v(x, k)\id_{(0, \sigma)}(k)}_{L^{2}(\field{R}{3})}\leq \sum_{j=1}^{N}\abs{e_{j}}\cdot\norm{\abs{k}^{\beta-3/2}\hat{\varphi}(k)\id_{(0, \sigma)}(k)}_{L^{2}(\field{R}{3})},\\ & \norm{\abs{k}^{\beta-3/2}\hat{\varphi}(k)\id_{(0, \sigma)}(k)}_{L^{2}(\field{R}{3})}^{2}=\frac{4\pi}{(2\pi)^{3}}\frac{\sigma^{2\beta}}{2\beta}. \end{split} \end{equation} Since $\Phi(\abs{k}v(x, k))$ is infinitesimally small with respect both to $\id\otimes\hfield$ and $\hfree$ (as it follows again from \eqref{inequalityphi}), the graph norm defined by $(\hfree)^{1/2}$ and the one defined by $(\hepsi)^{1/2}$ are equivalent, uniformly in $\epsi$ (theorem X.18, \cite{ReSi2}). This implies that, for $\sigma$ sufficiently small, \begin{equation*} \norm{\Phi\big(\abs{k}v(x, k)\id_{(0, \sigma)}(k)\big)\E^{-\I s\hepsi/\epsi}\Psi}_{\hilbert{H}} \leq\tilde{C}\sigma^{1/2}\big(\norm{(\hfree)^{1/2}\Psi}_{\hilbert{H}}^{2} + \norm{\Psi}_{\hilbert{H}}^{2}\big)^{1/2}, \end{equation*} which proves the statement. \end{proof} \begin{lemma}\label{cutoffenergysigma} Given a function $\chi\in\coinf(\field{R}{})$ then \begin{equation} \norm{\chi(\hepsi) - \chi(\hepsisigma)}_{\mathcal{L}(\hilbert{H})}\leq C\sigma^{1/2}\quad . \end{equation} \end{lemma} \begin{proof} Using Hellfer-Sj\"ostrand formula (see, e. g., \cite{DiSj} chapter $8$), given a self-adjoint operator $A$, we can write \begin{equation}\label{hellfersjostrand} \chi(A) = \frac{1}{\pi}\int_{\field{R}{2}}dx dy\ \bar{\partial}\chi^{a}(z)(A-z)^{-1}, \quad z:=x+\I y, \end{equation} where $\chi^{a}\in \coinf(\field{C}{})$ is an almost analytic extension of $\chi$ with the property that $\exists\, D_{\bar{N}}$ such that $\abs{\bar{\partial}\chi^{a}}\leq D_{\bar{N}}\abs{\Im z}^{\bar{N}}$, where $\bar{N}$ is fixed, but can be chosen arbitrarily large. Applying it to our case we get \begin{equation*} \chi(\hepsi) - \chi(\hepsisigma) = \frac{1}{\pi}\int_{\field{R}{2}}dxdy\, \bar{\partial}\chi^{a}(z)\big[(\hepsi-z)^{-1} - (\hepsisigma-z)^{-1}\big]. \end{equation*} Since both Hamiltonians are self-adjoint on $\hilbert{H}_{0}$ we have \begin{equation*}\begin{split} & (\hepsi-z)^{-1} - (\hepsisigma-z)^{-1} = (\hepsisigma-z)^{-1}(\hepsisigma - \hepsi)(\hepsi-z)^{-1} =\\ &=-(\hepsisigma-z)^{-1}\Phi(\abs{k}v\id_{(0, \sigma)})(\hepsi-z)^{-1}\\ &\Rightarrow \norm{\chi(\hepsi) - \chi(\hepsisigma)}_{\mathcal{L}(\hilbert{H})}\leq\\ & \leq\frac{1}{\pi} \int_{\field{R}{2}} dxdy\, \abs{\bar{\partial}\chi^{a}(z)}\norm{(\hepsisigma-z)^{-1}}_{\mathcal{L}(\hilbert{H})}\norm{\Phi(\abs{k}v\id_{(0, \sigma)})(\hepsi-z)^{-1}}_{\mathcal{L}(\hilbert{H})}. \end{split} \end{equation*} It holds also that \begin{equation}\label{resolvbounded} \norm{(\hepsisigma-z)^{-1}}_{\mathcal{L}(\hilbert{H})} \leq \frac{C}{\abs{\Im z}}. \end{equation} This follows because $\hilbert{H}_{0}$ is dense in the domain of $H^{\epsi=0, \sigma=0}=L^{2}(\field{R}{3N}, D(\hfield))$, and for every $\Psi\in\hilbert{H}_{0}$ \begin{equation*} \hepsisigma\Psi \to H^{\epsi=0, \sigma=0}\Psi, \quad (\epsi, \sigma)\to (0, 0) \quad . \end{equation*} According to theorem VIII.25 \cite{ReSi1}, this implies that \begin{equation*} (\hepsisigma-z)^{-1}\Psi \to (H^{\epsi=0, \sigma=0}-z)^{-1}\Psi, \end{equation*} therefore $\abs{\Im z}\norm{(\hepsisigma-z)^{-1}\Psi}$ is bounded for every $\Psi$ and the uniform boundedness principle gives \eqref{resolvbounded}. For the second norm we have \begin{equation*}\begin{split} & \norm{\Phi(\abs{k}v\id_{(0, \sigma)})(\hepsi-z)^{-1}}_{\mathcal{L}(\hilbert{H})}\leq\\ &\leq\norm{\Phi(\abs{k}v\id_{(0, \sigma)})}_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}\cdot\norm{(\hepsi-z)^{-1}}_{\mathcal{L}(\hilbert{H}, \hilbert{H}_{0})}\leq\\ & \leq \frac{C}{\abs{\Im z}}\norm{\Phi(\abs{k}v\id_{(0, \sigma)})}_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}, \quad z\in \mathrm{supp}\,\chi^{a}, \end{split} \end{equation*} because $\hepsi$ and $\hfree$ define the same graph norm uniformly in $\epsi$, as it follows from \eqref{inequalityphi}. From the same equation and from \eqref{normsigma} it follows also that \begin{equation*} \norm{\Phi(\abs{k}v\id_{(0, \sigma)})}_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}\leq \norm{\abs{k}^{1/2}v\id_{(0, \sigma)}}_{L^{2}(\field{R}{3})} + \sqrt{2}\norm{\abs{k}v\id_{(0, \sigma)}}_{L^{2}(\field{R}{3})}\leq D\sigma^{1/2}, \end{equation*} which proves the statement. \end{proof} \section{The approximate dressing operator}\label{constructionunitary} In this section we construct the approximate dressing operator $\uepsisigma$ for the infrared regularized Hamiltonian $\hepsisigma$. We recall at the beginning the formal procedure to build the so called \emph{superadiabatic projectors}, introduced first on a formal level by Berry \cite{Ber} for time-dependent Hamiltonians and generalized by Nenciu and Sordoni \cite{NeSo}. The basic idea is to construct a sequence of projectors $P^{\epsi}_{j}, j\in\field{N}{}$, whose commutator with the Hamiltonian is of order $\Or(\epsi^{j+1})$. They will then be almost invariant for the corresponding time evolution, in the sense that \begin{equation*} [\E^{-\I t\hepsi/\epsi}, P^{\epsi}_{j}]=\Or(\epsi^{j}\abs{t})\quad . \end{equation*} As we have briefly mentioned in the introduction, this procedure, which allows to build adiabatic dynamics beyond the leading order, has been implemented till now only for fibered Hamiltonians whose spectrum is made up of eigenvalues separated by a gap. Exploiting the fact that the infrared cutoff $\sigma$ gives rise effectively to a gap, we can however carry out the calculations also in our case. We proceed first in a formal way, using the superadiabatic projectors just as a formal tool to deduce the expression of the unitary intertwiner $\hilbert{U}\footnote[1]{To simplify the notation, we drop from now on the dependence of the unitary on $\sigma$ and $\epsi$, denoting it simply by $\uone$ instead that by $\uepsisigma$, as we did in the introduction.}$. We will then show that the expression we find for $\uone$ actually defines a unitary operator on $\hilbert{H}$ and we will prove some of its useful properties, which allow us to fully characterize the domain of the effective Hamiltonian, and to expand it in a series of powers of $\epsi$ which is convergent in $\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})$. The almost invariant subspaces will then be defined in term of $\uone$, just as the unitary transformed of the \emph{reference subspaces} defined by the $\qm$s. It is important to stress that, to characterize the transitions between the different almost invariant subspaces, we need a unitary which maps all of them {\em simultaneously} to the corresponding reference subspaces. For this reason, we cannot use the procedure based on Nagy formula proposed in \cite{MaSo} \cite{NeSo} and we employ instead a simpler one, based on the construction of the projectors. In what follows, $\sigma$ is considered a parameter independent of $\epsi$. Only at the end, we will choose it as a suitable function of $\epsi$. \subsection{Formal definition}\label{formaldefinition} Using the dressing operator $V_{\sigma}(x)$, whose expression is given in \eqref{vsigma}, we can define dressed $M$-particles projections: \begin{equation}\label{pizn} \pizn(x):= \vsigma(x)\qm \vsigma(x)^{*}. \end{equation} Note that, for $M=0$, we have $\proj{0}{0}(x)=|\Omega_{\sigma}(x)\rangle\langle\Omega_{\sigma}(x)|$, the projection onto the ground state of $\hfibsigma(x)$, but, for $M\neq 0$, $\pizn(x)$ is not a spectral projection of the fibered Hamiltonian. The standard construction we briefly mentioned above, which is used in the case with gap, is based on an iterative procedure. Starting from a zero order projection $\proj{}{0}$, {\em which corresponds to a spectral subspace of the fibered Hamiltonian}, it allows to build an approximate orthogonal projection $\proj{(n)}{}$ which satisfies \begin{equation}\label{pin}\begin{split} & (\proj{(n)}{})^{*}=\proj{(n)}{}, \qquad (\proj{(n)}{})^{2} - \proj{(n)}{} = \Or(\epsi^{n+1}),\\ & [\hepsi, \proj{(n)}{}] = \Or(\epsi^{n+1}). \end{split} \end{equation} As we already stressed, we avoid on purpose to be more precise on the sense in which equation \eqref{pin} holds and on which norm one should use on the right-hand side, because we will use it just as a formal tool to deduce the expression of $\uone$. To clarify the method we employ, we recall briefly the derivation of the first order almost projection $\proj{(1)}{}$ in the case with gap, i.\ e., we assume for the moment that $\hfibsigma(x)$ has a nondegenerate eigenvalue $E_{\sigma}(x)$ which is isolated from the rest of the spectrum. Our discussion is taken from \cite{PST}. Therefore, we start from a projection $\proj{}{0}(x)$, which projects onto the eigenspace associated to $E_{\sigma}(x)$, and we proceed formally, without worrying about specifying any regularity assumption. It holds then that \begin{equation}\label{pizeroproj} \proj{*}{0}=\proj{}{0},\qquad (\proj{}{0})^{2} = \proj{}{0}, \end{equation} \begin{equation} [\proj{}{0}, \hepsisigma ] = [\proj{}{0}, \frac{1}{2}\pop^{2}] = \I\epsi(\nabla\proj{}{0})\cdot\pop + \frac{\epsi^{2}}{2}\Delta\proj{}{0}, \end{equation} where the right-hand side of the commutator is of order $\Or(\epsi)$ when applied to functions of bounded kinetic energy. Applying the inductive scheme, we determine now the coefficient $\proj{}{1}$ in $\proj{(1)}{} = \proj{}{0} + \epsi\proj{}{1}$ by requiring that \begin{equation}\label{propertiespione}\begin{split} & (\proj{(1)}{})^{*}=\proj{(1)}{},\qquad (\proj{(1)}{})^{2} - \proj{(1)}{} = \Or(\epsi^{2}),\\ & [\proj{(1)}{}, \hepsi] = \Or(\epsi^{2}). \end{split}\end{equation} The first condition gives \begin{equation*} (\proj{}{0} + \epsi\proj{}{1})^{2} - (\proj{}{0} + \epsi\proj{}{1}) = \epsi(\proj{}{0}\proj{}{1} + \proj{}{1}\proj{}{0} - \proj{}{1}) + \Or(\epsi^{2}) \eqex \Or(\epsi^{2}), \end{equation*} so that we must have \begin{equation}\label{pioneoffdiag} \proj{}{1} = \proj{}{0}\proj{}{1} + \proj{}{1}\proj{}{0} + \Or(\epsi). \end{equation} Concerning the commutator, we have \begin{equation*} [\proj{(1)}{}, \hepsisigma] = \I\epsi(\nabla\proj{}{0})\cdot\pop + \epsi[\proj{}{1}, \hfibsigma] + \Or(\epsi^{2})\eqex \Or(\epsi^{2}), \end{equation*} so $\proj{}{1}$ has to satisfy \begin{equation}\label{commeq} [\proj{}{1}, \hfibsigma] = -\I(\nabla\proj{}{0})\cdot\pop + \Or(\epsi). \end{equation} It can be shown (\cite{Te3}, lemma $3.8$) that the unique solution, up to terms of order $\Or(\epsi)$, of \eqref{commeq} is given by \begin{equation}\label{solcommeq} -\I\proj{}{0}(\nabla\proj{}{0})(\hfibsigma - E)^{-1}\proj{\perp}{0}\cdot\pop\ +\ \textrm{adj.}\quad , \end{equation} where $\proj{\perp}{0}:= \mathbf{1} - \proj{}{0}$, and $+$ adj. means that the adjoint of everything which lies to the left is added. This ensures also that $\proj{}{1}$ is formally self-adjoint. If there is a gap, the reduced resolvent is certainly bounded in norm: \begin{equation} \norm{(\hfibsigma - E)^{-1}\proj{\perp}{0}}_{\mathcal{L}(\hilbert{H})} = \bigg[\inf_{x\in\field{R}{3n}}\textrm{dist}\bigg(E_{\sigma}(x), \sigma(\hfibsigma(x))\backslash\{E_{\sigma}(x)\}\bigg)\bigg]^{-1}, \end{equation} but in general the construction breaks down in the case without gap, because the reduced resolvent can be unbounded. A possible way out of this difficulty was proposed in \cite{Te1} and \cite{Te2} to prove the adiabatic decoupling for the zero order subspaces. It consists in introducing a second parameter $\delta>0$, by which the reduced resolvent is shifted into the complex plane: instead of $(\hfib-E)^{-1}\proj{\perp}{0}$, one considers $(\hfib-E-\I\delta)^{-1}\proj{\perp}{0}$. In our case, the infrared cutoff $\sigma$ plays an analogous role. Actually, at the end of our formal calculations, it will result that the reduced resolvent of $\hfibsigma$ is bounded. We carry out moreover another generalization of the standard construction, extending it to the situation when $M\neq 0$. To understand the way how we proceed it is useful to analyze more explicitly the case $M=0$. According to the above discussion we define \begin{equation*}\begin{split} & \proj{(1)}{0}(\epsi, \sigma):= \proj{0}{0} + \epsi\proj{0}{1}(\sigma),\\ & \proj{0}{1}(\sigma):= -\I(\nabla\proj{0}{0})\rfib(E_{\sigma})\cdot\pop\ +\ \textrm{adj.}\quad , \end{split} \end{equation*} where $\rfib(E_{\sigma}):= (\hfibsigma-E_{\sigma})^{-1}\proj{0\,\perp}{0}$. We have omitted the $\proj{0}{0}$ because of the well-known fact that $\nabla\proj{0}{0}$ is off-diagonal with respect to the block decomposition induced by $\proj{0}{0}$. Using the fact that $\nabla\proj{\textrm{0}}{0}=\I V_{\sigma}(x)[\Phi(\I\nabla_{x}v_{\sigma}), Q_{0}]V_{\sigma}(x)^{*}$, and introducing the notation $\rfield(0):= \hfield^{-1}Q_{0}^{\perp}$, we get then \begin{equation*}\begin{split} & \proj{0}{1} = \vsigma(x)[\Phi(\I\nabla_{x}v_{\sigma}), Q_{0}]\rfield(0) \vsigma(x)^{*}\cdot\pop +\\ & -\pop\cdot \vsigma(x)\rfield(0)[\Phi(\I\nabla_{x}v_{\sigma}), Q_{0}]\vsigma(x)^{*}=\\ &= -\vsigma(x)\big[\dgamma(\abs{k}^{-1}), [\Phi(\I\nabla_{x}v_{\sigma}), Q_{0}]\big]\cdot\pop\vsigma(x)^{*} + \Or(\epsi), \end{split} \end{equation*} where we have used the fact that the commutator between $\pop$ and a function of $x$ is of order $\Or(\epsi)$ (we stress again that we don't worry here about smoothness assumptions, the calculations should be considered as formal) and \begin{equation*} \rfield(0)Q_{\leq 1} = \dgamma(\abs{k}^{-1})Q_{\leq 1}, \end{equation*} together with \begin{equation*} Q_{0}\Phi(\I\nabla_{x}v_{\sigma})=Q_{0}\Phi(\I\nabla_{x}v_{\sigma})Q_{\leq 1}, \quad \Phi(\I\nabla_{x}v_{\sigma})Q_{0}=Q_{\leq 1}\Phi(\I\nabla_{x}v_{\sigma}) Q_{0}. \end{equation*} Calculating the commutator using equation \eqref{commutators}, we get \begin{equation*} \proj{0}{1} = \I\vsigma(x)[Q_{0}, \phisigma]\cdot\pop\vsigma(x)^{*}, \end{equation*} where we define \begin{equation}%\label{defadelta} \phisigma:= \Phi\bigg(\frac{\nabla_{x}v_{\sigma}(x, k)}{\abs{k}}\bigg). \end{equation} Since $\nabla_{x}v_{\sigma}(x, k)\abs{k}^{-1}\in L^{2}(\field{R}{3}, dk)$, $\proj{0}{1}$ is well-defined, and one can symmetrize it to get a symmetric operator (the fact that $\proj{0}{1}$ is not bounded due to the presence of $\pop$ will be dealt with below). It is now fairly clear how to proceed. For any $M$, we define \begin{equation}\begin{split} & \proj{(1)}{M}:= \proj{M}{0} + \epsi\proj{M}{1},\\ & \proj{M}{1}:= \frac{\I}{2}\vsigma\big\{[\qm, \phisigma]\cdot\pop+\pop\cdot[\qm, \phisigma]\big\} \vsigma^{*} \quad . \end{split} \end{equation} Formally $\proj{(1)}{M}$ satisfies \eqref{propertiespione} for any $M$. In fact, $\phisigma$ is adjoint, therefore $\I[\qm, \phisigma]$ is self-adjoint, and $(\proj{(1)}{M})^{*}=\proj{(1)}{M}$. Moreover, since $\proj{M}{1}$ is off-diagonal with respect to $\proj{M}{0}$, \begin{equation*} \proj{M}{0}\proj{M}{1}\proj{M}{0} = (\mathbf{1} - \proj{M}{0})\proj{M}{1}(\mathbf{1}-\proj{M}{0})=0, \end{equation*} equation \eqref{pioneoffdiag} holds exactly, without $\Or(\epsi)$ corrections. Concerning the commutator equation \eqref{commeq}, we get \begin{equation*}\begin{split} & [\proj{M}{1}, \hfibsigma] = \vsigma\big[[\qm, \I\phisigma]\cdot\pop, \hfield + E(x)\big]\vsigma^{*}+\Or(\epsi)=\\ &=\vsigma\big[[\qm, \I\phisigma], \hfield]\big]\cdot\pop \vsigma^{*} +\Or(\epsi)=\vsigma\big[[\hfield, \I\phisigma], \qm\big]\cdot\pop \vsigma^{*}+ \Or(\epsi). \end{split} \end{equation*} Applying again equation \eqref{commutators} we get, on $\fock_{\abs{k}, \mathrm{fin}}$, \begin{equation}\label{commhfa} [\hfield, \phisigma] = \Phi(\I\nabla_{x}v_{\sigma}), \end{equation} so that \begin{equation}\label{commeqsatisfied} [\proj{M}{1}, \hfibsigma] = -\I\nabla\proj{M}{0}\cdot\pop + \Or(\epsi). \end{equation} The next step is to find a unitary operator which intertwines the almost projections $\proj{(1)}{M}$ with the reference projections given by the $\qm$ up to terms of order $\Or(\epsi^{2})$. Employing a procedure analogous to the one we used for the projections, we assume that we can write an expansion \begin{equation} U^{(n)}:= \sum_{j=1}^{n}\epsi^{n}U_{n}, \end{equation} starting from a known $U_{0}$, and imposing that \begin{equation}\label{requnitary}\begin{split} & U^{(n)}U^{(n)*} - \mathbf{1}=\Or(\epsi^{n+1}),\qquad U^{(n)*}U^{(n)} - \mathbf{1}=\Or(\epsi^{n+1})\quad ,\\ & U^{(n)}\proj{(1)}{M}U^{(n)*}= \qm +\Or(\epsi^{n+1})\quad , \end{split} \end{equation} to deduce the coefficients $U_{n}$ iteratively. An obvious choice for the zero order unitary is given by \begin{equation} U_{0}(x):= \vsigma(x)^{*}, \end{equation} which satisfies $U_{0}(x)\proj{M}{0}(x)U_{0}(x)^{*}=\qm$. Without loss of generality, we assume that $U_{1}=TU_{0}$, for some operator $T$ on $\hilbert{H}$. The requirements \eqref{requnitary} then lead to \begin{equation*}\begin{split} & (U_{0}+\epsi U_{1})(U_{0}^{*}+\epsi U_{1}^{*}) = \mathbf{1} + \epsi(U_{0}U_{1}^{*} + U_{1}U_{0}^{*}) + \Or(\epsi^{2})=\mathbf{1} + \epsi(T^{*} + T) +\\ &+ \Or(\epsi^{2}) \eqex \Or(\epsi^{2}) \Rightarrow T^{*}+T=\Or(\epsi), \end{split} \end{equation*} and \begin{equation*}\begin{split} & (U_{0}+\epsi U_{1})(\proj{M}{0}+\epsi\proj{M}{1})(U_{0}^{*}+\epsi U^{*}_{1}) = (\mathbf{1}+\epsi T)(\qm + \epsi[\qm, \I\phisigma]\cdot\pop)(\mathbf{1}-\epsi T) +\\ &+ \Or(\epsi^{2}) = \qm +\epsi([\qm, \I\phisigma]\cdot\pop - [\qm, T]) + \Or(\epsi^{2}), \end{split} \end{equation*} therefore choosing \begin{equation*} T=\I\phisigma\cdot\pop \end{equation*} we satisfy both, for every $M$. The first order almost unitary is then given by \begin{equation} U^{(1)}=(\mathbf{1}+\I\epsi \phisigma\cdot\pop)\vsigma^{*}. \end{equation} \subsection{Rigorous definition and properties} To give a meaning to the till now formal expression for $U^{(1)}$ we have to deal with two problems. The first is due to the fact that the operator $\phisigma$ is unbounded. We introduce therefore in its definition a cutoff in the number of particles, replacing it by \begin{equation}\label{defamdelta} \phisigmaj:=\qless{J}\phisigma\qless{J}. \end{equation} The operator \begin{equation*} U^{(1)}_{\mathrm{J}}:=(\mathbf{1}+\I\epsi\phisigmaj\cdot\pop)\vsigma^{*} \end{equation*} is again formally almost unitary up to order $\Or(\epsi^{2})$, and intertwines the superadiabatic almost projectors for $M<\mathrm{J}$, \begin{equation} U^{(1)}_{\mathrm{J}}\proj{(1)}{M}U^{(1)*}_{\mathrm{J}}= \qm +\Or(\epsi^{2}), \qquad M<\mathrm{J}. \end{equation} This means that we can use $U^{(1)}_{\mathrm{J}}$ to study the transitions among superadiabatic subspaces up to a fixed, but arbitrary $\mathrm{J}$. The second problem, already mentioned in the introduction, is related to the presence of the unbounded momentum operator $\pop$. To make the whole expression bounded we introduce a cutoff in the total energy. More precisely, given a function $\chi\in\coinf(\field{R}{})$, we define \begin{equation}\label{uonechi}\begin{split} & U^{(1)}_{\mathrm{J}, \chi}:= \vsigma^{*}[\mathbf{1}+\I\epsi \vsigma\phisigmaj\cdot\pop \vsigma^{*} -\I\epsi(\mathbf{1}-\chi(\hepsisigma))\vsigma\phisigmaj\cdot\pop \vsigma^{*}(\mathbf{1}-\chi(\hepsisigma))]=\\ &=\vsigma^{*}[\mathbf{1} + \I\epsi\chi(\hepsisigma)\vsigma\phisigmaj\cdot\pop \vsigma^{*} + \I\epsi(\mathbf{1}-\chi(\hepsisigma))\vsigma\phisigmaj\cdot\pop \vsigma^{*}\chi(\hepsisigma)]\quad . \end{split} \end{equation} Note that $U^{(1)}_{\mathrm{J}, \chi}\tilde{\chi}(\hepsisigma)=U_{\mathrm{J}}^{(1)}\tilde{\chi}(\hepsisigma)$, for every $\tilde{\chi}\in\coinf(\field{R}{})$ such that $\chi\tilde{\chi}=\tilde{\chi}$. In the context of Born-Oppenheimer approximation, Sordoni \cite{So} applied a similar strategy in order to define bounded superadiabatic projections. Our aim in this section is first to prove that $U^{(1)}_{\mathrm{J}, \chi}\in\mathcal{L}(\hilbert{H})\cap\mathcal{L}(\hilbert{H}_{0})$. Once we have shown this, we can define a true unitary $\hilbert{U}$ through the formula \begin{equation*} \hilbert{U}:= U^{(1)}_{\mathrm{J}, \chi}[U^{(1)\,*}_{\mathrm{J}, \chi}U^{(1)}_{\mathrm{J}, \chi}]^{-1/2}. \end{equation*} We will then prove that both $\hilbert{U}$ and $\hilbert{U}^{-1}=\hilbert{U}^{*}$ belong to $\mathcal{L}(\hilbert{H})\cap\mathcal{L}(\hilbert{H}_{0})$, i.\ e., that $\hilbert{U}$ is a bijection on $\hilbert{H}_{0}$. We will in addition show that it can be expanded in a convergent power series both in $\mathcal{L}(\hilbert{H})$ and in $\mathcal{L}(\hilbert{H}_{0})$. \begin{lemma}\label{lemmav} $\vsigma$ and $\vsigma^{*}$ belong to $\mathcal{L}(\hilbert{H}_{0})$. Moreover \begin{equation} \norm{\vsigma}_{\mathcal{L}(\hilbert{H}_{0})}\leq C \quad . \end{equation} An analogous estimate holds for $\vsigma^{*}$. \end{lemma} \begin{proof}({\em We give the proof for $\vsigma$. The one for $\vsigma^{*}$ is the same up to some changes in the signs}). We can calculate the norm on a dense subset. It is known (\cite{ReSi1}, theorem VIII.33) that $H_{0}$ is essentially self-adjoint on a set of the form \begin{equation*} \mathcal{D}_{\mathrm{p}}\otimes\mathcal{D}_{\mathrm{f}}:=\mathcal{L}\{\psi\otimes\varphi: \psi\in\mathcal{D}_{\mathrm{p}}, \varphi\in\mathcal{D}_{\mathrm{f}}\}, \end{equation*} where $\mathcal{L}$ means ``finite linear combinations of'', $\mathcal{D}_{\mathrm{p}}$ is a core for $\pop^{2}$ and $\mathcal{D}_{\mathrm{f}}$ is a core for $\hfield$. We can choose then $\mathcal{D}_{\mathrm{p}}=\coinf(\field{R}{3n})$ and (\cite{ReSi2}, section X.7) \begin{equation}\label{df} \mathcal{D}_{\mathrm{f}}=\{\varphi\in\fock_{\mathrm{fin}}: \varphi^{(\N)}\in\otimes_{j=1}^{\N}\coinf(\field{R}{3})\cap L^{2}_{\mathrm{s}}(\field{R}{3\N})\}, \end{equation} where $\otimes_{j=1}^{\N}\coinf(\field{R}{3}):= \mathcal{L}\{\varphi_{1}\otimes\ldots\otimes\varphi_{\N}: \varphi_{j}\in\coinf(\field{R}{3}), j=1, \ldots, \N\}$. The vectors in $\fock_{\mathrm{fin}}$ are analytic vectors for $\vsigma(x)$ (\cite{ReSi2}, theorem X.41), so, if $\psi\in\mathcal{D}_{\mathrm{p}}$, $\varphi\in\mathcal{D}_{\mathrm{f}}$, we have \begin{equation*} \vsigma(\psi\otimes\varphi)(x) = \psi(x)\vsigma(x)\phi = \sum_{j=0}^{\infty}\I^{j}\frac{\Phi(\I v_{\sigma}(x, \cdot))^{j}}{j!}\psi(x)\varphi. \end{equation*} Moreover \begin{equation*}\begin{split} & (\pop^{2}\otimes\mathbf{1})\Phi(\I v_{\sigma}(x, \cdot))^{j}\psi(x)\varphi = -\epsi^{2}j\Phi(\I v_{\sigma})^{j-1}\Phi(\I\Delta v_{\sigma})\psi(x)\varphi+\\ & -\epsi^{2}j(j-1)\Phi(\I v_{\sigma})^{j-2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\Phi(\I\nabla_{x}v_{\sigma})\psi(x)\varphi+\\ & -2\I\epsi j\Phi(\I v_{\sigma})^{j-1}\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop\psi(x)\varphi + \Phi(\I v_{\sigma})^{j}\pop^{2}\psi(x)\varphi, \end{split} \end{equation*} where we have used \eqref{commphi} and the fact that $\Im\langle v_{\sigma}(x, \cdot), \nabla_{x}v_{\sigma}(x, \cdot)\rangle_{L^{2}(\field{R}{3})}=0$. >From the previous equations it follows that \begin{equation*} \sum_{j=0}^{J}(\pop^{2}\otimes\mathbf{1})\I^{j}\frac{\Phi(\I v_{\sigma}(x, \cdot))^{j}}{j!}\psi(x)\varphi \end{equation*} is convergent, so that $\vsigma(\psi\otimes\varphi)\subset D(\pop^{2}\otimes\mathbf{1})$ and \begin{equation}\label{commpv}\begin{split} & (\pop^{2}\otimes\mathbf{1})\vsigma(\psi\otimes\varphi)=-\I\epsi^{2}\vsigma\Phi(\I\Delta v_{\sigma})(\psi\otimes\varphi)+\epsi^{2}\vsigma\Phi(\I\nabla_{x}v_{\sigma})\cdot\\ &\cdot\Phi(\I\nabla_{x}v_{\sigma})(\psi\otimes\varphi)+2\epsi \vsigma\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop(\psi\otimes\varphi)+\vsigma(\pop^{2}\otimes\mathbf{1})(\psi\otimes\varphi). \end{split} \end{equation} This implies \begin{equation}\label{estimatepsquare}\begin{split} &\norm{(\pop^{2}\otimes\mathbf{1})\vsigma(\psi\otimes\varphi)}\leq \epsi^{2}\norm{\Phi(\I\Delta v_{\sigma})(\psi\otimes\varphi)}+\epsi^{2}\cdot\\ &\cdot\norm{\Phi(\I\nabla_{x}v_{\sigma})\cdot\Phi(\I\nabla_{x}v_{\sigma})(\psi\otimes\varphi)}+2\epsi\norm{\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop(\psi\otimes\varphi)}+\\ &+\norm{(\pop^{2}\otimes\mathbf{1})(\psi\otimes\varphi)}. \end{split} \end{equation} Using equation \eqref{inequalityphi} we can bound the first term with some constant independent of $\sigma$ and $\epsi$ times $\norm{\psi\otimes\varphi}_{\hilbert{H}_{0}}$. This happens because the constants involved in \eqref{inequalityphi} contain terms of the form $\norm{f/\sqrt{\abs{k}}}_{L^{2}(\field{R}{3})}$, but when one differentiates $v_{\sigma}$ with respect to $x$, one gets an additional $\abs{k}$, therefore all the terms of the form $\norm{\nabla_{x}v_{\sigma}/\sqrt{\abs{k}}}$ and so on are uniformly bounded in $\sigma$. The same reasoning applies also to all the estimates which follow. For the second term in \eqref{commpv}, using again \eqref{inequalityphi}, and the notation $\Psi=\psi\otimes\varphi$ and $f=\I\deriv{j}{}v_{\sigma}$, we get \begin{equation*}\begin{split} &\norm{\Phi(f)^{2}\Psi}^{2}_{\hilbert{H}}\leq 2\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\langle\Phi(f)\Psi, (\mathbf{1}\otimes\hfield)\Phi(f)\Psi\rangle_{\hilbert{H}} +\\ &+2\norm{f}^{2}_{L^{2}(\field{R}{3})}\norm{\Phi(f)\Psi}_{\hilbert{H}}^{2}\leq\\ &\leq \sqrt{2}\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\langle\Phi(f)\Psi, \big[a(\abs{\,\cdot\,}f)^{*}-a(\abs{\,\cdot\,}f)\big]\Psi\rangle_{\hilbert{H}} +\\ &+ 2\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\langle\Phi(f)^{2}\Psi, \hfield\Psi\rangle_{\hilbert{H}} +2\norm{f}^{2}_{L^{2}(\field{R}{3})}\norm{\Phi(f)\Psi}_{\hilbert{H}}^{2}\leq\\ &\leq \sqrt{2}\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\cdot\norm{\Phi(f)\Psi}_{\hilbert{H}}\cdot\norm{\big[a(\abs{\,\cdot\,}f)^{*}-a(\abs{\,\cdot\,}f)\big]\Psi}_{\hilbert{H}}+\\ &+2\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\cdot\norm{\Phi(f)^{2}\Psi}_{\hilbert{H}}\norm{\hfield\Psi}_{\hilbert{H}}+2\norm{f}^{2}_{L^{2}(\field{R}{3})}\norm{\Phi(f)\Psi}_{\hilbert{H}}^{2}\leq\\ &\leq \frac{a^{2}}{\sqrt{2}}\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\norm{\Phi(f)\Psi}^{2}_{\hilbert{H}}+\frac{1}{a^{2}\sqrt{2}}\norm{\big[a(\abs{\,\cdot\,}f)^{*}-a(\abs{\,\cdot\,}f)\big]\Psi}_{\hilbert{H}}^{2}+\\ &+b^{2}\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\norm{\Phi(f)^{2}\Psi}^{2}_{\hilbert{H}}+\frac{1}{b^{2}}\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\norm{\hfield\Psi}^{2}_{\hilbert{H}}+\\ &+2\norm{f}^{2}_{L^{2}(\field{R}{3})}\norm{\Phi(f)\Psi}_{\hilbert{H}}^{2},\qquad\forall a, b>0,\\ &\Rightarrow\norm{\Phi(f)^{2}\Psi}^{2}_{\hilbert{H}}\big(1-b^{2}\norm{f/\sqrt{\abs{\,\cdot\,}}}^{2})\leq\\ &\leq \frac{a^{2}}{\sqrt{2}}\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\norm{\Phi(f)\Psi}^{2}_{\hilbert{H}}+\frac{1}{a^{2}\sqrt{2}}\norm{\big[a(\abs{\,\cdot\,}f)^{*}-a(\abs{\,\cdot\,}f)\big]\Psi}_{\hilbert{H}}^{2}+\\ &+\frac{1}{b^{2}}\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\norm{\hfield\Psi}^{2}_{\hilbert{H}}+2\norm{f}^{2}_{L^{2}(\field{R}{3})}\norm{\Phi(f)\Psi}_{\hilbert{H}}^{2}. \end{split} \end{equation*} Each term on the right-hand side can be bounded by a constant independent of $\sigma$ times $\norm{\Psi}_{\hilbert{H}_{0}}$. For the third term in \eqref{estimatepsquare} we get \begin{equation*}\begin{split} & \norm{\Phi(f)\pop_{j}\Psi}_{\hilbert{H}}^{2}\leq 2\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\langle\pop_{j}\Psi, \hfield\pop_{j}\Psi\rangle + 2\norm{f}_{L^{2}(\field{R}{3})}^{2}\norm{\pop_{j}\Psi}^{2}_{\hilbert{H}}\leq\\ &\leq 2\norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\langle\Psi, \hfield\pop_{j}^{2}\Psi\rangle + 2\norm{f}_{L^{2}(\field{R}{3})}^{2}\norm{\pop_{j}\Psi}^{2}_{\hilbert{H}}\leq\\ &\leq \norm{f/\sqrt{\abs{\,\cdot\,}}}_{L^{2}(\field{R}{3})}^{2}\norm{(\pop_{j}^{2}+\hfield)\Psi}_{\hilbert{H}}^{2} +2\norm{f}_{L^{2}(\field{R}{3})}^{2}\norm{\pop_{j}\Psi}^{2}_{\hilbert{H}}, \end{split} \end{equation*} and we can again bound the right-hand side by a constant times $\norm{\Psi}_{\hilbert{H}_{0}}$. We have now to prove similar estimates for $\norm{(\mathbf{1}\otimes\hfield) V(\psi\otimes\varphi)}$, and we are done. For this we need that, on $\mathcal{D}_{\mathrm{p}}\otimes\mathcal{D}_{\mathrm{f}}$, \begin{equation*} [\hfield, \Phi(\I v_{\sigma})]=\I\Phi(\abs{k}v_{\sigma}), \end{equation*} and that \begin{equation*}\begin{split} & [\Phi(\abs{k}v_{\sigma}), \Phi(\I v_{\sigma})]=\I\Im\langle\abs{k}v_{\sigma}, \I v_{\sigma}\rangle=\I\int dk\,\abs{k}\abs{v_{\sigma}(x, k)}^{2}=-2\I E_{\sigma}(x)\\ &\Rightarrow [\Phi(\abs{k}v_{\sigma}), \Phi(\I v_{\sigma})^{j}]=-2\I j E_{\sigma}(x)\Phi(\I v_{\sigma})^{j-1}. \end{split} \end{equation*} We have then \begin{equation}\begin{split} & (\mathbf{1}\otimes\hfield)\Phi(\I v_{\sigma}(x, \cdot))^{j}(\psi\otimes\varphi)=\I\sum_{l=0}^{j-1}\Phi(\I v_{\sigma})^{l}\Phi(\abs{k}v_{\sigma})\Phi(\I v_{\sigma})^{j-l-1}(\psi\otimes\varphi)+\\ &+\Phi(\I v_{\sigma})^{j}(\mathbf{1}\otimes\hfield)(\psi\otimes\varphi)=\\ &=\I j\Phi(\I v_{\sigma})^{j-1}\Phi(\abs{k}v_{\sigma})(\psi\otimes\varphi) +E_{\sigma}(x)j(j-1)\Phi(\I v_{\sigma})^{j-2}(\psi\otimes\varphi)+\\ &+\Phi(\I v_{\sigma})^{j}(\mathbf{1}\otimes\hfield)(\psi\otimes\varphi). \end{split} \end{equation} This means, as for $\pop^{2}\otimes\mathbf{1}$, that $\vsigma(\psi\otimes\varphi)\subset D(\mathbf{1}\otimes\hfield)$, and that \begin{equation} (\mathbf{1}\otimes\hfield)\vsigma(\psi\otimes\varphi)=-\vsigma\Phi(\abs{k}v_{\sigma})(\psi\otimes\varphi)- E_{\sigma}(x)\vsigma(\psi\otimes\varphi)+\vsigma(\mathbf{1}\otimes\hfield)(\psi\otimes\varphi). \end{equation} Using equation \eqref{inequalityphi}, we can again bound each term by a constant independent of $\sigma$ times $\norm{\psi\otimes\varphi}_{\hilbert{H}_{0}}$. \end{proof} \begin{lemma}\label{lemmaamdelta}\begin{enumerate} \item For each $x\in\field{R}{3N}$ and $l=1, \ldots, 3N$, $\phisigmajl(x)\in\mathcal{L}(\fock)$, and $\phisigmajl(x)^{*}=\phisigmajl(x)$. Moreover, \begin{equation} \phisigmajl: \field{R}{3N}\to\mathcal{L}(\fock), \qquad x\mapsto \phisigmajl(x), \quad\in C_{\mathrm{b}}^{\infty}(\field{R}{3N}, \mathcal{L}(\fock)), \end{equation} and, for $\sigma$ small enough, \begin{equation} \norm{\phisigmajl}_{\mathcal{L}(\hilbert{H})}\leq C\sqrt{\mathrm{J}+1}\sqrt{\log(\sigma^{-1})}. \end{equation} Given $\alpha\in\field{N}{3N}$ with $\abs{\alpha}>0$, it holds instead \begin{equation} \deriv{x}{\alpha}\phisigmajl = \deriv{x}{\alpha}\phizerojl + \Or(\sigma^{\abs{\alpha}}\sqrt{\mathrm{J}+1})_{\mathcal{L}(\hilbert{H})}, \end{equation} where \begin{equation} \deriv{x}{\alpha}\phizerojl := (\deriv{x}{\alpha}\phisigmajl)_{|\sigma=0} \end{equation} is a well-defined bounded operator on $\hilbert{H}$. \item The statements of point $1$ (except for the self-adjointness of $\phisigmajl(x)$) remain true if $\fock$ is replaced by $D(\hfield)$. \end{enumerate} \end{lemma} \begin{corollary} The fibered operators $\deriv{x}{\alpha}\phisigmajl$ belong to $\mathcal{L}(\hilbert{H})\cap\mathcal{L}(\hilbert{H}_{0})$ $\forall\alpha\in\field{N}{3N}$. \end{corollary} \begin{proof}(Of the lemma) $1.$ The self-adjointness of $\Phi$ was already discussed in section \ref{nelsonmodel}. Concerning the other statements, we first remark that in general \begin{equation*} \norm{\qless{J}\Phi(f(x, k))\qless{J}}_{\mathcal{L}(\hilbert{H})}\leq 2^{1/2}\sqrt{\mathrm{J}+1}\cdot\sup_{x\in\field{R}{3N}}\norm{f(x, k)}_{L^{2}(\field{R}{3}, dk)}, \end{equation*} so that \begin{equation*} \norm{\deriv{x}{\alpha}\phisigmajl}_{\mathcal{L}(\hilbert{H})}\leq 2^{1/2}\sqrt{\mathrm{J}+1}\sup_{x\in\field{R}{3n}}\bigg\lVert\frac{\deriv{x}{\alpha}\deriv{l}{}v_{\sigma}(x, k)}{\abs{k}}\bigg\rVert_{L^{2}(\field{R}{3}, dk)}. \end{equation*} For $\alpha=0$ we have then \begin{equation*}\begin{split} & \deriv{j}{}v_{\sigma}(x, k)=\deriv{(j_{1}, j_{2})}{}v_{\sigma}(x, k)=\I\kappa_{j_{2}}\frac{\id_{(\sigma, \infty)}(k)\hat{\varrho}_{j_{1}}(k)}{\abs{k}^{1/2}}\E^{\I k\cdot x_{j_{1}}},\\ & \bigg\lVert\frac{\deriv{j}{}v_{\sigma}(x, k)}{\abs{k}}\bigg\rVert^{2}_{L^{2}(\field{R}{3}, dk)} \leq \tilde{C}\int_{\sigma}^{\Lambda}\, d\abs{k}\, \frac{1}{\abs{k}}\leq C\log(\sigma^{-1}). \end{split} \end{equation*} If $\abs{\alpha}>0$, we get instead \begin{equation*}\begin{split} &\deriv{x}{\alpha}\deriv{j}{}v_{\sigma}(x, k)=(\I\kappa_{1})^{\alpha_{j_{1}, 1}}(\I\kappa_{2})^{\alpha_{j_{1}, 2}}(\I\kappa_{3})^{\alpha_{j_{1}, 3}}\abs{k}^{\alpha_{j_{1},1}+\alpha_{j_{1},2}+\alpha_{j_{1},3}-1/2}\cdot\\ &\cdot(\I\kappa_{j_{2}})\id_{(\sigma, \infty)}(k)\hat{\varrho}_{j_{1}}(k)\E^{\I k\cdot x_{j_{1}}}, \end{split} \end{equation*} therefore we can split \begin{equation*} \frac{\deriv{x}{\alpha}\deriv{j}{}v_{\sigma}(x, k)}{\abs{k}} = \frac{\deriv{x}{\alpha}\deriv{j}{}v(x, k)}{\abs{k}} - \frac{\id_{(0, \sigma)}\deriv{x}{\alpha}\deriv{j}{}v(x, k)}{\abs{k}}, \end{equation*} where \begin{equation*} \sup_{x\in\field{R}{3N}}\bigg\lVert\frac{\deriv{x}{\alpha}\deriv{j}{}v(x, k)}{\abs{k}}\bigg\rVert_{L^{2}(\field{R}{3}, dk)}<\infty, \end{equation*} and \begin{equation*} \sup_{x\in\field{R}{3N}}\bigg\lVert\frac{\id_{(0, \sigma)}\deriv{x}{\alpha}\deriv{j}{}v(x, k)}{\abs{k}}\bigg\rVert_{L^{2}(\field{R}{3}, dk)}\leq C\sigma^{\abs{\alpha}}\quad . \end{equation*} $2.$ We remark that we can check that $\phisigmaj(x)$ is bounded on the dense set $\mathcal{D}_{\mathrm{f}}$ defined in \eqref{df}. Therefore, for $\varphi\in\mathcal{D}_{\mathrm{f}}$, we have \begin{equation*}\begin{split} &(\phisigmaj(x)\varphi)^{(M)}(k_{1}, \ldots, k_{M})=\frac{1}{\sqrt{2}}\sqrt{M+1}\int dk\,\frac{\nabla_{x}v_{\sigma}(x, k)^{*}}{\abs{k}}\cdot\\ &\cdot\varphi^{(M+1)}(k, k_{1}, \ldots, k_{M})+\frac{1}{\sqrt{2M}}\sum_{j=1}^{M}\frac{\nabla_{x}v_{\sigma}(x, k_{j})}{\abs{k_{j}}}\varphi^{(M-1)}(k_{1},\ldots, \hat{k}_{j},\ldots k_{M}). \end{split} \end{equation*} Because of the ultraviolet cutoff \begin{equation*} \abs{k}^{\alpha}\frac{\nabla_{x}v_{\sigma}(x, k)}{\abs{k}}\in L^{2}(\field{R}{3}, dk), \quad\forall\alpha>0, \end{equation*} so $\phisigmaj(x)\varphi\in D(\hfield)$ and \begin{equation*}\begin{split} & \hfield\phisigmaj(x)\varphi=[\hfield, \phisigmaj]\varphi + \phisigmaj(x)\hfield\varphi=-\I\qless{J}\Phi(\I\nabla_{x}v_{\sigma})\qless{J}\varphi +\\ &+\phisigmaj(x)\hfield\varphi, \end{split} \end{equation*} which is bounded by a constant independent of $\sigma$ times $\norm{\varphi}_{D(\hfield)}$. The same reasoning holds if we replace $\nabla_{x}v_{\sigma}(x, k)$ by $\deriv{x}{\beta}v_{\sigma}(x, k)$, for any $\beta\in\field{N}{3N}$, so the lemma follows. \end{proof} \begin{lemma}\label{lemmauonechi} $U^{(1)}_{\mathrm{J}, \chi}$ belongs to $\mathcal{L}(\hilbert{H})\cap\mathcal{L}(\hilbert{H}_{0})$ and \begin{equation}\label{normujchi} \norm{U^{(1)}_{\mathrm{J}, \chi}}_{\mathcal{L}(\hilbert{K})}\leq C(1+\epsi\sqrt{\log(\sigma^{-1})}), \end{equation} where $\hilbert{K}=\hilbert{H}$ or $\hilbert{H}_{0}$. The same holds for $U^{(1)\,*}_{\mathrm{J}, \chi}$. \end{lemma} \begin{proof} We start considering $U_{\mathrm{J}, \chi}^{(1)}$. We have already shown that $\vsigma$ and $\vsigma^{*}$ belong to $\mathcal{L}(\hilbert{H})\cap\mathcal{L}(\hilbert{H}_{0})$ with uniformly bounded norms, so we have to examine \begin{equation*} \epsi\chi(\hepsisigma)\vsigma\phisigmaj\cdot\pop \vsigma^{*}, \quad \epsi(\mathbf{1}-\chi(\hepsisigma))V\phisigmaj\cdot\pop \vsigma^{*}\chi(\hepsisigma) \quad. \end{equation*} Given $\Psi\in\mathcal{D}_{\mathrm{p}}\otimes\mathcal{D}_{\mathrm{f}}$ (for the definitions, see the proof of lemma \ref{lemmav}) we have \begin{equation*} \chi(\hepsisigma)\vsigma\phisigmaj\cdot\pop \vsigma^{*}\Psi=-\I\epsi\chi(\hepsisigma)\nabla_{x}\cdot(\vsigma\phisigmaj)\vsigma^{*}\Psi+\chi(\hepsisigma)\pop\cdot \vsigma\phisigmaj \vsigma^{*}\Psi. \end{equation*} Since $\nabla_{x}\cdot(\vsigma\phisigmaj)$ and $\chi(\hepsisigma)\pop$ are bounded operators (see lemma \ref{lemmav} and \ref{lemmaamdelta}) the left-hand side belongs to $\mathcal{L}(\hilbert{H})\cap\mathcal{L}(\hilbert{H}_{0})$. Moreover, from the previous lemma it follows that \begin{equation*} \norm{\phisigmaj}_{\mathcal{L}(\hilbert{K})}\leq C \sqrt{\log(\sigma^{-1})}\quad , \end{equation*} while all the other terms have uniformly bounded norms. For the second one, \begin{equation*} \vsigma\phisigmaj\cdot\pop \vsigma^{*}\chi(\hepsisigma)\Psi=\vsigma\phisigmaj\cdot[\pop, \vsigma^{*}]\chi(\hepsisigma)\Psi + \vsigma\phisigmaj\cdot \vsigma^{*}\pop\chi(\hepsisigma)\Psi. \end{equation*} $\pop\chi(\hepsi)\in\mathcal{L}(\hilbert{H})\cap\mathcal{L}(\hilbert{H}_{0})$, and so do all the other terms, as it follows from lemma \ref{lemmav} and \ref{lemmaamdelta}. Concerning $U_{\mathrm{J}, \chi}^{(1)\,*}$, from equation \eqref{uonechi} it follows that \begin{equation}\label{uonechistar} U_{\mathrm{J}, \chi}^{(1)\,*}=[\mathbf{1}-\I\epsi\chi(\hepsisigma)\vsigma\pop\cdot\phisigmaj \vsigma^{*}-\I\epsi(\mathbf{1}-\chi(\hepsisigma))\vsigma\pop\cdot\phisigmaj \vsigma^{*}\chi(\hepsisigma)]\vsigma, \end{equation} so one can apply the same reasoning as above. \end{proof} \begin{theorem}\label{theoremuone} Assume that $\sigma=\sigma(\epsi)$ and that conditions \eqref{conditionsigma} hold, then the operator \begin{equation}\label{uone} \hilbert{U}:= U^{(1)}_{\mathrm{J}, \chi}[U^{(1)\,*}_{\mathrm{J}, \chi}U^{(1)}_{\mathrm{J}, \chi}]^{-1/2} \end{equation} is well-defined and unitary, for $\epsi$ small enough. Both $\hilbert{U}$ and $\hilbert{U}^{*}$ belong to $\mathcal{L}(\hilbert{H})\cap\mathcal{L}(\hilbert{H}_{0})$, with the property that \begin{equation}\label{unormbounded} \norm{\hilbert{U}}_{\mathcal{L}(\hilbert{H}_{0})}, \quad \norm{\hilbert{U}^{*}}_{\mathcal{L}(\hilbert{H}_{0})} \quad \leq C \quad , \end{equation} where $C$ is independent of $\epsi$ and $\sigma$. Moreover we can expand them in a power series which converges both in $\mathcal{L}(\hilbert{H})$ and in $\mathcal{L}(\hilbert{H}_{0})$. \end{theorem} \begin{proof} Combining equations \eqref{uonechi} and \eqref{uonechistar}, and using the fact $\pop\cdot\phisigmaj=\phisigmaj\cdot\pop -\I\epsi\nabla\cdot\phisigmaj$, we have that \begin{equation*} U_{\mathrm{J}, \chi}^{(1)\,*}U_{\mathrm{J}, \chi}^{(1)}=\mathbf{1}+\epsi^{2}B_{\epsi}, \end{equation*} where \begin{equation}\label{bepsi}\begin{split} & B_{\epsi}:= B_{0}+\epsi B_{1},\\ \\ & B_{0}:= -\chi(\hepsisigma)\vsigma\nabla\cdot\phisigmaj\vsigma^{*} -(\mathbf{1}-\chi(\hepsisigma))\vsigma\nabla\cdot\phisigmaj\vsigma^{*}\chi(\hepsisigma)+\\ & + [\chi(\hepsisigma)\vsigma(\phisigmaj\cdot\pop)\vsigma^{*}+(\mathbf{1}-\chi(\hepsisigma))\vsigma(\phisigmaj\cdot\pop)\vsigma^{*}\chi(\hepsisigma)]^{2},\\ \\ & B_{1}:= -\I[\chi(\hepsisigma)\vsigma\nabla\cdot\phisigmaj\vsigma^{*}+(\mathbf{1}-\chi(\hepsisigma))\vsigma\nabla\cdot\phisigmaj\vsigma^{*}\chi(\hepsisigma)]\cdot\\ &\cdot[\chi(\hepsisigma)\vsigma(\phisigmaj\cdot\pop)\vsigma^{*}+(\mathbf{1}-\chi(\hepsisigma))\vsigma(\phisigmaj\cdot\pop)\vsigma^{*}\chi(\hepsisigma)]\quad . \end{split} \end{equation} >From lemma \ref{lemmauonechi} it follows that $B_{\epsi}$ is a bounded operator both on $\hilbert{H}$ and $\hilbert{H}_{0}$, self-adjoint on $\hilbert{H}$. From equation \eqref{normujchi} we have also that \begin{equation*} \norm{B_{\epsi}}_{\mathcal{L}(\hilbert{K})}= \Or\big(\log(\sigma^{-1})\big), \qquad \hilbert{K}=\hilbert{H}, \hilbert{H}_{0}\quad, \end{equation*} so, under conditions \eqref{conditionsigma}, we can assume that $\norm{\epsi^{2}B_{\epsi}}_{\mathcal{L}(\hilbert{K})}<1$. $U_{\mathrm{J}, \chi}^{(1)\,*}U_{\mathrm{J}, \chi}^{(1)}$ is therefore strictly positive, and the square root in \eqref{uone} is well-defined. Moreover, we can express it through a convergent power series, both in $\mathcal{L}(\hilbert{H})$ and in $\mathcal{L}(\hilbert{H}_{0})$: \begin{equation}\label{expansionsquareroot} [\uonemchistar\uonemchi]^{-1/2} = (\mathbf{1}+\epsi^{2}B_{\epsi})^{-1/2}=\sum_{j=0}^{\infty}(-1)^{j}\frac{(2j-1)!!}{(2j)!!}\epsi^{2j}B_{\epsi}^{j}\;\footnote[1]{$(2j-1)!!=1\cdot 3\cdot 5\cdots (2j-1)$; $(2j)!!=2\cdot 4 \cdot 6\cdots 2j = 2^{j}j!$}\quad . \end{equation} We can also explicitly calculate that \begin{equation*}\begin{split} & \hilbert{U}\hilbert{U}^{*}= \uonemchi[\uonemchistar\uonemchi]^{-1}\uonemchistar=\mathbf{1};\\ & \hilbert{U}^{*}\hilbert{U}= [\uonemchistar\uonemchi]^{-1/2}\uonemchistar\uonemchi[\uonemchistar\uonemchi]^{-1/2}=\mathbf{1}, \end{split} \end{equation*} where we have used in the first equation the fact that both $\uonemchi$ and $\uonemchistar$ are invertible, since they differ by a term of order $\Or(\epsi\sqrt{\log(\sigma^{-1}}))$ from a unitary operator. Putting together \eqref{uone} and \eqref{expansionsquareroot} we get in the end the expansion for $\uone$. \end{proof} \section{The effective Hamiltonian}\label{effectivehamiltonian} Using the results of the last section, we can define the effective Hamiltonian just as the unitary transform of $\hepsisigma$ \begin{equation}\label{heff} \heff:=\uone\hepsisigma\hilbert{U}^{*}. \end{equation} It follows from theorem \ref{theoremuone} that $\uone$ is a bijection on $\hilbert{H}_{0}$, so $\heff$ is self-adjoint on it. Moreover, since $\uone$ can be expanded in a power series in $\mathcal{L}(\hilbert{H}_{0})$, we can expand $\heff$ in $\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})$. However, putting directly the expansion for $\uone$ in \eqref{heff} we get $\epsi-$dependent coefficients, because of the $\epsi-$dependence in $\hepsisigma$. To get the correct expansion we need then to rearrange some terms, but the remainder we get will in the end be bounded in $\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})$. \begin{theorem}\label{thmheff} The expansion up to the second order of the effective Hamiltonian is given by \begin{equation} \heff = h_{0}+\epsi h_{1, \chi} + \epsi^{2}h_{2, \chi} + \Or(\sigma)_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})} +\Or(\epsi^{3}\big(\log(\sigma^{-1})\big)^{3/2})_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}, \end{equation} where $h_{i, \chi}\in\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})$, \begin{equation} h_{0}= \frac{1}{2} \pop^{2}+\hfield + E(x), \end{equation} $h_{1,\chi}$ is given in equation \eqref{honechi} and $h_{2,\chi}$ is given in equation \eqref{htwochi}. \end{theorem} \begin{remark} The coefficients in the expansion of $\heff$ depend explicitly on the cutoff function $\chi$ (and also on the cutoff $\mathrm{J}$ in the number of particles, even though we have not stressed it in the notation). However, as we have already mentioned, we expect the adiabatic decoupling to be meaningful only on states in the range of $\chi(\hepsisigma)$ (or $\chi(\hepsi)$, according to lemma \eqref{cutoffenergysigma} we can interchange the two), which becomes in the representation space $\chi(\heff)$. With this in mind, we will prove later that the effective dynamics on the range of $\qm\chi(\heff)$ ($M<\mathrm{J}+1$) is generated by an Hamiltonian independent of $\chi$ and $\mathrm{J}$. \end{remark} \begin{proof} To simplify the following reasoning we write \begin{equation*}\begin{split} & \uonemchi=\vsigma^{*}(\mathbf{1}+\epsi T),\\ & T:= \I\chi(\hepsisigma)\vsigma\phisigmaj\cdot\pop \vsigma^{*} + \I(\mathbf{1}-\chi(\hepsisigma))\vsigma\phisigmaj\cdot\pop\vsigma^{*}\chi(\hepsisigma);\\ & \uonemchistar=(\mathbf{1}-\epsi T +\I\epsi^{2}S)\vsigma,\\ & S:= \I\chi(\hepsisigma)\vsigma\nabla\cdot\phisigmaj\vsigma^{*}+\I(\mathbf{1}-\chi(\hepsisigma))\vsigma\nabla\cdot\phisigmaj\vsigma^{*}\chi(\hepsisigma). \end{split} \end{equation*} We omit the dependence on $\sigma$ and $\epsi$ in $T$ and $S$ to streamline the presentation. We can then write \begin{equation*} B_{\epsi}=\I S -T^{2}+\I\epsi ST . \end{equation*} Expanding then the square root according to formula \eqref{expansionsquareroot} we get \begin{equation*}\begin{split} &\heff = \uone\hepsisigma\hilbert{U}^{*}=\vsigma^{*}(\mathbf{1}+\epsi T)\bigg[\mathbf{1}-\frac{\epsi^{2}}{2}(\I S -T^{2})\bigg]\hepsisigma\cdot\\ &\cdot\bigg[\mathbf{1}-\frac{\epsi^{2}}{2}(\I S -T^{2})\bigg](\mathbf{1}-\epsi T +\I\epsi^{2}S)\vsigma + \Or\big(\epsi^{3}(\log(\sigma^{-1}))^{3/2}\big)_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}=\\ &=\vsigma^{*}\hepsisigma \vsigma +\epsi \vsigma^{*}T\hepsisigma \vsigma -\epsi \vsigma^{*}\hepsisigma T\vsigma +\I\epsi^{2}\vsigma^{*}\hepsisigma S\vsigma + \\ &- \frac{\epsi^{2}}{2}\vsigma^{*}\hepsisigma(\I S -T^{2})\vsigma -\epsi^{2}\vsigma^{*}T\hepsisigma T\vsigma-\frac{\epsi^{2}}{2}\vsigma^{*}(\I S -T^{2})\hepsisigma \vsigma + \\ &+ \Or\big(\epsi^{3}(\log(\sigma^{-1}))^{3/2}\big)_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}= \vsigma^{*}\hepsisigma \vsigma + \epsi \vsigma^{*}[T, \hepsisigma]\vsigma + \\ &+ \frac{\I\epsi^{2}}{2}\vsigma^{*}[\hepsisigma, S]\vsigma +\frac{\epsi^{2}}{2}\vsigma^{*}\big[[\hepsisigma, T], T\big]\vsigma +\Or\big(\epsi^{3}(\log(\sigma^{-1}))^{3/2}\big)_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}\quad. \end{split} \end{equation*} We examine now separately the terms coming from different powers of $\epsi$. \begin{equation}\begin{split}\label{vhepsiv} &(0)\qquad \vsigma^{*}\hepsisigma \vsigma= \frac{1}{2}\vsigma(x)^{*}\pop^{2}\vsigma(x) + \hfield + E_{\sigma}(x) = \frac{1}{2} \pop^{2}+\hfield + E_{\sigma}(x) +\\ &+ \frac{1}{2}\vsigma(x)^{*}[\pop^{2}, \vsigma(x)]= \textrm{(using \eqref{commpv})}\ \frac{1}{2} \pop^{2}+\hfield + E_{\sigma}(x) +\\ &+\epsi\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop + \frac{\epsi^{2}}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\Phi(\I\nabla_{x}v_{\sigma}) -\I\frac{\epsi^{2}}{2}\Phi(\I\Delta v_{\sigma})=\\ &= \frac{1}{2} \pop^{2}+\hfield + E(x) +\frac{\epsi}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop + \frac{\epsi}{2}\pop\cdot \Phi(\I\nabla_{x}v_{\sigma}) +\\ &+ \frac{\epsi^{2}}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\Phi(\I\nabla_{x}v_{\sigma}) + \Or(\sigma)_{\mathcal{L}(\hilbert{H})}, \end{split} \end{equation} where we have used \eqref{estimateex}. This implies then that \begin{equation*} h_{0}=\frac{1}{2} \pop^{2}+\hfield + E(x)\quad. \end{equation*} Summing the term of order $\Or(\epsi)$ coming from $(0)$ to the ones coming from the expansion of $\uone$ we get \begin{equation*} (1)\qquad \frac{\epsi}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop + \frac{\epsi}{2}\pop\cdot \Phi(\I\nabla_{x}v_{\sigma})+ \epsi \vsigma^{*}[T, \hepsisigma]\vsigma\quad . \end{equation*} \begin{equation*}\begin{split} & [T, \hepsisigma]=\I\chi(\hepsisigma)[\vsigma\phisigmaj\cdot\pop \vsigma^{*}, \hepsisigma] +\I(\mathbf{1}-\chi(\hepsisigma))[\vsigma\phisigmaj\cdot\pop \vsigma^{*}, \hepsisigma]\cdot\\ &\cdot\chi(\hepsisigma)=\\ &=\frac{\I}{2}\chi(\hepsisigma)[\vsigma\phisigmaj\cdot\pop \vsigma^{*}, \pop^{2}] + \frac{\I}{2}(\mathbf{1}-\chi(\hepsisigma))[\vsigma\phisigmaj\cdot\pop \vsigma^{*} , \pop^{2}]\chi(\hepsisigma)+\\ &+\I\chi(\hepsisigma)\vsigma[\phisigmaj\cdot\pop , \hfield + E_{\sigma}(x)]\vsigma^{*} + \I(\mathbf{1}-\chi(\hepsisigma))\vsigma[\phisigmaj\cdot\pop , \hfield + E_{\sigma}(x)]\cdot\\ &\cdot\vsigma^{*}\chi(\hepsisigma)\,\,. \end{split} \end{equation*} The commutator gives \begin{equation*}\begin{split} & [\phisigmaj\cdot\pop , \hfield + E_{\sigma}(x)]=[\phisigmaj, \hfield]\cdot\pop + \phisigmaj\cdot[\pop, E_{\sigma}(x)]=\\ &=\I\qless{J}\Phi(\I\nabla_{x}v_{\sigma})\qless{J}\cdot\pop -\I\epsi\phisigmaj\cdot\nabla E_{\sigma}(x)\quad, \end{split} \end{equation*} where we have used equation \eqref{commhfa} to calculate the first commutator. Putting together all the terms we have therefore \begin{equation*}\begin{split} &(1) = \frac{\epsi}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop + \frac{\epsi}{2}\pop\cdot \Phi(\I\nabla_{x}v_{\sigma}) -\epsi \chi(\vsigma^{*}\hepsisigma \vsigma)[\phim(\I\nabla_{x}v_{\sigma})\cdot\pop +\\ &-\epsi\phisigmaj\cdot\nabla E_{\sigma}(x)] - \epsi(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma \vsigma))[\phim(\I\nabla_{x}v_{\sigma})\cdot\pop +\\ &-\epsi\phisigmaj\cdot\nabla E_{\sigma}(x)]\chi(\vsigma^{*}\hepsisigma\vsigma) + \Or(\epsi^{2}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}, \end{split} \end{equation*} where the terms containing the commutator with $\pop^{2}$ give rise to higher order contributions and we have shortened $\phim(f):=\qless{J}\Phi(f)\qless{J}$. Analyzing the terms of order $\Or(\epsi^{2})$ we will see that they yield, as it happens for the zero order one, terms of the form $\pop\cdot\Phi$, which make the previous expression a symmetric operator. We have therefore in the end that \begin{equation}\begin{split}\label{honechi} & h_{1, \chi}= \frac{1}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop -\frac{1}{2} \chi(\vsigma^{*}\hepsisigma \vsigma)\phim(\I\nabla_{x}v_{\sigma})\cdot\pop +\\ &-\frac{1}{2}(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma\vsigma))\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \chi(\vsigma^{*}\hepsisigma\vsigma) +\\ &+\textrm{``}\pop\cdot\Phi\textrm{''}, \end{split} \end{equation} where the symbol ``$\pop\cdot\Phi$'' means that, associated to each term of the form $\Phi\cdot\pop$, there is another one of the form $\pop\cdot\Phi$ which makes the sum a symmetric operator. To calculate $h_{2}$ we follow the same route, \begin{equation*}\begin{split} & (2)\qquad \frac{\epsi^{2}}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\Phi(\I\nabla_{x}v_{\sigma}) +\epsi^{2}\chi(\vsigma^{*}\hepsisigma\vsigma)\phisigmaj\cdot\nabla E_{\sigma}(x) +\\ &+\epsi^{2}(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma\vsigma))\phisigmaj\cdot\nabla E_{\sigma}(x)\chi(\vsigma^{*}\hepsisigma\vsigma) +\frac{\I\epsi}{2}\chi(\vsigma^{*}\hepsisigma\vsigma)\cdot\\ &\cdot[\phisigmaj\cdot\pop, \vsigma^{*}\pop^{2}\vsigma] +\frac{\I\epsi}{2}(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma\vsigma))[\phisigmaj\cdot\pop, \vsigma^{*}\pop^{2}\vsigma]\chi(\vsigma^{*}\hepsisigma\vsigma)+\\ &+\frac{\I\epsi^{2}}{2}\vsigma^{*}[\hepsisigma, S]\vsigma +\frac{\epsi^{2}}{2}\vsigma^{*}\big[[\hepsisigma, T], T\big]\vsigma \quad . \end{split} \end{equation*} We examine separately the different terms. \begin{equation*}\begin{split} & \bullet\qquad \frac{\epsi^{2}}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\Phi(\I\nabla_{x}v_{\sigma})=\frac{\epsi^{2}}{4}\big[a(\I\deriv{j}{}v_{\sigma}) + a(\I\deriv{j}{}v_{\sigma})^{*}\big]\big[a(\I\deriv{j}{}v_{\sigma}) +\\ &+a(\I\deriv{j}{}v_{\sigma})^{*}\big]=\\ &=\frac{\epsi^{2}}{4}\sum_{j=1}^{3N}\big[a(\I\deriv{j}{}v_{\sigma})^{2} + a(\I\deriv{j}{}v_{\sigma})^{*\,2} + a(\I\deriv{j}{}v_{\sigma})^{*}a(\I\deriv{j}{}v_{\sigma})\big] + \norm{\nabla v_{\sigma}}_{L^{2}(\field{R}{3}, dk)\otimes\field{C}{3N}}^{2}\,\, . \end{split} \end{equation*} We remark that the last term, being $x$-independent, is a $c$-number. \begin{equation}\label{commaphi}\begin{split} & \bullet\qquad [\phisigmaj\cdot\pop, \vsigma^{*}\pop^{2}\vsigma]= [\phisigmaj\cdot\pop, \pop^{2}] + [\phisigmaj\cdot\pop, \vsigma^{*}[\pop^{2}, \vsigma]]=\\ &=2\I\epsi\deriv{j}{}\phisigmajl\pop_{l}\pop_{j} +\epsi[\phisigmaj\cdot\pop, \Phi(\I\nabla_{x}v_{\sigma})\cdot\pop + \pop\cdot \Phi(\I\nabla_{x}v_{\sigma})]+\Or(\epsi^{2})_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}=\\ &=2\I\epsi\deriv{j}{}\phisigmajl\pop_{l}\pop_{j} +\epsi[\phisigmajl, \Phi(\I\deriv{j}{}v)]\pop_{l}\pop_{j} + \epsi\pop_{l}\pop_{j}[\phisigmajl, \Phi(\I\deriv{j}{}v_{\sigma})] +\Or(\epsi^{2})_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}=\\ &=2\I\epsi\deriv{j}{}\phizerojl\pop_{l}\pop_{j} +\epsi[\phisigmajl, \Phi(\I\deriv{j}{}v_{\sigma})]\pop_{l}\pop_{j} + \epsi\pop_{l}\pop_{j}[\phisigmajl, \Phi(\I\deriv{j}{}v_{\sigma})] +\Or(\epsi^{2})_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})} +\\ &+ \Or(\epsi\sigma)_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}, \end{split} \end{equation} where we have used lemma \ref{lemmaamdelta}. \begin{equation*}\begin{split} &\qquad\quad \bullet\ [\phisigmajl, \Phi(\I\deriv{j}{}v_{\sigma})]= [\qless{J}\phisigmal\qless{J}, \Phi(\I\deriv{j}{}v)] = [\qless{J}, \Phi(\I\deriv{j}{}v_{\sigma})]\phisigmaj\qless{J}+\\ &+\qless{J}\phisigmaj[\qless{J}, \Phi(\I\deriv{j}{}v_{\sigma})]+\qless{J}[\phisigmaj, \Phi(\I\deriv{j}{}v_{\sigma})]\qless{J}=\\ &=\frac{1}{\sqrt{2}}Q_{\mathrm{J}}a(\I\deriv{j}{}v_{\sigma})\phisigmaj\qless{J} - \frac{1}{\sqrt{2}}a(\I\deriv{j}{}v_{\sigma})^{*}Q_{\mathrm{J}}\phisigmaj\qless{J}+\\ &+\frac{1}{\sqrt{2}}\qless{J}\phisigmaj Q_{\mathrm{J}}a(\I\deriv{j}{}v_{\sigma})- \frac{1}{\sqrt{2}}\qless{J}\phisigmaj a(\I\deriv{j}{}v_{\sigma})^{*}Q_{\mathrm{J}}+\\ &+\I\qless{J}\Re\big\langle\deriv{j}{}v_{\sigma}(x, \cdot), \frac{\deriv{l}{}v_{\sigma}(x, \cdot)}{\abs{k}}\big\rangle_{L^{2}(\field{R}{3}, dk)}=\\ &=:\mathcal{\tilde{R}}_{\mathrm{J}} +\I\qless{J}\Re\big\langle\deriv{j}{}v_{\sigma}(x, \cdot), \frac{\deriv{l}{}v_{\sigma}(x, \cdot)}{\abs{k}}\big\rangle_{L^{2}(\field{R}{3}, dk)}=\\ &=\mathcal{\tilde{R}}_{\mathrm{J}} +\I\qless{J}\Re\big\langle\frac{\deriv{j}{}v(x, \cdot)}{\abs{k}^{1/2}}, \frac{\deriv{l}{}v(x, \cdot)}{\abs{k}^{1/2}}\big\rangle_{L^{2}(\field{R}{3}, dk)} + \Or(\sigma^{1/2})_{\mathcal{L}(\hilbert{H})}\quad , \end{split} \end{equation*} where we have bounded the scalar product using the same procedure applied in lemma \ref{lemmaamdelta}. The remainder term $\mathcal{\tilde{R}}_{\mathrm{J}}$ vanishes when applied to states in the range of $\qm$ with $M<\mathrm{J}-1$. The scalar product can be written in a clearer way using the explicit expression of the function $v$ ($l_{1}, j_{1}=1, \ldots N$, $l_{2}, j_{2}=1, \ldots 3$): \begin{equation*}\begin{split} &\qquad\quad \bullet\ \Re\big\langle\frac{\deriv{(l_{1}, l_{2})}{}v(x, \cdot)}{\abs{k}^{1/2}}, \frac{\deriv{(j_{1}, j_{2})}{}v(x, \cdot)}{\abs{k}^{1/2}}\big\rangle_{L^{2}(\field{R}{3}, dk)}=\\ &=\Re\int_{\field{R}{3}}dk\, \hat{\varrho}_{l_{1}}(k)^{*}\hat{\varrho}_{j_{1}}(k)\frac{\E^{\I k\cdot(\vec{x}_{j_{1}}-\vec{x}_{l_{1}})}}{\abs{k}^{2}}\kappa_{l_{2}}\kappa_{j_{2}}, \quad \kappa:=\frac{k}{\abs{k}}\quad. \end{split} \end{equation*} When $l_{1}=j_{1}$, since the charge densities are spherically symmetric, we get \begin{equation*}\begin{split} & \Re\int_{\field{R}{3}}dk\, \frac{\abs{\hat{\varrho}_{l_{1}}(k)}^{2}}{\abs{k}^{2}}\kappa_{l_{2}}\kappa_{j_{2}}=\frac{1}{3}\int_{\field{R}{3}}dk\, \frac{\abs{\hat{\varrho}_{l_{1}}(k)}^{2}}{\abs{k}^{2}}\delta_{l_{2}, j_{2}}=\\ &=\frac{1}{12\pi}\int_{\field{R}{3}\times\field{R}{3}}d\vec{x}\ d\vec{y}\ \frac{\varrho_{l_{1}}(\vec{x})\varrho_{l_{1}}(\vec{y})}{\abs{\vec{x}-\vec{y}}}\delta_{l_{2}, j_{2}}=:\frac{1}{3}e_{l_{1}}\delta_{l_{2}, j_{2}}\quad . \end{split} \end{equation*} Putting all the terms together, we have in the end \begin{equation*}\begin{split} &\qquad \frac{\epsi}{2}[\phisigmaj\cdot\pop, V^{*}\pop^{2}V]= \I\epsi^{2}\deriv{l}{}\phizerojl\pop_{j}\pop_{l} +\I\epsi^{2}\qless{J}\sum_{l_{1}=1}^{N}e_{l_{1}}\popb_{l_{1}}^{2} +\\ &+\I\frac{\epsi^{2}}{2}\qless{J}\underset{(l_{1}\neq j_{1})}{\sum_{l_{1}, j_{1}=1}^{N}}\int_{\field{R}{3}}dk\, \frac{\hat{\varrho}_{l_{1}}(k)^{*}\hat{\varrho}_{j_{1}}(k)}{\abs{k}^{2}}\big[\E^{\I k\cdot(\vec{x}_{j_{1}}-\vec{x}_{l_{1}})}(\kappa\cdot\popb_{l_{1}})(\kappa\cdot\popb_{j_{1}}) +\\ &+ (\kappa\cdot\popb_{l_{1}})(\kappa\cdot\popb_{j_{1}})\E^{\I k\cdot(\vec{x}_{j_{1}}-\vec{x}_{l_{1}})}\big] +\epsi^{2}\mathcal{R}_{\mathrm{M}} + \Or(\epsi^{3})_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})} +\\ &+\Or(\epsi^{2}\sigma)_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}, \end{split} \end{equation*} where $\mathcal{R}_{\mathrm{J}}$ is a remainder term that vanishes when applied to states in the range of $\qm$, with $M<\mathrm{J}-1$, and $\popb_{l_{1}}$ denotes the three-dimensional momentum operator associated to each particle. \begin{equation*}\begin{split} & \bullet\qquad \frac{\I\epsi^{2}}{2}\vsigma^{*}[\hepsisigma, S]\vsigma= -\frac{\epsi^{2}}{2}\vsigma^{*}\chi(\hepsisigma)[\hepsisigma, \vsigma\nabla\cdot\phisigmaj \vsigma^{*}]\vsigma +\\ &- \frac{\epsi^{2}}{2}\vsigma^{*}(\mathbf{1}-\chi(\hepsisigma))[\hepsisigma, \vsigma\nabla\cdot\phisigmaj \vsigma^{*}]\chi(\hepsisigma)\vsigma=\\ &=-\frac{\epsi^{2}}{2}\chi(\vsigma^{*}\hepsisigma\vsigma)[\vsigma^{*}\hepsisigma\vsigma, \nabla\cdot\phisigmaj] - \frac{\epsi^{2}}{2}(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma\vsigma))\cdot\\ &\cdot[\vsigma^{*}\hepsisigma\vsigma, \nabla\cdot\phisigmaj]\chi(\vsigma^{*}\hepsisigma\vsigma)=\\ &=-\frac{\epsi^{2}}{2}\chi(\vsigma^{*}\hepsisigma\vsigma)[\hfield, \nabla\cdot\phisigmaj] - \frac{\epsi^{2}}{2}(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma\vsigma))\cdot\\ &\cdot[\hfield, \nabla\cdot\phisigmaj]\chi(\vsigma^{*}\hepsisigma\vsigma) +\Or(\epsi^{3})_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}=\\ &=\frac{\I\epsi^{2}}{2}\chi(\vsigma^{*}\hepsisigma\vsigma)\Phi^{\mathrm{J}}(\I\Delta v_{\sigma}) + \frac{\I\epsi^{2}}{2}(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma\vsigma))\Phi^{\mathrm{J}}(\I\Delta v_{\sigma})\chi(\vsigma^{*}\hepsisigma\vsigma) +\\ &+\Or(\epsi^{3})_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})} . \end{split} \end{equation*} This gives exactly the terms needed to make $h_{1}$ a symmetric operator. \begin{equation*} \bullet\qquad \frac{\epsi^{2}}{2}\vsigma^{*}\big[[\hepsisigma, T], T\big]\vsigma\quad . \end{equation*} Using the calculations for the first order part $h_{1}$, we get that \begin{equation*}\begin{split} & [\hepsisigma, T] = \chi(\hepsisigma)\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*} +\\ & +(\mathbf{1}-\chi(\hepsisigma))\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*}\chi(\hepsisigma) + \Or(\epsi\sqrt{\log(\sigma^{-1})}))_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}, \end{split} \end{equation*} we can keep therefore just the first two terms, omitting the remainder. \begin{equation*}\begin{split} & \bullet\qquad \frac{\epsi^{2}}{2}\vsigma^{*}\big[[\hepsisigma, T], T\big]\vsigma=\frac{\epsi^{2}}{2}\vsigma^{*}\big[\chi(\hepsi)\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*} +\\ &+(\mathbf{1}-\chi(\hepsisigma))\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*}\chi(\hepsisigma), T\big]\vsigma + \Or(\epsi^{3}\sqrt{\log(\sigma^{-1})})=\\ &=\frac{\I\epsi^{2}}{2}\vsigma^{*}\big[\chi(\hepsisigma)\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*},\ \chi(\hepsisigma)\vsigma\phisigmaj\cdot\pop \vsigma^{*}\big]\vsigma +\\ &+\frac{\I\epsi^{2}}{2}\vsigma^{*}\big[(\mathbf{1}-\chi(\hepsisigma))\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*}\chi(\hepsisigma),\ \chi(\hepsisigma)\vsigma\phisigmaj\cdot\pop \vsigma^{*}\big]\vsigma+\\ &+\frac{\I\epsi^{2}}{2}\vsigma^{*}\big[\chi(\hepsisigma)\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*},\ (\mathbf{1}-\chi(\hepsisigma))\vsigma\phisigmaj\cdot\pop \vsigma^{*}\chi(\hepsisigma)\big]\vsigma+\\ &+\frac{\I\epsi^{2}}{2}\vsigma^{*}\big[(\mathbf{1}-\chi(\hepsisigma))\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*}\chi(\hepsisigma),\\ &(\mathbf{1}-\chi(\hepsisigma))\vsigma\phisigmaj\cdot\pop \vsigma^{*}\chi(\hepsisigma)\big]\vsigma +\Or(\epsi^{3}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}\quad . \end{split} \end{equation*} Summing up, we get in the end that \begin{equation}\begin{split}\label{htwochi} & h_{2,\chi}= \frac{1}{4}\sum_{j=1}^{3n}\big[a(\I\deriv{j}{}v_{\sigma})^{2} + a(\I\deriv{j}{}v_{\sigma})^{*\,2} + a(\I\deriv{j}{}v_{\sigma})^{*}a(\I\deriv{j}{}v_{\sigma})\big] +\\ &+ \norm{\nabla v_{\sigma}}_{L^{2}(\field{R}{3}, dk)\otimes\field{C}{3n}}^{2} +\chi(\vsigma^{*}\hepsisigma\vsigma)\bigg\{\phisigmaj\cdot\nabla E(x)+\\ &- \frac{1}{2}\big(\deriv{j}{}\Phi_{0, l}\pop_{l}\pop_{j} + \pop_{l}\pop_{j}\deriv{j}{}\Phi_{0, l}\big) - \qless{J}\sum_{l_{1}=1}^{N}e_{l_{1}}\popb_{l_{1}}^{2}+\\ &-\frac{1}{2}\qless{J}\underset{(l_{1}\neq j_{1})}{\sum_{l_{1}, j_{1}=1}^{N}}\int_{\field{R}{3}}dk\, \frac{\hat{\varrho}_{l_{1}}(k)^{*}\hat{\varrho}_{j_{1}}(k)}{\abs{k}^{2}}\big[\E^{\I k\cdot(\vec{x}_{j_{1}}-\vec{x}_{l_{1}})}(\kappa\cdot\popb_{l_{1}})(\kappa\cdot\popb_{j_{1}}) +\\ &+ (\kappa\cdot\popb_{l_{1}})(\kappa\cdot\popb_{j_{1}})\E^{\I k\cdot(\vec{x}_{j_{1}}-\vec{x}_{l_{1}})}\big] +\mathcal{R}_{\mathrm{J}}\bigg\} + (\mathbf{1}-\chi(\vsigma^{*}\hepsisigma\vsigma))\bigg\{\cdots\bigg\}\cdot\\ &\cdot\chi(\vsigma^{*}\hepsisigma \vsigma)+\\ &+\frac{\I}{2}\vsigma^{*}\big[\chi(\hepsisigma)\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*},\ \chi(\hepsisigma)\vsigma\phisigmaj\cdot\pop \vsigma^{*}\big]\vsigma +\\ &+\frac{\I}{2}\vsigma^{*}\big[(\mathbf{1}-\chi(\hepsisigma))\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*}\chi(\hepsisigma),\ \chi(\hepsisigma)\vsigma\phisigmaj\cdot\pop \vsigma^{*}\big]\vsigma+\\ &+\frac{1}{2}\vsigma^{*}\big[\chi(\hepsisigma)\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*},\ (\mathbf{1}-\chi(\hepsisigma))\vsigma\phisigmaj\cdot\pop \vsigma^{*}\chi(\hepsisigma)\big]\vsigma+\\ &+\frac{1}{2}\vsigma^{*}\big[(\mathbf{1}-\chi(\hepsisigma))\vsigma\phim(\I\nabla_{x}v_{\sigma})\cdot\pop \vsigma^{*}\chi(\hepsisigma),\\ &(\mathbf{1}-\chi(\hepsisigma))\vsigma\phisigmaj\cdot\pop \vsigma^{*}\chi(\hepsisigma)\big]\vsigma \qquad , \end{split} \end{equation} where we have symmetrized $\deriv{j}{}\Phi_{0, l}\pop_{l}\pop_{j}$, which is possible up to terms of order $\Or(\epsi)$. The expression is fairly lengthy, but we will show below that many terms vanish when applied to a state in the range of $\qm\chi(\heff)$ ($M<\mathrm{J}-1$). \end{proof} \section{The effective dynamics}\label{effectivedynamics} We start with a number of lemmas we need to analyze the effective time evolution. \begin{lemma}\label{approxchi} Assume that $\sigma$ satisfies conditions \eqref{conditionsigma}, then $1.$\ Given a function $\tilde{\chi}\in\coinf(\field{R}{})$, we have \begin{equation} \tilde{\chi}(\vsigma^{*}\hepsisigma \vsigma)-\tilde{\chi}(h_{0})=\epsi\mathcal{R}^{\epsi}_{\chi}, \end{equation} where $\mathcal{R}^{\epsi}_{\chi}\in\mathcal{L}(\hilbert{H}, \hilbert{H}_{0})$, $\norm{\mathcal{R}^{\epsi}_{\chi}}_{\mathcal{L}(\hilbert{H}, \hilbert{H}_{0})}=\Or(1)$ and \begin{equation} \mathcal{R}^{\epsi}_{\chi}\qm=(\mathrm{Q}_{M+1}+\mathrm{Q}_{M-1})\mathcal{R}^{\epsi}_{\chi}\qm + \Or(\epsi^{2})_{\mathcal{L}(\hilbert{H}, \hilbert{H}_{0})}. \end{equation} $2.$\ Moreover, we have that \begin{equation}\label{diffheffhzero} \tilde{\chi}(\heff)-\tilde{\chi}(h_{0})=\Or(\epsi)_{\mathcal{L}(\hilbert{H}, \hilbert{H}_{0})}, \end{equation} and that \begin{equation}\label{diffqnheffhzero} \qm\tilde{\chi}(\heff)=\qm\tilde{\chi}(h_{0})\tilde{\tilde{\chi}}(\heff) +\Or(\epsi^{2}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H}, \hilbert{H}_{0})}, \end{equation} where $\tilde{\tilde{\chi}}$ is any $\coinf(\field{R}{})$ function such that $\tilde{\chi}\tilde{\tilde{\chi}}=\tilde{\chi}$ and $\tilde{\tilde{\chi}}\chi=\tilde{\tilde{\chi}}$, $M<\mathrm{J}-1$. \end{lemma} \begin{proof} Applying the Hellfer-Sj\"ostrand formula, equation \eqref{hellfersjostrand}, we get \begin{equation*} \tilde{\chi}(\vsigma^{*}\hepsisigma\vsigma)-\tilde{\chi}(h_{0})= \frac{1}{\pi}\int_{\field{R}{2}}dx dy\ \bar{\partial}\tilde{\chi}^{a}(z)\big[(\vsigma^{*}\hepsisigma\vsigma-z)^{-1} - (h_{0}-z)^{-1}\big]. \end{equation*} Since both Hamiltonians are self-adjoint on $\hilbert{H}_{0}$, we get \begin{equation*}\begin{split} & (\vsigma^{*}\hepsisigma\vsigma-z)^{-1} - (h_{0}-z)^{-1}=(h_{0}-z)^{-1}(h_{0}-\vsigma^{*}\hepsisigma \vsigma)\cdot\\ &\cdot(\vsigma^{*}\hepsisigma \vsigma-z)^{-1}=-(h_{0}-z)^{-1}\bigg[\frac{\epsi}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop + \frac{\epsi}{2}\pop\cdot \Phi(\I\nabla_{x}v_{\sigma})+\\ &+\frac{\epsi^{2}}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\Phi(\I\nabla_{x}v_{\sigma})\bigg](\vsigma^{*}\hepsisigma \vsigma-z)^{-1} +\Or(\sigma\abs{\Im z}^{-2})_{\mathcal{L}(\hilbert{H}, \hilbert{H}_{0})}\quad , \end{split} \end{equation*} where we have used equation \eqref{vhepsiv} to calculate the difference of the two Hamiltonians and estimated the integrand proceeding in the same way as in lemma~\ref{cutoffenergysigma}. Moreover, iterating the formula we get \begin{equation}\label{expansionresolvent}\begin{split} & (\vsigma^{*}\hepsisigma\vsigma-z)^{-1} - (h_{0}-z)^{-1}=(h_{0}-z)^{-1}(h_{0}-\vsigma^{*}\hepsisigma\vsigma)\cdot\\ &\cdot(\vsigma^{*}\hepsisigma\vsigma-z)^{-1}=-(h_{0}-z)^{-1}\bigg[\frac{\epsi}{2}\Phi(\I\nabla_{x}v)\cdot\pop +\\ &+\frac{\epsi}{2}\pop\cdot \Phi(\I\nabla_{x}v)\bigg](h_{0}-z)^{-1} + \Or(\epsi^{2}\abs{\Im z}^{-3})_{\mathcal{L}(\hilbert{H}, \hilbert{H}_{0})} , \end{split} \end{equation} so \begin{equation*}\begin{split} & \big[(\vsigma^{*}\hepsisigma \vsigma-z)^{-1} - (h_{0}-z)^{-1}\big]\qm = (\mathrm{Q}_{M +1}+\mathrm{Q}_{M-1})\big[(\vsigma^{*}\hepsisigma \vsigma-z)^{-1} +\\ &- (h_{0}-z)^{-1}\big]\cdot\qm +\Or(\epsi^{2}\abs{\Im z}^{-3})_{\mathcal{L}(\hilbert{H}, \hilbert{H}_{0})}. \end{split} \end{equation*} Concerning point $2$, equation \eqref{diffheffhzero} follows immediately from the fact that \begin{equation*}\begin{split} &(\heff-z)^{-1} - (h_{0}-z)^{-1}=(h_{0}-z)^{-1}(h_{0}-\heff)(\heff-z)^{-1}=\\ &=\epsi(h_{0}-z)^{-1}h_{1,\chi}(\heff-z)^{-1} +\Or(\epsi^{2}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})}, \end{split} \end{equation*} while, for equation \eqref{diffqnheffhzero}, we have by definition that \begin{equation*}\begin{split} &\qm\tilde{\chi}(\heff)=\qm\tilde{\chi}(\heff)\tilde{\tilde{\chi}}(\heff)=\qm\tilde{\chi}(h_{0})\tilde{\tilde{\chi}}(\heff) +\\ &+ \qm[\tilde{\chi}(\heff)-\tilde{\chi}(h_{0})]\tilde{\tilde{\chi}}(\heff)\quad . \end{split} \end{equation*} Proceeding as above we find \begin{equation}\begin{split}\label{estimatechi} & \qm[\tilde{\chi}(\heff)-\tilde{\chi}(h_{0})]\tilde{\tilde{\chi}}(\heff) = \frac{1}{\pi}\int_{\field{R}{2}}dx dy\ \bar{\partial}\tilde{\chi}^{a}(z)\qm\big[(\heff-z)^{-1} +\\ &- (h_{0}-z)^{-1}\big]\tilde{\tilde{\chi}}(\heff), \end{split} \end{equation} so, if we show that \begin{equation}\label{qnhonechichi} \qm h_{1,\chi}\tilde{\tilde{\chi}}(\heff)=\Or(\epsi\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})} \end{equation} we are done. >From equation \eqref{honechi} (omitting the ``$\pop\cdot\Phi$'' part, which can be treated in the same way) we get \begin{equation*}\begin{split} &\qm h_{1, \chi} \tilde{\tilde{\chi}}(\heff) = \frac{1}{2}\qm\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop \tilde{\tilde{\chi}}(\heff) -\frac{1}{2} \qm\chi(\vsigma^{*}\hepsisigma\vsigma)\cdot\\ &\cdot\phim(\I\nabla_{x}v_{\sigma})\cdot\pop\tilde{\tilde{\chi}}(\heff) -\frac{1}{2}\qm(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma\vsigma))\phim(\I\nabla_{x}v_{\sigma})\cdot\pop\chi(\vsigma^{*}\hepsisigma\vsigma)\cdot\\ &\cdot\tilde{\tilde{\chi}}(\heff), \end{split} \end{equation*} but it follows from point $1$, that we can replace $\chi(V^{*}\hepsi V)$ with $\chi(h_{0})$ up to terms of order $\Or(\epsi)$, and from equation \eqref{estimatechi} that we can replace $\tilde{\tilde{\chi}}(\heff)$ with $\tilde{\tilde{\chi}}(h_{0})$ up to terms of order $\Or(\epsi)$, therefore we have \begin{equation*}\begin{split} & \qm h_{1, \chi} \tilde{\tilde{\chi}}(\heff)=\frac{1}{2}\qm\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop \tilde{\tilde{\chi}}(h_{0}) -\frac{1}{2} \qm\chi(\vsigma^{*}\hepsisigma\vsigma)\phim(\I\nabla_{x}v_{\sigma})\cdot\\ &\cdot\pop\tilde{\tilde{\chi}}(h_{0}) -\frac{1}{2}\qm(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma \vsigma))\phim(\I\nabla_{x}v_{\sigma})\cdot\pop\tilde{\tilde{\chi}}(h_{0}) +\Or(\epsi)=\\ &=\frac{1}{2}\qm\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop \tilde{\tilde{\chi}}(h_{0}) -\frac{1}{2} \qm\phim(\I\nabla_{x}v_{\sigma})\cdot\pop\tilde{\tilde{\chi}}(h_{0})+\Or(\epsi)=\\ &=\Or(\epsi), \end{split} \end{equation*} so point $2$ is proved. \end{proof} The following lemma was proved in (\cite{FGS}, Appendix B) and we will use it to characterize the range of $\chi_{(-\infty, c)}(h_{0})$, where $\chi_{(-\infty, c)}$ is the characteristic function of the indicated interval. \begin{lemma}\label{lemmafgs} Let $\tilde{H}$ be an Hamiltonian of the form \begin{equation} \tilde{H}:= \mathbf{1}\otimes d\Gamma(\abs{k}) + H\otimes\mathbf{1}, \end{equation} acting on the Hilbert space $\tilde{\hilbert{H}}=\hilbert{H}\otimes\fock$, where $\fock$ is the bosonic Fock space over $L^{2}(\field{R}{d})$ and $\hilbert{H}$ is a generic Hilbert space. Then the set of all linear combinations of vectors of the form \begin{equation}\label{rangechi} \varphi\otimes a(g_{1})^{*}\cdots a(g_{\N})^{*}\Omega_{\mathrm{F}}, \quad \lambda + \sum_{j=1}^{\N}M_{j}< c, \quad (c>0), \end{equation} where $\varphi=\chi_{(-\infty, \lambda)}(H)\varphi$ for some $\lambda0$, we have \begin{equation}\label{energynotincreas} a\bigg(\frac{\I\deriv{j}{}v_{\sigma}(x, \cdot)}{\abs{k}}\bigg)\qm\chi(h_{0})=\mathrm{Q}_{M-1}\xi(h_{0})a\bigg(\frac{\I\deriv{j}{}v_{\sigma}(x, \cdot)}{\abs{k}}\bigg)\qm\chi(h_{0}) + \Or_{0}(\epsi^{\infty}), \end{equation} where $\xi\in\coinf(\field{R}{})$ and \begin{equation*} c_{\xi}=2d_{\chi} + E_{\infty}, \end{equation*} where \begin{equation*}\begin{split} & d_{\chi}:= 2c_{\chi}+E_{\infty}+\min\{c_{\chi}, \Lambda\},\\ & c_{\chi}:=\sup\{\abs{k}: k\in\mathrm{supp}\ \chi\},\\ & E_{\infty}:= \sup_{x\in\field{R}{n}}\abs{E(x)}, \end{split} \end{equation*} and we can choose $\sup\{\abs{k}: k\in\mathrm{supp}\ \xi\}$ arbitrarily close to $c_{\xi}$. An analogous statement holds for the creation operator. \end{corollary} \begin{proof} Applying twice lemma \ref{lemmapseudo}, the thesis will follow if we prove \eqref{energynotincreas} replacing $h_{0}$ with $\tilde{h}_{0}$. Applying lemma \ref{lemmafgs}, we have \begin{equation*} \chi_{(-\infty, c_{\chi})}(\tilde{h}_{0})\qm\chi(\tilde{h}_{0})\Psi=\qm\chi(\tilde{h}_{0})\Psi=\lim_{n\to\infty}\Psi_{n}, \end{equation*} where $\Psi_{n}$ is a linear combination of vectors of the form described in \eqref{rangechi}, with $\varphi=\chi_{(-\infty, \lambda)}(\pop^{2})\varphi$. We have therefore \begin{equation*} a\bigg(\frac{\I\deriv{j}{}v_{\sigma}(x, \cdot)}{\abs{k}}\bigg)\qm\chi(h_{0})\Psi=\lim_{n\to\infty}a\bigg(\frac{\I\deriv{j}{}v_{\sigma}(x, \cdot)}{\abs{k}}\bigg)\Psi_{n}. \end{equation*} Applying the operator to a vector of the form \eqref{rangechi} we get \begin{equation*}\begin{split} & a\bigg(\frac{\I\deriv{j}{}v_{\sigma}(x, \cdot)}{\abs{k}}\bigg)\varphi\otimes a(g_{1})^{*}\cdots a(g_{M})^{*}\Omega_{\mathrm{F}}=\varphi(x)\sum_{l=1}^{\N}\big\langle \frac{\I\deriv{j}{}v_{\sigma}(x,k)}{\abs{k}}, g_{l}(k)\big\rangle_{L^{2}(\field{R}{3}, dk)}\cdot\\ &\cdot a(g_{1})^{*}\cdots a(\overset{\vee}{g_{l}})^{*}\cdots a(g_{M})^{*}\Omega_{\mathrm{F}},\\ &\big\langle \frac{\I\deriv{j}{}v_{\sigma}(x,k)}{\abs{k}}, g_{l}(k)\big\rangle_{L^{2}(\field{R}{3}, dk)}= -\int dk\, \frac{\id_{(\sigma, \infty)}(k)\hat{\varrho}_{j_{1}}(k)^{*}}{\abs{k}^{3/2}}\E^{-\I k\cdot x_{j_{1}}}g_{l}(k)\kappa_{j_{2}}. \end{split} \end{equation*} The functions $g_{l}$ have by hypothesis compact support in $k$, with radius uniformly bounded by $c_{\chi}$. For the part depending on $x$, calculating the Fourier transform with the convolution theorem, we have \begin{equation*}\begin{split} & \mathcal{F}\bigg(\varphi(x)\big\langle \frac{\I\deriv{j}{}v_{\sigma}(x,k)}{\abs{k}}, g_{l}(k)\big\rangle\bigg)(p)=-\int dk\, \frac{\id_{(\sigma, \infty)}(k)\hat{\varrho}_{j_{1}}(k)^{*}}{\abs{k}^{3/2}}g_{l}(k)\kappa_{j_{2}}\hat{\varphi}(p + \tilde{k}_{j_{1}}), \end{split} \end{equation*} where \begin{equation*} \tilde{k}_{j_{1}} = (0, \ldots, k, 0, \ldots 0)\in\field{R}{3N}, \end{equation*} with the $k$ in the entry $j_{1}$. Now, since $\hat{\varrho}_{j_{1}}$, $g_{l}$ and $\hat{\varphi}$ have compact support, also the Fourier transform will have compact support in $p$, with radius bounded by $c_{\chi}+\min\{c_{\chi}, K\}=:d_{\xi}$. This means that \begin{equation*}\begin{split} & a\bigg(\frac{\I\deriv{j}{}v_{\sigma}(x, \cdot)}{\abs{k}}\bigg)\varphi\otimes a(g_{1})^{*}\cdots a(g_{M})^{*}\Omega_{\mathrm{F}}=\\ &=\chi_{(-\infty, d_{\xi})}(\tilde{h}_{0})a\bigg(\frac{\I\deriv{j}{}v_{\sigma}(x, \cdot)}{\abs{k}}\bigg)\varphi\otimes a(g_{1})^{*}\cdots a(g_{M})^{*}\Omega_{\mathrm{F}}, \end{split} \end{equation*} so, remarking that $\tilde{h}_{0}$ is a positive operator and that we can smooth $\chi_{[0, d_{\xi})}$ by a $\coinf$ function with arbitrarily close support, we prove the thesis. \end{proof} \begin{theorem}\label{zeroordertime}\emph{(Zero order approximation to the time evolution)} The following two estimates hold: \begin{equation} \norm{(\E^{-\I t\heff/\epsi} - \E^{-\I th_{0}/\epsi})\qm\tilde{\chi}(\heff)}_{\mathcal{L}(\hilbert{H})}=\Or\big(\sqrt{M+1}\abs{t}\epsi\sqrt{\log(\sigma(\epsi)^{-1})}\big), \end{equation} \begin{equation} \norm{\qm(\E^{-\I t\heff/\epsi} - \E^{-\I th_{0}/\epsi})\tilde{\chi}(\heff)}_{\mathcal{L}(\hilbert{H})}=\Or\big(\sqrt{M+1}\abs{t}\epsi\sqrt{\log(\sigma(\epsi)^{-1})}\big), \end{equation} for every $\tilde{\chi}\in\coinf(\field{R}{})$ such that $\tilde{\chi}\chi=\tilde{\chi}$. \end{theorem} \begin{corollary}\label{corzeroordertime} The subspaces associated to the $Q_{M}$s are almost invariant with respect to the dynamics generated by $\heff$ $\forall M\in\field{N}{}$, i.e., \begin{equation} \norm{[\E^{-\I t\heff/\epsi}, \qm]\tilde{\chi}(\heff)}_{\mathcal{L}(\hilbert{H})}=\Or\big(\sqrt{M+1}\abs{t}\epsi\sqrt{\log(\sigma(\epsi)^{-1})}\big). \end{equation} \end{corollary} \begin{remark}\label{adiabaticinvm}\emph{(Adiabatic invariance of $M$-photons dressed particles subspaces).} Using the unitary $\uone$ we can translate all the previous results from the representation space to the original dynamics. This means that if we define the perturbed dressed projectors \begin{equation} P_{M}^{\epsi}:=\hilbert{U}^{*}\qm\uone, \end{equation} which satisfy by construction \begin{equation*} \norm{P_{M}^{\epsi} - \pizn}_{\mathcal{L}(\hilbert{H})}=\Or(\epsi\sqrt{\log(\sigma(\epsi)^{-1})}), \end{equation*} we get that \begin{equation*} \norm{[\E^{-\I t\hepsisigma/\epsi}, P_{M}^{\epsi}]\tilde{\chi}(\hepsisigma)}_{\mathcal{L}(\hilbert{H})}=\Or\big(\sqrt{M+1}\abs{t}\epsi\sqrt{\log(\sigma(\epsi))^{-1}}\big), \end{equation*} and \begin{equation*} \norm{[\E^{-\I t\hepsisigma/\epsi}, \pizn]\tilde{\chi}(\hepsisigma)}_{\mathcal{L}(\hilbert{H})}=\Or\big((1+\abs{t})\epsi\sqrt{\log(\sigma(\epsi)^{-1})}\big), \quad \forall\tilde{\chi}\in\coinf(\field{R}{}) \quad . \end{equation*} For the original dynamics we have then \begin{equation*}\begin{split} & [\E^{-\I t\hepsi/\epsi}, P_{M}^{\epsi}]\tilde{\chi}(\hepsi) = [(\E^{-\I t\hepsi/\epsi}- \E^{-\I t\hepsisigma/\epsi}), P_{M}^{\epsi}]\chi(\hepsi) +\\ &+ [\E^{-\I t\hepsisigma/\epsi}, P_{M}^{\epsi}](\chi(\hepsi) - \chi(\hepsisigma)) + [\E^{-\I t\hepsisigma/\epsi}, P_{M}^{\epsi}]\tilde{\chi}(\hepsisigma)\quad . \end{split} \end{equation*} The first term can be bounded by $\Or(\sigma(\epsi)^{1/2}\epsi^{-1})=\Or(\epsi^{2})$ using proposition \ref{approxsigma} and equation \eqref{unormbounded}. The second one by $\Or(\sigma(\epsi)^{1/2})$ using lemma \ref{cutoffenergysigma}, and the third one has been just estimated above. Putting together these facts, equation \eqref{almostinvsub} is proved. For the particular case $M=0$, this result, together with the expression of the zero order Hamiltonian $h_{0}$, was already shown in \cite{Te2}, assuming a relativistic dispersion relation for the particles, which automatically implies that they have a bounded maximal velocity, and avoids therefore the introduction of cutoff functions. \end{remark} \begin{proof} First of all we remark that, employing lemma \ref{approxchi}, we can replace $\tilde{\chi}(\heff)$ with $\tilde{\chi}(h_{0})$ up to terms of order $\Or(\epsi)_{\mathcal{L}(\hilbert{H})}$, which are smaller than the error we want to prove. Since $\heff$ and $h_{0}$ are both self-adjoint on $\hilbert{H}_{0}$, we can apply Duhamel formula, and we get then \begin{equation*}\begin{split} &(\E^{-\I t\heff/\epsi}-\E^{-\I th_{0}/\epsi})\qm\tilde{\chi}(h_{0}) =-\frac{\I}{\epsi}\int_{0}^{t}ds\, \E^{\I (s-t)\heff/\epsi}(\heff-h_{0})\E^{-\I s h_{0}/\epsi}\cdot\\ &\cdot\qm\tilde{\chi}(h_{0})=-\I\int_{0}^{t}ds\, \E^{\I (s-t)\heff/\epsi}h_{1,\chi}\E^{-\I s h_{0}/\epsi}\qm\tilde{\chi}(h_{0}) +\\ &+ \Or(\epsi\sqrt{M+1}\sqrt{\log(\sigma^{-1})})=\\ &=-\I\int_{0}^{t}ds\, \E^{\I (s-t)\heff/\epsi}h_{1,\chi}\qm\tilde{\chi}(h_{0})\E^{-\I s h_{0}/\epsi} + \Or(\epsi\sqrt{M+1}\sqrt{\log(\sigma^{-1})}),\\ &\Rightarrow \norm{(\E^{-\I t\heff/\epsi}-\E^{-\I th_{0}/\epsi})\qm\tilde{\chi}(h_{0})}_{\mathcal{L}(\hilbert{H})}\leq\abs{t}\cdot\norm{h_{1,\chi}\qm\tilde{\chi}(h_{0})}_{\mathcal{L}(\hilbert{H})}+\\ &+\Or(\epsi\abs{t}\sqrt{M+1}\sqrt{\log(\sigma^{-1})}). \end{split} \end{equation*} We proceed now as in the proof of lemma \ref{approxchi}. From equation \eqref{honechi} it follows (omitting the terms of the form ``$\pop\cdot\Phi$'', which can be treated in the same way) \begin{equation*}\begin{split} & h_{1,\chi}\qm\tilde{\chi}(h_{0})= \bigg[\frac{1}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop -\frac{1}{2} \chi(\vsigma^{*}\hepsisigma\vsigma)\phim(\I\nabla_{x}v_{\sigma})\cdot\pop +\\ &-\frac{1}{2}\big(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma\vsigma)\big)\phim(\I\nabla_{x}v_{\sigma})\cdot\pop\,\chi(\vsigma^{*}\hepsisigma\vsigma)\bigg]\qm\tilde{\chi}(h_{0})=\\ &=\frac{1}{2}\Phi(\I\nabla_{x}v_{\sigma})\cdot\pop\,\qm\tilde{\chi}(h_{0})- \frac{1}{2} \chi(\vsigma^{*}\hepsisigma \vsigma)\phim(\I\nabla_{x}v_{\sigma})\cdot\pop\,\qm\tilde{\chi}(h_{0})+\\ &-\frac{1}{2}\big(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma\vsigma)\big)\phim(\I\nabla_{x}v_{\sigma})\cdot\pop\,\qm\tilde{\chi}(h_{0})+\\ &-\frac{1}{2}\big(\mathbf{1}-\chi(\vsigma^{*}\hepsisigma\vsigma)\big)\phim(\I\nabla_{x}v_{\sigma})\cdot\pop[\chi(\vsigma^{*}\hepsisigma\vsigma)-\chi(h_{0})]\cdot\\ &\cdot\qm\tilde{\chi}(h_{0})=\\ &=-\frac{1}{2}[\mathbf{1}-\chi(h_{0})]\phim(\I\nabla_{x}v_{\sigma})\cdot\pop[\chi(\vsigma^{*}\hepsisigma \vsigma)-\chi(h_{0})]\qm\tilde{\chi}(h_{0})+\\ &+\frac{1}{2}[\chi(\vsigma^{*}\hepsisigma\vsigma)-\chi(h_{0})]\phim(\I\nabla_{x}v_{\sigma})\cdot\pop[\chi(\vsigma^{*}\hepsisigma\vsigma)-\chi(h_{0})]\qm\tilde{\chi}(h_{0})\quad . \end{split} \end{equation*} Applying lemma \ref{approxchi}, we have immediately that the last term is of order $\Or(\epsi^{2})$. We examine therefore more closely the second one. Using equation \eqref{expansionresolvent} and the corollary to lemma \ref{lemmapseudo}, we get that \begin{equation*}\begin{split} &[\chi(\vsigma^{*}\hepsisigma\vsigma)-\chi(h_{0})]\qm\tilde{\chi}(h_{0})=(\mathrm{Q}_{M+1} + \mathrm{Q}_{M-1})\xi(h_{0})[\chi(\vsigma^{*}\hepsisigma\vsigma)+\\ &-\chi(h_{0})]\cdot\qm\tilde{\chi}(h_{0}) + \Or(\epsi^{2})_{\mathcal{L}(\hilbert{H})}, \end{split} \end{equation*} where the function $\xi$ has a support slightly larger than that of the function $\chi$. Applying again the corollary, we get that, if we choose the support of the function $\tilde{\chi}$ sufficiently smaller than the support of the function $\chi$, then \begin{equation}\label{honeqnchi} h_{1,\chi}\qm\tilde{\chi}(h_{0})= \Or(\epsi^{2}), \end{equation} so we get the first estimate. For the second one, we apply again Duhamel formula, but inverting the position of the two unitaries: \begin{equation*}\begin{split} &\qm(\E^{-\I t\heff/\epsi}-\E^{-\I th_{0}/\epsi})\tilde{\chi}(\heff) =-\frac{\I}{\epsi}\int_{0}^{t}ds\, \qm\E^{\I (s-t)h_{0}/\epsi}(\heff-h_{0})\cdot\\ &\cdot\E^{-\I s \heff/\epsi}\tilde{\chi}(\heff)=-\I\int_{0}^{t}ds\, \E^{\I (s-t)h_{0}/\epsi}\qm h_{1,\chi}\tilde{\chi}(\heff)\E^{-\I s \heff/\epsi} +\\ &+\Or(\epsi\sqrt{M+1}\sqrt{\log(\sigma^{-1})}). \end{split} \end{equation*} It follows from the proof of lemma \ref{approxchi} (in particular equation \eqref{qnhonechichi} and what follows) that \begin{equation*} \qm h_{1,\chi}\tilde{\chi}(\heff)=\Or(\epsi\sqrt{M+1}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H})}, \end{equation*} so also the second estimate is proved. \end{proof} \begin{lemma}\label{lemtruncheff} The truncated effective Hamiltonian \begin{equation} \hefftwochi:= h_{0}+\epsi h_{1,\chi}+\epsi^{2}h_{2,\chi} \end{equation} is self-adjoint on $\hilbert{H}_{0}$ for $\epsi$ small enough. \end{lemma} \begin{proof} By construction, the coefficients $h_{i,\chi}$ belong to $\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})$, and define symmetric operators on $\hilbert{H}_{0}$. Moreover $h_{0}$ is self-adjoint on $\hilbert{H}_{0}$, therefore \begin{equation*} \norm{(\hefftwochi -h_{0})\Psi}_{\hilbert{H}}\leq C\epsi (\norm{h_{0}\Psi}_{\hilbert{H}} + \norm{\psi}_{\hilbert{H}}), \quad\forall\Psi\in\hilbert{H}_{0}. \end{equation*} By a symmetric version of the Kato theorem (\cite{ReSi2}, theorem X.13) we get then the proof. \end{proof} \begin{theorem}\emph{(First order approximation to the time evolution)} Given a function $\tilde{\chi}\in\coinf(\field{R}{})$, \begin{equation}\label{firstordertime}\begin{split} & \E^{-\I t\heff/\epsi}\qm\tilde{\chi}(\heff)=\\ &= \E^{-\I t\hdiag/\epsi}\qm\tilde{\chi}(\heff)-\I\epsi\int_{0}^{t}ds\ \E^{\I(s-t)h_{0}/\epsi}h_{2, \mathrm{OD}}\E^{-\I sh_{0}/\epsi}\qm\tilde{\chi}(\heff) +\\ &+\Or(\epsi^{2}\abs{t}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H})} + \Or(\epsi^{2}\abs{t}^{2}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H})}, \end{split} \end{equation} where $[\hdiag, \qm]=0\,\,\, \forall\, M$, \begin{equation}\label{hdiag}\begin{split} &\hdiag:= \sum_{l=1}^{N}\frac{1}{2m_{l}^{\epsi}}\popb_{l}^{2}+E(x)+ \hfield +\\ &-\frac{\epsi^{2}}{4}\underset{(l_{1}\neq j_{1})}{\sum_{l_{1}, j_{1}=1}^{N}}\int_{\field{R}{3}}dk\, \frac{\hat{\varrho}_{l_{1}}(k)^{*}\hat{\varrho}_{j_{1}}(k)}{\abs{k}^{2}}\big[\E^{\I k\cdot(x_{j_{1}}-x_{l_{1}})}(\kappa\cdot\popb_{l_{1}})(\kappa\cdot\popb_{j_{1}})+\\ &+(\kappa\cdot\popb_{l_{1}})(\kappa\cdot\popb_{j_{1}})\E^{\I k\cdot(x_{j_{1}}-x_{l_{1}})}\big], \end{split} \end{equation} with $m_{l}^{\epsi}:=1/(1+\frac{\epsi^{2}}{2}e_{l})$, $\kappa:= k/\abs{k}$ and \begin{equation} e_{l}:= \frac{1}{4\pi}\int_{\field{R}{3}\times\field{R}{3}}dx\ dy\ \frac{\varrho_{l}(x)\varrho_{l}(y)}{\abs{x-y}} \quad . \end{equation} The off-diagonal Hamiltonian is defined by \begin{equation}\begin{split} & h_{2, \mathrm{OD}}:= \phisigma\cdot\nabla E(x) \quad . \end{split} \end{equation} \end{theorem} \begin{proof} We split the proof into two parts. In the first one, we show that equation \eqref{firstordertime} is true with an off-diagonal Hamiltonian $\tilde{h}_{2, \mathrm{OD}}$ defined by \begin{equation*}\begin{split} & \tilde{h}_{2, \mathrm{OD}}:= \phisigma\cdot\nabla E(x) - \frac{1}{2}\big(\deriv{j}{}\Phi_{0, l}\pop_{l}\pop_{j} + \pop_{l}\pop_{j}\deriv{j}{}\Phi_{0, l}\big) +\\ &+ \frac{1}{4}\sum_{j=1}^{3n}\big[a(\I\deriv{j}{}v_{\sigma})^{2} + a(\I\deriv{j}{}v_{\sigma})^{*\,2}] \quad . \end{split} \end{equation*} In the second one, we prove that the last two terms in $\tilde{h}_{2, \mathrm{OD}}$ give actually rise to higher order contributions to the time evolution, although their norm in $\mathcal{L}(\hilbert{H}_{0}, \hilbert{H})$ is not small. This is caused by the fact that they are strongly oscillating in $\abs{k}$, so that their behavior is determined by the value of the density of states in a neighborhood of $k=0$. For both terms, it however vanishes for $k=0$, uniformly in $\sigma$, and this implies that they are of lower order with respect to the leading piece $\phisigma\cdot\nabla E(x)$, whose density instead diverges logarithmically in $\sigma$ (we elaborate on this last observation in a corollary to this theorem). We start showing that we can, up to the desired error, replace $\heff$ by $\hefftwochi$. By lemma \ref{lemtruncheff}, $\hefftwochi$ is self-adjoint on $\hilbert{H}_{0}$ like $\heff$, therefore we can apply Duhamel formula, getting \begin{equation*}\begin{split} & \E^{-\I t\heff/\epsi}-\E^{-\I t\hefftwochi/\epsi} = -\frac{\I}{\epsi}\int_{0}^{t}ds\, \E^{\I (s-t)\heff/\epsi}(\heff-\hefftwochi)\E^{-\I s \hefftwochi/\epsi}=\\ &=\Or(\epsi^{2}\big(\log(\sigma^{-1})\big)^{3/2}), \end{split} \end{equation*} using theorem \ref{thmheff}. Moreover, using lemma \ref{approxchi}, we can replace $\qm\tilde{\chi}(\heff)$ by $\qm\tilde{\chi}(h_{0})\tilde{\tilde{\chi}}(\heff)$. Since the diagonal Hamiltonian $\hdiag$ is also self-adjoint on $\hilbert{H}_{0}$ for $\epsi$ sufficiently small (the proof can be given along the same lines of lemma \ref{lemtruncheff}), we apply again Duhamel formula, \begin{equation*}\begin{split} & (\E^{-\I t\hefftwochi/\epsi}-\E^{-\I t\hdiag/\epsi})\qm\tilde{\chi}(h_{0})\tilde{\tilde{\chi}}(\heff) =\\ &=-\frac{\I}{\epsi}\int_{0}^{t}ds\ \E^{\I(s-t)\hefftwochi/\epsi}(\hefftwochi-\hdiag)\E^{-\I s\hdiag/\epsi}\qm\tilde{\chi}(h_{0})\tilde{\tilde{\chi}}(\heff)=\\ &=-\I\int_{0}^{t}ds\ \E^{\I(s-t)\hefftwochi/\epsi}h_{1,\chi}\E^{-\I s\hdiag/\epsi}\qm\tilde{\chi}(h_{0})\tilde{\tilde{\chi}}(\heff)+\\ &-\I\epsi\int_{0}^{t}ds\ \E^{\I(s-t)\hefftwochi/\epsi}(h_{2,\chi}-h_{2, \mathrm{D}})\E^{-\I s\hdiag/\epsi}\qm\tilde{\chi}(h_{0})\tilde{\tilde{\chi}}(\heff). \end{split} \end{equation*} To analyze the first term, we remark that, proceeding as in lemma \ref{approxchi}, one can prove that \begin{equation*} \tilde{\chi}(h_{0})-\tilde{\chi}(\hdiag)= \Or(\epsi^{2}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H})}, \end{equation*} so \begin{equation*} [\E^{-\I t\hdiag/\epsi}, \tilde{\chi}(h_{0})] = \Or(\epsi^{2}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H})}, \end{equation*} therefore \begin{equation*}\begin{split} &-\I\int_{0}^{t}ds\ \E^{\I(s-t)\hefftwochi/\epsi}h_{1,\chi}\E^{-\I s\hdiag/\epsi}\qm\tilde{\chi}(h_{0})\tilde{\tilde{\chi}}(\heff)=\\ &=-\I\int_{0}^{t}ds\ \E^{\I(s-t)\hefftwochi/\epsi}h_{1,\chi}\qm\tilde{\chi}(h_{0})\E^{-\I s\hdiag/\epsi}\tilde{\tilde{\chi}}(\heff) +\\ &+ \Or(\epsi^{2}\abs{t}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H})}=\Or(\epsi^{2}\abs{t}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H})}, \end{split} \end{equation*} using equation \eqref{honeqnchi}. Concerning the second one, applying once again Duhamel formula, we have \begin{equation*}\begin{split} & -\I\epsi\int_{0}^{t}ds\ \E^{\I(s-t)\hefftwochi/\epsi}(h_{2,\chi}-h_{2, \mathrm{D}})\E^{-\I s\hdiag/\epsi}\qm\tilde{\chi}(h_{0})\tilde{\tilde{\chi}}(\heff)=\\ &= -\I\epsi\int_{0}^{t}ds\ \E^{\I(s-t)\hefftwochi/\epsi}(h_{2,\chi}-h_{2, \mathrm{D}})\E^{-\I sh_{0}/\epsi}\qm\tilde{\chi}(h_{0})\tilde{\tilde{\chi}}(\heff) +\\ &+ \Or(\epsi^{2}\abs{t}^{2}\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H})}, \end{split} \end{equation*} so we have to look at $\E^{\I(s-t)\hefftwochi/\epsi}(h_{2,\chi}-h_{2, \mathrm{D}})\qm\tilde{\chi}(h_{0})$. Following a procedure already employed several times, we first observe that, in the expression for $h_{2, \chi}$, equation \eqref{htwochi}, we can replace, making an error of order $\Or(\epsi)$, $\chi(\vsigma^{*}\hepsisigma \vsigma)$ with $\chi(h_{0})$. Using the corollary to lemma \ref{lemmapseudo}, we can then eliminate from $h_{2, \chi}\qm\tilde{\chi}(h_{0})$ all the terms containing $(\mathbf{1}-\chi(h_{0}))$. What remains is then \begin{equation*}\begin{split} & \E^{\I(s-t)\hefftwochi/\epsi}(h_{2,\chi}-h_{2, \mathrm{D}})\qm\tilde{\chi}(h_{0})=\\ &=\E^{\I(s-t)\hefftwochi/\epsi}\sum_{i=-2}^{2}\mathrm{Q}_{M+i}\xi(h_{0})\bigg\{\frac{1}{4}\sum_{j=1}^{3N}\big[a(\I\deriv{j}{}v_{\sigma})^{2} + a(\I\deriv{j}{}v_{\sigma})^{*\,2}]+\phisigmaj\cdot\nabla E(x)+\\ &-\deriv{j}{}\phizerojl\pop_{l}\pop_{j}-\frac{1}{2}\sum_{l_{1}=1}^{N}e_{l_{1}}\popb_{l_{1}}^{2}-\frac{1}{4}\underset{(l_{1}\neq j_{1})}{\sum_{l_{1}, j_{1}=1}^{N}}\int_{\field{R}{3}}dk\, \frac{\hat{\varrho}_{l_{1}}(k)^{*}\hat{\varrho}_{j_{1}}(k)}{\abs{k}^{2}}\cdot\\ &\cdot\big[\E^{\I k\cdot(x_{j_{1}}-x_{l_{1}})}(\kappa\cdot\popb_{l_{1}})(\kappa\cdot\popb_{j_{1}})+ (\kappa\cdot\popb_{l_{1}})(\kappa\cdot\popb_{j_{1}})\E^{\I k\cdot(x_{j_{1}}-x_{l_{1}})}\big]\bigg\}\qm\tilde{\chi}(h_{0})+\\ &+\frac{\I}{2}\E^{\I(s-t)\hefftwochi/\epsi}\sum_{i=-2}^{2}\mathrm{Q}_{M+i}\xi(h_{0})\big[\chi(h_{0})\phim(\I\nabla_{x}v_{\sigma})\cdot\pop,\ \chi(h_{0})\phisigmaj\cdot\pop \big]\qm\tilde{\chi}(h_{0}) + \\ &+\Or(\epsi\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H})}. \end{split} \end{equation*} Applying theorem \ref{zeroordertime}, we can now replace $\E^{\I(s-t)\hefftwochi/\epsi}$ with $\E^{\I(s-t)h_{0}/\epsi}$ making an error of order $\Or(\abs{t}\epsi\sqrt{\log(\sigma^{-1})})$. Concerning the last term, we get, expanding the commutator and applying again the corollary to lemma \ref{lemmapseudo}, \begin{equation*}\begin{split} &\big[\chi(h_{0})\phim(\I\nabla_{x}v_{\sigma})\cdot\pop,\ \chi(h_{0})\phisigmaj\cdot\pop \big]\qm\tilde{\chi}(h_{0}) = \sum_{i=-1}^{1}\mathrm{Q}_{M+i}\xi(h_{0})\cdot\\ &\cdot\big[\phim(\I\nabla_{x}v_{\sigma})\cdot\pop,\ \phisigmaj\cdot\pop \big]\qm\tilde{\chi}(h_{0})=\\ &=\sum_{i=-1}^{1}\mathrm{Q}_{M+i}\xi(h_{0})\big[\phim(\I\deriv{l}{}v_{\sigma}),\ \phisigmajl\big]\pop_{l}\pop_{j}\qm\tilde{\chi}(h_{0})+\Or(\epsi\sqrt{\log(\sigma^{-1})})_{\mathcal{L}(\hilbert{H})}. \end{split} \end{equation*} Using equation \eqref{commaphi}, we can check that this cancels exactly the diagonal terms in $h_{2, \chi}-h_{2, \mathrm{D}}$, so that \begin{equation*}\begin{split} & \E^{\I(s-t)\hefftwochi/\epsi}(h_{2,\chi}-h_{2, \mathrm{D}})\qm\tilde{\chi}(h_{0})=\E^{\I(s-t)h_{0}/\epsi}\tilde{h}_{2, \mathrm{OD}}\qm\tilde{\chi}(h_{0}) +\\ &+ \Or(\epsi\abs{t}\sqrt{\log(\sigma^{-1})}). \end{split} \end{equation*} We examine now separately the last two terms in $\tilde{h}_{2, \mathrm{OD}}$, to prove that they can be neglected. Putting $\Psi:= \qm\tilde{\chi}(\heff)\Psi_{0}$, we get for the first \begin{equation*}\begin{split} & \I\epsi\int_{0}^{t}ds\ \E^{\I(s-t)h_{0}/\epsi}\deriv{j}{}\Phi_{0, l}\pop_{l}\pop_{j}\E^{-\I sh_{0}/\epsi}\qm\tilde{\chi}(\heff)\Psi_{0}=\\ & =\I\E^{-\I th_{0}/\epsi}\frac{\epsi}{\sqrt{2}}\int_{0}^{t}ds\ \E^{\I sh_{0}/\epsi}\bigg[a\bigg(\frac{\deriv{l}{}\deriv{j}{}v(x, k)}{\abs{k}}\bigg)^{*} + a\bigg(\frac{\deriv{l}{}\deriv{j}{}v(x, k)}{\abs{k}}\bigg)\bigg]\cdot\\ &\cdot\pop_{j}\pop_{l}\E^{-\I sh_{0}/\epsi}\Psi=\\ &=-\I\E^{-\I th_{0}/\epsi}\frac{\epsi}{\sqrt{2}}\int_{0}^{t}ds\, \E^{\I sh_{0}/\epsi}\bigg[a\bigg(\kappa_{j_{2}}\kappa_{l_{2}}\E^{\I k\cdot x_{j_{1}}}\frac{\hat{\varrho}_{j_{1}}(k)}{\abs{k}^{1/2}}\bigg)^{*} +\\ &+ a\bigg(\kappa_{j_{2}}\kappa_{l_{2}}\E^{\I k\cdot x_{j_{1}}}\frac{\hat{\varrho}_{j_{1}}(k)}{\abs{k}^{1/2}}\bigg)\bigg]\pop_{(j_{1}, j_{2})}\pop_{(j_{1}, l_{2})}\E^{-\I sh_{0}/\epsi}\Psi \quad . \end{split} \end{equation*} The part with the annihilation operator gives \begin{equation}\label{estimateapp}\begin{split} & \E^{-\I th_{0}/\epsi}\frac{\epsi}{\sqrt{2}}\int_{0}^{t}ds\, \E^{\I sh_{0}/\epsi}a\bigg(\kappa_{j_{2}}\kappa_{l_{2}}\E^{\I k\cdot x_{j_{1}}}\frac{\hat{\varrho}_{j_{1}}(k)}{\abs{k}^{1/2}}\bigg)\pop_{(j_{1}, j_{2})}\pop_{(j_{1}, l_{2})}\E^{-\I sh_{0}/\epsi}\Psi=\\ & = \E^{-\I th_{0}/\epsi}\frac{\epsi}{\sqrt{2}}\sqrt{M}\int_{0}^{t}ds\,\E^{\I s\sum_{\mu=1}^{M-1}\abs{k_{\mu}}/\epsi}\E^{\I s h_{\mathrm{p}}/\epsi}\int_{\field{R}{3}}dk\, \E^{-\I k\cdot x_{j_{1}}}\frac{\hat{\varrho}_{j_{1}}(k)^{*}}{\abs{k}^{1/2}}(\kappa\cdot\popb_{j_{1}})^{2}\cdot\\ &\cdot\E^{-\I s h_{\mathrm{p}}/\epsi}\E^{-\I s\abs{k}/\epsi}\E^{-\I s\sum_{\mu=1}^{M-1}\abs{k_{\mu}}/\epsi}\Psi(x;k, k_{1}, \ldots, k_{M-1})=\\ &=\E^{-\I th_{0}/\epsi}\frac{\epsi}{\sqrt{2}}\sqrt{M}\int_{\field{R}{3}}dk\, \frac{\hat{\varrho}_{j_{1}}(k)^{*}}{\abs{k}^{1/2}}\int_{0}^{t}ds\, \E^{\I s h_{\mathrm{p}}/\epsi}\E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{2}\E^{-\I s h_{\mathrm{p}}/\epsi}\cdot\\ &\cdot\E^{-\I s\abs{k}/\epsi}\Psi(x;k, k_{1}, \ldots, k_{M-1})=\\ &=\E^{-\I th_{0}/\epsi}\frac{\epsi}{\sqrt{2}}\sqrt{M}\int_{\field{R}{3}}dk\, \frac{\hat{\varrho}_{j_{1}}(k)^{*}}{\abs{k}^{1/2}(1-\I \abs{k}\epsi^{-1})}(1-\I \abs{k}\epsi^{-1})\int_{0}^{t}ds\, \E^{\I s h_{\mathrm{p}}/\epsi}\cdot\\ &\cdot\E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{2}\E^{-\I s h_{\mathrm{p}}/\epsi}\E^{-\I s\abs{k}/\epsi}\Psi(x;k, k_{1}, \ldots, k_{M-1}). \end{split} \end{equation} Integrating by parts we have \begin{equation}\label{intparts}\begin{split} & -\I \abs{k}\epsi^{-1}\int_{0}^{t}ds\, \E^{\I s h_{\mathrm{p}}/\epsi}\E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{2}\E^{-\I s h_{\mathrm{p}}/\epsi}\E^{-\I s\abs{k}/\epsi}\Psi(x;k, k_{1}, \ldots, k_{M-1})=\\ &=\int_{0}^{t}ds\, \E^{\I s h_{\mathrm{p}}/\epsi}\E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{2}\E^{-\I s h_{\mathrm{p}}/\epsi}\deriv{s}{}\E^{-\I s\abs{k}/\epsi}\Psi(x;k, k_{1}, \ldots, k_{M-1})=\\ &=\E^{\I t h_{\mathrm{p}}/\epsi}\E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{2}\E^{-\I t h_{\mathrm{p}}/\epsi}\E^{-\I t\abs{k}/\epsi}\Psi(x;k, k_{1}, \ldots, k_{M-1}) +\\ & - \E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{2}\Psi(x;k, k_{1}, \ldots, k_{M-1})+\\ &-\frac{\I}{\epsi}\int_{0}^{t}ds\, \E^{\I s h_{\mathrm{p}}/\epsi}[\hp, \E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{2}]\E^{-\I s h_{\mathrm{p}}/\epsi}\E^{-\I s\abs{k}/\epsi}\Psi(x;k, k_{1}, \ldots, k_{M-1})=\\ &=\E^{\I t h_{\mathrm{p}}/\epsi}\E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{2}\E^{-\I t h_{\mathrm{p}}/\epsi}\E^{-\I t\abs{k}/\epsi}\Psi(x;k, k_{1}, \ldots, k_{M-1}) +\\ & - \E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{2}\Psi(x;k, k_{1}, \ldots, k_{M-1})+\\ &-\I\int_{0}^{t}ds\, \E^{\I s h_{\mathrm{p}}/\epsi}\bigg\{\bigg[-\I\nabla_{x_{j_{1}}}(\E^{-\I k\cdot x_{j_{1}}})\cdot\popb_{j_{1}} - \epsi\Delta_{x}(\E^{-\I k\cdot x_{j_{1}}})\bigg](\kappa\cdot\popb_{j_{1}})^{2} +\\ &+ \E^{-\I k\cdot x_{j_{1}}}[E(x), (\kappa\cdot\popb_{j_{1}})^{2}]\bigg]\bigg\} \E^{-\I s h_{\mathrm{p}}/\epsi}\E^{-\I s\abs{k}/\epsi}\Psi(x;k, k_{1}, \ldots, k_{M-1})=\\ &= \E^{\I t h_{\mathrm{p}}/\epsi}\E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{2}\E^{-\I t h_{\mathrm{p}}/\epsi}\E^{-\I t\abs{k}/\epsi}\Psi(x;k, k_{1}, \ldots, k_{M-1}) +\\ & - \E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{2}\Psi(x;k, k_{1}, \ldots, k_{M-1})+\\ &+\I\int_{0}^{t}ds\, \E^{\I s h_{\mathrm{p}}/\epsi}\bigg\{\bigg[\E^{-\I k\cdot x_{j_{1}}}\kappa\cdot\popb_{j_{1}} - \epsi\abs{k}\E^{-\I k\cdot x_{j_{1}}}\bigg](\kappa\cdot\popb_{j_{1}})^{2} +\\ &+ \E^{-\I k\cdot x_{j_{1}}}[E(x), (\kappa\cdot\popb_{j_{1}})^{2}]\bigg]\bigg\} \E^{-\I s h_{\mathrm{p}}/\epsi}\E^{-\I s\abs{k}/\epsi}\abs{k}\Psi(x;k, k_{1}, \ldots, k_{M-1})\quad . \end{split} \end{equation} We have to put now this expression back into \eqref{estimateapp} and estimate the result. We show how to proceed for the most singular term, i.\ e. the one containing \begin{equation*} \I\int_{0}^{t}ds\, \E^{\I s h_{\mathrm{p}}/\epsi}\E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{3}\E^{-\I s h_{\mathrm{p}}/\epsi}\E^{-\I s\abs{k}/\epsi}\abs{k}\Psi(x;k, k_{1}, \ldots, k_{M-1}), \end{equation*} the others can be treated in the same way. Putting this term back in \eqref{estimateapp}, and ignoring the unitary $\E^{-\I th_{0}/\epsi}$, which does not change the norm, we have to estimate in the end \begin{equation*}\begin{split} & \frac{\I\epsi}{\sqrt{2}}\sqrt{M}\int_{\field{R}{3}}dk\, \frac{\hat{\varrho}_{j_{1}}(k)^{*}}{\abs{k}^{1/2}(1-\I \abs{k}\epsi^{-1})}\int_{0}^{t}ds\, \E^{\I s h_{\mathrm{p}}/\epsi}\E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{3}\E^{-\I s h_{\mathrm{p}}/\epsi}\cdot\\ &\cdot\E^{-\I s\abs{k}/\epsi}\abs{k}\Psi(x;k, k_{1}, \ldots, k_{M-1}). \end{split} \end{equation*} Applying Cauchy-Schwarz inequality we get \begin{equation*}\begin{split} & \abs{\cdots}^{2}\leq \frac{\epsi^{2}\abs{t}^{2}M}{2}\int_{\field{R}{3}}dk\, \frac{\abs{\hat{\varrho}_{j_{1}}(k)}^{2}}{\abs{k}(1+\abs{k}^{2}\epsi^{-2})}\cdot\\ &\cdot\int_{\field{R}{3}}dk\, \int_{0}^{t}ds\, \big\lvert\E^{\I s h_{\mathrm{p}}/\epsi}\E^{-\I k\cdot x_{j_{1}}}(\kappa\cdot\popb_{j_{1}})^{3}\E^{-\I s h_{\mathrm{p}}/\epsi}\E^{-\I s\abs{k}/\epsi}\abs{k}\cdot\\ &\cdot\Psi(x;k, k_{1}, \ldots, k_{M-1})\big\rvert^{2}. \end{split} \end{equation*} The first integral gives \begin{equation*}\begin{split} & \int_{\field{R}{3}}dk\, \frac{\abs{\hat{\varrho}_{j_{1}}(k)}^{2}}{\abs{k}(1+\abs{k}^{2}\epsi^{-2})}= C\int_{0}^{\Lambda}dk\, \frac{k}{1 + k^{2}\epsi^{-2}}=C\epsi^{2}\int_{0}^{\Lambda/\epsi}\frac{k}{1 + k^{2}}=\\ &=\Or(\epsi^{2}\log(1/\epsi)), \end{split} \end{equation*} so, calculating the norm we get \begin{equation*}\begin{split} & \norm{\cdots}_{\hilbert{H}}^{2}\leq C\epsi^{4}\log(1/\epsi)\abs{t}^{2}M\int_{0}^{t}ds\, \big\lVert \E^{\I s h_{\mathrm{p}}/\epsi}\E^{-\I k\cdot x_{j_{1}}}\cdot\\ &\cdot(\kappa\cdot\popb_{j_{1}})^{3}\E^{-\I s h_{\mathrm{p}}/\epsi}\E^{-\I s\abs{k}/\epsi}\abs{k}\Psi(x;k, k_{1}, \ldots, k_{M-1})\big\rVert^{2}_{\hilbert{H}}=\\ &=C\epsi^{4}\log(1/\epsi)\abs{t}^{2}M\int_{0}^{t}ds\, \big\lVert(\kappa\cdot\popb_{j_{1}})^{3}\E^{-\I s h_{\mathrm{p}}/\epsi}\abs{k}\Psi(x;k, k_{1}, \ldots, k_{M-1})\big\rVert^{2}_{\hilbert{H}}\leq\\ &\leq C\epsi^{4}\log(1/\epsi)\abs{t}^{2}M\int_{0}^{t}ds\, \big\lVert(\kappa\cdot\popb_{j_{1}})^{3}\E^{-\I s h_{\mathrm{p}}/\epsi}\abs{k}\qm\tilde{\chi}(\heff)\big\rVert^{2}_{\mathcal{L}(\hilbert{H})}\norm{\Psi_{0}}^{2}_{\hilbert{H}}, \end{split} \end{equation*} which shows that the norm of this term is of order $\Or(\epsi^{2}\abs{t}\sqrt{\log(1/\epsi)})$ in $\mathcal{L}(\hilbert{H})$. The part with the creation operator gives \begin{equation*}\begin{split} & \E^{-\I th_{0}/\epsi}\frac{\epsi}{\sqrt{2}}\int_{0}^{t}ds\, \E^{\I sh_{0}/\epsi}a\bigg(\kappa_{j_{2}}\kappa_{l_{2}}\E^{\I k\cdot x_{j_{1}}}\frac{\hat{\varrho}_{j_{1}}(k)}{\abs{k}^{1/2}}\bigg)^{*}\pop_{(j_{1}, j_{2})}\pop_{(j_{1}, l_{2})}\E^{-\I sh_{0}/\epsi}\Psi=\\ & = \E^{-\I th_{0}/\epsi}\frac{\epsi}{\sqrt{2}}\int_{0}^{t}ds\, \E^{\I s\sum_{\mu=1}^{M+1}\abs{k_{\mu}}/\epsi} \E^{\I s\hp/\epsi}\frac{1}{\sqrt{M+1}}\sum_{\mu=1}^{M+1}\kappa_{\mu, j_{2}}\kappa_{\mu, l_{2}}\E^{\I\kappa_{\mu}\cdot x_{j_{1}}}\frac{\hat{\varrho}_{j_{1}}(k_{\mu})}{\abs{k_{\mu}}^{1/2}}\cdot\\ &\cdot \pop_{(j_{1}, j_{2})}\pop_{(j_{1}, l_{2})}\E^{-\I s\sum_{\nu=1(\nu\neq\mu)}^{M+1}\abs{k_{\nu}}/\epsi}\E^{-\I s\hp/\epsi}\Psi(x; k_{1}, \ldots, \hat{k}_{\mu}, \ldots, k_{M+1})=\\ &= \E^{-\I th_{0}/\epsi}\frac{\epsi}{\sqrt{2(M+1)}}\frac{\hat{\varrho}_{j_{1}}(k_{\mu})}{\abs{k_{\mu}}^{1/2}}\int_{0}^{t}ds\, \E^{\I s\hp/\epsi}\E^{\I\kappa_{\mu}\cdot x_{j_{1}}}(\kappa_{\mu}\cdot\popb_{j_{1}})^{2}\E^{-\I s\hp/\epsi}\E^{-\I s\abs{k_{\mu}}/\epsi}\cdot\\ &\cdot\Psi(x; k_{1}, \ldots, \hat{k}_{\mu}, \ldots, k_{M+1})=\\ &=\E^{-\I th_{0}/\epsi}\frac{\epsi}{\sqrt{2(M+1)}}\frac{\hat{\varrho}_{j_{1}}(k_{\mu})}{\abs{k_{\mu}}^{1/2}(1-\I\abs{k_{\mu}}\epsi^{-1})}(1-\I\abs{k_{\mu}}\epsi^{-1})\int_{0}^{t}ds\, \E^{\I s\hp/\epsi}\E^{\I\kappa_{\mu}\cdot x_{j_{1}}}\cdot\\ &\cdot(\kappa_{\mu}\cdot\popb_{j_{1}})^{2}\E^{-\I s\hp/\epsi}\E^{-\I s\abs{k_{\mu}}/\epsi}\Psi(x; k_{1}, \ldots, \hat{k}_{\mu}, \ldots, k_{M+1})\quad . \end{split} \end{equation*} We can now integrate by parts using equation \eqref{intparts}. As for the annihilation part, we examine just one term, \begin{equation*}\begin{split} & \frac{\epsi}{\sqrt{2(M+1)}}\frac{\hat{\varrho}_{j_{1}}(k_{\mu})\abs{k_{\mu}}^{1/2}}{(1-\I\abs{k_{\mu}}\epsi^{-1})}\int_{0}^{t}ds\, \E^{\I s h_{\mathrm{p}}/\epsi}\E^{\I k_{\mu}\cdot x_{j_{1}}}(\kappa_{\mu}\cdot\popb_{j_{1}})^{3}\E^{-\I s h_{\mathrm{p}}/\epsi}\cdot\\ &\cdot\E^{-\I s\abs{k_{\mu}}/\epsi}\Psi(x;k_{1}, \ldots, \hat{k}_{\mu}, \ldots, k_{M+1}). \end{split} \end{equation*} The norm squared is given by \begin{equation*}\begin{split} & \frac{\epsi^{2}}{2(M+1)}\int dx\int dk_{1}\ldots dk_{M+1} \frac{\abs{\hat{\varrho}_{j_{1}}(k_{\mu})}^{2}\abs{k_{\mu}}}{1+\abs{k_{\mu}}^{2}\epsi^{-2}}\bigg\lvert\int_{0}^{t}ds\,\E^{-\I s\abs{k_{\mu}}/\epsi}\E^{\I s\hp/\epsi}\E^{\I k_{\mu}\cdot x_{j_{1}}}\cdot\\ &\cdot (\kappa_{\mu}\cdot\popb_{j_{1}})^{3}\E^{-\I s\hp/\epsi}\Psi(x;k_{1}, \ldots, \hat{k}_{\mu}, \ldots, k_{M+1})\bigg\rvert^{2}=\\ &=\frac{\epsi^{6}}{2(M+1)}\int dx\int dk_{1}\ldots dk_{M+1} \frac{\abs{\hat{\varrho}_{j_{1}}(\epsi k_{\mu})}^{2}\abs{k_{\mu}}}{1+\abs{k_{\mu}}^{2}}\bigg\lvert\int_{0}^{t}ds\,\E^{-\I s\abs{k_{\mu}}}\E^{\I s\hp/\epsi}\E^{\I \epsi k_{\mu}\cdot x_{j_{1}}}\cdot\\ &\cdot (\kappa_{\mu}\cdot\popb_{j_{1}})^{3}\E^{-\I s\hp/\epsi}\Psi(x;k_{1}, \ldots, \hat{k}_{\mu}, \ldots, k_{M+1})\bigg\rvert^{2}\leq\\ &\leq \frac{\epsi^{6}\abs{t}}{2(M+1)}\int dk_{1}\ldots dk_{M+1} \frac{\abs{\hat{\varrho}_{j_{1}}(\epsi k_{\mu})}^{2}\abs{k_{\mu}}}{1+\abs{k_{\mu}}^{2}}\int_{0}^{t}ds\,\big\lVert\E^{-\I s\abs{k_{\mu}}}\E^{\I s\hp/\epsi}\E^{\I \epsi k_{\mu}\cdot x_{j_{1}}}\cdot\\ &\cdot (\kappa_{\mu}\cdot\popb_{j_{1}})^{3}\E^{-\I s\hp/\epsi}\Psi(x;k_{1}, \ldots, \hat{k}_{\mu}, \ldots, k_{M+1})\big\rVert^{2}=\\ &=\frac{\epsi^{6}\abs{t}}{2(M+1)}\int dk_{1}\ldots dk_{M+1} \frac{\abs{\hat{\varrho}_{j_{1}}(\epsi k_{\mu})}^{2}\abs{k_{\mu}}}{1+\abs{k_{\mu}}^{2}}\int_{0}^{t}ds\,\big\lVert\E^{-\I s\abs{k_{\mu}}}\E^{\I s\hp/\epsi}\E^{\I \epsi k_{\mu}\cdot x_{j_{1}}}\cdot\\ &\cdot (\kappa_{\mu}\cdot\popb_{j_{1}})^{3}\E^{-\I s\hp/\epsi}\Psi(x;k_{1}, \ldots, \hat{k}_{\mu}, \ldots, k_{M+1})\big\rVert^{2}_{L^{2}(\field{R}{3N}, dx)}=\\ &=\frac{\epsi^{6}\abs{t}}{2(M+1)}\int dk_{1}\ldots dk_{M+1} \frac{\abs{\hat{\varrho}_{j_{1}}(\epsi k_{\mu})}^{2}\abs{k_{\mu}}}{1+\abs{k_{\mu}}^{2}}\int_{0}^{t}ds\,\big\lVert(\kappa_{\mu}\cdot\popb_{j_{1}})^{3}\E^{-\I s\hp/\epsi}\cdot\\ &\cdot\Psi(x;k_{1}, \ldots, \hat{k}_{\mu}, \ldots, k_{M+1})\big\rVert^{2}_{L^{2}(\field{R}{3N}, dx)}\leq\\ &\leq \frac{C\epsi^{6}\abs{t}}{2(M+1)}\int_{0}^{\varrho_{0}/\epsi} d\abs{k_{\mu}} \frac{\abs{k_{\mu}}^{3}}{1+\abs{k_{\mu}}^{2}}\int d\Omega_{\mu}\int_{0}^{t}ds\,\big\lVert(\kappa_{\mu}\cdot\popb_{j_{1}})^{3}\E^{-\I s\hp/\epsi}\cdot\\ &\cdot\Psi(x;k_{1}, \ldots, \hat{k}_{\mu}, \ldots, k_{M+1})\big\rVert^{2}_{\hilbert{H}}=\\ &=\frac{C\epsi^{6}\abs{t}\varrho_{0}^{2}}{4(M+1)}[\epsi^{-2}-\log(\epsi^{-2}+1)]\int d\Omega_{\mu}\int_{0}^{t}ds\,\big\lVert(\kappa_{\mu}\cdot\popb_{j_{1}})^{3}\E^{-\I s\hp/\epsi}\cdot\\ &\cdot\qm\tilde{\chi}(\heff)\Psi_{0}\big\rVert^{2}_{\hilbert{H}}, \end{split} \end{equation*} so we get that this term gives a contribution of order $\Or(\epsi^{2}\abs{t})$ in the norm of $\mathcal{L}(\hilbert{H})$. We separate also in the second term of $\tilde{h}_{2, \mathrm{OD}}$ the annihilation and the creation part. For the annihilation one we get \begin{equation*}\begin{split} & -\frac{\I\epsi}{4}\int_{0}^{t}ds\ \E^{\I(s-t)h_{0}/\epsi}\sum_{j=1}^{3N}a(\I\deriv{j}{}v_{\sigma})^{2}\E^{-\I sh_{0}/\epsi}\Psi=\\ &=-\frac{\I\epsi}{4}\sqrt{M(M-1)}\E^{-\I th_{0}/\epsi}\sum_{j_{1}, j_{2}}\int d\xi\int d\zeta\, \hat{\xi}_{j_{2}}\hat{\zeta}_{j_{2}} \frac{\hat{\varrho}_{j_{1}}^{\sigma}(\xi)^{*}\hat{\varrho}_{j_{1}}^{\sigma}(\zeta)^{*}}{\abs{\xi}^{1/2}\abs{\zeta}^{1/2}}\cdot\\ &\cdot\int_{0}^{t}ds\ \E^{-\I s(\abs{\xi}+\abs{\zeta})/\epsi} \E^{\I s\hp/\epsi}\E^{-\I(\xi+\zeta)\cdot x_{j_{1}}}\E^{-\I s\hp/\epsi}\Psi \quad . \end{split} \end{equation*} We proceed now in the same way as we did for the first term. Integrating by parts we get \begin{equation*}\begin{split} & \frac{-\I(\abs{\xi}+\abs{\zeta})}{\epsi}\int_{0}^{t}ds\, \E^{-\I s(\abs{\xi}+\abs{\zeta})/\epsi} \E^{\I s\hp/\epsi}\E^{-\I(\xi+\zeta)\cdot x_{j_{1}}}\E^{-\I s\hp/\epsi}\Psi=\\ &=\E^{-\I t(\abs{\xi}+\abs{\zeta})/\epsi} \E^{\I t\hp/\epsi}\E^{-\I(\xi+\zeta)\cdot x_{j_{1}}}\E^{-\I t\hp/\epsi}\Psi - \E^{-\I(\xi+\zeta)\cdot x_{j_{1}}}\Psi +\\ &-\int_{0}^{t}ds\, \E^{-\I s(\abs{\xi}+\abs{\zeta})/\epsi} \E^{\I s\hp/\epsi}\E^{-\I(\xi+\zeta)\cdot x_{j_{1}}}[(\xi+\zeta)\cdot\popb_{j_{1}} - \epsi\abs{\xi+\zeta}^{2}]\E^{-\I s\hp/\epsi}\Psi. \end{split} \end{equation*} As in the previous case, we examine just one term, \begin{equation*}\begin{split} & \frac{\I\epsi}{4}\sqrt{M(M-1)}\int d\xi\int d\zeta\, \hat{\xi}_{j_{2}}\hat{\zeta}_{j_{2}} \frac{\hat{\varrho}_{j_{1}}^{\sigma}(\xi)^{*}\hat{\varrho}_{j_{1}}^{\sigma}(\zeta)^{*}}{\abs{\xi}^{1/2}\abs{\zeta}^{1/2}[1-\I(\abs{\xi}+\abs{\zeta})\epsi^{-1}]}\cdot\\ &\cdot\int_{0}^{t}ds\, \E^{-\I s(\abs{\xi}+\abs{\zeta})/\epsi} \E^{\I s\hp/\epsi}\E^{-\I(\xi+\zeta)\cdot x_{j_{1}}}(\xi+\zeta)\cdot\popb_{j_{1}}\E^{-\I s\hp/\epsi}\cdot\\ &\cdot\Psi(x;\xi, \zeta, k_{1}, \ldots, k_{M-2}). \end{split} \end{equation*} the others can be treated in the same way. Using Cauchy-Schwarz inequality we get then \begin{equation*}\begin{split} & \norm{\cdots}_{\hilbert{H}}^{2}\leq C\epsi^{2}\abs{t}M(M-1)\int_{\sigma}^{\Lambda} d\abs{\xi}\int_{\sigma}^{\Lambda} d\abs{\zeta} \frac{\abs{\xi}\abs{\zeta}(\abs{\xi}^{2}+\abs{\zeta}^{2})}{1+(\abs{\xi}+\abs{\zeta})^{2}\epsi^{-2}}\cdot\\ &\cdot \int_{0}^{t}ds\, \bigg\lVert \frac{\xi+\zeta}{\abs{\xi+\zeta}}\cdot\popb_{j_{1}}\E^{-\I s\hp/\epsi}\qn\tilde{\chi}(\heff)\bigg\rVert_{\mathcal{L}(\hilbert{H})}^{2}\norm{\Psi_{0}}_{\hilbert{H}}^{2}=\Or(\epsi^{4}\abs{t}^{2})_{\hilbert{H}}\quad, \end{split} \end{equation*} independently of $\sigma$. The creation part can be estimated in a way entirely analogous to the one already employed for the first term of $\tilde{h}_{2, \mathrm{OD}}$. \end{proof} \begin{corollary}\label{radiatedpiece} At the leading order, the radiated piece (i.\ e. the piece of the wave function which makes a transition between the almost invariant subspaces) for a system starting in the Fock vacuum, $\Psi_{0}(x)=\psi(x)\Omega_{\mathrm{F}}$, is given by \begin{equation} -\frac{\epsi}{\sqrt{2}}\E^{-\I th_{0}/\epsi}\sum_{j=1}^{N}\frac{\hat{\varrho}_{j}^{\sigma(\epsi)}(k)}{\abs{k}^{3/2}}\kappa\cdot\int_{0}^{t}ds\, \E^{\I s\abs{k}/\epsi}\opw(\ddot{\vec{x}}_{j}^{c}(s; x, p))\psi(x)\quad , \end{equation} where $\vec{x}_{j}^{c}$ is the solution to the classical equations of motion \begin{equation}\begin{split} & \ddot{\vec{x}}_{j}^{c}(s; x, p) = -\nabla_{\vec{x}_{j}}E(x^{c}(s; x, p)),\\ & \vec{x}_{j}^{c}(0; x, p)=\vec{x}_{j},\qquad \dot{\vec{x}}_{j}^{c}(0; x, p)=\vec{p}_{j},\quad j=1, \ldots, n\quad. \end{split} \end{equation} We get the leading order of the radiated piece corresponding to the original Hamiltonian $\hepsi$, for a system starting in the approximate dressed vacuum $\Omega_{\sigma(\epsi)}(x)$, applying to this wave function the dressing operator $V_{\sigma(\epsi)}(x)$. \end{corollary} \begin{proof} Applying equation \eqref{firstordertime} for the case $M=0$ we get at the leading order \begin{equation*}\begin{split} &Q_{0}^{\perp}\E^{-\I t\heff/\epsi}\psi(x)\Omega_{\mathrm{F}}=-\I\epsi\int_{0}^{t}ds\ \E^{\I(s-t)h_{0}/\epsi}h_{2, \mathrm{OD}}\E^{-\I sh_{0}/\epsi}\psi(x)\Omega_{\mathrm{F}}=\\ &=-\frac{\epsi}{\sqrt{2}}\int_{0}^{t}ds\, \E^{\I(s-t)\hp/\epsi}\E^{\I(s-t)\abs{k}/\epsi}\nabla E(x)\cdot \frac{\I\nabla_{x}v_{\sigma}(x, k)}{\abs{k}}\E^{-\I s\hp/\epsi}\psi(x)=\\ &=\frac{\epsi}{\sqrt{2}}\sum_{j=1}^{N}\frac{\hat{\varrho}_{j}^{\sigma}(k)}{\abs{k}^{3/2}}\kappa\cdot\int_{0}^{t}ds\, \E^{\I(s-t)\abs{k}/\epsi}\E^{\I (s-t)\hp/\epsi}\E^{\I k\cdot \vec{x}_{j}}\nabla_{\vec{x}_{j}} E(x)\cdot\\ &\cdot\E^{-\I s\hp/\epsi}\psi(x)=\\ &=\frac{\epsi}{\sqrt{2}}\sum_{j=1}^{N}\frac{\hat{\varrho}_{j}^{\sigma}(k)}{\abs{k}^{3/2}}\kappa\cdot\int_{0}^{t}ds\, \E^{\I(s-t)\abs{k}/\epsi}\E^{\I (s-t)\hp/\epsi}(\E^{\I k\cdot \vec{x}_{j}}-1)\nabla_{\vec{x}_{j}} E(x)\cdot\\ &\cdot\E^{-\I s\hp/\epsi}\psi(x)+\\ &+\frac{\epsi}{\sqrt{2}}\sum_{j=1}^{N}\frac{\hat{\varrho}_{j}^{\sigma}(k)}{\abs{k}^{3/2}}\kappa\cdot\int_{0}^{t}ds\, \E^{\I(s-t)\abs{k}/\epsi}\E^{\I (s-t)\hp/\epsi}\nabla_{\vec{x}_{j}} E(x)\E^{-\I s\hp/\epsi}\psi(x)=\\ &=\frac{\epsi}{\sqrt{2}}\sum_{j=1}^{N}\frac{\hat{\varrho}_{j}^{\sigma}(k)}{\abs{k}^{3/2}}\kappa\cdot\int_{0}^{t}ds\, \E^{\I(s-t)\abs{k}/\epsi}\E^{\I (s-t)\hp/\epsi}(\E^{\I k\cdot \vec{x}_{j}}-1)\nabla_{\vec{x}_{j}} E(x)\cdot\\ &\cdot\E^{-\I s\hp/\epsi}\psi(x)+\\ &-\E^{-\I t h_{0}} \frac{\epsi}{\sqrt{2}}\sum_{j=1}^{N}\frac{\hat{\varrho}_{j}^{\sigma}(k)}{\abs{k}^{3/2}}\kappa\cdot\int_{0}^{t}ds\, \E^{\I s\abs{k}/\epsi}\opw(\ddot{\vec{x}}_{j}(s; x, p))\psi(x)+\\ &+\Or(\epsi^{2}\abs{t})_{\mathcal{L}(\hilbert{H})}\norm{\psi}_{L^{2}(\field{R}{3n})}, \end{split} \end{equation*} where we have used Egorov's theorem to approximate $\E^{\I s\hp/\epsi}\nabla_{x_{j}} E(x)\E^{-\I s\hp/\epsi}$ (see, e. g., \cite{Ro}). To end the proof we have to show that the norm of the first term is small. For the $j$th term in the sum we get \begin{equation*}\begin{split} & \norm{\cdots}^{2}_{\hilbert{H}}\leq \frac{\epsi^{2}}{2}\int dx\, \int dk\, \frac{\abs{\hat{\varrho}_{j}^{\sigma}(k)}^{2}}{\abs{k}^{3}}\bigg\lvert\int_{0}^{t}ds\, \E^{\I s\abs{k}/\epsi}\E^{\I s\hp/\epsi}(\E^{\I k\cdot \vec{x}_{j}}-1)\cdot\\ &\cdot\nabla_{\vec{x}_{j}}E(x)\E^{-\I s\hp/\epsi}\psi(x)\bigg\rvert^{2}=\\ &=\frac{\epsi^{2}}{2}\int dx\, \int dk\, \frac{\abs{\hat{\varrho}_{j}^{\sigma}(\epsi k)}^{2}}{\abs{k}^{3}}\bigg\lvert\int_{0}^{t}ds\, \E^{\I s\abs{k}}\E^{\I s\hp/\epsi}(\E^{\I \epsi k\cdot \vec{x}_{j}}-1)\nabla_{\vec{x}_{j}}E(x)\cdot\\ &\cdot\E^{-\I s\hp/\epsi}\psi(x)\bigg\rvert^{2}\leq\\ &\leq \frac{\epsi^{2}\abs{t}}{2} \int dk\, \frac{\abs{\hat{\varrho}_{j}^{\sigma}(\epsi k)}^{2}}{\abs{k}^{3}}\int_{0}^{t}ds\, \bigg\lVert(\E^{\I \epsi k\cdot \vec{x}_{j}}-1)\nabla_{\vec{x}_{j}}E(x)\cdot\\ &\cdot\E^{-\I s\hp/\epsi}\psi(x)\bigg\rVert^{2}_{L^{2}(\field{R}{3N})}\leq\\ &\leq C\epsi^{4}\abs{t} \int_{\sigma}^{\Lambda/\epsi} d\abs{k}\, \frac{1}{\abs{k}}\int_{0}^{t}ds\, \bigg\lVert \vec{x}_{j}\nabla_{\vec{x}_{j}}E(x)\E^{-\I s\hp/\epsi}\psi(x)\bigg\rVert^{2}_{L^{2}(\field{R}{3N})}=\\ &=C\epsi^{4}\abs{t} \log\bigg(\frac{\Lambda}{\epsi\sigma(\epsi)}\bigg)\int_{0}^{t}ds\, \bigg\lVert \vec{x}_{j}\nabla_{\vec{x}_{j}}E(x)\E^{-\I s\hp/\epsi}\psi(x)\bigg\rVert^{2}_{L^{2}(\field{R}{3N})}\quad . \end{split} \end{equation*} \end{proof} \begin{remark}\label{remarkradiatedpiece} The norm squared of the leading part of the radiated piece is given by \begin{equation*}\begin{split} & \frac{2\epsi^{2}}{3\pi}\int dx \int_{0}^{\infty} dk\, \frac{\abs{\hat{\varphi}_{\sigma}(k)}^{2}}{k}\bigg\lVert \opw\bigg(\int_{0}^{t}ds\, \E^{-\I sk/\epsi}\sum_{j=1}^{N}\ddot{\vec{x}}^{c}_{j}(s; x,p)\bigg)\psi(x)\bigg\rVert_{\field{C}{3}}^{2}\geq\\ &\geq \frac{\epsi^{2}}{12\pi^{4}}\int dx \int_{\sigma}^{\Lambda} dk\, \frac{1}{k}\bigg\lVert \opw\bigg(\int_{0}^{t}ds\, \E^{-\I sk/\epsi}\sum_{j=1}^{N}\ddot{\vec{x}}^{c}_{j}(s; x,p)\bigg)\psi(x)\bigg\rVert_{\field{C}{3}}^{2}\geq\\ & \geq \frac{\epsi^{2}}{12\pi^{4}}\int dx \int_{\sigma(\epsi)\epsi^{-1}}^{\Lambda} dk\, \frac{1}{k}\bigg\lVert \opw\bigg(\int_{0}^{t}ds\, \E^{-\I sk}\sum_{j=1}^{N}\ddot{\vec{x}}^{c}_{j}(s; x,p)\bigg)\psi(x)\bigg\rVert_{\field{C}{3}}^{2}\quad . \end{split} \end{equation*} The symbol which appears in the Weyl quantization \begin{equation*} \int_{0}^{t}ds\, \E^{-\I sk}\sum_{j=1}^{N}\ddot{\vec{x}}^{c}_{j}(s; x,p) \end{equation*} is independent of $\epsi$ and for $k=0$ is a non null function, \begin{equation*} \sum_{j=1}^{N}[\dot{\vec{x}}_{j}^{c}(t; x, p) - \dot{\vec{x}}_{j}^{c}(0, x, p)]=\sum_{j=1}^{N}[\dot{\vec{x}}_{j}^{c}(t; x, p) - \vec{p}_{j}], \end{equation*} therefore the corresponding operator will be also non null. Therefore, we expect for a generic state $\psi$ that \begin{equation}\label{conditionpsi} \inf_{0\leq k \leq \Lambda}\bigg\lVert \opw\bigg(\int_{0}^{t}ds\, \E^{-\I sk}\sum_{j=1}^{n}\ddot{\vec{x}}^{c}_{j}(s; x,p)\bigg)\psi(x)\bigg\rVert_{L^{2}(\field{R}{3N})\otimes\field{C}{3}}>0 \quad . \end{equation} Then the norm of the radiated piece will be bigger than \begin{equation*}\begin{split} & \inf_{0\leq k \leq \Lambda}\bigg\lVert \opw\bigg(\int_{0}^{t}ds\, \E^{-\I sk}\sum_{j=1}^{N}\ddot{\vec{x}}^{c}_{j}(s; x,p)\bigg)\psi(x)\bigg\rVert_{L^{2}(\field{R}{3n})\otimes\field{C}{3}}\cdot\\ &\cdot\bigg(\frac{\epsi^{2}}{12\pi^{4}} \int_{\sigma(\epsi)\epsi^{-1}}^{\Lambda} dk\, \frac{1}{k}\bigg)^{1/2}=\Or\big(\epsi\log(\epsi\sigma(\epsi)^{-1})\big), \end{split} \end{equation*} which gives a lower bound on the transition of the same order of the upper bound. \end{remark} \begin{corollary} Let \begin{equation*} \omega(t):= \E^{-\I t\heff/\epsi}\omega_{0}\E^{\I t\heff/\epsi}, \end{equation*} where \begin{equation*} \omega_{0}:= \frac{|\qm\tilde{\chi}(\heff)\Psi_{0}\rangle\langle\qm\tilde{\chi}(\heff)\Psi_{0}|}{\norm{\qm\tilde{\chi}(\heff)\Psi_{0}}^{2}_{\hilbert{H}}}, \end{equation*} where $\Psi_{0}$ is any state in $\hilbert{H}$, and let $\omegap$ be the partial trace over the field states \begin{equation*} \omegap(t):= \mathrm{tr}_{\fock}\,\omega(t), \end{equation*} then \begin{equation}\begin{split} & \omegap(t) = \E^{-\I t\hdiagp/\epsi}\omegap(0)\E^{\I t\hdiagp/\epsi} + \Or(\epsi^{2}\abs{t}\sqrt{\log(\sigma(\epsi)^{-1})})_{\hilbert{I}_{1}(L^{2}(\field{R}{3N}))} +\\ &+\Or(\epsi^{2}\abs{t}^{2}\sqrt{\log(\sigma(\epsi)^{-1})})_{\hilbert{I}_{1}(L^{2}(\field{R}{3N}))}, \end{split} \end{equation} where \begin{equation*}\begin{split} &\hdiagp:= \sum_{l=1}^{N}\frac{1}{2m_{l}^{\epsi}}\popb_{l}^{2}+E(x) +\\ &-\frac{\epsi^{2}}{4}\underset{(l_{1}\neq j_{1})}{\sum_{l_{1}, j_{1}=1}^{N}}\int_{\field{R}{3}}dk\, \frac{\hat{\varrho}_{l_{1}}(k)^{*}\hat{\varrho}_{j_{1}}(k)}{\abs{k}^{2}}\big[\E^{\I k\cdot(\vec{x}_{j_{1}}-\vec{x}_{l_{1}})}(\kappa\cdot\popb_{l_{1}})(\kappa\cdot\popb_{j_{1}})+\\ &+(\kappa\cdot\popb_{l_{1}})(\kappa\cdot\popb_{j_{1}})\E^{\I k\cdot(\vec{x}_{j_{1}}-\vec{x}_{l_{1}})}\big], \end{split} \end{equation*} \begin{equation*} \omegap(0):= \mathrm{tr}_{\fock}(\omega_{0}), \end{equation*} and $\hilbert{I}_{1}(L^{2}(\field{R}{3N}))$ denotes the Banach space of trace class operators on $L^{2}(\field{R}{3N})$. \end{corollary} \begin{proof} The proof follows from the following three facts: \begin{enumerate} \item the term of order $\epsi$ in equation \eqref{firstordertime} is off-diagonal with respect to the $\qm$; \item the diagonal Hamiltonian $\hdiag$, defined in \eqref{hdiag}, is equal to $\hdiagp\otimes\mathbf{1}+\mathbf{1}\otimes\hfield$, so we have that \begin{equation*} \mathrm{tr}_{\fock}\big(\E^{-\I t\hdiag/\epsi}\omega_{0}\E^{\I t\hdiag/\epsi}\big)=\E^{-\I t\hdiagp/\epsi}\mathrm{tr}_{\fock}(\omega_{0})\E^{\I t\hdiagp/\epsi}\quad ; \end{equation*} \item the known inequality \begin{equation*} \norm{AB}_{\hilbert{I}_{1}(\hilbert{H})}\leq \norm{A}_{\hilbert{I}_{1}(\hilbert{H})}\cdot\norm{B}_{\mathcal{L}(\hilbert{H})}, \end{equation*} which holds for any Hilbert space $\hilbert{H}$. \end{enumerate} \end{proof} \begin{theorem}\label{densitymatrix} Given a macroscopic observable for the particles, $\opw(a)$, where $a$ is a smooth function bounded with all its derivatives, and a density matrix $\omega\in \hilbert{I}_{1}(P_{M}^{\epsi}\chi(\hepsi)\hilbert{H})$ whose time evolution is defined by \begin{equation} \omega(t):= e^{-\I t\hepsi/\epsi}\omega e^{\I t\hepsi/\epsi}, \end{equation} then \begin{equation}\label{trace}\begin{split} & \tr_{\hilbert{H}}\bigg(\big(\opw(a)\otimes\id_{\fock}\big)\omega(t)\bigg)=\tr_{L^{2}(\field{R}{3N})}\bigg(\opw(a)e^{-\I t H^{(2)}_{\mathrm{D, p}}}\tr_{\fock}\tilde{\omega}(0)e^{\I t H^{(2)}_{\mathrm{D, p}}}\bigg) +\\ &+\Or\big(\epsi^{2}\abs{t}\log(\sigma(\epsi)^{-1})\big) + \Or\big(\epsi^{2}\abs{t}^{2}\log(\sigma(\epsi)^{-1})\big), \end{split} \end{equation} where \begin{equation}\label{omegatilde} \tilde{\omega}(0):= V_{\sigma(\epsi)}^{*}\omega V_{\sigma(\epsi)}\quad . \end{equation} \end{theorem} \begin{proof} First of all we observe that, using proposition \ref{approxsigma} and lemma \ref{cutoffenergysigma}, we have \begin{equation*}\begin{split} & \tr_{\hilbert{H}}\bigg(\big(\opw(a)\otimes\id_{\fock}\big)\omega(t)\bigg) = \tr_{\hilbert{H}}\bigg(\big(\opw(a)\otimes\id_{\fock}\big)e^{-\I t\hepsisigmaepsi/\epsi}\omega_{\sigma(\epsi)}\cdot\\ &\cdot e^{\I t\hepsisigmaepsi/\epsi}\bigg) + \Or(\sigma(\epsi)^{1/2}) + \Or(\sigma(\epsi)\epsi^{-1}), \end{split} \end{equation*} where $\omega_{\sigma(\epsi)}\in \hilbert{I}_{1}(P_{M}^{\epsi}\chi(\hepsisigmaepsi)\hilbert{H})$. By the definition of the effective Hamiltonian and the cyclicity of the trace we have then at the leading order \begin{equation*}\begin{split} & \tr_{\hilbert{H}}\bigg(\big(\opw(a)\otimes\id_{\fock}\big)\omega(t)\bigg)\simeq \tr_{\hilbert{H}}\bigg(\hilbert{U}\big(\opw(a)\otimes\id_{\fock}\big)\hilbert{U}^{*}e^{-\I t\heff/\epsi}\hilbert{U}\omega_{\sigma(\epsi)}\cdot\\ &\cdot \hilbert{U}^{*}e^{\I t\heff/\epsi}\bigg) \quad . \end{split} \end{equation*} The transformed observable, using the definition of $\hilbert{U}$ and lemma \ref{approxchi}, is given by \begin{equation*}\begin{split} & \hilbert{U}\big(\opw(a)\otimes\id_{\fock}\big)\hilbert{U}^{*} = \vsigma^{*}\big(\opw(a)\otimes\id_{\fock}\big)\vsigma + \I\epsi\chi(h_{0})\phisigmaj\cdot\pop\vsigma^{*}\cdot\\ &\cdot\big(\opw(a)\otimes\id_{\fock}\big)\vsigma +\I\epsi\big(\id - \chi(h_{0})\big)\phisigmaj\cdot\pop\chi(h_{0})\vsigma^{*}\big(\opw(a)\otimes\id_{\fock}\big)\vsigma +\\ &-\I\epsi\vsigma^{*}\big(\opw(a)\otimes\id_{\fock}\big)\vsigma\chi(h_{0})\pop\cdot\phisigmaj -\I\epsi\vsigma^{*}\big(\opw(a)\otimes\id_{\fock}\big)\vsigma\big(\id - \chi(h_{0})\big)\cdot\\ &\cdot\pop\cdot\phisigmaj\chi(h_{0}) +\Or(\epsi^{2}\log(\sigma^{-1})) \quad . \end{split} \end{equation*} The operator $\vsigma$ can be considered as the Weyl quantization of the operator valued symbol \begin{equation*} (x, p) \to \E^{-\I\Phi(\I v_{\sigma}(x, \cdot))}\in\mathcal{L}(\fock) \quad. \end{equation*} It is not in the standard symbol classes because the derivative is an unbounded operator, but if we multiply it by $\qm$ we get a smooth bounded symbol. In the calculation of the trace $\vsigma$ is always multiplied by $\qm$ or $Q_{M\pm 1}$, therefore we can proceed as if it were in a standard symbol class. A simple application of the product rule for pseudodifferential operators (equation (A.11) \cite{Te3}) gives then \begin{equation*}\begin{split} & \vsigma^{*}\big(\opw(a)\otimes\id_{\fock}\big)\vsigma = \opw(a)\otimes\id_{\fock} + \vsigma^{*}[\big(\opw(a)\otimes\id_{\fock}\big), \vsigma] =\\ &= \opw(a)\otimes\id_{\fock} - \I\epsi\vsigma^{*}\opw\big(\{a, \vsigma\}\big) + \Or(\epsi^{3})_{\mathcal{L}(\hilbert{H})}, \end{split} \end{equation*} where $\{\cdot, \cdot\}$ denotes the Poisson bracket. We get therefore \begin{equation*} \{a, \vsigma\} = (\nabla_{p}a)\cdot\I\Phi(\I\nabla_{x}v_{\sigma})\vsigma, \end{equation*} and \begin{equation*}\begin{split} & \vsigma^{*}\big(\opw(a)\otimes\id_{\fock}\big)\vsigma = \opw(a)\otimes\id_{\fock} +\epsi\Phi(\I\nabla_{x}v_{\sigma})\big(\opw(\nabla_{p}a)\otimes\id_{\fock}\big) +\\ &+ \Or(\epsi^{2}\log(\sigma^{-1}))\quad . \end{split} \end{equation*} Using this expression we have \begin{equation*}\begin{split} &\hilbert{U}\big(\opw(a)\otimes\id_{\fock}\big)\hilbert{U}^{*} = \opw(a)\otimes\id_{\fock} +\epsi\Phi(\I\nabla_{x}v_{\sigma})\big(\opw(\nabla_{p}a)\otimes\id_{\fock}\big) +\\ &+\I\epsi\chi(h_{0})\phisigmaj\cdot\pop\big(\opw(a)\otimes\id_{\fock}\big)+\I\epsi\big(\id - \chi(h_{0})\big)\phisigmaj\cdot\pop\chi(h_{0})\big(\opw(a)\otimes\id_{\fock}\big) +\\ &-\I\epsi\big(\opw(a)\otimes\id_{\fock}\big)\chi(h_{0})\pop\cdot\phisigmaj -\I\epsi\big(\opw(a)\otimes\id_{\fock}\big)\big(\id - \chi(h_{0})\big)\cdot\\ &\cdot\pop\cdot\phisigmaj\chi(h_{0}) +\Or(\epsi^{2}\log(\sigma^{-1})) \quad . \end{split} \end{equation*} All the terms of order $\epsi$ in the previous expression are off-diagonal with respect to the $\qm$s, and the same holds for the term of order $\epsi$ in \eqref{firstordertime}. Therefore, they all vanish when we calculate the trace. Using point $2$ and $3$ of last corollary we get then \eqref{trace} with \begin{equation*} \tilde{\omega}(0) = \hilbert{U}\omega_{\sigma(\epsi)}\hilbert{U}^{*}. \end{equation*} Using again lemma \ref{cutoffenergysigma} and the fact that the terms of order $\epsi$ in the expansion of $\hilbert{U}$ are off-diagonal we get in the end also \eqref{omegatilde}. \end{proof} \section*{Acknowledgment} We thank Alessandro Pizzo for showing us the strategy to prove proposition \ref{approxsigma} and the German Research Foundation DFG for financial support. \begin{thebibliography}{Ab} \bibitem[AbFr$_1$]{AbFr1} W.\ K.\ Abou-Salem and J.\ Fr\"ohlich. \emph{Adiabatic Theorems and Reversible Isothermal Processes}, Lett.\ Math.\ Phys. {\bf 72}, 153--163 (2005). \bibitem[AbFr$_2$]{AbFr2} W.\ K.\ Abou-Salem and J.\ Fr\"ohlich. \emph{Adiabatic theorems for quantum resonances}, arXiv: math-ph/0607054 (2006). \bibitem[AvEl$_1$]{AvEl1} J.\ E.\ Avron and A.\ Elgart. \emph{Adiabatic Theorem without a Gap Condition}, Commun.\ Math.\ Phys.\ {\bf 203}, 445--463 (1999). \bibitem[AvEl$_2$]{AvEl2} J.\ E.\ Avron and A.\ Elgart. \emph{Smooth adiabatic evolutions with leaky power tails}, J.\ Phys.\ A: Math.\ Gen.\ {\bf 32}, L537--L546 (1999). \bibitem[Ar]{Ar} A.\ Arai. {\em Ground state of the massless Nelson model without infrared cutoff in a non-Fock representation}, Rev.\ Math. Phys. {\bf 13}, 1075--1094 (2001). \bibitem[Ber]{Ber} M.\ V.\ Berry. \emph{Histories of adiabatic quantum transitions}, Proc.\ R.\ Soc.\ Lond.\ A {\bf 429}, 61--72 (1990). \bibitem[Be]{Be} V.\ Betz.\ {\em Gibbs measures relative to Brownian motion and Nelson's model}, Dissertation, TU M\"unchen (2002). \bibitem[Bo]{Bo} F.\ Bornemann. {\em Homogenization in time of singularly perturbed mechanical systems}, Lecture Notes in Mathematics {\bf 1687}, Springer (1998). \bibitem[Da]{Da} E.\ B.\ Davies. {\em Particle-boson interactions and the weak coupling limit}, J.\ Math.\ Phys.\ {\bf 20}, 345--351 (1978). \bibitem[De]{De} J.\ Derezi\'nski. {\em Van Hove Hamiltonians---Exactly Solvable Models of the Infrared and Ultraviolet Problem}, Ann.\ Henri Poincar\'e {\bf 4}, 713--738 (2003). \bibitem[DiSj]{DiSj} M.\ Dimassi and J.\ Sj\"ostrand.\ {\em Spectral Asymptotics in the Semi-Classical Limit}, London Mathematical Society Lecture Note Series {\bf 268}, Cambridge University Press (1999). \bibitem[Fr]{Fr} J.\ Fr\"ohlich. {\em On the infrared problem in a model of scalar electrons and massless scalar bosons}, Ann.\ Ist.\ H.\ Poincar\'e {\bf 19}, 1--103 (1973). \bibitem[FGS]{FGS} J.\ Fr\"ohlich, M.\ Griesemer and B.\ Schlein. {\em Asymptotic Completeness for Rayleigh Scattering}, Ann.\ Henri Poincar\'e {\bf 3}, 107--170 (2002). \bibitem[HeRo]{HeRo} B.\ Helffer and D.\ Robert. {\em Calcul fonctionnel par la transformation de Mellin et op\'erateurs admissibles}, J.\ Funct.\ Anal.\ {\bf 53}, 246--268 (1983). \bibitem[Hi]{Hi} F.\ Hiroshima. {\em Weak coupling limit and removing an ultraviolet cutoff for a Hamiltonian of particles interacting with a quantized scalar field}, J.\ Math.\ Phys.\ {\bf 40}, 1215--1236 (1999). \bibitem[KuSp$_1$]{KuSp1} M.\ Kunze and H.\ Spohn. {\em Slow Motion of Charges Interacting through the Maxwell Field}, Comm.\ Math.\ Phys.\ {\bf 212}, 437--467 (2000). \bibitem[KuSp$_2$]{KuSp2} M.\ Kunze and H.\ Spohn. {\em Post-Coulombian Dynamics at Order $c^{-3}$}, J.\ Nonlinear Sci.\ {\bf 11}, 321--396 (2001). \bibitem[LaLi]{LaLi} L.\ D.\ Landau and E.\ M.\ Lifshitz {\em The classical theory of fields}, 4th rev.\ English ed., Pergamon Press (1975). \bibitem[LMS]{LMS} J.\ L\"orinczi, R.\ A.\ Minlos and H.\ Spohn. {\em Infrared Regular Representation of the Three-Dimensional Massless Nelson Model}, Lett.\ Math.\ Phys.\ {\bf 59}, 189--198 (2002). \bibitem[MaSo]{MaSo} A.\ Martinez and V.\ Sordoni. {\em On the time-dependent Born-Oppenheimer approximation with smooth potential}, Comptes Rendu Math. {\bf 334}, 185--188 (2002). \bibitem[Ne]{Ne} E.\ Nelson. {\em Interaction of Nonrelativistic Particles with a Quantized Scalar Field}, J.\ Math.\ Phys.\ {\bf 5}, 1190--1197 (1964). \bibitem[NeSo]{NeSo} G.\ Nenciu and V.\ Sordoni. {\em Semiclassical limit for multistate Klein-Gordon systems: almost invariant subspaces, and scattering theory}, J.\ Math.\ Phys.\ {\bf 45}, 3676--3696 (2004). \bibitem[Pi]{Pi} A.\ Pizzo. {\em Scattering of an Infraparticle: The One Particle Sector in Nelson's Massless Model}, Ann. Henri Poincar\'e {\bf 6}, 553--606 (2005). \bibitem[PST]{PST} G.\ Panati, H.\ Spohn and S.\ Teufel. {\em The Time-Dependent Born-Oppenheimer Approximation}, to appear in Mathematical Modelling and Numerical Analysis. \bibitem[ReSi$_1$]{ReSi1} M.\ Reed and B.\ Simon. {\em Methods of Modern Mathematical Physics. I:Functional Analysis}, Academic Press (1972). \bibitem[ReSi$_2$]{ReSi2} M.\ Reed and B.\ Simon. {\em Methods of Modern Mathematical Physics. II: Fourier Analysis, Self-Adjointness}, Academic Press (1975). \bibitem[Ro]{Ro} D.\ Robert. {\em Autour de l'Approximation Semi-Classique}, Progress in Mathematics {\bf 68}, Birkh\"auser (1987). \bibitem[So]{So} V.\ Sordoni. {\em Reduction Scheme for Semiclassical Operator-Valued Schr\"odinger Type Equation and Application to Scattering}, Commun.\ Partial Differ.\ Equ.\ {\bf 28}, 1221--1236 (2003). \bibitem[Sp]{Sp} H.\ Spohn. {\em Dynamics of charged particles and their radiation field}, Cambridge University Press (2004). \bibitem[Te$_1$]{Te1} S.\ Teufel. {\em A Note on the Adiabatic Theorem Without Gap Condition}, Lett. Math. Phys. {\bf 58}, 261--266 (2001). \bibitem[Te$_2$]{Te2} S.\ Teufel. {\em Effective $N$-Body Dynamics for the Massless Nelson Model and Adiabatic Decoupling without Spectral Gap}, Ann.\ H.\ Poincar\'e {\bf 3}, 939--965 (2002). \bibitem[Te$_3$]{Te3} S.\ Teufel. {\em Adiabatic Perturbation Theory in Quantum Dynamics}, Lecture Notes in Mathematics {\bf 1821}, Springer (2003). \end{thebibliography} \end{document} ---------------0702191110579 Content-Type: application/postscript; name="bornopp.eps" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="bornopp.eps" %!PS-Adobe-2.0 EPSF-2.0 %%Title: /home/lute/PaperNelson/bornopp.fig %%Creator: fig2dev Version 3.2 Patchlevel 5-alpha5 %%CreationDate: Mon Dec 18 15:43:24 2006 %%For: lute@bell (Luca Tenuta) %%BoundingBox: 0 0 719 442 %Magnification: 1.0000 %%EndComments /$F2psDict 200 dict def $F2psDict begin $F2psDict /mtrx matrix put /col-1 {0 setgray} bind def /col0 {0.000 0.000 0.000 srgb} bind def /col1 {0.000 0.000 1.000 srgb} bind def /col2 {0.000 1.000 0.000 srgb} bind def /col3 {0.000 1.000 1.000 srgb} bind def /col4 {1.000 0.000 0.000 srgb} bind def /col5 {1.000 0.000 1.000 srgb} bind def /col6 {1.000 1.000 0.000 srgb} bind def /col7 {1.000 1.000 1.000 srgb} bind def /col8 {0.000 0.000 0.560 srgb} bind def /col9 {0.000 0.000 0.690 srgb} bind def /col10 {0.000 0.000 0.820 srgb} bind def /col11 {0.530 0.810 1.000 srgb} bind def /col12 {0.000 0.560 0.000 srgb} bind def /col13 {0.000 0.690 0.000 srgb} bind def /col14 {0.000 0.820 0.000 srgb} bind def /col15 {0.000 0.560 0.560 srgb} bind def /col16 {0.000 0.690 0.690 srgb} bind def /col17 {0.000 0.820 0.820 srgb} bind def /col18 {0.560 0.000 0.000 srgb} bind def /col19 {0.690 0.000 0.000 srgb} bind def /col20 {0.820 0.000 0.000 srgb} bind def /col21 {0.560 0.000 0.560 srgb} bind def /col22 {0.690 0.000 0.690 srgb} bind def /col23 {0.820 0.000 0.820 srgb} bind def /col24 {0.500 0.190 0.000 srgb} bind def /col25 {0.630 0.250 0.000 srgb} bind def /col26 {0.750 0.380 0.000 srgb} bind def /col27 {1.000 0.500 0.500 srgb} bind def /col28 {1.000 0.630 0.630 srgb} bind def /col29 {1.000 0.750 0.750 srgb} bind def /col30 {1.000 0.880 0.880 srgb} bind def /col31 {1.000 0.840 0.000 srgb} bind def end save newpath 0 442 moveto 0 0 lineto 719 0 lineto 719 442 lineto closepath clip newpath -92.6 497.0 translate 1 -1 scale /cp {closepath} bind def /ef {eofill} bind def /gr {grestore} bind def /gs {gsave} bind def /sa {save} bind def /rs {restore} bind def /l {lineto} bind def /m {moveto} bind def /rm {rmoveto} bind def /n {newpath} bind def /s {stroke} bind def /sh {show} bind def /slc {setlinecap} bind def /slj {setlinejoin} bind def /slw {setlinewidth} bind def /srgb {setrgbcolor} bind def /rot {rotate} bind def /sc {scale} bind def /sd {setdash} bind def /ff {findfont} bind def /sf {setfont} bind def /scf {scalefont} bind def /sw {stringwidth} bind def /tr {translate} bind def /tnt {dup dup currentrgbcolor 4 -2 roll dup 1 exch sub 3 -1 roll mul add 4 -2 roll dup 1 exch sub 3 -1 roll mul add 4 -2 roll dup 1 exch sub 3 -1 roll mul add srgb} bind def /shd {dup dup currentrgbcolor 4 -2 roll mul 4 -2 roll mul 4 -2 roll mul srgb} bind def /$F2psBegin {$F2psDict begin /$F2psEnteredState save def} def /$F2psEnd {$F2psEnteredState restore end} def $F2psBegin 10 setmiterlimit 0 slj 0 slc 0.06299 0.06299 sc % % Fig objects follow % % % here starts figure with depth 50 % Polyline 0 slj 0 slc 7.500 slw gs clippath 12718 7320 m 12885 7320 l 12885 7260 l 12718 7260 l 12718 7260 l 12838 7290 l 12718 7320 l cp eoclip n 3510 7290 m 12870 7290 l gs col0 s gr gr % arrowhead n 12718 7320 m 12838 7290 l 12718 7260 l 12718 7320 l cp gs 0.00 setgray ef gr col0 s % Polyline gs clippath 3540 1052 m 3540 885 l 3480 885 l 3480 1052 l 3480 1052 l 3510 932 l 3540 1052 l cp eoclip n 3510 7290 m 3510 900 l gs col0 s gr gr % arrowhead n 3540 1052 m 3510 932 l 3480 1052 l 3540 1052 l cp gs 0.00 setgray ef gr col0 s % Polyline 2 slj n 3690 1125 m 3690 1126 l 3690 1129 l 3690 1134 l 3691 1142 l 3691 1155 l 3692 1171 l 3693 1192 l 3695 1219 l 3696 1251 l 3698 1289 l 3700 1333 l 3703 1383 l 3706 1439 l 3709 1502 l 3713 1570 l 3716 1644 l 3720 1723 l 3725 1807 l 3729 1895 l 3734 1988 l 3739 2084 l 3744 2183 l 3750 2284 l 3755 2388 l 3761 2492 l 3767 2598 l 3772 2704 l 3778 2810 l 3784 2916 l 3790 3021 l 3796 3124 l 3802 3226 l 3807 3327 l 3813 3425 l 3819 3522 l 3824 3616 l 3830 3707 l 3835 3796 l 3841 3883 l 3846 3967 l 3851 4049 l 3857 4127 l 3862 4204 l 3867 4277 l 3872 4349 l 3877 4417 l 3882 4484 l 3887 4548 l 3892 4610 l 3896 4670 l 3901 4727 l 3906 4783 l 3911 4837 l 3916 4889 l 3920 4940 l 3925 4989 l 3930 5036 l 3935 5082 l 3940 5127 l 3945 5171 l 3950 5213 l 3955 5255 l 3960 5295 l 3968 5358 l 3977 5419 l 3986 5477 l 3995 5534 l 4004 5590 l 4014 5643 l 4024 5695 l 4034 5746 l 4045 5795 l 4056 5843 l 4067 5889 l 4079 5934 l 4092 5977 l 4105 6019 l 4118 6060 l 4132 6099 l 4147 6136 l 4162 6172 l 4177 6207 l 4193 6240 l 4209 6272 l 4225 6302 l 4242 6330 l 4260 6357 l 4278 6383 l 4296 6407 l 4314 6430 l 4333 6452 l 4352 6473 l 4371 6492 l 4391 6510 l 4411 6527 l 4431 6543 l 4452 6559 l 4472 6573 l 4494 6587 l 4515 6600 l 4537 6613 l 4560 6626 l 4583 6638 l 4609 6651 l 4636 6663 l 4663 6676 l 4692 6688 l 4722 6700 l 4753 6712 l 4784 6724 l 4817 6735 l 4851 6747 l 4886 6758 l 4922 6768 l 4959 6778 l 4997 6788 l 5036 6798 l 5075 6807 l 5116 6816 l 5157 6824 l 5198 6831 l 5240 6838 l 5282 6845 l 5324 6851 l 5366 6856 l 5409 6861 l 5451 6865 l 5493 6868 l 5535 6871 l 5576 6873 l 5617 6874 l 5658 6875 l 5698 6875 l 5738 6875 l 5778 6873 l 5817 6872 l 5856 6869 l 5894 6866 l 5933 6863 l 5969 6858 l 6005 6854 l 6042 6848 l 6079 6843 l 6116 6836 l 6154 6829 l 6192 6822 l 6231 6814 l 6270 6805 l 6310 6796 l 6351 6786 l 6391 6776 l 6433 6765 l 6475 6754 l 6517 6742 l 6559 6730 l 6602 6718 l 6645 6705 l 6689 6693 l 6732 6680 l 6776 6666 l 6819 6653 l 6862 6640 l 6905 6626 l 6948 6613 l 6991 6600 l 7033 6587 l 7075 6574 l 7117 6561 l 7158 6549 l 7199 6536 l 7239 6524 l 7279 6512 l 7319 6501 l 7359 6490 l 7398 6479 l 7438 6468 l 7478 6458 l 7513 6448 l 7550 6439 l 7586 6430 l 7623 6422 l 7661 6413 l 7699 6404 l 7737 6396 l 7776 6388 l 7816 6380 l 7856 6371 l 7897 6363 l 7939 6356 l 7981 6348 l 8024 6340 l 8068 6333 l 8112 6325 l 8156 6318 l 8201 6311 l 8246 6304 l 8292 6297 l 8337 6290 l 8383 6283 l 8429 6277 l 8475 6271 l 8521 6265 l 8567 6259 l 8613 6253 l 8658 6248 l 8704 6242 l 8749 6237 l 8793 6232 l 8837 6227 l 8881 6223 l 8925 6218 l 8968 6214 l 9011 6210 l 9053 6206 l 9095 6202 l 9137 6198 l 9179 6195 l 9221 6191 l 9263 6188 l 9302 6184 l 9343 6181 l 9383 6178 l 9424 6175 l 9465 6172 l 9507 6169 l 9550 6166 l 9594 6163 l 9639 6160 l 9685 6157 l 9733 6154 l 9782 6151 l 9833 6148 l 9885 6145 l 9940 6142 l 9996 6138 l 10055 6135 l 10115 6132 l 10178 6128 l 10242 6125 l 10307 6122 l 10374 6118 l 10442 6115 l 10510 6111 l 10579 6108 l 10647 6104 l 10714 6101 l 10780 6098 l 10843 6095 l 10903 6092 l 10959 6089 l 11011 6086 l 11058 6084 l 11101 6082 l 11137 6080 l 11168 6079 l 11193 6078 l 11213 6077 l 11228 6076 l 11238 6076 l 11245 6075 l 11248 6075 l 11250 6075 l gs col0 s gr % Polyline n 3870 1080 m 3870 1081 l 3871 1084 l 3872 1089 l 3873 1098 l 3875 1110 l 3879 1127 l 3882 1149 l 3887 1176 l 3893 1209 l 3901 1248 l 3909 1293 l 3918 1344 l 3929 1402 l 3941 1465 l 3953 1535 l 3968 1610 l 3983 1691 l 3999 1777 l 4016 1867 l 4033 1962 l 4052 2059 l 4071 2160 l 4091 2264 l 4111 2369 l 4132 2476 l 4152 2584 l 4173 2692 l 4195 2799 l 4216 2907 l 4237 3013 l 4258 3119 l 4279 3222 l 4300 3324 l 4320 3424 l 4341 3521 l 4361 3617 l 4380 3709 l 4400 3799 l 4419 3887 l 4438 3971 l 4456 4053 l 4474 4132 l 4492 4209 l 4510 4282 l 4527 4353 l 4544 4422 l 4561 4488 l 4577 4552 l 4593 4613 l 4610 4672 l 4626 4729 l 4642 4783 l 4657 4836 l 4673 4887 l 4689 4936 l 4704 4983 l 4720 5028 l 4736 5072 l 4751 5115 l 4767 5156 l 4783 5196 l 4799 5235 l 4815 5273 l 4839 5325 l 4862 5376 l 4887 5425 l 4911 5472 l 4936 5517 l 4962 5560 l 4988 5602 l 5015 5643 l 5042 5681 l 5069 5718 l 5098 5754 l 5126 5788 l 5156 5820 l 5185 5851 l 5216 5880 l 5246 5908 l 5278 5935 l 5309 5960 l 5341 5983 l 5374 6005 l 5406 6025 l 5439 6044 l 5472 6062 l 5505 6078 l 5539 6093 l 5572 6107 l 5606 6119 l 5639 6130 l 5672 6140 l 5705 6148 l 5738 6156 l 5771 6163 l 5804 6168 l 5836 6173 l 5868 6177 l 5900 6180 l 5932 6183 l 5964 6185 l 5995 6186 l 6026 6187 l 6057 6188 l 6088 6188 l 6119 6188 l 6150 6188 l 6186 6187 l 6222 6186 l 6258 6185 l 6295 6183 l 6333 6182 l 6371 6180 l 6409 6178 l 6448 6175 l 6488 6173 l 6528 6170 l 6569 6167 l 6610 6163 l 6652 6160 l 6695 6156 l 6738 6152 l 6781 6148 l 6825 6143 l 6869 6139 l 6913 6134 l 6957 6129 l 7001 6124 l 7045 6119 l 7089 6114 l 7133 6109 l 7176 6104 l 7219 6099 l 7262 6094 l 7304 6088 l 7346 6083 l 7387 6078 l 7428 6073 l 7468 6068 l 7508 6062 l 7547 6057 l 7586 6052 l 7625 6047 l 7664 6042 l 7703 6038 l 7739 6033 l 7776 6028 l 7813 6024 l 7850 6019 l 7888 6014 l 7926 6010 l 7965 6005 l 8004 6000 l 8044 5996 l 8084 5991 l 8125 5986 l 8167 5982 l 8209 5977 l 8251 5972 l 8294 5967 l 8337 5963 l 8381 5958 l 8425 5953 l 8470 5949 l 8514 5944 l 8559 5940 l 8604 5935 l 8649 5931 l 8693 5927 l 8738 5922 l 8782 5918 l 8827 5914 l 8871 5910 l 8914 5907 l 8958 5903 l 9000 5900 l 9043 5896 l 9085 5893 l 9127 5890 l 9169 5886 l 9210 5883 l 9252 5881 l 9293 5878 l 9334 5875 l 9375 5873 l 9414 5870 l 9454 5868 l 9494 5865 l 9535 5863 l 9576 5861 l 9618 5859 l 9661 5857 l 9705 5855 l 9751 5853 l 9798 5851 l 9847 5849 l 9897 5847 l 9950 5845 l 10004 5843 l 10061 5841 l 10120 5839 l 10180 5837 l 10243 5835 l 10308 5833 l 10374 5831 l 10442 5828 l 10511 5826 l 10581 5824 l 10651 5822 l 10720 5820 l 10788 5819 l 10855 5817 l 10918 5815 l 10978 5813 l 11034 5812 l 11085 5810 l 11130 5809 l 11170 5808 l 11204 5807 l 11232 5807 l 11254 5806 l 11271 5806 l 11282 5805 l 11289 5805 l 11293 5805 l 11295 5805 l gs col0 s gr % Polyline n 4230 1035 m 4230 1036 l 4231 1039 l 4233 1044 l 4235 1052 l 4239 1064 l 4244 1080 l 4251 1101 l 4259 1127 l 4269 1158 l 4280 1195 l 4294 1238 l 4309 1286 l 4326 1341 l 4345 1401 l 4366 1466 l 4389 1537 l 4413 1612 l 4439 1692 l 4466 1776 l 4494 1863 l 4524 1954 l 4554 2046 l 4585 2141 l 4616 2237 l 4648 2334 l 4681 2432 l 4713 2529 l 4746 2626 l 4778 2722 l 4810 2817 l 4842 2911 l 4874 3003 l 4905 3092 l 4936 3180 l 4966 3266 l 4996 3349 l 5025 3430 l 5054 3508 l 5082 3584 l 5109 3657 l 5136 3728 l 5163 3796 l 5189 3861 l 5214 3925 l 5239 3985 l 5263 4044 l 5287 4100 l 5311 4154 l 5334 4207 l 5357 4257 l 5380 4305 l 5403 4351 l 5425 4396 l 5448 4439 l 5470 4481 l 5492 4521 l 5514 4560 l 5536 4597 l 5558 4633 l 5580 4668 l 5602 4702 l 5603 4703 l 5633 4747 l 5663 4790 l 5694 4831 l 5725 4871 l 5756 4909 l 5788 4946 l 5821 4982 l 5854 5016 l 5887 5049 l 5921 5081 l 5956 5112 l 5991 5141 l 6027 5169 l 6064 5196 l 6101 5222 l 6138 5247 l 6176 5271 l 6215 5293 l 6254 5314 l 6294 5334 l 6334 5353 l 6374 5371 l 6415 5387 l 6456 5403 l 6497 5417 l 6538 5430 l 6580 5442 l 6621 5454 l 6663 5464 l 6705 5473 l 6746 5482 l 6788 5490 l 6829 5496 l 6870 5502 l 6911 5508 l 6952 5513 l 6993 5517 l 7034 5520 l 7075 5523 l 7115 5526 l 7155 5528 l 7196 5530 l 7236 5532 l 7276 5533 l 7317 5534 l 7358 5535 l 7398 5536 l 7440 5537 l 7482 5537 l 7524 5538 l 7567 5538 l 7610 5538 l 7654 5539 l 7699 5539 l 7744 5539 l 7790 5539 l 7837 5539 l 7884 5538 l 7932 5538 l 7981 5537 l 8030 5537 l 8079 5536 l 8129 5536 l 8180 5535 l 8231 5534 l 8282 5533 l 8333 5532 l 8385 5531 l 8436 5530 l 8488 5529 l 8540 5527 l 8591 5526 l 8642 5525 l 8693 5523 l 8743 5522 l 8793 5521 l 8843 5519 l 8891 5518 l 8940 5516 l 8988 5515 l 9035 5513 l 9081 5512 l 9127 5510 l 9172 5509 l 9216 5507 l 9260 5506 l 9303 5504 l 9345 5503 l 9387 5502 l 9429 5500 l 9469 5499 l 9510 5498 l 9554 5496 l 9598 5495 l 9641 5493 l 9685 5492 l 9728 5490 l 9772 5489 l 9816 5488 l 9861 5486 l 9906 5485 l 9953 5483 l 10000 5482 l 10048 5480 l 10098 5479 l 10150 5477 l 10202 5476 l 10256 5474 l 10312 5473 l 10369 5471 l 10428 5469 l 10488 5468 l 10549 5466 l 10610 5464 l 10672 5462 l 10734 5461 l 10795 5459 l 10855 5457 l 10913 5456 l 10968 5454 l 11021 5453 l 11069 5451 l 11113 5450 l 11153 5449 l 11187 5448 l 11217 5447 l 11241 5446 l 11260 5446 l 11274 5446 l 11284 5445 l 11290 5445 l 11294 5445 l 11295 5445 l gs col0 s gr /Times-Roman ff 444.50 scf sf 11340 6435 m gs 1 -1 sc (E) col0 sh gr /Times-Roman ff 444.50 scf sf 11385 5940 m gs 1 -1 sc (E) col0 sh gr /Times-Roman ff 444.50 scf sf 11385 5940 m gs 1 -1 sc (E) col0 sh gr /Times-Roman ff 444.50 scf sf 11385 5490 m gs 1 -1 sc (E) col0 sh gr /Times-Roman ff 254.00 scf sf 11610 6570 m gs 1 -1 sc (1) col0 sh gr /Times-Roman ff 254.00 scf sf 11655 6075 m gs 1 -1 sc (2) col0 sh gr /Times-Roman ff 254.00 scf sf 11655 5580 m gs 1 -1 sc (3) col0 sh gr /Symbol ff 444.50 scf sf 1485 1755 m gs 1 -1 sc (s) col0 sh gr /Times-Roman ff 444.50 scf sf 1800 1710 m gs 1 -1 sc (\(H) col0 sh gr /Times-Roman ff 254.00 scf sf 2295 1800 m gs 1 -1 sc (BO) col0 sh gr /Times-Roman ff 444.50 scf sf 11745 6435 m gs 1 -1 sc (\() col0 sh gr /Times-Roman ff 444.50 scf sf 12150 6435 m gs 1 -1 sc (\)) col0 sh gr /Times-Roman ff 444.50 scf sf 11790 5940 m gs 1 -1 sc (\() col0 sh gr /Times-Roman ff 444.50 scf sf 12195 5940 m gs 1 -1 sc (\)) col0 sh gr /Times-Roman ff 444.50 scf sf 11790 5445 m gs 1 -1 sc (\() col0 sh gr /Times-Roman ff 444.50 scf sf 12195 5445 m gs 1 -1 sc (\)) col0 sh gr /Times-Roman ff 444.50 scf sf 3240 1665 m gs 1 -1 sc (\)) col0 sh gr /Times-Roman ff 444.50 scf sf 2700 1665 m gs 1 -1 sc (\() col0 sh gr /Times-Roman ff 444.50 scf sf 3105 1665 m gs 1 -1 sc (\)) col0 sh gr /Times-Roman ff 571.50 scf sf 11925 6480 m gs 1 -1 sc (r) col0 sh gr /Times-Roman ff 571.50 scf sf 11970 5985 m gs 1 -1 sc (r) col0 sh gr /Times-Roman ff 571.50 scf sf 11970 5490 m gs 1 -1 sc (r) col0 sh gr /Times-Roman ff 571.50 scf sf 2880 1710 m gs 1 -1 sc (r) col0 sh gr /Times-Roman ff 571.50 scf sf 12465 7875 m gs 1 -1 sc (r) col0 sh gr % here ends figure; $F2psEnd rs showpage %%Trailer %EOF ---------------0702191110579 Content-Type: application/postscript; name="spectrum.eps" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="spectrum.eps" %!PS-Adobe-2.0 EPSF-2.0 %%Title: /home/lute/PaperNelson/spectrum.fig %%Creator: fig2dev Version 3.2 Patchlevel 5-alpha5 %%CreationDate: Mon Dec 18 15:43:11 2006 %%For: lute@bell (Luca Tenuta) %%BoundingBox: 0 0 743 500 %Magnification: 1.0000 %%EndComments /$F2psDict 200 dict def $F2psDict begin $F2psDict /mtrx matrix put /col-1 {0 setgray} bind def /col0 {0.000 0.000 0.000 srgb} bind def /col1 {0.000 0.000 1.000 srgb} bind def /col2 {0.000 1.000 0.000 srgb} bind def /col3 {0.000 1.000 1.000 srgb} bind def /col4 {1.000 0.000 0.000 srgb} bind def /col5 {1.000 0.000 1.000 srgb} bind def /col6 {1.000 1.000 0.000 srgb} bind def /col7 {1.000 1.000 1.000 srgb} bind def /col8 {0.000 0.000 0.560 srgb} bind def /col9 {0.000 0.000 0.690 srgb} bind def /col10 {0.000 0.000 0.820 srgb} bind def /col11 {0.530 0.810 1.000 srgb} bind def /col12 {0.000 0.560 0.000 srgb} bind def /col13 {0.000 0.690 0.000 srgb} bind def /col14 {0.000 0.820 0.000 srgb} bind def /col15 {0.000 0.560 0.560 srgb} bind def /col16 {0.000 0.690 0.690 srgb} bind def /col17 {0.000 0.820 0.820 srgb} bind def /col18 {0.560 0.000 0.000 srgb} bind def /col19 {0.690 0.000 0.000 srgb} bind def /col20 {0.820 0.000 0.000 srgb} bind def /col21 {0.560 0.000 0.560 srgb} bind def /col22 {0.690 0.000 0.690 srgb} bind def /col23 {0.820 0.000 0.820 srgb} bind def /col24 {0.500 0.190 0.000 srgb} bind def /col25 {0.630 0.250 0.000 srgb} bind def /col26 {0.750 0.380 0.000 srgb} bind def /col27 {1.000 0.500 0.500 srgb} bind def /col28 {1.000 0.630 0.630 srgb} bind def /col29 {1.000 0.750 0.750 srgb} bind def /col30 {1.000 0.880 0.880 srgb} bind def /col31 {1.000 0.840 0.000 srgb} bind def end save newpath 0 500 moveto 0 0 lineto 743 0 lineto 743 500 lineto closepath clip newpath -50.3 516.7 translate 1 -1 scale /cp {closepath} bind def /ef {eofill} bind def /gr {grestore} bind def /gs {gsave} bind def /sa {save} bind def /rs {restore} bind def /l {lineto} bind def /m {moveto} bind def /rm {rmoveto} bind def /n {newpath} bind def /s {stroke} bind def /sh {show} bind def /slc {setlinecap} bind def /slj {setlinejoin} bind def /slw {setlinewidth} bind def /srgb {setrgbcolor} bind def /rot {rotate} bind def /sc {scale} bind def /sd {setdash} bind def /ff {findfont} bind def /sf {setfont} bind def /scf {scalefont} bind def /sw {stringwidth} bind def /tr {translate} bind def /tnt {dup dup currentrgbcolor 4 -2 roll dup 1 exch sub 3 -1 roll mul add 4 -2 roll dup 1 exch sub 3 -1 roll mul add 4 -2 roll dup 1 exch sub 3 -1 roll mul add srgb} bind def /shd {dup dup currentrgbcolor 4 -2 roll mul 4 -2 roll mul 4 -2 roll mul srgb} bind def /$F2psBegin {$F2psDict begin /$F2psEnteredState save def} def /$F2psEnd {$F2psEnteredState restore end} def $F2psBegin 10 setmiterlimit 0 slj 0 slc 0.06299 0.06299 sc % % Fig objects follow % % % here starts figure with depth 50 % Polyline 0 slj 0 slc 0.000 slw n 1260 3195 m 1260 1215 l 11925 1215 l 11925 3195 l gs 0.75 setgray ef gr % Polyline 2 slj n 1260 3195 m 1261 3195 l 1264 3196 l 1269 3197 l 1277 3200 l 1288 3203 l 1304 3207 l 1324 3213 l 1349 3220 l 1380 3229 l 1415 3239 l 1456 3251 l 1503 3265 l 1555 3280 l 1612 3297 l 1673 3316 l 1740 3336 l 1810 3358 l 1884 3381 l 1962 3405 l 2042 3431 l 2124 3457 l 2208 3485 l 2294 3513 l 2380 3542 l 2466 3572 l 2552 3602 l 2637 3632 l 2722 3663 l 2806 3694 l 2888 3725 l 2969 3756 l 3048 3788 l 3125 3819 l 3200 3851 l 3273 3882 l 3344 3914 l 3413 3946 l 3480 3978 l 3545 4010 l 3607 4042 l 3668 4075 l 3727 4107 l 3784 4141 l 3840 4174 l 3893 4208 l 3946 4242 l 3996 4277 l 4046 4313 l 4094 4349 l 4141 4387 l 4187 4425 l 4233 4463 l 4277 4503 l 4320 4544 l 4363 4586 l 4406 4629 l 4447 4672 l 4448 4673 l 4490 4719 l 4533 4767 l 4575 4816 l 4617 4866 l 4658 4917 l 4700 4969 l 4741 5023 l 4782 5077 l 4823 5132 l 4864 5189 l 4905 5246 l 4946 5304 l 4986 5363 l 5027 5423 l 5067 5484 l 5108 5544 l 5148 5606 l 5189 5668 l 5229 5730 l 5269 5792 l 5310 5854 l 5350 5916 l 5389 5978 l 5429 6039 l 5469 6100 l 5508 6161 l 5547 6221 l 5586 6279 l 5624 6337 l 5662 6394 l 5700 6450 l 5737 6504 l 5774 6557 l 5811 6608 l 5847 6658 l 5883 6706 l 5919 6752 l 5954 6796 l 5988 6838 l 6023 6878 l 6056 6916 l 6089 6952 l 6122 6985 l 6155 7016 l 6187 7045 l 6218 7071 l 6249 7094 l 6280 7116 l 6311 7134 l 6341 7151 l 6371 7164 l 6400 7175 l 6430 7184 l 6459 7189 l 6488 7192 l 6518 7193 l 6547 7190 l 6576 7185 l 6605 7177 l 6634 7166 l 6664 7153 l 6694 7137 l 6724 7119 l 6754 7098 l 6785 7074 l 6816 7048 l 6847 7020 l 6879 6989 l 6911 6956 l 6944 6920 l 6977 6882 l 7010 6842 l 7044 6800 l 7078 6756 l 7113 6709 l 7148 6661 l 7184 6611 l 7220 6559 l 7256 6505 l 7293 6450 l 7330 6394 l 7368 6336 l 7406 6277 l 7444 6217 l 7483 6156 l 7521 6094 l 7561 6032 l 7600 5969 l 7639 5906 l 7679 5842 l 7719 5778 l 7759 5715 l 7799 5651 l 7840 5587 l 7880 5524 l 7921 5461 l 7961 5399 l 8002 5337 l 8043 5276 l 8084 5216 l 8125 5157 l 8166 5098 l 8207 5041 l 8249 4985 l 8290 4929 l 8332 4875 l 8374 4822 l 8416 4770 l 8458 4720 l 8501 4670 l 8544 4622 l 8588 4575 l 8630 4531 l 8673 4488 l 8717 4446 l 8761 4405 l 8807 4365 l 8853 4326 l 8900 4288 l 8948 4251 l 8998 4215 l 9049 4180 l 9101 4145 l 9154 4111 l 9210 4078 l 9267 4045 l 9326 4012 l 9386 3980 l 9449 3949 l 9514 3917 l 9581 3886 l 9650 3855 l 9721 3825 l 9794 3794 l 9870 3764 l 9947 3734 l 10027 3704 l 10109 3674 l 10192 3644 l 10277 3615 l 10364 3585 l 10452 3556 l 10541 3528 l 10630 3499 l 10719 3472 l 10808 3445 l 10897 3418 l 10984 3392 l 11069 3368 l 11152 3344 l 11232 3321 l 11309 3300 l 11382 3280 l 11451 3261 l 11515 3244 l 11574 3228 l 11628 3214 l 11676 3202 l 11719 3191 l 11756 3181 l 11787 3173 l 11813 3166 l 11834 3161 l 11851 3157 l 11863 3154 l 11871 3152 l 11876 3151 l 11879 3150 l 11880 3150 l gs 0.75 setgray ef gr /Symbol ff 476.25 scf sf 6660 662 m gs 1 -1 sc (s) col0 sh gr /Times-Roman ff 476.25 scf sf 6956 662 m gs 1 -1 sc (\() col0 sh gr /Times-Roman ff 476.25 scf sf 7178 662 m gs 1 -1 sc (H) col0 sh gr /Times-Roman ff 317.50 scf sf 7547 810 m gs 1 -1 sc (fib) col0 sh gr /Times-Roman ff 571.50 scf sf 8370 5490 m gs 1 -1 sc (E\() col0 sh gr /Times-Roman ff 571.50 scf sf 9135 5490 m gs 1 -1 sc (\)=inf) col0 sh gr /Times-Roman ff 492.13 scf sf 7920 622 m gs 1 -1 sc (\() col0 sh gr /Times-Roman ff 492.13 scf sf 8423 622 m gs 1 -1 sc (\)) col0 sh gr /Times-Roman ff 492.13 scf sf 8574 622 m gs 1 -1 sc (\)) col0 sh gr /Times-Roman ff 508.00 scf sf 10727 5510 m gs 1 -1 sc (\() col0 sh gr /Times-Roman ff 492.13 scf sf 12248 5482 m gs 1 -1 sc (\)) col0 sh gr /Symbol ff 508.00 scf sf 10406 5510 m gs 1 -1 sc (s) col0 sh gr /Times-Roman ff 508.00 scf sf 10967 5510 m gs 1 -1 sc (H) col0 sh gr /Times-Roman ff 349.25 scf sf 11367 5670 m gs 1 -1 sc (fib) col0 sh gr /Times-Roman ff 492.13 scf sf 11745 5482 m gs 1 -1 sc (\() col0 sh gr /Times-Roman ff 492.13 scf sf 12399 5482 m gs 1 -1 sc (\)) col0 sh gr /Times-Roman ff 571.50 scf sf 8100 630 m gs 1 -1 sc (x) col0 sh gr /Times-Roman ff 571.50 scf sf 11925 5490 m gs 1 -1 sc (x) col0 sh gr /Times-Roman ff 571.50 scf sf 8910 5490 m gs 1 -1 sc (r) col0 sh gr /Times-Roman ff 571.50 scf sf 12015 2115 m gs 1 -1 sc (r) col0 sh gr % here ends figure; % % here starts figure with depth 48 % Polyline 0 slj 0 slc 7.500 slw gs clippath 6555 602 m 6555 435 l 6495 435 l 6495 602 l 6495 602 l 6525 482 l 6555 602 l cp eoclip n 6525 8190 m 6525 450 l gs col0 s gr gr % arrowhead n 6555 602 m 6525 482 l 6495 602 l 6555 602 l cp gs 0.00 setgray ef gr col0 s % Polyline gs clippath 12043 2280 m 12210 2280 l 12210 2220 l 12043 2220 l 12043 2220 l 12163 2250 l 12043 2280 l cp eoclip n 810 2250 m 12195 2250 l gs col0 s gr gr % arrowhead n 12043 2280 m 12163 2250 l 12043 2220 l 12043 2280 l cp gs 0.00 setgray ef gr col0 s % Polyline [60] 0 sd n 12195 3150 m 990 3150 l gs col0 s gr [] 0 sd /Times-Roman ff 412.75 scf sf 6705 3645 m gs 1 -1 sc (e) col0 sh gr /Times-Roman ff 317.50 scf sf 6885 3825 m gs 1 -1 sc (0) col0 sh gr % here ends figure; $F2psEnd rs showpage %%Trailer %EOF ---------------0702191110579--