1: %\documentstyle[12pt,jeep,amsmath,amssymb,amsthm,amsfonts,showlabels]{article}

2: %\documentstyle[12pt,jeep,showkeys]{article}

3: \documentstyle[12pt,jeep]{article}

4: %\usepackage{showlabels}

5: \pagestyle{plain}

6: \addtolength{\topmargin}{-5pc}

7: \addtolength{\textheight}{6pc}

8: \addtolength{\oddsidemargin}{-3pc}

9: %\addtolength{\textwidth}{3pc}

10: \numberbysection

11: \begin{document}

12: \title{ Selection of the Ground State for Nonlinear Schr\"odinger Equations}

13: \author{

14: A. Soffer \thanks{Department of Mathematics, Rutgers University, New

15: Brunswick, NJ} \hspace{.05 in}

16:  and 

17:  M.I. Weinstein \thanks{Mathematical Sciences Research, Bell Laboratories - Lucent

18:  Technologies, Murray Hill, NJ 07974 

19:  }} 

20: \baselineskip=18pt

21: \maketitle

22: %

23: %\renewcommand{\showlabelsfont}{\small\slfamily}

24: \newcommand{\ds}{\displaystyle}

25: \newcommand{\xtext}[1]{\mbox{#1}}

26: \newcommand{\ra}{\rightarrow}

27: \newcommand{\da}{\downarrow}

28: \newcommand{\wra}{\rightharpoonup}

29: \newcommand{\rf}{\widehat}

30: \newcommand{\nit}{\noindent}

31: \newcommand{\no}{\nonumber}

32: \newcommand{\be}{\begin{equation}}

33: \newcommand{\ee}{\end{equation}}

34: \newcommand{\ba}{\begin{eqnarray}}

35: \newcommand{\ea}{\end{eqnarray}}

36: \newcommand{\bom}{\mbox{$\Omega$}}

37: \newcommand{\Fee}{\mbox{$\Phi$}}

38: \newcommand{\del}{\mbox{$\nabla$}}

39: \newcommand{\delx}{\mbox{$\nabla_{x}$}}

40: \newcommand{\dely}{\mbox{$\nabla_{y}$}}

41: \newcommand{\ar}{\mbox{$\alpha$}}

42: \newcommand{\fee}{\mbox{$\varphi$}}

43: \newcommand{\dta}{\mbox{$\delta$}}

44: \newcommand{\lam}{\mbox{$\lambda$}}

45: \newcommand{\Lam}{\mbox{$\Lambda$}}

46: \newcommand{\eps}{\mbox{$\epsilon$}}

47: \newcommand{\gam}{\mbox{$\gamma$}}

48: \newcommand{\Gam}{\mbox{$\Gamma$}}

49: \newcommand{\al}{\mbox{$\alpha$}}

50: \newcommand{\bt}{\mbox{$\beta$}}

51: \newcommand{\va}{\mbox{$\bf a $}}

52: \newcommand{\bF}{\mbox{$\bf F $}}

53: \newcommand{\db}{\mbox{$\parallel$}}

54: \newcommand{\ov}{\overline}

55: \newcommand{\ud}{\underline}

56: \newcommand{\nW}{\left| W\right|_{\bf a}}

57: %\newcommand{\Xeta0}{\mbox{$[\eta_0]_X$}}

58: \newcommand{\D}{\partial}

59: %\newcommand{\px}{\mbox{$\partial_{xx}^{2}$}}

60: \newcommand{\bu}{{\bf u}}

61: \newcommand{\bXz}{{\bf X}_0}

62: \newcommand{\bX}{{\bf X}}

63: \newcommand{\bY}{{\bf Y}}

64: \newcommand{\bN}{{\bf N}}

65: \newcommand{\bR}{{\bf R}}

66: \newcommand{\valpha}{\vec\alpha}

67: \newcommand{\vPsi}{\vec\Psi}

68: \newcommand{\talpha}{\tilde\alpha}

69: \newcommand{\tOmega}{\tilde\Omega}

70: %\newcommand{\teta}{\tilde\eta}

71: \newcommand{\tetab}{\eta_b}

72: \newcommand{\tbeta}{\tilde\beta}

73: \newcommand{\tomega}{\tilde\omega}

74: \newcommand{\tA}{\tilde A}

75: \newcommand{\tE}{\tilde E}

76: \newcommand{\tLambda}{\tilde \Lambda}

77: \newcommand{\tGamma}{\tilde \Gamma}

78: \newcommand{\tchi}{\tilde\chi}

79: \newcommand{\tJ}{\tilde J}

80: \newcommand{\tn}{\tilde n}

81: \newcommand{\ts}{\tilde s}

82: \newcommand{\tp}{\tilde p}

83: \newcommand{\tpsi}{\tilde \psi}

84: \newcommand{\tTheta}{\tilde \Theta}

85: \newcommand{\eoY}{[\eta_0]_{\bf Y}}

86: \newcommand{\tG}{\tilde G}

87: \newcommand{\bT}{{\bf T}}

88: \newcommand{\bA}{{\bf A}}

89: \newcommand{\bAt}{{\bf A(t)}}

90: \newcommand{\bG}{{\bf G }}

91: \newcommand{\bH}{{\bf H }}

92: \newcommand{\bfe}{{\bf e }}

93: \newcommand{\bXYZ}{{\bf XYZ\ }}

94: \newcommand{\bTb}{\overline{\bf T}}

95: \newcommand{\cT}{{\cal T}}

96: \newcommand{\cO}{{\cal O}}

97: \newcommand{\cH}{{\cal H}}

98: \newcommand{\cI}{{\cal I}}

99: \newcommand{\cJ}{{\cal J}}

100: \newcommand{\cK}{{\cal K}}

101: \newcommand{\cN}{{\cal N}}

102: \newcommand{\cA}{{\cal A}}

103: \newcommand{\cB}{{\cal B}}

104: \newcommand{\cC}{{\cal C}}

105: \newcommand{\cD}{{\cal D}}

106: \newcommand{\cE}{{\cal E}}

107: \newcommand{\cF}{{\cal F}}

108: \newcommand{\cG}{{\cal G}}

109: \newcommand{\cQ}{{\cal Q}}

110: \newcommand{\cR}{{\cal R}}

111: \newcommand{\RR}{R^\#}

112: \newcommand{\SS}{S^\#}

113: \newcommand{\cS}{{\cal S}}

114: \newcommand{\cchi}{C_\chi}

115: \newcommand{\cTb}{\overline{\cal T}}

116: \newcommand{\Th}{\Theta}

117: \newcommand{\tH}{{\tilde H}}

118: \newcommand{\bx}{{\bf x}}

119: \newtheorem{theo}{Theorem}[section]

120: \newtheorem{defin}{Definition}[section]

121: \newtheorem{prop}{Proposition}[section]

122: \newtheorem{lem}{Lemma}[section]

123: \newtheorem{cor}{Corollary}[section]

124: \newtheorem{rmk}{Remark}[section]

125: \renewcommand{\theequation}{\arabic{section}.\arabic{equation}}

126: \def\R{{\rm \rlap{I}\,\bf R}}

127: \def\C{{\bf C} \llap{

128:  \vrule height 6.6pt width 0.6pt depth 0pt \hskip 4.7pt}}

129: \def\WB{W\tilde g_\Delta(H)}

130: \def\tg{\tilde g_\Delta (H)}

131: \def\tA{\tilde A}

132: \def\tmu{\tilde \mu}

133: \def\tf{\tilde f}

134: \def\3half{{3\over2}}

135: \def\half{{1\over2}}

136: \def\L2loc{L^2_{\rm loc}}

137: \def\khione{\chi_1}

138: \def\khitwo{\chi_2}

139: \def\Pc{{\bf P_c}}

140: \def\cO{{\cal O}}

141: %\def\RH{R_H}

142: \def\ta0{\tilde\alpha_0}

143: \def\tb1{\tilde\beta_1}

144: \def\t{\tau}

145: \def\nn{\nonumber}

146: \def\ol{\overline}

147: \def\bar{\overline}

148: \def\veps{\varepsilon}

149: \newcommand{\nm}[1]{\vert\vert {#1} \vert\vert}

150: \newcommand{\h}[1]{{#1}}

151: \def\la{\langle}

152: \def\ra{\rangle}

153: \def\jxs{\la x\ra^{\sigma}}

154: \def\jxms{\la x\ra^{-\sigma}}

155: \def\Jt{\la t\ra}

156: \def\jt0{\la t\ra_0}

157: \def\j#1{\la #1 \ra}

158: % local decay norm of \phi_d

159: \def\ld{[\phi_d]_{LD}}

160: \def\Xnm#1{\nm{#1}_X}

161: \def\e#1{e^{#1}}

162: \def\i{\int^t_0}

163: %\centerline{\bf DRAFT OF REVISION}

164: \medskip

165: \date

166: \begin{abstract}

167: \nit 

168: We prove for a class of nonlinear Schr\"odinger systems (NLS)

169: having two nonlinear bound states that the (generic) large time

170: behavior is characterized by decay of the excited state, asymptotic

171: approach to the nonlinear ground state and dispersive

172:  radiation. Our analysis elucidates the mechanism through which

173: initial conditions which are very near the excited state branch evolve

174: into a (nonlinear) ground state, a phenomenon known as {\it ground state 

175:  selection}. 

176:  Key steps in the analysis are the introduction of a particular 

177: linearization and the derivation of a normal form which reflects the dynamics

178: on all time scales and yields, in particular, nonlinear Master equations.

179:  Then, a novel multiple time scale dynamic stability theory 

180:   is developed.

181: Consequently, we give a detailed description of the asymptotic behavior of the 

182: two bound state NLS for all small initial data. The methods are general 

183:  and 

184:  can be extended to treat NLS with more than two bound states

185:  and more general nonlinearities

186: including those of Hartree-Fock type.

187: 

188: %The initial energy in the excited state, to leading order,

189: %is distributed equally between ground state and continuum modes. 

190: %Relaxation time,

191: %polarization time ....

192: \end{abstract}

193: \thispagestyle{empty}

194: \tableofcontents

195: %

196: %\input mbs-todo

197: %

198: %\vfil\eject

199: %

200: \section{Introduction and statement of main results} 

201: 

202: In this paper we study the detailed dynamics of the nonlinear 

203:  Schr\"odinger equation with a potential (NLS)

204: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

205: \footnote{ In a similar way, one can treat nonlinearities $\lambda K[ |\phi|^2]$

206:  is a nonlinear term which can be local or nonlocal. Typical examples are:

207:  $K[|\phi|^2] = |\phi|^p$ and  $K[|\phi|^2] = {\cal K}\star|\phi|^2$,

208:  for some convolution kernel, ${\cal K}$.}:

209: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

210: \be

211: i\D_t\phi = H\phi + \lambda |\phi|^2\phi.\ 

212: \label{eq:nls}\ee

213: Here, $H = -\Delta + V(x)$ is a self-adjoint operator on $L^2(\R^3)$ and 

214: $\lambda$ is a coupling parameter, assumed real and of order one.

215: We assume that $V(x)$ is a smooth potential, 

216:  which decays sufficiently 

217:  rapidly as $|x|$ tends to infinity (short range). Finally, we assume

218: that the operator $H$ has no zero energy resonance \cite{JK,Mu}, a condition

219: which holds for generic $V$.

220: 

221: NLS is a Hamiltonian system with conserved Hamiltonian energy functional:

222: \be

223: \bH_{en}[\phi]\ =\ \int |\nabla\phi(x)|^2 + V(x)|\phi(x)|^2 + 

224:  {\lambda\over2} |\phi(x)|^4 \ dx 

225: \label{eq:Ham}

226: \ee

227: and additional conserved integral

228: \be

229: \bN[\phi] \ =\ \int |\phi(x)|^2\ dx

230: \label{eq:Ndef}\ee

231: These conserved integrals are continuous in the $H^1(\R^3)$ topology.

232: An extensive discussion of the well-posedness theory can be found in the 

233: \cite{Cazenave,Kato,SulemSulem}. In particular, NLS is 

234:   well-posed  globally in time in the space

235:  $H^1(\R^3)$, for initial data, $\phi_0$, 

236:  which is sufficiently small in $H^1$.  

237: \medskip

238: 

239: \nit Throughout this paper we shall assume that $\phi_0$ has sufficiently

240:  small $H^1$ norm. We shall use the notation:

241: \be \cE_0 \ \equiv\ \|\phi_0\|^2_{H^1}\label{eq:cE0def}\ee

242: 

243: For $V(x)$ which decays sufficiently rapidly as $|x|$ tends to infinity (short

244:  range potentials)  the spectrum of $H$ \cite{RS4} 

245:  consists of discrete spectrum, $\sigma_d(H)$, consisting of a  

246:  finite number of negative point eigenvalues, 

247:   and continuous spectrum, 

248: $\sigma_c(H)=[0,\infty)$.

249:  The dynamics of solutions

250:  for $\lambda=0$ (linear Schr\"odinger)

251: is very well understood. Let $\psi_{j*}$ and $E_{j*}$ denote bound states and 

252: bound state energies of the linear Schr\"odinger operator $H$:

253: \be

254: H\psi_{j*}\ =\ E_{j*}\psi_{j*},\ \

255:   \la\psi_{j*},\psi_{k*}\ra_{L^2}\ =\ \delta_{jk}. 

256: \label{eq:linearbs}\ee

257: Arbitrary initial conditions in an appropriate Hilbert

258: space, evolve as $t$ tends to infinity 

259:  into a time-quasiperiodic part consisting of a superposition of

260: time periodic and  spatially localized states with frequencies given by the 

261: eigenvalues and a dispersive or radiative part,

262:   which decays to zero as $t$ tends to infinity 

263: in appropriate spaces, {\it e.g.} $L^p,\ p>2$, $L^2(\la x\ra^{-\sigma}\ dx)$. 

264: 

265:  In order to be more precise, introduce $P_{c*}$,

266:  the orthogonal 

267: projection onto the continuous spectral subspace of $H$:

268: \be P_{c*}f\ =\ f\ -\ \sum_j\ \la\psi_{j*},f\ra\ \psi_{j*}. 

269: \label{eq:Pc*def}

270: \ee

271: The solution of the linear Schr\"odinger equation can be expressed as

272: \be

273:  e^{-iHt}\phi_0\ =\ 

274:  \sum_j\ \la\psi_{j*},\phi_0\ra\ \psi_{j*}e^{-iE_{j*}t}\ +\ e^{-iHt}P_{c*}\phi_0.

275: \label{eq:LSsolution}

276: \ee

277: The time decay of the continuous spectral part of the solution can be expressed,

278:   under suitable

279: smoothness, decay and genericity assumptions on $V(x)$, in terms of

280: {\it local decay estimates} \cite{JK,Mu}:

281: \be

282: \|\ \la x\ra^{-\sigma}  e^{-iHt}P_{c*}\phi_0\ \|_{L^2(R^3)}

283: \ \le\ C\la t\ra^{-{3\over2}}\ \|\ \la x\ra^{\sigma} \phi_0\ \|_{L^2(R^3)},

284: \label{eq:LSlocaldecay}

285: \ee

286: $\sigma\ge\sigma_0>0$, and $L^1-L^\infty$ {\it decay estimates} \cite{JSS,Y}

287: \be

288: \|\ e^{-iHt}P_{c*}\phi_0\ \|_{L^\infty(R^3)}

289: \ \le\ C |t|^{-{3\over2}}\ \|\ \phi_0\ \|_{L^1(R^3)}.

290: \label{eq:LSLpLq} 

291: \ee

292: 

293: 

294: For $\lambda\ne0$ the bound states of the linear problem persist, 

295:  and bifurcate 

296: from the linear states at zero amplitude into branches of {\it 

297:  nonlinear bound states}

298:  \cite{RW88}. Of interest to us is the detailed dynamics of the nonlinear

299: problem on short, intermediate

300: and long time scales and, in particular, the manner

301: in which the nonlinear bound states participate in the dynamics. In \cite{RW88}

302:  variational methods were used to establish the existence and  

303: orbital Lyapunov stability of bound states which are local minimizers 

304:  of $\bH_{en}$ subject to fixed $\bN$; see also 

305:  \cite{MIW86,GSS87}. This result says that 

306: initial data which is close, modulo a phase adjustment, in $H^1$ 

307:  to the ground state remains $H^1$ close to a phase adjusted 

308:  ground state 

309: for all time. The $H^1$ norm is closely related to the conserved

310:  Hamiltonian energy

311: of the system and  

312:    is insensitive to dispersive phenomena. Therefore,

313:  the detailed dynamics is not addressed by this result. For example, 

314: could the large time dynamics consist of a nonlinear ground state

315: plus a small nonlinear excited state part? The main result of this paper

316: implies that this cannot occur. 

317: 

318: For the nonlinear problem, 

319:  the simplest question to consider is the case

320: where $H$ has only one simple eigenvalue and the norm of the solution is small.

321:  The detailed dynamics was studied in 

322:  \cite{SW88,SW9092,PW95,Weder}.

323:   Small norm initial data

324: are shown to evolve into an asymptotic nonlinear ground state and a radiative

325:  decaying part.

326: 

327: In this paper we study the multibound state variant of this question. 

328:  We consider

329: the specific case where $H$ has two simple eigenvalues, $E_{0*}$ and $E_{1*}$.

330: The linear Schr\"odinger equation then has two time-periodic 

331:  solutions $\psi_{0*}e^{-iE_{0*}t}$ and $\psi_{1*}e^{-iE_{1*}t}$, with

332:  $H\psi_{j*}=E_{j*}\psi_{j*},\ \psi_{j*}\in L^2$. Therefore \cite{RW88},

333:   NLS has

334: two branches of nonlinear bound states bifurcating from the zero state 

335: at the eigenvalues of $H$, $\Psi_{\alpha_0}e^{-iE_0t}$ and  

336:  $\Psi_{\alpha_1}e^{-iE_1t}$, with $\Psi_{\alpha_j}\in L^2$ satisfying 

337: \be H\Psi_{\alpha_j}+\lambda|\Psi_{\alpha_j}|^2

338:  \Psi_{\alpha_j} = E_j\Psi_{\alpha_j}. 

339: \label{eq:Psieqn}

340: \ee

341:  Here, $\alpha_j$ denotes a coordinate along 

342: the $j^{\rm th}$ nonlinear bound state branch and  

343: \be E_j = E_{j*} + \cO(|\alpha_j|^2).\nn\ee

344: 

345: In contrast to the linear behavior (\ref{eq:LSsolution})

346:  our main result

347:   is the following:

348: \begin{theo}\label{theo:maintheorem}

349: 

350: Consider NLS with $V(x)$ a short range potential supporting two bound 

351: states as described above. Furthermore, assume that the linear Schr\"odinger

352:  operator, $H$, has no zero energy resonance.

353: % XYZ add more on this

354: 

355: \nit(1) 

356:  Assume the initial data, $\phi(0)$,

357:  is small in the norm defined by:

358: \be

359: [\phi(0)]_X\ \equiv\  \|\la x\ra^\sigma\phi(0)\|_{H^k},

360: \label{eq:triplenorm}

361: \ee

362: where $k>2$ and $\sigma>0$ are sufficiently large.

363:   Let $\phi(t)$ solve the initial value 

364: problem for  NLS. 

365: 

366: Assume the 

367:  (generically satisfied)

368:   {\it nonlinear Fermi golden rule} resonance

369: condition\footnote{The operator $f\mapsto\delta(H-\omega_*)f$ 

370: projects $f$ onto the generalized eigenfunction of $H$ with generalized

371: eigenvalue $\omega_*$. The expression in (\ref{eq:omegastar}) is finite  

372: by local decay estimates (\ref{eq:LSlocaldecay}); 

373:  see {\it e.g.} \cite{SW98}.}

374: \be 

375: \Gamma_{\omega_*} \equiv \lambda^2

376: \pi\left\la\psi_{0*}\psi_{1*}^2,\delta(H-\omega_*)\psi_{0*}\psi_{1*}^2

377:  \right\ra>0\label{eq:omegastar1}\ee

378: holds, where 

379: \be \omega_*=2E_{1*}-E_{0*}>0. \label{eq:omegastardef}\ee

380: 

381: \nit Then, as $t\to\infty$ 

382: \be

383: \phi(t)\ \rightarrow\  e^{-i\omega_j(t)}\Psi_{\alpha_j(\infty)}\ + \ 

384:  e^{i\Delta t}\phi_+, \label{eq:asymp}\ee 

385: in $L^2$, where {\em either} $j=0$ or $j=1$. The phase $\omega_j$ satisfies

386: \be

387: \omega_j(t) = \omega_j^\infty t + \cO(\log t) 

388: \label{eq:log-corr}\ee

389: 

390: \nit Here, $\Psi_{\alpha_j(\infty)}$ is a nonlinear bound state (section

391:  \ref{sec:bs}), with frequency $E_j(\infty)$ near $E_{j*}$.

392: When $j=0$, the solution is asymptotic to

393: a nonlinear ground state, while in the case $j=1$ the solution

394:  is asymptotic to a nonlinear excited state.

395: 

396: \nit(2)

397:  More specifically, 

398: we have the following expansion of the solution $\phi(t)$:

399: \ba

400: \phi(t) &=& e^{-i\tomega_0(t)}\Psi_{\alpha_0(t)} +

401:  e^{-i\tomega_1(t)}\Psi_{\alpha_1(t)}\nn\\ 

402:  &+& \pi_1 e^{-i\cH_0(\infty)t}P_c\tilde\phi_+(t) + R_{loc}(t) + R_{nloc}(t),

403: \nn\ea

404: where as $t\to\infty$

405: \be \tomega_j(t) - \omega_j(t) \to 0, \alpha_j(t)-\alpha_j(\infty)\to0

406: \nn\ee

407: and such that for each initial state, $\phi(0)$,

408: \ba

409: |\alpha_0(\infty)|\cdot |\alpha_1(\infty)|\

410:  &=&\ 0,\nn\\ 

411:  \tilde\phi_+(t)\ &\to& \  \phi_+\ \ {\rm in }\ L^2,\nn\\

412: \| R_{loc}(\cdot,t)\|_2 \ &=&\ \cO(t^{-{1\over2}})\nn\\

413: \| R_{nloc}(\cdot,t)\|_\infty\ &=&\ \cO(t^{-{1\over2}}).\nn

414: \ea

415: Here, $\cH_0(\infty)$ is a small spatially 

416:  localized perturbation of the operator

417: $$\sigma_3(-\Delta + V(x))$$

418: and $P_c=P_c(\cH_0(\infty))$, 

419:   the projection onto its continuous spectral part. Finally, 

420:  $\pi_1$ maps the vector

421:  $(z_1,z_2)$ to $(z_1,0)$.

422: 

423: %\nit\bXYZ Furthermore, the case $E_0(\infty)$ is nongeneric.

424: \end{theo}

425: \bigskip

426: 

427: \begin{rmk}

428: 

429: \smallskip

430: 

431: \nit (A) 

432:  Theorem \ref{theo:maintheorem} implies the absence of small norm 

433:   time - quasiperiodic solutions for this class of nonlinear Schr\"odinger 

434: equations \cite{sigalcmp}. Intuitively, one can explain why

435: one expects only a {\it  pure state} in the limit $t\to\infty$ and

436: how the condition on $\omega_*=2E_{1*}-E_{0*}$ arises. Our intuition is based

437: on viewing the nonlinearity as linear time-dependent potential;

438:  see also \cite{sigalcmp,sigjapan}.

439:   An approximate superposition of a nonlinear ground

440:  state and excited state $\phi \sim

441:  \Psi_{\alpha_0}e^{-iE_0t} + \Psi_{\alpha_1}e^{-iE_1t} $ can be viewed as

442:   defining a self-consistent time-dependent potential:

443: \ba W(x,t) &=&

444:  \lambda|\Psi_{\alpha_0}e^{-iE_0t} + \Psi_{\alpha_1}e^{-iE_1t}|^2,\nn\\

445:  &=& \lambda|\Psi_{\alpha_0} + \Psi_{\alpha_1}e^{-i(E_1-E_0)t}|^2\nn\\

446:  &\sim& \lambda |\Psi_{\alpha_0}|^2 +

447:  2\lambda \cos((E_1-E_0)t + \gamma)\Psi_{|\alpha_0|}\Psi_{|\alpha_1|}

448: \label{eq:selfpotential}

449: \ea

450: for $|\alpha_1|<<|\alpha_0|$.

451: As shown in \cite{SW98jsp}, \cite{KW01}, \cite{KW02}

452:   data with initial conditions given

453: by the unperturbed excited state decay exponentially on a time scale of

454: order $\tau \sim \cO(|\alpha_0\alpha_1|^{-2})$ provided the forcing frequency,

455:  $E_1-E_0\sim E_{1*}-E_{0*}> -E_{1*}$ or $\omega_*=2E_{1*}-E_{0*}>0$.

456: 

457: \smallskip

458: 

459: \nit (B)

460:  Theorem \ref{theo:maintheorem}

461:   implies asymptotic stability and selection of the ground state

462: for generic small data.  Part 1 of Theorem \ref{theo:maintheorem} 

463:  implies a form of asymptotic completeness. 

464: 

465: \nit (C) Since we control the decay of solutions in $W^{k,\infty}$, our results

466:  imply global existence of small solutions in $H^s$ for all $s$ sufficiently 

467:  large. 

468: 

469: \nit (D)  The asymptotic state where $|\alpha_1(\infty)|\ne0$ (and therefore 

470:  $|\alpha_0(\infty)|=0$) is non-generic. This can be seen by linearization about 

471: the excited state. 

472:  The linearized operator, $\cH_1$, is a localized perturbation

473: of an operator having embedded eigenvalues in its continuous spectrum, 

474:  under our 

475:  hypothesis $\omega_*\equiv2E_{1*}-E_{0*}>0$.

476:  The connection between embedded eigenvalues in the continuous spectrum

477:  of an appropriate

478: linear operator and 

479:  the non-persistence of localized time periodic states

480: and between embedded eigenvalues in the continuous spectrum

481:  of an appropriate

482: linear operator was explored first in \cite{sigalcmp,sigjapan}.

483:  It is well known that embedded eigenvalues 

484:  in the continuous spectrum are unstable to generic

485:  perturbations; see, for example,

486:  \cite{CFKS,Grillakis,SW98}. 

487:  In this case, the embedded eigenvalues are perturbed to  complex

488: eigenvalues, with corresponding eigenstates whose evolution is 

489:   exponentially {\em growing} with time, under the condition 

490:  (\ref{eq:omegastar}). The perturbation to the linear operator with embedded

491:  eigenvalues is however both nongeneric (in that it comes from 

492: linearization of a Hamiltonian nonlinear term about a critical point

493: of the energy) and breaks self-adjointness 

494: with respect to the standard $L^2$ inner product. A second order perturbation

495:  theory calculation shows that if $\omega_*\equiv 2E_{1*}-E_{0*}>0$ 

496:  generically the embedded eigenvalue perturbs

497: to an exponential instability \cite{SW95notes}. This suggests the existence 

498: of an unstable manifold of solutions for the {\em nonlinear} equation. 

499:  The existence of such non-generic solutions 

500:  of NLS with $E_1(\infty)\ne0$ for the full nonlinear flow has recently 

501: been demonstrated \cite{TY01b}.

502: 

503: \nit (E) Theorem \ref{theo:maintheorem} is stated for the case of 

504:  two nonlinear

505: bound state branches. The technique  of proof, however, can be used to 

506: consider the more general case. We expect results which are analogous to 

507: those of our main theorem, but more complicated due to the presence of:

508:  direct bound state - bound state interactions, bound state - continuum 

509: interactions {\bf and} bound state - bound state interactions mediated 

510: by the continuum. Multimode Hamiltonian systems have been considered 

511: in the context of linear time almost periodic perturbations have been

512:  studied in \cite{KW02}.  

513: \end{rmk}

514: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

515: 

516: 

517: \nit{\bf Relation to other work}

518:  

519: \nit We now wish to further put our results in context.

520: Research on nonlinear scattering in the presence of bound states has

521: followed two related lines.

522: 

523: \nit{\bf A. Nonlinear dispersive waves in systems with defects, potentials,...}

524: 

525: \nit Our analysis centers around nonlinear

526: bound states which bifurcate from linear bound states of the operator, $H$,

527: obtained by linearizing about the zero solution. These bound states exist

528: at all sufficiently small amplitudes (measured in any $H^s$ norm, $s\ge0$).

529: The behavior of the ``bifurcation diagram'' for larger amplitudes depends

530: in a detailed way on the details of the nonlinearity, the spatial dimension

531: and the norm \cite{RW88}. Such nonlinear bound states are also called 

532: nonlinear defect modes, nonlinear localized modes or nonlinear pinned modes.

533: They are localized about or ``pinned'' to the support of the potential, 

534:  $V$, and  

535:  arise due to a  local deviation  from translation invariance

536: or a ``defect'' in the homogeneous background which acts as 

537: an attractive potential well. 

538: To get a more refined picture of the dynamics

539: than in the $H^1$ theory, one must consider the linearized evolution

540:   about the family of 

541: nonlinear ground states. This linearized operator has continuous   

542: and discrete spectral parts inherited from the linear bound state

543:   spectral structure. 

544:  In particular, the discrete spectrum contains an eigenvalue at zero

545:  corresponding to the ground state

546: and a pair of

547: eigenvalues (located symmetrically about zero) 

548:  corresponding to the excited state. Thus, at linear order a solution

549: infinitesimally close to the ground state formally appears to be 

550:  quasiperiodic in time - a ground state plus a small excited state oscillation.

551: However, at higher order in perturbation theory one finds nonlinear resonant

552: coupling of the neutral oscillatory 

553:  modes to the continuum and as a result these slowly

554: damp to zero; generically, 

555:  for very large time energy splits between the ground state and  dispersive

556: parts of the solution. This mechanism for relaxation to the ground state was

557: earlier considered 

558:  for the nonlinear Klein-Gordon wave equation with a potential,

559: where the decay of ``breather-like'' solutions was studied \cite{SW99}. 

560: In this work, small norm solutions relax to the zero solution via resonant

561: energy transfer out of the bound state to radiation modes and dispersive radiation

562: of energy to infinity;

563:   the zero solution plays the role of the ground state.

564:  Results concerning special classes of initial data are 

565: considered in work of S. Cuccagna \cite{Cucc2} and T.-P. Tsai and H.-T. Yau

566:  \cite{TY00,TY01a}. Similar results to those of Theorem \ref{theo:maintheorem} 

567:  are considered in \cite{TY02}. 

568: 

569: 

570: \nit {\bf B. Nonlinear dispersive translation invariant equations:} 

571:  A closely related line of research focuses on the translation invariant

572:  nonlinear Schr\"odinger equation. Here the equation is (\ref{eq:nls}) with

573:   $V$ taken to be  identically zero nonlinear coupling  parameter

574:  $\lambda<0$. In this case, the equation has solitary

575: wave solutions, obtainable by minimization of ${\bf \cH_{\rm en} }[\phi]$ 

576: subject to ${\bf \cN}[\phi]=\cN_0$. For NLS in dimension $n$ with cubic 

577:  nonlinearity replaced by the general power nonlinearity $|\phi|^{p-1}\phi$,

578: we have that if $p<1+{4\over n}$, the foregoing variational problem has a

579: unique (up to translation) radially symmetric ground state solution for 

580: any $\cN_0>0$. In the case when $V(x)\ne0$ is a potential supporting 

581: bound states, the small $\cN_0$ solutions agree with the  bifurcating bound

582: states discussed above \cite{RW88}. 

583:  As pointed out earlier, constrained energy minimizers are

584:  $H^1$ orbitally Lyapunov stable \cite{CL82,MIW86}.

585: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

586: %\footnote{

587: %{\bf XYZ to include or not to include}

588: %The variational problem for cubic nonlinear ($p=3$) in three space dimensions

589: %($n=3$) does not have a ground state in the above sense. It does, however, have

590: %a

591: % postive radial solitary wave solution, which is however unstable \cite{GSS87}.

592: %In fact, this particular NLS is such that generic finite energy solutions 

593: % develop

594: %singularities in finite time; see \cite{SulemSulem} and 

595: % references cited therein.

596: %Note however, that the nonlinear Hartree equation with {\it nonlocal}

597: % nonlinear term {\bf XYZ} does have a stable ground state and global in time

598: % well-posedness theory in three space dimensions {\bf XYZ citations}.}

599: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

600: An interesting feature of  solitary waves in the translation invariant case

601:  is the

602: presence spurious neutral oscillations. These are sometimes called 

603:  {\it internal modes} \cite{Kivshar}. To explain this, consider the 

604:  linearization about a 

605: ground state solitary wave ($p<1+{4\over n}$). 

606:  Due to the underlying symmetric group of the 

607: equation (translation invariance, phase invariance, Galilean invariance,..)

608:  this linearization has a (generalized) zero eigenvalue of multiplicity 

609: related to the dimension of equation's symmetry group. 

610: In the non-integrable cases ($n=1, \ p\ne3$ and $n\ge2$) the linearization

611: has additional neutral modes. These neutral modes approach

612: zero as $p$ approaches $1+{4\over n}$; 

613:  the dimension of the zero subspace jumps by two 

614: at $p=1+{4\over n}$, the critical case, corresponding to the larger group of 

615: symmetries and the existence of a pseudo-conformal invariant \cite{MIW85}.

616:   V.S. Buslaev

617:  and G. Perel'man \cite{BP2} considered the problem in one space dimension

618:  and showed that nonlinear resonance of these ``internal modes'' with the 

619: continuum is responsible for their damping on long time scales and the

620:  asymptotic stability of solitons. See also the recent work of V.S. Buslaev and 

621:  C. Sulem \cite{BS03}. Their analysis was restricted to one space 

622: dimension only in their use of explicit eigenfunction expansion methods 

623: to obtain the required local energy decay estimates. 

624:  S. Cuccagna \cite{Cucc1,Cucc2}

625:  extended 

626: their results to more general nonlinearities and general space dimensions. 

627: In his analysis, the required dispersive estimates are obtained by adapting

628: K. Yajima's \cite{Y} approach in which  the wave operators, which conjugate

629: the linearized operator on its continuous spectral part to the constant

630: coefficient ``free'' dispersive evolution, are  shown to be bounded on $W^{k,p}$

631: spaces. This method was also used in \cite{SW99}. 

632: 

633: Another feature, common to problems of type {\bf A.} and {\bf B.} is the use

634: of the method of normal forms. In the context of nonlinear scattering, 

635:  normal form ideas were used to obtain the local behavior

636:   in a neighborhood of a soliton in 

637:  \cite{BP2} and for the decaying breather-like state in \cite{SW99}. 

638:  In contrast 

639:  to the normal form for finite dimensional Hamiltonian systems, resonant interaction

640: with the {\it continuous} spectrum gives rise to a more general normal form

641:  which captures internal damping, due to  energy transfer  out of certain discrete

642: modes to the continuum modes; 

643: see the discussion in the introduction to \cite{SW99}. In the present work,

644:  we derive a 

645:  {\it nonlinear master equation},

646: %\footnote{{\bf XYZ} some motivation

647: %for this term??}, 

648: coupled equations for the renormalized (up to near identity

649: transformations on the complex discrete mode amplitdues) discrete mode square

650: amplitudes ("mode powers"),

651:  which governs generic dynamics on large intermediate and very

652:  long time scales.  Normal forms of this type, expected to be valid for 

653: very long times, were derived and studied 

654:  in the local analysis

655: about the steady and `` wobbling '' kink-like solutions of discrete 

656:  nonlinear wave equations in \cite{KeWe00}. 

657: A key feature of the normal form of the current work is our analysis  

658: of its behavior on different time scales and the analysis

659:  of its transitional behavior across time scales, for general initial data.

660:    

661: %XYZ

662: %\nit{\bf The future: Scattering theory, scattering in the presence of defects}

663: %Comments .... References -- physics, etc 

664: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

665: 

666: Finally, we point out that there 

667:  are  many important areas of application which motivate the 

668:  for study of the class of

669:  models we treated in this paper. We mention two.

670:   At the most fundmanental level the nonlinear

671: Schr\"odinger equation arises as a mean field limit model governing the

672: interaction of a very large number of weakly interacting bosons \cite{Hepp},

673: \cite{Spohn}, \cite{LSY99}, \cite{FTY}. At a macroscopic level, it has

674: been shown that equations of this type arise as the equation governing

675:  the evolution of the envelope of the

676:  electric field of a light pulse propagating in a medium

677: with defects. See, for example,

678:  \cite{GWH01,GHW01,GSW01}.

679: 

680: \bigskip

681: \noindent{\bf Acknowledgments:} This work was supported in part by grants

682:  from the National Science Foundation.

683: 

684: \section{Structure of the proof}

685: 

686: We now sketch our analysis. Certain notation is defined in the appendix of 

687: section \ref{sec:notation}.  In analogy with the approach

688:  introduced in \cite{SW88},\cite{SW9092} in the one bound state case, we 

689: represent the solution in terms of the dynamics of the bound state part,

690:  described through the evolution of the {\it collective coordinates} 

691:  $\alpha_0(t)$ and $\alpha_1(t)$, and a remainder $\phi_2$,

692:   whose dynamics is controlled by a  dispersive equation. 

693: In particular we have

694: \be

695: \phi(t,x) = 

696: e^{-i\int_0^t E_0(s) ds - i\tTheta(t)}\left(\  \Psi_{\alpha_0(t)}

697:  + \Psi_{\alpha_1(t)} + \phi_2(t,x)\ \right)

698: \label{eq:fansatz}

699: \ee

700:  

701: We substitute (\ref{eq:fansatz}) into NLS and use the nonlinear equations 

702:  (\ref{eq:Psieqn}) for

703:  $\Psi_{\alpha_j}$ 

704:  to simplify. Anticipating the decay of the excited state, we center

705: the dynamics about the ground state. 

706:  We therefore obtain for  $\Phi_2 \equiv (\phi_2,\ol{\phi_2})^T$

707:  the equation:

708: \be

709: i\D_t\Phi_2 = {\cal H}_0(t)\Phi_2 + 

710:  {\cal G}(t,x, \Phi_2;\D_t\valpha(t),\D_t\valpha,

711:  \D_t\tTheta(t))

712: \label{eq:Phieqn}

713: \ee

714: where, ${\cal H}_0(t)$ denotes the matrix operator which is the linearization 

715: about the time-dependent nonlinear ground state $\Psi_{\alpha_0(t)}$. 

716:  The idea is that

717: in order for $\phi_2(t,x)$ to decay dispersively to zero 

718:  we must choose $\alpha_0(t)$

719: and $\alpha_1(t)$ to evolve in such a way as to remove all secular resonance

720:  terms from ${\cal G}$. Thus we require,

721: \be P_c({\cal H}_0(t))\Phi_2(t) = 0,\label{eq:PcPhid0}

722: \ee where $P_c({\cal H}_0)$ denotes the continuous spectral projection 

723: of ${\cal H}_0$; see also condition (\ref{eq:phi1inM1}).

724: Since the discrete subspace of ${\cal H}_0(t)$ is four dimensional

725:  (consisting of

726:   a generalized null space of dimension two plus two oscillating 

727: neutral modes) (\ref{eq:PcPhid0}) is equivalent to four 

728:  orthogonality conditions implying to four  

729:  differential equations for $\alpha_0,\alpha_1$ 

730: and their complex conjugates. These equations are coupled to the 

731:  dispersive partial differential equation for $\Phi_2$.  

732: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

733: %%\input mbs-interlude1

734: %% At this stage we have that NLS is equivalent to ....

735: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

736: At this stage we have that NLS is equivalent to a dynamical system

737: consisting of a finite dimensional part (\ref{eq:a0eqn}-\ref{eq:a1eqn}),

738: governing $\valpha_j=(\alpha_j,\ol{\alpha_j}),\ j=0,1$,  

739:  coupled to an infinite dimensional

740: dispersive part governing $\Phi_2$:

741: \ba

742: i\D_t\valpha &=& {\cal A}(t)\valpha + \vec F_\alpha\nn\\

743: i\D_t\Phi_2 &=& {\cH_0}(t)\Phi_2+ \vec F_\phi

744: \label{eq:model}

745: \ea

746: We expect ${\cal A}(t)$ and $\cH_0(t)$ to have limits as $t\to\pm\infty$.

747: Our strategy is to fix $T>0$ arbitrarily large, and to study the dynamics

748: on the interval $[0,T]$. In this we follow the strategy of \cite{BP2,Cucc2}.

749:  We shall rewrite (\ref{eq:model}) as:

750: 

751: \ba

752: i\D_t\valpha &=& {\cal A}(T)\valpha + ({\cal A}(t)-{\cal A}(T))\valpha

753:  + \vec F_\alpha\nn\\

754: i\D_t\Phi_2 &=& {\cal H}(T)\Phi_2+ ({\cH_0}(t)-{\cH_0}(T))\Phi_2+ \vec F_\phi

755: \label{eq:model1}

756: \ea

757: and implement a perturbative analysis about the 

758:  {\it time-independent} reference linear, respectively, matrix and differential,

759:  operators ${\cal A}(T)$ and ${\cH_0}(T)$.

760: 

761: More specifically, we analyze the dynamics of (\ref{eq:model1}) by

762:  using (1) the eigenvalues of ${\cal A}(T)$ to calculate the key

763: resonant terms and (2) together with the dispersive  estimates of

764:  $e^{-i\cH_0(T) t}P_c(T)$ \cite{Cucc1}.

765: 

766: Note also  that $P_c(T)\Phi_2(t)\ne \Phi_2(t)$ because

767: $\Phi_2(t)\in\ {\rm Range}\ P_c(t)\ \ne\ {\rm Range}\ P_c(T)$.

768:   We therefore decompose $\Phi_2$

769: as:

770: \be

771: \Phi_2 = {\rm disc(t;T)}\ +\ \eta,\ \ \eta = P_c(T)\eta\nn\ee

772: where disc$(t;T)$ lies in the discrete spectral subspace of $\cH_0(T)$

773: and show that ${\rm disc(t;T)}$ can be controlled in terms of $\eta$.

774: 

775:  The expected generic behavior of this system is that $\alpha_1$ and 

776:  $\Phi_2$ decay with a rate $t^{-{1\over2}}$. This slow rate actually

777: leads to an equation for $\alpha_1$ with the character 

778:  $\D_t\alpha_1 \sim ( {1\over t}\rho  - \D_t\tTheta ) \alpha_1 + $ integrable

779:  in $t$; see (\ref{eq:tTheta}).

780: Thus, $\tTheta$ is chosen to satisfy $\D_t\tTheta \sim {1\over t}\rho$

781: ensuring that $\alpha_1$ has a limit. In this way a logarithmic correction 

782: to the standard phase arises; see (\ref{eq:log-corr}).  

783: 

784: Next we explicitly factor out the rapid oscillations from $\alpha_1$

785:  and show that, after a near identity

786:  change of variables $(\alpha_0,\alpha_1)\mapsto (\talpha_0, \tbeta_1)$,

787:  that the modified ground and excited state amplitudes satisfy the 

788: system:

789: \ba

790: i\D_t\ta0\ &=&\ (c_{1022}+i\Gamma_{\omega})\

791:  |\tb1|^4\ta0\ +\

792:  F_\alpha[\ta0,\tb1,\eta,t]\nn\\

793: i\D_t\tb1\ &=&\ (c_{1121}-2i\Gamma_{\omega})

794:  |\ta0|^2|\tb1|^2\tb1\

795:  +\ F_\beta[\ta0,\tb1,\eta,t];

796: \label{eq:nf-intro}\ea

797: see Proposition \ref{prop:normalform}.

798: It follows that a  {\it Nonlinear Master Equation}

799: governs 

800:  $P_j=|\talpha_j|^2$, the power in the $j^{th}$ mode: 

801: \ba

802: {dP_0\over dt} &=& 2\Gamma P_1^2 P_0 +  R_0(t)\nn\\

803: {dP_1\over dt} &=& -4\Gamma P_1^2 P_0 +  R_1(t).\label{eq:mastereqn}

804: \ea

805: Coupling to the dispersive part, $\Phi_2$, is through the source terms

806: $R_0$ and $R_1$. The expression "master equation" is used since the role

807: played by (\ref{eq:mastereqn}) is analogous to the role of master equations

808: in the quantum theory of open systems \cite{Davies76}.

809: 

810:  A novel multiscale

811:   Lyapunov argument is implemented in section

812:  \ref{sec:stab} characterizing

813: the behavior of the system (\ref{eq:mastereqn}) coupled to that of 

814:  the dispersive part on short, intermediate and long time scales.

815:  We consider  the system (\ref{eq:mastereqn}) on three time intervals: 

816:   $I_0=[0,t_0]$ (initial phase)\  $I_1=[t_0,t_1]$ (embryonic phase) 

817:  and 

818:   $I_2=[t_1,\infty)$ (selection of the ground state).  

819: 

820: 

821: For $t\ge t_0$, the terms $R_0(t)$ and $R_1(t)$ are shown 

822:  (Proposition \ref{prop:ODE}) to have the form

823: \ba

824: R_0(t)\ &\sim&\ \frac{b_0(t_0,\cE_0)}{\la t\ra^2}\ +\ \rho_0(\cE_0,t)P_0P_1^2 

825: \label{eq:R0rough}\\

826: R_1(t)\ &\sim&\ \frac{b_1(t_0,\cE_0)}{\la t\ra^2}\ +\ 

827:             \delta_m(t)\sqrt{P_0}P_1^m\ +\ \rho_1(\cE_0,t)P_0P_1^2,

828: \label{eq:R1rough}

829: \ea 

830: where 

831: \be b_0=\cO(\la t_0\ra^{-1})\ ,\ \ b_1=\cO(\la t_0\ra^{-{1\over2}})\

832: \label{eq:bjrough}

833: \ee

834:  $R_0(t)$ and $R_1(t)$ have parts which are local in time and nonlocal in 

835:  time. The handling of nonlocal terms is explained in  

836: section \ref{sec:stab}. 

837: 

838: We set (Proposition \ref{prop:propB})

839: \be

840: Q_0\ =\ P_0\ -\ \frac{\tilde{b_0}}{\la t\ra},\ \ 

841: Q_1\ =\ P_1\ +\ \frac{\tilde{b_1}}{\la t\ra}\label{eq:Q0Q1rough}

842: \ee

843: where $\tilde{b_0}$ and $\tilde{b_1}$ are positive and 

844:  satisfy (\ref{eq:bjrough}) as well.

845: \bigskip

846: 

847: \nit{\bf Discussion of time scales}

848: 

849: Estimates  (\ref{eq:R0rough}-\ref{eq:R1rough}) and the definitions

850:  (\ref{eq:Q0Q1rough}) imply an effective finite-dimensional 

851:  reduction to a system of 

852: equations for the ``effective mode powers'': $Q_0(t)$ and $Q_1(t)$, 

853: whose character on different time scales dictates the 

854:  full infinite dimensional  dynamics, in a manner analogous to role

855: of a center manifold reduction of a dissipative system \cite{Carr}.

856: 

857: \nit\underline{\it Initial phase --  $t\in I_0=[0,t_0]$:}

858:   Here, $I_0$ is the maximal interval on which 

859:  $Q_0(t)\le0.$ If $t_0=\infty$, then $P_0(t)=\cO(\la t\ra^{-2})$ and the 

860: ground state decays to zero. 

861:  In this case, we show in section \ref{sec:nongeneric} that the excited

862: state amplitude has a limit as well (with may or  may not be zero). This case 

863: is nongeneric.

864: 

865: \nit\underline{\it Embryonic phase -- $t\in I_1=[t_0,t_1]$:} 

866:  If $t_0<\infty$, then for $t>t_0$:

867: \ba

868: \frac{dQ_0}{dt}&\ge& 2\Gamma' Q_0Q_1^2\nn\\

869: \frac{dQ_1}{dt}&\le& -4\Gamma' Q_0Q_1^2\ +\ \cO(\sqrt{Q_0}Q_1^m). 

870:  \ m\ge4\label{eq:Q0Q1I1}

871: \ea

872: Therefore, $Q_0$ is montonically increasing; {\it the ground state grows}.

873:   Furthermore, if $Q_0$ is small

874: relative to $Q_1$, then

875: \be \frac{Q_0}{Q_1}\ {\rm is\  montonically\  increasing}, \nn\ee 

876: in fact exponentially increasing; {\it 

877:  the ground state grows rapidly relative to the 

878:  excited state.}

879: %\footnote{\bXYZ what about reverse 

880: %inequalities}

881: 

882: \nit\underline{\it Selection of the ground state $t\in I_2=[t_1,\infty)$:}

883:  There exists a time $t=t_1$, $t_0\le t_1<\infty$ at which 

884:  the $\cO(\sqrt{Q_0}Q_1^m)$ term in (\ref{eq:Q0Q1I1}) is dominated

885: by the leading ("dissipative") term. For $t\ge t_1$ we have

886: \ba

887: \frac{dQ_0}{dt} &\ge& 2\Gamma' Q_0Q_1^2\nn\\

888: \frac{dQ_1}{dt} &\le& -4\Gamma' Q_0Q_1^2.  

889:  \label{eq:Q0Q1I2}

890: \ea

891: It follows that $Q_0(t)\to Q_0(\infty)>0$  and $Q_1(t)\to0$

892:  as $t\to\infty$; {\it the ground state is selected.}

893: 

894: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

895: %\footnote{Connections to:

896: %(0) asymptotic resolution into bound states and radiation, 

897: %(.5) $\tilde P_0(\infty) = \tilde P_0 + {1\over2}\tilde P_1(0)$, nonlinear

898: % {\it master equation}

899: %(1) internal modes of nonintegrable systems,

900: %(2) selection of the ground state,

901: %(3) center manifold approach used by Pillet and Wayne for one bound state 

902: %(5) nonlocal operator with very high order resonance??

903: %}

904: %\footnote{\bXYZ - put in later section

905: %{ \bf Hypotheses:}

906: %\nit (FGR)\  $\Gamma_{\omega_*}>0$, where  

907: %\be \omega_* \equiv 2E_{1*}-E_{0*}>0 \label{eq:omegastar}\ee

908: %$\Gamma_{\omega_*}$ is always non-negative and is the generically strictly 

909: %positive; see section 5.

910: %This is a nonlinear  analogue of the  Fermi golden rule. See,  

911: %for example, \cite{SW98}, \cite{SW99}

912: %\nit (H2)\  $H$ has no zero energy resonance 

913: %} 

914: 

915: %

916: \section{ Linear and nonlinear bound states}\label{sec:bs}

917: 

918: In this section we introduce bound states of the linear ($\lambda=0$)

919:  and nonlinear ($\lambda\ne0$) 

920:  Schr\"odinger equation (\ref{eq:nls}).

921: \bigskip

922: 

923: \noindent{\bf Bound states of the unperturbed problem}

924: 

925: Let $H= -\Delta + V(x)$. We assume that $V(x)$ is smooth and

926: sufficiently rapidly decaying, so that $H$ defines a self-adjoint

927: operator in $L^2$. Additionally, we assume that the spectrum of $H$

928: consists of continuous spectrum extending from $0$ to positive infinity

929: and two discrete negative eigenvalues, each of multiplicity one.

930: \be

931: \sigma(H)\ =\ \{E_{0*}, E_{1*}\}\ \cup\ [0,\infty )

932: \label{eq:specH}\nn\ee

933: Therefore, there exist eigenstates, $\psi_{j*}\in {\cal D}(H),\ j=0,1$ such that

934: \be H\psi_{j*} = E_{j*} \psi_{j*}. \label{eq:linearevp}\ee

935: 

936: We also introduce spectral projections onto the discrete eigenstates and 

937: continuous spectral part of $H$, respectively:

938: \ba P_{j*}f\ &\equiv&\ \la\psi_{j*},f\ra\psi_{j*}, \ j=0,1

939: \nn\\

940:     P_{c*}\ &\equiv &  I - P_{0*}- P_{1*}\nn\ea 

941:      

942: \bigskip

943: 

944: \noindent{\bf Nonlinear bound states}

945: 

946: We seek solutions of (\ref{eq:nls}) of the form

947: \be \phi = e^{-iE_jt}\Psi_{E_j}\nn\ee

948: Substitution into (\ref{eq:nls}) yields

949: \be

950: H \Psi_{E_j} + \lambda |\Psi_{E_j}|^2 \Psi_{E_j} = E_j \psi_{E_j}

951: \label{eq:nonlinearevp}\ee

952: We introduce a bifurcation parameter, $\alpha_j$,

953:   for the $j^{th}$ nonlinear bound state branch and define

954: \be \valpha_j = (\alpha_j,\bar\alpha_j)\label{eq:boldalpha}\ee

955: \medskip

956: 

957: \begin{prop} \cite{RW88} \label{prop:bifurcate}

958: For each $j=0,1$ we have a one parameter family, 

959:  $\Psi_{\alpha_j}=\Psi_j$, of   

960: bound states depending on the complex parameter 

961: $\alpha_j=|\alpha_j|e^{i\gamma_j}$ and defined 

962:  for $|\alpha_j|$ sufficiently small:

963: \ba

964:   \Psi_j(x)&\equiv&\ 

965:  \ \alpha_j\psi_j(x;|\alpha_j|^2) 

966:    = e^{i\gamma_j}|\alpha_j|\psi_j(x;|\alpha_j|^2)\nn\\ 

967:   &=& \alpha_j\left(\ \psi_{j*}(x) + 

968:  \lambda |\alpha_j|^2\psi_j^{(1)}(x;|\alpha_j|^2)\ \right)\nn\\

969:  &=& \alpha_j\left(\ \psi_{j*}(x) +

970:  \lambda |\alpha_j|^2\psi_j^{(1)}(x;0) + \cO(\lambda^2|\alpha_j|^4\right).

971:  \label{eq:jbranch}

972: \ea

973: Here, 

974: \ba

975:  \psi_j^{(1)}(x;0) &=& -(H-E_{0*})^{-1}(I-P_{E_{j*}})\psi_{j*}^3

976:  \label{eq:psij10}\\

977: E_j &\equiv&\ E_j(\valpha_j)\ \equiv\  

978:  E_{j*} + |\alpha_j|^2 E_j^{(1)}(|\alpha_j|^2)\nn\\

979:  &=& E_{j*} + \lambda\int\psi_{0*}^4 dx\  |\alpha_j|^2 \ 

980:  +\ {\cal O}(|\alpha_j|^4).

981: \label{eq:Ej_expand}

982: \ea

983: The mapping $\valpha_j\mapsto \left(E_j(\valpha_j),\Psi_j(\cdot;\valpha_j)

984:  \right)$

985:  is smooth.

986: \end{prop}

987:  The proof uses standard bifurcation 

988: theory \cite{Nirenberg}, which is based on the implicit function

989:  theorem. The analysis extends to the case of nonlocal nonlinearities.

990: A variational approach can also be used to construct nonlinear bound 

991:  states. Variational approaches, though more global, do not directly

992:  yield the information we require concerning smooth variation 

993:  with respect to parameters.

994: 

995: \begin{rmk} $\Psi_j$ depends on $\alpha_j$ and $\bar\alpha_j$.

996: We shall  compute derivatives of the nonlinear bound states

997: $\Psi_{\alpha}$ with respect to $\alpha_j$ and $\ol{\alpha_j}$

998: and use the notation:

999: \be \vec\nabla_j = \left(\D_{\alpha_j},\D_{\ol{\alpha_j}}\right)

1000:  \equiv (\D_j,\ol{\D_j}).

1001:  \label{eq:nablaj}\ee

1002: \end{rmk}

1003: 

1004: In what follows we shall ``modulate'' these bound states. That is, we shall

1005: allow $\valpha$ to vary with time. For convenience, we shall use the notation:

1006: \ba

1007: \Psi_j(t,x)\ &=&\ \Psi_{\alpha_j(t)}(x), 

1008: \nn\\

1009: E_j(t)\ &=&\ E_j(|\alpha_j(t)|^2)\nn\ea

1010: \section{Linearization about the ground state}\label{sec:linearization}

1011: 

1012: Let $\Psi$ denote a nonlinear bound state (ground state or excited state)

1013:  of (\ref{eq:nls}); see section \ref{sec:bs}. Then,

1014: \be

1015: H\Psi\ +\ \lambda|\Psi|^2\Psi\ =\ E\ \Psi.

1016: \label{eq:boundstateeqn}

1017: \ee

1018: We first derive the linear stability problem.

1019: Let

1020: \be

1021: \phi\ =\ \left( \Psi\ +\ p \right) e^{-iEt},

1022: \label{eq:lin_pert}

1023: \ee

1024: where $p$ denotes the perturbation about $\Psi$. Substituting  

1025: (\ref{eq:lin_pert}) into (\ref{eq:nls}) and neglecting all terms which 

1026: are nonlinear in $p$ and $\ol{p}$, we obtain the {\it linearized perturbation

1027: equation}

1028: \be

1029: i\D_t p \ =\ \left(\ H-E+2\lambda |\Psi|^2\ \right)\ p +\ \lambda \Psi^2\ \ol{p} 

1030: \label{eq:lin_pert_eqn}

1031: \ee

1032: Since $\ol{p}$ appears explicitly in (\ref{eq:lin_pert_eqn}) it is natural

1033: to consider the system for 

1034: \be

1035: \vec p\ =\ 

1036: \left(\begin{array}{c} p\\ \ol{p} \end{array}\right):

1037: \label{eq:vec_p_def}

1038: \ee

1039: \be

1040: i\D_t\vec p\ =\ {\cal H}\vec p,

1041: \label{eq:vec_p_eqn}\ee

1042: where

1043: \be

1044: {\cal H}\ =\ \sigma_3

1045:  \left(\begin{array}{cc} H-E+2\lambda |\Psi|^2 & \lambda \Psi^2 \\ 

1046:  \lambda \ol{\Psi}^2 & H-E+2\lambda |\Psi|^2  \end{array}\right),

1047: \label{eq:H_def}\ee

1048: where 

1049: \be

1050: \sigma_3 = \left(\begin{array}{cc} 1 & 0 \\ 0 & -1\end{array}\right).

1051: \label{eq:sigma3}

1052: \ee

1053: 

1054: 

1055: Later in this paper we shall refer specifically to the linearization about 

1056:  a ``curve" of  bound states $(\Psi_j(t), E_j(t))$ and will denote by 

1057:  ${\cal H}_j(t)$ the operator (\ref{eq:H_def}) with $E$ replaced 

1058:  by $E_j(t)$ and 

1059:  $\Psi$ replaced by $\Psi_j(t)$.

1060: Our main focus will be on the operator family

1061: \be

1062:  \cH_0(t)\ =\ \cH_{E_0(t),\Psi_0(t)}.

1063: \nn\ee

1064: 

1065: The nonlinear bound state $\Psi$ is {\it linearly spectrally stable}

1066: if the spectrum of ${\cal H}$, $\sigma({\cal H})$, is a subset of the 

1067:  real line

1068: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1069: \footnote{

1070: The spectrum of $\cH$ has the symmetries one expects for Hamiltonian systems.

1071: The mappings $\lambda\mapsto -\lambda$ and $\lambda\mapsto\bar\lambda$ send 

1072:  points in the spectrum to points in the spectrum.

1073: Note that if $\xi$ is an eigenvector of ${\cal H}$ with eigenvalue $\mu$ then

1074:  $\ol{\sigma_1\xi}$ is an eigenvector of ${\cal H}$ with eigenvalue $-\mu$.

1075: Therefore, $\xi_{-\mu}= \ol{\sigma_1\xi_\mu}$. }. 

1076: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1077:  $\Psi$ is {\it linearly dynamically stable} if, in an appropriate

1078:  space, all solutions of the initial value problem for (\ref{eq:vec_p_eqn})

1079:  are bounded in time. That is, in some norm $e^{-i\cH_0t}$ is a bounded 

1080:  operator. Linear dynamical stability of the ground state $\Psi_0$ follows

1081: from \cite{MIW85}. 

1082:  For this result and the necessary stronger dispersive estimates on 

1083:  $e^{-i\cH_0t}$ \cite{Cucc1}, we require information on the 

1084:  discrete spectrum of ${\cal H}_0$ and the corresponding spectral subspaces.

1085: 

1086: Before stating these results we observe that the operator $e^{-i\cH_0t}$ 

1087:  can

1088: be expressed in terms of the operator treated explicitly in \cite{MIW85}, 

1089:  \cite{Cucc1}. 

1090:  To see this,

1091:   express the ground state as $\Psi_0 = |\Psi_0| e^{i\gamma}$ ($|\Psi_0|>0$), 

1092:  where $\gamma=\arg\alpha$ is a constant. 

1093:  Set $p=e^{i\gamma}q$.

1094: Then, by (\ref{eq:lin_pert_eqn}) we have:

1095: 

1096: \be

1097: i\D_t q \ =\ \left(\ H-E_0+2\lambda |\Psi_0|^2\ \right)\ q +\ 

1098:  \lambda \Psi_0^2\ \ol{q}

1099: \label{eq:lin_pert_eqn1}

1100: \ee

1101: Now let $q=u+iv$, where $u$ and $v$ are real. Then,

1102: \be

1103: \D_t \left(\begin{array}{c} u \\ v\end{array}\right)\ = \ J\tilde H_0

1104:  \left(\begin{array}{c} u \\ v\end{array}\right), \nonumber\ee

1105: where

1106: \be

1107: J= \left(\begin{array}{cc} 0 & -1 \\ 1 & 0 \end{array}\right) 

1108: \ {\rm and}\ 

1109:  \tilde H_0=\left(\begin{array}{cc} H_0-E_0-\lambda |\Psi_0|^2  & 0 \\ 

1110:                 0 & H_0-E_0-3\lambda |\Psi_0|^2  \end{array}\right)

1111: \label{eq:J_tH0_def}

1112: \ee

1113: Note that 

1114: \ba p(t)\ &=&\ \pi_1\vec p(t)\ =\ \pi_1 e^{-i{\cal H}_0t}\ \vec p_0\nonumber\\

1115:  &=&\ e^{i\gamma}\left(\ \pi_1e^{J{\tilde H}_0t}\ \Re q_0\ +\ 

1116: \ i\pi_2 e^{J{\tilde H}_0t}\ \Im q_0\right),

1117: \ea

1118: where $\pi_1(z_1,z_2)=(z_1,0)$ and  $\pi_2(z_1,z_2)=(0,z_2)$.

1119: Therefore, we have:

1120: \begin{prop}

1121: Estimates on $e^{-i{\cal H}_0t}$ are equivalent to those for

1122:  $e^{J{\tilde H}_0t}$

1123:  and are independent of $\gamma$.

1124: \end{prop}

1125: 

1126: We now turn to a detailed discussion of the spectral properties of $\cH_0$.

1127: \begin{prop}\label{prop:Hspectrum}

1128: Consider ${\cal H}_0$, the linearization about the ground state. Let 

1129:  $\alpha_0$ be sufficiently small.

1130: 

1131: \noindent (1) $\sigma({\cal H}_0)$ is a subset of the real line.

1132: 

1133: \noindent (2) $\sigma_{\rm discrete}({\cal H}_0)\ = \ \{-\mu, 0,\mu\}$,

1134:  where $0<\mu<|E_0|$.

1135: 

1136: \noindent (3) Zero is a generalized eigenvalue of ${\cal H}_0$.

1137:  The generalized null space, $N_g({\cal H}_0)$, is given by 

1138: \be

1139: N_g({\cal H}_0)\ =\  

1140: {\rm span}\left\{

1141:  \sigma_3\left(\begin{array}{c} \Psi_0\\ \ol{\Psi_0} \end{array}\right)\ ,

1142:  \ \left(\begin{array}{c} \D_{E_0}\Psi_0\\ \ol{\D_{E_0}\Psi_0} \end{array}\right)

1143: \right\}

1144: \label{eq:NgH}\ee

1145: 

1146: \noindent (4) $\pm\mu$ are simple eigenvalues. We denote their corresponding

1147:  eigenfunctions by $\xi_\mu$ and $\xi_{-\mu}$. For $|\alpha_0|$ small 

1148: we have the expansion:

1149: \ba \mu\ &=&\  E_1-E_0 + {\cal O}(|\alpha_0|^2)\label{eq:muexpand}\\

1150:     \xi_\mu\ &=&\ \left(\begin{array}{c} 1 \\ 0\end{array}\right)

1151: \psi_{1*} + 

1152: \left(\begin{array}{c} |\alpha_0|^2c_1(|\alpha_0|^2) \\ 

1153:                        \overline{\alpha_0}^2c_2(|\alpha_0|^2)

1154:  \end{array}\right)

1155: \label{eq:ximu_def}\\

1156:     \xi_{-\mu}\ &=&\ \overline{\sigma_1\xi_\mu}, 

1157: \label{eq:ximinusmu_def}

1158: \ea

1159: where $c_1(a)$ and $c_2(a)$ are real analytic functions in $a$.

1160: 

1161: \noindent (5) $\sigma({\cal H}_0) - \sigma_{\rm discrete}({\cal H}_0)\ =\ 

1162:  (-\infty, E_0] \cup [-E_0,\infty)$.

1163: \end{prop}

1164: The proof of Proposition \ref{prop:Hspectrum} is in the appendix of 

1165:  section \ref{sec:appB}.

1166: \bigskip

1167: 

1168: Note that   if $\omega$ is an eigenvalue of ${\cal H}$:

1169: \be \sigma_3{\cal L}g\ \equiv {\cal H}g\ =\ \omega g \label{eq:evpH}\ee

1170: then $\omega$ is an eigenvalue of ${\cal H}^*$ with corresponding eigenfunction

1171:  $\sigma_3 g$:

1172: \be

1173: {\cal L}\sigma_3 (\sigma_3 g) = {\cal H}^* (\sigma_3 g) = \omega\sigma_3 g.

1174: \label{eq:evpH*}\ee

1175: Therefore, we have

1176: \begin{prop}\label{prop:H*spectrum}

1177: \smallskip

1178: 

1179: \noindent (1) $\sigma({\cal H}_0^*)\ =\ \sigma({\cal H}_0)$

1180: 

1181: \noindent (2) 

1182: \be

1183: N_g({\cal H}_0^*)\ =\

1184: {\rm span}\left\{

1185:  \left(\begin{array}{c} \Psi_0\\ \ol{\Psi_0} \end{array}\right)\ ,

1186:  \

1187:  \sigma_3\left(\begin{array}{c} \D_{E_0}\Psi_0\\ \ol{\D_{E_0}\Psi_0} \end{array}

1188:  \right)

1189: \right\}

1190: \label{eq:NgH*}\ee

1191: 

1192: \noindent (3) 

1193: \be N({\cal H}_0^*\mp \mu) \ =\ {\rm span}

1194:  \left\{\sigma_3 \xi_\mu, \sigma_3\xi_{-\mu}  \right\}

1195:  \equiv \left\{\zeta_\mu, \zeta_{-\mu} \right\}

1196: \ee

1197: Here, 

1198: \be \zeta_\mu=\sigma_3\xi_\mu\ {\rm and}\ \zeta_{-\mu} = -\sigma_3\xi_{-\mu},

1199: \label{eq:zetamu_def}

1200: \ee

1201: where this choice of $\zeta_{-\mu}$ is taken so that 

1202: $\la \zeta_{-\mu},\xi_{-\mu}\ra = 1$ in the $|\alpha_0| \downarrow 0$ limit. 

1203: \end{prop}

1204: \medskip

1205: 

1206: \nit {\bf

1207:  Nondegenerate basis for the discrete subspaces of $\cH_0$ and 

1208:  $\cH_0^*$}

1209: 

1210: For fixed $E\ne E_0$, the basis of $N_g({\cal H}_0^*)$ displayed in 

1211:  (\ref{eq:NgH*}) is a natural basis 

1212: due to its direct connection to the symmetries of NLS. 

1213:  However, this basis is degenerate

1214:  and singular in the limit $E\to E_{0*}$, as we shall now see. 

1215: 

1216: Consider the basis of $N_g({\cal H}_0)$ displayed in (\ref{eq:NgH}).

1217: Beginning with the first element of this basis, explicitly we have:

1218: \be

1219: \sigma_3\left( \begin{array}{c}\Psi_0\\ \ol{\Psi_0} \end{array}\right) 

1220: = \sigma_3\left( \begin{array}{c}\alpha_0\psi_0(\cdot;|\alpha_0|^2)\\ 

1221:                  \ol{\alpha_0} \psi_0(\cdot;|\alpha_0|^2) \end{array}\right) 

1222: = \alpha_0 \sigma_3 G_0\left(\begin{array}{c} 1\\ 1\end{array}\right) F_0,

1223: \nn\ee 

1224: where  

1225: \be \alpha_0 = |\alpha_0|e^{i\gamma_0},\ 

1226:  \ {\bf G_0} = \left( \begin{array}{cc}e^{i\gamma_0} & 0\\ 0 & e^{-i\gamma_0} 

1227:             \end{array}\right),

1228: \label{eq:G0_def}

1229: \ee

1230: and 

1231: \ba F_0 \equiv \psi_0(\cdot;|\alpha_0|^2)

1232:  &=& \psi_{0*}+|\alpha_0|^2\chi(\cdot,|\alpha_0|^2)\nn\\

1233:  &=& \psi_{0*}+|\alpha_0|^2\chi(\cdot,0) + \chi_4^{(0)}.

1234: \label{eq:F0_def}

1235: \ea

1236: Here and subsequently we use the notation $\chi(\cdot,p)$

1237:   to denote a generic 

1238: real-valued localized function of $x$ with smooth dependence on a parameter, $p$,

1239:  and $\chi_k^{(j)}$ is localized in $x$ and $\cO(|\alpha_j|^k)$.

1240: 

1241:   A nonsingular element of the 

1242:  $N_g({\cal H}_0)$ is obtained by dividing out $\rho_0$. We therefore define 

1243: \be \xi_{01} \equiv \sigma_3 G_0\left(\begin{array}{c} 1\\ 1\end{array}\right) 

1244:            F_0(\rho_0^2)\label{eq:xi01_def}

1245: \ee

1246: 

1247: We now turn to the second element of the basis displayed in (\ref{eq:NgH}).

1248: First note that

1249: \ba {\D\over\D |\alpha_0|}\Psi_0 &=&

1250:  e^{i\gamma_0}

1251:   {\D\over\D |\alpha_0|}\left( |\alpha_0|\psi_0(\cdot;|\alpha_0|^2)\right)\nn\\

1252:  &=& e^{i\gamma_0}\left[ \psi_0 +

1253:        |\alpha_0|^2\chi(\cdot;|\alpha_0|^2)\right]

1254: \nn\ea

1255: 

1256: Differentiation of (\ref{eq:nonlinearevp}) with respect to $|\alpha_0|$ 

1257: yields: 

1258: \be

1259: (H-E)\D_{|\alpha_0|}\Psi_0 + 2|\Psi_0|^2\D_{|\alpha_0|}\Psi_0 + 

1260:  \Psi_0^2\overline{\D_{|\alpha_0|}\Psi_0} = (\D_{|\alpha_0|}E)\Psi_0

1261: \label{eq:Psi0primeeqn}

1262: \ee

1263: Taken together with the complex conjugate of (\ref{eq:Psi0primeeqn}), this 

1264: yields, after multiplication by $\sigma_3$:

1265: \ba

1266: \sigma_3{\cal H} 

1267:   G_0 \left(\begin{array}{c} 1\\ 1\end{array}\right)F_0' &=& 

1268:  (\D_{|\alpha_0|}E)\ 

1269:  \sigma_3 G_0

1270:     \left(\begin{array}{c} 1\\ 1\end{array}\right) |\alpha_0|\psi_0\nn\\ 

1271:  &=& |\alpha_0| \xi_{01}

1272: \label{eq:taken}

1273: \ea

1274: 

1275: We define

1276: \be \xi_{02} \equiv  G_0 \left(\begin{array}{c} 1\\ 1\end{array}\right)

1277:            F_0',\label{eq:xi02_def}

1278: \ee

1279: where  

1280: \be

1281: F_0' = 

1282:      {\D\over\D |\alpha_0|}\left( |\alpha_0|\psi_0(\cdot;|\alpha_0|^2)\right)

1283:   = \psi_{0*} + |\alpha_0|^2\chi(x,|\alpha_0|^2).

1284: \label{eq:F0prime_def}

1285: \ee

1286: By the above calculation $\xi_{02}$ lies in the null space of

1287:   $(\sigma_3{\cal H})^2$.

1288: Therefore, the pair of vectors: $\xi_{01}$ and $\xi_{02}$ spans 

1289:  $N_g({\cal H}_0)$ and is nonsingular as $E_0\to E_{0*}$. By a previous

1290: remark:

1291: \be \zeta_{01} \equiv \sigma_3\xi_{02}\ {\rm and}\ \zeta_{02} \equiv

1292:                           \sigma_3\xi_{01}

1293: \label{eq:zeta01}

1294: \ee

1295: form a nonsingular basis for ${\cal H}_0^*$.

1296:  This choice of basis will facilitate a {\it uniform} description

1297:  of the dynamics in a neighborhood of the origin.

1298: 

1299: The above construction and Proposition \ref{prop:Hspectrum} imply the following

1300: basis for the discrete subspaces of $\cH_0$ and $\cH_0^*$.

1301: \begin{prop}\label{prop:xis}

1302: \ba

1303: N_g({\cal H}_0)\ &=&\ {\rm span} \{ \xi_{01}, \xi_{02} \} 

1304: \label{eq:NgH1}\\

1305: N_g({\cal H}^*_0)\ &=&\ {\rm span} \{ \zeta_{01}, \zeta_{02} \}

1306: \label{eq:NgH1*}\\

1307: N(\cH_0\mp\mu)\ &=&\ {\rm span} \{ \xi_\mu, \xi_{-\mu}\}

1308: \label{eq:NHpmmu}\\ 

1309: N(\cH_0^*\mp\mu)\ &=&\ {\rm span} \{ \zeta_\mu, \zeta_{-\mu}\} 

1310:  = \{ \sigma_3\xi_\mu, -\sigma_3\xi_{-\mu}\}\label{eq:NH*pmmu}\\

1311: \la \zeta_a, \xi_b\ra &=& C_{ab}\delta_{ab} + {\cal O}(|\alpha_0|^2),

1312: \label{eq:normalization}

1313: \ea

1314: where $a$ and $b$ vary over the set $\{(01),(02),\mu,-\mu\}$.

1315: For $\alpha_0$ small we have the expansions

1316: \ba

1317: \xi_{01} &=& \sigma_3G_0\left(\begin{array}{c}1 \\ 1\end{array}\right)F_0

1318:  = \sigma_3G_0\left(\begin{array}{c}1 \\ 1\end{array}\right)

1319:  \left(\psi_{0*}+ |\alpha_0|^2\chi^{(0)}_0\right)\nn\\

1320: \xi_{02} &=& G_0\left(\begin{array}{c}1 \\ 1\end{array}\right)F_0'

1321:  = G_0\left(\begin{array}{c}1 \\ 1\end{array}\right)

1322:  \left(\psi_{0*}+ |\alpha_0|^2\chi^{(0)}_0\right)\nn\\

1323: \xi_\mu &=& \left(\begin{array}{c}1 \\ 0\end{array}\right)

1324:  \left( \psi_{1*} + |\alpha_0|^2\chi_0^{(0)}\right)

1325:  + \ol{\alpha_0}^2\left(\begin{array}{c}0 \\ 1\end{array}\right)\chi_0^{(0)}

1326: \label{eq:xi_exps}

1327: \ea

1328: and $\xi_{-\mu} =\ol{\sigma_1\xi_\mu}$.

1329: \end{prop}

1330: 

1331: Finally, we shall find it useful to note that 

1332: \be

1333: \zeta_{01} = \sigma_3\xi_{02} = \xi_{01} + |\alpha_0|^2\chi(x;|\alpha_0|^2).

1334: \label{eq:useful}\ee

1335: \medskip

1336: 

1337: \nit{\bf Estimates for the linearized evolution operator}\label{sec:linevolution}

1338: 

1339: \begin{theo}\label{theo:lin_stab} (Linear dynamical stability \cite{MIW85})

1340: Let

1341: \be {\cal M}\ \equiv\ N_g({\cal H}_0^*)^\perp\ \cap \left(H^1\times H^1\right).

1342: \label{eq:Mdef}\ee

1343: There exists $C>0$ such that for any $f\in {\cal M}$

1344: \be || e^{-i{\cal H}_0t } f ||_{H^1}\ \le\ C||f||_{H^1}\label{eq:H1est}\ee

1345: \end{theo}

1346: 

1347: \begin{theo} \label{theo:lin_asymp_stab} (Dispersive estimates) \cite{Cucc1}

1348: Let

1349: \be {\cal M}_1\ \equiv\ \left[

1350:  N_g({\cal H}_0^*)\oplus N(H^*\mp\mu)\right]^\perp 

1351: \label{eq:M1def}\ee

1352: and $P_c$ the associated continuous spectral projection.

1353: For any $q\ge2$ there exists $C_{1,q}>0$ such that

1354: \be || e^{-i{\cal H}_0t }P_c f ||_{L^q}\ \le\ C_{1,q}\ t^{-{n\over2}+{n\over q}}

1355:  ||f||_{L^p},\label{eq:Lpqest}\ee

1356: where $p^{-1}+q^{-1}=1$.

1357: \end{theo}

1358: \begin{rmk}

1359: We shall in later sections use the notation ${\cal M}(t)$ and ${\cal M}_1(t)$ to

1360: denote corresponding time-dependent subspaces relative to the time dependent

1361:  operator ${\cal H}_0^*(t)$.

1362: \end{rmk}

1363: 

1364: \begin{theo} (Local decay estimate) \label{theo:local_decay}

1365: Let $\sigma$ be sufficiently large. 

1366:  Let $\omega\in \{\ \nu\in\R:\ |\nu|> |E_0|\ \}$, 

1367:  the interior of the continuous spectrum of

1368:  $\cH_0$. Then, for $t>0$

1369: \ba

1370: ||\la x\ra^{-\sigma}

1371:   e^{-i{\cal H}_0t} P_c \left({\cal H}_0-\omega -i0\right)^{-l}

1372: \la x\ra^{-\sigma}||_{\cB(L^2)}\ &\le& C\la t\ra^{-{3\over2}},\ \ l=0,1

1373: \label{eq:ldest}\\

1374: ||\la x\ra^{-\sigma} {\tilde\cH_0}^k

1375:   e^{-i{\cal H}_0t} P_c 

1376: \la x\ra^{-\sigma}||_{\cB(H^{2k},L^2)}\ 

1377:  &\le& C\la t\ra^{-{3\over2} -k},\ 

1378: \label{eq:H0kldest}

1379: \ea

1380: where ${\tilde\cH_0}=\cH_0+E_0\sigma_3$.

1381: For $t<0$, the same estimates hold with $-i0$ replaced by $+i0$

1382:  in (\ref{eq:ldest}).

1383: \end{theo}

1384: Theorem \ref{theo:local_decay} follows from the technique of \cite{JK}

1385: and results in \cite{Cucc2}; see also \cite{JSS}. 

1386:  In particular, for a Schr\"odinger operator

1387: $H_0=-\Delta+V$, one has the  local decay estimate 

1388: \be

1389: ||\la x\ra^{-\sigma} 

1390:   e^{-iH_0t} P_c(H_0)

1391: \la x\ra^{-\sigma}||_{\cB(L^2)}\

1392:  \le C\la t\ra^{-{3\over2}},\

1393: \label{eq:JKldest}

1394: \ee

1395: where $\sigma$ is positive and  sufficiently large.

1396: The key to the proof is an analysis of the resolvent, $(H_0-\lambda)^{-1}$

1397: near $\lambda=0$. One uses the spectral theorem 

1398: \be e^{-itH_0}P_c(H_0)\ =\ \int_0^\infty\ e^{-it\lambda} E'(\lambda)d\lambda

1399: \nn\ee

1400: and an explicit expansion of $E'(\lambda)$, 

1401: which can be expressed in terms of the imaginary part of the resolvent.

1402: The expansion of the resolvent is valid in the space~ 

1403:  $\cB(L^2(\la x\ra^{\sigma}dx),L^2(\la x\ra^{-\sigma}dx)$

1404:   for $\sigma$ sufficiently large.

1405:  Time decay is arbitrarily fast for functions with

1406:   spectral support away from $\lambda=0$,

1407: while the $t^{-{3\over2}}$ decay results from the behavior near $\lambda=0$.  

1408:  The analogue of 

1409: (\ref{eq:H0kldest}) is an estimate for $e^{-itH_0}H_0^kP_c(H_0)$,

1410: which follows from the expansion of \cite{JK} applied to the formula:

1411: \be e^{-itH_0}H_0^kP_c(H_0)\ =\ \int_0^\infty\ e^{-it\lambda} \lambda^k\ 

1412:  E'(\lambda)d\lambda.

1413: \nn\ee 

1414: Finally, this result can be obtained in the matrix case using the 

1415: approach of \cite{Cucc2}. 

1416: 

1417: \section{Decomposition and Modulation Equations}\label{sec:dec-mod}

1418: 

1419: Consider the nonlinear Schr\"odinger equation

1420: \be

1421: i\D_t\phi\ =\ H\phi\ +\ \lambda |\phi|^2\phi.

1422: \label{eq:NLS2}\ee

1423: 

1424: In the regime of low energy we decompose the solution of NLS 

1425:  in the following form:

1426: \be

1427: \phi(t)\ =\ e^{-i\Theta(t)}\left[ \Psi_0(t) +\Psi_1(t) +\phi_2(t)\right].

1428: \label{eq:decomp}

1429: \ee

1430: Here, $\Psi_0(t)$ and $\Psi_1(t)$ represent motion along the ground state

1431: and excited state manifolds of equilibria and $\phi_2$ 

1432:  is a decaying correction

1433:  term, lying in an appropriate {\it dispersive subspace}. The phase,

1434:  $\Theta$ is divided into two parts:

1435: \be \Theta(t) = \Theta_0(t) + \tTheta(t),\label{eq:Theta_def}\ee

1436: where 

1437: \be \Theta_0(t)=\int_0^t\ E_0(s)\ ds.

1438:  \label{eq:Theta_0_def}\ee

1439: Thus, $\D_t\Theta_0(t) = E_0(t)$  

1440:  is the modulated ground state energy, and $\tTheta(t)$ is 

1441: a ``long range'' logarithmic correction, which is to be derived below

1442: %\footnote{{\bf XYZ} Is there a relation between long range phase  corrections

1443: %and Berry's geometric phase?}.

1444: 

1445: We begin by setting

1446: \be

1447: \phi\ =\ e^{-i\Theta(t)}\left[ \Psi_0(t) + \phi_1\right].

1448: \label{eq:firststep}

1449: \ee

1450: Substitution of (\ref{eq:firststep}) into (\ref{eq:NLS2}) yields the 

1451: following equation for $\phi_1$:

1452: 

1453: \ba

1454: i\D_t\phi_1\ &=&\ \left(H-E_0(t)\right)\phi_1\ +\ 

1455:  2\lambda |\Psi_0(t)|^2\phi_1\ +\ \lambda\Psi_0^2(t)\ol{\phi_1}\nn\\

1456: &+&\ \left(-\D_t\tTheta+2\lambda|\phi_1|^2\right)\Psi_0(t)\ 

1457:  +\ \lambda\ol{\Psi_0(t)}\phi_1^2\ +

1458: \ \lambda |\phi_1|^2\phi_1 - \D_t\tTheta(t)\phi_1\nn\\

1459: &-& i\D_t\Psi_0(t) 

1460: \label{eq:phi1_eqn}\ea

1461: 

1462: Next, we decompose $\phi_1$ into a part along the excited state manifold and a

1463: correction term:

1464: \be \phi_1\ \equiv\ \Psi_1(t)\ +\ \phi_2.\label{eq:phi1_expand}\ee

1465:  We have, using (\ref{eq:phi1_eqn}),

1466: \ba

1467: i\D_t\phi_2 &=& \left( H-E_0(t) \right)\phi_2 + 2\lambda |\Psi_0(t)|^2\phi_2

1468:  + \lambda \Psi_0^2(t)\ol{\phi_2} - \D_t\tTheta(t)\phi_2\nn\\

1469: &\ -& \left( E_{01}(t) +\D_t\tTheta(t) + i\D_t \right)\Psi_1(t)\ + \ 2\lambda |\Psi_0(t)|^2\Psi_1(t)

1470: \ +\ \lambda \Psi_0(t)^2\ol{\Psi_1(t)}\nn\\

1471:  &+&\ \left(-\D_t\tTheta(t) + 2\lambda \Psi_0(t)|\phi_1|^2\right)\Psi_0(t)

1472:  + \lambda \ol{\Psi_0(t)}\phi_1^2\nn\\

1473: &\ +&  \lambda\left( |\phi_1|^2 \phi_1 -

1474:  |\Psi_1(t)|^2\Psi_1(t)\right)\nn\\

1475: & -&\ i \D_t\Psi_0(t) 

1476: \label{eq:phi2_eqn}

1477: \ea

1478: Since equation (\ref{eq:phi2_eqn}) involves $\ol{\phi_2}$ it is natural to

1479: consider the system governing $\phi_2$ and $\ol{\phi_2}$.

1480: Let

1481: \be \Phi_2\ =\ \left(\begin{array}{c}\phi_2\\ \ol{\phi_2}\end{array}\right)

1482: \label{eq:Phi2_def}\ee

1483: and introduce the matrix linear operator:

1484: \be

1485: {\cal H}_0(t)\ \equiv\ \sigma_3

1486:  \left(\begin{array}{cc} H-E_0(t)+2\lambda|\Psi_0(t)|^2 & \lambda\Psi^2_0(t)\\

1487:   \lambda\ol{\Psi}^2_0(t) & H-E_0(t)+2\lambda|\Psi_0(t)|^2 \end{array}\right).

1488: \label{eq:H0_def}\ee

1489: Then we have (recall that $\phi_1 = \Psi_1 +\phi_2$) 

1490: \ba

1491: i\D_t\Phi_2\ &=&\ 

1492: {\cal H}_0(t)\Phi_2\ + \bF -i \D_t\left(\begin{array}{c}

1493:  \Psi_0(t)\\ c.c. \end{array}\right) 

1494: - \left(\ ( E_{01}(1)+\D_t\tTheta(t) ) \sigma_3 + i\D_t\ \right)\ 

1495:  \left(\begin{array}{c}

1496:  \Psi_1(t)\\ c.c. \end{array}\right),\nn\\

1497: && \label{eq:Phi2_eqn}

1498: \ea

1499: where

1500: \ba

1501: \bF &\equiv&\ 

1502: - \D_t\tTheta(t)\sigma_3

1503:  \left(\begin{array}{c}\Psi_0\\ c.c.\end{array}\right)\  

1504: -\ \D_t\tTheta(t)\sigma_3

1505: \left(\begin{array}{c}\phi_2\\ c.c.\end{array}\right)\

1506: \nn\\ 

1507: &&+ \lambda\sigma_3\left(\begin{array}{c}2|\Psi_0|^2\Psi_1+\Psi_0^2\ol{\Psi_1}

1508:  \\

1509:  c.c. \end{array}\right)\nn\\

1510: &+& \lambda\sigma_3\left(\begin{array}{c}2\Psi_0|\phi_1|^2 +

1511:  \ol{\Psi_0}\phi_1^2\\

1512:  c.c. \end{array}\right)

1513: \ +\ \lambda\sigma_3\left(\begin{array}{c}|\phi_1|^2\phi_1

1514:  - |\Psi_1|^2\Psi_1\\

1515:  c.c. \end{array}\right)

1516: \label{eq:bFdef}

1517: \ea

1518: \medskip

1519: 

1520: \subsection{ Modulation Equations} \label{sec:modulationeqns}

1521: 

1522: Motivated by the results of Theorem \ref{theo:lin_asymp_stab}

1523:   on dispersive decay, we shall require that

1524: \be \Phi_2(t)\in {\cal M}_1(t),\label{eq:phi1inM1}\ee

1525: where ${\cal M}_1$ is defined in (\ref{eq:M1def}). Equivalently,

1526:  \be P_c(\cH_0(t))\Phi_2(t) = \Phi_2(t),\nn\ee  

1527: where $P_c(\cH_0(t))$ denotes the continuous spectral projection of 

1528: $\cH_0(t)$. By Proposition

1529:  \ref{prop:H*spectrum} this imposes four orthogonality conditions on 

1530:  $\Phi_2(t)$:

1531: \be \la \sigma_3\xi_a(t),\Phi_2(t) \ra = 0,\label{eq:orthogonality}\ee

1532: where $a\in\{(01), (02), \mu,-\mu\}$.

1533: We impose (\ref{eq:orthogonality}) at $t=0$ and now derive 

1534: {\it modulation equations} for the coordinates $\alpha_0(t)$ and $\alpha_1(t)$

1535: ensuring that (\ref{eq:orthogonality}) 

1536:   persists for all $t\ne0$.

1537:  

1538: To derive the modulations we first take the inner product of 

1539: (\ref{eq:Phi2_eqn})

1540: with the adjoint vectors $\sigma_3\xi_a$ to obtain the identity: 

1541: \ba

1542: && \left\langle \sigma_3\xi_a\ ,\  

1543:  i\D_t\left(\begin{array}{c} \Psi_0 

1544:                         \\ c.c. 

1545:         \end{array}\right)

1546:  \right\rangle

1547: + \left\langle \sigma_3\xi_a,\left(\ (E_{01}+\D_t\tTheta)\sigma_3 +i\D_t\right)

1548:               \left(\begin{array}{c} \Psi_1\\ c.c.

1549:                     \end{array}\right)\right\rangle\nn\\

1550: &&=\ 

1551: \la \cH_0^*(t)\sigma_3\xi_a,\Phi_2\ra + \la\sigma_3\xi_a,\bF\ra

1552: \ + i\langle \D_t(\sigma_3\xi_a),\Phi_2\rangle 

1553:  -i\D_t\la\sigma_3\xi_a,\Phi_2\ra

1554: \label{eq:identity}

1555: \ea

1556: 

1557: The initial data for NLS is decomposed so that 

1558: $\la\sigma_3\xi_a(t),\Phi_2(t)\ra=0$ for $t=0$. In order for 

1559: this condition to persist for all time it is necessary and sufficient that

1560: the last term in (\ref{eq:identity}) vanish, or equivalently:

1561: 

1562: \begin{prop}\label{prop:ModulationEqns} 

1563: The condition that $P_c(\cH_0(t))\Phi_2(t)=\Phi_2(t)$ is equivalent to the

1564: following modulation equations for the coordinates $\alpha_0$ and $\alpha_1$

1565:  which specify the dynamics along the ground state and excited 

1566:  state manifolds  of equilibria:

1567: \ba

1568: &&\left\langle \sigma_3\xi_a\ ,\ 

1569:       i\D_t\vPsi_0\right\rangle

1570:  \ +\ \left\langle \sigma_3\xi_a\ ,\ 

1571:  \left( \sigma_3 (E_{01}+ \D_t\tTheta)+i\D_t \right)

1572:    \vPsi_1 

1573:    \right\rangle, 

1574: \nn\\

1575: &&= 

1576: -\lambda \left\langle \xi_a,

1577: \left(\begin{array}{c}2|\Psi_0|^2\Psi_1+\Psi_0^2\ol{\Psi_1}\\

1578:  c.c. \end{array}\right)\right\rangle\nn\\

1579: && +\ \left\la \xi_a,

1580:  \left(\ -\D_t\tTheta(t)+2\lambda|\phi_1|^2\ \right)

1581:  \left(\begin{array}{c}\Psi_0\\ c.c.\end{array}\right)\right\ra

1582: \nn\\

1583: &&+ \lambda\left\la \xi_a, \left(\begin{array}{c} \ol{\Psi_0}\phi_1^2\\

1584:  c.c. \end{array}\right)\right\ra

1585:  +\ \lambda \left\langle \xi_a,

1586: \left(\begin{array}{c}|\phi_1|^2\phi_1

1587: - |\Psi_1|^2\Psi_1\\ c.c. \end{array}\right)\right\rangle\nn\\  

1588: && - \D_t\tTheta(t)\la \xi_a,\Phi_2\ra +\ i\la \D_t(\sigma_3\xi_a),\Phi_2\ra.

1589: \label{eq:modeqns}

1590: \ea

1591: where $a\in\{(01), (02), \mu,-\mu\}$ and 

1592:  $\vec\Psi_j\equiv (\Psi_j,\ol{\Psi_j}),\ j=0,1$.

1593: \end{prop}

1594: 

1595: \begin{rmk} 

1596: \smallskip

1597: 

1598: \nit (A)  Note that the term $\la \cH_0^*(t)\sigma_3\xi_a,\Phi_2\ra$

1599: in (\ref{eq:identity}) vanishes by the orthogonality constraint and because 

1600:  $\cH_0^*(t)$ maps the discrete

1601:  subspace into itself.

1602: It therefore does not appear in 

1603:  (\ref{eq:modeqns}).

1604: 

1605: \nit (B) The last term in (\ref{eq:modeqns}) is present

1606:  due to the time-dependence of the

1607: eigenvectors $\xi_a(t)$. An important simplification

1608: of this ``commutator term'', which we require for $a=(01)$,

1609: is carried out in the appendix of section \ref{sec:commutator}.

1610: \end{rmk}

1611: 

1612: Initial data for the system (\ref{eq:modeqns}),

1613:  (\ref{eq:Phi2_eqn}), governing $\alpha_0, \alpha_1$ and 

1614: $\Phi_2$ are obtained as follows. Given data $\phi_0$ for NLS,

1615: we find $\alpha_0$ so as to minimize 

1616: \be \| \phi_0-\Psi_{\alpha_0}\|_2;\nn\ee

1617: see \cite{SW9092}.

1618: This $\alpha_0$ is used to define the initial Hamiltonian $\cH_0(0)$.

1619: Now decompose $\phi_0$ using the biorthogonal decomposition 

1620: associated with $\cH_0(0)$. This specifies $\alpha_0(0), \alpha_1(0)$

1621: and $\Phi_2(0)$. 

1622: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1623: \bigskip

1624: 

1625: \subsection{Conservation laws and 'a priori bounds}\label{sec:bounds}

1626: 

1627: In this subsection we obtain bounds on $\alpha_0$, $\alpha_1$ and $\phi_2$

1628: using the conservation laws of NLS, noted in the introduction.

1629: 

1630: By the $L^2$ conservation law: $\bN[\phi]=\bN[\phi_0]$ we have

1631: \ba

1632:  h(t)\ &\equiv&\ |\Psi_0(t)|^2 + |\Psi_1(t)|^2 + \int\ |\phi_2(t,x)|^2\ dx

1633: \nn\\

1634: &=&\ \bN[\phi_0]-2\Re\int\Psi_0\ol{\Psi_1} - 2\Re\int\Psi_0\ol{\phi_2}

1635:  - \Re\int\Psi_1\ol{\phi_2}

1636: \label{eq:con1}

1637: \ea 

1638: Note that $\int\ \Psi_0\ol{\Psi_1}\ =\ \cO(|\alpha_0|\ |\alpha_1|^2)

1639:  + \cO(|\alpha_0|^2\ |\alpha_1|)$. Furthermore,  we have

1640: the orthogonality relation (\ref{eq:orthogonality}) with $a=(01)$,

1641:  $\la\sigma_3\xi_{01},\Phi_2\ra=0$  or equivalently

1642:  $\Re\int\Psi_0\ol{\phi_2}\ =\ 0$. Therefore 

1643:  (\ref{eq:con1}) is equivalent to

1644: \be

1645: \left(\ 1\ +\ \cO(h(t))\ \right)\ h(t)

1646: \ =\ \bN[\phi_0]\ 

1647: \ -\ \Re\int\Psi_1\ol{\phi_2}

1648: \label{eq:con2}

1649: \ee 

1650: The orthogonality relation (\ref{eq:orthogonality}) with $a=\mu$ or

1651:  $\la\sigma_3\xi_\mu,\Phi_2\ra=0$ implies 

1652:  $\Re\int\Psi_1\ol{\phi_2}\ =\

1653:   \cO\left(|\alpha_1|^3\ +\ |\alpha_1|\ |\alpha_0|^2)\right)\|\phi_0\|_2$ 

1654:  and therefore if $\bN[\phi_0]$ is sufficiently small

1655: \be

1656: |\Psi_0(t)|^2 + |\Psi_1(t)|^2 + \int\ |\phi_2(t,x)|^2\ dx\ \le\ C\bN[\phi_0]

1657: \label{eq:L2componentbound}

1658: \ee

1659: In a similar way, the conservation law $\bH_{en}[\phi(t)]=\bH_{en}[\phi_0]$

1660:  implies, if $\cE_0$ is sufficiently small,  the bound:

1661: \be

1662: \int\ |\nabla\phi_2(t,x)|^2\ dx\ \le\ C\cE_0 

1663: \label{eq:phi2H1bound}

1664: \ee

1665: \section{ Toward a normal form - algebraic reductions and frequency analysis}

1666: We now embark on a detailed calculation leading to a form

1667: of this system, which though equivalent, is of a form to which

1668: normal form methods can be easily applied. 

1669: 

1670: \subsection{Modulation equations}

1671: 

1672: Equations (\ref{eq:modeqns}) are a coupled system for $\alpha_0$,

1673:  $\alpha_1$ and their complex conjugates. It is natural to write the system

1674: as one which is nearly diagonal. This can be done by taking 

1675:  appropriate linear combinations

1676: of the equations in (\ref{eq:modeqns}). 

1677: 

1678: The equation which essentially determines  $\alpha_0$ can be found by 

1679:  adding the two equations obtained from (\ref{eq:modeqns})

1680: by setting $a=(01)$ and $a=(02)$. This gives:

1681: \ba

1682: && \la \sigma_3(\xi_{01}+\xi_{02})\ ,\

1683: i\D_t\vPsi_0\ra

1684:  \ +\ \la \sigma_3(\xi_{01}+\xi_{02})\ ,\

1685:  \left( \sigma_3(E_{01}+\D_t\tTheta)+i\D_t \right)

1686:  \vPsi_1\ra

1687: \nn\\

1688: &&= 

1689: -\lambda \left\langle \xi_{01}+\xi_{02},

1690: \left(\begin{array}{c}2|\Psi_0|^2\Psi_1+\Psi_0^2\ol{\Psi_1}\\

1691:  c.c. \end{array}\right)\right\rangle\nn\\

1692: && +\ \left\la \xi_{01}+\xi_{02},

1693:  \left(\ -\D_t\tTheta(t)+2\lambda|\phi_1|^2\ \right)

1694:  \left(\begin{array}{c}\Psi_0\\ c.c.\end{array}\right)\right\ra

1695: \nn\\

1696: &&+ \lambda\left\la \xi_{01}+\xi_{02},

1697:   \left(\begin{array}{c} \ol{\Psi_0}\phi_1^2\\

1698:  c.c. \end{array}\right)\right\ra

1699:  +\ \lambda \left\langle \xi_{01}+\xi_{02},

1700: \left(\begin{array}{c}|\phi_1|^2\phi_1

1701: - |\Psi_1|^2\Psi_1\\ c.c. \end{array}\right)\right\rangle\nn\\

1702: &&- \D_t\tTheta(t)\la \xi_{01}+\xi_{02},\Phi_2\ra 

1703:  + i\la \D_t(\sigma_3(\xi_{01}+\xi_{02})),\Phi_2\ra.

1704: \label{eq:me1}

1705: \ea

1706: The difference of the $a=(01)$ and $a=(02)$

1707: equations is the complex conjugate of the equation (\ref{eq:me1}).

1708: \bigskip

1709: 

1710: The equation which essentially determines  $\alpha_1$ is equation 

1711:  (\ref{eq:modeqns}) with  $a=\mu$:

1712: \ba

1713: && \la \sigma_3\xi_\mu\ ,\

1714:       i\D_t\vPsi_0\ra

1715:  \ +\ \la \sigma_3\xi_\mu\ ,\

1716:  \left( \sigma_3(E_{01}+\D_t\tTheta) + i\D_t \right)

1717:       \vPsi_1\ra

1718: \nn\\

1719: &&=\ 

1720: -\lambda \left\langle \xi_\mu,

1721: \left(\begin{array}{c}2|\Psi_0|^2\Psi_1+\Psi_0^2\ol{\Psi_1}\\

1722:  c.c. \end{array}\right)\right\rangle\nn\\

1723: && +\ \left\la \xi_\mu,

1724:  \left(\ -\D_t\tTheta(t)+2\lambda|\phi_1|^2\ \right)

1725:  \left(\begin{array}{c}\Psi_0\\ c.c.\end{array}\right)\right\ra

1726: \nn\\

1727: &&+ \lambda\left\la \xi_\mu, \left(\begin{array}{c} \ol{\Psi_0}\phi_1^2\\

1728:  c.c. \end{array}\right)\right\ra

1729:  +\ \lambda \left\langle \xi_\mu,

1730: \left(\begin{array}{c}|\phi_1|^2\phi_1

1731: - |\Psi_1|^2\Psi_1\\ c.c. \end{array}\right)\right\rangle\nn\\

1732: && - \D_t\tTheta(t)\la \xi_\mu,\Phi_2\ra +\  i\la \D_t(\sigma_3\xi_\mu),\Phi_2\ra.\

1733: \label{eq:me2}

1734: \ea

1735: The equation corresponding to $a=-\mu$ is the complex conjugate of this 

1736:  equation.

1737: 

1738: Since

1739: \be

1740: \D_t \left(\begin{array}{c} \Psi_k\\ \ol{\Psi_k} \end{array}\right) =\

1741: \left(\begin{array}{c} \D_k\Psi_k \\ \ol{ \ol{\D_k}\Psi_k }

1742:  \end{array}\right) \D_t\alpha_k

1743: \ +\ \left(\begin{array}{c} \ol{\D_k}\Psi_k \\ \ol{\D_k\Psi_k}

1744:  \end{array}\right)\ \D_t\ol{\alpha_k},

1745: \label{eq:dd}

1746: \ee

1747: we see that the equations for $\valpha_j=(\alpha_0,\ol{\alpha_0})^T,\

1748:  j=0,1$ can be expressed in the form:

1749: \ba

1750:  i{\bf M_{00}}

1751:  \D_t\left(\begin{array}{c}\alpha_0\\ \ol{\alpha_0}\end{array}\right)

1752: +i{\bf M_{01}}

1753:  \D_t\left(\begin{array}{c}\alpha_1\\ \ol{\alpha_1}\end{array}\right)

1754: + (E_{01}+\D_t\tTheta){\bf N_{01}}

1755:          \left(\begin{array}{c}\alpha_1\\ \ol{\alpha_1}\end{array}\right)

1756: &=& \bF_0\label{eq:me1a}\\

1757:  i{\bf M_{10}}

1758:  \D_t\left(\begin{array}{c}\alpha_0\\ \ol{\alpha_0}\end{array}\right)

1759: +i{\bf M_{11}}

1760:  \D_t\left(\begin{array}{c}\alpha_1\\ \ol{\alpha_1}\end{array}\right)

1761: +(E_{01}+\D_t\tTheta){\bf N_{11}}

1762: \left(\begin{array}{c}\alpha_1\\ \ol{\alpha_1}\end{array}\right)

1763: &=& \bF_1

1764: \label{eq:me2a}\ea

1765: Here, for $k=0,1$:

1766: \ba

1767: {\bf M_{0k}} &=&

1768: {\bf G_0} \left(\begin{array}{cc}

1769: %%

1770: \left\la\sigma_3(\xi_{01}+\xi_{02}),

1771:  \left(\begin{array}{c} \D_k\Psi_k \\ \ol{ \ol{\D_k}\Psi_k }\end{array}\right)

1772:  \right\ra

1773: &  \left\la\sigma_3(\xi_{01}+\xi_{02}),

1774:   \left(\begin{array}{c}

1775:  \ol{\D_k}\Psi_k \\ \ol{\D_k\Psi_k}\end{array}\right)\right\ra\\

1776:  \left\la\sigma_3\ol{(\xi_{01}+\xi_{02})},

1777:  \left(\begin{array}{c} \ol{\ol{\D_k}\Psi_k} \\ \D_k\Psi_k

1778:  \end{array}\right)\right\ra &

1779: \left\la\sigma_3\ol{(\xi_{01}+\xi_{02})},

1780:  \left(\begin{array}{c}\ol{\D_k\Psi_k}\\ \ol{\D_k}\Psi_k\end{array}\right)

1781:  \right\ra

1782: %%

1783: \end{array}\right)\label{eq:M0k}\\

1784: \nn\\

1785: {\bf N_{01}} &=&

1786: {\bf G_0}\left(\begin{array}{cc}

1787: %%

1788: \left\la\sigma_3(\xi_{01}+\xi_{02}),

1789:  \left(\begin{array}{c}\psi_1\\ 0\end{array}\right)\right\ra &

1790: \left\la\sigma_3(\xi_{01}+\xi_{02}),\left(\begin{array}{c}0\\

1791:  \psi_1\end{array}\right)

1792: \right\ra\\

1793: \left\la\sigma_3\ol{(\xi_{01}+\xi_{02})},

1794:  \left(\begin{array}{c}0\\ \psi_1\end{array}\right)\right\ra &

1795: \left\la\sigma_3\ol{(\xi_{01}+\xi_{02})},

1796:  \left(\begin{array}{c}\psi_1\\ 0\end{array}\right)

1797: \right\ra

1798: \end{array}\right)

1799: \label{eq:N01}\\

1800: \nn\\

1801: {\bf M_{1k} } &=&

1802: {\bf G_0}\left(\begin{array}{cc}

1803: \left\la\sigma_3\xi_\mu,

1804:  \left(\begin{array}{c} \D_k\Psi_k \\ \ol{ \ol{\D_k}\Psi_k }\end{array}\right)

1805:  \right\ra

1806: & \left\la \sigma_3\xi_\mu,

1807:  \left(\begin{array}{c} \ol{\D_k}\Psi_k \\ \ol{\D_k\Psi_k}\end{array}\right)

1808:  \right\ra

1809:  \\

1810:  \left\la\sigma_3\ol{\xi_\mu},

1811:  \left(\begin{array}{c} \ol{\ol{\D_k}\Psi_k} \\ \D_k\Psi_k

1812:  \end{array}\right)\right\ra &

1813: \left\la\sigma_3\ol{\xi_\mu},

1814:  \left(\begin{array}{c}\ol{\D_k\Psi_k}\\ \ol{\D_k}\Psi_k

1815: \end{array}\right)\right\ra

1816:  \end{array}\right),\label{eq:M1k}\\

1817: \nn\\

1818:  {\bf N_{11} } &=&

1819: {\bf G_0}\left(\begin{array}{cc}

1820: \left\la\sigma_3\xi_\mu,\left(\begin{array}{c}\psi_1\\ 0\end{array}\right)

1821: \right\ra &

1822: \left\la\sigma_3\xi_\mu,\left(\begin{array}{c}0\\ \psi_1\end{array}\right)

1823: \right\ra\\

1824: \left\la\sigma_3\ol{\xi_\mu},

1825:  \left(\begin{array}{c}0\\ \psi_1\end{array}\right)\right\ra &

1826: \left\la\sigma_3\ol{\xi_\mu},

1827:  \left(\begin{array}{c}\psi_1\\ 0\end{array}\right)\right\ra

1828: \end{array}\right)

1829: \label{eq:N11}

1830: \ea

1831: 

1832: \subsection{Algebraic reductions and determination of $\tTheta(t)$}

1833: To express the modulation  equations (\ref{eq:modeqns}) in a tractable 

1834: form we shall make use of a number of notations and  relations

1835:  which now list for convenience; see also the section \ref{sec:notation}. 

1836: 

1837: \medskip

1838: 

1839: \nit $\chi_k^{(j)}$ denotes a spatially localized  function of order $|\alpha_j|^k$,

1840:  as $|\alpha_j|\to0$. 

1841: 

1842: \nit $\cO_k^{(j)}$ denotes a quantity which is of order

1843:  $|\alpha_j|^k$ as $|\alpha_j|\to0$.  Both $\chi_k^{(j)}$  and $\cO_k^{(j)}$

1844:  are invariant under the map 

1845: $\alpha_j\mapsto\alpha_je^{i\gamma}$. 

1846: 

1847: \nit $\cO_k^{(0,1)} = \cO_{k_1}^{(0)}\cO_{k_2}^{(1)},\ \ k=k_1+k_2$.

1848: 

1849: \ba

1850: \phi_1 &=& \Psi_1 + \phi_2\nn\\

1851: \alpha_0 &=& |\alpha_0| e^{i\gamma_0},\ \ol{\alpha_0}e^{2i\gamma_0} = \alpha_0

1852:  \nn\\

1853:  \left(\D_k\Psi_k, \ol{\D_k}\Psi_k\right) &=& 

1854:  \left(\psi_{k*}+|\alpha_k|^2\chi_0^{(k)}, \alpha_k^2\chi_0^{(k)}\right),

1855:  \ k=0,1\nn\\

1856: \xi_{01}+\xi_{02} &=& 

1857:  G_0\left(\begin{array}{c} F_0+F_0'\\ F_0-F_0'\end{array}\right)

1858:  = \left(\begin{array}{c}

1859: 2e^{i\gamma_0}(\psi_{0*} + \chi_2^{(0)}) \\ e^{-i\gamma_0}\chi_2^{(0)}

1860: \end{array}\right)\nn\\

1861: \ \xi_\mu &=& 

1862:  \left(\begin{array}{c} \psi_{1*}+\chi_2^{(0)}\\ \ol{\alpha_0}^2\chi_0^{(0)}

1863:  \end{array}\right)

1864: \nn\\

1865: F_0+F_0' &=& 2\psi_{0*}+|\alpha_0|^2\chi(\cdot;|\alpha_0|^2) =

1866:   2(\psi_{0*} + \chi_2^{(0)})\nn\\

1867:   F_0-F_0' &=& |\alpha_0|^2\chi(\cdot;|\alpha_0|^2)=\chi_2^{(0)}.

1868:  \nn\\

1869:  \la \psi_{0*},\psi_0^{(1)}(\cdot;0)\ra &=& 0,\label{eq:ids}

1870: \ea

1871: 

1872: Using (\ref{eq:ids}) in (\ref{eq:me1}) we get:

1873: \ba

1874: &&2i(1+\cO_4^{(0)})\D_t\alpha_0 + i\alpha_0^2\cO_0^{(0)}\D_t\ol{\alpha_0}

1875: + \la \sigma_3(\xi_{01}+\xi_{02})\ ,\

1876:  \left(\ \sigma_3(E_{01}+\D_t\tTheta)+i\D_t\ \right)

1877:            \vPsi_1 \ra

1878:  \nn\\

1879: &&= -2\left( \D_t\tTheta \la F_0',\Psi_0\ra-2\lambda\la F_0',|\phi_1|^2\Psi_0\ra

1880:  \right)\nn\\

1881: &&+\la F_0+F_0',2\lambda|\Psi_0|^2\Psi_1+\lambda\Psi_0^2\ol{\Psi_1}

1882:  +\lambda|\phi_1|^2\phi_1 -\lambda|\Psi_1|^2\Psi_1\ra

1883: -e^{2i\gamma_0}\la F_0-F_0', c.c.\ra\nn\\

1884: && + ie^{2i\gamma_0}\la \D_t[\sigma_3(\xi_{01}+\xi_{02})],\Phi_2\ra

1885: \label{eq:a0-1}

1886: \ea

1887: \nit{\bf Determination of $\tTheta(t)$}

1888: 

1889: We anticipate  that 

1890:  generically $|\phi_1|\sim |\alpha_1| \sim t^{-{1\over2}}$ for $t$ very large.

1891: $\alpha_0$ will have a limit as $t\to\pm\infty$

1892: if $\D_t\alpha_0$ is integrable. We ensure this by  

1893:  choosing $\D_t\tTheta$ to cancel 

1894: the terms which are of order $t^{-1}$ and {\it nonoscillatory}.

1895: Thus, we choose $\tTheta$ to satisfy

1896: \be

1897: \D_t\tTheta \la F_0',\Psi_0\ra - 2\lambda\la F_0',|\phi_1|^2\Psi_0\ra = 0.

1898: \label{eq:tTheta}

1899: \ee

1900: To leading order this gives:

1901: \be

1902: \D_t\tTheta \sim 2\lambda\la \psi_{0*}^2,|\phi_1|^2\ra.

1903: \label{eq:tTheta-lead}

1904: \ee

1905: In this way, a logarithmic correction to the standard phase, $\int_0^tE_0(s)ds$

1906:  arises.

1907: 

1908: 

1909: Equations (\ref{eq:a0-1}), (\ref{eq:tTheta}) together with Proposition

1910:  \ref{prop:commutators} imply

1911: \ba

1912: &&2i(1+\cO_4^{(0)})\D_t\alpha_0 + i\alpha_0^2\cO_0^{(0)}\D_t\ol{\alpha_0}

1913: + (E_{01}+\D_t\tTheta)\la\xi_{01}+\xi_{02}\ ,\vPsi_1\ra +

1914:  \la \sigma_3(\xi_{01}+\xi_{02}),i\D_t\vPsi_1 \ra

1915: \nn\\ 

1916: &&= -\D_t\tTheta\la \chi_0^{(0)},\phi_2\ra + \D_t\tTheta \alpha_0^2

1917:  \la \chi_0^{(0)}, \ol{\phi_2}\ra\nn\\

1918: && + \lambda\cO_0^{(0)}(|\alpha_0|^2\alpha_1+\alpha_0^2\ol{\alpha_1})

1919:  +\lambda\la \chi_0^{(0)},|\phi_1|^2\phi_1 -|\Psi_1|^2\Psi_1\ra

1920: -\lambda\alpha_0^2 \la \chi_0^{(0)}, |\phi_1|^2\ol{\phi_1} -

1921:  |\Psi_1|^2\ol{\Psi_1}\ra\nn\\

1922: && -\D_t|\alpha_0|^2\left[\ \la \chi_0^{(0)}, \phi_2\ra + 

1923:  e^{2i\gamma_0}\la \chi_0^{(0)},\ol{\phi_2}\ra\ \right] +

1924:  i |\alpha_0|^2 \D_t\gamma_0

1925: \left[\ \la \chi_0^{(0)}, \phi_2\ra + 

1926:  e^{2i\gamma_0}\la \chi_0^{(0)},\ol{\phi_2}\ra\ \right] 

1927:  \label{eq:a0-1a}

1928: \ea

1929: \bigskip

1930: 

1931: We now turn equation (\ref{eq:me2}). 

1932: Using (\ref{eq:ids}) equation (\ref{eq:me2}) can be written as:

1933: \ba

1934: &&(1+\cO_2^{(0,1)})\ i\D_t\alpha_1 +

1935: (\cO(\alpha_0^2)+\cO(\alpha_1^2) )\ i\D_t\ol{\alpha_1}\nn\\

1936: &&+ (\ (1+\cO_2^{(0,1)})\alpha_1 + \alpha_0^2\ol{\alpha_1}\cO_2^{(0,1)}\ )

1937:  (E_{01}+\D_t\tTheta) + \la\sigma_3\xi_\mu,i\D_t\vPsi_0\ra\nn\\

1938: && = -\D_t\tTheta(\ \la\chi,\phi_2\ra + \alpha_0^2\la\chi,\ol{\phi_2}\ra\ )

1939: -\D_t\tTheta \cO_0^{(0,1)}\alpha_0\nn\\

1940: && -\lambda\cO_2^{(0,1)}(|\alpha_0|^2\alpha_1+\alpha_0^2\ol{\alpha_1})

1941: +\lambda \cO_0^{(0)}\alpha_0\la\chi,|\phi_1|^2\ra\nn\\

1942: &&+\lambda \ol{\alpha_0}\la\chi,\phi_1^2\ra +

1943:  \lambda \alpha_0^3\la\chi,\ol{\phi_1}^2\ra\nn\\

1944: &&+\lambda \la\chi,|\phi_1|^2\phi_1-|\Psi_1|^2\Psi_1\ra

1945: + \lambda\alpha_0^2\la\chi,|\phi_1|^2\ol{\phi_1}-|\Psi_1|^2\ol{\Psi_1}\ra

1946: \nn\\

1947: && + i\la \D_t(\sigma_3\xi_\mu), \Phi_2\ra

1948: \label{eq:a1-1a}

1949: \ea

1950: We next write the systems for $\valpha_0$ and $\valpha_1$.

1951: \ba

1952: && i{\bf M_{00}}\D_t\valpha_0 + i{\bf M_{01}}\D_t\valpha_1 

1953:  + (E_{01}+\D_t\tTheta)\sigma_3{\bf N_{01}}\valpha_1\nn\\

1954: &&\ \  =\  \lambda\cO_0^{(0)}\sigma_3\left(\begin{array}{c}

1955:                        |\alpha_0|^2\alpha_1+\alpha_0^2\ol{\alpha_1}\\ c.c.

1956:                          \end{array}\right)\nn\\

1957: &&\ \ + \lambda\sigma_3\left(\begin{array}{c}

1958: \la\chi_0^{(0)},|\phi_1|^2\phi_1-|\Psi_1|^2\Psi_1\ra 

1959:  + \alpha_0^2\la\chi_0^{(0)},

1960:  |\phi_1|^2\ol{\phi_1}-|\Psi_1|^2\ol{\Psi_1}\ra \\ c.c.

1961:                          \end{array}\right)\nn\\

1962: &&\ \ -\D_t\tTheta

1963: \left(\begin{array}{c}

1964: \la\chi_0^{(0)},\phi_2\ra +\alpha_0^2\la\chi_0^{(0)},\ol{\phi_2}\ra

1965: \\ c.c.  \end{array}\right)

1966: \nn\\

1967: &&\ \ - \D_t|\alpha_0|^2\sigma_3

1968: \left(\begin{array}{c}

1969: \la\chi_0^{(0)},\phi_2\ra + e^{2i\gamma_0}\la\chi_0^{(0)},\ol{\phi_2}\ra 

1970:                          \\ c.c.

1971:                          \end{array}\right)\nn\\

1972: &&\  \ +i |\alpha_0|^2\D_t\gamma_0

1973: \left(\begin{array}{c}

1974: \la\chi_0^{(0)},\phi_2\ra + e^{2i\gamma_0}\la\chi_0^{(0)},\ol{\phi_2}\ra 

1975:                          \\ c.c.

1976:                          \end{array}\right)

1977: \label{eq:a0-sys}\\

1978: && i{\bf M_{11}}\D_t\valpha_1 + i{\bf M_{10}}\D_t\valpha_0 

1979: + (E_{01}+\D_t\tTheta)\sigma_3{\bf N_{11}}\valpha_1\nn\\

1980: &&\ \ =  \lambda\cO_0^{(0,1)}\sigma_3\left(\begin{array}{c}

1981:                        |\alpha_0|^2\alpha_1+\alpha_0^2\ol{\alpha_1}\\ c.c.

1982:                          \end{array}\right)\nn\\

1983: &&\ \ + \lambda\sigma_3\left(\begin{array}{c}

1984:  \la \chi,|\phi_1|^2\phi_1-|\Psi_1|^2\Psi_1\ra +\alpha_0^2

1985:       \la \chi,|\phi_1|^2\ol{\phi_1}-|\Psi_1|^2\ol{\Psi_1}\ra\\ c.c.

1986:            \end{array}\right)\nn\\

1987: &&\ \ +\lambda\sigma_3

1988: \left(\begin{array}{c}

1989: \alpha_0\la \chi,|\phi_1|^2\ra +\ol{\alpha_0}\la \chi,\phi_1^2\ra

1990:       +\alpha_0^3\la \chi,\ol{\phi_1}^2\ra\\ c.c.

1991:            \end{array}\right)\nn\\

1992: &&\ \ - \D_t\tTheta\ \sigma_3

1993: \left(\begin{array}{c}

1994: \la\chi,\phi_2\ra+\alpha_0^2\la\chi,\ol{\phi_2}\ra\\ c.c.\end{array}\right)

1995: + i\left(\begin{array}{c} \la\D_t(\sigma_3\xi_\mu),\Phi_2\ra\\ c.c.

1996:          \end{array}\right)

1997: \label{eq:a1-sys}

1998: \ea

1999: 

2000: The matrices ${\bf M_{jk}},\ {\bf N_{j1}},\  j,k=0,1$ are displayed in

2001:  equations (\ref{eq:M0k}-\ref{eq:N11}).

2002:   For the (generic) case where we expect $|\alpha_0|$

2003: to approach a nonzero limit and $\alpha_1$ and $\phi_2$ to decay to zero,

2004: we shall use the expansion: 

2005: 

2006: \ba

2007: {\bf M_{00}}

2008:  &=& \left(\begin{array}{cc}

2009:      2(1+\cO_4^{(0)}) & \alpha_0^2\cO_0^{(0)}

2010: \\   \ol{\alpha_0}^2\cO_0^{(0)} & 2(1+\cO_4^{(0)})

2011:          \end{array}

2012:    \right)\nn\\

2013: {\bf M_{11}}

2014:  &=& \left(\begin{array}{cc}

2015:      1+\cO_2^{(0,1)} & \cO(\alpha_0^2)+ \cO(\alpha_1^2) 

2016: \\    \cO(\ol{\alpha_0}^2)+ \cO(\ol{\alpha_1}^2) & 1+\cO_2^{(0)}

2017:          \end{array}

2018:    \right)\nn\\

2019: \ea

2020: The matrices ${\bf M_{10}}$ and ${\bf M_{10}}$ are higher order in $\alpha_0$

2021: and satisfy:

2022: \be

2023: {\bf M_{10}},\ {\bf M_{01}}

2024:  = \left(\begin{array}{cc}

2025:      \cO_2^{(0)} & \alpha_0^2\cO_0^{(0)}

2026: \\   \ol{\alpha_0}^2\cO_0^{(0)} & \cO_2^{(0)}

2027:          \end{array}

2028:    \right)\label{eq:M10M01}

2029: \ee

2030: The matrices ${\bf N_{01}}$ and ${\bf N_{11}}$

2031: satisfy:

2032: \ba

2033: {\bf N_{01}}

2034:  &=& \left(\begin{array}{cc}

2035:      \cO_2^{(0)}+\cO_2^{(1)} & \alpha_0^2(\cO_2^{(0)}+\cO_2^{(1)})

2036: \\   \ol{\alpha_0}^2(\cO_2^{(0)}+\cO_2^{(1)}) & \cO_2^{(0)}+\cO_2^{(1)} 

2037:          \end{array}

2038:    \right)\nn\\

2039: {\bf N_{11}}

2040:  &=& \left(\begin{array}{cc}

2041:      1+\cO_2^{(0)} & \alpha_0^2\cO_0^{(0,1)}

2042: \\   \ol{\alpha_0}^2\cO_0^{(0,1)} & 1+\cO_2^{(0)}

2043:          \end{array}

2044:    \right)\nn

2045: \ea

2046: %\begin{rmk} \bXYZ

2047: % Note that we shall require different expansions in the (nongeneric)

2048: % case where $|\alpha_0|\to0$, where we show $|\alpha_1|$ approaches a constant  

2049: %and $\phi_2$ decays to zero. See section \bXYZ.

2050: %\end{rmk}

2051: \bigskip

2052: 

2053: \nit{\bf Simplifications to equations (\ref{eq:me1}-\ref{eq:me2})}

2054: 

2055: \nit (1) Since ${\bf M_{10}}$ and ${\bf M_{01}}$ are higher order in $\alpha_0$

2056: we can eliminate $\D_t\alpha_1$ from (\ref{eq:a0-sys}) and 

2057:  $\D_t\alpha_0$ from (\ref{eq:a1-sys}).

2058: 

2059: \nit (2) Note also, that "commutator terms" with factors like

2060:   $\D_t|\alpha_0|^2$ or $\D_t\alpha_0^2$ can be eliminated via redefinition

2061: of the near identity 

2062:  matrix ${\bf M_{00}}$ through incorporation of a higher order correction.

2063: 

2064: \nit (3)  

2065: We can eliminate the term proportional to $|\alpha_0|^2\D_t\gamma_0$

2066: as follows.

2067: Consider the last two terms of (\ref{eq:a0-sys}). Since $\D_t|\alpha_0|^2

2068:  = \alpha_0\D_t\ol{\alpha_0} + \alpha_0\D_t\ol{\alpha_0}$ we can incorporate

2069: the second to last term of (\ref{eq:a0-sys}) as a higher order

2070:  correction to the near identity matrix $\bf{M_{00}}$, $\tilde{\bf{M_{00}}}$.

2071: Our goal is now to eliminate the eliminate $\D_t\gamma_0$ from the equation.

2072: The system now be written as form:

2073: \be

2074:  i\D_t\valpha_0

2075:  = \tilde{\bf{M_{00}}}^{-1}\left(\  ...\ +\

2076:  i \cO(|\alpha_0|^2)\D_t\gamma_0\right)

2077: \label{eq:hey1}\ee

2078: The first component has the form:

2079: \ba &&i\D_t\alpha_0 = { 1^{\rm st}\ {\rm component}\ {\rm of\ the\ vector:}}

2080:  \nn\\

2081: && \tilde{\bf{M_{00}}}^{-1}(...) +

2082:  i\cO(|\alpha_0|^2) \D_t\gamma_0 \tilde{\bf{M_{00}}}^{-1}

2083: \label{eq:hey2}

2084: \ea

2085: Since $\alpha_0=|\alpha_0|e^{i\gamma_0}$ we have

2086: \ba && i\D_t |\alpha_0| - |\alpha_0|\D_t\gamma_0

2087:  = e^{-i\gamma_0}\times

2088:  { 1^{\rm st}\ {\rm component}\ {\rm of\ the\ vector:}}\nn\\

2089: && \tilde{\bf{M_{00}}}^{-1}(...) +

2090:  i\cO(|\alpha_0|)\ |\alpha_0| \D_t\gamma_0 \tilde{\bf{M_{00}}}^{-1}

2091: \label{eq:hey3}

2092: \ea

2093: By taking the real part of (\ref{eq:hey3}), for $|\alpha_0|$ small,

2094:  we can solve for $|\alpha_0|\D_t\gamma_0$.

2095: This enables us to eliminate it from equation (\ref{eq:a0-sys}) as a higher 

2096:  order term.

2097: \bigskip

2098: 

2099: Implementation of these simplifications leads to the following

2100: \begin{prop}

2101: \ba

2102:  i\D_t\valpha_0 &=& 

2103:  {\bf M_{00}^\#}\left[

2104:  -(E_{01}+\D_t\tTheta)\sigma_3{\bf N_{01}}\valpha_1\right.\nn\\

2105: &&\  +\lambda\cO_0^{(0)}\sigma_3\left(\begin{array}{c}

2106:                        |\alpha_0|^2\alpha_1+\alpha_0^2\ol{\alpha_1}\\ c.c.

2107:                          \end{array}\right)\nn\\

2108: &&\ \ + \lambda\sigma_3\left(\begin{array}{c}

2109: \la\chi_0^{(0)},|\phi_1|^2\phi_1-|\Psi_1|^2\Psi_1\ra

2110:  + \alpha_0^2\la\chi_0^{(0)},

2111:  |\phi_1|^2\ol{\phi_1}-|\Psi_1|^2\ol{\Psi_1}\ra \\ c.c.

2112:                          \end{array}\right)\nn\\

2113: &&\ \left. -\D_t\tTheta

2114: \left(\begin{array}{c}

2115: \la\chi_0^{(0)},\phi_2\ra +\alpha_0^2\la\chi_0^{(0)},\ol{\phi_2}\ra

2116: \\ c.c.  \end{array}\right) \ \right]

2117: \nn\\

2118: \label{eq:a0eqn}\\

2119: i\D_t\valpha_1 &=& {\bf A}(t)\valpha_1\nn\\

2120: %%

2121: % {\bf M_{11}^\#}\left(\ -(E_{01}+\D_t\tTheta)\sigma_3{\bf N_{11}}\valpha_1

2122: %\right.\nn\\

2123: %&&\  +  \lambda\cO_0^{(0,1)}\sigma_3\left(\begin{array}{c}

2124: %                       |\alpha_0|^2\alpha_1+\alpha_0^2\ol{\alpha_1}\\ c.c.

2125: %                         \end{array}\right)\nn\\

2126: %%

2127: &&\ \ +{\bf M_{11}^\#}\left[\ \lambda\sigma_3\left(\begin{array}{c}

2128:  \la \chi,|\phi_1|^2\phi_1-|\Psi_1|^2\Psi_1\ra +\alpha_0^2

2129:       \la \chi,|\phi_1|^2\ol{\phi_1}-|\Psi_1|^2\ol{\Psi_1}\ra\\ c.c.

2130:            \end{array}\right)\ \right.\nn\\

2131: &&\ \ +\lambda\sigma_3

2132: \left(\begin{array}{c}

2133: \alpha_0\la \chi,|\phi_1|^2\ra +\ol{\alpha_0}\la \chi,\phi_1^2\ra

2134:       +\alpha_0^3\la \chi,\ol{\phi_1}^2\ra\\ c.c.

2135:            \end{array}\right)\nn\\

2136: &&\ \left. - \D_t\tTheta\ \sigma_3

2137: \left(\begin{array}{c}

2138: \la\chi,\phi_2\ra+\alpha_0^2\la\chi,\ol{\phi_2}\ra\\ c.c.\end{array}\right)

2139: \ \right]

2140: \label{eq:a1eqn}

2141: \ea

2142: Here, 

2143: \ba

2144: \bA\ &=&\ \left(\begin{array}{cc}

2145:  E_{10}+ \cO_0^{(0,1)}|\alpha_0|^2 &\cO_0^{(0,1)}\alpha_0^2\\

2146: -\cO_0^{(0,1)}\ol{\alpha_0}^2 & -E_{10}-\cO_0^{(0,1)}|\alpha_0|^2

2147: \end{array}\right)\nn\\

2148: \label{eq:Adef}\\

2149: \bA_{22} &=& -\bA_{11},\ \bA_{21}=-\ol{\bA_{12}},\ {\rm and}\nn\\

2150: \phi_1 &=& \Psi_1 + \phi_2\nn

2151: \ea

2152: ${\bf M_{00}^\#}$ and ${\bf M_{11}^\#}$ are near-identity matrices,

2153: whose deviations from the identity give rise to higher order terms which 

2154: are subordinate to the leading order behavior obtained in the analysis which

2155: follows.

2156: 

2157: \nit Finally, 

2158:  equations (\ref{eq:a0eqn}) and (\ref{eq:a1eqn}) are coupled to the equation

2159: for $\Phi_2$ given by (\ref{eq:Phi2_eqn}).

2160: \end{prop}

2161: 

2162: %\subsection{Interlude on strategy I.}

2163:  

2164: \subsection{Peeling off the rapid oscillations of  $\alpha_1$}

2165: \label{subsection:Peel}

2166: 

2167: Fix $T>0$ and large. We rewrite (\ref{eq:a1eqn}) centered about 

2168: the ground state at time $t=T$:

2169: \be

2170: i\D_t\valpha_1 =\bA(T) \valpha_1 + \left(\bA(t)-\bA(T)\right)\valpha_1 +

2171: \vec F,

2172: \label{eq:alpha1_eqnE}

2173: \ee

2174: where

2175: \be

2176: \bA(T)\ =\ \left(\begin{array}{cc}

2177:  E_{01}(T)+ \cO_0^{(0,1)}(T)|\alpha_0(T)|^2 &\cO_0^{(0,1)}(T)\alpha_0^2(T)\\

2178: -\cO_0^{(0,1)}(T)\ol{\alpha_0}^2(T) & -E_{01}(T)-\cO_0^{(0,1)}(T)|\alpha_0(T)|^2

2179: \end{array}\right)

2180: \label{eq:AT}

2181: \ee

2182: 

2183: Since $\bA(T)$ is a constant coefficient matrix it is a simple matter

2184: to obtain the fundamental matrix.

2185: \begin{prop}\label{prop:xdef}

2186: The system

2187: \be

2188: i\D_t\vec\alpha_1\ =\ \bA(T)\vec\alpha_1\label{eq:homogalpha1}\ee

2189: has a fundamental solution matrix:

2190: \ba

2191: && \bX(t)\ =\ \left(\begin{array}{cc} c_{11}^+ & c_{12}^-\\

2192:                                    c_{21}^+ & c_{22}^- \end{array}\right)

2193:          \left(\begin{array}{cc} e^{-i\lambda_+t} & 0\\

2194:                                  0   & e^{-i\lambda_-t}\end{array}\right)

2195: \nn\\

2196: &=& \left(\begin{array}{cc}

2197:  1  & \cO_0^{(0,1)}(T)\alpha_0(T)^2E_{10}^{-1}(T) \\

2198: \cO_0^{(0,1)}(T)\ol{\alpha_0(T)}^2 E_{10}^{-1}(T) & 1

2199: \end{array}\right)

2200:   \left(\begin{array}{cc} e^{-i\lambda_+t} & 0\\

2201:     0   & e^{-i\lambda_-t}\end{array}\right)

2202: \nn\\

2203: \label{eq:Xdef}

2204: \ea

2205: The eigenfrequencies, $\lambda_\pm(T)$ are given by $\lambda_+(T)$

2206: and $\lambda_-(T)=-\lambda_+(T)$, where:

2207: \be \lambda_+(T)\ =\ E_{10}(T)+

2208:  \cO_0^{(0,1)}|\alpha_0(T)|^2,\

2209: \label{eq:lambda_def}

2210: \ee

2211: where $E_{10}=E_1-E_0$, and provided $|\alpha_0(T)|^2/E_{10}(T)$

2212:   is sufficiently small.

2213: \end{prop}

2214: \medskip

2215: 

2216: We use the fundamental matrix, $X(t)$, to define a change of variables:

2217: \be

2218: \vec\alpha_1\ \equiv

2219:              \ \bX(t)\vec\beta\ =\ \bX(t)\left(\begin{array}{c}\beta_1\\

2220:                                                           \ol{\beta_1}

2221:                                            \end{array}\right)

2222: \label{eq:beta_def}\ee

2223: Therefore,

2224: \be

2225: \alpha_1 = e^{-i\lambda_+t}\beta_1 + \cO_0^{(0,1)}\alpha_0^2(T)E_{10}^{-1}(T)

2226:        e^{-i\lambda_-t}\ol{\beta_1}.

2227: \label{eq:a1-b1}\ee

2228: Then, $\vec\beta$ satisfies

2229: \be

2230: i\D_t\vec\beta\ = \ \bX^{-1}(t)\vec F(\bX(t)\vec\beta(t),t)

2231:  \label{eq:vecbeta_eqn}\ee

2232: Note that since the linear in $\alpha_1$ terms have been removed by the change

2233: of variables (\ref{eq:beta_def}), $\D_t\vec\beta_1\sim \cO(|\beta_1|^2)$.

2234: 

2235: Note that by (\ref{eq:Xdef})

2236: \be

2237: \bX^{-1}(t)\ =\

2238:         \left(\begin{array}{cc} e^{i\lambda_+t} & 0\\

2239:                                 0 & e^{i\lambda_-t}\end{array}\right)

2240: \left(\begin{array}{cc} 1+\cO_0^{(0,1)}(T)|\alpha_0(T)|^4 E_{10}^{-2}(T) &

2241:                                -\cO_0^{(0,1)}(T)\alpha_0^2(T)E_{10}^{-1}(T)\\

2242:                    -\cO_0^{(0,1)}(T)\ol{\alpha_0(T)}^2E_{10}^{-1}(T) &

2243:                1+\cO_0^{(0,1)}(T)|\alpha_0(T)|^2E_{10}^{-2}(T)

2244:                      \end{array}\right)

2245: \label{eq:Xinverse}\ee

2246: 

2247: Written out in detail, from (\ref{eq:vecbeta_eqn}) and the various definitions,

2248: we have

2249: \begin{prop}

2250: The equation for $\beta_1$ has the form

2251: \ba

2252: i\D_t\beta_1\ &=&\ \lambda e^{i\lambda_+t}

2253: \la \chi_0, 

2254: |e^{-i\lambda_+t}\beta_1\psi_{1*}+\phi_2|^2\ra

2255:  \alpha_0\nn\\

2256: &+&\ \lambda e^{i\lambda_+t}\langle \chi_0,

2257:  (e^{-i\lambda_+t}\beta_1\psi_{1*}+\phi_2)^2\rangle\ol{\alpha_0}\nn\\

2258: &+&\ \lambda e^{i\lambda_+t}\la \chi_0, (e^{i\lambda_+t}\ol{\beta_1}\psi_{1*}+

2259:  \ol{\phi_2})^2\ra\alpha_0^3\nn\\

2260: &+&\ \lambda\langle\chi_0 ,

2261:   \left[ |e^{-i\lambda_+t}\beta_1\psi_{1*}+\phi_2|^2

2262:         (e^{-i\lambda_+t}\beta_1\psi_{1*}+\phi_2) -

2263:               e^{-i\lambda_+t}|\beta_1|^2\beta_1\psi_{1*}^3\right]\ 

2264:  \rangle\nn\\

2265: &+& {\cal R}_\beta, \ {\rm where}

2266:  \label{eq:beta_eqn2}\\

2267: \cR_\beta &=& \bX^{-1}(t)\left[ (\bA(t)-\bA(T))\bX(t)\vec\beta_1+\cR_1\right]

2268: \label{eq:Rbeta_def}

2269: \ea

2270: \end{prop}

2271: 

2272: \begin{rmk}

2273: \be

2274: \bX^{-1}(t)[ \bA(t) - \bA(T) ] \bX(t) = 

2275:  {\rm \ Real \ Symmetric\ Diagonal}\ {\bf S}(t) + e^{-2i\lambda_+t} {\bf B}(t)

2276: \nn\ee

2277: Therefore, the nonoscillatory part involving {\bf S(t)} does not effect the 

2278: evolution of $|\beta_1|^2$.

2279: \end{rmk}

2280: 

2281: \subsection{Expansion of $\phi_2$}

2282: 

2283: The next step is to get an appropriate expansion of $\phi_2$,

2284:  which upon substitution into (\ref{eq:beta_eqn2}) can be used

2285:  to isolate the key

2286: resonant terms in the $\beta_1$ equation.

2287:   The analogous steps are then repeated for the $\alpha_0$ equation. Finally,

2288: a near-identity change of variables is constructed which maps the system for

2289: $\alpha_0$ and $\beta_1$ to a new system (a {\it normal form} plus corrections)

2290:  for which the dynamical behavior is more transparent.

2291: 

2292: $\Phi_2$ solves equation (\ref{eq:Phi2_eqn}). We shall require dispersive

2293: decay estimates for $\Phi_2$ and these are most naturally obtained relative to

2294: a time independent Hamiltonian.

2295:  Since $\alpha_0(t)$ is expected to tend to a limit as

2296:  $t\to\infty $ and since we are fixing a time interval $[0,T]$, it is natural

2297: to use as {\it reference Hamiltonian}, the operator ${\cal H}_0(T)$. We now make

2298:  use of the linear spectral theory of section 3 and decompose $\Phi_2$ into a

2299:  part lying in the ``discrete subspace of ${\cal H}_0(T)$'':

2300: \be

2301: N_g({\cal H}_0(T))\oplus N({\cal H}_0(T)-\mu(T)) \oplus N({\cal H}_0(T)+\mu(T))

2302: \label{eq:discrete_subspace}\ee

2303: and a part lying 

2304:  in ${\cal M}_1(T)$, the ``dispersive subspace of ${\cal H}_0(T)$'';

2305:  see (\ref{eq:M1def}).

2306: 

2307: Let \cite{BP2,Cucc2} 

2308: \be \Phi_2\ =\ k\ +\ n\ +\ \eta,\label{eq:Phi2kneta}\ee

2309:    where

2310: \ba

2311: k &\equiv&

2312:   \sum_{\xi_a\in N_g({\cal H}_0(T))}\langle\sigma_3\xi_a(T),\Phi_2\rangle\xi_a

2313:  \label{eq:k_def}\\

2314: n &\equiv&

2315:   \sum_{\xi_b\in N({\cal H}_0(T)\mp\mu(T))}\langle\sigma_3\xi_b(T),\Phi_2\rangle

2316: \xi_b

2317:  \label{eq:n_def}\\

2318: \eta &\equiv& P_c(T)\Phi_2\ =\ \Phi_2-k-n

2319: \label{eq:eta_def}

2320: \ea

2321: Since

2322: \ba \la\sigma_3\xi_a(t),\Phi_2(t)\ra&=&0,\ \xi_a\in N_g({\cal H}_0(t))\nn\\

2323: \la\sigma_3\xi_b(t),\Phi_2(t)\ra&=&0,\ \xi_b\in N_g({\cal H}_0(t)\mp\mu(t)),

2324: \nn\ea

2325: $\xi(t)$ may be replaced by $\xi(t)-\xi(T)$ in the definitions of $k$ and $n$.

2326: Inserting the expansion (\ref{eq:Phi2kneta}) into (\ref{eq:k_def})

2327:  and (\ref{eq:n_def}) and defining

2328: \ba p_{\rm null}(t,T) &=& \sum_{\xi_a\in N_g({\cal H}_0(T))}

2329:  \langle\sigma_3(\xi_a(T)-\xi_a(t)),\cdot\rangle\xi_a\label{eq:pnull_def}\\

2330: p_{\rm neut}(t,T) &=& \sum_{\xi_b\in N({\cal H}_0(T)\mp\mu(T))}

2331:  \langle\sigma_3(\xi_b(T)-\xi_b(t)),\cdot\rangle\xi_a\label{eq:pneutral_def}

2332: \ea

2333: we have that $k$ and $n$ may be expressed in terms of $\eta$ as follows:

2334: Then,

2335: \be

2336: \left[\ I\ -\ \left(\begin{array}{cc} p_{\rm null}(t,T) & p_{\rm null}(t,T)\\

2337:                                        p_{\rm neut}(t,T) & p_{\rm neut}(t,T)

2338:                      \end{array}\right)\ \right]

2339:         \left(\begin{array}{c} k \\ n\end{array}\right)\ =\

2340: \left(\begin{array}{c} p_{\rm null}(t,T)\eta \\ p_{\rm neut}(t,T)\eta

2341:    \end{array}\right).

2342: \label{eq:knofeta}

2343: \ee

2344: Therefore, we have

2345: \begin{prop}\label{prop:phi2kneta}

2346: There exists $\varepsilon_0>0$ such that if $|\alpha_0(t)-\alpha_0(T)|

2347:  <\varepsilon_0$ then the relation (\ref{eq:knofeta}) can be inverted and

2348: we have

2349: \be \Phi_2\ =\ \Phi_2[\eta]\ =\ k[\eta]+n[\eta]+\eta,\nn\ee

2350: where $\Phi_2$ is linear in $\eta$ and continuous in the weighted

2351:  (local decay)  norm

2352:   of $f\mapsto \|\la x\ra^{-\sigma}f\|_2$.

2353: \end{prop}

2354: The statement about continuity in the weighted norm

2355: follows from the spatial localization of the generalized eigenfunctions.

2356: 

2357: \begin{rmk}\label{rmk:knalpha}

2358:  Note that since $|\xi_a(T)-\xi_a(t)|\le

2359:  C\cE_0^{1\over2}|\alpha_0(T)-\alpha_0(t)|,$

2360: we have the simple estimate:

2361: \be |k|,\ |n|\ \le\ C |\alpha_0(T)-\alpha_0(t)|\  

2362:  ||\la x\ra^{-\sigma}\eta(t)||_2.\nn\ee

2363: Anticipating that for $t$ very large,

2364:   $|\alpha_0(T)-\alpha_0(t)|\sim t^{-{1\over2}}$ and

2365: $||\la x\ra^{-\sigma} \Phi_2(t)||_2\sim  t^{-{1\over2}}$, it follows that

2366: $|k|,\ |n| \sim t^{-1}$. Thus,  $|k|$ and $|n|$,  

2367:  are expected to decay faster than

2368:  $\eta$.

2369: \end{rmk}

2370: Finally, $\eta$ satisfies the following evolution equation obtained from

2371: (\ref{eq:Phi2_eqn}) by explicitly introducing the reference Hamiltonian,

2372:  ${\cal H}_0(T)$, and applying the projection  $P_c(T)$ to the equation.

2373: \ba

2374: i\D_t\eta\ &=&\ {\cal H}_0(T)\eta\nn\\

2375: &+& \left( {\cal H}_0(t)-{\cal H}_0(T) \right)\eta +

2376:   \left( E_1(t)-E_0(t) \right)P_c(T)\sigma_3

2377:              \left(\begin{array}{c}\Psi_1(t)\\ c.c.\end{array}\right)

2378: \nn\\

2379: &+&\ \lambda P_c(T)\sigma_3

2380:  \left(\begin{array}{c}2|\Psi_0|^2\Psi_1+\Psi_0^2\ol{\Psi_1} \\

2381:  c.c. \end{array}\right)\nn\\

2382: &+& \lambda P_c(T) \sigma_3\left(\begin{array}{c}2\Psi_0|\Psi_1+

2383:  \pi_1\Phi_2[\eta]|^2 +

2384:  \ol{\Psi_0}(\Psi_1+\pi_1\Phi_2[\eta])^2\\

2385:  c.c. \end{array}\right)\nn\\

2386: &+&\ \lambda P_c(T)\sigma_3\left(\begin{array}{c}|\Psi_1+\pi_1\Phi_2[\eta]|^2

2387:  (\Psi_1+\pi_1\Phi_2[\eta]) -

2388:  |\Psi_1|^2\Psi_1\\

2389:  c.c. \end{array}\right)\nn\\

2390: &-&\ i P_c(T)\left(\begin{array}{c}

2391:  \vec\nabla_1\Psi_1(t)\cdot\D_t\vec\alpha_1(t) +

2392:  \vec\nabla_0\Psi_0(t)\cdot\D_t\vec\alpha_0(t)\\

2393:  c.c. \end{array}\right)

2394: \label{eq:eta_eqn}

2395: \ea

2396: 

2397: \section{Normal Form and Master Equations}\label{sec:nfandmaster}

2398: 

2399: In section 4 we decomposed the solution, $\phi$, in terms of coordinates

2400:  $\alpha_0(t)$ and $\alpha_1(t)$ along manifolds of nonlinear bound states

2401: and $\phi_2$, a correction which lies in a time-dependent subspace, ${\cal M}_1(t)$,

2402:  of continuum modes. $\phi_2(t)$ was then decomposed into its discrete ($k$ and $n$)

2403:  and continuous ($\eta$) components with respect to a time \underline{independent}

2404:  Hamiltonian, ${\cal H}_0(T)$. We also observed that $k$ and $n$ are determined  

2405:  by and are expected to be more rapidly decaying than  $\eta$ (Proposition 

2406:  \ref{prop:phi2kneta} and Remark \ref{rmk:knalpha}).

2407:   Therefore, the evolution of $\phi$ is determined  

2408:  by $\alpha_0$, $\alpha_1$ and 

2409:  $\eta$. Finally, the fast oscillations of $\alpha_1$ are removed by the introduction

2410:  of $\beta_1\ =\ [ X(t)\vec\alpha_1 ]_1\sim e^{-i\lambda_+t}\alpha_1$. 

2411: 

2412: We now seek a form of the system for $\alpha_0$ and $\beta_1$ from which the large 

2413: time dynamics can be deduced. We obtain this ``normal form'' by first solving

2414:  for $\eta$ (see (\ref{eq:etaformal}),(\ref{eq:eta_eqn}) )

2415:  as a functional of $\alpha_0$, $\beta_1$ and the initial data 

2416:  $\eta(0)$

2417:   and then substituting an appropriate expansion (see sections 

2418:  \ref{sec:nfandmaster} and \ref{sec:stab}) 

2419: into  the equations for $\alpha_0$ and $\beta_1$. 

2420: 

2421: \begin{prop}\label{prop:normalform} {\bf The Normal Form}

2422: There exists a near identity change of variables

2423: \be

2424:  \left(\begin{array}{c}\tilde\alpha_0 \\ \tilde\beta_1\end{array}\right)

2425: \equiv \left(\begin{array}{c}\alpha_0 \\ \beta_1\end{array}\right)\ 

2426: +\ \left(\begin{array}{c}J_\alpha[\alpha_0,\beta_1,t] \\ 

2427:             J_\beta[\alpha_0,\beta_1,t]\end{array}\right)\

2428: \label{eq:ata}\ee

2429: where 

2430: \be J_k[\alpha_0,\beta_1,t]={\cal O}(|\alpha_0|^2+|\beta_1|^2),\ k=\alpha,\beta

2431:  \nn\ee

2432: and bounded uniformly in $t$, and such that

2433: \ba

2434: i\D_t\ta0\ &=&\ (c_{1022}+i\Gamma_{\omega})\ 

2435:  |\tb1|^4\ta0\ +\ 

2436:  F_\alpha[\ta0,\tb1,\eta,t]\nn\\

2437: i\D_t\tb1\ &=&\ (c_{1121}-2i\Gamma_{\omega})

2438:  |\ta0|^2|\tb1|^2\tb1\ 

2439:  +\ F_\beta[\ta0,\tb1,\eta,t].

2440: \label{eq:normalform}\ea

2441: The properties of $F_\alpha$ and $F_\beta$ are briefly discussed

2442: %\footnote{\bXYZ more detail}

2443: in the remark

2444:  \ref{rmk:approach} following Corollary

2445:  \ref{cor:nlPME} below and described in detail in

2446:  section \ref{sec:stab}.

2447: 

2448: Furthermore, 

2449: \ba \Gamma\ &=&\ \Gamma_{\omega_*}\ + {\cal O}(|\alpha_0(T)|^2) > 0,

2450:  \ {\rm where},\nn\\

2451:  \Gamma_{\omega_*} &\equiv& \lambda^2

2452: \pi\langle\psi_{0*}\psi_{1*}^2,\delta(H-\omega_*)\psi_{0*}\psi_{1*}^2\rangle

2453:  > 0\label{eq:Gamma}\nn\\

2454: \omega_*\ &=&\ 2E_{1*}-E_{0*}\label{eq:omegastar}

2455: \ea

2456: The coefficients $c_{klmn}=\cO_0^{(0,1)}$

2457:  are real constants multiplying monomials

2458: of the form $\alpha_0^k\bar\alpha_0^l\beta_1^m\bar\beta_1^n$.

2459: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2460: %\footnote{

2461: %\bXYZ 

2462: %For any $m\ge3$, there exist constants $C_\alpha$ and $C_\beta$ such that

2463: %\ba

2464: %|F_\alpha| &\le& 

2465: %C_\alpha\left(1+\cO(|\alpha_0|)\right)

2466: % \times \left(\ |\ta0||\tb1|^{2m}  + 

2467: % |\ta0|^2 |\tb1|^4\  +\  

2468: %  |\la\chi,|\eta_0| + |\eta_{2b}|\ra|\ |\tb1|^2\right.\nn\\

2469: %&+&

2470: %\left. |\la\chi,|\eta_0|^3 +|\eta_{2b}|^3\ra|

2471: %\right)

2472: %\label{eq:Falpha_est}\\

2473: %|F_\beta| &\le& C_\beta \left(1+\cO(|\alpha_0|)\right)

2474: %\times \left(\ |\ta0||\tb1|^{2m} + 

2475: %  |\ta0|^2|\tb1|^4\ +\ 

2476: % |\la\chi,\cH_0(T)\eta_0(t)\ra| \right.\nn\\

2477: % &+&\left. \la\chi, |\eta_{2b}|\ra|\tb1|^2\  

2478: % +\ |\la\chi,|\eta_0|^3 +|\eta_{2b}|^3\ra|\ 

2479: %\right)

2480: %\label{eq:Fbeta_est}

2481: %\ea

2482: %}

2483: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2484: \end{prop}

2485: 

2486: %\begin{rmk}\footnote{\bXYZ\ revise} A related {\it local }  

2487: %  normal form is derived in \cite{BP2}, \cite{Cucc2},

2488: % \cite{TY} which describes the dynamics in a small neighborhood of

2489: % a ground state. It does not give information on the behavior of solutions

2490: % with general initial data, {\it e.g.} 

2491: %  with large excited state and relatively small ground state.

2492: % In contrast, the normal form we have introduced describes

2493: %the dynamics in a full neighborhood of the origin.   

2494: %\end{rmk}

2495: \begin{rmk}

2496: Due to our choice of the phase correction, $\tTheta(t)$ 

2497:  (see (\ref{eq:tTheta}), a term of the

2498: form $c_{1011}\cO_0^{(0,1)}|\tb1|^2\ta0$,

2499:   is absent from the differential equation for

2500: $\ta0$ in (\ref{eq:normalform}).

2501: \end{rmk}

2502: 

2503: 

2504: Now  let

2505: \be P_0\equiv |\tilde\alpha_0|^2 \ {\rm and}\

2506:     P_1\equiv |\tilde\beta_1|^2

2507: \label{eq:power_def}\ee

2508:  denote the (renormalized) ground state and excited state powers.

2509: Then, by (\ref{eq:normalform}) we have the {\it Nonlinear Master Equation}:

2510: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2511: \begin{cor}\label{cor:nlPME}

2512: \ba

2513: {dP_0\over dt} &=& 2\Gamma P_0P_1^2\  +\  R_0

2514:  \label{eq:P0eqn}\\

2515: {dP_1\over dt} &=& -4\Gamma P_0P_1^2 \ +\ R_1.

2516: \label{eq:P1eqn}

2517: \ea

2518: where

2519: \ba R_0 &=& R_0[\ta0,\tb1,\eta,t]=  2\Im (\ol{\ta0}F_\alpha),\nn\\

2520:     R_1 &=& R_1[\ta0,\tb1,\eta,t]=  2\Im (\ol{\tb1}F_\beta)

2521: \label{eq:Rjdef}

2522: \ea

2523: \end{cor}

2524: A more precise and revealing variant of Corollary \ref{cor:nlPME} is 

2525:  Proposition \ref{prop:Master}, which is

2526: stated and proved in section \ref{sec:stab}.

2527: 

2528: \bigskip

2529: 

2530: \begin{rmk}\label{rmk:approach}

2531: The terms $F_\alpha$ and $F_\beta$ are such that $R_0$ and $R_1$ are 

2532: not small perturbations of the leading order terms in (\ref{eq:P0eqn},

2533: \ref{eq:P1eqn}) for all $t\ge0$. In fact there are three time intervals 

2534: defined in terms of transition times $t_0$ and $t_1$ 

2535:  (see section \ref{sec:stab}), 

2536:   in which we consider the system (\ref{eq:P0eqn}-\ref{eq:P1eqn}): $I_0=[0,t_0]$,

2537:  $I_1=[t_0,t_1]$ and $I_2=[t_1,\infty)$.

2538:  It is only for sufficiently large time, ($t\in I_2$), 

2539:  where $R_0$ and $R_1$ are negligible.

2540:  The behavior on short ($0\le t\le t_0$) and intermediate ($t_0\le t\le t_1$)

2541: time scales can be very different.

2542:  We go into the details

2543: of $R_0$ and $R_1$ in section~\ref{sec:stab} but wish to make some remarks 

2544: at this stage which indicate our approach. 

2545: 

2546: If we drop the terms $R_j$

2547: then we have a flow, which evolves in the first quadrant of the $P_0-P_1$

2548: plane according to:

2549: \ba

2550: {dp_0\over dt} &=& 2\Gamma p_0p_1^2\  

2551:  \label{eq:p0eqn}\\

2552: {dp_1\over dt} &=& -4\Gamma p_0p_1^2,

2553: \label{eq:p1eqn}

2554: \ea

2555:  where solutions for typical data converge to the $p_0$ axis with a rate

2556: $\la t\ra^{-1}$. In order for the  corrections coming from $R_0$ and $R_1$ to 

2557: be small, {\it intuitively} it is sufficient that

2558: \be

2559: R_j\ \sim \cE_0^\rho\left(\ P_0 P_1^2 + \la t\ra^{-3}\ \right),

2560: \label{eq:intuitI2}

2561: \ee

2562: where $\cE_0$ is small and $\rho>0$.  

2563: This is what we show for $t\ge t_1$. 

2564:   For the intermediate time range,

2565:  $t_0\le t\le t_1$, we show that the behavior is controlled

2566: by the system:

2567: \ba

2568: \frac{dP_0}{dt}\ &=& 2\Gamma P_0P_1^2 + \cO(\la t\ra^{-3})\nn\\

2569: \frac{dP_1}{dt}\ &=& -4\Gamma P_0P_1^2 + \cO\left(\sqrt{P_0}P_1^m\right)\ 

2570:  +\ \cO(\la t\ra^{-3}),

2571: \label{eq:I1nf}\ea

2572: where $m\ge3$.

2573: Therefore, for intermediate times we need to show: 

2574: \ba

2575: R_0\ &\sim& \cE_0^\rho\left(\ P_0 P_1^2 + \la t\ra^{-3}\ \right)\nn\\

2576: R_1\ &\sim& \cE_0^\rho\left(\ P_0 P_1^2 + \sqrt{P_0}P_1^m + \la t\ra^{-3}\ 

2577:  \right).

2578: \label{eq:intuitI1}

2579: \ea

2580: What makes the analysis subtle is the dependence of $R_j$ on $P_0$ and $P_1$

2581:  in a manner which is {\it nonlocal in time}.  That is, 

2582: \be

2583: \eta \sim \int_0^t e^{-i\cH_0(t-s)}P_c(\cH_0)\chi

2584:   \alpha_0^{m_1}\beta_1^{m_2}\ol{\alpha_0}^{m_3}\ol{\beta_1}^{m_4} \ ds 

2585: \nn\ee

2586: Local in time terms are simple to dominate by the leading terms.

2587:  However, nonlocal terms require

2588: careful analysis. Note in particular, that due to the 

2589:  ``history dependence'' of such terms, being expressed as 

2590: time integrals from $0$ up to $t$, an analysis of the effect of such  

2591: terms for $t\ge t_1$ requires use of estimates on other time regimes

2592:  $t\le t_0$ and $t_0\le t\le t_1$ as well. Furthermore, there are no

2593: decay estimates on either $P_0$ or $P_1$ in the intermediate interval

2594:  $t_0\le t\le t_1$ or on the size of this interval.

2595: \end{rmk}

2596: 

2597: The normal form of Proposition \ref{prop:normalform} is essentially

2598:  the Poincar\'e - Dulac normal form which can be constructed along

2599: the lines explictly implemented in \cite{SW99}; see also \cite{Arnold}.

2600: We now give a detailed outline of the procedure with explicit illustrative

2601:  detail of key points concerning the treatment of resonant and nonresonant

2602: terms.

2603: \medskip

2604: 

2605: \noindent{\bf Resonant terms and removal nonresonant terms:}

2606: %\bXYZ

2607: Here we illustrate, by way of a simple example, how non-resonant terms

2608: can be removed by near identity changes of variables. Consider the 

2609: scalar ordinary differential equation

2610: \be A'(t)\ =\ |A(t)|^2e^{i\Omega t}\label{eq:cODE}\ee

2611: where $A(t)$ is a complex valued function. We shall introduce a change of 

2612: variables $A\mapsto \tA = A + q_2(A,{\bar A}, t)$, where 

2613: $q_2(A,{\bar A},t)=\cO(|A|^2)$ and $q_2(A,{\bar A},t+{2\pi\over\Omega})

2614:  = q_2(A,{\bar A},t)$, which is therefore approximately the identity for 

2615: $|A|$ small, and such that 

2616: \be \tA'(t)\ =\ {i\over\Omega}|\tA(t)|^2\tA(t)+ 

2617:  {i\over\Omega} e^{2i\Omega t}|\tA(t)|^2 \bar{\tA(t)} + 

2618:  E_4(\tA(t),\bar{\tA(t)},t),

2619:  \label{eq:cODE1}\ee

2620: Here, $E_4$ is $2\pi/\Omega$ periodic in $t$ and 

2621: \be E_4(A(t),\tA(t),t)\ =\ \cO(|\tA(t)|^4)\label{eq:E4estimate}\ee

2622: 

2623: The change of variables can be derived by elementary means. Integration 

2624: of (\ref{eq:cODE}) gives:

2625: \ba

2626: A(t)-A(0)\ &=&\ \int_0^t |A(s)|^2e^{i\Omega s}ds\nn\\

2627:            &=&\ \int_0^t |A(s)|^2\frac{1}{i\Omega}\frac{d}{ds}e^{i\Omega s}ds

2628:                                                 \nn\\

2629:            &=&\ |A(s)|^2 \frac{1}{i\Omega}e^{i\Omega s}\left.\right|_0^t\ 

2630:  -\ \int_0^t \frac{1}{i\Omega}e^{i\Omega s}\frac{d}{ds} |A(s)|^2 \ ds\nn\\

2631:            &=&\ |A(s)|^2 \frac{1}{i\Omega}e^{i\Omega s}\left.\right|_0^t\

2632:  - \frac{1}{i\Omega}\int_0^t e^{i\Omega s}\bar{A(s)}|A(s)|^2e^{i\Omega s} ds

2633:  - \frac{1}{i\Omega}\int_0^t |A(s)|^2A(s) ds\nn\\

2634: &&\label{eq:manip}

2635: \ea

2636: Define $\tA(t)= A(t)-(i\Omega)^{-1}|A(t)|^2e^{i\Omega t}$. Then,

2637: $\tA$ satisfies the renormalized ODE, in which resonant quadratic terms

2638: have been removed. The process can be repeated; by introducing further 

2639: changes of variables $\tA\mapsto \tA_1=\tA +$\  higher order in $\tA$ and period

2640: in $t$, non-resonant (oscillatory) cubic terms  can be removed to obtain:

2641: \be

2642: \tA_1'(t)\ =\ \frac{i}{\Omega}|\tA_1(t)|^2\tA_1(t)+ ik|\tA_1(t)|^4\tA_1^(t)

2643:  +\ \dots,

2644: \label{eq:tA1eqn}\ee

2645: where $k$ is real. That the coefficients in the first to terms of this 

2646:  {\it normal form} are purely imaginary implies that, to this order, 

2647:  the amplitude $|\tA_1(t)|$ is independent in time. This is the 

2648:  typical situation of the norm form finite dimensional Hamiltonian systems, 

2649: in which resonances occur between isolated discrete frequencies. 

2650: 

2651: We next examine resonances between discrete and frequencies and the continuum of

2652:  frequencies, associated with the continuous spectral (dispersive) 

2653:  part of $\cH_0$. These can introduce {\it nonconservative} terms into 

2654:  the normal from (via coefficients with real as well as imaginary parts),

2655: which are responsible for energy transfer between discrete modes (bound states)

2656: and radiation. 

2657: 

2658: \medskip

2659: 

2660: \noindent{\bf Nonconservative resonant terms and energy transfer:}\  

2661:  We explain how to find the key resonant 

2662:  energy transfer terms, the leading terms in (\ref{eq:normalform}). 

2663:  These are terms

2664: responsible for the exchange of energy among the nonlinear ground and excited

2665: states mediated by interaction with continuums modes. We focus on the $\beta_1$

2666:  equation. Analogous considerations apply to the $\alpha_0$ equation.

2667: 

2668: Equation (\ref{eq:beta_eqn2}) can be written the following compact form:

2669: \be

2670: i\D_t\beta_1 = \sum_{p,q,r}C_{pqr}\beta_1^p\ol{\beta_1}^qe^{-i\omega_r t}

2671:  \la \chi_{pqr},\eta\ra + \cO(\eta^2) +...,

2672: \label{eq:b1formal}

2673: \ee

2674: where $C_{pqr}$ are of order $1$, $\alpha_0$ or higher order in $\alpha_0$,

2675:  $\omega_r\in\{\pm\lambda_\pm,\pm 2\lambda_\pm, 0\}$, and $\chi_{pqr}$ denote

2676: functions which are exponentially localized in space.

2677: The equation for $\alpha_0$ has a similar structure.

2678: The equation for $\eta = \eta_0 + \eta_1 + \eta_2$, can be formally solved

2679: giving:

2680: \ba

2681: \eta &=& \cO(\eta_0) +  \cO(\eta_0^2) + \cO(\eta^2)\nn\\

2682:  &+& \sum_{p_1, q_1, r_1} D_{p_1q_1r_1}\bG\left(\ \beta_1^{p_1}\ol{\beta_1}^{q_1}

2683:  e^{-i\nu_{r_1} s}\ \tchi_{p_1q_1r_1}\ \right)

2684: \label{eq:etaformal}

2685: \ea

2686: Here, $\bG$ denotes the operator

2687: \be

2688: f\ \mapsto\ \bG f\ \equiv -i\int_0^t e^{-i\cH_0(t-s)}\ P_c\ f(s)\ ds.

2689: \label{eq:G_def}

2690: \ee

2691: We insert the expansion for $\eta$, (\ref{eq:etaformal}) into

2692: the terms involving inner products $\la\chi_{pqr},\eta\ra$

2693:  in (\ref{eq:b1formal}) and in the corresponding equation for

2694:   $\alpha_0$. This yields

2695: a coupled system for $\alpha_0$ and $\beta_1$ which is closed up to 

2696:  higher order. The terms in the resulting equations  are of the form

2697: \be

2698: \sum_{p,q,r}\sum_{p_1,q_1,r_1} C_{pqr} D_{p_1q_1r_1}

2699:  \beta_1^p\ol{\beta_1}^q e^{-i\omega_r t}\left\la\chi_{pqr}\ ,\  

2700:  \bG\beta_1^{p_1}\ol{\beta_1}^{q_1}e^{-i\nu_{r_1} s}\ \tchi_{p_1q_1r_1}\right\ra.

2701: \label{eq:terms}

2702: \ee

2703: We now use the integration by parts lemma:

2704: \begin{lem}\label{lem:ibp}

2705: \ba

2706: \int_0^t e^{iAs} f(s)  ds &=&

2707:  \lim_{\delta\downarrow 0}\int_0^t e^{i(A\pm i\delta)s} f(s) ds\nn\\

2708: &=& -i(A\pm i0)^{-1}e^{iAt}f(t) +i(A\pm i0)^{-1}f(0)\nn\\

2709: &+&i(A\pm i0))^{-1}\int_0^t e^{iAs} f'(s) ds \label{eq:ibp}

2710: \ea

2711: \end{lem}

2712: Applying this lemma, we obtain

2713: \ba

2714: && e^{-i\omega_r t}\left\la\chi_{pqr}\

2715:  \bG\beta_1^{p_1}\ol{\beta_1}^{q_1}e^{-i\nu_{r_1} s}\ \tchi_{p_1q_1r_1}\right\ra

2716: \nn\\

2717: &=& - e^{-i(\omega_r+\nu_{r_1})t}\beta_1^{p_1}\ol{\beta_1}^{q_1}

2718:  \left\la\chi_{pqr}\ , (\cH_0(T)-\nu_{r_1}\mp i0)^{-1} P_c(T)

2719:  \tchi_{p_1q_1r_1}\right\ra

2720: \nn\\

2721: &+& \beta_1^{p_1}(0)\ol{\beta_1}^{q_1}(0)

2722:  \left\la\chi_{pqr}\ , (\cH_0(T)-\nu_{r_1}\mp i0)^{-1}e^{-i\cH_0(T)t}

2723:  P_c(T) \tchi_{p_1q_1r_1} \right\ra\nn\\

2724: &+& \int_0^t\left\la \chi_{pqr}\ , 

2725:  (\cH_0(T)-\nu_{r_1}\mp i0)^{-1}\ e^{-i\cH_0(T)(t-s)} e^{-i\nu_{r_1}s}

2726:   P_c(T) {d\over ds}

2727:  \left( \beta_1^{p_1}\ol{\beta_1}^{q_1}\tchi_{p_1q_1r_1} \right) ds

2728:  \right\ra

2729: \nn\\

2730: \label{eq:terms_exp}

2731: \ea

2732: We first focus on the first term in the expansion (\ref{eq:terms_exp}).

2733: This first contributes a resonant term, which cannot be transformed by  a

2734: near identity transformation  to higher order if

2735: \be \omega_r + \nu_{r_1} = 0.\label{eq:res1}\ee

2736: Now consider such a resonant term. We find that in the $\beta_1$ equation

2737:  they are of the form:

2738: \be

2739:  -|\alpha_0|^{2a_1}|\beta_1|^{2b_1} \beta_1

2740:  \left\la\vec\chi\ , (\cH_0 - \nu_{r_1}\mp i0)^{-1}P_c \vec\chi

2741:  \right\ra\label{eq:sampleres}

2742: \ee

2743: where $\nu_{r_1} = -\omega_r \in \{ \pm\lambda_\pm, \pm 2\lambda_\pm, 0 \}$.

2744: 

2745: There are two cases to consider:

2746: (I) $\nu_{r_1}$ {\it not in} the continuous spectrum of $\cH_0$ and

2747: (II) $\nu_{r_1}$ {\it in} the continuous spectrum of $\cH_0$

2748: \footnote{In case  (II)

2749:  we consider the generic case where if $\nu_{r_1}$ lies in the

2750: interior of the continuous spectrum of $\cH_0$.}.

2751: If $\nu_{r_1}$ is not in the continuous spectrum of $\cH_0$ then

2752: the inner produce in  (\ref{eq:sampleres}) does not involve a singular limit

2753: and we get the limit

2754: \be

2755: \left\la\vec\chi\ , (\cH_0-\nu_{r_1})^{-1}P_c \vec\chi

2756:  \right\ra.

2757: \ee

2758: In this case, the coefficient of $|\alpha_0|^{2a_1}|\beta_1|^{2b_1} \beta_1$ 

2759:  is {\it real}.  

2760: Such a term results only in a nonlinear distortion of the phase of $\beta_1$

2761:  and does not effect the amplitude.

2762: If $\nu_{r_1}$  is in the interior of the continuous spectrum of $\cH_0$

2763: then the limit is singular. We choose the plus sign ($+i0$) if we study

2764: the evolution for $t>0$ and the negative sign ($-i0$) for $t<0$. This choice

2765: related to the condition of outgoing radiation explained below; see

2766: also \cite{SW99}, for example. Evaluation of this singular limit

2767: gives:

2768: \ba

2769: && \left\la\chi\ , (\cH_0-\nu_{r_1}\mp i0)^{-1}P_c \chi

2770:  \right\ra

2771:  = \tLambda + i \tGamma,\ \ {\rm where}\nn\\ 

2772: && \tGamma = \pi\left\la\chi\ , \delta(\cH_0-\nu_{r_1}) \chi\right\ra, \ \

2773: \tLambda = \left\la\chi\ ,\ {\rm P.V.}\ (\cH_0-\nu_{r_1})^{-1}\chi\right\ra

2774: \nn\ea

2775: Contributions to the  imaginary part are therefore

2776:  responsible for a change in amplitude (here damping of $\beta_1$).

2777: 

2778: {\it Now, when does a frequency $\nu$ lie in the continuous spectrum of $\cH_0$?}

2779: By  Proposition \ref{prop:Hspectrum}, we must have $\nu>-E_0=|E_0|$

2780: or $\nu< E_0$. By (\ref{eq:lambda_def}), $\lambda_\pm \sim \pm(E_1-E_0)$.

2781: Since $\nu$ varies over the frequencies $0,\pm\lambda_\pm,\pm 2\lambda_\pm$

2782: we find that $\nu_{r_1}=\pm 2\lambda_\pm = -\omega_r$ resonances are in the

2783:  continuous spectrum and therefore are those

2784: giving rise to energy transfer, provided $2E_1-E_0>0$; 

2785:  see (\ref{eq:omegastar}). 

2786: We now embark on the details.

2787: 

2788: \subsection{Expansion of $\eta$}

2789: 

2790: We expand  $\eta$ as follows:

2791: \be \eta(t)\ =\ \eta_0(t) +\eta_1(t) + \eta_2(t),\label{eq:eta-exp}\ee

2792: where $\eta_0(t)$ corresponds to the linear homogeneous evolution 

2793: with initial data $\eta(0)=P_c(T)\Phi_2(0)$ and $\eta_1$ solves 

2794:  the inhomogeneous linear equation driven by  $\alpha_0$, 

2795:  $\alpha_1$ and $\eta_0(t)$.

2796: 

2797: \noindent\underline{Equation for $\eta_0(t)$:}

2798: \ba

2799: i\D_t\eta_0\ &=&\ {\cal H}_0(T)\eta_0\nn\\

2800: \eta_0(0)\ &=&\ P_c(T)\Phi_2(0)\label{eq:eta0ivp}\ea

2801: Thus,

2802: \be \eta_0(t)\ =\ e^{-i{\cal H}_0(T) t}P_c(T)\Phi_2(0)\label{eq:eta0}\ee

2803: 

2804: \noindent\underline{Equation for $\eta_1(t)$:}

2805: 

2806: \ba

2807: i\D_t\eta_1\ &=&\ {\cal H}_0(T)\eta_1\ +\ 

2808:  P_c(T)\left[{\cal H}_0(t)-{\cal H}_0(T)\right]\Phi_2[\eta_0] 

2809: +\ E_{10}(t)P_c(T)\sigma_3

2810:  \left(\begin{array}{c} \Psi_1\\ c.c.\end{array}\right) \nn\\

2811: &+&\  \lambda P_c(T)\sigma_3

2812: \left(\begin{array}{c} 2|\Psi_0|^2\Psi_1 + \Psi_0^2\ol{\Psi_1}

2813:       \\ c.c. \end{array} \right)\nn\\

2814: &+&\ \lambda P_c(T)\sigma_3

2815: \left(\begin{array}{c}

2816:  2\Psi_0|\Psi_1 +\pi_1\Phi_2[\eta_0]|^2 + 

2817:  \ol{\Psi_0}(\Psi_1+\pi_1\Phi_2[\eta_0])^2\\ c.c.

2818: \end{array}\right)\nn\\

2819: &+&\ \lambda P_c(T)\sigma_3

2820: \left(\begin{array}{c}

2821: |\Psi_1 +\pi_1\Phi_2[\eta_0]|^2 (\Psi_1 +\pi_1\Phi_2[\eta_0])

2822:  - |\Psi_1|^2\Psi_1\\ c.c.

2823: \end{array}\right)\nn\\

2824: &-&\ i P_c \left(\begin{array}{c}

2825:  \vec\nabla_1\Psi_1(t)\cdot\D_t\vec\alpha_1(t) +

2826:  \vec\nabla_0\Psi_0(t)\cdot\D_t\vec\alpha_0(t)\\

2827:  c.c. \end{array}\right)

2828: \label{eq:eta1_eqn}\ea

2829: The initial data for $\eta_1$ is $\eta_1(0)=0$.

2830: 

2831: \begin{rmk}

2832: A direct computation using the biorthogonal decomposition

2833: of the discrete subspaces of  

2834:  $\cH_0$ and $\cH_0^*$  of section \ref{sec:linearization}

2835: yields that the second term in (\ref{eq:eta_eqn}) is of a 

2836: higher order than is explicit:

2837: \be

2838: E_{10}(t)P_c(T)\sigma_3

2839:  \left(\begin{array}{c} \Psi_1\\ c.c.\end{array}\right)

2840:  = \cO\left( |\alpha_0|^2|\alpha_1| +|\alpha_1|^3 \right)\  

2841:  \left(\chi_0^{(0)} + \chi_0^{(1)} + \chi_0^{(0,1)} \right)

2842: \label{eq:direct}\ee

2843: for $|\alpha_0|<<1$ and $|\alpha_1|\downarrow 0$.

2844: \end{rmk}

2845: \noindent\underline{Equation for $\eta_2(t)$:}

2846: 

2847: \ba

2848: i\D_t\eta_2\ &=&\ {\cal H}_0(T)\eta_2\ +\

2849: P_c(T)\left[{\cal H}_0(t)-{\cal H}_0(T)\right]\Phi_2[\eta_1+\eta_2]\nn\\

2850: &+&\ \lambda P_c(T)\sigma_3

2851: \left(\begin{array}{c}

2852: 2\Psi_0\left[(\Psi_1+\pi_1\Phi_2[\eta_0])

2853: \overline{\pi_1\Phi_2[\eta_1+\eta_2]} + c.c.\right]\\ c.c.\end{array}\right)  

2854: \nn\\ &+&\ \lambda P_c(T)\sigma_3

2855: \left(\begin{array}{c}

2856: 2\Psi_0|\pi_1\Phi_2[\eta_1+\eta_2]|^2\ +\ 

2857:  2\ol{\Psi_0}

2858:  (\Psi_1+\pi_1\Phi_2[\eta_0]) \pi_1\Phi_2[\eta_1+\eta_2]

2859: \\ c.c. \end{array} \right)\nn\\

2860: &+&\  \lambda P_c(T)\sigma_3\left(\begin{array}{c}

2861: \ol{\Psi_0}(\pi_1\Phi_2[\eta_1+\eta_2])^2 \\ c.c.

2862: \end{array} \right)\nn\\

2863: &+&\ \lambda P_c(T)\sigma_3\left(\begin{array}{c}

2864: 2|\Psi_1+\pi_1\Phi_2[\eta_0]|^2\pi_2\Phi[\eta_1+\eta_2]\\

2865: c.c. \end{array} \right)\nn\\

2866: &+&\ \lambda P_c(T)\sigma_3\left(\begin{array}{c}

2867: (\pi_1\Phi_2[\eta_1+\eta_2])^2(\ol{\Psi_1}+

2868:  \overline{\pi_1\Phi_2[\eta_0]} )\ +\ 

2869: \overline{\pi_1\Phi_2[\eta_1+\eta_2]} 

2870:  (\Psi_1+\pi_1\Phi_2[\eta_0])^2\\ c.c.

2871: \end{array} \right)\nn\\

2872: &+&\ \lambda P_c(T)\sigma_3\left(\begin{array}{c}

2873: 2(\Psi_1+\pi_1\Phi_2[\eta_0]) |\pi_1\Phi_2[\eta_1+\eta_2]|^2\

2874:  +\ |\pi_1\Phi_2[\eta_1+\eta_2]|^2 \pi_1\Phi_2[\eta_1+\eta_2]\\ c.c.

2875: \end{array} \right)

2876: \nn\\ \label{eq:eta2_eqn}

2877: \ea

2878: 

2879: We expect that $\eta_2={\cal O}(\langle t\rangle^{-1})$. 

2880: Let

2881: \be \eta_2\ =\ \eta_{2a} + \eta_{2b}.\label{eq:eta2_def}\ee

2882: By construction we will show that 

2883:    $\eta_{2a}={\cal O}(\langle t\rangle^{-1})$

2884: and $\eta_{2b}={\cal O}(\langle t\rangle^{-{3\over2}})$.

2885: 

2886: \ba

2887: i\D_t\eta_{2a}\ &=&\ {\cal H}_0(T)\eta_{2a}\ +\ 

2888:  P_c(T)[{\cal H}_0(t)-{\cal H}_0(T)]\Phi_2[\eta_1]\nn\\

2889: &+&\ 2\lambda P_c(T)\sigma_3\left(\begin{array}{c}

2890: \Psi_0(\Psi_1\pi_1\ol{\Phi_2[\eta_1]} +

2891:  \ol{\Psi_1}\pi_2\Phi_2[\eta_1]) + \Psi_0|\pi_1\Phi_2[\eta_1]|^2\\ c.c.

2892: \end{array}\right)\nn\\

2893: &+&\ \lambda P_c(T)\sigma_3\left( \begin{array}{c} 

2894: 2\ol{\Psi_0}\pi_1\Phi_2[\eta_1](\Psi_1 + \pi_1\Phi_2[\eta_0])

2895:   + \ol{\Psi_0}(\pi_1\Phi_2[\eta_1])^2 \\ c.c.

2896: \end{array}\right)\nn\\

2897: &+& \lambda P_c(T)\sigma_3\left( \begin{array}{c}

2898: 2|\Psi_1+\pi_1\Phi_2[\eta_0]|^2\pi_1\Phi_2[\eta_1] 

2899: +(\Psi_1+\pi_1\Phi_2[\eta_0])^2\ol{\pi_1\Phi_2[\eta_1]}\\ c.c

2900: \end{array}\right)\nn\\

2901: &+& \lambda P_c(T)\sigma_3\left( \begin{array}{c}

2902: 2(\Psi_1+\pi_1\Phi_2[\eta_0])\ |\pi_1\Phi_2[\eta_1]|^2 

2903: +\ol{(\Psi_1+\pi_1\Phi_2[\eta_0])}\ (\pi_1\Phi_2[\eta_1])^2\\ c.c

2904: \end{array}\right)\nn\\

2905: &+& \lambda P_c(T)\sigma_3\left( \begin{array}{c}

2906: |\pi_1\Phi_2[\eta_1]|^2 \pi_1\Phi_2[\eta_1]\\ c.c.

2907: \end{array}\right)

2908:  \label{eq:eta2a_eqn}\ea

2909: \medskip

2910: 

2911: %The equation for $\eta_{2b}= \eta_2-\eta_{2a}$ is displayed in section

2912: % \bXYZ

2913: 

2914: \subsection{ Normal form and master equations}

2915: 

2916: Using (\ref{eq:direct})

2917:  and explicitly inserting in (\ref{eq:eta1_eqn}) the representation

2918: \be \alpha_1\ =\ \left( X(t)\beta(t)\right)_1

2919:  = e^{-i\lambda_+(T)t}\beta_1 + 

2920: \cO_0^{(0,1)}\alpha_0^2(T)E_{10}^{-1}(T) e^{-i\lambda_-(T)t}\ol{\beta_1}

2921: \label{eq:a1b1}\ee

2922: gives the following equation for $\eta_1$: 

2923: \ba

2924: i\D_t\eta_1\ &=&\ {\cal H}_0(T)\eta_1\ +\

2925:  P_c(T)\left( {\cal H}_0(t)-{\cal H}_0(T) \right)\Phi_2[\eta_0]\nn\\

2926: &+& E_{10}(t)P_c(T)\sigma_3 (\ |\ta0|^2\chi_0^{(0)} +\ |\tb1|^2\chi_1^{(1)}\ )

2927:  \nn\\

2928: && \times

2929:  \left(\begin{array}{c} e^{-i\lambda_+(T)t}\beta_1 +

2930:      \cO_0^{(0,1)}\alpha_0^2(T)E_{10}^{-1}(T)e^{-i\lambda_-t}\ol{\beta_1}

2931:          \\ c.c.\end{array}\right)\nn\\

2932: &+&  \lambda P_c(T)\sigma_3\psi_{0*}^2\psi_{1*}

2933: \left(\begin{array}{c}

2934: 2|\alpha_0|^2\beta_1 e^{-i\lambda_+(T)t} + \alpha_0^2\ol{\beta_1}

2935:  e^{i\lambda_+(T)t}\\ c.c. \end{array}\right)\nn\\

2936: &+& \lambda P_c(T)\sigma_3\psi_{0*}^2\psi_{1*}

2937: \left(\begin{array}{c}

2938: 2\alpha_0|\beta_1|^2 + \ol{\alpha_0}\beta_1^2e^{-2i\lambda_+(T)t}\\ c.c.

2939:  \end{array}\right) + \vec{\cal R}_{\eta_1}, 

2940: \label{eq:eta1_eqn2}\ea

2941: with initial condition $\eta_1(0)=0$.

2942: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2943: %\footnote{\bXYZ\ Are all terms quadratic in $\tb1$ and up to order $|\ta0|^2$ 

2944: % included?}

2945: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2946: \bigskip

2947: 

2948: Substitution of $\eta_1$ into the equations (\ref{eq:a0eqn})

2949:  and (\ref{eq:beta_eqn2}) for $\alpha_0$ and $\beta_1$ 

2950:  gives rise to terms which make explicit the resonant exchange of energy

2951: between the ground state and excited state. We next isolate the key terms 

2952:  in the expansion of $\eta_1$ relating to this energy exchange. 

2953: 

2954: Let's begin with the $\beta_1$ equation, (\ref{eq:beta_eqn2}). Written 

2955: out in greater detail we have:

2956: \ba

2957: i\D_t\beta_1 &=& 2\lambda\la\psi_{0*},\psi_{1*}^3\ra|\beta_1|^2\alpha_0         e^{i\lambda_+(T) t} + 

2958:  2\lambda\la\psi_{0*}\psi_{1*}^2,\pi_2\Phi_2\ra\beta_1\alpha_0\nn\\

2959: &+& \lambda\la\psi_{0*},\psi_{1*}^3\ra\beta_1^2\ol{\alpha_0}

2960:   e^{-i\lambda_+(T)t}

2961: + {\bf 2\lambda\la\psi_{0*}\psi_{1*}^2,\pi_1\Phi_2\ra} ({\bf \ol{\beta_1}\alpha_0 

2962:  e^{2i\lambda_+(T) t}} + \beta_1\ol{\alpha_0})\nn\\

2963: &+& 

2964: 2\lambda\la\psi_{0*}\psi_{1*},|\pi_1\Phi_2|^2\ra\alpha_0 e^{i\lambda_+(T)t}

2965: + 

2966:  \lambda\la\psi_{0*}\psi_{1*},(\pi_1\Phi_2)^2\ra\ol{\alpha_0}e^{i\lambda_+(T)t}

2967: \nn\\

2968: &+& \lambda\la\psi_{1*}^3,\pi_1\Phi_2\ra |\beta_1|^2e^{i\lambda_+(T)t} +

2969:  \lambda \la\psi_{1*}^2,(\pi_1\Phi_2)^2\ra\ol{\beta_1}e^{2i\lambda_+(T)t}\nn\\

2970: &+&\lambda\la\psi_{1*}^3,\pi_2\Phi_2\ra\beta_1^2 e^{-i\lambda_+(T)t}

2971:  + 2\lambda\la\psi_{1*}^2,|\pi_1\Phi_2|^2\ra\beta_1 e^{-i\lambda_+(T)t} \nn\\

2972: &+& \lambda\la\psi_{1*},|\pi_1\Phi_2|^2\pi_1\Phi_2\ra e^{i\lambda_+(T)t}

2973:  + {\cal R}_\beta,\label{eq:beta_eqn3}\ea

2974: where ${\cal R}_\beta$ is defined in (\ref{eq:Rbeta_def}).

2975: 

2976: We claim that the key term in (\ref{eq:beta_eqn3}) responsible for energy 

2977: transfer is the term  

2978: \be {\bf  2\lambda\la \psi_{0*}\psi_{1*}^2,\pi_1\Phi_2\ra\ol{\beta_1}\alpha_0 

2979:  e^{2i\lambda_+(T)t} }\label{eq:betarespo} \ee

2980: on the second line of (\ref{eq:beta_eqn3}).

2981: To see this we decompose $\eta_1$ into ``resonant'' and ``nonresonant'' parts

2982: \be \eta_1 = \eta_{1R} + \eta_{1NR}\nn\ee

2983: The part $\eta_{1NR}$ gives rise to the key resonant energy transfer 

2984:  term in the equation for $\beta_1$ and is the solution to the initial value 

2985:   problem

2986: \ba

2987: i\D_t\eta_{1R} &=& {\cal H}_0(T)\eta_{1R} + \lambda P_c(T)\sigma_3

2988:  \psi_{1*}^2\psi_{0*}\ol{\alpha_0}\beta_1^2 e^{-2i\lambda_+(T) t}

2989:  \left(\begin{array}{c}1\\ 0\end{array}\right)\nn\\

2990: \eta_{1R}(0) &=& 0.

2991: \label{eq:eta1R_eqn}\ea

2992: Solving (\ref{eq:eta1R_eqn}) using DuHamel's principle we have 

2993: \ba

2994: \eta_{1R}(t) &=& -i\lambda\int_0^t e^{-i{\cal H}_0(T)(t-s)}P_c(T)\sigma_3

2995:   \psi_{1*}^2\psi_{0*}\ol{\alpha_0(s)}\beta_1^2(s)e^{-2i\lambda_+(T)s}

2996:  \left(\begin{array}{c} 1 \\ 0\end{array}\right) ds

2997: \nn\\

2998: &=& -i\lambda e^{-i{\cal H}_0(T)t} \int_0^t e^{i[{\cal H}_0(T)-2\lambda_+(T)]s}

2999:  P_c(T)\sigma_3 \psi_{1*}^2\psi_{0*}\ol{\alpha_0(s)}\beta_1^2(s) 

3000: \left(\begin{array}{c} 1 \\ 0\end{array}\right) ds\nn\\

3001: \label{eq:eta1R_solved}\ea

3002: Recall that the continuous spectrum  of ${\cal H}_0(T)$ is given by points 

3003:  $\omega$  such that $|\omega|\ge|E_0|$. By Proposition \ref{prop:xdef}, 

3004: for $|\alpha_0|^2/E_{10}$ sufficiently small

3005: \be \lambda_+(T) = E_{10}(T) + \cO_0^{(0,1)}|\alpha_0(T)|^2.\nn\ee

3006: By the hypothesis (\ref{eq:omegastar}), $2E_{1*}-E_{0*}>0$, if $|\alpha_0|$ 

3007:  is sufficiently small then

3008:  $2\lambda_+(T)$ lies in the continuous spectrum of ${\cal H}_0(T)$. Therefore,

3009:  (\ref{eq:eta1R_eqn}) is a resonantly forced system. We expand the solution

3010: as follows. Let $\delta>0$ and set

3011: \be

3012: \eta_{1R}^\delta(t)\ =\ -i\lambda e^{-i{\cal H}_0(T)t} 

3013:  \int_0^t e^{i[{\cal H}_0(T)-2\lambda_+(T)-i\delta]s}

3014:  P_c(T)\sigma_3 \psi_{1*}^2\psi_{0*}\ol{\alpha_0(s)}

3015:  \beta_1^2(s) \left(\begin{array}{c} 1 \\ 0\end{array}\right) ds.

3016: \label{eq:eta1Rdelta_def}\ee

3017: Then, $\eta_{1R}=\lim_{\delta\to0}\eta_{1R}^\delta$.

3018: 

3019: We now apply Lemma \ref{lem:ibp} to 

3020: $\eta_{1R}$ with $A={\cal H}_0(T)-2\lambda_+(T)$.

3021: The result is

3022: \begin{prop}\label{prop:eta1R_expand}

3023:  The limit $\lim_{\delta\to0}\eta_{1R}^\delta=\eta_{1R}$ exists in ${\cal S}'$

3024:  and  

3025: \ba

3026: \eta_{1R}(t) &=& -\lambda e^{-i\lambda_+(T) t} \ol{\alpha_0(t)}\beta_1^2(t) 

3027:  \left({\cal H}_0(T)-2\lambda_+(T)-i0\right)^{-1}P_c(T)

3028:  \left(\begin{array}{c} 1 \\ 0\end{array}\right) \psi_{1*}^2\psi_{0*}\nn\\

3029: &+& \lambda\ol{\alpha_0(0)}\beta_1^2(0) e^{-i{\cal H}_0(T)t}

3030:   \left({\cal H}_0(T)-2\lambda_+(T)-i0\right)^{-1}P_c(T)

3031: \left(\begin{array}{c} 1 \\ 0\end{array}\right) \psi_{1*}^2\psi_{0*}\nn\\

3032: &+& \lambda e^{-i{\cal H}_0(T)t}

3033:  \left({\cal H}_0(T)-2\lambda_+(T)-i0\right)^{-1}\cdot\nn\\

3034: &&\ \ \  \int_0^t e^{i[{\cal H}_0(T)-2\lambda_+(T)]s}P_c(T)

3035:  \left(\begin{array}{c} 1 \\ 0\end{array}\right) \psi_{1*}^2\psi_{0*}

3036: {d\over ds}\left(\ol{\alpha_0(s)}\beta_1^2(s)\right) ds\nn\\

3037: &=& \eta_{1Ra} + \eta_{1Rb} + \eta_{1Rc}

3038: \label{eq:eta1R_expanded}

3039: \ea

3040: \end{prop}

3041: We substitute (\ref{eq:eta1R_expanded}) into the key term 

3042:  (\ref{eq:betarespo}) in (\ref{eq:beta_eqn3}).

3043: \be

3044: 2\lambda\la \psi_{0*}\psi_{1*}^2,\pi_1\Phi_2[\eta]\ra\ol{\beta_1}\alpha_0

3045:  e^{2i\lambda_+(T)t} = 

3046: 2\lambda\la \psi_{0*}\psi_{1*}^2,\pi_1\eta_{1Ra}\ra\ol{\beta_1}\alpha_0

3047:  e^{2i\lambda_+(T)t} + \bR.

3048: \label{eq:resp1}\ee

3049: Here, $\bR$ denotes rapidly dispersively decaying terms plus higher order

3050: terms in $|\alpha_0\beta_1|$. 

3051: By Proposition \ref{prop:eta1R_expand} and the Plemelj formula

3052: (\ref{eq:Plemelj})

3053: \ba

3054: && 2\lambda\la \psi_{0*}\psi_{1*}^2,\pi_1\eta_{1Ra}\ra\ol{\beta_1}\alpha_0

3055:  e^{2i\lambda_+(T)t}\nn\\

3056: &=& -2\lambda^2\left\la \psi_{0*}\psi_{1*}^2,

3057:  \pi_1({\cal H}_0(T)-2\lambda_+(T)-i0)^{-1}P_c(T)

3058:  \left(\begin{array}{c}1 \\ 0\end{array}\right)\psi_{0*}\psi_{1*}^2\right\ra

3059:  |\alpha_0|^2|\beta_1|^2\beta_1\nn\\

3060: &=& -2(\Lambda+i\Gamma) |\alpha_0|^2|\beta_1|^2\beta_1,

3061: \label{eq:resp2}\ea

3062: where (using that $E_{0}-2E_{10}=E_0-2(E_1-E_0)= 

3063:  \omega_*+{\cal O}(|\alpha_0(T)|^2)$)

3064: we have 

3065: \ba

3066: \Lambda &=& \Lambda_{\omega_*}\ \cO_0^{(0)} \nn\\

3067: &=& \lambda^2\la \psi_{0*}\psi_{1*}^2,

3068:  {\rm P.V.} ( H-\omega_*)^{-1} 

3069:  \psi_{0*}\psi_{1*}^2\ra\ \cO_0^{(0)}\label{eq:Lambda_def}\\

3070: \Gamma &=& \Gamma_{\omega_*}\ \cO_0^{(0)}\nn\\ 

3071: &=& \lambda^2\pi \la \psi_{0*}\psi_{1*}^2,

3072:   \delta(H-\omega_*) 

3073:  \psi_{0*}\psi_{1*}^2\ra\ \cO_0^{(0)} 

3074: \label{eq:Gamma_def}

3075: \ea

3076: %\footnote{\bXYZ\ correction terms are $\cO(|\alpha_0|^2)$ by local decay.}

3077: Recall that $\cO_0^{(0)}$ denotes a term of the form $1+\cO(|\alpha_0|^2)$.

3078: Returning to the equation for $\beta_1$ we have:

3079: 

3080: \ba i\D_t \beta_1\ &=&\ (\Lambda_{\omega_*} -i\Gamma_{\omega_*})

3081:  |\alpha_0|^2|\beta_1|^2\beta_1+....

3082: \label{eq:beta_eqn4}

3083: \ea

3084: 

3085: We now seek the key terms in the $\alpha_0$ equation, (\ref{eq:a0eqn}).

3086: Using that $\phi_1=\Psi_1+\phi_2$ 

3087:  and the representation (\ref{eq:a1b1}),

3088:  we have 

3089: %{\bf XYZ -- RECHECK}:  

3090: \ba

3091: 2i\D_t \alpha_0\ &=& 

3092:   \lambda\la\psi_{0*}^2,\psi_{1*}^2\ra e^{-2i\lambda_+(T)t}

3093:      \beta_1^2\ol{\alpha_0}

3094: \nn\\

3095: &+& 2\lambda\la\psi_{0*}^2\psi_{1*},\phi_2\ra e^{-i\lambda_+(T)t}

3096:  \ol{\alpha_0}\beta_1 + 

3097:  \lambda\la\psi_{0*}^2,\phi_2^2\ra \ol{\alpha_0}

3098: \nn\\

3099: &+& 

3100: %&+& \lambda\la\psi_{0*},\psi_{1*}^3\ra|\beta_1|^2\beta_1 e^{-i\lambda_+(T)t}

3101:  + 2\lambda\la\psi_{0*}\psi_{1*}^2, \phi_2\ra|\beta_1|^2\nn\\

3102: &+& \lambda\la\psi_{0*}\psi_{1*}, \phi_2^2\ra\ol{\beta_1}e^{i\lambda_+(T)t}

3103:  + {\bf 

3104:  \lambda\la\psi_{0*}\psi_{1*}^2, \ol{\phi_2}\ra\beta_1^2e^{-2i\lambda_+(T)t}

3105:     }

3106: \nn\\

3107: &+& 2\lambda\la\psi_{0*}\psi_{1*},|\phi_2|^2\ra\beta_1e^{-i\lambda_+(T)t}

3108:  + \lambda\la\psi_{0*},|\phi_2|^2\phi_2\ra\nn\\

3109: &+& {\cal R}_{\alpha_0}

3110: \label{eq:alpha0_eqn2}

3111: \ea

3112: 

3113: We first focus on the key resonant term in (\ref{eq:alpha0_eqn2}) which is 

3114:  responsible for the system settling onto the nonlinear ground state.

3115: We claim this term is:

3116: \be

3117: {\bf \lambda\la\psi_{0*}\psi_{1*}^2, \ol{\phi_2}\ra\beta_1^2

3118:  e^{-2i\lambda_+(T)t} }

3119: \label{eq:alpharespo}\ee

3120: In analogy with the previous calculation, we have

3121: \be

3122: \lambda\la\psi_{0*}\psi_{1*}^2, \ol{\phi_2}\ra\beta_1^2e^{-2i\lambda_+(T)t}

3123:  = \lambda\la\psi_{0*}\psi_{1*}^2, \pi_2\eta_{1Ra}\ra\beta_1^2e^{-2i\lambda_+(T)t}

3124:  + \bR,\label{eq:resp3}

3125: \ee

3126: where $\bR$ is as above.

3127: Therefore, applying Proposition \ref{prop:eta1R_expand} we get

3128: \be

3129: \lambda\la\psi_{0*}\psi_{1*}^2, \pi_2\Phi_2\ra\beta_1^2e^{-2i\lambda_+(T)t}

3130:  = (-\Lambda_{\omega_*} +i\Gamma_{\omega_*})|\beta_1|^4\alpha_0

3131:  + \bR 

3132: \label{eq:keyterm}

3133: \ee

3134: The expression $\bR$ denotes terms which are higher order, oscillatory type

3135: and dispersively decaying. See also Remark \ref{rmk:approach}. 

3136: 

3137: In summary, we have the following system for $\alpha_0$ and $\beta_1$:

3138: \begin{prop}\label{eq:prenormalform}

3139: \ba

3140: i\D_t \alpha_0 &=& (-\Lambda_{\omega_*} +i\Gamma_{\omega_*})|\beta_1|^4\alpha_0

3141:  + \bR \nn\\

3142: i\D_t \beta_1 &=&  2(\Lambda_{\omega_*} -i\Gamma_{\omega_*})

3143:  |\alpha_0|^2|\beta_1|^2\beta_1 + \bR

3144: \label{eq:pre_nf}\ea

3145: \end{prop}

3146: 

3147: The proof of Proposition \ref{prop:normalform} 

3148:  follows by constructing an appropriate near-identity change of 

3149: variables, transforming  (\ref{eq:pre_nf}) to 

3150:  (\ref{eq:normalform}). This is implemented as in \cite{SW99}. 

3151: \bigskip

3152: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3153: %\footnote{

3154: %The correction terms $R_0$ and $R_1$ satisfy the estimates:

3155: %\ba

3156: %&& |R_0| \le |\ta0|\ |F_\alpha|\ =\  

3157: % C_\alpha\left(1+\cO(|\alpha_0|)\right)\nn\\

3158: %&& \times \left(\ |\ta0|^2|\tb1|^{2m}  +

3159: % |\ta0|^3 |\tb1|^4\  +\

3160: %  |\alpha_0|\ |\la\chi,|\eta_0| + |\eta_{2b}|\ra|\ |\tb1|^2

3161: % +|\alpha_0|\ |\la\chi,|\eta_0|^3 +|\eta_{2b}|^3\ra|

3162: %\right)

3163: %\label{eq:R0_est}\\

3164: %&&|R_1| \le |\tb1|\ |F_\beta|\ =\ 

3165: %  C_\beta \left(1+\cO(|\alpha_0|)\right)\nn\\

3166: %&& \times \left(\ |\ta0||\tb1|^{2m+1} +

3167: %  |\ta0|^2|\tb1|^5\ +\

3168: % |\tb1|\ |\la\chi,\cH_0(T)\eta_0(t)\ra| 

3169: % + \la\chi, |\eta_{2b}|\ra|\tb1|^3\ 

3170: % +\ |\tb1|\ |\la\chi,|\eta_0|^3 +|\eta_{2b}|^3\ra|\

3171: %\right)\nn\\

3172: %\label{eq:R1_est}

3173: %\ea

3174: %}

3175: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3176: \section{Stability analysis on different time scales\ -\ 

3177:  overview}\label{sec:stab}

3178: 

3179: In Corollary \ref{cor:nlPME}  we obtained  coupled power equations

3180:  or nonlinear master equations governing the (renormalized) ground state and

3181: excited state square amplitudes: 

3182: \be

3183: P_0=|\talpha_0|^2 \ \ {\rm and}\ \  

3184: P_1=|\tbeta_1|^2

3185: \nn\ee

3186: If we neglect the correction terms $R_0$ and $R_1$, 

3187:  in (\ref{eq:P0eqn}-\ref{eq:P1eqn}) we obtain the simpler autonomous system

3188:  of differential equations: 

3189: \ba

3190: {dp_0\over dt} &=& 2\Gamma p_0p_1^2\label{eq:modelP0_eqn}\\

3191: {dp_1\over dt} &=& -4\Gamma p_0p_1^2\label{eq:modelP1_eqn}

3192: \ea

3193: Note that this  system is exactly solvable.  

3194: Addition of twice (\ref{eq:modelP0_eqn}) to (\ref{eq:modelP1_eqn}) yields that 

3195:  along any solution trajectory:

3196: \be 2p_0(t)+p_1(t)\ =\ 2p_0(0) + p_1(0).\label{eq:conserv}\ee 

3197: This relation can be used to eliminate $p_1$ from (\ref{eq:modelP0_eqn}) or 

3198:  $p_0$ from (\ref{eq:modelP1_eqn}). $p_0(t)$ and $p_1(t)$ are thus obtained by quadrature.

3199: %The phase portrait for the system system 

3200: % (\ref{eq:modelP0_eqn}-\ref{eq:modelP1_eqn}) is 

3201: %displayed in figure {\bf XYZ}. 

3202:  The dynamics of this 

3203:   finite dimensional reduced system

3204:  anticipates that an initial 

3205: state, arbitrarily close to but not exactly on the excited state branch, 

3206:  with energy distributed among the ground state and excited state, will evolve

3207: to a state with an increased ground state energy and no energy in the 

3208: excited state. While not strictly correct, since there are nongeneric 

3209:  data giving rise to solutions which converge to the excited 

3210:   state \cite{TY01b},  

3211:   this captures the generic very large time dynamics. 

3212:  The correction terms $R_0$ and $R_1$ 

3213:  in (\ref{eq:P0eqn}-\ref{eq:P1eqn}) lead to different

3214: transient behavior which may be quite different from that suggested

3215:  by the system (\ref{eq:modelP0_eqn}-\ref{eq:modelP1_eqn}).

3216: However, we show that eventually ($t\ge t_1$), this system dominates.

3217:  Moreover,  a  

3218: large class of data, for which  the system

3219: (\ref{eq:modelP0_eqn}-\ref{eq:modelP1_eqn}) controls the behavior is 

3220: that for which $P_0(0)>P_1(0)$ and sufficiently small initial 

3221:  dispersive part.

3222: 

3223: Before embarking on the details we give a brief overview of the strategy.

3224: Using the change of variables $(\alpha,\beta)\mapsto(\talpha,\tbeta)$

3225:  of Proposition \ref{prop:normalform}

3226: we have transformed away all {\it local in time} nonresonant terms. This

3227:  introduces contributions to $F_\alpha$ and $F_\beta$, and therefore

3228: contributions to $R_0$ and $R_1$ in equations (\ref{eq:P0eqn}-\ref{eq:P1eqn}),

3229: which are of two types: 

3230: 

3231: \nit (i) {\it local in time} terms depending on $\talpha_0$ and $\tbeta_1$,

3232:  which can be absorbed by the

3233:  leading terms in (\ref{eq:P0eqn}-\ref{eq:P1eqn}), with a small correction

3234: to the coefficient to the coefficient $\Gamma$ and are of order $b_0\Jt^{-2}$;

3235:  see Proposition \ref{prop:propA} below.

3236: 

3237: \nit {\it nonlocal in time} functions of $\talpha_0$

3238:  and $\tbeta_1$

3239: defined in terms of $\eta=\eta[\eta_0,\eta_1,\eta_2]$ in $F_\alpha$ and

3240:  $F_\beta$. These contribute

3241: terms to (\ref{eq:normalform}) with the same (anticipated) time-decay

3242: rate as the leading order terms in (\ref{eq:normalform}). Correspondingly,

3243: there are {\it nonlocal in time} functions of $\talpha_0$ and $\tbeta_1$

3244:  which contribute to $R_j$ in equations (\ref{eq:P0eqn}-\ref{eq:P1eqn}) which

3245: are of the same (anticipated) decay rate as the leading order terms

3246:  in (\ref{eq:P0eqn}-\ref{eq:P1eqn}).

3247:  The goal is to control these nonlocal terms,

3248:  to the extent possible,

3249:   by the leading order terms. However, due to the

3250:  different behavior

3251: of $\talpha_0$ and $\tbeta_1$ on different time scales the argument is somewhat

3252: tricky and we now explain our strategy.

3253: 

3254:  Let

3255: \be

3256: t_0 + 1\ \equiv

3257:  \sup_{\tau\ge0}\left\{

3258:  \tau:\ 0\le\tau'\le\tau,\ P_0(\tau')\le\frac{5}{2}

3259:  \frac{ [\eta_0]_X+\cE_0 }{ {\la \tau-1\ra} {\la\tau'\ra^2} }

3260:           \right\}\label{eq:t0def}

3261: \ee

3262: Propositions \ref{prop:biR0i}-\ref{prop:ODE} and \ref{prop:Riteta}

3263:  will justify this choice, by implying the inequalities (\ref{eq:below}).  

3264: If $t_0<\infty$, then we have  the bound:

3265: \be

3266: P_0(t) \le \frac{5}{2}

3267:  \frac{[\eta_0]_X+\cE_0}{ {\la t_0\ra} {\la t\ra^2} },\ 0\le t\le t_0.

3268: \nn\ee

3269: 

3270: Consider the system for $P_0(t)$ and $P_1(t)$,

3271:   (\ref{eq:P0eqn}-{\ref{eq:P1eqn}).

3272: Decomposing $R_j$ into local and nonlocal in time  parts we have:

3273: \ba R_j &=&

3274:  R_j^{l}(P_0,P_1,\eta_0,\eta) + \int_0^tK(t,s)r_j^{nl}(s) ds

3275: \nn\\

3276:  &=& R_j^{l}(P_0,P_1,\eta_0,\eta) + \int_0^{t_0}K(t,s)r_j^{nl}(s) ds

3277: \nn\\

3278: \ \ &+&\ \int_{t_0}^t K(t,s)r_j^{nl}(s) ds

3279: \nn\ea

3280: We have that

3281: \ba

3282: R_0^{l}(P_0,P_1,\eta_0,\eta) &\le& \cO(\cE_0^{\rho_0})\left(\ P_0P_1^2 +

3283:                                \frac{b_0}{\Jt^2}\right)

3284:                                 \nn\\

3285: R_1^{l}(P_0,P_1,\eta_0,\eta) &\le& \cO(\cE_0^{\rho_1})\left(\ P_0P_1^2 +

3286:                                \frac{b_1}{\Jt^2} +

3287:                                 \cO\left(\sqrt{P_0}P_1^m\right)\ \right),

3288: \label{eq:below}\ea

3289: where $\cE_0$ is the initial total energy and $\rho_j>0$. 

3290:  Therefore, for $\cE_0$ small

3291: \ba

3292: \frac{dP_0}{dt} &\ge& 2\Gamma' P_0P_1^2 - \frac{b_0}{\Jt^2}

3293:          + \int_{t_0}^t K_0(t,s)r_0^{nl}(s) ds\nn\\

3294: \frac{dP_1}{dt} &\le& -4\Gamma' P_0P_1^2 + \frac{b_1}{\Jt^2} +

3295:            \int_{t_0}^t K_1(t,s)r_1^{nl}(s) ds

3296:  + \cO\left(\sqrt{P_0}P_1^m\right),

3297: \nn\ea

3298: with $\Gamma'\sim\Gamma$. The reverse inequalities hold with $\Gamma'$

3299: replaced by $\Gamma''=\Gamma + o(\Gamma)$.

3300: 

3301: Next (Proposition \ref{prop:propB}) we introduce the auxiliary quantities

3302: \be Q_0 = P_0-k_0\Jt^{-1}\ {\rm and}\ Q_1 = P_1 + k_1\Jt^{-1},\nn\ee

3303: where $k_0(b_0,b_1,\cE_0)$ and $k_1(b_0,b_1,\cE_0)$ are chosen appropriately

3304: and derive equations of the form

3305: \ba

3306: \frac{dQ_0}{dt} &\ge& 2\Gamma'' Q_0Q_1^2

3307:          + \int_{t_0}^t K_0(t,s)r_0^{nl}(s) ds\nn\\

3308: \frac{dQ_1}{dt} &\le& -4\Gamma'' Q_0Q_1^2  +

3309:            \int_{t_0}^t K_1(t,s)r_1^{nl}(s) ds

3310:  + \cO\left(\sqrt{Q_0}Q_1^m\right),

3311: \label{eq:Q0Q1dp}\ea

3312: where $\Gamma''\sim\Gamma'\sim\Gamma$.

3313: 

3314: We then proceed with the following continuity argument. At $t=t_0$,

3315: \be

3316: \frac{dQ_0(t_0)}{dt} \ge 2\Gamma'' Q_0(t_0)Q_1^2(t_0),\ \

3317: \frac{dQ_1(t_0)}{dt} \le -4\Gamma'' Q_0(t_0)Q_1^2(t_0)  +

3318:   \cO\left(\sqrt{Q_0(t_0)}Q_1^m(t_0)\right).

3319: \label{eq:t0est}\ee

3320: Therefore, by continuity, the following inequalities hold for some time interval

3321:  $t_0\le t\le t_{0,1}$, with $\Gamma''$ replaced by ${\Gamma''\over2}$:

3322: \be

3323: \frac{dQ_0}{dt} \ge \Gamma'' Q_0Q_1^2,\ \

3324: \frac{dQ_1}{dt} \le -2\Gamma'' Q_0Q_1^2  +

3325:   \cO\left(\sqrt{Q_0}Q_1^m\right).

3326: \label{eq:t0t01}\ee

3327: Let

3328: $t^* \ \equiv\

3329:  \sup\{t\ge t_0:\ $ inequalities (\ref{eq:t0t01}) hold $\}$.

3330: We show that

3331: \be

3332:  \int_{t_0}^t K_1(t,s)r_1^{nl}(s) ds \le

3333: \cO(\cE_0^\rho)\left(\ Q_0Q_1^2 + \frac{b_j}{\Jt^2}\ \right)

3334: \label{eq:nlbound}\ee

3335: and therefore, up to renormalization of $Q_j$ (adding higher order terms

3336:  of order $\cE_0^\rho k_j\la t\ra^{-1}$ to the definition of $Q_j$)).

3337:  Use of this estimate

3338: in (\ref{eq:Q0Q1dp}) implies (\ref{eq:t0t01}), for $\cE_0$ sufficiently small.

3339:  The argument can be repeated   

3340: and therefore, $t^*=T$.

3341: 

3342: 

3343: \section{ Finite dimensional reduction and its analysis 

3344:  on different time scales}\label{sec:finite-dim-red}

3345: 

3346: We now begin our study of the generic case, where $t_0<\infty$ and

3347: the solution converges to the nonlinear ground state family as $t\to\infty$. 

3348: The following three propositions concern the various time-scales which

3349: enter the analysis. The first is a basic result, a normal form, which  

3350: is the point of departure for our analysis 

3351:  on all time scales.

3352:  

3353: \begin{prop}\label{prop:propA}

3354: \nit Let $m\ge4$. Let

3355: \ba

3356: b_0\ &=&\ \la t_0\ra^{-1}\left(\ [\eta_0]_X\ +\ c_*\cE_0^2\ \right),

3357: \label{eq:b0}\\

3358: \ b_1\ &=&\ \la t_0\ra^{-{1\over2}}

3359:  \left(\ [\eta_0]_X^{2\over3}\ +\ d_*\cE_0^2\ \right), \label{eq:b1}

3360: \ea

3361: for some order one constants $c_*$ and $d_*$.

3362: %\bXYZ

3363: 

3364:  If for some $t_0$, positive and finite,

3365: \be P_0(t_0)\ge \frac{ 3b_0\left(t_0,[\eta_0]_X\right)}{\la t_0\ra},

3366: \label{eq:thresh}\ee

3367:  then

3368: for $t\ge t_0$

3369: \ba

3370: {dP_0\over dt} &\ge& 2(1-\delta_1)\Gamma P_0P_1^2\ -\ {b_0\over \Jt^2}\ +\ J_0

3371: \label{eq:P0P1aa}\\

3372: {dP_1\over dt} &\le& -4(1-\delta_1)\Gamma P_0P_1^2\ +\ \cO(\sqrt{P_0}P_1^m)

3373: \ +\ {b_1\over \Jt^2}\ +\ J_1,

3374: \label{eq:P0P1ab}

3375: \ea

3376: where $J_0$ and $J_1$ are nonlocal in time terms, which have the form:

3377: \be

3378: J_j \ =\ \int_{t_0}^t\ K(t,s)\ r_j^{nl}(s)\ ds. \nn 

3379: \ee

3380: The terms encompassed in $J_j$ are derived and estimated in the coming sections. 

3381: %In this proposition, we simply note their presence.

3382: \end{prop}

3383: %

3384: \begin{rmk}

3385: \medskip

3386: 

3387: \nit The reverse inequalities of (\ref{eq:P0P1aa}- \ref{eq:P0P1ab}) 

3388: hold as well with a different constant $\delta_2\sim\delta_1$.

3389: \end{rmk}

3390: 

3391:  The proof of Proposition \ref{prop:propA} will be given

3392:  following the estimates on the remainder terms, $R_i$ (Proposition

3393:  \ref{prop:Master}).

3394: \bigskip

3395: 

3396: \begin{prop}\label{prop:propB}

3397: Assume that $t_0$ is positive and finite as in Proposition \ref{prop:propA}.

3398:  Then,  there exist $k_0=k_0(b_0,b_1,\cE_0)$ 

3399:   and $k_1=k_1(b_0,b_1,\cE_0)$, such that 

3400:  for $t\ge t_0$ the auxiliary functions

3401: \be Q_0(t)\ \equiv \ P_0(t) - { k_0\over \Jt},\ \ \  

3402:  Q_1(t)\ \equiv \ P_1(t) + { k_1\over \Jt}

3403: \label{eq:Q0Q1def}

3404: \ee

3405: satisfy

3406: \ba

3407: {dQ_0\over dt} &\ge& 2\Gamma' Q_0Q_1^2\ +\ J_0\ +\

3408:            \frac{\cE_0^2c_*}{\la t_0\ra\ \la t\ra^2}  \nn\\

3409: {dQ_1\over dt} &\le& -4\Gamma' Q_0Q_1^2  + \cO(\sqrt{Q_0}Q_1^m)\ 

3410:  +\ J_1\ - \ \frac{\cE_0^2d_*}{\la t_0\ra^{1\over2}\ \la t\ra^2},

3411: \label{eq:Q0Q1b}

3412: \ea

3413: for some positive free constants $c_*$ and $d_*$.

3414: In particular, $Q_0(t)$ is monotonically increasing for $t\ge t_0$.

3415: \end{prop}

3416: The next result shows that $c_*$ and $d_*$ can be chosen to control the terms 

3417:  $J_j$.

3418: 

3419: \begin{prop}\label{prop:propB.5}(Monotonicity of $Q_0$ for $\t\ge t_0$)

3420: There exist $c_*$ and $d_*$ of order one, such that for $t\ge t_0$.

3421: \ba

3422: {dQ_0\over dt} &\ge& 2\Gamma' Q_0Q_1^2\nn\\

3423: {dQ_1\over dt} &\le& -4\Gamma' Q_0Q_1^2  + \cO(\sqrt{Q_0}Q_1^m)\

3424: \label{eq:Q0Q1b.5}

3425: \ea

3426: \end{prop}

3427: The above three propositions are established in the next two sections. We complete 

3428: the current section by working out the consequences of the 

3429:  finite dimensional reduction (\ref{eq:Q0Q1b.5}).

3430: \bigskip

3431: 

3432: 

3433: The next proposition, shows that even if $Q_0$ very small at some stage,

3434:  it will eventually become large relative to $Q_1$ and the 

3435: $\cO(\sqrt{Q_0}Q_1^m)$ term in (\ref{eq:Q0Q1b}) will become negligible.

3436: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3437: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3438: \begin{prop}\label{prop:propC}

3439: Assume

3440: $t\ge t_0$ and suppose for some $r>0$ that 

3441: \be

3442: \frac{Q_0}{Q_1} \leq \cE_0^r.\label{eq:smallQ0}

3443: \ee

3444: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3445: %\footnote{\bXYZ $\cE_0$ here is the size of the energy. We also

3446: %use $Q_0,Q_1\le \cO(\cE_0)$, which should be spelled out}

3447: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3448: Then,

3449: \be \frac{Q_0(t)}{Q_1(t)}\ {\rm is\ increasing\ for\ } t\ge t_0 \ee

3450: and there exists $t_1$ such that

3451: \be \frac{Q_0(t_1)}{Q_1(t_1)}=\cE_0^r.\nn\ee

3452: Furthermore, for $t\ge t_1$:

3453: \ba

3454: &&\frac{d Q_0}{dt}\geq 2\Gamma' Q_0 Q_1^2\nn\\

3455: &&\frac{d Q_1}{dt} \leq - 4\Gamma' Q_0 Q_1^2

3456: \label{eq:Q0Q1c}\ea

3457: \nit Finally, for $t\ge t_1$:

3458: \be

3459: Q_1(t) \le {Q_1(t_1)\over 1+4\Gamma'' Q_1(t_1)(\inf_{[t_1,T]}|Q_0|)\cdot (t-t_1)}

3460: \label{eq:Qbound}

3461: \ee

3462: \end{prop}

3463: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3464: %\footnote{

3465: %\begin{rmk}

3466: %If the hypothesis of Proposition \ref{prop:odeprop}  is false, then for

3467: %all $t$ satisfying $0\le t\le T$, we have $P_0(t)\le 3b_0/\la t\ra$. 

3468: % We shall analyze this case 

3469: %below. 

3470: %\end{rmk}

3471: %}

3472: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3473: \nit{\bf Proof of Proposition  \ref{prop:propB}:}

3474: Define

3475: \be Q_0 \equiv P_0-k_0\la t\ra^{-1},\ \ 

3476:  Q_1 \equiv P_1+k_1\la t\ra^{-1},\label{eq:Q01_def}\ee

3477: where $k_0,k_1>0$ are to be appropriately chosen.  Note that 

3478: \be Q_0(t)\le P_0(t)\ {\rm and}\ Q_1(t)\ge P_1(t).\nn\ee 

3479:  Using (\ref{eq:P0P1aa}-\ref{eq:P0P1ab})

3480:  and some estimation we deduce a simplified system for $Q_0$ and $Q_1$.

3481: We calculate, omitting the terms $J_0$ and $J_1$, which

3482: are carried along passively.

3483:  

3484: We begin with $Q_0$.

3485: \ba

3486: {dQ_0\over dt} &=& {dP_0\over dt} + {k_0\over\la t\ra^2}\nn\\

3487:  &\ge& 2\Gamma P_0P_1^2 - \frac{b_0}{\la t\ra^2} + {k_0\over\la t\ra^2}\nn\\

3488:  &=& 2\Gamma \left(Q_0+{k_0\over\la t\ra}\right)

3489:         \left(Q_1-{k_1\over\la t\ra}\right)^2

3490:        -\frac{b_0}{\la t\ra^2} + {k_0\over\la t\ra^2}\nn\\

3491:  &\ge& 2\Gamma Q_0 Q_1^2 -4\Gamma {k_1\over\la t\ra} Q_0 Q_1

3492:        + 2\Gamma {k_1^2\over \la t\ra^2}\nn\\

3493:    &+& {k_0-b_0\over\la t\ra^2}.

3494: \label{eq:Q0a}

3495: \ea

3496: We estimate the second term on the right hand 

3497: side as follows. For any $s>0$,

3498: \be

3499: -4\Gamma {k_1\over\la t\ra} Q_0 Q_1 \ge -2s\Gamma Q_0Q_1^2

3500:  -2\Gamma Q_0 {k_1^2\over s\la t\ra^2}.\nn\ee

3501: Therefore,

3502: \be

3503: {dQ_0\over dt} \ge 2\Gamma(1-s)Q_0Q_1^2 + \left(

3504:   2\Gamma k_1^2 Q_0\left(1-{1\over s}\right)

3505: + k_0-b_0\right){1\over\la t\ra^2}

3506: \label{eq:Q0b}

3507: \ee

3508: Now set $s={1\over10}$ and assume 

3509: \be k_1\ge b_1.\label{eq:k1con}\ee

3510: Then,  using that $k_1=10b_1$, we have

3511: \be

3512: {dQ_0\over dt} \ge

3513: 2\cdot{9\over10}\Gamma Q_0Q_1^2 + \left(k_0 - b_0 

3514:  -18\Gamma b_1^2 Q_0\right){1\over\la

3515:  t\ra^2}

3516: \label{eq:Q0c}

3517: \ee

3518: If

3519: \be k_0 \equiv  b_0 +18\Gamma b_1^2 \sup_{t_0\le t\le T} Q_0\ +\ 

3520:  \frac{\cE_0^2c_*}{\la t_0\ra},

3521: \label{eq:k0_def}

3522: \ee

3523: we have  by (\ref{eq:Q0c}) and  (\ref{eq:Q1c})

3524: \be

3525: {dQ_0\over dt} \ge 2\cdot{9\over10}\Gamma Q_0Q_1^2\ +\ 

3526:  \frac{\cE_0^2c_*}{\la t_0\ra\ \la t\ra^2}

3527: \label{eq:Q0d}\ee

3528: In particular, $Q_0$ is increasing for $t\ge t_0$.

3529: \medskip

3530: 

3531: We now turn to $Q_1$, we have

3532: \ba

3533: {dQ_1\over dt} &=& {dP_1\over dt} - {k_1\over\la t\ra^2}\nn\\

3534:                &\le& -4\Gamma P_0P_1^2 + 

3535:  {b_1\over\la t\ra^2}- {k_1\over\la t\ra^2} 

3536:  + \cO(\sqrt{P_0}P_1^m)\nn\\

3537:                &=& -4\Gamma \left(Q_0+{k_0\over\la t\ra}\right)

3538:                 \left(Q_1-{k_1\over\la t\ra}\right)^2

3539:   + {b_1-k_1\over \la t\ra^2} + \cO(\sqrt{P_0}P_1^m)\nn\\

3540:                &\le & -4\Gamma Q_0\left(Q_1-{k_1\over\la t\ra}\right)^2 + 

3541:  {b_1-k_1\over \la t\ra^2} + \cO(\sqrt{P_0}P_1^m)\nn\\

3542:                &=& -4\Gamma Q_0Q_1^2 + 8\Gamma {k_1\over\la t\ra}Q_0Q_1

3543:                    -4\Gamma Q_0{k_1^2\over\la t\ra^2} + 

3544:   {b_1-k_1\over \la t\ra^2} + \cO(\sqrt{P_0}P_1^m)\nn\\

3545: && \label{eq:Q1a} 

3546: \ea

3547: The second term on the right hand side is estimated as follows. For any $r>0$

3548: we have, since $2ab\le ra^2 + r^{-1}b^2$, that

3549: \be

3550: 8\Gamma {k_1\over\la t\ra}Q_0Q_1 \le {4\Gamma k_1 r Q_0\over\la t\ra^2}

3551:  +{4\Gamma k_1 Q_0 Q_1^2\over r}.\nn\ee

3552: Therefore,

3553: \be

3554: {dQ_1\over dt} \le -4\Gamma(1-{k_1\over r}) Q_0Q_1^2 + 

3555:               4\Gamma k_1 Q_0 {r-k_1\over\la t\ra^2}

3556:                 + {b_1-k_1\over\la t\ra^2}

3557:  + \cO(\sqrt{P_0}P_1^m)

3558: \label{eq:Q1b}

3559: \ee

3560: Let

3561: \be  r\equiv 20k_1\ {\rm and}\ b_1\ \equiv\ k_1(1-76\Gamma k_1Q_0)\ -\ 

3562:  \frac{\cE_0^2d_*}{\la t_0\ra^{1\over2}},

3563:  \label{eq:k1b1}\ee

3564: which is consistent with the constraint (\ref{eq:k1con}).

3565: Then, 

3566: \ba

3567: {dQ_1\over dt} &\le& -4\cdot{19\over20}\Gamma Q_0Q_1^2 

3568:               \  -\  \frac{\cE_0^2d_*}{\la t_0\ra^{1\over2}\ \la t\ra^2} + 

3569:                    \cO(\sqrt{P_0}P_1^m).

3570: \label{eq:Q1c}

3571: \ea

3572: By (\ref{eq:Q0Q1b}), $Q_0$ is increasing

3573: %\footnote{\bXYZ\ We now have the $J_j$ terms.}

3574:  for $t\ge t_0$. Therefore, 

3575: \ba Q_0(t_0)&\ge& {1\over 100}\frac{k_0}{\la t_0\ra}\ \Longrightarrow\nn\\ 

3576:   P_0(t)&\le& Q_0(t) + \frac{k_0}{\la t\ra}

3577:  =  Q_0(t) + 100 \frac{k_0}{100\la t\ra}\nn\\

3578:  &\le& Q_0(t) + 100 Q_0(t_0)\le 101 Q_0(t) 

3579: \nn\ea

3580: Also, by definition $P_1\le Q_1$ and so

3581: \ba

3582: {dQ_1\over dt} &\le& -4\cdot{19\over20}\Gamma Q_0Q_1^2 +\cO(\sqrt{Q_0}Q_1^m) 

3583: \label{eq:Q1d}

3584: \ea

3585: This completes the proof of Proposition \ref{prop:propB}.

3586: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3587: 

3588: \nit{\bf Proof of Proposition \ref{prop:propC}: }

3589: \ba

3590: &&\frac{d}{dt}\left(\frac{Q_0}{Q_1}\right) = \frac{\dot Q_0 Q_1 - \dot

3591: Q_1 Q_0}{Q_1^2} \geq \frac{1}{Q^2_1} \left\{ \Gamma  Q_0Q_1^2Q_1 +

3592: 4\Gamma Q_0^2 Q_1^2 - C Q_0^{3/2} Q_1^m \right\}\nn\\

3593: &&

3594: \geq \frac{1}{Q^2_1} \left\{ 2\Gamma Q_0 \left[Q^3_1 - C Q_0^{1/2}

3595: Q_1^{-1/2} Q_1^{m + 1/2} \right] + 4\Gamma Q^2_0 Q_1^2\right\}\nn

3596: \ea

3597: $\frac{Q_0}{Q_1}\leq \cE_0^r$   then implies that  

3598: \ba

3599: \frac{d}{dt}\left(\frac{Q_0}{Q_1}\right)

3600:  &&\geq \frac{1}{Q_1^2}\left\{ 2\Gamma Q_0\left[Q_1^3 - \cE_0^{r/2} 

3601:  Q_1^{m+1/2}\right]+4\Gamma

3602: Q_0^2 Q_1^2\right\}\nn\\

3603: &&\geq \Gamma \left(\frac{Q_0}{Q_1}\right) Q_1^2 + 4\Gamma

3604: Q_1^2\left(\frac{Q_0}{Q_1}\right)^2 

3605: \nn\\

3606: &&\ge \Gamma \left(\frac{Q_0}{Q_1}\right) Q_1^2,\nn\\

3607: \ea

3608: for $m + \frac{1}{2} > 3$ and $|Q_1| \leq \cE_0 << 1$.

3609: Hence

3610: $$

3611: \frac{Q_0}{Q_1}\bigg|_t\geq \frac{Q_0}{Q_1}\bigg|_{t_0} 

3612: \exp(\Gamma\int^{t}_{t_0} Q^2_1(s)\ ds).

3613: $$

3614: Since $Q_1>0$, either $Q_1\downarrow 0$ or $\frac{Q_0}{Q_1}$ grows

3615:  exponentially with $t\ge t_0$. In either case, there exists

3616:  $t_1\ge t_0$, such that for $t=t_1$,  

3617: $ \frac{Q_0}{Q_1}\bigg|_{t_1} = \cE_0^r$. 

3618: Now whenever, 

3619: \be \frac{Q_0}{Q_1}\ge \cE_0^r,\label{eq:Q0overQ1}\ee

3620:  we have

3621: \ba

3622: \frac{dQ_1}{dt} &\leq& - 4\Gamma Q_0Q_1^2 + 

3623: + \cO(\sqrt{Q_0}Q_1^m)\nn\\

3624: &\leq& - 4\Gamma Q_0Q_1^2 + Q_0Q_1^{m-1/2}

3625: \cE_0^{-r/2}\nn\\

3626: &\leq& - 4\Gamma Q_0\left[Q_1^2 - C\cE_0^{-r/2} \cE_0^{m -5/2}

3627: Q_1^2\right]\nn\\

3628: &\leq& - 4\Gamma Q_0 Q_1^2\label{eq:ddtQ1bd}\ea

3629: for $m > \frac{5}{2} + \frac{r}{2}$.

3630: Therefore, by (\ref{eq:Q0overQ1}) we have

3631:  $\frac{dQ_1}{dt}\bigg|_{t_1}<0$. 

3632: Since $Q_0$ is increasing for $t\ge t_1\ge t_0$,

3633:  the inequality  $\frac{Q_0}{Q_1}\ge \cE_0^r$ persists 

3634:  and (\ref{eq:ddtQ1bd}) holds for all $t\ge t_1$.

3635: Hence

3636: \be

3637: \frac{dQ_1}{dt} \leq - 4\Gamma' Q_0 Q_1^2 \ {\rm and}\ 

3638:     \frac{dQ_0}{dt} \geq 2\Gamma'

3639: Q_0 Q_1^2.

3640: \label{eq:Q0Q1f}

3641: \ee

3642: 

3643: 

3644: Finally, for $t\ge t_1$

3645: \be {dQ_1\over dt} \le -4{19\over20}\Gamma (\inf_{[t_0,T]}Q_0)Q_1^2.

3646: \label{eq:Q1e}

3647: \ee

3648: Solving this scalar inequality 

3649: \be

3650: Q_1(t)\le {Q_1(t_1) \over 1+4\Gamma'' Q_1(t_1)

3651:  |Q_0(t_1)|(t-t_1)}

3652: \label{eq:Q1bound}

3653: \ee

3654: Since $P_1\le Q_1$, $P_1(t)$ decays to zero like $\la t-t_1\ra^{-1}$.

3655: This completes the proof of Proposition \ref{prop:propC}.

3656: 

3657: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3658: \section{Decomposition and estimation of the dispersion} 

3659: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3660: In this section, we revisit the decomposition of the dispersive part, $\eta$,

3661: which satisfies equation (\ref{eq:eta_eqn}). Here, we decompose $\eta$ 

3662: in a manner suitable for consideration of the solution on the various time scales.

3663: 

3664: \begin{prop}\label{prop:eta-decomp}

3665: 

3666: \be \eta(t)\ =\ \eta_0(t)\ +\ e_0(t)\ +\ \tetab(t).\label{eq:eta-decomp}

3667: \ee

3668: The three terms can be described as follows:\ 

3669: 

3670: \nit (i)

3671:   $\eta_0(t)$ is a dispersive wave generated by the data $\eta(0)=\eta_{\rm in}$:

3672: \be i\D_t\ \eta_0\ =\ \cH_0\ \eta_0,\ \eta(0)=\eta_{\rm in},

3673: \label{eq:eta0-eqn}\ee

3674: 

3675: \nit (ii) $e_0(t)$ is driven principally by $\eta_0(t)$:

3676: \be

3677: i\D_t\ e_0\ =\ \cH_0\ e_0\ + P_cS^{(0)}[e_0,\tetab;\eta_0],\ \ e_0(0)=0,

3678:  \label{eq:e0-eqn}

3679: \ee

3680: and

3681: 

3682: \nit (iii)

3683:  $\eta_b(t)$, which is driven by bound state dynamics: 

3684: \be

3685: i\D_t\ \tetab\ =\ \cH_0\ \tetab\ + P_cS^{(b)}[e_0,\tetab;\eta_0],\ 

3686:  \ \tetab(0)=0.\label{eq:etab-eqn} 

3687: \ee

3688: 

3689: We display expressions for $S^{(0)}$ and $S^{(b)}$ 

3690:  with the detail required

3691: in our analysis. We let $\chi$ denote a generic 

3692:  exponentially localized function of

3693:  position, $x$.

3694: \ba

3695: S^{(0)}\ 

3696: &\equiv&\ S^{(0)}_0+S^{(0)}_1e_0+S^{(0)}_2e_0^2+e_0^3\nn\\

3697: &\sim& \left(\alpha_0^2(t)-\alpha_0^2(T)\right)\chi\eta_0

3698:  + \eta_0^3 + \alpha_0\alpha_1\chi\eta_0 +\alpha_0\chi\eta_0^2  

3699: \nn\\

3700: &+&\ \left( \alpha_0\alpha_1+\alpha_1^2 \right)\chi e_0  +

3701: \left(\alpha_0+\alpha_1\right)\chi (\eta_0 +\tetab) e_0 + \eta_0\tetab e_0

3702: \nn\\

3703: &+&\ \left(\alpha_0^2(t)-\alpha_0^2(T)\right)\chi e_0  

3704: +\ (\alpha_0+\alpha_1)\chi e_0^2 + (\eta_0+\tetab)e_0^2\nn\\ 

3705: &+&\ e_0^3 

3706: \label{eq:S0-def}\\

3707: S^{(b)}\ &\equiv&\ S^{(b)}_0+S^{(b)}_1\tetab+S^{(b)}_2\tetab^2+\tetab^3\nn\\

3708: &\sim& \left(\alpha_0^2(t)-\alpha_0^2(T)\right)\chi \tetab\nn\\

3709: &+&  

3710:  i P_c(T)\left(\begin{array}{c}

3711:  \vec\nabla_1\Psi_1(t)\cdot\D_t\vec\alpha_1(t) +

3712:  \vec\nabla_0\Psi_0(t)\cdot\D_t\vec\alpha_0(t)\\

3713:  c.c. \end{array}\right)\nn\\ 

3714: &+& \left( E_1(t)-E_0(t) \right)P_c(T)\sigma_3

3715:              \left(\begin{array}{c}\Psi_1(t)\\ c.c.\end{array}\right) 

3716: \nn\\

3717: &+&\ \left(\ |\alpha_0|^2\ |\alpha_1|\ +\ |\alpha_0|\ |\alpha_1|^2\ \right)  \chi 

3718: \nn\\

3719: &+&\ \alpha_0\alpha_1\chi\tetab +\left(\alpha_0+\alpha_1\right)\chi\tetab^2 + 

3720:  \left(\eta_0+e_0\right)\tetab^2\nn\\

3721: &+&\ |\alpha_0|\chi\eta_0\tetab\ +\ \left(\eta_0^2+e_0^2\right)\tetab + \tetab^3 

3722: \label{eq:Sb-def}

3723: \ea

3724: \end{prop}

3725: 

3726: By  (\ref{eq:phi2H1bound}) we have the following $H^1$ bounds on $e_0$ 

3727: and $\tetab$ in terms of one another:

3728: \ba

3729: \| e_0(t)\|_{H^1}\ &\le&\ \cE_0 + \|\eta_0\|_{H^1} + \|\tetab(t)\|_{H^1}

3730:  \label{eq:e0H1}\\

3731: \| \tetab(t)\|_{H^1}\ &\le&\ \cE_0 + \|\eta_0\|_{H^1} + \|e_0(t)\|_{H^1}.

3732:  \label{eq:etab0H1}

3733: \ea

3734: 

3735: Using DuHamel's formula, both the $e_0$ and $\tetab$ equations can be written as

3736: equivalent integral equations: 

3737: \ba

3738: e_0(t)\ &=&\ -i\int_0^t\ e^{-i\cH_0(t-s)}\ P_c\ S^{(0)}(s)\ ds\label{eq:e0-int}\\

3739: \tetab(t)\ &=&\ -i\int_0^t\ e^{-i\cH_0(t-s)}\ P_c\ S^{(b)}(s)\ ds

3740: \label{eq:etab-int}

3741: \ea

3742: Therefore, in both cases we must estimate an expression of the form:

3743: \be

3744: w(t)\ =\ \int_0^t\ e^{-i\cH_0(t-s)}\ P_c\ S(s)\ ds.

3745: \label{eq:w-int}

3746: \ee

3747: For estimation, we shall require the following class of dispersive estimates:

3748: \begin{prop}\label{prop:regdecay}

3749: \nit (1)\ Let $2\le p<\infty,\ p',q\ge2,q'$ and $s$ be related by: 

3750: \be

3751: 3\left({1\over q}-{1\over p}\right) = s,\

3752: \  p^{-1} + (p')^{-1} = 1,\

3753:   q^{-1} + (q')^{-1} = 1.

3754: \nn\ee

3755: Then,

3756: \be

3757: \| e^{-i\cH_0t}P_cf\|_p\ \le\

3758: |t|^{-3({1\over2}-{1\over q})}

3759: \Jt^{-3({1\over q}-{1\over p})}

3760: \left(\ \| f\|_{p'}\ +\ \|\D^s f\|_{q'}\ \right)

3761: \label{eq:ge0}

3762: \ee

3763: 

3764: \nit (2) Assume $q\ge2$ and $s>3/q$. Then,

3765: \be

3766: \| e^{-i\cH_0t}P_cf\|_\infty\ \le\

3767: |t|^{-3({1\over2}-{1\over q})}

3768: \Jt^{-{3\over q}}

3769: \left(\ \| f\|_1\ +\ \|\D^s f\|_{q'}\ \right)

3770: \label{eq:ge0A}

3771: \ee

3772: \end{prop}

3773: {\bf Proof of Proposition \ref{prop:regdecay}:}

3774: We use the classical Sobolev inequality for functions

3775: defined on $\R^3$:

3776: \be

3777: \|f\|_p\le C\ \|\D^sf\|_q,\ \ \  3(q^{-1}-p^{-1}) = s

3778: \label{eq:sobolev}

3779: \ee

3780: and the $L^p$ -- $L^{p'}$ estimate (Theorem \ref{theo:lin_asymp_stab})

3781: \be \| e^{-i{\cal H}_0t } P_c f\|_p\ \le\ C_{1,p}\ t^{-{3\over2}+{3\over p}}

3782:  \|f\|_{p'},\ p^{-1}+(p')^{-1}=1.\label{eq:ge1}\ee

3783: 

3784: For $|t|\ge1$, we use (\ref{eq:ge1}). For $|t|\le1$, we have

3785: by (\ref{eq:sobolev}) that

3786: \ba

3787: \| e^{-i{\cal H}_0t}P_c f\|_p\ &\le&\ \|\D^s e^{-i{\cal H}_0t}P_c f\|_q\nn\\

3788:  &\le&\ C\ \| \cH_0^{s\over2}e^{-i{\cal H}_0t}P_c f\|_q\nn\\

3789:  &\le&\ C\ |t|^{-3({1\over2}-{1\over q})}\ \| D^s f\|_{q'},\label{eq:le1}\\

3790:  && q^{-1} + (q')^{-1}=1\nn

3791: \ea

3792: From estimates (\ref{eq:ge1}) and (\ref{eq:le1})

3793:  we obtain (\ref{eq:ge0}). Estimate (\ref{eq:ge0A})

3794:  is obtained similarly.

3795:  This completes the proof of Proposition

3796:  \ref{prop:regdecay}).

3797: \medskip

3798: 

3799: We now apply Proposition \ref{prop:regdecay} with $q=4$ and $s=1>3/4$

3800:  to the integral 

3801:  equation (\ref{eq:w-int}) and obtain the bound: 

3802: \be

3803: \|w(t)\|_\infty\ \le\ \int_0^t\ \frac{1}{|t-s|^{3\over4}}

3804:  \frac{1}{\la t-s\ra^{3\over4}}\ 

3805: \left(\ \|S(s)\|_1+\|\D S(s)\|_{4\over3}\ \right)\ 

3806:   ds

3807: \label{eq:w-est}

3808: \ee

3809: We shall use this bound as the first step in estimating $\|e_0(t)\|_\infty$

3810: and $\|\tetab(t)\|_\infty$.

3811: \medskip

3812: 

3813: \nit More specifically, our estimation strategy seeks a closed system of 

3814: inequalities for the following

3815: 

3816: \nit norms of $e_0$:\ 

3817:  {\bf 1.}\ $[e_0]_{\infty,\3half},\ \ 

3818:  {\bf 2.}\ [\D e_0]_{\L2loc;\3half},\ \ {\bf 3.}\ [e_0]_{H^1;0}$

3819: 

3820: \nit norms of $\tetab$:\ 

3821: $\ {\bf 4.}\ [\eta_b]_{\infty;0}, \ \ {\bf 5.}\  [\eta_b]_{H^1;0}$,  

3822: 

3823: \nit in terms of norms of the initial data 

3824: $[\phi(0)]_X$.

3825: % \ \ [\eta_0]_{W^{1,\infty};\3half}.

3826: \medskip

3827: 

3828: \centerline{\bf 1.\ Estimation of $\|e_0\|_\infty$}

3829: 

3830: By (\ref{eq:w-est}), we have

3831: \be

3832: \|e_0(t)\|_\infty\ \le\ \int_0^t\ \frac{1}{|t-s|^{3\over4}}

3833:  \frac{1}{\la t-s\ra^{3\over4}}\

3834: \left(\ \|S^{(0)}(s)\|_1+\|\D S^{(0)}(s)\|_{4\over3}\ \right)\

3835:   ds,

3836: \label{eq:e0-est}

3837: \ee

3838: so it sufficies to bound $\|S^{(0)}\|_1$ and 

3839:  $\|\D S^{(0)}\|_{4\over3}$. 

3840: \medskip

3841: 

3842: \nit\underline{\bf $\|S^{(0)}\|_1$:}

3843:  For any $t\ge0$, 

3844: \ba

3845: \|S_0^{(0)}\|_1\ &\le&\ C\la s\ra^{-{3\over2}}

3846:  \left(\ \cE_0[\eta_0(0)]_X + \sqrt{\cE_0} [\eta_0(0)]_X^2\ \right)

3847:  \le C(\cE_0)D(\eta(0))\ \la s\ra^{-{3\over2}}

3848: \nn\\

3849: \|S_1^{(0)}e_0\|_1\ &\le&\ C\la s\ra^{-{3\over2}}\ [e_0]_{\infty;{3\over2}}

3850:  \left(\ \cE_0 + \cE_0^{1\over2}[\tetab]_{\infty;0} + 

3851:  \cE_0^{1\over2}[\eta_0]_{\infty;0} + [\eta_0]_{\infty;{3\over2}}^2\ \right)\nn\\ 

3852:  &+& C\la s\ra^{-{3\over2}}\ [e_0]_{\infty;{3\over2}}

3853:   [\eta_0]_{\infty;{3\over2}} [\tetab]_{\infty;0}\ \le\

3854:  C\la s\ra^{-{3\over2}}

3855: \nn\\

3856: \|S_2^{(0)}e_0^2\|_1\ &\le&\ C\la s\ra^{-3}\ [e_0]_{\infty;{3\over2}}^2

3857:  \left(\ C\sqrt{\cE_0} + [\eta_0]_{\infty;{3\over2}} + [\tetab]_{\infty;0}

3858: \ \right)\nn\\

3859: \|e_0^3\|_1\ &\le&\ C\ \la s\ra^{-{3\over2}}[e_0]_{\infty;{3\over2}}

3860:                                       [e_0]_{H^1;0}^2

3861: \label{eq:S00L1}

3862: \ea

3863: 

3864: \nit\underline{\bf $\|\D S^{(0)}\|_{4\over3}$:}

3865:  For any $t\ge0$,

3866: 

3867: \ba

3868: \|\D S_0^{(0)}\|_{4\over3}\ &\le&\ C\la s\ra^{-{3\over2}}\ \left(\

3869: \cE_0 [\D\eta_0]_{\infty;{3\over2}}\ +\ 

3870: \cE_0^{1\over2} [\eta_0]_{\infty;{3\over2}} [\eta_0]_{H^1;0}\ 

3871: +\ \cE_0^{1\over4} [\eta_0]_{\infty,{3\over2}}^{3\over2}\ 

3872:  \right)\nn\\

3873: \|\D S_1^{(0)}e_0\|_{4\over3}\ &\le&\ C\la s\ra^{-{3\over2}}\left(\

3874: \cE_0 [\eta_0]_{\infty;{3\over2}}\ +\ \cE_0 [e_0]_{\L2loc;{3\over2}}

3875: \right) \nn\\ 

3876: &+& C\la s\ra^{-{3\over2}}\left(\

3877:  \cE_0^{1\over2} [\tetab]_{\infty;0} [\eta_0]_{\infty;{3\over2}}\ 

3878: +\ \cE_0^{1\over2} [e_0]_{\infty,{3\over2}}  [\eta_b]_{H^1;0}

3879:  + \cE_0^{1\over2} [\eta_b]_{\infty;0} [\D e_0]_{\L2loc;{3\over2}}

3880:  \right)\nn\\

3881: \|\D S_2^{(0)}e_0^2\|_{4\over3}\ &\le&\ 

3882: C \la s\ra^{-{3\over2}}  [e_0]_{\infty;{3\over2}} 

3883: \left(\

3884: \cE_0^{1\over2}[e_0]_{\infty;{3\over2}}+\cE_0^{1\over2}[e_0]_{H^1;0}\right)\nn\\ 

3885:  &+& C \la s\ra^{-{3\over2}} [e_0]_{\infty;{3\over2}}

3886:  \left( [\eta_0]_{H^1;0} + [\tetab]_{H^1;0} \right)\ [e_0]_{H^1;0} 

3887:  \nn\\

3888:  &+&\ [e_0]_{\infty;{3\over2}}^{3\over2}\ [e_0]_{L^2;0}^{1\over2}

3889:   \left( [\eta_0]_{H^1;0}+[\tetab]_{H^1;0}\right) 

3890: \nn\\

3891: \|\D e_0^3\|_{4\over3}\ &\le&\ C\la s\ra^{-{3\over2}}\ 

3892:  [e_0]_{\infty;{3\over2}}\ [e_0]_{H^1;0}^2

3893: \label{eq:DrS00L43}

3894: \ea

3895: 

3896: Use of the bounds (\ref{eq:S00L1}) and (\ref{eq:DrS00L43}) 

3897:  in (\ref{eq:e0-est}) implies

3898: \begin{prop}\label{prop:e0infty-est}

3899: \ba \|e_0(t)\|_\infty\ &\le&\ \la t\ra^{-\3half}\left(\

3900:  \cC(\cE_0,[\eta_0(0)]_X)\ +\ \cE_0^\half[\tetab(t)]_{\infty;0}

3901: \ +\ [\eta_0(0)]_X\ [\tetab(t)]_{\infty;0}\ [e_0]_{\infty;\3half}\ \right)\nn\\

3902: &+&\ \la t\ra^{-\3half}

3903:  \left(\ \cE_0^\half+[\eta_0(0)]_X\ +\ [e_0(t)]_{H^1;0}^2\ +\ 

3904:  [\tetab(t)]_{\infty;\3half}\ \right) 

3905:  \left(\ [e_0]_{\infty;\3half}+[e_0]_{\infty;\3half}^2\ \right)\nn\\

3906: &+&\ \la t\ra^{-\3half}\left(\ 

3907:  [e_0]_{\infty;\3half}\ [e_0]_{H^1;0}^2\ +\ \cE_0 [e_0]_{\L2loc;\3half}

3908: \ +\ \cE_0^{1\over2} [e_0]_{\infty;\3half}^2\ \right)

3909: \label{eq:e0-inftyest}

3910: \ea

3911: \end{prop}

3912: 

3913: \medskip

3914: 

3915: \centerline{\bf 2.\ Estimation of $[\D e_0(t)]_{\L2loc;{3\over2}}$}

3916: 

3917: \ba

3918: \|\D e_0(t)\|_{\L2loc}\ &\sim& \|\chi\D e_0(t)\|_2\ \ \ \left(\ 

3919:  \chi\ {\rm localized}\ \right)\nn\\ 

3920:  &\sim&\ \| \chi \la\cH_0\ra^{1\over2} e_0(t)\|_2

3921: \ \ \left(\ \D\ \la\cH_0\ra^{-{1\over2}}\in \cB(L^2)\ \right)\nn\\

3922:  &\le&\ \int_0^t\ \|\chi \la\cH_0\ra^{1\over2} e^{-i\cH_0(t-s)}P_cS^{(0)}(s)\|_2\ ds

3923: \ \ \ \left(\ {\rm by}\ (\ref{eq:e0-int})\ \right)\nn\\

3924: && \label{eq:ldest-start}

3925: \ea

3926: For the purpose of continuing the estimation, we regard $S^{(0)}$ as consisting

3927: of terms of two types: (i)\ terms having spatially localized (exponentially)

3928:  functions of $x$ as a factor, coming from $\Psi_j,\ j=0,1$, and (ii)\ terms like 

3929:  $e_0^3$ or $\tetab \eta_0 e_0$ and others which are not of this type.

3930:  It is convenient to refer to such terms as $S^{(0)}_{\rm LOC}$ and 

3931:  $S^{(0)}_{\rm NLOC}$ below. From 

3932: (\ref{eq:ldest-start})  we have:

3933: \ba

3934: \|\D e_0(t)\|_{\L2loc}\ &\le&\ 

3935:  \int_0^t \|\chi \la\cH_0\ra^{1\over2} e^{-i\cH_0(t-s)}

3936:  P_cS^{(0)}_{\rm LOC}(s)\|_2\ ds\nn\\

3937:  &+&\ \left(\int_0^{t-1}+\int_{t-1}^t\right)

3938:   \|\chi \la\cH_0\ra^{1\over2} e^{-i\cH_0(t-s)}

3939:  P_cS^{(0)}_{\rm NLOC}(s)\|_2\ ds \nn\\

3940: &\le&

3941:   \int_0^t \frac{C}{\la t-s\ra^{3\over2}}\ \|\la x\ra^\sigma S^{(0)}_{\rm LOC}\|\ ds

3942: \nn\\

3943: &+&\ \int_0^{t-1} \| e^{-i\cH_0(t-s)}

3944:  P_c\la\cH_0\ra^{1\over2}S^{(0)}_{\rm NLOC}(s)\|_\infty\ ds\ +\ 

3945:  \int_{t-1}^t \|\la\cH_0\ra^{1\over2} e^{-i\cH_0(t-s)}

3946:  P_cS^{(0)}_{\rm NLOC}(s)\|_2\ ds\nn\\

3947: &\le& \int_0^t 

3948:  \frac{C}{\la t-s\ra^{3\over2}}\ \|\la x\ra^\sigma S^{(0)}_{\rm LOC}\|\ ds

3949: \nn\\

3950: &+&\ C\ \int_0^{t-1} \frac{1}{|t-s|^{3\over2} }\ \|

3951:  \ \D\ S^{(0)}_{\rm NLOC}(s)\|_\infty\ ds\ +\ 

3952:  \int_{t-1}^t \|

3953:  \D S^{(0)}_{\rm NLOC}(s)\|_2\ ds\nn\\ 

3954: &=&\ {\bf A}\ +\ {\bf B}\ +\ {\bf C}

3955: \label{eq:ABCdef}

3956: \ea

3957: 

3958: We estimate the integrands of ${\bf B}$ and ${\bf C}$; that of ${\bf A}$ 

3959:  is similar.

3960: 

3961: \be

3962: \left|\ {\rm integrand\ of\  {\bf B}}\ \right|\ \le\ C\ \la t\ra^{-{3\over2}}\ 

3963:  \left(\ [e_0]_{\infty;{3\over2} }+[\eta_0]_{\infty;{3\over2} }\ \right)

3964: \ \left(\ 

3965:  [e_0]_{H^1;0}^2\ +\ [\tetab]_{H^1;0}^2\ +\ [\eta_0]_{H^1;0}^2\ \right)  

3966: \label{eq:Best}

3967: \ee

3968: 

3969: \ba

3970: \left|\ {\rm integrand\ of\  {\bf C}}\ \right|\ 

3971:  &\le&\ \la t\ra^{-{3\over2}}\left(\     

3972:   [e_0]_{\infty;{3\over2}}^2\left(\ [e_0]_{H^1;0}+[\tetab ]_{H^1;0}\ \right)

3973: \right.

3974: \nn\\

3975: &+&\ \left. [e_0]_{\infty;{3\over2}} [\eta_0]_{\infty;{3\over2}}[\tetab]_{H^1;0}

3976:  \right. \nn\\

3977: &+&\ \left. [\eta_0]_{W^{1,\infty};{3\over2}}\left(

3978:  [\tetab]_{\infty;0} [e_0]_{H^1;0}+[\eta_0]_{\infty;{3\over2}}[\tetab]_{H^1;0}

3979:  + [e_0]_{\infty;{3\over2}}[e_0]_{H^1;0} \right) \right.\nn\\

3980: &+& \left. [\eta_0]_{W^{1,\infty};{3\over2}}^2[e_0]_{H^1;0}\ \right)

3981: \nn\ea

3982: \medskip

3983:  

3984: Substitution of these bounds into (\ref{eq:ABCdef}), we have:

3985: 

3986: \begin{prop}\label{prop:De0localdecay}

3987: \be  \|\D e_0(t)\|_{\L2loc;{3\over2}}\ \le\  

3988:  C\la t\ra^{-\3half}\left(\ [e_0]_{\infty;\3half}\ +\ [\eta_0(0)]_X\ \right)

3989:  \ \left(\ [e_0]_{\infty;\3half}\ +\ [e_0]_{H^1;0}\ \right)

3990: \label{eq:De0loc-est}\ee

3991: \end{prop}

3992: 

3993:  

3994: \centerline{\bf 3.\ Estimation of $[e_0(t)]_{H^1;0}$}

3995: 

3996: \be

3997: \|e_0(t)\|_{H^1}\ \sim 

3998:  \|\cH_0^{1\over2}e_0(t)\|_2\ \le\ \int_0^t \|\D S^{(0)}(s)\|_2

3999:  \ ds\label{eq:e0H1start}\ee

4000: 

4001: \ba

4002: \|\D S_0^{(0)}\|_2\ &\le& C\la s\ra^{-{3\over2}}\ 

4003: \left(\ 

4004:  \cE_0 [\eta_0]_{W^{1,\infty};{3\over2}} + 

4005:  \cE_0^{1\over2}[\eta_0]_{\infty;{3\over2}}^2 \right.\nn\\

4006: &+&\ \left. [\eta_0]_{H^1;0}\ [\eta_0]_{\infty;{3\over2}}^2\ +\ 

4007:  \cE_0\ [\eta_0]_{W^{1,\infty};{3\over2}}\ \right)

4008: \nn\\

4009: \|\D S_0^{(1)}e_0\|_2\ &\le& C\la s\ra^{-{3\over2}}\

4010: \dots\nn\\

4011: \|\D S_0^{(2)}e_0^2\|_2\ &\le& C\la s\ra^{-{3\over2}}\

4012:  \left(\ \cE_0^{1\over2}[e_0]_{\infty;{3\over2}}^2 +

4013: \cE_0^{1\over2}[e_0]_{\infty;{3\over2}}[\eta_0]_{H^1;0}\ \right)\nn\\

4014: &+&\ C\la s\ra^{-{3\over2}}\ [e_0]_{\infty;{3\over2}}^2

4015:  \left( [\eta_0]_{H^1;0}+[\tetab]_{H^1;0} \right)\nn\\

4016: &+&\ C\la s\ra^{-{3\over2}}\ 

4017:  [e_0]_{\infty;{3\over2}} [\eta_0]_{W^{1,\infty};{3\over2}}

4018:  \left( [\eta_0]_{2;0}+[\tetab]_{2;0} \right)

4019: \nn\\

4020: \|\D e_0^3\|_2\ &\le&

4021:  C\la s\ra^{-3}\ [e_0]_{\infty;{3\over2}}^2\ [e_0]_{H^1;0}

4022: \nn\ea

4023: 

4024: 

4025: \begin{prop}\label{prop:e0H1} For $t\ge0$,

4026: \be

4027: \|e_0(t)\|_{H^1}\ \le\ \cC(\cE_0,[\eta_0(0)]_X)\ +\ 

4028:  [e_0]_{\infty;\3half}\left(\ [\eta_0(0)]_X\ +\ [e_0]_{H^1;0}\ \right) 

4029: \nn\ee

4030: \end{prop}

4031: \medskip

4032: 

4033: We now turn to the estimation of $\|\tetab(t)\|_\infty$ and 

4034:  $\|\tetab(t)\|_{H^1}$. By (\ref{eq:w-est}) we have

4035: \be

4036: \|\tetab(t)\|_\infty\ \le\ \int_0^t\ \frac{1}{|t-s|^{3\over4}}

4037:  \frac{1}{\la t-s\ra^{3\over4}}\

4038: \left(\ \|S^{(b)}(s)\|_1+\|\D S^{(b)}(s)\|_{4\over3}\ \right)\

4039:   ds,

4040: \label{eq:etab-est}

4041: \ee

4042: where $S^{(b)}$ is given by (\ref{eq:Sb-def}).

4043: 

4044: \begin{prop}\label{eq:SbQ}

4045: In terms of $Q_j,\ j=0,1$ we have

4046: \be

4047: S^{(b)}\ =\ S^{(b)}_0 + S_1^{(b)}\tetab + S_2^{(b)}\tetab^2 + \tetab^3 

4048: \nn\ee

4049: where

4050: \ba

4051: S_0^{(b)}\ &\sim&\ 

4052:  Q_0^{1\over2}Q_1 \chi\ +\ Q_0Q_1^{1\over2}\chi\nn\\

4053: S_1^{(b)}\tetab\ &\sim&\ Q_0^{1\over2}Q_1^{1\over2}\chi\tetab

4054:  + (\eta_0^2+e_0^2)\tetab + Q_0^{1\over2}\chi\eta_0\tetab\nn\\

4055: S_2^{(b)}\tetab^2\ &\sim&\ \left( Q_0^{1\over2}+Q_1^{1\over2}\right)\chi\tetab^2

4056:  + \left(\eta_0+e_0\right)\tetab^2

4057: \nn\ea

4058: \end{prop}

4059: 

4060: We now proceed with estimates of $S^{(b)}$ in $L^1$ and in $W^{1,{4\over3}}$.

4061: 

4062: Beginning with $\|S^{(b)}\|_1$ we have:

4063: \ba

4064: &&\|S^{(b)}\|_1\nn\\

4065: &\le& C\left(\  

4066:     Q_0^{1\over2}Q_1+Q_0Q_1^{1\over2}\  

4067:  +\ \ Q_0^{1\over2}Q_1^{1\over2}\|\tetab\|_\infty\ +\  

4068:   \|\tetab\|_\infty \left( \|\eta_0\|_2^2+\|e_0\|_2^2\right)

4069: \ \right)\nn\\

4070:  &+&\ C\left(\ 

4071:   Q_0^{1\over2}\|\eta_0\|_\infty\|\tetab\|_2\ 

4072:  +\  \ \left( Q_0^{1\over2}+Q_1^{1\over2}\right) \|\tetab\|_\infty^2\ 

4073: +\ \|\tetab\|_\infty\left(\ \|\eta_0\|\ \|\tetab\|_2\ +\ 

4074:  \|e_0\|_2\ \|\tetab\|_2\ \right)\

4075:   \right)\nn\\

4076: &+&\ C\|\tetab\|_\infty\|\tetab\|_2^2

4077: \label{eqw:L1Sb}\ea

4078: 

4079: We now turn to $\|\D S^{(b)}\|_{4\over3}$.

4080: \ba

4081: &&\|\D S^{(b)}\|_{4\over3}\nn\\

4082: &\le&\ C\left(\ 

4083: Q_0^{1\over2}Q_1 + Q_0Q_1^{1\over2}+

4084: Q_0^{1\over2}Q_1^{1\over2}\ \|\tetab\|_{H^1}

4085: \ \right)\nn\\

4086: &+&\ C\left(\ 

4087: \| (\eta_0^2+e_0^2)\D\tetab\|_{4\over3} + 

4088:  \|\eta_0\D\eta_0\tetab\|_{4\over3} + 

4089:  Q_0^{1\over2}\|\tetab\|_2\left(\|\eta_0\|_\infty+\|\D\eta_0\|_\infty

4090:  \right)\ 

4091: \right)

4092: \nn\\

4093: &+&\ C\left(\ 

4094: Q_0^{1\over2}\|\eta_0\|_\infty\|\D\tetab\|_2 +

4095:  \|(\eta_0+e_0)\tetab\D\tetab\|_{4\over3}+ 

4096: \|(\D\eta_0+\D e_0)\tetab^2\|_{4\over3}\ +\ \|\tetab^2\D\tetab\|_{4\over3}

4097: \ \right)\nn\\

4098: \label{eq:L43Sb}

4099: \ea

4100: To further estimate $\|\D S^{(b)}\|_{4\over3}$ we use the following estimates

4101: of individual terms:

4102: \ba 

4103: \|\eta_0^2\D\tetab\|_{4\over3} &\le& \|\eta_0\|_\infty\ \|\eta_0\|_2^{1\over2}\

4104:  \|\tetab\|_{H^1}\nn\\

4105: \|e_0^2\D\tetab\|_{4\over3} &\le& \|\eta_0\|_\infty^{3\over2}\ 

4106:  \|e_0\|_2^{1\over2}\ \|\tetab\|_{H^1}\nn\\

4107: \|\eta_0\D\eta_0\tetab\|_{4\over3} &\le& \|\D\eta_0\|_\infty\ 

4108:  \|\eta_0\|_\infty^{1\over2}\ \|\eta_0\|_2^{1\over2}\ \|\tetab\|_2\nn\\

4109: \|\eta_0\tetab\D\tetab\|_{4\over3} &\le& \|\tetab\|_\infty\ 

4110:  \|\eta_0\|_\infty^{1\over2}\ \|\eta_0\|_2^{1\over2}\ \|\tetab\|_{H^1}\nn\\

4111: \|e_0\tetab\D\tetab\|_{4\over3} &\le& \|\tetab\|_\infty\ 

4112:  \|e_0\|_\infty^{1\over2}\ \|e_0\|_2^{1\over2}\ \|\tetab\|_{H^1}\nn\\

4113: \|\D\eta_0\tetab^2\|_{4\over3} &\le& \|\tetab\|_\infty\ 

4114:  \|\D\eta_0\|_\infty^{1\over2}\ \|\D\eta_0\|_4\ \|\tetab\|_2\nn\\

4115: \|\D e_0\tetab^2\|_{4\over3} &\le& \|e_0\|_{H^1}\ 

4116:  \|\tetab\|_\infty^{3\over2}\ \|\tetab\|_2^{1\over2}\nn\\

4117: \|\tetab^2\D\tetab\|_{4\over3} &\le& \|\tetab\|_\infty\ 

4118:  \|\D\tetab\|_2\ \|\tetab\|_2 

4119: \label{eq:forDSb}

4120: \ea

4121: \bigskip

4122: 

4123: Recall that $\|\tetab(t)\|_{H^1}$ can be estimated in terms of

4124:   $\|e_0(t)\|_{H^1}$; see (\ref{eq:etab0H1}).

4125: 

4126: %\ba

4127: %\|\tetab(t)\|_\infty\ 

4128: %&\le&\ \cC(\cE_0,[\eta_0]_X)\ [\tetab(t)]_{\infty;0}

4129: % \ + \ \cC([e_0(t)]_{H^1;0})\ [\tetab(t)]_{\infty;0}]\nn\\ 

4130: % +\ C\int_0^t\ \frac{1}{ |t-s|^{3\over4}\la t-s\ra^{3\over4} }\ 

4131: %\left( Q_0^{1\over2}(s)Q_1(s)\ +\ Q_0(s)Q_1^{1\over2}(s)\ \right)\  ds\nn\\

4132: %\label{eq:etab-est1}

4133: %\ea  

4134: 

4135: Since $\tetab$ is driven by the bound state amplitudes ($Q_0$ and $Q_1$),

4136:  which have different behavior on the intervals $I_j$, we 

4137:   now estimate $\tetab(t)$ separately on 

4138: $I_0~=~[0,t_0],\ I_1~=~[t_0,t_1]$ and $I_2~=~[t_1,\infty)$. 

4139: 

4140: We now introduce appropriate norms on different time scales. Define

4141: \ba

4142: M_{I_0}(t)\ &\equiv&\ \sup_{0\le t\wedge t_0}\ 

4143:                     \la t\ra^\3half\ \|e_0(t)\|_\infty 

4144: \ +\ \sup_{0\le t\wedge t_0}\ \la t\ra\ \|\tetab(t)\|_\infty\ +\ 

4145:  \sup_{0\le t\wedge t_0}\ \|e_0(t)\|_{H^1}\nn\\

4146:          &+&\ \sup_{0\le t\wedge t_0}\ 

4147:  \la t\ra^\3half\ \|\D e_0(t)\|_{\L2loc} 

4148: \ +\ \sup_{0\le t\wedge t_0} \la t\ra^2\ |Q_0(t)|\ +\ 

4149:  \sup_{0\le t\wedge t_0}\ |Q_1(t)| 

4150:  \label{eq:MI0def}\\

4151: M_{I_1}(t)\ &\equiv&\ \sup_{t_0\le t\wedge t_1}\ 

4152:               \la t\ra^\3half\ \|e_0(t)\|_\infty 

4153: \ +\ \sup_{t_0\le t\wedge t_1}\ \|\tetab(t)\|_\infty\ +\  

4154:  \sup_{t_0\le t\wedge t_1}\ \|e_0(t)\|_{H^1}\nn\\

4155:       &+&\ 

4156:  \sup_{t_0\le t\wedge t_1}\ \la t\ra^\3half\ \|\D e_0(t)\|_{\L2loc}

4157:  \ +\ \sup_{t_0\le t\wedge t_1}\ |Q_0(t)|\ +\ 

4158:  \sup_{t_0\le t\wedge t_1}\ |Q_1(t)| \label{eq:MI1def}\\

4159: M_{I_2}(t)\ &\equiv&\ \sup_{t_1\le t\wedge T}\ 

4160:  \la t\ra^\3half\ \|e_0(t)\|_\infty 

4161: \ +\ \sup_{t_1\le t\wedge T}\ \la t-t_1\ra^\half\ \|\tetab(t)\|_\infty\ +\  

4162:  \sup_{t_1\le t\wedge T}\ \|e_0(t)\|_{H^1}\nn\\

4163:       &+&\

4164:  \sup_{t_1\le t\wedge T}\ \la t\ra^\3half\ \|\D e_0(t)\|_{\L2loc}

4165:  \ +\ \sup_{t_1\le t\wedge T}\ |Q_0(t)|\nn\\

4166: & +&\ 

4167:  \sup_{t_1\le t\wedge T}\ \la t-t_1\ra

4168: \ \Gamma'|Q_0(t_1)|\ |Q_1(t_1)|\  |Q_1(t)| \label{eq:MI2def} 

4169: \ea

4170: \medskip

4171: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4172: \begin{rmk}\label{rmk:e0control}

4173: By Propositions \ref{prop:e0infty-est}, \ref{prop:De0localdecay} and 

4174:  \ref{prop:e0H1}, the $e_0$ contributions to the norms $M_{I_k}$ are controlled

4175: in terms of the initial conditions. Therefore, 

4176:  to control $M_{I_k}$, it suffices to 

4177: bound $Q_j$ and $\tetab$. 

4178: \end{rmk}

4179:  

4180: The above estimates can be used together with the bounds on $Q_j$ of section

4181:  \ref{sec:finite-dim-red} to obtain the following three propositions, which

4182: give bounds for $\tetab$ on the intervals $I_0,\ I_1$ and $I_2$.

4183: 

4184: \begin{prop}\label{prop:etabonI0} ($\tetab(t)$ for  $t\in I_0$)

4185: Assume $t\in I_0=[0,t_0]$, {\it i.e.}\ 

4186: $Q_0(t)\le C\left([\eta_0]_X+\cE_0\right)\la t\ra^{-2}$ for $t\in [0,t_0]$. 

4187: Then, for $\|\phi(0)\|_X$ sufficiently small

4188: \be

4189: \|\tetab(t)\|_\infty\ \le\ \cC\left(\ \|\phi(0)\|_X,M_{I_0}(t_0)\ \right)

4190:   \ \la t\ra^{-1}\label{eq:etabI0}

4191: \ee

4192: \end{prop}

4193: 

4194: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4195: 

4196: \begin{prop}\label{prop:etabonI1} ($\tetab(t)$ for  $t\in I_1$)

4197: Assume $t\in I_1=[t_0,t_1]$.

4198: Then, for $\|\phi(0)\|_X$ sufficiently small

4199: \be

4200: \|\tetab(t)\|_\infty\ \le\ \cC\left(\ \|\phi(0)\|_X,M_{I_0}(t_0),

4201:             M_{I_1}(t_1)\ \right)

4202: \label{eq:etabI1}

4203: \ee

4204: \end{prop}

4205: 

4206: 

4207: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4208: \begin{prop}\label{prop:etabonI2} ($\tetab(t)$ for  $t\in I_2$)

4209: Assume $t\in I_2=[t_1,T]$.

4210: Then, for $\|\phi(0)\|_X$ sufficiently small

4211: \be

4212: \|\tetab(t)\|_\infty\ \le\ \cC\left(\ \|\phi(0)\|_X,M_{I_0}(t_0),

4213:             M_{I_1}(t_1), M_{I_2}(T)\ \right)\ \la t-t_1\ra^{-\half}

4214: \label{eq:etabI2}

4215: \ee

4216: \end{prop}

4217: 

4218: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4219: 

4220: %\begin{rmk} \bXYZ\ What figures in the estimation of $e_0(t)$ is 

4221: % $[\tetab(t)]_{\infty,0}=\sup_{t\ge T}\|\tetab(t)\|_\infty$.

4222: %\end{rmk}

4223: 

4224: 

4225: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4226: %removed to insert2 12/5

4227: %11/27/01 - 

4228: 

4229: In our estimates of section \ref{sec:BeginProof}, we shall use the following

4230:  result to estimate the size of correction terms in the system for $Q_0$ and 

4231:  $Q_1$ for $t\ge t_0$, where $Q_0$ is monotonically increasing.

4232: 

4233: \begin{prop}\label{prop:disip}

4234: Let $\zeta= \zeta(x,t),\ x\in \R^3,\ t\ge t_0 $, with $\zeta(x,t_0) = 0$ 

4235:  satisfy the following dispersive equation 

4236: \be

4237: i\partial_t \zeta = 

4238:  \cH_0\zeta + P_c\left(\    

4239:  S_1(t) + S_2(t)\zeta + \chi\zeta^2 + \zeta^3\ \right) ,\label{eq:zetaeqn}

4240: \ee

4241: where for all $k\ge0$ and $j=1,2$:

4242: \ba

4243: \|S_1(t)\|_{H^k} &=& \cO\left(\cC([\phi(0)]_X\ Q_0^{1\over2}(t)\right)\nn\\

4244:  \D_tQ_0&\ge& 0\  {\rm for }\ t\ge t_0,\ {\rm and }\  Q_0 \leq\cE_0.

4245: \label{eq:Sbound}\ea

4246: 

4247: Suppose

4248:  $\|\zeta(t)\|_{H^1} \leq \cC([\phi(0)]_X, M_{I_1}(t_1),M_{I_2}(T)\ )$ 

4249:  for all

4250: $t_0\le t\le T$, where  $[\phi(0)]_X$ is sufficiently small.

4251: Then,

4252: $

4253: \|\zeta(t)\|_\infty\ \le\ C\left(\ 

4254:  [\phi(0)]_X, M_{I_1(t_1)},M_{I_2}(T)\ \right)\ Q_0(t)^{1/2} $. 

4255: \end{prop}

4256: 

4257: %{\bf XYZ --- LOTS OF EDITING NEEDED TO MAKE CONSISTENT WITH ABOVE 

4258: % STARTING HERE}

4259:  

4260: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4261: %\footnote{\bXYZ shall we omit the more general argument? 

4262: %\ba

4263: %\|fg\|_{p'} &\leq& \| f \|_p \|g\|_{p^{''}} \quad

4264: %\frac{1}{p}+\frac{1}{p'} = 1 \quad \frac{1}{p'} = \frac{1}{p} +

4265: %\frac{1}{p^{''}}\nn\\

4266: %p^{''} &=& (1-\frac{2}{p})^{-1} = \frac{p}{p-2} > 1\nn\\

4267: %&&\|\chi\eta^2\|_{p'} + \|\eta^3\|_{p'} +\cE_0^{\bar m} Q_0^{1/2}

4268: %(s) \|\chi\eta\|_{p'}\leq\\

4269: %&&\leq c \|\eta\|^2_p + \|\eta\|^{2-\delta'}_2 \cdot

4270: %\|\eta\|^{1+\delta}_p + \cE_0^{\bar m} Q_0^{1/2}(s) \|\eta\|_p\\

4271: %&&\leq c\|\eta\|^2_p + c\epsilon_0^{2-\delta'} \|\eta\|_p^{1+\delta} +

4272: %\cE_0^{\bar m} Q_0^{1/2} (s) \|\eta\|_p\\

4273: %&& + \|S(s)\eta\|_{H^{\bar p}}.

4274: %\nn\ea

4275: %\ba

4276: %\|\eta^3\|_{H^{\bar p}} + \|\chi \eta^2\|_{\dot H^{\bar p}} &\leq&

4277: %c\|\eta^2\|_{L^2} + c \|D^{\bar p}\eta\|_{L^{p^{''}}}\|\eta\|_{L^p}\nn\\

4278: %&+& c\cE_0^{\bar m} Q_0^{1/2}(s) \|\eta\|_p + c\cE_0^{\bar m}

4279: %Q_0^{1/2} (s) \|D^{\bar p}\eta\|_2\nn\\

4280: %&+&\|D^{\bar p}\eta \|_{p^{'''}}\|\eta^2\|_{p/2}\nn\\

4281: % && \frac{1}{p^{'''}} = \frac{1}{2} - \frac{2}{p}

4282: %\nn\ea

4283: %We can always have 

4284: %\ba

4285: %\|D^{\bar p} \eta \|_{p^{'''}}&\leq& c\| D\eta\|_2(\ {\rm by\rm choosing\rm

4286: %}\ \frac{p}{2} \to \frac{p}{2} + \delta \ {\rm if\rm needed\ })\nn\\ 

4287: %\|D^{\bar p} \eta \|_{p^{'''}}&\leq& c\| D\eta\|_2  \ {\rm since }\ 

4288: %p^{''} < 2.

4289: %\nn\ea

4290: %Hence

4291: %$$\|\zeta\|_p\leq c \cE_0^{\bar m} Q_0^{1/2} (t) \int^t_0

4292: %\frac{ds}{\langle t-s\rangle^{\frac{3}{2} -\frac{3}{p}}} + c \int^t_0

4293: %\frac{ds}{\langle t-s\rangle^{\frac{3}{2}

4294: %-\frac{3}{p}}}\left\{\epsilon_0\|\zeta\|_p +

4295: %\epsilon_0\|\zeta_p\|_p^{1+\delta} + \epsilon_0\|\zeta\|_p^2\right\}.

4296: %$$

4297: %Assuming $t\leq T$ so that $\|\eta\|_p < 1$.

4298: %Then

4299: %$$

4300: %\sup_{t\leq T}\|\zeta\|_p \leq c \epsilon^{\bar m}Q_0^{1/2}

4301: %(T)\int^T_0\frac{ds}{\langle t-s\rangle ^{\frac{3}{2} - 3/p}} +

4302: %c\cE_0 \int^T_0\frac{ds}{\langle t-s\rangle^{\frac{3}{2} -

4303: %3/p}}\left(\sup_{t\leq T}\|\zeta\|_p\right).

4304: %$$

4305: %Hence

4306: %$$

4307: %\sup_{t\leq T}\|\zeta\|_p \leq c \cE_0^{\bar m}

4308: %Q_0^{1/2}(T)\int^T_0\frac{ds}{\langle t-s\rangle ^{\frac{3}{2} -

4309: %3/p}}.

4310: %$$

4311: %}

4312: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4313: \begin{cor}\label{cor:Corollary 1}

4314: 

4315: Let $p>6$.  $\cE_0^{\bar m}=\sup_{t'\leq t} \epsilon_1 (t'),$ then 

4316: $$

4317: \|\zeta\|_p \leq c Q_0^{1/2}(t)

4318:  \int^t_0\frac{\epsilon_1(s)}{\la t-s\ra^{3/2 - 3/p}}\ ds.

4319: $$

4320: \end{cor}

4321: 

4322: \begin{cor}\label{cor:Corollary 2} Let $p>6$. 

4323: $$

4324: |\la\chi,\zeta\ra| \leq cQ_0^{1/2}(t)

4325:  \int^t_0\frac{\epsilon_1(s)}

4326: {\la t-s\ra^{3/2 - 3/p}}\ ds.

4327: $$

4328: which follows by using

4329: $$

4330: \|\chi e^{-iHt} P_c\bar \chi g\|_2\leq c \langle t\rangle^{-3/2}

4331:  \| g\|_2.

4332: $$

4333: \end{cor}

4334: 

4335: \begin{cor}\label{cor:Corollary 3}

4336: $$

4337: \| \D^k\zeta\|_\infty\leq CQ_0^{1/2}(t)\int^t_0 \frac{\epsilon_1(s)}{\langle

4338: t-s\rangle^{3/2}}\ ds.

4339: $$

4340: \end{cor}

4341: \nit{\bf Proof:} This result follows from applying $\D^k$ to the equation

4342: for $\zeta$ and estimating, as above, in $L^p$ for any $\alpha$ and 

4343: $p>6$. By the Sobolev inequality this implies control derivatives

4344:   in $L^\infty$.   

4345: \bigskip

4346: 

4347: \begin{rmk}

4348: In our applications of Proposition \ref{prop:disip} and its corollaries,

4349:  $\epsilon_1(s)$ will be given by the source terms depending on $Q_0$

4350: and $Q_1^{1\over2}$. For $t>t_1$, ~$Q_1=\cO(\la t-t_1\ra^{-1})$.  

4351:  Therefore, since

4352:  the lowest order term in $Q_1$ contributing to $\epsilon_1(s)$ is 

4353:  $\cO(Q_1)$ (\ see equation (\ref{eq:eta1_eqn})\ ), it follows

4354: that for $t>t_1$

4355: \be

4356: \|\tetab\|_{W^{k,\infty}} = 

4357:  \cO(\la t-t_1\ra^{-{1\over2}}) + \cO(\la t-t_1\ra^{-1}) 

4358:  +\cO(\la t-t_1\ra^{-{3\over2}}). 

4359: \nn\ee

4360: 

4361: Since $\eta = \cO(\tetab) + \cO(\eta_0)$, the conclusion of the main 

4362: theorem for large $t$, $t>t_1$, follows. Namely, for 

4363: \be

4364: \|\eta\|_{W^{k,\infty}} = \cO(\la t-t_1\ra^{-{1\over2}}) + 

4365:  e^{-i\cH_0t}P_c[\eta(0) + \cE_0^2],

4366: \nn\ee 

4367: where the non-free wave part is coming from spatially localized source terms.

4368: \end{rmk}

4369: 

4370: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4371: 

4372: \section{Beginning of proof of Proposition \ref{prop:propA} }

4373: \label{sec:BeginProof}

4374: 

4375: The key to Proposition  \ref{prop:propA} is the following more detailed

4376: version of Corollary \ref{cor:nlPME}:

4377: 

4378: \begin{prop}\label{prop:Master}

4379: Let $t\ge 0$.

4380: The equations for $P_0, P_1$ can be written in the following form

4381: \ba

4382: \frac{d P_0}{dt} &=& 2\Gamma P_0 P_1^2 + R_0^{(0)}[\eta_0] + R_1^{(0)}

4383: [\tetab] + R_2^{(0)}[P_0, P_1]\nn\\

4384: \frac{d P_1}{dt} &=& -4\Gamma P_0 P_1^2 + R_0^{(1)}[\eta_0] + R_1^{(1)}

4385: [\tetab] + R_2^{(1)}[P_0, P_1].

4386: \label{eq:P0P1}

4387: \ea

4388: where

4389: 

4390: \nit (i) $R_0^{(i)}[\eta_0]$ are $\eta_0$-dependent terms only both local and

4391: nonlocal in time, and may also depend on $\talpha_0, \tbeta_1$.

4392: 

4393: \nit (ii) $\tetab$  is the bound state driven part of the dispersion;

4394:  see section \ref{sec:dec-mod}. 

4395:  

4396: \nit (iii) $R_2^{(0)}$ depends only on $P_0, P_1, t$, but not on $\eta$;

4397:   It is

4398: formally linear in $P_0$, of high order in $P_1$,

4399: and contains both local and nonlocal in time terms.

4400: 

4401: \nit  (iv)

4402:  $R_2^{(1)}$ depends only on $P_0, P_1, t$, but not on $\eta$. It is of

4403: high order in $P_1$, local and nonlocal terms included, but has terms

4404: which are linear in $\tilde \alpha_0 \sim \sqrt{P_0}$.

4405: \end{prop}

4406: 

4407: The proof uses repeated application of near-identity 

4408: transformations of the variables $\tilde \alpha_0, \tilde \beta_1$, 

4409:  derivable by integration by parts 

4410: ( see the 

4411: discussion of resonant and nonresonant terms in section

4412:  \ref{sec:nfandmaster} )  and  the

4413: decomposition of $\eta$: $\eta=\eta_0+e_0+\tetab$; see equation 

4414:  (\ref{eq:eta-decomp}).

4415: %(see prop. $(\tetab)$).

4416: 

4417: The proof is long so we break it up into three parts, which are presented

4418: in three different sections. The following is an overview.

4419: 

4420: \nit {\bf Part 1:} 

4421:   The terms $R_0^{(0)}[\eta_0]$ and $R_0^{(1)}[\eta_0]$ are forcing terms

4422:  in the ODE dynamics,

4423:  which are driven by the

4424:  dispersive part of the initial conditions. They are studied and  estimated 

4425:   in this section;

4426:  Proposition \ref{prop:biR0i}

4427: 

4428: \nit {\bf Part 2:}  The terms $R_1^{(0)}[\eta_b]$ and $R_1^{(1)}[\eta_b]$

4429: are studied and estimated 

4430:  in section \ref{sec:Rj1teta}; Proposition \ref{prop:Riteta}.

4431: 

4432: \nit {\bf Part 3:} The terms $R_2^{(0)}[P_0,P_1]$ and $R_2^{(1)}(P_0,P_1)$

4433: are studied and estimated 

4434:   in section \ref{sec:R120P0P1}; Proposition \ref{prop:ODE}

4435: 

4436: In Parts 1, 2 and 3, we require estimates for all $t\ge0$.

4437: On $I_0=\{t:0\le t\le t_0\}$ we use  

4438:   the {\it \'a priori} bound on $P_0(t)$, implied by the definition 

4439: of $t_0$; see eqn. (\ref{eq:t0def}). For $t>t_0$, we use the  

4440: 

4441: \nit{\bf Monotonicity Property Q}

4442: 

4443: \be 

4444: Q_0\ {\rm and}\ \frac{Q_0}{Q_1}\ {\rm are\ montonically\  increasing},

4445: \label{eq:PropertyQ}\ee

4446: where $Q_0$ and $Q_1$

4447: are  the {\it modified} bound state energies related to $P_0$

4448: and $P_1$ (\ see (\ref{eq:Q0Q1def}), (\ref{eq:b0}), (\ref{eq:b1})\ ). 

4449:  This monotonicity 

4450: property shown to hold at $t=t_0$ and is then 

4451:  shown to continue for all time, $t$,

4452:  by a continuity argument; see section \ref{sec:bootstrap}.

4453: Since there are many terms, we focus on those which are most problematic,

4454: namely, those which are {\it nonlocal} in time and of 

4455:  slowest time-decay rate. This calculations are very lengthy 

4456: and before embarking on them we present a calculation, related to the 

4457: normal form discussion in section \ref{sec:nfandmaster}, and which is

4458: repeatedly in order to exploit rapid oscillations in time.

4459: \bigskip

4460: 

4461: \noindent{\bf Expansion of oscillatory integrals, resonances  and improved 

4462:  time decay} 

4463: 

4464: In deriving and estimating the terms $R_j^{(i)}$ in (\ref{eq:P0P1}), 

4465:  we must frequently

4466: expand and/or estimate terms of the form:

4467: \be

4468:  \left\la \chi_1\ ,\int^t_0 e^{-i\cH_0(t-s)} P_c\  

4469:  \chi_2(s)\ e^{i\Omega_1 s}\ ds

4470:  \right\ra

4471: \label{eq:osc-int}

4472: \ee

4473: Here,  $\chi_1$ and $\chi_2(s)$ are localized functions of $x$, $\chi_1$ is

4474:  independent of $s$ and $\chi_2(s)$ depends on $s$ through its dependence on

4475:  $\alpha_0(s),\beta_1(s)$ or $\eta_0(s)$. 

4476: 

4477: Recall that $\cH_0$, define in (\ref{eq:H_def}) is of the form

4478: \be

4479: \cH_0\ =\ {\tilde\cH_0}\ +\ E_0\sigma_3

4480: \nn\ee 

4481: and ${\tilde\cH_0}=\sigma_3(-\Delta)$ plus a matrix potential which 

4482: decays to zero rapidly as $|x|$ tends to infinity. In 

4483: (\ref{eq:osc-int}) we would like to integrate by parts, exploiting the 

4484: oscillation of frequency $E_0$. However, ``peeling off'' these oscillations

4485: is a little tricky because $\sigma_3$ does not commute with ${\tilde\cH_0}$.

4486: We handle this as follows.

4487: 

4488: In \cite{Cucc1,Cucc2}, it is shown that 

4489: \ba

4490: Z\cH_0 W\ &=&\ \sigma_3\left(-\Delta-E_0\right), 

4491: \label{eq:ZH0W}\nn\\

4492: Ze^{-i\cH_0tP_c} W\ &=&\ e^{-i\sigma_3\left(-\Delta-E_0\right)t}

4493: \label{eq:ZexpH0tW}

4494: \ea

4495: where $W$ denotes the wave operator

4496: \ba W\ &=&\ \lim_{t\to\infty} e^{-i\cH_0t}e^{-i\sigma_3(-\Delta-E_0)t},\nn\\

4497:     Z\ &=&\ W^{-1}

4498: \label{eq:Waveop}

4499: \ea

4500: By (\ref{eq:ZexpH0tW}), we can rewrite (\ref{eq:osc-int}) as

4501: \ba

4502: && \left\la \chi_1\ ,e^{-i\cH_0t}\int^t_0 

4503:  We^{i\sigma_3\left(-\Delta-E_0\right)s} Z P_c\

4504:   e^{i\Omega_1 s}\ \chi_2(s)\ ds\right\ra\nn\\

4505: \ \ &=&\ 

4506:  \left\la \chi_1\ ,e^{-i\cH_0t}\int^t_0 We^{-i\sigma_3\left(-\Delta-E_0\right)s} Z P_c\

4507:   e^{i\Omega_1 s}\ \chi_2(s)\ ds\right\ra\nn\\

4508: \ \ &=&\ 

4509:   \left\la \chi_1\ ,e^{-i\cH_0t}\int^t_0 We^{i\sigma_3\Delta s}Z\ \cdot\

4510:   We^{i\sigma_3E_0s} 

4511:  Z P_c\

4512:  e^{i\Omega_1 s}\ \chi_2(s)\ ds\right\ra\nn\\

4513: \ \ &=&\

4514:   \left\la \chi_1\ ,e^{-i\cH_0t}\int^t_0 We^{i\sigma_3\Delta s}Z\ \cdot\

4515:   We^{i(\sigma_3E_0+I\Omega_1)s}

4516:  Z P_c\

4517: \chi_2(s)\ ds\right\ra\nn\\

4518: \ \ &=&\ \left\la \chi_1\ ,e^{-i\cH_0t}\int^t_0 We^{i\sigma_3\Delta s}Z\ \cdot\

4519:   (-i)\frac{d}{ds}W(\sigma_3E_0+I\Omega_1)^{-1} e^{i(\sigma_3E_0+I\Omega_1)s}

4520:  Z P_c\

4521: \chi_2(s)\ ds\right\ra\nn\\

4522: \ \ &\sim&\  -\ 

4523:  \left\la \chi_1\ ,e^{-i\cH_0t}\int^t_0 W\sigma_3\Delta e^{i\sigma_3\Delta s}Z\ \cdot\

4524:   W(\sigma_3E_0+I\Omega_1)^{-1} e^{i(\sigma_3E_0+I\Omega_1)s}

4525:  Z P_c\

4526: \chi_2(s)\ ds\right\ra\nn\\

4527: \ \ &\sim&\  -\ 

4528:  \left\la \chi_1\ ,\int^t_0 e^{i\cH_0 (t-s)}{\tilde \cH}_0

4529:   (\sigma_3E_0+I\Omega_1)^{-1} e^{i\Omega_1s} 

4530:   P_c\

4531: \chi_2(s)\ ds\right\ra

4532: \label{eq:int-by-pts}

4533: \ea

4534: In the previous string of equations we have used the notation $f\sim g$ to

4535: mean equality up to terms which are local in time. 

4536:  Note that $\sigma_3E_0+I\Omega_1$ is  invertible since its determinant

4537: is $\Omega_1^2+E_0^2$. We can therefore carry out this procedure any 

4538: finite number of times to arrange, up to local in time terms, an expression

4539: involves an operator of the form $\chi\exp(i\cH_0 (t-s)){\tilde\cH_0^k}\chi, 

4540: \  k\ge1$,

4541: where $\chi$ is spatially localized. Therefore, the enhanced 

4542:  local decay estimate 

4543: (\ref{eq:H0kldest}) of Theorem \ref{theo:local_decay} applies. We shall use

4544: these observations, together with the detailed dependence of $\chi_2(s)$ on  

4545: on $\alpha_0,\beta_1$ etc. to control certain terms in $R_j^{(i)}$. 

4546: \bigskip

4547: 

4548: \noindent{\bf Part 1: Estimation of 

4549: $\left|R_0^{(0)}[\eta_0]\right|$ and $\left|R_0^{(1)}[\eta_0]\right|$}

4550: 

4551: \begin{prop}\label{prop:biR0i}

4552: Assume either $t\le t_0$ or {\bf Montonicity Property  Q} on 

4553:  $[t_0,t_0+\delta_*]$. Then, 

4554: \ba

4555: \left|R_0^{(0)}[\eta_0]\right|,\ \left|R_0^{(1)}[\eta_0] \right| 

4556:  &\leq& b_i\left(t_0, [\eta_0]_X\right) \langle t

4557: \rangle^{-2} + \cO(\cE_0) 2\Gamma P_0P_1^2

4558: \label{eq:R0i}\\

4559: b_0\left(t_0,[\eta_0]_X\right) &=& \cO( \la t_0\ra^{-1}[\eta_0]_X )\nn\\

4560: b_1\left(t_0,[\eta_0]_X\right) &=& \cO( \la t_0\ra^{-{1\over2}} 

4561:  [\eta_0]_X^{2/3} )

4562: \nn\ea

4563: \end{prop}

4564: %\nit (\bXYZ: do we want $b_i \sim \cO(t_0\wedge t)\dots $?

4565: \medskip

4566: 

4567: \nit {\bf Proof of Proposition \ref{prop:biR0i}:}

4568: %Consider the $P_0$ equation, (\ref{eq:P0eqn}) {\bXYZ better eqn ref!}. 

4569: \medskip

4570: 

4571: \nit \underline{\it Proof of esimate (\ref{eq:R0i}) for $R_0^{(0)}[\eta_0]$:}\ 

4572:  The key terms  are those in the  $P_0$ equation, which are decaying most slowly 

4573:  with $t$. These are linear in $\eta_0$, since 

4574: $\|\eta_0\|_\infty = \cO(t^{-3/2})$.  We focus on the most difficult terms.

4575: These are {\it nonlocal} in $\eta_0(t)$.

4576: %

4577: %

4578: Recall equation (\ref{eq:alpha0_eqn2}) for $\alpha_0$ and  that

4579: the equation for $P_0$, (\ref{eq:P0eqn}), 

4580: is derived from the $\talpha_0$ equation

4581:  (related to (\ref{eq:alpha0_eqn2}) by a near-identity change of variables)

4582: by multiplication by $\talpha_0$ and taking the imaginary part of the

4583: $\partial_t\alpha_0$ equation. 

4584: 

4585: We consider the following representative ``most problematic'' terms in

4586:  $R_0^{(0)}[\eta_0]$, 

4587:  whose estimation

4588: introduces the necessary methods for treating them all:

4589: \ba

4590: &&(T1)\ 

4591:  \cO(\talpha_0) \left\la\chi,\int^t_0 e^{-i\cH_0(t-s)} P_c\Psi_0(s) \Psi_1(s)

4592: {\ol \eta_0}(s) ds \right\ra

4593: \cO\left(\ol{\talpha_0}\tbeta_1e^{-i\lambda_+t}+|\tbeta_1 |^2\right),

4594: \nn\\

4595: &&\label{eq:bi1}\\

4596: && (\ {\rm also\ with}\ \Psi_0\Psi_1{\ol\eta_0}\ {\rm replaced\ by}\

4597:   \Psi_0{\ol\Psi_1}\eta_0,

4598: \  {\ol\Psi_0}\Psi_1\eta_0\ )\nn\\

4599: &&(T2)\ 

4600: \cO(\talpha_0) \cO(\talpha_0\tbeta_1) \left\la

4601:  \chi, \int^t_0 e^{-i\cH_0(t-s)}

4602: P_c\tetab^2\eta_0 \right\ra ds.

4603: \label{eq:T2}

4604: \ea

4605: 

4606: %

4607: %Here,  $\chi\simeq\psi_{0*}$ is spatially localized.

4608: %

4609: 

4610: \noindent\underline{\it Estimation of T1:}

4611:  Since the time-integral is bounded by 

4612: $\cO(\cE_0) \la t\ra^{-3/2}\ [\eta_0]_X$, 

4613: %\Xeta0 = [\eta_0]_X   

4614: by the Cauchy-Schwarz inequality the second term in

4615:  (\ref{eq:bi1}) is bounded as follows:

4616: \be

4617: \bigg|\ \cO(\talpha_0) \cO(|\tbeta_1|^2)\left\la\chi,\int^t_0\cdots\right\ra

4618: \bigg|

4619: \le \cO(\cE_0) P_0 P_1^2 + \cO(\cE_0) \la t\ra^{-3} \|\eta_0\|^2_X.

4620: \nn\ee

4621: 

4622: We now  control the first term in (\ref{eq:bi1}). 

4623:  $\cO(\bar{\talpha_0}\tbeta_1)\cO(\talpha_0)$.

4624: We argue that the key contribution from this term which must be bounded 

4625: is of the form:

4626: \be

4627: \cO(\talpha_0) \bar{\talpha_0}\tbeta_1 e^{-i\lambda_+t}

4628:  \frac{\partial}{\partial t}

4629: \left\la \chi, \int^t_0 e^{-i\cH_0(t-s)}P_c \Psi_0(s)\Psi_1(s)

4630:  {\ol\eta_0(s)}ds\right\ra

4631: \nn\ee

4632: To see this, consider the term in the $\ta0$ equation which corresponds

4633: to the first term in (\ref{eq:bi1}):

4634: \be

4635:  \left\la \chi,\int^t_0 e^{-i\cH_0(t-s)} P_c\Psi_0(s) \Psi_1(s)

4636: {\ol\eta_0(s)} ds \right\ra

4637: \cO\left(\ol{\talpha_0}\tbeta_1e^{-i\lambda_+t}\right),

4638: \label{eq:a0-bi1}\ee

4639:  Next we integrate with respect to $t$,

4640: and integrate by parts, making using the oscillatory exponential factor. 

4641:  The result

4642: is a boundary term, which can be subsumed in the definition of $\ta0$, by a 

4643: near identity transformation, followed by a time-integral $0$ to $t$.

4644:  The latter contributes  terms to the 

4645:   $P_0$ equation (which has been modified 

4646:   due to the slight redefinition of $\ta0$) of the following type:  

4647: \ba

4648: &&\cO(\talpha_0) \cO(\tb1\D_t\ta0+\ta0\D_t\tb1)\ 

4649: \left\la \chi, \int^t_0 e^{-i\cH_0(t-s)}P_c \Psi_0\Psi_1{\ol\eta_0} ds

4650: \right\ra\label{eq:fst}\\ 

4651: &&\cO(\talpha_0) \cO(\bar{\talpha_0}\tbeta_1)\frac{\partial}{\partial t}

4652: \left\la\chi, \int^t_0 e^{-i\cH_0(t-s)}P_c \Psi_0\Psi_1{\ol\eta_0} ds\right\ra

4653: \label{eq:snd}\ea

4654: Since $\D_t\ta0 = \cO(|\ta0|^2|\tb1|)$ and  $\D_t\tb1=\cO(|\ta0|\ |\tb1|^2)$

4655: (\ref{eq:fst}) is bounded by

4656: \be

4657: \cO(\cE_0)\ \cO\left(\ta0^3\tb1^2\la t\ra^{-\frac{3}{2}}\ [\eta_0]_X\ \right),

4658: \nn\ee

4659:  where we have used

4660: \be\|\eta_0\|_\infty \leq C[\eta_0]_X\ \langle t \rangle^{-3/2}.\nn\ee

4661: By the Cauchy-Schwarz inequality this is bounded by

4662:  $\cO(\cE_0) (2\Gamma P_0

4663: P_1^2+ [\eta_0]_X^2 \langle t\rangle^{-3})$. Therefore, the contribution 

4664: from this first term satisfies the estimate (\ref{eq:R0i}).

4665: 

4666: Obtaining a bound on (\ref{eq:snd}) is more involved. 

4667: We use the local decay estimate of Theorem \ref{theo:local_decay}:

4668: \be \|\chi e^{-i\cH_0t} P_c{\tilde\cH_0}f\|_2\le\la t\ra^{-5/2}

4669:  \|\D^k\la x\ra^\sigma f\|_2.

4670: \nn\ee

4671: The expression (\ref{eq:snd}) bounded by: 

4672: \ba

4673: &=&\ \cO(\talpha_0) \cO(\bar{\talpha_0}\tbeta_1) \left( \la \chi,

4674: \Psi_0\Psi_1\eta_0\ra -i \la \chi, \int^t_0 e^{-i\cH_0(t-s)}

4675: P_c{\tilde\cH_0}\Psi_0\Psi_1{\ol\eta_0} ds \ra\right)\nn\\

4676: &\le&\ \cO(\cE_0)\left(\ \Gamma Q_0Q_1^2\ +\ \la t\ra^{-3}\ \right)\nn\\ 

4677: &&\ \ +\ \cO(\talpha_0^2 \tbeta_1)\ [\eta_0]_X\  

4678: \int^t_0\frac{ds}{\langle t-s\rangle^{5/2}} \frac{1}{\la s\ra^{3\over2}}

4679: |\talpha_0| |\tbeta_1 | .

4680: \label{eq:bi2}

4681: \ea

4682:  The latter integral requires detailed

4683: estimation on different time scales. 

4684: 

4685: To estimate the last integral, we split the range of integration

4686: into three regions:

4687: \ba I_0 \equiv \{s: 0\le s\le t_0\}\nn\\

4688:     I_1 \equiv \{s: t_0 <s\le t_1\}\nn\\ 

4689:     I_2 \equiv \{s: t_1 <s\le  T\}\label{eq:intervals}

4690: \ea

4691: \nit{\bf Estimate on $I_0$:}

4692:  Assume $t\le t_0$. Recall that by (\ref{eq:t0def}),  $|\talpha_0(s)|\le

4693:  \cO(\cE_0)\la s\ra^{-1}$.

4694: Using this we have

4695: \ba

4696: &&\cO(\talpha_0^2\tbeta_1)\ [\eta_0]_X\  

4697: \int_0^t \frac{1}{\langle t-s\rangle^{5/2}} \frac{1}{\la s\ra^{3\over2}}

4698: |\talpha_0|\ |\tbeta_1|\ ds\nn\\

4699: &=&\ \cO(|\ta0|^{3\over2}) \left(4\Gamma\talpha_0^{1\over2}\tbeta_1\right)

4700: \ \left(\frac{[\eta_0]_X}{4\Gamma}\ 

4701:  \int_0^t\ \frac{1}{\langle t-s\rangle^{5/2}} \frac{1}{\la s\ra^{3\over2}}

4702:  |\talpha_0|\tbeta_1|\ ds\right)

4703:  \nn\\

4704: &\le& \cO(\cE_0^{3\over4}) \left[ 2\Gamma P_0 P_1^2 + [\eta_0]_X^{4/3}

4705: \la t\ra^{-\frac{5}{2} \cdot \frac{4}{3}}\right],

4706: \nn\ea

4707: where we have used that $ab\le a^4/4 + 3b^{4\over3}/4$.

4708: \bigskip

4709: 

4710: \nit{\bf Estimate on $I_1$:}

4711: Let $t$ be such that $t_0\le t\le t_1$.

4712: We break the integral into an integral over $[0,t_0]$ plus an integral over

4713:  $[t_0,t]$.

4714:  Recall the definitions of $Q_j(t),\ j=0,1$ in terms of $|{\tilde\alpha}_j(t)$

4715:  displayed in (\ref{eq:Q0Q1def}).

4716:  Using that

4717: \begin{itemize}

4718: \item for $s\in I_0$, $|\ta0(s)|\le

4719:  \cO(\cE_0)\la s\ra^{-1}$ 

4720: and for

4721: \item $s\in I_1 $,

4722:  $Q_0$ is increasing

4723:  and $Q_0(s)\le \cE_0^r Q_1(s)$, 

4724: \end{itemize}

4725: we have 

4726: \ba

4727: &&\cO(|\ta0(t)|^2|\tb1(t)|) [\eta_0]_X\

4728: \left(\ \int_0^{t_0}+\int_{t_0}^t\ \right)

4729: \frac{1}{\la t-s\ra^{5/2}\la s\ra^{3/2}}

4730:  |\tb1(s)|\ |\ta0(s)|\ ds\nn\\

4731: &\le&\

4732: I_0\ {\rm type\ bound}\ +\ \cO(|\ta0|^2\ |\tb1|)\ [\eta_0]_X\ 

4733:  \int_{t_0}^t\ 

4734: \frac{|\tb1(s)|\ |\ta0(s)|}{\la t-s\ra^{5/2}\la s\ra^{3/2}}

4735: \ ds\nn\\

4736: &\le&\ I_0\ {\rm type\ bound}\ +\ \cO(|\ta0|^3\ |\tb1|\ \|\tb1\|_\infty)\ 

4737:  [\eta_0]_X\ 

4738:  \int_{t_0}^t\ \frac{1}{\la t-s\ra^{5/2}\la s\ra^{3/2}}

4739:  \ ds\nn\\ 

4740: &=&\ I_0\ {\rm type\ bound}\ +\ \cO\left(\cE_0^{1\over2}|\ta0|\cdot 

4741:  |\ta0|\cdot |\ta0| |\tb1|\right)\ [\eta_0]_X\  \la t\ra^{-{3\over2}} 

4742:  \ ds\nn\\

4743: &=&\ I_0\ {\rm type\ bound}\ +\ \cO\left(\cE_0\cdot 

4744:  Q_0^{1\over2}\cdot\cE_0^{r\over2}Q_1\right)\ [\eta_0]_X\ 

4745:  \la t\ra^{-{3\over2}} 

4746:  \nn\\

4747: &=&\ I_0\ {\rm type\ bound}\ +\

4748:  \cO\left(\cE_0^{{r+2\over2}}\right)\

4749:   Q_0^{1\over2}Q_1\ \la t\ra^{-{3\over2}}\nn\\

4750: &=&\ I_0\ {\rm type\ bound}\ +\

4751:  \cO\left(\cE_0^{{r+2\over2}}\right)\left(\ Q_0Q_1^2\ +\ \la t\ra^{-3}\ \right),

4752: \ea

4753: which is a bound of the type in (\ref{eq:R0i}).

4754: 

4755: 

4756: \nit{\bf Estimate on $I_2$:}

4757:  

4758: \ba

4759: && \cO(\talpha_0^2 \tbeta_1(t))

4760:  [\eta_0]_X \int^t_0

4761: \frac{|\talpha_0| | \tbeta_1 |}{\la  t-s\ra^{5/2} \la s\ra^{3/2}}\ ds

4762: \nn\\

4763: &=& \cO(\talpha_0^2\tbeta_1 (t) [\eta_0]_X) \left(\int^{t_0}_0 +

4764: \int^{t_1}_{t_0} + \int^t_{t_1}\right)\ 

4765: \frac{|\talpha_0| | \tbeta_1 |}{\la

4766: t-s\ra^{5/2} \la s\ra^{3/2}}\ ds\nn\\

4767: &=& I_0\ \&\ I_1\ {\rm type\ bounds}\ +\ 

4768: \cO(\talpha_0^2 \tbeta_1(t))

4769:  [\eta_0]_X\ \int^t_{t_1}

4770:  \frac{|\talpha_0| | \tbeta_1 |}{\la

4771: t-s\ra^{5/2} \la s\ra^{3/2}}\ ds\nn\\

4772: &&\label{eq:soit}\ea

4773: 

4774: The latter integral must be treated differently from the previous terms.

4775: For $t\in I_1$, we used that $Q_0$ monotonically increasing and 

4776: bounded by a small constant times

4777:  $Q_1$ to treat terms perturbatively. On $I_2$, $Q_0$ dominates $Q_1$

4778:  (which decays) and we must use a different argument. We return to 

4779: expression from which the last term in (\ref{eq:soit}) is derived: 

4780: \be

4781: \cO(\talpha_0) \cO(\bar{\talpha_0}\tbeta_1) \left( \la \chi,

4782: \Psi_0\Psi_1\eta_0\ra -i \la \chi, \int^t_0 e^{-i\cH_0(t-s)}

4783: P_c{\tilde\cH_0}\Psi_0\Psi_1{\ol\eta_0} ds \ra\right).

4784: \label{eq:soit1}

4785: \ee

4786: We need to expand and estimate time integral:

4787: \be

4788: \int_{t_1}^t e^{-i\cH_0(t-s)} P_c{\tilde\cH_0}

4789: \ta0(s)\tb1(s)e^{-i\lambda_+s}{\ol\eta_0(s)} ds\

4790:  =\ e^{-i\cH_0t}\ \int^t_{t_1} e^{i(\cH_0-\lambda_+)s} P_c\cH_0 

4791: \ta0(s)\tb1(s){\ol\eta_0(s)} ds\nn\ee

4792: which we do using integration by parts. We carry this out, 

4793:  then take the inner 

4794:  product of the result with a localized function, $\chi$, and 

4795: then finally multiply by 

4796: $\cO(|\ta0|^2\ |\tb1|)$. The result is of order

4797: \ba

4798: && |\ta0|^3\ |\tb1|^2\ \|\eta_0(t)\|_\infty\ +\ 

4799:  |\ta0|^3\ |\tb1|\ \left|\left\la\chi,e^{-i\cH_0t}

4800:  {\tilde\cH_0}\chi\right\ra\right|

4801: \nn\\

4802: &&+\ \cO(|\ta0|^2\ |\tb1|)\ \int_{t_1}^t e^{-i\cH_0(t-s)}\ P_c  

4803:  (\cH_0-\lambda_+)^{-1}{\tilde\cH_0}\chi e^{-i\lambda_+s}

4804: \frac{d}{ds}\left( \ta0(s)\tb1(s){\ol\eta_0(s)}\right)\ ds

4805: \nn\\

4806: \label{eq:soit2}

4807: \ea

4808: The first two terms in (\ref{eq:soit2}) are bounded as follows:

4809: \ba

4810: &&|\ta0|^3\ |\tb1|^2\ \|\eta_0(t)\|_\infty\ = \cO(\cE_0)\ (|\ta0|\ |\tb1|^2)\ 

4811: \|\eta_0(t)\|_\infty\nn\\

4812: && \le\ \cO(\cE_0)\left(\ Q_0Q_1^2\ +\ \|\eta_0(t)\|_\infty^2\

4813:  \right)\nn\\

4814: && \le\ \cO(\cE_0)\left(\ Q_0Q_1^2\ +\ \la t\ra^{-3}\ \right)

4815: \nn\ea

4816: \ba

4817: && |\ta0|^3\ |\tb1|\ \left|\left\la\chi,e^{-i\cH_0t}{\tilde\cH_0}

4818:  \chi\right\ra\right|

4819: \le\ C|\ta0|^{5/2}\ (|\ta0|^{1/2}\ |\tb1|)\ \la t\ra^{-5/2}\nn\\ 

4820: &&\le \cO(\cE_0^{5/2})\left( Q_0Q_1^2\ +\ \la t\ra^{-{5\over2}\cdot{4\over3}}\

4821:   \right)

4822: \ \le\ \cO(\cE_0^{5/2})\left( Q_0Q_1^2\ +\ \la t\ra^{-3}\ \right).

4823: \ea

4824: 

4825: We now turn to the nonlocal term, 

4826:   in (\ref{eq:soit2}),  which we denote $\cI_1$:

4827: \ba

4828: \cI_1\ \sim\  \cO(|\ta0|^2\ |\tb1|)\  &&\int_{t_1}^t e^{-i\cH_0(t-s)}\ P_c

4829:  (\cH_0-\lambda_+)^{-1}{\tilde\cH_0}\chi e^{-i\lambda_+s}\nn\\

4830:    && \left(\ \D_s\ta0(s)\tb1(s){\ol\eta_0(s)}+

4831:                    \ta0(s)\D_s\tb1(s){\ol\eta_0(s)}+

4832:                   \ta0(s)\tb1(s)\D_s{\ol\eta_0(s)}\ \right).\nn\\

4833: &&\label{eq:int1120}

4834: \ea

4835: Using that

4836: \be \D_t\ta0\sim\cO(\tb1^2\ta0),\ \D_t\tb1\sim\cO(\ta0\tb1^2),\

4837:  {\rm and}\ \D_t\eta_0\ =\ -i\cH_0\eta_0,\label{eq:Dt-ests}

4838: \ee

4839: we have

4840: \ba

4841: \cI_1\ \sim\  \cO(|\ta0|^2\ |\tb1|)\ &&\int_{t_1}^t e^{-i\cH_0(t-s)}\ P_c

4842:  (\cH_0-\lambda_+)^{-1}{\tilde\cH_0}\chi e^{-i\lambda_+s}\nn\\

4843:    &&\ \ \left(\ \ta0\tb1^3{\ol\eta_0}+

4844:                    \ta0^2\tb1^2{\ol\eta_0(s)}+

4845:                   \ta0(s)\tb1(s){\tilde\cH_0}{\ol\eta_0(s)}\ \right)\nn\\

4846: &=&\  \cI_{1a}\ +\ \cI_{1b}\ +\ \cI_{1c}

4847: \label{eq:I1abc}

4848: \ea

4849: Each of the three terms $\cI_{1j},\ j=a,b,c$ satisfies a bound of the form:

4850: \be

4851: \left|\cI_{1j}\right|\ \le\ C\ \cE_0^2\ [\eta_0(0)]_X\

4852:                         \frac{1}{\la t_0\ra}\frac{1}{\la t-t_1\ra^2}

4853: \label{eq:Ij-bound}

4854: \ee

4855: 

4856: We illustrate this by estimating $\cI_{1a}$; the other two terms

4857: are estimated similarly.

4858: 

4859: Using that $|\ta0|,\ |\tb1|\ =\ \cO(\cE_0^{1\over2})$ we have  

4860: \ba

4861: \left|\cI_{1a}\right|\ &\le&\ |\ta0|^2\ |\tb1|\

4862:   \int_{t_1}^t \frac{1}{\la t-s\ra^{5\over2}}\

4863:                |\tb1|^3\ |\ta0|\ |\eta_0|\ ds\nn\\

4864: &\le&\ \cE_0^2\ [\eta_0(0)]_X\ 

4865:  \left(\sup_{t\ge t_1} \la t-t_1\ra^{1\over2}|\tb1(t)|\right)^3

4866: \int_{t_1}^t \frac{1}{\la t-s\ra^{5\over2}}

4867:  \frac{1}{\la s-t_1\ra^{3\over2}}\frac{1}{\la s\ra^{3\over2}}

4868: \nn

4869: \ea

4870: Separate estimation of the contributions from the intervals

4871:  $[t_1,{1\over2}(t+t_1)]$ and $[{1\over2}(t+t_1),t]$ yields the bound:

4872: \be

4873: \left|\cI_{1a}\right|\ \le\ C\ \cE_0^2\ [\eta_0(0)]_X\ 

4874:    \left(\sup_{t\ge t_1} \la t-t_1\ra^{1\over2}|\tb1(t)|\right)^3                     \frac{1}{\la t_0\ra}\frac{1}{\la t-t_1\ra^2}

4875: \label{eq:I1a-bound}

4876: \ee

4877: 

4878:                    

4879: 

4880: 

4881: %%%%%%%%%%%%%%%%%%%%%%%%%

4882: %\nit{\bf Replace with new notes\ [\ }

4883: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4884: 

4885: \noindent\underline{\it Estimation of T2:}

4886: Consider the term

4887: \be

4888: {\rm T2}\ =\ \cO(\talpha_0(t)) \cO(\talpha_0(t)\tbeta_1(t)) \left\la 

4889:  \chi, \int^t_0 e^{-i\cH_0(t-s)}

4890: P_c\tetab^2(s)\eta_0(s) \right\ra\ ds.

4891: \label{eq:T2-def}

4892: \ee

4893: \medskip

4894: 

4895: \nit \underline{\it $t\in I_1\ \equiv\ [t_0,t_1] $}:\ 

4896: For $t\in I_1$,  

4897: \be

4898: |\talpha_0|^2\ |\tbeta_1| \leq\ c\cE_0^{r\over2} |\talpha_0|\ |\tbeta_1|^2;\

4899:  \ {\rm  Proposition\  \ref{prop:propC}}.

4900: \nn\ee

4901: Therefore,

4902: \ba

4903: |{\rm T2}|\ &\le&\ c\cE_0^{r\over2}\ |\talpha_0|\ |\tbeta_1|^2 

4904:  \left|\left\la\ \chi, 

4905:  \int^t_0 e^{-i\cH_0(t-s)}P_c \tetab^2 \eta_0 ds\ \right\ra\right|\nn\\

4906: &\le& 

4907:  c\cE_0^{r\over2}\ |\talpha_0|\ |\tbeta_1|^2 

4908:  \la t\ra^{-3/2} \left(\sup_{0\le s\leq t}\la s\ra^{3\over2}

4909:   \| \tetab^2 (s) \eta_0(s) \|_{W^{k,2}\cap L^1} \right)\nn\\

4910: &\le& c\cE_0^{r\over2}\ |\talpha_0|\ |\tbeta_1|^2\ \la t\ra^{-3/2}

4911:   \|\tetab\|^2_{W^{k,2}} [\eta_0]_X \nn\\

4912: &\le&\ c

4913: \cE_0^\rho [\eta_0]_X\ |\talpha_0|\ |\tbeta_1|^2\  \la t \ra^{-3/2}\nn\\

4914: &\le&\ c\cE_0^\rho [\eta_0]_X\left(\ P_0P_1^2\ +\ \la t \ra^{-3}\ \right)

4915: \nn\ea

4916: \bigskip

4917: 

4918: \nit \underline{$t\in I_2\equiv [t_1,\infty)$:}

4919: 

4920: \nit To see the relevant terms for $t> t_1$, we integrate by parts

4921: and obtain, besides easily estimable local terms and

4922:  terms with faster time decay, 

4923: \be

4924: O(\ta0^2\tb1) \left\la\ \chi, \int^t_0 e^{-i\cH_0(t-s)}

4925:   {\tilde\cH_0}P_c\tetab^2\eta_0\ ds\ \right\ra  \quad t> t_1.

4926: \label{eq:T2term}

4927: \ee

4928: Consider the contribution to the integral 

4929: in (\ref{eq:T2term}) coming from  

4930: $s\in[0,t_0]$: 

4931: \be

4932: \cO\left(\talpha_0^2(t)\beta_1(t)\right) 

4933: \left\la\chi, \int_0^{t_0} e^{-i\cH_0(t-s)}

4934:   {\tilde\cH_0}P_c\tetab^2\eta_0\ ds\ \right\ra \quad t> t_1>t_0

4935: \label{eq:T2terma}

4936: \ee

4937: Consider, the inner produce

4938: \be

4939: R(t)\ =\ \left\la \chi,\int_0^{t_0} e^{-i\cH_0(t-s)}

4940:   {\tilde\cH_0}P_c\tetab^2\eta_0\ ds\  \right\ra 

4941: \label{eq:Rdef}

4942: \ee 

4943: We have 

4944: \be \left|{\rm (\ref{eq:T2terma})}\right|\ \le\ 

4945:  C\ |\ta0(t)|^2 |\tb1(t)|\ |R(t)|\  \sim\ CQ_0(t) Q_1^{1\over2}(t) |R(t)|

4946: \label{eq:T2termz}

4947: \ee

4948: and therefore it suffices to prove that 

4949: \be |R(t)|\ \le\ CQ_1^{3\over2}(t)\label{eq:Rbound}\ee  

4950: To prove (\ref{eq:Rbound}), recall that for $t\ge t_1$ y 

4951: % by (\bXYZ backref)

4952: \ba

4953: \frac{dQ_0}{dt}\ &\asymp&\ \Gamma Q_0Q_1^2\nn\\

4954: \frac{dQ_1}{dt}\ &\asymp&\ -2\Gamma Q_0Q_1^2,\

4955: \label{eq:T2c}

4956: \ea

4957: and

4958: \be \cE_0^rQ_1(t_1)\ =\ Q_0(t_1).

4959: \label{eq:T2d}

4960: \ee

4961: Therefore,

4962: \be

4963: Q_1(t)\ \asymp\  \frac{2Q_1(t_1)}{1\ +\ \Gamma\int_{t_1}^tQ_0(s)ds}

4964: \label{eq:Q1asymp}

4965: \ee

4966: Therefore, to establish (\ref{eq:Rbound}) we need:

4967: \be

4968: |R(t)|\ \le\ C\  

4969: \left(\ \frac{2Q_1(t_1)}{1\ +\ \Gamma\int_{t_1}^tQ_0(s)ds}\ 

4970:      \right)^{3\over2}

4971: \label{eq:Rbound1}

4972: \ee

4973: We consider two cases

4974: 

4975: \nit Case 1: $\Gamma\ \sup_{s\in[t_1,t]}|Q_0(s)|\ |t-t_1|\ \le\ 1$,

4976: and

4977: 

4978: \nit Case 2: $\Gamma\ \sup_{s\in[t_1,t]}|Q_0(s)|\ |t-t_1|\ \ge\ 1$,

4979: where it suffices to prove

4980: \be

4981: |R(t)|\ \le\ \left(\ \frac{3Q_1(t_1)}{ \Gamma\int_{t_1}^tQ_0(s)ds}\ 

4982:      \right)^{3\over2} 

4983: \label{eq:case2bound}

4984: \ee

4985: In case 2 we prove the bound on $R(t)$, (\ref{eq:Rbound}), while 

4986: in case 1 we prove that the expression (\ref{eq:T2term}) of order

4987: $\la t-t_1\ra^{-{3\over2}}\la t-t_0\ra^{-{3\over2}}$.

4988: %\bXYZ. 

4989: \nit We first handle case 2, the bound (\ref{eq:case2bound}).

4990:   From (\ref{eq:T2c}) we have 

4991: \ba

4992: \frac{d}{dt}\left(\ Q_0\ +\ 2Q_1\ \right)\ &\asymp&\ 0,\ t\ge t_1

4993: \nn\\

4994: Q_0(t)\ +\ 2Q_1(t)\ &\asymp&\ Q_0(t_1)\ +\ 2Q_1(t_1)\nn\\

4995:                     &=&\ (2+\cE_0^r)\ Q_1(t_1)

4996: \label{eq:T2e}

4997: \ea

4998: Therefore, since

4999: \be

5000: Q_1(t)\ \asymp\ \frac{2Q_1(t_1)}{1\ +\ \Gamma\int_{t_1}^tQ_0(s)ds}

5001: \label{eq:Q1asym}

5002: \ee

5003: we have, as $t\to\infty$

5004: \ba

5005: Q_0(t)\ &\asymp&\ (2+\cE_0^r)Q_1(t_1)\ -\ 2Q_1(t)\nn\\

5006:         &\longrightarrow&\ (2+\cE_0^r)Q_1(t_1)

5007: \label{eq:T2f}

5008: \ea

5009: Therefore, for $t>t_1$ ($t-t_1$ large enough)

5010: \be

5011: (1+\cE_0^r)Q_1(t_1)\ \le\ Q_0(t)\ \le\ (2+\cE_0^r)Q_1(t_1)

5012: \nn\ee

5013: and therefore

5014: \ba

5015:  \frac{Q_1(t_1)}{1\ +\ \Gamma\int_{t_1}^tQ_0(s)ds}

5016:     &\ge&\ \frac{Q_1(t_1)}{1\ +\ \Gamma(2+\cE_0^r)\ |t-t_1|\ Q_1(t_1)}\nn\\

5017:     &\ge&\ \frac{1}{3}\frac{1}{\Gamma\ |t-t_1|}\ \ge 

5018:  \frac{1}{3}\frac{1}{\Gamma\ |t-t_0|}

5019: \label{eq:Q1lba}

5020: \ea

5021: On the other hand,

5022: \ba

5023: \left| R(t) \right|\ &\le&\

5024:  C\ 

5025:   \left\la \chi,\int_0^{t_0} \frac{1}{\la t-s\ra^{3\over2}}

5026: \| {\tilde\cH_0}\tetab^2(s)\eta_0(s)\|_{L^1}\ ds\ \right\ra \nn\\

5027:  &\le&\ C\ \sup_{\tau\in [0,t_0]}\|\tetab(\tau)\|_{H^2}^2\ 

5028:                                  [\eta_0]_X\ 

5029:       \int_0^{t_0} 

5030:  \frac{1}{\la t-s\ra^{3\over2}\la s\ra^{3\over2}} \ ds\nn\\

5031: &\le&\ C\ \cE_0^2\ [\eta_0]_X\  

5032:        \frac{1}{\la t-t_0\ra^{3\over2}}

5033: \label{eq:T2termb}

5034: \ea 

5035: The bounds (\ref{eq:T2termb}) and (\ref{eq:Q1lba}) imply (\ref{eq:case2bound}).

5036: \medskip

5037: 

5038: We now turn to case 1. In this case, 

5039:  $\Gamma\sup_{s\in [t_1,t]}Q_0(s)\ |t-t_1|\le 1$, 

5040: \be Q_0(t)\ \le\ {1\over\Gamma}\frac{1}{|t-t_1|}

5041: \label{eq:Q0ub}

5042: \ee

5043: and therefore by (\ref{eq:T2term}) and (\ref{eq:Q0ub})

5044: \be

5045: \left| {\rm (\ref{eq:T2term})}\right| 

5046: \le

5047: \cE_0^2[\eta_0]_X\ \sup_{s\in[t_1,t]}\left(\la s-t_1\ra^{1\over2}Q_1(s)\right)

5048:   \frac{1}{|t-t_1|}\frac{1}{\la t-t_1\ra^{1\over2}} 

5049:   \frac{1}{\la t-t_0\ra^{3\over2}}

5050: \label{eq:T2termc}

5051: \ee

5052: 

5053: %\footnote{

5054: We begin by noting that by Proposition \ref{prop:disip}

5055: and its corollaries, 

5056: %and the following Proposition \ref{prop:xeta},

5057:  for $p>6$

5058: \ba \|\tetab\|_{W^{k,p}}&\le&\cO(\cE_0) \la t\ra^{-1},\ \  t\leq t_0\nn\\

5059:  \|\tetab \|_{W^{k,p}} &\le& \cO(\cE_0^\rho) Q_0 (t)^{1/2},\ \    

5060:   t_0<t\le t_1\nn\\

5061: \|\tetab\|_\infty &=& \cO(\la t-t_1\ra^{-{1\over2}}), t\ge t_1

5062: \label{eq:tetabWkp}\ea

5063: 

5064: For $t< t_1$, the previous arguments with the known estimates on

5065: $\tetab$, and the facts that $|\talpha_0(t)| \leq \cO(\cE_0)/\la t\ra$

5066: on $I_0$ and 

5067:  $Q_0(t) \leq \cE_0^rQ_1$  on $I_1$

5068: imply the necessary bounds.

5069: %}

5070: %\footnote{For $t> t_1$, we need the following {\bXYZ ???}

5071: %

5072: %\begin{prop}\label{prop:xeta}

5073: %For $t\leq t_0$ and $\sigma<\frac{3}{2}$

5074: %$$

5075: %\| \la x\ra^\sigma \tetab(t) \|_\infty \leq \cO(\cE_0).

5076: %$$

5077: %\end{prop}

5078: %\nit{\bf Proof of Proposition \ref{prop:xeta}:}

5079: % The source term for $\tetab$ is 

5080: %$$

5081: %S(s) = O (\talpha_0\tbeta_1^m \chi).

5082: %$$

5083: %Hence, for $t\leq t_0$

5084: %$$

5085: %\| \la x\ra^\sigma S(s) \|_{H^k} \leq 

5086: %\cO(\cE_0)\frac{1}{\langle s \rangle}.

5087: %$$

5088: %

5089: %We have

5090: %\be

5091: %\la x\ra^\sigma\eta_b(t)\ =\ 

5092: % -i\ \int_0^t\ \la x\ra^\sigma\ e^{-iH_0(t-s)}\ P_c\ S(s)\ ds

5093: %\ +\ \int_0^t\ \la x\ra^\sigma\ e^{-iH_0(t-s)}\ P_c\ \eta_b^3(s)\ ds

5094: %\nn\ee

5095: %Taking the $L^\infty$ norm, we get

5096: %\ba

5097: %\|\la x\ra^\sigma\eta_b(t)\|_\infty\ &\le&\ 

5098: %C\int_0^t\ \frac{ds}{\la t-s\ra^{\frac{3}{2}-\sigma} }\cO(\cE_0)

5099: % \frac{1}{\la s\ra}\ +\ 

5100: % C\int_0^t\ \frac{ds}{\la t-s\ra^{\frac{3}{2}-\sigma} }\cO(\cE_0)

5101: % \frac{1}{\la s\ra}\  \|\eta_b(s)\|_X\nn\\

5102: % &=&\ \cO(\cE_0),\ {\rm for}\ \sigma<\frac{3}{2}

5103: %\nn\ea

5104: %Here, we used that \bXYZ

5105: %\be

5106: %\la x\ra^\sigma e^{-iHt}P_c\ =\ e^{-iHt}\la x\ra^\sigma\ 

5107: %    +\ \cO(t^\sigma)

5108: % e^{-iHt},

5109: %\nn\ee

5110: %smoothness and $\cO(s^{-1})$ decay of $\eta_b(s)$ for $s\le t_0$; 

5111: %see estimates above statement of Proposition \ref{prop:xeta}. 

5112: %}

5113: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

5114: %%%%%%%%% oldstuff 

5115: %\footnote{

5116: %We now turn to the second integral in (\ref{eq:soit}). 

5117: %We consider two cases:

5118: %\ba 2\Gamma|t-t_1|Q_0(t_1)Q_1(t_1)&\le& 100\ {\rm and}\nn\\

5119: % 2\Gamma|t-t_1|Q_0(t_1)Q_1(t_1)&\ge& 100\label{eq:cases}

5120: %\ea

5121: %}

5122: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

5123: Collecting all these, we have 

5124: $$

5125: \cO (\talpha^2_0\tbeta_1) \bigg|\langle \chi, \int^t_0 e^{i\cH_0(t-s)}P_c

5126: {\tilde\cH_0}\tetab^2 \eta_0 ds \rangle \bigg| \leq \cO(\cE_0) \left[2\Gamma

5127: Q_0 Q_1^2 + \langle t_0\rangle^{-1} \langle t \rangle^{-2}\right].

5128: $$  

5129: \medskip

5130: 

5131: \nit\underline{Proof of estimate (\ref{eq:R0i}) for  $R_0^{(1)}$:}

5132: The key terms to consider in the  $P_1$ equation are the slowest

5133: decaying nonlocal terms; see (\ref{eq:beta_eqn3}) and

5134:  (\ref{eq:eta1_eqn}).

5135: Since $b_1$ enters as $b_1^2$ in 

5136: %$k_0> 3(b_0 + c b_1^2)$,

5137:  (\ref{eq:k0_def}), we

5138: need to bound $R_0^{(1)}$ by $\cO(\cE_0)\left[\langle

5139: t_0\rangle^{-1/2} \langle t \rangle^{-2} + 2\Gamma P_0 P_1^2\right]$

5140: for $t> t_0$.

5141: 

5142: The slowest decaying nonlocal term in the $\beta_1$ equation arises from

5143: the balance:

5144: \be

5145: i\D_t\beta_1\ \sim\ 

5146: \lambda\la\psi_{1*}^3,\pi_1\Phi_2\ra |\beta_1|^2e^{i\lambda_+(T)t} 

5147: \nn\ee

5148: Since the $P_1$ equation is obtain by multiplying by $\ol{\beta_1}$ and 

5149: taking the imaginary part, we must estimate:

5150: \be

5151: \lambda\la\psi_{1*}^3,\pi_1\Phi_2\ra\ |\beta_1|^2\ol{\beta_1}\

5152:   e^{i\lambda_+(T)t},

5153: \nn\ee

5154: the leading order nonlocal part of which is

5155: \ba

5156: &\sim& |\tbeta_1|^2 \ol{\tbeta_1} \left\la \chi, \int^t_0 e^{-i\cH_0(t-s)}

5157: P_c\Psi_0\Psi_1 \eta_0 ds \right\ra e^{i\lambda_+ t}\nn\\

5158: &\sim& |\tbeta_1|^2 \ol{\tbeta_1} \left\la \chi, \int^t_0 e^{-i\cH_0(t-s)} P_c

5159: \talpha_0 \tbeta_1 e^{-i\lambda_+s} \chi \eta_0 (s) ds\right\ra

5160: e^{i\lambda_+ t}

5161: \nn\ea

5162: Integrating by parts (using the oscillatory factor $e^{i\lambda_+t}$) and 

5163:  removing nonresonant local in terms  by near identity transformations the 

5164: key term is

5165: \be 

5166:  |\tbeta_1|^2 \ol{\beta_1} 

5167: \left\la \chi, \int^t_0 e^{-i\cH_0(t-s)}{\tilde\cH_0}

5168: P_c \talpha_0\tbeta_1 \chi e^{i\lambda_+ s} \eta_0 (s) ds\right\ra

5169: e^{i\lambda_+ t}.\nn

5170: \ee

5171: 

5172: Suppose $t> t_1$.  Since we have no factor of $\talpha_0$ outside the

5173: integral (local in time $\alpha_0(t)$ factor), 

5174:  the estimate for $I_0=\{0 < s < t_0\}$ is

5175: the critical part and requires the $\cO(\la t\ra^{-1})$ bound on $I_0$ 

5176:  and Theorem \ref{theo:local_decay}:

5177: \ba

5178: &&\left|\left\la \chi, \int^t_0 e^{-i\cH_0(t-s)}{\tilde\cH_0} 

5179:  P_c \talpha_0 \tbeta_1 \chi

5180: \eta_0 e^{-i\lambda_+ s} ds \right\ra \right|\nn\\

5181: &\leq& C[\eta_0]_X 

5182:  \int^{t_0}_0 \frac{ds}{\la t-s\ra^{5/2}\la s \ra^{3/2}}|\talpha_0(s) |

5183: |\tbeta_1(s)|\  + \left(\ \int^{t_1}_{t_0} +

5184: \int^t_{t_1}\ \right)\cdots\ ds\nn\\

5185: &\leq& \cO(\cE_0) [\eta_0]_X

5186:  \int^{t_0}_0 

5187:   \frac{ds}{\la t-s\ra^{5/2}\la s \ra \la s\ra^{3/2}} 

5188: \ + [\eta_0]_X\ \int^{t_1}_{t_0}

5189:  \frac{ds}{\la t- s\ra^{5/2}\la s \ra^{3/2} }

5190: |\talpha_0(t_1)|\ |\tbeta_1(s)|  

5191:  + \int_{t_1}^t\cdots\ ds\nn\\

5192: &\leq& \cO(\cE_0) \left[\la t\ra^{-5/2} [\eta_0]_X +

5193:  \cO(\cE_0)[\eta_0]_X  |\talpha_0 (t) | \la t\ra^{-3/2} +

5194: [\eta_0]_X |\talpha_0(t)|\  

5195:  \la t-t_1\ra^{-{1\over2}}\la t\ra^{-{3\over2}} \right].

5196: \nn\ea

5197: Note that we can extract the factor $|\talpha(t)|$ from the integral for 

5198:  $s\ge t_0$ since in this range $Q_0(t)$ is monotonically increasing and 

5199:  $Q_0(t)\sim P_0(t)\sim |\talpha_0(t)|$ with correction terms which are 

5200:  which are rapidly decaying in time and which are 

5201:  therefore dominated by the

5202: first term. Multiplication by $\cO(|\tbeta_1|^3)$ prefactor gives the 

5203: bound

5204: \be

5205: C(\cE_0,[\eta_0]_X)\left[\

5206:  \left(\sup_{s\ge t_1}\la s-t_1\ra^{1\over2}|\tbeta_1(s)|\right)^3

5207:     \la t\ra^{-5/2}\la t-t_1\ra^{-3/2}\ +\ Q_0Q_1^2\ \right]

5208: \nn\ee

5209: %{\bf XYZ ???}

5210: This completes the proof of Proposition \ref{prop:biR0i}.

5211: \bigskip

5212: 

5213: \section{Local and nonlocal ODE terms: $R_2^{(j)}(P_0,P_1)$ of Proposition

5214:  \ref{prop:Master}}\label{sec:R120P0P1}

5215: 

5216: In this section we prove estimates on the terms $R_2^{(0)}(P_0,P_1)$ 

5217: and $R_2^{(1)}(P_0,P_1)$ of Proposition \ref{prop:Master}. 

5218: 

5219: \begin{prop}\label{prop:ODE}

5220: Assume either $t\le t_0$ or Monotonicity Property {\bf Q} on 

5221:  $[t_0,t_0+\delta_*]$. Then,

5222: For $m\ge4$,

5223: \ba

5224: &&(1)\ |R_2^{(0)}(P_0,P_1)|\le \cO(\cE_0^\rho)\Gamma P_0P_1^2 + 

5225:  \cO(\la t\ra^{-3})

5226: \  (P_0, P_1 << 1 )\nn\\

5227: &&(2)\ |R_2^{(1)}(P_0,P_1)|\le \cO(\cE_0^\rho)\Gamma P_0 P_1^2 + 

5228: \cO(\alpha_0)P_1^{2m} + \cO(\la t\ra^{-3}) , \ \   |\alpha_0|\sim\sqrt{P_0}. 

5229: \nn\ea

5230: \end{prop}

5231: 

5232: \subsection{Proof of part (1) of Proposition \ref{prop:ODE}:}

5233: 

5234: The most problematic terms are nonlocal, slowest

5235: decaying.

5236: The terms which are linear in $\eta$ in the $\alpha_0$

5237: equation contribute the slowest terms; these are nonlocal in time, $t$.

5238: 

5239: We have to consider terms arising in equation (\ref{eq:alpha0_eqn2})

5240:  of the type:

5241: \be

5242: \la\chi, \eta\ra e^{-i\lambda_+ t} \ol{\alpha_0} \beta_1,\   

5243: \la \chi, \eta\ra |\beta_1|^2,\  

5244: \la \chi, \eta\ra \beta_1^2 e^{-2i\lambda_+ t},

5245: \label{eq:typ1}

5246: \ee

5247: where $\phi_2\sim\eta$ and $\eta(t)=\eta_0(t)+e_0(t)+\eta_b(t)$;

5248: see (\ref{eq:eta-decomp}).

5249: 

5250: As calculated earlier, the 

5251:  last term is resonant and its contribution is 

5252: $+2\Gamma P_0 P_1^2$ to the $P_0$ equation; see (\ref{eq:keyterm}).

5253: 

5254: The leading ODE terms in $\eta$ are the source terms of the type

5255: in (\ref{eq:eta1_eqn}):

5256: \be

5257: |\Psi_0|^2\Psi_1, \Psi_0^2 \ol{\Psi_1}, \Psi_0 |\Psi_1|^2 , 

5258: \ol{\Psi_0} \Psi_1^2;

5259: \label{eq:typ2}

5260: \ee

5261: Recall that the $P_0$ equation is obtained from the $\tilde\alpha_0$

5262: equation by multiplication by $\ol{\tilde\alpha_0}$ and taking the  

5263:  imaginary part. 

5264: 

5265: Solving for $\tetab$ and plugging the source term contributions

5266: into the $\langle \chi, \eta\rangle$ terms in the $P_0$ equation

5267: gives, apart from the resonant term, terms of the type

5268: \be

5269: \ol{\alpha_0} \cO (\bar \alpha_0 \beta_1 e^{i\lambda_+ t} + |\beta_1|^2)

5270: \langle \chi, \int^t_0 e^{-i\cH_0(t-s)} P_c \left(\ |\Psi_0|^2 \Psi_1 +

5271: \Psi_0^2 \bar \Psi_1 + \Psi_0|\Psi_1|^2 + \bar \Psi_0 \Psi_1^2\ \right)

5272: ds

5273: \label{eq:typ3}

5274: \ee

5275: For $t>t_1$ the slowest terms are $\cO(|\Psi_0|^2 \Psi_1 + \Psi_0^2

5276: \bar\Psi_1)$ source terms. For $t_0<t<t_1$ the problematic terms are

5277: $\cO(\Psi_0 |\Psi_1|^2 + \bar \Psi_0 \Psi^2_1)$  source terms, since

5278: on this interval $\beta_1$ does not necessarily decay.

5279: Since $\Psi_1 \sim e^{-i\lambda_+ t} \beta_1 + \ {\rm  higher\ order

5280:  \ terms},$

5281: we can integrate by parts the $\cO(|\Psi_0|^2 \Psi_1 + \Psi_0^2 \bar

5282: \Psi_1)$ terms in (\ref{eq:typ3})

5283:  to get arbitrary order $\beta_1^m$ terms, which are nonlocal.

5284:  The remaining terms are local and its lowest order term is of the order

5285: \be

5286: \bar \alpha_0 \cO(\bar\alpha_0 e^{i\lambda_+t}\beta_1 + 

5287:  |\beta_1|^2) \left\la \chi,  \cO

5288: (|\alpha_0|^2 \beta_1 e^{i\lambda_+t} + \alpha_0^2 

5289:  \bar\beta_1 e^{-i\lambda_+t})\right\ra.

5290: \nn\ee

5291: The nonresonant local 

5292:  terms can be transformed to higher order by a near identity

5293: change of variables giving

5294: $$

5295: \cO(|\alpha_0|^3\beta_1^{2m} e^{i\Omega t}),\ \Omega>0\ {\rm and}\ 

5296: \  m\ {\rm arbitrary\  large},

5297: $$

5298: while resonant terms are of the type already derived but of higher order

5299:  (bounded by $\cO(\cE_0)Q_0Q_1^2)$).

5300: We are therefore left to consider the following nonlocal terms 

5301: \be

5302: \bar \alpha_0

5303:   \cO\left( \bar\alpha_0\beta_1 e^{-i\lambda_+ t} + |\beta_1|^2\right) 

5304: \left\la \chi,

5305: \int^t_0 e^{-i\cH_0(t-s)}P_c \alpha_0 \beta_1^{2m} e^{i\Omega_m s}\

5306:   \tilde \chi ds\right\ra

5307: \label{eq:nlc}

5308: \ee

5309: 

5310: Consider the first term in (\ref{eq:nlc}). 

5311:  Due to the oscillatory factor $e^{-i\lambda_+ t}$ in the 

5312:  $\cO(\bar\alpha_0

5313: \beta_1 e^{-i\lambda_+ t})$ term we can, 

5314:  by a near-identity transformation,  

5315: in the $\alpha_0$ equation, remove this term in exchange for one 

5316:  higher order

5317: \ba

5318:  && \cO(|\alpha_0|^2 \beta_1^3) \left\la \chi, 

5319:  \int^t_0\cdots \right\ra + 

5320:  \cO(\bar\alpha_0^2 \beta_1 e^{i\Omega s}) \partial_t \left\la \chi, \int^t_0 e^{-i\cH_0(t-s)}

5321: P_c\ \cdots\ \right\ra\nn\\

5322: && +\ {\rm higher\ order}\  + \ {\rm local\  terms}, 

5323: \nn\ea

5324: and by further near identity transformations, we are left 

5325:  with the following slowest term

5326: \be

5327: \cO(|\alpha_0|^2\beta_1) \partial^m_t

5328: \left\la \chi, \int^t_0 e^{-i\cH_0(t-s)}

5329: \ P_c\cdots\ ds\ \right\ra.

5330: \label{eq:slow}

5331: \ee

5332: All other terms are of order $o(\cE_0)\cO(P_0 P_1^2)$, 

5333:  $o(\cE_0)\la t\ra^{-3}$  

5334:  or higher order, for $\cE_0$ small.

5335: 

5336: \ba

5337: &&\partial_t \langle \chi, \int^t_0 e^{-i\cH_0(t-s)}

5338: P_c\ \alpha_0\beta_1^{2m}\tilde \chi ds\rangle\nn\\

5339: &=& \left\la\chi, \alpha_0(t) \beta_1^{2m}(t) \tilde\chi \right\ra -

5340: \left\la \chi, i \int^t_0 e^{-i\cH_0(t-s)} {\tilde\cH_0}

5341:   P_c \tilde\chi\alpha_0 \beta_1^{2m} e^{i\Omega s}

5342: ds\right\ra.

5343: \nn\ea 

5344: The first term is local and will contribute $o(1) \cO (P_0 P_1^2)$.

5345: It remains to estimate the  nonlocal term.

5346:  Note that 

5347: $$

5348: \int^t_0 e^{-i\cH_0(t-s)} e^{i\omega s} \tilde \chi \alpha_0\beta_1^{2m}ds 

5349:  = e^{-i\bar \omega t}

5350: \int^t_0 e^{-i(\cH_0-\bar\omega)(t-s)} e^{i(\omega+\bar\omega)s}

5351:  \tilde\chi \alpha_0\beta_1^{2m} ds

5352: $$

5353: 

5354: Consider now the second term in  (\ref{eq:nlc}).

5355: Using the oscillatory factor $e^{-i\bar \omega t}$, one can transform  

5356:  the $\cO(\bar\alpha_0|\beta_1|^2) \langle \chi,

5357: \int\cdots \rangle$ to higher order.

5358: 

5359: 

5360: We now turn to the second term in (\ref{eq:nlc}). First, let's

5361: consider the case where $t\ge t_1$.  

5362: \ba

5363: &&\cO(\bar\alpha_0 |\beta_1|^2) \partial^m_t \left\la\ \chi,

5364: \int_0^t e^{-i\cH_0(t-s)} \tilde \chi \alpha_0\beta_1^{2m}\ ds\ 

5365:  \right\ra\label{eq:typ4}

5366: \ea

5367: We consider (\ref{eq:typ4})

5368:   for $t\ge t_1$. For this we require the following

5369: 

5370: \begin{lem}\label{lem:Q1lbound}  

5371: For $t> t_1,$

5372: \be

5373: Q_1(t) \geq Q_1(t_1) \left[ 1+4(\Gamma' + \delta) Q_1 (t_1) Q_0 (t_1)

5374: (t-t_1) \right]^{-1}

5375: \label{eq:Q1lb}

5376: \ee

5377: \end{lem}

5378: {\bf Proof of Lemma \ref{lem:Q1lbound}:}  For $t\ge t_1$ 

5379: \be

5380: \frac{dQ_1}{dt} \geq - 4 (\Gamma' + \delta) Q_0 Q_1^2

5381: \nn\ee

5382: Therefore, 

5383: \be

5384: -\frac{dQ_1^{-1}}{dt} \geq - 4 (\Gamma' + \delta)Q_0(t)

5385:  \ge - 4 (\Gamma' + \delta)Q_0 (t_1)\ {\rm or}\ \

5386:   \frac{dQ_1^{-1}}{dt} \leq 4 (\Gamma' + \delta) Q_0 (t_1)

5387: \nn\ee

5388: since $Q_0 (t_1) \leq Q_0 (t)$ ($Q_0\uparrow$ for $t>t_0$).

5389: Integrating from $t$ to $t_1$, we get

5390: \be

5391: Q_1(t)^{-1} - Q_1(t_1)^{-1} \leq 4 (\Gamma' + \delta) Q_0 (t_1) (t-t_1),

5392: \nn\ee

5393: which is equivalent (\ref{eq:Q1lb}). This completes the proof of Lemma 

5394: \ref{lem:Q1lbound}. 

5395: 

5396: \nit{\bf Estimation of (\ref{eq:typ4}) for $t\ge t_1$} 

5397: Carrying out the differentiation in (\ref{eq:typ4}) we find that it suffices

5398: to bound terms of the type:

5399: \be 

5400: \cO\left(\alpha_0 |\beta_1|^2\right) 

5401: \left\la\chi, P_c\chi \right\ra\ \alpha_0\beta_1^{2m}

5402:  \ =\ \cO(\cE_0)Q_0Q_1^2\label{eq:1st}\\

5403: \cO\left(\alpha_0 |\beta_1|^2\right) 

5404:  \left\la\chi,

5405:  \int_0^te^{-i\cH_0(t-s)}P_c{\tilde\cH_0}^m \tilde\chi \alpha_0\beta_1^{2m}\ ds

5406: \right\ra

5407: \label{eq:2nd}\ee

5408: 

5409: We now consider (\ref{eq:2nd}), which we  break into the sum of 

5410: three integrals:   

5411: \be \cO(\bar\alpha_0 |\beta_1|^2) \left\langle\ \chi,

5412: \left(\ \int_0^{t_0}+\int_{t_0}^{t_1}+\int_{t_1}^t\right)

5413:  e^{-i\cH_0(t-s)}\ P_c\ {\tilde\cH_0}^k \tilde \chi \alpha_0\beta_1^{2m}\ 

5414:  ds\ \right\rangle\

5415: \label{eq:typ5}

5416: \ee

5417: \bigskip

5418: 

5419: \nit {\bf $\int_{t_1}^t:$}

5420: By the local decay estimate for $e^{-i\cH_0 t}\cH_0^k P_c$ of

5421:    Theorem \ref{theo:local_decay}, 

5422: the integral in (\ref{eq:typ5}) is bounded above by:

5423: \be

5424: \int^t_{t_1}\frac{ds}{\la t-s\ra^k} |\alpha_0 \beta_1^{2m} |

5425: \nn\ee

5426: This in turn is bounded above by 

5427: \be

5428: Q_0(t)^{1/2} \int^t_{t_1} \frac{ds}{\langle t-s\rangle^k} Q_1^m(s),

5429: \label{eq:typ6}\ee

5430: since $Q_0$ is increasing for $t\ge t_0$. 

5431: 

5432: 

5433: We now aim to  further bound (\ref{eq:typ6}) by ``extracting'' powers

5434: of $Q_1(t)$ from under the integral.  Recall that for $s\ge t_1$  

5435: \ba

5436: Q_1(s) &\le& \frac{Q_1 (t_1)}{1+2\Gamma Q_0(t_1)Q_1(t_1)|s-t_1|}\nn\\

5437:  &=& \frac{1}{2\Gamma Q_0(t_1)} \left[\frac{Q_1 (t_1) Q_0(t_1)

5438: 2\Gamma}{1+2\Gamma Q_0(t_1) Q_1 (t_1) |s-t_1|}\right]\nn\\

5439: &\le& \min \left\{

5440: Q_1(t_1), 2\Gamma^{-1} Q_0^{-1} (t_1) \langle s-t_1\rangle^{-1}\right\}

5441: \label{eq:Q1bound1}

5442: \ea

5443: and 

5444: \be

5445: Q_0^{-1} (t_1) \equiv Q_1(t_1)^{-1} \cE_0^{-r}, \ {\rm  and }\

5446:   Q_1 \leq \cE_0.

5447: \label{eq:Q0toQ1}\ee

5448: We write 

5449: \ba Q_1^m(s) &=& Q_1^{m-\bar k}(s) Q_1^{\bar k}(s)\ \le 

5450:  Q_1^{m-\bar k}(s)\frac{1}{\la s-t_1\ra^{\bar k}}

5451:                   \frac{1}{(2\Gamma Q_0(t_1))^{\bar k}}\nn\\

5452:              &=& Q_1^{m-\bar k}(s)\frac{1}{\la s-t_1\ra^{\bar k}} 

5453: \frac{\cE_0^{-r\bar k}}{(2\Gamma Q_1(t_1))^{\bar k}}\nn\\

5454:              &\le& Q_1^{m-2\bar k}(s) \cE_0^{-r\bar k} 

5455:                \la s-t_1\ra^{-\bar k},

5456: \label{eq:Q1mbound}\ea

5457: where we have used that $Q_1(t_1)\ge Q_1(s)$.

5458: 

5459: We consider separately the cases $t\ge 2t_1$ and $t_1\le t\le 2t_1$.

5460:  

5461: \nit{\bf $t\geq 2t_1$:}\  Take ${\bar k}=2$ and $m>2(r+2)$. Then, 

5462: the integral in (\ref{eq:typ6}) is bounded by 

5463: \be

5464: \cE_0^\rho\ Q_0(t)^{1\over2} \langle t \rangle^{-\min(k, \bar k)}, \ \rho>0.

5465: \nn\ee

5466: This implies, for $k$  sufficiently large, 

5467: \be

5468: \bigg|\cO(\alpha_0 |\beta_1|^2) \langle \chi, \int_0^t\cdots\rangle \bigg|

5469:   \leq

5470: \cO(\cE_0^\rho)

5471:  \left(\ Q_0 Q_1^2 + \langle t \rangle^{-4}\ \right).\  \rho>0

5472: \nn\ee

5473: \bigskip

5474: 

5475: \nit{\bf $t_1 \leq t < 2 t_1$:}

5476:  Let $t = t_1 + 2M,  M = (t-t_1) /2$ and rewrite the integral as

5477: \be

5478: \int^t_{t_1}\frac{ds}{\la t-s\ra^k}|\alpha_0\beta_1^{2m}| = 

5479:  \left(\int^{t_1 + M}_{t_1} + \int^{t_1 + 2 M}_{t_1 + M})\right)

5480:  \frac{ds}{\la t-s\ra^k}|\alpha_0\beta_1^{2m}|.

5481: \label{eq:split}

5482: \ee

5483: In a manner similar to the previous estimate, using that $Q_1$ is decreasing

5484:  we have, by (\ref{eq:Q1mbound}):

5485: \ba

5486: \int^{t_1 + M}_{t_1} \frac{ds}{\langle t-s\rangle^k} Q_1 (s)^{m/2}

5487: \langle s -t_1\rangle^{-\bar k}

5488: &\le& c \langle t - (t_1 + M)

5489: \rangle^{-k} Q_1(t_1)^{m/2}\nn\\

5490: &\le& c \langle t-t_1\rangle^{-k}Q_1(t_1) Q_1(t_1)^{{m\over2}-1}\nn\\

5491: &\le& c \langle t-t_1\rangle^{-k} \left[\ 1+4 (\Gamma' + \delta)

5492: Q_1(t_1) Q_0(t_1)|t - t_1|\ \right] Q_1 (t) Q_1 (t_1)^{m/2-1}\nn\\

5493: &\le& \cO(\cE_0)Q_1(t)

5494: \nn\ea

5495: for $k\ge1$.

5496: Hence

5497: \be

5498: \cO(\alpha_0|\beta_1|^2) Q_0^{1/2} \int^{t_1 + M}_{t_1} \cdots \leq \cO

5499:  (\cE_0) Q_0 Q_1^2.

5500: \nn\ee

5501: Furthermore, by (\ref{eq:Q1mbound}) the integral over $[t_1+M, t_1+2M]$ 

5502: is bounded by

5503: \be

5504: \int^{t_1+2M}_{t_1+M} 

5505:  \frac{ds}{\la t-s\ra^k \la s-t_1\ra^{\bar k}} Q_1(s)^{m/2}

5506:  \leq c Q_1(t_1)^{m/2} \langle

5507: M\rangle^{-\bar k}\leq c Q_1(t_1)^{m/2} \langle t-t_1\rangle^{-\bar k}

5508: \nn\ee

5509: and, as above, using the upper bound (\ref{eq:Q1lb}) for $Q_1(t_1)$

5510:  in terms of $Q_1(t)$ we have 

5511: \be

5512: \int^{t_1 + 2 M}_{t_1+M} \cdots \leq\cO (\cE_0) Q_1 (t) 

5513: \nn\ee

5514: for $\bar k \geq 1$.

5515: Therefore, for all $t > t_1$ the nonlocal (and local) ODE terms in the

5516: $Q_0$ equation are bounded by 

5517: \be

5518: \cO(\cE_0) \left[Q_0 Q_1^2 + \langle t \rangle^{-3}\right],

5519: \nn\ee

5520:  provided we

5521: control the integral $\int^{t_1}_0 \cdots\ ds$. 

5522: \bigskip

5523: 

5524: \nit{\bf $\int_0^{t_1}$:}

5525: Consider

5526: \be

5527: I(t) \equiv Q_0^{1/2} Q_1 \int^{t_1}_0\frac{ds}{\langle t-s\rangle^k}

5528: Q_0^{1/2} Q_1^{m-1}Q_1 ds.

5529: \label{eq:Idef}

5530: \ee

5531: By the mean value theorem, 

5532: $Q_1(s) = Q_1(s) - Q_1(t) + Q_1 (t) = \dot Q_1(\bar s) | t-s| + Q_1(t)$,

5533: where $s\le \bar s\le t$.

5534: Then,

5535: \ba 

5536: I(t) &\le& Q_0^{1\over2}(t)Q^2_1(t) 

5537:  \int^{t_1}_0\frac{ds}{\langle t-s\rangle^k}

5538: Q_0^{1\over2}(s) Q_1^{m-1}(s)\ ds\nn\\

5539: &+& I_1(t),\ {\rm where}\ 

5540: \label{eq:Ibound}\\

5541: I_1(t) &\equiv& c\ Q_0^{1\over2}(t) Q_1(t)

5542: \int^{t_1}_0\frac{ds}{\langle t-s\rangle^{k-1}}Q_0^{1\over2}(s)Q_1^{m-1}(s)

5543: \ Q^{1\over2}_0(\bar s) Q_1(\bar s),

5544: \label{eq:I1def}\ea

5545: where we have used 

5546: \be

5547: \dot Q_1 = \cO\left(Q_0^{1\over2} Q_1\right) + h.o.t. .

5548: \nn\ee

5549: %

5550: If $\bar s\le t_0$, then using that 

5551: $Q_0^{1\over2}(\bar s)\le \cE_0^\rho \la\bar s\ra^{-1}

5552:  \le \cE_0^\rho\la s\ra^{-1}$,

5553:  we have  the bound 

5554: \be

5555: I_1(t)\ \le\  \cO(\cE_0^\rho)Q_0^{1\over2}(t)Q_1(t)\Jt^{-2}

5556: \le \cO(\cE_0^\rho)\left( Q_0(t)Q_1^2(t) + \Jt^{-4}\right)

5557: %

5558: \nn\ee

5559: If $\bar s>t_0$, then since $Q_0(s)$ is monotonically increasing

5560: for $s\ge t_0$, and we have  

5561: \be

5562: I_1(t) \le 

5563: \frac{Q_0(t)Q_1(t)}{\la t-t_1\ra}\int_0^{t_1}

5564: \frac{ds}{\langle t-s\rangle^{k-1}}

5565:  Q_0^{1\over2}(s)Q_1^{m-1}(s)\  Q_1(\bar s)

5566: \nn\ee

5567: We now expand the latter factor of $Q_1$ in the integrand using the 

5568:  mean value theorem. Specifically,

5569:   there exists $s'$ with $t_0< \bar s\le s'\le t_1$

5570: such that:

5571: \ba

5572: Q_1(\bar s) &=& Q_1(t_1)+\dot Q_1(s')(s'-t_1)\nn\\

5573: &=&  Q_1(t_1) + \cO\left( Q_0^{1\over2}(s')Q_1(s') \right) |s'-t_1|

5574: \nn\\

5575: &\le& Q_1(t_1) + \cO\left( Q_0^{1\over2}(t_1)Q_1(s') \right) |\bar s-t_1|,

5576: \label{eq:mvt1}\ea

5577: where the last inequality follows by monotonicity of $Q_0$. 

5578: 

5579: Substitution into integrand in (\ref{eq:mvt1}) 

5580:  gives:

5581: \be

5582: I_1(t) \le

5583: c \cE_0^\rho \frac{Q_0(t)Q_1(t)}{\la t-t_1\ra}

5584:   Q_1^{1\over2}(t_1)\int_0^{t_1}

5585: \frac{ds}{\langle t-s\rangle^{k-3}}

5586:  Q_0^{1\over2}(s)Q_1^{m-1}(s),\label{eq:typ7}

5587: \ee

5588: where we have used (\ref{eq:Q0toQ1}) to replace $Q_0(t_1)$

5589: by $Q_1(t_1)$. 

5590: A higher order term (one proportional to $Q_1(t_1)$) is

5591: subsumed by the constant, $c$.   

5592: 

5593: From (\ref{eq:Q1lb}) we have

5594: \be

5595: Q_1(t_1)\ \le\   

5596: \left( 1+4(\Gamma'+\delta)Q_1(t_1)Q_0(t_1)(t-t_1)\right)\ Q_1(t)

5597: \nn\ee

5598: 

5599: \nit {\bf Case 1: $4(\Gamma'+\delta)Q_1(t_1)Q_0(t_1)(t-t_1)\le 100$} Here, 

5600: $Q_1(t_1)\le 101Q(t)$, and therefore

5601: \be

5602: I_1(t) \le

5603: c\cE_0^\rho \frac{Q_0(t)Q_1^{3\over2}(t)}{\la t-t_1\ra}

5604:   \int_0^{t_1}

5605: \frac{ds}{\langle t-s\rangle^{k-3}}

5606:  Q_0^{1\over2}(s)Q_1^{m-1}(s)

5607: \label{eq:typ8}\ee

5608: We now split the integral in (\ref{eq:typ8}) as 

5609:  $\int_0^{t_1}=\int_0^{t_0}+\int_{t_0}^{t_1}$.

5610: Using the $\la s\ra^{-1}$ decay of $Q_0^{1\over2}$ for $s\le t_0$

5611: we have

5612: \be

5613: \bigg|\int_0^{t_0} \cdot\bigg| \le \cE_0^{\rho}\la s\ra^{-1},

5614: \nn\ee

5615: and using the monotonicity of $Q_0$ for $t_1\ge s\ge t_0$ and the relation

5616: $Q_0(t_1)=\cE_0^rQ_1(t_1)\le 101\cE_0^rQ_1(t)$  we have

5617: \be

5618: \bigg|\int_{t_0}^{t_1} \cdot\ ds\bigg| \le \cE_0^\rho 

5619:  Q_0^{1\over2}(t_1) = \sqrt{101}\cE_0^{\rho-{r\over2}}Q_1^{1\over2}(t)

5620: \nn\ee

5621: Therefore, choosing $m$ sufficiently large, we get

5622: \ba

5623: I_1(t)\le c'\left( \cE_0^{\rho_1}Q_0(t)Q_1^{3\over2}(t) \la t\ra^{-1} 

5624:  + \cE_0^{\rho_2}Q_0(t)Q_1^2(t)\right) 

5625: \le c''\cE_0^{\rho_3}\left( Q_0Q_1^2 + \Jt^{-4}\right),\nn\\

5626: &&\label{eq:I1bound1}\ea

5627: where  $\rho_3=\min\{\rho_1,\rho_2\}$.

5628: 

5629: \nit {\bf Case 2: $4(\Gamma'+\delta)Q_1(t_1)Q_0(t_1)(t-t_1)\ge100$}  

5630:  Then, (\ref{eq:Q1lb}) and monotonicity of $Q_0$ for $t\ge t_0$ 

5631:  implies

5632: %\footnote{\bXYZ see earlier calculation}

5633: \be  

5634: \frac{1}{\la t-t_1\ra}\le 4(\Gamma'+\delta)Q_0(t)Q_1(t).

5635: \nn\ee

5636: This gives

5637: \be

5638: I_1(t)\le C\cE_0^\rho Q_0(t)Q_1^2(t)\nn\ee

5639: %\footnote{\bXYZ Probably delete from here on a bit

5640: %$$

5641: %I(t) \leq Q_0^{1\over2} Q^2_1\int^{t_1}_0 \frac{ds}{\langle t-s\rangle^k}

5642: %Q_0^{1/2} Q_1^{m-1} + Q_0^{1/2} Q_1 \int^{t_1}_0\frac{ds}{\langle

5643: %t-s\rangle^{k-1}}Q_0 Q_1^{m+m'}

5644: %$$

5645: %$$

5646: %\leq Q_0^{1/2} Q_1^2\int^{t_1}_0\frac{ds}{\langle t-s\rangle^k}

5647: %Q_0^{1/2} Q_1^{m-1} + \cO(\cE_0)\langle t\rangle^{-2} Q_0^{1/2}

5648: %Q_1 + Q_0^{1/2} (t)Q_1 (t) Q_0 (t_0)\times

5649: %$$

5650: %$$

5651: %\int^{t_1}_{t_0} \frac{ds}{\langle t-s\rangle^{k-1}} Q_1^{m+m'}.

5652: %$$

5653: %where we used that $Q_0^{1/2}(s) \leq \cO(\cE_0) \langle

5654: %s\rangle^{-1} {\rm  for } s\leq t_0$.

5655: %As before, we pull another power of $Q_1$ out of the integral terms:

5656: %$$

5657: %Q_0^{1/2} Q_1^2\int^{t_1}_0 \frac{ds}{\langle t-s\rangle^k} Q_0^{1/2}

5658: %Q^{m-1}_1 + Q_0^{1/2}(t)Q_1(t) Q_0 (t_0)

5659: %\int^{t_1}_{t_0}\frac{ds}{\langle t-s\rangle^{k-1}}Q_1^{m+m'} \leq

5660: %$$

5661: %$$

5662: %\leq Q_0^{1/2} Q^2_1\left[Q_1\int^{t_1}_0\cdots +

5663: %\int^{t_1}_{0}\frac{ds}{\langle t-s\rangle^{k-1}}Q^{1/2}_0

5664: %Q_0^{1/2} (\bar s) Q_1^{m-1} (s)Q^{m'}_1(\bar s)\right] \leq

5665: %$$

5666: %$$

5667: %+Q_0^{1/2} Q_1 Q_0(t_0) \left[ Q_1\int^{t_1}_{t_0} \cdots +

5668: %\int^{t_1}_{t_0} \frac{ds}{\langle t-s\rangle^{k-2}}

5669: %Q_1^{m+m'-1}Q_0^{1/2} (\bar s) Q_1^{m'} (\bar s)\right]

5670: %$$

5671: %$$

5672: %\leq Q_0^{1/2} Q^3_1 \cO(\cE_0) \langle t\rangle^{-1} + \cO

5673: %(\cE_0) Q_0^{1/2} Q_0^{1/2} (t_0) Q_1^3

5674: %$$

5675: %$$

5676: %+\cO(\cE_0) Q_0^{1/2}Q_1^2\left[\langle t\rangle^{-2} +

5677: %Q_0^{1/2} (t_0) \langle t\rangle^{-1}\right] + \cO(\cE_0)

5678: %Q_0^{1/2} Q^2_1 Q_0 (t_1)

5679: %$$

5680: %$$

5681: %+Q_0^{3/2} Q^2_1 \cO(\cE_0) + Q_0^{1/2} Q_1 Q_0(t_0) Q_0^{1/2}

5682: %(t_1) \cO(\cE_0)

5683: %$$

5684: %$$

5685: %\leq Q_0^{1/2} Q^2_1\cE_0^{-r} \cO(\cE_0) Q_0 (t) \langle

5686: %t\rangle^{-1} + \cO (\cE_0) \left[Q_0Q_1^2 + \langle

5687: %t\rangle^{-4}\right] \leq \cO (\cE_0) \left[Q_0Q_1^2 + \langle

5688: %t\rangle^{-4}\right];

5689: %$$

5690: %where we used that 

5691: %$$

5692: %Q_0(t_0) \leq Q_0 (t_1)\leq \cE_0^r Q_1(t_1) \leq \cE_0^r Q_1(t)

5693: %(1+4 \Gamma(1+\delta')Q_0(t_1) Q_1(t_1) |t-t_1|).

5694: %$$

5695: %and 

5696: %$$

5697: %Q_1(t) \leq \cE_0^{-r} Q_0(t)\ {\rm  for }\ t> t_1.

5698: %$$

5699: %}

5700: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

5701: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

5702: We now turn to (\ref{eq:typ4}) for $t\in [t_0,t_1]$.

5703: It suffices to estimate the nonlocal terms 

5704: \be

5705: \cO(\alpha_0|\beta_1|^2)\left\la\chi,\int_0^te^{-i\cH_0(t-s)}

5706:  P_c{\tilde\cH_0}^m

5707: {\tilde\chi}\alpha_0\beta_1^{2m}\ ds\right\ra

5708: \nn\ee

5709: for $t_0\le t\le t_1$.

5710: \ In this region $Q_0$ and $(Q_0/Q_1)$ are increasing functions.

5711:   Also $Q_0(t)

5712: \leq \cE_0^r Q_1 (t)$.

5713: The main difficulty is the need to pull a factor of $Q_1(t)$ out of

5714: the nonlocal term.  To this end we use the following proposition:

5715: 

5716: \begin{prop}\label{prop:I2}

5717: There exists a constant $\delta>0$ such that for $t\ge t_0$

5718: \be

5719: 2\Gamma\int^t_{t_0} Q_0Q_1^m ds \leq \delta 

5720:  Q_0^{1\over2}(t) Q_1(t)^{m-2}\ + \ Q_0(t)Q_1^{m-2}(t)\ +

5721: \ {\rm h.o.t.}

5722: \label{eq:I2int}\ee

5723: \end{prop}

5724: 

5725: \begin{cor}\label{cor:cor1}

5726: \be

5727: \int^t_{t_0} \frac{ds}{\langle t-s\rangle^{3/2}} 

5728: Q_0^{1/2}(s) Q_1^{m-1}(s)

5729:   \leq

5730: C(1 +\delta)Q_0^{3\over8} Q_1^{m-{1\over2}}.

5731: \nn\ee

5732: \end{cor}

5733: \nit The corollary follows from Proposition \ref{prop:I2} by 

5734:  the H\"older's  inequality. 

5735: 

5736: {\bf Proof of Proposition \ref{prop:I2}:}

5737: Recall that

5738: \be

5739: \frac{dQ_0}{dt} = 2\Gamma Q_0Q_1^2 + R_0,\ \  

5740: \frac{dQ_1}{dt} \leq 

5741:  -4\Gamma Q_0 Q_1^2 + \cO\left( Q_0^{1\over2} Q_1^m\right)

5742: \label{eq:Q0Q1sys}

5743: \ee

5744: 

5745: \nit\underline{Claim:}

5746: \be

5747: {1\over Q_0^{1\over2}}\int_{t_0}^t R_0 ds\ \le\

5748:  \cO(\cE_0^\rho)\left[ 1+\int_{t_0}^tQ_0(s)Q_1^{2m}(s)ds\right]

5749: \label{eq:claim}

5750: \ee 

5751: \nit\underline{Proof of claim:}

5752:  The leading order term of $R_0$, in the variable $\alpha_0$,

5753:   which is nonlocal, 

5754: is a term of the form (\ref{eq:nlc}). From this term we have  

5755: after integration by parts to obtain $e^{-i\cH_0(t-s)}\cH_0$ from

5756:  (\ref{eq:nlc}),

5757: \ba

5758: \int^t_{t_0} 

5759:  R_0 &\leq& Q_0^{1\over2}(t)\int^t_{t_0} Q_1 (t') \int^{t'}_{t_0}

5760: \frac{ds}{\langle t-s\rangle^{5/2}} Q_0^{1/2} Q_1^m ds dt'\nn\\

5761: &\leq& Q_0^{1\over2}(t) \int^t_{t_0} Q_1 (t') dt' \left[\int^{t'}_{t_0}

5762: \frac{ds}{\la t-s\ra^{4-\delta}} +\frac{1}{\la t-t'\ra^{1+\delta'}}

5763:  \int^{t'}_{t_0} Q_0 Q_1^{2m}

5764: ds \right]\nn\\

5765: &\leq& c Q_0^{1\over2}(t) \int^{t'}_{t_0} Q_1 (t') dt' \left[ \langle

5766: t-t'\rangle^{-3+\delta'} + \langle t-t_0\rangle^{-3+\delta'}+ \langle

5767: t-t'\rangle^{-1-\delta'}\int^{t'}_{t_0} Q_0Q_1^{2m} ds \right]\nn\\

5768: &\leq& cQ_0^{1\over2}(t) \left[ \cO(\cE_0) (1 + \int^t_{t_0} Q_0

5769: Q_1^{2m}\ ds)\right],

5770: \nn\ea

5771: thus proving the claim.

5772: \bigskip

5773: 

5774: We first rewrite the right hand side of (\ref{eq:I2int}) and 

5775: use the differential equation for $Q_0$ and the above claim  

5776: to integrate by parts:

5777: \ba

5778: &&2\Gamma\int^t_{t_0} Q_0 Q_1^m ds = \int^t_{t_0} 2\Gamma Q_0 Q^2_1

5779: Q^{m-2}_1 ds\nn\\ 

5780: &=& \int_{t_0}^t Q_1^{m-2}\frac{d}{ds}\left[ Q_0-\int_{t_0}^sR_0 \right]\ ds

5781: \nn\\

5782: &=& Q_1^{m-2}(t')\left[ Q_0-\int_{t_0}^sR_0 \right]\bigg|_{t_0}^t

5783:  -\int_{t_0}^t \frac{d}{ds}Q^{m-2}(s)

5784:  \left[ Q_0(s)-\int_{t_0}^{s}R_0 \right] \ ds\nn\\

5785: &=& Q_1^{m-2}(t)Q_0(t) - Q_1^{m-2}(t_0)Q_0(t_0) - 

5786: Q_1^{m-2}(t)\int_{t_0}^tR_0(s) ds\nn\\

5787: &&\ + (m-2)\int_{t_0}^t dt' Q_1^{m-1}(t')Q_0(t')

5788:        \left[ 4\Gamma Q_0(t')+

5789:  \cO\left(Q_0^{1\over2}(t')Q_1^{m-2}(t')\right)\ \right] \nn\\

5790: &&\ + (m-2)\int_{t_0}^t dt' Q_1^{m-1}(t')

5791:  \left[ 4\Gamma Q_0(t')+\cO\left(Q_0^{1\over2}(t')Q_1^{m-2}(t')\right)\right]

5792:  \ \left( \int_{t_0}^{t'}R_0(s) ds\right)

5793: \nn\ea

5794: The first term on the right hand side is of the form we want (the right

5795: hand side of (\ref{eq:I2int}). The second term is negative so we can drop it.

5796:  The third term is bounded by $\cO(\cE_0)Q_1^{m-2}(t)Q_0^{1\over2}(t)$

5797:  by the claim above, plus $\cO(\int_{t_0}^t\ Q_0Q_1^m\ ds)$, which

5798:  is of the above form   

5799:  and can be absorbed by the left hand by smallness of

5800:  $\cE_0$,  

5801: %{\bf XYZ Finish this off} 

5802: where we used $t_0<s<t_1,\  Q_0 \leq \cE_0^r Q_1$.

5803: This completes the proof of Proposition \ref{prop:I2}.

5804: \bigskip

5805: 

5806: Using the proposition and its corollary it is easy to  bound, 

5807:  for $t\ge t_0$,

5808: 

5809: \be

5810: \cO(\alpha_0 |\beta_1|^2 + |\alpha_0|^2\beta_1)

5811: \ \D^k_t\ \left\la \chi, \int^t_0 e^{-i\cH_0(t-s)}\ P_c\tilde \chi \alpha_0

5812: \beta_1^{2m} ds\ \right\ra

5813: \nn\ee

5814: by $\cO (\cE_0^\rho) \left( Q_0 Q_1^2 + \langle t\rangle^{-3}\right)

5815: + h.o.t.$

5816: \bigskip

5817: 

5818: It remains to consider $t\leq t_0$.  In this case we only know that

5819: $|\talpha_0| \le k_0 \langle t\rangle^{-1}$.

5820: 

5821: The first term of (\ref{eq:nlc})

5822: \be

5823: \bar \alpha_0

5824:   \cO\left(\bar\alpha_0\beta_1 e^{-i\lambda_+t} \right)

5825:  \langle \chi,

5826: \int^t_0 e^{-i\cH_0(t-s)}P_c \alpha_0 \beta_1^{2m} \tilde \chi ds\rangle

5827: \label{eq:nlc1}

5828: \ee

5829: is bounded above by $\cO(\la t\ra^{-3})$,

5830:  due to dispersive estimates and the decay of 

5831: the decay of $Q_0$ for $t\le t_0$.

5832: 

5833: In the second term of (\ref{eq:nlc}),

5834: \ba

5835: && \cO (\alpha_0|\beta_1|^2)\partial_t^k\langle \chi, \int_0^t\cdots

5836: \rangle\nn\\

5837: &\sim& 

5838: \cO\left(Q_0^{1\over2}(t)Q_1(t)\right)\ 

5839: \left\la \chi,\int_0^te^{-i\cH_0(t-s)}P_c{\tilde\cH_0}^k{\tilde\chi}

5840:  \talpha_0\tbeta_1^{2m} e^{i\Omega s} \ ds\right\ra,\ \Omega\ne0

5841: \nn\\

5842: &&\label{eq:typ9}

5843: \ea

5844: we need to pull $Q_0^{1\over2}(t)Q_1(t)$ out of the nonlocal (integral) term. 

5845: 

5846: By dispersive estimates, we have the bound 

5847: \be

5848: \cO\left(Q_0^{1\over2}(t)Q_1(t)\right)

5849:  \int_0^t\frac{1}{\la t-s\ra^{ {3\over2}+k} } |\talpha_0(s)|

5850: Q_1^m(s)\ ds\label{eq:typ10}

5851: \ee

5852: To pull the $Q_1$ term we proceed as earlier.

5853: By the mean value theorem, there exists $\bar s$, with 

5854: $s\le \bar s\le t\le t_0$ such that

5855: \ba

5856: Q_1(s) &=& Q_1(t) + Q_1(s) - Q_1(t) = Q_1 (t) - \dot Q_1 (\bar s) (t-s)\nn\\

5857:  &=& Q_1(t) + \cO\left(Q_0^{1\over2}(\bar s)Q_1(\bar s)\right) (t-s)\nn\\

5858:  &=& Q_1(t) + \cO\left(\cE_0^\rho\la\bar s\ra^{-1}\la t-s\ra\right), 

5859: \nn\ea

5860: where we have used that $\dot Q_1\sim Q_0^{1\over2}Q_1$ for $t\le t_0$.

5861: Therefore, the expression in (\ref{eq:typ10}) satisfies a bound of order 

5862: \ba

5863: &&Q_0^{1\over2}(t)Q_1^2(t)

5864: \int_0^t\frac{1}{\la t-s\ra^{ {3\over2}+k} }|\talpha_0(s)|

5865: Q_1^{m-1}(s)\ ds\nn\\

5866: && + 

5867: Q_0^{1\over2}(t)Q_1(t)

5868:  \int_0^t\frac{1}{\la t-s\ra^{ {1\over2}+k} }Q_0^{1\over2}(s)

5869: \cO\left(Q_0^{1\over2}(\bar s)Q_1^{m'}(\bar s)\right) Q_1^{m-1}(s)\ ds

5870: \nn\\

5871: &=&Q_0^{1\over2}(t)Q_1^2(t)

5872: \int_0^t\frac{1}{\la t-s\ra^{ {3\over2}+k} }\ Q_0^{1\over2}(s)

5873: Q_1^{m-1}(s)\ ds\nn\\

5874: &+& \cE_0^\rho\ Q_0^{1\over2}(t)Q_1(t) \frac{1}{\la t\ra^2}

5875: \label{eq:typ11}

5876: \ea

5877: where the last term, which is bounded by 

5878: $\cE_0^\rho \left( Q_0Q_1^2 + \la t\ra^{-4}\right)$,

5879:  is obtained using the decay of $Q_0(s)$ for $s\le t_0$. 

5880: 

5881: It remains to estimate the second to last term in (\ref{eq:typ11}).

5882: Estimating the convolution, using the $\la s\ra^{-1}$ decay of

5883:  $Q_0(s)$, we obtain the bound

5884: $\cO\left(Q_0^{1\over2}(t)Q_1^2(t)\right) \la t\ra^{-1}$ which

5885: is not bounded by the desired 

5886:  $\cO\left(\ \cE_0^\rho (Q_0Q_1^2 + \la t\ra^{-3})\ \right)$.

5887: 

5888: We will obtain the desired bound  by turning to an earlier expression,

5889:  derivable from 

5890:  (\ref{eq:typ9}). The expression we must consider is 

5891: \be 

5892: \cO\left(Q_0^{1\over2}(t)Q_1^2(t)\right)\

5893: \left\la \chi,\int_0^te^{-i\cH_0(t-s)}P_c{\tilde\cH_0}^k{\tilde\chi}

5894:  \talpha_0(s)\tbeta_1^{2m-2}(s) e^{i\Omega s} \ ds\right\ra,\ \Omega\ne0

5895: \label{eq:typ12}

5896: \ee

5897: We will show that this term is  

5898: of order $\cE^\rho Q_0(t)Q_1^2(t) \la t\ra^{-3/2}$, which implies the 

5899: desired bound.  

5900: We proceed as follows. First, by equation (\ref{eq:normalform})

5901: of  Proposition \ref{prop:normalform} the equation for $\talpha_0$ may

5902: be rewritten as:

5903: \be

5904: i\D_t\ta0(t) = (c+i\Gamma_\omega)|\tb1(T)|^4\ta0(t)\ +\ 

5905:  (c+i\Gamma_\omega)\left( |\tb1(t)|^4 - |\tb1(T)|^4\right)\ta0(t)

5906:  + F_\alpha\label{eq:alphaT}

5907: \ee 

5908: Introducing $\ta0^\#(t)$ via the equation

5909: \be \ta0(t) \equiv e^{-it(c+i\Gamma_\omega)|\tb1(T)|^4t}\ta0^\#(t),

5910: \label{eq:ta0sharpdef}

5911: \ee

5912: we have

5913: \be

5914: i\D_t\ta0^\#(t) = (c+i\Gamma_\omega)\left( |\tb1(t)|^4-|\tb1(T)|^4\right)

5915:  \ta0^\#(t)  +   e^{it(c+i\Gamma_\omega)|\tb1(T)|^4}F_\alpha

5916: \label{eq:ta0sharp-eqn}

5917: \ee

5918: The integral in (\ref{eq:typ12}) can be written, keeping only leading order

5919: terms, as  

5920: \ba

5921: &&\left\la \chi,\int_0^te^{-i\cH_0(t-s)}P_c{\tilde\cH_0}^k{\tilde\chi}

5922:  \ta0^\#(s)\tb1^{2m-2}(s) e^{i\tOmega s} \ ds\right\ra\nn\\

5923: &=&\ \left\la \chi,\int_0^te^{-i\cH_0(t-s)}P_c{\tilde\cH_0}^k{\tilde\chi_1}

5924:  \dot{\ta0}^\#(s)\tb1^{2m-2}(s) e^{i\tOmega s} \ ds\ +\ \dots\right\ra\nn\\

5925: &=&\ \left\la \chi,\int_0^te^{-i\cH_0(t-s)}P_c{\tilde\cH_0}^k{\tilde\chi_1}

5926:  \left(|\tb1(t)|^4-|\tb1(T)|^4

5927:    \right)\ta0^\#(s)\tb1^{2m-2}(s) e^{i\tOmega s} \ ds\right\ra\nn\\

5928: &\le&

5929:  \cO(\cE_0^2)\int_0^t\frac{ds}{\la t-s\ra^{{3\over2}+k}\la s\ra}

5930:                       \left(|\tb1(t)|^4-|\tb1(T)|^4 \right)\ +\ \dots

5931: \label{eq:typ13}

5932: \ea

5933: It suffices to show

5934: \begin{prop}\label{prop:P1diff}

5935: \be

5936: |P_1(t)-P_1(s)|\ \le\ \frac{C}{\la s\ra^{1\over2}},\ 0\le s\le t

5937: \label{eq:P1est}

5938: \ee

5939: \end{prop}

5940: For the proof we turn to the $P_1$ equation (\ref{eq:P1eqn}):

5941: \ba

5942: {dP_1\over dt} &=& -4\Gamma P_0P_1^2 \ +\ R_1\nn\\

5943: R_1 &=& R_1[\ta0,\tb1,\eta,t]=  2\Im (\ol{\tb1}F_\beta).

5944: \label{eq:P1eqna}

5945: \ea

5946: The key term in $\Im (\ol{\tb1}F_\beta)$ is the form

5947: $\tb1^m\ta0^\#e^{i\Omega t},\ \Omega\ne0$. We have, after two integrations

5948: by parts,  

5949: \ba

5950: &&\left| P_1(t)-P_1(s)\right|\ \le\ \int_s^t \tb1^m(s)

5951:              \ta0^\#(s')e^{i\Omega s'}ds'

5952: \ +\ \dots\nn\\

5953: &=&\ \tb1^m(s)\ta0^\#(s'){1\over i\Omega}e^{i\Omega s'}\left.\right|_s^t

5954:  \ -\ \int_s^t \frac{d}{ds'}\left(\tb1^m(s')\ta0^\#(s')\right)

5955:  {1\over i\Omega}e^{i\Omega s'}\ ds' +\ dots\nn\\

5956: &=&\ \cO\left(\frac{1}{\la s\ra}\right) + \cO\left(\frac{\log s}{\la s\ra}\right) +

5957:  \cO(\cE_0)\int_s^t\ \left|

5958:  \left(P_1(t)-P_1(s')\right)^2\ta0^\#(s')e^{i\tOmega s'} ds'\right|\nn\\

5959: &\le& \cO\left(\frac{1}{\la s\ra}\right) + 

5960:  \cO\left(\frac{\log s}{\la s\ra}\right) +

5961:  \cO(\cE_0)\sup_{s''}\la s''\ra^{1\over2}|P(t)-P(s'')|

5962:  \int_s^t \frac{1}{\la s'\ra^2} ds'\nn\\

5963: &=& \cO\left(\frac{1}{\la s\ra}\right) + 

5964:  \cO\left(\frac{\log s}{\la s\ra}\right) + 

5965:  \cO\left(\frac{\cE_0}{\la s\ra}\right)

5966:  \sup_{s''}\la s''\ra^{1\over2}|P(t)-P(s'')|. 

5967: \label{eq:P1diff}

5968: \ea

5969: Multiplication by $\la s\ra^{1\over2}$ and taking the supremum over

5970:   $s\le t$ implies Proposition \ref{prop:P1diff} and therewith

5971: the proof of part (a) of Proposition \ref{prop:ODE}.

5972: \bigskip

5973: 

5974: \subsection{Proof of part (2) of Proposition \ref{prop:ODE}}

5975: 

5976: The proof is similar to that of part (a) but simpler, 

5977:  since we can allow for the nonlocal terms to be controlled, in addition,  by 

5978: terms of order $Q_0^{1\over2}Q_1^m$.

5979: 

5980: The leading contributions are again nonlocal, linear $\tetab$

5981: contributions:

5982: 

5983: From (\ref{eq:beta_eqn3}) we read 

5984: \ba 

5985: i\partial_t \beta_1 &\sim& \cdots + 2 \lambda \langle \chi, \bar

5986: \Psi_1\tetab + \Psi_1 {\overline{\tetab}}\rangle e^{i\lambda_+ t}

5987: \alpha_0\nn\\

5988: &+& \lambda e^{i\lambda_+ t} 2 \langle \chi, \Psi_1 \tetab\rangle

5989: \bar \alpha_0\nn\\

5990: &&

5991: +\lambda e^{i\lambda_+ t} \langle \chi, \Psi^2_1 {\ol{\tetab}}

5992: + 2|\Psi_1|^2 \tetab\rangle + h.o.t.\\

5993: &+& X^{-1} (t) (A(t) - A(T)) X(t) \beta_1.\label{eq:beta4}

5994: \ea

5995: The first two terms on the righ hand side of (\ref{eq:beta4}) 

5996:  are easily seen to be of 

5997: order 

5998: \noindent $O(\alpha_0) P_1^{2m}$ or $\cO(\cE_0^\rho) P_0P_1^2$, 

5999:  by integration

6000: by parts over the ODE source terms in $\tetab$, as before.

6001: 

6002: The third term contributes to the $P_1$ equations, after normal form

6003: transformations and remaining resonant terms:

6004: 

6005: $$\Im\bar \beta_1 e^{i\lambda_+ t} \partial^n_t \langle \chi,

6006: |\beta_1|^2\int^t_0 e^{-i \cH_0(t-s)}P_c \alpha_0 |\beta_1 |^2 \tilde \chi

6007: ds\rangle

6008: $$

6009: and higher order/similar terms.

6010: \bigskip

6011: 

6012: The leading term, after integration by parts of the integral term

6013: (note that $H^{-1}P_c$ is bounded):

6014: 

6015: $$

6016: \bigg|\Im\bar \beta_1 e^{i\lambda_+ t}|\beta_1|^2\langle \chi,

6017: \int^t_0 e^{-i\cH_0(t-s)} P_c \cH_0^n\talpha_0\tbeta_1^{2m} 

6018:  \tilde \chi ds\bigg|

6019: $$

6020: $$

6021: \leq c |\beta_1|^3 \int^t_0 \frac{ds}{\langle

6022: t-s\rangle^k}Q_0^{1/2}(s) Q_1^m.

6023: $$

6024: To this end we repeat the argument of part (a) to estimate the above

6025: integral by

6026: $$

6027: Q_0^{1\over2}(t) Q_1^{m-1}(t) + \cO(t^{-3}) + h.o.t. (ODE).

6028: $$

6029: The main new type of term we need to control comes from the last term

6030: on the term $\cR_\beta$ in (\ref{eq:beta_eqn2}-\ref{eq:Rbeta_def}), 

6031: coming from the difference $\bA(t) - \bA(T)$.

6032: 

6033: This term contributes to the $P_1$ equation terms like 

6034: \be

6035: \Im \left(\ |\beta_1|^2 \left[\alpha_0^2(t) -

6036: \alpha_0^2(T)\ \right] {\bar \alpha}_0^2(t)\ \right), \Im \left[

6037: e^{i\omega t } \beta_1^2\ (P_0(t) - P_0(T)) \cO(\alpha^2_0(T))\right].

6038: \nn\ee

6039: %which follows from (6.39), (6.44), (6.36) \bXYZ????.

6040: 

6041: The term with phase $e^{i\omega t}$ can be integrated by parts, and

6042: gives higher order terms;

6043: 

6044: The term without a phase requires the estimate of 

6045: $$

6046: \alpha_0 (t)^2 - \alpha_0(T)^2 = 2\int^T_t\alpha_0

6047: \frac{d\alpha_0}{ds} ds.

6048: $$

6049: 

6050: Using that 

6051: $$

6052: \frac{d\alpha_0}{ds} = c\bar \alpha_0\beta^2_1 e^{-2i\lambda_+ t} + h.o.t.

6053: $$

6054: we can repeatedly integrate by parts to get

6055: $$

6056: \alpha_0^2 (t) - \alpha^2_0 (T) = \ {\rm local\ terms }\ + \int^T_t 

6057: \cO(\alpha_0^2) \beta_1^{2m} ds + h.o.t.

6058: $$

6059: where local terms = $\cO(\bar \alpha_0 P_1)$ and higher.

6060: 

6061: clearly, then

6062: $$

6063: |\beta_1|^2 \bar \alpha_0^2 (t) \left[\ { local\ terms }\ +

6064: h.o.t.\right] \leq \cO(\cE_0) \left[ P_0 P_1^2 + \langle t

6065: \rangle^{-3}\right].

6066: $$

6067: So, we need to estimate

6068: \be

6069: |\beta_1|^2 \bar\alpha_0^2 (t) \int^T_t \cO(\alpha_0^2) \beta_1^{2m} ds

6070: \ {\rm by}\ \cO(\cE_0^\rho)Q_0Q_1^2\ +\ {\rm h.o.t.}

6071: \nn\ee

6072: For $t\geq t_1 \quad  |\beta_1|^2 \leq Q_1$ is monotonic decreasing.

6073:   Hence 

6074: $$

6075: |\int^T_t \cO(\alpha_0^2) \beta_1^{2m} ds | \leq |\beta_1(t)|^2

6076: \int^T_t Q_0Q_1^{m-1} ds.

6077: $$

6078: 

6079: For $t> t_1, Q_1\downarrow$ is bounded by 

6080: $$

6081: Q_1(t) \leq Q_1(t_1) (1+2\Gamma Q_0(t_1) Q_1 (t_1) (t-t_1))^{-1} \leq c

6082: \langle t-t_1\rangle^{-1} Q_0 (t_1)^{-1}

6083: $$

6084: $$

6085: = cQ_1 (t_1)^{-1} \cE_0^{-r} \langle t-t_1\rangle^{-1}.

6086: $$

6087: since $Q_0(t_1) = \cE_0^rQ_1 (t_1)$.

6088: 

6089: So,

6090: $$

6091: \left[Q_1(t)^{1+r} Q_1(t)\right]^k \leq \cO(\cE_0) \langle

6092: t-t_1\rangle^{-k}.

6093: $$

6094: Hence, for $k>1$, or for $m> 2 + r + 1 = r+3$,

6095: $$

6096: \int^T_t Q_0 Q_1^{m-1} \leq \cO(\cE_0)

6097: $$so 

6098: $$

6099: \big|\int^T_t \cO(\alpha_0^2) \beta_1^{2m} \big| \leq \cO(\cE_0)

6100: Q_0Q_1^2.

6101: $$

6102: 

6103: If $t>t_0, t < t_1, Q_0 (t) \leq \cE_0^rQ_1 (t); \

6104:  {\rm  therefore }$

6105: $$

6106: \int^T_t Q_0 Q_1^{m-1} ds = \int^T_t Q_0Q_1^2 Q_1^{m-3} ds =

6107: \left(\frac{1}{2\Gamma} Q_0 (s) + R_0(s)\right) Q_1^{m-3} (s)

6108: \big|^T_{s=t}

6109: $$

6110: $$

6111: -\int^T_t \left(\ 

6112:  \frac{1}{2\Gamma} Q_0 + \int_{t_0}^sR_0(s)ds\ \right) Q_1^{m-4} (m-3)

6113: \frac{dQ_1}{ds} ds.

6114: $$

6115: The first term on the right hand side is local and high order.

6116: The second term on the right hand side  is bounded by 

6117: \ba

6118: c\int^{t_1}_t Q_0^2 Q_1^{m-2} ds + \int^T_{t_1} \left(\cdots\right) +

6119: \ {\rm h.o.t.}\ &\le& \cO(\cE_0) Q_1^2 +\cO(\cE_0^\rho)\la t\ra^{-3} + 

6120:  c\int^{t_1}_t Q_0^2 Q_1^{m-2} ds\nn\\

6121: &\le& \cO (\cE_0) \left[ Q_1^2 + \langle t

6122: \rangle^{-3}\right] + c \cE_0^r\int^{t_1}_t Q_0Q_1^{m-1} ds.

6123: \nn\ea

6124: 

6125: Hence,

6126: \be

6127: \int^T_t Q_0Q_1^{m-1} ds \leq \cO(\cE_0) \left[ Q_1^2 +

6128: \langle t\rangle^{-3}\right] + \cO(\cE_0) \int^T_t Q_0Q_1^{m-1} ds

6129: \nn\ee

6130: which implies that 

6131: \be

6132: \int^T_t Q_0 Q_1^{m-1} ds \leq \cO(\cE_0) \left[ Q_1^2 +

6133: \langle t \rangle^{-3}\right]

6134: \nn\ee

6135: for all $t\geq t_0$.

6136: 

6137: For $0< t < t_0$, we need to estimate 

6138: $$

6139: |\beta_1 |^2 \bar\alpha^2_0(t) \int^{t_0}_t O (\alpha_0^2) \beta_1^{2m}

6140: ds.

6141: $$

6142: Using that for $t< t_0, | \alpha_0 (t)|^2\leq k_0 \la t\ra^{-1}

6143:  \leq \cE_0 \langle t_0\rangle^{-1} \langle t\rangle^{-1}$ the 

6144:  above

6145: expression is bounded by 

6146: $$

6147: \cO(\cE_0) k_0(t_0) \langle t\rangle^{-1} (\ln

6148: \frac{t_0}{\langle t\rangle}) k_0 (t_0) \leq \cO(\cE_0)

6149: k_0(t_0) \langle t \rangle^{-1} \left(\frac{\langle t\rangle}{\langle

6150: t_0\rangle} \ln\frac{t_0}{\langle t\rangle}\right) \langle t

6151: \rangle^{-1}

6152: $$

6153: $$

6154: \leq \cO(\cE_0) k_0 (t_0) \langle t\rangle^{-2}.

6155: $$

6156: 

6157: This completes the proof of Proposition \ref{prop:ODE}. \bigskip

6158: 

6159: \section{$R_1^{(j)}(\tetab)$ terms of  Proposition \ref{prop:Master}}

6160: \label{sec:Rj1teta}

6161: 

6162: \begin{prop}\label{prop:Riteta} 

6163: Assume either $t\le t_0$ or Montonicity Property {\bf Q} on 

6164:  $[t_0,t_0+\delta_*]$. Then, the terms $R_1^{(i)}[\tetab],\ i=0,1$ 

6165:  in (\ref{eq:P0P1}) satisfy the estimates:

6166: \be

6167: \left|R_1^{(i)}[\tetab]\right| \leq \cO(\cE_0) \left[2\Gamma Q_0 Q_1^2 +

6168: \langle t_0\rangle^{-1} \langle t\rangle^{-2} \left(Poly \left[P_0(0), P_1(0),

6169: P_0(T), P_1 (T)\right]   \right)   \right]

6170: \label{eq:Rjitetabound}

6171: \ee

6172: where Poly $[\cdots]$ stands

6173: for polynomial in the bracketed variables.

6174: \end{prop}

6175: 

6176: {\bf Proof:}  The contributions to $R_1^{(i)}(\tetab)$ comes

6177: form linear and nonlinear terms in $\tetab$ in the $P_i$

6178: equations.

6179: 

6180: Consider first the nonlinear contributions:

6181: 

6182: In the $P_0$ equations we have terms like (7.41)

6183: \ba

6184: &&\bar \alpha_0^2\la \chi, \tetab^2\ra,\ \ 

6185:  \bar \alpha_0 \bar\beta_1 e^{i\lambda_+ t} \la \chi, \tetab^2\ra,\nn\\

6186: &&e^{i\lambda_+ t}\bar \alpha_0 \beta_1 \la\chi, |\tetab|^2\ra,\ \ 

6187:  \langle \chi, |\tetab|^2 \tetab\rangle\bar \alpha_0.

6188: \label{eq:termss}

6189: \ea

6190: Since for $t\in I_0, I_2$ we have time decay of either $\alpha_0$ or

6191: $\beta_1$ respectively, the main contribution is when $t\in I_1= [t_0,

6192: t_1]$.

6193: 

6194: For $t\in I_2$ the bound on $\tetab$ we have is

6195: $$

6196: \|\tetab(t)\|_{H^k} \leq c Q_0^{1/2}(t) \int^t_0 \frac{ds}{\langle

6197: t-s\rangle^{3/2}} Q_1^m (s).

6198: $$

6199: The second and third terms can be integrated by parts leaving terms of

6200: the type

6201: $$

6202: \cO(\beta_1^2)\ \cO(\bar\alpha_0)\ \partial^m_t\langle \chi, |\tetab|^2 + 

6203: \tetab^2  \rangle.

6204: $$

6205: We also need to integrate by parts the two other terms.  For this we

6206: need to pull out a phase factor from the leading nonlocal.

6207: 

6208: Pulling a phase as in the proof of Proposition \ref{prop:ODE}

6209:   we are left with

6210: estimating term of the type

6211: $$

6212: O(\alpha_0^2 + \alpha_0\beta_1) \langle \chi, 

6213: \tetab\partial^k_t\tetab\rangle + \langle \chi, |\tetab|^2

6214: \partial_t^k \tetab\rangle O(\alpha_0).

6215: $$  

6216: As in the estimates of Proposition \ref{cor:Corollary 3} , for $t\in I_1,$

6217: $$

6218: \|\partial^k_t\tetab\|_{H^s} \leq C

6219: Q_0^{1/2}(t)\int^t_0\frac{ds}{\langle t-s\rangle^{k'}}Q_1^m (s).

6220: $$

6221: with $k'$ large for $k$ large.

6222: 

6223: To this end we use the following.

6224: 

6225: \begin{prop}

6226: \label{prop:Lemma}

6227: 

6228: For $t\in I_1$:

6229: $$

6230: \int^t_{t_0}\frac{ds}{\langle t-s\rangle^{k'}} Q_1^{\bar m} (s) \leq

6231: \cO(\cE_0) Q^2_1 (t).

6232: $$

6233: \end{prop}

6234: 

6235: \nit{\bf Proof of Proposition \ref{prop:Lemma}:}

6236: $$

6237: \int^t_{t_0} \frac{ds}{\langle t-s\rangle^{k'}}Q_1^{\bar m} (s) =

6238: Q_1(t) \int^t_{t_0} \frac{ds}{\langle t-s\rangle^{k'}} Q_1^{\bar m -1}

6239: + \int^t_{t_0} \frac{ds}{\langle t-s\rangle^{k'}} (Q_1 (s) - Q_1(t))

6240: Q_1^{\bar m -1}.

6241: $$

6242: The second term on the right hand side can be estimated, using that

6243: $\dot Q_1=-\Gamma Q_0Q_1^2+\cO(Q_0^{1\over2}Q_1^m)$ and monotonicity of $Q_0$,

6244:  as follows:

6245:  $(s\leq \xi\leq t)$

6246: \ba

6247: &\leq& C \int^t_{t_0} \frac{ds}{\langle t-s\rangle^{k' -1}} \dot Q_1

6248: (\xi) Q_1^{\bar m-1}(s) \leq C Q_0 (t) \int^t_{t_0}

6249: \frac{ds}{\langle t-s\rangle^{k'-1}} Q_1^2 (\xi) Q_1^{\bar m-1}(s)\nn\\

6250: &+&CQ_0^{1/2}(t) \int^t_{t_0} \frac{ds}{\langle t-s\rangle^{k'-1}} Q_1^m

6251: (\xi) Q_1^{\bar m-1}(s)\nn\\

6252: &&

6253: \leq \cO(\cE_0)Q_1(t) \int^t_{t_0} \frac{ds}{\langle

6254: t-s\rangle^{k'-1}} Q_1^2 (\xi) Q_1^{\bar m-1} \nn\\

6255: &+& \cO

6256: (\cE_0)Q_1^{1/2} (t) \int^t_{t_0} \frac{ds}{\langle

6257: t-s\rangle^{k'-1}} Q_1^{m-1} (\xi) Q_1^{\bar m-1}.

6258: \nn\ea

6259: Repeating this argument, we have

6260: \ba

6261: \int^t_{t_0} \frac{ds}{\langle t-s\rangle^{k'}} Q_1^{\bar m} (s) &\leq&

6262: \cO(\cE_0) Q_0^j (t) \int^t_{t_0} \frac{ds}{\langle

6263: t-s\rangle^{k'-2j}} Q_1^{\bar m - 2j}\nn\\

6264: &\leq& \cO(\cE_0)Q_1^j(t).

6265: \nn\ea

6266: for $k' - 2j> 1$, which proves the Proposition \ref{prop:Lemma}. 

6267: 

6268: This Proposition together with the estimate 

6269: \be

6270: \| \tetab\|_{W^{k,\infty}} \leq \cO(\cE_0) Q_0^{1/2}

6271: \nn\ee

6272: implies that 

6273: \be

6274: R_1^{(0)} \leq \cO(\cE_0) 

6275:  \left[\ Q_0 Q_1^2 + {\rm higher\ order\ terms}\ \right].

6276: \nn\ee

6277: The estimates of $R_1^{(1)}$ are similar.

6278: 

6279: It remains to estimate the linear $\tetab$ terms in the $P_0,

6280: P_1$ eqs.

6281: 

6282: The leading order source term of $\tetab$ was estimated in

6283: Proposition \ref{prop:ODE}.  It remains to estimate the higher order

6284: corrections.

6285: 

6286: To this end we need to estimate terms of the following 

6287:  type appearing in the $P_0$

6288: equation, (\ref{eq:P0P1}):

6289: \ba

6290: &&\bar\alpha_0\left(|\beta_1|^2 + \alpha_0 \beta_1\right)

6291: \left\la \chi, \int^t_0 e^{-i\cH_0(t-s)} P_c \psi_0 \tetab^2 ds\ \right\ra \ 

6292:  {\rm and}\nn\\

6293: &&\left(|\beta_1|^2 \alpha_0 + \bar \alpha_0

6294: \alpha_0\beta_1\right)\left\la \chi, \int^t_0 e^{-i\cH_0(t-s)}

6295: \ P_c|\tetab|^2 \tetab ds\ \right\ra,

6296: \nn\ea

6297: and similar terms in the $P_1$ equation.

6298: 

6299: Again we focus on $t\in I_1$.

6300: Since for $0\leq s \leq t_0$ $\alpha_0$ and $\tetab$ are of order

6301: $\frac{\cO(\cE_0)}{\langle s \rangle}$,

6302: clearly these contributions are of order

6303: $$

6304: \cO(\cE_0) \left[ Q_0Q_1^2 + \langle t_0\rangle^{-3} \right],

6305: $$

6306: so it remains to estimate the $s$-integrals above on $I_1$.

6307: 

6308: But on $I_1, \|\tetab\| \leq \cO(\cE_0)Q_0^{1\over2}(t) \leq

6309: \cO(\cE_0) Q_1^{1\over2}(t)$

6310:  and since $Q_0$ is monotonic increasing on $I_1$, the above nonlocal

6311: term are bounded by 

6312: $$

6313: \cO(\cE_0)Q_0^{3\over2}(t).

6314: $$

6315: So,

6316: $$

6317: \cO(|\beta_1|^2\bar\alpha_0) \cO(\cE_0) Q_0^{3\over2}(t)\leq

6318: \cO(\cE_0) Q_1^2 Q_0

6319: $$

6320: $$

6321: O(|\alpha_0|^2\beta_1) \cO(\cE_0)Q_0^{3\over2} (t) \leq 

6322: \cO(\cE_0) Q_1^2 Q_0

6323: $$ 

6324: since $Q_0 \leq \cO(\cE_0) Q_1$ on $I_1$.

6325: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

6326: \section{Bootstrapping it all}\label{sec:bootstrap}

6327: 

6328: We assume that $t_0<\infty$, where $t_0$ is given by (\ref{eq:t0def}). 

6329: Consider the equations for $P_0$ and $P_1$, (\ref{eq:P0P1}) 

6330: displayed in Proposition \ref{prop:Master}. Explicit in (\ref{eq:P0P1})

6331: are terms which\ 

6332: 

6333: \nit (1)\ are driven by the dispersive part of the 

6334: initial data: $R_0^0[\eta_0]$ and $R_0^1[\eta_0]$

6335: 

6336: \nit (2)\ encompass interactions of the two bound

6337: states and dispersive waves: $R_1^0[\tetab]$\ and 

6338: 

6339: \nit (3) encompass interactions between bound states: $R_2^0[P_0,P_1]$. 

6340: 

6341: \nit  By Proposition \ref{prop:biR0i} the 

6342:  $R_0^j[\eta_0],\ j=0,1$ terms satisfy:

6343: \be

6344: R_0^j[\eta_0]\ =\ \cO\left(b_j(t_0,[\eta_0]_X,\cE_0\right)

6345:  \la t\ra^{-2}\ +\ \cO(\cE_0) 

6346:  2\Gamma P_0P_1^2

6347: \label{eq:R0jest}

6348: \ee

6349: Therefore, by Proposition \ref{prop:propB} and its proof

6350:  (see eqn. (\ref{eq:k0_def}) ), it is natural to introduce the functions: 

6351: \be

6352: Q_0(t)\ =\ P_0-\frac{k_0}{\la t\ra},\ \ Q_1(t)\ =\ P_1+\frac{k_1}{\la t\ra}

6353: \label{eq:Q0Q1}

6354: \ee

6355: where

6356: \ba

6357: k_0\ &=&\ b_0\ +\ \cO(\cE_0)b_1^2\ ,

6358:  \ \ \ k_1\ =\ 10b_1\label{eq:k0k1def}\\

6359: b_0\ &=&\ \la t_0\ra^{-1}\left(\ [\eta_0]_X\ +\ c_*\cE_0^2\ \right),

6360: \label{eq:b0a}\\ 

6361: \ b_1\ &=&\ \la t_0\ra^{-{1\over2}}

6362:  \left(\ [\eta_0]_X^{2\over3}\ +\ d_*\cE_0^2\ \right), \label{eq:b1a} 

6363: \ea

6364: where the constants $c_*$ and $d_*$ are to be chosen.

6365: We find, for any $m\ge4$ and all  $t\ge0$: 

6366: \ba

6367: \frac{dQ_0}{dt} &\ge& 2\Gamma Q_0 Q_1^2 + R_1^{0,\#}[\tetab] + R_2^{0,\#}[Q_0,Q_1]

6368:  \ +\ \frac{\cE_0^2c_*}{\la t_0\ra\ \la t\ra^2}

6369:  \nn\\

6370: \frac{d Q_1}{dt}&\le& -4\Gamma Q_0 Q_1^2 + R_1^{1,\#}[\tetab] + R_2^{1,\#}[Q_0,Q_1]

6371:  + \cO(\cE_0;m)\sqrt{Q_0}Q_1^m\ -\ 

6372:  \frac{\cE_0^2d_*}{\la t_0\ra^{1\over2}\ \la t\ra^2}.\nn\\  

6373: &&\label{eq:RRSS}

6374: \ea

6375: The analogous reverse inequalities hold as well with slightly different constants.

6376: 

6377: By the definition of $t_0$,  $Q_0(t_0)>0$.

6378: Furthermore, using the energy estimate on the bound state amplitudes

6379:  and (\ref{eq:Q0Q1}) of section \ref{sec:bounds}, we have

6380: \be  Q_0(t)+ Q_1(t)\leq C\cE_0,\ \ t>0.\label{eq:Q0Q1ebound}

6381: \ee

6382: 

6383: \bigskip

6384: 

6385: \bigskip

6386: 

6387: We now introduce a set of norms. The norm of

6388: $q(t)\equiv \left(\ Q_0(t),Q_1(t),\tetab(t)\ \right)$

6389: is defined as

6390: \be

6391: \|q(t)\|_\bY =  |Q_0 |_{y_0(t)} +

6392:  |Q_1|_{y_1(t)}  + \|\tetab\|_{y_2(t)}.

6393: \label{eq:bYdef}

6394: \ee

6395: The norm, $\|q(t)\|_\bY$, 

6396:  encodes all the estimates for $Q_0,Q_1$ and $\tetab$

6397:  in the intervals $I_0,I_1$ and $I_2$ through

6398: the following: 

6399: \ba

6400: |Q_0|_{y_0(t)} &\equiv& \sup_{0\le s\le t} |Q_0(s)|

6401:        \ +\  \sup_{0\le s\le \min\{t,t_0\} }\la t_0\ra\la s\ra |Q_0(s)|

6402: \label{eq:y0norm}\\

6403: |Q_1|_{y_1(t)} &\equiv& \sup_{0\le s\le t} |Q_1(s)|

6404:        \ +\  \sup_{t_1\le s\le t}|s-t_1|Q_1(s)\Gamma'Q_0(t_1)Q_1(t_1)

6405: \label{eq:y1norm}\\

6406: \|\tetab\|_{y_2(t)} &\equiv& \sup_{0\le s\le \min\{t,t_0\} }

6407:                         \la s\ra \|\tetab(s)\|_{W^{k,\infty}}

6408:   \ +\  \sup_{t_0\le s\le \min\{t,t_1\} } \|\tetab(s)\|_{W^{k,\infty}}\nn\\

6409:     &&\ \ +\ \sup_{t_1\le s\le t}\la s-t_1\ra^{1\over2}

6410:                          \|\tetab(s)\|_{W^{k,\infty}}

6411: \label{eq:y2norm}

6412: \ea

6413: {\it In these definitions we use the convention that terms for which the $s$-

6414:   range

6415:  empty are set to zero.} 

6416: \bigskip 

6417: 

6418:  By the $H^1$ {\it \'a priori} bounds

6419: \be

6420: \sup_{0\le s\le t} |Q_0(s)| \le \cE_0\left( 1+\cE_0^2\|\tetab\|_\infty\right)

6421: \nn\ee

6422: and by definition of $t_0$, (\ref{eq:t0def}),  for $t_0<\infty$,

6423: \be

6424: \sup_{0\le s\le t_0}\la t_0\ra\la s\ra |Q_0(s)|

6425:  \le {1\over2}\left(\ \cE_0+[\eta_0]_X\ \right).

6426: \nn\ee

6427: 

6428: In terms of these norms, we have bounds on $R_i^{j,\#}$. By Proposition

6429:  \ref{prop:Riteta}

6430: \be

6431: |R_1^{0,\#}[\tetab]|+|R_1^{1,\#}[\tetab]|\ \le\ 

6432:  \cO(\cE_0^\rho)\|q(t)\|_Y^{l_1}Q_0Q_1^2 +

6433:   C\ \frac{\cE_0^\rho\|q(t)\|^{l_2}_\bY}{\la t_0\ra\la t\ra^2}.

6434:  \label{eq:RRj}

6435: \ee

6436: By Proposition

6437:  \ref{prop:ODE}

6438: \ba

6439: |R_2^{0,\#}[Q_0,Q_1]|&\le&\cO(\cE_0^\rho)Q_0Q_1^2\ +\ 

6440:  C\ \frac{\cE_0^\rho+\|q(t)\|^l_\bY}{\la t_0\ra \la t\ra^2}\label{eq:RR2}\\

6441: |R_2^{1,\#}[Q_0,Q_1]|&\le&\cO(\cE_0^\rho)Q_0Q_1^2\ +\ 

6442: C\ 

6443: \frac{\cE_0^\rho+\|q(t)\|^l_\bY}{\la t_0\ra^{1\over2} \la t\ra^2}\label{eq:SS2}

6444: \ea

6445: where $l\ge2$.

6446: \bigskip

6447: 

6448: By the definition of $I_0,\ (0\le t\le t_0)$ and Propositions \ref{prop:biR0i}, 

6449:  \ref{prop:ODE} and \ref{prop:Riteta}, we have estimates

6450: (\ref{eq:RRj},\ref{eq:RR2},\ref{eq:SS2}).

6451: Therefore, for an appropriate choice of $c_*$ and $d_*$ 

6452: we have for $0\le t\le t_0$

6453: \ba

6454: \frac{dQ_0}{dt}\ &\ge&\ 2\Gamma' Q_0Q_1^2\ +\ 

6455:  \frac{c_*}{2}\frac{\cE_0^2}{\la t_0\ra\la t\ra^2}\nn\\

6456: \frac{dQ_1}{dt}\ &\le&\ -4\Gamma' Q_0Q_1^2\ +\ \cO(\cE_0;m)\sqrt{Q_0}Q_1^m\ -\ 

6457:  \frac{d_*}{2} \frac{\cE_0^2}{\la t_0\ra^{1\over2}\la t\ra^2},

6458: \label{eq:shp1}

6459: \ea

6460: where $m\ge4$.

6461: Note that by definition of $I_0$, $Q_0(t)<0$ for $t\in I_0$.

6462: 

6463: By continuity, (\ref{eq:shp1}) holds for $t_0\le t\le t_0+\delta$, for some 

6464:  $\delta>0$. It follows, using that $Q_0(t_0)>0$ and 

6465:   Propositions \label{prop:B.5} and \label{prop:C},

6466:   that (\ref{eq:PropertyQ}) ({\bf Monotonicity Property Q}) holds

6467:  on $t_0\le t\le t_0+\delta$. 

6468:   Therefore, by Propositions \ref{prop:biR0i}, \ref{prop:ODE} and 

6469:  \ref{prop:Riteta} the terms $J_0$ and $J_1$ in (\ref{eq:Q0Q1b})

6470: both satisfy the bound

6471: \be

6472: |J_k|\ \le \frac{\cO(\cE_0^{2+\rho})}{\la t_0\ra\la t\ra^2}

6473:  \ +\ \cE_0 Q_0Q_1^2,\ \ \rho>0,\ k=0,1.

6474: \label{eq:Jbounds}

6475: \ee

6476: Therefore,  for $\cE_0$ sufficiently small (\ref{eq:shp1}) holds

6477: with $c_*,d_*$ replaced by $c_*/2, d_*/2$.

6478: 

6479: Define

6480: \be

6481: T_*\ =\ \sup\{t\ge t_0:\ (\ref{eq:shp1})\ {\rm holds\ for\ some}\  c_*>0\

6482:  {\rm and}\ 

6483:   d_*>0\}

6484: \label{eq:T*def}\ee 

6485: For $t\in [0,T_*)$, $\|q(t)\|_{\bf Y}$ is small.

6486: We claim that $ T_*=\infty$. Suppose $t_0\le T_*<\infty$. Then, 

6487: for $t\in [t_0,T_*)$ we have, by Propositions  \ref{prop:B.5} and \ref{prop:C},

6488:  that the monotonicity property (\ref{eq:PropertyQ}) holds 

6489:  at $t=T_*$ and slightly

6490: beyond. Thus, the {\'a priori} bounds on $J_0,J_1$ of 

6491: Propositions \ref{prop:biR0i}, \ref{prop:ODE} and

6492:  \ref{prop:Riteta}, the $\tetab$ - bounds of Propositions \ref{prop:etabonI0},

6493:  \ref{prop:etabonI1} and \ref{prop:etabonI2}, and the smallness of

6494:    $Q_0$ and $Q_1$

6495: imply persistence of the inequalities (\ref{eq:shp1}), with perhaps a slightly

6496: smaller choice of positive constants $c_*$ and $d_*$. 

6497: This implies that $T_*=\infty$.   

6498: \bigskip

6499: 

6500: 

6501: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

6502: \section{Nongeneric behavior}\label{sec:nongeneric} 

6503: 

6504: Recall that $t_0$ is defined by (\ref{eq:t0def}).

6505: and consider the case where $t_0=\infty$ .

6506: We would like to show that 

6507: \ba

6508: P_0(t) &\to& 0 \ {\rm as }\ t \to \infty\nn\\

6509: P_1(t) \ && {\rm has \ a\ limit}.

6510: \nn\ea

6511: 

6512: The following is a consequence of the definition of $t_0$.

6513: \begin{prop}\label{prop:NG1}%{Proposition 1 (NG1)}

6514: Assume $t_0=\infty$. Then, 

6515: $P_0(t) = \cO\left([\eta_0]_X \la t\ra^{-2}\right).$

6516: Therefore, $\alpha_0\to0$ and the ground state decays.

6517: \end{prop}

6518: \bigskip

6519: 

6520: \begin{prop}\label{prop:NG2} %{Proposition 2 (NG2)}

6521:  Let  $t_0=\infty$. Then, $\beta_1$ has a limit as $t\to\infty$.

6522: \end{prop}

6523: States with this (nongeneric) behavior were constructed in \cite{TY01b}. 

6524: 

6525: \nit{\bf Proof of Proposition \ref{prop:NG2}:}

6526: The equation (\ref{eq:P1eqn}), together with the above estimate 

6527: \be P_0(t) = \cO(\ol{\eta_0})\la t\ra^{-2}

6528: \nn\ee

6529: implies

6530: \ba

6531: \frac{dP_1}{dt} &=& (-4\Gamma P_0 P_1^2 + \cO(\ol{\eta_0})\la

6532: t\ra^{-3/2})(1+\cO(P_0) + \cO(P_1))\nn\\

6533: &+& \Re (c e^{i\lambda_+ t} |\beta_1|^2\ol{\beta_1} \alpha_0) + \

6534:  {\rm h.o.t.}

6535: \ea

6536: To show that $P_1$ has a limit, we show that $\int_0^t\D_sP_2(s) ds$

6537: has a limit. 

6538: All terms other than the $\cO(\alpha_0)$ term, 

6539:  on the right hand side are absolutely integrable since

6540:  $P_0 = \cO(\la t\ra^{-2})$. 

6541: 

6542: It is left to integrate the $\cO(|\beta_1|^2 \bar\beta_1 \alpha_0)$ term.

6543: For $T$ given, let $\beta_T^2 \equiv \beta_1(T)^2$.  

6544:  Then, equation (\ref{eq:alpha0_eqn2}) reads

6545: \be 

6546: 2i\partial_t \alpha_0 = \lambda \la\psi_{0*},\Psi_1 (t)^2\ra\ol{\alpha_0}\ 

6547:   + \ {\rm integrable\ in\ t }.

6548: \nn\ee

6549:  Using 

6550: the expression for $\Psi_1(t)=\alpha_1\psi_1(\cdot,|\alpha_1|^2)$:

6551: \ba

6552: 2i\D_t\alpha_0&=& \lambda \la\psi_{0*}, \alpha_1 (t)^2 \psi^2_{1*}\ra

6553: \bar\alpha_0 + {\rm h.o.t.}\nn\\

6554: &=&\lambda \langle \psi_{0*}, \psi^2_{1*}\rangle e^{-2i\lambda_+

6555: t}\beta_1(t)^2\bar\alpha_0 (t) + O(\alpha^2_0 (T)\bar\beta_1) + {

6556: \rm h.o.t.}\nn\\

6557: &\equiv& \tilde\lambda e^{-2i\lambda_+ t} \beta_T^2 \bar\alpha_0 (t) +

6558: \tilde \lambda e^{-2i\lambda_+t}[\beta^2_1 (t) -

6559: \beta^2_1(T)]\bar\alpha_0(t) + O (\frac{1}{T^2})\nn\\

6560: &+& \ \ \ {\rm h.o.t.}

6561: \label{eq:inhom}\ea

6562: where $\tilde\lambda\equiv\lambda\langle \psi_{0*}

6563: \psi^2_{1*}\ra \ {\rm and}\    

6564: \beta_T^2 \equiv\beta^2_1 (T)$.

6565: 

6566: Solving the homogeneous part of (\ref{eq:inhom}):

6567: \be

6568: 2i\partial_t\hat\alpha_0 = \tilde\lambda e^{-2i\lambda_+t}

6569: \beta_T^2 \overline{\hat \alpha}(t)

6570: \label{eq:hom}\ee

6571: we have, using the Ansatz 

6572: $$

6573: \alpha_0(t) = A(t) e^{i(\lambda -a)t} + B(t) e^{i(\lambda + a) t}

6574: $$

6575: with

6576: $$

6577: \dot A \sim \tilde \lambda e^{-2i\lambda_+ t} [\beta^2_1 (t) -

6578: \beta^2_1(T)]\bar A e^{i\theta(t)}+ \ {\rm h.o.t.}.

6579: $$

6580: and a similar equation for $B(t)$.

6581: % XYZ explicit solution 

6582: 

6583: We have

6584: $$

6585: \frac{dP_1}{dt} =- 4\Gamma P_0P_1^2 + \Re \left(c

6586: e^{i\theta_A t }|\beta_1|^2\bar\beta_1 A 

6587:  + e^{i\theta_B t }|\beta_1|^2\bar\beta_1 B\ \right) + \ {\rm h.o.t.},

6588: $$

6589: $\theta_A, \theta_B\ne0$.

6590: Integration  of the above equation, 

6591: integration by parts (twice) of the $A$ and $B$, implies 

6592: \ba

6593: &&

6594: \langle t\rangle^{1\over2} |P_1(t) - P_1(T) | \leq C\langle

6595: t\rangle^{1\over2}\ \int^T_t |\beta_1|^2\bar\beta_1\langle

6596: t'\rangle^{-1}\langle t'\rangle [\beta^2_1 (t') -

6597: \beta^2_1(T)]^2\bar A(t') dt' + \ {\rm h.o.t. } \langle t\rangle^{1\over2}

6598: \nn\\

6599: &\leq& C\cE_0^m\left(\int^T_t \langle t'\rangle^{-{1\over2} - 1}

6600: dt'\right) \left( \sup_{0\leq t' \leq T} \langle t^{1\over2}\rangle

6601: |P_1(t') - P_1 (T) | \right)^2 + \sup_t\langle

6602: t\rangle^{1\over2}\ {\rm h.o.t. }\nn\\

6603: &&\Rightarrow | P_1(t) - P_1(T) |  \leq C\langle t\rangle^{-{1\over2}}

6604: \nn\ea

6605: which implies integrability of $\dot A(t)$ and limit of $P_1(t)$.

6606: 

6607: %\appendix

6608: \input mbs-appA % notation

6609: \input mbs-appB % Proof of Proposition Hspectrum

6610: \input mbs-appC % Commutator term in modulation eqns 

6611: \input mbs-bib

6612: %\bibliographystyle{amsplain}

6613: %\bibliography{mbs-bib}

6614: \end{document}

6615: 

6616: